The GenBank/EMBL/DDBJ ubiquilin cDNA accession number is AF176069, and the ubiquilin protein accession number is AAD49751 (see updated NM_013438).

The yeast two-hybrid (Y2H) procedure that utilizes the LexA/transactivation system was performed, as described previously (Stabler et al. 1999). The PS2 COOH-terminal 39-amino acids (residues K410–I448) (see Fig. 1a and Fig. b) were used as bait to screen a human fetal brain cDNA library. Out of ∼10⁷ primary yeast transformants, ∼50 clones were selected for further analysis based on growth in the absence of leucine and a substantial blue color change when grown on 5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside (X-gal) plates. False positives were subsequently eliminated based on their interactions with negative-control baits, such as LexA alone, nuclear lamin B, and CENP-C centromere protein. Plasmid DNAs were isolated from clones that fulfilled the above requirements, and the DNA sequences of their inserts were determined and compared with the GenBank/EMBL/DDBJ databases using the BLAST search program. Two of the interactors contained overlapping portions of the 3′ end of a novel human transcript, which we named ubiquilin. Interaction of ubiquilin proteins with various PS1 and PS2 baits were quantified in liquid culture assays by the enzymatic conversion of o-nitrophenyl-β-d-galactopyranoside (ONPG) to o-nitrophenol and d-galactose by β-galactosidase enzyme encoded by a Y2H reporter plasmid.

5′ RACE, using strategically designed ubiquilin-specific primers, was performed to obtain the full-length coding sequence of ubiquilin. The PCR reactions were carried out according to the instructions provided by the manufacturer, using an AP1 adaptor–ligated human adult brain cDNA library as template and the Advantage cDNA Polymerase Mix (CLONTECH Laboratories, Inc.). The procedure resulted in the isolation of 1,053 additional bps of the 5′ sequence, with respect to the longest original Y2H clone. The full-length ubiquilin open reading frame (ORF) was subsequently obtained by sequencing a human expressed sequence tag (EST) clone (American Type Culture Collection), which provided an extra 108 bps of 5′ sequence, including the ATG start codon, a Kozak consensus sequence, and an upstream in-frame stop codon. In addition, comparison of the sequence with the TIGR Tentative Human Consensus database returned another independent EST sequence (THC296552), which corroborated the ATG start codon.

A GeneBridge 4 radiation hybrid panel (Research Genetics) was performed with two ubiquilin gene-specific primers, RH1 (5′-GCAGCGACAACTTTTGTCTAACCCTG-3′) and RH2 (5′-CAGGCACCAAATTTGGCGCAGTAG-3′), and each of the 93 indexed human genomic DNA template pools. The conditions for the PCR reactions were similar to that of the 5′ RACE procedure, except that a longer initial denaturation step was used. Positive and negative reactions (1 or 0, respectively; 2 represents an ambiguous reaction) were compiled (0100010001 1002100110 0000011000 1011000000 1100000001 0010001001 0100010000 0111100000 1000000000 000), and the vector data was submitted to the Whitehead Institute/MIT Center for Genome Research website (http://carbon.wi.mit.edu:8000/cgi-bin/contig/rhmapper.pl) for comparison with their radiation hybrid database.

cDNA hybridization probes were ³²P-radiolabeled by random primer labeling of ∼100 ng of two ubiquilin cDNA restriction fragments, X (bases 1,132–1,860) and Y (bases 284–967), shown in Fig. 2 II. Each probe was hybridized to a different human mRNA Northern blot, Human Multiple Tissue Northern Blot I, and Human Brain MTN II, respectively (CLONTECH Laboratories, Inc.). After hybridization, the blots were washed twice with 2.0× SSC and 0.05% SDS at 50°C and then twice again with 0.1× SSC and 0.1% SDS at 50°C, before exposure to autoradiography film. Both blots were subsequently stripped in near-boiling temperature 0.5% SDS and then reprobed with radiolabeled human β-actin cDNA control probe (CLONTECH Laboratories, Inc.). The band intensities were quantified using GelExpert 97 2.0 (Nucleotech Corp.), divided by the β-actin control band intensity from the same lane, and then normalized against the lowest value.

To express full-length nonfusion ubiquilin protein in bacteria, an NcoI/XhoI cDNA fragment containing the entire ubiquilin ORF was subcloned into pET-15b (Novagen) and then transformed into Escherichia coli BL21(DE3). Ubiquilin protein expression was induced with 1.0 mM IPTG at 37°C for 2–3 h. Bacteria were lysed, and the protein extracts were separated by SDS-PAGE. Proteins were transferred onto nitrocellulose filters and immunoblotted with anti-ubiquilin pAb. In addition, various partial-length GST–ubiquilin fusion proteins were constructed, expressed, and purified using the GST-protein purification procedures described previously (Janicki and Monteiro 1997). The specific subregions of ubiquilin that were expressed are depicted in Fig. 2 III.

