| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
*
Department of Microbiology and Immunology, Vanderbilt University School of Medicine, Nashville, TN 37232;
Department of Microbiology and Immunology, Indiana University School of Medicine, Walther Oncology Center, Indianapolis, IN 46202; and
Institute of Pathology, Case Western Reserve University, Cleveland, OH 44106
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
14J15 natural T cell Ags. Therefore, we predict that cellular
lipids occlude the hydrophobic Ag-binding groove of CD1 during assembly
until they are exchanged for a glycolipid Ag(s) within the recycling
compartment for display on the plasma membrane. In this manner,
cellular lipids might play a chaperone-like role in the assembly of
CD1d1 in vivo, akin to the function of invariant chain in MHC class II
assembly. |
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
14J15 Ag receptor
-chain and those that express diverse TCRs. The CD1d1-restricted
V14J15 NKT cells are conserved in both humans and mice, and hence
are predicted to impart an evolutionarily conserved immune function.
This immune function, albeit elusive, is thought to be immunoregulatory
in nature. Upon activation in vivo through their Ag receptors, NKT
cells promptly elicit large amounts of immunoregulatory cytokines such
as IL-4, IFN-
, TNF-, and GM-CSF. Furthermore, V14J15 NKT
cell dysfunction in autoimmune-prone mice, as well as in individuals
afflicted with autoimmune diabetes, underscore their importance in the
physiology of normal immune responses (reviewed in Ref.
1). Thus, delineating the structure and function of CD1d
is essential to understanding the biology of NKT cells.
Humans express group I CD1a, CD1b, and CD1c and group II CD1d molecules
(2, 3, 4). Mice and rats express only CD1d (5, 6). Topologically, CD1 resembles the classical MHC-encoded
Ag-presenting molecules. Its domain organization and its association
with 
2-microglobulin
(2m)5
for complete assembly are similar to MHC class I molecules. A
distinguishing feature of CD1d1 is its exclusively nonpolar hydrophobic
Ag-binding groove. The groove consists of a large pocket A' that is
almost completely covered from all sides except for a narrow lateral
opening connecting it to a short apically exposed pocket F'
(7). Thus, CD1 molecules have evolved to present
hydrophobic Ags to the mammalian immune system. Indeed, structure
function studies pioneered by Brenner and colleagues (reviewed
in Refs. 8, 9, 10, 11) have revealed that human CD1b and CD1c present
lipid and glycolipid Ags to specific T cells reactive to mycobacterial
Ags. Ags presented by CD1b include mycolic acid, glucosylmonomycolate,
phosphatidylinositol (PI)-mannans, and lipoarabinomannans
(9, 10, 11). CD1b also presents self glycolipids such as GM1
to autoreactive T cells (12). CD1c, in contrast, presents
mycobacterial dolichylphosphorylmannose (DPM) to specific T cells
(13). Current evidence suggests that mouse and human CD1d
present -galactosylceramide (GalCer) and potently activate mouse
V14J15 and human V24JQ NKT cells, respectively
(14, 15, 16, 17). Additionally, CD1d1 also presents PI and
activates a small proportion of V14J15 NKT cells
(18). Thus, self and nonself lipids presented by CD1 are T
cell Ags.
The ability to study lipid CD1 interactions in vitro has illuminated
several physico-chemical aspects of lipid presentation and recognition
(reviewed in Ref. 1). Notwithstanding, all of the above
studies have relied on purification of mycobacterial or cellular lipids
and their ability to reconstitute functional CD1 molecules in vitro,
either on live cells or using a cell-free system, recognizable by
specific T lymphocytes. To elucidate the basis for CD1d1 function, our
approach has been to isolate and characterize the associated ligand by
biochemical methods. The data revealed that CD1d1 expressed in
mammalian cells assembled with a phospholipid, which was identified as
GPI (19). Consistent with this data, a recent study
demonstrated that a V14J15 NKT cell hybridoma recognized PI
presented by CD1d1 as its Ag (18). Notwithstanding, PI
does not represent a major NKT cell Ag (20) and, hence,
the role of phospholipid(s) in CD1d1 function remains elusive. Thus, to
clarify the role of CD1d1-associated PI and PI-glycans in CD1 function,
we have studied the assembly of this molecule in vivo. The data are
consistent with the idea that the natural CD1d1-associated ligands play
a chaperone-like role during its assembly in vivo akin to the function
of invariant chain in MHC class II assembly.
|
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The full-length
2ma cDNA from
pEE6-2m (21) was digested with
HindIII and BamHI, and the resulting
fragment was subcloned
intoHindIII-BamHI-digested pEE12
(CellTech, Slough, England). The resulting
pEE12-2m was checked for integrity by
restriction mapping. Full-length CD1d1 cDNA (pBluescript-mCD1d1; kindly
provided by Dr. S. Balk, Harvard Medical School, Boston, MA), was
subcloned into the pVL1393 vector (kindly provided by Dr. M. D.
Summers, Texas A&M University, College Station, TX) using the
XhoI-NotI restriction sites. The
EcoRI-EcoRV fragment from pVL1393-CD1d1
containing the first four exons of CD1d1 was subcloned into pCR3
(Invitrogen, Carlsbad, CA). Finally, the
EcoRV-NotI fragment containing exons 5 and 6 of
CD1d1 was subcloned into pCR3 containing exons 1–4 of CD1d1 resulting
in pCR3-mCD1d1; thus, the full-length cDNA encoding wild-type CD1d1 was
generated. The cDNA encoding soluble CD1d1 (pBluescript-sCD1d1; also
provided by Dr. S. Balk) was digested with XhoI and blunted
with Klenow polymerase. The XhoI blunt-NotI
fragment containing the sCD1d1 cDNA was subcloned into pCR3, resulting
in pCR3-sCD1d1. For cloning, the pCR3 vector was prepared by digesting
with HindIII and then was blunted with Klenow polymerase and
digested with NotI. The cDNA for an endoplasmic reticulum
(ER)-retained CD1d1 molecule was constructed by appending the ER
retention signal KDEL to the carboxyl terminus of sCD1d1. This
construct also contained a c-Myc tag between sCD1d1 and the ER
retention signal to facilitate detection of the expressed CD1d1. The
sCD1d1-er cDNA was constructed using a 70-bp-long DNA encoding the
c-Myc tag and KDEL. It entailed the annealing of the following four
oligonucleotides: a, 5'-AAGACTGAAATGGAGCAAAAGCTCATTTCTGAA-3'; b,
5'-GAGGACCTGAATTCGGAGAAGGATGAGCTCTGAAGATCTGC-3'; c,
5'-GCGGCCGCAGATCTTCAGAGCTCATCCTTCTCCGAATTCAG-3';
and d,
5'-TCCTCTTCAGAAATGAGCTTTTGCTCCATTTCAGTCTT-3'.
