We studied 30 ESRD patients on hemodialysis (HD) and 15 healthy control individuals. Among the patients, 15 were diabetic (self-reported diabetes mellitus and antidiabetic oral (20 %) or parenteral (80 %) medication). All subjects were studied cross-sectionally at one visit. The patients were selected from a single German non-profit out-patient dialysis center with ESRD for at least 3 months and an age above 18 years with stable sinus rhythm. Exclusion criteria were clinical signs of acute infection, active malignancy, CRP above 50 mg/L, presence of a cardiac pacemaker, atrial fibrillation, heart transplantation and immunosuppressive medication. Healthy controls had no known history of renal or heart disease and values within the normal range for serum creatinine, urea, CRP, BNP, pH and bicarbonate.

Patients and control individuals were assessed for age, body height, body weight and waist-to-hip ratio. A resting blood pressure was determined using an automated oscillometric system (dinamap, GE Systems). Patients were interviewed for dialysis vintage and residual diuresis and data were consolidated using the patient’s files. Ultrafiltration volume was taken from the dialysis monitor readings and the ultrafiltration rate was calculated from the ultrafiltration volume and the effective dialysis time.

The following parameters were measured by routine laboratory methods in an accredited diagnostic laboratory from blood samples drawn before the first dialysis of the week: creatinine, urea, cholesterol, HbA1c, CRP, interleukin-6 (IL-6), Bicarbonate, brain natriuretic peptide (BNP) and an automated blood differential count. HRV analysis was performed once within 2 weeks before to 2 weeks after the blood samples.

For analysis of monocytes, lymphocytes and B-cells, lithium heparinized whole blood was washed twice in phosphate-buffered saline/2 mM EDTA/0.5 % bovine serum albumin/0.07 % sodium azide. Monocyte staining was performed using anti-CD86 PeCy5 (clone IT2.2, Beckman Coulter, Krefeld, Germany), −CD14 PeCy7 (clone MPhi9, Becton Dickinson, Heidelberg, Germany), −CD143 FITC (clone 9B9, AbD Serotec, Düsseldorf, Germany), −CD16 APC (clone CB16, eBioscience), −CD19 (clone SJ24C1, eBioscience) and -CD15 eF450 (clone HI98, ebioscience) antibodies. Lymphocyte and B-cell staining comprised anti-CD3 FITC (clone BW264/56, Miltenyi Biotec, Bergisch-Gladbach, Germany), −CD56 V450 (clone B159, BD, Heidelberg, Germany), −CD19 APCeF780 (clone HIB19, ebioscience). For identification of AChR on peripheral blood leucocytes nicotinic acetylcholine receptor alpha subunit (clone G10, Biozol, Eching, Germany) was used. Specificity of this antibody has been evaluated earlier [27‐29]. The FACS staining protocols included fluorescence-minus-one-controls. To take unspecific binding by IgG into account, isotype controls were used to allow for better correction of gating. Specifity of staining was confirmed by use of isotypic IgG1 staining (Biozol, Eching, Germany). R-Phycoerythrin labelling of AChR as well as the corresponding Isotype control (IgG) was done using the LYNX RAPID RPE Antibody Conjugation Kit® (Abd Serotec, Düsseldorf, Germany).

By flow cytometry (MACSQuant, Miltenyi Biotech, Bergisch-Gladbach, Germany), monocytes were defined as CD15-CD19-CD86+. Subpopulations were divided according to the expression of CD14 and CD16 (CD14++CD16-: Mo1, CD14++CD16+: Mo2, CD14 + CD16+: Mo3). Absolute cell numbers were calculated from the relative numbers detected by flow cytometry and the absolute monocyte numbers detected by automated blood differential count. Expression density of ACE (CD143) was analyzed by measuring the median fluorescence intensity on Mo2 cells. Expression density of the AChR was analyzed by measuring the median fluorescence intensity on all cells that were positive for either CD14, CD3, or CD19. The gating strategy included doublets exclusion and characterisation of leukocyte subsets by staining with specific markers, i.e. anti-CD3 for T-cell, anti-CD19 for B-cell and anti-CD86 for monocyte gating (Figs. 2, 3 and 4).

