Myocardial infarction (MI) is a leading cause of morbidity and mortality worldwide. MI occurs when a coronary artery becomes occluded, leading to myocardial ischaemia and extensive cardiomyocyte death. The surviving myocardium subsequently undergoes compensatory remodelling and scarring, which often results in secondary complications, such as heart failure. Although MI can be successfully treated and managed (Anderson and Morrow, 2017), there are no approved therapies that promote repair of the damaged myocardium. Recent clinical trials have investigated immunomodulatory therapies that inhibit pleiotropic inflammatory pathways (Ridker et al., 2017; Tardif et al., 2019). These treatments lower the incidence of cardiovascular events post-MI but increase the risk of infections. Therefore, there is a need to explore treatments that specifically target myocardial inflammation and promote downstream cardiac repair mechanisms following MI.

Neutrophils are the first immune cell recruited to the myocardial infarct, where they phagocytose dead and dying cells (Dewald et al., 2004; Swirski and Nahrendorf, 2013). Neutrophils subsequently secrete inflammatory mediators to recruit monocytes, which later differentiate into macrophages (Dewald et al., 2005; Nahrendorf et al., 2007). Once the acute inflammatory response starts to resolve, most infiltrating neutrophils undergo apoptosis (Daseke et al., 2019). Apoptotic neutrophils are efferocytosed by inflammatory macrophages, triggering a series of anti-inflammatory pathways that promote cardiac repair (Savill et al., 2002; Schwab et al., 2007; Frangogiannis and Rosenzweig, 2012; Ma et al., 2013). Conversely, defective clearance of neutrophils augments inflammation, promoting cardiomyocyte apoptosis, infarct expansion and adverse structural remodelling (Frangogiannis et al., 2002; Vinten-Johansen, 2004; Garlichs et al., 2004; van Hout et al., 2015; Schloss et al., 2016). Indeed, blood neutrophilia is recognised as an indicator of adverse clinical outcomes following MI (Arruda-Olson et al., 2009; Chia et al., 2009). Removing cardiac-recruited neutrophils therefore has potential as a viable therapeutic strategy to improve myocardial repair post-MI.

Extensive work from our group and others has shown that cyclin-dependent kinase 9 (CDK9) inhibitor compounds selectively induce neutrophil apoptosis, reduce neutrophil infiltration, and promote the resolution of inflammation in vitro and in vivo (Rossi et al., 2006; Loynes et al., 2010; Leitch et al., 2012; Wang et al., 2012; Lucas et al., 2014; Hoodless et al., 2016). Unlike most other CDKs, CDK9 specifically regulates the transcription of primary inflammatory response genes via RNA polymerase II. These include genes encoding inflammatory cytokines and the neutrophil pro-survival protein Mcl1 (Sundar et al., 2021; Eyvazi et al., 2019; Lucas et al., 2014). Acute inhibition of CDK9 therefore provides a therapeutic opportunity to suppress the transcription of short-lived inflammatory disease drivers preferentially. However, owing to the conserved structure of CDKs, CDK9 inhibitor compounds may also target other kinases (Kryštof et al., 2012). Two potent CDK9 inhibitors, AT7519 and flavopiridol (FVP), have been widely used in clinical trials as anti-cancer therapies (Mahadevan et al., 2011; Chen et al., 2014; Luke et al., 2012; Awan et al., 2016). Our group has shown that AT7519 and FVP drive neutrophil apoptosis in a CDK9-dependent manner to resolve inflammation following tail fin transection in larval zebrafish (Hoodless et al., 2016). It is not yet understood how CDK9 inhibitors influence inflammatory and repair/regeneration responses following tissue wounding.

