Early Life Exposure to Maternal Insulin Resistance Has Persistent Effects on Hepatic NAFLD in Juvenile Nonhuman Primates

Nonalcoholic fatty liver disease (NAFLD) is the most common chronic liver disease in children and adults (1,2). The clinical discovery of NAFLD in children has led to speculation that its origins may lie in early intrauterine exposures. Indeed, human infants born to obese and gestational diabetic mothers have increased liver fat compared with infants of normal mothers (3,4). While the epidemiologic association between increased maternal BMI and early childhood metabolic diseases has been well described (5–7), the mechanisms remain unclear. Studies in rodents have shown that maternal high-fat (HF) diet and insulin resistance during pregnancy leads to offspring obesity and hepatic lipid accumulation (8–11). However, rodent studies are limited in their ability to explain whether and how differences in maternal metabolism and intrauterine exposures may affect human development.

To study the effect of maternal HF diet and maternal phenotype on the offspring, we have used a nonhuman primate (NHP) model where female Japanese macaques are maintained on a chronic HF diet (12–15). We took advantage of the fact that some females on a HF diet become insulin resistant (IR; HF+IR) while other females remain insulin sensitive (IS; HF+IS) compared with control (CON) females. We previously showed that all maternal HF diet–exposed fetuses had increased liver triglyceride content (12,14). Further, when mothers on the HF diet were switched to a CON diet during the next pregnancy, fetal hepatic triglycerides were reduced (12). These findings demonstrate that intrauterine exposure to maternal HF diet, independent of maternal metabolic effects, increases fetal hepatic steatosis. Whether this is reversible postnatally is unknown but has important clinical implications for understanding the pathophysiology of NAFLD since liver damage in utero may prime the liver for the progression of metabolic disease.

In this study, we hypothesized that exposure to maternal HF diet during pregnancy through weaning would produce long-lasting effects on the offspring liver and may induce an inflammatory response in addition to steatosis noted earlier in third-trimester fetuses. Further, we tested if switching juvenile offspring exposed to maternal HF diet to a CON diet after weaning might alleviate the adverse consequences caused by exposure to maternal HF diet. Our results suggest that regardless of postweaning diet, juvenile NHP offspring from HF+IR, but not HF+IS, mothers develop increased hepatic de novo fatty acid synthesis and increased activated hepatic macrophages in the absence of obesity. Juvenile body weight, adiposity, and early signs of insulin resistance increased modestly on the postweaning HF diet, yet liver triglycerides and inflammation were not increased further. These data demonstrate that early exposure to a maternal HF+IR environment provokes the development of juvenile NAFLD that is irreversible in nondiabetic, preobese juvenile NHP.

All animal procedures were in accordance with the guidelines of the Institutional Animal Care and Use Committee of the Oregon National Primate Research Center. Adult female Japanese macaques were maintained on a CON (14.6% calories from fat) or HF (36.6% calories from fat) diet for 1–5 years prior to pregnancy (12–15). Body weight and baseline insulin, glucose, and triglyceride measurements were collected, and intravenous glucose tolerance tests (GTTs) were performed during the third trimester of each pregnancy (12). The area under the curve (AUC) for glucose and insulin was calculated from zero. Mothers were classified during pregnancy as HF+IR if they had an insulin AUC greater than two SDs above the mean of the CON mothers (Fig. 1A). Mothers with insulin AUC less than two SDs relative to the CON group were classified as HF+IS (12,13).

Offspring born to CON, HF+IS, and HF+IR dams were maintained with their mothers until weaning (7–8 months). At weaning, CON, HF+IS, and HF+IR offspring were placed on a CON or HF diet (13,15) (Fig. 1B). Body composition analysis (DEXA) was performed at 10 and 12 months of age (13). Daily growth and lean and fat mass (gain/loss) rates were calculated. At 1 year of age, juvenile animals were killed (13,15). Plasma glucose, insulin, nonesterified fatty acids, triglycerides, and cytokine concentrations were measured (12,13,16). HOMA-IR was calculated (13).

Liver triglyceride content was measured (Infinity Triglyceride Reagent) following lipid extraction (12). In a subset of offspring, lipid mass spectrometry fatty acid profiling was performed to measure triglyceride, diacylglyceride, and phospholipid concentrations and fatty acid composition (17).

