Serum metabolic profile of postoperative acute kidney injury following infant cardiac surgery with cardiopulmonary bypass

Abstract

Objective:

To determine differences in the circulating metabolic profile of infants with or without acute kidney injury (AKI) following cardiothoracic surgery with cardiopulmonary bypass (CPB).

Methods:

We performed a secondary analysis of preoperative and 24hr postoperative serum samples from infants ≤120d/o undergoing CPB. Metabolic profiling of the serum samples was performed by targeted analysis of 165 serum metabolites via tandem mass spectrometry. We then compared infants who did or did not develop AKI in the first 72 hours postoperatively to determine global differences in the preoperative and 24hr metabolic profiles in addition to specific differences in individual metabolites.

Results:

A total of 57 infants were included in the study. Six infants (11%) developed KDIGO stage 2/3 AKI and 13 (23%) developed stage 1 AKI. The preoperative metabolic profile did not differentiate between infants with versus without AKI. Infants with severe AKI could be moderately distinguished from infants without AKI by their 24hr metabolic profile, while infants with stage 1 AKI segregated into two groups, overlapping with either the no AKI or severe AKI groups. Differences in these 24hr metabolic profiles were driven by 21 metabolites significant at an adjusted false discovery rate of <0.05. Prominently altered pathways include purine, methionine, and kynurenine/nicotinamide metabolism.

Conclusion:

Moderate-to-severe AKI after infant cardiac surgery is associated with changes in the serum metabolome, including prominent changes to purine, methionine, and kynurenine/nicotinamide metabolism. A portion of infants with mild AKI demonstrated similar metabolic changes, suggesting a potential role for metabolic analysis in the evaluation of lower-stage injury.

Introduction:

Acute kidney injury (AKI) is one of the leading complications of infant cardiac surgery, with incidence ranging from 25–64% depending on the specific population and definition used.[1–4] Postoperative AKI is associated with increased short-term mortality, with rates up to 12% in neonates and infants who develop severe AKI.[1, 5, 6] Infants with AKI also experience significant postoperative morbidity, including greater fluid overload [7], longer duration of mechanical ventilation, and increased length of hospital stay.[4, 5, 8] Despite the clinical importance of postoperative AKI following infant cardiac surgery, diagnosis of early or mild AKI continues to be challenging and the systemic effects of AKI remain incompletely understood.[9]

Metabolites are low molecular weight compounds (<1500 Daltons) that serve as the end products of gene/protein expression and determine the instantaneous cellular phenotype.[10–12] Collectively, the body’s metabolites are termed the metabolome. The circulating metabolome represents contributions from multiple metabolically active organs including the liver, endothelium, lung, intestines, brain, heart, and kidneys. The kidneys are particularly important as they can contribute both directly to the generation of circulating metabolites as well as indirectly through altered clearance.[13–17] Pre-clinical studies measuring changes in tissue, serum, and urine metabolites following ischemia-reperfusion injury have identified both novel markers of injury as well as potential pathologic metabolic pathways.[16, 18]

Individual metabolites such as creatinine and lactate are commonly assessed in the routine postoperative care of children undergoing congenital heart disease surgery. Recent improvements in mass spectrometry-based targeted metabolite assays now allow for analysis of larger metabolic screening panels.[15, 17, 19] Using these techniques, our group has previously shown a remarkable disruption of the circulating metabolome following infant cardiothoracic surgery with cardiopulmonary bypass (CPB).[20] Differences in the circulating metabolome associated with development of postoperative AKI, however, are not well defined in any patient population and to our knowledge there are no existing data in from pediatric congenital heart disease populations.

In this study we performed targeted serum metabolic profiling of infants undergoing CPB to determine the ability to differentiate between infants with and without postoperative AKI based on the circulating metabolome. As described in a recent review by Johnson, et al,[21] targeted metabolomic profiling represents the first stage of metabolomic study. The goal of this stage is to identify candidate metabolic pathways and metabolites for subsequent diagnostic and mechanistic study. We hypothesized that perioperative metabolic profiles could discriminate between patients with or without AKI and that patients with more severe AKI would demonstrate greater changes in their circulating metabolome. Furthermore, we sought to determine the alterations in specific metabolic pathways and individual metabolites that could be candidates for future comprehensive study in patients who develop postoperative AKI.