The zebrafish has proven to be an essential model for studying cardiac injury, repair and regeneration. Unlike adult mammalian hearts, zebrafish hearts regenerate rapidly following injury via cardiomyocyte proliferation (Poss et al., 2002; Jopling et al., 2010; Kikuchi et al., 2010). Adult zebrafish cardiac injury and regeneration studies have found that sustained neutrophil retention inhibits cardiomyocyte proliferation, promotes cardiomyocyte apoptosis and delays scar regression (Lai et al., 2017; Xu et al., 2019). The resolution of neutrophilic inflammation is therefore considered a prerequisite for timely and complete heart regeneration. We recently characterised neutrophil and macrophage migratory responses in larval zebrafish cardiac injury using bespoke live imaging (Taylor et al., 2019; Kaveh et al., 2020). We identified a conserved sequence of events marked by an early and acute phase of neutrophil recruitment followed by sustained macrophage recruitment (Kaveh et al., 2020). Importantly, the dynamics of the immune cell response in larval zebrafish closely recapitulates that of adult zebrafish and murine models of cardiac injury (Bevan et al., 2020; Epelman et al., 2015).

In this study, we use our established larval zebrafish cardiac injury model to investigate whether CDK9 inhibitor (CDK9i) treatment with AT7519 or FVP resolves neutrophil infiltration and examine whether this regulates downstream macrophage involvement and cardiac regeneration. We found that both AT7519 and FVP resolved neutrophilic inflammation via reverse migration. However, subsequent drug exposure caused adverse effects, which were avoided by shortening CDK9i treatment duration. Interestingly, transient (pulsed) treatment with AT7519, but not FVP, enhanced tnf expression in wound-associated macrophages, in turn promoting macrophage-dependent cardiomyocyte number expansion and the rate of myocardial wound closure. We show that, unlike FVP, AT7519 is a selective CDK9 inhibitor and thus a promising immunomodulatory treatment that could promote cardiomyocyte regeneration.

To understand better the differential phenotypes observed with AT7519 and FVP treatment, we next explored the selectivity of these two CDK9 inhibitor compounds in larval zebrafish. Drug screening studies in vitro have suggested that AT7519 is a more selective CDK9 inhibitor compared with first-generation CDK9 inhibitors such as FVP (Santo et al., 2010; Liu et al., 2012). We formally tested the Cdk9 selectivity of these inhibitors in vivo using stable cdk9 knockout zebrafish generated using CRISPR/Cas9 (Hoodless et al., 2016). Homozygous cdk9 mutant zebrafish larvae are phenotypically distinguishable at 3 days post-fertilisation (dpf) (Hoodless et al., 2016). Compared with their heterozygous and wild-type siblings, homozygous cdk9 mutants display a curved body axis, shorter body length and smaller eye diameter (Fig. 4A). No phenotypic differences were identified between heterozygous mutants and wild-type siblings up to 5 dpf (Fig. 4A); the genetic identity of these larvae was confirmed by genotyping (Fig. 4B). We reasoned that a truly selective CDK9 inhibitor would not have any effect on cdk9−/− knockout zebrafish larvae. To test this, we treated 3 dpf homozygous cdk9 mutants continuously with DMSO vehicle, AT7519 or FVP and quantified heart rate between treatments as a readout for overall health over a 48 h time period. We have shown that continuous CDK9i treatment causes wild-type larvae to develop bradycardia (Fig. S1) and heart rate is a recognised readout of drug-induced toxicity in larval zebrafish (Rubinstein, 2006; Kithcart and MacRae, 2017). Thus, a decline in heart rate with AT7519 or FVP compared with vehicle would suggest that the compounds are acting in a Cdk9-independent manner. First, larvae were treated with 1 μM concentrations of AT7519 or FVP (or DMSO vehicle) so that CDK9i treatments were fair and comparable. Between 2 hpt and 48 hpt, all treatment groups displayed a gradual reduction in heart rate, which was associated with decreased survival from 24 hpt (Fig. 4C,E). At 24 hpt, compared with the DMSO vehicle group, FVP-treated, but not AT7519-treated, mutant larvae displayed a significant reduction in heart rate (22.5±8.9 versus 80.4±6.6 and 63.2±8.6 versus 80.4±6.6, respectively), which was associated with increased mortality (64.3% versus 14.3%) (Fig. 4C,E). The heart rates of larvae treated with AT7519 displayed a more gradual reduction, similar to their DMSO vehicle-treated counterparts (Fig. 4C,E). To draw direct comparisons between drug selectivity and the differential phenotypes observed, we applied the drug concentrations established originally for resolving neutrophilic inflammation (50 μM for AT7519 and 3 μM for FVP). Using these concentrations, FVP-treated mutant larvae displayed significantly lowered heart rates from 2 hpt (90.0±6.7 versus 117.1±6.2) until 12 hpt (64.3±6.5 versus 99.6±4.9) compared with their DMSO vehicle-treated and AT7519-treated counterparts, which showed no differences across these time points (Fig. 4D). Until 12 hpt, this FVP-induced reduction in heart rate was not associated with a change in survival (Fig. 4F). At 24 h after FVP treatment, 14% of larvae survived, all of which displayed diminished heart rates (Fig. 4D,F). In contrast, 24 h after AT7519 treatment survival only decreased to 93%, but the heart rates of these larvae were lower compared with their DMSO vehicle-treated counterparts (60.0±9.0 versus 88.2±8.7) (Fig. 4D,F). At 48 hpt, no larvae that were treated with FVP or AT7519 survived, whereas 36% of DMSO vehicle-treated survived (Fig. 4D,F). In summary, we have developed a proof-of-concept assay using knockout larval zebrafish mutants to examine drug selectivity in vivo. The assay indicated whether AT7519 and FVP exhibit Cdk9-independent effects up to 48 hpt, with FVP displaying significant off-target effects from 2 hpt. These comparative zebrafish data suggest that AT7519 is a particularly selective CDK9 inhibitor in vivo.

