An APOO pseudogene on chromosome 5q is associated with LDL-C levels

Abstract

Background

Elevated levels of low-density lipoprotein cholesterol (LDL-C) are a major risk factor for CVD via its contribution to the development and progression of atherosclerotic lesions. While the genetic basis of LDL-C has been studied extensively, currently known genetic variants account for only ~20% of the variation in LDL-C levels.

Methods

Through an array-based association analysis in 1,102 Amish subjects, we identified a variant strongly associated with LDL-C levels. Using a combination of genetic analyses, zebrafish models, and in in vitro experiments, we sought to identify the causal gene driving this association.

Results

We identified a founder haplotype associated with a 15 mg/dL increase in LDL-C on chromosome 5. After recombination mapping, the associated region contained 8 candidate genes. Using a zebrafish model to evaluate the relevance of these genes to cholesterol metabolism, we found that expression of the transcribed pseudogene, APOOP1, increased LDL-C and vascular plaque formation.

Conclusions

Based on these data, we propose that APOOP1 regulates levels of LDL-C in humans, thus identifying a novel mechanism of lipid homeostasis.

Introduction

Low-density lipoprotein cholesterol (LDL-C) is a major risk factor for CVD, and genetic factors are estimated to account for 40 – 60% of the variation in this trait ^1–3. Monogenic forms of hypercholesterolemia, caused by mutations in LDLR, APOB and PCSK9 occur in approximately 1/250 individuals ⁴ while the vast majority of cases are multifactorial and polygenic, with SNPs at numerous loci influencing variation in LDL-C identified by combining genome-wide association studies (GWAS) across multiple populations. To date, and despite extensive GWAS and sequencing efforts, only ~20% of the variation in LDL-C levels is explained by all the common and rare loci identified by sequencing studies and large meta-analyses ^5–7. Some of the identified loci have been mapped to genes in pathways linked to lipoprotein particle metabolism while other loci map to regions with no apparent biological candidate genes. Uncovering new genes associated with LDL-C levels and the pathways in which these genes reside is of high clinical relevance because such discoveries could reveal new therapeutic targets aimed at LDL-C reduction and CVD prevention.

Recent efforts to identify additional lipid-related loci have focused on increasing sample sizes by performing larger and more robust meta-analyses. Studying isolated populations, in which there may be an enrichment of variants that are otherwise rare in outbred populations, offers a complementary strategy. The Amish, a founder population who immigrated to the U.S. from Europe 14 generations ago, is one such population for whom enrichment of several high penetrance variants influencing serum lipid levels has previously been reported ^8–11. In this report, we describe a founder haplotype in the Amish that is associated with a 15 mg/dl elevation in LDL-C levels and provide evidence from association mapping and functional data linking this novel LDL-C-associated locus to a pseudogene associated with a recently discovered apolipoprotein, apolipoprotein O, or APOO.

Methods

The whole genome sequencing data generated through the NHLBI Trans-Omics Precision Medicine (TOPMed) Consortium and used in this study are available on dbGaP (phs000956.v2.p1). Study materials and/or analytic methods available on request from the authors for distribution to other researchers for purposes of reproducing the results or replicating the procedure.

We performed an array-wide association of LDL-C levels using the Illumina Cardio-Metabochip (Ilumina, San Diego, CA)¹² on 1,102 Amish subjects who were participants in one of our prior community-based studies focusing on cardiovascular health ^13–15 and replicated the association in an additional 798 Amish subjects. IRB approval was obtained, according to the guidelines of University of Maryland, Baltimore, and informed consent was obtained from all study subjects. To fine map the causative locus on the LDL-C associated haplotype, we performed genetic association analyses and recombination mapping in a subset of 1,094 subjects for whom whole genome sequence data was available through TOPMed consortium (Expanded Methods, Supplemental Material).

To identify the gene on the associated haplotype driving association, we used wild type (Tubingen) zebrafish to evaluate the effect of knockdown or overexpression of human chromosome 5q genes (Expanded Methods, Supplemental Material). All procedures described were approved by and in accordance with guidelines of University of Maryland, Baltimore institutional IACUC. Morpholino antisense oligonucleotides (Gene Tools, LLC) were designed and validated to target each ortholog within the chromosome 5 fine-mapped locus. For the genes without identifiable orthologs in zebrafish, we generated expression constructs of human mRNA and injected them into embryos. Zebrafish embryos were cultured until 5 days post fertilization (dpf) and fed either a control or high cholesterol diet. At 7 dpf we performed in vivo visualization of cholesterol by fluorescent labeling and quantified the amount of vascular plaque formation in the caudal vein (Expanded Methods, Supplemental Methods).

To establish APOOP1-specfic expression, we performed RT-PCR and sequenced the RT-PCR product to confirm they are APOOP1 transcript-specific and not from amplifying genomic DNA or transcripts from an APOO-related gene (Expanded Methods, Supplemental Methods). We also performed in vitro allele-specific transcription assays using an enhancer-less pGL3 promoter vector (Promega, E1761) and Huh-7 cells following protocols recommended by Dual-Glo Luciferase Assay System (Promega, E1910). A previously reported liver enhancer served as the positive control ¹⁶. Primer sequences used during cloning sequences containing associated SNPs are available upon request. Chromosome conformation capture (3C) assays to detect potential promoter and SNP interactions were performed similarly to the protocol described by Hagège et al. ¹⁷ and Naumova et al. ¹⁸ using chromosomes from Huh-7 cells and one of 3 restriction enzymes: EcoRI, HindIII, or PstI. Detailed protocol and primers used to detect ligated fragments from potential promoter-SNP interactions are available upon request.

Statistical analysis

All genetic association analyses were run using the Mixed Models for Pedigree Analysis (MMAP) software program to account for the relatedness among study subjects ¹⁹. For qRT-PCR, samples and controls were run at least in triplicate on a Roche LightCycler 480 instrument and normalized relative to actin expression. Student’s t-test was used to determine significance. Bonferroni correction was used to adjust for multiple comparisons.

