PTENβ is an alternatively translated isoform of PTEN that regulates rDNA transcription

PTEN is a critical tumour suppressor that is frequently mutated in human cancer. We have previously identified a CUG initiated PTEN isoform designated PTENα, which functions in mitochondrial bioenergetics. Here we report the identification of another N-terminal extended PTEN isoform, designated PTENβ. PTENβ translation is initiated from an AUU codon upstream of and in-frame with the AUG initiation sequence for canonical PTEN. We show that the Kozak context and a downstream hairpin structure are critical for this alternative initiation. PTENβ localizes predominantly in the nucleolus, and physically associates with and dephosphorylates nucleolin, which is a multifunctional nucleolar phosphoprotein. Disruption of PTENβ alters rDNA transcription and promotes ribosomal biogenesis, and this effect can be reversed by re-introduction of PTENβ. Our data show that PTENβ regulates pre-rRNA synthesis and cellular proliferation. These results demonstrate the complexity of the PTEN protein family and the diversity of its functions.

T he PTEN gene was first identified and cloned in 1997, and PTEN protein has proved to be a powerful tumour suppressor 1,2 . It is one of the most frequently mutated or deleted genes in cancer, including cancer of the brain, breast, prostate, pancreas and ovary 3 . Germline mutations of the PTEN gene lead to inherited autosomal dominant hamartoma tumour syndromes, including Bannayan-Zonana syndrome 4 and Cowden disease 5 . Heterozygous Pten deletion leads to multiple tumours in mice, while the homozygote is embryonically lethal 6,7 . In the cytoplasm, PTEN antagonizes the PI3K-AKT-mTOR pathway by dephosphorylation of phosphoinositide-3,4,5-triphosphate (PIP3) 8 . Loss of PTEN function leads to uncontrolled activation of the PI3K/AKT pathway, which stimulates cell growth, cell proliferation and cell survival, and inhibits apoptosis. PTEN also accumulates in the cell nucleus through passive diffusion, Ran-MVP-mediated transport and mono-ubiquitinationregulated mechanisms 9 . In the nucleus, PTEN maintains genome integrity and regulates the process of DNA replication through regulation of RPA1, MCM2 and TOP2A (refs 10-13). In addition to its tumour suppressor function, PTEN is also involved in embryonic development, lipid metabolism 14,15 , Alzheimer's disease 16 and antiviral innate immunity 17 . However, the current concepts of PTEN function do not fully explain the multifarious activity of this protein.
In eukaryotes, the translation initiation codon is generally recognized and selected by the 5 0 cap structure dependent ribosomal scanning mechanism, as well as by 5 0 cap structure independent internal ribosome entry sites. This scanning complex rigorously controls the fidelity of initiation through recognition of the correct AUG triplet in the optimum context GCC(A/G)CCAUGG, referred to as the Kozak sequence, which includes a purine at the À 3 and a G at the þ 4 position (relative to the A of the AUG codon, which is designated þ 1) (refs 18,19).
Triplets that differ from AUG by only one nucleotide can also direct initiation of polypeptide chain synthesis in mammalian cells in an optimum context 20 . It is also reported that recognition of non-AUG codons is strongly stimulated by a downstream stem-loop (hairpin) structure that is separated from the preceding initiation codon by about 14 nucleotides 21 . Isoforms initiated from non-AUG codons are frequently endowed with functions additional to and differing from the canonical form of a given protein 22,23 .
Previously, we and others reported that a CUG codon in the 5 0 untranslated region (5 0 UTR) of PTEN mRNA initiates an N-terminal extended isoform of PTEN now known as PTENa (refs 24,25). In the current study we describe another PTEN isoform designated PTENb, which comprises an extended N-terminal extension of 146 amino acids (Homo sapiens) as compared with canonical PTEN protein. Our data show PTENb localizes specifically in the nucleolus and negatively regulates rDNA transcription. Disruption of PTENb leads to an abnormal increase in pre-rRNA synthesis and results in promotion of cellular proliferation. These findings characterize an N-terminal extended PTEN isoform with distinct localization and function, which results from alternative AUU translation initiation. Identification of PTENb further reveals the diversity of PTEN protein family function, and the correspondence of protein function with isoform subtype.

