Identification of genetic drivers of plasma lipoprotein size in the Diversity Outbred mouse population

Despite great progress in understanding lipoprotein physiology, there is still much to be learned about the genetic drivers of lipoprotein abundance, composition, and function. We used ion mobility spectrometry to survey 16 plasma lipoprotein subfractions in 500 Diversity Outbred mice maintained on a Western-style diet. We identified 21 quantitative trait loci (QTL) affecting lipoprotein abundance. To refine the QTL and link them to disease risk in humans, we asked if the human homologs of genes located at each QTL were associated with lipid traits in human genome-wide association studies. Integration of mouse QTL with human genome-wide association studies yielded candidate gene drivers for 18 of the 21 QTL. This approach enabled us to nominate the gene encoding the neutral ceramidase, Asah2, as a novel candidate driver at a QTL on chromosome 19 for large HDL particles (HDL-2b). To experimentally validate Asah2, we surveyed lipoproteins in Asah2−/− mice. Compared to wild-type mice, female Asah2−/− mice showed an increase in several lipoproteins, including HDL. Our results provide insights into the genetic regulation of circulating lipoproteins, as well as mechanisms by which lipoprotein subfractions may affect cardiovascular disease risk in humans.

High concentrations of LDL-C are associated with increased CVD risk.Interventions to reduce LDL-C result in improved cardiovascular outcomes.Small dense LDL particles are associated with CVD, including coronary artery disease and stroke (1), and are strong predictors of cardiovascular events (2).Likewise, larger HDL-2b particles are better predictors of coronary heart disease than small, dense HDL-3, LDL-C, or HDL-C levels (3)(4)(5).Variation in protein and lipid composition and particle size can affect lipoprotein function (6)(7)(8)(9).
Genetically diverse mouse populations, resulting from intercrossing or outcrossing inbred strains, can be used to discover the genetic drivers of lipoprotein abundance and composition.For example, intercross mouse populations identified gene loci affecting apolipoprotein (Apo) A2 (10).In an intercross study involving the RIIIS/J and 129S1/SvImJ mouse strains, eight unique loci affecting plasma cholesterol and causative genes within the loci were identified (11).Panels of recombinant inbred mouse strains have been used to leverage naturally occurring polymorphisms, which showed heterogeneity in lipoprotein size and apolipoprotein composition (12).The hybrid mouse diversity panel, a collection of 100 mouse strains, utilizes natural strain variation in a systems genetics approach to identify genetic drivers of phenotypes (13).A metaanalysis of nearly 5000 hybrid mouse diversity panel mice identified 26 significant loci associated with HDL-C.Several loci, including one for ApoA2, were consistent with previous reports, whereas other loci provided novel insights into gene-environment interactions (14).
The use of outcrossed mouse populations, including the collaborative cross (CC), brings additional genetic diversity to mouse genetic screens.In a study using 25 CC strains, increased adiposity and liver steatosis were associated with increasing total, HDL, and LDL cholesterol (15).Key genetic regulators of hepatic lipids were linked to diet-induced changes in liver steatosis severity and plasma lipid measures.Leveraging the genetic diversity of the diversity outbred (DO) mouse population, an outbred stock derived from eight founder strains of the CC, three loci were identified for plasma cholesterol (16).
In the present study, we utilized the DO mouse population to determine quantitative lipoprotein subclasses, identify quantitative trait loci (QTL) for several subclasses, and nominate candidate genetic drivers of lipoprotein particle sizes.We identified several known cholesterol-related genes, Apoa2 and Foxo1, as well as novel loci associated with plasma cholesterol.) and DO mice were purchased from Jackson Laboratories (Bar Harbor, ME) and maintained at the University of Wisconsin-Madison, as previously described (17,18).Briefly, founder mouse strains were fed standard laboratory chow (Formulab Diet 5008; LabDiets, Brentwood, MO) or given a high-fat, high-sucrose diet (HFHS; TD.08811; Envigo, Madison, WI) for 18 weeks.DO mice were maintained on the same HFHS diet as the founder mice for 16 weeks.Animals were euthanized at 22 weeks of age, and plasma was collected and stored at −80 • C.

