hiPSC lines used in this work were derived from fibroblasts obtained from healthy individuals after informed consent and were previously published (Fernandes et al., 2016; Kunze et al., 2018; Haenseler et al., 2017; Wang et al., 2017). Experiments followed international guidelines and local regulations. Details on the four hiPSC lines used in this study are summarized in the table shown in Figure 1B.

Derivation of the hiPSC lines SFC840-03-03 (840), SFC854-03-02 (854) and SFC856-03-04 (856), from fibroblasts of healthy donors was described previously (Haenseler et al., 2017; Fernandes et al., 2016). These hiPSC lines are also distributed by EBISC as STBCi026-A/RRID:CVCL_RB85 (hereafter referred to as “840”), STBCi066-A/RRID:CVCL_RC86 (hereafter referred to as “854”), and STBCi063-A/RRID:CVCL_RC81 (hereafter referred to as “856”). We maintained the 840, 854 and 856 hiPSCs in E8 medium on Geltrex-coated tissue culture treated vessels and passaged as small clusters of cells by splitting with 0.5 mM EDTA as described previously (Beers et al., 2012). We produced masterbatches of these hiPSCs, that we quality controlled, including SNP array for genome integrity, and started the differentiations of our experiments below five passages from thawing the hiPSCs.

We subcloned the previously described hiPSC line HMGU1 (alternative name ISFi001-A/RRID:CVCL_YT30) (Kunze et al., 2018) and worked with HMGU1 subclone 5. The hiPSC line HMGU1 subclone 5 was quality controlled and expanded on Geltrex-coated vessels by EDTA passage in different media depending on the differentiation protocol used: for cortical neuron differentiation in Essential8 (E8) medium (Thermo Fisher), for Gibco NSCs and cerebral organoids in StemFlex medium (Thermo Fisher) and for dopaminergic neurons in StemMACS iPSC Brew (Miltenyi Biotec).

We also worked with i3N hiPSC line NWTC11.G3-WT that has a doxycycline-inducible neurogenin 2 (NGN2) cassette stably inserted in the WTC11 hiPSC line (Wang et al., 2017), (hereafter referred to as “NGN2”). This line was maintained in E8 medium and quality controlled with a SNP array. The SNP arrays show no major chromosomal abnormalities in any of the lines (Supplementary Figures S1–S5).

The different differentiation protocols are summarized in the table shown in Figure 1C. Cortical neurons, cerebral organoids, dopaminergic neurons and i3 Neurons (NGN2) were differentiated following published protocols (Shi et al., 2012; Kriks et al., 2011; Lancaster and Knoblich, 2014; Fernandopulle et al., 2018). We also produced neuronal stem cells with the commercial PSC Neural Induction Medium (A1647801) from Gibco (“Gibco NSCs” hereafter). We further combined previously described methods to obtain uniform, mature low density neurons by applying a cocktail of proliferation inhibitors and maturation factors to cortical neurons from day 45 onwards (Kemp et al., 2016; Qi et al., 2017). All detailed differentiation protocols including all reagents used are provided in the Supplementary Material.

We applied standard fixation and immunofluorescent staining methods; all details for sample preparation and staining including antibodies and reagents used are provided in the Supplementary Material. Organoids were cryosectioned before staining (details in the Supplementary Material).

Microscopes used included an SP8 inverse confocal laser scanning microscope (Leica) [z-stacks with 0.3 µm step sizes with a ×63 oil objective (HC PL APO CS2, Leica)] or a BC43 spinning disc confocal (Andor) [0.3 µm step size, with a ×63 oil immersion objective (CFI Plan Apochromat Lambda D 60X Oil)].

For assessment of ciliation rate, we manually counted cilia on maximum intensity z-projected confocal images using the Fiji cell counter plugin. We counted nuclei, pericentrin (PCNT)+ basal bodies (centriole pairs were counted as one basal body) and ARL13B+ cilia (=ARL13B signal that protrudes from a PCTN+ basal body). We also counted INPP5E+, AC3+ and GPR161+ cilia (=INPP5E, AC3 or GPR161 signal that proceeds from a PCNT+ basal body). Given that in some conditions, it is difficult to attribute individual cilia to specific neurons/nuclei, we chose to quantify ciliation rate as the proportion of cilia over basal bodies. Ciliation rates are indicated as mean of a sample ± standard deviation, with n = number of samples unless stated otherwise, where one sample represents an individual cell culture well stained.

