Establishment of Patient-Derived Non–Small Cell Lung Cancer Xenografts as Models for the Identification of Predictive Biomarkers

Materials and Methods

Patients and tissue samples

One hundred and two tumor specimens were obtained at initial surgery from primarily diagnosed, early-stage NSCLC patients between July 2004 and January 2007. Written informed consent was obtained from each patient and the study was approved by the local ethical committee. From these samples, 25 passagable xenograft models were successfully established, which were analyzed in the study described here (Table 1 ). Four of the 25 patients had received chemotherapy or radiation therapy before surgery. The histologic type was determined according to WHO criteria. The collective consisted of 12 (48%) squamous cell carcinomas, 6 (24%) adenocarcinomas, and 4 (16%) carcinomas with a pleiomorphic phenotype. The remaining tumors were diagnosed as large-cell or dedifferentiated carcinomas, and one tumor was secondarily defined as small-cell lung carcinoma. The samples originated from 12 (44%) female and 13 (56%) male patients. With one exception (SQC 7298), the patients were long-term smokers. Pathologic staging was determined according to Mountain (29). All of the tumor samples were put into medium immediately after surgical resection under sterile conditions and transported without delay to the animal facility.

Establishment of xenografts

Surgical tumor samples were cut into pieces of 3 to 4 mm and transplated within 30 min s.c. to 3 to 6 immunodeficient NOD/SCID mice (Taconic); the gender of the mice was chosen according to the donor patient. Additional tissue samples were immediately snap-frozen and stored at -80°C for genetic, genomic, and protein analyses. All animal experiments were done in accordance with the United Kingdom Co-ordinating Committee on Cancer Research regulations for the Welfare of Animals and of the German Animal Protection Law and approved by the local responsible authorities. Mice were observed daily for tumor growth. At a size of about 1 cm³, tumors were removed and passaged to naive NMRI:nu/nu mice (Charles River) for chemosensitivity testing. Tumors were passaged no more than 10 times. Numerous samples from early passages were stored in the tissue bank in liquid nitrogen and used for further experiments. Several rethawings led to successful engraftment in nude mice.

Chemosensitivity testing

The chemotherapeutic response of the passagable tumors was determined in male NMRI:nu/nu mice. For that purpose, one tumor fragment each was transplanted s.c. to a large number of mice. At palpable tumor size (50-100 mm³), 6 to 8 mice each were randomized to treatment and control groups and treatment was initiated. If not otherwise mentioned, the following drugs and treatment modalities were used: etoposide (Vepesid; Bristol-Meyers Squibb) 10 mg/kg/d, qd 1-5, i.p.; carboplatin (Mayne Pharma Deutschland) 75 mg/kg/d, qd 1 and 8, i.p.; gemcitabine (Gemzar; Lilly Deutschland) 60-80 mg/kg/d, qd 1, 4, 7, and 10, i.p.; paclitaxel (Taxol; Sigma) 12.5 mg/kg/d, qd 1-5, i.v.; vinorelbine (Navelbine; Pierre Fabre Pharma) 1-5 mg/kg/d qd 1 and 8, i.p.; cetuximab (Erbitux; Merck) 50 mg/kg/d, qd 1-5, i.p.; and erlotinib (Tarceva; Hoffmann-LaRoche) 50 mg/kg/d, qd 1-5 and 8-12, p.o. Doses and schedules were chosen according to previous experience in animal experiments and represent the maximum tolerated or efficient doses. The injection volume was 0.2 mL/20 g body weight.

Tumor size was measured in two dimensions twice weekly with a caliper-like instrument. Individual tumor volumes (V) were calculated by the formula: V = (length + [width]²) / 2 and related to the values at the first day of treatment (relative tumor volume). Median treated to control (T/C) values of relative tumor volume were used for the evaluation of each treatment modality and categorized according to scores (− to ++++; see Tables 2 and 3 ). The mean tumor doubling time of each xenograft model was calculated by comparing the size between 2- and 4-fold relative tumor volumes. Statistical analyses were done with the U test (Mann and Whitney) with P < 0.05. The body weight of mice was determined every 3 to 4 days and the change in body weight was taken as variable for tolerability.