Methods:

Overall Project and Recruitment:

This study is a secondary analysis of a previously published cohort study that evaluated the role of alkaline phosphatase as a biomarker of postoperative outcomes following neonatal or infant congenital heart disease surgery. Methods for the parent cohort study have previously been published.[22] Subjects were enrolled in the parent study if they were ≤120 days of age and scheduled to undergo cardiothoracic surgery with CPB. Exclusion criteria for the parent study were weight <2kg (due to risk of excessive blood draws for research) and adjusted gestational age <34wks at the time of surgery (due to differences in alkaline phosphatase biology in premature infants). Subjects were included in this secondary analysis if a residual 24hr postoperative serum sample was available for metabolomic analysis. The Colorado Multiple Institution Review Board approved the protocol including this planned metabolomics analysis and informed written consent was obtained from subjects’ families prior to enrollment.

Clinical Data:

Baseline perioperative and postoperative clinical data were collected on all subjects including gender, preterm delivery, age and weight at surgery, Aristotle comprehensive complexity score, presence of postoperative single ventricle physiology, CPB time, aortic cross clamp time, deep hypothermic circulatory arrest/selective cerebral perfusion times. Clinical serum creatinine (obtained at least daily) and hourly urine output were collected for the first 72hrs postoperatively to determine the development of AKI using the Kidney Disease Improving Global Outcomes (KDIGO) criteria with the neonatal modification for subjects ≤28 days of age.[23, 24]

Sample collection and processing:

Serum samples were obtained preoperatively after induction of anesthesia but prior to first surgical incision and at 24 hours after return to the cardiac intensive care unit. Samples were collected in standard red top tubes without anticoagulant, preservative, or clot activator/serum separator. Following a 30-minute clotting period, the samples were centrifuged at 3000rpm (1734g) for 10 minutes. Then serum aliquots were placed in standard cryovials and stored at −70°C for batch analysis.

Metabolite analysis:

Our protocol for metabolite analysis has previously been published.[20, 25, 26] Briefly, we performed targeted metabolite analysis of 165 metabolites with relative quantification using liquid chromatography-tandem mass spectrometry (LC-MS/MS). Samples were analyzed using an Agilent 1200 series HPLC system (Agilent Technologies, Palo Alto, CA) interfaced with an ABSciex 5500 hybrid triple quadrupole/linear ion trap mass spectrometer (Concord, ON, Canada) equipped with an electrospray ionization source operating in the positive/negative switch mode.

Statistics:

Prior to data analysis, data distributions were assessed using the Shapiro-Wilk test. Clinical data were then summarized as either means (standard deviation-SD) or medians (interquartile range-IQR) for continuous variables depending on the data distribution or counts (percent) for categorical variables. Fisher’s exact test was used for comparison of categorical variables while Kruskal-Wallis/Wilcoxon rank-sum testing or one-way ANOVA/Student’s t-test were utilized for comparison of continuous variables as appropriate for the data distribution and number of groups. P-value <0.05 was considered statistically significant. JMP Pro 14.2.0 (SAS Institute Inc., Cary, NC) was used for clinical data analysis.

Metabolomic data were analyzed as previously published [20] using Metaboanalyst 4.0, a web-based metabolomics analysis tool.[27] Data were first log-transformed and auto-scaled (mean centered and divided by the square root of the standard deviation of each variable). We then utilized partial least squares-discriminant analysis (PLS-DA) to evaluate the ability to differentiate among AKI groups based on their metabolomic profile. R²Y (goodness of fit) and Q²Y (cross-validation) are reported for each model. The pathway analysis tool (based on the R-package GlobalTest) was utilized to assess for differences in specific metabolic pathways. Differences in individual metabolites were evaluated using Student’s t-test with an adjusted False Discovery Rate (FDR) of <0.05 considered significant. Figures were created using Metaboanalyst 4.0.

