The BCR-ABL1 Inhibitors Imatinib and Ponatinib Decrease Plasma Cholesterol and Atherosclerosis, and Nilotinib and Ponatinib Activate Coagulation in a Translational Mouse Model

Treatment with the second and third generation BCR-ABL1 tyrosine kinase inhibitors (TKIs) increases cardiovascular risk in chronic myeloid leukemia (CML) patients. We investigated the vascular adverse effects of three generations of TKIs in a translational model for atherosclerosis, the APOE*3Leiden.CETP mouse. Mice were treated for sixteen weeks with imatinib (150 mg/kg BID), nilotinib (10 and 30 mg/kg QD) or ponatinib (3 and 10 mg/kg QD), giving similar drug exposures as in CML-patients. Cardiovascular risk factors were analyzed longitudinally, and histopathological analysis of atherosclerosis and transcriptome analysis of the liver was performed. Imatinib and ponatinib decreased plasma cholesterol (imatinib, −69%, p < 0.001; ponatinib 3 mg/kg, −37%, p < 0.001; ponatinib 10 mg/kg−44%, p < 0.001) and atherosclerotic lesion area (imatinib, −78%, p < 0.001; ponatinib 3 mg/kg, −52%, p = 0.002; ponatinib 10 mg/kg, −48%, p = 0.006), which were not affected by nilotinib. In addition, imatinib increased plaque stability. Gene expression and pathway analysis demonstrated that ponatinib enhanced the mRNA expression of coagulation factors of both the contact activation (intrinsic) and tissue factor (extrinsic) pathways. In line with this, ponatinib enhanced plasma levels of FVII, whereas nilotinib increased plasma FVIIa activity. While imatinib showed a beneficial cardiovascular risk profile, nilotinib and ponatinib increased the cardiovascular risk through induction of a pro-thrombotic state.


Introduction
Background of the APOE*3Leiden.CETP mouse and response to treatment The APOE*3Leiden.CETP transgenic mouse model is an established model for hyperlipidemia and the development of atherosclerosis and was developed by cross-breeding of the APOE*3Leiden transgenic mice and human cholesterol ester transfer protein (CETP) transgenic mice.
The APOE*3Leiden mouse was initially developed as an animal model for Familial Dysbetalipoproteinemia (FD) or type III hyperlipoproteinemia, which is characterized by elevated levels of cholesterol and an increased ratio of cholesterol to triglycerides in the verylow-density-lipoprotein (VLDL) and intermediate-density lipoprotein (IDL) fractions, resulting in the appearance of β-VLDL particles 1 2 . Similarly as in FD patients, in E3L and E3L.CETP mice as a model for mixed dyslipoproteinemia, a major part of plasma cholesterol is contained in the VLDL and VLDL-remnant particles, leading to formation of β-VLDL particles, which further increases after cholesterol feeding. E3L mice were generated by the introduction of a DNA-construct containing the human apoe*3Leiden and apoc1 genes 1 . The primary effect of the dominant E*3-Leiden mutation is an impaired clearance of triglyceride-rich lipoproteins (chylomicron-and VLDL-remnants) caused by a reduced affinity of the apolipoprotein E*3Leiden for the LDLR 3 , whereas overexpression of apolipoprotein C1 inhibits lipolysis. While normal wild-type mice have a very rapid clearance of apoB-containing lipoproteins, APOE*3Leiden mice show an impaired clearance and are thereby mimicking the slow clearance observed in humans, particularly in patients with FD or type III hyperlipoproteinemia. Mice in contrast to humans do not possess CETP. CETP transgenic mice express the human cetp gene under control of its natural flanking regions 4 . CETP transfers cholesteryl ester from HDL to the apoB-containing lipoproteins in exchange for triglycerides, resulting in a more human-like lipoprotein metabolism in APOE*3Leiden.CETP mice. APOE*3Leiden.CETP mice are prone to develop hyperlipidemia and atherosclerosis upon feeding a western type diet containing saturated fat and cholesterol 5 6 . The model responds to all hypolipidemic drugs currently used in the clinic, such as statins, fibrates, ezetimibe, niacin and PCSK9 mAbs, at similar dosages and in a similar way to humans [7][8][9][10][11][12][13][14][15][16] . Furthermore, the model responds well to blood pressure lowering drugs 17 18 .
In the present study, female APOE*3Leiden.CETP mice are used because they are more susceptible to cholesterol-containing diets by having higher plasma cholesterol and triglyceride levels 19 and therefore develop atherosclerotic lesions.

Animals
Female APOE*3Leiden.CETP transgenic mice (9 to 14 weeks of age) were used. During the study, mice were housed under standard conditions with a 12-h light-dark cycle and had free access to food and water. Body weight, food intake and clinical signs of behavior were monitored regularly during the study. For the (cardio)vascular risk factor-atherosclerosis study, the number of animals per group was calculated using a probability of 0.05. Based on our experience from previous studies, we expected to have a variance of 15% (sigma 40%) in plasma lipids and a minimal effect of treatment of 30%, resulting in 15 animals per group.
Animal experiments were approved by the Institutional Animal Care and Use Committee of The Netherlands Organization for Applied Research under registration number 3557.

PK analysis and plasma drug concentrations
EDTA plasma samples collected during the 24-hour PK study and at week 16 of the (cardio)vascular risk factor-atherosclerosis study were subsequently stored at -80°C until analysis. Thawed plasma samples (50μL) were de-proteinized with two volumes of acetonitrile containing an appropriate src-inhibitor chemotype as internal standard for each analyte.

