Prenatal opioid exposure and maternal HCV infection impair microglia development and function: a patient-specific in vitro model

Abstract

Opioids are a class of pain-relieving drugs known to cross the placental and blood brain barriers, exposing the fetus in utero. Rates of opioid use disorder amongst pregnant individuals in the United States are on the rise, and intravenous routes of opioid administration are highly associated with hepatitis C (HCV) infection. Newborns with prenatal opioid exposure (POE) are more likely to be small for gestational age and have increased rates of neurodevelopmental delay. Microglia are brain-resident macrophages that originate from yolk-sac precursors that play critical role in neurodevelopment. However, our understanding of the impact of POE on microglia maturation and function remains limited due to the scarcity of adequate models. Here, we leveraged a model of induced microglia-like cells (iMGL) derived from umbilical cord blood mononuclear cells to uncover the mechanisms underlying the impact of POE ± maternal HCV infection on microglia morphology, phenotype, function, and transcriptional profiles. Our study revealed that iMGL are closely related to primary microglia. iMGL derived from pregnancies with POE and maternal HCV infection exhibited an ameboid-like phenotype, characterized by smaller area/perimeter and diminished ramifications. This was accompanied by dysregulated expression of key microglia markers, impaired phagocytic capacity, but increased secretion of inflammatory mediators. Finally, transcriptional analysis of iMGL with and without stimulation by LPS revealed that POE ± maternal HCV infection desensitized iMGL to LPS stimulation. This immune tolerance of iMGL in utero was reflected by altered expression of genes important for neurological and fetal development, phagocytosis, and antimicrobial responses with POE ± maternal HCV infection. Overall, these findings highlight the utility of iMGLs as an accessible patient-specific model to study preconditioning and development of fetal microglia and provide insight into mechanisms underlying adverse neurodevelopmental outcomes in newborns with POE in presence and absence of maternal HCV infection.

1.1. INTRODUCTION

In 2023, there were an estimated 81,083 overdose deaths resulted from opioid exposure in the United States ¹, largely driven by the influx of illicitly manufactured fentanyl ². While the opioid crisis has received extensive national attention, opioid use in pregnancy and mechanisms underlying adverse fetal outcomes have been relatively understudied ³. Rates of opioid use disorder (OUD) in pregnancy have increased dramatically since 1999 and disproportionally impact rural areas of the country, particularly Kentucky and the surrounding Appalachian region ⁴. Opioids are a class of drugs known to cross the placental barrier, exposing the fetus in utero ⁵. Opioid exposed pregnancies are further commonly complicated by maternal hepatitis C (HCV) infection, which is associated with intravenous routes of drug administration ⁶. Chronic HCV infection results in a sustained proinflammatory environment ⁷, which may influence immune cell maturation ⁸. Indeed, prenatal opioid exposure (POE) has been linked to stillbirth, fetal growth restriction, preterm birth (PTB), and neonatal opioid withdrawal syndrome (NOWS), predisposing surviving neonates to acute and long-term neurological deficits ^9–16. In utero opioid exposure and NOWS is associated with cognitive, speech, and motor deficits; higher rates of behavioral difficulties; impaired executive function; and poor school performance ¹⁷. Furthermore, children with POE have smaller brains, a higher incidence of microstructural brain injuries, and lower scores on cognitive tests ^{14, 18, 19}.

One proposed mechanism for how POE disrupts fetal brain development is by impairing microglia, resident macrophages of the central nervous system (CNS) ²⁰. Microglia originate from myeloid progenitors originating from the fetal yolk sac early in embryonic development where they cross the blood-brain barrier and migrate throughout the developing CNS ²¹. Here, microglia are critical mediators of healthy neurodevelopment by regulating the differentiation of neuron precursor cells and refining synaptic circuits ²². As microglia mature, they acquire their characteristic ramified morphology and maintain healthy brain homeostasis through regulation of neuronal survival and activity, synaptic pruning, and myelination ²³. Additionally, microglia play pivotal roles as immune sentinels of the brain and mediate neuronal repair ²⁴. Microglia are morphologically diverse and actively modify the shape of their projections in response to stimuli ²⁵. Thus, dysregulation of microglia development due to maternal environmental factors (such as toxins, infections, diet, or stress) can have lifelong consequences for offspring neurodevelopment ²⁶.

Murine models of prenatal neuroinflammation by targeting microglia toll-like receptors (lipopolysaccharide, bacterial endotoxins, or poly:IC) have shown that prenatal exposures can sensitize or precondition microglia to subsequent, postnatal insults ^{27, 28}. POE may mimic these effects, as opioids target opioid and toll-like receptors expressed by microglia ²⁹. However, a consensus is lacking on whether opioids prime microglia towards an immune-tolerant (desensitized) or immune-trained (sensitized) phenotype ^29–31. This discrepancy is likely due to variations in model used, the dose, duration, timing of exposure in microglia development, and/or route of opioid administration, and the complex interplay between various opioids and various opioid receptors ^{32, 33}. Additionally, ex vivo or in vivo studies of microglia function are limited to human post-mortem brain tissue or animal models ^{34, 35}. Immortalized cell lines ³⁶, primary isolated cell cultures ³⁷, and induced pluripotent stem cell (iPSC) derived microglia-like cells ³⁸ have been used in in vitro studies. However, these models present significant limitations, including reduced expression of key microglia genes, altered differentiation and phagocytosis, and increased expression of pro-inflammatory mediators ^{39, 40}. Recent advancements in reducing the differences between in vitro and in vivo experiments include co-cultures with other CNS resident cells, induced pluripotent stem cell-derived microglia, and patient-specific models of microglia function from peripheral blood ^{41, 42}.

Here, we used a recently established model of patient-specific umbilical cord blood mononuclear cell (UCBMC)-derived microglia-like cells (iMGL) ^{43, 44}, as UCBMCs and circulating fetal monocytes share erythromyeloid progenitor origins with microglia ⁴⁵. To infer the impact of POE ± HCV on the functional landscape of microglia, we obtained UCBMC from opioid-naïve (control) and opioid-exposed (POE) neonates, with and without maternal HCV infection. Differentiation of iMGL from UCBMC was confirmed by morphology and phenotyping, as well as by comparison of iMGL transcriptional profiles with established UCBMC and primary microglia bulk RNA sequencing datasets. We further quantified morphological features of skeletonized, z-stacked, immunofluorescence images of single-iMGL using ImageJ software. Finally, we assessed the functional landscapes of iMGL using multiplex Luminex assay, phagocytosis assay, and bulk RNA sequencing.

