Peptidoglycan recognition by Pal, an outer membrane lipoprotein.

Abstract

Peptidoglycan-associated lipoprotein (Pal) is a potential vaccine candidate from Haemophilus influenzae that is highly conserved in Gram-negative bacteria and anchored to the outer membrane through an N-terminal lipid attachment. Pal stabilizes the outer membrane by providing a noncovalent link to the peptidoglycan (PG) layer through a periplasmic domain. Using NMR spectroscopy, we determined the three-dimensional structure of a complex between the periplasmic domain of Pal and a biosynthetic peptidoglycan precursor (PG-P), UDP-N-acetylmuramyl-L-Ala-alpha-d-Glu-m-Dap-D-Ala-d-Ala (m-Dap is meso-diaminopimelate). Pal has a binding pocket lined with conserved surface residues that interacts exclusively with the peptide portion of the ligand. The m-Dap residue, which is mainly found in the cell walls of Gram-negative bacteria, is sequestered in this pocket and plays an important role by forming hydrogen bond and hydrophobic contacts to Pal. The structure provides insight into the mode of cell wall recognition for a broad class of Gram-negative membrane proteins, including OmpA and MotB, which have peptidoglycan-binding domains homologous to that of Pal.

Full text

Page 1

Peptidoglycan Recognition by Pal, an Outer Membrane Lipoprotein†,‡
Lisa M. Parsons, Florence Lin, and John Orban*
Center for AdVanced Research in Biotechnology, UniVersity of Maryland Biotechnology Institute, 9600 Gudelsky DriVe,
RockVille Maryland 20850
ReceiVed NoVember 1, 2005; ReVised Manuscript ReceiVed December 23, 2005
ABSTRACT: Peptidoglycan-associated lipoprotein (Pal) is a potential vaccine candidate from Haemophilus
influenzae that is highly conserved in Gram-negative bacteria and anchored to the outer membrane through
an N-terminal lipid attachment. Pal stabilizes the outer membrane by providing a noncovalent link to the
peptidoglycan (PG) layer through a periplasmic domain. Using NMR spectroscopy, we determined the
three-dimensional structure of a complex between the periplasmic domain of Pal and a biosynthetic
peptidoglycan precursor (PG-P), UDP-N-acetylmuramyl-L-Ala-R-D-Glu-m-Dap-D-Ala-D-Ala (m-Dap is
meso-diaminopimelate). Pal has a binding pocket lined with conserved surface residues that interacts
exclusively with the peptide portion of the ligand. The m-Dap residue, which is mainly found in the cell
walls of Gram-negative bacteria, is sequestered in this pocket and plays an important role by forming
hydrogen bond and hydrophobic contacts to Pal. The structure provides insight into the mode of cell wall
recognition for a broad class of Gram-negative membrane proteins, including OmpA and MotB, which
have peptidoglycan-binding domains homologous to that of Pal.
Gram-negative bacteria have a thin structural peptidogly-
can (PG)1layer located in the periplasm between a li-
popolysaccharide-containing outer membrane and an inner
cytoplasmic membrane. PG is a polymer composed of
N-acetylglucosamine (NAG) and N-acetylmuramic acid
(NAM) connected in an alternating fashion through ?-1,4-
glycosidic bonds. Short peptides are attached to the NAM
groups, and these can cross-link to form the cell wall matrix
(1). A number of proteins interact noncovalently with the
cell wall through a periplasmic PG-binding domain that is
widespread in Gram-negative bacteria (Figure 1a). Members
