Table 2 Recent investigations of solid tumors using PDX models

PDX models derived from a quadruple wild-type (KRAS, NRAS, BRAF, and PIK3CA) colorectal tumor are sensitive to EGFR blockade alone. EGFR/MEK combination blockade is superior to treatment with anti-EGFR antibodies, or to delivering a MEK inhibitor when resistance to cetuximab or panitumumab has already developed.

Fujii et al. established a colorectal tumor organoid library comprised of 55 organoids derived from 52 tumors and 43 patients, including hyperplastic polyps and sessile serrated adenoma/polyps. Pathohistological subtypes as well as differentiation hierarchies of CRCs were cell-intrinsically conserved regardless of environment (i.e., in patients, in vitro, and in PDX models). PDX models established by transplantation into the kidney subcapsules of immunocompromised mice developed the size of engrafted subrenal CRCs that secrete niche factors, including p38-MAPK, TGF-β, and EGF. In contrast to the robust engraftment efficiency in subrenal capsules, the metastatic capacity of spleen-injected CRC organoids was diverse.

Fior et al. generated and treated five zebrafish-based PDX (zPDX) models of colon cancer derived from different patients, and treated zPDXs with the FOLFOX regimen over 3 days. Two PDXs responded to treatment, as indicated by increased caspase 3 cleavage. These two sensitive zPDXs corresponded to patients in whom CEA levels remained stable 6 months after surgery without relapse. Contrastingly, among the three zPDXs in which FOLFOX was not effective, the corresponding patients developed increasing CEA levels and clinical evidence of relapse. Furthermore, Fior et al. investigated the predictive effect of the EGFR inhibitor Cetuximab in combination with FOLFIRI regimen, finding that resistant zPDX models are derived from tumors with mutations in either BRAF or KRAS.

Pieces of resected CRCs were subcutaneously implanted into NOG mice to generate the PDX model. The organoid cells were then extracted from the PDX model for tissue culture, and CRC organoids were infected with GFP lentivirus, allowing highly sensitive visualization of micrometastases (Fig. 2). Notably, lung micrometastases were detected in three out of four mice examined 2.5 months after orthotopic injection of GFP-labelled PDX-derived organoids. The implantation of organoids into the rectal submucosa of NOG mice resulted in metastatic dissemination into the lungs, but not the liver, presumably through the inferior vena cava.

Because insulin growth factor 1 receptor (IGF1R) is highly expressed in both pancreatic cancer cells and stromal fibroblasts, Zhou et al. developed nanoparticles with recombinant human IGF1 conjugated to magnetic iron oxide carrying anthracycline doxorubicin (IGF1-IONP-Dox), and demonstrated an enhanced therapeutic effect compared with conventional Dox treatment in the orthotopic pancreatic ductal adenocarcinoma (PDAC) PDX model. T2-weighted magnetic resonance imaging (MRI) revealed systemic delivery of IGF1R-targeted Dox after administration of IGF1-IONP-Dox. Notably, non-specific uptake of IGF1-IONP-Dox in the spleen did not cause apoptosis, as demonstrated by lack of active caspase 3.

PDX models of PDAC enable precision therapy, as resistance to MEK inhibitors is paradoxically associated with compensatory Akt signaling activation [78]. Witkiewicz et al. used this model to demonstrate that combination therapy with trametinib (MEK inhibitor) and dasatinib (tyrosine kinase Src inhibitor) significantly suppresses PDX tumor proliferation. Furthermore, the Bcl-2 inhibitor ABT737 and checkpoint kinase inhibitor AZD7762 induces a therapeutic response in cooperation with conventional anti-cancer drugs such as docetaxel and gemcitabine.

PDX models of PDAC respond more robustly to mitochondrial complex I inhibitors (phenformin and metformin) than to other metabolic inhibitors, including a glutaminase inhibitor, a transaminase inhibitor, and an autophagy inhibitor. Amino acids and metabolites involved in glycolysis such as lactate are decreased by complex I inhibitors, while oxidized glutathione is increased. There is no correlation between phenformin response and genetic abnormalities in TP53, PTEN, SMAD4, and KRAS. Although phenformin has not been applied clinically due to the risk of lactic acidosis [80], this study suggests the clinical promise of this biguanide agent.

Five PDX models of cholangiocarcinoma exhibited identical KRAS mutations to those of the tumors from which they were derived. Indeed, the oncogenic KRAS mutation frequently occurs as a point mutation in codon 12, and these mutations result in constitutive activation of the PI3K pathway and RAS/RAF/MEK/ERK axis, which promote cancer cell survival and proliferation [81]. Cell-cycle regulatory proteins, including Chk1 and E2F1, were selectively down-regulated by BET protein inhibitor JQ1 in JQ1-sensitive PDX models. While JQ1 failed to affect c-Myc expression in JQ1-insensitive PDX models, a BET inhibitor down-regulated c-Myc in JQ1-sensitive PDX tumors, which was accompanied by down-regulation of downstream transcriptional targets such as Chk1 and E2F1.