Results

Association of LDL-C with a 3.4 Mb founder haplotype

As part of our Amish Complex Disease Research Program, we genotyped the Cardio-Metabochip in 1,102 Amish subjects (Supplemental Table 1) and tested for association with LDL-C. We observed a significant association between LDL-C and rs17573010, with each copy of the minor allele A (allele frequency: 0.08) associated with a 17.7 mg/dL higher LDL-C (95% C.I.: 12.4 – 23.0; p = 8.5 × 10⁻¹¹) (Supplemental Figure 1). We genotyped this variant in an additional 798 subjects (Supplemental Table 1) and consistent with the Metabochip array analyses, found the G allele of rs17573010 to be associated with higher (8.8 mg/dL; p = 0.007) LDL-C levels. In the combined set of 1,900 Amish subjects each copy of the G allele of rs17573010 was associated with a 13.1 mg/dL higher LDL-C level (95% C.I.: 9.0 – 17.2; p = 9.1 × 10⁻¹⁰, Supplemental Table 2). In contrast, only 5 of the recently identified GWAS loci associated with LDL-C through large meta-analyses have effect sizes > 4 mg/dL per allele, including common variants at APOB (4.1 mg/dL per allele), SORT1 (5.6 mg/dL per allele), LDLR (7.0 mg/dL per allele), and APOE (7.1 mg/dL per allele) ²⁰. Interestingly, we did not detect an association to this region in our prior GWAS genotyped on the Affymetrix 500K/6M arrays ⁸ because rs17573010 is not present on this array and could not be imputed from European imputation reference panels due to its low minor allele frequency (MAF) (0.033 in 1000G EUR). Nor has this variant been associated with LDL-C in large general populations reported by others ^{5, 7, 21}, suggesting it might be marking a novel functional variant through extended linkage disequilibrium, as we have observed for other associations in the Amish ^{8, 9}.

Rs17573010 is an intronic variant that falls within HAVCR1 on chromosome 5q, the gene encoding hepatitis A virus cellular receptor 1. Prior GWAS analyses, most notably the Global Lipids Consortium ^{20, 21}, identified another SNP in this region (rs6882076) as being associated with LDL-C, albeit with a much smaller effect size, with the C allele of rs6882076 (allele frequency: 0.63) associated with a 1.7 mg/dL lower LDL-C per copy ²⁰. The estimated effect size of rs6882076 is slightly larger in the Amish (beta = 5.6 mg/dL lower for the same allele; p = 0.01). Rs6882076, which has MAFs of 0.35 and 0.21 in 1000G EUR and Amish, respectively, is located upstream of TIMD4. It is 87 kb away from the peak-associated SNP in the Amish and the two SNPs are not in linkage disequilibrium (r² = 0.02). This lack of linkage disequilibrium in the Amish implies that rs6882076 and rs17573010 track different causal variants. Though speculative, the most parsimonious interpretation is that both variants regulate the same gene, although we acknowledge that there could be multiple genes in the region influencing variation in LDL-C.

Among the 1,901 subjects genotyped for rs17573010 were seven individuals who were homozygous for the rs17573010 G allele. All seven had previously been genotyped using the Affymetrix 500K or 6.0 GWAS Array Set (Affymetrix, Santa Clara, California). Examination of these individuals revealed that all were homozygous for 6,500 or more SNPs across a 4.8 or larger Mb region that included rs17573010, strongly indicative of a founder haplotype that entered the population as a single copy, increased in frequency in the population over time through genetic drift, and then was passed down to both parents of these seven homozygotes.

To localize the region of association surrounding rs17573010, we performed association analyses of whole genome sequencing data obtained in a subset of 1,094 genotyped Amish subjects recently obtained through the NHLBI Trans-Omics Precision Medicine (TOPMed) consortium. Descriptive characteristics of this cohort are shown in Supplemental Table 1. Association results of variants in this region with LDL-C are displayed graphically in a zoom plot (Figure 1, panel A), with detailed results of the association analyses presented in Supplemental Table 3. As indicated in these analyses, there was a broad region of association extending over a 3.4 Mb region, from approximately 153.8 Mb to 157.2 Mb. The most strongly associated variant in this region is rs193010652 (position chr5:156,410,046), with the minor allele at this locus (MAF= 0.076) associated with a 15.7 mg/dL higher LDL-C level; p = 4.63 × 10⁻⁸, adjusted for age, sex, and APOB R3527Q genotype, a major and common determinant of LDL-C levels in this Amish population ⁸ (see Expanded methods, Supplemental Material). The region contains 35 additional variants in high linkage disequilibrium with rs193010652 (r² ≥ 0.80), all showing strong evidence for association with LDL-C levels (5.09 × 10⁻⁶ < p < 4.63 × 10⁻⁸). Notably, this region carries the signature of a founder haplotype in that nearly all subjects carrying the risk allele at rs193010652 (and rs17573010) also carry the same extended haplotype, presumably inherited from the same founder (see Supplemental Table 3).

Figure 1. Recombination mapping of the LDL-C associated locus.

A. SNPs within a 3.4 Mb region show strong association to LDL-C that span an extended haplotype. Enlarged version of Panel A provided as Supplemental Figure 1B. B. Recombination mapping based on individuals with partial LDL-C associated haplotypes narrows this region to a 442 kb region that includes 14 associated SNPs. The low and high LDL-C associated haplotypes are shown as white and black bars, respectively. LDL-C levels residualized following adjustment for age, sex, and APOB R3527Q genotype are normalized to subjects homozygous for the low LDL-C associated haplotype. See text for statistical comparisons between haplotype groups. C. The core haplotype, presumed to carry the risk variant. SNPs with a minor allele frequency <0.01 in the 1000G EUR are denoted by *.

Narrowing region of association through recombinant analysis

In addition to the 7 subjects homozygous for the high LDL haplotype, our sample also included 140 subjects who were heterozygous for this 36-SNP haplotype. Mean LDL-C levels in these individuals were 16.2 mg/dL higher than in the 862 subjects in the dataset not carrying this full risk haplotype (Figure 1, panel B and Supplemental Figure 1B). Our sample also contained 42 subjects with a recombination within the high-risk haplotype. The two most frequent haplotypes were one consisting of 19 SNPs at the centromeric part of the high-risk haplotype carried by 32 subjects (SNPs 1–19, positions 153,759,461 – 155,935,513 in Supplemental Table 3) and one carried by 7 subjects consisting of only 3 SNPs at the telomeric part of the high-risk haplotype (SNPs 34–36, chr5:156,994,242 – 157,184,358 in Supplemental Table 3). Compared to the 862 subjects not carrying any of the 36 SNPs on the risk haplotype, these subjects had mean LDL-C levels that were 5.9 mg/dL lower and 4.8 mg/dL higher, respectively, suggesting that the causative risk variant falls within a 442 kb region between rs187465743 (chr5:156,186,559) and rs182553698 (chr5:156,629,144); that is, the region containing SNPs 20–33 in Supplemental Table 3) (Figure 1, panel B). There were an additional 3 subjects carrying 17 SNPs at the telomeric part of the high-risk haplotype (SNPs 20–36, chr5:156,186,559 – 157,184,358), with LDL-C levels of 30.8 mg/dL higher than in subjects without the risk haplotype, also consistent with the localization of the high-risk haplotype to this 442 kb region.

