Ciliopathy protein HYLS1 coordinates the biogenesis and signaling of primary cilia by activating the ciliary lipid kinase PIPKIγ.

Mutation of ciliopathy protein HYLS1 causes the perinatal lethal hydrolethalus syndrome (HLS), yet the underlying molecular etiology and pathogenesis remain elusive. Here, we reveal unexpected mechanistic insights into the role of mammalian HYLS1 in regulating primary cilia. HYLS1 is recruited to the ciliary base via a direct interaction with the type Iγ phosphatidylinositol 4-phosphate [PI(4)P] 5-kinase (PIPKIγ). HYLS1 activates PIPKIγ by interrupting the autoinhibitory dimerization of PIPKIγ, which thereby expedites depletion of centrosomal PI(4)P to allow axoneme nucleation. HYLS1 deficiency interrupts the assembly of ciliary NPHP module and agonist-induced ciliary exit of β-arrestin, which, in turn, disturbs the removal of ciliary Gpr161 and activation of hedgehog (Hh) signaling. Consistent with this model of pathogenesis, the HLS mutant HYLS1D211G supports ciliogenesis but not activation of Hh signaling. These results implicate mammalian HYLS1 as a multitasking protein that facilitates ciliogenesis and ciliary signaling by coordinating with the ciliary lipid kinase PIPKIγ.

INTRODUCTION

The primary cilium, a sensory organelle on the surface of most mammalian cells, contains several subsegments including the basal body (BB), transition fibers (TFs), transition zone (TZ), and the microtubule axoneme ensheathed by the ciliary membrane contiguous with the plasma membrane (, ). Primary cilia process divergent extracellular signals vital for embryonic development and tissue homeostasis (–). Ciliary dysfunction leads to a panel of syndromic diseases collectively termed as ciliopathies (, ).

Hydrolethalus syndrome (HLS) is one of the deadliest ciliopathies (, ). Affected Finnish fetuses are linked to homozygosity for a missense mutation in the HYLS1 gene, encoding HYLS1 (hydrolethalus syndrome protein 1), and termed HLS1 (). We and others reported that HYLS1 orthologs in nematode, Xenopus, and Drosophila are required for ciliogenesis by regulating the organization of TFs and TZ, the docking of intraflagellar transport (IFT) particles to the ciliary base, and the ciliary gating function (–). However, how HYLS1 contributes to mammalian cilia remains uncharacterized. Although the HLS1 disease mutant of human HYLS1 was reported to mislocalize to the nucleus (), the corresponding nematode mutant localizes properly to the ciliary base and largely rescued ciliogenesis defects in hyls-1−/− worms (). Patients with HLS exhibit polydactyly, brain abnormalities, and cleft palate (), which are often caused by abnormal Hedgehog (Hh) signaling (). Likewise, some HLS cases are linked to mutations in KIF7 (), which encodes a conserved regulator of the Hh pathway KIF7 (–). Since Caenorhabditis elegans lacks the Hh pathway (), these observations suggest that mammalian HYLS1 may have evolved more complex functions than its nematode counterpart.

When ciliogenesis is induced, one key event is the serum starvation–induced recruitment of tau tubulin kinase 2 (TTBK2) to the distal end of the mother centriole (MC), which removes the microtubule-capping protein CP110 to allow the elongation of microtubule axoneme (). We found that the centrosomal phosphatidylinositol 4-phosphate [PI(4)P], produced by inositol polyphosphate 5-phosphatase INPP5E in nonciliated cells, prevents TTBK2 recruitment by inhibiting its interaction with the distal appendage/TF protein CEP164 (outer dense fiber protein 2) (). Once INPP5E is removed from the MC after serum starvation, the centrosomal type Iγ PI(4)P 5-kinase (PIPKIγ) exhausts PI(4)P and permits TTBK2 recruitment (). Yet, how PIPKIγ is regulated in vivo remains unclear. Our current work reveals a previously unknown physical and functional connection of mammalian HYLS1 with PIPKIγ at the ciliary base, which regulates both biogenesis and signaling of primary cilia via distinct mechanisms.

RESULTS

Ciliary localization and function of mammalian HYLS1

Superresolution three-dimensional structured illumination microscopy (3D-SIM) was used to characterize the subcellular localization of mammalian HYLS1. 3D-SIM revealed HYLS1 at the proximal end (marked by CEP152), but not the distal end (marked by CEP164), of the centrioles/BB in both nonciliated (Fig. 1A) and ciliated (Fig. 1B) human renal cortical tubular epithelial (RCTE) cells. This distribution appears to be different from nematode HYLS-1 that partially colocalizes with TFs (). Of note, the HYLS1 signal forms a ring-like pattern slightly bigger than CEP152 but similar to CEP164. No detectable HYLS1 signal was observed in other subcompartments of cilia (Fig. 1), suggesting that mammalian HYLS1 is highly enriched, if not restricted, to the proximal end of centrioles/BB. HYLS1 appeared at the same subcellular locales in human retinal pigment epithelial (RPE) cells (fig. S1, A and B). To determine how mammalian HYLS1 regulates primary cilia, we depleted HYLS1 from RCTE and RPE cells using two distinctive small interfering RNAs (siRNAs) (fig. S1, C and D). Similar to our observations in nematode (), HYLS1-depleted RCTE cells exhibited no cilia (Fig. 1C) or severely truncated cilia (Fig. 1D), which were also observed in HYLS1-depleted RPE cells (fig. S1E). Thus, the necessity for HYLS1 in ciliogenesis is conserved from C. elegans to human cells.

Fig. 1

Characterization of the localization and function of mammalian HYLS1.

(A and B) HYLS1 is localized to the proximal end of centrioles (A) and the ciliary BB (B) in mammalian cells. Renal cortical tubular epithelial (RCTE) cells, before (A) and after (B) 24-hour serum starvation, were subjected to indirect immunofluorescence (IF) labeling with indicated primary antibodies and analyzed with 3D-SIM. PolyE-tub, polyglutamylated tubulin. Scale bars, 0.5 μm. Images numbered as 0 and +5 indicate the bottom and top sections in a z stack of images showing HYLS1 and CEP164, respectively. A cartoon model presents the positioning of HYLS1 at the proximal end of the MC/BB. (C and D) Loss of HYLS1 inhibits ciliogenesis. RCTE cells were treated with negative control (siNC) or two distinct HYLS1-specific (siHYLS1-O1 and siHYLS1-O2) siRNAs for 48 hours, serum starved for 24 hours, and then subjected to IF microscopy using indicated antibodies. Scale bars, 2 μm. The percentage of ciliated cells (C) and the length of cilium (D) were quantified (n > 100). Results from at least three independent experiments were statistically analyzed and plotted as means ± SEM. ***P < 0.001.

