Paraoxonase 1 concerning dyslipidaemia, cardiovascular diseases, and mortality in haemodialysis patients

Paraoxonase 1 (PON1) is known for preventing atherosclerosis through lipid-modifying features, antioxidant activity, anti-inflammatory, anti-apoptosis, anti-thrombosis, and anti-adhesion properties. Uremic patients requiring haemodialysis (HD) are especially prone to atherosclerosis and its complications. We analysed the PON1 gene (PON1) polymorphisms and serum PON1 (paraoxonase) activity concerning dyslipidaemia and related cardiovascular diseases and mortality to show how they associate under uremic conditions modified by maintenance HD treatment. The rs662 AA + AG (OR 1.76, 95%CI 1.10–2.80, P = 0.018), rs854560 TT (OR 1.48, 95%CI 1.04–2.11, P = 0.031), and rs854560 AT + TT (OR 1.28, 95%CI 1.01–1.63, P = 0.040) contributed to the prevalence of atherogenic dyslipidaemia diagnosed by the triglyceride (TG)/HDL-cholesterol ratio ≥ 3.8. The normalized serum PON1 activity positively correlated with atherogenic dyslipidaemia (ẞ 0.67 ± 0.25, P = 0.008). The PON1 rs854560 allele T was involved in the higher prevalence of ischemic cerebral stroke (OR 1.38, 1.02–1.85, P = 0.034). The PON1 rs705379 TT genotype contributed to cardiovascular (HR 1.27, 95% CI 1.03–1.57, P = 0.025) and cardiac (HR 1.34, 95% CI 1.05–1.71, P = 0.018) mortality. All P-values were obtained in multiple regression analyses, including clinical variables. Multifaceted associations of PON1 with dyslipidaemia, ischemic cerebral stroke, and cardiovascular mortality in HD patients provide arguments for the consideration of PON1 and its protein product as therapeutic targets in the prevention of atherosclerosis and its complications in uremic patients.

). Table S3 demonstrates the expected sample sizes and odds ratios (ORs) for the likelihood of securing significance at 80% sample power in association analyses between PON1 SNVs and tested phenotypes. ORs lower than 1.6-1.8 could provide significance at sample power below 80%. PON1 SNVs, dyslipidaemia, and related diseases. There were no significant associations between tested PON1 SNVs and parameters of serum lipid profile. Although bearers of the PON1 rs662 low activity allele (A) showed lower serum HDL-cholesterol than the rs662 GG homozygotes (Table 2), this relationship did not persist in the multiple regression model including rs662 AA + AG vs. GG (P = 0.106), age (P = 0.721), male gender (P < 0.000001), diabetic nephropathy (P = 0.310), and lipid-modifying treatment (P = 0.005).  Table S4).

Probability of obtaining significant associations. Supplementary
The triglyceride (TG)/HDL-cholesterol ratio ≥ 3.8, indicating atherogenic dyslipidaemia 38 , was associated with PON1 rs662 (the dominant mode) and rs854560 (the additive mode) ( Table 3). The association of PON1 rs662 with atherogenic dyslipidaemia (P = 0.018) persisted in multiple regression analysis including age (P = 0.974), male gender (P = 0.641), diabetic nephropathy (P = 0.327), and lipid-modifying treatment (P = 0.002). In the model comprising the same clinical variables, PON1 rs854560 TT genotype significantly correlated with atherogenic dyslipidaemia (P = 0.031) together with lipid-modifying treatment (P = 0.015); PON1 rs854560 AT + TT was also associated with atherogenic dyslipidaemia (P = 0.040) with lipid-modifying treatment (P = 0.016) (Supplementary Table S5). PON1 rs705379 did not correlate with the frequency of the TG/HDL-cholesterol ratios ≥ 3.8 (Table 3). Table 4 shows no significant associations between tested PON1 SNVs and CHD or type 1 MI. PON1 rs854560 was significantly associated with ICS (Table 5). In logistic regression, the positive association between the rs854560 T allele and ICS (OR 1.375, 95% CI 1.024-1.847, P = 0.034) was shown together with the annual increase  Serum PON1 and patient data. In unadjusted analyses, the lower serum PON1 activity, the higher frequency of mixed dyslipidaemia by K/DOQI guidelines 39 , and the higher serum TG levels ( Table 6). The PON1/ HDL-cholesterol ratio correlated positively with male sex, cigarette smoking, presence of atherogenic dyslipidaemia, and values of the TG/HDL-cholesterol ratio ( Table 7). After adjustment for gender, cigarette smoking, urine output, living in a rural area, and serum phosphorus, significance persisted between normalized serum PON1activity and male sex, cigarette smoking, and atherogenic dyslipidaemia (and the TG/HDL-cholesterol ratio), as well as appeared for living in rural areas, urine output, and zinc supplementation.      