Tumor Suppressor RARRES1 Interacts with Cytoplasmic Carboxypeptidase AGBL2 to Regulate the A-Tubulin Tyrosination Cycle

Published OnlineFirst February 8, 2011; DOI: 10.1158/0008-5472.CAN-10-2294

Cancer Priority Report Research

Tumor Suppressor RARRES1 Interacts with Cytoplasmic Carboxypeptidase AGBL2 to Regulate the a-Tubulin Tyrosination Cycle

Ziad J. Sahab1, Michael D. Hall1, You Me Sung1, Sivanesan Dakshanamurthy1, Yun Ji1, Deepak Kumar2, and Stephen W. Byers1,2

Abstract Even though it is among the most commonly methylated loci in multiple cancers, the retinoic acid–induced tumor suppressor retinoic acid receptor responder 1 (RARRES1) has no known function. We now show that RARRES1 is lost in many cancer cells, particularly those with a mesenchymal phenotype, and is a transmembrane carboxypeptidase inhibitor that interacts with ATP/GTP binding protein-like 2 (AGBL2), a cytoplasmic carboxypeptidase. Knockdown of AGBL2 results in a failure of the cell to detyrosinate the C-terminal EEY region of a-tubulin and indicates that it is a candidate for the long sought-after tubulin tyrosine carboxypeptidase important in the regulation of microtubule dynamics. In contrast, knockdown of RARRES1 increases the level of detyrosinated a-tubulin consistent with a role as the cognate inhibitor of AGBL2. We conclude that RARRES1, its interacting partners AGBL2, Eg5/KIF11, another EEY-bearing protein (EB1), and the microtubule tyrosination cycle are important in tumorigenesis and identify a novel area for therapeutic intervention. Cancer Res; 71(4); 1219–28. 2011 AACR.

Introduction trolling aspects of stem cell biology, in which an inverse and direct relationship between LXN expression and the size of the Retinoic acid receptor responder 1 (RARRES1), also known hematopoietic stem cell population in mice has been reported as tazarotene-induced gene 1, was first identified as a novel (4). LXN was initially described as the only known mammalian retinoid-responsive gene in skin. It is induced in a retinoic acid carboxypeptidase inhibitor and is involved in the regional receptor–specific manner in a variety of human skin-related specification of neurons (5–8). Despite extensive evidence for systems (1). The RARRES1 promoter region is methylated in a tumor suppressor role of RARRES1, no mechanism for its primary prostate cancers compared with normal tissues or biological function has been determined. benign hyperplasias, and its decreased expression is asso- Here, we show that RARRES1 is a type III membrane protein ciated with an increase in the malignant potential of prostate found in a complex with several proteins involved in the carcinoma cells (2). RARRES1 is among the most commonly regulation of microtubule function and reveal that ATP/ methylated loci in multiple cancers and is often described as a GTP binding protein-like 2 (AGBL2; a carboxypeptidase) putative tumor suppressor gene (3). In addition, RARRES1 also and RARRES1 (an inhibitor) regulate the tubulin tyrosination plays a role in the proliferative/differentiative switch in adult cycle. Our findings point to a role for RARRES1, AGBL2, and adipose-derived mesenchymal stem cells. Notably, a RARRES1 the tubulin tyrosination cycle in cancer and identify a novel family member, latexin (LXN), has been implicated in con- avenue for potential therapeutic intervention.

Materials and Methods Authors' Affiliations: 1Georgetown-Lombardi Comprehensive Cancer Center, Georgetown University; and 2Department of Biological and Envir- onmental Sciences, University of the District of Columbia, Washington, DC Cells Immortalized human prostate epithelial PWR-1E cells (a gift Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/). from Dr. S.C. Chauhan, University of South Dakota), PC-3 Z.J. Sahab and M.D. Hall contributed equally to this work. human prostate cancer cells, SKBR-3 human breast cancer cells, and HEK 293 human embryonic kidney cells were main- Current address for Y. Ji: Hatfield Clinical Research Center, National tained according to the recommendation of American Type Cancer Institute, Bethesda, MD 20892. Culture Collection. Corresponding Author: Stephen W. Byers, Lombardi Comprehensive Cancer Center, Georgetown University, 3970 Reservoir RD NW Research Building Office E415A, Washington, DC 22057. Phone: 202-687-1813; Antibodies and reagents Fax: 202-687-7505. E-mail: [email protected] Primary antibodies targeting the following antigens were doi: 10.1158/0008-5472.CAN-10-2294 used: goat anti-human RARRES1 (catalogue no. AF4255, R&D 2011 American Association for Cancer Research. Systems), antityrosinated tubulin, and anti-Eg5 (Abcam),

www.aacrjournals.org 1219

Sahab et al.

