Accelerating the world's research.
`
`The 1.7 Å crystal structure of BPI: a
`study of how two dissimilar amino
`acid sequences can adopt the same
`fold
`
`Parag Mallick
`
`Journal of Molecular Biology
`
`Cite this paper
`
`Downloaded from Academia.edu 
`
`Get the citation in MLA, APA, or Chicago styles
`
`Related papers
`
`Download a PDF Pack of the best related papers 
`
`PFIT and PFRIT: Bioinformatic algorithms for detecting glycosidase function from structure …
`Parag Mallick
`
`Sequence and structural analysis of cellular retinoic acid-binding proteins reveals a network of conse…
`Lila Gierasch, Kannan Gunasekaran
`
`3DSIG 2010: The 6th Structural Bioinformatics and Computational Biophysics Meeting
`Mihir Solanki
`
`Exhibit 2069
`Page 01 of 17
`
`

`

`The 1.7 AÊ Crystal Structure of BPI: A Study of How
`Two Dissimilar Amino Acid Sequences can Adopt the
`Same Fold
`
`Gary Kleiger1, Lesa J. Beamer2, Robert Grothe1, Parag Mallick1
`and David Eisenberg1*
`
`1UCLA-DOE Laboratory of
`Structural Biology and
`Molecular Medicine, Molecular
`Biology Institute, UCLA
`BOX 951570, Los Angeles
`CA 90095-1570, USA
`2Biochemistry Department
`University of Missouri-
`Columbia, Columbia
`MO 65211, USA
`
`We have extended the resolution of the crystal structure of human bac-
`to 1.7 AÊ . BPI has two
`tericidal/permeability-increasing protein (BPI)
`domains with the same fold, but with little sequence similarity. To under-
`stand the similarity in structure of the two domains, we compare the
`corresponding residue positions in the two domains by the method of
`3D-1D pro®les. A 3D-1D pro®le is a string formed by assigning each pos-
`ition in the 3D structure to one of 18 environment classes. The environ-
`ment classes are de®ned by the local secondary structure, the area of the
`residue which is buried from solvent, and the fraction of the area buried
`by polar atoms. A structural alignment between the two BPI domains
`was used to compare the 3D-1D environments of structurally equivalent
`positions. Greater than 31 % of the aligned positions have conserved 3D-
`1D environments, but only 13 % have conserved residue identities. Anal-
`ysis of the 3D-1D environmentally conserved positions helps to identify
`pairs of residues likely to be important in conserving the fold, regardless
`of the residue similarity. We ®nd examples of 3D-1D environmentally
`conserved positions with dissimilar residues which nevertheless play
`similar structural roles. To generalize our ®ndings, we analyzed four
`other proteins with similar structures yet dissimilar sequences. Together,
`these examples show that aligned pairs of dissimilar residues often share
`similar structural roles, stabilizing dissimilar sequences in the same fold.
`# 2000 Academic Press
`
`*Corresponding author
`
`Keywords: BPI; X-ray crystallography; 3D-1D environment; domain; fold
`
`Introduction
`
`Proteins with sequence similarity also display
`structural similarity. However, many proteins with
`no apparent sequence similarity display the same
`folds. For example, the mitochondrial enzyme rho-
`danese contains two domains of similar structure,
`but little sequence similarity (Hol et al., 1983). This
`phenomenon is not rare. In fact, the database of
`distant aligned protein structures (DAPS) has over
`1000 examples of structurally similar proteins with
`less than 25 % sequence identity (Rice & Eisenberg,
`1997). Examples of both intra and inter-molecular
`
`fold similarity in the absence of amino acid simi-
`larity is given in Table 1.
`To study how dissimilar protein sequences
`adopt similar folds, we analyze the structure of the
`bactericidal/permeability-increasing protein (BPI).
`BPI has two domains with the same fold but with
`dissimilar
`sequences. Both BPI domains
`are
`twisted, anti-parallel b-sheet barrels capped by two
`a-helices. The domain main-chain atoms can be
`superimposed without
`signi®cant deformation
`(3.0 AÊ rmsd over 173 residues). The BPI domain is
`to date a unique fold. We take advantage of the
`apparent domain duplication in BPI to ®nd struc-
`turally conserved positions for the BPI domain.
`BPI is a mammalian protein located in polymor-
`phonuclear neutrophils, a cell of
`the
`innate
`immune response that protects the host during
`microbial
`infection (Elsbach & Weiss, 1995). BPI
`speci®cally binds lipopolysaccharides in the outer-
`membrane of Gram-negative bacteria. Although
`
`

