Arachidonate 12/15-Lipoxygenase–Induced Inflammation and Oxidative Stress Are Involved in the Development of Diabetic Cardiomyopathy

Recently, the incidence of diabetes has increased rapidly, due to worldwide changes in lifestyle. The epidemic of obesity together with sedentary lifestyles is projected to result in over 300 million people having diabetes by 2025 (1). According to the Framingham study, the relative rate of onset of heart failure in the diabetes is significantly high, and onset of heart failure leads to a poor prognosis for diabetic patients (2). Diabetes is an independent risk factor for heart failure. In addition, the number of cases of heart failure combined with diabetes continues to rise, and treatment will become a significant problem in the future (3). However, the detailed molecular mechanism underlying development of this pathological condition has not been elucidated yet (4,5).

Arachidonic acid (AA) is a free fatty acid that is released from the cell membrane in response to various cytokines, peptides, and growth factors that become active under inflammatory conditions (6). There are three families of enzymes involved in the oxidative metabolism of AA. These include the lipoxygenases (LOXs), which produce leukotrienes, hydroperoxyeicosatetraenoic acids, hydroxyeicosatetraenoic acids (HETEs), and hydroxyoctadecadienoic acids; the cyclooxygenases (COXs) COX-1 and COX-2, which produce prostaglandins including G2 and H2 as well as thromboxanes; and cytochrome P-450 monooxygenases, which produce epoxides and HETEs (7). The human LOX enzymes include 5-LOX, 12-LOX, and 15-LOX (8,9). LOX enzymes are named according to the specific carbon atoms of AA that are oxidized. Thus 12/15-LOX is a member of the LOX family that catalyzes the step from AA to 12(S)-HETE and 15(S)-HETE (10). 12/15-LOX was originally isolated from porcine leukocytes (11), but its tissue distribution is now known to be relatively wide, including blood vessels, the brain, and the kidney (12).

Several lines of evidence have suggested that 12/15-LOX may play an important role in the development of atherosclerosis, nephropathy, and neuropathy in diabetes (13–16). However, there is currently little evidence about a role played by 12/15-LOX in the development of diabetic cardiomyopathy. Therefore, we focused on the role of 12/15-LOX as a key molecule related to diabetic cardiomyopathy.

In the current study, we found that 12/15-LOX expression was significantly upregulated in the heart of diabetic mice. We have shown that diabetes-induced activation of cardiac 12/15-LOX increased inflammation and oxidative stress in the diabetic heart. Conversely, disruption of 12/15-LOX reduces inflammation, oxidative stress, and fibrosis in the diabetic heart, thereby improving systolic dysfunction. These findings suggest that inhibition of 12/15-LOX could be a novel treatment for diabetic cardiomyopathy.

All experimental protocols were approved by the Jikei University review board. Male C57BL/6 and Alox15-deficient (12/15-LOX KO) mice aged 7 weeks were used in this study. 12/15-LOX KO mice on a C57BL/6 background were purchased from The Jackson Laboratory. For the diabetic model, 7-week-old male C57BL/6 and 12/15-LOX KO mice were intraperitoneally injected with a single dose of streptozotocin (STZ) at 150 mg/kg body weight (Wako 545-00283). One week after induction of diabetes, the antioxidant N-acetylcysteine (NAC; SIGMA 9165) was administered to the three groups in the drinking water for 15 weeks (average 1.44 g/kg/day). 12/15-LOX inhibitor cinnamyl-3,4-dihydroxy-cyanocinnamate (CDC; BIOMOL International, LP) (17), 8 mg/kg once a day, was administered by subcutaneous injection for 4 weeks after induction of diabetes. We purchased the male db/db mice aged 11 weeks from CLEA Japan Inc. and used in this study.

Echocardiography was performed with a Vevo 770 high-resolution imaging system (VisualSonics Inc.). To minimize variability of the data, heart rate was ∼500–600 bpm when cardiac function was assessed. Average systolic pressure and heart rate were recorded by a photoelectric pulse device (Softron BP-98A, Softron) placed on the tail of conscious mice. Frozen 4-μm cross sections of the heart were fixed in 4% paraformaldehyde and subjected to immunofluorescence staining for 12/15-LOX (Abcam ab87353), 4-Hydroxy-2-nonenal (4-HNE) (Abcam ab20953), and wheat germ agglutinin. Nuclei were stained with Hoechst (H3570, Life Technologies, Japan). Pictures were taken on a Biorevo microscope (BZ-9000, Keyence). Cardiac tissue was fixed by perfusion with 20% formalin neutral buffer solution (136-10041, Wako), sectioned at 4-μm thickness, and stained. For measurement of the percentage area of cardiac fibrosis, we selected five fields at random and calculated the ratio of Masson-stained fibrosis area–to–total myocardium area with Photoshop CS5 (Adobe Systems Inc.) as described previously (18).

