Processing of serum specimens

Blood was drawn by venipuncture within 48 h of hospital admission. Whole blood samples were immediately sent to Institute of Laboratory Medicine of the hospital for further routine diagnostics (see Supplementary Table S1). Regarding serum uromodulin (sUmod), serum aliquots of the supernatants were stored at -80 °C, until analyses.

Assessment of serum uromodulin concentrations

Serum uromodulin concentrations were measured in serum samples using a sensitive uromodulin ELISA, certified for serum analyses (EUROIMMUN Medizinische Labordiagnostika AG, Luebeck, Germany) and based on monoclonal antibodies as reported earlier^34,43.

Laboratory parameters at baseline

Laboratory parameters of the COVID-19 cohort in serum and urine are summarised in Table S1. At baseline serum creatinine levels were 107.2±69.5 µmol/l with mean eGFR (CKD-EPI) of 66.9±27.5 ml/min/1.73m², and a mean value of serum uromodulin of 178.7 (±97.8)ng/ml. Serum uromodulin levels strongly correlated with eGFR (correlation coefficient 0.585, CI 0.514to0.648, p<0.001). Furthermore, sUmod levels were significantly lower in patients with CKD (91.8±60.7 ng/ml) compared to patients with normal kidney function (204.7±91.7 ng/ml; p<0.001; Fig. 1A).

Open in a separate window

Fig. 1

sUmod levels at admission in COVD-19 patients with and without chronic kidney disease (CKD) (A), with and without acute kidney injury (AKI) (B) and survival and death (C). ROC-Curve as well as corresponding AUC and confidence interval for the association between sUmod and AKI (D). sUMOD: serum uromodulin.

Association between uromodulin and acute kidney injury (AKI) in COVID-19 patients

In total, of the 378patients included in this study, 151patients (40.0%) presented with an AKI at the time of hospital admission or developed an AKI during hospitalisation. Among these patients, 110patients (72.8%) had AKI AKIN I, 22 (14.6%) had AKI AKIN II and 19patients (12.6%) patients had AKI AKIN III. 16patients (10.6%) eventually needed kidney replacement therapy. Of the 151AKI patients, 116patients (76.8%) had an AKI already at the time of hospital admission. The remaining 35patients newly developed an AKI during their hospital stay, and in average 7.3±6.7 days after admission.

COVID-19 patients suffering from AKI showed significantly lower levels of sUmod compared to patients without kidney impairment during hospitalisation ((124.8±79.5 ng/ml) vs 214.6±92.3 ng/ml; p<0.001; Fig. 1B). Univariable logistic regression models for the association between baseline variables and the occurrence of AKI found that sUmod at hospital admission was significantly associated with the occurrence of AKI (p<0.001; Table ​Table2).2). Patients with higher sUmod concentrations revealed a lower probability for AKI (OR 0.988, CI 0.985 to 0.991; Table ​Table2).2). These univariable results were confirmed by a multivariable logistic regression model with backward selection, which accounted for further potential confounding factors being significant in the univariable analyses (with p<0.05). Importantly, COVID-19 patients with higher sUmod values revealed a lower probability for AKI (OR 0.993, CI 0.993 to 0.989, Table ​Table2).2). The diagnostic ability of sUmod to predict AKI was evaluated by a ROC curve (Fig. 1D) and the corresponding area under curve (AUC; 0.774, CI 0.726 to 0.822).

Table 2

Odds Ratios (OR) and corresponding 95% Confidence Intervals (95% CI) as well as the p-value for the univariable models and multivariable models (with backward selection) for the probability for AKI.