For measurement of ciliary length, we used the ImageJ plugin CiliaQ for 3D reconstruction of cilia and extracted ciliary length and orientation in z-direction from the reconstructed data in CiliaQ (Hansen et al., 2021). Image processing with CiliaQ was applied to high resolution images acquired with an SP8 confocal microscope in the ARL13B channel. Cilia length represents the length of the ARL13B signal. We manually checked through the reconstruction data and only kept the reconstructed ARL13B+ objects that had a PCNT+ basal body. Cilia length is indicated as mean ± standard deviation, with n = number of cilia.

Cerebral organoid cryosections stained for ciliary markers were imaged using a Leica SP8 confocal microscope (at 63x with 7.5 optical zoom). We took images every 0.2 μm in the z-plane. Z-stacks from at least three ventricles of three different organoids in three independent organoid differentiations were analyzed (= at least nine images). Images were deconvoluted using Huygens Software and subsequently analyzed using the ImageJ/Fiji plug-in CiliaQ. The channel for ciliary length measurement (ARL13B staining) was used for segmentation and subsequent length assessment. The segmentation method ReyniEntropy was applied, with a Gaussian blur of 1.5. Cilia touching X/Y/Z borders of the image were excluded.

For analysis of additional markers (INPP5E, AC3), we assessed ciliation rate of ARL13B-positive cilia and ciliation rate of the additional ciliary marker separately and from these numbers calculated the percentage of INPP5E-positive cilia over ARL13B-positive cilia respectively AC3-positive cilia over ARL13B-positive cilia.

GraphPad Prism9 was used for data visualization and statistical analysis. The statistical test applied is indicated in the figure legends and/or the main manuscript. In the graphs for ciliation rates we displayed all Fields of View (FOVs) as single datapoints to illustrate the observed variability, but for statistical analysis we used the mean of the FOVs acquired for each individual sample unless stated otherwise. One sample is an individual immunofluorescence staining (of one individual cell culture well).

Before starting differentiation into neuronal cells, we first wanted to assess primary cilia in human induced pluripotent stem cells (hiPSCs). These are cycling cells and actively dividing cells are expected to lack a cilium during the division process. To assess the ciliation rate of human hiPSCs we performed immunofluorescence using antibodies against the ciliary membrane-associated protein ARL13B and against the centriolar protein pericentrin (PCNT) in four different hiPSC lines derived from healthy individuals (Figure 1B and 2A). We found that ciliation rate, as determined by manual count of ARL13B-positive signal correlating with PCNT-positive signal, was highly variable between different hiPSC lines and between independent experiments, with an average ciliation rate of 34.43 ± 12.21% across all samples [n = 37 samples; ± standard deviation (SD)] (Figure 2B). Interestingly, ciliation rate was not influenced by the proliferative status since we found no significant differences between lines when assessing proliferation by quantifying cells in S-phase using EdU or M-phase based on Hoechst staining, despite different ciliation rates (Supplementary Figure S6). Cilia on hiPSCs were short with an average length of 1.677 ± 1.071 µm (n = 138 cilia; average cilia length ± SD) (Figure 2C). We further determined that cilia were aligned towards the cell culture medium and away from the plastic (Figure 2A and Supplementary Figure S7).

The dual SMAD inhibition NSCs from the four different lines were further propagated to cortical neurons to assess their ciliation rates at day 40, 56 and 70 of differentiation.

At day 40, cortical neurons from all four hiPSC lines were similarly ciliated with an average ciliation rate of 30.79 ± 13.47% (n = 30 samples) (Figures 4A,B). Cilia had an average length of 1.743 ± 1.080 µm (n = 241 cilia) and were mostly found on the soma of the neurons without a clear orientation in the z plane (Figure 4A and Supplementary Figure S7).

