Novel Propeptide Function in 20 S Proteasome Assembly Influences β Subunit Composition*

The assembly of eukaryotic 20 S proteasomes involves the formation of half-proteasomes where precursor β-type subunits gather in position on an α-subunit ring, followed by the association of two half-proteasomes and β-subunit processing. In vertebrates three additional β-subunits (β1i/LMP2, β2i/MECL1, and β5i/LMP7) can be synthesized and substituted for constitutive homologues (β1/delta, β2/Z, and β5/X) to yield immunoproteasomes, which are important for generating certain antigenic peptides. We have shown previously that when all six β-subunits are present, cooperative assembly mechanisms limit the diversity of proteasome populations. Specifically, LMP7 is incorporated preferentially over X into preproteasomes containing LMP2 and MECL1. We show here that the LMP7 propeptide is responsible for this preferential incorporation, and it also enables LMP7 to incorporate into proteasomes containing delta and Z. In contrast, the X propeptide restricts incorporation to proteasomes with delta and Z. Furthermore, we demonstrate that the LMP7 propeptide can function in trans when expressed on LMP2, and that its NH2-terminal and mid-regions are particularly critical for function. In addition to identifying a novel propeptide function, our results raise the possibility that one consequence of LMP7 incorporation into both immunoproteasomes and delta/Z proteasomes may be to increase the diversity of antigenic peptides that can be generated.

Proteasomes are multisubunit, multicatalytic proteases responsible for the majority of non-lysosomal protein degradation in eukaryotic cells (1). They recognize and degrade ubiquitinated proteins, including those that are misfolded or damaged, as well as other regulatory proteins targeted for rapid turnover (2). Additionally, they are responsible for generating the majority of peptides presented by major histocompatibility complex class I molecules (3). The 26 S proteasome consists of a 20 S proteolytic core with a 19 S complex bound at one or both ends (4,5). The 20 S complex is composed of 28 subunits arranged in four stacked heptameric rings, with the outer two rings containing seven different ␣-type subunits (␣1-␣7) and the inner rings seven different ␤-type subunits (␤1-␤7) (6,7). Three ␤ subunits, delta (␤1), Z (␤2), and X (␤5), possess catalytic activity (8). In vertebrates there are three additional catalytic ␤-subunits, LMP2 (␤1i), MECL1 (␤2i), and LMP7 (␤5i), whose expression is induced by the pro-inflammatory cytokine IFN-␥ 1 (9). Each inducible subunit can substitute for its homologous constitutive counterpart (LMP2 for delta, MECL1 for Z, and LMP7 for X) during proteasome assembly to generate immunoproteasomes (10 -12). Two additional proteins induced by IFN-␥ (PA28a and PA28b) oligomerize to form an 11 S regulatory complex that can associate with the 20 S core (13)(14)(15). The precise role of the IFN-␥-inducible catalytic and regulatory subunits is not known. However, the prevailing concept is that immunoproteasomes capped with PA28 generate peptides whose length and COOH-terminal residue are better suited for binding to major histocompatibility complex class I molecules (16), and/or they increase the repertoire of peptides produced from a given antigen, and thus augment an immune response.
In the fully assembled 20 S proteasome, each ␣and ␤-subunit occupies a defined location relative to other subunits (7,17). Although the mechanism and orchestration of assembly is not well understood, it appears to occur in at least two stages, the first being the formation of a stable preproteasome complex (13-16 S) composed of an intact seven-member ␣-ring with at least three ␤ subunits (Z/␤2, C10-II/␤3, and C7-I/␤4) (18,19). Once the ␤ ring is complete, two half-proteasomes dimerize at the ␤ ring interface to form a 20 S complex. In yeast (and probably also in vertebrates (20)), assembly is facilitated by ump1p, a constituent of the preproteasome complex that is degraded once assembly is complete (21). In mammalian cells an added level of complexity results from the presence of IFN-␥-inducible catalytic ␤ subunits. If assembly were random as many as 36 different proteasome subsets could form (22). Instead, the IFN-␥-inducible subunits are incorporated cooperatively resulting predominantly (but not exclusively) in homogeneous subsets of immunoproteasomes and constitutive proteasomes (23). Cooperative assembly appears to involve coincorporation of LMP2 (␤1i) and MECL1 (␤2i) into preproteasomes (24) with ␤3 and ␤4 (19), thus excluding delta (␤1) and Z (␤2), with subsequent preferential incorporation of LMP7 (␤5i) rather than X (␤5) (23). Since LMP2 and MECL1 are adjacent in the ␤ ring their co-incorporation may be a result of physical interaction (25). However, the mature LMP7 subunit does not contact LMP2 or MECL1, either across one ␤ ring or the interface of the two ␤ rings, raising the question of how it is preferentially incorporated.