Animal husbandry-Asah2 mice
Asah2 mice were generated by Richard Proia (National Institutes of Health [NIH]) (19) and provided to the Summers/Holland laboratory.All animal procedures were performed in compliance with the protocols approved by the Institutional Animal Care and Use Committee at the University of Utah and adhered to NIH standards.Male (n = 5-6 per genotype) and female (n = 4-7 per genotype) mice were maintained under standard laboratory conditions at a temperature of 22-24 • C, in groups of 2-5 mice, with a 12 h light/ dark cycle.Mice were allowed ad libitum access to food and water unless fasting conditions were required for experimental procedures.Animals were fed a normal chow diet from the age of 4 weeks and transitioned to an HFD (60% total energy; D12492; Research Diets, Inc, New Brunswick, NJ) at 9 weeks of age for 16 weeks.At 25 weeks of age, mice were anesthetized with isoflurane, and blood collected by cutting the brachial artery.Whole blood was collected into vacutainers coated with 20% K 2 EDTA, centrifuged at 7,500 g for 7.5 min, and plasma was separated.Plasma was stored at −80 • C.

Plasma lipoprotein fractionation by ion mobility analysis
To analyze lipoprotein class size, plasma lipoproteins were separated by ion mobility analysis as previously described (20,21).Briefly, lipoproteins were harvested on paramagnetic particles, washed to remove free salt and proteins (e.g., IgG, albumin, and transferrin), and then resuspended in 25 mM ammonium acetate.Lipoproteins were then fractionated and quantified by summing the total number of particles within specific size ranges.Supplemental Table S1 shows the lipoprotein subclasses, their size ranges, and nomenclature.

QTL mapping of plasma lipoprotein phenotypes
Mapping of plasma lipoproteins for QTL analysis was performed as previously described (17).Briefly, 478 DO mice (236 females and 242 males) were obtained from Jackson Laboratories (Bar Harbor, ME) and maintained on a HFHS diet (TD.08811;Envigo) for 16 weeks, and plasma was collected for lipoprotein sizing by ion mobility analysis.Lipoprotein phenotype data were rankZ-transformed to achieve a normal distribution prior to mapping.Genetic mapping was performed using the R/qtl2 package with kinship correction to identify QTL using the GRCm38 genome build and Ensembl 75 for gene annotation.Genome scans used sex, mouse cohort (wave), and technical batch as additive covariates.As previously described, logarithm of odds (LOD) thresholds were defined through permutation testing to establish a genomewide family wide error rate for genome-wide QTL (22,23).A LOD greater than 6.0 was used as the threshold for identifying suggestive QTL, and a LOD greater than 7.4 identified significant QTL.

Identification of candidate causal genes
To identify candidate causal genes, we explored single nucleotide polymorphisms (SNPs) at each mouse locus, significant associations in human genome-wide association studies (GWAS) to cardiometabolic traits, and phenotyping data available from the International Mouse Phenotyping Consortium (IMPC; mousephenotype.org),as well as incorporating known lipoprotein biology from the published literature.We first surveyed the SNPs at each locus in our DO mouse dataset.Within SNP plots at a single locus, we identified the region with the strongest association between individual SNPs and the phenotype by looking at regions until ∼1.5 LOD drop.For example, if SNPs have a LOD of 6.0, the significant region would span all genes with SNPs having a LOD of 4.5-6.0.For genes that were contained within these regions, we searched the IMPC resource for available phenotypes.Very few genes had live mice produced or plasma lipid phenotypes available.We also attempted to utilize singletissue expression QTL datasets, including those for adipose and liver.However, for most lipoprotein QTL, we had difficulty identifying single-tissue expression QTL with allele effect patterns that matched those of our study QTL.Therefore, we chose to employ an analogous method of surveying syntenic regions in human GWAS for significant associations to cardiometabolic traits.A 2 Mbp region flanking the lipoprotein QTL was used to identify orthologous regions in the human genome using the LiftOver utility from UCSC Genomic Institute (genome.ucsc.edu/cgi-bin/hgLift-Over),and synteny was confirmed using the online Cinteny tool (24).Single nucleotide variants for cardiometabolic traits in these syntenic regions were harvested from GWAS Central (www.gwascentral.org)(25).By incorporating data from mouse SNPs, available mouse phenotyping, human single nucleotide variants, and searching literature for known roles in lipoprotein biology, candidate causal genes were nominated.

