Kinase requirements in human cells: II. Genetic interaction screens identify kinase requirements following HPV16 E7 expression in cancer cells

Abstract

Human papillomavirus (HPV) oncoproteins subvert cellular signaling pathways, including kinase pathways, during the carcinogenic process. To identify kinases targeted by the HPV16 E7 oncoprotein, shRNA kinase screens were performed in RKO colorectal carcinoma cell lines that differ only in their expression of HPV16 E7. Our screens identified kinases that were essential for the survival of RKO cells, but not essential for RKO cells expressing HPV16 E7. These kinases include CDK6, ERBB3, FYN, AAK1, and TSSK2. We show that, as predicted, CDK6 knockdown inhibits pRb phosphorylation and induces S-phase depletion, thereby inhibiting cell viability. Knockdown of ERBB3, FYN, AAK1, and TSSK2 induces a similar loss of cell viability through an unknown mechanism. Expression of the HPV16 E7 oncoprotein, known to bind and degrade pRb, relieves the requirement of these kinases. These studies demonstate that expression of a single oncoprotein can dramatically alter kinase sensitivity in human cells. The shRNA screens used here perform analogously to genetic interaction screens commonly used in genetically tractable organisms such as yeast, and thus represent an exciting method for unbiased identification of cellular signaling pathways targeted by cancer mutations.

Expression of an oncoprotein and loss of a tumor suppressor protein are well known changes that provide essential steps in the development of a cancer. We understand in detail many of the phenotypic changes that are induced by the activation of well studied oncogenes or the loss of tumor suppressor genes (reviewed in ref. 1). In many cases, molecular consequences of these tumorigenic events have been determined, but such studies are often limited to recording changes primarily in those pathways in which these proteins have been shown to act. It has been difficult to comprehensively determine the molecular alterations that are induced by specific oncogenic events. One potential approach to determine how cells are altered by expression of an oncoprotein or loss of a tumor suppressor protein in an unbiased, more comprehensive way, is the use of comparative RNAi screens. Such RNAi screens can be perceived as genetic interaction screens in which the action of the RNAi knockdown is analyzed in combination with the action of the cancer-promoting event. Here, we report a proof-of-principle genetic interaction screen in human cells in which the action of the human papillomavirus (HPV) 16 E7 oncoprotein is studied in conjunction with a kinase loss-of-function RNAi screen.

High-risk HPVs are causative agents of cervical carcinoma, a leading cause of death in women, worldwide. A number of other anogenital tract carcinomas, and a fraction of oropharyngeal carcinomas, are also associated with high-risk HPV infections (reviewed in (2, 3)). High-risk HPV E6 and E7 proteins are consistently expressed in cervical cancers, and their expression is necessary for induction and maintenance of the transformed state. The functional roles of the HPV E6 and E7 proteins have been studied extensively. The high-risk HPV E6 protein inactivates the tumor suppressor protein p53 by accelerating its degradation through the proteasome, whereas the HPV E7 protein inactivates the retinoblastoma tumor suppressor protein pRb and the related p107/p130 family members through a similar mechanism. In addition, high-risk HPV E6 and E7 oncoproteins have a number of additional cellular targets that contribute to their oncogenic activities (reviewed in ref. 4).

The p53 and pRb tumor suppressor pathways are commonly rendered dysfunctional in most human carcinomas, and their inactivation by high-risk HPV E6 and E7 proteins generates the functional equivalent of loss of p53 or pRb by mutation. The RKO colon carcinoma cell line used here has intact pRb and p53 tumor suppressor pathways, and previous work has shown that pRb functions are lost in the RKO cells that express HPV16 E7 (RKO E7) (5).