Results:

A total of 57 subjects were included in this secondary analysis: 55 with both preoperative and 24hr postoperative metabolomic data and an additional 2 with only 24hr metabolomic data (no preoperative sample available for analysis). Baseline and demographic data for the full cohort [22] and the metabolomics sub-cohort [20] have previously been published and there are no significant clinical differences between the full parent cohort and the metabolomics sub-cohort.[20] Thirteen subjects (23%) developed mild AKI (KDIGO stage 1) and six subjects (11%) developed moderate to severe AKI (KDIGO stage 2/3) within the first 72hr postoperatively. A single subject progressed from stage 1 AKI at 72hr postoperatively to stage 3 AKI at 7 days postoperatively, otherwise all subjects demonstrated their highest AKI stage within the first 72hr. All subjects met creatinine change as the criteria for their maximum KDIGO stage (no subject met a higher KDIGO stage based on urine output alone). Baseline characteristics by KDIGO stage are shown in Table 1. We found no significant differences in baseline or operative characteristics among subjects with stage 0, 1, and 2/3 AKI, although there was a trend towards increased bypass time and aortic cross clamp time in the subjects with AKI.

Table 1:

Baseline characteristics by postoperative KDIGO stage

Clinical Characteristic	KDIGO Stage 0 (n=38)	KDIGO Stage 1 (n=13)	KDIGO Stage 2/3 (n=6)	p-value
Female sex; n (%)	20 (53%)	4 (31%)	3 (50%)	0.38
Age at surgery-days; median (IQR)	8.5 (5, 71.5)	32 (4, 78.5)	47.5 (7.5, 97)	0.54
Preterm; n (%)	5 (13%)	2 (15%)	2 (33%)	0.38
Weight-kg; median (IQR)	3.5 (3.0, 4.3)	3.3 (2.8, 4.5)	3.6 (3.1, 5.3)	0.57
Aristotle score, comprehensive; mean (SD)	10.5 (3.7)	10.8 (3.0)	10.2 (2.8)	0.93
Cardiopulmonary bypass time-minutes; median (IQR)	133 (112, 176)	201 (108, 271)	200 (138, 332)	0.05
Aortic cross-clamp time-minutes; median (IQR)	75 (60, 99)	82 (54, 145)	90 (81, 152)	0.22
Deep hypothermic circulatory arrest time-minutes; median (IQR)	0 (0, 8)	0 (0, 9)	3 (0, 7)	0.98
Selective cerebral perfusion time-minutes; median (IQR)	0 (0, 0)	0 (0, 65)	0 (0, 24)	0.51
Single ventricle physiology; n (%)	12 (32%)	6 (46%)	3 (50%)	0.45
Preoperative serum creatinine (mg/dL); median (IQR)	0.43 (0.32, 0.53)	0.35 (0.26, 0.64)	0.30 (0.27, 0.39)	0.13

We first assessed the ability of the circulating metabolic profile to discriminate among patients by subsequent AKI severity. The preoperative metabolic profile did not discriminate well between subjects who did or did not subsequently develop AKI in the first 72hrs postoperatively (R²Y=0.41; Q²Y=−0.37). By 24hrs postoperatively, the metabolic profile provided moderate differentiation between subjects without AKI and those who subsequently developed moderate to severe AKI (R²Y=0.67; Q²Y=0.09) (Figure 1a).

Fig. 1.

A) Subjects with moderate to severe AKI (KDIGO stage 2/3-green) demonstrate a distinct metabolic profile compared to subjects with no AKI (KDIGO 0-red) by PLS-DA. B) Addition of KDIGO 1 subjects (KDIGO 0-red, KDIGO 1-green and light blue, and KDIGO 2/3-dark blue). KDIGO 1 subjects did not form a distinct third class but instead segregated into two groups overlapping with either the KDIGO 0 subjects (red outline-mild metabolic phenotype) or KDIGO 2/3 (blue outline-severe metabolic phenotype).

Interestingly, the metabolic profiles of subjects with mild AKI did not demonstrate a third distinct profile. Instead the metabolic profiles of stage 1 AKI patients segregated into two groups overlapping either with subjects demonstrating no AKI (subsequently referred to as mild metabolic phenotype) or moderate to severe AKI (subsequently referred to as severe metabolic phenotype) (Figure 1b). We therefore further assessed the clinical characteristics of mild AKI subjects, divided based on their overlap with the mild or severe metabolic phenotype. These results are shown in Table 2. Although all subjects met stage 1 AKI by creatinine change criteria, absolute peak postoperative creatinine was significantly higher in the group of subjects with the severe metabolic phenotype. This group included the subject who later progressed to stage 3 AKI by postoperative day 7. There was also a trend towards higher preoperative creatinine as well as a trend towards higher-risk perioperative characteristics (younger, lower weight, higher complexity score, longer bypass time) in the subjects with stage 1 AKI and the severe metabolic phenotype.

