`and amyloid
`Anthony L Fink
`
`Review R9
`
`Aggregation results in the formation of inclusion
`bodies, amyloid fibrils and folding aggregates.
`Substantial data support the hypothesis that partially
`folded intermediates are key precursors to aggregates,
`that aggregation involves specific intermolecular
`interactions and that most aggregates involve βb sheets.
`
`Address: Department of Chemistry and Biochemistry, University of
`California, Santa Cruz, CA 95064, USA.
`E-mail: enzyme@cats.ucsc.edu
`
`Folding & Design 01 February 1998, 3:R9–R23
`http://biomednet.com/elecref/13590278003R0009
`
`© Current Biology Ltd ISSN 1359-0278
`
`Introduction
`Protein aggregation can be merely a nuisance factor in
`many in vitro studies of proteins or it can cause major
`economic and technical problems in the biotechnology
`and pharmaceutical industries. Its effects can be lethal in
`patients who suffer from a variety of diseases involving
`protein aggregation, such as the amyloidoses, prion dis-
`eases and other protein deposition disorders [1,2]. This
`review focuses on the basic mechanism(s) of protein aggre-
`gation, the factors that determine whether it will occur,
`and the conformation of the protein molecules in the
`aggregate. Protein aggregation is intimately tied to protein
`folding and stability, and also, in the cell, to molecular
`chaperones. The prevalence of protein aggregation is
`probably much higher than generally realized — it is often
`ignored or worked around, and in protein folding experi-
`ments its presence may not even be realized [3]. The
`growing recognition of the critical importance of protein
`aggregation has resulted in a number of reviews [4–10].
`
`Unless specifically noted to the contrary, in this review
`the term aggregation will apply to aggregated protein
`involving the formation of insoluble precipitates that may
`be considered ‘pathological’ in nature. This is in contrast
`to the insolubility of the native state due to protein con-
`centrations exceeding the solubility limit (e.g. ‘salting
`out’), or the intermolecular association involved in the for-
`mation of native oligomers. It should be noted that in
`many such cases of pathological aggregation the initial
`material formed may be soluble aggregates, but these
`become insoluble when they exceed a certain size.
`
`It is convenient to classify protein aggregation according
`to the following categories: in vivo and in vitro, and
`ordered and disordered. Amyloid fibrils (both in vivo and
`
`in vitro) are examples of ordered aggregates, whereas
`inclusion bodies are examples of in vivo disordered aggre-
`gates. Corresponding disordered in vitro aggregates are
`those formed during the refolding of denaturant-unfolded
`protein at high protein concentrations, or under weakly
`native conditions at high protein concentration; these will
`be referred to as folding aggregates.
`
`Native, folded proteins may aggregate under certain con-
`ditions, most notably salting out and isoelectric precipita-
`tion (when the net charge on the protein is zero). Such
`precipitates of native protein are readily distinguished
`from pathological aggregates by their solubility in buffer
`under native-like conditions. In contrast, pathological
`aggregates dissociate and dissolve only in the presence of
`high concentrations of denaturant or detergent. In my lab-
`oratory, it has been shown that the native conformation is
`retained in salting out precipitates (Figure 1).
`
`Protein aggregation has usually been assumed to involve
`either unfolded or native states. Inclusion body formation
`and other aggregates formed during protein folding have
`been assumed to arise from hydrophobic aggregation of
`the unfolded or denatured states, whereas amyloid fibrils
`and other extracellular aggregates have been assumed to
`arise from native-like conformations in a process analo-
`gous to the polymerization of hemoglobin S [8]. Recent
`observations suggest that aggregation is much more likely
`to arise from specific partially folded intermediates,
`however. An important consequence of this is that aggre-
`gation will be favored by factors and conditions that favor
`population of these intermediates, and hence it is the
`properties of these intermediates that are important in
`determining whether aggregation occurs. Furthermore,
`the characteristics and properties of the intermediates may
`be significantly different from those of the native (and
`unfolded) conformation.
