CRISPR knockout screen implicates three genes in lysosome function

Defective biosynthesis of the phospholipid PI(3,5)P2 underlies neurological disorders characterized by cytoplasmic accumulation of large lysosome-derived vacuoles. To identify novel genetic causes of lysosomal vacuolization, we developed an assay for enlargement of the lysosome compartment that is amenable to cell sorting and pooled screens. We first demonstrated that the enlarged vacuoles that accumulate in fibroblasts lacking FIG4, a PI(3,5)P2 biosynthetic factor, have a hyperacidic pH compared to normal cells'. We then carried out a genome-wide knockout screen in human HAP1 cells for accumulation of acidic vesicles by FACS sorting. A pilot screen captured fifteen genes, including VAC14, a previously identified cause of endolysosomal vacuolization. Three genes not previously associated with lysosome dysfunction were selected to validate the screen: C10orf35, LRRC8A, and MARCH7. We analyzed two clonal knockout cell lines for each gene. All of the knockout lines contained enlarged acidic vesicles that were positive for LAMP2, confirming their endolysosomal origin. This assay will be useful in the future for functional evaluation of patient variants in these genes, and for a more extensive genome-wide screen for genes required for endolysosome function. This approach may also be adapted for drug screens to identify small molecules that rescue endolysosomal vacuolization.

each cell (Fig. 2c). Lysosomes in Fig4 null fibroblasts have an average pH of 4.1 ± 0.5 (mean ± sd), significantly more acidic than the normal lysosomes in wild-type cells (p < 10 −25 , t-test), whose observed pH of 5.0 ± 0.5 is comparable to previous measurements 20 . The acidity of the vesicles accumulating in FIG4 null cells is consistent with their origin from the endolysosomal compartment, and indicates that PI(3,5)P 2 deficiency, a known consequence of FIG4 loss, may lead to hyperacidification of mammalian lysosomes.  (b) Fluorescence emission at 535 nm was recorded for excitation wavelengths of 440 nm (pH insensitive, I (ex440) ) and 488 nm (pH sensitive, I (ex488) ). Ratiometric images were obtained as I (ex488) /I (ex440) , calibrated using separate, pH-clamped cells, and the calculated pH values were displayed as pseudocolored images. (c) Vesicles in Fig4 null fibroblasts are acidic, with an average pH 4.1 ± 0.5 (n = 132) compared with 5.0 ± 0.5 (n = 99) in wildtype fibroblasts (P ≤ 10 −25 , t-test). P 2 -deficient human cell line, we targeted FIG4 exon 6 using CRISPR/Cas9 in the haploid human cell line HAP1. We isolated two clonal FIG4 knockout cell lines, 1C2 and 3D4, transfected with independent sgRNAs. By visual inspection, and similar to FIG4 null mouse fibroblasts, the two clones contained a high proportion of vacuolated cells (Fig. 1b). Both clones contain frame-shifting indels in FIG4 predicted to result in premature stop codons (Fig. S1a). Vacuolization was rescued by transfection with wildtype FIG4 cDNA (Fig. S1b). These FIG4 null cells and the wild-type HAP1 line from which they were derived were used to develop and optimize the flow-sorting assay for vacuolated cells.
