Impaired Hepatic Glucose Metabolism Is Associated With Glucose Intolerance in Mice Carrying α₂ Isoform Na,K‐ATPase Mutation

ABSTRACT

Aim

The Na,K‐ATPase is important for energy demanding cellular processes, including essential components of substrate metabolism and metabolic flexibility. Mice heterozygous for the hemiplegic migraine‐related mutation, G301R, of the Na,K‐ATPase α₂ isoform (α₂ ^+/G301R) demonstrated altered metabolism in the heart. We hypothesized that mitochondrial dysfunction in α₂ ^+/G301R mice is associated with abnormal systemic glucose metabolism.

Methods

3‐ and 8‐month‐old α₂ ^+/G301R and wild type (WT) mice of both sexes were metabolically phenotyped. Glucose uptake in peripheral tissues was assessed by [¹⁸F]‐2‐fluoro‐2‐deoxyglucose (FDG) MRI. Mitochondrial function in liver was assessed with high‐resolution respirometry. Protein expression was assessed by proteomics and Western blot. Glycated hemoglobin (HbA1c) was measured with immunoassay.

Results

The 3‐month‐old α₂ ^+/G301R mice had normal responses to glucose tolerance test, but males showed a reduced insulin response. The 8‐month‐old α₂ ^+/G301R mice exhibited glucose intolerance associated with impaired glucose uptake in the liver while unaltered in skeletal muscle associated with unchanged GLUT‐4 and insulin receptor levels. Insulin secretion and insulin sensitivity were similar between 8‐month‐old genotypes. In the liver, 8‐month‐old α₂ ^+/G301R mice of both sexes showed reduced expression of the α₂ Na,K‐ATPase and several key mitochondrial proteins from the electron transport chain. Accordingly, mitochondrial oxygen consumption was reduced in 8‐month‐old α₂ ^+/G301R males, and 8‐month‐old α₂ ^+/G301R females showed a complex I‐mediated dysfunction in the liver mitochondria.

Conclusions

Migraine‐associated mutation of the Na,K‐ATPase α₂ isoform was associated with impaired mitochondrial respiratory capacity in the liver. Hence, α₂ ^+/G301R mice exhibited reduced capacity of ATP‐demanding metabolic pathways in the liver resulting in age‐dependent impaired glucose homeostasis.

Practioner Points

• The Na,K‐ATPase α₂ isoform (ATP1A2) plays an essential role in cellular ion homeostasis and energy metabolism. Genetics variants in ATP1A2 cause familial hemiplegic migraine type 2 (FHM2), and migraine with aura has been associated with systemic metabolic disturbances and impaired energy metabolism. However, whether ATP1A2 mutations influence peripheral metabolic tissues or contribute to systemic glucose dysregulation has not previously been investigated.

• Using mice heterozygous for the clinically relevant G301R mutation in ATP1A2, we demonstrate that this migraine‐associated variant causes age‐dependent glucose intolerance. The metabolic phenotype is driven by impaired hepatic glucose metabolism rather than altered skeletal muscle glucose uptake or insulin sensitivity. Proteomic and functional analyses further reveal reduced expression of mitochondrial electron transport chain proteins in the liver and impaired mitochondrial respiratory efficiency.

• These findings identify the liver as a previously unrecognized metabolic target of ATP1A2 dysfunction and suggest that impaired mitochondrial energy metabolism may link migraine‐associated ion transport defects to systemic metabolic alterations. This work expands the concept of migraine as a disorder involving disturbed energy metabolism and highlights potential metabolic consequences of ATP1A2 mutations beyond the nervous system.

1. Introduction

Familial hemiplegic migraine type 2, a type of migraine with aura, is caused by mutations in the ATP1A2 gene encoding the Na,K‐ATPase α₂ isoform, including the missense mutation, G301R [1, 2]. Interestingly, migraine with aura has been shown to be associated with increased prevalence of metabolic syndrome [3]. However, the underlying mechanisms are not fully understood.

The Na,K‐ATPase plays an essential role in membrane ion transport. Another important function of the Na,K‐ATPase is the regulation of several intracellular signaling pathways, including Src kinase signaling [4, 5, 6, 7]. The Na,K‐ATPase‐dependent Src signaling is involved in the modulation of electron transport chain function and mitochondrial uncoupling [8, 9]. Knock‐out of the Na,K‐ATPase α₁ isoform leads to amplified Src kinase signaling associated with impaired capacity to increase mitochondrial respiration to meet metabolic needs [8]. This imbalance in Na,K‐ATPase‐dependent signaling is suggested to be caused by dysfunction of electron transport chain complex I [8]. Dysregulation of the Na,K‐ATPase and downstream intracellular signaling contributes to complications in diabetes [10, 11, 12].

We recently characterized mice heterozygous for this mutation, α₂ ^+/G301R mice, with mitochondrial uncoupling in the heart associated with reduced cardiac function [9]. Mitochondrial uncoupling in α₂ ^+/G301R mice was associated with altered expression of the Na,K‐ATPase α isoforms and amplified Src kinase signaling. Interestingly, the cardiac phenotype was observed in adult but not young mice suggesting age‐dependent changes in cellular metabolism associated with imbalanced Na,K‐ATPase‐dependent intracellular signaling.

We hypothesized that the G301R mutation in the Na,K‐ATPase α₂ isoform leads to glucose intolerance caused by mitochondrial dysfunction and thus impaired glucose metabolism in the liver. We tested the hypothesis using glucose tolerance test and insulin tolerance test in α₂ ^+/G301R mice and wild type (WT) controls. We compared genotypes of both sexes and studied 3‐month‐old and 8‐month‐old mice, addressing a sex and age‐dependent metabolic phenotype. We used positron emission tomography/magnetic resonance imaging (PET‐MRI) and a whole‐body composition MR scanning to characterize the metabolic phenotype. These functional experiments were further strengthened by biochemical experiments and proteomics of liver tissue aligned with mitochondrial oxygen consumption and activity assessment using high‐resolution respirometry. With the use of these state‐of‐the‐art methods, we uncovered a mechanism that may underlie the metabolic phenotype in α₂ ^+/G301R mice.

2. Methods

2.1. Animal Experiments

Experiments in this study abided by the European Convention for the Protection of Vertebrate Animals used for Experimental and other Scientific Purposes, and reported in accordance with the ARRIVE (Animal Research: Reporting in vivo Experiments) guidelines. The Animal Experiments Inspectorate of the Danish Ministry of Environment and Food preapproved all conducted experiments (permit no. 2016‐15‐0201‐00982).

