HDL-proteome enrichment with apolipoprotein A-IV compensates for HDL loss of function in diabetes mellitus kidney disease

Background and aims: Diabetes mellitus kidney disease (DKD) is associated with lipid derangements worsening kidney function and enhancing cardiovascular (CV) risk. The management of dyslipidemia, hypertension and other traditional risk factors does not completely prevent CV complications bringing up the participation of untraditional risk factors such as advanced glycation end products (AGEs), carbamoylation and changes in HDL proteome and functionality. We analyzed HDL composition, proteome, chemical modication and functionality in non-dialytic DKD subjects categorized according to glomerular ltration rate (GFR) and urinary albumin excretion rate (AER). Methods:


Background
In diabetes mellitus (DM), abnormal kidney function is one of the most frequent complications being the leading cause of end-stage kidney disease. Besides, kidney function impairment increases the risk of cardiovascular disease (CVD) that is the major cause of mortality in both type 1 (DM1) and type 2 DM (DM2) [1] . Apart from traditional risk factors that are commonly observed in DM, including dylspidemia, hypertension and other components of the metabolic syndrome, untraditional risk factors -advanced glycation, carbamoylation and oxidation -contribute to macrovascular disease in DM kidney disease (DKD) [2][3][4] .
In DKD, the reduction in kidney function represented by the diminished glomerular ltration rate (GFR) is not unvariable accompanied in the same extension by elevation in the urinary albumin excretion rate (AER) [5] . In fact, many individuals with DKD with a marked reduction in GFR can still present normal (A1 stage) ou slightly reduced AER (A2 stage) and in some cases people in the A2 stage can revert to A1 [6] .
The incidence of CVD is positively related to the reduction in GFR as well as AER, and both parameters have additive effects on CV risk in any stage of abnormal kidney function [7] .
Advanced glycation end products (AGEs) are prevalent in DM and in chronic kidney disease (CKD) due to hyperglycemia, oxidative stress and detoxy cation failure of intermediate compounds of the glycation reaction. AGEs are independent predictors of CV risk by impairing the reverse cholesterol transport (RCT) and lipid metabolism, inducing in ammation and altering vasodilation [8][9][10] . In addition, the reaction of isocyanic acid derivated from urea with proteins leads to protein carbamoylation that is also related to atherogenesis [4] .
The reduction of high-density lipoprotein (HDL) cholesterol in plasma is a hallmark in DM although HDL dysfunction is also considered as having an important role in CV morbidity and mortality. This is especially reputable when analyzing clinical trials where the increment in HDL cholesterol did not contribute for CV risk improvement. HDL are antiatherogenic particles that mediate the removal of excess cholesterol from the arterial wall macrophages allowing its tra c to liver and excretion in feces by the RCT. Besides, HDL exert several others atheroprotective actions including antioxidant, antiin ammatory, vasodilation and anti-aggregant and improving glucose tolerance and insulin sensitivity. HDL is a cargo lipoprotein transporting many proteins and microRNAs that are able to control metabolism in different tissues and in the arterial wall [11] .
The HDL proteome that follows its complexity and functionality has been analyzed in CKD associated or not with DM, but in most of the studies, the individuals were on dyalisis that may have in uenced the results. In addition, the strati cation of CKD subjects by GFR or AER alone may not re ect changes in HDL proteome and function that takes place in an unusual evolution of chronic disease typically in DKD. We then analyzed the HDL proteome, composition and chemical modi cation by advanced glycation and carbamoylation, together with its functionality in non-dyalisis DKD subjects categorized according to the GFR and AER in comparison to age-matched control subjects. We found that the HDL proteome was enriched in apolipoprotein D (apo D) and apo A-IV and that HDL was modi ed by advanced glycation and carbamoylation, according to the reduction in GFR and increased AER. The ability of HDL in removing cellular cholesterol was reduced in DKD with GFR < 60 mL/min/1.73 m 2 plus A3 although its antioxidant activity was preserved and its capacity to prevent in ammation in macrophages was even increased.
This may be ascribed to the increment of HDL in apo A-IV that exerts antiatherogenic actions counteracting HDL loss of function that is reported in CKD. Subjects with other chronic diseases, rapid loss in GFR (> 3 mL / min / year), refractory hypertension, BMI < 18.5 Kg / m 2 , current smoke or alcohol abuse were not included. Blood was drawn after overnight fasting and plasma was immediately isolated in refrigerated centrifuge (4°C). Glycemia, triglycerides (TG), total cholesterol (TC), HDL cholesterol (HDLc), fructosamine, TSH, creatinine and urea were determined in plasma by enzymatic techniques after overnight fasting and albumin in 24h-urine. HbA1c was determined by high performance liquid chromatography (HPLC). DKD individuals were categoried according to the GFR above 60 mL/min/1.73 m 2 plus AER stages A1 (< 30 mg/g creatinine) and A2 (30 -300 mg/g creatinine) and GFR below 60 mL/min/1.73 m 2 plus stage A3 (> 300 mg/g creatinine). Control subjects presented GFR above 60 mL/min/1.73 m 2 plus A1.

