Patient‐derived xenograft model in colorectal cancer basic and translational research

Abstract

Colorectal cancer (CRC) is one of the most popular malignancies globally, with 930 000 deaths in 2020. The evaluation of CRC‐related pathogenesis and the discovery of potential therapeutic targets will be meaningful and helpful for improving CRC treatment. With huge efforts made in past decades, the systematic treatment regimens have been applied to improve the prognosis of CRC patients. However, the sensitivity of CRC to chemotherapy and targeted therapy is different from person to person, which is an important cause of treatment failure. The emergence of patient‐derived xenograft (PDX) models shows great potential to alleviate the straits. PDX models possess similar genetic and pathological characteristics as the features of primary tumors. Moreover, PDX has the ability to mimic the tumor microenvironment of the original tumor. Thus, the PDX model is an important tool to screen precise drugs for individualized treatment, seek predictive biomarkers for prognosis supervision, and evaluate the unknown mechanism in basic research. This paper reviews the recent advances in constructed methods and applications of the CRC PDX model, aiming to provide new knowledge for CRC basic research and therapeutics.

The establishment of PDX bio‐bank for precision medicine. Specimens from CRC patients enrolled in the study are submitted to PDX bio‐bank. Researchers can take advantages of the bio‐banks to screen drugs, uncover biomarkers, and study the basic mechanisms. The latent drugs will submit to animal tests using PDX in bio‐banks. The effective drugs and biomarkers will improve therapeutic effects.

1. BACKGROUND

Colorectal cancer (CRC), one of the most deadly cancers, is the fourth most common cause of cancer‐related deaths. ¹ Currently, surgery is the main treatment option for primary and metastatic CRC, and chemotherapy and targeted drugs are regularly used as adjuvant regimens. ² , ³ , ⁴ Despite improvements in screening and treatment, the number of individuals newly diagnosed is increasing rapidly. Moreover, metastasis, recurrence, and chemoresistance are common in CRC patients after treatment, which leads to the dismal prognosis and high mortality of CRC patients. ⁵ , ⁶ , ⁷ These issues lead to the need for more advancements in CRC treatment.

About 5% of CRCs are hereditary tumor syndromes, such as familial adenomatous polyposis (FAP) and Lynch syndrome. ⁸ Most of the CRCs are sporadic as the comprehensive effects of various factors, including somatic genetic alterations, chronic inflammation, lifestyle, and environmental stimuli. ⁹ These genetic and nongenetic risk factors work together to promote CRC development and progression. ¹⁰ , ¹¹ , ¹² , ¹³ To date, the mechanisms involved in the interactions of these factors driving CRC development and progression are not fully determined. More important, CRC is a significantly heterogeneous disease with distinct variations in molecular features, which cause personalized response to therapy. ¹⁴ , ¹⁵ Tumor microenvironment (TME), another reason for tumor heterogeneity, has been found to be closely associated with individual tumor progression. ¹⁶ TME plays functional and structural roles in tumor cell growth and proliferation. ¹⁶ , ¹⁷ TME exhibits the ability to limit the efficacy of therapeutic agents, modulate tumor metastasis, and prevent recurrence. ¹⁸ , ¹⁹ Therefore, TME is a fertile ground for the development of new treatments. ²⁰ , ²¹ TME possesses a dynamic composition, including various cell types and extracellular matrix, such as cancer‐associated fibroblasts (CAFs), regulatory T cells (Tregs), tumor‐associated macrophages, and myeloid‐derived suppressor cells, which coordinate with one another to promote tumor progression. ²²

The issues of heterogeneity suggest that personalized treatment is required for improving the prognosis of CRC patients based on the individual characteristics of cancer biology. ²³ The mutations of some crucial genes, such as KRAS, BRAF, and microsatellite, have been used as important markers for drug selection. ²⁴ Recently, strategies targeting TME have been developed and applied in the treatment of human cancers, including CRC. Daclizumab, a monoclonal antibody against the CD25 receptor, has been approved by the U.S. Food and Drug Administration and has the ability to effectively decrease circulating Tregs. ²⁵ The combination of oxaliplatin and programmed death ligand‐1 therapy to inhibit TME has been reported to be effective in CRC patients with microsatellite instability. ²⁶

2. TYPICAL CRC PRECLINICAL MODELS

Mice are frequently used in CRC research and drug screening. Currently, mouse models are divided into main two groups, primary tumor model and transplanted tumor model. The primary tumor models refer to spontaneous CRCs in mice or CRCs induced by genetic engineering or chemical reagents in mice. ²⁷ CRC transplantation models are built by transplanting human or mouse primary CRC tissues or cells into mice. ¹⁰ , ²⁸ Based on graft sources, the model can be divided into allografts and xenografts. Mouse‐derived CRC tissues or cells, such as MC38 mouse colon cancer cell line, are used in the construction of allografts. ²⁹ Generally, xenograft models are established by transplanting human CRC tissues or cells into mice. Transplanted CRC models present important advantages for screening drugs, evaluating drug effect, or predicting clinical effects of drugs. ³⁰ The primary tumor models are more suitable to evaluate the underlying mechanisms of CRC development or metastasis. The features and applications of these models are summarized in Figure 1.

FIGURE 1.