Mice were kept at a 12/12 h light/dark cycle with food and water available ad libitum. Heterozygous α₂ ^+/G301R mice on the C57BL/6J background were generated and bred as previously described [13, 14]. This is a global constitutive gene modification where the expression of Na,K‐ATPase α2 isoform is reduced in all tested tissues that express this isoform, possibly because of inefficient translation or quick degradation of the mutated enzyme [9, 13, 14]. The homozygote G301R mutation is lethal at birth [14] because of abolished respiration by the medullary respiratory center neurons [15]. Since diseases related to glucose metabolism often change with age [16], we studied both 3‐ and 8‐month‐old mice. Also, migraine is known to affect females more frequently than males [17]. Therefore, we included female and male mice and analyzed results for each sex and age group separately.

2.2. Western Blot

Mice were sacrificed by cervical dislocation. Middle liver lobe and soleus muscle were dissected in saline and snap frozen in liquid nitrogen immediately after dissection. Frozen tissue was mechanically homogenized (Tissuelyser II, Qiagen, Hilden, Germany) and placed in lysis buffer (Pierce IP lysis buffer, Thermo Scientific, Waltham, MA, USA) with Halt protease inhibitor cocktail (Thermo Scientific). Samples were stored at −80°C.

Total protein was measured using Pierce BCA protein assay kit (Thermo Scientific) and protein concentration was adjusted with 1:1 m‐Q water and 4× laemmli, with 50 mM DL‐Dithiothreitol sample buffer (Bio Rad, Hercules, CA, USA).

Protein samples were loaded on precast 4%–15% Criterion TGX stain‐free gels (Bio Rad, USA) and gel electrophoresis was run in 25 mM Tris/192 mM glycine/0.1% SDS, pH 8.3 running buffer. Total protein load on the stain‐free gels was detected with ultraviolet lighting imaging system (Azure c600, Azure Biosystems, Dublin, CA, USA). Proteins were blotted onto polyvinylidene difluoride membranes (Merck Millipore, Burlington, MA, USA) in 96 mM glycine/12 mM Trisbase/20% ethanol transfer buffer. The membranes were blocked in 3% nonfat dry milk (used for Na,K‐ATPase α₁ and α₂ isoform protein blots) or 3% bovine serum albumin (used for glucose transporter type 4 (GLUT‐4), insulin receptor β blots) in Tris‐buffered saline with 0.1% vol/vol Tween 20 (TBS‐T) for 2 h. The membranes were then incubated with primary antibody overnight at 4°C followed by washing in TBS‐T and 2 h incubation with secondary antibody. Bound antibodies were visualized with ECL (Amersham ECL Western blotting detection reagent, Cytiva, MA, USA) followed by chemiluminescence detection.

Secondary antibodies used, unless otherwise stated, were goat‐anti‐rabbit horseradish peroxidase (HRP) (1:2000, Cell Signaling Technologies, #7074, Danvers, MA, USA) or horse‐anti‐mouse HRP (1:2000, Cell Signaling Technologies #7076).

For skeletal muscle immunoblot protein detection, the following primary antibodies were used: Na,K‐ATPase α₁ isoform (1:2000 Novusbio, NBP2‐61137H, Centennial, CO, USA), Na,K‐ATPase α₂ isoform (1:2000, Chemicon, #AB9094). GLUT‐4 (1:500, Sigma Aldrich, #07‐1404, St. Louis, MO, USA) and insulin‐receptor β‐subunit (1:800, Cell Signaling Technologies #3025). For liver lysate immunoblot protein detection, the following primary antibodies were used: Na,K‐ATPase α₁ isoform (NASE, 1:3000), Na,K‐ATPase α₂ isoform (HERED, 1:1000) [18]. Both antibodies were gift from Professor Sandrine Pierre (Marshall University, Huntington, WV, USA). For intracellular signaling semi‐quantification in liver lysates, the following antibodies were used: bulk Src (1:500; sc‐8056; Santa Cruz Biotechnology Inc., Dallas, TX), Src phosphorylated at pY418 (1:200; No. 44660G; ThermoFisher Scientific), bulk MAPK (1:3000; No. CST4696; Cell Signaling Technology Inc., Danvers, MA), MAPK phosphorylated at Thr202/Tyr204 (1:2000; No. CST9101; Cell Signaling Technology Inc), bulk PLCγ (1:3000; No. CST2822; Cell Signaling Technology Inc) and PLCγ phosphorylated at Tyr783 (1:3000; No. CST2821; Cell Signaling Technology Inc) [9]. Samples for detection of the Na,K‐ATPase α₁ and α₂ isoforms were treated for denaturation step at 37°C for 30 min. The samples for detection of all other proteins of interest were boiled at the denaturation step.

The protein expression was quantified using ImageJ (NIH, USA) and normalized for total protein load detected with stain‐free gel.

2.3. Mitochondrial Respirometry

Animals were euthanized by cervical dislocation, and the liver was excised and immediately submerged in cold 0.9% KCl. After < 5 min, the liver was transferred to approximately 10 mL of medium A (250 mM sucrose, 10 mM Trizma base, 3 mM EGTA) and cut into smaller pieces. A homogenate was then made by transferring and mashing the tissue in a Potter‐Elvehjem homogenizer. Subsequently, the homogenate was centrifuged at 800g for 10 min. The supernatant was pipetted off, filtered through gauze, and centrifuged again at 7200g for 10 min to isolate mitochondria. After centrifugation, the pellet was resuspended in medium B (250 mM sucrose, 5 mM Trizma base) and cleaned in another centrifuge wash‐step (7200g, 10 min) before the final pellet was resuspended in 180–200 μL medium B. Samples were kept cool throughout.

For oximetry, 2–4 μL of liver sample was added to each chamber of an oxygraph (O2K High‐Resolution respirometry, Oroboros Instruments, Austria), and the protocol was run with corresponding software (DatLab7.4, Oroboros Instruments, Austria). Initially, the non‐adenylate non‐phosphorylating (LEAK) respiration state was obtained after addition of glutamate (10 mM) and malate (2 mM). Subsequently, adenosine diphosphate (ADP, 1.25 mM or until saturated), cytochrome C (10 μM), and succinate (10 mM) were added to evaluate complex I oxidative phosphorylation (OXPHOS), membrane integrity, and complex I + II OXPHOS respectively. Then, carbonyl cyanide 4‐(trifluoromethoxy)phenylhydrazone (FCCP, 0.5 μM in stepwise titration until saturated) was added to evaluate the maximum electron transport capacity (ETS) with complex I + II‐driven respiration, before obtaining complex II‐driven ETS after titration of rotenone. Results are given as oxygen flux (pmol oxygen/s/mg mass), normalized to protein content (mg) and correcting for the residual oxygen consumption at baseline. Protein content was measured using a BCA‐assay, as described above, diluting the mitochondrial sample 1:25 in medium B and adding 2.5 μL to the 100 μL wells. The measured protein concentration was adjusted according to the volume added in the oxygraph.