Mutational analysis

DNA extraction. Genomic DNA was isolated from xenograft samples with the QIAamp DNA Mini Kit (Qiagen) according to the manufacturer's instructions. DNA was quantified spectrophotometrically using the ND-100 (Nanodrop).

EGFR and p53 mutation analysis. Mutations in the tyrosine kinase domain (exons 18-21) of EGFR and exons 5 to 9 of p53 were detected with partially denaturing high-performance liquid chromatography as described previously (30, 31).

DNA sequencing. PCR products were purified with 2 μL ExoSAP-IT reagent (USB Corporation) at 37°C for 15 min followed by inactivation at 80°C for 15 min. Direct sequencing using BigDye Terminator version 3.1 Cycle Sequencing Kit (Applied Biosystems) was done on the 3100xl Genetic Analyzer (Applied Biosystems) following manufacturer's instructions. All variants were sequenced bidirectionally.

K-ras mutation analysis. Samples were screened for all types of mutations in K-ras codons 12 and 13 by pyrosequencing as described previously (32).

Genome-wide gene expression analysis

RNA extraction. RNA was extracted from NSCLC specimens and from 2 × 2 × 2 mm³ tumor xenograft samples (derived from different passages ranging from 1 to 6), which were taken from the sacrificed animals. Samples were snap-frozen and stored in liquid nitrogen until use. Total RNA of homogenized tumor samples was prepared with Trizol RNA extraction reagent (Invitrogen) followed by purification using the RNeasy Mini Kit (Qiagen) according to the manufacturer's recommendations. A DNase I (Qiagen) digestion step was included to eliminate genomic DNA. The quality of the total RNA was checked for integrity using RNA LabChips on the Agilent Bioanalyzer 2100 (Agilent Technologies) and the concentration was measured on the Peqlab NanoDrop. Only RNA with a RNA integrity number larger than 6.5 was used for cDNA synthesis.

Array hybridization. The one-cycle eukaryotic target labeling assay from Affymetrix was used according to manufacturer's instructions. Briefly, 2 μg high-quality total RNA was reverse transcribed using T7-tagged oligo-dT primer in the first-strand cDNA synthesis reaction. After RNase H-mediated second-strand cDNA synthesis, the double-stranded cDNA was purified and served as template for the subsequent in vitro transcription reaction, which generates biotin-labeled cRNA. The biotinylated cRNA was cleaned up, fragmented, and hybridized to GeneChip HGU133Plus2.0 expression arrays (Affymetrix), which contain 54,675 probe sets. The GeneChips were washed and stained with streptavidin-phycoerythrin on a GeneChip Fluidics Station 450 (Affymetrix). After washing on a GeneChip Fluidics Station 450, the arrays were scanned on an Affymetrix GeneChip 3000 scanner with autoloader and barcode reader (Affymetrix).

Data analysis. The quality of the hybridized arrays was analyzed with the Expressionist Pro 4.0 Refiner (GeneData) software. Here, based on raw intensities of individual oligonucleotide features (probes), the following analyzes are done: the experiments are grouped according to similarity and potential outlier experiments can be removed (or selected for rehybridization or refragmentation), and the quality of a particular experiment is compared with a virtual reference experiment, which is computed as average of all feature intensities of all arrays in that group. Moreover, defects on the arrays are masked. Here, for each array, the spatial signal distribution is compared with the reference experiment of the experiment group it belongs to. Regions with sharp boundaries, which have consistently higher or lower feature intensities compared with the reference experiment, are flagged as defects and excluded from further analysis. In addition, a signal correction (distortion and gradient) is done, the control gene statistics are calculated, and an overall classification of the quality of the experiments is provided. The probe intensities on each array were summarized with the MAS5.0 summarization algorithm and the refined and summarized data were loaded into the CoBi database (GeneData). The analysis of the probe set-specific signal intensities was done with the Expressionist Pro 4.0 Analyst (GeneData) software. The data set was median-normalized.

Hierarchical clustering of all experiments was done based on all probe sets represented on the HGU133Plus2.0 array with a quality P < 0.04 using positive correlation and complete linkage. Gene expression of primary tumors was compared with the median arrays of replicate tumors from each xenograft model in a paired t test. Probe sets differentially expressed at a P < 10⁻⁷ were subjected to pathway analysis using Metacore software (GeneGo).