Loss of mammalian HYLS1 interrupts the integrity of TFs and the TZ

The assembly of the primary cilium is a precisely controlled, multistep process. In nematode, HYLS-1 is essential for the proper formation of TFs, the docking of IFT particles, and the TZ organization (). Unexpectedly, most TF components, such as CEP164, SCLT1 (sodium channel and clathrin linker 1), CEP83, and ODF2 (outer dense fiber protein 2) (), were not affected by HYLS1 depletion in RCTE (Fig. 2A) or RPE (fig. S2A) cells. Because these proteins do not have homologs in the nematode genome, we specifically examined FBF1 (Fas binding factor 1), which is a unique TF functional component conserved across ciliated species and mislocalized in hyls-1 null worms (). After HYLS1 depletion, FBF1 signal at the MC/BB appeared normal in quiescent cells (fig. S3A); yet, in mitotic cells, FBF1 recruitment to the newly formed MC was impaired (fig. S3B). To investigate this phenomenon, we generated HYLS1-knockout RCTE cells using CRISPR-Cas9 (fig. S4, A to C). These HYLS1-edited cells, which grew very slowly with centrosome amplification (fig. S4D), exhibited a greater decrease in ciliogenesis (fig. S4E) and a significant loss of FBF1 at the centrosome/BB (fig. S4F). These data suggest that mammalian HYLS1 still plays a role in recruiting FBF1 to TFs but to a lesser extent compared to its nematode counterpart. Unlike in hyls-1–mutated worm, IFT140 (Fig. 2B) and IFT88 (Fig. 2B and fig. S2B) sustained their localization at the centrioles/BB in HYLS1-depleted cells, indicating that docking of IFT complexes to TFs is largely normal, likely due to sufficient FBF1 function. Thus, the defective ciliogenesis of HYLS1-depleted mammalian cells is likely caused by an unidentified, FBF1-independent mechanism.

Fig. 2

HYLS1 depletion interrupts the assembly of NPHP module at transition zone.

RCTE cells transfected with negative control (siNC) or HYLS1-specific (siHYLS1-O1 and siHYLS1-O2) siRNAs for 48 hours were serum starved for 24 hours and then analyzed by IF microscopy. (A) TF components (CEP164, SCLT1, ODF2, and CEP83) were not affected by HYLS1 depletion. (B) IFT proteins IFT88 and IFT140 were not affected by HYLS1 depletion. (C) TZ components (MKS1, TCTN1, MKS5, and CEP290) were not affected by HYLS1 depletion. (D) NPHP1 and NPHP4, the core components of the NPHP module, were absent from the BB in HYLS1-depleted cells. PolyE-tub antibody labeled both the BB and the axoneme of cilia. Arl13b antibody labeled the ciliary shaft. Scale bars, 2 μm. The percentage of cells with normal localization of the tested protein was quantified in control or HYLS1-depleted cells (n > 100). Results from at least three independent experiments were statistically analyzed and plotted as means ± SEM. N.S., no significant difference. ***P < 0.001.

We next examined TZ organization in cells lacking HYLS1. In mammalian cells, the TZ compartment contains at least three interconnected protein complex modules: the nephronophthisis (NPHP) module, the megakaryocytes (MKS) module, and the Joubert syndrome (JBTS) module (–). In HYLS1-depleted cells, MKS1 and TCTN1 (MKS components), CEP290 (JBTS component), and MKS5 all appeared normal (Fig. 2C and fig. S2C). However, NPHP1 and NPHP4, core components of the NPHP module, were lost at the TZ in HYLS1-depleted cells (Fig. 2D and fig. S2D), indicating that mammalian HYLS1 is necessary for integrating the NPHP module at the TZ.

Mammalian HYLS1 regulates the axoneme assembly similar to PIPKIγ

Because defects in TF and TZ architecture in HYLS1-depleted cells are unlikely to disrupt ciliogenesis, we extended our analyses to axoneme assembly. Intriguingly, HYLS1-depleted RCTE cells failed to recruit TTBK2 to the distal end of MC after serum starvation (Fig. 3, A and C), which is the key to licensing the nucleation of microtubule axoneme (). As a result, the microtubule-capping protein CP110 was retained at the MC (Fig. 3, B and C) and no axoneme elongation occurred. Similar defects were observed in HYLS1-depleted RPE cells (fig. S5, A to C). These phenotypes in HYLS1-depleted cells recapitulate defects observed in PIPKIγ-depleted cells (), implying an unexpected connection between HYLS1 and phosphoinositide signaling in the context of primary cilia. In support of this possibility, 3D-SIM images showed colocalization of HYLS1 with PIPKIγ at the proximal end of centrioles and BB in nonciliated and ciliated cells (Fig. 3D and fig. S5D), respectively. PIPKIγ depletion severely reduced HYLS1 signal at centrioles/BB (Fig. 3E and fig. S5E) without changing the global protein level of HYLS1 (Fig. 3F), indicating that PIPKIγ is required for HYLS1 targeting to the centrioles/BB. However, PIPKIγ does not need HYLS1 to target to these subcellular locales (Fig. 3G).

Fig. 3

HYLS1, recruited to the BB by PIPKIγ, is necessary for TTBK2 recruitment and CP110 removal.

(A and B) Loss of HYLS1 interrupts the recruitment of TTBK2 (A) and removal of CP110 (B) at the MC induced by serum starvation. RCTE cells were treated with negative control (siNC) or HYLS1-specific (siHYLS1-O1 and siHYLS1-O2) siRNAs for 48 hours, serum starved for 24 hours, and then subjected to indirect IF microscopy with indicated antibodies. Scale bars, 2 μm. (C) The percentage of cells with TTBK2-positive BB/MC (top) or with CP110 on both centrioles (bottom) was quantified (n > 100). (D) HYLS1 colocalizes with PIPKIγ at the centrosome or BB. RCTE cells before (+FBS) and after (−FBS) 24-hour serum starvation were stained for IF microscopy with indicated antibodies. Images were acquired with 3D-SIM. Scale bar, 0.5 μm. (E) PIPKIγ is indispensable for HYLS1 localization at the centrosome or BB. RCTE cells treated with control (siNC) or two distinct PIPKIγ-specific (siPIPKIγ-O1 and siPIPKIγ-O2) siRNAs for 48 hours. With (−FBS) or without (+FBS) 24-hour serum starvation, cells were analyzed by IF microscopy with indicated antibodies. 4′,6-diamidino-2-phenylindole (DAPI) was used to label nuclei. Scale bar, 5 μm. The percentage of HYLS1-positive centrosome/BB in control or PIPKIγ-depleted group was quantified in >100 cells. (F) The total protein level of HYLS1 was not affected by PIPKIγ depletion in RCTE cells. (G) HYLS1 is not required for PIPKIγ to target to the BB. Control (siNC) or HYLS1-depleted RCTE cells were serum starved for 24 hours and then subjected to IF microscopy to visualize PIPKIγ and cilia (polyE-tub). The percentage of PIPKIγ-positive BB was quantified in >100 cells. All statistical analyses were performed using results from at least three independent experiments and plotted as means ± SEM. N.S., no significant difference. ***P < 0.001.

