Voet D., Voet Ju.G. Biochemistry

Подождите немного. Документ загружается.

these, ⬃1200 have already been observed, with around

half of the proteins of known structure belonging to only

20 fold groups.

Although there are numerous ways in which domain

structures might be categorized (see, e.g., Section 8-3C),

perhaps the simplest way to do so is to classify them as

␣ domains (containing secondary structural elements that

are exclusively ␣ helices), ␤ domains (containing only

␤ sheets), and ␣/␤ domains (containing both ␣ helices and

␤ sheets).The ␣/␤ domain category may be further divided

into two main groups: ␣/␤ barrels and open ␤ sheets. In the

following paragraphs we describe some of the most com-

mon folds in each of these domain categories.

f. ␣ Domains

We are already familiar with a fold that contains only

␣ helices: the globin fold, which contains 8 helices in two

layers and occurs in myoglobin (Fig. 8-39) and in both the

␣ and ␤ chains of hemoglobin (Section 10-2B). In another

common all-␣ fold, two ␣␣ motifs combine to form a 4-helix

bundle such as occurs in cytochrome b

562

(Fig. 8-47a, oppo-

site). The helices in this fold are inclined such that their

contacting side chains intermesh and are therefore out of

contact with the surrounding aqueous solution. They are

consequently largely hydrophobic. The 4-helix bundle is a

relatively common fold that occurs in a variety of proteins.

However, not all of them have the up-down-up-down

topology (connectivity) of cytochrome b

562

. For example,

human growth hormone is a 4-helix bundle with up-up-

down-down topology (Fig. 8-47b). The successive parallel

helices in this fold are, of necessity, joined by longer loops

than those joining successive antiparallel helices.

Different types of ␣ domains occur in transmembrane

proteins. We study these proteins in Section 12-3A.

g. ␤ Domains

␤ Domains contain from 4 to ⬎10 predominantly an-

tiparallel ␤ strands that are arranged into two sheets that

pack against each other to form a ␤ sandwich. For example,

the immunoglobulin fold (Fig. 8-48), which forms the basic

domain structure of most immune system proteins (Section

35-2Be), consists of a 4-stranded antiparallel ␤ sheet in

face-to-face contact with a 3-stranded antiparallel ␤ sheet.

Note that the strands in the two sheets are not parallel to

one another, a characteristic of stacked ␤ sheets. The side

chains between the two stacked ␤ sheets are out of contact

with the aqueous medium and thereby form the domain’s

hydrophobic core. Since successive residues in a ␤ strand

alternately extend to opposite sides of the ␤ sheet (Fig.

8-17), these residues are alternately hydrophobic and hy-

drophilic.

The inherent curvature of ␤ sheets (Section 8-1C) often

causes sheets of more than 6 strands to roll up into ␤ barrels.

Indeed, ␤ sandwiches may be considered to be flattened

␤ barrels. Several different ␤ barrel topologies have been

observed, most commonly:

1. The up-and-down ␤ barrel, which consists of 8 suc-

cessive antiparallel ␤ strands that are arranged like the

staves of a barrel.An example of an up-and-down ␤ barrel

Section 8-3. Globular Proteins 251

Figure 8-48 The immunoglobulin fold. The X-ray structure of

the N-terminal domain of the human immunoglobulin fragment

Fab New shows its immunoglobulin fold.The peptide backbone

of this 103-residue domain is drawn in ribbon form colored in

rainbow order from N-terminus (red) to C-terminus (blue) with

its ␤ strands represented by flat arrows pointing toward the

C-terminus.The inset is the topological diagram of the

immunoglobulin fold showing the connectivity of its stacked

4-stranded and 3-stranded antiparallel ␤ sheets. [Based on an

X-ray structure by Roberto Poljak, Johns Hopkins School of

Medicine. PDBid 7FAB.]

JWCL281_c08_221-277.qxd 2/23/10 1:58 PM Page 251

occurs in retinol binding protein (Fig 8-49), which functions

to transport the nonpolar visual pigment precursor retinol

(vitamin A) in the bloodstream:

2. A fold consisting of two Greek key motifs, which

thereby constitutes an alternative way of connecting the

strands of an 8-stranded antiparallel ␤ barrel. Figure 8-50

indicates how two Greek key motifs in the C-terminal do-

main of the eye lens protein ␥-B crystallin are arranged to

form a ␤ barrel.

3. The jelly roll or Swiss roll barrel (which is named for

its topological resemblance to these rolled up pastries), in

which a 4-segment ␤ hairpin is rolled up into an 8-stranded

antiparallel ␤ barrel of yet different topology, as is dia-

grammed in Fig. 8-51a. The X-ray structure of the enzyme

peptide-N

-(N-acetyl-␤-D-glucosaminyl)asparagine ami-

dase F contains a domain consisting of a jelly roll barrel

(Fig. 8-51b).

