Brk/Protein Tyrosine Kinase 6 Phosphorylates p27^KIP1, Regulating the Activity of Cyclin D–Cyclin-Dependent Kinase 4

Abstract

Cyclin D and cyclin-dependent kinase 4 (cdk4) are overexpressed in a variety of tumors, but their levels are not accurate indicators of oncogenic activity because an accessory factor such as p27^Kip1 is required to assemble this unstable dimer. Additionally, tyrosine (Y) phosphorylation of p27 (pY88) is required to activate cdk4, acting as an “on/off switch.” We identified two SH3 recruitment domains within p27 that modulate pY88, thereby modulating cdk4 activity. Via an SH3-PXXP interaction screen, we identified Brk (breast tumor-related kinase) as a high-affinity p27 kinase. Modulation of Brk in breast cancer cells modulates pY88 and increases resistance to the cdk4 inhibitor PD 0332991. An alternatively spliced form of Brk (Alt Brk) which contains its SH3 domain blocks pY88 and acts as an endogenous cdk4 inhibitor, identifying a potentially targetable regulatory region within p27. Brk is overexpressed in 60% of breast carcinomas, suggesting that this facilitates cell cycle progression by modulating cdk4 through p27 Y phosphorylation. p27 has been considered a tumor suppressor, but our data strengthen the idea that it should also be considered an oncoprotein, responsible for cyclin D-cdk4 activity.

INTRODUCTION

Cyclin D1–cyclin-dependent kinase 4 (cdk4) complexes promote the G₀/G₁-phase transition, and as such their activity is tightly regulated by a variety of mechanisms, including the transcription and translation of the mitogen sensor cyclin D1 and positive and negative regulatory phosphorylation of cdk4 (1, 2). The best-characterized substrate of cyclin D-cdk4 is the G₁ gatekeeper, retinoblastoma (Rb), and deregulation of cdk4 potentially accelerates Rb phosphorylation and cell cycle transitioning, promoting cancer development (3). Cyclin D1 and cdk4 are overexpressed in a variety of human cancers, and in mouse models, loss of either cdk4 or cyclin D1 prevents the development of certain oncogene-driven tumors, further evidence of their involvement (4–6). However, the levels of cyclin D or cdk4 in a tumor may not be reliable measures of activity, due to the fact that a third protein, an assembly factor such as p27Kip1 or p21Cip1, is required both for the stabilization and then the subsequent activation of this complex (1, 7).

Independently of its ability to assemble cyclin D-cdk4 complexes, p27 acts as a bona fide “switch” turning cyclin D-cdk4 complexes on or off, which in turn modulates cell cycle entry or exit (8, 9). Tyrosine (Y) phosphorylation of p27 on residues Y74, Y88, and Y89 opens the cyclin D-cdk4-p27 ternary complex, rendering it able to phosphorylate substrates such as Rb (9–14). Cyclin D-cdk4-p27 complexes isolated from cells in G₀ lack Y phosphorylation on p27 and are catalytically inactive, while complexes isolated from proliferating cells are Y phosphorylated and active. Y88 and Y89 are part of the 3-to-10 helix, which has been shown to insert into the cdk ATP binding cleft (15). When not phosphorylated, residues Y88 and Y89 (Y88/Y89) sequester within this binding pocket and block cdk4 activity (p27 switched off). Results of nuclear magnetic resonance (NMR) analysis and other studies suggest that phosphorylation of Y88/Y89 induces a conformational change in p27, ejecting the Y88/Y89 loop, opening the cyclin D-cdk4 complex, and permitting both ATP access and the required phosphorylation on cdk4 residue T172 by cyclin-activating kinase (CAK), the latter causing activation of cdk4 (p27 switched on) (11, 12, 14, 16). Thus, p27's control of cyclin D-cdk4 makes it a key player in the regulation and integration of a cell's response to extracellular signals.

Members of the Src family of kinases (SFKs), including Src, Yes, and Lyn, have been shown to phosphorylate p27 in vitro (9). Moreover, distantly related kinases, such as the Abelson kinase Abl and the Janus kinase Jak2, also appear competent to phosphorylate p27 (11, 12, 17). The Src kinase family consists of 8 members: Src, Yes, Fyn, Fgr, Lyn, Hck, Lck, and Blk (18). Frk, Srm, Src42A, and Brk (breast tumor-related kinase)/protein tyrosine kinase 6 (PTK6) and Brk (breast tumor-related kinase) comprise a distantly related but distinct family (19, 20). All of these kinases share a domain organization comprising the tyrosine kinase domain (also termed SH1) plus one each of the protein-protein interaction modules SH2 and SH3, which bind to phosphotyrosine and proline-rich sequences (PXXP), respectively. The SH2 and SH3 domains recognize specific amino acid sequences within the SFK itself, thus adopting an autoinhibited state. Upon release from this inhibition by upstream signaling molecules, the SH2 and SH3 domains are free to bind downstream SFK target proteins (21).

We identified two SH3 domain recruitment sequences within p27 (22, 23) and confirmed that the p27 PXXP-SH3 interaction not only modulates Y88 phosphorylation by SFKs but also modulates cdk4 activity. We have also identified another kinase, PTK6/Brk (24–26), that functions as a high-affinity kinase, able to phosphorylate p27 more efficiently than other SFKs tested. Brk is an intracellular tyrosine kinase expressed in normal epithelial cells and overexpressed in 60% of breast cancers. It promotes signaling by several receptor-bound tyrosine kinases, including the ERbB receptor family, MET, and insulin-like growth factor 1 receptor (IGF-1R). Overexpression of Brk in vivo increases p27 phosphorylation, increases cdk4 activity, and increases resistance to specific cdk4 inhibition by the chemical inhibitor PD 0332991 in a kinase-dependent fashion. An alternatively (Alt) spliced form of Brk (Alt Brk) which contains the SH3 domain blocks p27 Y phosphorylation and acts as an endogenous inhibitor of cdk4, confirming the importance of the PXXP-SH3 interaction in vivo. Our data suggest that Brk is an important physiological kinase whose overexpression in cancer increases p27 Y phosphorylation, cdk4 activity, and proliferation.

MATERIALS AND METHODS

Antibodies.

The antibodies used in the study were as follows: mouse anti-p27 (Kip1) (BD Biosciences 610242); Cdk4 (DCS-35), p27 (N-20), C-terminal (C-term) Brk (C-18), Brk (D-6), N-terminal Brk (N-20), c-Src (SC-18), cyclin D1 (H 295), and ARHGDIA (A-20) (Santa Cruz Biotechnology); phosphotyrosine (P-Tyr-100) (Cell Signaling Technology); Cdk4 (C-term; catalog no. AP7520b) (Abgent); glutathione S-transferase (GST) (PRB-112C) (Covance); Flag (F3165) and actin (A2066) (Sigma-Aldrich); and phospho-Brk (Tyr342) (EMD Millipore). pY74, Y88, and pY89 phospho-specific antibodies were generated by immunization of rabbits with phospho-specific p27 peptides (Invitrogen). Negative- and positive-affinity chromatography with nonphosphorylated and phosphorylated peptides, respectively, was performed to purify the antibodies. The antibodies are specific only for Y88, Y89, and Y74 phosphorylation, respectively (data not shown; see Fig. 2E).

FIG 2.

Brk phosphorylates p27 in vitro. (A) GST-Brk or GST-Src was incubated with recombinant His-p27 or mutants and isolated using metal agarose chromatography (His), followed by immunoblot analysis with p27 and GST antibodies. A direct comparison between GST-Brk and GST-Src associations can be made. (B) WT p27 was incubated with equal microgram volumes (with approximately equivalent specific activities) of Abl, Src, or Brk in the presence of [γ-³²P]ATP. Phosphorylated products were run on SDS-PAGE gels and quantitated by autoradiography (n = 5). (C) WT p27 was incubated with equal amounts of Src or Brk in the presence of ATP. Phosphorylated products were run on SDS-PAGE gels and assayed by immunoblot analysis with p27, pY88, and pY74 antibodies. T,m., time in minutes. (D) Plot of densitometry from panel C. (E) WT p27 and mutants were incubated with Brk and ATP and separated on SDS-PAGE gels, and immunoblots were probed with p27, pY88, or pY74 antibodies. (F) p27-cyclin D-cdk4 ternary complexes were generated by incubation of recombinant proteins and isolated by immunoprecipitation with cdk4 antibodies. Those and monomeric p27 were phosphorylated with Brk in the presence of [γ-³²P]ATP. Phosphorylated products were run on SDS-PAGE gels and quantitated by autoradiography (n = 5). (G) p27-cyclin D-cdk4 ternary complexes were generated by incubation of recombinant proteins and then isolated by immunoprecipitation with cdk4 antibodies (lanes 1 to 4). These ternary complexes (lanes 1 to 4) and monomeric p27 variants (lanes 7 to 10) were phosphorylated with Brk in the presence of ATP, followed by immunoblot analysis with pY88 and p27 antibodies. (H) Increasing amounts of WT or ΔK1/K3 p27 were incubated with cyclin D-cdk4 and phosphorylated or mock phosphorylated by Brk and then used in Rb kinase assays in the presence of [γ-³²P]ATP. Phosphorylated Rb (Rb*) was isolated using GST-Sepharose and autoradiography. p27 was recovered by metal agarose chromatography, followed by immunoblot analysis with p27 and pY88 antibodies. Lane 9 is Brk without cyclin D-cdk4 or p27. (I) p27 and mutants were incubated in the absence (lanes 1 to 4) or the presence (lanes 6 to 9) of Brk and then incubated with cyclin D-cdk4 and used in Rb kinase assays in the presence of [γ-³²P]ATP. Lanes 5 and 10, cyclin D-cdk4 without p27. Note that the low level of cyclin D-cdk4 used in this experiment does not result in detectable Rb phosphorylation. Phosphorylated GST-Rb (Rb*) was isolated using GST-Sepharose followed by autoradiography. In parallel, cdk4-associated complexes were isolated by immunoprecipitation (Ip) and assayed by cyclin D immunoblot analysis. W, Western blot.