This resulted in a DNA fragment with a 5' blunt end and a 3'
NotI overhang. The oligonucleotides were phosphorylated and
ligated together to form the 
70-bp fragment. The resulting fragment
was cloned into the EcoRV-NotI site of
pCR3-mCD1d1, resulting in pCR3-sCD1d1-er. A CD1d1 mutant lacking its
internalization signal was constructed by substituting the
transmembrane and cytosolic region of CD1d1 with that of
H2Kb, resulting in CD1d1-Kbtail. The CD1d1-Kbtail
cDNA was constructed by subcloning an EcoRV-NotI
fragment containing exons 5–8 of H2Kb into the
EcoRV-NotI site of pCR3-sCD1d1. In sCD1d1 cDNA,
the EcoRV site lies downstream of exon 4, which encodes the
3 domain of the mature protein. cDNA encoding the soluble
H2Db was constructed by PCR amplification of
exons 1–4 using 5'-CACAAGCTTGGGAATTCCGGGGGCGATGGCTCCGCG-3' forward
and 5'-CGGGATCCCGTCACCATCTCAGGGTGAGGGG-3' reverse primers. The
resulting product was cleaved with HindIII and
BamHI and cloned into the HindIII and
BamHI site of pCR3. An authentic pCR3-Db-sol determined by
Sanger dideoxynucleotide sequence analysis was used for gene
transfer.
Cell lines
See Table I
for a description of
CD1- and H2 class I-expressing cell lines. O that expresses mouse
2ma was generated by
gene transfer into NS0 plasmacytoma as described
(21). Transfected cells were selected 24 h later in
L-glutamine-free medium. O cell lines expressing soluble
H2Db as well as wild-type and mutant CD1d1 were
generated by electroporation of the respective cDNA constructs.
Transfected cells were selected in the absence of
L-glutamine but in the presence of 0.8 mg/ml of geneticin
(G418; Life Technologies, Rockville, MD). O and derived lines were
maintained in L-glutamine-deficient DMEM (Media-Tech,
Herndon, VA) containing 10% dialysed FBS (HyClone Laboratories, Deer
Park, PA, or Life Technologies) and 0.5 mg/ml G418. Kb-high and Db-high
cells were maintained as described (21). Pig A-
(S49a) and Pig E- (BW5147e) mutants (22) were
from American Type Culture Collection (Manassas, VA) and maintained in
RPMI 1640 (Media-Tech) containing 5–10% FBS.
|
All cell lines used in this study express wild-type and mutant CD1d1 as well as H2Kb and H2Db molecules in the appropriate cell lines, which was confirmed by flow cytometry and/or immune precipitation methods using specific Abs (data not shown).
NKT cell hybridomas DN32.D3 and 431.A11 (gifts from A. Bendelac of Princeton University, Princeton, NJ) as well as N38-2C12 and N37-1A12 (generously provided by K. Hayakawa of Fox Chase Cancer Institute, Philadelphia, PA) were maintained in RPMI 1640 containing 10% FBS.
Flow cytometric analysis
To determine CD1d1 expression, 5 x
105 cells were reacted with 0.5 µg of
biotinylated 1B1, a CD1d1-reactive mAb, and biotinylated AF6, a
H2Kb-reactive mAb, and were detected with
streptavidin-CyChrome using a FACScan flow cytometer (BD Biosciences,
San Jose, CA). All reagents were from BD PharMingen (San Diego,
CA).
Labeling CD1d1-associated ligand with [3H]mannose, [32P]orthophosphate, [3H]inositol, [3H]ethanolamine, and [3H]mevalonic acid
At least 2 x 107 CD1d-expressing
cells or control cells were tritium labeled with 250 µCi of
[3H]2-mannose or
[3H]ethanolamine (American Radiolabeled
Chemicals, St. Louis, MO) as described (19). Cells were
labeled with 250 µCi of [3H]inositol (NEN,
Boston, MA) in inositol-free DMEM (Life Technologies) or with 250 µCi
of [3H]mevalonic acid (American Radiolabeled
Chemicals) in DMEM. [32P]Orthophosphate
labeling was accomplished using phosphate-free DMEM (Life Technologies)
as described (25). Culture supernatants from
tritium-labeled cells were collected for the purification of soluble
molecules. The harvested cells were solubilized in 2.5–3 ml of PBS
containing 1% (w/v)
3-[(3-cholamidopropyl)dimethylammonio]-2-hydroxy-1-propanesulfonate
(Sigma-Aldrich, St. Louis, MO). Cell lysates were clarified by
centrifugation. Membrane-bound CD1d1 and class I molecules were
isolated from the postnuclear fraction, whereas total cellular lipids
were extracted from the nuclear and membrane pellet.
