Modeling Lung Cancer Evolution and Preclinical Response by Orthotopic Mouse Allografts

Lung cancer is the leading cause of cancer death worldwide. Non–small cell lung cancer (NSCLC) accounts for about 85% of cases, with a 5-year survival rate below 15% (1). To improve control over this devastating disease, a variety of in vivo preclinical models have been used during the last decade. Among those, subcutaneous xenografts of human NSCLC cell lines have been extensively applied for drug sensitivity assays (2). Alternatively, intrapulmonary delivery of human cell lines in nude mice has provided a more clinically relevant system to study NSCLC biology (3). However, these methodologies still present some disadvantages. In particular, it is troublesome to obtain a single intraparenchymatous lung tumor that mimics the clinical situation and allows longitudinal follow-up. Indeed, localized intrapulmonary disease accounts for the majority of patients with NSCLC and is the basis of further tumor progression. In parallel, genetically engineered mouse models (GEMM) have become indispensable tools for our understanding of the mechanisms that contribute to tumor development as well as for treatment design and target validation demands. However, GEMM lung tumors induced by a single initiating oncogene are predominantly early-stage lesions, display limited histologic variation, and metastasize rather infrequently (4).

In sum, there is an urgent necessity to establish suitable animal models of NSCLC that faithfully recapitulate every salient aspect of the human disease. Here, we report a versatile and reproducible approach that circumvents important disadvantages of preexisting animal models of NSCLC, in particular those associated with low histologic heterogeneity and lack of spontaneous metastasis. This methodology is based on sequential orthotopic implantation of small solid fragments from an individual murine primary tumor into recipient mice. With this approach, we have generated an NSCLC orthoallobank: a comprehensive archived collection of serially passaged murine lung adenocarcinomas together with paired cell lines. This methodology enables full recapitulation of human NSCLC histopathologic features, including selection of mutations in relevant tumor suppressors together with metastatic dissemination. Our method offers a powerful tool to study the poorly characterized progression toward full-blown metastatic adenocarcinomas present at the time of diagnose in human patients. Moreover, it provides a rapid and standardized platform for the preclinical evaluation of therapeutic treatments in vivo.

The K-Ras^{lox/LSLG12Vgeo}; RERT^ert/ert strain has been previously described (5). Lung tumors were induced by intraperitoneal injection of 4-hydroxitamoxifen as previously described (5). EML4-ALK mice will be reported elsewhere (Voena and colleagues, manuscript submitted). In brief, a cDNA fragment encoding EML4-ALK (variant 1) was ligated to the surfactant protein-C (SP-C) promoter as well as to a polyadenylation signal. The expression cassette was injected into pronuclear stage embryos of FVB/N mice. The presence of the transgene was examined by PCR analysis after extraction of DNA from the tail of founder animals. RT-PCR analysis was performed to detect EML4-ALK mRNA to confirm the lung-specific expression of the transgene. Crl:NU-Foxn1^nu mice were purchased from Charles River. All animal experiments were approved by the Ethical Committee of CNIO and IDIBELL and performed in accordance with the guidelines for Ethical Conduct in the Care and Use of Animals as stated in The International Guiding Principles for Biomedical Research Involving Animals, developed by the Council for International Organizations of Medical Sciences (CIOMS).

