Personalizing non-small cell lung cancer treatment through patient-derived xenograft models: preclinical and clinical factors for consideration

Fresh tumor specimens (40 numbers) were obtained from NSCLC patients who underwent surgical removal at Soroka Medical Center, Beer-Sheva, Israel, between 2018 and 2021. After surgery, tumor specimens were sent for pathological assessment and subsequent confirmation and aliquoted for DNA extraction, cryopreservation, and implantation in NSG-SGM3 mice. The specimen process was performed within 2 h following surgical extraction. Written informed consent was obtained from all patients. The ethics committee of Soroka Medical Center (No. 0026-19-SOR) approved this study.

Non-obese diabetic/severe combined immune deficiency (NOD/SCID) (4–5-week-old) triple transgenic NSG-SGM3 mice were obtained from the Jackson Laboratory (Sacramento, CA, USA). Animals were housed in individually ventilated cages at an appropriate temperature (21–25 °C) with a 12-h light/dark cycle and free access to food and water. All experiments and protocols in mice were approved by the Institutional Animal Care and Use Committee of BGU (authorization number IL-59-11-2018-(A), and the Israel Ministry of Health (MOH). All experiments were performed in accordance with relevant guidelines and regulations. This study is reported in accordance with the ARRIVE guidelines.

PDX models were generated as previously described with minor modifications [19, 20]. Briefly, fresh tumor specimens (up to 2 h after surgical restriction) were divided into fragments of approximately 2–3 mm³ and implanted subcutaneously into the flanks of 6–7-week-old NSG-SGM3 mice. Each patient specimen was implanted into 3 NSG-SGM3 mice, which were then monitored for tumor growth for up to 180 days. When the tumor size was > 1000 mm³, the PDX mice were anesthetized with carbon dioxide, and the tumors were surgically removed. Extracted tumor tissue was used for passaging, DNA extraction, and histopathological examination.

Histological staining of tumor biopsies and PDXs was done as follows: 4 μm sections of the sample were fixed in 4% formaldehyde and embedded in paraffin using an automated staining device (Benchmark XT; Ventana Medical Systems, Tucson, AZ, USA). For hematoxylin and eosin (H&E) staining, a pathologist reviewed the pathological type and architecture between the original and xenograft tumors to confirm the diagnosis.

Sample sections were also incubated with the following antibodies: anti-Ki-67 (1:400 dilution; Abcam, Cambridge, UK); anti-TTF1 (Transcription Termination Factor 1) (1:200 dilution; Cell Signaling Technology, Danvers, MA, USA); anti-P40 (1:100 dilution; Cell Signaling Technology); and anti-PD-L1 (programmed death-ligand 1) (1:100 dilution; Abcam) to compare the immunophenotypic characteristics. Images were acquired using a Nikon Eclipse Ci microscope with a Nikon DS-Fi3 camera and NIS-elements software, version 5.21.00. All procedures were done by a medical pathologist in Soroka Medical Center.

DNA was extracted from 8 PDX tissue samples using the QIAcube Connect (Qiagen) with the DNeasy blood and tissue kit according to manufacture instructions (Cat No. 69504). Briefly, tissues were thawed to room temperature, and a piece of ~ 25 mg was cut and placed in a 1.5 ml tube. 180ul ATL buffer and 4 0ul of proteinase K were added to the tissue, and the samples were incubated on a thermomixer at 56 °C, 1000RPM overnight. The samples were then centrifuged at 3 °C at 10,000 × g, and the supernatant was used for the extraction. DNA concentration and quality were evaluated using NanoDrop One (Thermo Scientific). Next, 8 exome libraries were constructed simultaneously according to the manufacture protocol (Illumina DNA Prep with Enrichment, Illumina, Cat No. 20025523) using 700 ng DNA as a starting material. After the amplification of fragmented DNA, the quality of the libraries was determined using the TapeStation 4200 with the High Sensitivity D1000 kit (Agilent Cat No. 5067–5584), and the concentration of the libraries was measured using the Qubit dsDNA HS Assay Kit (Invitrogen, Cat No. Q32851). The libraries were then pooled together for the enrichment stage (200 ng per sample). The enrichment was performed using the xGen™ Exome Research Panel v2 (IDT 4 rxn kit). After the enrichment, the concentration of the library was measured using the Qubit dsDNA HS Assay Kit (Invitrogen, Cat No. Q32851), and the size was determined using the TapeStation 4200 with the High Sensitivity D1000 kit (Agilent Cat No. 5067–5584). Data generation was performed on the Illumina NextSeq2000, P2 200 cycles (Read1-101; Read2-101; Index1-10; Index2-10) (Illumina, Cat No. 20046812). Quality control was assessed using Fastqc (v0.11.8). Reads were trimmed for adapters, low quality 3`, and a minimum length of 20 using CUTADAPT (v1.10). The 100 bp paired-end reads were aligned to the human genome. The reference genome and known SNPs and INDELs databases for genome version hg38 (required by GATK) were downloaded from the GATK bundle.

