Renal tubule Cpt1a overexpression mitigates kidney fibrosis by restoring mitochondrial homeostasis

Chronic kidney disease (CKD) remains a major epidemiological, clinical and biomedical challenge. During CKD, renal tubular epithelial cells (TECs) suffer a persistent inflammatory and profibrotic response. Defective fatty acid oxidation (FAO), the main source of energy for TECs, contributes to kidney fibrosis. To determine if FAO gain-of-function (FAO-GOF) could protect from fibrosis, we generated a conditional transgenic mouse model with overexpression of the fatty acid shuttling enzyme carnitine palmitoyl-transferase 1 A (CPT1A) in TECs. Cpt1a knock-in mice subjected to three different models of renal fibrosis exhibited decreased expression of fibrotic markers, a blunted pro-inflammatory response and reduced epithelial cell damage. Mitochondrial number, oxygen consumption and ATP levels were restored after FAO-GOF. Studies in patients evidenced decreased CPT1 levels and increased accumulation of short and middle chain acylcarnitines, reflecting impaired FAO in human CKD. We propose that strategies based on FAO-GOF may constitute powerful alternatives to combat fibrosis inherent to CKD.


Introduction
Organ fibrosis constitutes a significantly prevalent pathological entity associated to high morbidity and mortality and a major biomedical challenge. While it may affect any tissue of the human body, its presence in the kidney generally indicates unrelenting progression to chronic renal failure, a condition associated to reduced expectancy and quality of life. Hence, fibrosis is a convergent pathway for prevalent pathologies underlying chronic kidney disease (CKD) such as diabetes, hypertension or nephrosclerosis 1 . Beyond proper blood pressure and glycemic controls, therapeutic options to revert or deter the progression of fibrosis are very limited. In the last few years, the understanding of major metabolic disturbances coexisting with kidney fibrosis have shed new light on the pathogenesis of fibrosis progression 2 . Among these, a drastic reduction in fatty acid oxidation (FAO) appears to be critical for the global energy failure occurring in the tubulo-interstitial compartment, contributing to immune cell infiltration and kidney fibrosis 3,4 . Here we tested whether specific metabolic gain-of-function in FAO in renal tubules was necessary and sufficient to counteract cellular and molecular changes associated to kidney fibrosis. Carnitine palmitoyl transferase 1A (CPT1A) is a rate-limiting and targetable enzyme in this pathway. We found that conditional overexpression of CPT1A resulted in significant mitigation of fibrosis and improvement of renal function in three different experimental models. This was related to enhanced mitochondrial mass, repaired architecture and bioenergetics recovery. The overexpression of CPT1A was also associated with a reduced inflammatory pattern and abrogation of TGF-beta-induced epithelial cell damage. Moreover, studies in patients with chronic renal failure showed that a reduction in the levels of CPT1A correlated with the degree of fibrosis. In addition, in a large cohort of diabetic patients with CKD a specific profile of increased plasma acylcarnitines was found, reinforcing the critical metabolic derangement of FAO associated to CKD.

Overexpression of tubular CPT1A results in phenotypic protection against kidney fibrosis.
To test the in vivo relevance of CPT1A and FAO for renal fibrosis, a mouse model with conditional, inducible expression of Cpt1a was engineered as described in Methods. Next, these mice were crossed with mice providing renal epithelial tissue specificity, Pax8-rtTA (extrarenal expression of Pax8 only occurs at the level of the thyroid without reported offtarget effects 5 ). The resulting mice expressed an optimized reverse tetracyclin-controlled transactivator (rtTA2s-M2) under the control of the Paxillin-8 gene promoter, which permits tissue specific expression of Cpt1a gene bearing TRE in proximal and distal tubules and collecting duct after doxycycline administration (Fig. 1a, b and Supplementary figure 1a, b). We next characterized the renal specific overexpression of CPT1A in this newly generated genetic mouse model based on the doxycycline inducible transgenic system Tet-On. CPT1A KI mice after doxycycline administration presented a 10-fold increase in CPT1A mRNA level in the whole kidney tissue compared with WT mice (Fig. 1c). This was accompanied by a marked augment of CPT1A protein expression in the kidney. Of note, no differences were observed when the liver tissue was analyzed (Fig.1d, e). Tubules were labeled by lotus tetragonolobus lectin (LTL), a marker for proximal tubules. Similarly, induced expression of CPT1A and GFP proteins co-localized in tubule segments as evaluated by immunohistochemistry (Fig. 1f). Importantly, overexpressed CPT1A presented a mitochondrial localization pattern, which was observed by double immunostaining by using the ATP synthase beta-subunit as a marker (Fig. 1g). To assess the magnitude of CPT1A overexpression on FAO, we analyzed the capacity to oxidize radiolabeled palmitate by the renal tissue. Functionally, the increase in CPT1A protein in the renal epithelium increased the capacity of kidney tissue to oxidize 14 C-palmitate as reflected in the levels of both 14 Cpalmitate-derived 14 CO2 as well as in 14 C-palmitate-derived acid-soluble products (Fig. 1h).

