Molecular profiling of patient-derived breast cancer xenografts

Breast cancer (BC) is the most commonly diagnosed cancer and remains the leading cause of worldwide cancer-related death in women [1]. A better understanding of BC biology is essential in order to identify new targeted therapies and tumors with molecular profiles that will respond to the targeted treatment. Gene expression profiling of invasive BC has defined three main tumor subtypes with very specific features (Luminal, Basal, human epidermal growth factor receptor 2 (HER2)) [2]. It is now common knowledge that the pathologic characteristics, array comparative genomic hybridization (aCGH) profiles, gene and miRNA expression profiles and activated pathways are radically different among these subtypes, supporting the view that BC is a disease composed of very different and independent molecular subgroups. These subtypes have also been shown to differ in terms of clinical presentation (that is, differences in axillary lymph node involvement, local and regional recurrence, metastatic patterns and overall prognosis), and in their sensitivity to systemic treatment [3].

Preclinical experimental models that reproduce the heterogeneity of this disease have become a major challenge in order to investigate the biology of each BC subtype and evaluate new targeted therapies. Such models are necessary to examine the treatment efficacy of potential new therapies and some have already contributed to the development of new human therapeutics [4, 5]. However, most of the existing in vivo models used for preclinical trials of anticancer drugs are based on a limited number of cell lines previously isolated from human tumors and selected by cell culture prior to implantation in immunodeficient animals. These models do not reproduce the architecture of the primary tumors [6, 7]. In contrast, tumor xenografts obtained by engraftment of tumor samples transplanted directly into animals seems to be able to reduce the biologic distance between the original patient tumor and the in vivo model. We have previously published a paper describing a large panel of BC xenografts that maintain the cell differentiation, morphology, architecture, vasculature, peripheral growth and some of the molecular features of the original patient's tumor [8].

However, new aberrations are expected to appear in xenografted tumors because of the selection pressure operated by the host animal, the loss of human stroma and the intrinsic genetic instability of breast tumors.

Genomic analyses showed that BC xenografts reflect the genomic profile of the patient's tumors, with additional DNA gains and losses. Gene expression profile analysis showed dynamic variations between tumor pairs (xenograft and primary tumor), with recurrent changes in the expression of a small group of stroma-related genes.

These data suggest that BC xenografts maintain the overall genetic profile of the original tumors, with additional changes that could be explained by adaptation of tumor cells to the new host.

The aim of the present study was to analyze the molecular profiles of a panel of human BC xenografts directly transplanted from patient tumor samples. Tumors transplanted into a host animal are subject to different stroma and selection pressures, which can impact on the tumor's molecular profile. We, therefore, compared the genomic and gene expression profiles of BC xenografts with their corresponding patient tumors. As illustrated by histologic and immunohistochemical examinations, and as previously reported, these xenografts maintained the biologic markers of the patient's primary tumors as well as their microscopic morphology [8].

Characterization and comparison of CNA and clustering analysis showed that the original genetic profile was generally maintained in the xenograft. Accurate quantification of homology is a complex procedure, as breast carcinomas display intratumor heterogeneity [19, 20] and xenografts are obtained by selecting only a small fragment of the tumor sample after surgery. The correlation between CNA profiles of pairs of xenografts and primary tumors was noteworthy, being greater than 0.5 for most of the paired samples (16/18). The similarity between genomic profiles appeared to be greater for triple-negative and luminal tumors than for HER2+ tumors, which could be explained by the fact that Her2 positive tumors are less markedly altered and are composed largely of stromal cells (see Figure 1), leading to underestimation of the correlation coefficient calculated on tumor DNA alterations. Analysis of the genomic profiles showed that the majority of tumors had a concordant distribution of chromosomal gains and losses, and maintained amplification regions. However, an enrichment of genomic rearrangements was observed in the xenografts, as previously demonstrated by our team, as well as by other authors [8, 21‐23].