Fig. 4.

AT7519 is a more selective CDK9 inhibitor than FVP in zebrafish. (A) Brightfield images of a cdk9+/+ (top), cdk9+/− (middle) and cdk9−/− (bottom) whole zebrafish at 4 dpf. Scale bars: 1 mm. (B) Restriction enzyme digest gel displaying cdk9 genotypes of zebrafish larvae. Hyperladder (HL) band and individual genotype bands (in order: cdk9−/−, cdk9+/− and cdk9+/+) are indicated. (C,E) Heart rate (beats/min) (C) and percentage survival (E) of cdk9−/− larvae at 2 hpt, 6 hpt, 12 hpt, 24 hpt and 48 hpt with 0.1%, DMSO vehicle, 1 μM AT7519 or 1 μM FVP treatment. (D,F) Heart rate (beats/min) (D) and percentage survival (F) of cdk9−/− larvae at 2 hpt, 6 hpt, 12 hpt, 24 hpt and 48 hpt with 0.3% DMSO vehicle, 50 μM AT7519 or 3 μM FVP treatment. (C,D) Error bars represent s.e.m., n=15 larvae, experimental n=3. **P<0.01, ***P<0.001, ****P<0.0001 (two-way ANOVA and Bonferroni post-hoc test for comparisons between DMSO vehicle or CDK9i treatment groups).

Fig. 4.

AT7519 is a more selective CDK9 inhibitor than FVP in zebrafish. (A) Brightfield images of a cdk9+/+ (top), cdk9+/− (middle) and cdk9−/− (bottom) whole zebrafish at 4 dpf. Scale bars: 1 mm. (B) Restriction enzyme digest gel displaying cdk9 genotypes of zebrafish larvae. Hyperladder (HL) band and individual genotype bands (in order: cdk9−/−, cdk9+/− and cdk9+/+) are indicated. (C,E) Heart rate (beats/min) (C) and percentage survival (E) of cdk9−/− larvae at 2 hpt, 6 hpt, 12 hpt, 24 hpt and 48 hpt with 0.1%, DMSO vehicle, 1 μM AT7519 or 1 μM FVP treatment. (D,F) Heart rate (beats/min) (D) and percentage survival (F) of cdk9−/− larvae at 2 hpt, 6 hpt, 12 hpt, 24 hpt and 48 hpt with 0.3% DMSO vehicle, 50 μM AT7519 or 3 μM FVP treatment. (C,D) Error bars represent s.e.m., n=15 larvae, experimental n=3. **P<0.01, ***P<0.001, ****P<0.0001 (two-way ANOVA and Bonferroni post-hoc test for comparisons between DMSO vehicle or CDK9i treatment groups).

