Puromycin-sensitive aminopeptidase. Sequence analysis, expression, and functional characterization.

Among the molecular mechanisms that control the cell division cycle, proteolysis has emerged as a key regulatory process enabling cells to pass critical check points. Such proteolysis involves a cascade of enzymes including a multisubunit complex termed 26S proteasome. Here we report on the analysis of a novel mouse cDNA encoding the puromycin-sensitive aminopeptidase (PSA) and on its expression in COS cells and 3T3 fibroblasts. PSA is 27-40% homologous to several known Zn(2+)-binding aminopeptidases including aminopeptidase N. Immunohistochemical analysis revealed that PSA is localized to the cytoplasm and to the nucleus and associates with microtubules of the spindle apparatus during mitosis. Furthermore, puromycin and bestatin both arrested the cell cycle, leading to an accumulation of cells in G2/M phase, and ultimately induced cells to undergo apoptosis at concentrations that inhibit PSA. Control experiments including cycloheximide further suggested that the induction of apoptosis by puromycin was not attributable to inhibition of protein synthesis. Taken together, these data favor the novel idea that PSA participates in proteolytic events essential for cell growth and viability.

Under basal metabolic conditions, the turnover of intracellular proteins is regulated mainly by non-lysosomal mechanisms involving 26S proteasomes. These multisubunit complexes are localized to the cytoplasm and to the nucleus (1)(2)(3)(4)(5), and they associate with the mitotic spindle apparatus (4 -6). Polyubiquitinated proteins and some other proteins are degraded by proteasomes in an ATP-dependent manner (7)(8)(9)(10)(11). Interestingly, in yeast, mutations in the proteasome subunit genes mts2, cim-3/SUG1, or cim-5/MSS1 arrest the cell division cycle during G 2 /M phase (12,13) suggesting essential roles for proteasomes during mitosis such as the ubiquitin-dependent degradation of cyclins (14 -16), the latter being required at the onset of anaphase to inactivate maturation promoting factor (17,18) and to complete mitosis (19). Additionally, proteasome-mediated proteolysis promotes sister chromatid segregation (20) and must further persist until early G 1 phase to allow G 1 cyclins to become activated (21). Taken together, these findings suggest that proteolysis is of fundamental importance for cell cycle regulation.
However, the mechanism by which the substrates of proteasomes are degraded into small peptides and amino acids is not yet fully understood. Possibly, serine or cysteine proteases are involved since cyclin degradation in oocyte extracts is inhibited by tosyl-lysine chloromethyl ketone (22). Tests with model peptides further indicated that proteasomes exhibit several distinct hydrolytic activities designated as chymotrypsin-like, trypsin-like, peptidyl-glutamyl peptide-hydrolyzing, branched chain amino acid-preferring, and small neutral amino acidpreferring activities, respectively (23). More recently, a novel proteolytic mechanism was identified in proteasomes of Thermoplasma acidophilum, which involves a threonine residue (24,25). Conservation of this threonine suggests that a similar mechanism may account for some of the activities of eukaryotic proteasomes. Finally, it has been assumed that exopeptidases act distally from endoproteinases to prevent accumulation of peptides and to allow recycling of amino acids (10).
Aminopeptidases hydrolyze N-terminal amino acids of oligopeptides. Their widespread distribution in plant and animal tissues as well as in bacteria and fungi suggest that they play important roles in various biological processes (for review see Ref. 26). The observation that aminopeptidase inhibitors can suppress cell proliferation (27,28) indicated that also cell growth may depend on aminopeptidase activity. However, most of the available data on aminopeptidases derive from studies on the metabolism of neuropeptides (for review see Ref. 29) such as enkephalins, where the Tyr 1 -Gly 2 bond is susceptible to aminopeptidase cleavage (30 -33). However, the aminopeptidase that primarily is relevant for degrading enkephalins and possibly other neuropeptides has not yet been identified unequivocally. A candidate is the ectoenzyme aminopeptidase N, which in brain was detected on synaptic membranes (34), although at a relatively low level. A puromycin-sensitive aminopeptidase (PSA) 1 has also been considered to play a role (31,35,36). PSA is approximately 100-fold more sensitive to puromycin than aminopeptidase N, and it is present in most organs, including brain, where it is most abundant and wherefrom it was purified (35,36). Moreover, PSA efficiently degrades enkephalins in vitro (35), and it is present in brain in much higher amounts than aminopeptidase N. However, a function for PSA in neuropeptide metabolism has been questioned since it is was found to be a cytoplasmic protein (37).