`
`Several observations indicate that transient aggregation
`occurring during in vitro protein refolding may be mis-
`taken for a transient intermediate [3]. Direct evidence for
`the transient association of partially folded intermediates
`during refolding has been obtained in small-angle X-ray
`scattering experiments of apomyoglobin [11,12], carbonic
`anhydrase and phosphoglycerate kinase [13]. The experi-
`ments show the rapid (milliseconds or less) formation of
`associated states that become monomeric on a slow time-
`scale (typically seconds to minutes or longer). Another
`approach indicative of transient aggregation is that involv-
`ing changes in the rate constants for refolding as a function
`
`MYLAN INST. EXHIBIT 1096 PAGE 1
`
`MYLAN INST. EXHIBIT 1096 PAGE 1
`
`
`
`folding aggregates and amorphous deposits); how the envi-
`ronmental conditions affect the rate and the amount of
`aggregation; and how the aggregation may be prevented.
`
`It seems likely from an evolutionary perspective that pro-
`teins have evolved to avoid sequences that result in a
`strong propensity to aggregate. It is also interesting to con-
`sider that many short peptide sequences containing several
`hydrophobic residues and a high tendency for β-sheet
`formation probably have a strong disposition to form
`aggregates and/or amyloid fibrils. It is only the flanking
`sequences, which are either quite polar and therefore
`increase the solubility limit or are sufficiently bulky to
`sterically prevent the required interactions, that result in
`the lack of aggregation and/or amyloid formation.
`
`Problems due to protein aggregation
`Protein deposition diseases
`Several dozen protein deposition diseases are known.
`The most familiar include the amyloid diseases (amyloid-
`oses), such as Alzheimer’s disease, and the transmissible
`spongiform encephalopathies (TSEs; prion diseases such
`as bovine spongiform encephalopathy, BSE, or Mad Cow
`disease and Creutzfeldt–Jacob disease, CJD, in humans).
`In both amyloid and prion diseases the aggregated pro-
`tein is usually in the form of ordered fibrils. Amyloid fibril
`formation has been observed to arise from both peptides
`and proteins. Several protein deposition diseases involve
`non-ordered protein deposits; some examples are inclu-
`sion body myositis, light-chain deposition disease and
`cataracts. Many thousands of people die each year from
`protein deposition diseases [14]. New diseases are added
`to this list every year, one of the latest being Huntington’s
`disease [15].
`
`Inclusion bodies
`Inclusion body formation is very common when proteins
`are overexpressed. This may facilitate their potential
`purification because inclusion bodies are usually highly
`homogeneous. The problem is that renaturation is fre-
`quently difficult, as a result of aggregation. Several tech-
`niques have been developed to help overcome the
`common problem of their re-aggregation during renatura-
`tion [16] and some are discussed later. Inclusion bodies
`and related insoluble non-ordered protein aggregates are
`also found in certain diseases.
`
`R10 Folding & Design Vol 3 No 1
`
`Figure 1
`
`Absorbance
`
`1700
`
`1680
`
`1640
`1660
`Wavenumber (cm–1)
`
`1620
`
`1600
`
`Native IL-2
`Ammonium sulfate precipitated IL-2
`Aggregated IL-2
`
`Folding & Design
`
`Precipitates of native protein, in this case interleukin-2 (IL-2) formed by
`‘salting out’ with ammonium sulfate, retain the native conformation. The
`figure shows the second derivatives of the FTIR spectra of the amide I
`region of native, ammonium sulfate precipitated IL-2 (dashed line) and,
`for comparison, aggregated (inclusion body) IL-2 (dotted line). IL-2 is
`an all-α protein, as indicated by the dominant band at 1654 cm–1 for
`the native conformation.
`
`of protein concentration [3]. This study indicates that tran-
`sient aggregates can easily be mistaken for structured
`monomers and could be a general problem in time-resolved
`folding studies. Because aggregation is sensitive to protein
`concentration, monitoring the kinetics as a function of
`concentration should reveal potential aggregation artifacts.
`
`Key questions relating to aggregation, many of which are
`not yet fully answered, include the following: the nature
`of the species responsible for aggregation: the detailed
`mechanism that leads to aggregation and the underlying
`kinetics scheme; the structure of the aggregates; the speci-
`ficity of the intermolecular interaction (e.g. are the aggre-
`gates homogenous?); why aggregation (even of the same
`protein) sometimes leads to ordered aggregates (amyloid)
`and sometimes to disordered aggregates (inclusion bodies,
`
`Protein drugs
`Protein aggregation is also a problem in a number of other
`aspects of biotechnology; for example, during storage or
`delivery of protein drugs. There are several reports that
`protein aggregation can occur during lyophilization of pro-
`teins or during their subsequent rehydration, depending
`on the conditions (e.g. the water content of the system is
`critical [17–22]). Because in some cases it appears that the
`dehydration of proteins, which occurs in lyophilization,
`
`MYLAN INST. EXHIBIT 1096 PAGE 2
`
`MYLAN INST. EXHIBIT 1096 PAGE 2
`
`
`
`results in denaturation [23,24], it is probably the ensuing
`rehydration that leads to aggregation, due to the forma-
`tion of partially folded intermediates during the refolding
`(see below).
