Selective Proteasomal Degradation of the B′β Subunit of Protein Phosphatase 2A by the E3 Ubiquitin Ligase Adaptor Kelch-like 15*

Background: Proteasomal degradation of PP2A regulatory subunits has been described, but responsible E3 ubiquitin ligases have remained elusive. Results: KLHL15 is an E3 ubiquitin ligase adaptor targeting the B′/B56β regulatory subunit for proteasomal degradation, promoting formation of alternative PP2A holoenzymes. Conclusion: KLHL15 contributes to brain-specific expression of B′β and modifies PP2A holoenzyme composition. Significance: E3 ligase-mediated B subunit degradation is a novel mechanism to remodel the PP2A heterotrimer. Protein phosphatase 2A (PP2A), a ubiquitous and pleiotropic regulator of intracellular signaling, is composed of a core dimer (AC) bound to a variable (B) regulatory subunit. PP2A is an enzyme family of dozens of heterotrimers with different subcellular locations and cellular substrates dictated by the B subunit. B′β is a brain-specific PP2A regulatory subunit that mediates dephosphorylation of Ca2+/calmodulin-dependent protein kinase II and tyrosine hydroxylase. Unbiased proteomic screens for B′β interactors identified Cullin3 (Cul3), a scaffolding component of E3 ubiquitin ligase complexes, and the previously uncharacterized Kelch-like 15 (KLHL15). KLHL15 is one of ∼40 Kelch-like proteins, many of which have been identified as adaptors for the recruitment of substrates to Cul3-based E3 ubiquitin ligases. Here, we report that KLHL15-Cul3 specifically targets B′β to promote turnover of the PP2A subunit by ubiquitylation and proteasomal degradation. Comparison of KLHL15 and B′β tissue expression profiles suggests that the E3 ligase adaptor contributes to selective expression of the PP2A/B′β holoenzyme in the brain. We mapped KLHL15 residues critical for homodimerization as well as interaction with Cul3 and B′β. Explaining PP2A subunit selectivity, the divergent N terminus of B′β was found necessary and sufficient for KLHL15-mediated degradation, with Tyr-52 having an obligatory role. Although KLHL15 can interact with the PP2A/B′β heterotrimer, it only degrades B′β, thus promoting exchange with other regulatory subunits. E3 ligase adaptor-mediated control of PP2A holoenzyme composition thereby adds another layer of regulation to cellular dephosphorylation events.

PP2A 3 is one of four major serine/threonine phosphatases (in addition to PP1, PP2B, and PP2C) (1) and comprises up to 1% of total protein in mammalian cells (2). Predominantly found as a heterotrimer containing a core dimer of catalytic (C) and scaffolding (A) subunits, PP2A is a widely distributed and critical mediator of many signaling pathways. A third, variable B subunit defines the substrates and localization of the holoenzyme in the cell (3)(4)(5). PP2A B regulatory subunits are subdivided into three gene families, B (PR55/PPP2R2), BЈ (B56/ PR61/PPP2R5), and BЉ (PR72/PPP2R3), with little sequence or structural similarity between them. In mammals each gene family is comprised of four to five highly similar genes. Additional complexity arises through alternative splicing. The five mammalian BЈ gene products (␣/␤/␥/␦/⑀) feature a conserved, tandem ␣-helical central domain that binds to the AC dimer (6,7). The variable N and C termini contain subcellular localization sequences and phosphorylation sites for isoform-specific regulation (8). Upon down-regulation of the A or C subunit, PP2A B and BЈ subunits are rapidly degraded by the proteasome (9 -11). However, whether proteasomal degradation constitutes a physiological PP2A regulatory mechanism involving dedicated E3 ligases has remained unclear.
We previously identified PP2A/BЈ␤ as a brain-specific PP2A holoenzyme that catalyzes the dephosphorylation and inactivation of tyrosine hydroxylase, the rate-limiting enzyme in dopamine synthesis (12), and uncovered BЈ␤ residues that mediate substrate specificity (7). Because PP2A/BЈ␤ could be a target for intervention into Parkinson disease, a hallmark of which is loss of dopamine in the striatum, we set out to discover novel interactors of this regulatory subunit by proteomic approaches.
