Parasympathetic neurons derived from human pluripotent stem cells model human diseases and development

Summary

Autonomic parasympathetic neurons (parasymNs) control unconscious body responses, including “rest-and-digest”. ParasymN innervation is important for organ development; and parasymN dysfunction is a hallmark of autonomic neuropathy. However, parasymN function and dysfunction in humans is vastly understudied, due to the lack of a model system. Human pluripotent stem cell-derived neurons can fill this void as a versatile platform. Here, we developed a differentiation paradigm detailing the derivation of functional human parasymNs from Schwann cell progenitors. We employ these neurons (i) to assess human autonomic nervous system development, (ii) to model neuropathy in the genetic disorder Familial Dysautonomia, and (iii) to show parasymN dysfunction during SARS-CoV-2 infection, (iv) to model the autoimmune disease Sjögren’s syndrome and, (v) to show that parasymNs innervate white adipocytes (WATs) during development, and promote WAT maturation. Our model system could become instrumental for future disease modeling and drug discovery studies, as well as for human developmental studies.

eTOC blurb

Wu et al. differentiate hPSC-derived Schwann cell precursors into functional parasympathetic neurons. These parasympathetic neurons are employed to study specific defects in the genetic autonomic disorder Familial Dysautonomia, during COVID-19 infection, and in the autoimmune disease Sjögren’s syndrome. The role of parasympathetic innervation of adipose tissue development is assessed.

Graphical Abstract

Introduction

The autonomic nervous system (ANS) regulates body homeostasis and involuntary responses, such as heart rate and blood pressure. The two major components of the ANS are the sympathetic nervous system (SNS) and the parasympathetic nervous system (PSNS). While postganglionic sympathetic neurons (symNs) release norepinephrine to stimulate the fight-or-flight response, increase heart beat and vasocontraction, postganglionic parasympathetic neurons (parasymNs) release acetylcholine (ACh) to trigger the opposite response called “rest-and-digest”¹. Autonomic dysfunction has been found in various diseases, such as Parkinson’s disease, Alzheimer’s disease, hypertension, multiple system atrophy, diabetic autonomic neuropathy, spinal cord injury, COVID-19, and familial dysautonomia (FD)^1–4. The “imbalance” of the ANS, resulting from either the increased sympathetic or decreased parasympathetic tone, or both, contribute to autonomic neuropathy in these diseases^3,5.

In the clinic, ANS function is assessed mostly by indirect methods, for example, skin conductance measurements indicating in/decreased sympathetic/parasympathetic activity^6,7. Additionally, ANS specific responses are difficult to observe without the interference from the central nervous system (CNS) and it is nearly impossible to extract primary human parasympathetic tissue for research. Human pluripotent stem cell (hPSC) technology provides an alternative platform to study the function of human ANS neurons provided that a defined neuron differentiation strategy is available^8,9. To our knowledge only two parasymN protocols have been described^10,11. Current research suggests that parasymNs develop from Schwann cell precursors (SCPs), which are derived from neural crest (NC) cells^12,13. This feature makes parasymNs developmentally different from symNs, which are differentiated directly from earlier NC cells¹⁴. Takayama et al., derived both symNs and parasymNs directly from autonomic progenitor cells¹⁰, which does not reproduce this developmental process. Furthermore, some, normally adrenergic, symNs can switch to becoming cholinergic upon innervation of a tissue that is not innervated by parasymNs¹⁵. Thus, Takayama’s parasymN-like cells may be cholinergic symNs instead. Goldsteen et al. differentiated cholinergic neurons via a vagal, enteric-destined NC progenitor and innervated lung tissue in co-cultures¹¹, however, parasymNs in vivo are derived from the SCP.

Here, we derived parasymNs from hPSCs via SCPs, reproducing the proper developmental paradigm in vitro. Our protocol generates parasymNs that exhibit typical molecular and physiological properties and can be derived from various hPSC lines. Together with the symN differentiation protocol we established earlier^16,17, these in vitro directed differentiation systems allowed us to identify key developmental regulators for human development of ANS paradigms. We further employed these parasymNs to model the genetic disorder Familial Dysautonomia (FD), where we showed that FD parasymNs are spontaneously hyperactive, and that the normal crosstalk with symNs is disrupted. We showed that parasymNs are not infected directly by SARS-CoV2, but are sensitive to infection-related metabolic changes, and parasymN conditional media displayed cardioprotective effects during COVID-19 infection. We further modeled the autoimmune disease Sjögren’s syndrome (SS), and identified parasymN hypoactivation caused by a patient-mediated IgG autoimmune response. When parasymNs were co-cultured with mouse salivary gland cells, dry mouth symptoms, from which patients are suffering, were mimicked. Lastly, we showed that in a parasymN and adipocyte co-culture system human parasymNs can functionally innervate and control adipocyte metabolism.

Results

Schwann cell progenitor-derived parasympathetic neurons

NC cells are SOX10⁺ early embryonic progenitor cells that are highly migratory and give rise to a large variety of cell types, including all peripheral neurons, melanocytes and peripheral glia¹⁸. We have previously established an efficient NC cell differentiation protocol^19,20 that yields sensory²¹, enteric²⁰ and symNs^16,17,22,23 (Fig. 1A). Recently, Majd et al. employed these NC cells to derive functional Schwann cells (SCs)²⁴ (Fig. 1A). Therefore, our goal was to redirect SCPs into parasymNs (Fig. 1A). After minor adaptations of the NC cell phase (to increase efficiency/reproducibility^17,21), we first aimed at capturing the time point when multipotent SCPs are induced, but not yet committed towards the SC fate. SCPs were present in a wide window during the culture²⁵ (day 10–30, Suppl. Fig. 1A). SCP markers (PAX3/GAP43) were downregulated from day 16–20, accompanied by an increase in immature SC markers (CD56), which were downregulated after day 24, while SC markers (MPZ, S100β) continued increasing (Suppl. Fig. 1B). Based on these observations, we performed bulk RNA sequencing on day 16 (Suppl. Fig. 1C). Principal component analysis (PCA) confirmed that day 16 SCPs were distinct from day 10 NC cells (Suppl. Fig. 1D). Gene ontology analysis identified enriched pathways and genes associated with SC development and SCP identity²⁶ in day 16 SCPs (Suppl. Fig. 1E–F). These results suggest that the proper time to start parasymN induction is between day 16–24. To differentiate SCPs into parasymNs, we performed a timepoint and factor screen, replating cultures every four days between day 16 and 24, and culturing with GDNF, BDNF, NGF, FBS and ciliary neurotrophic factor (CNTF) combinations for one week (Suppl. Fig. 2A, B). CNTF is crucial for supporting parasymN development and survival²⁷ and acts as the key trophic factor for parasymNs in target tissue innervation and communication²⁸. One week after replating, only the cultures replated before day 16 exhibited neuron-like cell morphology, and cells grown in condition 3 (GDNF, BDNF, CNTF, 1% FBS) displayed well developed neurite bundles (Suppl. Fig. 2C). We evaluated expression of peripheral neuron marker PRPH and the parasymN neurotransmitter choline acetyltransferase (ChAT, Suppl. Fig. 2D). Indeed, day 16 replated SCPs in condition 3 medium differentiated most efficiently into ChAT⁺ neurons (Suppl. Fig. 2E–F). By day 30, these parasymNs expressed autonomic markers ASCL1, PHOX2B, PRPH, and CHRNA3 by RT-qPCR (Fig. 1B). Parasympathetic, cholinergic markers ChAT, VAChT, ChT, and NPY2R were expressed (Fig. 1B). Notably, the neurons expressed HMX3²⁹ at day 30, which can distinguish parasymNs from other peripheral cholinergic neurons (Fig. 1B). Contamination of SOX10⁺ SCs was very low in our parasymN cultures (Fig. 1C), suggesting high differentiation specificity. We also compared the ratio of ChAT (cholinergic) to TH (adrenergic) expression between day 30 hPSC-derived parasymNs and symNs^16,17 and showed that parasymNs predominantly expressed ChAT, while symNs expressed TH (Fig. 1C, D). ParasymNs expressed the more mature markers ChAT, VAChT, NPY2R, and PHOX2B::GFP³⁰ (Fig. 1E). Image quantification showed a differentiation efficiency of ~50% ChAT⁺ parasymNs with, however, very high differentiation specificity, as almost all PRPH⁺ neurons expressed ChAT (Fig. 1F). We next performed bulk RNA sequencing of day 30 parasymN cultures and compared the expression profiles to published datasets of hPSC-derived prefrontal cortical neurons³¹ (PFC), posterior spinal motor neurons³¹ (MN), sensory neurons³² (SN), and symNs¹⁶. PCA analysis indicated that our parasymNs are distinct to those neuron types. As peripheral nervous system neurons, parasymNs, symNs and SNs were more distinctly separated from CNS neurons, PFC and MNs. As expected, parasymNs were more like SNs than symNs, reflecting their shared cholinergic nature (Fig. 1G). We also investigated the positional identity of our parasymNs by assessing the expression profile of HOX genes. In the human ANS the parasymN chain is derived mainly from cranial and vagal nerves, which are HOX 1–5⁺, but HOX2⁻, which represents cranial nerve V that gives rise to trigeminal and sensory motors, but not to parasymNs³³ (Fig. 1H). RT-qPCR analysis showed that parasymNs expressed HOX 1–5, but not HOX 2 (Fig. 1I). The cranial nerve III-derived parasymN ganglion is HOX null³³, and to our knowledge, there is no specific marker to identify parasymNs with cranial origin. Considering the variety of genes detected in hPSC-derived parasymNs, we suggest that these parasymNs display high differentiation specificity, robust parasymN identity, and may cover several parasymN subtypes in the body.

Figure 1.

SCP-derived neurons show parasympathetic identity. (A) Schematic illustration of the differentiation workflow to make symNs, SCs, and parasymNs from NCCs. (B) RT-qPCR analysis of day 30 parasymNs for parasymN markers. n=3–5 biological replicates. (C) RT-qPCR comparisons of SOX10 expression between SCs and parasymNs and ChAT/TH ratio between symNs and parasymNs on day 30. Student’s two-tailed t-test. n=3–6 biological replicates. (D) Western blot analysis of day 30 symNs and parasymNs. n=2 biological replicates. (E) Immunofluorescence images of hPSC-derived parasymNs for parasymN markers on day 30. (F) Quantification based on the immunofluorescence images. n=4–6 biological replicates. (G) PCA plot of parasymNs, symNs, SNs, MNs, and PFCs using bulk RNAseq. (H) Cartoon illustration shows the alignment of HOX genes in the parasympathetic and sympathetic nervous system. (I) RT-qPCR analysis of day 30 parasymNs for HOX genes. n=4 biological replicates. (J) Single cell transcriptomics defined the cellular composition in parasymN differentiation on day 30. Error bars=SEM. **p<0.01, ****P<0.0001. FC=fold change. See also Figure S1–S4.

