Lennarz W.J., Lane M.D. (eds.) Encyclopedia of Biological Chemistry. Four-Volume Set . V. 4

Подождите немного. Документ загружается.

Secondary Structure in

Protein Analysis

George D. Rose

The Johns Hopkins University, Baltimore, Maryland, USA

Proteins are linear, unbranched polymers of the 20 naturally

occurring amino acid residues. Under physiological conditions,

most proteins self-assemble into a unique, biologically relevant

structure: the native fold. This structure can be dissected into

chemically recognizable, topologically simple elements of

secondary structure:

-helix, 3

-helix,

-strand, polyproline

II helix, turns, and V-loops. Together, these six familiar motifs

account for , 95% of the total protein structure, and they are

utilized repeatedly in mix-and-match patterns, giving rise to the

repertoire of known folds. In principle, a protein’s three-

dimensional structure is predictable from its amino acid

sequence, but this problem remains unsolved. A related, but

ostensibly simpler, problem is to predict a protein’s secondary

structure elements from its sequence.

Protein Architecture

A protein is a polymerized chain of amino acid residues,

each joined to the next via a peptide bond. The

backbone of this polymer describes a complex path

through three-dimensional space called the “native

fold” or “protein fold.”

COVALENT STRUCTURE

Amino acids have both backbone and side chain

atoms. Backbone atoms are common to all amino

acids, while side chain atoms differ among the 20

types. Chemically, an amino acid consists of a

central, tetrahedral carbon atom, (

), linked cova-

lently to (1) an amino group (–NH

), (2) a carboxyl

group (–COOH), (3) a hydrogen atom (–H) and (4)

the side chain (–R). Upon polymerization, the amino

group loses an –H and the carboxy group loses an

–OH; the remaining chemical moiety is called an

“amino acid residue” or, simply, a “residue.” Resi-

dues in this polymer are linked via peptide bonds,

as shown in Figure 1.

DEGREES OF FREEDOM

IN THE

BACKBONE

The six backbone atoms in the peptide unit [C

(i)–CO–

NH–C

(i þ 1)] are approximately coplanar, leaving

only two primary degrees of freedom for each residue.

By convention, these two dihedral angles are called

and

(Figure 2). The protein’s backbone conformation

is described by the

-speciﬁcation for each residue.

CLASSIFICATION OF STRUCTURE

Protein structure is usually classiﬁed into primary,

secondary, and tertiary structure. “Primary structure”

corresponds to the covalently connected sequence of

amino acid residues. “Secondary structure” corresponds

to the backbone structure, with particular emphasis on

hydrogen bonds. And “tertiary structure” corresponds

to the complete atomic positions for the protein.

Secondary Structure

Protein secondary structure can be subdivided into

repetitive and nonrepetitive, depending upon whether

the backbone dihedral angles assume repeating values.

There are three major elements (

-helix,

-strand, and

polyproline II helix) and one minor element (3

-helix)

of repetitive secondary structure (Figure 3). There are

two major elements of nonrepetitive secondary structure

(turns and V- loops).

REPETITIVE SECONDARY STRUCTURE:

THE

-HELIX

When backbone dihedral angles are assigned repeating

-values near (2 608, 2 408), the chain twists into a

right-handed helix, with 3.6 residues per helical turn.

First proposed as a model by Pauling, Corey, and

Branson in 1951, the existence of this famous structure

was experimentally conﬁrmed almost immediately by

Perutz in ongoing crystallographic studies, well before

elucidation of the ﬁrst protein structure.

In an

-helix, each backbone N–H forms a hydrogen

bond with the backbone carbonyl oxygen situated four

residues away in the linear sequence chain (toward the

N-terminus): N–H(i)· · ·OyC(i 2 4). The two sequen-

tially distant hydrogen-bonded groups are brought into

spatial proximity by conferring a helical twist upon

the chain. This results in a rod-like structure, with the

hydrogen bonds oriented approximately parallel to

the long axis of the helix.

In globular proteins, the average length of an

-helix

is 12 residues. Typically, helices are found on the outside

of the protein, with a hydrophilic face oriented toward

the surrounding aqueous solvent and a hydrophobic face

oriented toward the protein interior.

Inescapably, end effects deprive the ﬁrst four amide

hydrogens and last four carbonyl oxygens of

Pauling-type, intra-helical hydrogen bond partners.

The special hydrogen-bonding motifs that can provide

partners for these otherwise unsatisﬁed groups are

known as “helix caps.”

In globular proteins, helices account for , 25% of

the structure on average, but this number varies.

