Transient primary cilia mediate robust Hedgehog pathway-dependent cell cycle control

Summary

The regulation of proliferation is a primary function of Hedgehog (Hh) signaling in development. Hh signal transduction requires the primary cilium for several steps in the pathway [1–5]. Many cells only build a primary cilium upon cell cycle exit, in G0. In those proliferating cells that do make a cilium, it is a transient organelle, being assembled in G1 and disassembled sometime prior to mitosis [6–9]. Thus, the requirement for primary cilia presents a conundrum: how are proliferative signals conveyed through an organelle that is present for only part of the cell cycle? Here we investigate this question in a mouse medulloblastoma cell line, SMB55, that requires cilium-mediated Hh pathway activity for proliferation [10]. We show that SMB55 cells, and the primary cerebellar granule neuron precursors (GNPs) from which they derive, are often ciliated beyond G1 into S phase, and the presence of the cilium in SMB55 cells determines the periods of Hh pathway activity. Using live imaging over multiple cell cycles, we demonstrate that Hh pathway activity in either G1-S of the previous cell cycle or G1 of the cell cycle in which the decision is made is sufficient for cell cycle entry. We also show that Cyclin D1 contributes to the persistent effects of pathway activity over multiple cell cycles. Together our results reveal that even though the signaling organelle itself is transient, Hh pathway control of proliferation is remarkably robust. Further, primary cilium transience may have implications for other Hh-mediated events in development.

Graphical Abstract

Results

To understand how primary cilia function in a proliferative context, we examined primary GNPs from the developing mouse cerebellum. GNP proliferation requires Hh signaling transduced by primary cilia, and overactive Hh pathway activity in GNPs causes medulloblastoma [11–16]. We isolated proliferating GNPs from wild-type postnatal day 7 mice and assessed ciliation in an asynchronous population. Not only did most cells have an Arl13b⁺ cilium, but also many PCNA⁺ S-phase cells were ciliated (Figure 1A). This result is contrary to the widely-held view that the mammalian primary cilium is limited to G0/G1 phase but consistent with recent findings that cilia in vivo could persist through S phase [8]. We were interested in further investigating the presence and proliferative function of these S-phase cilia, but primary GNPs pose a challenge for long-term culture and manipulation [11], and NIH3T3 cells, the mouse fibroblasts often used to study Hh signaling, do not require the Hh pathway for their proliferation (Figure 1B).

Figure 1: SMB55 cells model the Hh pathway-dependence and S-phase ciliation of GNPs.

(A) Asynchronous CD-1 P7 GNPs stained for Arl13b, PCNA, and DNA. Insets highlight ciliated cells (i) not in S phase (Not S), (ii) in early S, and (iii) in late S, distinguished by PCNA foci. Scale bar: 10 μm, inset: 5 μm

(B) Proportion of EdU⁺ GNP, SMB55, or NIH3T3 cells after 48-hour growth in Hh pathway active (Hh ON) or inactive (Hh OFF) conditions as indicated and 2-hour EdU labeling. Mean+SEM from 2 (GNP) or 3 (SMB55, NIH3T3) independent experiments, n≥200 cells in each.

Significance determined by unpaired Student’s t test.

(C) Asynchronous SMB55 population expressing eYFP-PCNA and stained for Arl13b and DNA. Insets highlight ciliated cells as in (A). Scale bar: 10 μm, inset: 5 μm

(D) Proportion of SMB55 cells ciliated after 24-hour treatment with 100 nM SANT-1, 2 mM HU, or 10 μM RO-3306 compared to an asynchronous population. Mean+SEM from 3 independent experiments, n≥100 cells in each. Significance determined by one-way ANOVA with Tukey’s post-hoc test.

See also Figure S1

GNP-derived medulloblastoma cells are an attractive alternative, but most such cells often lose their Hh-pathway dependence when cultured [17]. We found that the mouse medulloblastoma line SMB55 [10] grows well in culture while maintaining two critical features of GNPs: Hh pathway-dependent proliferation and ciliation during the cell cycle. GNPs require Sonic hedgehog (Shh) ligand for growth (Figure 1B) [11]. SMB55 cells, by contrast, lack a wild-type copy of the pathway repressor Ptch1 (Figure S1A), leading to constitutive, ligand-independent pathway activation. To assess Hh pathway-dependent proliferation in SMB55 cells, we blocked signaling with SANT-1, an inhibitor of the pathway activator Smoothened (Smo) [18], demonstrated by reduced expression of the transcription factor Gli1 (Figure S1B). After 48-hour SANT-1 treatment and 2-hour EdU labeling, there was a significant reduction in the proportion of EdU⁺ S phase cells (Figure 1B). Smo depletion similarly inhibited growth (Figure S1C–E). Consistent with this Hh-pathway dependence, we confirmed previous work [19] showing SMB55 cells require the primary cilium for proliferation (Figure S1F–I). Arl13b staining revealed that cilia are present on S-phase SMB55 cells as in GNPs (Figure 1C). These experiments establish SMB55 cells as a uniquely appropriate model to study Hh pathway-dependent proliferation.

