| Human Molecular Genetics | Pages |
©1999 Oxford University Press |
A direct interaction between EXT proteins and glycosyltransferases is defective in hereditary multiple exostoses
Introduction
Results
Yeast two-hybrid screens
Specificity of yeast two-hybrid interactions
Biological relevance of interactions
In vitro binding assays
Mammalian two-hybrid assays
Discussion
Materials And Methods
Yeast two-hybrid vectors
Yeast two-hybrid analysis
In vitro binding assays
TRAP1 and GalNAc-T5 extension and analysis
Mammalian two-hybrid vectors and analysis
Acknowledgements
References
A direct interaction between EXT proteins and glycosyltransferases is defective in hereditary multiple exostoses
Received May 12, 1999; Revised and Accepted August 16, 1999
Hereditary multiple exostoses (HME) is an autosomal dominant condition in which bony outgrowths occur from the juxtaepiphyseal regions of the long bones. In a few percent of cases these exostoses undergo malignant transformation to chondrosarcomas. HME results from mutations in one of two homologous genes, EXT1 and EXT2. These are members of a new gene family that is conserved from Caenorhabditis elegans to higher vertebrates. In humans this family comprises five genes which are most conserved at their C-termini, but they do not contain any discernible functional motifs and their function(s) is unclear. Indirect evidence suggests that EXT proteins are involved in glycosaminoglycan synthesis, act as tumor suppressors and affect hedgehog signaling. One recent study has also reported that these proteins co-purify with glycosyltransferase (GlcA and GlcNAc transferase) activity and on that basis it has been postulated that they are themselves glycosyl-transferases. We performed two-hybrid screens with a fragment of EXT2 from the region that is most highly conserved in the gene family and identified two interacting proteins: the tumor necrosis factor type 1 asso-ciated protein and a novel UDP-GalNAc:poly-peptide N-acetylgalactosaminyltransferase. Significantly, both these interactions were abrogated by a disease-causing EXT mutation, indicating that they are important in the etiology of HME. The EXT2-GalNAc-T5 interaction provides the first direct physical link between EXT proteins and known components of glycosamino-glycan synthesis.
INTRODUCTION
Hereditary multiple exostoses (HME) is an autosomal dominant skeletal disorder characterized by aberrant bone formation (exostoses) generally originating from the juxtaepiphyseal regions of the long bones (1). With the exception of the presence of exostoses, which can be surgically removed, most affected individuals are relatively healthy. However, 2-5% of HME patients suffer a much more severe consequence: the transformation of a benign exostosis to a malignant chondrosarcoma or osteosarcoma (2,3).
HME is a genetically heterogeneous disorder with at least three separate genes (EXT1-3) causing a clinically indistinguishable phenotype (4). EXT1 is located on chromosome 8q24.1 (5), EXT2 is located on chromosome 11p11-p13 (6,7) and EXT3 has been linked to chromosome 19p (4). EXT1 and EXT2 were isolated by positional cloning and represent two members of a new gene family (8,9). The EXT3 gene has not yet been isolated and genetic linkage to this locus has only been detected in a few pedigrees to date (4,10). Three additional members of the EXT gene family, designated EXTL1, EXTL2 and EXTL3, were also recently identified (11-13). These related genes have been designated as EXT-like because there is no evidence that defects in these genes result in any cases of HME.
The EXT and EXTL genes are highly conserved, in particular in their C-terminal regions, but their function(s) is unclear. However, at least EXT1 and EXT2 appear to act as tumor suppressors since loss of heterozygosity is frequently observed in sporadic and exostoses-derived chondrosarcomas (14,15). This has led to a classic tumor suppressor model for EXT1 and EXT2 in which an inherited haploinsufficiency for one allele results in the development of multiple exostoses and a subsequent somatic loss of the second copy of the gene moves the cell toward malignant transformation.
Indirect evidence indicates that EXT1 and EXT2 also participate in the synthesis of glycosoaminoglycans (GAGs). Overexpression of an EXT1 cDNA restored the ability of herpes simplex virus type 1 (HSV-1) to infect sog9 cells (16). These cells are unable to synthesize cell surface heparan sulfate GAGs and are thus normally resistant to HSV-1 infection. Although overexpression of an EXT1 cDNA in sog9 cells led to a partial restoration of GAG synthesis, it is not yet clear whether the primary defect in sog9 cells is in fact an EXT1 mutation or some other change in GAG synthesis that can be complemented by EXT1 overexpression. Nevertheless, these observations provide the first, more detailed, clues to EXT function.
The implied involvement of EXTs in GAG synthesis was taken one step further with the identification of EXT2 in a partially purified activity from bovine serum that retained D-glucuronyl (GlcA) and N-acetyl-D-glucosaminyl (GlcNAc) transferase activities (17). In the same study, overexpression of EXT2 in COS-7 cells led to a 4- to 10-fold increase in the measured levels of GlcA and GlcNAc transferase activities. On the basis of these observations it has been postulated that EXT proteins are themselves glycosyltransferases that directly mediate GAG synthesis (16,17). In a more recent study these observations have been supported by the finding that EXTL2 appears to encode a GlcNAc-T1 and GalNAc transferase activities (18). However, confusingly, none of the EXT proteins bears any sequence homology to the previously characterized family of five glycosyltransferases (19,20) which retain a recognizable degree of amino acid motif conservation (see below and Figure 1).
