[Cr] and hypertrophy in CrT-OE mice
A total of 44 male CrT-OE mice were screened by HPLC measurement of myocardial [Cr]. The three study groups were selected based on pre-defined clearly separated LV [Cr] values, which were estimated from RV creatine measurements (Table 1). LV function was evaluated by echocardiography, 1 week prior to tissue harvest. There were no significant differences between WT and medium [Cr] groups for any parameter. The high [Cr] group had significant LV hypertrophy (myocardial CSA) with mild LV dilatation (end-diastolic area), but preserved contractile function (Fractional area change; Table 1). In support of these changes, [Cr] correlated strongly with myocardial CSA (r
2 = 0.60, P = 0.0004). These findings are consistent with previous observations in CrT-OE mice at this age (Phillips et al. 2010).
A total of 34 differentially regulated peptides were identified in the pI6-9 and 33 in the pI4-7 experiments, respectively (Fig. 1; Suppl Tables 1–3). There is an overlap of four proteins between the two experiments, namely haemoglobin subunit β-1; isocitrate dehydrogenase [NADP] mito; elongation factor Tu, mito; β-enolase. PCA analysis showed efficient separation between groups (P ≤ 0.05). Differentially regulated proteins were analysed using a t test and the P value of <0.05 was set as the statistical significance threshold. These differences per group are listed in Suppl Table 1. We used n = 4 samples/group to allow an entire experiment to be run on a single gel, thereby reducing between-experiment variability. This strategy was effective since we were able to confirm changes in β-enolase, glutathione s-transferase and 3-hydroxyacyl-CoA dehydrogenase in response to augmented LV creatine, in agreement with our previous study (Phillips et al. 2010). Novel protein targets that changed dose-dependently with [Cr] were of particular interest and therefore, selected for follow-up.
Myozenin-2 and Nfatc1 pathway
Myozenin-2 is an endogenous calcineurin inhibitor. Reduced myozenin-2 results in unopposed calcineurin activity, relative activation of the nuclear factor of activated T-cells (Nfatc1) pathway and increased susceptibility to LVH (Diedrichs et al. 2004; Frey et al. 2004). Mutations in the myozenin gene are linked to human hypertrophic cardiomyopathy (Ruggiero et al. 2013). Myozenin-2 was down-regulated in the high [Cr] hearts (−21 %; P = 0.004 vs. Low; Fig. 2a, b) and this might explain the hypertrophic phenotype. However, this trend was not statistically significant by immunoblotting (Fig. 2c), which used a different set of tissue samples. This may also reflect the differences in sensitivity of the two experimental techniques (proteomics vs immunoblotting). Nevertheless, downstream activation of the Nfatc1 pathway can be inferred by elevated gene expression of regulator of calcineurin (Rcan1) (P = 0.017; Fig. 2d) and Foxo1 (Fig. 2e; P = 0.01 WT vs medium [Cr] groups). It should be noted that while increased gene expression of Rcan1 is considered a reliable indicator of calcineurin activation, it is not obligatory for myozenin-2 mediated hypertrophy (Ruggiero et al. 2013).
Alpha crystallin B chain
Alpha crystallin B is a heat shock protein and therefore, plays a role in stabilising proteins under stress conditions. Over-expression protects against necrotic and apoptotic death following ischaemia/reperfusion injury (Ray et al. 2001). Proteomic analysis showed that α-crystallin B chain increases in line with [Cr] (Fig. 2f, g) between WT and medium [Cr] (P = 0.0035) or between WT and high [Cr] groups (P = 0.0042). Activation of this protein was independently confirmed by immunoblotting for phospho α-crystallin B chain (CryAB) when normalised to total α-crystallin B VDAC and β-tubulin (Fig. 2h, i; P = 0.019). This change in α-crystallin B may contribute mechanistically to the protective effects of elevated [Cr] against ischaemia/reperfusion injury (Lygate et al. 2012; Zervou et al. 2016).
The endogenous antioxidant, catalase, was identified by two different peptides using 2DIGE and MS (Fig. 2j–l). In both cases, there was a drop in protein expression in response to augmentation of LV [Cr] and this consistent pattern invited further investigation (ANOVA P = 0.047 and P = 0.036, respectively for the two peptides). Catalase activity assays using LV lysates from the three groups of study showed a decrease between WT and medium [Cr] groups (P = 0.043) (Fig. 2m), but not between medium and high or WT and high [Cr]. A possible reason may be the discrepancy in assay sensitivity levels between proteomics analysis and enzyme activity. Creatine-supplementation has been attributed to direct antioxidant activity (Lawler et al. 2002; Sestili et al. 2006), although this has not been evident in the intact beating heart (Aksentijevic et al. 2014b). One speculative explanation could be that the net antioxidant activity has not changed due to compensatory reduction in catalase.
It is notable that approximately half the differentially regulated proteins identified in our experiments are targets for thioredoxin (Fu et al. 2009) (indicated by a on Suppl. Table 1), which in turn is negatively regulated by thioredoxin interacting protein (Txnip). We have previously reported that Txnip is upregulated in CrT-OE hearts acting as an endogenous inhibitor of further creatine uptake (Zervou et al. 2013).
