Endoplasmic reticulum stress drives proteinuria-induced kidney lesions via Lipocalin 2

In chronic kidney disease (CKD), proteinuria results in severe tubulointerstitial lesions, which ultimately lead to end-stage renal disease. Here we identify 4-phenylbutyric acid (PBA), a chemical chaperone already used in humans, as a novel therapeutic strategy capable to counteract the toxic effect of proteinuria. Mechanistically, we show that albumin induces tubular unfolded protein response via cytosolic calcium rise, which leads to tubular apoptosis by Lipocalin 2 (LCN2) modulation through ATF4. Consistent with the key role of LCN2 in CKD progression, Lcn2 gene inactivation decreases ER stress-induced apoptosis, tubulointerstitial lesions and mortality in proteinuric mice. More importantly, the inhibition of this pathway by PBA protects kidneys from morphological and functional degradation in proteinuric mice. These results are relevant to human CKD, as LCN2 is increased in proteinuric patients. In conclusion, our study identifies a therapeutic strategy susceptible to improve the benefit of RAS inhibitors in proteinuria-induced CKD progression.

C hronic kidney disease (CKD) is characterized by a progressive decline in renal function to end-stage renal disease that can occur, irrespective of the cause of renal damage, when a critical number of functional nephrons has been lost. CKD is now a worldwide concern: 410% of the US residents suffer from this disease 1 and similar rates are found in developing countries 2 . Understanding the pathophysiology of CKD progression is therefore a critical challenge for public health.
Most CKD are characterized by abnormalities of the glomerular filtration barrier, leading to increased glomerular permeability and abnormal filtration of macromolecules, such as albumin. Convergent evidences from clinical and experimental studies indicate that albuminuria and proteinuria are not simply markers of CKD progression, but active players in the evolution of the disease 3 . Mechanistically, it has been proposed that the proteins that escape into the glomerular filtrate have a toxic effect on tubular cells and that, once damaged, tubular cells lead to the development of interstitial fibrosis and inflammation 4-6 . Remarkably, several clinical studies have shown that the decline of renal function correlates more closely with the tubulointerstitial lesions than with the glomerular damage 7 . Hence, over the last 20 years, researchers have focused their efforts on the discovery of the molecular links between proteinuria and the development of tubulointerstitial lesions. Several candidates have been identified, that is, endothelin-1, MCP-1, RANTES or complement components 5,6,8 . However, so far, this did not lead to the development of novel therapeutic strategies susceptible to slow down CKD progression in humans. The only available strategy to counteract the deleterious effect of proteinuria is the renin-angiotensin system (RAS) inhibition, which reduces the leakage of proteins from the glomerular filtration barrier 9 . However, a residual proteinuria is observed in most patients under RAS blockade and the nephroprotective effect of RAS inhibitors may decline over time [10][11][12] . Moreover, attempts to further increase RAS blockade in proteinuric patients have revealed a high rate of severe side effects, such as hypotension or life-threatening hyperkalemia 13 . Thus, there is an urgent need to identify novel therapeutic targets susceptible to add benefit to RAS inhibition by preventing the toxic effect of residual proteinuria.
The endoplasmic reticulum (ER) has emerged as a signalling platform that responds to various cellular stresses by inducing a coordinated response, the unfolded protein response (UPR) 14 . During UPR, inositol-requiring enzyme 1a (IRE1a) promotes the phosphorylation of c-JUN and the specific splicing of UPR transcription factor X-box binding protein 1 (XBP1). Besides, protein kinase R-like kinase (PERK) phosphorylates eukaryotic translation-initiation factor 2a (eIF2a): this reduces general translation but promotes translation of activating transcription factor 4 (ATF4), which activates the CCAAT/enhancer-binding protein homologous protein (CHOP). If this adaptive response cannot overcome ER stress, it triggers apoptotic cell death. UPR and ER stress can be targeted by various therapeutic compounds either Food and Drug Administration approved or in preclinical studies 15 . Interestingly, a few studies showed that UPR is activated in tubular cells exposed to albumin [16][17][18] , but the pathophysiological role of such activation remains unknown 17,18 .
Here we combined in vivo and in vitro studies to dissect a novel molecular pathway in which albumin leads, via calciumdependent ER stress activation, to Lipocalin 2 (LCN2) overexpression, which in turn triggers tubular cell apoptosis and renal lesions. More importantly, we showed that inhibiting ER stress with 4-phenylbutyric acid (PBA) prevents proteinuria-induced renal lesions and LCN2 overexpression.

