Introduction

Endometrial cancer (EC) is the most common gynaecological malignancy in the developed world but remains significantly under-researched [1–3]. The incidence of EC is rising and, although traditionally viewed as a post-menopausal disease, this rise is particularly evident in the pre-menopausal population. The reason for this is multifactorial and risk factors include parity later in life, lower parity, increased lifespan, polycystic ovarian syndrome, lynch syndrome, and perhaps most notably, obesity [4–7]. Of the four histological subtypes, endometrioid is the most common EC, occurring in 75–80% of cases [8], and is the subtype associated with high body mass index (BMI) [6]. Total hysterectomy is the recommended treatment for all stages of EC including pre-malignant endometrial hyperplasia, with later stages often also receiving pelvic lymph node dissection and bilateral salpingo-oophorectomy. The trend towards being diagnosed at a younger age and the increased surgical risk for those with a high BMI necessitates research into fertility-preserving, conservative management options for endometrioid EC (EEC).

Around 90% of novel therapeutics fail in randomised control trials [9]. Up to 50% of these failures are due to lack of clinical efficacy, which can be interpreted as a lack of pre-clinical models that accurately reflect in vivo response [9,10]. Given its ability to mimic the 3D multicellular architecture of human disease, the ex vivo patient-derived explant (PDE) culture technique has recently emerged as a powerful preclinical platform for cancer research [11]. PDEs are typically small (1–3 mm in diameter) pieces of fresh tumour that are cultured without disintegration into the tissue’s individual cellular components. Although methods of PDE culture vary, these models appear to have strong clinical applicability [12], and have shown promise in correlating with patient outcomes and predicting drug resistance in non-small cell lung carcinoma, gastric and colon cancer[13]. Advantages of PDE models over other forms of 3D models such as organoids is that the PDE model retains the native 3D tissue architecture and can therefore account for spatial and microenvironmental effects such as the effect of stromal and tumour-associated immune cells [14]. An ex vivo PDE model for EEC could prove beneficial for fast tracking novel, conservative (surgery-sparing) therapeutics from conception to clinical trial.

Radiotherapy and the levonorgestrel-(LNG) releasing intra-uterine system (IUS) are the only non-surgical treatments available for EEC patients. LNG-IUS, also known as Mirena®, has traditionally been used as a long acting, reversible contraceptive. As an EC therapy, LNG-IUS is only feasible for early-stage cases of endometrioid histology and pre-malignant endometrial hyperplasia, both of which currently carry the primary treatment modality of total hysterectomy. LNG is a synthetic progestogen which mimics the naturally occurring hormone, progesterone. In the context of cancer therapeutics, LNG counterbalances estrogen-driven cell proliferation, which is often the molecular mechanism underlying EEC. However, response rates for treatment of early-stage EEC with LNG-IUS in the literature vary between 40%-80% [15,16]. Despite the variability in reported response rates, there are currently no approved biomarkers predicting LNG-IUS response. As a result, patients fitted with the LNG-IUS undergo biopsies every 3–6 months to assess treatment efficacy in real time [17]. The invasive nature of these biopsies creates an urgency for research into novel, clinically useful biomarkers to predict response to LNG-IUS therapy [18,19]. The ability to predict LNG-IUS resistance prior to administration would ensure this treatment option is only administered to patients who would benefit from it. We propose that the development of a clinically and biologically relevant model of EEC would provide a robust platform for assessing both biomarkers of LNG-IUS response and investigation of conservative treatment for EEC.

Materials and methods

Information regarding the materials and antibodies used can be found in S1 and S2 Tables, respectively.

