RES_ DUE^_T.N_PROTEINS_

Αi__ __& _l-____ -m2.- .m- -_l_. ___i--_ll L_ _l:

In both bacterial and eukaryotic cells, relatively long-lived proteins, whose half-lives are close to or exceed the cell generation time, coexist with proteins whose half-lives can be less than one percent of the cell generation time. Rates of intracellular protein degradation are a function of the cell's physiological state, and appear to be controlled differentially for individual proteins. In particular, damaged and otherwise abnormal proteins are metabolically unstable in vivo. Although the specific functions of selective protein degradation are in most cases still unknown, it is clear that many regulatory proteins are extremely short-lived ^n vi _; v . Metabolic instability of such proteins allows for rapid adjustment of their intracellular concentrations through regulated changes in rates of their synthesis or degradation. The few instances in which the metabolic instability of an intracellular protein has been shown to be essential for its function include the ell protein of bacteriophage lambda and the HO endonuclease of the yeast Saccharoijces H_ _____{.L §.k__^__. •

Most of the selective turnover of intracellular proteins under normal metabolic conditions is ATT-de endent and (In eukaryotes) τionlysosomal .

Recent biochemical and genetic evidence indicates that, in eukaryotes, covalent conjugation of ubi- quitin to short-lived intracellular proteins is essential for their selective degradation. The rules which determine whether a given protein is metabolically stable or unstable I.n y.iyo. were previously unknown.

Summary of _, "th& .Invention

This invention pertains to methods of engineer- ing the amino-terminus of proteins thereby con¬ trolling the metabolic stability and other proper¬ ties of a protein. Further, this invention provides a method for either in vivo or i.n vitro production of proteins with any of the twenty amino acid residues (or analogs thereof) at the protein's amino-terminus . The Invention Is based in part upon the striking discovery that the i.n iy half-life of an intracellular protein is a function of its amino-terminal amino acid residue and upon a novel (and more generally applicable) technique that allows one to generate proteins with specified amino-termini in y_iy£ or In yitr£. The invention also pertains to a newly Identified protease, ubiquitin-specific processing protease, which has properties that allow one to expose, either in yitι:£ ^or ΪS ZiX _> any desired amino acid residue, other than proline, at the amino-terminus of a protein of Interes .

The nature of the amino acid exposed at the amino- terminus of an intracellular protein was shown to be one crucial determinant that specifies whether a protein will be long- or short-lived i.n yjiv£. Individual amino acids can be categorized as either stabilizing or des abilizing amino acids with respect to the half-life that they confer upon a protein when exposed at the protein's amino- terminus . Destabilizing amino acid residues confer short half-lives, down to a few minutes for some of the destabilizing amino acids. Stabilizing amino acid residues confer long half-lives of many hours. This striking and newly discovered dependency of a protein's half-life on its amino- terminal residue is referred to herein as the N-end rule.

For some proteins, the presence of a destabili¬ zing amino acid at the amino terminus is necessary but not sufficient for des abilizatio . This is so because the complete amino- terminal degradation signal in a short-lived protein comprises two distinct de erminants, each of which is necessary, but each of which, by itself, is insufficient for efficient des abilization of a protein. One determinant, described above, is the amino - terminal residue of the protein. The second determinant, described below, is a specific internal lysine residue. The ability of this critical lysine residue to serve as the second determinant is to a significant extent independent of the amino acid

sequences surrounding the residue. Instead, an essential feature of this critical lysine residue Includes its spatial proximity to the protein's amino-terminus . Based upon the N-end rule, the amino-terminus of a protein can thus be designed or altered to change the intracell_lar half-life of ^" the protein and in this way the lifetime and/or activity of the protein in yLy can be regulated. This capability can be exploited for rational protein design in many different contexts. Natural proteins can be modified to render them more or less resistant to degradation in yi.y£. The design or alteration of the protein can be done at the protein level or at the genetic (DNA) level. For example, proteins can be modified by chemically altering or engineering the amino-terminus to provide for exposure at the amino-terminus of an amino acid residue of the stabilizing or destabilizing class. At the genetic level, genes encoding proteins can be made to encode an amino acid of the desired class at the amino - terminus so that the expressed protein exhibits a predetermined amino-terminal structure which renders it either metabolically stable or unstable with respect to the N-end rule pathway of proteolytic degradation. Amino -terminal regions can be engineered to provide appropriately located lysine residues in the context of a sufficiently seg- mentally mobile amino terminus to produce

destabilized protein. Furthermore, proteins can be expressed fused to a "masking" protein sequence which masks the engineered amino- terminus so that when unmasked the protein will exhibit the desired metabolic stability or other properties that depend on the nature of the protein's amino- terminal residue. In such constructs, for example, the junction between the two protein sequences can be designed to be cleaved specifically, for instance, by an endopro ease . Endoproteoly ic cleavage of the fused sequence unmasks the specifically engineered amino - terminus of the protein of interest and subjects the protein to degradation governed by the N-end rule. One specific and new way to engineer the protein's amino- terminus is provided in this invention by the identification of ubiquitin- specific processing protease and de ermina ion of its substrate specificity. Using this protease, fusions of ubiquitin with other proteins can be specifically processed either m y_it.r£ or m vι.v£ to generate proteins with desired amino - terminal residues .

A different, and also new way to specifically engineer short-lived proteins is provided in this invention by the discovery that ubiquitin- ro ein fusions, such as ubiquitin-Pro - β - alactosidase , that cannot be efficiently deubiquitinated, are metabolically unstable. Thus, by attaching the amino-terminal ubiquitin moiety ro a -pτorein in a

way that makes its removal either impossible or inefficient, one can destabilize proteins by a distinct technique that is not directly based on the N-end rule . In addition, variant cells can be developed which contain putative mutations in the "N-end" degrading proteasε(s) which either conditionally or noncoήditlonally stop degrading short-lived pro¬ teins. These cells can be used to overproduce proteins that ordinarily would be short-lived within the cell.

Brief Description of the Figures

Figure 1 shows the construction of ubiquitin- _ .__~S_? g ^ene fusions. Figure 2 shows experiments in which the half- lives of engineered 3-gal proteins are directly measured.

Figure 3 shows the changing of amino acid residues at the ubiquitin-/3-gal junction (A) using the newly discovered properties of ubiquitin- specific processing protease and the amino acid sequence in the vicinity of the junction (B).

Figure 4 shows the presence of multiple ubi¬ quitin moieties in metabolically unstable /3-gal proteins .

Figure 5 shows a series of /3-gal species containing ubiquitin in metabolically unstable 3-gal proteins.

Figure 6 shows that both prokaryotic and eukaryotic long-lived intracellular proteins have stabilizing amino acid residues at their amino- termini whereas secreted proteins exhibit a com¬ plementary bias.

Figures 7 and 8 show the construction of ubiquitin fusions with mouse dihydrofolate reductasε .

De_£ai.l_ d_D s_ rip_ _^£n_£f__th.e__I_nve_nt.i.£n The elucidation of the N-end rule is described in detail below. Briefly, this rule governing protein degradation was revealed by examining the vivo half-lives of the enzyme β- alactosidase having various amino acid residues at its amino - terminus and produced as a fusion protein with ubiquitin.

When a chimeric gene encoding a ubiquitin- β - alac - tosidase fusion protein is expressed in the yeast S_ . ___ _ --___ __ _\ ____ ■ ubiquitin is cleaved off the nascent fusion protein, yielding a deubiquitinated- β- galac - tosidase (/3gal) . With one exception, this cleavage takes place efficiently regardless of the nature of the amino acid residue of /3gal at the ubiquitin-/3gal junction, thereby making it possible to expose selectively different residues at the amino-termini of otherwise identical /3gal proteins. The /3gal pro¬ teins so designed exhibited strikingly different half-lives vivo, ranging from more than 20 hours to less than 3 minuter, depending upon the nature of

the amino acid at the amino-terminus of /3gal. Amino acids can be thus ordered according to the the half-lives they confer on /3gal when present at its amino-terminus . For example, the amino acids methlonine, serine, alanine , threonine, valine, glycine and cysteine confer a half-life of more than 20 hours. Phenylalanine, leucine, asparaglne, ani lysine yield half-lives of about three minutes. Arginine, the most destabilizing amino acid, confers a half-life of about two minutes. (See Table 1 below for complete list of amino acids and the correspond¬ ing half-lives)

35 A similar result is observed when S labelled proteins are synthesized in E. coli, isolated, and added to a mammalian cell lysate, specifically, the well characterized rabbit reticulocyte lysate system. In such a system, for example the following amino-terminal residues can be characterized as destabilizing: arginine, lysine, histldine, phenylalanine, leucine, tryptophan, tyrosine, alanine, serine, threonine, aspartic acid, glutamic acid, glutamine , cysteine and asparagine. Whether a particular amino acid Is destabilizing in any eukaryotic system can be determined. Through the course of these studies, it has been determined that the N-end rule has a hierarchical structure. Specifically, amino-terminal Glu and Asp (and also Cys in reticulocytes) are secondary destabilizing re-sidues

in that they are destabilizing through their ability to be conjugated to primary destabilizing residues such as Arg. Amino- terminal Gin and Asn are tertiary destabilizing residues in that they are des abilizing through their ability to be converted, via selective deamidation, into secondary des abilizing residues Glu and Asp.

Currently known amino- terminal residues in long-lived, noncompartmentalized intracellular proteins from both prokaryotes and eukaryotes belong virtually exclusively to the stabilizing class of amino acids, exactly as predicted by the N-end rule. This result stongly implicates the N-end rule in the selective degradation of intracellular proteins in general .

The appropriate amino - terminal amino acid appears to be an essential (though not necessarily a sufficient) requirement for the metabolic stability of a noncompartmentalized, intracellular protein. Thus, in order for a protein to be relatively stable intracellularly , a stabilizing amino acid should be present at the amino - terminus . The presence of a destabilizing residue at the amino - terminus of a protein is often, though not always, sufficient for its metabolic destabilization i.n vi.V£ . When such destabilization occurs to a relatively small extent, further analysis shows either an insufficient accessibility of the amino - terminus or a lack of the second determinant of the complete amino-terminal

degradation signal. In these instances, this second determinant, which by itself does not metabolically destabilize a protein, must be present In order for the half- life of a protein in vivo to be strongly dependent on the nature of its amino-terminal residue. The second determinant of the amino-terminal degradation signal was found to be a specific internal lysine residue. The ability of this critical lysine residue to serve as the second determinant was shown to be largely independent of unique amino acid sequences surrounding the residue. Instead, an essential features of the critical lysine residue includes Its spatial proximity to the protein's amino-terminus . The presence of a stabilizing amino acid at the amino- terminus at least in some cases (for instance, as observed for /3-gal) will confer stability upon the protein. However, a stabilizing amino acid at the amino -terminus may not always confer a long half-life because other degradative pathways may be Involved In determining the ultimate fate of the protein. For example, endoproteolytic cleavages (cleavages outside of terminal regions of the protein) may result In exposure of a destabilizing amino acid at the amino-terminus of a resulting product of the cleavage which is then rapidly degraded via the N-end rule pathway. Appropriate circumstances for use of a stabilizing amino acid can be ascertained empirically.

Although the N-end rule may be only one com¬ ponent (albeit a central one) of a more complex "half-life rule" which embraces other aspects of selective protein degradation i.n y.v , the N-end rule provides a rational, practical approach for de¬ signing or changing protein structure in order to produce proteins which are more or less resistant: to degradation by the N-end rule pathway than natural, unmodified protein. Proteins can be designed or modified at the protein or gene level to provide a desired amino acid of either the stabilizing or destabilizing class at their amino - terminus . Where necessary for destabilization , additional modifica¬ tions can be made to the amino - terminal region to provide appropriately located lysine residues. The ability to regulate the half-life of a protein will allow one to modulate the intracellular activity of the protein.