Primary tumors from inducible K-Ras^{lox/LSLG12Vgeo} or EML4-ALK mice were aseptically isolated and placed at room temperature in DMEM supplemented with 10% FBS plus 50 U/mL penicillin and 50 mg/mL streptomycin. Within 12 hours from surgical resection, tumors were implanted in Crl:NU-Foxn1^nu mice following previously reported procedures (3). Briefly, mice were anesthetized with a continuous flow of 1% to 3% isoflurane/oxygen mixture (2 L/min) and subjected to right thoracotomy. Mice were situated in left lateral decubitus position, and a small transverse skin incision (∼5–8 mm) was made in the right chest wall. Chest muscles were separated by a sharp dissection and costal and intercostals muscles were exposed. An intercostal incision of 2 to 4 mm on the third or fourth rib on the chest wall was made and a small tumor piece of 2 to 4 mm³ was introduced into the chest cavity. The tumor specimen was deposited following two alternative surgical procedures: (i) tumor specimens were deposited between the second and the third lung lobule and (ii) for the chemotherapeutic approach, tumor specimens were anchored to the lung surface with Prolene 7.0 suture. Next, the chest wall incision was closed with surgery staples, and finally chest muscles and skin were closed. Mice were inspected twice a week and monitored for the presence of breathing problems. On average, implanted lung tumours were harvested at humane endpoint, cut into small fragments, and serially transplanted into two to three new animals. These engrafted tumors (named orthoallografts) were also cryopreserved as 2 to 4 mm³ pieces in a solution of 90% FBS and 10% dimethyl sulfoxide and stored in liquid nitrogen for subsequent future implantations. In addition, a fresh small piece was used to derive matched cell lines in vitro when appropriate (see below).

K-Ras^{lox/LSLG12Vgeo} orthoallografts were implanted in Crl:NU-Foxn1^nu mice and randomly allocated into the treatment groups. Mice were intravenously treated at days 7, 12, and 17 (cisplatin, 3 mg/kg; paclitaxel, 20 mg/kg) and sacrificed at day 22 postimplantation.

EML4-ALK orthoallografts were implanted in Crl:NU-Foxn1^nu mice and randomly allocated into the treatment groups. Crizotinib (100 mg/kg) was administered daily by oral gavage starting on day 60 postimplantation and continued during 10 days. Upon completion of both treatments, hematoxylin and eosin (H&E)-stained lung sections were blindly assessed by a pathologist. Crizotinib (PF-02341066) was a kind gift from Pfizer Inc.

Images were acquired using eXplore Vista PET-CT (GE Healthcare). Mice were injected with 15MBq of ¹⁸F-FDG into the lateral tail vein in a volume of 100 μL. During imaging, mice were anesthetized with a continuous flow of 1% to 3% isoflurane/oxygen mixture (2 L/min). Forty-five minutes after radiotracer injection, micro-CT images were acquired at 400 projections and collected in one full rotation of the gantry in approximately 10 minutes. Micro PET scans were performed at 20 minutes per bed. CT images were reconstructed using filtered back projection with a Shepp–Logan filter and PET images with 3D OSEM reconstruction algorithm. ¹⁸F-FDG (2[¹⁸F]fluoro-2-deoxy-d-glucose) PET-CT images were analyzed using Amide Medical Image Data Examiner (AMIDE) software for evaluation of tumor ¹⁸F-FDG uptake. ¹⁸F-FDG uptake is indicated as maximal standardized uptake value (SUV Max), and was calculated by drawing a 3-dimensional (3D) region of interest (ROI) over the tumor applying the formula SUV = ROI radioactivity concentration (MBq/cc)/injected dose (MBq)/mice body weight (g).

Freshly collected lung tumor tissues coming from K-Ras^{lox/LSLG12Vgeo} and EML4-ALK orthotopic implants were minced with sterile scalpels. Single cells and clumps were transferred to cell culture plates and maintained in DMEM supplemented with 10% FBS plus 50 U/mL penicillin and 50 mg/mL streptomycin under standard culture conditions. When cell colonies with epithelial cell morphology were observed, cells were trypsinized and expanded. For the single-cell colony generation, cells were diluted with culture medium to obtain a concentration of approximately 1 cell/100μL and seeded in 96-well plates (100μL/well). After 12 hours, each well was carefully examined by bright-field microscopy. Wells containing single cells with epithelial morphology were marked, and the culture medium in such wells was changed every 3 days. Expanded cell populations derived from these colonies were subjected to further characterization. When appropriate, doxorubicin (Sigma-Aldrich) was used at 5 μg/mL.

