key: cord-0310960-xeef5r6i
authors: Kim, Hye Sung; Xiao, Yang; Chen, Xuejing; He, Siyu; Im, Jongwon; Willner, Moshe J.; Finlayson, Michael O.; Xu, Cong; Zhu, Huixiang; Choi, Se Joon; Mosharov, Eugene V.; Kim, Hae-Won; Xu, Bin; Leong, Kam W.
title: Chronic opioid treatment arrests neurodevelopment and alters synaptic activity in human midbrain organoids
date: 2021-06-02
journal: bioRxiv
DOI: 10.1101/2021.06.02.446827
sha: abeb144b11df3504025263b619dd863785bc7a68
doc_id: 310960
cord_uid: xeef5r6i

The impact of long-term opioid exposure on the embryonic brain is crucial to healthcare due to the surging number of pregnant mothers with an opioid dependency. Current studies on the neuronal effects are limited due to human brain inaccessibility and cross-species differences among animal models. Here, we report a model to assess cell-type specific responses to acute and chronic fentanyl treatment, as well as fentanyl withdrawal, using human induced pluripotent stem cell (hiPSC)-derived midbrain organoids. Single cell mRNA sequencing (25,510 single cells in total) results suggest that chronic fentanyl treatment arrests neuronal subtype specification during early midbrain development and alters the pathways associated with synaptic activities and neuron projection. Acute fentanyl treatment, however, increases dopamine release but does not induce significant changes in gene expressions of cell lineage development. To date, our study is the first unbiased examination of midbrain transcriptomics with synthetic opioid treatment at the single cell level.

The opioid epidemic has reached crisis levels across the globe, with opioid use disorder (OUD) affecting 40.5 million people worldwide 1 . The perverse misuse of prescription opioids and heroin combined with the emergence of extremely potent fentanyl derivatives triggered a 10-fold increase in overdose fatalities in the United states from 2013 to 2018 [2] [3] [4] [5] . Indeed, the negative social, economic, and health ramifications of COVID-19 have further worsened this situation, with significant increases in opioid-related use, misuse, and non-fatal/fatal overdose after the start of the pandemic 6, 7 . An unfortunate consequence of this accelerated opioid use includes a parallel surge in the use of prescription opioids among pregnant women 8, 9 . Survey data found that 6.6% of women reported prescription opioid use during pregnancy, with 21.2% of those women reporting misuse, and 22-30% of women filling at least one prescription for an opioid analgesic during pregnancy 9, 10 . In addition to increased health care costs and adverse maternal outcomes, opioid use during pregnancy leads to an inevitable increase in the incidence of neonates exposed to opioids in utero [11] [12] [13] .

Prenatal opioid exposure, which includes both the use and misuse of prescription and illicit opioid drugs, causes unusual and deleterious symptoms in neonates including neonatal abstinence syndrome (NAS), small head circumference 14 , decreased cerebral volume 15 , microstructural brain injury 9 , and low birth weight 8, 9, 16 . Extensive clinical data also suggest that in utero opioid exposure has adverse effects on the developing organ systems, including the central nervous system 9, 17, 18 . The impaired neurodevelopment of these neonates are correlated with long-term issues, with follow-up studies revealing that individuals who had in utero opioid exposure experience problems with cognitive, behavioral, and developmental outcomes in their childhood 19, 20 . However, the precise cellular and molecular mechanisms underlying opioid-related neurodevelopmental disruptions in humans are unclear, and a better understanding of potential risks of long-term opioid exposure is crucial. This is especially urgent because synthetic opioids (i.e. fentanyl, methadone, and buprenorphine) themselves are frequently used to treat both pregnant patients with OUD and neonates with NAS, and thus advancements in therapeutics are absolutely necessary 21, 22 .

Efforts to understand the mechanisms underlying human neurodevelopment in response to opioids are fraught with challenges, such as limited accessibility to fetal brain specimens, ethical conflicts, and complex individual drug abuse history. Although pre-clinical studies conducted with animal models have revealed the detrimental effects of opioid use on the neurodevelopment of offspring 23 , inherent species differences have led to a general gap in neuropsychiatric clinical translation and a subsequent failure in drug development 24 .

Pharmacological therapeutics developed using animal models have also been particularly ineffective for the treatment of opioid abuse in humans because: 1) different species vary in their neurodevelopmental trajectories, receptor expression, and central opioid pharmacokinetics, and 2) animal models cannot capture the broader network of symptoms and environmental/social factors that are fundamental to drug abuse and addiction 25, 26 .

Moreover, although human postmortem brain specimens have the ability to directly assess the human-specific differences in gene expression, biomarkers, and neuroanatomy in drug abuse [27] [28] [29] [30] , cellular and molecular responses cannot be evaluated in a time-dependent manner.

Brain organoids are thus excellent potential candidates to bridge the gap between animal models and human studies. Brain organoids are self-assembled three-dimensional aggregates generated from human pluripotent stem cells in vitro that resemble the human fetal brain 31 . Unlike conventional two-dimensional cell cultures, organoid models better recapitulate the fetal brain in cell type composition, 3D architecture, and similar lineage trajectory [32] [33] [34] [35] . Brain organoids also provide a platform for unparalleled manipulation, enabling systematic studies of human neurodevelopment, disease modeling, and drug screening 33, [36] [37] [38] [39] [40] . Further, recent advances in guided organoid generation methods using lineage-specific patterning factors allow for the generation of micro-tissue mimicking certain brain regions such as the cerebral cortex 41, 42 , hippocampus 43 , and midbrain 36 .