Rabbit anti-ubiquilin polyclonal antisera were generated against purified GST–ubiquilin fusion proteins B or C (Fig. 2 III) by Covance Research Products, Inc. Affinity-purified anti–ubiquilin-C antibodies were obtained by passing anti–C-serum first over a column containing GST alone coupled to CH-Sepharose 4B beads at 4°C overnight, to eliminate antibodies binding to GST, and then by mixing the flow-through fraction with ubiquilin–coupled CH-Sepharose 4B beads at 4°C overnight. After extensive washing, antibodies that bound to the ubiquilin-coupled beads were eluted dropwise with 0.1 M sodium citrate and 0.3 M sodium chloride solution, pH 2.2, and collected as ∼240-μl fractions into tubes containing 40 μl 1 M Tris-HCl, pH 9.1, to restore neutral pH.

[³⁵S]Methionine-radiolabeled ubiquilin or luciferase proteins were synthesized in rabbit reticulocyte lysates with an in vitro transcription and translation system (Promega) at 30°C for 90 min. [³⁵S]Methionine-radiolabeled PS2 and PS1 were synthesized using the same system, but with the addition of canine pancreatic microsomal membranes (Promega) during transcription and translation.

Various purified GST–ubiquilin fusion proteins already bound to glutathione-agarose beads were incubated with ∼2.5 μl [³⁵S]methionine-radiolabeled presenilin protein, in addition to protease inhibitors and 0.8% BSA at 4°C for 1 h with rocking. The beads were washed several times with pull-down buffer (0.5% NP-40 in 1.0× PBS), once with 200 mM KCl in pull-down buffer, and several more times with pull-down buffer alone. Supernatants and washes were discarded, whereas the bound proteins on the beads were mixed with sample loading buffer (Janicki and Monteiro 1999) and separated by SDS-PAGE. Then, the gel was stained with Coomassie blue to ensure equal loading of GST–ubiquilin fusion proteins. The gels were dried and exposed to autoradiography film.

Full-length ubiquilin was expressed using the pGEM-CMV mammalian expression vector (Janicki and Monteiro 1997) (see Fig. 2 IV). In addition, two ubiquilin fusion proteins were constructed within the same expression vector. In one, green fluorescent protein (GFP) was fused to the NH₂ terminus of ubiquilin (beginning at residue 20 [Ala]), whereas in the other, a myc epitope tag was fused to the COOH terminus of full-length ubiquilin.

HeLa cells were grown at 37°C in DME supplemented with 10% FBS and transfected with plasmid DNAs, using the calcium phosphate coprecipitation method.

Cells were lysed in buffer containing protease inhibitors (Monteiro and Mical 1996), and protein concentrations were determined by the bicinchonic acid (BCA) assay. Lysates containing known amounts of protein were mixed with sample loading buffer (Janicki and Monteiro 1999) and either heated at 100°C for 5 min or at 37°C for 15 min before separation by SDS-PAGE. The 37°C incubation was used to detect presenilin proteins by minimizing their tendency to aggregate and resist entry into the gels. After gel electrophoresis, proteins were transferred onto nitrocellulose filters and immunoblotted according to procedures described previously (Janicki and Monteiro 1997). In addition to our own rabbit anti-ubiquilin antibodies, we also used the following primary antibodies: goat anti-PS1, anti-PS2, and anti-BiP (NH₂ terminus specific; Santa Cruz Biotechnology, Inc.); rabbit anti-PS2 (NH₂ terminus or loop specific, raised to the corresponding GST–PS2 fusion proteins) (Janicki and Monteiro 1997, Janicki and Monteiro 1999) anti-ubiquitin (Dako), anti-lamin A/C (Mical and Monteiro 1998), anti-calnexin, and anti-calreticulin (StressGen); and mouse anti-GFP (Zymed Laboratories), anti–α-tubulin (Sigma-Aldrich), and anti-lamin B (MatriTect). Binding of primary antibodies were detected with ¹²⁵I-protein A or by incubation with appropriate secondary horseradish peroxidase–conjugated antibodies (New England Biolabs, Inc.) using autoradiography or chemiluminescence, respectively.