For X-Gal staining, cultured cells were fixed with 0.2% glutaraldehyde for 15 minutes and incubated in X-Gal staining solution (0.1 mol/L phosphate buffer, 2 mmol/L MgCl₂, 5 mmol/L potassium ferrocyanide, and 5 mmol/L potassium ferricyanide containing 1 mg/mL X-Gal [5-bromo-4-chloro-3-indolyl-β-d-galactosidase]) for 12 to 16 hours at 37°C.

Total cellular RNA (1 μg), extracted by RNeasy Mini Kit (QIAGEN), was reverse-transcribed by random primers using SuperScript Double-Stranded cDNA Synthesis Kit (Invitrogen), and the reverse transcription reaction (1 μL) was then subjected to PCR amplification using FastStart Universal SYBR Green Master (Roche). PCR signals were recorded on a StepOnePlus Real-time PCR System (Applied Biosystems) and analyzed using the StepOne version 2.2 software (Applied Biosystems). Primer sets were the following:

(i) Cytokeratin 13 Fw: TCGTGACTGGCATCTGAAAC

Cytokeratin 13 Rev: AGGATGATCCGGTTGTTGTC

(ii) Cytokeratin 19 Fw: GGGGGTTCAGTACGCATTGG

Cytokeratin 19 Rev: GAGGACGAGGTCACGAAGC

(iii) Vimentin Fw: CGGCTGCGAGAGAAATTGC

Vimentin Rev: CCACTTTCCGTTCAAGGTCAAG

(iv) E-cadherin Fw: TTTTCGGAAGACTCCCGATTCA

E-cadherin Rev: AGCTTGTGGAGCTTTAGATGC

(v) N-cadherin Fw: CTGATAGCCCGGTTTCACTTG

N-cadherin Rev: CAGGCTTTGATCCCTCTGGA

About 25 μg of protein extracts obtained from cell lysates was separated on SDS-PAGE (Bio-Rad), transferred to a nitrocellulose membrane, and blotted with primary antibodies raised against p53 (Cell Signaling Technology), p21 (Santa Cruz Biotechnology), and GAPDH (Sigma).

Cells (1 × 10⁶) were resuspended in 200 μL of sterile PBS and then inoculated by tail vein injection into Crl:NU-Foxn1^nu mice of 6 to 10 weeks of age. Animals were euthanized at different time points postinoculation, and whole lungs or individual lung tumors were collected.

For routine histologic study, lung lobes were fixed in 10% buffered formalin (Sigma) and embedded in paraffin. Paraffin sections (5 μm thick) were then processed for either H&E staining or immunohistochemistry. Images were acquired with a LEICA DM4000B microscope equipped with LEICA DC500 digital camera. Antibodies used for immunostaining included those raised against cleaved caspase-3 (R&D Systems), cytokeratin-7, and thyroid transcription factor 1 (Abcam).

To develop orthotopic lung cancer implants, we took advantage of the inducible K-Ras^G12V model that was engineered as a silent knock-in mutation preceded by a floxed transcription stop cassette (Lox-Stop-Lox or LSL). Activation of the K-Ras^LSLG12Vgeo allele is achieved upon Cre recombinase–mediated excision of the stop cassette, resulting in the development of NSCLC (6). To generate a more aggressive model, we used the K-Ras^{lox/LSLG12Vgeo} strain that includes a floxed version of the K-Ras wild-type allele. In this compound strain, delivery of the Cre recombinase results in expression of the oncogenic mutation together with the concomitant ablation of the wild-type copy of K-Ras (5). With this approach, we recapitulated the recurrent LOH event as found in 50% of human lung tumors with K-Ras activating mutations (7).