FACs assay to detect cells with an expanded acidic compartment. The fluorescent dye LysoSensor permeates the cell membrane and accumulates in lysosomes in a pH-dependent fashion 21 . After staining, vacuolated cells have elevated fluorescence when excited at 329 nm that is readily visualized in individual cells by epifluorescence microscopy (Fig. 3a). The difference between wildtype and mutant cells can be quantified within a population by fluorescence activated cell sorting (FACS). We applied stringent gating criteria that selected >60% of FIG4 null cells as LysoSensor positive, but <2% of wildtype cells (Fig. 3b). The reproducibility of sorting results obtained from 3 experiments each on the two independent FIG4 null clones is shown in Fig. 3c. A genome-wide screen for mutations causing accumulation of acidic vacuoles. We hypothesized that the FACS assay could be used as part of a mutagenic screen in order to discover novel genes which phenocopy FIG4-null vacuolization due to their roles in lysosome dynamics such as PI(3,5)P 2 production. We adopted a CRISPR/Cas9-mediated pooled screening approach in which each cell in a mixed population is a knockout for a different gene (reviewed in ref. 16 ). This screen involved four steps ( Fig. 4): first, HAP1 cells were transduced at low multiplicity of infection (MOI) with the GECKOv2 lentiviral library 22 encoding Cas9 nuclease and six sgRNAs per gene targeting 19,050 human genes; then, pooled cells carrying individual gene-disrupting mutations were stained with LysoSensor and vacuolated cells were captured by FACS; next, genomic DNA was isolated from pooled captured cells and integrated sgRNA sequences were amplified by PCR; finally, enriched sgRNAs were identified by sequencing of the PCR products and demonstration of elevated abundance in FACS-selected cells compared with the starting population.

Preliminary screen identifies 15 genes enriched in vacuolated cells. Wildtype HAP1 cells were
transduced with the pooled GeCKO library at an MOI of 0.2 and selected with puromycin. The mixed population was expanded as a pool to produce adequate material for downstream analysis. Sequencing sgRNAs from the plasmid pool and transduced cell population indicated that most guides were present at each step, and as expected, genes in cell-essential pathways were significantly depleted after transduction and outgrowth (Fig. S2). After staining with Lysosensor, the transduced population was sorted, using a stringent gating which captured <0.001% of cells from a wildtype cell population. From the genome-wide knockout cell population, 1.7 × 10 3 cells were captured out of ~10 7 sorted (0.017%), a substantial excess over the wild-type controls. These vacuolated KO cells were expanded as a pool and further sorted for one additional round to increase the purity of the vacuolated population. The resulting population contained predominantly vacuolated cells when assessed visually. Sequencing identified 16 sgRNA sequences targeting 15 genes which together accounted for 95% of the sequences (Table 1). Among these, the two most-highly enriched sgRNAs targeted VAC14, a gene required for PI(3,5)P 2 biosynthesis that is known to cause vacuolization and lysosome dysfunction 11,15,23,24 . One sgRNA targeted CLN8, a transmembrane protein responsible for a neurodegenerative lysosomal storage disorder 25,26 . Given the identification of bona fide lysosomal factors, we focused on the remaining 13 highly enriched genes not previously associated with lysosome function (Fig. 4b).
Functional validation of three selected genes. C10orf35, LRRC8A and MARCH7 were selected for validation by re-creating individual knock-out clones using CRISPR/Cas9 and two new sgRNA sequences (Table S1). VAC14 was included as a positive control. After isolation and clonal expansion of individual transfected cells,     (Table 1). www.nature.com/scientificreports www.nature.com/scientificreports/ sequencing revealed frame-shifting indel mutations predicted to result in premature truncation of each targeted gene (Figs 5 and S3). All derived mutant cell lines contained vacuolated cells visible by phase contrast microscopy (Fig. 6a), and all showed significantly greater LysoSensor staining by FACS analysis compared with wild-type cells (Fig. 6b). Both microscopy and FACS analysis indicated a lesser degree of vacuolation for C10orf35, LRRC8A and MARCH7 knockouts compared with FIG4 or VAC14, suggesting non-identical roles for these factors, or differential sensitivity to their loss. We observed complete absence of targeted protein expression in each mutant cell line by Western blotting (Fig. 7), indicating that the observed vacuolation phenotypes in these mutants result from loss of function. Finally, to confirm the endolysosomal origin of the vesicles in the mutant cell lines, we carried out immunostaining of the membrane protein LAMP2. The membrane surfaces of enlarged vacuoles in the C10orf35, LRRC8A and MARCH7 null clones are stained for LAMP2, while wildtype cells exhibit punctate staining of small lysosomes (Fig. 8). We conclude that mutations of C10orf35, LRRC8A and MARCH7 generate enlarged vesicles derived from the endolysosomal pathway, the same origin as those observed in FIG4 or VAC14-null cells. www.nature.com/scientificreports www.nature.com/scientificreports/

Discussion
Application of CRISPR-Cas9 mutagenesis to pooled screens provides a new tool for the study of gene function in cellular and animal models 16,27 . Although the capacity to programmably target entire coding or regulatory genomes represents a major advance over spontaneous or random mutagenesis, these pooled approaches share with classical forward genetic screens the requirement for an assay to identify and enrich cells with the phenotype of interest. As a result, while they can be readily applied as fitness-based selections (e.g., for increased cellular growth, or drug resistance), adapting them to other biological processes requires the development of cellular assays specific to the pathway or activity of interest. For instance, two recent screens have identified knockouts which perturb expression of an engineered reporter of Hedgehog signaling 28 , or which dysregulate native fetal hemoglobin expression 29 , necessitating specialized, sortable expression assays for each target. Here, we have developed a FACS-based assay which for the first time enables pooled screening for genes which disrupt lysosomal dynamics and result in an enlarged lysosomal compartment.