If cardiovascular survival probability was compared between genotypes of PON1 rs705379 (TT vs. CT vs. CC), the log-rank test revealed a significant difference (P = 0.025). Homozygotes TT of PON1 rs705379 revealed lower cardiovascular survival than the C allele bearers (Fig. 1a). A significance was shown for cardiac deaths in the recessive mode of inheritance (Fig. 1b) and mortality from CHD and its complications between the TT and CC genotypes (Fig. 1c) but not for vascular deaths, also if deaths from ICS were analysed separately (n = 98). These significances yielded P < 0.05 in the Cox proportional hazards models including age, gender, diabetic nephropathy, and rs705379 TT vs. CC + CT (HR 1.269, 95% CI 1.030-1.565, P = 0.025 for cardiovascular mortality; HR 1.343, 95% CI 1.053-1.713, P = 0.018 for cardiac mortality). Significantly associated with cardiovascular and cardiac mortalities were also age (HR 1.013, 95% CI 1.006-1.021, P = 0.006 for cardiovascular; HR 1.014, 95% CI 1.004-1.023, P = 0.004 for cardiac mortality) and diabetic nephropathy (HR 1.366, 95% CI 1.134-1.645, P = 0.001 for cardiovascular; HR 1.278, 95% CI 1.028-1.589, P = 0.027 for cardiac). In the Cox proportional hazards model, the difference between the PON1 rs705379 TT and CC genotypes concerning mortality from CHD and its complications was connected only with the TT homozygosity (HR 1.549, 95% CI 1.064-2.255, P = 0.023).
There were no associations between rs662 and rs854560 and cardiovascular, cardiac, and vascular mortality in HD patients.

Discussion
HD patients are burdened with dyslipidaemia, CHD, MI, and ICS more frequently than the general population [40][41][42] . This study shows several correlations between PON1 genetic polymorphisms (missense variants rs662 and rs854560, and promoter variant rs705379) and circulating functional protein PON1 concerning atherosclerotic phenotypes. Only associations, which remained significant among crucial demographic and clinical variables, were considered as having a meaningful contribution to dyslipidaemia and atherosclerotic complications.
The involvement of PON1 SNVs in dyslipidaemia, cardiovascular diseases, and mortality of the examined HD subjects may be summarized as follows: Low activity alleles of PON1 rs662 (A) and rs854560 (T) contribute to the higher prevalence of atherogenic dyslipidaemia diagnosed by the TG/HDL-cholesterol ratio ≥ 3.8 but are not related to untransformed parameters of serum lipid profile and dyslipidemic patterns established by K/ DOQI for kidney disease patients 39 . The PON1 rs854560 T allele is involved in the higher prevalence of ICS. The PON1 rs705379 genotype composed of low activity alleles (TT) contributes to cardiovascular mortality, cardiac deaths, and CHD mortality and its complications. The protein product of PON1 is paraoxonase. Normalized serum PON1 (paraoxonase) activity positively associates with atherogenic dyslipidaemia. All mentioned above findings are consistent, although not all previously found associations (with CHD, MI) were revealed in HD patients at a statistically significant level. www.nature.com/scientificreports/ Associations between PON1 SNVs and serum lipid parameters or dyslipidaemic profiles were evaluated among possible confounding variables like age, gender, diabetic nephropathy, and lipid-modifying treatment. It was shown that simvastatin increased the PON1 promoter's activity in a dose-dependent manner in expression cassettes transfected into HepG2 cells 29 . In HuH7 cells, statins (pravastatin, simvastatin, fluvastatin) downregulated the PON1 promoter by 30-50% in a dose-dependent manner. In contrast, the PON-1 secreted enzymatic activity and mRNA levels were increased by the active form of fenofibrate, fenofibric acid, approximately 70% 16 . In HD patients, statin treatment was independently positively related to PON1 concentration 12 . However, in another study on HD subjects, there were no significant differences in PON1 activity that could be related to statins 43 . In our study, the results of associations between PON1 SNVs and atherogenic dyslipidaemia seem to be not influenced by lipid-modifying therapy. Still, the correlation between rs662 and serum HDL-cholesterol became not significant in the model, including lipid-modifying treatment.
