Analyzing N-terminal Arginylation through the Use of Peptide Arrays and Degradation Assays*

Nα-terminal arginylation (Nt-arginylation) of proteins is mediated by the Ate1 arginyltransferase (R-transferase), a component of the Arg/N-end rule pathway. This proteolytic system recognizes proteins containing N-terminal degradation signals called N-degrons, polyubiquitylates these proteins, and thereby causes their degradation by the proteasome. The definitively identified (“canonical”) residues that are Nt-arginylated by R-transferase are N-terminal Asp, Glu, and (oxidized) Cys. Over the last decade, several publications have suggested (i) that Ate1 can also arginylate non-canonical N-terminal residues; (ii) that Ate1 is capable of arginylating not only α-amino groups of N-terminal residues but also γ-carboxyl groups of internal (non-N-terminal) Asp and Glu; and (iii) that some isoforms of Ate1 are specific for substrates bearing N-terminal Cys residues. In the present study, we employed arrays of immobilized 11-residue peptides and pulse-chase assays to examine the substrate specificity of mouse R-transferase. We show that amino acid sequences immediately downstream of a substrate's canonical (Nt-arginylatable) N-terminal residue, particularly a residue at position 2, can affect the rate of Nt-arginylation by R-transferase and thereby the rate of degradation of a substrate protein. We also show that the four major isoforms of mouse R-transferase have similar Nt-arginylation specificities in vitro, contrary to the claim about the specificity of some Ate1 isoforms for N-terminal Cys. In addition, we found no evidence for a significant activity of the Ate1 R-transferase toward previously invoked non-canonical N-terminal or internal amino acid residues. Together, our results raise technical concerns about earlier studies that invoked non-canonical arginylation specificities of Ate1.

In mammals such as mice, there are at least six isoforms of R-transferase, produced through alternative splicing of the Ate1 pre-mRNA ( Fig. 1, C and D) (9,78,79). At present, about 20 mammalian proteins, including natural protein fragments, have been identified as either constitutively or conditionally short-lived physiological substrates of the Ate1 R-transferase and the post-arginylation part of the Arg/N-end rule pathway. All of these substrates have in common the presence of a canonical (Nt-arginylatable) N-terminal residue, either Asp, Glu, or Cys ( Fig. 1A) (4, 29, 42-44, 49, 50, 73, 84, 85). In addition, there are hundreds of mammalian proteins, including natural protein fragments, that are predicted to be Nt-arginylated, based on the presence, in a protein or its natural fragment, of N-terminal Asp, Glu, or Cys (7,43,44,49,73,86).
Several publications over the last decade, largely by the laboratory of A. Kashina, have reported that Ate1 R-transferase can also arginylate, to a significant extent, a variety of "noncanonical" (non-Asp, non-Glu, non-Cys) N-terminal residues in cellular proteins (87)(88)(89)(90)(91)(92). The same group described evidence that some isoforms of R-transferase were specific for substrates bearing N-terminal Cys (93). Other recent publications by the same laboratory suggested, on the basis of MS anal-yses, that the Ate1 R-transferase is capable of arginylating not only ␣-amino groups of specific N-terminal residues but also ␥-carboxyl groups of internal (non-N-terminal) Asp or Glu (94,95). This suggestion followed a report, by another group, that described a modified 13-residue neurotensin hormone that FIGURE 1. The mammalian N-end rule pathway and isoforms of the mouse Ate1 R-transferase. See the Introduction for references and descriptions of the pathway's mechanistic aspects and functions. Amino acid residues are denoted by single-letter abbreviations. A yellow oval denotes the rest of a protein substrate. A, the mammalian Arg/N-end rule pathway. It targets proteins for degradation through their specific unacetylated N-terminal residues. Primary, secondary, and tertiary refer to mechanistically distinct classes of destabilizing N-terminal residues. Ntan1 and Ntaq1 are N-terminal amidases (Nt-amidases) that convert the tertiary destabilizing N-terminal residues Asn and Gln to Asp and Glu, respectively. The Ate1 R-transferase (arginyltransferase or Arg-tRNA-protein transferase) conjugates Arg, a primary destabilizing residue, to N-terminal Asp, Glu, and (oxidized) Cys (the "canonical" N-terminal residues). Type 1 and type 2 refer, respectively, to two sets of primary destabilizing N-terminal residues, basic (Arg, Lys, and His) and bulky hydrophobic (Leu, Phe, Trp, Tyr, Ile, and also Met, if the latter is followed by a bulky hydrophobic residue (⌽)). These sets of N-terminal residues are recognized by distinct substrate-binding sites of N-recognins (the pathway's E3 ubiquitin ligases) Ubr1, Ubr2, Ubr4, and Ubr5. B, the mammalian Ac/N-end rule pathway. It targets proteins through their N ␣ -terminally acetylated (Ntacetylated) residues. The red arrow on the left indicates the cotranslational removal of the N-terminal Met residue by Met-aminopeptidases (MetAPs). N-terminal Met is retained if a residue at position 2 is larger than Val. See the Introduction for references and additional aspects of the Ac/N-end rule pathway. C, the exons, including alternative exons (1A/1B and 7A/7B) of the mouse Ate1 gene, which encodes isoforms of R-transferase. The deduced lengths (in amino acid residues) of the encoded Ate1 exons are indicated at the top. D, mouse R-transferase isoforms that are produced through alternative splicing of Ate1 pre-mRNA. The terminology of these isoforms (Ate1 1A7A , Ate1 1A7B , Ate1 1B7A , Ate1 1B7B , Ate1 1A7AB , and Ate1 1B7AB ) (9, 79) is based on the presence or absence of the alternative exons 1A/1B and 7A/7B. Ate1 1A7AB and Ate1 1B7AB are minor isoforms that result from the retention of both variants of exon 7 (79). appeared to be arginylated at its internal Glu-4 residue (96). Thus far, studies by other laboratories have failed to find evidence for either a significant arginylation of non-canonical N-terminal amino acid residues or a significant arginylation of internal residues (85,(97)(98)(99).
In the present work, we used purified isoforms of the mouse Ate1 R-transferase, arrays of C-terminally immobilized synthetic peptides, and pulse-chase degradation assays to address the substrate specificity of arginylation by mouse Ate1. We show here that a sequence immediately downstream of a substrate's canonical (Nt-arginylatable) N-terminal residue, and particularly a residue at position 2, can significantly affect the rate of Nt-arginylation by R-transferase and thereby the rate of degradation of a substrate protein. We also show that the four major isoforms of the mouse R-transferase have similar substrate specificities in vitro, contrary to the earlier claim that some isoforms of Ate1 are specific for substrates bearing N-terminal Cys (93). Furthermore, and also contrary to the other cited reports (87-89, 94, 100, 101), we found no evidence for a significant activity of the Ate1 R-transferase toward the previously suggested non-canonical residues.
A caveat to otherwise direct contradictions between our findings and those of the earlier studies is that our results were produced in vitro (with purified Ate1 R-transferase and peptide arrays), whereas most reports about non-canonical arginylation of cellular proteins have been about analyses of natural, in vivo-produced proteins (87-92, 94 -96).

