Background
Acute kidney injury (AKI) occurs in approximately half of adult critically ill patients [
1‐
9]. Besides its recognized adverse effect on individual patient outcomes, both in the short- and long-term [
1,
2,
4‐
10], AKI causes an important socioeconomic burden ensuing from its relationship with the development of chronic kidney disease (CKD) [
11], and end-stage renal disease requiring renal replacement therapy (RRT) [
12].
Current diagnostic and staging criteria for AKI were defined by the Kidney Disease | Improving Global Outcomes (KDIGO) AKI work group (Additional file
1: Table S1) and require monitoring of two surrogate glomerular filtration rate (GFR) markers, i.e., serum creatinine (SCr) and urine output (UO), and of the intervention RRT [
13]. As renal stress and damage to the kidneys precede the observed decline in GFR [
14], diagnostic AKI biomarker research in the last decade has focused on detection of these early signals [
15‐
17]. Studies have shown that urinary biomarkers like neutrophil gelatinase-associated lipocalin (NGAL) [
18‐
22], and recently the panel tissue inhibitor of metalloproteinases 2 (TIMP-2) and insulin-like growth factor-binding protein 7 (IGFBP7) [
21,
23,
24], can detect AKI in critically ill patients earlier than SCr or UO, even when using the most sensitive KDIGO criteria. In addition, these biomarkers may also allow detection of other outcomes such as progression of AKI, use of RRT, development of CKD, and long-term mortality [
25,
26]. The complexity of the AKI syndrome and the interest in detecting other outcomes highly warrants evaluation of yet further candidate renal stress or damage biomarkers aiming to find either complementary or - less realistically - truly superior ones.
Recently, our group discovered chitinase 3-like protein 3 (CHI3L3) as a novel candidate biomarker for sepsis-induced AKI, by urinary proteomics [
27‐
29]. Validation with western blot analysis confirmed the presence of CHI3L3 in urine of septic mice with AKI, and its absence in urine of septic mice without AKI. In view of translational research [
30], two other members of the number-18 glycoside hydrolase-family [
31], i.e., chitinase 3-like protein 1 (CHI3L1) and acidic mammalian chitinase (CHIA), showed similar results. Subsequently, we found that CHI3L1 measured in urine was more discriminative for the presence of AKI in human septic patients than CHIA [
28].
The number-18 glycoside hydrolase-family is special in that it comprises catalytically inactive proteins such as chitinase-like proteins (CLP, e.g., CHI3L1) in addition to catalytically active proteins such as chitinases [
31]. These CLPs function as lectins because they can bind, but not hydrolyze, the glycan chitin, and therefore, represent the chi-lectin subfamily [
32,
33].
The objective of this study was first to evaluate the diagnostic performance of the urinary biomarker CHI3L1 for early detection of AKI stage ≥2 in adult critically ill patients, and then to compare this performance to that of NGAL, which was chosen as the reference urinary biomarker.
Methods
We followed recommendations for strengthening the reporting of observational studies in epidemiology (STROBE) (Additional file
1: Table S2) [
34]. Details of the methods are provided in Additional file
1: Text S1 and Tables S3A-F. The methods for additional analyses not included in the manuscript are provided in Additional file
1: Text S2 and Tables S4A and B.
We will refer to AKI that was diagnosed and classified by KDIGO as AKISCr/UO, while AKISCr will imply that the KDIGO UO criteria were discarded.
Study population
We conducted a prospective cohort study at the 22-bed surgical and 14-bed medical intensive care units (ICU) of Ghent University Hospital from September 2012 till August 2014. The inclusion and exclusion criteria are shown in Table
1.
Table 1
Inclusion and exclusion criteria of the study
Age ≥18 y | AKISCr/UO stage ≥2 at time of enrollmenta
|
Presence of both arterial and urinary catheter |
Expected ICU stay ≥48 h |
Respiratory SOFA score ≥2 (PaO2/FiO2 < 300) or cardiovascular SOFA score ≥1 (MAP <70 mmHg or on vasopressor(s) for at least 1 h) | CKD KDOQI stage 5 (GFR <15 ml/min/1.73 m2 or RRT)b
|
Written informed consent |
Ethics, consent and permissions
This study was approved by the Ethical Committee of the Ghent University Hospital (Belgian registration number of the study: B670201213147), and conducted in accordance with the declaration of Helsinki and in compliance with the Good Clinical Practice Guidelines. All patients or their legally authorized representatives provided written informed consent.
