Short and long TNF-alpha exposure recapitulates canonical astrogliosis events in human induced pluripotent stem cells-derived astrocytes

Astrogliosis comprises a variety of changes in astrocytes that occur in a context-specific manner, triggered by temporally diverse signaling events that vary with the nature and severity of brain insults. However, most mechanisms underlying astrogliosis were described using animal models, which fail to reproduce some aspects of human astroglial signaling. Here, we report an in vitro model to study astrogliosis using human induced pluripotent stem cells (iPSC)-derived astrocytes which replicates aspects temporally intertwined of reactive astrocytes in vivo. We analyzed the time course of astrogliosis by measuring nuclear translocation of NF-kB, secretion of cytokines and changes in morphological phenotypes of human iPSC-derived astrocytes exposed to TNF-α. It was observed the NF-kB nuclear translocation, increases either in the inflammation-related cytokines secretion and gene expression for IL-1β, IL-6 and TNF-α following 24 h TNF-α stimulation. After 5 days, human iPSC-derived astrocytes exposed to TNF-α exhibited increases in vimentin and GFAP immunolabeling, elongated shape and shrinkage of nuclei, which is typical phenotypes of astrogliosis. Moreover, about a 50% decrease in D-[3H] aspartate uptake was observed over the astrogliosis course with no evident cell damage, which suggests astrocytic dysfunction. Taken together, our results indicate that cultured human iPSC-derived astrocytes reproduce canonical events associated to astrogliosis in a time dependent fashion. Our findings may contribute to a better understanding of mechanisms governing human astrogliosis. Furthermore, the approach described here presents a potential applicability as a platform to uncover novel biomarkers and new drug targets to refrain astrogliosis associated to human brain disorders.

As a cell population with heterogeneous morphology and functioning, astrocytes from rodents and humans show distinct features regarding morphology, gene expression and functional competencies (Zhang et al., 2016). Roughly, the population of human cortical astrocytes is composed of larger and morphologically more complex cells when compared to cortical astroglia from rodents (Oberheim et al., 2009). Yet, multiple subclasses of astrocytes described in the human neocortex are not found in the same region of the murine brain (Oberheim et al., 2009).
Astrocytes are prompt to react after inflammatory insults, experiencing a continuum of temporally-ordered physiological changes that culminates in a process called astrogliosis. Initially, mechanical or pathological injuries in the CNS trigger NF-kB signaling in astrocytes leading to an increased production of NF-kB-dependent cytokines, which potentially activate these cells (Lattke et al., 2017;Saggu et al., 2016). Once activated, astrocytes show progressive upregulation of intermediate filaments GFAP and Vimentin (Liu et al., 2014) and heterogenous degrees of cell hypertrophy (Kang, Lee, Han, Choi, & Song, 2014). As long as the reactive phenotype is maintained, astrocytes may present impaired metabolic functions such as disrupted recycling of neurotransmitters (Schreiner, Berlinger, Langer, Kafitz, & Rose, 2013) and energy metabolism (Gavillet, Allaman, & Magistretti, 2008). These biological events are intertwined and occur within a time course which determines the extension of the lesion (Burda & Sofroniew, 2014). However, it is unclear whether human astrocytes exhibit such response patterns since most of our knowledge about astrogliosis comes from murine systems. Currently, divergent gene expression profiles found in mouse astrocytes were activated by distinct stimuli (Hamby et al., 2012;Zamanian et al., 2012), while mouse and human astrocytes did not share the same responses to lipopolysaccharide (LPS) and IL-1 (Tarassishin, Suh, & Lee, 2014). These results highlight interspecies differences regarding astrocytic responses.
Recently, murine astrocytes exposed to distinct activation stimuli were classified in A1 and A2 according to their transcriptomic profile. Interestingly, together with other cytokines, TNF-α was able to induce both profiles (Liddelow et al., 2017). Indeed, TNF-α emerges as a key player on induction, maintenance and profiling of astrogliosis.

