Development and Optimization of A High-Throughput 3D Rat Purkinje Neuron Culture to Study Paraneoplastic Cerebellar Degeneration

Improved understanding of the mechanisms involved in neurodegenerative disease has been hampered by the lack of robust cellular models that faithfully replicate in vivo features. Here, we present a rened protocol for generating age-dependent, well-developed and synaptically active rat Purkinje neurons in a 3D cell network culture which are responsive to a disease inducer. Using our model, we found that the application of autoantibody Yo, a paraneoplastic cerebellar degeneration (PCD) inducer, alters the structure of the dendritic arbour of cultured Purkinje neurons. The numbers of dendrites per branch-order, the branch-order in itself and the dendritic length were reduced by anti-Yo, proving a functional role for anti-Yo in the pathogenesis of PCD. Our new ex-vivo model is exible and can be used to investigate disease mechanisms that disturb Purkinje neuron function and communication in 3D. Since it is possible to use the approach in a multi-well format, this method also has high-throughput screening potential. both 0.5-1 µm intervals with the Zyla camera conguration (2048x2048) at the Andor Dragony microscope system using either a CFI Plan Apochromat Lambda S LWD 40x1.14 water objective (pixel size 151 nm), 60x1.20 oil objective (pixel size 103 nm) or CFI SR HP Apo TIRF 100x1.49 oil objective (pixel size 60 nm) to detect DAPI and CF TM 488/594/647 dye emission and superimposed with Fusion software (Oxford Instruments). 3D surface visualization of synapses was performed using Oxford Instruments analysis software IMARIS 9.3.1 and the lament tracer tool 22 Dendritic tree branch analysis. incubator to provide CO and balanced a sampling commercial to and

The rst question we addressed was which extracellular matrix is needed for maximal growth and survival. Initial attempts growing PNs directly on glass cover-slips coated with poly-D-lysine (PDL) and the extracellular matrix protein laminin failed. The yield of PNs per coverslip declined to zero from E18 to P10 at 21 days in vitro (DIV) (Figure 1a, non-3D-SCL). We reasoned that this was due to the lack of other cell types that provide the in vivo 3D-cell-network structure and cell-cell communication including paracrine factors. To overcome this, we developed a three-dimensional support cell layer (3D-SCL) approach by plating two cerebellar cell layers derived of either E18, P0 or P10 tissue onto PDL coated cover-slips. We also introduced a time-delay by plating the second cell layer 7 to 48 days later than the rst. We found that the tissue age of cells used to grow the 3D-SCL (E18 to P10) had no impact on the PN yield of the second layer, but there was a strong correlation between the in vivo age of the support cell layer and the tissue age of cells used to grow the second cell layer, the enriched PN layer. The highest survival rate of E18 derived-PNs was observed when plated onto the 3D-SCL at DIV14, for P0 derived-PNs at DIV21 and for P10 derived-PNs at DIV28 (Figure 1a). These ndings indicate that the older the starting tissue, the more mature the 3D-SCL has to be to achieve a high survival rate of PNs for a minimum of 21 to 28 DIV.
The use of a "double" cell layer was associated with higher metabolic demand than single layer cultures and led to non-physiological pH uctuations. This was associated with increased cell death when replacing half of the culture media once a week. By decreasing the interval und replacing the culture media either every 3.5-days (6 well) or every 2-days (12 and 24 well), we found this prevented pathological pH uctuations and gave a healthy well-developed neuronal network. Despite these improvements, PNs still had a poorly developed dendritic morphology compared to those in vivo, with fewer and shorter branches in E18 and P0 derived-PNs (Figure 1b-c, 1b upper panel).
