Genomic profile of metastatic breast cancer patient-derived xenografts established using percutaneous biopsy

Abstract

Background: Metastatic breast cancer (mBC) is a complex and life-threatening disease and although it is difficult to cure, patients can benefit from sequential anticancer treatment, including endocrine therapy, targeted therapy and cytotoxic chemotherapy. The patient-derived xenograft (PDX) model is suggested as a practical tool to predict the clinical outcome of this disease as well as to screen novel drugs. This study aimed to establish PDX models in Korean patients and analyze their genomic profiles and utility for translational research.

Methods: Percutaneous core needle biopsy or punch biopsy samples were used for xenotransplantation. Whole exome sequencing and transcriptome analysis were performed to assess the genomic and RNA expression profiles, respectively. Copy number variation and mutational burden were analyzed and compared with other metastatic breast cancer genomic results. Mutational signatures were also analyzed. The antitumor effect of an ATR inhibitor was tested in the relevant PDX model.

Results: Of the 151 cases studied, 40 (26%) PDX models were established. Notably, the take rate of all subtypes, including the hormone receptor-positive (HR +) subtype, exceeded 20%. The PDX model had genomic fidelity and copy number variation that represented the pattern of its donor sample. TP53, PIK3CA, ESR1, and GATA3 mutations were frequently found in our samples, with TP53 being the most frequently mutated, and the somatic mutations in these genes strengthened their frequency in the PDX model. The ESR1 mutation, CCND1 amplification, and the APOBEC signature were significant features in our HR + HER2- PDX model. Fulvestrant in combination with palbociclib showed a partial response to the relevant patient's tumor harboring the ESR1 mutation, and CCND1 amplification was found in the PDX model. AZD6738, an ATR inhibitor, delayed tumor growth in a relevant PDX model.

Conclusions: Our PDX model was established using core needle biopsy samples from primary and metastatic tissues. Genomic profiles of the samples reflected their original tissue characteristics and could be used for the interpretation of clinical outcomes.

Keywords: Metastatic breast cancer; Patient-derived xenograft; Whole-exome sequencing.

Fig. 1

Flow chart of WES and transcriptome data processing. Detailed information on the data mapping process and the variant calling process is given

Fig. 2

The statistics of PDX model establishment and fidelity between the PDX model and donor. a The total numbers of enrollment and take cases and those subdivided (enrollment cases and take cases) are represented by a bar graph. The take rate is also marked. b The take rate by tissue origin is represented by a bar graph. c The take rate by Ki-67 expression in patient tissues is represented by a bar graph. d The subtype composition of the established PDX model and its tissue of origin. The subtype was compared with PDX IHC analysis and clinical records. e The somatic mutation patterns of the patient tissue and its corresponding PDX model. The case number with IMT indicates the results of patient tissue analysis, and the case number with IMX indicates the results of PDX tissue analysis. The list of genes is ordered by mutational frequency, and the number of somatic mutations of each sample is marked in the figure. f CNV concordance between patient and PDX tissues. The correlation was determined by Pearson’s correlation analysis using the log2-fold change value of each sample

Fig. 2

The statistics of PDX model establishment and fidelity between the PDX model and donor. a The total numbers of enrollment and take cases and those subdivided (enrollment cases and take cases) are represented by a bar graph. The take rate is also marked. b The take rate by tissue origin is represented by a bar graph. c The take rate by Ki-67 expression in patient tissues is represented by a bar graph. d The subtype composition of the established PDX model and its tissue of origin. The subtype was compared with PDX IHC analysis and clinical records. e The somatic mutation patterns of the patient tissue and its corresponding PDX model. The case number with IMT indicates the results of patient tissue analysis, and the case number with IMX indicates the results of PDX tissue analysis. The list of genes is ordered by mutational frequency, and the number of somatic mutations of each sample is marked in the figure. f CNV concordance between patient and PDX tissues. The correlation was determined by Pearson’s correlation analysis using the log2-fold change value of each sample

Fig. 3

Genomic profile characteristics of the Korean metastatic breast PDX model. a Somatic mutations in our PDX model are arranged according to frequency. The samples are arranged by their subtype. The mutation burden of each sample is also presented. The genes are listed according to their mutation frequency, from high to low. b The mutation frequencies of TP53, PIK3CA, and ESR1 were analyzed by subtype and are represented by a bar graph (a comparison with the other datasets is also shown). c Mutation burden of each subtype. d The mutation frequency of actionable mutation genes. The mutations that were detected in the patient and PDX tumor samples derived from the same patient were compared

Fig. 3

Genomic profile characteristics of the Korean metastatic breast PDX model. a Somatic mutations in our PDX model are arranged according to frequency. The samples are arranged by their subtype. The mutation burden of each sample is also presented. The genes are listed according to their mutation frequency, from high to low. b The mutation frequencies of TP53, PIK3CA, and ESR1 were analyzed by subtype and are represented by a bar graph (a comparison with the other datasets is also shown). c Mutation burden of each subtype. d The mutation frequency of actionable mutation genes. The mutations that were detected in the patient and PDX tumor samples derived from the same patient were compared

Fig. 4

Significant features in the HR + PDX model. a ESR1 mutations in the HR + PDX model. b CCND1 amplification was analyzed by Fisher’s exact test in all of our PDX models. c The FPKM values of the CCND1 gene were analyzed, and those from ESR1-mutated samples and wild-type samples were compared. d Mutation signature analysis was performed, and the signatures were clustered. The signatures depict the 1st and 2nd highest cosine similarity. e, f APOBEC3B expression is presented as its FPKM value by subtype (e) or the PIK3CA gene mutation (f)

Fig. 5

Utility of PDX genomic data and in vivo drug tests for clinical implications. a ESR1 mutation site and its domain in the IMX-158 sample. b CNV in the IMX-158 sample. c PET scan of the donor of the IMX-158 sample. Baseline indicates before the initiation of palbociclib and fulvestrant treatment. The best response was observed four months after initiation. White arrows indicate hypermetabolic lesions in the liver and pelvic bone. d CNV in the X89 sample. e ATR mutation site and its domain in the X89 sample. g The X89 PDX model was treated with the ATR inhibitor AZD6738 for 4 weeks. The tumor volumes are presented as graphs

Fig. 5

Utility of PDX genomic data and in vivo drug tests for clinical implications. a ESR1 mutation site and its domain in the IMX-158 sample. b CNV in the IMX-158 sample. c PET scan of the donor of the IMX-158 sample. Baseline indicates before the initiation of palbociclib and fulvestrant treatment. The best response was observed four months after initiation. White arrows indicate hypermetabolic lesions in the liver and pelvic bone. d CNV in the X89 sample. e ATR mutation site and its domain in the X89 sample. g The X89 PDX model was treated with the ATR inhibitor AZD6738 for 4 weeks. The tumor volumes are presented as graphs