br and OH P A DHEA and DHEAS were negatively
and 17OH-P5, A4, DHEA and DHEAS were negatively correlated with BMI (data not shown).
3.2. Increased E2 plasma levels are associated with a high percentage of vis-ceral fat
The majority of circulating estrogens in postmenopausal women is produced from androgens via the enzyme CYP19A1 in the adipose tis-sue, which also influences the levels of other steroids like androgens and glucocorticoids. Since visceral/abdominal fat depots have different metabolic activities, and BMI alone disregards such information, we ex-plored the relationship between steroid levels in plasma and fat distri-bution. Of the 38 patients with data on steroid plasma concentration, 20 patients had available preoperative CT scans. For these patients CT-assessed obesity variables including subcutaneous abdominal fat vol-ume (SAV), visceral abdominal fat volume (VAV), total abdominal fat volume (TAV) and VAV% ([VAV/TAV]*100) were measured. There was a significant (moderate) positive correlation between visceral fat per-centage and E2-levels, both within the whole patient population (ρ = 0.47, p = 0.035, Table 3), and within the non-endometrioid subgroup (ρ = 0.59, p = 0.043, Table 3), also the visceral fat volume tended to the same within the whole patient population (ρ = 0.42, p = 0.068, Table 3). Neither BMI, TAV, SAV nor other steroids (including E1 and an-drogens) were found to be significantly correlated to the E2-levels.
3.3. High E2 concentration in plasma is associated with increased expres-sion of genes involved in Cyclosporin H related signaling in tumor
We performed whole exome RNA sequencing on tumor specimens from 38 patients of whom 33 overlapped with patients with available steroid profile. A number of genes had significantly different expression between patients with long and short survival (Supplementary Table S2). Expression of several of the top ranked genes was also vali-dated to have prognostic impact in a larger cohort including 256 EC pa-tients , and could potentially be useful prognostic biomarkers in EC (Supplementary Table S2). Also genes involved in local steroid metabo-lism were found to be associated with patient prognosis in the whole patient population, and depending on histologic subtype, these were however not among the top ranked genes (Supplementary Table S3).
We then explored whether circulating steroid levels were associated with specific RNA expression patterns indicating hormone signaling ac-tivation. To this end, genes with differential expression between pa-tients with the highest plasma level (upper quartile) of a specific steroid and patients with lower levels (three lower quartiles) were identified. When investigating genes overexpressed in the patient group with high E2 levels in a Gene Set Enrichment Analysis (GSEA), gene sets associated with estrogen related signaling were found to be significantly enriched in patients with high estradiol levels compared to low (Table 4, Supplementary Table S4). The levels of other steroids were not associated with specific gene signatures. r> 3.4. Local metabolism of steroid hormones
Steroids present in the blood, most of which have little activity, can be locally converted into active steroids through a complex enzymatic network, as illustrated in Fig. 2A. Cluster analysis of circulating steroids revealed that in patients with long survival, precursors or intermediates in the steroid metabolic pathway (like the progestogens P4, P5 and hy-droxylated forms and adrenal steroids DHEA and DHEAS) had higher levels than in patients with short survival (Fig. 2B). To examine to what extent EC tissues possess the enzymatic machinery to metabolize these blood steroids, we extracted the expression levels of all genes in-volved in the local steroid metabolism from the RNA seq data (Supple-mentary Fig. S2). STAR and CYP11A1 were detectable, although at very low level, suggesting that cholesterol may be converted to P5. En-zymes involved in the subsequent conversion of P5 to androgens
Fig. 1. Increased level of steroids in EC patients with long survival. The levels of DHEA (A), DHEAS (B), E1S (C), progesterone (E) and 21 OH progesterone (F) were increased in patients with long survival compared to short (all p-values b 0.05). For estradiol no significant difference in plasma level was observed between patients with long and short survival (D).
(i.e., CYP17A1, HSD3B2) were also detectable, at low levels, indicating that the substrates for these enzymes (P5, 17OH-P5 and DHEA), all highly abundant in the blood (Fig. 2A, left panel), can be converted into A4 in the tumor tissue. Further, the enzymatic machinery to metab-olize A4 into other androgens (T, DHT and AN) as well as estrogens is present at relatively high levels (HSD17Bs, AKR1Cs and SRD5As). P4 and 17OH-P4 could serve as substrates for local corticosteroid genera-tion, but the enzymes responsible for this pathway (i.e., CYP21A1 and CYP11Bs) are those with the lowest expression levels (undetectable for CYP11Bs); hence, local generation of corticosteroids in EC from pro-gestogens seems negligible. Nevertheless, cortisone/cortisol balance (both present at high circulating levels, in the nano molar range) can be regulated by the highly expressed HSD11Bs. No expression was de-tected for CY11B1 and 2, HSD17B8 and HSD3B1. Of interest, SRD5As and AKR1Cs, which are widely involved in the metabolism of both an-drogens and progestogens with AKR1Cs also having ample substrate specificity and catalytic activity, were highly expressed in EC. AKR1Cs show border line correlation with patient survival (p = 0.05; Supple-mentary Fig. S3).