Introduction

Ubiquitination alters stability, half-life, localization, activity, and conformation of proteins (Hussain et al. 2009). Therefore, ubiquitination and the removal of ubiquitin (deubiquitination) are likely to have important consequences in cells. The mammalian genome is thought to encode nearly 100 deubiquitinating enzymes (DUBs) of five related classes, one of which is the ubiquitin-specific proteases (USPs) (Nijman et al. 2005). One of the uncharacterized members of USPs is USP32, on 17q23, which harbors the peptidase signatures C-terminal aspargine, cysteine, histidine, and aspartic acid domains of USPs, suggesting that it plays a role in deubiquitination. This functional role, however, has not yet been confirmed.

USP32 is an ancient and highly conserved gene (Paulding et al. 2003) which, during evolution, gave rise to the hominoid-specific USP6 proto-oncogene (aka TRE2) on 17p13. In fact, USP6 evolved from the chimeric fusion of two genes, USP32 and TBC1D3, indicated by phylogenetic analysis (Paulding et al. 2003). Therefore, USP6 mRNA nucleotides 1–3193 have 89% sequence identity to TBC1D3 (TBC1 domain family, member 3) that functions in Rab GTPase signaling (Wiemann et al. 2001) and nucleotides 3194–6063 have 97% sequence identity to USP32. Interestingly, USP6 has also been indicated as a translocation partner with strong promoters or as part of chimeric transcripts in cancer cells (Oliveira et al. 2004, 2006; Panagopoulos et al. 2008). In addition to USP6, other members of DUBs have also been indicated in cancer-related pathways, such as USP10 deubiquitinating p53 and reversing the Mdm2-induced p53 nuclear exportation and degradation (Yuan et al. 2010). An interesting role for USP33 has recently been described: a Robo1 receptor binding DUB and causing its relocation to the plasma membrane in response to Slit and taking part in the inhibition of breast cancer cell migration (Yuasa-Kawada et al. 2009).

In addition, the USP32 gene maps to 17q23, which is amplified in breast cancers as well as other tumors (Haverty et al. 2008; Pärssinen et al. 2007; Sinclair et al. 2003). The structure of the breast cancer-associated amplicon is complex, with discontinuous regions of genomic amplification that cover several genes (Bärlund et al. 1997; Couch et al. 1999; Erson et al. 2001; Jonsson et al. 2005). Some members of the amplicon have already been extensively examined. Recently, a high-resolution study identified a discrete 17q23 region (Chr 17: 55,503–57,374 Kb) that is amplified in 20% of the HER2+ and 8% of the luminal breast tumors (Natrajan et al. 2009). This region harbors CA4, C17orf64, APPBP2, PPM1D, BCAS3, BRIP1, and USP32. In addition, an 81-gene copy-number signature to predict the metastatic capability of breast cancers identified USP32 (among other 17q23 genes) as a gene that has an increased copy number in the estrogen receptor (ER)-positive tumors (Zhang et al. 2009). USP32 was also reported as one of the upregulated transcripts in malignant breast epithelium compared to normal luminal epithelial cells (Grigoriadis et al. 2006). This suggests that upregulation of USP32 in mammary epithelial cells may be important in the pathogenesis of breast cancer and/or serve as a useful biomarker in breast cancer cells.

Given the fact that USP32 resides on an amplicon region important in breast cancers and has high sequence similarity to a known oncogene, we sought to better characterize USP32 to begin to gain insight into its potential role in normal and neoplastic cells.

Materials and methods

Tissue cDNA samples and cell lines

Human normal multitissue cDNA panel (Clontech), human normal breast total RNA (50 μg) (Ambion), and human breast cancer I tissue PCR array (Origene) were used. BT20, CAL51, MDA-MB231, MCF7, BT474, DU4475, SK-BR3, and HeLa cell lines were obtained from ATCC (Manassas, VA, USA). EFM-19, JIMT-1, HCC1937, HCC1143, HDQ-P1, and CAL85-1 cell lines were obtained from DSMZ (Braunschweig, Germany) and grown under recommended conditions. SUM-225, SUM-159, SUM-149, SUM-185, and SUM-1315 cell lines and the human papilloma virus (HPV)-immortalized nontumorigenic mammary cell line HPV11-21 were developed at the University of Michigan (Ethier et al. 1993, 1996; Ignatoski and Ethier 1999).

