IFNs Drive Development of Novel IL-15–Responsive Macrophages

The notion of immune quiescence during pregnancy forms the basis for the decades-old central tenet of reproductive immunology (1). However, countless pieces of evidence now support that the action of proinflammatory cytokines are critical to a healthy gestation for both mother and fetus (2). As in many other contexts, the proinflammatory response must be regulated to avoid an adverse outcome of pregnancy caused by damage to the developing fetoplacental unit.

Perturbation in homeostasis of the pleiotropic cytokine IL-15 has been linked to poor maternal and fetal outcomes of pregnancy in mice and humans. Mice lacking IL-15 (Il15^−/−) bear growth-restricted pups and exhibit higher rates of spontaneous resorptions than do IL-15–sufficient mice (3, 4). Mice deficient in either production of IL-15 or responsivity to IL-15 exhibit abnormal uteroplacental anatomy, including impaired remodeling of the uterine spiral arteries, a pathologic hallmark of life-threatening preeclampsia in humans (5). These data translate well to findings of reduced IL-15 in human placentae of pregnancies affected by preeclampsia (6). Conversely, IL-15 permits inflammatory-mediated fetal loss in mice (4). Indeed, spontaneous and recurrent miscarriage in humans is associated with unrestrained expression of IL-15 mRNA and protein (7).

IL-15 is abundant during normal pregnancy. In mice, IL-15 mRNA is expressed in the nonpregnant uterus, as well as in the uteroplacental unit throughout pregnancy, with a peak at midgestation (8). In humans, IL-15 mRNA is found in low abundance in proliferative phase endometrium and increases substantially in secretory phase endometrium and first-trimester decidua (9). IL-15 protein has been demonstrated in endometrial stromal cells, perivascular cells abutting uterine spiral arteries, and vascular endothelial cells, echoing the presumed roles for IL-15–responsive cells in vascular remodeling.

In the current model of IL-15 signaling, IL-15 is complexed with the high-affinity α subunit of the IL-15R (IL-15Rα) (10). Myeloid cells and nonhematopoietic cells then present the IL-15/IL-15Rα complex in trans to IL-15–responsive cells, typically killer lymphocytes bearing high levels of the β-chain of the IL-15R complex (CD122) and the common γ-chain (γ_c). Indeed, at the maternal–fetal interface, IL-15 is bound to the cell surface of CD14⁺ monocytes and macrophages in single-cell suspensions of human first-trimester uterine decidua (9), the uterine lining remodeled to accept an embryo. Consistent with these human data, pregnant mice lacking γ_c but reconstituted with bone marrow (BM) from Il15^−/− mice exhibit normal uterine vascular remodeling. These findings support that the dominant sources of IL-15 in mice also are nonhematopoietic cells and chemotherapy-resistant myeloid cells (11).

Altogether, these data show that appropriate signaling by IL-15 is critically important to healthy pregnancy. However, the targets and mechanisms of IL-15 signaling at the maternal–fetal interface are incompletely understood. NK cells are the prototypical IL-15–responsive cell type at the maternal–fetal interface. They are highly prevalent in the uterus and produce proinflammatory IFN-γ, without which uterine arteries do not remodel and preeclampsia develops in mice (12). Because uterine and other NK cells clearly depend on IL-15 for development (3, 13), it has been presumed that NK cells are responsible for all abnormalities of pregnancy associated with disrupted homeostasis of IL-15.

We now report that novel macrophages, termed CD122⁺ macrophages (CD122+Macs), express high levels of CD122 in the uterus of mice and humans. M-CSF permits type I and II IFNs to drive expression of CD122 on these macrophages. CD122+Macs respond to IL-15 by activating the ERK signaling cascade, and CD122+Macs stimulated with the TLR agonist CpG enhance production of proinflammatory cytokines in response to IL-15. CD122+Macs represent a new and unexpected target of IL-15 in the uteroplacental unit and, thus, may have significant clinical implications in healthy and complicated pregnancies.

All animals were housed, cared for, and euthanized in accordance with a protocol approved by the Institutional Animal Care and Use Committee of Children’s Hospital of Philadelphia. Wild-type mice were strain C57BL/6, and all knockout (KO) mice were on a C57BL/6 background. Ifnar^−/−, Ifngr^−/−, and Ifnar^−/−Ifngr^−/− mice, LysM-Cre and DTR-mCherry transgenic mice intercrossed to yield MM-DTR mice, and CD45.1 mice were all purchased from The Jackson Laboratory. All animals used were between 6 and 12 wk of age. Although it was only possible to use female mice to investigate uterine immune cells, both male and female mice were used for in vitro experiments, yielding similar results. For pregnant female mice, presence of copulation plugs were checked early each morning. Copulation was assumed to take place during the 12-h dark cycle (from 6 pm to 6 am in our facility). Thus, embryonic day 0.5 (E0.5) denotes the morning that a copulation plug was first detected.

Endometrial biopsies were obtained from patients and volunteers at Penn Fertility Care. M.A.M. has approval for the collection endometrial tissue (Institutional Review Board [IRB] no. 826813). Written informed consent was obtained from all subjects. All women were between the ages of 18 and 43, with regular menstrual cycles (between 25 and 35 d over the past 3 mo) and no significant medical history. Endometrial biopsies were obtained from known fertile women 8 d after a luteinizing hormone surge (secretory phase), which was detected in the urine by the Clearblue Ovulation Test. All biopsies were obtained using a Pipelle Endometrial Suction Curette (Cooper Surgical). All biopsies were dissociated and analyzed within 24 h of collection. Single-cell suspensions were obtained as below.

