For example, binding to the ubiquitin Keywords endoplasmic reticulum-associated degradation; Hrd1; Ube4b; ubiquitin ligase; valosin-containing protein Correspondence G.. We report a VCP-
Trang 1division cycle protein 48 binding interactions among
endoplasmic reticulum-associated degradation proteins Giacomo Morreale, Laura Conforti, John Coadwell, Anna L Wilbrey and Michael P Coleman Laboratory of Molecular Signalling, The Babraham Institute, Cambridge, UK
Valosin-containing protein (VCP⁄ p97) is an
AAA-ATPase associated with a variety of cellular activities,
most especially endoplasmic reticulum (ER)-associated
degradation (ERAD) [1], and its functional diversity
derives partly from its ability to bind a wide range of protein cofactors [2] Some bind directly to VCP in a mutually exclusive manner, targeting VCP to a particu-lar function For example, binding to the ubiquitin
Keywords
endoplasmic reticulum-associated
degradation; Hrd1; Ube4b; ubiquitin ligase;
valosin-containing protein
Correspondence
G Morreale, The Babraham Institute, B501,
Babraham Research Campus, Babraham,
Cambridge CB22 3AT, UK
Fax: +44 1223 496348
Tel: +44 1223 496251
E-mail: giacomo.morreale@bbsrc.ac.uk
Centro di Ricerca per la Viticoltura, Via
Casoni, 13/A, 31058 Susegana (TV), Italy
Fax: +39-0438-738058
Tel: +39-0438-73264
E-mail: giacomo.morreale@entecra.it
(Received 21 August 2008, revised 9
December 2008, accepted 16 December
2008)
doi:10.1111/j.1742-4658.2008.06858.x
Endoplasmic reticulum (ER)-associated degradation (ERAD) is a cell-autonomous process that eliminates large quantities of misfolded, newly synthesized protein, and is thus essential for the survival of any basic eukaryotic cell Accordingly, the proteins involved and their interaction partners are well conserved from yeast to mammals, and Saccharomyces cerevisiae is widely used as a model system with which to investigate this fundamental cellular process For example, valosin-containing protein (VCP) and its yeast homologue cell division cycle protein 48 (Cdc48p), which help to direct polyubiquitinated proteins for proteasome-mediated degradation, interact with an equivalent group of ubiquitin ligases in mouse and in S cerevisiae A conserved structural motif for cofactor bind-ing would therefore be expected We report a VCP-bindbind-ing motif (VBM) shared by mammalian ubiquitin ligase E4b (Ube4b)–ubiquitin fusion degra-dation protein 2a (Ufd2a), hydroxymethylglutaryl reductase degradegra-dation protein 1 (Hrd1)–synoviolin and ataxin 3, and a related sequence in
Mr78 000 glycoprotein–Amfr with slightly different binding properties, and show that Ube4b and Hrd1 compete for binding to the N-terminal domain
of VCP Each of these proteins is involved in ERAD, but none has an
S cerevisiae homologue containing the VBM Some other invertebrate model organisms also lack the VBM in one or more of these proteins, in con-trast to vertebrates, where the VBM is widely conserved Thus, consistent with their importance in ERAD, evolution has developed at least two ways
to bring these proteins together with VCP–Cdc48p However, the differing molecular architecture of VCP–Cdc48p complexes indicates a key point of divergence in the molecular details of ERAD mechanisms
Abbreviations
Atx-3, ataxin 3; Cdc48p, cell division cycle protein 48; DAPI, 4¢,6-diamidino-2-phenylindole; EGFP, enhanced green fluorescent protein; ER, endoplasmic reticulum; ERAD, endoplasmic reticulum-associated degradation; gp78, M r 78 000 glycoprotein; GST, glutathione S-transferase; Hrd1, hydroxymethylglutaryl reductase degradation protein 1; IBMPFD, inclusion body myopathy associated with Paget disease of bone and frontotemporal dementia; OTUD7a, OUT domain-containing protein 7; SMURF, Smad ubiquitination regulatory factor; Ube4b, ubiquitin ligase E4b; Ufd, ubiquitin fusion degradation protein; VBM, valosin-containing protein binding motif; VCP, valosin-containing protein; Wld S , slow Wallerian degeneration protein.