HeLa cells, 17 h after transfection with PS2 plasmid DNA, were collected in immunoprecipitation buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 0.5% NP-40, 2 mM EDTA) and homogenized by gentle strokes with a Dounce homogenizer. After centrifugation, the pellet was discarded, and the proteins were immunoprecipitated from the remaining supernatant by adding rabbit antisera (preimmune, anti–PS2-NH₂ terminus, or anti–PS2-Loop) and protein A–Sepharose CL-4B beads (Stabler et al. 1999). The beads were recovered by centrifugation and washed with 0.5% NP-40 in 1.0× PBS. They were boiled in sample loading buffer, and the bound proteins were analyzed by SDS-PAGE and immunoblotting. Blots were probed with primary anti–ubiquilin-B antibodies, which were subsequently detected with ¹²⁵I-protein A. Soluble, insoluble, Triton X-100–soluble, and Triton X-100–insoluble protein fractions of untransfected or PS2-transfected HeLa cells were prepared according to the procedure described previously (Stabler et al. 1999). Equal amounts of each fraction were separated by SDS-PAGE and, after electroblot transfer to nitrocellulose filters, were immunoblotted with either affinity-purified anti–ubiquilin-C antibody, anti–PS2-NH₂ terminus antibody, or anti-lamin A/C antibody (Mical and Monteiro 1998). The partitioning of endogenous lamin A/C proteins was monitored, since their cell fractionation behavior was known (Mical and Monteiro 1998; Stabler et al. 1999).

The procedure used for immunofluorescence staining of HeLa cells was described previously (Janicki and Monteiro 1997). Indirect immunofluorescent microscopy, digital image capture, and subsequent manipulation and merging of images were done with a Leica DM IRB microscope attached to a PowerMac computer running Signal Analytics IPLab Spectrum software (Stabler et al. 1999). Confocal images were captured on a Axiovert 100 microscope using LSM software (ZEISS). The primary antibodies used were: rabbit anti–ubiquilin-B and affinity-purified anti–ubiquilin-C, goat anti-PS1 and anti-PS2 (NH₂ terminus specific) (Santa Cruz Biotechnology, Inc.), and mouse anti-myc (Mical and Monteiro 1998).

Hippocampus brain tissue from six AD cases (aged 77–95) and five control cases (aged 19–71) were fixed in methacarn (6:3:1, chloroform/methanol/acetic acid) overnight. Locus coreuleus or substantia nigra brain tissue from three PD cases (aged 53–66) were fixed in formalin overnight, whereas frontal cortex brain tissue from two cases of diffuse Lewy body disease (aged 49 and 85) were fixed in either formalin or methacarn overnight. Samples were dehydrated and embedded in paraffin. Sections were cut 6-μm thick and mounted onto glass slides. After deparaffinization with xylene, sections were hydrated through graded ethanol treatment, and the endogenous peroxidase activity was eliminated by incubation in 3% hydrogen peroxide for 30 min. Nonspecific binding sites were blocked with 10% normal goat serum in Tris-buffered saline (50 mM Tris-HCl, 150 mM NaCl, pH 7.6) for 30 min before application of either preimmune sera, affinity-purified anti–ubiquilin-C antibody, or anti–ubiquilin-B antibody. Immunostaining was performed using the peroxidase–anti-peroxidase method, with DAB as cosubstrate (Sternberger 1986). Color images of the stained slides were captured using a Sony Digital Photo Camera DKC-5000 on a Nikon Eclipse E600 microscope using Adobe Photoshop^® software.

HeLa cells were transfected with expression constructs encoding ubiquilin protein (15 μg plasmid DNA), PS2 protein (7 μg plasmid DNA), or both. At 12 h after transfection, the cells were incubated for an additional 5–6 h in DME/FBS media in the absence or presence of proteasome inhibitors (20 μM synthetic lactacystin or 40 μM MG-132; Calbiochem-Novabiochem). Afterwards, the cells were lysed in buffer containing protease inhibitors and the protein concentrations were determined. Protein extracts totaling 100 μg were loaded per lane, separated by SDS-PAGE, and immunodetected with various antibodies. In a follow up experiment, HeLa cells were cotransfected with a constant amount of PS2 (7 μg plasmid DNA) and increasing amounts of either ubiquilin (0, 1, 2, 3, and 4 μg plasmid DNA) or GFP (0, 1, 2, 3, and 4 μg plasmid DNA) expression vectors.