RT-PCR
For liver gene analyses, liver samples were homogenized in Qiazol lysis buffer in a TissueLyser II and RNA isolated with RNeasy Mini Kit (Qiagen) following the manufacturer's protocols.Hepatic gene expression of Asah2 was normalized to β-actin (Actb), and fold change relative to wild-type controls for each sex was calculated using the 2 -ΔΔCt method (26).

Statistical analyses
Founder plasma lipoprotein data were analyzed in JMP Pro, version 15.0.0 (SAS Institute, Cary, NC).All data were logtransformed prior to statistical analysis.Chow-fed and HFHS-fed data were initially analyzed separately for strain and sex effects, prior to assessing the interaction effect of diet on strain and sex.Strain, sex, and diet interactions for each lipoprotein subclass were tested using a standard least squares model with P < 0.05 denoting a significant effect.Least square means differences with Tukey's honest significant difference post hoc analysis determined statistical differences between groups (P < 0.05).For instances where the interaction effect and one of the main effects failed to reach significance, a one-way ANOVA was used.Heritability calculations were conducted in R (version 4.3.1)using the "lme4" package to fit a linear mixed model with restricted maximum likelihood.Chow-fed and HFHS-fed mice were analyzed separately, with sex as the fixed effect and strain as the variable effect.Asah2 mouse data were first checked for a Gaussian distribution and log-transformed if not normally distributed.Statistical analyses of normally distributed data were performed by ANOVA followed by Tukey's post hoc analysis.Differences were considered significant at P < 0.05.

Strain and sex dependence of diet-induced alterations in lipoproteins
To estimate heritability (h 2 ) of lipoprotein phenotypes in DO mice, we first analyzed the lipoproteins of the eight founder strains of DO mice.The mice were fed a chow diet or a HFHS diet, and their plasma were analyzed by ion mobility analysis, a method that quantitates the various size categories of lipoproteins (supplemental Table S1) (27).
Mice fed a standard chow diet showed marked straindependent differences in their lipoprotein size distribution (Fig. 1A and supplemental Fig. S1), with significant interactions between strain and sex for all lipoprotein subclasses (P < 0.03).NOD, NZO, and PWK mice had increased lipoprotein abundance relative to the other strains (P < 0.005), with NZO having the highest concentrations of small LDL (LDL-IVa, -IVb, -IVc, P < 0.0001).PWK mice showed variation in classes by sex where females (F) had increased the small lipoprotein particles: HDL-3,2a, HDL-2b, and midzone relative to male (M) PWK mice (P < 0.007, P < 0.0001, P < 0.0001, respectively).NZO mice had increased concentrations of LDL particles compared with the other strains (P < 0.0001), whereas B6, WSB, and CAST mice had the lowest LDL particle concentrations (P < 0.03).Heritability was high for all particles, with strain explaining up to 77% of phenotypic variance (Fig. 1C, h 2 = 0.49-0.77).
When placed on the HFHS diet for 16 weeks, total lipoprotein concentrations were increased by ∼50% (25 ± 1 nM) in five of the eight mouse strains, compared with chow-fed mice (17 ± 1 nM, P < 0.0001).NZO, NOD, and PWK mice did not significantly change their total lipoprotein particle concentrations in response to the HFHS diet.However, there was no effect of sex or interaction between sex and strain on any of the individual subclasses in response to the HFHS diet (Fig. 1B and supplemental Fig. S1, P > 0.3).Therefore, we analyzed the particles by one-way ANOVA to determine differences between strains.
NOD mice had the lowest LDL-IIIb and LDL-IV particle concentrations relative to the other strains (P < 0.001).Large LDL-I and LDL-II particles did not vary by strain (P > 0.5).HDL-2b concentrations were highest in 129 mice and significantly different from CAST (P < 0.0004).Interestingly, the average LDL diameter was higher in chow-fed mice (204 ± 1 Å) compared with HFHS-fed mice (194 ± 1 Å, P < 0.0001).LDL peak diameter showed a strong strain dependence for chow-fed mice (P < 0.0001) that was not present for HFHS-fed mice (P > 0.6).Two mouse strains, A/J and B6, had the largest LDL diameter on the chow diet (P < 0.0001) but were not different from other strains on HFHS diet.Heritability for lipoprotein particle concentrations was diminished on the HFHS diet (Fig. 1D), with strain explaining ∼40% or less of the variance in particle concentrations (h 2 = 0-0.36).

