Renal ketogenesis protects against ischemic kidney injury

Abstract

Background:

Abnormal renal fatty acid oxidation in kidney disease suggests that dysregulated metabolism is a key component of kidney disease pathogenesis. While the liver is the main ketogenic organ, the rate-limiting enzyme for ketogenesis, mitochondrial Hydroxymethylglutaryl-CoA synthase 2 (HMGCS2), is induced in the proximal tubule of the kidney during fasting. We previously demonstrated that HMGCS2 induced in the kidney does not contribute to the circulating pool of ketones during fasting and cannot compensate for hepatic ketogenic deficiency. We hypothesized that kidney HMGCS2 may be acting locally within the kidney to maintain normal function during metabolic stress or injury.

Methods:

Mice with kidney or liver specific deletion of Hmgcs2 were subjected to ischemia/reperfusion injury (IRI). Kidney histology, metabolomics and lipidomics were analyzed. Mice were placed on a ketogenic diet for four days to increase plasma and kidney ketone content. Using novel mouse models with proximal tubular hemagglutinin-tagged mitochondria with or without Hmgcs2 deletion, proximal tubular-specific mitochondria were isolated and fatty acid oxidation capacity was measured after IRI.

Results:

Mice with kidney specific Hmgcs2 deletion had significantly more kidney injury after IRI compared to wild-type controls. Kidneys lacking HMGCS2 exhibited a decrease in ketone content and an increase in lipid droplet accumulation after IRI. Proximal tubular-specific mitochondria lacking HMGCS2 had significantly lower fatty acid oxidation capacity both at baseline and after ischemic injury. Administration of a ketogenic diet for four days prior to IRI was sufficient to decrease kidney injury and augment mitochondrial fatty acid oxidation in kidney Hmgcs2 knockout mice. Kidney tissue lipidomics revealed that the loss of kidney HMGCS2 was associated with a decrease in both arachidonic acid containing phospholipids and prostaglandin levels.

Conclusion:

Loss of renal HMGCS2 and resultant ketogenesis increased ischemia-induced injury and decreased mitochondrial fatty acid oxidation capacity, suggesting a role in renal ketogenesis in limiting acute kidney injury.

Introduction

Dysregulated metabolism, specifically abnormal fatty acid oxidation, is a key component of kidney disease pathogenesis^{1, 2}. β-oxidation of fatty acids results in the production of acetyl-CoA which can be used to produce ketones bodies, such as acetoacetate and β-hydroxybutyrate. As maintenance of fatty acid oxidation and related metabolic pathways are suggested to be important for kidney health, it is important to understand whether the ketogenic pathway is relevant to kidney health and disease. In addition to being an alternative fuel, ketones have been described to be antioxidants through other signaling or post-translational modification pathways³. While the liver functions as the main ketogenic organ⁴, expression of mitochondrial Hydroxymethylglutaryl-CoA synthase 2 (HMGCS2), the rate-limiting enzyme for ketogenesis, is induced in the proximal tubule of the kidney during fasting⁵.

Prior work has correlated renal HMGCS2 expression to an increase of circulating ketones⁶ and the pathophysiology of diabetic kidney disease⁷. However, we previously demonstrated that HMGCS2 induced in the kidney cannot compensate for hepatic ketogenic deficiency during fasting, suggesting the kidney does not significantly contribute to the pool of circulating ketones⁵. This contrasts with gluconeogenesis in which the kidneys can and will compensate for hepatic gluconeogenic deficiency^8–10. Thus, proximal tubular HMGCS2 is likely to have a local function. Moreover, transcriptomic analysis of chronic kidney disease (CKD) biopsies reveals that CKD is associated with a significant depression of renal HMGCS2 expression, supporting a potential role for HMGCS2 in kidney disease development (Table 1). Using a mouse model of kidney specific Hmgcs2 deletion, we sought to determine the function of renal HMGCS2 in response to ischemic kidney injury.

Table 1.

Kidney HMGCS2 expression was decreased in chronic kidney disease

Dataset	Kidney HMGCS2 Fold change (log2)	P-value
Lupus Nephritis vs Healthy Living Donor (ERCB Lupus TubInt)44	−2.58	5.75E-04
Vasculitis vs Healthy Living Donor (Ju CKD TubInt)44, 45	−2.01	4.62E-05
IgA Nephropathy vs Healthy Living Donor (Ju CKD TubInt)44, 45	−1.55	0.006
CKD vs Normal Kidney: Discovery Set (Nakagawa CKD Kidney)44, 46	−3.13	1.57E-12
CKD vs Normal Kidney: Validation Set (Nakagawa CKD Kidney)44, 46	−5.15	0.016
Acute Rejection vs Normal Kidney (Sarwal Transplant Kidney)44, 47	−1.62	0.013
Cadaveric Donor Group: Acute vs No Rejection (Flechner Transplant Kidney)44, 48	−1.74	0.019
CKD vs Normal Kidney: Kang et al Nat Med 20151	−1.76	Corrected P<0.05

Methods

In Vivo Studies

All animal experiments were performed in accordance with institutional regulations after protocol review and approval by Institutional Animal Care and Use Committees at the University of Texas Southwestern Medical Center. The Hmgcs2 conditional knockout (floxed) mouse model was generated by CRISPR-Cas9 methodology as previously described⁵. Six2-Cre mice (Tg(Six2-EGFP/cre)^1Amc/J, #009606) were purchased from Jackson Laboratory. Alb-CreERT2 mice were a generous gift from Dr. Pierre Chambon¹¹. MITO-Tag (RRID:IMSR_JAX:032290) and Ggt1-Cre (RRID:IMSR_JAX:012841) mice were purchased from Jackson Laboratory. All mouse strains were maintained on a C57BL/6 background and housed under standard laboratory conditions with a 12-hour light:dark cycle. Male mice were provided ad libitum access to water and standard chow (Harlan Teklad, 2916), except where noted by specific experimental protocols.

