Pre-clinical models to study abnormal uterine bleeding (AUB)

Another approach for studying AUB is the use of an ex vivo model, usually thinly cut whole tissue slices, mounted onto a porous membrane, and cultured under controlled conditions. This “ in vivo -like” model, preserves the natural architecture, cellular diversity and networks, cell viability, systemic gene and protein expression patterns, and pathway activity of the tissue of origin. 74 Often, this technique is employed for studying tumours of lung, prostate, gut, and breast, however, recent advancement also shows establishment of uterine fibroid (leiomyoma) organotypic cultures. 75 Salas et al describe uterine leiomyoma slices cultured on an alginate scaffold that respond to hormonal stimuli and express genes associated with leiomyoma development. The alginate scaffold has been constructed to contain microspheres with encapsulated drug, which offers a novel approach to investigate patient-specific drug responses. 75 The limitation of this model is accessibility to human-derived whole tissue biopsy and short culture maintenance of up to five days. 74 Despise this, the potential of organotypic ex vivo constructs as a personalised pre-clinical model for cytotoxicity studies to targeted therapies is widely recognised.

Data for this Review were identified by searches of the National Institute for Health and Care Excellence website, PubMed, Scopus, ScienceDirect and references from relevant articles using search terms “abnormal uterine bleeding”, “endometrial stromal/epithelial cell cultures”, “endometrial organoids”, “endometrial co-cultures”, “microfluidic”, “adenomyosis”, “leiomyoma “, and “models, animals”. Boolean operator “AND” and truncation strategies were used. Only articles published in English were included with focus on publications between 2015 and 2022. There are several exclusions where description of model development is discussed, and publication year precedes this period.

