Exhaled volatile substances in children suffering from type 1 diabetes mellitus: results from a cross-sectional study

Monitoring metabolic adaptation to type 1 diabetes mellitus in children is challenging. Analysis of volatile organic compounds (VOCs) in exhaled breath is non-invasive and appears as a promising tool. However, data on breath VOC profiles in pediatric patients are limited. We conducted a cross-sectional study and applied quantitative analysis of exhaled VOCs in children suffering from type 1 diabetes mellitus (T1DM) (n = 53) and healthy controls (n = 60). Both groups were matched for sex and age. For breath gas analysis, a very sensitive direct mass spectrometric technique (PTR-TOF) was applied. The duration of disease, the mode of insulin application (continuous subcutaneous insulin infusion vs. multiple daily insulin injection) and long-term metabolic control were considered as classifiers in patients. The concentration of exhaled VOCs differed between T1DM patients and healthy children. In particular, T1DM patients exhaled significantly higher amounts of ethanol, isopropanol, dimethylsulfid, isoprene and pentanal compared to healthy controls (171, 1223, 19.6, 112 and 13.5 ppbV vs. 82.4, 784, 11.3, 49.6, and 5.30 ppbV). The most remarkable differences in concentrations were found in patients with poor metabolic control, i.e. those with a mean HbA1c above 8%. In conclusion, non-invasive breath testing may support the discovery of basic metabolic mechanisms and adaptation early in the progress of T1DM.

pathophysiological processes. As VOCs are exhaled shortly after their production, they may deliver information quicker than invasive techniques and enable insights into short-term variations of metabolites. Apart from the typical sweet smell of breath in patients with severe diabetic ketoacidosis, hundreds of potentially endogenous volatile metabolites have been quantified in trace amounts (~ppbV-pptV range) in human exhalation under physiological and pathophysiological conditions [20][21][22][23][24][25][26][27] . Exhaled amounts of monomethylated alkanes and C 4 -C 20 alkanes have been suggested as markers of oxidative stress in diabetic adults 28 . Several pilot studies investigated the exhaled amounts of acetone, isoprene and a variety of other VOCs as surrogate markers of blood glucose and/ or as diagnostic tool for detection of hyper-or hypoglycemic events under well-defined experimental settings, i.e. intraveneous or oral glucose challenge, hypo-or euglycemic clamps [29][30][31][32][33][34] . While an association between metabolic state and certain VOCs could be detected under such circumstances, a single measurement of either acetone or isoprene failed to reflect blood glucose in pediatric T1DM patients 35 . Instead, a positive association between blood ketones and breath acetone was noted.
While breath gas analysis is certainly not suitable to assess the glycemic state, it might help to gain knowledge on metabolic adaption under conditions of real-life, i.e. simultaneous investigation of a broad panel of VOCs ("volatilome") in real-time. This can e.g. be done with proton-transfer-reaction time-of-flight mass spectrometry (PTR-ToF-MS). Very recently, we applied this technique in children with mild-to-moderate chronic kidney disease and healthy controls 36 . Within the frame of this study, we also invited children suffering from T1DM that were treated at our institution in order to examine volatile metabolites in the breath of pediatric T1DM patients and healthy controls.
Methods patients and controls. The study received appropriate ethics committee approval from the institutional Ethics Committee (University Medical Centre Rostock, Rostock, Germany) in accordance with the Declaration of Helsinki (approval number: A 2012 0103). All subjects and/or their parents gave written and informed consent for their participation in the study.
Both, healthy controls and T1DM patients were excluded in any case of febrile illness during the last three months, chronic inflammatory-/rheumatic disease (e.g. Crohn's disease, rheumatoid arthritis), hepatitis, HIV, glucocorticoid treatment, liver-, renal-, or cardiac failure, or hereditary dyslipidaemia.