Initial partial least squares discriminant analysis (PLS-DA) of the 1H-NMR results showed a good separation of the three groups for the aqueous metabolites (n = 6 each) (Suppl Fig. 1A). Representative NMR spectra are shown in Fig. 3.
Aqueous metabolites All detected metabolites are shown in Table 2 and corresponding correlations with LV [Cr] levels in Fig. 4. Creatine detected by HPLC and used for group stratification strongly correlated to the values obtained by 1H-NMR (Fig. 4a; P < 0.0001), providing independent validation for accuracy and reproducibility of estimating LV [Cr] from RV [Cr] measurements. Unexpectedly, we observed multiple strong correlations between metabolite levels and [Cr], for example, a [Cr]-dependent increase in glucose levels (P = 0.002; Fig. 4b) and accumulation of acetyl-carnitine (P = 0.001; Fig. 4d) and carnitine (P = 0.01; Fig. 4e), which suggests an abundance of mitochondrial acetyl-CoA levels (Longnus et al. 2001). In contrast, there was no change in myocardial triglyceride (Suppl Table 4) or lactate levels (Table 2) with elevated [Cr]. The Krebs cycle intermediates, fumarate and succinate, were both reduced with high [Cr] (Fig. 4j, k) and there were strong effects on metabolites associated with anaplerotic flux into the Krebs cycle, e.g. reduced levels of glutamate (Fig. 4h), alanine (Fig. 4f) and aspartate (Fig. 4i). It seems likely that the strong negative correlation with taurine (Fig. 4c) functions to balance the osmotic effect of creatine accumulation, since both are abundant osmolytes (Ito et al. 2008). Phosphocholine levels were reduced by 23 % in the high [Cr] hearts (Suppl Table 4), which could reflect the burden of increased creatine biosynthesis, since choline is a methyl donor for the co-factor s-adenosyl methionine. However, it has been estimated that 95 % of total [Cr] is located in skeletal muscle (Persky and Brazeau 2001), so even a fourfold [Cr] elevation in the heart will have minimal impact on whole body demands.
The increased levels of glutamine we observed could also theoretically contribute to anaplerotic flux via glutamate and α-ketoglutarate, however, this was not detected in the working rat heart perfused with 13C-labelled glutamine, even under pro-anaplerotic conditions (Lauzier et al. 2013). Instead, the authors detected changes in lipid metabolism that could be blocked by inhibiting the hexosamine biosynthetic pathway (HBP) (Lauzier et al. 2013). Glutamine is required for the first step in the HBP which leads to protein O-GlcNAcylation and thereby influences a multitude of cellular functions including known metabolic and cardioprotective proteins (Bond and Hanover 2015). Perfusion with glutamine has been shown to protect against ischaemia/reperfusion injury and again this could by blocked by addition of an HBP inhibitor (Liu et al. 2007). Further work is merited to test the hypothesis that elevated glutamine levels contribute to ischaemic protection in the CrT-OE mice via activation of the HBP.
Lipid metabolites Initial analysis by supervised PLS-DA showed a weak separation of the three groups (Suppl Fig. 1B). The full list of lipid metabolites as quantitated by 1H NMR, is shown on Suppl. Table 4. Some choline is apparently incorporated into additional phosphatidylcholine, which was elevated with high [Cr], while sphingolipid was reduced. Exogenous phosphocreatine has previously been shown to interact with phospholipids and stabilise the plasma membrane (Tokarska-Schlattner et al. 2012) and it is possible that this may also subtly alter the biochemical composition. Other lipid metabolites were not altered to a physiologically relevant level with the caveat that 1H-NMR has low sensitivity (Hinterwirth et al. 2014).
Integrating proteomics and metabolomics
Many of the differentially regulated proteins are involved in energy metabolism and we have attempted to integrate all the relevant proteomic and metabolomic data in a single diagram (Fig. 5). Although we observe clear [Cr] dose-dependency for many metabolites, indicating a continuum, significant changes are only observed in the high [Cr] group. Likewise, we observed reduced expression of multiple metabolic enzymes, but predominantly in the high [Cr] group. Thus, a general pattern emerges of impaired energy-generating pathways in mice with very high [Cr], as follows:
Glucose metabolism impaired glycolysis is in agreement with our previous study, which showed that reduced enolase expression impacted on capacity for lactate production, but only in mice with [Cr] >140 nmol/g protein (Phillips et al. 2010). Elevated PCr may reduce the need for glycolysis to power short-term increases in energy requirements (Safdar et al. 2008). It is notable that total glucose levels are elevated which may arise from either the intra or extracellular pool, suggesting that there is reduced glucose utilisation, consistent with our previous study. Excess glucose is probably converted to glycogen and we have previously shown that glycogen levels are positively correlated with myocardial [Cr] (Lygate et al. 2012). It is not possible to infer the impact (if any) on glucose oxidation since we observed changes in pyruvate dehydrogenase (PDH) subunits that are directionally opposed and lactate was unchanged. Ideally, we would have measured PDH activity biochemically or by hyperpolarised 13C-pyruvate, however, the former requires an entire mouse heart and the latter would require pre-stratification for [Cr] using in vivo 1H-MRS making it impractical and prohibitively expensive. As a surrogate measure, we quantified PDK4 mRNA, since this is a major regulator of PDH activity in the heart (Sugden and Holness 2006) and observed no differences in expression (WT 1.8 ± 0.38; medium 2.317 ± 0.23; high 1.8 ± 0.34, respectively). GLUT1-OE mice also have increased glucose uptake and glycogen stores and are protected against pressure-overload heart failure (Liao et al. 2002) and show improved tolerance to ischaemia (Luptak et al. 2007).