6
To evaluate the effect of the Cpt1a knock-in strategy on renal damage we first used the folic acid-nephropathy (FAN) injury model. Folic acid (FA) at high dosage induces the formation of crystals rapidly with tubular necrosis in the acute phase (1-14 days) and patchy interstitial fibrosis in the chronic phase (28-42 days). FAN is induced both by crystal obstruction and direct toxic effect to tubular epithelial cells. Animals were treated with doxycycline starting three weeks prior to FA injection. Histological parameters were evaluated in tissue samples from mice seven days after intraperitoneal (IP) injection of FA. Tubular atrophy and dilation were significantly attenuated and fibrosis was markedly ameliorated in Pax8-rtTAtg/0:tetO-Cpt1atg/0 compared with WT mice (Fig. 2a, left panel and Supplementary figure 2a, b and c). A significant reduction in the proportion of fibrosis as quantified by Sirius red was observed in the CPT1A KI model (Fig.2a, right panel). Renal function was reduced in FANtreated WT animals, as reflected by the increase in blood urea nitrogen (BUN) and creatinine, but this effect was blunted in mice overexpressing CPT1A (Fig. 2b, c). The protein expression of classical profibrotic markers, fibronectin (FN) and smooth muscle actin (aSMA) was also significantly reduced in the CPT1A KI mice (Fig. 2d, e). To confirm the beneficial effect of FAO-gain-of-function in kidney fibrosis, the latter was assessed by performing unilateral ureteral obstruction (UUO) for 3 and 7 days in mice overexpressing CPT1A in TECs. In this model, renal blood flow and glomerular filtration rate in the affected kidney are significantly reduced within 24 h, interstitial inflammation is maximal at 2-3 days while tubular dilation, tubular atrophy and fibrosis are evident after 7 days. Obstructed kidneys from WT animals showed significant tubulo-interstitial architectural and histological changes 7 days after UUO, while the extent of tubular atrophy and dilatation was markedly reduced in kidneys with increased levels of CPT1A (Supplementary figure 3a, left panel).
To determine the effect of CPT1A on fibrosis, Sirius red staining was performed to quantify the collagen content in the kidneys. Evaluation of renal lesions by light microscopy showed increased collagen deposition in the interstitial area after 7 days of the procedure in the 7 obstructed kidneys compared to the contralateral ones in WT mice. A significant protective effect of CPT1A overexpression was observed by a reduction of 20-40% in the collagen deposition (Supplementary figure 3a, right panel). As expected, circulating levels of BUN and creatinine were not different between WT and Cpt1a knock-in mice after 7 days of UUO due to the remaining functional kidney (Supplementary figure 3b, c). A less dramatic but still significant decrease in FN and SMA was also observed (Supplementary figure 3d).
To understand the mechanisms underlying the protective action of CPT1A on renal fibrosis, we performed expression analysis of whole kidney from WT or CPT1A KI mice subjected to FAN or 7-days UUO. Expression of genes related to critical cellular functions for the initiation and perpetuation of tubular dysfunction and chronic tissue damage was analyzed by using specific TaqMan probes. A reduced expression for this subset of genes was observed in the FAN model. QRT-PCR-based quantification displayed lower expression of epithelial injury (Havcr1-Kim-1) and fibrotic markers (collagens) in kidneys from CPT1A KI mice (Fig. 2f). and TGF-β involved in these cellular mechanisms related to kidney fibrosis (Supplementary   figure 5a, b). Data collected from the three models of CKD in the CPT1A KI mice strongly support that CPT1A is an enzyme, which, by itself, has a crucial impact on the outcome of fibrosis most likely due to its critical function in the facilitation of FAO.
CPT1A overexpression prevents mitochondrial dysfunction and restores FAO in the fibrotic kidney.

