Adapting an organ-on-chip device to study the effect of fetal sex and maternal race/ethnicity on preterm birth related intraamniotic inflammation leading to fetal neuroinflammation

Abstract

Problem:

Fetal neuroinflammation has been linked to preterm birth-related intraamniotic infection and inflammation; However, the contribution of fetal sex and maternal race/ethnicity is unknown. To determine if fetal sex and maternal race/ethnicity influence neuroinflammation, an organ-on-chip (OOC) model were established under normal or pathologic conditions utilizing amniotic fluid.

Method of study:

OOC is composed of two-cell culture chambers connected by Type IV collagen-coated microchannels. Human fetal astroglia (SVGp12) and microglia (HMC3) were co-cultured at an 80:20 ratio in the inner chamber. The outer chamber contained amniotic fluid (AF) from male and female fetuses of White Hispanic (WH) and African-American (AA) pregnant women with or without lipopolysaccharide (LPS-100 ng/ml) and incubated for 48 h. Glial migration (brightfield microscopy), activation (Immunocytochemistry), and cytokine production (Luminex assays) were quantified and compared (N = 4 for each category of sex and race/ethnicity).

Results:

In a pooled analysis, AF+LPS did not induce glial activation or inflammatory changes compared to AF alone. When stratified by sex, male AF+LPS promoted significant glial activation (high CD11b:p < 0.05; low Iba1:p < 0.01) compared to male AF without LPS; however, this was not associated with changes in pro-inflammatory cytokines. When stratified by race/ethnicity, AF+LPS induced glial activation in both groups, but a differential increase in pro-inflammatory cytokines was seen between WH and AA AF (WH-interleukin-1β: p < 0.05; AA-interleukin-8: p < 0.01).

Conclusion:

This OOC model of fetal neuroinflammation has determined that race/ethnicity differences do exist for perinatal brain injury. The fetal sex of neonates was not a determining factor of susceptibility to intraamniotic inflammation leading to neuroinflammation.

1 |. INTRODUCTION

Preterm birth (PTB; delivery prior to 37 weeks’ gestation) affects 1 in 10 infants born in the United States.¹ Although significant research and advancement have been made that improve pregnancy outcomes, it is still the leading problem in modern obstetrics. Babies born prematurely also have an increased risk of developing intraventricular hemorrhage, cerebral palsy, and retinopathy of prematurity, all of which are disorders found mostly amongst preterm infants.² Other diseases found in this vulnerable population include necrotizing enterocolitis, bron-chopulmonary dysplasia, and cerebral white matter disease.² Many risk factors have been identified for PTB, including low socioeconomic status, Black race, Hispanic ethnicity, tobacco use, infection, and maternal medical conditions, such as hypertension,³ but most research has been related to infection.

Intraamniotic infections are associated with pathogens such as Group B Streptococcus, Escherichia coli, Mycoplasma hominis, and Ureaplasma spp.⁴ can stimulate an inflammatory response in the amnion and chorion, chorioamnionitis, which is a well-established risk factor for preterm birth.⁵ The combination of chorioamnionitis and preterm birth also increases the risk of neurodevelopmental disorders encountered in neonates, including white matter injury, grey matter injury, cerebral palsy, and autism^6,7 (Figure 1A). Polam et al. reviewed extremely preterm infants with chorioamnionitis and found that they had a higher incidence of intraventricular hemorrhage and retinopathy of prematurity compared to their counterparts.⁸

FIGURE 1.