Upon further differentiation, the hiPSC lines 840 and 856 showed a significant reduction of the ciliation rate on day 56 and day 70, but the lines HMGU1 and 854 remained with an unchanged ciliation rate (Figure 4A,B). The length of cilia did not differ as much between lines or between day 40 and day 70, with an average length of 1.501 ± 0.9909 µm at day 70 (n = 454 cilia) (Figure 4C). Importantly, we observed variability in the strength of the ARL13B signal of cilia at day 70 of differentiation with this protocol: while some cilia continued to display a strong ARL13B signal, in many neurons the cilia were only weakly positive for this marker.

When comparing the neuronal cultures generated by the four lines, we noted that despite plating the neurons at the same density at day 35 (20,000 cells/cm²), the two lines with higher ciliation rate had much higher cell density. Indeed, at day 56 and day 70 the cultures appeared quite different in the neurons of the different cell lines: in lines HMGU1 and 854 (with stable higher ciliation), cell density was overall higher and the cells tended to cluster in large clumps, while in lines 840 and 856 (with decreasing ciliation) such clusters were not present (Figure 4D). This suggested that the lines yielding cultures with higher density at day 56 and 70 contained residual proliferating cells for a longer period than the lines with lower cellular density. We thus assessed proliferation using EdU and PAX6 to mark progenitor cells, demonstrating as expected a higher proportion of proliferating progenitors at day 40 in lines HMGU1 and 854 (Supplementary Figure S9). The higher ciliation rate in the two proliferating lines (HMGU1 and 854) could thus be explained by a higher proportion of NSCs in the culture and/or by higher density of the resulting neuronal cultures.

To evaluate cell identity of the resulting cultures at later timepoints, we next stained for MAP2 and TUBB3 at day 56 and day 70 which were highly positive in all four cultures, indicating neuronal identity in all four lines (Supplementary Figure S10). We further stained for the forebrain commitment marker FOXG1, which was positive in the majority of cells at day 56 in all four lines, indicating cortical identity (Supplementary Figure S10). Hence, despite persistence of proliferating cells in HMGU1 and 854, all four cultures yielded forebrain neurons at later stages. Interestingly, we observed that the distribution of cilia was particularly high in the regions with clustered neurons, suggesting that cell density plays a role for ciliation of neurons.

Taken together, we find that ciliation rate in these neuronal cultures varies strongly between cell lines, mirroring differences in overall behavior of the line: while starting at day 56 all lines contained cortical neurons, those lines with persistence of proliferating progenitors and subsequent higher culture density (HMGU1 and 854) had higher ciliation rates than the lines with lower culture density (840 and 856). However, ciliation rate was overall lower in more mature neuronal cultures than in NSCs for all lines.

Given this finding that ciliation was much higher on neurons of densely clustered regions compared to sparsely distributed neurons in the same cultures, we next evaluated whether neuronal density influences the ciliation rate. We plated cortical neurons at different densities and assessed ciliation rate at day 40. We found the lowest ciliation rate in cultures plated at the lowest density of 5,000 cells/cm² and significantly higher ciliation rates for densities of 40,000 cells/cm² and above, with a clear direct positive correlation between cell density and ciliation rate (Figure 5). We noted that at the highest neuronal density, accumulations (“clusters”) of cilia were found in all four cell lines.

We also tested if cell density influences the ciliation rate in hiPSC. For this we replated hiPSCs and assayed the samples at day 2 after replating, as in Figure 2, when hiPSCs were ∼50% confluent and had no overlapping nuclei, and at day 3 when hiPSC were over 90% confluent and cells were packed more densely so that nuclei started to overlap. Interestingly, higher hiPSC density led to divergent results depending on the hiPSC line used, with some lines displaying higher ciliation rates in more confluent cultures and others displaying lower ciliation with higher confluency, illustrating again the variability of ciliation rate in hiPSCs (Supplementary Figure S11).