Exponentially growing HeLa cells maintained in flasks were trypsinized and resuspended at a density of 2.68 × 10⁶ cells/ml in Opti-MEM medium (GIBCO BRL). Two 0.5-ml aliquots of the cell suspension were each electroporated with a mixture of either 7 μg PS2 and 15 μg of EGFP-C1 (CLONTECH Laboratories, Inc.) expression plasmid DNA or 7 μg PS2 and 15 μg of ubiquilin expression plasmid DNA. The aliquots containing the same DNA mixture were then combined, resuspended in 28 ml of OptiMEM medium (containing 10% FBS), and 2 ml of the suspension was plated in 14 separate wells (9.5 cm²). The cells were incubated for 7 h to allow for attachment to the dishes, and then the medium was removed and replaced with DME containing 10% FBS. Incubation was continued for an additional 14 h. The cells were starved for 45 min in methionine-deficient DME containing 10% dialyzed FBS. The medium was removed and the cells in each well were pulse labeled by adding 1 ml of the methionine-deficient medium containing 150 μCi of [³⁵S]methionine (1,000 Ci/mmol; Amersham Pharmacia Biotech) for 1 h. After labeling, the cells were washed twice with DME containing 1 mM nonradioactive l-methionine and 10% dialyzed FBS, and then they were chased with 2 ml of the same medium for 0–6 h. At appropriate time intervals, the cells were washed three times with ice-cold 1.0× PBS and lysed in buffer containing protease inhibitors (Monteiro and Mical 1996). The lysates were diluted 10-fold with immunoprecipitation buffer before PS2 proteins were immunoprecipitated, which was done by the addition of 5 μl rabbit anti-loop PS2 antibody (directed against GST-PS2 loop purified protein) (Janicki and Monteiro 1997) and 100 μl of a slurry of protein A–Sepharose (Amersham Pharmacia Biotech). The immunoprecipitated proteins attached to the Sepharose beads were eluted by incubation with protein sample loading buffer (lacking SDS), which contained 8 M urea, 10 mM DTT, and 10% β-mercaptoethanol (Janicki and Monteiro 1999), at 37°C for 15 min. The entire volume of samples were separated by SDS-PAGE on 8.5% gels. After electrophoresis, the gel was stained with Coomassie blue, impregnated with 2,5-diphenyloxazole, and dried (Stabler et al. 1999). The radioactivity of the immunoprecipitated PS2 band in each lane, corresponding to full-length PS2 protein, was quantified using a Molecular Dynamics Storm 840 Phosphoimager using ImageQuant software (Molecular Dynamics).

The half-life of endogenous HeLa ubiquilin protein was estimated by pulse labeling mock-electroporated HeLa cells for 1 h with 100 μCi of [³⁵S]methionine, followed by a chase with nonradioactive medium for 0–21 h. Ubiquilin protein was immunoprecipitated using 5 μl of rabbit anti–ubiquilin-C antibody, and the amount of radioactivity incorporated in the ubiquilin bands was determined after SDS-PAGE and phosphoimage analysis.

A novel human protein, which we named ubiquilin, for a protein with ubiquitin-related protein domains that interacts with the presenilins, was identified in a yeast two-hybrid screen (Y2H) for proteins that interact with the PS2 COOH-terminal sequence (Fig. 1a and Fig. b). Two overlapping clones encoding the COOH-terminal 203 and 218 residues of ubiquilin were isolated in the screen. To clone the entire ubiquilin ORF, 5′ RACE was performed. This yielded 1,053 bps of additional upstream sequence, including an in-frame methionine codon located at the beginning of a ubiquitin-like (UB) domain. Conceptual translation from this methionine predicted a 559-amino acid protein. However, this methionine was unlikely to be the authentic start codon since translation from it produced a protein that was noticeably smaller, as seen by SDS-PAGE, than endogenous ubiquilin detected using pAb raised to the COOH-terminal 218 residues of ubiquilin (data not shown). Subsequent EST database searches and sequencing of putative overlapping clones yielded a clone with a longer 5′ end, which included an upstream in-frame methionine codon (corresponding to a 595-amino acid protein). Several lines of evidence indicate that this methionine codon was the authentic start methionine of ubiquilin. (a) This start methionine contained a Kozak consensus sequence and an upstream in-frame stop codon. (b) Polypeptides translated from this start codon matched the endogenous ubiquilin protein in size, which was 66 kD in HeLa and other human cell types. (c) Mammalian overexpression of the complete ubiquilin ORF (see Fig. 3, D–F), in vitro transcribed and translated ubiquilin in rabbit reticulocyte lysates (see Fig. 3 G), and human ubiquilin protein synthesized in bacteria (see Fig. 3 H) all generated 66-kD proteins. These data indicated that we had cloned the complete ubiquilin ORF and that the protein is probably not extensively modified in mammalian cells, since bacterially and eukaryotically expressed human ubiquilin proteins have similar molecular masses when separated by SDS-PAGE.