Results
Proteinuria leads to UPR activation in vitro and in vivo. To investigate whether ER stress is a common response to protein overload, we first studied UPR activation in several complementary models of genetic and acquired glomerular diseases. We observed that both PERK and IRE1 pathways were activated in proteinuric WT1 þ /mut mice, which carry a mutation of Wt1 (ref. 19), a gene encoding an essential transcription factor of podocyte homeostasis (Fig. 1a) and Nphs2 D/ À mice, which display dramatic podocin downregulation on tamoxifen exposure 20 (Supplementary Fig. 1a). The activation of these pathways was confirmed in mice receiving doxorubicin (doxo) or serum bovine albumin, two widely used models of acquired proteinuric nephropathy 21 (Supplementary Fig. 1b,c). To ascertain that UPR was the direct consequence of proteinuria, we incubated renal tubular cells with albumin. Immunofluorescence experiments using albumin-fluorescein isothiocyanate showed the progressive (from 1 h) accumulation of albumin in endocytotic vesicles of mouse inner medulla collecting duct (mIMCD-3) cells, followed by a later co-localization of albumin with lysosomal markers (LAMP2; Fig. 1b). Nevertheless, albumin exposure was associated with phosphorylation of PERK, eIF2a and c-JUN from 5 min of incubation, which was followed by the subsequent overexpression of ATF4 and CHOP, and XBP1 splicing ( Fig. 1c-e).These results demonstrate that the activation of PERK and IRE1 pathways occurs before albumin internalization.
Albumin-induced UPR activation is calcium dependent. We then investigated the mechanisms by which albumin induces UPR in tubular cells. Previous studies have demonstrated that albumin exposure leads to reactive oxygen species (ROS) production and intracellular calcium increase, two biological processes known to induce ER stress 22,23 . Interestingly, within 30 min of albumin exposure, whereas we failed to detect any increase of ROS (Fig. 1f), we observed a rapid increase of cytosolic calcium concentration ([Ca 2 þ ] cyt ; Fig. 1g). Thapsigargin, a known inhibitor of the sarco/ER Ca 2 þ -ATPase that leads to ER Ca 2 þ release, and thereafter to a Ca 2 þ influx called store-operated calcium entry (SOCE), induced a very similar [Ca 2 þ ] cyt rise (Fig. 1g). Addition of albumin after thapsigargin stimulated a slight second [Ca 2 þ ] cyt increase, suggesting that albumin and thapsigargin share the same pathway to mobilize Ca 2 þ . To confirm the role of the SOCE in this mobilization, we repeated the same experiments in the presence of two SOCE-specific inhibitors gadolinium (Gd 3 þ ) and SKF96365 (ref. 24). As shown in Fig. 1h, Gd 3 þ and SKF96365 partially abolished the albumin-induced [Ca 2 þ ] cyt rise, demonstrating that albumin induces the rise of [Ca 2 þ ] cyt by triggering Ca 2 þ release from the ER, which in turn activates a SOCE. We then used 1,2-bis(oaminophenoxy)ethane-N,N,N,N'-tetraacetic acid (BAPTA), a calcium-specific aminopolycarboxylic acid, to chelate the extracellular Ca 2 þ . When BAPTA was added at the same time as albumin, chelation of extracellular Ca 2 þ fully prevented the [Ca 2 þ ] cyt rise ( Supplementary Fig. 2a), demonstrating that in the absence of free Ca 2 þ albumin can induce neither the Ca 2 þ release from the ER nor the subsequent SOCE. However, when we added BAPTA 5 min after albumin exposure, albumin triggered a rapid increase of [Ca 2 þ ] cyt ; however, as soon as BAPTA was added, [Ca 2 þ ] cyt started to slowly decrease ( Supplementary Fig. 2a), demonstrating that albumin required extracellular Ca 2 þ to sustain the rise of [Ca 2 þ ] cyt . Depletion of Ca 2 þ from the ER compartment is known to induce ER stress and UPR activation. To demonstrate that albumin-induced Ca 2 þ release is responsible for the observed UPR activation, we treated tubular cells exposed to albumin with either N-acetylcystein (NAC), an inhibitor of ROS generation, or cadmium (Cd 2 þ ) and Gd 3 þ , two inhibitors of Ca 2 þ transporters. Consistent with the observation that ROS did not increase after 30 min of albumin exposure, NAC treatment affected neither PERK nor eIF2a phosphorylation ( Fig. 1i and Supplementary Fig. 2b). In contrast, Cd 2 þ and Gd 3 þ administration completely prevented UPR activation ( Fig. 1i and Supplementary Fig. 2b), indicating that intracellular Ca 2 þ is the critical intermediate between albumin and UPR activation in this pathological setting.
Albumin-induced UPR leads to LCN2 overexpression. As all these results indicated that albumin leakage from damage glomeruli leads to UPR in tubular cells, we next aimed at identifying the potential targets of UPR activation during  | UPR is induced in tubular cells exposed to albumin in vivo and in vitro. (a) p-eIF2a, ATF4, CHOP and p-c-JUN evaluated in kidneys of WT1 þ /mut mice (n ¼ 10) and their wild-type littermates (Control, n ¼ 6) 6 weeks after birth (representative panels scale bar, 100 mm; insert scale bar, 10 mm). (b) Internalization of extracellular albumin in mIMCD-3 cells (exposed to 1% albumin-fluorescein isothiocyanate (Alb-FITC) for 30 min and 1, 3 and 24 h) evaluated using anti-LAMP2 antibodies (red), a marker of the lysosomal compartment. Pictures show endocytosis within 30 min to 1 h of the albumin and it's targeting to the lysosomal compartment after 24 h of treatment (n ¼ 3; scale bar, 20 mm). (c) pPERK, p-eIF2a and p-c-JUN expression during timecourse exposition of mIMCD-3 cells to 1% albumin (n ¼ 5). (d) ATF4 and CHOP expression in mIMCD-3 cells exposed to 1% albumin (Alb) for 24 h (upper panel) and splicing of XBP1 evaluated by reverse transcriptase-PCR (down) in mIMCD-3 cells exposed to 1% albumin for 3 h (lower panel; n ¼ 3).