anti–E-cadherin (BD-Transduction Laboratories), anti-cyclin Cell fractionation D1 (EMD-Calbiochem), anti-HA (Millipore), anti–histone-H4 Three-pool (membrane, cytoplasmic, and nuclear/cell deb- (Cell Signaling), anti–detyrosinated tubulin (AbD Serotec), ris fractions) cell fractionation was carried out as previously and anti-AGBL2, and anti–pan-cadherin (Sigma-Aldrich). described (9). The following inhibitory RNAs (RNAi; Dharmacon) were used: for RARRES1 knockdown, GUACACGGCUCAUCGAGAA and Cell lysis and tandem affinity purification AAAGAGGGAUGUAAAGUUC. For AGBL2 knockdown: GCA- A stable clone expressing a level of exogenous pGlue CACUUCUACCCAUAUA and UGGACAAGAUGUAGAUUUA. RARRES1 close to the endogenous level of this protein was grown in 5 dishes (150 mm each). At 90% confluency, medium siRNA, expression constructs, and transfection was discarded and each dish was lysed in 0.5 mL of a lysis Variant and full-length human RARRES1 isoforms were buffer, composed of 10% glycerol, 50 mmol/L HEPES-NaOH, directionally cloned into the BglII and HindIII sites in the pH 8.0, 100 mmol/L NaCl, 2 mmol/L EDTA, 0.1% NP-40, 2 pEYFP-N1 vector (Clontech). N-RARRES1 (þ121 to þ897) and mmol/L DTT, 10 mmol/L NaF, 5 nmol/L calyculin A, 50 mmol/ full-length RARRES1 (þ1toþ897) were cloned into the pGlue L b-glycerolphosphate, and 1 Complete Mini protease inhi- vector as codon-optimized versions by Genscript. AGBL2 bitor (Roche). Lysates were harvested by scraping and 2 (catalogue no. M-012937-00) and nontargeting control siRNAs freeze-thaw cycles were done to improve protein recovery. (catalogue no. D-001210-01-20) were from Dharmacon. Plas- Lysates were then centrifuged at 15,000 g for 15 minutes, mid DNA and siRNA constructs were introduced to cells either and supernatant were recovered and incubated for 4 hours by electroporation mediated by an Amaxa nucleoporator with 100 mL of streptavidin beads (Streptavidin Sepharose (Amaxa-Lonza) or by Fugene 6 transfection reagent (Roche). High Performance, GE Healthcare) prewashed 3 times with the lysis buffer. The slurry was then centrifuged at 1,500 g for Detection of AGBL family members by quantitative PCR 2 minutes, and the precipitate containing the streptavidin Total RNA was extracted from indicated cell lines with Trizol beads was recovered and washed 3 times with lysis buffer and reagent (Invitrogen) and isolated with an RNeasy purification 2 times with the TEV buffer, supplied with the AcTEV Protease kit (Qiagen). Single-stranded cDNA was prepared from 400 ng kit (catalogue no. 12575-015; Invitrogen). Streptavidin beads of RNA by TaqMan reverse transcription reagents (Applied were then incubated with 200 units of TEV protease in 150 mL Biosystems) following the manufacturer's guidelines. Real-time of TEV buffer overnight at 4C. The slurry was then centri- PCR (RT-PCR) was then carried out and monitored on a 7900 fuged at 1,500 g for 2 minutes, and the supernatant was HT system (Applied Biosystems), using TaqMan universal PCR recovered. The precipitated beads were washed twice with 200 master mix and the following inventoried primer/probe sets: mL of TEV buffer, centrifuged, and supernatants were pooled. GAPDH–Hs99999905_m1, AGBL1/Nna1–Hs00328701_m1, The final pool volume was diluted 1:1 (v/v) with calmodulin- AGBL2–Hs00417079_m1, AGBL3–Hs00227489_m1, AGBL4– binding buffer composed of 10 mmol/L b-mercaptoethanol, 10 Hs00262179_m1, and AGBL5–Hs00222447_m1 (Applied Bio- mmol/L HEPES-NaOH, pH 8.0, 150 mmol/L NaCl, 1 mmol/L systems). Plotted DCt values were determined by subtracting MgOAc, 1 mmol/L imidazole, 0.1% NP-40, and 2 mmol/L control [glyceraldehyde 3-phosphate dehydrogenase CaCl2. The mixture was then incubated for 90 minutes at (GAPDH)] cycle threshold values from each target cycle thresh- 4C with 100 mL of calmodulin beads (Calmodulin Sepharose old. Where targets did not return a threshold value after 40 4B; GE Healthcare) prewashed 3 times with calmodulin-bind- cycles of PCR, the transcript was determined to be absent. ing buffer. The slurry was then centrifuged at 1,500 g for 2 minutes, the supernatant was discarded, and the precipitated RARRES1 RT-PCR beads were washed twice with a calmodulin-rinsing buffer RARRES1 (full-length) forward 50-CAACAAGAGGAT- composed of 50 mmol/L ammonium bicarbonate, pH 8.0, 75 TACCTGCTTTACAAG-30 and reverse 50-GAGCAGAGTT- mmol/L NaCl, 1 mmol/L MgOAc, 1 mmol/L imidazole, and 2 0 CAGTGTGCATG-3 primers (generating a 630-base pair mmol/L CaCl2. Each wash was followed by centrifugation and amplicon). For b-actin, forward 50-CCACTGGCATCGTGATG- supernatant shedding. A total of 150 mL of a calmodulin- GAC-30 and reverse 50-GCGGATGTCCACGTCACACT-30 pri- elution buffer composed of 50 mmol/L ammonium bicarbo- mers (generating a 350-base pair amplicon). Thermal cycling nate, pH 8.0, and 25 mmol/L EGTA was then added to the for RARRES1 was done according to the following profile: 30 calmodulin resin. The slurry was then vortex mixed, centri- minutes at 50C, 15 minutes at 95C for the reverse tran- fuged at 1,500 g, and the supernatant was collected. This scriptase reaction, followed by PCR cycling of 30 seconds at elution step was repeated twice and supernatants were pooled 94C, 45 seconds at 52C, 1 minute at 72C, for indicated together (Supplementary Fig. S1). number of cycles with a subsequent final elongation at 72C for 10 minutes. Trypsin digestion and liquid chromatography/tandem mass spectrometric analysis PNGase F digestion The pull-down sample was vacuum-dried and reconstituted Whole cell lysates of untreated PWR-1E cells were made with 20 mL of a buffer composed of 500 mmol/L triethylam- with RIPA buffer, followed by digestion with the glycosidase monium bicarbonate, pH 8.5. The protein sample was then PNGase F, or no enzyme (NEB), according to the man- denatured by adding 1 mL of a 2% SDS solution followed by the ufacturer's instructions. addition of 2 mL of a reducing reagent composed of 50 mmol/L

