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

  1. Chun Hao Wong
  2. Steven W. Wingett
  3. Chen Qian
  4. J. Matthew Taliaferro
  5. Douglas Ross-Thriepland
  6. Simon L. Bullock

Reviewed by

Read the full article

Listed in

Log in to save this article

Abstract

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.

Article activity feed

  1. Review Commons

    Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

    Learn more at Review Commons

    Reply to the reviewers

    The authors do not wish to provide a response at this time

  2. Review Commons

    Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

    Learn more at Review Commons

    Referee #3

    Evidence, reproducibility and clarity

    In this manuscript, the authors performed an arrayed CRISPR loss-of-function screen targeting 18,253 genes with the goal of uncovering gene products that regulate cytoplasmic dynein-1 motor function. In order to assess the impact of gene knockout, the authors optimized a protocol for transfecting pools of cells with mRNA encoding Cas9 and scalably delivering arrayed pools of synthetic guides targeting a single gene to knock-out. In order to link gene knockouts to dynein-1 function the authors employed (1) a previously developed cell model U-2 OS PEX and (2) anti-EEA1 and anti-a-Tubulin antibodies and (3) hoechst as high-content fluorescent readouts for their genome-wide screen.

    The authors then picked a subset of genes to move forwards with that were deemed as hits. A secondary round of screening was performed on these hit genes and unsupervised phenotypic clustering was performed on the feature vectors derived from the high-content images. These analyses revealed several distinct phenotypic clusters that can be categorized by the dynein cargoes or other functional categories including proteostasis related functions. The authors identified the gene SUGP1, which has never previously been linked to dynein-dynactin functionality.

    The authors then show that targeting SUGP1 reduces the mRNA of both LIS1 and DYNC1l2 and the subsequent protein abundance of only LIS1.

    In summary, the authors provide an optimized method for performing what they have termed 'one-shot' genome wide arrayed screening with pools of synthetic guides. They additionally have generated a data resource for others interested in understanding early endosome pathways and dynein-dynactin functionality.

    The technical feat of generating such a large dataset and optimizing the protocol for arrayed synthetic guide pools will undoubtedly be useful for the community. However, this work has several limitations including (1) lack of adequate documentation for reproducing the analyses and (2) minimal mechanistic insight into the function of SUGP1.

    Major Comments:

    • The authors do not provide code or even pseudocode for the algorithms used to generate the features from the high-content images. If the authors are claiming that this would be a resource for the community to use then the authors need to provide an easy way for others to recreate their analysis.
    • The authors mention that they will make the images from their screen publicly available, which is an essential part of making their work a useful resource for the community. However, more details need to be provided about how they will share the results. While a "data dump" of images will be useful to a narrow group of computationally savvy scientists, the broader community will require an interactive interface to enable browsing of the data. The authors should establish such a platform and make it available to reviewers of the revised manuscript to evaluate its usefulness.
    • The authors highlight SUGP1 as an example for "novel mechanistic insights" - but the insights they provide are really minimal. If they authors want to claim mechanistic insights, they should experimentally address questions such as: Does SUGP1 physically interact with LIS1 mRNA? Which region of LIS1 mRNA confers regulation by SUGP1? Can the authors generate a version of LIS1 resistant to SUGP1 regulation to show that the effect of SUGP1 loss is mediated by LIS1 (and not additional factors?).

    Minor Comments:

    • Primary and Secondary antibody pairs are described nowhere in this paper. This would be impossible for anyone to recreate with just the list of primary and secondary antibodies used here.
    • The authors provide no description of how the segmentation was performed or any reference to the code that they used for segmentation regarding the definition of perinuclear region. Considering so many of the results are based on these values it is important that others are able to recreate these values.
    • Line 132: The authors do not explain what a min-max analysis is anywhere in the paper. This should be explained.
    • There is no discussion of how the authors quantify micronuclei formation. If they state that they are the first to do this and that this is a novel technique they at the minimum need to explain the methods for quantifying micronuclei.
    • Supplemental Fig 4C if a per cell intensity quantification is done I would like to see a metric for the segmentation accuracy on these cells overlaid with a cytoskeletal stain.
    • It would be nice to have examples of nuclei or morphology that were excluded from downstream analysis, perhaps in a supplemental figure.
    • Nowhere in the manuscript is it explained how the SUGP1 intensity measurement in Figure 6D is calculated, is this one a per well basis or a per cell basis?

    Significance

    The generation of the dataset described in this manuscript is impressive. However, to reach its full significance and usefulness for the scientific community, the authors should provide relevant technical details, in particular of their analysis pipeline, and share the screen results in an accessible, interactive interface. If they want to claim mechanistic insights into SUGP1, more mechanistic work is required.