Genetic association of lipoprotein classes in DO mice
We surveyed all lipoprotein subclasses in ∼500 DO mice genotyped at ∼69,000 genome-wide SNPs, enabling us to identify 30 QTL with LOD >6.0 (genome-wide P = 0.2) for association with plasma lipoprotein subclasses (Fig. 2A and supplemental Table S2).For several QTL, multiple lipoproteins comapped, including one on chromosome 10 (Chr 10) for LDL-I through LDL-IIIb.
Because DO mice derive from an outcross of eight founder strains, there are up to eight alleles segregating among DO mice across the genome.Through haplotype reconstruction, we could determine the association of strain-specific haplotype blocks with each phenotype and thus determine the directionality of each allele's influence on the phenotype, that is, their allele signatures.Figure 2B illustrates the allele signatures for all lipoprotein QTLs.
We defined lipoprotein QTL based on the genome location of the SNP having the highest LOD score and the allele signatures of comapping traits.For example, a single locus on Chr 10 at ∼57 Mbp that showed five LDL subclasses comapping in response to the same allele signature (low for B6 and high for 129 and WSB) was classified as a single QTL.After refining QTL based on genomic proximity and allele effects, we identified 21 unique QTL for all lipoprotein classes.Six QTL were identified for HDL-2b and five QTL for the HDL-2a,3 subclasses, the most for individual subclasses in our analyses (supplemental Table S2).Large VLDL had four QTL, whereas the remaining subclasses had 1-2 QTL.
To nominate causal genes at each QTL, we identified probable genes based on SNP association plots at each locus.One lipoprotein subclass, HDL-2b, significantly mapped to six loci, with four QTL having a LOD >6.0 (Fig. 3A).Each locus had a unique allele signature (Fig. 3B), indicating independent genetic regulation of this lipoprotein.By integrating mouse SNPs at each locus (supplemental Fig. S2 and supplemental Table S3), candidate genetic drivers were identified.The QTL on Chr 1 was located near Apoa2, the second most abundant apolipoprotein component of HDL (28).The QTL on Chr 9 was located near several apolipoproteins, including apolipoprotein A1 (Apoa1) as well as proprotein convertase subtilisin/kexin type 7 (Pcsk7).APOA1 is the most abundant apolipoprotein in HDL (28,29) and has been well characterized for its role in HDL function (30).The association of three genetic variants of PCSK7 with HDL-C and acute coronary syndrome has also been recently published (31).
Finding QTL at the Apoa2 and Apoa1/Pcsk7 loci demonstrated that our genetic screen could identify known drivers of cholesterol and lipoprotein levels.We next turned our attention to two distinct HDL-2b QTL on Chr 19.The first QTL includes a region between clusters of fatty acid desaturases (Fads1, Fads2, and Fads3) and membrane-spanning 4A (Ms4a) gene members (supplemental Fig. S2C), which have literature support for associations to lipid metabolism, total cholesterol, and HDL-C (32)(33)(34).The second QTL, on Chr 19, included two compelling genes (supplemental Fig. S2D): N-acylsphingosine amidohydrolase 2 (Asah2) and Apo-bec1 complementation factor (A1cf).Due to associations of ceramides with cardiovascular disease (35), we sought to investigate Asah2, which encodes a ceramidecatabolizing neutral ceramidase, as a causal gene for HDL-2b lipoproteins.
Given the increased concentration of apo-Bcontaining particles in Asah2 −/− mice, we asked if the abundance of the LDL receptor (LDLR) protein or mRNA was altered in Asah2 −/− mice.We assessed LDLR protein in liver by Western blot.In Asah2 −/− females, there was a small, but statistically significant, increase in LDLR protein compared with Asah2 +/+ females (supplemental Fig. S4F, H); males were not different by genotype (supplemental Fig. S4G, H).In addition, Ldlr and Pcsk9 expression in liver was similar between genotypes of the same sex (supplemental Fig. S4I).Thus, the mechanism by which the loss of Asah2 drives increased lipoprotein abundance is likely not because of downregulation of the LDLR protein.