Trang 2fusion degradation protein (Ufd) 1–Npl4 dimer targets
VCP to a function in ERAD, whereas binding to p47,
which competes with Ufd1–Npl4, targets VCP to a
role in homotypic membrane fusion [3,4] In other
cases, the function of the binding interaction is not
fully understood, but there are further examples of
mutual exclusivity [5–7] Thus, the principle that
cofac-tor binding determines functional specificity of VCP
may be more wide-ranging, perhaps targeting VCP to
different branches of the ubiquitin proteasome system
according to which ligase it binds [5]
We recently reported that VCP binds directly to the
N-terminal 16 amino acids of ubiquitination
fac-tor E4b (Ube4b) [8], a protein involved in
multiubiqui-tination and ERAD [9–12] Similar arginine-rich
sequences were subsequently identified in the
polyglu-tamine protein ataxin 3 (Atx-3) [13], which has
ubiqu-itin protease activity [14,15], and in the ER-resident
ubiquitin ligase Mr78 000 glycoprotein (gp78) (also
known as autocrine motility factor receptor) [5], a key
regulator of retrotranslocation during ER-associated
degradation [16,17] These sequences were respectively
termed VCP-binding motif (VBM) and
VCP-inter-acting motif [5,13]
VCP is a highly conserved protein whose functions
have been extensively explored in invertebrate
homo-logues In particular, studies of cell division cycle
protein 48 (Cdc48p) in Saccharomyces cerevisiae have
uncovered roles in many cellular processes, including
membrane fusion [18], ERAD [19] and spindle
disas-sembly [20], and have played a key role in
identify-ing the cofactors that direct Cdc48p to these
functions [3,21] These functions and binding
part-ners are well conserved in mammals, consistent with
the fundamental importance of these processes for
cell survival More specifically, S cerevisiae Cdc48p
interacts with Ufd2p, the homologue of Ube4b [11],
so S cerevisiae is used as a model for invstigating
the role of VCP and its associated ubiquitin ligases
during ERAD
However, despite the fundamental importance of
ERAD to cell survival, and despite good conservation
of the proteins and their binding partners, differences
in binding sites have begun to emerge For example,
S cerevisiaeUfd2p uses a C-terminal sequence to bind
Cdc48p [12], in contrast to the N-terminal sequence
used by mammalian Ube4b [8] Mammalian VCP and
S cerevisiae Cdc48p are also not functionally
inter-changeable, despite their strong homology [22] Thus,
to know how well the mechanism is conserved, it is
important to understand fully the differences in how
VCP–Cdc48p interacts with Ube4b–Ufd2p and with
other VBM-containing proteins
We hypothesized that additional ubiquitin-metaboliz-ing proteins would bind VCP through a similar VBM, and our search identified a functional VBM close to the C-terminus of the E3 ligase Hrd1, a protein involved
in retrotranslocation during ERAD [23–25] and in turnover of the important disease-related proteins p53, expanded polyglutamine and Pael receptor [26–28] Binding of Ube4b, Hrd1 and Atx-3 requires the sequence RXXR within a predicted a-helix However, neighbouring amino acids also influence binding, and a similar motif required for VCP binding in gp78 toler-ates substitution of these two arginines We map the site of binding of Ube4b to the N-domain of VCP and show that it competes for this site with Hrd1 Finally,
we investigate the evolutionary divergence of the VBM and discuss its consequences for mechanism
Results
Identification of the VBM in Hrd1 and its refinement in gp78
Our search for additional mammalian ubiquitin ligases that contain a sequence similar to the VBM of Ube4b led us to Hrd1–synoviolin, which binds VCP through its C-terminal cytosolic region (amino acids 236–626) [24] Within this region of mouse and human Hrd1, we identified four consecutive arginines close to the C-ter-minus at positions 599–602 [23] Using the glutathione S-transferase (GST)-fused sequence DAAELRRRRL-QKLESPVAH, we then showed that this sequence is sufficient to pull down 35S-labeled VCP expressed by
in vitro transcription and translation (Fig 1A, lane 5) Using a similar approach, we also confirmed an earlier report that the Atx-3-derived peptide MTSEELRKR-REAYFEK binds VCP [13] (Fig 1A, lane 4) We also identified arginine-rich motifs in three further ubiquitin-metabolizing proteins, the zinc finger deubiquitinating protein OUT domain-containing protein 7 (OTUD7a) [29], and the ligases Smad ubiquitination regulatory fac-tor (SMURF)1 and SMURF2, which are identical to each other in this region [30,31] Neither of these sequences bound VCP in the GST pulldown assay (Fig 1A, lanes 1 and 2), indicating both the importance
of neighbouring amino acids and the specificity of the binding to Hrd1, Ube4b and Atx-3 VBM sequences Another RING finger E3 ligase involved in ERAD, gp78, is homologous to Hrd1 [23] The homology is mostly in the N-terminal regions of these two proteins, but the C-terminal 12 amino acids of gp78 show 50% identity to the sequence around the Hrd1 VBM, and VCP binding has been mapped successively to the last
