PKCη/Rdx-driven phosphorylation of PDK1: a novel mechanism promoting cancer cell survival and permissiveness for parvovirus-induced lysis

Abstract

The intrinsic oncotropism and oncosuppressive activities of rodent protoparvoviruses (PVs) are opening new prospects for cancer virotherapy. Virus propagation, cytolytic activity, and spread are tightly connected to activation of the PDK1 signaling cascade, which delays stress-induced cell death and sustains functioning of the parvoviral protein NS1 through PKC(η)-driven modifications. Here we reveal a new PV-induced intracellular loop-back mechanism whereby PKCη/Rdx phosphorylates mouse PDK1:S138 and activates it independently of PI3-kinase signaling. The corresponding human PDK1phosphoS135 appears as a hallmark of highly aggressive brain tumors and may contribute to the very effective targeting of human gliomas by H-1PV. Strikingly, although H-1PV does not trigger PDK1 activation in normal human cells, such cells show enhanced viral DNA amplification and NS1-induced death upon expression of a constitutively active PDK1 mimicking PDK1phosphoS135. This modification thus appears as a marker of human glioma malignant progression and sensitivity to H-1PV-induced tumor cell killing.

Fig 1. MVM-induced activation of the PDK1/PKC/PKB signaling cascade.

As shown previously, MVM activates Rdx [22], PDK1, and PKCη [8] in permissive A9 mouse fibroblasts. This activation is accompanied by PDK1 and PKCη translocation from the plasma membrane to the perinuclear area, where PKCη co-localizes with Rdx [8,22]. (A) Asynchronously growing A9 cells were infected (or not) with CsCl-purified full MVM capsids (30 pfu/cell) or an equivalent amount of empty capsids (EC). Total cell extracts were prepared at the indicated times p.i. Activation of selected cell proteins by MVM was monitored by western blotting on the basis of the proteins’ capacity for auto-phosphorylation (PDK1phosphoS244; PKCηphosphoT655) or of their trans phosphorylation at residues known to be essential for activation (Rdx:phosphoT564, PKBphosphoT308 [PDK1 target] and PKBphosphoS473). These protein modifications were detected with phospho-specific antisera. Total amounts of PDK1 and PKB were determined in parallel, and α-tubulin was used as an internal control. It is noteworthy that PKCη activation (starting at 5 h p.i., red arrow) precedes activation of the slower migrating PDK1 form and PKB (starting at 15 h p.i., red arrows). In addition, like PDK1, a slower migrating form of Rdx becomes activated at 15 h p.i. coinciding with its phosphorylation by the NS1/CKIIα complex [22]. (B) Impact of MVM infection on the subcellular distribution of PDK1, PKCη, and radixin. A9 cells grown on spot slides were infected (or not) with CsCl-purified MVM (30 pfu/cell) and examined 36 h p.i. by confocal laser scanning microscopy to confirm colocalization of PDK1 (green), PKCη (red), and Rdx (blue). Colocalization appears white in the merge and was quantified with Image J software. Scale bars: 30 and 15 μm, as indicated.

Fig 2. Rdx interacts with PKCη and controls its activity and substrate specificity.

(A, B) A9 cells and derivatives expressing the gene encoding the indicated variant protein under the control of the NS1-inducible P38 promoter were infected with MVM (30 pfu/cell) and analyzed at the indicated times p.i. (A) Rdx interacts physically with PKCη inside cells. Left panel: Cell lines expressing MycPKCη (PKCη) alone or together with Flag-tagged CKIIαE81A (CKII), RdxT564A(Rdxa), RdxT564E (RdxE), EzT566A (EzA), EzT566E (EzE), or MoeT547A (MoeA), were harvested 36 h p.i. Co-immunoprecipitation assays were performed under non-denaturing conditions with mouse monoclonal Flag-tag-specific M2 antibodies. Immunoprecipitates (IPαFlag) and, for comparison, whole-cell lysates (Lys) were analyzed by western blotting with rabbit anti-Myc antibodies to detect MycPKCη. The percentage of Flag-positive cells in these lines was determined by immunofluorescence with M2 antibodies (% Flag⁺ cells). Arrows indicate the position of MycPKCη in CoIPs. n.d. stands for “not determined”. Right panel: A9, and cell lines expressing MycPKCη or MycCKIIα were harvested 36 h p.i. Co-immunoprecipitation assays were performed under non-denaturing conditions with anti-Myc antibodies. Immunoprecipitates (IPαMyc) and, for comparison, whole-cell lysates (Lys) were analyzed by western blotting with goat anti-Rdx antibodies to detect endogenous radixin. The percentage of Myc-positive cells in these lines was determined by immunofluorescence with anti-Myc antibodies (% Myc⁺ cells). Arrows indicate the position of Rdx in CoIPs (B) Rdx controls the activity of PKCη in MVM-infected A9 cells. A9 cells and derivatives expressing dominant-negative PKCηT512A (ηTA) or RdxT564A (RdxA) were harvested at the indicated times p.i. and analyzed by western blotting. As a measure of endogenous PKCη activity, the amount of PKCη auto-phosphorylated at T655 (ηP655) was estimated as compared to the total amount of the kinase (PKCη). The loading control was α-tubulin (Tubα). (C) Radixin controls the substrate specificity of PKCη. The MVM NS1 trans-activation domain, aa 545–672 (NS1_C) and C-terminally truncated PDK-1_N446 were phosphorylated in vitro by PKCη alone (PKCη) or with radixin (PKCη/Rdx) and their tryptic phosphopeptides were detected. Peptides labeled specifically in the presence of Rdx are indicated with arrows (presence) or dotted circles (absence).