Integration of mouse lipoprotein QTL with human GWAS
To determine the translational significance of our findings to humans, we analyzed the syntenic loci for significant traits in human GWAS.To determine synteny, a 2 Mbp flanking region was first identified for each mouse lipoprotein QTL.This 2 Mbp region was then used with the LiftOver utility from UCSC Genomic Institute (genome.ucsc.edu/cgi-bin/hgLift-Over) to yield the human syntenic locus for each QTL (supplemental Table S4).Synteny was further verified using the online Cinteny tool (24).We then asked if these loci were associated with cardiometabolic phenotypes, including cardiovascular, glycemic, blood lipid, or anthropometric (excluding height) traits in GWAS Central (www.gwascentral.org)(25).
Approximately 2,000 SNPs with cardiometabolic traits were identified within the syntenic regions (supplemental Table S5).SNPs syntenic with 20 of the 21 lipoprotein QTL identified in mice were strongly associated with metabolic traits in human GWAS (P < 10 −8 , Fig. 5).Fifteen of the 21 loci were highly enriched with associations for three or more trait categories surveyed, providing evidence of significant genetic associations in the human population.
To nominate causal genes for each locus, we considered lipid-associated SNPs in mouse and human data and searched for literature support and mouse phenotyping data through the IMPC (www.mousephenotype.org).At mouse QTL, we prioritized SNPs that fell within the genomic region with a 1.5 LOD drop.At the syntenic loci in human, we identified SNPs with significant associations (P < 10 −8 ).Next, we searched published literature for phenotypic or mechanistic support.Through this integrated pipeline of mouse lipoprotein QTL, human GWAS data, and published publicly available data, we nominated candidate driver genes for 18 of the 21 lipoprotein QTL (Table 1).
We identified Akna as the candidate gene for HDL-2b and Midzone lipoprotein classes at a QTL on Chr 4. Akna, an AT-hook transcription factor (36), promotes Cd40 expression (37), indicating a role in inflammatory processes.In human genetic association studies, AKNA variants have been linked to total cholesterol (38), HDL-C, and ApoA1 (39), and plasma sphingolipids (40).However, no studies have mechanistically linked Akna to cholesterol or lipoprotein metabolism.It is possible that Akna may reflect the role of HDL in inflammation and/or immune surveillance.
The second candidate driver for the HDL-2b locus on Chr 19 at ∼32 Mbp is A1cf (Apobec1 complementation Fig. 3. Genetic regulation of circulating HDL-2b.Genome-wide LOD profile for an HDL subclass (HDL-2b) identifies significant QTL on three chromosomes: 1, 9, and 19 (A).Allele effect plots illustrate distinct genetic architecture at each locus (B).Colored lines represent alleles derived from founder strains.Genomic position for peak SNP listed for each QTL.factor).A1cf works in conjunction with Apobec1 to catalyze the editing of Apob mRNA, which results in the introduction of a stop codon and production of a truncated protein product, ApoB48 (41,42).In human GWAS, there are strong associations between A1CF missense variants and plasma lipoprotein phenotypes  (total cholesterol, LDL-C, and serum ApoB) (25, https:// t2d.hugeamp.org/).
