Positive selection CRISPR screens reveal a druggable pocket in an oligosaccharyltransferase required for inflammatory signaling to NF-κB

Abstract

Nuclear factor κB (NF-κB) plays roles in various diseases. Many inflammatory signals, such as circulating lipopolysaccharides (LPSs), activate NF-κB via specific receptors. Using whole-genome CRISPR-Cas9 screens of LPS-treated cells that express an NF-κB-driven suicide gene, we discovered that the LPS receptor Toll-like receptor 4 (TLR4) is specifically dependent on the oligosaccharyltransferase complex OST-A for N-glycosylation and cell-surface localization. The tool compound NGI-1 inhibits OST complexes in vivo, but the underlying molecular mechanism remained unknown. We did a CRISPR base-editor screen for NGI-1-resistant variants of STT3A, the catalytic subunit of OST-A. These variants, in conjunction with cryoelectron microscopy studies, revealed that NGI-1 binds the catalytic site of STT3A, where it traps a molecule of the donor substrate dolichyl-PP-GlcNAc₂-Man₉-Glc₃, suggesting an uncompetitive inhibition mechanism. Our results provide a rationale for and an initial step toward the development of STT3A-specific inhibitors and illustrate the power of contemporaneous base-editor and structural studies to define drug mechanism of action.

Keywords: CRISPR screening; N-glycosylation; NF-κB; Toll-like receptors; inhibition mechanism; oligosaccharyltransferase.

Figure 1.. Validation of positive selection assay to identify genes required for NFκB signaling

(A) Reporter schematics. DCK* = deoxycytidine kinase. IRES= internal ribosomal entry site. GFP = green fluorescent protein (GFP). UBC = ubiquitin C gene promoter. (B) GFP intensity, as determined by fluorescence activated cell sorting (FACS), of the indicated NALM-6 reporter cells in which TNFAIP3 was (sgTNFAIP3) or was not (sgctrl) eliminated using CRISPR/Cas9. Where indicated the cell were treated with 10 μg/mL LPS for 24 hours prior to FACS. (C) Immunoblot analysis of reporter cells treated as in (B). Where indicated cells also underwent CRISPR-based editing with a sgRNA against MYD88 or control sgRNA. (D) Kill curves of cells in panel (B) after 7 day exposure to BVdU at the indicated concentrations and, where indicated, 10 μg/mL LPS. Data were normalized to cells unexposed to BVdU and LPS. n=3. (E and F) FACS-based competition assay of the indicated NALM-6 reporter cells. The cells were stably infected to produce an MYD88 sgRNA together with mCherry or a nontargeting control sgRNA together with blue fluorescent protein (BFP), mixed 1:99, respectively, and treated with 10 μM BVdU (or DMSO) and 10 μg/mL LPS for 20 days prior to FACS.

Figure 2.. Identification of genes required for NFκB pathway activity in response to various stimuli

(A) Volcano plots (left and right panels) showing genes whose sgRNAs were enriched or depleted over time after stable infection of the indicated reporter lines with a whole genome sgRNA library and 14 day exposure to 10 μM BVdU and 10 μg/mL LPS. Center panel highlights all genes scoring in the NFκB-DCK* arm of the screen with a p-value of less than 10⁻⁵. n = 2 biological replicates for each of the whole genome screens. (B) Schematic of NFκB pathway highlighting genes that scored in the positive selection whole genome screen. (C) Volcano plots showing genes whose sgRNAs were enriched or depleted over time after stable infection of the indicated reporter lines with a subgenomic library targeting the top 500 genes from NFκB arm of the whole genome screen in (A), plus controls, and exposure to 10 μM BVdU and either 10 μg/mL LPS, 1 ng/ ml TNFα, or 1 μg/mL CpG oligodinucleotide for 16 days. Genes highlighted in red originally scored with a p-value of less than 10⁻⁵ in the primary LPS-based whole genome screen. n = 2 biological replicates for each of the sublibrary screens. (D) GFP induction, as measured by FACS, of NFκB-DCK* reporter cells that underwent CRISPR/Cas9 knockout with the indicated sgRNAs (2 independent sgRNA/gene) and then treated with NFκB agonists as in (C) for 24 hours. Guide rank is from the NFκB arm of the whole genome screen in (A). (E) Quantitative real-time PCR of the indicated NFκB-responsive mRNAs in NALM-6 cells treated with 10 μg/mL LPS or vehicle for 24 hours. Fold-induction of the indicated transcripts was determined by quantitative RT-PCR.

