Lipidomic Profiling of Kidney Cortical Tubule Segments Identifies Lipotypes with Physiological Implications

Abstract

A detailed knowledge of the lipid composition of components of nephrons is crucial for understanding physiological processes and the development of kidney diseases. However, the lipidomic composition of kidney tubular segments is unknown. We manually isolated the proximal convoluted tubule (PCT), the cortical thick ascending limb of Henle’s loop, and the cortical collecting duct from 5 lean and obese mice and subjected the samples to shotgun lipidomics analysis by high-resolution mass spectrometry acquisition. Across all samples, more than 500 lipid species were identified, quantified, and compared. We observed significant compositional differences among the 3 tubular segments, which serve as true signatures. These intrinsic lipidomic features are associated with a distinct proteomic program that regulates highly specific physiological functions. The distinctive lipidomic features of each of the 3 segments are mostly based on the relative composition of neutral lipids, long-chain polyunsaturated fatty acids, sphingolipids, and ether phospholipids. These features support the hypothesis of a lipotype assigned to specific tubular segments. Obesity profoundly impacts the lipotype of PCT. In conclusion, we present a comprehensive lipidomic analysis of 3 cortical segments of mouse kidney tubules. This valuable resource provides unparalleled detail that enhances our understanding of tubular physiology and the potential impact of pathological conditions.

Graphical Abstract

Graphical Abstract.

Introduction

Lipids are vital components of the cell membrane and play crucial roles as well as in energy production and cellular signaling. Fatty acids (FAs) are the simplest lipids, which are carboxylic acids comprising an aliphatic (acyl) chain, that can comprise up to 24 carbon atoms in most cases. Molecular species with 16 and 18 carbons are the most prevalent. Most FAs require activation by acyl-CoA synthases to form fatty acyl-CoA esters for further metabolic pathways such as fatty acid oxidation (FAO), glycerophospholipids (GP), also known as phospholipids, production, or neutral lipid storage. FAs are esterified with glycerol to form triacylglycerides (TG), a neutral lipid that accumulates in cells as FA reserves in anhydrous lipid droplets (LD). FA can be released from LD in order to produce metabolic energy through FAO. Lipid droplets can also function as storage sites for FAs and sterol components required for the formation of cellular membranes.

The primary components of biological membranes are GP, which consist of a glycerol backbone esterified with an FA at sn-1 and sn-2 positions, and a phosphate derivative called a “head” at sn-3 position. The phosphate of this head can bind with choline to form phosphatidylcholine (PC), ethanolamine (PE), serine (PS), inositol (PI), glycerol (PG), or no molecule (phosphatidic acid, PA). Phospholipases can hydrolyze GP to produce FA, which have their own biological functions in the cell.

The second family of membrane components is the sphingolipids (SL), which are structurally based on sphingosine, a derivative of serine and palmitate. Most sphingosines are condensed with an amide-linked FA to form ceramide (Cer), a bioactive lipid that promotes cell death. When a phosphocholine is attached to the terminal hydroxyl group of the sphingosine, it is called sphingomyelin (SM). When instead a monosaccharide is attached, it is called a cerebroside. If additional neutral monosaccharides are added, a globoside (Gb) is formed, while if neuraminic acids are present, we end up with gangliosides (GM).

The membranes contain cholesterol, a 27-carbon, 4-ring molecule that enhances their fluidity. Cholesterol also acts as a precursor for several hormones and vitamins. When an FA is esterified to cholesterol, it forms a neutral lipid called cholesteryl ester (CE), which also works as a storage lipid.

The various headgroups and aliphatic chain compositions (number of carbons and degree of unsaturation, i.e., number of double carbon bonds) enable the presence of over 1000 distinct molecular species in every eukaryotic cell. In addition, there is considerable chemical and compositional variation in membrane lipids, which is especially apparent in organelles, plasma membranes of polarized cells, plasma membrane nanodomains, or membrane leaflets. This diversity influences various properties relating to lipid–lipid and lipid–protein interactions and membrane properties, such as fluidity and curvature.

The tubular system in the kidney comprises discrete segments with unique cellular and functional phenotypes that support exclusive physiological functions. There is established evidence at the molecular level that these segments possess their own distinct transcriptomic and proteomic identities.¹,² From a lipidomic standpoint, no study to date has provided a comprehensive characterization of the lipid composition of tubule segments in normal and diseased kidneys.

Lipidomic analysis in chronic kidney disease (CKD) is an emerging area of research that has already yielded valuable insights. Over the past decade, advances in analytical mass spectrometry have enabled more thorough analysis of the lipidome in plasma and kidney tissue. Studies have shown that changes in lipid metabolism associated with the progression of kidney disease involve variations in the elongation, saturation, synthesis, and lipolysis of a large number of intraclass lipids. Studies to date have shown that FAO and lipogenesis, previously thought to be independent processes occurring in separate cellular compartments, are interrelated. The available data on lipid biology in the kidney are heavily focused on energy metabolism in the proximal tubule, likely due to the significance of this segment in terms of cell mass and FA consumption.³ This should not preclude the existence of other biological processes involving lipid metabolism that are important in kidney physiology and can be uncovered through suitable techniques. Additionally, medical conditions leading to CKD, such as obesity and diabetes, have a substantial impact on energy and lipid metabolism, predominantly influencing the phenotype and tubular functions.^4–9 However, the lipidome from mouse kidney tubule segments has not been previously described, and the rapid development of a variety of analytical tools for lipid analysis based on mass spectrometry has made it possible to assess the lipid composition of membranes with great precision and depth.

Here, we present a comprehensive investigation of the lipid class and molecular species composition in 3 isolated tubular segments: proximal convoluted tubule (PCT), cortical ascending thick limb of Henle (CTAL), and cortical collecting duct (CCD) of both normal and obese mice. We present findings with key implications for comprehending kidney pathophysiology.

