Integrative Proteogenomics and Forward Genetics Reveal a Novel Mitotic Vulnerability in Triple-Negative Breast Cancer

Abstract

Triple-negative breast cancer (TNBC) is an aggressive subtype of breast cancer with few effective targeted therapies. Taxanes and other microtubule-targeting agents (MTA) are first-line chemotherapies for TNBC; however, the molecular mechanisms that underlie TNBC taxane sensitivity are largely unknown, preventing selection of taxane-responsive patients and development of more selective therapeutic strategies. In this study, we identified tumor-selective vulnerabilities in TNBC harboring inactivation of the tumor suppressor PTPN12 by integrating proteogenomic characterization and synthetic lethality screening. We discovered that PTPN12 inactivation drives mitotic defects through aberrant hyperactivation of the ubiquitin ligase complex APCFZR1, a critical regulator of the cell cycle. Consistent with the mitotic stress caused by PTPN12 inactivation in TNBC cell lines, tumors harboring loss of PTPN12 exhibit heightened sensitivity to taxane chemotherapy. Collectively, these data suggests that PTPN12 inactivation may drive chromosomal instability and favorable MTA response in TNBC-two prominent features of the disease with unclear mechanistic etiology.

Significance: Many TNBCs respond to MTAs, but the underlying cause(s) of this sensitivity remain poorly understood. Herein, we discover that the tumor suppressor PTPN12 regulates mitotic fidelity and MTA sensitivity in a large subset of patients with TNBC, which has significant implications for the use of MTAs in breast cancer.

Figure 1:. Integration of proteo-genomics and forward genetics to reveal novel vulnerabilities in TNBC.

To identify selective vulnerabilities of PTPN12-deficient TNBC, we integrated information from two orthogonal datasets. (A) First, we performed proteogenomic characterization of a cohort of TNBC PDX models that were classified into two groups based on their PTPN12 expression status. This allowed for the identification of pathways that are specifically dysregulated in PTPN12-deficient TNBC. (B) Second, we performed a genetic screen to identify genes that are selectively essential in PTPN12-deficient cells. Immortalized human mammary epithelial cells (HMECs) engineered with a doxycycline-inducible PTPN12 shRNA were transduced with an shRNA library targeting 2,569 genes. Genes that are selectively lethal in the PTPN12-deficient state represent candidate tumor-selective vulnerabilities of PTPN12-deficient TNBC. (C) These two datasets were integrated using geneset enrichment analysis (GSEA) to identify genesets that are mutually enriched and represent pathways that are both selectively dysregulated in PTPN12-deficient TNBC models and selective vulnerabilities of PTPN12-deficient cells.

Figure 2:. PTPN12-deficient models exhibit downregulation of cell cycle and mitotic pathways.

(A) ssGSEA was performed on transformed p-values obtained by moderated 2-sample t-test of PTPN12-normal and -deficient models. Heatmap shows normalized enrichment scores (NES) of significant gene-sets (FDR <0.05) enriched in RNA-seq, proteome and phosphoproteome datasets. (B) PTM-SEA analysis based on phosphoproteomics shows significant (p-value <0.01) upregulation of PTPN12-regulated tyrosine kinase signatures (e.g. EGFR, MET, ABL1) and downregulation of signatures of mitotic kinases including PLK1, CDK1, CDK2, and AURKB. Normalized enrichment score (NES) of each kinase signature in PTPN12-deficient models relative to PTPN12-normal models is plotted on the x-axis, while −log₁₀ transformed p-values of this enrichment are plotted on the y-axis. (C) Phosphotyrosine sites in PTPN12-regulated RTKs are hyperphosphorylated in PTPN12-deficient PDX models. Phosphoproteomic data was mined for phosphosites from RTKs that were previously shown to be regulated by PTPN12. The abundances of phosphotyrosine peptides from these RTKs across all 14 PDX models are plotted in the heatmap, with red indicating high abundance peptides and blue representing low abundance peptides. (D) Proteins involved in cell cycle progression and mitosis are downregulated in PTPN12-deficient PDX models. Heatmap of normalized protein abundance of cell cycle- and mitosis-associated proteins plotted for PTPN12-deficient and PTPN12-normal models. (E) Heatmap of normalized abundance of the 20 most significantly downregulated phosphopeptide substrates of cell cycle- and mitosis-associated kinases (PLK1, CDK1, CDK2, or AURKB) enriched in the PTM-SEA. All phosphopeptides shown are significantly downregulated in PTPN12-deficient models.

