Protocol adaptation and optimization for LUHMES differentiation in 3D
Differentiation of LUHMES cells in monolayer culture (Fig. 1a, b, 2D diff protocol) was well established and extensively characterized earlier (Scholz et al. 2011; Schildknecht et al. 2013), and the described changes in morphology, proliferation (Ki-67), and differentiation (Map2 and NF200) have been reproduced here (Supplementary Figure S1a). However, LUHMES cells differentiated in monolayer have limited life span. With increasing culture age, the interaction of the large neurite network with the extracellular matrix (plate coating) weakens, and the network either contracts into ganglion-like structures or fully detaches from the plate (Fig. 1b, last panel, d9). The brief survival of the differentiated cultures in 2D (which allows acute toxicity studies) is an obstacle for long-term, low-dose toxicological studies, as well as for cellular adaptation and resilience studies after toxicological stress. Therefore, we modified and adapted the LUHMES differentiation protocol for 3D. The 3D LUHMES model was prepared using the gyratory shaking technique as established for 3D rat primary aggregating brain cell cultures (Honegger and Monnet-Tschudi 2001; van Vliet et al. 2008) and iPSC microphysiological systems (Hogberg et al. 2013) with few modifications (Fig. 1a, c, 3D diff protocol).
First, the size of aggregates was monitored during differentiation. By adjusting initial cell number and shaker speed, we were able to control aggregate size (Fig. 1e, 3D diff). The cultivation of the aggregates from day 0 of differentiation under a constant shaking speed of 80 rpm allowed us to keep the size of aggregates within 300–425 μm in diameter through 21 days of differentiation, while under a gradually increasing speed from 68 to 80 rpm during the first 5 days, as originally established for rat primary cultures (Honegger and Monnet-Tschudi 2001), LUHMES aggregates reached 700 μm in diameter (Fig. 1e, 3D gradient).
Second, in order to test sufficient oxygen and nutrient supply, early apoptosis (Annexin V and caspase 3/7-positive cells) and necrosis (7-AAD-positive cells) were monitored by flow cytometry and fluorescence microscopy. Although a low percentage of caspase 3/7-positive cells were detectable on day 21 of differentiation (Fig. 2a), caspase 3/7-positive cells were distributed equally throughout the aggregates without any visible accumulation in the middle of the aggregates. No increase in Annexin V-positive cells was observed during the 21 days of differentiation in 3D (Fig. 2b). The percentage of Annexin V- and 7AAD-positive cells in the 3D cultures was comparable to those in monolayer undifferentiated LUHMES cultures, which were subjected to the same preparation procedure—both cultures were trypsinized for 30 min prior to Annexin V/7-AAD staining.
Third, we investigated compound penetration by staining the live, 12-day-old aggregates with DNA-binding blue fluorescent dye, Hoechst 33342 trihydrochloride, MW = 616 g/mol (Invitrogen) for 5, 15, 30, 60 min, and 6 h. For this experiment, LUHMES ubiquitously expressing GFP were used (Schildknecht et al. 2013). Hoechst 33342 dye penetration throughout the aggregates was advancing with increasing incubation time. Hoechst 33342 reached the middle of the aggregates after 1 h of treatment (Fig. 2c, Supplementary Figure S1b). This experiment ensured sufficient penetration of necessary small molecule factors for differentiation and nutrients, as well as toxicants. In addition, no visible apoptotic nuclear fragmentation accumulated in the middle of aggregates (Fig. 2c, 1 and 6 h). Thus, 3D cultures could be kept at least twice as long in culture than their 2D counterparts.
