Mouse genome-wide association studies and systems genetics uncover the genetic architecture associated with hepatic pharmacokinetic and pharmacodynamic properties of a constrained ethyl antisense oligonucleotide targeting Malat1

Antisense oligonucleotides (ASOs) have demonstrated variation of efficacy in patient populations. This has prompted our investigation into the contribution of genetic architecture to ASO pharmacokinetics (PK) and pharmacodynamics (PD). Genome wide association (GWA) and transcriptomic analysis in a hybrid mouse diversity panel (HMDP) were used to identify and validate novel genes involved in the uptake and efficacy of a single dose of a Malat1 constrained ethyl (cEt) modified ASO. The GWA of the HMDP identified two significant associations on chromosomes 4 and 10 with hepatic Malat1 ASO concentrations. Stabilin 2 (Stab2) and vesicle associated membrane protein 3 (Vamp3) were identified by cis-eQTL analysis. HMDP strains with lower Stab2 expression and Stab2 KO mice displayed significantly lower PK than strains with higher Stab2 expression and the wild type (WT) animals respectively, confirming the role of Stab2 in regulating hepatic Malat1 ASO uptake. GWA examining ASO efficacy uncovered three loci associated with Malat1 potency: Small Subunit Processome Component (Utp11l) on chromosome 4, Rho associated coiled-coil containing protein kinase 2 (Rock2) and Aci-reductone dioxygenase (Adi1) on chromosome 12. Our results demonstrate the utility of mouse GWAS using the HMDP in detecting genes capable of impacting the uptake of ASOs, and identifies genes critical for the activity of ASOs in vivo.