The regulation and mechanism of menstruation is ideally studied in humans. However, human heterogeneity and the inevitable disruption of endometrial architecture during tissue sampling may limit findings. Animal models of menstruation provide genetic homogeneity and allow genetic, environmental, and pharmacological manipulation to determine causation. Non-human primates menstruate and undergo spontaneous endometrial decidualisation. Rhesus macaques have similar uterine morphology and length of menstrual cycle to humans. 76 Both species display tightly co-ordinated spatial and temporal regulation of endometrial physiology at menstruation, e.g., increased MMPs and VEGF, vasoconstriction of spiral arterioles, and local tissue hypoxia. 77 , 78 These observations are preceded by progesterone withdrawal. 77 , 79 , 80 Macaques are reported to experience menstrual abnormalities (e.g. heavy menstrual bleeding) and may be fitted with tampons, and are excellent candidates for evaluating therapies for menstrual disorders. 78 Despite menstruating naturally, macaques may undergo ovariectomy and treatment with oestrogen and progesterone to create artificial menstrual cycles and accurate timing of endometrial sampling. Use of this model requires large experimental groups and long experimental times, resulting in significant costs. Therefore, many researchers now preferentially use rodent models to study menstruation. The mouse model of simulated menstruation was initially described in 1984 81 and further optimised in the 2000’s 82 ( Figure 5 ). Mice are ovariectomised and supplemented with exogenous oestrogen and progesterone to mimic the human hormonal endometrial environment. The model requires artificial induction of decidualisation, via a transcervical or surgical intrauterine injection of oil, reinforcing the importance of a decidualisation step prior to progesterone withdrawal in the physiology of menstruation. Once decidualisation has taken place, progesterone withdrawal results in histological and molecular changes analogous to those observed in the human endometrium at menstruation, with recruitment of leukocytes, 83 shedding of the luminal endometrium, and visible menstrual-like bleeding. 84 Endometrial tissue is then repaired and remodelled. Alternatively, simulation of menses may be achieved by inducing pseudopregnancy. In this model, female mice are mated with vasectomised males to mimic fertilisation events. Progesterone withdrawal may occur naturally, by ovariectomy, or by administration of a progesterone antagonist. 85 Figure 5 Mouse model of simulated menses. Ovariectomy is performed on female mice to deplete endogenous levels of ovarian hormones. After allowing 7 days for surgery recovery, mice are given daily subcutaneous injections of oestradiol (E 2 ) for three days (days 1-3). Seven days after the first E 2 injection, a progesterone (P 4 ) implant is subcutaneously inserted (day 7) along with a lower dose of E 2 (days 7-9). In order to trigger decidualisation, oil is transcervically administered at day 9. The removal of the P4 implant (t0) recapitulates menstrual events, with mice bleeding 8h after P 4 withdrawal (t8). Twenty-four hours after P4 withdrawal (t24), endometrial regeneration events can be observed. 81 , 82 Figure 5 Mouse model of simulated menses. Ovariectomy is performed on female mice to deplete endogenous levels of ovarian hormones. After allowing 7 days for surgery recovery, mice are given daily subcutaneous injections of oestradiol (E 2 ) for three days (days 1-3). Seven days after the first E 2 injection, a progesterone (P 4 ) implant is subcutaneously inserted (day 7) along with a lower dose of E 2 (days 7-9). In order to trigger decidualisation, oil is transcervically administered at day 9. The removal of the P4 implant (t0) recapitulates menstrual events, with mice bleeding 8h after P 4 withdrawal (t8). Twenty-four hours after P4 withdrawal (t24), endometrial regeneration events can be observed. 81 , 82 Despite endocrine (e.g., a shorter length of cycle and lack of spontaneous decidualisation) and immunological differences between mice and humans, 86 exogenous hormonal supplementation permits recapitulation of the events of human menstruation and decidualisation in this murine model. 5 , 87 , 88 Consistent with findings in the human endometrium, 89 the mouse model also has a critical period of progesterone withdrawal, where replacement of progesterone early after its withdrawal prevented menstrual-like bleeding and endometrial shedding. However, progesterone replacement at later time-points did not suppress menstruation. 90 It is also a popular model to investigate the dynamics of endometrial breakdown and repair and has been used to manipulate inflammation 91 and the endocrine environment 92 at menstruation, and to pharmacologically and genetically alter key factors involved at menses, for example, hypoxia 93 , 94 The xenograft mouse model provides an alternative model for in vivo examination of menstrual physiology and pathology. 95 , 96 , 97 In this model, fragments of human functional endometrium are xenografted to ovariectomised, immune-deficient mice, most commonly the severe combined immunodeficiency (SCID) mouse. Treatment with oestrogen and progesterone followed by removal of ovarian steroids resulted in menstrual breakdown of the xenografted human endometrium. Xenograft menstruation studies have examined endometrial regeneration and the role of ovarian steroids in regulating this process. 95 , 98 The advantages of this model are the ability to standardise ovarian hormone variations, manipulate the endometrium in a way that would be unethical in humans, and examine the local versus systemic leukocyte response by identification of human and mouse cell contributions. Limitations of the model include significant compromise of endometrial tissue architecture following transplantation, which may alter vascular and cellular responses. In addition, the necessary immunosuppressed state of the recipient mice may alter physiological inflammatory events in the menstrual endometrium. However, the commonly used SCID model aims to suppress T and B-cell mediated xenograft rejection without substantially affecting the innate immune response, and may be more relevant than other immunocompromised recipient mice. 95 The common spiny mouse ( Acomys cahirinus ) is the only known rodent to display spontaneous decidualisation and natural menstruation, with menstrual duration of 3 days and frequency of 6-10 days. 99 , 100 This rodent has previously been studied as an animal model for obesity and diabetes mellitus, and therefore, some laboratory reagents are currently available. The spiny mouse uterus is anatomically different to the human, but has physiological similarities, such as spiral arteriole remodelling in the perimenstrual phase. 99 In addition, endometrial decidualisation is tightly controlled, not compromising the structural integrity of the endometrial glands or the myometrium, as observed in other rodent models. Like women, the spiny mouse produces cortisol as its circulating glucocorticoid, as opposed to corticosterone in standard laboratory mice. This rodent may permit study of multiple successive menstrual cycles and examination of any pre-conditioning effects that menstrual cycles will have on endometrial physiology. Disadvantages include the variability of natural cycles, the current inability to genetically manipulate the model, limiting definitive mechanistic studies, and lack of specific antibodies and molecular biological reagents (e.g., primers). Adenomyosis occurs spontaneously in different mammals, including non-human primates, 101 mice, rats and rabbits. While these animal reports have proven to be useful for studying the development of adenomyosis, timing of onset and variability in the numbers affected limit use as a pre-clinical model for therapeutic intervention. In the 1980s, hormonal carcinogenic studies on mice revealed that exposure to certain endocrine-disrupting chemicals increased the risk of developing adenomyosis, alongside other uterine abnormalities such as cervical abnormalities, mammary carcinomas 102 and endometriosis. 