Aberrant Endometrial Features of Pregnancy in Diabetic NOD Mice

NOD/LtJ, NOD/LtSz-scid.scid (NOD.scid), C56Bl/6J (B6), and FVB/NJ (FVB) male and female mice were purchased from The Jackson Laboratory (Bar Harbor, ME). C57BL/6N (B6) mice and BALB/c mice were purchased from Taconic Farms (Germantown, NY). NOD and NOD.scid mice were housed in a barrier facility; B6 and FVB mice were housed conventionally. From 9 weeks of age, NOD mice were monitored weekly for hyperglycemia using tail venipuncture and OneTouch Ultra Glucose meter and strips (LifeScan, Burnaby, British Columbia, Canada). NOD mice were classified as normoglycemic (<11 mmol/l), pre-diabetic (11.1–14.9 mmol/l), and diabetic (≥15 mmol/l on sequential measurements). Experimental design precluded developing a diabetes incidence curve for females in our NOD colony. The diabetes incidence for males was >50% at 30 weeks of age (n = 8), consistent with published work (1).

Female NOD and NOD.scid mice with known blood glucose were paired with normoglycemic NOD males to obtain pregnancies. Copulation plug detection was named gd0. Gd-matched pregnancies in syngeneically mated B6, NOD.scid, FVB, and BALB/c mice were used as controls. Mice were killed at gd6 and gd8 for immunohistochemistry, gd7 for adhesion assays, or gd10 for morphometric analyses and quantification of IFN-γ. Peripartum placentas and offspring were weighed.

Avertin-anesthetized mice were killed by cervical dislocation. For histology, pancreas and uterus were dissected and fixed in fresh 4% neutral buffered paraformaldehyde (Sigma-Aldrich, Oakville, Ontario, Canada). For adhesion assays, these tissues plus multiple peripheral lymph nodes (cervical, submandibular, axial, mesenteric, iliac, superficial inguinal, and popliteal) were dissected, embedded in cryomatrix (Shandon Cryomatrix, Thermo Fisher Scientific, Waltham, MA), and snap frozen in liquid nitrogen. All animal usage complied with protocols approved by Queen’s University’s Animal Care Committee.

Paraformaldehyde-fixed implantation sites were processed using an automatic tissue processor (Triangle Biomedical Sciences, Durham, NC) and paraffin embedded using standard methods (28). For each group, three implantation sites from three dams were serial sectioned at 7 μm. At least 77 sections were cut at the center of each implantation site and mounted to glass slides. Alternate slides were stained with hematoxylin and eosin for vessel morphometry or periodic acid Schiff reagent for uNK cell enumeration. Eleven sections, at least 42 μm apart (to avoid duplicate enumeration of uNK cells), were scored per implantation site. UNK cells were enumerated within a standardized surface area in each of decidua basalis and mesometrial lymphoid aggregate of pregnancy (MLAp), a transient lymphoid structure between the circular and longitudinal smooth muscle coats of the uterus. Spiral artery wall and lumen diameters were measured and expressed as wall–to–lumen diameter ratios. When present in the tissue, three spiral artery cross-sections were measured on each of the 11 sections from each implantation site. All analyses were performed using an AxioImager M.1 microscope and Axiovision software (Carl Zeiss, Oberkochen, Germany).

Indirect antibody staining was performed for peripheral lymph node addressin (PNAd), MAdCAM-1, and VCAM-1 on paraffin-embedded sections of pancreas and uterus from virgin, gd6, or gd8 normoglycemic NOD mice; diabetic NOD mice; and B6 mice. Three viable implantation sites and six representative sections per antibody were used, plus isotype controls. Sections were deparaffinized, followed by antigen retrieval for MAdCAM-1 and VCAM-1 (citrate buffer and proteinase K methods, respectively). Endogenous peroxidase was quenched with 0.3% H₂O₂ (30 min), followed by blocking with 10% normal rabbit serum in 1% BSA (1 h). Primary antibodies were incubated overnight at 4°C (biotinylated rat anti-mouse PNAd, 1:100, MECA-79 [Biolegend, San Diego, CA]; LEAF-purified rat anti-mouse MAdCAM-1, 1:100, MECA-367 [Biolegend]; rat anti-mouse VCAM-1, 1:20, MK-2 [Antigenix America, Huntington Station, NY]). MAdCAM-1 and VCAM-1 sections were incubated with rabbit anti-rat IgG biotinylated secondary antibody (1:100; Vector Labs, Burlingame, CA) (30 min). All sections were subjected to amplification with Vectastain Elite ABC kit (Vector Labs) and diaminobenzidine substrate for visualization.

