Crystal structure of a designed tetratricopeptide repeat
module in complex with its peptide ligand
Aitziber L. Cortajarena
1
, Jimin Wang
1
and Lynne Regan
1,2
1 Department of Molecular Biophysics & Biochemistry, Yale University, New Haven, CT, USA
2 Department of Chemistry, Yale University, New Haven, CT, USA
Introduction
The basic tetratricopeptide (TPR) repeat comprises 34
amino acids that adopt a helix–turn–helix structure
[1,2]. We refer to the two tandem helices as the A-helix
and B-helix. In tandem arrays of TPR repeats, the
helices stack to form superhelical structures that dis-
play two surfaces: a concave binding face, and a con-
vex back face. The natural role of TPR proteins is to
mediate protein–protein interactions. Modules with
three tandem TPR repeats are by far the most com-
mon in nature, and presumably represent the minimal
functional binding unit [1]. The simple modular nature
of TPR proteins makes them ideal scaffolds for protein
design studies.
We designed a TPR protein, named CTPR3, com-
posed of three repeats of a consensus TPR sequence,
and solved its crystal structure at 1.6 A
˚
resolution [3].
The structural alignment of CTPR3 with natural
3-TPR domains clearly shows that its overall structure

Biophysics & Biochemistry, Yale University,
New Haven, CT 06520, USA
Fax: +1 203 432 5175
Tel: +1 203 432 9843
E-mail:
Website: />(Received 26 October 2009, revised 25
November 2009, accepted 16 December
2009)
doi:10.1111/j.1742-4658.2009.07549.x
Tetratricopeptide repeats (TPRs) are protein domains that mediate key
protein–protein interactions in cells. Several TPR domains bind the C-ter-
mini of the chaperones heat shock protein (Hsp)90 and ⁄ or Hsp70, and
exchange of such binding partners is key for the heat shock response. We
have previously described the design of a TPR protein that binds tightly
and specifically to the C-terminus of Hsp90, and in doing so, is able to
inhibit chaperone function in vivo. Here we present the X-ray crystal struc-
ture of the designed TPR domain (CTPR390) in complex with its peptide
ligand – the C-terminal residues of Hsp90 (peptide MEEVD). This struc-
ture reveals two interesting aspects of the TPR modules. First, a new pack-
ing arrangement of 3-TPR modules is observed. The TPR units stack
against each other in an unusual fashion to form infinite superhelices in the
crystal. Second, the structure provides insights into the molecular basis of
TPR–ligand recognition.
Abbreviations
ASU, asymmetric unit; Hsp, heat shock protein; TPR, tetratricopeptide repeat.
1058 FEBS Journal 277 (2010) 1058–1066 ª 2010 The Authors Journal compilation ª 2010 FEBS
mammalian cells, we showed that it inhibited Hsp90
function, presumably by preventing Hsp90 from form-
ing a complex with the TPR2A domain of Hsp-
organizing protein (HOP) [6].

model are good (Table 2), with 98.2% of all nongly-
cine residues located in the ‘most favorable’ region and
the remaining 1.8% nonglycine residues located in the
‘additionally allowed’ regions of the Ramachandran
plot.
Crystal packing – head-to-tail packing
The parent protein, CTPR3, crystallized as a monomer
with two molecules in the ASU. It was therefore some-
what surprising to find that CTPR390 forms ordered
superhelical structures in the crystal (Fig. 1B–D).
A superhelical arrangement has been previously
observed in the crystal forms of CTPR8 and CTPR20
[7]. The packing in CTPR390 crystals, however, is dif-
ferent. The CTPR390 units stack head to tail and form
continuous pseudoinfinite crystalline helical ‘fibers’,
which are arranged in a hexagonal symmetry lattice
(Fig. 2A,B). In the CTPR8 and CTPR20 crystal forms,
the ASU was composed of only part of the molecule
(two or four repeats), so the ends of the molecules
could not be located in the electron density map, and
the full-length structures were reconstructed by apply-
ing crystal symmetry and unit cell translations [7]. By
contrast, with CTPR390, we observed five molecules in
the ASU, and the discontinuity in the electron density
that defines the end of each CTPR390 molecule was
clear, allowing us to place the five individual units in
the ASU (Fig. 1B,D). Each CTPR390 unit is com-
posed of three TPR repeats (AB-helix pair) and an
additional C-terminal capping helix (A
cap

