Abnormal hepatic energy homeostasis in type 2 diabetes


  • Julia Szendroedi,

    1. Institute for Clinical Diabetology, German Diabetes Center, Leibniz Center for Diabetes Research, Heinrich Heine University Düsseldorf, Germany
    2. Department of Metabolic Diseases, Heinrich Heine University Düsseldorf, Germany
    3. Karl-Landsteiner Institute for Endocrinology and Metabolism, Medical University of Vienna, Vienna, Austria
    Search for more papers by this author
    • These authors contributed equally to this article.

  • Marek Chmelik,

    1. Karl-Landsteiner Institute for Endocrinology and Metabolism, Medical University of Vienna, Vienna, Austria
    2. MR Center of Excellence, Medical University of Vienna, Vienna, Austria
    Search for more papers by this author
    • These authors contributed equally to this article.

  • Albrecht Ingo Schmid,

    1. Karl-Landsteiner Institute for Endocrinology and Metabolism, Medical University of Vienna, Vienna, Austria
    2. MR Center of Excellence, Medical University of Vienna, Vienna, Austria
    Search for more papers by this author
  • Peter Nowotny,

    1. Institute for Clinical Diabetology, German Diabetes Center, Leibniz Center for Diabetes Research, Heinrich Heine University Düsseldorf, Germany
    2. Department of Internal Medicine 3, Medical University of Vienna, Vienna, Austria
    Search for more papers by this author
  • Attila Brehm,

    1. First Medical Department, Hanusch Hospital, Vienna, Austria
    Search for more papers by this author
  • Martin Krssak,

    1. Department of Internal Medicine 3, Medical University of Vienna, Vienna, Austria
    2. Department of Radiology, Medical University of Vienna, Vienna, Austria
    Search for more papers by this author
  • Ewald Moser,

    1. MR Center of Excellence, Medical University of Vienna, Vienna, Austria
    2. Center for Biomedical Engineering and Physics, Medical University of Vienna, Vienna, Austria
    Search for more papers by this author
  • Michael Roden

    Corresponding author
    1. Institute for Clinical Diabetology, German Diabetes Center, Leibniz Center for Diabetes Research, Heinrich Heine University Düsseldorf, Germany
    2. Department of Metabolic Diseases, Heinrich Heine University Düsseldorf, Germany
    3. Karl-Landsteiner Institute for Endocrinology and Metabolism, Medical University of Vienna, Vienna, Austria
    4. First Medical Department, Hanusch Hospital, Vienna, Austria
    • Head/Institute for Clinical Diabetology, Director/German Diabetes Center, Chair/Department of Metabolic Diseases, Heinrich Heine University, Auf‘m Hennekamp 65, D-40225 Düsseldorf, Germany
    Search for more papers by this author
    • fax: 0049 211 3382 69 1

  • Potential conflict of interest: Nothing to report.


Increased hepatocellular lipids relate to insulin resistance and are typical for individuals with type 2 diabetes mellitus (T2DM). Steatosis and T2DM have been further associated with impaired muscular adenosine triphosphate (ATP) turnover indicating reduced mitochondrial fitness. Thus, we tested the hypothesis that hepatic energy metabolism could be impaired even in metabolically well-controlled T2DM. We measured hepatic lipid volume fraction (HLVF) and absolute concentrations of γATP, inorganic phosphate (Pi), phosphomonoesters and phosphodiesters using noninvasive 1H/ 31P magnetic resonance spectroscopy in individuals with T2DM (58 ± 6 years, 27 ± 3 kg/m 2), and age-matched and body mass index–matched (mCON; 61 ± 4 years, 26 ± 4 kg/m 2) and young lean humans (yCON; 25 ± 3 years, 22 ± 2 kg/m 2, P < 0.005, P < 0.05 versus T2DM and mCON). Insulin-mediated whole-body glucose disposal (M) and endogenous glucose production (iEGP) were assessed during euglycemic-hyperinsulinemic clamps. Individuals with T2DM had 26% and 23% lower γATP (1.68 ± 0.11; 2.26 ± 0.20; 2.20 ± 0.09 mmol/L; P < 0.05) than mCON and yCON individuals, respectively. Further, they had 28% and 31% lower Pi than did individuals from the mCON and yCON groups (0.96 ± 0.06; 1.33 ± 0.13; 1.41 ± 0.07 mmol/L; P < 0.05). Phosphomonoesters, phosphodiesters, and liver aminotransferases did not differ between groups. HLVF was not different between those from the T2DM and mCON groups, but higher (P = 0.002) than in those from the yCON group. T2DM had 13-fold higher iEGP than mCON (P < 0.05). Even after adjustment for HLVF, hepatic ATP and Pi related negatively to hepatic insulin sensitivity (iEGP) (r =-0.665, P = 0.010, r =-0.680, P = 0.007) but not to whole-body insulin sensitivity. Conclusion: These data suggest that impaired hepatic energy metabolism and insulin resistance could precede the development of steatosis in individuals with T2DM. (HEPATOLOGY 2009.)