Expression analysis

RNA isolation and DNase treatment were performed as described (Erson et al. 2001). cDNA samples were synthesized with 1 μg DNase-treated RNA using RevertAid First Strand cDNA Kit (Fermentas). Various primer sets were used for USP32 and USP6 (full length = 210 and 213) RT-PCR and USP32 mutation analysis primer sequences are available upon request. For real-time RT-PCR (qRT-PCR) SYBR® Green Master (Roche) was used. USP32 (196 bp) F: 5′-CTGCAAGCAGGACACAACTGGTTT-3′, R: 5′-TCACGTAACTGAGGCTGCTTCCAA-3′ and B2M (270 bp) F: 5′-CCAGCAGAGAATGGAAAGTC-3′, R: 5′-CCTCATGATGCTGCTTACA-3′ primers were used. USP32 expression levels were calculated using the ΔΔCt method as previously described (Livak and Schmittgen 2001). The fold change for USP32 was normalized against the housekeeping control (B2M) and compared to normal breast tissue. For the already normalized primary breast cancer tissue PCR array, the ΔCt method was used (Livak and Schmittgen 2001).

In vivo deubiquitination assay

Ubiquitin-met-β-galactosidase (Ub-β-gal) in a pACYC184-based plasmid (a gift from Dr. M. Hochstrasser) and pGEX-4T-USP32 constructs were cotransformed into DH5α cells to assess deubiquitination activity as previously described (Papa and Hochstrasser 1993). DH5α cells were grown in 100 μg/ml ampicillin and 10 μg/ml chloramphenicol containing 2XYTA medium until A600 reached 0.6-0.8. Isopropyl-1-thio-β-d-galactopyranoside (IPTG) was then added to a final concentration of 0.2 mM for 4 h to induce USP32 expression from the pGEX-4T construct. Cells were then lysed in phosphate-buffered saline (PBS) containing 100 μg/ml lysozyme, protease inhibitor (Roche Complete Mini Protease Inhibitor Cocktail Tablet in 10 ml 1 × PBS buffer), and 10 U/ml DNase I, followed by multiple freeze/thaw cycles. The following antibodies were used for immunoblotting: monoclonal mouse anti-β-galactosidase antibody (1:1000; Cell Signaling) and goat anti-mouse IgG-HRP (1:2000; Santa Cruz Technologies). pGEX-Ubp3 (a gift from Dr. M. Hochstrasser) was used as positive control for DUB activity (Baker et al. 1992).

Fluorescence protection assay and subcellular localization

The full-length USP32 coding sequence (4.8 Kb, NM_032582) was PCR amplified with XhoI and Apa1 restriction enzyme site containing primers F: 5′-CCGCTCGAGATGGGTG CCAAGGAGTCAC-3′ and R: 5′-GGGCCCGCTGTAACACACAGTACTTTTTGTAATCAG-3′ and cloned into pEGFPN1 (Clontech). After sequence confirmation, HeLa cells were transfected with 2 μg USP32-GFP construct (USP32 in pEGFPN1) and 2 μg control pEGFPN1 in 6-well plates, using Fugene 6 (Roche) according to the manufacturer’s instructions. Twenty to 24 h post transfection, cells were washed three times for 1 min in KHM buffer (110 mM CH3COOK, 20 mM HEPES, and 2 mM MgCl2) at room temperature and were incubated with 20 μM digitonin for permeabilization (Lorenz et al. 2006). Pre and post-permeabilization images (20×) were captured and recorded using Zeiss LSM 510 (Central Lab, METU). To stain Golgi, BODIPY-TR (Molecular Probes) was used. Transfected cells were incubated with 5 μM BODIPY-TR in PBS/HEPES for 30 min at 4°C followed by washing steps. Cells were again incubated for 30 min at 37°C in fresh medium. After a final wash with the KHM buffer, cells were analyzed under the microscope (100×).

Stable USP32 silencing via shRNA

Synthetic sense (S) and anti-sense (AS) oligos (IDT) corresponding to USP32 cDNA nucleotide positions 1886-1904 were annealed and cloned into the pSUPERretroneoGFP (Oligo Engine) (S: 5′-GATCCCCCCAGTAAAGGCTACATCATTTCAAGAGAATGATGTAGCCTTT ACTGGTTTTTA-3′ and AS: 5′-AGCTTAAAAACCAGTAAA GGCTACATCATTCTCTT GAAATGATGTAGCCTTTACTGGGGG-3′) to form the shRNA. Similarly, control oligos with no homologies to human genome were annealed and cloned into pSUPERretroneoGFP: (S: 5′-GATCCCCGTACGTTACGCGTAACGTATTCAAGAGATACGTTACGCGTAACGTACTTTTTA-3′, AS: 5′-AGCTTAAAAAGTACGTTACGCGTAACGTATCTCTTGAATACGTTACGC GTAACGTACGGG-3′). Anti-USP32, control shRNA, and mock transfections were done with Fugene-HD (Roche). Positive mono- and polyclones in both HeLa and MCF7 cells were selected by 750 μg/ml G418 treatment. Silencing of USP32 was confirmed by qRT-PCR as described above, and statistical analysis of data was done by one-way ANOVA followed by Tukey’s multiple-comparison test (p < 0.05).