Statistical analyses were performed using GraphPad Prism 8.4.2. To determine clinical parameters that contributed to the establishment of the PDXs, logistic regression analysis was conducted to evaluate the correlation between the success rates and the clinical, demographic, and pathological parameters. A value of P < 0.05 was considered statistically significant. Statistical methods are indicated in the text and figure legends where applicable.

To study tumor stability and develop a preclinical platform for the treatment evaluation of NSCLC patients, we first established a PDX model platform in NSG-SGM3 mice. NSG-SGM3 mice were selected for our models due to their ability to engraft both myeloid and lymphoid-derived cells stably, thus providing a superior alternative to their backbone, commonly used NSG strain for humanization models. Fresh PDX samples from NSCLC donors were collected during surgical resections and, after pathological confirmation, engrafted in NSG-SGM3 mice. Transplant procedures were performed within two hours of biopsy collection. A total of 40 patient PDXs were engrafted, 20 of which successfully grew and formed propagating tumors in mice after 80 weeks, for a tumor engraftment rate of 50%. Successful PDX transplants were serially passaged in mice to create a "hot" sample bank and aliquoted and cryopreserved for histopathological and molecular analysis. A schematic presentation of the process is presented in Fig. 1A.

To study the correlation between PDX engraftment success rates and patient clinical characteristics, we obtained demographic and clinical data from our cohort of NSCLC donor patients (presented in Table 1). Our data showed that the ratio of males to females was 23:17, and the average age of patients was 62.6 years, ranging from 42 to 80 years. Twenty-seven (67.5%) patients had smoked regularly in the past. The tumor histologic types included adenocarcinoma (n = 27), squamous cell carcinoma (n = 6), neuroendocrine carcinoma (n = 1), and other histological types (n = 6). Samples were resected from lesions in the lung (n = 25), lymph nodes (n = 5), brain metastasis (n = 6), subcutaneous lesions (n = 2), spine (n = 1), and liver (n = 1). Figure 2 displays the clinical features of each participant, including the molecular alterations detected by clinical sequencing panels (if applicable) and the corresponding engrafted PDX take rate. Detailed information about each patient’s clinical background and treatment history are presented in Table S1.

Specific patient clinical features were compared to the engraftment ability of their corresponding PDXs to understand which parameters may affect successful PDX growth. We evaluated clinical metrics, including specific laboratory results on the day of tumor resection, treatment history, and tumor histopathological characteristics. We found no correlations between engraftment and patient levels of white blood cells (WBC count), albumin, and LDH (Fig. 1B). Nor was there a difference in engraftment success between patients who exhibited noticeable weight loss and those who did not. Patients with early-stage resectable tumors were less likely to form PDXs, compared to tumors from patients on advanced treatment lines following progression at the time of resection (Fig. 1B). In addition, a negative correlation was seen between prior chemotherapy treatment duration and engraftment success (Fig. 1B).

We further examined the relationship between tumor pathological characteristics and the rate of tumor engraftment in the PDX model. We did not find a significant difference between samples from the primary tumor and those from metastatic sites (Fig. 1C). Additionally, the histological subtypes did not affect tumor engraftment success rates (Fig. 1C). However, an advanced stage or grade of the disease was linked to higher chances of the successful growth of the PDX (Fig. 1C). Finally, we analyzed the overall survival of all patients and found that the probability of survival was lower among patients with successful PDX growth (Fig. 1D). The specific PDX take rates per patient are detailed in Table S1.

Successfully engrafted NSCLC PDXs were sequentially passaged up to 10 times (P10) in NSG-SGM3 mice, and tumor size was measured twice weekly to study PDX growth kinetics. Growth rates varied between the engrafted tumors, creating different time intervals for passing each PDX, as presented in Fig. 3. Diverseness in growth rates was associated with the implanted tumor's clinical behavior, indicating a positive correlation between the tumor aggressiveness and its ability to propagate successfully in mice and grow more rapidly.

Diverseness was most prominent in the time interval from implantation of the original tumor until the first passage, referred to as P1. This time interval includes the latency period, the time from tumor transplantation until the primary initiation of PDX propagation. Representative growth curves of 5 tumors from initial PDX growth in mice until the first passage are shown in Fig. 3B. Overall, the median time to the first passage was 24.0 weeks. In contrast, the second passage was required after a median period of 10.0 weeks, and a third passage was needed. A statistically significant difference was observed between the first passage and the following two passages (P value = 0.005, shown in Fig. 3C). Additionally, a significantly shorter interval to the first passage was observed in tumors of higher differentiation grades (Fig. 3D). Finally, a negative correlation was found between PDX growth speed until P1 and overall patient survival, as shown in Fig. 3E.