RNA was extracted from liver and omental adipose tissue samples, and gene expression was measured using real-time PCR (LightCycler 480, Roche) as described (18) (see Supplementary Table 1 for primers). Results were normalized to 18s rRNA.

Primary hepatocytes and hepatic macrophages were isolated from a subset of offspring (18). Hepatocytes were plated in complete DMEM (5.5 mmol/L glucose supplemented with 2 mmol/L glutamine, 2 mmol/L lactate, 1 mmol/L pyruvate, 1× nonessential amino acids, 100 units/mL penicillin–streptomycin, 1 nmol/L insulin, 100 nmol/L dexamethasone, and 10% FBS) on collagen-coated plates. After a 4-h attachment period, cells were washed and media was replaced with serum-free DMEM plus 0.2% BSA (SF) with or without 1 nmol/L insulin and 100 nmol/L dexamethasone, as described below.

For hepatic macrophage isolations, the supernatant from the hepatocyte isolation containing nonparenchymal cells was filtered and spun at 300× g to pellet nonparenchymal cells. Cells were resuspended in Histodenz (24%), and gradients were prepared and spun at 1,500× g for 20 min (19). Cells at the interface were collected, washed, and plated on tissue culture plates in complete DMEM with 55 μmol/L 2-mercaptoethanol. After 1–2 h attachment, nonadherent cells were aspirated and media was replaced on adherent cells, representing hepatic macrophages (20).

Hepatocytes incubated in SF media and hepatic macrophages incubated in complete media with 2-mercaptoethanol were studied 24 h after isolation. Cells were incubated with SF media (basal), lipopolysaccharide (LPS; 100 ng/mL), or palmitate (PALM; 200 nmol/L) for 3 h of treatment and then harvested for RNA isolation and gene expression.

Lipogenic gene expression was measured 24 h after isolation following 8 h of treatment with SF media (basal, 5.5 mmol/L glucose) or high-nutrient media (SF media + 25 mmol/L glucose, 500 nmol/L dexamethasone, 100 nmol/L insulin). Cells were collected for RNA isolation and analyzed for gene expression.

Maternal data were analyzed by mixed-model ANOVA with fixed effect of maternal diet (CON, HF+IS, HF+IR) and random effect of mother using SAS (PROC MIXED). Offspring data were analyzed by two-way ANOVA with fixed effects of maternal diet, postweaning diet, and interaction. Both models accounted for the inequality of variances between maternal groups. Individual posttest comparisons were made for comparisons of interest and when the interaction was significant (P < 0.15). Isolated cell data were analyzed similarly by one- or two-way ANOVA. Statistical significance was declared at P < 0.05 and P < 0.15 for interactions.

Using GTTs, we identified 25 pregnant females on the chronic HF diet demonstrating insulin resistance (HF+IR) (Fig. 1A). The remaining 10 females on the HF diet were IS (HF+IS). Maternal body weight was 20–30% higher in HF+IR mothers compared with HF+IS and CON mothers (Table 1). Fasting insulin concentrations were increased fourfold in HF+IR and 50% in HF+IS mothers compared with CON. Fasting glucose concentrations and glucose AUC during the GTT were similar between HF+IR and HF+IS mothers. Fasting triglyceride concentrations were increased in HF+IR mothers compared with HF+IS or CON mothers. Thus some mothers on the HF diet (HF+IR) have insulin resistance and higher triglycerides yet remain glucose tolerant.

Offspring from CON, HF+IS, and HF+IR females were maintained on their respective maternal diets or were switched to a CON or HF diet at weaning until 1 year of age (Fig. 1B). The postweaning HF diet had a modest effect on body weight and weight gain (Table 2). Offspring on the postweaning HF diet gained more weight per day compared with offspring on the CON postweaning diet. This was primarily due to increased lean rather than fat mass accretion. Retroperitoneal adipose mass tended to be increased in animals consuming the postweaning HF diet, independent of maternal group; however, body fat was low for all offspring at only 1–5% (Table 2).

Fasting plasma glucose, glycerol, triglyceride, and nonesterified fatty acid concentrations in juvenile offspring were similar among all groups (Table 2). Offspring fasting insulin concentrations and HOMA-IR calculations were twofold higher, and glucose:insulin ratios were 50% lower on the postweaning HF diet, with HF+IR/HF offspring having the highest insulin concentrations. Notably, HOMA-IR was improved in HF+IR/CON offspring, suggesting that insulin resistance was reversed by switching to the CON diet at weaning. Thus the postweaning HF diet had modest effects on inducing insulin resistance and increasing body weight.