Introduction
Antisense oligonucleotides (ASOs) are highly selective and potent therapeutic agents which has proven effective in treating a variety of disease states including cancer, viral infection and cardio-metabolic, inflammatory and neurological diseases [1][2][3][4][5][6][7]. Antisense technology uses short synthetic (12-24 mers), chemically modified DNA-like oligonucleotides to alter the intermediary metabolism of RNAs. The most widely exploited mechanism of ASO functionality is the degradation of complementary mRNA utilizing ribonuclease H1 (RNase H1), thereby preventing the translation of the associated protein(s) [8]. The evolution of this technology over the past 30 years has led to the development of a variety of ASO modifications resulting in greater stability, potency, affinity, and reduced toxicity [9,10].
While these therapeutic agents have proven to be effective against disease in the clinic, significant variability in ASO response has been reported in several of the drugs [11,12]. For example Kynamro, a second-generation 2´methoxyethyl ASO targeting human apoB, has been approved by the FDA for use as an adjunct with first-line therapies to reduce apoB and total cholesterol in homozygous familial hypercholesterolemia (FH) patients. In Phase 3 trials, FH patients who were administered 200 mg/week of the drug observed reductions in plasma LDL-C levels ranging from 2% to -82% [1]. Differences in observed efficacy is not unique to the ASO platform: genetic make-up may account for 20%-95% of overall variability of therapeutic agents in general [13]. Importantly, previous research has indicated that pathways leading to ASO tissue accumulation and pathways driving ASO activity may be segregated due to the presence of ASO 'sinks', trapping ASO within endocytotic compartments [14,15]. To date, no in vivo systematic interrogation of the role of genetic architecture on ASO accumulation and potency has been performed.
Genome-wide association studies (GWAS) provide 1) hypothesis-free mapping and 2) the identification of causal variants within the genome that contribute to the variability of a complex trait. The utilization of this system has allowed the detection of single-nucleotide polymorphisms (SNPs) leading to novel mechanistic insights and the identification of hundreds of genes potentially causal for human pathophysiological states such as diabetes, cancer and numerous cardiovascular diseases [16,17]. Association studies in mouse models provide multiple advantages over large scale analyses in human populations, including cost effectiveness, reproducibility of results and the reduced impact of environmental factors. The hybrid mouse diversity panel (HMDP) is one such resource. It consists of over 100 genetically unique inbred mouse strains-30 "classical inbred" strains, in addition to over 70 "recombinant inbred" strains. The recombinant inbred strains were derived from the F1 crosses of eight inbred "founder strains" and are able to offer important insights into genetically-derived differences in phenotype among inbred mice. The strains making up the HMDP are genotyped at 140,000 high quality SNPs [18][19][20], with sufficient power to detect traits contributing to 10% of overall phenotypic variance. GWAS carried out with the HMDP have identified significant SNPs in a wide variety of phenotype measurements such as NAFLD, bone mineral density, insulin resistance, obesity and heart failure [21][22][23][24][25].
To better understand the genetic variability and genomic variants that are associated with ASO uptake and potency, we employed GWAS and transcriptomic analysis of the HMDP using a generation 2.5 2´, 4´-constrained 2´-O-ethyl (cEt) ASO [9] targeting the murine long noncoding RNA metastasis associated lung adenocarcinoma transcript 1 (Malat1). Following a single dose of this ASO, both hepatic Malat1 expression and tissue accumulation were evaluated in 100 HMDP strains. Significant intra-strain variability was observed in both hepatic Malat1 ASO potency and accumulation. Using Factored Spectrally Transformed Linear Mixed Model (FastLMM) [26,27], we identified two loci on chromosomes 4 and 10 associated with hepatic Malat1 ASO accumulation. Systems genetic analysis determined these loci contributed to expression of Stabilin 2 (Stab2) on chromosome 10 and vesicle associated membrane protein 3 (Vamp3) on chromosome 4. Additionally, we identified three loci, two on chromosome 12 and one on chromosome 4, associated with variation in hepatic Malat1 ASO potency. Rho associated coiled-coil containing protein kinase 2 (Rock2) and Aci-reductone dioxygenase (Adi1) were identified as high confidence candidate genes regulating ASO PD in chromosome 12, while UTP11 small subunit processome component (Utp11l) was identified on chromosome 4. Additional in vitro studies validated Rock2 contributions to the hepatic potency of the Malat1 ASO. Our results demonstrate that genetic variation impacts ASO PK/PD and validates the use of the HMDP GWAS in advancing our understanding of the molecular mechanisms that contribute to ASO biology.

Genome-wide association analysis of hepatic ASO accumulation
To identify genomic regions associated with Malat1 ASO PK, 6-week-old male mice from the HMDP were administered a single 2 mg/kg dose of either the constrained ethyl (cEt) Malat1 ASO (ION 556089) or the cEt control ASO (ION 549144), which does not target any known coding gene in the mouse genome (S1 Fig). A wide spectrum in hepatic accumulation of the Malat1 ASO ( Fig 1A) was observed, ranging from 0.29 μg/g (Akr/J mice) to 2.17 μg/g (BXD31/ TyJ mice). This variability did not correlate to average body ( Fig 1B) or liver weights (Fig 1C).
These genes included Stab2, a type I transmembrane hyaluronan receptor involved in multiple cellular processes that has been previously implicated in affecting hepatic accumulation of phosphorothioate ASOs [28,29].

Systems genetics analysis of additional PK associated SNPs
Since the majority of SNPs identified in the HMDP occur in noncoding regions, it became imperative to not only identify genes within the same LD block, but also those with expression regulated by the identified SNPs. To this end, global gene expression microarrays conducted using the liver tissue of chow-fed male mice from 95 HMDP strains [30] (NCBI GEO GSE16780) were used in conjunction with the SNP data to generate a list of expression quantitative trait loci (eQTL). The eQTL analysis led to the identification of cis-(within 1 Mb of the peak SNP) and trans-(greater than 1 Mb from the peak SNP) regulated genes corresponding to the SNPs at Chromosome 4 (S1 Table) and Chromosome 10 (S2 Table). Vesicle associated membrane protein 3 (Vamp3) was determined to be strongly regulated in cis by identified chromosome 4 SNP rs32062485 (p = 2.4 x 10 −28 , S1 Table and Table).

