BDNF modulates heart contraction force and long-term homeostasis through truncated TrkB.T1 receptor activation

Introduction

Brain-derived neurotrophic factor (BDNF) is a growth factor widely expressed in the nervous system. Changes in its level have been found and correlated to the development of several human diseases including neurodegeneration, depression, psychiatric disorders, and obesity (Chao et al., 2006; Nagahara and Tuszynski, 2011; Lu et al., 2014). The direct relevance of normal BDNF signaling to human fitness has been validated by the use of animal models, which have also allowed the dissection of the molecular mechanism underlying BDNF function in vivo (Rios et al., 2001; Zuccato and Cattaneo, 2009; Baydyuk and Xu, 2014). The TrkB gene encodes BDNF high affinity receptors that are widely expressed in neuronal tissues (Klein et al., 1990; Escandón et al., 1994). This locus generates multiple TrkB isoforms all of which have the same extracellular domain but have different intracellular domains (Stoilov et al., 2002). The two main isoforms include a full-length receptor with a tyrosine kinase domain (TrkB.Kin) used for signaling and a truncated TrkB.T1 receptor lacking kinase activity (Klein et al., 1990; Dorsey et al., 2006). Although a role for TrkB.Kin has been established in neuronal development and function including differentiation, outgrowth, and synaptic plasticity, the physiological significance of TrkB.T1 intrinsic signaling is still unclear despite its high sequence conservation among species and its being the most highly expressed TrkB isoform in the mature animal. Mice lacking TrkB.T1 have increased anxiety-related behavior that is associated with structural alterations in neurites of the amygdala (Carim-Todd et al., 2009). Despite reports of TrkB.T1 signaling in isolated glia cells there are no obvious deficiencies in this cell population in TrkB.T1 mutant mice (Rose et al., 2003; Dorsey et al., 2006; Ohira et al., 2006; Carim-Todd et al., 2009). In the cardiovascular system, BDNF and its receptor TrkB have been described to have an early developmental role in cardiac endothelium formation (Anastasia et al., 2014). Interestingly, in the adult heart only TrkB.T1-specific polyadenylated mRNA has been reported, suggesting protein expression of this particular receptor isoform (Stoilov et al., 2002). Our analysis confirms this finding and we further show that it mediates BDNF inotropic function by regulating Ca²⁺ signaling. We found that specific deletion of TrkB.T1 in cardiomyocytes causes cardiomyopathy and that BDNF is the ligand activating TrkB.T1. We show that BDNF is secreted by cardiomyocytes and its specific deletion in cardiomyocytes causes a cardiomyopathy resembling that caused by TrkB.T1 deficiency. Our data unveil a novel nonneuronal function for BDNF and uncover the first physiologically relevant direct signaling activity of the TrkB.T1 receptor. These findings identify a new pathway regulating cardiac contractility and suggest that perturbation in BDNF and TrkB expression may cause cardiac pathological conditions.

Results

To address a potential role of BDNF in the mature cardiovascular system we first investigated the pattern of expression of its receptor TrkB in the adult mouse heart (Fig. 1). Although the full-length TrkB tyrosine kinase receptor is expressed in the cardiac endothelium during development (Donovan et al., 2000), we found that in the adult this isoform is virtually undetectable and only trace levels of its mRNA can be found by RT-PCR (Fig. 1, B and C). Instead, truncated TrkB.T1 protein is the dominant cardiac-expressed isoform as previously shown by RNA expression analyses (Fig. 1; Stoilov et al., 2002). The presence of TrkB mRNA and protein in the heart suggests an intrinsic role for this receptor independent of the nervous system cardiac innervation. Thus, we next tested whether BDNF plays a role in adult cardiac function in an ex vivo paradigm by perfusing isolated mouse hearts (Fig. 2). BDNF injected in the fluid streamline of a Langendorff-perfused mouse heart (Broadley, 1979) induced an increase in the cardiac contraction force as shown by an increase in systolic pressure and a decrease of the diastolic pressure (Fig. 2, A–D). The effect of BDNF appeared exclusively inotropic and lusitropic as it did not affect the spontaneous cardiac contraction frequency (Fig. S1). BDNF also did not influence the coronary flow (Fig. S2). Perfusion of the heart paced at 420 beats per minute (bpm) with 1 ng BDNF causes an increase in the left ventricle developed pressure (LVDP; EC₅₀ of 0.3 ng BDNF) with a timing similar to that elicited by the control compound caffeine (Fig. 2 E and Fig. S1). To test whether BDNF signaling is mediated by the TrkB tyrosine kinase receptor we attempted to use the TrkBF616A mouse model in which TrkB tyrosine kinase activity can be silenced by the specific phosphatase inhibitor 1-NMPP1 drug (Chen et al., 2005). However, in Langerdorff-perfused wild-type (WT) mouse hearts we found that 100 nM 1-NMPP1 induced a dramatic increase in the cardiac contraction force that blunted the BDNF effect, thus preventing the use of this model for this application (Fig. S3). Nevertheless, pretreatment with the K252a tyrosine kinase inhibitor did not prevent the increase in cardiac force elicited by BDNF, suggesting a function independent of TrkB kinase signaling (Fig. 2, B and D).