First-trimester decidual tissue was obtained from the Penn Family Planning and Pregnancy Loss Center. MAM has IRB approval for the collection of first-trimester extraembryonic tissue (IRB no. 827072). Written informed consent was obtained from all subjects. After tissue collection, decidua was manually separated from chorionic villi based on gross morphology. All samples were dissociated and analyzed within 24 h of collection. Single-cell suspensions were obtained as below.

Organs of interest were grossly dissected, weighed, minced finely with scissors, and digested in medium containing Liberase (Roche) at a final concentration of 0.28 Wünsch units/ml and DNase I (Roche) at a final concentration of 30 μg/ml for 30 min at 37°C with intermittent agitation. Cells were passed through a 70-μm filter, and RBCs were lysed with ammonium–chloride–potassium lysis buffer. Cells were again filtered and counted with a Countess hemocytometer (Thermo Fisher Scientific [TF]) prior to proceeding to Ab staining for flow cytometry or cell sorting.

Flow cytometry was performed on either an MACSQuant Analyzer 10 (Miltenyi Biotec), an LSRFortessa (BD Biosciences [BD]), or a CytoFLEX LX (Beckman Coulter). Cell sorting was performed on either an FACSAria Fusion (BD) or an MoFlo Astrios EQ (Beckman Coulter). Data were exported as FCS files and analyzed using FlowJo 10 software. All Ab staining was performed at 4°C in the dark. Prior to fluorescent Ab staining, cells were incubated with mouse Fc block (TruStain FcX; BioLegend [BL]) or human Fc block (BD). Fixable, fluorescent LIVE/DEAD viability dye (TF) was used in Blue, Aqua, Green, and near-infrared per manufacturer recommendations. The following anti-mouse Abs, colors, clones, and concentrations were used from BD, BL, R&D Systems, and TF: CCR2 (475301, PE, 1:100; R&D Systems), CD11b (M1/70, FITC, 1:200; BD), CD11c (N418, PE-Cy7, 1:200; BL), CD122 (TM-β1, BV421/PE, 1:100; BD), CD19 (6D5, APC-Cy7, 1:400; BL), CD3 (17A2, APC-Cy7, 1:100; BL), CD45.1 (A20, FITC, 1:100; BD), CD45.2 (104, PerCP-Cy5.5, 1:100; BL), CD49a (Ha31/8, PerCP-Cy5.5, 1:100; BD), CD49b (DX5, PE, 1:100; BL), CD64 (X54-5/7.1, PE-Cy7, 1:200; BL), Eomes (dan11mag, AF488, 1:200; TF), F4/80 (BM8, APC/PE, 1:100; BL), Ly-6C (HK1.4, AF700/Pacific Blue, 1:400; BL), Ly-6G (1A8, APC/Fire 750, 1:400; BL), MERTK (2B10C42, PE, 1:100; BL), MHC class II (MHCII) (M5/114.15.2, PerCP-Cy5.5, 1:200; BD), NKp46 (29A1.4, FITC, 1:100; TF), and T-bet (4B10, BV421, 1:100; BL). The following anti-human Abs were used: CD11c (B-ly6, PE-Cy7, 1:100; BD), CD122 (TU27, PE, 1:100; BL), CD14 (M5E2, V450, 1:100; BD), CD16 (3G8, AF700, 1:100; BL), CD19 (HIB19, BV605, 1:100; BL), CD3 (OKT3, BV785, 1:100; BL), CD45 (HI30, PerCP-Cy5.5, 1:100; BD), CD56 (5.1H11, BV650, 1:100; BL), CD64 (10.1, APC/Fire 750, 1:100; BL), granzyme B (GB11, Pacific Blue/FITC, 1:100; BL), HLA-DR (G46-6, V500, 1:100; BD), and perforin (δG9, AF647, 1:100, BD).

Sorted cells were pelleted, removed of medium, snap frozen on dry ice, and stored at −80°C. RNA was isolated using a Micro RNeasy mini total RNA kit (QIAGEN), performed according to the manufacturer’s instructions. MouseGene 1.0ST chips were used. Microarray data were normalized by the Robust Multichip Average algorithm and log₂ transformed using the oligo package in R (14). Linear modeling to obtain differentially expressed genes was performed with the limma package in R (15). A gene was considered significantly differentially expressed with a false discovery rate–adjusted p value <0.05. Gene ontology (GO) analysis was performed using DAVID (16, 17).

Single-cell suspensions of bulk BM were obtained from pooled femorae, tibiae, and humeri of donor mice. RBCs were lysed with ammonium–chloride–potassium lysis buffer, and cells were filtered through a 70-μm filter. Monocytes were enriched with a mouse BM monocyte isolation kit (Miltenyi Biotec) by magnetic bead–mediated depletion of nonmonocytes. Manufacturer instructions were followed exactly. Purity of transferred BM monocytes was routinely ∼90% prior to adoptive transfer. At least 10 million cells were injected i.v. per recipient.