49 and 30 amino acids of gp78 [5,17] Therefore, we
Trang 3used the GST pulldown assay to test VCP binding of
the peptide MLAAAAERRLQRQRTT, which spans
this region of homology to the Hrd1 VBM
Surpris-ingly, in view of the ability of homologous sequences
in Ube4b, Hrd1 and Atx-3 to bind VCP, this region of
gp78 was not sufficient for binding Instead, we found
that gp78 has a bipartite VCP-binding sequence,
requiring also a slightly more N-terminal arginine-rich
sequence (Fig 1B) Both arginine-rich sequences in
gp78, including the VBM-like sequence, are necessary
for binding, but neither is sufficient
Mutational analysis of VBM in Ube4b, Hrd1,
Atx-3 and gp78
We then refined the sequence requirements for VCP
binding in the homologous motifs of Ube4b, Hrd1,
Atx-3 and gp78 First, we extended our previous
dele-tion analysis of Ube4b [8] to show that amino acids
9–16 are necessary and sufficient to bind VCP, whereas
amino acids 1–8 were dispensable as long as other
amino acids supplied by the GST vector took their
place (Fig 1A, lane 8), possibly to maintain the
appro-priate secondary structure Complete removal of amino
acids 1–8 should disrupt a predicted a-helix spanning
amino acids 5–17 and may therefore alter binding
indi-rectly (data not shown) We then showed that alanine
substitution at Arg10 or Arg13, or at Leu14, disrupts
or severely weakens binding of VCP in this assay
with-out altering the predicted secondary structure
(Fig 2A,B) Alanine substitution at other sites had no
detectable effect, suggesting that Arg10, Arg13 and
Leu14 are important for contacting the binding site
on VCP Similar mutation analysis in Atx-3 and Hrd1 revealed analogous binding requirements The equivalent arginines were required for strong binding
in both proteins, and in Hrd1 a leucine equivalent to Ube4b Leu14 was also required (Fig 2B)
In gp78, although the analogous sequence is neces-sary for VCP binding (Fig 1B), mutation of amino acids aligning with Ube4b Arg10 and Arg13 (Fig 2A) did not disrupt VCP binding (Fig 2B) The gp78–VCP interaction was, instead, weakened by mutating the leucine that lies between them, and more severely affected by alanine substitution at Arg626 (Fig 2C, indicated as position b) Thus, the VBM consensus sequence RXXR is necessary for VCP binding in Ube4b, Atx-3 and Hrd1, but the influence of neigh-bouring amino acids is indicated both by the require-ment for an additional leucine in Ube4b and Hrd1 and
by the lack of VCP binding in OTUD7a and SMURF1⁄ 2
VBM dependence for VCP binding in intact proteins in vitro and in cells
In order to confirm a similar dependence on the VBM for interaction between intact proteins, we then repeated the VCP pulldown experiment using full-length Atx-3 and R282A Atx-3 (Fig 3A) As in the peptide experiments, there was a clear dependence on the VBM A VCP concentration of approximately
8 ngÆlL)1was sufficient to be pulled down by immobi-lized wild-type Atx-3 This is significantly lower than
Fig 1 An arginine-rich VBM common to several ERAD proteins (A) Left: table show-ing GST-fused peptides tested for their abil-ity to pull down VCP Each peptide is fused
to pGEX vector-encoded amino acids at both the N-terminus and C-terminus Right: western blot (top) and SDS ⁄ PAGE (bottom) showing precipitated VCP and GST peptides All except the OTUD7a- and SMURF1 ⁄ 2-derived peptides efficiently precipitated VCP Lane 8 refines our earlier mapping of the VBM in Ube4b [8] to amino acids 9–16 (B) Refining the VCP-binding sequence of gp78 using similar methods The table (left) shows N-terminal and C-terminal deletions within the GST-fused peptide used in (A), and the western blot (right) shows that neither of the two arginine-rich sequences alone is sufficient to bind VCP in this case.
Trang 4the average VCP concentration inside a cell, indicating
that these two proteins should also bind in vivo
Together with our earlier report that the removal of
the N-terminal, VBM-containing 16 amino acids of
Ube4b blocks binding of the otherwise intact protein
[8], this indicates similarities between the VCP-binding
sites of intact ERAD proteins and those indicated by
the peptide experiments above
Extrapolating these findings to physiological protein
levels is complex For example, it is possible that high
concentrations of VCP in the vicinity of the ER might still drive weak binding of mutated VBM-containing proteins Competitive or cooperative binding between the various VBM-containing proteins introduces another variable (see below) Thus, we tested whether VBM-dependent complexes do form in living cells Transiently transfected FLAG-tagged wild-type Atx-3 was able to coimmunoprecipitate VCP from HeLa cells, whereas R282A Atx-3, which lacks a functional VBM, could not (Fig 3B) It is not feasible to do this experiment with endogenous protein, as even a knockin mouse may not survive, due to ER stress, but