The first candidate gene arising from this screen, LRRC8A, encodes a component of the volume-regulated anion channel (VRAC) that regulates cell volume in response to changes in extracellular osmolarity [30][31][32] . Elucidation of the protein structure of LRRC8A revealed a transmembrane pore domain and a leucine-rich cytoplasmic domain [33][34][35] . Gene inactivation in a conditional knock-out mouse has demonstrated roles for LRRC8A a b Fluorescence signal (695 nm) www.nature.com/scientificreports www.nature.com/scientificreports/ in spermatogenesis and insulin secretion 36,37 . Our data indicate that LRRC8A is also involved in regulation of lysosome volume, and that loss of function leads to osmotic swelling of the lysosome. We previously proposed osmotic swelling as the basis for vacuolization in FIG4 and VAC14 null cells, resulting in those cases from impaired activation of lysosomal cation channels by PI(3,5)P2 (ref. 7 ). It will be of interest to compare the effects of neuronal knockout of LRRC8A with the neurodegeneration that results from neuronal inactivation of FIG4 (ref. 38 ).
The second gene, MARCH7, is a member of the ubiquitin ligase E3 class that targets proteins for trafficking to the multivesicular body for degradation in the proteosome. A mutation in another ubiquitin E3 ligase is responsible for the mouse mutant mahoganoid, which also exhibits accumulation of cytoplasmic vesicles and spongiform degeneration of the CNS similar to the FIG4 and Vac14 null mice 39 . The under-representation of null alleles of MARCH7 in the ExAC exome database 40 (pLI score of 0.98) indicates that haploinsufficiency of MARCH7 is deleterious and could result in dominantly inherited disease. Our screen also identified C10orf35, which encodes a predicted single-pass membrane protein whose cell localization and function have not previously been characterized.
Since it is based on LysoSensor fluorescence in an acidic environment, the FACS assay specifically detects enlarged lysosomes. The vesicles in FIG4 null cells were hyperacidic, with a pH of 4.1 compared with 5.0 in wildtype fibroblast cells. One mechanism that could account for hyperacidity would be dependence on PI(3,5)P 2 of a proton-permeable anion channel involved in proton countertransport. For example, impaired activation of ClC-7, proposed to play a role in escape of protons from the lysosome interior 41 , could lead to lower intralysosomal pH.
Although we functionally validated all three novel genes selected for follow-up, each was only captured once in this pilot study, suggesting that a number of additional genes with roles in lysosome dynamics remain to be discovered. We plan to expand the screen described here to achieve genome-wide saturation, to generate a catalog of genes with roles in lysosome regulation, and to further contribute to the understanding of lysosome biology in human cells. One limitation of this study is that it does not address potential tissue specificity of cytoplasmic vacuolization. While lysosomes are universal components of the mammalian cell, in vivo vacuolization in FIG4 and Vac14 mutant mice is restricted to neural tissue 17 , reflecting an apparent greater susceptibility. However, other cells derived the mutant mice exhibit vacuoles when cultured in vitro, including fibroblasts, osteoblasts, neurons, oligodendrocytes, and bone marrow macrophages. Additionally, as with other knock-out screens, genes which are essential for growth in this particular cell line (HAP1) would not be identified.