Although we anticipate many different nuances to emerge over time, the human cells used here behave much in the same way as yeast cells do in simple genetic screens. As reported in our preceding study (6), we found numerous differences in kinase requirements when we analyzed a wide range of human tumor cell lines. Although somewhat expected, it was interesting to note that kinase requirements were quite similar in cells from the same tissue type or in cell lines that differ only by expression of a single protein. Specifically, when the kinase requirements of RKO colorectal carcinoma cells expressing the HPV16 E7 oncoprotein were compared with the RKO parental cells, we observed that only a distinct subset of kinases were differentially required (6). These findings prompted us to characterize in greater detail the differential kinase requirements in cancer cells that differed only in expression of the HPV16 E7 oncoprotein. CDK6, ERBB3, FYN, AAK1, and TSSK2 were identified as essential kinases in RKO cells but were dispensable for viability of RKO E7 cells. The differential requirement of CDK6 in RKO versus RKO E7 cells most likely reflects the ability of HPV16 E7 to target the pRb tumor suppressor protein for degradation. ERBB3, FYN, AAK1, and TSSK2 may represent previously unidentified essential regulators of pRb activity in RKO cells, or these kinases may be rendered dispensable for RKO E7 viability as consequence of HPV16 E7 targeting other signal transduction pathways in these cells. Our results suggest a general method to study the changes in cellular signaling networks induced by a cancer-promoting event, specifically the expression of a single oncoprotein, and establishes a reliable method to do functional pathway mapping in human cells.

Results

Expression of the HPV16 E7 Oncoprotein Induces Differential Kinase Requirements in Cancer Cells.

To determine differential kinase requirements induced by expression of the HPV16 E7 oncoprotein, we performed essential kinase screens in both the RKO parental cell line and RKO cells stably expressing HPV16 E7 (RKO E7) (5). Both cell lines were infected with lentiviral vectors expressing 100 individual shRNAs targeting 80 unique kinases. These correspond to the top 100 kinase “hits” that are essential for viability of HeLa and 293T cells, as identified in our primary screen (6). Cells were subsequently assayed for viability 5 days after infection. Percentage of cell viability was determined using Alamar blue, a dye that measures mitochondrial fitness (7, 8). Values were calculated as a percentage of loss of viability, normalized to a scrambled control shRNA. Two independent screens were performed, each in quadruplicate, with a resulting correlation coefficient between screens of 0.95. Kinase signatures from both screens are displayed in a heat map; green denotes the greatest cell viability, and red denotes the least cell viability (Fig. 1). Each column represents data from a specific cell line, and each row illustrates results from individual kinase knockdown in each of the cell lines. Differential kinase requirements resulting from expression of the HPV16 E7 oncoprotein in RKO cells can be distinguished. Knockdowns of 19 unique kinases induced a 30% or greater differential in cell survival between RKO and RKO E7 cells (see Fig. 1). A complete list detailing percentage of loss of cell viability and the resulting survival differentials in RKO versus RKO E7 cells can be found in supporting information (SI) Table S1. These results demonstrate that shRNA screens can be used to determine specific differences in cellular kinase requirements induced by expression of the HPV16 E7 oncoprotein.

Fig. 1.

Expression of the HPV16 E7 oncoproteins alters the essential kinase signature in RKO colorectal carcinoma cells. RKO and RKO E7 cells were infected with each of 100 lentiviral shRNA expression vectors targeting 80 unique kinases and were subsequently assessed for cell viability using Alamar blue. Percentage of viability was determined after normalization to a scrambled shRNA control vector. Results from two independent shRNA screens, each performed in quadruplicate, are represented here in a heat map. Each column represents a cell line, each row represents an individual hairpin (hairpin names are given on the right). Green indicates the most cell viability and red indicates the least cell viability. Data are sorted using unsupervised clustering and Euclidean distance.

HPV16 E7 Expression Rescues the Lethality of Kinase Loss.

To verify and explore the differential kinase requirements identified in our initial screen, we chose to further investigate 8 of the 19 kinases exhibiting the largest survival differentials in RKO versus RKO E7 cells. In addition, three kinases exhibiting no significant survival differentials between the two cell populations were included as negative controls. First, to minimize the chance that the cell survival differentials were the result of off target effects, we tested four to five shRNAs from the original RNAi Consortium library targeting each of the selected kinases. For five of the eight kinases tested, we identified multiple shRNAs that resulted in a >30% survival differential. These kinases include CDK6, ERBB3, FYN, AAK1, and TSSK2. The remaining three kinases, TLK1, JNK3, and Cke1, did not pass this validation step and were consequently not included for further analysis. None of the available shRNAs targeting the three control kinases exhibited differential killing between RKO and RKO E7 cells (data not shown).