Each proteolytic ␤ subunit is expressed with an NH 2 -terminal propeptide that is removed autocatalytically during the final stages of assembly (26,27). Propeptide removal frees the active site NH 2 -terminal nucleophile allowing the subunit to become proteolytically active. Thus, one proposed function of the propeptide is to prevent catalytic activity before the active site can be sequestered in the proteasome core. The propeptide also protects the NH 2 terminus from acetylation and hence inactivation (28), and may serve as an intramolecular chaperone facilitating subunit folding (29).
The ␤5 propeptide is particularly important for proteasome assembly. In yeast, where there is only one ␤5 subunit, deletion of its propeptide disrupts assembly and ultimately viability (26). Interestingly, viability is restored when ump1p is deleted, suggesting that an interaction between the ␤5 propeptide and ump1p is normally required for assembly to proceed (21). Similarly, in mammalian cells LMP7 (␤5i) is not incorporated into proteasomes without its propeptide (30). Furthermore, a variant of LMP7 (LMP7E1) which differs from the predominant form (LMP7E2) only in its propeptide (31), is not incorporated (32). In the work presented here, we explore the role of ␤5 propeptides in the assembly of mammalian proteasome populations. We define two regions of the LMP7 propeptide critical for efficient incorporation of LMP7, and show that this propeptide can function "in trans" when fused to the mature LMP2 subunit. Most importantly, we provide evidence that differences between the LMP7 (␤5i) and X (␤5) propeptides influence which population of assembling proteasomes will incorporate each subunit.

EXPERIMENTAL PROCEDURES
Cell Cultures and Antibodies-Lymphoblastoid T2 cells (0.174 ϫ CEM R ) (33) obtained from P. Cresswell (Yale University, New Haven, CT) were grown in RPMI 1640 medium supplemented with 10% fetal bovine serum, L-glutamine, and antibiotics as described (23). MCP21 is a mouse monoclonal antibody that recognizes human C3 and immunoprecipitates both 20 S proteasomes and 13-16 S preproteasomes (34). The hybridoma was obtained from the European Collection of Animal Cell Cultures (Salisbury, Wiltshire, UK). Polyclonal antisera recognizing LMP2 and LMP7 were from rabbits immunized with recombinant mouse subunits (19). Anti-X (P93250) is a rabbit polyclonal antiserum raised against human X, and was obtained from K. B. Hendil (August Krogh Institute, University of Copenhagen, Copenhagen, Denmark).
DNA Constructs and Transfection-The cDNAs encoding human LMP7 (LMP7E2) and LMP7E1 were expressed using the episomal vector pCEP4 (hygromycin) (Invitrogen, Carlsbad, CA), and LMP2 was expressed using pCEP9 (neomycin r ), as described (23). pCEP9 was constructed from pREP9 and pCEP4, and is similar to pCEP4 except that it confers resistance to neomycin rather than hygromycin (23). The human X cDNA was obtained from T. Spies (Fred Hutchinson Cancer Research Center, Seattle, WA). Altered forms of LMP7 and X (see below) were expressed in pCEP4, and altered forms of LMP2 in pCEP9. T2 cells were transfected with cDNAs in episomal vectors, and selected with G418 and/or hygromycin, as described (23). To produce a T2 cell line overexpressing both LMP2 and LMP7 under one selection marker, LMP2 and LMP7 cDNAs were subcloned into pSG5 (Stratagene, La Jolla, CA), and linearized constructs (25 g each) were co-transfected with linearized pSV2.Neo (2.5 g) (CLONTECH, Palo Alto, CA) by electroporation (250 V and 500 microfarads). After selection with 0.8 mg (activity)/ml of G418, growth positive wells were expanded and screened for expression by immunoblotting. The cell line with the highest levels of LMP2 and LMP7 in immunoprecipitated proteasomes, and the lowest levels of delta and X (indicating replacement), were used for subsequent experiments.