Liver triglyceride content was 50% higher in animals from HF+IR mothers compared with offspring from CON and HF+IS mothers, regardless of postweaning diet (Fig. 2A). Similarly, cytoplasmic lipid droplets were visible in livers from HF+IR offspring even on a CON postweaning diet compared with the offspring livers from HF+IS or CON mothers (Fig. 2B). Likewise, hepatic diacylglycerides were higher in juveniles born to HF+IR mothers compared with CON offspring regardless of postweaning diet (Table 2). These data strongly implicate maternal metabolic phenotype as the primary driver of persistent juvenile hepatic lipid content.

Fatty acid profiling was performed to determine whether maternal or postweaning diet affected lipid composition in the juvenile liver. The composition of fatty acids in hepatic triglycerides and diacylglycerides (Supplementary Table 2) reflected the composition of the postweaning diet (14). Relative saturated and monounsaturated fatty acids were increased, polyunsaturated fatty acids were decreased, and the ratio of n-6:n-3 fatty acids was increased in hepatic triglycerides and diacylglycerides on the postweaning HF diet (Supplementary Table 2). Increased diacylglyceride saturation has been implicated in insulin resistance (17,21–23); however, animals from HF+IR mothers had lower total diacylglyceride saturation compared with juvenile livers from HF+IS and CON mothers (Supplementary Table 2). Hepatic phospholipids were increased by 20% in all offspring on the postweaning HF diet compared with the CON diet (Table 2). Thus hepatic lipid composition was primarily reliant on diet and occurred independently of maternal phenotype.

To investigate transcriptional mechanisms for hepatic steatosis in HF+IR offspring, we analyzed hepatic gene expression (Fig. 2C and D). Expression of the lipogenic transcription factor SREBP1 was twofold higher in the offspring of HF+IR mothers compared with HF+IS and CON offspring. The SREBP1 and insulin target genes for lipogenesis, FAS and ACC1, were also increased in HF+IR offspring compared with CON and HF+IS offspring (Fig. 2C). Other genes in the glycerol phosphate pathway for de novo diacylglyceride and triglyceride synthesis (24), including DGAT1, LPIN1, GPAT1, and AGPAT1, were increased in HF+IR offspring, regardless of postweaning diet (Fig. 2C). Expression of the fatty acid transporter CD36 was increased both by maternal and postweaning HF diet (Fig. 2C), suggesting increased lipid uptake. Expression of the lipid export gene APOB was also higher in HF+IR juvenile livers (Fig. 2C), which may reflect a compensatory mechanism in response to increased hepatic lipids.

We next determined whether the responsiveness to lipogenic stimuli was increased in isolated hepatocytes from juvenile animals born to HF+IR mothers. Hepatocytes were treated with high glucose, insulin, and dexamethasone. Since juveniles born to HF+IR mothers have increased hepatic lipids even on a CON postweaning diet, we studied hepatocytes from these offspring and CON/CON offspring. Compared with stimulated hepatocytes from CON offspring, stimulated HF+IR/CON hepatocytes had a greater increase in SREBP1, SREBP2, FAS, ACC1, DGAT1, and LPIN1 expression (Fig. 2D). Thus maternal insulin resistance has a long-lasting effect on juvenile liver lipogenic capacity that persisted in isolated hepatocytes and despite a healthy diet intervention after weaning.

HF+IR mothers may confer an effect on the adaptive immune system in the offspring liver that plays a role in the evolution of NAFLD from simple steatosis to nonalcoholic steatohepatitis. To explore this, we characterized inflammatory gene expression in the juvenile livers. In whole liver tissue, juvenile offspring born to HF+IR mothers had increased expression for CD45, a marker for immune cells; CD11B and CD68, both canonical macrophage markers; and CD1D expression, a marker for natural killer T (NKT) cells, suggestive of increased immune cell numbers (Fig. 3). Expression of these immune cell markers persisted despite switching to a healthy CON diet postweaning. Although there was no change in expression of MCP1, we did find increased expression of the MCP1 chemokine receptor, CCR2, in HF+IR livers, suggesting potential for increased macrophage recruitment (25).