Materials and Methods

Animal Experiments

Experiments were performed on 12-wk-old-male C57BL/6J wild-type or obese (B6.Cg-Lep^ob/J) mice obtained from the Janvier laboratory and maintained at the CEF (Centre d'Explorations Fonctionnelles of the Cordeliers Research Center, Agreement No. A75-06-12) until euthanasia for the isolation of kidney segments.

Tubule Segments Isolation

Isolation of PCT, CTAL, and CCD was performed according to localization in the kidney (cortex vs. medulla) and well-defined morphologic characteristics under binocular loupes after kidney treatment with Liberase (Sigma-Aldrich, St. Quentin Fallavier, France).¹⁰ Tubule length was measured using visilog software (Noesis, Courtabeuf, France), and pools of 150 mm (corresponding approximately to 150 segments) were transferred to 500 µL of PBS, rinsed 3 times, centrifuged (600 g, 5 min), and resuspended in 150 m m ammonium bicarbonate. Samples were frozen in liquid nitrogen and stored at −80°C until lipid extraction.

Polymerase Chain Reaction

RNA was extracted from a small portion of the segments using the RNeasy micro kit (Qiagen, Hilden, Germany). mRNA was then reverse transcribed into cDNA (Roche Diagnostics, France) according to the manufacturer’s instructions, and real-time PCR was performed on a LightCycler (Roche Diagnostics, France). No signal was detected in samples that did not undergo reverse transcription or in blank runs without cDNA. In each run, a standard curve was generated by serial dilution of the stock cDNA. The expression of the housekeeping gene Rpl26 was used to normalize the results. To verify the quality of our isolation process, we measured the expression of specific markers for each isolated segment, namely Cldn2 for PCT, Slc12a1 (=Nkcc2) for CTAL, and Aqp2 for CCD, using the following primers: Rpl26 (NM_009080) F_GCTAAT GGCACAA CCGTC and R_TCTCGA TCGTTTC TTCCTTGTAT; Cldn2 NM_016675 (F_CAGTATG TCCAGACTG CATTG and R_AGATGG CCTGAGA AGGG), Slc12a1 (NM_009194) F_GAGAT TGGCGTGGT CATAGTCAGAA and R_TGCTGCT GATGTTGCC GTCTTT; Aqp2 (NM_009699) F_GAGCG GGCTGGAT TCATGGAG and R_CCTGTGA CTGTGGCG TGCCTG). Primers for Acsl4 were F_CTT CCT CTT AAG GCC GGG AC and R_TCT CTT TGC CAT AGC GTT TTT AGA.

Shotgun Lipidomics

For lipidomics analysis, 150 mm of isolated tubules were spiked with 1.40 μL of internal standard lipid mixture containing 500 pmol of Chol-d6, 100 pmol of Chol-16:0-d7, 100 pmol of DG 17:0-17:0, 50 pmol of TG 17:0-17:0-17:0, 100 pmol of SM 18:1;2-12:0, 30 pmol of Cer 18:1;2-12:0, 30 pmol of GalCer 18:1;2-12:0, 50 pmol of LacCer 18:1;2-12:0, 300 pmol of PC 17:0-17:0, 50 pmol of PE 17:0-17:0, 50 pmol of PI 16:0-16:0, 50 pmol of PS 17:0-17:0, 30 pmol of PG 17:0-17:0, 30 pmol of PA 17:0-17:0, 40 pmol of Gb3 18:1;2-17:0, 25 pmol of GM3 18:1;2-18:0-d5, 25 pmol of GM2 18:1;2-18:0-d9, 25 pmol of GM1 18:1;2-18:0-d5, 30 pmol of LPA 17:0, 30 pmol of LPC 12:0, 30 pmol of LPE 17:1, and 30 pmol of LPS 17:1 and subjected to lipid extraction at 4°C, as described elsewhere.¹¹ Briefly, the sample was dissolved in 200 μL of 155 m m ammonium bicarbonate and then extracted with 1 mL of chloroform-methanol (10:1) for 2 h. The lower organic phase was collected, and the aqueous phase was re-extracted with 1 mL of chloroform-methanol (2:1) for 1 h. The lower organic phase was collected and evaporated in a SpeedVac vacuum concentrator. Lipid extracts were dissolved in 100 μL of infusion mixture consisting of 7.5 m m ammonium acetate dissolved in propanol: chloroform: methanol [4:1:2 (vol/vol)]. Samples were analyzed by direct infusion in a QExactive mass spectrometer (Thermo Fisher Scientific) equipped with a TriVersa NanoMate ion source (Advion Biosciences). A volume of 5 µL of sample were infused with gas pressure and voltage set to 1.25 psi and 0.95 kV, respectively.

Diacylglycerols (DG), TG, and ce were detected in the 10:1 extract by positive ion mode Fourier Transform Ion Cyclotron Resonance Mass Spectrometer (FTMS) as ammonium aducts by scanning m/z = 580–1000 Da, at R_m/z=200 = 280 000 with lock mass activated at a common background (m/z = 680.48022) for 30 s. Every scan is the average of 2 micro-scans, automatic gain control (AGC) was set to 1E6, and maximum ion injection time (IT) was set to 50 ms. For FA profiling of DGs and TGs, a parallel reaction monitoring was performed with an inclusion list of m/z = 580–1000 Da, at NCE of 20 and R_m/z=200 = 17 500 for 30 s. Every scan is the average of 2 micro-scans, AGC was set to 1E6, and maximum ion IT was set to 64 ms.