`
`Figure 2
`
`Native
`
`Review Protein aggregation Fink R11
`
`Ordered
`aggregates
`(amyloid fibrils)
`
`Amorphous
`aggregates
`(e.g. inclusion bodies)
`
`Folding & Design
`
`I
`
`Unfolded
`
`Mechanisms of aggregation
`One of the earliest and most prescient studies of protein
`aggregation was that of Goldberg and coworkers [25] on
`the enzyme tryptophanase, which revealed an intermedi-
`ate at moderate denaturant concentration that aggregated.
`Evidence for the potential specificity of aggregation was
`also observed in that the addition of other folded proteins
`did not affect the amount of aggregated tryptophanase.
`More recently, the idea that partially folded intermediates
`might be responsible for aggregation has been championed
`by King and coworkers [4,26,27] and Wetzel [6,8,10,28].
`Recent reports supporting the involvement of partially
`folded intermediates in the aggregation of several proteins
`suggests the generality of the phenomenon [29–31].
`
`Hypothesis
`Substantial data support the following model for the for-
`mation and structure of protein aggregates, in which spe-
`cific intermolecular interactions between hydrophobic
`surfaces of structural subunits in partially folded interme-
`diates are responsible for the aggregation. Key features of
`the model (Figure 2) are as follows. Protein folding
`involves intermediates, each consisting of an ensemble of
`closely related substates (the various substates of a given
`intermediate will be characterized by having common sec-
`ondary structure and most likely a common core of rela-
`tively native-like structure, with the remainder of the
`polypeptide chain disordered or in unstable structural
`units). The native state is formed by the sequential inter-
`action of substructural units, the building blocks (typically
`subdomains), which may be stable or metastable on their
`own, but are stabilized by the interactions with other such
`building blocks [32]. Formation of the native state
`involves the intramolecular interaction of the hydrophobic
`faces of structural subunits (Figure 3a). Specificity will
`arise from a variety of features, of which the geometric
`shape and extent of the hydrophobic patches, the con-
`straints of the polypeptide chain, and the presence of
`other structural subunits are probably the most important.
`Aggregation occurs when these hydrophobic surfaces
`interact in an intermolecular manner (Figure 3b). Thus,
`the initial stages of aggregation are quite specific in the
`sense that they involve the interaction of specific surface
`elements of the structural subunits of one molecule with
`‘matching’ hydrophobic surface areas of structural sub-
`units of a neighboring molecule. Three-dimensional prop-
`agation of this process leads to large aggregates. Initially,
`the aggregates (e.g. dimers and tetramers) will be soluble,
`but eventually their size will exceed the solubility limit.
`The fact that they may still have significant solvent-
`exposed hydrophobic surfaces would also minimize their
`
`The basic model for protein aggregation. The circled I represents a key
`partially folded intermediate, which can be populated either in the
`folding direction (from unfolded) or from the native state. The
`intermediate I has a strong propensity to aggregate, leading to either
`ordered or disordered (amorphous) aggregates. It is possible that
`there is very local order even in apparently (macroscopically)
`amorphous aggregates. Intermediate I will normally be an intermediate
`on both the in vivo and in vitro folding pathways.
`
`solubility. The intermediates are more prone to aggregate
`than the unfolded state because in the unfolded state the
`hydrophobic sidechains are scattered relatively randomly
`in many small hydrophobic regions, whereas in the par-
`tially folded intermediates there will be large patches of
`contiguous surface hydrophobicity, which will have a
`much stronger propensity for aggregation (these are the
`surfaces that ‘normally’ interact in an intramolecular
`manner to form the native conformation). The role of
`domain (or subdomain) swapping [33] in aggregation is
`unclear, but it is certainly possible that in some cases it
`may be an important factor.
