Local estrogen metabolism (intracrinology) in endometrial cancer: A systematic review

Endometrial cancer (EC) is the most common malignancy of the female gynaecological tract and increased exposure to estrogens is a risk factor. EC cells are able to produce estrogens locally using precursors like, among others, adrenal steroids present in the serum. This is referred to as local estrogen metabolism (or intracrinology) and consists of a complex network of multiple enzymes. Particular relevant to the final generation of active estrogens in endometrial cells are: steroid sulfatase (STS), estrogen sulfotransferase (SULT1E1), aromatase (CYP19A1), 17β-hydroxysteroid dehydrogenase (HSD17B) type 1 and type 2. During the last decades, a plethora of studies explored the level of these enzymes in EC but contrasting data were reported, which generated vigorous debate and controversies. Several reviews attempted at clarifying some of the debated issues, but published reviews are based on investigator-defined bibliography selection and not on systematic analysis. Therefore, we performed a systematic review of the literature reporting about the level of STS, SULT1E1, CYP19A1, HSD17B1 and HSD17B2 in EC. Additional intracrine enzymes and networks (e.g., HSD17Bs other than types 1 and 2, aldo-keto reductases, progesterone and androgen metabolism) were non-systematically reviewed as well.


Intent of this review
Local synthesis and degradation of steroids, also called intracrinology, represents the way tissues can control hormone levels and exposure in the local microenvironment. Several studies show that intracrinology is relevant for both homeostasis and physiological functions of distinct tissues, but can also lead to the establishment/maintenance of pathologic conditions (Konings et al., 2018a). The pathophysiology of the endometrium is regulated by estrogens, hence the intracrinology of these steroids controls the endometrial physiology and can cause disorders like infertility, endometriosis and endometrial cancer (EC) (Brosens et al., 2004;Cornel et al., 2012;Delvoux et al., 2014;Konings et al., 2018a).
During the last decades, intracrinology of EC has been thoroughly investigated but several aspects remain still controversial. Therefore, in an attempt to solve such controversies and for a better understanding of the impact of intracrinology in EC, we provide a complete and exhaustive summary of the current literature of EC intracrinology through a systematic review. and is related to increasing ageing, to lifestyle and increasing obesity (Kaaks et al., 2002;Morice et al., 2016). About 14% of all cases are diagnosed in premenopausal women (a minority of which may have not yet completed childbearing), whereas the vast majority of ECs occurs after menopause. Due to the presence of vaginal blood loss as an early symptom of disease in postmenopausal women, more than 75% of EC patients are diagnosed at an early stage, resulting in a favourable prognosis.
The leading risk factors for the development of EC are related to a prolonged estrogen exposure, as encountered in case of obesity, nulliparity, early menarche, late menopause (> 55 years), polycystic ovarian syndrome (PCOS), infertility, and treatment with tamoxifen (Kaaks et al., 2002;Luo et al., 2014;Lv et al., 2017;Renehan et al., 2008).

Diagnosis and treatment of EC
Several tools are used to diagnose EC, i.e., ultrasound, endometrial sampling and occasionally hysteroscopy, followed by additional diagnostic examinations to estimate the spread of the disease (chest X-ray, abdominal and thoracic CT scan; Colombo et al., 2016).
In case of clinical early stage disease, patients undergo hysterectomy with bilateral salpingo-oophorectomy. In patients with early stage EC and high-risk histological features, lymph node sampling as part of a surgical staging procedure is also performed.
After surgery, the final histopathological examination is performed and definitive disease stage is determined according to the International Federation of Gynecology and Obstetrics (FIGO) system (Werner et al., 2012).
Prognostic markers like depth of myometrial invasion, histological type, grade and lymph vascular space invasion can be used to guide the post-surgical care (Creutzberg et al., 2000;Holloway et al., 2017;Keys et al., 2004;Morice et al., 2016;Murali et al., 2014;Tejerizo-Garcia et al., 2013;Todo et al., 2010). Based on these prognostic markers and on the FIGO stage, adjuvant radiotherapy and/or chemotherapy can be advised (Colombo et al., 2016;Morice et al., 2016). In case of advanced or recurrent disease, endocrine treatment can also be considered (see below, paragraph 2.4).

Classification of EC
The most common and widely used method to classify EC consists of two different histological types: type I and type II (Morice et al., 2016;Remmerie, 2018). Type I is the most commonly diagnosed form (about 80% of the cases) and is related to estrogen exposure. It is further characterised by a well differentiated, endometrioid histology (e.g., with the presence of clear glandular epithelium), expression of hormone receptors, diploidy and is frequently diagnosed at a low stage and associated with a good prognosis. Type II EC is associated with high stage of disease, mostly non-endometrioid type (serous endometrial carcinoma, clear cell carcinoma or mixed carcinomas), high histological grade, aneuploidy, absence of hormone receptors, often TP53 mutations and poor prognosis (Morice et al., 2016;Remmerie, 2018). In recent years, the distinction between type I and II EC became highly debated because it does not accurately describe important patient characteristics, it is prone to errors and to a certain level of subjectivity. For instance, disease recurrence among good prognostic type I tumours occurs in a high percentage of the patients (20% of the cases). In contrast, among patients with type II tumours, expected to have very high recurrence rates, about 50% show no recurrence. Furthermore, there are subtypes of ECs with mixed characteristics that hardly fit in one of the two groups, like high-grade endometrioid ECs that histologically resemble type I and well-differentiated tumours, however they clinically behave like type II ECs (Goebel et al., 2018;Rutgers, 2015;Zannoni et al., 2010).
Recent efforts to obtain a more objective classification system were based on the analyses of molecular tumour features (Kamps et al., 2017;Kandoth et al., 2013;Remmerie, 2018;Talhouk et al., 2015). In particular, The Cancer Genome Atlas (TCGA) project used a multi-omics approach to comprehensively catalogue and characterise somatic mutations (whole genome sequencing), expression profiling (exome sequencing), microsatellite instability (MSI), and copy number variations among over 230 EC specimens. This analysis resulted in four EC clusters, three of which displayed endometrioid characteristics (polymerase ε -POLE -ultramutated, MSI hypermutated, copy-number low), whereas the forth group (copy-number high) showed in most cases a serous (type II) histology. Such classification correlates well with patient prognosis (Kandoth et al., 2013;Talhouk et al., 2015) and a molecular classification tool called ProMisE (Proactive Molecular Risk Classifier for Endometrial Cancer) has been designed and validated retrospectively (Kommoss et al., 2018;Remmerie, 2018).

Prognosis of patients with EC
Survival in EC patients is associated with FIGO stage, histology, and specific molecular marker profiles. The five-year overall survival is 74-94% for stage I-II, 57-66% for stage III and 20-26% for stage IV disease (Morice et al., 2016).
In cases of recurrent disease, the patient survival is related to the amount and localisation of recurrence and to the treatment possibilities (Colombo et al. 2016: ENITEC). Beside the recent TCGA molecular classification, several other efforts are on-going to define more accurate prognostic methods based on novel (combined) molecular and histologic markers (ENITEC; Talhouk et al., 2015).

