Introduction

Cholemic nephropathy (CN) is an under recognised cause of renal dysfunction in patients with acute on chronic liver failure (ACLF) as it requires a histological diagnosis1,2,3,4. Urine microscopy (UM) could be the most easy and cost-effective tool for identifying the spectrum of kidney injury, however, limited studies exist in the context of ACLF patients. UM identified a higher prevalence of structural AKI compared to acutely decompensated cirrhosis5. UM showed a higher prevalence of fine and coarse granular casts suggesting higher tubular injury. UM score derived from the number of renal tubule epithelial cells (RTEC) and/or granular casts predict the risk of structural tubular injury6. We showed in a cohort of critically ill patients with cirrhosis presence of RTEC and/or granular casts could determine non-resolution of AKI. Further, failure of mitochondrial function in the RTECs correlated with AKI progression and/or persistence. In a cohort study of ACLF patients, which also included data from postmortem kidney biopsies CN was the dominant histological diagnosis seen in almost two-thirds of the patients compared to acute tubular necrosis (ATN)3 The urine microscopy (UM) of CN patients demonstrated the presence of bilirubin crystals or bile casts in 100%, while it was seen in only 11% of patients with ATN and 78% of patients who had both ATN and CN. Therefore, the presence of bilirubin crystals or bile casts on UM could be a non-invasive and an indirect surrogate of CN in ACLF patients. However, currently, there are limited studies exploring the role of UM in ACLF patients. From a pathophysiological perspective, high serum levels of bilirubin and bile acids mediate toxicity to the renal tubules8,9, even though an association of higher bilirubin with CN is shown, the association of bile acids with CN has never been reported in ACLF patients. Currently, there are no studies exploring the role of UM and its correlation with biomarkers for delineating the spectrum of AKI in ACLF. UM could be most easily available tool which could aid the diagnosis of CN and it’s differentiation from ATN and HRS-AKI. The diagnosis of CN may enable therapeutic decisions of the use of extracorporeal therapies for instance plasma-exchange (TPE) in improving CN and cholestasis caused by bile acid toxicity amidst systemic inflammation10,11.

We aimed to evaluate whether UM can identify CN in patients with ACLF and differentiate from hepatorenal syndrome (HRS) and ATN. We performed a panel of biomarkers, serum bile acids, plasma metabolomics, and markers of systemic inflammation in patients stratified into distinct spectrum based on the urine microscopic stratification. In a subset of patients with biopsy-proven CN, metabolomics and mitochondrial function assessment was performed in the RTEC and these were compared to patients with ATN. We also investigated the response to therapeutic interventions and a clinical score to identify CN in ACLF patients.

Results

Altogether, of the 128 patients screened, a total of 68 patients were excluded, and 60 patients were included in the current study. (Supple Fig. 2) The patients were stratified into no AKI (n = 15), HRS-AKI (n = 15), CN (n = 15) and ATN (n = 15) based on the UM examination as described in methods. The comparison of baseline demographics, clinical, and biochemical characteristics is enumerated in Table 1 and Supple. Table 1. The groups were comparable in age and gender. Alcohol was the predominant etiology which constituted 80% of patients in no AKI group followed by 66% in ATN, 53% in HRS and 40% in CN. On UM, all patients had bilirubin crystals in CN and majority also had bile casts 12/15 i.80% of which 5 (33.3%) patients had 1, 5 (33.3%) had 2 and 2 (13%) had 4 bile casts/hpf. Mean total bilirubin was significantly higher in CN (27.05 ± 7.25 mg/dL) compared to HRS (21.96 ± 9.94), ATN (22.28 ± 9.47), and no AKI (15.86 ± 6.56) (p = 0.007). Patients with CN also showed a higher leucocyte counts compared to other groups. The groups had comparable severity of metabolic acidosis. However, we observed a significantly lower urine output (ml/hr) in the ATN (23.0 ± 11.5) group compared to HRS (66.0 ± 32.0) and CN (44.3 ± 12.7) (p < 0.001).

Table 1 Baseline characteristics in acute on chronic liver failure patients based on urine microscopy based stratification of acute kidney injury.
Full size table

Urine biomarkers in AKI phenotypes stratified by urine microscopy

Principal component analysis (PCA) of urine biomarkers documented a clear distinction among ACLF patients with and without AKI. When patients with UM-based stratification were compared, we observed patients with HRS and CN-AKI were closer, while patients with ATN-AKI showed marked distinction. (Fig. 1A) The overall comparison is depicted in Table 2 while intergroup comparisons are enumerated in Supple. Table 2. On random forrest analysis, NGAL, renin, alpha-2 microglobulin, collagen-IV, and GST-α were the top five biomarkers which showed highest mean decrease in accuracy (Fig. 1 B-C). It was interesting to observe that except GST-α (highest in CN), most of the injury markers i.e., renin, lipocalin-2, cystatin c, alpha-1 macroglobulin, albumin, TIMP-1, IP-10 and KIM-1 were significantly higher in ATN compared to other groups. (Supple. Fig. 3A and B, Table 2). The markers of renal repair i.e., EGF, osteopontin, and calbindin were significantly suppressed in ATN compared to CN and HRS. (Supple. Fig. 3B) Between HRS and CN, EGF and OPN was significantly higher in CN while calbindin in HRS. The markers of renal fibrosis, osteoactivin, collagen-IV was significantly higher in ATN but were comparable in HRS and CN (Table 2). Together, these data suggest a distinct profile of urine biomarkers in UM-based stratification of AKI.