Cohort

All patients undergoing hysterectomy after a diagnosis of early-stage EEC at Wellington Regional Hospital between 6^th July 2021 and 21^st June 2022 were invited to participate in the study and provided informed, written consent. Ethical approval was obtained from the Human Disability Ethics Committee (HDEC) (15/CEN/143) and the University of Otago Ethics Committee (H20_002). Consultation was undertaken with Research Advisory Group–Māori (RAG-M). All research was performed in accordance with the relevant guidelines/regulations and the Declaration of Helsinki. Authors had access to identifiable information from participants during the data collection in order to obtain final pathology reports. Following excision during surgery, the whole uterus was opened along the sagittal plane and the tumour was inspected by a pathologist and photographed by a medical photographer (L.G) (Fig 1A). A section of fresh tumour tissue, excess to diagnosis, was excised from the uterus into cold phosphate buffered saline (PBS).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Ex vivo cell culture set up.

(A) Example of a uterus used in the study, photographed from i) anterior, ii) posterior, and iii) open anterior view. Scale bars represent 5 cm. (B) Cell culture methodology. i) Explant tissue size. ii) Example of plate containing 2x explants per well. Black arrows point to individual explants sitting on top of a gelatin sponge. iii) Schematic of cell culture technique.

https://doi.org/10.1371/journal.pone.0301413.g001

Results

Development of a long-term, ex vivo culture model of endometrial cancer

Ten patients with EEC were recruited into this study. Their demographics and clinicopathological characteristics are listed in Table 1. Of the patients included in the study, two (Patients 1 and 3) had been previously treated with LNG-IUS for early-stage EC. Tumour was obtained from the uterus of patients undergoing hysterectomy (Fig 1A). Resected tumour tissue was transported to the laboratory in cold PBS and cultured on a media-soaked gelatin sponge as explants 1–3 mm in diameter for up to 21 days (Fig 1B). H&E staining and PR expression in explants from Day 0 (uncultured), Day 7, Day 14, and Day 21 showed explant architecture and tumour integrity was maintained over time in the ex vivo PDE model (Fig 2). A time course of explants from three other patients can be viewed in the supplementary material (S1–S3 Figs). Similarly, the proliferative capacity of tumour explants was maintained for the length of the culture period, as shown by positive IHC staining for two independent proliferation markers, Ki67 and PHH3 (Fig 2), although expression of these markers is lower in cultured explants compared to uncultured tissue (Fig 2 and S1–S3 Figs). The proportion of apoptotic cells (identified by CC3 and TUNEL) in explants from Patient 1 increased compared to uncultured tissue, however the rate of apoptosis remained stably elevated throughout the culture period (Fig 2). This pattern was also observed in Patient 8 explants (S1 Fig), while in Patient 9 explants, CC3 staining did not appear to change in cultured versus uncultured tissue (S2 Fig). De novo proliferation status was confirmed by positive nuclear BrdU staining at Day 21, indicating continued proliferation after 21 days in culture (Fig 3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Representative images from Patient 1 of H&E and IHC from Day 0 to Day 21 in ex vivo culture.

H&E and IHC for PR expression, proliferation (Ki67, PHH3) and apoptosis (CC3, TUNEL). In the TUNEL row, blue staining reflects DAPI indicating intact DNA and green reflects TUNEL, indicating apoptosis. H&E, haematoxylin and eosin; IHC, immunohistochemistry; PR, Progesterone Receptor; PHH3, Phosphohistone H3; CC3, Cleaved Caspase 3; TUNEL, Terminal deoxynucleotidyl transferase dUTP nick end labelling; DAPI, 4′,6-diamidino-2-phenylindole.

https://doi.org/10.1371/journal.pone.0301413.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. BrdU staining.

Representative images from three patients (P5, P6, P7) on untreated explants after 21 days in ex vivo culture with the addition of BrdU (BrdU was added to culture medium 24 h before fixing explants). H&E, Haematoxylin and Eosin; BrdU, Bromodeoxyuridine / 5-bromo-2’-deoxyuridine.

https://doi.org/10.1371/journal.pone.0301413.g003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Patient characteristics.