These screens identified the E3 ubiquitin ligase scaffold cul-lin3 (Cul3) and an uncharacterized protein, KLHL15. KLHL15 belongs to a large family of proteins with a characteristic domain architecture consisting of an N-terminal BTB (broadcomplex, tram track, and bric-a-brac) domain, a C-terminal

EXPERIMENTAL PROCEDURES
Cell Culture and Transfection-HEK293 cells were cultured (37°C and 5% CO 2 ) in Dulbecco's modified Eagle's medium (Invitrogen) with 10% (v/v) fetal bovine serum. Cells were grown to 60% confluency on collagen-coated plates and transfected using Lipofectamine 2000 (Invitrogen) following the manufacturer's protocol for transient transfection of adhered cells. 24 -48 h post-transfection, cells were washed in phosphate-buffered saline and harvested. For experiments including MG132 (Sigma), cells were incubated with 50 M concentrations of the drug (from a 50 mM stock in dimethyl sulfoxide, DMSO) for the final 12 h before lysis.
To silence endogenous KLHL15, four shRNAs were selected (25), subcloned downstream of the H1 promoter (26), and tested for efficient knockdown by coexpression with GFP-KLHL15. To replace endogenous with mutant GFP-KLHL15, the most effective H1-shRNA cassette (target, GGACAGT-GCTCAATAACAA) was subcloned into the Pci1 site of pEGFP-C1 KLHL15 after rendering the KLHL15 cDNA RNAiresistant by introducing four silent nucleotide changes into the target site (23).
Site-directed mutagenesis of KLHL15 and BЈ␤ was carried out using the QuikChange protocol (Invitrogen). KLHL15 mutations were introduced into the gene replacement plasmid expressing shRNA and RNAi-resistant GFP-KLHL15. All constructs were verified by sequencing.
Affinity Purification Coupled to Mass Spectrometry-pcDNA5-FLAG-BЈ␤ (PPP2R5B, BC045619) was stably transfected in T-REx Flp-In 293 cells (Invitrogen) and submitted to affinity purification with FLAG antibody coupled to magnetic beads as described previously (27). pcDNA3-FLAG-BЈ␤ was stably transfected into HEK293 cells and submitted to affinity purification with FLAG antibodies coupled to agarose beads (M2, Sigma) as described previously (28). Two biological replicates of the samples from the T-REx Flp-In 293 cells were analyzed on a LTQ mass spectrometer, and one HEK293 sample was analyzed on an LTQ-Orbitrap mass spectrometer. In parallel several negative control runs (cells expressing the tag alone and/or the tag fused to GFP) were analyzed on the respective mass spectrometers. All acquisition conditions have been described previously (29). A database search was performed using the Mascot search engine against the human and adenovirus complements of the RefSeq library (v45), allowing for deamidation (NQ) and oxidation (M) with trypsin fixed as an enzyme, and one missed cleavage allowed. For LTQ data, the MS tolerance was fixed at 3 Da, and the MS/MS tolerance was fixed at 0.6 Da. For LTQ-Orbitrap data, the MS tolerance was fixed at 12 ppm, whereas the MS/MS tolerance was fixed at 0.6 Da. Hits identified in any of the negative control runs were removed from the list of putative interactors, and only those hits detected in all three of the biological replicates with at least 2 unique peptides were kept. Furthermore, proteins identified in Ն4% of 275 AP-MS analyses performed under the same conditions were deemed potential frequent fliers and removed from further analysis. With the exception of the core PP2A enzymes, only two proteins were identified, CUL3 and KLHL15.
