A functional yeast survival screen of tumor-derived cDNA libraries designed to identify anti-apoptotic mammalian oncogenes

Abstract

Yeast cells can be killed upon expression of pro-apoptotic mammalian proteins. We have established a functional yeast survival screen that was used to isolate novel human anti-apoptotic genes overexpressed in treatment-resistant tumors. The screening of three different cDNA libraries prepared from metastatic melanoma, glioblastomas and leukemic blasts allowed for the identification of many yeast cell death-repressing cDNAs, including 28% of genes that are already known to inhibit apoptosis, 35% of genes upregulated in at least one tumor entity and 16% of genes described as both anti-apoptotic in function and upregulated in tumors. These results confirm the great potential of this screening tool to identify novel anti-apoptotic and tumor-relevant molecules. Three of the isolated candidate genes were further analyzed regarding their anti-apoptotic function in cell culture and their potential as a therapeutic target for molecular therapy. PAICS, an enzyme required for de novo purine biosynthesis, the long non-coding RNA MALAT1 and the MAST2 kinase are overexpressed in certain tumor entities and capable of suppressing apoptosis in human cells. Using a subcutaneous xenograft mouse model, we also demonstrated that glioblastoma tumor growth requires MAST2 expression. An additional advantage of the yeast survival screen is its universal applicability. By using various inducible pro-apoptotic killer proteins and screening the appropriate cDNA library prepared from normal or pathologic tissue of interest, the survival screen can be used to identify apoptosis inhibitors in many different systems.

Figure 1. Principle of the yeast survival screen.

A. Two S. pombe yeast strains were established with inducible expression of the pro-apoptotic proteins BAK and CED-4, following thiamine removal from the growth medium. Killer protein expression resulted in efficient yeast cell death upon plating onto thiamine-deficient yeast agar plates. Transformation of the yeast cells with a tumor-derived cDNA library led to survival of few killer protein-expressing yeast colonies from which the yeast cell death-inhibiting library cDNA insert was identified and analyzed for its anti-apoptotic potential in mammalian cells and expression levels in tumor biopsies. B,C. Inducible expression of human BAK (B) and C. elegans CED-4 (C) in yeast S. pombe. Total cell lysates were prepared from the parental yeast (wildtype), DSI- and HC4-strains cultured in thiamine-containing medium or after cultivation for 20 hours in thiamine-free medium. Subsequently, Western blot analysis was performed. Human BAK was detected with a polyclonal rabbit anti-BAK antibody (Santa Cruz, Heidelberg, Germany). Anti-CED-4 antiserum (9104.1, kindly provided by A. Gartner, University of Dundee, UK) was used at a 1∶500 dilution to detect C.elegans CED-4 expression. Ponceau S staining is presented as a protein loading control.

Figure 2. Statistics of the melanoma metastasis screen.

The cDNA library prepared from a lung metastasis of a melanoma patient was transformed into the BAK (DSI) and CED-4 (HC4)-expressing yeast strains for subsequent survival screenings. Of the 8.6×10⁵ plated DSI yeast colonies, 3,000 colonies survived the killer protein expression; of these, 116 were sequenced and further analyzed. Meanwhile, from the 5.9×10⁵ plated CED-4 yeast colonies, 1,800 colonies survived the killer protein expression, and of these, 133 were sequenced and analyzed.

Figure 3. Enhanced expression of selected candidate genes in various tumor entities.

A. mRNA expression analysis using the open access cancer microarray database Oncomine. The table displays the overall number of microarray analyses present in the database that show upregulation of PAICS, MALAT1 or MAST2 in the tumor type listed in the table. The Oncomine search was performed with the thresholds as follows: p-value < 1E-4, fold change >2, and gene rank  =  top 10%. B. An immunohistochemical staining of PAICS protein expression was performed with purified anti-hPAICS serum on 25 normal skin and 39 melanoma slides. Examples of a normal skin and a melanoma result are presented; the table at the bottom summarizes the results. The average expression level of PAICS protein in normal skin was low, while PAICS expression ranged between moderate (mod) and high in a majority of the melanoma biopsies. The unpaired, two-tailed Mann-Whitney test revealed significant differences between melanoma and normal skin IHC scores (p<0.0001). C. Super-SAGE expression analysis of normal and diseased pancreatic tissues revealed significant upregulation of MALAT1 in chronic pancreatitis, pancreatic adenocarcinoma and intraductal papillary mucinous tumors compared with normal pancreas. The experiment was performed with RNA pools from 4 healthy pancreas samples, 3 chronic pancreatitis biopsies, 4 ductal adenocarcinomas and 5 intraductal papillary mucinous tumors.

Figure 4. Increased apoptosis sensitivity and decreased proliferation in PAICS, MALAT1 and MAST2 knockdown cells.

