Figures
Abstract
Mouse models have been developed to investigate colorectal cancer etiology and evaluate new anti-cancer therapies. While genetically engineered and carcinogen-induced mouse models have provided important information with regard to the mechanisms underlying the oncogenic process, tumor xenograft models remain the standard for the evaluation of new chemotherapy and targeted drug treatments for clinical use. However, it remains unclear to what extent explanted colorectal tumor tissues retain inherent pathological features over time. In this study, we have generated a panel of 27 patient-derived colorectal cancer explants (PDCCEs) by direct transplantation of human colorectal cancer tissues into NOD-SCID mice. Using this panel, we performed a comparison of histology, gene expression and mutation status between PDCCEs and the original human tissues from which they were derived. Our findings demonstrate that PDCCEs maintain key histological features, basic gene expression patterns and KRAS/BRAF mutation status through multiple passages. Altogether, these findings suggest that PDCCEs maintain similarity to the patient tumor from which they are derived and may have the potential to serve as a reliable preclinical model that can be incorporated into future strategies to optimize individual therapy for patients with colorectal cancer.
Citation: Uronis JM, Osada T, McCall S, Yang XY, Mantyh C, Morse MA, et al. (2012) Histological and Molecular Evaluation of Patient-Derived Colorectal Cancer Explants. PLoS ONE 7(6): e38422. https://doi.org/10.1371/journal.pone.0038422
Editor: Alana L. Welm, Huntsman Cancer Institute, University of Utah, United States of America
Received: March 8, 2012; Accepted: May 9, 2012; Published: June 4, 2012
Copyright: © 2012 Uronis et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: This work was supported by grants from the Mentored Research Scholars Grant (119824-MRSG-10-195-01-TBG) from the American Cancer Society (www.cancer.org). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Introduction
Colorectal cancer (CRC) is the third most common cancer and the second leading cause of cancer death in the United States. In 2010, approximately 142,000 people were diagnosed with CRC, and about 40% of these patients presented with advanced disease [1]. Treatment for advanced CRC with chemotherapy is typically intended for disease control and palliation of symptoms only, and as a result, unresectable CRC remains an incurable disease. In order to improve clinical outcomes and develop new therapeutic approaches, the development of a reliable preclinical model to study CRC biology and drug sensitivities is required.
Mouse models of CRC remain one of the most useful tools to decipher the biological mechanisms underlying the oncogenic process. To date, a variety of genetically-engineered, carcinogen-induced and xenograft mouse models have been established [2], [3] and it is generally agreed that no one model is sufficient to elucidate all aspects of CRC etiology.
Genetically engineered mouse (GEM) models have been invaluable in establishing the role of many different genetic mutations and signal transduction pathways contributing to the oncogenic process and allow investigation in the context of an active immune system [2], [3]. However, many of these GEM models, primarily those involving mutation of the APC tumor suppressor gene, develop tumors in the small intestine rather than the colon. This makes longitudinal disease progression studies difficult in addition to lacking the genetic complexity observed in human cancers [2], [3].
Another widely used mouse models of CRC relies on the use of carcinogens to induce colorectal tumor development. Perhaps the most widely used carcinogen-based model is the Azoxymethane (AOM) model. Here, colorectal tumor development is initiated by AOM, a potent, colon-specific carcinogen through the formation of DNA adducts [4]. Colorectal tumors derived using this model recapitulate key human pathological features observed in humans and allow investigation of the early stages of CRC. However tumor initiation and development is a time consuming process, often taking up to 6 months with tumor multiplicity and penetrance depending heavily on the mouse strain [2], [5], [6].
