Elevated periimplantation uterine natural killer cell density in human endometrium is associated with impaired corticosteroid signaling in decidualizing stromal cells.

We routinely assess uNK cell densities by CD56 immunostaining of timed (d LH+7 to LH+10) endometrial biopsies from women suffering reproductive failure. Based on previous studies, normal uNK cell density is defined as 5% or less CD56+ cells in the stroma underlying the luminal epithelium ( 17 , 23 ). We speculated that an apparent excess of uNK cells in the periimplantation endometrium may reflect impaired 11βHSD1 expression and relative cortisol deficiency. To test this hypothesis, a TMA was constructed using biopsies with normal (n = 18) as well as elevated (n = 18) uNK cell levels. The TMA was stained with anti-CD56 and anti-11βHSD1 antibodies. As shown in Figure 1 A, a strong inverse correlation was observed between uNK cell density and 11βHSD1 immunoreactivity. This negative correlation was confirmed by semiquantitative image analysis ( Figure 1 B). Inverse correlation between uNK cell density in the subluminal epithelium and expression of 11βHSD1. A, Representative CD56 and 11βHSD1 immunostaining in midluteal endometrial biopsies deemed to have normal or elevated uNK cell count (≤5% or >5% CD56 + cells in the stroma, respectively). Original magnification, ×20. B, Relative 11βHSD1 immunostaining after image analysis of a TMA containing biopsies with normal (≤5%; n = 18) or increased (>5%, n = 18) uNK cell density in the subluminal stroma. ***, P < .001. Previous studies have shown that an aberrant decidual response, associated with reproductive disorders such as endometriosis and RPL, is maintained in culture ( 24 , 25 ). This prompted us to investigate whether uNK cell density in vivo could reflect the induction of 11βHSD1 in response to decidual cues. To this end, we divided timed endometrial biopsies and processed one part for CD56 immunostaining and the other part for primary HESCs. These cultures were passaged once, grown to confluency, and treated with 8-bromo-cAMP, P4, and E for either 4 or 8 days. As shown in Figure 2 , there was a striking inverse correlation between uNK cell density in vivo and the responsiveness of paired primary cultures to differentiation stimuli (Supplemental Table 1). This inverse correlation was apparent for the induction of not only HSD11B1 , the gene that encodes 11βHSD1 ( Figure 2 A) but also for PRL and IGFBP1 , two classical decidual marker genes ( Figure 2 , B and C). Inverse correlation between uNK cell densities in vivo and the induction of decidual markers in vitro. A, The uNK cell densities in midluteal biopsies correlated inversely to the induction of 11 β HSD1 transcripts in primary HESCs decidualized for 4 days (left panel) or 8 days (middle panel). Note the logarithmic Y-axis. The right panel shows the mean (±SEM) induction of 11 β HSD1 transcripts in biopsies deemed to have normal or elevated uNK cell counts. B, The uNK cell densities in midluteal biopsies correlated inversely to the induction of PRL transcripts in primary HESCs decidualized for 4 days (left panel) or 8 days (middle panel). The right panel shows the mean (±SEM) induction of PRL transcripts in biopsies deemed to have normal or elevated uNK cell counts. Note the logarithmic Y-axis. C, The uNK cell densities in midluteal biopsies correlated inversely to the induction of IGFBP1 transcripts in primary HESCs decidualized for 4 days (left panel) or 8 days (middle panel). The right panel shows the mean (±SEM) induction of PRL transcripts in biopsies deemed to have normal or elevated uNK cell counts. Note the logarithmic Y-axis. *, P < .05; **, P < .01. We also examined whether this association extended to IL-11 and IL-15, cytokines implicated in regulating uNK cells ( 11 – 13 ). Rather surprisingly, there was a significant trend toward higher levels of induction IL-15, but not IL-11, transcripts in decidualizing HESC cultures obtained from biopsies with normal uNK cell densities ( Figure 3 , A and B). Induction of IL11 and IL15 mRNA in decidualizing HESCs