Single-cell suspensions of BM were prepared as above. Adherent cells were removed by incubating bulk BM cells on tissue culture-treated plates for at least 2 h in complete DMEM (cDMEM; 10% FBS and 1% penicillin/streptomycin/glutamine). Nonadherent cells were then collected, counted, and plated at 5 × 10⁵ cells/ml in cDMEM containing 10% L929 cell–conditioned medium (made in our laboratory) for 5–7 d. GM-CSF/BM–derived macrophages (GM-BMDMs) were generated by culturing nonadherent BM cells in cDMEM supplemented with 50 ng/ml GM-CSF (PeproTech). CD122⁺ BMDMs were generated by adding 1 ng/ml IFN-α (BL) to the medium on day 5 of culture. When testing ability of IFN-γ (PeproTech) to derive CD122+Macs, doses indicated were added on day 5 of culture. Anti-IFNAR and anti-IFNGR (both Bio X Cell) were both used at 10 μg/ml in culture.

CD122⁺BMDMs were generated as above. Cells were washed on day 6 of culture, and medium was replaced with DMEM containing 3% serum for the final 16 h of culture with or without 1 μg/ml CpG 1826 (Integrated DNA Technologies), as indicated. To this medium, IL-15Rα alone (IL-15Rα Fc chimera protein; R&D Systems) or IL-15Rα precomplexed with IL-15 (PeproTech) was added, as indicated. Final concentrations of IL-15Rα and IL-15 were 100 and 50 ng/ml, respectively. Complexing of IL-15Rα with IL-15 was performed in PBS at 37°C for 30 min. OptEIA ELISA kits (BD) for mouse were used: IL-6 and TNF. Manufacturer instructions were followed precisely, and concentrations of cytokines were determined with a SpectraMax ELISA plate reader.

CD122⁺BMDMs were generated as above. Cells were washed on day 6 of culture, and medium was replaced with serum-free DMEM for 2 h. Medium was replaced with cDMEM containing IL-15Rα alone or IL-15Rα/IL-15 complex, as above, for 20 min. Whole-cell lysates were then prepared by washing cells in cold PBS and lysing in M-PER Mammalian Protein Extraction Reagent (TF) with protease/phosphatase inhibitor mixture (Cell Signaling Technology). Lysates were quantified by Bradford assay, normalized, reduced, and resolved by SDS gel electrophoresis on 10% Bis-Tris gels (Invitrogen). Resolved proteins were transferred to 0.45 μm nitrocellulose membranes (Bio-Rad Laboratories). Membranes were blocked in Odyssey Blocking Buffer (LI-COR Biosciences) and probed with the following Abs: pERK (1:1000; Cell Signaling Technology), total ERK (1:1000; Cell Signaling Technology), and GAPDH (1:2000; Novus Bio). Membranes were incubated with the following secondary Abs at a dilution of 1:20,000: Alexa Fluor 680 donkey anti-rabbit IgG (Invitrogen) or Alexa Fluor Plus 800 donkey anti-mouse IgG (Invitrogen). Proteins were detected and quantified using an LI-COR Odyssey.

For nonmicroarray data, statistical analyses were performed using GraphPad Prism 8. A (nonparametric) Wilcoxon matched-pairs signed rank test was used to compare effect of IL-15 treatment on matched samples. Mixed-effects analysis was used to compare paired observations with one-way ANOVA. Holm–Sidak multiple comparisons test was used in mixed-effects analyses. In all cases, p < 0.05 was considered significant.

To determine whether uterine leukocytes other than NK cells could be cellular targets of IL-15, we comprehensively examined expression of CD122 on immune cells in pregnant female mice. Populations of cells coexpressing CD122 and high levels of macrophage/monocyte-associated FcγRI (CD64) and F4/80 were evident in several organs but were particularly enriched in the uterus of pregnant and nonpregnant mice (Fig. 1A, Supplemental Fig. 1A). Other myeloid-phenotype cells, such as Ly-6G⁺ cells (neutrophils) and CD11c^hiMHCII^hi cells (conventional dendritic cells [DCs]), did not exhibit surface expression of CD122 (Fig. 1A). During gestation, these CD122+Macs were present throughout the maternal–fetal interface: in the myometrium, mesometrial lymphoid aggregate of pregnancy, decidua, and placenta (Fig. 1B).

Rarely are classical myeloid and lymphoid proteins coexpressed in the same cell. Recently, NK cells expressing canonical myeloid transcripts, including Csf1r, were identified in obese mice and humans (18). Thus, it was possible that CD122+Macs represented a subpopulation of so-called myeloid NK cells given the high abundance of NK cells in the early gestation uterus. To determine to which lineage uterine CD122+Macs cells belonged, we compared their morphology to other uterine cell types. Murine uterine CD122⁺F4/80^hi cells were morphologically large and hypervacuolated, similar to CD122⁻ conventional macrophages (cMacs; Supplemental Fig. 1B, 1C). The cell surface phenotype of CD122+Macs overlapped substantially with that of cMacs. Most CD122+Macs and cMacs were CD11c^int, Ly-6C^lo, CCR2^hi, MERTK^hi, CD206⁺, and CD209⁺, a constellation of findings typically associated with macrophages (Fig. 1C, 1D, Supplemental Fig. 1D). Aside from CD122, CD122+Macs and cMacs could be distinguished by expression of MHCII with CD122+Macs largely MHCII^lo and cMacs largely MHCII^hi (Fig. 1C).