Expression of the proinflammatory cytokines TNFA and IL1B were higher in the liver from HF+IR offspring compared with HF+IS or CON/CON offspring (Fig. 3). Intriguingly, TLR4 expression, an innate immune response receptor, was increased by 50% in the HF+IR juvenile liver compared with CON and HF+IS offspring and did not increase further with postweaning HF diet (Fig. 3). Expression of caspase 1 (CASP1) and SOCS3 was also increased in HF+IR liver, consistent with increased cytokine activation. Expression of SERCA2, a marker of ER stress, was increased in HF+IR livers (Fig. 3). Thus juvenile livers from HF+IR mothers showed increased activation of the innate immune system and classical macrophage activation as described in human NAFLD (26,27).

We also found evidence for alternatively activated macrophage pathways in the liver from animals born to HF+IR mothers (Fig. 3). Expression of the interleukin (IL)-4 receptor (IL4R) was increased in the HF+IR liver, suggesting increased responsiveness to IL-4, which induces alternatively activated macrophages (28). We also found increased expression of PPARG, PPARD, KLF4, IRF4, and JMJD3, factors promoting alternatively activated macrophages (28), in HF+IR offspring. We conclude that there was increased innate immune activation and alternative macrophage activation in the juveniles born to HF+IR mothers, independent of postweaning dietary manipulation.

Given the gene expression pattern in whole liver tissue suggesting increased macrophage activation, we wanted to determine the specific activation phenotype of hepatic macrophages to inflammatory signals. Hepatic macrophages and hepatocytes were isolated from the livers of CON/CON and HF+IR/CON juvenile offspring. Hepatic macrophages had higher expression of CD45, CD11B, CD68, TLR4 (Fig. 4A, Supplementary Table 3), IL1B, TNFA, IL6, and MCP1 (compare Fig. 4B and C) compared with hepatocytes. Moreover, hepatic macrophages from HF+IR/CON offspring had higher basal expression of CD45, CD68, and TLR4 compared with CON/CON cells (Fig. 4A) and tended to have higher expression of JMJD3, KLF4, PPARD, PPARG, CCR2 (Fig. 4A), and IL4R (Fig. 4B). These results suggest that offspring born to HF+IR mothers and weaned to a healthy diet have increased recruitment (CCR2), increased innate activation (TLR4), and an alternative activation macrophage phenotype (PPARG, PPARD, IL4R) (25,28).

Next we challenged hepatic macrophages in vitro with PALM or LPS to see whether maternal HF diet increased their responsiveness to stimuli. All cells had a robust increase in expression of the proinflammatory cytokines IL1B, TNFA, IL6, and MCP1 (Fig. 4B). Importantly, this increase in IL1B and TNFA was greater in hepatic macrophages from HF+IR compared with CON offspring, consistent with increased TLR4 expression in these cells (Fig. 4A). Expression of the cytokine signaling suppressor, SOCS3, tended to be higher in HF+IR/CON macrophages (Fig. 4B). Interestingly, in response to PALM and LPS, hepatic macrophages from HF+IR/CON juveniles had a fivefold increase in IL4R expression (Fig. 4B), suggesting increased alternative macrophage activation. These results suggest that maternal insulin resistance programs an inflammatory response that includes a combination of classic and alternative activation in hepatic macrophages.

In contrast to macrophages, PALM treatment in hepatocytes had a small effect on inducing IL1B and TNFA expression and lack of effect on IL6, MCP1, SOCS3, and IL4R expression, regardless of maternal diet group (Fig. 4C). In both CON/CON and HF+IR/CON hepatocytes, LPS, relative to PALM, elicited a greater inflammatory response as evidenced by increased expression of IL1B, TNFA, IL6, MCP1, and SOCS3; however, the fold induction was much less in hepatocytes compared with macrophages (compare Fig. 4B and C). These data support that exposure to the maternal HF+IR phenotype may prime hepatic macrophages, more so than hepatocytes, toward increased responsiveness to proinflammatory stimuli.

We measured adipose tissue inflammatory gene expression and circulating cytokines to assess systemic inflammation. In adipose tissue, expression of inflammation genes was not affected by maternal or postweaning diet (Table 3). Plasma cytokine concentrations were also not affected by maternal or postweaning diet in juvenile offspring (Table 3). These data suggest that maternal insulin resistance primes the liver for proinflammatory activation of resident hepatic macrophages, in the absence of adipocyte inflammation.