8
Renal mitochondrial abnormalities and dysfunction are common features in the pathogenesis of different forms of renal disease 6 . Cellular pathways promoting kidney damage can compromise mitochondrial homeostasis reflected in increased oxidative stress, apoptosis, microvascular loss and fibrosis, all of which contribute to renal function deterioration 7 . Thus, we then evaluated morphological alterations of mitochondria in cortical proximal tubules by transmission electron microscopy. Cells from tubular segments of healthy kidneys presented regular apical microvilli, intact basement membrane and basal infoldings. Mitochondria were very abundant; most of them presented an elongated shape and were localized in the basolateral part of the cells. They displayed a well-defined arrangement of well-preserved mitochondrial cisternae with a homogeneous inner matrix. In contrast, in FA-treated mice group, many epithelial cells were detached from the tubular basement membrane and showed disrupted basal infoldings. Mitochondrial structure was lost and mitochondria presented a fragmented, small and round appearance. Interestingly, most of these morphological alterations in mitochondria as well as the reduction in mitochondrial mass induced by FA were almost abrogated in renal epithelial cells overexpressing CPT1A (Fig. 3a). In the 3-days UUO and FAN models of kidney damage, CPT1A overexpression prevented the drop of mtDNA copy number (Data not shown and Fig. 3b). As previously shown in Fig.1e, the increase in CPT1A protein level in renal epithelium increased the capacity of kidney tissue to oxidize 14 C-palmitate by 1.5-fold reflected in the quantity of both 14 C-palmitate-derived 14 CO2 and ASP. As expected, defective FAO was observed in fibrotic kidneys from WT mice. However, in the 3-days UUO, FAN and ADN models, CPT1A overexpression counteracted this impairment, maintaining a FAO rate comparable to healthy kidneys ( Fig. 3c and Supplementary figures 6a and 6c). Closely related to the improvement in FAO by CPT1A, ATP content in whole kidney tissue increased from 50 to 80 µM/mg protein after CPT1A overexpression. In the 3-days UUO, FAN and ADN models, CPT1A overexpression rescued the drop in ATP levels ( Fig.3d and  Supplementary figures 6b and 6d). These results suggest that appropriate levels of CPT1A and metabolic function are necessary and sufficient to preserve adequate mitochondrial architecture and morphology.
CPT1A gain-of-function results in enhanced FAO-associated respiration of renal tubular epithelial cells even at the expense of glycolysis and AMPK activation.
TECs use glucose for anaerobic glycolysis. Metabolic alterations of these cells during kidney fibrosis not only involve a defect in FAO but also in glucose oxidation 3 . We found that in the FAN model there was a general downregulation trend in the expression of glycolysis-related genes, which was not recovered by CPT1A overexpression (Fig. 3e). By contrast, FANinduced repression in mRNA levels of the peroxisomal/mitochondrial function-related genes Acox1, Cpt2, Lrpprc, Sdha and Tfam was prevented in kidneys from CPT1A KI mice (Fig.   3f). In the UUO model we found that levels of the majority of the analyzed regulators of glucose utilization were not altered in obstructed kidneys compared to contralateral ones.
Only the increased expression of Ldh1 and Slc2a1 genes induced by UUO was prevented by CPT1A overexpression (Supplementary figure 6e). Similarly, CPT1A-gain-of-function did not induce a major shift towards the expression of peroxisome/mitochondrial-related genes in contralateral kidneys (Supplementary figure 6f).
To gain insight about quantitative metabolic changes at the cellular level we examined oxygen consumption (OCR) and extracellular acidification rates (ECAR) of primary TECs isolated from kidneys from CPT1A KI mice (Fig. 4a). Basal respiration was measured, followed by exposure to oligomycin, an inhibitor of ATP synthase, which allows determining ATP synthesis-coupled respiration and H+ leak. Then, the uncoupler carbonyl FCCP was added to measure the maximal respiratory capacity, followed by the Complex I inhibitor rotenone and complex III inhibitor antimycin A, which permits to estimate non-mitochondrial respiration (see Methods for details). We found that basal and maximum OCRs were markedly higher when palmitic acid was supplied to TECs, indicating that TECs efficiently metabolize palmitate. The increase in OCR was sensitive to the CPT1 inhibitor Etomoxir, confirming its specificity (Fig. 4b). FAO-associated OCR was also higher in primary kidney epithelial cells isolated from CPT1A KI compared to the ones isolated from WT mice (Fig.   4c). Cells treated with TGF-β1 had a lower baseline of oxygen consumption levels and showed a reduction in palmitate-induced elevation in OCR, indicating a low activity of fatty acid metabolism. CPT1A overexpression prevented the TGF-β1-induced bioenergetics derangement (Fig. 4c). Simultaneously, glycolytic function associated to palmitate consumption was determined by ECAR. Oligomycin-induced blockage of OXPHOS, allows determining the maximum glycolytic capacity. We found that CPT1A overexpression promoted the inhibition of basal glycolytic function both in the presence and absence of TGF-β1 (Fig. 4d).