`

`1020
`
`How Dissimilar Sequences Adopt the Same Fold
`
`Table 1. Examples of domains with similar folds but dissimilar amino acid sequences
`
`Protein 1
`
`Protein 2
`
`rmsd (AÊ )
`
`% Seq. ID
`
`Reference
`
`Hexokinase domain I
`Rhodanese domain I
`Rhizopuspepsin domain I
`Phosphoglycerate kinase
`dom. I
`Arabinose-binding protein
`dom. I
`Bovine F1-ATPase
`Bromoperoxidase A2
`b-amylase
`Cellobiohydrolase I
`Neuroglian
`
`Hexokinase domain II
`Rhodanese domain II
`Rhizopuspepsin domain II
`PGK domain II
`
`Arabinose domain II
`
`Rec A protein
`Triacylglycerol hydrolase
`Concanavalin B
`Serum amyloid component
`T cell antigen receptor
`
`2.8
`1.8
`3.0
`4.5
`
`3.2
`
`3.2
`2.7
`4.1
`3.4
`3.3
`
`11
`13
`13
`8
`
`7
`
`9
`14
`10
`10
`11
`
`Steitz et al. (1976)
`Ploegman et al. (1978)
`Subramanian et al. (1977)
`Banks et al. (1979)
`
`Quiocho et al. (1977)
`
`Abrahams et al. (1994); Story et al. (1992)
`Hecht et al. (1994); Uppenberg et al. (1994)
`Hennig et al. (1995); Mikami et al. (1994)
`Divne et al. (1994); Emsley et al. (1994)
`Bentley et al. (1995); Huber et al. (1994)
`
`The ®rst ®ve examples are similar folds where each domain is from the same protein (rms deviations are for superimposed
`main-chain atoms); the last ®ve examples are similar folds where each domain is from a different protein (rmsd are for superim-
`posed Ca atoms). This table has been adapted and expanded from Richardson (1981).
`
`BPI has two domains, a single BPI domain can
`adopt a stable fold: an N-terminal fragment has
`been expressed and retains the in vitro activity of
`wild-type BPI (Horwitz et al., 1996). The BPI X-ray
`structure was determined in our laboratory to
`2.4 AÊ resolution at room temperature (Beamer et al.,
`1997), and here is extended to 1.7 AÊ using cryo-
`crystallography diffraction data.
`Here, we use 3D-1D environment classes to
`identify the structurally equivalent positions in the
`two domains that have similar atomic environ-
`ments. 3D-1D environment classes have been used
`successfully for tasks such as protein sequence fold
`assignment, and assessing the quality of protein
`structures derived from NMR or X-ray crystallo-
`graphy experiments (Bowie et al., 1991; Luthy et al.,
`1992). The 3D-1D environment classes describe the
`protein structure, where each environment
`is
`de®ned by the local secondary structure, the sol-
`vent accessibility, and the fraction of local atoms
`that are polar for each residue. Our hypothesis is
`that structurally equivalent positions with identical
`environments are important
`for conserving the
`fold. While one might expect to ®nd positions with
`conserved structural environments in the core of
`each BPI domain, we show here that structurally
`equivalent surface positions with dissimilar residue
`types can conserve structural
`roles. With the
`environment classes determined from our 1.7 AÊ
`model, we compare structurally equivalent pos-
`itions between the two domains of BPI and exam-
`ine the physical-chemical properties that determine
`the unusual BPI structure.
`
`Results
`
`Extension of the resolution and refinement of
`the 1.7 AÊ structure of BPI
`
`Using cryo-crystallography as described in
`Materials and Methods, the resolution of the BPI
`structure was extended to 1.7 AÊ and re®ned to an
`R-factor of 0.198 with an Rfree of 0.249. Comparison
`
`of the high-resolution model of BPI with the room-
`temperature model reveals little structural change,
`with the exception of residues 42 to 48. In the new
`structure, several side-chains on one side of the
`loop now pack against the protein, whereas these
`side-chains in the previous model were mostly
`exposed to solvent. The loop rearrangement may
`be due to conditioning the crystals with 45 % PEG
`6000 for cryo-protection, freezing, or a combination
`of the two. Equivalent main-chain atoms for the
`two models superimpose with an rmsd of 0.9 AÊ .
`The ribbon diagram is shown in Figure 1(a).
`
`Comparison of the N-terminal and C-terminal
`domains of BPI using 3D-1D environments
`
`The structural alignment of the N-terminal and
`C-terminal BPI domains was used to generate a
`sequence alignment between the two domains
`(Figure 2). Conservation between the two domains
`was ®rst examined at the residue level. A total of
`21 pairs of residues are identical out of 164 aligned
`positions between the two domains, corresponding
`to a sequence identity of 13 %. This level of
`sequence identity is
`signi®cantly higher
`than
`would be expected for two independently gener-
`this length (Z-score
`ated random sequences of
`equals 3.6, see Materials and Methods). Even so,
`the sequence identity is too low to allow sequence
`alignment methods to predict an alignment, and in
`fact the structural identity of the two domains was
`unsuspected prior to the 2.4 AÊ resolution structure.
`To understand the fold identity of the two BPI
`domains in the absence of strong sequence simi-
`larity, positional similarities were examined using
`3D-1D environment classes. The environments for
`each position in the two domains are shown on the
`structure-derived sequence alignment of the two
`domains (Figure 2). We reasoned that structurally
`equivalent positions with similar environments
`might also have similar structural roles in each BPI
`domain. Of the 164 structurally aligned positions,
`51 have identical 3D-1D environments, correspond-
`
`Exhibit 2069
`Page 03 of 17
`
`