Neural cells derived from induced pluripotent stem cells (iPSC) have
increasingly been used to model human diseases. In the last few years, efforts have been made to improve differentiation protocols in order to obtain and characterize human iPSC-derived astrocytes (Chandrasekaran, Avci, Leist, Kobolak, & Dinnyes, 2016;Emdad, D'Souza, Kothari, Qadeer, & Germano, 2012;Shaltouki, Peng, Liu, Rao, & Zeng, 2013;Tcw et al., 2017;Yan et al., 2013). Owing to the uniqueness of human astrocytes as well as their responses to insults and inflammation, we evaluated human iPSCderived astrocytes regarding the canonical events associated with astrogliosis in response to short-and long-term TNF-α stimuli as a mimetic neuroinflammatory condition. These data were obtained following a circumspect improvement of the protocol for obtaining human iPSC-derived astrocytes described by Yan and colleagues, allowing us to show that these cells present typical astroglial functional features (Yan et al., 2013). In order to provide not only a characterization of human astrocyte-secreted proteins following inflammatory stimulus, we also evaluated and characterized major events related to the time course of astrogliosis such as i) TNF-α-mediated nuclear translocation of NF-kB, ii) increases in secretion and gene expression of inflammation-related cytokines, iii) alterations in the cell shape affecting the aspect ratio and decreasing the area of astrocyte nuclei and iv) dysfunctional aspartate/glutamate uptake. Although human iPSC-derived astrocytes have been shown to respond to inflammatory stimuli (Perriot et al., 2018;Santos et al., 2017), it was unclear whether human iPSC-derived astrocytes can reproduce in vitro the temporally-ordered physiological changes typical of astrogliopathology.

Generation of human iPSC-derived astrocytes
Human iPSC-derived astrocytes were differentiated from human iPSC-derived neural stem cells (NSC) obtained from iPSC of four healthy subjects. Neural stem cells (NSC) were obtained according to a protocol described previously (Yan et al., 2013). Three of these cell lines were used in other studies from our research group (Casas et al., 2018;Garcez et al., 2017 During this period, when reaching confluence, cells were passed at a ratio of 1:4 using Accutase (A6964 -Sigma Aldrich, MO, USA) to new Geltrex-coated flasks. AIM was changed on the day after cell passaging. By the end of the 21 days of differentiation, cells were exposed to astrocyte medium containing 10 % FBS in DMEM/F12. At this stage cells were named human iPSC-derived radial glia-like cells due to their positive labeling for radial glia markers PAX6 and phosphorylated Vimentin, which decrease with time in culture (Garcez et al., 2017). From then on, after reaching confluence, cells were passed at a ratio of 1:2 in the absence of Geltrex. Medium was changed twice a week and on the following day after every cell passage. Human iPSC-derived radial glialike cells were kept in culture for additional four weeks in order to expand the number of cells and to enhance maturing of astroglial functions. All experiments described in this paper were performed from day 49 after the beginning of astrocyte differentiation (Fig. 1a).

In vitro Astrogliosis
Human iPSC-derived astrocytes were seeded in non-coated multiwell plates or 75 cm 2 flasks at densities varying from 2 x 10 4 cells/cm 2 to 6.25 x 10 3 cells/cm 2 . Astrocyte medium was changed at day 1 and 3 after each passage.
At day 4, astrocyte medium was replaced by serum deprived medium (DMEM/F12) for 24 h and thereafter 1 -50 ng/ml of human recombinant TNFα (717904 -BioLegend, CA, USA) or vehicle (DMEM/F12) were added directly to the cells. Cells were then cultured for an additional 1 h, 24 h or 5 days and samples were collected/processed/analyzed according to the experiment (Fig. 1b).

Immunocytochemistry and cell morphology analyses
Human iPSC-derived astrocytes were seeded at a density of 6.25 x 10 3 cells/cm 2 on 96-multiwell Cell-Carrier plates (PerkinElmer, MA, USA). Four days after plating, cells were exposed to serum-deprived medium and in the following day, concentrations of TNF-α ranging from 1 -50 ng/ml were added.
For NF-kB translocation experiments, cells were fixed 1 h after TNF-α exposure. For GFAP, Vimentin and morphology analyses, cells were fixed 5 days after TNF-α stimulus. Cells were fixed with 4 % paraformaldehyde (Sigma Aldrich, MO, USA) in phosphate-buffered saline for 20 minutes, permeabilized with 0.3% Triton X-100 (Sigma Aldrich, MO, USA) and then A-11032 -Thermo Fisher Scientific, MA, USA). Nuclei were stained with 0.5 µg/mL 4′-6-diamino-2-phenylindole (DAPI) for 5 minutes. Images were acquired with the Operetta high-content imaging system using 10 x and 20 x objectives (PerkinElmer, MA, USA). Total number of cells was calculated by DAPI stained nuclei counting. Densitometry and cell morphology analyses were performed using software Harmony 5.1 (PerkinElmer, MA, USA). Eleven different fields from triplicate wells per experimental condition were used for quantification.