Mouse CD1d1-specific hAs
Polyclonal heteroantiserum (hAs) against CD1d1 was generated in
a rabbit immunized i.m. with 0.1 mg of Ni-affinity purified sCD1d1
(this Ag was >95% pure) emulsified in Ribi adjuvant. The sCD1d1
immune rabbit was boosted three times with 50 µg of sCD1d1 at 3- to
4-wk intervals. The immune rabbit was terminally bled 2 wk after the
last boost. The specificity of the resulting hAs was determined using
CD1d1-positive cell lines by flow cytometry as well as by immune
precipitation of
[35S]cysteine/[35S]methionine-labeled
proteins. It only precipitates CD1d1 from cell lines and no other
molecule. In fact, it does not even cross-react with the human homolog
CD1d or paralog CD1b (data not shown). Therefore, it is less likely to
cross-react directly with a lipid, lipoprotein, or another lipid
binding protein.
Biosynthetic labeling of proteins
Steady-state as well as pulse labeling of cells with [35S]cysteine/[35S]methionine and the chase of labeled proteins were performed as described (21). After preclearing with normal mouse serum, samples were immune precipitated with the anti-CD1d1 hAs or an appropriate control Ab as described. An H2Kb-specific (Y3) (26) or H2Db-specific (B22-249) (26) mAb was used when experiments were performed with mouse cell lines. W6/32 (generously provided by J. Yewdell of National Institute of Allergy and Infectious Diseases, National Institutes of Health, Bethesda, MD), an anti-HLA class I-specific mAb (27), was used when experiments were performed in human cell lines. Immune precipitates were separated by 15% SDS-PAGE and visualized by autoradiography. In pulse-chase experiments, the immune precipitates were subjected to 1–2 mU or varying concentrations of endoglycosidase H (endo H) for 14–16 h and were processed as above.
Affinity chromatography
CD1d1 molecules and the control class I molecules from each
sample were sequentially purified by immune affinity chromatography as
previously described (28). CD1d1-specific and class
I-specific affinity columns were prepared by prebinding 0.1–0.15 mg
of purified Ab to 1 ml of 50% protein A-Sepharose slurry (Repligen,
Needham, MA). Unbound Ab and nonspecifically bound proteins were washed
away with PBS and resuspended in an equal volume of the same buffer.
Each Ab-bound protein A-Sepharose was packed into a 5-ml disposable
column (Pierce, Rockford, IL) and used to purify CD1d1 and class I
molecules. Immune affinity chromatography was performed as described
earlier (28). Secreted CD1d1 from the tritium labeling
supernatants was purified using Hi-Trap metal chelating columns
(Amersham Pharmacia Biotech, Piscataway, NJ) according to the
manufacturer’s instructions. Ten 0.5-ml fractions were collected;
one-tenth of each fraction was dissolved in Econo-safe scintillation
fluid (RPI, Mount Prospect, IL) to measure radioactivity using a
scintillation counter (LS6500; Beckman Coulter, Fullerton,
CA).
Lipid extraction from affinity-purified proteins
Radioactive fractions from each affinity-purified sample were
pooled. An equal number of fractions not containing radioactivity were
pooled separately. Lipids were extracted by Bligh-Dyer method
(29) with two volumes of chloroform:methanol (2:1 C:M).
After thorough mixing, the top aqueous and the bottom organic phases
were allowed to separate. The organic phase was carefully transferred
into another tube. The aqueous phase was extracted two or three more
times. The resulting organic phase was pooled with the first and dried
under vacuum or a gentle stream of N2 gas. The
dried extract was redissolved in 0.5 ml of C:M. Radioactivity was
monitored by scintillation counting using 10–20 µl of the
extracted lipids.
Enzymatic digestions of lipids
C:M in two equal aliquots of the lipid extracts were evaporated
under vacuum or N2 gas. One aliquot was dissolved
in 50 µl of PBS (pH 7–7.4) and digested with 55 mU of
PI-specific phospholipase C (PI-PLC; Glyko, Novato, CA) at 37°C for
1 h. Likewise, the second aliquot was dissolved in 50 µl of
buffer containing 4.9 mM CaCl2 and 147 mM NaCl
(pH 8.9) and digested with 1 U of phospholipase
A2 (PLA2) (Sigma-Aldrich)
at 25°C for 1 h. The enzymatic reactions were stopped by C:M
extraction. The organic phase was evaporated, resuspended in 20–40
µl of C:M, and spotted onto a 20-cm x 20-cm K6 silica gel TLC
plate (Whatman, Clifton, NJ). Approximately 25 µg of PI was spotted
and served as a standard. Additionally, 0.025 µCi of
[3H]PI or [3H]DPM
(American Radiolabeled Chemicals) was used as radioactive standard.
TLC
TLC was performed in a chamber saturated with 100–150 ml of neutral mobile phase containing chloroform:methanol:water in 10:10:3 ratio (30). The plate was dried for at least 1 h, sprayed with EN[3H]ANCE (NEN Life Sciences) according to the manufacturer’s directions, and exposed to autoradiographic film or a phosphoimager TR plate specifically sensitive to tritium (Fuji Medical Systems, Stamford, CT). Phosphorimaging was facilitated by an FLA 2000 Fluorescent Image Analyzer (Fuji Medical Systems).
NKT cell stimulation and ELISA
Equal numbers (5 x 104 cells per
well) of stimulator cells and responder NKT cell hybridomas were
cocultured for 18–20 h at 37°C. Stimulation of hybridomas was
measured by monitoring IL-2 secretion by ELISA. ELISA was performed
using JES6-1A12 and JES5-5H4 (BD PharMingen) as IL-2 capture and
detection mAbs, respectively, according to the manufacturer’s
instructions.
|
Results |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Initial characterization of the CD1-associated ligand(s) relied on determining the mass of the compounds eluted from CD1d1. Interpretation of the mass spectral data suggested the natural ligand of CD1d1 to be GPI (19). To study CD1-lipid interactions in vivo as well as to characterize the natural CD1-associated ligand(s), a biochemical method was established. For this purpose, cells that express the soluble form of CD1d1 and H2Db class I molecules, sCD1d1 and Db-sol, respectively, were generated. Soluble molecules were used to establish the biochemical method because it provides an abundant source of CD1d1 and, hence, the associated ligand(s). Additionally, in a previous study we had demonstrated that [3H]2-mannose-labeled ligands specifically associated with sCD1d1 (19).