Enzymes.

GST-PTK6/Brk and GST-Src (SignalChem), His-Abl (New England BioLabs), and His-PTK6/Brk and His-Src (Invitrogen) were used according to the specifications of the manufacturers. Enzymes had approximately equivalent specific activities.

Phage enzyme-linked immunosorbent assay (ELISA).

Phage supernatants were generated and binding of SH3 phages to recombinantly produced His-tagged-p27 or GST-PXXP peptides was analyzed as described previously (27).

Construction of mutants and peptides.

Oligonucleotides encoding PXXP peptides K1, K2, and K3 were annealed and directly ligated into pGEX-KG expression vector for production of N-terminally GST-tagged peptides. The GST-, GST-Brk SH3-, and GST-Brk SH2-expressing plasmids used were previously described (28). Escherichia coli BL21 cells transformed with these plasmids were grown in LB-ampicillin until an optical density (OD) at 450 nm of 0.6 was reached, and protein production was induced by addition of 1 mM IPTG (isopropyl-β-d-thiogalactopyranoside). After 2 h, cells were harvested by centrifugation. Cell lysis and protein purification on GST-Sepharose were carried out according to the instructions in a GST-protein purification manual (GE Healthcare). Protein was eluted with an excess of glutathione and dialyzed against phosphate-buffered saline (PBS) for further use. Purified, C-terminal histidine (His)-tagged or N-terminal Flag-tagged p27's were generated from E. coli as described previously (12). Human p27 cDNA was used as a template in PCR mutagenesis with oligonucleotides carrying the following point mutations: PPPP91,92,94,95AAAA (ΔK1); PKKP188,189,190,191AAAA (ΔK3); and PPPP91,92,94, 95AAAA and PKKP188,189,190,191AAAA (ΔK1/K3).

Oligonucleotides used to generate F58-106 were as follows: forward primer, 5′-GGCCTCGAGCTAGCTCTCCTGCGCCG-3′; reverse primer, 5′GGGGTCTAGAGCCACCATGGACTACAAGGACGACGATGACAAGCGCAAGTGGAATTTCGATTTTC-3′.

The PCR fragments were ligated to the T7pGEMEX human His-p27 or T7pGEMEX human Flag-p27 plasmid for expression in E. coli. Mutants Y74F, Y88F, and Y88/89F were previously described (12). Flag-tagged p27 mutants were purified by Flag immunoprecipitation with Flag antibody (M-2; Sigma F-18C9) and eluted with Flag peptide (Sigma F-4799) according to the manufacturer's protocol. His-tagged p27 mutants were purified by fast protein liquid chromatography (FPLC) via His-trap affinity chromatography (His-Trap HP; GE Healthcare 71-5247-01). The affinity column was stripped according to the manufacturer's protocol and then washed with 5 column volumes of 100 mM CoCl₂. The crude material was applied with a loading buffer consisting of 6 M urea, 500 mM NaCl, 50 mM Tris-HCl (pH 7.5), and 20% glycerol. The material was washed with 500 mM NaCl, 50 mM Tris-HCl (pH 7.5), and 10% glycerol. The purified material was eluted with 500 mM imidazole, 20 mM HEPES (pH 7.4), and 1 M KCl. The protein was then dialyzed overnight in a solution of 25 mM HEPES (pH 7.7), 150 mM NaCl, 5 mM MgCl₂, and 0.05% NP-40. All purified proteins were analyzed by Coomassie staining and immunoblot analysis. The p27, ΔK1, ΔK3, ΔK1/K3, Y74F, and Y88/89F cassettes were cloned into the pTRE3G tetracycline-inducible retroviral expression construct using an In-Fusion gene cloning kit (Clontech). Alt Brk was generated by PCR using human Alt Brk in PCDNA3 vector (29) as a template, followed by cloning into the T7pGEMEX human Flag-tagged plasmid and pTRE3G using an In-Fusion cloning kit.

Recombinant cyclin D1-cdk4.

Recombinant His-cyclin D1-cdk4 was harvested from coinfected High5 cells and purified as described previously (12). Recombinant GST-Rb (86-kDa version) was purified and used in in vitro kinase assays as previously described (12).

In vitro phosphorylation of the p27-cyclin D1-Cdk4 ternary complex.

Recombinant His-p27 and mutants were incubated for 1 h at room temperature with cyclin D1-Cdk4–25 mM HEPES (pH 7.4). This ternary complex was immunoprecipitated with anti-Cdk4 antibodies (Santa Cruz; DCS 35) and protein G Dynabeads (Invitrogen; 10004D). The complex was then subjected to SFK phosphorylation and/or used in in vitro Rb kinase assays.

Cell lines.

MCF10A, MCF7, MDA MB 231, MDA MB 468, T47D, PC3, Mv1Lu, and HEK293 cell lines were purchased from ATCC and maintained according to vendor's instructions. Insulin levels were adjusted to 0, 10, or 50 μg/ml, and cells were grown for 2 weeks before being assayed as described. To arrest by contact, cells were grown to confluence and maintained for 6 days, including replenishing the medium every other day. Immunoprecipitation, immunofluorescence, and propidium iodide (PI) staining were performed as described in Materials and Methods. Fluorescence-activated cell sorter (FACS) analysis was performed as described previously (30). Cells were counted using an automated cell counter (Bio-Rad TC-20). Viability was measured by trypan blue staining and quantified using the cell counter.

Immunoprecipitation.

Cells were lysed with either Triton lysis buffer (25 mM HEPES [pH 7.4], 100 mM NaCl, 1 mM EDTA, 10% glycerol, 1% Triton X-100) or Tween lysis buffer (50 mM HEPES [pH 7.4], 150 mM NaCl, 1 mM EDTA, 2.5 mM EGTA, 10% glycerol, 0.1% Tween 20). The lysis buffers were supplemented with 1 mM phenylmethylsulfonyl fluoride (PMSF), 10 mM dithiothreitol (DTT), 1 mM NaV, 10 ng/ml leupeptin, and 1 ng/ml aprotinin. Lysates (1 mg) were precleared by incubation with Dynabeads (A or G; Life Technologies) for 1 h at 4°C. Immunoprecipitations proceeded as described previously (12).

Immunofluorescence.

Cell lines were split on day 0 into subconfluent conditions and fixed on day 2 in microwell plates using 4% paraformaldehyde–1× PBS (pH 7.4) for 15 min at room temperature. They were permeabilized with 0.1% Triton X-100 and blocked with 5% bovine serum albumin (BSA) for 1 h at room temperature. They were incubated with the first round of primary antibodies in PBS for 1 h at room temperature. The cells were washed with PBS and incubated with appropriate secondary antibodies (1:500) and diluted in 3% BSA–PBS for 1 h at room temperature. They were then washed with PBS and incubated with 0.02% Triton X-100–3% BSA for 30 min at room temperature to prepare them for a second round of incubation with antibodies. Cells were then washed with PBS and incubated with Hoechst stain (1 mg/ml)-PBS (1:5,000) for 15 min at room temperature. They were rinsed with water and mounted on a slide with 90% glycerol. Samples were incubated at 4–20°C before they were analyzed by confocal microscopy.

Rb in vitro kinase assay.

Cdk4-associated complexes were immunoprecipitated as described above. Recombinant Rb was added to the immunoprecipitates in the presence of the kinase buffer (50 mM) (HEPES [pH 7.4], 10 mM MgCl₂, 1 mM DTT, 2 mM EGTA, 3 mM β-glycerophosphatase, 100 μM ATP), and the mixture was incubated for 30 min before SDS-PAGE was performed. Immunoblot analysis was performed to probe for cdk4, Rb, and pRb-Ser780.

Inhibitor treatment.