Table 2:

Clinical characteristics of subjects with KDIGO stage 1 criteria divided based on their metabolic overlap with the stage 0 (mild metabolic phenotype) or stage 2/3 (severe metabolic phenotype) groups.

	KDIGO Stage 1 (n=13)		p-value
	Mild Metabolic Phenotype (n=6)	Severe Metabolic Phenotype (n=7)
Female sex; n (%)	2 (33%)	2 (29%)	0.99
Age at surgery-days; median (IQR)	49 (25, 101)	7 (4, 59)	0.22
Weight-kg; median (IQR)	3.9 (3.3, 4.7)	2.9 (2.4, 3.3)	0.06
Aristotle score, comprehensive; mean (SD)	9 (3)	12 (2.7)	0.11
Cardiopulmonary bypass time-minutes; median (IQR)	108 (81, 234)	212 (172, 274)	0.11
Aortic cross-clamp time-minutes; median (IQR)	74 (44, 149)	82 (61, 147)	0.67
Preoperative creatinine (mg/dL); median (IQR)	0.3 (0.19, 0.38)	0.64 (0.27, 0.75)	0.05
Peak postoperative creatinine (mg/dL); median (IQR)	0.47 (0.28, 0.55)	0.87 (0.48, 1.04)	0.04

We next sought to determine which individual metabolites and metabolic pathways that differed between the mild and severe circulating metabolic phenotypes. Twenty-one of the 165 metabolites measured (13%) differed significantly between the severe metabolic phenotype and the mild metabolic phenotype corrected for multiple comparisons with an FDR<0.05 (Table 3). Multiple circulating metabolites also showed at least moderate ability to identify subsequent moderate to severe clinical AKI. Over 20 metabolites demonstrated an area under the curve (AUC) of >0.75 for differentiating between subjects with stage 0 AKI versus stage 2/3 AKI. The combination of methylmalonic acid, kynurenic acid, 4-aminobenzoic acid (PABA), and pyrophosphate demonstrated the strongest differential ability (AUC=0.92; 95% CI 0.77–1), with other combinations of metabolites also achieving AUCs>0.90. While a comparison to protein biomarkers was not a primary aim of this study, it is notable that 24hr serum NGAL levels (previously published in the parent cohort [22]) demonstrated at most modest discriminating power between subjects with stage 0 versus stage 2/3 AKI in this secondary analysis cohort (AUC=0.67).

Table 3:

Differential metabolites between the mild and severe AKI metabolic phenotypes. Fold change defined as severe phenotype/mild phenotype.

Metabolite	Fold Change	Raw p-value	FDR
Purine Metabolism
Allantoate	1.8	2.3×10−4	0.008
Uric acid	0.79	3.7×10−4	0.008
Xanthine	0.67	0.002	0.026
Cysteine/Methionine/Taurine Metabolism
Serine	0.69	3.4×10−4	0.008
Cystathionine	2.0	0.003	0.027
S-methyl-5-thioadenosine	1.4	0.004	0.030
S-adenosyl-L-homocysteine	1.4	0.007	0.048
Taurine	0.62	0.003	0.026
Kynurenine/Nicotinamide Metabolism
Kynurenic acid	2.5	7.5×10−4	0.014
1-Methylnicotinamide	2.1	3.0×10−4	0.008
Trigonelline	1.8	0.003	0.027
Aspartate	0.72	0.001	0.016
Pyrimidine Metabolism
Methylmalonic acid	2.2	5.8×10−6	8.7×10−4
Uridine	0.70	0.004	0.032
Amino Acid Acetylation
N-acetyl-alanine	1.8	2.3×10−4	0.008
Acetyl-L-lysine	1.8	0.003	0.026
Arginine Metabolism
Putrescine	1.7	0.001	0.016
Creatinine	1.3	4.3×10−5	0.003
Other
4-aminobenzoic acid (PABA)	0.66	0.001	0.017
PPi	1.4	0.001	0.016