in culture and uNK cell densities in vivo. A and B, The uNK cell densities in midluteal biopsies were correlated to the induction of IL11 and IL15 transcripts, respectively, in corresponding primary HESCs decidualized for either 4 days (left panel) or 8 days (middle panel). The right panel shows the mean (±SEM) induction in biopsies deemed to have normal or elevated uNK cell counts. Next, we used the TMA to examine the expression of the cortisol-responsive receptors, GR and MR. As reported by others ( 26 ), GR and MR are both expressed in the endometrial stroma ( Figure 4 A). Semiquantitative analysis of the TMA showed no difference in stromal GR immunoreactivity in biopsies characterized by elevated vs normal uNK cell levels. In contrast, an increased percentage of CD56 + cell density was associated with significantly lower MR expression levels ( P < .05; Figure 4 B). Increased uNK cell levels are associated with impaired MR expression. A and B, The expression of GR and MR, respectively, in the subluminal stroma was assessed by immunostaining of a TMA containing biopsies with normal (≤5%; n = 18) or increased (>5%, n = 18) uNK cell density in the subluminal stroma. The left panel shows representative immunostaining (original magnification, ×40), whereas the right panel depicts semiquantitative image analysis of the immune staining. ***, P < .001. Previous knockdown experiments in decidualizing HESCs have shown that the 11βHSD1/MR axis regulates the expression of several enzymes involved in lipid metabolism and retinoid acid biosynthesis and storage ( 18 ). We measured the expression levels of two target genes, DHRS3 and RETSAT , to monitor the MR activity in decidualizing primary HESC cultures established from biopsies with normal or elevated uNK cell levels. Both genes were moderately induced upon treatment with 8-bromo-cAMP/P4/E in a time-dependent manner when the primary HESC cultures were established from samples with 5% or less CD56+ cells in the subepithelial stroma ( Figure 5 ). By contrast, this induction was significantly impaired in cultures established from high uNK cell samples. Elevated uNK cell density is associated with blunted expression of the MR-dependent genes in decidualizing HESCs. A, The uNK cell densities in 21 biopsies correlated inversely to the induction of DHRS3 transcripts in primary HESCs decidualized for 4 days (left panel) or 8 days (middle panel). The right panel shows the mean (±SEM) induction of DHRS3 transcripts in biopsies deemed to have normal or elevated uNK cell counts. B, The uNK cell densities in 21 biopsies correlated inversely to the induction of RETSAT transcripts in primary HESCs decidualized for 4 days (left panel) or 8 days (middle panel). The right panel shows the mean (±SEM) induction of RETSAT transcripts in biopsies deemed to have normal or elevated uNK cell counts. *, P < .05; **, P < .01. To validate these findings, we performed Western blot analysis of primary HESCs decidualized with 8-bromo-cAMP/P4/E for 4 days. As shown in Figure 6 A, 11βHSD1 and DHRS3 were abundantly expressed in primary cultures decidualized for 4 days. Induction of RETSAT, however, requires prolonged decidualization. This enzyme was barely detectable after 4 days of differentiation and only in primary cultures established from samples with 5% or less CD56+ cells in the subepithelial stroma. Semiquantitative analysis of the blots showed that elevated uNK cell density in vivo is associated with significantly lower 11βHSD1 expression in corresponding decidualizing primary HESC cultures and a trend toward lower DHRS3 levels ( P = .03 and P = .07, respectively; Figure 6 B). A, Composite figure showing 11βHSD1, DHRS3, and RETSAT protein expression in primary HESC cultures decidualized for 4 days. A total of eight primary cultures were established from biopsies with normal or elevated uNK cell densities. β-Actin served as a loading control. B, Semiquantitative analysis of 11βHSD1 and DHRS3 expression relative to β-actin. Because of the low level of expression, RETSAT expression was not quantified. *, P < .05.

