A genome-scale RNAi screen for genetic interactors of the dynein co-factor nud-2 in Caenorhabditis elegans

Cytoplasmic dynein 1 (dynein) is the predominant microtubule minus end-directed motor in animals and participates in a wide range of cellular processes, including membrane trafficking, nuclear migration, and cell division. Dynein's functional diversity depends on co-factors that regulate its subcellular localization, interaction with cargo, and motor activity. The ubiquitous co-factor nuclear distribution gene E (NudE) is implicated in many of dynein's functions, and mutations in NudE cause the brain developmental disease microcephaly. To identify genetic interactors of the Caenorhabditis elegans NudE homolog nud-2, we performed a genome-wide RNAi screen with the null allele nud-2(ok949), which compromises dynein function but leaves animals viable and fertile. Using bacterial feeding to deliver dsRNAs in a 96-well liquid format and a semi-automated fluorescence microscopy approach for counting parents and progeny, we screened 19762 bacterial clones and identified 38 genes whose inhibition caused enhanced lethality in nud-2(ok949) relative to the nud-2(+) control. Further study of these genes, many of which participate in cell division, promises to provide insight into the function and regulation of dynein.


Background & Summary
Animal cells rely on molecular motor proteins to generate forces and transport cargo along cytoskeletal tracks. Microtubule-based transport is critical for a wide range of cellular and developmental processes, and its dysfunction is recognized as an important contributor to neurological disease 1 . The mega-dalton motor complex cytoplasmic dynein 1 (dynein) is involved in virtually all processes that require microtubule minus end-directed motility, including intracellular trafficking, organelle positioning, and the organization of microtubule arrays 2 . During cell division, dynein is implicated in chromosome movement, assembly and positioning of the bipolar spindle, and the control of spindle assembly checkpoint signalling 3 . Dynein's functional diversity is facilitated by co-factors and cargo-specific adaptor proteins that regulate localization and activity of the motor 4,5 . For many dynein-dependent processes, the inventory of participating adaptors and regulators is likely incomplete. An important ubiquitous dynein co-factor is the conserved nuclear distribution gene E (NudE), which was first discovered in filamentous fungi 6,7 . NudE and the related protein NudE-like (Nudel) directly bind dynein and the dynein co-factor Lis1. NudE/Nudel and Lis1 contribute to many cellular functions of dynein, and mutant NudE causes the brain developmental disease microcephaly 8,9 . On the molecular level, NudE/Nudel help tether Lis1 to dynein to regulate force production of the motor and mediate interactions between dynein and cargo 4,5 (Fig. 1a). The nematode Caenorhabditis elegans possesses a single NudE/Nudel gene, nud-2, and animals in which nud-2 is inhibited by RNAi-mediated knockdown or by the genetic deletion nud-2(ok949) display phenotypes consistent with compromised dynein function [10][11][12] . The nud-2(ok949) deletion removes 87 % of the open reading frame and is therefore likely a null allele 10 , yet viability and fertility of mutant animals is only mildly affected. Importantly, reducing LIS-1 levels in nud-2(ok949) embryos results in enhanced lethality relative to nud-2(+) controls (Fig. 1b, c), suggesting that nud-2(ok949) represents a sensitized background that can be exploited for the discovery of dynein pathway genes. We therefore implemented an RNAi-based genetic interactor screen by targeting 87 % of annotated C. elegans open reading frames in nud-2(ok949) and nud-2(+) control animals using the bacterial feeding library from the Ahringer group 13 , adapted to a 96-well liquid format (Fig. 1d). The primary screen consisted of two technical replicates per bacterial clone for a total of 19762 clones. Synchronized L1 larvae were allowed to grow and have progeny by feeding on the bacteria for 92 h. Animals were marked with green fluorescent protein (GFP) expressed in the pharynx (myo-2::gfp), which facilitated automated counting of parents and larval progeny on a fluorescence microscopy platform at the end of RNAi treatment (Fig. 1e, f). The number of progeny per parent, termed reproductive fitness, could then be calculated for each bacterial clone. Genes that when targeted with dsRNA decreased reproductive fitness in nud-2(ok949) animals at least two-fold relative to nud-2(+) animals were chosen for secondary screening (1077 clones) (Fig. 2a-e). Secondary screening was more stringent, consisting of 3 biological replicates, each carried out in 2 technical replicates (Fig. 2f). We present a final list of 38 genetic interactors ( Fig. 2g; Table 1). The list includes 13 genes whose functions relate to the microtubule cytoskeleton, and 4 genes that code for known binding partners of dynein. Strikingly, 19 out of the 38 genes are implicated in embryonic cell division, where dynein is known to have multiple essential functions 14 . Further study of the nud-2 interacting gene set promises to provide insight into the regulatory mechanisms that underlie dynein's functional diversity. Furthermore, reanalysis of the screening data using different cutoffs may extend the list of putative dynein pathway genes. Finally, the data on reproductive fitness in the control strain may serve as a useful reference for other genome-wide RNAi screens.