Studies on non-uremic subjects are showing higher CHD risk associated with the PON1 rs662 A (Q192) and the PON1 rs854560 T (M55) alleles, which correspond with the low activity PON1 isoform 44,45 . On the other hand, the PON1 rs662 G (192R) and the PON1 rs854560 A (L55) alleles, which are related to the high activity PON1 isoform, were also found to be associated with atherogenic serum lipid profile 46 and common in atherosclerotic diseases (CHD, ICS) in the general population 47,48 . The PON1 rs662 G (192R) and the PON1 rs854560 A (L55) variants were suspected to be associated with CHD, particularly in diabetes mellitus, cigarette smoking, and older age 24,[49][50][51] . There are also studies not indicating the association between the PON1 rs662 and CHD 52 . In the examined HD patients, among three PON1 SNVs (rs705379, rs854560, rs662), being in a weak LD, none showed a significant relationship with CHD or type 1 MI. In studies by Garin et al. 24 and Imai et al. 47 on nonuremic patients, which documented the associations of rs854560 and rs662 with CHD, respectively, criteria for the CHD diagnosis were much more spectacular than in our study: transmural MI or positive coronary angiogram 24 or over 50% narrowing in at least one major coronary artery 47 . The analysis of HD cases with CHD diagnosed by such criteria could provide different results than currently presented.
In Polish HD patients, we revealed the association only between PON1 rs854560 and ICS in the dominant mode of inheritance. In the Han Chinese population without evident kidney diseases, rs705379 was significantly associated with ICS, also in the dominant way of inheritance 53 .
PON1 rs662 and PON1 rs854560 SNVs were associated with serum PON1 activity in HD patients in the previous studies. In Portuguese (n = 183) 43 and Hungarian (n = 20) 54 HD patients, the highest PON1 activity was shown in the PON1 rs662 GG homozygotes, the lowest-in the AA homozygotes. Concerning PON1 rs854560 in HD patients, Portuguese AA homozygotes showed the highest PON1 activity, whereas TT homozygotes-the lowest 43 . Japanese HD patients 14 , like Polish HD subjects in this study, did not differ in PON1 activity concerning PON1 rs854560 polymorphisms. To our knowledge, PON1 − 108C>T (rs705379) polymorphism was not examined in HD patients, but in healthy Japanese subjects, there was a decrease in PON1 activity for − 108T individuals when compared to those with the − 108C polymorphism 16 . Therefore, our study did not reveal significant differences in serum PON1 activity and the PON1/HDL-cholesterol ratio concerning PON1 SNVs previously associated with lower serum activity of anti-atherogenic PON1 1,43,54 . However, the low-expression alleles/genotypes of these SNVs corresponded with atherogenic dyslipidaemia (rs662, rs854560), ICS (rs854560), or cardiovascular mortality (rs705379). It places them among unfavourable genetic inheritance.
Serum  56 . PON1 activity was not associated with diabetes (diabetic nephropathy) in the examined HD patients, HD subjects studied by Varga et al. 57 , and Itahara et al. 58 . In another study, non-diabetic HD subjects showed higher PON1 activity than type 2 diabetics 59 .
Only the univariate analyses indicated inverse correlations between serum PON1 activity and mixed dyslipidaemia by K/DOQI and serum TG concentration. In HD patients, serum PON1 (concentration or activity) was already associated with serum lipids, mainly HDL-cholesterol and HDL subclasses 60,61 . HD patients showing CHD had lower PON1 activity than subjects without CHD 62 . In HD men but not in HD women, PON1 concentrations were lower in CHD subjects than in those without CHD, although serum PON1 concentration was not gender-dependent 13 . The normalized serum PON1 activity in healthy men was the highest in subjects with the lowest HDL-cholesterol values 63 . In our study, the PON1/HDL-cholesterol ratio positively correlated with the male gender. However, serum PON1 activities were similar in subjects with and without studied cardiovascular diseases, independently on gender (Supplementary Table S6). Although our results indicate that serum PON1 activity attenuates atherogenic serum lipid patterns, circulating PON1 was not associated with already developed comorbidities related to atherosclerosis (CHD, MI, ICS).