Total RNA was isolated from the hearts of mice with RNA-Bee reagent (Cosmo Bio Inc.). In brief, after preparing total RNA, first-strand cDNA was synthesized with QuantiTect reverse transcription (Qiagen Inc.). Real-time PCR was performed using Thermal Cycler Dice TP800 (Takara Bio Inc.) with SYBR Premix Ex Taq II (Takara Bio Inc.) according to the manufacturer’s instructions. The specific oligonucleotide primers for GAPDH, tumor necrosis factor (TNF)-α, collagen-1α2, and collagen-3α1 were selected using Primer3 (version 0.4.0) (http://frodo.wi.mit.edu/).

Whole cell lysates were prepared in radioimmunoprecipitation assay buffer (R0278, Sigma-Aldrich Japan Ltd.). Lysates (30 μg) were resolved by SDS-PAGE. Proteins were transferred to a polyvinylidene fluoride membrane (Life Technologies), which was incubated with the primary antibody, followed by anti-mouse, anti-rabbit IgG, light chain specific (The Jackson Laboratory). Specific proteins were detected using enhanced chemiluminescence (ECL Prime, GE Healthcare). The primary antibodies used for Western blotting were as follows: 4-HNE (Abcam ab20953), α-tubulin (sc-5286, Santa Cruz Biotechnology Inc.) and actin (A4700, Sigma-Aldrich Japan Ltd.), nuclear factor (NF)-κB p50, phosphor-NF-κB p65 (3035 and 3039, Cell Signaling Technologies), histone H3 (3638, Cell Signaling Technologies), Bip, and CHOP (3183 and 2895, Cell Signaling Technologies). ELISA was performed according to the manufacturer’s instructions to examine the levels of 12(S)-HETE and 15(S)-HETE (Assay Designs). We used the NE-PER (nuclear and cytoplasmic extraction kit, Thermo Fisher Scientific Inc.) for the extraction of nuclear proteins. ImageJ software was used (http://rsbweb.nih.gov/ij/) to analyze intensity of Western blot bands.

Cardiomyocytes and cardiac fibroblasts were prepared from neonatal rats and cultured as described previously (19). Cardiomyocytes and cardiac fibroblasts were exposed to a high glucose concentration (HG; 25 mmol/L glucose) and harvested at the indicated time points. Those were treated with normal glucose concentration (LG; 5.5 mmol/L glucose) and 19.5 mmol/L mannitol to control for osmolarity.

Cardiomyocytes were cultured in media with HG with or without 10 μmol/L CDC and 1 mmol/L NAC (Sigma-Aldrich Japan Ltd.) or with LG for 24–72 h at 37°C. Palmitic acids (100 μmol/L) were treated after 48 h after split of cardiomyocyte and incubated for 24 h.

We assayed reactive oxygen species (ROS) production by using chloromethyl-2,7-dichlorodihydro-fluorescein diacetate (CM-H2DCFDA; C6827, Invitrogen). Cardiomyocytes were cultured with 5 μmol/L CM-H2DCFDA at 37°C in the dark for 30 min and fixed the formation. CDC (10 μmol/L) was added 30 min before treatment of high glucose. Pictures were taken on a Biorevo microscope (BZ-9000, Keyence). Fluorescence intensity was assessed for image analysis of the histogram by Photoshop CS5.

The mitochondrial membrane potential (ΔΨm) was assessed using MitoTracker red. Briefly, cardiomyocytes were loaded with 0.5 μmol/L MitoTracker orange CMTMRos (M7510, Invitrogen) at 37°C for 30 min, washed, and fixed the formation. Fluorescence intensity was assessed by Photoshop CS5.