In contrast to CD122+Macs and cMacs, the morphology and flow cytometric profile of CD122⁺F4/80-CD64⁻ cells were consistent with classical large, granular, decidual NK cells (Fig. 1C, Supplemental Fig. 1B–F). Nearly all decidual NK cells express NKp46 and the tissue-resident NK cell marker CD49a, especially early in gestation (19). We found that CD122+Macs did not express CD49a, CD49b/DX5, T-bet, or Eomesodermin (Supplemental Fig. 1F), closely related master regulators of innate lymphoid fate that are abundantly expressed in uterine innate lymphoid cells (19–21). Morphologic and phenotypic data thus supported the notion that CD122+Macs cells are indeed macrophage-lineage cells and not NK cells, DCs, or neutrophils.

We next tested whether CD122+Macs were most like macrophages or another cell type at the level of the transcriptome. A monocyte–macrophage-specific reporter system, known as the MM-DTR mouse, fluorescently labels with mCherry monocytes and macrophages that transcribe both Lyz2 (LysM) and Csf1r (M-CSFR) (22). Expression of CD122 directly correlated with expression of mCherry in F4/80^hi cells in the uterus (Fig. 1E). We then sort-purified decidual CD122+Macs, cMacs, and NK cells to a purity of at least 95% (Supplemental Fig. 1G) and analyzed the transcriptome using microarrays. We found that 86.5% of the variance in gene expression among these populations could be attributed to the first principal component, PC1 (Fig. 1F). Clustering samples along PC1 clearly separated cMacs and CD122+Macs from NK cells. Similarly, hierarchical clustering analysis showed that all macrophages were closely related to each other but distantly related to NK cells (Fig. 1G). Among genes enriched in CD122+Macs relative to NK cells were those classically associated with macrophages, including numerous complement receptors, TLR, and Csf1r, Fcgr1, Lyz2, and Cx3cr1 (Supplemental Table I). Among genes enriched in NK cells relative to CD122+Macs were those typically associated with innate and killer lymphocytes, including Ncr1 (NKp46), transcripts encoding cytolytic mediators, Ly-49 receptors, and Eomes (Eomesodermin) and Tbx21 (T-bet). Of note, some cytolytic genes, including Prf1 and several granzymes, were modestly but significantly enriched in CD122+Macs relative to cMacs (Fig. 1H, Supplemental Table II). Altogether, these data support the notion that CD122+Macs and cMacs are highly related but distinct subtypes of uterine macrophages.

Given the unusual phenotype and restricted anatomic location of CD122+Macs in mice, we assessed whether such macrophages were also found in the human uterus. We therefore examined human secretory endometrium during the implantation window, as well as human first-trimester decidua, for the presence of human CD122+Macs (hCD122+Macs). We relied on expression of CD14 and CD64 to identify human CD14⁺ monocytes and macrophages by flow cytometry (Supplemental Fig. 2A). Prior work in the decidua and skin support that CD14⁺ cells in these organs are macrophages by phenotype, gene expression, and ontogeny (23, 24). Further, we confirmed in a publicly available single-cell RNA-sequencing database that the majority of decidual macrophages expressed Cd14 at the transcriptional level (25). We defined human NK cells as CD3⁻CD19⁻CD14⁻CD56^bright (uterine) or CD3⁻CD19⁻CD14⁻CD56^dim (blood). As expected, human NK cells were abundant in the uterus and expressed high levels of CD122.

We then identified CD14/CD64⁺ cells enriched in the human pre- and postimplantation uterus that expressed CD122 (Fig. 2, Supplemental Fig. 2A–G). These hCD122+Macs varied widely in abundance among individual samples but were present in all first-trimester deciduae tested (Fig. 2A, 2B, Supplemental Fig. 2A–D). hCD122+Macs were also present in over half of human secretory phase endometrial biopsies examined during the implantation window (Fig. 2C, Supplemental Fig. 2E–G). We found that hCD122+Macs often expressed CD16 and variably expressed CD11c, as did human conventional macrophages (Fig. 2B, 2C, Supplemental Fig. 2A–D, 2G). Both hCD122+Macs and human conventional macrophages expressed HLA-DR, but in some samples, hCD122+Macs expressed modestly less HLA-DR, similar to our observations of MHCII levels on CD122+Macs and cMacs in the mouse.

hCD122+Macs variably expressed CD56 (Fig. 2, Supplemental Fig. 2A–G). CD56 is typically associated with cytotoxic lymphocytes, but it has been found on myeloid cells in certain contexts (26). Consistent with this literature, hCD122+Macs expressing higher levels of CD56 appeared to express the cytolytic molecules perforin and granzyme B at the protein level (Fig. 2B, Supplemental Fig. 2C, 2D, 2G). These data agree with our transcript-level data in mice, showing enrichment of cytolytic mediators in CD122+Macs relative to cMacs (Fig. 1H, Supplemental Table II). Overall, hCD122+Macs bore striking resemblance to murine CD122+Macs, supporting the use of our murine model to obtain mechanistic insights into the biology of this novel uterine macrophage in humans.

Our morphologic, phenotypic, and transcriptomic data supported that CD122+Macs were macrophages. Myriad embryonic and adult progenitors give rise to macrophages in different tissues (27, 28). Monocytes are recruited in large numbers to the gravid uterus during gestation (29). We thus tested the hypothesis that CD122+Macs could derive from adult BM monocytes. Magnetic bead–purified Ly-6C^hi monocytes from adult mouse BM of MM-DTR mice were adoptively transferred into pregnant recipients during the peri-implantation period (Fig. 3). Recipients were then sacrificed at midgestation. We observed tissue-specific differences in phenotype of recovered donor cells with respect to CD64, F4/80, and CD122. Donor cells recovered from recipient blood, spleen, and BM were almost uniformly CD64^loF4/80^lo and FSC^loSSC^lo (Fig. 3B and data not shown), suggesting maintenance of monocyte fate (30). In contrast, donor cells recovered from recipient peritoneum, myometrium, and decidua included CD64^hiF4/80^hiFSC^hiSSC^hi cells, suggesting some conversion of adoptively transferred monocytes to tissue macrophages in these organs.