Interaction between ubiquilin and the presenilins was quantified in Y2H β-galactosidase liquid culture assays. A partial ubiquilin clone (encoding the COOH-terminal 218 residues), isolated from the Y2H screen, and a near full-length clone (encoding residues 37–595) both bound to the PS2-COOH-terminal bait >55-fold compared with negative-control baits (Fig. 1C and Fig. D, respectively). However, the longer ubiquilin clone interacted less strongly with the PS1–COOH-terminal bait. This interaction nevertheless was still >12-fold compared with the negative-control baits (Fig. 1 D). We also tested whether ubiquilin would interact with portions of the loop regions of the presenilins (Fig. 1a and Fig. b) that we had used in another Y2H study (Stabler et al. 1999). Interestingly, the near full-length ubiquilin construct also interacted well with both PS2 and PS1 loop sequences, and, surprisingly, the interaction was approximately twice as strong as with the PS2–COOH-terminal bait (Fig. 1 D). This suggested that two separate regions within the presenilin proteins could interact with ubiquilin.

Northern blot analysis of human tissues indicated that ubiquilin mRNA is widely expressed as a major 4.4-kb transcript, with several smaller differentially expressed minor transcripts (Fig. 2 II, X; Fig. 3 A). Comparison of ubiquilin band intensities after correction for mRNA loading (β-actin control bands) and normalized to the skeletal muscle value suggested that expression of the 4.4-kb transcript was highest in brain and pancreas (Fig. 3 C). Additional Northern blot analysis of mRNA from different subregions of the human brain revealed slight variation (1.0–1.5-fold) of ubiquilin expression with the same 4.4-kb transcript as the predominant species (Fig. 2 II, Y; Fig. 3B and Fig. C).

We determined the chromosomal location of the ubiquilin gene using PCR radiation hybrid mapping (see Materials and Methods; data not shown). This procedure resulted in an unambiguous assignment of ubiquilin nearby to two loci, CHLC.GATA22H04 and CHLC.GATA81C04, located on chromosome 9q22 close to the 9q21.3 border. Interestingly, the chromosome region to which ubiquilin maps is thought to contain a susceptibility gene(s) involved in late-onset AD (Kehoe et al. 1999).

The complete ubiquilin ORF consists of 595 residues, with a sequence rich in glutamines and serines, 10.9% and 11.1%, respectively, but lacking any cysteines (Fig. 4). The protein has a calculated isoelectric point (pI) of 4.97 and is predicted to be soluble, based on having a high hydrophilic profile using MacVector 6.5 software (Oxford Molecular Group; data not shown). Currently known protein structural motifs residing within ubiquilin include an NH₂-terminal UB domain and a COOH-terminal ubiquitin-associated (UBA) domain, both of which have been implicated in targeting and degradation of proteins by the ubiquitin-proteasome pathway. The predicted ubiquilin protein also contains numerous regularly spaced asparagine-proline (Asn-Pro) repeats of unknown function (Fig. 4). When first sequenced, no apparent mammalian homologue for ubiquilin had been deposited in GenBank/EMBL/DDBJ databases. However, there was significant homology to several full-length genes of unknown function: the C. elegans F15C11.2 gene (36% identical) and two related genes in Arabidopsis thaliana (34 and 35% identical). In addition, the protein had weak homology to the Saccharomyces cerevisiae DSK2 gene (31% identical), which can suppress a KAR1 defect (Biggins et al. 1996). KAR1 is an essential gene involved in the duplication of the yeast microtubule organizing center known as the spindle pole body. Interestingly, ubiquilin homology to DSK2 was primarily confined to the UB and UBA domains. However, since both UB and UBA domains are present in many other proteins, it is uncertain whether ubiquilin and DSK2 are true orthologs.

At about the time the ubiquilin sequence was deposited in GenBank/EMBL/DDBJ, a Xenopus laevis protein involved in inhibiting cyclin A–protein degradation, called XDRP1, and having 60% identity to ubiquilin had just been submitted (Funakoshi et al. 1999). Since then, two Mus musculus proteins that are thought to interlink integrin-associated proteins with the vimentin cytoskeleton, called PLIC-1 and PLIC-2 (84 and 69% identical, respectively) (Wu et al. 1999), and human Chap1 protein (renamed ubiquilin 2; 79% identical), which binds Hsp70-like Stch and can complement a DSK2 yeast mutant (Kaye et al. 2000), have been described. The high degree of homology among these protein sequences can be seen when aligned by the MacVector 6.5 ClustalW protein alignment program (Fig. 4). This comparison revealed that the human, mouse, and Xenopus sequences are more closely related to one another than the C. elegans, A. thaliana, and S. cerevisiae sequences, as would be expected for higher eukaryotes. The former group of proteins all shared extensive homology across the entire length of the ubiquilin polypeptide. The four regions of highest homology include the UB and UBA domains and two stretches rich in Asn-Pro repeats (residues G157–P326 and P375–P482) located between the two ubiquitin-related domains. Differences in the divergent regions are instructive, especially the insert between residues L485–T527 in Chap1, which is conserved in PLIC-2 but missing in ubiquilin, PLIC-1, and XDRP1. Based on the homology between human and mouse sequences, it appears that ubiquilin is more closely related to PLIC-1, whereas Chap1 is more closely related to PLIC-2. Collectively, the proteins are likely to be members of a related multigene family, since at least two independent genes are known to exist in humans, mice, and A. thaliana.