To determine whether NSCLC PDXs retain their original tumors' histopathological characteristics and if these are preserved throughout generations of mice, the original tumor and corresponding PDXs from serial passages were analyzed for histopathological features. Each tissue was stained for hematoxylin, eosin (H&E), and related clinical markers, including TTF1, P40, Ki67, and PDL1. The periodic histopathological results of the 4 selected PDXs are shown in Fig. 4.

We found that the histologic type and differentiation grade of NSCLC PDXs throughout mouse generations were preserved and remained stable compared with those of the initially engrafted tumor. Quantification of molecular biomarkers using standard scoring algorithms revealed few changes between tested biomarkers and PDX passages (Fig. 4C). The levels of the markers TTF1, Ki67, and PDL-1 were generally steady in the early passages of all PDXs; however, the levels were slightly altered in advanced passages (P7-8). Marker P40 showed the highest variation between the different passages of PDXs compared to the original tumor for PDX14 and PDX37.

In the next step, we aimed to determine the genomic fidelity of NSCLC PDXs throughout sequential passages. Whole exome sequencing (WES) was performed on selected PDXs at their first passage (P0) and an advanced passage (P6–P8). The representative results of three PDXs, PDX35, PDX37, and PDX40, are presented in Fig. 5. The genomic analysis showed 864, 485, and 907 non-synonymous SNVs detected in the first passaged PDX35, PDX37, and PDX40 xenografts, respectively. These included a maximum of 20 frameshift deletions in PDX40 P0, 10 frameshift insertions in PDX35 P0, 44 stop gain mutations seen in PDX35 P0, and a maximum of 1 stoploss mutation in PDX40 P0. In the advanced passaged PDXs, 910, 594, and 543 non-synonymous SNVs were detected in PDX 35 P6, PDX37 P7, and PDX40 P8, respectively, with a maximum of 19 frameshift deletions in PDX37 and a maximum of 7 frameshift insertions, 45 stop gain mutations, and 2 stop loss mutations all seen in PDX 35. Annotation distribution is presented in Fig. 5A.

Frequency analysis of the variants identified in the PDXs found that most alterations were rare and were detected in a frequency of less than 0.1. These results were consistent in all PDXs sequenced in their early and late passages. Paired analysis of early and late passage PDXs showed that only PDX35 retained a prominent number of alterations across the early and late passages consisting of 2735 matching mutations. In contrast, PDX37 and PDX40 had minimal similarity between their early and advanced passage samples, with 245 and 942 coequal variants, respectively (Fig. 5B). Although PDXs showed, in general, low-frequency variants, analysis of alterations that occurred in both early and late passages found higher frequency variations (Fig. 5C, D).

We checked whether specific target mutations were retained throughout PDX generations to gain a more clinically relevant perspective. As part of personalized treatment, molecular analyses using liquid or tissue biopsies were performed for all patients in various stages of the disease to detect clinically relevant mutations and follow tumor genomic dynamics. The molecular results of these panels were compared with the genomic results of PDXs' early and late passages.

The comparison analysis presented in Fig. 5E, F shows clinically relevant mutations in PDX35 in the diagnostic stage before tumor resection, including CCND3 amplification, DAXX C629fs*16, NKX2-1 amplification, TP53 F212fs*34, and VEGFA amplification. Only NKX2-1 and TP53 mutations were seen in the early and late passages of the engrafted PDX. As for PDX37, liquid biopsy sequencing found EGFR L858R, EGFR amplification, TP53, and MYC amplification. EGFR L858R mutation was found in the early corresponding PDX in addition to a new EGFR mutation found in exon 25. In the late PDX passage, only an intronic EGFR mutation was detected. TP53 mutations were present in the early and late PDX passages with different nucleotide variants. PDX40 analysis at the diagnostic stage comprised RB1 stop gain mutation, TP53 splice site variant, RASA1 stop gain mutation, and FH frameshift mutation, which all remained in the early but not in the late PDX generation, and KeaP1 mutations, which were not found in the early PDX generation, but a different mutation variant was found in the late PDX generation. TP53 stop gain mutation was seen in the late PDX generation. In addition, several mutations of unknown significance were detected in the clinical setting in PDX35 and PDX40, and they mainly were retained in the early (P0) and in the advanced (P6-8) PDX samples as shown in Fig. 5F.