Isolation of lipoproteins
Venous blood samples were drawn after overnight fasting and plasma immediately isolated in a refrigerated centrifuge. Preservatives were added to the plasma and density adjusted with bromide potassium to 1.21 g/mL. Low-density (LDL; d = 1.019-1.063 g/mL) and high-density lipoprotein (HDL; d = 1.063-1.21 g/mL) were isolated from plasma by discontinous density gradient ultracentrifugation (100 000g, 24 h, 4 °C, Sw40 rotor; Beckman ultracentrifuge). Samples were dialyzed against phosphate buffer saline containing EDTA (PBS).

HDL composition in lipids and apoA-I
The amount of lipids and apo A-I in HDL were determined, respectively, by enzymatic techniques [TC and TG; Labtest diagnóstica S. A., Minas Gerais, Brazil; phospholipids (PL); Randox Laboratorier LTD. Crumlin, Co. Antrem, United Kingdom] and immunoturbidimetry.

Determination of total AGE and pentosidine in HDL
The contents of total AGE and pentosidine were determined by uorescence measurement (Synergy HT Multi-Mode Microplate Reader, SpectraMax M5). Samples were excited at a wavelength of 370nm and the uorescence emitted at 440 nm and 378 nm, respectively, for total AGE and pentosidine.

Proteolytic digestion of HDL
The HDL protein concentration was determined by the Bradford assay (Bio-Rad, Hercules, CA, USA). Ten micrograms of HDL-protein was solubilized in 100 mM ammonium bicarbonate, dithiothreitol, and iodoacetamide, following digestion with trypsin (1:40, w/w Promega, Madison, WI,USA) for 4 h at 37 °C.
Trypsin was further added to samples (1:50, w/w HDL) and the incubation was done overnight at 37 °C. Samples were desalted using solid phase extraction (Oasis PRIME HLB SPE column; Waters) after acidic hydrolysis with 2% tri uoroacetic acid and kept frozen at -80°C until MS analyses. Prior to MS analysis, samples were resuspended in 0.1% formic acid ( nal protein concentration of 25 ng/μL).