Ad libitum fed male 12- to 16-week old mice were anesthetized with ketamine/xylazine and subjected to ischemia reperfusion injury (IRI) using a modified approach as previously described^{12, 13}. For experiments including contralateral nephrectomy, the right kidney pedicle was ligated followed by nephrectomy, while the left kidney pedicle was clamped for 25 minutes using a nontraumatic vascular clamp (Fine Science, 18055–05) for IRI or not clamped for sham controls. For unilateral IRI experiments, the left kidney pedicle was clamped for 30 minutes and contralateral kidney was not injured. For bilateral IRI experiments, both the left and the right kidneys were clamped for 30 minutes. Post-operatively, mice were returned to the cage with ad libitum access to chow pellets in the hopper as well as moist chow on the cage floor.

For glucose supplementation experiments, mice were orally gavaged with either glucose (50 mg in 100 μL sterile water per mouse, approximately 2g glucose/kg body weight) or water at 2, 7, and 23 hours after IRI. At 24 hours post-IRI, blood and tissue samples were collected.

For the ketogenic diet experiments, mice were fed a ketogenic diet (Ketogenic Diet AIN-76A-Modified, High Fat, Bio-Serv. F3666) beginning 4 days prior to IRI surgery. The diet was available continuously before and after surgery until tissue and plasma harvest.

RNA and protein measurements

Quantitative RT-PCR primers and antibodies are listed in Supplemental Tables 1 and 2, respectively. RNA and protein isolation methods for quantitative analysis are detailed in Supplemental Methods.

Histology

Kidneys were fixed and stained for histology according to subsequent downstream processing. Hematoxylin and eosin (H&E) or Picrosirius Red staining was performed by the Histology Core of the George M. O’Brien Kidney Research Core at UT Southwestern. Electron microscopy was performed by the UT Southwestern Electron Microscopy Core. Detailed methods of tissue fixation and staining are included in Supplemental Methods.

Immunoprecipitation of Tagged Mitochondria

Cell specific mitochondria were isolated as described in Bayraktar EC et al.¹⁴ Isolated mitochondria were then processed for either immunoblot analysis or respiratory analysis by Seahorse. Detailed methods are included in Supplemental Methods.

Plasma and Tissue Measurements

Plasma creatinine was measured by capillary electrophoresis by The George M. O’Brien Kidney Research Core at UT Southwestern. Measurement of other analytes is described in the Supplemental Methods.

Metabolomics, Lipidomics, and Prostaglandin Measurements

Kidney tissue metabolomics and lipidomics were performed by the UT Southwestern Children’s Medical Center Research Institute Metabolomics Core. Kidney prostaglandin content was performed at the UCSD Lipidomics Core. Methods are detailed in Supplemental Methods.

Statistics

Sample size calculation was based on the plasma creatinine difference between wild-type and Six2-Cre;Hmgcs2^fl/fl mice after IRI. Sham-operated animals were included at lower minimum required numbers because of their consistent lack of injury across experiments, in accordance with IACUC guidelines to minimize animal use. Sample sizes varied across experiments due to differences in tissue availability and technical feasibility. All available animals were included for analysis unless precluded by technical limitations. Exact n numbers for each experiment are indicated in the corresponding figure legends.

Statistical analyses were performed using Prism 10.0 (GraphPad). Student’s unpaired two-way t-test, one-way ANOVA, and two-way ANOVA with post hoc multiple comparison analyses were used when appropriate. Data were tested for normality and homogeneity of variance. Statistical analysis to account for non-normal data or heterogeneity of variance is specified in figure legends. Data are expressed as mean ± SD. A P value < 0.05 was considered statistically significant. Prism 10 (GraphPad), Canva, and BioRender.com were used to generate figures.

Results

Nephrectomy and ischemic injury induced kidney HMGCS2

Using a model of unilateral ischemia/reperfusion injury (IRI) and contralateral nephrectomy, we found that both IRI and sham surgery induced kidney HMGCS2 expression to levels comparable to those seen in response to fasting (Figure 1A,B). In a model of unilateral renal IRI, HMGCS2 expression was also upregulated in both the IRI kidney and the contralateral control (Supplemental Figure 1). As the mice were exposed to surgery and anesthesia, it is likely that a decrease in food intake could contribute to a semi-fasted state. When mice were given three doses of glucose (2g/kg body weight) post-operatively sufficient to increase circulating insulin levels, kidney HMGCS2 expression was partially suppressed in both sham and IRI kidneys (Figure 1C–E). Thus, a decrease in food intake and insulin signaling only partially contributed to renal HMGCS2 expression after sham or IRI surgery, suggesting an additional stress-induced component.

Figure 1. HMGCS2 expression in the proximal tubule of the kidney.

(A-B) Wild-type mice were either fed ad libitum, fasted for 24 hours or underwent contralateral nephrectomy and unilateral ischemia/reperfusion injury (IRI) or sham surgery.

(A) Whole kidney protein lysate immunoblotted for HMGCS2, n=3/group.

(B) Densitometry of (A) relative to total protein.

(C-E) Wild-type mice underwent contralateral nephrectomy and unilateral IRI or sham surgery and were gavaged water or glucose after IRI, n=4–6/group.

(C) Whole kidney protein lysate immunoblotted for HMGCS2.

(D) Densitometry of (C) relative to total protein.

(E) Plasma insulin measured by ELISA.

(F-H) Six2^Hmgcs2KO (KO) and WT littermates underwent contralateral nephrectomy and unilateral IRI or sham surgery. Kidney tissue harvested 24 hours later.

(F) Whole kidney protein lysate immunoblotted for HMGCS2, n=3/group.

(G) Densitometry of (F) relative to total protein.

(H) Immunostaining for HMGCS2 (red), lotus lectin (LTL, a proximal tubule marker, green), and DAPI (blue). Scale bar represents 100 μm.

Data shown mean ± SD. Lognormal one-way ANOVA with Holm-Sidak multiple comparisons test (B); two-way ANOVA with Sidak’s multiple comparisons (D,E,G).

In order to determine the role of renal HMGCS2 in ischemic injury, we generated kidney specific Hmgcs2 knockout mice using the Six2-Cre;Hmgcs2^fl/fl (Six2^Hmgcs2KO) mouse model, as previously described⁵ (Figure 1F,G). Similar to fasting, HMGCS2 protein was expressed in proximal tubule cells after sham and IRI surgery in WT mice and was lost in Six2^Hmgcs2KO kidneys (Figure 1H).

