Here we present a protocol to test the efficacy of targeted therapies selected based on the genomic makeup of a tumor. The protocol describes identification and validation of structural DNA rearrangements, engraftment of patients’ tumors into mice and testing responses to corresponding drugs.
We present here an integrative approach for testing efficacy of targeted therapies that combines the next generation sequencing technolo-gies, therapeutic target analyses and drug response monitoring using patient derived xenografts (PDX). This strategy was validated using ovarian tumors as an example. The mate-pair next generation sequencing (MPseq) protocol was used to identify structural alterations and followed by analysis of potentially targetable alterations. Human tumors grown in immunocompromised mice were treated with drugs selected based on the genomic analyses. Results demonstrated a good correlation between the predicted and the observed responses in the PDX model. The presented approach can be used to test the efficacy of combination treatments and aid personalized treatment for patients with recurrent cancer, specifically in cases when standard therapy fails and there is a need to use drugs off label.
Patient-derived xenografts (PDXs), which are generated from the implantation of patient tumor pieces into immunodeficient mice, have emerged as a powerful preclinical model to aid personalized anti-cancer care. PDX models have been successfully developed for a variety of human malignancies. These include breast and ovarian cancers, malignant melanoma, colorectal cancer, pancreatic adenocarcinoma, and non-small cell lung cancer1,2,3,4,5. Tumor tissue can be implanted orthotopically or heterotopically. The former, considered more accurate but technically difficult, involves transplantation directly into the organ of tumor origin. These types of models are believed to precisely mimic histology of the original tumor due to the “natural’ microenvironment for the tumor6,7. For example, orthotopic transplantation into the bursa of the mouse ovary resulted in tumor dissemination into the peritoneal cavity and the production of ascites, typical of ovarian cancer8. Similarly, injection of breast tumors into the thoracic instead of the abdominal mammary gland affected the PDX success rate and behavior9. However, orthotopic models require sophisticated imaging systems to monitor tumor growth. Heterotopic implantation of solid tumor is typically performed by implanting tissue into the subcutaneous flank of a mouse which allows for easier monitoring of tumor growth and is less expensive and time consuming7. However, tumors grown subcutaneously rarely metastasize unlike as observed in the case of orthotopic implantation10.
The success rate of engraftment has been shown to vary and greatly depend on the tumor type. More aggressive tumors and tissue specimens containing a higher percent of tumor cells were reported to have better success rates12,13. Consistent with this, tumors derived from metastatic sites were shown to engraft at frequencies of 50–80%, while those from primary sites engraft at frequencies as low as 14%12. In contrast, tissue containing necrotic cells and fewer viable tumor cells engraft poorly. Tumor growth can also be promoted by the addition of basement membrane matrix proteins into the tissue mix at the time of the injection into mice14 without compromising properties of the original tumor. The size and number of tissue pieces intended for implantation were also found to affect the success rate of engraftment. Greater tumor take-rates were reported for implantation in the sub-renal capsule compared to subcutaneous implantation due to the ability of the sub-renal capsule to maintain the original tumor stroma and provide the host stromal cells as well15.
Most studies use NOD/SCID immunodeficient mice, which lack natural killer cells16 and have been shown to increase the tumor engraftment, growth and metastasis compared to other strains14. However, additional monitoring is required as they may develop thymic lymphomas as early as 3-4 month of age13. In ovarian tumor transplants grown in SCID mice, the outgrowth of B cells was successfully inhibited by rituximab, preventing the development of lymphomas but without impacting the engraftment of ovarian tumors17.
More recently, NSG (NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ) mice, carrying a null mutation in the gene encoding the interleukin 2 receptor gamma chain18, became a frequently used strain for the generation of PDX models. Tumors from established PDX models passaged to future generations of mice are reported to retain histological and molecular properties for 3 to 6 generations19,20. Numerous studies have shown that the treatment outcomes in PDX models mimic those of their corresponding patients2,3,4,21,22,23. The response rate to chemotherapy in PDX models for non-small lung cancer and colorectal carcinomas was similar to that in clinical trials for the same drugs24,25. Studies conducted in PDX models, developed for patients enrolled in clinical trials, demonstrated responses to tested drugs similar to those observed clinically in corresponding patients2,3,4.
