A CRISPR/Cas9 screen reveals proteins at the endosome-Golgi interface that modulate cellular anti-sense oligonucleotide activity

Abstract

Anti-sense oligonucleotides (ASOs) are modified synthetic single-stranded molecules with enhanced stability, activity, and bioavailability. They associate with RNA through sequence complementarity and can reduce or alter mRNA expression upon binding of splice site positions. To target RNA in the nucleus or cytoplasm, ASOs must cross membranes, a poorly understood process. We performed an unbiased CRISPR/Cas9 knockout screen with a genetic splice reporter to identify genes that can increase or decrease ASO activity, resulting in the most comprehensive catalog of ASO-activity modifier genes. Here we reveal distinct targets, including AP1M1 and TBC1D23, linking ASO activity to transport of cargo between the Golgi and endosomes. AP1M1 absence strongly increases ASO activity by delaying endosome-to-lysosome transport in vitro and in vivo. Prolonged ASO residence time in the endosomal system may increase the likelihood of ASO escape. This insight into AP1M1 role in ASO trafficking suggests a way for enhancing the therapeutic efficacy of ASOs by manipulating the endolysosomal pathways.

Fig. 1. CRISPR/Cas9 KO whole genome pooled screen using the EGFP splice reporter identifies genes that modify ASO activity.

a Schematic representing splice switch reporter (HEK293-Cas9-EGFPsr) as a surrogate reporter for ASOs activity. ASO that enter the cell by gymnosis and escape the trafficking vesicles to the nucleus will bind to the cryptic exon of the splice reporter and cause the correct splicing and GFP expression. Created in BioRender. Roudnicky, F. (2025) https://BioRender.com/lhaszwvb Schematic representing whole genome knockout CRISPR/Cas9 screen. HEK293-Cas9 cell line expressing HEK293-Cas9-EGFPsr were transduced with a whole genome sgRNA library, after selection cells were treated with SS-ASO (25 nM), and top and bottom 10% GFP cells were FACS sorted (Round 1), after a week of recovery, treatment was repeated followed by sorting using same conditions (Round 2). SgRNAs were deep-sequenced, and sequences were analyzed to identify sgRNAs that became enriched in sorted populations (SORT1 and SORT2) relative to the unsorted population (BULK, populations that were NGS sequenced have an underline in the schematic). Created in BioRender. Roudnicky, F. (2025) https://BioRender.com/1qbrqw7. c Volcano plot of genes from the comparison of Round 1 of top 10% GFP+ vs unsorted population (BULK). d Volcano plot of genes from the comparison of Round1 of top 10% GFP- vs unsorted population (BULK). Both Volcano plots show the significance -log10 (rho-value) versus the log2 (Fold change). Colored genes have a rho-value < 0.0001 and Fold change > 1.5. Top 30 genes sorted based on the fold change are labeled. e Gene pathway analysis from significant genes (Fold change > 1.5 and p-adjusted value < 0.05) from the comparison of Round 1 of top 10% GFP+ (SORT1 GFP + ) vs unsorted population (BULK) and f from the comparison of Round1 of top 10% GFP (SORT1 GFP-) vs unsorted population (BULK) using g:Profiler with GO:MF (upper panel) and GO:CC (lower panel). Significant pathways are denoted with black bars (p-adjusted value < 0.05). Pathway enrichment significance was assessed using Fisher’s Exact one tailed Test, with p-values derived from the cumulative hypergeometric distribution. Multiple testing was corrected using g:SCS (Set Counts and Sizes) correction method. Source data are provided in corresponding Supplementary Data.