Thus, to characterize the natural CD1d1-associated ligand(s), both
sCD1d1 and Db-sol were radiolabeled with
[3H]2-mannose.
[3H]2-mannose was used in this experiment
because 1) its label is seldom lost to another sugar but fucose
(31) and 2) it labels GPI as well as DPM-glycans. sCD1d1
and control Db-sol were affinity purified from the supernatant of
[3H]2-mannose-labeled cells. The associated
ligand(s) were extracted by the Bligh-Dyer method, separated by TLC in
a neutral mobile phase along with nonradioactive mammalian PI as a
standard, and visualized by fluorography. The data revealed that
[3H]2-mannose was incorporated into a dominant
lipid species along with two minor ones that were associated with
sCD1d1 (Fig. 1, lane 1) but
not with Db-sol (Fig. 1, lane 2). Interestingly, the three
CD1d1-associated lipid species have a migratory pattern on the TLC that
was distinct from the nonradioactive PI standard (Fig. 1, arrow) with a
relative migration (Rf) suggestive of a
polar compound.
|
B) were it nonspecifically "mopping-up" the
numerous [3H]-labeled lipids spilled into the
tissue culture medium. This it does not,
and hence the [3H]-labeled lipids shown in Fig. 2A are specifically bound to CD1d1.
|
could be due to a glycan modifying a
lipid and/or a protein. For three reasons we believe that the
[3H]2-mannose-labeled compound is a glycolipid
and not a glycopeptide or glycoprotein. First, the control Db-sol, also
a glycoprotein with two N-linked glycans, processed in the
same manner as sCD1d1, does not contain the tritiated spots seen with
the CD1 sample (Fig. 1). Second, the CD1d1-associated
[3H]2-mannose-labeled ligands are PI-PLC sensitive as
described below. Third, sialic acid-containing glycans have poor
mobility on TLC and hence remain at or a bit above the origin in the
mobile phase used. For example, ceramide migrates to the front of the
TLC, whereas sialic acid containing glycoceramides remains close to the
origin (S. Joyce, unpublished data). Because each glycan moiety added
to proteins contains three sialic acid units, it is less likely to
migrate to the center of the TLC as did the
[3H]2-mannose-labeled ligand extracted from sCD1d1.
Therefore, we conclude that the
[3H]2-mannose-labeled compound associated with
CD1d1 is not its own glycan.
To determine the chemical nature of the
[3H]2-mannose-labeled lipid(s) associated with
CD1d1, in the second experiment the Bligh-Dyer extracted ligand(s) was
subjected to PI-PLC or PLA2 digestion. PI-PLC
specifically cleaves inositol-containing phospholipids; its enzymatic
activity is sensitive to acyl-modification of the inositol head group
(32). In contrast, PLA2 specifically
cleaves fatty acyl modification at C-atom 2 of glycerolipids; it is
sensitive to the stereochemistry of the asymmetric C-atom 2
(33). The enzymatic reaction was stopped by lipid
extraction; the products were separated by TLC in a neutral gradient
along with PI as well as [3H]PI and
[3H]DPM standards and were detected by
autoradiography. The data revealed that cleavage with PI-PLC resulted
in the loss of the [3H]2-mannose label from the
sCD1d1-associated ligand (Fig. 2A, lane 1). This
result is consistent with release of the water-soluble
[3H]2-mannose-labeled phosphoinositol-glycan
from mannosylated PI because the glycan does not partition into the
organic phase. In contrast, cleavage with PLA2
did not result in the loss of the [3H]2-mannose
label (Fig. 2A, lane 2). Both the
PLA2-cleaved and uncleaved
[3H]2-mannose-labeled PI partition into the
organic phase during Bligh-Dyer extraction. Therefore, it is unclear
whether the sCD1d1-associated mannosylated PI is sensitive to
PLA2 or not. Thus, together with the previously
published mass spectral data, we conclude that GPI is a major
[3H]2-mannose-labeled lipid associated with
CD1d1.
To determine the diversity of cellular lipids that incorporate
[3H]2-mannose, one-tenth of the postnuclear
detergent lysate of [3H]2-mannose-labeled
sCD1d1 and Db-sol cell lines was subjected to Bligh-Dyer extraction.
The organic phase of this extract was separated by TLC in a neutral
mobile phase and visualized by autoradiography. The data revealed that
both sCD1d1 and Db-sol cell lines incorporated
[3H]2-mannose into several glycolipids (Fig. 2B) whose identities, because they are not essential to this
study, were not determined. Thus CD1d1 selects a single dominant ligand
from a cellular pool of several
[3H]2-mannose-labeled glycolipids.
Stoichiometry of CD1d1-GPI association.
Previously, we reported that >90% of CD1d1 molecules are occupied by
GPI. Being a glycoprotein, [3H]2-mannose labels
the glycan moiety of CD1d1 -chain as well. Therefore, the ratio of
the [3H]2-mannose label associated with CD1d1
to that of the extracted ligands provides a measure of what percentage
of CD1d1 molecules is associated with GPI. To determine the number of
glycan groups on CD1d1, it was immune precipitated from cells pulse
labeled with
[35S]cysteine/[35S]methionine
and either not chased or chased for 2 h. The immune precipitates
were then subjected to digestion with varying concentrations of endo H.
