| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Autoimmune Disease Unit, Burns and Allen Research Institute, Cedars-Sinai Medical Center, Los Angeles, and the University of California, Los Angeles, California 90048
 |
Abstract |
|---|
 Top Abstract I. IntroductionII. TSHR Ectodomain StructureIII. Recombinant TSHR ExpressionIV. TSH and Autoantibody...V. TSHR Autoantibody AssaysVI. Monoclonal Autoantibodies to...VII. T Cell EpitopesVIII. Molecular MimicryIX. Animal Models of...X. Ectodomain Mutations and...XI. Extrathyroidal...Note Added in ProofReferences |
|---|
|
I. Introduction |
|---|
|
TopAbstractI. IntroductionII. TSHR Ectodomain StructureIII. Recombinant TSHR ExpressionIV. TSH and Autoantibody...V. TSHR Autoantibody AssaysVI. Monoclonal Autoantibodies to...VII. T Cell EpitopesVIII. Molecular MimicryIX. Animal Models of...X. Ectodomain Mutations and...XI. Extrathyroidal...Note Added in ProofReferences |
|---|
In 1992, during this logarithmic expansion of data on the TSHR,
Lefkowitz and co-workers (7) made the fundamental observation that
mutations in the cytoplasmic region of the 
1B-adrenergic receptor, a
member of the same G protein-coupled receptor (GPCR) family, led to
constitutive activation of the receptor in the absence of ligand.
Spontaneous mutations in the serpentine portion of the TSHR were soon
sought and found to be the cause of a significant number of toxic
thyroid adenomata in 1993 (8) and familial, nonautoimmune
hyperthyroidism in 1994 (9). Suddenly, the molecular biology of the
TSHR became logical and understandable. Spontaneously occurring TSHR
mutants provided exciting new insights into receptor structure and
function (reviewed in Refs. 10, 11).
The great contrast between the comprehensibility and clinical importance of experiments of nature (primarily in the TSHR transmembrane and cytoplasmic regions) and the experimental mutants (largely in the ectodomain) led many observers to despair of trying to understand the latter or, worse, to be so awed by the complexity of the field as to accept speculation or misinterpretation of data as fact. The present review is an attempt to correct this imbalance. This goal cannot be achieved by simple repetition of a mass of data. We, therefore, provide our interpretation of the subject, for which reason some bias is unavoidable. In particular, we wish to distinguish between conclusions that, at least in our view, are unequivocal and those that remain hypotheses or can be discarded. We have tried to be fair but accept that some of our interpretations may prove to be wrong in the future.
In some respects, investigation of the TSHR ectodomain is more difficult than that of the serpentine segment of the receptor. Thus, mechanistically informative, spontaneous mutations causing thyroid dysfunction are less likely to occur in the ectodomain. With some exceptions, major changes in TSHR binding and function are unlikely to be caused by single amino acid alterations in the ectodomain. Moreover, the lesser conservation of the TSHR ectodomain than of the serpentine region reduces the lessons that can be learned from other members of the GPCR superfamily. With the exception of its leucine-rich repeats (LRRs), the conformational structure of the TSHR ectodomain is an enigma. Nevertheless, despite these disadvantages, studies on the ectodomain cannot be avoided if we are to understand the actions of TSH and TSHR autoantibodies.
Even since its molecular cloning, the TSHR has lived up to its reputation of being a very difficult molecule to study. Expression of significant quantities of recombinant receptor in a form capable of binding TSH and autoantibodies has been an exceptionally difficult, frustrating, and time-consuming undertaking. Because the expression of many other recombinant proteins is a straightforward, mundane, and technically uninteresting subject, difficulties experienced with the TSHR are commonly attributed to lack of experience or suitable facilities. The conceptual hurdles in TSHR expression, including the highly conformational nature of the protein or the extremely high glycan content of the ectodomain, are not generally appreciated. It is also not well understood how great a handicap is the very low concentration of TSHR autoantibodies in patients’ sera, why the cloning of human autoantibodies from patients’ B cells has been far more difficult for the TSHR than for thyroid peroxidase (TPO), why it is difficult to generate antibodies in experimental animals that mimic the actions of autoantibodies and, finally, why it has been so difficult to develop an animal model of Graves’ disease by simply injecting recombinant antigen. In the future, determination of the three-dimensional structure of the TSHR ectodomain will also be exceptionally difficult and success is by no means assured.
On the other hand, it is precisely these difficulties and challenges that make the TSHR, especially the ectodomain, such a rewarding molecule to study. We are fortunate to have the opportunity to study the role of the ectodomain in TSHR activation by the immune system in Graves’ disease, a common clinical phenomenon only rarely observed with other members of the GPCR family. Recent evidence also now reveals that the TSHR subunit structure is unique to presently known members of the superfamily. Further improvements in autoantigen generation will no doubt lead to advances in the diagnosis of Graves’ disease and, possibly, to immunotherapy that will replace the present options of surgery, radioiodine ablation, or long-term antithyroid drug therapy.
|
II. TSHR Ectodomain Structure |
|---|
|
TopAbstractI. IntroductionII. TSHR Ectodomain StructureIII. Recombinant TSHR ExpressionIV. TSH and Autoantibody...V. TSHR Autoantibody AssaysVI. Monoclonal Autoantibodies to...VII. T Cell EpitopesVIII. Molecular MimicryIX. Animal Models of...X. Ectodomain Mutations and...XI. Extrathyroidal...Note Added in ProofReferences |
|---|
). Even excluding its carbohydrate
component (see below), the TSHR ectodomain represents nearly half the
mass of the holoreceptor, a very large size compared with the vast
majority of members of the GPCR superfamily with seven
membrane-spanning regions. The complexity and size of the TSHR
ectodomain derives from the evolutionary addition of nine exons (12) to
the single prototypic exon encoding receptors, such as for rhodopsin
and epinephrine, that have small ectodomains. This disparity in
ectodomain size is consistent with the difference in size of the
ligands for their cognate receptors, although some interesting
disparities exist (the Ca++ receptor has a very small
ligand and a large ectodomain). All introns in the TSHR are in phase 2,
meaning that splice junctions occur between the second and third bases
of each codon.
|
58–288), corresponding primarily to the
LRRs (see below). The less conserved N-terminal and C-terminal regions
of the TSHR are interesting in that they contain additional residues in
comparison to the LH/CG or FSH receptors. This difference is greatest
between the TSH and LH/CG receptors (LH/CGR). Thus, alignment of their
amino acid sequences reveals that these additional residues are
clustered in two approximate regions, giving the appearance of
"insertions" of 8 amino acids (residues 38–45) and 50 amino acids
(residues 317–366) (Fig. 1
). It must be emphasized that, given the low
homology between these regions of the TSHR and LH/CGR ectodomains, the
designated boundaries of the two insertions must be regarded as
approximate. The genomic structure of the TSHR ectodomain indicated that the division into exons corresponded to the limits of monomers or multimers of LRRs (12). Analysis of the amino acid sequence of this region suggested nine LRRs between amino acid residues 58–277 (13, 14). These nine repeated sequences show strong intersegment homology. The consensus sequence proposed was xLxxTxxLTxLPxxAFxxLxxLxxxL. LRRs 1–7 corresponded to the amino acids in exons 2–8 (12). LRRs 8 and 9 are situated in exon 9 (12). A variant dog TSHR cDNA has been described with deletion of amino acids 61–86 (15), corresponding to the LRRs encoded by exon 3. More recently, computer modeling based on ribonuclease inhibitor (16), the first molecule with LRRs to have its three-dimensional structure solved, has suggested a slightly different structure with eight LRRs between amino acid residues 54–254 corresponding to exons (17).