Summary association results for the 14 SNPs comprising the 442 kb haplotype showing peak evidence for association are provided in Table 1. All SNPs are strongly associated with LDL-C (3.46 × 10⁻⁶ < p < 4.63 × 10⁻⁸). Among these 14 SNPs, which include rs193010652, are 8 with MAF ≤ 0.01 in 1000G EUR. We speculate that the causative variant likely falls within this set, reasoning that any common SNPs present in the general population would have been well tagged in prior large lipid meta-GWAS studies and thoroughly tested for association with LDL-C in prior meta-analyses. These 8 SNPs are indicated in Figure 1, panel C with an asterisk. The remaining 6 SNPs have minor allele frequencies in 1000G EUR ranging from 0.015 to 0.098. In the Amish the frequencies of these SNPs range from 0.074 to 0.088, and the pairwise r² of all SNPs with rs193010652 ranges from 0.90–0.99, except for rs142538597, for which its r² with rs193010652 is 0.84.

Table 1.

Association of 14 most likely causal variants of increased LDL-C levels in the Amish^*

rsNum	Position (Build 37)	Risk allele	Alt allele	Type	Nearest Gene	Risk allele freq (Amish)	Risk allele freq (1000G EUR)	Per allele increase in LDL-C (mg/dL)	p-value
rs187465743	156,186,559	G	A	UTR3	SGCD	0.075	0.005	15.37	1.24E-07
rs36085389	156,196,392	T	G	intergenic	SGCD	0.075	0.034	15.37	1.24E-07
rs142538597	156,332,782	G	A	intergenic	TIMD4	0.088	0.015	12.40	3.46E-06
rs143855219	156,360,904	C	T	intronic	TIMD4	0.078	0.010	14.49	4.09E-07
rs898956003	156,369,777	C	T	intronic	TIMD4	0.075	Not found	15.37	1.24E-07
rs193010652	156,410,046	C	A	intergenic	TIMD4	0.076	0.002	15.72	4.63E-08
rs554775153	156,439,297	T	C	intergenic	HAVCR1	0.075	0.002	15.75	5.05E-08
rs557762686	156,447,673	C	A	intergenic	HAVCR1	0.076	0.002	15.28	1.15E-07
rs17573010	156,477,831	G	A	intronic	HAVCR1	0.081	0.033	15.05	8.02E-08
rs145465336	156,530,370	A	C	intronic	HAVCR2	0.079	0.028	14.85	1.58E-07
rs142557102	156,559,831	C	T	intergenic	MED7	0.079	0.009	14.85	1.58E-07
rs17575222	156,617,534	C	A	intronic	ITK	0.084	0.098	14.07	2.71E-07
rs78969277	156,617,894	T	C	intronic	ITK	0.084	0.098	14.07	2.71E-07
rs182553698	156,629,144	C	G	intronic	ITK	0.076	0.007	14.28	6.90E-07

^*adjusted for age, sex, and APOB R3527Q variant, n = 1,094 whole genome sequenced subjects

The 442 kb high risk haplotype encodes 7 protein-coding genes (SGCD, TIMD4, HAVCR1, HAVCR2, MED7, ITK, and FAM71B), 4 of which regulate various aspects of immune function (Supplemental Table 4). There are also 2 processed pseudogenes, APOOP1 and PPP1R2P3, that contain multiple stop codons and are not predicted to encode proteins. One of these, APOOP1, is transcriptionally active in liver (Supplemental Figure 2), a tissue highly relevant for cholesterol metabolism. We considered APOOP1 also to be worthy of follow-up analysis because of the important regulatory roles that some pseudogenes are known to play in gene expression, particularly for their parent genes ^22–24. Importantly, the parent gene (APOO) encodes a lipoprotein found in LDL and HDL particles ²⁵, and overexpression of ApoO in mouse liver leads to altered hepatic fatty acid metabolism ²⁶.

Functional or regulatory variants within the associated region

All 14 variants in the narrowed region (which notably include the 8 with MAF < 0.01 in 1000G EUR) are located in non-coding regions. Based on functional annotation data provided by ENCODE and Epigenome Roadmap, none of these SNPs are located in sequences with regulatory potential in liver. However, 2 SNPs, rs557762686 and rs17573010, are within sequences annotated as putative enhancers in hepatoma-derived cells, HepG2 (Supplemental Table 5). Nevertheless, we assessed whether these 14 non-coding SNPs could assert a regulatory effect on the 9 transcribed genes in this region by using chromosome conformation capture (3C) to detect physical proximity and potential interaction between promoters and SNPs via chromatin looping. Of the 126 (14 SNPs x 9 promoters) potential interactions, only 4 confirmed 3C products from SNP-containing and promoter-containing restriction fragments were detected: rs143855219 and APOOP1 (Figure 2A–D), rs142557102 with a fragment that contains the HACVR1 promoter, and rs182553698 with both MED7 and a fragment that contains the ITK and FAM71B promoters (Supplemental Figure 3). In addition to these functional candidates, SNP rs898956003 resides in a regulatory region predicted to enhance transcription based on histone modification in GM12878 lymphoblastoid cells. When the region containing this SNP was tested in an in vitro reporter assay, we detected a 1.9-fold higher transcription activity for the T compared to the C allele (Figure 2E). In contrast, no such allele-specific activity was found in other regions containing LDL-C associated SNPs (data not shown). In summary, four of the LDL-C associated SNPs with the strongest statistical evidence for association are located in sequences with regulatory potential and are excellent causal SNP candidates for this association.

Figure 2. Potential regulatory role of LDL-C associated SNPs.

A. The region containing TIMD4 and APOOP1 is shown with SNP rs143855219 (red star) and possible chromatin looping. B. Coordinates of the two EcoRI fragments that are linearly approximately 10 kb apart. C. PCR product demonstrates potential chromatin looping of regions containing rs143855219 and APOOP1 promoter (lane 4). Lanes 1, 2, and 3: no cross-linked DNA, no restriction enzyme, and no DNA ligase controls, respectively. D. The 3C ligation junction sequences were confirmed via sequencing. E. Allele-specific transcription activity of another LDL-C associated SNP, rs898956003, (green star in A). The region containing the T allele has 1.9-fold higher transcriptional activity compared to the C allele (p=0.007).