High–throughput genomic analyses of a patient tumor in conjunction with PDX models provide a powerful tool to study correlations between specific genomic alterations and a therapeutic response. These have been described in a few publications26,27. For example, therapeutic responses to the EGFR inhibitor cetuximab in a set of colorectal PDX models carrying EGFR amplification, paralleled clinical responses to cetuximab in patients28.
There are a few challenges associated with the development and application of PDX models. Among those is tumor heterogeneity29,30 that may compromise the accuracy of treatment response interpretation as a single cell clone with higher proliferative capacity within a PDX can outgrow the other ones31, thus resulting in a loss of heterogeneity. Additionally, when single tumor biopsies are used to develop PDX, some of the cell populations may be missed and will not be represented in the final graft. Multiple samples from the same tumor are recommended for implantation to resolve this issue. Although PDX tumors tend to contain all the cell types of the original donor tumor, these cells are gradually substituted by those of murine origin3. The interplay between murine stroma and human tumor cells in PDX models is not well understood. Nevertheless, stromal cells were shown to recapitulate tumor microenvironment33.
Despite these limitations, PDX models remain among the most valuable tools for translational research as well as personalized medicine for selecting patient therapies. Major applications of PDXs include biomarker discovery and drug testing. PDX models are also successfully used to study drug resistance mechanisms and identify strategies to overcome drug resistance34,35. The approach described in the present manuscript allows the researcher to identify potential therapeutic targets in human tumors and to assess the efficacy of corresponding drugs in vivo, in mice harboring engrafted tumors which were initially genomically characterized. The protocol uses ovarian tumors engrafted intraperitoneally but is applicable to any type of tumor sufficiently aggressive to grow in mice2,3,12.
Fresh tissues from consenting patients with ovarian cancer were collected at the time of debulking surgery according to a protocol approved by Mayo Clinic Institutional Review Board (IRB). All animal procedures and treatments used in this protocol were approved by Mayo Clinic Institutional Animal Care and Use Committee (IACUC) and followed animal care guidelines.
1. Mate pair sequencing and analyses
NOTE: Either fresh or flash frozen tissue must be used for mate pair (MPseq) sequencing. Paraffin embedded material is not suitable because it contains fragmented DNA.
2. Selection of therapeutic targets
3. Validation of genomic rearrangements by PCR and Sanger sequencing
4. Tumor engraftment and maintenance
5. Testing responses to genomically identified targets in PDX models
Tissue from resected ovarian tumors at the time of debulking surgeries were collected in accordance with IRB guidance and used for 1) genomic characterization and 2) engraftment in immunocompromised mice (Figure 1). Mate-pair sequencing protocol36,37 was used to identify structural alterations in DNA including losses, gains and amplifications. A representative genome plot illustrating a landscape of genomic changes in one tumor (designated as OC101) is shown in Figure 2. Typical for high grade serous subtype tumors, multiple gains (blue lines) and deletions (red lines) were found, indicating high levels of genomic instability, as well as chromosomal losses and gains, indicative of aneuploidy, were observed. On an average 300-700 alterations total are identified in high grade serous subtype tumors45. Subsequent analyses revealed a few DNA alterations that were potentially targetable with clinically relevant drugs. The top-ranked alteration for therapeutic intervention in the OC101 tumor was an amplification at chromosome 17 involving ERBB2 (Figure 2 and Figure 3A). ERBB2 is a gene which codes for HER2 receptor which is known upon dimerization with EGFR, HER3 or HER4 to activate RAS/ERK and PI3K/AKT signaling pathways and promote cell growth, cell migration and invasion. HER2 inhibitors (e.g., monoclonal antibodies pertuzumab and trastuzumab) are effective in treating breast cancer patients when tumors overexpress the HER2 protein. Anti-HER2 therapy for ovarian cancer, however, is not FDA approved.
Comparison of the genomic profile of donor patients’ tumor (Figure 1) to that of a corresponding PDX model (not shown) revealed a striking similarity, consistent with all previous studies reporting the molecular closeness of original tumors to their PDX derivatives.