Fig. 1
Fig. 1The alternative text for this image may have been generated using AI.
Full size image

(A) Partial least square- discriminant analysis (PLS-DA) plot showing clustering of samples in acute on chronic liver failure (ACLF) patients with different causes of acute kidney injury (AKI) i.e., acute tubular necrosis (ATN), cholemic nephropathy (CN), hepatorenal syndrome (HRS) and no AKI group. (B) Heat map showing differential expressed cytokines between ATN, CN, HRS and no AKI group. (C) Mean decrease in accuracy plot (MDA) showing expression of markers by random forest analysis highlighting the top two biomarkers for the differential diagnosis of AKI phenotype. (D) Heatmap showing plasma cytokines and growth factors distributed in different spectrum of acute kidney injury (AKI) in patients with ACLF (HRS- Hepatorenal syndrome, CN- cholemic nephropathy, ATN-acute tubular necrosis). (E and F) Bile acid production was also significantly high in patients with CN and DAMPS like HMGB1 (high mobility group box 1), LPS (lipopolysaccharide) and EDN (endothelin-1) were significantly high in patients with CN.

Table 2 Urinary renal biomarkers in acute on chronic liver failure correlated with urine microscopy based stratification of acute kidney injury.
Full size table

Urine microscopy-based diagnosis of CN is associated with higher serum bile acids and more severe systemic inflammation

It was very interesting to observe a distinct phenotype of most of the analysed cytokines and growth factors which were significantly higher (20 out of 29) in CN compared to ATN, HRS and no-AKI (Fig. 1D, Table 3). Patients with CN had significantly higher proinflammatory cytokines compared to ATN signifying intense systemic inflammation. The comparison between groups is depicted in Supple. Table 3. Random forest classification showed significant difference in the plasma levels of eotaxin, IL-15, IL3, IL-5, IL-6, IL-2, IL-4, IL-1beta, IL-1RA, IL-17A, TNF- alpha, MIP1-beta, MCP-1 and VEGF among the different phenotypes of AKI in ACLF patients and those with no AKI. (Supple Fig. 4A) We stratified the cytokines by the type of inflammatory response. Interestingly, the markers of type 1 inflammation were comparable between the groups, with the exception of IFN-γ which was markedly elevated in CN. Among the markers of type 2 and type 3 inflammation, the levels of IL-4 were lowest in CN. Of all the other cytokines, the levels of type 1 and II interferons i.e., interferon-α and interferon-γ were markedly elevated in CN compared to ATN, HRS and no AKI. Similarly, the levels of MCP-1 and MIP-1α, the key chemokines involved in migration and infiltration of monocytes/ macrophages and leucocytes respectively were significantly higher in CN compared to other groups. (Supple Fig. 4B) The levels of TNF-α were also significantly higher in CN while eotaxin levels were suppressed. (Supple Fig. 4C, Table 3) The plasma levels of VEGF, G-CSF, and GM-CSF were significantly high in CN, but were comparable in ATN and HRS-AKI. (Supple Fig. 4D).

Table 3 Inflammatory cytokines and growth factors in acute on chronic liver failure patients with urine microscopy based stratification of acute kidney injury.
Full size table

Serum bile acids: Quantitative serum bile acids were elevated in both CN and ATN compared to no-AKI and HRS. Between CN and ATN, patients with CN-AKI showed significantly higher levels of serum bile acids. The levels of endotoxin, and HMGB1 were markedly elevated in CN while endothelin-1 levels were suppressed (Fig. 1E-F).

Metabolomic profile of cholemic nephropathy (CN) is distinct in patients with ACLF

Hierarchical clustering analysis and PLSDA analysis showed a clear segregation of no AKI, ATN, HRS, and CN (Fig. 2A and B). When plasma metabotype of CN patients was compared to plasma metabotype of no AKI, 595 DEM’s were identified (294 down and 301 up; FC > 1.5, p < 0.05). (Fig. 2C). Comparative pathway enrichment of CN was compared to no AKI, CN patients have upregulated spermine and spermidine metabolism, arginine and proline metabolism, urea cycle and alpha linolenic metabolism, while downregulated folate and purine-pyrimidine metabolism (Supple Fig. 5A-B). When plasma metabotype of CN patients was compared to plasma metabotype of other AKI individuals 464 DEM’s were identified (194 down and 270 up; FC > 1.5, p < 0.05). CN patients compared to HRS showed 498 DEM’s (117 down and 381 up; FC > 1.5, p < 0.05) (Fig. 2D.) Comparative pathway enrichment of CN compared to HRS, showed patients with CN having an upregulated glycine and serine metabolism, methionine and ammonia recycling while downregulated pyrimidine metabolism, ketone body metabolism and others. (Supple Fig. 6A-B) When plasma metabotype of CN patients was compared to plasma metabotype of patients with ATN, 483 DEM’s were identified (301 down and 182 up; FC > 1.5, p < 0.05) (Fig. 2E). Comparative pathway enrichment when CN was compared to ATN, CN patients had upregulated arginine and proline metabolism, purine, and histidine metabolism while downregulated folate metabolism, aspartate metabolism and others. (Supple Fig. 7A-B).

Fig. 2
Fig. 2The alternative text for this image may have been generated using AI.
Full size image

(A and B) Hierarchical cluster analysis and partial least squares discriminant analysis (PLS-DA) showing clear segregation of cholemic nephropathy (CN), hepatorenal syndrome (HRS), acute tubular necrosis (ATN) and no acute kidney injury (AKI). (C–E) Volcano plots showing differentially expressed metabolites when CN were compared to no AKI; HRS and ATN. Y-axis corresponds to –log 20 p-value, and X-axis corresponds to log twofold change.