1.2. METHODS

1.2.1. Ethics approval statement

This study was approved by the Institutional Ethics Review Board of the University of Kentucky, Lexington, KY. (IRB# 43698 – Comprehensive Perinatal Substance Abuse Treatment)

1.2.2. Cohort characteristics

Umbilical cord blood samples were collected from participants with full-term, healthy pregnancies (N=12, controls) and those with prenatal opioid exposure (N=24, POE). Due to the high prevalence of HCV infection with intravenous drug use, we further stratified the POE group by HCV infection (N=12, POE HCV− and N=12 POE HCV+). Among those with positive HCV IgG, HCV viral loads were also measured in maternal blood collected at the time of delivery, indicating active HCV infection in 5 participants. HCV antibodies were not detected in any of the control participants. Other maternal and fetal characteristics were comparable among groups, including parity, maternal age, pre-pregnancy body-mass index (BMI), gestational age at delivery, fetal sex, and maternal race and ethnicity. Participant characteristics are summarized in (Table 1).

Table 1:

Cohort Characteristics

		Control	OUD HCV−	Control vs OUD HCV−	OUD HCV+	Control vs OUD HCV+	OUD HCV− vs OUD HCV+
	Number of enrollees	12	12		12
	DSM-V score	NA	8.67 ± 2.01	NA	9.58 ± 1.62	NA	p=0.233
	HCV IgG Antibody at Delivery (OD)	0.064 ± 0.04	0.01 ± 0.01	p=0.236	1.19 ± 0.42	p=<0.0001	p=<0.0001
	Parity	2.2 ± 1.9	1.5 ± 0.8	p=0.280	2.3 ± 1.6	p=0.819	p=0.128
	Maternal Age (years)	30.7 ± 6.2	29 ± 4.5	p=0.458	29.5 ± 5.1	p=0.619	p=0.800
	Pre-pregnancy BMI (kg/m 2 )	30.36 ± 7.82	29.01 ± 5.92	p=0.641	29.15 ± 4.80	p=0.654	p=0.952
	Gestational age at delivery (weeks)	39.10 ± 1.43	39 ± 1.25	p=0.822	38.5 ± 0.73	p=0.586	p=0.7545
	% Female newborn	50%	58.30%	p=0.698	58.30%	p=0.698	p=1
Race	White	75%	100%		91.67%
	Black	8.33%	0%		8.33%
	Asian	8.33%	0%		0%
	Mixed Race	8.33%	0%		0%
Ethnicity	Not Hispanic or Latino	100%	100%		100%

1.2.3. Sample collection and processing

Umbilical cord blood samples were collected in EDTA vacutainer tubes (BD Biosciences). UCBMCs were isolated after whole-blood centrifugation over LymphoPrep in SepMate tubes following manufacturer protocols (STEMCELL Technologies). UCBMCs were cryopreserved using 10% DMSO/FBS and Mr. Frosty Nalgene freezing containers (Thermo Fisher Scientific) at −80°C overnight and then transferred to a cryogenic unit until thawed and plated for induction to microglia outlined below in section 1.2.4.

1.2.4. Derivation of microglia-like cells (iMGL) from UCBMC

Induced microglia-like cells (iMGL) were derived from UCBMC using the protocol previously published by Sheridan et al.⁴³, with minor adaptations. Briefly, frozen UCBMC were thawed, and 1.0e⁶ cells were plated per well of a 24-well tissue culture plate previously coated with poly-D-lysine (50μg/ml) per manufacturer’s instructions. 500μl of media containing RPMI 1640, 10% FBS, and 1% penicillin/streptomycin was added to each well, and placed in a 37°C incubator with 5% CO₂ for 24 hours. After 24 hours, the media was replaced with induction media containing RPMI 1640, 1% Glutamax, 1% penicillin/streptomycin, 100 ng/ml IL-34 (Peprotech), and 10 ng/ml GM-CSF (Peprotech) and the plate returned to 37°C incubator with 5% CO₂ for an additional 12 days. On day 13, post-seeding, media was replaced with fresh induction media and returned to the incubator for a final 24 hours. On day 14 post-seeding, the presence of iMGL was confirmed by the appearance of ramifications by microscopy and harvested using cell scrapers and immediately used for downstream experiments.

1.2.5. Phenotyping

Thawed UCBMC from newborns of full-term, healthy pregnancies (N=9) or iMGL (N=27, 9 Controls, 9 POE HCV−, and 9 POE HCV+) were thoroughly washed with FACS buffer, and stained with a cocktail of the following surface antibodies at a ratio of 1:20 in 50μl/sample of Brilliant Stain Buffer (BD Biosciences): CD14, HLA-DR, CD16, CX3CR1, CD45, CD11b, P2RY12, TMEM119, TREM2, CD68, CD40, CD163, CD115, and CD86 ^{46, 47} as well as True-Stain Monocyte Block and Human TruStain FcX^™ (Fc Receptor Blocking Solution, BioLegend, 1:20). After incubation for 30 minutes at 4°C, UCBMCs or the iMGL cell pellet was washed with FACS buffer, fixed for 2 hours (Tonbo fix/permeabilization solution, 1:3), permeabilized (Tonbo permeabilization buffer), and stained with PU.1, IBA.1, and IRF8 intranuclear antibodies at a ratio of 1:20, overnight 4°C. The next day, stained UCBMCs/iMGLs were washed with FACS buffer, ran on the Cytek Aurora flow cytometer (Cytek Biosciences, 5-laser; 355 nm, 405 nm, 488 nm, 561 nm, and 640 nm) using the SpectroFlo Software v2.2.0.2. The cells were unmixed using stained beads with the autofluorescence extraction option enabled and extracted FCS files were analyzed using Flowjo 10.10.0 (Ashland, OR). Briefly, undesired events were removed by FlowAI 2.3.2 ⁴⁸. For supervised gating, cells expressing CD45+CD11b+P2RY12+TMEM119+ were designated as iMGL. The expression of surface and intracellular markers on UCBMCs/iMGL was then assessed.

1.2.6. Immunofluorescence staining

iMGL were differentiated in a 24-well plate on coverslips as described in section 1.2.4. On day 14, coverslips were carefully transferred to a new plate and fixed with 4% paraformaldehyde (BioLegend) for 30 minutes at room temperature. Following fixation, cells were washed three times with phosphate-buffered saline (PBS) for 5 minutes per wash. Cells were then permeabilized by incubating the coverslips in PBS with 0.2% Triton X-100 for 20 minutes at room temperature. After two additional PBS washes, blocking was performed in 5% bovine serum albumin (BSA) in PBS. Primary antibody IBA1 (1:1000, FUJIFILM Wako Pure Chemical 019-19741) or TMEM119 (1:50, ThermoFisher PA5-119902) incubation was conducted overnight at 4°C in a humidified chamber. The next day, cells were washed three times with blocking buffer (5% BSA-PBS) for 10 minutes each. Next, 0.5 μg/ml of secondary antibody Goat anti-Rabbit IgG-AlexaFluor Plus 594-Red (Invitrogen) in 1% BSA-PBS was added to cells and incubated for 1 hour at room temperature in the dark. To remove unbound antibodies, cells were washed an additional three times with PBS. Coverslips were then mounted onto glass slides using Vectashield^® antifade mounting medium with DAPI (H-1200-10) and sealed with clear nail polish. Slides were allowed to dry overnight and stored at 4°C until imaging.