Proliferation rate determination

Cell proliferation was measured by 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide assay (MTT) (Roche) according to the manufacturer’s instructions. A total of 5 × 103 cells were plated in 96-well plates in complete DMEM to be assayed at 24, 48, 72, and 96 h. The absorbance was measured using a Bio-Rad microplate reader at 570 nm. Data obtained for each cell group at 48, 72, and 96 h was normalized to the corresponding group’s OD values at 24 h to eliminate initial cell counting and/or plating differences. Experiments were done in triplicate. Statistical analysis of the data was done by one-way ANOVA followed by Tukey’s multiple-comparison test (p < 0.05).

Transwell migration assay

Transwell migration assay was performed as previously described (Cimen et al. 2009). HeLa cells (3 × 104) or MCF7 cells (10 × 104) were plated in the upper chamber of the transwell plates (8 μm pore size; Greiner Bio-one) in starvation medium (1% FBS). The lower side of the chamber was filled with 10% FBS containing medium. Cells were allowed to migrate for 24 h; nonmigrated cells were removed by scrubbing with sterile cotton swabs. The chambers were fixed in 100% methanol for 10 min, stained with Giemsa solution for 2 min, and washed twice in distilled water (Cimen et al. 2009). The migration of cells was quantified under a light microscope with 4× objective by counting five random fields per membrane. Statistical analysis of data was done by one-way ANOVA followed by Tukey’s multiple-comparison test (p < 0.05).

Results

USP32 as a ubiquitously expressed and membrane-bound DUB

Because USP6 is an oncogene with high sequence similarity to USP32 and is suspected to have alternatively spliced variants (Papa and Hochstrasser 1993) that may resemble the 3′ coding sequence of USP32, we first determined USP6 and USP32 expression patterns to examine the possibility of USP6 function overlapping with USP32. Therefore, we designed primers to amplify different-sized PCR products for USP32 and USP6 transcripts in a multitissue cDNA panel to distinguish their expression patterns. Earlier, USP6 was reported to be expressed only in testis (Paulding et al. 2003), but we also demonstrated USP6 transcript in ovary, whereas USP32 was ubiquitously expressed in the multitissue cDNA panel (Fig. 1a).

Fig. 1
figure 1

USP32 expression and deubiquitination function. aUSP32 (396 bp) and USP6 (245 bp) RT-PCR with a multitissue panel of spleen, testis, prostate, ovary, small intestine, colon, leukocyte, and thymus cDNAs. b Domain structure of partial USP32 constructs (I, II, III) and in vivo deubiquitination assay. Plasmid combinations (pGEX-USP32 and Ub-β-gal) were cotransformed into in DH5α cells. 1. Ub-β-gal plasmid only. 2. Empty pGEX + Ub-β-gal. 3. pGEX-USP32-I +Ub-β-gal. 4. pGEX-USP32-II + Ub-β-gal. 5. pGEX-USP32-III + Ub-β-gal. 6. pGEX-Ubp3 + Ub-β-gal [Ubp3 was used as a positive control for deubiquitination activity (Baker et al. 1992)]. Ub-β-gal is 125 KDa, β-gal is 117 KDa. c Conserved residues required for deubiquitination activity in USP32 and other DUBs [e.g., USP6, Doa4 (Y: yeast), Ubp3 (Y: yeast)] are shown. Bold letters show the conserved amino acids. * indicates active aspargine (738th amino acid of USP32), cysteine (743rd amino acid of USP32), histidine (1526th amino acid of USP32), and aspartic acid (1543rd amino acid of USP32) residues found in DUBs as indicated by ENTREZ

Full size image

USP32 (1604 aa) is predicted to be a ubiquitin protease due to the C-terminal aspargine, cysteine, histidine, and aspartic acid residues in the peptidase domains (Fig. 1b, c) conserved in DUBs. To confirm the predicted function of USP32, an in vivo deubiquitination assay (Papa and Hochstrasser 1993) was performed.