The frequently used mouse models in CRC (colorectal cancer) research. The features and applications of mouse models, including AOM (azoxymethane)/DSS (dextran sulfate sodium)–induced model, spontaneous CRC, GEMMs (genetically engineered mouse models), CDX (cell line‐derived xenograft), and PDX (patient‐derived xenograft) models, are summarized

A small number of old mice develop tumor, including CRC, under natural conditions. However, it is inconvenient to establish these models because of low success rate and more time consumption. Gene mutations using genetic engineering or chemical reagents contribute to an increase in the tumor incidence in mice to fit the needs of experiments. ³¹ , ³² Thus, these models have been widely used in CRC research. The genetically engineered mouse models (GEMMs) carrying the mutation in Apc gene have been found to suffer from spontaneously multiple intestinal neoplasia similar to patients with FAP, which were used to study the molecular mechanisms involved in the early stage of CRC. ³³ Subsequently, transgenic mice with alternative Apc mutations and/or in combination with other mutations of oncogene or tumor suppressor gene such as Ras, Pten, and Braf were established for further understanding of tumor biology. ¹¹ , ³⁴ The frequently used GEMMs are presented in Table 1. GEMMs with accurate gene mutations are helpful for evaluating the functions of specific genes as well as for evaluating the related molecular pathway of tumorigenesis. As the cost of GEMMs decreases, they are frequently applied in studying the pathogenesis of CRC. However, these models have limitations due to the great decrease in life span in mice with multiple mutations.

TABLE 1.

The development of murine intestinal GEMMs

Years	Target genes	Related disease in humans
1990	Apc min/+	Recapitulates the disease observed in FAP patients
1997	Apc 580S/580S	Colonic adenomas
2000+	Apc 580S	Invasive adenocarcinomas but not metastasis
	Apc, Kras G12V
	Apc, Pten−/−
	Apc, Pik3 CA
	Apc, TGFBR1/2/−/−
2005+	Kras, Braf mutation	Adenocarcinoma that had the capacity to metastasize

Abbreviations: FAP, familial adenomatous polyposis; GEMMs, genetically engineered mouse models.

Alternatively, the carcinogenic‐induced model was developed for CRC research. The mouse models are established by injecting azoxymethane combined with dextran sulfate sodium in a time‐ and dose‐specific manner to mimic the aberrant crypt foci‐adenoma‐carcinoma progress, which occurs in CRC patients. ³⁵ , ³⁶ , ³⁷ Compared with GEMMs, this model is much easier, more convenient, less time consuming, and inexpensive for researchers. Therefore, carcinogenic‐induced CRC model is one of the commonly used models in CRC research and drug screening. Although GEMMs and carcinogenic‐induced models are important tools in basic research and drug discovery of CRC, they still fail to remodel some critical features of clinical trials. ³⁸ These models lack genetic consistency, a diversified diet, varied lifestyle, and an equal microbiome similar to human CRCs, all of which affect the disease's onset and course. ³⁹ , ⁴⁰ Moreover, the absence of inter‐tumoral heterogeneity because of poor genetic heterogeneity in inbred mice compared to outbred humans also decreases the potential of results in translational clinical research. These differences urge us to develop new models that more exactly display the characteristics of human CRCs.

Human cancer cell lines are typical in vitro model systems commonly applied in drug discovery and basic research. ⁴¹ , ⁴² , ⁴³ , ⁴⁴ More than 100 cell lines have been established as CRC cell lines from cell line banks worldwide. Cell line–derived xenograft (CDX) models developed by implanting cancer cell lines into immunodeficient mice contribute to the discovery of underlying mechanisms and the development of cancer drugs. ¹⁰ , ⁴⁵ , ⁴⁶ The most common CRC cell lines used for establishing the CRC CDX model are presented in Table 2. The CDX model is very popular in CRC research and drug discovery globally because it consumes less time, produces high yield, and is inexpensive. However, this traditional model fails to simulate tumor genetic heterogeneity of the primary tumor because of losing original inheritance and missing relevant components of TME during in vitro passage. ⁴⁷ Moreover, the artificial culture conditions may cause alterations at genetic and epigenetic levels during serial passaging with enrichment for specific subclones. ⁴⁸ Recently, the U.S. National Cancer Institute (NCI) has deprioritized the NCI‐60 human cancer cell lines for most drug screening processes. ⁴⁹ Currently, patient‐derived xenograft (PDX) models have been established to overcome the limitations of the CDX model to improve our understanding of tumor biology and to better identify novel drugs for cancer treatment. ⁵⁰ , ⁵¹ , ⁵²

TABLE 2.

Frequently used cell lines in CRC research

Name	Source	Gene signature	Tumorigenicity in nude mice
HCT116	Colon cancer	CEA (+)	Yes
SW480	Rectum adenocarcinoma in situ	c‐myc (+), K‐ras (+), H‐ras (+), N‐ras (+), myb (+), sis (+), fos (+), TP53 mutation	Yes
SW620	Lymph node metastasis of rectum adenocarcinoma	Similar to SW480	Yes
HCT‐15	Colon cancer	CSAp (−)
SW1116	Colon cancer	c‐myc (+), K‐ras (+), H‐ras (+), myb (+), sis(+), fos (+)	Yes
LOVO	Metastatic colon cancer	c‐myc (+), K‐ras (+), H‐ras (+), N‐ras (+), myb (+), sis (+), fos (+)	Yes
HT‐29	Colon cancer	c‐myc (+), K‐ras (+), H‐ras (+), N‐ras (+), myb (+), sis (+), fos (+), TP53 mutation	Yes
LS180	Colon cancer	CEA (+)	Yes
LS174T	Colon cancer	c‐myc (+), K‐ras (+), H‐ras (+), N‐ras (+), myb (+), fos (+)	Yes
NCI‐H716	Colorectal cancer	CEA (−)	Yes
Caco‐2	Rectum cancer	Keratoprotein (+)	Yes
COLO205	Colon cancer	CEA (+)	Yes
DLD1	Colorectal cancer	c‐myc (+), K‐ras (+), H‐ras (+), N‐ras (+), myb (+), fos (+), CSAp (−)	Yes
RKO	Colon cancer	Wild‐type TP53	Yes

