Establishment of PDXs from NSCLC patient samples
In this study, we collected 62 tumor samples from 62 different patients with newly diagnosed NSCLC, who had not been previously treated. The average age of patients was 63.7 years, ranging from 44–79, 61% were females (38 out of 62 cases), and 27% were smokers (17 out of 62 cases). Most of the patients (71%, 44 out of 62 cases) were diagnosed in early stages of the disease (34 cases in stage I, 10 cases in stage II); and 85% had ADC (53 out of 62 cases), with acinar, mucinous, and solid ADC being the most prevalent subtypes. The primary tumors of these patients originated from various lobes of the lung, among which the left upper lobe (LUL) was the most commonly involved, while the right middle lobe (RML) was the least (Fig. 1).
Of the 62 engraftments, 13 led to the successful establishment of PDXs, which included eight ADCs, four SCCs, and one large-cell neuroendocrine carcinoma (LCNEC), representing a tumor take rate of 21% (Table 1). The generation harboring the patient-derived tumor tissue was termed P0, with subsequent generations numbered consecutively (P1, P2, P3 and so on). Next, we analyzed the relationship between engraftment rate of specimens and clinicopathological parameters in NSCLC patients, and found that histological subtype and clinical stage were significant factors affecting the PDXs engraftment. The success rate of ADC (8/53; 15%) was remarkably lower in comparison with the other subtypes (SCC, 4/7, 57%; LCNEC, 1/1, 100%), and the advanced stage (Ⅲ/Ⅳ; 7/18, 39%) was linked to higher chance of a successful engraftment compared with the early stage (I/II; 6/44, 13.6%). However, other factors, including age, sex, smoking status, primary tumor size, and lymphatic metastasis, did not correlate with the engraftment rate (Table 2). Moreover, patients with successful tumor engraftment had a significantly shorter DFS and OS than those without successful PDX construction (Fig. 2).
Table 1
Clinical and pathological characteristics of 13 NSCLC patients
Patient ID | Sex | Age (years) | Smoking habits | Histology | TNM | Stage | Site | Tumor size (cm3) | Metastasis | Driver oncogene status |
Patient#1 | Female | 50 | No | ADC | T2N3M1 | ⅣA | LLL | 13.50 | Yes | ALK+ |
Patient#3 | Female | 64 | No | SCC | T1bN0M0 | ⅠA2 | RLL | 1.52 | No | WT |
Patient#9 | Male | 68 | Yes | ADC | T1bN2M0 | ⅢA | RUL | 5.78 | Yes | WT |
Patient#18 | Female | 69 | No | ADC | T2aN2M0 | ⅢA | LUL | 12.80 | Yes | KRAS G12C |
Patient#19 | Male | 66 | Yes | LCNEC | T2aN0M0 | ⅠB | LLL | 13.50 | No | WT |
Patient#21 | Male | 58 | Yes | SCC | T2aN0M0 | ⅠB | RLL | 14.60 | No | WT |
Patient#30 | Female | 56 | No | ADC | T1cN0M0 | ⅠA3 | LLL | 19.60 | No | EGFR exon 21 L858R |
Patient#42 | Female | 78 | No | ADC | T1cN2M0 | ⅢA | RUL | 27.20 | Yes | EGFR exon 21 L858R |
Patient#43 | Female | 56 | No | ADC | T1cN0M0 | ⅠA3 | LUL | 27.40 | No | WT |
Patient#44 | Female | 44 | No | ADC | T4NXM1a | ⅣA | LUL | 27.80 | Yes | EGFR exon 19 del |
Patient#46 | Male | 63 | Yes | SCC | T2aN1M0 | ⅡB | LLL | 29.30 | Yes | KRAS G13C |
Patient#49 | Male | 63 | Yes | SCC | T4N0M0 | ⅢA | LUL | 30.34 | No | WT |
Patient#50 | Female | 59 | No | ADC | T2N2M0 | ⅢA | LUL | 31.97 | Yes | EGFR exon 21 L858R |
TNM, tumor node metastasis; ADC, adenocarcinoma; SCC, squamous cell carcinoma; LCNEC, large cell neuroendocrine lung cancer; LUL, left upper lobe; RLL, right lower lobe; RML, right middle lobe; LLL, left lower lobe; RUL, right upper lobe. |
Table 2
Correlation between clinical characteristics of NSCLC patients and establishment of PDXs
Variables | Engrafting | Non-engrafting | Total | Established rate (%) | P value |