Estrogen signalling in EC and its clinical significance
In the endometrium, 17β-estradiol (i.e. the most active estrogenic compound) controls, together with progesterone, the menstrual cycle, the reproductive functions and induces cell-proliferation (Groothuis et al., 2007;Robertshaw et al., 2016). Anti-estrogen hormonal treatments may have potential therapeutic effects, as is the case with other hormone dependent conditions like breast cancers. However, the use of endocrine drugs in EC patients is associated with low response rates and with the development of insensitivity, therefore, these drugs are currently indicated in case of advanced stage/recurrent disease, as palliative care, or for fertility preservation treatments (Colombo et al., 2016;Yamazawa et al., 2007;Yang et al., 2005). In case of advanced stage/recurrent EC, progestogens are the first line endocrine drug option, but also inhibitors of CYP19A1 (aromatase inhibitors, AIs) can be used as second/third line treatment. The use of AIs for EC in pilot studies is reported for different clinical settings, such as neoadjuvant, palliative or adjuvant therapy (Altman et al., 2012;Barker et al., 2009;Berstein et al., 2005a;Berstein et al., 2002;Quinn et al., 1981;Zhang and Cao, 1996). Phase II trials evaluating the use of an AI as single agent in advanced stage disease report response rates not higher than 10% (Lindemann et al., 2014;Ma et al., 2004;Rose et al., 2000;Steed and Chu, 2011;Thangavelu et al., 2013), whereas dual regimen, e.g., AI with an mTOR inhibitor, show more promising data (Slomovitz et al., 2015).
The blood levels of 17β-estradiol are controlled by the hypothalamus-pituitary-gonad axis before menopause and range between 5pM and 1000pM (depending on the phase of the menstrual cycle). In postmenopausal women, when the ovaries are no longer active, the systemic levels of 17β-estradiol drop to values that are lower than 80pM (Konings et al., 2018a;Mueller et al., 2015;Plant, 2015). However, the intra-tissue steroid levels (or intracrine levels) are not the same as those in the serum (Huhtinen et al., 2014;Tanaka et al., 2015) because intracrine enzymes expressed locally in endometrial cells use circulating blood substrates to generate and deactivate 17β-estradiol and other steroids (Konings et al., 2018a;Labrie, 2015).

Intracrine production and deactivation of 17β-estradiol
The last and central step in the intracellular production of 17β-estradiol consists in the reduction of the 17-keto moiety into the 17βhydroxyl of the steroid scaffold, thus converting estrone into 17β-estradiol ( Fig. 1). Since 17-keto steroids like estrone bind to the corresponding steroid receptor with low efficiency, estrone has much lower estrogenic activity compared with 17β-estradiol, which is a strong ER binder. The balance between estrone and 17β-estradiol is controlled by 17β-hydroxysteroid dehydrogenases (HSD17Bs). There exist 14 HSD17B types, and many of those have been implicated in estrogen metabolism. However, recent studies and knockout genetic mouse models indicate that HSD17B1 has the highest catalytic efficiency to reduce estrone to 17β-estradiol and is relevant in vivo, whereas HSD17B2 catalyses the opposite reaction, i.e., the deactivation of 17βestradiol to estrone (Hakkarainen et al., 2015;Konings et al., 2018a;Moeller and Adamski, 2009;Prehn et al., 2009;Saloniemi et al., 2010;Shen et al., 2009).
An in vivo role of HSD17B7 and 12 in the estrone/17β-estradiol redox balance cannot be completely excluded whereas that of other HSD17Bs is negligible. Nevertheless, HSD17Bs and other enzymes like aldo-keto reductases (AKR) are involved in the intracrine metabolism of androgens and progestogens that can in turn influence the estrogen signalling. These pathways are briefly discussed in section 6 ('Additional intracrine networks') and were recently thoroughly reviewed (Konings et al., 2018a).

Sulfatase and aromatase pathways
The most important source of estrone for the intracellular synthesis of 17β-estradiol is the blood, where estrone is present in a free state or in its sulfated form, i.e., estrone-sulfate (estrone-S). Due to their high water solubility, sulfo-conjugated steroids are abundant in the serum, and since they have longer half-life and stability than unconjugated forms, they represent a reservoir of steroids for the local synthesis (Mueller et al., 2015;Rizner, 2016). Estrone-S is converted into active (free) estrone by the enzyme steroid sulfatase (STS) and transported through the cell membrane with the aid of specific transporter proteins (Rizner et al., 2017), which will not be further discussed in this review. Intracellular estrone can be either activated to 17β-estradiol by HSD17B1 or can be converted back to estrone-S by estrogen sulfotransferase (SULT1E1).
The balance between STS and SULT1E1 influences the amount of intracellular substrate (estrone) available for further conversion into 17β-estradiol and together represent the sulfatase pathway (Mueller et al., 2015;Rizner, 2016;Fig. 1). Although the serum level of estrone-S in postmenopausal women is about 1000 fold lower than in women prior to menopause (2pM versus 2-5 nM; Konings et al., 2018a;Mueller et al., 2015), serum estrone-S can be considerably increased in women with EC (up to a few hundred pM) (Audet-Walsh et al., 2011).
The sulfatase pathway is active in various human tissues like lungs, vessels, thyroid, uterus, liver and testis (Foster et al., 2008;Konings et al., 2018a;Miki et al., 2002;Mueller et al., 2015;Purohit and Foster, 2012;Rizner, 2016). STS is also catalytically active on adrenal dehydroepiandrosterone-S (DHEA-S). Free DHEA and DHEA-S have serum concentrations in the range of nano-molar and micro-molar, respectively (Konings et al., 2018a;Mueller et al., 2015;Rizner, 2016), and are considered important precursors of most steroids through the conversion of DHEA into androstenedione by the enzyme 3β-hydroxysteroid dehydrogenase (Konings et al., 2018a). Sulfation of DHEA is catalysed by a DHEA specific SULT (Mueller et al., 2015). Neither DHEA nor its metabolism will be further discussed in this review.
Androstenedione and the other androgenic compound testosterone are also present systemically in women, with blood concentrations of about 1.2-1.8 nM and 0.2-1.0 nM, respectively and irrespective of the menopausal status (Audet-Walsh et al., 2011;Konings et al., 2018a;Mueller et al., 2015). Androgens are substrates for the intracrine synthesis of estrone and 17β-estradiol via the aromatase pathway ( Fig. 1). This reaction is catalysed by the enzyme aromatase (CYP19A1). CYP19A1 has restricted expression throughout human tissues. Among peripheral tissues (i.e., non-endocrine glands), adipose tissue shows the highest CYP19A1 expression. Here, produced estrogens are released in the blood circulation. Hence, CYP19A1 in the adipose tissue also contributes to the circulating (systemic) level of estrogens, especially in obese subjects. Low active estrone is converted to active 17β-estradiol by 17β-hydroxysteroid dehydrogenase type 1 (HSD17B1), whereas HSD17B2 catalyses the opposite reaction. A similar balance between androstenedione and testosterone is controlled by reductive HSD17B5 and the testis-specific HSD17B3, and by oxidative HSD17B2. Aromatase (CYP19A1) converts androgens into estrogens, whereas sulfatase (STS) hydrolyses 3-hydroxyl-sulfate moieties of estrone (as well as other steroids such as DHEA). Estrogen sulfotransferase (SULT1E1) catalyses the opposite reaction and is specific for estrogenic compounds. Chemical structures were downloaded from the Human Metabolome Database (Wishart et al., 2013).
It is interesting to note that the blood levels of 17β-estradiol and many additional steroid precursors (testosterone, androstenedione, DHEA, DHEA-S, estrone and estrone-S) are significantly increased in EC patients compared with healthy controls (Audet-Walsh et al., 2011;Lepine et al., 2010).

Systematic review on EC intracrinology
For the present systematic review, we focus on the enzymes that based on the current knowledge play a major role in the endometrial synthesis of 17β-estradiol and for which published literature describes controversial results: i.e., STS, SULT1E1, CYP19A1, HSD17B1 and HSD17B2 (Fig. 1). Additional enzymes and networks involved in the intracrine steroid metabolism are briefly discussed in section 6 and were recently reviewed (Konings et al., 2018a).

