Collecting duct cells show differential retinoic acid responses to acute versus chronic kidney injury stimuli

Retinoic acid (RA) activates RA receptors (RAR), resulting in RA response element (RARE)-dependent gene expression in renal collecting duct (CD). Emerging evidence supports a protective role for this activity in acute kidney injury (AKI) and chronic kidney disease (CKD). Herein, we examined this activity in RARE-LacZ transgenic mice and by RARE-Luciferase reporter assays in CD cells, and investigated how this activity responds to neurotransmitters and mediators of kidney injury. In RARE-LacZ mice, Adriamycin-induced heavy albuminuria was associated with reduced RA/RAR activity in CD cells. In cultured CD cells, RA/RAR activity was repressed by acetylcholine, albumin, aldosterone, angiotensin II, high glucose, cisplatin and lipopolysaccharide, but was induced by aristolochic acid I, calcitonin gene-related peptide, endothelin-1, gentamicin, norepinephrine and vasopressin. Compared with age-matched normal human CD cells, CD-derived renal cystic epithelial cells from patients with autosomal recessive polycystic kidney disease (ARPKD) had significantly lower RA/RAR activity. Synthetic RAR agonist RA-568 was more potent than RA in rescuing RA/RAR activity repressed by albumin, high glucose, angiotensin II, aldosterone, cisplatin and lipopolysaccharide. Hence, RA/RAR in CD cells is a convergence point of regulation by neurotransmitters and mediators of kidney injury, and may be a novel therapeutic target.


Scientific RepoRtS
| (2020) 10:16683 | https://doi.org/10.1038/s41598-020-73099-9 www.nature.com/scientificreports/ both 25 μM DEAB and 1 μM AGN193109 23 (Fig. 2a). In the mIMCD-3 mouse inner medullary CD cell line 23 ( Fig. 2b) and the M-1 mouse cortical CD cell line 48 (Fig. 2c), CD-derived Hoxb7 B2 mesenchymal cells (MSCs) 49 ( Fig. 2d), as well as a human cortical CD cell line 50 (Fig. 2e), clear endogenous RA/RAR activity repressed by both DEAB and AGN193109 was detected. In all these CD-derived cell cultures, 10 mg/ml albumin significantly repressed RA/RAR activity. In mIMCD-3 cells, we found that 10 mg/ml albumin, but not IgG and transferrin, repressed RA/RAR activity, indicating albumin specifically repressed RA/RAR activity (Fig. 3a). The effect of albumin was also dose-dependent at clinically relevant concentrations ranging from 0.3 mg/ml (equivalent to microalbuminuria) to 10 mg/ml (equivalent to heavy albuminuria) (Fig. 3b). To further confirm the specificity Figure 1. In Adriamycin nephropathy of RARE-LacZ mice, heavy albuminuria was associated with repressed RA/RAR activity. (a) RA/RAR activity in CD cells as a function of RARE-LacZ expression was reported as the ratio between number of β-gal-expressing CD cells and total number of DBA positive CD cells. Heavy albuminuria was associated with a 2.6-fold reduction of RA/RAR activity in CD cells; (b) Representative images of kidney tissues stained with an anti-β-galactosidase antibody (green), DBA-lectin (red, CD cells) and Topro-3 (blue, nuclei). In healthy control RARE-LacZ mice, RA/RAR activity (green signal) was physiologically confined to CD cells. In Adriamycin nephropathy mice, heavy albuminuria was associated with reduced RA/RAR activity in CD cells. *, ***: p < 0.05, p < 0.001, respectively. www.nature.com/scientificreports/ of albumin in repressing RA/RAR in CD cells and to shed light into the underlying mechanism, we found that overexpression of wild-type ALB, but not the gene truncated of the sequence encoding its RA-binding domain 40 , significantly repressed the endogenous RA/RAR activity (Fig. 3c). These data indicate that albumin excreted into urine could be reabsorbed by CD cells, leading to sequestration of RA and repression of RA/RAR activity in the cells, similar to what we reported previously in glomerular progenitor cells 40 . Although no RA/RAR activity was detected in any non-CD cells in the healthy kidney in RARE-LacZ mice 21,22 , we do not rule out the possibility that non-CD renal cells have functional RA/RAR machinery that is physiologically below the detection limit of X-gal assay for reporting RA/RAR-dependent