Amyotrophic lateral sclerosis (ALS) is the most common and devastating motor neuron (MN) disease. Its initial cause and the mechanisms of further progression are still enigmatic nowadays. The cerebrospinal fluid (CSF) seems to play a major role within the deathly cascade, which is particularly true concerning pathological protein aggregates and their spreading across the central nervous system [1
Several studies have demonstrated diverse pathological effects of ALS-CSF in different in vivo and in vitro models [3
]. These models mainly include primary murine cortical or spinal MN cultures and spinal cord neuroblastoma hybrid cell line NSC34 cultures. A multitude of phenotypes could be observed by this including aggregation of transactive response DNA binding protein 43 kDA (TDP-43) in the case of ALS-frontotemporal dementia (FTLD)-CSF [8
], neurofilament abnormalities [9
], gliosis [10
], endoplasmic reticulum (ER)-stress [6
], mitochondrial dysfunction [11
] and Golgi fragmentation [6
]. In addition, intrathecal infusion of ALS-CSF in an in vivo rat model mimicked neuropathological changes similar to the ones found in postmortem tissue of ALS patients [4
]. Furthermore, interesting is the fact that not only CSF but also extracellular mutant superoxide dismutase 1 (SOD1) itself was able to induce Golgi fragmentation and the inhibition of ER-Golgi trafficking [13
]. The term Golgi fragmentation describes a loss of the intactness of this cisternal structure and its dispersion across the cell [14
]. However, nothing is known about how ALS-CSF affects human patient-derived induced pluripotent stem cell (iPSC)-derived MNs, currently believed to be the gold standard for in vitro disease modeling of ALS. Additionally, effects on typical ALS aggregating proteins are barely reported.
Golgi fragmentation in MNs themselves has been reported in different in vitro and in vivo ALS models [16
]; reviewed by Sundaramoorthy and colleagues [15
]. It is also present in postmortem sections of ALS patients [18
]. Moreover, Golgi fragmentation is reported in several other neurodegenerative diseases like Alzheimer’s disease and Parkinson’s disease [25
]. There is convincing evidence that Golgi fragmentation is an early event in the disease process [23
] and not only a consequence of apoptosis [30
] but rather inducing it [26
]. Thus, similar to other neurodegenerative disease conditions [31
], Golgi fragmentation in ALS could be an event that takes place before seeding of the aggregates, axon degeneration and manifestation of clinical symptoms occurs [30
The main goal of our study was to establish a new model system for sporadic ALS by investigating the ability of CSF derived from sporadic ALS patients to induce motor neuronal degeneration in the current gold standard cell culture model of ALS, namely patient specific iPSC-derived MNs.
There is evidence that CSF from ALS patients can directly harm neurons and induce changes associated with ALS pathophysiology. However, data does not exist about such effects on human patient-derived MNs, nor have the effects on modern molecular markers of ALS been well investigated. By using iPSC-derived MNs from healthy controls and monogenetic forms of ALS we demonstrate a harmful effect of CSF from patients suffering from sporadic ALS on healthy donor-derived human MNs. Golgi fragmentation—a typical finding in lower organism ALS models and human post mortem tissue of ALS patients—was induced solely by the addition of ALS- but not control-CSF. Strikingly, these changes occurred predominantly in MNs, the cell type primarily affected in ALS [39
]. Thus, the presented system might serve as valuable human MN model system for sporadic ALS (Figure 6
). Interestingly, this was not obvious in SOD1 p.R115G mutant MNs most likely due to already slightly increased fragmentation rates in the SOD1 mutant MNs. It might also point towards genotype specific effects, which, however, needs to be proven in future systematic studies across a variety of ALS genes and different mutations.
