Utilizing patient-derived tumors in a subcutaneous preclinical model is an excellent way to study the efficacy of novel therapies, predictive biomarker discovery, and drug resistant pathways. This model, in the drug development process, is essential in determining the fate of many novel anti-cancer therapies prior to clinical investigation.
Patient derived tumor xenograft (PDTX) models provide a necessary platform in facilitating anti-cancer drug development prior to human trials. Human tumor pieces are injected subcutaneously into athymic nude mice (immunocompromised, T cell deficient) to create a bank of tumors and subsequently are passaged into different generations of mice in order to maintain these tumors from patients. Importantly, cellular heterogeneity of the original tumor is closely emulated in this model, which provides a more clinically relevant model for evaluation of drug efficacy studies (single agent and combination), biomarker analysis, resistant pathways and cancer stem cell biology. Some limitations of the PDTX model include the replacement of the human stroma with mouse stroma after the first generation in mice, inability to investigate treatment effects on metastasis due to the subcutaneous injections of the tumors, and the lack of evaluation of immunotherapies due to the use of immunocompromised mice. However, even with these limitations, the PDTX model provides a powerful preclinical platform in the drug discovery process.
Colorectal cancer (CRC) is a significant contributor to cancer deaths in the United States. In 2015, there were an estimated 132,700 new cases of CRC with 49,700 deaths 1. Although the prognosis in patients with localized disease is excellent, patients with advanced disease have poor outcomes, making this a major priority in the development of novel therapies. Despite standard of care chemotherapeutic regimens and newer biologics that are deployed against this disease, there has been only an incremental increase in overall survival. Accordingly, there is a significant effort in understanding the driver pathways involved in facilitating tumor growth in this disease. The Cancer Genome Atlas Network has recently identified numerous main pathways that are implicated in CRC dysregulation and include: WNT, phosphoinositide 3-kinase (PI3K), RAS, transforming growth factor-β (TGF- β) and TP53 2. Together, with investigations describing other pathways that potentiate growth in CRC have ignited the development of newer therapies aimed at significantly improving the survival in this patient population 3-5. Utilizing preclinical models in oncology drug development have been essential in this process in predicting the clinical activity of these novel compounds.
Various preclinical models have been utilized in the drug development process. Considering that preclinical transgenic animal models and immortalized cell lines have been unsuccessful in determining the clinical activity of novel oncology therapies, largely due to their inability to reflect the complexity of human tumors, patient-derived tumor xenograft (PDTX) models have been established. The greatest advantage of this model is that tumor heterogeneity remains intact and closely reflects the molecular characteristics and clonality of the originating patient tumor 6-9. PDTX models provide an excellent in vivo preclinical platform to study novel agents, drug resistance pathways, combinational strategies, and cancer stem cell biology 10.
A general overview of the PDTX process is illustrated in Figure 1. It begins in the clinic, consenting patients to allow some of their excess tumor tissue to be used for this research. Next, at surgery, a piece of the tumor is grossed by a pathologist and put into media to be transported to research personnel. Immediately after this, a section of the tumor is cut into small pieces and transplanted into immunodeficient mice subcutaneously. Once the tumor grows, it is passaged into different generations of mice in order to maintain the tumor10. Typically, after the F3 generation the tumor can be expanded into a treatment study where novel compounds and/or combinational therapies are evaluated. Utilizing Next Gen Seq (Exome Seq, RNA Seq and SNP array) potential predictive biomarkers are discovered that assist in the selection of patients that may derive benefit from a particular treatment.
The overarching goals of using PDTX models are to: 1) evaluate the efficacy of novel therapies as single agent or in combination and 2) identify predictive biomarkers of sensitivity or resistance prior to clinical investigation. In this manuscript, we provide the methodology in the initiation and maintenance of a CRC PDTX bank and provide the advantages and limitations of this model in drug development discovery.
Figure 1. Overview of the CRC PDTX Model Protocol. A patient derived tumor is received from surgery and immediately injected into athymic nude mice subcutaneously. Once the tumor grows it is expanded into subsequent generations and eventually expanded for treatment studies. Treatment responses are assessed and predictive biomarkers are identified that may aid in patient selection. Please click here to view a larger version of this figure.