Lee at al. developed GBM recurrent PDX models induced by temozolomide, in which the immediate peak stabilization of HIF1α in PDX cells after exposure to chemotherapy is expected to represent an essential step in the conversion of non-cancer stem cells into undifferentiated cancer stem cells (CSCs). In both GBM6 (classical, MGMT hypermethylated), and GBM43 (proneural, MGMT unmethylated) PDX models, HIF1α levels were increased in the CD133-positive CSC population both post-therapy and at disease recurrence.

Ectopic co-expression of SOX2, OLIG2, and ZEB1 transforms tumor-suppressor-deficient astrocytes into glioma-initiating cells in the absence of an upstream RTK oncogene. Among the three transcription factors, SOX2 expression is significantly upregulated in PDX GBMs. Histone H3 lysine 27 residue (H3K27) acetylation was present in more than 90% of the PDX SOX2 binding regions in the three analyzed patient GBM specimens, which indicates these regions as active cis-regulatory elements in patient GBM.

Fack et al. applied in situ metabolic profiling and LC-MS on brain sections of glioma PDX and human glioma samples with and without isocitrate dehydrogenase 1 (IDH1) mutations. Mass spectrometry imaging (MSI) and LC-MS analysis of orthotopic IDH-mutated glioma PDX models revealed IDH-specific adaptive mechanisms in metabolic pathways. Notably, cystathionine-β-synthase expression is a novel prognostic factor in the oligodendroglial glioma subtype.

Three group 3 medulloblastoma PDX models were established to evaluate the therapeutic efficacy of FTY720, an immunosuppressant that has currently been approved for treatment of multiple sclerosis [85]. FTY720 activates the tumor suppressor protein phosphatase 2A, and thus treatment of human medulloblastoma PDX cells with FTY720 arrested the cell cycle at the G1 phase and induced caspase-dependent apoptotic cell death.

In this study, the most metastatic PDX model (HCI-010) exhibited the highest percentage of basal/stem-like tumor cells, while the least metastatic model (HCI-002) had the lowest. This suggests that primary tumors contain a rare subpopulation of stem-like cells, and that the relative abundance of these cells correlates with metastatic potential. Higher-burden metastatic cells entered the cell cycle, expressing lower levels of quiescence and dormancy-associated genes and higher levels of cell-cycle-promoting genes, including CD24, CDK2, MMP1, and MYC, which have been associated with reactivation after dormancy. Dinaciclib, a CDK inhibitor that induces apoptosis in high MYC-expressing cancer cells via synthetic lethality [86, 87], significantly inhibits the metastatic potential of PDX models derived from drug-naïve patients.

More than 25 PDX models derived from triple-negative breast cancer (TNBC) tissues varied in the extent of PI3K and MAPK activation. PDXs were also heterogeneous in their sensitivity to chemotherapeutic agents; while PI3K, mTOR, and MEK inhibitors repressed growth but did not cause tumor regression, the PARP inhibitor talazoparib caused drastic regression in five of 12 PDXs. Notably, four out of five talazoparib-sensitive models did not harbor germline BRCA1/2 mutations, but several had somatic alterations in homologous repair pathways, including ATM deletion and BRCA2 alterations.

To elucidate the mechanism of 5-Aza-2'-deoxycytidine (decitabine) action on TNBC, Yu et al. investigated DNA methyltransferases (DNMTs) expression levels, which were correlated with response to decitabine in breast cancer organoid models based on PDX tumors derived from TNBC patients recruited to a prospective neoadjuvant study of anthracycline- and taxane-based chemotherapy. Organoids and PDX models expressing elevated levels of DNMT proteins were more sensitive to decitabine than those expressing low levels of DNMTs, suggesting that DNMT may be a predictive biomarker for treatment response to decitabine in TNBC. This effect was mediated by decitabine-induced ubiquitination and lysosomal degradation of all three DNMTs, which was dependent on the E3 ligase TNF receptor-associated factor 6 (TRAF6). Decitabine treatment of PDX tissues also increased TRAF6 transcription.

p21 protein–activated kinase 2 (PAK2) is activated by low C-terminal SRC kinase levels, which drives estrogen-independent tumor growth in patients with estrogen receptor (ER)–positive breast cancer resistant to endocrine therapy. Using a PDX model derived from ER(+)/PgR(+)/HER2(-) invasive ductal carcinoma, Xiao et al. confirmed that combination treatment with the PAK2 inhibitor FRAX597 and an ER antagonist synergistically suppressed breast tumor growth.

