Therapeutic response to CDK4/6 inhibition in breast cancer defined by ex vivo analyses of human tumors

Abstract

To model the heterogeneity of breast cancer as observed in the clinic, we employed an ex vivo model of breast tumor tissue. This methodology maintained the histological integrity of the tumor tissue in unselected breast cancers, and importantly, the explants retained key molecular markers that are currently used to guide breast cancer treatment (e.g., ER and Her2 status). The primary tumors displayed the expected wide range of positivity for the proliferation marker Ki67, and a strong positive correlation between the Ki67 indices of the primary and corresponding explanted tumor tissues was observed. Collectively, these findings indicate that multiple facets of tumor pathophysiology are recapitulated in this ex vivo model. To interrogate the potential of this preclinical model to inform determinants of therapeutic response, we investigated the cytostatic response to the CDK4/6 inhibitor, PD-0332991. This inhibitor was highly effective at suppressing proliferation in approximately 85% of cases, irrespective of ER or HER2 status. However, 15% of cases were completely resistant to PD-0332991. Marker analyses in both the primary tumor tissue and the corresponding explant revealed that cases resistant to CDK4/6 inhibition lacked the RB-tumor suppressor. These studies provide important insights into the spectrum of breast tumors that could be treated with CDK4/6 inhibitors, and defines functional determinants of response analogous to those identified through neoadjuvant studies.

Introduction

Breast cancer is a highly heterogeneous disease.^1-4 Such heterogeneity is known to influence patient response to both standard of care and experimental therapeutics. In regards to biomarker-driven treatment of breast cancers, it was initially recognized that the presence of the estrogen receptor α (ER) in a fraction of breast cancer cells was associated with the response to tamoxifen and similar anti-estrogenic therapies.^5,6 Since this discovery, subsequent marker analyses and gene expression profiling studies have further divided breast cancer into a series of distinct subtypes that harbor differing and often divergent therapeutic sensitivities.^1-3 While clearly important in considering the use of several current standard of care therapies, these markers, or molecular sub-types, do not necessarily predict the response to new therapeutic approaches that are currently undergoing clinical development. Thus, there is the continued need for functional analyses of drug response and the definition of new markers that can be used to direct treatment strategies.

Currently, all preclinical cancer models are associated with specific limitations. It is well known that cell culture models lack the tumor microenvironment known to have a significant impact on tumor biology and therapeutic response.^7-9 Xenograft models are dependent on the host response for the engraftment of tumor cells in non-native tissues, which do not necessarily recapitulate the nuances of complex tumor milieu.¹⁰ In addition, genetically engineered mouse models, while enabling the tumor to develop in the context of the host, can develop tumors that do not mirror aspects of human disease.¹⁰ Furthermore, it remains unclear whether any preclinical model truly represents the panoply of breast cancer subtypes that are observed in the clinic. Herein, we utilized a primary human tumor explant culture approach to interrogate drug response, as well as specific determinants of therapeutic response, in an unselected series of breast cancer cases.

Results

To determine the feasibility of using ex vivo primary human tissue for the analyses of therapeutic response, we focused on breast cancer. For these studies, primary tumor tissue that was not required for pathological diagnoses was employed. The tissues were macrodissected to ~1 mm³ and placed into culture on a semi-solid support that enabled uptake of media components by the tissue (Fig. 1A). Tissues were cultured for 72 h to monitor the expression of markers and/or drug response. Following the establishment of the methodology, 20 unselected breast cancer specimens were employed for explant analyses. Of these, 13 retained tissue architecture and degrees of cellularity comparable to that observed in the tumor section, as determined by histological analyses. Representative staining of the surgical specimens and corresponding explants are shown (Fig. 1B). The cultured explants retained similar expression of the key molecular markers, ER and HER2, which are used to base treatment decisions in breast cancer, compared with the original tumor specimen. In the analyses of the 13 primary surgical specimens that retained tissue architecture, nine cases were ER-positive (> 1% ER-positive tumor cells), four were Her2 positive (3+ Her2 staining), and two specimens were negative for both ER and Her 2. The explanted tissues retained these markers in 100% of cases over the culture window, although there was some reduction in the sensitivity for Her2 staining (1+ to 0 in two cases) (Fig. 1C). These findings demonstrate that this approach maintained the overall biomarker phenotype of the tumor.