Renal HMGCS2 deficiency led to increased ischemic injury and fibrosis

At baseline, Six2^Hmgcs2KO mice exhibited normal kidney function⁵. After a contralateral nephrectomy and unilateral IRI model in which wild-type animals sustain minimal injury, Six2^Hmgcs2KO mice exhibited more severe kidney injury, as evidenced by higher plasma creatinine levels, compared with wild-type littermates (Hmgcs2^fl/fl) and Six2-Cre^+/− controls (Figure 2A, Supplemental Figure 2A). Kidney injury molecule-1 (KIM-1) protein levels were also significantly higher in ischemic Six2^Hmgcs2KO kidneys (Figure 2B,C). Consistent with plasma creatinine and kidney KIM-1 levels, histologic tubular injury was also significantly worse in Six2^Hmgcs2KO mice (Figure 2D,E). Using a model of unilateral IRI, we next assessed how the loss of renal HMGCS2 contributes to the development of kidney fibrosis. Fourteen days after unilateral IRI, post-ischemic kidneys lacking HMGCS2 exhibited more Picrosirius red staining compared to post-ischemic wild-type kidneys (Figure 2F,G). Similarly, post-ischemic Six2^Hmgcs2KO kidneys exhibited higher expression levels of fibrotic markers such as Transforming Growth Factor Beta 1 (Tgfb1), fibronectin (Fn) and alpha I type I collagen (Col1a1), as well as an increase in inducible nitric oxide synthase 2 (Nos2) expression, suggesting increased inflammation (Figure 2H).

Figure 2. Lack of renal HMGCS2 led to increased ischemic injury and fibrosis.

(A-E) Six2^Hmgcs2KO (KO) and Hmgcs2^fl/fl (WT) mice underwent contralateral nephrectomy and unilateral IRI or sham surgery. Blood and kidney tissue harvested 24 hours later.

(A) Plasma creatinine, n=7–11/group.

(B) Representative immunoblot of whole kidney lysate for KIM-1.

(C) Quantification of (B) relative to total protein, n=3–11/group.

(D) Hematoxylin and eosin stained kidney sections. Scale bar represents 100 μm. Arrows highlight areas of tubular casts, tubular dilatation, and nuclear drop-out. Green arrow highlights a hyaline cast and the yellow arrows highlight proteinaceous casts.

(E) Tubular injury scoring of (D), n=3–4/group.

(F-H) Six2^Hmgcs2KO (KO) and Hmgcs2^fl/fl (WT) mice underwent unilateral IRI. Kidney tissue harvested 14 days later. CLC, contralateral control.

(F) Picrosirius red stained kidney sections. Scale bar represents 100 μm.

(G) Quantification of (F) percent area positive for Picrosirius red stain, n=4/group.

(H) Whole kidney mRNA expression shown relative to Rpl13a, n=7/group.

Data expressed as mean ± SD; two-way ANOVA with Sidak’s multiple comparisons (A,C,E,G,H).

Renal HMGCS2 promoted local renal ketone production

As the increased fibrosis in Six2^Hmgcs2KO kidneys is likely due to increased injury during the initial ischemic insult, we further investigated the early injury period. We performed global kidney metabolomic profiling to identify potential metabolic pathways regulated by renal HMGCS2. Global kidney metabolomics revealed only treatment differences (Supplemental Figure 3A, Supplemental Tables 3, 4), while there were no significant genotype differences nor interaction between genotype and condition (sham vs IRI). Pairwise comparisons between Six2^Hmgcs2KO IRI and WT IRI kidneys revealed an accumulation of urate and ceramides, SM (18:1,16:0) and SM (18:1,14:0), known to be associated with kidney disease^15–17 (Supplemental Figure 3B,E, Supplemental Table 5). Consistent with the loss of renal HMGCS2, kidney tissue 3-Hydroxybutyrate, or β-hydroxybutyrate, one of the main ketone bodies, was lower in Six2^Hmgcs2KO kidneys after IRI compared to WT IRI kidneys but did not reach statistical significance (Figure 3A, Supplemental Table 5). As the screening untargeted metabolomics data obtained by LC-MS/MS provided relative metabolite abundances and not absolute quantitative measurements, we used an enzymatic assay to validate β-hydroxybutyrate tissue content. There was significantly higher β-hydroxybutyrate levels within the wild-type kidney after IRI compared to sham kidneys, while the β-hydroxybutyrate content did not differ in Six2^Hmgcs2KO kidneys between sham or IRI (Figure 3B).

Figure 3. Kidney HMGCS2 promoted local renal ketogenesis.

(A) Six2^Hmgcs2KO (K-KO) and WT mice underwent bilateral IRI or sham surgery. Metabolomics performed on kidney cortex tissue. Metabolite abundance shown normalized to total ion count (TIC), n=3–4/group.

(B,C) Six2^Hmgcs2KO (K-KO), Hmgcs2^fl/fl (WT) mice underwent contralateral nephrectomy and unilateral IRI or sham surgery. Blood and kidney tissue harvested 24 hours later, n=6–7/group

(B) Kidney β-hydroxybutyrate (D-βOHB) tissue levels measured by enzymatic assay shown normalized to tissue weight.

(C) Plasma D-βOHB levels measured by enzymatic assay.

(D-F) Alb^Hmgcs2KO (L-KO), Hmgcs2^fl/fl (WT) mice underwent contralateral nephrectomy and unilateral IRI or sham surgery. Blood and kidney tissue harvested 24 hours later, n=6–7/group.

(D) Kidney D-βOHB tissue levels shown normalized to tissue weight. Mean group values are shown.

(E) Plasma D-βOHB levels. Mean group values are shown.

(F) Plasma creatinine levels.

(G-I) Six2^Hmgcs2KO (K-KO), Hmgcs2^fl/fl (WT) mice were placed on a ketogenic diet (KD) for 4 days, then underwent contralateral nephrectomy and unilateral IRI or sham surgery. Blood and kidney tissue harvested 24 hours later, n=4–7/group.

(G) Kidney D-βOHB tissue levels shown normalized to tissue weight. Dotted lines reflect mean kidney D-βOHB tissue levels in chow fed IRI mice (from Figure 3B).

(H) Plasma D-βOHB levels.