While GEM and carcinogen-based models have significantly enhanced our knowledge of the genetics and etiology of CRC, these models do not allow for accurate testing of cancer therapeutics to be used in the clinical setting [7]. The most widely utilized in vivo model for the testing of anti-cancer drug efficacy and combinations is the xenograft model. Historically, xenografts have been established through the subcutaneous injection of genetically-defined human-derived cell lines into immune-compromised nude mice [8]. However, to date, the majority of these cell line-based xenograft models have failed to generate drug sensitivity data that translates into clinically relevant information [7]. In addition, recent reports suggest that tumor-stroma interactions not present in cell line-based xenografts may represent an integral component in oncogenic potential and tumor drug response [9], [10]. Therefore, more recently, whole-tissue explants derived from human cancers including breast [11], lung [12], prostate [13] and colorectal cancer [14]–[16] have been established in an attempt to generate more clinically accurate and reliable xenograft models. However, these studies examined mainly early passage explants (<5 generations) from predominantly primary tumors and therefore there remains the need to further characterize these models and evaluate how well they retain important characteristics of the original human tumor especially in metastatic disease.
H&E stained sections of two independent well-differentiated adenocarcinomas (CRC039 and CRC075) show that tumor architecture remains similar after 11 passages in NOD/SCID mice. Images shown are at 20× magnification.
A. Representative PDCCE (CRC039) retains nuclear CDX2 expression after 11 generations in mice. Images shown are at 20× magnification. B. Early passage PDCCEs retain signet ring morphology observed in original patient colorectal tumor. Images shown are at 40× magnification. C. Xenografts generated from WiDr and HT29 CRC cell lines lack histological features consistent with patient-derived explants including the presence of stroma and the formation of glands. Images shown are at 20× magnification.
In this study, we have performed a more comprehensive molecular and histological analysis of a panel of 27 matched patient-derived colorectal cancer explants (PDCCEs) from both primary and metastatic sites as an extension of our previous work [17] in which we compared the gene expression profile of 14 matched PDCCEs and their corresponding human tumors. We now demonstrate that PDCCEs retain global gene expression patterns, oncogene mutation status and histological parameters present in the original human cancers. Altogether these findings suggest that PDCCEs have the potential to serve as a reliable preclinical model that can be used to develop and characterize new therapeutic targets for patients with CRC.
Materials and Methods
Tumor Samples/Ethics Statement
A total of 27 human samples were obtained for genomic and histological analysis. All patients provided written consent to have tissue stored and used for research. Samples used for analysis in the laboratory were de-identified and not linked with any personal health information (PHI). All parts of this study were approved by the Duke Institutional Review Board. All animal studies were performed under a Duke University Institutional Animal Care and Use Committee (IACUC) approved protocol.
Unsupervised cluster analysis of 27 patient tumor-PDCCE matched pairs show that 22 pairs (81%) fell within the same cluster and 18 matched PDCCE (66%) clustered directly in pairs with the original patient tumor (gray boxes). Sample names containing an X denote PDCCE (xenograft) samples. The number immediately following the X indicates the generation/passage number of that particular sample.
Generation of Patient-Derived Colorectal Cancer Explants (PDCCEs)
Colorectal tumors (both primary and metastatic) at time of surgery were collected under a Duke IRB approved protocol (Pro00002435). The tissues were washed with phosphate buffered saline (PBS) and then minced into pieces approximately ∼2 mm in size and injected into the flanks of 4-week-old NOD.CB17-PrkdcSCID-J mice obtained from Jackson Laboratories under a Duke IACUC approved protocol. Mice were observed and tumors measured with vernier calipers until the volume of the tumor ((V = L×2W×0.52 (L = longest diameter, W = shortest diameter)) reached ∼1,000 mm3. Tumors were then harvested, minced and re-implanted as described above until stable PDCCEs were established. At each generation, tumors were harvested and either fixed in 10% neutral buffered formalin (NBF), snap frozen in liquid nitrogen or frozen in optimal cutting temperature (OCT) medium on dry ice for further analysis.
Histological Preparation and Examination
Paraffin-embedded PDCCE tissues were sectioned in 6 µm intervals and stained with hemotoxylin and eosin (H&E). Each sample was evaluated by a trained pathologist for the following histological criteria: histologic type, CDX-2 positivity, and relative percentage of tumor, necrosis, stroma, tumor gland formation and CDX-2 positive nuclei. All tissues were examined using >10 high-powered fields per section. Tumor nuclei were evaluated for CDX-2 staining using a standard quantitative scale of 0, 1+, 2+ and 3+. Staining of tumor nuclei at 2+ and 3+ was considered positive and all cases considered positive exhibited at least 20% of tumor nuclei with staining.
