Patient-derived xenografts (PDX) can adequately reflect clinical drug efficacy. However, the methods for evaluating drug efficacy are not fully established. We selected five non–small cell lung cancer (NSCLC) PDXs with genetic alterations from established PDXs and the corresponding molecular targeted therapy was administered orally for 21 consecutive days. Genetic analysis, measurement of drug concentrations in blood and tumors using LC/MS-MS, and analysis of drug distribution in tumors using matrix-assisted laser desorption/ionization mass spectrometry were performed. Fifteen (20%) PDXs were established using samples collected from 76 patients with NSCLC with genetic alterations. The genetic alterations observed in original patients were largely maintained in PDXs. We compared the drug efficacy in original patients and PDX models; the efficacies against certain PDXs correlated with the clinical effects, while those against the others did not. We determined blood and intratumor concentrations in the PDX model, but both concentrations were low, and no evident correlation with the drug efficacy could be observed. The intratumoral spatial distribution of the drugs was both homogeneous and heterogeneous for each drug, and the distribution was independent of the expression of the target protein. The evaluation of drug efficacy in PDXs enabled partial reproduction of the therapeutic effect in original patients. A more detailed analysis of systemic and intratumoral pharmacokinetics may help clarify the mode of action of drugs. Further development of evaluation methods and indices to improve the prediction accuracy of clinical efficacy is warranted.

The developmental success rate for anticancer drugs is very low, with only 5% of the drugs approved for clinical application even despite a good preclinical response (1, 2). This is partly because traditional drug screening using cell lines is less predictive of clinical efficacy (3). In this context, the patient-derived xenograft (PDX) model, established by implanting a patient's tumor tissue directly into immune-deficient mice, has recently been considered as a possible solution. The PDX retains the original tumor's biological characteristics and heterogeneity as the tumor cells and its entire surrounding microenvironment are transplanted (4–6). Compared with cell lines or cell line xenograft models, PDXs better reflect the therapeutic effect with more than 80% similarity between clinical data and PDX responses in some instances (3, 7, 8).

Over the past two decades, the treatment landscape for non–small cell lung cancer (NSCLC) has changed dramatically owing to a better understanding of the disease biology and mechanisms of tumor progression (9). Specifically, the use of small-molecule tyrosine kinase inhibitors (TKI) and immunotherapy has improved survival rates in patients with NSCLC (10–12). The treatment strategy for lung adenocarcinoma is determined by gene profiling, such as EGFR mutation, anaplastic lymphoma kinase (ALK) fusion, ROS proto-oncogene 1 (ROS1) fusion, and v-raf murine sarcoma viral oncogene homolog B1 (BRAF mutation; ref. 13). It is indisputable that molecularly targeted therapies improve patient survival and quality of life and a more individualized treatment strategy is required (9, 14). Treatment strategies for patients with NSCLC should focus on the genetic alterations and the resistance mechanisms after drug administration (15). However, due to the diversity of these resistance mechanisms and the rarity of the disease, it is difficult to conduct prospective randomized clinical trials (16).

PDXs have been reported in NSCLC (17–22). A co-clinical trial generated PDXs in conjunction with clinical trials and reported the drug resistance mechanisms (23). In addition to their use in the exploration of drug resistance mechanisms and their application in drug efficacy evaluation, there exists a strong correlation between the treatment response in original patients and corresponding PDXs (18). The therapeutic strategies for lung cancer are increasingly fragmented on the basis of genetic alterations and resistance mechanisms; however, PDX is expected to be useful for drug development, for overcoming drug resistance, and for reducing the lag in clinical trials by enabling the conduction of early proof-of-concept testing in preclinical studies.

PDX demonstrates substantial potential utility for studying drug efficacy; however, the most remarkable limitation is the lack of a defined methodology for studying drug efficacy in PDX to predict clinical efficacy. It is necessary to ascertain various factors such as the number of mice to be used, criteria for commencement of the study, evaluation method, drug dosage, and administration schedule. In addition, the prediction of clinical efficacy as an endpoint must be evaluated; however, the criteria and indicators for evaluating the prediction are not determined. In this context, we propose that drug concentration and drug distribution are important factors that will enable PDX utilization for testing drug efficacy in humans. Phase I studies in humans are conducted after evaluation in animal models such as mice, dogs, and monkeys in the clinical development of anticancer drugs. The only standard indicator evaluated in these studies is the drug concentration in blood. Therefore, the blood concentration may be deemed one of the evaluation indices for comparing the therapeutic effects between PDX and humans.

Many drugs, including anticancer drugs, exert their pharmacologic effects in tissues, but not in plasma where the drug concentrations are usually measured. Plasma concentrations are not necessarily representative of the drug concentrations in the tissues. We previously reported the heterogeneous spatial distribution of anticancer drugs in several tumors, using matrix-assisted laser desorption/ionization mass spectrometry imaging (MALDI-MSI; refs. 24–27). The actual distribution and the concentration of drugs in the tumors are unclear, and it is often challenging to obtain human specimens after drug administration. The intratumor distribution of anticancer drugs is a potential evidence of the mode of action; therefore, a validated method to confirm whether drugs have reached the right target is an urgent need for drug development in oncology. PDX is considered an optimal platform for the simultaneous analysis of drug concentration and heterogeneous spatial distribution in tumors and the plasma drug concentration, because it preserves tumor heterogeneity and tissue structure in patients' tumors and enables repeated tumor sample collection at any instance after drug administration.

The aim of this study was to characterize the establishment of PDXs using NSCLC samples and to evaluate the utility of the PDX models in the original patients in terms of genetic alterations, protein expression, plasma drug concentration, and spatial distribution of the drug in tumor tissue (Fig. 1). We have discussed and provided description of the optimal use of PDX to predict clinical efficacy.

Figure 1.

CONSORT diagram and study design. The selection criteria for PDX used in this study are shown in A. Of the 55 lung cancer PDXs, 47 were NSCLC PDXs and 15 were driver gene positive. In this study, we used five of these PDXs. Five driver gene positive PDXs were used to evaluate the efficacy of molecularly targeted agents and to analyze blood concentration, intratumor concentration, and intratumor drug distribution (B). ALK, anaplastic lymphoma kinase; EGFR, epidermal growth factor receptor; HER2, human epidermal growth factor receptor type 2; NSCLC, non–small cell lung cancer; RET, rearranged during transfection; ROS1, c-ros oncogene; SCLC, small cell lung cancer.

Figure 1.

CONSORT diagram and study design. The selection criteria for PDX used in this study are shown in A. Of the 55 lung cancer PDXs, 47 were NSCLC PDXs and 15 were driver gene positive. In this study, we used five of these PDXs. Five driver gene positive PDXs were used to evaluate the efficacy of molecularly targeted agents and to analyze blood concentration, intratumor concentration, and intratumor drug distribution (B). ALK, anaplastic lymphoma kinase; EGFR, epidermal growth factor receptor; HER2, human epidermal growth factor receptor type 2; NSCLC, non–small cell lung cancer; RET, rearranged during transfection; ROS1, c-ros oncogene; SCLC, small cell lung cancer.

Close modal

PDX model

PDX samples were provided by the National Cancer Center J-PDX library. The protocol was approved by the Institutional Review Board (NCCRI: 2015–123), and all patients provided written informed consent. Lung cancer PDX models of cases with driver mutations were selected (Fig. 1A). The study was performed according to the precepts established by the Helsinki Declaration; its design and conduct complied with all applicable regulations, guidance, and local policies. Animal experiments were performed in compliance with the Institute for Laboratory Animal Research guidelines, National Cancer Center Research Institute (Tokyo, Japan) T17–073 and T19–008). The first engrafted PDX established by transplanting the patient's original tumor was designated transgeneration 1 (TG1) and subsequently TG2 and TG3. In the present study, we used PDXs from TG3 to TG4.

Reagents

The following drugs used in this study were purchased from Selleck Biotech: brigatinib (S8229), lorlatinib (S7536), gefitinib (S1025), osimertinib (S7297), afatinib (S1011), poziotinib (S7358), crizotinib (S1068), and entrectinib (S7998). Trastuzumab deruxtecan (T-DXd) was kindly provided by Daiichi Sankyo.

Patients' characteristics and clinical information

Patients' characteristics, including age, sex, Eastern Cooperative Oncology Group performance status, smoking history, and medical history, were collected. According to the Union for International Cancer Control classification for each tumor, tumor characteristics, including histology and tumor-node-metastasis (TNM) stage, were noted. Biomarker data, including gene analysis results of companion diagnostics and clinical sequencing, were collected from medical records. Prior treatment characteristics, including chemotherapy regimens, cycles of chemotherapy, and best response, were collected.