Increased uNK cell density in midluteal endometrium has been associated with reproductive failure, especially RPL ( 4 – 6 ) However, the mechanisms that account for cyclic recruitment of uNK cell precursors and subsequent proliferation and differentiation within the periimplantation environment are not well understood. In addition to IL-11 and IL-15, several other endometrial factors may be implicated in this process, including chemokine motif ligand 14, IL-12, and IL-33 ( 25 , 27 , 28 ). Different strands of evidence suggest that induction of a cortisol gradient upon decidualization of HESCs is also a key regulator of uNK cells in periimplantation endometrium. First, preconceptual prednisolone treatment markedly reduces uNK cell density in RPL patients ( 17 ). uNK cells express GR but not progesterone receptors ( 15 ), the inference being that glucocorticoids are likely to act directly on these cells. Second, we found a strong negative correlation between uNK cell densities and expression of 11βHSD1 in differentiating stromal cells in vivo. We have shown previously that 11βHSD1 expression and enzyme activity in decidualizing HESCs is driven by cAMP and P4 signaling ( 18 ). Furthermore, inhibition of 11βHSD1 activity with either carbenoxolone disodium salt or PF 915275 virtually abolishes the induction of HSD11B1 , indicating that local cortisol signaling reinforces the expression of this enzyme in decidualizing cells through an autocrine mechanism ( 18 ). By contrast, the expression of the type 2 isoform (11βHSD2), the dehydrogenase that converts cortisol into inactive cortisone, is low in both undifferentiated and decidualizing HESCs ( 18 ). Notably, the decidual process in the human endometrium is under tight spatiotemporal control ( 29 ). It is initiated in the midluteal phase of the cycle first in stromal cells surrounding the terminal spiral arteries and underlying the luminal epithelium. Thus, rather than the total tissue concentration of uNK cells, it is possible that excessive migration of uNK cells from their usual position in the basal and perivascular regions of the endometrium to the subluminal region is the hallmark of an abnormal decidual response that predisposes for early pregnancy loss. Finally, there is increasing evidence that the responsiveness of endometrial cells to differentiation signals is subject to epigenetic programming ( 30 ), which explains how an aberrant decidual response in vivo is maintained, at least partly, upon differentiation of purified HESCs in vitro ( 24 , 25 , 31 ). In agreement, we found that high uNK cell density in vivo is associated with blunted induction of HSD11B1 as well as decidual marker genes, such as PRL and IGFBP1 , in primary cultures. This strong inverse correlation suggests that the 5% threshold of uNK cell density is somewhat arbitrary. Whether increased uNK cells densities correlate to increased risk of pregnancy failure warrants further investigation. In addition to decidual marker genes, induction of IL-11 and IL-15 also tended to be lower in decidualizing cells established from biopsies with elevated uNK cell levels. This observation does not exclude that expression levels of these cytokines in situ correlate with uNK cell levels as reported for IL-15 in a recent study ( 32 ). Human uNK cells have been described as immature and inactive before pregnancy ( 1 ). How elevated levels of uNK cells prior to conception predispose to subsequent pregnancy failure is unclear. Ablation of these cells in mice has been shown to compromise spiral arteriole remodeling and maintenance of decidual integrity seen after midpregnancy ( 3 , 33 ). Yet uNK cell-deficient IL-15 −/− mice are fertile and have normal gestation lengths and litter sizes comparable with wild-type mice ( 34 ). Similarly, human uNK cells are implicated in spiral artery remodeling. They are a rich source of angiogenic growth factors, although paradoxically the endometrium of RPL patients is characterized by reduced expression of several key factors, including platelet-derived growth factor-BB, angiotensin-2, vascular endothelial growth factor-A, and vascular endothelial growth factor-C ( 35 ). Our data suggest