The present paper reports on the cloning and expression of PSA, its subcellular localization, and inhibitor studies that suggest a novel potential function of PSA. PSA is shown to localize both to the nucleus and to the cytoplasm and to associate with the spindle apparatus during mitosis. Finally, we show that a block of the cell cycle by PSA inhibitors is associated with an accumulation of cells in G 2 /M phase and with the induction of DNA fragmentation. These findings are discussed in the context of the novel idea that PSA may be involved in cell cycle regulation.

MATERIALS AND METHODS
Cloning and Sequencing of the Murine PSA cDNA-To isolate the PSA cDNA, a cDNA library from the murine neuroblastoma cell line Neuro2A (ATCC CCL 131) was constructed using a zap cDNA synthesis kit from Stratagene. Of this library, 1.5 ϫ 10 6 plaques were screened at low stringency using an NcoI/SpeI fragment (0.7 kilobases) of a putative human metallopeptidase cDNA that was isolated recently in our laboratory. 2 The inserts of isolated positive plaques were excised in vivo. After generating exonuclease III deletions (Stratagene manual) from the largest of these clones, overlapping sequences of both DNA strands were obtained using a Sequenase kit (U.S. Biochemical Corp.). Data base searches and sequence analysis were performed using the GCG programs FASTA and MOTIFS (Genetics Computer Group Inc., Madison, WI).
Northern Blot Hybridization-Mouse multiple tissue Northern blots were purchased from Clontech (Palo Alto, CA). A PSA probe was synthesized from the full-length mouse cDNA by random primer labeling and hybridized at 42°C overnight in 5 ϫ SSC containing 50% formamide, 50 mM sodium phosphate buffer pH 7.0, 1 ϫ Denhardt solution, and 1% SDS. Filters were washed twice at 65°C in 2 ϫ SSC for 5 min and twice at 62°C in 0.2 ϫ SSC containing 0.1% SDS. Filters were exposed to x-ray films for 2 days at Ϫ70°C.
Transient Expression of PSA in COS-7 Cells-The PSA cDNA was cloned into pSVL (Pharmacia Biotech Inc.) in two steps. First, the blunted NcoI/StuI fragment of the murine PSA cDNA was subcloned into the SmaI site of pSVL (Pharmacia). The resulting plasmid containing the 3Ј end of PSA was cut with XbaI and ligated with the XbaI fragment (5Ј end) of the PSA cDNA. For transfection experiments, COS-7 cells were grown to confluency in Dulbecco's modified Eagle's medium containing 10% fetal calf serum (Life Technologies, Inc.) and passaged 1:4. The following day, 3.9 ϫ 10 6 cells were resuspended in RPMI containing 10 mM dextrose and 0.1 mM dithiothreitol and were mixed with 5-7.5 g of plasmid DNA. Cells were electroporated (0.25 kV, 960 microfarads) in a Bio-Rad cuvette (width 0.4 cm) using a Bio-Rad gene pulser. Following electroporation, the cells were incubated in culture medium for 24 h, washed twice with Hanks' buffered salt solution, and grown for another 48 -72 h in fresh medium to prepare cell extracts.
Aminopeptidase Activity Assay-Subconfluent cell monolayers were washed twice with PBS and then incubated in PBS containing 0.5 mM EDTA to bring the cells into suspension. Floating cells were collected, washed once in PBS, and then incubated on ice for 10 min in 100 l of 10 mM Tris⅐HCl, pH 7.4. To obtain cell-free extracts, the cells were homogenized by ultrasonication, and the resulting lysates were centrifuged for 30 min at 13,000 ϫ g at 4°C. Supernatant was collected, and protein concentrations were determined using the Bio-Rad protein assay and bovine serum albumin (BSA) as standard. Aliquots of the extracts corresponding to 5-20 g of protein were diluted to 250 l with 10 mM Tris⅐HCl, pH 7.4, and reactions were initiated by adding an equal volume of 100 mM Tris⅐HCl, pH 7.4, containing 4 mM aminoacyl pnitroanilide (Sigma), 0.1 mg/ml BSA, and 1 mM dithiothreitol. After incubation at 37°C for 15 min, reactions were stopped by adding 500 l of 0.1 M sodium acetate, pH 4, and absorbance of the liberated pnitroaniline was measured spectrophotometrically at 405 nm. Activities were calculated assuming that 1 unit of enzyme liberates p-nitroaniline at a rate of 1 nmol min Ϫ1 (e ϭ 9500 1 ϫ mol Ϫ1 ϫ cm Ϫ1 ).