GST-fusion proteins containing the UBA domain and nine COOH-terminal residues (QHHSSISVS) of ubiquilin were necessary and sufficient to bind [³⁵S]methionine-radiolabeled PS2 and PS1 in a GST pull-down assay (Fig. 5A and Fig. B, respectively). Interestingly, slower migrating forms of both presenilins were much more tightly bound than the full-length forms. Whether the slower migrating forms are more ubiquitinated, extensively modified, or merely hydrophobic aggregates is not known. As expected, GST alone (Fig. 5A and Fig. B, lane L) did not bind either presenilin. No other region of ubiquilin tested was able to pull-down either of the presenilins, indicating that binding of presenilin to ubiquilin is mediated by sequences spanning the UBA domain and the COOH-terminal tail. These in vitro binding assay results were consistent with the interaction data of ubiquilin recovered in the Y2H screen.

To support the in vitro binding data, coimmunoprecipitation and cell fractionation experiments between PS2 and endogenous ubiquilin were also performed. PS2-transfected HeLa cell extracts were immunoprecipitated with two different anti-PS2 specific antibodies, separated by SDS-PAGE, and immunoblotted with anti–ubiquilin-B antibodies. Ubiquilin protein coimmunoprecipitated with both of the anti-PS2 antibodies, one raised against the loop and the other to the NH₂-terminal sequences, but did not coimmunoprecipitate when the preimmune sera was used (Fig. 5 C). Additional verification of ubiquilin binding to PS2 was obtained in cell fractionation studies. Untransfected and PS2-transfected HeLa cells were fractionated either in the absence or presence of Triton X-100 detergent into soluble and insoluble fractions. Equivalent amounts of the fractions were then immunoblotted to determine the amount of ubiquilin protein in the two fractions. In untransfected cells, almost all of the ubiquilin protein was present in the soluble fraction, independent of detergent treatment (Fig. 5 D, lanes S− and S+). This is consistent with the predicted hydrophilic nature of ubiquilin protein. However, in PS2-transfected cells, 30% more ubiquilin protein separated with the pellet fraction in the absence of detergent treatment (Fig. 5 F, lane P). This change in solubility is likely a consequence of ubiquilin binding to overexpressed membrane-bound PS2, which also fractionated in the insoluble membrane-containing pellet fraction (Fig. 5 E, lane P−). As expected for integral membrane proteins, Triton X-100 treatment caused the release of PS2 into the soluble fraction (Fig. 5 E, lane S+), which coincided with ubiquilin once again separating completely in the soluble fraction (Fig. 5 F, lane S+).

To determine the intracellular distribution of ubiquilin we used rabbit pAb that we had generated to two different GST–ubiquilin fusion proteins (see Materials and Methods). Both antisera detected endogenous and overexpressed 66-kD ubiquilin polypeptides, which were not detected by the preimmune sera (Fig. 3 D). Although one antiserum reacted with an unknown protein of ∼55 kD (Fig. 3 E), this reactivity was subsequently removed by affinity purification of the antibody (Fig. 3 F). Immunofluorescence staining of ubiquilin within cells was performed with the specific anti-ubiquilin antibodies. Compared with the preimmune sera (Fig. 6A and Fig. D), both antibodies revealed similar staining patterns (Fig. 6B and Fig. E). Indirect immunofluorescence, as well as laser confocal microscopy, revealed endogenous intracellular localization of ubiquilin in HeLa cells to both the nucleus and cytoplasm (Fig. 6B, Fig. E, and Fig. G). In some cells, endogenous ubiquilin staining was stronger in the nucleus (Fig. 6 E, arrowheads). Furthermore, in a small subset of cells, ubiquilin within the cytoplasm was localized to punctate structures (Fig. 6B and Fig. E, arrows). The frequency of these structures dramatically increased upon overexpression of full-length untagged ubiquilin (Fig. 6C and Fig. F). To corroborate the anti-ubiquilin staining pattern, we performed additional localization studies on myc- and GFP-tagged ubiquilin proteins in HeLa cells (Fig. 2 IV, O and N, respectively). Overexpressed myc-tagged ubiquilin stained with anti-myc mAb exhibited the same intracellular structures (Fig. 6 I) as overexpressed wild-type ubiquilin (Fig. 6C and Fig. F). Finally, the same fluorescent staining pattern was reproduced upon overexpression of GFP-ubiquilin (Fig. 6 H). Since the GFP-ubiquilin image in Fig. 6 H was captured from live and unfixed cells without antibodies, this established that the anti-ubiquilin structures seen previously were not artifacts of the paraformaldehyde fixation or immunofluorescence staining procedures.