expression of LacZ gene 21 . Using in vitro cellular models and quantitative RARE dual luciferase reporter assays, we found that cultured mouse and rat proximal tubular cell lines (mPTEC and NRK-52E) and rat renal fibroblasts (NRK-49F) had endogenous RA/RAR activity, equivalent to 1/3-1/4 of the activity of mIMCD-3 cells. As in mIMCD-3 cells, the RA/RAR activity in these non-CD cell lines was repressed by DEAB, AGN193109, as well as albumin (Supplementary Figs. 3 & 4). Thus, although this project focuses on CD cells, some of our findings may also have implications for other relevant renal cell types. For example, physiologically, proximal tubules and the CD reabsorb 71% and 3% of albumin filtrated from glomeruli, respectively 51 . This means that albumin may exert stronger physiological repression on RA/RAR activity in proximal tubules, likely contributing to lack of visible physiological RA/RAR activity in these tubules in RARE-LacZ mice 21 .
RA/RAR activity in CD cells was repressed by high glucose. Diabetes characterised by high blood and urinary glucose levels is a leading cause of CKD 52 . To examine whether high glucose may regulate RA/RAR activity, we compared the effects of 24 h exposure to 25 mM glucose (high glucose) and 5.5 mM glucose (normal medium control) on RA/RAR activity in mIMCD-3 cells, with 25 mM mannitol as an osmotic control for highglucose treatment. We found that high glucose, but not equal molar concentrations of mannitol, significantly repressed RA/RAR activity (Fig. 4a).

RA/RAR activity in CD cells was differentially regulated by mediators of hypertension and neurotransmitters.
Hypertension is another major risk factor for CKD 53 . To examine whether mediators of hypertension may regulate RA/RAR activity in CD cells, we compared the effects of four well-established mediators of hypertension on RA/RAR activity in mIMCD-3 cells. We found that angiotensin II (0.01-1 μM) and aldosterone (0.01-1 μM) dose-dependently repressed, while endothelin-1 (1-50 nM) and vasopressin (100 nM) dose-dependently induced RA/RAR activity (Fig. 4b-e). More recently, renal nerves have been shown to contribute to hypertension 54,55 , AKI and its transition to CKD [56][57][58][59] . We then examined whether neurotransmitters norepinephrine, calcitonin gene-related peptide (CGRP) and acetylcholine may also regulate RA/RAR activity in CD cells. As shown in Fig. 4f,g, norepinephrine (10 nM) and CGRP (1-100 nM) activated, while acetylcholine (0.1-10 µM) repressed RA/RAR activity in CD cells (Fig. 4h). These results suggest that mediators of hypertension and neurotransmitters may have opposite effects on the RA/RAR activity in CD cells, apart from their systemic effects on blood pressure and organ perfusion. In view that angiotensin II and aldosterone are wellestablished mediators of CKD progression 3,4,53 , while vasopressin and norepinephrine are vasopressors often used to boost organ perfusion and prevent AKI in shock patients 60 , it is possible that differential regulation of RA/RAR activity in the CD may be a novel determinant in their differing roles in AKI and CKD, in addition to their systemic effects on blood pressure. www.nature.com/scientificreports/ Of note, although we tested 0.1-10 nM norepinephrine and 1-100 nM vasopressin, only at the highest concentrations they induced RA/RAR activity. 10 nM norepinephrine is 4-20-fold higher than the plasma concentrations of healthy subjects 61 and comparable with plasma concentrations of patients with phaeochromocytoma 61 or those receiving norepinephrine as a vasopressor 62 . Vasopressin, however, primarily targets the renal CD. Thus, although its circulating concentrations in healthy subjects and patients are at picomolar levels 63,64 , it might be enriched at the CD and need higher local concentrations to regulate CD functions. In keeping with this notion, in mIMCD-3 cells, it took 10 nM vasopressin to activate the expression of the H + -K + -ATPase α 2 -subunit (HKα 2 ) gene cells 65 and 100 nM vasopressin was required to activate RA/RAR activity. Thus, both pathophysiological and pharmacological implications of our findings deserve further investigation.