Results and Discussion
Use of CelluSpots TM Peptide Arrays to Analyze the Substrate Specificity of R-transferase-CelluSpots arrays are produced by the synthesis of C-terminally immobilized peptides on a modified cellulose support that is dissolvable in a trifluoroacetic acid solution, thereby allowing the synthesized peptides to be spotted onto coated microscope slides. This technology makes it possible to produce a number of virtually identical copies of a peptide array from a single set of peptide-synthesizing reactions (102). CelluSpots arrays have been successfully used to examine the specificity of histone-modifying enzymes for amino acid sequences in histones; to analyze, in vitro, the binding specificity of antibodies; to characterize specific binding of SH2 (Src homology 2 domain) protein regions to phosphotyrosine-containing peptides; and to survey the specificities of phosphokinases and kinase inhibitors (103)(104)(105).
To analyze the substrate specificity of the mouse Ate1 R-transferase, we performed 14 C-arginylation assays using identical CelluSpots peptide arrays, each of which comprised 96 immobilized peptides that were largely 11 residues long (two peptides were 12 residues long) ( Fig. 2 and Table 1). To the best of our knowledge, CelluSpots have not been previously used to study arginylation. The unlimited sequence diversity of synthetic peptides, the availability of virtually identical copies of a given peptide array, and technically straightforward, reproducible conditions of arginylation assays make CelluSpots a particularly helpful method for analyzing the in vitro specificity of Ate1 R-transferase isoforms (3, 78 -81).
Amino acid sequences of 11-residue peptides in CelluSpots arrays were designed to examine several specificity aspects of the mouse R-transferase, including effects of a second-position residue of a substrate; effects of charge clusters immediately downstream of the (arginylatable) N-terminal residue; effects of the Pro residue (an imino acid) close to the N-terminal residue; and the influence of oxidation of N-terminal Cys on its efficacy as an acceptor of Arg in Ate1-catalyzed reactions. We also wished to address, using the conceptually independent and technically straightforward technology of peptide arrays, sev-FIGURE 2. Design and validation of CelluSpots arginylation assays. A, layout of two virtually identical copies of a CelluSpots 96-spot array of C-terminally immobilized 11-residue synthetic peptides that are produced and spotted onto coated microscope slides (102). Each peptide spot is denoted by a letter (A-D) and a number . For the complete sequences of these 96 peptides, see Table 1. For the five N-terminal residues of these 11-residue peptides, see the bottom of B. Each individual CelluSpots array (left panel versus right panel) was "bounded" through the use of a hydrophobic pen, making it possible for two arrays on a slide to be used for two separate enzymatic reactions. B, rectangle on the left, an arginylation assay was carried out with a CelluSpots peptide array containing a complete reaction mix (including [ 14 C]Arg; see "Experimental Procedures") save for the absence of Ate1 R-transferase. Note the negligible (essentially undetectable) levels of [ 14 C]Arg incorporation under these conditions. B, rectangle on the right, the otherwise identical assay was carried out in the presence of purified mouse Ate1 1B7A R-transferase. Note that "negative" peptide spots, containing 11-residue peptides bearing non-canonical (non-Asp, non-Glu, and non-Cys) N-terminal residues, exhibited negligible (undetectable) levels of [ 14 C]Arg incorporation at this level of autoradiographic exposure. Significantly longer autoradiographic exposures made background levels of [ 14 C]Arg incorporation detectable in the presence of Ate1 but still negligible (undetectable) in the absence of Ate1 (left). See "Results" for possible causes of very low but detectable background level [ 14 C]Arg incorporation in spots of peptides bearing noncanonical N-terminal residues. The bottom of B recapitulates the arrangement of 96 peptide spots in a rectangle of A and, in addition, cites the first five residues of each 11-residue peptide. Spot numbers are in green. The canonical (Nt-arginylatable) N-terminal residues Asp, Glu, and Cys (the latter a precursor of Nt-arginylatable oxidized Cys) are in red. Table 1 cites complete sequences of 11-residue peptides. (Two peptides, at C1 and C5, were 12 residues long.)  (87,94,96). As a part of initial quality controls, and also to verify that peptides on CelluSpots arrays could actually serve as efficacious substrates of purified mouse Ate1 R-transferase isoforms, the arginylation reaction was optimized on slides containing duplicate copies of a peptide array, in either the presence or the absence of a purified mouse Ate1 R-transferase isoform ( Fig. 2 and see "Experimental Procedures"). The expected "positive" peptides (bearing canonical N-terminal residues; see the Introduction) and the putative "negative" peptides (bearing non-canonical N-terminal residues) were interspersed with other peptides throughout a CelluSpots array (Fig. 2). Some peptides were included in duplicate, making it easier to assess the extent of variability of these arginylation assays within each assay. Spots A1-A24 and B1-B24 ( Fig. 2) were of 11-residue peptides whose sequences were variable over the first five residues and constant afterward, specifically Xaa1-Xaa2-Xaa3-Xaa4-Xaa5-Xaa6-Gly-Pro-Ala-Ser-Gly. The sequences of 11-residue peptides in other spots (C1-C24 and D1-D24) contained, in general, varying residues both before and after position 6, as described in Table 1. Two peptides, at C1 and C5, were 12 residues long ( Fig. 2 and Table 1).
Under the conditions used (see "Experimental Procedures"), the otherwise complete reaction mix that included L-[ 14 C]arginine ([ 14 C]Arg) but lacked the added Ate1 R-transferase resulted in a uniformly negligible level of [ 14 C]Arg incorporation (Fig. 2B, left panel versus right panel). In contrast, the same assay, with a virtually identical copy of the CelluSpots array but carried out in the presence of a purified Ate1 R-transferase isoform, such as Ate1 1B7A (Fig. 1C), yielded high levels of [ 14 C]Arg conjugation in spots whose peptides contained a canonical Nt-arginylatable residue, such as N-terminal Asp or Glu (e.g. see Fig. 2B (right), spots A7, A8, A11, A12, B6 -B11, C15, C23, D12, D13, D21, and D22). These and related results (Figs. 2-5) illustrate high signal-to-noise levels and a broad dynamic range of CelluSpots arginylation assays.
The autoradiogram in Fig. 2B (right) exhibited negligible (undetectable) levels of [ 14 C]Arg incorporation in "negative" spots (i.e. those containing 11-residue peptides with non-canonical (non-Asp, non-Glu, non-Cys) N-terminal residues). Significantly longer autoradiographic exposures of the same or independently produced CelluSpots 14 C-arginylation assays allowed the detection of (low) [ 14 C]Arg incorporation in the "negative" peptide spots of these arrays. Examples of background level but detectable [ 14 C]Arg incorporation that can be seen upon longer autoradiographic exposures are in Figs   that could be detected in "negative" peptide spots of CelluSpots arginylation assays upon relatively long autoradiographic exposures (e.g. Figs. 3A, 4B, and 5A; compare with Fig. 2B (right)) were likely to signify a low but actual Ate1-mediated conjugation of [ 14 C]Arg.
If so, what might be the reason for a uniformly low but still detectable (upon longer autoradiographic exposures) incorporation of [ 14 C]Arg in spots of peptides bearing non-canonical N-terminal residues? Two mechanistically different and nonalternative (mutually compatible) explanations of background level [ 14 C]Arg incorporation are as follows. In the first interpretation, R-transferase might be able to conjugate Arg, at a very low, presumably physiologically irrelevant efficacy (still detectable in 14 C-arginylation assays) to at least some non-canonical N-terminal residues. A different and non-alternative interpretation is that background level amounts of [ 14 C]Arg in spots of peptides bearing non-canonical N-terminal residues reflect the presence, in the corresponding peptide spots, of non-zero (but very low) admixtures of canonical Nt-arginylatable N-terminal residues, such as Asp and/or Glu. It would be interesting to rigorously distinguish between these interpretations by carrying out analogous arginylation assays with individual (i.e. nonarrayed) short peptides that would be synthesized from particularly pure amino acid stocks and would be purified and authenticated more extensively and rigorously than is possible with arrays of immobilized peptides.