B. Subunit structure
1. Single and two-subunit forms of the TSHR. Clarification of
the subunit structure of the TSHR is fundamental for understanding the
mechanism of action of TSH and TSHR autoantibodies. Before its
molecular cloning, studies too numerous to mention described a TSHR
with a variety of subunits with molecular masses ranging from 15
kDa to 200 kDa. The most credible data, obtained by
[125I]TSH cross-linking to thyroid plasma membrane
preparations or to cultures of FRTL-5 rat thyroid cells, provided two
schools of thought regarding TSHR subunit structure (reviewed in Ref.
18). On the one hand, the TSHR was suggested to be a 80-kDa heterodimer
with a 50-kDa hormone-binding A subunit linked by disulfide bonds to a
membrane-spanning 30-kDa B subunit (19, 20). Surprisingly, using
similar reagents, other investigators suggested that the TSHR contained
two subunits of 53 and 40 kDa (18), or three subunits of 31, 17, and 63
kDa (21), none of which were linked by disulfide bonds.
The molecular cloning of the TSHR in 1989 resolved this controversy but also generated puzzling and apparently contradictory observations that have led to considerable misunderstanding or skepticism. Very recently, important pieces of the puzzle have fallen into place and reveal a receptor more complex and fascinating than previously imagined. The following is our perspective of the series of advances and controversies in recent years.
1. Covalent cross-linking of radiolabeled TSH to membranes prepared from nonthyroidal mammalian cells expressing the recombinant human TSHR (22, 23) essentially confirmed the prior observations of Rees Smith and associates (19, 20), although these investigators slightly underestimated the sizes of the A and B subunits (see below).
2. The fact that the TSHR was encoded by a single mRNA (1, 2, 3, 4, 5, 6) established that TSHR subunits are formed by intramolecular cleavage of a single polypeptide chain. This situation contrasts with the subunits of its cognate ligand, TSH, which are encoded by separate genes.
3. The most surprising initial finding with the recombinant human TSHR was that cross-linking of [125I]TSH to intact cells in monolayer culture, rather than to membranes from cell homogenates, revealed two types of receptors on the cell surface. In addition to the two-subunit form previously observed in broken cell preparations, a considerable proportion of the receptors on the cell surface were present as a single, uncleaved chain without subunits (22). Both the single chain and the two-subunit forms of the TSHR bound TSH with similar high affinity. These data raised the possibility that the monomeric TSHR was the physiological form and that the two-subunit TSHR was present to a greater extent after cell homogenization because of lysosomal enzyme release (22). In retrospect, a single-chain TSHR had been observed on TSH cross-linking to intact, cultured rat (FRTL5) cells but had been considered a precursor (24). Support for the concept of a single-chain TSHR was also obtained in immunoblot studies of cell extracts in which only monomeric (25) or predominantly monomeric (26) forms of the TSHR were detected. As discussed below, however, it is difficult to interpret studies involving the immunodetection of the TSHR without analysis of the carbohydrate composition of the TSHR and without knowing whether the monomeric receptors detected are on the cell surface.
4. The development of murine monoclonal antibodies (mAbs) to the TSHR
greatly facilitated investigation of its subunit structure. Data
obtained with these reagents challenged the concept of a physiological
single-chain form of the TSHR, for the following reasons (27, 28): (i)
Only A and B subunits (renamed and ß), and not even a trace of
monomeric receptor, could be detected after TSHR affinity purification
from homogenized human thyroid tissue. (ii) TSHR cleavage into A and B
subunits (original terminology retained for reasons stated below)
occurred late in the synthetic process, either in the Golgi complex or
at the cell surface. Thus, kinetic studies of precursor-labeled
recombinant TSHR expressed in mouse L cells revealed that TSHR
intramolecular cleavage occurred only after maturation of N-linked
glycan moieties from the high-mannose to the complex form (28). Glycan
trimming and maturation to complex forms occur in the Golgi complex and
are required for protein trafficking to the cell surface (29). These
data suggested that TSHR A and B subunit formation was an ordered
physiological event rather than a proteolytic artifact consequent to
cell homogenization. (iii) Unlike in human thyroid tissue extracts,
homogenates of transfected mouse L cells contained large amounts of
TSHR monomers with immature, high-mannose carbohydrate (28). This
material was suggested to reflect the abnormal intracellular
accumulation of TSHR precursors consequent to receptor overexpression
overwhelming the protein-processing capacity of the cells.
Based on all of these findings, the perception has arisen that the monomeric TSHR is an unphysiological artifact of transfected nonthyroidal cells, unrelated to the situation in thyroid tissue in vivo. A reconciliation of these apparently contradictory data is provided below.
2. Loss of a C peptide during intramolecular cleavage of the TSHR
into two subunits. A major surprise in the past year has been the
realization that intramolecular cleavage of the TSHR into
disulfide-linked A and B subunits does not occur at a single site as
previously suspected, but at two sites (30). Consequently,
although it has not yet been directly identified, a putative C peptide
is released, analagous to the derivation of insulin from proinsulin
(Fig. 2). This remarkable phenomenon is
unique to presently known members of the GPCR superfamily. Three lines
of evidence suggested the presence of two cleavage sites in the TSHR:
|
2. The sum of the TSHR A and B subunit polypeptide chains is
5–7 kDa smaller than that of the uncleaved, monomeric TSHR. The
predicted and observed size of the single chain polypeptide backbone is
84.5 kDa. Yet, the enzymatically deglycosylated A subunit polypeptide
and the nonglycosylated B subunit are no more than 35 kDa and
42 kDa in size, respectively (27, 28, 30, 31). Although size
estimations cannot be very precise, they are sufficiently reproducible
among different laboratories using different methodologies for TSHR
detection to suggest that a piece of the TSHR has been lost during
intramolecular cleavage. In addition to the predominant subunit forms,
lesser amounts of A and B subunits of slightly larger size exist,
suggesting that there is a minor degree of incomplete cleavage at
both sites.
3. A c-myc epitope strategically inserted within the
putative C peptide region (Fig. 2) was lost in the cleaved, but not in
the uncleaved, monomeric receptor (30). Thus, despite the selective
inability of a c-myc mAb to detect the two-subunit TSHR,
this receptor was readily detected with another mAb to the N terminus
of the A subunit. Loss of the c-myc epitope by cleavage at a
single site within this epitope was unlikely for a number of reasons.
Thus (i) single cleavage within amino acid residues 338–349 (the
c-myc epitope) would generate a B subunit of 46–47 kDa,
clearly larger than the actual size (33–42 kDa) observed
experimentally (27, 28, 32). (ii) Chimeric receptor TSH-LHR-4 (33) and
a deletion mutant (residues 317–366) of the wild-type TSHR (34) lack
the region in which the c-myc epitope was inserted, yet
still cleave into two subunits (22, 23). (iii) The c-myc
epitope substitution substantially alters the sequence of the wild-type
receptor. Cleavage within the c-myc epitope would,
therefore, imply nonspecificity in the amino acid sequence at a TSHR
cleavage site.