To confirm the metabolic functional consequences of CPT1A overexpression in the human setting, we examined OCR and ECAR in the human tubular epithelial cell line, HKC-8.
Adenoviruses carrying CPT1A (AdCPT1A) or GFP (AdGFP) as a negative control were used to infect HKC-8 cells. CPT1A protein levels were 4-fold higher in CPT1A-expressing HKC-8 than in GFP control cells (Supplementary figure 7a, b). As expected, we found that the FAO rate was 2-fold higher in AdCPT1A-expressing HKC-8, as confirmed by measurement of 14C-palmitate-derived 14 CO2 (Supplementary figure 7c). In consistence with our observations in epithelial cells from murine kidneys overexpressing CPT1A, HKC-8 cells Depressed carbohydrate, amino acid and lipid oxidative pathways have been described in patients and animal models of CKD, leading to energy deprivation 6 . AMPK is a highly conserved sensor of the intracellular metabolic status and plays a critical role in systemic 11 energy homeostasis. The AMPK pathway is exquisitely sensitive to alterations in the [AMP]/[ATP] ratio and becomes active when this ratio is shifted towards less ATP. AMPK is activated by phosphorylation of its α-subunit residue Thr172 by upstream kinases. As a result, AMPK promotes catabolic pathways to generate more ATP and inhibits anabolic pathways. AMPK activation subsequently leads to acetyl-CoA carboxylase (ACC) inactivation (by phosphorylation of serine residues) and malonyl-CoA decarboxylase (MCD) activation, decreasing malonyl-CoA concentration and hence favoring FAO 8 . To determine if FAO gain of function in kidneys of CPT1A KI mice was associated to changes in AMPK activation as a consequence of increased ATP levels, phosphorylation of AMPK was analyzed by immunoblot. In the FAN model, obstructed kidneys presented increased AMPK phosphorylation protein levels compared with contralateral kidneys. Importantly, increasing CPT1A levels attenuated AMPK phosphorylation in fibrotic kidneys in the FAN model (Supplementary figure 8a, b). The levels of the ACC phosphorylated form changed accordingly. Similar data were obtained in the UUO and adenine models (Supplementary   figure 8c, d and 8e, f). Overall it appears that increased FAO associated to overexpression of CPT1A improves mitochondrial respiration in the context of reduced glycolysis. Moreover, it is likely that the enhancement of ATP production related to FAO reins in AMPK activation triggered by chronic kidney damage, with independence of the model employed.

TGF-beta associated epithelial cell damage.
Both studies in patients and in animal models show a strong correlation between infiltrated macrophage polarization and the extent of fibrosis 9 . In the FAN model, flow cytometry analysis revealed that CPT1A overexpression reduced the proportion of renal proinflammatory M1 subpopulation ( Fig. 5a and 5b). By contrast, the macrophage subpopulation positive for CD86 (M1) was higher than the positive for CD206 (M2) in fibrotic kidneys from the 3-and 7-days UUO models, while CPT1A overexpression did not affect this trend (Supplementary figure 9a, 9b, 9c and 9d). However, the M2 macrophage subpopulation was increased in fibrotic kidneys from CPT1A KI mice compared with the WT ones after 7 days UUO (Supplementary figure 9c and 9d). Infiltration of macrophage subpopulation positive for both CD86 and CD206 (M1/M2) observed in obstructed kidneys from WT mice 3 or 7 days after UUO was also enhanced by CPT1A overexpression (Supplementary figure 9b and 9d). In addition, the abundance of this M1 subpopulation was lower in FAN-induced fibrotic kidneys than in UUO-associated ones ( Fig. 5b and Supplementary figure 9b and 9d). These data suggest that the changes in the degree of macrophage infiltration and relative contributions of macrophage subpopulations are dependent on the model of kidney injury.
