Assessment of Requirements for IL-15 and IFN Regulatory Factors in Uterine NK Cell Differentiation and Function During Pregnancy ¹

Uterine NK (uNK) ³ cells, an NK cell subset, are the most frequent lymphocytes found within implantation sites during the first half of normal pregnancy in rodents, women, and nonhuman primates (1, 2). In mice, terminally differentiating uNK cells are first recognized at gestation day (gd) 5 (1, 3) in the last region of each implant site to show transformation of fibroblasts to decidual cells, the decidua basalis (DB). uNK cells rapidly increase in number by proliferation in DB, and then in a myometrial region called the mesometrial triangle. By gd 10 of the 19- to 20-day pregnancy, uNK cells form transient, pregnancy-associated mural structures at each implantation site called mesometrial lymphoid aggregates of pregnancy (MLAps) or metrial glands. The MLAps surround the branches of maternal uterine arteries and veins as they enter and leave implantation sites (4). uNK cells are the major but not sole sources of IFN-γ on the mesometrial aspect of each implantation site (5).

The absence of uNK cells in tgε26 (NK⁻, T⁻, B⁺), IL-2Rβ^−/− (NK⁻), and recombination-activating gene-2 (RAG-2)^−/−γ_c^−/− (NK⁻, T⁻, B⁻) mice, or absence of IFN-γ signaling in IFN-γ^−/−, IFN-γRα^−/−, and Stat-1^−/− mice during pregnancy is associated with structural stability rather than modification of the branches of the uterine arteries (called decidual spiral arteries (SAs)) that supply the conceptuses. Normal, pregnancy-induced reduction of the vascular smooth muscle coat, and arterial dilation and elongation are restricted (4, 5, 6, 7, 8). In addition to triggering vascular changes, uNK cell-derived IFN-γ promotes the terminal steps in uterine stromal cell transformation into decidua and controls uNK cell maturation (judged by diameter and cytoplasmic granularity) and senescence (7). Mechanisms by which uNK cell-derived IFN-γ exerts these actions are not defined.

In lymphoid organs, IL-15 plays a key role in the maturation of NK progenitor cells (9, 10, 11, 12). The mouse uterus initiates transcription of IL-15 following the onset of decidualization until gd 11, when it is lost (13). IL-15 has been demonstrated in the human uterus throughout the menstrual cycle. Human uterine IL-15 message and protein are elevated in the secretory phase of the menstrual cycle, when decidualization begins (2) and in early pregnancy when decidual tissue is greatly increased (14, 15, 16, 17). Human uterine IL-15 expression is most prominent in perivascular cells surrounding the decidual SA in the late cycle and in the endothelial cells of these arteries during early pregnancy (15). In lymphoid tissue, the transcription factors IFN-regulatory factor (IRF)-1 and IRF-2 regulate NK cell differentiation. IRF-1 binds to promoter sequences in the IL-15 gene, whereas IRF-2 appears to regulate the IL-15R. IRF-1 and IRF-2 are key regulators of IL-12-induced IFN-γ production by NK cells (18, 19, 20, 21). However, uterine IRF-1 is only abundantly expressed in lumenal and glandular epithelium and not uterine stroma in humans (22). This is the same site to which IRF-2 has been localized in the sheep uterus, the only species in which uterine IRF-2 has been investigated (23). Thus, there is circumstantial, but no direct evidence that IL-15 is important for uNK cell differentiation during pregnancy. This is important to establish, because unsuccessful clinical outcomes in human pregnancies are being attributed to imbalance in IL-15 (17). Mice lacking IL-15 (10), IRF-1, and IRF-2 (24) are reproductively competent but have profound reductions in numbers of peripheral NK cells. Implantation sites have not been assessed in these strains to characterize uNK cell differentiation or quantify SA modification. The present study was undertaken to define the role of IL-15 in supporting the differentiation of uNK cells within mouse implantation sites and to address the requirements for IRF-1 and IRF-2 in uNK cell differentiation.

C57BL/6J (B6) and IRF-1^−/− mice were purchased from The Jackson Laboratory (Bar Harbor, ME). Breeding pairs of IRF-2^−/− and RAG-2^−/−γ_c ^−/− were kindly provided by Drs. T. Mak (University of Toronto, Toronto, Ontario, Canada) and J. Di Santo (Pasteur Institute, Paris, France). Breeding pairs of IL-15^−/− mice were provided by Immunex (Amgen, Thousand Oaks, CA). All strains of knockout mice used in this study were of B6 background. B6 mice were housed under conventional husbandry. All other strains were bred under monitored, barrier husbandry in the University of Guelph (Ontario Ministry of Agriculture, Food, and Rural Affairs Isolation Unit (Guelph, Ontario, Canada)). Females were syngeneically mated. At specific gestation day, counting the copulation plug as day 0, timed-pregnant mice were euthanized by CO₂ inhalation, followed by cervical dislocation. Uterine horns were dissected and examined for fetal viability (implant size and color). Only viable implantation sites were chosen for further study. Some of the pregnant B6 and IRF-1^−/− mice were permitted to give birth, and their pups were weighed at birth (<12 h), and 21 and 49 days of age.