HYLS1 directly binds to and activates PIPKIγ

The colocalization of HYLS1 and PIPKIγ suggests a physical association between them. Either ectopically expressed (Fig. 4A) or endogenous (Fig. 4B) HYLS1 and PIPKIγ formed a protein complex in mammalian cells. Further investigation with in vitro protein pulldown assay using recombinant proteins purified from Escherichia coli demonstrated a direct interaction between HYLS1 and PIPKIγ (Fig. 4C). As shown in Fig. 4D, Flag-tagged HYLS1 was able to coprecipitate full-length (FL), N terminus of (NT), or C terminus–truncated (ΔCT) PIPKIγ, but not the CT alone, suggesting that HYLS1 binds to the NT of PIPKIγ. Although PIPKIγ NT is conserved among splicing variants, we did not find HYLS1 at other PIPKIγ-positive subcellular locales such as focal adhesions and adherens junctions (, ), suggesting that additional mechanism(s) confine HYLS1 to the centrosome/BB beyond PIPKIγ-mediated recruitment. Given that HYLS1 could not convert PI(4)P to PI(4,5)P2, that PIPKIγ generated more PI(4,5)P2 in the presence of HYLS1 in a dose-dependent manner (Fig. 4E) indicates that HYLS1 activates PIPKIγ. HYLS1 had no effect on the activity of PIPKIγ-ΔCT (Fig. 4F), although this peptide contains an intact NT and exhibits normal interaction with HYLS1 (Fig. 4D), suggesting that the CT of PIPKIγ is necessary for HYLS1 to activate PIPKIγ.

Fig. 4

HYLS1 binds to and activates PIPKIγ.

(A to C) HYLS1 directly interacts with PIPKIγ. (A) Human embryonic kidney (HEK) 293T cells were cotransfected with hemagglutinin (HA)–PIPKIγ and Flag-HYLS1 for 48 hours and then subjected to immunoprecipitation (IP) with anti-Flag antibody or regular mouse immunoglobulin G (IgG). (B) RCTE cells were used for IP with anti-HYLS1 antibody or regular mouse IgG. (C) Recombinant His-HYLS1 and His-PIPKIγ (0.5 μg), purified from E. coli, were incubated with protein G Sepharose and HYLS1 antibody or regular mouse IgG. The resulting precipitates (A to C) were analyzed by immunoblotting with indicated antibodies. (D) HYLS1 binds to the N terminus (NT) of PIPKIγ. Each of HA-tagged PIPKIγ polypeptides including the full length (FL), C terminus (CT, amino acids 435 to 694), NT (amino acids 1 to 150), and the CT-deleted fragments (ΔCT, amino acids 1 to 434) was coexpressed with Flag-HYLS1 in HEK293T cells. Co-IP assay was performed using anti-Flag antibody or regular mouse IgG. The resulted precipitates were analyzed by immunoblotting using Flag or HA antibody. A schematic illustration summarizes the interaction between HYLS1 and PIPKIγ. (E) HYLS1 activates the FL PIPKIγ (Iγ-FL) in a dose-dependent manner. Purified recombinant His-tagged FL (Iγ-FL) PIPKIγ (1 μg) was used for kinase assay with or without indicated amount of purified His-HYLS1. PI(4,5)P2 was separated on thin-layer chromatography and determined by autoradiography. PIPKIγ activity was represented by PI(4,5)P2 intensity and normalized against PIPKIγ alone. (F) HYLS1 has no effect on the CT-truncated PIPKIγ (Iγ-ΔCT). The PI4P 5-kinase activity of purified Iγ-FL or Iγ-ΔCT (1 μg) was determined with or without 10 μg of purified HYLS1 proteins. (E and F) Protein inputs were determined by Coomassie Blue staining. Results from three independent experiments were statistically analyzed and plotted as means ± SEM. ***P < 0.001. N.S., no significant difference.

HYLS1 binding disrupts the autoinhibitory, antiparallel dimer of PIPKIγ

Via in vitro pulldown assay using purified recombinant proteins, we noticed that the NT and CT of PIPKIγ bind to each other (Fig. 5, A and B). HYLS1 reduced the binding between PIPKIγ-CT and PIPKIγ-ΔCT (Fig. 5C), suggesting that HYLS1 competes with the CT of PIPKIγ for binding to its NT. The purified recombinant PIPKIγ-CT increased the activity of FL PIPKIγ (Fig. 5D), similar to the effect of HYLS1. Together, these data suggest a model where PIPKIγ maintains a low-activity status via a head-tail binding that masks its kinase domain, and HYLS1 binding to the NT of PIPKIγ interrupts this autoinhibitory conformation and activates PIPKIγ. To further distinguish whether the head-tail binding of PIPKIγ is an intra- or intermolecular interaction, we used size exclusion chromatography (SEC). As shown in Fig. 5E, the elution curve of PIPKIγ alone exhibited an extra peak (black arrow) compared to the mixture of PIPKIγ and HYLS1. This peak, which migrated the fastest, likely represents PIPKIγ dimer according to the protein size standards used with SEC, and the following elution peak likely represents PIPKIγ monomer (Fig. 5E, red arrow). The immunoblotting analyses of SEC elution fractions also indicated that PIPKIγ and HYLS1 were detected in common elution fractions in when mixed, supporting the conclusion that PIPKIγ and HYLS1 form a complex (Fig. 5E); this complex migrated faster than HYLS1 alone and slower than PIPKIγ alone in SEC. Together, our results support the existence of a PIPKIγ dimer in solution via a head-tail binding and suggest that the HYLS1 binding to the NT of PIPKIγ interrupts the dimerization or higher-grade oligomerization of PIPKIγ. By this means, HYLS1 blocks the autoinhibitory conformation of PIPKIγ and helps expose the kinase domain, thus enhancing PIPKIγ activity (Fig. 5F). This could be a unique on-site regulatory mechanism for PIPKIγ activity at the centrosome/BB.

Fig. 5

HYLS1 interrupts the autoinhibitory dimerization of PIPKIγ.