Western Blot Analysis-Crude cell extracts were separated by SDSpolyacrylamide gel electrophoresis (20 g of protein/lane) on a 7.5% separating gel and transferred to nitrocellulose membranes (Bio-Rad) in a Bio-Rad mini-trans-blot apparatus. The membranes were blocked overnight at 4°C in 10 mM Tris⅐HCl, pH 7.5, containing 150 mM NaCl, 0.1% Tween 20, 5% skim milk (Life Technologies, Inc.), 2% BSA, and 1% rabbit serum (Sigma). In the same buffer, goat anti-rat PSA antiserum was diluted 1:500 and incubated over night at 4°C. Subsequently, filters were washed 3 times in PBS containing 0.1% BSA (w/v) and 0.05% Tween 20 (v/v) and incubated for 1 h with biotinylated rabbit anti-goat IgG (DAKO, Glostrup, Denmark) diluted 1:10 4 in blocking solution containing 0.5% mouse serum. After washing the filters 3 times with PBS, bands were visualized using the Vectastain ABC-AP kit from Vector Laboratories (Burlingame, CA).
Immunocytochemistry-2.5 ϫ 10 4 cells/well were plated in LabTek chambers (Nunc, Naperville, IL). After 24 h, cells were washed in PBS and fixed for 15 min in a mixture of acetone and ethanol (1:1), which was precooled to Ϫ20°C. After rehydrating the cells in PBS for 5 min, the slides were transferred to a moist chamber and incubated for 30 min in blocking solution consisting of PBS, 1% BSA (w/v), and 2% rat serum (v/v) (Sigma). Goat anti-rat PSA IgG (37) was diluted in blocking solution to a concentration of 40 g/ml and preabsorbed for 1 h with either 10 g/ml rat aminopeptidase N or rat PSA (37). Subsequently, cells were incubated either with untreated or preabsorbed goat anti-rat PSA or with preimmune goat IgG (Sigma). After 45 min, cells were washed 3 ϫ in PBS and then incubated for 30 min with a sheep anti-goat Fab fragment coupled to alkaline phosphatase (Boehringer Mannheim), which was diluted 1:80 in blocking solution. Alkaline phosphatase reaction was developed for 20 -30 min according to the instructions of Boehringer Mannheim. To stop the staining reaction, cells were washed with PBS and mounted in Mowiol. Immunocytochemical analysis of detergent-extracted cells was performed as above, except that cells were extracted with detergent and fixed in PBS containing 3.7% formaldehyde (w/v) prior to staining (38). Briefly, cells were washed for 10 s in 0.08 M PIPES, pH 6.9, containing 10 mM EGTA, 1 mM MgCl 2 , 0.1 mM GTP, and 4% polyethylene glycol. Subsequently, the cells were transferred for 1.5 min to the same buffer containing 0.5% Triton X-100, followed by a 30-s wash in extraction buffer and fixation for 20 min at 25°C.
Cell Cycle Analysis-COS-7 cells were left untreated or were stimulated with bestatin (100 g/ml), puromycin (5 M), or VP16 (80 g/ml) for 14 h at 37°C. Subsequently, cells were trypsinized, fixed with 4% formaldehyde in PBS for 10 min on ice, lysed with 0.1% Triton X-100 in PBS for 10 min on ice, and stained with propidium iodide (50 g/ml) (39). Cells were analyzed by flow cytometry on an Epics analyzer, and Multicycle was used for cell cycle analysis.
Quantification of the Synthesis of DNA and Protein-To assess the effect of various aminopeptidase inhibitors on cell proliferation, COS-7 cells were plated in 96-well microtiter plates at a density of 5 ϫ 10 3 cells/well, each well containing 200 l of medium with or without aminopeptidase inhibitors. After incubating the cells for 24 h at 37°C, they were pulsed for another 16 h with [ 3 H]thymidine (1 Ci/well), and the incorporated radioactivity was measured using a liquid scintillation counter. Values represent means of triplicate measurements.
To quantitate protein synthesis, COS-7 cells were washed twice with PBS, resuspended at a density of 10 6 ml Ϫ1 in leucine-free RPMI medium containing 5% fetal calf serum, and mixed in aliquots of 0.5 ml with equal volumes of the same medium containing puromycin or cycloheximide at various concentrations. After preincubating these suspensions at 37°C for 45 min, 5 Ci of [ 3 H]leucine (144 Ci/mmol; Dupont NEN) was added to each sample for 30 min. Incorporated radioactivity was precipitated by adding 1 ml of ice-cold 10% trichloroacetic acid followed by incubation on ice for 30 min. Precipitates were collected on Whatman GF/C filters and washed 3 times with 3 ml of 5% trichloroacetic acid using a vacuum manifold (Millipore Corp., Bedford, MA). After air-drying the filters, bound radioactivity was quantitated by scintillation counting. Values were corrected by subtracting background (7-8 ϫ 10 3 cpm) as determined in the absence of cells. Values represent means of duplicate measurements. Stock solutions of bestatin (25 mg/ml), puromycin (25 mM), and cycloheximide (10 mg/ml) were prepared in deionized water. All inhibitors were purchased from Sigma.
