Structural and biochemical studies of the C-terminal
domain of mouse peptide-N-glycanase identify
it as a mannose-binding module
Xiaoke Zhou*†, Gang Zhao*†, James J. Truglio*†, Liqun Wang*†, Guangtao Li†, William J. Lennarz†,
and Hermann Schindelin*†‡§
*Center for Structural Biology and†Department of Biochemistry and Cell Biology, Stony Brook University, Stony Brook, NY 11794-5215; and‡Rudolf Virchow
Center for Experimental Biomedicine and Institute of Structural Biology, University of Wu ¨rzburg, Versbacher Strasse 9, 97078 Wu ¨rzburg, Germany
Edited by John Kuriyan, University of California, Berkeley, CA, and approved September 15, 2006 (received for review April 11, 2006)
The inability of certain N-linked glycoproteins to adopt their native
conformation in the endoplasmic reticulum (ER) leads to their
retrotranslocation into the cytosol and subsequent degradation by
the proteasome. In this pathway the cytosolic peptide-N-glycanase
(PNGase) cleaves the N-linked glycan chains off denatured glyco-
proteins. PNGase is highly conserved in eukaryotes and plays an
important role in ER-associated protein degradation. In higher
eukaryotes, PNGase has an N-terminal and a C-terminal extension
in addition to its central catalytic domain, which is structurally and
domain of PNGase is involved in protein–protein interactions, the
function of the C-terminal domain has not previously been char-
acterized. Here, we describe biophysical, biochemical, and crystal-
lographic studies of the mouse PNGase C-terminal domain, includ-
ing visualization of a complex between this domain and
mannopentaose. These studies demonstrate that the C-terminal
domain binds to the mannose moieties of N-linked oligosaccharide
chains, and we further show that it enhances the activity of the
mouse PNGase core domain, presumably by increasing the affinity
of mouse PNGase for the glycan chains of misfolded glycoproteins.
endoplasmic reticulum ? N-linked glycoproteins ? proteasome ?
protein degradation ? deglycosylation
Protein folding in the endoplasmic reticulum (ER) is assisted by
molecular chaperones, such as calnexin and calreticulin (1), that
use N-linked oligosaccharides attached to newly synthesized
proteins as tags to detect their folding status. The oligosaccha-
ride chains are attached via N-glycosidic bonds to the side-chain
amide groups of Asn residues and initially consist of a tetra-
decamer with the composition (GlcNAc)2(Man)9(Glc)3. Dy-
namic processing of the terminal glucose residues is essential for
proper folding. Correctly folded proteins are transported to the
Golgi complex for further carbohydrate modification, whereas
aberrantly folded proteins are retro-translocated to the cytosol
for degradation, which involves their ubiquitination, deglycosy-
lation, and proteolytic digestion by the proteasome.
Peptide-N-glycanase (PNGase) catalyzes the deglycosylation
of several misfolded N-linked glycoproteins (2, 3) by cleaving the
bulky glycan chain before the proteins are degraded by the
proteasome (4). PNGase is highly conserved in eukaryotes and
possesses a catalytic Cys, His, and Asp triad embedded in a
transglutaminase fold. Both mouse and yeast PNGase have been
reported to interact with HR23B?Rad23, a protein that is also
involved in DNA damage recognition (5–7). Recently the struc-
tures of yeast and mouse PNGase in complex with Rad23?
HR23B and an inhibitor (8), carbobenzyloxy-Val-Ala-Asp-?-
fluoromethyl ketone (Z-VAD-fmk), have been solved (9, 10).
The yeast protein corresponds to the central region of mouse
PNGase. The Z-VAD-fmk inhibitor covalently attaches to the
active site Cys, which otherwise carries out a nucleophilic attack
roteins need to acquire their native conformation after
protein synthesis to carry out their biological functions.
on the ?-N-glycosidic bond linking the Asn side chain and the
first GlcNAc residue, thereby cleaving the glycan chain from the
Early studies (11) revealed that mouse PNGase binds to free
glycan chains derived from its glycoprotein substrates, and that
this binding inhibits the activity of PNGase, thus suggesting that
mouse PNGase has a carbohydrate-binding activity. Moreover,
recent studies revealed that PNGase specifically acts on the
unfolded form of high-mannose type N-glycosylated proteins (4,
12, 13). However, how PNGase binds to glycan chains and how
it recognizes the high-mannose type substrates is unknown. In
this study, we present the structure of the C-terminal domain of
mouse PNGase, which is present in higher eukaryotes ranging
from Caenorhabditis elegans to humans with 30% sequence
identity between these two species. Biochemical, biophysical,
and crystallographic studies reveal that it contains a mannose-
binding domain, which presumably contributes to the oligosac-
charide-binding specificity of mouse PNGase. These findings
suggest that the C-terminal domain increases the binding affinity
between mouse PNGase and its substrates.