  3. Review Commons

    Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

    Learn more at Review Commons

    Referee #2

    Evidence, reproducibility and clarity

    In this manuscript the authors have conducted a genome-wide CRISPR loss-of-function screen in human cells to find regulators of cytoplasmic dynein, a microtubule-based motor that plays a major role in the transport of cargo towards microtubule minus ends. The screen was carried out to address how dynein is synthesised and assembled, and how its activity is controlled to enable the motor to selectively transport a wide variety of cargoes. Cells were fixed cells 72 hours after transfection and fluorescently stained for intracellular markers. Several read-outs were used in the screen, of which the major ones were the distribution of dynein-tethered peroxisomes and early endosomes. The authors used a 384 well format (61 unique 384-well plates) and a fluorescence microscopy-based imaging readout to gauge dynein activity. From a guide RNA library targeting 18,253 genes, the authors recovered 195 validated hits. For one gene (SUGP1) follow-up studies demonstrate that the protein encoded by this gene controls the levels of the dynein activator LIS1 and thereby promotes cargo trafficking by dynein. The dataset reported here represents a source for investigating proteins that might be involved in minus-end microtubule-based transport, as well as in other aspects of cellular organisation that were captured in the high-content imaging approach.

    I find this an interesting and well written resource manuscript, both from the perspective of how to conduct and analyse a high-content imaging screen, as well as from the dynein biology view. Results presented in this manuscript deserve follow-up studies. I do have a number of comments.

    Major comments

    1. On page 6 the authors state they used 61 x 384 wells. This equals 23,424 wells, but the authors state they analysed (8,150,065 cells from) 24,576 wells. What causes this difference in number? More importantly, the authors target 18,253 genes with four guides per gene. If I understand correctly these four guides per gene are present in a single well and the high-content imaging experiment was only done once. Although many cells were analysed per well (four fields of view per well; median of 345 cells analysed per well) and results are interesting and appear solid, I do think a replicate experiment is necessary.
    2. The screen was developed based on the U-2 OS PEX line, in which tethering of dynein to peroxisomes is achieved by addition of rapamycin acting via a split BICD2 protein. Thus, the assay depends on the BICD2 adapter. Is this limiting when one is looking for dynein regulators?
    3. Related to the question above, the authors do not recover BICD1 in their screen. Is this because BICD1 is not expressed in the cell systems used or is there another reaosn?
    4. It has very recently been shown (doi.org/10.1038/s41467-023-38116-1) that BICD2 phosphorylation by CDK1 in the G2 phase of the cell cycle promotes its interaction with PLK1. This is followed by PLK1 phosphorylation in the N-terminus of BICD2, which in turn facilitates interaction with dynein and dynactin, allowing the formation of active motor complexes. Thus, adaptor activation through phosphorylation regulates dynein activity. In the present manuscript the authors use PLK1 as a read-out of cell viability. However, PLK1 also appears to regulate dynein via BICD2 phosphorylation. Given the latest results would the authors interpret their PLK1 data differently? Would it be preferable to screen for regulators of dynein activity in non-dividing cells?
    5. Using the 377 genes listed in Supplementary Table 4 I performed a Metascape analysis. The results suggest that many of the hits are proteins involved in RNA metabolism or the cell cycle and that many of the encoded proteins form complexes. Based on this I wonder whether the screen yielded many proteins that are involved in controlling the steady state levels of dynein, microtubules, or of the dynein regulators. SUGP1 is an example of this. I suggest that the authors include an extensive Metascape analysis in a new version of the manuscript.
    6. On page 11 a UMAP plot is described, which is shown in Figure 4B. How were the "members of the same protein complexes, such as histones, ribosomal proteins, RNA polymerase II, the RUVBL and TRiC/CCT chaperonins, FAM160A2- AKTIP-HOOK3 and dynein-dynactin" identified?
    7. How do the complexes identified in Figure 4B relate to the MCODE-based complexes identified in Metascape?

    Significance

    I think the present manuscript is an interesting resource paper for the dynein community. The advance is technical rather than conceptual.

    I am a cell biologist with an interest in microtubules and how this cytoskeletal network controls cell shape and function. I analyse this using fluorescence microscopy and -omics approaches. I am not an expert in high-content imaging screens and analyses but the data presented here seem solid and novel to me.