Oncogene Mutation Analysis
Genomic DNA was isolated from snap frozen PDCCE tissues using a Qiagen genomic DNA isolation kit. Samples were diluted to 10 ng/µl and PCR was performed using the following primers for KRAS: forward 5′ GTGTGACATGTTCTAATATAGTCA 3′; reverse 5′ GAATGGTCCTGCACCAGTAA 3′ and BRAF: forward 5′ TCATAATGCTTGCTCTGATAGGA 3′; reverse 5′ GGCCAAAAATTTAATCAGTGGA 3′. Amplicons were sequenced by conventional methods using the forward primers.
Microarray Analysis
RNA was isolated from snap-frozen PDCCE tissues using a Qiagen RNA/DNA Allprep kit, converted to cDNA and labeled by one cycle IVT. IVT labeled cDNAs were prepared according to the manufacturer’s instructions, and targets hybridized to the Human U133A 2.0 GeneChip and read on an Affymetrix array scanner. Raw data was converted to. CEL files and RMA normalized. CEL files (GSE35144) are available at the Gene Expression Omnibus (GEO) data repository (http://www.ncbi.nlm.nih.gov/geo/). To check for sample outliers and batch effects, 3D principal components analysis of the global gene expression was performed. Batch effects were normalized using the ComBat algorithm (http://jlab.byu.edu/ComBat/) [18]. Unsupervised hierarchical clustering of the human tumors and matching PDCCEs was performed on the 20% of genes with the greatest coefficient of variation. Agglomerative clusters were generated using the pearson correlation coefficient and complete linkage using the R program (The R Foundation for Statistical Computing).
Software Used for Analysis
The R statistical software package is available at www.r-project.org. The Bioconductor R package is available at www.bioconductor.org. ComBat is available as an R script at http://jlab.byu.edu/ComBat/. Graphpad Prism is a product of Graphpad Software (La Jolla, CA, USA) and is available at www.graphpad.com/prism/prism.htm.
Results
Histological Evaluation of PDCCEs
A panel of 27 patient-derived colorectal cancer explants (PDCCEs) by direct transplantation of human colorectal cancer (CRC) tissues into NOD-SCID mice was created in this study. Table 1 shows the origin of the patient tumor and a total of 5 primary PDCCEs and 22 metastatic PDCCEs were generated. To assess the extent to which in vivo models of patient-derived colorectal cancer explants (PDCEEs) accurately recapitulate and can therefore serve as a model of the human condition, we investigated whether PDCCEs retain key biological features inherent to individual human colorectal cancers (CRC) over time. First, to evaluate the extent to which histological parameters are retained after xeno-transplantation, two independent PDCCEs were passaged through >10 generations and evaluated histologically. Both PDCCEs examined exhibited pathological features remarkably consistent with the original patient tumor through 11 generations (Figure 1). Next, a comprehensive histological evaluation performed on a sub-panel of 15 matched PDCCEs and original banked tissues revealed that 15/15 PDCCEs retained pathological features similar to those observed in the matched human tumor and were characterized as histologically identical to their matched original banked sample (Table 2). Even after 11 generations, PDCCEs retained the ability to form glands and contained CDX-2 positive nuclei comparable to the first generation PDCCEs (Figure 2). These data demonstrate that the histological features present in colorectal cancer, including the formation of glands and presence of stromal components are retained even in late passage explants, suggesting that unlike CRC cell line-derived xenografts, the PDCCE model provides us with a research tool that recapitulates the human condition generally not observed in other models.