Maternal insulin AUC during GTT, body weight, and plasma triglycerides were all positively related to offspring liver triglycerides (Fig. 5A). Interestingly, in the HF+IS and HF+IR offspring, maternal insulin AUC correlated more closely with offspring liver inflammatory markers, including IL1B, TLR4, and CD11B (R = 0.32–0.48) (Fig. 5B), than with hepatic triglyceride content (Fig. 5A) or SREBP1, SREBP2, or FAS expression (R = 0.11–0.16) (Fig. 5C). These relationships suggest the effects of maternal metabolism have more predictive power on the offspring’s hepatic immune system than on the de novo lipogenic pathway in development of juvenile NAFLD.

We took advantage of the fact that not all NHP mothers consuming a chronic HF diet are obese, and like humans, some are significantly overweight and have exaggerated insulin resistance (e.g., HF+IR mothers) while others consuming the HF diet remain relatively metabolically normal (e.g., HF+IS mothers). Unlike juvenile offspring from HF+IS mothers, which largely resembled CON offspring, offspring from HF+IR mothers showed a pattern of hepatic gene expression, cytokines, and lipogenic responses consistent with the early stages of NAFLD. Importantly, the NAFLD phenotype persisted despite the absence of postnatal obesity, insulin resistance, or systemic or local adipose tissue inflammation. These findings indicate that maternal metabolic dysfunction programs the NAFLD phenotype and suggests that postnatal diet may be incapable of reversing or correcting this disorder in juvenile offspring.

The livers of HF+IR NHP juvenile offspring showed evidence for increased de novo lipogenic capacity. The genes SREBP1, SREBP2, FAS, ACC1, LPIN1, and DGAT1 were higher in the HF+IR offspring, despite consuming a healthy diet after weaning. Because the turnover rate of the liver triglyceride pool is ∼40 days or less in humans (29), the 5-month period where offspring are on the healthy CON postweaning diet should be sufficient for any maternally derived lipid to dissipate. Further, in isolated hepatocytes, the same lipogenic gene pattern was increased upon hormone stimulation, supporting that livers from HF+IR offspring have a gene profile favoring lipid synthesis and storage that persists postnatally. Increased hepatic steatosis has been reported in postnatal offspring in other rodent models of maternal obesity (9–11,30,31); however, most of these offspring also have increased adiposity, making the early origins of steatosis difficult to discern. We observed that in the NHP HF+IR offspring, steatosis is present when the animals are lean with no detectable systemic inflammation. This demonstrates that the mechanism(s) driving excess hepatic fat storage may be different than those underlying adipose tissue expansion and suggests that the early NAFLD phenotype in HF+IR offspring may be an early risk factor for other metabolic diseases, including cardiovascular disease, cancer, and obesity (32,33).

NAFLD is often associated with insulin resistance, and whether insulin resistance is a cause or consequence of hepatic steatosis remains debatable (22,34,35). In our juvenile offspring, continued HF diet exposure after weaning increased fasting insulin concentrations and insulin AUC during the GTT (13); however, this hyperinsulinemia did not accelerate hepatic steatosis further in the HF+IR/HF juveniles. Likewise, HF+IR/CON offspring, when switched to a healthy diet at weaning, showed a persistent increase in steatosis and de novo lipogenic gene expression, despite lower insulin levels and normal body weight compared with HF+IR/HF offspring on the postweaning HF diet. Thus neither increased insulin concentrations nor insulin resistance in the offspring explain increased de novo lipogenesis in the HF+IR juvenile offspring. Overall, this suggests that hepatic steatosis found in fetuses from HF-diet mothers is resistant to changes in diet after weaning and, therefore, in utero exposures from an HF+IR mother may lay the foundation for an “unhealthy” liver.