Interestingly, pretreatment with isoproterenol, a nonselective β-adrenergic agonist, did not blunt the BDNF inotropic effect but was instead additive, suggesting that BDNF signaling does not overlap with the catecholamine pathways (Fig. 3, A and B). In addition, when we tested whether BDNF changes the phosphorylation level of some of the major molecular players involved in cardiac contraction, including CamKII, phospholamban (PLB), and troponin, we found that in Langendorff-perfused hearts at the peak of BDNF action phosphorylation levels were unchanged contrary to hearts treated with forskolin used as positive control (Fig. 3, C–F). These data suggest that BDNF exerts cardiac function through other pathways or this effect is modulatory and not immediately detectable with conventional biochemical analysis.

The effect on cardiac contraction force suggests that cardiomyocytes may be the cell type responding to BDNF with Ca²⁺ as the possible downstream effector. Therefore, we tested adult cardiomyocytes for TrkB expression and their response to BDNF (Fig. 1 A and Fig. 2, I–K). We found that isolated cardiomyocytes do express TrkB.T1 (Fig. 1 A) and when subjected to direct depolarization they elicited higher Ca²⁺ transients by treatment with BDNF compared with just vehicle (Fig. 2 I). Moreover, the Ca²⁺ transient increase was reversible because it was restored to the basal electrical stimulation level after BDNF removal (Fig. 2 J).

The fast action of BDNF and the finding that K252a does not block this function in heart (Fig. 2 B) and cardiomyocytes (Fig. 2 K) suggest that this role is not mediated by the TrkB kinase receptor but by the TrkB.T1 isoform, which is expressed in cardiomyocytes and has been shown to regulate Ca²⁺ signaling, at least in astrocytes (Rose et al., 2003). This prompted us to test whether this receptor mediates cardiac BDNF actions by using a specific TrkB.T1-deficient mouse model (Dorsey et al., 2006). Injection of BDNF in the perfusion system of TrkB.T1 knockout hearts failed to induce the rapid increase in LVDP observed in control hearts. However, their response to caffeine was normal (Fig. 2, F–H). Similarly, adult cardiomyocytes isolated from TrkB.T1 knockout mice failed to show the increase in Ca²⁺ transient evoked by BDNF in controls (Fig. 2, I–K). These data strongly suggest that the TrkB.T1 receptor isoform mediates the inotropic BDNF function in the adult heart and may have a role in normal cardiac physiology. To investigate the in vivo significance of the findings from the ex vivo and cardiomyocyte experiments we performed histological analysis of adult TrkB.T1-deficient hearts. TrkB.T1 knockout hearts showed a cardiac dilation characterized by reduced thickness of the left ventricular wall (Fig. 4, A and B). Moreover, in vivo echocardiography analysis showed an overall enlarged ventricular volume and thinner posterior wall thickness in TrkB.T1^−/− mice compared with WT controls (Fig. 4, C and D). To investigate the significance of the cardiac pathological condition caused by loss of TrkB.T1 we also looked for the presence of mechanisms of cardiac function adaptation including changes in L-type–mediated Ca²⁺ currents (Benitah et al., 2010). Electrophysiological recording of isolated cardiomyocytes detected an increase in L-type–mediated Ca²⁺ current density paired with a higher expression level of L-type Ca²⁺ channels as shown by the biochemical analysis of the CaV1.2 α1c subunit (Fig. 4, E–J). These data combined with the significant impairment in some functional parameters such as a reduced left ventricle-developed pressure measured in Langendorff-perfused mutant hearts suggest that TrkB.T1 is critical for normal cardiac function and its deletion induces secondary changes as a mechanism to compensate for the functional deficits (Table S1). Interestingly, the cardiomyopathy developed by TrkB.T1-deficient mice does not appear to originate during embryonic or early postnatal development as histological analysis of TrkB.T1 mutant mice at postnatal day 12 shows normal posterior wall thickness and no sign of left ventricle dilation (Fig. S4, A and B).