Renal histology and immunohistochemistry in biopsy-proven CN compared to ATN

Renal histology is considered as gold-standard for the diagnosis of CN and ATN. Post-mortem kidney biopsies were performed for confirmation of diagnosis, CN (n = 5) and ATN (n = 5) in patients who died during follow-up. On haematoxylin and eosin staining, the renal specimens demonstrated distinct histopathological features with presence of bile cast and bile pigment in renal proximal tubules and tubular epithelial cells in CN which were not identified in ATN; the latter showed sloughing and degeneration of renal proximal tubule epithelial cells (Fig. 3A). Further TUNEL staining of renal biopsy showed presence of endothelial injury in CN while ATN patients showed significant increase TUNEL positive endothelial and tubular epithelial cell, compared to CN. (Fig. 3B). On immunohistochemistry, patients with CN, in comparison to ATN, showed higher expression of aquaporin-2 (marker for fluid balance), SIRT-1 (marker for protective renal response to injury) and lower expression of SGLT2 (marker for failed renal repair) and HIF-1 (marker for hypoxia-induced injury) (Fig. 3C).

Fig. 3
Fig. 3The alternative text for this image may have been generated using AI.
Full size image

(A) Hematoxylin & eosin staining (H & E staining) showed higher necrosis in renal tissues of patients with acute tubular necrosis (ATN) compared to those with cholemic nephropathy (CN). (B) Tunel assay showing higher positivity with renal tubular epithelial cells (RTEC) in ATN compared to CN. (C) Renal histopathology and immunohistochemistry showed higher staining of aquaporin-2 (AQP2) and sirtuin-1 (SIRT-1), and lower staining of sodium-glucose co-transporter (SGLT) and hypoxia-inducible factor-1 (HIF-1) in patients with CN. On other hand, it showed lower intensity of AQP2 and SIRT-1 staining, and a higher intensity staining of SGLT and HIF-1 in ATN.

Urine and plasma metabolomics from CN showed a distinct metabolic profile compared to ATN

We performed untargeted metabolomics using liquid chromatography coupled to high-resolution mass spectrometry in RTEC in ACLF patients with biopsy-proven diagnosis of AKI (CN, n = 5, ATN, n = 5). Because of the lack of RTEC (due to lack of tubular injury) in patients with HRS and no AKI, only CN and ATN patients could be compared for RTEC. Metabolomic analysis of RTEC identified 196 (61 up and 135 downregulated) metabolites in RTEC of CN compared to ATN (Fig. 4A; FC > 1.5, p < 0.05). RTEC of CN patients showed increase in metabolites associated with nicotinamide adenine dinucleotide (NAD) metabolism, tricarboxylic acid cycle (TCA), amino acid (AA) metabolism, ketone bodies degradation and accumulation of primary bile acids (Figure-4B) Concomitantly, a decrease in the metabolites associated with tryptophan metabolism, purine metabolism, pantothenate and CoA biosynthesis, glyoxylate and dicarboxylate metabolism was noted in CN compared to ATN (Fig. 4C). CN patients also showed increase in the accumulation of various primary bile acids (including taurocholate, glycoursodeoxycholic acid, glycocholic acid and chenodeoxyglycocholate) both in plasma and RTEC compared to ATN, suggesting these are the main bile-acids causing damage to RTEC in CN (Fig. 4D). Together these data demonstrate an upregulation of pathways involved in NAD production and mitochondrial energy metabolism in the urine tubule epithelial cells in CN compared to ATN (Fig. 4D) (Supple. Tables 4 and 5).

Fig. 4
Fig. 4The alternative text for this image may have been generated using AI.
Full size image

(A) Volcano plots showing the results of pairwise comparisons of renal tubule epithelial cells (RTEC) (left) and plasma (right) metabolites levels in acute on chronic liver failure (ACLF) patients with cholemic nephropathy (CN) relative to ACLF patients with acute tubular necrosis (ATN). The vertical dashed lines indicate the threshold for the two-fold abundance difference. The horizontal dashed line indicates the p = 0.05 threshold. Between-group comparisons were performed using Student’s t test. (B) Up‐regulated metabolite pathway enrichment (bubble plot analysis) based on the HMDB database in CN compared to ATN in RTEC (Left) Plasma (right) (C) Down‐regulated metabolite pathway enrichment (bubble plot analysis) based on the HMDB database in CN compare to ATN in RTEC (Left) Plasma (right). (D) Diagram showing alteration in tryptophan metabolism and tricarboxylic acid cycle (TCA) cycle metabolites in RTEC of ACLF patients with CN (E) Real-time changes in oxygen consumption rate (OCR) with subsequent treatment with oligomycin (Oligo.) FCCP and rotenone and antimycin A (Rot. + AA.) in peripheral blood mononuclear cells (PBMC) of ACLF patients with CN and ATN (top) and bar graph (bottom) showing changes in basal mitochondrial respiration and glycolysis in PBMC of ACLF patients with CN and ATN. (F) Bar graph showing changes in maximal (left) and spare reserve (right) associated mitochondrial respiration in PBMC of ACLF patients with CN and ATN. (G) Bar graph showing relative expression (fold change) of given genes in RTEC of ACLF- CN with respect RTEC of ACLF-ATN.

