As insulin resistance and hypertriglyceridaemia are not features of T1DM, HDLs in individuals with T1DM tend to be the same size or larger than in healthy control individuals, and levels of HDL cholesterol are often increased in people with T1DM13,20,21. T1DM is, nevertheless, associated with increased cardiovascular disease risk22. This association raises the possibility that the cardioprotective functions of HDLs in T1DM might be reduced by post-translational modifications such as non-enzymatic glycation and oxidation of their lipid and apolipoprotein constituents, rather than by metabolic changes as is the case in T2DM. However, the evidence for this is conflicting, as outlined in the following section.
HDLs have several potentially cardioprotective functions, the best understood of which is their capacity to remove excess cholesterol from peripheral cells, including macrophages in the artery wall. HDLs also have a pivotal role in the reverse cholesterol transport pathway, whereby the cholesterol that effluxes from peripheral cells via ABCA1 to APOA1 and small spherical HDLs is esterified by LCAT. These cholesteryl esters are subsequently transported to the liver, where they are selectively removed from HDLs by scavenger receptor class B type 1 (SRB1) (Fig. 1a), incorporated into bile and excreted in faeces. HDLs are also important for maintaining endothelial function and they have anti-inflammatory and antioxidant activities.
Several investigators have reported that cholesterol efflux is associated with decreased cardiovascular risk and cardiovascular events, although this finding has not been confirmed by others. The efflux of cholesterol from cell membranes to lipid-free APOA1 and lipid-poor APOA1, discoidal HDLs, and small, spherical HDLs is mediated by ABCA1, and the related ATP-binding cassette transporter G1 (ABCG1) effluxes cholesterol to large, spherical HDLs (Fig. 1a).
ABCA1 expression, ABCG1 expression and cholesterol efflux to HDLs are decreased in T2DM (Fig. 1b). ABCG1 expression and cholesterol efflux to HDLs are also decreased, and cholesterol accumulation in macrophages is increased in the db/db mouse model of T2DM. Collectively, these findings suggest that at least some of the increased cardiovascular disease risk in T2DM can be attributed to reduced HDL cholesterol efflux capacity. However, other reports showing that the cholesterol efflux capacity of HDLs from patients with T2DM is similar to or increased relative to that seen in healthy controls call this interpretation into question. These discrepant results probably reflect variations in experimental conditions, including the use of different cell types as cholesterol donors, variations in the expression of ABCA1 and ABCG1 and their cellular distribution (intracellular versus within the cell membrane), as well as the use of isolated HDLs versus APOB-depleted serum and APOB-depleted plasma as cholesterol acceptors. These differences highlight a need for the development and adoption of standardized experimental conditions before the full effect of diabetes mellitus on cholesterol efflux capacity can be evaluated with confidence.
The cholesterol efflux capacity of serum from patients with T2DM and hypertriglyceridaemia is also increased relative to that seen in patients with T2DM and normal levels of triglycerides. This result can be explained by the expanded TGRLP pool in these patients acting as a sink for the core lipids that are transferred from HDLs to TGRLPs by CETP. This transfer depletes HDLs of core lipids and promotes the dissociation of lipid-free or lipid-poor APOA1, which can act as a cholesterol acceptor, from the particles. This observation is consistent with the report of a positive association of CETP activity with cholesterol efflux to APOB-depleted plasma in patients with T2DM.
Whether the persistently high blood levels of glucose in people with poorly controlled diabetes mellitus, as well as in db/db mice, also affect cholesterol efflux by generating reactive α-oxoaldehydes that non-enzymatically glycate HDL apolipoproteins has also been investigated. However, studies in multiple cell types, including the J774 macrophage cell line, the THP-1 monocytic cell line, human monocyte-derived macrophages, 3T3-L1 adipocytes and in individuals with T2DM are not conclusive; in these studies, non-enzymatic glycation of HDL apolipoproteins was associated with no or reduced cholesterol efflux to APOA1, isolated HDLs, APOA1-containing, synthetic reconstituted HDLs (rHDLs), or APOB-depleted serum and APOB-depleted plasma. Cholesterol efflux from J774 macrophages and the Fu5AH hepatoma cell line to APOB-depleted serum is also decreased, irrespective of glycaemic control, in T1DM and T2DM. On the other hand, non-enzymatic glycation of HDL apolipoproteins has no effect on reverse cholesterol transport in db/db mice. Overall, these outcomes indicate that, although non-enzymatic glycation of HDL apolipoproteins might affect cholesterol efflux under some circumstances, its influence on the removal of excess cholesterol from the body via the reverse cholesterol transport pathway is probably minimal.
