To improve treatment outcomes in non–small cell lung cancer (NSCLC), preclinical models that can better predict individual patient response to novel therapies are urgently needed. Using freshly resected tumor tissue, we describe an optimized ex vivo explant culture model that enables concurrent evaluation of NSCLC response to therapy while maintaining the tumor microenvironment. We found that approximately 70% of primary NSCLC specimens were amenable to explant culture with tissue integrity intact for up to 72 hours. Variations in cisplatin sensitivity were noted with approximately 50% of cases responding ex vivo. Notably, explant responses to cisplatin correlated significantly with patient survival (P = 0.006) irrespective of tumor stage. In explant tissue, cisplatin-resistant tumors excluded platinum ions from tumor areas in contrast to cisplatin-sensitive tumors. Intact TP53 did not predict cisplatin sensitivity, but a positive correlation was observed between cisplatin sensitivity and TP53 mutation status (P = 0.003). Treatment of NSCLC explants with the targeted agent TRAIL revealed differential sensitivity with the majority of tumors resistant to single-agent or cisplatin combination therapy. Overall, our results validated a rapid, reproducible, and low-cost platform for assessing drug responses in patient tumors ex vivo, thereby enabling preclinical testing of novel drugs and helping stratify patients using biomarker evaluation. Cancer Res; 77(8); 2029–39. ©2017 AACR.
Non–small cell lung cancer (NSCLC) is a leading cause of cancer death worldwide. Patients with stage I–III tumors are surgically resected and given adjuvant chemotherapy or radiotherapy. Patients with stage IV disease receive palliative chemotherapy only unless they can be stratified for targeted therapy. Most patients receive combination chemotherapy based on clinical parameters of cisplatin or carboplatin with at least one other drug such as vinorelbine, gemcitabine, or paclitaxel. Unfortunately, only approximately 5% of patients receiving adjuvant therapy show 5-year average survival benefit (1, 2). Therefore, more accurate methods for predicting chemotherapeutic benefit are urgently required to improve clinical outcomes.
The era of personalized medicine has heralded the development of targeted therapies for NSCLC, some of which rely on preselection of cancers according to genetic mutation. For example, selective EGFR inhibitors gefitinib and erlotinib provide clinical benefit over standard chemotherapy for NSCLC tumors bearing EGFR mutations (3, 4), whereas the ALK inhibitor crizotinib benefits ALK-mutated cases (5). A global industry is centered on assessing additional mono- or combinatorial treatments in NSCLC clinical trials. Despite this momentum, late-stage failures are a reality and there is less than 11% success in bringing a drug to market (6), attributable in part to nonpredictive preclinical drug platforms (7, 8). The incorporation of patient-derived xenograft (PDX) mouse models (9, 10) into preclinical studies has improved predictive accuracy somewhat (11, 12). However, PDX efficacy studies are expensive, requiring large numbers of mice. Furthermore, not all primary human tumors generate PDXs and, of those that do, serial propagation can select tumors that adapt to grow in an immunocompromised environment.
An alternative approach is to use 3-dimensional ex vivo culture of fresh human tumors. Methods for ex vivo culture of human tumors have been available for many years, and evidence shows that they can reliably reflect tumor growth in vivo (13–19). Here, we have developed and perfected an ex vivo culture method for NSCLC tumor samples that is both simple and reproducible. We have optimized culture conditions and show that tumor and stroma are retained intact and are viable. As proof-of-concept, NSCLC explant response to the standard-of-care chemotherapy drug cisplatin was examined, as well as response to the targeted agent TRAIL. We also illustrate how explants can be used to inform mechanisms of drug action by evaluating biomarkers of drug response. Together, our data show that the explant platform can effectively predict patient response to therapy and can be used for monitoring clinically relevant biomarkers.