103 Pre-natal oral exposure to diethylstilbestrol (DES), 102 dioxins 104 or phthalates 105 generated adenomyosis. However, the incidence is strain-dependent, with DES-exposed CD-1 mice being protected from developing adenomyosis, but prone to uterine malignant tumours, 106 and BALB/c mice showing complete protection from any uterine abnormalities. 102 Post-natal DES administration also displayed varying results depending on administration route. For example, using the same mouse strain, intraperitoneal administration of DES generated adenomyosis, 107 while subcutaneous administration showed development of uterine carcinomas. 108 Therefore, induction of adenomyosis using endocrine-disrupting chemicals is limited by lack of specificity and a long induction time (6 to 20 months after treatment) that introduces variability with age. One interesting finding in the DES model was the increased incidence of hyperprolactinemia observed in mice developing adenomyosis. 102 This observation triggered the development of adenomyosis models relying on hyperprolactinemia replication via implantation of pituitary grafts from other mice or pharmacological induction with dopamine agonists, or selective serotonin reuptake inhibitors. 109 , 110 Interestingly, the number of pituitary glands implanted, and the donor-recipient match/mismatch did not influence adenomyosis incidence. 109 However, the site of implantation did affect incidence, with transvaginal 111 or surgical uterine lumen implantation 112 displaying increased adenomyosis rates than kidney capsule implantation. 113 , 114 As in previous models, adenomyosis incidence was strongly strain-dependent, with BALB/c mice being protected compared to other strains. 115 In contrast to the endocrine-disrupting chemical models, this pituitary grafting model displays a shorter endpoint (∼3 months) and has also been validated in rats. 116 This model has contributed to research on novel therapeutics 117 , 118 and the study of molecular aspects of adenomyosis. 113 , 119 Limitations include the significant variability in adenomyosis induction rates, even within the same research group. 114 , 120 In addition, specific induction of adenomyosis remains elusive, with co-development of other uterine abnormalities. 112 The most popular current model of adenomyosis is the tamoxifen model. 121 Oral administration of this selective oestrogen receptor modulator from day 2 to 5 after birth results in an adenomyosis induction rate of 100%. 121 One of the reasons for its success is the short induction time of the lesions: at days 5-10 there is early evidence of adenomyosis 121 and by day 42 from birth, the model is completely established. 122 Despite its time advantage, the tamoxifen model presents similar limitations to the previous models, displaying strain and route of administration dependency with C57/BL6J mice protected 123 , and subcutaneous injection resulting in uterine carcinoma development. 123 Very recently, Hao et al. have developed two mouse models of adenomyosis by disrupting the interface between the endometrium and the myometrium. 123 This disruption is successfully achieved by either a mechanical or a thermal stimulus. With mechanical induction, a microcatheter is inserted into a uterine horn and the tip used to damage the endometrial-myometrial interface. For thermal induction, the microcatheter is attached to an electrosurgical scalpel, generating disruption by electrocoagulation. Although technically more complex, this model seems to be strain independent and provides an opportunity to adjust the severity of the lesions by increasing the damage generated . 123 Uterine fibroid (leiomyoma) formation in rodents occurs following exposure to endocrine-disrupting chemicals. 124 Rat neonatal exposure to DES, bisphenol A, 125 or tributyltin 126 promotes the development of leiomyoma, as well as other uterine abnormalities such as glandular hyperplasia 127 or adenomyosis. 107 Additionally, some mouse strains also report leiomyoma incidence when prenatally or neonatally exposed to some of these compounds. 128 Despite the non-specificity of their phenotype, these models are still in use for the exploration of potential treatments against leiomyoma formation and development. 129 However, genetic models such as the Eker rat model have gained popularity. Eker rats, which are defective for the tumour suppressor gene Tsc2, spontaneously develop uterine fibroids in approximately 65% of cases. 130 Alterations in this gene are also observed in some women with leiomyoma. 131 The course and composition of the lesions in rats are similar but not identical to human lesions. 132 Eker rats have proven useful to study the hormonal influence on the development of leiomyomas. 133 In mice, the Tsc2 mutation phenotype is not replicated but mutations in the Med12 gene do generate leiomyomas. 134 This gene is affected in up to 80% of women with leiomyoma, 135 offering another potential model of genetic predisposition. The most common critique for both these genetic models is that a single mutation may oversimplify this highly complex uterine condition. As a refinement option, the Eker rat model can be combined with endocrine-disrupting chemicals. Neonatal exposure to DES in Eker rats significantly increased leiomyoma formation from 65% to a 100%. 136 This combined model has been used to study the molecular triggers of uterine leiomyoma development as well as potential treatments. 137 The mouse xenograft model is an established pre-clinical model for developing novel therapeutics to suppress leiomyoma growth. 138 , 139 In the most conventional model, leiomyoma tissue derived from patients is cut in sections or disaggregated and further implanted on different immunosuppressed mice strains. The preferred graft location tends to be subcutaneous 140 , although some reports suggest that sub-renal implantation displays better engraftment. 141 Intrauterine transplantation has also been tested but displays less leiomyoma formation than other routes. 142 Tissue sections and freshly isolated cells derived from tissue disaggregation display similar reproducibility and level of engraftment. 143 Selection of recipient mice remains controversial, with some studies claiming that the optimal engraftment is achieved using SCID mice, 132 and other reports favour lymphocyte T, B, and NK cells-deficient mice (NOD/SCID mice). 141 The xenograft model has also been tested on rats where, rather than generating a genetic immunodeficiency, animals are treated with mycophenolate mofetil to avoid transplant rejection. 144 In every model, hormonal supplementation is essential for graft survival, with administration of both oestrogen and progesterone displaying the best results. 143 This xenograft model has been used to transfect tissue (or cells) with plasmids, 145 as well as microRNAs to study the impact on their development. 146 Very recently, a new xenograft model was validated, where human leiomyoma cell lines were 3D-cultured prior to subcutaneous implantation in immunosuppressed mice. 147 This removes the requirement for fresh tissue samples, reduces donor intra-variability, and showed longer graft survival (over 8 weeks). 132