Multiplex analysis of cytokines and BDNF
Cultures of human iPSC-derived astrocytes were seeded at a density of 2 x 10 4 cells/cm 2 in 75 cm 2 culture flasks. Four days after plating, cells were exposed to 24 h of serum-deprived medium. Then, 10 ng/ml TNF-α was added for an additional period of 24 h. After this incubation time, conditioned medium was collected, aliquoted and immediately frozen and stored at -80 o C.
ProcartaPlex™ bead-based multiplex immunoassays were performed for simultaneous detection and quantitation of multiple protein targets in cell culture supernatant. The platform MAGPIX™ was used for simultaneous detection of BDNF, INF-γ, IL-1β, IL-10, IL-13, IL-2, IL-4, IL-6, IL-8 and TNF-α in a single sample according to the manufacturer's instructions. Briefly, 50 μ l/well of coated magnetic beads solution (Luminex™ Corporation) were added to a 96-well plate. The plate was inserted onto a magnetic plate washer; beads were allowed to accumulate on the bottom of each well and were then washed. Samples or standards (50 μ l) were added to the wells and then incubated at room temperature for two hours on a Capp 18-X multiwell plate shaking platform (500 rpm). The plate was washed again twice and the detection antibody mixture was added and incubated at room temperature for 30 minutes on a shaking platform (500 rpm). The streptavidin solution was added to each well and incubated at room temperature for 30 minutes (500 rpm). After washing, reading buffer was added into each well and incubated at room temperature for 5 minutes (500 rpm). Data were acquired on MAGPIX™.

Quantitative Real time PCR
Cultures of human iPSC-derived astrocytes were seeded at a density of 2x10 4 cells/cm 2 in T-75 culture flasks. Four days after plating, cells were exposed to serum-deprived medium. After 24 h of serum deprivation, 10 ng/ml TNF-α was added for an additional period of 24 h. Cells were then detached with Accutase (Merck, Darmstadt, Germany), centrifuged at 300 g for 5 minutes after which the supernatant was discarded and the resulting pellet was submitted to RNA extraction. Total RNA was isolated using TRIzol™ reagent, All primers used in this study were validated by serial dilution to allow reaction efficiency calculation. Efficiencies from all tested primers ranged from 90%-110%, and the maximum allowed difference in efficiencies for primers used to estimate fold change of each gene was 10% (data not show). RT-qPCR data was calculated by the 2-ΔCt method using the geometric mean of RPLP0, GAPDH and IPO8 (IL-1β) or GAPDH and HPRT1 (IL-6) to normalize results.

D-Aspartate uptake
Cultures of human iPSC-derived astrocytes were seeded at a density of 2x10 4 cells/cm 2 on 24-well plates. Four days after plating, cells were exposed to serum-deprived medium.

Statistical Analyses
Statistical comparisons between two experimental groups were performed by Unpaired Student t-test. NF-kB translocation experiments were analyzed by One-way ANOVA followed by Tukey´s post hoc test. Aspartate uptake experiments were analyzed by One-way ANOVA followed by Dunnett´s post hoc test. Statistical significance was considered for P < 0.05. GraphPad Prism v8.02 (GraphPad Software, CA, USA) was the software used for data analyses and graphics.

