Aminoaciduria and Metabolic Dysregulation during Diabetic Ketoacidosis: Results from the Diabetic Kidney Alarm (DKA) Study

Abstract

Objective:

We examined changes in the excretion of various amino acids and in glycolysis and ketogenesis-related metabolites, during and after diabetic ketoacidosis (DKA) diagnosis, in youth with known or new onset type 1 diabetes (T1D).

Methods:

Urine samples were collected from 40 youth with DKA (52% boys, mean age 11±4 years, venous pH 7.2±0.1, blood glucose 451±163 mg/dL) at 3 time points: 0–8 hours and 12–24 hours after starting an insulin infusion, and 3 months after hospital discharge. Mixed-effects models evaluated the changes in amino acids and other metabolites in the urine.

Results:

Concentrations of urine histidine, threonine, tryptophan, and leucine per creatinine were highest at 0–8 hours (148.8±23.5, 59.5±12.3, 15.4±1.4, and 24.5±2.4% of urine creatinine, respectively), and significantly decreased over 3 months (p=0.028, p=0.027, p=0.019, and p<0.0001, respectively). Urine histidine, threonine, tryptophan, and leucine per urine creatinine decreased by 10.6±19.2, 0.7±0.9, 1.3±0.9, and 0.5±0.3-fold, respectively, between 0–8 hours and 3 months.

Conclusions:

In our study, DKA was associated with profound aminoaciduria, suggestive of proximal tubular dysfunction analogous to Fanconi syndrome.

INTRODUCTION

Diabetic ketoacidosis (DKA) is a common diabetes-related emergency that occurs in the setting of relative or absolute insulin deficiency [1]. DKA is also the leading cause of hospitalizations, morbidity, and mortality in children with type 1 diabetes (T1D) [2–4]. In the United States, up to 55% of individuals with new onset type 1 diabetes present with DKA [5–7], a significantly higher figure than other national registries [5, 6, 8].

While DKA-associated dehydration is a well-known risk factor for the development of acute kidney injury (AKI), DKA-related tubular injuries are poorly understood [9]. Previously, we demonstrated tubular injury during DKA, as evidenced by elevated tubular injury biomarkers including neutrophil gelatinase associated lipoprotein (NGAL), kidney injury molecule 1 (KIM-1) and interleukin 18 (IL-18) [10]. However, it is unclear whether DKA also associates with proximal tubular dysfunction, as reflected by inadequate reabsorption of amino acids, uric acid, and other metabolites in the proximal renal tubules and excessive excretion of these metabolites in the urine. Aminoaciduria, a pathogenetic feature of tubular dysfunction, was first described in a case report of new onset T1D from 1957 [11]. To date, there have been few prospective studies to examine the effects of DKA on proximal tubular function [12]. Additionally, tubular proteinuria may precede microalbuminuria in T1D [13], and studies have demonstrated that the proximal tubule [14] can be damaged early-on by high glucose concentrations [9, 15], raising the possibility of proximal tubular dysfunction as the earliest sign of diabetes-related kidney dysfunction [16]. Understanding tubular injury and damage during DKA is important to initiate early interventions and that could potentially mitigate the development of chronic kidney damage.

To address this knowledge gap, we measured urine amino acids and metabolites of glycolysis and ketogenesis in youth with either a known diagnosis of T1D or new-onset T1D at the time of DKA and then again 3 months post-hospitalization. Based on our previous observations of tubular injury during DKA, we hypothesized that DKA would associate with aminoaciduria and increased excretion of other metabolites of glycolysis and ketogenesis, suggestive of proximal tubular dysfunction.

METHODS

Participants and Study Design:

Forty youth were recruited from the Emergency Department (ED) at Children’s Hospital Colorado (CHCO) and all study participants underwent urine collection at three time points: once during the ED stay (0–8 hours), once following admission to the hospital (12–24 hours), and once 3 months post-hospital discharge (3 months). Inclusion and exclusion criteria were previously described [10]. New-onset T1D was defined as the episode of DKA occurring at the time of diagnosis of T1D. Known T1D was defined as the diagnosis of T1D preceding the evaluated episode of DKA.