Second, we performed a time course of kinase knockdown using multiple shRNAs targeting each of the five kinases that passed the first validation step. Cells were assayed for survival using Alamar blue at 1, 3, 5, and 8 days after infection with the shRNA lentiviruses, and were subsequently fixed and stained with crystal violet for microscopic evaluation. The 5-day time-point was determined to be optimal for determining cell survival differentials, as Alamar blue readings exhibited maximal differences in signal without overcrowding of cells in the well. Representative photomicrographs of the cells 5 days after infection with two shRNAs targeting each of the five kinases are shown and demonstrate cell survival differentials consistent with the Alamar blue measurements (Fig. 2A). Each of the shRNAs dramatically inhibited the growth of RKO cells without markedly affecting the proliferation of RKO E7 cells.

Fig. 2.

HPV16 E7 expression abrogates the requirement of CDK6, ERBB3, FYN, AAK1, and TSSK2 for viability of RKO cells. (A) Photomicrographs of RKO and RKO E7 cells after infection with two different shRNA expression vectors targeting each kinase. Cells were fixed and stained with crystal violet 5 days after infection. (B) Viability of RKO and RKO E7 cells after infection with different concentrations of shRNA expression vectors. Cells were infected with 1.25 to 20 μl of the corresponding shRNA expression vector and assayed by Alamar blue for cell survival 5 days after infection. Percentage of loss of cell viability resulting from kinase knockdown is graphed. Values were normalized to a scrambled hairpin control. Red lines with squares indicate values from RKO E7 cells, and blue lines with circles indicate values from RKO control cells. Results with two hairpins targeting different portions of each kinase are shown. (C) Efficiency of kinase mRNA knockdown in RKO and RKO E7 cells. Quantitative RT PCR analysis was used to determine expression of each kinase 60 h after infection with the corresponding shRNA expression vectors. Values shown are normalized GAPDH levels.

Next, we used a range of viral multiplicities of infection to deliver varying doses of shRNAs to the cells. Infections were performed using dilutions of the viral supernatants, from 1.25 to 20 μl, and cells were assayed by Alamar blue for cell viability 5 days after infection. Percentage of reduction in cell viability resulting from kinase knockdown was calculated after normalization to a scrambled shRNA control. Cell viability values are graphed for each shRNA at multiple viral concentrations. Red lines with squares indicate percentage of loss of cell viability in RKO E7 cells and blue lines with circles indicate percentage of loss of cell viability in RKO control cells. As in Fig. 2A, the results with two shRNAs targeting different portions of each kinase are shown (Fig. 2B). Each of the shRNAs inhibited the proliferation of RKO control cells by >30% over that of RKO E7 cells.

As our final validation step, we assessed kinase mRNA knockdown 60 h after infection by RT-PCR, normalizing all values to GAPDH mRNA levels. As a control, we included shRNAs against ALK, a kinase that did not demonstrate a significant percentage of survival differential between RKO and RKO E7 cells. TSSK2 mRNA levels could not be adequately assessed, because we could not identify effective primer pairs. Percentage of RNA knockdown was calculated and is shown in Fig. 2C. GAPDH levels were also measured as a report of cell fitness, and the relative mRNA levels of each of the kinases in RKO and RKO E7 cells were also measured (Fig. S1). GAPDH mRNA levels did not change significantly after viral infection, arguing that the change in kinase mRNA levels do not adversely affect overall cellular viability (Fig. S1A). As for the overall endogenous mRNA levels of five kinases tested in RKO cells, ALK, CDK6, and FYN kinase levels did not vary greatly with E7 expression. However, relative endogenous mRNA levels of AAK1 and ERBB3 are approximately fivefold less with E7 expression (see Fig. S1B). Taken together, these experiments demonstrate that HPV16 E7 can complement the requirement of CDK6, ERBB3, FYN, AAK1, and TSSK2 in RKO colorectal carcinoma cells.

S-Phase Depletion Induced by Kinase Knockdown is Prevented by Expression of HPV16 E7.