1220 Cancer Res; 71(4) February 15, 2011 Cancer Research

RARRES1 and Tubulin Tyrosination

TRIS-(2-carboxyethyl)phosphine. The mixture was then incu- minutes. The supernatant was recovered and mixed with 10 bated at 60C for 1 hour followed by the addition of 1 mLofa mL of primary antibody and incubated for 1 hour at 4C. cysteine-blocking reagent composed of 200 mmol/L methyl Twenty microliters of A/G bead slurry was then added and methane-thiosulfonate (MMTS) in isopropanol. Trypsin diges- incubated at 4C for 1 hour. Samples were then centrifuged, tion was then done by adding 10 mLof1mg/mL trypsin solution and supernatants were discarded. The precipitate was boiled in 80 mmol/L CaCl2. Samples were incubated overnight at for 3 minutes after adding 20 mL of SDS-PAGE sample buffer to 37C, vacuum dried, and then reconstituted with 10 mLofa2% release the complex from the beads. Western blotting was acetonitrile and 0.1% formic acid in distilled deionized water, then done as described above. as described previously (10). Nano liquid chromatography/ tandem mass spectrometry (LC/MS-MS) was done using a Q- Homology modeling Star Elite (Applied Biosystems), equipped with a nanoAcquity AGBL2 structure was predicted with human carboxypepti- UPLC system (Waters). Sample separations were done with a dase A1 (CPA1; PDB code: 1V77) as a template. The sequence 1.7-mm nanoAcquity BEH130 C18 column (100 mm 100 mm) identity of AGBL2 and CPA1 is 27%. The missing loops were at a flow rate of 400 nL/min. Tryptic digests were eluted by the built using the "loop model" building option in the Model- following gradient: 100% of solvent A [97.9% water, 2% acet- ler9v7. The model was refined further by molecular dynamics onitrile, 0.1% formic acid (v/v/v)] for 1 hour; then from 100% (MD) simulations, followed by energy minimization using solvent A to 100% solvent B [2% water, 97.8% acetonitrile, 0.1% SANDER module of AMBER 10.0. The quality of the refined formic acid (v/v/v)] in 2 hours; finally, A 100% solvent B flow model was checked with PROCHECK. Docking of the "EEY" was maintained for 1 hour, followed by a return to 100% of peptide motif was carried out with the SurFlexDock. MD solvent A flow in 15 minutes. Mass spectrometric settings simulations and energy minimization were done using the were as follows: ion spray voltage, 2,300 V; interface heater AMBER10.0 package. temperature, 220C; cone voltage, 20 V; and collision energy, 8V. Immunofluorescence To assess the levels of detyrosinated tubulin in control Protein identification versus AGBL2, RARRES1 knockeddown, and/or paclitaxel- Protein identification was done using ProteinPilot software treated HEK 293 cells, 50,000 cells were plated on cover slides with the following settings (11): sample type: identification; (Fisher brand microscope cover glass #12-545-100 18CIR-1). cysteine alkylation: MMTS; digestion: trypsin; instrument: After 24 hours, cells were treated with paclitaxel or dimethyl QSTAR ESI; species: Homo sapiens; Min S/N Filter: 10; pre- sulfoxide at a final concentration of 5 mmol/L for 2 hours. HEK cursor tolerance 75 ppm; maximum missed cleavage: 1; and 293 cells were then fixed in 3.7% paraformaldehyde/PBS for 10 MS/MS fragment tolerance of 0.3 Da. minutes at room temperature, followed by washing the cells 3 times with PBS. Postpermeabilization was done by adding Western blotting PBST (PBS/Tween 20) and incubating the cells for 5 minutes Western blotting was done as described previously (12, 13). at room temperature. Cells were then washed 3 times with PBS, Cells were lysed with RIPA buffer and loaded onto 4% to 12% followed by the addition of the primary antibody that consisted gradient polyacrylamide gel. Amounts of proteins loaded were of a rabbit polyclonal anti–detyrosinated tubulin (ABD Sero- 1 mg of total cell lysates for tubulin immunoblots and 10 mg for tec) with a final dilution of 1:250. Cells were then washed 3 all other immunoblots. Gels were electrophoresed at 100 V times with PBS before adding the secondary antibody that until the end of the separation. Proteins contained within the consisted of an Alexa 488–conjugated anti-rabbit IgG diluted gel were then electroblotted onto a nitrocellulose membrane 1:300 (Invitrogen). 40,6-Diamidino-2-phenylindole (DAPI) was (50 V for 50 minutes). Western blot analyses were accom- also added at a 1:50 dilution for nuclei detection. Cells were plished by a 1 mg/mL dilution of primary antibody, followed by washed 3 times with PBS, and images were obtained using a incubation with a horseradish peroxidase–conjugated second- 60 oil lens on the Olympus FV 300 confocal microscope. ary antibody against the appropriate species. Visualization of Consistent laser intensity or camera exposure levels for each the bands was then accomplished by the addition of a 1:1 ratio fluorescent marker in each experiment were used. For image of Super Signal West Pico-Stable Peroxidase Solution and analysis and quantification, measurements were made using Luminol/Enhancer Solution (Pierce) and by developing the Metamorph Image analysis software ver. 7.0. Average intensity chemiluminescent signal in the dark with the Kodak Scientific was calculated from integrated intensity and area for each Imaging film (Kodak catalogue no. 1651496), Fixer and Replen- selected area. Quantitation of fluorescence signals from 5 isher/Developer, and Replenisher (Kodak catalogue no. random fields for each treatment was done. An example of 1901859), according to the manufacturer's instructions. an original image used for quantitation is included in the supplementary materials (Supplementary Fig. S2). Immunoprecipitation HEK 293 cells were lysed with RIPA buffer and centrifuged Results at 14,000 g for 15 minutes. The supernatant was recovered and precleared by adding 1 mg of normal IgG premixed with 20 RARRES1 organization mL of A/G protein bead slurry. The mixture was incubated for RARRES1 is related to the putative carboxypeptidase inhi- 30 minutes at 4C and then centrifuged at 1,000 g for 5 bitor LXN, and both genes are adjacent to one another on

www.aacrjournals.org Cancer Res; 71(4) February 15, 2011 1221

Sahab et al.

