Introduction
Cardiovascular autonomic neuropathy (CAN), with its hallmark reduced heart rate variability (HRV), affects approximately 20% of people with diabetes and predicts an increased risk of major cardiac events and mortality [
1]. We recently demonstrated that lower HRV, indicating diminished cardiovagal tone, in recent-onset type 2 diabetes is associated with insulin resistance, hepatic steatosis and blunted cardiorespiratory fitness, suggesting that these components could play an important role in the early development of CAN, apart from well-known risk factors such as higher age, obesity, hypertension or poor glycaemic control [
2]. Indeed, it has been emphasised that hyperglycaemia cannot be considered as the sole factor responsible for the development of diabetic complications, particularly in patients with type 2 diabetes. Instead, an interplay of the aforementioned factors that have an impact on the adipose tissue fatty acid metabolism could underlie the onset and progression of diabetic microvascular complications including neuropathy [
3].
There is accumulating evidence suggesting that in addition to hyperglycaemia, dyslipidaemia could represent a contributing pathogenetic factor to neuropathy, particularly for type 2 diabetes. Since both hyperglycaemia and dyslipidaemia affect multiple cells in the peripheral nervous system, including neuronal axons, Schwann cells and dorsal root ganglia, deciphering the mechanisms by which perturbed glucose and lipid metabolism converge to result in nerve damage could foster the development of novel lipid-based disease-modifying treatments for diabetic neuropathy [
4]. Previous studies assessing large cohorts of individuals with type 2 diabetes point to a link between plasma lipid levels and diabetic neuropathies. The Fremantle Diabetes Study reported that the use of fibrates and statins was associated with lower risks of developing polyneuropathy over 5 years [
5]. Furthermore, the Anglo-Danish-Dutch Study of Intensive Treatment in People With Screen-Detected Diabetes in Primary Care (ADDITION)-Denmark study showed that hypertriacylglycerolaemia was associated with prevalent CAN [
6], while low HDL-cholesterol levels were predictors of incident polyneuropathy over 13 years [
7].
The past decade has witnessed novel technologies such as metabolomics aimed at the extensive characterisation and quantification of global metabolites from both endogenous and exogenous sources in the context of insulin resistance, the metabolic syndrome and type 2 diabetes [
8‐
11]. As a subfield of metabolomics, lipidomics emerged to investigate the relationship of dysregulation in lipid metabolism and pathogenesis of type 2 diabetes [
12] and diabetic neuropathy [
13].
To date, there are no studies that assess the relationship between cardiac autonomic function and metabolomic profiles including lipidomics in individuals with type 1 or type 2 diabetes. Using targeted fasting plasma metabolomic analysis, we previously reported both similarities and diabetes type-specific differences in the metabolite patterns when comparing participants with type 1 diabetes and type 2 diabetes from the German Diabetes Study (GDS) baseline cohort [
14]. Given the complex interplay between hyperglycaemia, insulin resistance, obesity and lipid metabolism as the putative factors contributing to diabetic neuropathy, we hypothesised that a distinct link exists between specific lipid metabolites and cardiac autonomic dysfunction in recent-onset type 2 diabetes as opposed to type 1 diabetes.
Discussion
The results of this study demonstrate that higher plasma levels of distinct lipid metabolites, namely phosphatidylcholines (five diacyl and one acyl-alkyl) and SMs (C16:0 and C16:1) are linked to cardiac autonomic dysfunction, particularly to lower cardiovagal tone, in individuals recently diagnosed with type 2 diabetes. In contrast, no associations of lipid metabolites with cardiac autonomic function were found in participants with recent-onset type 1 diabetes, suggesting a role for perturbed lipid metabolism in the early development of CAN specifically in type 2 diabetes. However, the 5 year follow-up period was too short to detect clinically relevant deterioration in HRV in excess of the physiological effect of ageing and thereby to allow for analyses of the predictive value of lipid metabolites in the development or progression of CAN.
There are no published studies available with which our findings could be directly compared. Previous cohort studies focused on the predictive value of various lipid metabolites, in particular phosphatidylcholine consumption, on incident type 2 diabetes as well as CVD and mortality risk. Phosphatidylcholines are a class of phospholipids that incorporate choline as a headgroup. They are a major component of biological membranes and can be easily obtained from a variety of readily available sources, such as egg yolk or soybeans, from which they are mechanically or chemically extracted using hexane. In the Nurses’ Health Study (NHS), NHS II and the Health Professionals Follow-up Study (HPFS), dietary intake of phosphatidylcholine was associated with an increased risk of incident type 2 diabetes [
21]. Moreover, in the NHS and HPFS, higher phosphatidylcholine consumption was associated with increased all-cause and cardiovascular mortality risk, especially in patients with diabetes, independent of traditional risk factors [
22]. Recent animal studies point to a mechanistic link between intestinal microbial metabolism of the choline moiety in dietary phosphatidylcholine and CVD through the production of a proatherosclerotic metabolite, trimethylamine-
N-oxide (TMAO), which is associated with an increased risk of incident major adverse cardiovascular events [
23]. However, the way in which TMAO could also play a role in the context of type 2 diabetes and CAN remains unknown.