Approximately 70% of patients who are obese or who have type 2 diabetes mellitus (T2DM) have nonalcoholic fatty liver (NAFL)1–5 of whom 20% show progression to nonalcoholic steatohepatitis (NASH).6 Nonalcoholic fatty liver disease (NAFLD) refers to hepatocellular lipid contents of >5.5% with varying degrees of inflammation in individuals consuming less than 20 g of alcohol per day.7–10

Liver fat content relates negatively to hepatic insulin sensitivity as assessed from impaired insulin-suppressed endogenous glucose production (iEGP).7, 11 Increased liver fat may precede and even predict the onset of T2DM12, 13 and cardiovascular disease.14 Those with T2DM typically not only feature steatosis and hepatic insulin resistance, but also whole-body insulin resistance, which in turn frequently associates with increased intramyocellular lipid contents (IMCL).15 Skeletal muscle of individuals with T2DM and their nondiabetic first-degree relatives shows reduced mitochondrial activity, which has been related to insulin resistance and may result from impaired function and/or density of mitochondria.16 The role of ectopic lipid deposition in skeletal muscle as cause or consequence of reduced oxidative capacity is yet unclear.17, 18

Interestingly, liver biopsies of patients with steatohepatitis also reveal mitochondrial dysmorphologies and reduced activities of respiratory chain enzymes.19, 20 The possible role of mitochondrial function in the development of liver steatosis and insulin resistance could be as follows15: increased free fatty acid availability increases hepatocellular acyl-coenzyme A molecules which impair insulin signaling thereby enhancing EGP and stimulating both storage of triglycerides and lipid oxidation. Increased lipid oxidation and citrate cycle activity will give rise to reactive oxygen species and diminish mitochondrial function. As a result, hepatocellular adenosine triphosphate (ATP) concentrations would decrease. Combined lipotoxicity, oxidative stress, and proinflammatory mediators, also key factors in the development of muscular insulin resistance,1 would further damage hepatic mitochondria leading to mitochondrial dysmorphologies and promoting the transition of steatosis to NASH.2, 11 In this context, hyperlipidemia was found to decrease both insulin sensitivity and mitochondrial function at the level of skeletal muscle.21 However, no data on the association between ectopic fat content, insulin sensitivity, and mitochondrial function are currently available at the level of the liver.

Thus, we hypothesized that humans with hepatic insulin resistance have decreased hepatocellular ATP concentrations and that the reduction of hepatic phosphorus compounds relates to hepatic insulin resistance and to lipid contents. To this end, we studied patients with T2DM, who present with hepatic insulin resistance, and tested the hypotheses that (1) patients with T2DM have decreased hepatic inorganic phosphate (Pi) and γATP and that these phosphorus compounds relate negatively to (2) hepatic lipid contents and (3) insulin sensitivity.

In order to address these hypotheses, we applied a recently developed technique using noninvasive phosphorus magnetic resonance spectroscopy (31P MRS). This method allows measurement of phosphorus metabolites in various regions of the liver using three-dimensional MRS imaging with an external reference standard for the absolute quantification of hepatocellular γATP, Pi, and phosphodiesters and phosphomonoesters (PDE, PME) in humans.22 Due to its low intraindividual variability, this technique makes it possible to trace subtle changes of hepatocellular phosphorus metabolites.


