Skip to Content
NCI Patient-Derived Models Repository (PDMR) - An NCI Precision Oncology Initiative Resource
Contact PDMR
Show menu
Search this site
Last Updated: 06/28/18

Limitations and Caveats

PDX Model-Associated Data:

  • Models are distributed only after the PDMR has performed multiple QC steps. As a general reminder, PDX fragments, sequence files, DNA, RNA, and protein vials are all generated from PDX tumors that are a mixture of human tumor and mouse stroma. The host mouse model used is NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ (NSG); the reference sequence can be found on the SOP page.
  • The provided PDX pathology, WES, RNASeq, and other model characterization information in the database are representative of that model. Recipients should perform their own analysis and characterization of the PDXs they grow from the distributed material.
  • A fragment from every PDX contributing to a distribution lot undergoes histopathology assessment to confirm diagnosis and PCR assessment for human:mouse DNA ratio. Models that have de-differentiated may also be reviewed by immunohistochemistry. Our standard IHC panel includes a human mitochondrial marker, cytokeratin, EBV, and CD45; additional markers may be used for certain cancer sub-types. Several models have stably low tumor content (20%-30%) relative to mouse stroma.
  • Identifiler (STR profile) is performed on a fragment from every PDX to confirm lineage identity.
  • Human pathogen assessment is performed using the hIMPACT panel from IDEXX. The PDMR does not distribute models that are positive for human immunodeficiency virus, Hepatitis B, or Hepatitis C.
  • Prior to distribution, all models must be proven to regrow from a viably cryopreserved fragment. For this QC step, five fragments are implanted and monitored for 300 days for regrowth since many models can take up to 200-250 days to regrow to 500-700 mm3 from a frozen sample. If the initial set of five implants fails to grow, five additional fragments from a different frozen lot are implanted and assessed. To date, very few models have failed this QC step, though some models have a higher percentage take-rate from frozen stock than others.
  • Many PDX models grow slowly, both when coming out of the freezer as well as between passages. For some models, you should expect initial implants to take as long as 200 days before tumor is of sufficient size for passage. Representative growth curves for each PDX model are available in the PDMR database in the specimen details.
  • Several models have stably low human tumor content (e.g., 20%-30%). The remaining content would be murine stroma.
  • The PDMR passages PDXs by lineage and tumor heterogeneity is observed in the models, we cannot guarantee which fragment a requestor will receive. For instance, there may be variation in differentiation level of the tumor or for a stomach adenocarcinoma in the signet ring cell content.

Model Expansion — Loss of PDX Material

  • The PDMR assesses every passage for the presence of xenograft-associated lymphoproliferative disorders (XALDs). We group the following PDX-associated issues into this classification for simplicity: human lymphoma outgrowth, murine tumor outgrowth, and graft-versus-host disease (GvHD).
  • The PDMR screens all PDX tumors that lose histological/morphological characteristics with an IHC panel including markers for human mitochondria, CD45, EBV, and cytokeratin or vimentin to confirm the presence of human tumor material (aka high-grade tumor) versus human lymphoma or murine tumor out-growth. We also perform PCR to assess the human:mouse DNA ratio on a fragment from all PDXs to verify presence or absence of human content.
  • Due to the nature of fragment passaging, EBV-driven human lymphomas occurring within a PDX model can develop, but not be identified during histopathologic assessment of one fragment. Not all fragments are created equal; some may maintain sufficient lymphoproliferative content to over-grow the PDX material.
    • Use of early-passage PDXs will carry this risk. The PDMR has observed XALD out-growth as late as passage 3 (P3) from fragment implantation and intermittently within sub-lineages of a model; these lineages are not distributed.
  • Outgrowth of a mouse tumor can also be observed during passaging, and in the PDMR’s experience is seen more often at later passages. Again, PDX tumors that lose histological/morphological characteristics are screened with an IHC panel and by PCR to confirm murine tumor out-growth has not occurred.

PDC and CAF Model-Associated Data:

  • Short tandem repeat (STR) profiles are generated for every in vitro distribution lot. STR profiles are reported in toto for a model. For example, if a model has a PDC and PDX model, all STR profile variants will be reported for that model. These profiles should be used by Recipients to validate the model at multiple points during the experimental process.
  • Even within the same histology, in vitro cultures can use different Complete Media recipes, require growth ±Matrigel-coated plates, and have different growth characteristics (adherent, suspension, clusters, single-cell). PDMR sequencing is performed on cultures grown in the recommended Complete Media + Y compound; adherent cultures on Matrigel-coated flasks and suspension cultures in uncoated flasks. Cross-comparison of RNASeq data from multiple models should take this into account prior to analysis.
  • Cancer associated fibroblasts (CAFs) have a finite lifespan in vitro. CAFs are guaranteed for experimental use for up to 3 passages when maintained on Matrigel-coated flasks in the recommended Complete Media + Y compound. Additional population doublings and subcultures are possible, but overall fitness of culture may deteriorate with subsequent passages. CAFs are not sequenced for the Repository, though they may be used as a germline surrogate for the model.
  • Patient-derived tumor cell cultures (PDCs) are tested for their ability to grow as a cell-line derived xenograft (CLX) in NSG host mice. CLXs are characterized by histopathology for human tumor/diagnosis sub-type confirmation but are not further analyzed.

Associated Sequence Files

  • While the sequencing pipeline removes mouse reads per the posted SOP, some murine contribution will likely remain. Review the SOPs for details. The host mouse model used is NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ (NSG); the reference sequence can be found on the SOP page.
  • The provided PDX pathology, WES, RNASeq, etc in the PDMR database are representative of the models. Recipients should perform their own analysis on PDXs they grow out to characterize them.
  • For a limited set of models, germline sequence has been included from either the patient’s peripheral mononuclear blood fraction or purified cancer-associated fibroblasts (CAFs). In these cases the germline sequence files are posted at the patient-level in the database and somatic calls have been made. For all other models, there has been no removal of somatic mutations from the sequences as tissue for germline sequencing was not obtained.
  • DNA, RNA, and protein vials are from PDX tumors which are a mixture of human tumor and mouse stroma; these are not pure human extractions. The host mouse model used is NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ (NSG); the reference sequence can be found on the SOP page.