H2Kb, which has one N-glycan
modification (34), was used as the control. The data
revealed five endo H-sensitive radioactive bands in the case of CD1d1
(Fig. 3A) and one, as
expected, in the case of H2Kb (Fig. 3B). Thus, all predicted glycosylation sites of CD1d1 are
modified by N-linked glycans.
|
15 mannose residues. GPI, in contrast, contains a
minimum of three mannose residues. Therefore, a 1:1 stoichiometry of
CD1d1 to GPI should yield five times as much radioactivity in the
aqueous phase compared with the organic phase. In over three separate
experiments, approximately seven to 10 times the radioactivity was
found in the aqueous phase compared with that in the organic phase
(data not shown). Thus, greater than half the total CD1d1 molecules
were associated with [3H]2-mannose-labeled GPI.
Considering 70–80% efficiency of Bligh-Dyer extraction, consistent
with the previous report (19), >90% of CD1d1 molecules
are associated with GPI. Calculations of stoichiometry require steady-state labeling of constituents being compared. Note that DPM and dolichylphosphorylglycans that contain mannose are precursors that feed into the biosynthesis of glycans added onto PI as well as glycoproteins. Their rates of synthesis are about the same within cells and, hence, the precursor pools should be the same after pulse with [3H]2-mannose, which in our experiments was for 24 h. Therefore, the results presented herein from our calculations will not differ from the actual stoichiometry of CD1d1 and the associated compound.
Select few cellular lipids associate with CD1d1
As noted above, the majority of sCD1d1 assemble with GPI. However, because [3H]2-mannose does not label all cellular lipids, it does not reveal the repertoire of ligands that associate with CD1d1. Additionally, our previous study revealed an ion pertaining to free PI and several unidentified ions. These observations raised the question regarding the diversity of the repertoire of CD1d1-associated cellular lipids. To address this question, cell lines expressing wild-type CD1d1 and control H2Kb were generated. These were radiolabeled with [32P]orthophosphate (a precursor of phospholipids), [3H]inositol (a precursor of PI and GPI), [3H]ethanolamine (a precursor of phosphatidylethanolamine and, to a lesser extent, of phosphatidylserine and phosphatidylcholine), or [3H]mevalonic acid (a precursor of dolichol and cholesterol) (35, 36). The associated ligand(s) was isolated and characterized as described above. Because the same lipid spots of very similar intensity were extracted from CD1d1 and the control H2Kb (data not shown), it was difficult to ascertain the phospholipid repertoire specifically associated with CD1d1 by [32P]orthophosphate labeling method.
CD1d1 associates with PI.
Therefore, the above experiments were repeated using labels that are
precursors of specific classes of phospholipids. The data revealed that
three [3H]inositol-labeled lipids were
associated with CD1d1 (Fig. 4A, lane 2). Two of
the [3H]inositol-labeled lipids were
specifically extracted from CD1d1, but the third faster migrating lipid
was also found in extracts of H2Kb (Fig. 4A, lanes 1 and 2). The two
[3H]inositol-labeled lipids were selected by
CD1d1 among several 8–10 biosynthetically labeled lipids synthesized
by the cell (Fig. 4C, lane 1). Of the two
[3H]inositol-labeled lipids associated with
CD1d1, one has the same Rf as the
[3H]PI standard, suggesting that this lipid may
be PI. The [3H]PI standard consists of stearic
and arachidonic acid at C-atoms 1 and 2 (according to the supplier),
the two fatty acids predicted from the mass spectral data of
CD1d1-associated ligand(s) reported previously (19).
|
A,
lanes 2 and 4). However, the slow migrating lipid
spot was partially resistant to PI-PLC (Fig. 4A, lane
4). PI-PLC activity on CD1d1-associated
[3H]inositol-labeled ligand is specific because
it cleaves [3H]PI standard (Fig. 4B,
lane 2) but not [3H]DPM (data not
shown). Also note that PI-PLC digestion of
[3H]PI results in a slow migrating spot (Fig. 4B, arrow) similar to that observed upon cleavage of
CD1d1-associated ligand with PI-PLC (Fig. 4A, arrow). This
new species could be [3H]inositolphosphate.
Together, the data suggest that the faster migrating CD1d1-associated
lipid that comigrates with the [3H]PI standard
is mammalian PI, whereas the identity of the second lipid species is
unknown. Thus, PI is another CD1d1-associated natural ligand. The
extent to which it occupies CD1d1 was not measurable in the current
experiment.
Of the two CD1d1-associated lipid spots, only the slowest migrating
species was partially susceptible to PLA2 (Fig. 4A, lanes 2 and 3). Resistance to
PLA2 was not due to inactivity of the enzyme,
because a larger quantity of [3H]PI standard
compared with the CD1d1-associated ligand was sensitive to
PLA2 (Fig. 4B, lane 3).
Because PLA2 is sensitive to the stereochemistry
of C-atom 2 of glycerophosphatides, the configuration of this atom in
CD1d1-associated PI may be distinct from the standard mammalian
PI.
It was puzzling that the [3H]inositol label was
not incorporated into GPI. If GPI was labeled, it, being more polar,
would migrate more slowly than PI upon TLC. Such a slow migrating lipid
with an Rf similar to the mannosylated-PI-glycan
observed in Fig. 1 was not present among CD1d1-associated
[3H]inositol-labeled lipids (Fig. 4A). The reason for this is currently unknown.
Dolichylphosphorylglycans and phosphatidylethanolamine do not
assemble with CD1d1.
Because diolichylphosphorylglycans and phoshatidylethanolamine
are indigenous to the ER, whether they also assemble with CD1d1 was
determined. The data revealed that neither CD1d1 nor
H2Kb contained any
[3H]mevalonic acid-labeled lipid (Fig. 5A), despite the fact that the
cell lines that express them contained numerous biosynthetically
labeled lipids (Fig. 5B). The fast migrating predominant
spot extracted from CD1d1 has an Rf predicted
for cholesterol; it was not found consistently. Additionally, this
lipid was also found in extracts of the control class I
molecules in the repeat of this experiment (data not shown).