Some proteins, like myoglobin, are predominantly

helical, while others, like plastocyanin, lack helices

altogether.

REPETITIVE SECONDARY STRUCTURE:

THE 3

-HELIX

When backbone dihedral angles are assigned repeating

-values near (2 508, 2 308), the chain twists into a

right-handed helix. By convention, this helix is named

using formal nomenclature: 3

designates three residues

per helical turn and 10 atoms in the hydrogen bonded

ring between each N–H donor and its CyO acceptor.

(In this nomenclature, the

-helix would be called a

3.6

helix.)

Single turns of 3

helix are common and closely

resemble a type of

-turn (see below). Often,

-helices

terminate in a turn of 3

helix. Longer 3

helices are

sterically strained and much less common.

FIGURE 1 (A) A generic amino acid. Each of the 20 naturally occurring amino acids has both backbone atoms (within the shaded rectangle) and

side chain atoms (designated R). Backbone atoms are common to all amino acids, while side chain atoms differ among the 20 types. Chemically, an

amino acid consists of a tetrahedral carbon atom (–C–), linked covalently to (1) an amino group (–NH

), (2) a carboxyl group (–COOH), (3) a

hydrogen atom (–H), and (4) the side chain (–R). (B) Amino acid polymerization. The

-amino group of one amino acid condenses with the

-carboxylate of another, releasing a water molecule. The newly formed amide bond is called a peptide bond and the repeating unit is a residue. The

two chain ends have a free

-amino group and a free

-carboxylate group and are designated the amino-terminal (or N-terminal) and the carboxy-

terminal (or C-terminal) ends, respectively. The peptide unit consists of the six shaded atoms (C

–CO–NH–C

), three on either side of the peptide

bond.

2 SECONDARY STRUCTURE IN PROTEIN ANALYSIS

REPETITIVE SECONDARY STRUCTURE:

THE

-STRAND

When backbone dihedral angles are assigned repeating

-values near (2 1208, 2 1208), the chain adopts

an extended conformation called a

-strand. Two or

-strands, aligned so as to form inter-strand

hydrogen bonds, are called a

-sheet. A

-sheet of just

two hydrogen-bonded

-strands interconnected by a

tight turn is called a

-hairpin. The average length of a

single

-strand is seven residues.

The classical deﬁnition of secondary structure found

in most textbooks is limited to hydrogen-bonded back-

bone structure and, strictly speaking, would not include

-strand, only a

-sheet. However, the

-sheet is

tertiary structure, not secondary structure; the interven-

ing chain joining two hydrogen-bonded

-strands can

range from a tight turn to a long, structurally complex

stretch of polypeptide chain. Further, approximately

half the

-strands found in proteins are singletons and

do not form inter-strand hydrogen bonds with another

-strand. Textbooks tend to blur this issue.

Typically,

-sheet is found in the interior of the

protein, although the outermost parts of edge-strands

usually reside at the protein’s water-accessible surface.

Two

-strands in a

-sheet are classiﬁed as either

parallel or anti-parallel, depending upon whether their

mutual N- to C-terminal orientation is the same or

opposite, respectively.

In globular proteins,

-sheet accounts for about 15%

of the structure on an average, but, like helices, this

number varies considerably. Some proteins are pre-

dominantly sheet while others lack sheet altogether.

REPETITIVE SECONDARY STRUCTURE:

THE POLYPROLINE II HELIX (P

)

When backbone dihedral angles are assigned repeating

-values near (2 708, þ 1408), the chain twists into

a left-handed helix with 3.0 residues per helical turn.

The name of this helix is derived from a poly-proline

homopolymer, in which the structure is forced by its

stereochemistry. However, a polypeptide chain can

adopt a P

helical conformation whether or not it

contains proline residues.

Unlike the better known

-helix, a P

helix has no

intrasegment hydrogen bonds, and it is not included in

the classical deﬁnition of secondary structure for this

reason. This extension of the deﬁnition is also needed in

the case of an isolated

-strand. Recent studies have

shown that the unfolded state of proteins is rich in

structure.

FIGURE 2 (A) Deﬁnition of a dihedral angle. In the diagram, the

dihedral angle,

, measures the rotation of line segment CD with respect

to line segment AB, where A, B, C, and D correspond to the x,y,z-

positions of four atoms. (

is calculated as the scalar angle between the

two normals to planes A–B–C and B–C–D.) By convention, clockwise

rotation is positive and

¼ 08 when A and D are eclipsed. (B) Degrees of

freedom in the protein backbone. The peptide bond (C

–N) has partial

double bond character, so that the six atoms, Ca(i)–CO–Ca(i þ 1), are

approximately co-planar. Consequently, only two primary degrees of

freedom are available for each residue. By convention, these two

dihedral angles are called

and

is speciﬁed by the four atoms C

(i)–

N–C

–C

(i þ 1) and

by the four atoms N(i)–C

–C

–N(i þ 1).