We identified a single locus associated with both IDL and LDL, mapping to Chr 6 (Figs. 1, 2).Interestingly, when we looked at the syntenic locus in humans, there were no significant associations for cardiovascular, lipid, anthropometric, or glycemic traits.However, both the mouse and human genomic region contain Apobec1 (apolipoprotein B mRNA-editing enzyme catalytic subunit 1).A recent study in DO mice also identified a QTL for atherosclerotic lesions at this locus (43).In Apobec1 −/− mice, plasma ApoB-100 is increased 176%, although this was not accompanied by changes to cholesterol concentrations in VLDL or LDL classes (44).The lack of genetic association in humans may be due to species differences where mice express Apobec1 in both liver and intestine (45), whereas human expression is restricted to the intestine (https://gtexportal.org/home/).It may also highlight the minimal contribution of plasma ApoB-48 concentration to total ApoB and the poor correlation to plasma cholesterol in humans (46,47).
We mapped a single locus spanning five LDL subclasses (Chr 10) and one for LDL diameter (Chr 6).Gja1, also known as connexin 43 (Cnx43), is our candidate gene for the LDL species mapping to Chr 10.Cnx43 +/− /Ldlr −/− mice have a reduction in atherosclerotic lesion formation, compared with Cnx43 +/+ /Ldlr −/− mice, without a change in plasma cholesterol or triglyceride (48).Adamts9, the candidate gene for LDL diameter, is a metallopeptidase with a type 1 thrombospondin motif.Adamts9 has not been directly studied for a role in lipoprotein metabolism but has been shown to be a suppressor of mammalian target of rapamycin pathway in cancer cell lines (49).Both thrombospondin-1 and thrombospondin-2, however, have been studied in cardiovascular disease (50,51) and mediate atherosclerotic plaque development in mice (52,53).
We identified seven QTL for the three major size ranges within VLDL particles.At these loci, we identified five candidate genes.Small VLDL, which mapped to Chr 19, includes the gene Gpam at the human syntenic locus.In mice, the founder strains CAST and PWK have splice and untranslated region variants (supplemental Table S5), consistent with those two strains carrying the low alleles for the Chr 19 QTL (supplemental Table S1).Within human GWAS, GPAM is strongly associated with HDL, LDL, and total cholesterol, with missense variants that are highly associated to each of the three traits (supplemental Table S4, https://t2d.hugeamp.org/).Gpam encodes for outer mitochondrial membrane glycerol-3-phosphate acyltransferase, an enzyme in the triglyceride synthesis pathway.Consistent with a role in lipoprotein metabolism, Gpam knockout mice have lower VLDL secretion rates and decreased liver triglycerides relative to wild-type mice (54).
We nominate Igf1 as the causal gene at a QTL on Chr 10 for large VLDL.In mice, Igf1 was shown to reduce liver cholesterol accumulation by activating Abca1 (55), and an association between plasma Igf1 cardiovascular disease risk has been found in human subjects (56).In bovine hepatocytes, Apob expression and VLDL secretion are increased with addition of exogenous IGF1 (57).
For two additional VLDL QTL (Chrs 1 and 8), we identified two lesser-known gene candidates: Ankrd44 and Cdh11, respectively.In a study looking for genetic drivers of stroke in cerebral arteries, Ankrd44 was downregulated in veins from rabbits with hypercholesterolemia alone and with combined hypercholesterolemia and hypertension (58).ANKRD44 genetic variants are associated with stroke risk in African Americans (59), though no mechanistic studies were found.Cadherin-11 (Cdh11), a cell adhesion protein, was nominated as the causal gene for intermediate-sized VLDL mapping to Chr 8. Cdh11 has been implicated in autoimmune disorders, aortic valve calcification, and recently, scarring following myocardial infarction through its effect on fibrosis and inflammation (60).In a mouse model of atherosclerosis, Cdh11 expression was increased in atherosclerotic plaques of ApoE −/− mice.ApoE −/− /Cdh11 −/− double knockout mice had altered immune cell populations and increased atherosclerosis (61).