A total of 46 primary NSCLC tumors were isolated from four K-Ras^{lox/LSLG12Vgeo} mice upon appearance of breathing difficulties (on average 11 months subsequent to oncogene induction). These primary lesions displayed limited histologic heterogeneity and were mainly classified as solid or papillary adenocarcinomas (Supplementary Table S1). All tumors were individually implanted in the right parieto-pleural lung region of Crl:NU-Foxn1^nu mice (hereafter nude mice) following a surgical procedure that can be easily implemented in any standard animal facility. Twenty-seven (∼60%) implanted tumors successfully engrafted and were able to grow as orthotopic implants through a period of time ranging from 98 to 150 days from implantation to humane endpoint (Supplementary Table S1). We observed that papillary primary tumors engrafted more efficiently (82%) when compared with solid tumors (50%). In 12 cases (26%), tumors did not grow through a follow-up period of 6 months, whereas the remaining seven animals had to be prematurely euthanized because of postoperative complications.

Passage #1 implants were used to perform consecutive mouse-to-mouse serial transplantations, resulting in faster tumor growth and consequently shorter lifespan (Supplementary Table S1). At the humane endpoint of every passage, all tumors were harvested and cryopreserved to create a comprehensive NSCLC orthoallobank. In addition, we embedded in paraffin and snap froze all implants upon conclusion of every passage to generate an inclusive archive of samples for future analysis. To check the viability of cryopreserved samples, we thawed and orthotopically reimplanted four different second-passage orthoallografts (oag-9, oag-12, oag-21, and oag-22) chosen to represent the three most frequent histologic subtypes (papillary, acinar, and solid; Supplementary Table S2). Importantly, these tumors not only successfully engrafted but also displayed median passage duration and histologic features indistinguishable from paired samples implanted without the cryopreservation step. Next, to test its suitability as a practical tool for the validation of preclinical treatments, we assessed whether the evolution of implanted tumors could be followed-up by standard imaging technology. To this end, we analyzed a cohort of animals orthotopically implanted with cryopreserved tumors by PET following injection of ¹⁸F-FDG. Tumor-specific FDG-PET signal was readily measurable upon implantation and increased during the course of the experiment becoming detectable in other lung lobules (Fig. 1A and B). This was further documented by post-mortem histopathologic analysis (Fig. 1C).

Finally, to prove whether this methodology could be extended to other driver oncogenes found in human NSCLC, we implanted murine primary lung tumors derived from EML4-ALK transgenic mice (Voena and colleagues, manuscript submitted; see Materials and Methods). Eleven tumors were orthotopically implanted in nude mice, resulting in four successful engraftments (36%). The observed frequency of engraftment increased to 77% when the procedure was replicated with EML4-ALK; p53^−/− primary tumors (Supplementary Table S3). All these lesions were also subjected to sequential passages (up to passage #3) and cryopreserved. This indicates that the methodology described above can be potentially extended to primary tumors from virtually any alternative GEMM.

In humans, invasive adenocarcinoma represents 60% to 70% of all surgically resected lung cancer cases. According to the predominant histologic pattern, adenocarcinomas can be further classified in differentiated (comprising lepidic, acinar, papillary, and micropapillary) and undifferentiated/solid subtypes (8). However, lung tumors in K-Ras mutant mice display much lower complexity than human NSCLC. Indeed, the heterogeneity of human disease is poorly represented in primary tumors from K-Ras^{lox/LSLG12Vgeo} mice, which predominantly display a differentiated pattern (Supplementary Table S1; see also ref. 9). In sharp contrast, orthoallografts obtained after a single passage in nude mice closely reproduced the diverse histologic appearance of human NSCLC. Unlike what we observed in the originating primary tumors, the most representative histologic subtypes found in human cancers appeared with variable frequency in the orthoallograft collection (Supplementary Fig. S1 and Supplementary Table S2).