There is also debate about whether HDL size regulates cholesterol efflux in people with diabetes mellitus. A study published in 2020 reported that small HDLs promoted more cholesterol efflux from baby hamster kidney cells in which ABCA1 expression was induced than did medium-sized HDLs or large HDLs. This study also reported that, compared with ABCA1-mediated cholesterol efflux to isolated small HDLs from people without T2DM, ABCA1-mediated cholesterol efflux to isolated small, spherical HDLs from people with T2DM was reduced. That reduction in cholesterol efflux was attributed to decreased association of serpin family A member 1 with small HDLs. This finding contrasts with an earlier report in which the main cholesterol acceptors in plasma from patients with T2DM were medium-sized HDLs. This discrepancy might also reflect that the presence of lipid-free or lipid-poor APOA1 in APOB-depleted plasma could promote cholesterol efflux independent of spherical HDLs.
Discrepant results have also been reported for SRB1. Selective uptake of cholesteryl esters from HDLs into macrophages via this receptor was increased in mice with streptozotocin-induced diabetes mellitus and in J774 cells exposed to high levels of glucose but was decreased in THP-1 cells exposed to high levels of glucose. These findings further highlight that variations in experimental design and conditions can considerably influence experimental outcomes. The use of conditions that recapitulate physiological situations wherever possible could minimize these discrepancies and increase confidence in identifying future directions that have the potential for therapeutic impact.
The antioxidant function of HDLs refers to their capacity to inhibit the formation of pro-atherogenic, oxidized low-density lipoproteins (oxLDLs) (Fig. 1a). HDLs do this by preventing the production of cytotoxic and pro-inflammatory lipid hydroperoxides, oxidized phospholipids, and reactive aldehydes and by accepting cytotoxic lipid hydroperoxides from low-density lipoproteins (LDLs) and converting them into inert lipid hydroxides. HDLs also limit oxidative stress in cells by suppressing reactive oxygen species (ROS) formation.
The antioxidant capacity of HDLs from patients with T1DM or T2DM is reported to be impaired. HDLs from individuals with T2DM and HDLs that have been non-enzymatically glycated in vitro by incubation with methylglyoxal both have a decreased capacity to protect against oxidative stress. However, this result was not confirmed in other studies, where no difference between the antioxidant function of HDLs from healthy control individuals and HDLs from patients with T2DM was observed. It is not known if methylglyoxal reduces the capacity of HDLs from individuals with T1DM to decrease oxidative stress.
Evidence that dyslipidaemia impairs the antioxidant function of HDLs comes from studies in which hypertriglyceridaemia was correlated with reduced antioxidant activity. One study of participants with T2DM and hypertriglyceridaemia reported impaired antioxidant activity only in large HDLs, whereas another study reported reduced antioxidant activity in small, dense HDLs but not in large HDLs. This discrepancy might be related to variations in glycaemic control between the patient groups and the small group sizes in these studies. Variation in the assays used for evaluating the antioxidant capacity of HDLs and the methods used for HDL isolation, as well as the extent of dyslipidaemia in the participants, might have also affected the study outcomes. Although the conclusions are limited by a lack of methodological consistency across the studies, they nevertheless suggest that improving glycaemic control and the use of first-line lipid-lowering therapies, where appropriate, might conserve and possibly enhance the antioxidant function of HDLs.
Studies of the effect of free fatty acids on the antioxidant function of HDLs are, by contrast, more consistent. Increased levels of oxidized free fatty acids in HDLs from patients with T2DM correlate inversely with the ability of the particles to suppress endothelial ROS production and prevent oxidized phospholipid formation.