Materials and Methods
Ex vivo explant culture
Fresh NSCLC tumors were collected from consented patients undergoing lung surgery (Ethical approval: LREC: 07/MRE08/07). Patients had no prior exposure to chemotherapy. Viable tumor areas were identified by frozen tissue sectioning and hematoxylin and eosin (H&E) staining. Tissue was placed in Hank balanced salt solution and cut into fragments of 2 to 3 mm3 using 2 skin graft blades on a dental wax surface. These were placed in fresh culture media (DMEM with 4.5 g/L glucose + 0%–5% FCS and 1% pen/strep); 9 fragments were randomly selected and placed on a 0.4-μm culture insert disc (Millipore) floated on 1.5 mL of media in a 6-well dish. Explants were incubated at 37°C and 5% CO2 for 16 to 20 hours. Discs were then transferred to new wells containing 1.5-mL fresh media, and drugs or carrier control were added to each well in a volume of 1.5 μL for 24 hours. Cisplatin (Sigma) was utilized over a dose range of 0 to 50 μmol/L (dissolved in dimethylformamide). TRAIL (20, 21) was utilized at 1 μg/mL, diluted in DMEM media from a stock of 1 mg/mL. After treatment, explants were washed with PBS and transferred to new wells containing 1 mL of 4% (w/v) paraformaldehyde for 20 hours. Explants were transferred onto sponges, pre-soaked in 70% (v/v) ethanol, and placed in histology cassettes. They were embedded into paraffin blocks from which 4-μm sections were generated.
H&E staining of formalin-fixed, paraffin-embedded (FFPE) material sections was generated by standard approaches and, for immunohistochemistry (IHC), sections were processed as described (22). The Novolink Polymer Detection System Kit (Leica Microsystems) was used according to the manufacturer's instructions. Primary antibodies were: cleaved PARP [E51]: Abcam 1:6,000, Ki67 Clone MIB-1: DAKO 1:2,000, p53 DO1: gift from David Lane 1:1,000, cytokeratin clone MNF116: DAKO 1:5,000. Antibodies were diluted in blocking solution made with 3% (w/v) BSA, 0.1% (v/v) Triton X-100 (Fisons) in TBS. Staining was visualized under a LEICA DM 2500 microscope and photographed with a LEICA DFC 420 camera.
Quantitation of IHC staining
Images of the tumor explants were taken at 10× magnification and merged using Adobe Photoshop CS5.1, generating a single image of one explant. Tumor area was determined using ImageJ analysis (23), excluding areas of necrosis and stroma. The labeling index was determined using ImmunoRatio (24), and a single value was obtained for all 9 explants derived from one treatment that was expressed as a percentage of the total tumor area.
Laser ablation inductively coupled plasma mass spectrometry
Sections of explants treated with cisplatin were subjected to Laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS) to produce elemental maps showing the spatial distribution of platinum in tissue sections (25). The method is described in Supplementary Fig. S1.
Significance of proliferation/death indices was determined by Wilcoxon matched pairs test and Jonckheere–Terpstra trend test, respectively. Unpaired data were compared by the Mann–Whitney U test or Kruskal–Wallis one-way ANOVA. Paired data were analyzed by the Page L test (Unistat Statistical Package, version 5.0, Unistat), and interrelationships were investigated by Spearman rank correlation (SPSS, version 22, IBM). The optimal cutoff point to determine the relationship between explant response and patient survival was examined using a plot of sensitivity against 1 − specificity as a receiver operator characteristic (ROC) curve (SPSS). Survival was investigated by Kaplan–Meier analysis (SPSS) of cell indices, which were compared by the log-rank Mantel–Haenszel (Peto) test, and by univariate and multivariate Cox regression (SPSS). P < 0.05 was considered statistically significant.
Histopathology of NSCLC tumors used for explants
Table 1 provides a summary of patient demographics, tumor type, and stage for all 45 samples utilized for this study. The histologic types and stages were broadly consistent with the known distribution of NSCLC cases in the United Kingdom (26). A proportion of NSCLC tumors is known to be necrotic (27), and an important first step was to exclude such tumors from analysis using H&E assessment. This led to the identification of 13 tumors (∼29%) that were excluded from explant generation. Supplementary Fig. S2 indicates the histologic type (Supplementary Fig. S2A) and stage (Supplementary Fig. S2B) of viable and nonviable tumors. The highest proportion of nonviable tumors was within the adenocarcinoma subtype [∼36% compared with ∼25% of squamous cell carcinoma (SCC) cases]. However, there was no correlation between tumor stage and viability.
Thirty-two viable tumors were processed for explant culture. Intrinsic levels of cell proliferation and cell death were first assessed in uncultured samples by Ki67 or cleaved PARP (cPARP) immunostaining (Fig. 1A and B). SCC tumors displayed significantly higher levels of proliferation than adenocarcinoma, whereas atypical carcinoid tumors were essentially indolent (Fig. 1A). These observations are consistent with several previous reports (28–30). With regard to cPARP staining, the majority of samples showed less than 20% of staining, indicating low levels of intrinsic cell death (Fig. 1A and B).
