Zebrafish were raised and handled according to standard protocols (zfin.org). Animal care and experimental protocols were approved by the Guangzhou Institutes of Biomedicine and Health Ethical Committee. Transgenic line Tg(kdrl:EGFP) was used in this study. For chemical treatment, embryos were cultured in 12-well plates (15 embryos per well in 2 ml E3 medium) and treated with the indicated compounds from 2 to 4 days post fertilization (dpf) with chorion removed. The maximum dose of each compound (in range 1–30 µM) that does not induce general developmental abnormality was determined in preliminary experiments and used in cardiovascular toxicity studies. DMSO was used as negative control. Treated embryos were examined at 4 dpf for morphology, circulation and blood vessel patterning. At least two biological repeats were performed and analyzed for each compound. The following compounds were used in this study: imatinib (Targetmol, Boston, MA, T6230), nilotinib (Targetmol, T1524), dasatinib (Targetmol, T1448), bosutinib (Targetmol, T0152), ponatinib (Targetmol, T2372), HQP1351 (Targetmol, T3071), asciminib (Targetmol, T5177), Src-IN-1 (Targetmol, T13061), infigratinib (Targetmol, T1975), AT9283 (Targetmol, T3068), danusertib (Targetmol, T2094), rebastinib (Targetmol, T2640), KW-2449 (Selleck, Houston, TX, S2158), tozasertib (Targetmol, T2509), azilsartan medoxomil (Targetmol, T6219), valsartan (Targetmol, T6716), captopril (Targetmol, T1462), benazepril (Selleck, S5938), amlodipine (Targetmol, T1385), acebutolol (Selleck, S4010), macitentan (Selleck, S8051), metolazone (Selleck, S1610), O-Dianisidine (Sigma, St. Louis, MO, 33430).

Live photos or videos of zebrafish embryos were taken with the Zeiss Axio Zoom. V16 stereo zoom microscope (Zeiss, Oberkochen, Germany). For confocal imaging of zebrafish blood vessels, fixed embryos or cross-sections were imaged with the Zeiss LSM 710 NLO inverted confocal microscope (Zeiss).

Zebrafish embryos were fixed in 4% PFA/PBS overnight at 4°C, washed three times with PBS then embedded in tissue freezing medium (Leica, Wetzlar, Germany, 14020108926). 20 μm-transverse sections were prepared using a cryotome (Leica, CM3050S). Sections were washed with PBS to remove optimal cutting temperature compound (OCT) and ready for confocal imaging.

Zebrafish embryos were fixed in 2.5% (v/v) glutaraldehyde and 2% (v/v) PFA overnight at 4°C, washed in 0.1 M PBS (pH 7.4) and post-fixed in 1% OsO₄ at RT for 1.5 h. Following gradient ethanol (50, 70, 80, 90, and 100%) and acetone dehydration, samples were embedded in Epon 812 and polymerized for 24 h at 40°C plus 24 h at 60°C. Cross sections (100 nm) were prepared with an ultramicrotome (Leica, EM UC7). Samples were stained with uranyl acetate and lead citrate then examined and photographed with the Tecnai G2 Spirit TEM (FEI).

Dechorionated wild type embryos cultured in 12-well plates (15 embryos per well in 2 ml E3 medium) were treated with ponatinib (3 µM), HQP1351 (1 µM) or asciminib (20 µM) from 2 to 4 dpf. For rescue experiments, azilsartan (30 µM) or valsartan (30 µM) was added at the beginning of TKI treatment. DMSO was used as negative control. Embryos were collected at 4 dpf and stained in dark for 15 min in staining solution [o-dianisidine, 0.6 mg/ml, 10 mM sodium acetate, 0.65% H₂O₂, and 40% (vol/vol) ethanol]. Stained embryos were cleared with DMSO 3 times then photographed with a stereo microscope (Lecia M205 FA). Experiments were repeated twice and at least fifteen embryos from each group were examined in each experiment.