Targeted proteomic analyses
Digested HDL proteins (50 ng protein) were quanti ed by parallel reaction monitoring (PRM), as previously described [12] . Brie y, an Easy-nLC 1200 UHPLC (Thermo Scienti c, Bremen, Germany) was used for peptide separation. Acquisition of the data was performed in an Orbitrap Fusion Lumos mass spectrometer (Thermo Scienti c, Bremen, Germany) using a nanospray Flex NG ion source (Thermo Scienti c, Bremen, Germany). A scheduled (3-min window) inclusion list containing m/z of precursor peptides of interest and corresponding retention times was generated using Skyline software [13] .
Selection of HDL peptides for targeted quanti cation PRM methodology was assembled from shotgun proteomics analyses as previously described [12] . Ninety-one proteins were identi ed, but this number was reduced to 47 proteins eliminating potentials contaminant proteins (keratin, proteins with <2 unique peptides and peptides with high interfering signal and mass error >10 ppm). Peptides susceptible to ex-vivo modi cation (e.g., methionine-containing peptides) were also avoided, and only peptides satisfactorily detected were included in the nal analysis.
After our exclusion criteria, 31 proteins remained. For statistical analysis, we considered the best peptide that represents each protein of interest. In order to nd these surrogate peptides, rstly the peptide pair with best Pearson's correlation coe cient was determined. From these 2 peptides, the peptide with the lowest CV was nally selected. The 31 surrogate peptides chosen for HDL proteins are highlighted in Supplementary Table 1.

Acetylation of LDL
LDL was acetylated as previously described by Basu et al [14] . Samples were extensively dialyzed before incubation with macrophages.

Measurement of 14 C-cholesterol e ux
Bone marrow-derived cells were isolated from mice and macrophages were differentiated as previously described [15] . Cells were overloaded with acetylated LDL (50 µg/mL DMEM) and 14 C-cholesterol (0.3 µCi/mL) for 48 h. HDL from controls and DKD subjects (50 µg/mL) were utilized as cholesterol acceptors in 6-h incubations and cholesterol e ux calculated as previous described [8] .

Measurement of the HDL antioxidant activity
The ability of HDL from controls and DKD individuals in inhibiting LDL oxidation was determined by incubation of LDL (40 µg/mL) isolated from a unique healthy plasma donor with CuSO 4 solution (1 mL) in the presence of HDL (80 µg/mL). Lipoproteins were dialyzed against PBS without EDTA prior incubations. The absorbance at 234 nm was continuously monitored every 3 min during 4 h and the lag time phase for LDL oxidation (min) and the maximum ratio of conjugated dienes formation calculated [15] . Measurement of the HDL antiin ammatory activity BMDM were isolated and cultured as previously described; then overloaded with acetylated LDL (50 µg/mL DMEM) and treated for 24 h with HDL from controls and DKD subjects. After washing, macrophages were incubated with lipopolyssacharide (LPS; 1 µg/mL DMEM) for 24 h. Medium was collected and the amount of TNF alpha and interleukin 6 (IL-6) determined by ELISA (R&D System-Duo Set, Minneapolis, EUA) [9] .

Statistical analysis
Statistical analysis was performed using the GraphPad Prism 5 program (GraphPad Software, Inc. 2007).
Comparisons were made by one-way analysis of variance (ANOVA) with Dunnett post-test, Student t test and Sperman linear correlation as appropriated. The value of p < 0.05 was considered statistically signi cant. Proteome data were compared by multiple comparisons.