When the chain is fully extended, as depicted here,

¼ 1808:

FIGURE 3 A contoured Ramachandran (

;

) plot. Backbone

angles were extracted from 1042 protein subunits of known structure.

Only nonglycine residues are shown. Contours were drawn in popu-

lation intervals of 10% and are indicated by the ten colors (in rainbow

order). The most densely populated regions are colored red. Three

heavily populated regions are apparent, each near one of the

major elements of repetitive secondary structure:

-helix (, 2 608,

2 408),

-strand (, 21208,1208), P

helix (, 2708, 1408). Adapted from

Hovmo

ller, S., Zhou, T., and Ohlson, T. (2002). Conformation of amino

acids in proteins. Acta Cryst. D58, 768–776, with permission of IUCr.

SECONDARY STRUCTURE IN PROTEIN ANALYSIS 3

NONREPETITIVE SECONDARY

STRUCTURE: THE TURN

Turns are sites at which the polypeptide chain changes

its overall direction, and their frequent occurrence is the

reason why globular proteins are, in fact, globular.

Turns can be subdivided into

-turns,

-turns, and

tight turns.

-turns involve four consecutive residues,

with a hydrogen bond between the amide hydrogen of

the 4th residue and the carbonyl oxygen of the 1st

residue: N–H(i)···OyC(i 2 3).

-turns are further

subdivided into subtypes (e.g., Type I, I

, II, II

, III,…)

depending upon their detailed stereochemistry.

-turns

involve only three consecutive, hydrogen-bonded resi-

dues, N–H(i)· · ·OyC(i 2 2), which are further divided

into subtypes.

More gradual turns, known as “reverse turns” or

“tight turns,” are also abundant in protein structures.

Reverse turns lack intra-turn hydrogen bonds but

nonetheless, are involved in changes in overall chain

direction.

Turns are usually, but not invariably, found on the

water-accessible surface of proteins. Together,

-and

reverse turns account for about one-third of the

structure in globular proteins, on an average.

NONREPETITIVE SECONDARY

STRUCTURE: THE V-LOOP

V-loops are sites at which the polypeptide loops back on

itself, with a morphology that resembles the Greek letter

“V” although often with considerable distortion. They

range in length from 6–16 residues, and, lacking any

speciﬁc pattern of backbone-hydrogen bonding, can

exhibit signiﬁcant structural heterogeneity.

Like turns, V-loops are typically found on the outside

of proteins. On an average, there are about four such

structures in a globular protein.

Identiﬁcation of Secondary

Structure from Coordinates

Typically, one becomes familiar with a given protein

structure by visualizing a model – usually a computer

model – that is generated from experimentally deter-

mined coordinates. Some secondary structure types are

well deﬁned on visual inspection, but others are not. For

example, the central residues of a well-formed helix are

visually unambiguous, but the helix termini are subject

to interpretation. In general, visual parsing of the

protein into its elements of secondary structure can be

a highly subjective enterprise. Objective criteria have

been developed to resolve such ambiguity. These criteria

have been implemented in computer programs that

accept a protein’s three-dimensional coordinates as

input and provide its secondary structure components

as output.

INHERENT AMBIGUITY IN

STRUCTURAL IDENTIFICATION

It should be realized that objective criteria for structural

identiﬁcation can provide a welcome self-consistency,

but there is no single “right” answer. For example, turns

have been deﬁned in the literature as chains sites at

which the distance between two

-carbon atoms,

separated in sequence by four residues, is not more

than 7A

, provided the residues are not in an

-helix:

distance[C

(i)–C

(i þ 3)] # 7A

and C

(i)–C

(i þ 3)

not

-helix. Indeed, turns identiﬁed using this deﬁnition

agree quite well with one’s visual intuition. However,

the 7A

threshold is somewhat arbitrary. Had 7.1A

been

used instead, additional, intuitively plausible turns

would have been found.

PROGRAMS TO IDENTIFY STRUCTURE

FROM

COORDINATES

Many workers have devised algorithms to parse the

three-dimensional structure into its secondary structure

components. Unavoidably, these procedures include

investigator-deﬁned thresholds. Two such programs are

mentioned here.