Validation of Stab2's role in hepatic ASO accumulation
Stab2, a previously determined key player in ASO accumulation [28,29,31], was identified within the LD block of peak chromosome 10 SNP rs29364476, and was further determined to be strongly regulated in cis by this SNP (p = 3.48x10 -6 , Fig 3A, S2 Table). Systems analysis of the peak SNP demonstrated a significant difference in hepatic ASO concentration dependent upon the genotype of SNP rs29364476 ( Fig 3B). Based on the variation of hepatic expression of Stab2 across the 100 strains, we next identified the strains expressing low and high levels of Stab2 (Fig 3C, S2 Fig) and compared the hepatic drug tissue accumulation between the two groups via LCMS. Mice with lower Stab2 expression accumulated significantly less drug compared to strains with higher Stab2 expression ( Fig 3C). To further confirm the role of Stab2, we utilized previously published Stab2 -/mice [28]. Compared to WT (Stab2 +/+ littermates), Stab2 -/mice demonstrated significantly less ASO accumulation in both liver and spleen after a single Malat1 ASO dose (Fig 3D and 3E). Importantly, there was a trend but no significant difference in the ASO efficacy between the Stab2 +/+ and Stab2 -/mice ( Fig 3F).

Validation of Vamp3's role in hepatic ASO accumulation
The Vamp3 gene product is a small soluble N-ethylmaleimide-sensitive factor attachment protein receptors with Arg/R residue (R-SNARE) highly expressed in the liver and known to play a vital role in providing specificity in catalyzing the fusion of vesicles to their target membrane [32]. Since Vamp3 plays a key role in vesicular transport, a process that has been implicated in ASO uptake, we investigated the putative role of Vamp3 in hepatic ASO PK. As with Stab2, systems analysis of the peak SNP rs32062485 demonstrated a significant difference in hepatic ASO concentrations based on the genotype distribution at that SNP ( Fig 4B). In order to isolate the effect of Vamp3 expression and reduce the input of confounding genes, we utilized the BXD subset of the HMDP panel, which is derived from a single founder pair cross and shown previously to provide sufficient power in GWAS [23]. Our analyses determined that BXD strains with relatively lower Vamp3 expression demonstrated significantly lower hepatic ASO concentrations ( Fig 4C, S3A Fig).
In order to experimentally validate the role of Vamp3 in ASO accumulation, we utilized a mouse hepatocellular SV40 large T-antigen carcinoma (MHT) cell line, which is capable of free ASO uptake. Cells were transduced with shRNA targeting either Vamp3 or a scrambled control, then assessed for Vamp3 expression reduction by QPCR ( Fig 4D). Cells were subsequently treated with 250 ug Malat1 ASO for 24 hours. LCMS of cells transduced with shRNA targeting Vamp3 demonstrated a significant decrease in ASO accumulation, validating Vamp3 as an important mediator of ASO uptake ( Fig 4D).

Genome-wide association analysis of hepatic ASO activity
There was also marked variability in hepatic Malat1 ASO activity (Fig 5A), ranging from 82% target reduction in BXA14/PgnJ mice to essentially no activity (0%) in 129X1/SvJ mice. These  To identify genomic loci responsible for the variation in potency, we performed FASTLmm analysis with the genome-wide significance threshold of 4.1 x 10 −6 and identified three loci ( Fig 6A, Table 4): two loci on chromosome 12 (rs29210579 p = 5.4 x 10 −6 and rs229212236 p = 5.5 x 10 −6 ), which were 8.4 Mb apart. rs29210579 contained 9 genes with in the LD block, 7 of which are expressed in liver, while rs29212236 contained 6 genes within the LD block with 4 expressed in liver (Tables 5 and 6). The remaining locus was one on chromosome 4 (rs27549337, p = 1.33x10-6), with one hepatic gene in the LD block (Table 7).