TNF-α induces NF-kB nuclear translocation in iPSC-derived astrocytes
First, we aimed to determine whether human iPSC-derived astrocytes were responsive to TNF-α by analyzing nuclear translocation of the transcription regulator NF-kB, a mandatory step on astrogliosis induction. It can be noted that cells incubated for 1 hour with TNF-α ranging from 1 to 50 ng/mL showed nuclear translocation of NF-kB in all tested conditions (Fig. 2a). The increase of NF-kB in the nucleus was concentration-dependent, with 1, 5, 10, 20 and 50 ng/ml TNF-α promoting increases of NF-kB immunoreactivity around 97%, 138%, 156%, 186% and 184%, respectively (Fig. 2b). When the whole cell area was analyzed, a trend toward increasing was observed for NF-kB in all concentrations of TNF-α tested, although it did not reach statistical significance (Fig. 2c). The translocation index of NF-kB (calculated by immunoreactivity of nuclei/cell area ratio), revealed that TNF-α induced a significant increase in the NF-kB translocation index in all concentrations tested. Similar to nuclei NF-kB immunoreactivity, the increase was dose dependent, as 1, 5, 10, 20 and 50 ng/ml of TNF-α promoted an increase of NF-kB nuclear translocation around 74%, 110%, 106%, 130% and 138%, respectively (Fig. 2d). Despite the fact that cells were responsive to 1 ng/ml TNF-α for NF-kB nuclei translocation, cells that received 10, 20 or 50 ng/ml TNF-α were even more effective on recruiting NF-kB to the nucleus. Based on these findings, 10 ng/mL TNF-a was chosen for further experiments in order to promote astrocytes activation.

Inflammation-related cytokines are increased after TNF-α stimulation
The release of cytokines is a hallmark of inflammatory response and also astrogliosis. Therefore, in order to map the secretion of immunomodulators by human iPSC-derived astrocytes activated by TNF-α, we performed a multiplex analysis of cytokines using conditioned media from these cells. Based on the literature, cytokines were classified into three major categories related to their primary biological roles on inflammation: pro-inflammatory, modulatory or antiinflammatory, and also brain-derived neurotrophic factor (BDNF). The secretion of pro-inflammatory cytokines IL-1β, IL-8 and IFN-γ was significantly stimulated by TNF-α. While IL-1β secretion was increased in 425% (from 12.32 ± 1.11 to 64.72 ± 14.05 pg/mL) and 254% for IL-8 (from 1.09 ± 0.16 to 3.85 ± 0.39 ng/mL), IFN-γ presented 890% increase (from 47.9 ± 22.5 to 474.3 ± 147.8 pg/ml) when compared to conditioned media from vehicletreated cells (Fig. 3a). As expected, IL-1β, IFN-γ and TNF-α were barely detectable in conditioned media of non-stimulated human iPSC-derived astrocytes. TNF-α was highly detected in the conditioned media from stimulated cells, probably due to its remaining content from the activation stimulus.

Inflammation-related cytokines display increased expression after TNF-α stimulation
Despite the fact that most cytokines have short half-life, increases in their secretion should impact gene expression and continuous production was expected. Thus, quantitative real time PCR (qPCR) was conducted to verify the expression of cytokines, which secretion was found to be increased in the conditioned medium (Fig. 4). Indeed, both IL-6 and IL-1β genes expression showed ~15.6-and ~ 6.0-fold increase, respectively, following TNF-α exposure (Fig. 4a,b). In addition, TNF-α expression was detected only in stimulated cells, but not in any replicates within the control group (Fig. 4c). In order to show that control samples were viable in terms of gene expression, we also analyzed the expression levels of housekeeping genes RPLP0, GAPDH and IPO8 from the same samples used in Figures 4b and 4c. No differences were found among experimental groups (Supplementary Figure   1).

TNF-α induced morphological alterations and upregulation of intermediate filaments Vimentin and GFAP
The incubation of astrocytes with 10 ng/ml TNF-α during five consecutive days (in serum-free conditions) triggered typical phenotypes of astrogliosis, as revealed by increased immunoreactivity for Vimentin and GFAP (Fig. 5a, b, c) as well as shrunken nuclei (Fig. 5f). The incubation with 10 ng/ml TNF-α did not change measurements of cell area (Fig. 5c). However, some vimentinstained astrocytes clearly showed a thin, process-devoid and polarized phenotype after treatment with 10 ng/ml TNF-α. To identify these cells, we calculated the aspect ratio (length to width ratio) from vimentin immunoreactivity. A greater length/width ratio would indicate cells with relatively thin shapes (Fig. 5f). We confirmed that TNF-α promotes an increase in cell polarization (25%), which is typically found under inflammatory conditions (Jang et al., 2013).