Despite conversion of monocytes to CD64^hiF4/80^hi cells in the peritoneum, we found that transferred monocytes converted to macrophages expressing high levels of CD122 only in the uterus (Fig. 3B). These data are consistent with our findings that CD122+Macs are enriched in the uterus in the steady-state. Of note, we performed this experiment with both pregnant and nonpregnant donors with identical results (data not shown), suggesting no intrinsic differences in potential of BM monocytes between pregnant and nonpregnant mice. Altogether, our data support a model in which BM monocytes are recruited to the uterus and differentiate into CD122+Macs.

The immune composition of the maternal–fetal interface changes dramatically throughout gestation. Consistent with prior reports (31), we found substantial declines in NK cells after midgestation (Fig. 4A, 4B, and data not shown). Similarly, populations of myeloid cells, including DCs, monocytes, and macrophages, are in constant flux at the maternal–fetal interface during gestation (29, 32). We thus investigated population dynamics of uterine CD122+Macs before and during pregnancy in the mouse. The myometrium exhibited progressive percent enrichment of total Macs cells over the course of gestation (Fig. 4), which agrees with and extends prior findings (29). Myometrial CD122+Macs were evident throughout gestation. The percentages of myometrial CD122+Macs, relative to total Macs, peaked after midgestation and declined during late gestation.

In contrast to Macs in the myometrium, the population of total decidual/placental Macs peaked prior to midgestation and declined thereafter, reaching a plateau in late gestation. Percentages of decidual/placental CD122+Macs remained relatively stable throughout gestation. We did, however, observe dynamic regulation of surface expression of CD122 by decidual/placental CD122+Macs. In the early postimplantation period through E10.5, CD122+Macs expressed moderate amounts of CD122 in the decidua and developing placenta (Fig. 4A, 4B). From E11.5 through E18.5, the last timepoint we examined before delivery, deciduo-placental CD122+Macs exhibited more robust expression of CD122 on a per-cell basis.

In the nonpregnant uterus, we observed CD122⁺F4/80^lo cells that persisted throughout pregnancy (Fig. 4A, 4B). These cells could have represented monocytes/macrophages or F4/80⁺ uterine eosinophils. Our data supported that presumptive eosinophils, with high side scatter properties and low to no expression of CD64, are abundant within the bulk F4/80^lo population in the nongravid uterus but nearly disappear during gestation (data not shown), consistent with prior reports (33, 34). However, these cells were uniformly negative for CD122. F4/80^loCD122⁺ cells, in contrast, exhibited lower side scatter properties and were CD64⁺, consistent with monocytes. These data reinforce a model in which monocytes and macrophages are the only uterine myeloid cells that express surface CD122. Further, these data provide evidence for a model in which populations of CD122⁺ monocytes and macrophages are dynamically regulated in a tissue layer-specific fashion in the prepregnant and gravid uterus.

Our data suggested that the uterus strongly favored development of CD122+Macs. Further, the drivers of the CD122+Mac fate appeared transiently strongest just after midgestation in the myometrium and persistently strong from midgestation through near term in the deciduo-placental unit. Total IFN activity at the mouse maternal–fetal interface precisely mirrors the population dynamics of CD122+Macs (35). The nonpregnant uterus transcribes type I IFN and IFN-stimulated genes (ISGs) (36), and type II IFN is apparent during the estrous phase (37). In the pregnant myometrium, total IFN activity is low until E10, peaks between E11 and E15, then returns to low levels after E15 (35). IFN activity is already robust in the developing placenta at E10 but spikes dramatically after E10 through to term.

To understand whether IFNs played a role in development of uterine CD122+Macs, we first compared the transcriptomes of decidual cMacs and CD122+Macs during early gestation, when both cell types were present and in similar abundance. Review of individual transcripts and GO analysis of transcripts significantly upregulated in CD122+Macs relative to cMacs showed strong enrichment of genes associated with IFN signaling and antiviral responses (Fig. 5A, Supplemental Table II). With this information, we next sought to determine whether IFNs were sufficient to drive expression of CD122 on decidual macrophages. F4/80⁺ cells in bulk single-cell suspensions of early gestation decidual capsules exhibited a dose-dependent upregulation of surface CD122 upon exposure to rIFN-α, although this effect was strongest in F4/80^lo cells (Fig. 5B and data not shown). To extend these findings, we cultured sort-purified uterine F4/80^lo monocytes in M-CSF with or without type I IFN-α. Consistent with the notion that culture in M-CSF results in endogenous production of type I IFN by macrophages (38), we did observe some expression of CD122 on uterine monocytes cultured in M-CSF alone (Fig. 5B, 5C). However, addition of IFN-α resulted in robust upregulation of CD122⁺ compared with culture in M-CSF alone.