RA/RAR activity in CD cells was differentially regulated by nephrotoxic agents and other inducers of kidney injury.
Depending on dosing regimens and duration of exposure, nephrotoxic agents aristolochic acid I and gentamicin, the chemotherapeutic drug cisplatin and endotoxin lipopolysaccharide (LPS) can induce AKI and/or CKD 1,66-69 . Although they are all broadly considered nephrotoxic, they are known to cause renal damage by different mechanisms 66 . For example, instead of direct cytotoxicity, cisplatin and LPS both induce AKI in a Toll-like receptor-4 (TLR4)-dependent manner 67,68 . As shown in Fig. 4i,j, aristolochic acid I (20-40 μM) and gentamicin (0.25-1 mg/ml) dose-dependently induced, while cisplatin (10-40 μM) and LPS (1-10 μg/ml) dose-dependently repressed RA/RAR activity in mIMCD-3 cells (Fig. 4k,l). Given the emerging evidence of a critical protective role for the RA/RAR activity in CD cells 9 , our results suggest that, LPS and cisplatin may disarm RA/RAR-mediated defensive mechanisms, leading to more progressive renal kidney injury, while activation of RA/RAR activity by aristolochic acid I and gentamicin may mount defence and mitigate injury.

RA/RAR activity in autosomal recessive polycystic kidney disease (ARPKD) renal cystic cells was significantly lower than in age-matched normal CD cells.
In ARPKD, all renal cysts derive from the CD 70 . To examine whether RA/RAR activity in CD-derived ARPKD renal cystic cells is abnormal, we compared the endogenous RA/RAR activity in primary cultures of embryonic (foetal 24 weeks), perinatal (newborn) and paediatric (18 months old) ARPKD renal cystic cells and their age-matched normal human CD cells. Endogenous RA/RAR activity was calculated in two forms: (i) Endogenous RA/RAR activity repressed by 1 μM RAR antagonist AGN193109 (EndoAct AGN ) = RARE-Luc activity (Vehicle) − RARE-Luc activity (AGN193109); (ii) Endogenous RA/RAR activity repressed by 25 μM of the RA biosynthesis inhibitor DEAB (EndoAct DEAB ) = RARE-Luc activity (Vehicle) − RARE-Luc activity (DEAB). As shown in Fig. 5a-c, CD-derived ARPKD renal cystic cells consistently had significantly lower EndoAct AGN and EndoAct DEAB than age-matched normal CD cells. These data suggest that RA/RAR activity in ARPKD renal cystic cells may be reduced and the biological importance of this change in ARPKD deserves further investigation. www.nature.com/scientificreports/ RA-568 was more potent than atRA in rescuing RA/RAR activity repressed by albumin, aldosterone, angiotensin II, cisplatin, high glucose and LPS. Given the hypothesised protective role for the RA/RAR activity in the CD 9 , to prevent CKD and its progression, it may be desirable to prevent RA/RAR repression by albumin, aldosterone, angiotensin II, cisplatin, high glucose and LPS in these cells. To this end, we propose that the RARα agonist RA-568 41 may be more suited than natural RAR pan-agonist atRA, because (i) it is structurally distinct from atRA ( Supplementary Fig. 3), RA-568 may not be sequestered by albumin as much as atRA 40,45 ; (ii) RA-568 activates RARα, but not RARβ and RARγ, nor RXR 41 , thus it will selectively activate RARα, the most abundant RAR isotype in CD cells 71 and the kidney (Supplementary Fig. 1), while minimising RARγmediated