(e) Quantifications of ATF4 and CHOP protein abundance and mRNA of CHOP or spliced XBP1 in mIMCD-3 cells exposed to 1% albumin for 24 or 3 h, respectively (n ¼ 6 and n ¼ 5 for protein and mRNA quantifications, respectively). (f) Measurement by dichlorodihydrofluorescein diacetate (DCFh-DA) of intracellular ROS generation in mIMCD-3 cells exposed for 30 min to 1% albumin or 100 mM H 2 O 2 (n ¼ 3). (g) Measurement of intracellular calcium in mIMCD-3 cells exposed to 1% albumin (black trace) or first exposed to 1.0 mM thapsigargin (red trace, red arrow 1) and then to 1% albumin (red trace, red arrow 2; n ¼ 3). (h) Measurement of intracellular calcium in mIMCD-3 cells exposed to 1% albumin for indicated times in control conditions (black trace), with gadolinium (Gd 3 þ , green trace) or with SKF96365 (red trace) (n ¼ 3). (i) Representative western blotting (upper panel) and quantification (lower panel) of pPERK evaluated in mIMCD-3 cells exposed to 1% albumin for 30 min in the presence or the absence of NAC, Cd 2 þ or Gd 3 þ (n ¼ 5 tubulointerstitial lesion development. Among the possible candidates, we focused on LCN2 (also known as NGAL, 24p3, siderocalin or uterocalin), a small, secreted iron-transporting protein, as (i) it has been recently reported that UPR activation is associated with LCN2 overexpression in cancer cells 25 and (ii) we previously showed that LCN2 is critically involved in the progressive deterioration of tubules following nephron reduction 26 . We observed that Lcn2 messenger RNA and protein expression progressively increased in proteinuric WT1 þ /mut mice as compared with control littermates (Fig. 2a-  (f) Representative western blotting (upper panel) and quantification (lower panel) of LCN2 protein abundance in mIMCD-3 cells exposed to 1% albumin for 24 h, treated with NAC, Cd 2 þ or Gd 3 þ (n ¼ 5). (g) LCN2 protein expression in mouse renal primary cultured cells exposed to 1% albumin or heatdenatured (denatured) 1% albumin for 24 h (n (m) LCN2 protein and mRNA expression in wild-type ATF4 þ / þ and ATF4 À / À MEFs exposed to 1% albumin or 0.5 mM thapsigargin (Th) for 24 h (n ¼ 3). (n) LCN2 mRNA expression in ATF4 À / À MEFs transfected with empty vector (EV) or ATF4 construct and exposed to 1% albumin or 0.5 mM thapsigargin (Th) for 24 h (n ¼ 3). Statistical analysis: one way analysis of variance followed by Tukey-Kramer test; Mann-Whitney test when only two groups are compared; *Po0.05, **Po0.01, ***Po0.001 versus controls. and few collecting ducts) but not in glomeruli ( Supplementary  Fig. 3a-c). Consistent with this observation, LCN2 was not found to be increased in glomeruli from WT1 þ /mut mice as compared with wild-type littermates ( Supplementary Fig. 3a,b). These results were confirmed in the other experimental models of proteinuria ( Supplementary Fig. 4a-c). More importantly, we observed that these findings were not restricted to mice: in patients with proteinuric nephropathies LCN2 immunoreactivity remarkably increased, but exclusively in tubular cells ( Supplementary Fig. 5), suggesting that LCN2 overexpression is a common response to protein overload. In favour of this idea, we demonstrated that the expression of LCN2 dramatically increased when cultured tubular cells were directly exposed to albumin (Fig. 2d,e). Inhibition of calcium entry by Cd 2 þ and Gd 3 þ completely prevented LCN2 upregulation (Fig. 2f), suggesting that UPR is an essential step in albumin-induced LCN2 overexpression. On the other hand, heat denaturation experiments showed that LCN2 induction was similar, regardless of the albumin state ( Fig. 2g), excluding the possibility that compounds bound to albumin account for the overexpression of LCN2, as reported for other biological outcomes 27 . Moreover, we showed that thapsigargin and tunicamycin, two compounds known to trigger ER stress, lead to a marked induction of Lcn2 gene expression both in vivo, in treated mice (Fig. 2h,i), and in vitro, in mIMCD-3 cells (Fig. 2j), demonstrating that ER stress per se induces LCN2 expression.
