T H E J O U R N A L O F C E L L B I O L O G Y
The Rockefeller University Press $30.00
J. Cell Biol. Vol. 182 No. 3 531–542
Correspondence to Richard D. Cummings: email@example.com; or Tongzhong
Abbreviations used in this paper: CD: cytoplasmic domain; DSS, disuccinimidyl
suberate; ERAD, ER-associated degradation; GalNAc, N-acetylgalactosamine;
Gal-T, galactosyltransferase; GlcNAc, N-acetylglucosamine; HPC, human pro-
tein C; PNS, Postnuclear supernatant; TMD, transmembrane domain.
The online version of this paper contains supplemental material.
Metabolic pathways are commonly regulated at key steps termed
branchpoints, in which several key pathways diverge from a
single precursor. Such a branchpoint occurs in the biosynthesis
of O-glycans of animal cell glycoproteins, in which the common
precursor to all mucin-type O-glycans, N-acetylgalactosamine
(GalNAc) ? 1-Ser/Thr (Tn antigen), may be modifi ed by one of
several enzymes to generate different core structures known as
core 1, core 2, core 3, etc. The core 1 O-glycan is generated by
the T-synthase, also known as the core 1 ? 3 galactosyltransferase
(Gal-T), which adds galactose to the Tn antigen to generate
Gal ? 1-3GalNAc ? 1-Ser/Thr (T-antigen; Ju et al., 2002a , b ). This is
a key precursor for all core 1 and 2 mucin-type O-glycans in
vertebrates and invertebrates.
The overall pathway of mucin glycosylation regulated by
the branchpoint T-synthase function is developmentally impor-
tant because disruption of the T-synthase in mice is embryonic
lethal ( Xia et al., 2004 ). Alternatively, another branchpoint
enzyme that utilizes the Tn antigen is the C3GnT (core 3 ? 3- N -
acetylglucosaminyltransferase), specifi cally expressed in the
GI tract, which generates the disaccharide N-acetylglucosamine
(GlcNAc) ? 1-3GalNAc ? 1-Ser/Thr. Disruption of the core 3
O-glycan biosynthesis by eliminating C3GnT in mice is associated
with increased susceptibility to colitis and colorectal tumors
( An et al., 2007 ).
We observed that some cells and tissues in patients with
Tn syndrome or in some tumor cells lack T-synthase activity,
although they have the T-synthase transcript, and express the Tn
antigen, indicating a lack of branchpoint enzyme activity. In de-
ciphering the regulation of the T-synthase, we discovered that
the T-synthase requires a unique and apparently client-specifi c
chaperone that we termed Cosmc (core 1 ? 3 – Gal-T specifi c
molecular chaperone), which is required for formation of active
T-synthase ( Ju et al., 2002a , b ). Thus, the expression of the Tn
antigen in patients with Tn syndrome and in human tumor cells
results from mutations in Cosmc ( Ju and Cummings, 2005 ; Ju
et al., 2008 ). These discoveries prompted us to explore the molec-
ular nature of the potential chaperone function for Cosmc and
its role in assisting the T-synthase in acquiring its active form.
Somatic mutations in Cosmc result in loss of T-synthase
activity, apparently because of degradation of newly synthesized
the T-synthase, which directs synthesis of the common
core 1 O-glycan structure (T-antigen), the precursor structure
for most mucin-type O-glycans in a wide variety of glyco-
proteins. Formation of active T-synthase, which resides in
the Golgi apparatus, requires a unique molecular chaper-
one, Cosmc, encoded on Xq24. Cosmc is the only molecu-
lar chaperone known to be lost through somatic acquired
egulatory pathways for protein glycosylation are
poorly understood, but expression of branchpoint
enzymes is critical. A key branchpoint enzyme is
mutations in cells. We show that Cosmc is an endoplasmic
reticulum (ER) – localized adenosine triphosphate binding
chaperone that binds directly to human T-synthase. Cosmc
prevents the aggregation and ubiquitin-mediated degra-
dation of the T-synthase. These results demonstrate that
Cosmc is a molecular chaperone in the ER required for
this branchpoint glycosyltransferase function and show
that expression of the disease-related Tn antigen can re-
sult from deregulation or loss of Cosmc function.
Regulation of protein O-glycosylation by the
endoplasmic reticulum – localized molecular
Tongzhong Ju , Rajindra P. Aryal , Caleb J. Stowell , and Richard D. Cummings
Department of Biochemistry, Emory University School of Medicine, Atlanta, GA 30322
© 2008 Ju et al. This article is distributed under the terms of an Attribution–Noncommercial–
Share Alike–No Mirror Sites license for the fi rst six months after the publication date (see
http://www.jcb.org/misc/terms.shtml). After six months it is available under a Creative
Commons License (Attribution–Noncommercial–Share Alike 3.0 Unported license, as de-
scribed at http://creativecommons.org/licenses/by-nc-sa/3.0/).
JCB • VOLUME 182 • NUMBER 3 • 2008 532
and mouse T-synthase, the functions of other Cosmc orthologues
have not been tested.
Interestingly, T-synthase orthologues were identifi ed in both
vertebrates and invertebrates such as C. elegans and D. melano-
gaster . Whereas T-synthase in some lower vertebrates has non-
conserved N-glycosylation sites, mammalian T-synthases lack
N-glycosylation sites. In contrast, invertebrate T-synthases, as in
C. elegans ( Ju et al., 2002a , 2006 ), contain multiple N-glycosylation
sites, some of which appear to be conserved. Expression of in-
vertebrate T-synthase in insect cells does not require Cosmc for
active protein formation ( Ju et al., 2006 ), which may suggest that
the presence of N-glycosylation on invertebrate T-synthases may
allow them to be folded in a Cosmc-independent pathway.
Human Cosmc localizes in the ER and
T-synthase localizes to the Golgi apparatus
in CHO K1 cells
To explore the subcellular localization of Cosmc and T-synthase,
we prepared a C-terminal HA-tagged Cosmc (Cosmc-HA)
and a C-terminal human protein C (HPC) 4 – tagged T-synthase
(T-synthase – HPC4). CHO K1 cells were transiently transfected
with both constructs and immunostained with anti-HA and anti-
HPC4 antibodies. Anti-HA staining revealed a perinuclear pattern
of expression for HA-Cosmc, and a merge of this image showed
colocalization with calnexin, an ER marker ( Fig. 1, A – C ).
trast, anti-HPC4 staining revealed a punctate pattern for T-synthase
that was coincident with the localization of ? – mannosidase II,
a Golgi marker ( Fig.1, D – F ). These results demonstrate that
human Cosmc localizes in the ER, whereas the T-synthase local-
izes to the Golgi apparatus.
We also stained with Alexa 488 – labeled anti-HA (Fig.
S3, green, available at http://www.jcb.org/cgi/content/full/jcb.
200711151/DC1) and Alexa 568 – labeled Anti-HPC4 (Fig. S3,
red) in the same cotransfected cells. Cosmc expression was peri-
nuclear, corresponding to ER localization, and T-synthase stained
in a punctate pattern, corresponding to Golgi localization (Fig. S3),
which is consistent with the results from Fig. 1 . This result further
demonstrates that Cosmc and T-synthase localize to different
cellular compartments. To confi rm the localization of Cosmc in
the ER and T-synthase in the Golgi apparatus, we stained human
293T and MDA-MB-231 cells with rabbit polyclonal IgG pre-
pared against peptides from human Cosmc and T-synthase.
The staining pattern for anti-Cosmc was similar to that for anti-
Calnexin (ER marker), whereas the staining pattern for anti –
T-synthase was similar to that for anti – mannosidase II (Golgi
marker; unpublished data). These results are consistent with our
observations using anti – epitope-tagged forms of Cosmc and
T-synthase and further support the conclusion that Cosmc is primar-
ily localized in the ER and T-synthase is in the Golgi apparatus.
[ID]FIG1[/ID] In con-
Cosmc is in heavy membrane fractions
corresponding to the ER and T-synthase
is in lighter membrane fractions
corresponding to the Golgi apparatus
To further establish the localization of Cosmc in the ER, we per-
formed subcellular fractionation on sucrose gradients of 293T
cells expressing full-length HPC4-Cosmc. T-synthase activity
T-synthase via proteasome-dependent pathways ( Ju and Cummings,
2002 ). Cosmc appears to be specifi c for the T-synthase because
expression of many other glycosyltransferase activities is un-
affected by loss of functional Cosmc ( Piller et al., 1990 ; Ju
and Cummings, 2002 ), and the only consequence in terms of
glycoconjugate biosynthesis in cells accompanying loss of func-
tion in Cosmc is an increase in Tn and Sialyl Tn antigen ex-
pression. However, the precise location and function of Cosmc
as a specifi c molecular chaperone in T-synthase biosynthesis
has not yet been elucidated.
In this paper, we report that Cosmc predominantly local-
izes in the ER and that the T-synthase is mainly localized in the
Golgi apparatus. Importantly, Cosmc has ATP-binding activity
that is consistent with its chaperone function for maturation of
T-synthase. When the T-synthase is expressed in cells expressing
a mutated Cosmc that results in loss of function or in insect cells
that constitutively lack Cosmc, the enzyme aggregates and is
subsequently degraded in the proteasome. Thus, Cosmc serves a
unique function in the ER as the key posttranslational regulator
for expression of the T-synthase and may represent a new type of
chaperone functioning in regulating protein glycosylation.