Methods
The systematic review was performed according to the PRISMA criteria (Liberati et al., 2009). Our primary outcome was descriptions and investigations on the intracrine estrogen metabolism focused on the enzymes: HSD17B1, HSD17B2, STS, SULT1E1 and CYP19A1. The search was performed on the 26th of September 2017 and conducted in PubMed (MEDLINE) and Embase databases. References in review articles were controlled for any missing article and eventually included.
Our key search terms included: estrogen metabolism AND endometrial cancer.
- The following limitations were set: publication year earlier than 1990 and languages other than English. This search returned a total of 2163 articles. Additionally, Embase and Medline were searched with broader terms: endometrial cancer AND estrogen metabolism AND 17beta hydroxysteroid dehydrogenase; endometrial cancer AND estrogen metabolism AND steroid sulfatase; endometrial cancer AND estrogen metabolism AND estrogen sulfotransferase; resulting in 122, 8 and 46 additional articles respectively which were not shown in the initial search. Total articles retrieved were 2339 (Fig. 2).

Article selection based on title and abstract
Article selection based on title was performed in a standardised method by two independent authors (KMCC, AR, Fig. 2). Inclusion criteria were: endometrial cancer (carcinoma) and/or normal endometrium and/or enzymes and/or receptors and/or serum steroid levels. Disagreements were resolved by discussing the title and abstract together. Based on title selection, 1445 articles remained for further selection by abstract. Inclusion criteria were the use of endometrial cancer tissue and description of one or more of the aforementioned enzymes. Exclusion criteria were the use of cell lines only, medication studies related to tamoxifen/SERMs, description of association between EC and genetic variants in STS, SULT1E1, CYP19A1, HSD17B1, HSD17B2, description of serum steroid levels not assessed by liquidchromatography tandem mass spectrometry (LC-MS) and the exploration of sarcomas and carcinosarcomas only.

Final outcome of the systematic analysis
After abstract selection, 86 articles remained for full-text analysis, of which 56 were excluded for the following reasons: 28 papers were review articles; eight articles described genetic variants; seven articles described diseases other than endometrial carcinomas; five studies described the effect of aromatase inhibitors; one study described the ER/ PR status only; six studies described steroid levels based on methods other than LC-MS; one study was performed in rodents. A total of 30 articles were left and included in our systematic analyses. References in these articles were controlled for missing inclusions. Six articles were included manually (Fig. 2), of which two were added after reference check (Cornel et al., 2017b;Lepine et al., 2010) and four were published after the search was performed (Konings et al., 2018b;Sinreih et al., 2017a;Sinreih et al., 2017b;Sinreih et al., 2017c). The 36 selected papers are listed in Supplemental Table 1, which reports the main characteristics and the study design. Two of these studies, though they did not focus specifically on EC, were not excluded from our analyses because one was the first describing STS activity in endometrium (the studied disease condition was leiomyoma (Yamamoto et al., 1990b), whereas the second paper studied endometrial hyperplasia in presence or absence of polycystic ovarian syndrome (PCOS), both risk factors for EC (Bacallao et al., 2008).
In the following part of this section, each enzyme will be discussed separately with respect to the mRNA, protein and enzyme activity levels. Results are overviewed in Tables 1-5.

STS
STS converts estrone-S, circulating in the blood, into free estrone. Thirteen articles were included for the present analyses (see Table 1).

STS mRNA expression
In 2004, Utsunomiya and co-workers detected STS mRNA expression using RT-sqPCR, but described no difference between normal endometrium and EC (Utsunomiya et al., 2004). Nevertheless, these authors found a significant positive correlation between STS immunereactivity and mRNA levels in the 21 EC tissues analysed with both techniques.
In 2006 Smuc et al. described a decreased STS mRNA level measured by RT-qPCR in 12 EC specimens compared with adjacent normal endometrium . However, in two later studies on larger (and overlapping) cohorts including 25 and 55 EC specimens and adjacent normal tissue, the same authors did not observe any difference between diseased and healthy tissue (Sinreih et al., 2017a;Smuc and Rizner, 2009). In the most recent among these studies (Sinreih et al., 2017a), the authors also assessed the STS mRNA levels in relation to the histopathological features of the specimens. In this case, STS mRNA was decreased in grade 3 EC (n = 5) compared with adjacent normal endometrial tissue, whereas it was unchanged in grade 1 EC (n = 32) or grade 2 EC (n = 8) compared with the adjacent control endometrium (Table 1).
A 2.2 fold increase of the mRNA expression of STS in EC (n = 48) compared with its adjacent normal endometrium (available for 33 samples) was described by Lepine and co-workers (Lepine et al., 2010; Table 1).
Cornel et al. explored the mRNA level of STS determined by microarray analysis on a large cohort of EC cases (n = 175; Table 1), and high STS mRNA levels were associated with endometrioid histology and high grade EC (Cornel et al., 2017b).
Utsunomiya and co-workers detected STS immune-reactivity in EC but not in normal endometrium (Utsunomiya et al., 2004; Table 1). Sinreih et al. compared the STS protein levels in EC and normal adjacent endometrium by western blotting (24 paired samples were analysed out of a total of 55 included in their study cohort) or immunohistochemistry (44 paired samples out of 55). Beside a high variability in the expression levels, these authors reported lower STS expression in EC compared with the normal adjacent endometrium (Table 1). Such decreased expression was restricted to high grade lesions when analysed by western blot, but was extended to the total group of specimens using immunohistochemistry (Sinreih et al., 2017a).

STS enzyme activity
STS activity was assessed in six studies using either organic-aqueous phase separation or TLC (Table 1). Four of these studies reported STS activity that ranged between 0.1 and 40 nMol/mg protein/hour (Tanaka et al., 2003;Utsunomiya et al., 2004;Yamamoto et al., 1993;Yamamoto et al., 1990b), whereas one study (Bochkareva et al., 2006b) detected significantly lower STS activity using TLC (in the range of fMol/mg protein/hour; Table 1). One paper could measure enzyme activity but did not report absolute values (Bacallao et al., 2008).
In EC tissues, Utsunomiya et al. detected STS enzyme activity that correlated with mRNA and protein levels in their cohort of 33 welldifferentiated ECs, 26 moderate and 17 poorly differentiated ECs (Utsunomiya et al., 2004; Table 1). These authors did not further compare the enzyme activities between normal endometrium and EC or between the different EC grades.
In 2003, Tanaka and colleagues studied the STS enzyme activity by thin layer chromatography (TLC) in a cohort of 24 EC tissues and normal premenopausal endometrium (n = 10) and found a decreased STS activity level in EC (Tanaka et al., 2003). This cohort consisted of EC specimens of all grades and no distinction was made between histopathological characteristics. A study assessing the STS activity in patients with endometrial hyperplasia associated or not with PCOS described lower STS activity in PCOS specimens but not in hyperplasia (Bacallao et al., 2008).
Four studies published prior to 1990, hence not retrieved by our search, reported increased STS activity in EC compared with normal endometrium (reviewed in: Rizner, 2016).