ED Visit and 3 Month Follow-up Visit:

For each participant, demographics as well as hospital visit-specific information including start time of the insulin infusion and duration of the hospital stay were recorded. Laboratory tests including a complete metabolic panel and venous blood gas, were obtained as standard of care assessments for DKA in the ED. Laboratory assessments and equipment utilized were previously reported [10]. Urine studies were collected in the ED within 0–8 hours and 12–24 hours of the participant being started on an intravenous insulin infusion. First morning urine studies were also done after an 8-hour fast at the CHCO or Barbara Davis Center for Diabetes (BDC) three months after discharge from the hospital. DKA severity was defined by ISPAD criteria: mild = venous pH ≥ 7.2 to <7.3, moderate = venous pH ≥ 7.1 to <7.2, and severe = pH <7.1 [17]. According to Kidney Disease: Improving Global Outcomes (KDIGO) guidelines, AKI was defined as a 1.5-fold increase in serum creatinine concentration in the first 24 hours of admission above standard values for serum creatinine concentration by age and sex [18].

Research Labs:

Urine samples from the 0–8 hour, 12–24 hour, and 3 month follow-up time points were assessed for a variety of biochemical measures. Assays were performed by standard methods in the Clinical Translation Research Center (CTRC), CHCO clinical, and University of Colorado Hospital (UCH) clinical laboratories. Measures assessed included serum cystatin C (immunoturbidimetric assay, Kamiya Biomedical, Tukwila, WA), serum creatinine (Jaffé method, Thermo Scientific, Waltham, MA), urine creatinine (enzymatic assay, Beckman Coulter, Brea, CA), serum uric acid (enzymatic assay, Thermo Scientific, Waltham, MA), and urine uric acid (enzymatic assay, Thermo Scientific, Waltham, MA). Urine pH was measured from supernatant by Accumet Basic AB15/15+ pH meter (Fisher Scientific, Hampton, NH). Urine osmolality was measured using freezing point method with Micro-Osmometer Model 3300 (Advanced Instruments, Norwood, MA).

Metabolic biomarkers and concentrations of the amino acids alanine, glutamine, glycine, histidine, threonine, tryptophan, isoleucine, leucine, and valine were quantified from urine using high-throughput proton nuclear magnetic resonance (NMR) metabolomics (Nightingale Health Limited, Helsinki, Finland) [19]. This method provides simultaneous quantifications of biomarkers from several metabolic pathways including markers of fluid balance, glycolysis-related metabolites, and low-molecular weight metabolites. Full metabolite and marker panel can be obtained from Nightingale Health Limited. Output was provided in both molar concentration units and as ratios to urine creatinine concentration (100*unadjusted value (mmol/L)/urine creatinine (mmol/L)) to adjust for urine concentration. In our analyses, we present the metabolites as ratios to urine creatinine concentration (% of urine creatinine).

Statistical Methods:

The distributions of all variables were examined prior to analysis. Descriptive statistics are reported as means ± standard deviation, unless otherwise specified. Nominal variables are expressed as counts and percentages. Biomarker concentrations were modeled by study participant over time using mixed-effects models with random intercepts. Models were first fitted using time as the primary predictor followed by models with interaction effects between time and each of the following covariates: DKA severity, AKI, and known vs. new onset T1D status. This model architecture allowed for the evaluation of the effect of covariates on each biomarker, as well as any effect modification over time. For models that included categorical covariates (i.e., DKA severity, AKI, T1D status), contrast statements were used to test group differences at each time point. The values from these effect models are expressed in standard deviations of the measure. As the data collected contained many more variables than participants, analyses were considered exploratory and hypothesis generating. Fold percent change for each variable was calculated from 0–8 hours to 3 months for each participant and then averaged. Adjustments for multiple comparisons were not employed. An α-level of 0.05 was used for tests of statistical significance. Analyses were performed in R (version 4.0.2; R Foundation for Statistical Computing).