Data are shown as the mean SEM. Multiple group comparisons were performed by one-way ANOVA, followed by Bonferroni test for comparison of means. Comparisons between two groups were made using the two-tailed unpaired Student t test or two-way ANOVA. In all analyses, P < 0.05 was considered statistically significant.

To determine whether 12/15-LOX is a key molecule in the development of diabetic cardiomyopathy, we examined the mRNA level of 12/15-LOX in the diabetic heart using real-time PCR (RT-PCR). We created STZ-induced diabetic mice (WT-STZ) and compared them with control (saline) mice (WT). Cardiac expression of 12/15-LOX was significantly upregulated, with a peak at 4 weeks after STZ treatment in WT-STZ (Fig. 1A). We next examined a production of 12(S)-HETE and 15(S)-HETE, a major metabolite of 12/15-LOX. Production of 12(S)-HETE and 15(S)-HETE was also significantly increased in the hearts of WT-STZ compared them with WT through the protocol until 16 weeks after STZ treatment (Fig. 1B). Immunohistochemistry showed that expression of 12/15-LOX was specifically upregulated in cardiomyocytes of the diabetic heart, but not in vascular cells and fibroblast cells (Fig. 1C).

To investigate the relationship between upregulation of 12/15-LOX and STZ-induced diabetic cardiomyopathy, we created WT-STZ using 12/15-LOX KOand compared them to WT-STZ. Animals were divided into four groups (WT, KO, WT-STZ, and KO-STZ). Blood pressure (systolic blood pressure and diastolic blood pressure), heart rate, plasma glucose levels and body weight of each animal were measured at 0, 8, and 16 weeks after saline (WT, KO) or STZ (WT-STZ, KO-STZ) treatment. Single injection of STZ (150 mg/kg i.p.) to adult male mice resulted in strongly high plasma glucose levels for a week after STZ treatment. Same dose of saline was injected as a control. High-dose STZ-induced model has been established as a diabetes model and recognized as a useful model of diabetic cardiomyopathy (13,20). There was no difference in blood pressure and heart rate among the four groups under normal or STZ-induced high glucose conditions. Plasma glucose levels remained around 10 mmol/L in WT and KO without STZ treatment. No difference in plasma glucose level was observed between WT and KO, WT-STZ, and KO-STZ at every week of age. Although body weight did not have a tendency to increase at advancing age in mice after STZ treatment than WT mice, there was no difference between WT and KO, WT-STZ, and KO-STZ (Table 1). The production levels of 12(S)-HETE and 15(S)-HETE in the hearts of KO and KO-STZ were equal to or less than those of WT (Supplementary Fig. 1).

Echocardiography was performed at 0, 8, and 16 weeks after saline or STZ treatment and compared between the four groups. Fractional shortening (FS) was gradually impaired and left ventricular systolic dimension (LVD) was increased in WT-STZ from 8 weeks after induction of diabetes compared with those of WT. In contrast, systolic dysfunction and left ventricular dilatation were not observed in KO and KO-STZ (Fig. 2A). Histological examination revealed that 16 weeks after induction of diabetes, the fibrotic area progressively extended to the perivascular and interstitial areas in the WT-STZ heart compared with the WT heart. This increase was significantly inhibited in the KO-STZ heart compared with that of WT-STZ (Fig. 2B). Quantitative analysis revealed that the percentage of cardiac fibrosis was significantly decreased in KO-STZ compared with those of WT-STZ (Fig. 2C). These results suggest that 12/15-LOX affects cardiac dysfunction of STZ-induced diabetic cardiomyopathy, and the increased expression of 12/15-LOX induced by hyperglycemia might cause cardiac dysfunction and fibrosis.

We next examined the expression of inflammatory cytokine genes by RT-PCR. We found that cardiac expression of TNF-α was upregulated in the WT-STZ heart compared with the WT heart (Fig. 3A). Cardiac collagen markers such as collagen-1α2 and collagen-3α1 were also increased in the WT-STZ heart compared with the WT heart every 4 weeks after STZ treatment; these increases were canceled by disruption of 12/15-LOX (Fig. 3B). Moreover, we examined the relationship between 12/15-LOX expression and activation of NF-κB in the WT-STZ heart and the KO-STZ heart by Western blot analysis. Cardiac activation of NF-κB was upregulated in the WT-STZ heart compared with the WT heart. This upregulation of NF-κB was also inhibited by disruption of 12/15-LOX (Fig. 3C and D), indicating that 12/15-LOX may induce cardiac inflammation, which was involved in the development of STZ-induced diabetic cardiomyopathy.