Phosphatidylcholine, PC-O, Cer, GlcCer, Lysophosphatidyl choline (LPC), and ether LPC (LPC-O) were detected as acetate adducts, while PG, PE, PE-O, lysophosphatidylethnolamine (LPE), and ether LPE (LPE-O) were detected as deprotonated adducts in the 10:1 extract, by negative ion mode FTMS, after polarity switch by scanning m/z = 420–1050 Da, at R_m/z=200 = 280 000 with lock mass activated at a common background (m/z = 529.46262) for 30 s. Every scan is the average of 2 micro-scans, AGC was set to 1E6 and maximum ion IT was set to 50 ms. For FA profiling of PC, PC O, PE, PE O, and PG, a parallel reaction monitoring was performed with an inclusion list of m/z = 590–940 Da at NCE of 35 and R _m/z=200 = 17 500 for 72 s. Every scan is the average of 2 micro-scans, AGC was set to 1E5 and maximum ion IT was set to 64 ms.

LacCer and Gb3 globoside (Gb3) were detected as protonated ions, and Gb4 was detected as ammoniated adduct in the 2:1 extract in positive ion mode FTMS by scanning m/z = 800–1600 Da at R _m/z=200 = 280 000 with lock mass activated at a common background (m/z = 1194.8179) for 30 s. GM1, GM2, and GM3 gangliosides were detected as deprotonated ions in the 2:1 extract in negative ion mode after polarity switch in FTMS by scanning m/z = 1100–1650 Da, at R_m/z=200 = 280 000 with lock mass activated at a common background (m/z = 1175.7768) for 30 s. Every scan is the average of 2 micro-scans, AGC was set to 1E6 and IT was set to 50 ms in both polarities.

Phosphatidic acid, PI, PS, LPA, and LPS were detected as deprotonated ions in the 2:1 extract in negative ion mode in FTMS by scanning m/z = 400–1100 Da at R_m/z=200 = 280 000 with lock mass activated at a common background (m/z = 529.4626) for 30 s. Every scan is the average of 2 micro-scans, AGC was set to 1E6, and IT was set to 50 ms. For FA profiling of PA, PI, and PS, a parallel reaction monitoring was performed with an inclusion list of m/z = 590–940 Da at NCE of 35 and R_m/z=200 = 17 500 for 84 s. Every scan is the average of 2 micro-scans, AGC was set to 1E5, and maximum ion IT was set to 64 ms.

The analytical process for lipidomics according to lipidomicstandards (https://lipidomicstandards.org/reporting_checklist/) is reported in the file report-lipidomics tubules as a Supplementary Data.

All data were acquired in centroid mode. All lipidomics data were analyzed with the lipid identification software, LipidXplorer.¹² Tolerance for MS and identification was set to 2  ppm. Data post-processing and normalization to internal standards were done manually in excel. The lipid contents are normalized to the total membrane lipid identified (excluding ce and TG) and expressed as a proportion of the whole lipid content in a sample. Data analysis was performed in the MetaboAnalyst 5.0 software.¹³ Each metabolite level was adjusted by autoscaling (mean-centered and divided by SD of each variable).

Immunohistochemistry

Normal mouse kidney slices were fixed in alcohol–formalin–acetic acid, dehydrated in ethanol and xylene, embedded in paraffin, and cut into 3 mm sections. Samples were then deparaffinized, rehydrated, and heated at 97°C for 20 min in citrate buffer. Endogenous peroxidase was inactivated by incubation in 0.3% H₂O₂ for 10 min at room temperature. Sections were incubated with PBS containing anti-ACSL4 (Sigma HPA005552). This antibody has been validated in the Human Protein Atlas Subsequently, the sections were incubated with anti-rabbit or anti-goat antibodies conjugated to peroxydase-labeled polymer and visualized with a peroxydase kit (Dako EnVision^®+ Dual Link System-HRP (DAB+), Agilent). Finally, tissue sections were counterstained with hematoxylin. The samples were visualized using a Nikon Eclipse Ti microscope.

Statistical Analysis

Graphs and statistical analyses were generated using GraphPad Prism 9 software (GraphPad Software, Inc.). Data are presented as mean ± standard error of the mean. Unpaired 2-sample t-tests were used to determine a significant difference between 2 groups and were 2-tailed. Multiple comparison within the same group were performed with 1-way followed by Tukey’s multiple comparison test. Multiple t-tests with a false discovery rate (FDR) control by the 2-stage step-up method of Benjamini, Krieger, and Yekutieli were performed for multiple comparisons between 2 groups. Two-way ANOVA with Šídák’s multiple comparison test was performed when more than 2 groups were compared. A P-value < .05 was considered a statistically significant difference.

Study Approvals

Animal experiments were conducted according to French veterinary guidelines and those formulated by the European Commission for experimental animal use (L358–86/609EEC). The animals were kept at CEF (Centre d'Explorations Fonctionnelles) of the Cordeliers Research Center, Agreement No. A75-06-12).

Results

The Three Tubular Segments of Lean Mice Have a Distinct Lipidomic Profile

Three tubular segments were isolated from 5 mouse kidneys using the liberase and protease perfusion and digestion protocol.¹⁰ Phase-contrast examination confirmed the origin of the 3 segments based on their morphologic characteristics: The PCT is large and convoluted, CTAL segments are thin, bright, and straight, while CCD appears dense with a cobblestone appearance¹⁴,¹⁵ (Figure 1A). Additionally, these 3 segments expressed specific differentiation markers. Cldn2 (encoding Claudin-2) and not Slc12a1 (encoding Nkcc2) or Aqp2 (encoding Aquaporin-2) were expressed in PCT, Slc12a1 and not Cldn2 or Aqp2 were expressed in CTAL, while Aqp2 and not Cldn2 or Slc12a1 were expressed in CCD, as illustrated in Figure 1B. The total membrane lipid contents extracted from tubule segments were 14.54 ± 3.83 nmol for PCT segments, 4.05 ± 0.84 nmol for CTAL segments, and 3.15 ± 0.29 nmol for CCD segments. The lipid class and species contents were therefore normalized to the total lipid identified and expressed as a proportion of the whole lipid content in a sample.