Inflammation is a major hallmark of renal fibrosis, especially in early stages 9 . Consistently, kidney infiltration by CD3+ positive T lymphocytes after FAN was reduced in CPT1A KI compared with WT mice (Supplementary figure 2c, d). Thus, we analyzed the profile of prototypical molecules related to inflammation in kidneys from mice overexpressing CPT1A in the context of renal damage. In the FAN model we found a lower expression of inflammation-related markers including the cytokines IL1b, IL6 and Tnfα (Fig. 5c), in contrast to findings in the 7 days UUO model (Supplementary figure 9e). Consistently, cells from kidneys of mice overexpressing CPT1A presented a reduced population of damaged epithelial cells (CD45-EPCAM+ CD24+) 10 compared to those of WT mice in both the FAN and 7-days UUO models (Figure 6a and 6b and Supplementary figure 10a and 10b).
Moreover, there was also a reduced expression of the apoptotic markers Apaf1, Bax and Bcl2 (Fig. 5d). A similar pattern regarding the expression of apoptotic markers was observed in the 7 days UUO model (Supplementary figure 9f).

13
Epithelial cell dedifferentiation is a process associated to the transformation of a terminal cellular phenotype into one with relatively higher potential of differentiation into more than one cell type 11 . Direct epithelial damage causes dedifferentiation and upregulation of mediators that may themselves orchestrate pro-inflammatory signaling. This dedifferentiation is characterized by the loss of epithelial markers and acquisition of mesenchymal features, a process known as epithelial-to-mesenchymal transition (EMT) 12 , the role of which in kidney fibrosis remains controversial. We studied the expression of Ecadherin, whose loss is considered as a representative feature of EMT, in the setting of CPT1A overexpression and found no significant differences in renal tubular expression in any of the three models employed (Supplementary figure 11a). A similar pattern was observed when further EMT-associated markers were analyzed in the FAN model. Only the EMT-associated increased expression of Snail was prevented in the kidneys of CPT1A KI mice (Supplementary figure 11b). We also tested whether enhancing renal epithelial FAO had a protective role in TGF-β-induced transformation of epithelial cells into a cellular fibrotic phenotype. Primary kidney epithelial cells isolated from CPT1A KI mice and treated with TGF-β1 for 48 h showed a marked reduction in the increase of these EMT-associated markers compared to the cells isolated from WT mice (Supplementary figure 11c). To explore the importance of FAO in a clinical context we analyzed a cohort of 686 patients with CKD and diabetes pertaining to the PREDIMED study 13 (see Supplementary Table 6 14 for details). In those where renal function parameters were available (n=686), we determined the levels of acylcarnitines. We found inverse correlation between GFR and short (c2-c7) and medium (c8-c14)-chain acylcarnitine levels (Fig. 7a, 7b and Supplementary Table 7 and 8) in patients with a GFR under 60 ml/min. Thus, CKD patients with a higher GFR showed less accumulation of short and medium-chain acylcarnitines (Fig. 7d, 7e, 7g) with no correlation in the case of long-chain (c16-c26) acylcarnitines (Fig. 7c, 7f). In a different cohort of CKD patients (see Supplementary Table 9 for details) we found a positive correlation between tubule CPT1A expression levels and eGFR (Fig. 7h, 7i) in a RNAseq study from different pathological backgrounds. The degree of fibrosis also correlated significantly with declining CPT1A levels (Fig. 7j). The increase in the levels of acylcarnitines and the reduction in the levels of the limiting step enzyme responsible for their metabolism (CPT1A) most likely reflect a decreased FAO capacity associated to CKD. Thus, these results reinforce the relevance of reduced FAO in CKD and support the importance of its gain-of-function to combat kidney fibrosis as demonstrated in our experimental model.

Discussion
Kidney fibrosis is critically linked to metabolic failure in tubular epithelial cells 14 . We demonstrate that genetic gain-of-function of FAO is sufficient to provide significant mitigation of fibrosis development in three different experimental models of chronic kidney damage.
Although FAO is a complex process depending on several biochemical steps and enzymatic systems, CPT1A is rate-limiting due to its key role for shuttling medium and long acyl CoA chains into the mitochondrial matrix with the concourse of L-carnitine 15 . To date, no drugs are available to specifically activate CPT1 and hence, pharmacological studies are not devoid of limitations. Homozygous CPT1A deficiency is lethal in the mouse 16 , while tissue specific CPT1A knockout models for the endothelium 17 , pancreatic α cells 18 and intestinal stem cells 19 have been generated. None of these studies addressed potential consequences 15 on organ fibrosis. The FAO gain-of-function genetic model herein described is likely one of the first direct demonstrations of the potential of FAO gain-of-function to treat kidney fibrosis.