if R282A Atx-3 does not form stable complexes with VCP even when it is overexpressed, it is even less likely
to bind strongly at endogenous protein levels We cannot rule out the possibility that weak complexes formed inside cells fall apart during the coimmuno-precipitation experiment, but it is clear that an intact VBM is required for high-affinity binding inside cells
We then showed that both Hrd1 and Ube4b colocal-ize with VCP in transfected cells in a VBM-dependent manner Binding of overexpressed Hrd1 to VCP has been shown to cause both proteins to accumulate in cytoplasmic aggregates [25] We confirmed this prop-erty in our study, but when we disrupted the VBM of Hrd1 with an R599A mutation, this aggregation no longer occurred, consistent with a model in which Arg599 is a critical mediator of VCP binding (Fig 3C–J) Mutant Hrd1 assumed a more reticular pattern, possibly reflecting binding to other ER pro-teins VCP can also be partially redistributed by trans-fection with the slow Wallerian degeneration protein (WldS), this time into discrete intranuclear foci [8] Although this is not the normal distribution of VCP, these foci do provide a site for specific colocalization studies, at least for proteins such as Ube4b, which enter the nucleus Therefore, we transfected HeLa cells with WldS to determine whether Ube4b colocalizes with VCP in these foci in a VBM-dependent manner FLAG-tagged Ube4b colocalized in most cells, but this was never seen with the R10A mutant Ube4b (Fig 3L–S) These experiments have some unavoidable limitations Both rely on mislocalized VCP, and the fact that Hrd1 is a multispanning ER membrane protein that also interacts with other VCP-binding
ER proteins [24] makes it difficult to confirm direct binding by coimmunoprecipitation Thus, the coimmunoprecipitation of VCP with Atx-3 remains the best evidence for VBM-dependent binding in cells, but these data are consistent with Hrd1 and Ube4b also binding in a VBM-dependent manner inside cells
Fig 2 Mutational analysis of VBMs in Ube4b, Hrd1, Atx-3 and
gp78 (A) Table showing alignment of the VBM-containing peptides
that were sufficient to bind VCP in Fig 1 Amino acids 9–16 of
Ube4b, which were shown to bind VCP in Fig 1, are underlined,
and the motif RXXR within this sequence is aligned with an
equiva-lent motif in the other three peptides, with these two arginines also
underlined Amino acids counting from the first of these arginines
are assigned positions e, f, g, h, i (bottom row), and the more
N-terminal arginine-rich stretch in gp78, shown to be required for
binding (Fig 1B), is assigned positions a, b and c, as shown (B, C)
Amino acids e–i in each protein (B), and a, b and c in gp78 (C),
were then mutated to alanine in turn, and the capacity for VCP
binding was retested and compared to that of the nonmutated
pep-tide (NM) Western blots of pulldown material indicate VCP binding.
Arginines at positions e and h are essential for binding in Ube4b,
Hrd1 and Atx-3, but in gp78, position b is the only individually
essential arginine.
Trang 5Mapping the Ube4b interaction domain in VCP
The binding sites of gp78, Atx-3, Hrd1 have all been
mapped to the N-domain of VCP (amino acids 1–199)
[13,25,32,33] As the VBM of Ube4b is very similar to
those of the above proteins, we hypothesized that
Ube4b should also bind to the N-domain However,
this conflicts with data from S cerevesiae, where the
N-domain of Cdc48p is neither necessary nor
suffi-cient, and instead the D1D2 domain binds Ufd2p [34]
Therefore, we tested directly whether the N-domain of
mammalian VCP is sufficient to bind the VBM of
Ube4b, here represented by WldS, a fusion protein that
shares its N-terminal 70 amino acids with Ube4b [35]
First, we confirmed that the N-terminal 16 amino acids
of WldS, identical to the Ube4b VBM, is the only VCP-binding site in this protein (Fig S1) We then found that GST-fused VCP1–199 was sufficient to pull down WldS, indicating that the N-terminal Ube4b-(and WldS)-binding site of VCP also resides within this region (Fig 4A) Thus, there are differences between mammals and S cerevisiae in the sequences mediating binding both on the Ube4b–Ufd2p side [8] and on the VCP–Cdc48p side
We then investigated whether the binding of each of these VBM-containing proteins to the N-domain of VCP is disrupted in disease Several mutations within the N-domain of VCP (positions 95, 155 and 191) cause inclusion body myopathy associated with Paget disease of bone and frontotemporal dementia
A
B
Fig 3 (A) Coomassie-stained gel showing pulldown of VCP using full-length recombi-nant Atx-3 fused to GST and a VCP concen-tration of 8 ngÆlL)1 This interaction is prevented by the mutation of Arg282 (B) Western blots showing coimmunoprecipita-tion of VCP with FLAG-tagged Atx-3 trans-fected into HeLa cells Again, the interaction