Affected metabolic pathways are shown in Figure 2. A total of 25 metabolic pathways differed significantly between the mild and severe metabolic phenotypes with the most prominent differences found in cysteine/methionine metabolism, kynurenine/nicotinamide metabolism, and purine metabolism (FDR=0.002). Pathway analysis includes all measured metabolites in the pathway and is influenced both by metabolites of high and intermediate significance. For example, significance of the cysteine/methionine pathway is not only determined by the metabolites listed in table 3, but also decreased methionine (FDR=0.13) and increased homocysteine (FDR=0.1), both individually of intermediate significance but collectively indicating modulation of this pathway in AKI.

Fig. 2.

Affected pathways in severe AKI metabolic phenotype. The x-axis and size of circles represent impact of differential metabolites within the pathway. The y-axis and color of circles represent statistical significance of the overall metabolic changes within the pathway.

Discussion:

Key Findings

In this study we present the novel findings of a distinct circulating metabolic profile associated with severe AKI following infant cardiothoracic surgery with CPB. While the targeted metabolic profiling strategy we employed does not support complete mapping of individual metabolic pathways, we did find preliminary evidence for differential modulation of multiple pathways in subjects with the severe metabolic phenotype. The most prominently affected pathways in these subjects included purine metabolism, cysteine/methionine metabolism, and kynurenine/nicotinamide metabolism. Furthermore, we found that our subjects with mild AKI by KDIGO criteria did not represent a third distinct metabolic phenotype. Instead they segregated into two groups, overlapping with the metabolic phenotypes of subjects with no AKI or moderate/severe AKI. These findings suggest heterogeneity among patients with mild clinical AKI, which could potentially be further characterized using a metabolomics approach.

Metabolomic Profiling in Pediatric Cardiac Surgery

Study of the metabolome in pediatric cardiology and cardiac surgery remains in its early stages, focused primarily on metabolomic profiling. Correia, et al performed a pilot study of the circulating metabolomic profile in 28 mixed-age children undergoing cardiac surgery.[12] They demonstrated an evolution of the global, untargeted metabolomic profile through 48hrs postoperatively, while a targeted panel of 15 metabolites provided moderate discrimination among subjects based on their level of postoperative support requirements (ICU-free days, inotrope score, and mechanical ventilation). Our group subsequently completed a larger study of infants undergoing cardiac surgery with CPB, examining a targeted panel of 165 circulating metabolites preoperatively, with rewarming from CPB, and at 24hrs postoperatively.[20] In this higher-risk age group, the circulating metabolome was highly disrupted, with 84% of metabolites differing significantly from baseline at 24hrs postoperatively. Pathway analysis provided evidence for an association between early disruption of aspartate/glutamate, nicotinamide, and tryptophan/kynurenine metabolism and both ICU length of stay and survival. Recently Simonato, et al performed untargeted metabolic profiling of urine from 14 neonates undergoing repair of d-transposition of the great arteries.[28] Interestingly, metabolites from the kynurenine arm of tryptophan metabolism again differed prominently between pre and postoperative samples, demonstrating the potential significance of this pathway, although the study was underpowered to assess associations with outcomes.

Based on our findings and those of Simonato, we undertook complete pathway mapping of the kynurenine arm of tryptophan metabolism through 48hrs postoperatively in a cohort of infants undergoing stage 2 palliation for single ventricle reconstruction (Glenn or hemi-Fontan).[29] We found early increases in the proximal pathway metabolites (kynurenine, kynurenic acid, anthranilate, and 3-OH-kynurenine), but relative deficiencies in the distal metabolites, including quinolate, the primary substrate for de novo NAD production. By 48hrs, most intermediates had normalized; however, both kynurenine and kynurenic acid remained significantly elevated, suggesting continued exposure to the vasodilatory and immunosuppressive effects of these metabolites. To our knowledge this study was the first application of the framework proposed by Johnson, et al [21] to pediatric cardiology, where an initial metabolomic profiling study led to a subsequent targeted validation study using comprehensive pathway mapping. In the current study we extend the concept of metabolomic profiling to the outcome of postoperative AKI, with the goal of identifying pathways and individual metabolites that are differentially disrupted in patients with postoperative AKI and are candidates for future targeted validation and pathway modulation studies.