1.2.7. Morphology analysis

Images of IBA1-stained iMGL were captured at 40X magnification with a step size of 1 μm using z-stack acquisition parameters. Captured images were then analyzed using ImageJ software to identify pixels with the highest intensity at any position in a z-stack. The 8-bit images were further processed using ImageJ MicrogliaMorphology and its R package ⁴⁹ with slight modifications tailored to in vitro microglia culture. Single-cell iMGL images were generated using the BioVoxxel Toolbox plugin. This semi-automated process provides recommended thresholding, and the highest quality threshold was selected to enhance the distinction between microglia ameboid and ramified forms. Skeletonization was then performed on the single cells, and fractal analysis was conducted using the FracLac plugin. Color coding of the images was achieved using the ColorByCluster feature. The resulting 27 morphology features were plotted to observe differences between the groups. These features were further processed using the MicrogliaMorphology R package and evaluated using Principal Component Analysis (PCA) and clustering analysis to capture cell heterogeneity and identify morphologies specific to each group.

1.2.8. Phagocytosis Assay

To quantify the phagocytic ability of UCB-derived microglia-like cells (iMGL), iMGLs (N=36, 12 Controls, 12 POE HCV−, and 12 POE HCV+) were incubated for 2 hours at 37°C in media containing 1 mg/mL pH-sensitive pHrodo E. coli BioParticles conjugates (ThermoFisher Scientific). Cell pellets were washed twice in FACS buffer, surface stained at 4°C with antibodies against CD2, CD20, CD45, CD14, HLA-DR, and P2RY12 at a ratio of 1:20, then resuspended in ice-cold FACS buffer. Samples were run using an Attune NxT and analyzed using FlowJo 10.10 software (Beckton Dickinson).

1.2.9. Luminex assay

Immune mediators in the supernatant of day-14 iMGL were measured using a custom 10-plex panel measuring levels of myeloid/microglia cell activation markers (IFNγ, IL-10, IL-1β, IL-4, IL-6, IL-8, MCP-1, MCP-3, MIP1α, and MIP1β) (Milliplex Human Cytokine/Chemokine/Growth Factor Panel A, HCYTA-60K). Standard curves were generated using a 5-parameter logistic regression using the xPONENT^™ software provided with the MAGPIX instrument (Luminex).

1.2.10. LPS Stimulation

Day 14, adherent iMGLs were stimulated with LPS (1 μg/ml) or RPMI-1640 only (non-stimulated controls) for 4 hours at 37°C. Cells were then harvested using cell scrapers and centrifuged for 5 minutes at 300g. iMGL cell pellets were resuspended in 700μl of QIAzol Lysis Reagent (Qiagen) and stored at −80°C until use for subsequent RNA isolation outlined below (section 1.2.11).

1.2.11. RNA Isolation/Library Building

Total RNA was isolated from iMGL ± LPS stimulation using an mRNeasy kit (Qiagen). Quality and concentrations were measured using Agilent 2100 Bioanalyzer. Libraries were generated using the NEBnext Ultra II Directional RNA Library Prep Kit (New England Biolabs). Briefly, following rRNA depletion, mRNA was fragmented for 7 minutes, converted to double-stranded cDNA, and adapter ligated. Fragments were then enriched by PCR amplification and purified. The size and quality of the library were verified using Qubit and Bioanalyzer. Libraries were multiplexed and sequenced on the NovaseqX platform (Novogene) to yield an average of 20 million 100 bp single end reads per sample.

1.2.12. iMGL bulk RNA-seq analysis

Quality control of raw reads was performed using FastQC retaining bases with quality scores of >20 and reads >70bp long. Reads were then aligned to the human genome (hg38) using splice-aware HISAT2 using annotations available from ENSEMBL (GRCh38.85) database. Quantification of read counts was performed using the GenomicRanges package in R and converted to transcripts per million (TPM) counts. Read counts were used to generate PCAs; then, all non-protein-coding genes were filtered out before differential gene expression analysis. Raw counts were used for testing of differentially expressed genes (DEGs) using edgeR ⁵⁰. Responses to LPS were modeled relative to unstimulated samples using negative binomial GLMs. DEGs were defined as having log2 FC +/− 1 and FDR <0.05. Functional enrichment of DEGs was performed using Metascape ⁵¹. Heatmaps of TPMs and bubble plots of enrichment of Gene Ontology (GO) terms were generated using ggplot in R. A sPLSDA for non-stimulated and LPS-stimulated cells was conducted using the MixOmics ⁵² package with the normalized counts.

For the comparison of transcriptional profiles between iMGL, monocytes, and primary microglia, we leveraged our previously published bulk-RNAseq dataset on human UCB monocytes ⁵³, as well as a human primary microglia data set by Galatro et al ⁵⁴. After merging datasets, all genes were annotated using the ENSEMBL (GRCh38.85) database. Quantification of read counts, PCA generation, and differential gene expression analysis were performed as outlined above. Module scores were generated by repurposing the AddModuleScore() function found in Seurat3 toward bulk RNA-seq data processed with DESeq2 ⁵⁵. The 135 input genes used to generate Module Scores were selected from human microglia gene sets in the Molecular Signatures Database (MSigDB) ⁵⁶ and are listed and described in Supplemental table 1. The DEGs produced by the primary microglia vs monocyte and iMGL vs monocyte comparisons were modelled using a Venn diagram using VennDiagram in R. Functional enrichment of the genes identified as common in the Venn diagram was performed using Metascape ⁵¹. Heatmaps of TPMs and bubble plots of enrichment of GO terms were generated using ggplot in R. The association between log2 fold change and FDR was modelled using EnchancedVolcano in R to generate a volcano plot. Gene lists used to for bulk-RNAseq analysis are outlined in Supplemental Tables 1–3.

1.2.14. Statistical analysis

We conducted a normality assessment using the Shapiro-Wilk test (alpha=0.05) and identified/removed outliers through ROUT analysis (Q=0.1%). For two-group comparisons, if the dataset showed adherence to a normal distribution, group differences were assessed using an unpaired T-test with Welch’s correction. In cases where the data did not meet Gaussian assumptions, group comparisons were conducted using the Mann-Whitney test. For comparisons of three independent variables (e.g. control vs POE HCV− vs POE HCV+) or four independent variables (e.g. UCBMC vs control vs POE HCV− vs POE HCV+), if the dataset showed adherence to a normal distribution, group differences were assessed using a One-way ANOVA with Fisher’s LSD test for independent comparisons. In cases where the data did not meet Gaussian assumptions, group comparisons were conducted using the Kruskal-Wallis test with uncorrected Dunn’s test for independent comparisons. Statistical analyses were performed using Prism (GraphPad).