Figure 3:. PTPN12-deficient cells are vulnerable to perturbation of mitotic regulators.

(A) Volcano plot illustrating results of synthetic lethal screen to identify PTPN12-selective growth modifiers. Combined PTPN12-selective growth effect (log₂) of each gene is plotted on the x-axis, and Fisher-transformed P value of each gene is plotted on the y-axis. Genes with at least three significant shRNAs (P value < 0.1 and abs(log₂ FC) > 0.5) are colored red. The size of each point represents the fraction of shRNAs targeting that gene that were significant. (B) PLK1 and BUB1 knockdown is PTPN12-synthetic lethal, while knockdown of multiple APC components is PTPN12-synthetic beneficial. PTPN12-selective growth effect of two-independent shRNAs from the genetic screen are plotted for each gene. shRNAs targeting PGK2, a non-essential testis-specific isoform of phosphoglycerate kinase, are shown as a negative control. (C) Network analysis of significant hits from the shRNA screen and proteomics. Nodes represent genes that are both PTPN12-selective growth modifiers and dysregulated in PTPN12-deficient PDX models, whereas edges indicate protein-protein or text-mining interactions between these genes obtained from STRING. The PTPN12-selective growth effect of each gene is indicated by the color of the node border (blue = lethal, red = beneficial), the difference in average protein abundance between PTPN12-normal and -deficient models is indicated by the fill color of the node (blue = down in PTPN12-deficient models, red = up in PTPN12-deficient models), and the size of each node represents the betweenness centrality as calculated by Cytoscape. A sub-network of genes involved in mitotic regulation are indicated by a black box. (D) Comparison of GSEA analysis of proteomic and genetic screen datasets showing enrichment of cell cycle- and mitosis-related pathways in both datasets. Transformed enrichment scores for gene sets in the genetic screen are plotted along the x-axis, whereas scores in the proteomic analysis are plotted on the y-axis. Transformed enrichment scores were calculated by taking the product of the log₂[abs(NES)] and abs[log₁₀(FDR)] for a given pathway. Pathways with a transformed enrichment score > 8 are not plotted. (E) REACTOME pathways related to regulation of cell-cycle progression and mitotic checkpoint signaling are enriched in both the genetic screen and proteomic datasets. The combined enrichment score of the top 20 mutually enriched pathways is plotted; 9 of the top 20 are related to cell cycle and mitotic regulation. Combined enrichment scores were calculated by taking the sum of the transformed enrichment scores in the genetic screen and proteomics from (D).

Figure 4:. PTPN12 deficiency causes aberrant degradation of APC^FZR1 substrates.