Abbreviations: CEA, carcinoembryonic antigen; CRC, colorectal cancer; CSAp: colon‐specific antigen.

3. DEVELOPMENT OF PDX MODELS

In an attempt to mimic the original tumor inter‐ and intra‐heterogeneity, the PDX model has been developed by directly implanting tumor tissues from patients into immunodeficient animals. ⁵³ , ⁵⁴ , ⁵⁵ In the 1950 s, the first case of tumor xenograft in mice was successfully established. Currently, the process of establishing PDXs in mice from primary or metastatic CRCs has been widely reported. ⁵³ , ⁵⁶ In the early stage, it was difficult to establish the CRC PDX model in most laboratories due to low success rate, complicated procedure, and high cost. With constant efforts, the simplified and standardized procedure for CRC PDX has been established, and PDX models have been widely applied in CRC research. ⁵⁷ , ⁵⁸ The timeline of PDX model development is shown in Figure 2. This model preserves the parental tumor construction and the existing communications between tumor cells and stromal and immune cells. ⁵⁹ , ⁶⁰ Many studies have confirmed the high consistency between PDX and the corresponding primary tumor tissues in histopathological and molecular characteristics even after several passages, indicating that PDX models keep genetic stability from the primary tumors. ⁶¹ , ⁶² , ⁶³ , ⁶⁴ The results from the consensus molecular subtype (CMS) classification of PDXs and the matched primary CRCs have indicated that no variation in CMS frequencies was found between the two groups. ⁶⁵ , ⁶⁶ The evaluation of carcinoembryonic antigen and cytokeratin also supports that PDX maintains histopathological features of the original human cancer tissues. ⁶⁷ , ⁶⁸

FIGURE 2.

The timeline of PDX (patient‐derived xenograft) model development. NOD‐SCID, nonobese diabetic/severe combined immunodeficiency; NOG, NOD.Cg‐Prkdc (scid) II2rg (tm1Sug)/JicCrl; NSG, NOD‐SCID IL2rg−/−; CIEA, Central Institute for Experimental Animals

Currently, standardized procedures have been constructed for the production, quality assurance, and application of CRC PDX (Figure 3). ⁶⁹ Generally, tumor tissues, obtained by surgery or biopsy procedures, are directly engrafted as pieces (~30 mm³) after resection or after cryopreservation. ⁷⁰ Optionally, tumor tissues can be pretreated in Matrigel (a solubilized basement membrane matrix) or combined with human fibroblasts before engraftment to improve implantation achievement. PDX models are developed either heterotopically, through subcutaneous implantation into the dorsal area of mice, or orthotopically, by direct implantation into the anatomical site of origin. For CRC, subcutaneous implantation is the most commonly used procedure to establish PDXs, because it is easy to operate, monitor, and resect with good tumor engraftment. The orthoxenografts are helpful to evaluate local invasive growth of primary tumors, study tumor–host communications in their anatomical context, and observe the site‐specific dependence on therapy. ⁷¹ , ⁷² Given that microsurgical skills and small animal imaging are required for generating and monitoring orthoxenografts, this method is more difficult as a preclinical model compared with heterotopic xenograft. The median overall success rate for constructing CRC PDX is 70%, and it often takes 2–4 months. ¹¹ , ⁷³ These parameters may vary based on sample type, tumor stage, and recipient strain. It is reported that epithelial subtypes displayed lower engraftment rate than microsatellite instability tumors. ⁶⁶ The engraftment success rate from metastatic samples is higher than that from primary tumors, indicating that the success is affected by tumor stage. ⁶⁶ , ⁷⁴ In terms of recipient strain, immune‐deficient mice such as NOD/SCID (nonobese diabetic/severe combined immunodeficiency) and NOD/SCID/IL2γ‐receptor null (NSG) are the most suitable hosts for PDX generation because of their lower immune rejection and higher engraftment rate. ⁵³ , ⁷⁵ , ⁷⁶ Alternatively, it is reported that zebrafish can be used for constructing the PDX model. ⁷⁷ , ⁷⁸ , ⁷⁹

FIGURE 3.

The establishment, storage, and application of CRC (colorectal cancer) PDX (patient‐derived xenograft). The tumor tissues derived from patients are implanted into mice with immunodeficiency. The model can be used in drug testing, basic research, and biomarker discovery

4. THE APPLICATION OF PDX IN CRC RESEARCH

PDX is a promising model for addressing clinically relevant issues such as drug screening, prognosis supervision, discovery of biomarkers, and evaluation of the involved mechanisms.