B Human_Latexin Rhesus_Monkey_XP001101220 Mouse_Latexin Rat_Latexin Human_TIG1_Long_Isoform Human_TIG1_Short_Isoform Chimpanzee_TIG1 Mouse_TIG1 Figure 1. RARRES-1 in cancer. Rat_TIG1 Xenopus_LOC398803 A, ideogram of human Chicken_Ovacalyxin_32 chromosome 3 in which Zebrafish_LOC494053 Opossum_XP001363241 RARRES1 and the paralogue LXN are adjacent at position 3q25. B, a C phylogenetic tree of selected RARRES1 and LXN orthologues, suggesting RARRES1 is the ancestral gene. C, schematic representation of the mRNA for the 6-exon RARRES1 isoform and 5-exon variant isoform (open D regions are coding, whereas shaded regions are untranslated). D, the putative variant protein contains an alteration in the amino acid sequence (underlined) which results in subsequent truncation. E, RARRES1 and its variant are induced by retinoic acid (RA) or E vitamin D3 (D3) in PWR-1E cells but not in PC-3 or other more aggressive prostate cancer cells.

β-Actin

chromosome 3, suggesting that they arose as a result of gene cells with a mesenchymal phenotype (Fig. 2A; Supplementary duplication (ref. 14; Fig. 1A). Similar molecules exist in all Table S1). HEK 293 cells express low but still detectable levels vertebrates examined; in zebrafish and several other more of RARRES1. In PWRE-1E cells, RARRES1 is further induced ancient vertebrates, only one orthologue exists, and this is both by retinoic acid and by vitamin D. more closely related to RARRES1 than to LXN (Fig. 1B). These data indicate that RARRES1 is the ancestral gene and in Identification of the RARRES1 interactome zebrafish, may, at least, fulfill the function of both LXN and We utilized a tandem affinity purification (TAP) LC/MS-MS RARRES1. We found that RARRES1 exists as 2 mRNA spliceo- approach to characterize the RARRES1 interactome in HEK forms: a low abundance variant and an abundant form which 293 cells. Nano-LC-Q-TOF/TOF mass spectrometry and a encodes for an additional 66 amino acids at the C-terminus ProteinPilot software search revealed 9 proteins (including (Fig. 1C–E). RARRES1) identified with a CI of more than 95% (Table 1). RARRES1 mRNA is decreased in prostate cancer and further Proteins that were also present in a complex isolated from in metastatic prostate cancer compared with normal prostate cells stably expressing the empty vector control were removed tissue and is induced after neoadjuvant therapy (2). RT-PCR from the analyses. Several proteins in the RARRES1 complex analysis shows that RARRES1 is expressed in normal prostate regulate microtubule function. These are as follows: the and breast cell lines and more differentiated breast cancer cell mitotic spindle-associated kinesin eg5/KIF11, an emerging lines and is very low in aggressive prostate and breast cancer target for cancer therapy; EB1, a microtubule plus end-binding

1222 Cancer Res; 71(4) February 15, 2011 Cancer Research

RARRES1 and Tubulin Tyrosination

Figure 2. RARRES1 and AGBL2 expression and interaction. A, RARRES1 is induced by retinoic acid (107 mol/L) after 30-cycle RT-PCR in SKBR-3 breast cancer cells and PWR-1E cells but not in PC-3 prostate cancer cells. R, retinoic acid; E, ethanol. B, immunoblot of RARRES1 in PC-3 (negative) and PWR-1E cells expressing endogenous (38 kDa), HA-DN-RARRES1 (45 kDa), and HA-RARRES1 (50 kDa, degradation product at 35 kDa). Immunoblot of RARRES1 in PWR-1E cells expressing variant-RARRES1-EYFP (57 kDa) and RARRES1-EYFP (64 kDa). Subcellular fractionation of PWR-1E cells indicates membrane-associated (38 kDa) and nuclear (50 kDa) pools of endogenous RARRES1. Endoglycosidase (PNGase F) digestion of PWR-1E whole cell lysates. C, immunoblot showing the presence of RARRES1, AGBL2, Eg5, and EB1 present in the RARRES1-pGlue complex after TAP, RARRES1 endogenous immunoprecipitation, and AGBL2 endogenous immunoprecipitation. Immunoblot of HA, contained in the pGlue vector, depicts the presence RARRES1-pGlue (45 kDa) in the sample versus the empty vector (11 kDa) in the control. D, 1/DCt values for AGBL family members (DCt represents the relative abundance of the different AGBL genes to the reference RNAse P; 1/DCt was used to draw the graph because the DCt parameter is inversely proportional to the gene level; *, an absent or undetectable cycle threshold for target transcript). protein that is regulated by RARRES1 (15); a novel human complex was confirmed by Western blot after TAP, RARRES1 cytosolic carboxypeptidase member of the ABGL/CCP family immunoprecipitation, and reverse AGBL2 immunoprecipita- (AGBL2), and a-tubulin. The presence of AGBL2, Eg5, and EB1 tion (Fig. 2C). Two other interactors, ANKRD26-like family in the exogenously and endogenously expressed RARRES1 member 1A and Crk-like protein kinase, are likely involved in