Cells were seeded on six-well plates in duplicate at 5.0 × 10⁴ per well. At 24 h postseeding, one well for each plate was treated with trypsin and counted using a Bio-Rad automated cell counter. At 48 h postseeding, another well was treated with trypsin, cells were counted, and the rest of the wells were treated with PD 0332991 (SelleckChem) at 50 nM, 100 nM, 200 nM, and 400 nM. Dimethyl sulfoxide (DMSO) was used as a negative control. Cells were counted again 24 and 48 h posttreatment. The concentrations of drug needed to inhibit proliferation by 50% (IC₅₀s) were determined by normalizing the number of viable cells treated with different concentrations of PD to the number of viable cells treated with DMSO for each cell line 48 h posttreatment. The number of viable cells treated with DMSO was considered to represent 100%. The log of the viability values was obtained, and the data were fitted to a nonlinear regression curve, which was used to generate the IC₅₀s using Graphpad Prism software.

Brk knockdown.

Lentiviral small interfering RNA (siRNA) particles (NM_005975.2-1064sc1 and NM_004383.x-2117s1c1) were purchased from Sigma-Aldrich. MCF7 cells were plated on day 0; on day 1, the medium was aspirated and the cells were infected with the siRNA lentiviral particles. Hexadimethrine bromide was used according to the manufacturer's instructions to enhance the infection efficiency. Cells were incubated overnight, medium was replenished on day 2, and the cells were incubated for 72 h and fixed with 4% paraformaldehyde and immunofluorescence analysis was performed as described above.

Expression in vivo.

Generation of wild-type Brk (WT-Brk), Brk KM (a catalytically inactive variant), and Brk YF (a constitutively active variant) has been described previously (31). Amphotropic retroviruses were generated by transfection using Lipofectamine 2000 (Life Technologies 11668-019) and HEK293 cells with pAmpho envelope and pBabe or pTRE3G tetracycline-inducible constructs. Following viral infection of MCF7 cells, stable integrants were isolated by puromycin selection. Colonies were pooled to generate stable, puromycin-resistant clones. Stable expression was verified by immunoblot and immunofluorescence analysis. Tetracycline-inducible expression was achieved by the addition of TetExpress (Clontech) to the medium.

Quantitative RT-PCR (q-RT-PCR).

RNA extraction was performed using TRIzol reagent (Life Technologies) as directed by the manufacturer's instructions. A 500-μg volume of RNA was subjected to reverse transcription (RT) using a Verso cDNA kit (Thermo Scientific). A 250-ng volume of RNA was mixed with cDNA primers and ABsolute Blue quantitative PCR (qPCR) SYBR green (Thermo Scientific) to perform qPCR. The following primers were used to perform qPCR: for actin, 5′-AAAATCTGGCACCACACCTTCTAC-3′ (forward) and 5′-TAGCACAGCCTGGATAGCAACG-3′ (reverse); for Brk, 5′-CCAAGTATGTGGGCCTCTGG-3′ (forward) and 5′-AAAGAACCACGGTTCCGACT-3′ (reverse); and for Alt Brk, 5′-GACGGTGGAGTCGGAACCTG-3′ (forward) and 5′-TAGTTCACAAGCTCGGGCAG-3′ (reverse).

Statistics.

The statistical analysis was performed using the Student t test, Welch's t test, and a 2-tailed type 3 test, due to the presence of unequal sample sizes with unequal variances.

RESULTS

Brk phosphorylates p27 in vitro.

p27 contains three putative SH3 recruitment sequences that contain the common PXXP core motif, designated K1, K2, and K3 (Fig. 1A). K1 contains a basic residue after the PXXP, thus qualifying it as a canonical type 2K SH3 target site (32). K2 is present only in the human orthologue of p27 and thus is unlikely to mediate conserved functions in cell cycle control. K3 is at the C terminus of p27, in a region that has been shown to be dispensable for cdk interaction (15). Based on the reported interactions of p27 with non-receptor-bound tyrosine kinases (SFKs), such as Src, Yes, and Lyn, we asked whether other members of the family might also interact with p27 and which recruitment sequences (K1, K2, and/or K3) are used. We tested 11 members of the SFK family as well as Abl, which has been reported to phosphorylate p27 in vitro and in vivo, for binding to either full-length p27 or GST-tagged K1, K2, or K3 peptides, using a phage ELISA procedure (27) (Fig. 1B and C).

FIG 1.

Brk binds to p27 with high affinity in vitro. (A) p27 sequence highlighting the proline tracts of the three putative SH3 domain recruitment sites (PXXP): K1 (90 to 96), K2 (114 to 117), and K3 (188 to 195). (B and C) Phage ELISA of SFK SH3 interactions with p27 (B) or p27's PXXP motifs (C). Data shown represent the means of the results of three independent experiments ± standard deviations after normalization and subtraction of the background binding to GST. (C) Recombinantly produced GST-K1, -K2, or -K3 or GST was immobilized in 96-well plates and analyzed for binding of the phages with the Brk-, Frk-, Yes-, or Abl-SH3 domain.

While the SH3 domains of most SFKs could interact with full-length p27, we found that the SH3 domain of Brk interacted strongly with full-length p27 (K_d [dissociation constant] = 250 nM) and associated better than either Src or Abl, two SFKs known to interact with p27 (Fig. 1B). This K_d value is reflective of the interaction between p27 and the phages, which contain many identical reiterated SH3 domains, which would enhance binding. We expressed the individual SH3 recruitment sites within p27 (K1, K2, and K3) as GST fusion peptides and tested them against the SH3 domain library (Fig. 1C). The GST domain expressed in the absence of any p27 sequence was used as a negative binding control (GST). Most SFK SH3 domains were not able to interact significantly with the individual PXXP-containing peptides (data not shown). Brk, however, interacted strongly with the K3 region and weakly with the K1 region (Fig. 1C). The related kinase, Frk, was the second-best binder to full-length p27 (Fig. 1B), but when Frk was tested against the individual PXXP domains, significant binding to the GST negative control was detected (Fig. 1C), so we could not conclude whether Frk's SH3 domain bound to the PXXP domain peptides. The SH3 domains of Abl and Yes interacted with the K3 domain of p27, although this interaction was reduced compared to that of Brk (Fig. 1C). No SH3 domains interacted with the K2 site (Fig. 1C).

To determine whether full-length Brk could interact with p27, we incubated GST-Brk or GST-Src with recombinant His-p27. p27-associated complexes were isolated by metal agarose chromatography (His) and then assayed for p27-associated Brk or Src by immunoblot analysis using GST antibodies (Fig. 2A). In the context of the full-length enzyme, GST-Brk and GST-Src associated to similar extents with p27, suggesting that additional contacts present outside the SH3 domain might increase the affinity.

To determine whether Brk's interaction with p27 led to phosphorylation, we incubated recombinant p27 with purified Brk in the presence of [γ-³²P]ATP (Fig. 2B). We included Abl and Src, as they have been shown to phosphorylate p27 in vitro. Consistent with Brk's high-affinity interaction with p27, we found that Brk phosphorylated p27 approximately 4-fold better than Abl and 8-fold better than Src (Fig. 2B), suggesting an increased specific activity toward this substrate. We additionally verified this using our p27 pY88 phospho-specific antibody (pY88) and our Y74 phospho-specific antibody (pY74) (Fig. 2C and D). These antibodies recognize Y88 or Y74 phosphorylation in vitro and in vivo and do not recognize phosphorylation on mutant Y88F or mutant Y74F, respectively (Fig. 2E, lanes 6 and 7) (14). We incubated full-length p27 with Src or Brk in the presence of ATP and either MgCl₂ or MnCl₂. In the presence of both cations, Brk phosphorylated p27 on both residues more efficiently (Fig. 2C and D). Brk also phosphorylated the adjacent residue Y89 in vitro (data not shown), but we focused on Y88 phosphorylation in the rest of this study.

The K1 site is essential for Brk's phosphorylation of p27 Y88 and activation of the cyclin D-cdk4 complex.

To determine whether Brk's phosphorylation of p27 was mediated through the K1 or K3 sites, we mutated the prolines in these domains to alanines to generate recombinant His-tagged mutants ΔK1 and ΔK3. Mutant ΔK1/K3 has lost both sites. All of these mutants contain an intact K2 site, which did not appear to associate with Brk in the interaction screen (Fig. 1). We incubated full-length GST-Brk or GST-Src with mutants ΔK1 and ΔK1/K3 and isolated complexes by metal agarose chromatography (His) as described above (Fig. 2A). The association of mutant ΔK1 or mutant ΔK1/K3 with Brk was reduced but still detectable (Fig. 2A, lanes 3, 4, 9, and 10), again suggesting that additional contacts outside the PXXP region mediated the full p27-Brk interaction. GST-tagged Brk did not associate with the metal agarose chromatography beads (data not shown).