Purification and characterization of the putative TSHR C peptide would provide proof for the two-cleavage site hypothesis. However, it is not presently feasible to detect the release of a small TSHR polypeptide fragment into the culture medium from the TSHRmyc cell line that is not overexpressing the TSHR. Very recently, the concept of two cleavage sites has been confirmed by mutagenesis of the TSHR (see Section II.B.5). Because there has been misunderstanding on this issue, the existence of two cleavage sites should not be construed as evidence for three TSHR subunits or evidence against a physiological monomeric receptor. Thus, a C peptide would not be a subunit, but a fragment released during the formation of the two-subunit (A and B) TSHR. Further, the release of a C peptide during intramolecular cleavage does not negate evidence for the simultaneous presence of other single-chain, monomeric TSHR on the cell surface.
3. TSHR subunit terminology. Because there is an evolving
change in the literature on the nomenclature of the TSHR subunits, it
is necessary to comment on our choice for describing the TSHR subunits
as A and B, rather than and ß. We suggest that the former is
preferable for three reasons. First, use of A and B gives credit to the
original discoverers of the two TSHR subunits (19). Second, a C peptide
(rather than a 
-peptide) conveys the concept well established for
conversion of proinsulin to insulin. Finally, TSH, the ligand of the
TSHR, contains - and ß-subunits, and use of a different
terminology will avoid confusion in future ligand-receptor structural
studies.
4. Shedding of the TSHR A subunit. There is evidence to suggest that the majority of TSHR on the cell surface shed their A subunits into the circulation (27). This hypothesis, which could have important implications in the pathogenesis of Graves’ disease, arose from the observation that 3-fold fewer A subunits than B subunits were recovered from human thyroid homogenates using TSHR-specific mAbs (27). Consistent with the disproportionate number of B subunits in human thyroid tissue, shed A subunits in the anticipated number were recovered from the culture medium of human thyrocytes, as well as from mouse L cells and Chinese hamster ovary (CHO) cells expressing the recombinant human TSHR (35). Moreover, A subunits were also reported to be detectable in human serum (35, 36, 37). Further evidence suggests that shedding of the A subunit is caused by cleavage by a matrix metalloprotease (35) accompanied by cell surface protein disulfide isomerase disruption of the disulfide bond(s) linking the A and B subunits (38).
Although a clearly established in vitro phenomenon, questions remain about the physiological nature of TSHR ectodomain shedding. Thus:
1. TSHR shedding is particularly evident when cells are "starved" in serum-poor tissue culture medium (35), conditions associated with decreased cell viability and autophagy of self-proteins (39).
2. The shed TSHR A subunit is smaller than the non-shed A subunit because of a reduction in its carbohydrate content. The protein backbone is of similar size.
3. No B subunit excess was apparent in CHO cells expressing a sufficient number of TSHR to permit direct subunit detection without the need for prior affinity purification (30).
5. Localization of the two cleavage sites in the TSHR ectodomain. Much progress has been made in characterizing the upstream and downstream cleavage sites, hereafter referred to as site 1 and site 2, respectively.
a. Site 1.
This site delineates the C terminus of the
glycosylated A subunit (Fig. 2). Initial efforts to deduce the location
of site 1 focused on the size of the A subunit. Because the size of the
TSH-binding A subunit was the same in the wild-type TSHR as in a mutant
receptor lacking amino acid residues 317–366, cleavage at site 1 was
likely to be closely upstream of residue 317 (22, 34). Subsequent data,
however, raised doubts as to this localization for site 1. Thus (i) It
was reported that the C terminus of the A subunit must be downstream of
residue 366 (26), rather than closely upstream of residue 317. The
basis for this conclusion was that the glycosylated, N-terminal portion
of the TSHR (the A subunit) was recognized by an antiserum to amino
acids 352–366, a region termed the "immunodominant peptide" (40, 41). (ii) It was suggested that the TSH cross-linking data were
potentially misleading because TSH bound with high affinity to the B
subunit (26). (iii) Mutagenesis of two Arg/Lys-rich regions closely
upstream to amino acid 317 failed to prevent TSHR intramolecular
cleavage and the formation of a normally sized A subunit (42). These
regions had been considered likely cleavage sites because of their
striking homology to subtilisin-related proprotein convertase motifs.
(iv) The A subunit polypeptide backbone, after enzymatic
deglycosylation, was observed to be 35 kDa (27, 28, 30) or 42 kDa (31)
in size, which would place site 1 in the region between amino acid
residues 330–390.
Surprisingly, despite all this evidence that site 1 (the C terminus of
the A subunit) was not upstream of residue 317, extensive mutagenesis
of the entire region downstream of residue 317 as far as cleavage site
2 (see below) failed to eliminate site 1 (43). To be informative, these
mutations were performed on a background of a TSHR with a mutation
(GQE367–369NET) that abolished cleavage at site 2.
Fortuitously, the clue to the location and nature of site 1 was
obtained in studies on the effect of the proteolytic enzyme, trypsin,
on TSHR structure. Thus, light trypsinization was observed to delete a
small (1–2 kDa) polypeptide fragment containing an N-linked glycan
moiety from the C terminus of the A subunit (44). This glycan moiety
could only be the most downstream (sixth) N-linked glycosylation motif
in the TSHR ectodomain (residue 302), the more proximal (fifth)
potential N-linked glycosylation site being far upstream at amino acid
198 (Fig. 3). Therefore, site 1 could be
localized to the small region between amino acids 303 and the vicinity
of residue 317, as originally suggested by the TSH cross-linking data
(22). Once again, however, extensive mutagenesis in this region failed
to prevent cleavage at site 1 (44).
|
b. Site 2.
Chimeric TSH-LH/CGRs were the key to localizing
this cleavage site. Indeed, chimeric receptors had also provided the
first clue for the presence of two separate cleavage sites in the TSHR
(see Section II.B.2.a). Obviously, mutations at one cleavage
site would only be informative (prevention of subunit formation) when
performed on a background of a receptor in which the other site was
lacking. The search for site 2 in one domain, therefore, involved
mutagenesis of this region using a chimeric receptor in which cleavage
at site 1 was prevented by replacement with the corresponding region of
the noncleaving LH/CGR.
Fortunately, mutations within the site 2 region involved a ‘mini’ chimeric receptor approach in which TSHR residues were replaced with the corresponding residues of the LH/CGR. This procedure rapidly identified TSHR amino acid residues 367–369 to be critical for cleavage at site 2. Remarkably, replacement of the same amino acid residues with Ala did not prevent cleavage at site 2. The simultaneous substitution of three LH/CGR residues (NET291–293) (single letter amino acid code) for the corresponding TSHR residues GQE367–369 were required to abolish TSHR cleavage at site 2. This mutation (GQE367–369NET) introduced a motif for an N-linked glycosylation site (N-X-S/T; where X can be a wide range of amino acids with variable efficiency for glycosylation) (45). Additional mutagenesis of these three residues provided further support for the likelihood that it was the lack of a glycosylation site in the TSHR, rather than the presence of a particular amino acid motif, that permitted cleavage at site 2. Thus, replacement of TSHR residues GQE with NQT (glycosylation motif), but not with NQE (incorrect motif), largely prevented cleavage at site 2. These mutations also confirmed the existence of the upstream site 1. Thus, when the glycosylation sequons NET or NQT were introduced into the wild-type TSHR, intramolecular cleavage was restored.
c. Summary.
From the above data, the following conclusions can
be made regarding the sites of TSHR intramolecular cleavage: (i)
Cleavage at site 1 occurs downstream of TSHR residue Asn302 and
(probably) upstream of residue 317. (ii) TSHR amino acids GQE at
positions 367–369 are critically related to cleavage at site 2. Thus,
cleavage is prevented by the introduction at this site of an N-linked
glycan motif, known to be glycosylated at the same site in the
noncleaving LH/CGR (46). (iii) Neither cleavage at site 1 nor at site 2
involve a specific amino acid motif. (iv) The TSHR 50-amino acid
insertion (residues 317–366) is required for cleavage at site 1, but
not at site 2.