We next asked whether the serial passaging of NSCLC orthoallografts could reveal a tendency to acquire increased malignancy over time. This is an important aspect, as the initial stages of malignization are difficult to be approached in humans. To this end, a group of representative primary tumors (n = 6) including predominant differentiated and undifferentiated histologic subtypes were passed mouse-to-mouse in the course of up to four sequential passages (Supplementary Table S4). As an initial evidence of increased malignancy, we first examined whether orthoallografts may have accumulated mutations with prognostic value. We focused our search to the tumor suppressor genes p53 and Lkb1 (also known as Stk11), as both present the highest frequency of co-occurrence with K-Ras activating mutations in advanced human NSCLC (10). While we failed to detect changes in the most frequently mutated p53 exons (exons 4, 6, and 7), on average, 10% of the tumor cells in two of the 13 orthoallografts analyzed presented aberrations in the Lkb1 gene, a frequently mutated gene in human NSCLC (11). We detected one nonsynonymous mutation (D162G) and one deletion (E354Δ) affecting the Lkb1 protein kinase and C-terminal regulatory domains, respectively (Fig. 2A). These regions concentrate most Lkb1 cancer–related mutations in humans (11). Importantly, Lkb1 aberrations were not identified in the originating passage #1 implants (Fig. 2A). Furthermore, the expression of the tumor suppressor p16 was decreased in late-passage samples, thus resembling the loss of p16 in human NSCLC progression (Fig. 2B; ref. 12). Finally, it has been proposed that expression of embryonic stem cell (ESC) markers correlates with undifferentiated NSCLC and is associated with worse prognosis (13). We took advantage of an RT-PCR array to measure the expression changes of a collection of established ESC markers during the orthotopic sequential transplantation (data not shown). Klf4, Oct4, c-Myc, CD24, and Lgr6 displayed a higher expression fold change in late- versus early-passage implants in a cohort of paired samples (n = 8; Fig. 2B).

We further collected other evidences that suggested an evolution toward a more malignant phenotype during the serial passaging of orthoallografts. First, we observed that the period from tumor implantation to humane endpoint was progressively shortened along the serial transplantation procedure. Yet, beyond passage #3, this period remained constant (Fig. 3A). In all cases and irrespectively of passage duration, when the initial implant predominantly displayed solid features, we did not observe major histologic changes through passages being the engrafted tumor histologically stable (data not shown). In contrast, those implants that initially showed a predominant differentiated pattern eventually displayed an increase in the incidence of the solid subtype (Fig. 3B). Importantly, despite of the increment of undifferentiated components, these tumors maintained some degree of histologic heterogeneity as also observed in human biopsies (Supplementary Table S3).

Notably, in addition to their close histologic resemblance to human NSCLC, orthoallografts displayed metastatic potential as early as passage #1. This is in sharp contrast to K-Ras mutant GEMM models that lack metastatic capacity unless additional genetic modifications are included (4). As mentioned above and akin to human disease, the engrafted implants frequently invade neighboring lung lobules and disseminate to various locations forming macroscopic implants in the diaphragm, ribs, chest cavity, lymph nodes, and liver (Fig. 3C). Furthermore, careful histopathologic scrutiny of brain and liver sections, two of the main metastatic target tissues of primary human NSCLC, revealed the presence of micrometastasis in 7 of 21 (33%) implant-bearing animals (Fig. 3D and Supplementary Table S5).

Thus far, it has been unfeasible to efficiently establish murine NSCLC cell lines that are p53-proficient. This is particularly relevant, as 50% of human NSCLC cases retain a wild-type p53 (14). In an attempt to establish such primary cell lines, orthotopic implants were resected and mechanically disaggregated immediately after passage #1 and maintained in culture under standard conditions. Following this approach, a panel of 35 murine K-Ras^G12V lung cancer (mKLC) cell lines was successfully derived from their corresponding orthoallografts. The process of cell line generation occurred with similar efficiency irrespectively of the predominant histologic subtype. Yet, the paired metastatic cell lines predominantly originate from solid orthoallografts (Supplementary Table S6). To assess whether these cell lines retained metastatic dissemination potential, a cohort of mice received a cell line derived from a diaphragmatic implant via tail vein injection. As expected because of the inoculation route, all animals developed pulmonary nodules. Remarkably, the injected animals also developed macroscopic tumors in various tissues such as lymph nodes, bone, liver, adrenal gland, ovary, skin, and salivary gland, thereby resembling the metastatic pattern reported in human patients with NSCLC (Fig. 4 and data not shown).