Different Ate1 Isoforms Exhibit Similar Substrate Preferences-A subset of 11-residue peptides in our CelluSpots arrays was designed to examine the effects of a second-position residue on the relative efficacy of Ate1-mediated arginylation. Peptides of that subset contained N-terminal Asp. This is a canonical (Nt-arginylatable) residue and apparently the most efficacious acceptor of Arg, among N-terminal Asp, Glu, or oxidized Cys (Fig. 1A) (9). N-terminal Asp was followed by a varying second-position amino acid residue (Xaa), by the Gly-Ser-Gly sequence at positions 3-5, and by the sequence (chosen quasi-randomly) Phe-Gly-Pro-Ala-Ser-Gly at positions 6 -11 (see Fig. 3 and its legend). In addition, the array's "control" peptide spots contained 11-residue peptides of otherwise identical sequences but bearing N-terminal Arg, a non-canonical residue, instead of N-terminal Asp (Fig. 3A). The invariant ]Arg autoradiograms of CelluSpots arginylation assays with the four purified Ate1 R-transferase isoforms, Ate1 1A7A , Ate1 1A7B , Ate1 1B7A , and Ate1 1B7B , and with the 11-residue peptides XZGSGFGPASG (where X represents Asp, Cys, or Glu and Z is Ser, Asp, Lys, or Pro). B, quantification of results in A, using [ 14 C]Arg autoradiography and normalizing the level of Nt-arginylation to Ntarginylation of the Asp-Ser-bearing 11-residue peptide. These and other quantified arginylation assays (Figs. 3 and 4) were carried out, independently, at least two times and yielded results within 10% of the data shown in A. C, [ 14 C]Arg autoradiograms of CelluSpots arginylation assays with the four purified Ate1 R-transferase isoforms, Ate1 1A7A , Ate1 1A7B , Ate1 1B7A , and Ate1 1B7B , and with the 11-residue peptides XKGLAGLPASC (where X represents Arg-Cys, Asp, Cys, or Cys-sulfonate). The sequence CKGLAGLPASC is the N-terminal sequence of mouse Rgs4 (see "Results").
sequence Gly-Ser-Gly-Phe-Gly-Pro-Ala-Ser-Gly at positions 3-11 is not significantly similar to sequences in natural mouse proteins. Arginylation assays with these peptide arrays indicated that all of the above peptides, which contained the canonical (Nt-arginylatable) N-terminal Asp followed by a varying position 2 residue (Xaa) and the sequence Gly-Ser-Gly-Phe-Gly-Pro-Ala-Ser-Gly (positions 3-11), could be arginylated by mouse R-transferase to different but significant extents, whereas none of the otherwise identical peptides bearing the non-canonical N-terminal Arg residue were arginylated above background levels (Fig. 3).
Ate1 1B7A , Ate1 1B7B , Ate1 1A7A , and Ate1 1A7B , the four major splicing-derived isoforms of the mouse Ate1 R-transferase, differ by the presence or absence of exon 1a versus the sequelogous (similar in sequence (106)) exon 1b and also exon 7a versus the sequelogous exon 7b (Fig. 1D) (3,78,79). Carrying out four identical arginylation assays with four Ate1 isoforms, we found that the relative activities of the purified Ate1 1B7A , Ate1 1B7B , Ate1 1A7A , and Ate1 1A7B R-transferases toward the 11-residue peptides with N-terminal Asp and a varying residue at position 2 were substantially similar (Fig. 3). Although one isoform, Ate1 1A7B , exhibited a significantly lower overall arginylation activity (per g of added purified R-transferase) than the other three isoforms, the substrate preferences of Ate1 1A7B , vis-à-vis the examined set of peptides, were substantially similar to those of the other three isoforms ( Fig. 3; see the legends to Figs. 3-5 for statistical aspects of [ 14 C]Arg quantification in these arrays).
These arginylation assays also showed that all four Ate1 isoforms exhibited a significant preference for arginylating N-terminal Asp-bearing peptides that contained either a hydrophobic residue other than Trp (specifically Ala, Val, Leu, Ile, Phe, or Tyr) or a basic residue (His, Lys, or Arg) at position 2, in comparison with either acidic residues (Asp or Glu) or the Pro residue at this position (Fig. 3). Pro and Trp at position 2 were the least favorable residues at position 2, in the otherwise identical sequence contexts of N-terminal Asp-bearing Asp-Xaa-Gly-Ser-Gly-Phe-Gly-Pro-Ala-Ser-Gly peptides (Fig. 3). It should be noted that the isoform Ate1 1B7B was more sensitive to the presence of Asp or Glu at position 2 than the other three Ate1 isoforms, a rare instance of a specificity difference among Ate1 isoforms (Figs. 3 and 4).
The Overall Charge of a Region Adjoining N-terminal Residue Affects the Efficacy of Arginylation-All four Ate1 R-transferase isoforms exhibited significantly decreased efficacies of arginylation with 11-residue peptides in which N-terminal Asp was followed by four basic (Arg) residues, in comparison with a single Arg at position 2 (Fig. 4). Three of the four Ate1 isoforms were also significantly less efficacious when N-terminal Asp was followed by four acidic (Asp) residues, in comparison with a single Asp at position 2 (Fig. 4). An interesting exception was the Ate1 1A7A isoform, which arginylated the N-terminal sequence DDGSG comparably with the sequence DDDDD (Fig. 4).
In contrast, a difference in the efficacy of arginylation, for all four Ate1 isoforms, was much smaller between the N-terminal sequences DAGSG and DAAAA, and the run of four Ala residues after N-terminal Asp was even more favorable for arginylation than a single Ala at position 2 (Fig. 5). Notably, even a single position 2 Asp residue made the resulting peptide a significantly less efficacious arginylation substrate than the otherwise identical peptide in which N-terminal Asp was followed by either a single positive (Arg) residue or a single small uncharged (Ala) residue ( Fig. 4; see the legends to Figs. 3-5 for statistical aspects of [ 14 C]Arg quantification in these arrays).
Negligible Efficacy of Unoxidized N-terminal Cys as a Substrate for Arginylation-Earlier studies of conditional N-degrons bearing N-terminal Cys have shown that the Ate1-mediated Nt-arginylation of Cys required its preliminary oxidation to Cys-sulfinate or Cys-sulfonate (3,4,29). (Although it is a priori likely that R-transferase can also Nt-arginylate the (less oxidized) N-terminal Cys-sulfinate, because its side chain "resembles" the side chain of Asp more than Cys-sulfonate does, there is no rigorous evidence, so far, that bears on this conjecture, inasmuch as Cys-sulfinate is readily oxidized in vitro to Cys-sulfonate.) In agreement with those findings, our results with CelluSpots peptide arrays indicated that the unmodified N-terminal Cys residue is either a very poor arginylation substrate of the Ate1 R-transferase or not a substrate at all (Fig. 2B, right, (Fig. 5, A  (columns 5-8), B, and C) resulted from a partial nonenzymatic oxidation of Cys in these peptides during preparation and handling of CelluSpots arrays.