The present study also shows associations between higher concentrations of SMs C16:0 and C16:1 and lower HRV indices, largely indicating diminished cardiovagal tone. Sphingolipids are complex lipids that are particularly abundant in nervous tissue and are implicated not only in a number of neurological diseases but also in insulin-resistant conditions such as diabetes or non-alcoholic steatohepatitis [
24,
25]. Higher concentrations of specific SMs have recently been found to predict incident type 2 diabetes in prospective sphingolipidomics studies over 6 and 11 years, respectively [
26,
27]. The formation of atypical neurotoxic deoxysphingolipids has been identified to play a causative role in the development of hereditary sensory and autonomic neuropathy type 1 (HSAN1) [
12]. Of note, 1-deoxysphingolipids have also been found to be elevated in individuals with type 2 diabetes and in non-diabetic individuals with the metabolic syndrome [
28]. Moreover, 1-deoxysphingolipid levels were also increased in individuals with type 2 diabetes who had polyneuropathy, when compared with healthy individuals, but there were no correlations between these levels and peripheral nerve function tests or the clinical neuropathy stages [
12]. In an open-label clinical trial in individuals with primary hypercholesterolemia or mixed dyslipidaemia, treatment with fenofibrate for 6 weeks resulted in lowering of plasma 1-deoxysphingolipid levels [
29]. An open-label trial in individuals with type 1 diabetes showed that supplementation with seal oil ω-3 polyunsaturated fatty acids over 12 years was associated with an increase in corneal nerve fibre length [
30]. In the Fenofibrate Intervention and Event Lowering in Diabetes (FIELD) study, treatment with fenofibrate for 5 years in individuals with type 2 diabetes was associated with a lower risk of amputations in the lower limbs, particularly minor amputations in the absence of peripheral arterial disease, which were predicted by neuropathy [
31]. Moreover, the de novo sphingolipid synthesis pathway is considered a promising target for pharmacological intervention in insulin resistance. It has been shown that inhibition of serine palmitoyltransferase, the first enzyme in the sphingolipid biosynthesis pathway, increases insulin sensitivity [
32]. Thus, there is accumulating evidence supporting the notion that lipidomics-based novel disease-modifying treatment approaches could merit further investigation in type 2 diabetes patients with polyneuropathy and CAN.
Other lipid metabolites contributing to an increased risk of diabetes and CVD include acylcarnitines [
33‐
35]. Carnitine is a quaternary ammonium compound involved in fatty acid metabolism, maintaining the balance between free and esterified CoA, which is crucial for normal cell function. In humans, carnitine exists as free active
l-carnitine or as acylcarnitines (i.e. esterified forms with various chain lengths) [
36]. Carnitine plays an important role in transporting long-chain fatty acids from the cytosol to the mitochondrial matrix, where β-oxidation takes place, and accumulation of acylcarnitines may reflect dysregulated fatty acid oxidation [
37]. In a Chinese study, a panel of acylcarnitines mainly involved in mitochondrial lipid dysregulation predicted incident type 2 diabetes beyond conventional risk factors [
33]. Among individuals with suspected stable angina pectoris, elevated serum even-chained acylcarnitines were associated with increased risk of cardiovascular death and to a lesser degree with acute myocardial infarction, again independent of traditional risk factors [
35]. Although we observed inverse associations of four acylcarnitines with several HRV indices (see ESM Table
4), statistical significance was lost after Bonferroni correction, while none of the ten NEFAs showed associations after rigorous adjustment (see ESM Table
1). However, since a type II error cannot be excluded, it is conceivable that if these associations were true, they would obviously be weaker than for the aforementioned lipid metabolites.
The potential source of bias resulting from the disparity in the use of lipid-lowering medications, the vast majority of which were statins, in 23% and 1% of participants with type 2 and type 1 diabetes, respectively, deserves comment. To address this aspect, all relevant analyses were adjusted for lipid-lowering medication. In epidemiological studies, the statin-mediated lipidomic changes in individuals with the metabolic syndrome or type 2 diabetes showed a significant shift towards the lipid profile of control individuals, indicative of a marked trend towards a normolipidemic phenotype [
38]. Moreover, administration of rosuvastatin for 3–8 weeks in individuals with hyperlipidaemia was associated with decreased levels of phosphatidylcholines and acylcarnitines and increased levels of polyunsaturated fatty acids, favouring an improvement of the atherogenic lipid profile [
39]. Thus, we would expect a favourable effect of statins towards a normalisation of the plasma lipidome which could rather attenuate the associations between HRV indices and lipid metabolites observed herein.