Additional data revealed that, similar to the
[32P]orthophosphate labeling experiments, the
[3H]ethanolamine label was nonspecifically
associated with lipids extracted from CD1d1 as well as
H2Kb. Thus, dolichylphosphorylglycans and
phosphatidylethanolamine may be minor, if at all, ligands of CD1d1.
|
It is clear that neither GPI nor PI is the natural Ag of NKT cells
(20). However, it is possible that GPI and PI have a
chaperone-like function in the assembly of CD1d1. If they do indeed
function as chaperones, then GPI and/or PI would assemble with CD1d1 in
the ER. Thus, two cell lines were generated. One line retains the
expressed CD1d1 in the ER by virtue of containing the KDEL signal
(37, 38) at its carboxyl terminus (sCD1d1-er). The second
line expresses a mutant CD1d1 that had its transmembrane and cytosolic
tail replaced with that from the nonrecycling
H2Kb (sCD1d1-Kbtail) and, hence, has lost its
endosome/lysosome-targeting signal. The expression and intracellular
trafficking patterns of the wild-type CD1d1 as well as sCD1d1-er and
sCD1d1-Kbtail mutants were determined. For this purpose, cell lines
expressing the wild-type and the mutant CD1d1 molecules were pulse
labeled with
[35S]cysteine/[35S]methionine
and chased for the indicated time periods. Cells were solubilized with
detergent, and the postnuclear fraction was subjected to immune
precipitation with CD1d1-specific hAs. The sCD1d1-er mutant was also
immune precipitated from the labeling supernatant. Immune precipitates
were subjected to endo H digestion. Both CD1d1 and sCD1d1-Kbtail
traffic through the secretory pathway with similar kinetics; their
t1/2 in the ER is between 1 and 2
h (Fig. 6A, upper two
panels). Thus, replacement of the transmembrane and the cytosolic
regions of CD1d1 with those of H2Kb did not
significantly alter the intracellular traffic of the mutant CD1d1
molecule.
|
4–8 h (Fig. 6A). The delayed egress of the sCD1d1-er mutant from the ER
is consistent with the function of the KDEL ER-retention signal
(37, 38). A reason that the ER retention signal was
appended to the secreted form of CD1d1 is that, if it did egress from
the ER, it would not be retrieved but secreted into the growth medium
so that it will not confound the experiment described in Fig. 6C. The sCD1d1-er mutant is indeed secreted into the
labeling supernatant upon achieving endo H resistance (Fig. 6A).
Whether sCD1d1-Kbtail had lost its ability to traffic through the
endosome/lysosome compartment was determined using CD1d1-restricted NKT
cell hybridomas as probes. One V14J15 NKT cell hybridoma,
DN32.D3, recognizes an endosomal/lysosomal Ag presented by CD1d1
(39) and, hence, should not be activated by sCD1d1-Kbtail.
A second CD1d1-restricted NKT cell hybridoma 431.A11, which expresses
the V3 receptor, recognizes an Ag present in the anterograde
secretory pathway (39). It should be activated by
sCD1d1-Kbtail. Thus, both DN32.D3 and 431.A11 recognize mCD1d1 (Fig. 6B). Moreover, as expected, DN32.D3 did not recognize,
whereas 431.A11 did recognize, sCD1d1-Kbtail (Fig. 6B).
Thus, sCD1d1-Kbtail inefficiently, if at all, traffics through the
endosomal/lysosomal compartments.
Thus, the wild-type and the two mutant CD1d1 molecules provide a good
system to determine the site of CD1d1 assembly with cellular lipids.
Cell lines expressing the three forms of CD1d1 as well as control
H2Kb were metabolically labeled with
[3H]inositol and solubilized in detergent, and
CD1d1 as well as H2Kb were sequentially affinity
purified from their postnuclear fractions. The eluted fractions were
collected and the amount of radioactivity in each of 10
fractions was measured. The data revealed that
[3H]inositol was incorporated into
molecules associated with CD1d1 but not significantly into
H2Kb (Fig. 6C). The parent cell line
used for ectopic expression of CD1d1 and H2Kb
expresses a small amount of endogenous CD1d1 (data not shown). Because
[3H]inositol is incorporated into PI, assembly
of CD1d1 with PI occurs in the ER. Moreover, GPI is associated with
sCD1d1, a form that does not enter the endosome/lysosome compartment.
Thus, we conclude that the assembly of PI and GPI with CD1d1 occurs in
the ER.
PI is sufficient for complete assembly and intracellular traffic of CD1d1
Because GPI is the major natural ligand of CD1d1 and PI assembles
with CD1d1 in the ER, it is possible that these cellular phospholipids
play a chaperone-like role in the assembly of CD1d1. Self-peptides
derived from the cytosol play a critical role in the assembly of MHC
class I molecules in the ER. Proper assembly of class I molecules is
inhibited by the lack of peptides in the ER (40). Thus, we
reasoned that the absence of GPI and PI in cells would adversely affect
CD1d1 assembly in GPI-deficient cell lines. Therefore, the assembly and
traffic of wild-type CD1d1 was studied in human GPI-positive K562 and
the derived GPI-deficient cell line IA. IA lacks functional Pig
A and, hence, they are deficient in the first glycosylated product
of PI (23). Upon CD1d1 cDNA transfer and selection, both
wild-type (K562-mCD1d1) and the Pig A mutant (IA-mCD1d1)
cells express similar levels of CD1d1 (Fig. 7A). This expression pattern
is consistent with the ability of V14J15 NKT cells and derived
hybridomas to recognize CD1d1 expressed by Pig A-deficient
mouse cell lines (Fig. 7B and Ref. 20).
|
C). The t1/2 of ER
residence is 2 h (Fig. 7C). Thus, CD1d1 assembles
normally in GPI-deficient cell lines. CD1d1 assembles with PI in GPI-deficient cells
To determine whether CD1d1 expressed in GPI-negative cells
assembles with a cellular lipid, K562-mCD1d1 and IA-mCDd1 cell lines
were labeled with [3H]inositol. CD1d1 from both
cell lines were affinity purified from postnuclear fractions of
detergent lysates, and the amount of radioactivity in each eluted
fraction was monitored by scintillation counting. The data revealed
that the [3H]inositol label was associated with
CD1d1 and not with H2Kb (Fig. 7D, a and d). Interestingly, about the
same amount of[3H]inositol-labeled compound
was associated with CD1d1 expressed by GPI-positive K562-mCD1d1 and
GPI-deficient IA-mCD1d1 cells (Fig. 7D, b
and c). Thus, CD1d1 assembles with an inositol-containing
lipid even in the absence of GPI.