2.4. Proteomics Analysis

Proteins from liver samples for mass spectrometry‐based proteomics were prepared as described above. Protein extract was precipitated using acetone followed by dissolution in 0.2 M tetraethylammonium bicarbonate (TEAB). Dissolved proteins were reduced (5 mM dithiothreitol, 50°C for 30 min), alkylated (15 mM iodoacetamide for 30 min at room temperature) and incubated overnight at 37°C after addition of 2 μg trypsin. The resulting peptide samples were labeled with two sets of 10‐plex Tandem Mass Tags (TMT) using the following tags: 127N, 127C, 128N, 128C, 129N, 129C, 130N, and 130C. A pool of all samples was labeled with mass tag 126 and served as a reference channel.

The labeled peptides sets were high‐pH fractionated followed by reversed phase nano‐liquid chromatography tandem mass spectrometry (RP‐nanoLC‐MS/MS) analysis of individual fractions as follows: the resulting TMTpro set was fractionated by high‐pH reversed‐phase fractionation with a linear gradient of 25 min from 10% solvent B (20 mM ammonium formate in 80% acetonitrile (ACN), pH 9.3) to 55% solvent B at a 6 μL/min flowrate on a Dionex Ultimate 3000 RSLCnano system inline coupled to a Dionex 3000 Ultimate UV detector (210 nm) and a Dionex Ultimate 3000 autosampler configured as a fraction collector (Thermo Scientific, Bremen, Germany), yielding seven fractions to be analyzed with nano‐LC‐MS‐MS.

Nano‐LC‐MS‐MS analysis of the fractionated samples was conducted using an Orbitrap Eclipse mass spectrometer (Thermo Fisher Scientific, San Jose, CA, USA) equipped with a FAIMS interface. Prior to analysis, the peptide fractions were separated using a nano‐HPLC interface (Dionex UltiMate 3000 nano‐HPLC, Thermo Scientific, Bremen, Germany). Briefly, peptide samples were trapped by a custom‐made fused capillary pre‐column (2 cm length, 360 μm OD, 75 μm ID packed with ReproSil Pur C18 5 μm resin [Dr. Maish, GmbH, Ammerbuch, Germany]) with a flow of 4 μL/min for 8 min followed by separation on a custom‐made fused capillary column (25 cm length, 360 μm OD, 75 μm ID, packed with ReporSil Pur C13 1.9 μm resin [Dr. Maisch, Ammerbuch‐Entringen, Germany]) using a linear gradient ranging from 88% to 86% solution A (0.1% formic acid, Fluka, Seetze, Germany) to 27%–32% B (80% acetonitrile [J.T. Baker, Gliwice, Poland] in 0.1% formic acid) over 119 min.

Mass spectra were acquired in positive ion mode by switching between CVs of −50 and −70 V, applying an automatic data‐dependent switch between an Orbitrap survey MS scan in the mass range of 400–1200 m/z, followed by peptide fragmentation applying normalized collisional energy of 40% in a 2‐s cycle. MS1 spectra were acquired at a resolution of 60 000 at 200 m/z by applying a dynamic exclusion of previously selected ions for 60 s. MS2 spectra were acquired at 50000 resolution at 200 m/z with a normalized AGC target of 200% using a 0.4 m/z isolation window. The maximum injection time was 86 ms. The MS1 AGC target was 400 000 ions, and the MS2 AGC target was 100 000 ions.

All raw data files were processed using the Proteome Discoverer software (v. 2.4.0.305) and searched with Sequest HT search algorithm. Sequest HT search parameters were set to default except for MS accuracy of 8 ppm, MSMS accuracy of 0.05 Da for HCD data, with two missed cleavages allowed. Fixed modifications were set to carbamidomethylation at cysteine residues, TMT 6‐plex N‐terminal and TMT 6‐plex on lysine residues. Variable modifications were set to methionine oxidation, deamidation of asparagine and glutamine and N‐terminal acetylation. Sequest HT searches were performed against the Uniprot mouse database (25252 entries, downloaded on 30th September 2019). Proteomics data were compared between genotypes with unpaired t‐test and used for Ingenuity Pathway Analysis (IPA) (Qiagen, Redwood City, CA). The differential protein expression was defined as p < 0.05 (Dataset S1).

2.5. Glucose and Insulin Tolerance Tests

Glucose tolerance was the study's primary outcome. Prior to glucose tolerance test, mice were fasted for 4 h from 6.30 AM. Glucose (2 g/kg body weight) was administered intraperitoneally in 3‐month‐old and 8‐month‐old α₂ ^+/G301R mice and WT littermates. In awake mice, glucose was assessed from mixed venous and arterial blood obtained from the tail and was measured immediately before and at 15, 30, 60, and 120 min. after the glucose infusion using Accu‐Chek glucometer (Aviva, London, UK). Mixed venous and arterial blood samples were obtained from the tail prior glucose injection and at 15 and 30 min after for determination of plasma insulin concentration using ultra‐sensitive insulin ELISA kit (Crystal Chemistry Inc., Downers Grove, IL, USA). In vehicle control experiments, 200 μL isotone saline (0.9% NaCl) was injected intraperitoneally.

In another series of experiments 0.5 IE pr. gram bodyweight insulin (Humulin 100 IE/mL, Eli Lilly, Indianapolis, IN, USA) was administered by intraperitoneal injection in 8‐month‐old α₂ ^+/G301R mice and WT littermates that were fasted for 4 h prior to the experiment. Blood glucose was measured as described for the glucose tolerance test protocol.