of this sequence family include the outer membrane proteins
Pal and OmpA, the inner membrane flagellar motor protein
MotB, and a number of other proteins with as yet unknown
function. Pal is linked to the outer membrane through an
N-terminal lipid group, while both OmpA and MotB have
additional integral membrane domains.
Pal is part of a larger network of proteins encoded by the
Pal/Tol operon, including TolA, TolB, TolQ, and TolR. TolA
is anchored to the periplasmic side of the inner membrane
by its N-terminal domain and forms a complex with the inner
membrane-spanning proteins TolQ and TolR (2). TolB is a
periplasmic protein that can bind both the C-terminal domain
of TolA (3) and Pal (4), thereby forming a possible structural
link between the inner and outer membranes. While this
system has been studied extensively, the exact function of
the Pal/Tol network remains unclear. Mutations in Pal lead
to vesicle formation and a loss of outer membrane integrity
(5, 6). Further, mutations in any of the Tol/Pal genes affect
the uptake of some organic compounds (7). Earlier work
showed that the Tol proteins are involved in protein import
as they have been parasitized to facilitate transport of
bacteriophage and colicins (toxins excreted by competing
bacteria) into the cell (2). More recently, it was demonstrated
that Pal and TolA are required for proper surface expression
of the cell surface O-antigen lipopolysaccharide in Escheri-
chia coli (8).
Pal is one of the components released by E. coli during
sepsis and may contribute to sepsis-induced inflammation
(9). In addition, Pal plays an important role in the virulence
of at least one species; it was shown that infecting human
subjects with bacteria containing a mutant Pal in a human
model of Haemophilus ducreyi infection affected the ability
of the bacteria to progress to the pustular stage of disease
(10). Humans can mount an immune response against Pal
from Haemophilus influenzae (also known as P6), and the
antibodies produced are bactericidal (11). Thus, a number
of efforts have been made and are currently underway to
utilize Pal to develop a vaccine against nontypeable H.
influenzae infections (12-14).
While crystal structures have been reported for a truncated
Pal from E. coli (PDB accession code 1oap) and for the
OmpA-like motif of RmpM from Neisseria meningitidis (15),
the mechanism by which these PG-binding domains recog-
nize the cell wall has remained unknown. We describe here
the solution NMR structure of the periplasmic domain of
Pal from H. influenzae bound to the peptidoglycan precursor
(PG-P), UDP-N-acetylmuramyl-L-Ala-R-D-Glu-m-Dap-D-
†Supported by NIH Grants GM57890 and 1S10RR15744 and the
W. M. Keck Foundation. L.M.P. is supported by a PRAT Fellowship
from NIGMS.
‡Structure coordinates for the Pal/PG-P complex are deposited in
the PDB (accession code 2aiz). NMR assignments for Pal, free PG-P,
and bound PG-P are deposited in BioMagResBank under the BMRB
accession codes 6465, 6856, and 6858, respectively.
* Corresponding author. Phone, 240-314-6221; fax, 240-314-6255;
e-mail, orban@umbi.umd.edu.
1Abbreviations: Pal, peptidoglycan-associated lipoprotein; PG,
peptidoglycan; PG-P, peptidoglycan precursor; NAM, N-acetylmuramic
acid; NAG, N-acetylglucosamine; OmpA, outer membrane protein A;
m-Dap, meso-diaminopimelate; H-bond, hydrogen bond.