In addition, brain organoids have revolutionized the characterization of brain development in combination with patient-derived iPSC engineering, genetic engineering, epigenetics, and single cell multiomics [44] [45] [46] [47] . Specifically, single-cell RNA sequencing (scRNA-seq) offers insights into underlying molecular signatures. Systematic RNA-seq analysis could identify early events of human neurogenesis and synaptogenesis that are limited to experimental access and validate cell types and neuronal activities 38, 48, 49 . For example, recent research identified 545 differentially expressed genes in the midbrains of opioid users compared to non-opioid users using bulk RNA-seq 27 . To date, however, cellular changes observed in distinct human neuronal subtypes upon opioid exposure have not been reported. Although multiple tested brain organoid generation models have been developed and validated, functional models are yet to be established to explore the temporal and spatial changes of neurons in response to a specific stressor, such as opioids. The susceptibility of newborns and adults to develop altered opioid responses thus remains under-investigated and poorly understood.

To our knowledge, this is the first study to investigate the influences of chronic opioid exposure on molecular and cellular alterations during neurodevelopment using a human midbrain organoid model. Based on a previously reported organoid generation protocol 36 , we established our midbrain organoids as a viable model to study neurodevelopment, as well as developed a framework for evaluating cell type-specific transcriptomic responses to fentanyl exposure in these organoids. Moreover, our comprehensive molecular characterization shows pseudo-time metrics of midbrain organoid development and offers an in-depth analysis of affected gene features and pathways under a chronic fentanyl exposure condition.

Interestingly, chronic fentanyl treatment arrests ectodermal cells at the stage of neural progenitors or neuroblasts from progressing into neurons, while acute fentanyl treatment does not cause significant transcriptomic changes associated with neurodevelopment. In addition, although less neurons were observed in the fentanyl-treated organoids, the average gene expression level of each neuron is higher in neuron projection, synapse assembly, and neurotransmitter dynamics. Furthermore, neurodevelopment resumed in lineage-related gene expression when fentanyl was withdrawn. Our investigation into the arrested neurodevelopment in fentanyl-treated organoids, in addition to further advancements in organoid technology, can help inform improved therapeutics for opioid abuse and prenatal opioid exposure.

We generated midbrain organoids using a protocol adapted from Kriks et al. and Jo et al. with minor modifications (Figure 1A , Methods) 36, 37 . Human iPSCs were dissociated to single cells to form uniformly-sized embryoid bodies (EBs) approximately 300m in diameter in nontreated U-bottom 96-well plates (Figure 1B) . At day 7-14, EBs showed uniform neuroectoderm formation along the outer surface of EBs where optically translucent and radially organized (Figure 1B (inset) ). These neuroectoderm-containing spheroids were cultured in N2 neuronal media supplemented with neurotrophic factors (BDNF, Ascorbic acid, GDNF, cAMP, TGF-b3, and DAPT) once neural induction was achieved in the first 9 days ( Figure 1A, Methods) . The organoids grew up to 0.8~1.2mm in diameter after 21 days ( Figure 1C) . The cytoarchitecture of the organoids were examined at day 21 by immunocytochemical analysis (Figure 1D) . The midbrain progenitors expressing OTX2 or FOXA2 proteins were located closer to the core of the organoids, with OTX2 expression specifically found in the apical surface and extending to the intermediate region of the neuroepithelia ( Figure 1D and Figure S1B ). In contrast, dopaminergic (TH positive) and

GABAergic (GABA positive) neurons were detected along the outer edge of the organoids, which is similar to the layering of the human midbrain development (Figure S1A ) 50 .

To further confirm whether organoids specifically developed towards the midbrain specification, we monitored the time course of a set of marker genes for 35 days by RT-PCR ( Figure S1C) . First, the expression of pluripotency markers, OCT4 and NANOG, drastically decreased upon neuronal induction, while pan-neuronal marker expression, TUJ1 and MAP2, gradually increased. In contrast, the expression of OTX2, a homeodomain transcription factor required for patterning the midbrain region 51 , increased after 7 days of floor plate induction.

The co-expression of the floor-plate (FP) marker FOXA2 and the roof plate marker LMX1A, a unique feature of midbrain development 37 , was also observed. Furthermore, we observed robust expression of various midbrain specific neuronal markers. For example, the expression of midbrain dopaminergic neuronal markers (TH, DAT and PITX3) was upregulated during the differentiation process. On the contrary, expression levels of PAX6, a forebrain marker, and TBR2 and GBX2, hindbrain markers, were low and not significantly changed. The immunocytochemical staining results also exhibited a decrease in the number of OTX2+ cells and an increase in the number of tyrosine hydroxylase (TH+) from day 7 to day 35 of the differentiation ( Figure S1A) . These results indicate that cells of the midbrain organoids gradually transitioned from proliferating neuroprogenitors into more mature midbrain-specific neurons.

To comprehensively determine the cell composition of the iPSC-derived midbrain organoids, single cell RNA sequencing was performed and annotated according to the cell type signatures of human fetal ventral midbrain (6-11 weeks human embryos) previously reported by Manno et al. 52 Cells at two differential time points, day 53 and day 77, were analyzed. At each time point, cells from three organoids of each condition were pooled. A total of 25,510 single cells was analyzed (Table S1) progenitor types decreased from 59% to 50%, whereas the percentage of neuronal and mesoderm-derived cell types increased from 35% to 38% and 6% to 12%, respectively ( Figure S2A ). This result further indicates the subsequent differentiation of neuroprogenitors into neurons in the developing organoids.

To determine the electrophysiological properties of the cells, at day 35 we immobilized the organoids on a Matrigel-coated coverslip and performed whole-cell current-clamp recordings from cells on or near the surface of the organoid. Biocytin dye was added to the patch pipette allowing to confirm in post-fixed organoids that recorded cells were TH+ ( Figure 1H ).