Metabolomic Profiling in AKI

Studies in small animal models have established that both isolated kidney ischemia-reperfusion injury (IRI) and sepsis result in significant changes in the kidney tissue, urinary, and circulating metabolomes.[18, 30–32] These metabolic changes were broadly characterized by the authors as related to perturbed energy metabolism, oxidative stress, osmotic regulation, and purine metabolism.[30, 32] Using a porcine model of E. coli sepsis-induced AKI, Izquierdo-Garcia, et al also demonstrated differences in the kidney, urinary, and circulating metabolic profiles by nuclear magnetic resonance (NMR) spectroscopy in septic animals compared to uninfected controls.[33] These preclinical studies, taken together, demonstrate the potential utility of metabolomic profiling to the study of AKI. However, differences in study design and metabolite analysis techniques make direct comparison across these studies or application to cardiac surgery-induced AKI challenging.

Additional research focused on cardiac surgery cohorts and large animal models of cardiac surgery/CPB is therefore needed to understand the metabolic changes associated with cardiac surgery-induced AKI. To our knowledge our study is the first to evaluate changes in the circulating metabolome associated with AKI after pediatric cardiac surgery. Zacharias, et al previously evaluated both circulating and urinary metabolic changes following adult cardiac surgery with CPB using ¹H NMR spectroscopy.[34] While these studies evaluated a limited number of metabolites, they demonstrated moderate capability at 24hrs postoperatively to identify subjects who would subsequently develop moderate or severe AKI by 72hrs postoperatively. The AUC for both urinary and circulating metabolites were noted by the authors to be higher than commonly used clinical AKI biomarkers including NGAL, cystatin C, and kidney injury molecule-1 (KIM-1). Our study similarly demonstrated high AUCs for individual metabolites and metabolite combinations in the identification of subjects who would progress to KDIGO stage 2/3 disease. Collectively, these studies provide evidence that metabolic profiling strategies could be used to identify patients at risk for subsequent KDIGO stage 2/3 disease following cardiac surgery, although further work is needed prior to translation into clinical medicine.

Interestingly, Zacharias, et al noted that the circulating metabolic profile of KDIGO stage 1 disease was not distinguishable from KDIGO stage 0 disease and attributed this to the mild nature of KDIGO 1 disease in adults.[34] In our infants, we also found that the circulating metabolic profiles of some patients with KDIGO 1 disease mirrored those of patients who did not experience postoperative AKI. However, a portion of our KDIGO stage 1 patients demonstrated a metabolic profile similar to patients who subsequently developed higher grade AKI. Notably, the KDIGO stage 1 patients demonstrating the more severe metabolic phenotype had significantly higher absolute creatinine levels. The complexities of interpreting early changes in postoperative creatinine have been well documented.[35] Our findings suggest that infants with KDIGO stage 1 disease and higher absolute peak creatinine may represent a different population, at risk for more significant metabolic disturbances compared to those with a rise in serum creatinine but a low absolute creatinine level. Further study is warranted to assess if additional metabolite testing can help improve prognostic testing for KDIGO stage 1 infants as well as the implications of this metabolic phenotype on outcomes.

Purine Metabolism

Uric acid is the end product of purine metabolism in humans.[36] Preclinical and adult studies demonstrate an association between higher uric acid levels and AKI.[37, 38] Potential mechanisms underlying this association include renal vasoconstriction, inflammation, and impaired autoregulation.[37] Surprisingly in our study, lower circulating levels of uric acid and its immediate precursor, xanthine, were associated with subsequent development of moderate/severe AKI. One possible explanation for this finding is increased postoperative oxidative stress in these subjects, leading to both uric acid depletion and end-organ injury. Uric acid is a potent antioxidant, making up between 35–65% of plasma antioxidant capacity.[39] It is capable of forming stable complexes with Fe²⁺, thereby inhibiting the Fenton reaction, or can be oxidized to allantoin, which is efficiently excreted in the urine.[39, 40] Uric acid levels decrease in kidney tissue following LPS-induced AKI in rats.[31] Interestingly, allantoate also increased in this model as it did in our subjects with the severe AKI metabolic phenotype and may represent further metabolism of allantoin produced through uric acid oxidation. To our knowledge, increased allantoate has not been previously reported in humans. Hypouricemia may also have direct impacts on cardiovascular risk through unfavorable modulation of endothelial function and resultant hypotension.[41] Additional studies appear warranted to assess the importance of uric acid in the modulation of oxidative stress and endothelial function after congenital heart disease surgery.