Figure 1. Structure and sequence analysis of rat and human GalNAc-T5. The rat GalNAc-T5 gene can be subdivided into four domains as indicated by the vertical bars and the descriptions on the schematic representation. The approximate location of the fragment of human GalNAc-T5 that interacts with the C-terminal end of EXT2 is indicated below the rat sequence. An amino acid sequence alignment of this fragment and the corresponding region of the rat GalNAc-T5 protein is shown below. The consensus sequence is derived from residues that are identical in both the rat and human GalNAc-T5 genes. These residues have also been shaded in the primary alignment. Residues within the highly conserved catalytic region are indicated by a large box.
The final indirect clue to the function of EXT proteins was the recent identification of a Drosophila homolog for EXT1, designated tout-velu (ttv) (21). The Ttv protein appears to be required for the diffusion of hedgehog (Hh) signals in Drosophila, perhaps by affecting the transduction of Hh signal across the surface of adjacent cells. Although the exact mechanisms underlying these effects are not understood, these studies suggest that one model for the effects of EXTs on bone growth in humans is that EXT1 mediates the rate of chondrocyte differentiation, and thus bone morphogenesis, by interfacing with Indian hedgehog (Ih) signaling (22).
In this report we describe the use of the yeast two-hybrid system (23) to isolate two proteins that specifically interact with EXT2: the tumor necrosis factor type 1 associated protein (TRAP1) (24,25) and a new isoform of human UDP-GalNAc:polypeptide N-acetylgalactosaminyltransferase (GalNAc-T5) (26). TRAP1 is a distantly related member of the hsp90 gene family and specifically interacts with the polypeptide encoded by the C-terminal ends of both EXT1 and EXT2, but fails to interact with the same domains from EXTL1 and EXTL3. Significantly, TRAP1 also fails to interact with EXT1 and EXT2 C-terminal polypeptides containing a biologically relevant histidine deletion. This histidine deletion appears to be the sole cause of HME in one previously studied family (27). In contrast to TRAP1, GalNAc-T5 interacts in vivo with the C-terminal end of only EXT2 and fails to interact with the corresponding regions from EXT1, EXTL1 or EXTL3. The interaction between EXT2, GalNAc-T5 and TRAP1 provides the first direct physical link between EXT proteins and GAG synthesis and suggests that these interactions are relevant to HME.
RESULTS
Yeast two-hybrid screens
The EXT gene family encodes proteins with variable N-terminal regions and highly conserved C-termini (shown diagrammatically in Fig. 2a and by partial alignment in Fig. 2b). The average amino acid sequence identity of EXT1, EXT2, EXTL1 and EXTL3 increases from 29 to 45% in the last 130 residues and this region, which has tentatively been designated the EXT domain (28), contains 35 invariant residues (Fig. 2b). We employed yeast two-hybrid screens (23) to identify proteins that interact with this conserved domain. The regions encoding the 130 C-terminal amino acids of EXT1 and EXT2 were cloned adjacent to the GAL4 DNA binding domain (BD) and the expression of fusion proteins was confirmed by western blot analysis with a GAL4 BD monoclonal antibody (Fig. 2c). Figure 2c also shows the expression of an EXT1-H627Del fusion protein which contains a deletion of one histidine codon at amino acid 627. This deletion has been previously shown to co-segregate with (and presumably cause) HME in one EXT1-linked family (27) and represents a biologically significant mutation within the C-terminal region we screened.
Figure 2. Schematic representation, nucleotide sequence alignment and western blot analysis of EXT genes. (a) EXT1, EXT2, EXTL1 and EXTL3 can be subdivided into variable N-terminal regions and an ~130 amino acid conserved C-terminal region, designated the EXT domain (shaded). The approximate location of the HME-causing deleted histidine residue within EXT1-H627del and the corresponding histidine in EXT2-H601del are indicated by solid bars within the EXT domain. (b) Amino acid sequence alignment of the EXT domains of EXT1, EXT2, EXTL1 and EXTL3. Shaded residues in the EXT sequences indicate amino acids conserved between two or more of the proteins. This information was used to derive the consensus sequence, which is also comprised of amino acids that are conserved between two or more of the EXT proteins. Shaded residues in the consensus indicate amino acids that are identical in all of the EXT proteins shown. The His residue deleted in EXT1-H627del is indicated by an asterisk. (c) Results of western blot analysis on constructs containing the last 130 amino acids of EXT1, EXT2 or EXT1-H627del cloned into the yeast two-hybrid vector pAS2-1 (lanes 1-3). The control plasmids pVA3-1, containing the murine p53 gene, and pAS2-1 were analyzed in lanes 4 and 5. Immunoblotting was performed with a monoclonal antibody to the GAL4 BD. The fusion proteins migrated with an apparent molecular weight of 37 kDa, with 22 kDa corresponding to the GAL4 BD and 15 kDa corresponding to the 3[prime]-end of either EXT1, EXT1-H627del or EXT2. The pVA3-1 control yielded a fusion protein of the expected molecular weight of 70 kDa. Fragments <25 kDa were not analyzed on this gel, thus the 22 kDa product from pAS2-1 is not visible.
A yeast strain expressing the EXT2 fusion protein was transformed with human keratinocyte and lymphocyte cDNA libraries fused downstream of the GAL4 transcriptional activation domain (AD). A total of 6 × 106 transformants was screened and this led to the isolation of 29 clones that co-activated both the HIS3 and lacZ reporter genes. The AD plasmids were recovered from these isolates, DNA sequenced and compared with each other and with GenBank. This revealed three discrete groups of proteins: two of these were TRAP1, 16 corresponded to the human GalNAc-T5 protein and 11 were false positives (see Materials and Methods).