PDCCEs Retain Basic Global Gene Expression Profiles Inherent to Human Colorectal Cancers
Next, to further evaluate the extent to which PDCCEs represent their primary human counterparts, we analyzed 27 matched patient tumors and PDCCEs by microarray analysis. Patient tumor and PDCCE gene expression data was first normalized using ComBat to minimize batch effects. Unsupervised hierarchical clustering analysis was then performed on the normalized data set and revealed three distinct clusters (Figure 3). Of the 27 matched patient tumor and PDCCEs, 22 pairs (81%) fell within the same cluster based on the dendrogram and 18 PDCCEs (66%) clustered directly with the original tumor sample. Altogether, these data suggest that basic global gene-expression patterns are preserved between PDCCEs and their original human counterparts.
Oncogene Mutation Status is Retained in PDCCEs
For patients with advanced colorectal cancer, the testing of mutation status of oncogenes such as KRAS is required for guiding therapy. Specifically, patients with KRAS mutations show no benefit from treatment with EGFR inhibitors such as cetuximab or panitumumab, while patients whose tumors are KRAS WT derive benefit from anti-EGFR based therapies [19], [20]. To determine whether these clinically-significant genomic parameters are maintained in PDCCEs, 27 matched PDCCEs and original patient samples were analyzed for KRAS and BRAF mutation status. Of the 27 matched pairs evaluated, 13 presented with activating KRAS mutations (codon 12 = 11; Codon 13 = 2) (Table 3). Of these 27 matched pairs, 26/27 PDCCEs (96%) matched their original human counterpart suggesting that human colorectal cancer tissues maintained as mouse PDCCEs are genetically stable and retain oncogenic mutation status critical to CRC pathophysiology. All samples tested negative for BRAF mutations. Altogether, these data suggest that PDCCEs maintain the biologically complex histological, gene expression and mutation-based characteristics observed in human CRCs.
Discussion
To date, a number of mouse xenograft models have been established to investigate CRC etiology and treatment. To a large extent, these models have been generated using late passage cell lines derived from human CRCs and while significant treatment-induced tumor responses have been observed in these models, they are rarely predictive of tumor response in human patients [7]. This is likely due in these models, at least in part, to the inherent lack of stroma in tumor-derived epithelial cell lines. Mounting evidence indicates that paracrine signaling and extracellular matrix components supplied by neighboring stromal cells play a significant role in the oncogenic potential of colorectal carcinoma and that modulation of these stromal interactions directly impact the efficacy of chemotherapy on tumor response [9], [10].
Recent attempts have been made to generate mouse xenograft models of CRC by direct transplantation of human colon tumors into immune-compromised mice [14]–[16], [21]. Poupon and coworkers reported that the passage of human colon cancer tissues through a xenograft stage significantly improves the success rate of cell line derivation from human CRC metastases [15]. More recently, Hohenadl and coworkers reported that histological characteristics and oncogene expression levels are retained in early passage CRC xenografts [21] while Fichtner et al., and Messersmith et al, used panels of 15- and 10 human CRC explants respectively to evaluate drug sensitivity [14], [16]. Additionally, these studies used patient-derived explants mainly from the primary site and as metastatic tumors tend to be more aggressive and are more likely to differentiate, it remains unclear if PDCCEs generated from metastatic sites would maintain similarity to the original tumor. While these studies have begun to underscore the value of explant models in CRC research, a more comprehensive histological and molecular analysis on a larger panel of human CRC explants are needed to justify their use as a preclinical model to perform accurate drug efficacy analysis and predictive biomarker identification.
In this study, we demonstrate that PDCCEs generated from human adenocarcinomas with varying histological features each retain the parameters of the tumor from which they were derived at the histological, global RNA expression and oncogene mutation levels. Despite the existence of differences in the percentage of tumor stroma present between the original human tissues and those xenografted into mice, our study focuses on the malignant epithelial cells. First, the histological architecture inherent to colorectal adenocarcinoma, primarily the ability to form dysplastic glands as well as the presence of CDX-2 positive nuclei is maintained in the PDCCEs throughout multiple passages (>10). Next, we compared the gene expression profiles between matched PDCCEs and its corresponding patient tumor and observed that 18/27 (66%) of the samples clustered directly together and 22/27 (81%) clustered within the major cluster as defined by the dendrogram.