Albumin induces LCN2 overexpression through ATF4. We then investigated the molecular links between UPR and LCN2 expression. UPR induces several transcription factors such as ATF4, CHOP, XBP1 or ATF6 (ref. 14). To determine whether one of them is involved in LCN2 overexpression, we transfected mIMCD-3 cells with vectors encoding each of these factors and studied Lcn2 mRNA and protein expression. Interestingly, only ATF4 and CHOP were able to induce the expression of the endogenous Lcn2 gene (Fig. 2k). However, when we used a Lcn2-luciferase reporter vector, we observed that ATF4, but not CHOP, significantly stimulated the Lcn2 promoter activity (Fig. 2l), indicating that the effect of CHOP on the endogenous Lcn2 promoter was either indirect or the result of a binding to a site located outside of the sequence used to drive the reporter gene. Consistent with the Lcn2-luciferase experiment, in silico analysis of the Lcn2 promoter confirmed the presence of a putative ATF4 binding site located at À 1,091/ À 1,070 upstream of the transcription start site (TSS: Chr2, 32243259). To ascertain that ATF4 is required for UPR-induced LCN2 expression, we exposed Atf4-null mouse embryonic fibroblast (MEF) to either albumin or thapsigargin. Experiments showed that albumin exposure induced ER stress in MEF as in tubular cells ( Supplementary Fig. 6a,b). We observed that in the absence of ATF4, UPR could no more stimulate LCN2 synthesis (Fig. 2m), despite the persistent overexpression of CHOP under ER stress ( Supplementary Fig. 6a). In addition, rescue experiments using an ATF4 expression vector ( Supplementary Fig. 6b,c) showed that ATF4 overexpression in Atf4-null MEF restored the albumin-and thapsigargin-induced LCN2 expression (Fig. 2n). Taken together, these results demonstrate that ATF4 is necessary for the efficient induction of Lcn2 gene expression during UPR.
LCN2 deficiency protects tubular cells from proteinuria. To investigate whether LCN2 is the critical effector of this novel molecular pathway, we generated double transgenic mice in which the Wt1 mutant allele was introduced in mice lacking the Lcn2 gene 26 . As expected, WT1 þ /mut XLcn2 þ / þ mice developed severe renal lesions 6 weeks after birth (Fig. 3a,b). However, the frequency and the severity of tubular lesions were significantly reduced in proteinuric WT1 þ /mut mice lacking Lcn2 (Fig. 3a). In contrast, neither the severity of glomerular lesions (Fig. 3b) nor the level of proteinuria (Fig. 3c) was affected by Lcn2 gene inactivation. Consistent with the persistence of severe proteinuria, the induction of UPR was similar in tubular cells of WT1 þ /mut mice, regardless of the Lcn2 genotype ( Supplementary Fig. 7a,b).
LCN2 deficiency reduces apoptosis via ROS modulation. We next wondered which were the cellular mechanisms involved in LCN2-induced tubular injury during proteinuria. Uncontrolled and prolonged ER stress leads to apoptosis 28 and LCN2 has been shown to favour apoptosis in multiple cellular types 29 . As tubular cell apoptosis has been shown to play a role in CKD progression in proteinuric nephropathies 30 , we combined in vivo and in vitro experiments, to investigate whether LCN2 might regulate tubular cell apoptosis during proteinuria. In vivo, we observed that the number of TUNEL (TdT-mediated dUTP nick end labelling)-positive tubular cells was significantly reduced in proteinuric WT1 þ /mut mice lacking Lcn2 as compared with WT1 þ /mut XLcn2 þ / þ littermates (Fig. 3e). Similarly, the expression of active caspase 3, a well-known marker of apoptosis, was significantly reduced in the kidneys of WT1 þ /mut XLcn2 À / À mice ( Supplementary  Fig.  8). Consistently, in vitro, the number of apoptotic cells was significantly decreased in Lcn2-silenced tubular cells when stressed with albumin ( Fig. 3f and Supplementary Fig. 9). The mechanisms of the pro-apoptotic role of LCN2 are still a matter of debate 29 . As ROS generation has been involved in ER stressinduced apoptosis 31 , we investigated whether LCN2 acts in a ROS-dependent manner in stressed cells. Our results showed that LCN2 deficiency significantly reduced ROS content in tubular cells after 24 h of albumin exposure (Fig. 3g). Consistent with this observation, Lcn2-silencing prevented albumin-induced overexpression of heme oxygenase 1 (HO-1), a gene involved in the anti-oxidative response ( Supplementary Fig. 10a). Notably, when NAC was added to albumin-treated cells, apoptosis was reduced at the same extent as those in cells lacking Lcn2 (Fig. 3h). As expected, we observed that NAC treatment prevented albumin-induced intracellular ROS accumulation ( Supplementary Fig. 10b). As NAC did not inhibit LCN2 expression (Fig. 2f), we concluded that LCN2 induces apoptosis after prolonged albumin exposure by promoting ROS generation.