Functional screening of regional candidate genes in zebrafish

Our genetic data and recombinant mapping narrowed the peak region of association with LDL-C levels to a small set of potential causal variants. Yet these studies were unable to identify the causal gene, because reported functions of annotated genes in the associated region offered no obvious biological evidence for association with LDL-C, with the potential exception of APOOP1 (see Supplemental Table 4). To address this further, we turned to a zebrafish system of genetically-induced hypercholesterolemia ²⁷ to test the potential contribution of genes in the region to regulation of LDL-C in vivo. We identified zebrafish orthologs for 5 of the 8 candidates in the region: HAVCR1, TIMD4, MED7, ITK, and SGCD (Supplemental Table 6). We first verified the expression of each ortholog during zebrafish embryonic and larval stages and confirmed that all were expressed by 5 days post fertilization (dpf; Supplemental Figure 4A). We then designed splice-blocking antisense oligonucleotide morpholinos (MOs) to suppress expression of each ortholog. Two genes and one pseudogene in the region, HAVCR2, FAM71B, and APOOP1 (Supplemental Table 6), did not have identifiable orthologs in the zebrafish. To investigate their function, we generated expression constructs for generation of full length capped human mRNA for overexpression in zebrafish embryos. We did not examine the other pseudogene, PPP1R2P3, because unlike APOOP1, it is not transcribed. Each MO or mRNA was injected into one- to two-cell stage wild type embryos that were subsequently grown to 5 dpf. We verified that MO treatment resulted in knockdown of the targeted transcript by qRT-PCR (Supplemental Figure 4B–C) and did not induce off-target toxicity, as detected by quantification of Δ113-p53 expression ²⁸ (Supplemental Figure 4D). Overexpression of mRNAs was also verified by qRT-PCR (Supplemental Figure 4E–F). At 5 dpf we introduced either a control larval diet (CD) or one supplemented with 4% w/w cholesterol (HCD) followed by quantification of LDL-C from homogenates of pooled larvae. We found that on either CD or HCD, no MO or mRNA induced significant upregulation of total LDL-C relative to animals injected with a non-specific control MO treated with the same diet (Figure 3A). Importantly, we did not see an effect on LDL-C for any of three separate MO concentrations or two separate mRNA concentrations injected for each gene. The sole exception to this was a significant increase observed by overexpression of APOOP1, which produced an increase in LDL-C on both diets, similar to our positive control, knockdown of ldlr (Figure 3A; p < 1 × 10⁻¹¹). Consistent with this observation, we observed a significant increase in expression of genes involved in cholesterol synthesis, hmgcr and srebp2, in livers dissected from 5 dpf larvae overexpressing APOOP1 mRNA that was not observed upon manipulation of other genes (Supplemental Figure 5). To preclude the possibility that overexpression of APOOP1 impacted LDL-C through effects on other genes at the chr5q locus, we examined expression of all other gene orthologs in larvae injected with APOOP1 mRNA. We found no significant changes in expression of any genes, suggesting that the effects on LDL-C are not mediated by any other genes in the region (Supplementary Figure 6).

Figure 3. LDL and plaque quantification in zebrafish.

Quantification of (A) LDL-C levels (n=100 larvae per experiment) or (B) Average number of vascular lipid plaques (n=20 larvae per experiment) in larvae treated with either control or high cholesterol diet. Embryos were injected either with a MO to target orthologs for knockdown or with mRNA for overexpression of human genes. Results were compared to non-targeting control MO or ldlr MO. Data represent the average of three experiments. (*, p<0.01; student’s t-test with Bonferroni correction)

Increased LDL-C is a major risk factor for formation of atherosclerotic plaques and in zebrafish is directly associated with accumulation of oxidized LDL in vascular plaques ²⁹. To test the potential contribution of any orthologs to vascular plaque formation, we treated MO- or mRNA-injected larvae with CD or HCD supplemented with BODIPY 576/598 C11 allowing for in vivo visualization of cholesterol by fluorescent labeling ^{27, 30}. Upon quantification of the average number of fluorescent plaques in a 438.6 μm region of the caudal vein (Supplemental Figure 7), we found that APOOP1 overexpression resulted in an approximately 20-fold increase in the average number of vascular plaques observed (Figure 3B; 4.03 plaques compared to 0.21 in control, p < 1 × 10⁻¹¹). This was comparable, albeit less dramatic, to the increase observed with injection of ldlr MO (Figure 3B; 7.56 plaques). An increase of this magnitude was not observed for any other MO or mRNA.

Suppression of apoO expression increases LDL-C and plaques

APOOP1 is an expressed pseudogene whose function is unknown. Since pseudogenes can modulate or interfere with parent gene function, we examined the phenotypic relationship between APOOP1 and apoO in zebrafish and tested whether zebrafish apoO could also play a role in cholesterol homeostasis. apoO is present in two copies in the zebrafish genome, apoOa and apoOb (Supplemental Figure 8). In qRT-PCR analysis of pooled livers dissected from 5 dpf larvae, we identified hepatic expression of both, though apoOb was expressed at significantly higher levels (Supplemental Figure 9A). We therefore designed splice-blocking MOs targeting both genes and injected them individually. After verification of either gene knockdown by qRT-PCR (Supplemental Figure 4C, 9B) and precluding off-target effects from either MO (Supplemental Figure 4D, 9C), we assessed LDL-C levels. We observed a significant increase in LDL-C with suppression of either gene (Figure 4A, Supplemental Figure 9D). We also observed significant formation of vascular plaques in MO-injected larvae treated with HCD (Figure 4B–C;19.23 uM LDL-C and 8.32 average plaques in apoOb MO embryos compared to 13.88 uM LDL-C and 0.21 average plaques in control; p < 1 × 10⁻⁵ and p < 1 × 10⁻¹¹, respectively).

Figure 4. Suppression of ApoO impacts LDL-C and vascular lipid accumulation.

Quantification of (A) LDL-C levels (n=100 larvae per experiment) or (B) average number of vascular plaques in larvae (n=20 per experiment) injected with MO targeting apoOb and fed either control or high cholesterol diet. Data represent averages across three experiments (*, p<0.01; student’s t-test with Bonferroni correction). (C) Representative brightfield (upper) and fluorescence (lower) microscopy images of caudal vein in larvae injected with either control MO, mRNA encoding human APOOP1, or apoOb MO and fed with high cholesterol diet containing cholesteryl conjugated BODIPY, observed in vascular plaques (arrowheads).