Hilary OD Critchley receives support from the Medical Research Council Centre for Reproductive Health (MRC CRH) Grant MR/N022556/1; and from Biotechnology and Biological Sciences Research Council (BBSRC; BB/S002995/1) Aleksandra O Tsolova receives support from the Medical Research Council Centre for Reproductive Health (MRC CRH) Grant MR/N022556/1 and student grant MR/P502030/1 Jacqueline A Maybin receives support from Wellcome Trust Fellowship 209589/Z/17/Z and from the Royal Society of Edinburgh 1077. Rocío Martínez Aguilar receives salary support from Wellcome Trust Grant 209589/Z/17/Z. The funders had no role in manuscript design, data collection, data analysis, interpretation, writing of this manuscript.

The first step in designing a pre-clinical model to study AUB is an understanding of the individual components of the uterus and the endometrium. The uterus comprises the outer serosa, myometrium, and the endometrium. The myometrium is a smooth muscle cell multilayer that undergoes contraction during labour and is the site of the structural causes of AUB, adenomyosis and fibroids (also known as leiomyomata or leiomyomas). 4 The endometrium is composed of two layers: the basal layer, containing progenitor cells, and a functional layer, which serves as the site for embryo implantation if pregnancy occurs ( Figure 2 , top). The function of both layers is regulated by circulating oestrogen and progesterone ( Figure 2 , bottom). In the absence of pregnancy, the functional layer of the endometrium sheds during menstruation and is then rapidly repaired through a complex sequence of events involving inflammation, angiogenesis, and tissue remodelling. 1 Figure 2 The menstrual cycle comprises three functionally distinct phases – proliferative, secretory, and menstrual, that are regulated by sex steroid hormones – oestrogen and progesterone . Figure 2 The menstrual cycle comprises three functionally distinct phases – proliferative, secretory, and menstrual, that are regulated by sex steroid hormones – oestrogen and progesterone . The endometrium is a multicellular tissue comprising stromal, epithelial, endothelial, and immune cells including uterine natural killer cells (uNK), neutrophils and macrophages. 1 , 5 Modern cellular techniques employ these endometrial cells for the development of several pre-clinical models to delineate endometrial pathophysiology.