Concluding remarks on STS
STS mRNA, protein and activity are detectable in endometrium and EC and this enzyme represents an important route of estrogen supply. Clear immunohistochemical signals were obtained by independent investigators and STS localises in endometrial cytoplasm and cell membranes. When STS mRNA or activity levels were compared between ECs and controls, contradicting results were reported. Two studies only quantified STS protein and reported opposite conclusions as well (Table 1). All but one published studies comparing STS activity between ECs and healthy endometrium used premenopausal control tissues, which is suboptimal since EC occurs predominantly after menopause. In one study, women with myomas were used as controls, with no further mention regarding their menopausal status (Yamamoto et al., 1993). When tumour grade/tumour histology was considered (Cornel et al., 2017a;Sinreih et al., 2017a;Sinreih et al. 2017b;Sinreih et al. 2017c;Smuc and Rizner, 2009), the mRNA levels of STS varied with no specific trend or association with patient characteristics and with opposite results between studies (Table 1).

SULT1E1
SULT1E1 catalyses the opposite reaction of STS and converts the intracellular estrone into estrone-S. Thirteen articles were included for our analysis (Table 2).

SULT1E1 protein expression
Utsunomiya and co-workers assessed SULT1E1 immune-reactivity among 76 EC specimens, and found expression in 29% of the cases ( Table 2). No further correlation with other histopathological characteristics was present, and no comparison with controls was performed due to the low number of normal endometrial samples used (n = 6) (Utsunomiya et al., 2004). A later study (Xu et al., 2012) compared the protein expression of SULT1E1 by immunohistochemistry in 30 specimens of paired EC and adjacent normal endometrium with known ERα status (15 ERα positive, and 15 negative; Table 2). SULT1E1 was expressed at detectable levels among the ERα positive EC tissues, but not among the ERα negative cases (Xu et al., 2012).
Sinreih and colleagues explored the SULT1E1 protein expression in EC specimens and paired adjacent normal endometrial tissue by western blotting and immunohistochemistry. Although western blot signal was very low (visible in one sample only out of the 24 tested), immunohistochemistry results on 31 paired specimens showed high subject variability with increased expression in EC versus paired normal tissue in six samples, decreased expression in 16 pairs and no expression in eight pairs. Overall, no trend towards increased/decreased expression in EC or normal tissue was observed (Sinreih et al., 2017a).

SULT1E1 enzyme activity
Four studies determined SULT1E1 activity using radioactive substrates (estrone, 17β-estradiol or PAPS) followed by TLC or aqueous/ organic phase separation and reported values that ranged from 0 to about 300 pMol/mg protein/hour (Table 2). In one study (Tanaka et al., 2003), the enzyme activity of SULT1E1 was assessed by [ 35 S]PAPS incorporation and TLC analysis in EC tissues (n = 24) compared with premenopausal controls (n = 10) and tended to be lower in EC, although this was not significant (Table 2). No additional differences between SULT1E1 activity and tumour characteristics were observed (Tanaka et al., 2003). In an earlier study (Yamamoto et al., 1993), no difference of SULT1E1 activity was detected between ECs (n = 15) and controls (n = 22; Table 2). Utsunomiya and colleagues assessed SULT1E1 enzyme activity in 76 EC specimens and observed a significant positive correlation between activity, immune-reactivity and mRNA levels (Utsunomiya et al., 2004).

Concluding remarks on SULT1E1
SULT1E1 is expressed in endometrium and EC and immune-reactivity is associated with the cytoplasm of endometrial glandular cells, but not with endometrial stromal cells.
Conflicting results exist with respect to the levels of this enzyme in ECs and controls. However, overall most studies on SULT1E1 mRNA expression report decreased or unchanged levels of this enzyme between ECs and controls, whereas all studies assessing SULT1E1 protein expression or activity report unchanged levels ( Table 2). All studies assessing SULT1E1 enzyme activity in EC compared diseased tissue with premenopausal controls, which can be a strong bias since SULT1E1 expression is influenced by the menstrual cycle (Konings et al., 2018a).

CYP19A1
The levels of CYP19A1, involved in the synthesis of estrogens from androgens, attracted significant attention and was explored in several studies because potent CYP19A1-inhibitors are already approved for human use for a long time (estrogen sensitive breast cancer). The role of CYP19A1 was also reviewed in several occasions in a non-systematic way (Bulun et al., 2007;Bulun et al., 2005a;Bulun et al., 2005b;Bulun et al., 1997;Bulun et al., 1999;Ito et al., 2011;Ito et al., 2006;Ito et al., 2007;Jongen et al., 2006;Sasano et al., 2000;Steed and Chu, 2011). For the present systematic analyses, reviews were excluded and 27 original articles were included (Table 3).

CYP19A1 mRNA expression
CYP19A1 mRNA was first detected in EC in 1994, using competitive RT-qPCR (Bulun et al., 1994) and these data were subsequently confirmed by other authors (Sasano et al., 1996;Watanabe et al., 1995). In one study on 19 EC specimens, mRNA detection of CYP19A1 was more prevalent among type II EC (detected in all six samples) than in type I EC (detected in 7 out of 13 specimens, Table 3; Berstein et al., 2005a). The same team revisited the same data later in a study enrolling 80 patients where enzyme activity was also assessed (see below; Berstein et al., 2005b). A subsequent study, however, examined CYP19A1 mRNA in relation to EC grade (Pathirage et al., 2006), and the authors observed a trend towards higher CYP19A1 levels in grade 1 disease compared with grades 2 and 3 EC, but also compared with normal tissue, which tested negative in 70% of the specimens (Table 3).
The mRNA level of CYP19A1 in EC specimens versus adjacent normal endometrium did not vary in a cohort of 25 paired specimens (Smuc and Rizner, 2009;Smuc et al., 2006). Similar data were reported also in an additional study published by the same team . When the same team expanded the analyses to 55 paired specimens, once more no significant difference was found, although a nonsignificant trend for an increased CYP19A1 expression in cancer tissue was identified (Sinreih et al., 2017a). Lack of significant differences between ECs (n = 49) and adjacent endometrial tissues (n = 36) was confirmed by Lepine and co-workers (Lepine et al., 2010). Jarzabek et al. (2013) examined a cohort of 51 EC tissues and 16 premenopausal endometrial tissues. CYP19A1 mRNA levels were extremely low in both cancerous and normal tissue and a non-significant trend towards increased CYP19A1 levels was observed in EC tissues (Table 3). In this study, CYP19A1 was not detected in approximately 10% of the EC specimens versus 50% of the normal endometrial samples (Jarzabek et al., 2013). Two studies from Cornel and co-workers found no association between CYP19A1 mRNA and disease status, histopathological features of the tumour or prognosis of the patients (Cornel et al. 2012(Cornel et al. , 2017b.
Some authors described associations between the CYP19A1 protein level and some histopathological features of patients (Table 3). A trend (non-significant) for an increased CYP19A1 expression in poorly-differentiated compared with well-differentiated tumours was reported by Watanabe and colleagues (Watanabe et al., 1995). The same authors and others (Fowler et al., 2005;Ito et al., 2001) reported no CYP19A1 expression in normal endometrium (Table 3). Subsequently, a study on 51 EC specimens and 17 normal premenopausal endometrial tissues partly confirmed these data, reporting a significantly increased CYP19A1 protein level in EC compared with normal tissue (Jarzabek et al., 2013), and a study on 55 EC patients found that CYP19A1 stromal cell positivity was associated with poor patient prognosis (Segawa et al., 2005).

