(Received for publication, September 18, 1996, and in revised form, October 18, 1996)
From the Department of Pathology, Thomas Jefferson University School of Medicine, Philadelphia, Pennsylvania 19107 and the § Department of Cell Biology, Duke University Medical Center, Durham, North Carolina 27710
In order to study the membrane topology, processing, and oligomerization of inositol trisphosphate receptor (IP3R) isoforms, we have utilized RNA templates encoding putative transmembrane domains to program a cell-free translation system of rabbit reticulocyte lysates supplemented with canine pancreas microsomes. In the absence of microsomes, translation of the RNA templates encoding all the putative transmembrane domains present in the C-terminal segment of the type I (1TM) and type III (3TM) IP3R isoforms resulted in a 62- and 59-kDa polypeptide, respectively. In both cases, an additional band approximately 3 kDa larger was observed upon the addition of microsomes. Both bands in the translation doublet were integrated into microsomal membranes and were full-length translation products, as shown by sedimentation through a sucrose cushion and immunoprecipitation with C-terminal isoform-specific antibodies. With both isoforms, N-glycopeptidase F digestion indicates that the upper band in the doublet corresponds to a glycosylated translation product. A 17-kDa protected fragment was observed after proteinase-K digestion of 1TM translated in the presence of microsomes. The pattern and size of protected fragments was consistent with the current six-transmembrane domain model of IP3R topology. Cotranslation of both 1TM and 3TM RNA templates in the presence of microsomes followed by immunoprecipitation with isoform specific antibodies revealed coimmunoprecipitation of translation products. This was not observed when the isoforms were translated separately and then mixed, suggesting that heteroligomerization occurs cotranslationally. A construct encoding only the first putative transmembrane domain of the type I isoform was found to be sufficient for integration into membranes but was unable to oligomerize with either 1TM or 3TM. Cotranslation experiments using additional constructs indicate that the major structural determinant for homoligomerization lies between putative transmembrane domain 5 and the C terminus. A second oligomerization domain involved in stabilization of heteroligomers is present within the first four transmembrane domains.
A family of specific receptors on intracellular membranes mediates the Ca2+ mobilization induced by the second messenger molecule D-myo-inositol 1,4,5-trisphosphate (IP3R).1 Full-length sequences of three distinct isoforms have been published (1-6). All receptor isoforms are believed to have a common domain architecture. The ligand-binding domain is located in the N-terminal region (7, 8), and the Ca2+ channel domain is located at the C-terminal portion of the receptor, which contains several putative transmembrane helices. All IP3Rs are also believed to be tetrameric structures (9, 10). From the analysis of deletion mutants, it is apparent that the transmembrane domains, in addition to forming the ion channel, are also essential for receptor oligomerization (11). The large intervening sequences between the N-terminal ligand-binding domain and the C-terminal transmembrane regions have been assigned a regulatory role. This region contains phosphorylation sites for cAMP and cGMP-dependent protein kinase as well as binding sites for ATP and calmodulin (7, 12-15).
The exact number of transmembrane helices in the C-terminal portion of the receptor and the transmembrane topology of this molecule have been matters of controversy (1, 16). The current six-transmembrane helix consensus model of the receptor arises in part from revised hydropapathy analysis and from the perception that this intracellular ligand-gated calcium channel may be structurally related to a superfamily of plasma membrane ion channels that includes the cGMP-gated cation channel and voltage-gated Na+,K+ and Ca2+ channels (6, 16, 17). All of these channels have six transmembrane domains, which are repeated four times within a single subunit or are present only once in an individual subunit of a tetrameric channel. An additional structural feature present in this superfamily of ion-channels is a hydrophobic region between transmembrane domains 5 and 6 (the pore domain), which is thought to be partially embedded in the membrane and to play an important role in controlling the ion selectivity of the channel (reviewed by Pongs (18)). A similar role has been ascribed to a hydrophobic region present between putative transmembrane domains 5 and 6 in the IP3R sequence (16, 17). Experimental support for the proposed topology of the IP3R is limited. A result that is consistent with the current model is the finding that the loop between transmembrane domains 5 and 6 is N-glycosylated at two sites (19), placing this region of the receptor within the intraluminal compartment of the endoplasmic reticulum. Experimental evidence to verify the exact number and sequence boundaries of transmembrane domains, cytosolic, and intraluminal loops as predicted from computer algorithms remains to be obtained.
Most cells contain more than one IP3R isoform (8, 20-22). Recently, it has been shown that IP3R isoforms have the capacity to form heterotetramers (23-26). These structures appear to assemble in the membrane during biosynthesis, since they are not observed when lysates containing different isoforms are mixed (23). An early step in the biosynthesis of IP3Rs must involve the insertion of the nascent IP3R polypeptide into the endoplasmic reticulum membrane and their subsequent folding and assembly into homo- and heteroligomers. However, very little information is available regarding these processes. In order to initiate such studies we have used an in vitro transcription/translation programmed with DNA constructs encoding the transmembrane domains of the IP3R with canine pancreas microsomes as acceptor membranes. The results indicate that newly synthesized C-terminal domains of type I and type III IP3R isoforms can be cotranslationally inserted into microsomal membranes and N-glycosylated. Protease protection assays have been used to analyze the topology of the inserted protein. Deletion constructs of the type I isoform have been used to show that the first two transmembrane domains are sufficient for membrane insertion but not for oligomerization. In addition, the data show that simultaneous translation of type I and type III IP3R mRNA results in the formation of heteroligomers in vitro. The structural requirements for homo- and heteroligomerization have been investigated.