Analysis of DNA Fragmentation-After incubating COS cells for the indicated time in the presence or absence of aminopeptidase inhibitors, cells were permeabilized with detergent as described previously (40,41) to extract fragmented but not intact DNA. Such cell extracts were extracted with phenol, and subsequently DNA was precipitated with EtOH and analyzed on 2% agarose gels after incubation of the samples for 30 min with RNase A (1 g/ml). In situ DNA end labeling of COS cells was performed in microtiter plates using terminal transferase to incorporate biotinylated deoxyuridine triphosphate (50 M), and streptavidin-alkaline phosphatase (all from Boehringer Mannheim) for detecting incorporation.

RESULTS
PSA Sequence Analysis-A fragment of a novel human cDNA encoding for a putative metallopeptidase 2 was used to screen a cDNA library derived from the mouse neuroblastoma cell line Neuro2A. Five clones were isolated, one being derived most likely from an incompletely spliced mRNA (not shown). All of the four remaining clones contained inserts of similar size (3 kilobases) and encoded an open reading frame of 920 amino acids (Fig. 1). The deduced protein sequence was 27-40% identical to that of several known aminopeptidases (Table I). The similarity was most significant in a core region (residues 150 -510) containing the motif HEXXH(X) 18  In the open reading frame (ORF), two potential translational start sites were identified at nucleotide positions 106 and 241, respectively, the former being immediately preceded by a stop codon (Fig. 1). Translational start at the first AUG is expected to give rise to a protein containing a hydrophobic N-terminal sequence of 26 residues, which is interrupted after 10 amino acids by two arginines. Separated from these two arginines by a spacer of 19 amino acids, a stretch of five additional basic residues is located further downstream. Together, these basic residues form a potential bipartite nuclear localization signal (NLS) (Fig. 1). Interestingly, four out of five amino acids at positions 727-731 are also basic in nature and thus may represent a second NLS. In addition to these two NLSs, one potential complementary NLS was identified at positions 111-115 consisting of a stretch of negatively charged residues. Furthermore, nine potential phosphorylation sites for casein kinase II and one for cAMP/cGMP-dependent kinase were identified, and two potential N-glycosylation sites are at positions 63 and 649, respectively (Fig. 1).
Further analyses of the predicted amino acid sequence revealed that the residues 269 -285 (termed PSA I) and 683-701 (PSA II) were 72 and 40% similar, respectively, to the proteasome motif described by Zwickl et al. (43) (Fig. 2b). In addition, these sequences were found to resemble also the microtubule binding sites in tau protein and two other microtubule binding proteins, MAP-2 and MAP-4 (44 -47) (Fig. 2a). Note that within PSA I and II, the only residues that are not similar to either the proteasome motif or the microtubule binding sites or to both are the two glutamates and the second proline residue within PSA II.
Interestingly, at positions 47-64 of the ORF the residues were almost identical to the N-terminal sequence of the purified rat puromycin-sensitive aminopeptidase (37). The only difference observed in the mouse sequence was Ala 57 , which is a threonine in position 9 of rat PSA (Fig. 2c). In addition, at position 46 of the mouse ORF, a methionine is present, which on the mRNA level is flanked by a Kozak consensus-like sequence. This suggests that translation may be initiated at the second AUG of the ORF, giving rise to a protein with an N terminus corresponding to that of rat PSA. That the mouse cDNA is in fact the homologue of the rat PSA became clearly evident from a comparison of the mouse cDNA sequence to the sequence of nine tryptic peptides isolated from rat PSA. As shown in Table II, there is only one amino acid difference among the nine tryptic peptides containing 118 amino acids (His 840 in the mouse sequence is Tyr in the rat sequence). These data, coupled with the N-terminal sequence data, indicate a greater than 98% sequence identity between the rat and mouse PSA.