To validate MPseq results at the DNA level, several sets of specific primers were designed for the edges of amplified region containing the ERBB2 gene, and PCR was carried out using DNA isolated from the original tumor as well as from a tumor propagated for several generations in the mice. A representative gel image of amplified products using two different sets of primers is shown in Figure 3B. No band was detected when normal pooled genomic DNA (designated as C), that did not contain amplification of the ERBB locus, was amplified. Purification of the products from the gel and Sanger sequencing (not shown) further confirmed the alteration predicted by MPseq. Further validation was conducted by examining the expression of HER2 protein in the corresponding PDX tumor using immunoblotting. The analysis revealed a high level of HER2 protein (the result is not shown), consistent with the observed amplification of the ERBB2 gene.
In the DNA of a different ovarian tumor (designated T14) numerous regional gains were observed. Those included AKT2 and RICTOR genes (Figure 3C). Both were of great interest from a therapeutic perspective as inhibitors of AKT2 and mTOR, which RICTOR associates with, are available and currently in clinical trials. Since there were no mate pair reads spanning the vicinity of either gene, simple PCR validation of the gain as detected by MPseq was not possible. We, therefore, tested the expression level of corresponding proteins by immunoblotting. High levels of AKT and RICTOR were observed (Figure 3D) suggesting that treatment with targeted drugs is warranted.
To test the sensitivity of this tumor to inhibitors of AKT and mTOR, PDX mice with intraperitoneally implanted T14 tumor were expanded and randomized to receive only chemotherapy (carboplatin/paclitaxel) or a combination of chemotherapy with pan-AKT inhibitor MK-220641 or mTOR inhibitor MK-866942. Chemotherapy was given to mice in the combination arm for 2 weeks prior to the addition of the targeted therapy (Figure 4). Ultrasound measurements were taken weekly to monitor tumor regression/growth.
Each treatment arm contained no less than 7 mice. This number was sufficient to observe differences in responses between the groups while keeping costs of the study at a considerably lower mark. Fewer mice (three to four) can be used in the control untreated group, as individual variations in tumor growth rate are negligible and growth is normalized to a tumor size at which treatment in other arms begins.
No difference was observed between chemo-treated and untreated groups in the first 3 weeks of observation (not shown). A significant reduction of tumor burden (58% median) was observed in the chemotherapy treated group by the end of week 6. An extra benefit over chemotherapy alone was observed in groups which received a combination of chemotherapy with targeted therapies. The difference became evident at week 4 and 3 for MK-8669 (Figure 4A) and MK-2206 (Figure 4B) respectively.
Animals were euthanized and tumor tissue was collected for molecular analyses of treatment response at the end of the treatment trial at week 7. For that purpose the amounts of total and phosphorylated S6 kinase (Figure 5A,B), AKT and mTOR (Figure 5C,D) were determined using immunoblotting. Ribosomal protein S6 kinase is a downstream messenger of the AKT-mTOR pathway, known to be up-regulated and phosphorylated upon stimulation of the AKT-mTOR axis by growth factors to promote cell survival and growth. Comparison of the levels of these proteins in untreated or treated with chemotherapy PDX tumors to mice that received AKT or mTOR inhibitors showed a marked decrease in the latter two (Figure 5), indicating effectiveness of the targeted therapy on the molecular level. Adjustments to the treatment regiment, as far as application timing for the combined drugs and therapy duration, should be made and tested to achieve better responses.
Primer Name | Sequence | Primer Mix | Primers combined | ||
OC101 17a-17b R1 | CTGGTCCTGGGAAATAGACACTAGATAAATCATC | 1 | OC101 F1 and R1 | ||
OC101 17a-17b R2 | GCTCAAGACAGTAACACCTAGTAGTTGATTCTGC | 2 | OC101 F1 and R2 | ||
OC101 17b-17a F1 | GGATTACGGGAACGTGCTACCTTG | 3 | OC101 F2 and R1 | ||
OC101 17b-17a F2 | AATGCCCTAGCAGTTCTATCCCCACTG | 4 | OC101 F2 and R2 |
Table 1: Primers used for the validation of the OC101chromosome 17 alteration.
Reagent | Quantity to add for 1 reaction (µL) | Quantity to add for 4 reactions (µL). [Multiply by 4.3 to make enough for each primer mix (4)] |
Nuclease-free water | 20.45 | 89.94 |
10x buffer (Easy A) | 2.5 | 10.75 |
dNTPs 10 mM | 0.5 | 2.15 |
Template DNA (concentration >10 ng/µL) | 0.3 | 1.29 |
Taq Polymerase | 0.25 | 1.08 |
Table 2: PCR setup to validate a junction.