EGP, endogenous glucose production; HCL, hepatocellular lipids; HLVF, hepatic lipid volume fraction; iEGP, insulin-suppressed endogenous glucose production; IMCL, intramyocellular lipid contents; jMRUI, Java-based Magnetic Resonance User Interface; mCON, age-matched and BMI-matched controls; M, insulin-mediated whole-body glucose disposal; MRSI, magnetic resonance spectroscopic imaging; NAFL, nonalcoholic fatty liver; NAFLD, nonalcoholic fatty liver disease; [P], concentrations of phosphorus metabolites; PCr, phosphocreatine; Pi, inorganic phosphate; Pia, γATPa, Pi and γATP concentrations adapted for HLVF; PME, PDE phosphomonoesters and phosphodiesters; Ra, rates of glucose appearance; r, Pearson correlation coefficients; R, Spearman correlation coefficients; T2DM, type 2 diabetes mellitus; yCON, young lean human controls.

Patients and Methods

Human Subjects.

Twenty-seven volunteers consisting of individuals with T2DM and age-matched and body mass index (BMI)-matched (mCON) and young nondiabetic human (yCON) controls were included in this study (Table 1). All participants were recruited by means of public advertisement and underwent complete medical history, examination, and laboratory tests to exclude viral hepatitis and other major diseases. Self-reported alcohol consumption was below 20 g/day. Patients presenting with islet cell antibodies and diabetes-related complications were excluded. The control groups had neither any family history of diabetes nor any medication use. None of the participants were trained or performed intense exercise. The study protocol conformed with the ethical guidelines of the 1975 Declaration of Helsinki as reflected by a priori approval by the appropriate institutional review committee. Informed consent was obtained from each volunteer after explanation of the purpose, nature, and potential complications of the study.

Table 1. Clinical Characteristics and Absolute Concentrations of Hepatic Phosphorous Metabolites (Means ± SD) of Patients with Type 2 Diabetes (T2DM), Matched Controls (mCON), and Young Controls (yCON)
  • Glycosylated hemoglobin A1c (HbA1c), alanine aminotransferase (ALT), inorganic phosphate (Pi), and γATP adapted for HLVF (Pia and γATPa), phosphomonoesters (PME) and phosphodiesters (PDE).

  • *

    P < 10−6 yCON versus T2DM and mCON.

  • P < 0.05 yCON versus T2DM and mCON.

  • P < 10−5 T2DM versus CONm and yCON.

  • §

    P < 0.05 versus T2DM.

Sex (female/male)3/64/53/6
Age (years)58 ± 661 ± 425 ± 3*
BMI (kg/m2)26.9 ± 3.425.5 ± 3.621.9 ± 1.7
HbA1c (%)7.1 ± 1.05.5 ± 0.35.1 ± 0.2
ALT (U/L)27 ± 1122 ± 619 ± 6
Pia (mmol/L)1.08 ± 0.251.50 ± 0.39§1.43 ± 0.23§
γATPa (mmol/L)1.88 ± 0.382.50 ± 0.61§2.26 ± 0.29§
PME (mmol/L)1.81 ± 0.582.43 ± 0.772.29 ± 0.29
PDE (mmol/L)10.36 ± 2.4612.24 ± 2.9311.9 ± 2.72
Pi/γATP0.58 ± 0.020.59 ± 0.020.64 ± 0.02
PME/γATP1.06 ± 0.241.10 ± 0.311.06 ± 0.17
PDE/ γATP6.20 ± 1.205.53 ± 1.055.40 ± 0.90

Experimental Protocol.

All participants refrained from any physical exercise and were on an isocaloric diet for 3 days and fasted for 12 hours before the start of the experiments. According to previous studies, sulfonylurea were withdrawn at one, metformin and alpha-glucosidase inhibitors at 3 days before the clamps.7, 9 Patients treated with insulin (glitazones), incretins, or lipid-lowering drugs were not included in the study. In the morning of the study day, hepatic lipids and phosphorus compounds were measured with 1H/31P MRS (n = 27). On a separate day, whole-body glucose disposal (M) and endogenous glucose production (EGP) were measured in 16 participants (T2DM: three female, four male; mCON: four female, three male; yCON: one female, one male) using clamp studies as described.7 Briefly, a primed-continuous infusion of D-[6,6-2H2]glucose (98% enriched; Cambridge Isotope Laboratories, Andover, MA) was commenced at 6:30 AM. The euglycemic-hyperinsulinemic clamp was started at 10:00 AM with a primed-continuous insulin infusion (Actrapid; Novo, Bagsvaerd, Denmark; 40 mU/m2 body surface area/minute) lasting for 240 minutes. A 20% dextrose infusion labeled with D-[6,6-2H2]glucose (2% enriched) was periodically adjusted to maintain euglycemia.7

Analytical Procedures.