Page 2

Ala-D-Ala (m-Dap is meso-diaminopimelate), a key inter-
mediate of PG biosynthesis. This represents the first three-
dimensional structure of such proteins complexed with a PG
unit. The structure shows that Pal interacts exclusively with
the peptide portion of the ligand. In particular, extensive
H-bond and hydrophobic contacts to the m-Dap group, which
is found in the cell walls of all Gram-negative bacteria,
indicate that Pal specifically recognizes un-cross-linked Dap
groups in the PG layer.
MATERIALS AND METHODS
Sample Preparation. The cloning and purification of the
H. influenzae Pal/PG-P complex was carried out as follows.
The Pal gene minus a 19 residue N-terminal signaling
sequence was amplified from genomic DNA using PCR and
placed into a pET-28a (Novagen) vector with a Tev-cleavable
His6-tagged GST fusion at the N-terminus. Protein was
expressed in E. coli BL21 (DE3) cells grown in LB or labeled
minimal media at 37 °C. Cells were harvested by centrifuga-
tion at 7000g for 10 min, resuspended in binding buffer (50
mM sodium phosphate, 300 mM sodium chloride, and 10
mM imidazole, pH 8.0), and lysed by sonication. The
resulting lysate was centrifuged at 25 000g for 25 min. Pal/
PG-P was purified by elution from a Ni-NTA column
(Qiagen) using standard procedures. The His6-tag was
removed by incubating the purified Pal/PG-P sample (in 20
mM Tris, 100 mM NaCl, 1 mM DTT, and 0.5 mM EDTA,
pH 8.0) with His6-tagged Tev protease overnight at 30 °C.
The GST tag and protease were then removed by passing
the solution over a Ni-NTA column. Samples for NMR
analysis were dialyzed into NMR buffer (50 mM sodium
phosphate and 50 mM NaCl, pH 6.8) and concentrated to
approximately 0.8 mM.
Samples of purified PG-P and peptidoglycan-free Pal were
also used to assist in the NMR assignment process, and their
preparation is described here. Supernatant containing13C/
15N-labeled Pal with bound PG-P was loaded onto a Ni-
NTA column following the manufacturer’s guidelines for
sample loading and washed with buffer (binding buffer with
20 mM imidazole). The column was washed with distilled
water (6 bed volumes) to remove any salt, then placed in a
47 °C water bath, and equilibrated for 7-10 min. While still
in the water bath, the column was washed with distilled water
at 47 °C (6 bed volumes) to elute pure PG-P (ca. 0.5 mg/L
of culture). Further washing of the column at room temper-
ature with wash buffer (6 bed volumes) followed by elute
buffer (binding buffer with 250 mM imidazole) gave
peptidoglycan-free Pal. The His6-tag was removed as de-
scribed above. The identity of PG-P was confirmed by
MALDI-TOF (Perseptive Biosystems) using a 2,5-dihy-
droxybenzolic acid (Aldrich) matrix. The mass spectrum was
acquired in negative ion linear mode with an acceleration
voltage of 20 kV and a delay time of 200 ns. Pure13C/15N-
labeled PG-P was lyophilized and redissolved in NMR buffer
either on its own or with unlabeled, peptidoglycan-free Pal.
Circular Dichroism. Circular dichroism measurements
were performed using a Jasco spectropolarimeter, model
J-720 and a water-jacketed quartz cell with a path length of
0.01 cm. The absorbance at 222 nm of 3 mg/mL of protein
in NMR buffer was recorded as the temperature increased
from 25 to 75 °C. Reversibility was confirmed by comparing
full spectra between 200 and 250 nm taken at 25 °C before
and after the melt.
NMR Spectroscopy. Spectra were collected at 298 K on a
Bruker DRX-600 equipped with 3-axis gradient probes and
recorded in States-TPPI mode. Pulsed-field gradients were
used for coherence selection and solvent suppression. Data
FIGURE 1: (a) Multiple alignment of Pal-like sequences. Top five
sequences: protein sequence alignment of five proteins from E.
coli containing the PG-binding domain motif. Residues in red are
invariant, while residues in yellow are similar. Percent identity of
each to Pal is OmpA 32%, YiaD 25%, YfiB 23%, and MotB 27%.
Sixth sequence: sequence of H. influenzae Pal colored as above
showing the conservation of Pal proteins aligned from 50 species.
Gaps were added or moved to align this sequence with the E. coli
protein sequences. The numbering and secondary structure is
according to the mature H. influenzae protein. Alignment was done
using ClustalW (37). (b) Solution structure of the Pal/PG-P NMR
complex. Pal is shown in ribbon form (gray) with residues invariant
in all six sequences labeled and highlighted (red). PG-P is shown
in green. (c) Backbone superposition of the 20 lowest energy
structures for the Pal/PG-P complex. ?-Strands are highlighted in
blue, R-helices are shown in red, loop or coil regions are gray, and
residues 4-7 of PG-P are in green. The ensemble conformations
for the m-Dap5 side chain are also shown in green. The orientation
is the same as for panel b.