		Univariable models		Multivariable models
Variable	Comparison	OR (95% CI)	p-value	OR (95% CI)	p-value
Age		1.055 (1.037–1.072)	<0.001		n.s
Sex	M vs F	1.481 (0.974–2.251)	0.066		n.i
Invasive ventilation	Yes vs. No	8.108 (3.787–17.36)	<0.001	5.784 (1.961–17.06)	0.002
CKD	Yes vs. No	5.240 (3.109–8.831)	<0.001		n.s
Diabetes	Yes vs. No	3.254 (2.095–5.055)	<0.001		n.s
WHO-score at admission		0.959 (0.746–1.232)	0.742		n.i
sUmod (ng/ml)		0.988 (0.985–0.991)	<0.001	0.993 (0.989–0.997)	<0.001
Serum creatinine (µmol/l)		1.045 (1.034–1.057)	<0.001		n.s
Urinary albumin (g/l)		1.854 (0.920–3.738)	0.084		n.i
Serum albumin (g/l)		0.914 (0.874–0.955)	<0.001		n.s
Serum urea (mmol/l)		1.467 (1.345–1.601)	<0.001	1.323 (1.195–1.466)	<0.001
UACR (g/g creatinine)		2.021 (0.961–4.249)	0.064		n.i
CRP (mg/l)		1.006 (1.003–1.009)	<0.001		n.s
IL-6 (ng/l)		1.012 (1.007–1.016)	<0.001	1.007 (1.001–1.013)	0.027
Leukocytes (Gpt/l)		1.100 (1.032–1.172)	0.003		n.s

Open in a separate window

Note that n.s. indicates that the variable was not significant in the multivariable model with backward selection and was therefore excluded from the models. Variables with n.i. were not included in the multivariable model because they were not significant already in the univariable models. CKD: chronic kidney disease, CRP: C-reactive protein, IL-6: Interleukin-6, sUmod: serum uromodulin, UACR: urine albumin-creatinine ratio.

Besides sUmod, invasive ventilation, serum concentrations of IL-6 and serum urea showed significant associations with AKI (Table ​(Table2).2). Patients receiving invasive ventilation showed a much higher risk for AKI (OR 5.784, CI 1.961 to 17.06, p=0.002). In more detail, 80.9% of patients receiving invasive ventilation had AKI. In contrast, AKI occurred in 34.2% of patients, who did not receive invasive ventilation. Patients with higher IL-6 and serum urea concentrations showed a higher risk to develop an AKI (OR 1.007, CI 1.001 to 1.013, p=0.027 and OR 1.323, CI 1.195 to 1.466, p<0.001, respectively. All other parameters did not remain significant after model selection.

Association between AKI and fatal outcome in COVID-19 patients

The in-hospital mortality rate in this cohort of COVID-19 patients was 15.3%. Among patients being dismissed alive, 6.9% died within 6 months. Out of the 47patients receiving invasive ventilation, 91.5% died during hospital stay, while out of the 330patients who did not receive invasive ventilation, 4.6% died. Univariable logistic regression models for the association between baseline variables and the occurrence of in-hospital death revealed that AKI in COVID-19 patients was significantly associated with in-hospital death (p<0.001; Table ​Table3).3). Patients with AKI showed a higher probability for in-hospital death (OR 10.113, CI 4.92 to 20.785; Table ​Table3).3). These results from univariable regression models were confirmed by a multivariable logistic regression model with backward selection, which was performed for all parameters being significant in the univariable analyses (with a p<0.05). Patients with AKI had a higher probability for in-hospital death (OR 5.6, CI1.76to17.881, p=0.004). In more detail, out of 151patients presenting with AKI, 31.8% died during hospital stay. In contrast, out of the 227patients not presenting with AKI merely 4.4% died.

Table 3

Odds Ratios (OR) and corresponding 95% Confidence Intervals (95% CI) as well as the p-value for the univariable models and multivariable models (with backward selection) for the probability for in-hospital death.