(I) Plasma creatinine levels. Dotted lines reflect mean plasma creatinine levels in chow fed IRI mice (from Figure 2A).

(J-L) Wild-type mice underwent contralateral nephrectomy and unilateral IRI or sham surgery and were gavaged water or glucose after IRI (as in Figure 1C–E), n=4–6/group

(J) Kidney D-βOHB tissue levels shown normalized to tissue weight. Mean group values are shown.

(K) Plasma D-βOHB levels. Mean group values are shown.

(L) Plasma creatinine levels.

Data expressed as mean ± SD, unpaired t-test performed on Metaboloanalyst (A), two-way ANOVA with Sidak’s multiple comparisons (A-L).

As the liver is the main source of circulating ketones, a large component of measured tissue ketones is liver-derived. To account for plasma-derived β-hydroxybutyrate, we analyzed kidney tissue β-hydroxybutyrate relative to plasma levels. A higher β-hydroxybutyrate kidney tissue content normalized to plasma levels suggests renal-derived ketone production. When normalized to plasma levels (Figure 3C), β-hydroxybutyrate was lower in Six2^Hmgcs2KO kidneys compared to WT kidneys after IRI (Supplemental Figure 4I). Indeed, during fasting conditions, ketone levels within the kidney were high, reflecting high circulating plasma levels (Supplemental Figure 4A–C). In liver Hmgcs2 knockout mice (Alb-CreERT2;Hmgcs2^fl/fl, or Alb^Hmgcs2KO)⁵ both plasma and kidney fasting β-hydroxybutyrate content were significantly reduced (Supplemental Figure 4A–B). However, when accounting for the plasma β-hydroxybutyrate levels, kidney-derived β-hydroxybutyrate was preserved (Supplemental Figure 4C). These data suggested that a significant portion of kidney β-hydroxybutyrate content was liver-derived when plasma β-hydroxybutyrate levels were high.

To further differentiate the role of circulating ketones and local kidney ketone production in protecting against IRI, we subjected Alb^Hmgcs2KO mice to renal IRI. Compared to WT IRI kidneys, β-hydroxybutyrate content trended lower in IRI kidneys from Alb^Hmgcs2KO mice (Figure 3D). This difference was due to plasma β-hydroxybutyrate levels which also trended lower in IRI Alb^Hmgcs2KO mice (Figure 3E). After IRI in wild-type animals, plasma levels of β-hydroxybutyrate were modestly increased compared to sham animals, albeit lower than fasting levels (Figure 3C, Supplemental Figure 4A). When normalized to plasma ketone levels, kidney ketones levels were not affected in Alb^Hmgcs2KO mice (Supplemental Figure 4G). Interestingly, Alb^Hmgcs2KO mice did not exhibit susceptibility to renal IRI (Figure 3F).

To further enhance circulating liver-derived ketones, we used a carbohydrate-restricted ketogenic diet which has been shown to protect against renal IRI¹⁸. Mice were given a ketogenic diet 4 days prior to renal IRI. The ketogenic diet significantly increased renal HMGCS2 expression (Supplemental Figure 5A). Compared to mice on standard chow, the ketogenic diet significantly increased β-hydroxybutyrate levels in plasma and kidney tissue of both wild-type and Six2^Hmgcs2KO mice, but not Alb^Hmgcs2KO mice, in both sham and IRI conditions (Figure 3G,H, Supplemental Figure 4D,E). Interestingly, the loss of renal ketogenesis was more apparent in the Six2^Hmgcs2KO mice on the ketogenic diet, as the kidney β-hydroxybutyrate levels were significantly lower than in wild-type kidneys, and specifically only in the IRI condition (Figure 3G, Supplemental Figure 4J). In contrast, the absolute kidney β-hydroxybutyrate content in ketogenic diet Six2^Hmgcs2KO kidneys was ~6.25-fold higher than Six2^Hmgcs2KO kidneys of mice on standard chow after renal IRI (Figure 3B vs 3G, Supplemental 4I,J). This increase in kidney β-hydroxybutyrate content was sufficient to mitigate kidney injury in Six2^Hmgcs2KO mice after IRI (Figure 3I).

In Alb^Hmgcs2KO mice, the ketogenic diet enhanced the difference in sham versus IRI renal ketone content that was independent of plasma liver-derived ketones, further supporting the effect of local ketogenesis induced by IRI (Supplemental Figure 4D–F,H). Notably, although glucose supplementation modestly decreased renal HMGCS2 (Figure 1C,D) and β-hydroxybutyrate content (Figure 3J), this did not result in worse kidney injury (Figure 3L). The decrease in kidney β-hydroxybutyrate content by glucose was driven by the decrease in plasma β-hydroxybutyrate (Figure 3K, Supplemental Figure 4L). When accounting for the effect of glucose on plasma β-hydroxybutyrate levels, the amount of kidney-derived β-hydroxybutyrate did not change (Supplemental Figure 4K).

Loss of renal ketogenesis increased lipid accumulation in the kidney after ischemic injury

To investigate the consequences of the loss of renal ketogenesis, we next assessed ketone related post-translational modifications (Supplemental Figure 6A). Ketones can act as histone deacetylase inhibitors, resulting in increased chromatin accessibility particularly for genes involved in antioxidant responses¹⁹. Beta-hydroxybutyrate has been described to directly modify chromatin, resulting in lysine beta-hydroxybutyrylation (Kbhb)²⁰. We assessed the global histone lysine acetylation (Kac) and Kbhb of sham and IRI kidneys and found no significant differences between Six2^Hmgcs2KO and WT kidneys (Supplemental Figure 6B,C), suggesting that the lack of renal HMGCS2 did not significantly affect Kac and Kbhb histone modifications.