6. Mechanism of TSHR cleavage into subunits. The thrombin receptor, another member of the GPCR family, is activated by the ability of its own ligand to proteolytically remove a fragment of the receptor ectodomain, exposing a "tethered ligand" (47). However, the attractive hypothesis that TSH could be the protease responsible for cleavage of its cognate receptor could be discarded because two-subunit TSHR was detected in transfected cells cultured in the absence of TSH (22, 28).
A second potential mechanism for TSHR intramolecular cleavage was involvement of one of a series of subtilisin-like endoproteases that convert many mammalian proproteins and prohormones into the active form by clipping at a pair of basic amino acid residues (48). Indeed, as mentioned above, such motifs were noted to be present in the vicinity of TSHR cleavage site 1 (22, 27). However, mutagenesis of these motifs fails to prevent TSHR cleavage, suggesting the absence of a role for classic subtilisin-related proprotein convertases (42).
More recently, matrix metalloproteinases (MMPs) have attracted much
attention for their role in releasing from the cell surface a large
number of membrane-anchored growth factors, receptors, adhesion
molecules, and proinflammatory cytokines, such as tumor necrosis
factor- (reviewed in Refs. 49, 50). This mechanism is
important in view of the potential pharmacological ability to influence
disease processes. Of interest from the point of view of the TSHR,
ligand-induced cleavage of the V2 vasopressin receptor
(also a member of the GPCR family) appears to be mediated by a plasma
membrane MMP (51). One MMP inhibitor, BB-2116, strongly inhibits TSHR
cleavage into A and B subunits (35), suggesting a role for MMPs in this
process. Unfortunately, MMPs do not interact with a specific amino acid
motif, precluding the ability to predict the TSHR residues with which
these enzymes may interact. The concept of MMP-mediated cleavage of the
TSHR is supported, but not proven, by the absence of specific amino
acid motifs for cleavage at sites 1 and 2 (43, 44). Another difficulty
regarding the MMP concept is that inhibitors of these enzymes are not
necessarily specific and may inhibit other types of proteases that are
active at the cell surface (52). At present, therefore, although
involvement of MMPs represents the most credible theory for TSHR
intramolecular cleavage, confirmatory data for this mechanism are
awaited.
Regardless of the actual enzyme involved in TSHR cleavage, the mutagenesis studies described above provide interesting new insight into, and generate new hypotheses regarding, the mechanism of cleavage. Thus (i) the presence or absence of glycosylation at cleavage site 2 represents a novel mechanism by which two closely related receptors have evolved to have a different subunit structure. (ii) The lack of a specific amino acid motif at TSHR cleavage site 2, together with the abrogation of this site by an N-linked glycan, suggests that carbohydrate may prevent the binding or action of a proteolytic enzyme such as a MMP. TSHR residues 367–369 may, therefore, not necessarily be the exact cleavage site. Precise identification of this site may require the purification of the TSHR B subunit and determination of its N-terminal sequence. (iii) Because the 50-amino acid insertion (residues 317–366) is necessary for cleavage at site 1 but does not, itself, appear to harbor the cleavage site, this region may also be a binding site or ‘anchor’ for a protease. Alternatively, amino acid residues 317–366 may, themselves, function as a protease and clip an adjacent region of the TSHR. (iv) Different proteases may be involved at sites 1 and 2. This possibility is raised by the observation that deletion of the 50-amino acid insertion (a potential protease or protease-binding site) abrogates cleavage at site 1 but not at site 2.
7. Role of the TSHR 50 amino acid ‘insertion’. Although the
precise boundaries of these additional 50 amino acids in the TSHR are
uncertain (because of low homology in adjacent regions of the
glycoprotein hormone receptors), this segment has been the subject of
intense study. The very hydrophilic nature of residues 317–366 led us
to speculate that it was a projection on the exterior of the TSHR
molecule, perhaps important in ligand specificity (34). Surprisingly,
however, its deletion had no effect on TSH binding or on TSH-mediated
signal transduction (34). The deduced superficial topography of TSHR
residues 317–366 was also the reason for selection of this region for
c-myc epitope tagging (53). The critical role of amino
acids 367–369 at TSHR cleavage site 2, as well as the localization of
site 1 to the proximate vicinity of residues 303–316, support the
original demarcation of the 50-amino acid insertion (2). Cleavage sites
at either end of residues 317–366, together with the absence
of Cys residues within this region, support the concept that this
insertion is, indeed, the putative C peptide lost during intramolecular
cleavage.
An interesting finding regarding the 50-amino acid insertion is that a
synthetic peptide corresponding to its C-terminal region (residues
352–367) is highly immunogenic when injected into rabbits with
adjuvant and is also reported to be recognized by TSHR autoantibodies
in the majority of Graves’ sera (41, 54). Further, an antiserum to a
closely related synthetic peptide (residues 352–366) recognizes the
TSHR A subunit in FRTL-5 rat thyroid cells, but very poorly in
transfected COS cells (26). These reports must now be reevaluated in
the light of the discovery that the immunogenic peptide is deleted from
the TSHR A subunit after intramolecular cleavage at site 1 (Fig. 2).
There is evidence that cleavage at sites 1 and 2 may not be complete
(30). Residues 352–366 could, therefore, be present on an incompletely
cleaved "large" A subunit resulting from cleavage at site 2 alone.
The 50-amino acid insertion also contains the epitope (residues 354–359) for a murine mAb to the TSHR (2C11) that recognizes the native receptor on the cell surface (55). Light trypsinization of intact CHO or COS-7 cells destroys antibody recognition of this epitope and also activates the TSHR (56). The potential role of intramolecular cleavage in constitutive TSHR activity is discussed below (Section II.B.9). Because its epitope appears to be on the TSHR C peptide, it is likely that mAb 2C11 is recognizing the single-chain TSHR and not the cleaved TSHR on the cell surface. Further, these data indicate that trypsin is removing at least part of the C peptide from the monomeric receptor. Support for this conclusion is the direct observation that trypsinization of intact cells converts all single-chain TSHR on the cell surface into cleaved forms with A subunits equal in size to, or smaller than, the A subunit formed by spontaneous cleavage (44).
8. Reconciliation and interpretation of a confusing literature. Strong opinions and prejudices exist as to the subunit structure of the physiological TSHR. The most widely accepted view is that only two-subunit TSHRs exist on thyroid cells in vivo and that single-chain TSHRs are, in large part, an artifact of transfected nonthyroidal cells (27, 28). Another camp argues that the TSHR subunits are an in vitro proteolytic artifact (26, 57). A third view (our opinion) is that both single-chain and two-subunit TSHRs are present and functional on the surface of thyroid cells.
To reconcile these diametrically opposing views, it is necessary to appreciate that the type of TSHR "seen" experimentally depends greatly on the methodology used. Thus, the TSHR or its subunits have been detected by (i) [125I]TSH covalent cross-linking; (ii) Western (immuno) blotting of TSHR forms separated on PAGE; and (iii) immunoprecipitation of [35S]methionine/cysteine precursor-labeled TSHR. In our view, TSH cross-linking is the most powerful (surprisingly underused) tool because it is the only one that specifically detects mature TSHR on the surface of intact cells. On the other hand, characterization of the TSHR by immunoblotting or immunoprecipitation using antibodies must be performed on broken cell preparations. Antibodies (unlike TSH) may recognize many forms of the TSHR including intracellular precursors and degradation products incapable of binding TSH. Finally, even using the same antibody for detection of the TSHR in homogenates from the same cell line, the TSHR can be seen as either predominantly single chain or predominantly two subunit, depending on the methodology employed (30).