Immunoprecipitation and Immunoblot Analyses-For immunoprecipitation, cells were lysed in buffer containing 20 mM Tris, pH 7.5, 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1 mM phenylmethylsulfonyl fluoride, 1 g/ml leupeptin, 1 mM benzamidine, and 1% (v/v) Triton X-100. Lysates were allowed to rotate at 4°C for 30 min to complete solubilization. Samples were then centrifuged at 13,000 ϫ g for 15 min to pellet debris. A portion of the supernatant was added directly to SDS-PAGE sample buffer to provide an input sample. HA-tagged and FLAG-tagged proteins were immunoprecipitated as indicated for each experiment from HEK293 cells overexpressing the indicated proteins via HA or FLAG antibody EZ-view Red Agarose gel (Sigma) for 4 h rotating at 4°C. EE-tagged proteins were immunoprecipitated with protein A/G-Sepharose (Santa Cruz Biotechnology) preincubated with 4 g of EE antibody. V5-tagged proteins were immunoprecipitated with goat anti-V5-agarose (Abcam). Beads were washed 4 times by centrifugation in lysis buffer and eluted by boiling in SDS-PAGE sample buffer. Samples were separated on 10% (w/v) SDS-PAGE gels, electrotransferred to nitrocellulose, and immunoblotted as indicated. Proteins were visualized using species-specific secondary antibodies conjugated to IRDye 680/800 (LI-COR) and a LI-COR Odyssey infrared scanner. Signals were quantified by densitometry using the gel analysis plugin of ImageJ (National Institutes of Health).
BЈ␤ Turnover Assays-Cycloheximide (Sigma) was dissolved in DMSO to a concentration of 100 mM. Cycloheximide was added directly to cultured HEK293 cells to a final concentration of 100 g/ml 1-10 h before lysis. Cells were washed in phosphate-buffered saline and lysed simultaneously in SDS-PAGE sample buffer containing 5% (v/v) ␤-mercaptoethanol. Lysates were sonicated with a probe tip to shear DNA and boiled for 5-10 min. Samples were then subjected to immunoblotting and analysis as described above.
All but the ␤-actin primers span introns, thus disfavoring amplification of any contaminating genomic DNA. Relative BЈ␤ and KLHL15 mRNA levels were determined using the ⌬⌬C t method using Pgk1, hypoxanthine phosphoribosyltransferase 1, and ␤-actin as internal controls with similar results.

RESULTS
KLHL15 Specifically Associates with BЈ␤-Mass spectrometry was carried out to identify proteins that associate with BЈ␤. To this end, FLAG epitope-tagged BЈ␤ was stably expressed in HEK293 cells followed by FLAG immunoprecipitation and elution of precipitated proteins with ammonium carbonate. Tryptic peptides were then subjected to liquid chromatography and tandem mass spectroscopy. BЈ␤-interacting proteins reliably identified in three independent experiments were the PP2A A and C subunits and Cul3 as well as KLHL15 (Table 1). KLHL15 is a member of the Kelch-like or BTB-Kelch protein family that is encoded by 39 genes in humans. As first shown for Keap1/ KLHL19 (16,17,30), these proteins function as substrate adaptors in Cul3-based E3 ubiquitin ligases (13,31,32).
To assess the specificity of the interaction between BЈ␤ and KLHL15, the human KLHL15 cDNA was isolated by RT-PCR, tagged at the N terminus with GFP and V5 epitopes, and transiently coexpressed with the five HA-tagged BЈ family subunits (␣-⑀). Remarkable considering the high sequence similarity between BЈ isoforms (60 -80%), KLHL15 selectively coimmunoprecipitated BЈ␤ (Fig. 1A). Further analysis confirmed that members of the distantly related B and BЉ subunits of PP2A do not interact with KLHL15 (data not shown). To test whether BЈ␤ discriminates among KLHL-family members, FLAG-BЈ␤ was coexpressed with GFP-tagged KLHL1, KLHL15, or KLHL24. Only KLHL15 coimmunoprecipitated robustly with BЈ␤ ( Fig. 1B), suggesting that KLHL15 functions as a highly selective E3 ubiquitin ligase adaptor for BЈ␤. Interrogating the fate of ubiquitylated BЈ␤, we immunoisolated the FLAG-tagged protein from HEK293 cells also expressing HA-tagged ubiquitin after incubation with chloroquine or MG132 (50 M,12 h) to inhibit lysosomal or proteasomal degradation, respectively. Proteasome, but not lysosome inhibition, resulted in the accumulation of heterogeneous, high molecular weight species positive for both FLAG and HA epitopes (Fig. 1C).