Results
An unidentified protein is recognized by a PTENa antibody. The protein PTENa functions in mitochondrial metabolism, and structurally distinct compounds such as acriflavin and aurin tricarboxylic acid (ATA) yield different results in regulation of PTENa synthesis initiated by a CUG start codon 25 . The CUG inhibitor acriflavin reduces PTENa expression in a dose-dependent manner without affecting canonical PTEN expression. We observed that a full length PTEN antibody reacted with an unidentified protein of molecular weight lower than PTENa (around 70 kDa), and expression of this protein showed dose dependent variation under acriflavin treatment parallel to that of PTENa (Fig. 1a). A panel of cancer cell lines was examined to determine whether differences in PTEN status affect this unidentified PTEN-like protein, and the protein was undetectable in PTEN-null PC3 cells (Fig. 1b). This protein was also recognized by a monoclonal antibody against the PTEN C-terminal domain, suggesting its C-terminal region is similar to or identical with canonical PTEN. In order to confirm this, we examined cancer cell lines with an anti-PTENa antibody raised in our laboratory by immunization with purified recombinant protein composed of the 173 N-terminal amino acids of PTENa (ref. 25). This unidentified protein was recognized by this anti-PTENa antibody (Fig. 1c), indicating that it was most likely another N-terminally extended PTEN variant generated by initiation at an in-frame initiation codon within the PTEN mRNA 5 0 leader. We designated this PTEN-like protein PTENb.
Alternative initiation of PTENb translation from AUU 594 . Evaluation of the 5 0 UTR of human PTEN mRNA for alternative translation start sites in frame with the AUG 1032 start codon revealed a total of four non-AUG alternative initiation codons in favourable Kozak contexts (Fig. 2a). Translation from AUU 594 or UUG 621 is expected to encode larger forms of PTEN of 549 or 540 amino acids respectively, with predicted molecular weights of 61.6-62.9 kDa. A protein of this size should migrate under gel electrophoresis at around 70 kDa, matching the observed PTENb band.
To determine whether one of these non-AUG codons initiates translation of this putative PTENb protein with upstream extension of the open reading frame (ORF), the coding sequence of PTENa was cloned into a plasmid with a C-terminal GFP tag (Fig. 2b, upper panel). As expected, the CUG 513 initiated PTENa ORF was translated into three distinct proteins with masses of 120 þ kDa (PTENa-GFP), 120 kDa (PTENb-GFP) and 90 kDa (PTEN-GFP) (Fig. 2b, lower panel). This suggested that PTENb is translated from a new ORF in the 5 0 UTR of PTEN mRNA upstream of the canonical AUG start codon.
To identify PTENb translation initiation sites, a plasmid was constructed to express CUG 513 -initiated PTENa with a C-terminal GFP tag, and mutations were created in this plasmid to determine which non-AUG codon is critical for PTENb expression (Fig. 2c). Mutation of CUG 513 into a non-initiating CUC triplet led only to disappearance of the most slowly migrating proteoform PTENa, thus excluding the possibility that PTENb is a cleavage product (Fig. 2d, lane 4). On the basis of CUG 513 mutation, potential non-AUG codons (AUU 594 , UUG 621 ) were mutated separately to a non-initiating CUC triplet. Anti-GFP immunoblots from these transfected lysates revealed that mutation of AUU 594 abolishes PTENb, whereas mutation of UUG 621 has no such effect (Fig. 2d, lane 6 versus lane 5). These results indicate that AUU 594 but not UUG 621 is essential for PTENb expression, and that AUU 594 is likely the translation initiation site for PTENb.
AUU 594 resides in a favourable Kozak context as noted (GUCACCAUU 594 ) (Fig. 2a). In order to determine whether this context influences translation initiation at AUU 594 , we constructed mutants to disrupt it and examined PTENb expression (Fig. 2e). Disruption of this Kozak context markedly reduced PTENb expression (Fig. 2f), strongly suggesting that translation initiation of PTENb is reliant on the Kozak context. Detailed evaluation of the 5 0 UTR of human PTEN mRNA revealed a 12 bp perfect palindromic sequence starting 18 bp downstream of AUU 594 , and phylogenetic analysis suggests that this secondary structure is evolutionarily conserved (Fig. 2g). As shown in Fig. 2h, disruption of this palindromic motif by mutation led to a marked reduction of PTENb without affecting canonical PTEN expression (lower panel, lane 4 versus lane 3), indicating this sequence is critical for translation initiation of PTENb.
These data demonstrate that PTENb synthesis relies on an alternative translation initiation at the AUU 594 codon, and expression of PTENb is influenced both by the Kozak context and by the secondary structure downstream of AUU 594 .
Mass spectrometry analysis of the PTENb sequence. Mass spectrometry was used for peptide sequencing to validate the PTENb translation initiation site. Purified human PTENb with a C-terminal His tag (Fig. 3a) was expressed in Sf9 insect cells for mass spectrometry (Fig. 3b). Mascot reports revealed three peptide fragments covering 39.04% of the N-terminal region of PTENb (from AUU 594 to AUG 1032 ) (Fig. 3b). LC-MS/MS captured MSRAGNAGE, which is the most proximal N-terminal peptide of PTENb (9 aa, MS/MS spectrum shown in Fig. 3c). Additional analysis showed that the first N-terminal amino acid of PTENb is acetylated. These mass spectrum data therefore verified AUU 594 is the PTENb translation initiation site.