Other important procedural differences that can influence the data obtained should be considered. For example (i) the homogenization procedure of cultured cells and very firm thyroid tissue can vary markedly in duration and vigor; and (ii) the need for prior affinity purification to detect a signal (27, 28) has the potential to introduce bias toward detection of one or another form of TSHR or TSHR subunit. This potential handicap is overcome by TSHR overexpression which, by providing a strong signal, eliminates the requirement for prior affinity purification (32).
Bearing in mind these methodological differences, we suggest the
following: (i) Regardless of how much immature, monomeric TSHR is
detected by immunological means in transfected, nonthyroidal cells,
when assessed by TSH cross-linking, there is no doubt that both
single-chain and two-subunit TSHRs are present on the surface of intact
thyrocytes (24) and transfected nonthyroidal cells (22), at least in
culture. (ii) When detected by TSH cross-linking, the proportion of
single-chain vs. two-subunit TSHRs detected on the cell
surface is remarkably similar (1:1) over a very wide range of
receptor numbers (15,000 or 2 million TSHRs per cell) (22, 32), and
irrespective of whether thyrocytes (24) or transfected CHO cells are
examined. These observations do not support the hypothesis that
single-chain TSHRs represent an unphysiological accumulation in
transfected cells. Taking all these data together, we believe that it
is very difficult to assess the physiological role of TSHR subunit
structure in broken cell preparations without consideration of the
methodology used and without focusing on the forms of TSHR expressed on
the cell surface, such as by analysis of their glycan composition (see
below).
We are also drawn to comment on the common perception (31, 35, 57) that we presently support the viewpoint that the two-subunit TSHR is a proteolytic artifact. In our original cross-linking data, we detected only two-subunit (and no monomeric) TSHR on TSH cross-linking to membrane preparations from homogenized, transfected CHO cells (22). In contrast, both two-subunit and single-chain TSHRs were detected on the surface of intact cells (22). On this basis, we raised the logical (in our view) hypothesis that the single-chain TSHR could be a physiological receptor in vivo, rather than being an immature precursor (24, 58), and that the two subunit TSHRs could be a degradation product, either artifactual or physiological. However, subsequent data on TSHR subunit structure reported by ourselves (23, 30, 32, 42, 43, 59) and others (27, 28, 31, 35) have clearly indicated that TSHR intramolecular cleavage is not an artifact. On the other hand, we stand by our original observation that a single-chain TSHR with high affinity for TSH also exists on the surface of intact cells (22). In our view, the single-chain vs. two-subunit controversy is moot; both receptor forms are present and functional. Rather, the outstanding question at present is whether or not the dual expression of single-chain and two-subunit TSHRs on the surface of cultured cells also occurs in thyroid cells in vivo. Obviously, this will be a very difficult question to answer.
9. The functional importance of TSHR cleavage. The evolutionary divergence of the TSHR into a receptor that cleaves into two subunits is unique and enigmatic. As mentioned above, TSH binds to both cleaved and uncleaved forms of the TSHR with similar high affinity (22). Moreover, TSH action does not require a cleaved receptor. Thus, TSH can activate chimeric TSH-LH/CGRs that do not cleave into two subunits (59). On the other hand, circumstantial evidence suggests that cleavage into subunits may influence the basal, as opposed to the TSH-induced, level of TSHR activity. Thus:
1. The TSHR ectodomain, particularly the region involved in intramolecular cleavage, plays a role in signal transduction (33, 60). Moreover, the spontaneous mutation of Ser281 in the TSHR ectodomain leads to increased constitutive activity (61, 62). This residue is very close to the C terminus of the A subunit that may be more exposed after cleavage at site 1 (44).
2. Unlike the noncleaving gonadotropin receptors, the TSHR is "noisy" in that it has significant constitutive (ligand-independent) activity (10, 63, 64, 65).
3. Light trypsinization of cells expressing the TSHR leads to receptor activation with the loss of an epitope at amino acid residues 354–359 (56), an epitope unlikely to be present in the two-subunit receptor (30, 43, 44). Trypsin also converts monomeric TSHR on the cell surface into two-subunit forms, as well as clipping the C terminus of the A subunit (44).
4. Deletion by mutagenesis of residues 339–367 is also reported to increase TSHR constitutive activity (66). However, the effect of this deletion on TSHR subunit structure is unknown.
Given the rapid progress in this field, an answer to the question of whether or not TSHR constitutive activity is related to intramolecular cleavage is likely to be available in the near future. Another possible consequence of TSHR intramolecular cleavage that will require future investigation relates to the pathogenesis of Graves’ disease. Thus, the release of a C peptide from the TSHR may be associated with the very common occurrence of disease-causing autoantibodies, a phenomenon rarely (if ever) encountered with other known members of the GPCR family that do not have a similar subunit structure.
C. Disulfide bonds in the TSHR ectodomain
As for any large protein, disulfide bonding is vital for the
correct folding of the TSHR and the maintenance of its tertiary
conformation. In addition, disulfide bonds are involved in the
quaternary structure of the TSHR A and B subunits (19, 20).
Determination of which Cys residues form pairs would provide important
information on the TSHR pending the solution of its three-dimensional
structure by crystal x-ray diffraction, a task that may never be
possible (see below). The most important present question, on which
there has been much speculation, is which Cys residues are involved in
the association of A and B subunits in the two-subunit form of the
TSHR.
There are 11 Cys residues in the TSHR ectodomain that could form five
pairs, with one residue remaining as the orphan. An additional two Cys
in the first and second extracellular loops are conserved in the GPCR
family, and, like other known members of the family, are likely to be
paired with each other. The 11 Cys in the TSHR ectodomain can be
classified into four groups (Fig. 4).
|
2. Group II. Also characteristic of the LRR family is a Cys cluster between residues 283–301 at the C terminus of the LRR region.
3. Group III. Three Cys are clustered between residues 390–408 at the extreme C terminus of the TSHR ectodomain, close to its entry into the plasma membrane.
4. Group IV. A solo Cys is present at residue 176, within the LRR region. Because of its isolation and the relative rigidity of the LRR structure, Cys176 is a good candidate to be the orphan.
At present, there are no direct data to indicate which Cys form the five pairs. However, a number of indirect lines of evidence permit informed speculation. Thus, (i) localization of the approximate sites of TSHR cleavage (see above) establishes Cys residues 390, 398, and 408 (group 3) to be the only ones on the B subunit. One or more of these residues must, therefore, be responsible for linkage to the A subunit. (ii) Of the residues in group 3 on the B subunit, Cys390 cannot be the sole link to the A subunit because its mutation reduces, but does not abolish, TSH binding (60, 68). Loss of the A subunit would be incompatible with TSH binding. TSH binding, on the other hand, is completely abrogated by mutation of Cys398 (60, 68) and Cys408 (68), making one or both of these residues the prime candidates for tethering of the A subunit. (iii) Two clusters of Cys residues on the A subunit could link with the B subunit; namely, the four Cys in group 1 or the three Cys in group 2. The three-dimensional model of the TSHR LRR region (17) makes Cys pairing between groups 1 and 2 very unlikely. (iv) All of the Cys residues in group 1 (68), except for Cys41 (68, 69), can be mutated without loss of high-affinity TSH binding. Again, because disruption of A and B subunit linkage is likely to be ‘fatal’ to high-affinity TSH binding, only Cys41 remains a candidate for this function. (v) Mutation of group 2 Cys residues, in particular Cys283 and Cys284, have a devastating effect on TSH binding (68), making these two residues (in our view) the most likely to be pairing with Cys398 and Cys408 in group 3. (vi) The hypothesis that Cys301 (group 2) and Cys390 (group 3) are paired (68) is supported by the recent evidence that intramolecular cleavage of the TSHR releases a C peptide approximately between these two residues (30, 43, 44). Further evidence for this likelihood is that mutation of either Cys301 or Cys390 produces the identical effect of reducing the affinity of TSH binding (60, 68). Disruption of a disulfide bond through either of its two residues would be expected to produce a similar phenotype. However, the fact that TSH binding persists despite these mutations indicates that a Cys301–390 bond cannot be the sole link between the A and B subunits. This conclusion is supported by the deletion of Cys301 by trypsin without the abolition of TSH binding (44).