A 12-lead resting 20-min ECG was recorded on a CardioControl Working Station (Welch Allyn, Delft, the Netherlands) after a resting period in the supine position of ≥20 min. Throughout the ECG, subjects were asked to breathe at 15 breaths/min guided by a visual metronome to standardize the influence of the respiratory rate on spectral HRV parameters. It is known that naturally occurring fluid shifts in dialysis patients within the interdialytic interval may influence ECG recordings. For example, it has been demonstrated that the LF/HF ratio changes over time when measured before, during, or after a dialysis session [16]. Therefore, ECG recordings were executed after the hemodialysis sessions with all patients being at dry weight. For the same reason, 20-min ECG-duration was chosen instead of conventional 24 h recordings. Although the use of time domain parameters has been recommended by the Task Force of The European Society of Cardiology and The North American Society of Pacing and Electrophysiology for long-term ECGs only [22], several studies found good ability of SDNN for prediction of mortality also from short-segment ECGs [30, 31].

All ECGs were processed by the Modular ECG Analysis System (MEANS) [32] to automatically obtain the location and type of the QRS complexes of the 20-min ECG. A visual check was performed by a medical student which was supervised by a cardiologist. For further analyses, HRV derived from the best 5-min segment of the 20-min ECG was used according to the following quality criteria: proportion of abnormal beats; mean and variance stationarity score of the tachogram. Information on QRS complexes was used to compute standard time and frequency domain parameters of HRV for 5-min segments of the ECG according to the current guidelines for the analysis of HRV [22]. Artefacts and ectopic beats were replaced by interpolated normal sinus beats. We used the standard deviation of normal intervals (SDNN in ms) and the number of pairs of adjacent NN intervals differing by more than 50 ms divided by the total number of all NN intervals (pNN50 in %) – both time domain parameters – and the frequency domain parameters low frequency power (LF) (0.04 to <0.15 Hz), high-frequency (HF) power (0.15–0.4 Hz), the ratio of LF to HF (LF/HF) and very low frequency (VLF) power (<0 · 04 Hz). To calculate frequency domain parameters, tachograms of RR intervals were adjusted for linear trends, tapered and zero-padded and a Fast Fourier transformation was employed.

Data were analysed using the IBM SPSS Statistics 22 software package (IBM Corporation 2013, Somer, NY, USA). Univariate analyses of metric outcomes were performed with the Kruskal-Wallis test with Dunn-Bonferroni post hoc test. Contingency tables were tested using Chi² tests. Results were adjusted for multiple testing. Data are presented as median [Inter quartile range (IQR)].

We further calculated gender-, age- and heart rate-adjusted means (±95 % confidence interval (CI)) of HRV parameters by diabetic HD, nondiabetic HD and healthy controls using linear regression models. The F-test was used to test the difference in adjusted means of HRV between diabetic HD, nondiabetic HD and healthy controls. Multifactorial analyses were undertaken using SAS 9.3 (SAS Institute, Cary, NC, USA).

Sample size estimation was based on the putative difference of SDNN between ESRD patients and healthy controls. According to previous findings a difference of 66 ms was established [33]. The study should be able to show this difference at a sample size of ≥ 6 in each observation group (controls, diabetic and non-diabetic HD patients) with a power of 0.95 and an α error probability of 0.05. For sample size estimation, G*Power Version 3.1.9.2 [34, 35] was used.