(A) PTPN12 status correlates with APC substrate abundance in TNBC PDX models. Quantitative proteomics was performed to measure abundance of validated APC substrates (https://slim.icr.ac.uk/apc/) across 14 PDX models. Mean substrate abundance across 3 biological replicates is shown, along with the PTPN12 expression levels (measured by IHC) and the correlation between PTPN12 expression and the abundance of each APC substrate. (B) PTPN12-deficient PDX models have decreased phosphorylation of specific FZR1 residues. Normalized abundance of FZR1 phosphopeptides (S40, T121, or mean of all sites) is plotted on the y-axis for PTPN12-deficient and -normal PDX models. (C) Regulators of CDK2 activity are PTPN12-selective growth modifiers. Barplots showing the PTPN12-selective growth effect (log₂ fold change in cell number) of two independent shRNAs targeting WEE1, CDC25A, or CDC25B in the PTPN12-synthetic lethal screen. Multiple shRNAs targeting each gene were found to be selectively lethal (CDC25A, CDC25B) or beneficial (WEE1) to PTPN12-deficient cells. shRNAs targeting PGK2, a non-essential testis-specific isoform of phosphoglycerate kinase, are shown as a negative control. (D) PTPN12 depletion impairs CDK2 activation. HMECs were engineered with a CDK2 activity reporter (DHB-mVenus) and monitored by live-cell imaging after transfection with control or PTPN12-targeting siRNAs. In control cells, CDK2 activity (mean nuclear mVenus intensity) accumulates steadily after NEB (vertical dashed line), but this activation is impaired in siPTPN12-transfected cells. (E) PTPN12 inactivation causes accumulation of G1 and early S phase cells. The proportions of CDT1+/GMNN− cells, which represent late G1 phase cells, and CDT1+/GMNN+ cells, which represent early S phase cells, are both increased upon siRNA-mediated PTPN12 depletion in HMECs. (F) PTPN12 depletion causes delayed accumulation of GMNN during S phase. HMECs engineered with AmCyan-GMNN were transfected with control or PTPN12-targeting siRNAs to induce depletion of PTPN12. 48 hours after transfection, cells were monitored by live cell imaging for 24 hours to track the abundance of GMNN throughout the cell cycle. Mean intensity of AmCyan-GMNN is plotted on the y-axis as a function of time since nuclear envelope breakdown (NEB). P-value determined using Student’s t-test at 13 hours post-NEB. (G) Delayed GMNN accumulation in PTPN12-deficient cells is suppressed by inhibition of WEE1. Data from the same experiment as (F) is shown, but in cells treated with 30nM of the WEE1 inhibitor adavosertib. (H) PTPN12 deficiency causes aberrant depletion of PLK1 that is rescued by APC inhibition. shPTPN12-HMECs were treated with vehicle or doxycycline for 72 hours and levels of PLK1 and pH3(S10) were then measured by immunofluorescence. Representative images are shown illustrating that PTPN12 deficiency leads to reduction of PLK1 expression specifically in pH3 positive cells and that PLK1 abundance is rescued by treatment with 20nM proTAME. (I) APC inhibition rescues the aberrant depletion of PLK1 caused by deficiency of PTPN12. Quantification of imaging from (H) comparing intensity of PLK1 in pH3-positive cells (>70^th percentile of pH3 intensity) treated with vehicle or 20nM proTAME.

Figure 5:. PTPN12-deficient cells exhibit mitotic defects dependent on APC^FZR1 hyperactivation.