Cryostat sections (12 μm) of substrate tissue (mouse lymph nodes, pancreas, and gd7 uterus) were cut immediately before use in the modified Stamper-Woodruff adhesion assay. Two sources of indicator lymphocytes were used: human peripheral blood lymphocytes or mouse splenocytes. Human blood in ACD anticoagulant (BD Biosciences, San Jose, CA) was provided by a healthy nulliparous noncycling adult female using depot-medroxyprogesterone acetate contraception. This donor had normal random and fasting blood glucose, insulin, and lipid profiles. Peripheral blood lymphocytes were isolated as previously described (24,29), adjusted to 5 × 10⁶ cells/μl in serum-free RPMI and incubated with anti-CD56 mAb (1:100; Coulter Immunology, Hialeah, FL).

Mouse splenocytes were dissociated mechanically, erythrocytes were lysed, and lymphocytes were harvested after centrifugation over Lympholyte-M (Cedarlane Laboratories, Burlington, Ontario, Canada). Aliquots of 5 × 10⁶ cells/100 μl in RPMI and labeled with either anti-CD49b mAb (DX5, 1:100; eBioscience, San Diego, CA) or CellTracker Blue CMAC (Molecular Probes; Invitrogen, Burlington, Ontario, Canada), a fluorescent intravital dye without effects on adhesive function (30).

To assess changes in endothelial cell function induced by diabetes, aliquots of PE-Cy5-CD56–labeled human peripheral blood lymphocytes were overlaid onto sections of normoglycemic or diabetic NOD pancreas, lymph nodes, or gd7 uterus or matched B6 tissues on a rotating table (112 rpm) and incubated (30 min, 4°C). Nonadherent cells were removed by dipping the slides into PBS; slides were then fixed (4% paraformaldehyde, 30 min), rinsed (PBS then dH₂O), and mounted. Fluorescent adherent cells were scored in 25 high-power fields (HPF)/specimen at ×400 magnification.

To assess changes in lymphocyte function with progression to diabetes, B6 substrate tissues were incubated with splenocytes from normoglycemic or diabetic NOD mice or B6 mice. The experiment was conducted as above.

Uteri were collected from virgin or gd10 normoglycemic, pre-diabetic, or diabetic NOD mice and from a gd10 BALB/c mouse. The MLAp and decidua basalis were dissected from healthy implantation sites; each was pooled by uterus in a microcentrifuge tube containing 100 μl RPMI and homogenized by Kontes micropestle (Fisher Scientific, Waltham, MA). Mesometrial tissue was dissected from virgin mice and handled similarly. After centrifugation (800g; 5 min), supernatants were collected and stored at −20°C until assayed for IFN-γ by enzyme-linked immunosorbent assay as previously described (31). The limit of detection of this assay was 0.01 IU IFN-γ.

Data were analyzed using the Prism 4.03 Statistical Software package (GraphPad, San Diego, CA) and are presented as means ± SE. Blood glucose values were analyzed by repeated-measures ANOVA, followed by both linear trend and Bonferroni’s test. Fetal and placental weights were analyzed by independent one-tailed t tests. Resorption rates, MLAp uNK counts, spiral artery lumen diameters, and adhesion assays were analyzed using ANOVA with Bonferroni’s post-test. The Kruskal-Wallis test was used to analyze uNK counts within the decidua basalis, IFN-γ quantification, and spiral artery wall–to–lumen ratios, followed by Dunn’s post-test. A P value <0.05 was considered statistically significant.