Fig. 2. Arrayed screens validate genes involved in retrograde trafficking and transport to and from Golgi compartment.

a Schematic of arrayed validation screens. The sgRNA were delivered via lentiviral transduction. Following antibiotic selection, cells were treated with SS-ASO (25 nM, 3 days), and quantified the GFP+ expression by flow cytometry or high content imaging. Created in BioRender. Roudnicky, F. (2025) https://BioRender.com/vr5pkra. b Percentage of GFP+ cells was measured with flow cytometry. The screen was performed three times as independent experiments using 3 to 4 sgRNA per gene. Data is plotted as mean with SD and normalized to non-targeting sgRNA. c Proportion of GFP+ cells was measured using high content imaging. The screen was performed as 2 independent experiments using 3 to 4 sgRNA per gene. Data is plotted as mean with SD. d Representative images of the most interesting hits from the screen performed in Fig. 2c. Cellular segmentation was performed according to the Hoechst counterstaining and is depicted as white outlines. GFP is depicted in green. e Protein-protein association network generated by STRING database performed on the hits presented on (c). The disconnected nodes are not displayed. The thickness of the line between nodes represents the confidence of the interaction, with the thickest lines showing the highest confidence. The nodes are color-coded in red or green according to their effect on ASO activity. This analysis identifies the AP1 adapter complex, as well as the TBC1D123/Fam91a1/WDR1 complexes, outlining the importance of vesicular transport between endosome/Golgi in determining ASO activity. Source data are provided as a Source Data file or in corresponding Supplementary Data.

Fig. 3. AP1M1 and TBC1D23 are strong modifiers of ASO activity.

a Percentage of CD81 + HEK293 was quantified by flow cytometry following 72 h of treatment with an ASO against CD81 (0.25 or 2.5 µM), or a control non- targeting ASO (CTRL, 2.5 µM). Horizontal lines show the median, each dot represents a technical replicate (n = 2) from 3 independent experiments. Different color shades represent different clones. Statistical testing was performed with two-way ANOVA with Dunnett’s post-test (exact p-value = 0.0008). b Percentage of CD81 + U2OS was quantified by flow cytometry following 72 h of treatment with an ASO against CD81 (0.25 and 2.5 µM), or a control non- targeting ASO (CTRL, 2.5 µM). Horizontal lines show the median, each dot represents a technical replicate (n = 2) from 3 independent experiments. Different color shades represent different clones. Statistical testing was performed with two-way ANOVA with Dunnett’s post-test. c Proportion of CD81 + HEK293 over time was quantified by flow cytometry, following 1–5 days time points of treatment with an ASO against CD81 (2.5 µM). Horizontal lines show the median, each dot represents a technical replicate (n = 2) from 2 independent experiments. Different color shades represent different clones. Statistical testing was performed with two-way ANOVA with Dunnett’s post-test (exact p-value = 0.0029). d Graph showing the proportion of CD81 + U2OS over time, as quantified by flow cytometry, following 1-3 days of treatment with an ASO against CD81 (2.5 µM). Horizontal lines show the median, each dot represents a technical replicate (n = 2) from 3 independent experiments. Different color shades represent different clones. Statistical testing was performed with two-way ANOVA with Dunnett’s post-test. e Representative Western blot experiment of U2OS (AP1M1 or TBC1D23 KO) cells expressing the rescue construct (inducible expression of AP1M1 or TBC1D23), following doxycycline (Dox) treatment for 3 days. We used antibodies against AP1M1, TBC1D23, and ACTIN (loading control). The Western blot shows a dose-dependent re-expression of our genes of interest. f Percentage of CD81+ in rescued U2OS as a function of Dox concentration, as quantified by flow cytometry, following 0.5 days of Dox treatment followed by 2.5 days of treatment with Dox and ASO against CD81 (2.5 µM). Dotted lines show the non-linear fitted curve (least squares regression method), each dot represents a technical replicate (n = 2) from 3 independent experiments. Different shades of color represent different clones. Statistical testing was performed with a linear regression (R² = 0.02, 0.31, 0.04 for wild-type, AP1M1 KO, and TBC1D23 KO, respectively). **p-value < 0.01; ***p-value < 0.001; ****p-value < 0.0001. Source data are provided as a Source Data file.