Results
Anthropmetric and biochemical data of control and DKD subjects are depicted in table 1. In the group GFR <60 + A3 there was a greater predominance of male individuals (72%) compared to the control (50%) and GFR> 60 + A1 and A2 (30%) groups. Age was similar among groups as well as the time of disease comparing both DKD grups. BMI and CVD history were higher in the group GFR > 60 + A1 and A2. Fructosamine and HbA1c levels were higher in DM groups as compared to controls but CML levels did not reached statistically difference among groups. Total cholesterol (TC) was lower in the group GFR > 60 + A1 and A2 as compared to controls.
Eighty percent of DKD subjects were in use of insulin and 70 % were on statins and beta-blockers.
Twenty-nine HDL-associated proteins were quanti ed by PRM using nanoscale liquid chromatography coupled with mass spectrometry (Nano-LC / MS / MS). Of these, 2 were more expressed in the GFR < 60 + A3 group as compared to the control group: apo A-IV ( gure 1 panel A) and apoD ( gure 1 panel B).
A positive correlation was observed between apoA-IV ( gure 2 panel A) and cystatin C (Figure 2 panel B) with the AER. On the other hand, a negative correlation was observed between apoA-I ( gure 2 panel C), serum amyloid protein A-4 (SAA4, gure 2 panel D) and apoC-IV ( gure 2 panel E) with the AER.
The HDL composition in lipids and apoA-I is shown in gure 3. The HDL content in TC ( gure 3 panel A) and TG ( gure 3 panel B) was similar between all groups. There was a decrease in PL ( gure 3 panel C) and apoA-I ( gure 3 panel D) in the HDL of the TFG group <60 + A3 compared to the control group. A positive correlation was observed between HDL-PL (Figure 3 panel E), and apoA-I (Figure 3 panel F) with the GFR.
Total AGE ( gure 4 panel A) and pentosidine ( gure 4 panel B) were determined in HDL, being, for both cases, higher in HDL isolated from individuals with GFR < 60 + A3 (80% and 93.7 %, respectively), as compared to the control group. Although the values of HbA1c and fructosamine were similar between groups, carbonyl stress as a function of albuminuria and GFR < 60 can be attributed to the renal changes that accompany macroalbumininuria and are re ected in the lower detoxi cation of glycation reaction precursors and greater oxidative stress. The modi cation of HDL by isocyanic acid that re ects uremic stress was greater in the group with GFR < 60 + A1 compared to the control group ( gure 4 panel C). A positive correlation was observed between total AGE (Figure 4, panel D) and pentosidine in HDL (Figure 4 panel E) with AER.
HDL was utilized as acceptors of 14 C-cholesterol from BMDM. As shown in the gure 5 (panel A), 14 Ccholesterol e ux mediated by HDL isolated from the TFG < 60 + A3 group was lower as compared to that mediated by the HDL from the control group. The antioxidant activity of HDL was determined by mesuring LDL oxidation with CuSO 4 along time. The lag phase for LDL oxidation determined by the presence of HDL was similar among all groups ( gure 5 panel B) as well as the maximum ratio of conjugated dienes formation in LDL ( gure 5 panel C). In the gure 5 (panel D and E), it is demonstrated the secretion of in ammatory cytokines, IL6 (panel D) and TNF alpha (panel E) in macrophages treated with HDL from controls and DKD patients with GFR < 60 + A3 groups and after challenged with LPS. For both cytokines, it was observed a very lower secretion when macrophages were exposed to DKD HDL in comparison to C-HDL.