Clinical relevance and project outline. (A) Schematic description of the fetal brain phenotype resulting from an intraamniotic inflammation that can lead to fetal neuroinflammation. (B) Schematic of the two-chamber OOC device showing the outer chamber, representing the intraamniotic cavity containing amniotic fluid with or without infectious stimuli, and the inner chamber, representative of the fetal brain containing astroglia and microglia. These two chambers are separated by arrays of microchannels filled with basement membrane collagen stained blue with Masson’s trichome stain. (C) Bright field microscopy images of SVGp12 and HMC3 cells morphology and cell-specific marker expression of Glial fibrillary acidic protein (GFAP) and Iba1. Green-cell specific stain and blue-nuclei. Polymerase chain reaction confirmed that both cell lines are male in origin by the presence of the SRY gene. These results were normalized by GAPDH. (D) Schematic showing the experimental workflow, biological variables, and endpoint assays used in the experimental design. A & D are created with BioRender program

Multiple animal models have evaluated the relationship between intraamniotic infection and neuroinflammation. In general, neuroinflammation in these models is characterized by microglial leading to cytokine release, maturation arrest of pre-oligodendrocytes, oligodendrocyte loss and dysfunction, and defective neuronal maturation and loss.⁶ Specifically, a mouse model by Burd et al. demonstrated that intrauterine inflammation leads to fetal brain injury and altered neuronal morphology in the setting of PTB.⁹ Kuypers et al., utilizing a sheep model, found that intrauterine infection with lipopolysaccharide led to inflammation and white matter and hippocampal injury of the fetus.¹⁰ At a cellular level, microglia and astrocytes have been shown to play critical roles in the development of neuroinflammatory disorders.^11–13 In vivo, it has been shown that activated glial cells increase their expression of activation marker cluster of differentiation receptors (CD)11b¹⁴ and decrease their expression of resting marker ionized calcium-binding adapter molecule 1 (Iba1).¹⁵ This change is associated with increased production of pro-inflammatory cytokines such as tumor necrosis factor-alpha,¹⁶ interleukin-1b (IL-1b),^17,18 IL-8,¹⁹ IL-6,²⁰ and interferon-gamma²¹ that disrupt cellular homeostasis and contribute to neuronal damage.²²

Two major factors that can contribute to differences in neuronal inflammation, is fetal sex and maternal race/ethnicity. Several epidemiologic, observational, and genetic have shown the contributions of these two variables to the onset of neuroinflammation; however, a model to test them in vitro as not yet emerged. The increasing rate of neuroinflammation, neonatal morbidity, and the autism spectrum of diseases following infection led us to postulate that fetal sex and maternal race/ethnicity may modify the risk of neuroinflammation. However, racial differences, for example, cannot be accounted for or adequately researched using animal models, creating a gap in the field. In vitro studies utilizing human cells to study the effects of intraamniotic infection and neuroinflammation are scarce. Additionally, 2D cell environments do not account for the dynamic environment of the human body, limiting the utility of current model systems.

Therefore, to determine if the risk of fetal neuroinflammation in-utero differs by fetal sex or maternal race/ethnicity, we utilized an ex-vivo two-chamber organ-on-chip (OOC) platform to model fetal brain exposure to healthy and pathologic amniotic fluid components. Using the OOC, we tested the effects of lipopolysaccharide (LPS) infected amniotic fluid from African-American (AA) and White Hispanic (WH), male and female, fetuses on cell migration, microglial activation, and production of pro-inflammatory cytokines.

2 |. METHODS

2.1 |. IRB approval

This study protocol is approved by the Institutional Review Board at The University of Texas Medical Branch (UTMB) at Galveston, TX, as an exempt protocol to use of discarded placenta after normal term cesarean deliveries (UTMB 11–251). Similarly, discarded amniotic fluid samples were collected at the time of the Cesarean section. No subject recruitment or consent was required for this study. The second set of bio-banked amniotic fluid samples was used to determine racial disparity in amniotic fluid cytokine levels.^23–25 The sample collections were approved by TriStar institutional review boards at Centennial Women’s Hospital Nashville, TN.

