Genome-scale requirements for dynein-based trafficking revealed by a high-content arrayed CRISPR screen

The cytoplasmic dynein-1 (dynein) motor plays a key role in cellular organisation by transporting a wide variety of cellular constituents towards the minus ends of microtubules. However, relatively little is known about how the biosynthesis, assembly and functional diversity of the motor is orchestrated. To address this issue, we have conducted an arrayed CRISPR loss-of-function screen in human cells using the distribution of dynein-tethered peroxisomes and early endosomes as readouts. From a guide RNA library targeting 18,253 genes, 195 validated hits were recovered and parsed into those impacting multiple dynein cargoes and those whose effects are restricted to a subset of cargoes. Clustering of high-dimensional phenotypic fingerprints generated from multiplexed images revealed co-functional genes involved in many cellular processes, including several candidate novel regulators of core dynein functions. Mechanistic analysis of one of these proteins, the RNA-binding protein SUGP1, provides evidence that it promotes cargo trafficking by sustaining functional expression of the dynein activator LIS1. Our dataset represents a rich source of new hypotheses for investigating microtubule-based transport, as well as several other aspects of cellular organisation that were captured by our high-content imaging.


Supplementary figure 3. Assay development for rapamycin-induced relocalisation of peroxisomes. A)
Representative images of results of applying a spot detection mask on raw images of rapamycin-treated U-2 OS PEX cells. Scale bar, 50 µm. B) Optimisation of rapamycin concentration and treatment duration in the peroxisome relocalisation assay. Note that, with these raw values, the number of GFP spots increases with perinuclear clustering because discrete puncta are otherwise relatively uncommon due to the dim signal, whereas RFP spot number decreases with perinuclear clustering as signals from multiple dispersed puncta coalesce at the MTOC. Data points represent mean aggregation at well level (minimum of 100 cells from at least three wells analysed per condition). Error bars, S.D.. A 2.5-h treatment with 2 nM rapamycin was selected for the genome-wide screen. C) Impact of number of seeded cells per well (384-well plate format) on the number of GFP-BICD2N-FRB and PTS-RFP-FKBP spots. Cells were seeded for 72 h prior to treatment with rapamycin (2 nM). Datapoints represent mean aggregation at well level (minimum of three wells analysed per condition). Error bars, S.D.. 1500 cells per well were seeded for the genomewide screen.
Supplementary figure 4. Assay scaling for high-throughput editing. A) Representative low magnification view of 384-well plate regions showing consistent editing in U-2 OS PEX cells treated with crLIS1, crDYNC1H1, and crPLK1. crLIS1 and crDYNC1H1 activity was assessed by immunostaining for the target proteins, whereas activity of crPLK1 was read out by a reduction in cell number (revealed by Hoechst staining). B -D) Violin plots (median, bold line; first/third quartile, dashed lines) of frequency of cells depleted for LIS1 (B) or DYNC1H1 (C), or the number of cells (D), after transfection with crLIS1, crDYNC1H1 or crPLK1, respectively. Datapoints represent mean aggregation at the well level (minimum of 100 cells from at least four wells analysed) for three individual plates (P).
Supplementary figure 5. Quality control of assay endpoints used in the genome-wide screen. Heatmap displaying rZ′ score of features used for hit calling from the screen data. All endpoints were normalised by NTC vs crLIS1 except for cell number, which was normalised by NTC vs crPLK1. The displayed rZ' scores represent the corrected rZ' for the individual features for the entire screening set. See Supplementary table 1 for the complete dataset.
Supplementary figure 6. Additional genome-wide screen endpoints. A) Effects of arrayed library crRNAs on cell viability and comparison to results from previous cell viability studies. Scatter plot of library results and corresponding violin plot of controls (median, bold dashed line; first/third quartile, dashed lines; colour code in key at bottom of figure) of proportion of inviable cells (gated based on nuclear morphology of NTC cells). Data points represent normalised values based on the neutral control (NTC, 0) and lethal editing control (crPLK1, 100). Dashed line on the y-axis represents 2.5*S.D. of crLIS1, the threshold for calling lethal crRNAs. Library copies of crPLK1 and crLIS1 are labelled with (L). The genes targeted by the lethal crRNAs were cross-referenced with their gene essentiality categorisation from the Cancer Dependency Map project (DepMap 22Q2 Public+Score; https://depmap.org/portal): 'common essential genes' are classed as essential for growth and survival in at least 90% of cancer cell lines; 'essential for U-2 OS' genes are those classed as not essential across multiple cell lines but essential in U-2 OS cells; 'non-essential' genes are those not identified as essential in a panel of sensitive cell lines; 'not screened' genes are those that are not represented in the DepMap screening dataset. Numbers in parentheses refer to percentage of essential genes in our dataset that were found in each Depmap category. B) Effects of arrayed library crRNAs on micronuclei incidence. Scatter plot of library results and corresponding violin plot of controls (median, bold dashed line; first/third quartile, dashed lines; colour code in key at bottom of figure) for number of micronuclei per cell (x-axis) and number of cells containing micronuclei (y-axis). Data points represent rZ normalisation (central reference = NTC). Dashed lines on the x-and y-axes represent 2.5*S.D. of crLIS1, the threshold for calling crRNAs that cause micronuclei abnormalities. Library copy of crPLK1 is labelled with (L). Labelled genes were functionally enriched (FDR ≤ 0.005) for gene ontology terms associated with regulation of chromosome segregation, including, 'nuclear division', 'mitotic sister chromatid segregation' and 'nuclear chromosome segregation' (http://bioinformatics.sdstate.edu/go/). C) Effects of arrayed library crRNAs on PTS-RFP-FKBP texture and number of PTS-RFP-FKBP and GFP-BICD2N-FRB spots. Scatterplot of library results and corresponding violin plots of controls (median, bold dashed line; first/third quartile, dashed lines; colour code in key at bottom of figure) of normalised values based on the neutral control (NTC, 0) and inhibition (crLIS1, -100). Dashed lines on both axes represent ± 2.5*S.D. of NTC and arrayed library, the threshold for hit calling. Both endpoints were generated from a linear discriminant analysis; PTS-RFP-FKBP texture was generated from six textural features based on filtered images, while 'GFP/RFP spots' was generated by two features (total number of GFP-BICD2N-FRB and PTS-RFP-FKBP spots). Core components of the dynein complex and dynactin complex that met the threshold for hit calling for either endpoint are labelled in purple and teal text, respectively. Library copy of crLIS1 is labelled with (L). Known peroxisome biogenesis genes are labelled in blue text (note that there is a duplicate of the PEX12 crRNA pool in the library). crPAFAH1B2 and crDNM1L are labelled as examples of crRNAs that affect PTS-RFP-FKBP texture in the opposite way to crRNAs against dynein-dynactin components.