Implantation sites were fixed in Bouin’s fixative (Fisher Scientific, Whitby, Ontario, Canada) overnight and in 70% ethanol until processed. All fixed tissues were processed into paraffin using standard methodology. Two or three paraffin-embedded implantation sites from each pregnant mouse (four to five pregnant females per study group) were serially sectioned (7 μm) and stained with H&E or periodic acid-Schiff (PAS), which stains the glycoprotein in uNK cell granules. All stained slides were examined by light microscopy. The center section of each serially sectioned implantation site was identified, and for each implantation site, five median tissue sections on each side of the central section were also analyzed (n = 11). Sections were at least 42 μm apart to avoid duplicate counting of individual uNK cells. The number of uNK cells (PAS positive) per square millimeter in the MLAp and in the DB was counted at ×500 magnification using an eyepiece micrometer grid of 1 mm². The circular smooth muscle was used as the dividing line between the MLAp and DB. Cross-sectional area measurements of the MLAp, decidua, and placenta, and ratios for vessel:lumen diameter line morphometry of the main decidual arteries, cut in cross-section, were measured on the same slides that were used for uNK cell numeration, using OPTIMAS image analysis software, version 6.2 (Optimas, Bothwell, MA).

Six- to 8-wk-old RAG-2^−/−γ_c^−/− or IL-15^−/− females were used as recipients of BM. Each recipient was pretreated with a single i.p. injection of 5-fluorouracil (150 mg/kg) 48 h before BM infusion. Six- to 8-wk-old IL-15^−/−, IRF-1^−/−, or B6 mice were used as BM donors. BM cells were pooled from femurs and humeri. Donor cells were depleted of RBC using lysis buffer. A total of 2 × 10⁷ viable BM cells was given i.v. to each 5-fluorouracil-treated female. Three weeks later, recipients were paired with RAG-2^−/−γ_c^−/− males for mating and euthanized at gd 12.

Total RNA was isolated from freshly dissected tissue using RNeasy mini kit (Qiagen, Hilden, Germany), according to the manufacturer’s instructions. Two micrograms of total RNA were reverse transcribed using first-strand cDNA synthesis kit (Amersham Pharmacia Biotech, Piscataway, NJ). cDNA were amplified using the following conditions: 94°C for 6 min; 30–35 cycles of 94°C for 45 s, 56°C for 45 s, and 72°C for 45 s; and an extra 7 min at 72°C following the last cycle. The PCR primers were as follows: IL-15–5′, 5′-GCCATAGCCAGCTCATCTTC-3′; IL-15–3′, 5′-GCAATTCCAGGAGAAAGCAG-3′; IRF-2–5′, 5′-ACACTGGAGGAAGAGGAGCA-3′; IRF-2–3′, 5′-CAACAACCACCAGGGAGAGT-3′; β-actin-5′, 5′-GCTACAGCTTCACCACCACA-3′; and β-actin-3′, 5′-ACATCTGCTGGTAGGTGGAC-3′.