KLHL15 Promotes BЈ␤ Ubiquitylation and Degradation-Because other KLHL proteins act as substrate recognition modules for an E3 ligase core complex composed of Cul3 and Roc1 (13), we sought to determine whether KLHL15 regulates BЈ␤ levels via ubiquitylation and proteasomal degradation. First, we generated KLHL15-directed polyclonal antibodies and effective shRNAs to confirm expression of endogenous KLHL15 in HEK293 cells ( Fig. 2A). To assess whether BЈ␤ is ubiquitylated in a KLHL15-dependent manner, FLAG-BЈ␤ was immunoprecipitated from HEK293 cells also transfected with GFP-KLHL15, KLHL15-directed shRNAs, or control, scrambled shRNAs. Proteasome inhibition by MG132 before cell lysis Mascot scores, spectral counts, number of unique peptides, and sequence coverage for the indicated proteins are listed (PPP2R5B, BЈ␤). In each cell, "͉" delimitates data from different biological replicates.

Gene name
Mascot score
To test whether KLHL15-mediated ubiquitylation promotes BЈ␤ protein turnover, we assessed BЈ␤ levels after inhibition of protein synthesis by cycloheximide (CHX) for up to 8 h (Fig. 2, C and D). As a loading control, BЈ␤ signals were normalized to extracellular signal-regulated kinase (ERK1/2) signals in the same lane, which were stable under these conditions (as compared with Ponceau S-stained total protein). Compared with cotransfection with a non-targeting shRNA, overexpression of KLHL15 lowered steady-state levels of BЈ␤ (0 h CHX) as well as decreased the half-life of BЈ␤ from ϳ8 to ϳ2.5 h. Conversely, silencing of endogenous KLHL15 increased BЈ␤ levels at steady state and stabilized the protein such that ϳ90% of the protein remained after 8 h in CHX (Fig. 2, C and D). Comparison of the area under the curve (AUC) of the BЈ␤ degradation time courses as well as the 5-and 8-h time points revealed significant differences between KLHL15 overexpression, silencing, and control (Fig. 2D). Together these data indicate that KLHL15 recruits a Cul3-based E3 ligase to promote polyubiquitylation and proteasomal degradation of specifically the BЈ␤ subunit of PP2A. Near complete stabilization of BЈ␤ in the absence of KLHL15 further demonstrates that BЈ␤ degradation is principally mediated by ubiquitylation via KLHL15, at least in HEK293 cells.
KLHL15 May Contribute to Brain-specific Expression of BЈ␤-We previously showed that BЈ␤ protein is abundantly expressed in the brain but undetectable in other tissues even after microcystin affinity purification of PP2A holoenzymes (12). We, therefore, wondered whether posttranslational regulation by KLHL15 may contribute to selective expression of BЈ␤ in the nervous system. Quantitative RT-PCR analysis showed that BЈ␤ mRNA is fairly uniformly expressed in a panel of nine rat tissues (Fig. 3A). KLHL15 mRNA expression is also ubiquitous, with the highest levels in lung, muscle, and spleen (Fig. 3A). Transcript profiles were similar regardless of whether ␤-actin, phosphoglycerate kinase 1 (Pgk1) or hypoxanthine phosphoribosyltransferase 1 mRNA was used for normalization ( Fig. 3A and not shown). Plotting the ratio of BЈ␤ to KLHL15 mRNA levels revealed that forebrain, brainstem, and cerebellum contain more BЈ␤ than KLHL15 mRNA, whereas the opposite holds for all non-neuronal tissues examined (Fig. 3B). Because BЈ␤ transcripts are ubiquitous, brain-specific expression of the PP2A/BЈ␤ protein (12) may, therefore, at least in part result from relatively low levels of KLHL15 in the brain.
The BTB Domain Mediates KLHL15 Dimerization and Cul3dependent BЈ␤ Degradation-The mass spectrometry-based identification of both KLHL15 and Cul3 in BЈ␤ immunoprecipitates is in line with previous evidence that BTB-Kelch proteins tether Cul3-containing E3 ubiquitin ligases to their targets (13,31,32). To confirm an interaction between Cul3 and KLHL15, immunoprecipitation was carried out from transfected HEK293 cells. HA-tagged Cul3, but not Cul2, immunoprecipitated GFP-KLHL15 (Fig. 4A). Even larger amounts of GFP-KLHL15 were recovered in immunoprecipitates of Cul3 ⌬Roc, a dominant-negative C-terminal deletion mutant unable to bind the E3 core component Roc1. GFP-KLHL1 also preferentially associated with Cul3 over Cul2, underscoring Cul3 preference as a feature shared among KLHL family proteins.