Insulin sensitivity was also assessed through baseline glucose and insulin concentrations using the homeostatic model assessment ($\text{HOMA}-\mathrm{IR}=\frac{\text{insuli}{\mathrm{n}}_{{\mathrm{t}}_0}\times 28.7\times \text{glucos}{\mathrm{e}}_{{\mathrm{t}}_0}}{22.5}$) and the quantitative insulin sensitivity check index ($\text{QUICKI}=\frac{1}{\left(\mathrm{Log}\left(28.7\times \text{insuli}{\mathrm{n}}_{{\mathrm{t}}_0}\right)+\mathrm{Log}\left(\text{glucos}{{\mathrm{e}}_{\mathrm{t}}}_0\times 18\right)\right)}$).

2.6. MicroPET‐MRI‐Scan

We used functional PET‐MRI 1T (Mediso Medical imaging Systems, Budapest, Hungary) to assess glucose distribution in key tissues in vivo. Prior PET‐MRI scan, 8‐month‐old mice were fasted for 4 h from 6.30 AM. Glucose (2 g/kg body weight) was administered intraperitoneally. After 30 min, a bolus of [¹⁸F]‐2‐fluoro‐2‐deoxyglucose (FDG) (∼15 MBq/animal) was injected via the tail vein before PET‐scan. During the experiment, mice were anesthetized with 2% isoflurane. Dynamic PET‐scanning was performed for 60 min after the ¹⁸FDG injection followed by 15 min MRI‐scan.

To reconstruct the static PET image, a three‐dimensional ordered subset expectation algorithm was used (Tera‐Tomo 3D; Mediso Medical Imaging Systems). Four iterations and six subsets with a voxel size of 0.4 × 0.4 × 0.4mm³ were obtained. Data were corrected for dead time, decay and randoms using a delayed coincidence window without corrections for attenuation and scatter.

Volume of interest was defined manually, using PMOD 3.6 software (PMOD Technologies LLC, Zürich, Switzerland) to evaluate the tracer activity in skeletal muscle (m. quadriceps region), liver, and cardiac tissue.

2.7. HbA1c Assessment in Blood

Glycated hemoglobin (HbA1c) was measured in mixed whole blood from the tail. HbA1c was measured using the Tina‐quant II immune turbidimetric method on the CobasC111 system (Roche, Basel, Switzerland) as described previously [19]. HbA1c was assessed as a fraction of total hemoglobin.

2.8. Body Composition

Body composition of non‐fasted awake mice was assessed by MR (EchoMRI Houston, TX, USA). Lean mass and adipose tissue mass were determined and their relative composition was calculated.

2.9. Data Analyses

GraphPad Prism software (v.9, GraphPad Software, San Diego, CA, USA) was used for graphing and statistical analysis. All data are summarized as the mean value ± standard error of the mean of the sample group. Where samples were run one more than once, and average was calculated and considered as a single experiment. In glucose tolerance experiments, glucose concentrations were compared by two‐way ANOVA analysis. All other comparisons were done with non‐paired Student's t‐test. In Western blot analyses, averaged data from WT mice were set to 100%. A probability (p) level of < 0.05 was considered significant.

3. Results

3.1. Eight‐Month‐Old α₂ ^+/G301R Mice Are Glucose Intolerant but Not Insulin Resistant

During glucose tolerance test, 3‐month‐old female mice showed blood glucose responses that were not different between genotypes (Figure 1A). This was associated with a similar insulin response 15 min after the glucose injection (Figure 1B). The 3‐month‐old male mice also showed similar blood glucose levels upon glucose tolerance test (Figure S1A), but the male α₂ ^+/G301R mice showed lower insulin response 15 min after glucose challenge compared to that observed in matched WT mice (Figure S1B). In contrast, the 8‐month‐old female and male α₂ ^+/G301R mice were glucose intolerant in comparison with matching WT mice (Figure 1C, Figure S1C). The insulin response was similar between genotypes of 8‐month‐old mice of both sexes (Figure 1D, Figure S1D). A time‐control experiment with injection of saline (0.9% NaCl) showed no significant changes in blood glucose in both genotypes (Figure S2).

FIGURE 1.

The 8‐month‐old, but not 3‐month‐old, α₂ ^+/G301R female mice showed glucose intolerance but similar insulin sensitivity. Changes in glucose during glucose tolerance test (GTT) were not statistically different between genotypes in 3‐month‐old females (A, n = 7), and the insulin response was similar 15 min after GTT initiation (B). The 8‐month‐old female α₂ ^+/G301R (C, n = 8) showed glucose intolerance during GTT in comparison with matched WT mice. Insulin responses were similar between genotypes for 8‐month‐old female mice (D). Both 3‐ and 8‐month‐old genotypes had similar insulin sensitivity assessed by the homeostatic model assessment (HOMA‐IR, E) and the quantitative insulin sensitivity check (QUICK) index (F). Intraperitoneal injection of insulin (0.5 IE pr. gram bodyweight) in 8‐month‐old female mice revealed similar sensitivities to insulin between genotypes (G, n = 8). *p < 0.05.

There was no difference between the genotypes of both sexes in each age group when comparing insulin sensitivity through baseline levels of insulin and glucose by the homeostatic model assessment (HOMA‐IR, Figure 1E, Figure S3A) and the quantitative insulin sensitivity check index (QUICK‐index, Figure 1F, Figure S3B). The insulin sensitivity of peripheral tissues was further assessed by insulin tolerance test. Intraperitoneal insulin injection was associated with a similar decline in blood glucose between genotypes, both in 8‐month‐old female (Figure 1G) and male mice (Figure S3C).

Glucose was injected before FDG‐PET scanning in 8‐month‐old female mice to assess the distribution of glucose disposal in peripheral tissues (Figure 2A–C). The α₂ ^+/G301R mice showed a reduced FDG load in the liver compared to WT mice (Figure 2D). There was no difference between genotypes in skeletal muscle (Figure 2E) and cardiac tissue ¹⁸FDG load (Figure 2F).

FIGURE 2.

The FDG‐PET scan revealed impaired hepatic glucose uptake upon glucose injection. Representative images from an FDG‐PET/MRI scan of 8‐month‐old female α₂ ^+/G301R mouse at 7 s (A), 7 min (B), and 1 h (C) after intravenous FDG injection, as indicated. Glucose tolerance test was started 30 min prior the beginning of the scan. Red and white arrows in (A–C) indicate the liver and the soleus muscles, respectively. The 8‐month‐old α₂ ^+/G301R mice showed impaired FDG load in the liver (D) whereas the loads in skeletal muscle (E) and cardiac tissue (F) were similar between genotypes. ¹⁸FDG load was quantified as an area under the curve (AUC) of the maximum standardized uptake value (SUV_max). **p < 0.01, n = 4.