`

`How Dissimilar Sequences Adopt the Same Fold
`
`1021
`
`Figure 1. (a) Ribbon representation of the 1.7 AÊ crystal structure of human BPI. The N-terminal domain is blue.
`The C-terminal domain is red. Residues 10 to 193 fold into a structural element called the N-terminal barrel. The bar-
`rel is composed of ®ve anti-parallel b-strands which twist about the barrel axis. Two a-helices complete the barrel by
`closing a gap in the b-sheet. Residues 260 to 430 fold into a similar structure called the C-terminal barrel. Amino acid
`residues 201 to 229, as well as 431 to 456, fold into the central b-sheet of six strands, located in the center of the mol-
`ecule, which interacts with both the N-terminal and C-terminal barrels. A linker of residues 230 to 250 (olive) con-
`nects the N-terminal and C-terminal domains. (b) Superposition of the N-terminal domain (blue) on the C-terminal
`domain (red). Residues 1-229 were structurally aligned to residues 251-456 using the algorithm ALIGN V2 (Cohen,
`1986). The two domains align with 3.0 AÊ root mean square deviation over the main-chain atoms of the 173 structu-
`rally corresponding residues. (c) Schematic of BPI showing its elongated shape and two-domain structure. The two
`domains are related by a pseudodyad perpendicular to the page. Secondary structure units are represented by arrows
`(b-strands) and rectangles (helices). The N-terminal domain (residues 1-229) is gray; the C-terminal domain (residues
`251 to 456) is black. Secondary structure units have been numbered, with the primes denoting the units in the
`C-terminal domain. Residue positions for the start and end of each secondary structure unit are shown. The three
`subdomains (N-terminal barrel, C-terminal barrel, and central sheet) are shown.
`
`ing to 31 % of the structurally equivalent positions
`in the BPI domain alignment. This level of conser-
`vation is highly signi®cant relative to an alignment
`of two independently generated random pro®les of
`this length (Z-score ˆ 12.9. A fuller statistical anal-
`ysis of
`the BPI domain alignment
`is given in
`Materials and Methods).
`
`Analysis of environmentally conserved
`positions in the BPI alignment as a function of
`environment class
`
`We then asked if certain environment classes are
`more conserved than others in the BPI alignment.
`We use a p-value which is the probability that at
`
`Exhibit 2069
`Page 04 of 17
`
`

`

`1022
`
`How Dissimilar Sequences Adopt the Same Fold
`
`Figure 2. Structure-based alignment of the amino acid sequences of the N-terminal and C-terminal domains of BPI.
`Both the sequence and corresponding 3D-1D environments are shown. The N-terminal domain (residues 1 to 229) is
`the top sequence; the C-terminal domain (residues 251 to 456) is the bottom of each dual line. If the 3D-1D environ-
`ments are identical for a given pair in the alignment, the two positions are considered structurally conserved and
`colored black. Consecutive stretches of 3D-1D-conserved residues are labeled blocks I-V. The nine positions that were
`removed from the initial alignment have stars located above the corresponding N-terminal position.
`
`least M out of N pairs of environments from the
`BPI domain alignment would be the same if the
`environments were paired at
`random. Small
`p-values indicate correlation between the environ-
`ments of structurally aligned positions; p-values
`for each environment class are shown in Table 2.
`Notice that residues important for stabilizing the
`cores of proteins tend to be hydrophobic and bur-
`ied in apolar environments,
`that
`is,
`in the B1
`
`environment class (see Materials and Methods and
`Table 5 for the de®nition of each environment
`class). We would then expect to observe a low
`p-value for B1-B1 pairs because the structural role
`of residues that pack in the protein core tend to be
`conserved. The p-values for the H:B1 and E:B1
`environment classes are 3  106 and 8  106,
`respectively. Therefore, one would expect
`to
`observe at least as many H:B1 matches for the BPI
`
`Exhibit 2069
`Page 05 of 17
`
`