Glutamate uptake is impaired in human iPSC-derived astrocytes exposed to TNF-α
One important functional feature of astrogliosis is the dysfunction of glutamate/aspartate transporters. In order to accurately measure uptake activity, the non-metabolized analog D-[ 3 H] Aspartate was used. Aspartate uptake was carried out 1 and 5 days after TNF-α incubation. At both time points, TNF-α was able to inhibit D-[ 3 H]aspartate uptake by human iPSCderived astrocytes (Fig. 6a,b). At 60 minutes of aspartate uptake, a reduction of ~47% was found in cells exposed to TNF-α for 1 day (Fig. 6a). When cells were exposed to TNF-α for 5 days, a reduction in aspartate uptake was observed at 5, 15, 30-and 60-minutes time points (61.5%, 72.4%, 73.3% and 51.5%, respectively) when compared to control cells (Fig. 6b). Also, human iPSC-derived astrocytes displayed a highly specific aspartate uptake activity, as evidenced by full inhibition with the competitive inhibitor of excitatory amino acid transporters DL-TBOA (Fig. 6a, b). In the presence of DL-TBOA, TNF-α had no effect on transporters activity (Fig. 6a, b). Since the astrogliosis induction was carried out in the absence of serum, we assessed astrocyte viability after five consecutive days of TNF-α exposure. The removal of fetal bovine serum from culture medium induced a small increment in cell death: 5% of cells incorporated ethidium bromide while only 0.7% were ethidiumpositive among serum-supplemented cells (Fig. 6c). Similarly, TNF-α in the absence of serum decreased astrocyte viability in 7.4%. No significant difference in cell death was observed between cells exposed to TNF-α and vehicle (Fig. 6c).