(A and B) The CT of PIPKIγ binds to the NT of PIPKIγ. Myc-tagged PIPKIγ CT (Myc-Iγ-CT) was coexpressed with HA-tagged FL [HA-Iγ-FL, (A)] or NT [HA-Iγ-NT, (B)] in HEK293T cells. Cells were used for co-IP with anti-HA antibody and analyzed by immunoblotting using indicated antibodies. (C) HYLS1 inhibits the head-tail binding of PIPKIγ. The association of Myc-Iγ-CT with HA-Iγ-ΔCT in HEK293T cells with or without Flag-HYLS1 was analyzed by co-IP and immunoblotting using indicated antibodies. Relevant intensity of the Myc-Iγ-CT band pulled down by HA-Iγ-ΔCT was quantified, and results from three independent experiments were statistically analyzed and plotted as means ± SEM. **P < 0.01. (D) The purified C-terminal polypeptide of PIPKIγ activates PIPKIγ in a dose-dependent manner. Protein inputs were determined by Coomassie Blue staining. Results from three independent experiments were quantified, statistically analyzed, and plotted as means ± SEM. ***P < 0.001. (E) HYLS1 disrupts PIPKIγ dimerization. His-PIPKIγ (30 μg) purified from E. coli was subjected to SEC with or without His-HYLS1 (100 μg). Size standards for molecular weight calibration (inset): myoglobin (17 kDa), ovalbumin (44 kDa), γ-globulin (158 kDa), and thyroglobin (670 kDa). Elution fractions were subjected to immunoblotting using indicated antibodies. (F) Working model: PIPKIγ forms antiparallel dimer via an intermolecular head-tail binding; HYLS1 binding interrupts this autoinhibitory dimerization of PIPKIγ and exposes its kinase domain for catalytic action.

HYLS1 promotes the licensing of axoneme elongation by facilitating PI(4)P removal

We next tested whether HYLS1-mediated activation of PIPKIγ contributes to the exhaustion of INPP5E-generated centrosomal PI(4)P after serum starvation, which is critical for TTBK2 recruitment and subsequently CP110 removal and axoneme elongation (). Because the PI(4)P antibody yields much less cytosolic background in immortalized mouse renal collecting duct epithelial cells (IMCD3) than in RCTE cells, we used IMCD3 cells for this study. While the PI(4)P pool at the distal end of MC in nonciliated cells () was unaffected [Fig. 6A, +FBS (fetal bovine serum)], a significant amount of PI(4)P was retained at centrosomes in HYLS1-depleted cells after serum starvation (Fig. 6A, −FBS). Further investigation showed that the kinetics of PI(4)P removal was significantly delayed in HYLS1-depleted cells (Fig. 6B, top), although the INPP5E removal induced by serum starvation was insensitive to HYLS1 depletion (Fig. 6B, bottom). As expected, HYLS1-depeleted cells showed markedly suppressed TTBK2 recruitment (Fig. 6C, top) and strongly increased CP110 retention (Fig. 6C, bottom) at the MC/BB. Thus, HYLS1 is necessary for the licensing of axoneme elongation by facilitating PI(4)P exhaustion at the MC/BB.

Fig. 6

HYLS1 is necessary for the efficient depletion of centrosomal PI(4)P in response to serum starvation.

(A to C) Loss of HYLS1 caused a retention of centrosomal PI(4)P in serum-starved cells. IMCD3 cells were transfected with control (siNC) or HYLS1-specific (siHYLS1) siRNAs for 48 hours, and then serum starved (−FBS) or not (+FBS) for variant times as indicated before subjected to indirect IF microscopy to visualize PI(4)P, INPP5E, TTBK2, and CP110. CEP135 represents the centrosome/BB. Representative images after 4 hours of serum starvation show that the centrosomal PI(4)P has been lost in control cells but not in HYLS1-depleted cells. The percentage of cells showing positive signal of centrosomal INPP5E (B, top), centrosomal PI(4)P (B, bottom), TTBK2 at the MC (C, top), or CP110 at the MC (C, bottom) at indicated time point was quantified in control or HYLS1-depleted cells. Results from three different experiments were statistically analyzed and plotted. (D) Codepletion of INPP5E (siINP) recovered the serum starvation–induced recruitment of TTBK2 to the BB, removal of CP110 from the MC, and ciliogenesis in HYLS1-depleted RCTE cells. (E) Overexpression of PIPKIγ compensated for HYLS1 depletion and recovered TTBK2 recruitment and ciliogenesis. RCTE cells were transfected with control or HYLS1 siRNA for 24 hours followed by transfection with empty vector or HA-tagged, PACT-fused PIPKIγ-ΔCT (Iγ-ΔCT) for another 24 hours and then serum starved for 24 hours. Cells were then labeled with indicated antibodies for IF microscopy. γ-tub, γ-tubulin. Scale bar, 2 μm. The percentage of TTBK2-positive MC/BB, CP110-positive MC/BB, or ciliated cells was quantified in each group (n > 100). All statistical analyses were performed using results from at least three independent experiments and plotted as means ± SEM. ***P < 0.001 and **P < 0.01.

Because HYLS1 depletion had no effect on INPP5E removal (Fig. 6B, top), we reasoned that HYLS1 promotes the exhaustion of PI(4)P by activating PIPKIγ. If so, then INPP5E deficiency should rescue ciliogenesis in HYLS1-depleted cells by reducing PI(4)P production at the MC () and consequently lessening the need for PIPKIγ activity. Codepletion of INPP5E with HYLS1 completely restored the TTBK2 recruitment, CP110 removal, and ciliogenesis in response to serum starvation (Fig. 6D). On the basis of our PIPKIγ autoinhibition model (Fig. 5F), we constructed a constitutively active, centrosome-targeting PIPKIγ by replacing its CT with the PACT (pericentrin-AKAP450 centrosomal targeting) motif (). As expected, HYLS1-depleted cells expressing PIPKIγ-ΔCT-PACT at the BB exhibited normal TTBK2 recruitment and ciliogenesis after serum starvation (Fig. 6E). Together, our results suggest that HYLS1 is required for the initiation of axoneme elongation by facilitating the recruitment of TTBK2 to the MC. HYLS1 achieves this by activating PIPKIγ at the ciliary base to promote PI(4)P exhaustion after INPP5E is removed from the MC by growth-arresting signals. Moreover, levels of HYLS1 in the whole cell, bound with PIPKIγ, or at the centrosome/BB are not affected by serum starvation (fig. S5, F and G), indicating that HYLS1-mediated activation of centrosomal PIPKIγ is constitutive.