The 5 0 UTR of Homo sapiens PTEN and Mus musculus Pten are highly homologous as shown in our recent study 25 , suggesting that PTENb expression in mice is conserved and has a similar alternative translation initiation mechanism. We previously generated a FLAG-knockin mouse model (Pten FLAG ), in which the C-terminus of the Pten gene was targeted for insertion of a FLAG-coding sequence 25 . To verify PTENb resides at the Pten gene locus, tissue samples were extracted from heterozygous Pten FLAG mice and wild-type control mice for FLAG pull-down followed by immunoblotting with an anti-PTENa antibody raised in our laboratory 25 . A protein band with a molecular weight lower than PTENa which was similar to the PTENb band identified in cancer cell lines was detected in the FLAG elute from Pten FLAG tissues, while no band was detected in wild-type tissues (Fig. 3d). The existence of in vivo PTENb was further confirmed by mass spectrometric analysis of a FLAG-purified 70 þ kDa band (molecular weight lower than PTENa) from Pten FLAG tissues. Mascot reports revealed four peptide fragments that cover 68.5% of the N-terminal region of PTENb (Fig. 3e). These results verified the existence of PTENb in vivo, and confirmed that PTENb and PTENa are translated in vivo from identical mRNA at the same gene locus. This Pten FLAG knockin animal model demonstrates the natural occurrence of alternative initiation that results in translation of PTENb.   PTENb is localized predominantly in the nucleolus. Longer protein isoforms resulting from alternative translation initiation frequently contain a signal for subcellular localization that is absent in shorter forms 26,27 . The length of the PTENb protein differs from PTENa and PTEN due to its N-terminal extension, which raised the possibility that PTENb has unique subcellular localization and special function. Immunofluorescence examination of the subcellular localization of PTENb was compared with PTENa and PTEN. Separate plasmids expressing C-terminal GFP-tagged PTEN, PTENa or PTENb were constructed, and to ensure expression of only a single isoform in each instance, we created an ATG 1032 4ATA mutation in GFP-tagged PTENa and PTENb (Fig. 4a). The subcellular distribution of PTENb was completely different from that of PTEN and PTENa as shown in Fig. 4b. PTEN was distributed ubiquitously in both the nucleus and cytoplasm, while PTENa predominately showed cytoplasmic localization. In contrast, strong PTENb fluorescence signals were distributed mainly in the relatively shallow DAPI staining areas of the nucleus which have been suggested to be nucleolar regions 28 , and little cytoplasmic PTENb localization was detected. In order to exclude the possibility protein localization was influenced by the GFP tag, PTEN, PTENa and PTENb expression plasmids were constructed without a tag ( Supplementary Fig. 1a), and immunofluorescence with PTEN monoclonal antibody in PTEN null cells revealed that these untagged isoforms have subcellular distribution patterns identical to the GFP-tagged molecules ( Supplementary Fig. 1b,c). UBF that is a well-known nucleolar marker was used to evaluate for co-localization of C-terminal GFP tagged PTENa, PTENb or PTEN within the nucleolus. There was prominent co-localization of PTENb with this marker in the nucleolus, but considerably less nucleolar co-localization was detected with PTEN or PTENa (Fig. 4c). This extensive distribution of PTENb in the nucleolus was confirmed through cell fractionation (Fig. 4d). Moreover, we demonstrated that the N-terminal extended sequence was solely responsible for PTENb accumulation in the nucleolus, and this was also true for PTENa localization on mitochondria as we reported previously ( Supplementary Fig. 2).
The nucleolar-localization signals (NoLSs) that govern nucleolar localization and retention are usually rich in arginine and lysine residues 29,30 . We noted a series of poly-arginine residues in the N-terminal extended domain of PTENb (aa 20-25), which is evolutionarily conserved (Supplementary Fig. 3). To determine whether this sequence acts as an NoLS and directs PTENb nucleolar localization, we established a PTENbDR 6 construct in which these six arginines were deleted. This DR 6 mutation was remarkable in that it completely abolished PTENb nucleolar accumulation (Fig. 4e), indicating that this poly-arginine sequence is critical for PTENb nucleolar localization, and showing that it functions as an NoLS. These results indicated that the distinctive N-terminal extension sequence of PTENb confers unique cellular localization, and raised the possibility that PTENb is involved in nucleolar function.
Nucleolin is a novel downstream target of PTENb. In order to identify potential PTENb targets and explore PTENb function in the nucleolus, mass spectrometry was used to analyse the results of S-tag-PTENb pull down. Results with S-tag-PTEN and S-tag-PTENa were analysed separately as controls. Mass spectrometry revealed that PTENb specifically interacts with a large number of nucleolar proteins ( Supplementary Fig. 4a), including nucleolin that is the most abundant multifunctional phosphoprotein of the nucleolus. Nucleolin is comprehensively involved in ribosome biogenesis 31,32 , and was thus considered to be a potential PTENb target (Fig. 5a). Interaction of PTENb and nucleolin was demonstrated by reciprocal immunoprecipitation (Fig. 5b,c). In addition, immunofluorescence showed extensive co-localization of PTENb with nucleolin in the nucleolus, whereas PTEN or PTENa did not (Fig. 5d), further confirming the specificity of the interaction of PTENb and nucleolin. To identify the functional domains by which PTENb binds to nucleolin, two truncated forms of PTENb were constructed with C-terminal S-tags ( Supplementary Fig. 5a). Immunoprecipitation demonstrated that the N-terminal extension domain of PTENb is essential for interaction with nucleolin ( Supplementary Fig. 5b). These data show PTENb and nucleolin co-localize and interact each other.