We then probed these mutants for phosphorylation using the pY88 and pY74 antibodies (Fig. 2E). We did not observe Y88 phosphorylation in the ΔK1 and ΔK1/K3 mutants (Fig. 2E, lanes 3 and 5), suggesting that an intact K1 site was necessary for efficient Y88 phosphorylation. pY88 was detected in the ΔK3 mutant, demonstrating a requirement for the K1 site, rather than the K3 site, in this assay. When the GST-SH3 domain of Brk was used in the phage ELISA (Fig. 1), the K3 peptide bound preferentially, suggesting that additional contacts increase the K1:Brk affinity in the context of the full-length Brk protein. When probed with the pY74 antibody, all of the mutants, with the exception of Y74F, were still phosphorylated (Fig. 2E). The phosphorylation of residue Y88 and that of residue Y74 appear to occur independently, as the singly mutated variants (Y88F and Y74F) were still phosphorylated on the intact sites, respectively (Fig. 2E, lanes 6 and 7). These data suggested that interaction of the K1 site with Brk is required for Y88 phosphorylation but that phosphorylation of Y74 is K1 and K3 independent (Fig. 2E).

In vivo, p27 is not detected as a monomer but rather appears to be complexed with cyclin-cdk complexes (12). To determine whether Brk could phosphorylate p27 when bound to cyclin D-cdk4, p27-cyclin D-cdk4 ternary complexes were generated by incubation with recombinant components and isolated by immunoprecipitation with cdk4 antibodies (Fig. 2F and G). Cyclin D-cdk4-p27 ternary complexes and monomeric p27 were then incubated with Brk and [γ-³²P]ATP, and total Y phosphorylation was monitored by autoradiography (p27*) (Fig. 2F). An approximately 3-fold reduction in phosphorylation was seen when p27 was associated with cdk4. A similar experiment was performed with the p27 mutants, isolating p27-cyclin D-cdk4 ternary complexes by cdk4 immunoprecipitation (Fig. 2G, lanes 1 to 4) and the use of monomeric p27 (Fig. 2G, lanes 7 to 10), followed by Brk and ATP incubation and immunoblot analysis with pY88 antibodies. When both monomeric and cdk4-associated p27s were assayed, phosphorylation was not detected in mutants ΔK1, ΔK1/K3, and Y88F (Fig. 2G, lanes 1, 3, 4, 7, 9, and 10). Mutant Y74F was phosphorylated as a monomer (Fig. 2G, lane 8) as well as when associated with cdk4 (Fig. 2G, lane 2), and the level seen with pY88 was reduced approximately 3-fold. Thus, Brk can still phosphorylate p27 when presented in more-physiological p27-cyclin D-cdk4 ternary complexes.

We and others have shown that loss of Y88 phosphorylation converts p27 into a cdk4 inhibitor in vitro and in vivo (12, 14), effectively locking the p27-cyclin D-cdk4 ternary complex into a closed conformation and preventing both CAK phosphorylation of cdk4 and ATP access to the catalytic site. Our data predicted that loss of the K1 site, which causes loss of pY88 phosphorylation, should also convert p27 into a cdk4 inhibitor. Therefore, we tested the ability of WT and p27 mutants to inhibit recombinant cyclin D-cdk4 in in vitro kinase assays, using recombinant Rb as a substrate (Fig. 2H and I). We found that while cyclin D-cdk4 phosphorylates Rb, this phosphorylation can be inhibited by increasing concentrations of nonphosphorylated p27 or, alternatively, that the specific activity of the complex is increased by Y88-phosphorylated p27 (Fig. 2H). WT or p27 mutants were not phosphorylated (Fig. 2I, lanes 1 to 4) or were phosphorylated with Brk (Fig. 2I, lanes 6 to 9) and were then mixed with cyclin D-cdk4. Recombinant Rb was added in the presence of [γ-³²P]ATP, and phosphorylation was monitored by autoradiography (Fig. 2I, Rb*). In this experiment, we used a low concentration of cyclin D-cdk4 that did not give detectable Rb phosphorylation (Fig. 2I, lane 5). In the absence of Brk phosphorylation, the addition of the WT and the p27 mutants still did not permit detectable cyclin D-cdk4 activity (Fig. 2I, lanes 1 to 4). The addition of the phosphorylated WT to cyclin D-cdk4, however, restored Rb kinase activity (Fig. 2I, lane 6), suggesting that it converted cyclin D-cdk4 into an active conformation and increased the specific activity of this complex. However, all of the other mutants, which were not phosphorylated on all residues, still inhibited cyclin D-cdk4 activity (Fig. 2I, lanes 7 to 9). As expected, mutants ΔK1 and Y88/89F, which lacked Y88 phosphorylation, inhibited Rb kinase activity (Fig. 2I, lanes 7 and 9). Mutant Y74F, which retained Y88 phosphorylation, also inhibited cyclin D-cdk4 activity (Fig. 2I, lane 8), suggesting that phosphorylation of both Y74 and Y88 is required to activate cyclin D-cdk4. To verify that the addition of phosphorylated p27 did not increase the assembly of residual, monomeric cyclin D and cdk4 components in the recombinant preparation, we immunoprecipitated cdk4 from the cyclin D-cdk4 and p27 incubations and probed for cyclin D levels. We found that the same amount of complex was present in the presence or absence of phosphorylated p27 (Fig. 2I, lower panel). Thus, the K1 site was required for Brk's phosphorylation of Y88, and loss of this phosphorylation converted p27 into a cdk4 inhibitor in vitro.

The isolated Brk SH3 domain or an isolated p27 K1 peptide was able to block Brk's phosphorylation of p27.

It appeared that the SH3-PXXP interaction was required for Y88 phosphorylation. To verify this, we attempted to block this interaction by the addition of a Flag-tagged K1 site containing peptide F58-106 (Fig. 3A) or the isolated SH3 domain (Fig. 3C) as a competitor in the p27-Brk interaction. Peptide F58-106 encompasses a portion of p27 (residues 58 to 106) that contains the K1 site only along with phosphorylation sites Y74 and Y88. While Brk phosphorylates p27 on both residue Y74 and residue Y88 (Fig. 3A, lane 3), it was able to phosphorylate F58-106 only on residue Y74 (Fig. 3A, lane 4), even though residue Y88 is intact in this peptide. When p27, Brk, and increasing amounts of the K1 site-containing competitor F58-106 were incubated together, phosphorylation on residue Y88 decreased (Fig. 3A, lanes 5 to 8). Y74 phosphorylation was reduced only slightly when the highest concentration of F58-106 was used (Fig. 3A, lanes 5 to 8).

FIG 3.

SH3-PXXP interaction is required for p27 Y88 phosphorylation. (A) p27 and Brk were incubated with ATP and with increasing concentrations of the PXXP competitor, F58-106, followed by immunoblot analysis with pY88, pY74, and p27 antibodies. In lane 5, “+” represents equimolar amounts of F58-106 and p27; in lane 6, “++” represents 2.5× F58-106; in lane 7, “+++” represents 5× F58-106; and in lane 8, “++++” represents 10× F58-106. (B) GST-Brk, p27, and F58-106 were incubated and immunoprecipitated (IP) with GST antibodies, followed by immunoblot analysis with GST or p27 antibodies. (C) Increasing amounts of GST-SH3, GST-SH2, or GST peptides were incubated with p27, followed by immunoblot analysis with pY88, pY74, p27, and GST antibodies. Lanes 3, 5, and 7, 2.5× competitor; lanes 4, 6, and 8, 5× competitor. (D) GST-SH3, GST-SH2, or GST peptides were incubated with p27 or ΔK1/K3, followed by metal agarose chromatography and immunoblot analysis with GST and p27 antibodies. Loading control: GST peptides were added to the reaction, and the data are not representative of association.

To demonstrate that the K1-containing peptide, F58-106, could interact with Brk, we incubated GST-Brk with p27 and/or F58-106, followed by immunoprecipitation with GST antibodies and immunoblot analysis with GST and p27 antibodies (Fig. 3B). F58-106 interacted with GST-Brk (Fig. 3B, lanes 5 and 6), demonstrating that this K1-containing peptide was sufficient to associate with full-length Brk.

We then performed the converse reaction, and increasing amounts of individual recombinant, GST-tagged pieces of Brk (GST-SH2 and GST-SH3) or GST peptides were incubated with p27 before the addition of full-length Brk and ATP. The reactions were probed by immunoblot analysis with pY88 and pY74 antibodies (Fig. 3C). We found that the prior addition of GST-SH3 was sufficient to reduce Y88 and, to a lesser extent, Y74 phosphorylation (Fig. 3C, lanes 3 and 4). The addition of the GST peptide (Fig. 3C, lanes 5 and 6) or the GST-SH2 peptide (Fig. 3C, lanes 7 and 8) did not reduce Y88 or Y74 phosphorylation (Fig. 3C, compare lanes 3, 5, and 7), with the exception of addition of a 10-molar excess of GST (Fig. 3C, lane 6). p27 and GST immunoblot analysis was also performed, as a measure of the total protein amounts added to the reactions, and the results did not represent their association.