Fig 3. MVM induces PDK1 activation through PKCη/Rdx-driven trans-phosphorylation at S138.

(A, B) A9 cell derivatives expressing MycPDK1_X alone or in the presence of the indicated dominant-negative effector protein were infected with MVM (30 pfu/cell) and, when indicated, metabolically labeled 24 h p.i. and further processed for analysis of the PDK1 tryptic phosphopeptide pattern. (A) PKCη/Rdx causes the appearance of a specific PDK1 phosphopeptide and controls the overall phosphorylation of PDK1. Top: Phosphorylated MycPDK1 was determined by metabolic ³²P-labeling of MycPDK1 after recovery by immunoprecipitation (P-MycPDK1). The total amount of MycPDK1 (MycPDK1) was measured 24 h p.i. by western blotting. Actin was used as a loading control. Bottom: As shown by Lachmann and coworkers, the tryptic phosphopeptide pattern of metabolically labeled MycPDK1 comprises peptides (a-d), which are autophosphorylated by PDK1 and absent in catalytically inactive PDK1S244A. In contrast, peptides (e, f), which are present in the profile of inactive PDK1S244A are targeted by (an)other kinase(s) [8]. Phosphorylation of “e” (arrow) was found to depend on both PKCη and Rdx (absence = dotted circles). dnCKII: CKIIαE81A, dnPKCη: PKCηT512A, dnEz: EzT566A, dnRdx: Rdxdl[P], dnMoe: MoeT547A). (B) Phosphorylation and activity of PDK1 mutants. Top: Overall phosphorylation of MycPDK1 measured after metabolic labeling 24 h p.i. (P-MycPDK1). Total (MycPDK1) and functionally active, autophosphorylated PDK1phosphoS244 (PDK1:P244) were quantified by western blotting. PKB was used as an internal loading control. Bottom: Tryptic phosphopeptide patterns of inactive (white) and active (black) PDK1 mutants. Specific loss of trans-phosphorylated peptides “e” (produced by PKCη/Rdx) and “f” is indicated by dotted circles. (C) PKCη/Rdx phosphorylates PDK1:S138 in vitro. PDK1_N446 was phosphorylated in vitro in the presence of PKCη alone (PKCη) or PKCη with radixin (PKCη/Rdx), and its tryptic phosphopeptide profile obtained. Phosphopeptides found specifically in the presence of Rdx are indicated with arrows (presence) or dotted circles (absence).

Fig 4. Detection of PDK1phosphoS135 in human cancer-derived cell lines.

(A) Whole-cell extracts of the indicated human tumor-derived cell lines were analyzed by western blotting. Top: PDK1phosphoS135 (human)/phosphoS138 (mouse) was detected with immunoaffinity-purified phosphospecific antisera from three individual immunizations (#39, #40, and #769). Mock- and MVM-infected A9 cells were used as controls of antiserum specificity. Bottom: The same extracts were analyzed to determine their PDK1 (PDK1), PDK1phospho308, and total PKB, PKCη, and Rdx contents. α-Tubulin was used as loading control. a, full-length PKCη; b, PKCη proteolytic cleavage product PKMη. The significant level of PKMη seen in some of the examined cancer cell lines could be indicative of upregulated kinase activity, as this fragment can result from cleavage of the regulatory domain and/or increased the turnover of activated protein [40]. (B) Impact of PKCη and Rdx on cell metabolic activity and survival. The indicated cell lines were transduced with a rAAV (10⁴ viral genomes/cell) expressing a dominant-negative (dn) or constitutively active (ca) form of the indicated signaling protein under control of the PV P4 promoter. 72 h post transduction, the cells were labeled for 30 min with Mitotracker and mitochondrial activity was measured by confocal laser scanning microscopy, quantified with Image J software as relative light intensity per cell, and expressed as a percentage of the value obtained for mock-treated cells (light gray columns). In parallel, proportions of dead cells (dark gray columns) and apoptotic cells (hatched columns) were measured, respectively, by PI (necrosis) and DAPI staining (detection of apoptotic bodies). The data presented are means with standard-deviation bars of three individual experiments, each involving > 200 cells per sample. dnPDK, PDK1K204M; dnPKCη, PKCηT512A; caPKCη, PKCηA160E; dnRdxA, RdxT564A; caRdxE, RdxT564E; dnRdxP, Rdxdl[P]; For comparison, viability was also measured 24 h after infection of A9 cells with MVM and of NCH149 and MRC-5 cells with H-1PV. Treatments that significantly (p<0,01) impaired cell metabolic activity and/or viability are indicated in black and marked by astericks. rAAV-mediated transduction efficiencies were checked by confocal microscopy (S7 Fig.).