To conclusively prove that the CD1d1-associated
[3H]inositol is indeed incorporated into PI,
the associated ligand(s) was extracted, separated by TLC, and detected
by phosphorimaging. The data revealed that one of the three detectable
lipids extracted from CD1d1 is indeed PI, because the
Rf of this lipid is similar to the mammalian
[3H]PI standard (Fig. 7D) and
because this spot extracted from mCD1d1 is sensitive to PI-PLC (Fig. 4A). Thus, CD1d1 assembles with PI in GPI-deficient
cells.
The association of CD1d1 with PI in the IA-mCD1d1 cell line could be
due to its assembly with human 2m. Therefore,
mouse cell lines S49a and Bw5147e deficient in Pig A and
Pig E (complementation group E mutant that is unable to
mannosylate glucosaminylated PI), respectively (22), were
labeled with [3H]inositol, and their endogenous
CD1d1 and H2 class I molecules were affinity purified. Scintillation
counting of the eluted fractions revealed that
[3H]inositol-derived radioactivity was
associated with affinity purified CD1d1 only but not with control class
I molecules (data not shown). Thus, the association of CD1d1 with PI in
IA-mCD1d is not due to assembly with human 2m.
Instead, the data suggest that CD1d1 use lipids indigenous to the ER to
assemble stable CD1d1 in vivo.
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Selective binding of lipids to CD1d1 in vivo
Among the numerous [3H]2-mannose-, [3H]inositol-, and [3H]mevalonic acid-labeled lipids, only PI and GPI assemble with CD1d1. The label in [3H]2-mannose does not transfer into another sugar but into fucose (31). Therefore, if a fucosylated lipid(s) is indeed synthesized by the cell, it is not associated with CD1d1 because such lipids are resistant to PI-PLC. Among the mannosylated lipids synthesized in the ER are GPI and DPM-glycan. The latter are precursors of N-linked glycans attached to glycoproteins (36). They were not associated with CD1d1 in the [3H]2-mannose and the [3H]mevalonic acid labeling experiments. Thus CD1d1 selects GPI over DPM-glycans.
Two [3H]inositol-labeled lipids specifically associate with CD1d1, one of which is PI-PLC-sensitive PI. The character of the second, slow-migrating, [3H]inositol-labeled compound is unknown because no inositol-containing lipid other than PI is known to exist in mammalian cells. That notwithstanding, inositol-containing ceramides are synthesized by yeast (41) in lieu of sphingomyelin (42). Although not described in mammalian cells, inositolceramide or a similar compound may be synthesized by certain mammalian cells (e.g., tumor cells such as in NS0 plasmacytoma and K562 erythroleukemia lines) used in the present study. Alternatively, because inositol can be converted to glucose, this novel compound associated with CD1d1 could be a glucose-containing lipid other than a PI-glycan. Additionally, this novel compound is resistant to both PI-PLC and PLA2. Thus, further biochemical studies are required to solve the structure of the novel CD1d1-associated compound.
Mammalian cells synthesize nonacylated and acylated inositol-containing GPI anchors for proteins (43), and protein-free GPI anchors although usually contain acylated inositol (44) can contain nonacylated inositol. The GPI extracted from CD1d1 is sensitive to PI-PLC. Therefore, the GPI associated with CD1d1 does not contain an acylated inositol. Furthermore, the PI associated with CD1d1 is resistant to PLA2 (45). PLA2 is sensitive to the stereochemistry of the asymmetric C-atom 2 of glycerophosphatides (33). Therefore, one plausible reason for the resistance of the CD1d1-associated ligand to PLA2 might be the stereochemistry of PI. Whereas the glycerol in cellular PI has its asymmetric C-atom 2 in L-configuration (45), that in the CD1d1 ligand might be in the D-configuration. Alternatively, C-atom 2 in CD1d1-associated PI may be modified by a PLA2-resistant ether linkage to an alkyl group as opposed to the PLA2-sensitive ester linkage. The CD1d1-associated PI is less likely to contain an alkyl group because its mass would be distinct from the observed mass of CD1d1-associated PI reported previously (19).
Thus, CD1d1 exhibits selectivity in the type of natural ligands it binds: selectivity is observed in the type of lipids that bind (e.g., CD1d1 selects PI and GPI over DPM-glycans and numerous other cellular lipids). CD1d1-associated PI is resistant to PLA2, and hence the C-atom 2 has a distinct configuration. Finally, CD1d1 associates with GPI distinct from the common mammalian GPI (i.e., it does not contain an acylated inositol common to mammalian GPI).
Model for CD1 assembly, intracellular traffic, and Ag presentation
The association of PI and GPI with CD1d1 raised interest in the
biological significance of this finding. A thorough search for the
function of CD1d1-GPI complex in vivo revealed that it is inert to the
immune system. Thus, NKT cells do not recognize the CD1d1-GPI complex
(Ref. 20 and data not shown). However, the possibility
remains that the CD1d1-PI can be recognized by some NKT cells.
Consistent with this possibility, one study reported the recognition of
PI in the context of CD1d1 by an autoreactive V14J15-positive NKT
cell hybridoma (18). Thus, this finding supports the
biological relevance of the CD1d1-associated PI described herein. That
notwithstanding, the prevalence of PI-reactive NKT cells in vivo and
their physiological role remain undefined.