CYP19A1 enzyme activity
Overall, CYP19A1 enzyme activity in endometrium and EC is low, with most studies measuring activities in the range of fMol/mg protein/ hour (Table 3). Initial studies during early 1990s detected CYP19A1 activity using high performance liquid chromatography (HPLC) in two  Study cohort overlaps with (Sinreih et al. 2013(Sinreih et al. , 2017b(Sinreih et al. , 2017cSmuc and Rizner, 2009;Smuc et al., 2006).
(continued on next page) K.M.C. Cornel et al. Molecular and Cellular Endocrinology 489 (2019) 45-65 EC specimens (Yamamoto et al., 1990a). In a subsequent study on 15 ECs and 22 normal endometrial specimens, the same authors described a higher CYP19A1 activity in EC compared with normal tissue (Yamamoto et al., 1993). This was confirmed by Watanabe and coworkers who found a significantly increased CYP19A1 activity in EC specimens (n = 42, including grades 1 to 3) compared with premenopausal controls (n = 7; see Table 3). The same authors also described a trend towards higher CYP19A1 activity levels in less differentiated tumours (Watanabe et al., 1995), in line with their immunohistochemistry data (Table 3). Three large studies from another team examined the enzyme activity of CYP19A1 in relation to genetic variants in the CYP19A1 gene (Berstein et al., 2004) as well as in relation to tumour type and other characteristics (Berstein et al. 2005a(Berstein et al. , 2005b. A good correlation between CYP19A1 mRNA level and enzyme activity was described in both patient cohorts examined (comprising 19 and 80 EC specimens, although the authors did not state whether these patient populations were overlapping; Table 3) and this correlation was strongest among type I ECs (Berstein et al., 2005a). An increased CYP19A1 activity was observed in type II compared with type I EC (Berstein et al. 2005a(Berstein et al. , 2005b. In two studies comprising 45 and 55 specimens, Bochkareva and coworkers described that CYP19A1 activity increased with increasing invasion in the myometrium and cervix (Bochkareva et al., 2006a;Bochkareva et al., 2006b) although it is unclear from these papers whether the same cohort of patients was used. Another study that explored CYP19A1 activity in 52 EC samples and eight postmenopausal control endometrial tissues did not find any significant difference between EC specimens and controls (Jongen et al., 2005).

Concluding remarks on CYP19A1
Despite the initial interpretation of CYP19A1 being associated with stromal cells only, it is now evident that this enzyme localises in both epithelial and stromal endometrial cells. CYP19A1 is expressed at very low levels in endometrium and EC. This is indicated by the fact that several authors report no mRNA or protein signals, or report positivity in a proportion of the samples only, suggesting that the CYP19A1 level is close to the detection limit of the method employed. CYP19A1 enzyme activity results are in line with these data, with most studies reporting that the activity ranges between 0.1 to around 20 fMol/mg protein/hour (Table 3; one study only reports activities in the range of pMol/mg protein/hour; Watanabe et al., 1995).
With regard to the levels of CYP19A1 in ECs and controls, among all studies assessing CYP19A1 mRNA by RT-sqPCR or RT-qPCR, most publications (5/6, including only the latest paper in those cases when the same patient cohort was explored in consecutive studies) found unchanged CYP19A1 levels (quantified expression or positivity). In contrast, CYP19A1 resulted increased in 4 out of 7 studies assessing the protein expression (quantification or positivity) and in 50% of the four studies assessing the enzyme activity (Table 3).
Most of the studies that explored the CYP19A1 levels in relation to tumour grade or histology described a trend for a higher CYP19A1 expression among poorly differentiated tumours or high stage of disease compared with well differentiated tumours and early stage ECs (Table 3). However, due to inconsistencies between studies and to the general small study populations investigated, these results need independent confirmation in large study cohorts.

HSD17B1
HSD17B1 is involved in the last step in the local activation of estrogens and converts low-active estrone into active 17β-estradiol. Fourteen articles were included for analysis (Table 4).

HSD17B1 mRNA expression
The first studies on HSD17B1 mRNA expression in EC date early 2000, and initially no HSD17B1 mRNA was detected using RT-sqPCR in 20 normal premenopausal endometrial specimens and 46 EC tissues samples (including all grades; Utsunomiya et al., 2001; Table 4).
Later studies detected mRNA signals for HSD17B1, but the overall expression was described as low (up to 1000 fold less than other steroidogenic enzymes; Sinreih et al., 2017a) and inconsistent data were described. In a cohort of 12 EC specimens and adjacent normal tissue,  (Sinreih et al. 2017b(Sinreih et al. , 2017c are considered as one publication, since same data are presented and discussed (one report is 'data in brief', i.e. is the published version of the supplementary data). g n.s: non-significant. h AKR: aldo-keto reductase (see section 6). K.M.C. Cornel et al. Molecular and Cellular Endocrinology 489 (2019) 45-65 the HSD17B1 mRNA level was lower in EC than in healthy endometrium  and the same conclusion was reported in a second study published in the same year by the same team . Once more these authors confirmed these results using the same patient cohort (enlarged to 25 paired specimens; Smuc and Rizner, 2009). However, when the same specimens were reanalysed later in a larger cohort (n = 55 pairs including the same specimens as previously reported; 27 were assessed at the mRNA level) no statistically significant difference in HSD17B1 mRNA levels between ECs and controls was found, although the majority of the EC specimens (15 out of 27) had decreased mRNA levels (Sinreih et al., 2017a). Findings on decreased HSD17B1 mRNA levels in EC tissues were supported by Lepine et al. who found a 2.2 fold lower level comparing 35 EC tissues with the corresponding normal endometrium adjacent to  (Sinreih et al. 2013(Sinreih et al. , 2017a(Sinreih et al. , 2017b(Sinreih et al. , 2017cSmuc and Rizner, 2009 Study cohort overlaps with (Sinreih et al. 2013(Sinreih et al. , 2017a(Sinreih et al. , 2017b(Sinreih et al. , 2017cSmuc et al., 2006). Controls include also premenopausal subjects.  (Sinreih et al. 2013(Sinreih et al. , 2017b(Sinreih et al. , 2017cSmuc and Rizner, 2009;Smuc et al., 2006). Controls include also 17 premenopausal subjects. (Cornel et al., 2017b)  (continued on next page) K.M.C. Cornel et al. Molecular and Cellular Endocrinology 489 (2019) 45-65 EC (Lepine et al., 2010) and by Bacallao et al. who described a lower level of HSD17B1 in endometrial hyperplasia associated with PCOS (n = 10) compared with control endometrium (n = 23; Table 4). However, these authors described the same difference in endometrium from PCOS patients without hyperplasia (n = 18), whereas the seven cases of hyperplasia alone had HSD17B1 levels similar to controls, indicating that the reduced HSD17B1 level is associated with PCOS and not with hyperplastic endometrium (Bacallao et al., 2008). In contrast to these findings, in a cohort of 56 EC tissues and 16 postmenopausal endometrial controls, the HSD17B1 mRNA expression level assessed by RT-qPCR was higher in grade 1 ECs (n = 29) compared with the controls, whereas no differences were found among grade 2 and 3 ECs (n = 18 and n = 11, respectively; Table 4) compared with controls (Cornel et al., 2012). In contrast to these findings, a recent study that explored the expression level of 38 steroidogenic genes in relation to demographic and histopathological characteristics in 51 EC specimens and adjacent normal tissue described an increased level of HSD17B1 mRNA in high grade EC but not in low grade lesions (Sinreih et al. 2017b(Sinreih et al. , 2017c. In a study on 175 EC specimens (including grade 1 to 3; Table 4) analysed by cDNA microarray, high levels of HSD17B1 mRNA correlated both with advanced stage disease and with poor patient prognosis (Cornel et al., 2017b), and the same research team recently confirmed the association between tumour stage and HSD17B1 mRNA level in an independent cohort of 47 EC specimens (Konings et al., 2018b).