(A) PTPN12 knockdown causes various types of mitotic defects. shPTPN12-HMECs were transduced with H2B-GFP to visualize mitotic chromosomes and imaged for 24hrs +/− doxycycline. Percentage of abnormal mitotic events; including metaphase arrest, segregation errors, and spindle defects, observed across 100 mitoses in each state is plotted on the y-axis. (B) PTPN12 knockdown causes mitotic defects in MDA-MB-231 TNBC cells. MDA-MB-231 cells, which have normal PTPN12 expression, were engineered with H2B-GFP and a dox-inducible shRNA targeting PTPN12. Live cell imaging was performed to quantify mitotic abnormalities 24 hours after treatment with 1μM doxycycline to induce PTPN12 knockdown. (C) Re-expression of PTPN12 rescues mitotic defects in PTPN12-deficient SUM159 TNBC cells. SUM159 cells, which have low expression of PTPN12, were engineered with H2B-GFP and a dox-inducible PTPN12 cDNA. Live cell imaging was performed to quantify mitotic abnormalities 24 hours after treatment with 500nM doxycycline to induce PTPN12 expression. (D) Metaphase arrest in PTPN12-deficient cells is caused by spindle assembly checkpoint (SAC) activation. Live-cell imaging of shPTPN12-HMECs was performed as in (A) in the presence and absence of 250nM reversine, an inhibitor of the critical spindle checkpoint kinase TTK. Treatment with reversine completely rescues the increase in frequency of metaphase arrest observed in PTPN12-deficient cells. (E) APC inhibition rescues the mitotic defects caused by depletion of PTPN12. H2B-GFP shPTPN12 HMECs were treated with vehicle or doxycycline overnight then treated with DMSO or 20nM proTAME and monitored by live cell imaging for 24 hours to quantify abnormal mitotic events. (F) Mitotic defects in PTPN12-deficient cells are driven by loss of PLK1. H2B-GFP HMECs engineered with a dox-inducible PLK1 cDNA were transfected with control or PTPN12-targeting siRNAs, treated with 250nM doxycycline for 24 hours to induce PLK1 expression, then monitored by live cell imaging to quantify the frequency of mitotic abnormalities. (G) PTPN12-deficient PDX models exhibit increased chromosome instability (CIN). Copy-number estimates based on whole-exome sequencing data were used to calculate the number of large copy-number variations (CNVs; >5Mb and abs(segment mean) > 1)) in each of the 14 PDX models that were proteogenomically characterized (mean +/− s.e. is plotted). PDX models with normal PTPN12 expression (IHC ≥2) have significantly fewer large CNVs than PTPN12-deficient models (IHC <2). (H) Heatmap illustrating the prevalence and distribution of large CNVs (>5Mb and abs(segment mean >1)) in PTPN12-normal and -deficient PDX models. Red segments represent large duplications/amplification and blue segments represent large deletions.

Figure 6:. PTPN12-deficient primary TNBC are sensitive to taxane chemotherapy.

(A) APC^FZR1 hyperactivation drives mitotic defects in PTPN12-deficient TNBC. The defects in cell cycle and mitotic progression observed in PTPN12-deficient cells suggests that they may exhibit increased sensitivity to other mitotic perturbations, such as taxane-based chemotherapy. (B) APC inhibition suppresses the increase in mitotic defects caused by taxane treatment in PTPN12-deficient cells. shPTPN12-HMECs engineered with H2B-GFP were assessed using live cell imaging to quantify the frequency of mitotic abnormalities. Treatment with .5nM docetaxel caused an increased number of abnormal mitotic events in PTPN12-deficient cells and treatment with 20nM proTAME suppressed these defects. (C) PTPN12-deficient PDXs exhibit greater response to docetaxel. 11 PDX models (5 PTPN12-deficient, 6 PTPN12-normal) were engrafted and grown to 200–300mm³, then randomized +/− 20mg/kg docetaxel. Representative response curves for 3 PTPN12-normal (top) and 3 PTPN12-deficient PDX models are shown (bottom). Tumor regressions demarcated with RECIST criteria. (D) Summary of docetaxel response of PDX models from (C). Docetaxel response for each model is expressed as mean log₂ fold-change in tumor volume at day 28 compared to day 0 relative to vehicle group. Dashed line indicates partial response according to RECIST criteria. (E) PTPN12 protein expression is lost in >25% of TNBC. Barplot showing the proportion of patient specimens with low PTPN12 expression (IHC <=2) in the ER+/HER2−, HER2+, and double-negative (DN) subtypes from CALGB9344. (F) Patients with PTPN12-deficient tumors show greater benefit from the addition of adjuvant paclitaxel treatment in CALGB9344. Clinical outcomes in response to a standard AC chemotherapy regimen or a standard AC regimen followed by several cycles of paclitaxel therapy were assessed in CALGB9344. PTPN12 status was retrospectively measured in patient specimens from this trial and the association between PTPN12 status and clinical outcome in response to each treatment regimen was analyzed using a Cox proportional hazards model. Patients with high expression of PTPN12 (IHC > 2) did not show an improvement in DFS after addition of paclitaxel therapy (left panel), whereas patients with PTPN12-deficient tumors (IHC <= 2) did exhibit a trend toward improved DFS after paclitaxel treatment (right panel), although it was not statistically significant due to the small number of TNBC cases available for analysis. This trend was especially pronounced at >4 years post-treatment.