Profiling human genes expressed in human NSCLC-derived xenograft models on nude mice using the human-specific Affymetrix array HGU133Plus2.0 will mostly uncover genes specifically expressed in human cancer cells in xenograft tumors composed of a mixture of human- and mouse-derived cells. Using the human-specific Affymetrix arrays in contrast to cDNA arrays has the advantage that 11 independent probes as perfect match oligonucleotides interrogate a particular gene. Unless a murine gene is 100% homologous to a human gene, this provides very high specificity compared with a continuous stretch of DNA on a cDNA array. In addition, the presence of a set of 11 mismatch oligonucleotides provides a further degree of specificity as only those probe sets are incorporated in the determination of a condensed signal intensity for which perfect match > mismatch. From histologic analyses of lung cancer xenografts, the amount of murine tissue in a human lung cancer xenograft tumor was estimated to be as low as 5% to 10%.⁶

Real-time PCR

cDNA was reverse transcribed using SuperScript III Reverse Transcription Kit (Invitrogen) and TaqMan quantitative real-time PCR done using cDNA corresponding to 25 ng RNA/reaction. Gene- and species-specific primer/probe pairs for COL3A1 and GAPDH and TaqMan Fast MasterMix obtained from Applied Biosystems were used according to the manufacturer's instructions and amplifications were carried out on the Applied Biosystems 7500 Real-time PCR system with 40 cycles. Results from analysis done in duplicates are shown as mean 40 - average C_T value. For samples with no detectable signal, C_T values were set to 40, so the results are displayed as 0.

Immunoblotting

Lysates for immunoblotting were prepared by adding lysis buffer [150 mmol/L NaCl, 20 mmol/L Tris, 1% Triton X-100, 0.5% sodium deoxycholate, 0.5% SDS, 2 mmol/L EDTA, 2.5 mmol/L sodium pyrophosphate, 1 mmol/L β-glycerophosphate (pH 7.7)] containing protease and phosphatase inhibitors (Sigma-Aldrich) to the tumor tissue homogenized in fluid nitrogen. The protein concentration was determined using Bio-Rad Protein Assay (Bio-Rad Laboratories). Tumor lysates (30 μg) were separated on 9% SDS-PAGE polyacrylamide gels and transferred to nitrocellulose membranes. Membranes were blocked and incubated with the primary antibodies overnight at 4°C. The secondary antibody (DAKO) was conjugated with horseradish peroxidase. Protein bands were visualized using the enhanced chemiluminescence detection system (GE Health Amersham Life Science). To show equal protein loading, the blots were stripped and reprobed for β-actin. As positive controls for the detection of phosphorylated proteins, lysates of A431 cells treated with EGF were used (Biosciences PharMingen).

The following antibodies were used for immunoblotting: AKT was purchased from Stressgen Bioreagents; pAKT (Ser⁴⁷³), PTEN (clone 138G), pPTEN (clone 44A7), STAT3, and pSTAT3 (Tyr⁷⁰⁵, clone D3A7) from Cell Signaling (New England Biolabs); EGFR (clone 1005) from Santa Cruz Biotechnology; pEGFR (clone 74), ERK (clone MK12), pERK1/2, phosphotyrosine (clone PY69), STAT1 (clone 1), and pSTAT1 (Tyr⁷⁰¹, clone 4a) from Biosciences PharMingen; β-actin (clone AC-15) from Sigma; and swine anti-rabbit and rabbit anti-mouse antibodies from DAKO.