We speculate that the 9 PDCCE samples that did not cluster directly with their corresponding original tumor may have been due to the inherent heterogeneous nature of CRC. It is plausible that the original CRC tumor samples corresponding to these 9 PDCCEs harbored small sub-populations bearing additional oncogenic events. This would in turn confer a growth advantage to these populations after being transplanted into the mouse, causing the PDCCE to have a different genetic composition than the original tissue from which it was derived. In support of this notion, it appears that most variation between the primary tumor and its PDCCE occur in early PDCCE passages and that less variation occurs through the process of passaging. For example, PDCCE CRC105 clustered with the original patient sample at PDCCE passages 1 and 11 while PDCCE CRC149 did not cluster with its original sample at either passage 1 or 5 suggesting that genetic changes occur predominantly in early passages and are maintained through later passages. It is also possible, that in these 9 samples, there may have been a greater stromal contamination resulting in a difference in their clustering pattern. These results suggest that there are indeed intrinsic differences between the matched patient tumor and PDCCE and extrapolations drawn from these models must be done so carefully. However, overall our findings suggest that the PDCCE model has the potential to be used in the investigation of new therapeutic agents that target both the malignant epithelial tumor architecture and/or stromal component and that PDCCEs can be maintained for 10 or more generations while retaining key histological parameters.
Finally, we evaluated the KRAS and BRAF mutation status of the PDCCEs and showed that the status of all but one of the oncogene mutations was retained. We observed one case (CRC020) in which a KRAS activating mutation was present in the PDCCE but was not detected in the original patient samples despite the fact that these samples clustered together by unsupervised cluster analysis. It is most likely that a small, undetectable population of KRAS mutant cells was present in the patients tumor at the time of surgical resection and that the growth advantage conferred by KRAS activation allowed for subsequent expansion of KRAS mutant cells during early PDCCE passages.
Although any single mouse model will never fully recapitulate actual findings in patients, the use of preclinical models is necessary and practical for the development of therapeutic agents and biomarkers and a crucial first step in bringing these agents to the clinic. We do realize the limitations of our model and that any finding must undergo rigorous testing to gauge its accuracy, reliability, and reproducibility and must also be retrospectively validated in multiple patient samples. Nevertheless we feel that our preclinical mouse model has the potential to be used to identify and test novel combinations of therapeutic agents and to also develop both predictive and prognostic biomarkers, which can then be systematically brought forth into the clinical setting.
Acknowledgments
The authors wish to thank the Duke microarray core facility for collecting the microarray data.
Author Contributions
Conceived and designed the experiments: TO CM MM HKL BC DH. Performed the experiments: JU SM XY. Analyzed the data: JU SM DH. Contributed reagents/materials/analysis tools: SM CM MM HKL BC DH. Wrote the paper: JU DH.
References
- 1. Mnejja M, Hammami B, Bougacha L, Jemal L, Chakroun N, et al. (2011) Internal jugular vein thrombosis by protein C activated resistance. Eur Ann Otorhinolaryngol Head Neck Dis 128: 269–271.
- 2. Uronis JM, Threadgill DW (2009) Murine models of colorectal cancer. Mamm Genome 20: 261–268.
- 3. Richmond A, Su Y (2008) Mouse xenograft models vs GEM models for human cancer therapeutics. Dis Model Mech 1: 78–82.
- 4. Druckrey H, Preussmann R, Matzkies F, Ivankovic S (1967) [Selective production of intestinal cancer in rats by 1,2-dimethylhydrazine]. Naturwissenschaften 54: 285–286.
- 5. Papanikolaou A, Wang QS, Delker DA, Rosenberg DW (1998) Azoxymethane-induced colon tumors and aberrant crypt foci in mice of different genetic susceptibility. Cancer Lett 130: 29–34.