We incubated the GST, GST-SH3, or GST-SH2 peptides with His-tagged versions of p27 and the ΔK1/K3 mutant, isolated complexes by metal agarose chromatography, and performed immunoblot analysis with GST antibodies (Fig. 3D). The GST-SH3 peptide interacted with p27 in a K1/K3-dependent fashion: it bound to WT p27 but did not associate with ΔK1/K3 p27 (Fig. 3D, lanes 4 and 5). We also found that the SH2 peptide bound to both WT p27 and ΔK1/K3 (Fig. 3D, lanes 2 and 3), suggesting that Brk had a second binding site on p27. However, this binding was independent of the PXXP K1 site (Fig. 3D, lane 3). Because this experiment was performed in the absence of ATP, the results demonstrated that this SH2 domain recognizes p27 in an atypical phosphotyrosine-independent fashion. Where on p27 the SH2 domain associates and/or the significance of this interaction is unclear, but this finding explains the results represented by Fig. 2A, where we found that full-length Brk was still able to interact with mutant ΔK1/K3 in vitro (Fig. 2A, lane 4).

These data suggest that the SH3 domain binds to p27 in a K1 site-dependent manner. The K1 site and the SH3 domain mediate p27 Y88 phosphorylation in vitro, and blocking this interaction is sufficient to prevent Y88 phosphorylation. While the SH2 domain binds to p27 and may contribute to Brk's association with p27, it does not appear responsible or required for Brk's phosphorylation of Y88, which is mediated by the SH3 domain.

Brk interacts with p27 in vivo.

While we found that Brk was a high-affinity kinase for p27, able to phosphorylate p27 in vitro, we wanted to demonstrate that this interaction was physiological. Brk was detected by immunoblot analysis with a C-terminal Brk antibody in several breast cancer cells (MDA MB 231 and MCF7) and in the MCF10A normal mammary epithelial cell line (Fig. 4A). As reported by others, it was not detected in the MDA MB 468 breast cancer line (33). To verify that Brk specifically interacted with p27 in vivo, we immunoprecipitated Brk and examined the immunoprecipitates for p27 association (Fig. 4B) or immunoprecipitated p27 and examined the immunoprecipitates for Brk association (Fig. 4C), demonstrating that this interaction could be seen in vivo under physiological conditions with endogenous proteins. Brk was not detected in p27 immunoprecipitates from MDA MB 468 cells, where endogenous Brk was not expressed (Fig. 4C, lane 4).

FIG 4.

Brk phosphorylates p27 in vivo. (A) Breast cancer cell lines (MDA MB 468, MDA MB 231, and MCF7) and a normal mammary epithelial cell line (MCF10A) were analyzed by immunoblot analysis with Brk antibodies. (B) Lysates were immunoprecipitated using the C-terminal Brk antibody, and immunoblot analysis was performed using Brk and p27 antibodies. (C) Lysates were immunoprecipitated with p27 antibodies and probed for Brk and p27. Immunoprecipitation with IgG served as a negative control. (D to G) Expression of the WT and p27 mutants was induced (+) by adding tetracycline to the medium, and the results were compared with those seen with the uninduced cells (–). (D) Immunoblot analysis was performed using p27 and actin antibodies. (E) Lysates were analyzed by PI staining and FACS analysis. (F) Brk was immunoprecipitated from the lysates, and immunoblot analysis was performed using Flag and Brk antibodies. Immunoprecipitation with IgG served as a negative control. (G) Flag-tagged WT or mutant p27 was immunoprecipitated from the lysates using Flag antibodies, and immunoblot analysis was performed using p27, pY88, and pY74. Immunoprecipitation with Flag antibodies (IgG) without the lysate served as a negative control.

To demonstrate that the in vivo p27-Brk interaction was mediated through the PXXP motifs (K1 or K3) of p27, we expressed, in a tetracycline-inducible manner, Flag-tagged variants of p27-WT, ΔK1, ΔK3, or ΔK1/K3 in MCF7 cells. In the presence of tetracycline, the mutants were greatly overexpressed relative to the endogenous p27 levels, which, however, cannot be seen in the corresponding figure panel (Fig. 4D, lanes 2, 4, 6, and 8). Expression of p27 and all the mutants caused G₁ growth arrest, as measured by PI staining and FACS analysis (Fig. 4E) and by monitoring proliferation by cell counting (data not shown). This was consistent with previous results, where we showed that p27 inhibits the other G₁ cdk, cdk2, in a manner independent of its Y88 phosphorylation status and that G₁ arrest is detected when cdk2 is inhibited (12).

We immunoprecipitated lysates with Brk antibodies and probed immunoblots with Flag antibodies to specifically detect the association of p27 mutants with endogenous Brk (Fig. 4F). We found that Brk associated with WT p27 (Fig. 4F, lane 2), but its association with the ΔK1 or ΔK1/K3 mutants was reduced (Fig. 4F, lanes 4 and 8). Brk continued to associate with the ΔK3 mutant, suggesting that in vivo, the K1 region was primarily responsible for p27's interaction with Brk (Fig. 4F, lane 6). The association of Brk with the p27 mutants was not completely lost upon deletion of the K1 and K3 sites, consistent with the results seen in the in vitro binding assays (Fig. 2A and 3D).

To determine whether the K1 or K3 site was responsible for Y88 phosphorylation in vivo, the mutants were isolated by Flag affinity chromatography and probed by immunoblot analysis with pY88, pY74, and p27 antibodies (Fig. 4G). Consistent with the results seen in vitro with the recombinant proteins, WT and ΔK3 were still phosphorylated on Y88, suggesting that the K1 site was primarily responsible. All of the mutants were still phosphorylated on residue Y74 (Fig. 4G), consistent with in vitro results that suggested that this phosphorylation event was K1 independent. Thus, while the K3 site was originally identified as the better interactor with Brk's SH3 domain, in the context of full-length p27 and full-length kinases, the K1 site mediates this interaction.

Modulating Brk levels in vivo modulates p27 Y88 phosphorylation.

Our data suggest that Brk phosphorylates p27 on Y88 in vitro, which leads to the model that modulation of Brk would modulate p27 Y88 phosphorylation. Brk activity has been reported to be insulin sensitive (34), so in order to increase or decrease the levels of endogenous Brk, we increased or decreased the level of insulin in the tissue culture medium for MCF7 cells (Fig. 5A). When the cells were grown in the standard 10 μg/ml insulin, expression of Brk, Src, cyclin D, and Cdk4 was detected by immunoblot analysis with the respective antibodies (Fig. 5A, +). However, when insulin was removed from the medium, Brk expression specifically decreased, while the expression of Src, cyclin D, and cdk4 remained unchanged (Fig. 5A, –). When the concentration of insulin was increased to 50 μg/ml in the medium, Brk expression increased, again without a concomitant change in the expression of Src, cyclin D, or cdk4 (Fig. 5A, +++). The modulation of Brk levels resulted in a corresponding decrease or increase in cell proliferation, as detected by cell counting (Fig. 5B, left panel), without a change in cell viability (Fig. 5B, right panel), suggesting cell cycle arrest rather than cell death. Decreasing insulin and Brk expression reduced p27 Y88 phosphorylation, while increasing insulin and Brk expression increased Y88 phosphorylation, as detected by immunoblot analysis with pY88 antibodies (Fig. 5C). This suggests that Y88 phosphorylation is dependent on Brk levels. Y74 phosphorylation was not changed under the increased or decreased Brk conditions, suggesting that this phosphorylation occurred more constitutively in vivo (Fig. 5C).

FIG 5.

Modulation of Brk protein levels modulates p27 phosphorylation. (A to C) MCF7 cells were cultured in the presence of 10 μg/ml insulin (+), no insulin (–), or 50 μg/ml insulin (+++). (A) Immunoblot analysis was performed with Brk, c-Src, cyclin D1, cdk4, and actin antibodies. (B) Treated cells were counted to monitor proliferation (left panel). Viability was determined using trypan blue staining (right panel). (C) p27 was immunoprecipitated, and immunoblot analysis was performed with p27, pY88, and pY74 antibodies. (D) Brk protein expression was knocked down by using two different siRNAs. Immunofluorescence was performed on the cells 72 h postinfection. Cells were probed for Brk, p27, pY88, pY74, and c-Src.

To directly demonstrate that loss of Brk affected p27 Y88 phosphorylation, we knocked down Brk using two different siRNAs that were previously shown to be directed against human Brk (35). Lentiviruses that expressed these siRNAs were used to infect MCF7 cells. Cells were allowed to recover for 72 h postinfection, and then we performed immunofluorescence analysis with Brk, p27, pY88, pY74, and Src antibodies (Fig. 5D, complete experiments with controls; see also Fig. S1 in the supplemental material). While expression was detected with all antibodies in the mock-infected (Mock) cells, both siRNAs effectively reduced Brk expression (Fig. 5D, lane 1), without altering either p27 expression (Fig. 5D, lane 2) or Src expression (Fig. 5D, lane 5). The localization of p27 became more nuclear in cells treated with both siRNAs (Fig. 5D, lane 2), presumably due to the growth arrest that would result from the absence of Brk (Fig. 5B). However, loss of Brk specifically reduced Y88 phosphorylation (Fig. 5D, lane 3), while leaving Y74 phosphorylation intact. This suggests that Brk is the physiological kinase responsible for p27 phosphorylation in vivo. While Src may be competent to phosphorylate Y88, it does not appear to compensate for Brk's loss in vivo. Phosphorylation on residue Y74 appears independent of Brk's expression.