2.2 |. Clinical samples

Amniotic fluid samples were collected from term, not in labor, cesarean deliveries with no documented pregnancy complications. Amniotic fluid samples were collected before cesarean section by puncture of intact membranes using a 22-gauge needle prior to artificial rupture of the membranes. Amniotic fluid was centrifuged immediately for 10 min at 2000 × g to remove cellular and particulate matter. Aliquots of amniotic fluid were stored at −80°C until analysis. Inclusion Criteria: Amniotic fluid samples from male and female neonates were collected from self-identifying AA or WH women with normal term birth and no identified pregnancy-related complications. (N = 4 from each category [male AA, female AA, male WH, female WH]). Exclusion Criteria: Term labor vaginal deliveries (>390/7 weeks) were excluded. Subjects with multiple gestations, placenta previa, fetal anomalies, and/or medical or surgeries (intervention for clinical conditions that are not linked to pregnancy) during pregnancy were excluded. Severe cases of preeclampsia or persistent symptoms (headache, vision changes, right upper quadrant pain) or abnormal laboratory findings (thrombocytopenia, repeated abnormal liver function tests, creatinine doubling or >1.2, or HELLP syndrome) or clinical findings (pulmonary edema or eclampsia) were also excluded. Subjects who had any surgical procedures during pregnancy or who were treated for hypertension, preterm labor, or for suspected clinical chorioamnionitis (reports on foul-smelling vaginal discharge, high levels of CRP, fetal tachycardia), positive GBS screening or diagnosis of bacterial vaginosis, behavioral issues (cigarette smoking, drug or alcohol abuse) and delivered at term were excluded from the control groups. Race in the second cohort of amniotic fluid was determined by subject interviews and identifying the race/ethnicity back to three generations from both the maternal and paternal sides.^23,26,27 Failure to identify three generations resulted in the exclusion of subjects from the consenting process. Maternal neuroinflammation-related illnesses were not screened in these samples and cannot be ruled out as a causal factor of neuroinflammation.

2.3 |. Human fetal brain cell line cultures

The fetal brain cells used in this study were male SVG p12 (astroglia) obtained from ATCC (ATCC cat. no. CRL-8621, Manassas, VA) and male HMC3 (microglia) (ATCC cat. no. CRL-3304, Manassas, VA). These cells were obtained from the fetal brain of 8–12 week fetuses. Immortalized SVG p12 and HMC3 cells were stored in liquid nitrogen and thawed in a water bath for 3 min, and incubated for 48 h in 5% CO₂ at 37°C in a T25 flask containing ATCC-formulated Eagle’s Minimum Essential Medium (Catalog No. 30–2003) with 10% fetal bovine serum.

2.4 |. Microfluidic OOC designs

The microfabrication procedure is similar to that previously outlined for the amnion membrane OOC device²⁸ (Figure 1B). In short, to form the master mold, a 2-step photolithography process was conducted using photosensitive epoxy (SU-8; MicroChem, Westborough, MA, USA), forming the first microchannel layer (24 channels; 5 μm deep, 600 μm long, 30 μm wide) and the second cell culture chamber layer (500 μm deep). The master mold was then coated with (tridecafluoro-1,1,2,2-tetrahydrooctyl) trichlorosilane (United Chemical Technologies, Bristol, PA, USA) to facilitate polydimethylsiloxane (PDMS) release from the master mold after replication. A soft lithography technique was utilized to make the OOCs out of PDMS. PDMS devices were replicated from the master mold by pouring PDMS pre-polymer (10:1 mixture, Sylgard 184; DowDuPont, Midland, MI, USA) onto the mold, followed by curing at 85°C for 45–60 min. The reservoirs to hold the culture medium were punched out from this PDMS layer using a 4 mm diameter punch. This PDMS layer was treated with oxygen plasma (Harrick Plasma, Ithaca, NY, USA) for 90 seconds to improve the bonding of the PDMS layer onto the glass substrate and to also make the device hydrophilic for easy cell and culture medium loading, followed by bonding onto a glass substrate (22 × 22 mm). A single device (2 cm × 2 cm) fits within a well of a 6-well culture plate.