HSD17B1 protein expression
All studies assessing HSD17B1 at the protein level reported very low expression, with some authors detecting no immunohistochemical signal (Utsunomiya et al., 2003;Utsunomiya et al., 2001). These data on low protein levels are in line with the first evidences of endometrial HSD17B1 protein expression obtained using radioimmunoassay on proliferative endometrial samples. The HSD17B1 level in endometrium was over 50 times lower than in control placenta tissue (Maentausta et al., 1990). Later, the same authors confirmed HSD17B1 protein expression in EC tissues (Maentausta et al., 1992; Table 4).
Subsequent studies reported weak cytoplasmic HSD17B1 staining but inconsistent results were described (Table 4), i.e., higher levels in grade 1 ECs compared with controls (Cornel et al., 2012), or unchanged levels between EC and adjacent normal tissue (Sinreih et al., 2017a).

HSD17B1 enzyme activity
Although HSD17B1 enzyme activity was described in endometrial preparations during the 1990s (Maentausta et al., 1990) an initial study that examined 12 EC cases using TLC did not detect any HSD17B1 enzyme activity  Table 4).
Cornel and co-workers used a HPLC method coupled with a fluorescence detection system and described an increased HSD17B1 enzyme activity level among grade 1 EC specimens (n = 29) compared with normal postmenopausal controls (n = 16), whereas there were no differences for grade 2 and 3 (n = 18 and n = 11, respectively; Table 4). This result was in line with the mRNA data in the same samples (Cornel et al., 2012). The measured HSD17B1 activity ranged in the fMol-nMol/ mg protein/hour magnitudes. The same team confirmed the presence of HSD17B1 enzyme activity among 47 ECs analysed in a later study (the two study cohorts partly overlapped; Table 4). These authors also showed that a high activity was more prevalent (non-significantly) among high stage disease (in line with the corresponding mRNA results), and that the enzyme could be blocked by a specific HSD17B1 inhibitor, which indicates the engagement of the HSD17B1 (Konings et al., 2018b).

Concluding remarks on HSD17B1
Overall, expression levels of HSD17B1 are low but present, indicating that the final activation of estrogens is relevant in the endometrium and EC. Immunohistochemical signals are detectable in the cytoplasm of epithelial cells.
With respect to HSD17B1 in ECs and controls, the HSD17B1 mRNA is described as unchanged in EC compared with normal endometrium or decreased in those studies comparing EC with adjacent normal endometrium (Table 4). An increased HSD17B1 activity in EC is described by one team only and needs further confirmation.  (Sinreih et al. 2013(Sinreih et al. , 2017aSmuc and Rizner, 2009;Smuc et al., 2006). Association with age. (Konings et al., 2018b) Cases only  (Sinreih et al. 2017b(Sinreih et al. , 2017c are considered as one publication, since same data are presented and discussed (one report is 'data in brief', i.e. is the published version of the supplementary data). g MPA: medroxyprogesterone acetate. h AKR: aldo-keto reductase (see section 6). i CDB: fluorophore 2-(4-carboxy-phenyl)-5,6-dimethylbenzimidazole. K.M.C. Cornel et al. Molecular and Cellular Endocrinology 489 (2019) 45-65 4.6. HSD17B2 HSD17B2, that inactivates 17β-estradiol into its precursor estrone, is detectable in healthy endometrium and associated pathological conditions. Thirteen articles were included for this analysis (Table 5).

HSD17B2 mRNA expression
HSD17B2 mRNA was first detected by RT-sqPCR among EC tissues from a cohort of 46 specimens, but levels were not further studied in relation to histopathological features . One research team studied EC specimens and adjacent normal tissues using the same patient cohort with increasing number of specimens in consecutive publications Sinreih et al., 2013;Sinreih et al., 2017b;Sinreih et al., 2017c;Smuc and Rizner, 2009;Smuc et al., 2006). These authors initially described an increased HSD17B2 expression in tumours compared with controls    (Sinreih et al. 2013(Sinreih et al. , 2017a(Sinreih et al. , 2017b(Sinreih et al. , 2017cSmuc and Rizner, 2009 (Sinreih et al. 2013(Sinreih et al. , 2017a(Sinreih et al. , 2017b(Sinreih et al. , 2017cSmuc et al., 2006). Controls include also premenopausal subjects. (Cornel et al., 2012) Case-control 15 women were premenopausal. Increased level restricted to premenopausal subjects. Study cohort overlaps with (Sinreih et al. 2017a(Sinreih et al. , 2017b(Sinreih et al. , 2017cSmuc and Rizner, 2009;Smuc et al., 2006 (Sinreih et al. 2013(Sinreih et al. , 2017b(Sinreih et al. , 2017cSmuc and Rizner, 2009; (continued on next page) K.M.C. Cornel et al. Molecular andCellular Endocrinology 489 (2019) 45-65 et al., 2006), but these results were not confirmed using a larger cohort of 25 paired specimens (Smuc and Rizner, 2009). Finally, based on a cohort of 47 paired samples, these authors reported that the increased HSD17B2 mRNA was restricted to premenopausal women (Sinreih et al., 2013;Table 5). When the expression level was assessed in relation to demographic and histopathological characteristics in a cohort of 51 EC specimens and adjacent normal endometrium, the HSD17B2 mRNA level was lower in high grade ECs (compared with its adjacent normal counterpart) but not in low grade lesions (Sinreih et al. 2017b(Sinreih et al. , 2017c. In contrast, Lepine et al. found an increased HSD17B2 mRNA expression of 4.4 fold in their 49 EC cases compared with adjacent endometrium (n = 36), which was more pronounced in type II EC (Lepine et al., 2010;Table 5). Similarly, increased HSD17B2 mRNA expression in high grade EC (grade 2 and 3 were pooled) was demonstrated by RT-qPCR in comparison with normal postmenopausal endometrium using a cohort of 19 controls and 58 ECs (of which 18 were grade 2 and 11 grade 3; Cornel et al., 2012). Such increased mRNA level among high grade versus low grade EC was further confirmed by the same authors using an independent cohort of 175 specimens analysed by microarray (Cornel et al., 2017b). In the same microarray study, the HSD17B2 level was also inversely associated with stage and low levels were non-significantly associated with a poorer patient prognosis (Cornel et al., 2017b). In a study on endometrial hyperplasia associated with or without PCOS (Table 5), a decreased HSD17B2 mRNA expression in endometrial tissue from patients with PCOS or hyperplasia alone compared with normal endometrium was observed (Bacallao et al., 2008).

HSD17B2 protein expression
Utsunomiya et al. were the first to describe immunohistochemical expression of HSD17B2 in EC, by exploring 20 normal premenopausal endometrium, 36 endometrial hyperplasia and 46 EC samples (23 grade 1, 14 grade 2 and 9 grade 3). In normal tissues, HSD17B2 expression was only seen in the cytoplasm of glandular cells during the secretory phase of the menstrual cycle and not in the proliferative phase. The highest expression level was seen in hyperplastic endometrium with 75% positive cases (a threshold of > 10% positive cells was considered as positive staining), whereas the percentage of positivity decreased in EC tissues, where 37% were HSD17B2 positive (Table 5). A subsequent study from the same team  reported similar results, but it is not clear from the paper whether the same cohort was used. In a subsequent study, the same research group once more assessed the expression of HSD17B2 in 16 ECs, of which 50% were positive ( Table 5). The authors further correlated HSD17B2 positivity with the response to progestogen treatment (Utsunomiya et al., 2003).
Sinreih and co-workers compared the expression of HSD17B2 in EC and the corresponding normal adjacent endometrium by western blot and immunohistochemistry and found an increased HSD17B2 protein expression in EC (using immunohistochemistry; Sinreih et al., 2017a).