is blocked by mutation of Arg282, confirm-ing dependence on the VBM for bindconfirm-ing in cells (C–J) Hrd1 redistributes VCP in a VBM-dependent manner FLAG-tagged Hrd1 (C–F) or R599A mutant Hrd1 (G–J) was tran-siently transfected into the PC12 subline TV, which expresses EGFP-tagged VCP (R599 corresponds to position e in Fig 2A) Over-expressed wild-type Hrd1 aggregates together with VCP, reflecting a binding inter-action as previously reported [25] In con-trast, the R599A mutant does not redistribute VCP, and itself assumes a more reticular pattern, consistent with the R599A mutant failing to bind VCP (L–S) HeLa cells were transfected with Wld S to partially redistribute VCP into intranuclear foci as pre-viously reported [8], so that these foci could
be used for specific colocalization studies Faint FLAG-tagged Ube4b signal colocalized with VCP in these foci (arrows, N), whereas the R10A mutant (R) did not [this can be better seen in Fig S3, where parts (L)–(S) are all equally enhanced by adjusting levels
in PHOTOSHOP]
Trang 6(IBMPFD) [36,37] Not only does the binding site of
each VBM-containing protein map to amino acids 1–
199, but that of Atx-3 has been mapped even closer to
the IBMPFD mutations at amino acids 143–199 [38]
Therefore, we tested whether disruption of VBM
bind-ing could be part of the pathogenic mechanism GST
constructs fused to VCP1–199 containing the point
mutations R155C, R155H, R155P, R159H, R159T and
R191Q were able to pull down bacterially expressed,
His-tagged WldS (Fig 4B) Together with a previous
report that mutation of Arg93 or Arg155 does not
block binding of Atx-3 [39], this suggests that binding
of VBM-containing proteins is unaltered by the
IBMPFD mutations In view of the highly basic nature
of the VBM, we also tested whether the acidic
sequence EDEEE(192–196) of VCP is required for
binding Individual point mutations in this VCP
sequence did not alter binding of WldS(Fig 4B)
Mutually exclusive binding of VBM-containing
proteins to VCP
As discussed above, several proteins bind the
N-domain of VCP in a mutually exclusive manner,
including the VBM-containing proteins Atx-3 and
Ube4b [13] Knowing which proteins compete with one
another for binding is an essential step towards
under-standing how VCP interacts with other proteins inside
a cell Thus, having identified Hrd1 as a new
VBM-containing protein (Fig 1), and having shown that
Ube4b binds the N-domain (Fig 4), like Hrd1 [25], we
then looked for competition between them for VCP
binding Preincubating VCP with WldS, which contains
the N-terminal region of Ube4b [35] to allow maximal
blocking of the VBM-binding sites on VCP, we found
an inverse relationship between the amount of WldSin the input and the amount of VCP that Hrd1 could pull down (Fig 5) Half-maximal inhibition corresponded
to a WldS⁄ VCP polypeptide ratio of approximately 2.4 (approximately 72 lg of VCP and 75 lg of WldS, with molecular masses of 97 and 43 kDa respectively) Thus, Hrd1 is excluded from binding VCP by increas-ing amounts of the Ube4b-derived VBM Precisely how closely this models protein concentrations in the vicinity of the ER is unknown, but as most of these proteins are abundant at the ER, there is likely to be significant competition between the various VBM sequences for binding the VCP N-domain
Evolutionary conservation of VBM-containing proteins
We previously reported that the VBM of Ube4b is located within an N-terminal extension that is absent in
S cerevisiae[8] We now show that the VBMs of Hrd1 and gp78 are also missing from their common S cere-visiae homologue, as we map them to sites that are C-terminal extensions in the mammalian proteins [23] Ube4b also docks at a different site on VCP from where Ufd2p binds Cdc48p (Fig 4), and Atx-3 apparently has
no close S cerevisiae homologue These observations indicate that there is evolutionary divergence in the molecular architecture of VCP–Cdc48p-containing complexes in ERAD, despite conservation of the princi-ple of VCP–Cdc48p binding To understand more about the evolutionary conservation of the VBM in each of these proteins, we looked for VBM-like sequences in a range of organisms (Fig 6 and Tables S1 and S2)
Fig 4 The VCP N-domain precipitates WldS
independently of pathogenic mutations (A)
Western blot (top) showing Wld S (and
prote-olytic fragments of WldSrecognized by the
same specific antibody) precipitated by
GST-fused VCP1–199 (SDS ⁄ PAGE, bottom) (B)
The same VCP fragment was able to
precip-itate WldSeven when pathogenic VCP
mutations were incorporated (top panel and
lane 1 of bottom panel) or when the highly
acidic sequence at amino acids 191–196
was mutated (bottom panel) Controls
shown on the top panel are valid for the top
and bottom panels (different gels from the
same experiment) The lane marked Wld S
was loaded with the input Wld S MM,
marker lane.