USP32 cDNA encoding the C-terminal peptidase domains was cloned as three overlapping inserts into pGEX vectors as GST fusion peptides. These constructs (I, II, and III) and Ub-β-gal plasmids were cotransformed into DH5α cells for ectopic expression. Deubiquitination assay tested the protease ability of the peptides expressed from pGEX based on ubiquitin removal from the Ub-β-gal. DH5α cells cotransformed with both pGEX-USP32 and pACYC184-Ub-β-gal constructs were lysed and USP32 peptide expression was confirmed by Western blotting (GST antibody, data not shown). A β-gal antibody showed the cleavage of Ub-β-gal into β-gal by both USP32 (construct II, has all three active peptidase domains) (lane 4) and the positive control Ubp3 (lane 6) (Baker et al. 1992), showing the deubiquitination activity of USP32 (Fig. 1b). Construct I (lane 3), which harbored only the first peptidase region (amino acids 733-911), and construct III (lane 5), which harbored only the second and the third peptidase regions (amino acids 1225-1318 and 1510-1565), had minimal enzymatic activity as faint β-gal bands could be visualized for darker exposures of the film (darker exposures not shown).

Next, we examined the cellular localization of USP32. Fluorescence protection assay relies on the time course loss of the fluorescence signal of any soluble protein after the permeabilization of the plasma membrane (Lorenz et al. 2006). We checked for but did not observe any photobleaching of transfected cells during the assay period (Fig. 2a). Transfected cells were then treated with digitonin to permeabilize the plasma membrane. The signal of USP32-GFP (pEGFP-USP32, third row) did not fade out, even 270 s after digitonin treatment as it did in GFP-only-expressing cells (pEGFP only, second row) in 90 s. This initial observation suggested a membrane- and/or organelle-bound localization of USP32, as also predicted by ENTREZ Gene (Maglott et al. 2007). The GFP signal for vector-transfected cells was undetectable after 145 s (Fig. 2a).

Fig. 2
figure 2

Fluorescence protection assay and subcellular localization of USP32 in HeLa cells. a First row: Empty pEGFP vector-transfected cells without digitonin treatment were imaged for 0, 40, and 170 s to demonstrate lack of photobleaching. Second row: pEGFP-transfected cells were treated with 20 μM digitonin and were imaged for 0, 55, and 145 s until the GFP signal faded. Third row: pEGFP-USP32-transfected cells were treated with 20 μM digitonin and cells were imaged for as long as 270 s (20 ×). b Three images of full-length USP32 (USP32-GFP) localization (100×). c Full-length USP32 (USP32-GFP) (first column), BODIPY-TR dye staining of Golgi (second column), and overlay of the two images (third column). Below are cells transfected with USP32-1-GFP (1), GFP-USP32-2 (2), and GFP-USP32 -3 (3) together with the structures of the constructs (100×)

Full size image

To further confirm these observations, we examined the subcellular localization of the full-length and partial USP32-GFP fusion peptides in HeLa cells. Full-length USP32 (USP32-GFP) seemed to localize to Golgi (stained with BODIPY-TR). Other N terminus constructs had similar localization, whereas construct 3 (USP32-3) always had a very clear cytoplasmic diffusion pattern (Fig. 2b, c). We have observed that localization of USP32-1 and USP32-2 was consistent with USP32-GFP, but cytoplasmic signals were also detected in some cells for these partial constructs, almost comparable to GFP alone (Fig. 2b).

USP32 silencing and its effect on proliferation rate

Anti-USP32 and control shRNA vectors were stably transfected into HeLa cells. Silencing of USP32, detected with qRT-PCR (Fig. 3a), resulted in a more than 30% reduction in the proliferation rate of transfected cells (both poly- and monoclonal) at 96 h post plating detected by MTT (Fig. 3b) and growth curve assays (Supplementary Fig. 1). Moreover, USP32-silenced HeLa cells had lower migration abilities compared to control cells as indicated by the transwell migration assay (Fig. 3c). However, no evident apoptosis or significant difference in the cell cycle profiles of USP32-silenced and control cells was observed when stained with propidium iodide for flow cytometry analysis (data not shown).