Fig 5. Impact of caPDK1 on the growth factor dependence of cell metabolic activity and survival.

Each indicated cell line was transduced with a rAAV (10⁴ rAAV genomes/cell) expressing mutant PDK1 under the control of the PV P4 promoter. 72 h post transduction, the cells were treated (or not) for 4 h with 0.5 μM wortmannin prior to labeling for 30 min with Mitotracker. Mitochondrial activity and cell death were measured as described in the legend of Fig. 4. The constitutively active mutant PDK1:S138E (in A9 cells PDK1:S138E and to a lesser extent PDK1:S237D) significantly (p<0,01) reconstituted metabolic activity and prevented cells from undergoing death through necrosis (indicated by astericks). Thus, PDK1:S138E (and at least in part in A9 PDK1:S237) appeared to render cell viability independent of growth factor signaling via the PI3 kinase (marked black). It should be noted that PDK1:S138E mimics PDK1phophoS135 modification detected in non-transduced NCH149 (Fig. 4A). Transduction efficiencies were checked by confocal microscopy (S7 Fig.).

Fig 6. Impact of activated PDK1 on cell permissiveness for PV infection.

(A) Normal human fibroblasts (MRC-5 and BJ-1) and human glioma cells (NCH149) were infected (or not) with H-1PV (30 pfu/cell) and harvested at the indicated times p.i. Activated PDK1 was detected by western blotting as PDK1phosphoS135 (PDK:P135) and as autophosphorylated PDK1phosphoS241 (PDK:P241). Total PDK1, viral NS1, PKCη, and Rdx were quantified in parallel. Actin was used as loading control. (B) MRC-5 and BJ-1 cultures grown on spot slides were transfected (or not) with rAAV:PDK1S138E (caPDK1) and infected 24 h thereafter with H-1PV. The proportion of cells having initiated virus replication was determined by measuring NS1 expression by immunofluorescence staining 24 h p.i. (C, D) Normal human cells were transfected (or not) with rAAV (10⁴ rAAV genomes/cell) expressing the indicated PDK1, PKCη, or Rdx variant under the control of the PV P4 promoter and infected with (30 pfu/cell) H-1PV 24 h after transduction. Accumulation of viral DNA in the transfected MRC-5 cells was determined by dot blot hybridization (C, upper panel) and expressed in light intensity units (C, lower panel). (D) The proportion of dead cells was measured by PI staining 48 h p.i. Data are presented as means with standard deviation bars of three independent experiments and significance of the changes compared to mock treated cells were determined by student’s test at p-values p<0,01 (*) and p<0,02 (**), respectively.

Fig 7. Detection of PDK1phosphoS135 in human cancer tissues.

(A) Cryosections of human brain tumors were analyzed for the presence of PDK1phosphoS135 by immunostaining with specific monoclonal antibodies (PDKp135) and counterstaining with DAPI. Normal human astrocytes, muscle tissue samples and a safety margin of healthy looking brain tissue of tumor #56 were used as negative controls while two glioma-derived cell lines (NCH149, GBM21) served as positive controls. Scale bar, 50 μm. (B) Quantitation of the IF microscopy data (see S6 Fig.) shows that 70% of the analyzed tumor samples (n = 36) contained PDK1phosphoS135-positive cells, 50% with a strong and 20% with a weak signal. (C) The presence of PDK1phosphoS135 in tumor T43 was confirmed by western blot analysis of whole-cell extracts with immunoaffinity-purified phosphospecific antiserum (Rabbit #769). Normal muscle tissue and Brain^Rg4 were analyzed in parallel for comparison. Total amounts of PDK1, PKCη/Rdx, and actin were also determined.

Fig 8. Interdependence of PV propagation and activation of the PDK/PKC/PKB signaling cascade.

Schematic representation of parvoviral interaction with the PDK1/PKC/PKB signaling cascade in permissive host and human cancer cells (left panel) versus non-permissive normal human cells (right panel). In permissive cells (e.g. tumor cells) PDK1-S135 phosphorylation stimulates parvovirus propagation and prolongs survival independently of growth factor signaling. In non-permissive normal human cells, due the failure of PVs to induce PDK1 stimulation, NS1 remains inactive.