(n=826,beta=–29 mg/dl,P= 1.3 × 10–^5 ). We
also found that the mean level of AST in^352 Ser
homozygotes was twofold higher than that
of^352 Asn homozygotes (35.8 versus 18.4 U/L,
respectively, normal range 10 to 35 U/L;n=
6443, recessive model,P=5.9×10–^17 ), further
supporting the functional role of this variant.
Although the higher level of AST might
originate from liver, we found no association
with other liver dysfunction measures, such
as alanine aminotransferase (ALT) and the
liver to spleen density ratio (a measure of
fatty liver), or with alkaline phosphatase or
inflammatory markers (table S4). Moreover,
(^352) Ser homozygotes (n=12)hadnormal
levels of gamma-glutamyl transpeptidase,
activated partial thromboplastin time, pro-
thrombin time, and internationalized nor-
malized ratio (table S5). Creatine kinase levels,
a marker of muscle damage, was higher, but
was not significantly higher and still with-
in the normal range among 12^352 Ser homo-
zygotes (121 U/L) compared with 13^352 Asn
homozygotes (86.5 U/L) and 12 heterozygotes
(75.5 U/L) (normal range, 44 to 196 U/L for
males and 29 to 143 U/L for females), sug-
gesting that the origin of increased AST levels
maybefrommuscle.
B4GALT1is associated with CAD protection
The Amish cohort studied was relatively healthy,
with a low prevalence of CAD. Therefore, to
test the impact of B4GALT1 on risk of CAD, we
leveraged data from two large exome-sequencing
datasets (totalN> 500 K): Geisinger Health
System’s DiscovEHR and the UK Biobank
(UKBB). Given the rarity of p.Asn352Ser in the
general population (MAF < 3.6 × 10–^5 ), directly
testing the variant for association was not
possible. Therefore, we focused on aggregating
B4GALT1-coding variants by creating a multi-
variant burden test of rare (MAF < 0.01%)
predicted loss-of-function (pLOF) or deleterious
missense variants, including the small number
of p.Asn352Ser carriers, which was then used
in analyzing both the DiscovEHR and UKBB
data, followed by fixed-effects meta-analysis
( 21 ). With this burden test, we largely reca-
pitulated the constellation of associations
observed for p.Asn352Ser in the OOA, includ-
ing lower LDL-C [beta =–0.195 (–8 mg/dL),
SCIENCEscience.org 3 DECEMBER 2021•VOL 374 ISSUE 6572 1223
Fig. 3. HumanB4GALT1^352 Ser homozygotes show decreased galactosylation and sialylation.
(AtoC) Levels of total galactosylation (A), sialylation (B), and fucosylation (C) in global plasma
glycoproteins, fibrinogen, ApoB100, and IgG (mean ± SEM,n= 12 opposite homozygote-matched pairs).
Pvalues are based onttest. Red color indicates individuals carrying the known familial hypercholesterolemia
p.Arg3527GlnAPOBvariant.
Fig. 2. Gene-based association analyses of rare
B4GALT1pLOF and predicted deleterious
missense variants in DiscovEHR and UK BioBank
show CAD protection.(A) Association with LDL,
HDL, AST, and ALT are similar to OOA results. All
Pvalues are based onttest using the additive
genetic model. (B) Association with CAD protection
(n= 544,955).Pvalues are based on the likelihood
ratio test using the additive genetic model.
Meta-analysisPvalues are based onZtest.
RESEARCH | RESEARCH ARTICLES