Because the action of CDK6 on the cell cycle through phosphorylation of pRb has been extensively studied and is well known (reviewed in ref. 9), we sought to compare the effects of CDK6 knockdown on the cell cycle in RKO cells, which contain functional pRb, and RKO E7 cells, where pRb function is compromised because of HPV16 E7-mediated degradation (5). Given the differential requirement of CDK6 in RKO and RKO E7 cells, we postulated that upon CDK6 knockdown, RKO cells would undergo G₁/S cell cycle arrest, whereas RKO E7 cells may continue to progress through the cell cycle. To test this hypothesis, cell cycle profiles were determined upon CDK6 knockdown in each of the cell lines. We also tested shRNAs targeting ERBB3 and included shRNAs targeting JNK3, which do not differentially affect viability of RKO and RKO E7 cells, as a negative control. Cells were collected 96 h after infection for FACS analysis, and the resulting profiles are shown (Fig. 3). For RKO cells expressing the E7 oncoprotein, no change in the cell cycle profiles was detected upon CDK6 or ERBB3 knockdown (see Fig. 3 Lower). In contrast, however, we observed a striking reduction of S-phase upon CDK6 and ERBB3 knockdown in RKO cells. As expected, JNK3 knockdown did not affect the cell cycle profiles of RKO or RKO E7 cells (see Fig. 3). These results indicate that expression of the HPV16 E7 oncoprotein abrogates cell cycle perturbations induced by loss of CDK6 and ERBB3.

Fig. 3.

HPV16 E7 expression in RKO cells abrogates S-phase depletion induced by CDK6 and ERBB3 knockdown. RKO and RKO E7 cells were infected with two shRNA expression vectors targeting CDK6, ERBB3, and JNK3 or a scrambled shRNA control vector. Cells were harvested 96 h after infection, stained with propidium iodide, and DNA content was analyzed by FACS. Percentage of cells in S-phase is given. Data shown is representative of three independent experiments.

CDK6 Depletion Reduces pRb Phosphorylation.

Given the striking reduction in S-phase in response to CDK6 knockdown in RKO cells, and CDK6's known function to phosphorylate pRb (reviewed in ref. 9), we sought to determine whether CDK6 knockdown altered phospho-pRb levels in RKO cells. A phospho-specific antibody was used to analyze pRb phosphorylation at serines 807 and 811, which had previously been identified as CDK4/CDK6-preferred phosphorylation sites (10). RKO and RKO E7 cells were infected with an shRNAs targeting CDK6 and a scrambled shRNA as a control. Cells were lysed for Western analysis 96 h after infection. Samples were analyzed for total pRb, phosphoserine 807/811-pRb, tubulin, CDK6, and E7 levels using a Kodak Imager (Fig. 4A). Total pRb and phosphoserine 807/811-pRb levels were normalized to the tubulin loading control, and relative protein levels after CDK6 knockdown were calculated as a percentage of control. Phosphoserine 807/811-pRb levels (P-Rb, gray bars) as a fraction of total pRb levels (Total Rb, black bars) were calculated and are indicated (Fig. 4B). The ratios of phosphoserine 807/811-pRb were as expected: CDK6 knockdown reduces phosphoserine 807/811-pRb levels in RKO cells.

Fig. 4.

HPV16 E7 expression in RKO cells prevents reduction of phsopho pRb levels upon CDK6 knockdown. (A) Immunoblot analysis of phospho pRb (Ser 807/811) and total pRb levels in RKO colorectal carcinoma cells 72 h after infection with CDK6 specific shRNA expression vectors or a scrambled shRNA control vector. CDK6, E7 and α-tubulin levels are also shown. (B) Relative levels of total pRb (black) versus phospho-pRb (Ser 807/811) (gray) upon CDK6 knockdown were quantified using a Kodak Imager.