Inflammation is an important link in the progression of NAFLD, and steatosis is associated with increased macrophage activation (36–42). Offspring born to HF+IR mothers, regardless of postweaning diet, had increased hepatic macrophage numbers (increased CD11B and CD68 mRNA), which may reflect increased recruitment (increased CCR2 mRNA) (25). Here, hepatic macrophages in HF+IR offspring present with increased classical and alternative activation. Increased classical activation of hepatic macrophages in HF+IR offspring is supported based on increased expression of TLR4 in HF+IR macrophages and a greater proinflammatory cytokine response (IL1B, TNFA, IL6) upon treatment with PALM and LPS. Our whole liver gene expression results support this based on increased IL1B, TNFA, CASP1, and TLR4 expression in HF+IR offspring. Increased alternative activation is evidenced by increased expression of IL4R, PPARG, PPARD, IRF4, JMJD3, and KLF4. Alternative activation of macrophages involves PPARG and PPARD, which fulfill a protective function to increase fatty acid oxidation (28). We speculate that alternative activation of hepatic macrophages in HF+IR offspring serves an adaptive or reparative role in response to steatosis. However, alternative activation of macrophages has also been associated with pathological tissue remodeling and fibrosis (42). Moreover, alternative activation is found in models of insulin resistance and obesity in liver and adipose macrophages (28,37,42–44). We also found increased expression of CD1D, suggesting increased NKT cell activation, which is seen in human NAFLD (45). Thus our results support that juveniles born to HF+IR mothers develop increased classical and alternatively activated hepatic macrophages and NKT cell activation that in combination with increased de novo fatty acid synthesis, supports a NAFLD phenotype.

The mechanisms whereby excessive maternal insulin resistance programs immune cells and NAFLD in the juvenile liver are unclear. Liver Kupffer cell populations are established prior to birth and self-maintained postnatally (46). Thus fetal exposures associated with maternal insulin resistance may prime these resident immune cells, rendering them more responsive to stimulation as proinflammatory effector cells and/or differentiation into alternatively activated macrophages. Although we previously found increased fetal liver triglycerides in offspring from HF diet–fed mothers (12), we found no evidence for increased hepatic cytokines, suggesting that immune cells may not have matured fully or that the stimulus for inflammation may occur in the final month of gestation, when fetal growth accelerates. We previously demonstrated that only HF+IR mothers and their fetuses have reduced uterine and fetal blood flow. Further, 13 of 36 cytokines were significantly upregulated in the umbilical, but not maternal, circulation of HF+IR maternal–fetal pairs (16). This suggests that high maternal insulin resistance may trigger fetal/placental hypoxia (16,47) and inflammatory cytokine production. Indeed, there is increased inflammation in cord blood of infants born to obese mothers (48) and in sheep fetuses exposed to overfeeding (49). Furthermore, we recently demonstrated that lowering placental inflammation in an obese pregnant transgenic mouse model prevented NAFLD (50). Together our results suggest a model whereby maternal insulin resistance on a HF diet may produce exposure to increased transplacental lipids and hypoxia/inflammatory stimuli in the fetus that drives inflammation and subsequent juvenile NAFLD (Fig. 5C).

Importantly, since not all obese women have babies born with increased birth weight or with fatty livers (3,4,7), focusing on the metabolic conditions that provoke excessive insulin resistance during pregnancy, rather than obesity per se, could be an important new therapeutic target for lessening the transmission of metabolic dysfunction from mother to child and preventing future NAFLD in the next generation.

Acknowledgments. The authors thank Diana Takahashi, Ashley Kostrba, Sarah Comstock, and the staff of the Oregon National Primate Research Center for assistance with animal studies. The authors also thank Rachel C. Janssen (University of Colorado Denver) for assistance with editing the manuscript.

Funding. Research support was provided by the National Institutes of Health (K01-DK-090199 to S.R.T., P51-OD-011092 as partial salary support for K.L.G., and R24-DK-090964 to K.L.G. and J.E.F.), the National Institutes of Health/National Center for Advancing Translational Sciences Colorado Clinical and Translational Science Institute grant UL1-TR-000154 from the Colorado Nutrition and Obesity Research Center (NORC) (P30-DK-048520), and the Academic Enrichment Fund of the University of Colorado School of Medicine (to K.C.E.K.).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. S.R.T. designed the experiments, performed liver and cell experiments, analyzed data, wrote the manuscript, and reviewed the manuscript. K.C.B. performed liver and cell experiments, analyzed data, performed lipid profiling, and reviewed the manuscript. S.A.N. performed liver and cell experiments, analyzed data, and reviewed the manuscript. K.C.E.K. reviewed the manuscript. B.C.B. and G.I.S. performed lipid profiling and reviewed the manuscript. K.L.G. designed the experiments and reviewed the manuscript. J.E.F. designed the experiments, wrote the manuscript, and reviewed the manuscript. J.E.F. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.