Discussion
The prevalence of kidney complications in DM is high and the AER together with the reduced GFR independently and sinergistically predict CV morbidity and mortality [16] . Changes in the HDL proteome and functionality may modulate the antiatherogenic actions of this lipoprotein and consequently the development of atherosclerosis [17] . In this study, we evaluated in DM2 patients with DKD, the composition, chemical modi cation and proteomics of the HDL particles and their ability to remove cholesterol from macrophages, inhibit LDL oxidation and macrophage in ammation.
Targeted proteomics quanti ed 29 proteins in HDL of all experimental groups, although only 2 were differentially expressed in DKD with GFR < 60 plus A3. Apo D is an atypical HDL apolipoprotein closely related to retinol-binding protein. It contributes for the HDL hydrophobic nucleus remodeling by facilitating the lecithin-cholesterol acyltransferase (LCAT) anchoring to the lipoprotein structure and carrying lysophosphatidylcholine, although it is not clear whether it has the potential to activate or inhibit the enzyme [18] . Some studies show an increase of apoD in the HDL proteome of individuals with coronary artery disease and in areas of human atherosclerotic lesion as well as in apolipoprotein E knockout mice plasma [18,19,20] . However, it is not clear yet whether this increase in apoD in these conditions refers to its role in inducing atherosclerosis or whether it represents a compensatory adaptive mechanism to changes observed in cardiovascular disease [18] .
Several studies have demonstrated that apoA-IV can be used as an early marker of kidney failure in individuals with CKD and in the general population, although further studies are needed to understand the pathophysiological basis of this association [21] . Besides, in individuals on hemodialysis, the increment in apoA-IV was associated with an increased risk of all causes mortality [22] . Our data agree with these ndings and an increased apoA-IV expression was observed in the HDL proteome of individuals with higher AER and reduced GFR as compared to controls. Also, apoA-IV was positively correlated to AER. This apolipoprotein is synthesized in the intestine and secreted in the mesenteric lymph, being transported by chylomicrons but mainly by HDL, being the third most abundant apolipoprotein in this lipoprotein [21,22] . ApoA-IV presents many anti-atherogenic functions, acting in the removal of cellular cholesterol [23] and promoting activation of lipoprotein lipase [24] , LCAT and cholesteryl ester tranfer protein (CETP) [25] . In addition, its anti-atherogenic activity is complemented by its anti-in ammatory and antioxidant properties [26,27] .
Similarly to other HDL-apolipoproteins, apoA-IV is modi ed by advanced glycation in DM and CKD.
Recently, our group demonstrated that E. coli recombinat apo A-IV submitted to advanced glycation in vitro mantains its ability in removing macrophage excess cholesterol, despite its large amount pyrraline, CML and argypyrimidine [28] . This may explain the fact that no major reductions in cholesterol e ux were observed in the present study, even with a signi cant increase in total AGE and pentosidine in the HDL of the group GFR < 60 + A3 as compared to controls. In addition, the in vitro glycation of apo A-IV only partially impaired its ability to inhibit the in ammatory response promoted by LPS in macrophages [280] .
On the contrary, apoA-I has its ability in removing cell cholesterol and antioxidant and anti-in ammatory properties severely impaired by advanced glycation [29] .
The HDL proteome has been analyzed in CKD, but the vast majority of studies were performed in individuals undergoing hemodialysis, which can interfere with the observed results. Also, many studies included CKD together with DKD indistinctively. Mangé et al [30] found 40 differently expressed proteins in HDL from CKD subjects. ApoC-II and apoC-III were higher while transtirretin and haptoglobin-related protein (HPTR) were lower in HDL from CKD subjects on hemodialysis compared to healthy subjects. In another study, individuals on hemodialysis, with or without DM, an increase in SAA1, albumin, phospholipase A2 and apoCIII was observed in HDL proteome. Along with the reduction in the content of PL and an increase in TG and lysophospholipids, these modi cations were linked to the reduction in the HDL-mediated cholesterol e ux [31] . In agreement HDL dysfunction was also related to changes in HDL proteome by Florens et al [32] . Shao et al [33] found proteins related to renal injury (beta 2 microglobulin, complement factor D, cystatin C, prostaglandin-H2 D-isomerase, retinol-binding protein 4 and AMBP) incfeased in HDL of individuals with CKD.Others were more present in the control group, among them: apoA-I, apoA-II, apoL-I, apoM and paraoxonase 1, confering greater damage to the anti-atherogenic activities of HDL in CKD.