PBA delays CKD progression during proteinuria. Several therapeutic strategies have been developed to modulate pathological ER stress, such as chemical chaperones that improve ER function, possibly by attenuating protein misfolding and stabilizing protein conformation 15 . Among them, the PBA is already used with a good safety profile in humans 32 . To provide the rational bases for a novel pharmacological strategy susceptible to counteract the toxic effect of proteinuria during CKD progression, we treated WT1 þ /mut mice with PBA. As expected, PBA treatment resulted in a marked reduction of ER stress in kidneys of proteinuric WT1 þ /mut mice, as judged by the decrease of p-eIF2a, ATF4, CHOP and p-c-JUN ( Supplementary  Fig. 11a). Remarkably, PBA administration was associated with a dramatic improvement of kidney lesions and function: tubular lesions (Fig. 4a), interstitial fibrosis (Fig. 4c) and plasma creatinine (Fig. 4b) were reduced in WT1 þ /mut PBA-treated mice as compared with WT1 þ /mut vehicle-treated littermates. Similarly, the number of tubular apoptotic cells was significantly reduced in WT1 þ /mut mice receiving PBA as compared with the vehicle-treated animals (Fig. 4d). It is worth noting that as observed in WT1 þ /mut mice lacking LCN2, the reduction of tubulointerstitial lesions was independent of the severity of glomerular injury and proteinuria ( Supplementary Fig. 11b-d). Importantly, the beneficial effect of PBA was associated with a dramatic reduction of LCN2 expression in both proteinuric WT1 þ /mut mice (Fig. 4e,f) and mIMCD-3 cells exposed to albumin (Fig. 4g,h). The decrease of LCN2 synthesis was associated with reduced ATF4 expression (Fig. 4g), suggesting that LCN2 downregulation by PBA was directly related to UPR inhibition. Consistently, PBA also prevented thapsigargininduced LCN2 expression ( Supplementary Fig. 12).
PBA decreases urinary LCN2 in a proteinuric patient. To finally explore the potential relationship between PBA and LCN2 in humans, we took advantage of a patient followed in our nephrology department for a lysinuric protein intolerance (a urea cycle disorder due to a SLC7A7 gene mutation), which was associated with amyloidosis and proteinuria. PBA, which is currently used in patients with urea cycle disorders as an ammonium-scavenging agent 32 , was recommended because of persistent hyperammoniema. A renal biopsy performed before PBA treatment revealed that the amyloid deposits involved exclusively the glomeruli and not the interstitial or vascular compartments. Proteinuria was severe (2.6 g per day), despite the use of a RAS inhibitor. Immunohistochemistry revealed that LCN2 expression was remarkably increased in renal tubular cells (Fig. 4i). Urinary LCN2 excretion was also dramatically increased as compared with healthy controls. Remarkably, serial urine collections revealed that PBA treatment resulted in a progressive reduction of urinary LCN2 excretion, despite sustained levels of residual proteinuria (Fig. 4j).

Discussion
Proteinuria is now recognized not simply as a marker but also as a major risk factor of CKD progression. Despite the fact that RAS inhibitors reduce the leakage of proteins in some patients, therapeutic strategies capable of counteracting the deleterious effect of residual proteinuria on the tubulointerstitial compartment are still lacking. By combining experimental models of proteinuric nephropathies with in vitro and in vivo genetic and pharmacological approaches, we uncovered a novel ER stress pathway. We demonstrated that proteinuria stimulates, via a calcium release-induced ER stress, the overexpression of LCN2, which in turn leads to tubular apoptosis and renal lesions (Fig. 6).
More importantly, we showed that inhibition of this pathway by PBA, a pharmacological strategy already used in humans, delayed renal deterioration in proteinuric mice. Collectively, these data identify a novel therapeutic strategy and suggest a crucial role for albumin/ER stress/LCN2 pathway in modulating the progression of CKD.
Our study showed that ER calcium release is a crucial event in the signalling pathway leading to albumin-induced ER stress. Indeed, we observed that extracellular albumin rapidly induces an ER calcium release followed by a store-operated calcium entry that maintains the cytosolic calcium increase. Interestingly, a previous study showed that albumin induces ER stress and [Ca 2 þ ] cyt increase in cultured podocytes through TRPC6 (transient receptor potential cation channel, subfamily C, member 6) activation 33 . Whether TRPC6, which is expressed in tubular cells 34 , plays a role in our experimental model remains to be elucidated. Furthermore, our experiments using BAPTA revealed that albumin requires extracellular calcium to induce ER calcium release. Taking all these data together, it is tempting to speculate that albumin binds to a protein receptor in a calcium-dependent manner, to trigger the intracellular events responsible for ER calcium release. Interestingly, LRP2/Megalin, a well-known albumin receptor, binds its ligands in an extracellular calcium-dependent manner 35,36 . It is worth noting that LRP2 is expressed in mIMCD3 cells (Supplementary Fig. 13a,b).
In the present study, we showed that AFT4 is mandatory for the efficient induction of Lcn2 gene expression in albuminoverloaded tabular cells. In fact, albumin failed to induce Lcn2 in Atf4 À / À cells and overexpression of ATF4 completely rescued the phenotype. These data are consistent with the observation that ATF4 directly binds to the Lcn2 promoter 31 . A previous study has reported that CHOP is able to bind to the Lcn2 promoter as well and is able to induce LCN2 expression, at least, in human cancer cells 37 . As a recent work has shown that CHOP and ATF4 interact to induce target genes during UPR 31 , it is possible that ATF4 and CHOP cooperate to induce LCN2 expression in albumin-overloaded tubular cells.