Specificity of yeast two-hybrid interactions
The specificity of the EXT2-TRAP1 and EXT2-GalNAc-T5 interactions were confirmed in the series of control transformations summarized in Table 1. Control constructs were tested for fusion protein production by western blotting to ensure that the correct products were being expressed (data not shown). In all cases these control transformations failed to activate HIS3 or lacZ expression, suggesting that the EXT2-TRAP1 and EXT2-GalNAc-T5 interactions were indeed specific. In addition to these standard controls, the TRAP1-AD and GalNAc-T5-AD constructs also failed to interact with the N-terminal 275 amino acids of either EXT1 or EXT2 cloned as BD fusion proteins (data not shown), indicating that the interactions are with the more conserved C-termini of the proteins. We also tested the ability of the EXT1 and EXT2 C-terminal fragments to homo- or heterodimerize by cloning them into the AD vector pACT2 and co-transforming them in various BD-AD combinations. Based on this limited assay, the last 130 residues of EXT1 or EXT2 failed to dimerize (data not shown).
Table 1. Summary of yeast two-hybrid transformations
| Binding domain pAS2-1 | Activation domain pACT2 | [beta]-gal activation |
| p53 | SV40 | + |
| EXT1-3[prime] | - | - |
| EXT1-3[prime] | SV40 | - |
| EXT1-3[prime] | TRAP1 | + |
| EXT1-3[prime] | GalNAc-T5 | - |
| EXT1-3[prime] H627Del | - | - |
| EXT1-3[prime] H627Del | SV40 | - |
| EXT1-3[prime] H627Del | TRAP1 | - |
| EXT1-3[prime] H627Del | GalNAc-T5 | - |
| EXT2-3[prime] | - | - |
| EXT2-3[prime] | SV40 | - |
| EXT2-3[prime] | TRAP1 | + |
| EXT2-3[prime] | GalNAc-T5 | + |
| EXT2-3[prime] H601Del | - | - |
| EXT2-3[prime] H601Del | SV40 | - |
| EXT2-3[prime] H601Del | TRAP1 | - |
| EXT2-3[prime] H601Del | GalNAc-T5 | +/- |
| EXTL1-3[prime] | - | - |
| EXTL1-3[prime] | SV40 | - |
| EXTL1-3[prime] | TRAP1 | - |
| EXTL1-3[prime] | GalNAc-T5 | - |
| EXTL3-3[prime] | - | - |
| EXTL3-3[prime] | SV40 | - |
| EXTL3-3[prime] | TRAP1 | - |
| EXTL3-3[prime] | GalNAc-T5 | - |
| - | TRAP1 | - |
| LAMIN | TRAP1 | - |
| p53 | TRAP1 | - |
| - | GalNAc-T5 | - |
| LAMIN | GalNAc-T5 | - |
| p53 | GalNAc-T5 | - |
We next tested whether TRAP1 and GalNAc-T5 interacted with the same C-terminal regions derived from other EXT genes, including EXT1, EXTL1 and EXTL3 (each separately cloned into pAS2-1). TRAP1 interacted with the C-terminal fragments of both EXT1 and EXT2 (Table 1), whereas GalNAc-T5 interacted only with the C-terminal fragment of EXT2. Although the average sequence identity of the EXT domain in the five known EXT proteins is 45%, both TRAP1 and GalNAc-T5 failed to interact with the C-termini of either EXTL1 or EXTL3. Western blot analysis (data not shown) revealed that these fusion proteins were indeed being expressed, indicating that a simple failure in expression of the constructs did not account for these negative results.
Biological relevance of interactions
We were next interested in determining the potential biological relevance of these interactions for HME. Most of the inherited missense mutations in EXT1 and EXT2 lie in the 5[prime]-half of the gene (29) which, as noted above, does not appear to interact with the two proteins we identified in these screens. However, in one family with multiple exostoses a single amino acid deletion (H627Del) in EXT1 is the only detected alteration (27). With the exception of EXTL2, this histidine is conserved in all the other members of the EXT gene family. We therefore tested the ability of the deleted protein to interact with either TRAP1 or GalNAc-T5. For these experiments we also constructed and tested a similar deletion in EXT2 (EXT2-H601Del) that contains a deletion of the corresponding histidine in EXT2 (marked by an asterisk in Fig. 2b). It should be noted that this particular EXT2 deletion has not yet been observed in HME families and its exact biological significance is unknown. The two histidine deletion constructs and constructs encoding C-terminal fragments from EXT1, EXT2, EXTL1 and EXTL3 were co-transformed along with either TRAP1 or GalNAc-T5. The resulting cell lines were quantitatively assayed for [beta]-galactosidase ([beta]-gal) activity (Fig. 3). When TRAP1 was tested in combination with either EXT1 or EXT2 it activated the transcription of lacZ (40-60 [beta]-gal units). However, the TRAP1 and EXT1-H627Del or EXT2-H601Del combinations produced only background levels of [beta]-gal activity, indicating that, since it failed to interact with the relevant mutation, TRAP1 may indeed have biological significance in HME. A similar negative result was obtained when TRAP1 was tested in combination with either EXTL1 or EXTL3. Neither of these EXT-like genes has been implicated in HME and the apparent specificity of TRAP1 for EXT1 and EXT2 (which, when altered, do cause HME) again suggests that TRAP1 may be relevant to the disease.