Western Blot Analysis of PSA-transfected COS Cells-To further substantiate that the present cDNA encodes mouse PSA, it was transfected into COS-7 cells by electroporation and expressed under the control of an SV40 promoter using the expression vector pSVL. Crude extracts of cells transfected either with pSVL-PSA or with vector alone were subjected to Western blot analysis. Using an antiserum against purified rat PSA (37), a 100-kDa band was identified in PSA-transfected cells and in control extracts (Fig. 3). The intensity of the band was clearly higher in the PSA-transfectants as compared with control extracts, indicating its induction. This was consistent with data obtained by Northern blot analysis showing that PSA mRNA is expressed in control cells and strongly induced in PSA-transfected cells (data not shown).
Induction of Aminopeptidase Activity in Transfected COS Cells-To prove that our cDNA encodes for a peptidase, crude Residues labeled with open squares are part of a potential complementary nuclear localization signal. The sequences that resemble both known microtubule-binding sites and a motif conserved among 26S proteasome subunits (for details see Fig. 3) are underlined. q, potential N-glycosylation sites; ϳ, potential casein kinase phosphorylation sites; triangle, potential phosphorylation site for cAMP/cGMP-dependent kinase.
cell extracts of transfected COS cells were assayed for aminopeptidase activity using various synthetic aminoacyl p-nitroanilides as substrates. Depending on the nature of the aminoacyl residue, aminopeptidase activity was induced 10 -20-fold in crude extracts of transfected cells as compared with mocktransfected cells (Fig. 4). In contrast, no significant induction of aminopeptidase activity was detected in the culture medium (data not shown). Moreover, of the substrates examined, methionyl-and leucyl-p-nitroanilide were most efficiently metabolized, followed by lysyl-and alanyl-p-nitroanilide as reported for rat PSA. Preincubating COS cell extracts either with puromycin or with PSA antiserum or with the aminopeptidase inhibitor bestatin resulted in profound inhibition of Leu-pnitroanilide peptidase activity (Table III). Inhibition was ob-served also in the presence of metal chelators such as EDTA and phenanthroline (Table III). In contrast, pepstatin, leupeptin, and phosphoramidon had only minor or no significant inhibitory effects. Taken together, these data clearly demon-  PSA  40  38  27  31  34  33  YSCII  70  28  34  32  32  Yeast AP1  28  40  30  30  Rat AP-N  23  33  33  Lacto-Ala-AP  25  25  Human AP-A  79 FIG. 2. Sequences within PSA align either with known microtubule-binding sites and with a proteasome motif or with the N-terminal amino acids of purified rat PSA, respectively. a, alignment of the sequence motifs PSA I and II to the microtubulebinding sites of tau protein and microtubule-associated protein-2 and -4. b, alignment of PSA I and II to the proteasome motif sequences of multiple proteasome subunits. c, alignment of PSA residues 47-64 to the N-terminal sequence of rat PSA (37). Note that mouse PSA differs only in a single residue (Ala instead of Thr) from the N terminus of the purified rat PSA.  strate that the present cDNA encodes the puromycin-sensitive aminopeptidase.
Tissue Distribution of PSA-A multiple tissue Northern blot was hybridized with a PSA cDNA probe to identify the organs expressing PSA. As shown in Fig. 5, a single PSA transcript of 4.5 kilobases was detectable in all organs examined, suggesting that its expression is widely distributed. Note that PSA mRNA was more abundant in the brain than in all other organs examined. This is consistent with the previously proposed hypothesis that PSA may act as a regulator of neuropeptide activity.
Subcellular Localization of PSA-The hypothesis that PSA is involved in neuropeptide metabolism at postsynaptic membranes, however, has been questioned since PSA has been shown to occur predominantly in a soluble cytoplasmic form. Only in the brain, a significant fraction of the protein is associated with membranes, and cell fractionation studies indicated that also the membrane-associated PSA is mainly an intracellular protein (37). Therefore, determining the subcellular sites of PSA expression should be helpful in understanding its functions. Using the IgG fraction of an antiserum against purified rat PSA, immunocytochemistry was performed in COS cells transfected either with the PSA expression vector or with the vector alone. As shown in Fig. 6a, PSA-transfected cells stained strongly positive. The staining was most concentrated in a perinuclear region, where the nuclei usually were flattened. Such flattening of the nucleus defines a cellular polarity and indicates the site where the cytoplasmic microtubule complex originates and where the Golgi complex is anchored to the cytoskeleton. Note that transient transfection with PSA significantly increased the intensity of staining in this perinuclear region as compared with cells that resisted transfection (Fig.  6a) or as compared with control transfectants (data not shown). Besides the cytoplasmic microtubule complex/Golgi compartment, cellular processes frequently were stained (data not shown). Interestingly, in mitotic cells the spindle apparatus also stained positive for PSA (not shown), and this staining was resistant to extraction of the cells with detergent (Fig. 6, c-e) when performed in a microtubule-stabilizing buffer according to the protocol of Deery et al. (38). Note that already during prophase PSA associated with the microtubules of arising as-ters and with the microtubule organizing centers (Fig. 6, c and  d). Some PSA-transfected cells also exhibited nuclear staining, which sometimes was concentrated in large vesicular inclusions but usually was distributed throughout the nucleus (data not shown). In contrast, no profound nuclear staining was observed in control-transfected or nontransfected cells. The

FIG. 6. Immunocytochemical localization of PSA in COS cells.