Discussion

We have identified a novel association in the Amish with elevated LDL-C on chromosome 5q33. These studies reinforce the value of studying population isolates to map genes for complex traits insofar as the LDL-C association tracks a founder haplotype that likely entered the Amish population on a single founder more than 14 generations ago and drifted upward in frequency. Thus, this haplotype is tagged by variants that are common in the Amish (MAF=8%) but very rare in non-Amish Europeans. Initial identification of the associated founder haplotype was made possible because the Cardio-Metabochip included a variant that closely tracked this haplotype. In addition to the significant clinical relevance of this haplotype in the Amish, it merits emphasis that its discovery provides a window into the biology of human cholesterol metabolism more generally by implicating a role for APOO in LDL-C regulation, potentially through its mediation by a pseudogene. We also highlight another striking feature of this haplotype made possible through whole genome sequencing. That is, despite being associated with ~15 mg/dL increase in LDL-C, none of the associated variants on the haplotype are linked to a causative variant in exons.

Several lines of evidence converge on the pseudogene, APOOP1 as the causal gene in the region of association on chromosome 5q33. First, recombination mapping narrowed the region of association to a short list of genes and this region included SNPs demonstrating the strongest evidence for association. Second, this narrowed region also includes a SNP previously associated with LDL-C levels, albeit with a much smaller effect size, but nevertheless suggesting that it could be independently regulating the same causal gene. Finally, functional evaluation of the genes within the narrowed region in zebrafish revealed that overexpression of APOOP1, as well as reciprocally knocking down apoOA and apoOB, results in significant increases in LDL-C levels and vascular plaque formation.

Our study corroborates a previous large-scale functional examination of 133 genes from 56 loci identified through genome-wide association studies of blood lipid-related phenotypes using a cell-based RNAi screening approach.³¹ In that study, siRNA targeting two genes, HAVCR1 and TIMD4, at the chromosome 5 locus did not show consistent or significant changes in LDL uptake or in free cholesterol. This is important because these two genes are the most proximal to the GWAS association with LDL-C previously reported at this locus, and the failure to detect a consistent phenotype in cell-based studies is consistent with our zebrafish results showing no phenotypic changes in the morpholino knockdown of these two genes. To our knowledge, our study is the first to systematically examine all genes within this locus in a whole animal system using whole body LDL-C as a phenotypic readout. Furthermore, it is among the few studies to have mapped a non-coding regulatory variant as the most likely causative variant.

Like other processed pseudogenes, APOOP1 lacks introns and contains multiple stop codons. In the past, pseudogenes were believed to be functionless because of their inability to produce proteins. However, several pseudogenes have recently been identified that are not only transcriptionally active, but also capable of influencing the activity of other genes and affecting the phenotype of the organism (reviewed by ²³). Approximately 20% of pseudogenes, especially those that are evolutionarily conserved, are transcribed ^{32, 33}. APOOP1 is one of five processed pseudogenes of APOO in the human genome. The other four are located on chromosomes 3, 4, 12, and 16. While APOOP1 is not conserved in mice and zebrafish, it is the only APOO pseudogene that is transcriptionally active in a tissue-specific manner (expression of other 4 pseudogenes are not detected, data not shown), adding further support to its potential functional significance.

Pseudogenes interfere with the expression of other genes by several well-defined mechanisms. Pseudogenes can encode for an antisense RNA that inhibits the translation of the parent gene or generate a siRNA that leads to gene silencing. Pseudogenes can also compete for regulatory factors, including transcription factors (via its promoter or other regulatory sites) and miRNA (via the 3′UTR of the non-coding transcript). It is not known how the expressed APOOP1 might alter the expression of APOO. Because APOOP1 is not expressed in peripheral tissues, such as blood, we cannot test whether individuals with different LDL-associated genotypes exhibit differential levels of APOOP1 expression, particularly in liver. However, the 3′UTRs of both APOO and APOOP1 contain microRNA binding sites for miR429/200b, miRNAs that reportedly play a role in diet-induced obesity, nonalcoholic fatty liver disease, and statin responses ^34,35,36. Therefore, it is possible that the APOOP1 mRNA influences APOO mRNA stability through microRNA binding and transcript stability, similar to what has been reported for PTEN and its transcribed pseudogene, PTENP1 ³⁷. It is not known if APOOP1 transcripts also interfere or regulate genes other than APOO, particularly those regulated by miR-200b/a/429, for which there are predicted target sites in many genes.

There are two isoforms of ApoO, a glycosylated and secreted 55 KD and non-glycosylated 22 KDa mitochondrial protein. APOO mRNA is upregulated in human diabetic heart ²⁵ and APOO levels are higher in the serum of patients with acute coronary syndrome ³⁸. siRNA knockdown of APOO in HepG2 cells also alters the expression of genes in fatty acid metabolism and in genes related to the inflammatory response ³⁹. Also known as MIC26, this lipoprotein is a component of MICOS complex, which consists of the mitochondrial contact site and cristae organizing system⁴⁰. ApoO overexpression in mouse liver alters the expression of lipid metabolism genes and mitochondrial ultrastructure ²⁶. Over-expression in mouse hearts impairs mitochondrial function and initiates cardiomyopathy ⁴¹. In contrast, overexpression of apoO in apolipoprotein A-1 transgenic mice does not alter the plasma lipoprotein profile or HDL functions such as protection against LDL oxidation and anti-inflammatory activity on endothelial cells ⁴². While its physiologic role is incompletely understood, these studies implicate a critical role for APOO in fatty acid metabolism, mitochondrial function, and inflammatory response.

The functional studies we have performed rest heavily upon the use of larval zebrafish as a model for lipid metabolism. The utility of the zebrafish for rapid interrogation of multiple genes make it uniquely suited to this study where examination of whole body physiology under the control of multiple different candidate genes allowed us to functionally fine map the human genomic locus. Using this approach, we identified a single candidate at the chromosome 5q locus impacting LDL-C levels. We also capitalized on the transparency of larval zebrafish to directly visualize vascular plaques in vivo, allowing for comprehensive analysis of both LDL-C levels and its consequences for atherosclerotic events. Multiple lines of evidence support the conservation of components of human lipid metabolism in zebrafish ⁴³. These include major classes of apolipoproteins ⁴⁴ as well as other proteins directly influencing LDL-C and atherogenesis, such as an ortholog of human CETP, which is not found in mouse ^{43, 45}. We have previously demonstrated the power of this combined genetic manipulation and in vivo visualization approach for modeling hypercholesterolemia and atherosclerosis in the context of a well-characterized familiar hypercholesterolemia gene, ldlr ²⁷. Here, we applied this model to identify a novel regulator of human hypercholesterolemia in APOOP1.