For IRF-1 Northern analysis, total RNA was isolated from freshly dissected tissue using the RNeasy mini or midi kits (Qiagen) according to the manufacturer’s instructions. Total RNA from B6 spleen cells, which had been activated with murine (m)rIFN-γ (100 U/ml) and LPS (200 ng/ml) for 6 h was used as positive control. Ten to 20 μg of total RNA were electrophoresed in 1% denaturing agarose-formaldehyde gels and then transferred onto Hybond nylon membranes (Amersham Life Sciences, Oakville, Ontario, Canada) according to standard protocols. Following electroblotting, the RNA was cross-linked by UV irradiation, and the nylon membranes were stored at 4°C until further processing. The mouse IRF-1 (574-bp) probe was generated from clone no. 777393 supplied by GenomeSystems (St. Louis, MO). Radioactive probes were prepared using a commercially prepared Rediprime random-primer DNA-labeling mixture (Amersham International, Buckinghamshire, U.K.). After dissolving the Rediprime mixture and DNA, 50 μCi of radioactive nucleotide [α-³²P]dCTP (specific activity, 3000 μCi/mmol; Amersham Life Sciences) was added. The mixture was incubated for 10 min at 37°C. Reactions were stopped by adding 5 μl of 0.2 M EDTA. The radioactive probe was purified by centrifuging 300 μl of TEN through a Quick Spin column (Bio-Rad Laboratories, Hercules, CA) containing 1 ml of Sephadex G-50 at 2000 rpm for 2 min. Membranes supporting the RNA samples were first prehybridized for 1 h at 42°C in a solution containing 5× sodium chloride/sodium phosphate EDTA, 0.5% SDS, 5× Denhardt’s solution, salmon sperm DNA (40 μg/ml), and 50% formamide. Then, the prehybridization buffer was replaced by hybridization buffer containing IRF-1 or 7S probes. Following overnight hybridization at 42°C, membranes were washed twice with low-stringency solution containing 2× SSC and 1% SDS at 42°C for 15 min, followed by a high-stringency wash with 0.1× SSC and 0.1% SDS at 50°C for 10 min. Blots were rinsed in 5× sodium chloride/sodium phosphateEDTA, and the hybridization signal was detected by autoradiography, using Kodak (Rochester, NY) x-ray film. For IL-15, total RNA extracted by the guanidinium isothiocyanate method (25) was analyzed by Northern blot hybridization as described, including stringent posthybridization washes (26). Each lane contained 10 μg of total RNA. [³²P]cRNA probes were generated from a ribovector containing mIL-15 cDNA, generously supplied by A. Troutt (Immunex, Seattle, WA).

In situ hybridization was performed on paraffin-embedded sections of nonpregnant gestation day-timed uteri from B6 and alymphoid mice. Clone no. 777393 (supplied by GenomeSystems) was used for the mouse IRF-1 (574-bp) probe. Digoxigenin-labeled IRF-1 sense and antisense probes were made using Northern starter kit, according to the manufacturer’s instructions (Roche, Montreal, Quebec, Canada). In situ hybridization on 8-μm sections was performed according to the nonradioactive in situ hybridization application manual (Roche). For the visualization of the RNA probe, the sections were incubated with 5-bromo-4-chloro-3-indolyl phosphate/nitroblue tetrazolium liquid substrate system (Sigma-Aldrich, Oakville, Ontario, Canada) for up to 24 h.

IRF-1^−/− mice received 500 ng of mrIL-15 (Research Diagnostics, Flanders, NJ) i.p., daily for 5 days, from gd 5. On gd 10, pregnant uteri were dissected and processed for analysis. The dosage and route of mrIL-15 were selected based on previous in vivo experiments (10, 27).

The means of experimental groups were compared using two-factor ANOVA. The between-mice and the between-implantation sites variances were considered as analytic factors (p < 0.05). Groups with significantly different means were identified using Tukey’s test (p < 0.05). Pup weights were compared by paired t test.

IL-15 is essential for NK cell development in lymphoid tissue. Implantation sites of IL-15^−/− mice differed from normal mice in that they completely lacked uNK cells. In addition, IL-15^−/− mice did not develop MLAp, had exceedingly hypocellular DB, and failed to modify SA (Fig. 1, a vs c). The latter features resemble those seen in alymphoid mice (Fig. 1 b) but are more pronounced. Placental cross-sectional areas in IL-15^−/− mice were similar to those in time-matched normal controls (data not shown), and no elevated fetal loss (fetal loss, 2.3%; mean litter size, 8.2) was observed.

To determine whether IL-15 was the sole limitation to uNK cell differentiation in IL-15^−/− mice, IL-15^−/− BM was transplanted into alymphoid mice, and B6 BM was transplanted into IL-15^−/− or alymphoid recipients. Transplanted females were mated to nontransplanted, syngeneic males, and gd 12 implantation sites were studied. uNK cells differentiated from IL-15^−/− BM and B6 BM in alymphoid mice to numbers equivalent to those found in normal mice. An MLAp developed, decidua became highly cellular, and arterial change was induced. However, transplanted B6 BM failed to give rise to uNK cells in pregnant IL-15^−/− hosts, no MLAp was present, decidual hypocellularity persisted, and arterial change was not induced (Figs. 2 and 3). Implantation sites of alymphoid mice were positive for IL-15 mRNA (Fig. 2 d)

Because IRF-1 is thought to be essential for IL-15 expression in BM (Ogasawara et al., 1998), and both IRF-1 and -2 are important for NK cell development in marrow, implantation sites of IRF-1^−/− and IRF-2^−/− mice were studied. They had no elevated fetal loss (fetal loss, 4.63%) compared with their congenic control. IRF-2^−/− sites did not differ histologically from implantation sites in normal mice (Fig. 4,b), although RT-PCR showed that IRF-2 was abundantly expressed in virgin and pregnant uteri of B6 and alymphoid mice (d). In contrast, in IRF-1^−/− mice after gd 6, implantation sites were smaller grossly and in histologic cross-sectional area than in time-matched B6 controls (Figs. 4,a and 5). uNK cells differentiated in IRF-1^−/− mice but were few in number, small, and hypogranular. In addition, MLAp regions were poorly developed, and SA failed to modify (Fig. 4 a).