Further, we could derive CD122+Macs in vitro from nonadherent BM cells. Traditional BMDMs cultured with M-CSF upregulated CD122 in a dose-dependent fashion in response to both type I IFN-α and type II IFN-γ (Fig. 5D, 5E). Although the ability to drive expression of CD122 on BMDMs was shared by both type I and II IFNs, we did observe differential effects of type I and II IFNs on expression of MHCII, with IFN-γ driving expression of MHCII on BMDMs in a dose-dependent manner (data not shown), in agreement with prior reports (39). Consistent with endogenous production of type I IFN by M-CSF–stimulated macrophages (38), blockade of the type I IFNR (IFNAR) nearly abolished expression of CD122 by BMDMs cultured in M-CSF alone (Fig. 5D, 5E). Blockade of the IFN-γR (IFNGR), however, had no effect on level of CD122 in BMDMs cultured in M-CSF alone, suggesting that BMDMs do not produce endogenous IFN-γ. Use of combined type I/type II IFNR KO BMDMs (Ifnar^−/−Ifngr^−/− double KO [DKO]) showed that DKO BMDMs appropriately do not upregulate CD122 with IFN treatment (Fig. 5F). However, we detected modest but nonzero levels of CD122 on DKO BMDMs by virtue of culture in M-CSF. Thus, IFNs enhance expression of CD122, but they are not strictly required for expression of CD122 on BMDMs.

To determine whether expression of CD122 by BMDMs in response to IFNs was a phenomenon universal to all macrophages, we generated GM-BMDMs by culturing nonadherent BM in medium supplemented with high-dose GM-CSF instead of M-CSF (40, 41). Consistent with prior reports (40), GM-BMDMs upregulated CD11c and MHCII to a greater extent than traditional BMDMs (data not shown). Also in contrast to traditional BMDMs, GM-BMDMs produce less endogenous type I IFN (41). Indeed, we observed that GM-BMDMs exhibited little to no expression of CD122 relative to traditional BMDMs (Fig. 5F). Although traditional BMDMs upregulated CD122 robustly in response to type I IFN, GM-BMDMs did not change expression of CD122 in response to exogenous type I IFN. Taken together, these data support a model in which M-CSF primes BM monocytes to develop into macrophages capable of upregulating CD122 in response to IFNs.

We next assessed whether responsivity to type I and/or II IFNs was required for development of uterine CD122+Macs in vivo. As CD122+Macs could derive from monocytes, we chose to first determine requirements for type I and II IFNs in generation of CD122+Macs with adoptive transfer of BM monocytes. Although recovery of adoptively transferred cells from mice not treated with irradiation or chemotherapy is low, this approach allows for testing of cell-intrinsic requirements for type I and II IFNs in development of uterine CD122+Macs in normal pregnancy. As discussed above, both type I and II IFNs are produced at the maternal–fetal interface (35–37). Recently, type III IFN-λ was shown to play a critical role in fetal protection against infection with Zika virus in mice and humans (42, 43). Thus, all three known families of IFN are actively produced at the maternal–fetal interface. Combined with our data that either type I or II IFN is sufficient to induce expression of CD122 on BMDMs in vitro, we hypothesized that there would be no specific IFN required for development of CD122+Macs. In other words, any class of IFN might be able to signal to uterine Macs to differentiate into CD122+Macs. Consistent with this hypothesis, donor-derived CD122+Macs were equally evident after transfer of BM monocytes from wild-type, type I IFNR–deficient (IFNAR KO), type II IFNR–deficient (IFNGR KO), and DKO mice (Fig. 6A, 6B). Further, we found similar abundance of uterine CD122+Macs in nonpregnant IFNAR KO, IFNGR KO, and DKO mice (Fig. 6C). Altogether, these data support that receptivity to type I and II IFNs, alone or in combination, is not required for BM monocytes to reach the uterus and develop into CD122+Macs.

Despite the observed expression of CD122 by CD122+Macs, if and how IL-15 signals to macrophages remained unknown. IL-15 has been shown to signal through CD122 and the γ_c, which results in phosphorylation of numerous downstream targets, including ERK (44). As numbers of native uterine CD122+Macs were limiting for use in biochemical and functional assays, we used CD122⁺BMDMs that we could generate by the millions in the following experiments. To test for phosphorylation of ERK in CD122+Macs exposed to IL-15, we first created CD122⁺BMDMs by culturing nonadherent BM cells in M-CSF plus IFN-α. Cytokines already present in culture, including exogenous IFN-α, were then removed by washing adherent cells thoroughly and adding fresh, cytokine-free medium. Washed CD122⁺BMDMs were then stimulated in the presence or absence of IL-15 precomplexed to the α-chain of the IL-15R (IL-15Rα), which most closely approximates true presentation of IL-15 in vivo and optimizes its activity (45). BMDMs stimulated with IFN express IL-15Rα (46), which can transduce IL-15 signals in the absence of CD122 in certain cell types (47). Use of a preassociated IL-15/IL-15Rα complex, rather than free IL-15, ensured that any effects seen by stimulating CD122⁺BMDMs with IL-15 would be mediated through CD122/γ_c and not IL-15Rα. After 30 min of stimulation, we observed robust phosphorylation of ERK1/2 in cells treated with the IL-15 complex (Fig. 7A, 7B). These data support that IL-15 signals through CD122 in macrophages using the ERK/MAPK cascade, similar to other IL-15–responsive cell types.

