Colorectal cancer (CRC) occurs when tumor suppressor genes (TSGs) are mutated. Identifying these genes in patients is important because CRC is a heterogeneous disease with much genetic and phenotypic variation, and identification of these genes may be helpful for determining therapeutic regimens. CRISPR-Cas9 screening of patient-derived primary tumor cells is not currently possible, but patient-derived tumor organoids serve as viable models for CRC. Colon organoids consist of patient-derived colonic epithelial cells (either healthy or tumor cells) which grow in a 3D matrix when treated with several exogenous factors including pharmacological inhibitors of transforming growth factor β (TGF-β). The TGF-β pathway is one of the most important cellular signaling pathways involved in colorectal cancer, and organoids are killed in the presence of TGF-β in the culture matrix. Michels et al. describe in this paper how they have developed a way to use such organoids for CRISPR-Cas9 screening with the goal of identifying tumor suppressors in patient tumor-derived cells.
Michels et al. began by demonstrating that CRISPR-Cas9-mediated knockout of the TGF-β receptor 2 (TGFBR2) rescued human colon organoids from TGF-β toxicity. They also showed that in organoids stably expressing Cas9, lentiviral expression of guide RNA (gRNA) targeting TGFBR2 at a multiplicity of infection (MOI) of less than 1 in the presence of TGF-β led to clonal enrichment of organoids expressing fluorescent reporters for gRNAs. Next, the researchers developed a training library of 6 known, important genes in the TGF-β pathway (positive controls) along with 94 “neutral control” genes and noncoding sequences. There were 20 gRNA sequences per gene in this library. Lentiviral treatment of organoids with this library caused enrichment of only a few of the positive controls, not all of them, demonstrating that organoid behavior in a CRISPR screen is not identical to that of cultured immortalized cells. Several algorithms used to predict gRNAs having high on-target scores for HepG2 cells did not result in high on-target rates in organoids. However, the researchers found that prescreening in HepG2 cells helped by identifying gRNA sequences with high on-target rates which could be used in organoids, decreasing gRNA library sizes needed for organoids.
Next, the researchers developed a model system in which they transplanted organoids into mice. These organoids, known as AK organoids (because they have loss of the APC gene and a mutation of the KRAS gene), would not grow into tumors unless knockout of TGFBR2 occurred, in which case they were called AKT organoids. The organoids were treated with a lentiviral gRNA library (2600 gRNAs) and then were injected into only a few (8–10) mice. Tumors grew, and genomic DNA was analyzed for known cancer drivers. This positive screen showed enrichment of TGFBR2 and other expected genes but also some false positive results from known neutral or essential genes, indicating that genetic neutral drift occurred during the time it took for tumors to grow.
To deal with the genetic neutral drift issue, the researchers developed a new approach to generating a gRNA library. Using IDT gBlocks gene fragments, containing internal 10 bp degenerated sequences, they constructed a plasmid vector which they then combined with IDT oPools oligo pools to couple a unique molecular identifier (UMI) to each gRNA sequence.
Lentiviruses were produced containing the UMI validation libraries. The lentiviruses were used to infect organoids which were then transplanted into mice, and tumors grew. By determining the UMI barcode incidence instead of only the gRNA sequence abundance, Michels et al. found that they could remove outlier clones from their analysis. This enabled them to eliminate false positives caused by clonal drift during tumor growth. Using the UMI library, they were also able to determine that the genes STK11 and TGFBR2 more frequently confer growth advantages to organoids than does the gene TP53, an unexpected finding.
Michels et al. concluded that using the UMI approach for clonal tracing allowed determination of the strengths of different genes to cause phenotypic changes in colon organoids that developed into tumors. This approach, unlike standard CRISPR-Cas9 screening, prevented partial confounding of data (false positives) resulting from genetic drift during long periods of tumor growth. This UMI positive screening approach was successful and affordable (requiring only a small number of mice). Thus, this result shows the feasibility of performing CRISPR-Cas9 screening in human colon organoids. Michels et al. stated that future applications of this approach could include screening patient-derived tumor organoids to determine specific genes of interest in their cancers, although they also stated that future research needs to include optimization of design algorithms for gRNA that would have high on-target efficiency specifically in organoids as opposed to cultured cells.