Table 4 Course of acute kidney injury during hospitalization with urine microscopy-based stratification.
Full size table

Mitochondrial function is preserved in CN compared to ATN

To understand the bioenergetic dysfunction at systemic level we compared the oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) of peripheral blood mononuclear cells (PBMCs) as readout for OXPHOS and glycolysis respectively in CN and ATN. In comparison to ATN, CN patients’ PBMCs showed significantly higher levels of OXPHOS (p < 0.01) and glycolysis (p < 0.001) (Fig. 4E). Further, analysis of different OXPHOS parameters showed significant increase in the maximum respiration (p < 0.001), ATP production (p < 0.001) and spare reserve capacity (p < 0.001) in CN PBMCs compared to ATN (Fig. 4F). The qRT-PCR analysis showed more than twofold expression of various genes (PGC1, COX-1, NR, SDH, and TFAM) associated with mitochondrial biogenesis in exfoliated RTECs of CN and ATN (Figure-4G).

Development of clinical score for the detection of cholemic nephropathy

The predictive power of various biomarkers and clinical indicators was evaluated for cholemic nephropathy (CN), with presence of bile casts (AUC = 0.90) and GST-alpha (AUC = 0.90) emerging as the top biomarkers for CN prediction. (Supple. Fig. 8) Both biomarkers exhibited the strongest associations with CN, demonstrating high predictive accuracy when compared to other clinical markers. The top 5 variables included OPN, GST-alpha, IL-3, IL-15, and bile acids (Supple. Fig. 9) for prediction of CN. After adjusting for multicollinearity, we created two multivariable logistic regression models, comprising GST-alpha, IL-15, and serum bilirubin (> 22 mg/dl) in Model-1 and AARC score in Model-2. (Supple. Table 5) Both models had excellent accuracy for discriminating cholemic from other AKI phenotypes (AUROC 0.96, 0.92–1.00) for Model-1 and (AUROC 0.97, 0.94–1.00) for Model-2 respectively. (Supple Fig. 10).

Course and outcome of cholemic nephropathy compared to other etiologies of AKI in ACLF patients and response to therapeutic interventions (Terlipressin, therapeutic plasma exchange and CRRT)

The resolution of AKI was significantly higher for HRS-AKI compared to CN and ATN (73.3% vs. 33.3% vs. 13.3%; p < 0.001) respectively. The response to terlipressin at day 4 was highest for HRS compared to CN (66.7% vs. 20%; p < 0.001). Therapeutic plasma exchange (PLEX) was performed for 9 (60%) of CN compared to 4 (26.6%) of HRS-AKI and 5 (33.3%) in ATN AKI (p = 0.005). Continuous kidney replacement therapy (CRRT) was performed in none of patients with HRS-AKI, compared to 8 (53.3%) vs. 5 (33.3%) in ATN (p < 0.001), and 2 (13.3%) patients in ATN underwent sustained low-efficiency dialysis (SLED). The 28-day survival was significantly higher for patients with HRS-AKI compared to CN and lowest for ATN (87% vs. 53% vs. 13%; p < 0.001) (Table 4).

We evaluated the therapeutic response of performing PLEX on pro and anti-inflammatory cytokines, growth factors, and bile acids in patients with CN and ATN. (Suppl. Fig. 11A-C) The heat map shows changes in plasma cytokines and growth factors after PLEX in ACLF patients with CN and ATN. A significant increase in the levels of most growth factors (EGF, G-CSF, VEGF) with the exception of GMCSF after PLEX in CN while in patients with ATN there was no significant change with the exception of a significant increase in the VEGF levels. A significant decrease in levels of almost all the pro-inflammatory cytokines [eotaxin, interferons (IFN-A2, IFN-G), monocyte chemoattractant protein-1 (MCP-1), macrophage inflammatory protein-1 (MIP-1α & MIP-1β), and tumor necrosis factor- alpha (TNFα)] and bile acids was demonstrated both in CN and ATN post-PLEX. However, a significant reduction in IL-6 was observed in CN but in ATN. For anti-inflammatory cytokines [interleukins (IL-1RA & IL-10)] were significantly increased after PLEX in CN but not in ATN. Together our data suggest a differential response of CN and ATN to TPE.

Discussion

The current investigation shows the UM could be an effective non-invasive diagnostic tool for the identification of AKI phenotypes in patients with ACLF. The presence of bilirubin crystals and or bile casts identified cholemic nephropathy (CN) as a distinct form of renal dysfunction in patients with ACLF compared to ATN and HRS. The distinction was noted in the panel of urine biomarkers, the severity of systemic inflammation, bile acids, and damage-associated molecular patterns. Overall, we observed patients with CN were more “inflamed” than the other patients. The levels of proinflammatory inducers were higher in patients with CN together with higher levels of a battery of inflammatory cytokines. Second, despite having more intense inflammation patients with CN do not develop the most severe form of renal pathology, probably because these patients actuate a program which is protective against immunopathology. In other words, these patients develop a program that make kidneys tolerant against the attack of inflammatory molecules. The underpinning mechanisms that may explain the induction of a protective program, however, still remain elusive but possibly related to preserved mitochondrial function in CN vs. ATN. Both CN and HRS demonstrated lower tubular injury, enhanced repair and decreased fibrosis at the level of the renal tubules compared to patients with ATN. In biopsy-proven CN, an increased metabolic adaption with respect to energy metabolism and mitochondrial biogenesis and function was observed compared to ATN. The clinical implication was observed in the response to terlipressin which was better in HRS-AKI compared to CN, signifying the need of possibly combining terlipressin with extracorporeal strategies for targeting systemic inflammation for management of CN. Survival was better for HRS and CN compared to ATN. Plasma exchange improved systemic inflammation and bile acids in CN and ATN. However, a significant increase in IL-1RA and growth factors (with the exception of GM-CSF), decrease in IL-6 and MIP-1Beta was noted in CN. The top 5 variables OPN, GST-alpha, IL-3, IL-15, and bile acids could distinguish CN from other AKI phenotypes. Further, a clinical score comprising GST-alpha, IL-15 and serum bilirubin or AARC score was more than 95% accurate in differentiating CN from other phenotypes of AKI.