Figure 1.

The 3 tubule segments possess a distinct lipidomic profile. (A) Phase contrast micrographs of isolated tubule segments at original magnification of × 40. (B) Relative expression levels of Cldn2, Slc12a1, and Aqp2 as markers of tubule segment origin using RT-qPCR. Rpl26 was used as a housekeeping gene, and 5 samples were taken per segment. P-values were calculated using 1-way ANOVA followed by Tukey’s multiple comparison test. (C) Histograms of the proportional distribution of major lipid classes in the cellular membrane composition of proximal convoluted tubule (PCT), cortical thick ascending limb of Henle (cTAL), and cortical collecting duct (CCD) isolated from 5 mouse kidneys. The percentage of total membrane lipids in a sample, expressed in mol%, is shown on the y-axis. Chol: cholesterol, PC: phosphatidylcholine, PE: phosphatidylethanolamine, PI: phosphatidylinositol, PS: phosphatidylserine, SM: sphingomyelin, PG: Phosphatidylglycerol, PA: Phosphatidic acids (D). Dimension reduction by principal component analysis on all identified lipid molecular species in the 3 segments from 5 mice. The resulting plot shows the 2D scores between the selected principal components that best explain the lipid variance. (E) Hierarchical clustering of the 100 most differentially expressed lipid molecular species among the 3 tubular segments (1-way ANOVA with 5 replicates). Each row represents the relative lipid molecular species composition, while each column represents either a replicate or a condition. Samples were normalized to sum, and each metabolite was autoscaled (mean-centered and divided by SD of each variable).

The distribution of the principal membrane lipid classes across the 3 tubule segments demonstrated an overall pattern consistent with lipid membranes composition of mammal cells.¹⁶ Specifically, Chol, PC, PE, and PI were found to be prevalent, while PS was less common (Figure 1C). The high concentration of SM in all 3 segments indicates a predominance of plasma membranes in the whole cell lysate preparation.¹⁶

Principal component analysis of all the lipid molecular species identified in the samples showed that each of them segregated into 3 distinct groups according to their segments of origin and that the replicates were highly consistent (Figure 1D). Hierarchical clustering of the 100 molecular species with the highest distribution variance among the 3 segments revealed the diverse composition of the segments (Figure 1E). These results indicate that each tubular segment analyzed has a distinct lipidomic signature that may contribute to its individual identity.

The CCD of Lean Mice Is Predominantly Enriched in Neutral Lipids

We performed a comparative analysis of the neutral lipid content in the 3 segments and observed a significant enrichment of TG and ce in CCD compared to PCT, with an intermediate proportion for CTAL (Figure 2A and B). Interestingly, the precursors of TG, namely PA and DG, showed a comparable incremental pattern (Figure 2C and D). In addition, based on a rat tubular proteome database (dataset identifier PXD016958), we found that perilipin-3, an envelope protein that is specific to LD, was predominantly expressed in CCD² (Figure 2E). This suggests that CCD and CTAL are more enriched in LD compared to PCT, which may reflect differences in FAO activity. In addition, the accumulation of TG precursors, PA and DG may occur as a result of increased metabolic flux generating TG in LD.¹⁷ Interestingly, the tubular proteomics database revealed increased expression of enzymes responsible for TG synthesis, including GPAT1 (which metabolizes glycerol-3-phosphate and 2 FA to LPA), AGPAT1 (which metabolizes LPA to PA), LIPIN1 (which metabolizes PA to DG), and DGAT1 (which metabolizes DG to TG), in the CCD compared to the PCT, with intermediate levels in the CTAL (Figure 2F).

Figure 2.

The cortical collecting duct (CCD) is predominantly enriched with neutral lipids. (A–D) Histograms of the distribution of different lipid classes, including triacylglycerides (TAG), cholesteryl ester (CE), phosphatidic acid (PA), and diacylglycerol (DAG), in the proximal convoluted tubule (PCT), cortical thick ascending limb of Henle (cTAL), and cortical collecting duct (CCD) of 5 mouse kidneys. The y-axis indicates the relative expression of each lipid class (mol%), which is the percentage of total membrane lipids in the sample. P-values were calculated using 1-way ANOVA followed by Tukey’s multiple comparison test. (E) Histograms of perilipin-3 expression in microdissected rat tubules. Data were obtained from public proteomic repositories (NCBI accession PDX016958 and publicly available data from the Kidney Tubule Expression Atlas website at https://esbl.nhlbi.nih.gov/KTEA/). (F) Schematic representation of the pathway for TAG synthesis, accompanied by histograms of the expression of TAG synthesis enzymes in microdissected rat tubules. Data were obtained from public proteomic repositories (NCBI accession PDX016958 and publicly available data from the Kidney Tubule Expression Atlas website at https://esbl.nhlbi.nih.gov/KTEA/). (G) Heat map of the fatty acids (FAs) profile of triacylglycerides (TG) molecular species relative expression in the 3 tubule segments (with n = 5 replicates per segment). FA molecular species are classified according to their chain size and degree of unsaturation. Each line corresponds to the proportional composition of a lipid molecular species after autoscaling (ie, normalizing the data by centering the mean and dividing by the SD of each variable). (H) Heat map of all identified ce molecular species in the 3 segments of the tubules (with n = 5 replicates per segment), classified by chain size and number of unsaturations. Each line corresponds to the proportional composition of a lipid molecular species after autoscaling (ie, normalizing the data by centering the mean and dividing by the SD of each variable).

We examined the composition of TG in detail with respect to the length of the FA chains and their degree of unsaturation. PCT was enriched in TG containing long chain polyunsaturated FA (PUFA), whereas CCD and the CTAL were enriched in TG containing FA with shorter and more saturated chains (Figure 2G). The simplest explanation for this observation is that FAO preferentially supports saturated FAs over unsaturated chains (FAO of unsaturated FA requires additional enzymes), which could lead to a relative accumulation of the latter in PCT TAGs. These differences do not appear in the composition of the ce (Figure 2H).