4.1. Drug discovery

It has been found that both the histology and expression of tumor‐associated markers (e.g., epithelial cell adhesion molecule, E‐cadherin, carcinoembryonic antigen, and Ki67) can be maintained during passage in nude mice. ⁸⁰ Importantly, the PDX model also displays a good response to multiple chemotherapeutic agents, consistent with the response in patients, indicating that this model can be used for drug screening. ⁶⁴ The establishment of large PDX cohorts originating from the same tumor sample helps analyze genotype‐response correlations and identify optimal drugs in patients. ⁸¹ Thus, the application of the PDX model provides more evidence for the potential drug before clinical studies start. Now, more studies employ the PDX model to uncover potential drugs for CRC treatment (Table 3).

TABLE 3.

Drug discovery in CRC research using PDX models

Drugs	Function	Reference
Interferon‐α	Enhances the antitumor effects of 5‐fluorouracil	Ref. 84
XMU‐MP‐2	Attenuates chemoresistance of CRC	Ref. 85
G6PD shRNA	Increases the antitumor effects of oxaliplatin	Ref. 86
Dasatinib	Sensitizes liver metastatic CRC to oxaliplatin	Refs. 87, 88
Anti‐TIMP‐2 antibody, U0126	Attenuates the drug resistance of CRC cells to 5‐fluorouracil	Ref. 89
AZ31	Enhances the antitumor effects of irinotecan	Ref. 90
GC1118	Exhibits therapeutic efficacy in KRAS mutation‐driven PDX	Ref. 94
TAK‐960	Displays active antiproliferative effect against CRC carrying KRAS mutation	Ref. 95
ABT‐263, NCB‐0846	Exhibits antitumor effects in KRAS/BRAF‐mutated CRC PDX	Ref. 97
Bortezomib, everolimus	Exhibits antitumor effects in KRAS‐mutated CRC	Ref. 98
RGX‐202	Suppresses CRC growth in KRAS‐mutated CRC PDX	Ref. 99
Volitinib, apatinib	Increases apoptosis in c‐Met overexpressing CRC	Ref. 101
ENMD‐2076	Exhibits a promising antitumor activity in CRC PDX	Ref. 102
SN38‐loaded nanoparticles	Enhances drug efficacy in CRC	Ref. 54
PORCN inhibitor	Suppresses growth of RNF43‐mutant cell‐derived PDX	Ref. 104
Statins	Displays potent roles for APC‐mutated CRC	Ref. 105
S63845	Overcomes intrinsic and acquired regorafenib resistance in CRC	Ref. 106

Abbreviations: CRC, colorectal cancer; G6PD, glucose‐6‐phosphate dehydrogenase; PDX, patient‐derived xenograft; TIMP‐2, tissue inhibitor of matrix metalloproteinase‐2.

Various chemotherapy administration methods have been applied for CRC treatment. The first‐line chemotherapeutic drugs in the treatment of CRC contain 5‐fluorouracil (5‐Fu), oxaliplatin, and irinotecan. ⁸² CRC often develops chemoresistance with progression, which leads to treatment failure. ⁵ , ⁸³ The discovery of new drugs to ameliorate CRC chemoresistance is helpful to improve the prognosis of CRC patients. It has been found that interferon‐α could function in enhancing the antitumor activity of 5‐Fu in CRC PDX. ⁸⁴ Pharmacological targeting protein tyrosine kinase 6 attenuates the chemoresistance of CRC cells by inhibiting Janus kinase 2/signal transducer and activator of transcription 3 pathway in the PDX model. ⁸⁵ Inhibition of glucose‐6‐phosphate dehydrogenase increases oxaliplatin‐based chemotherapy in the CRC PDX models. ⁸⁶ Dasatinib, an Src inhibitor, sensitizes liver metastatic CRC to oxaliplatin in CRC with high levels of phospho‐Src using PDX. ⁸⁷ Further study also shows that dasatinib exhibits modest antitumor effect only when combined with chemotherapy in CRC PDX. ⁸⁸ It is reported that TIMP‐2 (tissue inhibitor of matrix metalloproteinase‐2) is upregulated in 5‐Fu‐resistant CRC PDX, and inhibition of TIMP‐2 using an anti‐TIMP‐2 antibody or ERK/MAPK (extracellular regulated protein kinase/mitogen‐activated protein kinase) inhibition by U0126 suppresses TIMP‐2‐mediated 5‐Fu‐resistance. ⁸⁹ The ATM (ataxia telangiectasia‐mutated gene) inhibitor, AZ31, exhibits antitumor activity in CRC PDX resistant to irinotecan monotherapy. ⁹⁰ Alisertib, an inhibitor of aurora kinases, shows antitumor activity in combination with cetuximab or irinotecan. ⁹¹