The Database of Secondary Structure

Assignments in Proteins

This is the most widely used secondary structure

identiﬁcation method available today. Developed by

Kabsch and Sander, it is accessible on the internet, both

from the original authors and in numerous implemen-

tations from other investigators as well.

The database of secondary structure assignments in

proteins (DSSP) identiﬁes an extensive set of secondary

structure categories, based on a combination of back-

bone dihedral angles and hydrogen bonds. In turn,

hydrogen bonds are identiﬁed based on geometric criteria

involving both the distance and orientation between a

donor–acceptor pair. The program has criteria for

recognizing

-helix, 3

-helix,

helix,

-sheet (both

parallel and anti-parallel), hydrogen-bonded turns and

reverse turns. (Note: the

-helix is rare and has been

omitted from the secondary structure categories.)

Protein Secondary Structure Assignments

In contrast to DSSP, protein secondary structure assign-

ments (PROSS) identiﬁcation is based solely on back-

bone dihedral angles, without resorting to hydrogen

SECONDARY STRUCTURE IN PROTEIN ANALYSIS

bonds. Developed by Srinivasan and Rose, it is

accessible on the internet.

PROSS identiﬁes only

-helix,

-strand, and turns,

using standard

deﬁnitions for these categories.

Because hydrogen bonds are not among the identiﬁ-

cation criteria, PROSS does not distinguish between

isolated

-strands and those in a

-sheet.

Prediction of Protein

Secondary Structure from

Amino Acid Sequence

Efforts to predict secondary structure from amino acid

sequence dates back to the 1960s to the works of Guzzo,

Prothero and, slightly later, Chou and Fasman. The

problem is complicated by the fact that protein secondary

structure is only marginally stable, at best. Proteins fold

cooperatively, with secondary and tertiary structure

emerging more or less concomitantly. Typical peptide

fragments excised from the host protein, and measured in

isolation, exhibit only a weak tendency to adopt their

native secondary structure conformation.

PREDICTIONS BASED ON EMPIRICALLY

DETERMINED PREFERENCES

Motivated by early work of Chou and Fasman, this

approach uses a database of known structures to discover

the empirical likelihood, f ; of ﬁnding each of the twenty

amino acids in helix, sheet, turn, etc. These likelihoods

are equated to the residue’s normalized frequency of

occurrence in a given secondary structure type, obtained

by counting. Using alanine in helices as an example

fraction Ala in helix ¼

occurrences of Ala in helices

occurrences of Ala in database

This fraction is then normalized against the corre-

sponding fraction of helices in the database:

helix

Ala

fraction Ala in helix

fraction helices in database

occurrences of Ala in helices

occurrences of Ala in database

number of residues in helices

number of residues in database

A normalized frequency of unity indicates no pre-

ference – i.e., the frequency of occurrence of the given

residue in that particular position is the same as its

frequency at large. Normalized frequencies greater

than/less than unity indicate selection for/against the

given residue in a particular position.

These residue likelihoods are then used in combination

to make a prediction. When only a small number of

proteins had been solved, these data-dependent f-values

ﬂuctuated signiﬁcantly as new structures were added to

the database. At this point there are more than 22 000

structures in the Protein Data Bank (www.rcsb.org), and

the f-values have reached a plateau.

DATABASE-INDEPENDENT PREDICTIONS:

THE

HYDROPHOBICITY PROFILE

Hydrophobicity proﬁles have been used to predict the

location of turns in proteins. A hydrophobicity proﬁle is

a plot of the residue number versus residue hydropho-

bicity, averaged over a running window. The only

variables are the size of the window used for averaging

and the choice of hydrophobicity scale (of which there

are many). No empirical data from the database is

required. Peaks in the proﬁle correspond to local

maxima in hydrophobicity, and valleys to local minima.

Prediction is based on the idea that apolar sites along the

chain (i.e., peaks in the proﬁle) will be disposed

preferentially to the molecular interior, forming a

hydrophobic core, whereas polar sites (i.e., valleys in

the proﬁle) will be disposed to the exterior and

correspond to chain turns.

NEURAL NETWORKS

More recently, neural network approaches to second-

ary structure prediction have come to dominate the

ﬁeld. These approaches are based on pattern-recog-

nition methods developed in artiﬁcial intelligence.

When used in conjunction with the protein database,

these are the most successful programs available

today.

A neural network is a computer program that

associates an input (e.g., a residue sequence) with an

output (e.g., secondary structure prediction) through a

complex network of interconnected nodes. The path

taken from the input through the network to the output

depends upon past experience. Thus, the network is said

to be “trained” on a dataset.

