Multiomic profiling of breast cancer cells uncovers stress MAPK-associated sensitivity to AKT degradation

Abstract

More than 50% of human tumors display hyperactivation of the serine/threonine kinase AKT. Despite evidence of clinical efficacy, the therapeutic window of the current generation of AKT inhibitors could be improved. Here, we report the development of a second-generation AKT degrader, INY-05-040, which outperformed catalytic AKT inhibition with respect to cellular suppression of AKT-dependent phenotypes in breast cancer cell lines. A growth inhibition screen with 288 cancer cell lines confirmed that INY-05-040 had a substantially higher potency than our first-generation AKT degrader (INY-03-041), with both compounds outperforming catalytic AKT inhibition by GDC-0068. Using multiomic profiling and causal network integration in breast cancer cells, we demonstrated that the enhanced efficacy of INY-05-040 was associated with sustained suppression of AKT signaling, which was followed by induction of the stress mitogen-activated protein kinase (MAPK) c-Jun N-terminal kinase (JNK). Further integration of growth inhibition assays with publicly available transcriptomic, proteomic, and reverse phase protein array (RPPA) measurements established low basal JNK signaling as a biomarker for breast cancer sensitivity to AKT degradation. Together, our study presents a framework for mapping the network-wide signaling effects of therapeutically relevant compounds and identifies INY-05-040 as a potent pharmacological suppressor of AKT signaling.

Figure 1.. Design and characterization of INY-05–040.

(A) Chemical structures of INY-05–040 and the negative control compound INY-05–040-Neg. (B) Immunoblots for pan-AKT, phospho-PRAS40 (Thr²⁴⁶), total PRAS40, phospho-S6 (Ser^240/244), total S6, and vinculin in T47D cells treated for 5 h with INY-05–040 or INY-03–041 at the indicated concentrations. Data are from a single experiment. (C) Immunoblots for pan-AKT, phospho-PRAS40 (Thr²⁴⁶), total PRAS40, phospho-S6 (Ser^240/244), total S6, and vinculin in T47D cells treated with INY-05–040 (100 nM) or INY-03–041 (100 nM) for the indicated times. Data are from a single experiment. (D) Immunoblots for panAKT, phospho-PRAS40 (Thr²⁴⁶), total PRAS40, phospho-S6 (Ser^240/244), total S6, and vinculin in T47D cells treated with INY-05–040 or GDC-0068 at the indicated concentrations for 24 hours. Data are from a single experiment. (E) Immunoblots for pan-AKT, phospho-PRAS40 (Thr²⁴⁶), total PRAS40, phospho-S6 (Ser^240/244), total S6, and vinculin in T47D cells treated with INY-05–040 (100 nM) or GDC-0068 (100 nM) for 5 h followed by washout for the indicated times. Data are from a single experiment. (F) Immunoblots for pan-AKT, phospho-PRAS40 (Thr²⁴⁶), total PRAS40, phospho-S6 (Ser^235/236), total S6, and vinculin in BT-474 mouse xenograft tumors treated with vehicle (10% DMSO, 25% kleptose), GDC-0068 (12.5 mg/kg), INY-05–040 (25 mg/kg), or INY-03–041 (25 mg/kg) for 3 days, with a terminal treatment 4 h prior to protein harvest. N = 4–6 mice per group as shown. Panels are from the same membrane but have been cropped for clarity, with a solid white line denoting the location of the crop. Additional supporting data related to this figure are included in Fig. S1.

Figure 2.. Multiomic profiling of INY-05–040 and GDC-0068 in T47D breast cancer cells.