`

`How Dissimilar Sequences Adopt the Same Fold
`
`1023
`
`Table 2. Log-odds and fractional weighted log-odds (FWLO) values for 3D-1D environment classes (Env. Class) for
`the alignment of the N and C-terminal domains of BPI and for all alignments from the DAPS database
`
`BPI
`Env. Class
`
`H:E
`H:P2
`H:P1
`H:B3
`H:B2
`H:B1
`E:E
`E:P2
`E:P1
`E:B3
`E:B2
`E:B1
`C:E
`C:P2
`C:P1
`C:B3
`C:B2
`C:B1
`
`P-value
`
`5  102
`2  108
`NO
`NO
`NO
`3  106
`2  102
`1  105
`1  101
`NO
`NO
`8  106
`1  101
`2  101
`NO
`NO
`NO
`NO
`
`Log-odds
`
`FWLO
`
`DAPS database
`Log-odds
`
`FWLO
`
`1.7
`2.0
`NO
`NO
`NO
`3.1
`1.4
`1.0
`0.9
`NO
`NO
`1.3
`1.1
`0.9
`NO
`NO
`NO
`NO
`
`0.05
`0.24
`NO
`NO
`NO
`0.17
`0.06
`0.20
`0.04
`NO
`NO
`0.19
`0.03
`0.02
`NO
`NO
`NO
`NO
`
`1.4
`1.2
`1.0
`1.0
`1.0
`1.5
`2.1
`1.9
`1.3
`1.5
`1.3
`1.7
`1.4
`0.9
`0.9
`1.0
`1.1
`1.3
`
`0.03
`0.11
`0.06
`0.02
`0.02
`0.16
`0.01
`0.05
`0.04
`0.03
`0.03
`0.23
`0.07
`0.06
`0.03
`0.02
`0.01
`0.02
`
`The log-odds gives the log of the ratio of the observed to expected probability for the pairing of two identical environmental
`classes. It is a measure of the likelihood that any instance of an environment class is conserved in the alignment. The FWLO is the
`log-odds weighted by the observed probability for the pairing of two identical environment classes in fractional form. A given pair
`may have a positive log-odds value, but if it occurs rarely in the structural alignment, the FWLO value will be low. Notice that the
`FWLO values for the BPI alignment and the DAPS alignments follow similar trends; the H:B1, E:B1, and H:P2 have high FWLO
`values. See Table 5 for de®nitions of the environment classes. NO, no observations.
`
`domains in a random alignment only three out of
`one million times. Thus, positions that belong to
`the H:B1 or E:B1 classes are observed more often
`than expected at random, suggesting these pos-
`itions have conserved structural roles (a more
`thorough discussion of Table 3 is given in
`Materials and Methods.)
`The same argument can be applied to the surface
`of the protein. Positions which belong to the P2
`environment class tend to be solvent exposed at
`the protein's surface. If the environment at the sur-
`face is conserved, we would expect to observe low
`p-values for the P2 class. The p-values for the H:P2
`and E:P2 environment classes are 2  108 and
`1  105, respectively, which are as low as the
`values shown above for the H:B1 and E:B1 classes.
`While the structural conservation of residues in the
`protein core is well documented, the conservation
`of residues on the surface of proteins is not as well
`described. We next characterize the structural roles
`of residues in these positions, especially pairs from
`the BPI domain alignment which do not conserve
`residue type.
`
`The structural roles of 3D-1D environmentally
`conserved positions with dissimilar residues
`in the BPI domain alignment
`
`Thirty-one of the 51 3D-1D environmentally con-
`served pairs from the BPI domain alignment have
`similar or identical residues, de®ned by a positive
`substitution score from the GONNET matrix
`(Benner et al., 1994). However, 20 structurally
`equivalent
`positions with
`conserved
`3D-1D
`environments have dissimilar residues. We next
`
`wished to characterize the structural roles for resi-
`dues at these 20 positions and determine if any
`could be important for stabilizing the BPI fold.
`The structural roles for 3D-1D environmentally
`conserved positions with dissimilar residues were
`compared by analyzing the tertiary interactions
`they form with other residues. We analyzed favor-
`able contacts,
`including hydrogen bonds,
`salt
`bridges, disul®de bonds, van der Waals inter-
`actions, or aromatic ring stacking. We then deter-
`mined for each pair in the alignment whether the
`structural roles are similar.
`We ®nd that the structural roles for these pairs
`fall into three major categories: (i) conserved struc-
`tural roles; (ii) auxiliary structural roles; and (iii)
`different structural roles.
`
`Conserved structural roles
`
`The structural roles for a pair of dissimilar resi-
`dues are de®ned as conserved when a residue at
`position i in the N-terminal domain of BPI forms a
`tertiary interaction with residue j in the N-terminal
`domain, and residue i0 in the C-terminal domain
`forms a tertiary interaction with residue j0. Both
`residues i and i0 and j and j0 must be at structurally
`equivalent positions in the alignment for the struc-
`tural roles to be de®ned as conserved. The type of
`tertiary interaction does not have to be the same
`(Figure 3(a)).
`
`Auxiliary structural roles
`
`Residue i forms a tertiary interaction with resi-
`due j in the N-terminal domain, and residue i0 in
`
`Exhibit 2069
`Page 06 of 17
`
`

