Kinase requirements in human cells: V. Synthetic lethal interactions between p53 and the protein kinases SGK2 and PAK3

Abstract

Cervical carcinomas are initiated through a series of well-defined stages that rely on the expression of human papillomavirus (HPV) oncogenes. A panel of 100 small hairpin RNAs that target essential kinases in many tumor types was used to study the stepwise appearance of kinase requirements during cervical tumor development. Twenty-six kinases were commonly required in three cell lines derived from frank carcinomas, and each kinase requirement was traced to the specific stage in which the requirement emerged. Six kinases became required following HPV-induced immortalization, and the requirement for two kinases, SGK2 and PAK3, was mapped to the inactivation of p53 in primary human epithelial cells. Loss of the p53 tumor suppressor in other primary epithelial cells also induced dependence on SGK2 and PAK3. Hence, SGK2 and PAK3 provide important cellular functions following p53 inactivation, fulfilling the classical definition of synthetic lethality; loss of p53, SGK2, or PAK3 alone has little effect on cell viability, whereas loss of p53 together with either SGK2 or PAK3 loss leads to cell death. Whereas tumor suppressor gene mutations are not directly druggable, other proteins or pathways that become obligatory to cell viability following tumor suppressor loss provide theoretical targets for tumor suppressor-specific drug discovery efforts. The kinases SGK2 and PAK3 may thus represent such targets for p53-specific drug development.

Human papillomavirus (HPV)-associated cervical carcinogenesis is a multistep process. Cervical intraepithelial neoplasia (CIN) is an early clinical manifestation of persistent HPV infection. Progression of CINs to carcinoma in situ and invasive carcinoma is a slow and altogether inefficient process that can take decades to occur. HPV genome integration into a host cell chromosome is a frequent hallmark of malignant progression. It represents a terminal event for the viral life cycle that results in persistent, dysregulated HPV E6/E7 oncoprotein expression (1).

Cervical carcinomas have three well-studied characteristics that make them a unique and powerful system to study the early stages of cancer development. The first is that almost all cervical carcinomas are associated with HPV infections. Because of frequent HPV genome integration during malignant progression, only two viral oncoproteins, E6 and E7, are consistently expressed in cervical cancers (2). Hence, the initiating oncogenic insult, infection with a high-risk HPV, is common to almost all of these tumors. The second important feature of cervical carcinogenesis is that early stages of cervical carcinogenesis are molecularly defined and can be recapitulated in tissue culture; high-risk HPV E6/E7 expression facilitates immortalization of primary human epithelial cells, and organotypic cultures of HPV-immortalized primary human epithelial cells exhibit histopathological abnormalities similar to high-grade premalignant squamous intraepithelial lesions (3). Moreover, high-risk HPV E6/E7 expression in basal epithelial cells in conjunction with continuous low dose estrogen treatment is sufficient for cervical cancer development in transgenic mice (4).

The third feature is that high-risk HPV E6 and E7 oncoproteins target cellular signaling networks that are commonly mutated in the majority of non-HPV-associated solid human tumors. Among the important functions of high-risk HPV E7 proteins is the ability to bind and destabilize the retinoblastoma tumor suppressor protein (pRB) and other members of the pRB family (5, 6). The E6 oncoprotein associates with the ubiquitin ligase E6AP and targets the p53 tumor suppressor protein for proteasomal degradation (7). Through these and other mechanisms, high-risk HPVs subvert cellular regulatory mechanisms that inhibit inappropriate cell division. The E6/E7 oncoproteins also contribute to malignant progression through induction of genomic instability, and persistent HPV E6/E7 expression is necessary for maintenance of the transformed phenotype of cervical carcinoma cell lines (3).