Although mouse models have proven to be invaluable for understanding many human phenotypes, they are less suitable for studies requiring rapid genetic screening and in vivo visualization. Moreover, since wild-type mice have a predominantly HDL cholesterol profile and are naturally resistant to atherosclerosis, humanized mouse models are required, such as Ldlr^−/−; Apobec1^−/−; Tg-hApoB ⁴⁶ to produce a lipoprotein profile more closely resembling human. In zebrafish, however, evidence suggests a substantial LDL-C fraction, especially upon introduction of a high cholesterol diet ³⁰. Investigating the role of APOOP1 in LDL-C metabolism in mice is also complicated by the fact that there are 3 ApoO pseudogenes in the mouse genome, none of which are syntenic to the human APOOP pseudogenes, and one of which (Apoo-ps, chromosome 13) encodes a non-coding RNA with unknown function. Zebrafish, in contrast, do not have any known expressed pseudogenes to apoo that might interfere with the function of APOOP1 and complicate interpretation of our observations.

In conclusion, these studies implicate a new pathway regulating LDL-C levels that involves APOOP1 and APOO. We propose that a SNP highly enriched in the Amish alters transcription and/or expression of the nearby APOOP1 pseudogene, which, in turn, interacts with the parent APOO gene, ultimately influencing LDL-C. Based on our zebrafish results, one plausible speculation is that the associated SNP may increase transcription of APOOP1, and increased transcripts may compete for critical binding sites in APOO, thereby downregulating its expression and/or function. Further studies are necessary to decipher the regulatory interaction between APOOP1 and APOO, which may be apparent only at the level of translation of APOO. Our data, however, support a role for both the pseudogene and APOO itself in regulation of LDL-C, and suggest that manipulation of APOO as a novel therapy for hyperlipidemia.

What Is New?

• We have identified a novel haplotype influencing LDL-C levels in the Amish population of Lancaster County, PA that is associated with an ~15 mg/dl increase in LDL-C.

• We have linked the high LDL-C phenotype to APOOP1, a transcribed pseudogene, and have provided a possible regulatory link to its parent gene, APOO, that encodes a newly identified apolipoprotein, ApoO.

• We propose that APOOP1 regulates levels of LDL-C in humans, potentially through its interaction with APOO, thus identifying a novel mechanism of lipid homeostasis.

What Are the Clinical Implications?

• Uncovering new genes associated with LDL-C levels and the pathways in which these genes reside is of high clinical relevance because such discoveries could reveal new therapeutic targets aimed at LDL-C reduction and CVD prevention.

• These studies implicate a new pathway regulating LDL-C levels that involves APOOP1 and APOO.

• Manipulation of APOO, or other components of the APOOP1 regulatory network, may have potential as a novel therapy for hyperlipidemia.