Systematic identification, systems genetics analysis and validation of genes associated with Malat1 ASO activity
Consistent with the accumulation studies described above, we again performed transcriptomic (eQTL) analyses using hepatic microarray expression data, accessible on NCBI GEO (GSE16780). This effort resulted in a list of genes that were in cis and trans regulation by the peak SNP at chromosome 4 (rs27549337) and two on chromosome 12 (rs29210579 and rs29212236) (S3, S4 and S5 Tables).
Analysis of SNP rs27549337 on chromosome 4 indicated that variability of the ASO activity associated with this SNP was not associated with endogenous Malat1 mRNA expression variability, further validating the systems genetics approach (Fig 5B and 5C). Four genes were revealed to be regulated by that SNP in cis. Three of these genes are expressed in liver, and intriguingly the list includes Utp11l, small subunit processome component [33]. This protein is ubiquitously expressed and is involved in active pre-RNA processing complex, making it an appealing target [34]. Systems analysis of the chromosome 12 SNP rs29210579, as with the chromosome 4 SNP, indicated that the variability in Malat1 ASO activity associated with the SNP was not due to basal Malat1 mRNA expression level differences (Fig 6A and 6B). We identified Rock2 and Lpin1 as high-confidence candidate genes because they both were (i) expressed in the liver, (ii) present in the LD region (Table 6) and, (iii) strongly regulated in cis by rs29210579 (p = 1.27x10 -10 S4 Table).
To assess the function of Lpin1 in modulating ASO potency, we again utilized a mouse hepatocellular SV40 large T-antigen carcinoma (MHT) cell line [35], with non-silencing (NS) siRNA or Lpin1 siRNA for 48 h. Treatment with Lpin1 siRNA yielded a 58% reduction in target gene expression (S6A and S6B Fig). Subsequent exposure of siRNA treated cells to Malat1 ASO for 72 hours did not reveal differential ASO activity (S6C and S6D Fig) suggesting that Lpin1 is not playing a significant role in Malat1 ASO activity under these conditions.

Validation of Rock2's role in hepatic ASO activity
To inhibit the kinase activity of Rock, MHT cells were treated for 24 h with Y27632, a potent pharmacological inhibitor of Rock1 and Rock2. In the presence of Y27632 a significant decrease in phospho-myosin light chain 2 (MLC2) was observed (S6E Fig), consistent with inhibition of Rock kinase activity. Subsequent exposure of treated cells to Malat1 ASO revealed a significant difference in ASO activity in treated vs nontreated MHT cells (Fig 7D), validating our finding that Rock2 regulation impacts ASO activity.

Validation of Adi1's role in hepatic ASO activity
We also considered other cis-genes that may affect ASO PD. Aci-reductone dioxygenase (Adi1) was the gene most significantly regulated (p = 3.4x10 -38 , S5 Table and Fig 7A) in cis by  chromosome 12 SNP rs29212236. While Adi1 is primarily known for its role in methionine metabolism, the human orthologue has recently been implicated to a novel role in nuclear mRNA processing possibly by modulating splicing factor U1-70K-related functions [36]. Systems analysis of Adi1 reveals that strains with higher Adi1 had lower Malat1 ASO activity ( Fig  8B and 8C), thus suggesting that interactions between several genetic factors might affect the potency of ASO drugs. In order to futher validate the role of Adi1 in ASO activity, MHT cells were treated with either a scrambled siRNA or Adi1 siRNA for 24 hrs, washed, and subsequently exposed to increasing concentrations of Malat1 ASO (Fig 8D). Treatement with Adi1 siRNA showed a significant reduction in ASO activity compared to control siRNA, validating Adi1 as an important mediator of ASO efficacy in liver (Fig 7E).