Electrophysiological analysis revealed that ~20% of the recorded neurons (4 of 21 cells) fired spontaneous action potentials with an average frequency of 0.6 ± 0.3 Hz (Figure 1I) , which is lower than the ~4 Hz autonomous firing frequency of mature mouse DA neurons. Injection of a +10 pA step current (Figure 1J , black) produced a train of stable action potentials, but higher current steps (+30 pA (red) and +50 pA (blue)) yielded action potentials with a higher frequency and lower amplitude that quickly ceased. These results suggest that the dopaminergic neurons of the organoid are not fully matured at day 35 of differentiation.

Further, the average resting membrane potential of the neurons in the organoids was around -40mV, which is higher than that of mature neurons (~-60 mV) ( Figure 1K) . Similarly, dopamine synthesis and release were undetectable in untreated organoids, but increased rapidly in the presence of dopamine precursor L-dihydroxyphenylalanine (L-DOPA), demonstrating the ability of neurons to produce the neurotransmitter ( Figure 1E ). Taken together, these results suggest that midbrain organoids are similar to early-stage human fetal midbrains and could thus be used as a tool for understanding the influence of chronic opioid exposure on neurodevelopment.

First, we confirmed that organoids express opioid receptors, including mu (OPRM1), kappa (OPRK1), and delta (OPRD1). The mRNA expression levels of opioid receptors drastically increased until day 35 and saturated after that point (Figure 2A ). In addition, the opioid receptors were detected where TH+ or GABA+ neurons were located ( Figure 2B) , indicating that the midbrain-specific neurons of the organoid express opioid receptors.

Acute opioid administration occurs in many clinical situations, including as pain relief during labor and delivery. To investigate the impact of short-term opioid treatment on the fetal brain (i.e., acute opioid exposure in utero), the midbrain organoids were treated with synthetic opioids (fentanyl, oxycodone, and hydrocodone) for 4 hours ( Figure S4 ). The 4-hour treatment of opioids up-regulated the bulk mRNA expression levels of all three opioid receptors by 2 to 4-fold as compared to untreated organoids ( Figure 2C ). In addition, the intracellular and extracellular levels of dopamine also increased in organoids treated for 4 hours with fentanyl and for a subsequent hour with L-DOPA ( Figure 2D ). Measurements with Fluo-4 AM showed that the frequency and amplitude of spontaneous calcium transients were attenuated in cells in fentanyl-treated organoids ( Figure 2E , 2F and Movie S1), consistent with the mechanism underlying opioids' modulation of signal transmission and pain perception 53 .

Next, we performed single cell transcriptome analysis of fentanyl-treated organoids. We Interestingly, when we examined the single cell expression of previously reported opioid response genes 27, 54 , synaptic plasticity genes (SNCG, SV2A) were expressed at a lower level in the FTY group ( Figure 2G ). This indicates that synapse transmission was downregulated by acute opioid treatment. Gene set enrichment analysis, however, showed no significant change in biological pathways, possibly because we sequenced the organoids on the same day of the acute treatment and ground level changes in gene expression profiles

were not yet established. Overall, our results suggest that even a short-term opioid treatment elicits cell-type specific responses in the midbrain, but does not significantly alter the gene profiles in neuronal lineage specification and neuron projection.

Several neuroimaging and animal studies have shown the adverse effects of prenatal opioid exposure on neurodevelopment 26, 55 , including dysregulated functional connectivity and decreases in brain volume across multiple regions such as the midbrain, and many others have linked this exposure to clinical observations of delayed neurodevelopment 19, 20, 56 . Therefore, we sought to investigate the effects of chronic fentanyl treatment using our midbrain organoid model, in which organoids were treated with 74 nM fentanyl for 53-77days starting from the first day of floor plate induction ( Figure 1A ). We first examined the morphology of the midbrain organoid over its development. Our time course analysis of midbrain organoids indicated that chronic fentanyl exposure did not change the gross morphology nor the size of the organoids during 77 days of culture ( Figure 3A) , possibly because organoid growth is eventually limited to a certain size anyway without the support of a 3D hydrogel scaffold and perfusable vasculature. Furthermore, there were no significant changes in the distribution pattern of OTX2 and TH proteins in the midbrain organoids, forming rosette-like regions ( Figure 3B ).

Next, we explored whether the differentiation and maturation of various cell types in the midbrain organoids were affected by chronic opioid exposure. Interestingly, the expression level of both midbrain progenitor markers, FOXA2 and LMX1A, and pan-neuronal markers, MAP2, was largely lower in the chronic fentanyl treatment group than in the untreated organoids ( Figure 3C ). In contrast, there was no significant difference in the expression of other genes related to pluripotency (OCT4, NANOG), midbrain (DAT, PITX3), forebrain (PAX6), or hindbrain (TBR2, GBX2) regional identity at the bulk mRNA level, as measured by RT-qPCR ( Figure S6A ). All three opioid receptors in the chronic fentanyl exposure group were expressed much earlier than in untreated organoids ( Figure 3C) Furthermore, we calculated RNA dynamics (i.e. rates of transcription, splicing, and degradation of individual genes) based on the mapped spliced and unspliced mRNA reads with scVelo ( Figure S8A-B, Methods) . By quantifying the connectivity of cell clusters, partition-based graph abstraction (PAGA) provided a simple abstract graph of cell fate connectivity ( Figure 3H) . We double checked the pseudotime projection of our dataset and plotted out velocity streamlines of RNAs in both a normal brain organoid ( Figure S9A ) and a growth-impaired organoid ( Figure S9B ). In theory, quiescent stem cells and terminally differentiated cells (e.g. neurons) show close to zero velocity values as they are at a steady state of spliced and unspliced mRNA levels. The size of the stream line arrows was in proportion to the value of the cell's velocity. We observed low velocity fields in a subpopulation of Prog and Rgl, as well as in Gaba and DA (Figure S9A-B) , which is consistent with the findings of previous studies 48, 59 . In addition, phase portraits of neuronal specification markers and mature neuron function markers showed that chronic fentanyl exposure arrested cells at an early stage of neurodevelopment ( Figure 3J-K, S9C-D) . Although radial glia-like cells were identified in midbrain organoids, their specification and activity to give rise to certain

neurons are yet to be clarified.