Next, we examined inflammation and oxidative stress-related transcripts between WT and Six2^Hmgcs2KO kidneys after IRI and found no significant differences (Figure 4A, Supplemental Figure 2B). As Hmgcs2 is a nuclear-encoded mitochondrial gene, we next assessed whether the loss of HMGCS2 in kidney mitochondria affected mitochondrial properties. Using voltage-dependent anion channel (VDAC) as a surrogate for mitochondrial abundance, we found that there was no significant difference in VDAC protein abundances between sham and IRI conditions or between Six2^Hmgcs2KO and WT kidneys (Figure 4B,C). Similarly, mitochondrial DNA abundance relative to nuclear DNA abundance was also not significantly different (Figure 4D). We also assessed protein abundances of components of the oxidative phosphorylation machinery of the electron transport chain and found no significant differences (Figure 4E,F). While we did not find remarkable differences in mitochondrial morphology between Six2^Hmgcs2KO and WT kidneys on electron microscopy, there were notably more lipid droplets in Six2^Hmgcs2KO kidneys after IRI (Figure 4G). This increase in lipid droplets on electron microscopy was confirmed with Oil Red O staining (Figure 4H,I).

Figure 4. Loss of renal ketogenesis increased lipid accumulation in the kidney after ischemic injury.

(A-C) Six2^Hmgcs2KO (KO) and WT mice underwent contralateral nephrectomy and unilateral IRI or sham surgery. Kidney tissue harvested 24 hours later.

(A) Whole kidney mRNA expression shown relative to Rpl13a, n=5–8/group.

(B) Whole kidney protein lysate immunoblotted for VDAC, n=6/group.

(C) Quantification of (B) shown relative to total protein.

(D) Six2^Hmgcs2KO (KO) and WT mice underwent unilateral IRI. Kidney tissue harvested 24 hours later. Mitochondrial DNA (mtDNA) content shown relative to nuclear DNA (nDNA) content. Data shown compared to WT contralateral control (CLC), n=13–15/group.

(E-I) Six2^Hmgcs2KO (KO) and WT mice underwent contralateral nephrectomy and unilateral IRI or sham surgery. Kidney tissue harvested 24 hours later.

(E) Whole kidney protein lysate immunoblotted for OXPHOS proteins, n=4–5/group.

(F) Quantification of (E) shown relative to total protein.

(G) Transmission electron microscopy of kidney tissue. Scale bar represents 1 μm.

(H) Kidney tissue stained with Oil Red O. Scale bar represents 50 μm, n=5/group.

(I) Quantification of (H) by percent area stained for Oil Red O using ImageJ.

Data expressed as mean ± SD; two-way ANOVA with Sidak’s multiple comparison (A,C,D,F), two-sided, unpaired t-test (I).

Renal ketogenic deficiency decreased fatty acid oxidation in the kidney

Lipid accumulation can develop in the setting of increased uptake, increased storage, increased synthesis (i.e., de novo lipogenesis), and/or decreased utilization (i.e., fatty acid oxidation). The loss of HMGCS2 can cause an increase in acetyl-CoA, a substrate for de novo lipogenesis. While there was a slight trend of higher acetyl-CoA in Six2^Hmgcs2KO IRI kidneys compared to wild-type, acetyl-CoA levels overall were decreased after IRI in both wild-type and Six2^Hmgcs2KO kidneys (Supplemental Figure 3F). We also screened the expression of de novo lipogenesis genes and found no significant differences (Figure 5A).

Figure 5. Renal ketogenic deficiency decreased fatty acid oxidation in the kidney.

(A) Six2^Hmgcs2KO (KO) and WT mice underwent contralateral nephrectomy and unilateral IRI Kidney tissue harvested 24 hours later. Whole kidney mRNA expression shown relative to Rpl13a, n=10–11/group.

(B) Overview of MITO-Tag mouse model.

(C) Immunostain of Ggt1-Cre;MITO-Tag mouse kidney against GFP, HMGCS2, and LTL (proximal tubule).

(D) Ggt1-Cre;MITO-Tag and Ggt1-Cre;Hmgcs2^fl/fl;MITO-Tag mice were either fed ad libitum or fasted for 24 hours. Protein lysate of immuno-isolated mitochondria using anti-hemagglutinin (HA) beads were immunoblotted for HMGCS2.

(E-G) Ggt1-Cre;MITO-Tag and Ggt1-Cre;Hmgcs2^fl/fl;MITO-Tag mice were either kept on standard chow or placed on a ketogenic diet (KD) for 4 days, then underwent unilateral IRI. Anti-HA immuno-isolated mitochondria were isolated and ADP-stimulated oxygen consumption in the presence of palmitoylcarnitine was measured using Seahorse, n=7/group.

(E) Experimental overview for (F,G). Created in BioRender. Huen, S. (2025) https://BioRender.com/ll4i1q1

(F) ADP-stimulated oxygen consumption in the presence of palmitoylcarnitine in mice on chow diet.

(G) ADP-stimulated oxygen consumption in the presence of palmitoylcarnitine in mice on chow diet compared to KD.

Data expressed as mean ± SD; two-way ANOVA with Sidak’s multiple comparison (A,F,G)