Construction of Mutant and His-tagged Proteasome Subunits-The ExoIII/Mung Bean Nuclease Deletion Kit (Stratagene) was used to generate LMP7 propeptide deletion mutants ⌬ϩ1, ⌬-35, ⌬-49, and ⌬-67. The kit was used following the manufacturer's instructions. Briefly, the full-length human LMP7 cDNA was cloned into pBluescript II KS (Stratagene) between KpnI and BamHI restriction sites. An oligonucleotide linker was inserted between KpnI and NheI upstream of the LMP7 start codon to provide 5Ј susceptible (NheI) and 3Ј resistant (KpnI) overhangs that enable unidirectional digestion by ExoIII. The linker provides a start codon and translation control sequence upstream of the resistant KpnI site. After varying periods of ExoIII digestion, mung bean nuclease was used to create ligatable blunt ends. Candidate LMP7 propeptide deletion mutants were sequenced to identify and characterize in-frame deletions. Selected constructs were then subcloned into pCEP4 using NotI and BamHI restriction sites. ⌬-69 and ⌬-70 were generated using overlapping oligonucleotides extending from an internal EagI site at nucleotide 40 of the cDNA (the "A" of the ATG start codon represents nucleotide 1). ⌬-3.LMP7 was produced by removing a fragment between an NcoI restriction site introduced at the start codon by site-directed mutagenesis, and the NcoI restriction site at nucleotide 210. A small linker fragment was inserted to restore the correct reading frame. Pro7(⌬-23:-11).LMP7 was generated by removing the DNA fragment between SmaI and NcoI restriction sites, and substituting an oligonucleotide linker encoding the necessary residues.
To generate LMP7E1 with an LMP7E2 NH 2 -terminal region ((E2)7E1) the EagI to EcoRI fragment of LMP7E2 encoding 39 residues (Pro 58 -Phe 20 ) was deleted and replaced with overlapping oligonucleotides encoding the comparable region of LMP7E1 (Pro 56 -Phe 20 ). (X)7E1 was engineered by replacing the NH 2 -terminal coding region of (E2)7E1 (extending from the EagI restriction site) with an oligonucleotide sequence encoding Met 59 -Arg 44 of the X propeptide. The above constructs were generated in the pcDNA II vector (Invitrogen), and subcloned into pCEP4 with a reversed multiple cloning site using XhoI and HindIII restriction sites. LMP7 with an X propeptide (proX.LMP7) was generated by removing the XhoI to NcoI fragment beginning 30 nucleotides upstream of the start codon and ending at nucleotide 210, and replacing it with overlapping oligonucleotides encoding the X propeptide. X with an LMP7 propeptide (pro7.X) was engineered using the Seamless Cloning Kit (Stratagene) to directly clone the LMP7 propeptide cDNA fragment with the mature X cDNA. The COOH-terminal His-tags contain six histidines preceded by Arg⅐Gly⅐Ser. X.His and pro7.X.His were produced from X and pro7.X by cloning overlapping oligonucleotides into a BamHI restriction site at nucleotide 727. LMP7.His was generated by polymerase chain reaction using a 3Ј primer that encoded the His-tag. ProX.LMP7.His was generated by removing the XhoI to NcoI fragment from LMP7.His, and replacing it with the same restriction fragment from proX.LMP7.
To make pro7.LMP2 we first removed an internal NcoI site in the LMP2 cDNA without changing the encoded amino acids, using sitedirected mutagenesis (Altered Sites Kit, Promega, Madison, WI). A new NcoI site was then introduced by site-directed mutagenesis at the same place as found in LMP7, just 5Ј of the codon for Thr 1 . The LMP2 propeptide coding region was then replaced with the LMP7 propeptide using this new NcoI site and an upstream XhoI site. LMP2.His was made by polymerase chain reaction amplification of the LMP2 cDNA using a 3Ј primer that inserted six Histidine codons before the stop codon.
Every construct used in this study was sequenced to confirm anticipated changes, and to ensure that no other coding changes were introduced. The translational control sequences upstream of the start codon were preserved in all constructs.
Quantitative Immunoblotting-Gel electrophoresis and protein blotting was performed as described above except an ECF substrate (Amersham Pharmacia Biotech) was used. Relative chemifluorescence was quantitated using the Storm 860 PhosphorImaging system and Image-Quant software (version 1.2) (Molecular Dynamics, Sunnyvale, CA). For each protein band the volume in pixels was measured, and the background from an adjacent gel region was subtracted. In separate experiments we determined that sequential 2-fold sample dilution reduced the signal (number of pixels) approximately 2-fold, indicating linearity in the range of interest.

Propeptide Regions Necessary for LMP7 Incorporation into
Proteasomes-Previous studies have shown that deletion of the entire LMP7 propeptide prevents incorporation of the subunit into 20 S proteasomes (30). We were interested in whether certain regions of the propeptide might be more important for function than others. To test this, we generated a series of partial deletions from the NH 2 terminus, ranging from removal of one residue (⌬-70) to full propeptide truncation (⌬ϩ1) (Fig.  1A). The NH 2 -terminal methionine codon was preserved in all constructs. Each deletion mutant was expressed alone and with LMP2 in T2 cells, which lack endogenous LMP2 and LMP7 (33,35). Proteasomes were immunoprecipitated from cell lysates using MCP21, a monoclonal antibody that recognizes C3, a proteasome ␣-type subunit (34). Following immunoprecipitation, proteasomes were denatured and subunits separated by SDS-PAGE. The presence of specific subunits was determined by immunoblotting with subunit-specific antisera. Large deletions of the propeptide (to residue ϩ1, Ϫ3, or Ϫ35) dramatically reduce LMP7 incorporation. This is apparent when these altered forms of LMP7 are either expressed alone (Fig. 1A, lane 1 of each group), or when co-expressed with LMP2 (lane 3 of each group). In contrast, removal of a smaller region (to Ϫ70 or Ϫ49) does not prevent incorporation, although the efficiency is reduced compared with full-length LMP7. Deletion of two or four amino acid residues to Ϫ69 and Ϫ67, respectively, has an effect similar to removal of a single residue (to Ϫ70) (data not shown).