The study was performed in heterozygous mice in an attempt to avoid non-physiological scenarios due to CPT1A overdosing 20 . Mitigation of renal fibrogenesis was reflected in a reduction of fibrotic markers, histological improvement and amelioriation of renal function.
Slight differences in the magnitude of changes are most likely due to the variation in the intensity of the inflammatory or the fibrotic components among models. While no experimental model recapitulates with close fidelity human CKD, the fact that we found significant correlation between advanced kidney disease and two signatures of FAO reduction (increased acylcarnitines and low CPT1A) in two large patient cohorts attests to the clinical relevance of a FAO-related major metabolic disturbance in human CKD 21 . The fact that we could not evaluate levels of CPT1A and acylcarnitines simultaneously in any of the two cohorts poses some limitations for a clear-cut interpretation. We found correlation between GFR and short and medium but not long-chain acylcarnitines levels, as could be initially expected. In a detailed study, Afshinnia F et al. found higher plasma abundance of long chain free FAs coupled with a significantly lower long-to-intermediate acylcarnitine ratio (a marker of impaired β-oxidation) in severe stages of CKD, but CPT1A levels were not determined 21 . Increased levels of short and middle ACs in the diabetic cohort with CKD may be explained by alterations in the expression or activity of FAO enzymes aside CPT1A 3 as short and middle ACs may derive from long chain ACs incomplete fatty acid oxidation 22 .
Further, the systemic clearance of ACs may be differentially impaired in this population, also impacting on thei 23,24 .
The high energetic demand of renal tubular epithelial cells, required for their performance of multiple functions related to the homeostasis of the internal milieu, dictates their dependence on an intact mitochondrial function. We show that FAO gain-of-function, attained by overexpressing tubular CPT1A (in both animals and cells) restores mitochondrial mass and architecture, enhances OCR and increases ATP production in conditions of renal damage. This appears to occur at the expense of glycolysis as ECAR is significantly lower under conditions of CPT1A overexpression. The reciprocal regulation between FAO and glycolysis, the Randle cycle 25 , has been demonstrated in muscle and heart, our data now supporting that it also takes place in renal tubular cells. Alternatively, it is possible that TECs are consuming lactate as the main fuel for gluconeogenesis, a well-known process occurring in the kidney 26 . The fundamental question as to whether metabolism dictates phenotype is in part answered here, as our data support that an increased capacity of the tubular epithelial cell to meet its energy challenges is able to significantly deter the course of kidney fibrosis.
Thus, efforts directed at enhancing FAO in the early stages of fibrosis should very likely pay off to combat human CKD.

Isolation of primary kidney epithelial cells. Kidneys from CPT1A KI and wild-type (WT) mice
(3-to 5-week-old males) were collected after sacrifice by cervical dislocation and perfusion with PBS. The capsule was removed and the cortex from 2 kidneys was dissected, placed in 1 mL ice-cold phosphate buffered saline (PBS) (Corning, New York, NY), and minced into pieces of approximately 1 mm 3 . These pieces were digested with 10 ml HBSS containing 2 mg/mL collagenase I (Thermo Scientific, Rockford, IL) for 30 minutes at 37°C with gentle stirring and supernatants were sieved through a 100-μm nylon mesh. After centrifugation for 10 minutes at 3000 rpm, the pellet was resuspended in sterile red blood cell lysis buffer  Unilateral ureteral obstruction (UUO). UUO surgery procedure was performed as previously described 33 . Briefly, mice were anesthetized with isofluorane (3-5% for induction and 1-3% for maintenance) and divided into two experimental groups: the UUO group and the sham operation group. In the UUO group, mice were shaved on the left side of the abdomen, a vertical incision was made through the skin with a scalpel and the skin was retracted. A second incision was made through the peritoneum to expose the kidney. The left ureter was ligated twice 15 mm below the renal pelvis with surgical silk and the ureter was then severed between the two ligatures. Then, the ligated kidney was placed gently back into its correct anatomical position and sterile saline was added to replenish loss of fluid. The incisions were sutured and mice were individually caged. The sham operation was performed in a similar manner, but without ureteral ligation. Buprenorphine was used as an analgesic. A first dose was administered 30 minutes before surgery and then every 12 hours for 72 hours, at a dose of 0.05 mg / kg subcutaneously. In this model, renal blood flow and glomerular filtration rate become significantly reduced within 24 h and interstitial inflammation (peak at 2-3 days), tubular dilation, tubular atrophy and fibrosis are evident after 7 days. The obstructed kidney reaches maximal dysfunction around 2 weeks after the procedure. Mice were sacrificed by CO2 overdose and control and obstructed kidney and blood samples were harvested after perfusion with PBS at 3 and 7 days after UUO.  were seeded in 12-well dishes to reach a confluence of 70% and were infected with CPT1A adenoviruses as described in the adenovirus-mediated CPT1A overexpression procedure section. Then, cells were incubated in 500 μL of media containing 0.3% BSA/100 μM palmitate/0.4 μCi/mL 14 C-palmitate at 37 °C for 3 h. Each sample was assayed in triplicate.