2.5 |. Collagen loading into the OOC

Before the OOCs were used, they were washed with 70% ethanol for 10 min to sterilize, then washed three times with 1X phosphate-buffered saline (PBS). The microchannels were then coated with Matrigel (Corning Matrigel Basement Membrane Matrix, DEV-free; 1:25 in KSFM) by loading Matrigel in the outer chambers and applying suction pressure to the inner chamber. The device was then incubated at 37°C in 5% CO₂ environment for 4 h. This mimics the basement membrane of the blood-brain interface on-chip (Figure 1B).

2.6 |. Cell seeding and treatment in the OOC

After a 4-hour incubation, the devices were washed three times with 1X PBS before cell seeding. SVG p12 and HMC3 were seeded in the OOC in 80:20 ratio of microglia to astroglia in the human brain. 6000 SVG p12 and 1500 HMC3 were seeded in the center chamber of the OOC. Cell culture media was placed in both the center and outer chambers. Seeding densities were determined in previous cell loading titrations. The OOCs were incubated at 37°C with 5% CO2 overnight. The culture media was changed for all center chambers. In the outer chambers, either culture media and amniotic fluid and or 100 ng/ml LPS²⁹-treated culture media and amniotic fluid (i.e., race/ethnicity and sex-specific samples) in a 2:1 ratio was used. Cigarette smoke extract (CSE 1:25), a potent and validated inducer of oxidative stress,^30–39 was used as a positive control to induce a pathologic phenotype. After 48 h, the center-chamber media was saved for further testing, and cells underwent endpoint staining (Figure 1C).

2.7 |. Microscopy

Brightfield microscopy images were captured using a Nikon Eclipse TS100 microscope (4×, 10×, 20×) (Nikon). Three regions of interest per condition were used to determine the overall cell morphology. A Keyence All-in-one Fluorescence BZ-X810 microscope, ×4, ×10, and ×40 magnification) was used to determine cell barrier formation, document collagen in microchannels, and cell-specific marker expressions.

2.8 |. Staining of cell-specific resting and activation markers

Immunocytochemical staining for Glial fibrillary acidic protein (GFAP) (1:500; ab68428; Abcam, Cambridge, MA, USA) ionized calcium-binding adaptor molecule 1 (Iba1) (1:100; ab221003; Abcam, Cambridge, MA, USA), and CD11b (1:100; ab197701; Abcam, Cambridge, MA, USA) was performed after 48 h, as previously described.^46,47 Manufacturers’ instructions were followed for determining appropriate dilutions of antibodies to ensure specific and uniform staining. After 48 h, cells were fixed with 4% paraformaldehyde, permeabilized with0.5% Triton X, and blocked with 3% bovine serum albumin in 1× PBS, before incubation with conjugated antibodies for 2 hours at room temperature in the dark. Secondary rabbit (1:1000) was added for 1 h for GFAP. The OOCs were washed with 1× PBS, and then also treated with NucBlue^® Fixed ReadyProbes Reagent (R37606; Thermo Fisher Scientific, Waltham, MA) to stain the nucleus.

2.9 |. Masson trichrome stain

OOCs were fixed in 4% paraformaldehyde for 15 min and stained using the Masson Trichrome method to identify collagen within the microchannels. Three microscopic fields were captured at 10×.

2.10 |. Multiplex assays for inflammatory cytokine markers analyses using Luminex

To analyze changes in inflammatory mediators’ interleukin (IL)-8 and IL-1β were analyzed from the cell supernatants in the OOC after treatments (MILLIPLEX Human Cytokine/Chemokine/Growth Factor Panel A, Cat#3923332, Millipore Sigma, Burlington, Massachusetts). Supernatants were manually collected from the reservoirs of the devices after 48 h of LPS exposure. Standard curves were developed with duplicate samples of known quantities of recombinant proteins that were provided by the manufacturer. Sample concentrations were determined by relating the fluorescence values that were obtained to the standard curve by linear regression analysis.