The data on CN is very sparse and only comprises of case reports or case-series and animal data1,2,3,4. In an animal model of bile duct ligation CN was seen as a rapidly progressive renal injury mediated by bile acids which lead to tubulointerstitial fibrosis and was abrogated by norUDCA treatment12. In a series of 149 patients wherein a kidney biopsy was performed, 17.8% of the patients had evidence of CN13. The authors identified serum bilirubin, alkaline phosphatase, urinary bilirubin, and urobilinogen as predictors of CN. Similar to our observations, the diagnosis of CN was not associated with higher death in these patients and was associated with higher C-reactive protein levels, a marker of systemic inflammation. On the clinical front, these patients had more severe jaundice, higher urine output, compared to ATN. Urine output is now recognized as the most relevant surrogate of AKI progression in ACLF patients4. Together these data suggest CN to be associated with a better prognosis and therefore its identification and differentiation from ATN is imperative even though both cause structural damage to the renal tubules. Further, the response to terlipressin and outcomes of AKI were inferior compared to HRS. Biomarkers of tubular injury were markedly elevated in ATN compared to CN and HRS. NGAL is the most widely validated biomarker for identification of ATN and terlipressin non-response14,15,16. All tubular injury markers were markedly elevated in patients with ATN compared to HRS and CN, excepting KIM-1 and IP-10. Circulatory dysfunction and tubular damage could be key factors implicated in ACLF patients with AKI17. Patients with CN had higher levels of osteopontin, a glycoprotein essential for renal tubulogenesis and EGF which is a kidney-specific protein, restricted to the thick ascending loop and collecting ducts and is responsible for tubular repair and regeneration after injury18.

Interestingly, patients with CN had significantly higher systemic inflammation and bile acids, had preserved mitochondrial function, and improved outcomes. Previous studies have shown systemic inflammation can directly injure the RTEC by increased tryptophan (Trp) flux to kynurenine, accumulation of ketogenic amino acids and impaired mitochondrial function19,20,21,22. Moreau and colleagues had shown the defects in tryptophan metabolism as a key factor driving renal dysfunction and severity in ACLF patients22. Patients with increased tryptophan metabolites kynurenine had worse outcomes, higher risk of bacterial infections and more frequently developed ACLF22. Infact, increase in tryptophan flux to kynurenine due to the defect in de-novo NAD synthesis leads to loss of NAD pool, essential for regulating the physiological functions which include energy metabolism and immune modulation required for host tolerance to immune pathology20,21,22. Kidneys are important in regulating the synthesis, degradation and excretion of tryptophan, therefore, changes in renal function have a direct implication on tryptophan metabolism21,22. We found a significant enrichment of pathways related to tryptophan and nicotinamide adenine dinucleotide metabolism, tricarboxylic acid cycle, amino acid metabolism and bile acid synthesis in patients with CN compared to ATN. Systemic inflammation also causes endothelial dysfunction, loss of barrier integrity, formation of microthrombi causing alterations in blood rheology and organ dysfunction23. VEGF is a reliable marker of endothelial cell permeability and crucial for cell survival, chemotactic for monocytes, and inhibits the maturation of dendritic cells24. We found significantly high levels of VEGF in patients with CN. Concomitant to this, we also observed lower levels of endothelin-1 in CN compared to ATN suggesting decreased endothelial injury. Patients with HRS and no AKI had significantly lower levels of endothelin-125. Interestingly, IL-4 is a potent inducer of eotaxin which were markedly increased in patients with HRS-AKI. These are effectors of type-2 immune response which may be protective for the epithelia by their immunomodulatory action26. Aberrations in the nicotinamide pathway has been implicated in the host response to sepsis and viral pathogens. Cells of the macrophage/monocyte lineage contain the rate limiting enzyme indoleamine 2,3 dioxygenase which is implicated in de-novo nicotinamide synthesis. Monocyte/macrophages are central to disease progression in ACLF27, and levels of MCP-1 have shown correlation with the severity of inflammation in patients with ACLF28 We found MCP-1 to be significantly higher in patients with CN compared to ATN and HRS. We have previously shown that bioenergy failure drives the functional exhaustion of monocytes in ACLF28. Preserved mitochondrial function was observed in both RTEC and systemically by significantly higher glycolysis and OXPHOS in CN patients’ PBMCs compared to ATN.