Materials
T7 RNA polymerase, Taq polymerase, DNA ligase, RNasin ribonuclease inhibitor, rabbit reticulocyte lysate, and wheat germ lysate were purchased from Promega (Madison, WI). Pfu polymerase was obtained from Stratagene (La Jolla, CA). N-glycosidase F and the long range PCR kit were from Boehringer Mannheim. Biotin-labeled in vitro translation kit, Amplify, and ECL immunoblotting kits were obtained from Amersham Corp. Stabilized acrylamide solution (Protogel) for the preparation of SDS gels was obtained from National Diagnostics (Atlanta, GA). Tran35S-label was obtained from ICN Radiochemicals (Irvine, CA). Restriction endonucleases were purchased from Promega, Boehringer Mannheim, or New England Biolabs (Beverly, MA). The pCITE T-vector kit and the sequencing primers U-19, CITE, and T3 were obtained from Novagen (Madison, WI). The cDNA encoding the rat type I IP3R isoforms were kindly provided by Dr. Thomas Südhof (University of Texas Southwestern Medical Center, Dallas) and Dr. Greg Mignery (Loyola School of Medicine, Chicago). The cDNA encoding the rat type III IP3R isoform was kindly supplied by Dr. Graeme Bell (Univ. Chicago). Bovine preprolactin was transcribed and translated from the plasmid pGEMBP1 (27).
PCR Primers
The following primers were synthesized by the Nucleic Acid
Facility of the Jefferson Cancer Institute (restriction sites are underlined): 1TM-F, 5-CGTCTAGAGTCGACATGAACTGGCAGAAGAAA-3
;
1TM-Fb, 5
-GCG
AACTGGCAGAAGAAA-3
; 1TM-R,
5
-TAT
GGTACCTTAGGCTGGCTGCTGTGG-3
; 1TM1-R,
5
-GC
TCTAGTGTTCCTCCTCT-3
; 1TM1,2-R,
5
-GC
TTGGGCAGCGCAATGAC-3
; 1TM
1-F,
5
-AAAGGAGTGAGAGGAGGA-3
; 1TM
1-R, 5
-TCGGGCGCACCAGTACAA-3
; 1TM5,6-F, 5
-ACC
GATTTAGTGTACAGAGAGGAG-3
; 1TM4-R,
5
-CTGTAA
GTCTCCTCTCTGTACACTAA-3
; 1TM1-4,stop,
5
-GCG
TTACTCCTCTCTGTACACTAA-3
; 1TM,stop,
5
-GCG
TTATTGTTTCTGCTTCCTTTG-3
; 3TM-F,
5
-CGCCATGGATGGAGCAGATCGTGTTC-3
; 3TM-R,
5
-TATGAATTCTCAGCGGCTCATGCAGTT-3; 3TMs-F,
5
-CGAACA
GAACAGATGACGGAGCAG-3
; 3TMs-R,
5
-AGTCCAACGCGTGTCGATGATCACCCCAAA-3
.
Subcloning the Transmembrane Domain and Constructs into the CITE Vector
All constructs described below were subjected to automated DNA
sequencing to confirm that the products were in frame with respect to
the translation initiation site and to verify the sequence of the
insert. The amino acid boundaries of the constructs are detailed in
Fig. 1.
Transmembrane Domain of the Type I IP3R and Mutant Constructs
1TMThe pI7 clone encoding 6 kilobase pairs of the
C-terminal portion of rat type I IP3R (2) was used as a
template for PCR amplification of the putative transmembrane domains.
1TM-F and 1TM-R were used as primers to amplify nucleotides encoding
from methionine 2253 to the stop codon. The 1.6-kilobase pair PCR
product was ethanol-precipitated and purified from a 1% agarose gel
using a Quaiex kit (Qiagen, Chatsworth, CA). The 1TM PCR product was then directly ligated into the pCITE T-vector. Following transformation of Escherichia coli (DH5), positive clones were
identified by colony PCR using a combination of vector (CITE) and
insert-specific (1TM-R) primer.
The construct encoding the first putative transmembrane domain was amplified using 1TM-F and 1TM1-R as primers and cloned into the pCITE T-vector as described above for 1TM. The 1TM1,2 construct, which encodes the first two putative transmembrane domains, was amplified using 1TM-F and 1TM2-R as primers and cloned into the pCITE T-vector as described above for 1TM.
1TM1,tag and 1TM1,2,tagIn these constructs the antibody epitope tag for the type I IP3R was inserted at the C terminus to permit immunoprecipitation. Nucleotides encoding amino acids 2708-2749 were excised from the pCITE-1TM plasmid by digestion with BstEII (cutting at base pair 8450) and NotI (vector site). This "tag" fragment was gel-purified. The 1TM1-R and 1TM1,2-R primers were designed to contain a BstEII restriction site. To make the tagged constructs, the tag fragment was ligated into the BstEII/NotI-digested pCITE-1TM1 and 1TM1,2 plasmids.