A straightforward approach to modifying a protein in order to increase or decrease its meta¬ bolic stability or to modulate other properties of the protein is to directly engineer the amino- terminus of the protein at the protein level. To provide a desired amino- terminal amino acid, the amino - terminus of the protein of interest can be chemically altered, for example, by adding an amino acid of the stabilizing or destabilizing class to the amino- terminus of a protein or polypeptide, em¬ ploying an appropriate chemistry. Thus, for

example, an unstable protein can be rendered more stable by adding a stabilizing amino acid residue (e.g. me hlonine serine, alanine, threonine, valine glycine or cysteine) to the amino-terminus of the protein. Conversely, a stable protein can be destabilized by adding a destabilizing amino acid to the amino-terminus . One distinct way to modify the amino- terminus of a protein would be to employ spe¬ cific enzymes , amino acid-protein ligases , which catalyze posttranslational addition of a single amino acid to the protein's amino-terminus . Other methods for nongenetic alterations of the same type can readily be ascertained by those skilled in the art. In some proteins, the amino- terminal end is obscured as a result of the protein's conformation (i.e. , its tertiary or quaternary structure) . In these cases, more extensive alteration of the amino-terminus may be necessary to make the protein subject to the N-end rule pathway. For example, where simple addition or replacement of the single amino- terminal residue is insufficient because of an inaccessible amino-terminus , several amino acids (including lysine, the site of ubiquitin joining to substrate proteins) may be added to the original amino - terminus to increase the accessibility and/or segmental mobility of the engineered amino terminus. Modification or design of the amino-terminus of a protein can also be accomplished at the genetic

level. Conventional techniques of site - directed mutagenesis for addition or substitution of ap¬ propriate codons to the 5' end of an isolated or synthesized gene can be employed to provide a desired amino- terminal structure for the encoded protein. For example, so that the protein expressed has the desired amino acid at its amino - terminus the appropriate codon for a stabilizing amino acid can be inserted or built into the amino - terminus of the protein- encoding sequence. Where necessary, the DNA sequence encoding the amino-terminal region of a protein can be modified to introduce a lysine residue in an appropriate context. This can be achieved most conveniently by employing DNA constructs encoding "universal destabilizing segments". A universal dest bilizing segment comprises a DNA construct which encodes a polypeptide structure, preferably segmentally mobile, containing one or more lysine residues, the codons for lysine residues being positioned within the construct such that when the construct is inserted into the structural gene, the lysine residues are sufficiently spatially proximate to the amino - terminus of the encoded protein to serve as the second determinant of the complete amino - terminal degradation signal. An example of a des abilizing segment is shown in the exemplification below (see Figures 7 and 8) . The insertion of such constructs Into the 5' portion of

a structural gene would provide the encoded protein with a lysine residue (or residues) in an appropriate context for destabilization.

At the same time, expressed proteins are often naturally modified within a cell after translation. These modifications can include changes at the protein's amino -terminus . For example, the amino- terminus can be acted on by an aminopeptidase which cleaves one or several amino acids from the amino- terminus. Amino acids may also be added to the amino-terminus by post-translational processing. This invention provides a way to "by-pass" still undefined rules of amino-terminal protein processing to expose exactly and specifically the desired amino acid residues at the amino - terminus of a mature pro¬ cessed protein species. To minimize the impact of such posttranslational events on the ultimate structure of the amino-terminus of a protein of in¬ terest, specific fusion proteins can be designed wherein the amino-terminus of a protein of interest (designed to have the desired stabilizing or de¬ stabilizing structure) is preceded by a "masking" protein sequence fused to the amino-terminus . The fusion proteins are designed so that the masking protein sequence fused to the amino-terminus of the protein of interest is susceptible to specific cleavage at the junction between the two. Removal of the protein sequence thus unmasks the amino- terminus of the protein of interest and the half-

life of the released protein is thus governed by the predesigned amino - terminus . The fusion protein can be designed for specific cleavage i^n vivo, for example, by a host cell endoprotease or for specific cleavage in a _in y_i_tr_o_ system where it can be cleaved after isolation from a producer cell (which lacks the capability co cleave the fusion protein) .

Ubiquitin is a broadly useful fusion partner for construction of a fused protein with a protein of interest: the discovery that artificial ubiquitin-protein fusions can be cleaved precisely by a cytoplasmic eukaryotic protease with little or no dependence on the protein to which ubiquitin is fused can be applied both j.n ; ^* _iv and _in yi.tr_£ in protein engineering strategies, and is a major aspect of this invention. For example, the ubiquitin-protein fusion method can be used to artificially generate authentic amino-termini in proteins produced through artificial means. Thus, amino-termini characteristic of natural eukaryotic or prokaryotic proteins can be generated by m vitro cleavage of ubiquitin-protein fusions produced in a prokaryotic host.

A specific methodology for producing ubiquitin- β - alac osidase fusion proteins is described in detail below. Genes encoding any other proteins can be substituted for LacZ (the /3-gal gene) in this methodology .

In general, ubiquitin fusion proteins are expressed by a chimeric gene construct comprising, in 5' to 3' orientation, a ubiquitin gene linked to a gene encoding the protein of interest. The codon for the amino-terminal amino acid of the protein of interest is located immediately adjacent the 3 ^f end of the ubiquitin gene. The fused gene product is cleaved endoproteolytically either iri vivo or . vi.t£ (using either pure or partially purified ubIquitin-speci ic protease identified in the present invention) at the junction between ubiquitin and the protein of interest to generate the protein of interest having the desired amino acid at its amino-terminus . There are a number of specific uses for the described ability to specifically engineer the protein's amino- terminus . One such use is es¬ tablished by the fact that the Intracellular half- life of the released protein is governed by the principles of the N-end rule. Other applications of the specific method for engineering the protein's amino-terminus described herein range from adjusting the desired functional properties of a protein of interest, to modulating its antigenicity, and again, to other uses that can readily be ascertained by those skilled In the art.

This method of generating the desired amino acid residue at the amino-terminus of a protein of Interest involves two novel components: one,- the

use of ubiquitin-protein fusions, and the other, the use of ubiquitin-specific processing protease that has been identified, and whose striking substrate requirements were discovered, in this work. Al- though the initial identification of the ubiquitin- specific protease has been made m vjiy_£ , the enzyme is also relatively stable and active in v_i_tr£ (in extracts) , and can readily be purified to homogeneity by techniques known to those skilled in the art. Furthermore, the substrate specificity of the ubiquitin- specific processing protease is highly conserved in evolution, being the same in yeast and mammals. The enzyme can be purified chromato¬ graphically from a crude extract by sequential chromatography on phosphocellulose , DEAE cellulose, and SH-Sepharose among other methods known to those skilled in the art. Al ernatively, the gene for this protease can be cloned by those skilled in the art . The cloned protease gene can be used either _in yi_v£, or, alternatively, the gene can be over- expressed in a suitable host, the overexpressed ubiquitin-specific protease purified and used for the same or similar purposes i_n itro . The dis- covery of this enzymatic activity, and detailed characterization of its substrate specificity herein provides for the yitr£ and m y_iy£ use of this enzyme .

The use of ubiqultin-protein fusions to allow the generation of a desired amino acid residue at the amino-terminus of a protein of interest can be extended to facilitate the purification of such proteins from producer cells. A gene can be readily constructed that encodes a convenient marker pro¬ tein, such as stra avidin, linked to a ubiqultin- protein fusion construct described above. The resulting (marker protein) -ubiqultin-protein fusion can be simply Isolated from producer cells by using the preselected property of the marker protein, for instance, the known ability of streptavidin to be isolatable by affinity chromato raphy on a biotln column. Thus, purified (marker protein) -ubiquitln- protein fusion can then be specifically cleaved by the ubiquitin- specific protease described In this invention to generate the final product, a protein of interest with the desired amino acid residue at its amino-terminus . The codon for the amino-terminal amino acid of the protein of Interest can be made to encode the desired amino acid by, for example, site-directed mutagenesis techniques currently standard in the field. If the gene encoding the protein of interest is a synthetic gene the appropriate 5' codon can be built-in during the synthetic process. Alter¬ natively, nucleotides for a specific codon can be added to the 5' end of an isolated or synthesized gene by ligation of an appropriate DNA sequence to

the 5' (amino - terminus -encoding) end of the gene. DNA inserts encoding appropriately located lysine residues (such as the "universal des abilizing segments" described above) can be inserted into the 5' region to provide for the second determinant of the complete amino - terminal degradation.

Ubiquitin- like fusion partners capable of being cleaved by the ubiquitin- specific protease can also be used. In addition, fusion partners other than ubiquitin for masking the amino - terminus of a protein of interest can be used. For example, functional homologues of ubiquitin from eukaryotes or prokaryotes may be used. In appropriate cases, the fusion proteins can be designed to contain a proteolytic cleavage site for a restriction endo¬ protease which has sufficiently narrow specificity so that only one target site is cleaved in a fusion protein. A crucial property of such a protease must be a sufficiently relaxed requirement for the nature of the amino acid residue(s) abutting the carboxy- terminal side of the cleavage site. The target site for cleavage is the junction between the fusion partner and the amino - terminus of the protein of interest and thus the recognition site for the endoprotease is located to provide for cleavage at this location. The commerically available protease, complement factor X , exhibits these properties and thus can be used to directly generate proteins with predetermined amino acid residues in the ultimate

position of their amino-termini (see, K. Nogai and H.C. Thogersen Nature ^0_9_:810 (1984)) . The recog¬ nition site for the endoprotease can be engineered into the junction between the masking protein sequence and the 3' region encoding the amino- terminus of the protein of interest.

&_ different and distinct method for engineering short-lived proteins is provided in this invention by the discovery that ubiqultin-protein fusions, such as ubiquitin-Pro - β-galactosidase fusion (Table 1) , that cannot be efficiently deubiquitinated are metabolically unstable. Thus, by attaching the amino-terminal ubiquitin moiety to a protein in a way that makes its removal either impossible or inefficient, one can destabilize a protein by a distinct technique which is qualitatively different from the method of generating the desired amino- terminus of a protein according to the requirements of the N-end rule. Prevention of the efficient deubiquitination of a ubiquitin-protein fusion can be achieved in several ways, for instance, by using a proline residue at the ubiqultin-protein junction as shown in Table 1, or by changing the amino acid sequence of ubiquitin near its carboxy1-terminus in such as way that the ubiquitin moiety is no longer recognized by the ubiquitin-specific processing protease but can still be recognized by the rest of the degradative pathway. These and other ways to reduce the rate of deubiquitination of a ubiquitin-

protein fusion can be readily ascertained by those skilled in the art.

The methods of this invention can be employed, iS_- _. ___._i iiiϋ _{: >} f° ^r regulating the half-life of a protein intracellularly . There are many instances where this capability is useful. For example, when a gene is Introduced into a cell for expression therein, the expressed product can be designed for a long or short half-life depending upon the particu- lar need .

In general, destabilized proteins which have short half-lives are more amenable to regulation of intracellular levels of the protein. The ability to finely regulate the intracellular levels and ac - tivity of a protein can be useful in therapy or in the work with _in vjLt_r£ cell cultures. In gene therapy, for example, a gene may be introduced into a cell to compensate for a genetic deficiency or abnormality. The gene can be inserted under control of an inducible promoter. Induction results in enhanced expression of the gene product and con¬ sequently, higher levels of the product within the cell. If the gene is designed to encode an unstable protein, the intracellular concentration of the expressed protein will be more quickly responsive to a later reduction in the rate of its synthesis because it does not persist within the cell. In this way, the intracellular level and/or activity of

the protein encoded by the inserted gene can be more finely regulated.