Engraftment | 13 | 49 | 62 | 21.0% | |
Age | | | | | 0.541 |
≤ 60 | 6 | 18 | 24 | 25% | |
> 60 | 7 | 31 | 38 | 18.4% | |
Sex | | | | | > 0.999 |
Female | 8 | 30 | 38 | 21.1% | |
Male | 5 | 19 | 24 | 20.8% | |
Smoking status | | | | | 0.319 |
Never | 8 | 37 | 45 | 17.8% | |
Former/ current | 5 | 12 | 17 | 29.4% | |
Histology | | | | | 0.042* |
ADC | 8 | 45 | 53 | 15.1% | |
SCC | 4 | 3 | 7 | 57.1% | |
LCNEC | 1 | 0 | 1 | 100% | |
Others | 0 | 1 | 1 | 0 | |
Stage | | | | | 0.040* |
Ⅰ/Ⅱ | 6 | 38 | 44 | 13.6% | |
Ⅲ/Ⅳ | 7 | 11 | 18 | 38.9% | |
Tumor size (cm3) | | | | | 0.484 |
<11 cm3 | 2 | 14 | 16 | 12.50% | |
≥ 11 cm3 | 11 | 35 | 46 | 23.91% | |
Metastasis | | | | | 0.222 |
Negative | 6 | 32 | 38 | 18.8% | |
Positive | 7 | 17 | 24 | 29.2% | |
Site | | | | | 0.116 |
left lung | 9 | 20 | 29 | 31.03% | |
right lung | 4 | 29 | 33 | 12.12% | |
*Fisher’s exact test |
Tumor growth characteristics with increased passage in mice
To study the growth characteristics of the successfully implanted tumors, PDX tumors were passaged three generations in mice (P1-3), in addition to the original xenograft (P0). Our results showed that the time required for grafts from different patients to grow to 300 mm3 fluctuated between 42 and 177 days in P0 generation, with the average growth time of 116 days. However, the latency time of the subsequent passages became shorter, and the average growth time decreased to 101, 61, and 39 days in P1, P2, and P3 generation, respectively (Fig. 3A). In addition, the growth curves of xenografts from the same tumor tissue in specific passages were similar, but not entirely consistent, which might due to the heterogeneity of tumor cells from a single tumor tissue and the interindividual difference in the immune reactivity of mice. However, the growth curves of xenografts from different pathological tissues in specific passages were significantly different. Among them, LCNEC showed the fastest growth, followed by SCC and ADC (Fig. 3B-D).
Previous reports have shown that a small portion of PDXs may undergo a transformation during passage processes, resulting in the engraftments are of lymphocytic, rather than tumor origin[23, 24]. In the present study, lymphoproliferations occurred in 3 (4.8%) xenografts, despite none of these patients having a prior or subsequent clinical history of lymphoproliferative disease. These PDX tumors exhibited extremely rapid growth, and were strongly positive for EBV early RNA transcript (EBER) and human CD20 antigen, indicating these tumors were EBV-associated human diffuse large B cell lymphoma which was formed by clonal proliferation of human B-cell lymphocytes. In addition, the liver and spleen of these mice were significantly larger than normal mice (Figure S1).
PDX tumors preserve morphologic and genetic features of the primary tumors
To evaluate whether the established PDXs could retain histological characteristics with the primary tumors from patients, we performed histopathological and immunohistochemical (IHC) examinations using all successfully grafted PDX tumors and their corresponding patient tissues. Tumor sections were stained with hematoxylin and eosin (H&E), and immune-stained for clinically relevant biomarkers, including the primary markers for ADC (TTF1 and Napsina), SCC (P63), as well as LCNEC (CD56, Synaptophysin, Chromogranin). H&E staining showed that the PDX tumors were usually poorly differentiated and did not have some structure characteristics of the primary tumors, such as the acinus and nipples of ADC, and the keratinized structure of SCC. However, the expression of biomarkers was positive and coincident in patient and mice tissues, which maintained over multiple passages (Fig. 4).