1TM1-4,tagA construct encoding the first four putative transmembrane domains was amplified from the 1TM plasmid using 1TM-Fb and TM4-R. These primers included sites for NdeI and BstEII, respectively. The PCR fragment was digested with these enzymes and ligated into the CITE-1TM plasmid that had been gel-purified after being cut with NdeI and BstEII.
1TMInverse PCR (28) was used to obtain a mutant
construct in which the first putative transmembrane domain
(corresponding to nucleotides 7140-7218) was deleted from pCITE 1TM.
PCR amplification of the entire pCITE 1TM plasmid with the exception of
the region of the first transmembrane domain was carried out using
1TM1F and 1TM
1R primers and a long range PCR kit obtained from
Boehringer Mannheim. The PCR product was blunt-ended with
Pfu polymerase (29), gel-purified, and ligated.
A forward primer containing an NdeI restriction site (1TM5,6-F) and 1TM-R, which contains an EcoRI restriction site, was used to amplify the segment of the pCITE 1TM plasmid, which encodes the fifth putative transmembrane domain through to the C terminus. The PCR fragment was cut with NdeI and EcoRI, gel-purified, and ligated into NdeI/EcoRI-digested pCITE 1TM plasmid.
1TM1-4,stop, 1TM5,6,stop, and 1TM,stopTo study homoligomerization with 1TM, we designed constructs from which the antibody epitope had been removed and replaced with a stop codon. The alternative approach of using unique restriction sites within the coding sequence to generate truncated transcripts produced results that were difficult to interpret, since multiple translation products were observed. 1TM1-4,stop was amplified from 1TM1-4,tag DNA template using 1TM-Fb as forward primer and 1TM1-4,stop as reverse primer. 1TM5,6,stop was amplified from 1TM5,6 DNA template using 1TM5,6-F as forward primer and 1TM,stop as reverse primer. 1TM,stop was amplified from 1TM DNA template using 1TM-Fb as forward primer and 1TM,stop as reverse primer. All PCR inserts were cut with NdeI and EcoRI, gel-purified, and ligated into NdeI/EcoRI-digested pCITE 1TM plasmid.
Transmembrane Domain of the Type III IP3R and Mutant Constructs
3TMThe cDNA encoding the rat type III IP3R
was used as a template for the PCR amplification of the putative
transmembrane domains. 3TM-F and 3TM-R were used to amplify nucleotides
encoding from methionine 2132 to the stop codon. The PCR product was
made with Pfu polymerase, and overhanging 3-adenosines were
added by incubation of the PCR product with Taq polymerase
and dATP (30). The product was ethanol-precipitated and directly
ligated into the pCITE-T vector. Following transformation of E. coli (DH5
), positive clones were identified by colony PCR using
a combination of vector (CITE) and insert-specific (3TM-R) primers.
A construct in which amino acids 2519-2645 were
deleted from the C-terminal tail of the type III isoform was
constructed by inverse PCR using 3TMs-F and 3TMs-R primers. Both
primers were designed to encode an MluI restriction site.
The PCR was carried out with the long range PCR kit as described above
for 1TM1, and the product was ligated after digestion with
MluI. The inclusion of the MluI site results in a
conservative change in the amino acid at position 2646 from lysine
to arginine.
Cell-free Transcription
Plasmid DNA (5 µg) was linearized with EcoRI.
Capped transcripts were synthesized with T7 RNA polymerase using an
mRNA synthesis kit (Ambion, Austin, TX). Transcribed templates were
purified by phenol/chloroform extraction and ethanol precipitation. The sample was resuspended in 20 µl of diethylpyrocarbonate water and
stored at 80 °C.
Cell-free Translation and Translocation Assays
Routinely, cell-free translations were carried out for 1 h at 30 °C in a final volume of 25 µl and contained 10 µl of rabbit reticulocyte lysate, 0.5 µl of RNasin, 0.5 µl of 1 mM amino acids (minus methionine), 20 µCi of Tran35S-label, 1.5 µl of RNA template (1-3 µg), and 6 µl of buffer A (pH 7.2) containing 110 mM KOAc, 2 mM Mg(OAc)2, and 20 mM KHepes. Where appropriate, the translation reactions contained 3 µl of nuclease-treated canine pancreatic microsomes prepared as described (31). SDS-PAGE gels containing [35S]-labeled translation products were processed for fluorography with 1 M sodium salicylate (32) or Amplify and exposed to Kodak XAR-5 film. In some experiments the translation reactions were carried out using biotinylated lysyl-tRNA using a cell-free translation kit manufactured by Amersham.
For protease protection assays, aliquots of the translation reactions were diluted with 3 volumes of a buffer containing 120 mM KCl, 20 mM Tris-HCl (pH 7.4), 1 mM EDTA, and 100 µg/ml proteinase K. The reactions were incubated on ice for 1 h, and 3 mM phenylmethylsulfonyl fluoride was added for a further 10 min. The sample was then diluted 5-fold into 0.1 M Na2CO3 buffer (pH 11.0) and spun through a sucrose cushion (1.5 M sucrose, 10 mM TES, pH 7.2) at 60,000 × g for 20 min on a Beckman TLA-100 rotor. The pellet was resuspended in SDS-PAGE sample buffer and analyzed as described above.