Appendix: TOPMed Members

Name	Institution(s)
Abe, Namiko	New York Genome Center
Abecasis, Goncalo	University of Michigan
Albert, Christine	Massachusetts General Hospital
Allred, Nicholette (Nichole) Palmer	Wake Forest Baptist Health
Almasy, Laura	Children’s Hospital of Philadelphia, University of Pennsylvania
Alonso, Alvaro	Emory University
Ament, Seth	University of Maryland
Anderson, Peter	University of Washington
Anugu, Pramod	University of Mississippi
Applebaum-Bowden, Deborah	National Institutes of Health
Arking, Dan	Johns Hopkins University
Arnett, Donna K	University of Kentucky
Ashley-Koch, Allison	Duke University
Aslibekyan, Stella	University of Alabama
Assimes, Tim	Stanford University
Auer, Paul	University of Wisconsin Milwaukee
Avramopoulos, Dimitrios	Johns Hopkins University
Barnard, John	Cleveland Clinic
Barnes, Kathleen	University of Colorado at Denver
Barr, R. Graham	Columbia University
Barron-Casella, Emily	Johns Hopkins University
Beaty, Terri	Johns Hopkins University
Becker, Diane	Johns Hopkins University
Becker, Lewis	Johns Hopkins University
Beer, Rebecca	National Institutes of Health
Begum, Ferdouse	Johns Hopkins University
Beitelshees, Amber	University of Maryland
Benjamin, Emelia	Boston University, Massachusetts General Hospital
Bezerra, Marcos	Fundação de Hematologia e Hemoterapia de Pernambuco - Hemope
Bielak, Larry	University of Michigan
Bis, Joshua	University of Washington
Blackwell, Thomas	University of Michigan
Blangero, John	University of Texas Rio Grande Valley School of Medicine
Boerwinkle, Eric	University of Texas Health
Borecki, Ingrid	University of Washington
Bowler, Russell	National Jewish Health
Brody, Jennifer	University of Washington
Broeckel, Ulrich	Medical College of Wisconsin
Broome, Jai	University of Washington
Bunting, Karen	New York Genome Center
Burchard, Esteban	University of California, San Francisco
Cardwell, Jonathan	University of Colorado at Denver
Carlson, Sara	University of Washington
Carty, Cara	Women’s Health Initiative
Casaburi, Richard	University of California, Los Angeles
Casella, James	Johns Hopkins University
Chaffin, Mark	The Broad Institute
Chang, Christy	University of Maryland
Chasman, Daniel	Brigham & Women’s Hospital
Chavan, Sameer	University of Colorado at Denver
Chen, Bo-Juen	New York Genome Center
Chen, Wei-Min	University of Virginia
Chen, Yii-Der Ida	Los Angeles Biomedical Research Institute
Cho, Michael	Brigham & Women’s Hospital
Choi, Seung Hoan	The Broad Institute
Chuang, Lee-Ming	National Taiwan University
Chung, Mina	Cleveland Clinic
Cornell, Elaine	University of Vermont
Correa, Adolfo	University of Mississippi
Crandall, Carolyn	University of California, Los Angeles
Crapo, James	National Jewish Health
Cupples, L Adrienne	Boston University
Curran, Joanne	University of Texas Rio Grande Valley School of Medicine
Curtis, Jeffrey	University of Michigan
Custer, Brian	Blood Systems Research Institute UCSF
Damcott, Coleen	University of Maryland
Darbar, Dawood	University of Illinois at Chicago
Das, Sayantan	University of Michigan
David, Sean	Stanford University
Davis, Colleen	University of Washington
Daya, Michelle	University of Colorado at Denver
de Andrade, Mariza	Mayo Clinic
DeBaun, Michael	Vanderbilt University
Deka, Ranjan	University of Cincinnati
DeMeo, Dawn	Brigham & Women’s Hospital
Devine, Scott	University of Maryland
Do, Ron	Icahn School of Medicine at Mount Sinai
Duan, Qing	University of North Carolina
Duggirala, Ravi	University of Texas Rio Grande Valley School of Medicine
Durda, Peter	University of Vermont
Dutcher, Susan	Washington University in St Louis
Eaton, Charles	Brown University
Ekunwe, Lynette	University of Mississippi
Ellinor, Patrick	Massachusetts General Hospital
Emery, Leslie	University of Washington
Farber, Charles	University of Virginia
Farnam, Leanna	Brigham & Women’s Hospital
Fingerlin, Tasha	National Jewish Health
Flickinger, Matthew	University of Michigan
Fornage, Myriam	University of Texas Health
Franceschini, Nora	University of North Carolina
Fu, Mao	University of Maryland
Fullerton, Malia	University of Washington
Fulton, Lucinda	Washington University in St Louis
Gabriel, Stacey	The Broad Institute
Gan, Weiniu	National Institutes of Health
Gao, Yan	University of Mississippi
Gass, Margery	Fred Hutchinson Cancer Research Center
Gelb, Bruce	Icahn School of Medicine at Mount Sinai
Geng, Xiaoqi (Priscilla)	University of Michigan
Germer, Soren	New York Genome Center
Gignoux, Chris	Stanford University
Gladwin, Mark	University of Pittsburgh
Glahn, David	Yale University
Gogarten, Stephanie	University of Washington
Gong, Da-Wei	University of Maryland
Goring, Harald	University of Texas Rio Grande Valley School of Medicine
Gu, C. Charles	Washington University in St Louis
Guan, Yue	University of Maryland
Guo, Xiuqing	Los Angeles Biomedical Research Institute
Haessler, Jeff	Fred Hutchinson Cancer Research Center, Women’s Health Initiative
Hall, Michael	University of Mississippi
Harris, Daniel	University of Maryland
Hawley, Nicola	Yale University
He, Jiang	Tulane University
Heavner, Ben	University of Washington
Heckbert, Susan	University of Washington
Hernandez, Ryan	University of California, San Francisco
Herrington, David	Wake Forest Baptist Health
Hersh, Craig	Brigham & Women’s Hospital
Hidalgo, Bertha	University of Alabama
Hixson, James	University of Texas Health
Hokanson, John	University of Colorado at Denver
Hong, Elliott	University of Maryland
Hoth, Karin	University of Iowa
Hsiung, Chao (Agnes)	National Health Research Institute Taiwan
Huston, Haley	Blood Works Northwest
Hwu, Chii Min	Taichung Veterans General Hospital Taiwan
Irvin, Marguerite Ryan	University of Alabama
Jackson, Rebecca	Ohio State University Wexner Medical Center
Jain, Deepti	University of Washington
Jaquish, Cashell	National Institutes of Health
Jhun, Min A	University of Michigan
Johnsen, Jill	Blood Works Northwest, University of Washington
Johnson, Andrew	NIH National Heart, Lung, and Blood Institute
Johnson, Craig	University of Washington
Johnston, Rich	Emory University
Jones, Kimberly	Johns Hopkins University
Kang, Hyun Min	University of Michigan
Kaplan, Robert	Albert Einstein College of Medicine
Kardia, Sharon	University of Michigan
Kathiresan, Sekar	The Broad Institute
Kaufman, Laura	Brigham & Women’s Hospital
Kelly, Shannon	Blood Systems Research Institute UCSF
Kenny, Eimear	Icahn School of Medicine at Mount Sinai
Kessler, Michael	University of Maryland
Khan, Alyna	University of Washington
Kinney, Greg	University of Colorado at Denver
Konkle, Barbara	Blood Works Northwest
Kooperberg, Charles	Fred Hutchinson Cancer Research Center
Kramer, Holly	Loyola University
Krauter, Stephanie	University of Washington
Lange, Christoph	Harvard School of Public Health
Lange, Ethan	University of Colorado at Denver
Lange, Leslie	University of Colorado at Denver
Laurie, Cathy	University of Washington
Laurie, Cecelia	University of Washington
LeBoff, Meryl	Brigham & Women’s Hospital
Lee, Seunggeun Shawn	University of Michigan
Lee, Wen-Jane	Taichung Veterans General Hospital Taiwan