Table 1. RARRES1 interactome

Confidence % Coverage Accession Number Protein

100 30.8 sp|Q71U36|TBA1A_HUMAN Tubulin alpha-1A chain OS=Homo sapiens GN=TUBA1 A PE=1 SV=1 100 48.5 sp|P46109|CRKL_HUMAN Crk-like protein OS=Homo sapiens GN=CRKL PE=1 SV=1 100 40.5 gi| 166898073 ANKRD26-like family C member 1A (Prostate, ovary, testis-expressed protein on chromosome 2; POTE-2) 100 27.6 sp|Q15691|MARE1_HUMAN Microtubule-associated protein RP/EB Family member 1 OS=Homo sapiens GN=MAPRE1 PE=1 SV=3 100 46.7 sp|P10599|TH!0_HUMAN Thioredoxin OS=Homo sapiens GN=TXN PE=1 SV=3 100 40.4 sp|Q05639|EF1 A2_HUMAN Elongation factor 1-alpha 2 OS=Homo sapiens GN=EEF1A2 PE=1 SV=1 100 18.9 sp|P52732|KIF11_HUMAN Kinesin-like protein KIF11 OS=Homo sapiens GN=KIF11 PE=1 SV=2 98 45.9 sp|Q5U5Z8|CBPC2_HUMAN Cytosolic carboxypeptidase 2 GN=AGBL2 PE=2 SV=1 97 20.4 sp|P49788|TIG 1 _HUMAN Retinoic acid receptor responder protein 1 OS=Homo sapiens GN=RARRES1 PE=2 SV=2

NOTE: List of proteins identified at more than 95% confidence as being present in the RARRES1-TAP produced complex. Proteins that were also present in a complex isolated from cells stably expressing the empty vector control were removed from this list.

www.aacrjournals.org Cancer Res; 71(4) February 15, 2011 1223

Sahab et al.

Figure 3. AGBL2 predicted tertiary structure. A, homology model structure of AGBL2 (ribbon, magenta), and human carboxypeptidase (PDB: 1V77; ribbon, yellow) with their active site residues shown by ball and stick model. The catalytic zinc is shown in orange. B, human carboxypeptidase (PDB: 1V77; ribbon, yellow), with the shape of the catalytic site represented by spacefill atoms (purple). C , homology model structure of AGBL2 (ribbon, magenta), with the shape of the catalytic site represented by spacefill atoms (cyan). D, docking of the EEY peptide into the AGBL2 binding pocket.

the regulation of the membrane-associated actin cytoskeleton tionality once inserted into the membrane is unclear (www. (16). expasy.ch). We addressed RARRES1 orientation by examining its N-glycosylation status via incubation with PNGase F gly- RARRES1 is a type III membrane protein cosidase. Unlike E-cadherin, RARRES1 does not undergo a An affinity purified polyclonal antibody raised to residues 43 shift in molecular weight when exposed to PNGase F (Fig. 2B), to 294 of human RARRES1 protein recognizes an approxi- suggesting that, even though it possesses several putative sites mately 38-kDa protein from PWR-1E whole cell lysate which is for glycosylation, it is not N-glycosylated and that the C- present at much reduced levels in PC-3 cells, consistent with terminal likely faces the cytoplasm and not lumen of the mRNA data. The antibody recognizes TAP-tagged RARRES1, membrane compartment (a type III transmembrane protein). DN-RARRES1, and RARRES1-EYFP at their respective pre- This is consistent with its association with several cytoplasmic dicted molecular masses but does not detect the variant proteins (Table 1). isoform of RARRES1-EYFP. Cell fractionation analysis reveals that approximately 38-kDa RARRES1 is membrane bound; a Regulation of a-tubulin tyrosination second, higher molecular weight pool (50 kDa) is present in Other than its presence in GenBank, nothing is known of the nuclear/cell debris pool, indicating a form which is highly human AGBL2, though a likely family member, mouse Nna-1, modified and/or insoluble in nonionic detergents (Fig. 2B). is important in cerebellar degenerative disorders (17). Nna-1 Although RARRES1 has a transmembrane domain, its direc- knockout mice have high levels of tyrosinated tubulin in the