Given that these in vitro culture protocols involved replating the maturing cells at given intervals, we also investigated if replating has an influence on the ciliation rate. In the dual SMAD inhibition protocol (Shi et al., 2012), cells are frozen at day 28, then thawed and replated after 7 days (day 35) for final neuronal maturation. For this experiment, we skipped the replating at day 35: we thawed the cells at day 28 directly to the imaging plate and compared these to standard neurons replated at day 35, both being fixed and stained at day 36. We could not find significant changes in ciliation rates between these two conditions (Supplementary Figure S12). We also checked if replating at a later timepoint affects the ciliation rate. For this we compared neurons replated on day 55 to neurons replated at day 35 (standard protocol), both fixed at day 56. Also here, we could not find significant changes in ciliation rate (Supplementary Figure S12).

Taken together, these results indicate that culture density directly impacts ciliation rate in neuronal culture, but not in iPSCs, while replating does not affect ciliation rate.

Given the variability observed with the dual SMAD protocol, where cells remain unguided in neural maintenance medium (NMM) (Figure 6A), with variable persistent proliferation, we decided to apply other differentiation protocols. We first modified the dual SMAD inhibition protocol by applying proliferation inhibitors and neuronal maturation factors (MAT) from day 45 of differentiation onwards to obtain more uniform mature neurons. Using this “MAT protocol,” we obtained similar densities of neurons with consistent high levels of MAP2-positive cells and Beta4 spectrin (SPTBN4)-positive axon initial segments at day 70 from all four hiPSC lines (Figure 6B and Supplementary Figure S13). In contrast to the previously applied differentiation in NMM, we no longer observed clusters of cells or of cilia in day 70 neurons in lines HMGU1 and 854, and the four lines produced very similar cultures. Ciliation rate was now significantly lower than in NMM neurons for all hiPSC lines with an average ciliation rate of 9.179 ± 5.587% (n = 15 samples) (Figure 6C). Overall, day 70 MAT neurons had only few ARL13B-positive cilia and mostly with a weak ARL13B signal (Figure 6B).

To compare with a different protocol generating excitatory glutamatergic neurons, we assessed ciliation rate in NGN2 neurons, 12 and 20 days after NGN2 induction (Figures 6D–F). At both timepoints we obtained TUBB3- and MAP2-positive neurons and we found SPTBN4-positive axon initial segments in day 20 NGN2 neurons (Supplementary Figures S13 and S14). Ciliation rate was 5.936 ± 7.177% (n = 5 samples) at day 12 and 19.58 ± 8.338% (n = 6 samples) at day 20 (Figure 6F). Similar to the MAT protocol, the NGN2 neurons only had few ARL13B-positive cilia mostly with weak ARL13B signal.

To evaluate cilia on another type of hiPSC-derived neuron, we differentiated the HMGU1 line to midbrain patterned dopaminergic neurons and assessed ciliation rate at day 24 and 35 of differentiation (Supplementary Figure S15). Overall ciliation rate, when taking all nuclei into account, was 42.86 ± 9.8% (n = 4 samples) at day 24 and significantly reduced to 13.96 ± 5.214% (n = 3 samples) at day 35. Given that these cultures are not composed purely of dopaminergic neurons, we next restricted cilia counts to TH+ dopaminergic neurons and found even lower ciliation rates with 18.94 ± 4.648% (n = 4 samples) at day 24 and 7.032 ± 6.1% (n = 3 samples) at day 35. Length of cilia did not significantly differ between day 24 with 2.301 ± 1.005 µm (n = 28 cilia) and day 35 with 2.210 ± 1.771 µm (n = 48 cilia) (Supplementary Figure S15).

In summary, ciliation rate was overall lower in neurons produced with a variety of protocols than in NSCs.