1.3. RESULTS

1.3.1. In utero fetal exposure to maternal opioid use and HCV infection lead to adverse maternal-fetal outcomes.

Umbilical cord blood (UCB) from newborns with and without prenatal opioid exposure (POE) was collected from full-term (>37 weeks’ gestation) pregnancies. Maternal demographics are summarized in Table 1. At enrollment, all participants in the POE group were screened for severity of opioid use disorder per Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-V) criteria. All participants in the POE group reported a history of illicit use of one or more opioids (heroin, fentanyl, and/or non-prescribed oxycodone, hydrocodone, buprenorphine, and methadone products) (Figure 1A) and met the criteria for severe opioid use disorder, by meeting 4 or more DSM-V criteria at time of enrollment (Table 1). All participants received buprenorphine (a partial opioid agonist) throughout their pregnancies for management of opioid use disorder. Given the high prevalence of hepatitis C infection (HCV) with intravenous drug use, we stratified participants based on their HCV status, via self-reporting and confirmed by HCV IgG antibody detection at time of delivery (Table 1). Of the 12 participants positive for HCV IgG, 5 were also positive for viral load detected by RT-qPCR, indicating active HCV infection at time of delivery. Other maternal factors including parity, maternal age, pre-pregnancy body mass index (BMI), gestational age at delivery, race, and ethnicity were comparable between control, and POE ± HCV groups (Table 1).

Figure 1: History of maternal opioid use and newborn outcomes.

(A) Pie chart depicting self-reported maternal opioid use at time of enrollment in the past 1 year. (B-D) Bar plots of (B) newborn birthweight (grams) and length (centimeters), (C) newborn length of stay in intensive care units (NICU/NACU), and (D) highest Finnegan score measured. (E) Stacked bar graph depicting the frequency of newborns receiving pharmacological treatment (morphine) for NOWS among POE HCV− and POE HCV+ groups. #=p<0.1, *=p<0.05, ***=p<0.001.

POE exerts acute and long-term adverse outcomes for newborns, including growth restriction, developmental delays, and the increased likelihood of chronic diseases in later life ⁵⁷. Here, newborns with POE ± maternal HCV exposure were born smaller (Figure 1B) and more likely to require admission to intensive care units (NICU/NACU) for the management of neonatal opioid withdrawal syndrome (NOWS) (Figure 1C). The Finnegan Neonatal Abstinence Syndrome (FNAS) scoring is a commonly used clinical tool to guide pharmacological interventions in infants with NOWS ⁵⁸, with scores above 8 considered pathological and requiring treatment ⁵⁹. We report comparable Finnegan scores in newborns with POE regardless of maternal HCV status (Figure 1D). However, newborns with POE and maternal HCV exposure were significantly more likely to require treatment for severe NOWS symptoms compared to those with POE without maternal HCV exposure (Figure 1E). Overall, these clinical findings indicate that POE and maternal HCV infection are associated with adverse neonatal outcomes.

1.3.2. UCB-derived microglia-like cells (iMGL) as a model of prenatal opioid exposure (POE).

Microglia originate from erythromyeloid progenitors (EMP) in the fetal yolk-sac and are brain-resident macrophages that play a key role in neurodevelopment by regulating synapse formation and supporting neuronal proliferation ⁶⁰. Their early ontogeny during fetal development makes them exquisitely sensitive to changes in maternal health ^{61, 62}. Fetal monocytes also originate from EMP’s in the fetal yolk-sac as well as liver ⁴⁵. Given the shared EMP origin of microglia and fetal monocytes, we leveraged and optimized an established protocol to differentiate microglia-like cells from umbilical cord blood mononuclear cells (iMGL) ⁴³ to uncover how POE +/− HCV priming affects microglia development and sensitization. iMGL were harvested from culture, and then phenotype, morphology, function, and transcriptional profiles were assessed using flow cytometry, immunofluorescence imaging, immune mediator production, and bulk RNA-sequencing (Figure 2A).

Figure 2: UCB-derived microglia share similar profiles to human primary microglia.

(A) Experimental design. (B) Confocal imaging (40x) of DAPI (blue) and IBA1 (red, left) or (20x) TMEM119 (red, right) stained iMGL. (C) Principal Component Analysis (PCA) of the comparison of non-stimulated iMGL, isolated monocytes, and primary microglia samples. (D) Bar plot of module scores generated from a list of 135 key microglia and immune function genes across UCB monocyte, iMGL, and human primary microglia bulk-RNAseq datasets. (E) Venn diagram representing the unique and shared DEGs of iMGL and primary microglia compared to UCBMC. (F) Bubble plot of Gene Ontology (GO) terms associated with select genes from the 1189 common DEGs between iMGL and primary microglia represented in the Venn diagram in Figure 2E. The size of the bubble denotes the number of genes mapping to each GO term, and the intensity of the color denotes the statistical strength of prediction. (G) Heat map depicting genes mapping to select GO terms in Figure 2E. (H) Volcano plot of DEGs in iMGL compared to primary microglia. #=p<0.1, **=p<0.01, ***=p<0.001.

Successful derivation of iMGL was first confirmed by morphology and expression of canonical microglia markers IBA1 and TMEM119 by confocal microscopy (Figure 2B). iMGL morphology was reflective of the dynamic nature of primary microglia, including the appearance of long, branching processes (ramifications), rod/bipolar like projections, as well as small, dense, and circular “ameboid” structures (Figure 2B). Next, we assessed the transcriptomic profile of iMGL compared to publicly available human, UCB monocyte ⁵³ and primary microglia ⁵⁴ bulk-RNAseq datasets. Because no RNA sequencing datasets are available for healthy human fetal/newborn microglia, we selected a primary human microglia the dataset that had young healthy adults (~30-40 years old) without neurodegenerative diseases (e.g. Alzheimer’s disease, Parkinson’s disease, Amyotrophic lateral sclerosis (ALS)) ⁵⁴. Our principal component analysis (PCA) revealed that the iMGL transcriptome was comparable to primary microglia and distinct from UCB monocytes (Figure 2C). Similarities in gene expression patterns between iMGL and primary microglia relative to UCB monocytes were further highlighted by elevated module scores of 135 genes important for microglia differentiation and function identified in the molecular signatures database (MSigDB) ⁵⁶ (Figure 2D, Supplemental Table 1). Our analysis of differentially expressed genes (DEG) revealed 1189 DEGs shared between both iMGL and primary microglia relative to UCB monocytes (Figure 2E, Supplemental Table 1). DEGs that were upregulated by iMGL and primary microglia compared to UCB monocytes enriched to gene ontology terms (GO) important for synaptic signaling (TMEM37, APOE, MSR1), as well as the development of cell/neuron projections and axons (PARP1, TRIM32, LAMP2), and glial cell differentiation (NRROS, MYOF) (Figure 2F–G, Supplemental Table 1).