The LMP7 subunit is incorporated preferentially over X into proteasomes containing LMP2 and MECL1 (23). It is also incorporated into delta/Z proteasomes when LMP2 is absent. That this latter population might also form in the presence of LMP2 is suggested by pre-clearing experiments using anti-LMP2 (36), and greater overall incorporation of LMP7 when co-expressed with LMP2 (23, 31) (Fig. 1A). To confirm this directly we tagged LMP2 with polyhistidine at the COOH terminus (LMP2.His) where it should not interfere with incorporation (7,37), allowing us to selectively remove LMP2-containing immunoproteasomes from the entire pool of proteasomes using Ni 2ϩ -NTA-agarose. Indeed, when LMP7 and LMP2.His are co-expressed, LMP7 is found not only in LMP2-containing proteasomes removed by nickel precipitation (Fig. 1B), but also in proteasomes lacking LMP2 that can be precipitated with MCP21 after exhaustive removal of the His-tagged population (data not shown). To determine whether partially deleted forms of LMP7 are still incorporated into LMP2-containing immunoproteasomes, we used nickel precipitation to isolate this population from cells co-expressing ⌬-49.LMP7 and LMP2.His. As shown in Fig. 1B, this population contains LMP7, indicating that the partially truncated form is incorporated into immunoproteasomes. Taken together, these results suggest that removal of short segments from the NH 2 terminus of the LMP7 propeptide reduces subunit incorporation slightly, while further deletion of the region between residues Ser 49 and Arg 35 completely abolishes it. Notably, these deletions do not appear to selectively reduce incorporation of LMP7 into immunoproteasomes.
The LMP7 Propeptide Functions in Trans When Fused to LMP2-In yeast, the ␤5 subunit propeptide expressed as a separate polypeptide (in trans) restores incorporation of the truncated ␤5 subunit (26). Using a similar approach with the LMP7 propeptide, we did not observe any incorporation of truncated LMP7 (⌬ϩ1, ⌬-3, or ⌬-35) in T2 cells (data not shown). However, when the LMP7 propeptide is expressed as a fusion protein on LMP2 (replacing the LMP2 propeptide), incorporation of these truncated forms of LMP7 is at least partially restored (Fig. 1A, lane 2, of each group). Furthermore, this form of LMP2 is incorporated (data not shown), and replaces the constitutive homologous ␤ subunit (delta), as shown previously by Schmidtke et al. (27). These results demonstrate that the LMP7 propeptide can function in trans when expressed on another ␤ subunit. Its inability to function as a separate polypeptide may be due to ineffective competition with the endogenous ␤5 subunit (X). We have attempted to express LMP2, truncated LMP7, and the LMP7 propeptide together so that a pool of preproteasomes dependent on LMP7 for maturation would be created. However, these experiments have not been successful, at least in part due to difficulties obtaining stable triple transfectants.
The NH 2 -terminal Region Is Important for Propeptide Function-The two forms of LMP7 2 that the result from alternative first exon use (LMP7E1 and LMP7E2) encode polypeptides that differ only in the NH 2 -terminal two thirds of their propeptides ( Fig. 2A), yet LMP7E1 is not incorporated into proteasomes (23,32). The NH 2 -terminal regions of the two functional ␤5 propeptides (LMP7E2 and X) are homologous, and yet quite different from the same region of LMP7E1 ( Fig. 2A). Conse-2 Throughout this article all reference to LMP7 implies LMP7E2 except where specified. quently, to test whether this region confers functionality, we replaced the NH 2 -terminal region of LMP7E1 with residues from LMP7E2 or X to produce (E2)7E1 and (X)7E1, respectively (Fig. 2B). Both of these proteins are incorporated into proteasomes, although not as efficiently as LMP7E2 (Fig. 2B). There appears to be enhanced incorporation when they are coexpressed with LMP2, suggesting incorporation into immunoproteasomes. This is further supported by enhanced incorporation of LMP2 under these same conditions (data not shown).