To further confirm that modulation of Brk modulates p27 Y88 phosphorylation, we expressed WT Brk, Brk KM (a catalytically inactive variant), and Brk YF (a constitutively active variant) (31) in MCF7 cells (Fig. 6A). Constitutive expression of WT Brk increased proliferation, as measured by cell counting (Fig. 6B), without any change in viability as measured by trypan blue staining (data not shown). Expression of the YF or KM variants did not increase proliferation, and the proliferation of these lines was similar to that seen with the mock infection-expressing line (Fig. 6B). We immunoprecipitated p27 from the cell lines and performed immunoblot analysis, using pY88, pY74, and p27 antibodies (Fig. 6C). The levels of Y88 phosphorylation were similar in the mock, KM, and YF cells, but the level was increased in WT-expressing cells. Y74 phosphorylation levels were similar in all four lines. We also performed immunofluorescence analysis with Brk, p27, and pY88 antibodies (Fig. 6D; see also Fig. S2A in the supplemental material). We found that Brk was overexpressed in all three Brk-expressing lines (WT, KM, and YF) relative to the mock infection-expressing cells. Because these cells were analyzed only 24 h postplating, Brk was predominantly cytoplasmic (Fig. 6D). Its localization appears to have become more nuclear as cells recovered postplating (data not shown). Consistent with the immunoprecipitation results, we detected increased Y88 phosphorylation only in the WT cells. This suggests that increased p27 Y88 phosphorylation was dependent on the presence of a kinase-active Brk. It was unclear why p27 Y88 phosphorylation was not increased in the YF line, but the reason became apparent in the experiments described below (see Fig. 8E).

FIG 6.

Constitutive expression of Brk makes MCF7 cells more resistant to PD 0332991. (A to D) MCF7 Mock, Brk WT, Brk YF (constitutively active), or Brk KM (catalytically inactive) cells were analyzed by immunoblot analysis using Brk and actin antibodies (A), for proliferation rate by cell counting (B), by p27 immunoprecipitation followed by immunoblot analysis with pY88, pY74, and p27 antibodies (C), or by coimmunofluorescence analysis using Brk (green) and pY88 (red) antibodies (D). (C) Immunoprecipitation with IgG served as a negative control. (D) DNA with Hoechst staining is shown in blue and is merged with red and green staining. (E) Cdk4 was immunoprecipitated from the lysates, and recombinant (Rec.) Rb was added to perform an in vitro kinase assay in the presence of ATP. Immunoblot analysis was performed to probe for Rb, Ser780 pRb, and cdk4. Immunoprecipitation with IgG served as a negative control. (F) Cells were treated with different concentrations of PD 0332991 or DMSO as a control at day 2. Total cell numbers were analyzed by counting 24 and 48 h posttreatment to determine IC₅₀s. (G) Cells were treated with different concentrations of PD032291 or DMSO and subjected to PI staining and FACS analysis. Percent G₁ content is plotted. Data are representative of the results of n = 3 experiments.

FIG 8.

Alt Brk acts as an endogenous inhibitor of Brk. (A) Flag-tagged WT p27 was incubated with increasing amounts of Flag-tagged Alt Brk (lanes 5 to 7) in the presence of equal amounts of Brk. In lane 5, “+” represents equimolar Brk and Alt; in lane 6, “++” represents 2.5× Alt; and in lane 7, “+++” represents 5× Alt. Immunoblot analysis was performed using Flag, pY88, and pY74 antibodies. (B) Alt Brk-expressing cells were plated on day 0. They were induced (+) or left uninduced (–) on day 1, and cells were harvested on days 2, 3, and 4. Immunoblot analysis was performed using N-terminal Brk, C-terminal Brk, Rb, Ser780 pRb, Flag, or actin antibody. (C) p27 was immunoprecipitated from induced (+) or uninduced (–) Alt Brk-expressing cells, and immunoblot analysis was performed with p27, pY88, and pY74 antibodies. (D) Cells (Alt Brk or TRE3G vector alone) were treated with Tet (+) or left untreated (−Tet), and cell counts were used to monitor proliferation. (E) Lysates from MCF7 Mock, WT, YF (constitutively active), or KM (catalytically inactive) cells were used in immunoblot analysis with the N-terminal Brk antibody (Alt Brk) and C-terminal Brk (Brk), pY342 Brk (active Brk), and actin antibodies. (F) p27 was immunoprecipitated from MCF7 cells that were contact arrested (G₀), asynchronous (A), or serum starved with 1% and 0% serum for 48 h. Immunoblot analysis was performed to probe for p27, pY88, and pY74. IgG and Mv1Lu lysates served as a negative control. (G) Expression of WT or mutant p27 was induced by adding tetracycline to the culture medium. p27 was immunoprecipitated from cell lysates and probed for p27, pY88, and pY74. Immunoprecipitation with IgG served as a negative control.

Our data demonstrated that increasing Brk expression increased p27 Y88 phosphorylation, and we hypothesized that this would increase cdk4 kinase activity. To directly examine this, we immunoprecipitated cdk4-associated complexes from Mock, WT Brk, Brk KM, and Brk YF cell lines and performed in vitro Rb kinase assays (Fig. 6E). We added exogenous Rb substrate to these immunoprecipitates and then assayed for Rb phosphorylation by immunoblot analysis using a RB phospho-specific antibody that recognizes residue Ser780, which was previously shown to be specifically phosphorylated by cdk4 (36). Cdk4-associated kinase activity was detected in the mock infection-, Y-, and KM-expressing clones (Fig. 6E, lanes 1, 3, and 4) and represented endogenous cdk4 activity. However, increased cdk4-associated Rb phosphorylation was detected in the WT Brk-expressing cells (Fig. 6E, lane 2), consistent with the increase in p27 Y88 phosphorylation detected here.

We additionally treated the Brk-expressing cell lines with PD 0332991, a cdk4-specific inhibitor that causes a potent G₁ arrest (Fig. 6F) (37, 38). This small-molecule inhibitor is exquisitely specific for cdk4, stoichiometrically blocking catalytic activity, and, at the concentrations used in this assay, has no activity against other serine/threonine kinases (38). Cells were plated at day 0, treated with four different concentrations of PD 0332991 at day 2, and then harvested and counted at days 3 and 4 (24 and 48 h posttreatment). IC₅₀s, defined as the concentrations of drug needed to inhibit proliferation by 50%, were calculated (Fig. 6F). As seen by others, the IC₅₀s of MCF10A and the Mock MCF7 cells were 110 and 150 nM, respectively, and these lines are considered sensitive to PD 0332991 treatment. MDA MB 468 cells were resistant to PD 0332991 treatment, consistent with the results determined by others (39), with an IC₅₀ of 350 nM.

The MCF7 WT-expressing cells, with their increased p27 Y88 phosphorylation, were now resistant to PD 0332991 treatment, with IC₅₀s of greater than 600 nM (Fig. 6F). The MCF7 KM and YF cells, with nearly endogenous levels of Y88 phosphorylation, had IC₅₀s similar to those measured for the mock cells. The level of p27 Y88 phosphorylation correlated with cdk4 sensitivity: increased Y88 phosphorylation resulted in increased IC₅₀s. These data suggest that the WT line had more cdk4 kinase activity that necessitated more drug in order for inhibition to occur. By immunoblot analysis, we demonstrated that the levels of cyclin D and cdk4 did not increase significantly in the WT Brk-expressing cells in the presence or absence of PD 0332991 (data not shown), suggesting that the activity of the complex increased instead as a result of the increased Y88 phosphorylation seen in this line. FACS analysis confirmed that the arrest seen in the Mock, KM, and YF cells treated with PD 0332991 at up to 100 nM was due to an increase in G₁ content (Fig. 6G). The WT-expressing cells, however, maintained approximately 55% G₁ content at these concentrations. The viability of these cells was unchanged with all concentrations of the drug, suggesting that the PD 0332991 treatment was cytostatic and not cytotoxic (data not shown). These data suggest that Brk overexpression increases Y88 phosphorylation and increases cyclin D-cdk4 activity, rendering those cells more resistant to specific cdk4-inhibitor therapy.

Alt Brk acts as an endogenous inhibitor of p27 phosphorylation.