To further investigate cilium presence during the cell cycle, we arrested SMB55 cells for 24 hours in S phase using hydroxyurea (HU) or in G2 phase using RO-3306, a Cdk1 inhibitor. While the proportion of ciliated S phase-arrested cells was no different than the asynchronous control population, the proportion of ciliated G2-arrested cells was significantly reduced (Figure 1D). Therefore, SMB55 cells typically have a cilium during the cell cycle, including S phase, but lose the cilium by G2.

Because ciliary signaling has largely been studied in G0, we tested whether S-phase cilia are competent for Hh signal transduction. We treated NIH3T3 cells with Shh or SAG, a Smo agonist, for only 2 hours, reasoning that most cells would remain in the same cell cycle phase. We identified ciliated S-phase cells (PCNA⁺, Arl13b⁺) and measured ciliary Smo to assess pathway activation. For all treatments there was no significant difference between S-phase and non-S-phase ciliary Smo amount; pathway activators increased ciliary Smo in both populations (Figure 2A,A’). Similarly, ciliary Smo was similar in S-phase and non-S-phase SMB55 cells (Figure 2B,B’). We then asked whether cilium presence is temporally correlated with pathway activation within a cell cycle. We measured Gli1 mRNA in 24-hour HU-arrested SMB55 cells and found that these cells had high Gli1 mRNA, comparable to control cells (Figure 2C). The half-life of Gli1 mRNA is significantly shorter than 24 hours [20]. Therefore, the Hh pathway must be transcriptionally active in ciliated S-phase SMB55 cells. In contrast, RO-3306-arrested SMB55 cells, which typically do not have a cilium (see Figure 1D), had low Gli1 mRNA, comparable to SANT-1-treated cells (Figure 2C). This effect is likely specific to G2, as RO-3306 had no effect on Hh signaling itself (Figure S2). Thus, S-phase cilia are competent for Hh signal transduction, and in SMB55 cells, which have constitutively active Smo, presence or absence of a cilium determines pathway activity during the cell cycle.

Figure 2: Cilium presence determines Hh pathway activity during the cell cycle.

(A-B) Ciliary Smo intensity, indicating pathway activation, in (A) eYFP-PCNA NIH3T3 cells after 2-hour pathway stimulation with 100 nM Shh and (B) eYFP-PCNA SMB55 cells after 24-hour pathway inhibition with 100 nM SANT-1. Cells were stained for Smo and cilia (Arl13b), and PCNA foci distinguished S-phase cells. Scale bar: 10 μm, inset: 3 μm.

(A’-B’) Quantification of Smo intensity (Mean+SEM) at cilia from 3 independent experiments, n=30 cells in each. No significant difference in any condition between S-phase and non-S-phase cilia by two-way ANOVA with Tukey’s post-hoc test.

(C) Gli1 mRNA in SMB55 cells treated with SANT-1, HU, or RO-3306 for 24 hours. Gli1 mRNA amount was measured by RT-qPCR, normalized to Gapdh and expressed relative to the asynchronous control. Mean+SEM, n=3

See also Figure S2.

Having characterized Hh pathway activity during the SMB55 cell cycle, we asked when signaling is required to inform the proliferative decision. The Hh pathway controls proliferation by regulating cell cycle entry at G1/S [21]. Staining for markers of cell cycle arrest (p27) and commitment (phospho-Rb, pRb) showed that most SANT-1-treated SMB55 cells were p27⁺ and pRb^-, consistent with G0 arrest (Figure 3A). Thus, we asked at what point prior to G1/S Hh pathway activity must occur, using an approach analogous to that used to study regulation of the yeast cell cycle [22]. We tracked individual, asynchronously-growing cells through multiple cell cycles, using phase-contrast imaging over 48 hours. Addition of SANT-1 to the asynchronous population at the beginning of imaging allowed us to observe the effect of Hh pathway inhibition on cells starting at all points in the “previous cell cycle.” We then determined the outcome of the cell cycle entry/exit decision in the “decision cell cycle” by assessing whether each cell entered the ensuing mitosis (M₂) (Figure 3B). Because tracking cells in neurospheres, the typical mode of SMB55 growth, was not feasible, we performed imaging on 2-D SMB55 cultures, which also maintain Hh pathway-dependent proliferation (Figure S3A).

Figure 3: Hh pathway activity in either the previous cell cycle or the decision cell cycle is sufficient for cell cycle entry.

(A) SMB55 cells labeled with EdU and stained for pRb and p27 after 48-hour SANT-1 treatment. Scale bar: 10 μm

(B) Diagram of the cell cycle decision

(C) Effect of SANT-1 treatment in the previous cell cycle on SMB55 cell cycle entry. Based on the time of drug addition relative to M₁, the frequency of cells progressing to M₂ is shown, representing cell cycle entry frequency. Mean+SEM from 3 independent experiments, n=3–20 cells in each. Significance determined by two-way ANOVA and Tukey’s post-hoc test.

(D) Effect of SANT-1 washout at M₁ on cell cycle entry frequency. Mean+SEM from 3 independent experiments, n=10–40 cells in each. Significance determined by unpaired Student’s t test.

See also Figure S3, Video S1.