Since the LMP7 and X propeptides differ in length by 13 amino acids ( Fig. 2A), and X is primarily restricted to constitutive proteasomes, we questioned whether the ability of LMP7 to incorporate into immunoproteasomes was due to its longer propeptide. To test this we deleted 13 residues (Pro 23 through Glu 11 ) closer to the mature subunit, preserving the NH 2 -terminal region and mid-portion of the propeptide. In contrast to other deletions, this has no effect on overall LMP7 incorporation nor does it reduce LMP2 incorporation (Fig. 2C), implying that this region closer to the mature subunit is dispensable.
Taken together, the results in Figs. 1 and 2 suggest that there are at least two regions of the LMP7 propeptide that are required for full function. The NH 2 -terminal region restores partial function to an otherwise non-functional propeptide (LMP7E1), but when removed from LMP7E2 only causes a partial loss of function (⌬-49.LMP7). However, the middle region is also important as removal of residues Ϫ49 to Ϫ35 completely prevents subunit incorporation (⌬-35.LMP7). It should be noted that incorporation of LMP7 into immunoproteasomes does not appear to be determined solely by either of these regions, nor is it due to the greater length of the LMP7 propeptide.
The LMP7 Propeptide Facilitates Incorporation of X into LMP2-containing Proteasomes-Although there is considerable homology between the mature LMP7 and X proteins (ϳ68% identical), the propeptides of these two ␤5 subunits are considerably different with the exception of their NH 2 -terminal regions (12) (Fig. 2A). Thus, we hypothesized that differential incorporation of these two subunits is a property of their propeptides. To test this, we replaced the X propeptide with the LMP7 propeptide (pro7.X) and co-expressed pro7.X with LMP2 in T2 cells. Pro7.X dramatically enhances incorporation of LMP2 into proteasomes (Fig. 3A), comparable to the increase seen with LMP7 (data not shown), while overexpression of full-length X has little effect. Precursor LMP2 (pre-LMP2) often accumulates in preproteasomes when LMP7 is absent, and thus is seen in MCP21 immunoprecipitations (23) (Fig. 3A). However, this varies with growth conditions and the timing of cell harvest (Ref. 23, and data not shown). Pre-LMP2 accumulation also appears to be diminished when X is overexpressed (Fig. 3A). Nevertheless, considerably greater LMP2 incorporation with co-expression of pro7.X compared with full-length X is a consistent finding, and suggests that pro7.X is readily incorporated into LMP2-containing proteasomes. To address this more directly, pro7.X and X were tagged with COOH-terminal polyhistidine (pro7.X.His and X.His), enabling purification of proteasomes containing these subunits, and allowing us to distinguish His-tagged X from endogenous X on immunoblots. When LMP2 and pro7.X.His are co-expressed, LMP2 incorporation is greatly enhanced (Fig. 3B), as shown with the non-His-tagged subunit (Fig. 3A). More importantly the majority of LMP2 co-purifies with His-tagged proteasomes, indicating that pro7.X.His is indeed incorporating into LMP2-containing proteasomes. Diminishing amounts of C3 after three consecutive rounds of nickel precipitation indicates removal of the majority of His-tagged proteasomes (Fig. 3B). Subsequent immunoprecipitation with MCP21 reveals a substantial pool of proteasomes remaining that contain endogenous X and C3, but minimal LMP2 (Fig. 3B). We also find that pro7.X.His, like full-length LMP7, is not restricted to LMP2-containing proteasomes as it is readily incorporated in the absence of LMP2 (data not shown). Therefore, the LMP7 propeptide confers to the ␤5 subunit an ability to incorporate into both pools of proteasomes.
The X Propeptide Restricts Incorporation of LMP7 to Primarily Delta-containing Proteasomes-Since the propeptide of LMP7 appears to be the primary determinant allowing incorporation of ␤5 into LMP2 and delta-containing proteasomes, it seemed likely that the X propeptide would restrict incorporation to delta-containing proteasomes. To address this, we constructed LMP7 with an X propeptide (proX.LMP7), and ex- pressed it in the presence or absence of LMP2. When expressed alone proX.LMP7 is incorporated into proteasomes at a level comparable to wild-type LMP7 (Fig. 4, lane 2 versus lane 4). However, unlike full-length LMP7, proX.LMP7 does not enhance LMP2 incorporation (Fig. 4, lane 3 versus lane 5). In addition, proX.LMP7 incorporation is not increased as is typically seen with LMP7 when it is co-expressed with LMP2. This implies that while proX.LMP7 is competent to incorporate into delta-containing proteasomes, it is not capable of efficiently allowing the maturation of LMP2-containing proteasomes. These results suggest that the X propeptide restricts the ␤5 subunit to primarily constitutive proteasomes, preventing it from being incorporated into immunoproteasomes.