Resolving the inflammatory response is a promising therapeutic approach to promote tissue repair/regeneration following injury. CDK9 inhibitor compounds, currently deployed in clinical trials as anti-cancer treatments, can be applied experimentally to curtail early neutrophilic inflammation (Rossi et al., 2006; Lucas et al., 2014; Hoodless et al., 2016; Cartwright et al., 2019). This study is the first to examine the effect of CDK9 inhibitors (AT7519 and FVP) during the inflammatory and regenerative response following tissue injury in vivo. Using a larval zebrafish model of cardiac injury combined with heartbeat-synchronised imaging, we showed that AT7519 and FVP resolve neutrophilic inflammation at the injured heart via reverse migration, but differentially regulate macrophage polarisation and myocardial regeneration.

As previously shown in various models of injury and infection in vivo (Rossi et al., 2006; Loynes et al., 2010; Leitch et al., 2012; Lucas et al., 2014; Hoodless et al., 2016; Barth et al., 2020), we found that CDK9 inhibitors enhance the resolution of neutrophilic inflammation following heart injury in larval zebrafish. Numerous studies have demonstrated that CDK9 inhibitors induce neutrophil apoptosis via downregulation of Mcl1 (Moulding et al., 1998; Rossi, et al., 2006; Leitch et al., 2012; Wang et al., 2012; Lucas et al., 2014; Dorward et al., 2017). Here, we show that AT7519 and FVP promote the resolution of neutrophilic inflammation from the cardiac lesion via reverse migration (Fig. 1). Despite previously observing increased neutrophil apoptosis following tail fin transection with CDK9i treatment (Hoodless et al., 2016), we did not find any evidence of this at the injured heart. Reverse migration is the primary inflammatory-cell resolution mechanism in this model (Kaveh et al., 2020), most likely because of the size and sterility of the myocardial laser wound. An injury of such scale would release fewer chemoattractant signals, such as reactive oxygen species (e.g. hydrogen peroxide), cytokines (e.g. Il1β) and chemokines (e.g. Cxcr2/Cxcl8), responsible for regulating neutrophil wound retention (Yoo et al., 2010; Yan et al., 2014; Powell et al., 2017; Coombs et al., 2019; Isles et al., 2019). Expression of these inflammatory mediators could indeed be modulated in the presence of AT7519 or FVP, as documented in other studies (Santo et al., 2010; Yik et al., 2014). Consequently, this would alter the chemoattractant gradient, desensitising wound-swarming neutrophils and inducing their reverse migration. Similarly, other compounds that cause neutrophil apoptosis in mammalian systems have been shown to promote neutrophil reverse migration following tail fin wounding in larval zebrafish (Robertson et al., 2014). Further research is needed to understand better how CDK9 inhibitors regulate the aforementioned inflammatory mediators to induce immune cell reverse migration, particularly via chemokine signalling at sites of sterile injury (Isles et al., 2019; Coombs et al., 2019). Reverse migration may well be an important neutrophil resolution mechanism following cardiac injury in mammals, as shown following sterile liver injury (Wang et al., 2017). However, with live imaging proving extremely difficult in mammalian models of MI, it is not currently possible to visualise inflammatory cells non-invasively at high spatiotemporal resolution.