Trang 7Interestingly, in contrast to our earlier report [8],
the recent database submission CAC19740 indicates
that there is a VBM-like sequence in Ufd2p of
Schizosaccharomyces pombe (Table S2), whose
func-tionality we confirmed experimentally (Fig S2) This
difference from S cerevisiae probably reflects the
major divergence of present-day yeasts from a
common ancestor [40] Caenorhabditis elegans, in
contrast, has a putative VBM in Hrd1 but not in
Ufd2p, whereas in nearly every vertebrate that we
studied, there was a well-conserved putative VBM in
all four proteins (Table S1) Thus, VBMs in these
four ERAD proteins are very well conserved
among vertebrates, but only sporadically present in
invertebrates
Discussion
Our data indicate that molecular interactions
govern-ing ERAD diverge significantly between vertebrates
and many invertebrates, despite the essential nature
of this cell-autonomous process Despite good
conser-vation of most of the proteins involved, and strong similarities in the pattern of binding partners, the sequences that mediate these interactions are significantly different from those in mammals in
S cerevisiae and in many other invertebrate homologues Not only is the corresponding protein domain absent, but our characterization of essential amino acids for VCP binding indicates that the VBM does not appear elsewhere in these proteins These differences in molecular structure of the VCP ERAD complexes indicate a divergence point in this basic cellular mechanism that was not evident from earlier data Intriguingly, however, the fact that evolution has established more than one way for these proteins to interact shows how important it is that they do so
We have mapped the VCP-binding activity of Hrd1
to a C-terminal VBM, refined the VBMs of gp78 and Ube4b [5,8], and confirmed a functional VBM in Atx-3 [13] Each protein is important for ERAD Hrd1 and gp78 are E3 ligases ubiquitinating ERAD substrates [17,23] Haploinsufficiency for Ube4b, an E4
Fig 5 Hrd1 and Wld S bind VCP in a mutually exclusive manner (A) Fast Blue-stained SDS ⁄ PAGE gel (below) showing the effect of preincu-bating various quantities of Wld S bacterial extract with a bacterial extract containing recombinant VCP for 30 min at 4 C before precipitating with GST–Hrd1(VBM) bound to glutathione resin Note that other proteins present in the bacterial extracts help to block nonspecific protein interactions and binding of Wld S and Hrd1 to VCP is direct [8,24] Western blots below show VCP and Wld S input Densitometry values were zeroed for the background Note the decreasing interaction of VCP with GST–Hrd1(VBM) when it is preincubated with more Wld S (B) Com-parative histogram of VCP bound to GST–Hrd1(VBM); 750 lL of WldSwas taken to confer half-maximal inhibition of Hrd1–VCP binding, a point corresponding to a Wld S ⁄ VCP polypeptide ratio of approximately 2 : 4 (see text) Data points are mean values ± standard error; n = 4.
***P < 0.0001.
Trang 8ligase, triggers ER stress and neurodegeneration in
mice [10] Atx-3 inhibits retrotranslocation, probably
through deubiquitination [7] Despite excellent
conser-vation among vertebrates, each VBM is absent in
S cerevisiae, and Ube4b–Ufd2p binds to different sites
on VCP–Cdc48p in mammals and S cerevisiae
The sequence RXXR is almost invariant
through-out these VCP-binding sequences Human gp78
dif-fers in having a conservative lysine for arginine
substitution [23], in requiring a second arginine-rich
stretch for VCP binding, and in tolerating arginine
to alanine mutations in the sequence RXXR
How-ever, we class gp78 as a variant VBM, as this region
is still required for VCP binding (Fig 1) Mutational
analysis and comparison with other RXXR proteins
indicates that neighbouring amino acids also
influ-ence VCP binding
The binding motif that we define in Ube4b, Hrd1 and Atx-3 is similar but not identical to the motif (L⁄ I ⁄ V ⁄ Y)-R-(K ⁄ R ⁄ W)-(R ⁄ K ⁄ L)-R-X-X-(Y ⁄ F)-(F ⁄ K ⁄
L⁄ Y) reported in Atx-3 [13] We find more tolerance
of alanine substitution around the essential arginines, perhaps because different methods were used for muta-tion analysis Short synthetic peptides [13] may not preserve secondary structure as effectively as GST fusion proteins, which is important because each VBM lies within a predicted a-helix, in which the two con-served arginines would project positively charged side chains to the same face Alternatively, dimerization of the GST fusion proteins may influence the strength of VCP binding
Interestingly, the Hrd1, gp78 and Ube4b VBMs are located within C-terminal or N-terminal extensions missing from their S cerevisiae homologues [8,23] Atx-3 has no S cerevisiae homologue Thus, differ-ences in VCP–Cdc48 interaction between Ube4b and Ufd2p [8] can be generalized to a wider range of ERAD proteins This provides a structural explanation for the controversy regarding whether Hrd1 binds Cdc48–VCP directly Whereas the S cerevisiae pro-teins interact via mutual binding partner, Ubx2 [41,42], the corresponding mammalian proteins clearly exhibit
a direct interaction through a sequence that is lacking
in S cerevisiae These data do not exclude an addi-tional, direct binding site in yeast, and Ubx domains may also play a role in coordinating ERAD machinery
in mammals through initiating or strengthening these interactions [41,43,44] Mammalian VBM-containing proteins may also be linked to VCP through mutual binding partners [24,25] Thus, evolution has generated more than one way of recruiting VCP–Cdc48p to ubiquitin ligases, and some molecular details of ERAD may differ accordingly
Several proteins bind VCP in a mutually exclusive manner These include Ufd1–Npl4 and p47 [4], Ufd1– Npl4, SVIP and p47 [6], Ufd1–Npl4 and gp78 [5], and Ufd1–Npl4 and Atx-3 [7], and there is evidence that such competition can be important in regulating ERAD [45] We now extend this to Ube4b and Hrd1 (Fig 5), consistent both with their homology and with their shared use of the VCP N-domain for docking (Fig 4) [25] Interestingly, gp78 requires both the N-domain and D1-domain [5], mirroring the bipartite VCP-binding sequence that we report As these pro-teins bind differently to S cerevisiae Cdc48p, differ-ences in competition for binding are one way in which the ERAD mechanism could differ
In a biological context, two models are compatible with mutually exclusive binding: ternary complex and negative cooperativity (Fig 7) Hexameric VCP [46]
Fig 6 Evolutionary alignment of VBM of Ube4b–Ufd2, Hrd1, Atx-3
and gp78 among several vertebrate and invertebrate species VBM
or putative VBM sequences are indicated in bold Atx-3 alignment
is not shown for some species, as they lack a homologue to this
protein For further details, see Tables S1 and S2.