https://doi.org/10.1371/journal.pone.0301413.t001

LNG treatment of explants

After 21 days, explants were fixed, embedded and stained with H&E (S4 Fig) and assessed for viability (S5 Fig). High variability was observed in the viability of the explants, even in explants from one tumour sample cultured under the same conditions (Fig 4). Due to this variability, it was difficult to determine which patients were intrinsic responders to LNG treatment using this model. Therefore, the tumours that produced an explant with high tumour viability (>90% viable) after 21 days of treatment with 4 μg/mL LNG were deemed “quasi-resistant”. i.e. the explants treated with LNG for 21 days were examined by H&E stain and the viability of the epithelial compartment was assessed by our histopathologist (D.K.). If the viability was more than 90% then that tumour was considered likely to be resistant to LNG and was therefore quasi-resistant. The viabilities of explants selected for downstream IHC assessment are available in S3 Table. Patients 2 and 4 were excluded due to low viabilities, and Patient 3 was excluded due to a low volume of tumour tissue within the explants. Explants from Patient 1 were used to validate the ex vivo model prior to commencement of LNG treatment and thus were not included in the downstream LNG analysis.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Variability in viability of the tumour tissue within explant replicates after 21 days in ex vivo culture.

Explants from six patients were treated with LNG for 21 days in the ex vivo culture model. A histopathologist assessed the viability of the tumour tissue within each explant based on H&E stains. Patients 2 (P2) and 4 (P4) were excluded from downstream analysis due to low viability of 4 μg/mL LNG explants, and Patient 3 (P3) was excluded due to having a small epithelial tumour compartment. Red circles indicate the explants selected for downstream assessment. Explants were selected based on high viability and large epithelial tumour compartment. Neg, negative control (culture media alone); Vhc, vehicle control (culture media plus 0.08% DMSO); LNG, levonorgestrel (4 ng/mL or 4 μg/mL).

https://doi.org/10.1371/journal.pone.0301413.g004

TUNEL staining was performed on explants from quasi-resistant tumours to assess apoptosis (Fig 5A) and further demonstrated individual variability of response (Fig 5B). Additional representative images of TUNEL positivity staining, as well as a visual guide for TUNEL positivity assessment can be found in S6 Fig. Overall, apoptosis was not significantly increased in the quasi-resistant tumours that were treated with LNG compared to untreated (negative) and vehicle controls. Although apoptosis appeared to increase during ex vivo culture when comparing uncultured (Day 0) controls to explants cultured for 21 days, the majority of the tissue (>55%) was still viable after 21 days (Fig 5B).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Assessment of apoptosis in LNG quasi-resistant explants.

Visualised using TUNEL, in endometrial cancer explants after 21 days in ex vivo culture. (A) Representative images of TUNEL. Explants from Patient 5 which were treated with either vehicle control (culture media plus 0.08% DMSO) or 4 μg/mL LNG for 21 days and stained with TUNEL (green) and DAPI (blue). (B) TUNEL positive cells as a percentage of all cells from explants treated with either culture media (Negative), DMSO (Vehicle), 4 ng/mL LNG or 4 μg/mL LNG. Plot shows mean and standard error of the mean (SEM) bars. A mixed effects analysis showed no significant difference in TUNEL positivity between treatment groups. LNG, Levonorgestrel; H&E, Haematoxylin and Eosin; TUNEL, Terminal deoxynucleotidyl transferase dUTP nick end labelling; DAPI, 4′,6-diamidino-2-phenylindole.

https://doi.org/10.1371/journal.pone.0301413.g005

On Day 21, the most viable explants from quasi-resistant tumours were fixed and stained immunohistochemically for their expression of five candidate proteins previously identified in a 2D model of EEC as biomarkers of resistance to LNG treatment in our laboratory [21,22]. The candidate biomarkers were examined in the current study to demonstrate the feasibility of using this PDE model for biomarker validation. Protein expression was scored by three observers on a scale of 0–3 (Fig 6A) and the median score across the three observations was taken (Fig 6B). Necrotic and stromal regions of tissue were ignored. There was minimal staining of CD80 and SATB2 in the tumour compartment (Fig 6B). KLF4 had the highest degree of staining of the candidate biomarkers examined, and MAOA and ALDH1A1 were moderately expressed. There were no consistent observations made across treatment conditions.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Candidate biomarker immunohistochemical analysis.