Discussion
Antisense drugs have been successfully used as research tools and as therapeutic agents in the clinic for decades. While the basic uptake and activity properties of these drugs have been elucidated [37], ongoing research efforts seek to identify factors that affect the distribution, efficacy, safety and tolerability of these therapeutic agents [38]. Advances in ASO medicinal chemistry have drastically enhanced activity, allowing significantly lower doses of drug to be administered in the clinic. Recently, research has been increasingly directed toward understanding the mechanisms of oligonucleotide uptake [39]. Briefly, once ASOs reach the cell surface they are internalized via endosomal vesicles, ultimately reaching their intended mRNA target either in the nucleus or cytosol [14,39,40]. It is generally accepted that some pathways of internalization and trafficking are productive, i.e leading to a pharmacological effect, while others are pharmacologically non-productive 'sinks' [41,42]. Several structural, nucleic acid binding and chaperone proteins have been shown to bind to ASOs and influence intracellular localization and trafficking within the cell [42,43]. Most importantly, some of these interactions are known to impact ASO therapeutic potency [4].
Here we investigate the role of genetics in modulating hepatic Malat1 ASO uptake and activity in mice. Results from our studies demonstrate that after a single dose of a Malat1 cEtmodified ASO, hepatic distribution of the drug and activity vary significantly among the 100 genetically unique HMDP strains. This considerable variability in both parameters allowed us to perform GWAS and led to the identification of two genes, Stab2 and Vamp3, associated with variation in hepatic ASO deposition, and another two, Rock2 and Adi1, that are associated with variation in ASO potency. These results are particularly noteworthy as those four genes are involved in either internalization and endosomal/lysosomal trafficking processes known to be crucial for ASO uptake and subcellular localization [39,44,45] Previous work has demonstrated Stab2 is capable of binding and internalizing PS MOE ASOs in stabilin-expressing stable cell lines [28]. Additionally, immunostaining in wildtype mice displayed high ASO accumulation in tissues that expressed Stab2 [28,46]. Here we expand on these previous findings and confirm the role of Stab2 in hepatic ASO uptake. First, GWAS and systems analysis determined the Stab2 gene located within a peak SNP LD block, as well as locally regulated by the presence of strong cis-eQTLs associated with the SNP haplotype. Second, HMDP strains with lower Stab2 expression presented with significantly lower hepatic ASO accumulation when compared to strains with more highly expressed Stab2. Third, ASO immunohistochemistry has previously demonstrated that liver sinusoidal endothelial cells (LSECs), a site of high hepatic Stab2 expression, exhibited relatively high concentrations of ASO [47]. Finally, Stab2 -/mice accumulated significantly less hepatic and splenic Malat1 ASO when compared to WT mice. Significantly, the differences in Malat1 ASO uptake between Stab2 -/and WT were more pronounced in spleen, which has higher Stab2 expression [28]. While this evidence demonstrates that Stab2 expression is critical in a subset of cells for ASO uptake, that both the livers and spleens of Stab2 -/mice were capable of some ASO accumulation and the existence of cell types capable of ASO uptake which don't express Stab2 is evidence indicating the existence of Stab2-independent ASO uptake pathway(s) [48]. Surprisingly, when we assessed Malat1 antisense drug potency in the WT and Stab2 -/mice, we did not