Like canonical PTEN, PTENb contains an intact phosphatase domain, and we found that exogenous expression of PTENb, PTENa or PTEN in PTEN null cells all efficiently reduce p-AKT to a similar level ( Supplementary Fig. 6a). Moreover, missense mutations in the phosphatase domain of PTENb (G275E, analogous to PTEN (G129E)) and (C270S, analogous to PTEN (C124S)) both eliminated the lipid phosphatase activity of the protein ( Supplementary Fig. 6b). PTENb thus has phosphatase activity that is presumably similar to PTENa and canonical PTEN. In addition, we found that affinity for nucleolin binding was largely diminished in protein phosphatase activity deficient PTENb mutants (Y284L and C270S) (Fig. 5e, lane 3 and lane 5 versus lane 2 and lane 4), suggesting that PTENb protein phosphatase activity is necessary for interaction with nucleolin. This further indicated PTENb acts as a nucleolin phosphatase.
In order to clarify the relationship of PTENb with nucleolin and evaluate PTENb function in the nucleolus, somatic PTENb knockout was established with CRISPR-Cas9. CRISPR-Cas9mediated gene targeting eliminated PTENa and PTENb without affecting canonical PTEN expression, and we also designed a specific target sequence for sole knockout of PTENa as a control   PTENα P t e n + /+ P t e n + /+ P t e n + /+ M a r k e r P t e n + /+ P t e n + /+ P t e n + /+ P t e n + /+ P t e n + /+ P t e n + /+ P t e n + /+ P t e n + /+ P t e n F L A G /+ P t e n F L A G /+ P t e n F L A G /+ P t e n F L A G /+ P t e n F L A G /+ P t e n F L A G /+ P t e n F L A G /+ P t e n F L A G /+ P t e n F L A G /+ P t e n F L A G /+ P t e n F ( Fig. 5f,g). We found that PTENa and PTENb double knockout led to elevation of nucleolin phosphorylation levels at Thr84, which is one of the XTPXKKXX motifs located in the nucleolin N-terminal region important for nucleolin function (Fig. 5h). Moreover, exogenous expression of wild-type PTENb but not protein phosphatase activity abolished PTENb efficiently downregulated nucleolin phosphorylation levels at Thr84 (Fig. 5i). These results raised the possibility that PTENb acts as a nucleolin phosphatase. In vitro dephosphorylation assays further verified Cdk1/CyclinB1-phosphorylated nucleolin Thr84 is a target of PTENb for dephosphorylation (Fig. 5j, lane 5 versus lane 4). In addition, we found that the poly-arginine sequence deleted mutant that eliminated PTENb nucleolar localization also effectively abolishes its phosphatase activity on nucleolin. This indicates that PTENb nucleolin phosphatase activity is linked to its localization in the nucleolus (Supplementary Fig. 7). These results argue that PTENb acts as a nucleolin phosphatase, and show that the PTENb-nucleolin interaction results in dephosphorylation modification of nucleolin, which influences its function.

PTENb inhibits rDNA transcription and cell proliferation.
Nucleolin is involved in almost every step of ribosome biogenesis [33][34][35] , and a recent study showed the most prominent effect of nucleolin is regulation of rDNA transcription by promotion of the rDNA transcription complex assembly, in which the phosphorylation status of nucleolin is rigorously regulated 36,37 .
The fact that PTENb interacts with nucleolin and dephosphorylates nucleolin at Thr84 raised a question as to whether nucleolin mediates PTENb regulation of rDNA transcription, and thus represents a direct mechanism by which PTENb controls ribosomal biogenesis.
To determine the influence of PTENb on rRNA synthesis, total rRNA levels were evaluated by gel electrophoresis. Knockout of PTENa and PTENb without affecting PTEN expression led to elevation of 28S and 18S total rRNA levels, whereas knockout of PTENa alone had no such effect (Fig. 6a, lane 3 versus lane 2 and lane 1). Moreover, exogenous expression of PTENb but not PTENa in PTENa and PTENb null cells efficiently downregulated 28S and 18S rRNA levels (Fig. 6b, lane 3 versus lane 2 and Supplementary Fig. 8a). These results argue that PTENb plays a role in regulation of rRNA, and in particular is involved in regulation of pre-rRNA synthesis from the outset of ribosome biogenesis. Transcription levels of rRNA genes were therefore evaluated by analysing pre-rRNA with qPCR in PTENb wild-type and knockout cells. As expected, 45S pre-rRNA was increased in PTENb and PTENa double knockout cells as compared with PTENa only knockout or wild-type cells (Fig. 6c).