Taking all these data together, our working hypothesis is depicted in
Fig. 4. The essential feature is that Cys283,284 and Cys398,408 link
the A and B subunits. If correct, Cys41 remains the enigma. Thus, in
this schema, the four Cys in group 1 remain isolated from the remainder
of the ectodomain. Yet, only mutation of Cys41 abolishes TSH binding.
If disulfide bonding among these four Cys residues is important in the
three-dimensional structure of the TSHR, then mutation of the unknown
Cys paired to Cys41 should produce an identical effect. If Cys41 is not
paired within group 1, then it is likely to be paired with a Cys in
group 3. Therefore, our hypothesis that Cys41 does not link with the B
subunit is wrong, or Cys41 is, indeed, an important component in the
TSH-binding site (34).
D. Carbohydrate moieties in the TSHR ectodomain
Early reports on affinity enrichment using lectins of TSH-binding
activity from thyroid tissue revealed that the TSHR, like most other
cell surface proteins, was a glycoprotein (70, 71, 72, 73). TSH-binding
activity could be adsorbed by a wide selection of lectins and varied
depending on the TSHR species, making it very difficult to define the
glycan moieties involved other than that they were N linked (73).
Modest progress has been made since the molecular cloning of the TSHR.
The human TSHR ectodomain contains six potential N-linked glycosylation
sites (Fig. 1), but evidence on which of these sites actually contains
glycan is limited. All sites are on the A subunit of the cleaved TSHR,
the sixth and last being at Asn302.
Perhaps the biggest surprise has been the unexpectedly high ectodomain glycan content of the TSHR. Estimates of 10–14 kDa (31, 57) or even 18 kDa (27, 28) may be too low. In transfected mouse L cells (28) and CHO cells (30), 25–27 kDa of N-linked glycan has been detected. Even a truncated TSHR A subunit lacking N302 has 20 kDa of carbohydrate (74). Based on these estimates, the recombinant TSHR A subunit expressed in mammalian cells contains more than 40% glycan, a level nearly consistent with a proteoglycan rather than a glycoprotein! The suggestion that TSHR glycosylation is greater in transfected cells than in thyroid tissue (28) is a distinct possibility and needs further evaluation. Another possible explanation for the apparent discrepancy between the glycan content of thyroid tissue (19, 20, 27, 28) and cultured cells (28, 30) is that the glycan on Asn302 is removed during thyroid tissue homogenization, as it is with trypsin (44).
Regarding which of the N-linked glycosylation sequons contain glycan, estimates of only 12–14 kDa of TSHR glycan do not support a conclusion that all six sites are glycosylated (57). Only Asn302 has been determined directly to contain glycan (44). Mutagenesis studies on the TSHR have only provided limited information. Thus, four N-linked glycosylation motifs on the human TSHR (Asn residues 99, 177, 198, and 302) could be individually eliminated by mutagenesis without affecting trafficking of a functional receptor to the cell surface (75). Mutation of Asn77 and Asn113 completely abolished TSHR expression on the cell surface or generated a receptor with reduced affinity for TSH, respectively (75). Whether or not these effects are consequent to loss of glycan moieties or secondary to folding abnormalities in the polypeptide chain are unknown. However, the presence of more than 25 kDa of glycan suggests that most, if not all, potential sites contain glycan, at least in transfected cells. Surprisingly, mutation of the five potential N-linked glycosylation sites in the rat TSHR (Asn113 is absent) produced effects different from those observed with the human TSHR (41). Thus, mutation of only Asn198 reduced TSH binding affinity, apparently without alteration in the biological response to TSH stimulation.
The precise composition of the TSHR ectodomain glycan will be difficult to determine. Potential heterogeneity among glycans at different sites will require purification of polypeptide fragments containing the individual glycan moieties, a daunting task. In generic terms, however, it is clear that, as with most glycoproteins that reach the cell surface (29), the carbohydrate on the mature TSHR (both single chain and cleaved) has been transformed from the high-mannose variety in the endoplasmic reticulum to complex form. The possible role of TSHR ectodomain glycan in ligand and autoantibody binding will be discussed separately below.
E. TSHR dimerization
Dimerization of ligand-coupled receptors is a well recognized
phenomenon that may also apply to the TSHR. For example, at a
functional level, there is evidence that ‘blocking’ TSHR
autoantibodies may become stimulatory when linked by divalent anti-IgG
(76). It has also been reported that the recombinant TSHR ectodomain
generated in bacteria and in insect cells (77, 78) can form multimers
when detected by PAGE. Mammalian cell B subunits have also been
reported to form dimers (31). Further, high molecular weight complexes
consistent with receptor dimerization are typically detected on TSH
cross-linking studies. However, the sizes of these bands vary widely in
different experiments (22, 23, 30, 32, 79), and we are uncertain
whether these bands are physiological multimers or artifactual
aggregates. It is clear that the TSHR, even when fully glycosylated, is
a particularly ‘sticky’ protein that is prone to aggregate. In
particular, we confirm the observation (26) that heating TSHR to more
than 50 C before PAGE leads to marked aggregation. Such heating may
induce ‘caramelization’ of the very high TSHR glycan content.
|
III. Recombinant TSHR Expression |
|---|
|
TopAbstractI. IntroductionII. TSHR Ectodomain StructureIII. Recombinant TSHR ExpressionIV. TSH and Autoantibody...V. TSHR Autoantibody AssaysVI. Monoclonal Autoantibodies to...VII. T Cell EpitopesVIII. Molecular MimicryIX. Animal Models of...X. Ectodomain Mutations and...XI. Extrathyroidal...Note Added in ProofReferences |
|---|
5,000 receptors per thyrocyte (58), and instability during the
purification process have prevented receptor purification from this
source. Despite valiant efforts (73, 80, 81), thyroidal TSHRs have
never been purified sufficiently to be visualized as an unequivocal and
discrete electrophoretic band on direct protein staining. Recombinant
TSHR is, therefore, the only practical source for purification. Recombinant TSHRs expressed in nonthyroidal cells have been of value in the assay of TSHR autoantibodies (see below). However, full exploitation of this material for understanding the structure-function relationship of the TSHR, as well as for exploring the pathogenesis and possible immunotherapy of Graves’ disease, will require its high level expression and purification. A number of groups have devoted years of effort to this endeavor, admittedly only a means to an end and not, in itself, of direct value.
A. Approaches to TSHR expression
Two major decisions are required of investigators expressing the
recombinant TSHR.