Quantitative Immunoblotting-To obtain a more quantitative estimate of the effect of the ␤5 propeptide on immunoproteasome assembly as judged by LMP2 incorporation, an ECF substrate was used for immunoblotting (see "Experimental Procedures"). LMP2 was co-expressed with X.His, pro7.X.His, LMP7.His, or proX.LMP7.His, and His-tagged proteasomes were removed by nickel precipitation. Subsequent immunoprecipitation with MCP21 (after 3 rounds of nickel precipitation) demonstrated that all of the detectable LMP2 had been removed (Fig. 5). Quantitative analysis of the first nickel precipitation fraction reveals the that amount of LMP2 incorporated into proteasomes with pro7.X is over 7 times greater than with X, and approximately 5-fold greater for LMP7 compared with proX.LMP7 (Fig. 5). These numbers probably represent a minimal estimate of the quantitative effect, as there is likely to be additional material detected in the second and third rounds of nickel precipitation for pro7.X.His and LMP7.His, while X.His and proX.LMP7.His are already near background. These results confirm that the ␤5 propeptide has an important role in determining relative efficiency of ␤5 incorporation into LMP2containing proteasomes, and thus overall immunoproteasome assembly.
The LMP7 Propeptide Allows X to Compete Effectively with LMP7 for Incorporation-To further address the importance of the ␤5 propeptide, we expressed His-tagged ␤5 subunits in T2 cells overexpressing LMP2 and LMP7. Overexpression of these immunoproteasome subunits leads to virtually complete re-placement of delta and X (Fig. 6, and data not shown). When an additional His-tagged ␤5 subunit is expressed from an episomal vector, it competes effectively for incorporation into proteasomes only when an LMP7 propeptide is present (i.e. pro7.X or LMP7). Note that the His-tagged proteasomes contain LMP2 as expected, and some endogenous non-His-tagged LMP7, suggesting mixed proteasomes form. Furthermore, pro7.X.His appears to be incorporated as efficiently as LMP7.His, given the relative amounts of LMP2, C3, and LMP7 (non-His-tagged) that co-precipitate. These results emphasize that X can compete effectively with LMP7 for incorporation into immunoproteasomes if it has an LMP7 propeptide. DISCUSSION Our studies indicate that in mammalian cells, where constitutive and inducible catalytic ␤5 subunits can be co-expressed, the ␤5 propeptide determines which population of assembling proteasomes will incorporate this subunit. The X propeptide restricts ␤5 (X) primarily to constitutive proteasomes containing delta (␤1) and Z (␤2), whereas the LMP7 propeptide facilitates incorporation of ␤5i (LMP7) into immunoproteasomes containing LMP2 (␤1i) and MECL1 (␤2i), but also permits FIG. 3. The LMP7 propeptide allows X to incorporate into LMP2-containing proteasomes. A, proteasomes were immunoprecipitated from T2 cells expressing LMP2 alone (Ϫ), or co-expressing LMP2 with X or pro7.X. Material from 16 ϫ 10 6 cells was separated and immunoblotted with anti-LMP2. B, His-tagged proteasomes were isolated from T2 cells co-expressing LMP2 and X.His (left panel) or LMP2 and pro7.X.His (right panel), by three successive nickel precipitations (lanes 1-3). Remaining material was immunoprecipitated with MCP21 (IP). Proteasome subunits were separated and immunoblotted with subunit-specific antisera. Precipitates from 7 ϫ 10 6 cells were used for the X and C3 immunoblots, while 16 ϫ 10 6 cells were used for the LMP2 immunoblots. FIG. 4. The X propeptide on LMP7 reduces LMP2 incorporation and prevents increased LMP7 incorporation. Proteasomes were immunoprecipitated from untransfected T2 cells (Ϫ), or T2 cells expressing LMP7 without (Ϫ) or with (ϩ) LMP2, and proX.LMP7 without (Ϫ) or with (ϩ) LMP2. Proteasome subunits were separated and visualized by immunoblotting with subunit-specific antisera. Material from 5 ϫ 10 6 , 16 ϫ 10 6 , and 7 ϫ 10 6 cells was used for the LMP7, LMP2, and C3 immunoblots, respectively. A small amount of proX.LMP7 precursor (arrowhead), and the nonspecific band (*) are indicated.