HSD17B2 enzyme activity
Three articles described the HSD17B2 enzyme activity in EC. Cornel and co-workers detected no differences between the enzyme activity in postmenopausal controls and grade 1, 2 and 3 EC specimens using HPLC. Utsunomiya and co-workers reported only the presence of HSD17B2 activity measured by TLC, which correlated with protein and mRNA levels . HSD17B2 activity was lower in the endometrium of women with PCOS (n = 18) compared with the tissue of controls (n = 23) but it did not change in hyperplasia (n = 17; Bacallao et al., 2008). Overall, the HSD17B2 enzyme activity ranged between 0.1 and 1.2 nMol/mg protein/hour (Table 5; one study does not report absolute values; Bacallao et al., 2008).

Concluding remarks on HSD17B2
HSD17B2 is clearly detected in human endometrium at the mRNA, protein and enzyme activity levels and it is regulated by progesterone during the premenopausal period, as reviewed recently (Konings et al., 2018a). Immune signals are associated with the cytoplasm of epithelial cells. Although a number of studies reported lower levels of HSD17B2 in EC compared with normal endometrium especially at the protein level, indicative of a diminished rate of 17β-estradiol deactivation, the majority of the papers exploring the mRNA level of HSD17B2 (4/5, including only the latest paper in those cases when the same patient cohort was explored in consecutive studies) reported increased HSD17B2 expression in ECs compared with controls (Table 5). However, in these reports it is also clear that such feature is restricted to  (Sinreih et al. 2013(Sinreih et al. , 2017aSmuc and Rizner, 2009;Smuc et al., 2006). Association with age and menopausal status.  (Sinreih et al. 2017b(Sinreih et al. , 2017c are considered as one publication, since same data are presented and discussed (one report is 'data in brief', i.e. is the published version of the supplementary data). g MPA: medroxyprogesterone acetate. h AKR: aldo-keto reductase (see section 6). i CDB: fluorophore 2-(4-carboxy-phenyl)-5,6-dimethylbenzimidazole.
specific patient subgroups (premenopausal ECs, poorly-differentiated lesions). Most likely, this result is also related to the increased proliferation of the epithelial component (also when poorly differentiated) present in endometrial carcinoma.

Net sulfatase and 17β-hydroxysteroid balances
Since STS and SULT1E1 as well as HSD17B1 and HSD17B2 are enzymes catalysing opposite reactions in controlling the final availability of active 17β-estradiol, some authors also measured the ratios of STS:SULT1E1 and HSD17B1:HSD17B2 to estimate the net balance in these reactions. Additionally, these enzymes are expressed in endometrial epithelial cells, therefore, this kind of measurement also offers a technical normalisation method, correcting for biases derived from differences in the amount of stromal/epithelial cells or endometrial/myometrial cells present in each specimen.
The STS:SULT1E1 ratio was six fold increased at the mRNA level in ECs compared with normal adjacent tissues (Smuc and Rizner, 2009). When the STS:SULT1E1 activity ratio was compared between endometrium and endometrial hyperplasia in patients with or without PCOS, the activity ratio resulted lower in endometrium of PCOS patients (irrespective to the presence of hyperplasia) than in controls, but unchanged (non-significantly higher) in case of endometrial hyperplasia without PCOS versus controls (Bacallao et al., 2008). The same study also assessed the HSD17B1:HSD17B2 mRNA ratio, which resulted higher in endometrium of PCOS patients compared with controls, but it was unchanged in case of endometrial hyperplasia (irrespective of the presence of PCOS) compared with control endometrium (Bacallao et al., 2008). The increased HSD17B1:HSD17B2 ratio at the mRNA and enzyme activity level was described in a study on 16 controls and 58 ECs. This result was restricted to low grade ECs (n = 29), and it was confirmed in a subgroup analyses on 13 ECs versus adjacent normal endometrium (Cornel et al., 2012). In a large cohort of 175 EC cases analysed by microarray, high HSD17B1:HSD17B2 mRNA ratio correlated with poor patient prognosis, and the ratio was better predictive of patient prognosis compared with the mRNA level of HSD17B1 or HSD17B2 analysed separately (Cornel et al., 2017b).

Conclusive considerations on the intracrine estrogen metabolism in EC
The systematic analysis on the enzymes that ultimately account for the local generation of 17β-estradiol (e.g., STS, SULT1E1, CYP19A1, HSD17B1 and HSD17B2; section 4) as well as the non-systematic analysis on other intracrine enzymes (see section 6), revealed the existence of discordant results, which can be attributed to technical, methodological/study design, but also true biological reasons.

Weaknesses of past research
Technically, past studies present several limitations. Different authors made use of a wide variety of techniques and rarely results from one assay were confirmed using an alternative technique (see in Tables 1-5, few studies assessed mRNA, protein and activity of the same enzyme in the same samples). Gene expression levels were assessed by means of RT-sqPCR (end-point PCR products were quantified by band intensity in agarose/acrylamide gels), RT-qPCR (real-time device), and using different primer-pairs spanning over and amplifying different regions of the cDNA. Not always the same house-keeping genes were used and frequently, one house-keeping gene only was used for normalisation of signals, whereas two or three are recommended (Kozera and Rapacz, 2013). Furthermore, the presence of mRNA variants not all correlating with enzyme and protein activity (as well known for HSD17B1, reviewed by an accompanying paper published in this special issues; Heinosalo et al., 2018) can lead to biases in result interpretation. Protein levels were assessed in most cases by immunohistochemistry, which is semi-quantitative at best. In addition, immunohistochemistry outcome can be variable depending on the protocol, the antigen retrieval method, the experience of the team, the antibody, the fixation used to prepare the material (method and duration), the method used to visualise antigen/antibody interactions (with/without amplification; Rizner et al., 2016). Enzyme activity measurements also show differences based on the method used, radioactive, fluorescent, chromatography based (TLC or HPLC), release of tritiated water, aqueous/organic phase separation. Freezing and processing of the tissues, as well as the buffer used for the lysis, can affect the measurements of the enzyme activity.
In term of methodology/study design, some studies compared pre with postmenopausal samples (see Tables 1-5), which, since many enzymes vary during the menstrual cycle (Konings et al., 2018a), can bias the results. Many studies used cohorts of patients partially overlapping, and not always this is clearly stated in the papers, which does not allow define whether a result is reproduced by technical replication or by confirmation in independent patient cohorts. Some studies compared EC with adjacent tissue with no sign of malignancy. Although this approach is relevant to reduce potential inter-individual variability, it does not take into account the potential premalignant status of such control tissue. Finally, several studies combined EC of different types and histologies, and subgroup analyses were not thoroughly performed or were hampered by the small sample size.
In biological terms, intracrine networks are complex, intricate, show patient to patient variability, redundancy in the reactions catalysed and promiscuity in the preferred substrate. Some enzymes, despite showing biological activities, have levels of expression that are barely detectable (Konings et al., 2018a). In addition, the expression of several enzymes and the availability of substrates vary through the menstrual cycle or during different periods of life. Such heterogeneity makes it difficult to perform meta-analyses of the data.