		Univariable models		Multivariable models
Variable	Comparison	OR (95% CI)	p-value	OR (95% CI)	p-value
Age		1.068 (1.039–1.097)	<0.001	1.08 (1.025–1.139)	0.004
Sex	M vs F	1.766 (0.978–3.189)	0.059		n.i
AKI	Yes vs. No	10.113 (4.920–20.785)	<0.001	5.6 (1.76–17.811)	0.004
Invasive ventilation	Yes vs. No	225.750 (71.623–711.547)	<0.001	165.146 (45.142–604.165)	<0.001
CKD	Yes vs. No	4.129 (2.295–7.429)	<0.001		n.s
Diabetes	Yes vs. No	1.938 (1.101–3.413)	0.022		n.s
Malignom History	Yes vs. No	2.458 (1.111–5.440)	0.026		n.s
COPD	Yes vs. No	1.203 (0.475–3.049)	0.696		n.i
WHO-score at admission		1.337 (0.946–1.890)	0.1		n.i
sUmod (ng/ml)		0.993 (0.989–0.996)	<0.001		n.s
Serum creatinine (µmol/l)		1.010 (1.006–1.014)	<0.001		n.s
Urinary albumin (g/l)		2.312 (1.151–4.641)	0.018		n.s
Serum albumin (g/l)		0.884 (0.833–0.938)	<0.001		n.s
Serum urea (mmol/l)		1.134 (1.086–1.184)	<0.001		n.s
Serum sodium (mmol/l)		1.082 (1.017–1.150)	0.013		n.s
UACR (g/g creatinine)		1.850 (0.977–3.504)	0.059		n.i
CRP (mg/l)		1.009 (1.006–1.013)	<0.001		n.s
IL-6 (ng/l)		1.009 (1.005–1.013)	<0.001		n.s
Leukocytes (Gpt/l)		1.097 (1.022–1.176)	0.01		n.s
Urinary protein (mg/l)		1.001 (1.000–1.001)	0.013		n.s
PCT (µg/l)		1.101 (0.999–1.213)	0.051		n.i
LDH (µmol/l*s)		1.232 (1.109–1.368)	<0.001	1.356 (1.107–1.661)	0.003

Open in a separate window

Note that n.s. indicates that the variable was not significant in the multivariable model with backward selection and was therefore excluded from the models. Variables with n.i. were not included in the multivariable model because they were not significant already in the univariable models. AKI: acute kidney injury, CKD: chronic kidney disease, COPD: chronic obstructive pulmonary disease, CRP: C-reactive protein, IL-6: Interleukin-6, LDH: lactate dehydrogenase, PCT: procalcitonin, sUmod: serum uromodulin, UACR: urine albumin-creatinine ratio.

Association between serum uromodulin and fatal outcome in COVID-19 patients

Since sUmod concentrations at hospital admission of COVID-19 patients were significantly associated with the occurrence of AKI, the association between sUmod and in-hospital death was evaluated. Notably, COVID-19 patients who died during hospital stay, had significantly lower sUmod levels (129.14±89.56 ng/ml) compared to patients surviving hospitalisation (187.71±96,64 ng/ml; p<0.001; Fig. 1C). Univariable logistic regression models for the association between baseline parameters and in-hospital death showed a significant association of sUmod and the probability of in-hospital death (p<0.001; Table ​Table3).3). Patients presenting with higher sUmod levels had a lower probability for death (OR 0.993, CI 0.989 to 0.996) compared to patients with lower sUmod concentrations. However, sUmod did not remain significant in multivariable logistic regression model with backward selection, which was performed for all parameters being significant in the univariable analyses (with a p<0.05).

Besides health-related sUmod levels, patients receiving invasive ventilation showed the largest risk of death (OR165.146, CI: 45.14 to 604.17). In more detail, out of 47patients receiving invasive ventilation, 91.5% died during hospital stay, while out of 330patients who did not receive invasive ventilation, merely 4.6% died. Increasing age as well as patients having higher LDH values also showed a significantly larger probability of death (OR age: 1.08, OR LDH: 1.356). All other parameters including serum creatinine did not remain significant after model selection (see Table ​Table33).