1. The expression system to be used. Systems used to date have involved the transfer of TSHR cDNA in numerous vectors (plasmid and viral) into different types of cells including: (i) prokaryotic bacteria, (ii) eukaryotic insect cells, (iii) eukaryotic mammalian cells, such as CHO cells, mouse L fibroblasts, mouse SP2/0 myeloma cells, human embryonal kidney (HEK), and simian virus-40-transformed African green monkey kidney (COS) cells.
In addition, TSHR protein, or fragments thereof, can be generated as (i) short, synthetic peptides (typically 16–20 amino acids in length); (ii) cell-free translation of TSHR mRNA in an in vitro system.
2. Expression of the TSH holoreceptor, the TSHR ectodomain, or ectodomain fragments. The holoreceptor (743 amino acids after signal peptide deletion) includes the hydrophobic membrane-spanning segments. The smaller ectodomain (397 amino acids after signal peptide deletion) contains more hydrophilic amino acid residues. The decision on whether to express the holoreceptor or the ectodomain depends heavily on the intended use of the recombinant TSHR. Obviously, the membrane-spanning segments are necessary for studying TSHR signal transduction. On the other hand, if the goal is to study hormone or autoantibody binding independent of signal transduction, or to generate antisera to the TSHR, the ectodomain (or parts of the ectodomain) may suffice and may be advantageous. Expression of only the ectodomain carries the theoretical advantage of generating a more water-soluble product. In addition, the absence of a membrane-anchoring region offers the potential for generating a protein that is not retained in the membrane but is secreted into culture medium. Under these conditions, as has been accomplished with the ectodomain of TPO (82, 83), TSHR could theoretically be extracted from spent medium, harvested repeatedly from long-term mammalian cell cultures.
B. TSHR expression in prokaryotes
Much effort has been made to express the recombinant TSHR in
bacteria, probably because this system is technically the easiest to
apply and, at least for some proteins, a high level of expression can
be attained. Unfortunately, however, major difficulties have been
encountered, most likely explaining the lack of reports on TSH
holoreceptor expression in bacteria. Numerous groups have reported on
the prokaryotic expression of the more soluble TSHR ectodomain (27, 40, 77, 84, 85, 86, 87). In some systems, such as with the pMAL vector (Ref. 86 and
G. D. Chazenbalk and B. Rapoport, unpublished data) and with a
GST-expressing fusion vector (88), the TSHR ectodomain has been
generated in a (at least in part) soluble form, rather than as
insoluble inclusion bodies that require dissolving in high
concentrations of chaotropic agents such as urea or guanidine.
There is debate as to whether or not the nonglycosylated TSHR ectodomain of prokaryotic origin specifically binds TSH and/or TSHR autoantibodies. There are some reports that TSHR autoantibodies (77, 78, 88) and TSH (88) bind to this material, although the latter is "inefficient" (88). It is noteworthy that, on immunoblotting, 60% of patients’ sera containing high titers of TSHR autoantibodies recognize nonglycosylated, bacterial TSHR ectodomain (78). Others (85, 86), including ourselves (G. D. Chazenbalk and B. Rapoport, unpublished data) have not detected specific TSH or TSHR autoantibody binding to the TSHR ectodomain generated in bacteria.
C. TSHR expression as cell-free translates or as synthetic
peptides
As for TSHR expression in prokaryotes, cell-free translation of
TSHR mRNA using microsomal preparations (89) is technically
straightforward and generates products lacking carbohydrate moieties.
Modification of the system can generate glycosylated material, but not
with the quantity and quality of glycans present on glycoproteins
expressed on the surface of mammalian cells. TSHR holoreceptor (90) or
ectodomain (91, 92, 93) expression by cell-free translation has been
accomplished, however, with conflicting findings. Remarkably, TSH is
reported to bind to the holoreceptor (90), a finding that awaits
confirmation. TSHR autoantibodies either did (91, 92), or did not (93),
interact with material produced in similar systems. The role of
conformational integrity and glycan content in autoantibody recognition
of the TSHR ectodomain (discussed below), as well as other factors, may
explain these discrepant results.
Synthetic TSHR polypeptide synthesis is a technically demanding but routine procedure performed by commercial or specialized laboratories. The practical advantages of generating peptides is counterbalanced by their limited ability to reconstitute the binding sites on a native, conformational, and glycosylated protein. Nevertheless, if one or more peptides do interact specifically with a ligand or antibody, valuable information is immediately available because the sequence of the peptide is known. Data on TSH and auto-antibody interactions with the TSHR are, therefore, deferred to the section on epitopes and binding sites (Section IV).
D. TSHR expression in eukaryotic insect cells
Extremely high levels of proteins can be expressed in insect cells
infected with a baculovirus transfer vector modified to contain a gene
of interest (94). This positive feature, together with improved vectors
and the fact that proteins produced in these cells, unlike in
prokaryotic cells, are glycosylated, has made the baculovirus system a
major focus for TSHR expression. Attempts to express the active
TSH holoreceptor in insect cells failed (Refs. 95, 96, 97 and B. Rapoport,
unpublished observations). Only the TSHR ectodomain has been expressed
in this system. Initial reports (96, 97) on the generation of milligram
quantities of TSHR ectodomain using conventional baculovirus vectors
were optimistic, even when this material was largely insoluble, having
been expressed without a signal peptide (97). Thus, after
solubilization, the signal peptide-lacking TSHR ectodomain bound TSH
with moderately high affinity (10-9 M
Kd) (98) and was believed to be glycosylated (97). Most
important, solubilized insect cell TSHR ectodomain expressed with (96)
or without (99) a signal peptide was recognized on enzyme-linked
immunosorbent assay (ELISA) by autoantibodies in patients’ sera (96, 99). This material was, therefore, reported to provide a simple,
specific, and highly sensitive assay for TSHR autoantibodies (99).
Unfortunately, enthusiasm for TSHR ectodomain expression in insect cells using conventional vectors soon tempered. TSH binding to this material, even when generated with a signal peptide and when directly determined to contain 14 kDa of glycan, was not confirmed (96). Subsequently, it was determined that the signal peptide-lacking ectodomain, although previously reported to be recognized by autoantibodies (97), did not contain carbohydrate and did not interact with TSHR autoantibodies (100). Three modifications have so far been attempted to improve the quality of the TSHR ectodomain produced in insect cells.
1. Use of an earlier baculovirus promoter to enhance TSHR ectodomain glycosylation. The final stage after baculovirus infection of insect cells is host cell death with release of infective viral particles encapsulated in polyhedrin protein (94). In the few days before the host cell dies, the baculovirus usurps the host protein synthesis mechanism to make large quantities of polyhedrin protein. This diversion greatly reduces synthesis of posttranslational modification enzymes, such as glycosylases. To improve recombinant protein glycosylation, new baculovirus transfer vectors were introduced with the cDNA being driven by a promoter that functions at an earlier phase of the infective cycle than the classic polyhedrin promoter.
TSHR ectodomain generated using one of these vectors (pAcMP3) did,
indeed, increase the extent of TSHR glycosylation compared with that
generated with the conventional polyhedrin promoter (14 kDa
vs. 9 kDa of glycan) (79). However, the level of
TSHR expression with this system was very low, consistent with the
early stage of the viral cycle. Further, most of the TSHR ectodomain
remained within the cell in particulate form. Unfortunately, only trace
amounts of soluble TSHR ectodomain were secreted or were present in the
cytosol fraction of infected insect cells, too little to be of
practical value. Qualitatively, however, this material provided
important information. Thus, it completely neutralized TSHR
autoantibodies in patients’ sera but only marginally inhibited TSH
binding. These data suggested that TSHR glycosylation was important for
autoantibody recognition and supported the viewpoint (96, 101) that
autoantibodies bind better to the isolated TSHR ectodomain than does
TSH.