To gain further insight into the make-up of the parasymN cultures and to identify the contaminating, mis-differentiated cell types, we performed single-cell RNA sequencing (scRNA-seq) on day 30 hPSC-derived parasymNs. Using the PHATE analysis³⁴ to identify the different cell clusters (Fig. 1J, Suppl. Fig. 3A), we identified early progenitors on the right, progressing into the main two lineages, parasymNs and SCs, to the left. Other clusters, such as peripheral neurons (PNSN), cranial nerve progenitors³⁵ (CNP1 & 2), SCs (SC), SC-derived endoneurial fibroblasts^36,37 (EF), early NC^38–40 (eNC), endothelial cells (EC), osteoblasts (OB), early cardiomyocytes⁴¹ (eCM), and mesenchymal progenitors⁴² (MP), were also identified (Suppl. Fig. 3A–B). Of note, our autonomic neurons (symNs and parasymNs) form tight, ganglia-like clusters and die upon dissociation, whereas mis-differentiated cells and early progenitors tolerate the dissociation process much better. Hence, it is highly likely that the majority of mature parasymNs were lost during the scRNA-seq cell preparation step, and thus the primary focus of the scRNA-seq analysis was to identify the contaminating cell types. We concluded that these contaminating cell types in our day 30 parasymN cultures are unlikely to negatively affect the disease modeling approaches we show in this work.

We developed a method to enrich for parasymNs by treating developing neurons with a low dosage of cytosine arabinoside (AraC) for one week (Suppl. Fig. 4A). This significantly improved the neuron purity from ~45% to 80% (Suppl. Fig. 4B–C), aiding future studies.

We next examined the functionality of the hPSC-derived parasymNs. Using microelectrode array (MEA), we detected spontaneous action potentials in day 30 parasymNs (Fig. 2A), that could be further induced by stimulating the cholinergic receptors on the neurons by nicotine (Fig. 2B). We also detected the release of ACh, the main neurotransmitter of parasymN, in the culture medium by ELISA (Fig. 2C). ParasymNs mainly express cholinergic muscarinic receptors (MusR)¹. Thus, we assessed the expression of all MusR subtypes in parasymNs and found that the M2 and M4 MusRs to be expressed (Fig. 2D), consistent with findings in human and animal parasymN tissues^1,43,44. We then treated parasymNs with bethanechol (BeCh) and atropine, clinical MusR agonist and antagonist, respectively^45–47. As expected, BeCh activated, while atropine inactivated the parasymNs (Fig. 2E). Furthermore, we treated hPSC-derived parasymNs/symNs with 6-OHDA (100 μM), a neurotoxin that mainly targets catecholaminergic neurons (i.e. symNs) and can trigger apoptosis in 12 hours^10,48,49. As a result, early apoptotic signaling PUMA and cleaved Caspase-3 (c-Cas3) were significantly upregulated in symNs, but not in parasymNs (Fig. 2F). Accordingly, 12 hours after 6-OHDA treatment, symNs displayed early hyperactivity, a reported precursor of neurodegeneration⁵⁰. In contrast, parasymNs remained unaffected (Fig. 2G). 24 hours after 6-OHDA treatment, strong c-Cas3 signals, broken cell bodies and fragmented axon structures were observed in symN cultures as typical features of neurodegeneration⁵¹. Again, parasymNs remained intact (Fig. 2H). To test whether parasymNs can build functional connections to their target tissues, we established a co-culture system using hPSC-derived parasymNs and CMs. During development, SCPs migrate along the axons of preganglionic neurons to locations close to the target tissues and differentiate into parasymNs in ganglia¹. To replicate this developmental process in vitro, we replated day 16 SCPs on top of CMs⁵² (Fig. 2I). 10 days after co-culture, we identified TUJ1⁺/PRPH⁺/ChAT⁺ parasymNs targeting α-ACTININ⁺ CMs (Fig. 2I), which was evident by the formation of neurocardiac junctions observed as nodal structures⁵³ (Fig. 2I’ arrowheads). To confirm the functional connectivity, we activated the parasymNs in the co-cultures with nicotine and measured the beating rate of CMs. We found that the beating of CMs decreased after treatment with nicotine (Fig. 2J). To rule out the possibility of nicotine working directly on CMs, we confirmed this parasymN-mediated activation of CMs via optogenetics. We co-cultured CMs with parasymNs differentiated from an iPSC line⁵⁴ that expresses light-activated channelrhodopsin-2 (ChR2). Similar to nicotine, blue light activation of ChR2-parasymNs suppressed CM beating (Fig. 2J). This data suggests that our hPSC-parasymNs are functional.

Figure 2.

SCP-derived neurons display functional parasymN features. (A) Representative MEA heatmap of day 0 hPSCs and day 30 parasymNs. (B) ParaymN activity unstimulated or nicotine (1 μM) treated, measured by MEA. Student’s two-tailed t-test. n=4 biological replicates. (C) Acetylcholine (ACh) concentration in parasymN culture media measured by ELISA. n=3 biological replicates. (D) RT-qPCR of parasymNs for muscarinic receptors (MusRs). Student’s two-tailed t-test. Data of M2/4 were pooled and compared to M1/3/5. n=3–4 biological replicates. (E) ParaymN activity between control and BeCh (1 μM) or atropine (1 μM)-treated cells measured by MEA. Student’s two-tailed t-test. n=3 biological replicates. (F) Western blot analysis of 6-OHDA (100 μM) treated parasymN and symN 12 hr after treatment. n=3 biological replicates. (G) Changes of paraymN and symN electric activity over time after treatment with 6-OHDA. Values of parasymNs or symNs were normalized by their own vehicle treated groups, respectively (indicated by the red dotted line). Two-way ANOVA. n=3 biological replicates. (H) Immunofluorescence staining of 6-OHDA treated parasymN and symN 24 hr after treatment. n=4 biological replicates. (I) Top: Schematic illustration of hPSC-derived parasymNs and CMs co-culture. Bottom: Immunofluorescence images of the co-culture and the neural cardiac junction (NCJ, indicated by white arrows) highlighted in I’. (J) Cardiac action potential of the co-culture measured by MEA between control and nicotine (1 μM)-treated or blue light-treated cells. Student’s two-tailed t-test. n=6 for nicotine treatment and 4 for blue light biological replicates. Error bars=SEM. *p<0.05, **p<0.01, ***p<0.001, ****P<0.0001. FC=fold change. MFR=mean firing rate. See also Figure S5.

Lastly, we tested whether our protocol works with different hPSC lines. We differentiated parasymNs using another hESC line (MEL1) and four hiPSC lines (652⁵⁵, Cp1, hDFn, and C1²³). ParasymNs from all lines were positive for ChAT and PRPH (Suppl. Fig. 5A), with similar efficiencies (Suppl. Fig. 5B). We also showed that a commercial neural supplement BrainFast facilitated parasymN differentiation as more neurite bundles and higher level of CHAT expression were observed at an earlier stage (Suppl. Fig. 5C). Finally, we tested whether our differentiation strategy works on SCPs from other SC protocols. Kim, et al. published a hPSC-SC protocol (Suppl. Fig. 5D) in 2017, although their SCs did not display efficient functional myelination⁵⁶. We applied our hPSC-parasymN protocol to this SC protocol in the first week of SCP induction and successfully differentiated neurons that were positive for parasymN markers (Suppl. Fig. 5D, E). Taken together, our hPSC-parasymN protocol is highly specific, functional, flexible, and reproducible.

Studying human ANS development in vitro

Taking advantage of the directed differentiation protocols we established for parasymNs and symNs^16,17, we next sought to untangle the transcriptional network that governs ANS development. Compared to highly migratory SCPs, that are derived from late migratory NCCs⁵⁷, symNblasts are differentiated from earlier NC cells. However, at the symNblast stage they are not as migratory. In addition, postganglionic symN ganglia locate closer to the spinal cord, while postganglionic parasymN ganglia locate closer to the distal target tissues⁵⁸. Thus, we hypothesized that parasymN-destined SCPs and symN precursors (day 14 symNblast in our protocol^16,17) may differ in their migratory capacity, differentiation potency, and regulatory network for cell fate determination. First, we compared the migratory capabilities between day 16 SCPs and day 14 symNblasts by directly plating down the 3D spheroids. As expected, SCPs displayed higher migratory potential compared to symNblasts (Fig. 3A). We then examined whether day 14 symNblasts can be efficiently differentiated to parasymNs. The culture produced much fewer ChAT⁺ neurons than symNblast-derived symNs (Fig. 3B). Compared to SCP-parasymNs, symNblast-ChAT⁺ neurons barely expressed HMX3, suggesting that these neurons might not be parasympathetic (Fig. 3B). To gain a comprehensive understanding of parasymN and symN development, we compared day 14 symNblasts to day 16 SCPs using bulk RNA sequencing (Fig. 3C). The two neural precursors occupied distinct coordinates on the PCA plot (Fig. 3D). In accordance with the migration assay, genes that are involved in cell adhesion and neural development pathways were enriched in day 14 symNblasts over day 16 SCPs (Fig. 3E). PCA analysis with day 10 NCCs, day 16 SCPs and day 30 parasymNs delineated a developmental progression of parasymNs over time (Fig. 3F). We further compared day 30 parasymNs to day 16 SCPs. Among differentially expressed genes, we identified multiple parasymN markers^29,59 (including HMX2/3), that were upregulated in day 30 parasymNs, but not symN markers (Fig. 3G).

Figure 3.