Plasma glucose was measured by the glucose oxidase method (Glucose analyzer II; Beckman Instruments Inc., Fullerton, CA). Plasma triglycerides were assayed colorimetrically (Roche, Vienna, Austria) and insulin was quantified radioimmunometrically.7

31P/1H MRS.

Volunteers were scanned in the prone position with the 10-cm-diameter linearly polarized surface coil positioned under the lateral aspect of the liver in a 3-T whole-body scanner (Medspec S30/80; Bruker Biospin, Ettlingen, Germany). The 31P three-dimensional k-space–weighted MRS imaging (MRSI) localization technique (13 × 13 × 13 matrix; field of view: 20 cm × 20 cm × 20 cm; repetition time: 1 second) with adiabatic excitation pulse was used to acquire 31P spectra (Fig. 1). The protocol including setup required 45 minutes. Quantification was performed as described22 using an MRSI software tool developed in our laboratory23 and quantified in jMRUI (Java-based Magnetic Resonance User Interface)24 using AMARES (advanced method for accurate, robust and efficient spectral fitting of MRS data)25 on the basis of prior knowledge.26 Complete quantification of one MRSI data set lasted 20-30 minutes.

Figure 1.

(A) Axial MR image from the liver of an individual with T2DM depicting MRSI grid, resonance frequency coil, voxels used for quantification (crossed), and four selected voxels. (B) Corresponding spectra (highlighted voxels), acquired with three-dimensional MRSI within 34 minutes. Note: absence of the phosphocreatine (PCr) signal in the liver spectrum (top row) as compared with mixed muscle/liver spectra (bottom and middle rows).

Reproducibility of the assessment of hepatic lipids was determined in humans in different segments of the liver and resulted in variations of lower than 10%, indicating good reproducibility.27 Reproducibility of the measurement of phosphorous metabolites has been tested in vitro using phantoms and in vivo in humans. The standard error of the mean (SEM) was 0.85% of the mean of the reference concentration in a phantom containing K3PO4.22 SEMs were 1.55% and 7.88% of the means of ATP and Pi, respectively, from three measurements in one volunteer, each done after complete repositioning of the volunteer.22 Furthermore, SEMs were 0.40% and 2.16% of the means of the ATP and Pi concentrations, respectively, from measurement by three independent operators.

Localized 1H MRS was performed to measure hepatic lipid contents.7 Data were converted into absolute concentrations expressed as percent fat per volume,28, 29 using the equations by Longo et al.29 Hepatic lipid volume fraction (HLVF), calculated as HCL hepatocellular lipids/(1.138 − 0.339 HCL) is tightly correlated with biopsy-measured liver fat concentrations.29 Measured concentrations of phosphorus [P] metabolites were corrected for the volume captured by lipid droplets within hepatocytes using the equation [Pa] = [P]*(1 − HLVF)−1 to yield [P] adapted for liver fat infiltration.

Calculations and Statistics.

M values were assessed from steady-state glucose infusion rates during the last 30 minutes of the clamp. Rates of glucose appearance (Ra) were calculated by dividing the tracer infusion rate times tracer enrichment by percent of tracer enrichment in plasma and subtracting tracer infusion rate.7 Clamp-Ra was calculated using Steele’s non–steady state equations.7 EGP was calculated from the difference between Ra and mean glucose infusion rates. Insulin-suppressed EGP (iEGP) represents EGP during the last 30 minutes of the clamp.