see any significant changes in hepatic Malat1 ASO potency. In fact, a recent study has demonstrated that Scavenger Receptor B1 (Srb1) ASO had reduced potency in Stab2 -/when compared to wildtype mice [41,35]. A likely explanation for this discrepancy is the fact that, relative to Malat1, a larger proportion of hepatic Srb1 mRNA expression is in LSEC cells versus hepatocytes [49,39]. Thus, ablation of Stab2 dependent uptake into LSECs would be expected to have more deleterious effect on SRB1 ASO potency relative to the Malat1 ASO that targets a more hepatocyte localized mRNA.
While the current work does not address the mechanism for the disconnect between hepatic Malat1 ASO accumulation and activity (S2 Fig), it is tempting to speculate that the Stab2-mediated ASO uptake/internalization pathway is predominantly higher capacity and lower affinity than other productive pathways [31,41]. This is further validated by our data demonstrating that Stab2 is implicated in ASO uptake but not efficacy (Fig 3). These data are consistent with evidence that there are numerous uptake/internalization pathways that can be broadly categorized as productive and non-productive with regards to antisense pharmacology  [14]. To better understand the relationship between ASO accumulation and activity, we compared ASO liver accumulation and potency in the 100 strains and found that there was no correlation (S7 Fig). This lack of hepatic accumulation-activity correlation further supports the existence of multiple ASO uptake pathways, and that these pathways have varying levels of efficacy in delivering the ASO to its target RNA. Rho-kinase (Rock2) is one of the serine/threonine kinases functioning as a downstream effector molecule of small GTPase RhoA promoting contractile force generation and morphological changes in cells and organs. Rock2 has been shown previously to promote acto-myosin contractility and tubulin polymerization by either directly phosphorylating the myosin regulatory light chain (MLC) and the myosin binding subunit (MYPT1) of the MLC phosphatase or phosphorylating LIM kinases-1 and -2 (LIMK1 and LIMK2) [50][51][52][53]. Rock2 activity is known to play a role in the movement of endosomal and lysosomal vesicles [54,55]. One gene that has been shown to be significant in the productive endosomal trafficking of ASO is Anxa2, a phospholipid binding protein that is required for the biogenesis of late endosomes [56]. Studies have shown that Anxa2is able to bind ASOs, and upon incubation of ASOs with cells ANXA2 was enriched in late endosomes, suggesting that Anxa2is important for intracellular conveyance of ASOs [42]. Interestingly, Rock activity is essential for several ANAX2 signaling pathways [57,58]. Even though the interactions of Anxa1 and Rock2 on ASO activity remain unclear, it suggests that factors affecting endosomal trafficking are important for productive ASO uptake. In summary, genetic factors have been shown for the first time to be capable of altering the hepatic ASO accumulation and activity of Malat1 cET ASO in vivo. Here we establish the utility of HMDP in identifying genes affecting the uptake and potency properties of ASOs; future work must focus on translating these findings to human systems. Future studies will include using in vitro and in vivo model systems to understand the specific mechanism(s) of action of these identified genes in ASO uptake and/or intracellular trafficking. Finally, further studies evaluating the factors affecting potency and uptake in disease models will provide us with a more concise understanding of ASO response spectrum across populations and will help us design more effective drugs with improved therapeutic benefits.