Discussion

Mammalian cell biology has been limited by the absence of methods to perform genetic screens that have been immensely powerful in genetically tractable model systems. Components of mammalian signal transduction pathways are commonly identified by any number of approaches, including protein purification, homologies from model systems, and reverse genetics approaches, where interesting genes are identified from cloning of disease genes. Proteins identified using these approaches can be linked to molecular machines of all sizes by identifying associated proteins and integration into signaling pathways by finding immediate upstream and downstream partners. Additionally, their roles in contributing to general cellular phenotypes can be inferred either by analyzing how they affect cell behavior when their levels or activities are altered, or by determining their subcellular localization in comparison with other proteins of interest. These methods have provided the basis of our views of modern mammalian cell biology. However, it would be a powerful addition to these tools to be able to use genetic-type screens in mammalian cells. When done to saturation, genetic screens can provide a list that approaches an unbiased and comprehensive compendium of genes that play roles in a particular process or that affect the action of other proteins. We and others have commented previously that RNAi screens, done with various panels of siRNAs or shRNAs, provide approaches that closely resemble genetic screens for mammalian tissue culture cells (11, 12). In the studies reported here, we have performed the equivalent of a genetic interaction screen and determined how the expression of the HPV16 E7 oncoprotein has changed the patterns of kinase requirements of a colon carcinoma cell line. In our preceding study (6), we report the comparison of essential kinase signatures from tumor cell lines that carry various mutations acquired by the cells during carcinogenesis. Initially, we expected that these experiments might enable us to correlate specific alterations in essential kinase signatures to defined oncogenic mutations common among these cancer lines. However, the resulting kinase signatures were highly complex, because they presumably reflect the many different cellular variations acquired during tumor development and adaptation to culture. However, when we compared cell lines that differed only by expression of a single oncoprotein, HPV16 E7, the differences in kinase requirements were fewer and relatively straightforward to discern (6).

In the work reported here, we compared an established colorectal carcinoma cell line, RKO, to a derivative that expresses the HPV16 E7 oncoprotein, RKO E7. From a total of 80 kinases that were evaluated in this study, our initial screen tentatively identified 19 kinases essential for viability of the parental RKO line but dispensable for viability of RKO E7 cells. Somewhat surprisingly, this screen did not identify any kinases that showed a synthetic lethal interaction with HPV16 E7 expression in RKO cells. These results suggest that in this cellular background, subversion of cellular signaling pathways targeted by HPV16 E7, such as the pRb pathway, might not induce new sensitive points but rather relieve existing rate-limiting steps. This would be consistent with pRb's role as an inhibitor of proliferation and pRb family member's ability to act on a number of key transcription steps, making it unlikely that any one event might be sufficient to overcome the large number of important pRb/p107/p130 regulated events.

Because the 80 kinases evaluated in this study were originally identified as essential for the viability of HeLa or 293T cells, one might expect that HeLa and RKOE7 cells would respond similarly in essential kinase screens, because they are both cancer cell lines that express high-risk HPV E7. Although these cell lines did respond similarly to a majority of the kinases tested, we did observe a subset of kinases whose essentiality was altered by HPV16 E7 expression as compared with RKO parental cells and HeLa cells. The differences in kinase requirements we have observed in RKO E7 versus HeLa cells might be because there is no E6 expression in the RKO E7 cells, and we are thus observing the effects solely a result of E7 expression. Alternatively, these differences in kinase essentiality may be caused by differences in the genetic background of the cells. It is important to note that kinases targeted by the E7 oncoprotein in RKO cells might be very different from those identified when assayed in other cancer cell lines or in primary epithelial cells.

From the initial list of 19 kinases that were tentatively identified in the primary screen, a subset of 8 kinases that showed the most dramatic survival differentials in RKO versus RKO E7 cells were chosen for more detailed follow-up. For each of these kinases, we tested multiple shRNAs targeting different regions of the respective mRNA. To qualify for more detailed studies, similar cell survival differentials had to be detected with two or more shRNAs targeting each kinase. Of the eight kinases that were tested, five kinases, CDK6, ERBB3, AAK1, TSSK2, and FYN, continued to score with multiple shRNAs, demonstrating consistent and robust responses similar to what we observed in our initial screen. Those shRNAs that inhibited the proliferation of RKO cells also showed concomitant decreases in target kinase mRNA or protein levels (see Figs. 2 A and C and 4A). It is interesting to note that the relative endogenous mRNA levels of AAK1 and ERBB3 are reduced approximately fivefold in RKO E7 cells when compared with RKO control cells. Conversely, endogenous mRNA levels of CDK6 and FYN, in addition to CDK6 protein levels, do not significantly vary between the two cell lines (see Fig. 4A and Fig. S1B). These results demonstrate that the decreased sensitivity to knockdown of these kinases is not a consequence of high-level expression in RKO E7 cells. Indeed, the observed reduced expression of AAK1 and ERBB3 mRNA in RKO E7 may reflect the fact that their expression is not essential for viability of these cells. Even though the steady state mRNA levels of AAK1 and ERBB3 are reduced in RKO E7 cells, they are present at levels sufficient to observe a further decrease of each upon specific shRNA expression (see Fig. 2C).