During development, neuronal dendrites are generated by a series of processes: rst extension and retraction of dendritic branches, and subsequently stabilisation of existing dendrites through building of synaptic connections and neuronal calcium homeostasis 8 . Calciumdependent protein kinase C (PKC) subtypes, activated by synaptic inputs from parallel bres (granule cells) through metabotropic glutamate receptors (mGluR1/4), trigger functional changes as well as long-term anatomical maturation of the PN dendritic tree 9 . Altering the activity of calcium-dependent PKC subtypes using PKC antagonist K252a improved dendritic branching for E18 and P0 derived-PNs similar to in vivo, but had no effect on the branching characteristics of P10 derived-PNs (Figure 1b-c, 1b lower panel). Interestingly, PKC inhibition induced by K252a signi cantly improved cell survival rate observed for P0 and particularly for P10 derived-PNs in a concentration dependent manner ( Figure 1d). The survival rate in P0 derived-PNs was improved by a factor of 6 by blocking 20% of PKC activity (10 nM K252a), whereas in P10 derived-PNs, blocking PKC activity to 50% (25 nM K252a) increased the survival rate by a factor of 28. Inhibiting PKC activity had no effect on the survival rate of E18 derived-PNs ( Figure 1d).
Purkinje neuron survival and dendritic tree development are also highly dependent on paracrine factors such as progesterone, insulin and insulin-like growth factor 1 (IGF1) [10][11][12] . We therefore supplemented our cultures with 40 µM progesterone and found that this led to increasingly branched dendritic trees in E18 derived-PNs, but had no impact on the branch structure of P0 and P10 derived-PNs ( Figure 1e). Even though PN dendritic development was insu cient when either K252a inhibition or progesterone were not supplied, supplementation with insulin and IGF1 was su cient to maintain the long-term growth of the other cerebellar cell types including granule, Golgi, Lugaro, unipolar brush, stellate and basket cells (Figure 1f).
To prove that our PNs expressed functional synapses, we demonstrated the presence of pre-and postsynaptic biomarkers of functional synapses including voltage-gated calcium channels (VGCC), metabotropic glutamate receptor 1 (mGluR1), post-synaptic density protein 95 (PSD95), glutamate-decarboxylase 65 (GAD65), glycine transporter 2 (GlyT2), α-synuclein and bassoon using immunocytochemistry. All these markers were present indicating a level of maturity of both the PNs and the surrounding network ( Figure 1g).
Next, we elucidated the maturity of the these PNs by testing their functional activity. In vivo, PNs re spontaneous action potentials at frequencies of about 40-50 Hz with a complex trimodal pattern of tonic ring, bursting, and silent modes that depend on anatomical and functional maturity 13,14 . E18 derived-PNs cultured in a 24 well multielectrode array rst showed spontaneous bioelectrical activity on in vitro day 11. The spike rate increased constantly from 0.15 ± 0.03 Hz (DIV11) to 2.56 ± 0.59 Hz (DIV21). After DIV28, the spike activity become erratic with long periods of silence, but overall, a frequency of 2.79 ± 0.55 Hz was maintained until DIV63 ( Figure 1h). We observed both uniform and highly non-uniform spike intervals and trains with silent periods between bursts and spike frequencies of up-to 140 Hz within the burst. Exchanging the PNC media at DIV28 to one previously used in organotypic brain slice culture 15 , prevented the erratic spike activity and stabilized the spike frequency at 6.35 ± 1.85 Hz for up-to 63 DIV.
In addition to immunocytochemical and high-throughput electrophysiological studies, we found that this 3D PN model system was also useful for cell-type-speci c genetic engineering, for example, using lentiviral particles to express PN-speci c green uorescence protein (GFP) via implementation of the L7 promoter 16,17 . We applied L7-GFP inducing viral particles to dissociated PNs on the day of seeding and found PNs that express GFP with minimal off-target expression (<0.02%) after 3 days. At DIV14, 61.5% of the PN population were GFP positive and these cells did not differ in dendritic structure and stably expressed GFP for up-to 169 DIV ( Figure 1i). Using our culture system, we also found a su ciently high transfection rate of PNs when lentiviral particles were added to the culture at DIV14 and DIV28, however the rate of transfection and speed of expression fell progressively when genetic manipulation was performed later. The GFP positive PNs in the culture revealed a very similar development to in vivo, as we were able to observed the fusion phase (E17-P5), the phase of stellate cells with disoriented dendrites (P5-P7), as well as the phase of orientation and attering of the dendritic tree (P7-P21) 18,19 ( Figure 1i).