As single-cell transcriptional analysis of the response to kidney IRI identified a strong induction of lipid metabolism, including fatty acid oxidation²¹, we next evaluated whether loss of renal HMGCS2 affects fatty acid oxidation. HMGCS2 localizes to the mitochondria, which is also the site of fatty acid oxidation. To account for renal metabolic heterogeneity and proximal tubule-specific HMGCS2 expression, we generated a mouse model to isolate proximal tubule-specific mitochondria. We utilized the MITO-Tag mouse¹⁴, which when bred onto the Ggt1-Cre mouse will result in proximal tubular-specific hemagglutinin expression on the mitochondrial outer membrane (Figure 5B). Ggt1-Cre;MITO-Tag (or Ggt-MT) kidneys showed colocalization of MITO-Tag proximal tubule cells with HMGCS2 expression during fasting (Figure 5C). Anti-hemagglutinin immunoprecipitation with Ggt-MT kidneys specifically isolates viable and respirating proximal tubular mitochondria²². To specifically determine the role of proximal tubular HMGCS2 in fatty acid oxidation, we generated mice with both conditional Hmgcs2 deletion and MITO-Tag expression in the proximal tubule (Ggt1-Cre;Hmgcs2^fl/fl;MITO-Tag, Ggt^Hmgcs2KO-MT). Although the Ggt1-Cre;Hmgcs2^fl/fl mouse did not result in complete deletion of renal Hmgcs2 (Supplemental Figure 7), the Ggt^Hmgcs2KO-MT model can specifically isolate proximal tubular mitochondria lacking HMGCS2 (Figure 5D). Next, we subjected Ggt-MT (WT) and Ggt^Hmgcs2KO-MT (KO) mice to unilateral IRI, then immuno-isolated proximal tubule-specific mitochondria from the kidneys 24 hours later (Figure 5E). Using palmitoylcarnitine as a substrate, the fatty acid oxidation capacity of the immuno-isolated mitochondria was measured by Seahorse. Proximal tubular mitochondria isolated from IRI kidneys demonstrated a lower fatty acid oxidation capacity compared to those isolated from the contralateral uninjured control. Notably, proximal tubular mitochondria lacking HMGCS2 from Ggt^Hmgcs2KO-MT kidneys demonstrated significantly lower fatty acid oxidation capacity compared to WT proximal tubular mitochondria in both the uninjured and injured kidneys (Figure 5F). This difference in fatty acid oxidation capacity was independent of mitochondrial abundance or morphology, and independent of expression of fatty acid oxidation regulators such as carnitine palmitoyltransferase 1A (CPT1A) and peroxisome proliferator-activated receptor gamma coactivator 1-alpha (PGC1A), as well as expression of downstream PGC1A target genes known to be involved in oxidative phosphorylation (Supplemental Figure 2C–H). As the ketogenic diet mitigated the increased kidney injury in Six2^Hmgcs2KO mice, we next assessed whether the ketogenic diet could rescue the fatty acid oxidation defect in proximal tubular mitochondria lacking HMGCS2. Strikingly, pre-treatment with a ketogenic diet increased fatty acid oxidation capacity in proximal tubular mitochondria lacking HMGCS2 in both uninjured and injured kidneys, while the ketogenic diet only had a non-significant effect on WT mitochondria from injured kidneys (Figure 5G).

Loss of renal HMGCS2 was associated with decreased renal prostaglandin content

Given the increase in lipid accumulation in the kidney, we assessed the lipid content of the kidney. While global untargeted kidney lipidomics revealed no significant effects of treatment, genotype or interaction between genotype and condition (sham vs IRI) (Supplemental Figure 3C,D, Supplemental Table 6), pairwise comparisons between Six2^Hmgcs2KO IRI and WT IRI kidneys revealed decreases in several phospholipid species, many of which contained arachidonic acid (20:4), the precursor for prostaglandin synthesis (Figure 6A, Supplemental Table 7). As PGC1a has been suggested to promote ketone and prostaglandin production to limit ischemic injury²³, we next assessed the effect of renal HMGCS2 deficiency on prostaglandins after IRI. In sham conditions and after IRI, kidneys that lack HMGCS2 had significantly lower levels of several prostaglandins (PGD2, bicyclo PGE2, PGF2a) and arachidonic acid (Figure 6B, Supplemental Figure 8). Notably, levels of PGE2, which is known to be renoprotective^{23, 24}, were not significantly different in sham kidneys, but were significantly lower in ischemic Six2^Hmgcs2KO kidneys.

Figure 6. Loss of renal ketogenesis resulted in decreased renal phospholipids containing arachidonic acid and kidney prostaglandin content.

(A-B) Six2^Hmgcs2KO (KO) and WT mice underwent bilateral IRI or sham surgery.

(A) Lipidomics performed on kidney cortex tissue. Phospholipids containing arachidonic acid (20:4) abundances shown normalized to total ion count (TIC), n=4/group.

(B) Eicosanoid prostaglandin measurement by LC-MS/MS performed on kidney cortex tissue.

(C-D) Six2^Hmgcs2KO (K-KO), Hmgcs2^fl/fl (WT) mice were either kept on standard chow or placed on a ketogenic diet for 4 days, then underwent contralateral nephrectomy and unilateral IRI or sham surgery. Kidney tissue harvested 24 hours later, n=6/group.

(C) Whole kidney lysate immunoblotted for PLA2.

(D) Densitometry of (C) relative to total protein.

Data expressed as mean ± SD, two-way ANOVA with Sidak’s multiple comparisons (A, B, D).

To further investigate the mechanism underlying reduced prostaglandin levels in HMGCS2-deficient kidneys, we examined phospholipase A2 (PLA2) expression, the key enzyme responsible for liberating arachidonic acid from phospholipids. Interestingly, PLA2 expression was increased in kidneys of mice on a ketogenic diet compared to chow (Figure 6C,D). However, there was no significant difference in PLA2 expression between WT and Six2^Hmgcs2KO kidneys on either diet, suggesting that reduced prostaglandins in HMGCS2-deficient kidneys were unlikely due to impaired PLA2 expression.

Discussion

Coordination of metabolic programs in the kidney is critical for sustaining kidney function. While AKI and CKD are associated with diminished metabolic capacities, it is unclear whether the loss of these metabolic functions is primarily a consequence of kidney disease or also independently causative of kidney disease. Renal gluconeogenesis is a significant contribution to systemic fasting responses to maintain glycemia^{25, 26} and occurs during ammoniagenesis to maintain systemic acid-base balance²⁷. Both AKI and CKD are associated with altered renal gluconeogenesis^{28, 29}. Moreover, tubular deletion of the gluconeogenic enzyme phosphoenolpyruvate carboxykinase 1 (Pck1) leads to a defect in mitochondrial respiration and decreased kidney function³⁰. In contrast to gluconeogenesis, we previously found that the expression of the ketogenic enzyme HMGCS2 in the proximal tubule does not contribute to systemic ketosis. However, similar to gluconeogenesis, decreased renal expression of the ketogenic enzyme HMGCS2 is also associated with CKD (Table 1), suggesting an additional stress-sensitive metabolic program within the kidney.

In response to surgical stress, the proximal tubule increased HMGCS2 expression. HMGCS2 protein abundance in the kidney in both sham and IRI kidneys was comparable to that seen in fasting. Although decreased food intake after surgical stress and IRI could be contributing factors, we found that the administration of glucose sufficient to increase circulating insulin levels only partially decreased HMGCS2 expression in both sham and IRI kidneys. While the amount of glucose administered may not have been sufficient to fully suppress HMGCS2 expression via insulin signaling, it is possible that additional factors regulate renal HMGCS2 expression in response to surgical stress. Furthermore, it remains unclear why proximal tubular HMGCS2 was induced by both sham surgery and IRI surgery.