Trang 9could assume a central, organizing role in a ternary
complex where different VCP polypeptides bind
differ-ent cofactors Cofactors compete for each individual
site, but the six VCP subunits bring together various
ERAD proteins Substrates ubiquitinated by E3 ligases
Hrd1 and gp78 could be passed efficiently to a nearby
E4 (Ube4b) for further ubiquitination [9,11], and
indi-rect binding to more ligases via Ubx domain proteins
provides even more scope for the coordination of
ubiq-uitination in this way [43,44] VCP interacts with
Hrd1, derlin and VIMP in a ternary complex, although
multiple pairwise interactions complicate the analysis
[24,25], and gp78 and PNGase also bind VCP as a
ter-nary complex through their different binding sites [33]
In the negative cooperativity model, cofactor binding
to a single N-domain closes off all sites in the
hexamer, including unoccupied sites This allows a
single cofactor molecule to determine the role of the
VCP complex, contributing to functional diversity The
stoichiometry of one Ufd1–Npl4 dimer per VCP hexamer supports this model [47], as does the failure
or Ufd2 and Ufd3 to coimmunoprecipitate in S cerevi-siae[34]
VCP extracts ubiquitinated proteins from the ER and chaperones them to the cytoplasm for protea-some-mediated degradation The differences in molecu-lar interactions between S cerevisiae and mammals help to explain differences in ERAD For example, one key question is what recruits VCP to the ER
S cerevisiae Ubx2 optimizes this process by binding both Hrd1p and Cdc48p [41,42], whereas mammalian gp78 and Hrd1 both bind VCP directly, recruiting it to the ER and influencing ERAD [17,24] In both pro-teins, we now map VCP binding to the extreme C-ter-minus In Hrd1, the two critical arginines are located
11 and 14 amino acids from the C-terminus Interest-ingly, Atx-3 has an opposite effect on retrotransloca-tion, inhibiting it in a VCP-binding dependent manner [7] Thus, competition for a VCP-binding site between Atx-3 and Hrd1–gp78 could regulate the retrotranslo-cation process
The VBM joins a growing list of VCP-binding sequences [2] The Ubx domain of p47 [48] also occurs
in many other proteins [2,43,44,49,50], and Ufd1–Npl4 binds similarly, despite lacking homology [51] The PUB domain [52] is structurally different from VBM and Ubx, and, unlike both, binds C-terminally in VCP [53] Interestingly, PUB domains are often found in higher eukaryotes but are also absent in S cerevisiae, similarly to the VBM [52] Finally, Ufd2p binds Cdc48 directly, despite lacking the VBM of its mammalian homologue Ube4b, so an alternative binding sequence exists [11,12]
Intriguingly, although S cerevisiae Cdc48p does not use a VBM to bind the corresponding Ufd2p, Cdc48p can still bind the mammalian VBM (Fig S2) Thus, there is an evolutionary pressure to maintain the VBM-binding site in S cerevisiae Cdc48p that may come from other, as yet unidentified, binding partners The VBM of Ube4b is shared with WldS, a mutant, chimeric protein that uniquely delays axon degenera-tion and is a fusion of Ube4b sequence with the NAD+-synthesizing enzyme Nmnat1 [35,54] Ube4b sequences are required for the full phenotype [55], pos-sibly by competing for VCP binding with wild-type Ube4b Our data show that competition with other VBM-containing proteins is another possibility
In summary, the strong conservation of VBMs among four mammalian proteins involved in ERAD contrasts strikingly with the complete absence in
S cerevisiae homologues, and poor conservation in other invertebrates The docking site on VCP also
Fig 7 Ternary complex (A) and negative cooperativity (B) models
for binding between VCP hexamer and VBM-containing proteins (A)
In the ternary complex model, different VCP polypeptides in the
com-plex are able to bind different VBM proteins simultaneously, because
competition operates only at the level of each individual N-domain.
Thus, VCP coordinates formation of a ternary complex in which
VBM-containing protein 1 (e.g Ube4b) and VBM-containing protein 2
(e.g Hrd1) are brought into close proximity with one another (B) In
the negative cooperativity model, binding of one VBM-containing
protein to one VCP polypeptide (left) closes off all sites in the VCP
hexamer through conformational change VCP can only bind VBM
protein 2 when no other VBM protein is bound (right).