Ethics statement
Ionis is AAALAC accredited and follows the 8th Ed. Of the Guide for the Care and Use of Laboratory Animals and the 2013 AVMA guidelines for the euthanasia of animals. All animals in this study were anesthetized with Isoflurane and euthanized via cervical dislocation. The Ionis IACUC-approved protocol is # P-0225. This protocol was approved on 5/28/2014. Oligonucleotide synthesis and delivery 2´, 4´-constrained 2´-O-ethyl (cEt) ASOs were synthesized at Ionis Pharmaceuticals (Carlsbad, CA) as described previously [59]. The Malat1 (ION 556089) and control (ION 549144) ASOs were formulated in saline, and injected via subcutaneous administration into the animals. In order to monitor reproducibility of the subcutaneous dosing, we used C57BL/6J mice as the control strain for each experiment. Similar levels of Malat1 inhibition in the livers were observed for each experiment (S8 Fig).

Mice
Mice were obtained from The Jackson Laboratory and housed at Ionis Pharmaceuticals (Carlsbad, CA), maintained on a chow diet and entered into studies before they exceeded 7 weeks of age. Stabilin-2 knockout (Stab2 -/-) mice, developed as described in Hirose et.al. 2012 [60], were purchased from Riken BioResource Center. At the end of each experiment, mice were anesthetized, euthanized by cervical dislocation and blood was collected by cardiac puncture. Spleens and livers were harvested and immediately snap frozen in liquid nitrogen for mRNA expression analysis and histology. As many antisense drugs target liver [37,61], our main analytical endpoints were hepatic drug accumulation and reduction of Malat 1 liver mRNA expression. Despite the hepatic abundance of Malat1, its loss of function has been found to be phenotypically silent, with no effect on global gene expression, phosphorylation, splicing factor levels or pre-mRNA splicing events [46,62]. A dose response study of Malat1 ASO performed in 3 classic inbred strains (S9 Fig) determined the ED 50 to be approximately 2 mg/kg. ASO tolerability and hepatotoxicity were assessed by measurement of serum aminotransferase (ALT and AST) and spleen weights (S10A, S10B and S10C Fig). Although significant baseline variability was observed in the serum ALT/AST levels and spleen weights, no post-dose elevations in either parameters were observed in any strains.

RNA analysis
Total mRNA was isolated using a QIAGEN RNAeasy kit (QIAGEN, Valencia, CA, USA). Reduction of target mRNA expression was determined by qPCR using StepOne RT-PCR machines (Applied Biosystems) as described previously [35]. Relative levels of Malat1 and Lpin1 were normalized to Cyclophilin.

Liver pharmacokinetic analysis
Pieces of whole liver were minced, weighed into individual wells, then homogenization buffer [63] was added to those corresponding wells. Control tissue homogenate was made by adding homogenization buffer at a 9 to 1 ratio to weighed amount of mouse untreated liver. Aliquots and appropriate amounts of calibration standards were added into wells. Internal standard and approximately 0.25cm 3 granite beads were added and then extracted via a liquid-liquid extraction with ammonium hydroxide and phenol: chloroform: isoamyl alcohol (25:24:1). The aqueous layer was then further processed via solid phase extraction utilizing a Phenyl plate. Eluates had a final pass through a protein precipitation plate before dry down under nitrogen at 50˚C. Dried samples were reconstituted in water containing 100 μM EDTA. These samples were injected into an Agilent LCMS instrument consisting of a 1260 binary pump, a 1200 isocratic pump, a column oven, an auto sampler, and a 6130 single quadrupole mass spectrometer for analysis (Agilent, Wilmington, DE, USA).

Association analysis
Since genetic association studies in inbred mice produce inflated false positive results owing to population structure and genetic relatedness, it is essential to use appropriate statistical tests. Efficient mixed model association (EMMA) utilizes linear mixed model with single dimensional optimization and phylogenetic control based kinship analysis to account for population stratification [26]. EMMA enabled correction for the spurious associations helps to cut down on the data processing time manifold. We identified genetic associations using the FastLMM, which is a reformulated mixed model analysis that performs linearly in run time and memory footprint for GWAS in very large data sets [27]. We retrieved the genotypes from UCLA systems genetics database along with chromosomal locations of linkage disequilibrium (LD) blocks. The average LD block size for the HMDP is 2Mb. The blocks were calculated here using Plink2's implementation of the Haploview algorithm and appear to be smaller than average (S6 Table). The global gene expression microarrays which were generated from liver of chow fed male mice from 95 HMDP strains, were obtained from the UCLA systems genetics database and are available at NCBI GEO (GSE16780). These data were used to perform the cisand trans-analysis as previously described [30]. FastLMM on ASO accumulation and activity properties was performed with a genome-wide significance threshold of 4.1 x 10 −6 as described [21][22][23][24][25]30].

Statistics
Data are reported as means ± SEM. The statistical tests are mentioned in the figure legends. Statistical significance was considered when p<0.05. Unpaired Student's t-test or two-way ANOVA and Bonferroni's multiple comparison test have been used to determine significance relative to control groups. Lee.