Results and Discussion
Overall Structure of the Mouse PNGase C-Terminal Domain. Due to
difficulties in obtaining large single crystals, this domain (resi-
dues 451–651) was expressed as either an intein fusion or a
His-tagged protein. Both purified proteins yielded crystals that
diffracted to ?2 Å; however, the space groups differed (P3221
and C2, respectively). The structure of the intein-tagged mouse
PNGase C-terminal domain was solved by using single isomor-
phous replacement and anomalous scattering with the aid of a
Hg derivative (Table 2, which is published as supporting infor-
mation on the PNAS web site) and was refined at 1.9-Å
resolution (Table 1) to an R factor of 0.167 (Rfree? 0.217). The
N-terminal residues 451–471 are disordered in this structure.
Subsequently, the structure of His-tagged form of the protein
was refined at 2-Å resolution (Table 1) to an R factor of 0.15
(Rfree? 0.207). The two structures are very similar as reflected
in a rms deviation in the C?positions of 0.24 Å. However, in the
His-tagged protein model, residues 454–463 could be visualized,
Author contributions: X.Z., W.J.L., and H.S. designed research; X.Z., G.Z., L.W., and H.S.
performed research; G.Z. and G.L. contributed new reagents?analytic tools; J.J.T. and H.S.
analyzed data; and X.Z., W.J.L., and H.S. wrote the paper.
The authors declare no conflict of interest.
This article is a PNAS direct submission.
Abbreviations: PNGase, peptide-N-glycanase; Z-VAD-fmk, carbobenzyloxy-Val-Ala-Asp-?-
fluoromethyl ketone; ER, endoplasmic reticulum; ITC, isothermal titration calorimetry.
Data deposition: The atomic coordinates and structure factors have been deposited in the
Protein Data Bank, www.pdb.org [PDB ID codes 2G9F (intein-tagged protein), 2G9G
(His-tagged protein), and 2I74 (complex)].
§To whom correspondence should be addressed. E-mail: hermann.schindelin@virchow.
© 2006 by The National Academy of Sciences of the USA
November 14, 2006 ?
vol. 103 ?
thus leaving only residues 451–453 and 464–472 as unassigned
in the electron density maps. Due to the presence of these
additional residues, discussion in this article focuses on the
His-tagged model, which is shown in Figs. 1 and 4.
The mouse PNGase C-terminal domain is a slightly elongated
molecule and displays a ?-sandwich architecture, which is com-
posed of two layers, containing nine and eight antiparallel
?-strands, respectively, and three additional short helices (Fig. 1).
on the orientation shown in Fig. 1A), deviates strongly from a
(?-strands 4–6, 11, 14, and 17), including a very long strand (?11)
that is also involved in the formation of a second four-stranded
of the molecule. ?-Strands 16 and 17, which reside in the four-
stranded and six-stranded ?-sheets, respectively, are separated by a
nine-residue-long loop, whereas the loops connecting ?4 and ?5 in
the six-stranded sheet, as well as ?8 and ?9 in the four-stranded
sheet, completely disrupt this front layer. The back layer (Fig. 1A)
displays a more traditional architecture, with a five-stranded anti-
edge strand (?7), which leaves sufficient room to allow a short
three-stranded ?-sheet (?1–3) also to hydrogen bond in a parallel
fashion with ?10. The three helices are distributed throughout the
structure, with the longest helix (?1) at one end of the molecule,
which will be referred to as the proximal end because it is adjacent
to the N and C termini, and the shortest helix (310-2) at the distal
end of the molecule. The first 310helix (310-1) links ?-strands 6 and
7 and is located between the proximal and distal ends.
of the three-dimensional structure of the protein reveals that a
depression between two loop regions and the adjacent ?-strands
(?8 and ?9 and the ?15 and ?16 junctions) is one of the two most
highly conserved regions (Fig. 1 B and C). The residues decorating
the saddle-shaped depression at the distal end include two trypto-
phans, Trp-532 and Trp-624, which sit on opposite sides on the
ridges flanking the saddle and are separated by 11 Å. Three
additional residues, Tyr-536, Phe-525, and Lys-527, are located on
the concave side of the saddle.