Fig. 4. The effect of AP1M1 is downstream of gymnotic uptake.

a Percentage of CD81 + HEK293 cells was quantified by flow cytometry, following gymnotic or RNAiMAX delivery of CD81 ASO (0.25 and 2.5 µM) for 24 h. Horizontal lines show the median, each dot represents a technical replicate (n = 2) from 3 independent experiments. Different color shades represent different clones. Statistical testing was performed with a two-way ANOVA with a Dunnett’s post-test (*p-value < 0.05, exact p-value = 0.0125; ****p-value < 0.0001). Dashed line depicts the percentage of CD81+ in untreated HEK293 clones. b Percentage of CD81 + U2OS cells was quantified by flow cytometry, following gymnotic or RNAiMAX delivery of CD81 ASO (0.25 and 2.5 µM) for 24 h. Horizontal lines show the median, each dot represents a technical replicate (n = 2) from 3 independent experiments. Different color shades represent different clones. Statistical testing was performed with a two-way ANOVA with a Dunnett’s post-test. Dashed line depicts the percentage of CD81+ in untreated U2OS clones. c Fluorescence Mean Intensity (FMI, from Flow cytometry analysis) of HEK293 cells treated with a fluorescently labeled control non-targeting ASO (CTRL), over time (from 30 min to 72 h). Each dot represents the average value of a clone measured with 2 replicates from 2 independent experiments. Statistical analysis was performed with a two-way ANOVA with Dunnett’s post-test at 72 h. d Vesicles containing the fluorescently labeled ASO (spots) per U2OS cell were quantified from 30 min to 72 h by imaging. Each dot represents the average ± SEM of a clone measured with 2 replicates in 3 independent experiments. Statistical testing was performed with two-way ANOVA with Dunnett’s post-test at 72 h. e Representative images of the time course uptake experiment quantified in Fig. 4d. ASO is depicted with magenta while the nucleus is counterstained with Hoechst and depicted in cyan. Scale bar corresponds to 50 µm. Source data are provided as a Source Data file.

Fig. 5. Lysosomal function and morphology is not disrupted in AP1M1 KO cells.

a The number of spots per cell, the area covered by these spots, and their intensity was quantified by imaging of U2OS cells treated with the following lysosomal markers: SiR Lysosome (SiR Lyso), Cytofix Red Lysosome (Cytofix Lyso) and Magic Red. Every condition had 2 technical replicates measured in 3 independent experiments. Different color shades represent different clones. Statistical testing was performed with the unpaired two-tailed t-test (***p-value < 0.001, exact p-value = 0.0014). b Representative images of U2OS cells stained with lysosomal markers, as quantified in Fig. 5a. The different lysosomal markers are depicted in magenta, while the nucleus is counterstained with Hoechst and depicted in cyan. Scale bar corresponds to 40 µm. c Quantification of the proportion of cells that contain at least 1 galectin3-mCherry punctate (Gal3-mCherry) per cell was performed by imaging in U2OS clones overexpressing Gal3-mCherry. Horizontal lines show the median; 5 technical replicates per condition were measured in 2 independent experiments. Different color shades represent different clones. Statistical analysis was performed using an unpaired two-tailed t-test. d Representative images of (c). Gal3-mCherry fluorescence is depicted in magenta, while the nucleus is counterstained with Hoechst and depicted in cyan. Scale bar corresponds to 100 µm. Source data are provided as a Source Data file.

Fig. 6. AP1M1 KO cells show perturbed trans-Golgi network and show delayed transport of cargo to the lysosomes.