The majority of CDK9 inhibitors act by competitively inhibiting the ATP-binding domain, which is conserved between all CDKs (Kryštof et al., 2012). Consequently, long-term exposure to CDK9 inhibitor compounds can cause undesirable effects as a result of inhibition of other CDKs, many of which are cell-cycle regulators, such as CDK2 (De Azevedo et al., 1996; Wyatt et al., 2008). We showed that continuous AT7519 or FVP treatments result in developmental and injury-associated adverse effects, including reduced cardiomyocyte number expansion, cardiac function and macrophage wound retention, in addition to neutropenia (Fig. 2, Figs S1, S2). Continuous FVP treatment has previously been shown to inhibit cardiomyocyte proliferation in larval zebrafish (Matrone et al., 2015), suggesting that the same anti-proliferative effect could be occurring, although cardiomyocyte apoptosis may also contribute to the reduction in cardiomyocyte numbers. We postulated whether the adverse effects associated with continuous CDK9i treatment were due to non-specific binding. To test this, we developed a larval zebrafish CDK9 inhibitor selectivity assay using homozygous knockout cdk9 mutants and heart rate as a surrogate measurement for overall health. The assay revealed that AT7519 and FVP accelerate the decline of heart rate in knockout mutants from 1 day post-treatment (Fig. 4). This reduction in heart rate coincides with the onset of adverse phenotypes (Fig. 2), indicating that from 1 day post-treatment both compounds were acting in a Cdk9-independent manner. However, the assay does not rule out the adverse effects being partially Cdk9 dependent, as vehicle-treated knockout mutants also displayed a decline in health, albeit more gradual. Of the two CDK9 inhibitors, FVP showed marked off-target effects in the selectivity assay (Fig. 4), a likely cause for the prominent adverse phenotypes observed (Fig. 2), which has also been reported in vitro (Garriga et al., 2010; Liu et al., 2012). Indeed, this larval zebrafish knockout screening approach could be applied to other druggable targets and used to identify uniquely selective inhibitors in a high-throughput manner and across short time scales (≤2 h) in vivo.

By limiting the CDK9i treatment period to a 2-h window, we were able to enhance the resolution of neutrophilic inflammation while avoiding all adverse effects. Using the transient treatment, wound macrophage accumulation was unaffected (Fig. 3), suggesting that prolonged neutrophil swarming is not required for macrophage recruitment/retention. Furthermore, we observed an unexpected difference between CDK9i treatments whereby AT7519, but not FVP, increased the polarisation of wound macrophages to a tnf+ phenotype (Fig. 3). The selectivity assay revealed that from 2 hpt (the duration of transient treatment), FVP exhibited significantly less Cdk9 selectivity compared with AT7519 (Fig. 4). Additionally, FVP has been shown to inhibit TNF activation and signalling in other models of inflammation (Takada and Aggarwal, 2004; Haque et al., 2011; Schmerwitz et al., 2011), whereas AT7519 does not disrupt TNF activity (Lucas et al., 2014). Overall, these findings suggest that FVP suppressed tnf upregulation in wound-associated macrophages. As our selectivity assay enables high-throughput assessment of individual animals across short time scales live in vivo, it is not suited for gene expression analysis. Therefore, how AT7519 and FVP differentially influence the expression of inflammatory response genes could be further investigated by RNA sequencing.

Cellular mechanisms regulating immune cell activity after wounding have been largely characterised in murine models and are not entirely recapitulated in zebrafish. For example, neutrophil apoptosis, subsequent macrophage efferocytosis and polarisation have not been reported following wounding in larval zebrafish (Starnes and Huttenlocher, 2012; Robertson et al., 2014; Loynes et al., 2018; Kaveh et al., 2020). Instead, the role of immune cells is more dynamic and closely coupled to molecular signalling (Loynes et al., 2018; Coombs et al., 2019; Tsarouchas et al., 2018; Sanz-Morejón et al., 2019). Larval zebrafish studies have demonstrated tnf+ macrophages to have pro-regenerative roles following tissue wounding (Nguyen-Chi et al., 2017; Tsarouchas et al., 2018; Gurevich et al., 2018; Cavone et al., 2021). Our data show that cardiac-injured larvae transiently treated with AT7519, but not FVP, exhibit enhanced cardiomyocyte number expansion at 2 days post-injury (Fig. 5), 1 day after the peak tnf+ macrophage response (Fig. 3). Importantly, we found macrophages to be required for the improved regenerative response following AT7519 treatment (Fig. 6). One molecular mechanism for this macrophage-dependent effect could be that tnf+ macrophages express/secrete mitogenic factors, such as Vegf, as is the case during muscle wounding angiogenesis (Gurevich et al., 2018). Similarly, Tnf itself could act as a mitogen via activation of histone genes in progenitor cells, as described during spinal cord regeneration (Cavone et al., 2021). Single-cell RNA sequencing of wound-dwelling macrophages has recently been performed in the spinal cord and skeletal musculature of larval zebrafish by Cavone et al. (2021) and Ratnayake et al. (2021), respectively. In both studies, the pro-regenerative macrophage subpopulation identified expresses traditional M1 and M2 markers and shares mitogenic factors, specifically Tnf and Hbegf (Cavone et al., 2021; Ratnayake et al., 2021). Given that tnf+ macrophages have pro-regenerative properties in larval zebrafish, tnf is likely one of many differentially regulated genes in macrophages that could be promoting cardiomyocyte regeneration in our model. Interestingly, in adult zebrafish tnf+ macrophages promote scar deposition following cardiac injury (Bevan et al., 2020), suggesting a transition in tnf+ macrophage function during zebrafish development.