Human Cosmc and T-synthase are
expressed coordinately in human tissue
To defi ne whether Cosmc and the T-synthase are coordinately ex-
pressed, we examined expression using Northern blot analysis of
multiple human tissue samples. Cosmc is expressed in all tissues
examined (Fig. S1 A, available at http://www.jcb.org/cgi/content/
full/jcb.200711151/DC1), with the highest expression in heart,
skeletal muscle, kidney, liver, and placenta, which mirrors the
expression of T-synthase (Fig. S1 B; Ju et al., 2002a ). The co-
ordinate expression of Cosmc and T-synthase in human tissues is
consistent with a close relationship between these two proteins.
Cosmc only exists in vertebrates and is
highly conserved across species
Upon BLASTP searching in the National Center for Biotech-
nology Information database using human Cosmc, we identi-
fi ed Cosmc orthologues in monkey, cow, mouse, rat, dog, bird,
frog, and zebrafi sh (Fig. S2, available at http://www.jcb.org/cgi/
content/full/jcb.200711151/DC1). No orthologues were found in
Caenorhabditis elegans or Drosophila melanogaster . All Cosmc
orthologues are predicted to be type II transmembrane proteins,
with a short N-terminal cytoplasmic domain (CD) and trans-
membrane domain (TMD) and a large C-terminal domain (Fig. S2).
Human and primate Cosmc both have 318 aa, and there is only
a single amino acid difference at V191I. In contrast, rodent
Cosmc contains 316 aa and has > 95% identity to human Cosmc
but has a 2-aa gap between positions 33 and 34. These 2 aa are
located at the beginning of the lumenal domain, which we show
to be the functional domain of Cosmc. Frog Cosmc is a 317-aa
protein with a gap in the sequence at position 33 compared with
human. Zebrafi sh Cosmc lacks 3 aa at the C terminus compared
with human Cosmc, ending at position 315. Except for mouse
Cosmc, which can act as a molecular chaperone for both human
533LOCALIZATION AND FUNCTION OF COSMC • Ju et al.
encoding a soluble Cosmc with a C-terminal KDEL-ER reten-
tion sequence (HPC4-sCosmc-KDEL). Cosmc function is gen-
erally rate limiting for T-synthase activity ( Ju and Cummings,
2002 ). Therefore, we cotransfected human 293T cells with either
HPC4-tagged full-length T-synthase alone or in addition to
the Cosmc constructs. Endogenous T-synthase activity was only
slightly promoted by HPC4-sCosmc, whereas coexpression
with the HPC4-sCosmc-KDEL signifi cantly increased T-synthase
activity ( Fig. 2 A ).
the media, whereas HPC4-sCosmc-KDEL was mainly retained
in cells ( Fig. 2, B and C ). These results show that in the ab-
sence of the CD, TMD, and an ER retention signal, Cosmc is
secreted from cells. The slight promotion of T-synthase activity by
HPC4-sCosmc may result from the small amount of protein that
transiently resides in the ER during its biosynthesis. Importantly,
both HPC4-sCosmc and HPC4-sCosmc-KDEL were functional
in insect cells in rescuing T-synthase activity (unpublished data).
As a control, we coexpressed T-synthase with sCosmc lacking
its engineered signal sequence. That form of the protein, which
[ID]FIG2[/ID] The HPC4-sCosmc was mainly secreted into
was mainly recovered in fractions 12 – 14, corresponding to the
fractions containing the Golgi enzyme marker ? 4 – Gal-T ( Fig. 1 G ).
In contrast, HPC4-Cosmc was recovered primarily in fractions
6 – 9, corresponding to fractions containing the ER markers
GRP78 and GRP94 ( Fig. 1 H ). These results indicate that Cosmc
is primarily localized in the ER, whereas the T-synthase is pri-
marily localized in the Golgi apparatus, which is consistent with
the immunofluorescent data ( Fig. 1, A – F ). Some Cosmc was
observed in light membrane fractions and may refl ect a small
fraction that shuttles between the ER and Golgi during potential
retrieval, but this remains to be explored.
Soluble Cosmc with an engineered ER
retention signal promotes T-synthase
activity equivalently to wild-type Cosmc
To test whether the TMD of Cosmc was important to localization
and function, we made two constructs: one encoding a soluble
N-terminal HPC4-tagged Cosmc (HPC4-sCosmc) and the other
Figure 1. Localization of human Cosmc and T-synthase . (A – F) Immuno-
fl uorescent staining. CHO K1 cells cultured on chambered slides were tran-
siently transfected with Cosmc-HA or with T-synthase – HPC4 and stained
with rat anti-HA IgG1 (green) and rabbit anti-calnexin IgG (red; A – C; bars,
4 μ m) or with mouse anti-HPC4 (green) and rabbit anti – ? -ManII (red; D – F;
bars, 8 μ m). (G and H) Sucrose gradient subcellular fractionation. 293T
cells transiently transfected with HPC4-Cosmc were harvested and homog-
enized. The postnuclear supernatant (PNS) was loaded onto a sucrose
gradient. After ultracentrifugation, 16 fractions ( ? 0.6 ml/fraction) were
collected and measured for both T-synthase and ? 4 – Gal-T activity (G) and
analyzed on Western Blot with anti-HPC4 and anti-KDEL (H).
Figure 2. KDEL-tagged soluble form of Cosmc functions as wtCosmc and
localizes in ER. (A – C) Function of KDEL-tagged soluble form of Cosmc.
293T cells were transiently transfected with plasmids encoding T-synthase –
HPC4, HPC4-sCosmc, HPC4-sCosmc-KDEL, and wtCosmc. Cell extracts
were prepared and T-synthase activity was measured. Bar 1 represents a
single value, bar 5 represents a mean of duplicate values with error bar
shown, and the remaining bars represent a mean of triplicate values with
error bars shown. The pink line indicates the endogenous level of T-synthase
activity and the green line represents the activity after transfection of T-synthase
alone (A). T-synthase and Cosmc in cell extracts and media were also
analyzed by Western blot with anti-HPC4 (B and C). Black lines indicate
that intervening lanes have been spliced out. (D – F) Localization of sCosmc-
KDEL. CHO K1 cells cultured on chambered slides transiently transfected
with HPC4-sCosmc-KDEL were immunofl uorescently stained with anti-HPC4
(green; D) and rabbit anti-calnexin IgG (red; E) and merged (F). The im-
ages were collected by confocal microscopy. Bars, 8 μ m.
JCB • VOLUME 182 • NUMBER 3 • 2008 534
Cosmc binds ATP
Many chaperones have ATP binding or ATPase activity, some with
low affi nity. Thus, we tested whether Cosmc has ATP binding
activity ( Itoh et al., 1995 ). Cosmc binds to ATP-Sepharose and
can be eluted with ATP, although the protein does not bind well
( Fig. 3 A ).
binant T-synthase to ATP-Sepharose. To further test whether
Cosmc binds ATP directly, we performed photolabeling using
8-azido ? -[ 32 P]ATP. Under ultraviolet exposure, the activated
azido group directly cross-links to the peptide bond in the ATP
binding region of Cosmc. Purifi ed Cosmc, but not T-synthase,
could be photo – cross-linked with 8-azido ? -[ 32 P]ATP. This bind-
ing is specifi c because it was signifi cantly inhibited by excesses
of cold ATP or GTP ( Fig. 3 B ). Typically, either ATP or GTP can
inhibit specifi c ATP binding to molecular chaperones ( Csermely
and Kahn, 1991 ; Saira Mian, 1993 ; Soti et al., 2002 ). These re-
sults demonstrate that Cosmc can bind ATP, which is similar to
other chaperones, whereas T-synthase does not bind ATP.
[ID]FIG3[/ID] In contrast, there was no signifi cant binding of recom-
Proteasome inhibitors cause accumulation
of full-length T-synthase expression but fail
to restore activity
Our previous study showed that T-synthase in Jurkat cells, which
have a mutated Cosmc , accumulates upon treatment of the cells
with lactacystin, a proteasome inhibitor ( Ju and Cummings,
2002 ), suggesting that the degradation of misfolded T-synthase
is targeted to the proteasome. Human carcinoma LSC cells
would be expected to remain in the cytosol, had no effect on
T-synthase activity either in mammalian cells or insect cells
(unpublished data). Together, these results show that the CD
and TMD of Cosmc are important for ER retention but are not
important in chaperone function and that a C-terminal KDEL
retention signal is as effective as the full-length protein in its
chaperone function toward T-synthase.
To confi rm the localization of HPC4-sCosmc-KDEL in the
ER, CHO K1 cells were transiently transfected and stained with
Alexa 488 – labeled HPC4 mAb. The HPC4-sCosmc-KDEL was
expressed in a perinuclear pattern that generally overlapped that
of the ER marker calnexin, a transmembrane protein ( Fig. 2, D – F ;
Wada et al., 1991 ; Bergeron et al., 1994 ). Some lack of overlap
in HPC4-sCosmc-KDEL expression and calnexin could result
from the fact that the HPC4-sCosmc-KDEL is soluble, and some
escapes to the Golgi and is secreted to media ( Fig. 2 C ).
Cosmc contains a single potential site of N-glycosylation
in its extreme C terminus at N313. Thus, we examined whether
Cosmc was N-glycosylated. About one-third of HPC4-sCosmc and
HPC4-sCosmc-KDEL were N-glycosylated (unpublished data),
based on sensitivity to Peptidyl N-glycanase F, an enzyme that
removes N-glycans. However, the normal transmembrane form of
Cosmc was not detectably N-glycosylated because there was no
change of electrophoretic mobility upon treatment with Peptidyl
N-glycanase F (unpublished data). Ineffi cient N-glycosylation of
Cosmc may result from the fact that the potential N-glycosylation
site is very close to the C terminus and is surrounded by acidic
amino acids, which can impair the effi ciency of N-glycosylation
( Whitley et al., 1996 ; Nilsson and von Heijne, 2000 ).