In conclusion, our comparative lipidomic analysis suggests that CCD has a higher concentration of neutral lipids compared to PCT, which may indicate differences in biosynthetic activities, and PCT is enriched in long chain PUFA-containing TG. Indeed, LDs should form under conditions where the kidney uptake of FA exceeds the FAO.⁵ Interestingly, while FAO primarily produces ATP in the proximal tubule, the distal segments of the tubule may exhibit higher glycolytic activity to generate energy.⁷,¹⁴,¹⁵,¹⁸

The Cortical Thick Ascending Limb of Lean Mice is Enriched in Arachidonic Acid-Containing Phospholipids

Examination of the distribution of molecular species of PC and PE provided important elements for kidney physiology. Notably, the membranes of CTAL cells showed higher levels of PE and PC containing FA 20:4 (arachidonic acid, AA) compared to the other segments (Figure 3A and Figure S1). Membrane GP are sources of FA following the action of phospholipase A2. Our observation suggests that PE and PC present in CTAL cells may serve as a source of mobilizable AA that is further metabolized to prostaglandins, among others, by cyclooxygenase. The importance of this AA reservoir function becomes clear when considering the role of CTAL and AA metabolites in the regulation of juxtaglomerular cell activity.¹⁹

Figure 3.

The cortical thick ascending limb is enriched with fatty acids (FAs) that contain arachidonic acid. (A) Hierarchical clustering of all identified phosphatidylcholine (PC) molecular species in the 3 tubular segments (n = 5 replicates). Each row in the resulting graph shows the relative composition of lipid molecular species, while each column indicates a specific condition or replicate. Samples were normalized by sum, and data were autoscaled (mean-centered and divided by SD of each variable) to ensure accurate analysis. (B) Histograms of the expression of Acsl4 in microdissected rat tubules. Data were obtained from public proteomic repositories, including NCBI accession PDX016958 and publicly available data on the Kidney Tubule Expression Atlas website (https://esbl.nhlbi.nih.gov/KTEA/). (C) Representative photomicrograph of immunohistochemical staining for Acsl4 in a mouse kidney with no histological lesion. Antibody is HPA005552 (Sigma-Aldrich). Stars denote the macula densa. Original magnification x10. (D) Hierarchical clustering of all identified molecular species of ether PC (PC O) and ether PE (PE O) in the 3 tubule segments (n = 5 replicates). Each row shows the proportional composition of each lipid molecular species, while each column represents either 1 condition or 1 replicate. Samples were normalized by sum, and data were autoscaled. (E) Histograms of ether PC distribution among proximal convoluted tubule (PCT), cortical thick ascending limb of Henle (cTAL), and cortical collecting duct (CCD) from 5 mouse kidneys. The y-axis (mol%) indicates the percentage expression of a lipid category, i.e., the proportion of total membrane lipids in a sample. P-values were determined by 1-way ANOVA followed by Tukey’s multiple comparison test.

Long-chain acyl CoA synthase (ACSL) family members consist of 5 distinct isoforms that facilitate the formation of acyl-CoA by catalyzing FA with chain lengths ranging from 12 to 22 carbon atoms.²⁰ ACSL4 catalyzes the transformation of long chain PUFA into fatty acyl-CoA products, with a preference for AA.²¹ According to the tubular proteome database cited above, ACSL4 expression was high in CTAL, almost absent in PCT, and moderate in CCD, consistent with the increased presence of AA within membrane PE and PC in CTAL (Figure 3B). Moreover, we performed an immunohistochemistry study showing that Acls4 is predominantly expressed in the macula densa, a segment cTAL in the juxtaglomerular apparatus, suggesting that ACSL4 may have a function in the global function of the juxtaglomerular apparatus, by enriching membranes in AA (Figure 3C). This suggests that there is a metabolic bias in this part of the tubule that ensures precise physiological function.²²

The PCT of Lean Mice Contains a High Concentration of Docosahexaenoic Acid and Ether Phospholipids

When examining the FA composition of PE and PC, we found a preferential enrichment of PCT in FA 22:6 (docosahexanoic acid, DHA), an omega-3 PUFA (Figure 3A and Figure S1), suggesting that PCT is a reservoir for DHA. In addition, ether phospholipids, such as docosahexanoate-containing (sn-1ether) phosphatidylcholine (PC O) and PE (sn-1ether) phosphatidylethanolamine (PE O), were found exclusively in PCT cells (Figure 3D). The physiological functions of DHA, particularly with regard to its renal metabolism, are complex and not fully understood, but it is thought to have anti-inflammatory and anti-fibrotic properties, particularly in the context of dietary prescriptions.²³ DHA metabolism serves as a precursor for the biosynthesis of specialized anti-inflammatory and pro-resolving mediators with nephroprotective effects in several kidney disease models.²⁴

An intriguing finding in the lipid composition of PCT is the pronounced enrichment of ether phospholipids (Figure 3E). Although ether phospholipids have a similar structure to GP, they differ in having at least 1 ether linkage on carbon 1 of glycerol instead of an ester bound. Ether phospholipids can affect the physical properties of biological membranes in several ways.²⁵ For example, ether PC has several effects, such as membrane rigidity, reduced fluidity, stabilization of membrane domain formation, and stabilization of negatively curved surfaces. Most importantly, compared to PC, they decrease the permeability of cell membranes to ion fluxes and maintain transmembrane ion gradients, including potassium.²⁶,²⁷ Thus, the selective accumulation of ether phospholipids within the PCT may play a role in maintaining the transmembrane potassium gradients that are critical for the reabsorptive functions of this tubule segment. In addition, ether phospholipids are targets for damage by reactive oxygen species (ROS), and animal cells that produce ROS are enriched in ether phospholipids.²⁸ By acting as antioxidants, ether phospholipids may mitigate cell damage caused by oxidative stress in PCT cells, which exhibit high oxidative phosphorylation activity and ROS production.