The method of this invention can also be used to expand the uses of selectable markers by shorten- ing the time necessary for a phenotype related to the marker to become manifest. Toward this end, a product encroded, by a marker gene can be destabilized by altering its amino-terminus according to the N-end rule. In this way selection for the negative phenotype can be facilitated because the product of the marker gene will be more quickly extinguished after the function of the gene encoding the marker is abolished. An example Is the thymidine kinase (tk) gene. The tk gene can be engineered to encode a less stable enzyme by introducing an appropriate destabilizing amino acid at the amino - terminus . Gene mutation resulting in tk phenotype will be more quickly manifested by cells because residual tk will be more quickly degraded. This can be especially useful In slow growing cells where more time Is required to "dilute out" tk synthesized prior ^' to transformation to the tk type.

The principles of protein modification based upon the N-end rule may also be employed in the design of cytotoxlns . Proteinaceous cytotoxins can be designed as unstable proteins degradable by the N-end rule pathway so that they do not persist after their toxic action has been exerted on a target

cell. Reducing the lifetime of the toxin reduces the likelihood of killing nontargeted cells.

Discovery of the N-end rule pathway of degra¬ dation allows development of mutant cells having mutations in genes encoding essential components of the N-end rule pathway. For example, cells can be produced that either permanently or condi ionally are unable to efficiently degrade otherwise short¬ lived proteins. These cells can be used to produce desired proteins that ordinarily would be unstable within a cell .

The invention is illustrated further by the following detailed description of the elucidation of the N-end rule .

H _tho_d_s_

_ ——_ ----:—--.-_.9.— __ ——__.— -. ϋ.• £££ _. __.Yi_Li_L_ _: cells carrying pUB23 (Fig. 1) , which encodes ub-Met-/3gal (Fig. 3A) , were labeled

35 with [ S] methionine, followed by extract pre- paration, immunoprecipitation of /3gal and elect- rophoresis as described below. The wet polyacryla- mide gel was subjected to autoradiography, the band of /3gal was excised, and the eleetroeluted 3gal was subjected to six cycles of radiochemical sequencing by Edman degradation. The sequencing was carried out by W. Lane at the MicroChem Facility of Harvard University .

ii£i_._:iΞ£_.__^_ H£ ££Ξ _.i£ pUB23 (figure 1) was treated sequentially with

Ace I, the Klenow fragment of pol I, and Bam HI. A fragment containing the Xho I site was purified and inserted between a filled-in Hind III site and a BAM HI site of the M13mp9 phage DNA. (J . Messing and J. Vieira, Gene 19, 2-63 (1982.)) . SIte-directed uta- genesis (M. Smith, Annu. Rev. Genet. 19, 423 (1985)) was carried out as described by Kramer, W. et a 1 . Nucl. Acl.d_s Res. 12., 9441 (1984) using a synthetic 25-resldue oligodeoxyribonucleotide , containing ten bases on the 5' side and twelve bases on the 3' side of the Met codon of gal. All four bases were allowed to occur at the original Met codon positions during synthesis. Primary phage plaques were screened by hybridization (Wood, N.I. e_t al. PNAS " 8_2^ 1585 (1985)) , with the use of a 12-residue oligonucleotide probe spanning the region of codon changes and hybridizing to the original sequence. Nonhybridizing plaques containing inserts of the expected size were sequenced by the chain termina¬ tion method. (Sanger, F. et a.1 . , PNAS 7J. 5463 (1977)) . To transfer the desired constructs into the pUB23 background, replicative form DNA of mutant phages was digested with Xho I and Bam HI, and added to the same digest of the plasmid pLGSD5-ATG (see Fig. 1 and L. Guarente , Meth.£cis Enzymol . , 10. 181 (1983)) . The ligated mixture was used to transform the E. coli strain KC1061. (M.J. Casadaban and S.N.

Cohen, J. Mol. B to1. , 13.8. 179 (1980)) . Colonies containing plasmlds of interest (in which the open reading frame of 3gal had been restored) were recognized by their light blue color on X-/3gal plates .

l__ _ι_ __._l_ι_i_ --_—Α--ι-i_ E -i.__.__--_-.

S: . ££revisiae cells of the strain BWG-9a-l (MAT ]_ii>A S_i._L_i _• _§_£. 6.) , transformed (F. Sherman e_ ajL . M t-h£d _^n_Ye_a _G ne_ti.£ Cold Spring Harbor Labo- ratory, N.Y. , 1981)) with plasmids of interest were grown at 30 C to A- _nn of approximately 5 in a medium of 2 percent galactose, 0.67 percent Yeast Nitrogen Base without amino acids (DIFCO) , adenine (10 μg/ml) and amino acids including methionine (Sherman, F. et -al. , s.up_ra) . Typically, cells from a 5 ml culture were harvested by filtration through the well of a Millipore microtiter filtration plate, washed several times on the filter with the same medium lacking methionine and resuspended in 0.3 ml of 1 percent galactose, 50 mM potassium phosphate buffer

35 (pH 7.4) . [ S]methionine (50 to 100 μCi) was then added for 5 minutes at 30 C; the cells were col¬ lected by filtration and resuspended on 0.4 ml of the growth medium containing cycloheximide at 0.5 mg/ml . Samples (0.1 ml) were withdrawn at indicated times, and added to 0.75 ml of cold buffer A (see below for buffer composition) containing leupeptin, pepstatin A, antipain, aprotinin and chymostatin

(Sigma) , (each at 20 μg/ml) in addition to 0.4 ml of glass beads. Immediately thereafter, the cells were disrupted by vortexing for approximately 3 minutes at 4 C ; the extracts were centrifuged at 12,000g for 3 minutes and the radioactivity of acid-insoluble

35

S in the supernatants was determined. Aliquots of the supernatants containing equal amounts of the total acid-insoluble S were processed for immuno- precipitation with a monoclonal antibody to /3gal. Ascitic fluid containing a molar excess of the antibody (at least tenfold) was added to each aliquot, with subsequent incubation at 4 C for 2 hours; protein A-Sepharose (Pharmacia) was then added, the suspension was incubated with rocking at 4°C for 30 minutes and centrifuged at 12,000g for 1 minute. The protein A-Sepharose pellets were washed three times in buffer A (see below) containing 0.1 percent sodium dodecyl sulfate (SDS) , resuspended in an SDS, dithiotreitol (DTT) -containing electro- phoretic sample buffer (U.K. Laemmli, Nature 227 6S0 (1970)) , heated at 100 C for 3 minutes, and cen¬ trifuged at 12,000g for 1 minute. Equal aliquots of the supernatants were subjected to electrophoresis in a 7 percent discontinuous polyacrylamide-SDS gel (15 by 15 by 0.15 cm) with subsequent flourography. In some experiments, the above protocol was not used, but the extracts were prepared by boiling cells directly in the presence of SDS, with es¬ sentially the same results.

Anal_ ^ _£f_ub^^£ l__£r£ i.n _£^ duced_i_n_E_ _; C_£l_ _. i

Plasmid pUB23 (Figs. 1 and 3) was introduced into DS410 , a minicell-producing E. £U strain. (N. Stoker, ej: a].. , in Transcri.£^i.on_arιd_Trans_ _j1

~ ~~ i_Ui£S- A_£ta t ^. i. al. -• • -" • Harnes and S. J. Higgins, Eds. , IRL press, Oxford, 1984, p.153) . Minicells were prepared and labeled for 60 minutes at 36 C with [ S]methionine ( 600 Ci/mmole, Amersham) as described by N. Stoker et al. , su£.ra . 0 Labeled minicells were centrifuged, resuspended in 2 percent SDS, 10 mM DTT, 10 mM Na-HEPES (ph 7.5) and heated at 100 C for 3 minutes. After centri- fugation at 12,000g for 1 minute the supernatant was diluted 20-fold with buffer A (1 percent Triton 5 X-100, 0.15 M NaCl, 5 mM Na-EDTA, 50 mM Na-HEPES, _P H 7.5) , followed by the addition of phenylme thy1 - sulfonyl fluoride (PMSF) and N- e hylmalei ide to 0.5 mM and lOmM, respectively. After 4 hours at 4 C , the sample was dialyzed against buffer A containing 0 0.5 mM PMSF overnight at 4 C, and processed for immunoprecipitation (as described above) .

Ana^y i _^ _£f_ub^^ al _^ _ r£t i.ns._ r£du e_d_j _: n_y as.£

S.. cetevisiae. cells carrying plasmids of interest were grown in 800 ml of a uracil-deficient medium, 5 then harvested and disrupted with glass beads in buffer A containing leupeptin, pepstatin A, anti- pain, aprotinin and chymostatin (each at 3 μg/ l) . The extract was centrifuged at 12,000g for 3

inutes . Saturated ammonium sulfate was added to the supernatant to a final concentration of 57 percent. After overnight incubation at 4 C , the precipitated protein was collected by centrifugation at 23,000g for 30 minutes. The pellet was redis- solved in buffer A containing protease inhibitors. After clari ication at 12,000g for 3 minutes, the sample was passed through an affinity column which had been prepared by crosslinking an IgG fraction from an ascitic fluid (containing a monoclonal antibody to gal to Affi-Gel 10 (Bio-Rad) . The IgG fraction used for crosslinking had been purified from the ascitic fluid by affinity chromatography on protein A-Sepharose. After washing with buffer A lacking Triton X-100, the antibody-bound proteins were eluted with 0.25 M glycine-HCl (pH 2.6) . The eluate was immediately adjusted to pH 7.5 with 1 M Na-HEPES (pH 8.5) , and thereafter made 0.1 percent in SDS. The sample was concentrated by ultrafil- tration in Centricon 30 (Amicon) , and subjected to electrophoresis in a 7 percent discontinuous poly- acrylamide-SDS gel (U.K. Laemmli, Nature_.(London) 22:7., 680 (1970)) . Electroblotting of proteins to nitrocellulose, and immunoblot analysis with a peptide-mediated antibody to ubiquitin were per¬ formed as described by P.S. Swerdlow, D. Finley and A. Varshavsky, Anal t. Biochem. 156, 147 (1986) . The same results were obtained with a different

antibody to ubiquitin obtained from A. Haas (Univ of Milwaukee Med. School) .

• ££JUϊ£H£.ii£Ξ_£i._.._ £U_^iES ^. £S__! ^ Ξ^.__iS£ - S-j

T n y__Ub^ ^£al._Fu i£rι_Pr£ ems. Four of the pKKUb -X- ^* gal vectors (those encoding Ub-MET-/3gal , Ub-Gin- 3gal , Ub -Arg-/3gal , and Ub -Pro -/3gal) were constructed as follows. Site- directed mutagenesis [M. Smith, Annu^ _______._ _: ^ S£^._

19_, 423 (1985) ; T. Maniatis, et a l , "Molecular Cloning", (Cold Spring Harbor Laboratory, N.Y. , 1982) ; "Current Protocols in Molecular Biology", F.M. Ausubel, e_t a 1_^, (Wiley-lnterscience, N.Y. , 1987] was used to insert the sequence GTAC between the first and second codons of the ubiquitin reading frame in the yeast expression vector pUB23 (which encodes Ub-Met-/3gal) and in its derivatives, (A. Bachmair, _ a.1 _ , S.£i £ £ 2_ 4 , 179 (1986) , encoding Ub-Arg-/3gal , Ub -Gin-/3gal , and Ub - Pro -jSgal . The insertion created a Kpn I site positioned such that when the vector is cut with Kpn I and the ends blunted by mung bean nuclease, the second codon of the ubiquitin reading frame starts precisely at one of the fragment's ends. Thus, digestion of each of the above four vectors with Kpn I and Tth 1111, followed by treatment with mung bean nuclease, yielded four fragments which contained the corresponding Ub -X- 3gal- coding sequences but lacked the first (ATG) codon of the ubiquitin reading frame. These fragments were subcloned into an E^

coli expression vector pKK233-2 [E. Amann and J. Brosius, Gene 40_, 183 (1985)] which had been prepared by digesting it with Nco I and filling in staggered ends using Klenow fragment of Pol I [M. Smith, Annu. Rev. Genet. 1 , 423 (1985) ; T.