2.11 |. Statistical analyses

All data were analyzed using Prism 7 software (GraphPad Software, La Jolla, CA, USA). The Shapiro-Wilk test for normality was conducted to check for the normality of the data. The paired student’s t-test (e.g., AF vs. AF+LPS; or Male vs. Male+LPS) and unpaired student’s t-test (e.g., Male vs. Female) were used to comparing means between two groups for normally distributed data. Data are shown as mean ± SEM.

3 |. RESULTS

3.1 |. Establishing an OOC platform to study intraamniotic inflammation and fetal neuroinflammation

A previously established two-chamber organ-on-chip device²⁸ was utilized to develop a fetal neuroinflammation model (Figure 1A–B). The OOC device consists of an inner chamber and an outer chamber separated by microchannels (Figure 1B). The microchannels that separated the two chambers were coated in Type 4 collagen, which makes up a significant portion of the basement membrane found in the brain (Figure 1B). The inner chamber contained an 80:20 ratio of fetal-derived astroglia cell lines (SVGp12) and microglia (HMC3) with growth media. These cells maintained their cell-specific markers glial fibrillary acidic protein (GFAP) and Ionized calcium-binding adaptor molecule 1 (Iba1) in culture (Figure 1C). The male sex of these cells was confirmed by a polymerase chain reaction of the Y chromosome marker sex-determining region Y (SRY) gene (Figure 1C). The outer chamber served as a model for the amniotic cavity and contained either standard growth media or treatments modeling infectious or sterile inflammation (Figure 1D).

Prior to understanding the role of fetal sex and maternal race/ethnicity-specific amniotic fluid on fetal brain neuroinflammation, the OOC model was validated under control, infectious inflammation (LPS), and noninfectious inflammation due to oxidative stress (induced by cigarette smoke extract – CSE) conditions to develop a model of a pathologic phenotype. In this report, we define the pathologic phenotype as neuronal cell migration, glial activation, and an increase in inflammatory cytokines, as seen in-utero.⁶ To determine the number of cells that migrated from the inner to outer chamber through microchannels after 48 h, fluorescent images of the OOCs were quantified under each condition. Migratory cells were quantified as cells within the microchannel array or residents in the outer chamber. Neither LPS nor CSE treatment in the outer chamber affected astroglia cell migration after 48 h (Figure 2A). In addition to cell migration, microglial activation was tested, as it is a prominent marker of neuroinflammation in-utero.⁷ Microglia-specific markers, Cd11b (green) and Iba1 (red), were imaged to identify activated and resting HMC3, respectively. Oxidative stress (induced, by CSE^{30,32,34–36,39,40}), produced a pathologic phenotype characterized by an increase in microglial activation mark CD11b (green-white arrows) (p = 0.02) and a decrease in resting marker Iba1 (red) (p = 0.003) compared to control (Figure 2B). LPS alone in the OOC device did not produce an effect suggesting that amniotic fluid (as reported below) as a medium is required for its propagation and to cause a pathologic effect, or the time required for LPS to migrate through the microchannels coated with collagen is substantially higher than CSE diffusing through this layer. However, oxidative stress-induced inflammation, in the absence of an infection, could play a larger role in the development of glial-associated neuropathology than infectious inflammation alone. These data also suggest that the microchannels used in this report may perform a true barrier function.

FIGURE 2.

Validation of physiological and pathologic models to study fetal brain inflammation on-chip. (A) Fluorescent microscopic images of cell migration and glial cell activation within the OOC. Migrating cells were localized by blue nuclear stain and marked by pink (control), light blue (LPS), or yellow dots (CSE) outside the inner chamber (yellow dashed circle). Migratory cells were defined as cells that had migrated into the microchannel or resided in the outer chamber. Glial activation was determined by staining for activation marker Cd11b (green) and resting marker Iba1 (red). White arrows highlight Cd11b positive cells. Nuclei are stained blue with DAPI. Representative images are shown for n = 3. (B) Bar graphs representing the number of migrated cells, number of activated microglia (CD11b positive), and Iba1 expression intensity in each condition. CSE induced a significant amount of glial activation as shown by a high quantify of CD11b positive cells (p = 0.02) that also contained low Iba1 expression (Control vs. CSE: p = 0.003; CSE vs. LPS: p = 0.002). Graphs show mean ± SEM values and contain three replicates