For deglycosylation of the translation products, microsomes were spun
through a sucrose cushion and resuspended in 0.3% SDS, 100 mM -mercaptoethanol, 50 mM sodium phosphate
(pH 8.6), and 1 mM EDTA. The sample was boiled for 3 min
and supplemented with a final concentration of 1.5% octylglucoside.
Two aliquots were incubated overnight at 37 °C in the presence or
absence of 1 unit of N-glycosidase F.
Immunoprecipitation and Immunoblotting
The translation mixture was diluted into 500 µl of a solubilization buffer containing 150 mM NaCl, 50 mM Tris-HCl (pH 7.8), 1% Triton X-100 (w/v), 1 mM EDTA, 0.5 mM phenylmethylsulfonyl fluoride, and 5 µg/ml each of aprotinin, soybean trypsin inhibitor, and leupeptin. The samples were precleared with 20 µl of Pansorbin (Calbiochem). After removal of Pansorbin by centrifugation (10,000 × g for 10 min), IP3R isoform-specific antibody and 50 µl of a 20% (v/v) slurry of protein A-Sepharose was added, and the sample was incubated for 2 h at 4 °C. Immune complexes, precipitated with protein A-Sepharose, were washed three times in solubilization buffer and analyzed by SDS-PAGE. In some experiments the polypeptides in the gel were transferred to nitrocellulose, which was autoradiographed and then immunoblotted with isoform-specific antibodies to locate the receptor. The polyclonal type I IP3R antibody used in this study has been described previously (33). The type III polyclonal antibody was raised against the C-terminal peptide corresponding to residues 2657-2670. The peptide was synthesized with an additional N-terminal cysteine, which was used for conjugation to keyhole limpet hemocyanin (34). Antibody was raised in rabbits by Cocalico Biologicals (Reamstown, PA). The antibody was affinity-purified using the peptide coupled to Ultralink Iodoacetyl beads as described by the manufacturer (Pierce).
PCR was used to amplify the C-terminal segment
of the type I and type III IP3R isoforms containing the
putative transmembrane domains. Fig. 1 shows the
sequence boundaries of the DNA constructs used in the present study.
The PCR fragments were cloned into a plasmid behind a modified
5-untranslated viral sequence (CITE) that is known to enhance
translation efficiency in the rabbit reticulocyte system (35, 36). Fig.
2A shows the results of translating the RNA
template encoding the 1TM construct using an amino acid mixture labeled
with [35S]methionine and [35S]cysteine. In
the absence of microsomes, the 1TM construct yielded a single labeled
polypeptide having a molecular mass of 62.3 ± 2.3 kDa (mean ± S.E.; n = 3). The predicted molecular mass of the
translation product is 58 kDa, taking into account 13 amino acids
derived from the vector and PCR primer (MATTHMDSSRVE) and the 498 amino
acids of the construct itself. Control experiments showed that
translation of preprolactin mRNA in the presence of microsomes led
to the appearance of a smaller product, corresponding to cleavage of
preprolactin by signal peptidase, verifying the efficient processing
activity of the microsomal membranes (Fig. 2A, lanes
1 and 2). Translation of 1TM mRNA in the presence
of microsomes resulted in the appearance of both the 62-kDa band and an
additional translation product that was approximately 3 kDa larger.
Both bands in the doublet corresponded to full-length translation
products, since they were both immunoprecipitated by an antibody
recognizing the 18 amino acids at the C terminus (Fig. 2A,
lanes 5 and 6).
In order to assess if one or both translation products are integrated into the microsomal membrane, the translation reactions were diluted with buffer, and the microsomal membranes were recovered by centrifugation through a sucrose cushion (Fig. 2B). Both translational products were recovered in the pellet fraction in these experiments. This was also the case when the translation products were first diluted into a buffer containing Na2CO3 (pH 11), incubation conditions that strip peripherally attached proteins from membranes (37). From these data we conclude that the doublet of translation products was integrated into the microsomal membrane. This occurs cotranslationally, since insertion into membranes was not observed when microsomes were added after translation had been terminated with cycloheximide (data not shown).
Cell-free Translation of the Transmembrane Domains of the Type III IP3RThe type III transmembrane domain construct
amplified by PCR corresponds to amino acids 2132-2670 of the rat
sequence (5). Fig. 3A shows that the RNA
template encoding this construct translated, in the absence of
microsomes, as a single labeled polypeptide having a molecular mass of
59.2 ± 0.6 kDa (mean ± S.E.; n = 3). The
predicted molecular mass of the translation product is 63 kDa, taking
into account 10 amino acids derived from the vector and PCR primer
(MATTHMDSPW) and the 539 amino acids of the construct itself. As
observed with 1TM mRNA, translation of 3TM mRNA in the presence
of microsomes resulted in the appearance of a doublet of translation
products consisting of both the 59-kDa band and an additional
translation product that was approximately 3 kDa larger. Both bands in
the doublet were specifically immunoprecipitated by type III antibody,
which did not immunoprecipitate the 1TM translation products (Fig.
