The eNAMPT/TLR4 inflammatory cascade drives the severity of intra-amniotic inflammation in pregnancy and predicts infant outcomes

Introduction: Intra-amniotic inflammation (IAI) or chorioamnionitis is a common complication of pregnancy producing significant maternal morbidity/mortality, premature birth and neonatal risk of chronic lung diseases such as bronchopulmonary dysplasia (BPD). We examined eNAMPT (extracellular nicotinamide phosphoribosyltransferase), a critical inflammatory DAMP and TLR4 ligand, as a potential therapeutic target to reduce IAI severity and improve adverse fetal/neonatal outcomes. Methods: Blood/tissue samples were examined in: 1) women with histologically-proven chorioamnionitis, 2) very low birth weight (VLBW) neonates, and 3) a preclinical murine pregnancy model of IAI. Groups of pregnant IAI-exposed mice and pups were treated with an eNAMPT-neutralizing mAb. Results: Human placentas from women with histologically-proven chorioamnionitis exhibited dramatic NAMPT expression compared to placentas without chorioamnionitis. Increased NAMPT expression in whole blood from VLBW neonates (day 5) significantly predicted BPD development. Compared to untreated LPS-challenged murine dams (gestational day 15), pups born to eNAMPT mAb-treated dams (gestational days 15/16) exhibited a > 3-fold improved survival, reduced neonate lung eNAMPT/cytokine levels, and reduced development and severity of BPD and pulmonary hypertension (PH) following postnatal exposure to 100% hyperoxia days 1–14. Genome-wide gene expression studies of maternal uterine and neonatal cardiac tissues corroborated eNAMPT mAb-induced reductions in inflammatory pathway genes. Discussion: The eNAMPT/TLR4 inflammatory pathway is a highly druggable contributor to IAI pathobiology during pregnancy with the eNAMPT-neutralizing mAb a novel therapeutic strategy to decrease premature delivery and improve short- and long-term neonatal outcomes. eNAMPT blood expression is a potential biomarker for early prediction of chronic lung disease among premature neonates.

Immunostaining analysis. Digital images obtained from Confocal software were exported to ImageJ. The fluorescence intensity associated with each pixel was determined in sections 750 × 750 μm that included 4 sections per animal and 5 animals per group. Excitation and acquisition parameters were adjusted to fully eliminate pixel saturation, and all images were collected under identical settings.

Murine model of intrauterine inflammation.
All mice were housed under standard conditions (12h light-dark cycle, 25-27°C, ~40% humidity) in autoclaved micro-isolator cages with free access to food and water throughout the duration of the experiments. All animal care procedures and experiments were approved by the Institutional Animal Care and Use Committee (University of Arizona).
Timed-pregnant C57BL6 mice (aged 8-10 weeks) with the vaginal plug were documented as day 0 of gestation and confirmed by serial physical examinations and abdominal ultrasound. LPS (E. Coli , 055:B5, Sigma, 50 ug/mouse) was administered IP to pregnant mice at day 15 of gestation to produce high-grade placental inflammation (43), intrauterine inflammation, maternal systemic inflammatory response, a well-established model of clinical intrauterine inflammation (44, 45) that results in severe IAI 80-90% of dams and moderate to high rates of fetal loss (46) with premature abortion within 24-48 hrs (44,45,47). Treated dams group received two doses of eNAMPT mAb (10ug/mouse IP) on day GD15 and GD16 S9upp. Fig. 1).
Ultrasound evaluation of fetal viability. The initial abdominal ultrasound was performed prior to LPS challenge to define the number of sacs in each horn of each dam. The second abdominal ultrasound was performed 48 hr after LPS administration and was designed to determine the number of total fetuses (dead or alive) in each dam. After birth, the number of surviving pups in each group was recorded to determine the survival rate.
Western blotting for eNAMPT protein expression. Snap-frozen lung tissues were homogenized in RIPA buffer (50 mmol/L Tris-HCl pH 7.4, 150 mmol/L NaCl, 0.5 % sodium deoxycholate, 0.1 % SDS, 1% NP-40, 5 mmol/L EDTA) supplemented with complete protease/phosphatase inhibitor cocktail (Cell Signaling Cat #5872S) using tissue grinder with glass pestles (VWR Cat #26307-606). After centrifugation (15,000 g for 20 min at 4°C), protein concentration of homogenates was determined by Bio-Rad DC protein assay (cat #5000112). Following incubation 5 min at 90°C in loading buffer, aliquots containing equal amounts of protein (25-30 ug) were subjected to sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE). Subsequently, proteins were transferred to PVDF membranes and probed with specific primary, then by secondary antibodies. Proteins were visualized using an ECL system (Pierce West Pico cat #34580) and ChemiDoc MP imaging system (Bio-Rad). Densitometric analysis was performed using Bio-Rad Image Lab 6.01. Assessment of murine BPD. As shown before (48), two hit model, with prenatal inflammation and postnatal hyperoxia produced a greater degree of lung injury, with significantly enlarged alveoli, pulmonary fibrosis, and macrophage infiltrate, than either a prenatal or postnatal insult alone. In the model of intraamniotic inflammation described above, surviving pups were exposed to hyperoxia FiO2 85% for 14 days, then housed on room air. Exposed groups were studied for long term outcome. To avoid O2 toxicity in the dams and to eliminate maternal effects between the groups, the nursing dams were rotated between their hyperoxic and room air litters every 24 h. In a subgroup of pups born to treated dams, an additional eNAMPT mAb dose was administered IP after hyperoxia exposure on day 14-15 of life (one dose of eNAMPT mAB, 10ug/mouse IP).