RESULTS

Participants:

Participant clinical characteristics are summarized in Table 1. The cohort had a mean age of 11±4 years with a mean HbA1c 12.7±1.9 % at DKA diagnosis. The mean venous pH, HCO₃⁻, and beta-hydroxybutyrate (BHOB) at diagnosis were 7.19±0.08, 8.90±4.75 mmol/L, and 7.55±2.25 mmol/L, respectively. Forty-eight percent of the participants were girls. Fifty-eight percent of the participants were diagnosed with T1D at the time of their DKA episode and 42% had an established diagnosis of T1D at the time of DKA. According to ISPAD criteria, 16 participants had mild DKA, 15 had moderate DKA, and 9 had severe DKA.

Table 1:

Participant demographics and baseline laboratory assessments

	N=40
Sex	21 (52.5%) boys
	19 (47.5%) girls
Race	28 (70%) Non-Hispanic White
	5 (12.5%) Hispanic White
	3 (7.5%) Black
	4 (10%) More than one race
Age (years)	11 ± 4
Height (cm)	148.9 ± 23
Weight (kg)	44 ± 19
BMI (kg/m2)	18.7 ± 4.5
New type 1 diabetes diagnosis	23 (57.5%)
Systolic blood pressure (mm Hg)	116 ± 14
Diastolic blood pressure (mm Hg)	76 ± 12
Heart rate (beats per minute)	106 ± 18
DKA severity	Mild: 16 (40%)
	Moderate: 15 (37.5%)
	Severe: 9 (22.5%)
HbA1c (%)	12.7 ± 1.9
GADA positive	26 (65%)
mlAA positive	18 (45%)
IA-2 positive	24 (60%)
ZnT8 positive	22 (55%)
Venous pH	7.2 ± 0.1
Serum HCO3− (mEq/L)	8.9 ± 4.7
Serum BHOB (mg/dL)	78.1 ± 22.9
Serum Na+ (mEq/L)	143 ± 7
Serum K+ (mEq/L)	4.7 ± 0.8
Serum Cl− (mEq/L)	103 ± 6
Serum creatinine (mg/dL)	0.70 ± 0.28
Blood urea nitrogen (mg/dL)	15 ± 7
Serum glucose (mg/dL)	451 ± 163
Serum phosphorus (mg/dL)	5.2 ± 2.0
Serum calcium (mg/dL)	10.5 ± 1.7
Urine glucose (mg/dL)	510 ± 213
Urine osmolality (mOsm/kg)	804.1 ± 235.1

Results are expressed as mean ± standard deviation, or count and percentage, as appropriate. Serum Na⁺ has been corrected by glucose.

Abbreviations: BHOB, beta-hydroxybutyrate; DKA, diabetic ketoacidosis; GADA, glutamic acid decarboxylase antibodies; HCO3⁻, bicarbonate; IA-2, islet tyrosine phosphatase 2 antibodies; mIAA, micro-insulin autoantibodies; T1D, type 1 diabetes; ZnT8, zinc transporter 8 antibodies.

Urine Glycolysis and Ketogenesis-Related Metabolites:

Propylene glycol and 3-hydroxyisovalerate were highest at 0–8 hours with mean concentrations of 161.8±21.5 and 37.6±2.2 % of urine creatinine, respectively (Figure 1). These concentrations lowered at 12–24 hours (mean 81.6±23.6 and 16.1±2.2 % of urine creatinine, respectively) and further at 3 months post (8.0±21.5 and 11.6±2.0 % of urine creatinine, respectively). Urine lactate and glucose were highest at 0–8 hours of onset of DKA with mean concentrations of 791.8±77.2 and 33,070.7±4,107.9 % of urine creatinine, respectively, but had recovered to values similar to those observed at 3 months by the 12–24 hour timepoint (mean 76.5±80.1 and 8730.1±3817.9 % of urine creatinine). Lactate, 3-hydroxyisovalerate, and propylene glycol decreased by 0.3±0.8, 0.4±0.4, and 0.2±0.4-fold, respectively (p<0.0001 for all).