To explore the role of the KLHL15 BTB domain in regulation of PP2A/BЈ␤, conserved amino acids were targeted by site-directed mutagenesis. Guided by a threading model of the BTB domain dimer (Fig. 4B), we mutated two buried residues near the predicted dimerization interface (D32A, H45L) as well as three adjacent exposed residues (Ile-72, Leu-73, Lys-74; ILK72AAA). To avoid confounds due to dimerization of endogenous with transfected, mutant KLHL15, we used a single plas-

KLHL15 Down-regulates PP2A/B␤
mid, RNAi-based gene replacement strategy (23). To this end, H1 promoter-driven shRNA targeting KLHL15 ( Fig. 2A) was expressed from the same plasmid as the mutant GFP-KLHL15 cDNA, which was rendered RNAi-resistant by silent mutagenesis of the RNAi target site. HEK293 cells coexpressing FLAG-BЈ␤, HA-Cul3, and GFP-KLHL15 were subjected to FLAG immunoprecipitation to elucidate sequence determinants for recruitment of Cul3 to BЈ␤. Both KLHL15 and Cul3 were detected in BЈ␤ immune complexes when wild-type, D32A, or H45L mutant KLHL15 was expressed (Fig. 4C). In contrast, KLHL15 ILK72AAA interacted with BЈ␤ but did not recruit Cul3 into the complex. As a specificity control, neither KLHL15 nor Cul3 associated with BЈ␤ when the KLHL15 Kelch domain was mutated (R318E, see below).
The Top of the KLHL15 ␤-Propeller Mediates BЈ␤ Association-According to high confidence structure modeling templated on the Kelch repeats of Keap1/KLHL19 (38,39), the C terminus of KLHL15 folds into a cone-shaped ␤-propeller with six propeller blades composed of four antiparallel ␤-strands each (Fig. 5A). Keap1/KLHL19 was previously shown to interact with a peptide derived from its substrate Nrf2 via multiple, blade-spanning residues that form the top cavity of the Kelch ␤-propeller (39). The corresponding residues are perfectly conserved among KLHL15 orthologs in vertebrates but are distinct in other KLHL family members (Fig. 5B) and thus may function as substrate specificity determinants. A subset of these putative substrate binding residues mapping to Kelch blades 1 and 2 (asterisks in Fig. 5B) was assessed for effects on BЈ␤ interaction and down-regulation. As with BTB domain mutants, Kelch domain mutant GFP-KLHL15 replaced endogenous KLHL15 via shRNA expressed from the same plasmid. R318E, E335R/ L337A, and E371R substitution strongly impaired association of GFP-KLHL15 with FLAG-BЈ␤ by coimmunoprecipitation (Fig. 5C) and KLHL15-mediated degradation of BЈ␤ (t1 ⁄ 2 Ͼ 8 h, Fig. 5, D and E). On the other hand, mutation of a residue of similar evolutionary conservation but predicted localization at the bottom of the KLHL15 ␤-propeller (E518R) had no interaction phenotype (Fig. 5C). Thus, conserved residues unique to KLHL15 form a substrate binding pocket predicted to reside on top of the Kelch ␤-propeller.
The BЈ␤ N Terminus Interacts Directly with KLHL15-BЈ subunits share a highly conserved central portion composed of tandem ␣-helical repeats that interacts with the PP2A A and C subunit (6). Because KLHL15 associates with BЈ␤, but not other BЈ isoforms (Fig. 1A), we surmised that instead, the divergent N-or C-terminal extensions of BЈ␤ contain the KLHL15 binding domain. In support, initial coimmunoprecipitation experiments with truncated BЈ␤ cDNAs mapped critical residues to the N terminus (amino acids 31-65, Fig. 6A). Interestingly, truncation of the first 30 residues consistently enhanced the interaction of BЈ␤ with KLHL15, suggesting perhaps that the extreme N terminus folds back to occlude an adjacent KLHL15 binding domain.