Figure 3. Quantitation of yeast two-hybrid interactions. The yeast strain Y190 was co-transformed with fragments of either GalNAc-T5 (a) or TRAP1 (b) that were recovered from yeast two-hybrid screens and with various experimental or control plasmids. GalNAc-T5 was recovered and expressed as a fusion protein with the GAL4 AD in the plasmid pACT2. TRAP1 was recovered and expressed as a fusion protein with the GAL4 AD in the plasmid pACT. The control plasmids used for these experiments were pAS2-1 (empty), pVA3-1 (murine p53), pTD1-1 (SV40 large T-antigen) and pLAM5[prime]-1 (human lamin C). All of the experimental plasmids were C-terminal fragments of EXT proteins fused to the GAL4 DB of the vector pAS2-1. With the exception of the first bars, which are the p53 and SV40 positive controls, the remaining samples in the left and right panels were co-transformed with the appropriate control or experimental plasmids, indicated below the chart, and the GalNAc-T5 or TRAP1 plasmids, respectively. Three isolates from each transformation were assayed for [beta]-gal activity using the substrate CPRG.
The results obtained with GalNAc-T5 were not all as definitive as those obtained with TRAP1. Together, the two wild-type constructs (EXT2 plus GalNAc-T5) activated the transcription of lacZ to the same extent as EXT2-TRAP1. EXT1, EXTL1 and EXTL3 again failed to interact with GalNAc-T5. However, the EXT2-H601Del-GalNAc-T5 combination only reduced activation of lacZ to about half the level of wild-type EXT2 (Fig. 3a). This is in contrast to the effect that this same deletion had on TRAP1 interactions, in which it failed entirely to activate lacZ expression. Thus, it would appear that the histidine deletion only partially abrogates the EXT2-GalNAc-T5 interaction but completely abrogates the EXT2-TRAP1 interaction. As noted above, the EXT2 histidine deletion mutation is not known to cause HME. However, the fact that it has the same effect on TRAP1 interactions as the biologically relevant EXT1 histidine deletion argues that the conserved histidine marks the location of a biologically important conserved domain in both proteins. The failure of the EXT2 histidine deletion to completely abrogate the GalNAc-T5 interaction may indicate that, at least in the family that carries the EXT1-H627Del mutation, GalNAc-T interactions are less relevant to HME than is the TRAP1 interaction.
In vitro binding assays
The specificity of the EXT2-GalNAc-T5 interaction was examined in vitro. EXT1 and EXT2 GST fusion proteins were bound to glutathione beads and then mixed with 35S-radiolabeled GalNAc-T5 protein (produced by coupled in vitro transcription and translation). The specificities of the interactions were measured by SDS-PAGE and autoradiography to detect labeled proteins that had bound to the GST constructs. An empty GST construct that retained only the short GST epitope, but no EXT sequences, failed to pull down any detectable labeled proteins, and this served as one negative control (Fig. 4, track 5). The other negative control in this experiment consisted of radiolabeled luciferase protein that was mixed into all of the pull-downs. Figure 4, track 1 shows the input labeled luciferase and track 2 the input labeled GalNAc-T5. As expected, the GST-EXT fusion proteins failed to pull down labeled luciferase. However, surprisingly in light of the yeast two-hybrid data, both EXT1 and EXT2 pulled down GalNAc-T5. This observation held true over a wide range of binding and washing conditions, including binding and washing at >37°C. However, at higher temperatures the amounts of EXT1 pulled down appeared to decrease, whereas the amounts of EXT2 remained very similar to those shown in Figure 4 (data not shown). This may reflect some in vitro preference for EXT2 binding by GalNAc-T5, but we have not as yet derived pull-down conditions that will absolutely discriminate between EXT1 and EXT2 binding. It is clear that in vitro both EXT1 and EXT2 are capable of binding the novel glycosyltransferase GalNAc-T5. However, in the much more stringent in vivo situation (in the yeast two-hybrid system) only EXT2 interacts at a detectable level. This will be discussed further below.
Figure 4. In vitro binding analysis of the EXT-GalNAc-T5 interactions. Lanes 1 and 2 contain aliquots (2 µl) of the input [35S]methionine-labeled luciferase and GalNAc-T5 proteins, respectively. Two proteins were visible in the GalNAc-T5 in vitro transcription and translation reaction (lane 2) due to the presence of a second in-frame methionine. These proteins all corresponded to their correct predicted molecular weights. Approximately 3 µg of GST fusion proteins containing the C-terminal conserved regions of EXT1 or EXT2 and GST only was combined with equal volumes (15 µl) of the in vitro transcribed and translated GalNAc-T5 and luciferase proteins. Following incubation and washing, the samples were boiled in loading buffer and analyzed by SDS-PAGE. As expected, the GST-EXT fusion proteins and GST alone failed to pull down labeled luciferase (lanes 3-5). Whereas GST alone failed to interact with any detectable proteins, the C-terminal ends of both EXT1 and EXT2 pulled down the GalNAc-T5 protein.
