Generation of three-dimensional human neuronal cultures: application to modeling CNS viral infections

Background A variety of neurological disorders including neurodegenerative diseases and infection by neurotropic viruses can cause structural and functional changes in the central nervous system (CNS), resulting in long-term neurological sequelae. An improved understanding of the pathogenesis of these disorders is important for developing efficacious interventions. Human induced pluripotent stem cells (hiPSCs) offer an extraordinary window for modeling pathogen-CNS interactions, and other cellular interactions, in three-dimensional (3D) neuronal cultures that can recapitulate several aspects of in vivo brain tissue. Methods Herein, we describe a prototype of scaffold-free hiPSC-based adherent 3D (A-3D) human neuronal cultures in 96-well plates. To test their suitability for drug screening, A-3D neuronal cultures were infected with herpes simplex virus type 1 (HSV-1) with or without acyclovir. Results The half maximal inhibitory concentration (IC50) of acyclovir was 3.14 μM and 3.12 μM determined using flow cytometry and the CX7 High Content Screening platform, respectively. Conclusions Our A-3D neuronal cultures provide an unprecedented opportunity for high-content drug screening programs to treat human CNS infections. Electronic supplementary material The online version of this article (10.1186/s13287-018-0881-6) contains supplementary material, which is available to authorized users.


Background
Investigations to understand the pathogenesis of neurological and neuropsychiatric diseases have been hampered by the absence of satisfactory human neuronal models. The experimental approaches to modeling central nervous system (CNS) disorders and infections have changed profoundly thanks to human induced pluripotent stem cell (hiPSC) technologies. For the first time, it is possible to generate and manipulate nearly limitless numbers of hiPSC-derived neuronal lineage cells reprogrammed from specific individuals. Such an approach is essential to push the boundaries of personalized medicine since individuals show differences in susceptibility to common environmental, genetic, or infectious stressors. Thus, hiPSC-based models can be used to investigate the pathogenesis of neurotropic viruses and neuropsychiatric and neurodegenerative diseases at the cellular and molecular levels, modeling individual and cell-type specific differences in susceptibility to a given mutation, injury or pathogen, and as cellular platforms for drug screens.
Conventional drug screens employ two-dimensional (2D) cell cultures to enable simple, fast, and costeffective workflows, but they do not mimic the three-dimensional (3D) architecture of brain tissue in vivo. In recent comparisons between 2D cell monolayers and cells cultured as 3D cellular aggregates, the latter were better predictors of drug responses in vivo [1,2]. Indeed, cellular responses in 2D monolayer cultures differ from their in vivo analogs for several reasons: 1) impaired cell-cell communication, essential for cells to sense and respond to the environment; 2) reduced intercellular contacts, necessary to forward signaling information to groups of cells; and 3) lack of appropriate extracellular matrix (ECM)-cell adhesion interactions, which play a pivotal role in regulating many physiological processes [3]. The relatively low predictive value of the 2D culture system could account for a portion of drug failures during clinical trials [4]. Hence, there is an urgent need for the development of preclinical models that can more accurately predict clinical outcomes, saving both time and cost.
The first step in the generation of physiologically relevant cellular models is the addition of the "third dimension" to the classic 2D cultures. This is a "conditio sine qua non" dictated by the evidence that physiological responses, cell-to-cell signaling, and gene expression profile of 3D cultures more closely resemble in vivo conditions. Over the last decade, substantial literature has demonstrated the superiority of 3D culture systems over the classic 2D monolayer to provide physiological models and predict drug response in vivo [2,5]. These 3D culture models allow cell-cell and cell-ECM communication, both necessary for an adequate drug response. The use of hiPSCs to generate 3D culture models could revolutionize drug screening. Over the last few years, 3D neuronal cultures have been generated through hiPSC differentiation in scaffolds made of collagen [6], poly(lactic-co-glycolic acid)/poly(L-lactic acid) [7], collagen and hyaluronan [8], alginate [9], polycaprolactone, chitosan, cellulose acetate, and polyethersulfone nanofibers [10,11]. Furthermore, 3D neuronal cultures have been established within matrices, such as matrigel [5], or as spherical free-floating spheres known as neurospheres [12,13]. Notably, cerebral organoids, 3D aggregates that recapitulate multiple layers of brain-like regions, have been grown in matrigel droplets [14,15]. The ability of stem cells to spontaneously aggregate and organize has been reported and is believed to generate multiple neuronal microenvironments in cerebral organoids [16][17][18]. Cerebral organoids have become popular for modeling aspects of neurodevelopmental disorders, and have great potential to replace the classical 2D cellular platforms employed for drug screening to increase the predictability of drug response in vivo. Despite rapid progress in the development of 3D neuronal cell platforms, their applicability in moderate-to highthroughput drug screening is hindered by the technical complexity of the differentiation procedure and cost. Hence, the replacement of 2D cultures by 3D cultures in drug screening campaigns will require rapid, robust technologies amenable to analysis in a reliable manner, ideally using standard equipment. In this study, we investigated the feasibility of generating adherent 3D (A-3D) neuronal cultures from hiPSC lines derived from two different individuals. Using herpes simplex virus type 1 (HSV-1) infection as an example, we further demonstrate the utility of A-3D cultures grown in 96-well plates for CNS antiviral drug screening.