Page 3

were processed using nmrPipe (16) and analyzed with Sparky
(T. D. Goddard and D. G. Kneller, UCSF). Backbone and
side-chain assignments for Pal and bound PG-P were made
using standard triple resonance experiments with a sample
where both molecules were13C/15N-labeled (17). To assign
free PG-P, a15N-labeled sample in H2O was used to collect
the two-dimensional (2D)15N-HSQC (18) and 2D HN-C′
plane of the three-dimensional (3D) HNCO (19). A13C/15N-
labeled PG-P sample in D2O was used to collect the 2D13C-
HSQC and 2D TOCSY (20). Distance restraints in the Pal/
PG-P complex were obtained from a 3D15N-edited NOESY
(21) spectrum, 3D13C-edited NOESY (22) spectra in H2O
and D2O, and a 2D NOESY (23) in D2O. To assist in the
assignment of intermolecular NOEs between ligand and
protein, a 3D13C-edited NOESY was also acquired on a
mixed sample (13C/15N-labeled PG-P with unlabeled Pal, 1:1).
Mixing times for NOESY spectra were 120 ms and 150 ms.
3JHNHR values were acquired from a 3D HNHA (24)
experiment. Hydrogen-deuterium exchange data on the
complex were obtained from a15N HSQC spectrum recorded
after 17 min at 25 °C in D2O. Residual dipolar coupling
(RDC) data were collected for N-HNand C′-N using the
IPAP-HSQC (25) and quantitative HNCO (26) experiments
on a sample with 18.7 mg/mL of Pf1 phage added (27). T1
and T2measurements were made on the Pal/PG-P complex
using the method of Farrow et al. (28).
Structure Calculations. Structures were calculated with
CNS1.1 (29) using standard simulated annealing and torsion
angle dynamics protocols. Prochiral groups were given
floating stereospecific assignments until they could be
unambiguously assigned from the structure. Initial NOE
restraints were automatically assigned by NOEID (http://
orban.umbi.umd.edu/NOEID) from a model based on the
crystal structure of the truncated E. coli Pal protein (1oap).
Distance restraints were based on peak intensities categorized
as strong (1.8-3.0 Å), medium (1.8-4.0 Å), weak (2.7-
5.3 Å), medium-weak (1.8-4.2 Å), medium-strong (1.8-
3.5 Å), or very weak (3.0-6.0 Å). Dihedral restraints were
obtained from TALOS (30) and
restraints were 1.7-2.3 Å for rHN-Oand 2.5-3.1 Å for rN-O
and were assigned based on deuterium exchange data and
calculated structures. Final force constants used were 1000
kcal mol-1Å2for bond lengths, 500 kcal mol-1rad-2for
angles and improper torsions, 40 kcal mol-1Å-2for
experimental distance restraints, 200 kcal mol-1rad-2for
dihedral restraints, and 4.0 kcal mol-1Å-4for the van der
Waals repulsion term. A force constant of 0.3 kcal mol-1
Å-2was used for the protein RDC restraints, while 0.5 kcal
mol-1Å-2was used for the PG-P RDC restraints. The 20
best structures were chosen based on a low total energy, no
NOE violations greater than 0.5 Å, no dihedral violations
greater than 5°, and other diagnostic indicators of structure
quality as summarized in Table 1. Structures were analyzed
using PROCHECK (31) and displayed using PyMol (32).
3JHNHR values. H-bond
RESULTS
Ligand Purification and Identification. While making the
backbone NMR assignments of Pal (17), signals due to five