In our previous studies, we have developed high-throughput screens to identify functional differences in kinase requirements among human cells. We have been particularly interested in understanding how the roles of kinases vary from one tumor cell type to another (8, 9). The long-term goal of these studies is to identify kinases that may serve as useful targets for antineoplastic drug discovery. An initial screen of ∼2,000 small hairpin (sh)RNAs targeting ∼430 kinases identified kinases essential for proliferation/survival in two well-studied human cell lines, 293T and the HPV18 positive cervical carcinoma cell line HeLa. From this proof-of-concept screen, a list of 100 shRNAs representing 86 kinases plus controls was compiled (“top 100 hits”), whose loss led to a ≥50% inhibition of proliferation/viability in either of the two cell lines (8).

Studies using this collection of shRNAs have shown that essential kinase signatures are remarkably different when comparing cell lines representing various tumor types (8, 9) and similarities are detected only in particular settings. For example, comparison of primary cells from the same tissue and of the same lineage, irrespective of the individual donor or the date of collection, yields a very similar pattern of kinase requirements (8). Comparison of cells that are identical except for the expression of a single gene, for example, an oncogene or a tumor suppressor gene, reveals distinct changes in kinase requirements (10, 11). Whereas most tumor cells, even those isolated from the same site, have different kinase requirements, we discovered a limited number of examples of tumor cells from the same site with closely related patterns of kinase requirements. In particular, the HPV18 positive adenocarcinoma cell line HeLa and the HPV16 positive squamous cell carcinoma line CaSki were among the most closely related tumor cell lines, whereas the HPV16 positive squamous cell carcinoma line SiHa showed a more distinct pattern of kinase sensitivity in these initial experiments (8).

In this study we examined how kinase requirements change during tumor development. We show that SGK2 and PAK3 become important for cell proliferation/viability as primary epithelial cells lose p53 tumor suppressor activity and remain important during tumor development. Because loss of p53 tumor suppressor activity is the most common hallmark of human tumorigenesis, the synthetic lethality with SGK2 and PAK3 loss suggests that these two proteins may be evaluated as potential chemotherapeutic targets.

Results

Identification of Kinases That Are Important for Proliferation/Survival of HPV-Positive Human Cervical Cancer Cell Lines.

On the basis of previous studies with cervical carcinoma lines (8, 9), we investigated whether there was a common set of kinases that were required for proliferation/survival of three cervical carcinoma cell lines but were dispensable for primary human foreskin keratinocytes (HFKs). Cells were infected with the appropriate lentiviral shRNA expression vectors, and cell proliferation/survival was assessed by Alamar blue staining. Alamar blue is a redox-sensitive dye that interrogates mitochondrial fitness of cells, and these assays provide a readout for cell proliferation/viability. The raw values were normalized to a scrambled control shRNA and are presented as percentage decreases in proliferation/viability. We designated kinases as “essential candidates” (i) when an shRNA inhibited proliferation/viability ≥50% on average in the three cervical cancer lines and (ii) when the shRNA scored as ≥50% more effective in suppressing proliferation/viability as compared to the average response in two populations of HFKs. From the tested set of 86 kinases plus controls, we identified 26 kinases (represented by 27 shRNAs) as essential candidates by these criteria (Fig. 1 and Table S1).

Fig. 1.

Identification of protein kinases that become important as a consequence of HPV oncoprotein expression in primary human keratinocytes. HeLa, SiHa, and CaSki cervical carcinoma (CxCa), high passage HKc/DR (HP), and low-passage HKc/HPV16 (LP) HPV16 immortalized keratinocytes and HFKs expressing the entire HPV16 early coding region (ER) or E6 and/or E7 oncoproteins were infected with lentiviral vectors expressing shRNAs to individual kinases. The percentages of the decrease in cell proliferation/survival normalized to a scrambled control shRNA and compared to HFKs as determined by Alamar blue assays are shown. The numbers represent averages of 2–4 independent experiments, each performed in quadruplicate. CxCa represent averages of the 3 cervical carcinoma lines tested. Only kinases that show average differences of ≥50% (CxCa and HPV-immortalized HFKs) or ≥40% (HPV-oncogene expressing HFK populations) are shown. Kinases that scored in CxCa as well as HKc/DR and HKc/HPV16 are highlighted in orange; those also scored in HPV oncogene expressing HFKs are highlighted in yellow.