Figure 1. Explants retain tissue architecture and clinical biomarkers. (A) Schematic of the explant procedure utilized in this study. (B) Representative hematoxylin/eosin, Her2 and ER staining between explant and surgical specimens. (C) Histological scoring of staining of primary tumor tissue and explants with Her2 and ER show concordance of staining.

Breast tumors exhibit widely divergent proliferative indices that are related to tumor subtype and overall pathophysiology. To determine if the explanted cultures retained the heterogeneous proliferative nature of the primary tumor, the marker Ki67 was evaluated (Fig. 2A). Among primary tumors, a range (18–80%) of Ki67-positive tumor cells was observed (Fig. 2B). This observation is in accordance with the known range of Ki67 staining in clinical specimens. There was an exceedingly strong positive correlation in Ki67 index between the original tumor and the corresponding explanted tissue in all cases (R = 0.9385, p < 0.0001 for concordance). These data indicated that in the context of such ex vivo culture, the proliferative potential remained similar to that of the primary tumor.

Figure 2. Proliferative index in explants is consistent with that observed in the surgical specimen. (A) Representative Ki67 staining from surgical specimen and associated explants. (B) Quantification of Ki67 staining from matched explant and surgical specimens demonstrating concordance between the two. Data was analyzed by paired t-test. Correlation coefficient 0.9385 and p < 0.0001 for concordance in the measures.

The preceding findings suggested that treatment of explanted tissue with therapeutic agents could reveal sensitivities that would be inherent in the primary tumor. To specifically interrogate this concept, explants were treated with the CDK4/6 inhibitor PD-0332991, a potent cytostatic agent that is in phase II clinical trials for multiple malignancies, including breast cancer.^11-13 Treatment with PD-0332991 revealed highly profound suppression of Ki67 staining. This effect was specific to the drug and was not observed in tissues treated with DMSO (vehicle) for the same time period (Fig. 3A, left panels). The dramatic suppression of Ki67 positivity was also not routinely observed following treatment with either doxorubicin or PARP inhibitor (ABT-888) in parallel with these studies (data not shown). Interestingly, only a subset of explanted tumors responded effectively to PD-0332991 (Fig. 3A, left panels). Specifically, ~85% (11/13) of cases exhibited a greater than 5-fold suppression of Ki67 upon exposure to PD-0332991 (Fig. 3). In contrast, approximately 15% (2/13) of cases failed to respond to PD-0332991 treatment, as indicated by minimal or no reduction in Ki67 staining after exposure to the CDK4/6 inhibitor (Fig. 3B). Of note, the response to PD-0332991 did not depend on ER or HER2 status. These analyses suggest that there is variability in the response to PD-0332991 in the clinical population of breast cancer patients. The failure to respond to PD-0332991 was specific to the tumor cells, as normal mammary epithelial cells present within the explants effectively responded, as evidenced by complete suppression of Ki67 (Fig. 3C). Additionally, from the same tumor resection, we could demonstrate that lymph-node metastasis were also resistant to CDK4/6 inhibition (Fig. 3C). Thus, the failure to respond to PD-0332991 is a consequence of tumor-specific genetic events and not due to distinctions in intrinsic drug sensitivity of an individual.

Figure 3. Differential response to CDK4/6 inhibition in breast cancer explants. (A) Representative Ki67 of explants treated with DMSO (vehicle) or PD-0332991. (B) Quantification of Ki67 from PD-0332991 treated cultures. (C) Example of normal tissue, matched tumor and lymph-node metastasis stained for Ki67.