Retinol

h. ␣/␤ Barrels

In ␣/␤ domains, a central parallel or mixed ␤ sheet is

flanked by ␣ helices. The ␣/␤ barrel, which is diagrammed

in Fig. 8-19b, is a remarkably regular structure that consists

of 8 tandem ␤␣ units (essentially 8 overlapping ␤␣␤ mo-

tifs) wound in a right-handed helical sense to form an inner

8-stranded parallel ␤ barrel concentric with an outer barrel

of 8 ␣ helices. Each ␤ strand is approximately antiparallel

to the succeeding ␣ helix and all are inclined at around the

same angle to the barrel axis. Figure 8-52 shows the X-ray

structure of chicken triose phosphate isomerase (TIM), de-

termined by David Phillips, which consists of an ␣/␤ barrel.

This is the first known structure of an ␣/␤ barrel, which is

therefore also called a TIM barrel.

The side chains that point inward from the ␣ helices in-

terdigitate with the side chains that point outward from the

␤ strands. A large fraction (⬃40%) of these side chains are

those of the branched aliphatic residues Ile, Leu, and Val.

The side chains that point inward from the ␤ strands tend

to be bulky and hence fill the core of the ␤ barrel (contrary

to the impression that Figs. 8-19b and 8-52 might provide,

␣/␤ barrels, with one known exception, do not have hollow

cores). Those side chains that fill the ends of the barrel are

in contact with solvent and hence tend to be polar, whereas

those in its center are out of contact with solvent and are

therefore nonpolar. Thus, ␣/␤ barrels have four layers of

polypeptide backbone interleaved by regions of hydrophobic

252 Chapter 8. Three-Dimensional Structures of Proteins

Figure 8-49 Retinol binding protein. Its X-ray structure shows

its up-and-down ␤ barrel (residues 1⫺142 of this 182-residue

protein). Its peptide backbone is drawn in ribbon form colored in

rainbow order from N-terminus (red) to C-terminus (blue). Note

that each ␤ strand is linked via a short loop to its clockwise-adjacent

strand as seen from the top. The protein’s bound retinol molecule

is represented by a gray ball-and-stick model.The inset is the

protein’s topological diagram. [Based on an X-ray structure by T.

Alwyn Jones, Uppsala University, Uppsala, Sweden. PDBid

1RBP.]

JWCL281_c08_221-277.qxd 2/23/10 1:59 PM Page 252

Section 8-3. Globular Proteins 253

Figure 8-50 X-ray structure of the C-terminal domain of

bovine ␥-B crystallin. (a) A topological diagram showing how its

two Greek key motifs are arranged in a ␤ barrel. One Greek key

motif (red) is formed by ␤ strands 1 to 4 and the other (blue) is

formed by ␤ strands 5 to 8. [After Branden, C. and Tooze, J.,

Introduction to Protein Structure (2nd ed.), p. 75, Garland (1999).]

(b) The 83-residue peptide backbone displayed in ribbon form.

fold

7812 3456

Here the members of an antiparallel pair of ␤ strands in a Greek

key motif are colored alike with the N-terminal Greek key colored

red (strands 1 & 4) and orange (2 & 3) and the C-terminal Greek

key colored blue (5 & 8) and cyan (6 & 7).The N-terminal

domain of this two-domain protein is nearly superimposable on

its C-terminal domain. [Based on an X-ray structure by Tom

Blundell, Birkbeck College, London, U.K. PDBid 4GCR.]

Figure 8-51 X-ray structure of the enzyme peptide-N

(N-acetyl-␤-

D-glucosaminyl)asparagine amidase F from

Flavobacterium meningosepticum. (a) A diagram indicating how

an 8-stranded ␤ barrel is formed by rolling up a 4-segment ␤

hairpin.A topological diagram of the jelly roll barrel is also shown.

[After Branden, C. and Tooze, J., Introduction to Protein Structure

(2nd ed.), pp. 77–78, Garland (1999).] (b) A ribbon diagram of the

domain formed by residues 1 to 140 of this 314-residue enzyme.

Here the two ␤ strands in each segment of the ␤ hairpin are

colored alike, with strands 1 & 8 (the N- and C-terminal strands)

red, strands 2 & 7 orange, strands 3 & 6 cyan, and strands 4 & 5

blue. [Based on an X-ray structure by Patrick Van Roey, New York

State Department of Health,Albany, New York. PDBid 1PNG.]

(a)

(b)

(a)

(b)

JWCL281_c08_221-277.qxd 2/23/10 1:59 PM Page 253

side chains. In contrast, both ␣ domains and ␤ domains

consist of two layers of polypeptide backbone sandwiching

a hydrophobic core.