3.2 |. Modeling the impact of fetal sex and maternal race/ethnicity during intraamniotic inflammation

To recreate the intraamniotic environment within the OOC device, amniotic fluid previously collected from non-laboring cesarean sections from both male and female fetuses of AA and WH patients was used in the experiments. To do this, OOCs were loaded with cells as described above, and the outer chamber was filled with normal growth media (control), amniotic fluid from either male or female fetuses from either maternal race/ethnicity, either spiked with or without LPS. End-points such as cell migration, glial activation, and inflammation were monitored to determine the development of a pathologic phenotype (Figure 1C).

Results from a pooled analysis (regardless of fetal sex and maternal race/ethnicity) showed a significant difference in cell migration (p = 0.04) (Figure 3A) but not glial activation (number of CD11b positive cells) (Figure 3B) between amniotic fluid and amniotic fluid+LPS groups after 48 h. Compared to amniotic fluid alone, the addition of LPS caused a decrease in glial resting marker Iba1 (p = 0.02). However, that did not result in glial activation nor the production of pro-inflammatory cytokines IL-1β and IL-8 (Figure 3C).

FIGURE 3.

Amniotic fluid containing LPS does not induce a pathologic phenotype. (A) Fluorescence microscopy images showing migratory cells outside of the inner chamber as yellow dots. The clustering of data (AF or AF+LPS co-treatment) showed a difference between the number of migrated cells (p = 0.04). Graphs show mean ± SEM values (n = 16). (B) Fluorescence microscopy images showing glial cells expressing resting marker Iba1 (red) and activation marker CD11b (green) under AF and AF+LPS conditions. Data showed no differences between the number of activated glial cells (CD11b positive cells), but there was a significant downregulation of Iba1 (p = 0.02). Data are shown as mean ± SEM values(n = 16). (C) Multiplex cytokine analysis showed no differences between AF and AF+LPS in IL-1β and IL-8 concentrations in media samples from the inner chamber. Data are shown as mean ± SEM values (n = 16)

To determine if any underlying variables were masking the development of pathologic phenotypes, stratified analyses were performed based on fetal sex, and maternal race/ethnicity. Neither fetal sex nor maternal race/ethnicity was a determining factor for the induction of cell migration, as no differences were noted between any groups (Figure 4A). When analyzed based on fetal sex alone, the addition of LPS to amniotic fluid from male fetuses showed a significantly higher Cd11b expression compared to male amniotic fluid with no LPS (p = 0.02) (Figure 4B-Top graphs). No changes were observed between treated and untreated female samples (Figure 4B-Top graphs). When analyzed by maternal race/ethnicity alone, no differences in glial activation were observed between AA and WH amniotic fluid samples. However, the addition of LPS to WH amniotic fluid induced significant Cd11b expression compared to all other groups (AA LPS vs. WH LPS: p = 0.01; WH vs. WH LPS: p = 0.03) (Figure 4B-Top graphs). These data suggest that male fetuses, compared to female, and WH race/ethnicity, compared to AA, are more vulnerable to microglia activation in response to an intraamniotic inflammation.

FIGURE 4.