Figure 3.. STT3A is required for glycosylation and trafficking of TLR4 to the cell surface

(A) Flow cytometry for cell surface TLR4 in TNFAIP3−/− NFκB-DCK* NALM-6 cells overexpressing exogenous TLR4 (left panel) and in parental NALM-6 cells with endogenous TLR4 expression (right panel) after lentiviral CRISPR/Cas9 knockout of indicated genes. MFI = median fluorescence intensity. (B) Immunoblot analysis of whole cell extracts and cell surface proteins from TNFAIP3−/− NFκB-DCK* NALM-6 cells after CRISPR/Cas9 gene knockouts with the indicated sgRNAs. Cell surface proteins were captured on streptavidin agarose after surface biotinylation. (C) Immunoblot analysis of cell extracts from the cells in (B) after the extracts were treated with the N-glycosidase PNGase F (PNGase) or endoglycosidase H (EndoH) ex vivo. (D) Immunoblot of HEK293T clones that exogenously express HA-tagged TLR4 and, where indicated, wereCRISPR/Cas9 edited to lack STT3A or STT3B. Data shown are representative of three independent clones for each genotype. GLUT1 = glucose transporter 1. (E) GFP induction, as measured by FACS, of NALM-6 NFκB-DCK* reporter cells with knockout of the indicated genes, with or without treatment with the NFκB agonist LPS at 10 μg/mL for 24 hours. (F and G) Immunoblots of RAW264.7 murine macrophage cell line stably transduced with Cas9 and the indicated sgRNAs. In (F) cells were treated, where indicated, with 10 μg/mL LPS for 24 hours. (H) NALM6 TNFAIP3−/− NFkB-DCK* reporter cells stably overexpressing either TLR4 or CD16-TLR4 were stimulated or not with LPS 10 μg/mL for 24 hours, followed by FACS for GFP intensity as a marker of NFkB pathway activity. The left panel shows data from STT3A wild-type cells while the right panel shows data from STT3A knockout cells. MFI = median fluorescence intensity

Figure 4.. Regulation of TLR4 and NFκB by OST-A requires STT3A catalytic activity and can be inhibited with drug-like molecules

(A) Immunoblot of HEK293T cells stably infected to express HA-tagged TLR4 and the indicated HiBIT-tagged STT3A proteins after CRISPR-based editing with the indicated sgRNAs. (B) Immunoblot of TNFAIP3−/− NFκB-DCK* NALM-6 cells overexpressing TLR4-HA and treated with NGI-1. (C) Immunoblot of RAW264.7 cells treated with NGI-1 and/or LPS. (D) GFP induction, as measured by FACS, of TNFAIP3−/− NFκB-DCK* reporter cells treated with the indicated concentration of NGI-1 for 48 hours followed by the addition of 10 μg/mL LPS for 24 hours. GFP values were normalized to values for cells not treated with LPS or NGI-1. (E) Kill curves of TNFAIP3−/− NFκB-DCK* reporter cells that were pretreated, where indicated, with NGI-1 for 24 hours and then treated with 10 μM BVdU in the presence or absence of 10 μg/mL LPS. Data were normalized to cells unexposed to BVdU and LPS. n=3. (F) Immunoblot analysis of whole cell extracts and cell surface proteins from TNFAIP3−/− NFκB-DCK* NALM-6 cells after CRISPR/Cas9 gene knockouts with the indicated sgRNAs. Cell surface proteins were captured on streptavidin agarose after surface biotinylation. Where indicated cells were treated with NGI-1 (1 μM or 10 μM) or tunicamycin 1 μM for 24 hours. Mobility shifts in proteins between the whole cell fraction and the surface fraction may be due to different loading buffers used for these fractions (see Methods). (G) Structures of NGI-1 and NGI-235. (H and I) Anti-HA immunoblot analysis (H) and anti-TLR4 FACS (I) to detect cell surface TLR4 of TNFAIP3−/− NFκB-DCK* NALM-6 cells expressing exogenous TLR4-HA and treated with the indicated concentrations of either NGI-1 or NGI-235 for 24 hours. GLUT1 = glucose transporter 1, MFI = median fluorescence intensity.