Animal experiments

Four-to-six weeks old female SCID-Beige mice (CB17.Cg-PrkdcscidLystbg-J/CrlCrlj, Charles River Laboratories) were purchased. Mice were housed in sterile filter-capped polycarbonate cages, maintained in a barrier facility under a 12-hour light/dark cycle, and provided sterilized food and water. All invasive procedures were performed by intraperitoneal administration of three types of mixed anesthesia (medetomidine hydrochloride, Meiji Seika Pharma; midazolam, Maruishi Pharmaceutical; betorfal tartrate, Meiji Seika Pharma) or inhalation of isoflurane (Zoetis Japan) to reduce the pain of the experimental animals. After receiving TG2 or TG3 frozen PDX tumor samples, the tumors were thawed and implanted into both flanks of SCID-Beige mice under anesthesia. When the tumors grew, they were transplanted into the right flank of multiple mice and used for drug efficacy tests. Mouse body weight and tumor growth were assessed twice weekly. Tumors were measured with a caliper, and tumor volumes were calculated using the formula: tumor volume (mm3) = (tumor length (mm) × [tumor width (mm)]2)/2. When the tumor growth reached approximately 500–1,000 mm3, they were cut into 2 mm cubic and transplanted into multiple mice for drug efficacy testing.

When the implanted tumors grew to approximately 150–250 mm3, grouping of the PDX mice was conducted using software calculations (EXSUS, CAC Croit) to minimize tumor volume differences between groups. Each PDX drug efficacy study consisted of three or more groups (4–10 mice per group): a control group and two or more drug treatment groups. The correspondence between the PDX and the drugs administered is shown in Fig. 1B. T-DXd was administered intravenously via tail vein on day 0, and all other drugs including vehicle (normal saline) were administered orally once a day for 21 consecutive days. The dosage of each drug was set at an appropriate amount based on previous reports (28–32). All mice were euthanized due to cervical dislocation, and the blood and tumor tissue were collected at 24 hours after conduction of the last drug administration on day 21. Tumor samples were snap frozen by liquid nitrogen after collection and stored at −80°C until analysis. Blood samples were collected in EDTA-2K tubes, centrifuged (1,200 × g for 10 minutes at 4°C), and then the plasma was stored at −80°C until analysis.

The antitumor effect was shown as %Growth (%Gr) by comparing the tumor growth increment in the control at day 0 and treatment groups at day 21.

  • ΔC = (Tumor volume of the control group at day 21) − (Tumor volume of the control group at day 0)

  • ΔT = (Tumor volume of treatment group at day 21) − (Tumor volume of treatment group at day 0)

  • %Gr = 100 × (ΔTC)

Histologic and IHC analysis

PDX tissues were fixed in 10% buffered formalin and then subjected to hematoxylin and eosin (H&E) staining, and a qualified pathologist confirmed NSCLC diagnoses and histopathologic characteristics. Frozen tumor sections of xenografts were also stained with H&E after MSI analysis. For IHC staining of LC-001, LC-002, and LC-004, frozen sections were fixed in 10% buffered formalin for 10 minutes at 25°C. Sections were then incubated with 0.3% H2O2 in methanol at 25°C after washing with 1× TBS. Sections were incubated in serum-free protein block solution (X0909; Dako) for 15 minutes at 25°C, followed by incubation with primary antibody overnight at 4°C: ALK (D5F3) XP Rabbit mAb (#3633, 1:500; Cell Signaling Technology), ROS1 (D4D6) Rabbit mAb (#3287, 1:250; Cell Signaling Technology), or HER2 (4B5) (Roche). After washing with 1× TBS, signal stain boost IHC detection reagent (#8114, anti-rabbit; Cell Signaling Technology) was added and incubated for 30 minutes at 25°C. Sections were reacted with diaminobenzidine (#8059S; Cell Signaling Technology) and counterstained with hematoxylin. A BZ-X710 microscope (Keyence) was used for histologic photography.

Nucleic acid extraction

Nucleic acids were extracted using pleural fluid (Patient B and D) or formalin-fixed, paraffin-embedded (FFPE) samples (Patient A and C) from patient tumors and fresh frozen tumors from PDX tumors. Genetic analysis was performed using patient tumors and PDX tumors of TG1, TG2, and TG3 in LC-001 to LC-003, patient tumors and PDX tumors of TG1 in LC-004, and patient tumors and PDX tumors of TG1 and TG2 in LC-005. DNA and RNA were extracted from FFPE samples using a GeneRead DNA FFPE Kit (Qiagen) and an RNeasy FFPE Kit (Qiagen), and from pleural fluid and fresh frozen tumors using either a KingFisher Cell and Tissue DNA Kit (Thermo Fisher Scientific) or a MagMAX mirVana Total RNA Isolation Kit (Thermo Fisher Scientific).

Whole-exome sequencing, including mutation calls and copy-number analysis

Whole-exome sequencing (WES) samples were prepared using a Twist Comprehensive Exome Kit (Twist Bioscience). Library concentration was measured by a Qubit 2.0 Fluorometer (Thermo Fisher Scientific). Library quality was considered sufficient if a single peak between 250–350 bp was detected by the Agilent 2200 TapeStation system (Thermo Fisher Scientific). Sequencing analysis was performed on an Illumina NovaSeq 6000 system (Illumina). Paired-end WES reads were aligned to the combined human (hg38) and Mus musculus (mm10) reference genome using Burrows-Wheeler Aligner (27). Both somatic synonymous and nonsynonymous mutations were called using our in-house caller and two publicly available mutation callers: Genome Analysis Toolkit (https://gatk.broadinstitute.org/hc/en-us) MuTect2 and VarScan2 (http://varscan.sourceforge.net/). Mutations were discarded if any of the following criteria were met: the total read number was <20, the variant allele frequency in tumor samples was <0.05, the mutant read number in the germline control samples was >2, the mutation occurred in only one strand of the genome, or the variant was present in normal human genomes in either the 1000 Genomes Project dataset (https://www.internationalgenome.org/) or our in-house database. Gene mutations were annotated by SnpEff (https://pcingola.github.io/SnpEff/). The copy-number status was analyzed using our in-house pipeline, which determines the log R ratio (LRR) as follows: (i) SNP positions in a homozygous state [variant allele frequency (VAF): ≤0.05 or ≥0.95] or heterozygous state (VAF: 0.4–0.6) in the genomes of respective normal samples were selected from the 1000 Genomes Project database (ii); normal and tumor read depths at the selected position were adjusted on the basis of the GpC percentage of a 100-bp window flanking the position (33) (iii), the LRR was calculated as the |{\log _2}( {{t_i}/{n_i}} )$|⁠, where |{n_i}$| and |{t_i}$| represent the normal and tumor-adjusted depths at position |i$|⁠, respectively; and (iv) each representative LRR was determined by the median of a moving window (1 Mb) centered at position |i$|⁠. The LRRs of the copy numbers of alleles, the major allele, and the minor allele were determined for every region of the genome.

Transcriptome sequencing, expression analysis, and detection of fusion genes

Total RNA was treated with DNase I (Thermo Fisher Scientific) and subjected to poly (A)-RNA selection prior to cDNA synthesis. The library used for RNA sequencing was prepared using a NEBNext Ultra Directional RNA Library Prep Kit (NEB), following the manufacturer's protocol.

Sequencing was conducted from both ends of each cluster using a NovaSeq 6000 (Illumina). RNA-sequencing reads were aligned to the merged hg38 and mm10 reference genome using STAR (https://github.com/alexdobin/STAR) with the annotated human (release 31) and Mus musculus (release M24) GTF files distributed by GENCODE (https://www.gencodegenes.org/).

Pharmacokinetic analysis

LC/MS-MS was employed to measure the concentration of each drug (free fraction). For plasma preparation, briefly, 20 μL mouse plasma was mixed with 100 μL internal standard solutions in MeOH for protein precipitation. All samples were vortexed for 10 seconds and centrifuged at 12,000 × g for 10 minutes at 4°C. The supernatants were separated and applied for LC/MS-MS analysis. For tumor samples, three consecutive tumor sections, before and after the section for MSI, respectively, were collected in two polypropylene tubes. Tumor sections were homogenized in 50 μL methanol by vortex mixing for 1 minute. Then, the internal standard solution in MeOH was added and vortex mixed for 10 seconds. After centrifuging at 12,000 × g for 10 minutes, the supernatant was collected. A portion of the supernatant was applied for LC/MS-MS analysis, and the rest was used for protein quantification using a Qubit 2.0 Fluorometer (Qubit Protein Assay Kit, #Q33211, Thermo Fisher Scientific). The drug concentration in the tissue section was calculated by dividing drug quantity (measured by LC/MS-MS) by the protein amount in the corresponding tissue extract.