Kynurenine and Nicotinamide Metabolism

Kynurenine and nicotinamide metabolism are linked metabolic pathways that ultimately result in de novo production of nicotinamide adenine dinucleotide (NAD). The kynurenine pathway (KP) serves as the primary route for tryptophan catabolism, resulting in production of quinolinic acid, the precursor of the nicotinamide pathway. KP metabolites are highly active with a host of biologic properties including immune suppression, inhibition of nitric oxide (NO) synthase, regulation of oxidative stress, modulation of vasomotor tone, alteration of mitochondrial function, and both NMDA agonism and antagonism. Our group has previously demonstrated a significant activation of the KP after congenital heart disease surgery,[29] and elevated circulating kynurenic acid was associated with increased ICU length of stay and cardiovascular failure (cardiac arrest, ECMO, or death) in our patients.[20] Activation of the KP has previously been demonstrated in clinical studies of AKI, although lack of comprehensive pathway mapping makes comparison of studies challenging. Aregger, et al identified urinary kynurenic acid as the single best independent predictor of kidney recovery in critically ill adults with AKI,[42] while Dabrowski, et al found that failure to clear plasma kynurenic acid with continuous veno-venous hemofiltration was strongly associated with mortality in septic patients.[43] Kynurenine 3-monooxygenase (an intermediate KP enzyme) knockout mice are protected from ischemia reperfusion AKI, with increased kynurenic acid production and decreased downstream 3-hydroxykynurenine.[44] Direct administration of kynurenic acid in a rat model of ischemic AKI was also protective,[45] and it has been suggested that these collective findings may be the result of kidney NMDA receptor antagonism by kynurenic acid but agonism by downstream KP metabolites including quinolinic acid.[46] Further studies are needed to fully define the key KP metabolites involved in the response to ischemia-reperfusion related AKI.

Nicotinamide metabolism occurs downstream of the KP and is responsible for NAD production via three complementary routes: de novo synthesis from the KP, salvage from circulating nicotinamide, and conversion from dietary niacin. NAD is the rate limiting factor for oxidative metabolism and is critical for kidney health, particularly in the proximal tubules where significant energy production is required to meet solute reabsorption demands.[47] Experimental AKI leads to tissue NAD depletion through a combination of increased consumption and decreased de novo synthesis.[48] While our metabolite screening panel does not include NAD itself, multiple circulating metabolites in the nicotinamide pathway are altered after infant cardiac surgery. We previously have shown significant depletion of nicotinamide, methylnicotinamide, niacin, and trigonelline accompanied by an increase in NADPH at 24hrs postoperatively.[20; supplemental data file] Interestingly, however, it was the subset of patients with higher methylnicotinamide and trigonelline levels that were at highest risk for subsequent clinical AKI in our study. Methylnicotinamide and trigonelline are end products of nicotinamide and nicotinate respectively, methylated to allow urinary excretion. Methylnicotinamide has previously been studied as a marker of both glomerular filtration and proximal tubular secretion,[49] and it is likely that the increase in circulating levels of these metabolites reflect early changes in their clearance. The importance of increased circulating levels, however, is unclear, as these metabolites were long thought to be biologically inert. Recent studies have shown that methylnicotinamide and trigonelline have significant biologic activities as vasodilators and modulators of immune function through NO production, activation of COX-2, and prostaglandin synthesis[50] as well as regulators of NAD production.[51] Therefore, in addition to serving as potential biomarkers, it is possible that they could modulate the pathologic postoperative state that leads to AKI. Translational studies to measure changes in this pathway at the tissue level and to evaluate the effects of circulating pathway intermediates on the kidneys would be useful to further understand the importance of this pathway on the development of postoperative AKI.