HYLS1 is required for activation of Hh signaling

We showed above that a small fraction of HYLS1 siRNA–treated cells could assemble truncated cilia (Fig. 1D), likely due to the remaining residual HYLS1. Yet, these truncated cilia lack the NPHP module at the TZ (Fig. 2D and fig. S2D). The TZ is an essential functional domain as part of the ciliary gate that controls protein entry and exit (–). Thus, HYLS1 depletion may interrupt ciliary trafficking and/or signaling. To test this possibility, we examined the well-characterized, cilium-dependent Hh pathway (, ). Compared to control cells, the HYLS1-deficient cilia exhibited a significant decrease in Gli3 (glioma-associated oncogene homolog 3) accumulation at the ciliary tip after treatment with the Smoothened agonist (SAG) (Fig. 7A). A few of HYLS1−/− RCTE cells that formed severely truncated cilia showed stronger suppression of SAG-induced Gli3 accumulation at the ciliary tip (fig. S4G). Moreover, depletion of HYLS1 interrupted the SAG-induced Gli1 transcription in RCTE cells (fig. S7C). In mouse embryonic fibroblast (MEF) cells that have the complete Hh pathway, SAG-induced transcription of Gli1 (glioma-associated oncogene homolog 1) and patched homolog 1 (Ptch1) were both suppressed by HYLS1 depletion (Fig. 7B). These results suggest that HYLS1 is necessary for the activation of Hh signaling. Yet, the SAG-induced ciliary entry of Smoothened (fig. S6S) () and ciliary tip accumulation of KIF7 (fig. S6B) (, ) were not affected by HYLS1 depletion. The ciliary level of TULP3 (tubby-like protein 3), a PI(4,5)P2-binding trafficking protein that mediates the ciliary entry of Hh suppressor Gpr161 (G protein-coupled receptor 161) (), also remained the same after HYLS1 depletion (fig. S6C). Consistent with this observation, the level of Gpr161 in resting cilia was comparable with or without HYLS1 (Fig. 7C). However, the HYLS1-depleted cilia were significantly less efficient in removing Gpr161 from cilia after SAG treatment (Fig. 7C). These observations suggest that HYLS1 deficiency suppresses Hh activation by inhibiting the SAG-induced ciliary exit of Gpr161. This effect is different from the inactivation of Hh signaling caused by Inpp5e deletion, which increases ciliary PI(4,5)P2 to promote the ciliary entry of TULP3 and Gpr161 in resting cilia (, ). Depletion of HYLS1 or PIPKIγ did not suppress the increase of TULP3 in Inpp5e-null cilia (fig. S6D), suggesting that the augmented PI(4,5)P2 in the INPP5E-deficient ciliary membrane may not be supplied by PIPKIγ. It is also plausible that the residual HYLS1 or PIPKIγ following siRNA treatment is sufficient to support the ciliary level of PI(4,5)P2 in INPP5E-deficient cells.

Fig. 7

HYLS1 promotes the activation of Hh pathway via NPHP1.

HYLS1 depletion (A to C) and NPHP1 depletion (D to F) both interrupted the Hh signaling. The ciliary tip accumulation of Gli3 (A and D), transcription of Gli1 and Ptch1 (B and E), and ciliary exit of Gpr161 (C and F) induced by SAG were determined. (A, C, D, and F) RCTE cells were transfected with control (siNC), HYLS1 (siHYLS1-O1 and siHYLS1-O2), or NPHP1 (siNPHP1) siRNAs for 48 hours, serum starved for 24 hours with (+SAG) or without (−SAG) 500 nM SAG, and then subjected to IF microscopy. γ-tub, γ-tubulin; Ac-tub, acetylated tubulin. The percentage of cells with Gli3 at cilia tip (A and D) or without Gpr161 in cilia shaft (C and F) was quantified. Scale bar, 2 μm. (B and E) MEF cells were transfected with siNC, siHYLS1 (B), or siNPHP1 (E) for 48 hours, serum starved for 24 hours with or without 500 nM SAG, and then subjected to quantitative polymerase chain reaction. Relative levels of Gli1 or patched homolog 1 (Ptch1) mRNA in each group were normalized against the control group (siNC and −SAG). (G) Overexpression of the kinase-dead PIPKIγ still recruits HYLS1 and NPHP1 to centrosome/BB. RCTE cells were transfected with siNC or human PIPKIγ-specific siRNA (siPIPKIγ) for 24 hours and then transfected with empty vector (mock), HA-tagged, and PACT-fused wild-type (Iγ) or kinase-dead (Iγ-KD) mouse PIPKIγ for 24 hours, followed by 24-hour serum starvation. These cells were subjected to IF microscopy to determine the percentage of cells with HYLS1-positive (top) or NPHP1-positive (bottom) centrosome/BB. All statistical analyses were performed using results from at least three independent experiments and plotted as means ± SEM. N.S., no significant difference. *P < 0.05, **P < 0.01, and ***P < 0.001.

Most HYLS1-depleted cilia showing defective SAG-induced Gli3 accumulation in the ciliary tip (fig. S7A) and Gpr161 removal from cilia (fig. S7B) were NPHP1 negative. NPHP1 depletion (fig. S7, D and E) suppressed SAG-induced Gli3 accumulation (Fig. 7D), transcription of Gli1 and Ptch1 (Fig. 7E), and Gpr161 removal from cilia (Fig. 7F), similar to effects caused by HYLS1 depletion (Fig. 7, A to C). Together, our results suggest that HYLS1 supports the activation of the Hh pathway by ensuring the integration of the NPHP module at the TZ. Both wild-type and kinase-dead PIPKIγ-ΔCT-PACT could recover the loss of HYLS1 and NPHP1 at the ciliary base in PIPKIγ-depleted cilia () to a comparable extent (Fig. 7G). This suggests that the kinase activity of PIPKIγ is dispensable for HYLS1 targeting to the BB and mediating the NPHP1 positioning. Thus, besides facilitating FBF1 assembly, HYLS1 has at least two more fundamental functions in mammalian cells: activating PIPKIγ at the BB and assembling the NPHP module at the TZ, with the former vital for the initiation of axoneme elongation and the latter crucial for the ciliary removal of Gpr161 and the activation of Hh signaling.

A recent study suggested that β-arrestin proteins () are required for the retrograde trafficking of Gpr161 in response to SAG (). SAG stimulated a loss of β-arrestin from the ciliary shaft, but an accumulation of β-arrestin at the ciliary base, in normal cells (Fig. 8A). This β-arrestin translocation was significantly diminished in HYLS1-depleted cilia, as well as in NPHP1-depleted or PIPKIγ-depleted cilia (Fig. 8, A and B). Collectively, our results suggest a functional axis comprising PIPKIγ, HYLS1, and the NPHP module at the proximal compartment of the cilium that is necessary for the SAG-induced removal of β-arrestin from cilia, which mediates the retrograde trafficking of Gpr161 and ensures the activation of Hh signaling. This process appears fulfilled without, or with a minimal amount of, PIPKIγ activity.

Fig. 8

NPHP1, HYLS1, and PIPKIγ are required for the SAG-induced ciliary exit of β-arrestin.