Study investigations.
All patients were seen in the afternoon and breath sampling was done after a routine visit in our outpatient clinic. Demographic and clinical data including the history of disease, daily insulin dosage related to body weight as well as results from routine laboratory analysis were gathered by interview and chart review, respectively. The mean HbA 1c during the last 12 months was considered as surrogate marker of long-term metabolic control and a mean HbA 1c above 8% was considered as poor.
A trained physician measured weight and height via electronic scales and a fixed stadiometer. Blood pressure (BP) was measured according to the updated Task Force Report on high blood pressure by using an oscillometric device (Dinamap 1846SX; Critikon, Tampa, USA). Calculations of individual age-and sex-related standard deviation scores (SD scores) for height, weight, BMI and BP were done as previously described 7,37-39 . Patients were classified as hypertensive in case of BP values above the height-and sex-related 95th percentile.
Breath sampling and analysis. Breath was sampled continuously during a five minute period and VOCs were analysed in real time by high-resolution proton-transfer-reaction-time-of-flight-mass-spectrometery (PTR-ToF-MS). For this purpose, a PTR-ToF-MS 8000 (Ionicon Analytik GmbH, Innsbruck, Austria) was used essentially as described before 36,40,41 . All participants breathed evenly through the mouth into a sterile T-piece that did not introduce a breathing resistance. The T-piece enabled sampling in side-stream mode of exhaled and inhaled breath (sampling flow rate was 20 ml/min) into an inert and heated (75 °C) transfer line (silcosteel, ID 0.75 mm, Restek, Bellafonte, USA). The transfer line was connected to the drift tube (75 °C, 610 V, 2.3 mbar) of the PTR-ToF-MS. Within the drift tube, the proton transfer reaction between VOCs and H 3 O + occured prior to transfer of the protonated VOCs into the high resolution reflectron time-of-flight mass spectrometer (Tofwerk AG, Thun, Switzerland). The mass and time resolution were ~4000 m/Δm and 200 ms, respectively. The mass scale was recalibrated every minute using the H 3 O + -isotope (21.023), NO + (29.998) and C 3 H 6 O (59.049). For quantification of individual VOCs, the intensity measured in counts per seconds (cps) during the third minute of the sampling interval was averaged and normalized onto primary (H 3 O + ) ion count. A liquid calibration unit (constant total gas flow 1000 ml min −1 , water flow 0.05 ml min −1 , 75 °C) was used for external calibration of isoprene, isopropanol, C 1 -C 10 aldehydes, acetone, ethanol and dimethylsulfide 36 . Data processing. A custom made Matlab-based data processing algorithm ("breath tracker", Matlab version 7.12.0.635, R2011a) enabled the automatic recognition of expiratory and inspiratory phases. Acetone was used as tracker mass 36,40,41 .
Seven VOCs that are either related to metabolic or physiologic processes relevant for T1DM (e.g. glucose metabolism) or have been proposed in the literature as potential markers for metabolic processes and thus may be relevant for diabetes, were selected for further data analysis. VOCs from healthy controls and T1DM patients were tested for differences related to anthropometric and clinical characteristics. In addition, we investigated our findings relative to the mode of insulin therapy (MDII vs. CSII) and the mean HbA 1c , i.e. a mean HbA 1c below and above 8% was used to discriminate between good and poor long-term metabolic control.

Statistics.
For statistical analysis and visualization of results SPSS statistical package 22 (SPSS Inc. Chicago, Illinois, USA), and Sigma Plot Version 10 (Jandel Scientific Inc.) were used. Normal distribution of data was www.nature.com/scientificreports www.nature.com/scientificreports/ evaluated by the Kolmogorov-Smirnov test and comparison between groups was done using Student's t-test or Mann-Whitney U test, if appropriate. For computation of correlations, Spearman's rho test was used. All p-values were two-sided and a p-value below 0.05 was considered significant. Data are given as mean ± standard deviation (sd) or median and range, where appropriate.