Results

Engraftment of xenografts. A total of 102 tumor specimens were transplanted to immunodeficient mice. From these, 25 passagable xenograft lines were obtained (Table 1). The successful growth in mice was not dependent on the clinicopathologic characteristics mentioned. Both stage T₂ and T₃ and adenocarcinomas (6 of 25) and squamous carcinomas (12 of 25) were successfully engrafted. The growth rate of the established xenografts (depicted in Table 2 as tumor doubling time) was a unique feature of the different models and not related to any patient characteristic or to the tumor type. A comparison of the H&E-stained original sample with xenografted tumors after few passages revealed the same histologic architecture of the primary NSCLC and the xenograft tumor (data not shown). The staining with the anti-EpCAM or Ki-67 antibodies revealed comparable pictures (not shown). In general, the tumors revealed a more homogeneous structure during murine passaging due to the replacement of human accessory cells (fibroblasts, endothelial, and immune cells) by murine tissues (data not shown). After two to four passages in mice, the composition of the xenograft tissue did not change any further. At that stage, xenograft transplants were archived in the tumor bank, if required retransplanted and used for chemosensitivity testing, mutation analysis, and RNA and protein analyses.

Chemosensitivity. To further characterize the established xenograft models, extensive therapeutic experiments with five cytotoxic agents, used also in the clinic for treatment of NSCLC, were done. Table 2 gives an overview on the results obtained and Fig. 1 presents exemplified growth curves of three models after different treatments. Paclitaxel, gemcitabine, and carboplatin proved to be the most efficient agents in NSCLC xenografts with 4 of 25, 3 of 25, or 3 of 25 xenografts responding with tumor shrinkage. Etoposide (1 of 25 complete remission) and vinorelbine (0 of 11 complete remission) were less efficient.

It became obvious that some models had a high responsiveness toward the majority of agents tested (e.g., SQC 7343 and ADC 7462), whereas others were intrinsically resistant toward a variety of drugs (e.g., PLC 7064 and ADC 7198). In general, each tumor had its individual pattern of response to the agents tested, which was independent of the tumor type (e.g., adenocarcinoma or squamous cell carcinoma). A correlation analysis comparing overall sensitivity of xenografts with their tumor doubling time revealed a higher responsiveness of models with faster growth rate (r = 0.591). Table 3 gives an overview on the responsiveness of the xenograft models to anti-EGFR therapies. Twelve of 25 models tested showed significant responses to cetuximab (3 of 25 tumor shrinkage) and 6 of 25 to erlotinib (1 of 25 tumor shrinkage). The response rates were quite different among the tumors tested, ranging from no response (∼100% T/C in models ADC 7700 and SQC 7612) to complete remissions (∼0% T/C in models ADC 7462 to both anti-EGFR therapies and DDC 7668 to cetuximab). The responses to anti-EGFR therapies were not related to the histologic type of NSCLC or clinical variables. Interestingly, there was a clear dose-response effect to erlotinib between 25 and 75 mg/kg, whereas this was not obvious for cetuximab (Fig. 1C). One model (SQC 7177) was moderately sensitive to erlotinib (T/C 28%) but resistant to cetuximab, a phenomenon whose mechanism needs further elucidation.

Mutational analysis. Sequence variants in the EGFR genes were observed in 6 of 25 NSCLC xenografts. In all these cases, the changes were either silent or represented intronic changes not considered to be functional. Mutations in K-ras and p53 were detected in 6 of 25 (24%) and 13 of 25 (52%) of models, respectively. These data are consistent with previous reports on frequencies of mutations in these genes in NSCLC (33).

None of the EGFR or p53 mutations investigated correlated alone or in combination with the response to anti-EGFR therapies. Interestingly, 5 of 6 tumors with K-ras mutations were resistant to erlotinib. However, the same K-ras mutation in codon 12 was associated with either sensitivity (ADC 7462) or resistance (PLC 7336) to anti-EGFR therapies.

Gene expression analysis. In total, RNA was isolated from 17 primary NSCLC specimens and from 2 to 5 vehicle control (PBS) tumors and hybridized to Affymetrix HGU133Plus2.0 arrays. High-quality gene expression data from the primary NSCLC specimen and the corresponding xenograft model derived thereof were available for 17 models and were used for a comparative analysis.

Unsupervised hierarchical clustering of samples (Fig. 2A ) based on the expression of all genes revealed that 9 of the 17 primary tumors clustered directly with the xenografts derived thereof. These primary samples consisted of at least 25% tumor tissue. The other 8 xenografts formed a distinct cluster from primary tumors. Five of these primary tumors (SQC 7860, ADC 7406, ADC 7700, PLC 7064, and SQC 7177) contained <10% tumor tissue. We calculated the correlation coefficient between primary cancer and the xenograft derived thereof for all 17 cases. The correlation coefficient ranged between 0.945 and 0.782 with 10 of 17 > 0.900 indicating a high degree of similarity between the primary cancer and the corresponding xenograft model. The correlation coefficients are listed in Supplementary Table S1.