`

`1024
`
`How Dissimilar Sequences Adopt the Same Fold
`
`Table 3. Classi®cation of the structural roles for equivalent positions with conserved 3D-1D environment classes and
`dissimilar residue types. Comparison of structural roles for the N-terminal (n) and C-terminal (c) domains of BPI.
`
`Residue (n)
`
`Atom
`
`Structural role
`
`Residue (c)
`
`Atom
`
`Structural role
`
`N-terminal domain
`
`C-terminal domain
`
`Ser112
`
`OG
`
`A. Conserved structural roles
`Ser79
`OG
`H bond to Ser69 OG through
`a water molecule
`H bond to bb carbonyl group
`of His138 through a water
`molecule
`H bond to Ser134 OG
`H bond to Thr216 OG1
`
`Asp116
`Thr219
`
`OD2
`OG1
`
`Tyr336
`
`His361
`
`Ser365
`Phe446
`
`OH
`
`CE1
`
`OG
`-
`
`H bond to Ser327 OG
`
`VdW with Leu385 CB and
`CD1
`
`H bond to Lys380 NZ
`Ring stacking with His443
`
`H bond to bb carbonyl group
`of Phe262
`VdW with Glu272 CD and
`CG
`H bond to bb amide group of
`Met 312
`H bond to Gln329 NE2
`VdW with Met360 CB
`
`VdW with Arg416 CZ and
`Leu414 CB
`Salt bridge to Arg416 NH2
`Solvent exposed
`VdW with Val453 CG1
`VdW with Leu440 CB and
`CD1 and Val368 CG1
`VdW with Tyr436 CE1
`
`H bond with Arg252 NH1
`and Tyr255 OH
`VdW w/Pro241 CG,CD
`H bond to bb carbonyl group
`of Ala370
`VdW with His361 CE1
`VdW with Leu447 CD1
`
`B. Auxiliary structural roles
`Tyr16
`OH
`H bond to Gln20 NH2
`
`Thr265
`
`OG1
`
`Glu19
`
`Asp36
`
`Lys77
`Ala83
`
`Asn180
`
`Ser184
`Pro188
`Asn206
`Val222
`
`Glu227
`
`NE2
`
`H bond to Gln20 OE1
`
`Leu268
`
`CG,CD1
`
`OD1
`
`H bond to bb amide of Ser55
`
`Thr294
`
`NZ
`CB
`
`ND2
`
`OG
`-
`CG
`CG1
`
`OE1
`
`Salt bridge to Asp116 OD1
`VdW with Leu63 CD2 and Ile
`80 CG2
`H bond to Ser184 OG
`
`H bond to Asn 180 ND2
`Solvent exposed;
`VdW with Lys225 CB and CG
`VdW with Leu209 CG and
`CD1
`Salt bridge to Lys225 NZ
`
`Thr334
`Val339
`
`Pro415
`
`Glu419
`Lys423
`Val433
`Phe449
`
`Val454
`
`OG1
`
`OG1
`CG1
`
`CB,CD
`
`OE2
`-
`CG1, CG2
`CD1,CE1,CE2
`
`CB
`
`C. Different Structural Roles
`Val1
`-
`Solvent exposed
`
`Asp251
`
`OD1
`
`Lys12
`Gly120
`
`His138
`Thr215
`
`NZ
`-
`
`-
`OG
`
`Salt bridge with Asp452 OD1
`Solvent exposed;
`
`Solvent exposed
`H bond to ordered water
`
`Tyr261
`Ser369
`
`Leu385
`Pro442
`
`CG,CD1,CD2,CE2,CZ
`OG
`
`CB,CD1
`CB,CG
`
`H bond, Hydrogen-bond interaction; VdW, a van der Waals interaction; backbone, bb.
`
`the C-terminal domain forms a tertiary interaction
`with residue k0. Residues i and i0 must occupy
`structurally equivalent positions, but residues j and
`k0 do not have to be aligned (Figure 3(b)).
`
`Different structural roles
`
`Residues i and i0 belong to the same environ-
`ment class but one or both of the residues have no
`corresponding structural role, or one or both resi-
`dues are involved in stabilizing the interface
`between the domains in the BPI structure rather
`than the domain itself (Figure 3(c)). A summary of
`the structural roles for the 20 pairs of dissimilar
`residues is given in Table 3.
`Four pairs of 3D-1D environmentally conserved
`positions with dissimilar residue types fall into the
`conserved structural roles category. For example,
`Ser79 and Tyr336 both belong to the E:P2 environ-
`ment class, however the substitution of serine for
`tyrosine in the GONNET matrix (Benner et al.,
`1994) is not favorable. The conserved structural
`roles of both Ser79 and Tyr336 are to stabilize the
`domain by connecting adjacent b-strands. These
`residues are located on b-strands 4 and 40 respect-
`
`ively (see Figures 1(c) and 4(a)). Ser79 (N-terminal
`domain) forms a 2.7 AÊ hydrogen bond to a water
`molecule which is hydrogen bonded to Ser69,
`located on b-strand 3. The crystallographic tem-
`perature factor for this water molecule is 27 AÊ 2
`suggesting that it is well ordered. In the C-terminal
`domain, Tyr336 forms a 2.8 AÊ hydrogen bond to
`Ser327, which lies on b-strand 30. Ser69 and Ser327
`are also paired in the BPI domain alignment.
`Therefore, Ser79 and Tyr336 interact with structu-
`rally equivalent
`residues
`in their
`respective
`domains, conserving the structural role for these
`dissimilar residues.
`A similar example involves Asp116 and Ser365,
`and 50,
`respectively
`located on b-strands
`5
`(Figure 4(b)). The conserved structural roles of
`Asp116 and Ser365 connect adjacent b-strands in
`their respective domains. Asp116 forms a 2.9 AÊ
`hydrogen bond to Ser134 on b-strand 6. Ser365
`forms a 3.1 AÊ hydrogen bond to Lys380 on
`b-strand 60. Ser134 and Lys380 are paired in the
`BPI domain alignment.
`Conserved structural roles do not necessarily
`involve identical tertiary interactions. For example,
`Thr219 and Phe446 are structurally equivalent resi-
`
`Exhibit 2069
`Page 07 of 17
`
`