A small increase over background levels of [ 14 C]Arg incorporation could be observed with 11-residue peptides bearing unmodified N-terminal Cys when the second residue was basic (Lys) (Fig. 5, A (columns 5-8) and B). It remains to be determined whether this small effect resulted from a weak increase in the otherwise negligible activity of R-transferase toward unoxidized N-terminal Cys when the second residue was basic (Lys) or whether, non-alternatively, this effect stemmed from a (possibly) increased propensity of N-terminal Cys to be slowly and nonenzymatically oxidized to Cys-sulfinate and/or Cyssulfonate by oxygen in the air if a residue at position 2 was basic.
Oxidation of N-terminal Cys Allows Its Arginylation-To compare Ate1 R-transferase isoforms by their ability to arginylate the unmodified versus oxidized N-terminal Cys residue, we used an 11-residue peptide whose sequence, CKGLAGLPASC, was identical to the N-terminal sequence of wild-type mouse Rgs4. We also used the otherwise identical peptides that differed at position 1 (Fig. 5C). Rgs4, a regulator of specific G proteins, has been shown to be a conditionally Nt-arginylated (and therefore conditionally short-lived) physiological substrate of the Arg/N-end rule pathway (3,4,29,107). N-terminal positions of examined 11-residue peptides contained either an unmodified (i.e. not deliberately oxidized) N-terminal Cys (the same as in wild-type Rgs4) or the oxidized N-terminal Cys (Cyssulfonate), which was substituted for unmodified Cys during the synthesis of these peptides. Two other (control) peptides contained, instead of N-terminal Cys, either N-terminal Asp (an unconditionally Nt-arginylatable (canonical) N-terminal residue; see the Introduction) or the 2-residue sequence Arg-Cys, bearing N-terminal Arg, a non-canonical residue. The lat-ter peptide was a 12-residue one, in contrast to 11-residue peptides in the other spots, due to the addition of N-terminal Arg (Fig. 5C).
As expected (given the previous evidence about oxidation of N-terminal Cys being a prerequisite for its Nt-arginylation (3,4,29)), we found that the peptide bearing N-terminal Cys-sulfonate (incorporated during peptide synthesis) was a significant arginylation substrate. In contrast, the arginylation, with four examined Ate1 isoforms, of the otherwise identical peptide bearing the unoxidized (more accurately, not deliberately oxidized) N-terminal Cys residue was at background levels ( Fig.  5C, column 4 versus column 3). Also, as expected, the otherwise identical peptide bearing the canonical N-terminal Asp residue (instead of Cys) was arginylated much more efficaciously than even the oxidized N-terminal Cys residue, whereas the otherwise identical 12-residue peptide bearing the 2-residue N-terminal sequence Arg-Cys (N-terminal Arg is a non-canonical residue) was not arginylated above background levels ( Fig. 5C; see also Fig. 2B (right), spots C1-C4).
We found that all four examined isoforms of the mouse Ate1 R-transferase exhibited similar relative efficacies of arginylation of peptides in this N-terminal Cys-based setting (Fig. 5). These and other results of this section disagreed with data of an earlier report indicating that some isoforms of Ate1 are nearly specific for N-terminal Cys (93). In particular, none of the four examined Ate1 isoforms in CelluSpots assays could arginylate the unoxidized N-terminal Cys residue above background levels of [ 14 C]Arg incorporation (Fig. 5). In addition, these Ate1 isoforms were similar in their relative ability to arginylate the oxidized N-terminal Cys (Cys-sulfonate) residue (Fig. 5C).
The Rate of Ate1-dependent Protein Degradation Is Influenced by the Overall Charge of a Region Adjoining the N Terminus-We asked whether the observed influence of second-position residues and adjoining clusters of charged residues on the efficacy of Nt-arginylation in CelluSpots assays (Fig. 4) was also relevant to the rate of Ate1-dependent protein degradation. Pulse-chase assays were carried out in rabbit reticulocyte extract, a previously characterized transcriptiontranslation-degradation system that contains the Arg/N-end rule pathway (44,49,73). The Arg/N-end rule reporter substrates were derivatives of Rgs4, a regulator of G proteins that has been identified as an Ate1-dependent physiological substrate of the Arg/N-end rule pathway (4,29,107). Wild-type mammalian Rgs4 proteins bear N-terminal Cys, after the cotranslational removal of their N-terminal Met by Metaminopeptidases. The N-terminal Cys residue of Rgs4 has been shown to be oxidized, by NO and oxygen, either in vivo or in reticulocyte extract to (ultimately) Cys-sulfonate, which can be Nt-arginylated by the Ate1 R-transferase. The resulting Nt-arginylated Rgs4 is processively destroyed to short peptides by the post-arginylation part of the Arg/N-end rule pathway (Fig. 1A) (3, 4).
Degradation assays in reticulocyte extract employed the Ub reference technique (URT), derived from the Ub fusion technique (Fig. 6A) (1, 108). In this method, the cotranslational cleavage of a URT-based protein fusion by deubiquitylases in reticulocyte extract produces, at the initially equimolar ratio, a test protein with a desired N-terminal residue and a "reference" protein, in the present case f DHFR-Ub K48R , a flag-tagged derivative of the mouse dihydrofolate reductase (Fig. 6A). In URTbased assays, a pulse-labeled test protein is quantified by measuring its level relative to that of a stable reference protein, f DHFR-Ub K48R , at each time point of a chase. The URT-based, "built-in" reference protein increases the accuracy of degradation assays (Fig. 6A) (108 -110).
To simplify the mechanics of initial targeting (i.e. to bypass the N-terminal Cys oxidation step) (it is required for degradation of wild-type Rgs4 (3, 4, 29)), the natural N-terminal Cys of Rgs4 was replaced by Asp in the context of URT fusion (Fig.  6A). In addition, to place an uncharged residue at position 2 and FIGURE 6. Pulse-chase degradation assays using Rgs4-based Arg/N-end rule substrates and the ubiquitin reference technique. A, URT. For its description and references, see "Results." B, 35 S-pulse-chase analyses, in rabbit reticulocyte extract, of XZGSG-Rgs4 (where X represents Arg, Asp, or Val and Z is Ala), a set of mouse Rgs4-based test proteins. Rgs4 is a regulator of G proteins and a conditionally short-lived physiological substrate of the Arg/Nend rule pathway (see "Results" for references and additional information). In these test proteins, the sequence Gly-Ser-Gly (GSG) was inserted at positions 3-5 to make possible its replacements, in test proteins of C, with the sequences RRR or DDD (see below). C-terminally flag-tagged XZGSG-Rgs4 f proteins were coexpressed, in reticulocyte extract, with the N-terminally flagtagged f DHFR-Ub reference protein, initially as a single Ub fusion (see A), and were labeled for 10 min at 30°C with [ 35 S]Met/Cys, followed by a chase for 45 min, immunoprecipitation with anti-flag antibody, SDS-PAGE of immunoprecipitated proteins, and autoradiography. The bands of 35  to allow changes of residues 2-5 without altering the length of the reporter, the (charged) Lys residue at position 2 of wild-type mouse Rgs4 was replaced by Ala, and the Gly-Ser-Gly sequence was inserted at positions 3-5. As described below, the set of constructed and examined Rgs4 derivatives (Figs. 6 and 7) involved alterations of residues at positions 1-5. The C-terminally flag-tagged DAGSG-Rgs4 f (specifically f DHFR-Ub K48R -Asp-Ala-Rgs4 f , before the cotranslational cleavage, by deubiquitylases, at the Ub K48R -Asp junction; Fig. 6A) (108 -110)) was Nt-arginylated by Ate1 in reticulocyte extract (Fig. 1A). The sequence of the DAGSG-Rgs4 f test protein was identical to that of wild-type Rgs4, save for the replacement of wild-type Cys-Lys by Asp-Ala and the insertion of Gly-Ser-Gly.