  4. Review Commons

    Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

    Learn more at Review Commons

    Referee #1

    Evidence, reproducibility and clarity

    Cytoplasmic dynein-1 (dynein) is the predominant minus-end directed microtubule-based motor involved in the transport of numerous cellular cargoes in addition to mitotic functions. Although in vitro analysis of dynein assembly and interactions are becoming more common, cell biological studies that aim at identifying different dynein cargoes and functions are lacking. To shed light on dynein's function, Wong et al., performed a genome-wide CRISPR loss-of-function screen using peroxisome-tethered and endosome localization assays as readouts. Detailed data analysis and supervised and unsupervised phenotypic clustering of targets from an extensive RNA gene library (~20,000 genes) revealed ~200 genes disrupted in cargo trafficking processes. The authors also showed that a novel gene SUGP1, identified in their screen, promotes the expression of a critical dynein activator, LIS1. The generated datasets provide a rich library of genes that can be further mined by other researchers in the field.

    Major comments:

    This manuscript reads well, and the conclusions are mostly supported by experimental data. The authors went to great lengths to optimize their high throughput assay setup by testing different cell lines and transfection conditions and included different positive and negative controls, which is a strength of the study. The use of two functional readouts (early endosome and BICD2-driven peroxisome distribution) in the initial screen followed up by a validation screen of a smaller subset of genes using readouts for dynein disruption phenotypes (Golgi fragmentation, lysosome clustering) is another strength. In addition, the follow-up identification of an RNA-binding protein SUGP1 as a regulator of LIS1 mRNA levels provides an interesting new way of regulating dynein function.

    I have a few concerns about the experimental design and conclusions.

    1. The images in Fig. 1E and 1G for the dynein control (crDYNHC1) show some clustering around the cell nucleus while crLIS1 knockout shows no perinuclear clustering. Is this expected? Shouldn't dynein knockout prevent perinuclear clustering? Is it possible that crDYNHC1 does not lead to a complete knockout? Given that this is a proof-of-principle control in the assay, a more detailed validation of DYNHC1 knockout using western blotting and RT-qPCR, in addition to the validation shown in Fig. 1C would strengthen the claim that this control works as expected. These experiments should be fast and easy to do.
    2. The use of multiple crRNAs together to target a single gene can increase off-target effects, however, the authors never test for off-target phenotypes or address the possibility of off-targeting. Can the authors show using a few examples that their approach does not lead to significant off-targeting? This should also be addressed in the text.
    3. It is unclear to me how the authors established the limits for the quantification of localization ratios. As described in methods, the perinuclear region was defined as having an outer limit of 7 μm from the nuclear envelope. However, the cells are not the same size (also seen in representative images), which could skew the calculation of ratios solely based on fixed distance limits. Have the authors considered taking into account cell size? Perhaps a more accurate calculation would be to measure the distance from the nucleus to the cell periphery for each cell and normalize this value to the cell size to account for cell size differences. The perinuclear region could then be defined as the percentage of the distance from the nuclear envelope of the normalized cell radius. It is also unclear how the size and intensity of each "spot" are accounted for in the analysis as this is an important aspect of the quantification given that the "spots" are not the same size/intensity. Redoing this analysis would not require the authors to collect any new data but could help in gene identification, especially given that the authors only identified ~50% of the known dynein-dynactin complex components to be disrupted in their assay. These genes might have more subtle phenotypes that could be amplified by doing more precise image analysis and quantification.

    Minor comments:

    1. It would be helpful if the authors could change gene names to a bold or brighter font in scatter plots in all figures. It is hard to read the names the way they are right now.
    2. Line 185 the authors say: "We also analysed the induction of micronuclei (Supplementary figure 6B), which to our knowledge has not been assessed in earlier screens." What screens are the authors referring to? Could you add references here?
    3. Line 256: "Each cargo was assayed in two independent screens, in which there was good agreement in general between the effects of the crRNA pools (Supplementary figure 8)." The authors also indicate in the legend for Supplementary figure 8 that "The only metric with a poor R2 score (proportion of cells with two γ- Tubulin puncta) was not used for hit calling." However, the EEA1 localization ratio also shows poor R2 score, shouldn't this screen also be excluded? In general, what was the cutoff for R2 score? This information should be included.

    Significance

    This work is the first genome-wide loss-of-function CRISPR study (to my knowledge) aimed at identifying dynein-driven trafficking disruption phenotypes. In general, the data generated in this study will enrich the field's understanding of how dynein is regulated and how it achieves its broad cargo and functional specificity.This manuscript will also provide a resource and experimental setup for the design of other genome-wide loss-of-function CRISPR studies.

    I have broad expertise in the cytoskeleton field with a detailed understanding of dynein's function from a mechanistic and functional perspective. I have minimal experience with high throughput screening, but I am experienced with CRISPR-based assays and cell imaging techniques.

  5. Version published to 10.1101/2023.03.01.530592v1 on bioRxiv