As overexpressed presenilin proteins are localized predominantly to the ER, and ubiquilin protein in HeLa cells is found throughout the nucleus and cytoplasm, we determined if ubiquilin colocalized with the presenilins when the latter were overexpressed. Indeed, there was a dramatic change in ubiquilin staining in HeLa cells coexpressing either PS1 or PS2 and ubiquilin, resulting in almost complete colocalization of the presenilin and ubiquilin intracellular staining patterns, as seen by laser confocal immunofluorescence microscopy (Fig. 7A and Fig. B). Careful examination of the colocalized proteins revealed ubiquilin staining to be punctate, whereas presenilin staining could be traced to an ER-like pattern. Particularly striking was colocalization of ubiquilin with irregularly shaped presenilin-enriched structures within the cytoplasm of overexpressing cells. In contrast, there was little colocalization in the nucleus where ubiquilin, but not presenilin, staining could be found.

Since Y2H data suggested that ubiquilin interacts with both the loop and COOH-terminal domains of the presenilins, we examined whether removal of these regions from PS2 would abolish colocalization of the proteins in vivo. To assess this possibility, two PS2 constructs containing progressively longer COOH-terminal deletions were made. In the first, PS2(ΔC), the 39-amino acid COOH-terminal domain was deleted (Fig. 1 A, arrowhead), whereas in the second, PS2(ΔLC), the large hydrophilic loop and all sequences downstream were deleted (Fig. 1 A, arrow). When expressed in HeLa cells, PS2(ΔC) still colocalized with ubiquilin (Fig. 7 C), presumably since the PS2 loop region that remained could bind ubiquilin, as suggested by the strong interaction between the two in Y2H assays (Fig. 1 D). However, once the hydrophilic loop region and COOH terminus were both removed, PS2(ΔLC) no longer colocalized with ubiquilin (Fig. 7 D), suggesting that the NH₂ terminus and the first six transmembrane domains of PS2 do not contain additional ubiquilin binding sites. A notable feature of the overexpression of PS2(ΔLC) was its accumulation into large spherical aggregates that failed to stain positively for ubiquilin.

Since ubiquilin contains multiple ubiquitin-related structural motifs, we investigated whether its association with presenilin would have an effect on presenilin-protein modification and/or stability. Treatment of mock-transfected and PS2-transfected HeLa cells with proteasome inhibitors lactacystin (20 μM) or MG-132 (40 μM), as expected, increased the overall amount of ubiquitinated proteins in cells (Fig. 8 A, lanes 2 and 3, 5 and 6, and 8 and 9 versus lanes 1, 4, and 7, respectively). However, these conditions did not markedly increase PS2 protein levels (Fig. 8 B, lanes 4–6). Remarkably, coexpression of ubiquilin with PS2 caused a substantial (>10-fold) increase in PS2 protein accumulation (Fig. 8 B, lanes 7–9) compared with overexpression of PS2 alone (Fig. 8 B, lanes 4–6). This effect was independent of treatment with lactacystin or MG-132, as it also occurred just as robustly in cells not treated with the proteasome inhibitors. Particularly striking was the increased accumulation in SDS-PAGE gels of full-length uncleaved PS2 protein (Fig. 8 B, arrowhead) and higher molecular weight complexes that PS2 forms. Increasing the amount of transfected ubiquilin plasmid DNA, whereas maintaining constant levels of transfected PS2 plasmid DNA, revealed a dose-dependent effect by ubiquilin on PS2 accumulation within the cells (Fig. 8 E, lanes 1–5). This effect was specific for ubiquilin expression, since cotransfection of increasing amounts of GFP plasmid DNA with PS2 did not increase PS2 protein accumulation (Fig. 8 F, lanes 1–5). The blot in Fig. 8 E was exposed for a shorter time than that in Fig. 8 F to optimally illustrate the subtle changes in PS2 protein levels. In these experiments, we routinely observed the accumulation of full-length PS2 protein, but not the endoproteolytic cleaved forms of the protein (Haass and De Strooper 1999). In fact, recent reports have indicated that cleavage of presenilins is not essential for biological function of the presenilins, namely by its ability to rescue PS activity in C. elegans (Jacobsen et al. 1999; Steiner et al. 1999). The absence of detectable endoproteolytic cleaved forms of PS2 is more clearly illustrated in Fig. 8 G, which shows that an antibody specific for PS2-loop sequences downstream of the proposed endoproteolytic cleavage site recognized the same 54 kD and higher PS2 species (as the PS2 NH₂-terminal specific antibody). Interestingly, though ubiquilin caused a dose-dependent increase in presenilin accumulation, it did not alter the accumulation levels of various other endogenous HeLa proteins, such as lamin B, calreticulin, calnexin, BiP, and α-tubulin (Fig. 8 H). Similar experiments repeated with PS1 indicated that ubiquilin also promoted a dose-dependent increase in PS1 protein accumulation in HeLa cells (data not shown).

