Systematic discovery of mutation-directed neo-protein-protein interactions in cancer

Abstract

Comprehensive sequencing of patient tumors reveals genomic mutations across tumor types that enable tumorigenesis and progression. A subset of oncogenic driver mutations results in neomorphic activity where the mutant protein mediates functions not engaged by the parental molecule. Here, we identify prevalent variant-enabled neomorph-protein-protein interactions (neoPPI) with a quantitative high-throughput differential screening (qHT-dS) platform. The coupling of highly sensitive BRET biosensors with miniaturized coexpression in an ultra-HTS format allows large-scale monitoring of the interactions of wild-type and mutant variant counterparts with a library of cancer-associated proteins in live cells. The screening of 17,792 interactions with 2,172,864 data points revealed a landscape of gain of interactions encompassing both oncogenic and tumor suppressor mutations. For example, the recurrent BRAF V600E lesion mediates KEAP1 neoPPI, rewiring a BRAF^V600E/KEAP1 signaling axis and creating collateral vulnerability to NQO1 substrates, offering a combination therapeutic strategy. Thus, cancer genomic alterations can create neo-interactions, informing variant-directed therapeutic approaches for precision medicine.

Keywords: BRET(n); cancer genomics; cancer target; driver mutations; interactome; neoPPI; oncogene; protein-protein interaction; systems biology; tumor suppressor.

Figure 1.. Schematic illustration of qHT-dS platform for systematic differential PPI discovery.

See also Figure S1–2. (A) Cancer driver mutations are translated to mutated residues that may lead to mutation-directed differential PPIs and pathway rewiring. (B) Components of the qHT-dS platform to identify mutation enhanced Go-PPIs, mutation-reduced Lo-PPI and common PPIs. (STAR Method). (C) Square pie chart illustrating the tumor types by the distribution of genes in OncoPPI v2 library. (D) Bubble plot showing the tumor driver mutations, and their frequency and tumor lineage distributions in OncoMut v1 library.

Figure 2.. Evaluation of qHT-dS platform performance

(A) Representative BRET saturation curves from qHT-dS using a known PPI, SPOP ^WT/CUL3, as an example. (B) AUC analysis of the BRET saturation curves. (C) Positive discovery rate of the known WT PPIs using CARINA comparing to the single point analysis. (D) Identification of statistically significant (SS) positive PPIs for both WT and MUT.

Figure 3.. Systematic analysis to prioritize differential PPI to advance neoPPI candidates

(A) Flow chart of differential PPI identification and neoPPI candidate prioritization (B) Volcano plot of the differential score vs. p-values for each WT/MUT PPI pair shows the identified HS-Go- and HS-Lo-PPI. (C) Similarity analysis of HS-Go-PPI binding partners between alleles. (D) Clustering analysis reveals recurrent and unique HS-Go-PPI binding partners. (E) Commonly re-wired pathways that are significantly enriched for the recurrent HS-Go-PPI binding partners from the pathway analysis. (F) Go-PPI network links various driver mutations of diverse genes with commonly rewired oncogenic pathways, with supporting evidence from qHT-dS and GST pulldown validation.

Figure 4.. neoPPI candidates induced by mutations of tumor suppressors and oncogenes.

See also Data S1. (A) Spoke diagram of BRAF^V600E allele-selective neoPPI hubs with experimental evidence from BRET, GST pulldown, NanoPCA, semi-IP of epitope-tagged mutant with endogenous binding partners, and co-IP of endogenous mutant and binding partner complexes. (B) GST pulldown results of the selected Go-PPI, comPPI and/or Lo-PPI in HEK293T cell overexpressing the GST-tagged BRAF^V600E and its comPPI binding partner Venus-Flag-tagged 14–3-3β and NRAS, Lo-PPI with MEK1, and neoPPI with KEAP1. (C) NanoPCA results of BRAF^V600E/KEAP1 PPI in A375 and H1299 cells. The data are presented as mean±SD from four replicates of the NanoPCA screen. *** p≤0.001. (D) Semi-IP of flag-BRAF^V600E and endogenous binding partners, such as KEAP1, in H1299 cells. H1299 cells expressing flag-BRAF^V600E or empty-vector were subjected to anti-flag immunoprecipitation and SDS-PAGE analysis as indicated. (E) Co-immunoprecipitation (co-IP) of endogenous BRAF^V600E and binding partners in cancer cell lines. (i) Co-IP of endogenous BRAF^V600E in an isogenic genetically engineered RKO cells with flag-tag knockin at the endogenous BRAF^V600E loci (flag-RKO: V600E CBP-DYKDDDDK/V600E/+). Parental untagged RKO cells were used as control. (ii) co-IP of KEAP1 with BRAF^V600E in RPMI-7951 cells carrying V600E mutation (left); and co-IPs of endogenous untagged BRAF^V600E with VHL and BCL2L1 in parental RKO cells. The IP complexes were precipitated with indicated antibodies with anti-IgG as control. (F-H) (upper) Spoke diagram of (F) AKT1^E17K, (G) SMAD4^G386D and (H) SPOP^F133L allele-selective neoPPI hubs with experimental evidence as indicated; (lower) GST pulldown validation of selected PPIs as indicated.

Figure 5.. Validation of BRAF^V600E interaction with KEAP1.