To determine whether rDNA transcription is regulated by PTENb, a pHrD-IRES-Luc or pGL3-basic control vector was transfected into PTENb and PTENa double knockout cells, and the luciferase reporter activity of the rDNA promoter was measured. Introduction of PTENb, but not PTENa, efficiently downregulated luciferase activity ( Fig. 6d and Supplementary  Fig. 8b). These results indicate that PTENb negatively regulates rDNA transcription, which may act at least in part through PTENb dephosphorylation of nucleolin.
Nucleolar morphology is closely related to ribosome biogenesis, and enlarged nucleoli are typically found in association with activation of ribosome biogenesis. To evaluate alterations in nucleolar morphology, silver staining was used for the identification of argyrophilic nucleolar organizer regions (AgNOR). PTENb and PTENa double knockout resulted in greater average areas of AgNOR staining, while PTENa only knockout did not have this effect (Fig. 6e). This suggested that elimination of PTENb leads to upregulation of ribosomal biogenesis. Stimulation of cell proliferation is a characteristic biologic feature of nucleolin, and is typically associated with elevated ribosomal biogenesis. We therefore investigated biologic changes resulting from regulation of ribosomal biogenesis by PTENb. Growth curves showed significant increases in cell growth rate resulting from PTENb knockout (Fig. 6f). This activation of cell proliferation resulting from PTENb deprivation was further confirmed by S phase analysis (Fig. 6g). These results argue that PTENb controls cellular proliferation and may act through a mechanism involving negative regulation of pre-rRNA synthesis.

Discussion
In eukaryotic cells, the ribosome typically initiates protein synthesis at the AUG codon closest to the 5 0 end of the mRNA 38 . However, there is now evidence including observations from our previous study 25 that demonstrates initiation of protein translation can occur at most codons which differ from AUG by no more than a single nucleotide (non-AUG initiation) 20 . More than 50 instances of non-AUG-initiated N-terminal extensions have been predicted or verified experimentally in mammals 39,40 . In this study, we report identification of a translational variant of PTEN, designated PTENb. This protein is synthesized from an alternative translation start site 438 bp upstream of the AUG initiation sequence, which adds 146 N-terminal amino acids to the canonical PTEN ORF. We found that PTENb specifically localizes in the nucleolus and inhibits pre-rRNA synthesis, and our experimental evidence strongly suggests this inhibition acts through the mechanism of nucleolin dephosphorylation.
In previous studies both our group and Hopkins et al. independently identified a translational variant of PTEN with N-terminal extension, designated 'PTENa' in our study, and 'PTEN-long' by Hopkins et al. 24 . These have proven to be the same isoform that is translated from a CUG codon 519 bp upstream of the AUG initiation sequence of canonical PTEN (refs 24,25). In addition to the initiation sites for PTENa and PTENb (CUG 513 , AUU 594 ), we found another two alternative non-AUG codons in favourable Kozak contexts, which may be sites for initiation of other as yet unidentified isoforms. As multiple additional forms of PTEN may exist, we consider it to be prudent to designate PTEN isoforms as a sequential series, using a, b, g, d as labels.
In eukaryotes, the efficiency of initiation depends on the Kozak context 41 which encompasses the AUG codon. Our data show that initiation of PTENb relies on the Kozak context surrounding AUU 594 , which suggests this context is also essential for efficient initiation of other non-AUG codons. We identified a 12 bp perfect palindromic sequence starting 18 bp downstream of AUU 594 , and showed that disruption of this conserved palindrome sharply downregulates expression of PTENb. The crucial role this downstream palindromic sequence plays in AUU-mediated initiation, together with discovery of the fact in our previous study that the CUG centred palindromic sequence may be a signature motif for alternative Leu-tRNA initiation 25 , underscores the importance of secondary structure for non-AUG initiation. This is consistent with the concept that a strong secondary structure downstream of the non-AUG codon may significantly increase initiation efficiency as reported by Kozak 42 . We previously showed that PTENa localizes in mitochondria and functions in mitochondrial metabolism 25 . Hopkins et al. reported that PTEN-long (designated PTEN-L) is secreted into adjacent cells, which relies on an N-terminal secretion sequence 24 . These findings indicate that the N-terminal extended PTENa isoform is endowed with capability for subcellular localization differing from canonical PTEN. Data in the current study show that there is partial co-localization of PTEN and PTENa in the cytoplasm, and this is consistent with findings in our previous study that PTEN interacts and co-localizes with PTENa in mitochondria 25 . However, PTEN isoforms localize for the most part in a mutually exclusive manner (Supplementary Fig. 9). PTENb specifically localizes in the nucleolus, whereas PTEN and PTENa do not, and an evolutionarily conserved poly-arginine sequence has proven to be an NoLS of PTENb. The PTENb N-terminus is 27 amino acids shorter than PTENa, and PTENb therefore does not contain the N-terminal secretion signal sequence, which may in part account for the differences in cellular distribution of PTENb and PTENa. Moreover, protein conformation may be altered by differences in the length of the N-terminal domain, leading to differing affinities for interacting proteins or nucleotide sequences 27,43 . This may in some measure explain the diversity of PTEN isoform distribution.