Overall, our trajectory maps illustrated that: 1) neuron specification and maturation was 

To find the master regulators that control cell fate decisions, we performed regulatory network inference and clustering on scRNA data using SCENIC R package 49 (Methods).

Instead of looking at individual genes in the differentially expressed matrix, this method examines gene networks, taking cis-regulatory motifs and transcription factors into account.

In brief, we scored the activity of each regulon in all single cells and identified the top transcription factor networks ( Figure 4A, Figure S10 ). Both untreated and fentanyl-treated samples showed a significant enrichment in CREB5 (25 predicted target genes) and MAX (10 predicted target genes) regulons ( Figure 4B, Figure S5 ). CREB5 is a cyclic-AMP response element binding (CREB) protein that has roles in neuronal survival, differentiation, memory, and drug addiction 60 . MAX regulates more than 180 genes in the human telencephalon and contributes to the development of all regions in the brain 61 

We analyzed the heterogeneous cell type-specific response to chronic fentanyl exposure based on single cell RNA sequencing results. Differentially expressed genes between D77_UT and D77_FTY were identified by using the "FindMarkers" function in Seurat. Both samples expressed high levels of VGF, SCG2, VAMP2, and RAB3A in various neuronal types (DA, Gaba, OMTN, Sert, RN) ( Figure S11, S12) , which are secretory-or vesicle-associat ed proteins that regulate neurotransmitter release 64, 65 In the D77_FTY sample, gene sets related to synaptic activity, neuron projection, and neurotransmitter transport were highly enriched in neurons (DA, Gaba, OMTN). Differential responses between DA and Gaba were not observed (Figure 4I, S11) , likely because neural circuits were not yet well established in organoids. In addition, progenitors (Rgl, Prog, NProg, Nb) in D77_UT exhibited high regulon activity in neuron specification and migration, while progenitors in D77_FTY showed high activity in neurogenesis, differentiation, and axongenesis ( Figure S10B ). This indicates that progenitors in the D77_FTY arrested at a much earlier stage than those in D77_UT.

Mesoderm-derived cells (pericytes, endothelial cells, and microglias) in untreated midbrain organoids (D77_UT) were enriched with genes that mediate endothelium development, angiogenesis, and extracellular structure organization. In chronic fentanyltreated organoids, we observed an elevated expression in genes related to immune response (PLCG2, TFF3, CRYAB) and homeostasis (EPYC, SPINK6) ( Figure S13) . Interestingly, pericytes were highly responsive to fentanyl treatment, which is demonstrated by the significant transcriptome changes as compared to untreated pericytes (Figure S3A, S12A) . Furthermore, the extracellular dopamine level of organoids with the chronic fentanyl treatment was higher than that of untreated organoids (Figure S6B) , but not significantly. The organoids with chronic fentanyl treatment also showed less neurons with calcium signaling ( Figure S6C and Movie S2), indicating that cellular excitability and signal transmission were affected by fentanyl.

To determine if and to what degree the impact of fentanyl exposure on neurodevelopmental program can be reversed when fentanyl is withdrawal after chronic exposure, we alleviated organoid stress by withdrawal of fentanyl for 2 days. We performed similar differential expression analysis between the untreated sample (D77_UT) and withdrawal sample (D79_WD), and found that regulon activities (FOXJ1, RFX4, SOX2, BARHL2) of neuronal differentiation and subtype specification were substantially restored in progenitor cells (Prog, NProg, Rgl) (Figure 5A-B) . We also observed an increased expression of ventral midbrain progenitor genes (LMX1A), subtype specification genes (NKX6-1, FOXA2), and mature neuron markers (RBFOX3, NFE2L1, HMGN3) in the D79_WD sample as compared to D77_FTY (Figure 5C ). Gene set enrichment analysis showed that the synaptic activity and neuron projection in D79_WD were more similar to D77_UT than D77_FTY (Figure 5D) , indicating that the synaptic activity decreased to a basal state in response to drug withdrawal. All the above data suggest that neural development resumes in transcriptomes when opioids are withdrawn.

In utero opioid exposure has been reported to impair neurodevelopment and potentially cause poor neurocognitive, behavioral, and developmental outcomes in childhood 19, 20 .

However, the mechanism underlying this dysregulation is unclear, and more focused research must be done. One of the primary adult neural circuits involved in opioid action is the mesolimbic reward system. This system generates dopamine signals that arise in the ventral tegmental area (VTA) in the midbrain and propagate throughout the rest of the brain, playing a role in learning and reward pathways. Drugs of abuse, including opioids, take advantage of this system, inducing dopamine surges and relieving pain through acute opioid exposure 66 . Chronic opioid exposure, however, does not simply have an exaggerated effect -it disrupts dopamine signaling and causes transcriptional and epigenetic changes in the brain regions within the mesolimbic system, promoting addiction and vulnerability to relapse 66, 67 . Although extensive clinical and preclinical data have demonstrated that opioid addiction is strongly associated with the mesolimbic system in an adult brain, little is known about the particular association in the developing human fetal brain. We therefore aimed to use midbrain organoids to investigate the molecular and cellular alterations caused by chronic opioid exposure in the midbrain region, which is the largest dopamine-producing area in the brain and heavily involved in the mesolimbic reward system, during neurodevelopment.