We present a 3D rat model for growing Purkinje neurons that is independent of derived tissue age, and which provides a complex and robust system with great experimental exibility. By combining a 3D network structures and optimized concentrations of hormones, paracrine factors and activity regulators (progesterone, insulin, IGF-1, K252a) provided at optimal time points, this system creates the ideal conditions to grow a balanced cerebellar network in miniature ( Figure 2a).
Furthermore, we show how this 3D rat PNC model responds to disease inducer such as the paraneoplastic autoantibody anti-Yo. Anti-Yo is associated with paraneoplastic cerebellar degeneration (PCD) a disease process showing marked Purkinje neuron death 20 . We found that anti-Yo antibodies signi cantly altered the structure of the dendritic arbour of PNs over time. Anti-Yo minimizes rst the numbers of dendritic branches by half for branch order 6 to 8 at 48h and 5 to 11 at 96h; second the branch-order in itself (Ctrl: 20; Yo 1 : 14, Yo 2 : 13) and third the dendritic length for branch order 4-8 at 48h (Ctrl 23.0 ± 2.6 µm, Yo 1 13.6 ± 1.1 µm, Yo 2 13.8 ± 1.2 µm) and 3-12 at 96h (Ctrl 22.0 ± 2.7 µm, Yo 1 13.1 ± 1.2 µm, Yo 2 12.1 ± 1.3 µm) (Figure 2b-c). These ndings correlate well with previous observations using organotypic cerebellar slice culture 15 , and con rm a functional role for anti-Yo in the pathogenesis of PCD. The long-term stability and neuronal complexity of our culture will facilitate further studies of cell-and network-dependent mechanisms of cerebellar degeneration related to PCD or other cerebellar disorders.

Material And Methods
Neuronal culture preparation. Protein kinase C inhibitor K252a (Alomone, # K-150; IC 50 25 nM). In long-term cultures that were maintained for more than 28 days in vitro the IGF1 and progesterone concentration were reduced to 10 ng/mL and 20 µM, respectively. K252a was supplemented for 21 days before the washout process started, its optimal concentration was experimental evaluated for each tested culture type. Half of the culture medium was replaced every 3.5 (6 well) and 2 (12/24 well) days, respectively. All experiments testing the Purkinje neuron yield dependent on derived tissue age, in vitro age of the 3D-SCL and K252a concentration were performed randomly, containing 3 to 6 probes per experimental setting and 5 independently repeats for each group and condition.
L7 promoter (full length 1005 bp) were custom cloned by SBI System Bioscience into construct pCDH-L7-MCS-copGFP (#CS970S-1) and viral particle with a yield of 2.24 x 10 9 ifus/mL were produced. Freshly prepared Purkinje neurons of E18 or P0 cerebellum suspended in growth media containing no serum were incubated for 10 minutes at 37°C with 1.22 x 10 6 viral particle/mL before seeded onto the supplement structure layer containing cover-slip or live cell imaging µ-dish (#80136, 35 mm, Ibidi). Media was changed after 3 days and transfection e ciency evaluated by live cell imaging microscopy 24h post transfection, daily up to 21 days and weekly up to 169 days in culture, respectively. Additional, lentiviral transfection of Purkinje neurons in culture were performed 1 day after feeding at DIV15 and DIV29 by applying 2.5 x 10 6 viral particle/mL to evaluate the e ciency and effects of age-dependent genetic manipulations. The neuronal development of the GFP expressing Purkinje neurons was followed by obtaining 10 independent 3x3 tile scan using the Zyla camera con guration (2048x2048) with the CFI Plan Apochromat Lambda dry objective 10x0.45 (pixel size 603 nm) or 20x0.75 (pixel size 301 nm) at the Andor Dragon y microscope system (Oxford Instruments company). The experiments of DIV0, DIV15 and DIV29 were repeated three times.
Immunohistochemical cell type characterisation.
Purkinje neuron count and imaging.