In reticulocyte extract, DAGSG-Rgs4 f was rapidly degraded (t1 ⁄ 2 Ͻ 10 min; Fig. 6, B (lanes 5-8) and D). The otherwise identical RAGSG-Rgs4 f , with the N-terminal sequence Arg-Ala, was degraded even faster (t1 ⁄ 2 Ͻ 1 min; Fig. 6, B (lanes 1-4) and D), most likely because in this case the preliminary step of Ntarginylation was neither possible nor necessary prior to the recognition of RAGSG-Rgs4 f by a mammalian N-recognin such as Ubr1 (Fig. 1A). In contrast, the otherwise identical VAGSG-Rgs4 f (bearing the N-terminal sequence Val-Ala, which is not recognized by the Arg/N-end rule pathway) was stable during the chase (t1 ⁄ 2 Ͼ Ͼ 45 min; Fig. 6, B (lanes 9 -12) and D). In addition, the initial, time 0 (pre-chase) levels of DAGSG-Rgs4 f and (particularly) RAGSG-Rgs4 f were much lower than the relative level of VAGSG-Rgs4 f , indicating a substantial early (prechase) degradation of DAGSG-Rgs4 f and RAGSG-Rgs4 f but not of VAGSG-Rgs4 f (Fig. 6, B and D). In sum, this set of oth-erwise identical Rgs4-based test proteins, which differ solely at their N terminus-proximal positions, is a particularly suitable setting for examining effects of N-terminal changes on the rate of proteolysis by the Arg/N-end rule pathway (Fig. 6, B and D; see the legends to Figs. 6 and 7 for statistical aspects of quantifying protein degradation in these assays).
We also compared degradation rate effects of a second-position basic residue (Lys) versus either an acidic residue (Asp) or an imino residue (Pro) in DXGSG-Rgs4 f test proteins (where X represents Asp, Pro, Lys). In agreement with a significantly less efficient arginylation of N-terminal Asp-bearing CelluSpots peptides containing either an acidic (Asp) or Pro residue at position 2 (in comparison with a basic residue at that position), the degradation of DKGSG-Rgs4 f , with Lys at position 2, was much faster than the degradation of either DDGSG-Rgs4 f or DPGSG-Rgs4 f (Fig. 7; see the legends to Figs. 6 and 7 for statistical aspects of quantifying protein degradation in these assays). In sum, the identity of a position 2 residue, in an N-degron that requires Nt-arginylation, can influence the degradation rate of a corresponding protein, at least in part by altering the rate of Nt-arginylation (Fig. 1A).
We also constructed and examined derivatives of DAGSG-Rgs4 f that contained either one or four Asp residues at positions 2-5 or, alternatively, either one or four Arg residues at the same positions. (These changes of Rgs4 were done through substitutions of its wild-type residue 2 and of the previously inserted Gly-Ser-Gly at positions 3-5 (see above) by either four Asp or four Arg residues.) Interestingly, DDDDD-Rgs4 f , in which a stretch of four acidic (Asp) residues followed N-terminal Asp, was found to be long-lived in reticulocyte extract (t1 ⁄ 2 Ͼ Ͼ 45 min; Fig. 6, C (lanes 13-15) and E), despite the fact that N-terminal Asp, when it is present in a less disfavorable downstream sequence context (e.g. in DAGSG-Rgs4 f ), is a short-lived substrate of the Arg/N-end rule pathway (Fig. 6, B (lanes 1-4) and D; see the legends to Figs. 6 and 7 for statistical aspects of quantifying protein degradation in these assays). Predictably, the N-terminal sequence, Asp-Ala, was a good substrate for Ntarginylation in CelluSpots peptide arrays (e.g. see Fig. 3A). The absence of significant degradation of DDDDD-Rgs4 f was in qualitative agreement with the relative extents of Nt-arginylation of N-terminal Asp in the context of CelluSpots peptide arrays, in which the N-terminal DDDDD . . . sequence was arginylated much less efficaciously than the N-terminal DDGSG . . . sequence (Fig. 4).
When four basic (Arg) residues followed N-terminal Asp, the rate of degradation of the resulting DRRRR-Rgs4 f was nearly identical to that of DR-Rgs4 f , which contained a single Arg residue after N-terminal Asp (followed by the wild-type GLA sequence of Rgs4 at positions 3-5) (Fig. 6, C (lanes 5-8 versus  lanes 1-4) and E). These degradation assay results were in qualitative agreement with the extents of Nt-arginylation of N-terminal Asp in the context of CelluSpots assays, in which both the DRGSG . . . and DRRRR . . . peptides were arginylated, with the DRGSG . . . peptide being a more efficacious Ate1 substrate than the DRRRR . . . peptide (Fig. 4).
Together, the results of arginylation assays on peptide arrays and protein degradation assays in reticulocyte extract (Figs. 2-7) indicated that multiple acidic residues immediately down-  A and B), the relative efficacy of Nt-arginylation of N-terminal Asp is a function of the identity of a residue at position 2 (also see "Results"). B, quantification of data in A. These and other quantified pulse-chases (Fig. 6) were carried out, independently, at least three times, and yielded results within 10% of the data values shown in A.
stream of a canonical (arginylatable) N-terminal residue such as Asp could significantly attenuate its Ate1-mediated Nt-arginylation and, consequently, the rate of degradation of a corresponding N-terminal Asp-bearing test protein by the Arg/Nend rule pathway. Although multiple basic residues (Arg) downstream of N-terminal Asp also decreased the efficacy of Nt-arginylation, this decrease was less severe than the one of several acidic (Asp) residues and did not strongly influence the rate of degradation of a corresponding N-end rule substrate (Figs. 4 -6).

Use of CelluSpots Assays to Address Reports about Arginylation of Non-canonical N-terminal Residues and Specific
Internal Residues of Cellular Proteins-As mentioned in the Abstract and Introduction, several publications, over the last decade, have suggested, largely on the basis of MS-based analyses of isolated cellular proteins (digested by trypsin or other proteases), that the arginylation of specific proteins or their natural fragments can involve, to a significant extent, non-canonical (non-Asp, non-Glu, non-Cys) N-terminal residues, including N-terminal Met, Leu, Phe, Val, Ala, Thr, Gly, Asn, Lys, and Pro (87)(88)(89)(90)(91)(92). In addition, a 2005 study (96) described a modified 13-residue neurotensin hormone that appeared to be arginylated, in vivo, at its internal Glu-4 residue. Neurotensin is produced through the processing of a larger precursor protein in the endoplasmic reticulum and Golgi. These compartments are distinct from the currently known locations of the cytosolic/ nuclear Ate1 R-transferase (96). A subsequent publication suggested that Ate1 is capable of arginylating not only the ␣-amino groups of N-terminal residues but also ␥-carboxyl groups of internal (non-N-terminal) Asp and Glu in natural proteins and described several proteins, including actin, dystrophin, myosin, titin, and tubulin, that appeared to contain, according to MSbased data, internal Asp or Glu residues conjugated to Arg (94).