Causes of ESRD were: 15 patients: diabetic nephropathy, 4: glomerulonephritis, 3: nephrosclerosis, 3: interstitial disease, 3: unknown origin, 1: polycystic kidney disease, 1: thrombotic microangiopathy. One diabetic patient had to be excluded from analysis as his ECG registrations could not be evaluated for technical reasons. We excluded one further patient from HRV analyses because of erratic heart rhythm, which was detected by poincaré plot analysis [36]. Subject demographics are listed in Table 1. All patients were treated thrice weekly for at least 4 h with similar single-use synthetic high-flux dialyzers and ultrapure dialysate and heparinate anticoagulation. Further treatment parameters and cardiovascular assessments are reported in Table 2.

As expected, there was a large difference in creatinine and urea values. However, also CRP and IL-6 values were significantly higher in dialysis patients. Furthermore, a large range was noted for BNP values but none of the patients had clinical signs of heart failure. Of note, the diabetic dialysis patients had a balanced glucose metabolism and HbA1c was within the goal range in most of the participants. All results are listed in Table 3.

Time-domain parameters (SDNN, pNN50) and frequency-domain parameters (VLF, LF, HF as well as the LF/HF ratio) are reported in Table 4. There were marked differences in all ECG parameters between patients and healthy controls. However, diabetic and non-diabetic hemodialysis patients did not differ with regard to the ECG parameters. After correction for age and gender, differences in heart rate, SDNN and VLF remained significant (Additional file 1: Table S1). Also, after additional correction for heart rate, results were unchanged (Table 5).

Results of total monocyte number, monocyte subpopulation quantification (Mo1, Mo2, Mo3) and monocyte expression density of CD143 (angiotensin converting enzyme) are reported in Table 6. Mo2 and Mo3 populations were expanded in diabetic and non-diabetic dialysis patients. There was a higher expression of CD143 on Mo2 and Mo3 cells in dialysis patients compared to healthy individuals, albeit it did not reach statistical significance.

Table 7 reports AChR density on monocytes (as defined by the coexpression of CD86 and CD14), B-lymphocytes (CD19+) and T-lymphocytes (CD3+). Monocytes showed a higher expression of the AChR in dialysis patients compared to controls. In contrast, the expression of AChR was not different on B- or T-lymphocytes.

Dialysis treatment parameters as well as hemodynamic parameters before and after a dialysis session were not related to the ECG analysis results. Also, ECG-derived markers of autonomous regulation were only weakly correlated to the monocyte subpopulation measurements (Table 8).

ECG analysis revealed strong signs of autonomous dysfunction. In both diabetic and non-diabetic dialysis patients the time-domain parameters of HRV analysis were severely impaired. This means that there is strong autonomous neuropathy that seems to be independent from the presence of diabetes mellitus.

In addition, the frequency-domain parameters were also strongly impaired. Customarily, LF was thought to reflect sympathetic influence on cardiac function whereas HF variation should be a marker of parasympathetic influences. Therefore, the ratio of LF to HF was considered to reflect the balance between sympathetic and vagal activity. Recently, this interpretation has been challenged and it was suggested that the HRV power spectrum is mainly determined by the parasympathetic system [37]. LF power may rather provide an index of baroreflex function [23].

Earlier studies showed that patients with chronic renal failure have a strong sympathetic overactivity [12] but little is known about parasympathetic activity in these individuals. Our data indicate that parasympathetic activation seems to be strongly diminished. Most likely, our findings result from the unexpectedly strong autonomous neuropathy, which renders the frequency-domain analysis inconclusive. Of note, the LF/HF ratio was not different between dialysis patients and controls, not even when analyzing only diabetic dialysis patients. This supports the notion that the global changes by renal failure disable the frequency-domain analysis. However, it has been revealed in a validation study with pharmacological blocking that LF/HF ratio is only weakly correlated with robust measures of autonomic tone [38]. Therefore, these results should be interpreted with caution. Additionally, the use of frequency-domain parameters has been criticized recently. Time-domain parameters are considered more robust since they can be estimated with smaller bias and considerably smaller variability than frequency domain parameters [39].