Instruments used for LC/MS-MS analysis were the Nexera X2 UHPLC system (Shimadzu) connected with a QTRAP5500 mass spectrometer (AB SCIEX). The electrospray ionization interface in the positive mode was used to perform MS-MS. The multiple reaction monitoring transitions for all compounds (Supplementary Table S1A), the LC conditions for separating (Supplementary Table S1B), and the optimized MS-MS parameters for each compound (Supplementary Table S1C) are presented. The developed methods were validated in terms of specificity, linearity, accuracy, and precision using blank mouse plasma spiked with standard solutions.

MSI

For quantification drug concentrations and imaging drug distributions in tumor tissues, the frozen tissues were successively sliced into 8-μmol/L-thick slices at −20°C with a cryomicrotome (Leica CM 1950; Leica Biosystems) and placed on the adhesive glass slides coated with an indium tin oxide (SI0100N; Matsunami Glass). We used matrix-assisted laser desorption/ionization mass spectrometry imaging (MALDI-MSI) in this study because MALDI-MSI is the most widely used visualization technique in pharmacokinetic studies (24). α-CHCA was applied to the aforementioned section tissue surface at 250°C for 8 minutes using a sublimation apparatus (SVC-700TMSG/7PS80; Sanyu Electron). Subsequently, 10 mg/mL α-CHCA in ultrapure water/acetonitrile/IPA/TFA = 60/30/10/0.1, v/v/v/v was sprayed stepwise onto the sections using a sprayer (PS270; GSI Creos; refs 24, 27). MALDI-MSI analysis was performed in positive ionization mode using iMScope (Shimadzu), which consisted of an optical microscope and an ion trap time-of-flight analyzer with an atmospheric pressure MALDI source consisting of a 335 nm Nd:YAG laser. The protonated molecules of drugs were fragmented by low-energy collision-induced dissociation with Ar gas; the isolation width was 3 Da. Fragment patterns are illustrated in Supplementary Fig. S1. One fragment ion each drug was selected, and that ion was used for imaging of drugs. The mass tolerance was 50 mDa. Selected ions and the setting values of laser power and collision energy are listed in Supplementary Table S1D. The laser was fired 200 shots at a repetition rate of 1,000 Hz per spot; the pitch of the laser shot was 50 μmol/L. Imaging data were obtained using Imaging MS Solution (ver. 1.20; Shimadzu) and analyzed using Imaging IMAGEREVEAL (ver. 1.12; Shimadzu). The chromatogram for each compound is shown in Supplementary Table S2.

Statistical analysis

All data were analyzed using GraphPad Prism ver. 8.3 (GraphPad software) and SAS ver. 9.4 (SAS Institute). For drug testing, statistically significant differences between control and each group at day 21 were tested as follows: (i) Bartlett's test was used as a test for the equality of k variances (ii); If the k sampled populations had equal variances (P > 0.05 by Bartlett test), Dunnett multiple comparison test was performed; otherwise, Steel test was used. P < 0.05 was considered statistically significant.

Data availability

The data generated in this study are publicly available in National Bioscience Database Center (NBDC)/the Japan Science and Technology Agency (JST) under the accession number JGAS000413.

Establishment of lung cancer PDXs

From August 2018 to May 2020, a total of 245 lung cancer specimens were registered for J-PDX, 21 for small cell lung cancer (SCLC), and 224 for NSCLC (34). In total, 55 PDXs were established, including eight SCLC PDXs and 47 NSCLC PDXs. Of these NSCLC PDXs, 15 PDXs reported genetic aberrations in the clinical information: seven EGFR mutations, five ALK fusion, one ROS1 fusion, one RET fusion, and one HER2 amplification (Fig. 1A). The overall establishment rate was 22%, and by histologic subtype, SCLC accounted for 38% and NSCLC accounted for 21%. The establishment rate was 16% for adenocarcinoma and 35% for squamous cell carcinoma, and the establishment rate tended to be higher in chemotherapy-resistant specimens than in chemotherapy-naïve specimens (Supplementary Table S3A). The establishment rate by genetic alteration is shown in Supplementary Table S3B. The establishment rate of EGFR mutation-positive cases was 16% (7/43) in all and 19% (7/36) in the TKI-resistant samples. In contrast, the establishment was not found in the TKI-naïve samples. The establishment rates of ALK fusion, ROS1 fusion, HER2 amplification, MET skipping, and RET fusion were 21% (5/24), 33% (1/3), 100% (1/1), 0% (0/1), and 25% (1/4), respectively. Overall, NSCLC with genetic aberrations tended to have a higher establishment rate in treatment-resistant specimens than in treatment-naïve specimens.

NSCLC PDXs with driver gene alteration

We used five of the 15 established NSCLC PDXs and tested their efficacy against TKIs and several unapproved drugs administered to patients. Table 1 shows the clinical characteristics of the 4 patients, PDX sample origins, and collection points of tumors from patients and corresponding PDXs. The clinical driver gene for each patient is as follows: Patient A, ALK fusion; Patient B, EGFR exon 21 L858R mutation; Patient C, ROS1 fusion; and Patient D, HER2 amplification. Two PDXs, LC-001 and LC-002, were established by two longitudinal pleural fluid samples in patient A, and a single PDX was established in each of the other patients.

Table 1.

Patient characteristics.

TNMSmokingPDX samplePt sequence sample
PatientAgeSexHistologyClinical driver geneStagingStatusTreatmentPDX No.(Collection point)(Collection point)Gene alteration
           Pt: MDM2 amp (CN = 19.6) 
       First line: alectinib  Pleural effusion Pleural effusion TG1: EML4-ALK, MDM2 amp (CN = 23.7) 
        LC-001    
       Second line: CBDCA+PEM  (Before fifth line) (Before fifth line) TG2: MDM2 amp (CN = 21.7) 
     T2aN3M1a  Third line: ceritinib    TG3: EML4-ALK, MDM2 amp (CN = 23.2) 
70 Adeno EML4-ALK (IHC)  Never      
     IV  Fourth line: TS-1    Pt: MDM2 amp (CN = 19.6) 
       Fifth line: lorlatinib  Pleural effusion Pleural effusion TG1: EML4-ALK, MDM2 amp (CN = 22.8) 
        LC-002    
       Sixth line: alectinib  (After fifth line) (Before fifth line) TG2: EML4-ALK, MDM2 amp (CN = 19.8) 
           TG3: EML4-ALK, MDM2 amp (CN = 20.6) 
           Pt: EGFR ex21 L858R (29.5%) 
    EGFR T3N3M0  First line: gefitinib  Bronchoscopic specimen FFPE TG1: EGFR ex21 L858R (53.9%), EGFR amp (CN = 6.6), MET amp (CN = 6.0) 
42 Adeno   Never  LC-003    
    ex21 L858R (Dako) IIIB  Second line: clinical trial  (After second line) (Time of diagnosis) TG2: EGFR ex21 L858R (50.2%), EGFR amp (CN = 6.3), MET amp (CN = 6.1) 
           TG3: EGFR ex21 L858R (51.8%), EGFR amp (CN = 6.1), MET amp (CN = 4.9) 
       First line: CDDP+PEM     
    EZR-ROS1 TxN3M1a  Second line: nivolumab  Pleural effusion FFPE Pt: PIK3CA (21.6%) 
46 Adeno   Former  LC-004    
    (RT-PCR) IVa  Third line: TS-1  (After third line) (Time of diagnosis) TG1: EZR-ROS1, PIK3CA (100%) 
       Fourth line: crizotinib     
       First line: CBDCA+nab-PTX     
           Pt: TP53 (33.4%), SMARCA4 (43.8%) 
    HER2 T2bN2M1a  Second line: atezolizumab  Pleural effusion Pleural effusion  
59 Adeno   Never  LC-005   TG1: TP53 (100%), SMARCA4 (100%) 
    Amplification (OCA) IVa  Third line: nivolumab  (After second line) (After second line)  
       Fourth line: TS-1    TG2: TP53 (100%), SMARCA4 (100%) 
TNMSmokingPDX samplePt sequence sample
PatientAgeSexHistologyClinical driver geneStagingStatusTreatmentPDX No.(Collection point)(Collection point)Gene alteration
           Pt: MDM2 amp (CN = 19.6) 
       First line: alectinib  Pleural effusion Pleural effusion TG1: EML4-ALK, MDM2 amp (CN = 23.7) 
        LC-001    
       Second line: CBDCA+PEM  (Before fifth line) (Before fifth line) TG2: MDM2 amp (CN = 21.7) 
     T2aN3M1a  Third line: ceritinib    TG3: EML4-ALK, MDM2 amp (CN = 23.2) 
70 Adeno EML4-ALK (IHC)  Never      
     IV  Fourth line: TS-1    Pt: MDM2 amp (CN = 19.6) 
       Fifth line: lorlatinib  Pleural effusion Pleural effusion TG1: EML4-ALK, MDM2 amp (CN = 22.8) 
        LC-002    
       Sixth line: alectinib  (After fifth line) (Before fifth line) TG2: EML4-ALK, MDM2 amp (CN = 19.8) 
           TG3: EML4-ALK, MDM2 amp (CN = 20.6) 
           Pt: EGFR ex21 L858R (29.5%) 
    EGFR T3N3M0  First line: gefitinib  Bronchoscopic specimen FFPE TG1: EGFR ex21 L858R (53.9%), EGFR amp (CN = 6.6), MET amp (CN = 6.0) 
42 Adeno   Never  LC-003    
    ex21 L858R (Dako) IIIB  Second line: clinical trial  (After second line) (Time of diagnosis) TG2: EGFR ex21 L858R (50.2%), EGFR amp (CN = 6.3), MET amp (CN = 6.1) 
           TG3: EGFR ex21 L858R (51.8%), EGFR amp (CN = 6.1), MET amp (CN = 4.9) 
       First line: CDDP+PEM     
    EZR-ROS1 TxN3M1a  Second line: nivolumab  Pleural effusion FFPE Pt: PIK3CA (21.6%) 
46 Adeno   Former  LC-004    
    (RT-PCR) IVa  Third line: TS-1  (After third line) (Time of diagnosis) TG1: EZR-ROS1, PIK3CA (100%) 
       Fourth line: crizotinib     
       First line: CBDCA+nab-PTX     
           Pt: TP53 (33.4%), SMARCA4 (43.8%) 
    HER2 T2bN2M1a  Second line: atezolizumab  Pleural effusion Pleural effusion  
59 Adeno   Never  LC-005   TG1: TP53 (100%), SMARCA4 (100%) 
    Amplification (OCA) IVa  Third line: nivolumab  (After second line) (After second line)  
       Fourth line: TS-1    TG2: TP53 (100%), SMARCA4 (100%) 