Preclinical data suggest that RB is required for the cytostatic response to PD-0332991.^11,12,14 RB is the proto-typical tumor suppressor that functions downstream of CDK4/6 to modulate cell cycle progression.^15-17 Tumors that are RB-deficient express exceedingly high levels of p16ink4a.¹⁷ Therefore, to determine the basis for the differential response to PD-033299 in the explanted tissues, both the primary tumors and explants were stained for p16ink4a and RB. These data revealed that within the cohort of cases evaluated, 23% (3/13) exhibited high levels of p16ink4a, and 15% (2/13) exhibited RB loss (Fig. 4A and B). These markers were conserved between the primary tumors and the explants. Importantly, of the three tumors with high (3+) immunostaining for p16ink4a, two failed to respond to PD-0332991, while both of the tumors that exhibited loss of RB failed to respond to PD-0332991 (Fig. 4B, red box). These studies indicate that the analyses of p16ink4a and RB status can be used prospectively to evaluate the response to CDK4/6 inhibitors in clinical specimens and suggest an innovative approach for developing markers for targeted therapies in advance of clinical trials.

Figure 4. p16ink4a and RB status are associated with the response to PD-0332991 in tumor specimens. (A) Representative staining of p16ink4a and RB in surgical specimens and explants. (B) Association of p16ink4a and RB status in response to PD-0332991 among explanted tissue cultures. Cases that did not respond to PD-0332991 are boxed in red.

Discussion

Defining the basis of therapeutic response in cancer is routinely hampered by the limitations of existing preclinical models. Here, we utilized primary tumor tissue, cultured ex vivo, to directly interrogate drug response and determinants of that response. These analyses provided data similar to that obtained through serial biopsying in a clinical setting and could represent a key innovation to develop markers of therapeutic response in patient populations as observed in the clinic.

While breast cancer is extensively studied, and there is a wealth of models, it is clear that not all models accurately recapitulate the complexity or heterogeneity of human disease.¹⁰ Particularly, most cell culture and xenograft models have been developed from metastatic disease and, by definition, are selected for their ability to grow in culture or in animal hosts. As a result, such models reflect only a small spectrum of selected disease. By using primary tumors in ex vivo culture, relatively unbiased analyses of the type of disease that is treated in the clinic can be evaluated. In keeping with other similar approaches,¹⁸ the cultured tissue explants retained the associated stromal components that are known to play important roles in the pathogenesis and therapeutic sensitivities of cancer.

Defining the spectrum of cases that will respond to a given therapy in the clinic is essential for directing therapy to patients efficaciously. The analyses performed here indicate that approximately 85% percent of cases will respond to CDK4/6 inhibition. While these results combined with previously published pre-clinical studies demonstrate a clear acute cytostatic response, prior studies have also described the ability of extended CDK4/6 inhibition to drive increased rates of cellular senescence, or permanent cell cycle exit, while not increasing rates of apoptosis.¹³ Such results demonstrate a sustained/durable cytostatic response, which has also been observed in clinical trials examining the efficacy of PD-0332991. Specifically, one such study examining the effectiveness of CDK4/6 inhibition among a population of teratoma patients demonstrated continued efficacy (delayed disease progression) over a period of two years.¹⁹ Clinically, the endpoint of delayed disease progression implies a state of tumor growth stasis, which can be easily quantified through the use of Ki67 immunohistochemical staining, a widely used clinical marker of cell proliferation. Due to the well-characterized cytostatic mechanism of PD-0332991, Ki67 is the most relevant response that would be employed clinically.

Importantly, such preclinical studies have also demonstrated that RB loss is associated with the bypass of CDK4/6 mediated therapies. However, whether this is really a determinant of therapeutic response in patients has never been documented. Here, we find that tumors that were RB-negative and p16ink4a-high at surgical resection were unresponsive to PD-0332991. Interestingly, while high levels of p16ink4a are in general a strong surrogate marker for RB loss,¹⁷ the association is not 100%. We have confirmed this finding in the analyses of DCIS and breast cancer cases (in preparation). Regardless, the current data confirm the validity of the preclinical data and strongly indicate that RB status is the primary marker of drug response, followed by intense (3+) immunostaining for p16ink4a. Therefore, both markers should be employed in the clinic to direct treatment with CDK4/6 inhibitors for breast cancer.