Although dyslipidaemia is increasingly recognised as an important factor contributing to the pathogenesis of neuropathy, particularly in type 2 diabetes [
3,
4], it is currently not well understood whether specific lipid classes and levels in the nerve are impacted [
40]. It cannot be determined from this study whether the observed increase in systemic lipid levels in relation to lower HRV mirrors the local content in the peripheral nerves in diabetes. However, there is emerging evidence suggesting that the nerve concentrations of lipid metabolites such as phosphatidylcholines and SMs are elevated in mouse models of type 2 diabetes [
41]. Recently, a dysregulation of lipid pathways with an increase in triacylglycerols containing saturated fatty acids and increased expression of diacylglycerol acyltransferase 2 (DGAT2), the enzyme required for the last step in triacylglycerol synthesis, was identified. Increased DGAT2 expression was present not only in nerves assessed in murine models of type 2 diabetes but also in sural nerve biopsies from hyperlipidaemic individuals with diabetes and peripheral neuropathy. These findings support the hypothesis that abnormal nerve–lipid signalling is an important factor in the pathogenesis of neuropathy in type 2 diabetes [
40]. However, to date, no such experimental and clinical evidence is available for autonomic nerves.
This study has several strengths. First, it has a relatively large sample size of individuals with well-controlled type 1 and type 2 diabetes who underwent comprehensive metabolic characterisation assessed by state-of-the-art methodology. Second, HRV was determined during a 3 h hyperinsulinaemic–euglycaemic clamp, avoiding the impact of confounding factors such as blood glucose fluctuations [
42]. Third, rigorous adjustment for numerous possible confounders and for multiple testing was applied. Yet this study also has some limitations. First, the cross-sectional design for the lipid metabolites and the relatively short follow-up period without deterioration in HRV after 5 years do not provide insight into the temporal sequence of the observed associations. Second, although a control group with normal glucose tolerance and HRV measurements is part of the GDS, lipidomics data for comparison with the diabetes groups were not available. Thus, it cannot be unequivocally determined from this study whether the observed associations of higher plasma levels of specific phosphatidylcholines and SMs with lower HRV indices can be attributed specifically to type 2 diabetes. However, indirect evidence suggests that such a scenario is conceivable, since on the one hand no such associations were found in recent-onset type 1 diabetes and on the other hand the systemic concentrations of both phosphatidylcholines and SMs predict the development of type 2 diabetes [
11,
26,
27]. Third, statistical power of the prospective analysis was limited due to the relatively high number of dropouts at 5 years.
In conclusion, using targeted lipidomics we demonstrated that higher plasma levels of phosphatidylcholines and SMs are closely associated with lower cardiac vagal tone in individuals recently diagnosed with type 2 diabetes as opposed to those with type 1 diabetes. However, since cardiac autonomic function did not deteriorate over 5 years, analysis of the predictive value of lipid metabolites for the progression to CAN was not readily feasible. Further follow-up of the GDS participants over 10 and more years will presumably reveal deterioration in HRV in both participants with type 1 diabetes and type 2 diabetes, a prerequisite to allow for a more promising prediction analysis. Thus, plasma lipid panels could not only be useful to improve the prediction of the longer-term development or progression of CAN but also may allow for the clinical stratification of patients early in the course of the disease to target interventions in a more individualised approach to particularly susceptible patients. It has been suggested that given the key role of lipids in the pathophysiology of type 2 diabetes and CVD, lipidomics in general has the potential to improve prediction of future disease risk, inform on mechanisms of disease pathogenesis, identify patient groups responsive to particular therapies, and more closely monitor response to therapy. The ultimate utility of lipidomics to clinical practice will depend on: (1) its ability to predict future risk of morbidity and mortality when incorporated into conventional clinical risk engines; and (2), for widespread application, lipidomic-based measurements must be practical and accessible through standard pathology laboratories [
43]. It remains to be established whether targeted lipidomics could be helpful in developing novel, potentially disease-modifying lipid-lowering treatment modalities.
Acknowledgements
We appreciate the voluntary contribution of all study participants. The authors wish to thank the staff of the Research Group Neuropathy, Institute for Clinical Diabetology at the German Diabetes Center, Düsseldorf, Germany, especially F. Battiato, M. Schroers-Teuber and J. Schubert for their excellent work. Some of the data were presented as an abstract at the EASD meeting in 2020. The GDS Group consists of A. E. Buyken (University of Paderborn, Paderborn, Germany), J. Eckel (German Diabetes Center, Leibniz Center for Diabetes Research at Heinrich Heine University, Düsseldorf, Germany), G. Geerling (Heinrich Heine University, Düsseldorf, Germany), and H. Al-Hasani, C. Herder, A. Icks, J. Kotzka, O. Kuss, E. Lammert, W. Rathmann, J. Szendroedi, S. Trenkamp, D. Ziegler and M. Roden (speaker) (all at the German Diabetes Center, Leibniz Center for Diabetes Research at Heinrich Heine University, Düsseldorf, Germany).
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.