Fatty acid oxidation (FAO) the expression of three key enzymes involved in β-oxidation were reduced in high [Cr] compared to WT hearts (but not in moderate [Cr]), strongly suggesting a deficit in the ability to utilise fatty acids. Presumably, there is commensurate reduction in fatty acid uptake since we do not observe accumulation of lipids, with a trend for reduced triglycerides and significantly lower sphingolipids in the high [Cr] group.
Tricarboxylic acid (TCA) cycle Our findings also suggest an imbalance in the TCA cycle with potential restrictions at the level of isocitrate dehydrogenase and malate dehydrogenase. Changes in metabolite levels suggest altered anaplerotic flux, which might represent a response to these bottlenecks (e.g. glutamate feeding into α-ketoglutarate; aspartate conversion to oxaloacetate). It is surprising that acetyl-carnitine is elevated in high [Cr] hearts since this indicates that substrate availability is higher than demand. Excess acetyl-CoA is converted to acetyl-carnitine and exported out of the mitochondria, to be broken down to its constituent parts in the cytosol, where it may inhibit fatty acid uptake (Longnus et al. 2001). Overall, this suggests that despite acetyl-CoA production via glycolysis and FAO likely to be reduced in the high [Cr] hearts, the rate of acetyl-CoA production still outstrips the capacity of the TCA cycle.
Our analysis represents a snap-shot of protein expression and metabolite concentrations and may not represent dynamic flux through the system. It would have been informative to quantify substrate preference, e.g. by radiolabeled uptake experiments or 13C-hyperpolarisation studies. However, male mice with very high [Cr] are rare (~8 % of all offspring from heterozygote mating) and it took us several years to breed sufficient mice for the current study. Thus, complex experiments requiring high animal numbers are not practicable.
Theoretically, we might have gained further insights using computer modelling. However, the fact that we observed such large and widespread changes in metabolite concentrations is a truly surprising result, particularly given that creatine is at the terminus of energy-generating pathways with few known feedback mechanisms. Hence, existing in silico models are of limited utility in understanding these findings, which suggest deep interconnections that are not part of current metabolic models. For example, the CardioNet metabolic flux model we have used previously (Aksentijevic et al. 2014a) is insensitive to altered creatine levels.
A further potential limitation is that [Cr] correlates closely with LV hypertrophy (Wallis et al. 2005), which has its own metabolic sequelae, raising the possibility that the correlations observed between [Cr] and other metabolites are simply an epiphenomenon. The metabolic response to LVH is characterised by an increase in glycolytic and anaplerotic flux, manifesting as elevated lactate, alanine and aspartate in aortic banded mice and rats (Kolwicz et al. 2012; Sorokina et al. 2007). A similar pattern was observed in hyperthyroid-induced LVH where glucose and glycogen levels were also reduced (Atherton et al. 2011). All these responses are directionally opposite to what we observe in high [Cr] hearts suggesting that the presence of LVH is not driving the metabolic phenotype. It should also be noted that the severity of LVH in our study is relatively mild. Myocardial cross-sectional area increased by 17 % in high vs normal [Cr], compared with an increase of 56 % in the transverse aortic constriction model (Lygate et al. 2007), therefore, the driving force for metabolic remodelling is not as strong.
Our model of augmented cardiac creatine is a result of transgenesis and therefore, a ‘forced’ metabolic phenotype that does not occur naturally. Whether there is metabolic feedback at physiological creatine levels that is lost under the pathologically low [Cr] levels observed in the failing heart remains to be established. Nevertheless, our findings are highly informative when considering target levels for therapeutic [Cr] aimed at improving cardiac energetics, most likely via pharmacological activation of the creatine transporter (Zervou et al. 2016). Only the high [Cr] group had a severe metabolic phenotype, which supports the concept of a safe window for creatine elevation between 20 and 100 % above wild-type levels (i.e. corresponding to the medium [Cr] group), which we previously demonstrated does not impact on in vivo function and protects against ischaemia–reperfusion injury (Lygate et al. 2012).
Finally, we took a non-biased approach to identify proteomic and metabolic adaptations in response to elevated myocardial creatine levels in vivo. In mice with very high [Cr], we observed reduced expression in multiple proteins involved in energy generation, implying impairment of glycolysis, fatty acid oxidation and the TCA cycle, resulting in a substrate rich, but energy-poor heart. Surprisingly, strong correlations were observed between creatine tissue levels and many key metabolites suggesting the existence of hitherto unsuspected feedback mechanisms. The potential link between creatine and glucose uptake is of particular interest for future study.