Our data suggest that the level of Brk dictates the level of p27 Y88 phosphorylation, since modulating Brk levels modulates Y88 phosphorylation. We had previously demonstrated that p27 Y88 phosphorylation was lost in contact-arrested cells, suggesting that this was one way by which cdk4 activity was inhibited under this condition (12). The MCF10A, MDA MB 231, and MCF7 breast cell lines could all be contact arrested when grown to confluence and maintained for 6 days in the presence of replenished serum, as shown by increased G₀/G₁ content (Fig. 7A, bottom panel, %G0/G1). We immunoprecipitated endogenous p27 from asynchronously growing (A) breast cancer cells and contact-arrested (G₀) breast cancer cells, followed by immunoblot analysis with pY88, pY74, and p27 antibodies (Fig. 7A, bottom panel). As expected, Y88 phosphorylation was detected only in the proliferating cells and was absent in the G₀-arrested cells. Y74 phosphorylation was still detected in the G₀-arrested cells, consistent with our observations that this phosphorylation was more differentially regulated than Y88 phosphorylation.

FIG 7.

Alt Brk is upregulated in G₀ cells. (A) Lysates from asynchronous (A) or contact-arrested (G₀) cells were analyzed for Brk expression using immunoblot analysis (top panel). p27 was immunoprecipitated, followed by immunoblot analysis with p27, pY88, and pY74 antibodies (bottom panel). Data represent percent G₀/G₁ as determined by PI staining and FACS analysis (top panel, %G0/G1). Immunoprecipitation with IgG without the lysate served as a negative control. (B) Brk was immunoprecipitated using the C-terminal Brk antibody and recombinant p27, and ATP was added to perform an in vitro kinase assay. Immunoblotting was performed to probe for p27, Brk, pY88, and pY74. Lane 3, no recombinant p27 added; lane 4, IP with IgG negative control; lane 5, no immunoprecipitate added. (C and D) Lysates were subjected to immunoblot analysis with the N-terminal Brk antibody. (E) RNA was extracted from A and G₀ cells, and q-RT-PCR was performed to probe for Brk and Alt-Brk mRNA expression. The ratio of Alt Brk to Brk is plotted.

However, when we examined the levels of Brk using a C-terminal Brk antibody, we found that, in fact, Brk expression did not decrease in the G₀ cells but, rather, increased (Fig. 7A, top panel). To verify that this Brk was catalytically active, we immunoprecipitated Brk from A and G₀ cells and then incubated the immunoprecipitates with recombinant p27 and ATP in vitro (Fig. 7B). Brk immunoprecipitated from both A and G₀ cells was able to phosphorylate recombinant p27 (Fig. 7B, lanes 1 and 2). We immunoprecipitated this Brk under stringent conditions, so endogenous p27 was not recovered and phosphorylation was detected only in the reactions where recombinant p27 was added (Fig. 7B, compare lanes 1 and 2 to lane 3).

Thus, Brk was present and active in contact-arrested (G₀) breast cells but p27 was not phosphorylated on residue Y88 (Fig. 7A, bottom panel). An alternative splice variant of Brk, Alt Brk, which lacks expression of exon 2 and encodes a shorter, 15-kDa protein, has been reported in the T47D breast cancer cell line and in several prostate and colon cancer cell lines (29, 40). This Alt Brk shares the N-terminal SH3 domain with Brk and has a unique proline-rich carboxy terminus but lacks the catalytically active SH1 kinase domain (29). Given our results where we demonstrated that the addition of exogenous SH3 peptides could block Brk's phosphorylation of recombinant p27 (Fig. 3C), we hypothesized that Alt Brk might compete with full-length Brk for binding to p27 and function as an endogenous inhibitor of p27 phosphorylation. Using an N-terminal-specific antibody of Brk which would recognize both full-length and Alt Brk, we detected Alt in G₀ cells (Fig. 7C and D). We performed q-RT-PCR on lysates derived from proliferating (A) or contact-arrested (G₀) cells, and found that the ratio of Alt Brk to Brk was significantly increased in G₀ cells (Fig. 7E).

These data suggested that the presence of increased Alt could lead to its increased association with p27, which might block the interaction of full-length Brk or essentially outcompete Brk for p27's association. To directly verify this, we expressed a Flag-tagged Alt Brk in bacteria and purified this 15-kDa protein (Fig. 8A). When p27 was incubated with Brk, Y88, and Y74, phosphorylation was detected by immunoblot analysis (Fig. 8A, lane 3). When increasing concentrations of Alt Brk were added to the reaction, Y88 phosphorylation decreased (Fig. 8A, lanes 5 to 7). Y74 phosphorylation was not affected. This suggested that Alt Brk was able to function as an inhibitor, similarly to the small SH3 peptide used in Fig. 3C, blocking Brk's phosphorylation of residue Y88.

To verify this in vivo, we expressed Flag-tagged Alt Brk in a tetracycline-inducible manner in MCF7 cells (Fig. 8B to D). When tetracycline was added to the culture medium, Alt Brk expression was detected using the N-terminal Brk and Flag antibodies (Fig. 8B), and cells were arrested in G₁ phase, as detected by cell counting (Fig. 8D) and FACS analysis (data not shown). When endogenous p27 was immunoprecipitated from the Alt Brk-expressing cells, p27 Y88 phosphorylation was reduced, while Y74 phosphorylation was unchanged (Fig. 8C), suggesting that Alt Brk functions in vivo as an endogenous inhibitor of p27 Y88 phosphorylation. We examined Rb phosphorylation on residue Ser780 as a measure of cdk4 kinase activity (Fig. 8B). Rb phosphorylation was reduced consistent with the loss of p27 Y88 phosphorylation, which would translate into a reduction in cdk4 activity and growth arrest.

Our model suggested that increasing Brk would increase p27 Y88 phosphorylation, which in turn would increase cdk4 activity and PD 0332991 resistance. However, Alt Brk, functioning as an endogenous inhibitor of Brk's phosphorylation of p27, could dampen this cascade. We returned to examine the MCF7 cells that overexpressed Brk YF, the catalytically active variant, as described in the Fig. 6 legend. It had been unclear why this mutant did not cause an increase in p27 Y88 phosphorylation even though exogenous Brk was detected by both immunoblot analysis (Fig. 6A and 8E) and immunofluorescence analysis (Fig. 6D). When we examined the YF-expressing cells using an antibody that recognizes active Brk (pY342), we found that the WT- and the YF-expressing cells had more active Brk than either the mock infection- or KM-expressing cells (Fig. 8E). However, using the N-terminal Brk antibody, we now detected an increase in Alt Brk expression specifically in the YF cell line (Fig. 8E, lane 3), suggesting that even though exogenous Brk was expressed, its inhibitor was also expressed. This explained why, in the YF line, we had not detected a change in p27 Y88 phosphorylation (Fig. 6C and D) and had not seen an increase in PD 0332991 resistance (Fig. 6E). Our data demonstrated that the ratio of Alt Brk to Brk dictates the status of p27 Y88 phosphorylation, which in turn regulates cdk4 activity and PD 0332991 sensitivity. Thus, Alt Brk, with its SH3 domain, functions as an endogenous cdk4 inhibitor.

While Alt Brk was able to block Y88 phosphorylation, it did not affect Y74 phosphorylation, consistent with our results that suggested that Y74 phosphorylation is differentially regulated. Y74 phosphorylation was not affected by loss of Brk expression (Fig. 5D, lane 4) or in cells arrested by contact (Fig. 7A, bottom panel), and loss of the K1 site did not prevent Y74 phosphorylation in vitro (Fig. 2E), and in vivo (Fig. 4G, lanes 4 and 6). We did see loss of Y74 phosphorylation when MCF7 cells were grown in the absence of serum (Fig. 8F, lanes 3 and 4), suggesting mitogen dependence. We examined the phosphorylation state of p27 and mutants Y74F and Y88/89F overexpressed in MCF7 cells in a tetracycline-inducible manner. In the presence of Tet, these mutants were detected by immunoblot analysis using p27 antibodies, while endogenous p27 levels were too low to be detected (Fig. 8G, lanes 2, 4, and 6). WT p27 was phosphorylated on both residues. Mutant Y88/89F was not phosphorylated on residue 88 or residue 89 due to mutation but was still phosphorylated on residue Y74 (Fig. 8G, lane 6). Mutant Y74F was not phosphorylated on residue Y74, as expected, or on residue Y88 (Fig. 8G, lane 4). Together, these data suggest that phosphorylation on Y74 is Brk independent but may be required for efficient phosphorylation on Y88 by Brk. However, Y88 phosphorylation, which is regulated by Brk in an SH3-PXXP-dependent manner, is the switch that modulates cdk4 activity.

DISCUSSION

We have identified Brk/PTK6 as an authentic p27 kinase that can activate the p27 on/off switch and that is thus able to modulate cyclin D-cdk4 activity. While many SFKs appear competent to phosphorylate p27, Brk phosphorylates p27 more efficiently, and its SH3 domain associates with a higher affinity. Reducing Brk expression by siRNA in vivo eliminates p27 Y88 phosphorylation, even though Src is expressed, demonstrating that Brk is the physiological kinase in these cells. It is striking that Brk is overexpressed in many of the same cancers that appear dependent on cdk4 kinase activity. Increased expression of Brk in breast cancer cells would increase p27 Y phosphorylation and increase resistance to cdk4 inhibition in a kinase-dependent fashion, suggesting that a limiting factor in this type of therapy is the level of active cdk4. These data lead to the following model: at least in some tumors, Brk expression regulates p27 Y phosphorylation, which in turn regulates cdk4 activity and cell cycle progression. In ongoing work, we are actively looking for this direct connection in breast tumors. We have detected both Brk overexpression and Y88 phosphorylation in primary breast tumors embedded in paraffin (unpublished data). It remains to be determined whether p27 Y phosphorylation or Brk expression will serve as a marker for cdk4 activity and in turn for cdk4 inhibitor sensitivity.