We previously showed that LCN2 plays a crucial role in CKD progression during nephron reduction and cystogenesis, by mediating the mitogenic effect of epidermal growth factor receptor 26 . In the present study, we demonstrated that LCN2 acts rather by triggering apoptosis in protein-overloaded tubular cells. Interestingly, albumin overload did not lead to epidermal growth factor receptor phosphorylation ( Supplementary Fig. 14), suggesting that LCN2 might act as a central integrator of multiple signalling pathways leading to CKD progression. We also observed that LCN2 triggers apoptosis through ROS generation. The mechanisms by which LCN2 promotes oxidative stress in this pathological setting are still unknown. It is possible that an increase in intracellular iron content plays a role, as LCN2 has been shown to transport iron and modulate mitochondrial apoptosis via Bcl-2like protein 11 (BIM) expression 38,39 . In favour of this idea, LCN2 endocytosis has been shown to increase ROS generation through an intracellular accumulation of iron leading to apoptosis 40 . Alternatively, LCN2 might act by inhibiting the nuclear factor-erythroid 2 p45-related factor 2 antioxidant pathway, as LCN2 has been also shown to sequestrate catechol 41 , a small metabolite product that could activate nuclear factor-erythroid 2 p45-related factor 2 (ref. 42). Whether the secreted or the intracellular form of LCN2 is involved in ROS-induced tubular apoptosis remains to be elucidated.
The substantial increase of the survival rate in proteinuric WT1 þ /mut mice lacking LCN2 might appear bigger than expected, given the difference in renal function. However, the experimental design of the study could explain, at least in part, this discrepancy. In fact, kidney lesions and apoptosis were evaluated at 6 weeks, whereas the difference in mortality rate was discernable 4 weeks later. Considering the natural history of CKD progression, it is very likely to be that both kidney lesions and renal failure were much greater at this time point than at 6 weeks. In favour of this idea, it has been shown that this model progresses very rapidly towards the complete destruction of the kidney 19 . Nevertheless, as we studied mice bearing a germline Lcn2 inactivation, we cannot rule out the possibility that LCN2 deficiency affected the lifespan of WT1 þ /mut mice via a systemic effect. It is known that CKD is associated with increased cardiovascular mortality 43 Figure 6 | Proposed model of proteinuria-induced tubulointerstitial damage. This study shows that by inducing Ca 2 þ -dependent ER stress, albumin activates ATF4 expression, which in turn stimulates LCN2 production. LCN2 then leads to increased apoptosis and tubulointerstitial damage.
NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10330 ARTICLE Whether LCN2 inhibition might reduce the cardiovascular lesions and mortality associated with CKD progression is an interesting question that deserves further investigation. Urinary LCN2 excretion has been shown to be a valuable biomarker of CKD progression 45 , including in proteinuric nephropathies 46 . Despite the fact that proteinuria is now recognized as a major risk factor of CKD progression, the rate of progression varies among proteinuric patients 47 . By showing that LCN2 is a critical target of proteinuria, our data suggest that urinary LCN2 could be helpful to mark the proteinuric patients with the higher risk of CKD progression.
Chemical chaperones have recently received increasing interest as a novel therapeutic approach for the treatment not only of rare genetic diseases leading to misfolded proteins but also of more frequent diseases, that is, metabolic diseases or cancers, characterized by ER stress 48 . Besides, multiple new therapeutic compounds targeting specific UPR branches are currently under development 15 . In the present study, we showed that pharmacological modulation of ER stress is an efficient treatment for preserving renal function and morphology in proteinuric mice, at least in part, by decreasing LCN2 expression. The observation, however, that PBA administration, but not LCN2 deficiency, decreased the severity of interstitial fibrosis in WT1 þ /mut mutant mice suggests that PBA might act by modulating additional actors of CKD progression. Notably, the benefit was observed despite the persistence of severe proteinuria. More importantly, our clinical observation that PBA dramatically reduced LCN2 overexpression in a proteinuric patient suggests that this treatment might be valuable in humans. Although we are aware that these promising results must be confirmed, they provide new therapeutic perspectives for patients who display residual proteinuria despite RAS inhibition 11 or develop adverse side effects on RAS inhibitors 13 .
In conclusion, by identifying a novel therapeutic strategy able to counteract the toxic effect of proteinuria on tubular cells, our study provides the rational for the development of an innovative ready-to-use multitarget therapy for proteinuric CKD patients, susceptible to inhibit not only proteinuria leakage but also its tubular toxicity. We expect that adding molecular chaperones will significantly improve the benefit of RAS inhibitors in delaying the progression of CKD, a major public health problem.

Methods
Animals. Mice were on FVB/N background unless otherwise specified. Mutant mice used for these studies were as follows: (i) mutant Wt1 knock-in mice (WT1 þ /mut mice) bearing a heterozygote Wt1 mutation 19 , (ii) mutant Nphs2 mice (Nphs2 D/ À mice), bearing a floxed Nphs2 exon 2 allele, a-null Nphs2 allele and a podocyte-expressed, tamoxifen-responsive Cre recombinase 20 , and (iii) Lcn2 À / À mice 26 . Animals were fed ad libitum and housed at constant ambient temperature in a 12-h light cycle. Animal procedures were approved by the Departmental Director of 'Services Vétérinaires de la Préfecture de Police de Paris' and by the ethical committee of the Paris Descartes University (approval number: A75-15-34).
For Nphs2 mice, experiments were performed on 6-week-old male and female triallellic Nphs2 lox/ À ,Cre þ mice and their control littermates (Nphs2 lox/ À ,Cre _ mice). Cre recombinase was induced by intraperitoneal administration of tamoxifen (33 mg kg À 1 per day for 5 days; Sigma) and mice were killed 3-4 weeks after the last injection (n ¼ 6 for each genotype).