The FACS assay for vacuolization described here can be applied in the future to evaluation of novel patient mutations in these genes, addressing the challenge of interpreting 'variants of unknown significance' identified by exome and whole genome sequencing [42][43][44] . As a result of this preliminary screen, disruptive variants in C10orf35, LRRC8A and MARCH7 found by clinical sequencing should be considered candidate genes for neuromuscular and lysosomal disorders of unknown origin. This functional assay could also be adapted to drug screening for therapeutic small molecules that reverse the vacuolization phenotype.

Methods
Generation of FIG4 null HAP1 cells. Human HAP1 cells (Horizon Discovery, #C631) were originally derived from human hematopoetic cells 45,46 . HAP1 and HAP1-derived cell lines and pools were maintained in Iscove's Modified Dulbecco's Medium (IMDM) supplemented with 10% FBS and 100 units/mL penicillin, 100 µg/mL streptomycin (Invitrogen). Guide sequences targeting FIG4 exon 6 were selected using E-CRISP 47 and screened for potential off-target sites with Cas-OFFinder 48 . Guides were ordered as oligonucleotides from IDT, annealed, and cloned into plasmid pSpCas9(BB)-2A-GFP 49 to express the sgRNA along with SpCas9 and GFP C10orf35 clone 2 MARCH7 clone 2 LRRC8A clone 5 HAP1 wildtype VAC4 clone G9 FIG4 clone 3D4 www.nature.com/scientificreports www.nature.com/scientificreports/ genes. HAP1 cells (~1 million) were transfected with 12 ug sgRNA plasmid using 6 ul Fugene reagent (Promega). After 18 hrs of growth, individual GFP-positive cells were flow sorted into a 96 well plate. Resulting colonies were examined under a light microscope and those with vacuolated appearance were expanded. Induced mutations in the targeted region (FIG4 exon 6) were identified by Sanger sequencing. Clones 1C2 and 3D4 were generated with two different sgRNAs and used in subsequent experiments.
Measurement of lysosomal pH. Ratiometric fluorescence of 10kD Oregon Green Dextran (OGDx) was carried out as previously described 19 . OGDx was included in the cell culture media for 2 hours followed by a 24 hour chase to permit accumulation of the fluorophore in lysosomes. Fluorescence emission at 535 nm was recorded for the excitation wavelengths 440 nm (pH insensitive) and 488 nm (pH sensitive). The ratio of fluorescence emission at these wavelengths was compared to a standard curve generated by incubating OGDx-labeled cells in buffers of known pH in the presence of the ionophores valinomycin and nigericin. Small punctate lysosomes and a subset of enlarged vacuoles were labeled by OGDx in FIG4-null cells. For each cell, a single average pH value was obtained for all of the OGDx-containing lysosomes.

FACS assay using accumulation of fluorescent dye to separate vacuolated and wildtype cells.
Eighteen hours prior to FACS analysis, HAP1 cells were plated in 100 mm plates at 40,000 cells per cm 2 in IMDM containing 10% FBS with a final concentration of 100 units/mL penicillin, 100 µg/mL streptomycin, and 0.25 µg/mL of Gibco Amphotericin B ("full medium"). Cells were incubated in a humidified incubator at 37 °C with 5% CO 2 . The endolysosomal compartment was labeled by incubation for 15 min with 5 μM LysoSensor Yellow/Blue DND-160 dye (LysoSensor) (Molecular Probes). LysoSensor fluorescence is excited at 329 nm with emission peaks at 440 and 540 nm. Labeling medium was removed and cells were washed 3 times at room temperature with PBS. Cell monolayers were removed from the plates by treatment at 37 °C for 5 minutes with TrypLE Express Enzyme Without Phenol Red (Thermo Fisher Cat #12604013). Cells were suspended in PBS containing 2% FBS and placed on ice. Propidium iodide, final concentration 1.5 ug/ml, was added to the cell suspension as a viability marker.