The antioxidant function of HDLs might also be regulated by the particle proteome. Although reports of reduced levels and reduced activity of the HDL-associated antioxidant enzyme paraoxonase 1 (PON1) in people with T2DM are consistent, this is not the case for T1DM. The level and activity of HDL-associated PON1 in patients with T1DM have been variously reported as increased, decreased and unchanged relative to healthy control individuals. These discrepant outcomes might reflect the use of different assays to measure PON1 activity, variations in HDL isolation protocols and differences in disease duration, ranging from 1 to 18 years. As non-enzymatic glycation of HDL apolipoproteins in patients with poorly controlled diabetes mellitus is also associated with reduced HDL PON1 activity, it follows that this modification needs to be considered when assessing the antioxidant function of HDLs in T1DM as well as T2DM.
Modification of HDL apolipoproteins by the pro-oxidant enzyme myeloperoxidase (MPO), and the ability of MPO to degrade thiocyanate into cyanate, which carbamylates HDL-associated proteins, also reduces PON1 activity and the antioxidant function of HDLs (Fig. 1b). MPO and PON1 additionally form ternary complexes with HDLs, whereby PON1 reduces MPO activity and MPO oxidatively inactivates PON1 (ref. ). In this system, an imbalance between levels of MPO and levels of PON1 in favour of MPO further exacerbates the loss of HDL antioxidant function in T2DM (Fig. 1b). Levels of MPO are also correlated with reduced HDL capacity to suppress ROS production in endothelial cells in T2DM and increased arterial stiffness in children with T1DM. Collectively, these studies provide strong evidence that the antioxidant function of HDLs is impaired in T1DM and T2DM.
Endothelial dysfunction, an initiating factor in atherosclerotic lesion development, is prevalent in people with diabetes mellitus. This association has been attributed to two key features of diabetes mellitus, hyperglycaemia and elevated levels of free fatty acids, which cause mitochondrial dysfunction, ROS production, activation of protein kinase C (PKC) and reduced synthesis of the endothelial vasoprotective molecule nitric oxide (NO) (Fig. 1b). Endothelial dysfunction is also associated with decreased HDL function in adolescents with T1DM and renal dysfunction and is correlated with systemic inflammation and impaired NO bioactivity. Hyperglycaemia and the binding of non-enzymatically glycated HDL apolipoproteins to advanced glycation end product (AGE) receptors further exacerbates endothelial dysfunction by increasing pro-inflammatory signalling and adhesion molecule expression.
There is mounting evidence that unmodified HDLs can mitigate these adverse effects of diabetes mellitus on endothelial function and vasorelaxation by binding to SRB1 (ref. ). This binding induces the non-receptor tyrosine kinase SRC and activates the phosphatidylinositol 3-kinase (PI3K)-AKT and PI3K-mitogen-activated protein kinase (MAPK) signalling pathways. This activation, in turn, activates endothelial NO synthase (eNOS) and generates NO (Fig. 1a).
Direct in vivo evidence of the vasoprotective effects of HDLs comes from restoration of NO bioavailability and increased endothelium-dependent vasorelaxation in patients with T2DM after a single rHDL infusion. This benefit has been attributed to the binding of bioactive lipids, including sphingosine-1-phosphate (S1P), to HDLs, which activates eNOS and increases NO production. It is also in line with the association of low levels of HDL-associated S1P with reduced eNOS activity and endothelial dysfunction in patients with T2DM and restoration of endothelial function by supplementing non-enzymatically glycated rHDLs with S1P. These observations are consistent with the results of a cross-sectional study of patients with T2DM in which decreased levels of HDL-associated S1P were correlated with reduced glycaemic control and eNOS activation. Decreased plasma levels of S1P and levels of HDL-associated S1P have also been confirmed in T2DM and T1DM, although this finding was not confirmed by others. These discrepant findings suggest that regulation of HDL function by S1P in diabetes mellitus is dependent on multiple factors, including HDL composition, disease duration and glycaemic control. The mechanistic basis of these observations might be related to reduced cross-linking of S1P to non-enzymatically glycated APOM, the apolipoprotein with which S1P associates on the HDL surface. This reduction has been reported to promote preferential partitioning of S1P-APOM complexes into large, less dense HDLs, and is associated with decreased activation of the MAPK and PI3K-Akt signal transduction pathways and impairment of the ability of HDLs to inhibit inflammation in endothelial cells.