We next tested whether IL-15 could act on CD122+Macs to modulate production of cytokines. CD122⁺BMDMs stimulated with the TLR9 agonist CpG produced abundant TNF-α and IL-6 (Fig. 7C, 7D). Those CD122⁺BMDMs costimulated with CpG and IL-15 precomplexed to IL-15Rα exhibited enhanced production of TNF-α and IL-6 (Fig. 7C, 7D), but we did not observe changes in levels of IL-10 (data not shown). Finally, consistent with our data showing that native uterine CD122+Macs did not express NK lineage markers by flow cytometry or transcriptionally (Supplemental Tables I, II), CD122⁺BMDMs did not acquire NK cell markers, such as NKp46, on CD122⁺BMDMs (data not shown). Overall, these data support that CD122+Macs are novel cellular targets for IL-15.

Macrophages are critical for successful pregnancy. Severe abnormalities of pregnancy are found in osteopetrotic (op/op) mice that lack functional M-CSF and exhibit substantially reduced uterine macrophages (48). Implantation of the embryo is compromised and litter sizes are small relative to M-CSF–sufficient mice. The op/op dams cannot be mated with op/op sires, as the complete absence of M-CSF is incompatible with gestation of live litters. These data complement a later study showing that inducible deletion of macrophages early in pregnancy causes complete fetal loss, because macrophages sustain the vasculature of ovarian corpus luteum, required to produce progesterone and maintain early pregnancy (49).

Macrophages are abundant at the maternal–fetal interface and have been shown to directly abut blood vessels and NK cells (50). Fetal trophoblasts of the developing placenta are thought to communicate bidirectionally with macrophages, but few studies have formally addressed this (51, 52). These findings suggest that uterine and placental macrophages serve numerous critical functions in pregnancy but are likely heterogenous. Several groups have investigated macrophage heterogeneity during pregnancy in mice and humans (23, 53), but subsets of uterine macrophages remain incompletely described. In this work, we address a critical unmet need to gain a deeper understanding of signals that drive phenotype and function of macrophages at the maternal–fetal interface.

We found novel and unexpected uterine macrophages that express CD122 and respond to IL-15 under direction of M-CSF and IFNs. Although IL-2 also signals to cells expressing CD122 and γ_c, IL-2 is absent from the maternal–fetal interface in the steady-state (8). These data suggest that CD122+Macs respond specifically to IL-15 during steady-state pregnancy. GO analysis confirmed that the transcriptome of CD122+Macs is enriched in ISGs. IFNs are classically produced in response to viral infections, but IFNs are abundant in the cycling uterus and at the maternal–fetal interface in the steady-state. Type I and II IFNs were sufficient to drive expression of CD122 on uterine and BM-derived monocytes and BMDMs. However, neither type I nor type II IFN was required, alone or in combination, to generate CD122+Macs in the uterus of KO mice or from BM monocytes adoptively transferred into pregnant hosts. Further, M-CSF alone could drive modest expression of CD122 on cultured BMDMs independently of IFN. These data suggest that a combination of factors, likely including some we have yet to investigate, promote development of CD122+Macs in the uterus. We also acknowledge the possibility that type III IFN could play a role in shaping the fate of CD122+Macs, although this remains to be tested. Type III IFN is the most recently described IFN, and extremely limited information is available about its presence at the maternal–fetal interface (43). Like type I and II IFN, type III IFN signals through STAT1 and activates networks of canonical ISGs (54, 55). Of note, type III IFN appears to activate a network of genes most similar to that of type I IFNs (56). Future investigations are needed to address the effects of type III IFN on uterine macrophages.

Although both type I and type II IFNs are capable of directing macrophages to upregulate CD122 in vitro, it remains to be determined which IFNs do so in vivo. It is possible that several different IFNs at several different timepoints in pregnancy act on macrophages, as observed developmental kinetics of CD122+Macs mirrored that of total IFN activity in the uterus and placenta (35). Mouse and human uterine glandular epithelial cells are an established reservoir of type I IFN (57, 58). For example, the type I IFNε is produced by uterine epithelium under direction of estrogen (58). Although a population of CD122+Macs exists in the nonpregnant uterus of IFNAR KO mice under direction of M-CSF and a non–type I IFN and/or additional signal, IFN-ε may still contribute to development of CD122+Macs in the normal nonpregnant uterus.

Regulation of IFN-γ during murine pregnancy has been thoroughly described (59). Levels progressively rise to a peak at midgestation, after which levels halve but plateau to near term. The cells upon which we performed gene expression profiling were obtained on E7.5, when levels of uterine IFN-γ are known to be relatively low. Consistent with those data, gene expression and GO analyses of CD122+Macs appeared most consistent with a type I or III IFN signature. The transcriptome of CD122+Macs was dominated by ISGs, including Mx1, that are induced preferentially in response to IFN-I and -III (54, 60). Further, CD122+Macs are largely MHCII^low in vivo. Type II IFN has been shown to enhance expression of MHCII in APC, whereas IFN-I has been shown to oppose IFN-II-mediated upregulation of MHCII (39). We also showed that treatment of BMDMs with IFN-γ in vitro led to upregulation of CD122 and MHCII, whereas treatment of BMDMs with IFN-α led to upregulation of CD122 but not MHCII. The IFN system is complex and redundant, particularly the type I and type III families of IFNs, each composed of several members. This makes isolating individual actors in vivo difficult. Future work will be directed at dissecting this complex system to determine the in vivo IFN requirements for generation of CD122+Macs during pregnancy.