Maternal race/ethnicity, but not fetal sex, predisposes neonates to neuroinflammation under inflammatory conditions. (A) Bar graphs representing the number of migrated cells based on nuclei localization. Stratification of groups by fetal sex and maternal race/ethnicity did not show any significant difference in the number of migratory cells. Data are shown as mean ± SEM (n = 4). (B) Data analysis after stratification based on fetal sex and maternal race/ethnicity. Graphs represent the number of activated glial cells (CD11b positive cells) and Iba1 expression based on fluorescent image analysis. Top graphs -LPS in the male AF induced a significant amount of activated glial cells (p = 0.02) compared to male AF with no LPS. LPS in the WH AF induced a significant amount of activated glial cells (p = 0.03) compared to WH AF alone; these values were also significantly higher than AA AF co-treated with LPS (WH LPS vs. AA LPS: p = 0.01). Data are shown as mean ± SEM values. Bottom graphs – LPS in the male AF induced a significant decrease in Iba1 expression (p = 0.001) compared to male AF with no LPS; these results were significantly higher than LPS co-treated female AF (p = 0.009). LPS co-treatment of AA AF induced a significant decrease in Iba1 expression (p = 0.03) compared to AA AF alone; LPS co-treatment of WH AF was significantly lower than LPS treated AA AF (p = 0.003). Data are shown as mean ± SEM (n = 4). (C) Multiplex cytokine analysis measured media concentrations of pro-inflammatory cytokines IL-1β and IL-8 in the inner chamber after 48 hours of treatment. Stratified analysis of data showed no difference in cytokine levels when analyzed based on fetal sex; however, LPS co-treatment of WH AF induced significant production of IL-1β (p = 0.03) compared to WH AF alone; these values were also significantly higher than AA AF co-treated with LPS (p = 0.03). Conversely, LPS co-treatment of AA AF induced significant production of IL-8 (p = 0.02) compared to AA AF alone; these values were also significantly higher than WH AF co-treated with LPS (p = 0.009). Data are shown as mean ± SEM (n = 4)

To further confirm the activation status of microglia, the expression of resting marker Iba1, was also measured by fluorescence microscopy and stratified by fetal sex and maternal race/ethnicity. Corresponding with Cd11b activation status, the addition of LPS to male amniotic fluid samples induced a significant decrease in Iba1 expression (p = 0.001) compared to male amniotic fluid with no LPS. The addition of LPS induced a significant decrease in Iba1 within OOCs containing male compared to female amniotic fluid (p = 0.009) (Figure 4B-Bottom graphs). The addition of LPS to AA amniotic fluid induced a significant decrease in Iba1 expression (p = 0.01) compared to AA amniotic fluid with no LPS. LPS within amniotic fluid from WH also induced significantly lower Iba1 than AA amniotic fluid containing LPS (p = 0.003) (Figure 4B-Bottom graphs). These data further validate pathologic glial activation in this model and show that there are fetal sex and maternal race/ethnicity-specific differences in Iba1 expression.

Neuroinflammatory status was assessed by multiplex analysis of pro-inflammatory cytokines IL-1β and IL-8 after 48 h of culture, and results were stratified by fetal sex and maternal race/ethnicity (Figure 4C). When stratified by fetal sex, no significantly different production of pro-inflammatory cytokines was found between groups (Figure 4C). Conversely, maternal race/ethnicity did affect cytokine production. LPS co-treatment of WH amniotic fluid induced significantly higher levels of IL-1β compared to WH amniotic fluid alone (WH vs. WH LPS: p = 0.03; AA LPS vs. WH LPS: p = 0.03), while LPS co-treatment of AA amniotic fluid induced significantly higher levels of IL-8 compared to AA amniotic fluid alone AA vs. AA LPS: p = 0.02; AA LPS vs. WH LPS: p = 0.009) (Figure 4C). Taken together, these results suggest that while fetal sex does induce glial activation, maternal race/ethnicity is a stronger determining factor of neuroinflammation.

4 |. DISCUSSION

In this study, we modeled intraamniotic infectious inflammation and its impact on fetal neuroinflammation using a novel in-vitro microphysiological system. The OOC device contained a chamber for neuronal-glial cells and an amniotic fluid chamber separated by a collagen barrier. To mimic infectious inflammation, LPS was added to the amniotic fluid compartment, and experiments were performed. Our pooled analysis did not show an effect; hence, stratified analyses were performed. We report that fetal sex, specifically male sex, and maternal race, specifically WH, can modify neuroinflammation as indicated by cellular level changes (i.e., glial cell activation) along with higher cytokine response(i.e., pro-inflammatory cytokine IL-1β or IL-8). To note, the impact of male sex on neuroinflammation is not as pronounced as we observed with race and this very well may be a limitation due to our sample size as the trend, even in the cytokine response, was indicative of the vulnerability of the male fetal sex.