Prevalence of AKI in COVID-19 patients

AKI commonly occurs in individuals hospitalised with COVID-19^16,46. Our findings align with and contribute to the evolving body of literature on kidney complication associated with SARS-CoV-2 infection. In the present study, 40% of the entire cohort of COVID-19 patients experienced AKI. Notably, recent meta-analyses reported lower AKI rates ranging from 8 to 28%^17,47,48. Varying rates of AKI can be explained by the heterogeneity of COVID-19 cases. Variations in viral loads, host-related nephrotoxic susceptibility, undetected preexisting risk factors, a diverse “disease spectrum”, and the potential influence of autoimmunity and autoinflammation can collectively contribute to the observed variety in AKI rates among individuals with COVID-19⁴⁹. Our study enrolled hospitalised patients, inherently encompassing a higher proportion of individuals with severe COVID-19 cases. This distinction in patient profiles, with a focus on hospitalised cases compared to outpatient clinic and intensive care unit (ICU) cohorts, could contribute to the observed differences in AKI prevalence rates. Another reason for the observed lower prevalence of AKI in other studies could be a potential underreporting of AKI. In accordance with our findings, an investigation in 3993 hospitalised COVID-19 patients revealed a prevalence of 46% for AKI⁴⁶. Notably, the majority of AKI cases in our study were categorised as AKIN stage I (72.8%), deviating from prevalences of 42% and 44% documented in other reports^17,46. Here, a potential source of error may also lie in diagnostic discrepancies between physicians and nephrological experts, as the missing cases likely correspond to AKIN stage I instances. Our study expands on existing knowledge by exploring the timing of AKI onset. We observed that almost three quarters of AKI cases were already present at the time of hospital admission. Nevertheless, a considerable portion of AKI cases (23.2%) developed during the hospital stay, with an average of 7.36 days post-admission. This temporal analysis adds granularity to the understanding of AKI progression in COVID-19 patients, offering insights into potential windows for preventive measures.

Advantages of sUmod as a diagnostic marker for AKI in the context of COVID-19

Our study presents insights into the pathological involvement of specific and vulnerable segments of the human nephron in COVID-19 related kidney injury. Epithelia of the TAL segment were shown to be highly susceptible towards hypoxia, various toxins, oxygen radicals, disturbance in haemodynamics etc., all common factors characteristic in a clinical setting of intensive care increasing the risk of an impaired synthesis of uromodulin (Heyman et al. 1991, 2010; Steubl et al., 2017, Thielemans et al. 2023). A distinctive aspect of our investigation is its pioneering role as the first study to assess sUmod in COVID-19 patients, shedding light on the substantial engagement of the medullary epithelia in virus-mediated kidney injury. Compared to conventional urinary biomarkers, sUmod emerges as a remarkably robust biomarker. This robustness is particularly noteworthy considering the challenges associated with urine processing, as highlighted by previous studies on the timing and stability of proteins/peptides^42,62. Importantly, sUmod remains stable even under conditions such as water loading, distinguishing it from certain limitations associated with urine-based markers⁶³. In contrast to other ‘kidney biomarkers’, sUmod stands out for its specificity to kidney tissue components. Our study contributes to the understanding that a broad ‘biomarker panel’, encompassing urinary proteins and proinflammatory interleukins such as NGAL, KIM-1, EGF, urinary uromodulin, and albumin, showed relatively low diagnostic relevance for assessing kidney damage in COVID-19 patients⁶². Furthermore, our investigation revealed a compelling correlation between sUmod and other markers, such as serum creatinine and cystatin C. Despite the lack of significance in the multivariable model to predict fatal outcome, the initial univariable association suggests sUmod may have potential implications for risk stratification and early intervention in COVID-19 patients. Clinicians should be attentive to the dynamic nature of biomarkers and their utility in predicting patient outcomes. However, its initial association warrants further exploration in larger cohorts or specific patient subgroups.

The authors thank technical personnel of EUROIMMUN Medizinische Labordiagnostika AG, Department of Experimental Immunology, for supporting uromodulin analyses, which was funded by EUROIMMUN Medizinische Labordiagnostika AG, Luebeck.