In our study, we did not observe significant neuronal network degeneration, MN-loss or protein aggregates containing SOD1, FUS or TDP-43 (Figure 3
and Figure 4
. Indeed, there is a large debate as to whether protein aggregations can be detected in iPSC-derived motor neurons [37
], but there is at least clear evidence that it is found in FUS-ALS models [34
]. Thus, even though it might not be surprising that we could not induce any kind of cytoplasmic aggregation of TDP-43 and/or SOD1, it was also not visible in the FUS-ALS model. Another explanation might be the protocol used in the study, which is known to induce structural neurodegeneration only at later stages [35
]. Since Golgi fragmentation is expected to be an early event in the pathophysiological cascade [23
], it is plausible that it occurs prior to further changes, which could develop either later or under additional stress conditions [34
]. Furthermore, ALS-CSF might not be “toxic enough” to induce the complete spectrum of pathological changes seen in postmortem tissue, which is particularly true for inducing pathological aggregates via CSF [8
]. Concentration of disease relevant proteins and/or aggregates in patient-derived CSF might be too low to act as agent to induce further aggregation (seeding). This would fit with a recent study by Ding and colleagues using CSF from ALS and additionally ALS-FTLD patients in which only ALS-FTLD-CSF was able to induce TDP-43 aggregates in target cells. It is well known that ALS-FTLD patients have a much more widespread TDP-43 pathology compared to ALS patients without co-occurrence of FTLD [42
]. The main differences from our study were the use of a glioblastoma cell line (U251) with obvious differences in general cell metabolism and a longer incubation time with CSF (21 days vs. 6 days). Of note, we additionally did not observe pathological protein mislocalization or aggregate formation when treating monogenetic ALS-patient-derived MNs with ALS-CSF (Figure 4
), which are known to have a higher propensity to serve as seeding acceptors [43
]. Furthermore, the seeding ability of the mutant protein also differs significantly between different mutations [1
], which might explain why we did not find seeding events in patient-derived MNs. Since we modelled spinal MNs, we cannot exclude that this might differ in cortical neuronal subtypes. Finally, the employed methods might not have been able to visualize smaller aggregates or oligomers of misfolded proteins. However, we intended to analyze the samples as similar as possible to human postmortem sections in which cytoplasmic protein localization and aggregation is the hallmark pathology.
We could not observe a significant reduction of MNs or neurons in general by exposition to ALS-CSF compared to control-CSF, when healthy donor-derived MNs were treated. Several reasons could explain this difference from results from murine primary cortical/spinal motor neuron cultures [3
], including human vs. murine cell culture system, medium composition or preanalytic CSF preparation. Another major concern regards the cellular age. Because iPSC-derived neurons do resemble fetal neurons and are not as old as the ones normally affected in ALS patients. Even though all primary neuron cell culture models are derived from fetal brains as well—and thus this cannot be the sole reason for the differences—human development takes much longer, and hence neuronal maturation affects phenotype appearance more severely especially in stem cell derived models [34
]. The reduction of neurons and MNs by exposition to control CSF with no differences to ALS-CSF (Figure 2
a,b) indicated a disease-independent effect on human iPSC-derived neurons. This finding can likely be caused by a reduction of specific medium components in case of CSF-concentrations above 10%. Interestingly, this seems to be similar in adult human neural precursor cells in which CSF promotes stemness [48
]. Our data point more towards a decreased relative amount of MNs and neurons compared to total cells due to a proliferative effect on NPCs rather than neuronal cell death.
Our study has some limitations. First, the amount, duration and time point of CSF treatment was limited due to the significant proliferation of NPCs induced by CSF and thus an overgrowth after longer incubation times. This effect was independent of the CSF-type. One additional limitation of our study is the use of pooled CSF. However, this was necessary due to the long time culture conditions and the protocol for human iPSC-derived MNs with the need for medium change every other day (Figure 1
). Thus, we cannot answer the question as to whether single ALS patient-derived CSF might vary depending on clinical parameters such as ALS subtype, disease stage or even family background [47
]. However, the study rationale was to investigate whether CSF from sporadic ALS patients could induce an MN-like degeneration in the dish, and thus we intended to analyze group differences rather than ALS subtype or individual patient’s specific differences. Furthermore, the model system presented represents one aspect of ALS (namely a model for spinal cord degeneration), and thus there might be differences when analyzing other neuronal cell populations affected, e.g., cortical neurons. The control group were not healthy controls but consisted of patients suffering from non-inflammatory, non-neurodegenerative neurological disorders (e.g., headache), who underwent CSF puncture during clinical routine conditions. However, CSF samples depicting signs of inflammation and/or classical neuropathological markers were excluded (see Table 1
). Additionally, we used the marker SMI-32 for the staining of MNs. However, as described in the methods section, following this particular protocol for MN differentiation, all SMI-32 stained MNs were additionally positive for specific motor neuronal markers including Hb9 (Homeobox gene 9), Cholinacetyltransferase (ChAt) and ISLET1 (ISL LH4 homeobox 1) [34
]. We clearly show, for the first time, that sporadic ALS-patient-derived CSF is sufficient to induce typical signs of neurodegeneration (Golgi fragmentation) and that human iPSC-derived motor neurons, the predominantly affected cell type in ALS, seem to be specifically vulnerable.