`
`Aggregation often appears to be irreversible, but this is
`usually a reflection of the very slow rates of disaggregation
`and the fact that the equilibrium lies far in favor of the
`aggregate rather than its soluble monomeric form. Under
`certain conditions, aggregates, including in vivo amyloid
`deposits, can be reversed [34,35]. In practice, however,
`once insoluble aggregates form, the process is effectively
`irreversible under native-like conditions.
`
`A number of observations suggest that specific inter-
`actions are involved in protein aggregation. Among the
`clearest evidence for such specificity is the work of
`Brems and coworkers [36–40] with bovine growth hor-
`mone (bGH), which showed that a peptide fragment
`of bGH could inhibit aggregation and that mutations,
`which increased the hydrophobicity of the domain inter-
`face, increased the propensity for aggregation. Another
`strong indication of specificity in aggregation is the
`homogeneity of inclusion bodies [41], which are normally
`highly homogeneous once any contaminating material has
`
`MYLAN INST. EXHIBIT 1096 PAGE 3
`
`MYLAN INST. EXHIBIT 1096 PAGE 3
`
`
`
`R12 Folding & Design Vol 3 No 1
`
`Figure 3
`
`(a)
`
`Unfolded
`
`Intermediates
`
`(b)
`
`Unfolded
`
`Proposed mechanisms for aggregation
`involving structural building blocks
`(subdomains). See text for more details.
`(a) The normal folding situation, in which
`hydrophobic surfaces (black) of the structural
`units interact in an intramolecular manner to
`form the native conformation. Each identifiable
`intermediate along the pathway will consist of
`an ensemble of substates, having in common
`the compact structural unit(s), with the
`remainder of the chain in a disordered or
`unstable structured form. (b) The situation in
`which aggregation occurs by intermolecular
`interaction of the structural building blocks,
`again via interaction of complementary
`hydrophobic surfaces. The common
`intermediate in folding and aggregation can
`be seen.
`
`Native
`
`Aggregates
`
`Folding & Design
`
`been removed [41]. Similar arguments apply to amyloid
`fibril formation.
`
`Factors favoring aggregation
`Circumstances that lead to the population of partially
`folded intermediates, especially if their concentration is
`high, are thus likely to lead to aggregation; these circum-
`stances include mutations that lead to differential destabi-
`lization of the native state relative to the partially folded
`intermediate, as well as environmental conditions. Conse-
`quently, the major factors that determine whether a
`protein will aggregate, and the extent and rate of the
`aggregation, are: the protein amino acid sequence, the
`pH, the temperature and ionic strength, the concentration
`of the protein, the presence of cosolutes (e.g. denaturants
`such as urea, other chaotropes or kosmotropes — includ-
`ing osmolytes — and ligands that interact selectively with
`
`either the native or non-native conformations of the pro-
`tein or the aggregated form), and the presence (or absence)
`of various molecular chaperones. In fact, a major role of
`chaperones is to prevent the potential aggregation of newly
`synthesized proteins, or those denatured through stress.
`
`In most instances of aggregation, especially in vivo, there
`is a kinetic competition between aggregation and other
`processes, such as folding (Figure 4). The environmental
`conditions and the protein concentration significantly affect
`the degree and rate of intermolecular association. The
`protein concentration enters the equation because aggre-
`gation minimally involves a second-order kinetic process
`(although the growth of amyloid fibrils, and other aggre-
`gates, may appear first order). In the context of physiologi-
`cal aggregation, the potential role of post-translational
`processing may be critical. It is likely that in many cases
`
`MYLAN INST. EXHIBIT 1096 PAGE 4
`
`MYLAN INST. EXHIBIT 1096 PAGE 4
`
`
`
`Figure 4
`
`Degradation (proteolysis)
`
`Aggregates
`
`Partially
`folded
`intermediate(s)
`
`Native
`protein
`
`Nascent
`polypeptide
`
`Binding to
`chaperones
`
`Folding & Design
`
`Aggregation usually involves kinetic competition. During in vivo protein
`synthesis, for example, partially folded intermediates are divided
`between pathways leading to spontaneous folding and the native state,
`aggregation, binding to chaperones and proteolytic degradation.
`
`when aggregation occurs from a solution of the native
`protein it is the partially folded intermediates in equilib-
`rium with the native state that are the immediate precursors
`of the aggregates.