To determine if the ubiquilin-induced increase in presenilin accumulation is due to a change in the rate of presenilin protein turnover, we carried out pulse-chase experiments of HeLa cells in which PS2 protein was coexpressed with either ubiquilin or GFP (as a control). PS2 protein in the pulse-chase lysates was immunoprecipitated using an anti–PS2-loop antibody and resolved by SDS-PAGE (Fig. 9 A). Phosphoimage analysis of the immunoprecipitated full-length PS2 band (∼54 kD), indicated that the half-life of PS2 protein to be approximately similar. 3.1 and 2.9 h when coexpressed with GFP or ubiquilin, respectively (Fig. 9A and Fig. B). Apart from the prominent PS2 band at 54 kD, a smear of bands with varying intensity was resolved after SDS-PAGE, using the PS2-loop antibody for immunoprecipitation. Some of these bands appear to be PS2 related, especially the higher molecular weight species, as they corresponded to similar PS2 immunoreactive species using this and other PS2 antibodies (Fig. 8E and Fig. G) (Janicki and Monteiro 1997, Janicki and Monteiro 1999). Phosphoimage analysis of the smear of bands above the 97-kD marker indicated that the rate of turnover of these larger PS2 complexes was also similar in both GFP- and ubiquilin-coexpressing cells (data not shown).

A notable difference of these pulse-labeling experiments was the increased amount of immunoprecipitated PS2-labeled proteins in cells that were overexpressing ubiquilin, compared with GFP. Despite using similar amounts of PS2 DNA for electroporation within parallel-labeling experiments, we routinely obtained a 1.6–2.0-fold increase in PS2-labeled protein when PS2 was coexpressed with ubiquilin compared with GFP. This effect was seen in three separate experiments (data not shown). Overall, these results suggest that the ubiquilin-induced elevation in presenilin accumulation does not involve a substantial change in the rate of presenilin–protein turnover, but instead may facilitate increased PS2 protein synthesis. Finally, to determine the rate of endogenous ubiquilin protein turnover, we performed another pulse-chase experiment of mock-transfected HeLa cells and immunoprecipitated the endogenous ubiquilin protein using our anti–ubiquilin-C antibody (Fig. 9 C). Phosphoimage analysis indicated that pulse labeled ubiquilin decays exceedingly slowly, only 15% over the 21 h chase period, from which we estimate the half-life of the protein to be 90 h.

Since neuropathological lesions in AD and PD have been found to be highly immunoreactive to ubiquitin antibodies, we investigated whether these structures may also contain ubiquilin immunoreactivity. Immunohistochemistry of adult human brain sections was performed with anti–ubiquilin-B and anti–ubiquilin-C antibodies. The former was raised against ubiquilin polypeptides that were devoid of the UB domain, whereas the latter was raised against ubiquilin polypeptides lacking the UBA domain as well (Fig. 2 III, B and C, respectively). Compared with the preimmune sera, which showed no specific staining (Fig. 10A and Fig. C), both antibodies strongly stained neurons with little, if any, staining of glia (Fig. 10B, Fig. D, and Fig. E). In general, the anti–ubiquilin-C antibody stained neurons with greater clarity and intensity. As with cultured cells, anti-ubiquilin staining in neurons was to the nucleus and the cytoplasm. In addition, there is clear evidence for ubiquilin staining within long axonal processes (Fig. 10 D). Interestingly, neurons containing neurofibrillary tangles (NFTs) from AD affected brains seemed to stain more intensely compared with control neurons (Fig. 10 E and F). Moreover, examination of ubiquilin staining in brains afflicted with PD (not shown) and diffuse LB disease revealed robust staining of Lewy bodies (Fig. 10 H).