Our midbrain region-specific organoids were generated by a guided method using midbrain patterning molecules. Dual-SMAD inhibition factors, LDN 193189 and SB431542, and a Wnt pathway activator, CHIR 99021, instructed iPSC differentiation toward floor plate lineages. Subsequent treatment of a small molecule agonist, purmorphamine, and recombinant SHH boosted floor plate cells toward ventral mesencephalic lineages. We compared the gene expression profiles of organoids in multiple independent batches ( Figure   S1 ) that were generated using two different iPSC lines (i.e. FA11 and QR19). No significant variations were found, suggesting that our guided method could generate reliable and consistent midbrain organoids. In addition, our method does not involve embedding the organoids in matrices (e.g. Matrigel) for high reliability and reproducibility of organoid generation 41, 68 .

These midbrain organoids also exhibited the cytoarchitecture and compositional features unique to the developing human ventral midbrain region in vivo 52, 69 . The electrophysiological properties of the organoids at day 35 indicate that neurons located at the surface of the organoids are not fully mature, as evidenced by their depolarized resting membrane potential, slow or absent pacemaking activity, and inability to maintain long action potential trains compared to mature neurons in an adult human brain 41, 70, 71 . This is not surprising, however, as organoids model early stages of human embryonic development, in which fully realized electrophysiological properties, such as action potentials and spontaneous synaptic transmission, are not yet observed and ontogenetic changes still occur 72 . Indeed, electrophysiological properties of neurons in midbrain organoids can gradually become more mature 71 , and cortical organoids have been shown to parallel the timeline of neurodevelopment in vivo, using genome-wide analysis of the epigenetic clock and transcriptomics to demonstrate that organoids reached postnatal stages around day 300 of culture in vitro 68 . We believe that midbrain organoids could develop and mature over time to similarly reflect in vivo neurodevelopment.

Once midbrain organoids were established as viable models to investigate midbrain neurodevelopment, fentanyl, a potent synthetic opioid, was used to treat the organoids. Acute 

Challenges remain in uncovering native brain functionality and correlating organoid-based findings to clinical readouts. Primarily, brain organoids still lack the full cellular maturation, heterogeneity, architecture, vasculature, and microenvironmental niche seen in the actual human brain. 44, 75, 76 This limits the extent to which organoids can be a fully realistic model and demands that organoids undergo further characterization to better understand these limitations. Recent advances in spatial multi-omics, along with high-throughput transcriptomic and proteomic profiling, offer great opportunities to characterize the developmental and functional dynamics of brain neurogenesis and synaptogenesis in organoids. [77] [78] [79] Future studies could also validate these models through human studies or animal models.

Another challenge is the scope of region-specific organoids. Neuropsychiatric disorders, including substance abuse, affect several brain regions and implicate complex neural circuits. [80] [81] [82] Midbrain organoids, although providing valuable information regarding the effects on the midbrain, lack the functional neural circuits found in the native central nervous system (CNS) and can inherently only tell part of the story. Until further advancements in brain organoid development, systematic studies using several different region-specific organoids are important to fill out the picture.

A final challenge is that brain organoids can only reveal the effects of a single drug abuse and identify those potential biomarkers and therapeutic targets. In the event of polydrug abuse, these models are unable to match between specific downstream effects and their drug of origin. Multiples studies must be done in parallel, each investigating an individual drug abuse, to appreciably gain a mechanistic understanding of each drug.

Overall, long-term exogenous opioid treatment elicits massive transcriptomic profile changes in neurodevelopment and neuronal communication. Future studies in correlating the clinical readouts with transcriptomic markers would establish comprehensive evaluation of drug abuse outcomes and offer opportunities in therapeutic targets.

In midbrain organoids, a short fentanyl treatment induced an increase in dopamine release without significant changes in the gene expressions of cell lineage development. In contrast, chronic fentanyl exposure impaired cell subtype specification and altered synaptic activities in neurons, indicating arrested neurodevelopment. Upon fentanyl withdrawal, the neurodevelopment resumed to normal at the transcriptomic level. Overall, our study dissected the molecular landscape of opioid responses in neural and non-neural cells and unveiled the affected pathways that may inform strategies to boost treatments in substance abuse and neonatal abstinence syndrome.

The FA11 iPSC line was generously provided by the Columbia Stem Cell Initiative with an approved IRB. FA11 cells were derived from human dermal fibroblasts of a healthy donor 83 .

The iPSCs were maintained under feeder-free conditions over Matrigel-coated plates in mTeSR Plus media (Stemcell Technologies). The media was changed daily and the iPSC cultures were split into 1:6-1:10 every 5 days using ReLeSR (Stemcell Technologies).

iPSCs before passage 30 were used to generate the organoids. To form embryoid bodies (Figure 1A) . Detailed product information could be found in Table S2 .

On day 0, we used neuronal induction medium, which composed of 15% Knockout serum replacement (Gibco), 1% GlutaMax (Gibco), 1% minimum essential media-nonessential amino acids (MEM-NEAA) (Gibco), and 0.1% -mercaptoethanol (Gibco) in Knockout DMEM/F12 (Gibco), supplemented with 100nM LDN193189 (Stemgent) and 10M SB431542 (Tocris Bioscience). On day 2, we changed to the neuronal induction media that was supplemented with LDN193189, SB431542, 100ng/ml SHH (R&D Systems) and 2M Purmorphamine (Calbiochem). On day 4, 3M CHIR99021 (Stemgent) was added to the medium. From day 6 to day 8, the basal media containing LDN193189 and CHIR99021 was gradually replaced to a differentiation media containing DMEM/F12: Neurobasal (1:1) (Gibco), N2 supplement (Gibco), B27 supplement without vitamin A (Gibco), 1% GlutaMAX, 1% MEM-NEAA, 0.1% -mercaptoethanol and 1% penicillin-streptomycin (Gibco). On day 9, the differentiation media was supplemented with CHIR99021, 10ng/ml BDNF (Prospec), 0.2mM ascorbic acid (Sigma), 20ng/ml GDNF (Prospec), 0.2mM dibutyryl-cAMP (Calbiochem), 1ng/ml TGF-3 (R&D Systems), and 10M DAPT (Tocris). From day 11, the organoids were cultured in the same media without CHIR99021. Media was then changed every 3-4 days.