Recently, anti‐EGFR (epidermal growth factor receptor) monoclonal antibody has been used as the first‐line targeted therapy for CRC. KRAS and NRAS mutations, occurring in 45% of CRC, cause primary resistance to anti‐EGFR therapy. ⁹² , ⁹³ Thus, it is important to screen potential drugs for KRAS or NRAS mutant CRC using the PDX platform. GC1118, an anti‐EGFR antibody, exhibits promising therapeutic efficacy against KRAS mutation‐driven PDX. ⁹⁴ TAK‐960, the inhibitor of Polo‐like kinase 1, is an active antiproliferative agent against PDX models carrying KRAS mutation. ⁹⁵ Research from a novel individualized CRC PDX system revealed that the combination of the inhibitors toward EGFR and MEK (mitogen‐activated protein kinase 1) or CDK4/6 (cyclin‐dependent kinase 4/6) pathway displays a synergistic inhibitory effect against CRC using high‐throughput drug screening. ⁹⁶ CRISPR screenings uncover a novel combination treatment targeting BCL‐XL and WNT signaling for KRAS/BRAF‐mutated CRC PDX. ⁹⁷ The combination of bortezomib and everolimus (RAD001) is efficacious against mutant KRAS CRC PDX. ⁹⁸ RGX‐202, an oral small‐molecule SLC6A8 (solute carrier family 6, member 8) transporter inhibitor, suppresses CRC growth in KRAS mutant CRC PDX. ⁹⁹ Metformin selectively inhibits metastatic CRC with the KRAS mutation in the PDX model, and retrospective clinical studies further supported the role of metformin. ¹⁰⁰

Other inhibitors also showed the potent antitumor activities using CRC PDX. c‐Met inhibitor in combination with anti‐VEGF synergistically reduces microvessel density, suppresses proliferation, and increases apoptosis in c‐Met‐overexpressing CRC PDX. ¹⁰¹ ENMD‐2076, a novel orally active small‐molecule multikinase inhibitor, exhibits a promising antitumor activity in CRC PDX. ¹⁰² A nanoparticle‐based agent targeting LRP‐1 (low‐density lipoprotein‐related protein 1) can be used to promote the efficacy of neo‐adjuvant radiotherapy using CRC PDX. ¹⁰³ SN38‐loaded nanoparticles are capable of enhancing drug efficacy in CRC. ⁵⁴ PORCN inhibitor significantly suppresses RNF43‐mutant cell‐derived PDX CRC development. ¹⁰⁴ Statins, the HMG‐CoA reductase inhibitors, play potent roles in APC‐mutated CRC. ¹⁰⁵ Mcl‐1 inhibitors overcome intrinsic and acquired regorafenib resistance in CRC. ¹⁰⁶

Advancements in high‐throughput analysis and bioinformatics contribute to cancer research, ¹⁰⁷ , ¹⁰⁸ enabling PDX to become a powerful preclinical tool for large‐scale drug discovery. ¹⁰⁹ Recent studies have suggested that PDXs faithfully recapitulate human tumor biology and predict drug response by directly comparing drug responses in patients and their corresponding xenografts in high‐throughput analysis. ⁴⁹ , ¹¹⁰ , ¹¹¹ , ¹¹² High‐throughput drug screening in CRC PDX suggested that the sequential administration of XPO1 (exportin 1) and ATR (ATM and Rad3 related) inhibitors upregulates the therapeutic outcome in TP53‐mutated CRC. ¹¹³ Bortezomib, the NF‐κB inhibitor, was selected to render BETi‐resistant cells more sensitive to BETi. ¹¹⁴ PDX is also a promising tool used in uncovering novel immunotherapy methods such as chimeric antigen receptor T cells, which are genetically modified T cells recognizing cancer cells. ¹¹⁵ , ¹¹⁶

Overall, these studies identify PDX as a useful model for conducting in vivo testing of more precise therapy for CRC. Most of these studies employed PDX as a preclinical model for drug screening in the laboratory. Clinical trials to analyze the utility of PDX models are ongoing in human cancers, including CRC (Table 4). These studies will uncover the potential effect of leveraging annotated PDX models matched with the patients in a clinical trial, and this approach will accelerate the development of new drugs and improvement of CRC therapeutics.

TABLE 4.

Clinical trials using PDX model

Number	Interventions	Conditions	Status
NCT03358628	In vivo drug testing	Osteosarcoma	Not yet recruiting
NCT04602702	Drug paneling for personalized medicine	Kidney neoplasm, renal cell carcinoma, metastatic solid tumor	Recruiting
NCT02247037	Chemotherapy	Triple negative breast cancer	Completed
NCT02720796	Drug sensitivity testing	Sarcoma	Terminated
NCT02910895	Tumor biopsy	Soft tissue sarcoma	Recruiting
NCT02752893	Biospecimen collection	Breast cancer	Completed
NCT04410302	Biospecimen collection	Bladder carcinoma, gastric carcinoma, liver and intrahepatic bile duct carcinoma	Recruiting
NCT02572778	Local biopsy in the tumor	Squamous cell carcinoma of the head and neck	Recruiting
NCT03786848	MiniPDX group	Prostate cancer	Unknown
NCT04703244	Chemotherapy or endocrine therapy for breast cancer	Breast cancer, residual	Recruiting
NCT02732860	In vivo drug testing in pPDX	Colorectal neoplasms, colorectal cancer, breast cancer, breast neoplasms, ovarian cancer, ovarian neoplasm	Recruiting
NCT03219047	Ibrutinib	Recurrent mantle cell lymphoma, refractory mantle cell lymphoma	Unknown
NCT03134456	Pembrolizumab injection	Metastatic non‐small cell lung carcinoma	Recruiting
NCT03551457	Diagnostic test: biopsy	Urogenital neoplasms	Completed
NCT02752932	Drug testing on PDX	Head and neck squamous cell carcinoma	Completed
NCT03336931	Molecular profiling and drug testing	Childhood cancer, childhood solid tumor, childhood brain tumor, childhood leukemia, refractory cancer, relapsed cancer	Recruiting
NCT01750164	Biospecimen collection	Metastatic breast cancer	Terminated
NCT05092009	Tissue and blood	Lung cancer	Recruiting

Abbreviation: PDX, patient‐derived xenograft.