Interestingly, while the HMGCS2 protein abundance was similar between sham and IRI wild-type kidneys, the kidney β-hydroxybutyrate content was higher in the IRI kidney compared to the sham kidney, indicating enhanced local ketone production. This increased β-hydroxybutyrate content in the IRI kidney was not seen in Six2^Hmgcs2KO kidneys. The mechanism by which IRI increased kidney ketogenesis remains to be determined. Many factors such as potential post-translational modifications, substrate availability, hormonal signaling, and cellular energy demand could differentially regulate HMGCS2 activity³. Future work is needed to determine the mechanisms by which renal HMGCS2 activity is differentially regulated to generate more renal ketones in IRI than in sham conditions despite similar HMGCS2 protein abundance (Figure 1A, Supplemental Figure 5B).

Kidney ketone content is composed of both liver-derived and kidney-derived β-hydroxybutyrate. As it is difficult to differentiate liver-derived and kidney-derived β-hydroxybutyrate in the kidney, we used Alb^Hmgcs2KO mice to determine the role of liver-derived ketones. Hepatic ketogenic deficient Alb^Hmgcs2KO mice with preserved renal HMGCS2 did not demonstrate increased susceptibility to IRI injury. Thus, protection provided by renal ketogenesis did not require hepatic-derived ketones, supporting an intrinsic renal response. Although kidney β-hydroxybutyrate levels were only slightly and non-significantly increased after IRI in Alb^Hmgcs2KO mice, this was largely due to the loss of plasma liver-derived contribution to kidney ketone content. Thus, the overall abundance of kidney-derived ketones was small compared to liver-derived ketones. Moreover, succinyl-coenzyme A (CoA):3-ketoacid CoA transferase, the rate-limiting enzyme for ketolysis, is highly abundant in the kidney^{31, 32} and may rapidly metabolize locally produced ketones, further limiting detectable changes in whole tissue ketone measurements.

Local kidney ketogenesis in response to IRI was further magnified in the context of a ketogenic diet. Despite having similarly high circulating ketones in both sham and IRI mice on a ketogenic diet, the β-hydroxybutyrate content in IRI kidneys was considerably higher than in sham kidneys. Conversely, the β-hydroxybutyrate content in Six2^Hmgcs2KO IRI kidneys on the ketogenic diet was comparable to both WT and Six2^Hmgcs2KO sham kidneys, likely reflecting mostly plasma-derived β-hydroxybutyrate. Moreover, Alb^Hmgcs2KO mice which could not develop ketosis on a ketogenic diet also demonstrated higher kidney ketone content in IRI kidneys compared to sham kidneys. These data support a model whereby IRI induces HMGCS2-dependent intrarenal ketogenesis.

The renoprotective effects of the ketogenic diet and β-hydroxybutyrate administration have been previously reported^{18, 33}. While we were unable to reproduce the model of β-hydroxybutyrate delivery by minipump as previously described³³ (data not shown), we found the ketogenic diet to be effective in generating sustained ketosis and protection from IRI. Six2^Hmgcs2KO mice which have intact hepatic HMGCS2, developed significant ketosis on the ketogenic diet comparable to their wild-type counterparts. While the levels of hepatic-derived circulating ketones in mice on standard chow were insufficient to rescue the lack of kidney ketogenesis, raising the kidney β-hydroxybutyrate content with systemic ketosis diminished kidney injury in Six2^Hmgcs2KO mice. These data indicated that in the absence of endogenous renal ketogenesis, maintaining renal ketone body levels was sufficient for renoprotection, regardless of the source. These findings suggest that patients with renal ketogenic deficiency may benefit from a ketogenic diet and avoiding excessive carbohydrate exposures to enhance hepatic ketogenesis and promote ketosis, maximizing kidney ketone content. Additional studies are needed to determine whether these preclinical findings will translate to patients with AKI.

We show here that the loss of renal HMGCS2 impaired fatty acid oxidation and in the context of ischemic kidney disease, resulted in increased tubular damage. Dysfunctional fatty acid oxidation in ketogenic deficient kidneys likely contributed to decreased lipid utilization and the observed increase in lipid droplets in Six2^Hmgcs2KO kidneys after IRI. Excessive lipid accumulation can lead to lipotoxicity resulting in tubular damage³⁴. As fatty acid oxidation is upstream of ketogenesis, a defect in fatty acid oxidation due to the loss of renal ketogenesis is unexpected. However, consistent with our findings, recent studies by Queathem et al showed that hepatic ketogenic deficiency impairs hepatic fatty acid oxidation³⁵. Queathem et al proposed that free coenzyme A trapping or reduced mitochondrial matrix redox potential in the setting of ketogenic insufficiency could contribute to a defect in fatty acid oxidation. However, in the context of ketogenic deficiency, the decrease in fatty acid oxidation does not fully explain the development of fatty liver disease. Similarly, the specific mechanism by which tubular HMGCS2 and ketogenesis supports kidney fatty acid oxidation remains unknown. While we did not assess free CoA levels, acetyl-CoA levels were decreased after IRI in both wild-type and renal ketogenic deficient kidneys (Supplemental Figure 3F). In addition, the overall redox potential as assessed by NADH/NAD⁺ ratios were not significantly different (Supplemental Figure 3G). Fatty acid oxidation defects in HMGCS2-deficient proximal tubular mitochondria were rescued by a ketogenic diet, indicating that increased intrarenal ketone content can support mitochondrial fatty acid oxidation capacity. Future work is needed to determine how increased kidney ketone content modifies mitochondrial fatty acid oxidation capacity independent of mitochondrial structure and morphology, and whether the impairment of fatty acid oxidation is the primary driver of kidney injury in the context of renal ketogenic deficiency.