Our data demonstrate that PTENb is involved in regulation of pre-rRNA synthesis, and disruption of PTENb results in elevated ribosomal biogenesis and cellular proliferation. Our findings suggest that this PTENb regulatory mechanism may act through dephosphorylation of nucleolar protein nucleolin. It has been reported that PTEN acts on ribosomal biogenesis through the PI3K-mTOR pathway by downregulation of 45S pre-rRNA and that PTEN localizes to the nucleolus to regulate nucleolar homeostasis [44][45][46] . However, our results indicate there is extensive distribution of PTENb in the nucleolus while little nucleolar PTEN can be detected. The PTENb coding sequence overlaps the sequence of full-length canonical PTEN and PTENb can be efficiently recognized by PTEN antibodies. It is therefore possible that the nucleolar immunofluorescence signals detected by the PTEN monoclonal antibody in this study are derived from recognition of this previously unidentified PTENb isoform instead of from canonical PTEN. Furthermore, it must be noted that most of the observations from the studies cited above were made in PTEN knockout cells or with Pten deficient mice. As PTEN and its isoforms PTENa and PTENb are coded by the same gene, current PTEN knockout mouse models are essentially equivalent to at least triple deletion of PTEN and its isoforms. The phenotypic deficiencies of PTEN knockout models may thus be partially attributed to loss of PTEN or one of its isoforms.
By use of specific disruption of PTENb, our study reveals that PTENb plays an essential role in rDNA transcription and regulation of cellular proliferation. Several recent studies have described a diversity of nucleolar functions in addition to ribosome biogenesis, such as modification of small RNAs, regulation of the cell cycle and control of aging 47,48 , and nucleolar perturbations have been observed in various diseases ranging from auto-immunity to cancer 49,50 . Our results show that in addition to nucleolin, PTENb may physically interact with a number of other nucleolar proteins (Supplementary Fig. 4b), which raises the possibility that it participates in nucleolar functions apart from ribosome biogenesis.
In this study, we identify PTENb as a new isoform of PTEN, which negatively regulates rDNA transcription and cellular proliferation. Based on previous studies that show imbalance of rDNA transcription closely correlates with tumour progression 51,52 , identification of PTENb may provide a new track for therapeutic drug design. These results together with our previous observations regarding PTENa illustrate the complexity of PTEN function and enrich our understanding of the PTEN family. Mice. Tissues of C57BL/6 mice 25 were lysed in lysis buffer (50 mM Tris-pH 7.5, 150 mM NaCl, 1% NP40) and subjected to immunoblotting or immunoprecipitation. All animals were maintained in a special pathogen-free facility, and the animal study protocols used were approved by the ethics committee of Peking University Health Science Center (approval number bjmu20110301).

Methods
Antibodies and reagents. Primary antibodies used for western blot, immunoprecipitation and immunofluorescence are listed in Supplementary Table 1. The anti-PTENa (ref. 25) and anti-nucleolin antibodies were raised in our laboratory ( Supplementary Fig. 10). The reagent Acriflavin was from Sigma, A8126.