4.2. Biomarker discovery

Targeted therapy is helpful to improve the survival of many patients with malignant CRC. ¹¹⁷ The PDX model is developed to detect biomarkers that can effectively predict suitable drugs for specific subgroups of CRC patients. ⁵⁸ , ⁷³ , ¹¹⁸ A cohort of CRC PDX showed that IGF2 (insulin growth factor 2) may be used as a biomarker of reduced tumor sensitivity to anti‐EGFR therapy and a determinant of response to combined IGF2 and EGFR targeting in CRC patients. ¹¹⁹ Drug screening in a cohort of 49 CRC PDX models revealed that the expression of EGFR ligands was associated with the sensitivity of CRC toward cetuximab. ¹²⁰ Moreover, the downregulation in classical EGFR‐mediated signaling and the upregulation in antiapoptotic signaling are associated with acquired cetuximab resistance. ¹²⁰ A recent study reported the dynamic alterations of biomarkers in KRAS G13D mutant CRC PDX model treated with cetuximab. ¹²¹ The expression of EGFR ligands is associated with the sensitivity of CRC to cetuximab, and the gene copy number of some oncogenes (BRAF, EGFR, and KRAS) is correlated positively with cetuximab and erlotinib sensitivity. ¹²⁰ , ¹²² The implementing systems and molecular images were established to predict the efficacy of BCL‐2 inhibition in CRC based on a PDX cohort. ¹²³ PP2AC expression accurately indicates therapeutic response to p38 inhibitor in CRC PDX. ¹²⁴ APC and TP53 mutation as convenient biomarkers predicts the sensitivity to EGFR‐targeted therapies. ¹²⁵ As the key markers are continuously screened out to evaluate the related treatment in CRC PDX models, more targeted drugs are also used more precisely.

4.3. Basic research

PDX models are frequently used in basic research to study the related mechanism in CRC development and progression. Based on PDX metastasis models, researchers found that N6‐methyladenosine modification of circNSUN2 promotes colorectal liver metastasis. ¹²⁶ METTL3 (methyltransferase like 3) promotes tumor progression via an m⁶A‐IGF2BP2‐dependent mechanism in CRC. ¹²⁷ LncRNA LINRIS stabilizes IGF2BP2 (IGF2 binding protein 2) and enhances aerobic glycolysis in CRC. ¹²⁸ PCK1 (phosphoenolpyruvate carboxykinase 1) and DHODH (dihydroorotate dehydrogenase) participate in CRC liver metastatic colonization by promoting nucleotide synthesis. ¹²⁹ Inhibition of long noncoding RNA SNHG29‐mediated YAP activation promotes antitumor immunity in CRC. ¹³⁰ METTL3‐mediated m⁶A modification of LBX2‐AS1 regulates CRC proliferation, migration, and invasion acting as a ceRNA to sponge miR‐422a. ¹³¹ These studies confirmed the roles of RNA‐associated modulation in CRC development and progression using the PDX model. Vesicle transporter GOLT1B (golgi transport 1 homolog B) induces the cell membrane localization of DVL2 (disheveled segment polarity protein 2) and PD‐L2 to promote CRC metastasis. ¹³² Eukaryotic initiation factor 4A2 promotes cancer cell metastasis and oxaliplatin resistance in CRC. ¹³³ Hypoxia‐induced exosomal miR‐135a‐5p is involved in the development, clinical severity, and prognosis of CRC liver metastases through the pre‐metastatic niche. ¹³⁴ LncRNA LVBU promotes CRC progression by regulating urea cycle/polyamine synthesis axis. ¹³⁵ These papers studied the mechanism of CRC metastasis using the PDX model. Overall, PDX models are becoming more common in CRC basic research.

5. ADVANTAGES OF CRC PDX MODEL

Currently, many studies have reported that the CRC PDX model maintains the genetic consistency with the primary tumors. ¹³⁶ , ¹³⁷ , ¹³⁸ , ¹³⁹ , ¹⁴⁰ The comprehensive genetic characterization of PDXs in CRC uncovered that the frequency of key genetic mutations in PDXs is similar to that reported in primary tumors. The common genetic mutations and their frequencies, such as APC (75.9%), TP53 (70.9%), KRAS (55.7%), NRAS (5.1%), BRAF (15.2%), PIK3CA (26.6%), PIK3R1 (6.3%), and CTNNB1 (3.8%), are consistent with the data reported in human CRC tumors in a cohort of 79 CRC PDX models by whole‐exome and RNA sequencing. ¹³⁶ This consistency is important for drug screening, biomarker identification, and basic research. To some extent, the PDX model can also simulate the interactions between cancer cells and TME, which plays pivotal roles in tumor progression. ⁷³ Recent studies observed one of the molecular subtypes (CMS4), which is characterized by the activation of pathways related to EMT and stemness, such as TGF‐β and integrin, and is mostly regulated by prominent stromal cell infiltration of adjacent cancer tissue, particularly CAFs in CRC. ¹³⁷ , ¹⁴¹ , ¹⁴² The high expression of TME signatures plays an important role in inducing resistance of cancer cells to chemotherapies and targeted agents in CMS4 CRC. ¹⁴¹ , ¹⁴² , ¹⁴³ Moreover, the TGF‐β signaling inhibitors exhibit the ability to block the cross‐talk between cancer cells and the microenvironment and reduce the number of metastases in CRC PDX models, indicating a dependency on TGF‐β signaling in stromal cells during metastasis in CRC. ¹⁴³ , ¹⁴⁴ The retrospective analysis of a randomized clinical study also suggested that the CMS4 subtype exhibits a lack of benefit from adjuvant oxaliplatin‐based chemotherapy in CRC patients. ¹⁴⁵ Thus, PDX models can be used to evaluate the strategies targeting TME.