Worm growth and RNA interference
The genome-wide RNAi screen was performed using a modified version of the screening protocol in the 96-well liquid format developed by Lehner et al. 16 , and an automated screening platform was implemented based on the strategy previously reported by Maia et al. 15 (Fig. 1d). Bacteria from the Ahringer library glycerol stock 17 (Source BioScience) harboring expression vectors for gene-specific RNAi were transferred into 100 μL of LB medium (supplemented with 100 μg/mL ampicillin and 12.5 μg/mL tetracycline hydrochloride) in 96-well plates (#82.1582.001, Sarstedt) using a 96-pin replicator (Boekel) and grown overnight. The first column of pins was removed from the replicator to leave the first column of each plate free for inclusion of positive and negative controls. The negative controls were empty dsRNA vector pL4440 and hil-5 dsRNA vector, which exhibited 10-20 % embryonic lethality in the nud-2 (ok949) background. plk-1 dsRNA vector, which causes 100 % embryonic lethality, was used as a control for RNAi efficiency, and lis-1 dsRNA vector was used as a positive control for enhanced lethality in the nud-2(ok949) background (Fig. 1b, c). 5 μL of the saturated bacterial cultures were used to inoculate 400 μL of LB (with ampicillin and tetracycline) in a deep-well 96-well plate (#260251, Thermo Scientific  Graph shows the number of larval progeny per adult, termed reproductive fitness (RF), in the nud-2(+) control and the null allele nud-2(ok949), after targeting genes with dsRNA using the bacterial feeding method in a 96well liquid format. Empty dsRNA vector and hil-5(RNAi) are negative controls, plk-1(RNAi) is a control for the efficiency of RNAi, and lis-1(RNAi) is the positive control that results in enhanced lethality in nud-2(ok949) relative to nud-2(+). Error bars represent the SEM with a 95 % confidence interval, and n denotes the number of adults whose progeny were counted. (c) The RF ratio, defined as the RF of nud-2(ok949) divided by the RF of nud-2(+) for the conditions shown in (b). Error bars represent the SEM with a 95 % confidence interval.    92 h, which was sufficient time for the L1 larvae to grow until adulthood, lay eggs, and for these eggs to develop into larvae. Each 96-well library plate with bacteria was assayed in two technical replicates per strain.

Image acquisition and automated counting of parents and progeny
Prior to image acquisition, 70 μL of worm paralyzing solution (50 μL complete S Medium, 20 μL of 0.7 % tricaine / 0.07 % tetramisole) was added to all wells. Brightfield and fluorescence images of each well were acquired with the IN Cell Analyzer 2000 (GE Healthcare) using a 2x / 0.1 NA Plan Fluor objective (Nikon). This allowed the entire well to be captured in a single image (Fig. 1f). Automated counting of adult worms (parents, larger pharynxes) and larvae (progeny, smaller pharynxes) was performed using the IN Cell Investigator Developer Toolbox software (GE Healthcare), and visual data inspection was performed with Spotfire DecisionSite (GE Healthcare) (Fig. 1f). Automated counting involved two steps: 1) recognition of the pharynx by detection of objects with an area larger than 0.383 μm 2 , a mass (sum of all pixels within the shape) smaller than 300000 pixels, and a perimeter smaller than 1 mm; 2) distinction between larvae and adults by calculating the total intensity of pixels inside the object and applying a threshold. Only wells with 3-13 parent worms were considered for further analysis, as the number of progeny correlates linearly with the number of adults within this range (Fig. 1e).