Ethics Statement: Patient-derived colorectal adenocarcinoma tumor specimens were obtained from consenting patients at the University of Colorado Hospital in accordance with a protocol approved by the Colorado Multiple Institutional Review Board (08-0439). All animal work was performed under animal protocols approved by the University of Colorado Denver Institutional Animal Care and Use Committee (IACUC, Protocol # 51412(06)1E and 96813(04)1E).
1. Receiving and Preparing Patient Blood
2. Receiving and Preparing Patient Tumor Sample
3. Injection of Patient Derived Tumor Xenografts
4. Maintenance of Patient Derived Tumor Xenograft Bank
5. Developmental Therapeutics with Patient Derived Tumor Xenografts
Note: Most tumors at F3 generation have good growth kinetics (grow faster and more consistent), therefore, proceed to PDTX drug efficacy studies.
6. Organization of a PDTX bank
Similarities of Common Mutations in the CRC PDTX Models and the TCGA
We investigated whether the percentage of common mutations (KRAS, NRAS, BRAF, PIK3CA, APC, CTNNB1 and TP53) in the CRC PDTX bank were representative to the mutation frequency seen in the CRC patient population. As shown in Figure 2A (TCGA) and B (CRC PDTX bank), the frequency of mutations in these genes were very similar between the TCGA (n= 276 patients) and the CRC PDTX bank (n= 59 CRC patients). The biggest difference observed was in the APC gene whereby a 23 % difference was seen. These results suggest that common mutations observed in the CRC patient population is well represented in the CRC PDTX model.
Evaluation of the Stability of Treatment Responses Between Different Generations
In this CRC PDTX model, we set out to determine whether treatment effects were similar between different generations. Tumors were expanded in athymic nude mice (10 tumors/group) and the efficacy of the standard of care agents such as cetuximab (0.4 mg/mouse IP twice per week) and irinotecan (15 mg/kg IP once per week) were examined in 4 unique CRC PDTX models in two separate generations. As illustrated in Figure 3A and B, CRC026 (F3) was more resistant to cetuximab treatment, while CRC010 (F6) exhibited treatment sensitivity. Similar findings were observed when these CRC explants were treated in different generations; CRC026 (F9) was resistant and CRC010 (F7) was sensitive to cetuximab. To determine the efficacy of irinotecan in the PDTX model, we investigated treatment effects on tumor growth in 2 CRC PDTX models. While CRC098 (F8) was resistant to irinotecan treatment, CRC036 (F5) exhibited sensitivity (Figure 3C and D). As was observed with cetuximab, treatment with irinotecan in different generations did not change the treatment response in these tumors; CRC098 (F12) was resistant and CRC036 (F10) was sensitive to irinotecan (Figure 3C and D). Although the growth kinetics of untreated tumors was sometimes different between generations, the treatment responses to irinotecan and cetuximab remained the same, indicating the stability of this model in evaluating anti-cancer therapies.
Investigation of the Stroma Component in the CRC PDTX Model
Next, we were interested in determining if the stroma component within this CRC explant model was comprised of human and/or mouse derived cells. We used a dual-color Cot-1 FISH assay that consisted of mouse Cot-1 DNA (green fluorophore) and human Cot-1 DNA (red fluorophore) to determine mouse and human cells in 10 separate CRC explants between F0 and F1 generations 25. As shown in Figure 4A, in the F1 generation the human stroma is replaced by the mouse stroma, since the human tumor is now surrounded by all mouse stroma. These findings were evident in all 10 CRC explants that were subjected to Cot-1 FISH between the F0 and F1 generations. In addition to Cot-1 FISH, we investigated protein levels of mouse and human HGF and VEGF ligands in the F0 and F1 generations using mouse and human ELISAs for these ligands. As displayed in Figure 4B and C, whereas all human derived HGF in the F0 generation is replaced with mouse HGF in the F1 generation, VEGF ligand consisted of both human and mouse in the F1 generation. Together these experiments suggest that in the CRC PDTX mouse model that the mouse stroma overtakes the human stroma in the F1 generation and the ligands secreted may vary with respect to being human and/or mouse derived.
Figure 2. Comparison of Common Mutations in the CRC PDTX Bank and the TCGA. We observed very similar mutation frequencies between the TCGA (A) and the CRC PDTX model (B) with respect to KRAS, NRAS, BRAF, PIK3CA, CTNNB1 and TP53. Please click here to view a larger version of this figure.