Identification of Human Kinases That Become Important at Distinct Stages of HPV-Mediated Human Cervical Carcinogenesis.

Because HPV-associated carcinogenesis can be modeled in vitro, we next analyzed two HPV16-immortalized HFK lines that model different stages of cervical carcinogenesis. The two cell lines, HKc/HPV16 and HKc/DR, are derived from a single piece of foreskin epithelium that was transfected with a head-to-tail dimer of the HPV16 genome (12). HKc/HPV16 represent freshly immortalized cells, whereas HKc/DR have been selected for resistance to differentiation and failure to growth arrest in response to TGF-ß (13). Whereas both cell lines are nontumorigenic, mRNA expression analysis has shown that HKc/DR are more similar to cervical carcinoma cells than HKc/HPV16 cells (14). As in our experiments with cervical cancer lines, we designated kinases as “essential candidates” when their depletion yielded ≥50% difference in proliferation/survival relative to HFKs. A total of 19 kinases that met these criteria were identified for HKc/DR. Six of these, CDK7, HER3, JNK3, MELK, PAK3, and SGK2, also scored in cervical carcinoma lines. For HKc/HPV16, we identified 28 kinases that met these criteria. Ten of these, CDK7, EPHB1, HER3, JNK3, KHS1, MELK, MYO3B, PAK3, ROS, and SGK2, also scored in cervical cancer lines. Eighteen of the 19 essential candidate kinases for HKc/DR also scored in HKc/HPV16. Six of these 18 kinases, CDK7, HER3, JNK3, MELK, PAK3, and SGK2, scored in HKc/HPV16 and HKc/DR as well as the cervical carcinoma cell lines (Fig. 1 and Table S2).

Identification of Human Kinases That Become Important as a Consequence of HPV Oncogene Expression.

To identify kinases that become important as a consequence of HPV16 gene expression, we next analyzed two independent sets of donor/passage matched HFK populations that expressed the HPV16 early region or the HPV16 E6 and/or E7 oncogenes. Expression of HPV16 E7, pRB, and p53 was assessed by Western blotting. Decreases in p53 and pRB steady state levels served as a surrogate marker for HPV16 E6 or E7 expression, respectively (Fig. S1). Each of these keratinocyte populations was transduced with the collection of 100 shRNAs as above. On the basis of our previous experience with screening for kinase differences in near isogenic cell populations (10, 11), we classified kinases as essential candidates when they showed ≥40% decreased proliferation/viability relative to normal cells in each matched set. Six kinases (ADCK4, BTK, HUNK, PAK3, ROS, and SGK2) met these criteria in HPV16 early region expressing HFKs, one (SGK2) in HPV16 E6/E7 expressing HFKs, three (PAK3, SGK2, and SURTK106) in HPV16 E6 expressing HFKs, and none scored in HPV16 E7 expressing HFKs. Only PAK3 and SGK2 consistently met these criteria in HPV16 E6, early region expressing HFKs, HKc/HPV16, and HKc/DR as well as in the cervical carcinoma cell lines (Fig. 1 and Tables S1 and S2). These results demonstrate that HPV16 E6 expression in primary HFKs induces synthetic lethality upon loss of SGK2 and PAK3 expression, and this is retained in HFKs expressing the entire HPV16 early region, HPV16-immortalized HFKs, and cervical carcinoma lines.

Validation Experiments.

To establish quantitative comparisons of the SGK2 and PAK3 responses and determine whether additional shRNAs specific for each of the kinases yielded similar results, we tested four different shRNA expressing lentiviruses for each of the two kinases in titration experiments by using CaSki, SiHa, and HeLa cervical carcinoma cells and HFKs. These experiments revealed that multiple SGK2 and PAK3 specific shRNAs and at a variety of titers inhibited cell proliferation/viability in each of the cervical carcinoma lines but to a lesser extent in HFKs (Fig. 2A).

Fig. 2.