3A, lanes 4-6). Almost all of the translation product doublet was recovered in the pellet fraction when the microsomes were stripped with Na2CO3 buffer and
then centrifuged through a sucrose cushion (Fig. 3B). These
results indicate that the 3TM translation products are also integrated
into the microsomal membrane.
Enzymatic Deglycosylation of the Translation Products
It is
known that the type I IP3R is N-glycosylated,
and two potential sites of glycosylation have been identified in the
intraluminal loop between putative transmembrane domains 5 and 6 (19).
Although there is no direct experimental evidence that the type III
IP3R is a glycoprotein, one of the glycosylation sites in
the type I IP3R is conserved in the intraluminal loop of
the type III isoform. Therefore, a possible explanation for the
appearance of a higher molecular weight band in the presence of
microsomes is that it represents the glycosylated translation product.
Fig. 4 shows the results of experiments in which the 1TM
and 3TM translation products obtained in the presence of microsomes
were deglycosylated with N-glycopeptidase F. Only a single
translation product was observed after deglycosylation of the
translation products of both isoforms (lanes 2,
4, and 6). We conclude that the higher molecular
weight band seen with both IP3R isoforms in the presence of
microsomes represents a glycosylated translation product.
Protease Protection Assays
The current topological model of
the IP3R predicts the presence of three intraluminal loops
(6, 16, 17). Two of these loops are relatively small. The third loop,
present between putative transmembrane domains 5 and 6, is much larger
and encompasses the N-glycosylation sites and the proposed
pore-forming domain. Protease cleavage of the translation product
inserted into microsomal membranes would be expected to result in the
appearance of protected fragments whose size can be predicted from the
topological model. In initial trials, we found that labeling the
translation product with biotinylated lysine rather than
[35S]methionine gave a lower background and a clearer
signal in protease protection assays. Fig. 5A
shows that proteinase K cleavage of the biotinylated 1TM translation
product produces several protected bands (compare lanes 1 and 2). All of the protected polypeptides become
accessible to proteinase K in the presence of detergent (Fig.
5A, lane 3). The most prominent of these
protease-protected bands (Fig. 5A, lane 2,
arrow) has a molecular mass of 16.8 ± 0.5 kDa
(n = 3). This molecular mass is in agreement with the value of 16.9 kDa calculated from the primary sequence for putative transmembrane domains 5 and 6 and the intervening large intraluminal loop. A doublet of protected bands above the 19 kDa marker was reproducibly observed in these experiments, although their intensity was variable (e.g. Fig. 5, compare A and
B, brackets). Enzymatic deglycosylation of the
proteinase K-digested 1TM translation product resulted in a decrease in
the doublet of bands above 19 kDa and an increase in the amount of
16-kDa polypeptide (Fig. 5B, lane 3). This
suggests that the doublet and the 16-kDa polypeptide correspond,
respectively, to the glycosylated and nonglycosylated forms of the
large intraluminal loop of the receptor. The presence of a doublet of
glycosylated bands may indicate heterogeneity in the occupation of the
two available consensus glycosylation sites or heterogeneity in the
carbohydrate composition of glycan chains. The two protected bands of
the lowest molecular weight seen in Fig. 5A may correspond
to the two smaller intraluminal loops expected from the predicted
topology, but they were not reproducibly observed in every experiment
(e.g. Fig. 5, compare A and B).
Role of the First Transmembrane Domain
The IP3R
is an example of a polytopic transmembrane protein that does not
contain a cleavable signal sequence. The first transmembrane domains of
such proteins are believed to act as topogenic signals for membrane
insertion. The sequential arrangement of transmembrane domains that act
as "signal anchor" and "stop-transfer" sequences are believed
to determine the final topology of the protein (38). We have examined
the role of the first transmembrane domain in membrane insertion and
receptor oligomerization by making a construct that contains only the
first putative transmembrane domain (1TM1) or a construct in which only
the first putative transmembrane domain has been selectively deleted
(1TM1). In order to facilitate immunoprecipitation of the 1TM1
translation product, the construct was engineered to contain the 42 amino acids present at the C terminus of the type I IP3R
(1TM1,tag), which includes the epitope recognized by the type
I-specific antibody. Fig. 6A shows the translation of 1TM1,tag RNA in the presence or absence of microsomes. A
translation product with an apparent molecular mass of 13.2 kDa was
observed (predicted mass is 12.1 kDa). As expected, the size of the
product did not change upon the addition of microsomal membranes,
although the translation reaction was more efficient under these
conditions. The 1TM1,tag translation product was found in the pellet
fraction after treatment of the microsomes with Na2CO3 (pH 11.0) and centrifugation through a
sucrose cushion (Fig. 6B, lane 1), indicating
that the presence of the first transmembrane domain was sufficient to
permit insertion into the microsomal membrane. Translation of the
1TM
1 construct in which the first putative transmembrane domain has
been selectively deleted is shown in Fig. 6C. This product
was also integrated into the microsomal membranes, as shown by recovery
in the pellet fraction after Na2CO3 treatment
and centrifugation through a sucrose cushion (Fig. 6C, lane 3). Unlike the full-length 1TM construct, which
translates as a doublet, the 1TM
1 translation product appears as a
prominent single band together with diffuse bands of lower mobility
(Fig. 6C, lane 1). Digestion of the 1TM
1
translation product with N-glycopeptidase F did not shift
the mobility of the prominent band but did remove the diffuse bands
(data not shown). We interpret the results in Fig. 6 to indicate that
in the absence of the first transmembrane domain, additional topogenic
sequences in putative transmembrane domains 2-6 can direct membrane
insertion. The five transmembrane domains of 1TM
1 could adopt many
possible orientations in the membrane with variable degrees of
glycosylation, which could account for the diffuse 1TM
1 translation
products. The prominent N-glycopeptidase F-insensitive
1TM
1 band may reflect an aberrant orientation that exposes the loop
containing the glycosylation sites to the extravesicular space.