Maniatis , e . al . , "Molecular Cloning", (Cold Spring Harbor Laboratory, N.Y. , 1982) ; "Current Erotocols in Molecular Biology", F.M. Ausubel, t: al^, (Wiley-Interscience, N.Y. , 1987)] . This step yielded the complete Ub-X-/3gal sequence (in which the ATG codon was supplied by the pKK233-3 vector) , optimally positioned downstream of the regulatable Ptrc promotor of the vector. To construct the remaining sixteen pKKUb-X-/3gal expression vectors, pKKUb-Arg-βgal was digested with Sail and BamHI. One of the two BamHI sites in pKKUb -Arg-/3gal is located at the junction between the ubiquitin- and /3gal-coding sequences; the other BamHI site, present in the Initial pKK233-2 vector [E. Amann and J. Brosius, Gene 40_, 183 (1985)] , was removed in a preliminary construction step. The small Sall/BamHI fragment (containing the P _r£ promoter, the complete ubiquitin-coding sequence, and the Arg codon at the Ub-/3gal junction) was subcloned into a M13rαp9 vector [M. Smith, Annu. Rev. Genet.. 19_, 423 (1985) ; T.

Maniatis , e_t a.1^ , "Molecular Cloning" , (Cold Spring Harbor Laboratory, N.Y. , 1982) ; "Current Protocols in Molecular Biology", F.M. Ausubel, t a.!^ , (Wiley-Interscience , N.Y. , (1987); J. Messing- and J.

Vieira, Gene 1.9, 263 (1982)] . A BstXI/BamHI fragment of this construct that contained a portion of the ubiquitin- coding sequence and the Arg codon at the Ub-/3gal junction, was then exchanged for the sixteen otherwise identical BstXI/BamHI fragments [from the previously made, M13mp9-based constructs A. Bachmair e a 1^ , Sαience 2.3.4, 179 (1986) ] , which differed exclusively in a codon at the Ub-/3gal junction. The resulting sixteen M13 -mp -based constructs were treated with Sail and BamHI, and the small Sall/BamHI fragments containing, the ubiquitin-coding sequence and different single codons at the Ub-/3gal junction were cloned back into pKKUb -Arg-/3gal , replacing the original Sall/BamHI fragment, and yielding the remaining sixteen pKKUb-X-/3gal expression vectors. In all cases, the identity of the amino acid encoded at the Ub-/3gal junction of a final pKKUb -X-/3construet was verified by subcloning into M13 and nucleotide sequencing by the chain termination method [M. Smith, Annu._Rev^_ Genet^ 1.9, 423 (1985) ; T. Maniatis, et a1^, "Molecular Cloning", (Cold Spring Harbor Laboratory, N.Y. , 1982) ; "Current Protocols in Molecular Biology", F.M. Ausubel, e_t a 1^ , (Wiley-Interscience, N.Y. , (1987)] .

35 Z£Ξ U£££ £S_£ _i_i _l i ---_ϊ- _-^--_^≤£i_ £ £ i _I£ 51

E . co l i

An overnight culture (1ml) of E_. ££U JM101 cells bearing one of the twenty pKKUb-X-/3gal expression vectors was diluted into 50 ml of Luria broth supplemented with ampicillin at 40 μg/ml, and the cells were grown with shaking for approximately 2 hours at 37°C. The cells were harvested by centrifugation at 4,000g for 10 minutes, washed twice with M9 buffer, and resuspended in 25 ml of M9 minimal medium supplemented with glucose (0.22%, w/v) , thiamine (18 μg/ml) , ampicillin (40 μg/ml) , 0.5mM isopropylthiogalactoside (IPTG), and 0.15 ml of 10.5% (w/v) Methionine Assay Medium (Difco) . After incubation with shaking for one hour at 37° C,

0.5 to 1.0 MCi of ³⁵ S-Translabel (ICN:-85% [ ³⁵ S]

35 methionine, -15% ^' [ S] cysteine) was added and shaking was continued for 5 minutes . Unlabeled

L-methionine was than added to ImM and shaking was continued for another 10 minutes. Cells were harvested, washed twice with M9 buffer, and resuspended in 0.5 ml of 25% (w/v) sucrose, 50mM Tris-HCl (pH8.0) . Thereafter, 0.1 ml of lysozyme (lOmg/ml, Sigma) in 0.25 M Tris-HCl (pH 8.0) was added, and the mixture was incubated at 0° C for five minutes, followed by the addition of 0.1 ml of 0.5 M Na-EDTA (pH 8.0) and further incubation at O'C for five minutes. The cell suspension was then added to a lysis solution (0.8 ml H_0, 50 μl of 1 M Tris-Hcl (pH 8.0), 125 μl of 0.5 M Na-EDTA (pH 8.0) , 10 μl of 10% (w/v) Triton X-100) , and gently mixed.

The lysate was centrifuged at 40,000g for one hour, and Ub-X-3gal was purified from the supernatant by affinity chromatography on an aminopheny1 hio - pyranogalactoside- garose (APTG- garose) column as described (A. Ullman, Gene 2£ , 27 (1984) . Ubiquitin-X- 3gal was eluted from APTG-agarose with 10 M 2 -mercap oe hano1 , 0.1 M Na-borate (pH 10.0) , dialyzed overnight at ° C against 50% (v/v) glycerol, 0.1 mM EDTA, 1 mM dithiothreitol (DTT) , 40 mM Tris-HCl (pH 7.5) , and stored at -20°C in the same buffer. Control experiments showed that the transient exposure of Ub-X-/3gal purified by the above proceudre were 0.5 - 1 m , with enzymatic

4 activity of 4 - 6 x 10 units/mg and specific radioactivity of 1 - 2 X 10 cpm/μg. Unlabeled Ub-X-/3gal was prepared essentially as described above except that after two hours of growth in Luria broth with ampicillin, IPTG was added to 0.5 mM, and the cells were grown for one more hour before harvesting and lysis.

2£S££^ :£i£S_£_[—∑ _i-i—-^S££_ιiS_L

Washed reticulocytes from phenylhydrazine - treated rabbits were purchased from Green Hectares (Oregon, Wisconsin) , and shipped overnight at O'C. The reticulocytes were washed three times with 3-4 volumes of standard phosphate -buffered saline (PBS) (centrifugations at lOOOg for 10 minutes at ° C) .

To deplete intracellular ATP [J. Etlinger and A. Goldberg, Proc. Natl. Acad. Sci. USA, 74, 54 (1977) ; A. Hershko , t al_. , Proc. Natl^ Acad. Sci . USA 77: 1783 (1980) ; Hershko et al . , J_ _; __Bi£l^_C em^, 258 , 8206 (1982)] , the cells were incubated for 90 minutes at 37° C in Krebs -Ringer phosphate buffer containing 0.2 mM 2, 4-dinitrophenol and 20 M 2-deoxyglucose , and then washed three times in PBS. Pelleted reticulocytes were then lysed at 0°C by resuspending the pellet in 1.5 volumes of 1 mM DTT. After -10 minutes at 0° C , the sample was centrifuges at 80,000g for 90 minutes at 4°C. The supernatant was removed, divided into aliquots, and stored under liquid nitrogen. Only once-frozen aliquots were used in all experiments. Unless stated otherwise, the ATP-depleted reticulocyte extract was used directly after thawing, without further processing. In some experiments, the thawed extract was at first dialysed overnight at 4°C against 1 mM DTT, 10 M Tris-HCl (pH 7.5) in dialysis tubing with a m.w. cutoff of -3 kD . Fraction II was prepared by DEAE chromatography of ATP-depleted reticulocyte extract as previously described [D. Finley and A. Varshavsky, Trends _Biochem. Sci. 10, 343 (1985) ; A. Herschko and A. Ciechanover, H_Z2.__l _z ._- _ : ___:£_-__. _-

Mol. Biol. 3..3, 19 (1986) ; S. Pontremoli and E. Me1oni , _Annu. Rev^Biochem._ 55, 455 (1986) ; M. Rechsteiner, Annu. Rev. Cell . Biol . !3 _> (1987); J.S. Bond and P.E. Butler, Annu. Rev. Biochem.. 56,

333(1987) ; J.F. Dice, FASEB_J^ 1, 349 (1987) ; J. Etlinger and A. Goldberg, Proc_ _L _Natl __Acad_ _L _Sci_ _L US. , 74, 54 (1977) ; A. Hershko, et a1^ , Proc^ a l^ A£ad^_S.ci^_US.A_22: ¹⁷⁸³ (1980) ; Hershko et_a1^, J^ 2 £i._≤tl S-_- > HH ' 8206 (1982)] and stored under liquid nitrogen. Reaction mixtures for assaying the degradation of test proteins in either the total reticulocyte extract of Fraction II contained (final concentrations) 5% (v/v) glycerol, 1 mM DTT, 5 mM MgCl ₂ , 50 mM Tris-HCl (pH 7.5) , 70% (v/v) reticulocyte extract (or Fraction II at 6 mg/ml of

35 the total protein) , [ S]Ub-X- gal fusion protein at

20 μg/ml, and when present, 0.5 mM ATP and an

ATP- regenerating system (10 mM creating phosphate, 0.1 mg/ml creatine phosphokinase) . Reaction mixtures were prepared as follows: a mixture complete except for ATP and ATP-regenerating system was incubated for ten minutes at 37 " C ; to allow for the deubiquitination of a Ub-X-/3gal fusion protein ATP and ATP-regenerating system were then added to start the ATP- dependent reactions in the extract and the 37°C incubation continued. Control reactions with the ATP-depleted extract were performed identically except that ATP and ATP- regenerating system were omitted. The ATP - dependent degradation of 125I-labeled bovine serum albumin, hen lysozyme, and cytochrome c from 2_ £ £ Y ^i£ (purchased from

Sigma, St. Louis, Missouri, and labeled using the chloramine T method (A. Ciechanover, et al^_ , Proc^.

Natl. .Acad. Sci. USA 77., 1365 (1980) was assayed as described above except that the 10-minute preincubatlon of the test protein at 37°C in the ATP-depleted reticulocyte extract was omitted. To follow the degradation of test proteins , aliquots were taken from the reaction mixture at the indicated, times, and either" assayed, for the amount of 5% TCA-soluble radioactivity present, or analyzed by SDS-PAGE [U.K. Laemuli, Nature 227.: 680 (1970)] (8% polyacrylamide, 0.05 bisacrylamide, 15x15x0.15 cm gels) , with subsequent fluorograph .

2__i U£_-_^£_L£_ii£_ii£S—£_-_£_- _Zi££_i _L

Figure 1 shows construction of a ubiquitin-la£Z gene fusion. pUB2 , a pBR322-based genomic DNA clone (E. Ozkaynak, t al.. Nature 3JL__ _> ⁶⁶³ (1984) contains six repeats of the yeast ubiquitin-coding sequence (open boxes) together with the flanking regions (jagged lines) . pUB2 was modified as shown in the diagram by placing a Bam HI site six bases down- stream from the first ubiquitin repeat. This allowed the construction of an In-frame fusion (confirmed by nucleotide sequencing) between a single ubiquitin repeat and the la Z gene of the expression vector pLGSD5-ATG (called G2 in L. Guarente, Methods Enzymol. 10_1 181 (1983)) . The term "2μm" denotes a region of the pLGSD-ATG that contains the replication origin and flanking se¬ quences of the yeast plasmid called 2μm circle (See

L. Guarente , su£ra) . Figure 3B shows the amino acid sequence of the fusion protein in the vicinity of the ubiquitin-/3gal junction.