Abbreviations: ALK, anaplastic lymphoma kinase; CBDCA, carboplatin; CDDP, cisplatin; Dako, Dako EGFR pharmDx; EGFR, epidermal growth factor receptor; F, female; HER2, human epidermal growth factor receptor type 2; IHC, immunohistochemistry; M, male; nab-PTX, nab-paclitaxel; OCA, oncomine comprehensive assay ver3; PEM, pemetrexed; Pt, patient; ROS1, c-ros oncogene 1; RT-PCR, real-time polymerase chain reaction; TNM, tumor–node–metastasis.

Gene analysis

To confirm changes in genetic alterations from passaging tumor grafts, we performed WES and RNA sequencing of clinical samples and PDX tumors. The patient tumor samples used in the analysis, the time of collection, and its contrast to PDX tumors with representative genetic abnormalities are shown in Table 1. LC-001 sequencing analysis showed ALK fusions in TG1 and TG3 PDX tumors, but not in the patient tumor or TG2 PDX tumor, possibly due to the specimen collection site, quantity, or quality. LC-004 showed EZR-ROS1 fusion in PDX tumors. Although HER2 amplification was reported in the clinical sequence by Oncomine Comprehensive Assay v3 (Thermo Fisher Scientific) in LC-005, neither HER2 amplification nor gene mutation was observed in the patient tumors or PDX tumors in this study. On the other hand, HER2 immunostaining of PDX tumors consistently showed strong protein expression of HER2, and RNA sequencing also showed high expression levels in both the patient and PDX tumors (Supplementary Fig. S2). Overall, the clinical driver genes were found to be largely maintained in PDX tumors. Somatic mutations with allele frequencies of 10% or higher and genes with copy-number amplifications of six or higher in WES are shown in Supplementary Tables S4 and S5, respectively.

Histopathologic comparison of PDX models by generation

We compared the histologic findings from the original patient tumors and each passaged PDX tumor (TG1–3; Fig. 2). Morphologically, in LC-005 PDX, the original and PDX tumors had similar morphology. In the other PDX tumors, some of the original tumor morphology was preserved, but there was a tendency for PDX tumors to be less differentiated histologically than the original tumors.

Figure 2.

Histologic findings of patient tissue and PDX tissue. H&E staining of patient tissue and PDX (TG1-3) tissue. Overall, the original morphology was maintained after establishment and passage of PDXs, but increased malignancy was observed in some PDXs. The H&E staining of Patient A's specimen at diagnosis is shown as the patient tumor in LC-001 and LC-002. X20. TG, transgeneration.

Figure 2.

Histologic findings of patient tissue and PDX tissue. H&E staining of patient tissue and PDX (TG1-3) tissue. Overall, the original morphology was maintained after establishment and passage of PDXs, but increased malignancy was observed in some PDXs. The H&E staining of Patient A's specimen at diagnosis is shown as the patient tumor in LC-001 and LC-002. X20. TG, transgeneration.

Close modal

Patients and PDX models

We conducted an in vivo drug test using the five NSCLC PDXs with driver gene alterations. Figure 3 shows the original patient's clinical course and the drug efficacy in the corresponding PDXs. The drugs selected were either TKIs administered to the original patient or administered after the PDX sample collection, and drugs with suggested efficacy against the respective gene alterations.

Figure 3.

Evaluation of the drug treatment in vivo. The treatment course of the original patient in each PDX, the time of PDX specimen collection, and the results of PDX drug efficacy studies are shown. The top band shows the chemotherapy regimen, best response, and progression-free survival (month). A, shows LC-001 and 002 generated from Patient A (ALK fusion), B shows LC-003 generated from Patient B (EGFR mutation), C shows LC-004 generated from Patient C (ROS1 fusion), and D shows LC-005 generated from Patient D (HER2 amplification). N means the number of mice used in each group. P value indicates Dunnett multiple comparison adjusted P value comparing between control and each group at day 21 after two-way ANOVA. BW (%) indicates percent bodyweight change in each group at day 0 and day 21. Error bars represent the mean ± SEM. CBDCA, carboplatin; CDDP, cispatin; nabPTX, albumin-bound formulation paclitaxel; NE, not evaluable; PD, progressive disease; PEM, pemetrexed; PR, partial response; SD, stable disease.

Figure 3.

Evaluation of the drug treatment in vivo. The treatment course of the original patient in each PDX, the time of PDX specimen collection, and the results of PDX drug efficacy studies are shown. The top band shows the chemotherapy regimen, best response, and progression-free survival (month). A, shows LC-001 and 002 generated from Patient A (ALK fusion), B shows LC-003 generated from Patient B (EGFR mutation), C shows LC-004 generated from Patient C (ROS1 fusion), and D shows LC-005 generated from Patient D (HER2 amplification). N means the number of mice used in each group. P value indicates Dunnett multiple comparison adjusted P value comparing between control and each group at day 21 after two-way ANOVA. BW (%) indicates percent bodyweight change in each group at day 0 and day 21. Error bars represent the mean ± SEM. CBDCA, carboplatin; CDDP, cispatin; nabPTX, albumin-bound formulation paclitaxel; NE, not evaluable; PD, progressive disease; PEM, pemetrexed; PR, partial response; SD, stable disease.

Close modal

Patient A: EML4-ALK fusion

Patient A received first-line treatment for ALK-positive lung adenocarcinoma with alectinib, followed by various cytotoxic agents and ceritnib, and then lorlatinib was administered as fifth-line therapy (Fig. 3A). LC-001 was established from the pleural effusion at the early stage of lorlatinib treatment, and LC-002 was established from the pleural effusion at the time of disease progression of lorlatinib. ELM4-ALK fusion was maintained in almost all PDX tumors. However, there was no secondary mutation in the ALK gene or other reported resistance mechanisms. The drug test was conducted using LC-001 and LC-002 with either lorlatinib, a clinically administered ALK inhibitor, or brigatinib, a novel ALK inhibitor not approved at the time. Both lorlatinib and brigatinib showed a constant efficacy in LC-001, but LC-002 showed a decreased sensitivity to brigatinib. Clinically, there was a rapid increase in pleural effusion within a short period of 2 weeks of lorlatinib administration, suggesting a discrepancy between the PDX results and the clinical response.

Patient B: EGFR ex21 L858R with EGFR and MET amplification

Patient B received gefitinib as the first-line treatment for EGFR mutation-positive lung adenocarcinoma with a short-term progression of 2.5 months and presented with negative results for the T790M mutation (Fig. 3B). The patient was subsequently enrolled in a clinical trial and demonstrated an exacerbation at 6 months. After the second exacerbation, LC-003 was established from a bronchoscopic specimen and tested for efficacy with gefitinib, a clinically refractory drug, and osimertinib, a novel third-generation EGFR-TKI that had not been administered previously. Gene analysis of tumor tissue at diagnosis and PDX tumors showed that EGFR ex21 L858R was maintained at a constant allele frequency, whereas ex20 T790M was not found. On the other hand, copy-number analysis showed EGFR and MET amplification, suggesting that these amplifications resulted in EGFR-TKI resistance. In vivo, consistent with the clinical effect, no antitumor effect was observed with gefitinib treatment, whereas osimertinib showed some therapeutic effect.