Heteroligomerization of the 1TM and 3TM Translation Products
Sucrose density gradients were used to investigate the
oligomerization status of in vitro translated products. Fig.
7 shows the sedimentation profile of the
35S-labeled 1TM translation product after lysis of
microsomal membranes in Triton X-100 or SDS. Analysis of Triton X-100
lysates revealed a peak of radioactivity at the top of the gradient and
a second broader peak located in the middle of the gradient. The broad second peak was absent when the microsomal membranes were lysed in SDS.
This suggests that the transmembrane domain of the type I
IP3R can homoligomerize when translated in
vitro.
Experiments were then performed to determine if cotranslation of both
1TM and 3TM would lead to heteroligomerization. The coimmunoprecipitation of two different translation products possessing heterologous immunological tags has previously been used as an experimental method to study heteroligomerization of other ion channel
proteins in a cell-free translation system (36, 39-42). The results
using this approach are shown in Fig. 8. When 1TM and
3TM RNA templates were translated separately in the presence of
microsomes and then immunoprecipitated, it was clear that each of the
antibodies was highly specific for its respective isoform (Fig. 8,
lanes 1-4). When both templates were translated together, each Ab immunoprecipitated its respective doublet of translation products and also coimmunoprecipitated a proportion of the other isoform. Since 1TM and 3TM translation products overlap in their migration, three 35S-labeled bands are observed when both
templates are cotranslated and immunoprecipitated with either type I or
type III specific Abs (Fig. 8, lanes 5 and 6).
When the templates were translated separately and then mixed before
immunoprecipitation, each Ab only immunoprecipitated its cognate
antigen (Fig. 8, lanes 7 and 8). From these data
we conclude that coimmunoprecipitation of both isoforms reflects the
cotranslational assembly of heteroligomers in the microsomal membranes
and that this process does not occur when the lysates containing
individual isoforms are mixed. This conclusion is consistent with
previous studies showing that heteroligomers were not formed when
different cell lysates containing individual IP3R isoforms
were mixed (23, 25). Although complicated by the presence of the heavy
chain of IgG, the occurrence of coimmunoprecipitation in the
translation reactions was confirmed by analysis of the immunoprecipitates using immunoblotting with isoform-specific Abs (data
not shown).
The translation products obtained from 1TM and 3TM are closely spaced
on SDS-PAGE gels. In order to accentuate differences in the molecular
weight of the two isoforms, we prepared the 3TM "small" (3TMs)
construct in which 127 amino acids were deleted from the C-terminal
tail, leaving the antibody epitope at the C terminus in place (Fig. 1).
Like 3TM, the 3TMs construct was also translated as a doublet but ran
on SDS-PAGE at a distinctly smaller molecular weight than the 1TM
translation product (Fig. 9). When 1TM and 3TMs RNA
templates were translated together and microsomal lysates were
immunoprecipitated with type I IP3R Ab, a clear doublet
corresponding to coprecipitating 3TMs translation products could be
observed (Fig. 9, lane 5). Again, it could be shown that
this was not due to cross-reactivity of the Ab (Fig. 9, lane
3), and it was absent when individual templates were translated separately and lysates were mixed after the translation was completed (Fig. 9, lane 7). Attempts to observe 1TM translation
product coimmunoprecipitated by type III Ab from cotranslated 1TM and 3TMs templates were unsuccessful (Fig. 9, lane 8), even with
more prolonged exposure of the autoradiographs (data not shown). The reason for this lack of reciprocal immunoprecipitation is not clear at
present. One possibility is that truncation of the C-terminal tail
hinders access of the type III Ab to 3TMs when 3TMs is present as a
component of heteroligomers with 1TM.
Structural Features of 1TM Required for Heteroligomerization
In an attempt to localize the regions of the
IP3R that may be important for heteroligomerization, we
prepared constructs encoding limited regions of 1TM and tested each for
their ability to associate with 3TM in coprecipitation assays. The
results obtained with a construct encoding only the region from
transmembrane segment 5 to the C terminus (1TM5,6; see Fig. 1) is shown
in Fig. 10. 1TM5,6 translates in the presence of
microsomes as a lower band of approximately 36 kDa, above which lies a
series of closely spaced bands that merge together at the
autoradiograph exposures shown (Fig. 4, lane 5, and Fig.