Numerous factors influencing serum PON1 activity may mask its association with the genetic background or clinical and laboratory phenotypes. Erythropoietin, widely used in HD subjects, was shown to elevate PON1 activity in predialysis patients 31 . In HD patients, zinc supplementation 32 or decaffeinated green tea extract (catechins) 33 increased the activity of PON. Nandrolone decanoate decreased PON activity in HD patients 34 . Acrolein, both air pollution and cigarette smoke component 36 , inactivates PON1 37 . The acrolein levels are usually low in outside air (0.20 ppb in urban air and 0.12 ppb in rural air). However, in large cities, acrolein pollution reaches 5.6 ppb. 36 . Thus, places of settlement (city, village), different in air pollution, may influence PON1. Age 13,58 , dialysis duration 30,58,64 , serum urea and creatinine 30,58 , vitamin C administration 33,65 , and body mass index 57,66 yielded ambiguous results concerning their influence on PON1 activity. Lipid-modifying treatment 12,16,29 , advanced glycated end-products 61 , and C-reactive protein (CRP) level 66 are also mentioned among possible modifiers of PON1 activity. An impact of secondary hyperparathyroidism on PON1 is also possible. Paraoxonases have two binding-calcium sites. One of two calcium atoms and a phosphate ion lie at the bottom of the PON1active-site cavity 67  www.nature.com/scientificreports/ inactivation of PON1 activity. Human serum PON1 requires calcium for enzymatic activity, and calcium is needed for maintaining PON structural stability 35 . This study checked several factors mentioned above for association with serum PON1 activity and the PON1/ HDL-cholesterol ratio. In HD patients, paraoxonase concentration was not associated with gender in multiple regression analysis 13 . Our study did not reveal a relationship between serum PON1 activity and gender, but such an association was demonstrated for normalized PON1 activity (the PON1/HDL-cholesterol ratio). Male gender and cigarette smoking, being traditional risk factors for CHD and MI 68 , were positively related to the higher PON1/HDL-cholesterol ratios. Unexpectedly, village settlement also appeared to be positively associated with the higher PON1/HDL-cholesterol ratio. Maybe unhealthy rural residents (HD patients) prefer living in indoor air that contains more acrolein than outdoor air (< 0.02 to 12 ppb but can be higher if residents smoke tobacco at home) 36,69 . On the other hand, positive correlations of the PON1/HDL-cholesterol ratio with village settlement as well as zinc supplementation could be related to their relatively higher positive influences, if any, on PON1 activity than on serum HDL-level. Correlations between these variables need further studies. In HD patients, the higher urine output represents better residual renal function and is generally connected with healthier cardiovascular status 70,71 . It is well established that the decreased kidney function is associated with reduced basal and stimulated PON1 activity 72 . Following this finding is the inverse association of the PON1/HDL-cholesterol ratio with urine output in the examined HD patients. Calcium-phosphorus parameters, disturbed in the examined HD patients according to serum levels of calcium, phosphorus, and parathyroid hormone, did not correlate with serum PON1 at the taken method of the significance validation.
Ikeda et al. 14 have found that PON1 concentration (but not paraoxonase activity or PON1 rs662 and PON1 rs854560) was involved in cardiovascular mortality in HD patients. In the examined HD patients, the PON1 rs705379 TT genotype was associated with cardiovascular mortality, specifically with cardiac deaths and mortality due to CHD and its complications. PON1 rs705379 is the main contributor to serum PON1 variation, accounting for about 13% of the disparity in arylesterase activity 73 . In the study by Gupta et al. 74 , the PON1 rs705379 CT and TT genotypes corresponded in non-diabetics with the lower PON1 activity and CHD. However, this significance did not persist in multiple regression analysis. The TT genotype was also associated with low serum PON1 activity and an increased CHD risk in type 2 diabetic patients 75 . In the examined HD patients, associations between rs705379 and CHD were not demonstrated and did not explain cardiac deaths. The TT genotype of PON1 rs705379 independently correlated with type 2 diabetic nephropathy as a cause of end-stage renal disease 76 . This study shows that the TT rs705379 genotype and type 2 diabetic nephropathy are independent predictors of cardiovascular (and cardiac) mortality in HD patients. Maybe, attenuated antioxidant, anti-inflammatory, anti-apoptosis, anti-thrombosis, and anti-adhesion properties of PON1 2-4 are involved in overall cardiovascular mortality, not only in deaths due to CHD and its complications.