Analysis of recurrent changes observed in xenografts showed enrichment of CNA in certain chromosomes. The majority of these regions are known to be associated with chromosomal imbalances in BC cells derived from either tumor samples or cell lines. In the present analysis, negative differences corresponded to deletions that were not detected in the patient tumor or DNA gains that were lost in the xenograft. Conversely, positive values represented gains observed in xenografts that were not detected in the patient tumor or deletions present in the patient tumor but lost in the xenograft. The greater number of DNA rearrangements observed in the xenografts compared to the patient tumors (Figure 5A) suggests that the CNA differences observed were due to gains or losses occurring in xenografts and not present in the patient tumors, rather than losses and gains in the patient tumors that returned to a normal status in the xenograft. In addition, the majority of regions presenting high frequencies of CNA changes were associated with chromosomal imbalances in BC. This result is concordant with those of a recent paper that analyzed the chromosomal aberrations in a panel of nine patient-derived models of sarcoma. Kresse et al. showed that many CNA changes found in xenografts are frequently observed in sarcoma patients, suggesting that xenografts may in some way represent the genomic rearrangement intrinsic to tumor progression [23]. In the case of breast cancer, Ding et al. studied the pattern of genetic differences between a patient's tumor and the corresponding xenograft. Although their conclusions are based on analysis of a single patient's tumor, this study elegantly demonstrated that many of the mutations detected in the xenograft were also observed in brain metastases derived from the same patient [24]. The fact that genomic alterations are conserved for several years, without any major changes, as demonstrated by sequential CGH analysis after multiple passages in vivo, suggests that genomic profiles remain relatively stable over time despite new selection pressures and loss of human stroma. This finding was also demonstrated by clustering analysis, in which the similarity between different xenograft passages was greater than the similarity between the xenograft and the original tumor, indicating that selection of breast tumor cells at the time of the first tumor engraftment is the major source of genetic variability. This observation can be explained by a selective tumor cell process during in vivo transplantation. In addition, loss of the human stromal compartment in xenografts results in enrichment of human tumor cells, and consequently enrichment of DNA alterations not detected in the patient's tumor.

The amplifications present in the patient's tumors were generally conserved in the xenografts, except for the HBC-8 tumor, in which the changes were limited to DNA gains. This suggests that these genomic regions do not undergo any major variations after grafting tumors into immunocompromised mice. More than 40 (42) amplicons were detected in the panel of BC xenografts (8p11.2-p12, 8q24, 11q13.3, 17q12-q21 and 20q13.3). Several of them have been frequently described in BC patients and carry potential oncogenes, the overexpression of which may be important for initiating, survival and/or development of breast tumors [25, 26].

In terms of gene expression, less than 20% of probe sets showed significant variations on a single comparison and these genes were not enriched in stroma-related components. Investigation of genes with recurrent changes (common to 10 to 13 pairs) identified a small group of 205 genes associated with stromal gene ontology annotations, consistent with changes in the tumor environment from human to mouse. Another study compared gene expression between two BC xenografts and patient tumors [21]. Bergamaschi et al. reported a significant variation in the expression of human extracellular matrix-related genes that were down-regulated in xenografts compared with primary tumors. Interestingly, we also found that about one-half of the genes that were down-regulated in xenografts were correlated with breast carcinoma prognosis (data not shown). This set of genes is enriched in the immune system and is composed of immune response-related genes. Inversely, the set of genes not correlated with prognosis was highly enriched in wound response, cell adhesion, blood vessel development, extracellular matrix organization and cell migration related genes. These results indirectly emphasize the major role of the stromal compartment and the immune system in the prognosis of breast carcinoma [27].

The gene expression profile data may provide important information on pathway activation or cellular targets for novel anticancer agents, and it may also contribute to the identification of genes affecting response to treatment. Additional gene expression studies and pathway signaling analyses will be necessary to complete the molecular characterization of BC xenografts, especially in the context of preclinical development of molecular targeted agents.

These analyses support the use of the primary BC xenografts as preclinical models to study the effect of new anticancer drugs and to identify biologic factors associated with drug response.