Next, four mKLC lines were further characterized in vitro. To confirm expression of K-Ras^G12V and the absence of stromal contaminants, we took advantage of the surrogate marker β-geo, inserted as a bicistronic transcript at the 3′–untranslated region (UTR) of the K-Ras locus (6). As expected, all cell lines displayed strong X-Gal staining, indicative of K-Ras^G12V expression (Supplementary Fig. S2A). Although all cell lines showed a characteristic morphology evocative of an epithelial origin, we nevertheless assessed by RT-PCR the expression of prototypical epithelial markers such as cytokeratin 13, cytokeratin 19, and E-cadherin along with the mesenchymal markers vimentin and N-cadherin. All tested mKLC cell lines expressed these epithelial markers (Supplementary Fig. S2B). Interestingly, we also observed increased levels of vimentin and N-cadherin when compared with primary tumors, suggesting the acquisition of a mesenchymal-like phenotype as previously reported for other epithelial cancer cell lines in vitro (15).

p53 is known to be stabilized in response to DNA damage and other stress signals, resulting in the induction of several downstream targets that mediate a variety of cellular responses such as cell-cycle arrest or apoptosis (16). To assess p53 functionality, we exposed the mKLC lines to the genotoxic agent doxorubicin. This treatment resulted in clear p53 stabilization together with an induction of the p53 transcriptional target p21, indicating that mKLC cell lines derived by an orthoallograft procedure can be efficiently established preserving a functional p53 response (Supplementary Fig. S2C).

Considering that undifferentiated histopathologic subtypes are associated with poor prognosis in human patients with NSCLC (17), we set out to assess the capacity of our mKLC cell lines to fully reconstitute tumor heterogeneity in vivo. To this end, we used grafting by tail vein injection to favor lung colonization and performed a time course study to investigate tumor progression. Injection of 1 × 10⁶ cells resulted in tumor formation that recapitulated the histologic heterogeneity as well as time-dependent evolution observed in the orthoallografted tumors. Whereas early stages (12 days) were characterized by small nodules of tumor cells with focal acinar differentiation, tumor progression invariably resulted in the appearance of adenocarcinomas with predominant solid features at humane endpoint (25 days; Supplementary Fig. S3). Importantly, the presence of a variable contribution of other histologic subtypes was also identified in a cell line–dependent manner (data not shown). We next aimed to assess whether this histologic heterogeneity, also observed in the majority of orthotopic implants as well as in human disease, was the result of a mosaic composition of cells of different nature or instead a reflection of inherent cellular plasticity in vivo. To this end, we generated single-cell clones starting from mKLC mass cultures and injected them via tail vein into recipient mice. Remarkably, single-cell clones were able to generate lung lesions displaying different histologic subtypes with variable composition depending on the individual clone (Supplementary Fig. S4).

We next wanted to assess whether the orthotopic implants could be used to anticipate therapeutic response in the clinic. To this end, we subjected nude mice implanted with K-Ras^G12V orthoallografts to the current standard of care for patients with NSCLC consisting of cisplatin-based chemotherapy (cisplatin plus paclitaxel). As shown in Fig. 5A, the cohort receiving chemotherapy showed a reduction in tumor burden when compared with those receiving vehicle alone. This was accompanied by the appearance of abnormal nuclei in the majority of cells, displaying enlarged, irregular, or multinucleated morphology (Fig. 5B). Overall, this response and the resulting histopathologic changes recapitulate the clinical outcome observed in patients with NSCLC subjected to this chemotherapeutic regime (18, 19).