Recent evidences have demonstrated that PDX models exhibit good abilities to predict the outcomes of both traditional and novel antitumor therapeutics, revealing that PDX could be used in “co‐clinical trials” for screening drugs. ¹¹⁸ , ¹⁴⁶ After the PDX models were given the same treatment as the primary patients, they accurately replicated the patients' clinical outcomes, even as the patients underwent several additional cycles of therapy over time. ¹⁴⁷ Thus, if the investigations in in vivo tests and in clinical trials are performed in parallel, we can identify therapeutic targets more rapidly, especially for some rare types of cancers. Furthermore, the methods of “co‐clinical trials” can be used to seek more appropriate strategies for cancer patients based on the drug sensitivity of the PDX model. In this situation, PDX models can be applied as an “avatar model”. ¹⁴⁸ Because gene signature alone is insufficient to uncover therapeutic options for the majority of patients with advanced disease, the generation of the PDX platform contributes to the discovery of novel therapeutic approaches that potentially provide personalized therapeutic options for individual patients. ¹⁴⁹ , ¹⁵⁰ These studies identified the PDX system as an important tool for precise medicine because it can highly reproduce the outcomes of antitumor agents in patients.

6. LIMITATIONS OF CRC PDX MODEL

One problem in the development of the PDX model is engraftment failure. The rate of successful PDX engraftment is relatively high for CRC, ranging from 56% to 87.5% according to published studies, ⁵⁸ , ¹⁵¹ , ¹⁵² , ¹⁵³ compared to other cancers, such as epithelial ovarian cancer (48.8%) and breast cancer (27%). ¹⁵⁴ , ¹⁵⁵ The success rate of CRC PDX engraftment is higher in surgical specimens (36/50, 72%) versus in biopsy specimens (14/40, 35%). ¹⁵¹ The selection of mouse model is also important, with more severely immunosuppressed strains such as NOD/SCID or NOD/SCID/IL2gnull (NSG) being easier for PDX construction, due to their lower immune rejection to engraftment. ⁶⁰ , ⁷⁶

Although the CRC PDX platform is highly informative, some issues need to be further addressed. It is likely that PDX cannot fully represent the heterogeneity of the primary tumor. It has been shown that human tumors consist of multiple subclones carrying distinct genetic and epigenetic alterations, especially for metastatic CRCs. ¹⁵⁶ , ¹⁵⁷ , ¹⁵⁸ It is impossible to cover this intra‐tumoral heterogeneity because it is difficult to avoid the sampling bias. Moreover, the human TME is rapidly replaced in a time‐ and size‐dependent manner by murine stromal cells, including CAFs, inflammatory cells, extracellular matrix, and blood/lymphatic vessels over time. ¹⁵⁹ , ¹⁶⁰ The strategy for overcoming this problem is to develop the humanized PDX model. Briefly, the human peripheral blood cells, hematopoietic stem cells, or human tumor‐infiltrating lymphocytes are implanted into immune‐deficient mice after ablating the endogenous immune system for humanization. ¹⁶¹ , ¹⁶² In recent years, tumor immunotherapy using different approaches, including specific antibodies, immune checkpoint blockades, cancer vaccines, and adoptive cell therapies, has been widely recognized as an effective and promising method with fewer side effects. ¹⁶³ , ¹⁶⁴ , ¹⁶⁵ The PDX model lacks a complete immune system and is not suitable for immunotherapy research. Humanized mice are developed to address this issue. However, these models simulate only some aspects of the human immune system, failing to reproduce a functional and complete human immune system.

Orthotopic models provide a more suitable platform to study the biology of metastasis and response to therapy in CRC. ¹¹ , ¹⁶⁶ Orthotopic implantation into the colorectal wall of the host requires considerable technical skill. Thus, the difficulty of orthotopic transplantation also limits the capacity and reproducibility. The establishment and management of large PDX cohorts need vast labor and resource. It generally takes 1–4 months to construct a PDX model and 4–8 months to expand a sufficient number of mice to perform an interpretable preclinical study. ¹⁶⁷ Typically, it is impossible for patients with advanced cancer to wait 4–8 months for clinical decision making based on PDX model drug tests.