Validation of SGK2 and PAK3. (A) Multiple PAK3 and SGK2 lentiviral shRNA expression vectors inhibit proliferation/viability of CaSki, SiHa, and HeLa cervical carcinoma cells more efficiently than in primary HFK at multiple concentrations. Cell proliferation/viability was assessed by Alamar blue staining. (B) Multiple PAK3 and SGK2 lentiviral shRNA expression vectors cause decreases in PAK3 and SGK2 mRNA levels in CaSki cells. Messenger RNA levels were determined by quantitative reverse transcription PCR analysis at 30 h after infection with the indicated shRNA vectors; control denotes infection with a vector encoding scrambled shRNA. Bar graphs represent averages and standard deviations of 3 independent experiments and are normalized for GAPDH expression. (C) Multiple PAK3 and SGK2 shRNA expression vectors inhibit proliferation/viability of HPV16 E6 expressing HFKs more efficiently than matched control HFKs. Cells were stained with crystal violet and photographed.

To confirm kinase knockdown, we infected CaSki cells with multiple PAK3 and SGK2 specific shRNA expression vectors and analyzed mRNA levels by quantitative reverse transcription PCR at 30 h postinfection because the commercial antibodies do not allow consistent detection of these kinases by Western blotting. These experiments demonstrated significant knockdown of PAK3 and SGK2 with each of the corresponding shRNAs (Fig. 2B).

Last, we also assessed the ability of multiple PAK3 and SGK2 specific shRNAs to suppress proliferation/viability in HPV16 E6 expressing HFKs as compared to matched control HFKs. Multiple shRNAs that target different regions of SGK2 or PAK3 mRNA inhibited cell proliferation/survival of HPV16 E6 expressing HFKs but did not markedly affect control HFKs (Fig. 2C).

Hence, SGK2 and PAK3 are each important for cell proliferation/viability of HPV-positive cervical cancer cell lines, HPV16-immortalized HFKs, and HPV16 E6 oncogene expressing HFKs.

Synthetic Lethality Caused by SGK2 or PAK3 Depletion in HPV16 E6 Expressing Cells Is a Consequence of p53 Inactivation.

The best-known cellular target of HPV16 E6 is the p53 tumor suppressor. HPV16 E6 associates with the cellular ubiquitin ligase E6AP, and the E6/E6AP complex targets p53 for proteasomal degradation (7). To determine whether the observed sensitization of HPV16 E6 expressing HFKs to SGK2 or PAK3 depletion was because of p53 degradation, we generated HFKs expressing HPV16 E6 or the HPV16 E6 I₁₂₈T mutant that is defective for association with the E6AP ubiquitin ligase and thus p53 degradation (15). Donor/passage matched vector transduced HFKs were used as controls. SGK2 and PAK3 depletion markedly inhibited cell proliferation/survival of HPV16 E6 expressing cells, whereas HFKs expressing the HPV16 E6 I₁₂₈T mutant were less sensitive to SGK2 or PAK3 depletion (Fig. 3A).

Fig. 3.

Inhibition of cell proliferation/viability by SGK2 and PAK3 depletion is related to loss of p53 tumor suppressor activity. (A) HFK transduced with control vector (HFK-c), wild-type HPV16 E6 (HFK-16E6), or the p53 degradation defective HPV16 E6I₁₂₈T mutant (HFK-16I₁₂₈T) were infected with lentiviral vectors encoding scrambled (control 1 and 2), SGK2 specific, and PAK3 specific shRNAs, and cell proliferation/viability was assessed by Alamar blue assays. A Western blot documenting p53 degradation in HFKs expressing wild-type HPV16 E6 but not the HPV16 E6I₁₂₈T mutant is shown on the right. (B) HFKs infected with a control or p53 specific shRNA expression vector (3756) were infected with shRNA expression vectors encoding scrambled, SGK2, PAK3, or MAP3K8 specific shRNAs. Photomicrographs are shown in the left panel, a Western blot documenting p53 depletion is shown in the middle panel, and quantification of Alamar blue assays is shown in the right panel.