irritating skin syndrome 19 and RXR-mediated pro-fibrotic effects of atRA metabolites 72 . Indeed, in both mouse (mIMCD-3) and human (HCD) CD cell lines, we found that RA-568 was at least 100-fold more efficacious than atRA in restoring RA/RAR activity repressed by 10 mg/ml albumin. Though not totally unexpected, it was striking to find that atRA at concentrations as high as 100 ng/ml had no effect on the RA/RAR activity repressed by 10 mg/ml albumin. This was likely because high-concentration albumin sequesters both endogenous and exogenous atRA. In contrast, in both mIMCD-3 mouse CD cells and HCD human CD cells, RARα agonist RA-568 efficaciously restored RA/RAR activity repressed by 10 mg/ml albumin. At 1-10 nM, it restored RA/RAR loss of activity; and at 100 nM the RA/RAR activity was significantly higher than the vehicle control group (Fig. 6a,b). Further, as shown in Supplementary Fig. 5, RA/RAR activity induced by 100 nM RA-568 in both mIMCD-3 and HCD cells was not repressed by increasing concentrations of albumin (0.3-10 mg/ml), in contrast to 100 nM atRA, which neither boosted RA/RAR activity nor prevented RA/RAR activity from being repressed by albumin in a dose-dependent manner (1-10 mg/ml). These data suggest that selectively activating RARα by RA-568 is likely a better approach than pharmaceutical atRA to restoring and boosting RA/RAR activity in CD cells. We next examined the effects of atRA and RA-568 in preventing RA/RAR repression by LPS and cisplatin in mIMCD-3 cells. As shown in Fig. 6c, RA/RAR activity repression by either LPS (10 μg/ml) or cisplatin (50 μM) was not affected by 1-100 nM atRA but was significantly and dose-dependently prevented by RA-568, at 10-100 nM and 1-100 nM, respectively. Similarly, we compared effects of atRA and RA-568 in preventing RA/ RAR repression by angiotensin II, aldosterone and high glucose. As shown in Fig. 6d-f, RA/RAR repression by either angiotensin II, aldosterone or high glucose in mIMCD-3 cells was prevented by 100 nM RA-568; 10 nM RA-568 also effectively prevented RA/RAR activity repression by aldosterone and high glucose. In contrast, 1-100 nM atRA did not prevent angiotensin II, aldosterone and high glucose from repressing RA/RAR activity. The ineffectiveness of atRA in all these settings was in stark contrast to the excellent efficacy of 1-100 nM atRA in preventing RA/RAR repression by 25 μM DEAB, an inhibitor of RA synthesis (Figs. 2a, 6d) 21 .

Discussion
AKI and CKD are multifactorial disorders [2][3][4]7 . This study highlights the RA/RAR activity in CD cells as a convergence point of regulation by the multiple factors involved in AKI and CKD. Intriguingly, it responds in an opposite fashion to mediators of kidney injury (induced by gentamicin and aristolochic acid I; repressed by cisplatin and LPS), mediators of hypertension (induced by endothelin-1, vasopressin and norepinephrine; repressed by aldosterone and angiotensin II) and neurotransmitters (induced by CGRP and norepinephrine; repressed by acetylcholine). Further investigations into the mechanisms and biological consequences of RA/RAR activity changes induced by these factors may facilitate deeper understanding of the heterogeneity and multifactorial nature of AKI and CKD and guide refined therapies.
