A CRISPR-based screen for Hedgehog signaling provides insights into ciliary function and ciliopathies

Abstract

Primary cilia organize Hedgehog signaling and shape embryonic development, and their dysregulation is the unifying cause of ciliopathies. We conducted a functional genomic screen for Hedgehog signaling by engineering antibiotic-based selection of Hedgehog-responsive cells and applying genome-wide CRISPR-mediated gene disruption. The screen can robustly identify factors required for ciliary signaling with few false positives or false negatives. Characterization of hit genes uncovered novel components of several ciliary structures, including a protein complex that contains δ-tubulin and ε-tubulin and is required for centriole maintenance. The screen also provides an unbiased tool for classifying ciliopathies and showed that many congenital heart disorders are caused by loss of ciliary signaling. Collectively, our study enables a systematic analysis of ciliary function and of ciliopathies, and also defines a versatile platform for dissecting signaling pathways through CRISPR-based screening.

Figure 1. Development of a Hedgehog pathway reporter-based screening strategy

a) A transcriptional reporter combining 8 copies of the GLI binding sequence (Gli-BS) with a minimal promoter (P_min) to convert Hh signals into blasticidin resistance. b) Blasticidin resistance was assayed across a range of concentrations in stimulated (+ShhN) and unstimulated 3T3-[Shh-BlastR;Cas9] cells. Representative curves of 5 independent experiments performed in duplicate. c) Overview of the Hh pathway, with key negative and positive regulators shown in red and green, respectively (left). Effects of control sgRNAs on blasticidin resistance in stimulated and unstimulated 3T3-[Shh-BlastR;Cas9] cells (right). Bars show mean inhibitory concentration 50 (IC₅₀) values and circles show IC₅₀ values from N = 2 (for gene-targeting sgRNAs) or 5 (for no sgRNA and negative control (Ctrl) sgRNA) independent experiments performed in duplicate. d) Overview of the screening strategy. Cells receiving a negative control sgRNA, a positive regulator-targeting sgRNA, and a negative regulator-targeting sgRNA are shaded grey, green, and red, respectively. e) Flow cytometry histograms of cell mixtures showing the fraction of GFP positive (Smo sgRNA-2, green) cells either in the absence of selection (left) or after two rounds of signaling and selection (right). Representative results from three independent experiments. f) Quantification of cell depletion as in (e).

Figure 2. Overview of genome-wide screen results

a) Scatter plot showing log₂ of normalized sgRNA counts in selected versus unselected cell pools, with sgRNAs targeting select genes highlighted. b) Volcano plot of casTLE P values versus effect sizes for all genes (after filtering; see Methods), with select Hh pathway components highlighted. Green area indicates P-value cutoff corresponding to 10% false discovery rate (FDR); combined green and yellow areas indicate 20% FDR, with the number of genes in each area indicated. c) Schematic illustration of a primary cilium, with known structural features and select protein products of hit genes shown. Proteins shown are grouped by protein complex membership or localization, with select newly identified hits highlighted in bold, italic font. d) For the indicated categories, proteins encoded by hit genes identified are listed in order of statistical confidence. In addition to select gene ontology terms enriched among screen hits, the top 100 hits not otherwise listed in panels a–c are shown.

Figure 3. Evaluation of screen performance

a) Assessment of screen performance using 130 positive and 1386 negative reference genes, as determined by precision-recall analysis (left) and ROC curve (right), with the area under each curve (AUC) shown. Dashed lines indicate performance of a random classification model. b) Analysis of hit gene detection for select gene categories (N = number of genes in each category), with the fraction of hits detected at 10% or 20% FDR, not detected, or not determined shown; see Supplementary Table 3 for details. The NPHP category includes genes mutated exclusively in NPHP and not other ciliopathies. Abbreviations: SRTD (short rib thoracic dysplasia), JATD (Jeune asphyxiating thoracic dysplasia), OFD (Oral-Facial-Digital Syndrome). c) Hit gene identification is compared for the indicated datasets. Pie charts show the fraction of N=88 genes detected as hits across all genes included in part (b), except the NPHP-specific category; see Supplementary Fig. 3a for detail among individual categories.

Figure 4. Characterization of FAM92A and TTC23 as transition zone and EvC zone components