Comparison of parasymN and symN development. (A) Cell migration of day 16 SCP and day 14 symNblast was compared by plating down spheroids and assessing neuron migration out of the spheroid after 24 hr. n=4 biological replicates. (B) Left: Representative immunofluorescent images of neurons after day 14 symNblasts were differentiated in the respective symN or parasymN medium. Top right: Quantification of neural numbers (TH⁺ for symNs, ChAT⁺ for parasymNs) from both media. Unpaired Student’s t-test. n=3 biological replicates. Bottom right: Comparison of HMX3 mRNA level between SCP- or symNblast-derived neurons in parsymN medium. Unpaired Student’s t-test. n=3 biological replicates. (C) Sample distance plot of day 16 SCP and day 14 symNblast by bulk RNAseq. (D) PCA plot of day 16 SCP and day 14 symNblast by bulk RNAseq. (E) Pathway enrichment analysis showed GO terms upregulated in day 14 symNblast compared to day 16 SCP. (F) PCA plot of day 10 NC, day 16 SCP, and day 30 parsymN. (G) Heatmap showed expression profile of parasymN and symN developmental signatures in day 16 SCP and day 30 parsymN. (H) Protein-protein interaction analysis of upregulated genes in parasymN and symN differentiations. (I) Representative immunofluorescent images and mRNA levels of CD274 in parasymNs and symNs. Unpaired Student’s t-test. n=3–4 biological replicates. *p<0.05, **p<0.01.

To further discover the regulatory networks that regulate parasymN and symN differentiation, we performed protein-protein interaction network analysis by analyzing genes that are upregulated in parasymNs over SCPs, or in symNs over symNblasts using the STRING database⁶⁰. Both interaction networks identified the epidermal growth factor (EGF) signaling with high centrality (Fig. 3H–I). Indeed, the importance of EGF in promoting neural development of NCCs⁶¹ has been reported. In the symN differentiation network, we identified multiple symN developmental and functional signatures such as GATA2/MEIS1/PHOX2A/PHOX2B/HAND2/ISL1/TH (Fig. 3H). In contrast, in the parasymN differentiation, parasymN signatures, such as PHOX2B/ASCL1/PRPH/HMX2/NOS1/CHRM2/ACHE, were identified and linked to pathways related to SC development, including GFAP/GAP43/MBP/EGR1/MPZL1 (Fig. 3H). Together with our results in Fig. 3B, these data suggest that human parasymN development runs parallel to SC development as they are both SCP derivatives, but distinct from symN development. Interestingly, we also identified CD274 in parasymN network analysis (Fig. 3H). CD274, or also known as PD-L1, is a transmembrane protein that is typically expressed in immune cells⁶², but also found in brain and peripheral neurons, including trigeminal neurons, symNs, and sensory neurons⁶³. We confirmed CD274 membrane expression both in hPSC-derived parasymNs and hPSC-derived symNs, as well as CD274 mRNA levels in both neurons at similar levels (Fig. 3I). This suggests that CD274 might be a useful surface marker to purify parasymNs via flow cytometry. Taken together, our directed differentiation protocols for autonomic neurons enabled us to gain better molecular insight into parasymN and symN cell fate specification.

Parasympathetic neuropathology in Familial Dysautonomia

We sought to model diseases with parasympathetic neuropathy, especially those not fully understood due to the limitation of human model systems. We first studied Familial Dysautonomia (FD), a genetic disease that mainly affects the ANS, causing abnormalities in controlling body temperature, gland secretion, and blood pressure⁶⁴. One of the most life-threatening symptoms in FD is the dysautonomic crisis. Induced by stress, FD patients display extreme cardiovascular dysregulation, hypertension, high blood pressure, vomiting attacks, and diffused sweating⁶⁴. Although both the SNS and PSNS are affected in FD patients, the level of involvement and the defects in parasymNs remain understudied^65,66. Clinical observations suggest both hypoactivity and hyperactivity of the FD PSNS^67–69. Recently, we showed that FD symNs are spontaneously hyperactive¹⁶, thus here we sought to better understand the FD parasymN pathology and its interaction with symNs. We differentiated parasymNs from healthy control subjects (hPSC-ctrl-H9 hESC and iPSC-ctrl-C1, hereafter data from these lines was combined) and FD (iPSC-FD-S2²³) patients. Day 30 FD parasymNs showed similar neural morphology (Fig. 4A, inset images) and gene expression of parasymN markers (Suppl. Fig. 6A–B) as control parasymNs. MEA analysis showed that FD parasymNs were hyperactive across time points, compared to control (Fig. 4A), but the level of parasymN hyperactivity was not as high as symN hyperactivity in FD (Fig. 4B). These results reflect parasymN phenotypes in FD patients, which are subtler than symN phenotypes^65–67. RT-qPCR analysis confirmed that ELP1 splicing is defective in FD parasymNs (Fig. 4C), as expected given that all FD patients carry this mutation, and it leads to most severe splicing defects in neural cells⁷⁰. To test if hyperactive FD parasymNs affect different target tissues, we performed two co-culture systems with hPSC-derived CMs¹⁶ and hPSC-derived smooth muscle cells^20,71 (SMCs, Fig. 4D, Suppl. Fig. 7A–B). Generally, ACh signaling leads to decreased beating of cardiac muscles, but increased contraction of smooth muscle (Fig. 4D), due to the expression of different ACh receptors⁷². Differentiated parasymNs were observed after co-culturing hPSC-derived SCPs with hPSC-derived SMCs (Suppl. Fig. 7C), with direct contacts to SMCs evident by synaptophysin⁺ (SYP), as well as the nodal (bouton en passant) structures⁷³ of neuromuscular junctions (Suppl. Fig. 7D, arrows). Indeed, hyperactive FD parasymNs lowered CM beating, while triggering an increase in SMC contraction compared to control parasymNs (Fig. 4E). Our data confirmed a parasymN specific phenotype in FD, which may contribute to altered body homeostasis in this disease.

Figure 4.

Parasympathetic hyperactivity and impaired autonomic crosstalk in FD. (A) Inset: Immunofluorescence images of control and FD parasymNs for PRPH. MEA analysis of control and FD parasymNs. Two-way ANOVA. n=3–5 biological replicates. (B) Fold changes of neural activity of FD symNs and parasymNs relative to control symNs and parasymNs. Student’s two-tailed t-test. n=4–11 biological replicates. (C) RT-qPCR analysis for ELP1 splicing in ctrl and FD parasymNs. n=4 biological replicates. (D) Cartoon illustration of the effects of ACh released by parasymNs to CMs and SMCs. (E) Muscular activities of hPSC-CMs and hPSC-SMCs co-cultured with control or FD parasymNs. Student’s two-tailed t-test. n=4 biological replicates. (F) RT-qPCR of control and FD parasymNs for signal transduction markers. Multiple unpaired Student’s t-test. n=3–5 biological replicates. (G) Sample distance plot of ctrl and FD parasymNs by bulk RNAseq. (H) PCA plot of day ctrl and FD parasymNs by bulk RNAseq. (I) Pathway enrichment analysis showed GO terms upregulated in FD parasymNs compared to ctrl. (J) Left: Schematic illustration of conditional medium treatments. Right: MEA analysis of FD symNs and parasymNs treated with conditional media from control and FD symNs and parasymNs. One-way ANOVA. n=5–9 biological replicates. (K) Cartoon illustration shows that in FD PNS, both PSNS and SNS are hyperactive, but SNS hyperactivity is stronger. It also shows that the crosstalk between PSNS and SNS is impaired. Error bars=SEM. *p<0.05, **p<0.01, ****P<0.0001. FC=fold change. MFR=mean firing rate. See also Figure S6–S7.

To understand the mechanisms of parasymN hyperactivity in FD better, we analyzed genes related to the regulation of neural activity. When comparing markers of signaling receptors, we found that FD parasymNs expressed higher nicotinic receptors (CHRNA3), which receive the preganglionic signals, but lower MusRs (CHRM2), which autoregulate ACh release (Fig. 4F). Accordingly, ChAT expression increased in FD parasymNs, which was in line with the observed hyperactivity (Fig. 4F). These findings indicate that FD parasymNs might be hypersensitive to the external stimuli, in addition to lacking the necessary autoregulatory machinery. We further compared healthy and FD parasymNs using bulk RNA sequencing (Fig. 4G). PCA analysis indicated that the transcriptional profile of FD parasymNs is distinct from healthy control (Fig. 4H). Compared to healthy parasymNs, FD parasymNs exhibited higher expression of genes associated with oxidative stress and neural activity related signaling pathways, which may explain parasymN degeneration found in patients (Fig. 4I).

We next investigated how symNs and parasymNs communicate in FD. ParasymN activity can be downregulated through expressing the neuropeptide Y (NPY) receptor (NPY2R), which is targeted by NPY released from symNs as a cofactor of NE^74–76. In cardiovascular diseases, such as hypertension, oversecreted NPY from symNs inhibits parasymN activity, therefore leading to an unbalanced ANS⁷⁶. To test this model, we first compared NPY2R levels in control and FD parasymNs; no significant difference was found (Suppl. Fig. 6C). We then performed reciprocal conditional medium treatments. Conditional media from control and FD symNs were given to FD parasymNs, and vice versa (Fig. 4J, left). Interestingly, control media from parasymNs or symNs had the ability to suppress the activity of their FD counterparts, an effect that was abolished when FD conditional media was added (Fig. 4J, right). Our findings revealed that in FD, parasymNs, and symNs are both hyperactive, though symN hyperactivity is stronger than parasymN hyperactivity, overall likely leading to a hyperactive state in the patient. The important crosstalk and mutual regulation between the two systems is impaired, which explains the general oversensitivity and overreaction of FD patients’ ANS (Fig. 4K).

hPSC-parasymNs for COVID-19 research

COVID-19 has caused significant damages to human lives in the past four years. Respiratory failure and cardiovascular pathologies are the major causes of death after SARS-CoV-2 infection⁷⁷. As the key regulator of respiratory and cardiovascular functions, SNS hyperactivation has been observed in patients with COVID-19^2,4. Moreover, symNs also express angiotensin-converting enzyme 2 (ACE2), the membrane receptor/enzyme, which is the gateway for SARS-CoV-2 invasion, suggesting that symNs could be a target of SARS-CoV-2⁷⁸. Recently, we have shown that our hPSC-symNs express ACE2, and that hPSC-symNs were hyperactivated upon infection with SARS-CoV-2 pseudovirus⁷⁹. However, the role of the PSNS in COVID-19 remains unclear.

ACE2 belongs to the renin-angiotensin-aldosterone system (RAAS), the key system for blood pressure and blood volume maintenance. For a healthy RAAS, the proper balance of its two arms is crucial. Angiotensin II (Ang II) induces vasocontraction, which is counteracted by angiotensin 1–7 that is hydrolyzed from Ang II by ACE2⁷⁷. Unbalanced RAAS, due to compromised ACE2 by the binding of SARS-CoV-2 is a major cause of multiple organ damage in patients⁸⁰.