In all cases the xenograft tumors of one model clustered tightly together, showing high consistency between replicate tumors. A paired t test between the primary tumors and xenografts revealed 193 differentially expressed probe sets with P < 10⁻⁷ (Supplementary Table S2). Clustering based on these 193 probe sets showed a clear distinction between primary tumors and xenografts (Fig. 2B), with most probe sets (155 of 193) being down-regulated in the xenograft models and only a very small number being up-regulated (38 of 193). The 193 probe sets correspond to 134 unique genes and main pathways represented in this data set were cell adhesion and immune response according to GeneGo Metacore pathway analysis (Supplementary Fig. S1).

We did a quantitative TaqMan PCR for one of the most strongly differentially expressed genes, the gene COL3A1, a collagen that is strongly expressed in connective tissue. We used primer/probe pairs specific for the human and the murine gene, respectively, and analyzed the gene expression level of the human and murine genes in the pairs of primary lung cancer and xenograft model derived thereof (Fig. 2C). For comparison, we measured expression levels of the housekeeping gene GAPDH. Human COL3A1 was detectable in the primary lung cancer but at much lower levels (higher CT values) in the xenograft samples. In contrast, the murine COL3A1 was strongly expressed in the xenografts but not in the primary lung cancer samples. Human GAPDH transcripts were strongly detectable in the primary tumors and the xenografts, whereas, as expected, the murine GAPDH was only expressed in xenograft samples. Moreover, results from control samples, mouse Lewis lung tumors, and human A549 lung cancer cells confirmed the species specificity of the TaqMan primer/probe pairs. In summary, this analysis shows that for the COL3A1 gene the results from global gene expression analysis are species specific; indeed, instead of human connective tissue, mouse stroma can be detected in the xenograft samples. Similar results were obtained when using primer/probe pairs specific for human and mouse COL1A2 and PECAM1 encoding for the CD31 vascular endothelial marker (data not shown).

Protein analyses. The protein expressions of EGFR, STAT1, STAT3, ERK, AKT, and PTEN as well as their corresponding phosphorylated forms were examined by immunoblotting in the 25 NSCLC xenograft models. Quantification of the immunoblots showed that the protein expression pattern of the 12 selected markers did not correlate with the response to the two EGFR-directed therapeutics. Individual, tumor-specific expression patterns were found for EGFR, ERK1, STAT1, STAT3, AKT, and the corresponding phosphorylated forms, whereas all tumors expressed similar high levels of PTEN (data not shown).

The analysis of xenograft tumors taken after 2 weeks of therapy with cetuximab or erlotinib compared with vehicle-treated controls revealed a down-regulation of the total EGFR content in 2 sensitive xenograft models (ADC 7462 and ADC 7466) of 6 analyzed. But in none of the resistant tumors investigated (exemplified for ADC 7387 and SQC 7343) a change in expression level was found (Fig. 3 ). In the xenograft model ADC 7462, additionally, a down-regulation of ERK1 after cetuximab treatment was observed. The other proteins, pEGFR, pSTAT1, pSTAT3, and pERK showed no regulation relating to sensitivity or resistance of the xenografts.

Discussion

NSCLCs are one of the most frequent tumor types, which present with a very bad prognosis and only minor therapeutic success had been made in the last years (34, 35). Therefore, an intensive search for better treatment modalities including compounds targeting tyrosine kinases was initiated. Several clinical trials with therapies directed against the EGFR have been done (IDEAL, INTACT, and TALENT) showing a differential but in general not a promising outcome (36–39).

The outcome of these studies revealed an urgent need for a better stratification of patients. Thus far, no biomarker for prediction of response or outcome has been installed for routine use (40). One of the reasons may be that clinical samples used for the analysis of biomarkers were mainly pretreatment samples, and only available in minute amounts resulting in nonsufficient material for an extended study on the protein level. Further, clinical samples are relatively heterogeneous; they contain a variable percentage of cancer cells and other infiltrating cells.