Endometrial function through the menstrual, proliferative, and secretory phases is tightly regulated by hormonal endocrine signalling. Circulating ovarian sex steroids oestrogen and progesterone are crucial for this regulation. Oestrogen stimulates endometrial cell proliferation and regeneration following menstruation; progesterone opposes oestrogen-stimulated proliferation and stimulates cell differentiation in preparation for implantation. If pregnancy does not occur, progesterone is withdrawn, and this triggers menstruation as the new menstrual cycle begins. 1 Dysregulation of these processes may lead to abnormalities in endometrial breakdown, bleeding, repair, and regeneration. Normal uterine bleeding is defined using four parameters: frequency, duration, volume, and regularity. A typical menstrual cycle (i) has a frequency of 24–38 days, (ii) a regularity (cycle to cycle variation) of 2–20 days, (iii) has a duration of 4.5–8 days, and (iv) has a volume of 5–80 millilitres. Abnormal Uterine Bleeding (AUB) is a highly prevalent symptom of an underlying condition/s experienced by one in three women of reproductive age. AUB, which includes heavy menstrual bleeding (HMB), causes significant morbidity and affects every aspect of the sufferer's life. A clear definition, description of the causes, and signs of AUB has been established, which “ should assist in providing a solid basis for the standardisation of international research and clinical manuscripts addressing the diagnosis, pathogenesis, and management of AUB ”. 2 , 3 The International Federation of Obstetrics and Gynaecology (FIGO) Menstrual Disorders Committee (MDC) published the two FIGO systems (terminology; System 1 and classification; System 2) in 2011 with revisions in 2018. 4 System 1 defines AUB as a symptom presenting values of menstrual bleeding frequency, duration, volume, and regularity outside of the normal range described above. 2 Acute AUB is an episode of heavy menstrual bleeding (HMB) that requires immediate intervention. Chronic AUB is AUB present for most of the past 6 months that does not require urgent medical attention. 3 FIGO system 2 classifies AUB using the acronym PALM-COEIN [ pahm-koin ] describing two groups of AUB aetiologies – “ PALM ” describes the structural, and “ COEIN ” - the non-structural causes ( Figure 1 ). Figure 1 FIGO classification of abnormal uterine bleeding (AUB) – structural (PALM) and non-structural (COEIN) causes (4) . Figure 1 FIGO classification of abnormal uterine bleeding (AUB) – structural (PALM) and non-structural (COEIN) causes (4) . Abnormal uterine bleeding can be caused by none, one, or multiple PALM-COEIN aetiologies. Some may be present but asymptomatic (AUB-L, see Figure 1 ), and others can only be diagnosed by exclusion of other identifiable factors (such as underlying endometrial anomaly, AUB-E). 4 This classification system is essential for scientific and clinical progress in this area. Non-structural causes such as AUB-E require delineation of mechanism of action for treatment improvement, whereas structural causes require delineation of endometrial impact as a cause (or lack) of AUB. Only by gaining understanding of the above, can we then address the complexity of the presence of more than one cause of AUB in a single patient. Advancement in our understanding of endometrial pathophysiology has been possible through use of modern cellular and molecular techniques, as well as animal models. This has facilitated investigation of therapeutic targets and new treatment options in the wider goal of improving precision for patient management. 1