Sample collection, sample handling, and data collection
Blood and urine were collected at enrollment. The large majority of patients i.e., 89 % was enrolled on either the first (28 %) or second (61 %) ICU day, while the minority i.e., 11 % was enrolled on either the third (9 %) or fourth (2 %) ICU day. Each subject was sampled a second time on the day of enrollment (d1) at 6 pm if the first collection was before noon. The subsequent sampling times were at 6 am and 6 pm on d2–4, and at 6 am on d5–7 (Additional file
1: Table S3A). This is similar to the methodology used in the hallmark study by Kashani et. al. [
21]. If the patient was discharged from the ICU before d7, the study stopped.
These paired blood and urine samples were collected by standard methods and centrifuged by standard protocols. Serum and urine supernatants were stored at −80 °C and thawed at room temperature immediately prior to analysis. Clinical data needed to complete the individual clinical research files were extracted from the hospital records by study coordinators. Clinical data and samples were anonymized. JDL had access to the anonymized SCr and serum C-reactive protein data in order to determine the appropriate sample dilution for CHI3L1 measurement by enzyme-linked immunosorbent assay (ELISA) (Additional file
1: Table S3B). All other technicians were blinded to clinical data.
Biomarker measurements
Creatinine and urinary NGAL (UNGAL) analyses were performed externally. Creatinine concentrations were measured with a kinetic rate-blanked Jaffé assay (commercial reagents, Roche Diagnostics, Basel, Switzerland) on a Cobas c502, while UNGAL concentrations were measured with a particle-enhanced turbidimetric immunoassay (ST001-3CA, BioPorto, Hellerup, Denmark) on a Modular P. The concentration of CHI3L1 was determined in-house with a sandwich ELISA (DC3L10, R&D Systems, Minneapolis, MN, USA).
Both urinary CHI3L1 (UCHI3L1) and UNGAL concentrations were statistically analyzed as such and after correction for urine dilution by using the ratio to urinary creatinine (UCr). The relative change in SCr measured at enrollment was defined as the ratio of the enrollment SCr to reference SCr. The UO after enrollment, defined as the mean UO in the first valid 6-h period after enrollment, was determined as the mean of the 6 UO values that were calculated each h in the first valid 6-h period after enrollment.
Primary endpoint
The primary endpoint of the study was the development of AKI
SCr/UO stage ≥2 within 12 hours (h) after enrollment (Additional file
1: Table S1). Reference SCr was defined as the lowest SCr value within the last 3 months (mo) prior to enrollment. The details for calculation of UO are outlined in Additional file
1: Text S1.
Secondary endpoints
Secondary endpoints of the study were: AKI
SCr/UO stage ≥2 within 24 h and 7 days after enrollment; AKI
SCr stage ≥2 within 12 h, 24 h and 7 days after enrollment (Additional file
1: Table S1).
UCHI3L1 response to AKI
We compared samples that were collected in the 24 h preceding diagnosis of the first episode of AKISCr/UO stage ≥2 to those that were not followed by a first episode of AKISCr/UO stage ≥2 within the next 24 h. For this analysis, we excluded all samples collected in the period starting from diagnosis of the first episode of AKISCr/UO stage ≥2 till the end of the study.
For all 21 patients who developed AKISCr/UO stage ≥2 (reference time 0 h) within 7 days after enrollment, we documented the UCHI3L1 concentrations corresponding with the time points 24 h before, 12 h before, 12 h after, and 24 h after diagnosis of the first episode of AKISCr/UO stage ≥2. This allowed us to investigate the distribution of UCHI3L1 over time in patients with AKISCr/UO stage ≥2.
We also studied the distribution of UCHI3L1 in samples corresponding with different stages of severity of AKISCr/UO. If the total study period of 7 days was completed, 11 serum and 11 urine samples were available per ICU patient. All available UCHI3L1 concentrations were classified according to their AKISCr/UO stage at that moment. As such, UCHI3L1 concentrations were divided into four groups: no AKISCr/UO at the time of sampling, and AKISCr/UO stages 1, 2, or 3 at the time of sampling.