a Representative electron microscopy images of wild type and AP1M1 KO U2OS cells. Empty yellow arrowheads point to healthy Golgi stacks. Dashed yellow lines surround Golgi anomalies. N depicts the nucleus. Golgi anomalies were rare in the wild type samples (n = 11) but observed in most of the AP1M1 KO cells (n = 30). Scale bar corresponds to 500 nm. b, c TGN46 was immunostained in wild type and AP1M1 KO U2OS and quantified by imaging for staining intensity (b) and the number of spots per cell (c). Five technical replicates were measured in each of the 4 independent experiments. Different color shades represent different clones. Statistical testing was performed using an unpaired two-tailed t-test. d Representative images of TGN46 immunostaining (depicted in magenta), and nuclei counterstaining by Hoechst (depicted in cyan) in wild type and AP1M1 KO U2OS cells. e Proportion of mCherry+ cells in wild-type and AP1M1 KO cells, expressing AP1M1 doxycycline-(Dox) inducible construct, were untreated or treated with Dox (0.01 µg/ml for 72 h) followed by quantification by imaging. In AP1M1 Dox-inducible construct, AP1M1 expression is linked to mCherry through an IRES element (Supplementary Fig. 5a) and is regulated by Dox. Every condition had 2 technical replicates measured in 3 independent experiments. Different color shades represent different clones. Statistical testing was performed using one-way ANOVA with Tukey post-test (exact p-value = 0.0069). f Intensity of TGN46 staining was quantified by imaging in cytoplasm of wild type and AP1M1 KO cells expressing AP1M1 Dox-inducible construct, treated with Dox (0.01 µg/ml for 72 h). Every condition had 2 technical experiments measured in 3 independent experiments. Different color shades represent different clones. Statistical testing was performed using one-way ANOVA with Tukey post-test (exact p-value = 0.0027). g Representative images of TGN46 immunostaining (depicted in magenta) in wild type and AP1M1 KO rescue cells treated with Dox (0.01 µg/ml for 72 h). MCherry (depicted in yellow) serves as a readout of the extent of the AP1M1 expression. Nuclei are counterstained with Hoechst (depicted in cyan). The far right panel shows the same images as the middle panel, without the mCherry signal, but with mCherry positive cells marked with a yellow asterisk. h Number of de-quenched DQ-red BSA positive spots per cell (wild type and AP1M1 KO U2OS) was quantified by imaging over time (75–375 min) after treatment with DQ-red BSA. Horizontal line represents the median. Every condition had 2 technical replicates measured in 3 independent experiments. Different color shades represent different clones. Statistical testing was performed using one-way ANOVA with Dunnett’s post-test (exact p-value = 0.0062). i Representative images from (h) showing the fluorescent signal (dequenched DQ-red BSA in lysosomes, depicted in magenta) in wild type and AP1M1 KO cells. Nuclei are counterstained with Hoechst (depicted in cyan). Scale bar corresponds to 50 µm. **p-value < 0.01; ***p-value < 0.001; ****p-value < 0.0001. Source data are provided as a Source Data file.

Fig. 7. Ap1m1 expression modulates ASO activity in vivo.

a Schematic representing the in vivo study design. Adult transgenic mKate2.ssEGFP.HBB mice carrying the splice reporter were injected subcutaneously with the GalNAc-siRNA targeting Ap1m1 or Ahsa1 (15 mg/kg) for 15 days before the administration of the SS-ASO, also via subcutaneous route. After 3 days, mice were sacrificed, and the liver and kidneys were collected for subsequent analysis. Created in BioRender. Malong, L. (2025) https://BioRender.com/gy7da4l. b Fold change of Ap1m1 and Ahsa1 mRNA, in liver (left panels) and kidney samples (right panels) collected at day 18, was quantified by qRT-PCR and normalized by housekeeping gene Gapdh. Horizontal lines show the mean, and each dot represents an animal. N = 16 animals per group, separated to 4 independent experiments. Statistical testing was performed with two-way ANOVA with Dunnett’s post-test. c GFP protein expression, from the splice reporter (normalized by total protein level) was quantified at day 18. To account for the variability in Ap1m1 KD levels, the animals injected with Ap1m1 siRNA have been segregated into insufficient KD (less than one third knockdown) or sufficient KD (more than two third knockdown) groups. Each dot represents an animal, horizontal lines show the means. N = 12 animals per group, separated to 4 independent experiments. Analysis was performed with one-way ANOVA with Dunnett’s post-test (exact p-value = 0.0095). *p-value < 0.05; **p-value < 0.01 ****p-value < 0.0001. Source data are provided as a Source Data file.