Figure 1: Trajectories of urine amino acids and metabolic biomarkers over time.

Solid-colored lines represent the trend of the study variables means over time, stratified by DKA severity (green=mild and red=moderate/severe). Single colored points represent the respective value of each participant for that outcome. Units are ratios to urine creatinine concentration (100*unadjusted value (mmol/L)/urine creatinine (mmol/L)). P-values indicate whether the change over the 3 timepoints was significant.

Abbreviations: DKA, diabetic ketoacidosis.

Urine Amino Acids:

Concentrations of urine histidine, threonine, and tryptophan were highest during DKA, showing similar levels at 0–8 and 12–24 hours. Peak concentrations of urine histidine, threonine, and tryptophan were at 0–8 hours with mean concentrations of 148.8±23.5, 59.5±12.3, and 15.4±1.4 % of urine creatinine, respectively (Figure 1). Concentrations of urine glycine were highest at 12–24 hours with mean concentrations of 237.7±44.4 % of urine creatinine. Urine leucine concentrations were highest at 0–8 hours with a mean concentration of 24.5±2.4 % of urine creatinine. These concentrations lowered at 12–24 hours (mean 12.0±2.3 % of urine creatinine) and further at 3 months post (mean 6.1±2.2 % of urine creatinine). Histidine, threonine, tryptophan, and leucine changed 10.6±19.2, 0.7±0.9, 1.3±0,9, and 0.5±0.3-fold, respectively. The decrease in glycine (p=0.048), histidine (p=0.028), threonine (p=0.027), tryptophan (p=0.019), and leucine (p<0.001) over time were significant (Figure 1). Further, there was a significant decrease in leucine from 0–8 to 12–24 hours, (p=0.001).

Effects Modification by DKA Severity:

Within 8 hours of DKA onset, participants with moderate or severe DKA exhibited higher urine 3-hydroxyisovalerate (p<0.001), urine lactate (p<0.001), and urine glucose (p=0.007) per creatinine concentrations than those with mild DKA. There was no significant impact of DKA severity on any amino acids measured.

Relationships between AKI and Urine Metabolites:

Participants with AKI had a 75.2 % higher urine threonine (106.4±31.8 vs. 31.0± 13.3 % of urine creatinine, p<0.05) and a 10.3 % higher urine tryptophan (22.3±3.4 vs. 12.0±1.4 % of urine creatinine, p<0.05) than those without AKI at 12–24 hours (Figure 2).

Figure 2: Effect size of acute kidney injury on urine amino acids and metabolic biomarkers.

Effect size estimates from linear models are represented by bars with standard error around the estimated effect and interpreted as effect of y-axis variable on x-axis variable at 0–8 hours (top) and 12–24 hours (bottom). Effect size units are reported as change in z-score per 1 standard deviation change in the x-axis variable.

Relationships between New-Onset Type 1 Diabetes Status, and Urine Metabolites:

Participants with new-onset T1D had a 20.8 % higher urine leucine (32.8±2.7 vs. 12.0±3.2 % of urine creatinine, p<0.01) than individuals with known T1D at 0–8 hours (Figure 3).

Figure 3: Effect size of new onset type 1 diabetes on urine amino acids and metabolic biomarkers.

Effect size estimates from linear models are represented by bars with standard error around the estimated effect and interpreted as effect of y-axis variable on x-axis variable at 0–8 hours (top) and 12–24 hours (bottom). Effect size units are reported as change in z-score per 1 standard deviation change in the x-axis variable.