To test whether N-glycosylation might be important to
Cosmc function, we eliminated the N-glycosylation sequon by
substitution of N313 to Q313 (N313Q-Cosmc). The chaperone
function of N313Q-Cosmc was similar to that of the wtCosmc
both in mammalian cells and insect cells (unpublished data).
Thus, N-glycosylation of Cosmc is not required for its function,
which is consistent with the fact that membrane form or full-
length Cosmc is ineffi ciently N-glycosylated.
Cosmc itself does not have T-synthase
A previous study ( Kudo et al., 2002 ) indicated that the gene
we initially reported as Cosmc ( Ju and Cummings, 2002 ) pos-
sibly encoded a second T-synthase (termed core 1 ? 3 – Gal-T2),
based upon some sequence identity ( ? 22%) with T-synthase
and apparent restoration of T-synthase activity in human Jurkat
and LSC cells, which lack T-synthase activity. To clarify this
point, we directly tested whether HPC4-sCosmc contained
core 1 ? 3 – Gal-T activity. Cosmc lacked any detectable activity
(Fig. S4, A and B, available at http://www.jcb.org/cgi/content/
full/jcb.200711151/DC1), whereas the soluble form of T-synthase
contained high enzymatic activity (Fig. S4, A and B). Thus, Cosmc
is not a second T-synthase. This is consistent with our recent
analysis of mouse null mutants lacking the T-synthase, in which
all T-synthase activity in the homozygous animals is missing
( Xia et al., 2004 ). Although Cosmc may have an evolutionary
sequence relationship to T-synthase, Cosmc functions as a chaper-
one, and not as a second T-synthase.
Figure 3. Cosmc has ATP-binding activity . (A) Hi-5 insect cells were in-
fected with Baculovirus encoding either Cosmc-HA or HPC4-sT-synthase plus
wild-type Cosmc. Cells for Cosmc-HA and media from cell coexpressing
HPC4 – sT-synthase and wild-type Cosmc were harvested. The cell extracts
and media were loaded on ATP-Sepharose column for chromatography
and eluted with 50 mM ATP, respectively. The washes and eluates were
analyzed by Western blot with anti-HA for Cosmc and anti-HPC4 for T-syn-
thase, as indicated. (B) HPC4-sCosmc and HPC4 – sT-synthase (coexpressed
with wtCosmc) were expressed in Hi-5 cells and purifi ed from the media.
3 μ g of Cosmc and T-synthase were photolabeled with 8-Azido ? -[ 32 P]ATP,
analyzed on SDS-PAGE, and transferred to nitrocellulose membrane for
radioautogram and Western blot with anti-HPC4. Black line indicates that
intervening lanes have been spliced out.
535 LOCALIZATION AND FUNCTION OF COSMC • Ju et al.
In contrast, the steady-state level of T-synthase expression in
LSB cells and activity of the protein remained unchanged.
The truncated N-terminal peptide in LSC cells could result in
a soluble T-synthase, which might be secreted into the media.
Thus, the media was Western blotted with HPC4 mAb. There was
no T-synthase detected in the media from LSC cells; however, a
substantial amount of soluble T-synthase was secreted into the
media from LSB cells (Fig. S5, available at http://www.jcb.org/
cgi/content/full/jcb.200711151/DC1). These data indicate that
T-synthase synthesized in LSC cells is rapidly degraded by the
proteasome because degradation was blocked by MG-132. Inter-
estingly, truncated T-synthase from LSC cells was the same size
as the soluble form of T-synthase from LSB cell media, suggest-
ing that the truncation of T-synthase in LSC cells results from
loss of the cytoplasmic and TMDs.
Truncated and full-length T-synthase from
LSC cells localizes in rough ER, whereas
T-synthase from LSB cells localizes in the
As shown in the previous section, T-synthase is a transmembrane
protein localized in the Golgi, where it functions to synthesize
express the Tn antigen because of a lack of T-synthase activity
( Brockhausen et al., 1998 ), whereas LSB cells contain T-synthase
activity, which implies that LSC cells lack Cosmc function.
In other studies, we have found that LSC cells have dysfunctional
Cosmc, caused by an insertional mutation in the Cosmc gene ( Ju
et al., 2008 ). To investigate T-synthase in LSC cells, both LSC
and LSB cells were stably transfected with C-terminal HPC4-
tagged full-length human T-synthase. Human recombinant
T-synthase from LSB cell extract was enzymatically active and
full length ( ? 45 kD); however, T-synthase from LSC cell extracts
was smaller ( ? 38 kD), inactive, and there was a reduced protein
expression ( Fig. 4 A ), potentially caused by truncation or cleav-
age of the peptide from its N terminus by an unknown protease.
The low level of protein expression could be caused by either a low
rate of protein expression or high rate of degradation. Interestingly,
the C. elegans T-synthase, which does not require mammalian
Cosmc, is expressed at equal levels recombinantly in both LSB
and LSC cells ( Ju et al., 2006 ). Therefore, to explore whether
degradation was involved in reduced T-synthase expression in
LSC cells, the cells were treated with the proteasome inhibitor
MG-132. In the presence of MG-132, more full-length T-synthase
was recovered, although it was still inactive ( Fig. 4, A and B ).
Figure 4. Characterization of T-synthase ex-
pressed in LSC cells containing a dysfunctional
Cosmc. (A and B) MG-132 causes accumula-
tion of full-length T-synthase protein in LSC cells
but failed to restore its activity. LSC and LSB
cells stably expressing T-synthase – HPC4 were
treated with 10 μ M MG-132 for 12 h and har-
vested. Cell extract was made and one portion
was used for measuring T-synthase activity (A),
whereas the other portion was analyzed on
SDS-PAGE (30 μ g protein/lane) by Western
blot with anti-HPC4 (B). Error bars represent
± 1 SD from the mean. Black lines indicates
that intervening lanes have been spliced out.
(C – H) T-synthase expressed in LSC cells resides
mainly in heavy membrane fractions (RER),
whereas T-synthase from LSB cells is in light
membrane fractions (Golgi). LSC and LSB cells
as in A and B treated with 10 μ M lactacys-
tin were homogenized and cell homogenate
was fractionated on a sucrose gradient by ultra-
centrifugation. The fractions were collected
and the activities of ? 4 – Gal-T and T-synthase
in fractions were determined (C). The fractions
were also analyzed by Western blotting with
anti-HPC4 (D – F), anti-20S proteasome ? 1-sub-
unit (G), and anti-KDEL (GRP78 and GRP94; H).
In F, lane 2 represents the combination of
fractions 1 and 2. (I – O) T-synthase expressed
in LSC cells resides mainly in the lumen of the
RER and was associated with GRP78. The cell
PNS of LSC and LSB cells as in A and B was
digested with 40 μ g/ml trypsin in the presence
or absence of 0.2% Triton X-100, and T-syn-
thase was analyzed under reducing SDS-PAGE
by Western blotting with HPC4 mAb (I and J).
As controls, calnexin and 20S proteasome ? 1-
subunit were also analyzed by Western blot-
ting with their respective antibody (K and L).
T-synthase was purifi ed from LSC and LSB cells
and Western blotted with anti-HPC4 (M) and
anti-KDEL (GRP78; N). GRP78 in the cell ex-
tract corresponding to one-tenth of the material
was detected to confi rm comparable starting
amounts of material (O).
JCB • VOLUME 182 • NUMBER 3 • 2008 536
folded T-synthase in the ER of LSC cells might be a result of
nonproductive association with known molecular chaperones.
Many chaperones, such as calnexin/calreticulin and the UDP-
Glc:glycoprotein glucosyltransferase system, retain misfolded
N-glycosylated proteins in the ER and help them fold correctly.
However, this system cannot function with human T-synthase
because it lacks N-glycans. Thus, we tested whether the GRP78
or BiP machinery might associate with the inactive T-synthase
in the ER of LSC cells. We conducted coimmunoprecipitation
experiments in which the HPC4-tagged T-synthase expressed
in LSC cells was purifi ed, and this material was analyzed by
Western blot to probe for coprecipitating chaperones. GRP78
was coimmunoprecipitated with T-synthase from LSC cells but
not from LSB cells ( Fig. 4, M – O ). We did not detect HSP40 in
this coprecipitation, but it is possible that it might be present
because the antibody to HSP40 has low affi nity. More compre-
hensive proteomic approaches will be needed to identify all the
proteins that coprecipitate with inactive T-synthase. Neverthe-
less, these results show that inactive T-synthase in LSC cells is
associated with a protein chaperone, but that association clearly
cannot productively result in active folded enzyme.
T-synthase purifi ed from LSC cells treated
with MG-132 is partially ubiquitinated
The results in the previous section show that misfolded T-synthase
is degraded by an ER-associated degradation (ERAD) pathway,
which usually requires ubiquitination. To probe whether the
misfolded T-synthase is ubiquitinated, we performed a Western
blot using anti-ubiquitin antibody against the affi nity-purifi ed
full-length HPC4-tagged T-synthase from LSC cells treated
with MG-132. The gels were intentionally overloaded to en-
hance detectability of ubiquitinated species. No ubiquitinated
protein was detected in T-synthase purifi ed from LSB cell ex-
tracts with or without MG-132 treatment ( Fig. 5, A and B ).