In particular, this report establishes that RA/RAR activity responds differentially to a wide range of neurotransmitters and mediators of kidney injury and thus is unlikely a bystander in AKI and CKD. This rationalise www.nature.com/scientificreports/ the next step of studies to test the exact roles for the RA/RAR in the CD in different AKI and CKD models, e.g. by selectively up-and down-regulating RA/RAR activity and its downstream genes in the CD cell lineage using Hoxb7 or AQP2 promoters and then examining whether this changes the outcomes of AKI and CKD models in mice 9 .
The main limitation of this work is that evidence we have presented so far is mainly from in vitro models. Although our in vitro and in vivo data both support the capacity of heavy albuminuria to repress RA/RAR activity in CD cells (Fig. 1, 2, 3), we are acutely aware that regulation of the RA/RAR pathway in the CD must be more complex in vivo than in cultured CD cells, and eventually, in vitro, in vivo and clinical studies must be analysed together to fully understand the regulation of the RA/RAR pathway in the CD. For example, high glucose has been reported to repress RAR expression in mIMCD-3 cells 71 . Thus, reduced RAR expression may, at least in part, contribute to high-glucose repression of RA/RAR activity in mIMCD-3 cells. In diabetic mouse kidneys, atRA biosynthesis and atRA-dependent gene expression are compromised 73 , at least in part, due to o-GlcNAcylation of proteins crucial for retinoid binding, metabolism and signalling 74 . Furthermore, a biphasic regulation of renal RA/RAR signalling has been suggested by transcriptomic analysis of renal biopsy tissues of early-and late-stage diabetic nephropathy, respectively 75 .
Similar to our findings in mIMCD-3 cells, LPS has been previously reported to repress RA/RAR signalling in hepatic stellate cells 76 and gentamicin has been reported to induce RA/RAR activity in renal tubules in zebra fish 39 . Given the potential importance of RA/RAR signalling in the CD 9 and the potential differences in vitro and in vivo, it is desirable to clarify how renal RA/RAR activity is regulated in different types of AKI and CKD in RARE-reporter animals 21,22,39,40 and in related clinical settings. To know exactly how this pathway is regulated will pave the way for further mechanistic studies, which will guide the development of intervening strategies, if indicated.
Taking ARPKD, a disease of the focal adhesion complex and the primary cilia due to mutations of fibrocystin (encoded by PKHD1) 70 , for example, we found that ARPKD renal cystic cells had significantly lower endogenous RA/RAR activity than age-matched control CD cells. Although we have three age-matched pairs leading to exactly the same findings, larger sample size may be needed to further confirm our findings. To examine a possible causal relation, it may be necessary to examine whether genetic mutations of PKHD1 in normal CD cells repress RA/RAR signalling and, if so, its role in ARPKD pathogenesis. Based on RA/RAR-dependent gene profiling in CD cells 21 , we hypothesise that decreased RA/RAR activity in CD cells may reduce expression of focal adhesion complex components tensin (Tns1) and α2-integrin (Itga2), and the cilium-specific transcription factor Foxj1 (Foxj1), impair signalling through focal adhesion complexes and cilia, leading to disease progression. If www.nature.com/scientificreports/ this is true, restoring the RA/RAR activity in the CD cell lineage may curb progression of this devastating disease and becomes a cost-effective strategy to improve ARPKD prognosis. The inefficacy of atRA in preventing RA/RAR repression by albumin, high glucose, cisplatin, LPS, aldosterone and angiotensin II was an important finding. Since in parallel studies atRA was always highly efficacious in preventing RA/RAR repression by the atRA biosynthesis inhibitor DEAB, RA/RAR repression by the aforementioned mediators of kidney injury may not be simply due to lack of the bioactive ligand atRA. Thus, further mechanistic studies are clearly needed to guide development of both retinoid-and non-retinoid-based rescuing strategies. The good news is that, in contrast to the poor efficacy of atRA, the synthetic RARα agonist RA-568 has shown excellent effect in preventing RA/RAR repression. As this compound has an established safety profile in preclinical studies in mice and rats 41 , it could be an ideal drug lead for AKI and/or CKD, by boosting or restoring RA/RAR activity in kidneys, particularly in the CD. Thus, our in vitro data justify further translational studies comparing the efficacy and safety of RA-568 and atRA as potential renal therapeutics.