RCTE cells were treated with control (siNC), HYLS1-specific (siHYLS1), NPHP1-specific (siNPHP1), or PIPKIγ-specific (siPIPKIγ) siRNAs for 48 hours and then serum starved for 24 hours with (+SAG) or without (−SAG) SAG. Cells were then subjected to indirect IF microscopy to visualize β-arrestin in cilia. (A) Representative images of control (siNC), HYLS1-depleted (siHYLS1), NPHP1-depleted (siNPHP1), and PIPKIγ-depleted (siPIPKIγ) cells. Scale bar, 2 μm. Relative intensity of β-arrestin in cilia (top) or at cilia base (bottom) was quantified in >100 cells. (B) Cells exhibiting β-arrestin translocation from the ciliary sheath to the ciliary base after SAG treatment was quantified in indicated experimental groups (n > 100). Results from three independent experiments were statistically analyzed and plotted as means ± SEM. **P < 0.01 and ***P < 0.001.

The etiology of the common HLS1 mutation

HLS1, a fetal autosomal recessive ciliopathy, is commonly caused by an aspartic acid to glycine mutation (p.D211G) in HYLS1 (). Despite being reported to mislocalize to the nucleus (), HYLS1D211G localizes properly to centrioles and the BB in nonciliated (Fig. 9A) or ciliated (Fig. 9B) RCTE cells, respectively. HYLS1D211G pulled down PIPKIγ normally (Fig. 9C) and recovered the defective TTBK2 recruitment and ciliogenesis caused by depletion of endogenous HYLS1 (Fig. 9D), indicating that this mutation does not interrupt the function of HYLS1 in activating PIPKIγ. However, unlike its wild-type counterpart, HYLS1D211G failed to replace the endogenous HYLS1 to support the SAG-induced Gli3 accumulation or Gpr161 removal from cilia (Fig. 9E). Likewise, HYLS1D211G could not fully recover NPHP1 at the TZ in HYLS1-depleted cells (Fig. 9F). Our results suggest that the p.D211G mutant of HYLS1 maintains normal ciliogenesis but cannot support the assembly of the NPHP module and TZ integrity, thus causing defective Hh signaling.

Fig. 9

The HLS pathogenic mutant supports ciliogenesis but not the activation of Hh signaling.

(A and B) HYLS1D211G remains normal subcellular locations. RCTE cells were transfected with Flag-tagged HYLS1 (WT) or HYLS1D211G (DG). Before (A) or after (B) 24-hour serum starvation, cells were subjected to IF microscopy to visualize indicated proteins. γ-tub, γ-tubulin; Ac-tub, acetylated-tubulin. Scale bars, 2 μm. (C) HYLS1D211G binds to PIPKIγ normally. HEK293T cells cotransfected with HA-PIPKIγ and Flag-tagged HYLS1 or HYLS1D211G were subjected to co-IP using anti-Flag antibody. Resulted immunoprecipitants were analyzed by immunoblotting using indicated antibodies. (D) HYLS1D211G supports ciliogenesis. RCTE cells expressing empty vector (mock) or RNAi-resistant HYLS1 (WT or DG) were transfected with siNC or siHYLS1-O1 (siHYLS1) for 48 hours, followed by 24-hour serum starvation. The percentage of ciliated cells (left) or TTBK2-positive basal bodies (right) was determined by IF microscopy and quantified in >100 cells. (E) HYLS1D211G is defective to activate Hh signaling. RCTE cells in (D) were serum starved with or without 500 nM SAG for 24 hours. Gli3 accumulation at cilia tip (left) or loss of Gpr161 in cilia shaft (right) after SAG treatment was analyzed by IF microscopy and quantified in >100 cells. Results from >3 independent experiments were statistically analyzed and plotted. (F) HYLS1D211G cannot recruit NPHP1 to the TZ. RCTE cells in (D) were subjected to IF to determine NPHP1-positive TZ (n > 100). (D to F) Results from >3 independent experiments were statistically analyzed and plotted as means ± SEM. *P < 0.05, **P < 0.01, and ***P < 0.001. (G) HYLS1 facilitates axoneme elongation by activating PIPKIγ. (H) Left: SAG induces the β-arrestin–mediated ciliary exit of Gpr161 and the accumulation of Gli3 to cilia tip. Right: There events are interrupted in HYLS1-depleted, NPHP1-mislocalized cells.

DISCUSSION

Here, we demonstrated a previously unidentified PIPKIγ-HYLS1 axis in mammalian cells that ties HYLS1, the protein mutated in lethal HLS (), to both cilia biogenesis and signaling. First, HYLS1 binds to and activates PIPKIγ at the proximal end of centriole/BB and thereby restrains centriole overduplication (fig. S4D) () and regulates ciliogenesis by promoting CP110 removal and axoneme elongation (Fig. 9G) as PIPKIγ. Although ciliogenesis is triggered by growth arrest, the global or centrosomal levels of PIPKIγ or HYLS1 remained comparable before and after serum starvation (fig. S5, F and G). The association between PIPKIγ and HYLS1 appeared unchanged before and after serum starvation (fig. S5F). Yet, INPP5E was removed from the distal end of MC shortly after serum starvation (Fig. 6B). Thus, we suggest that the serum starvation–triggered INPP5E removal from the proximal end of MC is the key event that interrupts the local balance of PI(4)P and PI(4,5)P2 production, whereas HYLS-dependent PIPKIγ activation is constitutive. Alternatively, there may be an unknown PIPKIγ-inhibiting protein that is disassociated from the PIPKIγ-HYLS1 complex after serum starvation. To determine how the ciliary phosphoinositide pathway is regulated spatiotemporally during ciliogenesis, centrosomal proteins that differentially associate with PIPKIγ, HYLS1, or INPP5E before and after serum starvation need to be identified and investigated in future studies.

In addition to regulating the initiation of axoneme elongation, the PIPKIγ-HYLS1 axis plays a unique role in activating Hh signaling by supporting NPHP assembly at the TZ (Fig. 9H). This requires PIPKIγ to recruit HYLS1 to the centriole/BB, but not its kinase activity. HYLS1−/− RCTE cells exhibit severe centriole overduplication (fig. S4D) similar to PIPKIγ-depleted cells () but not hyls-1 null worms (), further supporting the functional connection between mammalian HYLS1 and PIPKIγ. Our data suggest that mammalian HYLS1 has evolved centrosomal/ciliary localization and functions compared to its orthologs in C. elegans (–) and Drosophila (), likely via its unique interaction with PIPKIγ NT, which is not conserved in nematode or fly homolog/ortholog of mammalian PIPKIγ. To be noted, mammalian CP110 also lacks clear functionally conserved counterpart in worm and fly (, ).