On day 21, the organoids were transferred into non-treated 24-well plates by using a 200l tip with a wide bore opening. The organoid growth was monitored by bright-field microscopy (Nikon Eclipse) for 55 days, and the diameter was measured using Image J. nM, 19 nm, 37 nM and 74 nM, respectively, for 4 hours ("acute" opioid treatment). The cytotoxicity was assessed by CellTiter-Glo ® Luminescent Cell Viability Assay (Promega). The changes in the mRNA expression of opioid receptors and the dopamine release levels caused by opioid treatment were evaluated by real-time PCR and HPLC analysis, respectively, as described below. Fentanyl (74 nM in culture media) was also used to treat organoids starting on day 1 of the neuronal induction ("chronic" opioid treatment). On day 77, to analyze the effects of opioid withdrawal, media was replaced to fresh media without fentanyl, and the organoids were incubated for two days prior to single cell RNA sequencing analysis. (Table   S1 )

The organoids were fixed in 4% paraformaldehyde for 1 hour, saved in a 30% sucrose solution in PBS overnight, and subsequently embedded in O.C.T. compound (Tissue-Tek) for cryosectioning. Frozen organoids were cryosectioned at a thickness of 20m. We heated the slides up to 95 ºC in a citrate buffer (10mM Sodium Citrate, 0.05% Tween 20, pH 6.0) for antigen retrieval. For immunohistochemistry, the slides of organoid cryosections were permeabilized with 0.2% TritonX-100 in PBS 20min at room temperature, and then blocked with 3% Bovine Serum Albumin (BSA, Sigma) with 0.1% Triton X-100 in DPBS for 1 hour at room temperature. The sections were incubated with primary antibodies overnight at 4 ºC and with secondary antibodies for 1 hour at room temperature (Table S2) . All sections were stained with DAPI (Sigma) for cell nuclei. Images were taken on a confocal microscope (LSM 710, Zeiss).

Gene expression profiles of the organoids during development were evaluated for 35 days by real-time (RT) PCR. Total RNAs were isolated from organoids using TRIzol ® reagent (Invitrogen) and reverse-transcribed using iScript TM cDNA synthesis kit (Bio-Rad) to produce cDNA. Quantitative RT-PCR was performed using PowerUp TM SYBR TM Green Master mix (ThermoFisher) and QuantStudio TM 3 Real-Time PCR System (Applied Biosystem). ∆∆Ct method was applied to normalize expression levels of each gene to that of GAPDH. The sequences of primers were described in Table S3 .

Whole-cell patch clamp recordings were taken from organoids as previously described [87] [88] [89] . Recordings were performed with an Axopatch 700B amplifier (Molecular Devices) and digitized at 10 kHz with an ITC-18 (HEKA Instruments). Data analysis was performed using Clampfit 10 software (Molecular Devices, Sunnyvale, CA) and Matlab 8.0 (MathWorks, Natick, MA). In each neuron, input resistance, resting membrane potential, and spontaneous firing frequencies were monitored throughout the recording, and only cells with neuronal morphology and a stable baseline activity that continued for >5 minutes were counted as tonically active. In order to determine the identity of recorded neurons, neurons were labeled with biocytin (Sigma-Aldrich) added to the patch pipette saline for >10 minutes, fixed in 4% PFA for 1 hour and immunostained with anti-TH antibodies.

On day 180 of culture, both fentanyl-treated and untreated organoids were used for calcium imaging. The organoids were incubated with the cell-permeable calcium indicator Fluo-4 AM (ThermoFisher) for 30 min at 37°C. Time-lapse changes in Ca 2+ levels in live organoids were imaged using a fluorescence microscope (Eclipse TS100, Nikon). To see the instant effect of the fentanyl treatment on the calcium channel, untreated organoids were recorded before and after the addition of fentanyl (74 nM in culture media). The organoids were imaged every 2 seconds at room temperature. The fluorescence intensity was determined by Image J and the data were normalized to ∆F/F0 using the following equation: y = (F∆sec -F0sec)/F0sec.

Intracellular and extracellular dopamine production from the midbrain organoids was Organoid dopamine levels were then determined by HPLC with electrochemical detection, as previously described. 90, 91 In brief, samples were separated on a VeloSep RP-18, 3 M, 100x3.2mm column (PerkinElmer, Waltham, MA) with a Gilson 307 HPLC piston pump set to a flow rate of 0.7 mL/min and a mobile phase containing 45 mM NaH2PO4, 0.2 mM EDTA, 1.4 mM HSA, 5% methanol, and pH 3.2. Dopamine was detected on an ESA Coulochem II electrochemical detector at 350 mV oxidation potential. Data was then collected using Igor software, and dopamine concentration was calculated from areas under the HPLC peaks using calibration curves.

Five independent samples of midbrain organoids (Table S1) Table S5 .

The Seurat package (V3.0) in R (V3.6.3) was used to normalize the expression matrix and identify differentially expressed genes. The Glmnet package (V4.0) was used to assign cell types based on cells' likelihood to an annotated human fetal midbrain dataset 52, [92] [93] [94] [95] . First, we transformed both the experimental and reference datasets' gene counts separately using the "sctransform" function in Seurat.Secondly, we used Seurat's CCA integration method to account for technical differences between both datasets, and co-clustered both datasets in UMAP 58 .Then, we trained a multinomial logistic regression classifier (Glmnet R package) on the reference's integrated data excluding the "unknown" ("Unk") cell type and predicted the cell type of both datasets. In non-Unk reference cells, the classifier's accuracy was 99%. To validate the cell types assigned, we compared the gene signatures of each cell type between each dataset using highly correlated or anti-correlated genes to predict the cell types in our samples. Our results showed similar gene patterns compared to the reference dataset.