Despite the focus on CD cells of the present study, we acknowledge the possibility that RA/RAR activity in other renal cell types could be activated in pathological conditions. If so, how renal RA/RAR activity pathologically re-distributes in different AKI and CKD, whether such pathologically activated signalling in other renal cells, such as proximal tubular cells and infiltrating cells in ischaemia-reperfusion AKI 39 and renal progenitor cell in Adriamycin-induced nephropathy 40 , has the same roles as the RA/RAR in the CD, and how all these RA/RAR signalling interacts are important questions to answer. Eventually, all such knowledge should need be integrated to fully understand the pathological roles for the RA/RAR signalling in AKI and CKD and to devise intervention strategies.
In conclusion, we have found that RA/RAR activity in cultured CD cells responds to a wide range of neurotransmitters and mediators of kidney injury, either being induced or repressed. When repressed, the activity is poorly rescued by exogenous atRA, but is effectively reversed by RARα agonist RA-568. Further studies revolving around this newly found convergent point of regulation in AKI and CKD may help us better understand these interconnected syndromes and lead to novel therapeutics.

Methods
Studies on animal tissues. Immunofluorescence confocal microscopy. No new animal study was performed in this study. Instead, we performed confocal microscopy (LSM510 META, Carl Zeiss, Jena, Germany) on 10 μm frozen sections of renal tissues of RARE-LacZ mice (CD1 background, 6-to 10-week-old, female) collected in our previously published project, including nine with trace albuminuria at dip-stick tests (approximately 0.1-0.2 mg/ml albumin or albumin:creatinine ratio < 0.3) and six with high levels of albuminuria, with albumin:creatinine ratios of > 7 (17.7 ± 2.99), induced by two retro-orbital injections of 14 mg/kg Adriamycin in phosphate-buffered saline (PBS), as well as five healthy controls treated by two successive injections of PBS 40 . Animal experiments were approved by the University of Florence Animal Studies Review Board, performed in accordance with institutional, regional, and state guidelines, and adhered to the National Institutes of Health Guide for the Care and Use of Laboratory Animals. Dolichos biflorus agglutinin (DBA)-lectin (Vector Laboratories, Peterborough, UK) was used to stain CD cells. Chicken anti-β-galactosidase polyclonal antibody (Abcam, Cambridge, UK) and AlexaFluor 488-labelled donkey anti-chicken antibody (Jackson ImmunoResearch, West Grove, PA) were used to stain LacZ-expressing cells and Topro-3 (Molecular Probes, Invitrogen, Carlsbad, CA) was used for nuclear staining. Sections were also stained with secondary antibodies only to serve as negative control, as we previously reported 40,77 . To quantify the number of LacZ-expressing cells in the CD, β-gal-and DBA-lectin-positive cells were counted in at least 4 renal cortex sections of each mouse and percentage of dual positive cells was calculated. Five fields were counted on each section. Cell biology studies. Reagents. Acetylcholine, AGN193109, aldosterone, aristolochic acid I, cisplatin, CGRP, DEAB, dexamethasone, D-glucose, endothelin-1, gentamicin, human albumin, human angiotensin II, human IgG, human transferrin, LPS, mannitol, norepinephrine, tetracycline, atRA and vasopressin were purchased from Sigma-Aldrich Company Ltd, Gillingham, UK. RA-568 was synthesised by Sygnature Discovery, Nottingham, UK. All reagents dissolved in ethanol and/or dimethylsulphoxide (DMSO) were first diluted to 1000 × stock solution and then diluted 1000 × with culture medium to the working concentrations. Control groups were treated with 0.1% ethanol and/or DMSO.  24,48,78 . They were