Next, we aimed to extend our approach to a targeted therapy currently used in clinical practice with patients with NSCLC. Because there are no targeted treatments available for K-Ras mutant tumors, we studied the EML4-ALK model for which a crizotinib-based therapy is already in use in the clinic. Patients with NSCLC harboring an EML4-ALK inversion show an initial response accompanied by substantial tumor regression but this is inevitably followed by the acquisition of resistance (20). Interestingly, mice carrying EML4-ALK implants displayed a substantial tumor burden reduction when subjected to crizotinib treatment (Fig. 5C), with viable tumor areas scattered among necrotic and fibrotic tissue that may be responsible for future relapse (Fig. 5D and Supplementary Table S7).

We describe a novel and reproducible methodology based on sequential orthotopic implantation of primary tumors in recipient mice that circumvents most of the disadvantages associated with murine models of NSCLC such as the presence of low histologic heterogeneity and lack of spontaneous metastasis. With this approach, we have been able to document cancer evolution toward an increased malignant phenotype. This time-dependent evolutionary trend had been previously postulated for the progression of human NSCLC but lacked formal experimental evidence. In our model, the progressive acquisition of malignant traits is supported by several observations:

1. Serial passaging results in increased local invasion together with acquisition of distant metastatic potential displaying the same dissemination tropism observed in humans.

2. Similarly to advanced human NSCLC, serial passaging results in the acquisition of tumor-promoting mutations as well as loss of expression of relevant tumor suppressor genes. These alterations were not found in the original primary lesions, suggesting that they are progressively enriched as a consequence of fitness advantage and selective pressure.

3. Marked intratumor heterogeneity is generated by NSCLC cell lines originating form single cells and evolves over time toward an increase in more undifferentiated histologic subtypes that resemble those associated with poor prognosis in humans.

4. Both K-Ras^G12V- and EML4-ALK–driven implants are responsive to chemotherapy and crizotinib, respectively, but invariably display viable tumor cells posttreatment.

5. Serial transplantation allowed for the first time the generation of p53-proficient murine NSCLC primary cell lines.

The orthoallobank is a powerful tool to allow a comprehensive study of NSCLC evolution (Fig. 6). Our approach is well suited for the study of tumor features from a cell-autonomous perspective such as target validation exercises (both genetically and pharmacologically based), appearance of resistance mechanisms, and the contribution of stromal components. However, it is now accepted that the immune system plays both host-protective and tumor-promoting functions that influence various important aspects from tumor onset to metastatic dissemination (21, 22). In this context, the lack of a functional adaptive immune response represents an obvious limitation. The implementation of this technology in syngeneic mice is ongoing and will facilitate the analysis of the immune system contribution, resulting in a more comprehensive approach (data not shown). The collection of archived samples may be used to characterize by deep sequencing technologies the order of genomic events that eventually lead to the establishment of aggressive NSCLC. This is particularly relevant, as in humans, many of the co-driver mutations that trigger or influence the progression of the disease remain poorly defined. They appear with variable frequency and are often difficult to set apart from the substantial number of passenger mutations. Because there is no fitness advantage associated with passenger mutations, their accumulation during tumor development is largely proportional to time (23). Because of the relatively short period from implantation to malignization, this orthotopic allograft approach may diminish the chances to accumulate passenger mutations. Instead, those that play a relevant role during tumor development will be selected for, thus facilitating their identification. Indeed, we have identified mutations in the tumor suppressor gene Lkb1 identical to those previously reported in human NSCLC. Among those, the D162 mutation has been shown to be important for the structural integrity of human LKB1, and mutations affecting this residue have been reported in Peutz–Jeghers syndrome and other types of cancer (24). Therefore, this methodology will yield the causal and temporal relationships between specific genotypic aberrations acquired during multistage carcinogenesis and the resulting lung cancer phenotype. For instance, our results suggest that mutations affecting p53 arise later than those in Lkb1 or p16 silencing in K-Ras^G12V–driven NSCLC. Interestingly, unlike the originating primary tumor models, the serial transplantation of both K-Ras^G12V and EML4-ALK implants results in the acquisition of metastatic potential. Thus, this approach offers an invaluable platform to potentially characterize the metastatic dissemination process in vivo as well as the mutations, gene expression variations, or epigenetic changes responsible for the acquisition of the invasive capacity.