sequentially connected residues not belonging to the protein
were found in the 3D HNCACB and CBCACONH experi-
ments. A literature search identified this sequence as being
part of UDP-N-acetylmuramyl-pentapeptide, a biosynthetic
peptidoglycan precursor (PG-P). The PG-P was purified from
Pal for further characterization using a modified protocol
that also enabled purification of peptidoglycan-free Pal
without significant degradation of the protein. Previous
methods to purify Pal from the mature PG layer required
that the protein preparation be heated to 65 °C for 30 min
then centrifuged (33). Here, Pal was heated at 47 °C while
bound to an affinity column and washed with heated buffer.
Pal melts reversibly at 47 °C, but about 15% of the protein
does not refold correctly. Higher temperatures and longer
exposure to heat noticeably decreased the amount of protein
recovered from the column.
The assignments of UDP-NAM-L-Ala-γ-D-Glu-L-Lys-D-
Ala-D-Ala from Anabaena cylindrica, a Gram-positive
bacterium, have been reported previously (34), and these
were used to identify the UDP-NAM part of the PG-P from
the13C-HSQC spectrum. The pentapeptide portion of PG
can differ between Gram-negative and Gram-positive bacteria
in two ways. The second amino acid in the pentapeptide can
be either γ-D-Glu or R-D-Glu. Also, the third amino acid is
m-Dap in all Gram-negative bacteria and Gram-positive
bacilli and L-Lys in most other Gram-positive bacteria (35).
Since the backbone could be assigned sequentially by through
bond connectivities to R-carbons, the second residue is
unambiguously an R-D-Glu. Further, the molecular weight
determined by mass spectrometry matches the calculated
weight for PG-P with m-Dap instead of Lys (measured m/z
1200, calculated m/z 1202, for a15N-sample), consistent with
expression in Gram-negative E. coli. NMR assignments of
the third amino acid also indicated a Dap rather than Lys
chemical shift pattern with the H?resonance at 4.23 ppm.
On the basis of the combination of these data, the primary
structure of the PG-P co-purified with Pal was identified as
UDP-NAM-L-Ala-R-D-Glu-m-Dap-D-Ala-D-Ala (Figure 2a).
Solution Structure of Pal. Pal forms a monomeric R/?
sandwich with the secondary elements arranged in the order
R-?-R-?-R-?-R-? (Figure 1b,c). The first three
strands, ordered ?1, ?4, and ?2, are antiparallel with ?3 lying
parallel to ?2. Helices R1 and R4 are each seven-residues
FIGURE 2: (a) Structure of PG-P. (b) Evidence supporting PG-P
contact sites with Pal. Gray columns represent the ratio of15N-
relaxation rates (R2/R1) in PG-P. White columns show the sum of
the differences of N, HN, CR, HR, C?, H?, and C′ chemical shifts
between bound and unbound PG-P. To make them comparable to
proton shifts,15N and13C shifts were adjusted by multiplying them
by 0.154 and 0.276, respectively. Black columns represent the
intermolecular NOE restraints per residue. For comparative pur-
poses, values are shown as a percent of the total for each category
(e.g., m-Dap has 18/28 ) 64% of the total intermolecular restraints).
The line plot shows the average backbone (C′, N, and CR) rmsd
values per residue.