KLHL15 Down-regulates PP2A/B␤
Because the first 65 residues are dispensable for incorporation of BЈ␤ into the PP2A heterotrimer (7), we next asked whether KLHL15 discriminates between free and holoenzymeassociated BЈ␤. To this end, charge reversal substitutions were incorporated into BЈ␤ at tandem ␣-helical repeat 1 (KR103DE) and 4 (RK232ED) to block A and C subunit association (Fig. 6F and data not shown). Highly conserved across species and BЈ subunit isoforms, but not contacting A or C subunits directly (40,41), these residues instead appear to be necessary for proper folding of the two tandem ␣-helical repeats. Both monomeric BЈ mutants efficiently associated with KLHL15; furthermore, the endogenous PP2A C subunit was detected in KLHL15 immunoprecipitations when wild-type BЈ␤ was coexpressed (Fig. 6F). Therefore, KLHL15 targets BЈ␤ regardless of whether or not the regulatory subunit is incorporated into the PP2A heterotrimer.
The BЈ␤ N terminus Is Sufficient for KLHL15-mediated Degradation-We next inquired whether the BЈ␤ N terminus contains a degron, i.e. sequence determinants sufficient for deg- radation. To this end we assessed turnover of the BЈ␤ N terminus (amino acids 1-82) fused to GFP after overexpression or knockdown of KLHL15 (Fig. 7A). About 40% of GFP-BЈ␤ 1-82 degraded when protein synthesis was inhibited by CHX for 8 h. In contrast, GFP fusions of the BЈ⑀ N terminus (amino acids 1-68) or the BЈ␤ C terminus (amino acids 476 -497) were stable over this time course (Fig. 7, A-D). KLHL15 silencing prevented degradation of GFP-B␤ 1-82 , whereas KLHL15 overexpression accelerated degradation. Neither silencing nor forced expression of KLHL15 had an impact on the stability of GFP-BЈ␤ 476 -497 or GFP-BЈ⑀ 1-68 . Therefore, the N terminus contains both E3 ligase docking and ubiquitylation sites sufficient for degradation of BЈ␤.
KLHL15 Does Not Degrade the PP2A Holoenzyme but Influences Its Composition-Because KLHL15 can associate with the BЈ␤-containing PP2A holoenzyme (Fig. 6F), we examined whether KLHL15 mediates wholesale degradation of the PP2A/ BЈ␤ heterotrimer. In cells transfected with BЈ␤, neither endogenous nor epitope-tagged A and C subunit levels were modulated by KLHL15 overexpression or knockdown (data not shown). However, this negative result could be explained by BЈ␤ competing with endogenous regulatory subunits for association with A and C subunits such that only a small pool of A and C subunits becomes subject to proteasomal degradation via KLHL15. To provide BЈ␤ with privileged access to the A subunit, we employed an A subunit point mutant (DW139RR) that interacts with BЈ but not B or BЉ subunits (22). EE-tagged A␣ DW139RR, FLAG-BЈ␤, and either GFP-KLHL15, control shRNA, or KLHL15-directed shRNA were coexpressed in HEK293 cells. Immunoblotting for the EE epitope revealed little to no effect of KLHL15 levels on A subunit stability in the presence of CHX (Fig. 8, A and B), whereas an inverse relation-  ship between KLHL15 and BЈ␤ levels was confirmed by blotting for FLAG (Fig. 8, A and C). Thus, even though KLHL15 can bind to the PP2A/BЈ␤ holoenzyme, the E3 adaptor only ubiquitylates and degrades the regulatory subunit.