Day 0 tumour tissue samples were taken on the day of surgery and tumour explants were fixed after 21 days of culture in culture media alone (Negative), culture media plus 0.08% DMSO (Vehicle), or LNG (4 ng/mL, 4 μg/mL). IHC stains were scored by three independent authors (H.V, C.H, D.K) on a four-point scale of 0–3 based on the proportion of positive tumour cells within the explant and the staining intensity. (A) Representative images of the 0–3 point scale for all markers. (B) Heatmap of the median stain score for each biomarker against treatment for three patients. LNG, Levonorgestrel; IHC, Immunohistochemistry.

https://doi.org/10.1371/journal.pone.0301413.g006

Discussion

In this study, we developed a long-term (21 day), ex vivo, patient-derived explant (PDE) culture model of early-stage EEC and utilised it to test the current non-surgical treatment, LNG-IUS, and the expression of candidate biomarkers of LNG resistance. Patients who are fitted with an LNG-IUS for EEC treatment are generally followed up on after 3–6 months [17]. The model established here can be maintained for 21 days. It therefore has the potential to model response to therapeutics in a way that more closely resembles in vivo biology. Although the long-term culture of early-stage EEC PDEs presented here does not have a 100% success rate, there is a need for long-term cultures which demonstrate an effect of long-term surgery-sparing treatments such as the LNG-IUS. Shorter term cultures, which may have a higher success rate, are unable to provide information on these long term changes post treatment. In this study we demonstrated the preservation of native tumour 3D architecture and proliferative capacity on unmanipulated EEC explants for up to 21 days in culture, although markers for apoptosis tended to be increased in cultured explants compared to uncultured (Day 0) controls. Tumours with explants that survived 21 days of 4 μg/mL LNG treatment were deemed quasi-resistant to LNG. These tumours could not unequivocally be considered resistant because the length of time given for the tumour to respond in vivo is up to 6 months [17] and the length of our culture was three weeks. Quasi-resistant tumours were investigated for their expression of candidate biomarkers of LNG resistance across treatment groups. CD80 and SATB2 were minimally expressed across the explants analysed, while MAOA and ALDH1A1 showed moderate staining, and KLF4 expression was high. This model could therefore be used to validate 2D model-based biomarker discovery studies.

PDE models have been established in a range of cancers including pancreatic, prostate, breast and ovarian [20,23–26]. Recently, a short-term (48 h) PDE model of EC was employed to explore tumour response to therapeutics [27]. In lieu of gelatin sponges, authors utilised Millipore PVDF filter supports onto which tumour explants were placed. Explants maintained native architecture as well as intrinsic cellular parameters after 48 h in culture, demonstrated by H&E and IHC staining for proliferation and death markers. The model was also used to assess explant sensitivity to standard-of-care chemotherapeutics carboplatin and paclitaxel, and results corresponded with the expected proportion of responders to these treatments [27]. Previous evidence showed that ex vivo models can be used to predict patient response to therapeutics [28], and the research by Collins et al demonstrated that EC PDE models also have this potential in the short term (less than 48 h) [27]. We demonstrate that these cultures can be extended to 21 days in some cases. In our PDE model, proliferating cells in the S phase of the cell cycle are present after 21 days, evidenced by BrdU staining (Fig 3). When comparing with Ki67, which enables visualisation of cells in all phases of the cell cycle except resting (G0), there was a slight reduction of Ki67 expression in cultured compared to uncultured tissue, but a large proportion of proliferating cells were maintained to Day 21 (Fig 1). There was increased apoptosis in cultured explants compared to uncultured tissue, however apoptosis did not increase noticeably between day 7 and 21 (Fig 1 and S1–S3 Figs).