Several studies have reported a putative role for BDNF in central nervous system nuclei controlling cardiac function, peripheral innervation, and the vasculature, suggesting that the phenotype observed in TrkB.T1 mutant mice could be the result of indirect alterations in the microvasculature and/or the sympathetic tone (Donovan et al., 2000; Clark et al., 2011; Anastasia et al., 2014; Wan et al., 2014). Although our ex vivo and in vitro analysis support a direct role for BDNF on cardiomyocyte contraction, to address this point we selectively deleted TrkB.T1 in cardiomyocytes by using a transgenic mouse expressing cre under the promoter of the murine α myosin heavy chain polypeptide 6 (Myh6; Fig. S4 C; Agah et al., 1997). As shown in Fig. 5, specific deletion of TrkB.T1 in cardiomyocytes completely abolished the inotropic effect of BDNF in Langendorff-perfused hearts. Interestingly, deletion of TrkB.T1 in endothelial cells using a cadherin5-specific cre mouse line as a control had no impact, as these hearts responded to BDNF as well as WT controls (Fig. 5, C and F). Most importantly, Myh6-cre/TrkB.T1 mutant hearts showed dilated cardiomyopathy with significantly reduced left ventricle posterior wall thickness, confirming an intrinsic cardiomyocyte role for TrkB.T1 in normal cardiac physiology (Fig. 5, G–K). Next we sought to determine whether BDNF is the ligand for TrkB.T1 in cardiomyocytes or whether other neurotrophins or Trk receptors could exert such a cardiac inotropic role. Thus, we evaluated the contraction force of perfused hearts in the presence of NGF or NT3. Whereas NGF had no effect on the heart contractile force, NT3 elicited a response, though slightly reduced compared with BDNF (Fig. S5, A–C). Because NT3 can also bind TrkB this result was not surprising (Coppola et al., 2001). Nevertheless, the lack of an NT3 inotropic effect in hearts lacking TrkB.T1 strongly suggests that TrkB.T1 is the only neurotrophin receptor mediating this function. To further elucidate the source and identity of the neurotrophin activating TrkB.T1 we tested whether cardiomyocytes could secrete a biologically active neurotrophin. Isolated embryonic cardiac myocytes were grown in culture for 7–10 d and the supernatant was tested for its ability to induce phosphorylation of TrkB kinase expressed in a cell line. In fact, although TrkB full length is not present in cardiomyocytes, a ligand that activates it will also activate TrkB.T1 because they have an identical extracellular ligand-binding domain (Fig. 6). This approach was used because of its biological sensitivity and specificity over other assays (Coppola et al., 2001). We found that cardiomyocyte-conditioned medium induced in vitro TrkB phosphorylation (Fig. 6 C, WT). This result suggested that cardiomyocytes produce and secrete a biologically active neurotrophin. To test if BDNF was the neurotrophic factor secreted in the medium we performed an immunoprecipitation experiment from medium conditioned by cardiomyocytes derived from a mouse model with an HA-tagged BDNF gene (Yang et al., 2009). Immunoprecipitation with an anti-HA antibody revealed the presence of BDNF in the medium, suggesting that BDNF can be secreted by cardiomyocytes. Importantly, medium conditioned by BDNF knockout cardiomyocytes was unable to induce TrkB phosphorylation, indicating that although NT3 can cause the inotropic effect in isolated hearts, BDNF is the TrkB.T1 physiological ligand. This result also excluded the possibility that NT4, the other neurotrophin binding TrkB with high affinity, was the neurotrophin secreted by cardiomyocytes in the media, a result further supported by the very low level of NT3- and NT4-specific mRNA detected by RT-PCR in heart and cardiomyocytes (Fig. S5, D–F). Western blot analysis of adult hearts isolated from mice with the HA-tagged BDNF gene also confirmed BDNF expression, again suggesting that this is the physiological ligand source of TrkB.T1 in cardiomyocytes (Fig. 6 A). Therefore, we tested whether specific deletion of BDNF in cardiomyocytes caused cardiomyopathy. Hearts from two- to three-month-old Myh6-cre-BDNF knockout mice showed dilated cardiomyopathy with significantly reduced posterior wall thickness compared with littermate controls (Fig. 5, G–K). Strikingly, the similarity between the phenotype observed in Myh6-cre-BDNF– and Myh6-cre-TrkB.T1–deficient mice suggests that BDNF-TrkB.T1 signal is intrinsic to cardiomyocytes and functions by an autocrine/paracrine mechanism.