Assessment of murine pulmonary hypertension.
Neonatal lungs were collected after 3 weeks of age for lung tissue morphology, fixed in 10% Neutral Buffered formalin, embedded in paraffin, sectioned, mounted onto slides, and finally stained with hematoxylin-eosin (H&E) (40). Neonatal lungs were also examined by western blotting for eNAMPT expression as an index of inflammatory process, for CD31 as an index of angiogenesis, and for SNAIL1, as an index of PH. RT-PCR was also performed for p-SMA (2, 88-90); eNOS and STAT3 as marker of PH.
Hemodynamic measurements were evaluated 1-2 weeks post hyperoxia exposure as we have described (49) utilizing a Millar catheter inserted in right jugular vein and into the right ventricle. Right ventricular systolic pressure (RVSP) was measured and recorded using a computerized hemodynamic recording system (HAEMODYN, Harvard Apparatus, MA, USA). RV/S + LV ratio was expressed as a ratio of the weight of the right ventricle to that of the septum plus left ventricle (RV/S + LV), (30, 50).
Quantification of micro-vessel density and vascular wall thickness. The left lung was perfused with 4% paraformaldehyde (PFA), inflated by infusion of 4% PFA at a constant pressure of 25 cm H2O through a cannula inserted in the trachea, fixed in 4% PFA overnight at 4°C and then embedded in OCT, Subsequently 5-μm-thick sections were taken and stained with hematoxylin and eosin (H&E) Images of individual pulmonary arteries were captured using a digital camera, mounted on a light microscope, and linked to a computer. Microvessel density was quantified by counting the percentage area of positive pixels per image with at least 21 images per sample (5 animals, 3 samples per animal, and 7 sections per sample as described previously) (49).

Vascular wall thickness. Vascular wall thickness was measured using a Zeiss Axiovert 200M
light microscope -CCD camera AxioCam (mRm) color camera and expressed as the percentage of total vessel size. Percent wall thickness was calculated as (2 x wall thickness)/external diameter. External diameter and internal diameter of 50 alveolar vessels (with an external diameter of 100-200 μm) per animal were determined and recorded by an independent investigator blinded to the treatment regimen. The ratio of vessel wall area to total area (WA%) and the ratio of pulmonary arteriole wall thickness to vascular external diameter (WT%) were measured using Zeiss axial program of 3 random wall sections.
Echocardiographic studies. Echocardiography was performed using a Vevo 3100 High Resolution Imaging System (Visual-Sonics, Toronto, Canada) with an MX550D (mouse) designed for rodent cardiac imaging as we have reported (33). Following anesthetic induction in 3% isoflurane, animals were placed in a supine position on a heated platform to maintain body temperature of 37°C. Anesthesia was maintained with 1.5-3% isoflurane (USP, Phoenix) in 100% oxygen and echo images collected and stored as digital cine loops for off-line calculations. Standard imaging planes, M-mode, Doppler, and functional calculations were obtained according to American Society of Echocardiography guidelines and as described in our previous studies (33).
Whole microarray data analysis in VLBW neonates. From a total of 33,252 genes per patient a total of 20,697 genes were included for analysis after filtering out genes with expression levels <50% of total expression. Outliers were weighted according to Ritchie et al (91) and genes were considered significant if the false discovery rates (FDRs), by Benjamini and Hochberg-adjusted P values, were less than 5%. Two NAMPT gene probes meeting the FDR threshold when comparing no BPD to BPD or severe BPD were included in the microarray data.
RNA sequencing of murine tissue samples. Mouse uterine and heart RNA was extracted and RNA QC assessed by RIN value, 28S/18S and fragment length distribution (Aligient 2100 Bio analyzer, Agilent RNA 6000 Nano Kit). Following library construct, RNA was sequenced using Illumina Hiseq (NovaSeq) PE150 platform; averagely generating 6 Gb raw data per sample. RNaseq data bioinformatic analyses pipeline included data quality control, calculation of Pearson correlations of all genes expressed to reflect the correlation of gene expression between samples. The HISAT2 and (Hierarchical Indexing for Spliced Alignments of Transcripts) Bowtie2 programs were utilized to align and clean reads to the reference genome and to the reference genes (12,92,93). Abundance and distribution of transcripts were assessed obtaining expected number of Fragments Per Kilobase er Millions base pairs (FPKM) (94). Correlation analysis to asses variation between samples was performed by Pearson correlation. DEseq2 algorithms were used to detect DEGs with Bioconductor software packages (95). To control for multiple testing error, the adjusted P-value False Discovery Rate (FDR) was utilized (54). Enrichment analysis for Gene ontology (GO) classification was performed focused on biological process and pathway classification with KEGG and Reactome sources (55).

Statistical analysis.
Continuous data were compared using nonparametric methods and categorical data by chi square test. Where applicable, standard one-way ANOVA was used and groups were compared using the Newman-Keuls test. Differences between groups were considered statistically significant when p values were less than 0.05 (p <0.05). Two-way ANOVA was used to compare the means of data from two or more different experimental groups. If significant differences were present by ANOVA (p <0.05), a least significant differences (LSD) test was performed post hoc. Between group differences were considered statistically significant when p <0.05. Statistical tests were performed using GraphPad Prism version 7.00 for Windows, (GraphPad Software, La Jolla California USA).
To determine whether the expression of NAMPT could predict BPD, we randomly split the human microarray data into 2/3 for training and 1/3 for testing. We fit a single gene logistic regression model with BPD as our dependent variable and NAMPT as our independent variable. The area under the receiver operating characteristic curve (AUC) was calculated, wherein 70% was deemed as clinically relevant. The same approach was utilized for the AUC of no BPD against severe BPD. Genomic and human cohort analyses were performed in R v.4.1.0 using the packages GEOquery, limma, DESeq2, pROC, ggpubr, and gt summary.