Around 10% of known enzyme structures contain an

␣/␤ barrel, making it the most common fold assumed by

enzymes. Moreover, nearly all known ␣/␤ barrel proteins

are enzymes. Intriguingly, the active sites of ␣/␤ barrel en-

zymes are almost all located in funnel-shaped pockets

formed by the loops that link the C-termini of the ␤ strands

to the succeeding ␣ helices and hence surround the mouth

of the ␤ barrel, an arrangement that has no obvious struc-

tural rationale. Thus, despite the observation that few ␣/␤

barrel proteins exhibit significant sequence homology, it

has been postulated that all of them descended from a

common ancestor and are therefore (distantly) related by

divergent evolution. On the other hand, it has been argued

that the ␣/␤ barrel is structurally so well suited for its enzy-

matic roles that it independently arose several times and

hence that ␣/␤ enzymes are related by convergent evolu-

tion (i.e., nature has discovered the same fold on several

occasions). Convincing evidence supporting either view

has not been forthcoming, so that the nature of the evolu-

tionary relationships among the ␣/␤ barrel enzymes re-

mains an open question.

i. Open ␤ Sheets

We have previously encountered examples of an open ␤

sheet in the structures of carboxypeptidase A (Fig. 8-19a)

and the N-terminal domain of glyceraldehyde-3-phosphate

dehydrogenase (Fig. 8-45). Their topological diagrams are

drawn in Fig. 8-53.The X-ray structures and topological di-

agrams of two other such proteins, those of the enzymes

lactate dehydrogenase (N-terminal domain) and adenylate

kinase, are shown in Fig. 8-54. Such folds consist of a cen-

tral parallel or mixed ␤ sheet flanked on both sides by ␣ he-

lices that form the right-handed crossover connections be-

tween successive parallel ␤ strands (Fig. 8-20b).The strands

of such a ␤ sheet do not follow their order in the peptide

sequence. Rather, the ␤ sheet has a long crossover that re-

verses the direction of the succeeding section of sheet and

turns it upside down, thereby putting its helical crossovers

on the opposite side of the sheet from those of the preced-

ing section (Fig. 8-55). Such assemblies are therefore also

known as doubly wound sheets (in contrast to singly

wound ␣/␤ barrels, whose helices are all on the same side of

their ␤ sheets). Doubly wound sheets consist of three lay-

ers of polypeptide backbone interspersed by regions of hy-

drophobic side chains (in contrast to four such layers for

␣/␤ barrels and two for ␣ domains and ␤ domains). Note

that both types of domains containing parallel ␤ sheets are

hydrophobic on both sides of the sheet, whereas antiparal-

lel sheets are hydrophobic on only one side.This additional

stabilization of parallel ␤ sheets probably compensates for

the reduced strength of their nonlinear hydrogen bonds

relative to the linear hydrogen bonds of antiparallel ␤

sheets (Fig. 8-16).

254 Chapter 8. Three-Dimensional Structures of Proteins

Figure 8-53 Topological diagrams of (a) carboxypeptidase A

and (b) the N-terminal domain of glyceraldehyde-3-phosphate

dehydrogenase. The X-ray structures of these proteins are

diagrammed in Figs. 8-19a and 8-45.The thin black vertical

arrows mark the proteins’ topological switch points.

Figure 8-52 X-ray structure of the 247-residue enzyme triose

phosphate isomerase (TIM) from chicken muscle. The protein is

viewed approximately along the axis of its ␣/␤ barrel.The

peptide backbone is shown as a ribbon diagram with its successive

␤␣ units colored, from N- to C-terminus, in rainbow order, red to

blue. The inset is its topological diagram (with ␣ helices

represented by rectangles).Two other views of this protein are

drawn in Figures 8-19b and 8-19c [Based on an X-ray structure

by David Phillips, Oxford University, U.K. PDBid 1TIM.]

4123 8567

3 1

(a)

(b)

5 4

JWCL281_c08_221-277.qxd 2/23/10 1:59 PM Page 254

There are few geometric constraints on the number of

strands in open ␤ sheets; they have been observed to con-

tain from 4 to 10 ␤ strands, with 6 ␤ strands being the most

common. Since the position where the chain reverses its

winding direction, the so-called topological switch point,

can occur between any consecutive ␣ helix and ␤ strand,

doubly wound sheets can have many different folds. More-

over, some ␤ strands may run in an antiparallel direction to

yield mixed sheets (e.g., carboxypeptidase A; Figs. 8-19a

and 8-53a), and in several instances, there is more than one

topological switch point (e.g., adenylate kinase; Fig. 8-54b).

The open ␤ sheet is the most common domain structure

that occurs in globular proteins. Moreover, nearly all such

Section 8-3. Globular Proteins 255

Figure 8-54 X-ray structures of open ␤ sheet–containing

enzymes. (a) Dogfish lactate dehydrogenase, N-terminal domain

(residues 20–163 of this 330-residue protein), and (b) porcine

adenylate kinase (195 residues).The peptide backbones are

shown as ribbon diagrams with successive ␤␣ units colored, from

N- to C-terminus, in rainbow order, red to blue. In Part b,

structural elements that are not components of the open ␤ sheet

are gray. The insets are topological diagrams of these proteins,

with the thin black vertical arrows marking their topological

Figure 8-55 Doubly wound sheets. (a) A schematic diagram of

a doubly wound sheet indicating how the long crossover (green)

between its N- and C-terminal segments (red and blue) reverses

the direction of the sheet’s C-terminal segment and places its ␣

helical crossover connections on the opposite side of the sheet.

(b) The corresponding topological diagram, with the thin black

vertical arrow marking the topological switch point. [After

Branden, C. and Tooze, J., Introduction to Protein Structure

(2nd ed.), p. 49, Garland (1999).]