Overall, we found that the cell type composition of the reference dataset and our own dataset were similar, with some variations in percentages of progenitor cells, neuroblasts, and oligodendrocyte progenitor cells (OPCs) (Figure 1F, S2) . We grouped the sub-clusters of a same cell type into one major type. For instance, any cell type with hDA in the name was mapped to the DA major type. Any cell type with Prog in the name was mapped to the Prog major type. Any cell type with hNb in the name was mapped to the Nb major type. A detailed table of cell metadata table was deposited in Gene Expression Omnibus (GEO) under accession number GSExxxxxx (This number will be available following the acceptance of the manuscript).

Differentially expressed genes between samples or cell types were identified by using the "FindMarkers" function in Seurat. The following parameters were used: min.pct = 0.25, logfc.threshold = log(2). The top genes in each comparison group were selected with a prefiltering step (p_val_adj < 1e-10, pct.1 > 0.1, pct.2 > 0.1, avg_logFC>5 or <-5) to remove low count and insignificant genes, and then ranked by the absolute value of log fold change.

Heatmap of top differentially expressed genes were shown in Figure S3 .

We performed ssGSEA to identify the activated pathways in our samples using the pipeline previously described in identifying subtypes with enriched gene sets 96 

The Scenic package (V1.2.2) in R(V3.6.3) was used with default settings to identify key cell fate decision transcription networks. 99 In brief, sets of genes that were co-expressed with transcription factors were identified using the GENIE3 module. We kept the top five transcription factors for each target gene. Then, the RcisTarget module sought out the enriched cis-regulatory motifs of candidate transcription factors and predicted the candidate target genes. Lastly, we used the AUCell algorithm in the SCENIC package to score the activity of each regulon in each cell. We grouped the scores by cell types and plotted their RegulonAUC values in the heatmaps to show the regulon activities of master regulators ( Figure 4A-B, 5A, Figure S5 ). This method is robust against normalization methods and cell dropouts, as we got consistent results after testing multiple runs of the same datasets under different normalization pipelines.

The single-cell RNA sequencing data will be deposited in the Gene Expression Omnibus (GEO) following the acceptance of the manuscript. 

Editorial: Challenges to Opioid Use Disorders During COVID-19

Regional differences in the drugs most frequently involved in drug overdose deaths: United States

Notes from the field: overdose deaths with carfentanil and other fentanyl analogs detected‫01½°ﻗ‬ states

Drug overdose deaths involving fentanyl

Drug and opioid-involved overdose deaths‫½°ﻗ‬United States

The Opioid Epidemic During the COVID-19 Pandemic

The Opioid Epidemic Within the COVID-19 Pandemic: Drug Testing in 2020

Public Health Surveillance of Prenatal Opioid Exposure in Mothers and Infants

Prenatal opioid exposure: The next neonatal neuroinflammatory disease

Vital Signs: Prescription Opioid Pain Reliever Use During Pregnancy -34

Prescription opioid epidemic and infant outcomes

The opioid epidemic and neonatal abstinence syndrome in the USA: a review of the continuum of care

Identification of Prenatal Opioid Exposure Within Health Administrative Databases

Impact of maternal drug dependency on birth weight and head circumference of offspring

Volumetric cerebral characteristics of children exposed to opiates and other substances in utero

The prevalence and impact of substance use disorder and treatment on maternal obstetric experiences and birth outcomes among singleton deliveries in Massachusetts

Association of fentanyl with neurodevelopmental outcomes in very-low-birth-weight infants

The neurodevelopmental impact of neonatal morphine administration

Cognitive and Behavioral Impact on Children Exposed to Opioids During Pregnancy

Early life stress and environmental influences on the neurodevelopment of children with prenatal opioid exposure

Treating pregnant women dependent on opioids is not the same as treating pregnancy and opioid dependence: a knowledge synthesis for better treatment for women and neonates

Modeling prenatal opioid exposure in animals: Current findings and future directions

Neuroimaging in infants with prenatal opioid exposure: Current evidence, recent developments and targets for future research

Differentially expressed gene networks, biomarkers, long noncoding RNAs, and shared responses with cocaine identified in the midbrains of human opioid abusers

A molecular profile of cocaine abuse includes the differential expression of genes that regulate transcription, chromatin, and dopamine cell phenotype

Transcriptome organization for chronic alcohol abuse in human brain

Substance-specific and shared transcription and epigenetic changes in the human hippocampus chronically exposed to cocaine and alcohol

Bioinspired Hydrogels for 3D Organoid Culture

Modeling Development and Disease with Organoids

The use of brain organoids to investigate neural development and disease

Brain organoids: advances, applications and challenges

Modeling SARS-CoV-2 infection in individuals with opioid use disorder with brain organoids

Midbrain-like Organoids from Human Pluripotent Stem Cells Contain Functional Dopaminergic and Neuromelanin-Producing Neurons

Dopamine neurons derived from human ES cells efficiently engraft in animal models of Parkinson's disease

Establishing Cerebral Organoids as Models of Human-Specific Brain Evolution

Generation of human striatal organoids and cortico-striatal assembloids from human pluripotent stem cells

Functional cortical neurons and astrocytes from human pluripotent stem cells in 3D culture

Reliability of human cortical organoid generation

Generation of functional hippocampal neurons from selforganizing human embryonic stem cell-derived dorsomedial telencephalic tissue

Building Models of Brain Disorders with Three-Dimensional Organoids

Cell diversity and network dynamics in photosensitive human brain organoids

Synthetic Analyses of Single-Cell Transcriptomes from Multiple Brain Organoids and Fetal Brain