4. Materials and Methods
4.1. Patient Material
We included patient material for (i) CSF derivation and (ii) cell derivation.
Add (i): CSF was taken from 11 ALS patients and 8 control subjects (who suffered from non-inflammatory, non-neurodegenerative neurological disorders) under clinical routine conditions and independent of this study. The use of CSF was approved by the local Ethics committee (EK393122012, EK 49022016). CSF was immediately centrifuged (2100 rpm at room temperature for 10 min) and supernatant was stored at −80 °C until use. We included patients with definite, probable or possible amyotrophic lateral sclerosis, according to the revised El Escorial criteria [49
]. Patients suffering from genetically proven spinal muscular atrophy, spinal bulbar muscular atrophy, genetic ALS and FTD overlap syndromes were excluded. Details on patients and CSF characteristics are shown in Table 1
Add (ii): We included cell lines carrying FUS (p.R521C), TDP-43 (p.S393L), SOD1 (p.R115G) mutations and systematically compared them to four control iPSC lines from healthy volunteers (Table 2
). All procedures were in accordance with the Helsinki convention and approved by the local Ethical Committee of the Technische Universität Dresden, Germany (EK45022009, approved 21 April 2009; EK393122012 approved 11 March 2013). Patients as well as controls gave their written informed consent prior to skin biopsy.
4.2. Generation and Expansion of iPSCs, in Vitro Differentiation of Embryoid Bodies, AP Staining and Immunofluorescence on iPSC Colonies and Derivation of iPSC-Derived Neuroprecursor Cells
These procedures were performed as previously described by Nauman et al., Japtok et al., and Kreiter et al. [34
4.3. Motor Neuron Differentiation
The basal medium (N2B27) for the final MN differentiation consisted of 48.75% DMEM/F12 (Invitrogen, Carlsbad, CA, USA), 48.75% Neurobasal medium (Invitrogen, Carlsbad, CA, USA), 1% penicillin/streptomycin/glutamine (Thermo Fisher Scientific, Waltham, MA, USA), 1% B27 supplement without vitamin A (Invitrogen, Carlsbad, CA, USA) and 0.5% N2 supplement (Invitrogen, Carlsbad, CA, USA). NPCs were taken from the liquid-nitrogen-storage (−196 °C), then seeded and expanded using N2B27 medium containing additional 150 µM ascorbic acid (Sigma Aldrich, St. Louis, MO, USA), 3 µM CHIR99021 (Cayman chemical company, Ann Arbor, MI, USA) and 0.5 µM purmorphamin (Cayman chemical company, Ann Arbor, MI, USA). After at least two more passages during NPC expansion, patterning was started after reseeding, using N2B27 medium with 1 µM purmorphamin and 1 µM retinoic acid (Sigma Aldrich, St. Louis, MO, USA). After 9 days cells were re-seeded with a density of 50,000 cells per well (26.316 cells/cm²) on PLO/laminin-coated coverslips and differentiation was started using N2B27 medium additionally containing 10 ng/mL rhBDNF (Promega, Madison, WI, USA), 500 µM dbcAMP (Sigma Aldrich, St. Louis, MO, USA) and 10 ng/ mL rhGDNF (Sigma Aldrich, St. Louis, MO, USA). Following this protocol all SMI-32+ cells correspond to MNs and were positively co-stained for Hb9 (Homeobox gene 9), Cholinacetyltransferase (ChAt) and ISLET1 (ISL LH4 homeobox 1) [35
]. All further analyses were performed according to this protocol.
4.4. CSF Treatment (See also Figure 1a)
Since CSF caused a significant proliferation of NPCs in our cell cultures (Figure 1
), we had to find a suitable protocol for our experiments. We therefore analyzed the optimal time point of CSF application. We tested different options (immediately after seeding, 24 h, 48 h and 10 days later) using different concentrations (10%, 20%, 50%) of test-CSF. We had the best results initiating the treatment at day 10 after seeding for differentiation, using CSF concentrations of 10% and 20%. We then evaluated the duration of CSF-treatment (24 h, 48 h, 72 h, 96 h, 6 d) and the optimal concentration (10%, 20%, 50%). For this part of our experiment we started using ALS- and control-CSF. This evaluation led to the decision of treating with 20% CSF for 6 days, the highest possible concentration with no extensive NPC proliferation. For the final experiments, CSF from 11 ALS patients and 8 control patients was mixed in equal parts. CSF treatment was started 10 days after the reseeding, blending the differentiation-medium with CSF in the proportion 1:5. Cells were either treated with 20% ALS-, 20% control-, or no CSF. Since the number of seeded cells was constant throughout all experiments, the ratio of CSF to the initial amount of cells was the same in all experiments. Medium was changed every second day throughout the whole protocol. After 6 days of treatment, cells were fixated for immunofluorescence.