Page 4

long and located on the same side of the ?-sheet at opposite
ends of ?1. The longer R2 and R3 helices angle across the
?-sheet, R2 from the C-terminus of ?1 to the N-terminus of
?2 and R3 from the C-terminus of ?3 to the N-terminus of
?2. The first 15 residues are flexible and unstructured as
evidenced by the low15N R2 values and a lack of NOE
restraints (Figure 3a,b). The ?1-R2, ?2-R3, and ?3-R4
loops appear to be slightly more mobile than the regions of
secondary structure based on decreased15N R2values in these
regions (Figure 3a). Consistent with these data is the
observation that some residues in these loops do not give
strong NOE signals. Residues in the loop regions had fewer
NOE restraints due to their surface exposure or difficulty in
making assignments due to overlapping chemical shifts
(Figure 3b). Because of this, loop residues, particularly in
?2-R3 and ?3-R4, have a higher rmsd than residues in
R-helices or ?-strands (Figure 3c). The two smaller helices
(R1 and R4) have higher rmsd values than the two larger
helices (R2 and R3) due to fewer NOE restraints and their
surface exposure. In addition, R1 is in close proximity to
the base of a large flexible tail that is probably contributing
to its higher rmsd values. On average, 17.3 NOE restraints/
residue were used for residues 16-134 (Figure 3b). The
structure statistics are summarized in Table 1.
The solution NMR structure of the 134-residue periplasmic
domain of H. influenzae Pal is similar to the crystal structure
of a truncated 108-residue homologue from E. coli reported
in the PDB (1oap). These two structures have a backbone
rmsd of 1.2 Å and 68% sequence identity over residues 29-
134 (residues 48-152 in E. coli). The main difference is
that a larger segment of the N-terminal region after the
natural signal sequence cleavage site is deleted in the E. coli
structure, so the R1 helix is not present. This does not affect
the overall fold, however. The solution structure also has
similarities to the crystal structure of the OmpA-like domain
of RmpM from N. meningitidis (1r1m, (15)) with an rmsd
of 2.3 Å over residues 30-134 (residues 76-203 in N.
meningitidis) and a sequence identity of 30%. The structure
of RmpM also lacks the N-terminal R1 helix.
Interaction between Pal and PG-P. The binding interface
between Pal and PG-P was initially characterized by compar-
ing the chemical shifts of free and bound PG-P. No shift
changes were seen for the signals of UDP and muramic acid,
while the resonances of the pentapeptide exhibited increas-
ingly large differences centered on the m-Dap residue (Figure
2b).15N R2 values were higher for m-Dap5, D-Ala6, and
D-Ala7 than for other residues in the PG-P, suggesting that
these were the most ordered residues in bound PG-P. The
combination of these results indicated that residues 4-7 of
the pentapeptide make the most contact with Pal. Likewise,
Pal residues in the ?1-R2, ?2-R3, and ?3-R4 loops
experienced chemical shift and peak intensity changes when
PG-P was removed (data not shown), indicating they were
involved in the interaction with PG-P.
The detailed structure of the binding interface was
determined by identification of 28 intermolecular NOEs
between Pal and PG-P. The majority of these NOE contacts
were from m-Dap5 to Pal, consistent with the chemical shift
perturbation and relaxation data (Figure 2b). The m-Dap5
residue has extensive NOEs to Y78, F36, and L82 (Figure
4a). Additionally, D-Glu4 has NOEs to Y78 and D-Ala7
contacts F36. There were a total of 18 NOE restraints from
m-Dap5 to Pal, 6 from D-Glu4, and 4 from D-Ala7. Upon
FIGURE 3: (a) Ratio of15N-relaxation rates (R2/R1) for backbone
amides in Pal. Error bars represent ( 1 SD. (b) Number of NOE
restraints per residue for Pal. Restraint types are shown from the
bottom of each column as follows: intraresidue (gray), sequential
(white), medium range (light gray), and long range (black). (c)
Average rmsd per residue for the backbone N, CR, and C′ atoms.
The secondary structure is shown at the bottom of the plot.
Table 1: Statistics for the Ensemble of 20 Structures
Pal PG-P
(A) Experimental Restraints
total NOE
intraresidue
sequential (|i - j| ) 1)
medium-range (|i - j| e 4)
long-range (|i - j| > 4)
intermolecular
hydrogen bond
total dihedral angle restraints
?
ψ
total RDCs
2056
793
489
318
493
67
25
10
4
0
28
8
5
3
2
12
160
200
101
99
212
(B) Structure Statistics
average rmsd (Å) to meana
all non-hydrogen atoms
backbone atoms (N, CR, C′)
Ramachandran distribution (residues 16-134)b
most favored (%)
additionally allowed (%)
generously allowed (%)
overall G-factorb,c
number of bad contacts for residues 16-134b
residues 16-134 residues 4-7
0.99 ( 0.05
0.47 ( 0.03
1.92 ( 0.15
1.60 ( 0.10
82.7 ( 2.7
17.5 ( 2.9
0.5 ( 0.6
0.1 ( 0.2
6.9 ( 1.7
aStatistics quoted are for the mean ( 1 SD.bObtained using
PROCHECK.cDefined as the log-odds score based on each residue’s
?-ψ, ?1-?2, and ?1 values. Low (negative) values suggest unusual
geometry.