We observed that only macrophages, not other uterine myeloid cells, express CD122. Further, only CD122+Macs, not CD122⁺ NK cells, experienced a boost in expression of CD122 on a per-cell basis that correlated with increased IFN activity during gestation. One interpretation is that only macrophages are in close enough proximity to the local source of IFN to express CD122 under direction of IFNs. An alternative explanation is that upregulation of CD122 in response to IFN is a mechanism unique to monocytes and macrophages. In support of this latter hypothesis, we provided evidence in vitro that M-CSF–derived macrophages, not GM-CSF–derived macrophages, responded to IFN by upregulating CD122. These data point to unique regulation of Il2rb in macrophages, allowing it to be expressed rapidly upon signaling by IFNs. Relatively little is known about transcriptional regulation of Il2rb (61, 62), and further investigations into why macrophages uniquely respond to IFN by upregulating CD122 are needed. This is especially true in light of the fact that CD122+Macs express neither Eomes nor T-bet, which have been previously shown to drive expression of Il2rb in killer lymphocytes (63). In summary, our data support the notion that CD122 is a novel, macrophage-specific, ISG.

Our data show that CD122+Macs can derive from BM monocytes, but the precise precursors of CD122+Macs have yet to be determined. We found evidence of CD122⁺F4/80^lo cells in the nonpregnant uterus that likely represent CD122⁺ monocytes. These cells may represent an intermediate through which CD122+Macs transit during gestation. It has long been appreciated that M-CSF is critical for pregnancy and is positively regulated by sex hormones (48, 64, 65). M-CSF concentrations are at or below the limits of detection in the uterus of nonpregnant mice but increase dramatically in the myometrium during gestation, reaching peak concentration at term (29, 64, 65). Indeed, we showed that large, F4/80^hi macrophages were absent from the nonpregnant uterus. The appearance of these cells in the pregnant uterus suggests they may derive from uterine monocytes exposed to M-CSF. In contrast to the myometrium, concentration of M-CSF is modest and constant in the decidua and placenta as gestation progresses (64). Reinforcing these data, our data and additional work showed maintenance of myeloid cells in the growing myometrium and decline of myeloid populations in the decidua as gestation progresses (29, 32). These data are consistent with our observations that BM monocytes adoptively transferred into pregnant recipients were recovered as F4/80^hi cells mainly from the myometrium late in gestation. Altogether, these data are consistent with a model in which BM monocytes are directed to develop into CD122+Macs by M-CSF and IFNs.

Finally, we found that CD122+Macs exhibit a biochemical and functional response to stimulation with IL-15. Whether CD122+Macs respond to IL-15 in the same manner as classical IL-15–responsive killer lymphocytes, however, remains to be determined. It has long been appreciated that IL-15 enhances cytotoxicity of NK cells (66), and we found it intriguing that CD122+Macs expressed a number of cytolytic transcripts. IL-15 may be responsible for the expression of granzymes and perforin observed in CD122+Macs, as IL-15 has been implicated in driving expression of cytolytic mediators in human plasma cells (67), another nontraditional responder to IL-15.

We have yet to understand the specific stimuli responsible for activating uterine and placental macrophages in vivo during pregnancy, but it is feasible that CD122+Macs respond to IL-15 by enhancing production of proinflammatory cytokines that may favor or threaten a healthy pregnancy. For instance, decidual CD122+Macs were enriched in Il18 transcript (encoding IL-18), which is known to support production of IFN-γ at the maternal–fetal interface. Establishing this link may have major clinical relevance. Mice deficient in IFN-γ signaling fail to remodel uterine spiral arteries into low-resistance, high-capacitance vessels (12). In the mouse, this has been attributed to defective production of IFN-γ by NK cells. Further, IL-12 and IL-18 appear to play a key role in production of IFN-γ at the maternal–fetal interface, as mice deficient in either or both of these cytokines exhibit similarly abnormal uterine artery remodeling as does the IFN-γ KO mouse (68). Thus, these mice develop the correlate of human preeclampsia (12, 69). At the same time, hyper-stimulation of CD122+Macs with an over-abundance of IL-15 may contribute to similar adverse outcomes of pregnancy, as seen in a mouse model of spontaneous preeclampsia characterized by an abundance of IL-15 and a relative paucity of NK cells (70). Investigations into the cytokine secretome of CD122+Macs, as well as into other IL-15–dependent and IL-15–independent functions of CD122+Macs in pregnancy, such as phagocytic activity, are ongoing and may shed light on new therapies for preeclampsia.

Lending validation to our findings in the mouse, we discovered that the human uterus contains a population of CD122+Macs. We observed that these cells phenocopied mouse CD122+Macs, with expression of CD122 and modest levels of NK cell markers, such as CD56, perforin, and granzyme. hCD122+Macs were not as abundant as mouse CD122+Macs early in gestation, but we have yet to fully explore how this population changes over gestation in humans. We also have yet to formally determine whether these human womb-associated, killer-like macrophages develop, signal, and function like their murine counterparts. Prior reports suggest that activation of cultured human monocytes in vitro with high-dose IFN-γ or LPS drive expression of CD122 (71, 72). Our findings greatly extend and provide critical context for these prior data. How IFNs signal to human monocytes and macrophages is a key area of ongoing investigation with clear implications for human pregnancy.

In summary, we have revealed that M-CSF and IFNs act on uterine macrophages to generate an entirely new and unexpected IL-15-responsive cell type at the maternal–fetal interface, the CD122+Macs. Given the importance of IL-15 in pregnancy, modulation of CD122+Macs may represent a novel therapeutic target to support pregnancies threatened by fetal growth restriction, fetal loss, and preeclampsia.

The authors have no financial conflicts of interest.