Cysteine/Methionine Metabolism

Alterations in methionine metabolism to cysteine via homocysteine and cystathionine have previously been described in both chronic kidney failure in humans [52] as well as in animal models of AKI.[53] The most common pattern includes normal to low levels of methionine and serine and increased levels of S-adenosylhomocysteine, homocysteine, and the products of homocysteine transsulfuration (cystathionine and cysteine). Our findings are consistent with this pattern, with the exception that cysteine was not analyzed on our screening profile. While it was originally thought that elevated circulating levels of homocysteine were purely secondary to decreased clearance, the alternative hypothesis of a mechanistic link between methionine metabolism and kidney injury is gaining increased traction.[52] One possible explanation is a block in homocysteine remethylation to methionine due to altered folate metabolism in kidney failure, resulting in increased flow through the transsulfuration pathway.[52] Additionally experimental models of AKI have demonstrated downregulation of transsulfuration pathway enzymes including cystathionine γ-lyase, the final step of the pathway, potentially leading to accumulation of more proximal metabolites.[52, 53] These intermediate metabolites, particularly homocysteine and S-adenosylhomocysteine, contribute to endothelial dysfunction, and may result in increased cardiovascular risk across a range of pathologies including kidney injury.[52] Based on our findings, additional studies of methionine metabolism in AKI appear warranted to further explore the role of this pathway in injury progression and associated vascular pathology.

Limitations and Future Directions

Our study has several limitations to acknowledge when interpreting the results and determining future studies. First, we recognize that this study design, like all observational and biomarker studies, only identifies associations and cannot definitively evaluate either the cause of altered metabolite levels or the kidney/systemic effects of these changes. Follow-on clinical and translational studies using pathway mapping, organ-specific measurements, and pathway modulation are needed to determine the mechanism, directionality, and clinical impact of these metabolic changes. Second, our sample size is relatively small, especially with regards to severe AKI patients. While the fact that we identified this many significantly dysregulated metabolites in AKI patients even with such a small sample size highlights the large differences in the circulating metabolome associated with severe AKI, we were undoubtedly underpowered to detect more subtle metabolite differences with this sample size. Third, this is a single center study and it is possible that perioperative management could vary across institutions leading to differential postoperative metabolic responses. Additional studies including other major pediatric cardiac centers are needed to determine generalizability of the results. Generalizability to other patient populations including adult cardiac surgery patients and non-surgical ICU patients also is unknown. Finally, since this study was a planned secondary analysis of a parent cohort study, the parent cohort study was not designed to comprehensively evaluate the metabolic phenotype of postoperative AKI. Specifically, future studies should include additional time points (both earlier and later in the postoperative course), older subjects, and increased number of subjects with AKI in order to increase study power. Additional study is also warranted to understand the mechanism and clinical importance of elevated preoperative creatinine in a subset of our neonatal patients.

From a metabolite analysis standpoint, we chose to use a targeted MS-based metabolite profiling approach. While this strategy has many strengths including analysis of only verified metabolites, increased sensitivity for the identification of low-concentration metabolites, focus on key metabolic pathways, and increased statistical power through reduced number of comparisons, it also has drawbacks including relative rather than absolute quantification, incomplete pathway mapping, and difficult comparison to both untargeted and NMR-based studies. Fully quantified pathway mapping of the metabolic pathways identified in this study represents a critical next stage of translational research in this population. Finally, our study analyzed only blood samples from patients, and we could therefore only assess circulating metabolites. We were also unable to assess the effects of metabolite pathway modulation on AKI. Additional clinical and translational studies of pediatric cardiac surgery are needed to both analyze metabolite changes in different compartments (blood, kidney tissue, and urine) and to modulate these pathways, providing key mechanistic insights and identifying future metabolic biomarker and therapeutic targets.

Conclusions

Development of moderate-to-severe clinical AKI in infants undergoing CPB is preceded by significant changes in the metabolome at 24hrs postoperatively. This severe metabolic phenotype includes prominent changes to purine, methionine, and kynurenine/nicotinamide metabolism. These metabolic pathways warrant further study both as predictive biomarkers of severe AKI and as potential mechanistic targets to prevent AKI and/or mitigate the systemic metabolic repercussions of AKI. A portion of infants with mild AKI demonstrated similar metabolic changes, suggesting a potential role for metabolic analysis in the evaluation of lower-stage injury.