(A) Principal component analysis (PCA) projection of the transcriptomic dataset, comprising n=3 biological replicates per treatment (DMSO; degrader: 100 nM INY-05–040; 500 nM GDC-0068; NegCtrl: 100 nM INY-05–040-Neg) and time point (5 h and 10 h). Ellipses are drawn around each group at 95 % confidence level. The first three independent axes (PCs) of highest variation are shown. (B) Number of differentially up- and downregulated transcripts (absolute fold-change ≥ 1.3) following differential gene expression analysis (FDR ≤ 0.05) across the indicated comparisons. Comparisons are relative to the corresponding DMSO-treated control; for example, Neg.Ctrl.10h refers to the effect of 10 h treatment with INY-05–040-Neg vs 10 h treatment with DMSO. The exception is “diff.time.DMSO” which evaluates differential expression as a function of time in culture (treatment with DMSO for 10 h versus treatment with DMSO for 5 h). (C and D) Gene set enrichment analysis (GSEA) on the mSigDb HALLMARK collection, based on the ranked t values from all genes for the indicated treatments relative to the corresponding DMSO-treated controls. Gene sets are labelled if the absolute normalized enrichment score (NES) exceeds 1 and the adjusted p-value falls below 0.05 (FDR). (E) Spearman’s correlation analysis of transcription factor (TF) activity predictions from RNAseq data in cells treated for 10 h with either degrader or GDC-0068. TF footprint analyses were performed with DoRothEA. SREBF1 (protein name: SREBP1) and SREBF2 (protein name: SREBP2) activity predictions are highlighted due to their divergence between degrader and GDC-0068-treated cells, with lower activity predictions observed only in GDC-0068-treated cells. (F) Spearman’s correlation analysis of GSEA-derived NES for individual HALLMARK gene sets, based on RNAseq data from cells treated for 10 h with either degrader or GDC-0068. “CHOLESTEROL HOMEOSTASIS” and “ANDROGEN RESPONSE” hallmark gene sets are highlighted as having positive and negative NES in Degrader- and GDC-0068-treated cells, respectively. (G) Spearman’s correlation analysis of transcription factor (TF) activity predictions from PRO-seq data in T47D cells treated for 5 hours with either degrader and GDC-0068 relative to DMSO-treated control. TF activity predictions were calculated from t values from all genes following differential gene expression analysis (FDR ≤ 0.05; n = 2 biological replicates per treatment). (H) Spearman’s correlation analysis of GSEA-derived NES for individual HALLMARK gene sets, based on PROseq data from (G). (I) Hierarchical clustering (Euclidean distance) of differential metabolite abundance (FDR ≤ 0.05) following 24-h treatments of T47D with either AZD 5383 (capivasertib; catalytic pan-AKT inhibitor; 2 μM), degrader (INY-05–040; 100 nM), GDC-0068 (catalytic AKT inhibitor; 500–750 nM), MK2205 (allosteric pan-AKT inhibitor; 1 μM) or NegCtrl (INY-05–040-Neg; 100 nM). Differential abundance analysis was performed relative to DMSO-treated controls (n = 9 replicates per treatment, from 3 biological replicates with 3 technical replicates each). More than 85% of the observed differences in metabolite abundance for a given treatment corresponded to at least a 20% change relative to DMSO-treated cells. Metabolite levels that were changed only upon treatment with Degrader are highlighted. Additional supporting data related to this figure are included in Figs. S2, S3, S4.

Figure 3.. COSMOS-based integration of transcriptomic and metabolomic datasets to identify treatment-specific networks.

(A) Schematic illustrating the principle of COSMOS and the datasets used for multiomic integration and predictions of treatment-specific signaling networks. (B and C) Top degree nodes from degrader- and GDC-0068-specific networks plotted in increasing order. MAPK14 (protein: p38𝛂) is highlighted as a degrader network-specific top degree node. The raw COSMOS networks are included in Fig. S5 (n = 11 independent runs using degrader data; n = 8 independent runs for GDC-0068 data). (D) Complementary GSEA analyses using stress MAPK-related gene sets (mSigDb C2 collection), based on the ranked t values from all genes for the indicated treatments relative to the corresponding DMSO treatment. Gene sets are labelled if the absolute normalized enrichment score (NES) exceeds 1.

Figure 4.. Validation of COSMOS-generated prediction of MAPK stress kinase signaling.