Statistical analyses were performed using SPSS 6.0 software (SPSS Inc., Chicago, IL). Data are given as means ± standard deviation in text and tables and means ± SEM in figures. Group comparisons were done using analysis of variance with Tukey post hoc testing or Kruskal-Wallis test for parameters with skewed distribution (HLVF, [Pa], glycosylated hemoglobin [HbA1C]). Spearman correlations were used for HLVF and HbA1C, and Pearson product-moment correlations were used for all other variables. Differences and correlations were considered significant at P < 0.05 for phosphorus metabolites and EGP, and at P < 0.01 for other parameters to correct for interrelated comparison. Multiple linear regression analysis was performed for the dependent variables Pi and γATP, including the following parameters: age, BMI, M, iEGP, HbA1C, HLVF, plasma concentrations of aminotransferases, triglyceride, and low-density lipoprotein cholesterol/high-density lipoprotein cholesterol.


Clinical Characteristics.

Individuals in the T2DM and mCON groups were older and more obese than those in the yCON group. They were metabolically well-controlled based on analysis of HbA1C. All other parameters were comparable between the groups (Table 1). Only one male T2DM had an alanine aminotransferase (ALT) level of 46 U/L, exceeding the upper limit of ALT for males of <41 U/L.

Insulin Sensitivity and Liver Fat.

The T2DM group had 42% and 63% lower M values than the mCON and yCON groups, respectively (Fig. 2A). T2DM featured 13-fold higher iEGP than mCON (0.48 ± 0.12 versus −0.04 ± 0.15 mg/kg body weight/minute; P < 0.05) indicating hepatic insulin resistance. HLVF was not different between the T2DM and mCON groups, but was higher (P = 0.002) than in the yCON group (Fig. 2B). Steatosis, as defined by HLVF of >5.5%, was present in six individuals with T2DM, three from the mCON group, and none of those in the yCON group.

Figure 2.

(A) Insulin-stimulated glucose disposal (M value), (B) hepatic lipid volume fraction, and hepatic concentrations of (C) γATP and (D) inorganic phosphate (Pi) in patients with type 2 diabetes (T2DM), matched controls (mCON), and young controls (yCON). Data are means ± SEM.

Hepatic Phosphorus Metabolites.

Also, γATP in T2DM was 26% and 23% lower compared with the mCON and yCON controls, respectively (Fig. 2C). Absolute Pi concentrations were 28% and 31% lower in T2DM than in mCON and yCON, respectively (Fig. 2D). Likewise, PIa and γATPa showed similar alterations, whereas PME, PDE, and derived ratios were comparable (Table 1).

Hepatic Pi and γATP were not associated with M values, but related negatively to hepatic insulin resistance (iEGP) (Fig. 3), HbA1C, BMI, and HLVF (Table 2). Hepatic Pi correlated positively with plasma ALT levels. Adjustment for HLVF abolished these correlations except for that between Pi and γATP and iEGP (r = −0.680, P = 0.007; r = −0.665, P = 0.010). In parallel, Pia and γATPa related negatively to iEGP (R = −0.650, P = 0.009; R = −0.639, P = 0.010) only.

Figure 3.

Correlation between hepatic inorganic phosphate (Pi) and γATP versus insulin-mediated endogenous glucose production (iEGP) as a measure of hepatic insulin resistance in patients with type 2 diabetes (T2DM: triangles, n = 7), matched controls (mCON: circles, n = 7), and young controls (yCON: squares, n = 2).

Table 2. Association Between Absolute Phosphorus Metabolites and Other Parameters
 iEGP (n = 16)HbA1c (n = 27)BMI (n = 27)HLVF (n = 27)ALT (n = 27)
  1. Pearson correlation coefficients (r) and Spearman correlation coefficients (R) for the association of inorganic phosphate (Pi) and γATP with insulin-suppressed endogenous glucose production (iEGP), HbA1c, body mass index (BMI), hepatic lipid volume fraction (HLVF), and alanine aminotransferase (ALT).

[γATP]r = −0.718R = −0.496r = −0.429R = −0.551n.s.
 P = 0.002P = 0.010P = 0.025P = 0.004
[Pi]r = −0.670R = −0.567r = −0.535R = −0.577r = −0.497
 P = 0.004P = 0.003P = 0.004P = 0.003P = 0.010

Multiple linear regression analysis for the dependent variable Pi and γATP including clinical characteristics (see Patients and Methods section) and HLVF, identified iEGP as the single significant independent predictor explaining 57% of Pi and γATP variances (P = 0.001) (Table 3).