Compared with treatment-naïve TNBC–derived PDX, chemoresistant TNBC–derived PDX highly expressed Wnt10B-related molecules, including non-phosphorylated active β-catenin, Axin2, CD44, and HMGA2. ICG-001, a canonical Wnt signaling inhibitor, reduced tumor growth and lymph node metastatic burden. The combination of doxorubicin and ICG-001 efficiently repressed lung dissemination after tail vein injection of tumor cells that had dissociated from chemoresistant TNBC PDX. This synergistic effect was mediated by Bcl-2 downregulation.

PDX models derived from lung squamous cell carcinoma enabled Weeden et al. to identify FGFR1 mRNA level rather than FGFR1 amplification as a precise predictor of response to an FGFR tyrosine kinase inhibitor. Seventeen percent of PDXs evaluated (six out of 36 cases) exhibited FGFR1 amplification, and FGFR inhibition decreased cell proliferation and enhanced differentiation, which decreased tumor growth rate and only modestly increased apoptotic cell death. Treatment with the FGFR inhibitor BGJ398 blocked both AKT and ERK signaling, and combination therapy with BGJ398 and a PI3K inhibitor was beneficial only in PDXs in which FGFR inhibition did not alter PI3K signaling.

Because collection of circulating tumor cells (CTCs) enables non-invasive serial tumor sampling, Drapkin et al. established more than 30 PDX models of small cell lung cancer (SCLC) derived from not only biopsy tissues but also from CTCs. SCLC PDX models retain stable genome somatic alterations between initial PDX model generation and serial passages. Surprisingly, CTC-derived PDX models reflect the evolving chemotherapy sensitivities of the original tumor at the time of CTC collection. Etoposide resistance is positively correlated with activation of the Myc-associated gene signature, which is consistent with results obtained using the genetically engineered mouse model of SCLC [92, 93]. In comparison to other malignancies, SCLC PDX models exhibit increased clonal homogeneity and genomic stability.

In general, patients with wild type EGFR (EGFRwt) do not respond to treatment with EGFR tyrosine kinase inhibitors (TKIs), while most patients with EGFR activating mutations initially respond to EGFR TKIs, but inevitably develop secondary resistance to TKI treatment. Simultaneous inhibition of EGFR and TNF prevents development of this acquired EGFR resistance. Notably, combination treatment with EGFR TKI (erlotinib) plus thalidomide was highly effective in inhibiting tumor growth in an EGFRwt PDX model, while EGFR inhibition or thalidomide alone was ineffective.

The TUM622 cell line established from PDXs of lung squamous cell carcinoma has increased spheroid-forming capacity due to overexpression of the stem cell factor SOX2, and shows a hyperplastic to dysplastic change in its acinar phenotype, in which apical-basal cell polarity is disrupted. Wnt/β-catenin signaling and SOX2, which contribute to normal lung development, are involved in acinar morphogenesis of TUM622 cells in three-dimensional cultures. Chen et al. reported enhanced epithelial/mesenchymal plasticity of TUM622 cells after incorporating cancer-associated fibroblasts (CAFs) into a spheroid culture system, and expanding this system to model not only cancer cell ECM, but also cancer cell-CAF interactions during tumorigenesis. CAFs antagonized oncogenic SOX2 to restore the formation of luminal structures and promote invasion.

Intravital imaging of BRAF-mutant melanoma cells containing an ERK/MAPK biosensor revealed how the tumor microenvironment affects responses to BRAF inhibition by PLX4720. Indeed, BRAF-mutant melanoma cells respond to PLX4720 heterogeneously, and BRAF inhibition activates CAFs, leading to FAK-dependent melanoma survival signaling. Fibronectin-rich matrices with 3-12 kPa elastic modulus are sufficient to induce PLX4720 tolerance. The PDX model revealed that co-inhibition of BRAF and FAK abolishes ERK reactivation in tumor stroma, leading to more effective control of BRAF-mutant melanoma.

To facilitate the advancement of pre-clinical in vivo modeling, Krepler et al. established more than 450 PDX models. Half (55%) of all analyzed samples were BRAF hotspot mutants, 20% were NRAS mutants, 7% were NF1 mutants, 2% were KIT mutants, 1.4% were GNAQ/GNA11 mutants, and 18% were WT, all of which were consistent with the TCGA database. The simultaneous inhibition of MDM2 and ERK is highly effective in PDXs refractory to BRAF inhibitors.

SREBP-1 protects vemurafenib-resistant melanoma cells from lipid peroxidation. Fatostatin treatment alone of PDX MEL006 inhibits tumor growth more potently than vemurafenib. Importantly, combined vemurafenib/fatostatin co-treatment has a greater anti-tumor effect than either monotherapy regimen.