The method is based on the observation that amino

acid substitutions follow a pattern within a family of

homologous proteins. Therefore, if the sequence of

interest has homologues within the database of known

structures, this information can be used to improve

predictive success, provided the homologues are recog-

nizable. In fact, a homologue can be recognized quite

successfully when the sequence of interest and a putative

homologue have an aligned sequence identity of 25%

or more.

Neural nets provide an information-rich approach

to secondary structure prediction that has become

increasingly successful as the protein databank has

grown.

SECONDARY STRUCTURE IN PROTEIN ANALYSIS 5

PHYSICAL BASIS OF

SECONDARY STRUCTURE

An impressive number of secondary structure prediction

methods can be found in the literature and on the web.

Surprisingly, almost all are based on empirical like-

lihoods or neural nets; few are based on physico-

chemical theory.

In one such theory, secondary structure propensities

are predominantly a consequence of two competing

local effects – one favoring hydrogen bond formation in

helices and turns, and the other opposing the attendant

reduction in sidechain conformational entropy upon

helix and turn formation. These sequence-speciﬁc biases

are densely dispersed throughout the unfolded polypep-

tide chain, where they serve to pre-organize the folding

process and largely, but imperfectly, anticipate the native

secondary structure.

WHY AREN’T SECONDARY STRUCTURE

PREDICTIONS BETTER?

Currently, the best methods for predicting helix and

sheet are correct about three-quarters of the time. Can

greater success be achieved?

Several measures to assess predictive accuracy are in

common use, of which the Q3 score is the most

widespread. The Q3 score gives the percentage of

correctly predicted residues in three categories: helix,

strand, and coil (i.e., everything else):

Q3 ¼

number of correctly predicted residues

total number of residues

£ 100

where the “correct” answer is given by a program

to identify secondary structure from coordinates,

e.g., DSSP. At this writing, (Position-Speciﬁc

PREDiction algorithm) PSIPRED has an overall Q3

score of 78%.

Is greater prediction accuracy possible? It has

been argued that prediction methods fail to achieve a

higher rate of success because some amino acid

sequences are inherently ambiguous. That is, these

“conformational chameleons” will adopt a helical

conformation in one protein, but the identical sequence

will adopt a strand conformation in another protein.

Only time will tell whether current efforts have encoun-

tered an inherent limit.

FURTHER READING

Berg, J. M., Tymoczko, J. L., and Stryer, L. (2002). Biochemistry, 5th

edition. W.H. Freeman and Company, New York.

Holm, L., and Sander, C. (1996). Mapping the protein universe.

Science 273, 595–603.

Hovmo

ller, S., Zhou, T., and Ohlson, T. (2002). Conformation of

amino acids in proteins. Acta Cryst. D58, 768–776.

Jones, D. T. (1999). Protein secondary structure based on position-

speciﬁc scoring matrices. J. Mol. Biol. 292, 195– 202.

Mathews, C., van Holde, K. E., and Ahern, K. G. (2000). Biochemi-

stry, 3rd edition. Pearson Benjamin Cummings, Menlo Park, CA.

Richardson, J. S. (1981). The anatomy and taxonomy of protein

structure. Adv. Prot. Chem. 34, 168–340.

Rose, G. D., Gierasch, L. M., and Smith, J. A. (1985). Turns in peptides

and proteins. Adv. Prot. Chem. 37, 1–109.

Voet, D., and Voet, J. G. (1996). Biochemistry, 2nd edition. Wiley,

New York.

BIOGRAPHY

George Rose is Professor of Biophysics and Director of the Institute for

Biophysical Research at Johns Hopkins University. He holds a Ph.D.

from Oregon State University. His principal research interest is in

protein folding, and he has written many articles on this topic. He

serves as the consulting editor of Proteins: Structure, Function and

Genetics and as a member of the editorial advisory board of Protein

Science. Recently, he was a Fellow of the John Simon Guggenheim

Memorial Foundation.

6 SECONDARY STRUCTURE IN PROTEIN ANALYSIS

Secretases

Robert L. Heinrikson

The Pharmacia Corporation, Kalamazoo, Michigan, USA

Secretases are proteolytic enzymes involved in the processing

of an integral membrane protein known as Amyloid precursor

protein, or APP.

-Amyloid (A

) is a neurotoxic and highly

aggregative peptide that is excised from APP by secretase

action, and that accumulates in the neuritic plaque found in the

brains of Alzheimer’s disease (AD) patients. The amyloid

hypothesis holds that the neuronal dysfunction and clinical

manifestation of AD is a consequence of the long-term

deposition and accumulation of A

, and that this peptide of

40–42 amino acids is a causative agent of AD. Accordingly, the

secretases involved in the liberation, or destruction of A

are

of enormous interest as therapeutic intervention points toward

treatment of this dreaded disease.