We demonstrated the pathogenetic role of elevated bile acids particularly taurocholate, glycoursodeoxycholic acid, glycocholic acid, and taurochenodeoxycholate in patients with CN. Intriguingly, patients in all groups were deeply jaundiced, however, not all patients of ACLF develop renal dysfunction secondary to CN, suggesting bile acids as the key drivers of this entity28. Our study identifies the data on both quantitative and the type of bile acids which were distinctly upregulated in biopsy-proven CN. None of the previous reports on CN had investigated the role of bile acids. Taurochenodeoxycholic and taurodeoxycholic acid cause activation of the muscarine-like receptors, causing significant vasodilatation and hemodynamic effects by causing stimulation of endothelial derived nitric oxide synthase which could explain the impact of these bile acids in causing defects in renal perfusion29. Bile acids also impact the cardiac myocardium which was not investigated in our study29,30. Therefore, the data suggests that measurement and targeted reduction of the number and type of bile acids could ameliorate the CN. The data of PLEX showed significant reduction of all pro-inflammatory cytokines, an increase in IL-1RA and decrease in bile acids. In a previous analysis, we showed patients with therapeutic response to PLEX had a significant elevation of IL-1RA10. Apart from this a significant decline in IL-6 and MIP-1beta, was noted in CN but not ATN. MIP-1beta, is an inflammatory chemokine and multiple studies have explored the role of MIP-1b inhibition for renoprotection of podocytes in the context of diabetic kidney disease. Interestingly, this was not observed for patients with ATN, suggesting a differential response of PLEX in CN compared to ATN patients. Apart from PLEX, other extracorporeal therapies like molecular adsorbent recirculation system and bilirubin or plasma-adsorption perfusion systems may ameliorate CN by combating systemic inflammation and hyperbilirubinemia/ bile acids which could be explored in future studies31,32,33.

On immunohistochemistry of the kidney biopsies, we showed increased expression of SIRT-1 in CN and AQP-2. SIRT1, which is upregulated during inflammation and hypoxia, provides renoprotective effects. It is highly expressed in medullary tubular cells and moderately in cortical proximal tubular cells. SIRT1 relies on NAD+ to modify substrates such as FOXO, p53, histones, PGC-1α, and NF-κB, thereby inhibiting kidney cell apoptosis, reducing inflammation, improving mitochondrial function, and lowering oxidative stress. In the RTEC we demonstrated increased expression of genes associated with mitochondrial biogenesis including PGC-1α. Therefore, SIRT-1 may regulate mitochondrial biogenesis indirectly via PGC-1α33. Further, because RASi regulates the IL-17 and T-lymphocytes, while SGLT-2 is known to act via the inflammasome pathway and macrophage polarization, its possible that renin-angiotensin-system inhibitors (RASi), sodium-glucose cotransporter 2 inhibitors (SGLT2i) and the vasopressin receptor antagonists (VRA) could have renoprotective effects by regulating the maladaptive hyperfiltration of the glomeruli during immunological injury34.

In RTEC, a preserved mitochondrial function, with a two-fold higher upregulation of the genes involved in mitochondrial biogenesis was observed. Mitochondrial function as a potential pathomechanism driven by systemic inflammation is demonstrated for ACLF. We showed preserved mitochondrial function both systemically in PBMCs and in RTECs in CN patients’ compared to ATN. The upregulation of repair mechanisms, lower mitochondrial injury despite higher levels of bile acids signifies the tolerant host response in driving CN in patients with ACLF.

Last but not the least, wherein UM is not available we developed a clinical score comprising- GST-alpha, IL-15 with either serum bilirubin or AARC score with an accuracy of more than 95% for differentiating CN from other phenotypes of AKI. GST family are phase II enzymes, primarily involved in the detoxification by formation of reduced glutathione (GSH) and combating the oxidative stress. Several studies have demonstrated that GSTs play a critical role in protecting kidney cells from oxidant-mediated injury35. Interestingly, GST-alpha is expressed in proximal tubules of the kidneys, as against NGAL which is specific to injury to the distal tubules suggesting nephron-specific biomarkers for CN and ATN respectively15,35. IL-15 is produced by both infiltrating immune cells and resident kidney cells and has cytoprotective effects on kidney cells. In animal models of AKI, it suppresses pro-apoptotic cell-death signalling, epithelial to mesenchymal transition in the RTECs and prevents renal fibrosis. Recombinant IL-15 is being investigated as a therapeutic agent to prevent renal fibrosis36. In previous studies, we showed serum bilirubin above 22 mg/dl and AARC score above 10 determined terlipressin non-response and the need of dialysis in ACLF patients37. It would be worthwhile to validate the 3-component model we developed in future studies.

Our study has many strengths. This is the first report detailing the role of UM as an important tool for the possible diagnosis of CN in patients with ACLF. The stratification of UM-based assessment correlated with analysis of urine biomarkers and serum cytokines and biopsy-proven diagnosis of CN and ATN. We had a confirmed diagnosis of CN in 40% patients based on renal histology. The value of UM has been underreported in these patients. The UM was seen by two expert pathologists. The distinctness was clearly appreciated in the panel of biomarkers and cytokine data which were performed in these patients. The diagnostic accuracy was further confirmed by a series of experiments that we performed in histologically proven patients with CN. We also elegantly demonstrated the metabolic adaptation of the host, intense systemic inflammation, and bile acids in driving distinct phenotypes of AKI in ACLF patients. The limitations of our study remain the lack of histological diagnosis of CN in all patients. This could be performed only as postmortem biopsies in patients who had died. Performing a kidney biopsy is extremely difficult during the course of AKI in ACLF patients. It may have been unethical to perform kidney biopsies for the diagnosis and subjecting them to a very high risk of complications for the diagnosis of type of AKI wherein currently there are no guidelines on therapeutic management. Further, kidney biopsies even if performed need to target the distal nephron segments which has a higher risk of bleeding. The other limitation is the lack of dynamic assessment of urine microscopy, and exclusion of patients with mixed picture on urine microscopy as in the real-world scenario most AKI in the context of critically ill are multifactorial and dynamically evolve. The current study was however designed as a discovery and dileneation of the pure phenotype of cholemic and whether it is distinct from the other phenotypes based on urine microscopy and histologically proven diagnosis. We have developed a clinical score for the diagnosis of the same, however, this needs validation before routine clinical application. We would propose future studies validating the interesting data on urine microscopic, based phenotyping of AKI in ACLF patients. Combination of clinical criteria, UM supported by renal biomarkers and the development of clinical scores, could be evaluated for the diagnosis of CN in large prospective studies. Considering, worse response of CN to vasoconstrictors compared to HRS, and overall higher AKI resolution rate compared to ATN, signifies the relevance of making a clinical diagnosis. Further, extracorporeal therapies targeting systemic inflammation and bile acids could be useful adjuncts in the management of these patients38.