Trang 10differs for at least one of these proteins These
differ-ences in the molecular architecture of VCP–Cdc48p
complexes indicate divergence in ERAD mechanisms
that is not apparent from previous data Differences
are likely in how proteins compete to bind VCP and in
the relative orientation of proteins and their key
func-tional domains within the complex Future studies now
need to address the extent of these structural
differ-ences, the consequences for the mechanism, and when
and why key steps in the evolution of ERAD took
place
Experimental procedures
Bioinformatics methods
blastp2.2.10 was used to search for the motif
EIRRRRLARLA, using a local mouse database The
sig-nificance cutoff was set at 1000 to allow for the shortness
of the search string In the absence of a protein sequence,
the existence of a homologous gene was inferred from
ensemblwhere possible (Table S1) clustal-w [56] was used
for multiple sequence alignment of selected proteins
Constructs
Plasmid constructs were prepared using standard
recombi-nant techniques [57] All VBM motif sequences tested were
derived from murine sequences and cloned into pGEX5T1
via EcoRI–XhoI WldS and R10A WldS were cloned into
pET28a via BamHI–HindIII Flagged Ube4b, flagged R10A
Ube4b, flagged Hrd1 and flagged R599A Hrd1 were cloned
into pHbApr-1 via HindIII–BamHI A list of templates,
primers and plasmids used for this work is available in
Table S3
Expression of GST fusion proteins and other
recombinant proteins
Transformed Escherichia coli BL21 cells were cultured in
liquid LB medium (pGEX vectors with 50 lgÆmL)1
ampicil-lin and pET vectors with 25 lgÆmL)1 kanamycin) at 37C
to D600 nm= 1 Expression was induced by addition of
1 mm isopropyl-thio-b-d-galactoside and further shaking at
30C for 12 h
In vitro transcription and translation
Radioactive recombinant VCP was produced using the
pGBKT7 construct [8] pGBKT7 was in vitro transcribed
and translated, incorporating [35S]methionine, using the
TNT T7 Reticulocyte Lysate Coupled Transcription⁄
Translation kit from Promega (Promega Ltd, Southampton,
UK)
Binding assays GST fusion proteins were purified and coupled to glutathi-one–Sepharose 4B according to the manufacturer’s protocol (GE Healthcare, Little Chalfont, UK) For further affinity experiments, these purified proteins were mixed with various amounts of bacterial protein extracts in a 1.5 mL tube at
4C, and unbound protein was washed out using NaCl ⁄ Pi plus 0.01% Triton X-100 The glutathione–Sepharose 4B beads were analysed by SDS⁄ PAGE, the gel was dried and, when radioactive recombinant VCP was used, it was exposed directly to autoradiography film overnight
Coimmunoprecipitation Flagged Atx-3 and flagged R282A Atx-3 expression vectors were generated using standard cloning procedures, and veri-fied by restriction enzyme analysis and DNA sequencing The coding regions of Atx-3 and R282A Atx-3 were PCR-amplified using primers harbouring appropriate restriction enzyme sites and FLAG-expressing sequences, with Pfu Polymerase (Promega Ltd.), and ligated into pCDNA3.1 (Invitrogen, Paisley, UK) HeLa cells were transfected with either flagged Atx-3 or flagged R282A Atx-3 expression vectors After 24 h, cells were washed with NaCl⁄ Pi, and harvested by adding lysis buffer [20 mm Tris, pH 7.5,
137 mm NaCl, 1 mm EGTA, 1% (v⁄ v) Triton X-100, 10% (v⁄ v) glycerol, and 1.5 mm MgCl2, supplemented with Complete Mini protease inhibitor cocktail tablets (Roche Diagnostics, Lewes, UK) and scraping cells after 20 min of incubation on ice Lysates were subsequently collected and cleared by centrifugation by centrifugation for 30 min at
14 000 g at 4 C Protein concentrations were determined
by the Bio-Rad protein assay (Bio-Rad, Hemel Hempstead, UK) FLAG-tagged proteins were immunoprecipitated from equal amounts of total protein by incubating with EZview Red ANTI-FLAG M2 affinity gel (Sigma-Aldrich Ltd., Gillingham, UK) for 2 h at 4C The beads were washed three times with lysis buffer, and analysed by SDS⁄ PAGE followed by western blotting using VCP antibody (BD Biosciences, Oxford, UK) and monoclonal antibody against FLAG (M2) (Sigma-Aldrich Ltd.)
Western blotting Proteins were separated by SDS⁄ PAGE and semidry blot-ted onto nitrocellulose (Bio-Rad) Blocking, washing and incubation with primary antibodies and suitable horserad-ish peroxidase-conjugated secondary antibodies [either sheep anti-(mouse IgG) (1 : 3000; GE Healthcare Ltd.) or anti-rabbit IgG (1 : 3000; GE Healthcare Ltd.) were per-formed in NaCl⁄ Pi plus 0.02% Tween-20 and 5% low-fat milk Proteins were visualized using the ECL detection kit (GE Healthcare Ltd.) according to the manufacturer’s instructions