To further investigate whether the PDXs preserved the genetic profiles of the primary tumors, we performed whole exome sequencing (WES) analysis on tumors from three representative patients (one ADC (Patient#18), one SCC (Patient#21), and one LNENC (Patient#19)) and their corresponding PDX tumors (P3 generation). The total mutation information detected was integrated into the Circos plots (Fig. 5A). Compared to patient tumors, the detected number of single nucleotide polymorphisms (SNPs), insertions, and deletions (INDELs) in the matched PDX tumors were dramatically increased. A total of 18,335, 18,449, and 24,731 SNPs were detected in the tumors of Patient#18, Patient#21, and Patient#19, respectively. However, 31,175, 30,122, and 30,739 SNPs were detected in their matched PDX tumors. Similarly, 1,906, 1,945, and 4,285 INDELs were detected in the tumors of Patient#18, Patient#21, and Patient#19, respectively, while 6,015, 5,662, and 6,689 INDELs were detected in their matched PDX tumors. Approximately 88–99% of SNPs found in the patient tumors were retained in their matched PDXs. Mutations in 40 cancer-associated genes were generally preserved between PDX and patient tumors. Correlation analysis indicated the similarity of somatic mutation in each pair of PDX and their corresponding primary tumor, with R-values all greater than 0.91 (Fig. 5B and 5C). Together, these results show that PDXs retain the genetic profiles of their primary tumors.
Efficacy validation of chemotherapy and optimal treatments selection in PDXs
To further assess the value of our PDXs in clinical individualized treatment for NSCLC patients, three typical NSCLC patients (Patient#18, Patient#21, and Patient#19) with established corresponding PDXs (P3 generation) were evaluated. Patient#18 was a 69-year-old non-smoking female, computed tomography (CT) revealed a 3.4*1.8 cm mass in the left upper lobe, which was pathologically diagnosed as low-differentiated ADC (T2aN2M0, stage ⅢA), accompanied by Kirsten RAS (KRAS) G12C mutation. Patient#21 was a 58-year-old smoking male, CT results showed a 3.1*2.6 cm mass in the right lower lobe. Pathological examination showed a moderately differentiated SCC (T2aN0M0, stage ⅠB). After surgery, both patients received 4 cycles of paclitaxel plus carboplatin treatment, and achieved clinical complete response. Patient#19 was a 66-year-old smoking male, CT scan demonstrated a mass of 2.6*2.2 cm in the left lower lobe, and the pathological diagnosis was LCNEC (T2aN0M0,ⅠB stage). After surgery, this patient received 6 cycles of etoposide plus nedaplatin treatment, and also achieved clinical complete response (Figure S2).
In PDXs, we first validated the therapeutic efficacies of abovementioned conventional chemotherapy regimens, and found that both treatments significantly inhibited the growth of PDX tumors, with the responses were entirely consistent with those of their corresponding patients. However, the mice experienced significant weight loss, especially in ADC and SCC groups, which indicated a high toxicity of these conventional chemotherapeutic drugs. Therefore, we simultaneously investigated the efficacy and toxicity of sotorasib and anlotinib in the ADC and SCC PDX, respectively, based on the genetic status of tumors. Sotorasib is a novel KRAS-G12C inhibitor, and approved for the treatment of adult patients with KRAS G12C-mutated locally advanced or metastatic NSCLC who have received at least one prior systemic therapy[25]. Anlotinib is a multiple TKI and approved for the treatment of patients with locally advanced or metastatic NSCLC who have undergone progression or recurrence after ≥ 2 lines of systemic chemotherapy[26]. As expected, these two targeted drugs showed dramatically better efficacy and lower toxicity compared with the conventional chemotherapeutic agents, and were the optimized treatments for these two patients (Fig. 6). Taken together, these results indicate that the treatment response of PDXs is similar to the clinical results, and can serve as a co-clinical study model for patients to select the optimized regimens.
The involvement of MAPK-ERK signaling pathway in acquired resistance to osimertinib in NSCLC
In the present study, we established an osimertinib-resistant PDX model by continuous oral treatment with osimertinib, which was derived from a 78-year-old non-smoking female with L858R mutation in exon 21 of EGFR (Patient#42). CT scan demonstrated a mass of 4.3*3.3 cm in the right upper lobe, and the pathological diagnosis was ADC (T1cN2M0, ⅢA stage). After surgery, started taking osimertinib 80 mg/day. There were no significant side effects, the tumor shrank and the patient had a clinical complete response (Figure S2). Osimertinib has been effective for 18 months to date.