Cell lines
Generation of A-3D neuronal cultures from iPSCs hiPSC lines 73-56,010-02 and PPMI-51625 were obtained from RUCDR and PPMI as described above. Monolayer cultures of neural precursor cells (NPCs) were derived from hiPSCs 73-56,010-02 and PPMI-51625 as previously described [19]. 3D neuronal cultures were generated in optical quality (Nalgene Nunc, 400 μl volume, CC2 surface, sterile) 96-well plates treated with matrigel. To induce neuronal differentiation, matrigel was removed and hiPSC-derived NPCs were seeded at a density of 2 × 10 5 cells/well. NPCs were cultured in neurobasal medium supplemented with 2% B27, BDNF 10 ng/ml, GDNF 20 ng/ml, CHIR9901 3 μM, forskolin 10 μM, dorsomorphin 1μM, 50 U/ml penicillin G, and 50 mg/ml streptomycin. After 4 days, CHIR990, forskolin, and dorsomorphin were withdrawn and cells were cultured for an additional 6 weeks. Half of the culture medium was changed every other day. Cell viability in A-3D cultures was determined by flow cytometry, as described below. Cell viability was also analyzed using propidium iodide (PI) staining; cells were counterstained with Hoechst 33,342 for normalization.

Generation of 2D neuronal cultures from hiPSCs
NPCs derived from hiPSC lines 73-56,010-02 were seeded in matrigel-coated 96-well plates at a density of 5 × 10 4 cells/well and cultured in neurobasal medium supplemented with 2% B27 and BDNF 10 ng/ml for 6 weeks. Half of the culture medium was changed every other day.

Immunocytochemistry
Immunocytochemistry was performed as previously reported [19]. Cells were fixed in 4% paraformaldehyde and permeabilized with 0.2% triton-X. Primary antibodies used were mouse monoclonal anti-βtubulin III antibody (clone TUJ1, R&D System, 1: were used for detection. Counterstaining was performed with TO-PRO-3. A Leica SP5 CW-STED confocal (in normal confocal mode) was used to collect a virtual stack of optical sections taken from the sample. A 40× oil objective was used for acquisition with a magnetic resonance scanner, with sequential scanning for proximal wavelengths, at 0.5 μm optical section thickness, a line average of 8 and frame average of 6, 1024 × 1024 resolution, using conservative laser power settings and HyD detectors. The resultant image z-stacks were compiled into 3D images using LAS AF software, version 4.0 (Leica).