CDK6's role in directly phosphorylating pRb at the G1/S phase transition of the cell division cycle is well established (reviewed in ref. 9); therefore, we tested whether CDK6 knockdown in RKO cells affected cell cycle progression. As predicted, CDK6 knockdown blocked RKO cells from entering S phase with a concomitant decrease in pRb phosphorylation on serine residues 807 and 811 (see Figs. 3 and 4). These results argue that CDK6 has an obligatory function in the regulation of G1/S transition in RKO cells and that this function is abrogated in RKO E7 cells, presumably because of HPV16 E7-mediated degradation of pRb family members. The exquisite sensitivity of RKO cells to CDK6 knockdown is somewhat surprising because CDK4 is generally thought to provide redundancy for CDK6 in pRb phosphorylation. However, evidence that these kinases are not entirely functionally redundant and may have distinct and overlapping functions has been reported in previous studies (13–17). The results described here also support the notion that CDK4 must provide only limited redundancy for CDK6 loss in RKO cells.

Similar to what we observed for CDK6, ERBB3 knockdown causes S-phase depletion in RKO cells, but to a much lesser extent in RKO E7 cells (see Fig. 3). ERBB3 is a member of the EGF receptor tyrosine kinase family and is not directly involved in pRb phosphorylation. Nonetheless, expression of cyclin D1, a regulatory subunit of CDK4 and CDK6, is regulated by growth factor signaling (18), and it is therefore conceivable that ERBB3 knockdown in RKO cells may indirectly affect CDK6 activity and thus pRb phosphorylation. Alternatively, ERBB3 may also modulate other signal transduction pathways involved in the G1/S cell cycle transition that are abrogated by HPV16 E7 expression through mechanisms other than pRb inactivation. The latter possibility is supported by the finding that ERBB3 knockdown results in a dramatic increase of cells with subG1 DNA content, suggestive of apoptotic cell death, whereas CDK6 knockdown did not have the same effect.

HPV16 E7 expression in RKO cells can likewise rescue loss of viability as a consequence of FYN, AAK1, and TSSK2 knockdown. However, we do not know whether this is related to ability of HPV16 E7 to target pRb family members or whether it reflects inactivation of other cellular HPV16 E7 targets (Fig. 5).

Fig. 5.

Kinases required for survival of RKO cells that may act upstream of Rb. CDK6, ERBB3, FYN, AAK1, and TSSK2 are necessary for the survival of RKO colorectal carcinoma cells. This requirement is relieved by expression of the HPV16 E7 oncoprotein and subsequent reduction of Rb levels. CDK6 loss inhibits phosphorylation of Rb, induces S-phase loss, and decreases cell survival. Knockdown of ERBB3, FYN, AAK1, and TSSK2 results in a similar loss of cell survival through an unknown mechanism. These kinases may function through direct or indirect action on the Rb protein or alternatively through other cell cycle control pathways that are subverted by HPV16 E7 expression.

The proof-of-concept experiments that are reported here provide evidence that unbiased shRNA screens can be productively used to determine how cellular signal transduction pathways change during carcinogenesis. Moreover, we demonstrate that shRNA screens can reveal changes in kinase requirements resulting from expression of a single oncoprotein. Previous methods have been effective in identifying cancer mutations and characterizing the corresponding oncoproteins or tumor suppressor proteins. The approach exemplified here provides a practical method to identify major cellular changes that occur well beyond the immediate biochemical alterations caused by the cancer mutations themselves. Genetic interaction screens performed in closely related cellular backgrounds raise the possibility of identifying essential proteins whose requirements are induced by the action of a distantly acting oncoprotein or tumor suppressor. Such proteins would be difficult to identify through standard biochemical methods. It will be interesting to determine how these newly identified protein kinases are wired to the molecular targets of HPV16 E7 and whether they may be useful as therapeutic cancer targets.

Materials and Methods

Tissue Culture.

RKO cells stably expressing the pcDNA3 control vector or HPV16 E7 have been described in ref. 5 and were maintained in McCoy's medium supplemented with 10% calf serum, 100 units/ml penicillin, 100 μg/ml streptomycin, and 0.5 mg/ml G418 (19).

100 “Hits” DNA and Virus Production.