NSCLC xenografts used thus far for therapeutic studies were derived mainly from established cancer cell lines (21). These cell lines have been cultivated for a long time in monolayers on plastic dishes and lost most of their original tumor tissue specificity. To circumvent these issues, we used a large panel of xenograft models that were established directly from patient-derived tumor specimens in immunodeficient mice. We (41) and others (42) have shown previously that colon cancer xenografts in early passages are very similar to the primary colon cancer from which they were derived with regard to histology, karyotype, and chemoresponsiveness.

Thus far, only few studies used patient-derived lung xenograft models (26–28). The study of Judde et al. (26) investigated the response of 5 xenograft models with wild-type EGFR to gefitinib in combination with cisplatin-based chemotherapeutic protocols. The study of Perez-Soler et al. (27) using NSCLC heterotransplants revealed similar engraftment rates (34 of 100) as ours. They also confirmed morphologic identity between original and transplanted tumors and showed a similar response rate to paclitaxel (21%) as we did (16%).

We were able to establish 25 routinely passagable NSCLC xenograft models starting from 102 NSCLC specimens. The engraftment rate of about 23% was in a similar range as the one reported by Cutz et al. (43) and is lower than the ones observed in our group for colon cancer xenografts with 43% (41) or acute lymphatic leukemias with 64% (44) but higher than for breast carcinomas with 9% (45). The NSCLC xenografts showed a high concordance with the corresponding clinical sample both concerning histology and immunohistologic staining for Ki-67 and EpCAM. Similar results of high comparativeness of xenografts with the original tumor were described by us (41) and others (27, 42, 46, 47).

The genome-wide RNA expression patterns of the clinical NSCLC specimens and the corresponding early-passage xenografts were analyzed by microarray analysis. In about half of the investigated cases (9 of 17), where global gene expression profiling data are available, the primary NSCLC specimens clustered together with the corresponding xenograft model (Fig. 2). The other 8 cases of primary lung cancers that clustered separately contained a lower amount of tumor cells in the clinical samples. However, the calculated correlation coefficients (Supplementary Table S1) showed a high concordance of gene expression of the xenografts with the corresponding clinical sample. We found 134 genes (<1% of the genome) that were differentially expressed between the clinical specimen and the xenograft derived thereof (Supplementary Table S2). In a similar genomic study with melanomas, a much lower correlation of only 75% with 1,374 differentially expressed genes between primary patient tumors and the corresponding xenografts was found (48). In the majority, the differentially expressed genes in our study belonged to the human stroma and immune cells, which are replaced during the early passaging by mouse tissue and are therefore not detected on the gene expression arrays that are specifically designed to detect human genes. This is also reflected by the fact that >80% of the differentially expressed genes are down-regulated in the xenografts compared to primary tumors.

The chemoresponsiveness of our panel of NSCLC xenografts revealed a relative sensitivity for paclitaxel, gemcitabine, and carboplatin, drugs that also give the highest response rates in the clinical situation (49). Most of our NSCLC donors were not chemotherapeutically pretreated so that no individual comparison of clinical and xenograft outcomes could be done. However, the differential responses of the 25 established xenografts to the single drug treatments (Tables 2 and 3) would highly justify an individual prediction of responsiveness.

In addition, we included cetuximab, a monoclonal antibody that binds to the extracellular domain of the EGFR, and erlotinib, a TKI in the chemosensitivity testing. Although in most of the published studies gefitinib was used for treatment, we decided for erlotinib as it has shown better antitumor effect in NSCLC (50).

The two therapies directed against the EGFR exerted a positive response (T/C values below 50%) in 12 of 25 and 6 of 25 of the xenografts, respectively (Table 3), although only in few cases a shrinkage of established tumors could be shown. Cetuximab treatment resulted in a stronger and dose-independent inhibition compared with erlotinib (Fig. 1C). Seven NSCLC xenograft models were remarkably growth inhibited by cetuximab but not by erlotinib (Table 3). The observed differential responsiveness of the xenograft models reflects the situation in the clinic: very few complete or partial responders and quite several intrinsically resistant NSCLC toward therapy with cetuximab have been reported (51). An individual prediction of responsiveness toward these therapies would therefore be highly warranted; thus far, no validated biomarkers have been developed for routine use.