To provide direct evidence that KLHL15 can impact PP2A holoenzyme composition, we assessed competition of BЈ␤ and BЈ␦ for incorporation into the PP2A heterotrimer as a function of KLHL15 levels. HA-tagged BЈ isoforms were coexpressed with wild-type EE-tagged A subunit, and cell lysates were sub-jected to EE-tag immunoprecipitation and immunoblotting (Fig. 8D). Association of both endogenous and HA-tagged BЈ␤ with the A subunit increased when KLHL15 was silenced (1.4 -2.2-fold, n ϭ 2), whereas PP2A holoenzyme incorporation of BЈ␦ increased when KLHL15 was overexpressed (1.6 -2.0-fold, n ϭ 2). Therefore, by virtue of degrading BЈ␤, KLHL15 creates a window of opportunity for other regulatory subunits with distinct subcellular localization and substrate specificity to incorporate into the PP2A heterotrimer. FIGURE 6. Residues within the divergent N terminus of B␤ mediate direct association with the KLHL15 ␤-propeller. A, shown is a diagram of BЈ␤ and Nand C-terminal truncation mutants, with the region sufficient for PP2A holoenzyme association shown in black (7). The KLHL15 binding domain in BЈ␤ (stippled box, amino acids 31-65) is aligned with the corresponding region of BЈ⑀. Tyr-52 (arrow) is conserved in BЈ␤ orthologs from mammals (mam), Xenopus laevis (frog), and Danio rerio (fish) but is replaced with a Ser in BЈ⑀. B, the indicated FLAG-BЈ␤ truncation and point mutants were coexpressed with GFP-KLHL15 in HEK293 cells and tested for association by FLAG immunoprecipitation (IP). Deletion of the N terminus (⌬1-65) and Tyr-52 substitution abrogates the interaction. C and D, the indicated BЈ␤ mutants were expressed in HEK293 cells and assessed for turnover by endogenous KLHL15 in the presence of CHX (0 -10 h). C shows representative blots, and D shows quantification (means Ϯ S.E., n ϭ 3 experiments, see the Fig. 2 legend). E, GST tagged BЈ␤ and BЈ⑀ N termini as well as hexahistidine (His 6 )-tagged KLHL 255-604 (Kelch repeats) were purified from E. coli and used for GST pulldown (pulld'n) analysis. Shown are GST and KLHL15 immunoblots. GST-BЈ␤ 1-82 , but not BЈ⑀ 1-68 , interacts directly with wild type but to a lesser extent with E371R mutant KLHL15 Kelch . F, GFP/V5-KLHL15 was coexpressed with the indicated point mutants of FLAG-BЈ␤ and subjected to immunoprecipitation via the V5 tag (the left two and right two lanes are from the same blots but not adjacent). KLHL15 interacts with PP2A holoenzyme-incorporated BЈ␤ as indicated by coimmunoprecipitation of endogenous PP2A/C when wild-type BЈ␤ is expressed as well as with monomeric (KR103DE, RK232ED) BЈ␤. Molecular mass marker positions are shown in kDa. *, p Ͻ 0.05; **, p Ͻ 0.01 by Student's t test.

DISCUSSION
BTB/Kelch domain-containing (KLHL, Kelch-like, BBK) proteins comprise a large family of evolutionarily conserved proteins with a characteristic domain architecture. Although Kelch and BTB domains are common protein-protein interaction modules found in proteins ranging from metabolic enzymes to transcription factors and ion channels, their presence in the same polypeptide chain defines the largest group of substrate adaptors for E3 ubiquitin ligases, with 39 Kelch-like genes in the human genome (14). The present report assigns a function to a previously uncharacterized member of this gene family, identifying the KLHL15-Cul3-Roc1 complex as the first E3 ligase for PP2A regulatory subunits.
The characterization of KLHL15 as an E3 ligase adaptor adds to a growing list of reports establishing KLHL family proteins as specificity determinants for the ubiquitin-proteasome system. Best characterized at both structural and functional levels is proteasomal degradation of Nrf2, which mediates transcriptional responses to oxidative stress via Keap1/KLHL19 (16). Similarly, KLHL9 and KLHL13 degrade Aurora B, the kinase component of the chromosomal passenger complex, to enable normal mitotic progression (21). Other examples of critical KLHL proteins include Gigaxonin/KLHL16, a ubiquitously expressed protein mutated in giant axonal neuropathy that binds and promotes ubiquitylation of MAP1B (42). KLHL12 targets Dishevelled, an essential mediator of Wnt/␤-catenin signaling, antagonizing the pathway in cultured cells (15).
Site-directed mutagenesis guided by structure prediction supports the consensus view of KLHL proteins as modular adaptors, with the N-terminal BTB domain mediating homodimerization of KLHL15 and Cul3 recruitment via distinct interaction surfaces and the top cavity of the C-terminal Kelch ␤-propeller docking to the substrate, BЈ␤. Predictably, interfering with either BЈ␤ or Cul3 interaction abrogated KLHL15-mediated down-regulation of BЈ␤. Unexpectedly, a point mutation in the BTB domain that specifically blocked KLHL15 self-association (D32A) had the same effect (Fig. 4, E and F). Consistent with this finding, missense mutations in KLHL10 linked to male sterility were recently shown to interfere with homodimerization of the protein (43). Perhaps substrate ubiquitylation by KLHL proteins is an obligatory trans reaction, with one subunit of the dimer directing Cul3-Roc1 for ubiquitin transfer to a substrate bound by the other subunit.