Next, we identified genes that were differentially expressed between iMGL and primary microglia. Genes downregulated in iMGL relative to primary included those playing a role in brain resident microglia (SIGLEC8, SALL1, P2RY12, CX3CL1, HLA-DPA1), DNA repair mechanisms (PPP1R10, MDC1), cell proliferation (FLOT1) and inflammatory responses (GPSM3, HCG22) (Figure 2H, Supplemental Table 1). Genes upregulated in iMGL relative to primary microglia are involved in microglia function (STAC, SARPC, GPC4, CRABP2, CD109, and TIMP3) (Figure 2H, Supplemental Table 1). Finally, given that opioids interact with both opioid and toll like receptors on myeloid cells, we assessed the general expression of opioid receptors (OPRD1, OPRK1, OPRM1, OPRL1) and TLR4 (Supplemental Figure 1A). iMGL express OPRL1 and TLR4, although to less extent than observed by primary microglia (Supplemental Figure 1A). Overall, these findings support that iMGL provide a useful in vitro model of microglia development and express receptors that are targetable by opioids and bacterial endotoxins such as lipopolysaccharide (LPS).

1.3.3. Altered phenotype and morphology of iMGL with POE and maternal HCV infection

To evaluate the impact of POE ± maternal HCV infection on iMGL morphology, we processed iMGL confocal microscopy images using the ImageJ MicrogliaMorphology and R package ⁴⁹. The extracted 27 morphology features were plotted to observe the overall differences between the groups. Comparison of extracted features from review of individual microglia per group showed lower iMGL “area”, “perimeter”, “mean radius”, and “average/maximum branch length” with POE and HCV infection (Figure 3A). Additional morphology features that describe area and territory span were also lower in the POE HCV+ group compared to controls (Supplemental Figure 1B), highlighting that microglia are smaller, more circular and dense with POE.

Figure 3: POE and maternal HCV alter iMGL structure and phenotype.

(A-C) Bar graphs depicting (A) Select microglia morphology features from IBA1-stained cells imaged at 40X and processed using ImageJ macro MicrogliaMorphology and MicrogliaMorphologyR package ⁴⁹, (B) Cluster frequency of ramified, bipolar/rod, and ameboid iMGL, and (C) MFI of myeloid/monocyte activation markers measured by full spectrum flow cytometry. * = p < 0.05, ** = p < 0.01, ***=p<0.001, ****=p<0.0001

We next used dimensionality reduction and clustering analysis using MicrogliaMorphologyR 1.1 to identify correlations between principal components and features to characterize iMGL clusters. This analysis of the first ten principal components generated 3 distinct iMGL clusters (Supplemental Figure 1C). Cluster 1 microglia were named “Ramified” due to characteristics including larger size (area, foreground pixels, # of slab voxels, height of bounding rectangle) with features of extensive and complex branching patterns (number of end point voxels, quadruple points, branches, junction voxels, junctions, triple points) (Supplemental Figure 1D). Cluster 2 microglia were named “Bipolar/Rod” due to features of rod-like shape (radii from hull’s center of mass, branch length, span/hull ratio, and relative variation in radii from hull’s center of mass) (Supplemental Figure 1D). Finally, Cluster 3 microglia exhibited features of ameboid morphology, including overall smaller size and increased circularity and density. We next scaled average values for all morphology features among clusters and highlighted individual cells using the ColorByCluster function in MicrogliaMorphology (Supplemental Figure 1E). The frequency of iMGL that fall within cluster 3 was significantly increased with POE (Figure 3B).

Finally, iMGL phenotype was assessed by spectral flow cytometry. We used a panel including CD45 (a pan marker for immune cells), canonical microglia markers (CD11b, IBA1, CX3CR1, PU.1, PR2Y12, and TMEM119) as well as additional markers of differentiation and function (CD14, CD16, CD68, CD163, TREM2) (Supplemental Figure 2A). Previous studies have identified CD11b, P2RY12, and TMEM119 as key markers for identifying microglia in early embryonic and postnatal development ⁴⁴. While CD11b can be expressed by both macrophages and microglia, P2RY12 and TMEM119 are largely microglia specific ⁴⁴. Therefore, the iMGL population was defined as CD45+CD11b+P2RY12+TMEM119+ (Supplemental Figure 2A). We report a decrease in CD45+CD11b+ population as well as the differentiated P2RY12+TMEM119+ iMGL in both POE groups, that is more pronounced in the POE HCV− group (Supplemental Figure 2A–B), in line with reports of P2RY12 and TMEM119 downregulation and impaired microglia differentiation in disease-associated and reactive states ⁶³. Furthermore, a modest decrease in IBA1 expression in the POE HCV+ group was noted compared to control (Supplemental Figure 2B) while CX3CR1 expression remained consistent among iMGL groups (Supplemental Figure 2B). The expression of P2RY12, TMEM119, IBA1, and CX3CR1 was negligible in undifferentiated UCBMC (Supplemental Figure 2B). With the CD45+CD11b+P2RY12+TMEM119+ subset as the parent, iMGL, population, we compared median fluorescence intensity (MFI) of key markers with POE ± HCV (Figure 3C). Relative to controls, iMGL derived from the POE HCV− group had increased MFI of PU.1, HLA-DR, and IRF8, which are important regulators of microglia inflammatory responses ⁶⁴ (Figure 3C). On the other hand, the MFI of CD16, TREM2, CD68, and CD40 was all lower in both POE groups compared to control (Figure 3C) and have been associated with attenuated neuroinflammatory responses and phagocytosis ⁶⁵ as well as neurodevelopmental disorders ^66–68.

1.3.4. POE and maternal HCV infection dysregulate iMGL function

Given the altered expression of markers important for phagocytosis by iMGL in the POE groups, we assessed iMGL function by measuring the phagocytosis of pHrodo^™ Red E. coli BioParticles. Our results indicate a significant decrease in iMGL phagocytic capacity with POE compared to controls, that is further exacerbated by maternal HCV infection (Figure 4A). Next, we measured spontaneous cytokine production by iMGLs by Luminex assay. Levels of inflammatory mediators IL-1β, IL-6, and IL8 were modestly elevated in the POE groups compared to controls, regardless of HCV status (Figure 4B). However, the higher concentration of IL-6 and IL-1β was less significant in the POE HCV+ group (Figure 4B). Overall, these findings suggest altered phagocytic capacity and inflammatory signaling pathways by iMGL with POE.