Patient C: EZR-ROS1 fusion

Patient C received platinum combination therapy and an immune checkpoint inhibitor for lung adenocarcinoma, with ROS1 fusion identified during the course of the therapy (Fig. 3C). LC-004 was established from the pleural effusion before crizotinib administration. Patient C responded well to crizotinib, but an explained sudden death on the 20th day after initiation of crizotinib made it impossible to evaluate the treatment efficacy. The drug testing of the clinically administered crizotinib and the recently approved drug for ROS1 fusion, entrectinib, showed significant in vivo antitumor activity.

Patient D: HER2 amplification

Patient D had a HER2 amplification-positive lung adenocarcinoma treated with a platinum combination and immune checkpoint inhibitors (Fig. 3D). LC-005 was established from a pleural effusion after second-line atezolizumab exacerbation. We conducted a drug testing using the pan-HER inhibitors afatinib and poziotinib and the anti-HER2 antibody–drug conjugate T-DXd, which recently showed dramatic efficacy against HER2-positive breast and gastric cancer (35, 36). For pan-HER inhibitors, poziotinib showed a significant promising effect, whereas afatinib showed no antitumor effect. On the other hand, T-DXd showed significant tumor shrinkage in a dose-dependent manner.

Drug exposure and target delivery

For the eight TKIs tested in this study, we analyzed drug concentrations in plasma and tumor samples to explore the reasons for the discrepancy between the clinical and PDX models (Table 2). The 21-day dosing period in humans is the potential required duration for plasma concentrations to reach a steady-state phase; however, in the PDX model, the mouse plasma concentrations of the eight drugs administered were low for all drugs and below the lowest limit of quantitation (2 ng/mL) for osimertinib, gefitinib, and afatinib. The tumor drug concentrations, which were normalized to protein amounts in tumor sections, tended to be higher than the plasma level for brigatinib, crizotinib, and entrectinib but below plasma concentration levels for the other drugs. These results suggest that it is difficult to infer the therapeutic relationship between antitumor activity in PDXs and clinical efficacy from the blood and intratumor drug concentrations.

Table 2.

Drug concentration in PDX mice plasma and tumor samples.

TumorPlasma
PDX No.Drug(mean ± SD, ng/mg)(mean ± SD, ng/mL)
LC-001 Brigatinib 25 ± 6.2 5.5 ± 1.0 
 Lorlatinib 6.2 ± 4.0 39 ± 25 
LC-002 Brigatinib 40 ± 27 5.6 ± 2.9 
 Lorlatinib 2.4 ± 0.8 18 ± 5.7 
LC-003 Osimertinib 3.7 ± 1.6 <2 
 Gefitinib 2.1 ± 0.8 <2 
LC-004 Crizotinib 253 ± 119 43 ± 33 
 Entrectinib 156 ± 124 55 ± 19 
LC-005 Afatinib ND ND 
 Poziotinib ND 1.2 ± 0.6 
TumorPlasma
PDX No.Drug(mean ± SD, ng/mg)(mean ± SD, ng/mL)
LC-001 Brigatinib 25 ± 6.2 5.5 ± 1.0 
 Lorlatinib 6.2 ± 4.0 39 ± 25 
LC-002 Brigatinib 40 ± 27 5.6 ± 2.9 
 Lorlatinib 2.4 ± 0.8 18 ± 5.7 
LC-003 Osimertinib 3.7 ± 1.6 <2 
 Gefitinib 2.1 ± 0.8 <2 
LC-004 Crizotinib 253 ± 119 43 ± 33 
 Entrectinib 156 ± 124 55 ± 19 
LC-005 Afatinib ND ND 
 Poziotinib ND 1.2 ± 0.6 

Abbreviations: ND, not detectable; PDX, patient-derived xenograft; SD, standard deviation.

Visualized drug distribution

Because LC/MS measurement is based on homogenized tumor tissue and greatly affected by blood adhesion, it cannot assess the drug's spatial distribution. Therefore, MSI was performed to assess intratumoral structures and spatial distribution of drugs. All eight TKIs used in this study were examined, and the three drugs, brigatinib, crizotinib, and entrectinib, were effectively analyzed. H&E staining, IHC for each TKI target protein (ALK protein in brigatinib, ROS1 protein in crizotinib and entrectinib), and MSI are shown in Fig. 4A. The intratumoral distribution of each drug was independent of the localization of the corresponding target protein. Comparison of the region of interest between the tumor and non-tumor sites showed a clear accumulation at tumor sites for brigatinib but no obvious difference by tissue structure for crizotinib and entrectinib (Fig. 4B; Supplementary Table S6).

Figure 4.

MSI. A, H&E staining, immunostaining of target proteins, and MSI images of PDX tumor tissues treated with each drug. B, Comparison of drug intensities at tumor and non-tumor sites in the H&E images. Brigatinib shows accumulation at tumor sites, whereas crizotinib and entrectinib show overall accumulation.

Figure 4.

MSI. A, H&E staining, immunostaining of target proteins, and MSI images of PDX tumor tissues treated with each drug. B, Comparison of drug intensities at tumor and non-tumor sites in the H&E images. Brigatinib shows accumulation at tumor sites, whereas crizotinib and entrectinib show overall accumulation.

Close modal

In this study, we focused on NSCLC PDXs with genetic alterations and explored the factors that influence the establishment rate of PDXs. We found that PDXs were more likely to be established in patients with a history of treatment with molecular targeted drugs. Among the 15 NSCLC PDXs with genetic alterations that could be established, we selected five PDXs generated from 4 patients and performed genomic analysis. The genetic alterations observed in clinical tests in the patients were maintained in PDXs after passaging, confirming that the PDXs inherited the driver gene alterations. When we tested the efficacy of clinically administered drugs and molecularly targeted therapies against each target, we found that some PDXs were correlated with the therapeutic effects of the patients (in the case of gefitinib and crizotinib), while others were not (in the case of lorlatinib). To elucidate this difference, the blood and intratumor drug concentrations were examined and found to be comparably low on day 21. MSI was performed to analyze the tissue structure and drug distribution in the tumor in detail. The tumor distribution could be evaluated for three of the eight drugs, and accumulation in the tumor site could be confirmed for each drug. Nevertheless, no relationship with drug efficacy could be found.

The establishment rate of lung cancer PDXs has previously been reported to be approximately 30%–50% (18, 19, 37, 38). Several reports have associated the establishment rates with advanced stage, squamous cell carcinoma, and lack of genetic mutations (19, 38, 39). In our study, 55 of 245 (22%) PDXs were established with lung cancer as a whole, and 14 of 76 (20%) PDXs were established with genetically altered NSCLC based on clinical information (Supplementary Table S3). PDXs generated in tumor specimens from patients who did not receive TKIs had particularly low establishment rate percentages, and notably, no PDX was established from TKI-naïve specimens with EGFR mutation (seven implantations), ALK fusion (12 implantations), or MET skipping (one implantation). On the other hand, several PDXs could be established in tumors that became TKI resistant. We previously analyzed the entire J-PDX library using multivariate analysis to examine factors that affected the establishment rate. Patients with a short survival time following tumor tissue collection and advanced disease demonstrate a higher establishment rate (34). In the current study, we focused on the effect of TKI treatment history in lung cancer cases and there was a higher establishment rate in TKI-resistant cases compared with that in the TKI-naïve cases; however, it was not statistically significant (resistant vs. naïve: 25% vs. 11%, P = 0.15). This could be attributed to the shorter survival time in TKI-resistant cases compared with that in the TKI-naïve cases and the faster tumor growth occurring after resistance. The mechanisms of resistance after subjection to the administered molecularly targeted therapies are subdivided, considering the EGFR mutations and ALK fusion genes; the therapeutic strategies following resistance development are explored in each subpopulation. In the case of genetic abnormalities with small populations, such as ROS1 and RET fusion genes, and EGFR and ALK resistance subpopulations, it is extremely difficult to confirm the efficacy in the individual population in large-scale clinical trials. The evaluation of drug efficacy using PDX is extremely useful for such rare gene abnormalities and rare resistance mechanisms among major cancers; hence, we believe that active creation of PDX in the drug-resistant phase will be useful for future drug development and for overcoming resistance.