- 6. Kaiser S, Park YK, Franklin JL, Halberg RB, Yu M, et al. (2007) Transcriptional recapitulation and subversion of embryonic colon development by mouse colon tumor models and human colon cancer. Genome Biol 8: R131.
- 7. Kerbel RS (2003) Human tumor xenografts as predictive preclinical models for anticancer drug activity in humans: better than commonly perceived-but they can be improved. Cancer Biol Ther 2: S134–139.
- 8. Kendall SD, Adam SJ, Counter CM (2006) Genetically engineered human cancer models utilizing mammalian transgene expression. Cell Cycle 5: 1074–1079.
- 9. Loeffler M, Kruger JA, Niethammer AG, Reisfeld RA (2006) Targeting tumor-associated fibroblasts improves cancer chemotherapy by increasing intratumoral drug uptake. J Clin Invest 116: 1955–1962.
- 10. Harris LG, Samant RS, Shevde LA (2011) Hedgehog signaling: networking to nurture a promalignant tumor microenvironment. Mol Cancer Res 9: 1165–1174.
- 11. Marangoni E, Vincent-Salomon A, Auger N, Degeorges A, Assayag F, et al. (2007) A new model of patient tumor-derived breast cancer xenografts for preclinical assays. Clin Cancer Res 13: 3989–3998.
- 12. Fichtner I, Rolff J, Soong R, Hoffmann J, Hammer S, et al. (2008) Establishment of patient-derived non-small cell lung cancer xenografts as models for the identification of predictive biomarkers. Clin Cancer Res 14: 6456–6468.
- 13. Pretlow TG, Wolman SR, Micale MA, Pelley RJ, Kursh ED, et al. (1993) Xenografts of primary human prostatic carcinoma. J Natl Cancer Inst 85: 394–398.
- 14. Arcaroli JJ, Touban BM, Tan AC, Varella-Garcia M, Powell RW, et al. (2010) Gene array and fluorescence in situ hybridization biomarkers of activity of saracatinib (AZD0530), a Src inhibitor, in a preclinical model of colorectal cancer. Clin Cancer Res 16: 4165–4177.
- 15. Dangles-Marie V, Pocard M, Richon S, Weiswald LB, Assayag F, et al. (2007) Establishment of human colon cancer cell lines from fresh tumors versus xenografts: comparison of success rate and cell line features. Cancer Res 67: 398–407.
- 16. Fichtner I, Slisow W, Gill J, Becker M, Elbe B, et al. (2004) Anticancer drug response and expression of molecular markers in early-passage xenotransplanted colon carcinomas. Eur J Cancer 40: 298–307.
- 17.
Kim M, Osada T, Barry WT, Yang X, Freedman JA, et al. (2012) Characterization of an Oxaliplatin Sensitivity Predictor in a preclinical Murine Model of Colorectal Cancer. Molecular Cancer Therapeutics in press.
- 18. Johnson WE, Li C, Rabinovic A (2007) Adjusting batch effects in microarray expression data using empirical Bayes methods. Biostatistics 8: 118–127.
- 19. Adelstein BA, Dobbins TA, Harris CA, Marschner IC, Ward RL (2011) A systematic review and meta-analysis of KRAS status as the determinant of response to anti-EGFR antibodies and the impact of partner chemotherapy in metastatic colorectal cancer. Eur J Cancer 47: 1343–1354.
- 20. Douillard JY, Siena S, Cassidy J, Tabernero J, Burkes R, et al. (2010) Randomized, phase III trial of panitumumab with infusional fluorouracil, leucovorin, and oxaliplatin (FOLFOX4) versus FOLFOX4 alone as first-line treatment in patients with previously untreated metastatic colorectal cancer: the PRIME study. J Clin Oncol 28: 4697–4705.
- 21.
Mischek D, Steinborn R, Petznek H, Bichler C, Zatloukal K, et al. (2009) Molecularly characterised xenograft tumour mouse models: valuable tools for evaluation of new therapeutic strategies for secondary liver cancers. J Biomed Biotechnol 2009: 437284. 437284 p.