10A, lane 3) but resolve into three distinct
bands at lower exposures (data not shown). All of the upper bands are removed upon digestion with N-glycopeptidase F (Fig. 4,
lane 6), and they must therefore correspond to
differentially glycosylated 1TM5,6 translation products. When 1TM5,6 is
cotranslated with 3TM and microsomal lysates are immunoprecipitated
with type I specific Ab, an additional band is seen that corresponds to
the glycosylated 3TM translation product (Fig. 10A,
lane 1). Similarly, immunoprecipitation of the lysates with
type III-specific Ab shows the presence of coprecipitating 1TM5,6 (Fig.
10A, lane 2). The control lanes, in which 1TM5,6
and 3TM were translated separately and then mixed, show no evidence of
coprecipitation (Fig. 10A, lanes 3 and
4). The same experiment was also carried out with a 1TM
construct encoding the first four putative transmembrane domains
attached to the type I epitope tag (1TM1-4,tag; see Fig. 1). This
construct also showed evidence of coprecipitation with 3TM (Fig.
10B, compare lanes 1 and 3). The
results of cotranslating 3TM with smaller segments of the receptor
encoded by 1TM1,tag and 1TM1,2,tag are shown in Fig. 10C.
Although both of these constructs were effectively translated when
combined with 3TM, neither product had the ability to associate with
3TM as judged by the absence of coprecipitation when tested with either
type I- (Fig. 10C, lanes 4 and 5) or
type III-specific Abs (Fig. 10D, lanes 4 and
5). These results suggest that the region of the receptor
containing the first four transmembrane domains and the last two
transmembrane domains can each independently interact with 3TM.
However, the first two transmembrane domains do not have the ability to
heteroligomerize.
Structural Features Required for Homoligomerization
In order
to determine if the same structural requirements apply for
homoligomerization of receptors we modified the coprecipitation assay
so that 1TM template was combined with truncated 1TM constructs from
which the antibody tag had been deleted and replaced with a stop codon
(1TM,stop, 1TM5,6,stop, and 1TM1-4,stop; see Fig. 1). In initial
experiments we first verified that homoligomerization could be
demonstrated by this procedure by combining 1TM with 1TM,stop. In this
case both templates when translated are expected to yield products of
similar molecular weights. However, it is clear that the slightly
smaller doublet of bands corresponding to 1TM,stop coprecipitates with
1TM when the two templates are cotranslated (Fig. 11,
lane 2) but not when the templates are translated separately
and then mixed (Fig. 11, lane 1). Similar evidence for coprecipitation was observed when 1TM5,6,stop was combined with 1TM
(Fig. 11, lane 4). However, in contrast to the results
observed with heteroligomerization assays, there was no indication of
coprecipitation of 1TM1-4,stop (Fig. 11, lane 6). This was
not due to impaired translation of template, since both translation
products were present in samples derived directly from the translation
mixture (Fig. 11, lane 7). Similar experiments combining
1TM1 and 1TM are shown in Fig. 12. Immunoprecipitation
of lysates containing both 1TM and 1TM1 templates with type I
IP3R-specific antibody resulted in the appearance of the
1TM translation product only (Fig. 12, lane 6). Directly
quenched translation assay mixtures showed that both templates
had indeed been translated (Fig. 12, lane 3). The same
experiment carried out with a combination of 1TM1,2 and 1TM also showed
no coprecipitation (data not shown). The failure of the 1TM-1, 1TM-1,2,
and 1TM1-4,stop translation products to coimmunoprecipitate with 1TM
suggests that the sequence between transmembrane domain 5 and the
C terminus is required to facilitate homoligomerization.
In the present study we have utilized a rabbit reticulocyte lysate system to study membrane integration, processing, and assembly of IP3Rs. Only RNA templates encoding C-terminal segments were used in these experiments, since the putative transmembrane domains of the receptor are confined to this region (11). Our studies show that the translation products of the 1TM and 3TM RNA templates are cotranslationally integrated into canine pancreas microsomal membranes and appear as doublets consisting of a glycosylated and a nonglycosylated band. 1TM possesses two consensus N-linked glycosylation sites (asparagines 2475 and 2503) in the large intraluminal loop between putative transmembrane helices 5 and 6 (17). Whether both sites in 1TM acquire sugar residues during in vitro translation is not clear. Although only a single glycosylated band is observed in 1TM, the small mobility differences between single and double glycosylated products may be difficult to distinguish on our gel systems. However, in the case of the smaller 1TM5,6 construct it is evident that heterogeneous glycosylated products can be formed. 3TM has only one consensus glycosylation site (asparagine 2404), which is clearly utilized during in vitro translation. The type III IP3R cDNA has been expressed in COS cells, but the glycosylation status of the recombinant protein has not been examined (4-6). Blotting of type III Ab immunoprecipitates derived from WB rat liver epithelial cell lysates with biotinylated concanavalin A indicates that this protein is glycosylated in the cell.2
An exhaustive analysis of the membrane topology of all the constructs used for cell-free translation has not been attempted in the present study. Protease protection analysis was carried out only on the 1TM construct. The data obtained were consistent with the six-transmembrane domain model proposed by others (16, 17). A minimal IP3R construct containing only the first putative transmembrane segment was capable of membrane insertion. When this domain was selectively deleted from 1TM, the remaining five transmembrane domains retained the ability to integrate into membranes, but the translation product was not glycosylated. Studies with other polytopic membrane proteins have shown that individual transmembrane segments can have signal sequence activity, stop-transfer activity, or both, depending on their sequence context (43, 44). Our data show that the first transmembrane span of the IP3R has signal sequence activity and that, when deleted, downstream domain(s) can substitute for this activity. Of the remaining domains, at least transmembrane segment 5 must have signal sequence activity when expressed as part of the 1TM5,6 construct, since this translation product is also inserted and glycosylated.