We identified the first division (M₁) after SANT-1 addition for each cell. Based on the time of M₁ relative to that of SANT-1 addition, we determined how long the pathway had been inhibited in the previous cell cycle. We then tracked each cell from M₁ onward to determine whether it entered the decision cell cycle, and thus continued to M₂ (Video S1). Remarkably, SMB55 cells treated with SANT-1 for up to 11 hours before M₁ entered the cell cycle as frequently as control cells (Figure 3C). Only when the pathway was inhibited for more than 12 hours before M₁ did cells fail to enter the decision cell cycle. Given that 4-hour SANT-1 treatment is sufficient to inhibit the pathway (Figure S3B), the pathway would not have been active in the decision cell cycle in any of the tracked cells. Thus, Hh pathway activity in G1 of the decision cell cycle is not necessary for cell cycle entry; signaling in the previous cell cycle is sufficient.

Since SMB55 cells typically bear a cilium in G1-S of the previous cell cycle and G1 of the decision cell cycle, we investigated whether pathway activity in G1 of the decision cell cycle is also sufficient for cell cycle entry. We treated SMB55 cells with SANT-1 for 17 hours and then washed out the inhibitor, assaying cell cycle entry in those cells at M₁ when washout occurred. Significantly more cells entered the cell cycle after washout compared to those remaining in SANT-1 past M₁ (Figure 3D), with no change in cell cycle length (Figure S3C). Therefore, Hh pathway activity in either the previous cell cycle or G1 of the decision cell cycle is sufficient for cell cycle entry.

How might Hh pathway activity in the previous cell cycle effect cell cycle entry in the decision cell cycle? We considered the example of serum-dependent proliferation in which mitogens in the previous cell cycle are similarly sufficient for cell cycle entry [23], and Cyclin D1 mediates persistence of mitogenic signal [24–26]. Although Cyclin D1 functions in G1 and is degraded in S phase, the protein can accumulate in G2 [25]. Given that Cyclin D1 is a well-established transcriptional target of the Hh pathway [21,27,28], and we observed its expression to be Hh pathway-dependent in SMB55 cells (Figure 4A,B), we predicted that Cyclin D1 might also accumulate in G2 in SMB55 cells. Consistent with our hypothesis, we observed a low level of Cyclin D1 protein in S phase-arrested cells but a high level in G2-arrested cells (Figure 4C). Next we synchronized cells in S phase with HU and then released the cells into G2. Cyclin D1 protein increased in the G2 cells compared to the S phase-arrested cells (Figure 4D). Treatment with SANT-1 throughout the experiment reduced the level of Cyclin D1 in G2 (Figure 4D), showing that its accumulation in G2 is Hh pathway-dependent. Thus, Cyclin D1 accumulation in G2 is also a feature of Hh pathway-dependent proliferation and likely contributes to the persistence of information about previous pathway activation into the decision cell cycle.

Figure 4: Cyclin D1 accumulation in G2 is a feature of Hh pathway-dependent proliferation.

(A) Ccnd1 mRNA after 24-hour SANT-1 treatment. Ccnd1 mRNA amount was measured by RT-qPCR, normalized to Gapdh, and expressed relative to the 0 nM control. Mean+SEM, n=3

(B-D) Cyclin D1 protein levels assessed by western blot with α-tubulin as a loading control. (B) 24 or 48-hour SANT-1 treatment, (C) 24-hour treatment with HU or RO-3306, (D) 24-hour HU then 6-hour RO-3306, all in the presence or absence of SANT-1. Under these conditions, Cyclin D1 runs as a doublet.

(E) Diagram of possible cell division decisions for sister cells vs. independent cells; M₁ and M₂ refer to the mitotic divisions in Figure 3B.

(F) Frequency with which observed (Obs.) control or SANT-1-treated sister-cell pairs made the same cell cycle entry decision compared to the expected (Exp.) frequency for independent cell pairs. Mean+SEM from 3 independent experiments, n=7–20 pairs each. Significance determined by unpaired Student’s t test.

A prediction following from Cyclin D1 protein accumulation in G2 is that sister cells derived from a given parental cell would inherit the same soluble cell cycle regulators and thus be expected to make the same decision about cell cycle entry (Figure 4E). To test this prediction, we considered the sister-cell pairs present in our live imaging experiment in Figure 3C. If the cells in those pairs made their cell cycle entry decision independently of each other, the percentage of sister-cell pairs expected to make the same decision would derive simply from the overall frequency of cell cycle entry at the same time points. We found instead that sister cells made the same decision more often than this expected frequency (Figure 4F). This result suggests that sister cells experience similar proliferative inputs, supporting a role for Cyclin D1 as one such input.

Discussion

The Hh pathway requires the primary cilium, but the cilium is present only transiently on dividing mammalian cells. We demonstrate that SMB55 cells, which require Hh pathway activity for proliferation, have two opportunities for Hh pathway activity prior to cell cycle entry based on cilium presence: G1-S of the previous cell cycle, and G1 of the decision cell cycle. Remarkably, Hh pathway activity in either signaling window is sufficient for cell cycle entry, supporting a model in which Hh pathway activity that has occurred over more than one cell cycle influences the cell cycle entry decision, i.e., that some element of the signal outlasts the signaling organelle itself.