`
`The hypothesis that aggregation arises from partially
`folded intermediates explains the apparent lack of correla-
`tion between protein stability and aggregation that is some-
`times observed [42]. Thus, if the decreased stability of the
`native state of a mutant form also destabilizes the partially
`folded intermediate that is responsible for aggregation then
`a correlation may be observed. On the other hand, if desta-
`bilization of the native conformation increases the popula-
`tion of a partially folded intermediate that aggregates then
`increased aggregation will be observed and there will be no
`apparent correlation with the native-state stability.
`
`The propensity for a given protein to aggregate, either in
`vivo or in vitro, may well be determined in part by the
`lifetime of partially folded intermediates. Longer lived
`intermediates are more likely to lead to aggregation for
`two reasons: first, there is a greater chance of interaction
`with another such partially folded intermediate, and
`second, in the in vivo situation, the molecular chaper-
`ones involved in preventing aggregation by sequestering
`the partially folded intermediate may become saturated,
`and thus there will not be enough free chaperones avail-
`able to bind to additional newly synthesized protein.
`Several observations indicate that in vivo and in vitro
`aggregation during folding give rise to similar aggregates,
`suggesting that it is likely that a common partially folded
`
`Review Protein aggregation Fink R13
`
`intermediate is responsible for both types of aggregates
`[26,30,43,44].
`
`In vivo aggregates: inclusion bodies
`It is noteworthy that inclusion body formation is found
`not only in prokaryote and eukaryote cells, but also for
`both heterologous and homologous overexpression. This
`emphasizes that it is the overexpression itself that is
`responsible for the aggregation. For example, in vivo
`aggregation of β-lactamase is only observed when the rate
`of expression exceeds 2.5% of the total protein synthesis
`rate [45,46]. This mirrors the observations with in vitro
`refolding systems, which show that aggregation increases
`as the protein concentration increases [44]. Despite the
`fundamental and practical importance of inclusion bodies,
`rather little is known about their structures and mecha-
`nisms of formation. No correlation has been found
`between inclusion body formation in recombinant pro-
`teins and a wide variety of factors, including size and
`hydrophobicity, although some correlation with average
`charge and fraction of turn-forming residues has been
`observed [47]. Inclusion bodies are frequently refractory
`to renaturation. To obtain functional protein requires
`denaturation and solubilization, often with disulfide
`reducing agents, and subsequent renaturation. It is fre-
`quently observed that the only way to renature significant
`amounts of soluble material is to use extremely low pro-
`tein concentrations to avoid aggregation. As a result, inclu-
`sion bodies are a major problem in biotechnology and the
`development of protein drugs [48].
`
`Although many inclusion bodies are refractile in phase
`contrast microscopy, some large insoluble aggregates in
`Escherichia coli are not refractile, and have been called
`‘floccule-type’ inclusion bodies [49]. At least in some
`cases, these appear to be less tightly packed, and more
`easily solubilized than classical inclusion bodies. It is pos-
`sible that these aggregates arise from native protein of
`very limited solubility, rather than partially folded inter-
`mediates. We have noticed that there is a range in the
`denaturant solubility of in vivo insoluble proteins: classi-
`cal inclusion bodies are relatively resistant to solubiliza-
`tion, whereas some insoluble material is much more
`readily dissolved in denaturant and is clearly morphologi-
`cally different (S. Seshadri and A.L.F., unpublished
`observations). Morphological differences have also been
`observed between cytoplasmic and periplasmic inclusion
`bodies, reflected in differences in denaturant solubility
`and protease resistance [50].
`
`Several factors have been suggested to lead to inclusion
`body formation, including: the high local concentration
`of protein; a reducing environment in the cytoplasm; lack
`of post-translational modifications; improper interactions
`with chaperones and other enzymes involved in in vivo
`folding; and intermolecular crosslinking via disulfides
`
`MYLAN INST. EXHIBIT 1096 PAGE 5
`
`MYLAN INST. EXHIBIT 1096 PAGE 5
`
`
`
`R14 Folding & Design Vol 3 No 1
`
`(although inclusion bodies can form from proteins
`lacking cysteine). Three possible mechanisms for inclu-
`sion body formation have been proposed: aggregation of
`native protein of limited solubility; aggregation of the
`unfolded state; and aggregation of partially folded inter-
`mediate states. Until recently, there was no evidence for
`any regular structure in inclusion bodies and the prevail-
`ing view was that inclusion body formation was an aggre-
`gation process mediated by non-specific interactions in
`the unfolded state [10].