In contrast, a portion of the inactive T-synthase from LSC cells
the core 1 structure (T-antigen). To explore the localization of
the inactive T-synthase in LSC cells, we performed subcellular
fractionation using sucrose gradient ultracentrifugation. Both
LSC and LSB cells were transfected to express human T-synthase-
HPC4. Because of the low expression of T-synthase in LSC
cells and less clear/organized images of the ER/Golgi in these
tumor cells, we used sucrose gradient subcellular fractionation
to localize Cosmc and the T-synthase in these cells. Subcellu-
lar fractionation studies, by Western blotting against the HPC4
epitope, showed that both the protein and activity of T-synthase
from LSB cells were in the light membrane fractions (fractions
10 – 13) as is the Golgi marker ? 4 – Gal-T ( Fig. 4, C and D ).
Interestingly, truncated T-synthase from LSC cells was recov-
ered in the heavy membrane fractions (fractions 5 – 8; Fig. 4 E ),
along with the ER markers 20S proteasome ? 1-subunit, GRP94,
and GRP78 ( Fig. 4, G and H ). In addition, the full-length
T-synthase from LSC cells treated with lactacystin was mainly
retained in heavy membrane fractions, with a small amount found
in the light membrane fractions corresponding to the Golgi
( Fig. 4 F ). These data support the conclusion that Cosmc func-
tions as an ER chaperone to assist the folding and/or acquisition
of T-synthase activity.
Truncated T-synthase from LSC cells
resides in the lumen of the rough ER
associated with GRP78
The soluble, truncated, and inactive forms of T-synthase from
LSC cells were recovered in heavy membrane fractions; how-
ever, they could be present in the lumen of the ER or associated
with the cytosolic face of the ER membrane. To address this
question, we treated the membranes with trypsin in the absence
or presence of Triton X-100. We anticipated that the soluble
truncated T-synthase in LSC cells might be associated with
the proteasome on the cytosolic side of the ER. Unexpectedly,
T-synthase from LSC cells, which was recovered in heavy ER
membranes, was not cleaved by trypsin in the absence of Triton
X-100 but was susceptible to trypsin action in the presence of
Triton X-100 ( Fig. 4 I ). As expected, T-synthase from LSB
cells, which is Golgi localized, was insensitive to trypsin in the
absence of Triton X-100 but was quantitatively lost by trypsin
treatment in the presence of Triton X-100 ( Fig. 4 J ). Calnexin, a
type-I ER membrane protein containing a cytoplasmic tail of
87 aa, was cleaved by trypsin with a shift of ? 10.5 kD and re-
tained the intact transmembrane and lumenal domains ( Fig. 4 K ).
In contrast, the 20S proteasome ? 1-subunit, which associates
with the cytosolic side of the ER, was sensitive to trypsin ( Fig. 4 L ).
Similarly, calnexin and the 20S proteasome ? 1-sunbunit from
both of the cell types were digested by trypsin in the presence of
Triton X-100 ( Fig. 4, I – L ). T-synthase from LSC cell extracts
(in the presence of Triton X-100) was decreased signifi cantly
even without trypsin treatment, which might be caused by par-
tial digestion by cellular (perhaps lysosomal) proteases. Collec-
tively, these results show that T-synthase from LSC cells, which
is inactive, is mainly retained within the ER lumen.
In spite of the data in the previous paragraph, T-synthase
lacks any described ER retention signals ( Gomord et al., 1999 ).
Thus, we considered whether retention of the inactive and mis-
Figure 5. T-synthase expressed in LSC Cells is partially ubiquitinated . LSC
and LSB cells stably expressing T-synthase – HPC4 were treated with MG-
132 for 12 h and subsequently harvested. T-synthase from 200 μ l of cell
extract (500 μ g of protein total) was purifi ed and analyzed by Western
blot with anti-ubiquitin (A) fi rst, and then with anti-HPC4 (B) after stripping
of the membrane.
537LOCALIZATION AND FUNCTION OF COSMC • Ju et al.
treated with MG-132 was observed as high molecular mass spe-
cies. Importantly, high molecular mass species from LSC cells
in the presence of MG-132 were stained with ubiquitin anti-
body. These results demonstrate that a fraction of the T-synthase
that accumulates in LSC cells in the presence of a proteasome
inhibitor is ubiquitinated.
T-synthase is present as disulfi de-bonded
covalent oligomers in the absence
Many molecular chaperones function to prevent protein oligo-
merization and subsequent degradation. To further investigate
the nature of T-synthase synthesized in the absence of Cosmc,
purifi ed T-synthase from LSC and LSB cells were compared by
Western blot under reducing conditions and after cross-linking
with disuccinimidyl suberate (DSS). The T-synthase from LSB
cells, which have normal wtCosmc, appears mainly as a mono-
mer under reducing conditions and as both monomers and dimers
in nonreducing gels ( Fig. 6 A ).
meric form of T-synthase is seen under reducing conditions
from LSC cells ( Fig. 6 A ). More striking is the presence of high
molecular mass oligomeric forms of T-synthase in LSC cells ob-
served in nonreducing gels ( Fig. 6 A ). These results demonstrate
that in the absence of Cosmc, T-synthase cannot form dimeric
enzyme but appears to form oligomeric complexes.
We explored the oligomeric state of T-synthase in the ab-
sence of Cosmc using the noncleavable cross-linker DSS. The
T-synthase from LSB cells was mainly found in dimeric form in
the presence of DSS ( Fig. 6 B , lanes 7 and 8). However, most of
the T-synthase from LSC cells was recovered as high molecular
mass aggregates in the presence of DSS ( Fig. 6 B , lanes 3 and 4).
To confi rm these results, we explored the oligomeric state of
T-synthase generated in an insect expression system, which lacks
a Cosmc orthologue. We examined T-synthase under reducing
conditions in both intact Sf-9 cells and cell extracts treated with
permeable DSS. In the absence of DSS or Cosmc, T-synthase is
recovered primarily as a monomer in cells and extracts ( Fig . 6 C ,
lanes 2 and 7). When coexpressed with Cosmc, we also recov-
ered T-synthase mainly as a monomer ( Fig. 6 C , lanes 4 and 9).
In contrast, treatment with DSS of cells expressing T-synthase
but lacking Cosmc coexpression resulted in a quantitative shift
of T-synthase to high molecular mass complexes ( Fig. 6 C ,
lanes 3 and 8). However, coexpression of Cosmc and T-synthase
rescued much of the T-synthase from cross-linking and a signif-
icant amount of T-synthase monomer was recovered ( Fig. 6 C ,
lanes 5 and 10). These results indicate that in the absence of
Cosmc, T-synthase appears in oligomeric and inactive complexes,
whereas in the presence of Cosmc, T-synthase appears as a mono-
mer and dimer, demonstrating that Cosmc helps to prevent non-
productive oligomerization of T-synthase.
We also explored the effects of blocking proteasomal deg-
radation on the oligomeric state of T-synthase in LSC and LSB
cells. When LSB cells were treated with lactacystin, there was
little change in the amount of monomeric T-synthase recovered
( Fig. 6 B , lane 9), and after DSS treatment of cells, there were
increased amounts of the dimeric form, as expected because
natural T-synthase is a homodimer, and a small amount of oligomers
[ID]FIG6[/ID] In contrast, mainly the mono-
Figure 6. T-synthase from cells in the absence of a functional Cosmc is
mainly in covalent oligomers. T-synthase – HPC4 expressed in LSC and LSB
cells treated with 10 μ M lactacystin was affi nity purifi ed and analyzed
on SDS-PAGE under reducing and nonreducing conditions and Western
blotted with anti-HPC4 (A). Cell extracts were made and cross-linking was
performed by adding 5 mM DSS. The recombinant T-synthase was puri-
fi ed and analyzed by Western blotting under reducing conditions (B). Sf-9
cells infected with Baculovirus-expressing T-synthase-HPC4 or coinfected
with Baculovirus encoding wtCosmc were harvested and one portion was
made for cell extracts. The cell extracts and the other portion of intact cells
were treated with DSS. T-synthase was purifi ed from the cell extracts and
analyzed by Western blot with anti-HPC4 (C). Black line indicates that
intervening lanes have been spliced out.
form perhaps because of limited endogenous Cosmc ( Fig. 6 B ,
lane 10). In contrast, treatment of LSC cells with lactacystin
caused a larger increase in total T-synthase ( Fig. 6 B , lane 5),
JCB • VOLUME 182 • NUMBER 3 • 2008 538
Expression of the Tn and sialyl-Tn antigens in human tumors
are associated with poor prognosis, including human breast car-
cinoma, and colorectal and esophageal cancer ( Springer, 1984 ;
Nakagoe et al., 2001 ; Schietinger et al., 2006 ). However, until
now the nature of Cosmc function in providing active T-synthase
was unknown. We originally considered that Cosmc might be
a “ subunit ” of the T-synthase, but we found that purifi ed active
T-synthase from rat liver and purifi ed recombinant soluble form
of T-synthase from 293T cell media were fully active and devoid
of Cosmc ( Ju et al., 2002a , b ). Thus, Cosmc is not a required co-
factor or subunit of active T-synthase enzyme.
Our studies demonstrate that Cosmc is a resident ER pro-
tein that binds ATP, whereas the T-synthase is a resident Golgi
protein. In the absence of Cosmc, the T-synthase is retained
in the ER lumen and accumulates as an enzymatically inactive
oligomeric disulfi de-bonded complex associated with GRP78.