High quality DNA preparations for 100 “hits” were obtained using a large-scale plasmid purification kit (Qiagen). For high-throughput lentiviral production in a 96-well format, 293T packaging cells were cotransfected with shRNA-encoding replication deficient viral vectors and the necessary helper plasmids for virus production. The virus was pseudotyped with the envelope glycoprotein from vesicular stomatitis virus (VSV-g), as previously described (20–22).

Infections in 96-well and 6-well Formats.

For cell survival assays, RKO and RKO E7 cell lines were seeded into four 96-well plates each at 2,200 cells per well using in a final volume of 50 μl per well. For FACS and Western analysis, cells were seeded in 6-well dishes at 30% confluency in a final volume of 1.5 ml per well. Twenty-four hours after plating, viral supernatants were added (ranging in amounts from 1.25 to 20 μl for the 96-well format, and 150 to 250 μl for the 6-well format), in the presence of 8 μg/ml Polybrene. Plates were centrifuged at 1,178 × g for 22 min. Media was replaced 12 h after infection. For cell survival assays, 2 μg/ml puromycin was added to two of the four 96-well plates. Cells were harvested 5 days after infection from 96-well plates for Alamar blue measurements and for crystal violet image acquisition. Cells were harvested from 6-well plates 72 h after infection for Western analysis and 96 h after infection for cell cycle analysis.

Alamar Blue Assay.

Five days after infection, media was removed from each well of the 96-well plates, and 100-μl Alamar blue reagent (Biosource/Invitrogen, diluted 10-fold in supplemented McCoy's medium) was added to each well (7, 8). Plates were incubated for 2 to 4 h at 37°C then read at 595 nm on a microtiter well plate reader, Spectrafluor Plus (Tecan). Raw values were normalized to a scrambled control.

Crystal Violet Staining.

Five days after infection, media was removed from each well of the 96-well plates. Cells were washed with PBS and fixed for 20 min with 10% acetic acid and 10% methanol. Cells were washed once with PBS, incubated for 30 min with 0.4% crystal violet in 20% ethanol, and were subsequently washed with PBS twice. Phase contrast Images were acquired with an inverted microscope at a magnification of 100× (Nikon).

Cell Cycle Analysis.

Infected cells were trypsinized and centrifuged with their media, then washed with PBS. After washing, cells were resuspended in 0.5 ml PBS. Cells were fixed by adding 0.5 ml of 80% ethanol and were incubated at 4°C overnight. After fixation, cells were washed with PBS and resuspended in 200 μl of 0.5 mg/ml RNAseA in PBS. After incubation at 37°C for 30 min, 200 μl of 70-μM propidium iodide in 38-mM sodium citrate, pH 7.3, was added and incubated at 37°C for an additional 30 min. Cells were stored at 4°C overnight and were analyzed on FACS Calibur (Becton Dickinson).

Western Analysis.

Cells were scraped in RIPA buffer containing protease inhibitor mixture (Roche) and 25 mM NEM. Fifty-microliter samples were prepared in 5× SDS loading buffer, resolved on 4 to 12% Nupage bis-Tris gradient gels (Invitrogen), and transferred onto PVDF membrane (Perkin–Elmer). Antibodies: ErbB3 (1B2 from Cell Signaling), Rb (M-153 from Santa Cruz), phospho-Rb (Ser-807/811 from Cell Signaling), alpha-tubulin (Sigma), CDK6 (DCS83 from Cell Signaling), HPV 16 E7 (a mixture of 8C9 from Zymed and ED17 from Santa Cruz), antirabbit and antimouse HRP conjugates (GE Healthcare). Images were captured and quantified on a Kodak 4000R digital imager.

RT-PCR.

Lentiviral vectors expressing specific shRNAs were produced and RKO and RKO E7 cells were infected as described above in a 96-well format. RNA was isolated using the RNeasy 96 Kit (Qiagen) at 36 and 60 h time points after infection. cDNA synthesis was performed using the SuperScript III Reverse Transcriptase Kit (Invitrogen) with 5 μl of RNA. Primer-probe sets were designed using the Assay Design Center from Roche. Gene specific primers were obtained from Integrated DNA Technologies, and gene specific probes were obtained from Roche. Real-time PCR reactions were performed in duplicate in a 96-well format using the LightCycler 480 (Roche).