Fig. 3
figure 3

Stable USP32 silencing and its effect in HeLa cells. a Relative expression of USP32 in mock (empty vector), control shRNA (C), and anti-USP32 shRNA-transfected cells was determined by qRT-PCR (calculated by the ∆∆Ct method). shRNA-1 M: USP32-silenced monoclonal cells; shRNA-1P: USP32-silenced polyclonal cells (same shRNA used for both). The baseline for the mock-transfected HeLa cells was set to 1. ** indicates significant difference between C and shRNA-1P cells, p < 0.05 (Tukey’s multiple-comparison test). b 5000 mock, control shRNA (C), and anti-USP32 shRNA-transfected cells [monoclonal (shRNA-1 M) and polyclonal (shRNA-1P)] were plated and assayed for proliferation at 24, 48, 72, and 96 h by MTT. Data obtained for each cell group at 48, 72, and 96 h were normalized to the corresponding cell group’s OD values at 24 h to eliminate cell counting and plating differences. *** indicates significant difference between C and shRNA-1P cells, p < 0.05 (Tukey’s multiple-comparison test). c Transwell migration assay was done in the presence of serum as a chemoattractant. USP32-silenced shRNA-1 M and shRNA-1P cells migrated through the 8-μm pores of the transwell chamber in significantly less numbers (*** p < 0.05) compared to the control and mock-transfected cells

Full size image

USP32 overexpression in breast cancer cells

Based on USP32’s effect on the proliferation and migration rate and the fact that it maps to 17q23, which is commonly amplified in breast cancers, we expanded our USP32 transcript analysis to breast cancer cell lines and primary tumors compared to normal breast tissue and immortalized nontumorigenic mammary cells (HPV11-21) by qRT-PCR. USP32 was found to be overexpressed more than twofold in 9 of 18 breast cancer cell lines compared to normal breast tissue and HPV11-21 (Fig. 4a) and in 22% of (9 of 41) primary breast tumors compared to the mean of 7 normal breast tissue samples, indicated as C1-7 (Fig. 4b). As a representative cell line for USP32 overexpression, all 34 exons of USP32 in MCF7 cells were PCR amplified and sequenced to determine if the wild type or mutated transcript was being produced. No mutations were detected in the coding sequence or intron-exon junctions, a result consistent with the COSMIC (Catalogue of Somatic Mutations in Cancer) database (http://www.sanger.ac.uk/genetics/CGP/cosmic/) which reports the lack of mutations in USP32 in the examined samples (Forbes et al. 2008). Moreover, no USP6 or its known alternatively spliced variants [known as Clones 210 and 213 (Nakamura et al. 1992)] were detected in the normal breast tissue and in two of the representative breast cancer cell lines with high and normal USP32 expression (MCF7 and MDA-MB-231, respectively) (Supplementary Fig. 2 and data not shown). Therefore, we eliminated the possibility of the presence of the USP6 transcript that may have overlapping expression profiles and/or functions due to sequence similarity with USP32 in normal breast and the examined breast cancer cells.

Fig. 4
figure 4

USP32 overexpression in breast cancer cells. a qRT-PCR amplification of the USP32 transcript in 18 breast cancer cell lines compared to normal breast tissue and HPV11-21 (calculated by the ΔΔCt method). Dotted bars indicate cell lines with USP32 expression more than twofold compared to controls (normal breast tissue and HPV11-21). b qRT-PCR amplification of the USP32 transcript in 41 primary breast tumors compared to 7 normal breast (C1-7) (calculated by the ΔCt method). Dotted bars indicate primary tumors, with USP32 expression more than twofold compared to C1-7

Full size image

Since USP32 silencing caused a decrease in the proliferation and migration properties of HeLa cells, we also stably transfected MCF7 cells with anti-USP32 and control shRNA vectors (Fig. 5a) to examine if the same effect would be observed in breast cancer cells. A reduction in the proliferation rate of MCF7 cells was detected by MTT analysis (Fig. 5b) at the end of 96 h. Moreover, a more than 30% reduction in the migration ability of USP32-silenced MCF7 cells was observed in the transwell migration assay (Fig. 5c).

Fig. 5
figure 5

Stable USP32 silencing and its effect in MCF7 cells. a Relative expression of USP32 in mock (empty vector), control shRNA (C), and anti-USP32 shRNA-transfected cells was determined by qRT-PCR (calculated by the ∆∆Ct method). shRNA-1 M: USP32-silenced monoclonal cells; shRNA-1P: USP32-silenced polyclonal cells (same shRNA used for both). The baseline for the mock-transfected MCF7 cells was set to 1. ** indicates significant difference between C and shRNA-1P MCF7 cells, p < 0.05 (Tukey’s multiple-comparison test). b 5000 mock, control shRNA (C), and anti-USP32 shRNA-transfected cells (shRNA-1 M and shRNA-1P) were plated and assayed for proliferation at 24, 48, 72, and 96 h by MTT. Data obtained for each cell group at 48, 72, and 96 h were normalized to the corresponding cell group’s OD values at 24 h to eliminate cell counting and plating differences. *** indicates significant difference between C and shRNA-1P cells, p < 0.05 (Tukey’s multiple-comparison test). c Transwell migration assay showed significantly (** p < 0.05) less migration of USP32-silenced cells through the 8-μm pores of the transwell chamber compared to controls