For the doxo experiments, Balb/c females were injected intravenously with doxo (10 or 12 mg kg À 1 ; Sigma) or vehicle (saline) and killed either 14 or 25 days after injection (nZ5 per group). PBA or vehicle was given in drinking water as described above, either from the day of doxo injection (Day 0, preventive study) or from 4 days after doxo injection (Day 4, intervention study), and mice were killed 2 weeks later (n ¼ 6, n ¼ 10, n ¼ 10 and n ¼ 5 in control, vehicle-treated doxo-injected, PBA-D0 doxo-injected and PBA-D4 doxo-injected group, respectively).
For BSA experiments, FVB/N females were uninephrectomized at 12 weeks and then intraperitoneally injected with BSA (25 mg g À 1 ; Sigma) or vehicle (saline) for 4 weeks (n ¼ 4 per group).
For all the studies, urine and plasma samples were collected at the time of killing and the kidneys were removed for morphological, protein and mRNA studies.
Primary cultures of renal cortical tubular cells were prepared as previously described 49 . Briefly, kidney cortex tubules from 3-to 4-week-old mice were micro-dissected, dissociated by collagenase treatment, washed in Hank's solution by centrifugation and filtered to obtain an enriched preparation of dissociated tubular cells. Cells were then cultivated to reach subconfluency. Cell cultures were exposed to 1% fatty acid-free BSA for 24 h and then cells were lysed in RIPA buffer for protein extraction.
Transfections and Luciferase reporter gene assays. Luciferase gene reporter assays were carried out in 24-well plates. In brief, subconfluent mIMCD-3 cells were co-transfected with pGL3 control vector or pGL3-24p3/Lcn2-Luc containing murine Lcn2 gene promoter (Addgene Inc., ID 25463) in combination with plasmids expressing protein of interest. pRL-SV40 Renilla control vector (Promega) was co-transfected in each condition. The luciferase activity was measured 36 h after transfection using a dual-luciferase reporter assay system (Promega) as recommended. Luciferase activity in each cellular lysates was normalized to Renilla luciferase activity. Results are presented as changes of transactivation of the murine LCN2 promoter constructs relative to the original activity of this promoter constructs in cells transfected with empty vector control plasmid.
Clinical samples. Renal biopsies were routinely performed in proteinuric (41 g per day) patients with diabetic nephropathy, membranous nephropathy or IgA nephropathy (n ¼ 3 per group). Renal biopsies from patients with resolute kidney injury that did display neither significant proteinuria nor histological lesions were used as controls.
The patient treated with PBA was a 33-year-old man, who developed proteinuric CKD associated with lysinuric protein intolerance. Blood pressure was 116/84 mm Hg, measured glomerular filtration rate (iohexol clearance) was 51 ml min À 1 per 1.73 m 2 , serum albumin levels were 38 g l À 1 and proteinuria 2.6 g per day. The patient was treated with irbesartan, hydrochlorothiazide, citrullin and sodium benzoate, and successively PBA (3 g per day) was added. Morning spot urine samples were obtained before PBA and every week after the beginning of the treatment for 2 months. Urine samples were stocked at À 80°C after centrifugation. No other therapeutic modification was made once PBA was introduced. Informed consent was obtained from all patients. Our institutional review board approved the study (approval number: DC-2009-955).
Urine and plasma analysis. For mice samples, urinary albumin and creatinine levels were measured using an Olympus multiparametric autoanalyser (Instrumentation Laboratory). Plasma creatinine was measured using Konelab Analyzer (Thermo Fischer Scientific).
For human samples, urinary LCN2 concentration was evaluated by enzyme-linked immunosorbent assay (R&D Systems) according to the manufacturer's instructions. Proteinuria and urinary creatinine was measured using a Hitachi 917 analyser (Roche Diagnostics).
Renal morphology. For morphology analysis, kidneys were fixed in 4% paraformaldehyde, paraffin embedded and 4 mm sections were stained with periodic acid-Schiff or Sirius red. The degree of glomerular lesion was evaluated using a semiquantitative score methodology as previously described 50 . The degree of tubular lesions was evaluated using semiquantitative score methodology as previously described 51 , with minor modifications in the evaluation of kidney lesions (0 ¼ normal, 1 ¼ involvement of o10% of the cortex, 2 ¼ involvement of 10-25% of the cortex, 3 ¼ involvement of 25-50% of the cortex, 4 ¼ involvement of 450% of the cortex). The degree of fibrosis was automatically quantified using Sirius red staining in eight representative fields (magnification Â 100) with a Nikon digital camera Dx/m/1200 and NIS software (Nikon), and expressed as the ratio between the Sirius red-positive surface and the total section area, as previously described 26 .
Apoptosis was detected in 4 mm sections of paraffin-embedded kidneys by TUNEL assay using the In Situ Cell Death Detection kit (Roche) according to the manufacturer's protocol. The number of apoptotic cells was determined as the number of TUNEL-positive nuclei per tubule in eight representative fields (magnification Â 400).