It has been suggested that the PSNS may be a key player in COVID-19-related illness for the following reasons. First, given the fact that human parasymNs express Angll receptors, parasymNs could be sensitive to changes caused by an unbalanced RAAS system⁸¹ (Fig. 5A). Indeed, increased Ang II levels may decrease the activity of the PSNS, which worsens the unbalanced ANS, since symNs are hyperactivated^79,81. Second, the anti-inflammatory effect mediated by ACh is one of the essential regulators of inflammation in the PNS⁸². Decreased parasymN activity during SARS-CoV-2 infection, therefore, may not only cause unbalanced ANS, but also weaken the defense system in the body (Fig. 5A).

Figure 5.

hPSC-derived parasymNs in COVID-19 mimicking studies. (A) Cartoon illustration of how SARS-CoV-2 infection leads to an imbalanced RAAS, which leads to an imbalanced ANS that worsens the cardiovascular complications observed in COVID-19 patients. (B) RT-qPCR of parasymNs for ACE2 and AGTR1/2. n=3 biological replicates. (C) Representative immunofluorescence images of parasymNs for AGTR1/2. (D) Schematic illustration of SARS-CoV-2 infection. (E) Representative immunofluorescence images of neurons after 72 hrs infection with WA1 SARS-CoV-2. (F) MEA analysis comparing neural activity of control and AngII-treated parasymNs. Student’s two-tailed t-test. n=10 biological replicates. (G) Schematic illustration of the potential anti-inflammatory effect of parasymN conditional media in an anti-inflammation assay. (H) ROS level of hPSC-derived CMs with each treatment was shown as CM-H2DCFDA intensity measured by ELISA. One-way ANOVA. n=4 biological replicates. (I) Cardiac activity of hPSC-derived CMs with each treatment measured by MEA. One-way ANOVA. n=3 biological replicates. (J) Beating variability of hPSC-derived CMs with each treatment measured by MEA. One-way ANOVA. n=3 biological replicates. Error bars=SEM. *p<0.05, **p<0.01, ***P<0.001. FC=fold change. MFR=mean firing rate.

Herein, we aimed to evaluate whether our hPSC-parasymNs can be used to study the COVID-19 related parasymN responses. We first analyzed and confirmed the expressions of Ang II receptors, AGTR1/2, in hPSC-parasymNs (Fig. 5B–C). The levels of AGTR1/2 were significantly higher than ACE2 (Fig. 5B), which may imply that SARS-CoV-2 infection would affect parasymN function not by direct infection, but through metabolic changes in the circulation. To evaluate this theory in our system, we infected hPSC-parasymNs with WA1/2020 strain SARS-CoV-2 (Fig. 5D). Notably, it has been demonstrated that cholinergic SNs can be infected by SARS-CoV-2^83–85. We, therefore, used hPSC-SNs^21,86 as a positive control. Remarkably, SARS-CoV-2 preferentially infected SNs, but not parasymNs (Fig. 5E), supporting our hypothesis that the PSNS is unlikely to be affected by SARS-CoV-2 through direct infection. Treating hPSC-parasymNs with Ang II decreased parasymN activity (Fig. 5F), suggesting that hPSC-parasymNs are functionally responsive to COVID-19-related homeostatic changes. We next sought to assess the anti-inflammatory function of hPSC-parasymNs in vitro. Recently, Yang, et al. showed that TNF-α and IL-6 released by immune cells are one of the main mechanisms of cardiovascular complications in COVID-19 using a hPSC-CM model⁸⁷. We, therefore, performed an anti-inflammation assay, in which hPSC-CMs were challenged by TNF-α and IL-6, with or without conditional media from healthy hPSC-parasymNs (Fig. 5G). 24 hours after treatment, the ROS level of CMs was analyzed using the CM-H2DCFDA oxidative stress indicator⁸⁸. Consistent with previous research, TNF-α/IL-6 significantly increased ROS levels in CMs (Fig. 5H). MEA analysis showed that TNF-α/IL-6 decreased the beating rate yet increased the beating variability of CMs (Fig. 5I–J), suggesting that CM function is impaired in the inflammatory state. With the addition of parasymN media, ROS level, CM beating, and beating variability were significantly restored to the levels found in healthy CMs (Fig. 5H–J). These results demonstrate parasymN involvement in COVID-19 pathology and may be a useful tool to further research possible treatment options in the future.

hPSC-parasymNs model Sjögren’s syndrome

SS is a female-predominant autoimmune disease affecting ~3 million people in the US, that is often a precursor to more deleterious autoimmune disorders⁸⁹. SS patients suffer from difficulties in producing tears and saliva, which are controlled by the ANS⁹⁰. There is no FDA-approved therapy available for this disease. Clinically, primary SS (pSS) is distinct from secondary SS (sSS), where patients experience autoimmune symptoms and rheumatic diseases⁹⁰. pSS patients usually exhibit high levels of anti-SSA/Ro and/or anti-SSB/La antibodies⁹¹. Years of case studies have suggested the involvement of parasympathetic neuropathy in SS, especially pSS. For instance, autoantibodies targeting MusRs are found in pSS patients, impairing ACh signaling⁹²; and prolonged fatigue, possibly caused by PSNS hypoactivation was reported⁹³. Vagal nerve stimulation has been shown to improve the symptoms in pSS patients⁹³. ParasymNs express MusR^1,43,44 (Fig. 2D), which may make them a target of autoantibodies in pSS. Moreover, autoantibodies that target the ganglionic AChRs in SS patients have also been reported⁹⁴. Thus, we asked whether intrinsic parasymN hypoactivity in addition to the blockage of downstream signaling due to the autoimmune response is underlying clinical symptoms in SS patients. To test this theory, we performed an antibody-based, complement-dependent cytotoxicity assay to model SS in vitro. Total IgG was purified from sera from two SS patients, and added to healthy hPSC-derived parasymNs together with complement-active healthy human serum (Fig. 6A). Healthy total human IgG was used as the control. After 72 hours, parasymNs were targeted by autoantibodies from both SS patients (Fig. 6B–C), suggesting that parasymNs might be a direct target of the autoimmune response. ParasymN activity decreased upon treatments with SS IgGs compared to control, suggesting that direct parasympathetic dysfunction might partially contribute to the symptoms of SS (Fig. 6D). Accordingly, we expected to find decreased extracellular ACh levels in parasymN media. However, upon SS IgG treatment of parasymNs, we found an unexpected increase in extracellular ACh levels compared to control-IgG treated neurons (Fig. 6E). We analyzed expression of ACh synthesis, ACh signaling, and neural plasticity-related genes, but none of them was changed significantly after SS IgG treatment (Suppl. Fig. 8A). Another pathway to increase ACh in the synapse is to suppress ACh degradation via acetylcholinesterase (AChE). We found that the activity of AChE was significantly decreased in SS-IgG treated parasymN media (Fig. 6F). In addition, ROS levels were elevated in SS patient IgG-treated parasymNs, indicating that SS-IgG treated parasymNs underwent a cell stress response (Fig. 6G).

Figure 6.

hPSC-parasymN model of Sjögren’s syndrome (SS). (A) Schematic illustration of the antibody-based complement-dependent cytotoxicity assay. (B) Immunofluorescence images using anti-human IgG secondary antibody showed SS IgG antibodies attached to parasymN clusters. Comparisons of parasymNs treated with or ctrl or SS IgG are: (C) Quantification of positive cell numbers in neural clusters. Three clusters in each biological replicate were analyzed from 3–4 biological replicates. One-way ANOVA. (D) MEA analysis of parasymNs. One-way ANOVA. n=4 biological replicates. (E) ACh release in parasymN cultures. One-way ANOVA. n=3 biological replicates. (F) AChE activity from parasymNs. One-way ANOVA. n=4 biological replicates. (G) ROS level of parasymNs was compared using CM-H2DCFDA. One-way ANOVA. n=3 biological replicates. (H) Percentage of changes in Ca²⁺ dynamics in salivary cells upon nicotine stimulation and SS patient IgG treatments in the co-culture compared to control IgG. n=4 biological replicates. (I) Cartoon illustration summarizing that in SS patients, parasymNs can be targeted by auto-Abs, which decreases parasymN activity. ACh extracellular level is increased, and AChE is reduced, possibly due to the autoimmune response. Error bars=SEM. *p<0.05, **p<0.01, ***p<0.001, ****P<0.0001. FC=fold change. MFR=mean firing rate. See also Figure S8–S9.

Furthermore, we sought to model the dry mouth symptom in SS using a hPSC-derived parasymN and mouse salivary gland cell co-cultures. Salivary tissue was obtained from Sox10-Cre/RFP mice⁹⁵. In culture, almost all cells were SOX10⁺ marking the salivary epithelium marker⁹⁵, and among these cells, over 80% cells were MIST1⁺ saliva producing acinar cells⁹⁶ (Suppl. Fig. 9A). After 10 days of co-culture, parasymNs were identified on the basal layer formed by the salivary cells, and the neural cell bodies were aggregated, mimicking typical ganglia structures (Suppl. Fig. 9A). Secretory granules are crucial for maturation and function of salivary acinar cells⁹⁷. In our culture system, granules were auto fluorescent and could be observed in acinar cells in the RFP channel (Suppl. Fig. 9A–B). We found that the number of acinar cells that produce granules was significantly increased when they are co-cultured with parasymNs (Suppl. Fig. 9B–C), implying that parasymNs may promote salivary gland maturation. We next assessed fluid secretion capacity of acinar cells under the regulation of parasymNs, using calcium imaging as an indicator⁹⁸. Ca²⁺ flux was excitable in the co-culture, when parasymNs were activated by nicotine, while salivary cells alone remained unaffected (Suppl. Fig. 9D), in line with previous understanding that nicotine affects salivary secretion mainly through the activation of ANS⁹⁹. On the other hand, parasymN activation does not affect the level of secretory proteins¹⁰⁰, which was also observed in the co-culture (Suppl. Fig. 9E). Finally, treating the parasymN-salivary gland co-cultures with SS patient IgG suppressed nicotine-induced Ca²⁺ flux (Fig. 6H), which may capture the feature of dry mouth symptoms in SS patients. Together, these findings suggest that parasymN function is directly affected in SS patients, and that the increased ACh level is a result of inhibited AChE in response to the autoimmune response (Fig. 6I).

hPSC-derived parasymNs innervate white adipocytes in vitro

In addition to disease modeling, we sought to investigate current biological conundrums using hPSC-derived parasymNs. It has been recognized that the SNS plays critical roles in lipid metabolism. SymNs innervate white, brown, and beige adipocytes¹⁰¹. Activation of symNs stimulates lipolysis and increases glucose uptake in WAT¹⁰². A functional co-culture model of symNs and white adipocytes has been established using primary symNs and 3T3-L1 mouse preadipocyte-derived white adipocytes, a widely used in vitro model system to study WAT¹⁰³. The role of the PSNS in WAT (assessed mostly in animal models), in contrast, is not fully understood and somewhat controversial^104–107.