HSV-1 infections
We used a genetically engineered HSV-1 construct carrying enhanced green fluorescent protein (EGFP) under the control of the viral promoters ICP0, and monomeric red fluorescent protein (RFP) whose expression is driven by the viral promoter glycoprotein C [20]. EGFP expression in infected cells indicates that HSV-1 has entered lytic cycles, while RFP expression indicates commitment to viral DNA replication. Cells were pretreated for 24 h with increasing concentrations of acyclovir (ACV) ranging from 0.1 μM to 50 μM, after which cell-free virus was adsorbed onto cells for 2 h at a multiplicity of infection (MOI) of 0.3. The inocula were removed, cells washed, and then cultured in the presence of ACV. Cells were analyzed at 48 h postinfection (hpi).

Flow cytometry analysis
At 48 hpi, cells in A-3D and 2D cultures in 96-well plates were washed with phosphate-buffered saline (PBS) and treated with accutase (Sigma) for 15 min and 3 min, respectively, at 37°C. Single-cell suspensions were generated by pipetting up and down. The cells were transferred into conical 96-well plates and pelleted by spinning down for 15 min at 2500 rpm. The supernatants were discarded, and the cells were washed with PBS, pelleted, and stained with 'fixable' viability dye (LIVE/DEAD Fixable Aqua Dead Cell stain kit; Molecular Probes) for 20 min at room temperature, and then washed. The cells were pelleted and resuspended in 2.5% formalin. The percentage of fluorescent cells and cell viability were determined by flow cytometry (FC) using a LSR-Fortessa cell analyzer (Becton, MD).

High-content imaging
At 48 hpi, cells in A-3D cultures cells were washed with PBS and fixed in 4% paraformaldehyde (PFA) in PBS for 20 min at room temperature. The nuclei were counterstained with Hoescht 33,342. The multiwell plates were loaded into a Thermo Fisher Scientific Cellinsight CX7 high-content imaging platform for automated confocal image capture and analysis. The system is based on an inverted epifluorescence microscope featuring a nipkowspinning disc confocal that automatically focuses and scans fields in individual wells using a motorized stage. Fluorescent images were produced using a dichroic filter and corresponding emitter filters that pass the fluorescent light onto a high-resolution charge-coupled device camera for image acquisition. Objects were identified as cells if they had valid nuclei and cell body measurements based on size, shape, and fluorescence intensity. Acceptable ranges for these parameters were determined in healthy control wells to ensure that aggregated cells and noncellular particles were excluded from analysis. The acquired images were then analyzed simultaneously by the high-content screening (HCS) Studio software, using the compartmental analysis bioapplication to quantify puncta "spot" detection for each valid cell in the image.