Likewise, the orthoallobank is a versatile and rapid platform for the preclinical assessment of therapeutic regimes including those dealing with the appearance of resistance mechanisms. K-Ras^G12V- and EML4-ALK–driven implants respond to chemotherapy and crizotinib, respectively, and display histopathologic features that could facilitate the understanding of the clinical outcome in patients subjected to similar regimes. While this response is comparable to that observed in primary murine tumors (20, 25), the orthotopic approach is substantially faster, reproducible, and homogenous, thus resulting in a more reliable and straightforward alternative. Furthermore, selected tumors from the archived orthoallobank that carry particular sets of co-driver mutations could be treated simultaneously to assess potential differences in the therapeutic response that may depend on, or be influenced by, a particular genetic composition. Also, proteomic profiling of plasma samples could potentially be applied for the identification of predictive biomarkers of treatment efficacy (26).

In sum, we show a new systematic approach to study the evolution of cancer that provides a powerful strategy not only to tease out the molecular events that stimulate NSCLC progression but also to identify novel therapeutic strategies to fight this devastating disease.

A. Villanueva is founder of the Spin-off of XenOPAT S.L. No potential conflicts of interest were disclosed by the other authors.

Conception and design: C. Ambrogio, F.J. Carmona, A. Vidal, M. Barbacid, D. Santamaría, A. Villanueva

Development of methodology: C. Ambrogio, F.J. Carmona, A. Vidal, O.A. Romero, T. Poggio, M. Sánchez-Céspedes, M. Esteller, F. Mulero, C. Voena, D. Santamaría, A. Villanueva

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C. Ambrogio, F.J. Carmona, A. Vidal, M. Falcone, P. Nieto, O.A. Romero, S. Puertas, M. Vizoso, M. Sánchez-Céspedes, M. Esteller, R. Chiarle, M. Barbacid, A. Villanueva

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C. Ambrogio, F.J. Carmona, A. Vidal, S. Puertas, M. Esteller, F. Mulero, R. Chiarle, D. Santamaría, A. Villanueva

Writing, review, and/or revision of the manuscript: C. Ambrogio, F.J. Carmona, A. Vidal, S. Puertas, E. Nadal, F. Mulero, R. Chiarle, D. Santamaría, A. Villanueva

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A. Vidal, M. Vizoso

Study supervision: C. Ambrogio, F.J. Carmona, M. Barbacid, D. Santamaría, A. Villanueva

The authors thank Samuele Marro, Maria Luigia De Bonis, and Manuel Serrano for critical reading of the article.

This work was supported by grants from the European Research Council (ERC-AG/250297-RAS AHEAD), EU-Framework Programme (LSHG-CT-2007-037665/CHEMORES, HEALTH-F2-2010-259770/LUNGTARGET, and HEALTH-2010-260791/EUROCANPLAT- FORM), Spanish Ministry of Economy and Competitiveness (SAF2011-30173), and Autonomous Community of Madrid (S2011/BDM-2470/ONCOCYCLE) to M. Barbacid; European Research Council (ERC-StG-2009-242965-LUNELY), Associazione Italiana per la Ricerca sul Cancro (AIRC) grant IG-12023, and Association for International Cancer Research (AICR) grant 12-0216 (R. Chiarle); Fondo de Investigaciones Sanitarias (PI10-0222, PI13-01339), INNPACTO (ORALBEADS-IPT-2011-0754-900000), and Generalitat de Catalunya (2005SGR00727) to A. Villanueva. E. Nadal was supported by a Juan Rodés contract (JR13/0002) from the Carlos III Institute of Health. C. Ambrogio is the recipient of a postdoctoral fellowship from the Spanish Association Against Cancer (AECC).

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.