Our experiments reveal that the cell cycle entry decision is robust to perturbations in pathway activity. Robustness is a well-established feature of Hh morphogen gradient interpretation [29,30], necessary in part because of fluctuations in ligand. However, since neither cilium assembly nor disassembly is synchronous [9,31–34], ciliation itself might also impose variability on the signaling level a given cell can attain. Illustrating this robustness, most control cells entered the cell cycle even though only about half of the G1-S population was ciliated. It is possible that the cilium is either present for only a portion of G1-S or does not assemble every cell cycle; the persistence of information would still allow such cells to proliferate.

Our work demonstrates that Hh pathway activation through cilia in G1-S of the previous cell cycle is sufficient to drive Cyclin D1 protein accumulation in G2 and cell cycle entry in the decision cell cycle. SMB55 cells can acquire sufficient signaling information for cell cycle entry as early as halfway through the previous cell cycle. We would define this point as the restriction point for Hh pathway activity [35] and note that recent work has similarly placed the restriction point for non-ciliary mitogen signaling in the previous cell cycle [23]. A restriction point in the previous cell cycle requires that some element of the pathway persists into the decision cell cycle, through G2, when the cells lack a cilium. The Hh-dependent accumulation of Cyclin D1 protein in G2 implicates Cyclin D1 in this persistence. Characterizing the mechanisms that underlie the persistence, such as inheritance of stable mRNA [24], or epigenetic changes [36], requires further work.

Although SMB55 cells are cancer cells, they reflect primary GNPs in their dependence on Hh pathway activity and their ciliation. Further, given the prevalence of Ptch1 mutations in human medulloblastoma [12,37], these Ptch1 mutant cells have potential clinical relevance. We cannot exclude the possibility that some aspect of proliferative control differs between SMB55 cells and GNPs and expect that future studies will extend these findings into developmental contexts.

Here we have used medulloblastoma cells to study Hh pathway-dependent proliferation and shown that this proliferative control is robust to the transient nature of the cilium during the cell cycle. Our work highlights the importance of considering cilia dynamics during the cell cycle when investigating cilium-dependent signaling events. It will be particularly interesting to test the implications of these findings on the role of Hh as a morphogen during developmental decisions.

STAR Methods

Resource Availability

Lead Contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Tim Stearns (stearns@stanford.edu).

Materials Availability

All shRNA plasmids generated in this study are available upon request.

Data and Code Availability

This study did not generate/analyze datasets or code.

Experimental Model and Subject Details

SMB55 cell culture

SMB55 cells were a gift of R. Segal (Dana Farber Cancer Institute, Harvard Medical School). SMB55 cells were cultured in DMEM/F-12 without L-Glutamine (Corning) supplemented with 2% B-27 without retinoic acid (Gibco) and 1% penicillin/streptomycin (HyClone). Cells were maintained as neurospheres in 3-D culture in untreated 96 well plates (Corning), changing media every 2 days. Cells were dissociated every 6 days with Accutase (Sigma-Aldrich) and plated at 150,000 cells/mL. All experiments, except for the live imaging experiments, were performed in 3-D culture. Cells were frequently tested to confirm that growth was dependent on Hh pathway activity using SANT-1 treatment and EdU incorporation as described below. SMB55 cells were genotyped to confirm the absence of the wild-type Ptch1 allele using primers recognizing the wild-type (Ptch-WT3, Ptch-WT4) and mutant alleles (NeoF3, PtchR3) as described in [12]. Genotyping PCR was performed using Taq polymerase on genomic DNA extracted from SMB55 cells using ZymoBead™ Genomic DNA kit (Zymo). Genomic DNA from Ptch1^−/− mouse embryonic fibroblasts and a wild-type mouse were used as a control for the mutant and wild-type alleles respectively. All cells were grown in a humidified incubator at 37 °C, 5% CO₂.

Other cell culture

NIH3T3 cells were obtained from ATCC and cultured in DMEM (Corning) plus 10% FBS (Gemini). To maximize ciliation of S-phase cells, NIH3T3 cells were split from dense plates and plated in dense (but not confluent) conditions. To induce ciliation in NIH3T3 cells, confluent cultures were starved in 0.5% FBS for 24 hours. HEK293T/17 cells were obtained from ATCC and cultured in DMEM plus 10% CCS (Hyclone). All cells were grown in a humidified incubator at 37 °C, 5% CO₂.

Mice

Pregnant wild-type CD-1 female mice or wild-type CD-1 litters were obtained from Charles River. The litter of one female (10–16 mice) was sacrificed at seven days (P7) for GNP isolation. All procedures were approved by the Institutional Animal Care and Use Committee in accordance with established animal care guidelines.