We examined the relationship between 12/15-LOX and cardiac oxidative stress in the diabetic heart. Immunohistological staining and Western blotting showed that expression of cardiac 4-HNE, a major marker of oxidative stress, was upregulated in myocardium in WT-STZ heart compared with WT heart, and this enhancement was significantly inhibited by disruption of 12/15-LOX (Fig. 4A–C). We next examined the NAD phosphate (NADPH) oxidase isoforms by RT-PCR at 4 weeks after induction of diabetes. Among the NADPH isoforms, expression of NADPH oxidase 4 (Nox4) but not NADPH oxidase 2 (Nox2) was upregulated in the WT-STZ heart compared with the WT heart and the KO heart. This increase was also inhibited in the KO-STZ heart (Fig. 4D).

To investigate the subcellular mechanism of the increase in ROS in the diabetic heart, intracellular ROS levels in cardiomyocytes were estimated under HG and LG by fluorimetry in vitro. Cardiomyocytes under HG showed enhancement of fluorescence intensity of DCFDA by sevenfold at 24 h compared with findings for those under LG (Fig. 4E and F). Treatment with 12/15-LOX inhibitor (CDC) inhibited the enhancement of DCFDA fluorescence under HG condition (Fig. 4E and F). To further investigate the subcellular origins of ROS production in the diabetic heart, we used MitoTracker red to evaluate the effect of HG on the mitochondrial membrane potential in cardiomyocytes. The result showed that there was loss of mitochondrial membrane potential (ΔΨm) as indicated by a decrease in the fluorescence intensity of MitoTracker red in cardiomyocytes under HG. This decrease in the fluorescence intensity was improved by treatment with CDC (Fig. 4E and F).

These in vivo and in vitro results suggest that activation of NADPH oxidase and mitochondrial membrane abnormalities were occurring in the diabetic heart and the 12/15-LOX pathway is involved in the process of production of ROS and oxidative stress in the diabetic heart.

To investigate the relationship between cardiac oxidative stress and cardiac inflammation, we examined the hearts of WT-STZ who were administered the antioxidant NAC in their drinking water (WT-STZ NAC) throughout the experimental period and compared them to the heart of WT, KO, WT-STZ, and KO-STZ. NAC was administered to diabetic mice for 15 weeks, and histological analysis and cardiac expressions of TNF-α, collagen-1α2, and collagen-3α1 were examined. Histological examination revealed that cardiac perivascular fibrosis was improved in the WT-STZ NAC heart as well as the KO-STZ heart compared with findings in the WT-STZ heart (Fig. 5A). Quantitative analysis revealed that the percentage of cardiac fibrosis was significantly decreased in KO-STZ and WT-STZ NAC hearts compared with those of WT-STZ hearts (Fig. 5B). Moreover, RT-PCR showed that increased expression of TNF-α, collagen-1α, and collagen-3α1 in WT-STZ NAC and KO-STZ hearts were significantly inhibited compared with the expressions in WT-STZ hearts (Fig. 5C and D). These results indicate that cardiac oxidative stress has a major role in promoting cardiac inflammation in the STZ-induced diabetic heart.

To further investigate the role of 12/15-LOX in the heart under HG, we cultured neonatal cardiomyocytes and cardiac fibroblasts under LG and HG using mannitol to adjust osmotic pressure for 24, 48, and 72 h. RT-PCR showed that hyperglycemia upregulated the expression of 12/15-LOX and production of 12(S)-HETE and 15(S)-HETE in cardiomyocytes but not in fibroblasts (Fig. 6A and B and Supplementary Fig. 2).

Furthermore, RT-PCR demonstrated the upregulation of TNF-α, collagen-1α2, and collagen-3α1 in cardiomyocytes under HG after 24 h (Fig. 6C and D). To determine whether increased expression of cardiac 12/15-LOX affects the upregulation of TNF-α, collagen-1α2, and collagen-3α1 in cardiomyocyte, we cultured neonatal cardiomyocytes with CDC under HG and examined the expression of each marker. Increased expression of TNF-α, collagen-1α2, and collagen-3α1 under HG were inhibited by treatment with CDC (Fig. 6C and D). These results suggest that cardiac 12/15-LOX induced by hyperglycemia implicated the upregulation of inflammatory cytokines such as TNF-α, collagen-1α2, and collagen-3α1.