Figure 2 shows that the m v.iv£ half-life of /3gal is a function of its amino- terminal residue, (lane a) Minicells isolated from an E. £ U strain carrying ρUB23, the initial, ub-l £Z fusion (Figs. 1

35 and 3B) , were labeled with [ S ]methionine for 60 minutes at 36 C, with subsequent analysis of /3gal as described. The same result was obtained when the labeled minicell SDS extract was combined with an unlabeled yeast SDS extract before immunoprecipita- tion of /3gal. (lane b) S_. £ r£Y_.i .ia cells carrying pUB23 (Fig. 1) , which encodes ub-Met-/3gal (Fig. 3B) ,

35 were labeled with [ S] methionine for 5 minutes at

30 C, with subsequent analysis of β ~ ~ t . The same result was obtained with the lengths of the

35 [ S]methionine labeling periods from 1 to 30 minutes, and with yeast extracts produced either by mechanical disruption of cells in the presence of protease inhibitors or by boiling the cells directly in an SDS - containing buffer. (lane c) Same as lane a but with E. U cells carrying the control plasmid pLGSD5 (called Gl in L. Guarente, ££££.) which encodes β ~ ~ l . (lanes d to g) S. • c e e-v± s_ a. e cells carrying pUB23 (Fig. 1) , which encodes ub-Met-

35 /3gal (Fig. 3A) , were labeled with [ S ] methionine for 5 minutes at 30 C (lane d) followed by a chase in the presence of cycloheximide for 10, 30, and 60

minutes (lanes e to g) , extraction, immunopreclpita- tion, and analysis of jSgal. (lanes h to j ) Same as lanes d to f, but with ub-Ile-/3gal (see Fig. 3A) . (lanes k to m) Same as lanes h to , but with ub-Gln-/3gal . (lanes n to q) Same as lanes d to g, but with ub-Leu-/3gal . (lanes r to u) Same as lanes d co- g, but with ub-Arg-/3g:aL. Designations: ori; origin of the separating gel; ub , ubiquitin; ySgal, an electrophoretic band of the /3gal protein con- taining a specified amino-terminal residue; in this terminology, the Met- 3gal portion of ub-Met-/3gal Is designated as 3gal. Arrowheads denote a metabolically stable, about 90kD degradation product of /3gal which is formed apparently as the result of an ι_n viv£ endoproteolytic cleavage of a proportion of short-lived gal proteins such as Leu-3gal and Arg-3gal (lanes n to u) .

Figure 3 shows the changing amino acid residues of gal at the ubiquitin-/3gal junction. (A) The initial plas id, pUB23 (Fig.l) , which encodes ub-Met-jSgal , was mutagenized as described above to convert the original Met codon ATG at the ub- gal junction Into codons specifying 19 amino acids other than Met. (The original round of mutagenesis shown in Figure 3, produced 15 out of 19 possible sub¬ stitutions. The remaining four substitutions were produced later (see Table 1)) . The arrowhead in¬ dicates the site of the deubiquitinating _^ vivo cleavage in the nascent fusion protein that occurs

with all of the fusion proteins except ub-Pro-/3gal (see text) . All of the constructions shown encode His as the second gal residue. In addition, in some of the constructions (ub-Me -His -Gly-/3gal , ub-Met-Gln-Gly-/3gal , and ub -Met-Gin-His -Gly-βgal , the last one produced by an insertion mutation, see Table 3) , either His or Gin were following Met at the ubiquitin-/3gal junction, with indistinguishable consequences for the metabolic stabilities of the corresponding /3gal proteins. (B) The amino acid sequence (in single -letter abbreviations) of ub-Met- /3gal, the initial fusion protein (Fig. 1) , in the vicinity of the ub-/3gal junction. S ingle - letter amino acid abbreviations: A, Ala; C, Cys; D, Asp; E, Glu; F, Phe; G, Gly; H, His; I, lie; K, Lys; L, Leu; M, Met; N, Asn; P, Pro; Q, Gin; R, Arg; S, Ser; T, Thr; V, Val; W, Trp ; Y, Tyr.

Figure 4 shows that ubiquitin-/3gal is short¬ lived if not deubiquitinated . (lanes a to g) S_ . vi ₁ s_i.a cells carrying plasmids encoding ub - X-/3gal fusion proteins in which X is the residue indicated at the top of each lane, were labeled for 5 minutes at 30 C with [ S ] methionine , followed by extraction, immunoprecipitation and analysis of /3gal. Fluorographic exposures for these lanes were several times longer than those for similar patterns in Figure 2 to reveal the multiple ubiquitination of short-lived β g a.1 proteins. (lanes h, i) Fluoro- graphic overexposure of lanes n, o in Fig. 2 to

reveal the "ladder" of multiply ubiquitinated Leu-/3gal proteins in a pulse-chase experiment (zero and 10 minutes chase, respectively) , (lane j) Same as lanes a to g, but with ub-Pro-/3gal . (lane k) Same as lane j , but with ub-Gln-/3gal. (lane 1) Same as lane j . (lanes m to p) S_- £ __£ i_Li_- cells carrying a plasmid- encoding ub-Pro- 3gal were labeled for 5 minutes at 30 C with [ S ]methionine (lane m) followed by a chase in the presence of cycloheximide for 10, 30, and 60 minutes (lanes n to p) . The upper small arrow to the right of lane p denotes ub-Pro-^gal, a small proportion of which is still present after 1 hour chase. The lower small arrow Indicates an apparently deubiquitinated Pro- 3gal that slowly accumulates during chase and is meta¬ bolically stable. The dot to the left of lane m denotes an endogeneous yeast protein that is pre¬ cipitated in some experiments by the antibody used. Square brackets denote the multiply ubiquitinated /3-gal species (see Fig. 5). Other designation are as in Figure 2.

Figure 5 shows the "ladder" /3gal species containing ubiquitin. (lane a) S_. cerevisiae cells carrying a plasmid which encodes ub-Gln-/3gal , were grown and disrupted, and the extracts processed for isolation of /3gal proteins by affinity chromato¬ graphy on a column with immobilized antibody to /3gal. The /3gal proteins thus obtained were electro- phoresed in a polyacrylamide-SDS gel, transferred to

nitro -cellulose , and probed with an antibody to ubiquitin. (lane b) Same as lane a, but with ub-Pro - 3gal . (lane c) Same as b but a longer autoradiographic exposure. (lane d) 2- £ S i^ ^ cells carrying a plasmid which encodes ub-Leu-/3gal

35 were labeled with [ S]methionine for 5 minutes, with subsequent extraction, immunoprecip itation and electrophoresis of /3gal (the same sample as in

Figure 4, lane f) . Square brackets denote the multiply ubiquitinated Gln-/3gal species detected with antibody to ubiquitin. The arrow indicates the band of ub - Pro -/3gal , the initial fusion protein seen in lanes b and c. The arrowheads indicate the position of the band of deubiquitinated /3gal (de- tectable by either Coomassie staining or metabolic labeling, but not with antibody to ubiquitin) derived from the ub-Gln-/3gal fusion protein.

Figure 6 shows both prokaryotic and eukaryotic long-lived intracellular proteins have stabilizing amino acid residues at their amino-termini, whereas secreted proteins exhibit a complementary bias.

(A) 208 long-lived, directly sequenced, intra¬ cellular (noncompartmentalized) proteins with unblocked amino-termini from both prokaryotes (77 proteins) and eukaryotes (131 proteins) were distri¬ buted into three groups according to the nature of their amino - terminal residues as defined by the N-end rule (Table 1) . All of the long-lived intra¬ cellular proteins examined bear exclusively

stabilizing residues at their amino-termini. In panels B to D , analogous diagrams are presented for 243 secreted eukaryotic proteins (B) , for 37 light and heavy immuno lobulin chains (C) , and for 94 secreted eukaryotic toxins (D) . Entries in C and D are subsets of entries in B. For proteins in B to D, the amino-termini compiled correspond, whenever the assignment is possible, to the most processed form of a protein that is still located within a secreting cell. The data in A to D were manually compiled from the entire set of complete protein sequences available before 1981. The same con¬ clusions have been recently reached after a more detailed and extensive, computer-assisted tabulation of protein amino-termini using the current National Biomedical Research Foundation database. The amino-terminal residues of Asn, Cys, His, and Trp were excluded from the compilation because in vivo half-lives of the corresponding 3gal proteins are still unknown (see, however, the legend to Table 1) . Inclusion of the residues (Table 1) into a recently compilation of the s-ame type did not change the original conclusion. Although the amino-terminal Pro was also excluded from the compilation, Pro appears to be a stabilizing residue for /3gal (Table 1) , consistent with the frequent presence of Pro at the amino-termini of long-lived noncompartmentalized proteins .

Figure 7 shows the construction of ubiquitin fusions with mouse dihydrofolate reductase.

Figure 8 shows a greater range of specific dihydrofolate reductase (DHFR)-based protein con- structs which have in common the DHFR moiety, and differ exclusively in the specific amino- terminal extensions attached to DHFR. Structure I is an initial DHFR construct. Structure II contains the 40-residue amino- terminal extension derived from the amino - terminus of /3gal (see Figure 3) . Structures III-V are variants of Structure II in which either one or both of the lysine residues (denoted as K in the s ingle- letter code, see the legend to Figure 3) were replaced by the arginine residues (denoted as ^" R) . Structures V-X are variants of the Structure II with increasing deletions in the carboxyl- terminal half of the /3gal-derived extension. Structures XI-XIII are variants of Structure II with increasing deletions in the amino-terminal half of the /3gal- derived extension. S ingle - letter amino acid desig¬ nations of the amino-termini of Structure I-XIII indicate variants of these protein constructs that differ exclusively in their amino - terminal residues. These variants were obtained through the use of the ubiquitin-protein fusion approach described herein (see Figure 3) . Each of the Structures I-XIII was constructed at the DNA level using s ite -directed mutagenesis and other methods of recombinant DNA technology currently standard in the field. T ese

DNA constructs were introduced into the yeast S.. £ £ Zi i£ > ^anc ^ ^tne half-lives (left column of Figure 8) of the corresponding proteins I-XIII were directly determined using the methods described above for /3gal and a monospeclfic antibody to DHFR.

Results and Discussion

a£id_i.n_v^ £_d ub^ u^ ^n t^£n_£ _a_na cent_ub -_ ££iϋis_ _l ≤ i__££ϊ££_£_i£_- is

Branched ubiquitin conjugates in which the carboxyl-terminal glycine of ubiquitin moieties is joined via an isopeptide bond to the α-amino groups of internal lysine residues in proteins apparently comprise the bulk of ubiquitin conjugates in eu¬ karyotic cells. Joining of ubiquitin to the amino- terminal α-amino groups of target proteins, to yield linear ubiquitin conjugates, may also be chemically feasible. See A. Hershko, et al.. , PNA-i 2£__ _-i ^{: 7021} (1984) . Whether or not linear ub-iquitin-protein fusions are actually synthesized i.n vry_£ through posttranslational enzymatic conjugation of ubiquitin to protein amino-termini, such proteins can also be produced by constructing appropriate chi eric genes and expressing them in vivo. Construction of one such gene, which encodes yeast ubiquitin linked to 3 al of Esche.ri.chla c£li^, is shown in Figure 1. When this gene is expressed in E. £ U, the resulting /3gal-containing protein has an apparent

molecular mass which is approximately 6 kD greater that that of the control /3gal, a value consistent with the presence of ubiquitin in the protein encoded by the chimeric gene. In contrast, when the same gene is expressed in yeast, the corresponding β ga.1 protein is elec rophoretically indis inguish¬ able- from the control 3gal. This result is in-

35 dependent of the length of the [ S]methionine labeling period (between 1 and 30 minutes) . Furthermore, determination of the amino-terminal residue in the putative Met-3gal (half-life, - - _/ - 20 hours) by Edman degradation of the in vivo - labeled, gel-purified /3gal (Figure 2, lane d) directly confirmed the presence of the expected Met residue (Figure 3A and Table 1) at its amino - terminus .