Finally, withdrawal of FGF and the addition of tetracycline, cAMP, and GDNF should rapidly induce exit from the cell cycle and differentiation to postmitotic mature neurons. We suggest that the observed continuous increase in the size of aggregates during differentiation could be due to a prolonged proliferation in 3D differentiating cultures. Higher cell density and increased cell-to-cell interactions may stimulate signaling between the cells within the aggregates that impedes exit from the cell cycle. Therefore, we quantified the expression of Ki-67, a proliferation marker, in undifferentiated cells as well as in 2D and 3D cultures. As expected, undifferentiated LUHMES were 98 ± 2 %-positive for Ki-67. Induction of differentiation in 2D reduced the expression of Ki-67 to 16 % by day 6, while in 3D cultures 49 ± 13 % of the cells were still Ki-67-positive on day 6 and 47 ± 12 % on day 12 (Fig. 3a, Supplementary Figure S1c). Therefore, we optimized the 3D diff protocol further to accelerate the exit from cell cycle in 3D and induce homogeneous differentiation. In the first step, we evaluated whether pre-differentiation in 2D for 48 h before 3D differentiation (Fig. 1a, 3D pre-diff protocol) would decrease proliferation. No differences in the size of aggregates (Fig. 1e, 3D pre-diff), as well as no change in Ki-67 expression (data not shown), were observed, compared to the 3D diff protocol; this protocol, therefore, was not followed further. In the second step, we tested whether increasing the tetracycline concentration would reduce the proliferation rate. LUHMES were differentiated according to the 3D diff protocol in the presence of 2, 4, and 10 μg/ml tetracycline. Although the highest tetracycline concentration reduced the proportion of proliferating Ki-67 cells (Supplementary Figure S1d), it appeared to be cytotoxic for the cultures (observation based on aggregate morphology, data not shown). Next, we applied treatment with the mitotic inhibitor taxol (also known as paclitaxel). The supplementation of neural differentiation media with anti-proliferation drugs, such as cytosine arabinofuranoside (AraC), is common and broadly used in primary neuronal cultures to block the proliferation of neuroprogenitors and astroglia without affecting postmitotic neurons (Gerhardt et al. 2001; Volbracht et al. 2006) After optimization experiments, 10 nM taxol for 48 h, from days 3 to 5 of differentiation, was chosen as a treatment scheme (Fig. 1a, 3D + T10 protocol). Treatment with taxol led to a reduction in aggregate size (250–300 μm on average, Fig. 1d, e, 3D + T10) and significant decreased in Ki-67-positive cells to 6 ± 6 % on day 6 and 2 ± 2 % on day 12 of differentiation (Fig. 3a, Supplementary Figure S1c). In addition, we analyzed the expression of the Ki-67 gene prior to and six and 12 days after induction of differentiation following either 3D diff or 3D + T10 protocols by real-time RT-PCR (Fig. 3b), which confirmed our flow cytometry data. The effects of taxol on Ki-67-positive cells were confirmed morphologically by immunocytochemistry, where whole aggregates were fixed at different stages of differentiation and stained with antibodies against Ki-67 and the postmitotic neuronal marker, NeuN (Fig. 3c). Fewer Ki-67-positive cells and higher number of NeuN-positive cells were found in 3D + T10 samples on days 6 and 12 of differentiation in comparison with 3D diff samples. Supplementation of LUHMES differentiation medium with taxol for 48 h selectively blocked proliferation without any negative effects on neuronal cells, increased the homogeneity of the cell population, and did not interfere with further toxicological studies since taxol was washed from the cultures before toxicant exposures. Thus, we favored the 3D + T10 protocol over other differentiation conditions, and this protocol was followed as a standard differentiation protocol for further experiments.
Characterization of LUHMES differentiation in 3D
The differentiation in 3D was characterized by immunocytochemistry. In addition to Ki-67 and NeuN stainings (described above), aggregates were stained with further neuronal markers (MAP2 for dendrites and apical part of axons, neurofilament (NF200) for axons, and synaptophysin for synapses) at different stages of differentiation with (3D + T10) and without (3D diff) taxol treatment. Induction of differentiation in 3D induced the expression of MAP2, NeuN, and synaptophysin; reduced the expression of Ki-67; and changed the morphology of neurofilament- and MAP2-positive neurites (Figs. 3c, 4). Interestingly, treatment with taxol not only inhibited proliferation, but significantly enhanced maturation, dendritic morphogenesis, and arborization as shown for MAP2 and synaptophysin stainings (Figs. 4, 5a). For more detailed visualization of long neurites protruding from the differentiated neurons, high magnification of MAP2/NF200 and synaptophysin/NF200 stainings of taxol-treated aggregates is shown (Fig. 5b). These findings are in agreement with publications showing that low taxol concentrations promote lamellipodial protrusions, stabilize microtubules, and increase spine formation (Buck and Zheng 2002; Gu et al. 2008).
Real-time PCR was performed to analyze induction of neuronal genes during differentiation of LUHMES in 3D. Expression of general neuronal markers (β-III-tubulin, NeuN, synapsin1), marker genes specific for dopaminergic neurons [tyrosine hydroxylase (TH), dopamine transporter (DAT), and vesicular monoamine transporter member 2 (VMAT2)], as well as proliferation and neural precursor markers Ki-67 and Nestin, were analyzed in course of 3D + T10 differentiation and normalized to the expression levels at day 0 (Fig. 5c). Ki-67 and Nestin were down-regulated during differentiation, while expression of neuronal markers was significantly induced (one-way ANOVA test followed by Dunnett’s post hoc test). The expression levels of these marker genes were similar to those in 2D differentiated LUHMES (Fig. 5d) with slightly higher expression of TH in 3D cultures versus 2D. Note that the expression level of all genes plateaued on day 6 of differentiation, suggesting complete differentiation.