The Mouse PNGase C-Terminal Domain Is Similar to the Sugar-Binding
Domain of Fbs1. The structure of the C-terminal domain was
compared with a nonredundant set of proteins from the Protein
Data Bank by using the Dali server (14), which identified the
sugar-binding domain of Fbs1 (15) as its closest structural homolog
with a Z score of 9.1. Fbs1 is an F-box protein, a component of the
SCF E3 ubiquitin ligase, which is composed of the Skp1, Cul1,
Roc1?Rbx1, and F-box proteins (15). The F-box proteins are the
substrate-binding components of this E3 complex, and in the case
of Fbs1, it was shown to recognize N-linked glycans, especially the
chitobiose core via its sugar-binding domain (16, 17).
Although the C-terminal domain of mouse PNGase shares only
10% sequence identity with the sugar-binding domain of Fbs1, the
two proteins are very similar in their tertiary structures and can be
superimposed with an rms deviation of 3.7 Å for 128 aligned
residues of 184 present in Fbs1 (Fig. 5, which is published as
domain of mouse PNGase, the sugar-binding domain of Fbs1
features a ?-sandwich architecture, and its front sheet also displays
the strong curvature. The chitobiose-binding region of Fbs1 has
been mapped to two loop regions of its sugar-binding domain,
which connect the two ?-sheets at the distal end of the elongated
molecule (15). The surrounding loops in this area adopt dissimilar
the sugar-binding domain of Fbs1, which result in completely
different surface models of the two proteins (data not shown).
despite the differences in the region encompassing the ligand-
binding site of the sugar-binding domain of Fbs1, the C-terminal
domain of mouse PNGase may also be involved in carbohydrate
binding, with the saddle-shaped depression being the most likely
binding site based on sequence conservation. The types of residues
located in this putative binding pocket are entirely consistent with
those commonly observed in carbohydrate binding (18, 19). In
biochemical properties. Very recently, Yoshida et al. (20) reported
that Fbs1 interacts with N-glycoproteins, especially denatured
glycoproteins, in agreement with the fact that PNGase acts on
misfolded N-glycosylated proteins (12, 13, 20). In the same study,
also interacts with mouse PNGase (21). p97 is an AAA ATPase
that functions as an extractor for misfolded proteins from the ER
which may sequentially bind to the same N-glycoprotein immedi-
ately after its extraction from the ER by the retrotranslocon. In this
model, the denatured glycoprotein is immediately recognized by
SCFFbs1, ubiquitinated by the SCF E3 ligase once it emerges from
the ER lumen, and then forwarded to PNGase for deglycosylation,
which suggests that the ER-associated degradation pathway is
highly cooperative and more processive than previously realized.
Table 1. PNGase refinement statistics
His-tagged native Intein-fused native
Resolution limits, Å
Number of reflections
Number of protein?solvent atoms
Deviations from ideality
Bond distances, Å
Bond angles, °
Chiral volumes, Å3
Planar groups, Å
Torsion angles, °
Average B factor, Å2
7.12, 32.44, 15.34
6.95, 36.77, 13.56
6.59, 36.55, 12.53
Rcryst? ???Fo? ? ?Fc?????Fo?, where Foand Fcare the observed and calculated structure factor amplitudes. Rfreeis
the same as Rcrystfor 5% of the data randomly omitted from refinement. Ramachandran statistics indicate the
Ramachandran diagram, as defined by PROCHECK [CCP4 suite (35)].
Zhou et al.
November 14, 2006 ?
vol. 103 ?
no. 46 ?
The Mouse PNGase C-Terminal Domain Binds to Oligomannose
Carbohydrates. To investigate whether the C-terminal domain of
mouse PNGase indeed binds to carbohydrates, isothermal titration
calorimetry (ITC) experiments were performed with two oligosac-
charides (?3,?6-mannopentaose and N,N?-diacetylchitobiose) as
(Fig. 2). The C-terminal domain was shown to bind to mannopen-
taose with a dissociation constant (Kd) of ?67 ?M. Full-length
mouse PNGase was found to have a binding affinity very similar to
that of the C-terminal fragment, whereas the mouse PNGase
fragment without the C-terminal domain (1–450) displayed no
the C-terminal domain of mouse PNGase is at least primarily
responsible for the binding of mouse PNGase to mannopentaose.
On the other hand, neither the C-terminal domain of mouse
PNGase nor the full-length enzyme showed any affinity toward
chitobiose in ITC experiments. This unexpected result indicates
that mouse PNGase does not engage in high-affinity interactions
with the first two acetylglucosamine residues of the glycan chain,
but instead may require the peptide part of the substrate for a
high-affinity interaction. For comparison, the same ITC experi-
ments were also carried out with yeast PNGase (data not shown).