Our study indicates that PON1 SNVs correlate with HD patients' clinical parameters (dyslipidaemia, ICS, cardiovascular mortality). These data suggest that abnormalities resulting from long-term atherosclerotic disturbances, like ICS or cardiovascular deaths, are satisfactorily associated with PON1 SNVs in HD patients. Associations of non-normalized serum PON1 activity with simultaneously shown clinical variables were limited to mixed dyslipidaemia by K/DOQI guidelines and serum TG levels in unadjusted analyses. Several studies 21,[77][78][79][80] have demonstrated the usefulness of the PON1 status in the evaluation of clinical PON1 associations. The PON1 status includes simultaneous determination of paraoxonase and diazoxonase activities 21,77-79 or PON1 activity and concentration 80 . Futurę studies incorporating the PON1 status are warranted in establishing PON1 relationships in end-stage renal disease patients treated with HD.
Our study shows multifaceted associations of PON1 with dyslipidaemia, ICS, and cardiovascular mortality in HD patients providing arguments for the consideration of PON1 as a therapeutic target in the prevention of atherosclerosis and its complications in uremic patients. At present, by implementing therapeutic lifestyle changes (aerobic exercises) and niacin, at least in men with metabolic syndrome, we can increase PON1 activity and PON1 concentration 81 . In the future, gene therapy may be a solution retaining multifunctional PON1 capacity.

Conclusions
1. In HD patients, there are associations of PON1 SNVs with the prevalence of atherogenic dyslipidaemia diagnosed by the TG/HDL-cholesterol ratio ≥ 3.8 (higher for low activity alleles of rs662 and rs854560), a frequency of ICS (higher for the rs854560 low activity allele), and cardiovascular mortality, specifically with cardiac deaths, as well as mortality due to CHD and its complications (higher in PON1 rs705379 low activity homozygotes). 2. The normalized serum PON1 activity (the PON1/HDL-cholesterol ratio) positively associate with atherogenic dyslipidaemia, male gender, and cigarette smoking, while a negative correlation occurs with urine output. 3. Correlations between the PON1/HDL-cholesterol ratio, living in the rural area, and zinc supplementation need further studies. 4. Multifaceted associations of PON1 with dyslipidaemia, ICS, and cardiovascular mortality provide arguments for the consideration of PON1 as a therapeutic target in the prevention of atherosclerosis and its complications in uremic patients. www.nature.com/scientificreports/ excluded. In total, we included 1407 patients. In all these subjects, we recorded demographic and clinical parameters, including dates of birth, the start of RRT, and death, gender, cause of end-stage renal disease, and cause of death, as appropriate. Concerning diabetes mellitus, we included only patients with type 2 diabetes as a cause of end-stage diabetic nephropathy. Evidence for dyslipidaemia, lipid-modifying therapy, CHD, including MI, and ICS was also gathered in all patients, if possible. To be enrolled in the dyslipidaemia study, patients had to possess serum lipid data obtained when they presented stable clinical status at least six weeks before serum lipid measurement and did not have the following exclusion criteria: blood or plasma transfusion as well as more significant surgery during three months preceding blood sampling for lipids. The demographic, clinical, and laboratory data and the mode of lipid-modifying therapy were taken at the time of blood collection for lipids. The patients' treatment was provided by nephrologists in the dialysis centres and was not influenced by the study. CHD, MI, and ICS data we gathered using the entire available period of RRT. Data were collected and identified by patients' names and surnames except from results obtained in two dialysis facilities where the code system existed.

Patients and methods
From the cohort mentioned above, we enrolled patients for testing PON1 activity. Therefore, these subjects fulfilled the enrolment criteria described for the entire HD population genotyped for PON1 SNVs. Blood samples for serum PON1 activity were collected in two collaborating dialysis centres for adult patients (≥ 18 years). In one centre, low-flux dialyzers were used; in another one-high flux dialyzers were applied. Patients were dialyzed in May-June 2019, three times a week. They were randomly selected from currently available stable patients who gave written consent (n = 93).