All neurons described so far were generated through 2D differentiation protocols. Since an increasing number of studies are applying 3D protocols, which are closer to reproducing the complex tissue architecture, we also analyzed cilia in undirected cerebral organoids (Lancaster and Knoblich, 2014). As described with this protocol, we observed the formation of “cortical units” in 45-day-old cerebral organoids with proliferating progenitors organized around a lumen in a “ventricular zone” (VZ) and mature MAP2-positive neurons away from this lumen. The majority of ARL13B-positive cilia observed were found on the progenitor cells and were projecting into the lumen. In addition, we found more sparsely distributed ARL13B-positive cilia at the basal side of the cortical units on MAP2-positive neurons away from the VZ (Figure 7A). The ciliation rate of neuronal progenitors within the VZ was 61.81 ± 25.04% (n = 3 organoids) and the average cilium length was 1.465 ± 0.7189 µm (n = 272 cilia) (Figure 7B–7D), findings which are consistent with those described above for NSCs generated with 2D protocols. Given that cerebral organoids generated with this protocol contain a variety of cells outside of the VZ, which are challenging to comprehensively define with limited microscopy channels, and the difficulty in confidently assigning single cilia to MAP2+ cells, we refrained from trying to quantify ciliation rates outside of the VZ. This illustrates the challenges associated with analysis of these more complex 3D models.

Given that neurons in vivo are generally thought to be ciliated, we were intrigued by the reduction of ARL13B-positive cilia in the matured neurons. Since work in mouse suggested that neuronal cilia can be ARL13B-negative (Brewer et al., 2024), this raised the possibility that our analyses so far relying on ARL13B as a ciliary marker could have “missed” the presence of cilia. We were thus wondering if we can find cilia that are negative for ARL13B, but positive for other ciliary markers on these neurons. We systematically looked at expression of Inositol polyphosphate-5-phosphatase E (INPP5E) (Figure 8A) and Adenylate Cyclase3 (AC3) (Figure 9A) in combination with PCNT and ARL13B staining in the different neuronal differentiation protocols. We only occasionally observed apparently INPP5E-positive/ARL13B-negative or AC3-positive/ARL13B-negative cilia, but even in these cases, weak ARL13B signal overlapping with the INPP5E respectively AC3 signal could be detected upon further scrutiny, indicating that all cilia on neurons up to day 70 in culture analyzed here were at least weakly positive for ARL13B. Likewise, all cilia in GIBCO NSCs, day 20 cortical NSCs, day 40 cortical neurons and cerebral organoids, were positive for INPP5E, which was uniformly strong in all conditions (Figure 8B and Supplementary Figure S16). In day 45 cerebral organoids, INPP5E signal fully overlapped with ARL13B signal at the ventricular zone and on the cilia found on cells away from the VZ (Supplementary Figure S17). In contrast, AC3 signal was rarely seen on cilia in hiPSCs, but increased at day 5 of the cortical differentiation to about 50% positivity. On more mature day 70 NMM or MAT cortical neurons, as well as on NGN2 neurons, the majority of cilia were AC3 positive (Figure 9B and Supplementary Figure S18). In day 45 cerebral organoids we found AC3-positive as well as a few AC3- negative cilia, as defined by ARL13B-positivity, both within the ventricular lumen and at the basal side of the VZ (Supplementary Figure S19).

The G-protein-coupled receptor 161 (GPR161) is known to disappear from primary cilia upon SHH stimulation in the developing neuronal tube (Mukhopadhyay et al., 2013). We found GPR161 expressed in the majority of ARL13B-positive cilia in the Gibco NSCs, the dual SMAD inhibition cortical neuronal differentiations, the MAT and NGN2 neurons (Supplementary Figure S20). After 24 h of SHH stimulation, we observed a significant decrease of GPR161-positive cilia in the Gibco NSCs and in cortical differentiations at days 20 and 40 (Figure 10A–C). Importantly, the number of cilia, based on ARL13B staining, was not affected by SHH stimulation. These results indicate that cilia on NSCs and young neurons respond to SHH signaling as expected by removing GPR161 from their cilia. Interestingly, when we performed the same 24 h stimulation on more mature neurons (day 56 and 70 cortical neurons and MAT neurons), we observed a higher proportion of cilia with residual GPR161-positive cilia (Figures 10B,C and Supplementary Figure S20), suggesting a decrease in responsiveness to SHH pathway stimulation.