PDX models derived from metastatic melanoma were established to assess heterogeneous responses after treating tumors with karonudib, which inhibits the oxidized nucleotide-sanitizing enzyme MTH1 [98]. Comparison of the mutation profile between response groups revealed that karonudib has a cytotoxic effect in melanoma PDX models, irrespective of the mutation statuses of the most common driver genes in melanoma. Importantly, high expression of ABCB1 is a potential resistance biomarker.

The PDX model revealed that the LOX inhibitor BAPN attenuated ECM remodeling and tumor stiffness with collagen I bundles by reducing EGFR tyrosine kinase activity of cancer cells. The Ca²⁺ channel blockers phenylalkylamine verapamil and nondihydropyridine diltiazem, used for the treatment of hypertension and arrhythmia for decades, are effective in PDX models for the prevention of collective tumor cell invasion by significantly downregulating phosphorylated myosin light chain 2. Mechanistically, tumor-derived ECM stiffness and activated EGFR signaling enhances intracellular Ca²⁺ concentration mediated by the L-type Ca²⁺ channel Cav1.1 in squamous cancer cells.

Liu et al. established PDX models of ovarian carcinoma derived from floating tumor cells in the ascites of irradiated nude mice, and ascites in established PDX models were then implanted intraperitoneally into NSG mice. Fresh ascites-derived tumor cells from these PDX tumor-bearing NSG mice were transfected with lentivirus encoding firefly luciferase and mCherry for bioluminescence imaging (BLI) of PDX models. Cohorts of NSG mice with luciferized PDX tumors were treated with carboplatin, or paclitaxel, either as a monotherapy or in combination, followed by weekly BLI measurements, the results of which positively correlated with those of plasma CA125 and cell-free DNA assays. Notably, PDX models derived from platinum-refractory patients demonstrated significant resistance to carboplatin.

Kim et al. investigated the therapeutic effects of a PARP inhibitor (PARPi; olaparib), an ATR inhibitor (ATRi; AZD6738), and a CHK1 inhibitor (CHK1i; MK8776) for the treatment of BRCA-mutant high-grade serous ovarian cancer (HGSOC) cells. Although PARPi and CHKi as single agents modestly suppressed tumor growth, the addition of ATRi/CHK1i to PARPi in BRCA-mutant ovarian PDX models significantly decreased tumor volume compared with single-agent therapies. Remarkably, significant differences were observed in responsiveness to the PARPi-ATRi and PARPi-CHK1i combinations. While PARPi-CHK1i combination treatment significantly increased PDX tumor suppression over single-agent treatments, PARPi-ATRi combination treatment significantly increased the incidence of PDX tumor regression.

PDX models derived from BRCA-mutant HGSOCs pathologically exhibited solid, pseudoendometrioid, and transitional cell carcinomas, referred to as SET features. Because approximately 50% of HGSOC patients have defects in homologous recombination (i.e., loss-of-functional mutations of BRCA1/2) [102], blocking cell cycle checkpoints could potentially induce synthetic lethality in HGSOC. Indeed, PDX models of BRCA2-mutant HGSOC exhibited higher levels of phosphorylated CHK1 than BRCA-intact HGSOC. PET imaging of PARP-1 with the PARP inhibitor analogue [¹⁸F] FTT was performed in PDX models, which revealed that PARP inhibitor treatment resulted in tumor suppression but not complete tumor regression, similar to the response observed in clinical settings.

Choi et al. examined the therapeutic effect of the antifungal itraconazole in PDX models of serous adenocarcinoma and carcinosarcoma. Combination therapy with paclitaxel and itraconazole significantly inhibited tumor growth and suppressed angiogenesis. Combination therapy was more effective in decreasing CD31, VEGF receptor, Gli1-mediated Hh, and mTOR signaling than paclitaxel monotherapy.

The response of PDX models to the PARP inhibitor rucaparib largely depends on BRCA1 promoter methylation zygosity. PDX models were derived from 12 HGSOC patients, ten of whom were chemotherapy-naïve, and two of whom had received multiple lines of therapy. A heterozygous therapeutic response to rucaparib was observed among BRCA1/2-mutant HGSOC PDX models. Kondrashova et al. observed altered zygosity of BRCA1 methylation, characterized by homozygosity in the chemo-naïve patient-derived samples and heterozygosity in the previously treated HGSOC patient–derived samples. Loss of homozygous BRCA1 methylation is likely responsible in part for acquired rucaparib resistance.

Lee et al. developed an optimized methodology to convert bladder tumor organoid lines into orthotopic PDX models with up to 80% efficiency using ultrasound-guided implantation of organoids between the bladder urothelium and lamina propria. Tumor evolution is common in organoids, even in the absence of drug treatment. Patient-derived bladder cancer organoids maintained heterogeneity, leading to clonal evolution. Notably, the basal organoid phenotype reversibly changed into the luminal phenotype in PDX models, likely due to cellular plasticity. PDX models retain drug responses to trametinib (MEK inhibitor) and gemcitabine.