2 4 5 6

12 5

3,4

1,2

2,3

(b)

2,3

1,2

3,4

(a)

switch points. [Based on X-ray structures by (a) Michael

Rossmann, Purdue University, and (b) Georg Schulz, Institut für

Organische Chemie und Biochemie, Freiburg, Germany. PDBids

(a) 6LDH and (b) 3ADK.]

(a)

(b)

JWCL281_c08_221-277.qxd 2/23/10 1:59 PM Page 255

proteins are enzymes, many of which bind mono- or dinu-

cleotides. In fact, the fold exemplified by lactate dehydro-

genase (LDH; Fig. 8-54a) is known as the dinucleotide-

binding fold or Rossmann fold (after Michael Rossmann,

who first pointed out its significance). This is because

mononucleotide units are commonly bound by ␤␣␤␣␤

units, of which LDH (which binds the dinucleotide NAD

⫹

)

has two. In some proteins, the second ␣ helix in a ␤␣␤␣␤

unit is replaced by a length of nonhelical polypeptide. This

occurs, for example, between the ␤5 and ␤6 strands of

glyceraldehyde-3-phosphate dehydrogenase (Figs. 8-45

and 8-53b), which also binds NAD

⫹

At the topological switch point of an open ␤ sheet, the

loops emerging from the C-terminal ends of the flanking ␤

strands go to opposite sides of the sheet and thereby form

a crevice between them. In nearly all of the ⬎100 known

structures of open ␤ sheet–containing enzymes, as Carl-

Ivar Brändén first pointed out, this crevice forms at least

part of the enzyme’s active site.Thus the active sites of both

␣/␤ barrels and the open ␤ sheet enzymes are formed by

loops emanating from the C-terminal ends of the ␤ strands.

In contrast, the active sites of enzymes that have other

types of domain structures exhibit no apparent regularities

in the positions of their active sites.

C. Structural Bioinformatics

In Section 7-4, we discussed the rapidly developing field of

inquiry known as bioinformatics as it applies to the se-

quences of proteins and nucleic acids, that is, how sequence

alignments are determined and how phylogenetic trees are

generated. An equally important aspect of bioinformatics,

which we discuss below, is how macromolecular structures

are displayed and compared.

a. The Protein Data Bank

The atomic coordinates of most known macromolecular

structures are archived in the Protein Data Bank (PDB).

Indeed, most scientific journals that publish macromolecu-

lar structures require that authors deposit their structure’s

coordinates in the PDB. The PDB contains the atomic co-

ordinates of ⬃70,000 macromolecular structures (proteins,

nucleic acids, and carbohydrates as determined by X-ray

and other diffraction-based techniques, NMR, and electron

microscopy) and is growing exponentially at a rate that is

presently ⬃7500 structures per year. The PDB’s Website

(URL), from which these coordinates are publicly avail-

able, is listed in Table 8-4.

Each independently determined structure in the PDB is

assigned a unique four-character identifier (its PDBid) in

which the first character must be a digit (1–9) and no distinc-

tion is made between uppercase and lowercase letters (e.g.,

1 MBO is the PDBid of the myoglobin structure illustrated

in Fig. 8-39c, although PDBids do not necessarily have a re-

lationship to the corresponding macromolecule’s name). A

coordinate file begins with information that identifies/de-

scribes the macromolecule, the date the coordinate file was

deposited, its source (the organism from which it was ob-

tained),the author(s) who determined the structure,and key

journal references. The file continues with a synopsis of how

the structure was determined together with indicators of its

accuracy and information that could be helpful in its inter-

pretation, such as a description of its symmetry and which

residues, if any, were not observed. The sequences of the

256 Chapter 8. Three-Dimensional Structures of Proteins

Table 8-4 Structural Bioinformatics Websites (URLs)

Structural Databases

Protein Data Bank (PDB): http://www.rcsb.org/

Nucleic Acid Database: http://ndbserver.rutgers.edu/

Molecular Modeling Database (MMDB): http://www.ncbi.nlm.nih.gov/Structure/MMDB/mmdb.shtml

Most Representative NMR Structure in an Ensemble: http://www.ebi.ac.uk/msd-srv/olderado

PISA (Protein Interfaces, Surfaces and Assemblies): http://www.ebi.ac.uk/msd-srv/prot_int/pistart.html

Proteopedia: http://www.proteopedia.org

Molecular Graphics Programs/Plug-ins

Cn3D: http://www.ncbi.nlm.nih.gov/Structure/CN3D/cn3d.shtml

FirstGlance: http://molvis.sdsc.edu/fgij/

Jmol: http://jmol.sourceforge.net/

KiNG: http://kinemage.biochem.duke.edu/software/king.php

Swiss-Pdb Viewer: http://spdv.vital-it.ch

Structural Classification Algorithms

CATH (class, architecture, topology, and homologous superfamily): http://www.cathdb.info/index.html

CE (combinatorial extension of optimal pathway): http://cl.sdsc.edu/

Pfam (protein families): http://pfam.sanger.ac.uk/ or http://pfam.janelia.org/

SCOP (structural classification of proteins): http://scop.mrc-lmb.cam.ac.uk/scop/

VAST (vector alignment search tool): http://www.ncbi.nlm.nih.gov/Structure/VAST/

JWCL281_c08_221-277.qxd 2/23/10 1:59 PM Page 256

structure’s various chains are then listed together with the

descriptions and formulas of its so-called HET (for hetero-

gen) groups, which are molecular entities that are not among

the “standard” amino acid or nucleotide residues (e.g., or-

ganic molecules such as the heme group, nonstandard

residues such as Hyp, metal ions, and bound water mole-

cules). The positions of the structure’s secondary structural

elements and its disulfide bonds are then provided.