Thus, we tested whether hPSC-parasymNs can innervate WAT and/or regulate WAT metabolism. Day 16 SCPs were co-cultured with 3T3-L1-derived white adipocytes on day 10 (Fig. 7A and Suppl. Fig. 10A). Differentiated adipocytes alone showed accumulating lipid droplets (Suppl. Fig. 10B–C), expression of typical adipogenic markers (Suppl. Fig. 10D) and displayed lipolytic activity when treated with β-adrenergic receptor agonist isoproterenol (Suppl. Fig. 10E). 10 days after co-culture, ChAT⁺ parasymNs were observed on FABP4⁺ adipocytes, and were positive for SYP staining in the nodal structures of axons (Fig. 7B and F). We also detected the expression of ACh secretory vesical marker VAChT, a synaptic marker in the co-culture, but not in WAT alone (Fig. 7C). These results suggest that hPSC-parasymNs innervate mouse WAT in vitro. We then examined whether the parasymN innervation affects adipocyte maturation. We found that expression of adipocyte maturation marker FABP4 was significantly increased in adipocytes co-cultured with parasymNs (Fig. 7D). Next, we compared the level of adipogenesis using the lipid droplet dye LipidSpot and found that adipocytes co-cultured with parasymNs displayed higher adipogenesis activity (Fig. 7E). Moreover, it has been shown that the size of nuclei decreases in mature adipocytes¹⁰⁸. Indeed, smaller nuclei were found in co-cultured adipocytes compared to adipocytes cultured alone (Fig. 7F). These data show that parasymNs promote white adipocyte maturation in vitro. To test whether adipose functions can be regulated by parasymN activity, we compared their glucose uptake and lipolytic activity. Co-culturing adipocytes with parasymNs increased medium glucose levels, which was in accordance with reduced intracellular glucose assessed by a fluorescent glucose probe, suggesting decreased glucose uptake in adipocytes (Fig. 7G). For lipolytic activity, we found that either in adipocytes only or in the parasymN co-cultured conditions, lipolysis levels remained unchanged, even when parasymNs were stimulated by nicotine (Fig. 7H). However, parasymNs counteracted the effect on increasing lipolytic levels induced by isoproterenol (mimicking adrenergic signaling, Fig. 7I), and lipolysis markers HSL and MGL were downregulated after nicotine treatment (Fig. 7J), suggesting that parasymNs may negatively regulate WAT lipolysis by counteracting sympathetic signaling. Taken together, our results demonstrate that human parasymNs target WAT, and regulate WAT maturation and functionality in vitro.

Figure 7.

hPSC-derived parasymNs target white adipocytes in vitro. (A) Schematic illustration of human parasymN and mouse WAT co-culture. (B) Immunofluorescence images of the co-culture. Comparisons of WAT with or without parasymNs are: (C) RT-qPCR analysis of WAT using human VACHT primer. Data was shown in Ct value. The Ct for of the detection limit is 40. n=5 biological replicates. (D) Left: Immunofluorescence images compare FABP4 expression in the adipocytes. Images were taken at the same exposure. Right: Image quantification of FABP4 staining. Student’s two-tailed t-test. n=6 biological replicates. (E) Left: Fluorescence images compare adipogenicity in the adipocytes using LipidSpot staining. Right: Image quantification of LipidSpot intensity. Student’s two-tailed t-test. n=4 biological replicates. (F) Left: Immunofluorescence images for Tuj1, FABP4, and DAPI in adipocyte cultures. Right: Quantification of nucleus size. 10 nuclei in each biological replicate were analyzed from 3 biological replicates. Student’s two-tailed t-test. (G) Glucose concentration in the culture media, as well as glucose uptake (intracellular) level in WAT. Student’s two-tailed t-test. n=4. (H) Glycerol release from adipocytes representing the lipolytic activity measured by ELISA. 1 μM nicotine was applied to activate parasymNs. Multiple unpaired Student’s t-test. n=4–6 biological replicates. (I) Glycerol release upon treatment with 10 μM isoproterenol. One-way ANOVA. n=4 biological replicates. (J) RT-qPCR for lipolytic markers using mouse primers in co-culture. Multiple unpaired Student’s t-test. n=4 biological replicates. Error bars=SEM. *p<0.05, **p<0.01, ****P<0.0001. FC=fold change. See also Figure S10.

Discussion

In this study, we developed the first SCP-originated postganglionic parasymN differentiation strategy from hPSCs. ParasymN identity and functionality is displayed and the neurons can be co-cultured with various cell types to show functional regulation of corresponding tissues. Our parasymN differentiation medium contains no NGF. Given that NGF is essential for symN survival¹⁰⁹, our method may exclude potential symN contamination from not fully differentiated NCCs. Moreover, during embryonic development, early differentiating parasymNs and symNs express both adrenergic/cholinergic genes until the postmitotic stage¹⁴. The ChAT/TH ratio (Fig. 1C) therefore not just shows low symN contamination but may also suggest that our hPSC-parasymNs are already at an early stage of maturation.

Using transcriptomic analysis, we showed that our parasymNs have distinct transcriptional signatures from similar neuron types and express specific parasymN markers such as HMX2/3²⁹ (Fig. 1B, 1G, and 3G), suggesting the specificity of our differentiation strategy. Furthermore, our hPSC-based parasymN and symN protocols^16,17 enabled us to perform a side-by-side comparison in order to gain more insight into the development and the regulatory network of ANS development (Fig. 3). Recently, it has been reported in zebrafish that SCPs can differentiate into symNs¹¹⁰. Preliminary data from our lab showed a similar result, when we subjected SCP to symN medium (data not shown). However, we find that parasymNs can only be generated through SCPs, not symN progenitors (Fig. 3B). Considering the low efficiency and the lack of critical parasymN specific markers, it is likely that the ChAT⁺ neurons generated from symNblasts in Fig. 3B are cholinergic symNs, SNs, or enteric neurons.

It has been shown that in FD the PSNS, in addition to the SNS, is defective. However, onset and extent of parasympathetic phenotypes in FD may be later and more subtle than the sympathetic phenotypes^65–67. Using indirect cardiovascular measurements, such as the cold face test, heart rate, blood pressure, and respiratory time, clinical studies have suggested that lower parasympathetic drive may be underlying FD pathologies^68,111. Interestingly, bradycardia has been suggested as an explanation of sudden death among FD patients^112–115, which could be a consequence of a hyperactive parasympathetic tone. Adding to the controversy, FD patients also display pupillary hypersensitivity to parasympathomimetic treatments¹¹⁶, and parasympathetic hypersensitivity induced lung stretching is also proposed as a reason for the breath-holding spells in FD⁶⁹. Here, we compared the behavior between healthy and FD hPSC-derived parasymNs and found that FD parasymNs are hyperactive, although this phenotype is more subtle compared to the FD symN phenotype in vitro (Fig. 4A–B). We also revealed the impaired crosstalk within the FD ANS (Fig. 4J–K), which may explain the instability and unpredictability of FD symptoms. Together, our system provides a platform of human parasymNs to study parasympathetic neuropathology in FD.

With regards to SARS-CoV-2 infection, the interaction between the RAAS and the SNS has been extensively studied^78,117,118. The understanding of the regulatory role of RAAS on the PSNS, in contrast, needs further work, despite studies showing that RAAS may interfere with parasympathetic drive and baroreflex through Ang II⁸¹. Here, we found that our hPSC-derived parasymNs exhibit functional reactivity to Ang II level changes but are not permissive to direct SARS-CoV-2 infection (Fig. 5B–F). We also showed the anti-inflammatory activity of hPSC-derived parasymNs in a model that simulates heart damage in COVID-19 (Fig. 5G–J). Our in vitro model therefore might be applied to study parasymN responsiveness to treatments and manipulations of other RAAS components, such as Ang 1–7 and MasR. The results could then be used for mechanistic studies and compound screenings to activate the PSNS in COVID-19 patients as a therapeutic strategy in the future.

Autoantibodies against type 3 MusR have been proposed as a reason for ANS dysfunction in SS⁹². Here, we found that such antibodies also recognize and hypo-activate presynaptic parasymNs (Fig. 6). We also identified excessive extracellular ACh, caused by reduced AChE activity, when cells were treated with SS patient IgGs, possibly suggesting an intrinsic parasympathetic response to inflammation¹¹⁹ (Fig. 6E–F). It is also possible that the reduced parasymN activity in response to SS patient IgGs is a result of cellular toxicity from excessive ACh taken up via the MusRs (Fig. 6H).

Dual innervation of the ANS is commonly found in most tissues in the body, but to date it remains unclear whether adipose tissues have parasympathetic innervation in addition to sympathetic innervation. In 2002, Kreier et al. reported that the WAT is innervated by parasymNs, detected using the retrograde pseudorabies virus (PRV) tracing technique in the rat model^104–106. Later in 2006, Giordano et al., reported the opposite result using PRV and immunohistochemistry in hamsters¹⁰⁷. Such differences among species prompted us to test WAT innervation mediated by our human parasymNs. Our results indicated that hPSC-derived parasymNs can innervate and regulate the maturity as well as functionality of 3T3-L1-derived white adipocytes (Fig. 7), although there are some limits to the interpretation of these results due to their in vitro and cross-species nature.

In summary, we describe a parasymN differentiation platform and demonstrate the versatility and utility of this model by employing it to model various human ailments.

Limitations of the Study

The process of scRNA-seq led to significant cell loss of parasymNs. Perhaps single nuclei RNA-seq might solve this problem in the future, although at a significant loss of sensitivity. The transcriptional analysis could lead us to identify novel genetic regulatory mechanisms to switch SCP identity into parasympathetic neurons at the expense of SCs, which may provide novel findings.

STAR methods

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Dr. Nadja Zeltner (nadja.zeltner@uga.edu)

Materials availability

Research materials generated in this article will be distributed upon reasonable request via MTA.