We performed WES and RNA sequencing on the five PDXs used in the study. Previous reports have shown that genetic alterations in PDXs increase copy-number variations by passaging, but alterations that favor tumor growth, such as driver gene alterations, are retained (40). Among the PDXs analyzed in this study, the failure to confirm ALK fusion in LC-001 may be due to the specimen used for analysis. We could not confirm HER2 amplification in LC-005 by analyzing the clinical specimens and PDX tumors; however, robust HER2 expression was maintained as verified by IHC and RNA sequencing. This suggests that HER2 positivity, in this case, is potentially caused by a mechanism other than genetic mutation or amplification. A large dataset is not available for determining the number of times a genetic analysis should be performed in PDX and for ascertaining the frequency of occurrence of genetic variations with each successive generation. In the current study, we performed genetic mutation analysis of up to three passages in five PDXs and observed that at least driver mutations were retained. However, mutations such as gene deletions and point mutations could occur through passaging. Hence, in the future, it will be necessary to comprehensively evaluate genetic changes that occur through passaging. Overall, our results showed that the driver gene alterations observed in clinical practice were generally maintained in PDXs, indicating the validity of evaluating drug efficacy in PDXs established from patients with genetic alterations.

In the drug testing of the selected five PDXs, lorlatinib resistance in clinical practice was not reproduced in LC-001 and LC-002, which was established over time before and after lorlatinib administration. On the other hand, brigatinib sensitivity was altered in these PDXs, suggesting that tumor characteristics were altered by the short lorlatinib treatment period of 2 weeks. LC-003 showed resistance to gefitinib, consistent with the clinical response. Genetic analysis showed amplification of EGFR and MET genes, which may be involved in resistance. Interestingly, osimertinib, for which EGFR and MET gene amplification has been reported as a mechanism of resistance, showed efficacy in LC-003, suggesting that it may have had some clinical efficacy (41, 42). Furthermore, LC-004 was administered with crizotinib for ROS1 fusion-positive lung cancer and showed good antitumor efficacy as in clinical practice (43, 44). In LC-005, pan-HER inhibitors were used to treat HER2-overexpressing PDXs, with poziotinib showing tumor reduction, while afatinib showed no efficacy. Besides, we evaluated the efficacy of T-DXd, a recently approved novel ADC for HER2, and found a dose-dependent tumor reduction (35, 36). These results support the clinical efficacy of T-DXd for HER2-positive NSCLC, which was recently demonstrated in DESTINY-LUNG01 (NCT03505710; ref. 45). In summary, our results were correlated with the clinical efficacy of certain drugs, such as gefitinib for LC-003 and crizotinib for LC-004, but also differed from clinical efficacy in other drugs, such as lorlatinib for LC-001 and LC-002.

Analysis of blood and intratumor drug concentrations was conducted to reveal factors that correlate between PDX drug efficacy and clinical response. To the best of our knowledge, this is the first report on the evaluation of blood levels of TKIs in PDX mice. Surprisingly, both the blood and intratumor concentrations of all drugs were low at 24 hours after conduction of the last drug administration. We could not find an evident association with drug efficacy; however, we confirmed that even low trough concentrations were sufficient to show efficacy in PDX mice models. We also explored the distribution of drugs in tumors and successfully visualized findings with brigatinib, crizotinib, and entrectinib. We previously analyzed the spatial distribution of various anticancer agents in tumors using MSI analysis (24–27, 46–48). The spatial distribution of drugs is heterogeneous and the intratumor distribution of alectinib is not influenced by the expression of its target proteins (26). In the current study, crizotinib and entrectinib were homogenously distributed throughout the tissues, while brigatinib tended to be heterogeneously concentrated in the tumor site; the drug distributions were markedly not influenced by the expression of the target proteins. For other drugs, imaging using MSI was difficult, possibly due to ionization of the drug or the low effective intratumor drug concentration. The blood concentration of TKIs could be evaluated using PDX and the spatial distribution in tumors could be analyzed using MSI. No evident relationship with drug efficacy was observed; this could be attributed to the setting of the specimen collection points. To evaluate the relationship between drug efficacy and drug distribution level, it is important to understand the relationship at an optimized point over time. Analysis of the spatial distribution of drugs is useful for clarifying the mode of action of drugs. Evaluation of the differences in the distribution patterns among drugs and the effects of target protein expression or tumor stroma is helpful for drug design and for formulation of treatment strategy. We previously performed longitudinal needle biopsies at 4, 9, and 14 days after alectinib administration in a cell line transplant model and successfully assessed MSI in biopsy specimens (27). In PDX, evaluation of concentration and spatial distribution using blood and tumor samples over time will enable pharmacokinetic analysis under conditions similar to those in humans. It is necessary to evaluate systemic and intratumor pharmacology using PDX to accurately estimate the pharmacokinetics in humans.

In all five PDX models, drug efficacy was assessed to contrast the clinical responses of the 4 donor patients with PDX models. Four to 10 mice were used in each group, and a huge variation in tumor growth rate was observed among PDXs. This may have been due to the heterogeneity of each tumor fragment. Conventionally, a 2 mm cubic tumor fragment is transplanted into a single mouse. If the tumor tissue is separated and the number of cells to be transplanted equalized, it could possibly reduce this heterogeneity. However, separation may disrupt the tumor microenvironment and result in loss of cell–cell interaction. Therefore, we transplanted the tumor “as is”. Previously, Townsend and colleagues used acute lymphoblastic leukemia PDXs to compare the number of mice per group (between 1 and 3) and reported that 1 mouse could predict the same treatment effect (49). However, in the case of solid tumors in our current study, there is more heterogeneity than in hematologic tumors, encouraging the use of a larger number of mice. Our results showed a large variation in the tumor growth rate, and it would be desirable to use a large number of mice (preferably 6 mice per group) to evaluate the efficacy of drugs. Another important factor to consider is the location of the tumor implantation. Considering the merits of tumor growth, orthotopic models in which the tumor is transplanted into the same organ of origin are desirable. However, as in lung cancer, transplantation of tumors into the body is difficult due to problems of technique and evaluation of tumor size. For this reason, we used an ectopic model as the basis of our study. However, this aspect should be examined in the future.

PDXs, in which the genetic alterations of original patients are maintained, are useful for screening drug development. In addition, the drug exposure level differs greatly between humans and mice. Cell line–derived xenograft models are often administered at the MTD in mice, resulting in blood concentrations that cannot be achieved in clinical practice; the drug is often assessed to be effective in vivo (50). In humans, the recommended dose in the phase I trial is set much lower than that in in vivo dosing due to potential adverse events with increasing doses. Whether the drug should be administered at a higher dose to match cell line–derived xenografts or rather at a lower dose to match humans in drug efficacy studies in PDXs will require further investigation.

In conclusion, we have successfully reproduced the treatment effect in patients with NSCLC PDXs with genetic alterations for some drugs. Further investigation is needed to establish appropriate evaluation methods and tools to reproduce and predict clinical outcomes with higher accuracy. We believe it is important to understand the blood concentration, intratumor concentration, and distribution in tumors using PDX.

S. Yagishita reports grants from Japan Society for the Promotion of Science and Japan Agency for Medical Research and Development during the conduct of the study; grants from Boehringer Ingelheim Japan outside the submitted work. Y. Hayashi reports personal fees from CMIC Pharma Science during the conduct of the study. Y. Matsumoto reports grants from Hitachi, Ltd., Hitachi High-Technologies; personal fees from Olympus, AstraZeneca, Novartis, COOK, Thermo Fisher Scientific, Erbe Elektromedizin GmbH, Fujifilm, Chugai, and Eli Lilly outside the submitted work. H. Horinouchi reports personal fees from Lilly, Taiho, Novartis, KyowaKirin; grants and personal fees from AstraZeneca, Chugai, MSD, BMS, and ONO; grants from Daiichi Sankyo Abbvie outside the submitted work. Y. Ohe reports grants and personal fees from AstraZeneca, Amgen, Bristol-Myers Squibb, Chugai, Eli Lilly, Janssen, Kyorin, MSD, Nippon Kayaku, Novartis, ONO Pharmaceutical, Taiho, Takeda Pharmaceutical, and Pfizer; personal fees from Boehringer Ingelheim, Celtrion; grants from Kissei outside the submitted work. N. Motoi reports grants from ONO and NEC; personal fees from AstraZeneca, Roche Diagnostics, Chugai Pharm, Daiichi-Sankyo Pharm, Takeda Pharm, MSD, Beckton Dickinson Japan, Covidien Japan Inc, and Novartis outside the submitted work. Y. Yatabe reports personal fees from AstraZeneca and MSD, grants and personal fees from Thermo Fisher Science, and grants from Invitae outside the submitted work. H. Mano reports other support from CureGene Co. Ltd., Ono Pharmaceutical, Daiichi Sankyo, PFDeNA, Konica Minolta, Ambry Genetics, Chordia Therapeutics, and non-financial support and other support from CMIC Pharmaceutical outside the submitted work. A. Hamada reports grants from Japan Agency for Medical Research and Development (AMED), Cyclic Innovation for Clinical Empowerment (CiCLE), National Cancer Center Research and Development Fund, and non-financial support from Daiichi Sankyo co. ltd during the conduct of the study; grants from Daiichi Sankyo co. ltd, Konica Minolta, Chugai pharmaceutical, Chordia Therapeutics, CMIC Pharma Science, Eisai, Eli Lilly, Healios, LSI medience, and Tosoh outside the submitted work. No disclosures were reported by the other authors.