Biochemical analyses and blood pressure
EDTA plasma samples were collected throughout the study. Plasma cholesterol and triglycerides were determined every 4 weeks using enzymatic kits (Roche/Hitachi) according to the manufacturer's protocols and average plasma levels over week 4 to 16 were calculated.
HDL-C was measured after precipitation as described previously 20 . The distribution of cholesterol over plasma lipoproteins was determined in group wise-pooled plasma by fast protein liquid chromatography (FPLC) 9 . The inflammatory markers SAA, E-selectin and MCP-1 were measured using the ELISA kits from Tridelta (SAA) and R&D (MCP-1, E-selectin) according to the manufacturer's instruction. Plasma ALT and AST were determined using a spectrophotometric assay (Boehringer Reflotron system) in group wise-pooled samples. Blood pressure (SBP, DBP) and heart rate were measured using the tail cuff method in 8 mice per group at 2 and 15 weeks 18 .

BAL and urinary albumin:creatinin
The lungs were flushed two times with 750 µl PBS into the trachea using a BD 20G angiocatheter to collect broncho-alveolar lavage (BAL) fluid. Protein and albumin content in BAL fluid were determined in the supernatant using the protein determination kit from Pierce and the mouse albumin ELISA kit (ALPCO, Salem, USA). Urinary albumin and creatinin levels were determined using the mouse albumin ELISA kit (ALPCO, Salem, USA) and the creatinin kit (Exocell, Philadelphia, USA) according to manufacturer's instruction.

Flow cytometric analysis
White cell profiling was performed via FACS using the BD FACS Canto II apparatus (Becton Dickinson, Franklin Lakes, New Jersey, USA). After 12 weeks of treatment, peripheral blood mononuclear cells (PBMCs) were isolated from fresh blood samples of 8 mice per group, and were sorted into CD11b+/CD11c-(monocytes), and further divided into CD11b+/Ly6C low and CD11b+/Ly6C high monocytes. The following conjugated monoclonal antibodies, all from eBiosciences, were used: CD11b-FITC, CD11c-PE/Cy7, Ly6C-APC.

Coagulation factor VII and VIIa
Total clotting FVII and FVIIa activity were measured on a STA compact apparatus

Histological assessment of lung morphology and atherosclerosis
Tissues were isolated, fixed in formalin, and embedded in paraffin. The caudal lung was cross  24 . Necrotic area and cholesterol clefts were measured after HPS staining 13 20 22 . Lesion stability index was calculated as described previously 13 20 . In each segment used for lesion quantification, the number of monocytes adhering to the endothelium were counted after immunostaining with AIA 31240 antibody (1:1000; Accurate Chemical and Scientific, New York, New York, USA) 14 20 .

Gene expression analysis
Messenger RNA was isolated from liver of 8 mice per group, using the NEBNext Ultra RNA sample Prep Kit. After fragmentation of the mRNA, cDNA synthesis was performed. The 7 quality and yield after sample preparation was measured with the Fragment Analyzer.
Clustering and DNA sequencing was performed using the Illumina Nextseq 500. The genome reference and annotation file Mus_Musculus.GRCm38 was used for analysis in FastA and GTF format. The reads were aligned to the reference sequence using Tophat 2.0.14 combined with Bowtie 2.1.0, and based on the mapped read locations and the gene annotation HTSeq-count version 0.6.1p1 was used to count how often a read was mapped on the transcript region.
Calculated P-values <0.01 were used as threshold for significance. Selected differentially expressed genes (DEGs) were used as an input for pathway analysis through Ingenuity Pathway Analysis (IPA) suite (www.ingenuity.com, accessed 2015). Gene set enrichment analysis was used to highlight the most important processes and pathways. Relevance of these pathways and processes is indicated as p-value and visualized in a graph by calculating the -log(p-value). Tables   2.1 Supplementary Tables   Table S1. Safety aspects of TKI treatment. Body weight , food intake (per cage) plasma ALT (pooled), and plasma AST (pooled) at baseline and after 16 weeks of treatment with imatinib (150 mg/kg, BID), nilotinib (10 and 30 mg/kg) or ponatinib (3 and 10 mg/kg). ALT, Alanine transaminase; AST, aspartate transaminase (n = 2-4 mice per cage) (n = 13-15 per group). Data are presented as means ± SD (n=13-15 per group) or means (cage or group level).   figure 3. To identify the most relevant processes affected by TKI treatment, we calculated the canonical biological processes/pathways affected by imatinib 150 mg/kg (A), nilotinib 10 mg/kg (red bars) and 30 mg/kg (blue bars) (B) and by ponatinib 3 mg/kg (red bars) and 10 mg/kg (blue bars) (C). The relevance of each process is indicated by a p-value of overlap. The p-value of overlap is calculated based on Fisher's exact test which is set standard for overlap analysis in IPA-software. For visualization purposes the -log of the p-value of the top 20 processes are plotted on the x-axes (n = 8 per group).

A B
Supplemental figure 4. TKI treatment regulates many genes related to atherosclerosis signaling, with the most pronounced effect by imatinib. The heat map shows all significantly upregulated (red) and downregulated (green) genes involved in atherosclerosis signaling of mice treated with imatinib (150 mg/kg), nilotinib (30 mg/kg) or ponatinib (10 mg/kg) as compared to control mice (A). Molecular response of imatinib (150 mg/kg) on atherosclerosis signaling visualized by subpathways (B). P-values of <0.01 were used as cutoff (n = 8 per group).