Method Details

GNP isolation and culture

10–16 P7 littermates were used for each GNP isolation experiment. Cerebella were dissected, and GNPs were isolated using a Percoll gradient as described previously [38], an approach that results in a population that is 95–99% GNPs [39]. Briefly, cerebella were minced with a scalpel and incubated at 37 °C for 30 minutes in a digestion buffer containing 1x Hanks’ Balanced Salt Solution (HBSS) with 10 U/mL papain (Worthington Biochemical) and 250 U/mL deoxyribonuclease (DNase I type II, Sigma-Aldrich). Following digestion, the papain solution was aspirated and replaced with 1x HBSS containing 8 mg/ml Ovomucoid (Worthington Biochemical), 8 mg/ml bovine serum albumin (Gemini), and 250 U/ml DNase. The cells were triturated with a Pasteur pipette to form a single cell suspension, centrifuged, resuspended in HBSS containing 0.02% bovine serum albumin, and passed through a 100 μm cell strainer. GNPs were then purified from the suspension using a step gradient of 35% and 65% Percoll (Sigma). Cells were collected from the 35%/65% interface, washed in HBSS containing 0.02% bovine serum albumin, and resuspended in Neurobasal media (Gibco) supplemented with 2% B-27 without retinoic acid (Gibco), 1% Sodium Pyruvate (Hyclone), 1% penicillin/streptomycin (HyClone), 1% L-glutamine (Gemini), and 60 μg/mL N-acetyl cysteine (Sigma-Aldrich). Human Sonic hedgehog (Shh, described below) was added to the media at 3 μg/mL. Cells were plated at 500,000 cells/mL in untreated 24 well plates (Corning) and then transferred to laminin-coated coverslips for immunofluorescence within 48 hours.

EdU incorporation assays

SMB55 cells were plated at 150,000 cells/mL in an untreated 96 well plate. Drug was added at the time of plating. At 24 hours, fresh drug was added and SMB55 cells were dissociated with Accutase. At 48 hours, Click-iT EdU (10 μM) was added to cell culture media and incubated for 2 hours. After the labeling period, the EdU was washed out and SMB55 cells were allowed to adhere to laminin-coated coverslips for approximately 3 hours before fixation, still in the presence of drug. The assay was performed similarly in GNPs and NIH3T3 cells except GNPs were plated at 500,000 cells/mL in untreated 24 well plates, NIH3T3 cells were plated on coverslips at 40,000 cells/well of a 24 well plate, and cells were not dissociated at the 24 hour timepoint.

Hedgehog pathway manipulation

The Smo antagonist SANT-1 (100 nM) and agonist SAG (100 nM) were used to inhibit or activate the Hh pathway for the indicated length of time. DMSO was used as a vehicle control in all assays involving SANT-1 or SAG. Human Sonic hedgehog (Shh) carrying two isoleucine residues at the N-terminus and a hexahistidine tag at the C-terminus was expressed in Escherichia coli Rosetta2(DE3)pLysS cells (Novagen) and purified by immobilized metal-affinity chromatography followed by gel-filtration chromatography as described previously [40]. The purified protein was used in signaling assays at 100 nM.(1.89 μg/mL) and for GNP growth at 158 nM (3 μg/mL).

Cell cycle arrest

Cell cycle arrest in SMB55 cells was performed by adding Hydroxyurea (2 mM) or RO-3306 (10 μM) at the time of plating, treating for 24 hours, and then collecting cells for downstream analysis. DMSO was used as a vehicle control in all assays involving RO-3306. For hydroxyurea washout, cells were washed 3 times with PBS for 5 min before adding new media containing RO-3306 for 6 hours.

Immunofluoresence and Click-it labeling

For fluorescence microscopy experiments, cells were fixed on poly-D lysine-coated (Sigma) glass #1.5 coverslips (Electron Microscopy Sciences). For SMB55 cells, these coverslips were coated overnight with 10 μg/mL laminin dissolved in DMEM/F-12 without L-Glutamine. For GNPs, these coverslips were coated overnight with 10 μg/mL laminin dissolved in Neurobasal media. The SMB55 and GNP neurospheres were allowed to adhere to the coverslips for 3–24 hours before fixation. NIH3T3 cells were grown on coverslips for the duration of the experiment. Cells were either fixed in 4% formaldehyde in PBS for 15 minutes at room temperate or 100% methanol for 15 minutes at −20 °C. Fixed cells were washed 3 times for 5 minutes in PBS and then blocked in 1% Normal Donkey Serum (Jackson Immuno Research), 0.1% Triton X-100 in PBS for 30 minutes. Primary antibody was diluted in blocking solution and incubated on coverslips for 1 hour at room temperature. Coverslips were washed with PBS, and then AlexaFluor conjugated secondary antibody (Invitrogen) was diluted 1:1000 in blocking solution and incubated on coverslips for 1 hour at room temperature. Coverslips were washed with PBS again and incubated with DAPI (Sigma-Aldrich) for 10 minutes. After a final wash, coverslips were mounted with Mowiol 4–88 (Calbiochem) in glycerol containing 1,4,-diazobicycli-[2.2.2] octane (DABCO, Sigma-Aldrich) antifade. The click reaction between EdU and Alexafluor-594 azide was performed with the Click-iT 594 Labeling Kit according to the manufacturer’s instructions. Cells were imaged on either a widefield Axiovert 200M microscope (Zeiss) with PlanApoChromat 63x/1.4 NA objective and OrcaER CCD camera (Hamamatsu) or a AxioObserver microscope (Zeiss) with a confocal spinning disk head (Yokogawa, Japan), PlanApoChromat 63x/1.4 NA objective, and a Cascade II:512 EMCCD camera (Photometrics). All images were acquired with Micro-Manager software [41]. Images of primary cilia were taken in 4–5 μm stacks (Widefield: 1 μm slices, Confocal: 0.5 μm slices) to ensure all cilia in the field of view were included.