The reaction was stopped by the addition of 200 µl of 1 M perchloric acid. The rate of palmitate oxidation was measured as released 14 CO2 trapped in a filter paper disk with 20 μL of 1 M NaOH in the top of sealed vials. 14 CO2 traps were transferred to scintillation vials to determine 14 C-palmitate-derived 14 CO2. The remaining acid solution for each sample was centrifuged at 14,000 g for 10 min at 4 °C and 400 μL of supernatant was also added to scintillation vials for the quantification of 14 C-palmitate-derived acid-soluble metabolites.
Control incubations without cells or tissue samples were also run in parallel. 14 C products were counted in an LS6500 liquid scintillation counter (Beckman Instruments Inc., Brea, CA).
Scintillation values were converted to mmol 14 CO2 or acid-soluble metabolites multiplying the specific activity and normalized to the protein content. Measurements of oxygen consumption rate. Fatty acid oxidation-associated oxygen consumption rate (OCR) (ligated to oxidative phosphorylation) and extracellular acidification rate (ECAR) (associated with lactate production and glycolysis) were studied using the Seahorse Bioscience metabolic analyzer according to the manufacturer's instructions 39 .
HKC-8 cells were seeded in a p60 plate and microRNA transfection or adenovirus-mediated CPT1A overexpression was performed (as shown in the transfection procedure and adenovirus-mediated overexpression sections, respectively) when they reached a confluence of 70% and 48 hours later, HKC-8 cells were treated with 10 ng/ml TGF-β1 for 48 h. In the case of primary kidney epithelial cells, they were seeded in a p60 plate and when they reached a confluence of 70%, they were treated with 10 ng/ml TGF-β1 for 48 h.
In all cases, cells were then seeded at 2 × 10 4  Measurements of oxygen consumption rate. Fatty acid oxidation-associated oxygen consumption rate (OCR) (ligated to oxidative phosphorylation) and extracellular acidification rate (ECAR) (associated with lactate production and glycolysis) were studied using the Seahorse Bioscience metabolic analyzer according to the manufacturer's instructions 39 .
HKC-8 cells were seeded in a p60 plate and microRNA transfection or adenovirus-mediated CPT1A overexpression was performed (as shown in the transfection procedure and adenovirus-mediated overexpression sections, respectively) when they reached a confluence of 70% and 48 hours later, HKC-8 cells were treated with 10 ng/ml TGF-β1 for 48 h. In the case of primary kidney epithelial cells, they were seeded in a p60 plate and when they reached a confluence of 70%, they were treated with 10 ng/ml TGF-β1 for 48 h. Clinical data. Acylcarnitines were evaluated based on the PREDIMED trial. The PREDIMED is a primary prevention multicenter trial conducted in Spain. A detailed description of this trial is described elsewhere 13 . Liquid chromatography-tandem mass spectrometry was used to semi-quantitatively profile acylcarnitines in plasma samples. Our analysis is based on a subsample of 686 participants with metabolomics data and measures of kidney function including glomerular filtration rate and urinary albumin-creatinine ratio 45 . Studies of CPT1A expression in kidney tubules were performed in a cohort of 433 patients (described in 46 and hereon named CKD cohort), whose demographic and clinical features are summarized in

Supplementary table 9.
Statistical analysis. Data in experimental models were analyzed using nonparametric tests except where indicated. The difference between two independent groups was examined with Mann-Whitney test, while more than two groups were compared with Kruskal-Wallis test. A P-value of 0.05 or less was considered statistically significant ( */# : P < 0.05, * */## : P < 0.01, ***/### : P < 0.001). Data were analyzed using GraphPad Prism 6.0 (GraphPad Software, La Jolla, CA). Data are reported as mean ± standard error of mean (SEM).
