Intraductal patient derived xenografts of estrogen receptor positive (ER+) breast cancer recapitulate the histopathological spectrum and metastatic potential of human lesions

Estrogen receptor α positive (ER+) or “luminal” breast cancers were notoriously difficult to establish as patient-derived xenografts (PDXs). We and others recently demonstrated that the microenvironment is critical for ER+ tumor cells; by grafting them into milk ducts >90% take rates are achieved and many features of the human disease are recapitulated. This intra-ductal (ID) approach holds promise for personalized medicine, yet human and murine stroma are organized differently and this and other species specificities may limit the value of this model. Here, we analyzed 21 ER+ ID-PDXs histopathologically. We find that ID-PDXs vary in extent and define four histopathological patterns: flat, lobular, in situ, and invasive, which occur in pure and combined forms. The ID-PDXs replicate earlier stages of tumor development than their clinical counterparts. Micrometastases are already detected when lesions appear in situ. Tumor extent, histopathological patterns, and metastatic load correlate with biological properties of their tumors of origin. Our findings add evidence to the validity of the intraductal model for in vivo studies of ER+ breast cancer and raise the intriguing possibility that tumor cell dissemination may occur earlier than currently thought. Conflict of interest statement: The authors declare no conflict of interest.


Introduction
Breast cancer (BC) is a frequent disease worldwide [1]. Over 75% of BCs express the estrogen receptor a (ER) in >1% of the tumor cells by immunohistochemistry (IHC) [2]. ER+ BCs largely overlap with the luminal subtypes defined by global gene expression [3,4]. Luminal BC is heterogeneous and further subdivided into luminal A and B BC, with low versus high proliferative indices and distant recurrence rates [5]. Twenty percent of patients experience distant recurrence and cancer-related death [6]. Overtreatment of the early stages of the disease and endocrine resistance are additional problems in this subgroup [7]. A lack of preclinical models hampered progress in understanding the biology of these tumors and the development of new therapeutic approaches. Genetically engineered mouse models mostly develop hormone receptor negative tumors, and few ER+ BC cell lines grow in vivo, and only if supplemented with nonphysiological amounts of estrogen [8]. PDXs are increasingly used for preclinical studies but difficult to establish from ER+ tumors [8]. We and others showed that the microenvironment is a major determinant of luminal BC cells and that take rates increase dramatically when hormone sensitive BC cells are grafted to mouse milk duct [9]. They grow in the absence of estrogen supplementation, recapitulate biological and molecular features of their clinical counterpart and respond similarly to therapy [9,10]. Yet, mammary stroma and endocrine milieu differ beween women and mice. To assess the impact of the mouse host on the biology of the engrafted human BC cells of this increasingly used model, we analyzed 21 ID-PDXs histopathologically.

Materials and Methods
The study was approved by the Commission cantonale d'éthique de la recherche sur l'être humain (CER-VD 38/15), patients signed an informed consent and all animal experiments were performed in accordance with protocols approved by the Service de la Consommation et des Affaires Vétérinaires of Canton de Vaud, Switzerland. Within 30 minutes of surgical excision, after inking of margins and macroscopic assessment, a fraction of tumor tissue was taken by the pathologist (MF) and transported to the laboratory in DMEM/F12. Information on patient age, menopausal status, primary tumor (PT) size, nodal status, histological type according to WHO 2012 classification [11], ER, progesterone receptor (PgR), Ki67, HER2 status, was collected from clinical charts and anonymized (Table S1).
Tumor pieces were mechanically and enzymatically dissociated to single cells, lentivirally transduced with luciferase-GFP and injected into the teats of 10-week-old immunocompromised female mice (NSG) [9]. In vivo growth was monitored every two weeks by measuring bioluminescence. Engrafted glands were dissected, fixed in buffered formalin for 2 hours and paraffin-embedded. Fifteen 4 µm sections were cut and nr 1, 7, and 15 stained with Hemalun Eosin (H&E). ALU staining was used to unequivocally identify human cells when required. IHC was performed under standard clinical conditions with Discovery Ventana ULTRA [9]. Metastatic load was calculated as percentage of bioluminescence positive organs of organs collected post mortem.