We next investigated the relationship between the oxidative stress and inflammation induced by hyperglycemia in cardiomyocytes. In treatment with NAC under HG, RT-PCR showed that upregulation of TNF-α, collagen-1α2, and collagen-3α1 in cardiomyocytes was inhibited (Fig. 7A and B). Moreover, Western blotting revealed that nucleus cardiac activation of NF-κB using anti-NF-κB p50 antibody and anti-phospho-NF-κB p65 (Ser468) antibody was upregulated under HG compared with under LG after 24 h. This activation of NF-κB was significantly inhibited by treatment with NAC in cultured cardiomyocytes (Fig. 7C and D).

In the current study, we demonstrate that the arachidonate 12/15-LOX is an important molecule involved not only in the onset and development of diabetic cardiomyopathy, but also in increased inflammation and oxidative stress in the heart.

It has been known for a long time that after STZ administration, alternations occur in the enzymatic activity of cardiac contractile proteins, such as myosin and actomyosin, in rats and mice and that cardiac contractile function gradually deteriorates (4,21). This finding has provided a basis for the well-known diabetic cardiomyopathy model, where STZ administration quickly increases plasma glucose levels and high plasma glucose levels are maintained for a long time, as shown in our experiments as well. In the hearts of wild-type mice exposed to increased plasma glucose levels, cardiac 12/15-LOX expression peaked at 4 weeks, and its upregulation persisted until 16 weeks after administration. Immunohistochemistry showed that the expression of 12/15-LOX in the heart was specifically upregulated in cardiomyocytes but not in vascular and fibroblast cells. Cardiac function gradually deteriorated from 8 weeks onward after STZ administration, whereas this deterioration was suppressed in 12/15-LOX KO. These results suggest that 12/15-LOX is involved in the onset and development of STZ-induced diabetic cardiomyopathy.

AA metabolites are known to be involved in the development of tissue inflammation and fibrosis in noncardiovascular diseases such as pulmonary fibrosis (22,23), suggesting an important role for AA as a potential pathway in the pathogenesis of cardiac fibrosis. A number of papers have demonstrated that the 12/15-LOX pathway induces inflammation in a variety of tissues. For example, overexpressed 12/15-LOX in macrophage-like J774.1 cells and human vascular smooth muscle cells increased production of inflammatory cytokines such as interleukin-6 and TNF-α (24,25). In an experimental asthma model, 12/15-LOX inhibition reduced airway inflammation and attenuated cell injury (26). Furthermore, 12/15-LOX caused cell growth of cardiac fibroblasts and matrix production in the heart (27), and mice with cardiac overexpression of 12/15-LOX showed increased the expressions of MCP-1 and TNF-α, thereby inducing cardiac fibrosis and dysfunction (28). TNF-α is an important molecule that triggers inflammation and cell injury in the heart (29), which is known to induce cardiac fibrosis as a consequence (30,31). Its suppression has been shown to improve diabetic cardiomyopathy, reduce cardiac fibrosis, and recover cardiac function (21,32,33). In agreement with these findings, our results showed that the expressions of TNF-α and collagen markers were elevated accompanied by an increase in the expression of 12/15-LOX in the STZ-induced diabetic heart, whereas these increases were suppressed in 12/15-LOX KO, leading to cardiac function being preserved. Moreover, administration of a 12/15-LOX inhibitor (CDC) suppressed the upregulation of TNF-α associated with high blood glucose levels in vitro. Our findings suggest that 12/15-LOX–induced TNF-α may have a major role in the development of cardiac fibrosis in STZ-induced diabetic cardiomyopathy.