3.2. Long‐Term Blood Glucose and Body Composition Changes With Age in the Male α₂ ^+/G301R Mice

Long‐term blood glucose was assessed by the concentration of hemoglobin (Hb) chemically linked to sugar, HbA1c, relative to the total Hb. The 3‐month‐old male, but not female, α₂ ^+/G301R mice showed a decreased level of HbA1c compared to WT controls (Figure 3A, Figure S4A). In 8‐month‐old mice, there was no statistical difference between the genotypes of both sexes. The body weight was similar in both genotypes for both sexes in 3‐, 6‐, and 8‐month‐old mice (Figure 3B, Figure S4B). Relative body composition was measured by MR scanning. The 3‐month‐old male α₂ ^+/G301R mice showed an increased amount of lean mass and had less adipose tissue than matched WT mice (Figure S4C,D). In contrast, 8‐month‐old α₂ ^+/G301R male mice showed reduced lean mass and increased fraction of adipose tissue. These age‐dependent differences in body composition were not observed in female mice (Figure 3C,D).

FIGURE 3.

Similar HbA1c and body composition between genotypes in female mice. Whole blood concentration of glycated hemoglobin (A; HbA1c; n = 9–10) was assessed as a fraction of total hemoglobin (Hb). There was no difference in body weight between genotypes at 3, 6, and 8 months (B). The body composition of lean tissue (C) and adipose tissue (D) were assessed using magnetic resonance scanning and was similar between genotypes in 3‐month‐old (n = 10–12) and 8‐month‐old female mice (n = 4–9). For comparison of male mice, please refer to Figure S4.

3.3. Impaired Mitochondrial Coupling Efficiency in Liver Tissue of 8‐Month‐Old Female α₂ ^+/G301R Mice

Mitochondrial respiratory capacity was evaluated in isolated mitochondria from 8‐month‐old male and female mice of both genotypes. In females, the α₂ ^+/G301R mice did not have alterations in absolute oxygen consumption values (Figure S5A). Residual oxygen consumption was similar between female genotypes (Figure 4A). Basal LEAK respiration tended to be higher in α₂ ^+/G301R females but did not reach statistical significance (p = 0.097; Figure 4B). Absolute oxygen consumption values were not different in OXPHOS, neither through complex I, nor the combined complex I + II‐stimulated respiration (Figure 4C,D). Consequently, respiratory control ratios for both complex I and complex I + II were reduced in α₂ ^+/G301R females (Figure 4E,F), indicating reduced coupling of electron transport to ATP synthesis, likely due to the elevation in LEAK respiration. Maximal electron transport capacity (ETS) was unchanged (Figure 4G), but inhibition of complex I produced a disproportionately larger decline in residual ETS respiration in α₂ ^+/G301R females compared with WT (RatioRot; Figure 4H). Thus, there is no alteration in the capacity for electron transport, but the liver mitochondria from female α₂ ^+/G301R mice consume more oxygen for electron transport through complex I, than the WT. Hence, female α₂ ^+/G301R liver mitochondria exhibit impaired coupling efficiency characterized by reduced phosphorylation efficiency and altered complex dependency in electron transport. The overall consumption of oxygen in liver mitochondria from the α₂ ^+/G301R males was reduced throughout the protocol compared with WT (Figure S5B). However, the complex‐specific changes in mitochondrial respiration, which were observed in female mice, were not seen in 8‐month‐old α₂ ^+/G301R male mice (Figure S6A–H).

FIGURE 4.

Impaired mitochondrial efficiency in liver tissue of 8‐month‐old female α₂ ^+/G301R mice. All oxygen fluxes were corrected for residual oxygen consumption (A). Difference in basal LEAK respiration supported by glutamate + malate measured before ADP addition did not reach statistical significance between genotypes (B). Adenosine diphosphate (ADP)‐stimulated OXPHOS supported by complex I (C) and OXPHOS supported by complexes I + II after succinate addition (D) were similar between genotypes. Respiratory control ratio (RCR) in complex I (E) and complexes I + II (F) was reduced in α₂ ^+/G301R females compared with wild type (WT). Maximal electron transport (ETS) capacity was similar between genotypes (n = 3–5, G), but the relative residual ETS after complex I inhibition was lower in α₂ ^+/G301R mice compared with WT mice (n = 3–5; RatioRot, H). Data are shown as mean ± SEM with individual data points overlaid. Genotypes were compared using t‐test; *p < 0.05. See Methods for respirometry protocol and definitions of flux control ratios and Figure S2 for hepatic respirometry data from 8‐month‐old male mice.

3.4. Altered Expression of Oxidative Phosphorylation and Glycogen‐Regulatory Proteins in Liver Tissue of α₂ ^+/G301R Mice

Proteomics analysis of liver lysates from 8‐month‐old α₂ ^+/G301R and WT mice of both sexes identified 5462 proteins (Dataset S1). There were some significant sex differences in the expression profiles even within the groups with same genotype; 952 proteins were found to be differentially expressed when females and males of WT mice were compared (Dataset S1). When different sexes of α₂ ^+/G301R mice were compared, 891 differentially expressed proteins were detected (Dataset S1). Because of this sex difference, the genotype effect was first assessed separately for each sex.

Out of all mapped proteins, 172 proteins were differentially expressed between female α₂ ^+/G301R and WT liver lysates, with 132 proteins being downregulated and 40 upregulated in α₂ ^+/G301R mice (Dataset S1, Figure S7A). Ingenuity pathway analysis suggested several canonical pathways where changes in mitochondrial translation and posttranslational protein modification were highlighted (Dataset S2). Comparison of α₂ ^+/G301R and WT male liver lysates identified 361 differentially expressed proteins, with 156 downregulated and 205 upregulated in α₂ ^+/G301R mice (Dataset S1, Figure S7B). Ingenuity pathway analysis of male data also suggested changes toward mitochondrial dysfunction associated with abnormal protein handling (Dataset S2).

Among key proteins involved in hepatic glucose handling and glycogen metabolism, the expression of GLUT‐2 (SLC2A2) was similar between genotypes in both sexes (Dataset S1). Glycogen synthase kinase‐3α (GSK3a) was upregulated in female α₂ ^+/G301R livers. Strikingly, GSK3a was not detected in any of the liver tissue samples from males of both genotypes, consistent with lower GSK3a expression previously reported in male mice [20]. Glycogen synthase kinase‐3β (GSK3b) was similar between genotypes (Dataset S1). The glucose‐6‐phosphatase (G6PC) expression was unchanged between genotypes in females, but it was found to be reduced in male α₂ ^+/G301R livers in comparison with WT. Western blot analysis confirmed increased hepatic GSK3 expression in 8‐month‐old α₂ ^+/G301R mice, reaching statistical significance in males, while females exhibited a similar trend (p = 0.054; Figure S8A–D). Western blot did not identify changes in α₂ ^+/G301R livers for G6PC expression in both sexes (Figure S8E,F).