GeCKo library and lentiviral transduction. GeCKO v2 pooled knock-out libraries "A" and "B" were obtained from Addgene (#1000000048; gift of Feng Zheng). Each half-library was transformed in bulk into Endura strain E. coli (Lucigen), expanded in 150 ml culture, and isolated by maxiprep (Zymo Research #D2403). Lentiviral particles were produced by the University of Michigan Vector Core by co-transfecting HEK-293T cells with packaging plasmids psPAX2 and pMD2.G. For bulk transduction, wild-type HAP1 cultures were expanded to 12.5 × 10 6 cells in T175 flasks. Viral supernatant (0.25 ml, ~10 7 IFU/ml) was added to reach an approximate multiplicity of infection of 0.2. After 24 hours, puromycin (2.5 ug/ml final concentration) was added to the media to select for transduced cells.
Cell sorting and sequencing. Pooled transduced HAP1 cultures were expanded to ~10 8 cells (6-7 doublings). Cells were plated at 40,000 cells/cm 2 , stained with LysoSensor as above, and approximately 10 7 cells were sorted. Captured cells were plated, expanded (~15 doublings), and subjected to an additional round of sorting by the same procedure to further enrich for vacuolated cells. After outgrowth (~4 doublings), genomic DNA was prepared using the Gentra Puregene kit (Qiagen #158767). Integrated sgRNAs were amplified by PCR for 24 cycles as previously described 50 , with modified amplification primers which included a randomized 'stutter' sequence (Table 1) to mitigate artifacts on the Illumina platform from low-diversity sequences. For sgRNA sequencing, 2 ug of gDNA per PCR reaction was used as template, and four reactions were pooled per sample to ensure sufficient representation of integrated sgRNAs. The resulting amplicon was diluted and further amplified in a second-round PCR to add Illumina adaptor sequences and sample-specific index barcodes. All libraries were pooled and sequenced on MiSeq and HiSeq instruments using 50-bp single-end reads. single gene validation knock-out. sgRNAs were targeted to constitutively included exons of C10orf35, LRRC8A and MARCH7 using Ensembl gene models. Potential sgRNA sequences were designed and cloned into pSpCas9(BB)-2A-Puro, transfected into wild-type HAP1 cells, and clonal KO cells were isolated as described above for FIG4. Western blotting. Protein was isolated from cells grown in a monolayer to 70~80% confluence using RIPA buffer and protein inhibitor cocktail (Thermo Scientific). Samples were prepared for electrophoresis under reducing conditions with Laemmli sample buffer (Bio Rad) containing 2-mercaptoethanol (Sigma Aldrich). Twenty ug of protein per lane was loaded onto acrylamide gels of varying concentration, depending on the molecular weight of the protein. The primary antibodies were mouse anti-C10orf35 (Abcam, 1/5,000 dilution), rabbit anti-LRRC8A (Thermo Scientific, 1/2000 dilution), and rabbit anti-MARCH7 (Sigma Aldrich, 1/1500 dilution). Membranes were subsequently incubated with peroxidase-labeled secondary antibodies. Chemiluminescence was detected with SuperSignal West Femto chemiluminescence reagent (Pierce, Thermo Scientific) and HyBlot-CL autoradiography film (Denville Scientific).

Microscopy of cultured cells.
To visualize vacuoles, cells were grown in full medium for 36 hours after replating, and phase contrast images were taken with the EVOS FLc system (Life Technologies). For LAMP2 immunostaining, cells were grown for 36 hours on 4-compartment slides (LAB-TEK, Rochester, NY), washed 3x with cold PBS, and fixed with cold methanol for 5 min at −20 °C. Cells were permeabilized by incubation for 10 min in 0.1% Triton X-100 in PBS and then blocked for 1 hour using 5% goat serum in PBS. Slides were processed for indirect immunofluorescence by incubation with LAMP2 antibody (Developmental Studies Hybridoma Bank, University of Iowa), 1/4000 dilution, for 2 hours at room temperature, then washed with PBS