See also Data S1ii. (A) BRETⁿ saturation curve of BRAF^V600E/KEAP1 neoPPI from qHT-dS. The data is presented by combining four replicates from the primary qHT-dS in HEK293T cells. (B) Venus-PCA shows the cytoplasm localization of BRAF^V600E/KEAP1 neoPPI using CHL-1 melanoma cell line, transfected with N-Venus-tagged BRAF WT or V600E and C-Venus-tagged KEAP1. Green: reconstituted Venus signal. Blue: nuclear stained with Hoechest. Venus-PCA signal was presented as the normalized fluorescence intensity. (C) Endogenous interaction of BRAF^V600E with KEAP1. The BRAF^V600E/KEAP1 complex was co-IPed with KEAP1 antibody from a pair of isogenic colon cancer cells, parental RKO cells harboring BRAF^V600E alleles and isogenic V600E-knockdown RKO (+/−/−) counterparts, with anti-IgG as control. (D) GST pulldown assay with BRAF and KEAP1 domain fragments. Cell lysate from HEK293T cells expressing the corresponding truncation constructs were subjected to GST pulldown and western blot as indicated. (E) BLI validation of the direct interaction between KEAP1 KELCH domain and BRAF^V600E kinase domain using human recombinant proteins. Interaction between 14–3-3ζ and BRAF^WT kinase domain was used as positive control, with 14–3-3ζ^K49E and BRAF^WT as negative control. (F) Effect of BRAF^V600E kinase inhibition by vemurafenib on BRAF^V600E/KEAP1 neoPPI. HEK293T cells transfected with GST-BRAF^V600E and Venus-flag-KEAP1 were treated with vemurafenib at indicated concentrations for 24 hours.

Figure 6.. Neo-interaction of BRAF^V600E with KEAP1 and its collateral vulnerability

(A-B) BRAF^V600E stabilizes endogenous NRF2. Immunoblot (A) and densitometry analysis (B) showing NRF2 levels upon CHX chase in HEK293T cells overexpressing BRAF WT or V600E. (C) BRAF^V600E activates NRF2 transcriptional activity. HEK293T cells were co-transfected with the NRF2-ARE luciferase reporter and either WT or V600E BRAF. Relative luciferase activity was measured, normalized to internal Renilla luciferase control. The data are presented as mean±SD of three replicates from a representative experiment. ***p<0.001. (D) BRAF^V600E increases NRF2 and its target gene NQO1 protein levels in HEK293T cells transfected with GST-BRAF^V600E versus WT. (E) Effect of BRAF^V600E on NRF2 mRNA levels in HEK293T cells transfected with flag-NRF2 and GST-BRAT WT or V600E. ^nsp>0.05. (F) KEAP1-dependency study of BRAF^V600E-induced increase of NRF2 protein levels. Melanoma cells, CHL-1 (left) and HMCB (right), were transfected with KEAP1-targeting siRNA and BRAF WT or V600E plasmids as indicated. NRF2 protein expression was evaluated using western blot as indicated. (G-H) Representative blots (G) and densitometry analysis (H) of the correlation of NRF2 and its target gene NQO1 protein levels with BRAF genetic status in six melanoma cell lines with WT or V600E BRAF. The data are presented as mean±SD from the densitometry analysis of three representative experiments. **p<0.01. (I) Violin plot of the correlation between NQO1 mRNA levels and BRAF genetic status in 967 cell lines from CCLE dataset. The lines indicate mean±SD. ***p<0.001. (J-K) Effect of BRAF inhibitor, vemurafenib (J), or MEK1 inhibitor, selumetinib (K), on NRF2 and NQO1 protein levels in WM3482 melanoma cell line with BRAF^V600E mutation. (L) Competitive binding between BRAF^V600E and NRF2 to KEAP1. GST pulldown of GST-KEAP1 complex from lysate of HEK293T cells co-transfected with flag-NRF2, and with increasing amounts of flag-BRAF^V600E. (M) Violin plot showing the CERES scores for CRISPR knockout of KEAP1 in 342 cancer cell lines from CCLE dataset. The lines indicate mean±SD. ***p<0.001. (N) Parallel chemogenomic screening in a pair of isogenic MCF10A cell lines. Data were presented as percentage of inhibition in parental MCF10A cells with BRAF WT versus its V600E knock-in counterpart. (O) Chemical structure of deoxynyboquinone (DNQ). (P) AUC analysis of DNQ-induced dose-dependent growth inhibition of twelve cell lines with BRAF WT or V600E. WT: CHL-1, HMCB, MCF10A, MeWO, WM3311 and RKO^+/−/−; V600E: A2058, A375, MCF10A^BRAF-V600E, WM3482, SK-MEL-5 and RKO. Each dot represents one cell line and the data are presented as mean±SD. *p<0.05. (Q) Representative DNQ-induced dose-dependent growth inhibition of CHL-1 and WM3482 cell lines. The experiments were repeated independently three times. The data are presented as mean±SEM from triplicates from a representative experiment. (R) Sequential combination effect of DNQ and vemurafenib in growth inhibition of WM3482 cells carrying BRAF^V600E mutation. DNQ-induced dose-dependent growth inhibition was tested in three conditions: (1) DNQ alone, (2) pretreatment with 100nM vemurafenib for 24h followed by DNQ for 3 days (vemurafenib (1st)+DNQ (2nd)), and (3) pretreatment with DNQ for 24 h followed by 100nM vemurafenib for 3 days (DNQ(1st)+vemurafenib(2nd)). The experiments were repeated independently three times. Data are presented as mean±SEM from triplicates from a representative experiment. (S) AUC analysis of the combination effect of DNQ and vemurafenib in three melanoma cell lines, A2058, SK-MEL-5, and WM3482, with BRAF^V600E mutation. The experiments were repeated independently three times. Data are presented as mean of triplicates from a representative experiment. Each dot represents a cell line and the lines indicate mean±SD. *p<0.05 from paired t-test. (T) Working model of BRAF^V600E/KEAP1 neoPPI in re-wiring KEAP1/NRF2/NQO1 ROS pathway and generating vulnerability to NQO1 substrate.