Page 5

application of the intermolecular NOE restraints in structure
calculations, the side-chain oxygen atoms of m-Dap5 were
consistently within H-bonding range of the main-chain amide
group of the conserved D37 residue. The amide proton of
D37 is in slow-exchange in D2O, and as there are no other
suitable acceptors within H-bonding distance in the structure,
a hydrogen bond restraint was included between the D37
amide and the side-chain carboxyl group of m-Dap5 in the
final stages of refinement. Similarly, the ?-ammonium group
of m-Dap5 was consistently found to be within H-bonding
distance of both the strictly conserved D71 side-chain
carboxyl group and the backbone oxygen of conserved R73
(Figure 4a). This was supported by NOEs from the assigned
m-Dap5 ?-ammonium group to the backbone amide protons
of D71 and R73. Therefore, H-bond restraints were also
included for these contacts in the later stages of structure
refinement. It is possible that other H-bonds are formed,
perhaps involving the side chain of R73. However, the NOE
data were not sufficient to assign any further H-bond
interactions between Pal and PG-P unambiguously.
The overall structure of the Pal/PG-P complex indicates
that the PG-P binding site is located at a surface cavity in
the Pal structure formed by the loop between ?1 and R2,
the loop between ?2 and R3, and R4 (Figure 4b). The
m-Dap5 residue of PG-P occupies this pocket in the complex
and forms stabilizing interactions through H-bonding to its
side chain ?-ammonium and ?-carboxylate moieties as well
as through hydrophobic contacts to its side-chain methylene
groups (Figure 4a). Residues D-Glu4 and D-Ala7 from PG-P
also contribute to the binding interface with Pal through
mostly hydrophobic interactions, but the majority of contacts
come from m-Dap5.
DISCUSSION
The Pal family of proteins is widespread in Gram-negative
bacteria with sequence homologues in more than 100
organisms. This sequence family includes other proteins such
as the outer membrane protein, OmpA, and the inner
membrane flagellar motor protein, MotB. Both contain two
domains, one of which is a periplasmic PG-binding domain.
Multiple alignment analysis indicates that six residues, F34,
G68, D71, G74, L82, and R86, are invariant in the wider
family of Pal-related proteins (Figure 1a). Three of these
residues, D71, G74, and L82, are at or near the PG-P binding
site (Figure 1b). The carboxylate group of D71 is involved
in an H-bond interaction with the m-Dap5 side chain
ammonium group, while L82 is one of several residues that
have stabilizing hydrophobic contacts with PG-P (Figure 4a).
As G74 does not interact directly with PG-P, its conservation
may be due to steric reasons whereby substitution with any
other amino acid would perhaps decrease the size of the
binding pocket adversely. Residue R86 is buried in the
hydrophobic core, extending from the middle of the R3 helix
toward the PG-P binding surface, while neighboring F34 and
G68 are also buried and likely to be important for maintain-
ing the orientation of R86 (Figure 1b). Although the
guanidinium group of R86 is close to the m-Dap5 carboxy-
late, it is not within H-bonding distance in any of the 20
lowest energy structures. It is noteworthy that the crystal
structure of E. coli Pal reported in the PDB (1oap), which
was determined in the absence of PG, has both the main-
chain amide of D37 (numbered D55 in 1oap) and the side-
chain guanidinium of R86 (numbered R104 in 1oap)
H-bonded to a sulfate ion. The location of this sulfate ion is
very similar to the position of the m-Dap5 ?-carboxylate
group in the Pal/PG-P complex, which also H-bonds to the
D37 backbone amide. In addition to these invariant amino
acids, position 78 is always found to be either tyrosine or a
hydrophobic residue.
A number of surface residues near the PG-P binding site
are conserved in Pal sequences (Figure 1a, bottom sequence)
but not in the broader Pal-related family (Figure 1a, top five
sequences). These include D37 in the ?1-R2 loop; R73 and
T75 in the ?2-R3 loop; Y78 and R85 from R3; E112, K113,
and P114 in the ?3-R4 loop; Y124 from R4; and R128 from
the R4-?4 loop (colored blue in Figure 4b). At the PG-P
binding site, only F36 is not conserved, suggesting that this
residue is not critical for PG binding despite the observation
of NOE contacts to the PG-P. The extent of surface
conservation around the PG-P site present only in Pal
sequences suggests that this particular family of proteins may
have a further binding function in addition to PG recognition.
This is consistent with results suggesting that another member
of the Pal/Tol operon, TolB, interacts with Pal in a region
that overlaps with the PG recognition site (4). The precise
function of this Pal-TolB complex is not yet known.
FIGURE 4: (a) Detailed view of the Pal/PG-P binding interface.
Carbon atoms of Pal residues with NOEs or H-bonds to PG-P are
shown as orange sticks with black labels. Carbon atoms of PG-P
residues are in green with blue labels. The backbone of Pal is shown
as a gray ribbon. H-bonds from m-Dap5 to Pal are represented by
dashed lines. Oxygen atoms are red, nitrogen atoms are blue, and
hydrogen atoms on the ?-ammonium group of m-Dap5 are white.
(b) Surface representation of Pal with bound PG-P (green),
demonstrating the conservation of residues around the binding site.
The view is directly into the binding pocket (from the left side of
Figure 1b). Residues colored in red are invariant in both Pal proteins
and E. coli proteins containing the PG-binding motif. Residues in
blue are conserved only in Pal proteins. The PG-P is oriented with
the UDP/NAM groups at the top and D-Ala7 at the bottom of the
figure.