Pregnant mammals develop mild glucose intolerance at mid-gestation to elevate fetal glucose availability. In normal mice, blood glucose increases at gd4.5 (32). All pregnant NOD mice (normoglycemic, n = 5; pre-diabetic, n = 7; diabetic, n = 7) displayed significant increases in blood glucose from their individual baselines (three prior measurements; Fig. 1) (P < 0.005). Mean gd10 blood glucose values were significantly increased only in mice overtly diabetic (19.30 ± 4.352 vs. normoglycemic 6.885 ± 0.8947 and pre-diabetic 7.950 ± 1.698) before mating (P < 0.005).

The impact of diabetes on NOD fetuses was scored grossly (Table 1). Diabetic mice (n = 4) had significantly more resorption sites than normoglycemic (n = 3), pre-diabetic (n = 6), and NOD.scid (n = 3) mice at gd10 (P = 0.021). However, gross study underestimated losses in diabetic NOD mice, since additional resorptions were found histologically. Weights of gd18 fetuses and placentas from one overtly diabetic NOD dam (n = 1) were compared with fetal weights from FVB dams (n = 3). NOD placental weight was significantly greater than FVB placental weight (0.13 ± 0.0080 vs. 0.086 ± 0.0031 g, P < 0.0001). Although NOD placental weights were greater, morphometry revealed a smaller, more compact placental area than FVB placentas (data not shown).

Fetuses from the diabetic NOD dam were smaller than those from FVB dams (0.98 ± 0.065 vs. 1.22 ± 0.068 g, P < 0.05). FVB fetuses included one that was very underweight but had no gross abnormalities and another fetus with agnathia. In contrast, NOD fetuses had gross neural tube defects (spina bifida and exencephaly, n = 3) and growth restriction (Fig. 2). Further necropsy revealed additional defects including situs inversus, diaphragmatic hernia, agnathia, anopthalmia, cardiac hypertrophy, pancreatic hyperplasia, asplenia, and an occult spina bifida. Each NOD fetus had at least one defect, whereas most, as reported by others (10,11), had multiple defects.

In NOD.scid (n = 3) and normoglycemic NOD mice (n = 3), uNK cell numbers in the MLAp (site of most uNK cell proliferation) and decidua basalis (site dominated by postmitotic uNK cells) were consistent with published numbers in normal mice (33). Pre-diabetic (n = 3) and diabetic NOD mice (n = 3) had reduced uNK cell numbers in the MLAp (81.9 ± 1.18; 80.8 ± 1.02) compared with NOD.scid mice (87.5 ± 2.22) (P < 0.05, Fig. 3). Diabetic NOD mice had drastically reduced decidual uNK numbers compared with all other groups (26.1 ± 0.67) (P < 0.001). Decidual uNK cells were small and had limited cytoplasmic granularity, suggesting immaturity.

To further assess uNK cell function in NOD mouse implantation sites, IFN-γ was quantified in the MLAp and decidua basalis. gd10 BALB/c (n = 1) MLAp and decidua basalis were positive controls, and IFN-γ concentrations matched a previous report for immune-competent mice (9.04 and 12.92, respectively) (31). Mesometrial uterus from a virgin normoglycemic NOD (n = 1) mouse lacked IFN-γ, a finding typical of virgin uteri from normal mice. IFN-γ concentrations in decidua basalis and MLAp from normoglycemic (n = 3) and pre-diabetic NOD (n = 3) mice did not differ. However, IFN-γ concentration within the decidua basalis of diabetic NOD mice (n = 3) was significantly higher than in decidua basalis from normoglycemic NOD mice (34.01 ± 4.47 vs. 9.35 ± 3.74, P < 0.05). No difference in IFN-γ concentration within the MLAp was found between normoglycemic NOD, pre-diabetic NOD, and diabetic NOD mice.