FIG. 5. Quantitation of the ␤5 propeptide effect on LMP2 incorporation. LMP2 was co-expressed with His-tagged X or LMP7 containing different ␤5 propeptides as indicated. His-tagged proteasomes were isolated by nickel precipitation (3 successive rounds), and remaining proteasomes immunoprecipitated with MCP21. Proteasome subunits were separated, and LMP2 visualized by immunoblotting with anti-LMP2 antisera and ECF substrate. Material from the first nickel precipitation (Ni) and the immunoprecipitation (IP) is shown. Relative chemifluorescence was quantitated as described under "Experimental Procedures," and is expressed as the total number of pixels (volume) for each band after background subtraction using an adjacent gel region (bar graph). Starting material was 13 ϫ 10 6 cells for each cell line.
incorporation into the population containing delta and Z. Thus, the LMP7 propeptide accounts for the contribution of this subunit to cooperative immunoproteasome assembly (23). These results confirm the previously recognized role of the ␤5 propeptide as an intermolecular chaperone (26), and further demonstrates for the first time that it influences the differential incorporation of homologous ␤5 subunits into discrete proteasome subsets. Other important functions of the ␤5 propeptide may include preventing premature proteolytic activity and active site acetylation (28,38), as well as assisting subunit folding (intramolecular chaperone) (29).
Several observations indicate that the NH 2 -terminal and mid-regions of the LMP7 propeptide are necessary for complete function. First, deleting the NH 2 -terminal half of the propeptide (to Ser 35 ) abrogates LMP7 incorporation, while removing one to as many as 23 residues (to Ser 49 ) has an intermediate effect. Second, in contrast to NH 2 -terminal deletions, removing 13 residues from the COOH-terminal region (Pro 23 through Glu 11 ) has no effect. Third, replacing the NH 2 -terminal region of a non-functional propeptide (LMP7E1), with the comparable region from LMP7E2 or X partially restores function. These deletions and switches of the NH 2 -terminal region did not reveal a domain conferring proteasome subset specificity, as there was no major effect on differential incorporation of LMP7 into immunoproteasomes versus delta-containing proteasomes. In contrast, full propeptide substitutions reversed the specificity of subunit incorporation. Thus, rather than a particular region of the LMP7 propeptide enhancing incorporation into immunoproteasomes, there may be a region of the X propeptide that negatively affects incorporation into immunoproteasomes. Further experiments will be necessary to identify such a region.
Rhodococcus proteasome ␤-subunit propeptides appear to facilitate subunit folding (29), although a similar function has not been demonstrated for eukaryotic proteasome propeptides. Our propeptide deletion experiments do not distinguish between deleterious effects on subunit folding that could indirectly diminish incorporation, versus removal of a region directly involved in incorporation. The observation that truncated forms of LMP7 can be incorporated when the LMP7 propeptide is expressed in trans (fused to LMP2) could indicate that some of the truncated subunits fold appropriately without the influence of an attached full-length propeptide. However, it is also possible that the propeptide of LMP7 attached to LMP2 could interact with truncated forms of LMP7 in trans and facilitate folding. It is worth noting that any propeptide effect on folding is unlikely to be subunit or propeptide-specific, since switching the LMP7 and X propeptides affects which population will accept the subunit, not whether it can be incorporated. Similarly, LMP2 with an LMP7 (27) or delta 3 propeptide can be incorporated. Therefore, while we cannot rule out effects of the ␤5 propeptide on subunit folding, such an effect is not likely to be responsible for differential subunit incorporation.
Accumulating evidence suggests the ␤5 propeptide and subunit occupy a singular position in proteasome assembly and function (28,38). For example, in yeast, the ␤5 propeptide is required for incorporation of this subunit and for viability (26). In mammals, LMP7 (␤5i) is also dependent on its propeptide for incorporation, while less is known about the constitutive subunit X (␤5). In contrast, ␤1 and ␤2 propeptides are not essential for incorporation or viability in yeast, although perturbations in cell growth are observed when these propeptides are deleted, particularly with ␤2 (28,38). Truncated forms of LMP2 (␤1i) are incorporated into proteasomes, albeit with reduced efficiency 3 (25), while the role of the MECL1 (␤2i) propeptide has not been investigated. Although the ␤5 propeptide influences proteasome subset formation as we have shown here, this is not the case with ␤1. Replacing the LMP2 (␤1i) propeptide with the delta (␤1) propeptide does not change its dependence on LMP7 for incorporation, suggesting it still incorporates primarily into immunoproteasomes. 3 Furthermore, partial deletion, mutation, or complete truncation of the LMP2 (␤1i) propeptide does not significantly affect co-incorporation with MECL1 (␤2i), suggesting that this results from interactions between the mature subunits rather than being a function of their propeptides (25). Finally, the ␤5 propeptide is critical for completion of proteasome assembly in yeast, apparently via interaction with ump1p, a key mediator of this process (21).
