Tumor organoids and cell polarity: models and mechanisms for cancer research

Division Of Basic Sciences - Nci

Investigators

Linked publications, trials & patents

Abstract

Goal 1: Identify cell polarity proteins that promote resistance to therapeutic drugs and T-cell-mediated killing and define their mechanisms of action Pancreatic cancer is among the deadliest malignancies, with a five-year survival rate of just 13%. Although gemcitabine is a first-line treatment, resistance arises rapidly. Resistance mechanisms include increased drug efflux, decreased deoxycytidine kinase (necessary for gemcitabine activation), increased cytidine deaminase (which inactivates gemcitabine), enhanced DNA repair, EMT, stemness, and tumor microenvironment changes (e.g., stroma, hypoxia), affecting cell cycle regulation. Multiple polarity proteins are overexpressed in pancreatic cancers and correlate with poor survival, suggesting that polarity proteins may be repurposed in drug resistance, similar to ER+ breast cancer. These proteins may regulate EMT or stemness, but they might also influence protein trafficking, asymmetric division, or metabolism, all of which promote resistance. Generation of polarity protein CRISPR library: We created a human polarity gene CRISPR knockout library targeting 54 polarity proteins and 7 MHC trafficking-related proteins (5 guides per gene), plus 50 non-targeting controls-355 sgRNAs total. These were cloned into a lentiviral backbone, and lentivirus was produced in HEK293T cells. The lentivirus was then concentrated, titrated, and confirmed by sequencing for integrity and equal representation of sgRNA. Dropout screen and identification of RACGAP1: Cas9-inducible Mia PaCa2-GemR cells were infected with the CRISPR library, selected with puromycin, and treated with/without doxycycline and gemcitabine for 10 days. gDNA was extracted, the CRISPR region was PCR-amplified, and the sgRNA abundance was assessed by next-generation sequencing. Genes essential for viability (lost with doxycycline alone) were excluded. Those whose loss sensitized cells to gemcitabine were identified by comparing sgRNA dropout in gemcitabine-treated cells ± doxycycline. MAGeCK and edgeR analysis generated volcano plots visualizing sgRNA dropouts. Five polarity genes showed significant dropout under gemcitabine treatment, indicating their loss re-sensitized GemR cells. Clinical dataset analysis showed only RACGAP1 strongly correlated with poor survival (hazard ratio 1.9), while FZD1, NUMA1, CDH1, and ATP1A1 lacked significance. RACGAP1 overexpression correlated with poor outcomes in multiple cancers (breast, esophageal, gastric, liver, ovarian, lung), consistent with reports linking it to poor prognosis. Mechanistically, RACGAP1 inactivates Rac GTPase and activates RhoA during integrin signaling; it also engages PI3K/AKT and Hippo/YAP pathways. How RACGAP1 promotes gemcitabine resistance remains unknown. Goal 2: Identify cell surfaceome changes associated with resistance to therapeutic drugs and understand their mechanisms Traditional proteomics struggle with cell surface proteins due to low abundance and poor solubility. New enrichment methods like cell surface biotinylation (CSB) and cell surface capture (CSC) overcome these issues by covalently tagging surface proteins using biotin. CSB utilizes a biotin-NHS ester that targets lysines, whereas CSC targets glycans, oxidizing sialic acids to form aldehydes for biotin coupling. These methods enable affinity purification and identification via mass spectrometry (MS). Using both techniques broadens the detectable surfaceome, as glycosylation patterns vary. We are using these approaches to characterize the surfaceomes of drug-sensitive and resistant ER+ breast cancer cell pairs, aiming to identify shared and drug-specific surface proteins for therapeutic targeting or biomarker development. We selected eight drug-sensitive and drug-resistant ER+ breast cancer cell line pairs with known resistance mechanisms. For instance, palbociclib resistance in MCF7 involves the inhibition of cyclin E and CDK2 activity; abemaciclib resistance also involves the inhibition of CDK2. Tamoxifen resistance is associated with the EZH2 and RAS/MAPK or PI3K pathways. CDK4/6 resistance may also involve metabolic changes, quiescence-to-senescence transitions, geroconversion, or altered antigen presentation. Unbiased surfaceome profiling could uncover new resistance mechanisms. Optimization of CSB and CSC: To improve reproducibility and scalability, we optimized CSB by refining biotin labeling timing/temperature, improving neutravidin column handling, adding chloroform-methanol washes to remove lipids, and streamlining MS prep. Protein overlap across 3 biological replicates improved from 22-55% to 82-83%. CSC labeling time and sample recovery steps were also optimized, yielding a 10-fold increase in peptides and 40% more identified proteins. Data analysis pipeline: We developed a custom pipeline using a ranking algorithm within Proteome Discoverer, integrating an R script, cell surface ranker (CS-RankR). This pipeline ranks candidate proteins based on PSMs, p-value, fold change (resistant vs. sensitive), and a surface prediction consensus (SPC) score, which is weighted twice as heavily. The SPC score is derived from four surfaceome databases compiled in SurfaceGenie and indicates the likelihood a protein is surface-localized. We also set cutoffs for consistent comparison across all datasets. Progress and preliminary analysis - MCF7 lines: Each line is being analyzed in triplicate using CSB and CSC. An interim analysis of CSB data from MCF7 pairs identified proteins that were upregulated greater than 4.0-fold in resistant cells (SPC greater than 3, PSM greater than Q1). The number of candidates varied; fulvestrant-resistant (FulvR) cells had the fewest. No candidate was shared across all pairs, but some overlapped between pairs: 2 between palbociclib-resistant (PalbR) and TamR, and 2 between AbemaR and FulvR. Notably, SLC7A5 and FGFR1 were elevated in PalbR and TamR, respectively, aligning with prior findings. To assess method complementarity, we compared CSB and CSC data from PalbR MCF7 cells. CSB identified ~4,000 proteins, CSC ~3,500, with ~50.9% overlap. Using our analysis pipeline, CSC yielded six surface candidates, while CSB yielded 22, with one shared: beta6 integrin. This confirms that combining both methods enhances coverage and uncovers non-redundant candidates.