Fewer spiral arteries were found in implantation sites from pre-diabetic (n = 3) and diabetic NOD (n = 3) mice than in normoglycemic NOD (n = 3) or NOD.scid mice (n = 3). Spiral artery lumen diameters shortened as blood glucose increased. For arteries in pre-diabetic (26.45 ± 0.62 μm) and diabetic dams (23.36 ± 0.57 μm), values were statistically different compared with normoglycemic NOD (36.82 ± 1.04 μm) or NOD.scid mice (34.43 ± 0.81 μm) (P < 0.001) (Fig. 4A). Spiral arterial wall-to-lumen diameter ratios for normoglycemic NOD mice and NOD.scid mice were greater than reported for B6 mice (34). This may indicate a basal arterial defect in the NOD strain. Spiral arteries from pre-diabetic (3.33 ± 0.092) and diabetic NOD mice (3.82 ± 0.11) had even greater wall-to-lumen ratios than normoglycemic NOD (2.39 ± 0.058) and NOD.scid mice (1.82 ± 0.044), suggesting progressive loss of capacity for spiral artery modification with increasing hyperglycemia (P < 0.01) (Fig. 4B).

To determine if decidua basalis of diabetic NOD mice expressed addressins aberrantly, PNAd, MAdCAM-1, and VCAM-1 were localized at gd6 and gd8. PNAd was not significantly detected in any implantation sites. MAdCAM-1 was localized to the lateral sinusoids in B6 (n = 3), while, in NOD mice (n = 3), reactivity was found in antimesometrial decidua. VCAM-1 was strongly expressed in the central decidua, mesometrial, and antimesometrial decidua of B6 mice (Fig. 5A). VCAM-1 was similarly localized in NOD mice, but staining intensity was weaker (Fig. 5B).

PNAd, MAdCAM-1, and VCAM-1 expression was absent from pancreatic islets from pregnant B6 mice (Fig. 5C). Moderate expression of PNAd, MAdCAM-1, and VCAM-1 was observed in islets from pregnant normoglycemic NOD mice. These molecules were strongly expressed in some of the islets remaining in pregnant diabetic NOD mice (Fig. 5D).

To assess endothelial homing receptor function, CD56-tagged indicator lymphocytes were evaluated for adhesion to various tissues. Adhesion to lymph nodes was used as a positive control and was similar for normoglycemic NOD (n = 8), diabetic NOD (n = 10), and B6 (n = 6) substrates (Fig. 6). Adhesion to diabetic NOD pancreas (5.28 ± 1.03 cells/HPF) was greater than to B6 pancreas (1.58 ± 0.19 cells/HPF) (P < 0.05). Low adhesion to B6 pancreas is consistent with published data. Adhesion to B6 uterus was localized to decidua basalis and low, concurring with previous reports for donors not at the ovulatory stage of their menstrual cycle (35). Adhesion to normoglycemic NOD uterus (13.9 ± 3.3 cells/HPF) was greater than to diabetic NOD (6.7 ± 0.50 cells/HPF) or B6 uterus (3.46 ± 0.54 cells/HPF) (P < 0.05). CD56⁺ cells showed less localized binding to decidua basalis in decidual sections from diabetic NOD mice: cells bound to uterine glands, antimesometrial decidua, and myometrium. Comparisons to controls used only cells bound to decidua basalis.

Fluorescently tagged splenic NK cells from diabetic NOD mice (n = 2) showed significantly more adhesion to normal B6 (n = 2) peripheral lymph nodes (6.28 ± 0.24 vs. 3.46 ± 0.38 cells/HPF) (P < 0.05) (Fig. 7A). Adhesion of splenic NK cells from NOD mice (diabetic NOD, 4.80 ± 0.0050 cells/HPF; normoglycemic NOD, 3.39 ± 0.43 cells/HPF) to B6 pancreatic endothelium, regardless of blood glucose, was increased over NK cells from B6 mice (1.84 ± 0.0040 cells/HPF) (P < 0.05). There were no differences between the three splenic NK cell populations in adhesion to gd7 B6 uterus.