Mutagenesis studies have suggested that the transmembrane segments play a primary role in the assembly of IP3R tetramers (11). In common with what is known regarding the architecture of voltage-gated ion channels (45, 46), it is believed that a functional inositol trisphosphate-gated Ca2+ channel is tetrameric and that transmembrane domains from each monomer contribute to the formation of a central ion-conducting pore. The precise domains that line the pore of IP3R and the types of interactions holding the tetramer together are unknown. In the case of the Shaker K+ channel, ionic interactions have been proposed to occur between negatively charged residues in transmembrane segments 2 and 3 and positively charged residues in transmembrane segment 4 (47, 48). Analysis of the six putative transmembrane segments of all three of the IP3R isoforms indicates the presence of only one highly conserved positively charged residue in putative transmembrane segment 3 (lysine). By this criterion, electrostatic interactions between the membrane embedded segments of the receptor are unlikely to play a major role in maintenance of the tetrameric structure. Hydrophobic interactions, interactions between charged residues located on cytosolic or lumenal loops, and intersubunit disulfide bridges (49) may all contribute to stabilization of the oligomer.
Heteroligomerization between closely related subunits has been described for many ion channel proteins, including the amilioride-sensitive Na+ channel (50), cyclic nucleotide-gated cation channel (51, 52), G-protein-gated inward rectifying K+ channel (53, 54), and voltage-gated K+ channels (36, 55, 56). Several studies have documented the presence of heteroligomers of IP3R isoforms in cell lysates (23-26). In the present series of experiments, we have provided evidence for heteroligomerization between 1TM and 3TM constructs in an in vitro translation system. The ability of a single specific antibody to coimmunoprecipitate both translation products was observed only when both isoforms were cotranslated and not when they were translated separately and then mixed. This indicates that association between isoforms is not an artifact of the detergent solubilization and immunoprecipitation procedures. In general, it was easier to observe 3TM coimmunoprecipitated by type I-specific Ab than to observe 1TM coimmunoprecipitated by type III-specific Ab. The lack of reciprocal coimmunoprecipitation was particularly evident when using truncated 1TM constructs. The results were not dependent on the particular type of antibody used, since different type I- and type III-specific antibodies gave similar results in these experiments (data not shown). Other factors that may contribute to this result are the relative translation efficiency of each RNA template when added in combination, differential occlusion of C-terminal antibody epitopes in homo- and heteroligomers, and the possibility that there is a preferred stoichiometry for heteroligomers favoring type I subunits. Because of these factors, the results obtained from coprecipitation assays are necessarily qualitative, and it is difficult to estimate the exact fraction of translation products that exist as heteroligomers. However, the data indicate that heteroligomers are a small fraction of the total translation product in the in vitro system. Analyses of cell lysates have produced estimates of heteroligomerization that range from involvement of only a minor fraction in WB rat liver epithelial cells (24) to the entire fraction of type I IP3Rs being heteroligomerized in RINm5F cells (26). The regulatory mechanisms that favor (or prevent) the formation of heteroligomers during receptor biosynthesis in the endoplasmic reticulum of different cell types remains to be defined.
An important advantage of the cell-free translation system is that it allows the structural requirements for oligomerization to be investigated using truncated RNA templates. Table I shows a summary of the data obtained from coprecipitation experiments using truncated 1TM constructs to study homo- and heteroligomerization. Constructs containing 1TM1 and 1TM1,2 proved incapable of homoligomerization or heteroligomerization. Thus, the first transmembrane domain, while necessary for membrane insertion, does not play a role in oligomerization. Our experiments indicate that 1TM5,6 is the smallest truncated 1TM construct that associates with 1TM or 3TM. We therefore conclude that this region contains the primary oligomerization domain of IP3Rs. The 1TM5,6 construct encompasses the proposed pore domain of the channel and the glycosylation sites. The putative transmembrane segments 5 and 6 are highly conserved among the different isoforms and also share a high degree of homology to corresponding domains in the ryanodine receptor (9). Inward rectifying voltage-dependent K+ channels have a similar topology consisting of two transmembrane segments with an intervening pore domain. These channels have also been shown to form homo- and heterotetramers (53, 54).
|
A surprising observation in these experiments was the finding that the 1TM construct containing the first four transmembrane segments could associate with 3TM but was inactive in homoligomerization assays. Since both 1TM5,6 and 1TM1-4,tag can independently associate with 3TM, this suggests that a secondary oligomerization domain is located in the TM1-4,tag construct. This domain would function to specifically stabilize heteroligomeric IP3R subunit interactions. Multiple sites of interaction have also been proposed to be involved in the self-assembly of tetrameric voltage-dependent K+ channels (36, 39, 48, 56, 57). Further mapping of the oligomerization domains and analysis of transmembrane topology should provide valuable information on the structure and assembly of this important class of ion channel proteins.