Independent evidence that ubiquitin cleavage of the fusion protein occurs immediately after the last Gly residue of ubiquitin is presented below. We con¬ clude that in yeast, ubiquitin is efficiently cleaved off the nascent ubiqui in-/3gal fusion protein, yielding a deubiquitinated /3gal . The absence of the deubiquitination reaction in E. ££l_i is consistent with other lines of evidence in¬ dicating that prokaryotes lack both the eukaryotic ubiquitin and ubiquitin- specific enzymes. At the same time, the possibility remains that a functional counterpart of ubiquitin exists in bacteria but is different in its amino acid sequence from that of eukaryotic ubiquitin. The present invention clearly

applies not only to the extremely close amino acid homologs of ubiquitin such as those found in eukaryotes but also to functional homologs of ubiquitin such as those that may exist in bacteria. The ubiquitin-3gal junction encoded by the chimerLc gene,. Gly-Met (Figures 1 and 3B) , is Identical to- the junctions- between; adj^acent repeats in the polyubiquitin precursor protein, which is efficiently processed into mature ubiquitin. Thus it is likely that the same protease, as yet un- characterized biochemically, is responsible both for the conversion of polyubiquitin into mature ubi¬ quitin and for the deubiquitination of the nascent ublquitin-/3gal protein. If so, one potential way to inhibit the in vivo deubiquitination of the ubi- quitin-/3gal (and thereby to allow analysis of metabolic consequences of a stable ubiquitin attach¬ ment to β ga.1 ) would be to convert the Met residue of 3gal at the ubiquitin- ?gal junction (Figure 3B) into other amino acid residues (Figure 3A) . The unex¬ pected results of such an approach are described below.

ϋl _ S_ZiZ£_l_-ϋ-ii-- — ———_-_-_-___ill— __ ^■ —i£££t.i£n_£f_i.t amino-terminal residue^ The ATG codon which speci- fies the original Met residue of gal at the ubi¬ quitin junction (Figure 3B) was converted by site- directed mutagenesis into codons specifying 19 other amino acids (See Figure 3A and Table 1) . The≤e

constructions differ exclusively in the first codon of /3gal at the ubiquitin-3gal junction (Figure 3A) . After each of the 16 plasmids thus designed was introduced into yeast, analysis of the corresponding /3gal proteins pulse- labeled in vivo led to the following results (Figures 2, 4, and Table 1) :

1) With one e-x.cept.ion (see below) , the effi¬ cient deubiquitination of the nascent ubiquitin- gal occurs irrespective of the nature of the amino acid residue of /3gal at the ubiquitin- 3gal junction. Thus, the apparently ubiquitin- specific protease that cleaves the original ubiquitin-/3gal protein at the Gly-Met junction is generally insensitive to the nature of the first residue of /3gal at the junction (Figure 3A and Table 1). This result, in effect, makes it possible to expose different amino acid residues at the amino-termini of the otherwise identical /3gal proteins produced in vivo.

2) The in vivo half-lives of the /3gal proteins thus designed vary from more than 20 hours to less than 3 minutes, depending on the nature of the amino acid residue exposed at the amino - terminus of /3gal (Figures 2, 4, and Table 1) . Specifically, de- ubiquitinated /3gal proteins with either Met, Ser, Ala, Thr, Val , Cys or Gly at the amino - terminus have relatively long in vivo half-lives of 20 hours or more (figure 2, lanes d to g, and Table 1) , similar to the half-life of a control /3gal whose gene had not been fused to that of ubiquitin. In striking

contrast, the 3gal proteins with either Arg, Lys, Phe, Leu, Asp or Trp at the amino-terminus have very short half-lives , between approximately 2 minutes for Arg-/3gal and approximately 3 minutes for Lys - /3gal, Phe- 3gal, Leu-/3gal, Asp-/3gal, Asn-/3gal and

Trp- gal (Figure 2, lanes n to u, and Table 1) . The harlf-Ilfe of /3gaL proteins with amino-terminal residues of either Gin, His or Tyr is approximately 10 minutes (Figure 2, lanes k to m, and Table 1) , while an amino-terminal lie or Glu confers on β ga.1 a half-life of approximately 30 minutes (Figure 2, lanes h to , and Table 1) . Both pulse-chase and continuous labeling techniques were used in these experiments and yielded similar results. The set of individual amino acids can be ordered with respect to the half-lives that they confer on /3gal when exposed at its amino- terminus . The resulting rule (Table 1) is referred to as the "N-end rule" .

Table 1: The N-end rule

In vivo

Ra< dius of deubiquitination

Residue X in gy ration of nascent ub-X- _/ 3gaL of X-(A) ub-X-^gal t of X-£gal

Met 1. .80 + Ser 1. .08 + Ala 0. .77 + Thr 1. ,24 + >20 hou s Val 1. .29 + Gly 0 Cys

He 1.56 + Glu 1.77 + -30 minutes

Tyr 2.13 + Gin 1.75 + —10 minutes His

Phe 1.90 + Leu 1.54 + - 3 minutes Trp + Asp 1.43 + Asn + Lys 2.08 +

Arg 2.38 + - 2 minutes

Pro 1.25 -* - 7 minutes

*The rate of in vivo deubiquitination of ub-Pro-/3gal is extremely low. The t.. _/r> shown is that of the inltrla-l ub-Pro-/3gal fusion protein (see Fig. 4, Lanes j tro p ) .

Le£e_nd_t.£_Tabl_ _l_

The N-end rule. In vivo half-lives of /3gal proteins in the yeast S_. ££ievisiae were determined either by the pulse-chase technique (for short-lived /3gal's; see below) or by measuring the enzymatic activity of y3gal in crude extracts. For the measurements of /3gal activity, cells growing in a galactose -containing medium were transferred to an otherwise identical medium lacking g-alacto-se and containing 10 percent glucose. After further growth for at least 5 hours at 30 C, the ratio of 3gal activities per cell before and after shift to glucose was determined for each of the /3gal pro¬ teins. [GAL promoter- driven expression of the fusion genes (Figs. 1 and 3) is repressed in glucose medium] . For shorter- lived /3gal proteins (t _η „ 1 hour) , the pulse-chase technique was used as Well (Figs. 2 and 4) . Electrophoretic bands of /3gal proteins labeled with [ S]methionine in pulse- chase experiments were cut out from scintillant- impregnated. dried gels similar to those of Figs. 2 and 4, and S in the bands was determined. The in vivo decay of short-lived /3gal proteins deviated from first-order kinetics in that the rate of degradation was lower when measured at later (1 hour) time points of the chase, the lower rate reflecting either a time -dependent toxic effect of cycloheximide or intrinsic characteristics of the in vivo degradation process. [Arrest of translation is required for an efficient short-term chase in 2- £ S£Xi i because of. the amino acid pool equili¬ bration problems related to the presence of vacuoles in this organism] . The half-life values listed below were determined for the first 10 minutes of chase. Several lines of evidence (see description of Figs. 4 and 6) suggest that Pro is a stabilizing residue. The listed radii of gyration of amino acids are from. M. Levitt, J. M£ 1 . Bi_£l_. 10_4 : 59 (1976) .

peubiquitination of Ub-X^ffgal f sion_ρr£tein in ATP-depleted reticulocyte extract. Each of twenty 35S-labelled Ub-X-j3gal proteins prepared in E. coli as described above, was added to an extract prepared from ATP-depleted rabbit reticuloeytes (Etlinger t_al._ _: _, Proc_. N.__.___L_ _.£ -___ _.

Sci. ITSA 74: 54 (1977) ^* Hershko etr al. , Proc. Natl. Acad. Sci. USA 77: 1783 (1980); Hershko et al . , J^_ Biol. Chem. 25.8.: 8206 (1982)), and the fates of the added proteins were followed by SDS -polyacrylamide gel electrophoresis (SDS-PAGE). As had been observed in_vivo with the same ubiquitin fusions in yeast, an apparently ubiquitin-specific protease in reticulocyte extract deubiquitinated the added Ub-X-y3gal fusion proteins to yield the corresponding X-/3gal test proteins. The deubiquitination of 19 out of the 20 Ub-X-/3gal proteins in the ATP-depleted extract was more than 90% complete in 5 minutes at 37° (Table 2) . The single exception, both in yeast and in reticuloeytes, Is Ub-Pro-/3gal, which was deubiquitinated approximately 20 times more slowly than were the other Ub-X-3gal proteins.

Amino acid sequencing (by Edman degradation) of deubiquitinated -/3gal proteins reisolated from either the reticulocyte extract or yeast cells showed that, in every case tested, the proteolytic cleavage occured precisely at the Ub-/3gal junction (Table 2) . Although sequencing revealed that the amino termini of some X- 3gal proteins underwent specific

-53

modificiations (see Table 2) , in no case did these modifications involve proteolytic cleavages beyond the amino - terminal residue X.

All of the deubiquitinated X-/3gal proteins were metabolically stable in the ATP-depleted reticulocyte extract as judged from SDS-PAGE analysis and rom' the. ne-gligible- production of acid-soluble radioactivity in the extract. Thus, preincubation of Ub-X- 3gal fusion proteins in the ATP-depleted reticulocyte extract makes it possible to generate twenty X-/3gal test proteins which differ exclusively at the amino - terminal residue X.

Half-life of a 3gal protein in ATP!-supplemented

£.£^ ££i Σ^£_ ≥^Σ ^a £^_i _ ^a __-__.S ^_Li£ _£^__£_- _-i--£ ---L amino-terminal _. t£sidue.

While all of the twenty X-/3gal proteins were metabo.lically stable in the ATP-depleted reticulocyte extract., most of them became short-lived upon addition of ATP to the extract (Table 2). We refer to an amino-terminal residue as stabilizing if the corresponding X-/3gal is relatively long-lived in the ATP-supplemented extract (less than 10% degradation in 2 hours at 37°C) , and as destabilizing if the degradation of the corresponding X-/3gal in the extract exceeds 15% under the same conditions (Table 2) . The time courses of degradation for several X-/3gal proteins showed reproducible initial lags. However, semilogarithmic plots of the time courses showed that, after the initial lags, the degradation of X-/3gal in the ATP-supplemented reticulocyte extract obeyed first-order kinetics for at least the first two hours, making It possible to compare the degradation of different X-/3gal proteins by comparing their -half-lives in the extract (Table 2) . The range of /3gal half-lives in the reticulocyte extract encompasses more than two orders of magnitude, from approximately 50 minutes for Gln-βgal to approximately 100 hours for Val-βgal (Table 2) . The half-lives of X-jSgal proteins-

bearing stabilizing amino-terminal residues range from approximately 20 hours for lle-/3gal to approximately 100 hours for Val-/3gal. Half-lives of the metabolically unstable X- 3gal proteins in the reticulocye extract were comparable to the half-lives of other proteolytic substrates (iodinated serum albumin, lysozyme, and cytochrome c) in the same extract. These latter test proteins have been used in earlier studies of ubiquitin- dependent protein degradation in reticulocye extract (Finley e_£_a 1^__, T e_nd _Bi _^ £ h.£m_ _: _ 2£i^ i . ^{: 343} (1985) ; Etlinger et_a1^, l ._l__.___ ^______

A£ d^_£ci.^_USA J I : 54 (1977)) . Recently, at least some of these proteins have been shown to be targeted for degradation via their destabilizing amino - terminal residues [Reiss et_al^., J_. \_%_o_ __

Chem _.3. ^: 2693 (1988)] as defined by the N-end rule .

5

Table 2 rhe N-end rule in yeast and in mammalian reticuloeytes.

_^Ξ! S£-li £!SiS :i_i£££U£S_ i—£S_£ΞiS£_ £i≤!:—i_L £ £ ial__f£r_i.t. _£ff £t_£n_^g.al__ha].f_^l ₁ if _as _^ _t. S.t_£d. iS_Σ £ £

Site -directed mutagenesis was employed to insert a codon specifying a "stabilizing" amino acid (in this experiment, the Met residue) before the first codon of β ga.1 at the ubiquLtin-/3gal junction (Table 3) . Insertion of a stabilizing residue (Met) before either another stabilizing residue (Thr) or a variety of destabilizing residues (Gin, Lys, and Arg) at the ubiquitin-/3gal junction invariably results in a long-lived deubiquitinated /3gal (Table 3) . Furthermore, in contrast to ubiquitin- Pro -/3gal which is not only short-lived but also resistant to deubiquitination (Figure 4, lanes j to p, and Table 1) , ubiquitin-Met- Pro - 3gal is efficiently deubi¬ quitinated in vivo to yield a ongi ived Met-Pro- /3gal (Table 3) . These results show that both the identity of amino acid residue and its amino- terminal location (presumably the presence of a free α-amino group) are essential for its effect on /3gal half-life. In addition, these results (Table 3) further support the expectation that ubiquitin- specific cleavage of the fusion protein occurs immediately after the last Gly residue of ubiquitin (Figure 3A) .