routinely grown in DMEM-F12 medium containing 100 μg/ml penicillin and 100 μg/ml streptomycin (thereafter known as antibiotics) (PAA Laboratories Ltd, Somerset, UK), supplemented with 5% foetal calf serum (FCS; Life Technologies Ltd, Paisley, UK). Both mIMCD-3 and M-1 cells were 100% positive for aquaporin-2 (AQP2), a CD principal cell marker, and mPTEC cells were 100% positive for proximal tubular cell marker aquaporin-1 (AQP1). NRK-52E and NRK-49F cells (LGC Standards) are epithelial and fibroblast-like cell clones from kidneys of an adult non-inbred Osborne-Mendel rat 79 . They were routinely grown in 5% FCS DMEM with antibiotics (PAA). NRK-49F cells stained positive for vimentin and negative for cytokeratin and factor VIII. RNAs extracted from M-1, mIMCD-3 and mPTEC were selectively amplified using mouse specific primers, while RNA extracted from NRK-49F cells was selectively amplified using rat specific primers only. HCD cells are a well-characterised human cortical CD cell line 50  www.nature.com/scientificreports/ Dr Joan Li (University of Queensland, Queensland, Australia) and fully characterised as reported before 49 . They were cultured in 20% FCS α-MEM medium with antibiotics (PAA). Primary human CD cells and their agematched embryonic (24 weeks), newborn and paediatric (18-month old) ARPKD renal cystic cells were fully characterised by Professor Patricia Wilson [80][81][82] and were obtained from the Polycystic Kidney Disease Charity Biobank. Both foetal and paediatric ARPKD renal CD cells and their age-matched normal CD cells were routinely grown in 3% FCS 50% Click's medium: 50% RPMI 1640 media (Gibco, Life Technologies) supplemented with antibiotics and 5 µg/ml human transferrin, 5 × 10 -8 M dexamethasone, 2 mM GlutaMax (Gibco). Cells were plated in 6-well plates coated with 4.1 mg/ml collagen type-I (Corning, NY) and left to attach overnight before further studies. All cell cultures were maintained at 37 °C and 5% CO 2 .
Transient transfection and RARE dual luciferase assays. Three tetracycline-inducible gene expression plasmids were procured from VectorBuilder, Santa Clara, CA for transient transfection studies: (i) ALB-EGFP, a plasmid with DNA inserts of enhanced green fluorescence protein (EGFP) and the wild-type ALB gene (Accession NM_000477.5); (ii) mutant ALB-EGFP, a plasmid with DNA inserts of EGFP and mutated ALB genes, encoding human albumin deleted of amino acids 139-189, including 11 of 12 amino acids critical for binding atRA 45 ; and (iii) a control vector, with the EGFP DNA insert only. The maps of these plasmids are shown in Supplementary  Fig. 6. Lipofectamine LTX, Plus reagents and Opti-MEM I Reduced Serum Medium (Gibco) were used in transient transfection of plasmid DNAs according to the manufacturer's instructions. For RARE dual luciferase assay, pGL3-RARE-luciferase (https ://www.addge ne.org/13458 /) 83 and PRLSV40 (https ://www.addge ne.org/vecto r-datab ase/3949/) plasmids (Promega, Southampton, UK) were used to co-transfect cells at a ratio of 5:1 to report RARE-dependent expression of Firefly luciferase and constitutive expression of Renilla luciferase. Cells were lysed with Reporter Lysis Buffer (Promega) and luminescence signal was detected with Luciferase Assay System and Beta-Glo Assay System (Promega). "Background signal" was determined as the average readout from wells treated with transfection medium only without any plasmid DNA and this was subtracted from readouts of other wells. The reporter assay results were expressed as relative unit (RLU, ratios of readouts of the two reporters). Alternatively, RLU of treatment group was normalised to vehicle control group and expressed as fold-change. Because pGL3-RARE-Luc and the control PRLSV40 plasmid both have the SV40 minimal promoter, the