Temp (C) | Time | |
94 | 5 min | |
94 | 40 s | 35 cycles |
59 | 40 s | |
72 | 2 min | |
72 | 5 min | |
4 | Hold |
Table 3: Cycling conditions for PCR to validate a junction.
Figure 1: Schematic representation of the strategy for genomically-guided therapy testing using PDX models for serous ovarian carcinoma. H&E staining of ovarian tumor is shown (top). Scale bar=100 mm. Please click here to view a larger version of this figure.
Figure 2: Genomic characterization of ovarian tumors using MPseq. Genome plot showing the landscape of structural alterations and copy number changes as detected by MPseq. The X axis spans the length of the chromosome with chromosome position number shown. Each chromosome is indicated on the right and left Y-axis. The height of the horizontal traces for each chromosome indicates the number of reads detected for 30k base pair windows. DNA copy numbers are indicated by color, with grey representing the normal 2N copy state, red corresponding to deletions and blue-to gains. Connecting black lines correspond to chromosomal rearrangements. Alterations at ERBB2 locus are depicted. Please click here to view a larger version of this figure.
Figure 3: Selection of targetable alterations and their validation. (A) A close-up segment of the genome plot shown in Fig.1 illustrating an amplification of the ERBB2 gene on chromosome 17 (in blue). Chromosome numbers are as indicated. (B) Validation analysis of breakpoints at ERBB2 locus as identified by MPseq using PCR amplification. C is a pooled genomic DNA control, OC101 is DNA from patient tumor, M is a DNA ladder. (C) A close-up of segments of the genome plot for another ovarian tumor showing gains (indicated by blue line) at the AKT locus (top) and at the RICTOR gene (bottom). Chromosomes are as indicated. (D) Validation analysis of expression of proteins of AKT/mTOR pathway by immunoblotting using tumor tissue from PDX (T14), genomic alterations for which are depicted in C. 30 mg of total protein and specific antibodies to AKT, RICTOR, p-mTOR were used. MAPK served as a loading control. Pos con is an independent tumor used as a positive control. Please click here to view a larger version of this figure.
Figure 4: The comparison of responses to chemotherapy alone and combined with targeted therapy in tumor harboring gains at AKT2 and RICTOR genes with corresponding drugs. Treatment responses to a combination of chemotherapy and anti-mTOR drug MK-8669 (A) or inhibitor of pan-AKT MK2206 (B) versus chemotherapy alone. The time of administration for each treatment and the duration are shown by the arrows. Volumes are expressed as percent of initial volume at the start of the treatment as mean +/- SD. Chemo is chemotherapy. Please click here to view a larger version of this figure.
Figure 5: Comparison of molecular changes elicited by each treatment as determined by immunoblotting analysis. (A) Levels of S6 and phospho-S6, the downstream effector of AKT-mTOR pathway are shown. 30 mg of total protein and specific antibody to S6 and p-S6 were used. GAPDH was used as a loading control. Quantification of protein levels normalized to GAPDH level is shown in (B) (C) Levels of mTOR, p-mTOR, AKT and p-AKT as detected by immunoblotting. GAPDH was used as a loading control. (D) Quantification of protein levels normalized to GAPDH level. NT is not treated, chemo is chemotherapy. Please click here to view a larger version of this figure.
We describe the approach and protocols we used to conduct a “clinical trial” in PDX models that takes advantage of molecular characteristics of the tumor as obtained by genomic profiling to determine the best choice of drugs for testing. Multiple sequencing platforms are currently used for genomic characterization of primary tumors including whole genome sequencing, RNAseq and customized gene panels. For high grade serous ovarian carcinoma, MPseq to identify structural alterations, DNA rearrangements and copy number changes, is particularly useful because of the high degree of genomic instability observed in this type of tumor. The second advantage of the MPseq platform is that it covers the entire genome but costs significantly less than other comprehensive sequencing technologies. MPseq, however, is not suitable for point mutation detection as base coverage is not sufficient, reaching only 8-10x. One of the limitations of using MPseq alone for genomic characterization of a tumor is the presence of complex clustered chromosomal rearrangements, analysis of which does not predict the expression of impacted genes of interest. A junction detected by MPseq predicted to create a putative fusion gene that validates in the DNA by PCR may not be expressed because of a frame shift or small deletions and insertions in the promoter region. Similarly, chromosomal gains and amplifications for potential therapeutic targets should be carefully assessed and validated on the RNA or protein level to ensure expression of the gene of interest.