Mammalian two-hybrid assays
When similar types of pull-down experiment were conducted using in vitro labeled TRAP1 we observed very little specific in vitro interaction with EXT1 and EXT2, even under low stringency binding and washing conditions (data not shown). This may reflect a failure of the TRAP1 protein to correctly fold in vitro. On the other hand, this observation might indicate that the `interaction' is indirect or is merely an artifact of the yeast two-hybrid system. In order to resolve this we decided to employ a more stringent in vivo test of specific interactions: the mammalian two-hybrid system. This system is based on the activation of a firefly luciferase reporter construct (30). Induction occurs as a result of specific interaction between a GAL4-EXT1 and VP16-TRAP1 transcription complex. The last 130 amino acids of EXT1 and EXT1-H627del were fused to the GAL4 DB in the plasmid pM1. TRAP1 was cloned so as to fuse the transactivation domain of the herpes simplex virus VP16 protein upstream of TRAP1. An interaction between the GAL4-EXT1 and VP16-TRAP1 fusion proteins would then result in the transcriptional activation of the reporter plasmid G5E1bLUC, which contains the TATA element from the adenovirus E1b gene and five GAL4-responsive UASG sites upstream of the firefly luciferase gene. Pairwise combinations of the plasmids were transiently co-transfected into embryonal kidney 293 cells and assayed for luciferase activity. Transfection of the individual constructs with empty pM1 or pVP-FLAG5 plasmids resulted in only background induction of luciferase activity (Fig. 5), whereas transient co-transfection of the GAL4-EXT1 and VP16-TRAP1 fusion proteins resulted in a 3- to 4-fold increase in luciferase activity. The EXT1-H627Del construct failed to interact with TRAP1 and induced only basal levels of luciferase activity. These data confirm the yeast two-hybrid results, suggesting that EXT1 and TRAP1 specifically interact in mammalian cells and that the interaction is ablated by a disease-associated mutation (H627Del) of EXT1.
Figure 5. Mammalian two-hybrid results. Pairwise combinations of the plasmids pM1, pM1-EXT1 (EXT1-3[prime]), pM1-EXT1 (H627del), pVP-FLAG5 or pVP-FLAG5-TRAP1 (TRAP1) were transiently co-transfected into embryonal kidney 293 cells and assayed for luciferase activity. pM1 encoded the GAL4 DB whereas pVP-FLAG5 encoded the transactivation domain of herpes simplex virus VP16. Protein interactions led to the functional reconstitution of GAL4-dependent transcriptional activation, which was monitored with the reporter plasmid G5E1bLUC. The positive control for these experiments consisted of the previously reported BRCA1-BARD1 interaction (40; data not shown). The vertical axis indicates the plasmid combinations and the horizontal axis shows the averaged luciferase activity from the transfections (n = 2).
DISCUSSION
The detection of a physical interaction between EXT2 and a specific glycosyltransferase is in agreement with previous inferences about EXT function. Our observations explain the restoration of GAG synthesis to sog9 cells(16) and provide one explanation for the derivation of glycosyltransferase activity when EXT2 protein is partially purified (17). Our observation of a very strong in vitro interaction between EXT2 and GalNAc-T5 suggests that the two proteins may be difficult to purify away from each other. We have shown here that GalNAc-T5 will even bind to EXT1 in the highly artificial conditions of an in vitro pull-down experiment, but fails to interact with EXT1 in vivo. This is not surprising, since the primary amino acid sequences of EXT1 and EXT2 are so conserved in the EXT domain. Does this in vitro observation suggest that the specificity for EXT2 seen in the two-hybrid experiments is, in fact, due to some failure in the EXT1 constructs and that GalNAc-T5 actually interacts with EXT2 and EXT1 in vivo? This type of two-hybrid false negative appears unlikely, because our EXT1 yeast two-hybrid experiments included a good positive control: the specificity of TRAP1 for the same EXT1 proteins expressed from the same constructs. It appears more likely that the in vivo folding of GalNAc-T5 and/or EXT1 increases the specificity of the interaction in the in vivo situation. The pull-down experiments nevertheless point to an inherent affinity between the EXT domain and GalNAc-T proteins at the primary sequence level and suggest that another member of the GalNAc-T family may be partnered with EXT1. Nevertheless, our observations do not explain all of the activities ascribed to EXT2 (17) and EXTL2 (18). Thus, GalNAc-T5 appears to use only polypeptides as acceptors, whereas a true heparan sulfate polymerase would require oligonucleotide acceptors (31). Likewise, GalNAc-T5 alone does not appear to use activated GlcNAc or GlcA as monosaccharide donors, as would a true heparan sulfate polymerase (31). Thus, it remains possible that EXT proteins either comprise one part of a larger multi-enzyme GAG synthesis complex or that GalNAc-T5 may in fact be an enzyme that post-translationally modifies EXT2. In this model GalNAc-T5 could function as either an activase or an inhibitor of heparan sulfate synthesis through its action on EXT2.
GalNAc-T5 is the latest addition to the family of UDP-GalNAc:polypeptide N-acetylgalactosaminyltransferases that transfer the N-acetylgalactosamine (GalNAc) from UDP-GalNAc to the hydroxyl group of either serine or threonine residues. Overall, members of the GalNAc family can be subdivided into four conserved and recognizable domains: N-terminal, hydrophobic, stem and catalytic. Sequence analysis of the fragment we recovered from our screens revealed a 318 amino acid fragment that was 81% identical to a recently identified rat GalNAc-T5 protein (26; Fig. 1). This sequence was also similar to a human Unigene cluster (Hs.55968). We derived the DNA sequence from one of these ESTs (IMAGE 1663513), which extended the human GalNAc-T5 sequence to 1878 bp. The 605 residue protein that is predicted from this sequence is 83% identical to the rat GalNAc-T5 sequence and the sequence identity rises to 95% in the catalytic domain. As shown in Figure 1a, the fragment of GalNAc-T5 that interacts with the C-terminal end of EXT2 overlaps with both the stem and catalytic domains of the rat GalNAc-T5 protein and this serves to at least partially localize the EXT terminal interacting region to within these two domains.