`

`How Dissimilar Sequences Adopt the Same Fold
`
`1025
`
`Figure 3. Pairs of 3D-1D environmentally conserved
`positions with dissimilar residues fall into three major
`categories. Each category describes the similarity of the
`structural role for each residue in the pair. (a) Con-
`served structural roles: residue i from the N-terminal
`domain interacts with residue j, and residue i0
`in the
`C-terminal domain interacts with residue j0. Both resi-
`dues i and i0 and j and j0 must be at structurally equival-
`ent positions in the alignment. (b) Auxiliary structural
`roles: residue i from the N-terminal domain interacts
`with residue j, and residue i0 in the C-terminal domain
`interacts with residue k0. Residue i and i0 must occupy
`structurally equivalent positions. (c) Different structural
`roles: either residue i or i0 is solvent exposed, or residue
`i from the N-terminal domain interacts with residue l0
`from the C-terminal domain.
`
`dues with identical environment classes (E:P2).
`Thr219 forms a 2.7 AÊ hydrogen bond with Thr216
`which stabilizes a tight turn from b-strand 8 to
`b-strand 9 (Figure 4(c)). In the C-terminal domain,
`the aromatic ring of Phe446 is approximately 4 AÊ
`from the ring of His443. Ring stacking of phenyl-
`alanine and histidine residues has been shown to
`be an energetically favorable interaction (Mitchell
`et al., 1994). Therefore, the interaction of Phe446
`with His443 stabilizes a tight turn from b-strand 70
`to b-strand 80, similar to the interaction found in
`the N-terminal domain.
`Eleven 3D-1D environmentally conserved pos-
`itions with dissimilar residue types have auxiliary
`structural roles in the BPI domains. The structural
`roles of these residue pairs are auxiliary because
`they help stabilize the fold, even though the struc-
`tural roles are not strictly conserved. An example
`is Lys77 and Thr334, located in b-strand 4 and 40
`respectively (Figure 1(c)). Lys77 forms a salt bridge
`
`Figure 4. Examples of pairs of 3D-1D environmentally
`conserved residues with dissimilar residues yet con-
`served structural roles from the BPI domain alignment.
`Residues and secondary structure elements from the
`N-terminal domain are blue; residues and secondary
`structure elements from the C-terminal are red. The
`relationship of the secondary structure elements to the
`entire BPI molecule is shown in Figure 1(c). Hydrogen
`bonds are shown as broken lines between the hydrogen
`bond donor and acceptor atoms.
`
`with Asp116 located on b-strand 5. Thr334 forms a
`hydrogen bond to Gln329 located on b-strand 30.
`These interactions still support the fold of each BPI
`domain by connecting adjacent b-strands.
`Five of the 3D-1D environmentally conserved
`positions with dissimilar residue types have differ-
`ent structural roles in the two domains. Some of
`these positions, such as His138, are completely
`exposed to bulk solvent and make no contacts
`with any ordered atoms in the BPI structure,
`whereas the structurally equivalent residue from
`the C-terminal domain, Leu385, makes van der
`Waals contacts with His361.
`Another example of a pair of residues with
`different structural roles is Lys12 and Tyr261.
`Lys12 forms a salt bridge with Asp452, connecting
`
`Exhibit 2069
`Page 08 of 17
`
`