Data in clinical studies for acylcarnitines: Means and standard deviations (SD) were used to describe continuous variables and percentage to describe categorical values. Metabolite measures were normalized using a logarithmic transformation and scaled to multiples of SD using z-score standardization. Three scores were created according to the number of carbons: short (2-7), medium (8)(9)(10)(11)(12)(13)(14) and long (16-26) acylcarnitine scores. The sum of the log-transformed values of each carnitine was computed to calculate these scores and a zscore standardization was also applied. Chi-squared and Student´s t-test were used to compare categorical and quantitative variables, respectively. Univariate and multivariable linear regression models were conducted to assess the association between the glomerular filtration rate (GFR) and both the acylcarnitine scores and individual metabolites. The covariates used in the multivariable models were age, sex, type-2 diabetes and albumin/creatinine ratio.
Data in clinical studies for CPT1A expression: ANOVA test was performed to assess the significance across different disease groups and cor.test() function in R was used to get the Pearson's correlation and the corresponding P-values. The CBMSO receives institutional support from Fundación "Ramón Areces". The     Table 1 Radiolabeled palmitate-derived CO2 and acid-soluble products (ASP) were determined after 41 incubation of 14 C-palmitate with kidney tissue from WT or Pax8-CPT1A mice after doxycycline treatment. Bar graphs represent the mean ± s.e.m (n = 4 mice per group). *P < 0.05 compared to kidneys from WT mice. Statistical significance between two independent groups was determined using non-parametric two-tailed Mann-Whitney test, while more than two groups were compared with Kruskal-Wallis test. WT and Pax8-CPT1A mice after doxycycline induction. Bar graphs represent the mean ± s.e.m. of fold changes (n = 6 mice per group). *P < 0.05, **P < 0.01 compared to their corresponding control (CT) kidneys; ## P < 0.01 compared to kidneys from WT mice with the same experimental condition. Statistical significance between two independent groups was determined using non-parametric two-tailed Mann-Whitney test, while more than two groups were compared with Kruskal-Wallis test. For gene nomenclature see Supplemental Table   4. represent the mean ± s.e.m. of fold changes (n = 6 mice per group). *P < 0.05, **P < 0.01 compared to their corresponding control (CT) kidneys; # P < 0.05, ## P < 0.01, compared to kidneys from WT mice with the same experimental condition. Statistical significance between two independent groups was determined using non-parametric two-tailed Kruskal-Wallis test. For detailed gene nomenclature see Supplemental Table 4.  (C and D) *P < 0.05, **P < 0.01compared to their corresponding control (CT) kidneys; # P < 0.05, ## P < 0.01 compared to kidneys from WT mice with the same experimental condition.
Statistical significance between two independent groups was determined using non-45 parametric two-tailed Mann-Whitney test, while more than two groups were compared with Kruskal-Wallis test. For detailed gene nomenclature see Supplemental Table 4.  of fold changes corresponding to densitometric analyses (n = 6 mice per group).
Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was used for normalization purposes. **P < 0.01 compared to their corresponding contralateral (CT) kidneys; # P < 0.05, ## P < 0.01 compared to kidneys from WT mice with the same experimental condition. (F) mRNA levels of fibrosis-associated genes were determined by qRT-PCR using TaqMan qPCR probes in contralateral (CT) and obstructed (UUO) kidneys from WT and Pax8-CPT1A mice subjected to UUO for 7 days after doxycycline induction. Bar graphs represent the mean ± s.e.m. of fold changes (n = 9 mice per group). ***P < 0.01 compared to their corresponding contralateral (CT) kidneys; # P < 0.05 compared to kidneys from WT mice with the same experimental condition. Statistical significance between two independent groups was determined using non-parametric two-tailed Mann-Whitney test, while more than two groups were compared with Kruskal-Wallis test. For detailed gene nomenclature see Supplemental Table 4. inflammation-associated (E) and apoptosis-associated (F) genes were determined by qRT-PCR using TaqMan qPCR probes in contralateral (CT) and obstructed (UUO) from kidneys of WT and Pax8-CPT1A mice subjected to UUO for 7 days after doxycycline induction. Bar graphs represent the mean ± s.e.m. of fold changes (n = 9 mice per group). **P < 0.01, ***P < 0.001 compared to their corresponding contralateral (CT) kidneys; # P < 0.05 compared to kidneys from WT mice with the same experimental condition. Statistical significance between two independent groups was determined using non-parametric two-tailed Mann-Whitney test, while more than two groups were compared with Kruskal-Wallis test. For detailed gene nomenclature see Supplemental Table 4. AdGFP with the same experimental treatment. Statistical significance between two independent groups was determined using non-parametric two-tailed Kruskal-Wallis test.