To conclude, our study shows UM-based stratification could diagnose CN as a distinct form of kidney dysfunction in patients with ACLF compared to HRS and ATN. Lesser tubular injury, higher renal repair, and preserved mitochondrial function form the hallmark of kidney dysfunction in these patients which makes them distinct from ATN and closer to HRS. A preserved host response to systemic inflammation and toxicity of bile acids are key pathomechanisms implicated in CN patients.

Methods

This was a prospective cohort study conducted at the Institute of Liver and biliary sciences, New Delhi from January 2019 to October 2021. The study was approved by the Institutional Ethics Committee (IEC)/Institutional Reviewer Board (IRB), Institute of Liver & Biliary Sciences (F25/5/107/ILBS/AC/2016/11252/01) and initiated after obtaining written informed consent from each patient or the next of kin The study was conducted in accordance with the Declaration of Helsinki and International Council for Harmonisation Good Clinical Practice (GCP) guidelines. The protocol is summarized in the Supplementary Appendix. Written informed consent was obtained from patients or their authorized representatives. ACLF was diagnosed in accordance with the definition by Asia Pacific association for the study of liver (APASL)11. At 48 h, patients with volume non-responsive AKI were categorized as either hepatorenal syndrome (HRS, n = 15), ATN (n = 15) or CN (n = 15) based on UM. Patients with bland urine sediment were classified HRS-AKI if they in addition had a therapeutic response to terlipressin and intravenous albumin, patients with bilirubin crystals or bile casts with or without RTECs as CN while patients with muddy brown granular casts were classified as ATN. Majority of patients with CN were managed as HRS-AKI. (Supple. Fig. 1) These patients were prospectively followed for the progression or resolution of AKI and the response to the therapeutic interventions. We excluded patients with mixed phenotype on UM i.e. with presence of bile casts and granular casts, other active sediment like dysmorphic RBC, waxy casts and pus cells or those with active infection. We also excluded patients with hepatocellular carcinoma (HCC), advanced cardiopulmonary disease, pregnancy, obstructive renal pathology, chronic kidney disease, acute or sub-acute liver failure and human immunodeficiency virus infection. Post-mortem renal biopsies were performed including detailed experiments in the biopsy-proven CN and ATN patients. (Supple. Fig. 2).

Urine and plasma cytokines, bile acids, DAMPs and markers of endothelial function

A detailed assessment of a panel of urine biomarkers, inflammatory cytokines, serum bile acids, endotoxin assay, endothelin-1, markers of systemic inflammation was performed in the urine and plasma of these patients. A panel of 17 renal biomarkers (MILLIPLEX MAP Kit; HKI1MAG-99 K and HKI2MAG-99 K; MilliporeSigma; Burlington, Massachusetts, USA), lipocalin-2/neutrophil gelatinase-associated lipocalin (NGAL), cystatin-C; renin; alpha-1 microglobulin; glutathione-S-transferase-alpha (GST-alpha); fatty acid binding protein-1(FABP-1); interferon inducible protein-10 (IP-10); tissue-inhibitor of metalloproteinase-1 (TIMP-1); kidney injury molecule-1(KIM-1); albumin, collagen IV; trefoil factor-3 (TFF-3); epidermal growth factor (EGF); osteoactivin; osteopontin (OPN); calbindin; clusterin was performed.

A panel of 29 different pro and anti-inflammatory cytokines/chemokines, and growth factors (MILLIPLEX MAP Kit; HCYTMAG-60 K; MilliporeSigma; Burlington, Massachusetts, USA) including interferon-gamma and alpha (IFN-γ, IFN-α), interleukins, (IL-4, IL-12,IL-15,IL-2,IL-17,IL-5, IL-6, IL-7, IL-8 , IL-10, interleukin-1 receptor antagonist (IL-1RA), macrophage inflammatory protein (MIP-1α , MIP-1β), granulocyte colony stimulating factors (G-CSF), granulocyte macrophage colony stimulating factors (GM-CSF), tumor-necrosis factor alpha and beta (TNF-α, TNF-β), eotaxin, monocyte chemoattractant protein (MCP-1, MCP-3), regulated on activation normal T-cell expressed and secreted (RANTES), interferon gamma-induced protein (IP-10), epithelial growth factor (EGF) , transforming growth factor (TGF ) and vascular endothelial growth factor (VEGF) were analyzed.