Optimization of explant culture
Our approach for the NSCLC explant culture system was based on previous experience with breast cancer samples (MacFarlane, unpublished observations; ref. 31;). As a first step in implementing protocols for NSCLC, we first investigated the effects of culture time and FCS concentration.
Explants were routinely allowed to recover for a period of 16 to 20 hours after their initial generation; viability was assessed over a time range of 24 to 72 hours after recovery for 5 tumors. As shown in Fig. 2A, a trend of decreasing cell proliferation and increasing cell death with increasing time of culture was observed, suggesting ex vivo explant cultures are more viable in short-term culture. Varying FCS concentration, from 0% to 5%, at 24 hours of culture after the initial 16 to 20 hours of culture recovery showed no statistically significant difference in levels of proliferation or cell death (Fig. 2B).
The data from above suggest that the 24-hour time point gives the greatest viability but that FCS concentration is not a significant factor. Subsequent analyses of drug responses were therefore performed for 24 hours, using 1% as the standard FCS concentration. Pooled data for 21 explants under these conditions are shown in Fig. 2C. Overall, there is a small but significant effect of cultivation on explant viability, with there being approximately 10% decrease in proliferation and almost 10% increase in cell death compared with the uncultured but freshly fixed native tumor.
Explant responses to cisplatin
Adjuvant chemotherapy for NSCLC usually consists of cisplatin often in combination with another chemotherapy drug. Clinical trial data have shown that cisplatin is the dominant drug, with the drug given in combination with cisplatin not modifying the effect of chemotherapy on overall survival (1, 32). Given this evidence, we focused on examining explant responses to cisplatin alone.
Twenty-one explants were treated with a dose range of cisplatin (0–50 μmol/L) for 24 hours following the initial recovery of 16 to 20 hours. Data for individual cases are shown in Supplementary Fig. S3. Levels of cell proliferation were only marginally affected by the drug (Supplementary Fig. S3B), and therefore the emphasis was placed on assessing cell death responses (Supplementary Fig. S3A). Cell death response for each tumor was calculated as fold induction relative to the control over the dose range (Fig. 3A). Of 21 tumors, 9 (43%) showed less than 2-fold induction in cell death in response to the drug, whereas the remaining 12 of 21 (57%) showed cell death induction ranging from 2- to 25-fold. The majority of these tumors only showed a response at high levels of cisplatin (50 μmol/L), with only 2 tumors responding at the lower dose of 10 μmol/L.
In addition to the 21 explants treated with a dose range of cisplatin, a further 9 were treated with a single dose of 50 μmol/L cisplatin. We obtained clinical and histopathologic information on all 30 patients and their tumors (Supplementary Table S1). Cell death difference compared with control in response to cisplatin is included alongside this information. One tumor was excluded from the analysis due to complex histopathology and 3 atypical carcinoids were excluded because of their different biologic behavior compared to adenocarcinomas and SCCs. For the remaining 26, a ROC curve was used to determine the threshold for resistance/sensitivity to cisplatin and this analysis gave an area under the curve of 0.6485 ± 0.1122 SE, a likelihood ratio of 3.30 and identified 28.45% as the optimal cutoff (Supplementary Fig. S4A).
We then categorized each explant into being either sensitive or resistant to cisplatin (Supplementary Fig. S4B and Supplementary Table S1). Using clinical information on corresponding patients (Supplementary Table S1), the relationship of cisplatin sensitivity/resistance in explant culture to patient survival post-surgery was determined (Fig. 3B). The data show a statistically significant relationship (P = 0.006) with sensitive cases demonstrating a mean survival time (MST) of 28 months and resistant cases an MST of 14 months. To rule out an effect of tumor stage, we separated stage I/IIA and IIB/III cases (Fig. 3C). There is a statistically significant relationship between cisplatin sensitivity and patient survival for both stage I/IIA cases (P = 0.02) and for stage IIB/III cases (P = 0.05), indicating the correlation is independent of stage; importantly, this relationship was also shown to be independent of stage in a multivariate Cox survival analysis. Of the 26 patients, 12 received adjuvant therapy (Supplementary Table S1), 8 received platinum-based chemotherapy, 3 received radiotherapy, and 1 received taxane-based chemotherapy. In those cases receiving any form of adjuvant therapy, there is a significant correlation with survival of patients and response to cisplatin in explants (P = 0.01; Fig. 3D, left). However, there was no relationship between cisplatin sensitivity in explants and survival for patients reported to have lymph node involvement (P = 0.13; Fig. 3D, right). Overall, the data show a strong relationship between patient survival and explant sensitivity to cisplatin, indicating that the explant platform is predictive of disease recurrence and response to adjuvant therapy.