Amino- terminal location of an amino acid is essen¬ tial for its effect on βgal half-life. The inser¬ tion mutants were obtained essentially as described for the initial set of mutants except that a 32- residue oligonucleotide, 5 ' -CCCGGGATCCGTGC (G/C/ T/) (G/T)CATACCACCTCTTAG was used, containing 14 bases on the 5 ' side and 15 bases on the 3' side of the ambiguous codon inserted behind the Met codon. Bases in parentheses denote ambiguities at the positions 16 and 17 in the sequence. Half-lives of the corresponding βgal proteins were determined as described In the legend to Table 1.

A_l^£ng^.l.iv d_ .l avag _£r du£t _; _ f_^£ l_i _formed

_i£Ξ S_i__^ £ Σ_££_ ^££i-liiZ ^—_^S-i _-E_i£_i S_L_.

The electrophoretic patterns of short-lived

(but not of long-lived) βgal proteins invariably contain a specific, about 90 kD cleavage product of βgal (Figure 2, lanes n to u) which, unlike the parental βgal species, accumulates, during the postlabeling (chase) period (Figure 4, lanes m-p) .

The 90 KD βgal fragment constitutes a relatively small proportion of the initial amount of the pulse- labeled βgal. Nonetheless, its existence implies that an in vivo endoproteolytic cleavage can rescue a protein fragment from the metabolic fate of its short-lived parental protein. It remains to be seen whether the resulting possibility of multiple half-lives within a single protein species is exploited in the design of naturally short-lived proteins .

__[£iS£i _:_?£ i_i _ ££r _:iiY d _^ ____ll_____i____

Ubiquitin-Pro -βgal , the only ubiquitin- gal fusion that is not deubiquitinated in vivo (Fig. 4, lanes j to p) , has a half-life of approximately 7 minutes (Table 1) which is less than 1 percent of the half-life of metabolically stable βgal proteins (Table 1) . One interpre ation of this result is that a metabolically stable ubiquitin attachment to protein amino-termini is sufficient to signal

degradation of acceptor proteins. This inter¬ pretation is consistent with earlier biochemical and genetic evidence that ubiqultination of short-lived proteins In a mammalian cell is essential for their degradation. At the same time, all ubiquitin-βgal fusion proteins other than ubiquitin-Pro-βgal are rapidly deubiquitinated In vivo (Table 1) . Thus, the posttranslational amino-termal ubiquitination of proteins may not be Involved in an initial recogni- tion or commitment step that designates proteins for degradation in vivo. Whether posttranslational amino-terminal ubiqultination (if it actually occurs in vivo) is essential for later stages of the degradation pathway remains to be determined. Earlier in vitro experiments indicated that prefer¬ ential chemical modification of amino-termini of proteolytic substrates inhibits their degradation in an in vitro ubiquitin-dependent proteolytic system. Based on these data, it was proposed that amino-ter- minal ubiqultination of proteins is essential for their degradation. An alternative interpretation of the same results is that chemical blocking of proteins' amino-termini prevents the recognition of their amino-terminal residues by the "N-end rule" pathway whose i£iti.al stages are not necessarily ubiquitin-dependent.

_L__^£££^ iZ£^_^S i_£_i££ iS_L_^Σ _IS£i£i_£iY_£ i££i_iiS_i;_: -!: iS_Xi∑£•

Overexposures of the pulse-chase fluorograms

(Fig. 2) reveal that the major band of a deubi- quitinated, short-lived βgal protein coexists with a

"ladder" of larger molecular mass, βgal -containing bands irregularly spaced at 4 to 7 kD intervals.

(Fig. 4, lanes c to g) . No such larger species appear when the fluorograms of long-lived βgal proteins are similarly overexposed (Fig. 4, lanes a and b) . Immunological analysis with both antibodies to βgal and antibodies to ubiquitin demonstrates that the "ladder" βgal species contain ubiquitin

(Fig. 5) .

A_m£del__f r__£h _ l_ i_y _d g£ada£^£n_£a hway .

With the exception of natural or engineered ubiquitin fusion proteins (Fig. 1 and Table 1) , nascent proteins apparently lack ubiquitin moieties. The in vivo amino - terminal processing of nascent noncompartmentalized proteins generates their mature amino-termini via the action of amino - terminal peptidases whose substrate specifici ies have been partially characterized. (See Tsunasawa, S. e t a 1 . 1• U£I• Chem . 2_ 0 5382 (1985) ; Boissel, J.P. et al. PNAS USA 8_2_, 8448 (1985)) . We suggest that the > amino-termini thus generated are recognized by an "N- end- reading" enzyme. One specific model is that a commitment to degrade a protein molecule is- made

as a result of the recognition of its amino-terminal residue by a stochastically operating enzyme whose probability of "clamping" at the target's amino- terminus is determined by the N-end rule (Table 1) . Once the commitment is made, it is followed by a highly processive ubiqultination of the target protein which in the case of βgal is- conjugated to more than 15 ubiquitin moieties per molecule of βgal (Fig. 4 _r lanes c to g, and Fig. 5) . The multiply ubiquitinated target protein is then degraded by a "down stream" enzyme (1) for which the ubiquitin moieties of the target serve as either recognition signals or denaturation (unfolding) devices, or both. The ubiquitin-containing "ladder" βgal species (Fig. 4, lanes c to 1, and Fig. 5) consist of apparently branched ubiquitin moeities joined to the α-amino groups of internal lysine residues in βgal. Surprisingly, the "ladder" βgal species derived from ubiquitin-Pro-βgal are electrophoretically indistin¬ guishable from the analogous species of βgal whose amino-terminal ubiquitin is cleaved off the nascent fusion protein (Fig. 4, lanes j to 1, and Fig. 5) . If the electrophoretically indistinguishable ublqui- tinated βgal species are indeed structurally homo¬ logous, these results would be compatible with two alternative models in which, immediately after the first ubiquitins are branch-conjugated to βgal, either a branch-ubiquitinated ubiquitin-Pro-βgal

undergoes amino- terminal deubiquitination or, alternatively, an analogous βgal species lacking the amino- terminal ubiquitin moiety reacquires it. Experimental resolution of this ambiguity may establish whether the posttranslational amino- terminal ubiqultination of proteins (if it occurs in vivo) plays a role in selective protein turnover-.

Although both prokaryotic and eukaryotic proteins appear to follow the N-end rule (see below) , bacteria apparently lack the ubiquitin system. Thus it is possible that th.e hypothetical N-end- recognizing protein is more strongly conserved between prokaryotes and eukaryotes than is the rest of the selective degradation pathway. Interest- ingly, the properties of a mammalian protein E3 whose presence is required for ubiqultination of proteolytic substrates by ubiquitin-conjugating enzymes in vitro are consistent with it being a component of the N-end-recognizing protein.

Th _N^ nd_ ul_ _a d_£h _kn£wn_a^ iS££ £ UHi Σ_EΞ££ 5 •

The unblocked amino - terminal residues in metabolically stable, noncompartmentalized proteins from both prokaryotes and eukaryotes are exclusively

(Fig. 6A) of the stabilizing class (Met, Ser, Ala,

Gly, Thr, Val) , that is, the class that confers long in vivo half-lives on βgal (Table 1) . The one short-lived intracellular protein for which the

mature amino-terminus is known is the ell protein of phage lambda, the central component of a trigger that determines whether lambda grows lytically or lysogenizes an infected cell. (Y.S. Ho, D. Wulff, M. Rosenberg, in Regulation of Gene Expression, I. Booth and C. Higgins, Eds. (Cambridge Univ. Press, London, 1986) , p. 79; F. Banuett, M.A. Hoyt, L.

McFarlane, H. Echols , I. Herskowitz , J.^_M£l_ _: Binl.^

18_7. _> 2.1- (1986) ; M.A. Hoyt, D.M. Knight, A. Das, H.I. Miller, H. Echols, Cell 3.1, 565 (1982) ; K. Nasmyth, Nature (london) 3_20_, 670 (1983)). The half-life of ell in lambda-infected E. ££.1.1 is less than 3 minutes. Strikingly, the mature amino- terminus of ell starts with Arg (Ho, Y.W. e_t al. , J. _-Ϊ£i- Chem- 25.7.. 9128 (1982)) , the most destablizing residue in the N-end rule (Table 1) .

While the destabilizing amino acids can be either hydrophobic, uncharged hydrophilic or charg¬ ed, they share the property of having larger radii of gyration than any of the stabilizing amino acids except Met (Table 1) .

Amino-terminal residues_In_ compartmentalized pro- t in _are_lar£ely_£f_th _d£ ab^].i.z_i.n^_£l.a. .

Figure 6 illustrates a striking difference between the choice of amino-terminal residues in long-lived, noncompartmentalized intracellular proteins (A) and in compartmentalized proteins, such as secreted proteins (B) , many of which are also

long-lived in their respective extracellular com¬ partments. One implication of this finding is that a single intracellular degradation pathway operating according to the N-end rule could be responsible both for the diversity of in vivo half-lives of intracellular proteins and for the selective de¬ struction of compartmentalized proteins that are aberrantly introduced into the intracellular space. Some miscompartmentalized proteins may be more harmful to the cell than others. It is therefore of interest that secreted eukaryotic toxins contain strongly destabilizing residues (Arg, Lys, Leu, Phe, Asp) at their amino-termini more often than the general population of secreted proteins (Fig. 6, panels B to D) .

The above consideration also suggest that, if the topological outside of a cell, such as lumens of the endoplasmic reticulum and golgi, and the extra¬ cellular space, were to have degradation pathways analogous to the N-end rule pathway, they could be based on "inverted" versions of the N-end rule in which the amino- terminal residues that are de¬ stabilizing inside the cell are now the stabilizing ones and vice versa. Thus, the methods of the present invention should also be useful for manipu¬ lating the metabolic stability and other properties of compartmentalized proteins, including secreted ones .

Possible role of the N-end rule pathway in the turn£ r_£f_l£ng^l^v d_ES£t in .

Long-lived intracellular proteins with destabilizing (Table 1) penultimate residues generally retain their initial amino-terminal methionine residue.

The amino-terminal residues in long-lived intracel¬ lular proteins that do undergo amino-terminal processing are invariably of the stabilizing class (Table 1) . An interesting possibility that would involve the N-end rule pathway in the turnover of long-lived proteins is that the rate-limiting step in the in vivo degradation of long-lived proteins may be a sl_£w aminopeptidase cleavage that exposes a destabilizing residue, followed by ragi.d degradation via the N-end rule pathway. Note that fine-tuning of the rate of degradation may in this case be a function of the rate of aminopeptidase cleavage exposing a destabilizing residue rather than a function of the residue's destabilizing capacity according to the N-end rule.

The_N^end rule__and_selectlye_degradation _of _shqrt- llvedL and damaged proteins .