Data and code availability

All data generated or analyzed in this study are included in this article and its supplementary data file. Raw data points are available from the corresponding author on a reasonable request. Bulk RNA sequencing data is accessible through NCBI Gene Expression Omnibus, accession number GSE253235. scRNA sequencing data is accessible as samples GSM8136369 and GSM8136370 under GSE253235. This paper does not report original code. Raw data were deposited on Mendeley at [10.17632/v462ccvfg8.1].

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Cell lines

Stem cell lines used here are listed in Zeltner, 2016²³ and Wu, 2022⁷⁹. hESC lines: WA09 and H9 Phox2B::GFP³⁰ (female, NIH 0062), MEL1 (male, NIH 0139). hiPSC lines: healthy 652, 11 y.o. female, healthy Cp1, 21 y.o. female, hDFn (neonatal foreskin fibroblast -derived iPSCs, Gibco #C0045C), healthy C1, 35 y.o. female²³, Famililal Dysautonomia FD-S2²³. Other cell lines: 3T3-L1 (ATCC, #CL173). Detailed stem cell maintenance was previously described¹⁷. Cells were maintained in Essential 8 medium (Gibco, A15170–01) on vitronectin coated (Thermo Fisher/Life Technologies, A14700, 5 μg/ml) cell culture plates, and passaged using EDTA (Sigma, ED2SS). This study employed human embryonic stem cell lines (WA09), the use of which was approved to the Zeltner lab by WiCell. All iPSCs employed in this work were reprogrammed from human samples obtained through the public repository Coriell Research Institute.

METHOD DETAILS

In vitro differentiations

NC differentiation.

The detailed protocol is described in Wu, 2020¹⁷. On day 0, hPSCs were replated on Geltrex (Invitrogen, A1413202)-coated plates at 125×10³ cells/cm². Day 0–1 medium contains Essential 6 medium (Gibco, A15165–01), 0.4 ng/ml BMP4 (PeproTech, 314-BP), 10 μM SB431542 (R&D Systems, 1614) and 300 nM CHIR99021 (R&D Systems, 4423). From day 2 on, cells were fed with medium containing Essential 6 medium, 10 μM SB431542 and 0.75 μM CHIR99021. Note that BMP4 activity varies from batch to batch. It is highly recommended to perform BMP4 titrations for each batch/lot to get the best differentiation efficiency.

ParasymN differentiation.

To differentiated parasymNs from SCPs²⁵, day 10 NCCs were dissociated using accutase (Coring, AT104500) and aggregated to make spheroids on ultra-low attachment plates (Corning, 07 200 601 and 07 200 602) in SCP medium containing Neurobasal medium (Gibco, 21103–049), B27 (Gibco, 17502–048), L-Glutamine (Thermo Fisher/Gibco, 25030–081), 3 μM CHIR99021, 10 ng/ml FGF2 (R&D Systems, 233-FB/CF), and 10ng/ml NRG1 (PeproTech, 100–03). On day 16, SCP spheroids were dissociated using accutase and replated on PO (Sigma, P3655)/LM (R&D Systems, 3400–010-01)/FN (VWR/Corning, 47743–654)-coated plates at 100×10³ cells/cm² in parasymN differentiation medium that contains Neurobasal medium, B27, L-Glutamine, 1% FBS (Atlanta Biologicals, S11150), 25 ng/ml GDNF (PeproTech, 450), 25 ng/ml BDNF (R&D Systems, 248-BD), 25 ng/ml CNTF (R&D Systems, 257-NT), 200 μM ascorbic acid (Sigma, A8960), 0.2 mM dbcAMP (Sigma, D0627) and 0.125 μM retinoic acid (Sigma, R2625, add freshly every feeding). ParasymNs are differentiated in two weeks and can be maintained in parasymN differentiation medium by halfway feeding weekly. To increase the purity of neurons, 1 μM AraC (Sigma, C1768) can be optionally added on day 23 until day 30.

SymN differentiation.

A video protocol can be found in our previous publications^16,17. Day 10 NCCs were dissociated using accutase and aggregated to generate spheroids on ultra-low attachment plates in medium containing Neurobasal medium, B27, L-Glutamine, 3 μM CHIR99021 and 10 ng/ml FGF2. On day 14, spheroids were dissociated using accutase and replated on PO/LM/FN coated plates at 100×10³ cells/cm² in symN differentiation medium that contains Neurobasal medium, B27, L-Glutamine, 25 ng/ml GDNF, 25 ng/ml BDNF, 25 ng/ml NGF, 200 μM ascorbic acid, 0.2 mM dbcAMP and 0.125 μM retinoic acid (add RA freshly every feeding).

SC differentiation.

The protocol is adopted from Majd, 2023²⁴ with slight modifications. Day 10 NCCs were replated as spheroids in SCP medium until day 24. Spheroids were plated on PO/LM/FN coated plates without dissociation in SC differentiation medium that contains Neurobasal medium, B27, L-Glutamine, 10 ng/ml FGF2, 100 μM dbcAMP, and 20ng/ml NRG1. SC identity was evaluated on day 30.

CM differentiation.

The protocol was adopted from a previous study⁵². 80–90 % confluent H9 ESCs were replated at a 1 to 5 ratio on Matrigel (Corning, 1:20)-coated plates. When cells reached 90 % confluency, medium was changed to CDBM base medium, that contains DMEM/F12 (Gibco, 11320033), 64 mg/L ascorbic acid, 13.6 μg/L sodium selenium (Sigma, S5261), 10μg/ml transferrin (Sigma, T3309) and Chemically Defined Lipid Concentrate (Gibco, 11905031). 5 μM CHIR99021 was added to CDBM medium on day 0. 0.6 U/ml heparin (STEMCELL Technologies, 07980) was added to CDBM medium on day 1, 5, 6. 0.6 U/ml heparin and 3 μM XAV were added to CDBM medium on day 2, 3, 4. After day 7, medium was changed to CM maintenance medium that contains CDBM base, 20 μg/ml insulin (Sigma, I-034), and 2% FBS. CMs were replated on day 10 using accutase at 200×10³ cells/cm² on Matrigel-coated plates. For parasymN co-culture, day 16 SCPs were replated on day 15 CMs in medium that was mixed with parasymN differentiation medium and CM maintenance medium at 1:1 ratio.

SMC differentiation.

The protocol is adopted from previous studies^20,71. H9 ESCs were replated at 45×10³ cells/cm² on Matrigel-coated plates in Essential 8 medium. On day 1–4, cells were fed with medium that contains Neurobasal medium, N2, B27, DMEM/F12, 8 μM CHIR, and 25 ng/ml BMP4. On day 4–6, cells were fed with medium that contains Neurobasal medium, N2, B27, DMEM/F12, 10 ng/ml PDGF-BB (Shenandoah Biotechnology, 100–18), and 2 ng/ml Activin (Peprotech, 120–14). Cells were replated on day 6 at 35×10³ cells/cm² on Matrigel-coated plates in medium that contains Neurobasal medium, N2, B27, DMEM/F12, 2 μg/ml heparin, and 2 ng/ml Activin. For parasymN co-culture, day 16 SCPs were replated on day 12 SMCs in parasymN differentiation medium.

SN differentiation.

hPSCs were dissociated using EDTA and resuspended in Essential 6 Medium (E6) containing 10 μM SB431542, 1 ng/mL BMP4, 300 nM CHIR99021, and 10 μM Y-27632. Dissociated cells were plated at 200,000 cells/cm² density on vitronectin plates (as day 0). Next day, the cells were fed with the same medium. On day 2, cells were fed with D2–12 E6-based medium containing 10 μM SB431542, 0.75 μM CHIR99021, 2.5 μM SU5402 (Biogems, 2159233), and 2.5 μM DAPT. Cells were fed every two days between until day 12. On day 12, cells were dissociated with Accutase for 20 min and resuspended in neurobasal medium-based SN medium containing N2, B-27, 2 mM L-glutamine, 20 ng/mL GDNF, 20 ng/mL BDNF, 25 ng/mL NGF, 600 ng/mL laminin-1 and fibronectin, 1 μM DAPT, and 0.125 μM retinoic acid. Cells were replated at 200,000 cells/cm² on PO/LM/FN coated plates. Cells were fed every 3 days until day 30 for SARS-CoV-2 infection. A step-by-step protocol is also described in our previous publication^21,86.

Bulk RNA sequencing

RNA from parasympathetic neurons at different stages (day 10, 14, 16, 20, and 30) was isolated and purified as described above. Library preparation and sequencing was performed by Novogene. mRNA was purified using the NEBNext UltraII RNA Library Prep Kit for Illumina (NEB). Briefly, mRNA was isolated using poly-T oligo-attached magnetic beads followed by mRNA fragmentation. The first strand cDNA was then synthesized using random hexamer primers followed by the second strand cDNA synthesis. Library was analyzed using Qubit and real-time PCR for quantification and bioanalyzer for size distribution detection. Sequencing was performed using NovaSeq (Illumina) with a sequencing depth of approximately 40 million reads. All subsequent analyses were done on the Galaxy web platform (usegalaxy.org)¹²⁰. Quality control of the reads were assessed using FastQC (version 0.74, http://www.bioinformatics.babraham.ac.uk/projects/fastqc/). Low quality bases were trimmed from sequencing reads using Trimmomatic (version 0.38.1)¹²¹. Reads were then mapped to the human genome (GRCh38.p14) using HISAT2 (version 2.2.1)¹²² and QualiMap BamQC (version 2.2.2c)^123,124 was used to assess quality of the aligned reads. Reads per gene were counted using HTSeq (version 0.9.1)¹²⁵ with a minimum alignment quality value of 10. The raw count matrix was then processed using DESeq2 (version 2.11.40.8)¹²⁶ with default settings to analyze differential expression between sample groups, perform principal component analysis, and measure sample-to-sample distances. Genes that were significantly downregulated or upregulated (p-adj <0.05) were subjected to gene ontology analysis using the DAVID functional annotation tool (https://david.ncifcrf.gov/). Significant (FDR < 0.05) GO terms were plotted. To generate PCA plots, normalized counts of genes were plotted using ggplot2 (version 3.4.0, https://ggplot2.tidyverse.org/index.html and https://cran.r-project.org/web/packages/GetoptLong/index.html)¹²⁷. Additional RNAseq databases used were: sympathetic neurons (GSE212255)¹⁶, sensory neurons (GSE161530)¹²⁸, PFC (GSE146760)³¹, and motor neurons (GSE205718)³². Protein-interacting networks were obtained by analyzing highly upregulated genes (Log2 FC > 1) using STRING⁶⁰ in Cytoscape (version 3.10.1). Clusters were obtained using Markov Clustering Algorithm (MCL) using an inflation parameter of 2. Nodes showing critical genes involved in parasympathetic and sympathetic neuron development were further selected and their networks were graphed.