˚
) 1.1001
Resolution (A
˚
) 50–2.85 (2.95–2.85)
R
merge
(%)
a
7.5 (39.7)
I ⁄ rI
a
21.18 (1.16)
Completeness (%)
a
99.4 (99.7)
Redundancy
a
5.18 (5.28)
v
2a
1.180 (0.928)
Total reflections 28 406
Unique reflections 13 180
a
Values in parentheses correspond to the highest-resolution bin.
Table 2. Model refinement statistics.
CTPR390–Hsp90
Resolution (A
˚

superhelices formed by the CTPR proteins and the
superhelix formed by the TPR domain of the enzyme
O-linked GlcNAc transferase [8], showing that the two
superhelices are similar [7]. The superhelix in
CTPR390, even though it is similar to that previously
observed in CTPR8 and CTPR20, is more compressed,
and presents a larger curvature, with one fewer repeat
per superhelical turn. These differences are clear when
the first three repeats of the CTPR390 superhelix are
superimposed onto the three N-terminal repeats of
CTPR8, as shown in Fig. 3C. The N-terminal repeats
align well, with an rmsd value of 0.897 A
˚
, but because
of the differences in the superhelical twist, the two
structures differ more and they do not overlap well
towards the C-terminal repeats. The fact that 3-TPR
units from CTPR390 align well with 3-TPR units of
CTPR8 or CTPR20 indicates that, rather than the
inter-repeat packing, the intermolecular packing is
probably responsible for the pitch and diameter differ-
ences between the two structures.
Structure of individual CTPR390 molecules
Considering the individual 3-TPR units, the structure
of CTPR390 is almost identical to the structure of the
parent protein, CTPR3 [3]. CTPR3 is the consensus
protein, which contains no binding residues. CTPR390
has Hsp90-specific residues ‘grafted’ onto the binding
surface of CTPR3 [5]. The pairwise backbone align-
ment of CTPR3 (Protein Data Bank ID: 1Na0) and

Fig. 1. Crystal structure of CTPR390–Hsp90 peptide complex. (A)
The ASU is shown in ribbon representation, with each CTPR390
unit colored differently (chain A, green; chain B, cyan; chain C,
magenta; chain D, yellow; chain E, orange). The chains are labeled
in the figure with their identification letters. The five Hsp90 peptide
ligands (G, H, I, J, and K) are shown as gray ribbons. (B) Ribbon
representation of a superhelix formed by five CTPR390 subunits,
reconstructed by applying crystal symmetry and unit cell transla-
tions. The color code for the different CTPR390 chains is the same
as in (A). (C) Axial view of the superhelix in (B). (D) Schematic rep-
resentation of the CTPR390 subunits packing in the infinite supe-
rhelices in the crystal form [same color code as in (A–C)].
Structure of designed TPR module–ligand complex A. L. Cortajarena et al.
1060 FEBS Journal 277 (2010) 1058–1066 ª 2010 The Authors Journal compilation ª 2010 FEBS
CTPR390 has an rmsd value of 0.738 A
˚
(Fig. 3D).
When we calculate pairwise alignments of CTPR390
molecules within the ASU, we obtain rmsd values in
the range 0.433–0.682 A
˚
, only slightly smaller than the
values observed for the CTPR390–CTPR3 compari-
son. The conformation of CTPR390 with the Hsp90
peptide ligand bound is thus very similar to the
ligand-free CTPR3 structure. This result lends strong
support to our hypothesis that CTPR3 is a stable
framework onto which we can introduce mutations to
change the binding specificity without affecting the
structure of the protein. In addition, this result con-