The recognition of polypeptide chain folding patterns or of local chemical features that target an otherwise long-lived but damaged protein for selective degradation in vivo is unlikely to be mediated directly by the N-end rule pathway. Instead, we suggest that specific proteases

(analogous in function to nucleases that recognize specific lesions in DNA) cleave a targeted protein so as to expose a destabilizing residue at the amino-terminus of one of the two products of a cut. One testable prediction of this model is that the initial cleavage -products of the degradation pathway should bear destabilizing residues at their N-ter- mini . The preferential exposure of destabilizing residues at the amino-termini of products of the initial protein cleavages may be due either to intrinsic specificities of the proteases involved or simply to the fact that a majority of the amino acids belong to the destabilizing class (Table 1) . Furthermore, initial cleavages of a protein would be expected to destabilize aspects of its original conformation, thus increasing the probability of further internal cuts. Whether the initial cleavage products of a protein would be degraded exclusively via the N-end rule pathway or would have to be processed further by additional internal cleavages should depend on several factors, such as the exposure of des abilizing residues at the amino- termini of initial cleavage products, and the relative rates of introduction of internal cuts. In this model, the N-end rule pathway should be es¬ sential for degradation of most of the metabolically unstable proteins, from chemically damaged, pre¬ maturely terminated, improperly folded and miscom- partmentalized ones to those that cannot assemble

into native multisubunit aggregates , and finally to otherwise normal proteins that are short-lived in vivo. Thus, the metabolic instability of a protein may be mediated not only by the exposure of a destabilizing residue at its amino-terminus , but also by local con ormational and chemical features of its- polypeptide chain that result in proteolytic cleavages exposing destabilizing residues at the amino-termini of cleavage products. For any given protein, a variety of factors in addition to the N-end rule may combine to modulate its half-life in vivo. Among such factors may be the flexibility and accessibility of the protein's amino-terminus (Thornton, J.M. and Sibanda, B.L. , J. Mol . Bi.£. 167 443 (1983)), the presence of chemical¬ ly blocking amino-terminal groups such as the acetyl group, the distribution of ubiquitinatable lysine residues near the amino-terminus , and other variabl¬ es, such as the structure of the carboxy-terminus . Since amino-terminal regions of multisubunit pro¬ teins are commonly involved in the interfaces between subunits (Thornton, J.M. and Sibanda, B.L. , __• ^£ • £Ϊ£- -\_-\-l ^~^~ (1983)) , quarternary structure of proteins is yet another parameter that is ex- pected to modulate the impact of the N-end rule pathway on protein half-lives in vivo. Finally as suggested above, the N-end rule pathway may also be essential for the degradation of proteins whose Initial recognition as targets for degradation is

independent of the structures at their amino- termini .

∑£S££ £S _ iS fi£ S£ _£i_E£ ^USS i^;£i£S i_5;d^ l.

£ £S_£^_ Si5£— £i^ _££_ SiS£__.£££SiSi_£i_2S£_i iS_L ^• It has been known for many years that in both bacteria and eukaryotes there exists an unusual class of enzymes, aminoacyl- transfer RNA-protein transferases , which catalyze posttranslational conjugation of specific amino acids to the mature amino-termini of acceptor proteins in vitro (R.L.

Soffer, in Trans_f£r_RNA _Bi _^ £].£^i ₁ £a.l._A ££££ , D. Soil, J.N. Abelson, P.R. Schimmel, Eds. (Cold Spring Harbor Laboratory, Cold Spring Harbor, NY 1980) , p493;C. Deutch, Methods_Enz^mo1. 106. _' 198(1984) : A. Kaji, H. Kaji, G.D. Novelli, ^._Bi£l^_Chem . 2.40, 1185 (1965)) . The posttranslational addition of amino acids to proteins in vivo dramatically ac¬ celerates in a stressed or regenerating tissue, for example, after physical injury to axons of nerve cells (S. Shyne-Athwal, R.V. Riccio, G. Chakraborty, N.A. Ingolia, Science 23.1, 603 (1986) ; N.A. Ingolia £ al.. I^_Neur£ i. 3, 2463 (1983)) . The N-end rule provides an explanation for this phenomenon. We suggest that selective changes in metabolic sta- bility of otherwise undamaged, longlived proteins that may be required by a changed physiological state of the cell are brought about by posttransla- tional addition of destabilizing amino acids to the

amino-termini of target proteins in vivo. Striking¬ ly, the known reactions of posttranslational addi¬ tion of amino acids to proteins (R.L. Soffer, in l_i_!:S __£.__£__-_I_-±_- £i£S £ i_AsE££ts.> D. Soil, J.N. Abelson, P.R. Schimmel, Eds. (Cold Spring Harbor Laboratory, Cold Spring Harbor, NY 1980) , p493;C. Deutch, Methods ΞnzrymoL. 106., 198(1984) : A. Ka i, H. Kaji, G.D. Novell!, J. Biol. Chem. 240, 1185 (1965) ; S. Shyne-Athwal, R.V. Riccio, G. Chakraborty, N.A. Ingolia, Science 23.1, 603 (1986) ; N.A. Ingolia et al . , J_. Neur£scL 3_ _■ 2463 (1983)) involve largely those amino acids (Arg, Lys, Leu, Phe, and Tyr) that are destabilizing according to the N-end rule (Table 1) . Physiological states in which addition of destabilizing amino acids to proteins could be expected to occur include entry to and exit from the cell cycle, responses to chemical or physical stress, and specific differentiation events, such as erythroid differentiation and spermatogenesis , in which a proportion of preexisting, otherwise long- lived intracellular proteins is selectively de¬ graded.

The in vitro degradation of some proteolytic substrates in a ubiquitin-dependent system from mammalian reticuloeytes has recently been shown to depend on the presence of certain aminoacyl- tRNAs (Ferber, S. and Ciechanover, A. , J. 15i.£1.. C_h m. 261 3128 (1986)) . We suggest that this phenomenon also reflects a requirement for posttranslational ^*

addition of specific destabilizing amino acids to the amino-termini of proteolytic substrates. The initial proteolytic substrates in question have amino - terminal residues of Asp or Glu, both of which are destabilizing according to the N-end rule

(Table 1) . This raises an interesting and testable possibility that certain amino - terminal residues in proteins may not be directly destabilizing as such but only through their ability to be conjugated to other destabilizing residues.

H i££i£iS_i£ i££ __-i£_l_--i_lΣ---i£_-£i ££_£ _-£££-- £

In a set of Ub fusions constructed with mouse dihydrofolate reductase (DHFR) , a monomeric ~20-kd protein whose structure is known at atomic resolu- tion, the mature amino-terminus of the "natural"

DHFR is extended by 7 residues due to a construction route taken (Figure 7) . After cleavage of Ub from the nascent ubiquitin-DHFR fusion proteins m yi.v£, the deubiquitinated DHFR proteins differ exclusively at their amino- terminal residues. These construc¬ tions are analogous to the set of βgal test proteins (Figure 3) . As expected, the DHFR proteins bearing those amino-terminal residues that are stabilizing according to the N-end rule (Table 1) are long-lived in yeast (Figure 7 and data not shown) . Although the presence of a residue that is destabilizing according to the N-end rule at the amino-termini of an otherwise identical TλHFR protein does destabilize

it in vi_£, the extent of destabilization is small (Figure 7A) in comparison to the results with βgal of analogous design (Table 1) . To address the mechanistic significance of these findings, a 40-residue amino-terminal region of βgal was positioned upstream of the original DHFR's a ino- terminus (Figure 7) . The DHFR proteins ^" bearing a destabilizing residue followed by the βgal-derived extension are approximately as short-lived in yi.v£ as their unstable βgal counterparts, in striking contrast to the otherwise identical DHFR proteins that lack the βgal- specific amino-terminal extension (Figure 7B and data not shown; cf. Figure 7A) . Furthermore, the extension-bearing DHFR proteins that have £abfllz .ng residues at their amino- termini are long-lived in yiv£ (Figure 7B) . This latter result proves that the βgal-specific ex¬ tension £_ U£h, in the absence of a destabilizing amino-terminal residue, does not confer a short half-life on DHFR. These findings also indicate that the reason for the striking difference between half-lives of the DHFR's that either lack or contain the βgal-specific extension (and bear identical, destabilizing amino- terminal residues) Is due to differences in ami.n£-1ermi. a1. targeting elements in these proteins and not to differences between the overall structures of DHFR and βgal.

When DHFR is fitted with a 26-residue, βgal- derived amino- terminal extension instead of the original 40-residue extension, the dependence of the iS Y.iY.£ half-life of the resulting protein on the nature of its amino- terminal residue is intermediate between that of the original DHFR and that of the DHFR bearing a 36-residue βgal-derived extension (Figure 7C; cf. Figure 7B) . Thus, the sequences required for the effect of the original βgal- specific extension are not confined to a short stretch within the extension but are distributed over the length of the extension. These insights indicate that the complete amino - terminal degrada¬ tion signal contains a distinct determinant addi^. tion l to the determinant represented by the amino- terminal amino acid residue. To address the nature of the second determinant in greater detail, a number of otherwise identical DHFR-based proteins bearing different variants of the βgal-derived extension and either a stabilizing or a de¬ stabilizing amino- terminal residue were expressed in the yeast . cerevisiae and their half-lives determined (Figure 8) . The first conclusion from the data shown in Figure 8 is that the two lysine (K) residues present in the βgal extension, although by themselves they do not render the protein metabolically unstable, are absolutely essential for conferring sensitivity to the N-end rule upon the test protein. Indeed, while the conversion of just

one of the two lysine residues into a similarly charged arginine (R) residue still results in a protein whose half life is a strong function of its amino-terminal residue (structures II-IV in Figure 8) , the conversion of both lysine residues Into arginine residues results in a long-lived test protein whose half-life is essentially insensisltlve to the nature of its amino-terminal residue (Structure V in Figure 8). At the same time, lysine residues are the only amino acid residues in pro¬ teins that can be posttranslatlonally joined to the carboxy-terminus of ubiquitin, with the formation of branched ubiquitln-protein conjugates. Strikingly, our direct determination of the positions of ubiquitin moieties in multiply ubiquitinated, short-lived proteins of the type shown in Figure 8 has shown that all of the multiple ubiquitin moieties attached to a given molecule of the test protein reside in branched Ub-Ub structures attached to one of the two lysine residue that have been identified above by genetic methods as being es¬ sential components of the complete amino-terminal degradation signal. What then distinguishes the above lysine residues (Structures II-IV) from the numerous other lysine residues in the rest of the

DHFR test protein? A clue to the unique role of the lysine residues as the second determinant of the amino- terminal degradation signal is provided by the

fact that, due to the design of the original ex¬ pression vector used in our work (see Figure 1) , our βgal test proteins bear a 45-residue amino-terminal extension derived from an iE £rn.al. sequence of the l_a.£ repressor encoded by the lacl gene. Thus the "βgal-derived" amino- terminal extension discussed above (Figures 7 and 8) is derived not from the amino- terminal sequence of the wild-type βgal but from an unrelated sequence present at the amino- termini of our βgal test proteins. It is likely that the l_a£ repressor- specific extension at the amino-termini of these βgals is more disordered (segmentally mobile) than the amino - terminal region of the wild-type βgal. If so, this extension, while not metabolically destabilizing βgal as such, could allow the observed extreme dependence of the βgal's half-life on the nature of its amino - terminal residue (Table 1) , and thereby, in hindsight, could have greatly facilitated the discovery of the N-end rule. The disordered (segmentally mobile) state of the βgal extension provides an explanation for the unique nature of lysine residues within the ex¬ tension versus the lysine residues in the spatially ordered DHFR portion of the test protein (Figure 8) . Thus, the simplest interpre ation of this and related evidence is that the complete amino- terminal degradation signal comprises not one but two dis¬ tinct determinants, each of which is necessary but by itself not sufficient to render the protein

metabolically unstable. One determinant, described in the first part of this application, is the protein's amino-terminal residue. The second determinant, described Immediately above, is a specific internal lysine residue. As Indicated by the data of Figure 8 and the considerations above, the ability of this critical lysine residue to serve as the second determinant is to a significant degree independent of the unique amino acid sequences sur- rounding the lysine residue. Instead, an essential feature of the critical lysine residue includes Its spatial proximity to the protein's amino-terminus .

E ui.v _lent:s

Those skilled in the art will recognize, or be able to ascertain using no more than routine experi¬ mentation, many equivalents to the specific embodi¬ ments of the Invention described herein. Such equivalents are intended to be encompassed by the following claims.