This study raises the possibility that one function of EXT proteins is to regulate the initiation and specificity of GAG synthesis through interactions with separate members of the GalNAc-T family of enzymes. With the exception of EXTL1, the EXT genes are ubiquitiously expressed (11), whereas the GalNAc-T genes show distinct patterns of tissue-specific expression (20,32). Interestingly, our preliminary analysis of human GalNAc-T expression indicates that at least one member of the family is transcribed in exostoses (unpublished data). Perhaps these specific interactions and the tissue specificity of GalNAc-T proteins are the reasons why the EXT genes do not appear to complement each other (9,33). In this regard, it will be important to determine whether any of the other four EXT proteins specifically interacts with any of the other four known GalNAc transferases (32,34,35). At first glance this is an attractive model, but it leaves several facts unexplained. The first is that GalNAc transferase activity has been convincingly localized to the Golgi (36), whereas EXTs are believed (from studies of epitope-tagged artificial constructs) to be localized to the endoplasmic reticulum (16). Are EXT proteins translocated to the Golgi, and could TRAP1 be the mediator of this movement? The second conundrum is that neither TRAP1 nor GalNAc-T5 appears to bind to the N-terminal regions of EXT proteins within which the great majority of HME-causing missense mutations have been mapped. Does this indicate that additional, and potentially more important, proteins specifically interact at the N-temini of EXT proteins or does it reflect the possibility that inappropriate folding of the N-terminal fragments has destroyed their ability to bind TRAP1 or GalNAc-T5? The third confusing fact that does not fit perfectly with a model in which the EXT-GalNAc-T interaction is central to EXT function is that the GalNAc-T5 interaction with EXT2 is only partially abrogated by the EXT2-H601Del protein. Admittedly, the direct biological relevance of this single amino acid deletion in EXT2 is unknown, but this observation is in stark contrast to the situation with TRAP1, in which either of the EXT2 or EXT1 histidine deletions completely abrogates TRAP1 binding. The lack of TRAP1 interaction with the biologically significant EXT1 mutant protein appears to point to TRAP1 as a significant player in the biology of HME. Again, one potential model that would accommodate this observation is that TRAP1 mediates folding or translocation events that are critical to EXT1 and EXT2 function.
We chose to use the C-terminal EXT domain in this study on the supposition that the sequence conservation within this region would reveal common functional interactions for all EXT genes. Given the conserved nature of this EXT domain, it is particularly interesting that the TRAP1 interaction appears to only be specific to the two EXT genes that are known to cause HME (EXT1 and EXT2) and that TRAP1 does not interact with EXTL1 or EXTL3 (EXTL2 was not tested in these experiments). A comparison of the amino acid sequences of these four proteins (Fig. 2b) reveals some interesting differences. For example, two Lys residues (at positions 27 and 64 in Fig. 2b) are conserved between EXT1 and EXT2 but substituted for either Arg or Gln in EXTL1 and EXTL3. These residues, and the previously mentioned His deletion, will provide future targets for further elucidating the specificity and function of TRAP1 binding.
TRAP1 was first identified from yeast two-hybrid screens as a protein that appeared to interact with the intracellular domain of the type 1 receptor for tumor necrosis factor (TNFR-1IC) (25) and, in a separate study, with retinoblastoma protein (Rb) (24). Under heat shock conditions, TRAP1 appeared to be shuttled from a fairly ubiquitous cytoplasmic distribution to the nucleus and form a complex with Rb. It has also been reported that TRAP1 is capable of refolding denatured Rb protein (24).TRAP1 contains a unique LxCxE motif which partially mediates the binding of several proteins to Rb (35). Mutating this sequence to LxMxE appeared to reduce the binding of TRAP1 to Rb (24). However, we introduced this same mutation into TRAP1 and observed no effect on the binding of TRAP1 to EXT2 (data not shown).
We have shown that TRAP1 only interacts with the HME-causing genes EXT1/EXT2 and fails to interact with EXT1-H627Del, EXT2-H601Del and EXTL1/EXTL3. Based on these data, TRAP1 appears to have a more complex role than as a general molecular chaperone for all of the members of the EXT gene family. This raises the question of just how related TRAP1 is to the hsp family of proteins. In the previous studies that identified TRAP1 it was classified as a heat shock protein of the hsp90 family. In addition, as mentioned above, there is some limited evidence that it may have chaperonin activity (24). Although the published sequence of TRAP1 contains a long open reading frame that spans >90% of its message, the first methionine is ~400 bp from the immediate 5[prime]-end of the clone (24,25). To resolve this discrepancy, we isolated and DNA sequenced additional TRAP1 cDNA clones from a human testis cDNA library. This added 13 bp to the TRAP1 sequence, revealing a new in-frame methionine. Initiation of protein synthesis from this site would result in a 704 amino acid protein with a predicted molecular weight of 80 kDa, i.e. much closer to the molecular weight of an hsp90. Use of this start codon would add 140 residues to the previously reported sequence, but the additional sequence is beyond the region of heat shock protein homology and bears no similarity to any nucleotide or protein sequences currently in GenBank. Overall, the amino acid sequence of TRAP1 protein would be only 35% identical to hsp90 and 15% identical to hsp70, suggesting that it is only a distantly related member of the heat shock gene family.