In these experiments, total (CMAC-stained) splenocyte binding was enumerated on adjacent tissue sections to enable calculation of NK cell enrichment (Fig. 7B). About 10% of the splenocytes from B6 and normoglycemic NOD mice adhering to B6 lymph nodes and pancreas were DX5⁺ NK cells. In diabetic NOD mice, the proportional binding of NK cells increased to B6 lymph nodes (30.7 ± 4.27%) and was significant in pancreas (51.5 ± 18.1%) (P < 0.05). The proportion of DX5⁺ B6 splenocytes binding to B6 uterine endothelium was high (67.0 ± 0.88%). This proportion was lower for splenocytes from normoglycemic NOD mice (33.4 ± 9.12%) and was significantly different for diabetic NOD mice (27.1 ± 2.81%) (P < 0.05). This suggests that NK cells from diabetic mice are reduced in ability to engage with uterine endothelium for extravasation and that other cell types compete for these receptors. This may skew the immune cell populations at a diabetic maternal-fetal interface.

This study addressed whether reduced uterine homing and/or function of NK cells occurs in diabetic pregnancy that would impair pregnancy-associated spiral arterial modification and link endometrial lymphocytes to the pregnancy complications of diabetic women. Spontaneous disease in type 1 diabetic NOD mice was used as a model for gradual disease progression. Unexpectedly, defective spiral arterial development and modification were found in normoglycemic NOD and NOD.scid mice. This pathology is absent in implantation sites of pregnant scid mice, which lack T- and B-cells (36). NOD.scid mice have similar NK cell defects to NOD mice but do not develop insulitis or diabetes. Of particular note was the paucity of spiral arteries, which are major vessels, in NOD implantation sites. This has not been reported in other strains. Reduced neovascularization has been reported in NOD mice subjected to femoral artery ischemia. This was attributed to lower induction of vascular endothelial growth factor (VEGF) in ischemic NOD tissue (37). VEGF (38) and the VEGF family member placental growth factor (PlGF) (39) are major uNK cells products. In other tissues, VEGF regulates the expression of chemokine ligands that position recruited bone marrow–derived circulating cells (40) and promote flexible (i.e., not developmentally hard-wired) angiogenesis. Endometrium requires flexible angiogenesis, since the number of arriving blastocysts is variable between reproductive cycles. The vascular histopathology was more severe in pre-diabetic and diabetic NOD mice (5,41).

Progesterone, the essential hormone of pregnancy, and IFN-γ are among the regulators of VEGF transcription (42,43). Progesterone is absolutely required for terminal uNK cell differentiation, and uNK cells are the major source of mouse implantation site IFN-γ during the first half of gestation. Because spiral artery modification in mice is normally initiated by IFN-γ released from uNK cells (36,44), this cytokine was quantified in implantation sites. As in other strains, IFN-γ was not detected in virgin uterus but was present between gd6 and gd10 in NOD implantation sites. Levels were comparable to those in a normal mouse and consistent with our earlier published data (31), except in the decidua basalis of overtly diabetic gd10 females. This tissue, in which uNK cells are reduced, unexpectedly showed elevated IFN-γ. This finding suggests that the primary deficit in spiral artery development in NOD mice is not overcome by the influx of IFN-γ–producing uNK cells. The elevated IFN-γ that we saw may relate to the disease stage we studied. Rodacki et al. (45) showed, in type 1 diabetic patients, numbers and activation status of blood NK cells vary with disease stage. At onset, NK cells are reduced but unusually activated producers of IFN-γ. With longer-standing disease, both patients and NOD mice lose NK cell activation receptors (45,46). Because spontaneously diabetic NOD mice are fragile, usually older than mice used for first pregnancy studies, and prescribed for removal from breeding colonies, we bred our mice as soon as they became diabetic. This may have been in the interval of elevated NK cell activation. We attempted support of diabetic females by subcutaneously implanted slow release insulin rods before mating. This normalized blood glucose only until mating. By gd10, all insulin-treated females had reverted to hyperglycemia (n = 6; S.D.B., B.A.C., unpublished data).