Lentivirus production and infection

To make recombinant lentivirus, HEK293T/17 cells were cotransfected with the respective transfer vector and second-generation lentiviral cassettes (packaging vector psPAX2 and envelope vector pMD2.G) using 1 mg/mL polyethylenimine (PEI; Polysciences). The medium was changed 24 hours after transfection, and viral supernatant was harvested after an additional 48 hours. To concentrate the lentivirus, supernatant was filtered through a 0.45 μm filter and viral particles were pelleted by centrifugation (2:20 hours, 20°C, 50,000 × g). The viral pellet was then resuspended in PBS at 4 °C overnight. For lentiviral infection of SMB55 cells, concentrated lentivirus was added to cells at the time of plating in the presence of 4 μg/mL Hexadimethrine bromide (Polybrene, Sigma-Aldrich) for an infection efficiency of about 80%. Media was changed 24 hours post infection. For infection of NIH3T3 cells, concentrated lentivirus was added to cells 24 hours after plating in the presence of 8 μg/mL Polybrene for an infection efficiency of about 80%. Media was changed 24 hours post infection. For experiments involving lentiviral expression of pTRIP eYFP-PCNA, infected cells were selected in 400 μg/mL G418 (Gemini) beginning 2 days post infection for 2 weeks. A clonal NIH3T3 eYFP-PCNA line was isolated by limiting dilution and used for all experiments.

shRNA knockdown

For shRNA knockdown of Smo, miR-30a oligos containing 22-mers targeting Mouse Smo and Firefly Luciferase (targeting sequences in Key Resources Table) were cloned into the XhoI and EcoRI restriction sites of the lentiviral shRNA expression vector pMCB281/pMK1209 [42]. shRNAs targeting the same region of Firefly Luciferase have been used previously [43]. Knockdown efficacy was tested by immunofluorescence at 3 days post infection and proliferation was measured using the CellTiter 96 AQueous One Solution Cell Proliferation Assay (Promega) at 6 days post infection. For shRNA knockdown of Cep164, oligos containing 19-mers targeting Mouse Cep164 or Firefly Luciferase (targeting sequences in Key Resources Table) were cloned into the HpaI and XhoI restriction sites in pSICOR [44]. The empty, pSICOR plasmid was used as an additional control. Knockdown efficacy and effect on ciliation were tested by immunofluorescence at 4 days post infection. Proliferation was measured by a 2-hour EdU incorporation 4 days post infection.

Key Resources Table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
rabbit anti-phospho-Rb Ser807/811, clone D20B12 (1:1000)	Cell Signaling	Cat#8516 RRID:AB_11178658
mouse IgG1 anti-p27Kip, clone 57 (1:500)	BD	Cat#610241
mouse IgG1 anti-polyglutamylated tubulin, clone GT335 (1:500)	Adipogen	Cat#AG-20B-0020 RRID:AB_2490210
mouse IgG2a anti-Arl13b, clone N295B/66 (1:1000)	Neuromab	Cat#75-287 RRID:AB_2341543
rabbit anti-Arl13b (1:200)	Proteintech	Cat#17711-1-AP RRID:AB_2060867
rabbit anti-Smo (1:500)	T. Stearns lab, described in [3]
mouse IgG1 anti-gamma tubulin, clone GTU-88 (1:5000)	Sigma-Aldrich	Cat#T5326 RRID:AB_532292
mouse IgG2a anti-PCNA, clone PC10 (1:500)	BioLegend	Cat# 307902 RRID:AB_314692
rabbit anti-Cep164 (1:500)	T. Stearns lab, described in [45]
rabbit anti-Cyclin D1, clone SP4 (1:500)	Abcam	Cat# ab16663 RRID: AB_443423
mouse IgG1 anti-alpha tubulin, clone DM1A (1:10,000)	Sigma-Aldrich	Cat#05-829 RRID:AB_310035
AlexaFluor-conjugated secondary antibodies (1:1000 for IF)	Invitrogen
IRDye-conjugated secondary antibodies (1:20,000 for western)	LI-COR
Chemicals, Peptides, and Recombinant Proteins
SANT-1 (100 nM) (in DMSO)	Millipore	Cat#559303
SAG (100 nM) (in DMSO)	Enzo Life Sciences	Cat#ALX-270-426-M001
Human Sonic Hedgehog (Shh) (100 nM)	T. Stearns lab
Hydroxyurea (2 mM) (in water)	Sigma-Aldrich	Cat#H8627
RO-3306 (10 μM) (in DMSO)	AdipoGen	Cat#50-464-735
Mouse Laminin (10 μg/μl)	Gibco	Cat#23017-015
Critical Commercial Assays
Click-iT 594 Labeling Kit	Invitrogen	Cat#C10339
Maxima First Strand cDNA Synthesis Kit for RT-qPCR	Thermo Scientific	Cat#K1641
TaqMan™ Gene Expression Master Mix	Applied Biosystems	Cat#4369016
CellTiter 96 AQueous One Solution Cell Proliferation Assay	Promega	Cat#G3582
Experimental Models: Cell Lines
SMB55	R. Segal lab
NIH3T3	ATCC	Cat#CRL-1658 RRID:CVCL_0594
HEK293T/17	ATCC	Cat#CRL-11268 RRID:CVCL_1926
Experimental Models: Organisms/Strains
Mouse: Crl:CD1(ICR)	Charles River	Cat#:CRL:022 RRID:IMSR_CRL:022
Oligonucleotides
Mouse Gapdh VIC-MGB-PL probe	Applied Biosystems	Mm99999915_g1
Mouse Gli1 FAM-MGB probe	Applied Biosystems	Mm00494645_m1
Mouse Cyclin D1 FAM-MGB probe	Applied Biosystems	Mm00432359_m1
shRNA targeting sequence: Mouse Smo 5’-GGCTCAGTATGAGAAGAAGAAA-3’	This paper
shRNA targeting sequence: Firefly Luciferase (22-mer) 5’-ACCGCCTGAAGTCTCTGATTAA-3’	[43]
shRNA targeting sequence: Mouse Cep164 5’-GGTGATCTTTACTATTTCA-3’	This paper
shRNA targeting sequence: Firefly Luciferase (19-mer) 5’-TGAAGTCTCTGATTAAGTA-3’	This paper
NeoF3 Genotyping: 5’-TGGGGTGGGATTAGATAAATGCC-3’	[12]
PtcR3 Genotyping: 5’-TGTCTGTGTGTGCTCCTGAATCAC-3’	[12]
Ptch-WT3 Genotyping: 5’-CTGCGGCAAGTTTTTGGTTG-3’	[12]
Ptch-WT4 Genotyping: 5’-AGGGCTTCTCGTTGGCTACAAG-3’	[12]
Recombinant DNA
pTRIP eYFP-PCNA	J. Ferrell lab
psPAX2	Addgene	Cat#12260 RRID:Addgene_12260
pMD2.G	Addgene	Cat#12259 RRID:Addgene_12259
pSICOR	Addgene	Cat#11579 RRID:Addgene_11579
pMCB281/pMK1209	M. Bassik lab
Software and Algorithms
FIJI	ImageJ	https://fiji.sc/
Prism 8	Graphpad	https://www.graphpad.com/