Figure 5.. Base editor screens identify mutations in STT3A that render it resistant to NGI-1

(A) Workflow of base editor screen done in TNFAIP3−/− NFκB-DCK* NALM-6 cells, highlighting the dual approach using a library of guides tiling STT3A that were co-introduced with either an A→G base editor or a C→T base editor. (B and C) Enrichment (y-axis) of STT3A sgRNAs [placed on the x-axis according to their position along the STT3A gene (or arbitrary number for control guides)] recovered from the top 1% GFP+ cells in the NGI-1 treated arm as compared to the top 1% GFP+ cells in the DMSO treated arm for the STT3A A→G base editor screen (B) and STT3A C→T base editor screen (C). (D and E) NGS of genomic DNA after introduction of the top-scoring sgRNA in the A→G base editor screen (D) or C→T base editor screen (E) into the screening cell line, TNFAIP3−/− NFκB-DCK* NALM-6 cells. The pie chart shows percentage of NGS reads that contained substitutions and the nucleotide bar graph shows the frequency of individual substitutions at the indicated nucleotides, along with canonical amino acid sequence above and validated mutated amino acid sequence below. Note that the sgRNAs in both panels align with the complementary DNA strand so substitutions introduced are T→C and G→A in this orientation. A naturally occurring A/G SNP (rs2241502) is present in panel (D) as well. Analysis done with CRISPResso2. (F and G) Anti-TLR4 FACS (F) to detect cell surface TLR4 and anti-HA immunoblot analysis (G) of STT3A−/− HEK293T cells expressing exogenous TLR4-HA that were rescued with the indicated STT3A variants and treated with NGI-1 for 24 hours.

Figure 6:. Cryo-EM structure of OST-A with NGI-1 bound

(A) OST-A structure is shown in ribbon representation. Subunits are colored individually and labeled. Bound NGI-1 is shown as orange spheres. Bound LLO (Dol-PP-GlcNAc₂Man₉Glc₃) is shown as black sticks. The inset shows a close-up view of NGI-1 and LLO binding sites, with STT3A shown in ribbon. NGI-1 is shown in orange sticks, and bound LLO in black sticks. The dolichyl tail and the glycan moiety of the LLO are indicated and labeled. Manganese (II) ion is shown as a pink sphere. (B) Close-up view of the bound LLO in the presence of NGI-1. OST subunits are shown in ribbon representation with subunits colored as in (A). NGI-1 and LLO are shown as the inset in (A), with the dolichyl tail and glycan parts indicated and labeled. The water molecule is shown as a red sphere and the manganese (II) ion as a pink sphere. EM density is shown as a blue mesh. (C) Ribbon representation of STT3A with bound LLO and NGI-1. The three clusters identified by the base-editor screen are indicated in boxes and labeled. NGI-1 and LLO are shown as in (A). Transmembrane helix 9 (TM9) is represented as a green cylinder and labeled. (D) Close-up of NGI-1 binding to the STT3 subunit. Residues involved in binding are shown as sticks. EM density, NGI-1, LLO, water and manganese (II) ion are shown as in (B) and labeled. (E) Anti-TLR4 FACS to detect cell surface TLR4 after 24 hours of NGI-1 treatment of STT3A−/− 293T cells exogenously expressing TLR4-HA and the indicated STT3A variants.

Figure 7:. Inhibition mechanism of NGI-1

(A) Comparison of LLO structure in the LLO-primed state (left, light blue, PDB:8AGB), in a ternary complex with LLO and peptide bound (middle, light brown, PDB:8AGC) and in the NGI-1 and LLO-bound, inhibited state (right, green, this study, PDB:8PN9). LLO is shown as sticks and the GlcNAc₂ moiety is indicated and labeled. Residues interacting with the glycan are shown as sticks and labeled. Manganese (II) ion is shown as a pink sphere. (B) Superposition of LLO in LLO-primed (PDB:8AGB) and NGI-1 bound (this study, PDB:8PN9) structures. Branch A of the LLO is shown in sticks (light blue for LLO-primed and black for NGI-1 bound). On the right, a schematic representation of the LLO glycan moiety is shown using the official nomenclature for glycans (SNFG). Branches A, B and C are labeled. (C) Proposed mechanism for OST-A inhibition by NGI-1. In active OST, binding of an acceptor peptide to LLO-primed enzyme leads to a ternary complex proceeding in catalysis (left). In contrast, binding of NGI-1 to LLO-primed OST-A displaces the first glycan units of the LLO, inducing a conformation that is not optimal for catalysis (right).