To the best of our knowledge, none of the MS-based evidence about significant levels of protein arginylation either at noncanonical N-terminal residues or at specific internal residues (87-92, 94 -96) has been examined, thus far, using independent, non-MS methods. We addressed the possibility of noncanonical arginylation by asking, using C-terminally immobilized 11-residue CelluSpots peptides and [ 14 C]arginylation assays, whether specific amino acid sequences that encompass the previously reported non-canonical arginylation sites (87-92, 94 -96) are capable of being significantly arginylated by the purified Ate1 R-transferase.
One set of [ 14 C]arginylation assays was carried out with the peptide ALYENKPRRPY and its derivatives at position 4. The 11-residue sequence ALYENKPRRPY is identical to the 13-residue sequence of mature neurotensin (QLYENKPRRPYIL) (96), save for the absence of the C-terminal Ile-Leu sequence of neurotensin and the presence of N-terminal Ala, instead of wild-type N-terminal Gln, in the 11-residue CelluSpots peptide. The Gln 3 Ala N-terminal replacement was necessary due to the possibility of a nonenzymatic deamidation of wild-type N-terminal Gln (76), because it would yield N-terminal Glu, a canonical substrate of Ate1 R-transferase (Fig.  1A). The CelluSpots arrays contained not only the above 11-residue ALYENKPRRPY peptide but also the otherwise identical peptides that differed by bearing, at position 4, either Asp, Asn, or Gln, instead of wild-type Glu (Fig. 8, B  (columns 17-20) and C). The earlier MS-based evidence (96) for the in vivo arginylation of the neurotensin's internal Glu-4 residue was summarized above.
We found that none of the neurotensin-based peptides (including the one containing the wild-type Glu-4 residue) generated a [ 14 C]arginylation signal above background levels of CelluSpots assays with four individually examined purified Ate1 R-transferase isoforms, indicating the inability of any one among these peptides to serve as Ate1 substrates at least in vitro (Fig. 8, B (columns 17-20) and C (items 17-20)). Given this robustly negative and reproducible result, we did not use Cel-luSpots assays to examine analogous claims, by another laboratory, about arginylation of internal Asp or Glu in actin, myosin, and other cellular proteins (94,95).
To address a related but different set of earlier reports about non-canonical specificities of Ate1, we carried out CelluSpots arginylation assays with 11 C-terminally immobilized 11-residue peptides (Table 1, entries 53-56 and 74 -83) whose sequences were identical to the N-terminal sequences of some among nearly 50 proteins that have been described previously as in vivo Ate1 substrates that were Nt-arginylated at non-canonical (non-Asp, non-Glu, non-Cys) N-terminal residues (87). In most cases, Nt-arginylation of this kind was reported to involve non-canonical N-terminal residues of presumed (i.e. neither isolated nor otherwise characterized) natural protein fragments, generated in vivo by unknown proteases (87).
The putative Nt-arginylation substrates listed below were selected from a much larger set of claimed non-canonical Ate1 substrates (87) to encompass a broad range of reported noncanonical N-terminal residues, specifically Met, Leu, Pro, Ala, Lys, Gly, Asn, Phe, His, and Val. The 11 resulting 11-residue peptides were identical to the N-terminal sequences of either a full-length protein or (presumed) protein fragments that have been reported as substrates of non-canonical Nt-arginylation (87). The CelluSpots arginylation assays with these peptides also contained, as items 1 and 12, a positive (N-terminal Aspbearing) control.
1. DKGLAGLPASC: the Rgs4-derived, 11-residue peptide, a positive control. It contains N-terminal Asp, a canonical (Ntarginylatable) residue, and was present in the array at two positions, as described in Fig. 8, A (items 1 and 12) and B (columns 1 and 12).
2. MCDRKAVIKNA: the light chain of cytoplasmic dynein (Dynnl1); NP_062656.2. The reported Nt-arginylated residue is the initial Met-1 of Dynnl1 (87). We note that the N-terminal Met residue, in the sequence Met-Cys, is expected to be cotranslationally cleaved off, from a nascent protein, by Metaminopeptidases (9). The authors of Ref. 87, which reported Nt-arginylation of the Met-1 residue of the full-length Dynnl1, did not mention this issue.
3. LEVLNFFNNQI: a presumed fragment of the Ras suppressor protein 1; NP_033131.2. The Nt-arginylated residue is Leu-65 (i.e. the authors (87) reported that their protein preparation contained a (presumed) fragment of this protein that was Nt-arginylated at its N-terminal Leu-65).
6. KPVYDELFYTL: a presumed fragment of the EH domain containing protein Mpast2; AAH12272.1. The arginylated residue is Lys-448 (i.e. the authors (87) reported that their protein preparation contained a (presumed) fragment of this protein that was Nt-arginylated at its N-terminal Lys-448).
7. GDDGAEYVVES: a presumed fragment of the protein similar to GAPDH; XP_979290.1. The Nt-arginylated residue is Gly-43 (i.e. the authors (87) reported that their protein preparation contained a (presumed) fragment of this protein that was Nt-arginylated at its N-terminal Gly-43).
12. MDPLNDNIATL: the same protein as in item 10 above, but a (presumed) different fragment of the non-muscle myosin Myh9 (NP_071855.1). The Nt-arginylated residue of this fragment is Met-589 (i.e. the authors (87) reported that their protein preparation contained a (presumed) fragment of this protein that was Nt-arginylated at its N-terminal Met-589).
We found that 10 of these 11 CelluSpots peptides, corresponding to N-terminal sequences of some among the previously reported non-canonical substrates of the Ate1 R-transferase (87), generated background levels of [ 14 C]Arg incorporation with four individually examined purified Ate1 isoforms, indicating the inability of these peptides to serve as significant Ate1 substrates at least in vitro (Fig. 8, A (items 2-11  and 14) and B (columns 2-11 and 14)).