Recent studies in yeast demonstrate a hierarchy of importance of the catalytic subunits, with ␤5 maintaining the most crucial proteolytic activity (38,39). For example, mutation of ␤5 affects degradation of ubiquitinated substrates and cell growth more profoundly than mutation of either ␤1 or ␤2. Interestingly, studies from knockout mice also support a particularly critical role for ␤5i (LMP7). Major histocompatibility complex class I expression is reduced (ϳ50%) on lymphoid cells from LMP7-deficient (LMP7 Ϫ/Ϫ ) mice (40), while it is unchanged in cells from LMP2 Ϫ/Ϫ mice (41). Whether these differences are attributable to the absence of LMP7 alone, or are compounded by reduced LMP2 and MECL1 incorporation in LMP7 Ϫ/Ϫ mice (23), is not clear. However, PA28b Ϫ/Ϫ mice have normal class I expression despite a defect in immunoproteasome assembly (42). Interestingly, while LMP2 and MECL1 are not incorporated into proteasomes in these mice, LMP7 is incorporated, albeit at reduced levels. Likewise, LMP7 is readily incorporated in LMP2 Ϫ/Ϫ mice (23,41). The presence of LMP7containing proteasomes in both PA28b Ϫ/Ϫ and LMP2 Ϫ/Ϫ mice suggests that high level class I expression may be more dependent on the presence of LMP7 than on immunoproteasomes (e.g. LMP2/MECL1/LMP7) per se. This is consistent with previous findings in vitro where expression of LMP7 without LMP2 increases production of peptides with hydrophobic COOH-terminal residues suitable for class I binding (16,43). FIG. 6. The LMP7 propeptide expressed on X or LMP7 enables the subunit to compete with overexpressed LMP7 for incorporation into immunoproteasomes. T2 cells overexpressing LMP2 and LMP7 were prepared as described under "Experimental Procedures." His-tagged X and LMP7 with different propeptides (X.His, pro7.X.His, LMP7.His, and proX.LMP7.His) were then expressed using an episomal vector. His-tagged proteasomes were isolated by three successive nickel precipitations (Ni), then remaining proteasomes immunoprecipitated with MCP21 (IP). Precipitated material was electrophoresed and proteasome subunits visualized by immunoblotting. For nickel precipitates only the first fraction is shown. Material from 5 ϫ 10 6 cells was used for the LMP7 and C3 immunoblots, while 15 ϫ 10 6 and 7 ϫ 10 6 cells were used for the LMP2 and X immunoblots, respectively. His-tagged X (filled arrowhead), His-tagged LMP7 (open arrowhead), and other subunits are indicated on the immunoblots. The evolution of a ␤5 propeptide that enables the formation of delta/Z/LMP7 as well as LMP2/MECL1/LMP7 proteasomes raises the question of whether the delta/Z/LMP7 population is immunologically relevant. This subset is most likely to form when there is incomplete replacement of delta/Z by LMP2/ MECL1, such as in unstimulated dendritic cells (44) and other lymphoid tissues, or with suboptimal induction by IFN-␥. It is plausible that the primary role of LMP7 may be to increase the overall production of peptides that can be presented by class I molecules, and that further digestion by delta/Z active sites in addition to LMP2/MECL1, could enhance the overall diversity of peptides available for presentation.
A major challenge will be to determine how differences in the ␤5 propeptide sequence confer specificity for incorporation into proteasome subsets. Several mechanisms are worth considering. First, the ␤5 propeptide is large enough (59 -72 amino acids) to directly contact ␤1 or ␤2 across the ring, so that specificity could be based on differential interaction with LMP2 or MECL1. This would be unlikely to result from ␤1/␤1i propeptide differences, as the nature of the LMP2 propeptide seems to have little effect on the requirement for LMP7 for efficient incorporation. 3 A second possibility is the existence of two mammalian homologues of yeast ump1p, with one directing assembly of constitutive proteasomes and the other immunoproteasomes. We recently identified human and mouse homologues of ump1p by searching EST data bases (20), and find no evidence for a second related protein, making this prospect seem unlikely. A third possibility is that other accessory proteins are involved in immunoproteasome assembly. In this regard, it was recently shown that PA28a and PA28b associate with preimmunoproteasomes, and in the absence of PA28b (PA28b Ϫ/Ϫ mice) recovery of these complexes is reduced, as is the incorporation of immunoproteasome subunits (42). This raises the possibility that the role of PA28 in proteasome assembly may involve binding to the ␣-ring and inducing a conformational change that makes preproteasomes more receptive to LMP2 and/or MECL1. It should be noted that these scenarios are not mutually exclusive, nor do they readily explain why the LMP7 propeptide directs more promiscuous incorporation.
In summary, it appears that the two mammalian ␤5 propeptides are important for influencing the formation of distinct proteasome subsets. It will be critical in future studies to determine the precise function of each of these subsets including LMP7/delta/Z, to establish how IFN-␥-inducible subunits influence the immune response.