RLU (ratios of readouts of the two reporters) always comprises two parts, one dependent on RA/RAR and RARE (RA/RAR activity) and the rest being "baseline signal" mediated by the minimal promoter in the pGL3-RARE-Luc plasmid. The expected effects of DEAB and AGN193109 on the reporter assay are as follows: (i) DEAB represses RA/RAR activity induced by endogenously synthesised RA but does not repress the effects of the trace amount of RA in 1% FCS in the culture medium and any exogenous RAR agonists, if applicable, nor does it repress any RA-independent transcription activity of RAR mediated by the AF-2 domain of the nuclear receptor 84 ; (ii) AGN193109 is not only a potent pan-RAR antagonist (antagonising effects of RAR agonists and inhibiting RA/RAR-mediated AF1-mediated gene expression) but also a RAR inverse agonist (actively recruiting corepressors to RAR and thus repressing RAR agonist-independent AF2-mediated transcriptional activity of RAR) 84 . At the concentrations used in this study, both DEAB and AGN193109 almost completely switched off RA/RAR-dependent Dhrs3, Sprr1a and Ppbp mRNA expression, but AGN193109 was consistently more potent in pressing these gene expression (− 99%, − 95% and − 93%) than DEAB (− 96%, − 85% and − 86%) 23 . Thus, DEAB incomplete repression of RAR activation by endogenous RA, DEAB inability to repress effects of RA in 1% FCS cell culture medium and AGN193109 incomplete repression of RA/RAR activity may not play any major role in the RLU values unrepressed by DEAB and AGN193109 in the RARE dual luciferase results in this project, although some minor roles cannot be completely ruled out. Nonetheless, this study particularly focuses on endogenous RA/RAR activity, which is defined as net RLU values repressed by both DEAB and AGN193109.
For ALB overexpression studies, mIMCD-3 cells were seeded into a 24-well plate at 4 × 10 4 cells/well in 500 μl 5% FCS-DMEM-F12 without antibiotics. After culturing for 16 h, cells were co-transfected with tetracyclineinducible ALB-EGFP, mutant ALB-EGFP or the EGFP-only control plasmid, together with pGL3-RARE-luciferase and PRLSV40 plasmids for 6 h. Medium was then changed to 1% FCS DMEM-F12. After an overnight incubation, medium was changed to 1% FCS DMEM-F12 supplemented with 2 μg/ml tetracycline for 24 h to switch on transgenes. www.nature.com/scientificreports/ For reporter assays involving primary human CD cells and age-matched ARPKD renal cystic cells, cells were seeded into 96-well plates, 3.5 × 10 4 /well, in 500 μl 3% FCS 50% Click's medium: 50% RPMI 1640 without antibiotics. After culturing to about 70% confluence, cells were co-transfected with pGL3-RARE-luciferase and PRLSV40 plasmids and cultured for another 6 h. Medium was then changed to fresh 1% FCS medium. After an overnight incubation, medium was changed to 1% FCS medium with 1 μM AGN193109, 25 μM DEAB, 10 mg/ ml albumin or vehicle and cultured for 24 h. Data analysis. Immunofluorescence confocal microscopy data are expressed as percentage of LacZ-expressing DBA-positive cells over total DBA-positive cells per renal cortex section. All cell biology experiments were performed in quadruplicate wells for each group, with reproducible results in at least two independent experiments. Results are presented as mean ± SEM. For comparison of two and more than two groups, t-tests (nonpaired, two-tail; with Welsh's correction if any data of comparison failed Shapiro-Wilk normality tests) and oneway ANOVA analysis (including Tukey's multiple comparisons) were conducted, respectively. For comparison of fold-changes, data were log-transformed before statistical analysis. Statistical significance was determined as p < 0.05. Analyses were performed using GraphPad Prism, Version 8.0 (GraphPad Software Inc., CA).