4.5. Immunofluorescence of Spinal Motor Neurons
For immunofluorescence staining, cells were washed twice with phosphate-buffered-saline (PBS) without Ca2+/Mg2+ (LifeTechnologies, Carlsbad, CA, USA) and fixed with 4% paraformaldehyde in PBS for 10 min at room temperature. Paraformaldehyde was aspirated and cells were washed three times with PBS. Fixed cells were first permeabilized for 10 min in 0.2% Triton X solution and subsequently incubated for 1 h at room temperature in blocking solution (1% bovine serum albumin, 5% donkey serum, 0.3M glycine (Carl Roth GmBH & Co. KG, Germany) and 0.02% Triton X-100 (Thermo Fisher Scientific, Waltham, MA, USA) in PBS). Following blocking, primary antibodies were diluted in blocking solution and cells were incubated with primary antibody solution for 12 h at 4 °C. The following primary antibodies were used: chicken anti-SMI-32 (1:10.000, Covance, Princeton, NJ, USA), mouse anti-FUS (1:5000, Sigma Aldrich, St. Louis, MO, USA), rabbit anti-beta-III-Tubulin (1:3000, Covance, Princeton, NJ, USA), mouse anti-GM-130 (1:200, BD Pharmingen, Franklin Lakes, NJ, USA), mouse anti-SOD1 (1:100, Cell Applications Inc., San Diego, CA, USA), rabbit anti-TDP-43 (1:400, Abcam, Cambridge, UK), rabbit anti-GFAP (1:1000 Chemicon, Temecula, CA, USA), mouse anti-vimentin (1:100, Sigma Aldrich, St. Louis, MO, USA). The nuclei were counterstained using Hoechst 33342 (LifeTechnologies, Carlsbad, CA, USA).
4.6. Quantification and Statistics
After immunofluorescence staining, cells were counted, and relative amounts were calculated. We took at least 15 pictures of each condition at random fields of view across the respective well. This resulted in from 350 to up to 3000 cells counted in each condition for one individual experiment, which we always performed at least three times. Furthermore, the morphology of the Golgi apparatuses was analyzed using a Zeiss Axio Observer Z1 microscope (Carl Zeiss AG, Oberkochen, Germany) and classified in different categories (Figure 5
). One category was defined as Golgi fragmentation (loss of the intactness of the cisternal structure and dispersion of the vesicular compartments across the cell; see Figure 5
) and relations were calculated using this number. We analyzed the Golgi apparatus of at least 140 neurons per condition of each individual experiment, which we again always performed at least three times. For the analysis of aggregates containing disease relevant protein such as FUS, TDP-43 and SOD1, we examined at least 200 cells per condition of each experiment, which we performed at least four times. Statistical analyses were done using either one-way ANOVA with Tukey post-hoc test, independent samples t-test, or a two-way ANOVA with Bonferroni post-hoc test using GraphPad prism 7 (Version 7.03, GraphPad Software, San Dieago, CA, USA). In case of non-normal distributed data, analyses were performed using Kruskal-Wallis-test with Dunn-Bonferroni post-hoc test. Analysis of normal distribution was performed using Shapiro-Wilk test. Data are depicted as mean ± SEM from at least three independent experiments each (independent differentiation batches).
4.7. Analysis of Neuronal Network Degeneration
We performed the analysis as previously described by Glaß et al. [50
]. After imaging a neuron or neuronal network, we converted the picture into a 2-bit image using the “threshold” function from Fiji (ImageJ-win64, open source). Following this, we transformed the obtained mask into a skeletonized network using the skeletonize plugin. A full intact axon without any disruption will yield a ratio of perimeter/section-length of 2. A disrupted axon will have an increased perimeter/section-length of up to 4, when the axon is solely represented by alternating pixel/blank spaces. We then rescaled the value range from 2–4 to 0–1. We normalized the data within each experiment to control-no-CSF condition. For the analysis we took at least 15 pictures at random spots across the respective well of each condition of one individual experiment, which we always performed at least three times.