Changes in patterns of vascular addressins are found in diabetic NOD mice (4,47), but effects of pregnancy on these molecules have not been addressed nor have comparisons of expression been made between gestational endometrium and pancreas. We found alterations in the normal pattern of MAdCAM-1 and VCAM-1 expression within the early gestational uterus of diabetic NOD mice. MAdCAM-1 was expressed in an aberrant location, and VCAM-1 expression was less intense. VCAM-1 is thought to be the key addressin in early uNK cell recruitment to deciduas basalis (48). The successful localization of many uNK cells in diabetic NOD pregnancy supports this postulate (48). Homing receptor elevation underlies insulitis, which is present in all adult NOD mice, regardless of diabetic state. Assays of lymphocyte adhesion using NOD pancreas showed that, in pregnancy, as in nonpregnant NOD mice, hyperglycemia elevates islet adhesion (49). The pattern of lymphocyte adhesion to decidualized uterus was less straightforward. There was a striking increase in the level of adhesion to uterus from normoglycemic NOD mice relative to controls. One interpretation of these data is that endometrial endothelium in gestational uterus compensates against pancreatic endothelial cell recruitment signals early in disease but cannot sustain this state as disease progresses. Adhesion assays using NOD splenocytes on B6 tissue showed diabetes-promoted gains in the ability of lymphocytes to adhere to pancreas and peripheral lymph nodes but not to uterus. This is consistent with the relatively normal numbers of uNK cells quantified within gd6–8 implantation sites. It also suggests there is a large reserve of circulating NK cells with tissue-homing potential enabling NK cell elevation in one tissue without decoy homing from another tissue. Previous studies of normal lymphocytes reported NK cell enrichment among cells adhering to gd7 uterus (26). Similar enrichment was not observed in assays using splenocytes from diabetic NOD mice. Splenic NK cells from diabetic NOD mice were enriched, however, by adhesion to pancreas, suggesting preferential homing to this organ over pregnant uterus.

In human diabetic pregnancy, placental and neonatal sizes are usually correlated; diabetic women frequently having abnormally large placentas with macrosomic infants. The diabetic NOD placentas were abnormally heavy, but fetal trophoblastic areas were greatly reduced relative to maternal decidua basalis in all pregnant NOD and NOD.scid mice. Human pregnancies complicated by both diabetes and intrauterine growth restriction or preeclampsia often display hypercellular placental histopathology accompanying small size. Human pregnancies complicated by diabetes have various vasculopathies, which tend to perturb spiral artery modification. Most frequently, long nonspiraling arteries persist that remain responsive to vasoactive substances. In pathological diabetic pregnancies (i.e., comorbid preeclampsia or intrauterine growth restriction), spiral arteries are incompletely modified, resulting in placenta hypoxia. Depending on the timing of the insult, placental insufficiency and/or hypoxia can result in growth restriction or abnormal organ development, intrauterine death, or fetally programmed risk for postnatal health impairment. Fetuses of overtly diabetic NOD dams showed early and late resorption, growth restriction, and a high rate of developmental defects. Our study shows there are major endometrial contributions associated with pathogenesis that include poor vascular bed development and altered endometrial lymphocyte recruitment and function, particularly of proangiogenic uNK cells. Thus, gestation in spontaneously diabetic NOD females appears to model intrauterine growth restriction and preeclampsia in human diabetic pregnancy. Links between NK cells and autoimmunity are beginning to be understood. Their relationships during physiological stress such as pregnancy require further study.

This work was supported by the Canadian Institutes for Health Research. B.A.C. is a Tier 1 Canada Research Chair.

We thank Drs. E. Leiter (The Jackson Laboratory, Bar Harbor, ME), M. van den Heuvel (University of Western Ontario, London, Ontario, Canada), C. McKerlie and Mr. C. Fleming (Mount Sinai Hospital, Toronto, Ontario, Canada) for invaluable advice, and M. Bilinski and R. Watering for technical support.