Because PDX trials are technically cumbersome, time consuming, and expensive, in vitro cultures of patient‐derived cancer cells show the potential to be more easily expanded for genetic manipulations and high‐throughput screenings, which are called patient‐derived organoids (PDOs) (Figure 4). ¹¹ , ¹⁶⁸ Recently, it has been optimized how to establish PDOs. ⁶³ , ¹⁶⁹ , ¹⁷⁰ , ¹⁷¹ This approach allows self‐organizing three‐dimensional structures that reconstitute multiple functional structures of the primary tumors. ¹⁷⁰ , ¹⁷² It is reported that primary or metastatic CRC PDOs can be directly established from patient specimens (biopsy or surgical resection) or PDX explants. ¹⁷³ , ¹⁷⁴ , ¹⁷⁵ The success rates of CRC PDOs range between 60% and 90% according to the reports, and CRC PDOs inherit many features of the original tumor tissues, including histopathologic, genetic, and transcriptomic profiles. ¹⁷⁶ , ¹⁷⁷ , ¹⁷⁸ , ¹⁷⁹ , ¹⁸⁰ Thus, PDOs might be an alternative option to the PDX model. With the development of novel preclinical models such as PDX and PDO, CRC treatment will eventually move to the era of individualized treatment.

FIGURE 4.

The establishment of PDOs (patient‐derived organoids). The tumor tissues derived from CRC (colorectal cancer) patients are cultured under three‐dimensional conditions. The organoids are then obtained, which have many features similar to the original tumor tissues in patients. Thus, these organoids can be used to mimic the tumors in patients

7. PROSPECTS

CRC is a multistep disease caused by the combination of various genetic and epigenetic alterations and environmental risk factors. The development of new therapeutic methods for CRC remains an ongoing challenge because of the inter‐ and intra‐tumoral phenotypic heterogeneity characterized by distinct molecular traits and responses to therapy. ⁹ However, phenotypic heterogeneity is not reflected in traditional CRC models such as tumor cell lines and GEMMs. The PDX system is a promising tool to study the development and progression of CRC and uncover the latent mechanisms. Ideally, the preclinical model in oncology should reach some criteria to help clinical decisions: availability in number, variations in genetic background, maintenance of the characteristics of human tumors, and consistency in therapy response similar to patients. ⁷³ , ¹¹⁸ The large, high‐quality PDX biobanks are helpful for supporting further studies (Figure 5). Currently, some international groups are engaged in establishing the PDX biobank. These groups include EurOPDX, the NCI Repository of Patient‐Derived Models, the Innovative MODels Initiative (IMODI) Consortium, the Pediatric Preclinical Testing Consortium, the Children's Oncology Group Cell Culture and Xenograft Repository, the Public Repository of Xenografts, and the Jackson Laboratory PDX Resource. ⁶⁴ , ¹⁸¹ These great efforts will promote the application of PDX models in drug discovery and basic research in the context of population‐scale studies to improve the prognosis of cancer patients. ¹⁸² Although a large number of Chinese groups have made efforts to develop the PDX model for single‐center research, ¹²⁶ , ¹³⁰ , ¹⁸³ , ¹⁸⁴ , ¹⁸⁵ integrative cohorts containing multicenter data should be established and maintained in the future.

FIGURE 5.

The establishment of PDX (patient‐derived xenograft) biobank for precision medicine. Specimens from CRC patients enrolled in the study are submitted to PDX biobank. Researchers can take advantage of the biobanks to screen drugs, uncover biomarkers, and study the basic mechanisms. The latent drugs will submit to animal tests using PDX in biobanks. The effective drugs and biomarkers will used to improve the therapeutic effects.

Standardizing protocols for constructing and biobanking PDX models are critical to maximize the impact of studies using these publically available PDX resources. Currently, the fact that underlying mechanisms about the occurrence and progression of CRC are not fully elucidated causes the difficulty in the standardization of the CRC PDX model system. In future studies, with the development of high‐throughput sequencing and other technologies, ¹⁸⁶ it will be much easier to construct the CRC PDX model for simulating the heterogeneity of human CRC tumors. In addition, it is important to study how to reduce the time to establish the CRC PDX model. Improved transplantation technologies, such as ultrasound‐guided fine needle aspiration, may be helpful. Clinically, ultrasound‐guided fine needle aspiration technique is commonly employed for tissue acquisition. ¹⁸⁷ The ultrasound monitor is applied as a reference to orthotopically implant CRC cells into mice. Orthotopic implantation contributes to increase in the success rate of the PDX model. Because of its minimally invasive feature, this method decreases the risk of complications in orthotopic transplantation. Compared with the traditional orthotopic transplantation model, the occurrence of inflammation is reduced and the animal's healing time is reduced, which significantly upregulate the survival rate of mice. Currently, attempts have been made to establish the PDX model of some types of cancers using ultrasound monitor, such as lung cancer, pancreatic cancer, and melanoma. ¹⁸⁸ , ¹⁸⁹

In conclusion, the PDX model has emerged as a promising approach in the discovery of oncologic drugs and basic research in CRC. The PDX model is widely used in various CRC research studies. However, this model has certain limitations, including the use of animals, limited engraftment efficiency, and high costs. These issues need to be further addressed. Now, the PDO model is established and can be used to overcome these limitations to some extent.

AUTHOR CONTRIBUTIONS

Xiaofeng Liu and Zechang Xin prepared the manuscript. Kun Wang reviewed the manuscript.

CONFLICT OF INTEREST

The authors have no conflicts of interest. Xiaofeng Liu is an editorial board member of AMEM and a coauthor of this article. To minimize bias, he was excluded from all editorial decision making related to the acceptance of this article for publication.

ETHICS APPROVAL

All authors read and approved the final manuscript.

Liu X, Xin Z, Wang K. Patient‐derived xenograft model in colorectal cancer basic and translational research. Anim Models Exp Med. 2023;6:26‐40. doi: 10.1002/ame2.12299