Discussion
In this study, human iPSC-derived astrocytes were characterized with regard to their responsiveness to short and long-term inflammatory conditions. Thus, human iPSC-derived astrocytes from healthy subjects were challenged with the pro-inflammatory cytokine TNF-α, which is able to initiate and sustain long-term features of astrogliosis (Cui, Huang, Tian, Zhao, & Zheng, 2011).
In glial cells, NF-kB remains in its physiologically latent cytoplasmic form, being translocated to the nucleus under inflammation (Kaltschmidt & Kaltschmidt, 2009). We noticed that human iPSC-derived astrocytes were capable of responding to short term TNF-α exposure through the translocation of NF-kB to the nucleus, a predicted event that occurs within an hour after inflammatory stimuli (Nelson et al., 2004;Romano, Freudenthal, Merlo, & Routtenberg, 2006).
Secretome analysis revealed that high levels of TNF-α were detected in the conditioned media 24 h after exposing cells to this cytokine, but it was difficult to distinguish between residual levels arising from the stimulus or secretion.
However, TNF-α was almost undetectable in the conditioned medium and cell extracts from non-stimulated astrocytes, confirming that TNF-α is essentially produced by stimulated astrocytes. Since TNF-α via NF-kB activation triggers inflammatory events such as cytokines production, we also found that upon short-term TNF-α stimulation (24 h), either modulatory cytokines IL-2, IL-4 and IL-6 or the chemotactic and inflammatory cytokine IL-8 had their levels increased from pg/mL to ng/ml range, which is typical for inflammatory responses or pathological conditions (Stenken & Poschenrieder, 2015).
However, IFN-γ and BDNF were also increased in the conditioned medium from astrocytes exposed to TNF-α, but to a lesser extent. It is important to take into account that over the course of acute inflammatory stimuli many factors and cytokines are expressed in a coordinated manner. Since IFN-γ is a pro-inflammatory factor, increases in its secretion may be important to induce and sustain astrogliosis (Yong et al., 1991). In a previous report, cultured rat astrocytes were exposed to different inflammatory stimuli, but only TNF-α was able to increase expression and secretion of BDNF via NF-kB activation (Saha, Liu, & Pahan, 2006). Even though the activity-dependent secretion BDNF is well documented in neuronal cells, the secretory nature of astrocytes suggests that increased BDNF secretion by TNF-α might be an attempt of stimulated-astrocytes to provide a neurotrophic factor that will rescue viability following acute inflammatory stimuli.
Interestingly, relatively high levels for IL-6 and IL-8 were detected in the conditioned medium from non-stimulated astrocytes (1.6 and 1.0 ng/ml, respectively). Of note, astrocytes are a major source of IL-6, which is important for synapse formation (Wei et al., 2011), maturation of dendritic spines (Wei et al., 2012) and sprouting of glutamatergic connections (Menezes et al., 2016), while IL-8 regulates angiogenesis by enhancing survival and proliferation of endothelial cells (Li, Dubey, Varney, Dave, & Singh, 2003).
As expected for mature astrocytes, almost all human iPSC-derived astrocytes were immunoreactive for GFAP and vimentin, though not all astrocytes necessarily express GFAP in the human brain tissue (Kettenmann & Verkhratsky, 2011;Kimelberg, 2004). Both GFAP and vimentin are considered markers of astrocytes and their overexpression characterizes primary features of reactive astrogliosis, since their deletion reduces reactive gliosis and promotes synaptic regeneration (Wilhelmsson et al., 2004). TNF-α is fairly known to induce overexpression of GFAP and vimentin (Perriot et al., 2018;Zhang et al., 2016) and our human iPSC-derived astrocytes were responsive to long-term TNF-α exposure by increasing GFAP and vimentin.
The long-term exposure to TNF-α was also able to alter astrocyte morphology in vitro, since they displayed an elongated shape, namely polarized astrocytes, resembling those in response to a wound (Peng & Carbonetto, 2012) or recruited to mold the glial scar around injured sites (Adams & Gallo, 2018).
It is well documented the crucial role of astrocytes in maintaining glutamate homeostasis by continuous removal of extracellular glutamate from the synaptic cleft by glutamate transporters, a mechanism that avoids glutamate excitotoxicity (Rothstein et al., 1996). Recent evidences have shown that extracellular glutamate stimulates glutamate release from astrocytes in order to coordinate neuronal activity, pointing to an emerging role for astrocytes in modulating both glutamate uptake and release and extending their therapeutic potential (Mahmoud, Gharagozloo, Simard, & Gris, 2019). It could be noted that iPSC-derived astrocytes exposed to short term TNF-α had already showed impairment in the aspartate uptake. As astrogliosis progresses due to long term exposure to TNF-α, the aspartate uptake impairment worsened.
Importantly, the impairment of glutamate uptake by prolonged exposure to TNF-α was not related to astrocyte viability, which confirms that they were dysfunctional indeed. Since atrophic astrocytes have been proposed to be dysfunctional, the impairment of glutamate uptake triggered by TNF-α intertwines both events in the context of astrocytopathy (Pekny & Pekna, 2016). Astrogliosis is considered a complex response, which varies from hypertrophy, atrophy, alteration in astrocytes biomarkers and dysfunction. Our findings show that human iPSC-derived astrocytes can be functionally activated, mimicking previously reported physiological aspects of astrogliosis The inflammatory processes share similarities and differences between animal and human models, causing intense debate over the theme (Cauwels, Vandendriessche, & Brouckaert, 2013;Warren et al., 2015). For example, genomic responses to inflammatory challenges are divergent between both species, although the same dataset has been subject of contradictory interpretations in two different studies (Kodamullil et al., 2017;Mestas & Hughes, 2004;Seok et al., 2013;Takao & Miyakawa, 2015). Besides, signaling pathways of neuroinflammation share similarities between rodents and humans, but substantial differences were found for molecular and cellular interactions (Kodamullil et al., 2017).  Tables   Table 1 -Cell lines used in the study. Human neural stem cells (NSC) were differentiated from induced pluripotent stem cells (iPSC) and they were exposed to astrocyte induction medium    and ΔCt (c) since the basal levels of TNF-α were below the limit of quantification in non-stimulated cells. Experiments were performed in triplicates from 4 cell lines. *P < 0.05; ****P < 0.0001; N.D. non detected.