Statistical analysis
The primary analysis was based on comparison of the areas under the receiver-operating characteristics curves (AUC-ROC) of UCHI3L1 with those of UNGAL for predicting the defined endpoints, which was performed in MedCalc 15.2.1 (MedCalc Software, Oostende, Belgium). We also calculated Spearman’s coefficients of rank correlation with this program. In SPSS 22 (IBM, Armonk, NY, USA) we performed (1) mixed model analysis with log
10(UCHI3L1) as the outcome variable; diagnosis of the first episode of AKI
SCr/UO stage ≥2 within 24 h after sampling, as the predictor variable; and patient as the random factor; (2) Fisher’s exact or the chi-square test - the 95 % confidence interval (CI) of a proportion was calculated with the Wilson procedure without correction for continuity [
35,
36] - and the Mann–Whitney
U test; (3) the Wilcoxon matched-pair signed-rank test; and (4) related-samples Friedman’s two-way analysis of variance by ranks. Box and whisker plots were generated in GraphPad Prism 5 (GraphPad Software, San Diego, CA, USA). For all analyses, two-sided
P values <0.05 were considered statistically significant.
In Additional file
1: Text S1 and Tables S3C-F, we provide all details and also describe how the urinary biomarkers were introduced into the statistical programs.
Discussion
We found that UCHI3L1 was a good biomarker for early detection of AKI stage ≥2 in adult critically ill patients admitted to an ICU, with a performance similar to that of UNGAL.
These findings may have important clinical and socioeconomic impact. Increasing severity of AKI is associated with increased risk of worse patient and kidney outcomes [
1,
2,
4‐
10]. Importantly, observational and also intervention studies showed that early AKI management can counteract AKI deterioration, and is associated with lower mortality and less RRT dependence at discharge [
37‐
44]. Consequently, even earlier identification of AKI using a biomarker may have a much stronger effect on these outcomes.
Both UCHI3L1 and UNGAL better predicted AKI stage ≥2 that was defined based on SCr alone versus based on SCr or UO. These two urinary proteins are biomarkers of renal stress or damage [
45,
46], while SCr and UO are GFR surrogates. However, UO is much more sensitive to decline in GFR, and therefore is probably associated with less renal stress or damage than SCr, which is supported by studies reporting that UO-based AKI classes are associated with a lower ICU/hospital mortality than SCr-based ones [
9,
47,
48]. This may explain the better AUC-ROCs when considering SCr alone for diagnosis. The findings by Macedo et al. [
49], who reported similar ICU mortality for exclusively UO
+ AKI patients (8.8 %) and (non)oliguric SCr
+ AKI patients (10.4 %), appear contradictory to previous findings [
9,
47,
48]. However, severity of AKI was greater in exclusively UO
+ patients: >60 % of these patients were stage 2, while >70 % of the (non)oliguric SCr
+ patients were stage 1 [
49]. We also observed a partial overlap in UCHI3L1 between AKI
SCr/UO stage 1 and stage 2 samples, indicating heterogeneity of AKI severity within KDIGO classes, which can be partly explained by the different impact of meeting the defined criteria for either UO alone, or SCr alone, or both SCr and UO [
9]. This could also clarify the decreased performance of UCHI3L1 and UNGAL at enrollment for prediction of AKI
SCr/UO stage ≥2 within the next 24 h. The majority of the extra AKI
SCr/UO stage ≥2 patients in the 24 h observation period fulfilled UO criteria only. These patients, therefore, probably had less renal stress or damage, and consequently a low biomarker signal. Another explanation could be that the hit leading to AKI is following the biomarker measurement. This may more likely occur when the observation period is longer [
21].
The observation that the AUC-ROC for the individual KDIGO parameters, i.e., SCr and UO, were similar to those of UCHI3L1 for detection of AKI stage ≥2, warrants discussion. First, when comparing the AUC-ROC for UO and UCHI3L1, we must take into account that although the measurement of UO started at enrollment, it was only completed 6 h later than the time at which UCHI3L1 was measured. Second, renal stress or damage may not always be reflected by decline in GFR; vice versa, a decline in GFR may not always reflect renal stress or damage. This may lead to underestimation of the diagnostic performance of UCHI3L1 in our study.
We found a trend for increased UCHI3L1 concentrations in the 24 h preceding AKI, and for decreased concentrations afterwards. However, it should be emphasized that after meeting AKISCr/UO stage ≥2, the individual time–concentration profiles of AKISCr/UO differed widely between patients: some remained in the same severity stage, some deteriorated and others ameliorated (data not shown). The number of patients observed in this pilot study also precludes firm conclusions.