a) Induction of Hh pathway luciferase reporter is shown for cells transduced with the indicated sgRNAs and transfected with plasmids encoding Fam92a-3xFLAG (Fam92a) or GFP-FKBP (GFP). Cells were untreated or stimulated with SAG. Bars show mean of 4 replicate measurements (circles); one of two representative experiments. b) Analysis of cilia in 3T3-[Shh-BlastR;Cas9] cells transduced with the indicated sgRNAs. Bars show mean percentage of ciliated cells; dots show ciliated percentage in each of two independent experiments (>200 cells analyzed per datapoint). Scale bar: 5 μm c) Mass spectrometry analysis of FAM92A-associated proteins purified from IMCD3 cells. The normalized spectral abundance factor (NSAF), the percent of each protein covered by identified peptides, and the percentile rank of the corresponding gene in the screen are indicated. d) FAM92A localizes to the transition zone of IMCD3 cells, distal to centrioles (γ-tubulin) and proximal to the ciliary membrane (ARL13B). One of two independent experiments (five fields of view each). Scale bars: 5 μm and 1 μm (insets). e) Mass spectrometry analysis of TTC23-associated proteins purified from IMCD3 cells. f) TTC23-LAP co-localizes with IQCE, distal to FAM92A, in IMCD3 cells. Line plots show normalized intensity along the length of the cilium; tick marks are 1 μm intervals. Representative images are shown from two independent experiments (five fields of view each). Scale bars: 5 μm and 1 μm (insets). g) The median and interquartile range of ciliary IQCE levels are shown for cells transduced with the indicated sgRNAs; one of two independent experiments. h) Ciliary TTC23-LAP and IQCE signals were analyzed following introduction of Iqce-targeting or control (Ctrl) siRNAs. The median, interquartile range (box boundaries), 10–90% percentile range (whiskers), and outliers are plotted for N=390 (Ctrl) and N=300 (Iqce) cilia. One of two (IQCE) or four (GFP) replicate experiments. Scale bars: 1 μm.

Figure 5. Insights into ciliopathies from previously uncharacterized screen hits

a) TXNDC15, with the transmembrane domain (TMD), thioredoxin domain, and MKS-associated mutation indicated (top). Luciferase reporter levels were measured for cells transduced with the indicated sgRNAs and transfected with plasmids encoding GFP-FKBP (GFP), wildtype Txndc15 (Tx-WT), or mutant Txndc15 (Tx-mut). Bars show mean of 3 replicates (circles); one of two representative experiments. b) Cilia were analyzed in 3T3-[Shh-BlastR;Cas9] cells transduced with the indicated sgRNAs. Bars in graph show percentage of cells with cilia that are normal, distorted, or absent. Each bar per condition represents an independent experiments with >200 cells counted. Scale bars: 5 μm and 1 μm (insets). c) Cilia were analyzed in 3T3-[Shh-BlastR;Cas9] cells transduced with the indicated sgRNAs. Representative images are shown at top. At bottom, the median cilium length, interquartile range (box boundaries), 10–90% percentile range (whiskers), and outliers are plotted. One of three independent experiments. Scale bars: 5 μm. d) Analysis of ARMC9-FLAG localization relative to centrioles (ninein) and ciliary membrane (ARL13B) is shown for IMCD3 cells treated as indicated. Scale bar: 1 μm. e) ARMC-FLAG intensity along the length of the cilium from base (position 0) to tip (position 1.0) was measured for IMCD3 cells treated as indicated. The mean and standard deviation are plotted after normalizing the total intensity in each cilium to 1.0; one of three representative experiments. f) Fluorescence intensity of GLI3 at the cilium tip was measured for the indicated cells in the presence or absence of ShhN. Mean and standard error of the mean are shown for each of N=3 independent experiments (at least 250 cilia analyzed per condition). g) Table showing select clinical features in canonical ciliopathies and their observation in the context of specific mutations and syndromes. Colors indicate high (red), moderate (orange) and low (yellow) prevalence.

Figure 6. TEDC1 and TEDC2 form a tetrameric complex with δ- and ε-tubulins

a) IMCD3 cells stably expressing TEDC2-LAP were immunostained with antibodies to centrin2 and γ-tubulin to visualize centrioles. Scale bar: 5 μm (2 μm for insets). Representative images are shown for three independent experiments. b) Mass spectrometry analysis of TEDC2-associated proteins purified from IMCD3 cells reveals TEDC1, ε-tubulin, δ-tubulin in nearly stoichiometric amounts, as well as α/β-tubulin and CENPJ. For each protein, the normalized spectral abundance factor (NSAF), the percent of the protein covered by identified peptides, and the percentile rank of the corresponding gene in the screen dataset are indicated. c) Binding of TEDC1 and TEDC2 was assessed via co-immunoprecipitations performed in HEK293T cells transfected with plasmids encoding the indicated proteins. Recovered proteins were analyzed by Western blot. Representative blots are shown for two independent experiments. See also Supplementary Fig. 6b and Supplementary Fig. 8.

Figure 7. The TED complex is required for centriole stability

a) Cells transduced with the indicated sgRNAs were stained with antibodies to ninein (centrioles) and polyglutamylated tubulin (GT335, centrioles and cilia). Scale bar: 5 μm. One of three representative experiments. b) Hierarchical clustering of relative growth scores across the indicated cell lines reveals that TEDC1, TEDC2, and TUBE1 share a similar pattern of relative fitness. Blue and red shading indicates decreased and increased proliferation relative to the average behavior across all cell lines. c) For cells transduced with the indicated sgRNAs, centrioles were visualized by staining with antibodies to centrin3 and γ-tubulin. Insets show centrin3 staining in mitotic cells, marked by yellow arrowheads. Scale bars: 5 μm (2 μm for insets). See also Supplementary Fig. 6h. One of three representative experiments. d) Centrioles marked by centrin3 and γ-tubulin were counted in cells at the indicated cell cycle stages. Statistically significant differences in centriole counts are shown for select conditions (*, P < 1×10⁻⁶; **, P < 1×10⁻¹⁰; ***, P < 1×10⁻⁶⁰, determined by two-sided Fisher’s exact test).