Given the specific interactions observed in our study, GalNAc-T5 or TRAP1 would appear to provide attractive candidate genes for the elusive EXT3 locus. However, TRAP1 has been mapped to human chromosome 16 and we have localized GalNAc-T5 to chromosome 2 (unpublished data). A preliminary literature search suggests that, in contrast to EXT1 and EXT2, frequent loss of heterozygosity for the short arm of chromosomes 2 and 16 has not generally been observed in chondrosarcomas or osteosarcomas (15,37,38). However, given that TRAP1 has been reported to interact with four proteins that have roles in tumorigenesis (TNF1R, Rb, EXT1 and EXT2), it will be interesting to explore the possible role of TRAP1 in a subgroup of these malignancies or in HME-related chondrosarcomas.
MATERIALS AND METHODS
Yeast two-hybrid vectors
The C-terminal conserved regions of EXT1, EXT2, EXTL1 and EXTL3 were RT-PCR amplified from placental RNA with the following oligonucleotides:
EXT1 REV BamHI (CTGGATCCTGATGAGTGGATCTGCACTG);
EXT2 FOR EcoRI (CGGAATTCACGAATGAAGTGTCCATGGTG);
EXT2 REV BamHI (CTGGATCCCTACTCTGACATCAGAGTGC);
EXTL1-3[prime] FOR EcoRI (CGGAATTCTCCATGGTTCTCACCAC);
EXTL1-3[prime] REV BamHI (CGGGATCCGATAGGTTTTACCGGTGTCC);
EXTL3-3[prime] FOR EcoRI (CGGAATTCTCCATGGTGCTGACAGGTGCT); and
EXTL3-3[prime] REV BamHI (CGGGATCCGTTTCCTTCCACTCCTCCTGC).
Amplification products were purified using a High Pure PCR product purification kit (Boehringer Mannheim, Norristown, PA), restriction digested with EcoRI and BamHI and cloned into the corresponding sites of pAS2-1 (Clontech, Palo Alto, CA). Unless otherwise noted, all vectors were propagated by transformation into MAX efficiency DH5[alpha] competent cells (Gibco, Baltimore, MD). Plasmids were purified using a Qiagen (Santa Clarita, CA) QIAprep spin miniprep kit and the entire DNA sequence of the fusion proteins was confirmed by dye terminator cycle sequencing using an ABI 377 automated fluorescence sequencer (Applied Biosystems, Foster City, CA). The EXT1-H627Del and EXT2-H601Del constructs were generated by PCR site-directed mutagenesis using the following oligonucleotides:
EXT1 HIS DEL REV (GTAGTGATAATATTTGTAAATAGCAGCTCC);
EXT2 HIS DEL FOR (GGGGCAGCTTTTTATAAGTATTTTAATTAC); and
EXT1 HIS DEL REV (GTAATTAAAATACTTATAAAAAGCTGCCCC.
The LxCxE motif in TRAP-1 was modified to LxMxE by similar methods (24).
Yeast two-hybrid analysis
Yeast two-hybrid screening was performed using the strains, media, vectors and protocols from the Matchmaker two-hybrid system 2 (Clontech). The EXT1-pAS2-1 and EXT2-pAS2-1 plasmids were initially transformed into Y190 and selected by plating on synthetic dropout (SD)/Trp- medium. Expression of the fusion proteins was confirmed by western blot analysis with a GAL4 BD monoclonal antibody (Clontech). These cells were then transformed with 25 µg of both the keratinocyte (pACT2) and lymphocyte (pACT) cDNA libraries (Clontech). Approximately 1 × 106 transformants were spread on SD/Leu-/
Trp-/His-/3-AT+ (25 mM) plates. After a 7-10 day incubation at 30°C, positive clones were patched on SD/Leu-/Trp-/His-/3-AT+ (25 mM) plates, incubated for 3 days at 30°C and assayed for [beta]-gal activity using a colony-lift filter assay. Quantitative [beta]-gal assays were performed using CPRG (Boehringer Mannheim). Plasmid DNA from the positive clones (Leu+/Trp+/His+) was recovered by electroporation of 1 µg of total yeast DNA into DH10B bacterial cells (Gibco). False positives from these screens failed to activate lacZ expression in subsequent assays (n = 2), were discarded based on DNA sequence analysis (e.g. mitochondrial, n = 2) or failed to yield pACT or pACT2 plasmids (n = 7).
In vitro binding assays
The C-terminal conserved regions of EXT1 and EXT2 were PCR amplified from the previously described pAS2-1 constructs, cloned into the EcoRI and SalI sites of pGEX-5X-1 (Pharmacia, Piscataway, NJ) and propagated in the bacterial strain BL21 or DH5[alpha]. Single colonies were picked into 5 ml cultures of 2× YTG (containing 100 µg/ml ampicillin), grown overnight at 30°C and used to inoculate 1 l cultures of the same medium. Large-scale cultures were grown for 3-8 h (until OD600 nm = 0.6-0.8) at 30°C and induced with 0.1 mM IPTG for an additional 2 h at 30°C. Fusion proteins were purified using standard procedures (39) with 2 ml of a 50% slurry of glutathione-agarose beads (Sigma, St Louis, MO). Purified DNA from the yeast two-hybrid clones was used as the template for PCR amplifying fragments of the TRAP1 and GalNAc-T5 genes with the following primers:
TRAP1 REV (CTGTCATCTGTGGTGTCAGTC);
T7-GalNAc-T5 FOR (GAATTCTAATACGACTCACTATAGGGAGACCACCATGCAAA-AGGCAGACCCCAAAGAG); and
GalNAc-T5 REV (CTTAAGCAGACTCTGATCCCAAG).