This report also defines a novel regulatory mechanism for PP2A, one of the major Ser/Thr phosphatases. With at least 12 genes encoding regulatory subunit in mammals, PP2A comprises a family of signaling enzymes with unique substrate specificity, localization, and regulation by second messengers (3)(4)(5)8). We found that KLHL15 displays exquisite specificity for binding and down-regulation of BЈ␤, one of five members of the BЈ/B56 family of regulatory subunits. PP2A/BЈ␤ is enriched in the nervous system, where it was shown to dephosphorylate and inactivate two important enzymes, 1) calcium/calmodulin-dependent protein kinase II, a key mediator of long term potentiation and spatial memory (44), and 2) tyrosine hydroxylase, the ratelimiting enzyme in dopamine synthesis (12). According to qRT-PCR, BЈ␤ mRNA is relatively uniformly expressed among across tissues (Fig. 3A), whereas BЈ␤ protein is only detectable in the brain (12). Because brain is the only tissue with higher BЈ␤ than KLHL15 mRNA levels, we propose that the brain-specific expression of the BЈ␤ protein is at least in part a result of efficient KLHL15-mediated proteasomal degradation in other tissues.
KLHL15 likely not only determines cell and tissue specificity of BЈ␤ expression but may also regulate abundance of this PP2A regulatory subunit dynamically and in response to signaling events. This could for instance occur through BЈ␤ phosphorylation at Tyr-52, an essential KLHL15 binding determinant (Fig.  6, B-D), or at nearby Ser-44, Ser-46, and Ser-47, Ser residues that have been identified in unbiased phosphoproteomic screens (PhosphoSitePlus).
The ␤-propeller of KLHL15 directly engages the divergent N terminus of BЈ␤, a region of the regulatory subunit necessary and sufficient for proteasomal degradation (Figs. 6, B and E, and  7, A and B). Accordingly, KLHL15 targets both monomeric and holoenzyme-incorporated BЈ␤ (Fig. 6F). However, we found no evidence that KLHL15 degrades the A and C subunit within the PP2A/BЈ␤ heterotrimer. Instead, the E3 ligase adaptor appears to abstract BЈ␤ from the AC dimer, allowing other regulatory subunits to take the place of BЈ␤ (Fig. 8D). Selective ubiquitylation and proteasomal degradation of BЈ␤ by KLHL15 may, therefore, be a mechanism for dynamic regulation of PP2A holoenzyme composition and specific substrate dephosphorylation.
On a final note, our results raise the intriguing possibility that each of the dozen or more PP2A regulatory subunit is paired with a dedicated E3 ubiquitin ligase. Indeed, substrates for the FIGURE 7. The B␤ N terminus is sufficient for KLHL15-mediated turnover. A-D, the BЈ␤ N (residues 1-82) and C terminus (476 -497) and the BЈ⑀ N terminus (1-68) were fused to GFP and cotransfected with plasmids expressing GFP-KLHL15, control shRNA, or shRNA targeting KLHL15. Stability of GFP-BЈ fusion proteins was then assessed in the presence of CHX for up to 8 h. A shows representative blots from a BЈ␤ 1-82 turnover experiment, whereas panels B-D show quantification of BЈ␤ 1-82 (B), BЈ␤ 476 -497 (C), and BЈ⑀ 1-68 (D) turnover (means Ϯ S.E., n ϭ 3 experiments, see Fig. 2 legend). GFP-BЈ␤ 1-82 is less stable than the other GFP fusion proteins and was selectively regulated by KLHL15 overexpression and silencing. Molecular mass marker positions are shown in kDa. *, p Ͻ 0.05; **, p Ͻ 0.01 by Student's t test. AUC, area under the curve. majority of the 39 human KLHL family proteins remain to be identified. KLHL protein-catalyzed exchange of PP2A regulatory subunits could thus shape dephosphorylation profiles in every cell.