Generation of A-3D neuronal cultures
The strategy for the generation of A-3D neuronal cultures relies on the ability of neural stem cells/neural progenitor cells to self-assemble into aggregates. We refer to the neural stem cells/neural progenitor cells collectively as neural precursor cells (NPCs). NPCs are seeded at high density and, because of their excessive number compared to the available surface in each well, the cells self-assemble in a multilayer fashion. We hypothesized that: 1) the self-assembling and the self-organization of the NPCs and their derived progeny in a 3D space would be accompanied by production of ECM; and 2) the "spheroidalization" of the aggregating cells (leading to the formation of spheres heterogeneous in size) could be prevented by seeding NPCs into matrigel-treated wells. NPCs were generated from hiPSCs line 73-56,010-02 and PPMI-51625 as previously described [19] and seeded on matrigel-treated optically clear 96-well plates at a density of 2.5 × 10 5 cells/well. The differentiation procedure of NPCs to generate A-3D neuronal cultures is described in the Methods section. During the differentiation process, a higher density of NPCs was observed toward the center of the culture wells, with a consequent increase in the thickness of the 3D cell aggregates (Fig. 1). Our differentiation procedure generated multilayered cell aggregates showing comparable size (Fig. 2), whose thickness ranged from approximately 25 μm to 60 μm after 4 weeks of NPC differentiation (Fig. 1). Furthermore, HCS analysis indicated comparable average cell body total intensity (mean = 69.26 RFU, standard deviation (SD) = 13.66) and average neurite total length (mean = 48,391 μm, SD = 11,111.62) among different culture wells (Fig. 2). After 4 weeks of NPC differentiation, the composition and the distribution of cells in the A-3D cultures was investigated. Cell viability, estimated by flow cytometry using a 'fixable' viability dye or by exclusion of propidium iodide, was 80 to 90%.
Confocal immunofluorescence microscopy of A-3D cultures differentiated from a healthy donor [21] (73-56,010-02) showed a multilayered structure in which neuronal cells expressing the neuronal marker TUJ1 were sandwiched between cell layers expressing the NPC marker nestin (Fig. 1a-c). The production of ECM from differentiating cells in our A-3D cultures was evaluated by analyzing the expression of proteoglycans, the main component of the ECM in the CNS [22]. Figure 1d-f shows abundant staining with chondroitin sulfate proteoglycans (types A and C) which play an important role in the neuronal migration [22], indicating that the ECM produced by the differentiating cells sustains the 3D architecture. The presence of structures resembling developing human cortex was next investigated. A large fraction of cells in the layer above the dense region of MAP2-positive neuronal cells showed glial fibrillary acidic protein (GFAP)-positive (Fig. 1g-i) and vimentinpositive extensions (Fig. 1j) [23,24] resembling radial glial cells, which are located in the ventricular zone [24]. The mean fraction of cells staining for MAP2 and vimentin were 72.4% (SD = 12.51) and 30.6% (SD 3.37), respectively (Fig. 1p). Staining with the progenitor cell marker CUX-2 (a homeodomain transcription factor expressed in the subgranular zone (SGZ)) was mainly observed in the NPC layer of our 3D cultures (Fig. 1k). CUX2 has several regulatory functions, including the proliferation of the NPCs in the SGZ that will generate upper layers neurons, dendritic growth, and the integration of the upper cortical layers II-III neurons [25]. Western blot analysis indicated that NPCs differentiated into glutamatergic, GABA-ergic, and dopaminergic neurons (Fig. 3). Furthermore, besides the expression of GFAP, Western blot analysis revealed the expression of the microglial marker ionized calcium binding adaptor molecule 1 (Iba1) (Fig. 3). A subpopulation of neurons stained for calbindin (Fig. 1m), a cytosolic calciumbinding protein with greater expression in layer III of pyramidal cells in the dorsolateral prefrontal cortex (DLPFC) compared with layer V pyramidal cells in the same region [26]. Immunoreactivity for CTIP2, a zinc finger protein expressed in layer V neurons, was detected in a subset of neurons [27] (Fig. 1n). Tau protein, which progressively segregates into the axons during neuronal differentiation [28], was abundantly expressed (Fig. 1o). The peripheral area between the 3D cell structure and the edge of the culture well was mainly occupied by processes extending from differentiating cells. In summary, these A-3D cultures showed features of a developing cortex.
To further demonstrate the utility of this method to study CNS disease pathogenesis, we employed this differentiation procedure to generate A-3D cultures using PPMI-51625, a hiPSC line derived from a Parkinson's disease patient carrying the G2019S mutation in the LRRK2 gene. This line also spontaneously formed a multilayered organoid structure in which neuronal cells are sandwiched between GFAP-expressing glial cells (Additional file 1: Figure S1). Successful generation of A-3D cultures from both healthy individuals and patients with disease-causing mutations underscores the potential uses of this approach to study cellular-, molecular-, and tissue-based mechanisms and define new neuroprotective therapies for a wide range of neurological disorders. We utilized the neurotropic virus HSV-1 to demonstrate suitability of the A-3D culture system for high-throughput drug screening. A robust, accurate, and rapid analysis of multilayer cellular structures is central to successful employment of 3D cellular platforms for drug screening. We have previously developed 2D models of lytic and latent HSV-1 infections in hiPSC-derived neuronal cells [20,21,29]. Our 3D culture models also exhibit lytic infection with HSV-1 (Fig. 4). We therefore tested the antilytic effects of a range of concentrations (0.1 μM to 50 μM) of ACV against a recombinant HSV-1 strain based on KOS virus expressing EGFP and RFP under the control of the viral promoters ICP0 and glycoprotein C (gC), respectively [20], using A-3D cultures of neurons generated from the hiPSC line 73-56,010-02. Cells were pretreated with ACV for 24 h, after which they were infected at a MOI of 0. The drug concentration that reduced the number of fluorescent cells in cultures infected with HSV-1 by 50% (IC50) was estimated using the four-parameter log-logistic function model in the drc (V3.0-1) package in R (3.4.0). The IC50 determined by FC and HCS was 3.144 μM and 3.121 μM, respectively (Fig. 4). The similarity between these values derived from traditional FC and high-throughput screening indicates that robust, accurate, and rapid analysis of complex brain cell structures can be achieved using the A-3D culture system.