Plasmids and cloning strategies. The plasmids pCMV-tag-2b, pSA, pEGFP-N1, pDsRed2-N1 and pFastBac1 were purchased from Addgene. pX330 was purchased from BEIJING IDMO CO., LTD. pcDNA3.1-Cas9 was a gift from Dr Jianzhong Xi in the College of Engineering, Peking University. PTEN, PTENa, PTENb, or nucleolin or truncations and mutations of these molecules were inserted into these plasmids. pHrD-IRES-Luc was a gift from Dr Xiaojuan Du in the Department of Cell Biology, Peking University Health Science Center 53 . (b) C-terminal GFP-tagged PTENb or GFP-tagged mock was transfected, followed by immunoprecipitation with GFP antibody and immunoblotting with an anti-nucleolin antibody. (c) FLAG-tagged nucleolin or mock plasmid was co-transfected with GFP-tagged PTENb, followed by immunoprecipitation with FLAG antibody and immunoblotting with GFP antibody. (d) Images of immunofluorescence staining for PTENb and nucleolin. Hela PTEN null cells were transfected as in Fig. 4b, followed by staining with an anti-nucleolin antibody, and were imaged by confocal microscopy. The scale bars represent 5 mm. (e) PTENb protein phosphatase activity is required for its interaction with nucleolin. Wild-type PTENb or PTENb mutants as indicated in Supplementary Fig. 6b was transfected separately, followed by immunoprecipitation with GFP antibody and immunoblotting with an anti-nucleolin antibody.   Mass spectrometry. For in vitro PTENb identification experiments, the PTENb coding sequence (CUG 513 -UGA 2240 , PTENa initiation codon CUG 513 mutated to CUC and canonical PTEN initiation codon AUG 1032 mutated to AUA) was subcloned with a C-terminal His-tag into a pFastBac1 vector for expression in Sf9 insect cells. His affinity chromatography was used for purification of PTENb-His before SDS-PAGE resolution of a 70-kDa band. Immunoprecipitated proteins were digested in-gel by endoproteinase Glu-C (Promega) following the manufacturer's instructions. Peptides were separated by online reversed-phase nanoscale capillary liquid chromatography (Easy-nLC 1,000, Thermo Scientific). The data dependent mass spectra were acquired with the LTQ-Orbitrap Elite mass spectrometer (Thermo Scientific) equipped with a nanoelectrospray ion source (Thermo Scientific). Raw files were searched by Proteome Discoverer (Version 1.4.1.14, Thermo Scientific) against the UniProt Human database supplemented with N-terminal extended PTEN sequence ( Supplementary  Fig. 11a). Search parameters were set as follows: enzyme with semi-Glu-C; up to two missed cleavages; carbamidomethyl cysteine as fixed modification; methionine oxidation and peptide N-terminal acetylation as variable modifications. The false discovery rates were set at 0.01 (ref. 25). For endogenous PTENb identification experiments, a FLAG-purified 70 kDa band (molecular weight lower than PTENa) from Pten FLAG tissues was treated in the same way as described above. Raw files were searched with Proteome Discoverer (Version 1.4.1.14, Thermo Scientific) against the UniProt Mouse database supplemented with the N-terminal extended PTEN sequence (Supplementary Fig. 11b). For peptide analysis of pull down experiments, proteins were digested by endoproteinase Trypsin (Promega) in gel following the manufacturer's instructions. The following process was the same as described above. Raw files were searched by Proteome Discoverer (Version 1.4.1.14, Thermo Scientific) against the UniProt Human database. Search parameters were set as follows: enzyme with Trypsin; up to two missed cleavages; carbamidomethyl cysteine as fixed modification; methionine oxidation as variable modifications. The false discovery rates were set at 0.01.
CRISPR-CAS9-mediated somatic PTENb knockout. Sequences (ACCTCCCGCTCCTGGAGCGG or AGTCGCCTGTCACCATTTCC) of the human PTEN gene in exon1 were targeted separately. Oligos were purchased from Tsingke and ligated into the U6-sgRNA plasmid. Hela cells were seeded onto 24-well plates containing 500 ml DMEM medium at a density of 1,00,000 cells per ml of cell suspension. 200 ng of U6-sgRNA plasmid and 200 ng of Cas9 plasmid in total were transfected into cells using PEI according to the manufacturer's instructions. Medium was replaced at around 12 h post transfection. Cells were maintained for another 72 h to allow sufficient time for genomic engineering mediated by the CRISPR-Cas9 system. Transfected cells were then treated with 500 mg ml À 1 of G418 for 3 days. After G418 selection, cell clones were selected and amplified for mutation sequencing.
Co-immunoprecipitation. Immunoprecipitation was performed as previously described 12 . In brief, cells were extracted and lysed in lysis buffer (50 mM Tris-pH 7.5, 150 mM NaCl, 0.5% NP40) freshly supplemented with 1 mM PMSF and protease inhibitor. Cell lysate (500 mg) was incubated with antibody against GFPtag or FLAG-tag for 8 h, followed by incubation with protein A/G agarose for 1 h. The protein-bead complex mixture was washed in washing buffer containing 0.1% NP40 and subjected to western blot to evaluate protein interaction.
Immunofluorescence and confocal microscopy. HeLa cells were seeded on cover glass in DMEM medium followed by transfection with C-terminal GFP-tagged PTEN, PTENa or PTENb. After transfection for 24 h, cells were fixed with 4% PFA, permeabilized with 0.2% Triton X-100 and blocked for at least 30 min using 4% BSA dissolved in PBS þ 0.1% Tween-20 (PBS-T). Primary antibody (diluted in blocking solution) was then applied for 1 h, followed by five washes in PBS-T, and fluorophore-conjugated secondary antibody was applied for 45 min. Coverslips were mounted and cells were evaluated with fluorescence microscopy. A Nikon A1 microscope was used for confocal microscopy.