LeFaive, Jonathon	University of Michigan
Levine, David	University of Washington
Levy, Dan	NIH National Heart, Lung, and Blood Institute, National Institutes of Health
Lewis, Joshua	University of Maryland
Li, Yun	University of North Carolina
Lin, Honghuang	Boston University
Lin, Keng Han	University of Michigan
Liu, Simin	Brown University, Women’s Health Initiative
Liu, Yongmei	Wake Forest Baptist Health
Loos, Ruth	Icahn School of Medicine at Mount Sinai
Lubitz, Steven	Massachusetts General Hospital
Lunetta, Kathryn	Boston University
Luo, James	NIH National Heart, Lung, and Blood Institute, National Institutes of Health
Mahaney, Michael	University of Texas Rio Grande Valley School of Medicine
Make, Barry	Johns Hopkins University
Manichaikul, Ani	University of Virginia
Manson, JoAnn	Brigham & Women’s Hospital
Margolin, Lauren	The Broad Institute
Martin, Lisa	George Washington University
Mathai, Susan	University of Colorado at Denver
Mathias, Rasika	Johns Hopkins University
McArdle, Patrick	University of Maryland
McDonald, Merry-Lynn	University of Alabama
McFarland, Sean	Harvard University
McGarvey, Stephen	Brown University
Mei, Hao	University of Mississippi
Meyers, Deborah A	University of Arizona
Mikulla, Julie	National Institutes of Health
Min, Nancy	University of Mississippi
Minear, Mollie	National Institutes of Health
Minster, Ryan L	University of Pittsburgh
Mitchell, Braxton	University of Maryland
Montasser, May E.	University of Maryland
Musani, Solomon	University of Mississippi
Mwasongwe, Stanford	University of Mississippi
Mychaleckyj, Josyf C	University of Virginia
Nadkarni, Girish	Icahn School of Medicine at Mount Sinai
Naik, Rakhi	Johns Hopkins University
Natarajan, Pradeep	The Broad Institute, Harvard University, Massachusetts General Hospital
Nekhai, Sergei	Howard University
Nickerson, Deborah	University of Washington
North, Kari	University of North Carolina
O’Connell, Jeff	University of Maryland
O’Connor, Tim	University of Maryland
Ochs-Balcom, Heather	University at Buffalo
Pankow, James	University of Minnesota
Papanicolaou, George	National Institutes of Health
Parker, Margaret	Brigham & Women’s Hospital
Parsa, Afshin	University of Maryland
Pattison, Jessica Tangarone	University of Michigan
Penchev, Sara	National Jewish Health
Peralta, Juan Manuel	University of Texas Rio Grande Valley School of Medicine
Perez, Marco	Stanford University
Perry, James	University of Maryland
Peters, Ulrike	Fred Hutchinson Cancer Research Center, University of Washington
Peyser, Patricia	University of Michigan
Phillips, Larry	Emory University
Phillips, Sam	University of Washington
Pollin, Toni	University of Maryland
Post, Wendy	Johns Hopkins University
Powers Becker, Julia	University of Colorado at Denver
Preethi Boorgula, Meher	University of Colorado at Denver
Preuss, Michael	Icahn School of Medicine at Mount Sinai
Prokopenko, Dmitry	Harvard University
Psaty, Bruce	University of Washington
Qasba, Pankaj	National Institutes of Health
Qiao, Dandi	Brigham & Women’s Hospital
Qin, Zhaohui	Emory University
Rafaels, Nicholas	University of Colorado at Denver
Raffield, Laura	University of North Carolina
Ramachandran, Vasan	Boston University
Rao, D.C.	Washington University in St Louis
Rasmussen-Torvik, Laura	Northwestern University
Ratan, Aakrosh	University of Virginia
Redline, Susan	Brigham & Women’s Hospital
Reed, Robert	University of Maryland
Regan, Elizabeth	National Jewish Health
Reiner, Alex	University of Washington
Rice, Ken	University of Washington
Rich, Stephen	University of Virginia
Roden, Dan	Vanderbilt University
Roselli, Carolina	The Broad Institute
Rotter, Jerome	Los Angeles Biomedical Research Institute
Ruczinski, Ingo	Johns Hopkins University
Russell, Pamela	University of Colorado at Denver
Ruuska, Sarah	Blood Works Northwest
Ryan, Kathleen	University of Maryland
Sakornsakolpat, Phuwanat	Brigham & Women’s Hospital
Salimi, Shabnam	University of Maryland
Salzberg, Steven	Johns Hopkins University
Sandow, Kevin	Los Angeles Biomedical Research Institute
Sankaran, Vijay	Harvard University
Scheller, Christopher	University of Michigan
Schmidt, Ellen	University of Michigan
Schwander, Karen	Washington University in St Louis
Schwartz, David	University of Colorado at Denver
Sciurba, Frank	University of Pittsburgh
Seidman, Christine	Harvard Medical School
Sheehan, Vivien	Baylor College of Medicine
Shetty, Amol	University of Maryland
Shetty, Aniket	University of Colorado at Denver
Sheu, Wayne Hui-Heng	Taichung Veterans General Hospital Taiwan
Shoemaker, M. Benjamin	Vanderbilt University
Silver, Brian	University of Massachusetts Memorial Health Center
Silverman, Edwin	Brigham & Women’s Hospital
Smith, Jennifer	University of Michigan
Smith, Josh	University of Washington
Smith, Nicholas	University of Washington
Smith, Tanja	New York Genome Center
Smoller, Sylvia	Albert Einstein College of Medicine
Snively, Beverly	Wake Forest Baptist Health
Sofer, Tamar	Brigham & Women’s Hospital
Sotoodehnia, Nona	University of Washington
Stilp, Adrienne	University of Washington
Streeten, Elizabeth	University of Maryland
Sung, Yun Ju	Washington University in St Louis
Sylvia, Jody	Brigham & Women’s Hospital
Szpiro, Adam	University of Washington
Sztalryd, Carole	University of Maryland
Taliun, Daniel	University of Michigan
Tang, Hua	Stanford University
Taub, Margaret	Johns Hopkins University
Taylor, Kent	Los Angeles Biomedical Research Institute
Taylor, Simeon	University of Maryland
Telen, Marilyn	Duke University
Thornton, Timothy A.	University of Washington
Tinker, Lesley	Women’s Health Initiative
Tirschwell, David	University of Washington
Tiwari, Hemant	University of Alabama
Tracy, Russell	University of Vermont
Tsai, Michael	University of Minnesota
Vaidya, Dhananjay	Johns Hopkins University
VandeHaar, Peter	University of Michigan
Vrieze, Scott	University of Colorado at Boulder, University of Minnesota
Walker, Tarik	University of Colorado at Denver
Wallace, Robert	University of Iowa
Walts, Avram	University of Colorado at Denver
Wan, Emily	Brigham & Women’s Hospital
Wang, Fei Fei	University of Washington
Watson, Karol	University of California, Los Angeles
Weeks, Daniel E.	University of Pittsburgh
Weir, Bruce	University of Washington
Weiss, Scott	Brigham & Women’s Hospital
Weng, Lu-Chen	Massachusetts General Hospital
Willer, Cristen	University of Michigan
Williams, Kayleen	University of Washington
Williams, L. Keoki	Henry Ford Health System
Wilson, Carla	Brigham & Women’s Hospital
Wilson, James	University of Mississippi
Wong, Quenna	University of Washington
Xu, Huichun	University of Maryland
Yanek, Lisa	Johns Hopkins University
Yang, Ivana	University of Colorado at Denver
Yang, Rongze	University of Maryland
Zaghloul, Norann	University of Maryland
Zhang, Yingze	University of Pittsburgh
Zhao, Snow Xueyan	National Jewish Health
Zhao, Wei	University of Michigan
Zheng, Xiuwen	University of Washington
Zhi, Degui	University of Texas Health
Zhou, Xiang	University of Michigan
Zody, Michael	New York Genome Center
Zoellner, Sebastian	University of Michigan