a, transfected cells (arrows) are readily identified by increased PSA staining of the Golgi-cytoplasmic microtubule complex region as compared with cells resisting transfection (arrowheads). Fixation was performed in a mixture of acetone and ethanol. b, inhibition of PSA staining by preabsorbing the anti-PSA IgG with purified rat PSA. c, in cells entering M phase of the cell cycle, endogenous PSA is detected in association with arising asters. d, the same cell as in panel c is shown under Nomarsky optics to reveal the presence of microtubules and to identify the two most intensely stained spots as microtubule organizing centers (arrows). e, in metaphase cells, PSA is associated with the spindle apparatus. In panels c-e, the cells were extracted with detergent prior to fixation in formalin in order to substantiate that the association of PSA and microtubules is specific. Note that some unidentified microsomes also stained positive for PSA even after extraction by detergent (arrows in panel e). Size bars, 20 m. pattern of staining was specific since it was efficiently inhibited by an excess of purified rat PSA (Fig. 6b) but not by rat aminopeptidase N (not shown). Likewise, no background staining was detectable if the primary antibody was replaced by a preimmune goat IgG fraction (data not shown).
The finding that in COS cells the subcellular localization of PSA was heterogeneous led to the hypothesis that it may be regulated in a cell cycle-specific manner. To further evaluate this possibility, endogenous PSA was localized within cells of the nontransformed murine fibroblast cell line Swiss 3T3. Like all other cell lines examined so far, 3T3 fibroblasts express PSA mRNA (data not shown). As shown in Fig. 7, 3T3 cells exhibited similar staining patterns as observed in untransfected COS cells, a difference being the nuclear staining observed in a   7. Immunocytochemical localization of endogenous PSA in Swiss 3T3 fibroblasts. a, exponentially growing cells exhibiting cytoplasmic (arrowhead) and, in a subpopulation, also a punctate nuclear staining (boldface arrows). Most intense staining is seen in M phase cells (open arrows). The bridge connecting cells during cytokinesis is also stained (thin arrow). b, background staining after preabsorption of the anti-PSA IgG with purified rat PSA. c, some cells exhibiting nuclear staining appeared to be in prophase as based on signs of nuclear membrane disintegration and chromatin condensation (arrowhead). In a neighboring early prophase cell (short arrow), nuclear staining is concentrated in the center of a growing aster (long arrow). d, during metaphase, PSA is detected in association with the spindle apparatus and in the cytoplasm. e, PSA staining during anaphase. f, late telophase cells (arrowheads) undergoing cytokinesis. The bridge is out of focus. Size bars, 20 m. subpopulation of 3T3 cells (Fig. 7a). Such nuclear staining appeared to be associated with cells that were in prophase of the cell cycle (Fig. 7c). During metaphase, PSA staining was associated with the spindle apparatus (Fig. 7d), and it became concentrated around the chromosomes during anaphase (Fig.  7e). Most intense staining was observed in the newly formed nuclei and in the cytoplasm during telophase (Fig. 7f). As in COS cells (Fig. 6, c-e), in 3T3 cells the microtubule-associated staining resisted extraction with detergent (not shown). Taken together, these data support the conclusion that PSA binds specifically to mitotic spindles and may play a role in proteolysis during mitosis.
Inhibitors of PSA Arrest the Cell Cycle during G 2 /M Phase-To examine whether PSA activity may be essential during mitosis, the DNA content of COS cells was assessed by flow cytometry after treating the cells for 14 h with either puromycin or the aminopeptidase inhibitor bestatin or with the etoposide VP16, which reacts with topoisomerase II and is known to induce a G 2 block (42). As shown in Fig. 8, both puromycin and bestatin lead to a significant accumulation of cells in G 2 /M phase, the effect being comparable with or even exceeding that of VP16, respectively. This suggests that puromycin and bestatin inhibit the cell cycle during G 2 or M phase, or both.