bon) and CTPR8 (blue ribbon) superhelices. Backbone alignment of
the first three N-terminal repeats of CTPR8 and CTPR390 chain C.
The N-termini and C-termini of the superhelices are labeled. The
A-helix and B-helix of the first repeat are also labeled. (D) Pairwise
alignment of the CTPR390 structure (chain C in magenta) and the
CTPR3 structure (Protein Data Bank ID: 1Na0 in blue). The N-ter-
mini and C-termini of the proteins and the A-helices and B-helices
of the three repeats are labeled.
AB
CD
Fig. 4. X-ray crystal structure of CTPR390 in complex with the
C-terminal peptide of Hsp90. (A) CTPR390–Hsp90 complex (protein
chain C and peptide chain I). The backbone of CTPR390 is shown
as a ribbon representation, and the side chains of the TPR residues,
which directly interact with the peptide, are displayed as yellow
sticks. The C-terminal Hsp90 peptide is shown as sticks in purple.
(B) 2F
o
– F
o
electron density maps for two of the peptide chains in
the ASU: peptide chain I. (C) Overlay of the five peptide chains
(G, H, I, J and K chains). The peptide backbones are aligned, giving
an rmsd value of 0.298 A
˚
. (D) Overlay of two peptide chains (I in
magenta, and J in yellow) bound to two CTPR390 molecules in the
ASU (C and D, respectively). The two views are related by 90° rota-
tion about a vertical (y) axis. Only the protein chain backbones, and
not the peptide chains, were overlayed, giving an rmsd value of

different peptide chains relative to the TPR domains.
This result implies that the peptide chains can reorient
as rigid bodies in the binding pocket. The average
B-factor for atoms in the peptides is higher than the
average B-factor for atoms in the protein (Table 2),
which again may be a reflection of the mobility of the
peptide chain in the binding pocket. This conforma-
tional variability could exist because not all of the
TPR–peptide interactions that are seen in the TPR2A–
peptide complex are reproduced in the CTPR3–peptide
complex. Such interactions are discussed in detail in
the next section [5].
Atomic details of the CTPR390–Hsp90 interaction
Analysis of the detailed interactions in the CTPR390–
Hsp90 complex is presented for one of the complexes
in the ASU: chains C (TPR) and I (peptide) (Fig. 4A),
for which the electron density for the peptide is the
clearest, and the confidence in the conformation of the
peptide within the complex is the highest.
The dissociation constant for the CTPR390–MEE-
VD interaction is  200 lm [5], whereas the dissocia-
tion constant of the TPR2A–MEEVD interaction
 11 lm [9]. A comparison of the cocrystal structures
of TPR2A and CTPR390 in complex with the MEE-
VD peptide provides an explanation for the lower
affinity of the designed protein.
The backbone overlay of CTPR390 and TPR2A
protein chains gives an rmsd value of 1.632 A
˚
, and

˚
(Fig. 5A), as compared with the rmsd value of 1.632 A
˚
when the entire TPR domains are aligned. When the
‘clamp residues’ are aligned, the superposition of the
two Hsp90 peptides shows that the C-terminal residues
of the peptides align reasonably well and present the
same overall conformation. At the N-terminus of the
peptide, the alignment diverges more, with the major
difference being that the N-terminal Met is signifi-
cantly further away from the binding cleft in the
CTPR390–MEEVD complex than in the TPR2A–
MEEVD complex (Fig. 5A).
Figure 5B,C shows detailed schematic diagrams of
the TPR–ligand interactions for CTPR390 and
TPR2A, respectively, generated using ligplot [10]. The
electrostatic interactions and hydrogen-bonding inter-
actions mediated by the conserved carboxylate clamp
residues for the TPR2A–Hsp90 and CTPR390–Hsp90
complexes are tabulated and compared in Table 3.
The CTPR390–Hsp90 complex reproduces most of
the key interactions present in the TPR2A–Hsp90
complex. In the CTPR390–Hsp90 structure, the water
molecules cannot be located clearly, so the interactions
present in the TPR2A–Hsp90 complex mediated by
water molecules could not be placed in the CTPR390–
Hsp90 complex (which does not mean that they are
not present). Additionally, for most of the interactions,
the distances between the interacting atoms are greater
in the CTPR390–Hsp90 complex than in the TPR2A–