Results
Characteristics of patients and controls. A total of 53 T1DM patients (32 males, 21 females) was enrolled in parallel to 60 healthy controls (28 males, 32 females). All patients were in a stable metabolic state and breath gas analysis was done adjacent to a regular consultation in our outpatient clinic. The categorization of patients according to sex, mode of therapy (MDI vs. CSII), long-term metabolic control (mean HbA1c below or above 8%) and duration of disease (less than 5 years vs 5 years or longer) is presented in Fig. 1. Standardized anthropometric data and clinical characteristics of patients and controls are given in Table 1. None of these parameters were related to sex, duration of disease, or mode of therapy. However, patients with poor long-term metabolic control showed significantly higher cholesterol (4.76 ± 1.02 mmol/l vs 3.95 ± 0.74 mmol/l; p < 0.001), LDL-cholesterol (2.73 ± 0.88 mmol/l vs 2.03 ± 0,60 mmol/l; p < 0.001), triglycerides (1.47 ± 1.22 mmol/l vs 0.78. ± 0.52 mmol/l; p < 0.001) and higher blood pressure (standardized BP sys : 2.28 ± 1.07 vs 1.13 ± 0.96; p < 0.001; standardized BP dias : 1.35 ± 1.34 vs 0.83 ± 0.97; p < 0.001) (Supplement 1).

VOC analysis and breath profiles.
More than 300 VOCs were detectable in the breath of pediatric T1DM patients and healthy controls. Only VOCs with expiratory concentrations higher than inspiratory concentrations were considered in detail. Figure 2 shows a heatmap of 33 potential VOCs that fulfilled these criteria.  www.nature.com/scientificreports www.nature.com/scientificreports/ Out of these 33 VOCs, we focused on seven volatile metabolites with a known or postulated relation to metabolic processes which could be relevant within patients suffering from T1DM.
In While noticable differences between controls and patients were observed for ethanol, isopropanol, DMS, isoprene, and pentanal, comparable concentrations of acetone and limonene were exhaled by both groups (Table 2 and Fig. 3). Further, we discriminated patients according to long-term metabolic control, i.e. a mean HbA 1c below and above 8%. Compared to healthy controls, either cohort of T1DM patients exhaled significantly higher amounts of ethanol, DMS, isoprene, and pentanal. Additionally, the alveolar concentrations of isopropanol and pentanal were significantly different in patients with adequate and poor long-term metabolic control.
In T1DM patients, the exhaled amount of isopropanol was positively related to the HbA 1c at the time of examination and the mean HbA 1c (Fig. 4). Furthermore, the exhaled amounts of isopropanol and pentanal were associated and either VOC showed an association with the cholesterol and LDL-cholesterol concentration, respectively (Fig. 5).
Within the patient cohort, positive relations between acetone and limonene (r = 0.51, p < 0.001) as well as between acetone and ethanol (r = 0.61, p < 0.001) were noted. By contrast, in healthy controls the alveolar concentrations of the VOCs under study were virtually unrelated to each other. (Fig. 6).