Inhibition of Proliferation by Puromycin and Bestatin but Not by Cycloheximide Is Associated with Induction of DNA
Fragmentation-Since puromycin, besides inhibiting PSA, is known also to interfere with protein synthesis, its effect was compared to those of cycloheximide, another inhibitor of protein synthesis, and of bestatin, using a DNA synthesis assay as a measure for cell proliferation. As expected, both puromycin and cycloheximide as well as bestatin dose-dependently inhibited proliferation (Fig. 9a). However, unlike cycloheximide, puromycin significantly inhibited proliferation even at concentrations that did not detectably inhibit protein synthesis (Ͻ1 M) (Fig. 9b), and the ID 50 of puromycin was comparable with that required for half-maximal inhibition of PSA activity (1 M; Table III).
Interestingly, cells that were treated with puromycin or bestatin showed the morphological characteristics of apoptosis such as chromatin condensation around the margin of the nuclei, membrane blebbing and DNA fragmentation (Fig. 10, B and C). Fragmentation of DNA was assessed by in situ DNA end-labeling in COS cells that were treated with bestatin (100 g/ml) or puromycin (5 M); DNA breaks were detected in 76 (Ϯ5)% and 81 (Ϯ3)% of the cells, respectively. This contrasts with cycloheximide (5 g/ml), which ultimately leads to cell death without inducing more than 3% of the cells to undergo DNA fragmentation. These data were confirmed by gel electrophoretic analysis showing significant DNA fragmentation in puromycin-and bestatin-treated cells but not in cells exposed to cycloheximide (Fig. 10E). These findings indicate that aminopeptidase inhibitors can induce cells to die via apoptosis. DISCUSSION Recently, we have isolated from a human fetal brain cDNA library a novel cDNA that was assumed to encode for a Zn 2ϩbinding aminopeptidase as based on sequence information. 2 To allow subsequent characterization of this novel putative aminopeptidase in the mouse, the corresponding murine cDNA was isolated and sequenced. Here we report that the deduced protein sequence is similar to known aminopeptidases such as aminopeptidase N, this similarity being most concentrated in a region that contains the sequence motif HEXXHX 18 E. Similar motifs have been identified in several metallopeptidases where they coordinate a zinc ion at the active site (48). Thus, the HELAH sequence encoded by our novel clone presumably has the same function. Interestingly, both the N terminus and a number of tryptic peptides of rat PSA were essentially identical to this clone. Therefore, the novel clone was concluded to encode the murine homologue of rat PSA. Consistently, the mouse protein cross-reacted with an anti-rat PSA antiserum if expressed in COS cells, and it had the expected molecular mass of 100 kDa. However, the mouse ORF appeared to be extended by 46 amino acids as compared with purified rat PSA (37). This indicates either that rat PSA is derived from a precursor or that its N terminus results from translational initiation at an internal AUG. In fact, translational initiation at the second AUG in the mouse ORF would give rise to an N terminus corresponding precisely to that of rat PSA. The observation that only the second but not the first AUG codon of the ORF is flanked by a Kozak consensus-like sequence further supports the idea that translation at least under certain conditions is initiated preferentially at the second AUG. It is unlikely that the N terminus of rat PSA is generated by cleavage of a precursor since the in vitro translated mouse protein is not susceptible to cleavage by microsomal membranes. 3 Likewise, the finding that the N-terminal extension is conserved also in the 3 D. B. Constam, unpublished observation. human ORF 2 suggests that it does not simply reflect a species difference.
Two potential NLS are present in the PSA sequence, which may account for the observed nuclear localization of PSA in a subpopulation of cells. Furthermore, a potential complementary NLS was identified. In proteasomes, such complementary NLS have been speculated to account for the inhibition of nuclear translocation since several proteasome subunits harbor NLS and cNLS motifs in cis (49). According to this model, tyrosine phosphorylation of proteasome subunits may cause a conformational change, which in turn disrupts NLS-cNLS interactions and allows proteasomes to be translocated to the nucleus (49). Interestingly, PSA also harbors several potential phosphorylation sites. It will be important to determine whether PSA undergoes phosphorylation and whether this is a mechanism to regulate its subcellular distribution. Also, it should be addressed whether the localization of PSA within a cell may be regulated by alternative translational initiation since one of the putative NLS resides at the N terminus, which apparently is not necessarily translated.