Importantly, the CelluSpots arrays containing these 11-residue peptides were flanked by spots of the positive-control peptide DKGLAGLPASC (item 1 above), which contained N-terminal Asp, a canonical (Nt-arginylatable) residue. As expected, FIGURE 8. Use of CelluSpots assays to address reports about arginylation of non-canonical N-terminal residues and specific internal residues of cellular proteins. A, list of 11-residue CelluSpots peptides whose sequences were identical to the N-terminal sequences of either a fulllength protein (item 2) or presumed natural fragments of other proteins (items 3-11) that have been reported to be Nt-arginylated by Ate1 at their non-canonical (non-Asp, non-Glu, non-Cys) N-terminal residues (87)(88)(89)(90)(91)(92). See "Results" for a detailed list of these peptides, which contains additional information and relevant references. Items 1-12 refer to a positivecontrol peptide that contained N-terminal Asp, a canonical (Nt-arginylatable) residue. B, [ 14 C]Arg autoradiograms of CelluSpots arginylation assays with the four purified Ate1 R-transferase isoforms, Ate1 1A7A , Ate1 1A7B , Ate1 1B7A , and Ate1 1B7B , and with 20 peptides shown in A (items 1-12) and C (items [13][14][15][16][17][18][19][20]. Column 1, the positive-control peptide bearing the canonical (Nt-arginylated) N-terminal Asp residue (see item 1 in A). Columns 2-11, the sequences of these peptides are shown, respectively, in items 2-11 in A. Column 12, the same positive-control peptide (bearing the canonical N-terminal Asp residue) that was also present in peptide spots of column 1 (see item 1 in A). , the sequences of these peptides shown in items 13-16 in C. Column 17, the peptide shown in item 17 in C. The sequence of this peptide, ALYENKPRRPY, was identical to the 13-residue sequence of the mature neurotensin hormone (QLYENK-PRRPYIL) (96), save for the absence of the C-terminal Ile-Leu sequence of neurotensin and the presence of N-terminal Ala, instead of wild-type N-terminal Gln. A fraction of the QLYENKPRRPYIL neurotensin was reported to be arginylated at its internal Glu-4 residue (96); hence, the use of CelluSpots arginylation assays and the four purified isoforms of the DKGLAGLPASC was efficaciously Nt-arginylated by Ate1 isoforms, in striking contrast to the absence of significant arginylation of the other 11-residue peptides (Fig. 8, A (items 1 and 12) and B (columns 1 and 12); compare with Fig. 8, A (items 2-11  and 14) and B (columns 2-11 and 14)). Possible causes of very low but detectable background level arginylation in CelluSpots peptide arrays, as distinguished from negligible (undetectable) levels of [ 14 C]Arg incorporation, are discussed above.
The single and weak potential exception, in which a slightly above-background level of [ 14 C]Arg incorporation was observed with two of the four examined Ate1 isoforms, was the peptide KPVYDELFYTL, corresponding to the (presumed) internal fragment of the Mpast2 protein (see item 6 above) (Fig.  8, A (item 6) and B (column 6)). The near-background levels of [ 14 C]Arg incorporation with the peptide KPVYDELFYTL (Fig.  8B, column 6; compare with columns 1 and 12) were significantly lower, in absolute amounts of incorporated [ 14 C]Arg, than the levels of [ 14 C]Arg vis-à-vis 11-residue peptides bearing the unmodified N-terminal Cys residue (Fig. 5, A and C). As described above, these N-terminal Cys-bearing peptides were classed as non-arginylatable, in agreement with earlier data (4,7,29).
We also tried to determine whether the report about noncanonical Nt-arginylation of, for example, the N-terminal Metbearing sequence MDPLNDNIATL (87) (see item 12 above) could have stemmed from in vivo oxidation of N-terminal Met to either Met-sulfoxide or Met-sulfone, a change that might have converted the non-arginylatable N-terminal Met into a possible substrate of Ate1. However, we found that the 11-residue peptides M(sulfoxide)DPLNDNIATL and M(sulfone) DPLNDNIATL were just as incapable of arginylation by the four individually examined purified Ate1 isoforms as was the otherwise identical unmodified MDPLNDNIATL peptide, thereby precluding the above possibility at least in vitro (Fig. 2B, right, spots C6 -C8; Fig. 8, B (columns 14 -16) and C (items 14 -16)). In addition and as expected, the 12-residue negativecontrol peptide RMDPLNDNIATL, which contained the noncanonical N-terminal Arg residue (added during peptide synthesis to the 11-residue MDPLNDNIATL), also yielded background levels of [ 14 C]Arg incorporation in these assays (Fig. 2B, right, spot C5; Fig. 8, B (column 13) and C (item 13)).
It was previously suggested that the proposed ability of the Ate1 R-transferase to arginylate a variety of non-canonical N-terminal as well as internal residues in cellular proteins may be conferred by (unknown) Ate1 cofactors that can be active both in vivo and in cell extracts (87,91). If correct, this hypothesis may at least partly reconcile the direct and extensive discrepancy between the cited earlier reports (87)(88)(89)(90)(91)(92) and our arginylation data with peptide arrays and purified Ate1 (Figs. 2-5 and 8). Therefore, we modified the CelluSpots arginylation assays by adding an extract from mouse cells (the N2a cell line) to reaction mixes, either in the absence or the presence of the exogenous purified Ate1 1B7A isoform. This modification of arginylation assays did not alter any of our findings.
In particular, CelluSpots 14 C-arginylation assays with extract alone, containing only the endogenous Ate1, produced arginylation patterns similar to those generated in the presence of both the extract and the added purified Ate1 1B7A isoform (data not shown). Because the absolute levels of 14 C-arginylation were significantly lower in N2a extract-only assays (in the absence of added exogenous Ate1), we cannot rule out the possibility that procedures for preparing the extract either damaged or diluted the postulated (unknown) Ate1-modifying factors. With that caveat, although explanations of the above discrepancy based on the presence, in vivo, of (unknown) factors that may alter the specificity of Ate1 remain formally possible, they were not supported by the above assays.
Conclusions-In the present study, CelluSpots arrays of C-terminally immobilized synthetic 11-residue peptides, were used, in conjunction with protein degradation assays, to analyze the specificity of arginylation by purified isoforms of the mouse Ate1 R-transferase, a component of the Arg/N-end rule pathway ( Figs. 1 and 2). We showed, among other things, that amino acid sequences immediately downstream of a substrate's canonical (Nt-arginylatable) N-terminal residue (Asp, Glu, or oxidized Cys), and particularly a residue at position 2, can affect the rate of Nt-arginylation by R-transferase and thereby the rate of degradation of a substrate protein (Figs. 3-7).
Other CelluSpots arginylation assays of the present study ( Fig. 8) were designed to examine specific interpretations of MS-based data that have been published over the last decade, largely by the Kashina laboratory, about the ability of the Ate1 R-transferase to arginylate a variety of non-canonical (non-Asp, non-Glu, non-Cys) N-terminal residues and also specific internal (non-N-terminal) Asp or Glu residues in many cellular proteins (87-92, 94 -96). The chief reason for concerns about the validity of the cited reports is a technical one, inasmuch as the evidence for non-canonical arginylation specificities of Ate1 remained almost entirely MS-based (87-92, 94 -96), without a significant attempt, until the present study with Cellu-Spots arrays, to examine interpretations of MS data by other, independent methods.
We addressed the 2005 report by the Sillard laboratory about arginylation of the hormone neurotensin at its internal Glu-4 residue (96) by carrying out CelluSpots arginylation assays using four purified Ate1 isoforms and 11-residue analogs of the 13-residue neurotensin. The robustly negative results of these arginylation assays (Fig. 8, B (columns [17][18][19][20] and C (items 17-20)) did not contradict the above findings (96) directly, given, in particular, the in vitro nature of the CelluSpots approach. Nevertheless, our results (Fig. 8) do raise concerns about the validity of MS-based interpretations that suggested the in vivo arginylation of internal (as distinguished from N-terminal) Asp or Glu residues in a number of cellular proteins (94 -96).
In addition and independently, our analyses, using Cellu-Spots arginylation assays and four purified Ate1 isoforms, of Nt-arginylation of non-canonical N-terminal residues encompassed 11 different 11-residue peptides. These peptides were identical to the N-terminal sequences of either full-length proteins or (presumed) protein fragments that have been reported to undergo Nt-arginylation at non-canonical N-terminal residues (87). Positive controls, a part of the CelluSpots arginylation assays, were 11-residue peptides bearing a canonical (Ntarginylatable) N-terminal residue, such as Asp (Figs. 2, 3, and 8,  A (items 1 and 12) and B (columns 1 and 12)). As described above, the results of these arginylation assays were reproducibly and robustly negative. Specifically, we observed background levels of 14 C-arginylation of the 11 examined 11-residue peptides (Fig. 8). Their sequences were identical to the N-terminal sequences of 11 among nearly 50 proteins that have been previously described as substrates of Ate1 that were Nt-arginylated at non-canonical (non-Asp, non-Glu, non-Cys) N-terminal residues (87).