Single-cell RNA sequencing

Day 30 parasymNs were dissociated using 0.25% Trypsin-EDTA (Gibco, 25200056) and 1mg/ml collagenase NB4 (Nordmark, S17454) for 40 minutes at 4 °C. Single cell RNA-seq library was prepared using the PIPseqTM T20 3’ Single Cell RNA Kit v4.0 from Fluentbio. A total of 40,000 cells were loaded for each PIP (Particle-templated instant partition) reaction. Single cells were encapsulated in the droplet emulsion containing beads decorated with barcoded templates. After breaking the emulsion, cDNA was synthesized and amplified with 12 cycles. cDNAs are isolated from PIPs using SPRI (Solid-phase reversible immobilization) purification. Quality of the amplified cDNA was tested by the Fragment AnalyzerTM Automated CE System in the Georgia Genomics and Bioinformatics Core (GGBC) in the University of Georgia. Afterwards, 100ng of amplified cDNA was processed for fragmentation, A-tailing, adapter ligation, and indexing. The quality of the final library was tested by the Fragment AnalyzerTM Automated CE System in GGBC. Libraries were pooled using 15%PhiX and sequenced with the NovaSeq 6000 in the Emory Integrated Genomics Core (EIGC). Sequencing depth was 40,000 reads per cell. The Novaseq runs produced reads of 51 bp. We added electronically 3 T’s to the end of each read 1 sequence and a ! to the quality score. Afterwards, the FASTQ files were processed and mapped to reference genome GRCh38–2022.04 and single cells clustered using the PIPseeker v02.01.04 package from Fluentbio. The data samples provided as output from the PIPseeker software (https://www.fluentbio.com/products/pipseeker-software-for-data-analysis/) underwent standard quality control and normalization as outlined by [Seurat]¹²⁹ via the [SCANPY]¹³⁰ package^131,132. These high quality data underwent dimensionality reduction via the [PHATE]³⁴ algorithm, which preserves both global and local distances, to facilitate both visualization and [leiden]¹³³ clustering. Detecting cluster cell types from which, was aided by pairwise differential expression analysis between clusters and known marker genes. Gene ranking was performed using the Wilcoxon Rank-Sum (WRS) method. The mean expression of which, alongside the respective clusters were utilized by [SCANPY]’s `dotplot` function to produce Suppl. Fig. 3. To annotate the clusters, highly expressed genes in each cluster compared to all other clusters combined were used.

RT-qPCR

At least 0.5×10⁶ cells were collected using Trizol (Invitrogen, 15596026) for each sample. Reverse transcription was performed from 1 μg total RNA using iScript^™ Reverse Transcription Supermix (Bio-Rad, 170884). SYBR green (Bio-Rad) RT-qPCR was performed using CFX96 Touch Deep Well Real-Time PCR Detection System (Bio-Rad) and analyzed by CFX Maestro. For primers used in this study, please see Supplemental Table 1.

Western blot

ParasymNs and symNs on 6-well plates were prepared for sample collection. One well for each sample. Cells were washed with PBS and scraped after incubating with lysis buffer containing RIPA (Sigma-Aldrich, R0278, 1x), PMSF protease inhibitor (Thermo Fisher Scientific, 36978, 1 mM) and PhosSTOP phosphatase inhibitor (Sigma-Aldrich, 4906845001). Protein concentration was measured using Bradford reagent (Bio-Rad, 5000006). Proteins were loaded as 20 μg/well and ran in 12% acrylamide gel. After the transfer, nitrocellulose membranes were blocked by 5% skim milk in TBST. Membranes were incubated with primary antibodies overnight at 4 °C. The next day, membranes were washed by PBST and incubated with secondary antibodies for one hour at room temperature. Images were taken using the iBright western blot imager (Invitrogen, FL1500). For primary and secondary antibodies used in this study, please see KEY RESOURCES TABLE.

Key resources table.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
α-actinin	Sigma	A7811
αSMA	Sigma	A5228
AGTR1	Proteintech	25343-1-AP
AGTR2	Novus Biologicals	NBP1-77368
Anti-hm IgG	Proteintech	SA00003-12
Calponin	Abclonal	A3734
Cas3	Cell Signaling	9662
c-Cas3	Cell Signaling	9661
CD274	eBioscience	14-5983-82
ChAT	Millipore	AB144P
DAPI	Sigma	D9542
dsRNA	Millipore	MABE1134
FABP4	Abcam	ab92501
GAPDH	Cell Signaling	97166S
GFP	Abcam	ab13970
Healthy human IgG	Sigma	I4506
MIST1	Novus Biologicals	NBP2-22478
NPY2R	Lifespan Biosciences	LS-C120758
PHOX2B	Abcam	ab183741
PRPH	Santa Cruz Biotechnology	SC-377093/H0112
PUMA	Proteintech	55120-1-AP
SYP	Santa Cruz Biotechnology	sc-17750
SARS-CoV NP	BioVision	A2066
TH	Pel-Freez	P40101- 150
TUJ1	Biolegend	802001
VAChT	Abclonal	A16068
Bacterial and virus strains
SARS-CoV-2	BEI Resources	USA-WA1/2020
Biological samples
Sjögren’s syndrome patient serum_patient 1	Central BioHub GmbH	patient ID: 138237
Sjögren’s syndrome patient serum_patient 2	Central BioHub GmbH	patient ID: 153001
Healthy complement-active human serum	Innovative Research	ICSER
Chemicals, peptides, and recombinant proteins
Geltrex	Invitrogen	A1413202
Human fibronectin (FN)	VWR/Corning	47743-654
Mouse laminin I (LM)	R&D Systems	3400-010-01
Phosphate-buffered saline (PBS)	Gibco	14190-136
Poly-L-ornithine hydrobromide (PO)	Sigma	P3655
Vitronectin (VTN)	Thermo Fisher/Life Technologies	A14700
Accutase	Innovation Cell Technologies	AT104500
Ascorbic acid	Sigma	A8960-5G
B27 supplement	Thermo Fisher/Life Technologies	12587-010
BDNF	R&D Systems	248-BD
BMP4	R&D Systems	314-BP
CHIR99021	R&D Systems	4423
CNTF	R&D	257-NT-050
Collagenase	Tribioscienc	TBS2116-01
dbcAMP	Sigma	D0627
Dexamethasone	Sigma	D2915
Dispase II	Tribioscience	TBS2117-01
DMSO	Thermo Fisher/Life Technologies	BP231-100
DMEM	Thermo Fisher/Life Technologies	10829-018
DMEM/F12	Thermo Fisher/Life Technologies	11330-057
Fetal bovine serum (FBS)	Atlanta Biologicals	S11150
GDNF	PeproTech	450
E8 medium	gibco	A15169-01
E8 supplement	gibco	A15171-01
E6 medium	gibco	A15165-01
IBMX	Cayman	13347
L-glutamine	Thermo Fisher/Gibco	25030-081
N2 supplement	Thermo Fisher/Life Technologies	17502-048
Neurobasal medium	gibco	21103-049
NRG1	Peprotech	100-03
Oil Red solution	Sigma	O1391
Primocin (antibiotics)	InvivoGen	ANTPM1
recombinant FGF2	R&D Systems	233-FB/CF
Retinoic acid	Sigma	R2625
SB431542	Tocris/R&D Systems	1614
Trizol	Invitrogen	15596026
EDTA	Sigma	ED2SS
Y27632	R&D Systems	1254
Critical commercial assays
NEBNext UltraII RNA Library Prep Kit	NEB	E7775
iScript™ Reverse Transcription Supermix	Bio-Rad	170884
CM-DCFDA	Invitrogen	C6827
NAb™ Spin Kits	Thermo Scientific	89978
Acetylcholinesterase Assay Kit	AAT Bioquest	11400
Choline/Acetylcholine Quantification Kit	Sigma	MAK056
Amylase Assay Kit (Colorimetric)	Abcam	ab102523
Adipolysis Assay Kit	Cayman	10009381
Glucose Uptake Probe-Green	Dojindo	UP02-10
Deposited data
RNA-seq	This study	NCBI GEO: GSE253235
scRNA-seq	This study	NCBI GEO: GSE253235
Raw data	This study	Mendeley [10.17632/v462ccvfg8.1]
Experimental models: Cell lines
H9-ESC (WA09)	WiCell	NIH 0062
H9 Phox2B::GFP (WA09)	Oh et al.30	NIH 0062
H9 EF1::RFP (WA09)	Fattahi et al.20	NIH 0062
MEL1	Stem Cells Ltd	NIH 0139
652-hiPSC	Miller et al.55	GM01652
Cp1-hiPSC	Home-made	N/A
hDFn	Gibco (fibroblast)	C0045C
C1-hiPSC	Zeltner et al.23	AG02602
S2-hiPSC	Zeltner et al.23	GM04899
3T3-L1	ATCC	CL-173
Vero E6	ATCC	C1008
Experimental models: Organisms/strains
C57BL6 mice	The Jackson Laboratory	RRID:IMSR_JAX:00066
Oligonucleotides
Recombinant DNA
Software and algorithms
Fiji	https://fiji.sc	https://fiji.sc
Galaxy web platform	Afgan et al.120	https://usegalaxy.org
FastQC (v0.74)	Babraham Bioinformatics	http://www.bioinformatics.babraham.ac.uk/projects/fastqc/
Trimmomatic (v0.38.1)	Bolger et al.121	http://www.usadellab.org/cms/?page=trimmomatic
HISAT2 (v2.2.1)	Kim et al.122	http://daehwankimlab.github.io/hisat2/
QualiMap BamQC (v2.2.2c)	Garcia-Alcalde123 and Okonechnikov124 et al.	http://qualimap.conesalab.org
HTSeq (v0.9.1)	Anders et al.125	https://htseq.readthedocs.io/en/latest/
DESeq2 (v2.11.40.8)	Love et al.126	https://www.bioconductor.org/packages/release/bioc/vignettes/DESeq2/inst/doc/DESeq2.html
DAVID functional annotation tool	https://david.ncifcrf.gov/	https://david.ncifcrf.gov/
ggplot2 (v3.4.0)	Tang et al.127	https://ggplot2.tidyverse.org/index.html and https://cran.r-project.org/web/packages/GetoptLong/index.html
STRING (v3.10.1)	Szklarczyk et al.60	https://string-db.org
PIPseeker (v02.01.04)	Fluent BioSciences	https://www.fluentbio.com/products/pipseeker-software-for-data-analysis/
SCANPY package	Wolf et al.130	https://scanpy.readthedocs.io/en/stable/
PHATE algorithm	Moon et al.34	https://dburkhardt.github.io/tutorial/visualizing_phate/
Other

Immunohistochemistry

Cells were fixed by 4% paraformaldehyde for 20 minutes and washed twice by PBS. Fixed cells were permeabilized by 0.3% Triton, 1% BSA and 3% goat or donkey serum in 1x PBS for 20 minutes, and incubated with primary antibodies overnight at 4 °C. The next day, cells were washed and incubated with secondary antibodies for one hour at room temperature. Lastly, cells were washed and incubated with DAPI (Sigma, D9542, 1:1000) for 15 minutes at room temperature. Fluorescent images were taken using the Lionheart FX Automated Microscope. For primary and secondary antibodies used in this study, please see KEY RESOURCES TABLE.