Collectively, the findings within this paper indicate that using ex vivo tumor explant culture approaches could provide insights similar to neoadjuvant clinical trials. Additionally, in principle, ex vivo analyses of diagnostic biopsies could be used to immediately determine the sensitivity of the primary tumor to the therapeutic agent.

Methods and Materials

Tissue source and ethics statement

Excised breast tumor tissues were provided by a pathologist in accordance with Thomas Jefferson University Institutional Review Board (IRB) approval (Control #09D.54) and under written, informed consent. All cases represented invasive ductal carcinoma, and none of the patients received neoadjuvant chemotherapy. Post-excision and pathological examination tissues were placed into 10 ml of complete culture medium for transport to the laboratory. Tissues were maintained in this medium at 4°C until processing. All source tissues were plated as explants within 12 h of receipt from pathological examination.

Breast tumor explant culture medium and conditions

Culture media consisted of RPMI 1640 base (Sigma) supplemented with 10% Fetal Calf Serum (FCS). Antibiotic/antimycotic solution was added to a final concentration of 1X (Sigma). Hydrocortisone (Sigma) was resuspended in 95% EtOH and supplemented at a concentration of 1 mg/100 ml media. Lastly, recombinant human insulin was added at a final concentration of 1 mg/100 ml media (Gibco). Upon plating, all tissues were cultured at 37°C and 5% CO₂ until harvest.

Dissection and plating

Under sterile conditions and 1 hour prior to tissue dissection, 1 cm³ hemostatic gelatin dental sponges (Vetspon, Novartis) were hydrated in complete explant media at 37°C. While sponges were hydrating, tissues were transferred to a 10 cm cell culture dish with 10 ml complete explant media for dissection. Using a sterile scalpel blade and microdissection scissors, any apparent fat tissue was removed from the surface of the specimen and discarded. Samples were then cut into 1 mm³ sections. Upon complete hydration of gelatin sponges, sponges were transferred to individual wells of a 12-well cell culture dish, and 1.5 ml of complete explant media was added to each well containing one sponge. Each treatment condition was examined in duplicate, requiring two sponges and two culture dish wells for each condition in total. Next, three 1 mm³ dissected specimen sections were chosen at random and placed on top of each sponge using sterile forceps. Tissues were placed near the edges of the sponge to allow for media perfusion in between tissue sections. Explant cultures were incubated at 37°C and 5% CO₂.

Treatments

All explants were maintained in complete explant media for 24 h. At 24 h, media was aspirated at low velocity to prevent disturbance of plated tissues, and 1.5 ml of media containing 500 nM PD0332991 or equal volume vehicle (DMSO) was then added to the corresponding treatment wells. Culture dishes were then returned to 37°C and cultured for 48 h. After 48 h, tissues were gently removed from sponges and placed into embedding cassettes and fixed for 24 h in 10% neutral buffered formalin, followed by storage in 70% EtOH until embedding in paraffin for sectioning and immunohistochemical analysis.

Immunohistochemistry

Paraffin-embedded tissues were cut to 5 µM sections for histological and immunohistochemical analysis. For explants, staining for ER (clone SP1), Ki67 (clone 30–9) and Her2 (clone 4B5) were performed using established clinical protocols and controls for immunohistochemical staining. For clinical specimens, ER, HER2 and Ki67 IHC were performed using the same antibodies on the BenchMark XT Slide Preparation System (Ventana Medical Systems). Tissues were scored using ASCO/CAP guidelines.²⁰ p16ink4a and RB were stained and scored as previously described.⁷

Image analyses and scoring

Immunohistochemically stained primary tumor and explant slides were scanned using an Aperio ScanScope®CS instrument at 20×, (Aperio Technologies). Stains were quantified using FDA approved analysis tools developed by Aperio. Nuclear algorithms were used to evaluate ER, PR and Ki67 stain, and a membranous algorithm was applied for Her2 quantification. The mean staining intensity and percentage of staining cells were recorded for ER, PR and Ki67. For Her2, the score (range 0–3) was reported.

Statistical analysis

All statistical analyses were performed using graphpad Prism and the Student t-test for significance. Results with p < 0.05 were considered significant.

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.