Determination of the IC50 for ACV using 2D cultures
To investigate whether ACV exhibits differential efficacy between 2D and 3D cultures, hiPSC-derived monolayer cultures of neurons were generated in 96-well plates (see Methods section). Cells were pretreated with ACV at different concentrations ranging from 0.1 μM to 50 μM and infected with the previously described HSV-1 construct at MOI 0.3. The percentage of fluorescent cells was analyzed using FC. The IC50 of ACV in our 2D neuronal cultures determined using FC was 0.5 μM (Fig. 4).

Discussion
We have developed a reproducible procedure to differentiate hiPSCs into 3D neuronal cultures in a 96-well plate format that can be used to model CNS infections or genetic contributions to neurological diseases. This A-3D culture system enabled us to model HSV-1 infection and test the antiviral potency of ACV, the gold standard therapy for HSV-1 infections. We showed that the generation of A-3D neuronal cultures does not require that cells be embedded in matrices or seeded into artificial scaffolds. Instead, after an initial period of NPC self-assembly, their differentiating progeny organize themselves in multilayered cell structures showing features of the ventricular/ subventricular zone and neuronal cells with some features of human cortical layers. The abundant secretion of ECM, highlighted by immunostaining with chondroitin sulfate (Fig. 1d-f), provides another important advantage with our A-3D platform, considering the influence of ECM on drug activity [3]. We provide evidence for acceptable wellto well variability in our 3D cultures (Fig. 2). We then used two sensitive and quantitative platforms, flow cytometry (FC) and high-content screening (HCS), to determine the IC50 of acyclovir in HSV-1-infected A-3D cultures. The HCS platform yielded estimates comparable to FC (3.1 μM, Fig. 4). The IC50 of ACV estimated in 2D neuronal cultures using FC was 0.5 μM, indicating higher potency of ACV in 2D cell cultures when compared with 3D cell cultures. This may be due to known activities of ECM in binding and regulating the bioavailability of soluble mediators such as drugs and trophic factors, or to a decreased penetrance of ACV in the A-3D cultures.

Conclusions
Our A-3D human neuronal cultures, paired with a novel confocal-based HCS platform (CX7), provide an unprecedented opportunity to develop a rapid and robust drug screening tool for CNS diseases. There are two distinct advantages of the CX7 HCS technology over FC to analyze 3D cultures. The first comes from the accurate and rapid acquisition of confocal z-stacks from multilayer cellular structures, eliminating laborious and time-consuming manual steps for cell dissociation and staining required for FC. The second advantage is rapid data analysis. To our knowledge, this represents the first evidence for the feasibility of antiviral drug screening in 3D hiPSC-derived neuronal platforms. Our work provides several avenues for future development. Besides additional characterization of A-3D models of CNS infections or neurodegeneration, the scalability of our platform needs to be investigated for optimal drug screening.

Additional file
Additional file 1: Figure S1.