Because the first description of mutations in the EGFR and a correlation with response to EGFR-targeted therapy (9, 52, 53), numerous investigations have been done to use this variable as prediction marker (8, 54, 55). It became obvious that mutations mainly arise in patients of Asian origin, females, nonsmokers, and adenocarcinomas, a cohort that also showed best responses to gefitinib (56). From these facts, it is not surprising that in our NSCLC xenograft collection originating from a Caucasian population with a low percentage of never smokers (1 of 25) and of both genders no relevant mutations in the EGFR gene were found. Independent of the lack of functional mutations, 12 of 25 xenografts were significantly growth inhibited by cetuximab and 6 of 25 xenografts by erlotinib pointing to the fact that other markers besides mutations in the target gene might determine therapeutic responsiveness. Our results agree with a report of Steiner et al. (57) that showed efficacy of cetuximab in several cell line-derived NSCLC xenograft models independent of mutations in the EGFR.

K-ras mutations have been described to cause intrinsic resistance toward TKI in NSCLC (55, 58–60). In our panel of NSCLC xenografts, 6 tumors presented with K-ras mutations from which 5 showed intrinsic resistance to erlotinib, confirming the reported observations. Only one xenograft with a K-ras mutation in codon 12 (ADC 7462) was highly responsive to both EGFR-targeted therapies. A 52% frequency of p53 mutations found in our xenograft panel is comparable with the incidence of p53 mutations in NSCLC patients (61). There was no correlation between p53 mutations and therapeutic response. From these mutational analyses, it can be concluded in general that single gene alterations cannot alone or exclusively determine the sensitivity of a NSCLC toward an EGFR-targeted therapy.

To test the hypothesis that blocking the EGFR could result in changes of protein expression or phosphorylation status of the EGFR itself as well as downstream signal transduction proteins and as these changes might be pharmacodynamic markers, we studied protein expression of the EGFR signal transduction proteins in the xenograft models. We did not find a correlation between constitutive EGFR protein expression and response to EGFR-targeted therapies. Even after treatment, only minor changes were observed in accordance with drug response. Two tumors (ADC 7462 and ADC 7466) being sensitive to EGFR-targeted therapies showed a lower expression level of the EGFR protein after cetuximab and to a lower extent also after erlotinib treatment (Fig. 3). That reflects data reporting an internalization of EGFR protein after antibody treatment (62).

Whereas several preclinical studies did also not find a clear correlation between EGFR expression and gefitinib sensitivity (21, 63–65), others have reported a positive correlation (66). It was shown that NSCLC with high levels of phosphorylated AKT and mutated EGFR were more sensitive to gefitinib treatment and had a significantly longer time to progression and overall survival in advanced or metastatic stage (67–69). We were not able to confirm these results as in our models neither gain-of-function mutations in the EGFR nor constitutively active phosphorylated AKT correlated with treatment response. As potential reason for these partially conflicting results, the stage of our tumor samples should be considered. In many studies, advanced or metastatic NSCLC from patients or highly dedifferentiated cell lines have been used. In contrast, we included in our study exclusively primary tumors of patients diagnosed at early stages (T₂ and T₃) in which mutations of the EGFR or constitutively active downstream pathways (phosphoinositide 3-kinase or AKT) are rare events. Further, it is known that regulations of the EGFR protein only can be observed when high levels of ligands (e.g., EGF) are present. We suggest that, in contrast to cell culture experiments, in vivo physiologically only minor amounts of ligands are available, so that potential changes in protein expression are below the detection limit.

In summary, a large panel of well-characterized patient-derived NSCLC xenografts is now available whose concordance with the clinical situation has been shown. These xenograft models allow to perform preclinical studies evaluating novel targeted therapies and uncovering the expression and regulation of biomarkers. Further investigations comparing the therapeutic response of NSCLC xenografts to the expression of resistance markers and growth factor ligands are ongoing and will be reported separately.

Disclosure of Potential Conflicts of Interest

I. Fichtner is employed by EPO-GMBH. The authors report no other conflicts.