The conclusion that the specific balance between sympathetic and vagal tone in hemodialysis patients cannot be reliably measured by ECG analysis while reduced heart rate variability is a useful measure of autonomic dysfunction in general is compatible with recent publications on this topic. Kurata et al. [33] found high catecholamine levels as signs of sympathetic overactivation together with reduced heart rate variability (low SDNN).

A study by Fukuta et al. addressed heart rate variability in 120 chronic hemodialysis patients with similar age and dialysis vintage as in our patients [40]. They used a different ECG analysis technique based on 24 h Holter monitoring on the day between dialysis treatments. Therefore, the absolute measurements for time- and frequency-domain parameters cannot be compared to our data. Nevertheless, their findings are important to our study, since they demonstrated impaired R-R variation and low values for LF and HF variation in their patients. Furthermore, patients with the strongest impairment in heart rate variability were more likely to die from cardiovascular causes within the observation period of 26 months.

A recent study [41] showed that the initiation of hemodialysis therapy in advanced stage 4–5 CKD patients improves several measures of heart rate variability. Thus, uremic intoxication in general seems to play an important role.

The lack of echocardiographic data might appear as a significant limitation. However, none of the patients of the study had any sign of heart failure. Additionally, BNP measures in dialysis patients with echocardiographically diagnosed heart failure were more than 5 times higher than in our population [42] and BNP levels in our population are within a range that was identified as beneficial in a prospective 2-year mortality analysis [43].

Preliminary data from Psychari et al. [44] suggested a relation between markers of autonomous dysfunction and systemic inflammation. They demonstrated a correlation between SDNN values and IL-6 plasma levels in patients with CKD 3–4. In addition, the expression of adrenoceptors and AchRs on monocytes and the multiple effects that sympathetic [5, 6, 8, 9, 45] as well as parasympathetic [11] stimulation has on these cells, further supports this hypothesis. We could confirm earlier data on chronic inflammation and its extent in dialysis patients [4, 46, 47]. The inflammation markers CRP and IL-6 were greatly enhanced in both patient groups and the pro-inflammatory monocyte subpopulations were expanded. Furthermore, the expression of the angiotensin converting enzyme ACE on the surface of monocytes tended to be higher in dialysis patients compared to healthy controls. This also confirms earlier findings on subpopulation alterations and their enhanced expression of ACE [48‐50]. Since ACE expression on Mo2 cells is related to adverse outcome and enhanced cardiovascular disease in dialysis patients, these cells are thought to play a causal role in the progression of atherosclerosis. However, our study is limited by the different cardiovascular medication in cases and controls. The high burden of cardiovascular disease in ESRD makes it unlikely to recruit dialysis patients without cardiovascular medication. Therefore, confounding immunomodulatory effects can not be entirely ruled out. Also, heparinate anticoagulation has been discussed to posess anti-inflammatory effects. There is some low level clinical evidence in asthma patients with mixed results. Effects in other inflammatory diseases have been equivocal [51, 52]. However, it cannot be entirely ruled out that heparinate anticoagulation might influence the immune system, especially in vitro and in supraphysiological doses higher than in the dialysis circuit.

Interestingly, expression of the AchR is significantly elevated in CD14+ Monocytes in ESRD patients compared to controls. This is not the case for other cell populations such as CD3+ T-Lymphocytes and CD19+ B-Lymphocytes. It might appear as a limitation of our study, that antibody specificity was not tested with alternative approaches such as knockout cells/mice in addition to previous investigations [27‐29].

However, our analysis showed that there was no relation between markers of autonomous dysfunction and the studied markers of systemic and cellular inflammation. Additionally, we have at least no hint, that autonomic nerve dysfunction might relevantly influence monocyte subpopulation composition in dialysis patients. Although negative, this is an important clarification in the scientific debate. Monocyte subpopulation composition may be influenced by several conditions in chronic renal failure. The altered balance of sympathetic and parasympathetic tone can now be excluded as a major regulator of monocyte differentiation.