Human cerebral organoids recapitulate gene expression programs of fetal neocortex development

Generalizing RNA velocity to transient cell states through dynamical modeling

SCENIC: single-cell regulatory network inference and clustering

The embryonic midbrain directs neuronal specification of embryonic stem cells at early stages of differentiation

Molecular Diversity of Midbrain Development in Mouse, Human, and Stem Cells

Single cell transcriptomics reveals opioid usage evokes widespread suppression of antiviral gene program

Prenatal opioid exposure is associated with smaller brain volumes in multiple regions

Retrospective review of neurodevelopmental outcomes in infants treated for neonatal abstinence syndrome

The single-cell transcriptional landscape of mammalian organogenesis

Uniform Manifold Approximation and Projection for Dimension Reduction

Heterogeneity of Human Neuroepithelial Cells and Early Radial Glia

CREB signalling in neural stem/progenitor cells: recent developments and the implications for brain tumour biology

Associating transcription factors and conserved RNA structures with gene regulation in the human brain

BARHL2 transcription factor regulates the ipsilateral/contralateral subtype divergence in postmitotic dI1 neurons of the developing spinal cord

Reciprocal autoregulation by NFI occupancy and ETV1 promotes the developmental expression of dendrite-synapse genes in cerebellar granule neurons

Recapitulating physiological and pathological shear stress and oxygen to model vasculature in health and disease

Mesolimbic dopamine signaling in acute and chronic pain: implications for motivation, analgesia, and addiction

Neurobiologic processes in drug reward and addiction

Long-term maturation of human cortical organoids matches key early postnatal transitions

The role of developmental transcription factors in adult midbrain dopaminergic neurons

Assembly of functionally integrated human forebrain spheroids

Brain-region-specific organoids using mini-bioreactors for modeling ZIKV exposure

A single-cell RNA-seq survey of the developmental landscape of the human prefrontal cortex

Quadrato G, Brown J and Arlotta P. The promises and challenges of human brain organoids as models of neuropsychiatric disease

Modeling neurological disorders using brain organoids

Simultaneous epitope and transcriptome measurement in single cells

Highly sensitive spatial transcriptomics at near-cellular resolution with Slide-seqV2

High-Spatial-Resolution Multi-Omics Sequencing via Deterministic Barcoding in Tissue

Diagnosis and treatment of neuropsychiatric disorders

Neurobiology of substance use in adolescents and potential therapeutic effects of exercise for prevention and treatment of substance use disorders

The neurobiology of addictive disorders

Highly efficient neural conversion of human ES and iPS cells by dual inhibition of SMAD signaling

Wnt antagonism of Shh facilitates midbrain floor plate neurogenesis

Generation of regionally specified neural progenitors and functional neurons from human embryonic stem cells under defined conditions

Parkin and PINK1 Patient iPSC-Derived Midbrain

Dopamine Neurons Exhibit Mitochondrial Dysfunction and alpha-Synuclein Accumulation

alpha-Synuclein-Dependent Calcium Entry Underlies Differential Sensitivity of Cultured SN and VTA Dopaminergic Neurons to a Parkinsonian Neurotoxin

iPSC-derived dopamine neurons reveal differences between monozygotic twins discordant for Parkinson's disease

Alpha-synuclein overexpression increases cytosolic catecholamine concentration

L-3,4-dihydroxyphenylalanine increases the quantal size of exocytotic dopamine release in vitro

Integrating single-cell transcriptomic data across different conditions, technologies, and species

Comprehensive Integration of Single-Cell Data

Regularization Paths for Generalized Linear Models via Coordinate Descent

Regularization Paths for Cox's Proportional Hazards Model via Coordinate Descent

Tumor Evolution of Glioma-Intrinsic Gene Expression Subtypes Associates with Immunological Changes in the Microenvironment

Complex heatmaps reveal patterns and correlations in multidimensional genomic data

RNA velocity of single cells

SCENIC: single-cell regulatory network inference and clustering

STAR: ultrafast universal RNA-seq aligner

Spatial reconstruction of single-cell gene expression data

The dynamics and regulators of cell fate decisions are revealed by pseudotemporal ordering of single cells

Reversed graph embedding resolves complex single-cell trajectories

We thank Drs. Barbara Corneo, Peter Sims, Tingting Wu, and Dantong (Danielle) Huang for scientific discussion and suggestions. We thank Erin Bush for help on single cell sequencing. 

The authors declare no conflict of interests. (H) Recorded cells were labeled with biotin dye (red) added to the patch pipette. After fixation, organoids were immunostained for TH (green). All recorded neurons (n=21) were confirmed to be TH+ (scale bar, 100m). (I) Representative whole-cell recording from spontaneously active DA neuron. (J) Evoked action potentials following step current injections of +10pA (black), +30pA (red), and +50pA (blue). (K) Average resting membrane potential of DA neurons in organoids. release in response to 4-hour opioids treatment (n=3 organoids). Statistical analysis was determined by Oneway AVOVA as compared to untreated control (*p<0.05, **p<0.01, ***p<0.001, and ****p<0.0001). (E and F) Calcium imaging using Fluo4-AM of Day 180 organoids showing decreased frequency of calcium transients after fentanyl treatment (Movie S1). Statistical analysis of frequency and amplitudes was determined by t-test (*p<0.05, **p<0.01, and ****p<0.0001). Frequencies of all cells were significantly decreased after the FTY treatment (***p<0.001). (D) Single sample gene set enrichment analysis (ssGSEA) showed pathways relating to neuron projection (yellow), synapse assembly and signal transduction (blue), and neurotransmitter dynamics (purple) restore basal states in fentanyl withdrawal samples. The pattern was more similar to untreated samples (D77_UT, right heatmap of Fig. 4I)