LUHMES 3D model for neurotoxicity testing
Next, we analyzed the performance of the 3D model for neurotoxicity testing by applying two well-known neurotoxicants, MPP+ and rotenone. Both chemicals are mitochondrial complex I inhibitors and cause Parkinsonism (Betarbet et al. 2000; Franco-Iborra et al. 2015). MPP+ is specific for dopaminergic neurons, because of its selective uptake by them (Langston et al. 1984), while rotenone has broader toxicity. LUHMES neuronal aggregates were treated with increasing concentrations of both compounds for 24 and 48 h. First, cell viability assay, based on mitochondria metabolic capacity, was performed to generate concentration–response curves (Fig. 6a, b). Second, a cytotoxicity assay, based on the measurement of membrane integrity, was conducted in the same samples using LDH release assay (Supplementary Figure S2). As expected, mitochondria impairment was measured at concentrations at which the cellular membrane was still intact (low LDH activity in the media), confirming the mitochondria selectivity of the test compounds by the higher sensitivity of the resazurin reduction-based assay. The concentrations (5 μM MPP+ and 0.1 μM rotenone) with slight mitochondria impairment after 24- and 48-h exposure were used for further gene expression and washout experiments.
As a proof of concept, compound washout experiments were performed to address counter-regulation responses after short-term exposure in comparison with long-term chronic exposure. LUHMES were differentiated in 3D and exposed to 0.1 μM rotenone and 5 μM MPP+ for 12, 24, 48, or 192 h. On day 8 of differentiation, after 12, 24, and 48 h of exposure, compounds were washed out, and aggregates seeded into new plates and cultivated further until day 15. In case of 192-h exposure, aggregates were exposed to the toxicants continuously from day 7 until day 15. Cell viability (resazurin reduction assay) was assessed for all exposure conditions on day 15 of differentiation (Fig. 6c). Continuous exposure (192 h) to 5 μM MPP+ was 100 % toxic for LUHMES, while exposure to 0.1 μM rotenone for 192 h reduced cell viability by 70 %. Interestingly, after MPP+ wash out, around 80 % of cells were lost by day 15. This suggested either that MPP+ accumulates in the aggregates and continues to affect mitochondria after wash out, or that processes initiated by 5 μM MPP+ cannot be reversed, and cells cannot recover from the primary hit (at least at this concentration) (Fig. 6c dark gray line).
The washout effect was different for varying durations of 0.1 μM rotenone exposures (Fig. 6c light gray line). Exposure for 12 and 24 h further reduced viability by 34 % in total, while cells treated with 0.1 μM rotenone for 48 h were more strongly affected (47 % decrease in viability). However, although 5 μM MPP+ was less toxic than rotenone immediately after the hit (day 8), its withdrawal could not rescue cells from ongoing cell death, while in samples treated with 0.1 μM rotenone, cell viability continued to decline but to lesser extent than in MPP+ samples.
Since mitochondria are the primary target for rotenone, we further evaluated the effects of rotenone on mitochondrial membrane potential in individual aggregates. LUHMES were differentiated following 3D + T10 protocol, exposed to 0.05, 0.1, and 0.5, 1 and 10 μM rotenone or DMSO from d6 to d8 of differentiation. After 48-h exposure to rotenone, the aggregates were stained with the MitoTracker dye, to allow its accumulation in mitochondria according to the magnitude of their membrane potential. The mean fluorescence intensity values were then estimated in individual aggregates by fluorescence microscopy and normalized to DMSO controls (Fig. 6d, n ≥ 3, independent experiments with 10–20 aggregates assayed per experiment). Mitochondrial activity was significantly reduced in rotenone-treated samples. High correlation between data from resazurin and MitoTracker assays was observed for lower rotenone concentrations (0.05, 0.1, and 0.5 μM), which was not as closely related for the higher cytotoxic concentrations (1 and 10 μM), where changes in morphology and size of the aggregates prohibited precise microscopic evaluation of MitoTracker samples.
There are some limitations in imaging 3D cultures. Since 3D aggregates are ≥200 μm thick, imaging them using conventional fluorescence microscopy is very challenging due to issues with light scattering and penetration depth. Advanced confocal microscopy and/or two-photon microscopy in combination with optical clearing by treatment of the tissue with Scale clearing solution (Hama et al. 2011) prior to imaging overcome these limitations. Previously, it has been shown that rotenone perturbs neurite integrity in 2D LUHMES cultures (Schildknecht et al. 2013; Krug et al. 2013). To confirm these findings and to optimize the imaging of neurite integrity in 3D cultures, RFP-expressing LUHMES were used. Wild-type LUHMES were mixed with RFP-expressing LUHMES in the ratio 49:1 on day 0 of differentiation and differentiated following the 3D + T10 protocol. It was shown previously that RFP is only expressed in viable cells (Schildknecht et al. 2013). After rotenone treatment from days 6 to 8 of differentiation, RFP-expressing viable cells within the aggregates were imaged using confocal microscopy for neurite quantification. Exposure of LUHMES aggregates to 0.1 μM rotenone significantly affected the neurite integrity in comparison with DMSO controls (Fig. 7a, b). Thus, application of the fluorescent cell line mixed with wild-type cells helped to overcome the limitation of image quantification in these highly compact three-dimensional cultures.