Figure 4. Impaired iMGL function with POE and maternal HCV.

(A-B) Bar plots depicting A) iMGL phagocytosis determined by uptake of pHrodo E. coli (MFI) by flow cytometry, and B) Concentration of spontaneous cytokine production in iMGL culture supernatant by Luminex. # = p < 0.1, * = p < 0.05, ** = p < 0.01.

1.3.5. POE and maternal HCV alter the transcriptional profile of iMGL

To further uncover changes in iMGL function with POE ± maternal HCV, we treated iMGLs with and without LPS treatment for 4 hours. We chose LPS as it is a commonly used stimulus to study microglia function ^69–71 and acute microglia responses in vitro are detected 4 hours post LPS stimulation ^{69, 72}. Furthermore, opioids have been reported to weakly activate TLR4 thereby inhibiting LPS-induced TLR activation ^{73, 74} and murine models of microglia immune training have shown that prenatal LPS treatment can sensitize microglia to subsequent insults postnatally ^{73, 74}. Therefore, we used bulk-RNA-sequencing to assess the transcriptional profiles of LPS treated iMGLs to distinguish possible preconditioning effects of POE ± maternal HCV on iMGL function. First, differential gene expression (DEG) among groups, revealed the largest number of DEGs between HCV− and HCV+ samples with and without LPS treatment (Figure 5A, Supplemental Table 2). DEGs identified among non-stimulated POE ± maternal HCV groups enriched to gene ontology (GO) terms including LPS response (IL1B, IDO1, PTGS2), positive regulation of cytokine production (EGR1, EREG, PTGS2), and negative regulation of cell proliferation (PELI1, ADM) (Figure 5A–C, Supplemental Table 2). However, DEGs enriched to GO terms important for viral responses (IRF1, CXCL9, IFITM3) among LPS stimulated POE ± maternal HCV groups (Figure 5A–C, Supplemental Table 2).

Figure 5. Transcriptional profiles of iMGL at baseline and following LPS stimulation are altered by POE and maternal HCV.

(A) Bar plot of number of differentially expressed genes (DEG) by group (control, POE HCV−, POE HCV+) and treatment [Non-stimulated (NS) or LPS stimulated (LPS)]. (B) Bubble plot of gene ontology (GO) terms associated with the select genes expressed by iMGL in the HCV+ group compared to HCV−, of non-stimulated (left) and following LPS stimulation (right). The size of the bubble denotes the number of genes mapping to each GO term, and the intensity of color denotes the statistical strength of prediction. (C) Heatmap of average gene expression from comparison of DEGs (control, POE HCV−, POE HCV+) and treatment (NS or LPS stimulated). (D) Three-way Venn diagram representing unique and shared DEGs among control, POE HCV−, and POE HCV+ groups with LPS stimulation. (E) Bubble plot of GO terms associated with select genes expressed by iMGL in control, POE HCV+, and shared between both control and POE HCV+ groups with LPS stimulation. (F) Heatmap of average gene expression supporting the GO terms selected in the bubble plot in panel E.

Next, we compared DEGs post-LPS stimulation based on POE exposure and maternal HCV status (Figure 5D, Supplemental Table 2). In the control group, upregulated DEGs enriched to GO terms related to neurodevelopment and function (NCAM1, KIF5C, GRIK2, GAD1, SEMA5A, ROBO1, and ITGA2), while downregulated genes mapped to functions important for the regulation of IL-6 production (IL16, NOD1, SPON2) (Figure 5E–F, Supplemental Table 2). Upregulated DEGs also enriched to shared GO terms across control and HCV+ groups, including response to wounding (ERBB2, BCL2, EREG), innate immune response (C3, CASP4, CAMP), and cellular response to cytokines (FAS, ICAM1) (Figure 5E–F, Supplemental Table 2). Upregulated DEGs shared between POE groups enriched to GO terms including myeloid cell differentiation (CSF3, LYN, MT1G) with LPS (Figure 5E–F, Supplemental Table 2). In contrast, DEGs upregulated only in the HCV+ group in response to LPS uniquely mapped to GO terms associated with viral response (IRF1, TLR2, HIF1A), cell activation (CCR7, IL6, SERPINE2), learning or memory (RCAN1, SRC, FOSL1), and modulation of synaptic transmission (CAMK2A, EDN1, KMO) (Figure 5E–F, Supplemental Table 2).

Finally, a sparse partial least squares differential analysis (sPLSDA) of the non-stimulated iMGL further highlights the impact of POE and maternal HCV status on the transcriptome of non-stimulated iMGL (Figure 6A). Notable genes that delineated the POE HCV− group were associated with various neurological pathologies and fetal development (HDX, MN1, HSF5, GAREM1) (Figure 6B, Supplemental Table 3), whereas genes that define the POE HCV+ group were involved in neuroinflammation and neuropathies (CCDC57, MTX1, PRR5, PPP2R5D, SSU72) (Figure 6B, Supplemental Table 3). The sPLSDA analysis of the LPS-stimulated iMGL (Figure 6C) showed increased expression of genes involved in cell proliferation/growth (TRIM23, MRPL13, ODC), signaling (ASB8, CSE1L), and metabolism (BSC1L, LIN54) in the control group (Figure 6D, Supplemental Table 3). Expression of CFAP298, a gene involved in neurogenesis, was increased in the POE HCV− group, whereas the expression of several other genes that play a role in regulating microglia activation and inflammatory responses (FSTL1, HDAC5, GMCL1), DNA repair mechanisms (ZNF689), and neurodevelopmental disorders (TARS2, B3GNT8) were higher in both POE HCV− and POE HCV+ groups relative to control (Figure 6D, Supplemental Table 3). However, expression of genes involved in cell migration/signaling (CHCHD4, PTK7, SLC13A5), and nervous system development (RNF165, TBX1) were significantly higher in the POE HCV+ group (Figure 6D, Supplemental Table 3). Overall, our analysis of iMGL transcriptional profiles suggests that POE and maternal HCV infection are associated with a heightened inflammatory iMGL state at baseline. These analyses were performed on total cell populations, therefore, the varied degrees of iMGL differentiation in the POE ± HCV groups are possible drivers of differences in gene expression and LPS responses among iMGL.

Figure 6. Transcriptional profile of NS and LPS-stimulated iMGL by POE and maternal HCV infection.

(A) PCAs from the sPLSDA comparisons of the non-stimulated (NS) iMGL displayed in 1-2 (top) and 3-4 (bottom) space. (B) Heatmap of the average gene expression from selected genes from the sPLSDA of the NS iMGL. (C) PCA from the sPLSDA comparison of lipopolysaccharide (LPS) stimulated iMGL. (D) Heatmap of the TMP values of selected genes from the sPLSDA of LPS-stimulated iMGL.