The downstream pathway analysis was made with pooled female and male datasets, to improve power of the analysis. In this pooled dataset, 248 proteins were differentially expressed; thereof, 80 proteins were downregulated and 168 proteins upregulated in α₂ ^+/G301R livers in comparison with WT controls (Dataset S1, Figure 5). The analysis identified reduced expression of key proteins in electron transport chain complexes suggesting impaired oxidative phosphorylation in livers of α₂ ^+/G301R mice (Dataset S2, Figure S9). Impaired function of complex I in α₂ ^+/G301R mice was based on reduced expression of NADH dehydrogenase where most subunits showed either a significant reduction in expression or a tendency toward reduced expression. The NADH:ubiquinone oxidoreductase subunits A8 and B5 in complex I was reduced in α₂ ^+/G301R mice. Within complex II, the reduced expression of succinate dehydrogenase flavoprotein (SDHA) in α₂ ^+/G301R livers suggested impaired oxidation of succinate to fumarate resulting in impaired passing of electrons to ubiquinone. Proteins involved in assembly and stabilization of electron transport chain complexes including NDUFAF family members, supercomplex assembly factor, and cytochrome c oxidase subunit 7A‐related protein were not differentially expressed. Seven of 14 ATP synthase (complex V) subunits were reduced in α₂ ^+/G301R mice, and all remaining subunits, which were detected, trended toward lower expression. Several proteins implicated in mitochondrial membrane proton leak were assessed in the proteomics dataset. Tafazzin and uncoupling protein 2 (UCP2), both important for mitochondrial membrane proton leak, were not detected (Dataset S1). Another key mitochondrial membrane proton leak‐associated protein, adenine nucleotide translocase 1 (ANT1), showed a trend toward increased abundance (p = 0.06) in males and was not detected in females. ANT2 was not detected.

FIGURE 5.

Proteomics analysis of liver tissue suggests that the α₂ ^+/G301R mice have mitochondrial dysfunction. Vulcano plot for proteomic comparing α₂ ^+/G301R and WT mice of both sexes (n = 8). Some differentially expressed proteins are indicated on the plot showing that many of these proteins are important for mitochondrial function. See also Dataset S1, Figure S5. Blue, red and black circles indicate downregulated, upregulated proteins and proteins that were not significantly different between the groups. t‐test, as appropriate.

3.5. Reduced Na,K‐ATPase α₂ Isoform Expression in the Liver and Skeletal Muscles From α₂ ^+/G301R Mice

Western blot analysis of liver lysates demonstrated the expression of both the α₁ and α₂ isoforms of the Na,K‐ATPase, with the α₁ isoform being visually predominant (Figure S10). No differences in Na,K‐ATPase α₁ isoform expression were detected in liver tissue between genotypes of both sexes (Figure S10A,B). In contrast, α₂ isoform expression was significantly reduced in the livers of both male and female α₂ ^+/G301R mice compared with their respective wild‐type controls (Figure S10C,D).

We have questioned whether these changes in the expression profile correlate with the Na,K‐ATPase related intracellular signaling pathways alterations. However, no differences in phosphorylation of Src (Figure S11A), phospholipase C (PLCγ) (Figure S11B) and mitogen‐activated protein kinase (MAPK) (Figure S11C) were detected.

The expression of both α₁ isoform (Figure S12A,B) and α₂ isoform (Figure S12C,D) of the Na,K‐ATPase was reduced in the soleus muscle from 8‐month‐old α₂ ^+/G301R female mice in comparison with WT controls. The insulin receptor β (Figure S12E,F) and GLUT‐4 (Figure S12G) expressions were similar in both genotypes.

4. Discussion

In this study, we demonstrate for the first time that a clinically relevant mutation in the Na,K‐ATPase influences whole‐body glucose homeostasis. We found that α₂ ^+/G301R mice were glucose intolerant, a phenotype that was only observed in 8‐month‐old mice, whereas 3‐month‐old mice were similar between genotypes. The disturbance in glucose homeostasis was not associated with changed insulin secretion or insulin resistance. Instead, impaired glucose tolerance was associated with reduced hepatic glucose uptake in 8‐month‐old α₂ ^+/G301R mice, a phenotype associated with inefficient hepatic mitochondrial function.

4.1. Impaired Hepatic Glycogenesis May Underlie Glucose Intolerance in α₂ ^+/G301R Mice

In mice, the liver and skeletal muscles are the major sites for post‐prandial glucose disposal [21]. A major part of this glucose is converted through glycogenesis to glycogen, a process that requires two ATP molecules per cycle [22, 23]. Another major consumer of cellular ATP is the Na,K‐ATPase. Importantly, the expression of Na,K‐ATPase α₂ isoform was consistently reduced in α₂ ^+/G301R mice. Surprisingly, the expression profile of the α₁ isoform differed between the liver and skeletal muscle. In skeletal muscle, α₂ ^+/G301R mice exhibited reduced expression of the Na,K‐ATPase α₁ isoform, whereas α₁ expression remained unchanged in the liver. Although the proteomics analysis suggested a reduction of the α₁ isoform in the livers of male α₂ ^+/G301R mice only, this finding could not be confirmed by Western blot analysis. We have previously reported similar tissue‐specific differences in the consequences of the G301R mutation in Atp1a2. Although the mutation consistently reduces α₂ isoform protein expression, it is also associated with reduced α₁ isoform abundance in vascular endothelial and smooth muscle cells [13], whereas α₁ expression is increased in cardiac muscles [9]. The mechanisms underlying these tissue‐specific effects of the Atp1a2 mutation on α₁ isoform expression remain unclear but may depend on the cell type‐specific composition and relative abundance of co‐expressed Na,K‐ATPase isoforms. In skeletal muscle, the α₂ isoform is the predominant Na,K‐ATPase isoform [24], and its reduction may compromise the overall membrane ion transport capacity, leading to secondary downregulation of the α₁ isoform. In contrast, hepatocytes predominantly express the α₁ isoform [25], whereas the relatively weak α₂ signal detected by Western blot most likely originates from non‐parenchymal cells, including Kupffer cells and cells of the hepatic vasculature, which also express the α₂ isoform [5, 26]. Consequently, reduced α₂ expression in these neighboring non‐parenchymal cell populations would not be expected to directly influence α₁ isoform expression in hepatocytes.