Amplification products were purified and labeled with [35S]methionine (Amersham, Piscataway, NJ) using the TNT T7 Quick Coupled Transcription/Translation System (Promega, Madison, WI). Approximately 3 µg of each GST fusion protein was washed for 5 min at 4°C in 500 µl of binding solution [50 mM NaCl, 25 mM Tris-HCl, pH 7.5, 0.5% Nonidet P-40, 1 mM EDTA, 1 mM dithiothreitol, 1 mg/ml bovine serum albumin and Complete protease inhibitor (Boehringer Mannheim)]. Binding assays were performed for 2 h at 4°C in 200 µl of binding solution containing 15 µl of the GalNAc-T5 and luciferase in vitro translated proteins. Following incubation, the beads were washed six times in 1 ml of the same solution. Samples were boiled in 30 µl of sample buffer and directly analyzed by SDS-PAGE (12.5%).
TRAP1 and GalNAc-T5 extension and analysis
TRAP1 was removed from the vector pACT using EcoRI and BglII, separated by agarose gel electrophoresis, excised and recovered using a QIAquick gel extraction kit (Qiagen). The resulting 2.0 kb TRAP1 fragment was radiolabeled and extended by hybridization to 1 × 106 clones from a human testis cDNA library (Gibco). The primary and secondary cDNA screens were performed under standard conditions (40). DNA sequences of the updated TRAP1 and GalNAc-T5 genes (GenBank accession nos AF154107 and AF154108) were analyzed with Sequencher (Gene Codes, Ann Arbor, MI).
Mammalian two-hybrid vectors and analysis
The C-terminal end of EXT1 was PCR amplified from first strand placental cDNA using the oligonucleotides EXT1 FOR EcoRI and EXT1 REV XhoI (CCGCTCGAGTGATGAGTGGATCTGCACTG). The resulting fragment was digested with EcoRI and XhoI and cloned into the EcoRI and SalI sites of pM1 (40), a mammalian expression vector fusing the GAL4 BD to the C-terminal end of EXT1. The EXT1-H627Del PCR fragment derived from the oligonucleotides EXT1 FOR EcoRI and EXT1 REV BamHI was digested with EcoRI and BamHI and directly cloned into the corresponding sites of pM1. Plasmid DNA from TRAP1 in pACT was digested with EcoRI and BglII and cloned into the EcoRI and BclI sites of pVP-FLAG5, a mammalian expression vector fusing the transactivation domain of the herpes simplex virus VP16 with TRAP1 (30). The SV40 promoter drives expression of both GAL4-EXT and VP16-TRAP1. The pVP-FLAG5 plasmid was propagated in the Dam and Dcm methylation-deficient strain SCS110 (Stratagene, La Jolla, CA). Two additional vectors, G5E1bLUC, which contains the TATA element from the adeno-viral E1b gene and five GAL4-responsive UASG sites upstream of the firefly luciferase gene, and pSV-[beta]-GAL, which constitutively expresses the [beta]-gal gene to normalize for transfection efficiency, were used in this assay (41). Detailed protocols for the vector construction, mammalian two-hybrid transfections and luciferase assays were as previously described (30,42). Briefly, 3 µg of purified DNA from each of the GAL4-EXT, VP16-TRAP1 and G5E1bLUC plasmids and 1 µg of purified DNA from the pSV-[beta]-GAL plasmid were transiently co-transfected into embryonal kidney 293 cells and assayed for luciferase activity.
ACKNOWLEDGEMENTS
We thank Jennifer Ashley, Teresa Gallardo, Lurdes Queimado, Amy Wandstrat and Julia T. Tsan for reagents and advice. We thank Richard Baer and Anne Bowcock for helpful discussions and critical reading of the manuscript. This work was supported by a grant from Texas Scottish Rite Hospital for Children (to M.L.).
REFERENCES
+Present address: Division of Human Genetics, Department of Genetics, Washington University Medical Center, St Louis, MO 63110-8232, USA
§To whom correspondence should be addressed. Tel: +1 314 747 3261; Fax: +1 314 747 2489; Email: lovett{at}genetics.wustl.edu
This page is run by Oxford University Press, Great Clarendon Street, Oxford OX2 6DP, as part of the OUP Journals
Comments and feedback: jnl.info{at}oup.co.uk
Last modification:
Copyright© Oxford University Press, 1999.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  A. Leskovar, H. Wegele, N. D. Werbeck, J. Buchner, and J. Reinstein The ATPase Cycle of the Mitochondrial Hsp90 Analog Trap1 J. Biol. Chem., April 25, 2008; 283(17): 11677 - 11688. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. C. Lemos, P. Kotanko, P. T. Christie, B. Harding, T. Javor, C. Smith, R. Eastell, and R. V. Thakker A Novel EXT1 Splice Site Mutation in a Kindred with Hereditary Multiple Exostosis and Osteoporosis J. Clin. Endocrinol. Metab., September 1, 2005; 90(9): 5386 - 5392. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 T. Morita, A. Amagai, and Y. Maeda Translocation of the Dictyostelium TRAP1 homologue to mitochondria induces a novel prestarvation response J. Cell Sci., November 15, 2004; 117(24): 5759 - 5770. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||