1.4. DISCUSSION

In early pregnancy, yolk sac-derived fetal myeloid progenitors seed in the fetal brain and mature into microglia, critical regulators of neurodevelopment. As such, programming of microglia differentiation and function begins before birth and can be influenced by changes in the maternal environment. Rodent models of prenatal exposure to LPS have shown that maternal inflammation influences microglia development and function ^{27, 28, 75, 76}. Similarly, previous studies have shown increased inflammation and immune activation in both maternal and neonatal circulation in rodent models of prenatal opioid exposure (POE) ⁷⁷. This “priming” or “conditioning” prenatally can lead to immune tolerance (desensitization) or training (sensitization) of microglia and consequently altered responses to subsequent challenges postnatally ^{27, 28}. However, few studies have addressed the mechanisms underlying altered microglia function and neurodevelopment in newborns with POE in humans. The few studies completed in rodent models of POE have demonstrated reduced microglia frequencies ⁷⁸ and altered functional capacity including dampened surveillance properties including phagocytosis and sculpting of neuron circuits by opioid exposed microglia ²⁰. Therefore, dysregulation of microglia development and function are a possible mechanism underlying adverse fetal outcomes and stunted neurodevelopment in newborns with POE. Consistent with prior studies highlighting adverse infant outcomes with maternal opioid use, our data show that POE is associated with smaller newborns who require long-term stays in intensive care units for management of NOWS ¹². These findings underscore the need for additional models to better understand the mechanisms underlying adverse outcomes of newborns with POE.

Rodent models are not able to fully recapitulate the biology of human microglia ⁷⁹. Therefore, patient-derived in vitro models of microglia function are warranted to recapitulate features of human neurodevelopment. Indeed, human microglia-like cells have been generated from human somatic cells and validated against primary cells ^{42, 43}. Umbilical cord blood mononuclear cells share erythromyeloid progenitor origins with microglia ⁴⁵, and therefore offer an accessible window towards early microglia development and priming. Here, we used a recently established model of patient-specific UCBMC-derived microglia-like cells (iMGL) ⁴³ to recapitulate in vivo microglia development with maternal inflammation/immune activation. To support the use of iMGL as an in vitro surrogate to study microglia function, our data show that iMGL have distinct phenotypic and transcriptional profiles from monocyte precursors, are similar to that of other in vitro differentiated microglia-like cells ^{43, 80, 81}, and share key transcriptional features with primary human microglia ^{54, 56}.

In vivo, microglia morphology can provide insight into microglia function, as they protrude and retract their projections to survey and respond to environmental cues ⁸². We identified three clusters of iMGL based on their morphology designated as ramified, bipolar/rod, and ameboid. In vivo, ramified microglia are considered homeostatic, or surveillant, and are characterized by extensive branching with multiple primary and secondary processes ²⁵. Microglia shift to unramified forms in response to inflammatory or damage signals in their environment, thought to increase motility to sites of inflammation where they phagocytose debris and release immune mediators to repair damage or resolve infection ²⁵. We report an increased abundance of ameboid/unramified iMGL that have decreased phagocytic capacity and increased spontaneous cytokine production with POE and HCV compared to controls, suggesting that these microglia are primed towards an inflammatory phenotype but are functionally defective. Other in vitro models of iPSC-derived microglia have highlighted that ameboid microglia persist in cases of neuron/astrocyte death and remodeling ⁸³, while others have identified ameboid microglia as under-differentiated or immature ^{22, 84}. Our findings indicate a shift towards a more pro-inflammatory, and potentially damaging phenotype often associated with neurological deficits in primary microglia ^{60, 85–87}.

Given the increased ameboid morphology of iMGL with POE and HCV infection and the association of ameboid microglia with underdevelopment and potentially damaging neuroinflammation ²⁵, we next assessed iMGL phenotype and function. Our flow cytometry analysis revealed a decrease in the abundance of “differentiated” P2RY12+TMEM119+ iMGL with POE and maternal HCV infection. Given the in vitro nature of this model, these findings suggest that POE may be interfering with iMGL development. However, downregulation of both P2RY12 and TMEM119 have been associated with damage associated microglia (DAM) phenotypes in vivo. This DAM phenotype has been further described by loss of TREM2 expression, a key sensor for damage and associated with neurodevelopmental defects ^{44, 88}. Among differentiated iMGL (defined as CD45+CD11b+P2RY12+TMEM119+), we note significant shifts in the expression of key microglia reactivity and differentiation markers with POE ± maternal HCV. Notably, a downregulation of TMEM2, CD68, and CD40 which are all important for maintaining neuron development and mediating responses to inflammation or damage in the CNS ^{22, 65, 67, 89}. Although iMGL from both POE groups secreted high levels of inflammatory IL-1β, IL-6, and IL-8 at baseline, phagocytic capacity was decreased compared to controls. Impaired phagocytic capacity and therefore poor clearance of cellular debris lead to accumulation of toxic substances in the brain contributing to neurological impairment ^90–93. Furthermore, PU.1, HLA-DR, and IRF8 expression were all significantly higher in the POE HCV− group compared to controls, all of which have been associated with regulation of microglia development and maintenance ^{89, 94}. Notably, PU.1 and IRF expression are important regulators of microgliogenesis, phagocytosis, and inflammatory responses ^95–97. However, elevated PU.1 and IRF8 expression can render microglia more resistant to cell death and become more hyperinflammatory, which can be detrimental to cells of the surrounding CNS ^96–98. Additionally, the elevated expression of antigen presentation protein, HLA-DR, by primary microglia has been associated with damage-associated pathologies ^{99, 100} and increased microglia reactivity ¹⁰¹. Overall, these findings suggest that POE primes iMGL precursors towards a reactive phenotype but are functionally impaired.

Normally, microglia are highly dynamic cells that survey the brain parenchyma for signs of damage, maintaining synaptic health, and responding to insults ¹⁰². Given the altered phenotype and function of iMGL with POE and maternal HCV infection, we further interrogated the transcriptional profiles of non-stimulated iMGL. Although fewer DEGs were noted among POE HCV− and HCV+ groups compared to controls, DEGs common to the POE groups mapped to pathways important for antimicrobial responses, cytokine production, and negative regulation of cell proliferation in non-stimulated iMGL. These results suggest that dynamic surveillance and homeostatic maintenance properties of microglia, are disrupted with POE. Our sPLSA analysis further indicates the expression of genes implicated in various neurological pathologies with POE in the absence of HCV. Aberrant expression of transcriptional regulators HDX and MN1 by human microglia has been linked to neurodegenerative diseases and poor synaptic plasticity ¹⁰³ while upregulated HSF5 by neuronal cells in mice is associated with neural inflammation and impaired responses to acute stress, such as hypoxia/ischemia ¹⁰⁴. iMGL exposed to POE and HCV infection exhibited the expression of many genes associated with neuro-developmental and -inflammatory conditions ^{105, 106}. Notably, PRR5 and LGALS9B are upregulated in primary microglia in response to disease associated signals leading to increased phagocytosis and secretion of pro-inflammatory cytokines ^{107, 108}. Our bulk-RNAseq analysis of non-stimulated iMGL indicate that support that POE induces a reactive phenotype that is further exacerbated by maternal HCV infection.