that increased density of CD56 + cells in the subluminal endometrial stromal compartment may be an indirect marker of local corticosteroid deficiency. Furthermore, our data indicate that the 11βHSD1/MR axis in target cells may be particularly affected as exemplified by the impaired induction of DHRS3 and RETSAT transcripts in decidualizing cultures established from high uNK cell biopsies. These enzymes are involved in lipid metabolism and retinoid acid (RA) biosynthesis and storage. Both shortage and excess of RA contribute to fetal malformation, suggesting that retinoid metabolism must be regulated closely at the fetomaternal interface ( 36 ). Alcohol dehydrogenase and nicotinamide adenine dinucleotide phosphate oxidase-dependent short-chain dehydrogenases/reductases, including DHRS3, oxidize retinal to retinol and promote its storage as retinyl esters. Likewise, RETSAT is involved in the regulation of retinoid storage as lipid droplets ( 37 ). Interestingly, RA inhibits decidualization of HESCs, and excess levels of RA or retinal are cytotoxic ( 38 ). Thus, high uNK cell density in the periimplantation endometrium may be associated with perturbations in the retinoid metabolism pathway in the stromal compartment, which in turn predisposes for an impaired decidual response and compromise histiotrophic support of the early conceptus. Thus, it seems likely that complex and dynamic gradients of chemoattractants and chemorepellents control the spatiotemporal distribution of uNK cells in the periimplantation endometrium. In addition to glands and other immune cells, decidualizing stromal cells play a major role in governing this process, at least partially, by inducing a cortisol gradient that establishes a nutritive environment essential for postimplantation embryo development and fetal growth. Our data suggest that excessive uNK cells in the subluminal stromal compartment prior to conception may serve as a potential biomarker for a suboptimal decidual response in pregnancy. A recent pilot randomized, double-blind controlled clinical trial suggested an improvement in live birth rate with prednisolone in women with RPL and high midluteal uNK cell density ( 23 ), although this finding needs validating in a larger trial. In addition, assessment of uNK cell density varies greatly from laboratory to laboratory and intercycle variation has been reported ( 22 ). To be clinically useful, international standardization of uNK cell assessment is urgently needed.

The study was approved by the local ethics committee (1997/5065). Subjects were recruited in the Implantation Clinic, a dedicated research clinic at University Hospitals Coventry and Warwickshire National Health Service Trust for patients suffering RPL or recurrent in vitro fertilization treatment failure. Written informed consent was obtained prior to tissue collection. Endometrial biopsies were timed between 7 and 10 days after the preovulatory LH surge. Samples were obtained using a Wallach Endocell sampler (Wallach) under ultrasound guidance, starting from the uterine fundus and moving downward to the internal cervical ostium. Each biopsy was divided, with one part fixed in formalin for immunohistochemistry and the other processed for primary cell culture. The demographic details of participating subjects are summarized in Supplemental Tables 1 and 2, published on The Endocrine Society's Journals Online web site at http://jcem.endojournals.org . HESCs were isolated, cultured, and maintained as described ( 20 ). Primary cultures were passaged once, allowed to grow to confluency, and then decidualized with 0.5 mM 8-bromoadenosine cAMP (8-bromo-cAMP; Sigma), 1 μM progesterone (P4; Sigma), and 0.1 μM cortisone (E; Sigma). Cortisone, which is inactive, was added to decidualizing HESC cultures as the substrate for endogenous conversion by 11βHSD11 to cortisol ( 18 ). Five-micrometer-thick formalin-fixed, paraffin-embedded tissue sections were labeled with antibody to CD56 (NCL-CD56-1B6; Novacastra) using standard methods and detection systems ( 3 ). The uNK cell density was determined as the percentage of uNK cells within the stromal cell population. Because uNK cell density varies with