Although these reports implicate HDL S1P content in the regulation of vascular function in T1DM and T2DM, it is not known if this relationship is causal. Moreover, we do not know whether targeting sphingolipid metabolism can restore the capacity of HDLs to improve endothelial function in people with T1DM or T2DM. Improved understanding of this area has the potential to identify novel interventions for reducing diabetes mellitus-associated cardiovascular complications.
APOA1-containing rHDLs also repair damaged endothelium by stimulating endothelial progenitor cell proliferation and function, something that HDLs from patients with T1DM or T2DM are unable to do. HDLs from patients with T1DM are additionally unable to reverse the impaired endothelium-dependent vasorelaxation that is induced by exposure to oxLDLs, and non-enzymatically glycated HDLs do not prevent oxLDLs from impairing endothelium-dependent vasorelaxation (Fig. 1b). Downregulation of eNOS expression and endothelial levels of NO by non-enzymatically glycated HDLs is also correlated with a reduction in key HDL-associated antioxidant enzymes and increased endothelial ROS production (Fig. 1b). This observation is consistent with a 2022 report in which HDLs from patients with diabetes mellitus increased levels of oxidized fatty acids and ROS production in cultured endothelial cells. HDLs can also prevent endothelial-to-mesenchymal transition, a form of endothelial dysfunction in which endothelial cells trans-differentiate into mesenchymal cells, which do not have cardioprotective functions, when exposed to high levels of glucose. Collectively, these findings indicate that increasing levels of unmodified HDLs, whether pharmacologically, with other interventions or with rHDL infusions, has the potential to improve endothelial function and reduce pro-oxidant burden in T1DM and T2DM.
Diabetes mellitus is associated with elevated plasma levels of inflammatory mediators and biomarkers, including C-reactive protein, soluble intercellular adhesion molecule 1 (ICAM1), soluble vascular cell adhesion molecule 1 (VCAM1) and pro-inflammatory cytokines such as IL-6, IL-1β and tumour necrosis factor (TNF). HDLs from healthy individuals inhibit several of these pro-inflammatory mediators in different cell types. For example, they reduce the cytokine-activated expression of ICAM1 and VCAM1 (Fig. 1a) on the surface of endothelial cells by inhibiting activation of the master regulator of inflammation nuclear factor-κB (NF-κB) in human vascular endothelial cells and coronary artery endothelial cells. HDLs also inhibit macrophage polarization towards a pro-inflammatory phenotype and decrease pro-inflammatory cytokine secretion from macrophages.
However, HDLs in patients with T2DM and end-stage renal disease, obesity and peripheral artery disease have a reduced capacity to inhibit NF-κB activity in cytokine-activated endothelial cells. In addition, HDLs from patients with T2DM do not inhibit endothelial expression of ICAM1 and VCAM1 in New Zealand White rabbits and in human coronary artery and human umbilical vein endothelial cells with acute vascular inflammation. This observation is consistent with a report showing that non-enzymatic glycation of APOA1 suppresses expression of the monocyte integrin CD11b and increases monocyte adhesion to endothelial VCAM1 and endothelial ICAM1 (refs. ). Loss of the anti-inflammatory function of HDLs is also associated with reduced levels of PON1, reduced PON1 activity and increased levels of the acute-phase protein serum amyloid A. Increased levels of MPO and apolipoprotein carbamylation and increased activation of NF-κB in endothelial cells in patients with T2DM further decrease the anti-inflammatory function of HDLs in these individuals.