Oxidative stress has been well-known to be involved in the development of diabetic cardiomyopathy, and increased ROS production is shown to induce various cardiovascular complications, including cardiac dysfunction (32). Accumulating evidence suggests that mitochondria and NADPH oxidase important sources of ROS production in the diabetic heart (34–36), where the crosstalk between mitochondria and NADPH oxidase is important and represents a feed-forward vicious cycle of ROS production (37,38). It is important to elucidate the cellular mechanisms of ROS increase in the diabetic heart. Our histological examination revealed that cardiac production of 4-HNE was increased in the diabetic heart and DCF fluorescence showed that intracellular ROS levels in cardiomyocytes were increased under high glucose conditions. We also showed the cardiac expression of NOX4, one of the NADPH homologs, is elevated in the diabetic heart, and the mitochondrial membrane potential is decreased in cardiomyocytes exposed to high glucose levels in vitro. Moreover, disruption of 12/15-LOX decreased the production of 4-HNE and expression of NOX4 in the diabetic heart, while 12/15-LOX inhibitor (CDC) decreased intracellular ROS levels in cardiomyocytes and restored the mitochondrial membrane potential. These results suggest that oxidative stress is increased in the diabetic heart and production of ROS in cardiomyocyte generated from mitochondria and NADPH oxidase plays a pivotal role in the onset and development of diabetic cardiomyopathy. Moreover, our results suggest that 12/15-LOX is an important molecule involved in the process of production of ROS and oxidative stress in the diabetic heart.

Recently, diabetic patients with hypertension have increased in numbers, with the risk of cardiovascular disease thought to be increased in these patients. Research into molecules involved in both diseases is of critical importance, given that any such molecule will likely represent a viable therapeutic target in the future. Diabetes not only causes hyperglycemia and increases oxidative stress, but also exhibits a complicated disease condition associated with concomitant hypertension, eventually leading to atherosclerosis and organ injury (39). Persistent hyperglycemia associated with diabetes, insulin resistance, and postprandial hyperglycemia induces inflammation and oxidative stress and causes myocardial injury (40,41). We previously confirmed by microarray analysis that 12/15-LOX was upregulated in the heart of a hypertensive heart failure model and demonstrated that it is involved in the inflammation of the heart and development of cardiac fibrosis (28). Our present in vivo and in vitro studies demonstrate that 12/15-LOX is involved in the development of disease not only in the hypertensive heart failure model but also in the STZ-induced diabetic cardiomyopathy model, suggesting that 12/15-LOX is an important molecule affecting the heart in both diseases.

It has been shown that 12/15-LOX is involved in the onset of a variety of diseases associated with type 1 and type 2 diabetes, such as cardiovascular disease, hypertension, renal disease, and neurodegenerative disorders. Furthermore, the role of 12/15-LOX has been suggested in humans as well. Indeed, the expression of 12/15-LOX was shown to be an independent risk factor for atherosclerosis in 828 patients with diabetes (42), and 12/15-LOX inhibition is shown to lead to reductions in the size of cerebral infarction in humans (43–46).

We acknowledge several limitations of our study. First, not all groups were evaluated over the entire course of cardiac fibrosis development in the study. Second, we could not entirely exclude the effects of the pharmacological agents used, such as CDC and NAC, especially on cardiac fibrosis, due to lack of several control groups in the study. Furthermore, while we demonstrated that the active involvement of ROS in high glucose leads to the upregulation of 12/15-LOX, which, in turn, results in cardiac fibrosis, the precise mechanisms linking 12/15-LOX to profibrotic signaling still remain largely unknown. Despite these limitations, however, we believe we have provided important new insights into diabetic cardiomyopathy.

In conclusion, this study demonstrated the mechanisms through which the expression of 12/15-LOX in the heart, associated with persistent hyperglycemia, leads to the development of diabetic cardiomyopathy via cardiac inflammation and oxidative stress. Moreover, it was suggested that inhibition of 12/15-LOX could potentially be a useful treatment for not only diabetic cardiomyopathy but also diabetes complications.

Acknowledgments. The authors thank Y. Inada and Y. Takada for research assistance and technical support (Division of Diabetes, Metabolism and Endocrinology, Department of Internal Medicine, Jikei University School of Medicine).

Funding. This work was supported by a Grant-in-Aid for Scientific Research from the Ministry of Education, Culture, Sports, Science and Technology of Japan and from Uehara Memorial Foundation (to Y.K.).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. H.S. conducted experiments, performed data analyses, and wrote the manuscript. Y.K. and M.S. designed research, analyzed data, and wrote the manuscript. H.I., I.S., T.Y., and D.K. contributed to experimental work (staining analysis and animal care). T.N., K.T., and M.Y. contributed to discussion. T.M. and K.U. designed research and contributed to the review/editing of the manuscript. K.U. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.