Figure 3. The Effects of Cetuximab and Irinotecan Treatment on Tumor Growth. (A) CRC026 displayed resistance to cetuximab when evaluated in the F3 and F9 generations and (B) CRC010 exhibited sensitivity to cetuximab in the F6 and F7 generations. (C) CRC098 showed resistance to irinotecan in the F8 and F12 generations, while (D) CRC036 was sensitive to irinotecan in the F5 and F10 generations. Each data point represents an average of 10 tumors per treatment group. Mice were treated with cetuximab (100 μl IP 400 μg/mouse) twice a week and irinotecan (100 μl IP 20 mg/kg) was dosed once a week. Data presented as mean ± SEM. Please click here to view a larger version of this figure.
Figure 4. Evaluation of Mouse and Tumor Cell Components in the F0 and F1 Generations. (A) Dual-color FISH for the human cot-1 DNA (red) and mouse cot-1 DNA (green) was used to investigate differences between tumors in the F0 and F1 generation. Human stroma (F0) is replaced by mouse stroma (F1) in CRC098 and CR174 (scale = 20x). Necrotic cells were identified by reduced or lack of DAPI intercalation and red fluorescence. Investigation of mouse and human HGF and VEGF ligand expression by ELISA (mouse and human ELISAs) between F0 and F1. (B) Human HGF was replaced by mouse in the F1 generation and (C) human and mouse VEGF were evident in the F1 generation (picograms/milliliter [pg/ml]). Data presented as mean ± SEM. Please click here to view a larger version of this figure.
The PDTX drug discovery platform offers an improved model to the shortcomings of other preclinical models that are unreliable in predicting clinical activity of novel compounds. Importantly, tumors in this model are biologically stable, retain metastatic potential, and exhibit similar drug responsiveness from generation to generation. In this model, patient derived tumors are injected into athymic nude mice, passaged, and subsequently used in therapeutic evaluation. There are several critical steps for a successful PDTX bank that include: 1) a cohesive clinical team to identify/consent patients and for the removal and grossing of tumor tissue for the PDTX model and 2) a strong research group with excellent laboratory and animal technical skills for injecting tumors, organization and maintenance of the PDTX bank and monitoring the health of mice. A significant advantage in our PDTX models is injecting tumors with the trocar procedure. The alternative method of cutting a tumor pocket and then suturing or using skin clips is more time consuming for personnel, requires more pain medication for the mice, and monitoring of the sutured or wound clipped area. The trocar procedure requires less training, is very quick, the mice are under anesthesia for less time, and less pain medications are needed. Therefore, in our experience injecting tumors with trocars is the best method versus alternative methods. These factors will significantly influence the overall success of the PDTX bank and in the drug discovery process. Although this in vivo model is considerably more costly than testing agents in cancer cell line cultures, PDTX models offer a more clinically relevant approach in testing oncology compounds.
In the CRC PDTX bank, we have received 99 tumor samples that were injected into athymic nude mice. There were 68 out of 99 (68.7% take rate) tumors that grew in mice and were passed into multiple generations. There are several reasons why some tumors that we received did not grow in mice. For instance, we sometimes received only very small pieces of tissue that only allowed us to inject into only one mouse decreasing the chances of establishing a tumor. Another issue was that at times we received patient tissue that was normal and did not contain tumor cells evident on the H&E slide. In addition, some poor tissue quality can be due to receiving already necrosed tumor where there was a patient response to treatment. Therefore, it is important to have a reliable surgical and pathology team when grossing the tumor. Considering some of these issues, we have been able to establish one of the largest CRC PDTX banks of fully annotated tumors with respect to mutations, gene expression, and clinical characteristics, making this a valuable model for the evaluation of novel therapies with the ultimate goal of improving patient outcomes.
Numerous biological and combinational therapies have been investigated in this model with the objective of determining efficacy, drug resistance mechanisms, as well as treatment effects on the cancer stem cell population. Our group has examined the efficacy of numerous novel biological pathway inhibitors using this preclinical model 12-18, which has provided valuable insight into further clinical development of these compounds. Many of these studies have identified predictive biomarkers 12-14,17,18 that may aid in patient selection in future clinical trials. In addition, other studies have further determined mechanisms of treatment resistance using this model, which has led to the development of rational combinations 19-24. For instance, Bardelli and colleagues 19 demonstrated that cetuximab treatment induced MET amplification and that Met may be an underlying mechanism of treatment resistance to cetuximab in CRC. 19In a separate study, Bertotti et al. 20 identified Her2 as a target in CRC tumors that were resistant to cetuximab. Finally, we and another group have shown that treatment with a Notch pathway inhibitor in combination with irinotecan reduced the CRC cancer stem cell population and tumor recurrence after treatment was discontinued 25,26. Together, these studies demonstrate the potential power of utilizing PDTX models in the drug development process that may significantly impact the further development of novel compounds.