DISCUSSION
Lipoproteins that are isolated based on size or buoyant density represent snapshots of dynamic processes involving the transfer of lipids and proteins between particles and between tissues and lipoprotein particles.Genetic linkage and association studies help to identify the genes that affect these dynamic processes.The heterogeneity of the major lipoprotein classes (VLDL, LDL, and HDL) in the DO founder strains suggested that we might find QTL that explain this heterogeneity (Fig. 1).
Prior to embarking on the genetic study in DO mice, we first estimated heritability (h 2 ) of these distinct lipoprotein subclasses in the eight founder strains fed either chow diet or HFHS diet.Heritability estimates were reduced by approximately half for mice maintained on the HFHS diet (Fig. 1C, D), which in part may reflect increased intrastrain variance in particle concentrations (supplemental Fig. S1).Previous reports have highlighted an interaction between diet and genetics that alters heritability of metabolic phenotypes.In pedigreed baboons, high fat and/or high cholesterol diets increased variance and reduced h 2 in plasma HDL-C, median HDL size, and ApoA1/ApoB protein abundance compared with a low-fat, low-cholesterol diet (62).In a survey of 13 inbred mouse strains fed a Western diet, up to 75% of variance in lean body weight could be attributed to genetics (h 2 ), whereas other phenotypes (e.g., plasma glucose) showed reduced heritability (h 2 < 20%) (63).Finally, a panel of 22 inbred CC strains showed that some traits (e.g., body weight and plasma cholesterol) showed higher variation than other traits (e.g., body fat) between high-protein diet or HFD (64).Our results are consistent with these previous findings, showing that a metabolically challenging diet can increase nongenetic variance and thus reduce estimated h 2 of metabolic phenotypes.
Despite the reduction in estimated heritability observed for mice maintained on the HFHS diet, we identified 21 QTL for 16 lipoprotein subclasses in DO mice.Some of the loci contain genes encoding wellknown apolipoproteins (e.g., the locus harboring apoA1, C3, A4, A5, and PCSK7, and another locus containing apoA2), inspiring confidence in the ability of our screen to detect relevant loci.Nearly all these QTL, when lifted over into the human genome, are associated with lipid traits, as well as other cardiometabolic phenotypes, including cardiovascular disease risk, body weight, and diabetes.
At the HDL-2b locus on Chr 19 at ∼32 Mbp, we identified two candidate drivers: Asah2 and A1cf.We chose to investigate the relationship between Asah2 and lipoproteins by phenotyping Asah2 −/− mice.In female mice fed an HFD, Asah2 deletion resulted in increased HDL, midzone, large LDL, IDL, and small VLDL particles (Fig. 4).A similar trend was seen in midzone and IDL-2 particles in male mice.Asah2 is highly expressed in the intestine and functions as a neutral ceramidase.Although sphingolipids, including ceramides, can be carried on LDL and VLDL particles (65), to our knowledge, this is the first time a ceramidase has been shown to affect lipoprotein abundance.
Ceramides have recently been recognized as a cholesterol-independent lipid biomarker of cardiovascular disease risk (66), with the potential for being a predictor of endothelial dysfunction and early atherosclerosis (67).In a mouse model of atherosclerosis, pharmacological inhibition of hypoxia-inducible factor 1α in adipose tissue decreased ceramide formation through a neutral sphingomyelinase and was associated with decreased plasma cholesterol and delayed atherosclerotic plaque progression (68).Moreover, targeted disruption of ceramide synthesis similarly blunts plaque formation.Specifically, genetic deletion (69) or pharmacological inhibition (70,71) of serine palmitoyltransferase, which catalyzes the committed step in de novo ceramide synthesis, also blunts plaque formation.Asah2 is required for intestinal degradation of dietary sphingolipids (19).As ASAH2 expression is predominantly in the intestine, its activity may alter intestinal absorption of cholesterol.Indeed, strategies that reduce ceramides in the gut can blunt cholesterol absorption (72,73).Moreover, ASAH2 mediates the body's response to microbial sphingolipids (19,74).Deletion of Asah2 alleviates diet-induced nonalcoholic steatohepatitis/nonalcoholic fatty liver disease via downregulation of stearoyl-CoA desaturase (Scd1) and reduces cholesterol accumulation (75).When fed a very low-fat diet, Scd1 −/− mice have increased plasma cholesterol, especially in the LDL and VLDL fractions (76).When fed a standard chow diet (13% calories as fat), Scd1 −/− mice have increased plasma HDL-C (77).Therefore, one might hypothesize that Asah2 modulates lipoprotein metabolism through its effect on Scd1 expression.
For many years, elevated HDL was considered to be protective against atherosclerosis.However, Mendelian randomization studies and mouse knockout experiments showed that this concept was overly simplistic (78).Rather, opposing dynamic processes affect HDL and atherosclerosis risk by affecting the dynamics of cholesterol transport.For example, mutations in ABCA1 reduce the ability of cells to transport cholesterol and phospholipids out of cells (79,80).This results in lower HDL and increased atherosclerosis (81).In contrast, mutations in SRB1 decrease the transport of cholesterol esters from HDL into cells (82)(83)(84).This leads to increased HDL and increased atherosclerosis.
We are not aware of any genetic association between ASAH2 and atherosclerosis.Its modulation of HDL could increase or decrease atherosclerosis.Alternatively, ASAH2 could affect atherosclerosis through its effects on sphingolipids, independent of its effect on HDL, through their effects on inflammatory pathways (85).Indeed, ceramides predict coronary artery disease independently of cholesterol (86).Pharmacological inhibition of ceramide synthesis via the serine palmitoyltransferase inhibitor, myriocin, reduces atherosclerotic plaque formation in APOE-deficient mice (71,73).Moreover, ceramidases can also increase the formation of sphingosine-1-phosphate, an antiatherogenic lipid that is largely carried in ApoMcontaining particles (87).
Interestingly, a recent mouse genetic study identified a QTL for atherosclerotic lesions at the Asah2 locus in male mice (88).This study utilized a DO-F1 mouse population, generated by breeding DO mice to cholesteryl ester transfer protein;ApoE3-Leiden mice to produce an F1 generation with increased susceptibility to atherosclerosis.While their QTL on Chr 18 overlapped in position with the Asah2 locus in our study, the allele effects did not match between the two studies.It is possible that the discrepancies in allele effects are due to breeding strategies or reflective of differing responses to dietary compositions.Nevertheless, it is intriguing that this locus has been linked to both atherosclerosis and lipoprotein biology in two mouse genetic screens.
In summary, we mapped mouse lipoprotein subclasses to identify 21 unique QTL, which we then crossreferenced with the human syntenic locus to determine candidate genes associated with cardiometabolic traits in human GWAS.By integrating mouse data with human GWAS at the syntenic locus, we nominated candidate genes for 18 unique lipoprotein QTL.Deletion of Asah2, a novel candidate driver for HDL-2b, resulted in increased HDL, midzone, large LDL, IDL, and small VLDL particles in plasma from female mice.Similar validation experiments can be performed to explore the candidate genes we have nominated at the remaining QTL.