Pre-clinical models have significantly advanced our understanding of endometrial function and AUB. While every pre-clinical model offers research advantages, each has specific limitations as discussed. For this reason, careful consideration and selection criteria are required when choosing the right model system or combination of models for a specific research question. Correct research approaches can answer scientific questions, lead to therapeutic target identification/validation and inform clinical trial design. It is an exciting time for research in the reproductive field and recent advancements have the potential to identify new approaches to ameliorate abnormal uterine bleeding (AUB).

Broadly, pre-clinical modelling is any experimental preparation in a laboratory setting which enables an in-depth investigation of a complex biological phenomenon. Essential to the development of a pre-clinical model is the consideration of the model's intended purpose. This will establish validation requirements of the model. In the context of studying human menstrual disorders, i.e., AUB and endometrial function, a pre-clinical model must present the same normal physiological response to sex steroid hormones as is observed physiologically . To study endometrial cell-cell interactions, the model needs construction with endometrium-derived cells with preserved phenotype when extracted from their natural microenvironment. In other cases, the model requires the use of undifferentiated progenitor cells, and their phenotype must be induced in vitro to reflect the tissue of origin. In vitro models have advanced to offer sophisticated approaches to mimic normal endometrial physiology in 3D environments. This makes them a preferred method when studying cell interactions and function as their scientific potential lies in the deconvolution of the system and ease of manipulation. If the purpose of the model, however, is to explore physiological functions or systemic interactions between organs, whole organism models such as mouse or non-human primates may be better suited or complement the in vitro systems.

The evolution of in vitro models of endometrial-derived, multi-cellular, and three-dimensional constructs have shown their potential for uncovering the function of individual cell types, inter-cellular interaction, and cell interaction with its microenvironment. The advancement of the complexity of these cultures has served as a proof-of-concept that establishment of an in vivo -like endometrial microenvironment is possible. Forward-looking, modern gene and protein discovery techniques can now begin to describe in a non-biased manner the key cellular and molecular responses of the uterus that result in structural and non-structural causes of AUB. Non-structural causes such as AUB-E require delineation of mechanism of action for treatment improvement, whereas structural causes such as uterine fibroids require delineation of endometrial impact as a cause (or lack) of fibroid-associated AUB and vice versa . Gaining such understanding can be possible with the emerging in vitro models, which can help address the complexity of the presence of more than one cause of AUB in a single patient. In addition, in vivo models are still necessary to study the organ-to-organ interactions which may equally impact normal and abnormal uterine function.

Hilary Critchley (HC) has received clinical research support for laboratory consumables and staff from Bayer AG and provides consultancy advice (paid to Institution) for Bayer AG, Gedeon Richter, Vifor Pharma UK Ltd,; Myovant Sciences GmbH. HC has received royalties from UpToDate for article on abnormal uterine bleeding. Jacqueline A Maybin receives Wellcome Trust Clinical Research Career Development Fellowship (salary support) - To institution; Tenovus Scotland - To institution; Royal Society of Edinburgh - To institution; Wellbeing of Women - To institution.

All authors made significant contributions to manuscript preparation and read and approved the final versions and revisions of the manuscript. Aleksandra O Tsolova: literature search, conceptualisation, figures, writing – original draft, and writing – review & editing. Rocío Martínez Aguilar: literature search, figures, writing – original draft, and writing – review & editing. Jacqueline A Maybin: writing – original draft, and writing – review & editing. Hilary OD Critchley: conceptualisation, original draft; writing — review & editing.