However, HDLs from patients with T2DM and non-alcoholic fatty liver disease (also known as metabolic dysfunction-associated steatotic liver disease) and HDLs from healthy control individuals have been reported to reduce inflammation in human vascular endothelial cells to a similar extent. The ability of HDLs isolated from patients with T1DM with poor glycaemic control to inhibit inflammation in adipocytes has also been reported as similar to that of HDLs from healthy control individuals. The reasons for these discrepancies in relation to the studies outlined in the preceding paragraph are not known, but it is noteworthy that many of the studies are small and that some of the outcomes might have been chance findings. It is additionally important to note that infusion of rHDLs that have not been exposed to high levels of glucose considerably reduces endothelial cell inflammation in patients with T2DM and dyslipidaemia. This finding, in addition to the results of other studies, suggests that in vivo oxidative modifications of HDL lipid constituents and non-enzymatic glycation of HDL apolipoproteins might contribute, at least in part, to loss of the anti-inflammatory function of HDLs in diabetes mellitus.
All HDL apolipoproteins are susceptible to non-enzymatic glycation. However, as reaction intermediates and AGEs are rapidly degraded in vivo, it is difficult to obtain an accurate picture of the cumulative effect of reaction intermediates, such as Schiff bases and Amadori products, on HDL function. This difficulty has led to the use of static, in vitro systems in which supraphysiological concentrations of reactive α-oxoaldehydes are used to study the functional effect of non-enzymatic glycation of HDL apolipoproteins. This in vitro approach enables glycation products to be characterized in the absence of competing reactions and is considered to be a surrogate for 'lifetime' in vivo exposure of HDL apolipoproteins to reactive α-oxoaldehydes.
In addition to the functional changes outlined earlier in the section 'Effect of diabetes mellitus on HDL function', non-enzymatic glycation inhibits the ability of APOA1 to activate LCAT, which can impair the reverse cholesterol transport pathway. The reduced capacity of non-enzymatically glycated APOA1 to activate LCAT has been attributed to a change in its conformation that restricts access to the LCAT activation site as well as to competitive inhibition of LCAT-mediated cholesterol esterification. Non-enzymatic glycation of HDL apolipoproteins might also exacerbate atherosclerotic lesion formation by increasing the activity of hepatic lipase and accelerating free fatty acid release, reducing the anti-inflammatory function of HDLs and increasing HDL degradation.
Glycation of APOA2, APOCI and APOE has also been reported in people with diabetes mellitus, and the glycation of APOE is isoform-specific. The main functional consequence of these modifications is impaired receptor-mediated lipoprotein uptake and impaired binding of very-low-density lipoproteins to cell-associated heparin. The interaction of glycated APOE in HDLs with heparin is also impaired.
APOA4 has anti-diabetic properties. The effect of non-enzymatically glycated APOA4 on HDL metabolism in diabetes mellitus has not been investigated systematically but it is positively correlated with coronary artery disease burden and atherosclerotic lesion progression in humans and in APOE-deficient mice. Non-enzymatic glycation does not affect the capacity of lipid-free APOA4 to accept cholesterol from bone marrow-derived macrophages that have been activated by lipopolysaccharides, and non-enzymatically glycated APOA4 reduces expression and secretion of pro-inflammatory cytokines in these cells. Lipid-free, non-enzymatically glycated APOA4 increases inflammation in aortic endothelial cells in vitro; however, whether this is also the case for HDL-associated APOA4 is not known.
Non-enzymatic glycation of the minor HDL apolipoprotein, APOC1, which inhibits CETP activity, also has functional consequences. For example, it decreases the ability of APOC1 to inhibit CETP activity, and this reduction is not corrected by improving glycaemic control in patients with T1DM. This reduction might be due to the irreversible AGE-mediated cross-linking of APOC1, which prevents APOC1 from blocking binding of HDLs to the N-terminal domain of CETP, which is the initiating step for cholesteryl ester and triglyceride transfers between different lipoprotein classes.
As knowledge of the effect of non-enzymatically glycated apolipoproteins on HDL function is limited, further studies are necessary before the implications of these modifications are fully understood. Investigations into the structural and functional consequences of non-enzymatic glycation of APOA2, APOC1, APOE and APOA4 could provide insights into the progression of cardiovascular diseases and identify therapeutic targets with the potential to slow disease progression and improve management of established disease.