In patients tested for serum PON1 activity, we additionally recorded a place of settlement (city, village), a history of parathyroidectomy, actual dry body mass, ESA administration, treatment with cinacalcet hydrochloride and phosphate binders, cigarette smoking, zinc supplementation, green tea intake, vitamin C and D supplementation, and nandrolone decanoate usage. Blood samples for serum PON1 activity were collected in these 93 patients together with routine laboratory parameters, which included serum concentrations of lipids, urea, creatinine, haemoglobin, CRP, total Ca, P, intact parathyroid hormone, and activity of liver enzymes.
All study subjects were Caucasians. We diagnosed dyslipidaemia according to the K/DOQI guidelines, which were elaborated taking into account the specificity of lipid abnormalities in chronic kidney disease subjects, including HD individuals 39 . Patients diagnosed as dyslipidaemic by serum LDL-cholesterol ≥ 100 mg/dL, we referred to as hyper-LDL-cholesterolaemic.
Those showing non-HDL-cholesterol ≥ 130 mg/dL and TG ≥ 200 mg/dL were described as hyper-TG/hypernon-HDL-cholesterolaemic. HD subjects showing dyslipidaemia by serum LDL cholesterol ≥ 100 mg/dL and simultaneously occurring non-HDL cholesterol ≥ 130 mg/dL and TG ≥ 200 mg/dL are referred to as having mixed dyslipidaemia 82 . The remaining patients were diagnosed as non-dyslipidaemic by K/DOQI criteria. We used the atherogenic index (the TG/HDL-cholesterol ratio) to interpret the atherogenic pattern of dyslipidaemia. The ratio of ≥ 3.8 was considered as indicating atherogenic dyslipidaemia because this ratio was reliable for identifying atherogenic LDL phenotype B in men and women 38 . HD subjects with the TG/HDL-cholesterol ratio < 3.8, we described as patients without atherogenic dyslipidaemia.
CHD was diagnosed based on medical history, electrocardiograms, exercise stress tests, and, in some cases, coronary angiography or computed tomography. From CHD patients, we selected subjects who underwent MI, diagnosed using medical history, characteristic electrocardiographic abnormalities, and elevated levels of cardiac markers of cardiomyocyte damage. Patients with ST-elevation MI and non-ST-elevation MI were included in this study. Type 1 MI 83 was recognized in all of them. Clinical data and computed tomography were used for the determination of ICS.
Blood sampling. Blood probes for PON1 genotyping and serum PON1 activity were taken before a midweek dialysis session when HD patients had collected blood for routine periodical laboratory testing. Monovette tubes (SARSTEDT, Nümbrecht, Germany) were used for venous blood sampling. Tubes containing the EDTA anticoagulant were applied for DNA analyses and blood morphology. If serum was needed, blood samples were drawn to Monovette tubes allowing the blood clotting (no anticoagulant). PON1 activity, cholesterols, TG, creatinine, urea, CRP, albumin, calcium, phosphorus, alkaline phosphatase (ALP), and parathyroid hormone were determined in serum. PON1 genotyping. DNA was extracted from blood lymphocytes by the salt-out extraction method. Coded DNA samples were stored at -75 ° C and genotyped.
PON1 SNVs designated as rs662 (Q192R, 575A>G), rs854560 (L55M, 163A>T), and rs705379 (− 108C>T) were selected for genotyping. PON1 rs662 was genotyped using a high-resolution melting curve analysis on the Light Cycler 480 system (Roche Diagnostics, Germany). Analysis of PON1 rs854560 and rs705379 was performed using predesigned TaqMan SNV Genotyping Assay according to the manufacturer's instructions provided by Applied Biosystems (Applied Biosystems, Foster City, CA). Serum PON1 activity. We determined circulating PON1 activity assessing catalytic efficiency for paraoxon hydrolysis (paraoxonase activity) with the use of a commercially available kit produced by Rel Assay Diagnostics, Mega Tıp (Gaziantep, Turkey, REF: RL0031, LOT: NN19064P), also applied by other investigators in clinical studies 84,85 . Two reagents are used in this kit: one contains a Tris buffer and Ca ion, the second-a stabile substrate solution. According to the manufacturer, the paraoxonase assay coefficient of validation (CV%) was 4.1 for high activity sera pool, 1.7-for medium activity sera pool, and 1.5-for low activity sera pool.  Germany). Blood morphology parameters were determined by flow cytometry (Sysmex, Kobe, Japan).