Unlike INPP5E that restricts the ciliary level of Hh-pathway suppressor Gpr161 in resting cilia by suppressing PI(4,5)P2 and TULP3 levels in cilia (, ), the PIPKIγ/HYLS1-dependent NPHP module promotes the SAG-induced ciliary exit of Gpr161 by mediating β-arrestin translocation out of cilia (, ). In line with this mechanism, Hh signaling defects were associated with NPHP and JBTS, ciliopathies caused by NPHP1 and other NPHP gene mutations (). We observed that HYLS1D211G, the HLS1-causing HYLS1 mutant, disrupts Hh signaling without affecting ciliogenesis (Fig. 7, D to F). This explains the phenotypic similarity between HLS1 and HLS2 () caused by KIF7 mutations, which also interrupt Hh activation (–), such as polydactyly and cleft palate. Although HYLS1 deficiency interrupted Gpr161 trafficking out of cilia, it did not affect polycystin-2, another ciliary membrane protein mutated in polycystic kidney disease (PKD). This is consistent with the lack of renal cysts in HLS1 fetuses () and the recent finding that cell-autonomous Hh signaling is dispensable for renal cyst progression in PKD mouse models ().

Loss of the lipid kinase activity of PIPKIγ is lethal in various organisms (, –). A kinase-dead mutant of PIPKIγ was identified as pathogenic in the type 3 lethal congenital contractual syndrome (LCCS3), which features lethal skeletal dysplasia and muscle wasting (). LCCS10 in the same phenotypic series was defined as a ciliopathy (). In addition, dynamin 2 () and adenylyl cyclase 6 () mutated in LCCS5 and LCCS8, respectively, play important roles in the context of primary cilia (, , ). To this end, it will be interesting to determine whether PIPKIγ-associated LCCS3 is also a ciliopathy in future studies.

Recent studies demonstrated the importance of subcompartmentalization of unique PI species in the ciliary membrane (). It is noteworthy that many TF and TZ proteins contain lipid-binding C2 or B9 domain (), such as C2CD3 (C2 domain containing protein 3), CC2D2A (coiled-coil and C2 domain-containing 2A), NPHP1, NPHP4, MKS1, and AHI1 (Abelson helper integration site 1) (–). Although it is mostly unclear whether these proteins bind PIs and what the functional consequences are, it was suggested that nematode MKS-5 (mammalian RPGRIP1L ortholog) establishes a TZ subdomain to limit the abundance of ciliary PI(4,5)P2 via its C2 domain (). Our current work, by elucidating the unexpected connection between the ciliopathy protein HYLS1 and PIPKIγ during ciliogenesis and ciliary signaling, sheds light on complex interactions between PI species, PI enzymes, and PI regulators and effectors in the context of primary cilium. However, further studies are needed to determine the composition of this cilium-specific PI signaling pathway and its correlation with ciliopathies. Examination of these mechanistic questions will provide fundamental information to comprehend the development, tissue/organ homeostasis, and the etiology and pathogenesis of ciliopathies, and thus will be a fascinating direction to explore in the future.

MATERIALS AND METHODS

Antibodies and reagents

Rabbit polyclonal antibodies: CEP83 (HPA038161), CEP164 (SAB3500022), and TTBK2 (HPA018113) are from Sigma-Aldrich; CEP290 (ab84870) and CEP135 (ab75005) are from Abcam; hemagglutinin (HA) (51064-2-AP), Arl13b (17711-1-AP), FBF1 (11531-1-AP), TCTN1 (15004-1-AP), SCLT1 (14875-1-AP), MKS1 (16206-1-AP), INPP5E (17797-1-AP), GPR161 (13398-1-AP), MKS5 (55160-1-AP), KIF7 (24693-1-AP), TULP3 (13637-1-AP), NPHP4 (13812-1-AP), IFT140 (17460-1-AP), and IFT88 (13967-1-AP) are from Proteintech; CEP152 (A302-480A) and CP110 (A301-343A) are from Bethyl Laboratories; and polycystin 2 antibodies are provided by the Baltimore PKD Research and Clinical Core Center. Rabbit monoclonal antibody against PIPKIγ was described previously (, ).

Mouse monoclonal antibodies: acetylated α-tubulin (T7451), γ-tubulin (T6557), FLAG (F1804), and HA (H3663) are from Sigma-Aldrich; β-actin (sc-47778), Smoothened (SMO) (sc-166685), and HYLS1 (sc-376721) are from Santa Cruz Biotechnology; β-arrestin is from BD transduction Laboratories (610550); c-Myc is from Invitrogen (VB299757A); polyglutamylated tubulin is from Enzo Life Sciences (ALX-804-885-C100); ODF2 is from Abnova (H00004957-M01); and PI(4)P antibody is from Echelon (Z-P004). Mouse monoclonal NPHP1 antibody (immunoglobulin G1) was generated by Abmart as described previously (). Goat polyclonal antibody: GLI-3 is from R&D Systems (AF3690). SAG (566660) was purchased from Millipore.

For Western blot, primary antibodies were 1:5000 diluted except for β-actin (1:20,000) and HYLS1 (1:500). For immunofluorescence, primary antibodies were 1:500 diluted except for those against NPHP1 (1:50), acetylated α-tubulin (1:2000), and PI(4)P (1:100).

DNA constructs and siRNAs

HA-tagged PIPKIγ-CT, PIPKIγ-NT, PIPKIγΔCT, and PACT-fused PIPKIγΔCT were amplified from our published FL PIPKIγ_i3 complementary DNA (cDNA) () and subcloned into pcDNA3-HA or pcDNA3-myc vectors. His-tagged PIPKIγ in pET28a () was used to create His-PIPKIγΔCT in pET28a by polymerase chain reaction (PCR) and subcloning. The HYLS1 cDNA was obtained from RCTE cells and subcloned into pcDNA3-Flag or pET28a vectors. HYLS1-D211G was generated using a QuikChange site-directed mutagenesis kit (Agilent Technologies). All constructs were verified by DNA sequencing. siRNA SMARTpool targeting human INPP5E (L-020852-00-0005), mouse HYLS1 (L-172366-00-0005), mouse NPHP1 (L-044125-00-0005), and RNA interference (RNAi) negative control were purchased from GE Healthcare Dharmacon. Human PIPKIγ, HYLS1, and NPHP1 siRNA oligonucleotides were obtained from Invitrogen: PIPKIγ-O1, 5′-GCGTGGTCAAGATGCACCTCAAGTT-3′; PIPKIγ-O2, 5′-GCTACTACATGAACCTCAACCAGAA-3′; HYLS1-O1, 5′-GAGAAGGAATGGGCTCTCCAGCTTA-3′; HYLS1-O2, 5′-CCTATCATTGGTCTTTCCTAGCTAT-3′; and NPHP1 siRNA, 5′-GATGGTTGGTGGATAGCTA-3′.