`
`In a series of elegant studies on the tailspike endorhamno-
`sidase of phage P22, King and coworkers [43,51–55] have
`shown that partially folded intermediates are responsible
`for the aggregation of tailspike protein as well as inclusion
`body formation in this unique system. Their results are
`consistent with a model in which aggregate formation,
`both in vivo and in vitro, occurs by the specific association
`of a partially folded intermediate that associates preferen-
`tially in an intermolecular fashion to form aggregates,
`rather than intramolecular association leading to the native
`conformation. A series of multimeric partially folded inter-
`mediates, representing early stages of the aggregation
`pathway for the P22 tailspike protein, have been trapped
`in the cold and isolated by nondenaturing polyacrylamide
`gel electrophoresis [26,27]. It should be noted that the P22
`tailspike protein system is an unusual one and not typical
`of most globular proteins, although similar ladders of mul-
`timers were observed with the P22 coat protein and car-
`bonic anhydrase II [27]. Monoclonal antibodies against
`tailspike chains that discriminate between folding inter-
`mediates and native states were used to distinguish pro-
`ductive folding intermediates and off-pathway aggregation
`intermediates. The aggregation intermediates displayed
`epitopes in common with productive folding intermedi-
`ates, but were not recognized by antibodies against native
`epitopes [55]. Because many other protein expression
`systems also show increased inclusion body formation as
`the temperature is raised, it is likely that ‘thermolabile’
`intermediates are common precursors to inclusion bodies
`at elevated temperatures.
`
`There are several protein systems in which it has been
`shown that point mutations may dramatically affect the
`amount of aggregate (inclusion body) formation; these
`include the P22 tailspike protein [53], interferon-γ [56],
`colicin A [57], Che Y [58], and interleukin-1β (IL-1β)
`[56,59]. For example, with the human IL-1β expression
`system in E. coli, the amount of inclusion body ranges
`from close to 0% for the native to close to 100% for the
`Leu10→Thr and Lys97→Val mutants [60]. No strong
`correlations were observed between the extent of inclu-
`sion body formation and either thermodynamic or thermal
`stability. This and other investigations suggest that the
`tendency of at least some proteins to form inclusion
`bodies is related to the stability or solubility of folding
`
`intermediates rather than native states [42]. The in vivo
`effects of sequence and growth temperature of IL-1β, as
`manifested by inclusion body formation, were quantita-
`tively reproduced in an in vitro system, indicating not
`only that inclusion body formation is based on the intrin-
`sic properties of the protein sequence, but also that inter-
`actions with chaperones or other cellular factors were not
`significant [59].
`
`Consistent with the proposed model, point mutations in
`the hydrophobic face of the immunoglobulin light chain,
`in which polar residues were introduced, significantly
`decreased the fraction of inclusion body formed in an
`E. coli expression system [61]. The potentially strong
`dependence of inclusion body formation on point muta-
`tions may also be a reflection of specificity in aggregation.
`
`We have used attenuated total reflectance (ATR) FTIR
`to examine the structure in various aggregated protein
`forms, including inclusion bodies [30]. These included
`all-β, all-α and α+β proteins. Some common features
`were observed. First, both the inclusion bodies and the
`folding aggregates exhibited substantial secondary struc-
`ture, typically 50–70% of that of the native conformation.
`We interpret this to mean that the inclusion bodies arose
`from the association of partially folded intermediates con-
`taining substantial native-like structure. Second, the
`structure of a given protein in inclusion bodies, refolding
`aggregates, and thermal aggregates (formed by heating a
`concentrated protein solution to a temperature just below
`the beginning of the thermal denaturation transition) is
`the same (Figure 5). Thus, we conclude that a given
`protein will have one partially folded intermediate that is
`particularly prone to aggregate, and consequently most
`aggregates of that protein will arise from that intermediate
`and thus have similar structures.
`
`Furthermore, in all cases, even for all-β proteins, signifi-
`cant new β structure, compared to that in the native confor-
`mation, was observed; typically this amounted to ~20–25%
`of the total secondary structure. We take this to indicate
`that the intermolecular interactions leading to the aggrega-
`tion involve β-sheet-like interactions. The exact nature of
`the intermolecular interactions is unknown, and could be
`different in different aggregates; for example, the number
`of strands in the β sheet may vary. What is clear, however,
`is that aggregation usually leads to an increase in the
`amount of secondary structure with IR components in the
`1623–1637 cm–1 region, which corresponds to the region
`where β-sheet structure is observed. Interestingly, the
`amount of secondary structure in the inclusion body varies
`from one protein to another, as does the amount of disor-
`dered structure. Figure 5 shows the spectra of the native
`inclusion body and folding aggregate forms of IL-2, illus-
`trating the dramatic increase in β-sheet content compared
`to the native conformation.