Sphingolipid Compositional Diversity of Kidney Tubular Segments from Lean Mice

Notable variations have been observed in certain classes of SLs, such as the globosides Gb4, the gangliosides GM3, and SM. Understanding their role in kidney physiology is challenging because their functions in the kidney are poorly understood, but it appears that these lipids are associated with various models of kidney disease.²⁹,³⁰ However, the idea that the SMs produced in vivo have detrimental bioactive properties has been called into question because of analytical issues.

We found that CTAL lacked Gb4 compared to PCT and CCD (Figure 4A ). Gb4 corresponds to the P antigen, the major glycolipid class present in the cell membrane of human erythrocytes. In addition, Gb4 function as a receptor for parvovirus B19 and as an endogenous ligand molecule for the TLR4 receptor.³¹,³² The Stx1a and Stx2a subtypes of Shiga toxins show a preference for binding to Gb4.^31–33

Figure 4.

Compositional diversity of tubule segments in sphingolipids (SL). (A) Histograms of the distribution of Gb4 globosides in proximal convoluted tubule (PCT), cortical thick ascending limb of Henle (cTAL), and cortical collecting duct (CCD) isolated from 5 mouse kidneys. The y-axis (mol%) indicates the percentage expression of a lipid category, i.e., the proportion of total membrane lipids in a sample. P-values were determined by 1-way ANOVA followed by Tukey’s multiple comparison test. (B) Histograms of the distribution of GM3 gangliosides in the PCT, CTAL, and CCD portions of 5 mouse kidneys. The y-axis (mol%) indicates the percentage expression of a lipid category, i.e., the proportion of total membrane lipids in a sample. P-values were determined by 1-way ANOVA followed by Tukey’s multiple comparison test. (C) Histograms of the distribution of sphingomyelins (SM) in the PCT, CTAL, and CCD isolated from 5 mouse kidneys. The y-axis (mol%) indicates the percentage expression of a lipid category, i.e., the proportion of total membrane lipids in a sample. P-values were determined by 1-way ANOVA followed by Tukey’s multiple comparison test.

We also found that GM3 was present at high levels in the CCD and very low levels in the PCT and CTAL (Figure 4B). Of note, among gangliosides, we only identified GM3 in the tubules.³⁴ GM3 inhibits cell growth and growth factor receptor function by direct interaction.³¹,³²,³⁵

Finally, we observed high expression of SM in PCT and comparatively lower expression in CTAL and CCD (Figure 4C). This group of lipids performs various functions such as cell signaling, lipid raft formation concentrated in desmosomes and tight junctions, caveolar endocytosis, and apoptosis.^36–39

This comparative analysis indicates a significant compositional diversity among the SL classes of PCT, CTAL, and CCD. Given the cellular functions of these molecular species, it is highly likely that they are involved in kidney physiological processes, although the nature and mechanisms of these processes remain to be elucidated.

Tubular Segments Lipid Composition Is Affected by Obesity

Prevalent medical conditions like obesity and diabetes affect tubular homeostasis.⁸,⁴⁰,⁴¹ We investigated whether they affect the lipid composition of tubular cells. To this end, we performed lipidomic profiling of PCT, CTAL, and CCD of B6.Cg-Lep^ob/J obese mice, which are homozygous for a mutation of the leptin gene and exhibit obesity and insulin resistance.⁴² Notably, these obese mice are from the same genetic background than the lean mice (C57B6/J).⁴² Hierarchical clustering using the top 100 molecular species with the greatest distribution variability among the 3 segments revealed that tubules still possess a unique lipidomic identity that is unaltered by the metabolic perturbations associated with obesity (Figure S2). However, comparative analysis of lipid class composition revealed unexpected differences between tubules from lean and obese mice, and these contrasts remained consistent across the 3 segments (Figure 5A). For example, the lipid classes with the lowest representation in all 3 segments of obese mice were GM3 and Gb3 (Figure 5B). Furthermore, GM3 was the class of lipids the most significantly underrepresented in the PTC in obese mice, with a fold change that decreased in the more distal segments. In contrast, the content of PIs was significantly increased in obese tubules compared to lean mice in all 3 segments.

Figure 5.

Impact of obesity in the lipid composition of the tubule segments. (A) Dimension reduction using principal component analysis (PCA) of all lipid molecular species identified in 3 segments [proximal convoluted tubule (PCT), cortical thick ascending limb of Henle (cTAL), and cortical collecting duct (CCD)] from 5 lean and obese mice. The resulting plot shows the 2D scores between the selected principal components that best explain the variance of the lipids. (B) Volcano plot comparing lipid classes in obese and lean mice. The x-axis shows log2 [fold change for obese/lean], while the y-axis shows −log10 (P-value) obtained by a -2-sided unpaired t-test with correction for multiple comparisons. (C–F) Histograms of the distribution of triacylglycerides (TG), diacylglycerides (DAG), cholesteryl esters (ce), and cholesterol in the PCT, CTAL, and CCD isolated from 5 mouse kidneys. The y-axis (mol%) indicates the relative expression of a lipid class as a percentage of total membrane lipids in a sample. P-values were computed with 2-way ANOVA with Šídák’s multiple comparison test.