Brk protein is detected in both the cytoplasm and nucleus of normal human mammary cells, but it appears to be catalytically inactive. However, Brk is overexpressed in more than 60% of human breast carcinomas (41, 42), and in high-grade human breast tumors, it is both overexpressed and active at the plasma membrane (42, 43), suggesting constitutive signaling in these tumors. Its expression promotes proliferation and tumor growth in human mammary epithelial cells, although the direct substrate(s) required for this tumor-promoting effect had not been identified. Brk lacks the amino-terminal myristoylation/palmitoylation typical of Src family members and, as such, has a wider area of localization and binding partners (44–46). Several Brk substrates have been identified (24), including β-catenin, p190RhoGAP, Paxillin, PSF, STAT3, Sam68, SLM1, SLM2, AKT, p130CAS, and FAK (47), but the identification of p27 as a direct phosphorylation target provides new insights about Brk's role in proliferation control and directly links it to cdk regulation. Others have suggested that Brk has additional roles in p27 regulation: in MDA MB 231 cells and Src, Yes, and Fyn null mouse embryonic fibroblast (MEFs), Brk overexpression transcriptionally downregulates p27 (47, 48).

Brk appears to phosphorylate p27 on residue Y88 in an SH3-dependent manner. Loss of the K1 site prevents Y88 phosphorylation in vitro and in vivo, and addition of either a K1-containing peptide or an SH3-containing peptide is able to prevent Y88 phosphorylation. The importance of the PXXP-SH3 interaction is further demonstrated by the effect of Alt Brk expression. Alt Brk contains Brk's SH3 domain, but lacks the kinase domain, and is able to inhibit Brk's phosphorylation of p27 in vitro and in vivo. Alt Brk appears to function as a competitive inhibitor, and our data consistently demonstrated that the ratio of Alt Brk to Brk dictated the status of p27 Y88 phosphorylation, which in turn regulates cdk4 activity and PD 0332991 sensitivity. Alt levels were increased in cells arrested by contact, but the regulation of Alt with respect to Brk and its potential role in tumors remain to be determined. We found an increase in Alt Brk levels when constitutively active YF Brk was overexpressed, suggesting that the expression of Alt may be due in part to self-regulation by Brk itself. Alt has been found to associate in vivo with additional other Brk substrates, and in its presence, phosphorylation of Brk itself and overall phosphotyrosine levels were reduced (29). Thus, exogenous expression of Alt may affect the activity of other substrates, which might contribute to the associated growth arrest. However, our data clearly demonstrated that Alt Brk functions as an endogenous inhibitor of p27 Y phosphorylation. This suggests that we have identified a novel, potentially targetable domain and that blocking the PXXP-SH3 interaction might be a viable strategy to inhibit p27 Y phosphorylation and cdk4 activity, which should be explored therapeutically.

p27 is a well-characterized tumor suppressor whose loss or reduction appears to be required to activate oncogenic cdk2. However, because of its role as an activator of cyclin D-cdk4 complexes, p27 may also function as an oncoprotein (49, 50). In the ErbB2 breast cancer model, tumorigenesis is accelerated in p27^+/− mice compared to p27^+/+ animals, while tumorigenesis is blocked in the complete absence of p27 (p27^−/−) (51, 52, 57). Thus, in this breast cancer model, while p27 levels must be reduced to release oncogenic cdk2, residual p27 is required to accelerate tumor formation, via assembly and activation of cyclin D-cdk4. In humans, p27 is rarely mutated or silenced, suggesting that a similar requirement for residual p27 levels may exist (52). p27 levels are reduced by accelerated proteolysis or cytoplasmic mislocalization, likely related to increased oncogenic signaling. Decreased but residual p27 levels correlate with more-aggressive phenotypes, high proliferation indices, increased invasive behavior, and high mortality (52). Thus, while p27 levels are reduced in human tumors, the residual p27 that remains would be switched on to activate cdk4. This would imply that in tumors, a decreased level of p27, with a concomitant increased level of pY88, would be oncogenic. Direct evidence for an oncogenic role of the Cip/Kip proteins has been demonstrated only for p21 in a glioblastoma model, where loss of the homologous Y phosphorylation site in p21 prevented platelet-derived growth factor (PDGF)-dependent tumor formation (53). A formal description of p27's oncogenic role in animal models is still lacking but is being actively pursued.

Characterization of p27's oncogenic function is important because cdk4 has been a highly sought-after therapeutic target for decades, given that cdk4 and cyclin D are frequently overexpressed in many tumor types. PD 0332991, now known as palbociclib, is currently in clinical trials for multiple myeloma and breast cancer and has shown promising results (54). This is the first cdk4 inhibitor with the required specificity to provide a therapeutic benefit, extending median progression-free survival (PFS) for metastatic breast cancer patients from 10.2 months with letrozole alone to 20.2 months with the combination of letrozole and palbociclib (55). In this study, cyclin D amplification and/or loss of p16 did not correlate with sensitivity. The best biomarker of response to date is RB positivity, but even that is not full proof, suggesting that a marker for cdk4 activity itself, such as p27 Y phosphorylation, will be required. For example, RB⁺ pancreatic ductal adenocarcinomas (PDAC) were a priori considered targetable tumor types, dependent on cdk4 activity, due to the early loss of the cdk4 inhibitor p16 (INK4A) and activation of RAS seen in approximately 80% of cases. However, most PDAC cells appear resistant to CDK4/6 inhibition. Recently, it was shown that PD 0332991 synergizes with IGF-1R inhibitors to repress the growth of PDAC (56). Given Brk's insulin sensitivity, it is interesting to speculate that reducing IGF-1R signaling might reduce Brk activity and decrease p27 Y phosphorylation, which in turn might reduce cdk4 activity to a level that would then be inhibited by PD 0332991. Brk-specific inhibitors or p27 Y phosphorylation-specific inhibitors, such as Alt Brk, may synergize with PD 0332991, decreasing the amount of active cdk4 and increasing the efficacy of this type of therapy. Screening for p27 Y phosphorylation might serve as an indicator of tumors that would be responsive to cdk4-specific inhibition. If the tumors did not contain p27 pY88, they would likely not respond to palbociclib, but if they contained too much p27 pY88 and too much cdk4 activity, they might be resistant.

The full significance of Y74 phosphorylation is still to be determined, but it appears to occur in an SH3-PXXP-independent fashion, and in the absence of Brk, additional Y kinases can compensate in vivo. As a monomer, Brk is able to phosphorylate p27 on Y88 and Y74 independently (Fig. 2E), but we found that Y74 may serve as a prerequisite to allow Y88 phosphorylation in vivo (mutant Y74F was not phosphorylated on either residue Y88 or residue Y74). Phosphorylation on both Y88 and Y74 is required to activate cyclin D-cdk4, so even in contact-arrested cells, where Y74 phosphorylation is still detected, cdk4 is inactive. This leads to the suggestion that p27 may be phosphorylated by multiple kinases in a mitogen-dependent fashion on residue Y74, with the complex remaining inactive until regulated SH3-dependent Y88 phosphorylation occurs to convert the complex to a higher-specific-activity form. Thus, the model of p27 activation of cdk4 might eventually be refined even further, where p27Off, p27Low, and p27High represent the unphosphorylated, singly phosphorylated, and fully phosphorylated p27 versions, which in turn would translate into different levels of cdk4 activity and would be generated in the presence of different levels or types of Y kinase activity. Y88 phosphorylation is the final switch that permits the p27 cyclin D-cdk4 ternary complex to be turned on, allowing both CAK activation and ATP coordination.

In summary, we have identified SH3 domain recruitment sequences within p27 that modulate Y88 phosphorylation and, therefore, cdk4 activity and as such have identified a new and potentially targetable regulatory region required for cdk4 activation and cell cycle progression. We have identified Brk/PTK6 as a physiological p27 kinase that can modulate cyclin D-cdk4 activity and whose overexpression in breast cancer cells renders them resistant to cdk4-specific inhibition. We have also identified an endogenous inhibitor, Alt Brk, which adds to the regulation of cdk4 activity and supports the idea of the importance of this interface as a bona fide therapeutic targeting site. p27 has long been considered a tumor suppressor, but our data further strengthen the idea that it should also be considered an oncoprotein, responsible for cyclin D-cdk4 activity. As the use of cdk4 inhibitor PD 0332991 continues to move from the bench to the bedside, a further understanding of cdk4 activity and biomarkers of its activation will be required.