2. Denaturation of insoluble material followed by refolding. Another approach taken to overcome the deficiencies of the baculovirus system with respect to the TSHR expression was to accept the largely insoluble nature of the protein and its lesser degree of glycosylation, but to attempt refolding of the protein into its correct conformation (78, 98). As with refolding of prokaryotic TSHR ectodomain (77), this process generated electrophoretic bands interpreted to represent folded and unfolded monomers and tetramers. Remarkably, glycosylated insect cell TSHR material was recognized less well than nonglycosylated bacterial ectodomain by the same panel of sera with high titers of TSHR autoantibodies (78). Moreover, these data suggested that, for the majority of sera, the native, folded state was not an important factor in autoantibody binding (78).
Important and provocative conclusions were made from these studies (77, 78). First, TSHR conformation and glycosylation were not major factors in autoantibody recognition of the TSHR, and linear epitopes were clearly recognized by many TSHR autoantibodies. Second, the apparent polyclonality of autoantibody recognition of different forms of TSHR ectodomain (folded vs. unfolded; glycosylated vs. non-glycosylated) implied that T cells, not autoantibodies, are of primary importance in Graves’ disease. We suggest some caution in making these conclusions because of the very high concentration at which sera were studied (1:20), the extremely low titer in serum of TSHR autoantibodies (102, 103), and the known oligoclonality of TSHR autoantibodies in serum (104, 105, 106). In addition, as pointed out for TPO autoantibodies (107), qualitative recognition by a serum of a particular form of antigen (or epitope) does not indicate what proportion of the antibodies in the serum interact with this antigenic form (or epitope). Adsorption studies or quantitative competition with mAbs to different epitopes or antigenic forms are required to address this issue.
3. Replacement of the mammalian TSHR signal peptide with an insect cell signal peptide. Very recently, to overcome the handicap of protein expression without a signal peptide, or the potential deleterious effect of a mammalian signal peptide, the human (100) and mouse (108) TSHR ectodomains have been expressed in insect cells using an insect cell signal peptide (part of the baculovirus envelope glycoprotein 67). Although such a strategy frequently does not work for other proteins (109), the insect cell gp signal peptide enhanced TSHR ectodomain expression relative to use of the TSHR signal peptide (100) and produced a more heavily glycosylated protein (100, 108). However, as occurred with TSHR expression in mammalian cells (see below), the TSHR ectomain was still not secreted by insect cells and was largely insoluble, requiring solubilization and refolding (100, 108).
As observed previously with the more heavily glycosylated TSHR ectodomain generated in insect cells using an earlier promoter (79), as well as with the glycosylated ectodomain retained within mammalian cells (see below) (110), the new form of insect cell TSHR ectodomain neutralized Graves’ patients’ TSHR autoantibodies (100). However, the amount of TSHR required for autoantibody neutralization and the nature of the glycan on the neutralizing material were not determined and remain important unanswered questions with respect to the role of carbohydrate in autoantibody recognition (discussed below). In contrast to these neutralization studies with crude insect cell homogenates, purified mouse TSHR ectodomain produced using an insect cell signal peptide was recognized on ELISA by only a minority of Graves’ sera (108). All studies on more heavily glycosylated TSHR ectodomains produced in insect cells agree on one point. TSH binds very poorly (79, 108) or not at all (100) to this material. The specific and relatively high-affinity binding of TSH to nonglycosylated TSHR ectodomain (98), recently confirmed (100), remains an enigma.
E. TSHR expression in mammalian cells
Theoretically, as described above, expression of the TSHR
ectodomain may be sufficient for studying interaction with ligand or
autoantibodies, as opposed to studying the functional activity of the
TSHR. In practice, however, until recently, it was easier to study TSH
and autoantibody binding to the TSHR ectodomain, not as an isolated
fragment, but as part of the holoreceptor. For the purpose of
generating large amounts of recombinant TSHR in mammalian cells, stable
integration of cDNA into the genome is preferable to transient
transfection, although improvements in the latter, particularly the
application of new viral vectors (111, 112, 113), may alter this balance in
the future. This review will, therefore, focus on TSHR expression in
stably transfected mammalian cells.
1. Stable TSH holoreceptor expression.
a. Mammalian cells are most effective for generating
conformationally intact TSHR.
There has been no question since its
molecular cloning that the TSHR expressed in mammalian cells, unlike in
other systems (see above), is completely functional. Stable expression
of the human TSH holoreceptor in CHO mammalian cells has been reported
by many investigators (2, 95, 114, 115, 116, 117, 118). In addition to CHO cells,
stable TSHR expression has also been achieved in SP2/0 mouse myeloma
cells (116). The latter have the advantage of growing well in
suspension culture in a fermentor, permitting the propagation of large
numbers of cells. Cells stably expressing the TSHR transduce a signal
after TSH (2, 114) and TSHR autoantibody (34, 114, 119, 120)
stimulation. Intact cells (121) or membranes produced from these cells
(119, 122) can be used in a TSH binding inhibition assay for TSHR
autoantibodies. A typical level of TSHR expression in stably
transfected cell is 90,000/cell, as in the commonly used JPO9 line
(122). Unfortunately, although this level of expression is far higher
than estimated in thyroid tissue (2,000–5,000 receptors per cell)
(58), it is still too low to make TSHR purification practical.
b. Role of TSHR mRNA 5'- and 3'-untranslated regions.
Parenthetically, an interesting story emerged from studies
performed years later to explore the inferior sensitivity of the
original CHO cell line developed in our laboratory compared with cell
lines such as JPO9. We noted that whereas we had transfected the
full-length (4-kb) TSHR cDNA, other cell lines such as JP26 (114) and
JPO9 (122) had been transfected with only the coding region of the cDNA
(2.3 kb). Subsequent studies revealed that the 5'- and 3'-untranslated
regions of the TSHR cDNA reduced by nearly 10-fold the level of TSHR
expression in transfected CHO cells, from 150,000 to 16,000
receptors per cell (123). This number of TSHR, far lower than we
originally estimated (124), explained the lesser sensitivity of our
original cell line. This observation is also of potential
pathophysiological importance. Thus, the TSHR untranslated ends are
obviously present in vivo and may possibly explain the low
level of TSHR expression reported in thyroid tissue (58). It remains to
be seen whether mutations in the TSHR untranslated regions can
derepress this suppressive effect and lead to increased TSHR
expression. The TSHR is "noisy" in the absence of ligand (10, 63, 64, 65). Consequently, as reported in mice transgenic for the
ß2-adrenergic receptor (125), constitutive expression of enhanced
numbers of TSHR could be a cause of goiter, or even hyperthyroidism.
Such a mechanism could complement the already demonstrated effect of
activating mutations in the coding region of the TSHR that occur
without a change in TSHR number (8, 9).
c. TSHR overexpression.
Even with as many as 100,000 TSHRs per
cell, a level readily achieved in CHO and SP2/0 cells (see above),
purification of significant amounts of TSHR would be a formidable task.
For example, at this level of expression, about six confluent 10-cm
dishes of cells would yield only 1 µg of TSHR, assuming 100%
recovery. Clearly, overexpression of greater numbers of TSHR on each
cell would be desirable. Such amplification has recently been attained
using a dihydrofolate reductase minigene system. Thus, CHO cells
expressing approximately 2 million TSHRs have been generated (32).
Without such a high level of expression and even with the availability
of mouse mAbs to the TSHR (27, 55, 126, 127), the recombinant TSHR
ex