H. Jo: Conceptualization, data curation, software, formal analysis, validation, investigation, visualization, methodology, writing–original draft. S. Yagishita: Conceptualization, resources, data curation, software, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. Y. Hayashi: Software, formal analysis, validation, investigation, visualization, methodology. S. Ryu: Software, formal analysis, validation, investigation, visualization, methodology, writing–original draft. M. Suzuki: Formal analysis, validation, investigation, visualization, methodology. S. Kohsaka: Software, formal analysis, supervision, validation, investigation, visualization, methodology, writing–original draft. T. Ueno: Software, formal analysis, validation, investigation, methodology. Y. Matsumoto: Resources. H. Horinouchi: Resources. Y. Ohe: Resources. S.-i. Watanabe: Resources. N. Motoi: Resources, supervision, visualization, methodology. Y. Yatabe: Resources, formal analysis, supervision, visualization. H. Mano: Resources, supervision, funding acquisition. K. Takahashi: Conceptualization, resources, supervision. A. Hamada: Conceptualization, resources, formal analysis, supervision, funding acquisition, methodology, writing–original draft, project administration, writing–review and editing.

We are grateful to all the patients, patients' family, oncologists, allied healthcare professionals, research concierges, and research assistants who participated in this study. We would like to thank Daiichi Sankyo Co. Ltd., for providing T-DXd. We also thanks to Hideaki Kakinuma and Ako Takahashi for cooperation. This work was partly supported by the Japan Agency for Medical Research and Development (AMED), Cyclic Innovation for Clinical Empowerment (CiCLE) under grant number 17pc0101011h0001 (H. Mano and A. Hamada), AMED Research on Regulatory Science of Pharmaceuticals and Medical Devices under grant number 20mk0101181h0001 (S. Yagishita), JSPS KAKENHI grant number 18K15296 (S. Yagishita), and National Cancer Center Research and Development Fund (31-A-3, A. Hamada), Japan.

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.