At 12–24 hours, participants with new-onset T1D had a 5.8 % higher urine tryptophan (16.5±1.9 vs. 10.7±1.9 % of urine creatinine, p<0.05, a 75.3 % higher urine threonine (106.4±31.8 vs. 31.1±13.3 % of urine creatinine, p<0.05), and a 7.2 % higher urine isoleucine (11.5±1.8 vs. 4.2±1.6 % of urine creatinine, p<0.05), than those individuals with known T1D. Participants with new-onset T1D also had 13.4 % higher urine histidine concentrations (mean 101.7±44.2 vs. 88.3±33.2 % of urine creatinine, p<0.05) and 8.0 % higher urine leucine (mean 16.0±2.8 vs. 8.0±2.9 % of urine creatinine, p<0.05) compared to those with known T1D at the 12–24 hour time point.

DISCUSSION

In this study, we report high urinary excretion of numerous amino acids during DKA, with improvements in aminoaciduria demonstrated at 3 months post-hospitalization. The aminoaciduria present in the setting of DKA was accompanied by both glucosuria and uricosuria, analogous to Fanconi syndrome and suggestive of proximal tubular dysfunction [20].

Amino acids are commonly classified as either glucogenic, ketogenic, or mixed amino acids. Glucogenic amino acids are converted to glucose through the process of gluconeogenesis, while ketogenic amino acids are catabolized into acetyl-CoA, a precursor for the development of ketone bodies (Supplemental Figure 1). Due to insulin deficiency, DKA is typically accompanied by a combination of enhanced lipolysis and protein catabolism which leads to increased ketogenesis as the main source of cellular energy [1]. Branched chain amino acids (BCAAs) including leucine (ketogenic), isoleucine (mixed), and valine (glucogenic) are increasingly demonstrated to have associations with diabetes and possibly DKA [12, 21–25]. Insulin resistance, a critical feature of both type 1 and type 2 diabetes in youth and adults [26, 27], suppresses the catabolism of BCAAs by reducing the enzymatic activity of branched chain α-keto acid dehydrogenase complex [21, 28], a plausible etiology of the increase in BCAAs in obesity and diabetes. In animal studies, it has been shown that the supplementation of BCAAs attenuate oxidative stress, proteinuria, and alleviate diabetic kidney injury [38]. In our study, urinary excretion of the BCAA leucine and the non-BCAAs threonine (mixed), tryptophan (mixed), and histidine (glucogenic) were highest at 0–8 hours of DKA, highlighting possible dysfunction in the proximal tubule as all types of amino acids are excreted.

As a ketogenic amino acid, leucine is metabolized into the ketone body 3-hydroxyisovaleric acid which significantly decreases over time with resolution of DKA [29]. We demonstrate significantly elevated urinary excretion of leucine in participants with severe DKA vs. mild DKA, as well as elevated urinary excretion of tryptophan in moderate DKA vs. mild DKA. Additionally, participants with known T1D had significantly elevated urinary excretion of leucine at 0–8 hours and isoleucine and tryptophan at 12–24 hours. DKA is a metabolically perturbed state, and the severity of DKA is likely more influential on the urine concentrations than known vs. new-onset T1D. Nevertheless, these results together suggest a role for the progressive metabolic consequences of severe DKA in the development of proximal tubular dysfunction and subsequent aminoaciduria. In contrast, urinary excretion of threonine and histidine did not differ significantly between severity categories of DKA. Our findings suggest a potential role of 3-hydroxyisovalerate and leucine, in addition to the well-studied β-hydroxybutyrate in the management of DKA.