To directly assess the involvement of p53, we depleted p53 in HFKs by infection with a lentiviral shRNA. Codepletion of p53 and PAK3 or SGK2 resulted in a dramatic decrease in cell proliferation/viability, whereas depletion of an unrelated kinase, MAP3K8, which does not score as synthetic lethal with HPV16 E6 expression, had similar effects in control and p53-depleted HFKs (Fig. 3B).

To determine whether the observed synthetic lethality was specific to human foreskin derived keratinocytes or could be seen in epithelial cells derived from other human tissues, we depleted p53 in primary human mammary and prostate epithelial cells. Similar to what we observed in HFKs, p53 loss caused synthetic lethality with SGK2 and PAK3 depletion in mammary (Fig. 4A) and prostate epithelial cells (Fig. 4B).

Fig. 4.

Depletion of p53 causes synthetic lethality with SGK2 and PAK3 loss in primary human epithelial cells derived from multiple tissues. (A) Primary human mammary epithelial cell infected with a control or p53 specific shRNA expression vector were infected with shRNA expression vectors encoding scrambled, SGK2, or PAK3 specific shRNAs. Photomicrographs are shown in the left panel, and quantification of Alamar blue assays is shown in the right panel. (B) Primary human prostate epithelial cells infected with a control or p53 specific shRNA expression vector were infected with shRNA expression encoding scrambled, SGK2, or PAK3 specific shRNAs. Quantification of Alamar blue assays is shown.

Hence, functional inactivation of p53 induces cellular changes that cause synthetic lethality with SGK2 and PAK3 depletion in primary human epithelial cells.

Mechanisms of Synthetic Lethality.

Whereas our experiments document a block to proliferation/survival in p53-deficient cells upon depletion of PAK3 and SGK2, we wanted to gain some insight into the potential mechanism. A decrease in cell number as a consequence of kinase knockdowns may result from apoptosis, autophagy, senescence, or cell cycle block. Hence, we performed immunofluorescence experiments with antibodies for cleaved, active caspase 3, a marker of apoptosis, and LC3, a marker of autophagy, in HeLa cells with knockdown of SGK2 or PAK3. Cells were counterstained with Hoechst and phalloidin to visualize nuclei and actin microfilaments, respectively. These experiments suggest that the mechanisms of synthetic lethality in HeLa cells were different for the two kinases; SGK2 depletion caused autophagy, whereas PAK3 knockdown resulted in caspase 3 activation, suggestive of apoptosis. Moreover, PAK3 depletion caused marked disruption of actin filament staining, indicative of a collapse of the actin cytoskeleton, whereas no such changes were observed with SGK2 depletion (Fig. 5).

Fig. 5.

Depletion of SGK2 and PAK3 in HeLa cells is associated with autophagy and apoptosis, respectively. HeLa cells were infected with lentiviral vectors encoding scrambled (Left), SGK2-specific (Middle), and PAK3-specific shRNAs (Right). Cells were stained with antibodies for the autophagy marker LC3 (Top) and active caspase 3 (Bottom) and counterstained with Hoechst 33258 and phalloidin dyes to visualize nuclei and actin cytoskeletal structures, respectively.

Discussion

Synthetic lethal screens are one example of a larger group of genetic tests in which two genes can be shown to coordinately modify a particular phenotype and thus must have related functions within an organism. The terms “synthetic lethal” and “synthetic lethalities” were coined in 1946 by Dobzhansky (16). In the simplest terms, synthetic lethality is scored when either of two mutations in different genes has no effect on their own but in combination they have a lethal phenotype. Two logical premises have been proposed to explain how synthetic lethality can be achieved. In one scenario, two pathways are redundant and loss of either pathway alone has no effect on the cell phenotype. However, combining the two mutations leads to a lethal phenotype by removing both pathways and depriving a cell of an essential function. In the second scenario, one protein acts upstream of the second, and loss of either has no effect. One mutation occurs in a positively acting step and the other in a negative one. Because the two proteins functionally balance one another, losing one will tip the balance slightly, but losing both is catastrophic. With a little imagination, one can conceive of other machinations that follow analogous rules and give similar outcomes.