The bulk of a PDB file consists of a series of ATOM (for

“standard” residues) and HETATM (for heterogens)

records (lines), each of which provides the coordinates for

one atom in the structure. An ATOM or HETATM record

identifies its corresponding atom according to its serial

number (usually just its sequence in the list), atom name

(e.g., C and O for an amino acid residue’s carbonyl C and O

atoms, CA and CB for C

␣

and C

␤

atoms, N1 for atom N1 of

a nucleic acid base, C4* for atom C4¿ of a ribose or deoxyri-

bose residue), residue name [e.g., PHE, G (for a guanosine

residue), HEM (for a heme group), MG (for an Mg

2⫹

ion),

and HOH (for a water molecule)], chain identifier (e.g.,A,

B, C, etc., for structures consisting of more than one chain,

whether or not the chains are chemically identical), and the

residue sequence number in the chain. The record then

continues with the atom’s Cartesian (orthogonal) coordi-

nates (X,Y, Z), in angstroms relative to an arbitrary origin,

the atom’s occupancy (which is the fraction of sites that ac-

tually contain the atom in question, a quantity that is usu-

ally 1.00 but, for groups that have multiple conformations

or for molecules/ions that are only partially bound to a pro-

tein, may be a positive number less than 1.00), and its

isotropic temperature factor (a quantity that is indicative

of the atom’s thermal motion, with larger numbers denot-

ing a greater degree of motion). The ATOM records are

listed in the order of the residues in a chain. For NMR-

based structures, the PDB file contains a full set of ATOM

and HETATM records for each member of the ensemble

of structures that were calculated in solving the structure

(Section 8-3A; the most representative member of such a

coordinate set can be obtained from http://www.ebi.ac.uk/

msd-srv/olderado). PDB files usually end with CONECT

(connectivity) records, which denote the nonstandard

connectivities between ATOMs such as disulfide bonds

and hydrogen bonds as well as connectivities between

HETATMs.

A particular PDB file may be located according to its

PDBid or, if this is unknown, by searching with a variety of

criteria including a protein’s name, its source, the author(s),

keywords, and/or the experimental technique used to de-

termine the structure. Selecting a particular macromole-

cule in the PDB initially displays a Structure Summary

page with options for interactively viewing the structure,

for viewing or downloading the coordinate file, and for

classifying or analyzing the structure in terms of its geo-

metric properties and sequence (see below).

b. The Nucleic Acid Database

The Nucleic Acid Database (NDB) archives the atomic

coordinates of structures that contain nucleic acids. Its co-

ordinate files have substantially the same format as do

those of the PDB, where this information is also kept. How-

ever, the NDB’s organization and search algorithms are

specialized for dealing with nucleic acids. This is useful, in

part, because many nucleic acids of known structure are

identified only by their sequences rather than by names, as

are proteins (e.g., myoglobin), and consequently could eas-

ily be overlooked in a search of the PDB.

c. Viewing Macromolecular Structures in

Three Dimensions

The most informative way to examine a macromolecu-

lar structure is through the use of molecular graphics pro-

grams that permit the user to interactively rotate a macro-

molecule and thereby perceive its three-dimensional

structure. This impression may be further enhanced by si-

multaneously viewing the macromolecule in stereo. Most

molecular graphics programs use PDB files as input. The

programs described here can be downloaded from the In-

ternet addresses listed in Table 8-4, some of which also pro-

vide instructions for the program’s use.

Jmol, which functions as both a Web browser–based ap-

plet or as a standalone program, allows the user to display

user-selected macromolecules in a variety of colors and

formats (e.g., ball and stick, backbone, wireframe, space-

filling, and cartoon).The Interactive Exercises on the Web-

site that accompanies this textbook (http://wiley.com/

college/voet/) all use Jmol (this site also contains a Jmol tu-

torial). FirstGlance uses Jmol to display macromolecules

via a user-friendly interface. KiNG, which also has Web

browser–based and standalone versions, displays the so-

called Kinemages on this textbook’s accompanying Web-

site. KiNG provides a generally more author-directed user

environment than does Jmol. Macromolecules can be dis-

played directly from their corresponding PDB page using

Jmol, KiNG, and several other viewers. The Swiss-Pdb

Viewer (also called DeepView), in addition to displaying

molecular structures, provides tools for basic model build-

ing, homology modeling, energy minimization, and multi-

ple sequence alignment. One advantage of the Swiss-PDB

Viewer is that it allows users to easily superimpose two or

more models. Proteopedia is a 3D interactive encyclopedia

of proteins and other macromolecules that resembles

Wikipedia in that it is user edited. It uses mainly Jmol as a

viewer.

d. Structural Classification and Comparison

Most proteins are structurally related to other proteins.