MEA assay

For neural activity measurements, day 16 SCPs were plated on PO/LM/FN coated MEA plates (Axion BioSystems, BioCircuit or CytoView) at 100×10³ cells/cm². For hPSC-SMs or hPSC-SMCs co-cultured with parasymNs, day 10 CMs or day 6 SMCs were plated on Matrigel-coated MEA plates at 200×10³ cells/cm² or 35×10³ cells/cm², respectively. Activity of differentiated parasymNs or CMs/SMCs on desired timepoints was measured using a MEA plate reader (Axion BioSystems, Maestro Pro) under the neural detection mode or cardiac detection mode, respectively, according to manufacturer’s instruction.

SARS-CoV-2 infection

SARS-CoV-2 (USA-WA1/2020) was obtained from BEI Resources. Viral stocks were generated by infecting Vero E6 cells (ATCC, C1008) at ~95% confluency in 150-cm² flasks with SARS-CoV-2 at an MOI of 0.1 PFU per cell. At 68 h after infection, supernatants were collected, pooled and centrifuged at 400g for 10 min. The resulting stock was aliquoted, titered and stored at −80 °C for further use. All work with live SARS-CoV-2 was performed inside a certified Class II Biosafety Cabinet in a Biological Safety Level (BSL)-3 laboratory in compliance with all state and federal guidelines and with the approval of the University of Georgia Institutional Biosafety Committee. ParasymNs and SNs were infected with SARS-CoV-2 at a multiplicity of infection (MOI) of 1 in respective neuron media and incubated at 37°C/5% CO2. On day 3 pi, virus was removed, cells were gently washed in PBS and fixed in 4% formaldehyde for 3 days at 4°C. 3d post-fixation, wells were rinsed with PBS and stained to detect dsRNA and SARS-CoV Nucleoprotein.

ROS measurement

Cells were washed twice by PBS, and incubated with 10 mM CM-DCFDA (Invitrogen, C6827) in PBS for 45 minutes at 37 °C. After the incubation, cells were washed twice by PBS, and the fluorescent signal was measured using a fluorometer (BioTek) at 492 nm excitation and 520 nm emission.

Sjögren’s syndrome patient serum

Serum samples were purchased from Central BioHub GmbH. Considering that SS occurs mostly in women, only female samples were used for this study. One patient with peripheral nervous system involvement (patient ID: 138237, ICD-10 code: M35.06) and another with high anti-SSA level (patient ID: 153001, ICD-10 code: M35.0) were selected.

Antibody-based complement-dependent cytotoxicity assay

Total IgG was purified using the Nab^™ Spin Kits (Thermo Scientific, 89978) according to manufacturer’s instruction. Day 30 parasymNs were treated with 2% healthy complement-active human serum (Innovative Research, ICSER) and 200 nM SS IgG for 72 hours. Healthy human IgG (Sigma, I4506) was used as control.

AChE activity

After the antibody-based complement-dependent cytotoxicity assay, cell culture medium was collected for the assay or stored at −80 °C for long-term storage. AChE level was measured using the Amplite^™ Colorimetric Acetylcholinesterase Assay Kit (AAT Bioquest, 11400) according to manufacturer’s instruction.

ACh ELISA

For fig. 2c, parasymN culture medium was collected after 24 hours. For the antibody-based complement-dependent cytotoxicity assay (fig. 5e), cell culture medium was collected at the end of the assay. Medium samples were used directly for the assay or stored at −80 °C for long-term storage. ACh level in the medium was measured using the Choline/Acetylcholine Quantification Kit (Sigma, MAK056) according to manufacturer’s instruction. Media only control was conducted and subtracted from the presented data as blank.

Salivary acinar cell primary culture

Cell dissociation for 8-week-old Sox10-Cre/RFP mice was performed as described⁹⁵. Briefly, Collagenase (1 mg/ml, TBS2116–01, Tribioscienc) and Dispase II (2.5 mg/ml TBS2117–01, Tribioscience) enzymes were injected into the posterior tongue’s subepithelial space and incubated for 15 min at 37C. After the incubation, the epithelium was separated, and the von Ebner’s glands were carefully removed. These von Ebner’s gland tissues were dissected into tiny pieces before further incubating in 3 ml of 0.25% Trypsin EDTA (1 ml /mouse: #25200056, Fisher Scientific, Hampton, NH) for 30 min at 37 C. Trypsin treatment was stopped by carefully pipetting up and down with 1 ml of 10% FBS in PBS (Ca²⁺ and Mg²⁺ free, (#14190144, Fisher Scientific, Hampton, NH). The cell suspension was centrifuged at 450 g for 10 min at 4C. One-fourth of the supernatant was removed without disturbing the pellet. The remaining cell suspension was then filtered through 70 μm (#352350, Falcon, Fisher Scientific, Hampton, NH) and 35 μm (#64750–25, Electron Microscopy Science) filters. Cells were counted and plated at 15,000/cm2 density on tissue culture treated plates in medium that contains DMEM/F12, 1% FBS, and B27. A week after plating, a layer of salivary cells were formed, and day 16 SCPs were plated on the salivary cells. The co-culture was maintained in parasymN medium for 10 days and ready for further experiments.

Calcium imaging

Salivary cell and parasymN co-culture was washed with PBS and treated with 2 μM of Fluo-4 AM (TOCRIS, 6255) in fresh parasymN medium at 37 °C for 20 min. Cells were washed three times with PBS and incubated at 37 °C for another 30 min with fresh medium. After the incubation, cells were read using a fluorescence microplate reader (BioTek) at 440 nm excitation and 520 nm emission. The result was confirmed using the Lionheart FX Automated Microscope with EF1::RFP H9 hESC-derived parasymNs.

Amylase assay

Amylase activity was measured using the Amylase Assay Kit (Colorimetric) (Abcam, ab102523) according to manufacturer’s instruction. Cell lysate for each sample was collected from one well of 24-well plate by roughly scrapping the well using P1000 tips and pipetting for couple times in 200 μl of assay buffer. Samples were measured immediately or stored at −80C. Before measurement, samples were centrifuged at 300g for 5 min.

Adipocyte differentiation

3T3-L1 were seeded at 20×10³ cells/cm² and fed until 80% confluency, which was defined as day 0. From day 0–3, cells were fed with the differentiation medium containing DMEM (Gibco, 11965118), 10 % FBS, 0.5 mM IBMX (Cayman, 13347), 1 μM dexamethasone (Sigma, D2915), and 10 μg/ml insulin (Sigma, I9278). From day 3–6, cells were fed with post-differentiation medium containing DMEM, 10 % FBS, and 10 μg/ml insulin. After day 6, cells were maintained in DMEM with 10 % FBS until the desired timepoints.

Oil Red staining

Cells were fixed by 4% paraformaldehyde for 20 minutes and washed twice with PBS. Fixed cells were then incubated with 60 % isopropanol at room temperature for 5 minutes. After removing the isopropanol, cells were air dried at room temperature for 1–5 minutes, and incubated with the Oil Red solution (Sigma, O1391) at room temperature for 10–15 minutes. After staining, cells were washed with PBS for 2–5 times until no extra Oil Red was rinsed out. For Oil Red concentration measurement, Oil Red inside the cells was extracted by 100 % isopropanol for 15 minutes and measured by ELISA at 492 nm.

Lipolysis assay

To begin with, adipocyte culture medium was replaced with prewarmed lipolysis medium containing 0.5 % BSA in DMEM, and cells were incubated for 4 hours at 37 °C. After 4 hours, the lipolysis medium was replaced with fresh prewarmed lipolysis medium with test drugs and incubate for 90 minutes at 37 °C. Then, the cell culture medium was collected for the assay or stored at −20 °C for long-term storage. Glycerol release in adipocyte cultured medium was measured using Adipolysis Assay Kit (Cayman, 10009381) according to manufacturer’s instruction.

Glucose measurement

Medium samples from the lipolysis assay were used for glucose measurement. Glucose levels in the medium were measured and recorded by the glucose meter (Roche, ultra 2) according to manufacturer’s instruction.

Glucose uptake assay

To measure the efficiency of glucose uptake in WAT, cell culture was incubated in serum free and glucose free DMEM (Gibco, 11966025) with or without nicotine for 15 min. Cells were then fed by serum free and glucose free DMEM that contains the fluorescent probe, Glucose Uptake Probe-Green (Dojindo, UP02–10) and incubated for another 15 min. Intracellular glucose probe intensity was measured using a fluorescence microplate reader (BioTek) immediately at 440 nm excitation and 520 nm emission or imaged using a Lionheart FX Automated Microscope. The specificity of glucose signals in WAT was confirmed by differentiating co-cultured parasymNs from EF1::RFP H9 hESCs.

QUANTIFICATION AND STATISTICAL ANALYSIS

Data was collected from at least (or more) three independent experiments (biological replicates), with multiple technical replicates each. Biological replicates¹³⁴ are defined as independent experiments conducted several days apart either started from a new frozen vial of that particular cell line ¹³⁵, or started from a consecutive passage number of that cell line. Multiple clones derived from one patient line have been analyzed previously²³ and they were shown to not have significant variability, thus here we used one clone per iPSC line. Data is shown as mean ± SEM. Statistical analysis is described in figures. All the analyses and graphs were processed using Prism 9.

Highlights.

• Generation of functional, SCP-derived parasymNs from hPSCs

• ParasymNs and symNs mimic human ANS development

• Functional defects of parasymNs are characterized in models of FD, COVID-19, and SS

• Human parasymNs regulate adipose maturation and function in vitro