To assess whether small placental size affected fetal and/or postnatal development, IRF-1^−/− litters were weighed at birth, weaning (21 days), and adulthood (49 days). Neonates were significantly smaller than barrier husbandry-reared, age-matched B6, and weight deficits persisted into adulthood (Fig. 5).

Microarray (GEM; Incyte, St. Louis, MO) data, based upon RNA collected from mesometrial triangles of B6 mice on gd 6 and MLAps on gd 10, showed no IRF-1 expression (our unpublished observations). This was confirmed by Northern analyses conducted on microdissected mesometrial triangle (gd 6) and MLAp (gd 10) regions from B6 and alymphoid mice (Fig. 4,c). To localize the microdomains of IRF-1 expression within implantation sites, in situ hybridization was conducted on virgin and pregnant uteri from B6 and alymphoid mice. In both strains, signals in virgin uteri were localized to the lumenal and glandular epithelia (Fig. 4, e and e1). At gd 8, in both strains, very strong hybridization was found in trophoblast giant cells (Fig. 4, j and k). At gd 10, giant cells remained heavily labeled (Fig. 4,g). The now mature placenta was also positive, as were the uterine epithelial cells adjacent to and at the antimesometrial aspect of the implantation sites (Fig. 4,f). The uNK cell-rich MLAp was negative, but signal was observed in dispersed decidual cells and SA walls (Fig. 4 h).

To determine whether IRF-1^−/− uNK precursors are competent for normal maturation and numerical expansion in an IRF-1^+/+ uterine environment, alymphoid mice were engrafted with BM from IRF-1^−/− mice. This established terminally differentiated uNK cells in recipients’ implantation sites (Fig. 6,a; compare with c for normal control, d for nonengrafted host, and f for donor implantation sites). Reconstitution of alymphoid mice by IRF-1^−/− BM induced MLAp development and reversed the anomalies in the decidua and its arteries seen in implantation sites of untreated hosts (Fig. 6, compare b with e). Quantification showed that IRF-1^−/− BM reconstituted uNK cell numbers to those observed in B6 (Fig. 6 g).

Because BM cells from IL-15^−/− and IRF-1^−/− mice differentiate into uNK cells in alymphoid mice, presence of IL-15 in the recipient uteri was anticipated. Uterine mRNA from virgin and pregnant alymphoid, IRF-1^−/− and B6 mice was assessed for IL-15. RT-PCR, Northern (Fig. 7), and RNase protection assays (not shown) all indicated that mRNA for IL-15 was induced after uterine decidualization in the three strains. Attempts to address IL-15 at the protein level were unsuccessful.

To further address whether anomalies at the implantation sites of IRF-1^−/− mice were due to IL-15 protein deficiency, mrIL-15 was administered for 5 days to pregnant IRF-1^−/− mice. Treatment did not induce changes to IRF-1^−/− implantation sites. The uNK cell population remained immature and was not increased in number. MLAp was not induced, and no significant changes were induced in SA or placental size from mrIL-15 treatment of IRF-1^−/− mice compared with untreated controls (Table I).

These studies clearly demonstrate that IL-15 is absolutely essential for the support of NK cell differentiation in the decidualizing uterus. Analyses of implantation sites in IL-15^−/− mice revealed a complete absence of uNK cells as well as pathology consistent with that seen in other strains severely or totally deficient in uNK cells. The features included unmodified SA structure, poor development of decidua, and absence of MLAp development within the uterine wall. These features do not compromise fetal viability or postnatal survival under microbiological barrier husbandry and are assumed to negatively reflect the major functions of uNK cells. No measures of fetal stress or postnatal physiology were performed.