Cisplatin sensitivity in explants was also correlated with tumor stage and histologic type (Fig. 3E). There was a significant negative trend between difference in percentage of cPARP staining compared with control in response to cisplatin and increasing tumor stage (P = 0.007), suggesting that more advanced tumors are more resistant to the drug. There was also a significant correlation between cisplatin sensitivity and tumor type (P = 0.0004), with SCC cases demonstrating greater cisplatin sensitivity than adenocarcinoma subtypes (Fig. 3E).
Cisplatin sensitivity is linked to drug accumulation in tumor areas
A number of mechanisms have been reported to render cells resistant to cisplatin including reduced drug uptake, enhanced export, drug deactivation, increased repair of DNA damage, or alterations in apoptosis (33, 34). To examine drug uptake/export, we investigated Pt ion distribution across explant tissue using LA-ICP-MS (see Supplementary Fig. S1) imaging (Fig. 4 and Supplementary Fig. S5). For resistant cases, Pt ions were depleted from areas corresponding to tumor cells but were present in the stroma (Fig. 4A and Supplementary Fig. S5). In contrast, for sensitive cases, Pt ions were present throughout the tumor and stromal areas of the explant, indicating widespread cisplatin uptake (Fig. 4B and Supplementary Fig. S5). Thus, while cisplatin is available to the resistant explants, there is decreased intracellular drug concentration in tumor cells.
TP53 expression in the explants
The TP53 gene is frequently mutated in NSCLC (35). Wild-type TP53 protein is induced by DNA-damaging agents such as cisplatin, whereas mutated TP53 is either not expressed or is constitutively expressed. We utilized IHC to gain an indication of TP53 function in the 30 tumors, identifying 3 categories: (i) WTTP53 tumors (12 tumors), (ii) MUTTP53 tumors with constitutively high TP53 levels (16 tumors), and (iii) MUTTP53 tumors expressing undetectable TP53 (2 tumors). IHC TP53 staining of positive and negative tumors is shown in Fig. 5A, whereas Fig. 5B indicates induction of TP53 following treatment of a WTTP53 tumor with a dose range of cisplatin and quantitation of the staining. Overall, 40% of tumors were WTTP53 and 60% MUTTP53 based on IHC criteria (Supplementary Table S1). This is approximately consistent with the known mutation rate of TP53 in human NSCLC (34).
As expected, TP53MUT tumors had significantly higher intrinsic levels of proliferation compared with TP53WT tumors (Fig. 5C), and the majority of TP53MUT cancers were of the SCC subtype (Fig. 5D). In terms of response to cisplatin, TP53MUT samples had significantly higher levels of cell death induction compared with TP53WT samples (Fig. 5D, left) and significantly higher levels of suppression of cell proliferation (Fig. 5D, right). These data counteract the view that TP53MUT tumors are defective in their apoptotic response to DNA damage induced by cisplatin.
Explant responses to TRAIL
TRAIL is a death receptor ligand that has been developed for therapy, although clinical trials have been disappointing (36). It is thought that preclinical in vitro studies using cell lines have not faithfully represented the clinical situation. This is supported by data demonstrating that the majority of primary human tumor cells are resistant to TRAIL receptor agonists (36–38). To investigate TRAIL sensitivity in NSCLC, 12 explants were treated with TRAIL either as a single agent or in combination with cisplatin (Fig. 6A). TRAIL alone did not elicit a strong response, except for one case (LT22) that demonstrated approximately 4-fold induction of cell death. Similarly, TRAIL did not enhance the effects of cisplatin in the majority of cases, except for one tumor (LT83) for which slightly greater cell death induction (6-fold) than cisplatin alone (4-fold) was detected (Fig. 6A and B).
Predicting drug response in patients with cancer is a major challenge in the clinic. Cell line xenograft mouse models have been extensively used for preclinical drug testing, but while these models can provide an initial indication of in vivo drug efficacy, data are often not predictive of patient outcome (7, 39). Although the advent of mouse PDX models has opened up the possibility of tailoring drugs to a tumor with a specific genetic lesion (11, 40), in practice, these models are expensive and lose the characteristics of the original human tumor microenvironment over time. Here, we have perfected a rapid and low-cost platform that relies on the in situ assessment of drug responses within real human tumors. We validate this platform by showing that explant response to the standard-of-care chemotherapy drug cisplatin is related to survival of patients (P = 0.006), indicating that explant response to cisplatin is predictive of disease progression. Responses to the targeted agent TRAIL are also more consistent with clinical outcomes than standard cell line model systems (36–38). We demonstrate how the explant platform can be used to inform mechanisms of drug action by biomarker monitoring.