We found that there was significant heterogeneity in survival of explants using the PDE model outlined in this study which resulted in fewer patients being included in the final analysis, and which made interpreting results difficult. The variable explant viability was likely due to a combination of sampling and handling techniques, the low grade and stage of the tumours, as well as tumour heterogeneity, which has been noted previously with organoid viability [29]. For example, tumour samples were taken fresh at the time of hysterectomy. Some patients had minimal or non-pedunculated tumours; some tumours were often hard to see on a fresh (versus formalin fixed) specimen. This may have contributed to the amount of solid tumour removed in the process of sample retrieval. This necessitated the generation of multiple explants per treatment which may impose practical challenges in a high-throughput laboratory or with small volume samples. EEC is typically a slow-growing cancer and resecting early-stage tumours could have led to inconsistent viability of the explants due to its non-aggressive nature. Moreover, explants from early-stage tumour are likely to contain non-cancerous epithelium and stromal cells. While we were interested primarily in the tumour compartment, these other cell types comprise the tumour microenvironment and could have had varying impacts on explant survival. Other cancer types where the long-term PDE model has been established, such as pancreatic [20] and ovarian [26] cancer, are highly aggressive and have high stromal content compared to EEC, which tends to be quite mucinous. As stroma are known to provide signals aiding tumour growth, invasion, and metastasis [30], PDEs from pancreatic or ovarian tumour would be more likely to survive ex vivo. This explains the lower degree of apoptosis observed in explants generated from pancreatic cancer compared to early-stage EEC [20]. Notwithstanding, the current study provides a method to maintain EEC PDEs that survive 21 days in culture in some cases. As stated for PDE models in ovarian cancer, one of the major limitations of this model is the small volume of tumour necessary to culture the tissue. This means that the explant itself is going to be only a small representation of the heterogeneity of the whole tumour [31]. High tumour heterogeneity, where very similar regions of tumour tissue contain tumour cell clones with differing resistance to LNG, may also have affected our explant viabilities and supports the notion that multiple explants from a single patient should be generated. Although this is a low-throughput setup, the EEC PDE model poses advantages over alternative higher throughput options like organoids, the greatest of which is a better reflection of in vivo tissue architecture. The PDE model could be used to complement high-throughput discovery work conduced in organoid models as well as 2D cell line models.

Most PDE cultures in the literature are short-term (between 48 h and 7 days), although we and others have developed long-term cultures (up to 30 days) [24,26,28]. Methods to increase nutrient and oxygen diffusion to maintain explant survival over time vary. For example, Karekla et al placed fragments of non-small cell lung carcinoma tissue onto a culture insert disc floating on cell culture media [28], whereas Abreu et al placed ovarian carcinoma PDEs in a flask with constant orbital shaking [26]. The most common technique, also utilised in the current study, is the use of a haemostatic gelatin sponge, which acts as the mediator between nutrient media and tissue [20,24].

As our aim was to establish a PDE model for testing surgery-sparing treatments, which are only considered for early-stage EEC, we focussed our sampling on stage 1 EEC (stage 1A and stage 1B). It is worth noting that although in some cases the post-surgical pathology assessment increased the patient’s grade or stage to outside these parameters, all patients had a pre-surgical assessment of early stage and/or low grade which would qualify them for LNG treatment. Two patients had been previously treated with LNG-IUS for EEC. One (Patient 3) was excluded from biomarker analysis due to a low amount of tumour within the explants, possibly the result of LNG-IUS treatment, and explants from the other (Patient 1) were used to validate the ex vivo model prior to commencement of LNG treatment.

Early-stage EEC is often driven by estrogen stimulation with deficient progesterone opposition [32,33]. As estrogen was not added to the culture media, this molecular driver of cell proliferation could have been lost and would explain the drop in Ki67 expression and increase in apoptosis from uncultured tumour compared to cultured explants. As we were assessing LNG, which is a progestogen, it was not feasible to include estrogen in the culture media as it could have interfered with LNG treatment. Due to the small size of the explants that represent different pieces of the original excised tissue, it was difficult to determine whether there were natural pockets of non-proliferating cells present within explants or whether cell proliferation was affected by culturing. We observed CC3 expression across all timepoints of ex vivo culture in Patient 1 –including at Day 0, and both CC3 and TUNEL staining was higher in cultured explants compared to uncultured tissue. This is unsurprising due to the significant environmental changes that the tissue experiences when being cultured using an ex vivo method, however it is a limitation that could be addressed in the future. In the other three time courses presented in the supplementary material there is variability in CC3 expression. Patient 8 displayed a similar profile to Patient 1 (S1 Fig), Patient 9 had CC3 positive cells at Day 0 which were maintained at each timepoint (S2 Fig), and Patient 10 had very little CC3 staining at Day 0, a high degree of staining at Day 7 and then appeared to recover by Day 14 as the CC3 positivity decreased (S3 Fig). Based on this variance, it is likely that the generation of successful long-term PDEs for early-stage EEC is patient-dependent.