The results from this study correlate to in-vivo and clinical findings relating to risk factors of neuroinflammation. Both racial/ethnic and fetal sex differences have been observed in rates of PTB and its related outcomes. African-Americans (14.2%) and Hispanic infants (9.8%) have higher incidences of PTB compared to Caucasians (9.2%).¹ Male infant has worse outcomes after fetal hypoxic events related to PTB and slower cognitive recovery.⁴¹ Dada et al. characterized fetal sex differences in mouse models in relation to neuroinflammation and PTB. Male mice were found to have decreased hippocampal volume and increased macrophage and infiltrating cells compared to female mouse models documenting biological sex disparities In vivo.⁴² This observed sex difference supports clinical evidence that male neonates born preterm have worse outcomes than female neonates.⁴¹ In correlation to the literature, male AF did induce significant cellular level changes associated with neuroinflammation.

Primary screening of amniotic fluids prior to their use in our experimental models showed a significant increase in pro-inflammatory cytokines IL-6 and IL-8 in WH female samples compared to male AF (WH male vs. WH female IL-6: p = 0.03; WH male vs. WH female IL-8: p = 0.01) (data not shown). AF from female fetuses contained higher anti-inflammatory cytokine IL-10 regardless of race/ethnicity (AA male vs. AA female: p = 0.04; WH male 0.7 ± 0.0 ng/ml vs. WH female0.87 ± 0.1 ng/ml) (data not shown). The minimal neuroinflammation observed in female samples in our experiments may be attributed to the higher IL-10 content that can minimize resident or induced inflammation. This data correlates with the literature that has shown that there are sex differences in AF inflammatory environment, with female infants having a greater anti-inflammatory environment compared to males.¹¹

Race/ethnicity and sex differences have hampered the understanding of various disease pathologies leading to the failure of universal interventions. Our data not only support some of the reported findings in neuroinflammation but also provide an experimental approach to test neuroinflammation and generate mechanistic explanations. Peltier et al. showed that race/ethnicity-specific fetal membrane explants cultured with matched AF in the presence of infection-induced differential cytokine responses between AA and European-Americans.⁴³ In the presence of LPS, a similar response was induced by AA and WH AF samples in the production of IL-8 or IL-1β, respectively, by fetal glial cells. These data suggest that regardless of fetal organ, race/ethnicity plays a role in the response of the fetus to intraamniotic infection and or inflammation.

The adaptation of microfluidic OOC devices allows us to study the effect of risk factors of neuroinflammation in a novel way. The integration of human male fetus-derived glial cells removes the species-specific variabilities that can be found in in-vivo small or large animal models, allowing us to study clinically relevant changes due to fetal sex or maternal race/ethnicity. While this model has advantages over current in-vitro systems, limitations do exist, such as the cellular simplicity of this model (i.e., lacks neurons and oligodendrocytes), its inability to recreate brain regions to model white or grey matter disease, and the use of only male derived cells. The lack of neuronal cell migration in our model may partly be a limitation of OOC; however, we speculate that this may be due to the integrity of the barrier in this model, providing credence to our approach. Future modifications to microphysiological systems should be pursued to better model this complex interface.

In conclusion, we have adapted an OOC device to create a novel in-vitro model to study in-utero risk factors of neuroinflammation. We observed no significant effect of fetal sex on neuroinflammation; however, male fetuses are more vulnerable than female fetuses to neuroinflammation in the presence of intraamniotic infection or inflammation. Maternal race/ethnicity is a stronger determining factor of neuroinflammatory predisposition under adverse pregnancy conditions than fetal sex.

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.