Discussion

We have identified a new unexpected role of BDNF in adult cardiac physiology. This action is mediated by truncated TrkB.T1 receptors expressed in cardiomyocytes and is independent of BDNF function in the nervous system. Although truncated TrkB.T1 has an intracellular domain that is extremely conserved among species, so far there is no data suggesting physiological relevance of its direct signaling in vivo. In isolated glial cells, TrkB.T1 evokes Ca²⁺ signal, but mice lacking this receptor do not show any obvious defect in this cell population (Carim-Todd et al., 2009). The lack of a phenotype could be a result of compensatory mechanisms, redundancy of TrkB.T1-mediated Ca²⁺ signaling in this cell type, or a function required during specific physiological challenges. However, we found that in cardiomyocytes Ca²⁺ regulation by TrkB.T1 is critical as deletion of this receptor causes dilated cardiomyopathy. BDNF activation of TrkB.T1 appears modulatory of cardiac function as in a Langendorff-perfused heart it causes only a 5% increase in cardiac contraction force; yet loss of this pathway leads to cardiomyopathy over the life of the animal. Indeed, it is not surprising that even small perturbations in the proper control of Ca²⁺ levels are deleterious to normal cardiac function as Ca²⁺ regulation is central to the function of cardiomyocytes. Moreover, we provide evidence that BDNF is the ligand activating TrkB.T1 in an autocrine/paracrine fashion as it is expressed and released by cardiomyocytes and its deletion in this cell type causes a cardiomyopathy resembling that caused by TrkB.T1 knockout. Most recently, a parallel study has suggested that TrkB is required for normal cardiac contraction and relaxation (Feng et al., 2015). Contrary to our study that shows that TrkB.T1 is the receptor for BDNF in the heart, Feng et al. (2015) suggested that instead this function is mediated by the TrkB kinase receptor isoform. This finding is somewhat surprising because we and others have found that in the heart TrkB.T1-encoding mRNA is highest, whereas the TrkB kinase messenger RNA is virtually undetectable (Stoilov et al., 2002). The discrepancies in protein analysis data can be explained by the use of different antibodies and appropriate controls. Nevertheless, the most important difference relates to the cardiac phenotype observed by Feng et al. (2015) in mouse models with specific inactivation of the kinase domain either by a chemical genetic (i.e., the TrkBF616A/1-NMPP1 model) or by a conditional deletion approach. We found that in Langendorff-perfused WT mouse hearts 100 nM 1-NMPP1 induced a dramatic increase in the cardiac contraction force and blunted the BDNF effect. Although Feng et al. (2015) reported that 1-NMPP1 did not alter basal cardiomyocyte function they did not include a control to test whether 1-NMPP1 blunts BDNF action as we found (Fig. S3). Thus, it is possible that such an omission leads to misinterpreting the lack of BDNF inotropic function in the TrkBF6161A/1-NMPP1 model as a specific phenotype caused by loss of TrkB kinase activity.

Although, we do not have a definitive explanation for the results obtained with a conditional deletion of a TrkB tyrosine kinase–specific exon, it has been reported that targeting the kinase domain of TrkB causes a dramatic down-regulation of the TrkB.T1 isoform protein (Klein et al., 1993). Although Klein et al. (1993) had no explanation for the disregulation, this finding suggests the possibility that a similar targeting of the kinase region by Feng et al. (2015) generates a TrkB.T1 hypomorph. In support of this scenario is the fact that targeting of the kinase domain causes only reduced systolic function insufficient to trigger chamber dilation (Feng et al., 2015), whereas the complete deletion of TrkB.T1 achieved in our study causes a more severe pathology that includes cardiac dilation developed as a consequence of impaired Ca²⁺ signaling. Importantly, targeting of the TrkB.T1-specific exon does not cause changes in the expression of the TrkB kinase receptor, again suggesting specificity of the phenotype (Fig. 1; Dorsey et al., 2006). Nevertheless, the study by Feng et al. (2015) is important because it supports the finding that TrkB has an intrinsic role in cardiomyocyte function that is independent of cardiac innervation. Moreover, despite the possible explanations for the discrepancies between the two analyses, we cannot entirely exclude that a very small level of the TrkB tyrosine kinase receptor may in part be responsible for the BDNF effect in the heart.