In the skeletal muscle tissue, glucose transport is facilitated by translocation of GLUT‐4 to the plasma membrane in an insulin‐dependent manner [27]. The α₂ ^+/G301R and WT mice displayed similar expression of the insulin receptor and GLUT‐4 in skeletal muscle. Accordingly, similar insulin sensitivity was observed between genotypes, and the FDG‐PET results suggested similar glucose uptake in skeletal muscle tissue upon glucose tolerance test in both genotypes. In contrast, reduced hepatic FDG trapping was observed in α₂ ^+/G301R female mice, suggesting the cause of their glucose intolerance. In the liver, glucose is primarily transported into the hepatocytes through the insulin‐independent GLUT‐2 [28] although insulin signaling is important for glucose metabolism in the liver, for example, glycogenesis [29]. Importantly, glycogen metabolism in the liver is modulated by the Na,K‐ATPase that promotes glycogen synthesis via signal transduction pathways [30, 31] suggesting that reduced Na,K‐ATPase abundance may impair glucose disposal.

Glycogenesis is ATP‐demanding [22, 23], and post‐prandial glycogen synthesis therefore depends on efficient mitochondrial ATP generation [28, 32] suggesting an alternative pathway behind the observed glucose intolerance in α₂ ^+/G301R mice. Respirometry of liver mitochondria from 8‐month‐old female α₂ ^+/G301R mice revealed reduced respiratory control ratios driven by increased LEAK respiration. This mitochondrial phenotype was associated with reduced expression of components in complexes I and II, and most prominently, with decreased expression of ATP synthase subunits in complex V. The observed increase in expression of the negative regulator of glycogen synthesis, GSK3, in liver tissue of α₂ ^+/G301R may contribute to suppressing glycogen formation. In fact, the antagonized action of the Na,K‐ATPase for GSK3 signaling has been reported [30] and thus, the reduced expression of Na,K‐ATPase may promote GSK3 signaling. The increased GSK3 expression was consistently observed across Western blot and proteomics analyses, but its role as an adaptive response versus a causal factor in impaired glycogen synthesis remains uncertain. Although ADP‐stimulated OXPHOS and maximal ETS capacity were largely maintained, complex I inhibition caused a disproportionally large loss of residual ETS capacity in female α₂ ^+/G301R liver mitochondria. This mechanism may contribute to reduced complex II activity because of the observed reduction of SDHA, which would restrict succinate oxidation through complex II and consequently impair oxidative phosphorylation efficiency. Overall, mitochondrial irregularities could cause an energy deficit and subsequent dysfunction of glycogenesis and overall energy storage.

Furthermore, proteomics suggests reduction in G6PC in male α₂ ^+/G301R mice. G6PC is part of glucose‐6‐phosphatase catalyzing the terminal step of gluconeogenesis. Its deficiency is reported to lead to defect in glycogen storage, fasting hypoglycemia and other metabolic disturbances [33]. This may contribute to reduced HbA1c and reduced body lipids in young male α₂ ^+/G301R mice. These changes in G6PC have to be further studied as it was not validated with Western blot.

Prior work has established a functional association between Na,K‐ATPase and mitochondrial function. Hence, knockout of the Na,K‐ATPase α₁ isoform in skeletal muscle leads to mitochondrial uncoupling associated with reduced glycogen content and decreased exercise endurance [8]. Although the molecular mechanisms connecting altered Na,K‐ATPase function to mitochondrial inefficiency remain incompletely understood, accumulating evidence suggests that non‐canonical signaling functions of the Na,K‐ATPase are involved. In particular, loss of Na,K‐ATPase‐mediated inhibition of Src kinase leads to enhanced Src activity [34, 35], which has been linked to alterations in mitochondrial respiration and oxidative phosphorylation [8, 9, 36, 37]. Tissue‐dependent alterations in Src kinase expression have previously been reported in α₂ ^+/G301R mice, with reduced Src expression observed in cerebral arteries [13] and increased expression in cardiac tissue [9]. In the present study, proteomics and Western blot analyses showed that Src kinase expression in the liver was similar between α₂ ^+/G301R and WT mice. We have also examined additional putative Na,K‐ATPase‐dependent signaling pathways [38, 39], including PLCγ and MAPK. None of these signaling pathways was altered in the liver from α₂ ^+/G301R mice. These findings are consistent with the predominant expression of the Na,K‐ATPase α₁ isoform in hepatocytes [25], making these cells less directly susceptible to changes in α₂ isoform expression. Instead, reduced α₂ expression is more likely to affect the function of non‐parenchymal cell types, particularly macrophages and cells of the hepatic vasculature, where the α₂ isoform is also expressed [13, 40]. Whether Na,K‐ATPase‐dependent intracellular signaling pathways are altered in non‐parenchymal cells of α₂ ^+/G301R mice remains unclear. It is possible that such changes were not detected because the signals were diluted by the predominance of hepatocytes in the whole‐liver lysates. This question will require future spatially resolved expression and signaling analyses.

It remains unclear whether the observed changes in hepatic mitochondrial function result from reduced expression of the small residual pool of the Na,K‐ATPase α2 isoform in hepatocytes or are secondary to intercellular signaling from non‐parenchymal cells, such as Kupffer cells, impaired hepatic perfusion, or other cell type‐specific interactions. Indeed, several vascular beds in α₂ ^+/G301R mice have been reported to exhibit increased vascular resistance and reduced tissue perfusion [13], which could lead to chronic reductions in oxygen and substrate delivery, increased reactive oxygen species production, and low‐grade inflammation in the liver. A contribution from resident macrophages also cannot be excluded. Although previous studies suggest that the Na,K‐ATPase α₂ isoform is not directly involved in macrophage cytokine production [41], other macrophage‐dependent mechanisms may still contribute to the observed phenotype. Further studies are therefore required to determine how reduced Na,K‐ATPase α₂ isoform expression is mechanistically linked to altered mitochondrial function in hepatocytes.