Background

Proteolytic enzymes play crucial roles in a wide variety

of normal and pathological processes in which they

display a high order of selectivity for their substrate(s)

and the speciﬁc peptide bonds hydrolyzed therein. This

article concerns secretases, membrane-associated pro-

teinases that produce, or prevent formation of, a highly

aggregative and toxic peptide called

-amyloid (A

This A

peptide is removed from a widely distributed

and little understood Type I integral membrane protein

called amyloid precursor protein (APP). The apparent

causal relationship between A

and AD has fueled an

intense interest in the secretases responsible for its

production. Herein will be discussed the current under-

standing of three of the most-studied secretases,

and

-secretases. A schematic representation of the A

region of APP showing the amino acid sequence of A

and the major sites of cleavage for these three secretases

is given in Figure 1.A

is produced by the action of

- and

-secretases, and there is an intense search

underway for inhibitors of these enzymes that might

serve as drugs in treatment of Alzheimer’s disease (AD).

The

-secretase cleaves at a site near the middle of A

and gives rise to fragments of A

that lack the

potential for aggregation; therefore, ampliﬁcation of

-secretase activity might be seen as another approach

to AD therapy.

-Secretase

The activity responsible for cleavage of the Lys

-Leu

bond within the A

region (Figure 1) is ascribed to

-secretase. This action prevents formation of the 40–42

amino acid residue A

and leads to release of soluble

APP

and the membrane-bound C83-terminal fragment.

-Secretase competes with

-secretase for the APP

substrate, but the

-secretase product, soluble APP

(pathway A, Figure 1) is generated at a level about 20

times that of the sAPP

released by

-secretase (pathway

B). Because

-secretase action prevents formation of the

toxic A

peptide, augmentation of this activity could

represent a useful strategy in AD treatment, and this has

been done experimentally by activators of protein kinase

C (PKC) such as phorbol esters and by muscarinic

agonists. The speciﬁcity of the

-secretase for the

Lys

- # -Leu

cleavage site (Figure 1) appears to be

governed by spatial and structural requirements that

this bond exist in a local

-helical conformation and be

within 12 or 13 amino acids distance from the membrane.

-Secretase has not been identiﬁed as any single

proteinase, but two members of the ADAM (a disintegrin

and metalloprotease) family, ADAM-10 and ADAM-17

are candidate

-secretases. ADAM-17 is known as TACE

(tumor necrosis factor-

-converting enzyme) and TACE

cleaves peptides modeled after the

-secretase site at the

Lys

- # -Leu position. This was also shown to be the case

for ADAM-10; overexpression of this enzyme in a human

cell line led to several-fold increase in both basal and

PKC-inducible

-secretase activity. As of now, it remains

to be proven whether

-secretase activity derives from

either or both of these ADAM family metalloproteinases,

or whether another as yet unidentiﬁed proteinase carries

out this processing of APP.

-Secretase

The enzyme responsible for cleaving at the amino-

terminus of A

-secretase (Figure 1). In the mid-1980s,

when A

was recognized as a principal component of AD

neuritic plaque, an intense search was begun to identify

the

-secretase. Finally, in 1999, several independent

laboratories published evidence demonstrating that

-secretase is a unique member of the pepsin family of

aspartyl proteinases. This structural relationship to a

well-characterized and mechanistically deﬁned class of

proteases gave enormous impetus to research on

-secretase. The preproenzyme consists of 501 amino

acids, with a 21-residue signal peptide, a prosegment of

about 39 residues, the catalytic bilobal unit with active

site aspartyl residues at positions 93 and 289, a

27-residue transmembrane region, and a 21-residue

C-terminal domain. The membrane localization of

-secretase makes it unique among mammalian aspartyl

proteases described to date. Another interesting feature

of the enzyme is that, unlike pepsin, renin, cathepsin D,

and other prototypic members of the aspartyl proteases,

it does not appear to require removal of the prosegment

as a means of activation. A furin-like activity is

responsible for cleavage in the sequence Arg-Leu-Pro-

Arg- # -Glu

of the proenzyme, but this does not lead to

any remarkable enhancement of activity, at least as is seen

in recombinant constructs of pro-

-secretase.