Cell fractionation. Nucleoli were isolated in Hela cells as previously described with some modifications 54 . Isolated nucleoli were lysed with RIPA buffer. Equal amounts of whole cell lysate, cytosolic fraction, nuclear plasma fraction and nucleolar fraction were evaluated with western blot analysis. Briefly, Hela cells (one 15-cm dish) were scraped into ice-cold PBS and washed three times with ice-cold PBS. Cell pellets were then re-suspended in hypotonic buffer (10 mM Hepes, pH 7.9, 10 mM KCl, 1.5 mM MgCl 2 , 0.5 mM DTT), allowed to swell and subjected to Dounce homogenization. The crude nuclear fraction was recovered by centrifugation at 218g for 5 min, and the cytoplasmic supernatant was retained. Nuclei were re-suspended in 0.35 M sucrose plus 0.5 mM magnesium acetate and then subjected to brief sonication to release the nucleoli. Pellets were purified by two cycles of centrifugation through 0.88 M sucrose, and the supernatant from the first sucrose gradient was recovered as a crude nucleoplasmic fraction. The pellets consisted of nucleoli.
S-tag pull down assay. HEK 293T cells were transfected with pSA, pSA-PTEN, pSA-PTENa, pSA-PTENb or pSA-NCL plasmids and harvested 36 h after transfection. Cells were lysed with lysis buffer (50 mM Tris pH 7.5, 150 mM NaCl, 2 mM EDTA, 1% NP40 and 1 mM NaF). Equal amounts of protein were incubated with S protein agarose (Novagen) for 4 h. The protein-bead complexes were washed four times with washing buffer containing 0.1% NP40. After boiling at 100°C in loading buffer, proteins were loaded onto NuPAGE 4-12% gels (Invitrogen) and visualized with silver staining (Pierce Silver Stain Kit) or were subjected to western blot. The potential interacting proteins in specific bands were evaluated with mass spectrum analysis.
In vitro kinase and phosphatase assay. His-nucleolin was expressed in Sf9 cells and purified using Ni-NTA agarose (Qiagen). Control FLAG peptides, FLAG-CDK1 and FLAG-PTENb were purified with ANTI-FLAG M2 Affinity Gel (F2426) and 3 Â FLAG Peptide (F4799) from HEK 293T cells after transfection with pCMV vector, pCMV-CDK1 or pCMV-PTENb for 30 h. For kinase assay, purified His-nucleolin on Ni-NTA agarose was incubated with CDK1 in kinase buffer (50 mM Tris-HCl, 10 mM MgCl 2 , 2 mM DTT, 1 mM EGTA and 0.2 mM ATP) for 30 min at 30°C. For the following phosphatase assay, phosphorylated His-nucleolin was incubated with control FLAG peptide or purified FLAG-PTENb in dephosphorylation buffer (20 mM HEPES, 1 mM MgCl 2 , 1 mM EDTA, 1 mM DTT and 0.1 mg ml À 1 BSA) for 1 h at 37°C. The phosphorylation levels of nucleolin sites with potential for alteration were evaluated with western blot.
Functional pathway enrichment analysis and visualization. Gene Ontology (GO) and KEGG enrichment analysis were performed using ClueGo (ref. 55), which is a Cytoscape plug-in. During analysis, over-represented terms in KEGG and the GO Biological Process and Cellular Component were enriched, and the GO level was set at three to five. The GO term fusion option was selected to reduce redundancy. In order to account for multiple testing, P values were corrected with Benjamini-Hochberg correction. Only terms of Po0.05 were displayed. The selection criteria for the terms that have associated genes from the uploaded pull down list was set as a minimum of two genes per term, and a minimum of 4% of total genes associated with the term for the analysis of the PTENb pull down network. The Kappa score was set at 0.4.
Luciferase reporter assay and real-time PCR. For luciferase reporter assays, after transfection of various combinations of control, reporter or protein expression vectors for 24 h, cells were lysed with a passive lysis buffer (Promega). Luciferase activity was measured with the Dual Luciferase Assay System (Promega) following the manufacturer's protocol. For quantification of 45S pre-rRNA, total RNA was extracted from cells followed by reverse transcription and evaluated with qPCR. The qPCR primers used are listed in Supplementary Table 4.
Silver staining assay. Silver staining assays were performed as previously described 56 . Briefly, cells were seeded onto glass coverslips overnight and were then fixed in 2% glutaraldehyde, followed by postfixation in a 3:1 ethanol-acetic acid solution. Cells were stained with a 0.33% formic acid-33.3% silver nitrate solution in 0.66% gelatin and mounted on slides with Vectashield (Vector Labs). Nucleolar area was measured in cells with cellSens software provided by Olympus.
BrdU incorporation and cell cycle analysis. For Brdu incorporation, cells were stained using the APC Brdu Flow Kit (BD Bioscience) according to the manufacturer's instructions and subjected to analysis using FACSVerse (BD Biosciences). Experiments were each performed three times, and at least 10 4 cells were analysed per sample.
Statistical analysis. Prism GraphPad software v5.0 was used for analysis. The statistical significance of differences between different groups was calculated with the two-tailed paired t-test. P values of 0.05 or less were considered significant.