Full size image

In a cDNA microarray experiment, USP32 was reported among a group of genes that were responsive to estrogen (E2) and antiestrogen treatments in ER-positive ZR-75.1 breast cancer cells (Scafoglio et al. 2006). To determine experimentally if any canonical and/or noncanonical E2 response elements exist in the USP32 promoter, we cloned a 2.7-kb immediate upstream promoter region of USP32 into a dual luciferase reporter system but did not detect any E2 responsiveness in these regions (Supplementary methods and Supplementary Fig. 3) in MCF7 (ER +) cells. Although in silico prediction programs suggested the presence of estrogen response elements within this 2.7-kb region, these sites were not conserved between mice and human USP32 upstream sequences (data not shown).

Discussion

It has long been known that DUBs have a role in the pathogenesis of many diseases, including cancer. They are known to take part in protein degradation, receptor endocytosis, and possibly other processes. The role of DUB enzymes in the Golgi apparatus is also just beginning to be understood as more evidence points to regulatory roles of ubiquitin in the sorting of proteins at the trans-Golgi network (Piper and Luzio 2007) and Golgi membrane dynamics (Meyer 2005; Wang et al. 2004). Ubp3p, for example, together with Bre5p has roles in the regulation of COPI and COPII complexes in endoplasmic reticulum to Golgi trafficking (Cohen et al. 2003a, b) and autophagy (Kraft et al. 2008).

In this study we confirmed the DUB activity of USP32. To get initial evidence on what this DUB might be doing in cells, we performed a fluorescence protection assay to determine if USP32 is a soluble or membrane/organelle-bound protein. Contrary to soluble GFP alone, the GFP-fused USP32 signal did not leave the cell when the membranes were permeabilized with digitonin, whose activity is limited to only the cholesterol-rich plasma membranes (Lorenz et al. 2006). Subcellular localization studies also confirmed these results and suggested a possible Golgi colocalization. Localization of the partial proteins (USP32-1 and USP32-2) was consistent with the full-length USP32, but we also noticed cytoplasmic signals in some transfected cells for N-terminal partial constructs, almost comparable to GFP alone. This may be explained by the presence of a Golgi localization or retention signal around the C terminus of USP32-1 and the N terminus of USP32-2, lowering the efficiency of localization of these proteins. It is also possible that localization pattern inconsistencies for these two partial peptides are due to a different cell cycle status of the transfected cells or abnormal folding of partial peptides. The roles of USP32 in connection to Golgi, its potential partners, and the functions of the N-terminus EF-hand domains remain to be investigated.

We also showed USP32 overexpression in breast cancer cell lines and primary breast tumors. No mutations were detected in MCF7 cells as a representative cell line for USP32 overexpression, indicating that it is the wild-type transcript that is overexpressed. The fact that no mutations have been identified in USP32 in cancer cells to date (Forbes et al. 2008) suggests that the increased expression observed in cancer cells is most likely associated with amplification or other epigenetic mechanisms.

No USP6 or its variants (excluding as of yet undetected isoforms) were found in normal breast or the examined breast cancer cell lines. It was crucial to know the USP6 expression status in breast cells because of the high sequence similarity between USP6 and USP32.

Although E2-induced expression of USP32 has been reported, we failed to detect any E2 responsive regions in the 2.7-kb upstream region, but we cannot exclude any other indirect effects of E2 or further upstream regions harboring E2 response elements. Such effects will need to be investigated for E2 responsiveness of USP32.

In short, USP32 is a novel DUB that maps to 17q23 which harbors an amplicon in breast cancer cells (Erson et al. 2001; Sinclair et al. 2003). Since the amplicon structure is complex and gene rich, a functional understanding of genes in this region and how they may contribute to breast tumorigenesis is crucial. Together with its effect on proliferation and migration as well as its overexpression in breast cancer cells, further investigation of USP32 function in Golgi will be interesting. Specific roles of USP32 (by itself or with other 17q23 genes) in breast carcinogenesis will be important to investigate next.