endometrial depth, counting of CD56+ cells was confined to the stroma underlying the luminal epithelium. Five randomly selected high-power magnification fields per biopsy were assessed using ImageJ software (Rasband, W. S., ImageJ, National Institutes of Health) to minimize interobserver variability ( 21 , 22 ). Normal uNK cell density was defined as 5% or less CD56 + cells in the stroma underlying the luminal epithelium ( 17 , 23 ). A Mirax Midi slide scanner was used to scan bright-field sections with a ×20 objective with a resolution of 0.23 μm/pixel. This produces images that can be dynamically manipulated within the viewer software, allowing optical magnifications up to ×20 and digital magnification to ×200. Areas of interest, ie, subepithelial regions, were spotted and tissue microarrays comprising duplicate 0.6-mm cores from 18 cases in each group were constructed using Alphelys TMA Designer R 2 version 1.0.0.8. Sections (3 μm) were cut from completed array blocks and transferred to silanized glass slides. Sections from these arrays were then stained for CD56, 1:200 (NCL-CD56-1B6; Novocastra); 11βHSD1, 1:300 (AB83522; Abcam); MR, 1:400 (H-300: SC-11412; Santa Cruz Biotechnology); and GR, 1:200 (E-20: SC-1003; Santa Cruz Biotechnology). Semiquantative analysis was performed using a Panoramic viewer to capture images. High-power (×400) images were analyzed with Image J ( http://rsbweb.nih.gov/ij/ ) using the color deconvolution plugin and thresholding to assess the percentage of strongly immunopositive endometrial cells. The observers were blind as to the origin of the samples. Total RNA was extracted with RNA STAT-60 from primary HESC cultures. After treatment with amplification-grade deoxyribonuclease I (Invitrogen Ltd), cDNA was generated using the SuperScript II first-strand synthesis system for RT-PCR kit (Invitrogen). Template quantification was performed with an ABI Step One system (Applied Biosystems) using Power SYBR Green PCR master mix (Applied Biosystems). RNA input variances were normalized against the levels of the L19 housekeeping gene, which encodes a ribosomal protein. All measurements were performed in duplicate. Specific primer pairs were designed using Primer3 software ( http://frodo.wi.mit.edu ): L19 sense, 5′-GCG GAA GGG TAC AGC CAA T-3′, L19-R antisense, 5′-GCA GCC GGC GCA AA-3′; 11βHSD1 sense, 5′-AGC AAG TTT GCT TTG GAT GG-3′, 11βHSD1 antisense, 5′-AGA GCT CCC CCT TTG ATG AT-3′; decidual prolactin (PRL) sense, 5′-AAG CTG TAG AGA TTG AGG AGC AAA C-3′, decidual PRL antisense, 5′-TCA GGA TGA ACC TGG CTG ACT A-3′; IGF-binding protein-1 (IGFBP1) sense, 5′-CGA AGG CTC TCC ATG TCA CCA-3′, IGFBP1 antisense, 5′-TGT CTC CTG TGC CTT GGC TAA AC-3′; IL-11 sense, 5′-CTC GAG TTT CCC CAG ACC CTC GG-3′, IL-11 antisense, 5′-TGT CAG CAC ACC TGG GAG CTG TAG-3′; IL-15 sense, 5′-TGG CTG CTG GAA ACC CCT TGC-3′, IL-15 antisense, 5′-CCC TGC ACT GAA ACA GCC CAA AA-3′; DHRS3 sense, 5′-AGC GCG GCG CCA GAA AGA TT-3′, DHRS3 antisense, 5′-TCA CCC ACC TTC TCC CGG ACG-3′; and RETSAT sense, 5′-CGC TGC CTG CCA GGT GTG AAG-3′, RETSAT antisense, 5′-AGA CGT AGC GCT CCA TCG CC-3′. Whole-cell protein extracts were obtained by direct lysis in Laemmli buffer heated to 100°C. Proteins resolved by SDS-PAGE were transferred to a polyvinyl difluoride membrane (GE Healthcare) and probed with antibodies raised against 11βHSD1, 1:1000 (AB83522; Abcam); DHRS3, 1:1000 (15393-1-AP; ProteinTech Group); RETSAT, 1:1000 (SAB1407586; Sigma); and β-actin, 1:100 000 (A1978; Sigma). After incubation with horseradish peroxidase-conjugated secondary antibodies diluted 1:2000 (DAKO), immunoreactivity was visualized using the ECL + chemoluminescent detection kit (Amersham). Data were analyzed with the statistical package GraphPad Prism (GraphPad Software Inc). A Student's t test and a Mann-Whitney U test were used when appropriate. Logarithmic transformations were used when data were not normally distributed. Pearson's correlation coefficient ( r ) was used to assess the correlation between uNK cell densities in vivo and the induction of various genes upon decidualization of corresponding primary HESC cultures. Statistical significance was assumed when P < .05.