(A and B) Immunoblots for panAKT, phospho-PRAS40 (Thr²⁴⁶), total PRAS40, phospho-p38α (Thr¹⁸⁰/Tyr¹⁸²), total p38α, phospho-c-Jun (Ser⁷³), total c-Jun, phospho-S6 (Ser^240/244), total S6, and vinculin after treatment of BT-474 (A) or T47D (B) cells for the indicated times with DMSO, 100 nM INY-05–040, or 750 nM GDC-0068. Data are from a single experiment. (C) Quantification of total AKT (normalized to vinculin), c-Jun (normalized to vinculin), phospho-c-Jun (pJun) Ser⁷³ (normalized to vinculin), phospho-p38 (pP38) Thr¹⁸⁰/Tyr¹⁸² (normalized to total p38), phospho-PRAS40 (pPRAS40) Thr²⁴⁶ (normalized to total PRAS40), phospho-S6 (pS6) Ser^240/244 (normalized to vinculin) and total S6 (normalized to vinculin), including normalization to the respective DMSO control signal for each time point and cell line. Note that phospho-c-Jun and phospho-S6 were normalized to vinculin given changes in the levels of the respective total proteins. Stippled white lines are added to aid interpretation of samples loaded on the same membrane; white blocks separate samples run on different membranes. Supporting data for additional cell lines (MCF7 and MD-MB-468) are included in Fig. S6. (C) Cytotoxicity index assayed using CellTox Green, in BT-474 or T47D cells treated for 24 h with either DMSO or 50 nM JNK-IN-8, followed by 120-h co-treatment with either DMSO, INY-05–040 (100 nM) or GDC-0068 (750 nM). The cytotoxicity index represents cytotoxicity values corrected for background fluorescence and normalized to total signal following chemical permeabilization (used as proxy measure for total cell number). The data are displayed as Cumming plots following bootstrap-coupled estimation of effect size for each condition relative to DMSO. The upper plots display the raw data alongside standard deviations indicated with gapped lines. The plots beneath display the estimated effect sizes, sampling distribution and bootstrap 95% (percentile) confidence intervals. For accurate interpretation, please note differences in y-axis scaling. The data are from a single experiment performed with four technical replicates per condition; data from a separate experimental replicate, including a JNK-IN-8 dose curve, are in Fig. S7.

Figure 5.. Integration of cell line screen data with publicly available omics datasets to identify sensitivity biomarkers for INY-05–040.

(A) Analytical workflow for cell line screen processing and subsequent integration of the growth response metric (GI50adj) with publicly available cell line omics data from the DepMap project. A total of 288 cancer cell lines were profiled with GDC-0068, INY-03–41 and INY-05–040, with the full set of responses included in Fig. S8A. Subsequent integrative analyses focused on breast cancer cell lines only. Note that the applied growth response metric (GI50adj) takes into account cell line growth which is a known confounder in drug sensitivity measurements. The final output corresponds to the concentration of drug that results in 50 % cell growth inhibition. (B) PCA on breast cancer-specific transcriptomics and proteomics data, with coloring according to sensitivity to INY-05–040 (sensitive: GI50adj < 0.5 μM; intermediate: 0.5 μM ≤ GI50adj ≤ 1 μM; resistant: GI50adj > 1 μM; see also Fig. S8B). The PAM50 subtype of each cell line is specified by shape. Transcripts and proteins contributing the most to the observed data structure alongside PC1 and PC2 are labelled. (C and D) Spearman’s correlation analysis of PC1 values for each cell line and the corresponding GI50adj value for INY-05–040. A linear regression line with 95% confidence intervals (shaded area) is included in each analysis, demonstrating that cell line-specific PC1 scores can be used as proxy measures for INY-05–040 sensitivity (meaning the higher the PC1 score, the more resistant the cell line). (E) GSEA (mSigDb HALLMARK gene sets) using transcript and protein loading values alongside the respective PC1, a proxy measure for sensitivity to INY-05–040; FDR ≤ 0.05. NES: normalized enrichment score. (F) A plot of all gene sets that were significantly and positively enriched across PC1 loadings from the DepMap transcriptomic data, and the corresponding NES from the T47D dataset following 10 h treatment with INY-05–040 (see also Fig. 2). Highlighted gene signatures were also statistically significant (FDR ≤ 0.05) in the T47D dataset. (G to I) Spearman’s correlation analysis of JNK1 mRNA expression (G), pJNK1 (T183/Y187) (H) and p-cJun (S73) with the cell line-specific GI50adj value for INY-05–040. A linear regression line with 95% confidence intervals (shaded area) is included in each analysis. Reverse phase protein phosphorylation (RPPA) data were obtained from the DepMap project and subset for the signals of interest. Additional supporting data related to this figure are in Figs. S8 and S9.