Recently, the HDL proteome was analyzed in 191 individuals with DM1, in the Diabetes Control and Complications Trial (DCCT) study. Eight proteins were associated with proteinuria, although only 1 was simultaneously associated with AER and coronary calcium content [34] . Wang et al [35] demonstrated enrichment in 8 proteins related to in ammation and lipid metabolism (serum amyloid A1, A2, and A4; hemoglobin beta, HPTR, CETP, PLTP and apo E) in HDL isolated from individuals in short-term as compared to long-term dyalisis therapy.
Although not quantitatively changed in our proteomics analysis, cystatin C levels in HDL wers positively correlated with the AER. Cystatin is directly linked to impaired renal function and is also considered by some investigators as a marker of coronary atherosclerosis and cerebrovascular events [36] . On the other hand, SAA4 and apoC-IV were inversely correlated with AER. Higher levels of SAA associated to HDL have been shown as correlated with a reduced ability of HDL in removing cell cholesterol [31] and are associated with CV mortality [37] .
ApoA-I determined by immunoturbidimetry was reduced in HDL of the TFG group < 60 + A3 with a negative correlation with AER. Therefore, it is likely that in vivo, cellular cholesterol removal may be even more compromised than that observed in the present study due to the lower content of apo A-I in individuals with established DKD. In our e ux assay HDL concentration was matched by the protein component that is mainly represented by apoA-I, making more di cult to observed differences.
A decrease in the HDL-PL was also observed in the GFR group < 60 + A3 with a positive correlation with GFR. Notably, PLs are positive modulators of cell cholesterol removal, as they facilitate the interaction of HDL with plasma membrane domains that guarantee cholesterol exportation by diffusion and / or mediated by speci c receptors.
Reduced ability in removing cell cholesterol was previously reported in CKD individuals stage 3 and 4 as compared to healthy controls although not independently of age [380] . In this sense, there are studies showing the role of aging on the ability of HDL in mediating cholesterol e ux that may is a bias for many studies dealing with controls and CKD subjects at different ages [39,40] . In dialytic individuals, Yamamoto et al [41] reported a reduced cholesterol e ux mediated by HDL from DM with or without DKD in comparison to controls.
Diferently from our work, there are many studies that measured the cholesterol e ux mediated by the apoB depleted serum. Although, HDL is the only lipoprotein in that serum, there are many other components including albumin, cytokines, haptoglobin and insulin that could affect cholesterol removal.
Besides, the measurement of cholesterol removal mediated by apo B-depleted serum mitigates variations in HDL cholesterol level that are frequently altered in DM and DKD [42,43] . Ganda et al [44] did not nd any alteration in the cholesterol e ux mediated by the apo B-serum isolated from CKD subjects (stages 4 and 5), although the expression of Abca1 in monocytes isolated from those subjects was reduced compromising the cholesterol e ux to apoA-I. Chemical modi cation of HDL by advanced glycation and carbamoylation that was observed in the present study was related to the stage of AER and GFR. In individuals in hemodialysis or peritoneal dialysis a greater amount of pentosidine was found in plasma that correlated to the progression of kidney disease. Besides, pentosidine levels were higher in CKD stage 5 subjects as compared to stage 1 [45] . In CKD subjects undergoing hemodialysis it was demonstrated elevated concentration of carbamoylated HDL, which is dysfunctional impairing the RCT [46,47] .

Conclusion
The antioxidant capacity of HDL assessed by the lag time for LDL oxidation was similar among groups, while the HDL anti-in ammatory ability was greatly increased. This was demonstrated by the reduced secretion of in ammatory cytokines by macrophages challenged by LPS. These results may be ascribed to the enhanced amount of apo A-IV in HDL that seems to compensate for chemical modi cations of HDL that take place in established DKD [48] . The exact role of apoA-IV enhancement in HDL proteome should be investigated in detail in order to clarify its contribution for CVD prevention in DKD. Competing interests: there are no con icts of interest. All authors have read and approved the submission of the manuscript; the manuscript has not been published and is not being considered for publication elsewhere, in whole or in part, in any language.

Funding
The authors would like to thank the nancial support from Fundação de Amparo à Pesquisa do Estado   Figure 4 HDL modi cation by advanced glycation and carbamoylation. The amount of total AGE (panel A), and pentosidine (panel B) was determined in HDL by measuring the absorbance in the uorescence range at 440 nm (total AGE) and 378 nm (pentosidine) and carbamoylation (panel C), by ELISA. The results were compared by one-way ANOVA with Dunnett's post-test. Associations between HDL chemical modi cation with AER (panels D and E) were performed by Spearman´s correlation.