Albumin uptake in mIMCD-3 cells. Cells exposed to 1% fluorescein isothiocyanate-albumin (Sigma) for different time were exposed to rat anti-LAMP2 antibody (Abcam, Ab 13524). The secondary antibody was fluorescent goat anti-rat antibody (Invitrogen). Images were acquired using a Leica TCS SP5 AOBS microscope with a 40/1.25 (differential interference contrast (DIC)) (Leica Microsystems). Resulting digital images were acquire with LAS AF software and analysed using ImageJ.
Cell death measurement. Cellular death was assessed based on propidium iodide (PI) staining (Sigma) and Annexin V/PI staining (Molecular Probes) according to the manufacturer's protocols. Cell fluorescence was measured by flow cytometry. PI-negative cells were defined as viable. Flow cytometry analysis was performed on a FACSCalibur (BD Biosciences) followed by analysis using FlowJo. mRNA analysis. mRNAs were quantified in mouse kidneys and cultured cells by quantitative reverse transcriptase-PCR using an ABI PRISM 7700 Sequence Detection system (Applied Biosystems). Primers were as follows: Lcn2 (fwd) The splicing of XBP-1 mRNA was analysed by PCR using the following protocol: 94°C for 3 min, 29 cycles of 94°C for 30 s, 58°C for 30 s, 72°C for 30 s and 72°C for 3 min. We used the following primers: XBP-1 (fwd) 5 0 -ACAC GCTTGGGAATGGACAC-3 0 and (rev) 5 0 -CCATGGGAAGATGTTCTGGG-3 0 .
RPL13, SDHA and HPRT were used as the normalization controls as previously described 52 .
Western blotting. Western blottings were performed as previously described 26 using the following: a goat anti-mouse LCN2 antibody (R&D systems) at 1:1,000, a rabbit anti-mouse CHOP antibody (Santa Cruz Biotechnology) at 1:500, a rabbit anti-mouse p-c-JUN antibody (Cell Signaling Technology) at 1:1,000, a rabbit anti-mouse pPERK antibody (Cell Signaling Technology), a rabbit anti-mouse ATF4 antibody (Santa Cruz Biotechnology) at 1:500, a rabbit anti-mouse p-eIF2a antibody (Cell Signaling Technology) at 1:1,000 and a rabbit anti-mouse cleaved caspase 3 antibody (Cell Signaling Technology) at 1:1,000, followed by either a rabbit HRP-conjugated anti-goat antibody at 1:10,000 (Dako) or a donkey HRP-conjugated anti-rabbit antibody at 1:10,000 (Amersham). Mouse monoclonal anti-a-tubulin or anti-b-actin antibody (Sigma) was used as loading control. The uncropped versions of western blottings are shown in Supplementary Fig. 15.
ROS measurement. Two techniques were used to measure intracellular ROS content. mIMCD-3 cells were loaded with dichlorodihydrofluorescein diacetate (10 mM; LifeTechnologies) and incubated for 30 min at 37°C. After incubation, cells were washed twice with PBS, lysed and acquired on a microplate reader (Tristar LB 941, Berthold). For analysis of intracellular ROS content, CellROX Deep Red Reagent was used according to the manufacturer's instructions (LifeTechnologies) and analysed either by immunofluorescence or flow cytometry. Flow cytometry analysis was performed on a FACSCalibur (BD Biosciences) followed by analysis using FlowJo.
Calcium spectrofluorimetry. Intracellular Ca 2 þ concentration variations were studied using spectrofluorimetry. Recordings were made in a HEPES buffered saline medium (HBS) containing (in mM): 135 NaCl, 5.9 KCl, 2 CaCl 2 , 1.2 MgCl 2 , 11.6 Hepes, 11.5 glucose pH 7.3 adjusted with NaOH. mIMCD-3 cells were grown on glass coverslips in Petri dishes. Before confluence, coverslips were washed three times with HBS medium and then loaded with 4 mM Fluo-4 AM (Molecular Probes). After loading, coverslips were washed twice with Fluo-4 AM-free HBS medium and maintained in this medium. Just before recordings, a coverslip was introduced in a 1 cm width-3 ml quartz cuvette, containing 2 ml of HBS medium and inserted in a spectrofluorimeter (RF-1501 Shimadzu Corporation, Japan). Fluo-4 was excited at 494 nm and fluorescence was measured at 516 nm. The baseline fluorescence was recorded during 10 min before adding BSA. The [Ca 2 þ ] cyt was expressed as a ratio value of the fluorescence intensity divided by the baseline average fluorescence intensity.
Data analysis and statistics. Data were expressed as means ± s.e.m. Differences between the experimental groups were evaluated using analysis of variance followed by the Tukey-Kramer test. The log-rank test was used for survival analysis. When only two groups were compared, Mann-Whitney test was used. Assumptions for statistical analyses were met (that is, normal distribution and equal variance). Replicates used were biological replicates, which were measured by using different samples derived from distinct mice. The results are representative examples of more than three independent experiments. We estimated the sample size considering the variation and mean of the samples. No randomization or NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10330 ARTICLE NATURE COMMUNICATIONS | 7:10330 | DOI: 10.1038/ncomms10330 | www.nature.com/naturecommunications blinding was done. No animals were excluded from the analysis. The statistical analysis was performed using Graph Prism Software.