Exposure of the 3D LUHMES model to rotenone and MPP+ alters the expression of genes involved in transsulfuration and one-carbon metabolic pathways
The performance of the 3D model for toxicological studies was analyzed by gene expression. We have chosen a panel of candidate genes which were shown to be involved in cellular adaptation to MPP+ exposure in 2D LUHMES cultures by regulating central carbon metabolism and amino acid turnover (ASS1, argininosuccinate synthase, SHMT2, serine hydroxymethyl transferase), transsulfuration pathway [CTH, cystathionase (cystathionine γ-lyase)], oxidative stress and DNA replication and repair [TYMS, thymidylate synthetase MLF1IP, centromere protein U (MLF interacting protein)] in earlier studies (Krug et al. 2014). LUHMES were differentiated following the 3D + T10 protocol and exposed to 0.1 μM rotenone and 5 μM MPP+ for 12 and 24 h on day 7 (Fig. 8a). In agreement with our earlier studies (Krug et al. 2014), on day 8—immediately after exposure—we observed the same regulation trends of those genes by rotenone and MPP+ [Fig. 8b, c, dark bars (rotenone), Supplementary Figure S3 (MPP+)]. Gene expression analysis was performed in three to four independent experiments (up to 12 technical replicates) and normalized to DMSO-treated controls. ASS1 was the most strongly up-regulated gene by MPP+ (FC = 3.3, 24 h) and rotenone (FC = 2.4, 24 h). ATF4, activating transcription factor four, was identified as upstream regulator of the cellular cascades initiated by MPP+ (Krug et al. 2014) but was less up-regulated in our 3D model by rotenone (FC = 1.5, 24 h), though 2.3 times increased by 24-h MPP+ treatment. CTH and SHMT2 were more up-regulated by MPP+ than by rotenone. MLF1IP and TYMS were significantly down-regulated in the 3D system following MPP+ and rotenone treatment. As proof-of-concept experiments—to study cellular counter-regulation—rotenone was washed out on day 8 of differentiation, and cells were kept in culture for further 7 days. Since washout of 5 μM MPP+ did not prevent cell death, we analyzed the expression of the same panel of genes on day 15 only in rotenone-treated samples (Fig. 8b, c, light bars). Interestingly, ASS1, CTH, and SHTM2, which were up-regulated immediately after exposure, were down-regulated 7 days later after rotenone withdrawal (Fig. 8b), while down-regulated genes (MLF1IP and TYMS) were further repressed with an even stronger effect (Fig. 8c). ATF4 was only slightly up-regulated on day 8 and returned to control level 7 days after recovery (Fig. 8b). This observation suggests that certain genes and signaling pathways are counter-regulated and/or may be responsible for cellular recovery after the primary hit, while other processes cannot be restored and, thus, might be a part of the new cellular homeostasis.
Altered expression of mir-7 miRNA after exposure of 3D LUHMES to rotenone
Finally, miRNAs involved in mitochondrial functions and relevant for PD were analyzed after exposure of LUHMES aggregates to 0.1 μM rotenone. In order to test whether miRNAs are involved in the recovery process, miRNA expression was assessed on day 8 as the reaction to the primary toxicant hit and on day 15, 7 days after rotenone withdrawal (refer to Fig. 8a for treatment and sampling scheme). In agreement with the literature showing down-regulation of mir-7 in PD models (Junn et al. 2009; Fragkouli and Doxakis 2014), we observed a reduction of miR-7 expression as early as 12 h after rotenone treatment (Fig. 8b, dark bar), while known pro-apoptotic miR-16 remained at control level (Fig. 8c). No changes were observed in expression of miR-210 (hypoxia-sensitive miRNA, involved in mitochondrial respiration (Chan et al. 2012, data not shown), suggesting mir-7 as a primary rotenone miRNA target prior to mitochondria-mediated apoptosis. On day 15 after rotenone washout, however, mir-7 expression went back to control levels, grouping this miRNA together with other counter-regulated genes (Fig. 8b), suggesting a possible role of this miRNA in cellular adaptation and recovery. In addition, brain-specific miRNA, mir-124, was unchanged on day 8 of differentiation and was upregulated on day 15 after washout (data not shown).