A number of organotypic culture systems have been previously developed for human tumors (13–18). In most of these previous systems, viability of tumors has been demonstrated for up to 7 days (13–19). Here, we identified a mild effect of cultivation after 24 hours of culture (Fig. 2C), but tissue architecture was maintained intact for up to 72 hours. Our preference is to examine drug responses immediately after cultivation to minimize any effects of culture. Correlation of organotypic culture data with patient outcomes has been previously reported for the Histoculture Drug Response Assay system (15–17). However, a disadvantage of this technique is that the endpoint requires enzymatic digestion of tissue, thus preventing assessment of the specific cell type affected by the drug. This disadvantage can be overcome by using our in situ FFPE/IHC approach.
Our data show that the majority of the cisplatin-resistant tumors are of a higher stage (Fig. 3 and Supplementary Table S1), but the ability to induce cell death in response to cisplatin does not correlate with intact TP53 (Fig. 5). In fact, we have found that TP53-mutated NSCLC cases are more sensitive to cisplatin in explants thanWTTP53 cases (Fig. 5D). Previous studies have investigated whether TP53 mutations are of prognostic value in predicting response to chemotherapy in NSCLC; the results are controversial (41). In a 35-patient study, the presence of mutant TP53 was highly indicative of resistance to cisplatin (P = 0.002) (42), while in a study involving 253 patients, TP53-positive patients had a significantly greater survival benefit from adjuvant chemotherapy compared with TP53-negative patients (43). Another report in the International Adjuvant Lung Cancer Trial (IALT), a randomized trial of adjuvant cisplatin-based chemotherapy, found no correlation between TP53 mutation and outcome in 524 patients (44). Overall, it will be important to extend analysis to a greater number of explants/patients to robustly determine the prognostic value of TP53 mutation. Lack of response to cisplatin does, however, correlate with exclusion of the drug from tumor areas (Fig. 4). Cisplatin import is mediated by the copper transporter CTR1, whereas the copper transporters ATP7A and ATP7B regulate the efflux of cisplatin (45). Resistance to cisplatin has been associated with alterations in the expression status of these transporters (46) and so it will also be important to evaluate these transporters in the explant system used here.
In summary, the explant platform provides a patient-relevant model system for the preclinical evaluation of novel anticancer agents. When combined with tumor stratification approaches, the platform has the potential for personalizing drug treatment. The technology is low-cost, rapid, and achievable within an integrated cancer translational research setting. An important next step will be to conduct a clinical study aimed at determining which patients would best respond to chemotherapy prospectively; such a study is currently being developed within our center.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Conception and design: B. Sharp, C. Pritchard, M. MacFarlane, J.H. Pringle
Development of methodology: E. Karekla, B. Sharp, J. Le Quesne, C. Pritchard, M. MacFarlane, J.H. Pringle
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): E. Karekla, W.-J. Liao, J. Pugh, M. MacFarlane
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): E. Karekla, B. Sharp, J. Pugh, H. Reid, J. Le Quesne, D. Moore, C. Pritchard, M. MacFarlane, J.H. Pringle
Writing, review, and/or revision of the manuscript: E. Karekla, B. Sharp, J. Pugh, H. Reid, C. Pritchard, M. MacFarlane, J.H. Pringle
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): E. Karekla, H. Reid, J.H. Pringle
Study supervision: C. Pritchard, M. MacFarlane, J.H. Pringle
Other (histopathologicinterpretation): J. Le Quesne
This work was supported by a Medical Research Council (MRC) Doctoral Training Grant to E. Karekla, the MRC Toxicology Unit (MC A/600), and the Leicester Experimental Cancer Medicine Centre (C325/A15575 Cancer Research UK/UK Department of Health). C. Pritchard was supported by a Royal Society-Wolfson merit award.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
We thank Andrew Wardlaw for providing the ethical framework for this project; Hilary Marshall and Will Monteiro for support in tissue collection; thoracic surgeons at Glenfield Hospital, Leicester, for providing clinical samples; and Chris Baines for providing clinical data. We also thank Angie Gillies for assistance with histology.
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).
- Received April 18, 2016.
- Revision received January 10, 2017.
- Accepted January 30, 2017.
- ©2017 American Association for Cancer Research.