Proteins that are misfolded are eliminated from the ER by an
ERAD pathway, in which misfolded proteins are targeted for
reverse translocation out of the ER, where they are ubiqui-
tinated, and then disposed of in the cytoplasmic proteasome
( Ahner and Brodsky, 2004 ). The inactive T-synthase is sub-
jected to degradation by the ERAD pathway, as indicated by
the presence of ubiquitinated T-synthase, which accumulates
in the presence of proteasome inhibitors. The ER retention
of Cosmc requires its transmembrane and cytosolic domains
because soluble Cosmc is secreted from cells. However, soluble
Cosmc engineered to contain a KDEL ER-retrieval signal is
retained in the ER and functions as a chaperone for T-synthase.
Human Cosmc and T-synthase are coordinately expressed in
human tissues, indicating the close relationship between these
with enhanced recovery of high molecular mass aggregates of
T-synthase in DSS-treated cells ( Fig. 6 B , lane 6). These results
show that T-synthase in cells expressing functional Cosmc is
primarily in active monomeric and dimeric forms, whereas in
the absence of Cosmc, the T-synthase is partly degraded and re-
coverable in high molecular weight complexes, and no dimeric
active forms of enzyme are present, indicating that proteolysis
occurs in the context of the oligomeric complexes.
Cosmc directly interacts with T-synthase
Many molecular chaperones function to prevent protein oligo-
merization by directly interacting with their client proteins.
To investigate the nature of Cosmc and T-synthase interaction,
we purifi ed soluble HPC4-tagged T-synthase and His-tagged
Cosmc from Hi-5 cells and tested for their direct interaction
in vitro. As a control, we used galectin-3, which would not be
expected to bind Cosmc. After mixing Cosmc with either of
these proteins, Cosmc was quantitatively captured by adsorp-
tion on Ni-NTA beads and eluted with imidazole. The input
(loading), unbound, and bound materials were analyzed by
Western blot with anti-Cosmc, anti-HPC4, or anti – galectin-3
antibodies. Cosmc was quantitatively bound by Ni-NTA beads,
whereas T-synthase did not bind to Ni-NTA beads ( Fig. 7,
A and B ).
a fraction of T-synthase coprecipitated with Cosmc ( Fig. 7,
A and B ). We tested whether this interaction was specifi c to the
client T-synthase using galectin-3 as a potential client. Galectin-3
did not bind to Ni-NTA beads and also did not coprecipitate
with Cosmc on Ni-NTA beads ( Fig. 7, C and D ). These results
show that Cosmc directly binds to T-synthase and that this inter-
action is specifi c, supporting the conclusion that Cosmc is a
specifi c molecular chaperone for its client T-synthase and that
they directly interact.
[ID]FIG7[/ID] In contrast, from the mixture of Cosmc with T-synthase,
Our interest in T-synthase and its regulation as a branchpoint
enzyme arose from observations that the lack of T-synthase ac-
tivity, which leads to expression of the truncated Tn antigen,
is correlated with several autoimmune diseases including IgA
nephropathy ( Allen et al., 1997 ) and Tn-Syndrome ( Berger, 1999 ).
In addition, the Tn and sialyl-Tn antigen are recognized as tumor-
associated antigens in mucins and other glycoproteins from
many human tumors ( Springer, 1984 ; Nakagoe et al., 2001 ;
Schietinger et al., 2006 ). In examining human Jurkat cells that
lacked T-synthase activity, we found that expression of the
T-synthase transcript was normal but, surprisingly, active forms
of recombinant T-synthase could not be expressed in cells. This
led to our fi nding of Cosmc as a required molecular chaperone
for active T-synthase formation and the fi nding that the X-linked
Cosmc gene [Xq24] is mutated in cells lacking T-synthase activ-
ity ( Ju and Cummings, 2002 ). Thus, acquired somatic or poten-
tially inherited mutations in Cosmc could account for the defects
in core 1 O-glycan biosynthesis and expression of Tn antigen.
We recently showed that mutations in Cosmc are associated with
Tn antigen expression in patients with Tn syndrome ( Ju and
Cummings, 2005 ) and in human tumor cells ( Ju et al., 2008 ).
Figure 7. Cosmc directly interacts with T-synthase. The N-terminal His-
tagged soluble Cosmc were expressed in Hi-5 insect cells. The human
N-terminal HPC4 epitope-tagged soluble T-synthase was expressed by co-
transfecting WT Cosmc in Hi-5 insect cells. Proteins were purifi ed directly
from the media and used in coprecipitation experiments to test physical as-
sociation. Galectin-3 was used as a negative control for specifi city. Cosmc,
T-synthase, Cosmc and T-Synthase, Galectin-3, Cosmc and Galactin-3 were
incubated with Ni-NTA Superfl ow. Ni-NTA Superfl ow beads were washed
and proteins were eluted and analyzed by Western blot with mAb to the
HPC4, anti-Galectin-3, and anti-Cosmc. Cosmc interactions with T-synthase
were determined (A and B). As a control, Cosmc and Galectin-3 inter-
actions were determined (C and D). Black line indicates that intervening
lanes have been spliced out. L, loading; U, unbound; B, bound.
539LOCALIZATION AND FUNCTION OF COSMC • Ju et al.
binds directly to ATP, thus indicating that Cosmc is a chaperone.
In other experiments, when we purifi ed the T-synthase from rat
liver microsomes, which included the ER and Golgi compart-
ments, we found that both Cosmc and HSP40 copurifi ed in early
steps with the T-synthase (unpublished data). Although this fur-
ther supports a role for Cosmc as a T-synthase chaperone, it also
indicates that HSP40 could be a cochaperone for Cosmc as it is
for HSP70 ( Kleizen and Braakman, 2004 ).
In order for the misfolded T-synthase to be ubiquitinated
by ubiquitination system, it must be retrotranslocated from the
lumen back to the cytosol. However, little is known about the
mechanisms of retrotranslocation for type II membrane proteins
and how such unfolded proteins that accumulate in the ER are
degraded by ERAD pathway. The degradation of T-synthase,
the fi rst identifi ed enzyme with a defi ciency in its ER chaperone
protein that leads to disease, may provide a novel system to ex-
plore these pathways.
Thus, Cosmc represents a protein-specifi c quality control
factor and joins a growing number of such chaperones in the
ER ( Hendershot and Bulleid, 2000 ; Ellgaard and Helenius,
2003 ). Cosmc is not a T-synthase subunit, lacks glycosyltrans-
ferase activity, and, among glycosyltransferases, it is only known
to associate with the inactive T-synthase. We noted that Cosmc
does not associate with fully folded T-synthase ( Ju and Cummings,
2002 ), but it can be coimmunoprecipitated with a subset of
T-synthase proteins, likely representing folding intermediates
in the pathway to fully active enzyme. Most importantly, we
have shown that purifi ed recombinant Cosmc can directly in-
teract with purifi ed T-synthase. As expected, only a small frac-
tion of recombinant T-synthase coprecipitated with purified
recombinant His-tagged Cosmc, which could result from the
possibility that only a fraction of purifi ed recombinant T-synthase
might be in unfolded state upon purification. This fraction
therefore transiently interacts with active recombinant His-
tagged Cosmc, supporting the fact that Cosmc is an ER molec-
ular chaperone, but this issue will require extensive studies,
including exploring the role of ATP in these interactions. Loss-
of-function of Cosmc results in the formation of misfolded aggre-
gates of T-synthase polypeptide, which would be predicted to
be retrotranslocated back to the cytosol where ubiquitination
two proteins. Active T-synthase cannot be expressed without
the presence of functional Cosmc ( Ju and Cummings, 2002 ).
More importantly, purifi ed recombinant Cosmc directly inter-
acts with purifi ed recombinant T-synthase ( Fig. 7 ). Thus, Cosmc
is an ER-localized molecular chaperone that prevents degrada-
tion and aggregation of mammalian T-synthase, and the overall
proposed pathway is shown in Fig. 8 .
Cosmc itself does not have any T-synthase activity, which is in
contrast to a previous study that erroneously concluded that the
Cosmc gene encoded a second core 1 ? 3 – Gal-T (T-synthase) in
humans named C1Gal-T2 ( Kudo et al., 2002 ). The T-synthase
activity is encoded by a single gene, which is consistent with
our studies in this paper showing that loss of Cosmc function is
associated with a complete loss of T-synthase activity, that dele-
tion of T-synthase in mice eliminates all T-synthase activity and
results in embryonic death ( Xia et al., 2004 ), and that mutations
in Cosmc are associated with loss of T-synthase in patients with
Tn syndrome ( Ju and Cummings, 2005 ).
The T-synthase is not a glycoprotein and has no canonical
motifs for N-glycosylation, which is in contrast to most glycosyl-
transferases. Thus, the T-synthase, unlike other newly synthe-
sized glycoproteins in the ER, cannot interact with the known
quality control systems composed of several lectin-based chaper-
ones, such as the calnexin/calreticulin system, and other glycan-
recognizing or modifying proteins ( Ruddock and Molinari, 2006 ;
Schroder, 2006 ; Otsu and Sitia, 2007 ). The absence of such inter-
actions may underscore the molecular function of Cosmc in
specifi cally preventing oligomerization and inactivation of the
T-synthase and provides the fi rst specifi c evidence for the quality
control of nonglycosylated protein folding in the ER.