RT-qPCR

For measurement of mRNA transcript amount by RT-qPCR, 300,000 SMB55 cells were treated for the indicated amount of time and then collected by centrifugation. Cells were resuspended in 1 mL Trizol reagent (Invitrogen) and total RNA was extracted according to the manufacturer’s instructions. The reverse transcription reaction was performed using the Maxima First Strand Synthesis kit with 185 ng of RNA. The product of this reaction was then assessed by qPCR in the Applied Biosystems Fast 7500 using the Taqman™ Gene Expression Master Mix and multiplexed Taqman™ probes. All reactions were performed in triplicate. Target gene values were normalized to Gapdh levels from the same reaction and expressed relative to control samples using the ΔΔCt method.

Immunoblotting

For immunoblotting, 1–2×10⁶ SMB55 cells were treated as indicated and then collected by centrifugation. Cells were lysed by shaking in 50 μL RIPA buffer (25 mM Tris 7.4, 150 mM NaCl, 2% NP-40, 0.25% Sodium Deoxycholate, 1 mM EDTA, 1 mM DTT) supplemented with 1 mg/ml each leupeptin, pepstatin, and chymostatin, and 1 mM phenylmethylsulfonyl fluoride for 1 hour at 4 °C. Soluble lysate was then isolated by centrifugation (1 hour, 4 °C, 16,100 × g). Protein concentration was measured by BCA and then 30 μg of protein was boiled in sample buffer and run on an SDS-PAGE gel. After transfer to nitrocellulose membrane (BioRad), membranes were blocked in 5% milk in TBST for 30 min, and then incubated in primary antibody diluted in TBST overnight at 4 °C. Membranes were washed and then incubated in secondary antibody solution for 1 hour at room temperature. IRDye conjugated secondary antibodies (LI-COR) were diluted 1:20,000 in TBST. Membranes were washed again and scanned on a LI-COR Odyssey scanner.

Live Imaging

For live imaging of SMB55 cells, 525,000 SMB55 cells were plated as neurospheres in 3D culture. After 2 days of growth, spheres were plated in fresh media on 35 mm Fluorodish glass imaging dishes (World Precision Instruments) that had been coated with poly-D lysine and then laminin as described for immunofluorescence microscopy. SANT-1 was added at the time of plating. Cultures were imaged by phase-contrast microscopy on a Keyence BZ-X700 all-in-one fluorescence microscope with a Plan Fluor 20x/0.45 NA Ph1 objective (Nikon) for 48 hours at 5-minute intervals in a humidified, temperature-controlled (37 °C), and CO₂-controlled (5%) chamber (Tokai Hit). For washout experiments, cells on imaging dishes were treated for 16.5 hours with SANT-1 beginning at the time of plating. To washout the SANT-1, three 5-minute washes were performed with fresh media. In the washout condition, this media contained DMSO only as a vehicle control. In the non-washout condition, this media still contained SANT-1. The imaging was started at the end of this procedure and the washout was considered to begin at 17 hours post SANT-1 addition. We selected a 17-hour SANT-1 treatment prior to M₁ because without washout this condition largely prevented cell cycle entry (see Figure 4C). Cultures were then imaged for 48 hours. At the end of all live imaging experiments, cells were labeled with EdU for 2 hours before fixation to confirm the effect on SANT-1 on the population’s proliferation before performing analysis of single cells.