Data analysis and hit selection
ScreenSifter software was used to visualize and analyze the data 18 . Reproductive Fitness (RF) was defined as the number of progeny per parent 15 . The RF ratio for each dsRNA vector was calculated by dividing the RF of the nud-2(ok949) strain (GCP68) by the RF of the nud-2(+) control strain (GCP95) (Fig. 1b, c). After genome-wide screening with two technical replicates per strain (Data Record 1), a threshold for hit selection was defined based on the RF ratios for the positive control (lis-1 dsRNA vector), which was included in each 96-well plate: threshold ¼ μ p þ 2σ p , where μ p is the mean and σ p is the standard deviation of the positive control RF ratios (Fig. 2c, d). The resulting threshold RF ratio was 0.43. dsRNA vectors that reduced viable progeny by more than 85 % in the nud-2(+) control strain were excluded, as the RF ratio could not be reliably calculated. Using these criteria, a total of 1077 dsRNA vectors with an RF ratio ≤ 0.5 were selected for secondary screening (Data record 2). dsRNA vectors with a mean RF ratio ≤ 0.5 from three biological replicates (2 technical replicates each) were included in a final list of genetic interactors ( Fig. 2g; Table 1). All dsRNA vectors in the final hit list were sequenced with the M13 forward primer (5′ GTAAAACGACGGCCAGT 3′) to confirm the identity of the targeted gene (Table 1).

Data Records
The data associated with this work is available as two downloadable spreadsheets from FigShare (Table 2; Data Citation 1).

Data record 1primary screen
Raw data from the primary genome-wide RNAi screen is shown in five separate worksheets. The "all_data" worksheet shows the results for every bacterial clone. The worksheet "controls" contains the data for the positive and negative controls. The worksheet "undetermined" identifies the wells whose RF ratio could not be determined for technical reasons. The worksheet "lethal" lists genes that caused more than 85 % embryonic lethality in the nud-2(+) control strain and were therefore excluded from further analysis. The worksheet "hits" shows the dsRNA vectors selected for secondary screening and their location in the re-arrayed plates.
Well -Well coordinates.
RF_R1 / R2 / Avg -Reproductive Fitness of each replicate and the mean of both. Occasionally the RF of a replicate could not be determined for technical reasons (marked "OUT"). In these cases, the value in RF_Avg corresponds to the RF value of the single replicate that worked.
Gene Symbol -Approved name of the gene targeted by the dsRNA vector in each bacterial clone.

Data record 2secondary screen
Data obtained from the secondary RNAi screen is shown in three different worksheets. Analogous to Data Record 1, the worksheet "all_data" shows the results for every bacterial clone that was re-screened. The worksheet "controls" contains the data for the positive and negative controls. The worksheet "RF ratios" summarizes the RF ratios obtained for all clones in each biological replicate, and the average of all replicates. Genes that scored as "enhancers" are highlighted in green.

Technical Validation
Replicate correlation Correlation analysis of replicates in the primary and secondary screen attests to the reproducibility of the screening procedure. In the primary screen, the Pearson correlation coefficients for the technical replicates were 0.80 for nud-2(+) and 0.75 for nud-2(ok949) (Fig. 2a

Controls
Controls were included in every 96-well plate, and plotting of their RF ratios demonstrates high consistency throughout the screening procedure with a clear separation between positive and negative controls (Fig. 2c, d). The robust performance of controls justifies the use of the positive control RF ratios as the basis for setting the threshold for hit identification (RF ratio ≤ 0.5; Fig. 2c, d).

Genes in the final hit list with functional links to dynein
The final hit list is enriched in genes whose functions relate to the microtubule cytoskeleton (13 out of 38) and to cell division (19 out of 38), where dynein is known to play multiple essential roles 14 . In addition, we found two subunits of the dynein complex (dli-1 and dlc-1) and two known dynein regulators (lis-1 and zyg-12). This demonstrates that the screen was successful in identifying genes that functionally intersect with the dynein pathway.

Usage Notes
The raw data of the two-stage RNAi screen is made accessible for several reasons (Data Records 1 and 2). First, it provides a list of genetic nud-2 interactors, whose further study should advance our understanding of dynein regulation and function. Second, users can apply their own normalization strategies and thresholds to compare the nud-2(+) and nud-2(ok949) strains, which may extend the list of candidate dynein pathway genes. Third, while this study focused on genes whose inhibition reduces reproductive fitness of nud-2(ok949), some dsRNA vectors resulted in decreased lethality in nud-2(ok949) (RF ratio > 1; Fig. 2e, f). These genes are also potential interactors of nud-2. Fourth, the data on reproductive fitness in the control strain may serve as a useful reference for other genome-wide RNAi screens using the 96-well liquid format.