1224 Cancer Res; 71(4) February 15, 2011 Cancer Research

RARRES1 and Tubulin Tyrosination

Figure 4. RARRES1 and AGBL2 regulate a-tubulin tyrosination in HEK 293 cells. A, AGBL2 immunoblot depicting the full-length protein (104 kDa) and other predicted variants (or degradation fragments) at 72, 64, 38, 33, and 31 kDa. Myc-AGBL2 transfection results in a low amount of full-length exogenous protein expression. AGBL2 knockdown results in the disappearance of all AGBL2 variants and fragments. B, RARRES1 immunoblot depicting exogenous pGlue-RARRES1 (45 kDa) and endogenous RARRES1 (35 kDa), using a mouse monoclonal antibody. C and D, immunoblot for detyrosinated a-tubulin in HEK 293 cells in which RARRES1 or AGBL2 was knocked down or exogenously expressed. Control Myc, control pGlue, Myc-AGBL2, and RARRES1-pGlue transfections were done in duplicate. Loading control was done using a monoclonal anti–a-tubulin antibody. degenerative cerebellum, and it was suggested that Nna-1 RARRES1 (Fig. 4B) were detected using a mouse monoclonal might have the characteristics of a tubulin carboxypeptidase antibody (Ab92884; Abcam). Consistent with their role in (18). However, it is now thought that Nna-1 is involved in the tubulin detyrosination, AGBL2 knockdown and RARRES1 degradation of proteosomally generated peptides (19, 20). overexpression reduced the level of detyrosinated tubulin. Although significant levels of AGBL2 and other family mem- In contrast, knockdown of RARRES1 and exogenous expres- bers exist in HEK 293, PC-3, and PWR-1E cells, Nna-1 sion of AGBL2 increased the level of detyrosinated tubulin, (AGTPBP) was not detectable by quantitative PCR (Fig. 2D) consistent with RARRES1 inhibition of endogenous AGBL2 and neither itself nor the other AGBL family members were activity (Fig. 4C and D). Similar results were found following found in the RARRES1 complex (Table 1). the analysis of detyrosinated and tyrosinated tubulins by mass To gain some insight into its potential carboxypeptidase spectrometry (not shown). These findings were further vali- activity, we predicted AGBL2 structure using human CPA1 dated using immunofluorescent staining of detyrosinated (PDB code: 1V77) as template. Although the overall sequence tubulin (Fig. 5A, Supplementary Fig. S3). The relative abun- identity is only 27% (active site 21%), the predicted tertiary dance of detyrosinated tubulin was significantly upregulated structure of AGBL2 is remarkably similar to CPA1 (Fig. 3A). after RARRES1 knockdown and expression of exogenous Importantly, the residues predicted to interact with the zinc AGBL2, and significantly downregulated after exogenous coordination atom and the folding in this catalytic site is expression of RARRES1 or AGBL2 knockdown, when com- almost identical. Although the structure of the catalytic site is pared with the appropriate control (Fig. 5B and C). Although very similar, the substrate binding cavity of AGBL2 is lengthy knockdown of AGBL2 did not completely abolish detyrosi- and narrow whereas it is bulky and wide for CPA1 (Fig. 3B and nated tubulin, as measured by immunocytochemical quanti- C). These structural predictions strongly indicate that AGBL2 tation of whole cells, examination of individual micrographs is likely to have carboxypeptidase activity. Importantly, exten- reveals a more significant loss of microtubule-associated sive molecular simulations indicate that the putative C-term- detyrosinated tubulin. Diffuse cytoplasmic staining observed inal a-tubulin peptide EEY substrate can be docked precisely under all conditions makes the threshold of this particular into the predicted AGBL2 binding pocket (Fig. 3D). assay rather high in HEK 293 cells. The association between To further test whether AGBL2 has the characteristics of a detyrosinated tubulin and microtubules is shown in Supple- tubulin carboxypeptidase, we carried out Western blots using mentary Figure S3. Interestingly, treatment of HEK 293 cells antibodies specific for the tyrosinated and detyrosinated with paclitaxol does not result in a major increase in the levels forms of a-tubulin on proteins extracted from HEK 293 of detyrosinated tubulin, as observed previously in HeLa cells control cells and cells in which AGBL2 was knocked down. (21). This may be a result of the known drug-resistant phe- Several predicted AGBL2 variants and/or possibly degradation notype of HEK 293 cells, an embryonically derived cell line. fragments are detected in Figure 4A. AGBL2 siRNA resulted in Although immunoblots show an increase in detyrosinated marked knockdown of all AGBL2 forms, except a presumably tubulin on transfection of AGBL2, immunofluorescence stain- nonspecific band at approximately 20 kDa (Fig. 4A). Exogen- ing reveals a more significant increase at the single cell level. ous (45 kDa) and endogenous (38 kDa) long isoforms of Taken together, these data provide strong evidence that

www.aacrjournals.org Cancer Res; 71(4) February 15, 2011 1225

Sahab et al.

Figure 5. RARRES1 and AGBL2 regulate the detyrosination cycle of a-tubulin in HEK 293 cells. A, immunofluorescence signal of Alexa 488 probing the detyrosinated a-tubulin (green) and DAPI probing the nucleus (blue) in HEK 293 controls, RARRES1 knockdown, AGBL2 knockdown, RARRES 1 overexpression, and AGBL2 overexpression with and without taxol treatment. The color balance and the contrast of all the images combined were optimized using Photoshop CS3 to allow for visualization of both DAPI and detyrosinated tubulin staining on an 8-bit resolution monitor. An example of the original image used for quantitation is included in the supplementary materials (Supplementary Fig. S1); B, average signal intensities of detyrosinated a-tubulin in HEK 293 cells following knockdown or exogenous expression of RARRES1 with and without taxol treatment (*, P < 0.05 and #, P < 0.001); C, average signal intensities of detyrosinated a-tubulin in HEK 293 cells, following knockdown or exogenous expression of AGBL2 with and without taxol treatment (*, P < 0.05; **, P < 0.01; #, P < 0.001; ##, P < 0.0001).