-Secre-

tase has been referred to by a number of designations in

the literature, but the term BACE (

-site APP cleaving

enzyme) has become most widely adopted. With the

discovery of the

-secretase, it was recognized that there

was another human homologue of BACE with a

transmembrane segment and this has now come to be

called BACE2. This may well be a misnomer, since the

function of BACE2 has yet to be established, and it is

not clear that APP is a normal substrate of this enzyme.

At present, BACE2 is not considered to be a secretase.

There is considerable experimental support for the

assertion that BACE is, in fact, the

-secretase involved

in APP processing. The enzyme is highly expressed in

brain, but is also found in other tissues, thus explaining

the fact that many cell types can process A

. Use of

antisense oligonucleotides to block expression of BACE

greatly diminishes production of A

and, conversely,

overexpression of BACE in a number of cell lines leads

to enhanced A

production. BACE knockout mice show

no adverse phenotype, but have dramatically reduced

levels of A

. This not only demonstrates that BACE is

the true

-site APP processor, but also that its

elimination does not pose serious consequences for

the animal, a factor of great importance in targeting

BACE for inhibition in AD therapy.

FIGURE 1 A schematic overview of APP processing by the

-, and

-secretases. The top panel shows the amino acid sequence of APP

upstream of the transmembrane segment (underlined, bold), and encompassing the sequences of A

1–40

and A

1–42

–V

, and D

–A

respectively). The

-secretase cleaves at D

and Y

; the

-secretase at Lys

, and the

-secretase at Val

and/or Ala

. Below the sequence is a

representation of APP emphasizing its membrane localization and the residue numbers of interest in

- and

-secretase processing. Panel A

represents the non-amyloidogenic

-secretase pathway in which sAPP

and C83 are generated. Subsequent hydrolysis by the

-secretase produces a

p3 peptide that does not form amyloid deposits. Panel B represents the amyloidogenic pathway in which cleavage of APP by the

-secretase to

liberate sAPP

and C99 is followed by

-secretase processing to release

-amyloid peptides (A

1–40

and A

1–42

) found in plaque deposits.

8 SECRETASES

Much of the evidence in support of the amyloid

hypothesis comes from the observation of mutations

near the

- and

-cleavage sites in APP that inﬂuence

production of A

and correlate directly with the onset of

AD. One such mutation in APP, that invariably leads to

AD in later life, occurs at the

-cleavage site where Lys-

Met

is changed to Asn-Leu

(Figure 1). This so-

called Swedish mutation greatly enhances production of

, and as would be expected,

-secretase hydrolyzes

the mutated Leu-Asp

bond in model peptides , 50

times faster than the wild-type Met-Asp

bond. It is

important to recognize that BACE cleavage is required

for subsequent processing by the

-secretase; in this

sense, a BACE inhibitor will also block

-secretase.

Another BACE cleavage point is indicated in Figure 1 by

the arrow at Y

# E

; the A

11 – 40 or 42

subsequently

liberated by

-secretase action also forms amyloid

deposits and is found in neuritic plaque.

In all respects, therefore, BACE ﬁts the picture

expected of

-secretase, and because of its detailed

level of characterization and its primary role in A

production, it has become a major target for develop-

ment of inhibitors as drugs to treat AD. Great strides in

this direction have become possible because of the

availability of three-dimensional (3-D) structural infor-

mation on BACE. The crystal structure of BACE

complexed with an inhibitor is represented schemati-

cally in Figure 2. Homology with the pepsin-like

aspartyl proteases is reﬂected in the similar folding

pattern of BACE, with extensive

-sheet organization,

and the proximal location of the two aspartyl residues

that comprise the catalytic machine for peptide bond

cleavage. The C-terminal lobe of the molecule is larger

than is customarily seen in the aspartyl proteases, and

contains extra elements of structure with as yet

unexplained impact on function. In fact, before the

crystal structure was solved, it was thought that this

larger C-terminal region might contribute a spacer to

distance the catalytic unit from the membrane and to

provide mobility. This appears not to be the case. As

denoted by the arrow in Figure 2, there is a critical

disulﬁde bridge linking the C-terminal region just

upstream of the transmembrane segment to the body

of the molecule. Therefore, the globular BACE molecule

is proximal to the membrane surface and is not attached

via a mobile stalk that would permit much motion. This

steric localization would be expected to limit the

repertoire of protein substrates that are accessible to

BACE as it resides in the Golgi region. Crystal structures

of BACE/inhibitor complexes have revealed much about

the nature of protein-ligand interactions, and infor-

mation regarding the nature of binding sites obtained by

this approach will be of critical importance in the design

and development of inhibitors that will be effective

drugs in treatment of AD.