1.
Mullard
A
. 
Parsing clinical success rates
.
Nat Rev Drug Discov
2016
;
15
:
447
.
2.
Paul
SM
,
Mytelka
DS
,
Dunwiddie
CT
,
Persinger
CC
,
Munos
BH
,
Lindborg
SR
, et al
How to improve R&D productivity: the pharmaceutical industry's grand challenge
.
Nat Rev Drug Discov
2010
;
9
:
203
14
.
3.
Hidalgo
M
,
Amant
F
,
Biankin
AV
,
Budinska
E
,
Byrne
AT
,
Caldas
C
, et al
Patient-derived xenograft models: an emerging platform for translational cancer research
.
Cancer Discov
2014
;
4
:
998
1013
.
4.
Girotti
MR
,
Lopes
F
,
Preece
N
,
Niculescu-Duvaz
D
,
Zambon
A
,
Davies
L
, et al
Paradox-breaking RAF inhibitors that also target SRC are effective in drug-resistant BRAF mutant melanoma
.
Cancer Cell
2015
;
27
:
85
96
.
5.
Bertotti
A
,
Migliardi
G
,
Galimi
F
,
Sassi
F
,
Torti
D
,
Isella
C
, et al
A molecularly annotated platform of patient-derived xenografts (“xenopatients”) identifies HER2 as an effective therapeutic target in cetuximab-resistant colorectal cancer
.
Cancer Discov
2011
;
1
:
508
23
.
6.
Siolas
D
,
Hannon
GJ
. 
Patient-derived tumor xenografts: transforming clinical samples into mouse models
.
Cancer Res
2013
;
73
:
5315
9
.
7.
Tentler
JJ
,
Tan
AC
,
Weekes
CD
,
Jimeno
A
,
Leong
S
,
Pitts
TM
, et al
Patient-derived tumour xenografts as models for oncology drug development
.
Nat Rev Clin Oncol
2012
;
9
:
338
50
.
8.
Gao
H
,
Korn
JM
,
Ferretti
S
,
Monahan
JE
,
Wang
Y
,
Singh
M
, et al
High-throughput screening using patient-derived tumor xenografts to predict clinical trial drug response
.
Nat Med
2015
;
21
:
1318
25
.
9.
Herbst
RS
,
Morgensztern
D
,
Boshoff
C
. 
The biology and management of non-small cell lung cancer
.
Nature
2018
;
553
:
446
54
.
10.
Camidge
DR
,
Doebele
RC
,
Kerr
KM
. 
Comparing and contrasting predictive biomarkers for immunotherapy and targeted therapy of NSCLC
.
Nat Rev Clin Oncol
2019
;
16
:
341
55
.
11.
Arbour
KC
,
Riely
GJ
. 
Systemic therapy for locally advanced and metastatic non-small cell lung cancer: a review
.
JAMA
2019
;
322
:
764
74
.
12.
Ettinger
DS
,
Wood
DE
,
Aggarwal
C
,
Aisner
DL
,
Akerley
W
,
Bauman
JR
, et al
NCCN guidelines insights: non-small cell lung cancer, version 1.2020
.
J Natl Compr Canc Netw
2019
;
17
:
1464
72
.
13.
Duma
N
,
Santana-Davila
R
,
Molina
JR
. 
Non-small cell lung cancer: epidemiology, screening, diagnosis, and treatment
.
Mayo Clin Proc
2019
;
94
:
1623
40
.
14.
Dagogo-Jack
I
,
Shaw
AT
. 
Tumour heterogeneity and resistance to cancer therapies
.
Nat Rev Clin Oncol
2018
;
15
:
81
94
.
15.
Gerlinger
M
. 
Targeted drugs ramp up cancer mutability
.
Science
2019
;
366
:
1452
3
.
16.
Billingham
L
,
Malottki
K
,
Steven
N
. 
Research methods to change clinical practice for patients with rare cancers
.
Lancet Oncol
2016
;
17
:
e70
e80
.
17.
Jiang
Y
,
Zhao
J
,
Zhang
Y
,
Li
K
,
Li
T
,
Chen
X
, et al
Establishment of lung cancer patient-derived xenograft models and primary cell lines for lung cancer study
.
J Transl Med
2018
;
16
:
138
.
18.
Kang
HN
,
Choi
JW
,
Shim
HS
,
Kim
J
,
Kim
DJ
,
Lee
CY
, et al
Establishment of a platform of non-small-cell lung cancer patient-derived xenografts with clinical and genomic annotation
.
Lung Cancer
2018
;
124
:
168
78
.
19.
Kita
K
,
Fukuda
K
,
Takahashi
H
,
Tanimoto
A
,
Nishiyama
A
,
Arai
S
, et al
Patient-derived xenograft models of non-small cell lung cancer for evaluating targeted drug sensitivity and resistance
.
Cancer Sci
2019
;
110
:
3215
24
.
20.
Morgan
KM
,
Riedlinger
GM
,
Rosenfeld
J
,
Ganesan
S
,
Pine
SR
. 
Patient-derived xenograft models of non-small cell lung cancer and their potential utility in personalized medicine
.
Front Oncol
2017
;
7
:
2
.
21.
Nakajima
T
,
Geddie
W
,
Anayama
T
,
Ko
HM
,
da Cunha Santos
G
,
Boerner
S
, et al
Patient-derived tumor xenograft models established from samples obtained by endobronchial ultrasound-guided transbronchial needle aspiration
.
Lung Cancer
2015
;
89
:
110
4
.
22.
Zhang
F
,
Wang
W
,
Long
Y
,
Liu
H
,
Cheng
J
,
Guo
L
, et al
Characterization of drug responses of mini patient-derived xenografts in mice for predicting cancer patient clinical therapeutic response
.
Cancer Commun
2018
;
38
:
60
.
23.
Kim
HR
,
Kang
HN
,
Shim
HS
,
Kim
EY
,
Kim
J
,
Kim
DJ
, et al
Co-clinical trials demonstrate predictive biomarkers for dovitinib, an FGFR inhibitor, in lung squamous cell carcinoma
.
Ann Oncol
2017
;
28
:
1250
9
.
24.
Nishidate
M
,
Hayashi
M
,
Aikawa
H
,
Tanaka
K
,
Nakada
N
,
Miura
SI
, et al
Applications of MALDI mass spectrometry imaging for pharmacokinetic studies during drug development
.
Drug Metab Pharmacokinet
2019
;
34
:
209
16
.
25.
Nishimura
M
,
Hayashi
M
,
Mizutani
Y
,
Takenaka
K
,
Imamura
Y
,
Chayahara
N
, et al
Distribution of erlotinib in rash and normal skin in cancer patients receiving erlotinib visualized by matrix assisted laser desorption/ionization mass spectrometry imaging
.
Oncotarget
2018
;
9
:
18540
7
.
26.
Ryu
S
,
Hayashi
M
,
Aikawa
H
,
Okamoto
I
,
Fujiwara
Y
,
Hamada
A
. 
Heterogeneous distribution of alectinib in neuroblastoma xenografts revealed by matrix-assisted laser desorption ionization mass spectrometry imaging: a pilot study
.
Br J Pharmacol
2018
;
175
:
29
37
.
27.
Ryu
S
,
Ohuchi
M
,
Yagishita
S
,
Shimoi
T
,
Yonemori
K
,
Tamura
K
, et al
Visualization of the distribution of nanoparticle-formulated AZD2811 in mouse tumor model using matrix-assisted laser desorption ionization mass spectrometry imaging
.
Sci Rep
2020
;
10
:
15535
.
28.
Cross
DA
,
Ashton
SE
,
Ghiorghiu
S
,
Eberlein
C
,
Nebhan
CA
,
Spitzler
PJ
, et al
AZD9291, an irreversible EGFR TKI, overcomes T790M-mediated resistance to EGFR inhibitors in lung cancer
.
Cancer Discov
2014
;
4
:
1046
61
.
29.
Zhang
S
,
Anjum
R
,
Squillace
R
,
Nadworny
S
,
Zhou
T
,
Keats
J
, et al
The potent ALK inhibitor brigatinib (AP26113) overcomes mechanisms of resistance to first- and second-generation ALK inhibitors in preclinical models
.
Clin Cancer Res
2016
;
22
:
5527
38
.
30.
Robichaux
JP
,
Elamin
YY
,
Tan
Z
,
Carter
BW
,
Zhang
S
,
Liu
S
, et al
Mechanisms and clinical activity of an EGFR and HER2 exon 20-selective kinase inhibitor in non-small cell lung cancer
.
Nat Med
2018
;
24
:
638
46
.
31.
Yun
MR
,
Kim
DH
,
Kim
SY
,
Joo
HS
,
Lee
YW
,
Choi
HM
, et al
Repotrectinib exhibits potent antitumor activity in treatment-naive and solvent-front-mutant ROS1-rearranged non-small cell lung cancer
.
Clin Cancer Res
2020
;
26
:
3287
95
.
32.
Ardini
E
,
Menichincheri
M
,
Banfi
P
,
Bosotti
R
,
De Ponti
C
,
Pulci
R
, et al
Entrectinib, a Pan-TRK, ROS1, and ALK inhibitor with activity in multiple molecularly defined cancer indications
.
Mol Cancer Ther
2016
;
15
:
628
39
.
33.
Yoon
S
,
Xuan
Z
,
Makarov
V
,
Ye
K
,
Sebat
J
. 
Sensitive and accurate detection of copy number variants using read depth of coverage
.
Genome Res
2009
;
19
:
1586
92
.
34.
Yagishita
S
,
Kato
K
,
Takahashi
M
,
Imai
T
,
Yatabe
Y
,
Kuwata
T
, et al
Characterization of the large-scale Japanese patient-derived xenograft (J-PDX) library
.
Cancer Sci
2021
;
112
:
2454
66
.
35.
Shitara
K
,
Bang
YJ
,
Iwasa
S
,
Sugimoto
N
,
Ryu
MH
,
Sakai
D
, et al
Trastuzumab deruxtecan in previously treated HER2-positive gastric cancer
.
N Engl J Med
2020
;
382
:
2419
30
.
36.
Modi
S
,
Saura
C
,
Yamashita
T
,
Park
YH
,
Kim
SB
,
Tamura
K
, et al
Trastuzumab deruxtecan in previously treated HER2-positive breast cancer
.
N Engl J Med
2020
;
382
:
610
21
.
37.
Izumchenko
E
,
Paz
K
,
Ciznadija
D
,
Sloma
I
,
Katz
A
,
Vasquez-Dunddel
D
, et al
Patient-derived xenografts effectively capture responses to oncology therapy in a heterogeneous cohort of patients with solid tumors
.
Ann Oncol
2017
;
28
:
2595
605
.
38.
John
T
,
Kohler
D
,
Pintilie
M
,
Yanagawa
N
,
Pham
NA
,
Li
M
, et al
The ability to form primary tumor xenografts is predictive of increased risk of disease recurrence in early-stage non-small cell lung cancer
.
Clin Cancer Res
2011
;
17
:
134
41
.
39.
Stewart
EL
,
Mascaux
C
,
Pham
NA
,
Sakashita
S
,
Sykes
J
,
Kim
L
, et al
Clinical utility of patient-derived xenografts to determine biomarkers of prognosis and map resistance pathways in EGFR-mutant lung adenocarcinoma
.
J Clin Oncol
2015
;
33
:
2472
80
.
40.
Ben-David
U
,
Ha
G
,
Tseng
YY
,
Greenwald
NF
,
Oh
C
,
Shih
J
, et al
Patient-derived xenografts undergo mouse-specific tumor evolution
.
Nat Genet
2017
;
49
:
1567
75
.
41.
Nakatani
K
,
Yamaoka
T
,
Ohba
M
,
Fujita
KI
,
Arata
S
,
Kusumoto
S
, et al
KRAS and EGFR amplifications mediate resistance to rociletinib and osimertinib in acquired afatinib-resistant NSCLC harboring Exon 19 Deletion/T790M in EGFR
.
Mol Cancer Ther
2019
;
18
:
112
26
.
42.
Leonetti
A
,
Sharma
S
,
Minari
R
,
Perego
P
,
Giovannetti
E
,
Tiseo
M
. 
Resistance mechanisms to osimertinib in EGFR-mutated non-small cell lung cancer
.
Br J Cancer
2019
;
121
:
725
37
.
43.
Shaw
AT
,
Riely
GJ
,
Bang
YJ
,
Kim
DW
,
Camidge
DR
,
Solomon
BJ
, et al
Crizotinib in ROS1-rearranged advanced non-small-cell lung cancer (NSCLC): updated results, including overall survival, from PROFILE 1001
.
Ann Oncol
2019
;
30
:
1121
6
.
44.
Shaw
AT
,
Ou
SH
,
Bang
YJ
,
Camidge
DR
,
Solomon
BJ
,
Salgia
R
, et al
Crizotinib in ROS1-rearranged non-small-cell lung cancer
.
N Engl J Med
2014
;
371
:
1963
71
.
45.
Smit
EF
,
Nakagawa
K
,
Nagasaka
M
,
Felip
E
,
Goto
Y
,
Li
BT
, et al
Trastuzumab deruxtecan (T-DXd; DS-8201) in patients with HER2-mutated metastatic non-small cell lung cancer (NSCLC): interim results of DESTINY-Lung01
.
J Clin Oncol
38
:
15s
, 
2020
;
38
(
suppl; abstr 9504
).
46.
Hayashi
Y
,
Ohuchi
M
,
Ryu
S
,
Yagishita
S
,
Hamada
A
. 
A procedure for method development and protein binding ratio as the indicator of sensitivity with anticancer agents on MALDI mass spectrometry imaging
.
Drug Metab Pharmacokinet
2021
;
38
:
100385
.
47.
Nishidate
M
,
Yamamoto
K
,
Masuda
C
,
Aikawa
H
,
Hayashi
M
,
Kawanishi
T
, et al
MALDI mass spectrometry imaging of erlotinib administered in combination with bevacizumab in xenograft mice bearing B901L, EGFR-mutated NSCLC cells
.
Sci Rep
2017
;
7
:
16763
.
48.
Tsubata
Y
,
Hayashi
M
,
Tanino
R
,
Aikawa
H
,
Ohuchi
M
,
Tamura
K
, et al
Evaluation of the heterogeneous tissue distribution of erlotinib in lung cancer using matrix-assisted laser desorption ionization mass spectrometry imaging
.
Sci Rep
2017
;
7
:
12622
.
49.
Townsend
EC
,
Murakami
MA
,
Christodoulou
A
,
Christie
AL
,
Koster
J
,
DeSouza
TA
, et al
The public repository of xenografts enables discovery and randomized phase II-like trials in mice
.
Cancer Cell
2016
;
29
:
574
86
.
50.
Thompson
J
,
Stewart
CF
,
Houghton
PJ
. 
Animal models for studying the action of topoisomerase I targeted drugs
.
Biochim Biophys Acta
1998
;
1400
:
301
19
.