Characterization of genetic intratumor heterogeneity in colorectal cancer and matching patient-derived spheroid cultures.

Patient-derived in vitro cultures of colorectal cancer (CRC) may help guide treatment strategies prior to patient treatment. However, most previous studies have been performed on a single biopsy per tumor. The purpose of this study was to analyze multiple spatially distinct biopsies from CRCs and see how well intratumor heterogeneity (ITH) was recapitulated in matching patient-derived spheroids. Three to five biopsies were collected from six CRC tumors. Each biopsy was split in two; one half was used for spheroid culturing, while the other half was used for DNA and RNA purification. For two patients, lymph node metastases were analyzed. Somatic mutations were called from whole exome sequencing data. Each tumor contained mutations shared across all biopsies and spheroids, including major CRC drivers such as APC, KRAS, and TP53. At the same time, all tumors exhibited ITH on both mutation and copy number level. The concordance between biopsies and spheroids ranged between 40 and 70% for coding mutations. For three patients, the biopsy and spheroid from matching areas clustered together, meaning that the spheroid resembled the area of origin more than the other areas. However, all biopsies and spheroids contained private mutations. Therefore, multiple cultures from spatially distinct sites of the tumor increase the insight into the genetic profile of the entire tumor. Molecular subtypes were called from RNA sequencing data. When based on transcripts from both cancer and noncancerous cells, the subtypes were largely independent of sampling site. In contrast, subtyping based on cancer cell transcripts alone was dependent on sample site and genetic ITH. In conclusion, all examined CRC tumors showed genetic ITH. Spheroid cultures partly reflected this ITH, and having multiple cultures from distinct tumor sites improved the representation of the genetic tumor subclones. This should be taken into account when establishing patient-derived models for drug screening.