Live imaging analysis

All live imaging analysis was performed in FIJI. To analyze image sequences for the live imaging experiment, all mitotic events in the first 4–19 hours after drug addition were identified. The frame containing cytokinesis was used to define the time of mitosis relative to drug addition. Daughter cells of each mitotic event were then tracked manually through the image sequences. If the cell divided, the time of cytokinesis was recorded and the cell was considered to have entered the cell cycle. If the cell was tracked to the end of the image sequence without dividing, it was counted as not dividing and thus not entering the cell cycle. Cells that left the field of view, were not able to be tracked with certainty, or had an aborted or multipolar mitosis were not included in the analysis. For sister cell analysis, only pairs in which both cells could be defined as entering the cell cycle or not were included. In the washout experiment, cells born from mitotic events only in the first 2 hours after washout were tracked and analyzed, in order to capture cells in which SANT-1 was washed out as close to M₁ as possible. Tracking was performed similarly as above, except that cells were considered to not enter the cell cycle if they did not divide within 30 hours. Cell cycle length was determined by the cytokinesis-cytokinesis time in cells that did enter the cell cycle and divide.

Image quantification

All image quantification was performed in FIJI. To quantify EdU incorporation, the number of EdU⁺ nuclei was expressed relative to the total number of nuclei (identified by DAPI). Proportion of cells ciliated, Smo intensity, and Cep164 intensity were quantified from maximum projections of a 5 μm z-stack (1 μm slices) taken on the Axiovert. The proportion of cells ciliated was determined by counting the number of cilia in a field of view and dividing that count by the number of nuclei. Cilia were always defined by two markers- either two axoneme markers (Arl13b and poly-glutamylated tubulin) or an axoneme and a basal body marker (Arl13b and gamma tubulin). SMB55 cilia are relatively short and so to avoid counting ciliary vesicles as cilia, only elongated Arl13b⁺ structures were counted. To quantify intensity of Smo at cilia, the cilium was defined by drawing a region of interest around the Arl13b⁺ region. Then average Smo pixel intensity in this region was measured. Smo intensity in a similarly-sized region of interest next to but not overlapping with each cilium was also measured and subtracted to normalize for background fluorescence. Similarly, Cep164 intensity was measured by drawing a region of interest around the gamma tubulin focus, measuring the average Cep164 intensity in this region, and subtracting background intensity of a nearby region of similar size. Cep164 was measured only at the Cep164⁺ centrosome (containing the mother centriole). If the mother centriole was not obvious (as in the case of Cep164 knockdown), the gamma tubulin focus with the higher Cep164 intensity was selected.

Quantification and Statistical Analysis

Statistical analysis

All statistical analysis was performed in Prism 8 (Graphpad). Details for statistical tests used can be found in the figure legends. Figure legends also indicate the number of independent replicates performed and the number of cells analyzed for each condition of each replicate (n). All graphs show the mean and SEM. In all cases, significance was defined as *p-value ≤ 0.05, **p-value ≤ 0.01, ***p-value ≤ 0.001, and ****p-value ≤ 0.0001. ns indicates no significance.

Analysis of sister cell fate

To test hypothesis that sister cells usually make the same decision about cell cycle entry, we compared the frequency with which sister cell pairs made the same decision about cell cycle entry to the frequency we would expect if the two cells made an independent decision. We first identified all sister cell pairs from the live imaging data set in Figure 3C for which the cell cycle entry outcome was known for both sisters. Each pair was given an observed value of 1 (if the cells made the same decision- both cells divide or both cells arrest) or 0 (if the cells made a different decision- one divides and one arrests). Each pair was also given an expected value determined by calculating the probability (between 0 and 1) that the two cells would make same decision if they behaved independently based on the cell cycle entry frequency for cells at the same time point (as measured in Figure 3C). Then, the observed values were added up for all pairs in an experiment and divided by the number of pairs, giving the frequency with which sister cells were observed to make same decision. Similarly, the expected values were added up for all pairs in an experiment and divided by the number of pairs, giving the frequency with which independent cells would be expected to make the same decision.

How does a transient primary cilium sustain Hedgehog (Hh)-dependent proliferation? Ho et al. show that despite the transience of the cilium, cell cycle entry can be driven by Hh activity in the previous cell cycle. Thus, the effects of signaling outlast the signaling organelle itself, making Hh pathway control of proliferation remarkably robust.

• GNPs and medulloblastoma cells are ciliated through G1 into S phase

• Cilium presence during the cell cycle determines periods of Hh pathway activity

• Hh pathway activity in the previous cell cycle drives cell cycle entry

• Cyclin D1 mediates Hh pathway signal persistence over multiple cell cycles