The data presented herein support the idea that the lipids associated
with CD1d1 play a chaperone-like role in the assembly of CD1d1. Being a
type I integral membrane protein, akin to the classical Ag-presenting
molecules for which much is known regarding assembly in vivo, the
assembly of CD1d begins as it is cotranslationally inserted into the
ER. Soon thereafter it binds calnexin, a step inferred from the
latter’s association with a group I CD1 synthesized in the absence of
2m (46) as well as with MHC class
I molecules (47, 48, 49). Calnexin probably prevents
aggregation of unassembled CD1 H chains. The monomeric H chain then
assembles with 2m, forming a heterodimer that
is receptive to lipids of the ER and/or the downstream vesicular
compartment(s).
The ability to extract PI from both wild-type CD1d1 and derived mutants
that do not negotiate the recycling compartment (19)
suggests that the association of the cellular ligand with CD1 occurred
in the secretory pathway. The association of lipids with CD1d1 during
biosynthesis probably satisfies two structural requirements: 1)
protecting the deep, nonpolar hydrophobic Ag binding groove from
collapse, and 2) occupying that groove with a ligand that might be
easily exchanged for an Ag in a late secretory vesicle. Thus, PI and
GPI might play a chaperone-like role (Fig. 8) akin to the function of MHC class
II-associated invariant chain in class II assembly in vivo
(50).
|
14J15-positive and -negative NKT cells recognize CD1d1 expressed
by the mutant as effectively as they recognize those expressed by
wild-type cells. Thus, GPI, albeit a major natural ligand of CD1d1, is
neither essential for the assembly of CD1d in vivo nor the natural Ag
of NKT cells (20). Therefore, in the absence of GPI, PI or
other lipids of the ER may assist CD1d1 assembly in vivo. After
assembly of CD1d in the ER, it negotiates the secretory pathway and
arrives at the MHC class II-enriched vesicles either directly from the
trans-Golgi or via the plasma membrane. Targeting to the class
II-enriched vesicles depends on the Yxx
internalization sequence
found at the cytosolic tail of CD1d (39, 51). Here, CD1d
meets the naturally processed Ag(s) whence they arrive at the plasma
membrane for presentation to NKT cells (Fig. 8) (39, 51).
Because human CD1 show structural (11, 52, 53, 54, 55, 56) and
functional (57, 58, 59) similarities to CD1d1, we predict that
the above assembly and intracellular traffic mechanism may be a common
feature of this family of lipid Ag-presenting molecules.
To support our model that PI and PI-glycans act as chaperones, an in
vitro competition assay was performed. sCD1d1, when plate bound, does
not activate NKT cell hybridomas. The addition GalCer but
not PI to plate-bound sCD1d1 specifically activates V14J15 NKT
cell hybridomas. This stimulatory activity of GalCer can be
inhibited by titrating amounts of PI (A. K. Stanic, L. Van Kaer, and S.
Joyce, unpublished data). Considering that PI and GalCer have
similar affinities for CD1d1 (19, 60), the reverse can
also occur. Thus, PI and PI-glycans assembled with CD1d1 in the ER can
be exchanged for another lipid.
Two recent reports indicate that CD1 molecules can be assembled in vitro by oxidative refolding chromatography (61, 62). Albeit an inefficient process, these reports suggest that CD1 can assemble in the ER of mammalian cells without cellular lipids; i.e., CD1 assemble as empty molecules in the ER. However, it is possible that the lipids contained within the H and L chain inclusion bodies could have been included in the in vitro assembly process, thus achieving a low level of folding of CD1. A similar argument can be levied against our data as well, in that lipids resulting from cell death could have contributed to the elution of radio-labeled lipids from CD1d1. However, this co-elution is less likely because, as discussed above, CD1d1 did show selectivity in associating with cellular lipids and did not bind all lipids that incorporated the test label. Additionally, the previously discussed selectivity also rules out the possibility that lipids nonspecifically associate with CD1d1 because they cosegregate within detergent micelles during CD1 extraction.
How lipid loading onto CD1d in the ER is accomplished and how these lipids are exchanged for Ag in endocytic vesicles remain important unsolved questions. Because lipids are membrane embedded, they need to be "plucked out" of the membrane and loaded into the Ag-binding groove of CD1d. This event might require an elaborate molecular machinery similar to the peptide loading complex essential for peptide Ag assembly with MHC class I and class II molecules (40, 50). Solution of the mechanism should provide insights into the basis for lipid-protein interactions in vivo as well as into the evolution of the lipid Ag-presenting system in vertebrates.
In conclusion, we have developed a biochemical method to study protein lipid interactions in vivo. Using this method, the role of cellular lipids in CD1d1 assembly was studied. The results revealed that cellular lipids play a chaperone-like role during the assembly of CD1 family of proteins. Finally, and most importantly, the development of this method sets the stage for the isolation and characterization of the elusive natural Ag(s) of CD1d-restricted NKT cells.
|
Acknowledgments |
|---|
|
Footnotes |
|---|
2 Current address: Department of Microbiology and Immunology, Albert Einstein College of Medicine, Yeshiva University, Bronx, NY 10461.
3 A.D.D. and J.-J.P. contributed equally to this research work.
4 Address correspondence and reprint requests to Dr. Sebastian Joyce, Department of Microbiology and Immunology, Vanderbilt University School of Medicine, Nashville, TN 37232. E-mail address: Sebastian.Joyce{at}mcmail.vanderbilt.edu
5 Abbreviations used in this paper: 2m, 2-microglobulin; PI, phosphatidylinositol; DPM, dolichylphosphorylmannose; ER, endoplasmic reticulum; hAs, heteroantiserum; endo H, endoglycosidase H; C:M, chloroform:methanol; PI-PLC, PI-specific phospholipase C; PLA2, phospholipase A2; GalCer, -galactosylceramide; Rf, relative migration.
Received for publication August 24, 2001. Accepted for publication November 1, 2001.
|
References |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
+ T cells. Nature 372:691.[Medline]