An important new contribution of our study relates to the fact that we have identified the source and identity of the neurotrophin activating TrkB.T1 in the heart. This is relevant because other neurotrophins expressed in the heart could be responsible for this function (Scarisbrick et al., 1993; Emanueli et al., 2014). Moreover, our data suggest that BDNF produced in loco is a significant determinant of TrkB.T1 cardiac function. This was even more surprising considering the relatively low level of BDNF present in the heart compared with the brain (only ∼5% per gram of tissue; Fig. 6 A). In the future it will be of interest to investigate whether BDNF produced by peripheral districts (i.e., striatal muscle) can affect cardiac function and whether loss of either cardiac or peripheral BDNF is sufficient to cause cardiomyopathy.

A major question still unresolved relates to how TrkB.T1 transduces BDNF inotropic function. The fact that the BDNF effect is additive to that of isoproterenol, a nonselective β-adrenergic agonist, suggests that BDNF signaling does not overlap with the catecholamine-activated pathway. Moreover, in an initial analysis of some of the key players in cardiac excitation–contraction coupling, including CamKII, PLB, and troponin I, we found no changes in their level of activation in response to BDNF treatment. This result appears to contrast the finding by Feng et al. (2015) that BDNF increases PLB, CamKII, and RyR receptor phosphorylation. However, their result was obtained after a prolonged treatment (10 or 20 min) of cardiomyocytes with BDNF, raising the possibility that the increase in the phosphorylation of these excitation–contraction coupling players may be secondary because it is detected after BDNF inotropic effect has already subsided. In our paradigm we found that at the peak of BDNF inotropic function in Langendorff-perfused hearts there was no increase of PLB or CamKII phosphorylation, suggesting that they are not directly involved in BDNF contraction in cardiomyocytes.

In glia cells TrkB.T1 mediates inositol-1,4,5-trisphosphate (IP₃)–dependent calcium release from intracellular stores. It was proposed that this mechanism involves an as yet unidentified G protein that stimulates PLC production of IP₃, Ca²⁺ release from stores, and Ca²⁺ entry from the extracellular space. It is tempting to speculate that a similar mechanism is at work in cardiomyocytes where TrkB.T1 induces Ca²⁺ release from the intracellular stores and increases systolic Ca²⁺ transients (Fig. 2). Indeed, IP₃ can contribute to systolic Ca²⁺ transients through a mechanism by which IP₃ receptors are opened by a combination of IP₃ and Ca²⁺ binding (Missiaen et al., 1994; Bootman et al., 1995). According to such a model, IP₃ receptors can sit with IP₃ bound and wait for an activating Ca²⁺ signal to derive from voltage-activated channels or neighboring RyRs (Kockskämper et al., 2008). In turn, activated IP₃ receptors would cause further release of Ca²⁺ from the ER, increasing the systolic Ca²⁺ transients. Importantly, several studies point toward a relevant role of PLC–IP₃ signaling in the development and progression of cardiac hypertrophy and dilation, further supporting a link of BDNF to this pathway (Mende et al., 1998; Kockskämper et al., 2008). In the future, it will be important to investigate through pharmacological and genetics approaches whether this is indeed the TrkB.T1-activated pathway in cardiomyocytes.

Lastly, considering the importance of BDNF/TrkB signaling in the control of food intake, mood disorders, and neurodegenerative diseases there is a significant effort to develop pharmacological agents to harness the activation of this pathway (Nagahara and Tuszynski, 2011; Lu et al., 2014). Although most studies have been focused on the generation of BDNF mimetic molecules, which can activate all TrkB receptors (Longo and Massa, 2013; Rosenthal and Lin, 2014), the present results suggest that any such effort should consider possible effects on the function of the cardiovascular system. Nevertheless, our data indicate that pharmacological activation of intracellularly activated TrkB kinase pathways such as by a transactivation mechanism may achieve central nervous system activation of TrkB kinase without causing cardiac toxicity (Wiese et al., 2007; Yanpallewar et al., 2012).

Materials and methods

Supplementary Material