Our data indicate that increased cardiomyocyte number expansion following transient AT7519 treatment correlates with accelerated myocardial wound closure, to the point of almost complete regeneration (Fig. 5). This was not, however, associated with enhanced cardiac function, which recovered rapidly in both AT7519-treated and control larvae. Cardiomyocyte proliferation is a prerequisite for cardiac regeneration in many animal models (Godwin et al., 2017; Chablais et al., 2011; Curado et al., 2007; Porrello et al., 2013), suggesting that cardiomyocyte proliferation, enhanced by macrophages in our model, could be driving myocardial wound closure. Furthermore, 4D LSFM imaging during myocardial regeneration revealed that wound-bordering cardiomyocytes protrude into and subsequently bridge across the wound, gradually sealing it (Fig. 5, Movie 7). Cardiomyocyte bridging has previously been reported following transplantation of neonatal rat cardiomyocytes to infarcted hearts in vitro (Sekine et al., 2006); however, to our knowledge, this is the first time such an event has been observed live in the beating heart. Extracellular matrix proteins, such as collagen, could form a scaffold to facilitate cardiomyocyte wound bridging, as similar mechanisms occur in the injured hearts of adult zebrafish (Simões et al., 2020). Future studies could include high-resolution live imaging and complementary sequencing experiments in larval zebrafish to unravel the cardiac/immune cell types and signalling molecules regulating cardiac regeneration.

As our findings are from a developing zebrafish model, it will be important to validate AT7519 treatment in an adult MI model that more closely mimics human disease. It will be particularly important to corroborate our findings by determining whether AT7519 polarises macrophages to a reparative phenotype and how this regulates cardiac fibrosis, cardiac function, scar resolution and angiogenesis. Furthermore, it will be necessary to identify whether AT7519 affects other immune cell types absent in our model, namely monocytes, B cells, T cells and eosinophils – all of which play important roles during myocardial injury and repair (Hofmann and Frantz, 2015; Toor et al., 2020). Nevertheless, we have shown that the timing, duration and selectivity of CDK9 inhibitor treatment is imperative when targeting the acute inflammatory response to promote tissue repair/regeneration. AT7519 treatment could be particularly effective in a clinical setting where MI is followed by prolonged coronary reperfusion injury, as there is a profound secondary influx of neutrophils (Vinten-Johansen, 2004; Niccoli et al., 2009; Mangold et al., 2015). This can occur following percutaneous coronary intervention (PCI), a standard clinical procedure for opening an acutely occluded coronary artery following MI. Thus, AT7519 could be administered at the time of PCI to resolve locally recruited neutrophils and promote downstream mechanisms that positively modulate myocardial repair.

In summary, we have shown that AT7519 and FVP resolve neutrophil infiltration by inducing reverse migration from the cardiac injury site. However, AT7519, unlike FVP, showed promise as a selective CDK9 inhibitor by augmenting macrophage polarisation and promoting cardiomyocyte regeneration. As such, future research should establish whether selective CDK9 inhibitors, such as AT7519, have analogous reparative effects on macrophage polarisation and infarct healing in adult models of MI associated with neutrophilic inflammation. This could ultimately reveal the clinical potential of selective CDK9 inhibition as an immunomodulatory therapy for MI.