Cell culture and transfection

Human embryonic kidney (HEK) 293T, IMCD3, RCTE, and hTERT-RPE-1 cells were purchased from ATCC (American Type Culture Collection). Inpp5e+/− and Inpp5e−/− MEF cells were generated as reported and generously shared by J. Reiter (University of California, San Francisco) (). HEK293T, IMCD3, and MEF cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) containing 10% FBS. RCTE and hTERT-RPE-1 cells were cultured in DMEM/F-12 containing 10% FBS. IMCD3 and MEF cells were cultured in DMEM containing 10% FBS. For plasmid transfection, X-tremeGENE 9 (Roche) was used following the manufacturer’s manual. Lipofectamine RNAiMAX (Invitrogen) was used for siRNA transfection following the manufacturer’s manual. Typically, 48 hours after initial transfection, cells were collected for further analysis by immunofluorescence or Western blot.

Immunofluorescence microscopy

For indirect immunofluorescence labeling of centrosomal or BB proteins, cells were fixed with prechilled methanol at −20°C for 20 min. For ciliary membrane proteins, cells were fixed with 4% paraformaldehyde at room temperature for 20 min and then permeabilized in 0.2% Triton X-100 in phosphate-buffered saline (PBS) for 10 min. Cells were then blocked in 3% bovine serum albumin and incubated with appropriate primary antibodies overnight at 4°C. Fluorescence images were acquired using Nikon TE2000-U with NIS-Elements (Nikon). Immunofluorescence staining of PI(4)P was performed as described (). 3D-SIM analyses were performed using an ELYRA Super-Resolution Microscopy system (Zeiss) equipped with an alpha “Plan-Apochromat” 100×/1.46 Oil DIC oil immersion objective. An Andor iXon 885 electron-multiplying charge-coupled device camera was used to acquire raw images. Sections were acquired at 0.125-mm z-steps. Color channels were aligned using alignment parameter from control measurements with 0.5-μm-diameter multispectral fluorescent beads (Zeiss). Structured illumination reconstruction and image processing were performed with the ZEN software package (Zeiss). Final image processing was done using Adobe Photoshop (Adobe).

Immunoprecipitation and protein pulldown

HEK293T cells were lysed at 4°C for 1 hour using IP buffer [20 mM Hepes-KOH (pH 7.2), 10 mM KCl, 1.5 mM MgCl2, 1 mM EDTA, 1 mM EGTA, 150 mM NaCl, and 0.5% NP-40], with 1× Complete Protease Inhibitor Cocktail (Roche) and 1× PhosSTOP Phosphatase Inhibitor Cocktail (Roche). Cell lysates were cleaned by centrifugation (12,000 rpm, 10 min) at 4°C, incubated with antibodies and protein A–conjugated Sepharose 4B resins at 4°C for 4 hours or overnight. The protein A beads were pelleted by centrifugation and washed with IP buffer for three times. Proteins pulled down by protein A beads were eluted using 1× SDS loading buffer and then analyzed by Western blot. Constructs in pET28 expression vectors were transformed into BL21(DE3) competent cells (Novagen) and proteins were purified using His-Bind resin following the manufacturer’s instructions (Novagen). Purified His-HYLS1 and His-PIPKIγ (0.5 μg each) were incubated in PBS–0.5% Triton X-100 at 4°C for 2 hours. Then, HYLS1 antibody and protein A beads were added to pull down HYLS1 and proteins associated with HYLS1.

PIPKIγ kinase assay

PIPK assay was performed as described previously (). In brief, the reaction occurred in 40 μl of kinase buffer containing 10 mM MgCl2, 0.5 mM EDTA, 50 mM tris (pH 7.5), and 25 μM diC16-PI4P (P-4016, Echelon). Immediately after adding 10 μl of adenosine 5′-triphosphate (ATP) solutions (50 μM cold ATP and 5 μCi of P-33 ATP), the reaction was incubated at 25°C for 5 min and then terminated using 100 μl of 1 M HCL and 200 μl of chloroform/methanol solution (volume ratio, 1:1). The organic portion was collected, extracted by 80 μl of chloroform:1 M HCL (volume ratio 1:1) and then subjected to thin-layer chromatography plates (Thermo Fisher Scientific) using a solvent containing chloroform, methanol, ammonium hydroxide, and water (volume ratio, 90:90:7:22). Labeled phosphoinositides were visualized using autoradiography.

Size exclusion chromatography

His-fusion proteins of HYLS1 and PIPKIγ were purified following standard protocols. The active fractions were further purified through ion exchange column and then subjected to the gel filtration using the Superose 6 10/300 GL (GE Healthcare) in PBS. Fractions derived from the column were analyzed by immunoblotting.

CRISPR-Cas9 gene editing

HYLS1 guide RNA (gRNA) was designed using an online tool (https://zlab.bio/guide-design-resources) and subcloned into pSpCas9(BB)-2A-GFP (px458), in which a green fluorescent protein (GFP) was fused to Cas9. RCTE cells were transfected with the px458-HYLS1 gRNA construct for 48 hours and then subjected to flow cytometry to sort single GFP-positive cells. Single clones were expanded for further confirmation of the on-target cleavage in HYLS1 using genomic PCR followed by Sanger sequencing. gRNA sequence was as follows: GCTGCCAACATTCGTTCTTC.

Quantitative PCR

Total RNA was isolated from RCTE or Inpp5e+/− MEF cells using the TRIzol reagent (Invitrogen, #15596-026). cDNA was generated using the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems, #00351115). Quantitative PCR was performed using the 2× All-in-One qPCR Mix (GeneCopoeia, #HmiRQP2641) on a CFX384 Real-Time system (Bio-Rad). Relative expression levels were calculated using the 2−ΔΔCt method and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) as reference gene. Forward and reverse primers are as follows: human/mouse GAPDH-forward, 5′-TCCTGCACCACCAACTGCTT-3′; human/mouse GAPDH-reverse, 5′-GTCTTCTGGGTGGCAGTGAT-3′; human Gli1-forward, 5′-CCATTCCAATGAGAAGCCGT-3′; human Gli1-reverse, 5′-GACCATGCACTGTCTTGACA-3′; mouse Gli1-forward, 5′-GGTGCTGCCTATAGCCAGTGTCCTC-3′; mouse Gli1-reverse, 5′-GTGCCAATCCGGTGGAGTCAGACCC-3′; mouse Ptch1-forward, 5′-CTCTGGAGCAGATTTCCAAGG-3′; and mouse Ptch1-reverse, 5′-TGCCGCAGTTCTTTTGAATG-3′.

Statistical analyses

Results are presented as means ± SEM, as specified in each figure legend. Statistical comparisons were made by using an unpaired two-tailed t test with GraphPad Prism (Graphpad, La Jolla, CA). P values <0.05 were considered supportive to statistically significant differences.