decrease in body weight; 0.001 <P< 0.04
and 0.004 <P< 0.03 for heterozygotes and
homozygotes, respectively; analysis of variance
(ANOVA) was used for all animal group com-
parisons, andPvalues were calculated based
onFtest] (fig. S9A). Serum chemistry anal-
ysis revealed a 38% decrease of LDL-C in
(^353) Ser/ (^353) Ser mice (0.01 <P< 0.06; Fig. 5A),
and this reduction remained significant at
18 and 25 weeks after adjusting for body
weight. Consistent with the human pheno-
type, an increase of AST, but not ALT, was
observed in both^353 Asn/^353 Ser and^353 Ser/
(^353) Ser mice (14 and 50%, respectively,P< 0.02;
Fig. 5B and fig. S9F). No significant differences
were observed in total cholesterol, ApoB-100,
HDL-C, triglycerides, or non-esterified fatty
acids(figs.S9,BtoE,andS10).Finally,protein
analysis of plasma showed a 20% trending
decrease of fibrinogen (P< 0.09) and a trend
toward increased circulating IgG levels in
(^353) Ser/ (^353) Ser mice (P= 0.2; Fig. 5, C and D).
B4galt1knockdown in liver of adult mice
decreases circulating LDL-C levels
We further investigated whether down-regulation
ofB4galt1in mouse liver could recapitulate the
human phenotype. CRISPR-Cas9 was used to
knock down endogenous expression ofB4galt1
in livers of adult mice ( 21 ). Adult mice con-
stitutively expressing Cas9 enzyme were trans-
duced with an AAV8 vector containing guide
RNAs (gRNAs) targeting exon 2 ofB4galt1.
This approach resulted in an ~50% gene-editing
rate ofB4galt1in the liver, where edits included
deletions of 2 bp (40.4%), 1 bp (8.3%), and 3 bp
(2.8%) in exon 2 (Fig. 6, A and B). Further se-
quencing analysis of the mRNA from the liver
revealed that the majority of the persisting
transcripts (57.7%) had the 2-bp deletion seen
in the genomic DNA, and only 20.7% of the
transcripts were wild-type for B4galt1 (Fig. 6C).
These results suggested that the knockdown
ofB4galt1in the liver was greater than the
50% suggested by the genomic DNA editing
rate. Indeed, TaqMan analysis of B4galt1 in
liver showed a 50% decrease in mRNA levels
(Fig. 6D), with no changes inB4galt1mRNA
expression in the spleen (fig. S11). Minimal
B4galt1gene editing was observed in the spleen
(2%), a secondary target of AAV8 (fig. S12). We
then measured circulating LDL-C levels starting
at 2 weeks from viral transduction and through-
out a 12-week study period. An overall 50%
decrease of LDL-C was detected throughout
the duration of the study (0.001 <P< 0.06;
Fig. 6E). No significant changes in HDL-C or
total cholesterol were observed (fig. S13, A to D).
Additionally, there was a trend toward increased
circulating AST enzyme activity and no con-
sistent changes in ALT (fig. S13, E and F).
These findings were confirmed by two other
independent gRNAs designed against exon 2
ofB4galt1(figs. S14 and S15).
Finally, to ascertain whether the decrease
in the LDL-C was mediated specifically by
B4galt1knockdown, we generated a liver
knockdown of complement factor B (Cfb).
Cfbwas chosen as control because of its high
expression in liver and no known function
in LDL-C and cholesterol metabolism or gly-
cosylation. Although we achieved similar lev-
els of liver-specific gene editing (~50% in liver
versus <1% in spleen) (fig. S16), LDL-C levels
were not affected byCfbliver knockdown
(Fig. 6D). These results further corroborate the
functional link betweenB4galt1and LDL-C
metabolism and suggest that liver-specific
modulation ofB4galt1expression may be a
useful approach to reducing LDL-C and per-
haps CAD.
Discussion
Large genome-wide analyses of ~600,000 indi-
viduals have identified 386 loci associated with
lipid traits, none of which identified the
B4GALT1gene ( 36 ). The identified p.Asn352Ser
variant is conserved across 100 vertebrate
species [Genomic Evolutionary Rate Profiling
(GERP) score = 5.9] and is located in the
flexible long C-terminal region of the protein
that undergoes conformational changes to
allow for the exchange of the sugar molecule
during glycosylation ( 37 ). Therefore, as sup-
ported by our enzymatic assays, a mutation in
this region may impede the necessary confor-
mational change and affect glycosylation ef-
ficiency. Several studies support a role of protein
glycosylation in lipid metabolism ( 7 , 38 – 46 )
and CVD ( 46 , 47 ). Our results suggest a pre-
viously uncharacterized role ofB4GALT1in
LDL metabolism.
We also observed decreased fibrinogen levels
in^352 Ser individuals. Although Mendelian
randomization studies have not provided evi-
dence for a causal effect of fibrinogen on CVD
( 48 ), increased fibrinogen can increase CVD risk
through multiple pathways ( 4 ), and a synergistic
effect for LDL and fibrinogen on CVD cannot be
ruled out.