Table 3. Multiple Regression Analysis for the Dependent Variables Inorganic Phosphate (Pi) and γATP with Physiologic Parameters
Dependent VariableModelIndependent VariableR2Adjusted R2Standardized Coefficient (β)P Value
  1. Multiple linear regression analysis was performed including the following parameters: hepatic (iEGP, Insulin-Suppressed Endogenous Glucose Production) and Whole-Body Insulin Sensitivity (M, Insulin-Mediated Whole-Body Glucose Disposal), Body Mass Index, age, liver fat content (HLVF, hepatic lipid volume traction) plasma concentration of aminotransferases, triglycerides, and low-density lipoprotein cholesterol/high-density lipoprotein cholesterol.



This study demonstrates that individuals with metabolically well-controlled T2DM have markedly lower hepatocellular concentrations of Pi and γATP than both age-matched and BMI-matched and young lean nondiabetic humans. We employed a recently developed method for absolute quantification of hepatic phosphorus compounds and for assessment of their relationships to insulin resistance and ectopic lipid deposition.22 Before this method, various tracer dilution methods were applied to assess surrogate parameters of human hepatic mitochondrial function.30 However, these methods rely on a range of assumptions, involve whole-body protein turnover, and do not assess hepatic energy metabolism directly. The time course of ATP resynthesis upon fructose infusion has been used as a more direct marker of hepatic mitochondrial function but apparently becomes less efficient with increasing BMI.31

Previous liver 31P MRS studies were limited by the inability to estimate absolute concentrations, making it necessary to express the intensities of phosphorus signals as ratios to that of another metabolite and/or to the assumed concentration of a metabolite. Consequently, qualitative MRS would fail to detect these alterations in the presences of parallel increases or decreases of metabolites as shown in alcoholic liver disease.32 Therefore, we conclude that it is highly recommended to use absolute quantification based on external references rather than peak ratios.

The measurement on a 3-T scanner provides for greater signal-to-noise ratio and thereby improved accuracy compared to MRS studies performed on 1.5-T to 2-T MR scanners,33, 34 which is necessary for an accurate measurement of metabolite concentration. Further advantages include improved spatial resolution, speed-up of the measurement due to k-space–weighted acquisition, B1 homogeneity–insensitive excitation, the possibility to quantify a very high number (on average 55±5) of spectra per experiment, and high reproducibility of the metabolite concentrations.

Overall, the absolute concentrations measured here are in agreement with that of 1.4-2.8 mmol/L for Pi and 1.6-3.8 mmol/L for γATP.32–37 In addition to MRS-dependent differences, the great interindividual variability within and across these studies could be due to the nutritional status and variations of age and BMI. To overcome such influences, we controlled for dietary intake and matched the controls for age and BMI. Of note, the high intraindividual and interobserver reproducibilities of γATP and Pi made it possible to investigate more subtle changes in phosphorus metabolites across a range of hepatic insulin resistance and fat contents.22

The decrease in absolute levels of γATP in viral and alcoholic hepatitis32, 37 and obesity38 has been commonly interpreted as “energy deficit” or impaired “ATP homeostasis”. Decreased γATP could result from hepatocellular γATP depletion due to increased ATP utilization by energy-demanding processes such as Na+/K+ adenosine triphosphatases, lipogenesis, or gluconeogenesis. However, stimulation of catabolism would imply increase of Pi levels and/or Pi/γATP ratios, which reflect the bioenergetic state of the cell and have been used as a marker for reduced phosphorylation status39 in livers of humans suffering from obesity,38 viral cirrhosis,40 or malignancies.41 In contrast, individuals from our T2DM group showed comparable reductions of Pi and γATP, indicating balanced anabolic/catabolic rates. However, one cannot rule out that they had decreased mitochondrial ATP synthesis arising from impaired glycolysis, fat oxidation, or oxygen consumption.

Alternatively, parallel reduction of γATP by 25%-36% and of Pi by 37%-50% in alcoholic hepatitis with ALT elevation has been explained by loss of functional hepatocytes due to necrosis and replacement with fat and collagen.32 In our T2DM group, similar reductions of γATP and Pi were observed without any clinical or laboratory evidence of liver damage, which does not rule out excessive hepatocellular fat deposition as a cause of lower phosphorus metabolites. However, even after correcting the absolute γATP and Pi concentrations for hepatocellular lipid volume, γATP and Pi remained lower in the T2DM group, indicating reduced energy metabolism independent of liver fat content.