It is interesting that although the ER contains many cha-
perones that can promote proper protein folding, in the absence
of Cosmc these chaperones are not successful in promoting
folding of the T-synthase. We found that GRP78 (BiP) could be
coimmunoprecipitated with inactive T-synthase generated in the
absence of Cosmc. GRP78 is known to associate with unfolded
proteins in the ER to aid in their exit ( Kleizen and Braakman,
2004 ; Ni and Lee, 2007 ). The inactive oligomers of T-synthase
accumulate within the ER in the absence of Cosmc and reside
within disulfi de-bonded complexes. We also found that Cosmc
[ID]FIG8[/ID] It is important to note that
Figure 8. Working model for Cosmc function.
Human Cosmc is localized predominantly in
the ER where it interacts with the nascent poly-
peptide of human T-synthase. Cosmc, along with
other ER chaperones such as HSP70 (BiP)/
HSP40 and protein disulfi de isomerase (PDI),
assists its folding properly. Native T-synthase,
which occurs mainly as a homodimer, exits
to the Golgi apparatus, where it functions in
synthesizing core 1 O-glycans (T-antigen). When
Cosmc is mutated and dysfunctional, the nascent
polypeptides of T-synthase form inactive aggre-
gates or oligomers, which are associated with
GRP78 (BiP), and subsequently proceed to the
ERAD pathway where they are retrotranslocated
from the ER to the cytosol, ubiquitinated, and
subsequently degraded by the proteasome.
JCB • VOLUME 182 • NUMBER 3 • 2008 540
of pcDNA3.1(+) containing wild type Cosmc ( Ju et al., 2002a ). A con-
struct encoding a soluble N-terminal HPC4 epitope-tagged Cosmc (HPC4-
sCosmc) was made using the similar strategy to making soluble N-terminal
HPC4 epitope-tagged T-synthase (HPC4 – sT-synthase; Ju et al., 2002a ).
The construct in pcDNA3.1(+) or pVL1393 with an HPC4-epitope tag was
fused inframe with the lumenal domain of Cosmc. HPC4-sCosmc with the
KDEL-retrieval signal at its C terminus (HPC4-sCosmc-KDEL) was constructed
using PCR and subcloning into pcDNA3.1(+). The PCR primers are listed
in Table I .
The construct for expression of a soluble N-terminal 6 × His-tagged
Cosmc was prepared using the following strategy. The N-terminal cytoplas-
mic and TMDs of Cosmc were replaced with transferrin N-terminal signal
sequence followed by 6 × His tag. The lumenal domain of Cosmc was pre-
pared by the digestion of full-length Cosmc cDNA in pcDNA3.1(+) with
BsmI – XbaI. The Vector with pVL1393 backbone and transferrin N-terminal
signal sequence was obtained by digestion of the plasmid encoding hu-
man soluble T-synthase with EcoNI – XbaI. The DNA oligos (IDT [Integrated
DNA Technologies]) encoding 6 × His coding sequence with EcoNI and
BsmI at each site were synthesized and would fi ll in the gap between the
EcoNI site from the vector and the BsmI site from the Cosmc fragment. The
DNA oligos were the following: sense strain, 5 ? -CCCATCACCATCACCAT-
CACGACGATGACGATAAGAGGATTGGTCATGGAAATAGAATGCA-3 ? ;
and the complementary strain, 5 ? - CATTCTATTTCCATGACCAATCCTCT-
TATCGTCATCGTCGTGATGGTGATGGTGATGG-3 ? . Then the vector fragment,
the annealed oligos, and the Cosmc fragment were ligated and the con-
struct was made and confi rmed by sequencing.
293T, CHO K1, LSC, and LSB cells were cultured in DME plus 10% FBS at
37 ° C in 5% CO 2 . Sf-9 cells were cultured in TNM-FH, whereas Hi-5 cells
were cultured in Ex-Cell 405 media at 27 ° C.
Transfection and glycosyltransferase activity
Mammalian cells were transfected with Fugene6 transfection reagent and
insect cells were transfected with a BaculoGold Transfection kit ( Ju and
Cummings, 2002 ). T-synthase activity was measured using GalNAc- ? -phenyl
as the acceptor ( Ju et al., 2002b ). ? 4 – Gal-T activity was measured using
pNP- ? -S-GlcNAc as the acceptor ( Kawar et al., 2002 ).
Immunofl uorescent staining of CHO K1 cells
CHO K1 cells were cultured on a chambered slide and transiently trans-
fected with the expression constructs using Fugene 6 transfection reagent
according to the manufacturer ’ s protocol. At 48 h after transfection, cells
were washed with PBS and fi xed with 4% PFA on ice for 45 min and per-
meabilized with 0.1% Triton X-100 for 30 min on ice. After blocking with
1% BSA in PBS for 1 h at RT, the cells were incubated with primary anti-
bodies for 1 h at RT. The cells were washed with PBS three times and
incubated with Alexa Fluor – labeled secondary antibodies at RT for 1 h.
In some cases, the transfected cells were stained with Alexa Fluor – labeled
primary antibodies directly, as indicated in the fi gures. Cells were then
washed four times with PBS and mounted with Prolong Antifade Media
(Invitrogen). After drying at RT for 12 – 16 h, cells were visualized on a
confocal microscope (TCS NT; Leica) at RT under 40 × Plan Fluotar 1.0 NA
oil immersion or 100 × Plan APO 1.4 NA oil immersion objective lenses.
The images were maximum projection collected with a pinhole of 1 using
0.5- μ m step size. Images were analyzed using the TCS and Volocity soft-
Preparation of cell extracts
Cell extracts were prepared as previously described ( Ju and Cummings,
2002 ). Protein concentration was determined by BCA assay according to
manufacturer ’ s protocol.
occurs, and the ubiquitinated protein is then degraded in the
proteasome, as shown in the model ( Fig. 8 ). Overall, our study
shows that Cosmc is a key regulator of the expression of functional
T-synthase, which represents a novel mechanism for controlling
a branchpoint glycosyltransferase in O-glycan biosynthesis.
Materials and methods
GalNAc- ? -phenyl, UDP-Gal, GlcNAc- ? -S-pNp, and ATP-Sepharose were
obtained from Sigma-Aldrich. UDP-6-[ 3 H]Gal (40 – 60 Ci/mmol) was ob-
tained from American Radiolabeled Chemicals, Inc. Human embryonic
kidney cell line (HEK293T), Chinese hamster ovary cells (CHO K1), and
insect cells (Hi-5 and Sf-9) were obtained from American Type Culture Col-
lection. Human colon carcinoma cell lines LSC and LSB were provided by
S. Itzkowitz (Mount Sinai School of Medicine, New York, NY). Sep-Pak
C18 Cartridges were obtained from Waters Corporation. Restriction en-
zymes were obtained from New England Biolabs, Inc. FuGENE6, Taq
DNA polymerase, and rat anti-HA mAb were obtained from Roche. TNM-FH
and EX-Cell 405 media, BaculoGold Transfection kit, vector pVL1393,
mouse anti – human calnexin mAb (IgG1), mouse anti- – human HSP40,
and anti-ubiquitin mAb (IgG1) were purchased from BD Biosciences.
Galactin-3 antibody was purchased from Santa Cruz Biotechnology, Inc.
Rabbit anti – human calnexin antiserum and mouse anti-KDEL (GRP78 and
GRP94) mAb (10C3) were purchased from Assay Designs. Alexa Fluor –
labeled secondary antibodies were purchased from Invitrogen. Peroxidase-
labeled secondary antibodies were obtained from KPL. Proteasome
inhibitors MG-132 and lactacystin and rabbit anti – human proteasome
20S ? -type1 subunit (IgG) were purchased from EMD. Vector pcDNA3.1(+),
PCR TOPO4 cloning kit, SuperScript One-Step RT-PCR kit and SDS-PAGE
gels were obtained from Invitrogen. Ni-NTA Superfl ow, Plasmid Purifi -
cation kit, and QIAquick Gel Extraction kits were obtained from QIAGEN.
8-Azido ? -[ 32 P]ATP was purchased from Affi nity Labeling Technologies.
Chemiluminescent Substrate, BCA protein assay kit, DSS, and UltraLink
Support Media were purchased from Thermo Fisher Scientifi c. Rabbit anti –
mannosidase II antibody was provided by K. Moremen (University of Georgia,
Northern blot of human Cosmc and T-synthase
Human 12-lane multiple tissue Northern blot (Clontech Laboratories, Inc.)
was incubated with 32 P-labeled probes for full length of human Cosmc
(970 bp) as previously described ( Ju et al., 2002a ) at 68 ° C overnight
using PerfectHyb solution (Sigma-Aldrich). After washing, the blot was ex-
posed to BioMax x-ray fi lm (Kodak) for 36 h and developed. ? -Actin in the
same blot was probed with 32 P-labeled human Actin cDNA as previously
described ( Ju et al., 2002a ).
Preparation of expression constructs
A construct encoding C-terminal HA-tagged Cosmc (Cosmc-HA) was made
by introducing HA-epitope (YPYDVPDYA) into wild-type Cosmc at its C ter-
minus by PCR. The product was cloned into PCR3.1. The insert was cut
with BamHI (partially) – XbaI and cloned into pcDNA3.1(+) or pVL1393.
The construct expressing C-terminal HPC4 epitope-tagged T-synthase was
made as previously described ( Ju and Cummings, 2002 ). A construct en-
coding N-terminal HPC4-tagged Cosmc (HPC4-Cosmc) was made using a
similar strategy to Cosmc-HA. The HPC4 epitope tag was introduced into
the N terminus of Cosmc by PCR. The PCR product was digested with
BamHI and ligated with the vector fragment generated by BamHI digestion
Table I. PCR primers used for making expression constructs
Constructs Forward primerReverse primer
Cosmc-HA5 ? -GCGGATCCACCATGCTTTCTGAAAGCAGC
5 ? -GCGGATCCACCATGCTTGAGGACCAGGTGGA
AAGCAGCTCC-3 ? (containing HPC4-tag sequence)
5 ? -TCCTTGCACGCCCCACTACG-3 ?