Synthetic lethal screens have been in use for nearly 100 years in genetically tractable model systems but historically have not been possible in mammalian cells because of their diploid or polyploidy/aneuploid genomes. However, several reports of synthetic lethal interactions have been reported in mammalian cells in recent years. The simplest example has been the recognition that mutations in two members of the metalloprotease family have a lethal phenotype only when both are inactivated (17). Screens for synthetic lethalities have become possible when methods that inactivate a gene product not through mutagenesis at the level of DNA but either through inactivation using small molecule inhibitors (18, 19) or through down-regulation of the mRNA by RNAi (20–23) have been used. Several reviews have called for the use of RNAi screens to exploit synthetic lethality in the study of cancer (24–29). By definition, such a screen should lead to identification of highly selective therapeutic targets (30, 31). One promising strategy might be to focus on tumor suppressors or oncogenes that are frequently mutated in many different tumor types, many of which are not or only poorly druggable. Several studies have impressively validated this concept and led to identification of multiple, potentially druggable proteins that become essential in tumors as a consequence of K-Ras oncogene activation (32–34).

By using a limited shRNA screen, we have identified two synthetic interactions with loss of p53 tumor suppressor activity. Whereas we did not seek synthetic lethalities directly in the original screen, it was designed to interrogate synthetic lethalities arising as a consequence of loss-of-function changes as a consequence of HPV oncoprotein expression. Therefore, the demonstration that the loss of SGK2 or PAK3 was lethal only when coupled to the loss of p53 was an ideal outcome from the screen. The synthetic interactions between p53 and SGK2 or between p53 and PAK3 have been confirmed by several criteria. p53 loss through two methods, expression of the HPV E6 protein, or p53 depletion each cooperate with SGK2 or PAK3 loss to generate cell death. The synthetic interactions between p53 and SGK2 loss or between p53 and PAK3 loss are not limited to foreskin keratinocytes but are seen in primary epithelial cells from mammary or prostate tissues. We do not yet know whether this relationship is specific for cases in which p53 loss occurs as an initiating event in tumor development, but SGK2 and PAK3 depletion also caused cell death in several human cancer cell lines, many of which contain p53 mutations (8). Further work will be needed to systematically investigate this relationship in other tumor types.

It is interesting to note that the cell fate changes following p53 and SGK2 or PAK3 loss are different. SGK2 depletion in p53 null cells may lead to reduction in cell proliferation/survival via autophagy, whereas PAK3 depletion in p53 null cells causes apoptosis. These findings suggest that SGK2 and PAK3 are not components of the same pathway but represent two independent types of sensitivities that are initiated following p53 loss. Further studies will be needed to conclusively address this issue. Because the test set of kinases used here was limited to only 86 kinases, full examination of the kinome and the entire proteome will expand the list of synthetic lethal interactions and may also cover the pathways that lead to synthetic lethality of SGK2 and PAK3 loss in p53 null cells.

Little is known about the functional roles of SGK2 and PAK3. SGK2 is the most poorly studied member of the SGK family. SGK1 and SGK3 play roles in the regulation of a series of plasma membrane events including several channel proteins, the insulin signaling network, and some transporter proteins (35, 36). Similar data are not available for SGK2, which also lacks the amino-terminal membrane-binding domain of SGK1 and SGK3, and hence its role in plasma membrane regulatory events is suspect. PAK3 has been primarily studied in neurons, where it interacts with CDC42 and RAC and is essential for dendrite formation. PAK3 has also been implicated in actin filament regulation in proliferating cells (37, 38). This role is supported by our work, where PAK3 down-regulation leads to actin filament collapse. PAK3 mutations have been detected in lung adenocarcinomas (39), and potentially more relevant to the work presented here, inhibition of the small GTPase Rac1, a key regulator of PAK kinases, was shown to cause cell death specifically in p53-deficient B- and T-cell lymphomas (40).