Purkinje neurons were counted manually and blind by screening the cover-slips using a Leitz Diaplan Fluorescence microscope equipped with CoolLED pE-300white. For dendritic tree branch analysis and determination of maturity and synaptic interaction, 10 Purkinje neuron Zstack images per cover-slide were collected in 5 independent and randomized experiments at 0.5-1 µm intervals with the Zyla camera con guration (2048x2048) at the Andor Dragon y microscope system using either a CFI Plan Apochromat Lambda S LWD 40x1.14 water Patient sera Sera were obtained from two untreated patients with gynecological cancer and PCD who had Yo antibodies against CDR2 and CDR2L (anti-Yo 1−2 ) but lacked P/Q-type VGCC antibodies 15 . A pool of sera from 100 healthy donors (non-hCDR 100p ) were used as controls. Sera were not heat-inactivated before use. The sera were stored at the PND Biobank #133/2015 or the Biobank for diagnostic cancer marker #188.05 with approval of the regional ethics committee, Western-Norway.

E x vivo PCD model
Twenty-eight days post-seeding, the culture medium was replaced with medium containing; human serum positive for Yo antibodies (anti-Yo; hCDR2/2L; 4 µL/mL). Purkinje neuron culture was collected 2 or 4 days after commencement of treatment to evaluate the antibody effects ( Figure. 2b-c). Each independent experiment included treatment (anti-Yo 1−2 ) and positive (non-hCDR 100p ) control to account for variations in cell survival between culture preparations. All treatments were performed in triplicate and 5 PNs were analyzed each.  Tables   Table 1 | Primary antibodies. The signal to noise ratio for the antibodies were evaluated for the following conditions: 4% PFA at pH 7.2 diluted in 100 mM PBS; 1.5% PFA at pH 6 diluted in 100mM natrium acetate buffer (NaAcB)); without heat-induced antigen retrieval (HIAGR);

Abbreviations
and with HIAGR either TRIS-based (pH 9) or citric acid-based (pH 6). The best conditions for each used antibody are described below.  synaptic: α-synuclein (α-syn) -marker of glutamatergic synaptic terminals from granule cells (parallel bres) and unipolar brush cells (type I/II); GAD65-marker of axon terminals from stellate and basket cells; bassoon -marker of the active zone of mossy bre terminals and parallel bre terminals between Golgi cells and granule cells, and between basket cells and Purkinje neurons; and synapsin I -synaptic vesicle phosphoprotein of mature CNS synapses; Nuclei staining DAPI (blue). Scale bar, 20 µm; (h) MEA recorded spike patterns (10s) with a cut-out (1s) at day 21 in vitro following Purkinje neuron maturity. (i) Live-cell imaging of E18 derived-Purkinje neuron expressing lentiviralinduced GFP from day of seeding (DIV0) up to 2 months (DIV53). The Purkinje neuron development to maturity was very similar to in vivo, as the fusion phase (E17 -P5 ≈ DIV0 -DIV7), the phase of stellate cells with disoriented dendrites (P5 -P7 ≈ DIV7 -DIV9), as well as the phase of orientation and attering of the dendritic tree (P7 -P21 ≈ DIV9 -DIV23) were observed. Scale bar, 50 µm Optimized 3D rat Purkinje neuron culture protocol. (a) Each tested culture desired different conditions of support and activity interdependent of the starting tissue age. Whereas the supplementation of insulin-like growth factor 1 (IGF1) and progesterone (PROG) induced a stable environment to obtain high survival rates of Purkinje neurons, PKC activity modulation mainly shaped the dendritic tree development, with the exception of P10 tissue derived neurons where the survival was highly dependent on the inhibition of PKC but not their dendritic tree development. The optimized protocol for all tested tissues relies on the time point of placing the second cell layer, the Purkinje neuron enriched layer, and media that is supplemented with IGF1, progesterone and K252a, where K252a starting concentration is altered dependent on the used tissue to start the culture as follow; DIV1-10: E18 -5 nM, P0 -10 nM, P10 -25 nM; DIV10-22: the K252a concentration is raised to 25 nM for E18 and P0 until the dendritic tree is well-developed and mature; DIV22-28: washout phase, K252a supplementation is stopped (DIV22-24: 12.5 nM, DIV24-26: 6.75 nM, DIV26-28: 3.35 nM). At DIV 28 the IGF1 and progesterone concentration is reduced by factor, 2.5 and 2, respectively, to proceed to long-term culture conditions. The developed protocol allows to grow a stable Purkinje neuron 3D culture for up