When comparing the individual classes of neutral storage lipids without correction for multiple comparisons as previously performed with the volcano plots shown in Figure 5B, we found that the contents in TG, DG, and ce did not differ between obese and lean mice. However, non-esterified cholesterol levels were significantly higher in the 3 segments analyzed in obese mice (Figure 5C through F). Cellular accumulation of non-esterified cholesterol may play a critical role in kidney injury and progression of associated diseases.⁵,⁴³,⁴⁴

Proximal Convoluted Tubules of Obese Mice are Enriched in Phosphatidylethanolamine Species Sensitive to Ferroptosis

ACSL4 is a critical determinant of ferroptosis susceptibility, a regulated cell death pathway associated with the dysregulation of lipid oxidative metabolism that can occur in the proximal tubule during acute kidney injury and CKD.^45–49 Indeed, ACSL4 preferentially enriches cell membranes with PE molecular species with FA 18 at the sn-1 position and AA or FA 22:4 (adrenic acid, AdA) at the sn-2 position, the latter of which are oxidation substrates for ferroptosis execution.⁴⁹,⁵⁰ AA and AdA are critical determinants of susceptibility to ferroptosis because, when incorporated into membranes, they are highly susceptible to peroxidation due to the presence of weak C-H bonds between adjacent C = C double bonds.⁵¹

Therefore, we performed a detailed analysis of the AA and AdA composition of individual PE molecular species in PCT from lean and obese mice. Strikingly, we found that AA and AdA at the sn-2 position were significantly more abundant in PE in obese PCT cells than in PCT from lean mice. Furthermore, we did not observe a similar enrichment of ferroptosis-sensitive PE species in CTAL and CCD, suggesting a particular susceptibility of PCT to this lipid remodeling (Figure 6  B and C). Interestingly, an enrichment at the sn-2 position with FAs with shorter chains, such as C18s, which are not activated by ACSL4, was not observed in obese mice (Figure 6D). We measured Acsl4 transcripts expression in PCT isolated from the kidneys of obese and lean mice and found that expression was more elevated in the obese PCT compared to the lean PCT, which is not the case in CTALs and CCDs and corroborates the specific lipidomic profile of obese PCTs enriched in AA- and AdA-carrying PE (Figure 6E). This is consistent with ACSL4’s preference for highly unsaturated FAs such as AA and AdA as substrates.¹⁸ Thus, the enrichment of cell membranes with AA and AdA-PE species, which can be directly oxidized during ferroptosis, suggests that the PCT in obese mice may have an increased susceptibility to ferroptosis due to the activity of ACSL4.

Figure 6.

Proximal convoluted tubules (PCTs) of obese mice are enriched in phosphatidylethanolamine molecular species sensitive to ferroptosis. (A) Histograms of the distribution of PE 18:0/20:4 in the PCT, cortical thick ascending limb of Henle (cTAL), and cortical collecting duct (CCD). The y-axis (mol%) indicates the relative expression of a lipid class as a percentage of total membrane lipids in a sample. P-values were computed with multiple t-tests with false discovery rate (FDR) control. (B) Histograms of the distribution of PE 18:1/20:4 in the PCT, CTAL, and CCD. The y-axis (mol%) indicates the relative expression of a lipid class as a percentage of total membrane lipids in a sample. P-values were computed with multiple t-tests with FDR control. (C) Histograms of the distribution of PE 18:1/22:4 in the PCT, CTAL, and CCD. The y-axis (mol%) indicates the relative expression of a lipid class as a percentage of total membrane lipids in a sample. P-values were computed with multiple t-tests with FDR control. (D) Histograms of the distribution of PE 18:0/18:1, PE 18:0/18:2, PE 18:1/18:2, and PE 18:2/18:2 in the PCT of lean and obese mice. The y-axis (mol%) indicates the relative expression of a lipid class as a percentage of total membrane lipids in a sample. P-values were computed with multiple t-tests with FDR control. (E) Acsl4 transcripts measured by RT-qPCR of mRNA from PCT, CTAL, and CCD isolated from 12-wk-old obese mice and lean littermates (n = 5 mice per group). P-values were computed with multiple t-tests with FDR control.

Therefore, the enrichment of cell membranes with AA and AdA containing PE, which can be directly oxidized during ferroptosis, suggests that the PCT in obese mice may have an increased susceptibility to ferroptosis due to the activity of ACSL4. Dedicated experiments are needed to demonstrate that PCT in obese kidneys has an increased susceptibility to lipid peroxidation in situations where ferroptotic activity is increased.

Discussion

We present, for the first time, a comprehensive analysis of the lipidome of 3 cortical segments of mouse kidney tubules. Our findings provide an unprecedented level of detail and are critical for a deeper understanding of tubular physiology and the implications of pathological conditions. This research paves the way for numerous avenues of investigation in the field of tubular lipid metabolism. Access to state-of-the-art lipidomic analysis platforms and the use of microdissection or isolation methods provide opportunities for comparative studies of tubular lipid composition in various conditions, ranging from primary tubular pathologies to those secondary to glomerulopathies, systemic metabolic abnormalities, or therapeutic interventions targeting the tubule.

The most remarkable finding of the study is the identification of different compositions among the 3 tubular segments, which are considered as true signatures. These inherent lipidomic features, likely supported by a specific proteomic program and undoubtedly involved in the regulation of precise physiological functions, are consistent with the hypothesis of a lipotype assigned to a tubular segment. Lipotypes are both a result and a component of differentiation programs that lead to the emergence of different cell types and can define different anatomical structures.⁵² According to this hypothesis, lipid distribution should precisely replicate the most delicate aspects of tissue anatomy. Changes in lipid metabolism could alter cell type composition and cause defects in tissue configuration. Lipotypes are critically considered to be bidirectionally linked to transcriptional cell states, with lipid composition being a consequential driver of cell fate. Therefore, multi-omics integration (proteomics and lipidomics data) is worth to mention as an exciting new possibility for system-level type of studies.