The yeast enzyme neither binds to mannopentaose, which is
consistent with the absence of the C-terminal domain in yeast
of the ?-sandwich is colored in cyan, and the back sheet and loops are in gray with helices in orange. ?-strands have been labeled. Figs. 1 A and B, 2C, 3 A and
B, and 4 have been generated with PyMOL (36). (B) Sequence conservation of the C-terminal domain in the context of its three-dimensional structure. Strictly
conserved residues have been mapped in red and conserved residues in light orange onto a surface representation of the molecule. A and B differ by a rotation
of ?90° around the vertical axis. (C) Multiple sequence alignment of PNGase C-terminal domains (Homo sapiens, human; Pan troglodytes, chimpanzee; Canis
and residue numbers refer to the mouse protein. This figure was generated with the program ESPript (37). Blue stars, Ala substitutions of these residues abolish
Structure and multiple sequence alignment of the mouse PNGase C-terminal domain. (A) Ribbon representation of the crystal structure. The front layer
www.pnas.org?cgi?doi?10.1073?pnas.0602954103Zhou et al.
PNGase, nor interacts with chitobiose as one would expect based
on the close structural relationship between the core domain of
mouse PNGase and full-length yeast PNGase (9, 10).
Because ?3,?6-mannotriose has been reported to inhibit the
function of mouse PNGase (11), it was also used as substrate in the
ITC binding assay. Mannotriose showed a binding affinity identical
to that of the C-terminal domain of mouse PNGase as mannopen-
taose (data not shown), indicating that this branched structure may
be the minimal binding unit required for interactions with this
domain. Mouse PNGase may mainly recognize the high-mannose
type of oligosaccharide substrates at the second branch site, which
the complex structure described below.
Mannose-Binding Site of the Mouse PNGase C-Terminal Domain.
Based on the conservation of solvent-exposed residues in the
C-terminal domain and its structural similarity with the sugar-
binding domain of Fbs1, site-directed mutagenesis was carried out
to probe the role of selected residues in oligomannose binding.
These residues were individually replaced with Ala, and their
mannopentaose-binding affinities were investigated by using ITC
web site). These studies revealed that the K527A, E529A, W532A,
Y536A, W624A, Q625A, Q628A, and R631A substitutions abol-
ished binding of the mouse PNGase C-terminal domain to man-
nopentaose, whereas the F525A and E541A mutants resulted in
slightly reduced binding affinities. Two additional substitutions,
K530A and K540A, at residues that are not highly conserved
revealed no detectable effects on binding. To confirm that the
substitutions do not affect the overall structure of the C-terminal
domain, the corresponding variants were analyzed by CD spectros-
copy (Fig. 6A, which is published as supporting information on the
PNAS web site).
From an analysis of the crystal structure, it became clear that all
of these residues are located on the concave part of the saddle,
except Lys-530 and Lys-540, which are pointing away from the
groove (Fig. 2C). The oligomannose-binding site is apparently
formed by ?-strands 8, 9, and 16, which provide the concave part
and the second 310helix, which form the ridges on either side of the
saddle. Trp-532 resides in the loop between ?-strands 8 and 9 and
Trp-624 in the 310-2 helix, and these residues are on either side of
the binding site.
Structure of the Mouse PNGase C-Terminal Domain in Complex with
Mannopentaose. The crystal structure of this domain in complex
with mannopentaose in space group P21containing two molecules
(A and B) in the asymmetric unit was refined at 1.75-Å resolution
to an R factor of 0.172 (Rfreeof 0.208). Three of the five mannoses
(Man2–4) are well defined in the electron density maps (Fig. 3A),
whereas Man1 is rather flexible, and Man5 is completely disor-
dered. Man1 represents the reducing end of the mannopentaose
in hydrogen bonds with the C-terminal domain (Fig. 3B). Man2,
Man3, and Man4 lie within the binding groove and form several
hydrogen bonds with the protein. Man2 is located at the center of
the binding groove and hydrogen-bonds to the side chains of
Asp-531, Trp-532, and Gln-625. Man3 has the best defined density
and hydrogen-bonds to Glu-529, Gln-625, Gln-628, and Arg-631.
Man4 interacts only with Lys-527. In addition to these direct
protein–substrate interactions, there are water-mediated interac-
tions (Fig. 3B), which differ in number between the two complexes
present in the asymmetric unit. Compared with the apo structure,
the loop between ?8 and ?9 moved ?1.4 Å toward the binding
pocket. At the same time, rotations of the side chains of Asp-531
and Trp-532 in this loop allow interactions with the substrate.