A caveat to the otherwise direct contradictions between our findings and those in the cited earlier studies is that our results were produced in vitro (with purified Ate1 R-transferase and peptide arrays), whereas most reports about non-canonical arginylation of cellular proteins were about analyses of natural, in vivo-produced proteins (87-92, 94 -96). Nevertheless, given the clear absence of in vitro support for MS-based interpretations that suggested significant levels of non-canonical arginylation of both N-terminal and internal residues in cellular proteins, it would be particularly important to verify the above interpretations definitively, using methods that do not rely on MS alone.

Experimental Procedures
Plasmids, cDNAs, and Primers-The NEB Turbo Escherichia coli strain (New England Biolabs, Ipswich, MA) was used for cloning and maintaining plasmids. Phusion High-Fidelity DNA polymerase (New England Biolabs) was used for PCR. All constructed plasmids ( Table 2) were verified by DNA sequencing. pKP496, the parental plasmid used for constructing URT-based plasmids, has been described previously (44).
1-liter cultures of BL21 (DE2) E. coli transformed with one of the above plasmids were grown to an A 600 of ϳ0.6 at 37°C in LB containing 50 g/ml ampicillin. Cultures were cooled on ice for 45 min and then induced by the addition of isopropyl-␤-D-thiogalactoside to a final concentration of 0.25 mM for 6 h at room temperature. Cells were harvested by centrifugation at 2,000 ϫ g for 10 min at 4°C. Pelleted cells were resuspended in 15 ml of PBS containing 30% (v/v) glycerol, 0.3 M NaCl, 12 mM imidazole, 20 mM ␤-mercaptoethanol, and 1 mg/ml lysozyme, followed by freezing in liquid nitrogen. Samples were then thawed slowly on ice and centrifuged at 27,000 ϫ g for 30 min at 4°C. The N-terminal His 10 -Ub fusions of Ate1 isoforms were then purified by nickel-nitrilotriacetic acid chromatography, with elution by 0.3 M imidazole. Eluted proteins were dialyzed overnight against 5% (v/v) glycerol, 0.3 M NaCl, 2 mM ␤-mercaptoethanol, and 50 mM Na 2 HPO 4 -NaH 2 PO 4 (pH 8.0), containing 1ϫ Complete EDTA-free protease inhibitor mixture (Roche Applied Science). The N-terminal His 10 -Ub moiety was cleaved off by incubating dialyzed samples for 1 h at 37°C with purified Usp2cc deubiquitylase (111). The latter was added at the 1:10 molar ratio. The resulting samples were dialyzed against 5% (v/v) glycerol, 0.15 M NaCl, 10 mM ␤-mercaptoethanol, 50 mM Tris (pH 7.4), followed by further purification of (deubiquity-   CelluSpots-based Arginylation Assays-CelluSpots peptide arrays spotted on glass slides were synthesized by Intavis (Köln, Germany). Each slide contained two side-by-side identical copies of the array (Fig. 2, A and B). Mouse Ate1-dependent arginylation assays were performed as described previously (79), with slight modifications. Individual (rectangular) peptide arrays were "outlined" with an ImmEdge hydrophobic barrier pen (Vector Laboratories, Burlingame, CA) to minimize the required reaction volume. 90-l reaction samples containing 5.8 M [ 14 C]Arg (PerkinElmer Life Sciences, NEC267E050UC; 346 mCi/mmol), purified total E. coli tRNA (0.6 mg/ml), and E. coli aminoacyl-tRNA synthetases (800 units/ml; Sigma-Aldrich) in 5 mM ATP, 0.15 M KCl, 10 mM MgCl 2 , 1 mM DTT, 50 mM Tris-HCl (pH 8.0) were preincubated at 37°C for 15 min to allow the formation of Arg-tRNA. Thereafter, a specific purified Ate1 isoform (2 g) was added, and the resulting mixtures (final volumes of 0.1 ml) were added to each array. Arginylation was allowed to proceed for 1 h at 37°C in a humidified chamber with gentle rocking. Reactions were quenched by rinsing the arrays twice with a large volume excess (10 ml) of Tris-buffered saline (TBS) (1.15 M NaCl, 50 mM Tris-HCl, pH 7.5) containing, in addition, 1% Tween 20 and RNase A (10 g/l). Arrays were then washed three times for 20 min each in the same buffer (including RNase A) at 37°C, followed by an overnight (ϳ16-h) wash, at room temperature, in TBS containing 1% Tween 20 but no RNase A. The resulting slides were washed once with 10 mM Tris, pH 7.5, and were allowed to air-dry, followed by autoradiography for varying times (usually ϳ2 weeks), using a Phos-phorImager-type screen. The exposed screens were scanned using a Storm-860 scanner (Molecular Dynamics, Caesarea, Israel). Quantification of 14 C spot intensities was performed using the array analysis module of ImageQuant TL software (GE Healthcare, Little Chalfont, UK).
In Vitro Transcription-Translation-Degradation Assays-The TNT T7 coupled transcription/translation system (Promega, Madison, WI) was used to carry out transcription-translation-degradation assays, largely as described previously (44,50,73) (see also the Tansey Laboratory Protocol, "Ultimate Mammalian Denaturing IP," available on the World Wide Web). Nascent proteins in reticulocyte extract were pulse-labeled with L-[ 35 S]methionine (0.55 mCi/ml, 1,000 Ci/mmol; MP Biomedicals, Santa Ana, CA) for 10 min at 30°C in the total volume of 30 l. The labeling was quenched by the addition of cycloheximide and unlabeled methionine to final concentrations of 0.1 mg/ml and 5 mM, respectively, bringing the total volume to 40 l. Samples of 10 l were removed at the indicated time points. The reaction was immediately terminated by the addition of 80 l of TSD buffer (1% SDS, 5 mM DTT, 50 mM Tris-HCl, pH 7.4). A sample was snap-frozen in liquid nitrogen and was stored at Ϫ80°C until further processing, which comprised heating the samples at 95°C for 10 min, followed by dilution with 1 ml of TNN buffer (0.5% Nonidet P-40, 0.25 M NaCl, 5 mM Na-EDTA, 50 mM Tris-HCl, pH 7.4) containing 1ϫ Complete protease inhibitor mixture (Roche Diagnostics). Samples were thereafter immunoprecipitated using 10 l of anti-flag M2 magnetic beads (Sigma-Aldrich), with rocking at 4°C for 4 h, followed by four washes in TNN buffer, resuspension in 20 l of SDS-sample buffer, and heating at 95°C for 10 min. Immunoprecipitated proteins were fractionated by SDS-4 -15% PAGE, followed by autoradiography. Quantification of autoradiograms was carried out using a Storm 860 scanner (Molecular Dynamics).
Author Contributions-B. W., K. I. P., C. S. B., and A. V. designed the experiments. B. W. performed the experiments with participation by K. I. P., and C. B., B. W., and A. V. wrote the paper. All authors discussed the results and commented on the manuscript.