All methods shown above were routinely used.

Statistical analysis.
To ensure adequate power (80%) for detection of associations between tested SNVs and dyslipidaemia by K/DOQI, atherogenic dyslipidaemia, CHD, MI, or ICS, an HD patients' risk of these diseases was taken into the calculation of the desired sample size at different ORs (Supplementary Table S7). Allele frequencies for these calculations we obtained from gnomAD Exomes, European Non-Finnish. We tested the distribution of continuous variables by the Shapiro-Wilk test. For the presentation of nonnormally distributed variables, we used a median and range (minimum-maximum). Normally distributed continuous results are shown as a mean ± standard deviation. Categorical variables are presented as a percentage of the total number.
Departure from HWE was determined by Chi-squared analysis (df = 1, P ≥ 0.05 for accordance). Modes of inheritance were created concerning alleles known as associated with low PON1 activity or concentration as the risk alleles: A for rs662, T for rs854560, and T for rs705379 86 . To test PON1 SNVs for associations with categorical variables (types of dyslipidaemia, CHD, type 1 MI, and ICS) in the additive genetic model, we used the Cochran-Armitage test for trends 87 . As this model does not indicate a type of inheritance and may appear unpowered, especially when the suitable mode is recessive 88 , we additionally applied the dominant and recessive modes.
The Kruskal-Wallis and Mann-Whitney U tests were used for the comparison of continuous variables. Dichotomous variables were compared using Fisher's exact test. OR with 95% confidence interval (CI) and P-values were computed to show significance in the odds of tested genotype occurrence in the case group to the odds of this occurrence in the control group. Genetic associations, yielding the P < 0.05 indicating the fifth class association by the Better Associations for Disease and GEnes (BADGE) system 89 , were verified in the multiple regression analysis or logistic regression, as appropriate, comprising clinical variables (age, gender, diabetic nephropathy, and lipid-modifying treatment when tested variables were associated with serum lipid profile; age, diabetic nephropathy, sex, and RRT duration if the association with ICS was determined).
The linear regression was used to determine the associations among serum PON1 activity or the PON1/ HDL-cholesterol ratio and patient characteristics. The obtained results were adjusted for traits associated with PON1 activity or the PON1/HDL-cholesterol ratio at a P-value < 0.1 in unadjusted analyses. Multiple regression was used for adjustment. Due to a relatively small number of subjects tested for serum PON1 activity, we have chosen five variables (gender, cigarette smoking, urine output, living in a rural area, and serum phosphorus), which could influence a relationship between PON1 activity and the examined phenotypes. In the regression analyses, there was no normality for all included data. However, the outstanding values did not exceed the 3 sigma limit in any case, so we used this statistical approach. The results are presented as the regression coefficient (β) ± standard error (SE).
Survival analyses included patients who died between the start of their RRT and October 21, 2019. We evaluated cardiovascular mortality and separately cardiac and vascular mortalities. Among cardiac mortality, CHD deaths and CHD complications (heart failure, MI, sudden death) were analysed. We also computed deaths separately due to ICS. For survival analysis, we applied the Kaplan-Meyer method with the log-rank test. Mortality was calculated between genotypes and using modes of inheritance. In the case of significance, the Cox proportional hazards model was performed. If the latter was also significant, we applied the Cox proportional hazards model that included age, gender, and diabetic nephropathy.
Linkage disequilibrium (LD) between tested SNVs was estimated using the Haploview 4.2 software (http:// www. broad. mit. edu/ mpg/ haplo view/). P < 0.05 were chosen as the statistically significant to facilitate decisions which data may be relevant for further analyses. Finally, only associations that yielded P < 0.05 also among clinical variables were discussed as significant.
Ethics approval and consent to participate. All patients or their parents, as appropriate, whose DNA samples were collected and are stored in the Department of Biochemistry and Molecular Biology, Poznan University of Medical Sciences, Poznań, Poland, gave a written consent informing that all their data and DNA samples may be anonymously used in the future studies concerning uremia without additional agreements. Informed consent we obtained from all study participants tested for serum PON1 activity. The Institutional Review Board of the Poznan University of Medical Sciences, Poland, approved the research design. The study conformed to the principles set out in the WMA Declaration of Helsinki and the Department of Health and Human Services Belmont Report.