Background
Breast cancer (BC) is the most frequently occurring type of malignancy in women and also the leading cause of cancer related deaths among women in high-income countries [
1]. BC remains a serious issue in current healthcare, although diagnostic and therapeutic strategies have improved over the past decades. In addition to first line chemotherapeutic treatments, targeted therapies for cancers overexpressing the estrogen, progesterone and Her2 receptor already led to improved patient survival over the past decade [
2]. Nevertheless, a subgroup of BC patients either does not respond to first line chemotherapy, or develops resistance. Furthermore, no tailored therapy is currently available for tumors without expression of specific receptors. Therefore, improved personalization strategies for BC treatment are urgently needed.
The most important goal of personalized medicine is to dedicate the most appropriate treatment to the individual patient. This could lead to a situation where the percentage of non-responders and the high proportion of adverse effects of classical chemotherapeutics could be minimized. The success of this approach depends on extensive characterization of individual tumors and their sensitivity to chemotherapy. The majority of preclinical research for treatment efficacy in BC has been performed using established BC cell lines and mouse models. These models are usually not generated from primary BC and resemble only a subset of the diverse types of tumors observed in primary BC. In addition, BC cell lines have been in culture now for decades and have acquired several changes that could affect their biological behavior and therefore they do not faithfully reflect the tumor of origin [
3]. Hence, a short-term primary culture derived directly from the tumor is necessary for better characterization and generation of chemotherapy sensitivity profiles in breast tumors from patients.
Various strategies have been applied to generate primary cultures from individual tumors which include: a) 2D culture of dissociated tumor cells, b) 3D spheroid cultures, c) patient derived mouse xenograft (PDX) cultures and d) organotypic tumor slice cultures [
4‐
7]. These strategies all have their own advantages and limitations depending on the specific research purpose.
Tumor dissociation and 2D cultures are suitable for certain tumor types only, because not all primary tumors grow in a monolayer ex vivo and dissociation strategies are very challenging [
8,
9]. Moreover, in 2D, tumor architecture is completely lost and this method of culturing causes high selection of tumor cells that grow out [
10]. Especially in very heterogeneous cancers, such as BC, tumor cell selection limits the usefulness of this culture option for optimal drug response testing.
Primary tumor cells can also be cultivated ex vivo in 3D in a gelatinous protein mixture, mimicking the extracellular matrix. The generation of 3D tumor spheroid cultures can be applied to more tumor types and tumor cells can be expanded in great numbers for high throughput drug testing. The major disadvantage of this approach is that this expansion of tumor cells takes months of culturing and is therefore not optimal for diagnostic purposes [
5].
Another method to reliably assess responses to cytotoxic treatments is the generation of PDX models,. [
11,
12] which are generated by implanting pieces of fresh human tumor tissue subcutaneously in immune-deficient mice [
7]. However, the successful engraftment rate of breast tumors is less than 25 % and outgrowth of the engraftment takes months. Therefore, this method is not optimal for studying cytotoxic drug responses for personalized cancer treatment.
The organotypic slice method turns out to be ideal for the purpose of short-term primary cultures. It can be applied to most solid tumors and the tissue processing is relatively fast compared to other methods which demand much longer waiting times for ex vivo tumor proliferation [
6,
13,
14]. It does not involve selective outgrowth of tumor cells and short-term assays that could predict clinical drug responses can be readily performed making this method in principle ideal for studies on personalized BC treatment [
14‐
16].
Nevertheless, ex vivo assays for personalized treatment based on the tumor tissue slice model are delicate because it is a low throughput assay and methodological developments are challenging [
16]. Moreover, tumor heterogeneity requires advanced analytical tools to faithfully categorize tumor responses to drug treatment, especially in the case of BC.
We improved on previously reported organotypic tumor tissue slice methods and optimized it for the ex vivo culture of primary BC. Using reliable markers of cell proliferation and cell death we developed a robust analytical system for breast tumor slices ex vivo. Using these methods we characterized cytotoxicity responses of individual breast tumor slices to chemotherapy.
Discussion
We optimized the ex vivo culture system to preserve tumor morphology and tumor cell proliferation, while minimizing culture induced tumor cell death for up to seven days. We did not characterize later time points, but there are no signs of tissue deterioration after six days, suggesting that extended incubations may be possible if required for a specific functional assay. We specifically found that the culture of breast tumor slices is highly dependent on rotational movement during incubation, most probably by increasing nutrient exchange. Automated tissue slicing also increases ex vivo lifespan of tumor slices, as thinner slices allow better penetration of nutrients in the absence of active blood circulation. Finally, composition of the culture medium is highly important to maintain tumor slice viability with sufficient numbers of replicating cells for several days.
As traditional chemotherapies especially target cells in S phase or mitosis, tumor cell proliferation should be maintained during ex vivo culture of tissue slices to investigate chemotherapy responses. EdU incorporation is probably the most reliable biomarker for proliferation because it allows real time measurement of DNA synthesis in very short time intervals compared to other proliferation markers such as Ki-67, which remains positive for days after proliferation has ceased [
23]. Cyclin A is a slightly better marker than Ki-67, but it also stains G2 phase cells and would therefore also mark cells that are blocked at the G2/M checkpoint. In our ex vivo tumor slice culture system active EdU incorporation is preserved for up to seven days, which is sufficient to detect most differences in drug response.