OurB4galt1 353Sergermline knock-in mouse
model phenocopied the LDL-C and fibrinogen
phenotype identified in humans. Additional
investigation of this mouse model and deeper
human phenotyping studies will be necessary
to better understand the mechanisms underlying
these observations.
The p.Asn352SerB4GALT1missense variant
does not appear to be associated with any
severe phenotype. OOA^352 Ser homozygotes,
UKBB heterozygotes for pLOF and deleterious
variants, and^353 Serb4galt1knock-in mice
all had increased serum AST levels without
elevations in other liver dysfunction biomarkers.
Further investigation of the source and conse-
quences of this observation will be required
whenB4GALT1is considered further as a
potential therapeutic target.
Previous mass spectrometry studies of N-
glycans from plasma and hepatic membrane
glycoproteins ofB4galt1-knockout mice re-
vealed a higher level of sialylation and ga-
lactosylation inB4galt1−/−and a shift from
type 2 to type 1 glycan chains ( 49 , 50 ), which
waslikelycausedbyenhancedactivityof
B3galt proteins ( 51 ). Our preliminary inves-
tigation showed that the level of type 1 chain
(corresponding to the beta-galactose 1,3 link-
age) was low and comparable between^352 Asn
and^352 Ser homozygotes (table S11). This ob-
servation suggests that our missense mutation
did not cause a glycan chain type switch, as was
observed in mice ( 51 ), and that the compensa-
tory action ofB3GALT1may differ between
mouse and human.
We provide human genetic and mouse
modeling evidence for an important role of
B4GALT1 and protein glycosylation in the
regulation of lipid metabolism and fibrino-
gen levels. Our data suggest that modula-
tion of B4GALT1 expression and/or activity
may have pleiotropic effects for cardiopro-
tection. Further understanding of the under-
lying mechanisms, including potential adverse
consequences, may divulge new targets and
pathways for the treatment and prevention
of CVD.
REFERENCESANDNOTES
- E. J. Benjaminet al.,Circulation 137 , e67–e492 (2018).
- R. H. Nelson,Prim. Care 40 , 195–211 (2013).
- T. W. Meade,J. Clin. Pathol. 50 , 13–15 (1997).
- R. Vilar, R. J. Fish, A. Casini, M. Neerman-Arbez,Haematologica
105 , 284–296 (2020). - L. A. Langeet al.,Am. J. Hum. Genet. 94 , 233– 245
(2014). - S. Suet al.,Atherosclerosis 200 , 213–220 (2008).
- I. Surakkaet al.,Nat. Genet. 47 , 589–597 (2015).
- T. J. Hoffmannet al.,Nat. Genet. 50 , 401–413 (2018).
- P.S.deVrieset al.,Hum. Mol. Genet. 25 , 358– 370
(2016). - A. Helgadottiret al.,Nat. Genet. 48 , 634–639 (2016).
- C. Sidoreet al.,Nat. Genet. 47 , 1272–1281 (2015).
- I. Moltkeet al.,Nature 512 , 190–193 (2014).
- R. L. Minsteret al.,Nat. Genet. 48 , 1049–1054 (2016).
- T. I. Pollinet al.,Science 322 , 1702–1705 (2008).
- H. Shenet al.,Arch. Intern. Med. 170 , 1850–1855 (2010).
- R. B. Horensteinet al.,Arterioscler. Thromb. Vasc. Biol. 33 ,
413 – 419 (2013). - E. Daleyet al.,J. Bone Miner. Res. 25 , 247–261 (2010).
- J.S.Albertet al.,N. Engl. J. Med. 370 , 2307– 2315
(2014). - F. K. Welty, J. Ordovas, E. J. Schaefer, P. W. Wilson,
S. G. Young,Circulation 92 , 2036–2040 (1995). - M. E. Montasseret al.,Circulation 138 , 1343–1355 (2018).
- Materials and methods are available as supplementary
materials. - D. Taliunet al.,Nature 590 , 290–299 (2021).
- R. Vaser, S. Adusumalli, S. N. Leng, M. Sikic, P. C. Ng,
Nat. Protoc. 11 ,1–9 (2016). - I. Adzhubei, D. M. Jordan, S. R. Sunyaev,Curr. Protoc. Hum. Genet.
Chapter 7, Unit 7.20 (2013). - J. H. Sul, B. Han, E. Eskin,J. Comput. Biol. 18 , 1611– 1624
(2011). - J. M. Schwarz, D. N. Cooper, M. Schuelke, D. Seelow,
Nat. Methods 11 , 361–362 (2014). - Y. Choi, A. P. Chan,Bioinformatics 31 , 2745– 2747
(2015). - B. Hansskeet al.,J. Clin. Invest. 109 , 725–733 (2002).
- M. Guillardet al.,J. Pediatr. 159 , 1041–1043.e2 (2011).
1226 3 DECEMBER 2021¥VOL 374 ISSUE 6572 science.orgSCIENCE
RESEARCH | RESEARCH ARTICLES