5 ? -GCTCTAGACTAAGCGTAGTCTGGGACGTCGTATGGGTAG
TCATTGTCAGAACCATTTG-3 ? (containing HA-tag sequence)
5 ? -GGTCTCCAGATTTTATAGTGTGGC-3 ?
HPC4-sCosmc-KDEL5 ? -GCTCTAGAGTTCATCTTTGTCATTGTCAGAACCATTTG-3 ?
(containing KDEL sequence)
541 LOCALIZATION AND FUNCTION OF COSMC • Ju et al.
DSS and 1% DMSO at RT for 30 min. The other portion of Sf-9 cells was
treated with 5 mM DSS at RT for 30 min and the cell extracts were made
as previously described. The reaction was quenched by adding 50 mM
Tris-HCl, pH 7.5. After adding 1 mM CaCl 2 to all cell extracts, the T-synthase –
HPC4 was purifi ed and analyzed on Western blot with anti-HPC4.
In vitro Cosmc and T-synthase coprecipitation
Approximately 2 μ g each of His-tagged soluble Cosmc, HPC4-tagged sol-
uble T-synthase, and soluble Galectin-3 (provided by S. Stowell, Emory
University, Atlanta, GA) were used. His-Cosmc and HPC4-T-synthase were
mixed in buffer (5 mM Tris-HCl and 30 mM NaCl, pH 7.85). Similarly,
His-Cosmc and Galectin-3 were mixed. For controls, His-Cosmc and
HPC4 – T-Synthase were individually prepared in buffer. All preparations
were incubated at RT ( ? 23 ° C) for 25 min. One-fi fth of each mixture was
allocated for input (loading). The remaining part was diluted 25 × with
Ni-NTA washing buffer (50 mM imidazole, 50 mM NaH 2 PO 4 , 300 mM
NaCl, and 0.1% Triton X-100, pH 7.8). 20 μ l Ni-NTA Superfl ow beads
(QIAGEN) were added to the Ni-NTA washing buffer. Proteins were incu-
bated with beads overnight while rotating at 4 ° C. The beads were col-
lected using bench centrifugation and the beads were washed fi ve times
with 400 μ l Ni-NTA washing buffer. The bead-bound material was eluted
fi ve times with 20 μ l of elution buffer (50 mM NaH 2 PO 4 , 300 mM NaCl,
250 mM imidazole,0.1% Triton X-100). Input, unbound, and bound frac-
tions (in equal proportions) were electrophoresed on an SDS/PAGE (4 – 20%)
and transferred to nitrocellulose membrane (Bio-Rad Laboratories). After
blocking with 1% milk, Western blots were performed, first for HPC4 –
T-synthase and Galectin-3, using 4 ml of 4 μ g/ml mAb to HPC4 and 4 ml
of 0.2 μ g/ml of antibody to Galectin-3. The membranes were incubated
for 1 h at RT. For HPC4 blotting, the membrane was washed fi ve times
with 20 mM Tris-HCl, 300 mM NaCl, and 1 mM CaCl 2 , pH 7.2. For
Galectin-3, the membrane was washed fi ve times with 20 mM Tris-HCl,
300 mM NaCl, and 0.05% Tween-20, pH 7.2. After washing, the mem-
branes were incubated with 4 ml of 0.17 ng/ml HRP-conjugated goat
anti – mouse IgG at 23 ° C for 45 min. Membranes were washed fi ve times
with 20 mM Tris-HCl, 300 mM NaCl, and 1 mM CaCl 2 , pH 7.2, and in-
cubated with 3 ml of SuperSignal West Pico Chemiluminescent Substrate
at RT for 1 min. The blot was exposed to fi lm (Denville Scientifi c, Inc.) for
1 min, and developed by autoradiography. The membranes were stripped
by washing with 2% SDS and 1% 2-mercaptoethanol in deionized water
for 30 min, followed by washing three times with 20 mM Tris-HCl and
300 mM NaCl, pH 7.2. The membranes were blocked with 1% milk and
washed three times with 20 mM Tris-HCl and 300 mM NaCl, pH 7.2.
Membranes were incubated with polyclonal chicken IgG against Cosmc
using 4 ml of 0.2 μ g/ml antibody for 1 h. Membranes were washed fi ve
times with 20 mM Tris-HCl and 300 mM NaCl, pH 7.2, and incubated
with 4 ml of 0.17 ng/ml HRP-conjugated goat anti – chicken IgG at 23 ° C
for 45 min. Membranes were washed with 20 mM Tris-HCl and 300 mM
NaCl, pH 7.2, and developed accordingly.
Online supplemental material
Fig. S1 shows that human Cosmc is coordinately transcribed with human
T-synthase in Northern blot. Fig. S2 shows that protein sequences of Cosmc
are highly conserved across species. Fig. S3 shows that Cosmc and T-synthase
localize in different cellular compartments. Fig. S4 shows that Cosmc
itself does not have any T-synthase activity. Fig. S5 shows that truncated
T-synthase protein in LSC cells is not secreted into the media. Online sup-
plemental material is available at http://www.jcb.org/cgi/content/full/
We thank Dr. Steven Itzkowitz for kindly providing human colon carcinoma cell
lines LSC and LSB. We thank Dr. Kelley Moremen for his generosity in provid-
ing rabbit anti-Man II antisera and Sean Stowell for recombinant Galectin-3.
We also thank Dr. Baoyun Xia, Dr. Jamie Heimburg-Molinaro, Ms. Connie Arthur,
and Mr. Anthony Luyai for suggestions and critiques of this manuscript.
This work was supported by a National Institutes of Health RO1 grant
(RO1 GM068559-01A2) to R.D. Cummings. The authors declare that there is
no confl ict of interest.
Submitted: 29 November 2007
Accepted: 16 July 2008
Ahner , A. , and J.L. Brodsky . 2004 . Checkpoints in ER-associated degradation:
excuse me, which way to the proteasome? Trends Cell Biol. 14 : 474 – 478 .
Western blotting with anti-HPC4, Calnexin, GRP78 (BiP), HSP40, or 20S
proteasome ? 1-subunit was performed as previously described ( Ju and
Cummings, 2002 ) or as otherwise stated.
About 5 × 10 6 293T cells transiently transfected with HPC4-Cosmc for
48 h or LSB and LSC cells stably expressing T-synthase – HPC4 were har-
vested and washed with cold PBS. Cells were homogenized in 25 mM
Hepes, pH 7.5, containing 250 mM sucrose. Then PNS was made by cen-
trifugation at 20,000 g for 30 min. Then the concentration of sucrose in
PNS was adjusted to 50% (wt/vol) and loaded on 70, 60, 40, and 20%
sucrose gradient. After centrifugation at 100,000 g for 20 h, 16 fractions
were collected from the bottom of the tube. The fractions were analyzed by
Western blotting with antibodies indicated in the results and also measured
for both T-synthase and ? 4 – Gal-T activity.
T-synthase activity of soluble N-terminal HPC-tagged Cosmc purifi ed
293T cells in a T75 fl ask were transiently transfected with constructs encod-
ing soluble HPC4-tagged Cosmc with or without KDEL-signal and soluble
HPC4-tagged T-synthase. The epitope-tagged proteins were absorbed on
HPC4-Ultralink resin and washed. Some of the sample was used for T-synthase
activity assay directly and the other part was eluted for Western blotting by
ATP-Sepharose chromatography of Cosmc
Hi-5 insect cells were infected with Baculovirus encoding either Cosmc-HA
or soluble N-terminal HPC4 epitope-tagged T-synthase plus wild-type Cosmc.
At 96 h after infection, cells expressing Cosmc-HA were harvested and the
media from cells expressing soluble T-synthase and wild-type Cosmc was
collected. The cell extract and media were chromatographed on a 0.1-ml
ATP-Sepharose column preequilibrated with 50 mM Tris-HCl buffer, pH 7.8,
containing 100 mM NaCl and 5 mM MgCl 2 , respectively. The column was
washed with 0.4 ml of the buffer six times. Bound proteins were eluted with
50 mM ATP in the buffer (0.1 ml/fraction). The loading, fl owthrough, washes,
and eluates were analyzed by Western blot with either anti-HA and or anti-
HPC4 mAbs, as indicated.