Because p53 pathway mutations are common in human tumors, the potential synthetic lethality with SGK2 and PAK3 has particular interest. Our results do not indicate whether the kinase activity of SGK2 or PAK3 is important for the synthetic interactions with p53 loss. However, if this were the case, small molecule inhibitors of these kinases may show differential effects on cell viability depending on the p53 status. Our results raise the tantalizing possibility that synthetic lethal interactions like those described here may provide a class of highly selective therapeutic targets for cancer drug development. An inhibitor that could kill a cancer cell by blocking the roles of proteins such as SGK2 or PAK3 in a p53-dependent manner but spare normal cells has many properties that will be of interest to the cancer research community.

Materials and Methods

Cell Culture.

Normal HFKs were obtained from neonatal foreskins and cultured as described previously (8). HPV oncogene expressing cell populations were generated by transfection of appropriate ß-actin expression plasmids (41) by using nucleofection (AMAXA). HPV16 E7 expression was assessed by Western blotting; decreased p53 expression was used as a surrogate marker for HPV16 E6 expression. The HPV16 E6I₁₂₈T mutant was obtained from Elliot Androphy (University of Massachusetts, Worcester, MA). HFKs with p53 knockdown were obtained by infection with appropriate lentiviral shRNA vectors followed by selection in 2 μg/mL puromycin. All experiments were performed with donor/passage matched cells. HKc/HPV16 and HKc/DR cells are derived by transfection of foreskin epithelial cells derived from a single donor with a head-to-tail dimer of the HPV16 genome (12). HKc/HPV16 represent freshly immortalized cells, whereas HKc/DR have been selected for resistance to differentiation and failure to arrest growth in response to TGF-ß (13). They were obtained from Lucia Pirisi/Kim Creek (University of South Carolina School of Medicine, Columbia, SC) and grown in K-SFM (Gibco). HeLa, CaSki, and SiHa cells were grown in DMEM supplemented with 1% penicillin-streptomycin and 10% calf serum. Primary human mammary and prostate epithelial cells were purchased from Clonetics/Lonza and grown in the media supplied.

Infections with shRNA Expressing Lentiviruses.

Lentiviruses expressing shRNAs were produced as previously described (42), and 2,000–3,000 cells were seeded per well in 96-well plates. Cells were infected at 24 hours after plating as previously described (10). Viability assays using Alamar blue were performed after puromycin selection at 5 days postinfection or stained with crystal violet for image acquisition as previously described (10).

Quantitative RT-PCR.

RNA was isolated by using the RNeasy 96 kit (Qiagen). Quantitative RT-PCR analysis was performed by using the QuantiTect SYBR Green RT-PCR Kit (Qiagen) on an Applied Biosystems 7300 Real Time PCR System. Primers were 5′-GCTCGACTATGTCAACG-3′ (forward) and 5′-CCAAGAGAATGTTCTCTGG-3′ (reverse) for SGK2 and 5′-CCAGATCACTCCTGAGC-3′ (forward) and 5′-CCAGATATCAACTTTCGGACC-3′ (reverse) for PAK3.

Immunofluorescence.

Three days postinfection with appropriate lentiviruses, cells were washed with PBS, fixed for 15 min with 4% paraformaldehyde in PBS, permeabilized for 15 min with 0.2% Triton X-100 in PBS, and incubated for 2 h with either LC3 rabbit polyclonal antibody (Santa Cruz) or Cleaved Caspase-3 rabbit polyclonal antibody (Cell Signaling Technology) diluted in blocking buffer consisting of 0.5% BSA in PBS. The secondary antibody was a goat anti-rabbit antibody conjugated with Alexa Fluor 488 (Molecular Probes/Invitrogen) and the final 2-h incubation step also contained rhodamine, phalloidin, and Hoechst 33258 dyes (Molecular Probes/Invitrogen). Fluorescent images were acquired with an inverted fluorescence microscope at a magnification of 200× (Zeiss).