As decrease of EdU positive cells may be reversible when chemotherapy treatment has ended, more unidirectional response markers such as induction of cell death are important to define cytotoxic response to the given treatment. TUNEL staining is often used as a way of measuring cell death, although quantitative assessment of TUNEL staining remains a challenge. Measuring the number of individual TUNEL positive cells is hardly possible, because of the proximity of individual nuclei, which frequently even overlap in thin sections of organotypic BC slices. As an alternative option we determined the number of DAPI pixels that are TUNEL positive. The major advantage of this analytical method is that it can be performed automatically in a standardized high-throughput manner. This method may not be useful to distinguish subtle increases but it is suitable to detect major differences in TUNEL signal.
Similar culture methods have been used for ex vivo breast tumor slices in previous studies [
15,
16,
20]. These report that viability and proliferation was retained for at least four to seven days. However, extensive validation of culture conditions for multiple tumors where not performed. Also, comparative analysis of different culture conditions is lacking from these reports. The tumor slice studies performed on head and neck, colon and lung tumors also lack some of the characterizations that have been performed in this study [
6,
13]. Furthermore, the markers used as surrogates for treatment response, such as Ki-67, may not be the most adequate analysis for treatment response. Therefore, we propose to use EdU incorporation as the most direct and sensitive marker of proliferation in tumor tissue slices.
The responses to FAC treatment in our study were quantified based on morphologic examination, EdU incorporation and TUNEL assay. Interestingly, sensitivity classification based on the individual analytical methods did not differ significantly: the two least sensitive tumors always cluster together (Fig.
5). As morphologic examination and TUNEL assay both indicate tumor cell death, one of these methods would probably be sufficient, but more extensive characterization will be necessary to corroborate this conclusion.
Therapy resistance was arbitrarily defined in our cohort based on studies performed in BC patients receiving neo-adjuvant chemotherapy, which report that 10-20 % of primary breast tumors are resistant to treatment [
24‐
26]. Interestingly, this arbitrarily defined threshold is consistent with the result obtained from one clinically proven therapy resistant tumor suggesting that our thresholds for therapy resistance may resemble the clinical outcome.
Actual clinical data regarding the response to FAC treatment are not yet available for our collected samples and it is therefore not yet possible to determine the predictive value of our ex vivo analysis method. Clinical responses to treatment will be monitored closely in the future in order to assess the predictive value of the ex vivo assay. However, this assessment may take years. Clinical data on the response to neo-adjuvant FEC treatment was only available for the patient having the least sensitive tumor in our analysis.
Ideally, the ex vivo analysis should be done with biopsies obtained prior to neo-adjuvant treatment, because this would enable determination of in vivo response of the tumor in a relatively short time frame. On the other hand, tumor heterogeneity may pose additional challenges to develop tumor biopsy based cytotoxicity assays into valid predictive tests. Moreover, the currently described methodology concerning this assay might not allow direct implementation of this assay in a clinical setting. Fresh tumor material is needed, timings are important and procedures are quite laborious. Therefore, automatization and high-throughput possibilities should be explored for this assay.
As this ex vivo assay does not take pharmacokinetics into account, it can only predict intrinsic sensitivity of tumor cells to a given treatment. In other words, an ex vivo resistant tumor is likely to also be resistant in vivo, but ex vivo sensitivity may not faithfully predict clinical sensitivity. Therefore, this cytotoxicity assay is primarily expected to be an effective tool to prevent unnecessary treatment of patients that harbor a therapy-resistant tumor, especially in advanced metastatic BC, where resistance to FAC chemotherapy is much more frequently observed than in primary BC. Similar assays can be developed to investigate other drugs, e.g. tamoxifen sensitivity.
Competing interests
The authors declare that they have no competing interests.
Authors’ contribution
KN: Conceived of the study, generated the first draft, performed tissue slicing, tissue culturing and immunostaining experiments, performed imaging and image analysis. NS: Conceived of the study, generated the first draft, performed tissue slicing, tissue culturing and immunostaining experiments. HS: Performed imaging and image analysis. CvD: Coordinated protocols for the use of patient specimens and collected tumor specimens for research. MdB: Coordinated protocols for the use of patient specimens and collected tumor specimens for research. JH: Acquired funding and supervised the research group, conceived of the study, generated the first draft. RK: Acquired funding and supervised the research group, conceived of the study, generated the first draft. MV: Acquired funding and supervised the research group, conceived of the study, generated the first draft. AJ: Acquired funding and supervised the research group, conceived of the study, generated the first draft, coordinated protocols for the use of patient specimens and collected tumor specimens for research. DvG: Acquired funding and supervised the research group, conceived of the study, generated the first draft. All authors were involved in design of experiments, analysis and interpretation of data and in revising the manuscript. All authors read and approved of the final version of the manuscript.