`

`1026
`
`How Dissimilar Sequences Adopt the Same Fold
`
`the N-terminal domain of BPI with the C-terminal
`domain. Tyr261 forms van der Waals contacts with
`Pro241, located in the domain linker. The structural
`roles for these two residues are different because
`Lys12 stabilizes the interface between the two
`domains and Tyr261 interacts with the domain
`linker.
`
`Examples of conserved structural roles in
`proteins other than BPI with similar folds yet
`dissimilar sequences
`
`The database of distant aligned protein struc-
`tures (DAPS) has over 1000 pairs of structurally
`similar proteins with less than 25 % sequence iden-
`tity. To determine if other proteins of similar struc-
`ture and dissimilar sequence also share residues
`with conserved structural roles, ®ve examples were
`chosen from DAPS and compared in the same
`manner as the domains of BPI. These examples are
`shown as the last ®ve entries in Table 1.
`The ®rst example aligns
`the a subunit of
`the bovine mitochondrial F1-ATPase with the
`Escherichia coli Rec A protein. These proteins share
`only 9 % sequence identity over the aligned resi-
`dues, yet 26 % of the positions are 3D-1D environ-
`mentally conserved. Like the domains of BPI, we
`®nd pairs of 3D-1D environmentally conserved
`positions with dissimilar residues which share con-
`served structural roles. For example, Tyr300 of
`F1-ATPase forms a 3.5 AÊ hydrogen bond to
`Arg304 (Figure 5(a)). These residues are located on
`an a-helix. Tyr300 and Arg304 are separated by
`four residues corresponding to approximately one
`turn of the helix. This serves to align the two resi-
`dues so that they can form a hydrogen bond stabi-
`lizing
`the
`helix. Likewise,
`the
`structurally
`equivalent residues in Rec A, Arg169 and Gln173,
`also form a 3.5 AÊ hydrogen bond. Tyr300 and
`Arg169 are paired in the structure-based sequence
`alignment and are not favored to substitute for one
`another according to the GONNET matrix (Benner
`et al., 1994). Arg304 and Gln173 are also paired in
`the alignment, demonstrating that the structural
`roles in both proteins are conserved.
`The second example aligns bromoperoxidase A2
`with triacylglycerol hydrolase. The sequence iden-
`tity is 14 % over the aligned positions while 25 % of
`the aligned positions are 3D-1D environmentally
`conserved. Gln42 in bromoperoxidase forms a
`2.7 AÊ hydrogen bond with the carbonyl group oxy-
`gen atom of Leu260 (Figure 5(b)). The structurally
`equivalent
`residue in triacylglycerol hydrolase,
`Trp52, forms a 3.1 AÊ hydrogen bond with the car-
`bonyl group oxygen atom of Leu228. Leu260 and
`Leu228 are paired in the alignment of these two
`proteins. The structural role of both Gln42 and
`Trp52 is a long-range tertiary interaction serving to
`connect two helices. It is noteworthy that this inter-
`action is between two residues
`separated by
`approximately 200 residues.
`A third example compares T cell antigen recep-
`tor with neuroglian. The sequence identity is 11 %
`
`over the aligned positions. Of the aligned pos-
`itions, 16 % have 3D-1D environmentally con-
`served positions. Trp162 of the antigen receptor is
`structurally aligned to Gln650 of neuroglian.
`Trp162 forms a 3.4 AÊ hydrogen bond to Gln213,
`which also forms a 2.7 AÊ hydrogen bond to
`Arg211. Gln650 forms a 2.8 AÊ hydrogen bond to
`Arg687, which is paired with Arg211 in the align-
`ment of neuroglian and T cell antigen receptor.
`This example demonstrates how hydrogen bonds
`through
`intermediate
`residues
`can
`promote
`conserved structural roles in proteins of similar
`structure.
`In our fourth example, we align b-amylase with