Fig. 1 .
Fig. 1.Genetics and diet exert a strong influence on circulating lipoproteins.Profile of lipoprotein subclasses in female (F) and male (M) mice of the eight DO founder strains maintained on standard rodent chow diet (A) or a Western-style diet high in fat and sucrose (B).Heritability (h 2 ) estimates for lipoprotein particles in mice fed a chow diet (C) versus a Western-style diet (D).

Fig. 2 .
Fig. 2. Genetic architecture of circulating lipoproteins.Heatmaps illustrate average Z-scores across all mice for 3-5 mice per sex/ strain.Genome-wide QTL for lipoprotein subclasses in ∼500 DO mice maintained on Western-style diet (A).LODs greater than 7.4 met the genome-wide threshold for significant QTL, whereas LODs greater than 6.0 are considered suggestive QTL.Supplemental Table S2 lists all QTL, their genomic positions, LOD scores, and allele effect values.Allele effect values are illustrated for 30 QTL for individual lipoprotein subclasses (B).Blue depicts alleles associated with reduced lipoprotein values, red depicts for increased values.Founder strains are listed on the left, whereas lipoproteins and their QTL (Chr and Mbp location) are shown along the top and bottom, respectively.Loci with similar allele effect patterns (e.g., Chr 10 at ∼57 Mpb for LDLIIIb-LDLIIa and HDL-3,2a) are considered one QTL.

Fig. 5 .
Fig.5.Mouse lipoprotein QTL are syntenic to regions associated with cardiometabolic traits in humans.Heatmap illustrates maximum enrichment (−log 10 P value) for single nucleotide polymorphisms that are present within regions syntenic to mouse lipoprotein QTL and associated with lipid, glycemia, anthropometric (excluding height), or cardiovascular phenotypes in human GWAS.

TABLE 1 .
Candidate genes for plasma lipoprotein QTL