A clinical test for LNG resistance in the form of a biomarker would be hugely beneficial. The most recent clinical trial reported complete pathological response to LNG-IUS treatment for EEC of 42% [15]. This means that over half of the patients who receive LNG-IUS treatment are not likely to respond. Follow up appointments to assess response occur roughly every 3–6 months [17]. There are many challenges with the treatment pathway for patients undergoing conservative management of EEC. Patients undergo pipelle or hysteroscopy dilation and curettage (D&C) to sample the endometrium in order to assess response to therapy. Patients require two negative biopsies to confirm response to treatment. Anecdotally, we observed high non-attendance for pipelle biopsies, and frequent distress or anxiety around in-clinic pipelles. This route of sampling can also result in ‘insufficient tissue’; pathologists cannot determine whether there is simply not enough cancer because of the sample method, or because the cancer has regressed. Furthermore, LNG can have a metaplastic effect on endometrial cancer, making it difficult for pathologists to diagnose ‘treatment effect’ or regression in comparison to persistent disease. Women who do not respond to LNG-IUD experience a delay of effective treatment which can decrease survivability [34]. An approved biomarker for LNG response would ensure that only those likely to respond are offered this treatment, reducing avoidable cancer progression and increasing timely treatment of disease.

To demonstrate the practicality of the ex vivo PDE model generated in this study, LNG quasi-resistant EEC tumours which were treated and not treated with LNG were investigated for their expression of candidate biomarkers. The quasi-resistant state of the explants was reflected in Fig 5B, where the degree of apoptosis identified by TUNEL was not significantly increased in the LNG treated explants compared to the untreated (negative) and vehicle controls. Previous cell line work in our laboratory generated EEC cells resistant to LNG [21]. Ensuing whole genome RNA sequencing and bioinformatic analysis revealed a number of differentially expressed genes in LNG resistant compared to LNG sensitive cell lines [22]. Many of these proteins had not been previously investigated in the context of EEC. Five of these proteins were further investigated using the PDE model in the current study. There was low SATB2 and CD80 protein expression in the explants studied. As the explants used in the current study represent a sample of quasi-resistant tumours (non-responders) it would be interesting to assess SATB2 and CD80 expression in tissue from LNG responders to determine their clinical applicability as biomarkers of LNG response. The remaining three candidate biomarkers investigated are involved in maintaining cancer stem cells (CSC), which are known to be more invasive and treatment resistant than standard cancer cells [35–38]. ALDH1A1 is a cytosolic enzyme which catalyses the production of retinoic acid from retinal, and is part of the reaction series that clears ethanol from the liver [39]. ALDH1A1 overexpression can be oncogenic, and it is thought to exert this effect by maintaining the CSC population [40]. ALDH1A1 was variably expressed in the quasi-resistant explants examined and may hold value as a predictive biomarker for LNG resistance in EEC. MAOA is a mitochondrial enzyme with primary function in the brain due to its involvement in the catabolism of monoamine neurotransmitters [41,42]. As such it has been implicated in mental illness and stress [43,44], but it also has a role in cancer progression. For example, in a mouse model of prostate cancer MAOA knockout mice had reduced risk of invasive cancer and a reduced population of CSCs [45]. Expression of MAOA in most of the explants warrants further investigation of this candidate biomarker. KLF4 belongs to the zinc-finger protein family of transcription factors and has diverse functionality including cell cycle regulation, p53 regulation, and induction of pluripotent stem cells [46–48]. KLF4 methylation has been investigated as a predictor of EC with increased CpG methylation being associated with increased risk of EC [49]. The consistent elevated expression of KLF4 in the explants examined combined with its expression in LNG-resistant endometrial cancer cell lines and primary cells [21] provides evidence for KLF4 as a potential biomarker for LNG resistance, and further investigation of this protein in a larger patient cohort is warranted. Encouragingly, ALDH1A1, MAOA and KLF4 were all expressed to some degree at Day 0 and after 21 days of both low and high doses of LNG treatment (Fig 6). This indicates that there is sustained expression of these markers long-term in tumours that are resistant to LNG treatment. Proteins that are contributing to LNG resistance would necessarily sustain their expression in order for resistance to persist. This provides further evidence for warranting closer examination of these candidate biomarkers in patient diagnostic biopsies [15,17,34].