Photolabeling of Cosmc by 8-Azido ? -[ 32 P]ATP
HPC4-sCosmc and HPC4 – sT-synthase coexpressed with wtCosmc were
expressed in Hi-5 cells and affi nity purifi ed from the media. 3 μ g Cosmc
and T-synthase were photolabeled in 20 μ l of reaction containing 50 mM
Tris-HCl, pH 7.8, 3 mM MgCl 2 100 mM NaCl, and 1 mM ATP plus 1 μ Ci
of 8-Azido ? -[ 32 P]-ATP in the absence or presence of 10 mM ATP and
10 mM GTP for 30 min on ice. Then the reactions were exposed under UV
(wavelength ? 254 nm) for 90 s on ice followed by immediate addition of
250 μ l TBS containing 50 mM DTT. The reactions were dialyzed with
Centricon Concentrator (3 kD cutoff) with TBS and concentrated down to
50 μ l. 20 μ l of sample were analyzed on SDS-PAGE (4 – 20%) and trans-
ferred to nitrocellulose membrane for radioautogram and Western blot
Proteasome inhibitors treatment
About 10 6 LSC and LSB cells were seeded in T75 fl asks and grown for
24 h. The cells were treated with 10 μ M MG-132 or lactacystin (dis-
solved in 100% DMSO at 2 mM stock) or 0.5% DMSO in complete media
for 12 – 14 h. The cells were harvested for T-synthase activity assay and
Trypsin digestion experiment and coimmunoprecipitation of GRP78
About 2 × 10 6 of LSC and LSB cells stably expressing T-synthase – HPC4
were harvested and cell after nuclear homogenate was obtained. One
portion the homogenate was solubilized with 0.2% Triton X-100 to make
the extract. The cell homogenate and extract (containing 50 μ g of pro-
tein) were treated with 40 μ g/ml trypsin at 37 ° C for 4 h, respectively,
and then analyzed by Western blotting with HPC4, mouse anti – human
20S ? 1-subunit mAb, and mouse anti – human Calnexin mAb. The T-synthase
from cell extracts was also purifi ed and Western blotted with anti-KDEL
LSC, LSB, and Sf-9 cells expressing C-terminal HPC4-tagged T-synthase
with and without Cosmc were harvested and washed twice with PBS. Cell
extract was made from one portion of the cells and incubated with 5 mM
JCB • VOLUME 182 • NUMBER 3 • 2008 542
Schroder , M. 2006 . The unfolded protein response. Mol. Biotechnol. 34 : 279 – 290 .
Soti , C. , A. Racz , and P. Csermely . 2002 . A Nucleotide-dependent molecular
switch controls ATP binding at the C-terminal domain of Hsp90. N-terminal
nucleotide binding unmasks a C-terminal binding pocket. J. Biol. Chem.
277 : 7066 – 7075 .
Springer , G.F. 1984 . T and Tn, general carcinoma autoantigens. Science .
224 : 1198 – 1206 .
Wada , I. , D. Rindress , P.H. Cameron , W.J. Ou , J.J. Doherty II , D. Louvard , A.W.
Bell , D. Dignard , D.Y. Thomas , and J.J. Bergeron . 1991 . SSR alpha and
associated calnexin are major calcium binding proteins of the endoplas-
mic reticulum membrane. J. Biol. Chem. 266 : 19599 – 19610 .
Whitley , P. , I.M. Nilsson , and G. von Heijne . 1996 . A nascent secretory protein
may traverse the ribosome/endoplasmic reticulum translocase complex as
an extended chain. J. Biol. Chem. 271 : 6241 – 6244 .
Xia , L. , T. Ju , A. Westmuckett , G. An , L. Ivanciu , J.M. McDaniel , F. Lupu , R.D.
Cummings , and R.P. McEver . 2004 . Defective angiogenesis and fatal
embryonic hemorrhage in mice lacking core 1 – derived O-glycans. J. Cell
Biol. 164 : 451 – 459 .
Allen , A.C. , P.S. Topham , S.J. Harper , and J. Feehally . 1997 . Leucocyte beta
1,3 galactosyltransferase activity in IgA nephropathy. Nephrol. Dial.
Transplant. 12 : 701 – 706 .
An , G. , B. Wei , B. Xia , J.M. McDaniel , T. Ju , R.D. Cummings , J. Braun , and
L. Xia . 2007 . Increased susceptibility to colitis and colorectal tumors in
mice lacking core 3 – derived O-glycans. J. Exp. Med. 204 : 1417 – 1429 .
Berger , E.G. 1999 . Tn-syndrome. Biochim. Biophys. Acta . 1455 : 255 – 268 .
Bergeron , J.J. , M.B. Brenner , D.Y. Thomas , and D.B. Williams . 1994 . Calnexin:
a membrane-bound chaperone of the endoplasmic reticulum. Trends
Biochem. Sci. 19 : 124 – 128 .
Brockhausen , I. , J. Yang , N. Dickinson , S. Ogata , and S.H. Itzkowitz . 1998 .
Enzymatic basis for sialyl-Tn expression in human colon cancer cells.
Glycoconj. J. 15 : 595 – 603 .
Csermely , P. , and C.R. Kahn . 1991 . The 90-kDa heat shock protein (hsp-90)
possesses an ATP binding site and autophosphorylating activity. J. Biol.
Chem. 266 : 4943 – 4950 .
Ellgaard , L. , and A. Helenius . 2003 . Quality control in the endoplasmic reticulum.
Nat. Rev. Mol. Cell Biol. 4 : 181 – 191 .
Gomord , V. , E. Wee , and L. Faye . 1999 . Protein retention and localization in the
endoplasmic reticulum and the golgi apparatus. Biochimie . 81 : 607 – 618 .
Hendershot , L.M. , and N.J. Bulleid . 2000 . Protein-specifi c chaperones: the role
of hsp47 begins to gel. Curr. Biol. 10 : R912 – R915 .
Itoh , H. , R. Kobayashi , H. Wakui , A. Komatsuda , H. Ohtani , A.B. Miura , M.
Otaka , O. Masamune , H. Andoh , K. Koyama , et al . 1995 . Mammalian
60-kDa stress protein (chaperonin homolog). Identifi cation, biochemical
properties, and localization. J. Biol. Chem. 270 : 13429 – 13435 .
Ju , T. , and R.D. Cummings . 2002 . A unique molecular chaperone Cosmc re-
quired for activity of the mammalian core 1 beta 3-galactosyltransferase.
Proc. Natl. Acad. Sci. USA . 99 : 16613 – 16618 .
Ju , T. , and R.D. Cummings . 2005 . Protein glycosylation: chaperone mutation in
Tn syndrome. Nature . 437 : 1252 .
Ju , T. , K. Brewer , A. D ’ Souza , R.D. Cummings , and W.M. Canfi eld . 2002a .
Cloning and expression of human core 1 beta1,3-galactosyltransferase.
J. Biol. Chem. 277 : 178 – 186 .
Ju , T. , R.D. Cummings , and W.M. Canfi eld . 2002b . Purifi cation, characteriza-
tion, and subunit structure of rat core 1 Beta1,3-galactosyltransferase.
J. Biol. Chem. 277 : 169 – 177 .
Ju , T. , Q. Zheng , and R.D. Cummings . 2006 . Identifi cation of core 1 O-glycan
T-synthase from Caenorhabditis elegans . Glycobiology . 16 : 947 – 958 .
Ju , T. , G.S. Lanneau , T. Gautam , Y. Wang , B. Xia , S.R. Stowell , M.T. Willard ,
W. Wang , Y. Xia , R.E. Zuna , et al . 2008 . Human Tumor Antigens Tn and
Sialyl Tn Arise from Mutations in Cosmc . Cancer Res. 68 : 1636 – 1646 .
Kawar , Z.S. , I. Van Die , and R.D. Cummings . 2002 . Molecular cloning and
enzymatic characterization of a UDP-GalNAc: GlcNAcbeta-R beta1,4-
N-acetylgalactosaminyltransferase from Caenorhabditis elegans . J. Biol.
Chem. 277 : 34924 – 34932 .
Kleizen , B. , and I. Braakman . 2004 . Protein folding and quality control in the
endoplasmic reticulum. Curr. Opin. Cell Biol. 16 : 343 – 349 .
Kudo , T. , T. Iwai , T. Kubota , H. Iwasaki , Y. Takayma , T. Hiruma , N. Inaba , Y.
Zhang , M. Gotoh , A. Togayachi , and H. Narimatsu . 2002 . Molecular
cloning and characterization of a novel UDP-Gal:GalNAc(alpha) peptide
beta 1,3-galactosyltransferase (C1Gal-T2), an enzyme synthesizing a core 1
structure of O-glycan. J. Biol. Chem. 277 : 47724 – 47731 .
Nakagoe , T. , T. Sawai , T. Tsuji , M. Jibiki , A. Nanashima , H. Yamaguchi , N.
Kurosaki , T. Yasutake , and H. Ayabe . 2001 . Circulating sialyl Lewis(x),
sialyl Lewis(a), and sialyl Tn antigens in colorectal cancer patients: multi-
variate analysis of predictive factors for serum antigen levels. J. Gastroenterol.
36 : 166 – 172 .
Ni , M. , and A.S. Lee . 2007 . ER chaperones in mammalian development and
human diseases. FEBS Lett. 581 : 3641 – 3651 .
Nilsson , I. , and G. von Heijne . 2000 . Glycosylation effi ciency of Asn-Xaa-
Thr sequons depends both on the distance from the C terminus and on
the presence of a downstream transmembrane segment. J. Biol. Chem.
275 : 17338 – 17343 .
Otsu , M. , and R. Sitia . 2007 . Diseases originating from altered protein quality
control in the endoplasmic reticulum. Curr. Med. Chem. 14 : 1639 – 1652 .
Piller , V. , F. Piller , and M. Fukuda . 1990 . Biosynthesis of truncated O-glycans in
the T cell line Jurkat. Localization of O-glycan initiation. J. Biol. Chem.
265 : 9264 – 9271 .
Ruddock , L.W. , and M. Molinari . 2006 . N-glycan processing in ER quality
control. J. Cell Sci. 119 : 4373 – 4380 .
Saira Mian , I. 1993 . Sequence similarities between cell regulation factors, heat
shock proteins and RNA helicases. Trends Biochem. Sci. 18 : 124 – 127 .
Schietinger , A. , M. Philip , B.A. Yoshida , P. Azadi , H. Liu , S.C. Meredith , and H.
Schreiber . 2006 . A mutant chaperone converts a wild-type protein into a
tumor-specifi c antigen. Science . 314 : 304 – 308 .