A. MelJuSo melanoma cells with stable pGIPZ shRNA-mediated PAICS knockdown (sh1-3PAICS) and control shRNA (shctr)-transduced cells were incubated for 24 hours with 0.2 µM staurosporine. Apoptosis was quantified by FACS using the Nicoletti protocol , and data are presented as the mean ± SEM, n = 4. Efficient knockdown of PAICS was confirmed in Western blot analysis using an anti-PAICS antiserum (inlet). B. Cell expansion kinetics of MelJuSo cells upon pGIPZ shRNA-mediated PAICS knockdown (sh1-3PAICS). Viable cells were quantified using a CASY cell counter, and cell numbers were compared with cells transduced with a non-targeting control shRNA (shctr). Data represent the mean values ± SEM, n = 3. One-way-ANOVA testing with Bonferroni multi-comparison correction was performed. The significance is indicated by stars for the comparison of shctr versus sh1-3PAICS (*: p-value <0.05; ***: p-value <0.001). The PAICS knockdown efficiencies were analyzed via immunoblotting (see inlet Fig. 4A ). C. Apoptosis assays using the Nicoletti FACS protocol were performed with parental A549 cells (wt), control GFP cells (ctr) and two zinc finger nuclease technology (ZFN)-mediated MALAT1 knockout cell clones (ko1 and ko2) . Results are shown for both untreated cells and cells incubated for 16 hours with either 1 µM staurosporine or 400 µM cisplatin. Data are presented as the mean ± SEM (n = 6), One-way-ANOVA testing with Bonferroni multi-comparison correction was applied, and statistical significance is indicated as **: p<0.01, ***: p<0.001. Confirmation of the MALAT1 knockout is presented in Fig. S6A. D. Two stable pGIPZ-mediated MAST2 knockdown cell lines were established from parental U87 cells (for evaluation of the knockdown efficiencies, see Fig. S6B), and following incubation of the cells with DMSO (solvent control), recombinant TRAIL (250 ng/µl), MG132 (2,5 µM), epoxomicin (50 nM), TRAIL plus MG132, and TRAIL plus epoxomicin for 16 h, apoptosis was quantified in a CASPASE-3 activity assay. Control shRNA-transduced U87 cells served as a control. Data are presented as the mean ± SEM (n = 4), One-way-analysis of variance (ANOVA) testing with Bonferroni multi-comparison correction was applied, and statistical significance is indicated as ***: p<0.001.

Figure 5. Xenograft tumor growth of MAST2 and PAICS knockdown in U87 and MelJuSo cells, respectively.

A. The subcutaneous xenograft tumor growth of stable sh2MAST2 knockdown U87 cells was quantified with a manual caliper 2-3 times per week and compared with non-targeting control shRNA (shctr)-transduced cells. Data are presented as the mean ± SEM. 5 mice/group were injected into the right flank. The unpaired, two-tailed t-test was performed individually for each day after injection of the tumor cells and revealed the following p-values: day 23, 0.0063; day 25, 0.0063; day 28, 0.0169; day 30, 0.0127; day 32, 0.0219; day 34, 0.0496. B. pTRIPZ sh2PAICS-transduced MelJuSo were injected subcutaneously into the right flank of NOD-SCID mice. When the xenografted tumors reached a size of 50 mm³, mice received drinking water containing 2 mg/ml doxycycline and 10 g/l sucrose ad libitum until the end of the experiment to induce the PAICS shRNA knockdown. The graph displays the relative tumor growth (tumor volumes from day x/day 0 of doxycycline treatment). Injected mice per group: shctr  = 10, sh2PAICS  = 11. Data are presented with the mean values and error bars indicate SEM. The unpaired, two-tailed t-test was performed individually for each day after the start of doxycycline treatment and revealed significant differences for day 4 to day 28 with doxycycline treatment (day 4, 0.0028; day 7, 0.0163; day 11, 0.0024; day 14, 0.0082; day 18, 0.0414; day 21, 0.0429; day 25, 0.0311; day 28, 0.0414). In C, the relative tumor size is shown (calculated as tumor volume of sh2PAICS MelJuSo cells divided by tumor volume of shctr cells). D. Three xenograft tumors per shRNA used (shctr, sh2PAICS) were isolated, and PAICS protein levels were analyzed via Western blot. In contrast to the MelJuSo cells grown in cell culture (see Fig. 4B ), an additional 40-kDa band appeared in the anti-PAICS Western Blot when the lysates were prepared from MelJuSo tumors grown in mice. As with the original 37-kDa PAICS band, this larger signal disappeared in the PAICS knockdown cells, thereby arguing that it represents a posttranslationally modified PAICS isoform.