`concanavalin B. These proteins
`share
`10 %
`sequence identity. Of the aligned positions, 22 %
`are 3D-1D environmentally conserved. Leu78 of
`b-amylase is structurally aligned to Gln68 of conca-
`navalin B. Unlike the above examples, these resi-
`dues are buried in the core of each protein and
`belong to the H:B2 environment. Leu78 forms van
`der Waals contacts with other hydrophobic resi-
`dues in the core, while Gln68 forms both van der
`Waals contacts with other residues as well as two
`hydrogen bonds to water molecules near the sur-
`face of the protein. The hydrogen bonding poten-
`tial of
`the amide group hydrogen and oxygen
`atoms are thus satis®ed, even though Gln68 is bur-
`ied in the core. While this example does not satisfy
`the de®nition for
`conserved structural
`roles
`because each residue forms many non-speci®c con-
`tacts, it shows how dissimilar residues can stabilize
`the core of proteins.
`
`Clustering of 3D-1D environmentally
`conserved positions in both the BPI fold
`and the lipid-binding pockets
`
`Positions with conserved 3D-1D environments
`tend to cluster in the BPI fold. The location of the
`51 3D-1D environmentally conserved pairs are
`highlighted in Figure 6(a). These positions tend to
`cluster around the core of each domain and the
`phospholipid binding pockets, while the two tips
`of the molecule contain very few 3D-1D environ-
`mentally conserved residues. Clustered positions
`predominantly belong to classes with mostly bur-
`ied and apolar environments (H:B1 or E:B1) or
`mostly solvent accessible environments (H:P2, or
`E:P2).
`The degree of clustering was assessed by calcu-
`lating Ca-Ca distances for all 3D-1D environmen-
`tally conserved positions in each domain. For this
`analysis, a position is considered a tertiary neigh-
`bor of another position if its Ca atom is less than
`7 AÊ away but is not within two residues on the
`peptide chain. We ®nd 3D-1D environmentally
`conserved positions have at
`least one other
`conserved tertiary neighbor 41 and 44 % of the
`time for the N-terminal and C-terminal domains,
`respectively. In contrast, positions that were not
`3D-1D environmentally conserved has ``tertiary
`neighbors'' that were conserved only 24 and 28 %,
`
`Exhibit 2069
`Page 09 of 17
`
`

`

`How Dissimilar Sequences Adopt the Same Fold
`
`1027
`
`Figure 5. Four examples of pairs of 3D-1D environmentally conserved residues with dissimilar residues yet con-
`served structural roles. Each example is based on an alignment chosen at random from the DAPS database. (a) Rec A
`protein and F1-ATPase. (b) Bromoperoxidase A2 and triacylglycerol hydrolase. (c) Neruoglian and T-cell antigen
`receptor. (d) b-amylase and concanavalin B.
`
`con®rming that 3D-1D environmentally conserved
`positions tend to cluster.
`3D-1D environmentally conserved positions in
`the lipid-binding pockets of each domain also
`cluster. The program CAST (Liang et al., 1998) was
`
`used to identify all positions in contact with at
`least one lipid atom. Fifty-one positions contribute
`to the N-terminal lipid-binding pocket and 43 pos-
`itions contribute to the C-terminal pocket. A total
`of 11 of these positions from the N-terminal pocket
`
`Exhibit 2069
`Page 10 of 17
`
`

`

`1028
`
`How Dissimilar Sequences Adopt the Same Fold
`
`Figure 6 (legend opposite)
`
`are 3D-1D environmentally conserved; 15 positions
`from the C-terminal pocket are 3D-1D environmen-
`tally conserved. While the number of 3D-1D envir-
`onmentally conserved positions between the
`domains of BPI must be equal, the number of con-
`served positions in the lipid-binding pocke