Although, the molecular make-up of the tumor is largely preserved after propagation in mice for several generations, changes in expression levels of key genes may occur over time either reflecting tumor evolution, clonal selection or adaptive response to murine environment. Thus, validation of an alteration intended for therapeutic intervention in both original donor tumor and PDX tumor is critical. For tumor isolated from PDX models both RNA and protein can be used to interrogate the expression level as the material is abundant. Either one can be chosen to cross-check the expression in the original tumor, depending on the availability and type of the stored tissue, and availability of antibody for the corresponding protein detection. The PDX model is specifically invaluable in testing combination therapies as the in vivo setting allows the monitoring of adverse effects, as well as dosage or duration adjustments in the treatment regimen.
The choice between orthotopic and subcutaneous engraftment can be made depending on the specific questions being addressed in the study. However, it is important to keep in mind that the sensitivity of xenografted tumors to therapeutics may be modulated by the site of implantation46. On the other hand, no evidence has been reported yet on the discovery of drugs showing a therapeutic response in orthotopic models but absent in subcutaneously implanted PDX9.
The authors have nothing to disclose.
We thank the members of the Mayo Clinic Center for Individualized Medicine (CIM) Dr. Lin Yang and Faye R. Harris, MS, for the help in conducting experiments. This work was supported by Mr. and Mrs. Neil E. Eckles’ Gift to the Mayo Clinic Center for Individualized Medicine (CIM).
3M Vetbond | 3M, Co. | 1469SB | |
anti-AKT antibody | Cell Signaling Technologies, Inc. | 9272 | |
Anti-GAPDH antibody(G-9) | Santa Cruz Biotech. Inc. | sc-365062 | |
Anti-MAPK antibody | Cell Signaling Technologies, Inc. | 9926 | |
Anti-phospho-AKT antibody | Cell Signaling Technologies, Inc. | 9271 | |
Anti-mTOR antibody | Cell Signaling Technologies, Inc. | 2972 | |
Anti-Phospho-mTOR antibody | Cell Signaling Technologies, Inc. | 2971 | |
Anti-Phospho-S6 antibody | Cell Signaling Technologies, Inc. | 4858 | |
Anti-Rictor antibody | Cell Signaling Technologies, Inc. | 2114 | |
Anti-S6 antibody | Cell Signaling Technologies, Inc. | 2217 | |
Captisol | ChemScene, Inc. | cs-0731 | |
Carboplatin | NOVAPLUS, Inc. | 61703-360-18 | |
DMEM | Mediatech, Inc. | 10-013-CV | |
Easy-A Hi-Fi PCR Cloning Enzyme | Agilent, Inc. | 600404-51 | |
Lubricant | Cardinal Healthcare | 82-280 | |
Matrigel | Corning, Inc. | 356234 | |
McCoy's media | Mediatech, Inc. | 10-050-CV | |
MK-2206 | ApexBio, Inc. | A3010 | |
MK-8669 | ARIAD Pharmaceuticals, Inc. | AP23573 | |
Nair Sensitive Skin | Church & Dwight Co. | Nair Hair Remover Shower Power Sensitive | |
NOD/SCID mice | Charles River, Inc. | NOD.CB17-Prkdcscid/NCrCrl | |
Paclitaxel | NOVAPLUS, Inc. | 55390-304-05 | |
PEG400 | Millipore Sigma, Inc. | 88440-250ML-F | |
Perjeta | Genetech, Co. | Pertuzumab | |
Rituximab | Genetech, Co. | Rituxan | |
RPMI1640 | Mediatech, Inc. | 10-040-CV | |
SCID mice | Harlan Laboratories, Inc. | C.B.-17/IcrHsd-PrkdcscidLystbg | |
SLAx 13-6MHz linear transducer | FUJIFILM SonoSite, Inc | HFL38xp | |
SonoSite S-series Ultrasound machine | FUJIFILM SonoSite, Inc | SonoSite SII | |
Tween 80 | Millipore Sigma, Inc. | P4780-100ML |