Indeed, as we shall see in Section 9-6, evolution tends to

conserve the structures of proteins rather than their se-

quences. The computational tools described below facili-

tate the classification and comparison of protein structures.

They can be accessed directly via their websites (Table 8-4)

and, in some cases, accessed directly from the PDB. Studies

using these programs yield functional insights, reveal dis-

tant evolutionary relationships that are not apparent from

sequence comparisons (Section 7-4B), generate libraries of

unique folds for structure prediction, and provide indica-

tions as to why certain types of structures are preferred

over others.

Section 8-3. Globular Proteins 257

JWCL281_c08_221-277.qxd 2/23/10 1:59 PM Page 257

1. CATH (for Class, Architecture, Topology, and Homol-

ogous superfamily), as its name suggests, categorizes pro-

teins in a four-level structural hierarchy. (1) “Class,” the

highest level, places the selected protein in one of four cate-

gories of gross secondary structure: Mainly ␣, Mainly ␤, ␣/␤

(having both ␣ helices and ␤ sheets), and Few Secondary

Structures. (2) “Architecture” is the description of the gross

arrangement of secondary structure independent of topol-

ogy. (3) “Topology” is indicative of both the overall shape

and connectivity of the protein’s secondary structures. (4)

“Homologous superfamily” is those proteins of known

structure that are homologous to (share a common ancestor

with) the selected protein. For 1 MBO (sperm whale myo-

globin), the CATH classification is Class (C): Mainly alpha;

Architecture (A): Orthogonal bundle; Topology (T): Glo-

bin-like; and Homologous superfamily (H): Globin. CATH

permits the user to navigate up and down the various hierar-

chies and thereby structurally compare them.

2. CE (for Combinatorial Extension of the optimal

path) finds all proteins in the PDB that can be structurally

aligned with the query structure to within user-specified

geometric criteria. The amino acid sequences of these pro-

teins are also aligned on the basis of this structural align-

ment rather than sequence alignment (Section 7-4B). The

structurally aligned proteins can be simultaneously dis-

played by a Java applet called Compare3D that displays

both the aligned C

␣

backbones and the structurally aligned

sequences. CE can likewise align and display two user-

selected structures. The atomic coordinates of the aligned

structures can also be downloaded in PDB format for use

with other programs such as display by Jmol and KiNG.

3. Pfam (for Protein families) is a database of nearly

11,000 multiple sequence alignments of protein domains

(called Pfam families). Using Pfam, one can analyze a pro-

tein for Pfam matches (74% of proteins have at least one

match in Pfam), view Pfam family annotations and align-

ments, determine the domain organization of a protein

based on its sequence or its structure, find groups of related

Pfam families (called clans), examine the phylogenetic tree

of a Pfam family, and view the occurrence of a protein’s do-

mains across different species.

4. SCOP (Structural Classification Of Proteins) classi-

fies protein structures based mainly on manually generated

topological considerations according to a 6-level hierarchy:

Class [All-␣,All-␤, ␣/␤ (having ␣ helices and ␤ strands that

are largely interspersed), ␣⫹␤ (having ␣ helices and ␤

strands that are largely segregated), and Multi-domain

(having domains of different classes)], Fold (groups that

have similar arrangements of secondary structural ele-

ments), Superfamily (indicative of distant evolutionary

relationships based on structural criteria and functional

features), Family (indicative of near evolutionary relation-

ships based on sequence as well as on structure), Protein,

and Species. For 1 MBO these are Class: All-␣; Fold:

Globin-like; Superfamily: Globin-like; Family: Globins;

Protein: Myoglobin; and Species: Sperm whale (Physeter

catodon). SCOP permits the user to navigate through its

treelike hierarchical organization and lists the known

members of any particular branch. Thus, with 1MBO,

SCOP displays a list of the 174 structures in the PDB that

contain sperm whale myoglobin (a protein that has re-

ceived far more structural study than most).

5. VAST (Vector Alignment Search Tool),a component

of the National Center for Biotechnology Information

(NCBI) Entrez system, reports a precomputed list of pro-

teins of known structure that structurally resemble the

query protein. The VAST system uses the Molecular Mod-

eling Database (MMDB), an NCBI-generated database

that is derived from PDB coordinates but in which mole-

cules are represented by connectivity graphs rather than

sets of atomic coordinates.VAST displays the superposition

of the query protein in its structural alignment with a user-

selected list of the related proteins using Cn3D [a molecu-

lar graphics program that displays MMDB files and that is

publicly available for a variety of computer platforms

(Table 8-4)]. VAST also reports the structure-based se-

quence alignment of these proteins.

