polio vaccine and defining the effects of X-ray radiation on human cells [14], researchers
realized their value and set out to derive cell lines from lung, breast, ovarian, colon,
prostate, skin, renal, CNS, and other tissues [15]. There are now more than 1,300 cancer
cell lines [15], which are still routinely used by scientists around the world. Sequencing of
these has revealed the extent to which cell lines resemble the human tumor, aiding model
selection, and has described the mutational signatures present in human tumors as well as
their dynamics [11]. However, recent criticism has focused on the fact that the majority of
these cell lines are derived from European-descent populations, which may mean that
disease dynamics in other genetic ancestries may not be well-modeled by these cells [15].
Another in vitro model recently developed are 3D organoids [16]. Organoids are
“artificially grown masses of cells or tissue that resemble an organ” [17], and have been
found to faithfully reproduce the original tumors. Particularly, in an early effort to generate
a “living biobank” of organoids from colorectal cancer (CRC) patients, exome sequencing
found that organoids have a similar mutation bias (CpG > T transitions) to that of earlier,
large-scale CRC sequencing efforts, as well as having a similar frequency of hypermutated
tumors and maintaining common driver mutations [18]. RNA analysis of the same
organoids identified differential expression of cancer-associated genes such as PROX1
and PTCH1, as well as having similar expression profiles to other CRC tumors. Follow-up
studies on lung, esophagus, pancreas, and other tissues have shown similar results [16],
which demonstrates the power of this system to model cancer evolution. These have been
used successfully to investigate the link between infectious agents and cancer [16], and to
describe the mutational signatures in healthy stem cells that lead to malignant transformation [19]. Nevertheless, these are still not fully reproducible, showing a great deal of
variation, limiting their current applicability [16].
In vivo models have also been used to explore cancer biology in the context of a full host
organism. Particularly, patient-derived xenografts (PDXs) have emerged as a faithful
preclinical model to recapitulate tumor histology, genome, transcriptome, and heterogeneity, as well as drug response. These are models in which fresh tumor tissue is directly
transplanted into immunocompromised rats or mice either subcutaneously or orthotopically
[20]. Whole-exome sequencing of PDX models has been used to identify targetable
genomic alterations, and their transcriptomic characterization at the single-cell level has
identified subpopulations of cells that provide drug resistance in melanoma [20]. Large
repositories, such as the one in The Jackson Laboratory, which comprises 455 PDX models
from more than 30 primary sites, are being genomically and transcriptomically
characterized to maximize their utility for translational studies [21]. Even though their
utility in preclinical research has been extensively recognized, PDX model generation
requires high technical skills and can have low success rates, and can take several months
to establish [20], which represent important limitations. Additionally, as the host animals
are immunodeficient, a different model would need to be chosen for the study of tumor–
immune cell interactions, a very important contributor to tumor dynamics.
26
C. Molina-Aguilar et al.
realized their value and set out to derive cell lines from lung, breast, ovarian, colon,
prostate, skin, renal, CNS, and other tissues [15]. There are now more than 1,300 cancer
cell lines [15], which are still routinely used by scientists around the world. Sequencing of
these has revealed the extent to which cell lines resemble the human tumor, aiding model
selection, and has described the mutational signatures present in human tumors as well as
their dynamics [11]. However, recent criticism has focused on the fact that the majority of
these cell lines are derived from European-descent populations, which may mean that
disease dynamics in other genetic ancestries may not be well-modeled by these cells [15].
Another in vitro model recently developed are 3D organoids [16]. Organoids are
“artificially grown masses of cells or tissue that resemble an organ” [17], and have been
found to faithfully reproduce the original tumors. Particularly, in an early effort to generate
a “living biobank” of organoids from colorectal cancer (CRC) patients, exome sequencing
found that organoids have a similar mutation bias (CpG > T transitions) to that of earlier,
large-scale CRC sequencing efforts, as well as having a similar frequency of hypermutated
tumors and maintaining common driver mutations [18]. RNA analysis of the same
organoids identified differential expression of cancer-associated genes such as PROX1
and PTCH1, as well as having similar expression profiles to other CRC tumors. Follow-up
studies on lung, esophagus, pancreas, and other tissues have shown similar results [16],
which demonstrates the power of this system to model cancer evolution. These have been
used successfully to investigate the link between infectious agents and cancer [16], and to
describe the mutational signatures in healthy stem cells that lead to malignant transformation [19]. Nevertheless, these are still not fully reproducible, showing a great deal of
variation, limiting their current applicability [16].
In vivo models have also been used to explore cancer biology in the context of a full host
organism. Particularly, patient-derived xenografts (PDXs) have emerged as a faithful
preclinical model to recapitulate tumor histology, genome, transcriptome, and heterogeneity, as well as drug response. These are models in which fresh tumor tissue is directly
transplanted into immunocompromised rats or mice either subcutaneously or orthotopically
[20]. Whole-exome sequencing of PDX models has been used to identify targetable
genomic alterations, and their transcriptomic characterization at the single-cell level has
identified subpopulations of cells that provide drug resistance in melanoma [20]. Large
repositories, such as the one in The Jackson Laboratory, which comprises 455 PDX models
from more than 30 primary sites, are being genomically and transcriptomically
characterized to maximize their utility for translational studies [21]. Even though their
utility in preclinical research has been extensively recognized, PDX model generation
requires high technical skills and can have low success rates, and can take several months
to establish [20], which represent important limitations. Additionally, as the host animals
are immunodeficient, a different model would need to be chosen for the study of tumor–
immune cell interactions, a very important contributor to tumor dynamics.
26
C. Molina-Aguilar et al.