The time course for transcription of mouse uterine IL-15 expression (13) mirrors that for development of uNK cells. Rare cells are first found at gd 5 with significant numbers on gd 6 when IL-15 message is first reported. uNK cells proliferate within mouse uterus from gd 6 to midgestation, achieve relatively stable peak numbers between gd 10 and 12, and then gradually decline in numbers (3, 28). IL-15 message in decidua is found to gd 11 and is then absent to term (13). Withdrawal of IL-15 would be the simplest explanation for the disappearance of uNK cells. IL-15 withdrawal is likely gradual due to transendosomal cycling providing persistent surface-bound IL-15 in decidual stroma (29). Just before parturition, ∼10% of the peak uNK cell number remains. These cells are shed with the placenta (28). Induction of IL-15 transcription in SA endothelium, as seen in women, may also occur in mice. This could account for the highly localized appearance of intravascular uNK cells within the component of the SA coursing through the decidua (1, 30, 31) and for uNK cell absence from more superficial segments of the same vessels within the MLAp (32). This positional relationship may be of major functional importance in elongation and dilation of SA, because both mouse and human uNK cells express endothelial cell mitogens and other vasoactive molecules such as inducible NO synthase (33, 34)

If IL-15 is the only growth factor essential for uNK cell differentiation, as suggested by our marrow transplants, IL-15 regulation within uterus becomes an important question. In marrow, IRF-1 appears to be a key transcriptional activation factor for IL-15 (20). Implantation sites in IRF-1^−/− mice differ from those in any other mouse strain reported to date. The uNK cell lineage was established, but it failed to fully differentiate numerically or morphologically. Furthermore, a major role for IRF-1 in trophoblast, particularly giant-cell, biology was defined. Over a period of 3 years, no drift was seen in the IRF-1^−/− phenotype of small placentae and small size of offspring. Small placentae may be the primary cause of the intrauterine growth retardation, or there may be additional primary deficits in fetuses as themselves during development. The lasting effect of this gene deletion into adulthood suggests that IRF-1^−/− mice maybe a convenient model to address postnatal outcomes in growth-restricted humans (35). Consistent with other in vitro and in vivo models evaluating differentiation of NK cells, moving marrow cells from IRF-1^−/− to a normal stromal environment, showed that the gene deletion did not impair the differentiation capacity of uNK progenitor cells.

Limited differentiation of uNK cells in the IRF-1-free environment suggested that some IL-15 was present in the uteri of IRF-1^−/− mice that could interact with the high-affinity IL-15R on uNK progenitor cells (6, 36). This was supported at the level of IL-15 gene transcription and by the inability of exogenous IL-15 treatment to normalized IRF-1^−/− implantation sites. Thus, IRF-1 in a normal mouse may regulate expression of genes involved in recruitment of uNK cell precursors to the uterus (37), or it may act indirectly in stromal-derived cytokine support of terminal uNK cell differentiation steps. Vascular addressin VCAM could be a key IRF-1-regulated recruitment molecule (38). VCAM has unique expression in DB and is postulated to be the major addressin homing NK cells to the uterus (31, 37). IRF-1 also contributes to vascular remodeling events where it is antagonized by IRF-2 (39, 40). IRF-2 appears to be an enhancer of VCAM-1 expression in myogenic events outside of the vasculature (41).

Endocrine control of IRF-1 and IL-15 gene expression has been proposed in human uterus and may account for the absence of IRF-1 regulation of murine uterine IL-15. Prolactin, a pituitary hormone, is thought to enhance IRF-1 expression (42), whereas progesterone and PG, derived from ovary and placenta, are proposed as regulators of IL-15 expression (14, 15, 16). However, these mechanisms fail to explain regulation of NK cell differentiation in decidua, because the hormone effects are present systemically. Thus, additional unique regulatory mechanisms must be involved in IL-15 expression in decidualizing uterus.

The results of this study suggest that uNK cell development and maturation have aspects similar to NK cell development in other tissues but, in addition, have distinct differences. The absence of IL-15 leads to the absence of both uNK and NK cells, whereas absence of IRF-1 results in partial deficiency of uNK and complete deficiency of NK cells. In the absence of IRF-2, uNK cells are normal in number and function, but there is a severe deficiency of peripheral NK cells. This may suggest tissue-specific regulation of IL-15 in uterine stromal cells. Unique, IRF-1-regulated decidual cytokines may account for the skewing of NK cells features seen in uterine compared with peripheral sites.

We thank Immunex, Inc., for providing the IL-15 cDNA probe and permission to receive breeding pairs of IL-15^−/− mice from Dr. M. Caligiuri (Ohio State University, School of Medicine, Columbus, OH); Dr. T. W. Mak for providing IRF-1^−/− and breeding pairs of IRF-2^−/− mice; Lorena Montoya, Martha Brackin, and the Ontario Ministry of Agriculture, Food, and Rural Affairs Isolation staff for their technical support.