˚
2
).
Met1 of the Hsp90 peptide is also engaged in tight
hydrophobic interactions with a cavity mainly formed
by the side chains of Tyr236 and Glu271 of TPR2A
(Fig. 5C). However, in the CTPR390–Hsp90 complex,
although an equivalent Tyr is present (Tyr55), there is
a Lys (Lys55) at the Glu271 position that pushes the
Met outside of the binding pocket. Therefore, Met
does not contribute to the binding, resulting in a
weaker binding affinity (Fig. 5A,B).
A comparison of the average B-factors for the resi-
dues in the MEEVD peptide show that the C-terminal
Asp has a B-factor of 84, whereas the N-terminal Met
has a B-factor of 125. These values provide additional
support for the notion that the Met is not engaged in
specific interactions with the protein. Therefore, the
Met probably has more conformational flexibility than
A
B
C
Fig. 5. CTPR390–Hsp90 interactions and comparison with the
TPR2A–Hsp90 complex. (A) Overlay of the five carboxylate clamp
residues of the CTPR390–Hsp90 (magenta) and TPR2A–Hsp90
(blue) complexes. The side chains of the protein residues and the
two Hsp90 peptides are shown in stick representation. The identi-
ties of the residues in both the CTPR390 (top) and TPR2A (bottom)
domains and the N-termini and C-termini of the peptides are indi-
cated. (B) Schematic 2D diagram of CTPR390–Hsp90 peptide inter-

Hsp90 complex from 11 lm to 90 lm [9]. Therefore,
the lack of this interaction in the CTPR390–peptide
complex will partially contribute to the moderately
weak binding affinity of the designed TPR module.
Discussion
In this article, we present the cocrystal structure of a
designed TPR domain with its partner peptide.
We show that this 3-TPR domain can adopt a
superhelical structure in the crystal similar to those
reported for long TPR arrays [7]. This result illustrates
the natural tendency of TPR domains to stack head to
tail and self-assemble into an ordered macrostructure
in crystals. We have seen no evidence for such associa-
tion in solution.
We previously showed that, by grafting the binding
residues from a given natural TPR domain onto a con-
sensus scaffold, we could incorporate the binding
activity in the newly designed domain. This structure
proves that the new domain obtained using this ‘graft-
ing’ strategy mimics not only the binding activity [5,6],
but also the interactions at a molecular level between
the protein and the ligand. This result confirms the
TPR domains as a stable protein scaffold where, by
grafting the binding residues, one can interchange the
binding activities between domains.
Additionally, this work allows us to compare the
structure of the consensus CTPR3 domain without
ligand and the designed CTPR390 (with a total of only
12 mutations relative to the parent CTPR3) with
ligand bound. These two structures overlap almost per-

˚
) Residue in TPR Residue in peptide Distance (A
˚
)
K229 D5 (OXT) 2.68 K13 D5 (OXT) 3.09
N233 D5 (OXT) 2.83 N17 D5 (OXT) 3.99
N264 D5 (OXT) 2.83 N48 D5 (OXT) 2.94
D5 (NH) 2.96 D5 (NH) 3.18
H
2
O–D5 (OD2) 2.68–3.03 – –
K301 D5 (OD1) 2.63 K78 D5 (OD1) 2.87
D5 (OD2) 3.04
R305 E3 (O) 2.73 R82 E3 (O) 2.64
H
2
O–E3 (NH) 3.13–2.71 – –
E2 (OE1) 2.78 E2 (OE1) 2.79
E2 (OE1) 3.10
Structure of designed TPR module–ligand complex A. L. Cortajarena et al.
1064 FEBS Journal 277 (2010) 1058–1066 ª 2010 The Authors Journal compilation ª 2010 FEBS
CTPR3 scaffold, DY DEAIEYYQKALEL. Underlining indi-
cates the solvent-exposed charged residues [5].
Cloning of the CTPR390 gene
The gene encoding CTPR390 was constructed as previously
described and cloned into the pProEx-HTA vector to incor-
porate a cleavable N-terminal His-tag (GibcoBRL,
Gaithersburg, MD, USA) [5,12]. The identity of the
construct was verified by DNA sequencing (W.M. Keck
Facility, Yale University, New Haven, CT, USA).