Finally, recent studies on primary murine microglia cells co-cultured with endogenous opioid peptide, enkephalin, have shown that opioids weakly activate TLR4 and significantly inhibit LPS-induced TLR4 activation ^{73, 74}. Our data suggest that POE and HCV are conditioning microglia precursors prenatally thereby impacting responses to subsequent challenges. In response to a secondary challenge, LPS, iMGL from the control group generated the largest transcriptional responses with DEG that mapped to multiple pathways important for healthy brain development, nervous system function, and inflammatory responses. Acute responses to LPS are critical for immune surveillance mechanisms and neuroprotective functions of microglia ^{71, 109}. In contrast, iMGL from the POE HCV+ group lack signatures associated with brain development and inflammatory responses. Previous studies in rodents have shown that sustained LPS stimulation results in a microglial refractory period resulting in synaptic and cognitive impairment ^{70, 110, 111}. Given that iMGL were derived from UCBMC chronically exposed to opioids and/or maternal HCV induced inflammation throughout gestation and that opioids have been shown to interfere with LPS-induced TLR4 activation ^{73, 74}, our findings suggest that POE preconditions microglia to be desensitized to challenges by tissue damage or inflammation. This immune tolerant iMGL phenotype with POE and maternal HCV is further supported by heightened inflammation at baseline (increased secretion of pro-inflammatory cytokines) but impaired functional responses (decreased phagocytic capacity and LPS responses) compared to control. Impaired microglia responses to stimulation can, in turn, impede remodeling of the surrounding CNS leading to compromised neurodevelopment ¹¹².

1.5. CONCLUSIONS

In summary, our study highlights the value of using a patient-specific model of UCB-derived microglia-like cells to interrogate the impact of maternal inflammation, such as POE and maternal HCV infection, on microglia development. This model could be used to interrogate mechanisms underlying altered neurodevelopmental outcomes with various maternal environmental exposures. We posit that POE primes early microglia precursors in utero towards an inflammatory state that impairs microglia development. Impaired microglia development is reflected by morphology (minimal ramifications, circularity, density) and the lower abundance of induced microglia-like cells in the POE groups. Furthermore, POE and HCV conditioning of iMGL are shown by elevated expression of markers important for microgliogenesis and inflammatory responses (PU.1, HLA-DR, IRF8) and higher production of pro-inflammatory cytokines (IL-1β, IL-6, IL-8). However, iMGL in both POE groups showed poor phagocytic ability, decreased expression of key markers of microglia function (TREM2, CD68, CD40, CD16) and decreased activation in response to LPS, suggesting immune-tolerant (desensitized) iMGL. These findings are further driven by maternal HCV status, supporting the need for early HCV interventions in pregnancy.

1.6. LIMITATIONS

While this model leveraging patient-specific iMGL provides an accessible window towards fetal neurodevelopment, some limitations exist. We were able to show that iMGL share a similar phenotype and transcriptional profile to primary microglia. However, this model is limited by its inability to fully recapitulate the microenvironment of the CNS. Indeed, recent studies have used neuron-microglia co-cultures to help bridge this gap ^113–115, however, further studies are needed to implement these co-culture models in the context of POE and fetal neurodevelopment. Furthermore, our study is limited by relatively lower cell counts compared to studies performed using cell lines. A limitation of our transcriptional profiling analysis by bulk-RNAseq is the inability to analyze our data on a single cell resolution. Future studies should implement single-cellRNA seq methods to better define differences in iMGL differentiation between groups. Finally, studies in the field of substance use disorders are challenged by the prevalence of polysubstance abuse, making the delineation of the impact of opioids difficult. While we were unable to quantify the level of opioid exposure to the fetus, we attempted to mitigate these variances by only including pregnant women diagnosed with severe opioid use disorder per DSM-V criteria and were stabilized on opioid replacement therapy (buprenorphine products). Future studies should be refined by using additional methods to quantify POE, including newborn urine/meconium testing and HPLC methods to detect the presence of opioid subtypes in the offspring.

Supplementary Material

Supplemental Figure 1

Supplemental Figure 1:

(A) Bar plots depicting the transcripts per million (TPM) of opioid and toll-like receptor genes across UCB monocyte, iMGL, and human primary microglia bulk-RNAseq datasets. (B) Bar graphs depicting additional microglia morphology features from IBA1-stained cells imaged at 40X and processed using ImageJ macro MicrogliaMorphology and MicrogliaMorphologyR package ⁴⁹. (C) Principal components analysis (*abs(R)>0.55; p<0.05) of iMGL clustered into three classes using k-means clustering. (D) Heatmap of cluster-specific measures on average; the measures were scaled across clusters. (E) Representative image of iMGL from each group highlighted to reflect cluster identification (blue = ameboid, red = ramified, yellow = biploar/rod. Any white structures showing in the image are either background or microglia structures cutoff in the image and not included in analysis. *=p<0.05, **=p<0.01, ***=p<0.001, ****=p<0.0001.

Supplemental Figure 2

Supplemental Figure 2

(A) Gating strategy used for UCBMC and iMGL phenotyping including representative gating from control (top), POE HCV− (middle), and POE HCV+ (bottom) iMGL groups. Cell frequencies shown in red were calculated from the single cell population. (B) Bar graphs depicting percent expression of canonical microglia markers by UCBMC and iMGL by flow cytometry. #=p<0.1, **=p<0.01, ***=p<0.001, ****=p<0.0001.

Supplemental Table 3

Supplemental Table 3: Gene lists supporting the sPLSDA analysis from bulk-RNA-seq data by POE ± HCV and with and without LPS treatment in Figure 6.

Supplemental Table 2

Supplemental Table 2: Gene lists supporting the iMGL bulk-RNA-seq analysis by POE ± HCV and with and without LPS treatment in Figure 5.

Supplemental Table 1

Supplemental Table 1: Gene lists supporting the UCB monocyte vs primary microglia vs iMGL bulk-RNAseq comparisons in Figure 2.

DATA AVAILABILITY

The datasets supporting the conclusions of this article are available on NCBI’s Sequence Read Archive: PRJNA970759 (UCB monocytes) ⁵³, PRJNA1216256 (UCB-monocyte derived microglia-like cells), and PRJNA387182 (primary human microglia) ⁵⁴.