In addition, several “Structure Analysis” tools can be in-

voked from a PDB Structure Summary page. The “Se-

quence Details” page provides the sequence of each chain

in the structure and, for polypeptides, indicates the second-

ary structure of each of its residues.

e. The Structural Genomics Project

As we shall see in Section 9-3B, proteins with similar se-

quences are likely to have similar three-dimensional struc-

tures.Yet, of the ⬃7 million polypeptides of known sequence,

only ⬍40,000 have known structures comprising ⬃1200 of

the estimated 8000 different protein folds. Consequently,

⬃40% of the open reading frames (ORFs; DNA sequences

that appear to encode proteins) in known genomes specify

proteins whose structures and functions are unknown. The

need to better characterize such proteins has led to the struc-

tural genomics project,a loosely organized international con-

sortium of structure determination centers dedicated to elu-

cidating the structures of representative proteins from every

protein family and hence making the structures of most pro-

teins readily obtainable from their gene sequences.

Traditionally, the determination of protein structures by

X-ray and NMR techniques has been carried out through

hypothesis-driven research, that is, the proteins are being

studied to solve specific biochemical problems, and hence

these proteins tend to be functionally well characterized. In

contrast, structural genomics endeavors to determine the

structures of large numbers of uncharacterized proteins

(except for their sequences) and has a long-range goal of

determining the structures of all members of selected pro-

teomes (the collections of all proteins encoded by the

corresponding genomes). To effectively do so required that

the rate of protein structure determination be greatly accel-

erated. Hence the first phase of the structural genomics

project concentrated on developing high-throughput (i.e.,

robotic) methods of expressing, purifying, crystallizing, and

structurally elucidating the proteins to be studied.This has

been followed, since 2005, by a production phase in which

large numbers of protein structures have been determined.

258 Chapter 8. Three-Dimensional Structures of Proteins

JWCL281_c08_221-277.qxd 2/23/10 1:59 PM Page 258

Indeed, ⬃14% of the structures in the PDB are products of

the structural genomics project (a percentage that is in-

creasing) with many of these proteins having novel folds.

Unfortunately, the knowledge of a protein’s structure does

not necessarily reveal its function. Hence even though the

cost of determining a protein’s X-ray structure by the

structural genomics project appears to be significantly less

than that by hypothesis-driven research (although the lat-

ter are adopting high-throughput methods), it is still a mat-

ter of debate as to whether the structural genomics project

is a cost-effective way of obtaining biochemically useful in-

formation relative to hypothesis-driven research.

4 PROTEIN STABILITY

Incredible as it may seem, thermodynamic measurements

indicate that native proteins are only marginally stable enti-

ties under physiological conditions. The free energy re-

quired to denature them is ⬃0.4 kJ ⴢ mol

⫺1

of amino acid

residues, so that 100-residue proteins are typically stable by

only around 40 kJ ⴢ mol

⫺1

. In contrast, the energy required

to break a typical hydrogen bond is ⬃20 kJ ⴢ mol

⫺1

.The

various noncovalent influences to which proteins are sub-

ject—electrostatic interactions (both attractive and repul-

sive), hydrogen bonding (both intramolecular and to wa-

ter), and hydrophobic forces—each have energetic

magnitudes that may total thousands of kilojoules per

mole over an entire protein molecule. Consequently, a pro-

tein structure arises from a delicate balance among powerful

countervailing forces. In this section we discuss the nature

of these forces and end by considering protein denatura-

tion, that is, how these forces can be disrupted.

A. Electrostatic Forces

Molecules are collections of electrically charged particles

and hence, to a reasonable degree of approximation, their

interactions are determined by the laws of classical electro-

statics (more exact calculations require the application of

quantum mechanics). The energy of association, U, of two

electric charges, q

and q

, that are separated by the dis-

tance r is found by integrating the expression for

Coulomb’s law, Eq. [2.1], to determine the work necessary

to separate these charges by an infinite distance:

[8.1]

Here k ⫽ 9.0 ⫻ 10

J ⴢ m ⴢ C

⫺2

and D is the dielectric con-

stant of the medium in which the charges are immersed

(recall that D ⫽ 1 for a vacuum and, for the most part, in-

creases with the polarity of the medium;Table 2-1).The di-

electric constant of a molecule-sized region is difficult to

estimate. For the interior of a protein, it is usually taken to

be in the range 3 to 5 in analogy with the measured dielec-

tric constants of substances that have similar polarities,

such as benzene and diethyl ether.

Coulomb’s law is only valid for point or spherically sym-

metric charges that are immersed in a medium of constant

U ⫽

D. However, proteins are by no means spherical and their

internal D values vary with position. Moreover,a protein in

solution associates with mobile ions such as Na

⫹

and Cl

⫺

which modulate the protein’s electrostatic potential. Con-

sequently, calculating the electrostatic potential of a pro-

tein requires mathematically sophisticated and computa-

tionally intensive algorithms that are beyond the scope of

this text. These methods are widely used to calculate the

surface electrostatic potentials of proteins using a program

called GRASP (for Graphical Representation and Analy-

sis of Surface Properties) written by Anthony Nicholls,

Kim Sharp, and Barry Honig. Figure 8-56 shows a GRASP

diagram of human growth hormone in which the protein’s

surface is colored according to its electrostatic potential.