The Tg(kdrl:EGFP) is a zebrafish transgenic report line that expresses GFP in endothelial cells and it is widely used in studies related to vessel development and function. We tested whether CML TKI treatment in this transgenic fish line induces cardiovascular defects. Embryos were treated with the listed concentration of TKI at 2 dpf (Figure 1), a time point when major trunk vessels (dorsal aorta, posterior cardinal vein) and intersegmental vessels have formed and active circulation has been established, and embryonic morphology and vasculature were examined at 4 dpf. We found that embryos treated with 20 µM imatinib have normal morphology, vasculature and blood circulation (Figure 1 and Supplementary Videos S1, S2). 20 µM of nilotinib treatment leads to mild cardiac edema (Figure 1A) and mildly reduced circulation (Supplementary Video S3) while the truck/intersegmental vessels appear normal (Figure 1B). In addition to cardiac edema, bosutinib (20 µM) or dasatinib (10 µM) treatment results in stenosis of dorsal aorta (red arrow in Figure 1B) while posterior cardinal vein (white arrow in Figure 1B) and intersegmental vessels are less affected. Circulation is either reduced (Bosutinib, Supplementary Video S4) or barely detectable (Dasatinib, Supplementary Video S5) in the treated embryos. We found ponatinib to be the most toxic CML TKI in this assay. Ponatinib as low as 3 µM induces severe cardiac edema (Figure 1A) and constriction of dorsal aorta and posterior cardinal vein, inhibits the formation of subintestinal vein (yellow arrow in Figure 1B) and no circulation is detectable in the treated embryos (Supplementary Video S6). The constriction of both trunk vessels was confirmed in cross-sections (Figure 1C). We then performed transmission electron microscope analysis to further characterize ponatinib-induced vascular defects. In control embryos, a thin layer of squamous endothelial cells with well-formed cell junctions (white arrow in Figure 1D) is observed. In ponatinib treated embryos, endothelial cells collapse and cell junctions are disrupted to form a thicker and irregular lining layer of vessel. Together, these data indicate that ponatinib disrupts endothelial lining which then leads to vessel constriction in zebrafish embryos.

In addition to BCR-ABL1, CML TKIs have distinct inhibitory activities to other kinases and the differences in substrate spectrum might contribute to their various cardiovascular toxicities. Previous studies reveal that SRC kinase and FGFRs are common targets for ponatinib, dasatinib and bosutinib but they are not inhibited by imatinib and nilotinib (Moslehi and Deininger, 2015; Rossari et al., 2018; Zeng and Schmaier, 2020; Lee et al., 2021) which is consistent with our observation that ponatinib, dasatinib, bosutinib but not imatinib and nilotinib induce dorsal aorta stenosis in zebrafish (Figure 1B), so we hypothesized that inhibition of SRC or FGFRs is involved in the vasculopathies of CML TKIs. We investigated this hypothesis by testing the effect of SRC inhibitor (Src-IN-1) or FGFR inhibitor (infigratinib) on vessel integrity. We found that Src-IN-1 treatment (5 µM) does not induce stenosis of dorsal aorta, however, a combination of Src-IN-1 (5 µM) with imatinib either at 10 or 20 µM clearly induces stenosis of dorsal aorta (Figure 2). Similarly, combination of Src-IN-1 with nilotinib (2 µM + 20 µM or 5 µM + 5 µM) is able to induce dorsal aorta constriction (Figure 2). On the other hand, infigratinib (20 µM), either alone or in combination with imatinib or nilotinib, is not sufficient to induce constriction of dorsal aorta (Figure 2), indicating inhibition of FGFRs may not be involved in vessel toxicity induced by CML TKIs. Src-IN-1 is known to inhibit SRC or LCK. As lck is a T-cell specific gene and it is absent from vasculature from 2 to 4 dpf, the vascular effect of Src-IN-1 is most likely due to the inhibitor of SRC. These results suggest that a combinatory inhibition of ABL1 and SRC may phenocopy ponatinib-induced vascular defects.