This is the first translational study demonstrating that UCHI3L1 predicts the occurrence of AKI stage ≥2 in adult critically ill patients [
28]. Schmidt et al. independently showed that UCHI3L1 predicts the occurrence of delayed graft function in adult patients who receive deceased-donor kidney transplants [
46]. In their preclinical study, these authors reported that the transcription of the
CHI3L1 gene is significantly upregulated in the mouse kidney after ischemia/reperfusion (I/R) injury with increased excretion of its protein in urine. These mRNA and protein levels correlated with the degree of kidney injury and were at earliest measured on the first day after I/R, when SCr values had already peaked. Recently, the same group also studied a cohort of hospitalized patients who had AKI, and found that UCHI3L1 was associated with the composite outcome of AKI progression and in-hospital death [
50].
We must speculate on the source of CHI3L1. Upon renal stress or damage, this protein is secreted by macrophages within the kidney [
46], while NGAL is secreted by specific cells of the distal nephron [
45]. Another source for the urine component of NGAL is the circulating plasma pool [
51]. We speculate that the same is true for CHI3L1 as this protein has an apparent molecular weight of ±39–40 kDa [
52,
53], and as within the group of patients with no AKI (in the 7-d prediction window) a concomitant very high level of SCHI3L1 was observed more in those with an increased than with a normal UCHI3L1 at enrollment (Additional file
1: Table S8). Additionally, we speculate that CHI3L1 binds to the megalin receptor for tubular reabsorption. This implies that NGAL and CHI3L1 can each indirectly affect the urinary concentration of the other, as they are then competitors [
51].
Similar to NGAL, CHI3L1 is also stored in the secondary granules of circulating neutrophils [
54‐
56]. This could implicate that in the urine of patients with a urinary tract infection (UTI), CHI3L1 is increased too [
57]. Although data on UCHI3L1 in UTI patients are missing, proteome profiling of human neutrophils suggests that this issue is less relevant for UCHI3L1 [
56], which agrees with the reported 12 pg NGAL and 0.16 pg CHI3L1 per leukocyte [
57,
58].
Surprisingly, only in 2013 He et al. investigated the possibility that the CLP CHI3L1 mediates its biological effects through receptor binding, and identified interleukin-13 receptor α2 as the binding partner [
59]. These biological effects include inhibition of apoptosis in renal epithelial cells [
46,
59], and inhibition of pyroptosis and interleukin-1β production in macrophages [
59,
60]. These innate immune cells play an important role in both kidney injury and repair [
61].
Our study has important limitations. First, this is a single-center study conducted in surgical and medical ICUs. Although the baseline characteristics of patients, the observed outcomes, and the NGAL cutoff based on the Youden index suggest that the patients included are representative of ICUs in developed Western countries, these data remain to be confirmed in other centers and in different types of ICU. Second, only a limited number of patients reached the primary endpoint, which can be partly explained by selection bias, i.e., not asking (legally authorized representatives of) the most critically ill patients for consent. Yet, this is a typical and hence, rather unavoidable feature of prospective studies like this [
21]. The restricted period for observation of AKI stage ≥2 certainly contributes to the low event rate as well. Therefore, we included all 21 patients (12 %) who developed AKI
SCr/UO stage ≥2 within 7 days after enrollment in the mixed model analysis. Third, following the KDIGO guidelines [
13], reference SCr was defined as the lowest SCr value within the last 3 mo prior to enrollment. This method is prone to bias, as blood draws for SCr measurement tend to be performed more often when patients are in hospital or sick, thereby not reflecting true baseline kidney function. Fourth, we did not measure urinary [TIMP-2]•[IGFBP7], a two-biomarker panel found to be superior to UNGAL [
21,
23,
24], because it was not available at the start of our study.
Competing interests
A patent application was filed on 4 April 2011 by Ghent University with E. Meyer and B. Maddens as inventors. The international patent application has been published as WO2012/136548. E. Hoste received a speaker’s fee from Astute Medical. For the remaining authors no competing interests were declared.
Authors’ contributions
Design and conduct of the study: EH designed the study in conjunction with EM, JDL and LN. EH was the Principal Investigator. Data collection: EH, JD and the other staff members of the Surgical and Medical Divisions of Intensive Care enrolled subjects. JDL and KD performed the laboratory analyses and analytically interpreted the results in conjunction with EM. EH and JDL clinically interpreted the data. Management: clinical data were managed by the study coordinators of the Ghent University Hospital Division of Intensive Care, under supervision of EH. Statistical analysis: JDL and EH performed the statistical analysis for the study. Interpretation of the data: all authors reviewed the data and participated in discussions related to interpretation. Preparation, review or approval of the manuscript: JDL wrote the paper. All authors reviewed and edited the paper and have seen and approved the final draft.