Intracrine estrogen metabolism: conclusions from past research
Despite all these limitations, past literature underscores few relevant aspects of endometrial intracrinology and indicates that estrogens (and other steroids, see section 6) are intracrinally modified in the endometrium. Provided on the one hand that it is difficult to compare the mRNA, protein and enzyme activity levels of distinct enzymes between studies because PCR/primers and antibodies can have different performances, enzyme assays (always in vitro/cell-free) can have different efficiencies and behave differently from in vivo, and provided on the other hand that low levels of a protein can still be associated with in vivo biological activity and significance (Cornel et al., 2017b;Konings et al., 2018a), the following conclusion can be drawn.
The STS and SULT1E1, combined as the sulfatase pathway, represent a major route of estrogen supply and removal in endometrial cells. The activity of STS is few magnitudes higher than that of SULT1E1 (nMol versus pMol mg protein/hour ranges). This is confirmed at both the mRNA and protein level as well and indicates that the balance of the pathway is shifted towards the formation of free estrone, also suggested by other authors (Rizner, 2016) The aromatase pathway is also active in the endometrium as suggested by several authors (reviewed in: Bulun et al., 2007;Ito et al., 2011). However, since the enzyme activity of CYP19A1 ranges in the fMol/mg protein/hour and mRNA and protein signals are not detectable by all authors (even using different primers and antisera), this pathway seems less relevant than the sulfatase pathway as a route of estrogen supply for the endometrium, whereas it is relevant in other tissues like the bone or the lungs (Konings et al., 2018a). Intracrine CYP19A1 may be relevant for specific subjects with particular high protein levels (overexpression) and, in addition, CYP19A1 may contribute to the levels of circulating estrogens via its activity in the adipose tissue in obese subjects.
The HSD17B1/2 balance is the last step in the estrogen activation/ deactivation and is relevant to endometrial pathophysiology, as reviewed recently (Konings et al., 2018a). Although it resulted negligible in some studies, the presence of HSD17B1 in endometrial tissues is demonstrated not only by mRNA, protein and activity signals, but also by the use of specific HSD17B1 inhibitors able to block the estrogen 17hydroxy reducing activity in endometrial preparations (Delvoux et al., 2014;Konings et al., 2018b). Although the enzyme activity of HSD17B1 is low (in the range of fMol-nMol/mg protein/hour) and is one or two magnitudes lower than the activity of HSD17B2, the redox balance in this reaction is also driven by cofactor availability, and the reductive metabolism is predominant in vivo (Delvoux et al., 2009).
Finally, the inter patient variability and the complexity of intracrinology are factual evidences, and even after excluding the contribution of the aforementioned technical weaknesses, the occurrence of contrasting literature data is likely to represent true findings also in the future. This aspect has important consequences in future clinical scenarios and underscores the need to preselect patients responsive to endocrine drugs prior to start a treatment, as described in section 7.

Additional intracrine networks
The enzymes described in section 4 represent the final and core network for the synthesis of active 17β-estradiol. However, as stated earlier, 14 HSD17Bs exist (with unpublished data referring to a 15th; Konings et al., 2018a) and the role of at least HSD17B7 and HSD17B12 in estrogen metabolism cannot not be excluded. In addition, extra intracrine pathways exist and offer alternative ways to metabolise steroids other than estrogens. In particular progesterone and androgens are attracting significant attention in recent years for their role in endometrial physiology and pathology.
Progesterone has well-known anti-estrogenic actions in the endometrium and PR positivity is a good prognostic marker for EC patients (Tangen et al., 2014). Beside its oxidative action on 17β-estradiol, HSD17B2 has also 20-hydroxysteroid dehydrogenase activity and in combination with HSD17B5 (and other AKRs) controls the level of active progesterone (Konings et al., 2018a). A disturbed progesterone metabolism is postulated to play a role in EC Sinreih et al., 2014;Sinreih et al., 2017c;Smuc and Rizner, 2009).
In case of androgens, although their role is less clear than that of progestogens, they possess anti-proliferative effects in the endometrium and androgen receptor expression is a favourable prognostic marker in EC (Ito et al., 2016;Tangen et al., 2016). In addition, androgen metabolism can have relevant influence on the activation of the estrogen signalling. Primarily because it makes endometrial cells devoid of estrogens via the action of the enzymes HSD17B5 and 5α-reductase. It is postulated that the formation of dihydrotestosterone through the conversion of androstenedione to testosterone by HSD17B5, followed by further 5α-reductase activity, would diminish the availability of testosterone and, in turn, its aromatisation to 17β-estradiol in EC. Tanaka and co-workers found a correlation between the tissue level of dihydrotestosterone and the expression of 5α-reductase that was also associated with a favourable prognosis in patients (Tanaka et al., 2015). Secondarily, steroids in the androgen metabolism like A-diols (comprising androstenediol and the 5α-reduced forms 3α-and 3β-diols) are weak ER binders and can therefore activate the estrogen signalling. Adiols are generated from DHEA or dihydrotestosterone via the action of various HSD17Bs and AKRs, as recently thoroughly reviewed (Konings et al., 2018a;Moeller and Adamski, 2009;Prehn et al., 2009). Finally, AKRs (that also include HSD17B5) are enzymes that posses various 3, 17, 20 steroid keto-reductase/hydroxyl-dehydrogenase activities and can use various substrates like androgens, progesterone (and other 21carbon derivatives), thus making the intracrine network extremely flexible (Konings et al., 2018a;Moeller and Adamski, 2009;Prehn et al., 2009).
The expression of most of the enzymes involved in these metabolisms (HSD17Bs other than HSD17B1 and 2, and AKRs) was explored in EC. Table 6 gives a non-systematic overview of the outcomes of these studies that also shows conflicting data (like for HSD17B5, 7 and 12). One study only explored recently the level of HSD17B14 that has putative steroid oxidative catalytic activity and may use estrogens as substrate. HSD17B14 was reported down regulated in EC compared with adjacent normal tissue (Sinreih et al. 2017a(Sinreih et al. , 2017b. Also HSD17B6 has steroid oxidative catalytic activity, with important role in androgen metabolism (although likely not on estrogens), but this enzyme was never studied in EC.

Intracrine drug targets for patient care, a future opportunity
As overviewed in paragraph 2.4, the use of hormonal treatments for EC is associated with low response rates. However, it is generally accepted that there is room for improvements, starting with the preselection of responsive patients (Derbyshire et al., 2017;Gao et al., 2014). Rendering cells devoid of local estrogens by targeting intracrine enzymes can be a relevant future opportunity. Inhibition of CYP19A1 (AIs) has been tested in various trials (see paragraph 2.4), but since CYP19A1 is highly expressed in adipose tissue and may affect also the systemic estrogen levels, it is not possible to discriminate between systemic and local (intracrine) effects of AIs.
In contrast, inhibition of STS directly decreases the intracellular estrogen level. Potent inhibitors of STS were tested in phase I/II clinical trials for some indications like endometriosis (Konings et al., 2018a; Table 6 Main studies reporting the expression of HSD17B4,5,7,8,12,and 14 in EC (non-systematic search).

Conclusions
Technical issues and sub-optimal study-design account in part for the discrepancies in the results reported by published studies. Also, the fact that only recently the serum and tissue steroid levels can be assessed accurately by LC-MS has hampered to draw robust conclusions so far. A world-wide effort between professionals to harmonise protocols and to collect multi-centred large and stratified patient cohorts aimed at studying tissue intracrinology is highly needed.
True biological differences and patient to patient variability account however for another significant part of the contradictions described in various studies. In combination with the complexity of the intracrine networks, this implies that the clinical exploitation of this knowledge to identify drug targets for future patient care is possible but needs a preselection of responsive patient for each drug (personalised-medicine). To this end, we also need to learn how to preselect responsive patients. Cancer classification, even the most advanced TCGA classification system, helps only marginally in this (only patients with low receptor levels and with likely low chances to respond to hormonal drugs can be identified; Kandoth et al., 2013). Therefore, retrospective and prospective studies to identify tumour and serum markers and tumour features that correlate with clinically measured response to endocrine drugs in EC patients are needed.

Declaration of interest
The authors have no conflict of interest that could be perceived as prejudicing the impartiality of the research reported.