Discussion
Within this study, we investigated 53 pediatric T1DM patients in parallel to children suffering from chronic kidney disease and healthy controls. PTR-ToF-MS was used for direct analysis of exhaled VOCs. The findings in children with impaired renal function relative to healthy controls have been published recently 36 . Here we focused on the exhalation of ethanol, acetone, isopropanol, dimethylsulfide, isoprene, pentanal, and limonene by pediatric T1DM patients.
These compounds were selected as they have been previously associated to the disturbed glucose homeostasis or reflect a metabolic link to T1DM-related comorbidities, i.e dyslipidimia and oxidative stress.   www.nature.com/scientificreports www.nature.com/scientificreports/ The most famous volatile metabolite related to an impaired glucose metabolism is acetone and several studies with diabetic patients point to significantly enhanced acetone exhalation in diabetic patients 34,[42][43][44][45] . To our surprise, this was not the case in our patient cohort and most probably reflects differences in the experimental setting, i.e. sampling of breath in a real-life setting when patients are in a stable metabolic state. In line with this point of view, no association between blood glucose and acetone was detectable. However, our T1DM patients exhaled significantly higher isopropanol concentrations than their healthy peers and isopropanol was associatied with HbA 1c as well as with cholesterol and LDL-cholesterol levels in blood. While the exspiration of isopropanol might mirror endogeneous uptake secondary to the use of disinfectant, this will hardly translate into the associations mentioned above. Instead, the reduction of acetone to isopropanol by means of a nicotinamide adenine dinucleotide dependent redox-reaction has to be considered 46 . The formation of isopropanol contributes to acetone metabolism and this pathway may be activated in T1DM patients and thus contributing to acetone elimination and indicating metabolic adaptation.  www.nature.com/scientificreports www.nature.com/scientificreports/ Isoprene (2-methyl-1,3 butadiene) and limonene (1-Methyl-4-(prop-1-en-2-yl)cyclohex-1-ene) both belong to the family of monoterpenes. Isoprene is supposed to be linked to the biosynthesis of cholesterol and is most probably stored in fat tissue 47,48 . While limonene did not show differences between healthy controls and T1DM patients, isoprene concentrations were elevated in the T1DM group. Previous studies describe similar isoprene concentrations in healthy children and children with T1DM 42 or even elevated concentrations in adults during hypoglycemic phases 49 . However, the number of participants was low in both cases and no controlled alveolar sampling was applied. As isoprene is known to be largely effected by physiological variations 41,47,50,51 , this may have influenced these results.
Endothelial dysfunction is known to be present in T1DM patients [52][53][54] and recent research suggests that endothelial dysfunction may already emerge in diabetic children, even if no atherosclerotic structural changes have been observed yet 38,55 . Increased isoprene concentrations in T1DM patients thus may reflect damages at the cellular level that will translate in altered physiological properties of the endothelium. Furthermore, long-term hyperglycemia might cause vasoconstriction and an inflammatory state with concomitant oxidative stress, which in turn is mirrored by an exhalation of elevated amounts of pentanal. Such structural damage may further influence substance distribution and transport between compartments, leading to alterations in VOC exhalation.
An increase of aldehyde concentrations in T1DM patients, as observed for pentanal in our data, might indicate an elevation of oxidative activity. Oxidative stress is associated with chronic microinflammation and may be to blame for many if not for all of the long-term sequelae seen in diabetic patients 7,56 . However, it is difficult to verify oxidative stress and it has yet to be clarified if elevated pentanal concentrations originate from oxidative stress or an impairment of the antioxidant response.
Dimethylsulfide mainly originates from bacterial activity 57,58 and may be a breakdown product of methionine 59 . Elevated concentrations in T1DM patients thus may be explained by alterations of the gut microbiome under permanent insulin-deficiency. This might also result in an increased endogeneous production of ethanol, which was elevated in T1DM patients, too 60   www.nature.com/scientificreports www.nature.com/scientificreports/ concentrations in the breath of healthy subjects directly after ingestion of 75 g of glucose, mirroring the increasing blood glucose concentration, However, ethanol returned to baseline quicker than blood glucose. Authors hypothesized that bacterial ethanol production may be supported by systemic hyperglycemia 61 . This explanation might be valid in our study as well, since ethanol concentrations were highest in patients with poor metabolic control (mean HbA 1c above 8%). Simic et al. noted elevated endogenous ethanol levels in the blood of patients with diabetes mellitus which would be directly reflected in elevated breath concentrations and hence corroborate this hypothesis 62 .
In conclusion, differences in VOC profiles can be observed between healthy controls and T1DM patients, even in children, where long term effects such as co-morbidities have not emerged yet. Those differences were most pronounced in patients with poor metabolic control, i.e. those with a mean HbA 1c above 8%.
Monitoring of poor metabolic control as well as early observation of metabolic adaptation through a non-invasive window via breath profiles in T1DM patients may support the discovery of basic metabolic mechanisms and adaptation early in the progress of T1DM.