1226 Cancer Res; 71(4) February 15, 2011 Cancer Research

RARRES1 and Tubulin Tyrosination

preclude it too from interacting with the lysosomal CPAs. This suggests that its cognate carboxypeptidase(s) is likely active within the cytoplasm. Although little is known about cytoplasmic carboxypeptidases or their substrates, removal of the C- terminal tyrosine of a-tubulin by an unknown carboxypeptidase (the tyrosination cycle) is important in several aspects of microtubule function including kinesin interactions, spindle dynamics, mitosis, and neuronal specification (25). Microtu- bules containing large amounts of detyrosinated a-tubulin are more stable and resistant to depolymerization by nocodazole and other destabilizing agents (26). Detyrosinated a-tubulin is elevated in aggressive breast and prostate cancers which are often resistant to microtubule-targeted chemointerventions Figure 6. Model for RARRES1/AGBL2 function. The C-terminal EEY motif (27). Elevated levels of detyrosinated a-tubulin and D2-tubulin, a of -tubulin is detyrosinated by AGBL2. Tyrosination, in the case of soluble a highly stable version generated from the precursor detyro- a-tubulin, is mediated by tubulin tyrosine ligase (TTL). RARRES1 blocks AGBL2 activity, presumably through direct inhibition, increasing the sinated form, predominate in normal brain tissue but are also proportion of tyrosinated tubulin and altering the accessibility of associated with tumors in other tissues, further suggesting a microtubules (MT) to certain kinesins. role for the tubulin tyrosination cycle in tumorigenesis (28). Recently, a role for detyrosinated tubulin in epithelial to AGBL2 and RARRES1 are a-tubulin tyrosine carboxypeptidase mesenchymal transitions, important in development, stem cell and carboxypeptidase inhibitor, respectively. differentiation, and tumor invasion has been proposed (29). Remarkably, although the tubulin tyrosination cycle has been Discussion known to occur for decades, the identity and regulation of tubulin carboxypeptidase have remained a mystery (30, 31). Previous studies indicated a role for RARRES1 in some Our demonstration that AGBL2 is a RARRES1-interacting aspect of cellular proliferation and tumorigenicity (3). How- protein that regulates the tubulin tyrosination cycle implicates ever, these groups utilized the variant version of RARRES1 in both molecules in the regulation of this evolutionarily ancient their studies (the only form identified at the time), and we process and identifies it as a novel target for intervention have shown that the variant mRNA species is present in very (Fig. 6). small quantities and cannot verify that this transcript is translated into a functional protein. Consequently, it is likely Disclosure of Potential Conflicts of Interest that the major function of RARRES1 is carried out by the full- length isoform. No potential conflicts of interest were disclosed. The RARRES1 paralogue LXN, was initially described as a Acknowledgments carboxypeptidase inhibitor in the rodent brain and regulates hematopoietic stem cell numbers and lifespan in mice (4, 22). The authors thank the Lombardi Cancer Center for the following core RARRES1 also regulates proliferation and differentiation in facilities (NIH P30 CA51008): microscopy, tissue culture, proteomics, genomics, adipose-derived mesenchymal stem cells and proliferation and and epigenomics. motility in HK1 cells, suggesting a role for this family of genes in the differentiation of stem cells from several different tissues Grant Support (23). Recombinant LXN can inhibit the activity of the lysosomal This study was funded by NIH R01CA129813, NIH 1 P01 CA130821, R01 carboxypeptidase A (CPA) family members in vitro and can be DK58196 (S.W. Byers), and NIH U56 LCCC/UDC Partnership (D. Kumar and S.W. cocrystallized with them; however, structural analyses and Byers).). Y. Ji was partly supported by an NIGMS T32 grant. localization patterns of LXN clearly suggest that it is a cyto- The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in plasmic protein and thus it is not likely to interact with accordance with 18 U.S.C. Section 1734 solely to indicate this fact. members of the lysosomal CPA family (24). Because RARRES1 is a type III transmembrane protein, the cytoplasmic localiza- Received June 28, 2010; revised December 13, 2010; accepted December 15, tion of its putative carboxypeptidase inhibitor domain would 2010; published OnlineFirst February 8, 2011.

References 1. Nagpal S, Patel S, Asano AT, Johnson AT, Duvic M, Chandraratna 3. Jing C, El Ghany MA, Beesley C, Foster CS, Rudland PS, Smith P, et al. RA. Tazarotene-induced gene 1 (TIG1), a novel retinoic acid Tazarotene-induced gene 1 (TIG1) expression in prostate carcinomas receptor-responsive gene in skin. J Invest Dermatol 1996;106: and its relationship to tumorigenicity. J Natl Cancer Inst 2002;94:482–90. 269–74. 4. Liang Y, Jansen M, Aronow B, Geiger H, Van Zant G. The quantitative 2. Youssef EM, Chen XQ, Higuchi E, Kondo Y, Garcia-Manero G, Lotan trait gene latexin influences the size of the hematopoietic stem cell R, et al. Hypermethylation and silencing of the putative tumor sup- population in mice. Nat Genet 2007;39:178–88. pressor Tazarotene-induced gene 1 in human cancers. Cancer Res 5. Arimatsu Y. Latexin: a molecular marker for regional specification in 2004;64:2411–7. the neocortex. Neurosci Res 1994;20:131–5.

www.aacrjournals.org Cancer Res; 71(4) February 15, 2011 1227

Sahab et al.