Despite the major advantages using patient derived tumors in determining the efficacy of new compounds, there are several limitations in this model. As shown experimentally in this paper, the human stroma from the originating tumor (F0) is replaced with the mouse stroma at the F1 generation in the CRC PDTX model. Depending on the particular drug target, this may be an issue when mouse ligands are unable to activate the receptor(s) on human tumor cells. For instance, we show that in the F1 generation, HGF is derived from the mouse stroma and studies have determined that mouse HGF is incapable of functionally activating the human c-Met receptor 27,28. As a result, c-Met inhibitors may not exhibit anti-tumor growth effects in these PDTX models. In fact, humanized HGF SCID mice have been developed to address this potential problem 28. Another disadvantage of using subcutaneous tumors is the inability to study treatment effects on the metastatic potential of tumors. Using orthotopic models, although more technically difficult to establish and image tumor growth, will likely provide better insight into the treatment effects on metastasis. A final limitation of this model is the inability to investigate the role of the immune system in potentiating tumor growth and its function in facilitating resistance to treatment. Moreover, with the excellent activity of immunotherapies recently demonstrated in the clinic, using immunodeficient mice prevents the investigation of immune targeted agents. Accordingly, humanized mouse models have been developed to address these limitations, which may provide value in understanding the fundamental role of immune-tumor interactions as well as allow for the investigation of immunotherapies in combination with other novel and approved compounds.
In conclusion, we provide a method on developing and maintaining a CRC PDTX model that is invaluable in determining therapeutic efficacy, predictive biomarkers and drug resistance pathways of novel anti-cancer agents. Although this model has inherent challenges, the utility of this model is that it closely recapitulates the tumor heterogeneity of the original tumor, which offers a more precise and clinically relevant investigation of novel therapies prior to clinical evaluation. Future evaluation of the clinical impact in drug development of this preclinical PDTX model will ultimately determine the power of this model in predicting the clinical activity of therapies in cancer.
The authors have nothing to disclose.
This work was supported by grant 1R01CA152303-01.
RPMI or DMEM | Corning | 10-040-CV | |
Penicillin-Streptomycin | Corning | 30-002-CI | |
Non-essential Amino Acids | Corning | 25-025-CI | |
Fetal Bovine Serum | Corning | 35-010-CV | Thaw in -4 °C, then activate for 30 minutes at 60 °C water bath |
CPT blood tube | BD vacutainer | 362761 | |
Microcentrifuge tube | Surelock | A-7002 | |
Phospate-Buffered Saline | Corning | 21-040-CV | |
Cyrogenic vials | Cyroking | C0732901 | |
Plastic tumor cutting dish | Trueline | TR4001 | |
Scissors | Roboz | RS-5881 | |
Forceps | Roboz | RS-5135 | |
Matrigel (gelatinous protein mixture) | Corning | 354234 | Store at -20 or -80 °C, then thaw on ice, do not leave at room temperature |
10% Formalin cups | Protocol | 032-059 | |
Liquid Nitrogen Dewar Storage | Thermolyne | CY50900 | |
Portable liquid nitrogen dewar | Nalgene | 4150-2000 | |
Dimethyl Sulfoxide | Fischer | 67-68-5 | |
Freezing container: Mr Frosty | Nalgene | 5100-0001 | |
Isopropyl Alcohol | Decon | 64-17-5 | |
Trocars | Innovative Research of America | MP-182 | |
Anesthesia machine | Patterson Veterinary | none | |
Anesthesia box | Patterson Veterinary | none | |
Isoflurane | Vet one | 1038005 | |
F-Air Canister | Bickford Omnicon | 80120 | |
Meloxicam | Vet one | 5182-90C | |
Calipers | Fowler | 54-100-167 | |
Weight scale | Ohaus | Scout Pro SP601 |