Ponatinib is the only approved third generation TKI that can overcome T315I mutation but prominent cardiovascular toxicity limits its application. To meet this unsatisfied medical need, other small chemical inhibitors that can target T315I mutation have been developed while their potential cardiovascular toxicity remains to be investigated. So we used our zebrafish model to evaluate potential cardiovascular toxicities of several clinical stage TKIs. We found that up to 20 µM of danusertib (Gontarewicz and Brummendorf, 2010), AT9283 (Tanaka et al., 2010), KW-2449 (Shiotsu et al., 2009) or tozasertib (MK-0457) (Giles et al., 2013) treated embryos have normal morphology and blood vessel architecture (Figures 3A,B). On the other hand, 5 µM of rebastinib (DCC-2036) (Eide et al., 2011) induces cardiac edema and stenosis of dorsal aorta, indicating potential cardiovascular toxicity. HQP1351, which shares similar structure and inhibitory repertoire with ponatinib (Ren et al., 2013), induces severe cardiovascular toxicity: HQP1351 at 1 µM is able to induce severe cardiac edema and stenosis of dorsal aorta (Figures 3A,B) which is the most toxic one among all TKIs we examined. Asciminib treatment (up to 20 µM) does not lead to noticeable cardiovascular toxicity (Figures 3A,B). These data reveal dramatically different cardiovascular toxicities of those T315I targeting TKIs in zebrafish model and suggest that precaution should be taken during further development of those TKIs with clear vasculopathies.

Small chemicals are widely used in zebrafish studies to modify development or physiology of embryos. So we performed proof-of principle studies to investigate whether or not ponatinib-induced vascular constriction can be rescued by small chemical compounds. We first tested commonly used anti-hypertensive drugs since many of them have vasodilation function. Embryos were treated with ponatinib (3 µM) together with the indicated anti-hypertensive drugs at 2 dpf and the vascular morphology and circulation of treated embryos were examined at 4 dpf. We found ARBs such as valsartan (30 µM) and azilsartan (30 µM) fully rescue ponatinib-induced vascular stenosis (Figure 4A) and improve circulation (Supplementary Videos S7, S8). On the other hand, endothelin receptor antagonist (macitentan, 5 µM), angiotensin-converting enzyme (ACE) inhibitors (benazepril and captopril, 30 µM), beta blocker (acebutolol, 30 µM), calcium channel blocker (amlodipine, 30 µM) and diuretic (metolazone, 30 µM) are not effective in rescuing ponatinib-induced cardiovascular toxicity (Figure 4A) and defective circulation (Supplementary Videos S9–S14). We then tested whether these compounds can rescue HQP1351 induced cardiovascular toxicity. We found that valsartan and azilsartan are effective in rescuing HQP1351 induced vascular constriction and circulation defects (Figure 4B and Supplementary Video S15). In addition, we found that the ACE inhibitor benazepril and the diuretic metolazone are able to reverse HQP351 induced vascular defects while they fail to do so in the majority of ponatinib treated embryos. Macitentan, captopril, acebutolol and amlodipine fail to rescue both ponatinib and HQP1351 induced vascular defects (Figure 4B).

Ponatinib is recently reported to induce cerebral thrombosis in a zebrafish ischemic stroke model (Zhu et al., 2020). We found that HQP1351 (1 µM) but not asciminib (20 µM) also induces cerebral thrombosis in the same assay (Figure 5) which is consistent with their different activities to induce vessel stenosis. Since azilsartan and valsartan can rescue ponatinib or HQP1351 induced dorsal aorta stenosis and restore trunk circulation, we analyzed whether they are able to rescue cerebral thrombosis as well. We found that co-treatment with azilsartan (30 µM) or valsartan (30 µM) fully rescues ponatinib (3 µM) or HQP1351 (1 µM) induced cerebral thrombosis (Figure 5). Together, these results suggest that ponatinib/HQP131 may induce vessel constriction and cerebral thrombosis via over-activation of the angiotensin signaling pathway and ARBs could have clinical value in preventing ponatinib induced vascular occlusive events in patients.