The plasma cytokines were classified as type 1, type 2 and type 3 and other cytokines and growth factors13. Quantitative assessment of serum bile acids was performed using spectrophotometry by diazyme manufacturer with a detection range of 1–180 µmol/L, lower limit of sensitivity 0.47 µmol/L with no cross reactivity with an intra-assay co-efficient of variation (CV) of 3.9%, an inter-assay CV of 2.9% and reference range of 0–10 µmol/L. High mobility group protein B1(HMGB1) was measured in plasma of same group of patients using ELISA based commercial kits from SunRed Diagnostics Pvt. Ltd. (Shanghai, China) according to manufacturer’s protocol. Plasma endotoxin levels were measured in the same group of patients using Limulus amoebocyte lysate (LAL) assay, a chromogenic quantification assay (Elabscience, USA) according to the manufacturer’s protocol. The data was represented with the arbitrary unit EU/ml. Endothelin-1 levels were measured using enzyme linked immunosorbent assay (ELISA).

Metabolomics in the plasma was performed using an ultra-high-performance liquid chromatographic system (HPLC) followed by high-resolution orbitrap mass spectrometry (HRMS) (details of experimental methods have been provided in the Supple. Appendix).

Experiments in biopsy-proven CN compared to ATN

Renal histology: Detailed analysis was performed in biopsy-proven CN (n = 5) and ATN (n = 5) for the following.

Morphological analysis: The diagnosis of CN and ATN was rendered based on histology.

Immunohistochemistry

Immunohistochemistry was performed for mitochondrial biogenesis marker sirtuin-1 (SIRT-1-Abcam, ab32441, 1:100), hypoxia-inducible factor-1 (HIF-1-(Novus Biologicals, NB100-479, 1:200),, sodium-glucose linked cotransporter-1 (SGLT-1 Abcam, ab14685, 1:250) and aquaporin-2 (AQP-2- Sigma, SAB4503085, 1:200).

Terminal deoxynucleotidyl transferase nick end labelling (TUNEL) assay

Extent of renal injury cell death analysis was performed by in-situ labeling of apoptosis-induced DNA strand breaks (TUNEL assay) using the in-situ cell death detection kit (Roche; CAT No. 11684795910) as per the manufacturer’s protocol.

Metabolomics and quantitative reverse transcriptase polymerase chain reaction (qRT-PCR) in renal tubule epithelial cells (RTEC). The exfoliated RTECs of biopsy proven CN (n = 5) and ATN (n = 5) were subjected to high resolution mass spectrometry. We performed the quantitative expression of the following five genes associated with mitochondrial biogenesis (peroxisome proliferator-activated receptor-gamma coactivator-1alpha (PGC-alpha); cyclo-oxygenase-1 (COX-1); nuclear receptor factor (NRF); succinate dehydrogenase (SDH); transcription factor A mitochondrial (TFAM) by qRT-PCR in RTEC.

Leucocyte energy metabolism: Systemic bioenergetic adaptation energy metabolism of freshly isolated peripheral blood mono nuclear cells (PBMC) was performed using Agilent Seahorse XFe24 Extracellular Flux Analyzer. (Details in supple. Appendix).

Management protocol: Patients were managed in accordance with the standard of care.

(Details of methodology and patient management is provided in supple. Appendix).

Definitions: Hepatorenal syndrome (HRS) in this study was diagnosed in patients with acute-on-chronic liver failure (ACLF) using established clinical and laboratory criteria as defined by international club of ascites criteria. At the 48-h evaluation point after initial volume resuscitation, patients demonstrating acute kidney injury (AKI) with persistent elevation in serum creatinine and a bland urine sediment were considered for HRS-AKI and subsequently these patients were initiated on terlipressin. The process was strictly by exclusion and we ruled out other AKI causes such as:​

  1. 1.

    Obstructive uropathy or structural renal pathology using ultrasonography where indicated

  2. 2.

    Overt infections and sepsis through clinical and microbiological screening

  3. 3.

    Chronic kidney disease by careful review of past renal function and morphology

  4. 4.

    Urine microscopy actively excluded patients with granular casts (indicative of ATN), dysmorphic RBCs (active glomerulonephritis), or waxy casts/pus cells.

Statistical methods

For the comparison of continuous and categorical variables, appropriate statistical tests were employed. Specifically, the Student T-test or the Mann–Whitney U-test was applied to continuous variables, depending on the distribution of the data. For categorical variables, Pearson’s chi-square test was used. The data were summarised based on their distribution. Normally distributed data were expressed as mean values along with the standard deviation (± SD). In cases where the data were not normally distributed, the results were presented as median with the corresponding range. Qualitative data were reported as numbers with percentages.

Bioinformatics and multivariate analysis

Metaboanalyst was utilised for further analysis of the data. Principal Component Analysis (PCA) and Partial Least Square Discrimination Analysis (PLS-DA) were performed to segregate biomarkers and to delineate the spectrum of acute kidney injury (AKI).

Metabolites that demonstrated independent associations were ranked according to their p-values and false discovery rates. Pathways that were either protective or associated with clinical outcomes were subsequently explored in greater detail.

Random forest and heat map analyses

Random Forest Analysis (RFA), a supervised classification technique based on an ensemble of decision trees, was conducted. The “mean decrease in accuracy” (MDA) was calculated to assess the importance of each variable.Heat map analysis was also performed to classify cytokines and growth factors in relation to the AKI spectrum.

Development of clinical score based on binary logistic regression analysis

For the creation of a clinical scoring system, we used univariable and multivariable analysis using binary logistic regression analysis. The variables significant on univariable analysis were assessed for multicollinearity. Considering, small sample size, two multivariable models including 3 parameters using backward stepwise logistic regression analysis was used. P value less than 0.05 was considered significant.