Supplementary figure 12. Readouts of the orthogonal cargo localisation screen. A, B) Bar graphs showing
readouts generated from the screen with VBC-generated crRNAs. Quantification of (A) localisation ratio (perinuclear vs peripheral) and total number for GFP-BICD2N-FRB and PTS-RFP-FKBP spots in U-2 OS PEX cells treated with rapamycin and (B) localisation ratio (perinuclear vs peripheral) of EEA1, TGN46 or LAMP1 spots in unmodified U-2 OS cells. '(V)' and '(D)' indicate crRNAs synthesised based on the VBC score or from the original Discovery set, respectively. Data points represent mean aggregation from at least three independent experiments (minimum of 100 cells analysed per well, four wells analysed per condition). EEA1 ratio values were log-transformed for normal distribution. Bold lettering indicates crRNAs that were novel components of the 'dynein-dynactin' cluster of phenotypic profiles. Error bars signify S.D.. *p<0.05, **p<0.01, ***p<0.001 (one-way ANOVA with Dunnett's multiple comparison against NTC). C) Representative images of LAMP1 staining in U-2 OS cells treated with the indicated crRNAs. Scale bar, 20 µm. FAM86B2 and RNF183 crRNAs. A, B) Indel distribution derived from sequences of target regions in unmodified U-2 OS cells transfected with VBC-derived FAM86B2 (A) or RNF183 (B) crRNAs. Corresponding sequences from cells transfected with NTC crRNAs were used as a reference. Insufficient targeting of crFAM86B2 pool may be due to all four crRNAs within the pool targeting overlapping regions, and therefore competing with each other. Efficiency scores are out of 100. C) Incomplete targeting with the crRNF183 pool in U-2 OS cells may be associated with an overlapping target region and/or spanning of a common polymorphism (rs3750534) within a target sequence for crRNF183 (V) #3 and crRNF183 (V) #4, which have particularly low activity. Supplementary figure 16. Supplementary data for differential splicing analysis. A, B) Venn diagrams showing overlap of genes that undergo differential splicing (A) and differential splicing events (B) in the datasets, as determined with rMATs (note that some genes have more than one differential splicing event). Threshold for classifying an event as differential was: absolute IncLevelDifference ≥ 0.2, total read count (inclusion count + skipping count) ≥ 10 and FDR ≤ 0.05. See Supplementary table 9 for full results. C) Classes of alternative splicing events common to both comparisons (i.e. NTC vs crSUGP1 or crXCR1 vs crSUGP1), as identified by rMATS. Blue and green represent events that were enriched in control and crSUGP1 samples, respectively. rMATS reports five splicing categories: i) alternative 3' splice sites (A3SS); ii) alternative 5' splice sites (A5SS); iii) mutually exclusive exons (MXE); iv) retained introns (RI), and v) skipped exons (SE). The A3SS and A5SS events involve the splicing together of two exons separated by a single intron. For A3SS events, alternative splicing causes a downstream exon to extend partially into neighbouring intronic sequence. A5SS is defined by an alternative splicing event causing an upstream exon to extend partially into the adjoining intron. Mutually exclusive exons describe the splicing of adjacent exons (separated by a single intron) in which one exon is retained but the other is excluded, or vice versa. The graph reports MXE events in which the upstream exon was selected. Events classified as RI are those in which an intron is not spliced out and hence is retained in the mature transcript. SE denotes splicing events in which an exon is skipped over and not included in the processed RNA molecule. D) Venn diagrams showing overlap of differential 3′-end usage events in the datasets, as determined by LABRAT. LABRAT quantifies alternative polyadenylation (APA) sites and reports upstream or downstream shifts in the usage of those sites for each gene as compared to the control (see Supplementary table 10 for full results). The threshold for classifying an event as differential was Dψ ≥ 0.05 and FDR ≤ 0.05.