Results
Tumor cells from 21 patients were ID grafted to 88 mice in 220 glands after lentiviral transduction with luciferase-GFP ( Figure 1A, B, Table S1). Mice were sacrificed when radiance >10E8. Presence of micrometastases was determined by ex vivo radiance measurements on distant organs revealing bones as the most frequent site of tumor cell seeding followed by lungs, brain and liver ( Figure 1C (Table S1). Overall, the distribution of ER+ and PgR+ cells in PDXs and PTs were similar ( Figure S2A, B) but in 5 and 16 pairs, respectively, positivity was discrepant ( Figure S2C, D).
HER2 status was IHC 3+ in 2 of 3 PDXs corresponding IHC 3+ PTs including one confirmed by FISH.
A PT with a IHC 2+ score and focal gene amplification was negative in the PDX suggesting clonal outgrowth [12].  [11,12] was characterized by a monolayer lining of variably dilated mouse ducts ( Figure 2A) with large columnar tumor cells showing mild nuclear pleomorphism and abundant eosinophilic cytoplasm forming apical "snouts" ( Figure 2B). It was associated with foci of intraductal proliferation of low nuclear grade tumor cells in a cribriform pattern similar to atypical ductal hyperplasia [11] ( Figure 2E).
Strong and diffuse ER ( Figure 2D) and PgR expression ( Figure 2E) were observed in the F pattern, interpreted as an early step of "colonization" of the murine ducts by human tumor cells.
The LOB pattern was characterized by tumor cell growth within the ductal wall, like pagetoid spread of lobular carcinoma in situ (LCIS) ( Figure 2F, G) [11,13]. It was associated with intracellular clear, mucin-like vacuoles bestowing a signet-ring cell-like appearance on the tumor cells ( Figure 2F). Although the tumor cells were derived from invasive primary tumors, only in 1/3 of the PDXs was of INV pattern with tumor cells detected outside the host ducts, either isolated or in small clusters ( Figure   2N-P). Thus, intraductally xenografted tumor cells replicate earlier stages than the original PT.
Notwithstanding the early developmental stages, micrometastases were present in 87.5% of the mice and in all the 16 analyzed PDXs (Table S1) Figure 3A). In luminal B-like cases, all NST, the F pattern was observed in 9/14 and only 2 were pure ( Figure 3A). The INV pattern was detected in 6/14, IS and INV patterns were combined in 4/14. Thus, there was a trend for the F pattern to be present in PDXs obtained from luminal A-like BC.
The IS and INV patterns on the contrary were observed, with or without associated F pattern, in PDXs obtained from the more aggressive luminal B-like subtype. Metastatic burden was significantly higher in mice engrafted with luminal B-like than luminal A-like PTs ( Figure 3B).
Tumor extent did not correlate with the time the grafted cells spent in the hosts ( Figure 3C) nor with the number of cells injected ( Figure 3D) but with in vivo growth rates ( Figure 3E) and Ki67 index in PT ( Figure 3F). The proliferative indices in PTs also correlated with in vivo growth rates ( Figure 3G) and metastatic load ( Figure 3H). Thus, tumor extent relates to PT biology rather than to engrafment modalities; growth rates and metastatic burden reflect Ki67 index and hence patient prognosis.

Discussion
ID-PDXs of ER+ BCs exhibit a phenotypic diversity, which reflects the entire spectrum of human disease progression from precursor lesions to in situ disease and invasive cancers [11,13,17]. Our findings indicate that critical histopathologic features are intrinsic to tumor cells and not determined by the stroma or systemic factors, which differ between human and mice. The observation that two-thirds of the ID-PDXs derived from invasive cancers represent earlier lesions, like in situ carcinoma or flat epithelial atypia suggest that engrafted tumor cells re-start their growth intraductally, recapitulating different steps with, for some of them, a prolonged stage of non-invasive growth as they likely did in the patient's breast years before diagnosis. This phenomenon is important to consider when using the model in personalized medicine [9].
While the typical F, LOB, IS patterns are readily identified, some morphological aspects of ER+ ID-PDX are more difficult to classify. This concerns dispersed foci of human cells lining the wall of small mouse ducts, with neither apical snouts nor cystic duct dilation. We classified these as F but speculate that they will develop into a different pattern.