2
PO
4
, 20% (w ⁄ v) poly(ethylene
glycol) 20000, and 50 mm Caps (pH 10.0). The well solution
was mixed in equal volumes (2 lL) with a protein–peptide
complex solution (1 : 4 molar ratio) at 30 mgÆmL
)1
protein
concentration. Crystals appeared within a week at 20 °C,
and reached sizes of approximately 80 · 80 · 50 lm within
2 weeks. Crystals were flash-cooled under a nitrogen gas
stream (100 K). Data were collected to 2.85 A
˚
resolution at
the NSLS beamline X12C (Brookhaven National Labora-
tory). The data collection statistics are shown in Table 1.
Structure determination and refinement
We used hkl2000 [15] to index, scale and integrate the
data. The protein crystallized in space group R3 with unit
cell dimensions of a = b = 100.67 A
˚
, c = 161.57 A
˚
, and
a = b =90°, c = 120°. The CTPR390 structure was
solved by molecular replacement using molrep [16] in the
ccp4i suite [17]. The structure of the consensus TPR with-
out the solvating helix was used as search model [CTPR3
(Protein Data Bank ID: 1NA0] [3]. There were five TPR

) = 27.1 (28.2). The geometry and stereochemical
properties of the model were checked with molprobity [22].
Crystallographic statistics are shown in Table 2.
Coordinates
The X-ray structure of the CTPR390–Hsp90 peptide com-
plex has been deposited in the Protein Data Bank as 3KD7.
Acknowledgements
We thank members of staff at NSLS beamlines X12C
and X6A, BNL, where data were collected. The high-
throughput crystal screening service of the Hauptman-
Woodward facility assisted in identifying initial
crystallization conditions. We thank T. Kajander for
his advice during the crystallization process and data
collection. We thank staff members and users of the
Yale Center for Structural Biology for valuable
insights during the structure-solving and refinement
process. We thank R. Collins, T. Grove, R. Ilagan, M.
Jackrel, L. Kundrat and G. Pimienta-Rosales for valu-
able discussions and comments on the manuscript.
References
1 D’Andrea L & Regan L (2003) TPR proteins: the versa-
tile helix. Trends Biochem Sci 28, 655–662.
A. L. Cortajarena et al. Structure of designed TPR module–ligand complex
FEBS Journal 277 (2010) 1058–1066 ª 2010 The Authors Journal compilation ª 2010 FEBS 1065
2 Blatch GL & Lassle M (1999) The tetratricopeptide
repeat: a structural motif mediating protein–protein
interactions. Bioessays 21, 932–939.
3 Main ERG, Xiong Y, Cocco MJ, D’Andrea L & Regan
L (2003) Design of stable alpha-helical arrays from an
idealized TPR motif. Structure 11, 497–508.

Consensus design as a tool for engineering repeat
proteins. Methods Mol Biol 340, 151–170.
13 Pace CN, Vajdos F, Fee L, Grimsley G & Gray T
(1995) How to measure and predict the molar absorp-
tion coefficient of a protein. Prot Sci 4, 2411–2423.
14 Luft JR, Collins RJ, Fehrman NA, Lauricella AM,
Veatch CK & DeTitta GT (2003) A deliberate
approach to screening for initial crystallization condi-
tions of biological macromolecules. J Struct Biol 142,
170–179.
15 Otwinowski Z & Minor W (1997) Processing of X-ray
diffraction data collected in oscillation mode. Methods
Enzymol 276, 307–326.
16 Vagin AA & Teplyakov A (1997) MOLREP: an auto-
mated program for molecular replacement. J Appl Crys-
tallogr 30, 1022–1025.
17 Collaborative Computational Project, Number 4
(1994) The CCP4 suite: programs for protein crystal-
lography. Acta Crystallogr D 50, 760–763.
18 Bru
¨
nger AT, Adams PD, Clore GM, DeLano WL,
Gros P, Grosse-Kunstleve RW, Jiang JS, Kuszewski J,
Nilges M, Pannu NS et al. (1998) Crystallography &
NMR system: a new software suite for macromolecular
structure determination. Acta Crystallogr D 54,
905–921.
19 Murshudov G, Vagin A & Dodson E (1997) Refinement
of macromolecular structures by the maximum-likeli-
hood method. Acta Crystallogr D 53, 240–255.