The difficulties associated with both defining clinical LNG response and obtaining tissue biopsies for this PDE model outside of hysterectomy means that PDE models and other forms of in vitro models derived from this surgery are best used for studying mechanisms of LNG resistance in order to inform clinical decision-making/guidelines for recommending LNG-IUS to future patients. For personalising decision-making around proceeding with an LNG-IUS versus hysterectomy, a biopsy derived from a clinically routine pipelle targeting biomarkers, once established, would be more appropriate than a PDE.

This was a pilot study with the overall aim of developing and validating a PDE model of early-stage EEC. The sample size chosen was based on availability of patient tumour, length of time of the culture, and the survivability of the explants generated. The variability of response identified represents ‘real world’ patient variability and heterogeneity of the tumour. Further validation or subsequent studies that use the explant model would benefit from larger sample sizes. The explant model was used as a proof of concept to investigate LNG biomarkers and therefore the small sample size of the study limits the generalisability of the results. We recommend further investigation of these markers in larger cohorts, in particular in patient pipelle samples before and after treatment with the LNG-IUD. In this respect, the current study can be used to guide the feasibility of further examination of the candidate biomarkers examined. The development of long-term ex vivo PDE models has been successful in a number of cancer types. Overall, the benefits of using a biologically relevant PDE model can provide a useful tool to complement existing preclinical models. The amount of intra- and inter-patient variability in explant viability observed in the current study suggests that the use of a long-term ex vivo PDE explant culture for studying early-stage EEC would be best in low-throughput laboratories. Future studies could seek to improve this limitation. Although Collins et al reported that neither culture media nor serum supplementation had a significant effect on the viability of their EC explants [27], modifying culture conditions such as with the addition of growth factors used to sustain organoid cultures, or the EEC cell proliferation molecule estrogen could improve explant viability. An alternate culturing system with continuous media flow or orbital shaking could also be adopted [26]. Although we did not notice any major change in media colour over the weekend, PDE culture protocols for more aggressive cancers such as pancreatic recommend replenishing the nutrient media daily, which could improve the number of viable explants long-term [20]. Nevertheless, it is likely that this PDE model for early-stage EEC is better suited for shorter periods of time. Higher grade/stage EEC tumours, or those of a more aggressive histological subtype, such as serous, could be more likely to survive reliably long-term using the same methodology. However, this would not represent the disease state for LNG-IUD treatment. The model can be used to investigate protein expression for biomarker validation. Of the five candidate biomarkers of LNG resistance evaluated in this paper, KLF4 expression was high, while SATB2 and CD80 expression was low. Further investigation of ALDH1A1, MAOA, and particularly KLF4 in a patient cohort is warranted, and their involvement in maintaining cancer stem cell populations could lead to novel areas of investigation for non-surgical treatment of EEC and LNG resistance.

Supporting information

Acknowledgments

The authors wish to acknowledge the patients who donated their tissue. Additionally, we acknowledge Victoria University of Wellington and the Gillies McIndoe Institute for the use of their facilities and histology equipment, as well as Wellington Southern Community Laboratories and Medlab Central for performing the indicated immunohistochemistry.