6. Hatanaka Y, Uratani Y, Takiguchi-Hayashi K, Omori A, Sato K, Miya- 18. Kalinina E, Biswas R, Berezniuk I, Hermoso A, Aviles FX, Fricker LD. A moto M, et al. Intracortical regionality represented by specific novel subfamily of mouse cytosolic carboxypeptidases. FASEB J transcription for a novel protein, latexin. Eur J Neurosci 1994;6: 2007;21:836–50. 973–82. 19. Chakrabarti L, Neal JT, Miles M, Martinez RA, Smith AC, Sopher BL, et al. 7. Normant E, Gros C, Schwartz JC. Carboxypeptidase A isoforms ThePurkinjecelldegeneration5Jmutationisasingleaminoacidinsertion produced by distinct genes or alternative splicing in brain and other that destabilizes Nna1 protein. Mamm Genome 2006;17:103–10. extrapancreatic tissues. J Biol Chem 1995;270:20543–9. 20. Berezniuk I, Sironi J, Callaway MB, Castro LM, Hirata IY, Ferro ES, 8. Normant E, Martres MP, Schwartz JC, Gros C. Purification, cDNA et al. CCP1/Nna1 functions in protein turnover in mouse brain: cloning, functional expression, and characterization of a 26-kDa Implications for cell death in Purkinje cell degeneration mice. FASEB endogenous mammalian carboxypeptidase inhibitor. Proc Natl Acad J 2010;24;1813–23. Sci U S A 1995;92:12225–9. 21. Fonrose X, Ausseil F, Soleilhac E, Masson V, David B, Pouny I, et al. 9. Orford K, Crockett C, Jensen JP, Weissman AM, Byers SW. Serine Parthenolide inhibits tubulin carboxypeptidase activity. Cancer Res phosphorylation-regulated ubiquitination and degradation of beta- 2007;67:3371–8. catenin. J Biol Chem 1997;272:24735–8. 22. Arimatsu Y, Ishida M. Early patterning of the rat cerebral wall for 10. Sahab ZJ, Man YG, Semaan SM, Newcomer RG, Byers SW, Sang regional organization of a neuronal population expressing latexin. QXA. Alteration in protein expression in estrogen receptor alpha- Brain Res Dev Brain Res 1998;106:71–8. negative human breast cancer tissues indicates a malignant and 23. Kwok WK, Pang JC, Lo KW, Ng HK. Role of the RARRES1 gene in metastatic phenotype. Clin Exp Metastasis 2010;27:493–503. nasopharyngeal carcinoma. Cancer Genet Cytogenet 2009;194:58–64. 11. Sahab ZJ, Iczkowski KA, Sang QXA. Anion exchange fractionation of 24. Takiguchi-Hayashi K, Arimatsu Y. Restricted expression of latexin in serum proteins versus albumin elimination. Anal Biochem 2007; dorsal midline cells of developing rat forebrain. Neuroreport 1995; 368:24–32. 6:281–3. 12. Sahab ZJ, Suh Y, Sang QXA. Isoelectric point-based prefractionation 25. Konishi Y, Setou M. Tubulin tyrosination navigates the kinesin-1 motor of proteins from crude biological samples prior to two-dimensional gel domain to axons. Nat Neurosci 2009;12:559–67. electrophoresis. J Proteome Res 2005;4:2266–72. 26. Khawaja S, Gundersen GG, Bulinski JC. Enhanced stability of micro- 13. Sahab ZJ, Semaan SM, Sang QXA. Methodology and applications of tubules enriched in detyrosinated tubulin is not a direct function of disease biomarker identification in human serum. Biomark Insights detyrosination level. J Cell Biol 1988;106:141–9. 2007;2:21–43. 27. Mialhe A, Lafanechere L, Treilleux I, Peloux N, Dumontet C, Bremond 14. Aagaard A, Listwan P, Cowieson N, Huber T, Ravasi T, Wells CA, et al. A, et al. Tubulin detyrosination is a frequent occurrence in breast An inflammatory role for the mammalian carboxypeptidase inhibitor cancers of poor prognosis. Cancer Res 2001;61:5024–7. latexin: relationship to cystatins and the tumor suppressor TIG1. 28. Lafanechere L, Job D. The third tubulin pool. Neurochem Res Structure 2005;13:309–17. 2000;25:11–18. 15. Sahab ZJ, Hall MD, Zhang L, Cheema AK, Byers SW. Tumor sup- 29. Whipple RA, Matrone RA, Cho EH, Balzer EM, Vitolo ME, Yoon JR, pressor RARRES 1 regulates DLG2, PP2A, VCP, EB1, and Ankrd26. J et al. Epithelial-to-mesenchymal transition promotes tubulin fetyrosi- Cancer 2010;1:14–22. nation and microtentacles that enhance endothelial engagement. 16. Hanks SK, Polte TR. Signaling through focal adhesion kinase. Bioes- Cancer Res 2010;70:8127–37. says 1997;19:137–45. 30. Argarana CE, Barra HS, Caraway H. Release of [C-14]tyrosine from 17. Fernandez-Gonzalez A, La Spada AR, Treadaway J, Higdon JC, Harris tubulinyl-[C-14]tyrosine by brain extract—separation of a carboxypep- BS, Sidman RL, et al. Purkinje cell degeneration (pcd) phenotypes tidase from tubulin-tyrosine ligase. Mol Cell Biochem 1978;19:17–21. caused by mutations in the axotomy-induced gene, Nna1. Science 31. Barra HS, Arce CA, Argarana CE. Posttranslational tyrosination/detyr- 2002;295:1904–6. osination of tubulin.Mol Neurobiol 1988;2:133–53.

1228 Cancer Res; 71(4) February 15, 2011 Cancer Research

Tumor Suppressor RARRES1 Interacts with Cytoplasmic Carboxypeptidase AGBL2 to Regulate the α-Tubulin Tyrosination Cycle

Ziad J. Sahab, Michael D. Hall, You Me Sung, et al.

Cancer Res 2011;71:1219-1228. Published OnlineFirst February 8, 2011.

Updated version Access the most recent version of this article at: doi:10.1158/0008-5472.CAN-10-2294

Supplementary Access the most recent supplemental material at: Material http://cancerres.aacrjournals.org/content/suppl/2011/02/07/0008-5472.CAN-10-2294.DC1

Cited articles This article cites 31 articles, 9 of which you can access for free at: http://cancerres.aacrjournals.org/content/71/4/1219.full#ref-list-1

Citing articles This article has been cited by 8 HighWire-hosted articles. Access the articles at: http://cancerres.aacrjournals.org/content/71/4/1219.full#related-urls

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://cancerres.aacrjournals.org/content/71/4/1219. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.