Overall there is an excellent agreement between the cocrystal
structure and the binding studies involving altered residues (com-
pare Figs. 2C and 3B).
The C-Terminal Domain Enhances the Catalytic Activity of Mouse
PNGase. To investigate whether the C-terminal domain contributes
to the deglycosylation activity of mouse PNGase (Fig. 3C and Fig.
6B), we compared the enzymatic activity of full-length mouse
PNGase, mouse PNGase ?C (residues 1–450), and the mouse
PNGase K527A variant, a mutant in which the binding of the
(Lower) Fit of the experimental data (black squares) with a one-site binding model (thin line). (B) Schematic representation of the first seven residues of the
by ?-1,6-glycosidic linkages, whereas mannoses 2 and 4 and mannoses 1 and 5 are connected by ?-1,3-glycosidic bonds. (C) Close-up view into the putative
affinity to mannopentaose after mutation to Ala are colored in yellow, whereas those of residues that retain partial binding affinity are colored in blue. Carbon
atoms of residues that have no effect after substitution with Ala are shown in cyan.
Zhou et al.
November 14, 2006 ?
vol. 103 ?
no. 46 ?
ITC studies. RNase B, a high-mannose type N-glycosylated protein
and well characterized PNGase substrate, was used as a substrate
(12). The assay showed that the half life (t1?2) of glycosylated RNase
it is ?80 min in the presence of mouse PNGase ?C. This dramatic
difference between the full-length protein and the mouse PNGase
?C truncation confirmed that the C-terminal domain is very
important for mouse PNGase activity. The mouse PNGase K527A
mutant displayed an intermediate activity in this assay with a t1?2of
?5 min. This finding indicates that although no binding to man-
nopentaose of the corresponding mouse PNGase C-terminal do-
main variant could be detected in our ITC experiments, a residual
the core domain. In conclusion, although the catalytic core domain
of mouse PNGase exhibits de-N-glycosylation activity in the ab-
accelerates substrate turnover.
Relative Arrangement of the C-Terminal and Catalytic Domains of
PNGase. Recently, the structure of the core domain of mouse
PNGase in complex with the XPC binding domain of HR23B has
been determined (10). This structure, especially the complex with
the inhibitor Z-VAD-fmk, provides additional information on how
PNGase recognizes its substrate. As demonstrated here, substrate
binding by mouse PNGase involves not only the core domain, but
also the C-terminal domain. Because the relative arrangement of
the two domains is important in understanding the substrate
specificity of PNGase from higher eukaryotes, possible domain
orientations were investigated by manually positioning the struc-
tures of the mouse PNGase core and C-terminal domains while at
the same time fulfilling two conditions: (i) The C terminus of the
core domain and the N terminus of the C-terminal domain have to
be in proximity because there are only four residues in between.
because both ends are flexible. (ii) Because the active site cysteine
(Cys-306) in the core domain has to be in close proximity to the
N-glycosidic bond of the substrate, the distance between its side
chain and the center of the mannose-binding motif in the C-
terminal domain was required to be within 25 Å, the maximum
length of four pyranoses.
In the resulting model (Fig. 4), a continuous cleft is visible that
the mannose residues omitted from molecule B. The carbon atoms of the
2B. (B) Stereoview of the hydrogen-bonded interactions between the C-
of interacting residues are shown in yellow, potential hydrogen bonds are
indicated as dashed lines in magenta, and water molecules are shown as red
spheres. (C) RNase B digestion by PNGase. Time course of the reactions
involving full-length mouse PNGase, mouse PNGase ?C (residues 1–450), and
The curves were obtained by densitometric analysis of SDS?PAGE gels and
represent the average of two experiments. The error bars indicate the result-
ing standard deviations; however, in some instances, the standard deviations
are smaller than the geometric symbols that represent the data points.
Mannopentaose binding activity of the C-terminal domain. (A)
in green and the C-terminal domain in cyan. In the core domain, residues of
the catalytic triad are colored in red, and other conserved residues in the
binding pocket are colored in light orange. The mannose-binding residues in
the C-terminal domain are displayed in yellow. The C terminus of the core
domain is labeled C, and N terminus of the C-terminal domain is labeled N. (B)
Close-up view in the same orientation as in A, with the protein in a surface
those of the mannotetraose are colored in green. The carbon atoms of the
PNGase inhibitor Z-VAD-fmk are shown in magenta.
Hypothetical model of the relative arrangements of the core and
www.pnas.org?cgi?doi?10.1073?pnas.0602954103 Zhou et al.