Such diagrams are useful for assessing how a protein might

associate with charged molecules such as other proteins,

nucleic acids, and substrates. Similar computations are used

to predict the pK’s of protein surface groups, which can

have significant application in the elucidation of an en-

zyme’s mechanism of action (Section 15-1).

a. Ionic Interactions Are Strong but Do Not Greatly

Stabilize Proteins

The association of two ionic protein groups of opposite

charge is known as an ion pair or salt bridge. According to

Eq. [8.1], the energy of a typical ion pair, say the carboxyl

Section 8-4. Protein Stability 259

Figure 8-56 A GRASP diagram of human growth hormone.

The diagram shows the protein’s surface colored according to its

electrostatic potential, with its most negative areas dark red, its

most positive areas dark blue, and its neutral areas white. The

protein’s orientation is the same as that in Fig. 8-47b. [Based on

an X-ray structure by Alexander Wlodawer, National Cancer

Institute, Frederick, Maryland. PDBid 1HGU.]

JWCL281_c08_221-277.qxd 2/23/10 1:59 PM Page 259

group of Glu and the ammonium group of Lys, whose

charge centers are separated by 4.0 Å in a medium of

dielectric constant 4, is –86 kJ ⴢ mol

⫺1

(one electronic charge

⫽ 16.0 ⫻ 10

⫺19

C). However, free ions in aqueous solution

are highly solvated, and the formation of a salt bridge has

the entropic penalty of localizing the salt bridge’s charged

side chains. Consequently, the free energy of solvation of

two separated ions is about equal to the free energy of for-

mation of their unsolvated ion pair.Ion pairs therefore con-

tribute little stability toward a protein’s native structure. This

accounts for the observations that although ⬃75% of

charged residues occur in ion pairs (e.g., Fig. 8-57), very few

ion pairs are buried (unsolvated), and ion pairs that are ex-

posed to the aqueous solvent tend to be but poorly con-

served among homologous proteins.

b. Dipole–Dipole Interactions Are Weak but

Significantly Stabilize Protein Structures

The noncovalent associations between electrically neu-

tral molecules, collectively known as van der Waals forces,

arise from electrostatic interactions among permanent

and/or induced dipoles. These forces are responsible for

numerous interactions of varying strengths between non-

bonded neighboring atoms. (The hydrogen bond, a special

class of dipolar interaction, is considered separately in Sec-

tion 8-4B.)

Interactions among permanent dipoles are important

structural determinants in proteins because many of

their groups, such as the carbonyl and amide groups of

the peptide backbone, have permanent dipole moments.

These interactions are generally much weaker than the

charge–charge interactions of ion pairs. Two carbonyl

groups, for example, each with dipoles of 4.2 ⫻ 10

⫺30

C ⴢ m

(1.3 debye units) that are oriented in an optimal head-to-

tail arrangement (Fig. 8-58a) and separated by 5 Å in a

medium of dielectric constant 4, have a calculated attrac-

tive energy of only ⫺9.3 kJ ⴢ mol

⫺1

. Furthermore, these en-

ergies vary with r

⫺3

, so they rapidly attenuate with distance.

In ␣ helices, however, the negative ends of the dipolar

amide and carbonyl groups of the polypeptide backbone

all point in the same direction (Fig. 8-11), so that their in-

teractions and bond dipoles are additive (these groups, of

course, also form hydrogen bonds, but here we are con-

cerned with their residual electric fields). The ␣ helix

therefore has a significant dipole moment that is positive

toward the N-terminus and negative toward the C-terminus.

Consequently, in the low dielectric constant core of a pro-

tein, dipole–dipole interactions significantly influence pro-

tein folding.

A permanent dipole also induces a dipole moment on a

neighboring group so as to form an attractive interaction

(Fig. 8-58b). Such dipole–induced dipole interactions are

generally much weaker than are dipole–dipole interactions.

Although nonpolar molecules are nearly electrically

neutral, at any instant they have a small dipole moment re-

sulting from the rapid fluctuating motions of their elec-

trons. This transient dipole moment polarizes the electrons

in a neighboring group, thereby giving rise to a dipole mo-

ment (Fig. 8-58c) such that, near their van der Waals con-

tact distances, the groups are attracted to one another (a

quantum mechanical effect that cannot be explained in

terms of only classical physics). These so-called London

dispersion forces are extremely weak. The 8.2-kJ ⴢ mol

⫺1

heat of vaporization of CH

, for example, indicates that the

attractive interaction of a nonbonded contact

between neighboring CH

molecules is roughly ⫺0.3

kJ ⴢ mol

⫺1

(in the liquid, a CH

molecule touches its 12

nearest neighbors with ⬃2 contacts each).

London forces are only significant for contacting groups

because their association energy is proportional to r

⫺6

Nevertheless, the great numbers of interatomic contacts in

the closely packed interiors of proteins make London forces

260 Chapter 8. Three-Dimensional Structures of Proteins

Lys 77

Arg 45

Asp 60

Glu 18

–

Figure 8-57 Examples of ion pairs in myoglobin. In each case, oppositely charged

side chain groups from residues far apart in sequence closely approach each other

through the formation of ion pairs.

JWCL281_c08_221-277.qxd 2/23/10 1:59 PM Page 260