-Secretase

-Secretase activity is produced in a complex of proteins

and is yet to be understood in terms of the actual catalytic

entity and mechanism of proteolysis. This secretase

cleaves bonds in the middle of the APP segment that

traverses the membrane (underlined and boldface in

Figure 1), and its activity is exhibited subsequent to

cleavages at the

-or

-sites. In Figure 1, the

-secretase

cleavage sites are indicated by two arrows. Cleavage at

the Val

-Ile

bond liberates the more abundant

40-amino acid residue A

1–40

). Cleavage at Ala

Thr

produces a minor A

species, A

1–42

, but one that

appears to be much more hydrophobic and aggregative,

and it is the 42-residue A

that is believed to be of

most signiﬁcance in AD pathology. As was the case for

APP

-site mutations, there are human APP mutants

showing alterations in the vicinity of the

-site, and these

changes, powerfully associated with onset of AD, lead to

higher ratios of A

1–42

Central to the notion of the

-secretase is the

presence of presenilins, intregral membrane proteins

with mass , 50 kDa. There are a host of presenilin

mutations in familial AD (FAD) that are associated

with early onset disease and an increased production of

the toxic A

1–42

. This correlation provides strong

FIGURE 2 Schematic representation of the 3-D structure of the

BACE (

-secretase) catalytic unit as determined by x-ray crystallo-

graphy. Arrows and ribbons designate

-strands and

-helices,

respectively. An inhibitor is shown bound in the cleft deﬁned by the

amino- (left) and carboxyl- (right) terminal halves of the molecule. The

C-terminus of the catalytic unit is marked C to indicate the amino acid

residue immediately preceding the transmembrane and cytoplasmic

domains of BACE. These latter domains were omitted from the

construct that was solved crystallographically. The arrow marks a

disulﬁde bridge, which maintains the C-terminus in close structural

association with the body of the catalytic unit. The catalytic entity as

depicted sits directly on the membrane surface, thereby restricting its

motion relative to protein substrates. (Courtesy of Dr. Lin Hong,

Oklahoma Medical Research Foundation, Oklahoma City, OK.)

SECRETASES 9

support for the involvement of presenilin in AD, and its

presence in

-secretase preparations implies that it is

either a proteolytic enzyme in its own right, or can

contribute to that function in the presence of other

proteins. In fact, much remains to be learned about the

presenilins; it has been difﬁcult to obtain precise

molecular and functional characterization because of

their close association with membranes and other

proteins in a complex. Modeling studies have predicted

a variable number of transmembrane segments (6–8),

but presenilin function is predicated upon processing by

an unknown protease to yield a 30 kDa N-terminal

fragment (NTF) and a 20 kDa C-terminal fragment

(CTF). These accumulate in vivo in a 1:1 stoichiometry

within high molecular weight complexes with a variety of

ancillary proteins. Some of the cohort proteins

identiﬁed in the multimeric presenilin complexes dis-

playing

-secretase activity include catenins, armadillo-

repeat proteins that appear not to be essential for

-secretase function, and nicastrin. Nicastrin is a Type I

integral membrane protein with homologues in a variety

of organisms, but its function is unknown. It shows

intracellular colocalization with presenilin, and is able to

bind the NTF and CTF of presenilin as well as the C83

and C99 C-terminal APP substrates of

-secretase.

Interestingly, down-regulation of the nicastrin homol-

ogue in Caenorhabditis elegans gave a phenotype similar

to that seen in worms deﬁcient in presenilin and notch.

Evidence that nicastrin is essential for

-secretase

cleavage of APP and notch adds to the belief that

nicastrin is an important element in presenilin, and

-secretase function. Efforts to delineate other protein

components of

-secretase complexes and to understand

their individual roles in the enzyme function represent a

large current research effort. Recently, two additional

proteins associated with the complex have been identiﬁed

through genetic screening of ﬂies and worms. The aph-1

gene encodes a protein with 7 transmembrane domains,

and the pen-2 gene codes for a small protein passing twice

through the membrane. Both of these putative members

of the

-secretase complex are new proteins whose

functions, either with respect to secretase activity or in

other potential systems, remain to be elucidated.

At present, it is still unclear as to how

-secretase exerts

its function. What is known, however, is that

-secretase

is able to cleave at other peptide bonds in APP near the

-site in addition to those indicated in Figure 1, and is

involved with processing of intra-membrane peptide

bonds in a variety of additional protein substrates,

including notch. This lack of speciﬁcity is a major concern

in developing drugs for AD targeted to

-secretase that do

not show side effects due to inhibition of processing of

these additional, functionally diverse protein substrates.