The glucose intolerance phenotype in the current study was only observed in 8‐month‐old α₂ ^+/G301R mice, whereas 3‐month‐old α₂ ^+/G301R mice showed glucose challenge responses similar to those observed in WT mice. An age‐dependent phenotype in α₂ ^+/G301R mice has also been reported in cardiac tissue where mitochondrial uncoupling is associated with impaired cardiac function in 8‐month‐old, but not 3‐month‐old, α₂ ^+/G301R mice [9]. Together, these findings suggest that chronic alterations in Na,K‐ATPase function progressively compromise mitochondrial efficiency rather than inducing an acute bioenergetic defect, leading to an age‐dependent mitochondrial phenotype and tissue‐specific functional decline.

4.2. Age and Sex‐Dependent Change in Body Composition in α₂ ^+/G301R Mice

The body composition changed significantly with age in the male α₂ ^+/G301R mice. Hence, reduced HbA1c in 3‐month‐old male α₂ ^+/G301R mice was associated with reduced adipose tissue relative to lean mass compared to WT controls. This may potentially be due to increased glucose absorption by the liver causing reduced pancreatic stimulation of insulin production. In contrast, adipose tissue was increased relative to lean mass in 8‐months‐old male α₂ ^+/G301R mice compared to matched WT controls. Notably, there was no difference in body composition between genotypes neither in 3‐month‐old nor 8‐month‐old female mice. Also, female α₂ ^+/G301R mice did not exhibit reduced level of HbA1c as was observed in 3‐month‐old males. The differences between sexes remain elusive and require further investigation. In the present study, we did not observe any difference in body weight between genotypes of both sexes at different ages. A previous study compared body weight of C57BL/6 mice carrying heterozygous knockout of the Na,K‐ATPase α₂ isoform, i.e., α₂ ^+/− mice [42]. In line with the findings from the present study, male α₂ ^+/− mice show similar body weight compared to WT mice measured at different ages; body composition of male α₂ ^+/− mice was not assessed. However, elderly female α₂ ^+/− mice show higher body weight compared to WT mice; a difference between genotypes that is not observed when mice are exposed to high‐fat diet. The difference in observations of body weights of female α₂ ^+/− mice in the previous study and female α₂ ^+/G301R mice in the present study might be caused by different genotypes or different composition of standard diets.

To our knowledge, no study to date has specifically investigated whether FHM2 is associated with systemic glucose intolerance or any other metabolic phenotypes. Although a caution should be given in translating patient phenocopies to rodent studies, patients with migraine with aura show an increased prevalence of metabolic syndrome [3], and migraine is increasingly regarded as a disorder involving impaired cerebral and systemic energy metabolism [43]. In this context, the present study provides new experimental evidence that an FHM2‐associated Na,K‐ATPase mutation can disrupt peripheral metabolic processes, thereby linking altered energy handling outside the brain to the metabolic dimension of migraine. This may further implement the Na,K‐ATPase in some drug‐induced insulin resistance, as it is seen for sevoflurane [44] having a complex modulatory action on the Na,K‐ATPase [45].

5. Conclusion

Hemiplegic migraine‐associated mutation of the Na,K‐ATPase was associated with reduced expression of key mitochondrial proteins in the liver suggesting impaired capacity of ATP‐demanding pathways, e.g., gluconeogenesis and glycogenesis. This outcome was associated with reduced long‐term glucose levels and altered lean and fat mass in 3‐month‐old α₂ ^+/G301R male mice, as well as glucose intolerance in 8‐month‐old α₂ ^+/G301R mice of both sexes. Mitochondrial oxygen consumption and function was altered in 8‐month‐old α₂ ^+/G301R mice of both sexes. This study calls for future investigation translating these pre‐clinical findings into familial hemiplegic migraine type 2 patients carrying the Na,K‐ATPase mutation.

6. Limitations

The study has some limitations that must be considered. First, we studied the familial hemiplegic migraine‐associated G301R missense mutation in the Atp1a2 gene encoding the Na,K‐ATPase α₂ isoform [1, 46]. There are several other Atp1a2 mutations that are associated with the same migraine phenotype [47], but it remains to be studied whether these other mutations also lead to the metabolic phenotype characterized in the present study. Furthermore, this is a global constitutive mutation that affects multiple organ systems, either directly through alterations in Na,K‐ATPase activity or indirectly via compensatory mechanisms. Notably, the α₁ isoform is the predominantly expressed Na,K‐ATPase isoform in the liver [48], suggesting that liver dysfunction associated with the Atp1a2 mutation might result from indirect changes in the α₁ isoform expression. Second, there are inherited limitations in the ability of mice to represent metabolic disturbances in humans [21]. It therefore remains to be investigated whether familial hemiplegic migraine type 2 patients carrying the same mutations as α₂ ^+/G301R mice exhibit the same metabolic phenotype. Third, although most of the experiments were done in the awake mice, PET‐MRI experiments required mice to be anesthetized, which may alter glucose metabolism [49].

Author Contributions

Hans Christian Beck: methodology, investigation, writing – review and editing, formal analysis. Christian Staehr: conceptualization, investigation, writing – original draft, methodology, writing – review and editing, formal analysis, project administration, data curation, supervision, visualization. Niels Jessen: writing – review and editing, methodology. Tina Myhre Pedersen: investigation, methodology, writing – review and editing, formal analysis. Tue Duy Nguyen: investigation, writing – original draft, methodology, writing – review and editing, formal analysis, visualization. Alexander V. Chibalin: conceptualization, writing – review and editing. Mads Kjolby: writing – review and editing, methodology, formal analysis. Per Bendix Jeppesen: investigation, methodology, writing – review and editing, formal analysis. Elena V. Bouzinova: investigation, writing – review and editing. Mikkel Vendelbo: visualization, validation, methodology, supervision, formal analysis, writing – review and editing. Vladimir V. Matchkov: conceptualization, investigation, funding acquisition, writing – original draft, methodology, writing – review and editing, project administration, supervision. Rajkumar Rajanathan: investigation, writing – review and editing.

Funding

Vladimir V. Matchkov received support from the Lundbeck Foundation (R344‐2020‐952 and R412‐2022‐449), the Novo Nordisk Foundation (NNF18OC0052021 and NNF19OC0056371) and the Independent Research Fund Denmark (3101‐00103B).

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Data S1: apha70291‐sup‐0001‐DataS1.xlsx.xlsx.

Data S2: apha70291‐sup‐0002‐DataS2.xlsx.xlsx.

Figures S1‐S12: apha70291‐sup‐0003‐FiguresS1‐S12.docx.docx.

Data Availability Statement

Proteomics data are uploaded separately in Datasets S1 and S2. Other data will be shared by the corresponding authors upon reasonable request.