Reduction of Pi and γATP could also be a feature of hepatic insulin resistance preceding hepatocellular lipid accumulation. Increased circulating lipids derived from dietary intake and impaired inhibition of lipolysis stimulate lipid oxidation and production of reactive oxygen species, giving rise to lipid peroxidation and damage of mitochondrial proteins and DNA, reduction of oxidative capacity, and subsequent lipid accumulation.42, 43 In support of this contention, patients with NASH exhibit paracrystalline inclusions in megamitochondria,11, 44 increased mitochondrial DNA mutations,45 and lower respiratory complex activities in the liver.20 But even obese subjects already may have lower liver ATP resynthesis upon fructose administration.31 In our study, HLVF was similar in patients with T2DM and mCON with only five individuals from all groups with excessive steatosis at HLVF of >10% and only one man with T2DM with slightly elevated ALT. This indicates that our patients had steatosis and were unlikely to be suffering from NASH. Of note, HLVF related inversely to Pi and γATP, which strongly relied on the negative association of hepatic insulin resistance with Pi and γATP, which was identified as an independent predictor of the concentrations of these metabolites.

A recent study reported higher PME/Pi and PME/ATP ratios along with lower insulin sensitivity in overweight patients with NAFLD, as diagnosed with ultrasound, compared to lean subjects.46 Because PME not only comprises phosphocholine, phosphoethanolamine and adenosine monophosphate but also glucose-6-phosphate and 3-phosphoglycerate,47 the higher PME signal was interpreted as a surrogate of increased gluconeogenesis. More likely, the increase in PME/Pi or PME/ATP simply resulted from reduction of the absolute concentrations of Pi or γATP. Of note, increased PME ratios have been repeatedly found in patients with severe liver diseases including hepatitis, liver cirrhosis, or neoplasia and therefore rather reflect extensive membrane remodeling.48–50 In this study, PME concentrations and ratios did not differ between the groups, which further supports the absence of liver damage in our T2DM individuals.

Up to 60% of the general population shows defects of the hepatic respiratory chain despite normal liver function.51 Because the majority of this group is older than 50 years, aging-induced mitochondrial damage could contribute to the observed alterations of phosphorus metabolites in our T2DM individuals. Reduced energy metabolism could be due to loss of functional hepatocytes or reduction of mitochondrial contents and/or fitness.15 However, reduction of absolute Pi and γATP concentrations were found in T2DM but not in age-matched and BMI-matched healthy humans. In the face of age-dependent decline of whole-body insulin sensitivity across the groups, these data suggest a diabetes-related or even diabetes-specific effect causing the reduction of hepatic Pi and γATP. This is corroborated by a negative relationship between Pi and γATP and glucometabolic control as assessed from HbA1C and iEGP, which not only reflects hepatic insulin sensitivity but also defines plasma glucose concentrations.

Some limitations need to be taken into account. First, no biopsies were done, and thus exact information on inflammatory status is lacking. This was due to the fact that biopsies are not considered ethical in subjects without any evidence of liver disease. Second, not all participants were available for the clamp test. Nevertheless, the overall number of participants allowed detection of associations with clinically relevant parameters. Finally, our T2DM cohort comprised normal weight (n = 3), overweight (n = 3), and obese (n = 3) patients resulting in a mean BMI of ∼27 kg/m2. Because T2DM cohorts mostly have higher BMI values, the results obtained in our cohort cannot be necessarily extrapolated to all patients with T2DM. In Europe, however, Caucasian cohorts with T2DM and/or NAFLD may present with mean BMI values below 30 kg/m2.52–54 Likewise, HLVF values did not cover a broad range from normal fat content to excessive steatosis. However, we detected alterations in phosphorus metabolites even within the normal range.

In conclusion, patients with metabolically well-controlled T2DM already show reduced hepatic ATP concentrations which relates to hepatic insulin resistance independent of hepatic lipid contents. These data further suggest that impaired hepatic energy metabolism and hepatic insulin resistance could precede the development of steatosis in patients with T2DM.