This is important in the context of the results: There may be a more or less favorable environment for a particular signaling event due to differences in membrane composition. For example, we found that the tubular composition of obese mice lacked GM3, which is critical for growth factor signaling in membranes. GM3 inhibits cell growth through direct interaction with growth factors.³¹,³²,³⁵ In addition, the decrease in GM3 specifically in the proximal tubule cells of obese mice compared to other sections is consistent with findings suggesting that the nephromegaly associated with obesity predominantly targets the PCT.⁴⁰

Our data also provide novel insights into tubular lipid remodeling in the kidney tubule during obesity. For example, we observed that the PCT of obese mice is enriched in certain PE molecular species, particularly those sensitive to ferroptosis, probably driven by ACSL4 activity. The involvement of ACSL4 in the pathophysiology kidney diseases is beginning to be described, and the emerging picture associates it with an injured phenotype and cell death.⁵³ Our observations are in keeping with a body of evidence that the most important risk factor for acute kidney injury is preexisting CKD, which increases risk by as much as 10 times, as compared with the absence of CKD. The mechanisms by which this occurs are yet poorly understood, and in a pathophysiological point of view, our results make it possible to propose a hypothesis according to which an epithelium engaged in a maladaptive and chronically injured response, promotes the expression of ACSL4. As an increased activity of ACSL4 sensitizes TEC to cell death, it is possible to consider that this permanently injured epithelium offers less resistance during exposure to a ferroptotic signal with consequent cell death. This is consistent with the recent observation that lipid anabolism in PCT is increased during CKD to support membrane synthesis and cell growth.⁹ Critically, this study identified a number of lipid-activated transcription factors (some of which regulate ACSL4 expression) that are involved in determining proximal tubule cell phenotype in chronically diseased kidneys. Taken together, these observations raise the possibility that lipid biosynthesis during CKD may be characterized by a diversification of molecular species. As a result of qualitative changes in the composition of membranes, PCT phenotype and function may be affected, as well their response to stress.

Notably, B6.Cg-Lepob/J obese mice become obese but are only mildly hyperglycemic and do not develop kidney lesions characteristic of human diabetes. In this context, our results are important because they indicate that obesity produces marked alterations in lipid metabolism, differentially affecting different tubular segments. Critically, these alterations occur without obvious tissue damage, suggesting that they may precede phenotypic changes associated with diabetic/obese nephropathy. Comprehensive lipidomic analyses of early-stage diabetic nephropathy have been performed on kidney tissue from diabetic mice, and pathologically relevant lipid species were identified by comparing the renal lipidome between normal and diabetic rats. For example, in an 8-wk high fat diet feeding combined with low-dose streptozotocin injection model, glycerides, including TG, DG, FA, Chol, and ce, were significantly increased in the kidneys of the diabetic nephropathy group.⁵⁴ Neutral lipids comprised the majority of the diabetic nephropathy kidney lipidome, with most having a higher degree of unsaturation and linoleic acid side chains, which may serve as potential markers for diabetic nephropathy. Significant changes were observed in the least abundant lipid classes, particularly phospholipids and SL such as Cer and GM3. However, a direct comparison with our results is not possible, as the data obtained in the study cited above were generated from renal cortices, which erase the metabolic heterogeneity that exists in each segment of the nephron. In addition, there is a systemic (ie, blood) lipidomic signature of obesity that may be associated with risk of progression to CKD.⁵⁵ Whether this signature reflects or correlates with intra-renal lipidomic changes is an open question, but if so, it would provide a form of liquid biopsy for assessing the metabolic reprogramming that occurs in obese kidneys.

The lipid compositional diversity of the kidney’s elementary structures has been suggested, in particular through the use of lipid mass spectrometry imaging techniques, such as Matrix-assisted laser desorption/ionization (MALDI) imaging mass spectrometry (IMS).⁵⁶ This allows detailed study on the architecture of the human kidney, which can be taken as a kind of molecular histology. Using this technology, it is possible to determine lipid distribution maps that describe the histology of the tissue from a metabolic point of view. However, a limitation of this technique is the difficulty in identifying the architecture of the tissue with a level of detail. This technique and our approach are not mutually exclusive, especially since we have been able to identify over 500 lipid species in perfectly identified tubular segments and have been able to generate a significant number of pathophysiological hypotheses, which is the bit of these non-targeted approaches.

We performed a compositional analysis of membrane lipids, and consequently, there is lack of report on non-esterified or free FAs. This can be seen as a methodological limitation, given that pathophysiologic changes in cell signaling and damage are attributed to free FAs. In addition, while our analyses cover only 3 cortical segments, it is important to examine additional segments and conditions, particularly to identify lipotypes of poorly studied kidney tubule segments, but also the effects of kidney disease or treatments.

In conclusion, we offer an extensive lipidomic analysis of 3 cortical sections of mouse kidney tubules. This valuable resource provides unmatched information that advances our understanding of tubular physiology and the potential influence of pathological conditions.

Author Contributions

L.C., V.P., J.L.S., and G.C.: conducted experiments; N.P. and G.C.: designed research studies, acquired data, analyzed data, and provided reagents. N.P.: wrote the manuscript.

Funding

This work was funded by grants from l’Agence de la Biomédecine (N.P.), La Fondation du Rein (N.P.), and the Agence Nationale de la Recherche (A.N.R.) project ANR-21-CE14-0040-01 (G.C.).

Conflict of Interest

The authors have no conflict of interest to disclose.

Data Availability

The analytical process for lipidomics according to lipidomic standards (https://lipidomicstandards.org/reporting_checklist/) has been deposited and is publicly available as of the date of publication (https://doi.org/10.5281/zenodo.10889906).

Raw data of all lipids identified with their m/z values is available in the file untitled “kidney_tubules_paper_data” in the Supplementary Data.

The rat tubule proteome database is deposited to the ProteomeXchange Consortium (http://proteomecentral.proteomexchange.org) via the PRIDE partner repository with the dataset identifier PXD016958.