Finally, the relationship between ketone production, fatty acid oxidation, and prostaglandin levels remains to be determined (Supplemental Figure 9). The depletion of arachidonic acid–containing phospholipids in ketogenic deficient kidneys supports the hypothesis that renal ketogenesis may be involved in maintaining substrate availability for prostaglandin synthesis. In Six2^Hmgcs2KO kidneys, free arachidonic acid levels were lower in both sham and IRI suggesting that renal ketogenesis may be involved in regulating arachidonic acid synthesis from linoleic acid at homeostasis. After IRI, PGE2, a prostaglandin known to be renoprotective, became notably lower in ketogenic deficient kidneys. Recent work on hepatic ketone metabolism suggests that hepatic ketogenesis supports the synthesis of polyunsaturated fatty acids such as arachidonic acid, the precursor of prostaglandins, through acetoacetate activation by cytosolic acetoacetyl-coenzyme A synthetase³⁶. There is also evidence that fatty acid oxidation itself could increase mitochondrial membrane phospholipid content, including arachidonic acid containing cardiolipin³⁷. It remains to be determined whether restoring arachidonic acid containing phospholipids or prostaglandin levels can attenuate the increased kidney injury in Six2^Hmgcs2KO mice. As changes in dietary lipid composition can remodel the membrane lipidome³⁸, future work will determine whether the ketogenic diet alone can remodel the phospholipid content in the kidney, including arachidonic acid containing phospholipids. While renal PLA2 expression was not different between wild-type and Six2^Hmgcs2KO kidneys, the increase in PLA2 expression in the ketogenic diet indicated the possibility that an increase in kidney ketone content could enhance the capacity for arachidonic acid liberation from phospholipids and incorporation into prostaglandins independent of renal HMGCS2. Future investigation is needed to determine whether the renoprotection of a ketogenic diet in Six2^Hmgcs2KO kidneys is directly due to the increase in kidney ketone content from hepatic-derived circulating ketones and/or the upregulation of kidney PLA2 expression. Whether the effect of renal ketogenesis on fatty acid oxidation and arachidonic acid metabolism are directly related remains unknown. Both arachidonic acid and prostaglandins have been shown to have wide ranging effects on mitochondrial function and metabolism^39–43, suggesting the possibility that local renal ketogenesis could limit kidney injury after ischemic injury by supporting the synthesis of arachidonic acid and downstream eicosanoids such as prostaglandins.

Limitations.

First, our study used an IRI model which causes mild injury in wild-type mice to highlight the increased susceptibility of kidney Hmgcs2 KO mice to kidney injury. As the injury model is mild, we did not observe any remarkable changes in mitochondrial morphology which is typically observed in more severe renal IRI models. The increased kidney injury in Six2^Hmgcs2KO kidneys after mild IRI suggests a scenario in which mitochondrial functional dysfunction such as fatty acid oxidation impairment could occur without overt mitochondrial structural damage or changes in mitochondrial abundance. Future work is needed to assess the role of renal ketogenesis in the context of a more severe kidney injury and at later time points after IRI. Second, to assess fatty acid oxidation, we used a model capable of isolating proximal tubule specific mitochondria and used palmitoylcarnitine as a substrate. The advantage of this approach focuses on the analysis of proximal tubule specific mitochondria which express HMGCS2 and directly on fatty acid oxidation. However, this approach does not address whole cell fatty acid metabolic flux including fatty acid uptake. Future studies are needed with isolated proximal tubules to address the upstream metabolic steps involved in regulating fatty acid oxidation. Third, the ketogenic diet induces many metabolic changes in addition to increased ketone production, including decreased insulin production and signaling, as well as changes in lipid metabolism. Thus, the protection provided by the ketogenic diet may not be exclusively ketone dependent. Fourth, the screening of untargeted metabolomics and lipidomics data were used for hypothesis generation and are not mechanistically conclusive. β-hydroxybutyrate was validated with an alternative means of quantitative measurement. While we did not quantitatively validate the arachidonic acid containing phospholipids, quantitative prostaglandin measurements were performed with internal standards. Additional investigation is required to determine whether renal HMGCS2 expression and/or renal ketogenesis are directly related to the abundance of arachidonic acid-containing phospholipids and subsequent prostaglandin synthesis. Finally, our studies focused on β-hydroxybutyrate and did not distinguish the roles of acetoacetate and β-hydroxybutyrate.

Conclusion.

The ketogenic enzyme HMGCS2 is expressed in the proximal tubule during states of stress. Loss of tubular HMGCS2 resulted in proximal tubular ketogenic deficiency, impaired renal fatty acid oxidation, reduced renal prostaglandin content, and increased susceptibility to ischemic kidney injury. Thus, tubular HMGCS2 deficiency may not only be a biomarker of kidney disease but could also participate in acute and chronic kidney disease pathogenesis.

Supplementary Material

1. Supplemental Methods

2. Supplemental Figure 1. HMGCS2 expression in kidney ischemia/reperfusion

3. Supplemental Figure 2. Inflammation, oxidative stress, and Ppargc1a gene expression in renal ischemic injury.

4. Supplemental Figure 3. Kidney metabolomics and lipidomics after ischemic injury

5. Supplemental Figure 4. Kidney ketone content normalized for plasma ketone levels.

6. Supplemental Figure 5. Effect of ketogenic diet on renal HMGCS2 expression

7. Supplemental Figure 6. Loss of renal HMGCS2 did not alter ketone-related post-translational modifications.

8. Supplemental Figure 7. Ggt1-Cre;Hmgcs2^fl/fl mouse model kidney HMGCS2 protein expression.

9. Supplemental Figure 8. Kidney prostaglandin content.

10. Supplemental Figure 9. Working model of the effects of kidney HMGCS2 deficiency on renal ketogenesis, fatty acid oxidation, and prostaglandin synthesis.

11. Supplemental Table 1. Quantitative RT-PCR primers

12. Supplemental Table 2. Antibody and staining reagents.

13. Supplemental Table 3. Kidney metabolomics data (file)

14. Supplemental Table 4. Kidney metabolomics ANOVA results (file)

15. Supplemental Table 5. Kidney metabolomics comparison KO IRI vs WT IRI (file)

16. Supplemental Table 6. Kidney lipidomics data (file)

17. Supplemental Table 7. Kidney lipidomics comparison KO IRI vs WT IRI (file)

18. Supplemental References

Key Points:

1. Proximal tubular expression of HMGCS2 during ischemic injury increases kidney ketone and prostaglandin content.

2. Kidney HMGCS2 deficiency emerged as both a biomarker and contributor to kidney disease.

3. Renal ketogenesis supported the maintenance of proximal tubular fatty acid oxidation.