In order to induce acquired resistance to osimertinib in PDXs, the PDXs (P1 generation) were passaged into six mice, which were randomly divided into osimertinib and control groups. Treatment started when subcutaneously implanted tumor volume ≥ 100mm3. After osimertinib treatment, tumors grew more slowly than those in control group, and began to shrink after 4 weeks. However, after about 15 weeks of continuous osimertinib treatment, tumors increased exponentially, confirming the development of acquired resistance to osimertinib (Fig. 7A). In order to explore the resistance mechanisms and alternative therapies, tumors from osimertinib-resistant and -sensitive PDXs were subjected to WES analysis. Results showed that the detected number of mutations in osimertinib-resistant PDX tumors was dramatically increased, which was about 10 times more frequent than sensitive tumors. Many wild-type genes in sensitive tumors underwent mutation during the development of osimertinib resistance, such as dual-specificity phosphatase 6 (DUSP6), Ras p21 protein activator 1 (RASA1), ATRX, SETD8, etc. However, the common genetic alterations involved in osimertinib resistance, such as EGFR C797S, MET amplification, and BRAF mutation, were not detected in osimertinib-resistant PDXs (Fig. 7B). Then we performed Gene Ontology (GO) terms and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses for the variant genes occurred within osimertinib-resistant PDX tumors. The results suggest that these genes were significantly clustered in the protein phosphorylation term and MAPK-ERK signaling pathway. Interestingly, both DUSP6 and RASA1 can decrease the protein phosphorylation and activity of MAPK-ERK pathway[27, 28], and their genetic mutations might abolish their functions[29, 30]. M62I mutation of DUSP6 was identified in all three osimertinib-resistant PDX tumors, while T846A, N850S, or I931T mutation of RASA1 was detected in at least one osimertinib-resistant PDX tumor. Therefore, the overactivation of MAPK-ERK signaling pathway might play a pivotal role in the development of osimertinib resistance in Patient#42, and the specific inhibitors for this pathway could be considered as an alternative treatment after osimertinib failure.
DUSP6 M62I mutation reduces osimertinib sensitivity in NSCLC
To further clarify the activation status of the MAPK-ERK signaling pathway in osimertinib-sensitive and -resistant PDX tumors, we performed IHC analysis to determine the expression levels of ERK1/2 and phosphorylated ERK1/2 (p-ERK1/2). As expected, we observed no significant difference in the ERK1/2 level between two groups, while the level of p-ERK1/2 was increased in the resistant tumors compared to the sensitive ones (Fig. 8A-C).
It is well known that MAPK-ERK signaling pathway plays a pivotal role in various biological events, including metabolic reprogramming, cell proliferation, survival, and differentiation[31]. DUSP6, a broadly expressed dual-specificity phosphatase protein, has been assumed to bind and dephosphorylate ERK, leading to decreased ERK activity [29]. In our osimertinib-resistant PDX tumors, we found a consensus missense variant, M62I, located within the ERK-binding domain of DUSP6. Previous research has suggested that M62I mutation could reduce the interaction between DUSP6 and ERK, resulting in increased ERK phosphorylation and ERK activity[29]. However, it has not been evaluated so far whether the DUSP6 M62I mutation could influence the sensitivity of osimertinib. Therefore, we overexpressed wild-type (WT) DUSP6 and M62I mutant DUSP6 in PC9 and H1975 cell lines, and found that the effect of osimertinib on p-ERK1/2 inactivation was dramatically enhanced after DUSP6 overexpression, while it was attenuated by M62I DUSP6 mutation. (Fig. 8D-F). Moreover, overexpression of DUSP6 showed a synergistic anti-viability effect with osimertinib in NSCLC cells, but DUSP6 M62I mutation significantly decreased the cellular osimertinib sensitivity. (Fig. 8G-H). Finally, we selected trametinib, a specific inhibitor of the MAPK-ERK signaling pathway[32], to treat osimertinib-resistant PDX mice, and the results showed that the tumor growth was significantly slowed down after combined application of osimertinib and trametinib (Fig. 8I). Collectively, these data demonstrate that the DUSP6 M62I mutation-induced MAPK-ERK pathway overactivation is an important mechanism and therapeutic target of osimertinib resistance in NSCLC.