`
`MYLAN INST. EXHIBIT 1096 PAGE 6
`
`MYLAN INST. EXHIBIT 1096 PAGE 6
`
`
`
`Review Protein aggregation Fink R15
`
`the inclusion bodies were similar regardless of whether
`they were localized in the cytoplasm or periplasm.
`
`In vivo aggregates: amyloid
`A number of human diseases involve the deposition of pro-
`tein aggregates; a subset of these, known as amyloidoses,
`are lethal diseases involving the extracellular deposition of
`amyloid fibrils and plaques. Amyloid plaque is usually
`defined by three features: characteristic birefringent stain-
`ing using Congo Red, fibrous morphology observed by
`electron microscopy, and a distinctive X-ray fiber diffrac-
`tion pattern similar to that of certain insect silks, consistent
`with high β-sheet content (the so-called cross-β structure).
`Amyloid fibrils are typically 7–12 nm in diameter and can
`be dissociated by high concentrations of denaturant.
`
`The molecular mechanisms leading to pathological amy-
`loid formation are not well understood. In most forms of
`amyloidosis the formation of amyloid appears to be caused
`by a combination of factors, including high concentrations
`of protein, proteolysis, mutations, and other unknown
`factors. Certain other molecules are frequently found in
`association with amyloid plaques, namely serum amyloid
`P component [63], proteoglycans [64], apolipoprotein E
`[65] and metal ions. The role of these molecules, if they
`have one, in amyloid deposition is unclear at present.
`There have been reports that these and other ligands may
`increase or decrease aggregation (e.g. [66,67]).
`
`Evidence to support partially folded intermediates as
`amyloid fibril precursors has been found in at least two
`systems. Transthyretin (TTR) is involved in two amyloid
`diseases, familial amyloidotic neuropathy and senile sys-
`temic amyloidosis. A partially folded monomeric interme-
`diate, populated at low pH, has been implicated in in vitro
`amyloid formation from transthyretin [68,69]. Recently,
`two naturally occurring variants of human lysozyme have
`been found to be amyloidogenic [70]. Both mutants are
`significantly less stable than the wild-type protein and
`have populated partially folded intermediates at 37°C that
`are assumed to be the precursors of amyloid. The ther-
`mally aggregated mutants showed an increased level of
`β structure compared to the native conformation.
`
`Some mutations leading to amyloid-fibril formation (in
`transthyretin [71,72], lysozyme [70] and immunoglobulin
`variable light, VL, chain domains [73]) are also observed to
`result in decreased stability of the native state. Presumably,
`in such cases there is differential destabilization of the
`native conformation compared to the aggregating intermedi-
`ate conformation that results in population of the aggregat-
`ing conformation. It has been shown, in an in vitro system,
`that point mutations may convert a non-amyloidogenic pro-
`tein into an amyloidogenic one [29,73]. In Alzheimer’s
`disease, the Aβ40 peptide is significantly less amyloidogenic
`than the Aβ42 peptide [74]. The intermediacy of partially
`
`Figure 5
`
`Absorbance
`
`β
`
`α
`
`1700
`
`1680
`
`1640
`1660
`Wavenumber (cm–1)
`
`1620
`
`1600
`
`Inclusion bodies
`Folding aggregates
`Native protein
`
`Folding & Design
`
`Aggregation results in a substantial increase in β structure compared
`to the native conformation. This figure shows the second derivatives of
`the FTIR spectra of the amide I region of IL-2. The aggregated forms,
`inclusion bodies and folding aggregates, show 25% more β structure
`than the native state. The spectra of the two aggregated forms are very
`similar, indicating similar underlying secondary structure.
`
`Our results, along with those of Georgiou and coworkers
`[44,62] on β-lactamase (see below), suggest that the gener-
`ation of insoluble aggregates during refolding from denat-
`urant can be considered to be an in vitro model system for
`inclusion body formation.
`
`The secondary structure of β-lactamase (an α+β protein)
`inclusion bodies in E. c