With respect to the potential function of PSA, two other motifs, termed PSA I and II, appeared to be interesting. PSA I and II show significant similarity to the microtubule binding sites of MAP-2, MAP-4, and tau protein (44 -47) as well as to a sequence that is highly conserved among ␣-type proteasome subunits of various species (43). PSA I more closely resembled the proteasome motif than did PSA II, whereas PSA II Cterminally contained the residues PGEG, which aligned better to microtubule binding sites than did the corresponding resi-dues of PSA I. Given these similarities with microtubule binding sites, PSA I and PSA II are candidates to mediate the observed association of PSA with microtubules. Likewise, it is tempting to speculate, whether PSA may functionally interact with proteasomes given that PSA I and II resemble a proteasome motif that is thought to play a role in the assembly of the proteasome multisubunit complex. Subcellular localization of PSA by immunocytochemistry in COS-7 and in Swiss 3T3 fibroblasts further revealed several interesting features. First, in both cell types the staining of individual cells was heterogeneous, some cells showing PSA staining exclusively in a cytoplasmic compartment where the origin of the cytoplasmic microtubule complex and the Golgi complex are localized, others exhibiting nuclear staining as well. A fraction of those cells showing nuclear staining were identified to be in prophase of the cell cycle as based on chromatin condensation and breakdown of the nuclear membrane and on the identification of newly arising asters. Furthermore, PSA was found to associate with the spindle apparatus during mitosis and remained detectable at a high level in mitotic and early G 1 cells. Note that microtubule-bound PSA was detectable even after extracting soluble PSA by detergent. In contrast to microtubule-associated PSA, cytoplasmic PSA was eluted almost completely by extraction with detergent, the residual staining being associated with some unidentified vesicular structures. This observation is consistent with the previous prediction that PSA may belong to a class of amphitropic proteins that exist both in soluble and membrane-associated forms (37). Interestingly, 26S proteasomes, like PSA, have been lo- calized to the cytoplasm and to the nucleus and to the spindle apparatus in a variety of cell types and by several independent laboratories (1)(2)(3)(4)(5). This is consistent with the idea that PSA may contribute to or act downstream of proteolytic processes known to involve proteasomes.
In an attempt to assess whether PSA may be essential for cell division, COS cells that had been treated with the PSA inhibitors puromycin or bestatin were analyzed by flow cytometry for their content of DNA. A significant accumulation of cells in G 2 /M phase was observed within hours, indicating that puromycin-sensitive and aminopeptidase activities are required to complete mitosis. The idea that PSA is involved in cell cycle-regulating proteolysis is consistent with previous reports demonstrating that the aminopeptidase inhibitor bestatin, at concentrations similar to those used in the present study, inhibits the proliferation of primary hepatocytes (27,28). However, according to these authors, bestatin blocked primary hepatocytes reversibly at the G 1 /S transition upon their release from serum starvation, and cell death was not reported. In contrast, we observed that growth inhibition by bestatin is associated with profound and rapid induction of apoptosis. A possible explanation for this discrepancy is that at G 1 /S, additional aminopeptidase activities may be required that are sensitive to bestatin but not to puromycin. Support for this idea arises from the fact that bestatin, unlike puromycin, significantly reduced the number of cells in S phase also in our asynchronously growing cultures. Given the need for relatively high concentrations of bestatin, though, we cannot fully rule out the possibility that some nonspecific toxicity may account for this reduction. On the other hand, the fact that bestatin like puromycin induced a block preferentially during G 2 /M phase argues that bestatin does act specifically and that pharmacological concentrations of bestatin may rather be required due to its limited cellular uptake.
The observation that puromycin, but not cycloheximide, induced DNA fragmentation similar to that observed in the presence of bestatin further suggests that bestatin and puromycin may both affect cell viability by specifically inhibiting PSA activity or a similar aminopeptidase. Moreover, the finding that puromycin induces apoptosis is consistent with previous results reported by Davidoff and Mendelow (50,51) for a human erythroid leukemia cell line. Although these authors did not discuss the possibility that puromycin-induced apoptosis might result from inhibition of PSA activity, this view is supported by the present data showing that the half-maximal effective dose of puromycin is on the order of 1 M both for suppressing DNA synthesis and inhibiting PSA activity, whereas no detectable inhibition of protein synthesis is observed at 1 M puromycin.
An interesting but maybe too simple explanation for the common effects of puromycin and bestatin to induce apoptosis would be that both of these inhibitors interfere with the proteolytic degradation of some important cell cycle regulators. For example, inhibition of cyclin degradation may result in excessive activation of the kinase p34cdc2, which in turn can induce apoptosis (52). Likewise, accumulation of ornithine decarboxylase, which is normally degraded by proteasomes, can lead to the induction of apoptosis (53). Alternatively, inhibition of PSA-like activity may result in the accumulation of a peptide that by itself may become toxic above a certain threshold concentration.