In addition to the intrinsic elevations of ketogenic acids inherent to DKA, pre-renal effects secondary to dehydration are common and lead to high prevalence of AKI. Co-occurrence of aminoaciduria in the setting of AKI plus DKA provides further insight into the possibility of additional injury sustained to the proximal tubule of the nephron in the setting of DKA. Consequently, the ability to predict who will develop AKI during DKA is important, as timely detection of AKI is essential to initiate early intervention and thereby prevent lasting kidney damage. At 3 months post-hospitalization, we demonstrated an improvement in the aminoaciduria seen in DKA, potentially secondary to the resolution of acidosis. Remission of the biochemical manifestations of proximal tubular dysfunction frequently occurs with the resolution of acidosis in individuals with concurrent distal tubular injury [30, 31] and our observations of resolution of aminoaciduria post DKA are consistent with this pattern.

Few published studies to date have fully described the urinary excretion patterns of amino acids in children with diabetes experiencing DKA. One prior study described aminoaciduria in DKA and found that the urinary excretion of BCAAs histidine, serine, and threonine was elevated in poorly controlled youth with T1D and DKA when compared to youth with well-controlled T1D [12]. This study from 1991 was done on a cohort of 10 children with type 1 diabetes admitted in severe DKA with a mean blood pH of 7.15 and 10 children with well-controlled type 1 diabetes and included 10 controls. The amino acids were assessed using a different method on the Biotronic LC 2000 amino acid analyzer. This study also found changes in serum amino acids and linked them to metabolic alterations and renal dysfunction occurring during DKA. Few other studies of individuals with diabetes not in DKA corroborate our findings. In a study of youth with T1D without microalbuminuria, Durá-Gúrpide et al found a lower excretion rate of urinary ketogenic and mixed amino acids but a higher excretion rate of urinary glucogenic amino acids in children with T1D vs. controls [32]. Another study demonstrated increased urinary leucine, lysine, isoleucine, and valine excretion in children with T1D compared to controls without T1D [33]. We provide additional evidence of silent proximal tubular injury during DKA as evidenced by increased aminoaciduria in youth with T1D and DKA, which supports a potential central role for proximal tubular injury in the pathogenesis of kidney dysfunction and diabetic kidney disease in T1D. Proximal tubular injury and consequent aminoaciduria may serve as a marker of early functional and structural disturbances in the kidney, in addition to known functional abnormalities like hyperfiltration [34–37], and may represent a critical early indicator of kidney disease development and progression in T1D.

This study has several important strengths. These include its prospective study design, extensive inclusion and exclusion criteria that necessitated the presence of T1D autoantibodies for all enrolled youth with a diagnosis of T1D, the large number of biomarkers analyzed, and the use of serial measurements of urine biochemical parameters. This study also has several notable limitations. These include a relatively small sample size, a short follow-up time of 3 months, use of estimated rather than measured GFR assessments, limited data regarding kidney function prior to this DKA episode, and a lack of healthy age- and sex-matched controls to define normative values for the analyzed tubular injury biomarkers.

Additionally, there are no universally accepted standards of aminoaciduria. The measurements collected three months after discharge provide surrogate normative values, although it is notable that they could be influenced by the long-term effects of DKA on tubular function in T1D. Furthermore, because the analyses presented here were exploratory, adjustments for multiple comparisons were not employed. Future longitudinal and flux studies are necessary to further elucidate the mechanisms of proximal tubular dysfunction in DKA.

CONCLUSION

We previously demonstrated that DKA associates with proximal tubular injury [10]. Here we extend these observations by showcasing further evidence of proximal tubular dysfunction in the setting of DKA by demonstrating significant aminoaciduria which is suggestive of a proximal tubular dysfunction. Understanding the tubular dysfunction and subsequent aminoaciduria present in DKA may allow for the development of novel screening tests that can identify early kidney dysfunction well before the current clinical markers of diabetic kidney disease such as microalbuminuria occur.

Highlights:

• Diabetic ketoacidosis (DKA) in children with T1D is characterized by aminoaciduria.

• Aminoaciduria in DKA resolved by 3 months post hospitalization.

• Aminoaciduria provides further evidence of proximal tubular dysfunction during DKA.