Predicted molecular structure of the mammalian cell entry proteinMce1A of Mycobacterium tuberculosis

Amit Kumar Das,a Devrani Mitra,a Morten Harboe,b Bidisha Nandi,a Robin E. Harkness,c

Debabrata Das,a and Harald G. Wikerd,e,*

a Department of Biotechnology, Indian Institute of Technology, Kharagpur 721 302, West Bengal, Indiab Institute of Immunology, Rikshospitalet University Hospital, NO-0027 Oslo, Norway

c Aventis Pasteur, Connaught Campus, Toronto, Ont., Canada M2 3T4d National Veterinary Institute, Oslo, Norway

e Department of Environmental Immunology, Norwegian Institute of Public Health, NO-0403 Oslo, Norway

Abstract

The proposed role of the mammalian cell entry protein 1A (Mce1A) ofMycobacterium tuberculosis is to facilitate invasion of host

cells. The structure of Mce1A was modelled on the basis of the crystal structure of Colicin N of Escherichia coli by fold prediction

and threading. Mce1A, as the model predicts, is an a=b protein consisting of two major (a and b) domains, connected by a long ahelix. The model further revealed that the protein contains 12 helices, 9 strands, and 1 turn. The final model of Mce1A was verified

through the program VERIFY 3D and more than 90% of the residues were in the favourable region. A mouse monoclonal antibody,

TB1–5 76C, is directed to an epitope within a 60-mer peptide that has been shown to promote uptake of bacteria in mammalian cells.

We show here that the epitope could be narrowed down to a core of 4 amino acids, TPKD. Upstream flanking residues, KRR also

contributed to binding. Mce2A does not promote uptake in mammalian cells and sequence comparison of Mce1A and Mce2A

indicates that the epitope mediates uptake. The epitope was located at the surface of the Mce1A model at the distal b strand-loopregion in the b domain. The localization of this epitope in the model confirms its potential role in promoting uptake of M. tu-

berculosis in host cells.

� 2003 Elsevier Science (USA). All rights reserved.

Keywords: Mammalian cell entry protein; Molecular model; Mycobacterium tuberculosis; Monoclonal antibody; Epitope

Arruda et al. [1] described a DNA fragment of My-

cobacterium tuberculosis which conferred to Escherichia

coli the ability to enter into mammalian cells and to

survive inside macrophages. The mammalian cell entry

(mce) gene was subsequently localized in the M. tuber-

culosis H37Rv genome and termed mce1 [2]. Analysis of

the whole genome revealed a striking structure: Four

copies of mce were found, situated in operons eachconsisting of eight genes, and these operons were orga-

nized in a similar manner with extensive homology. In

each case the genes preceding mce coded for two integral

membrane proteins, YrbE [3], while mce and the fol-

lowing five genes were predicted to encode proteins with

hydrophobic stretches at the N-terminus probably rep-

resenting signal sequences. These features are consistent

with the location of these proteins at the mycobacterial

cell surface and the proposed role of mce in facilitating

invasion of host cells. InM. tuberculosis the genes of the

mce1 operon are now designated yrbE1A, yrbE1B,

mce1A, mce1B, mce1C, mce1D, mce1E, and mce1F, re-

spectively [3].

It is widely recognized that detailed knowledge of the

three-dimensional structure of individual proteins is es-sential to understand their function. During recent years

with a rapidly increasing number of whole genome se-

quences, structural biologists have developed new

strategies to predict the molecular structure of new

proteins and protein domains based on the amino acid

sequence [4]. The protein data bank (pdb) is widely

recognized as a source of information about structurally

solved proteins, both experimentally and theoretically.There are many different algorithms and methods for

characterization of structural features of proteins that

Biochemical and Biophysical Research Communications 302 (2003) 442–447

www.elsevier.com/locate/ybbrc

BBRC

* Corresponding author. Fax: +47-22-04-26-86.

E-mail address: harald.g.wiker@fhi.no (H.G. Wiker).

0006-291X/03/$ - see front matter � 2003 Elsevier Science (USA). All rights reserved.

doi:10.1016/S0006-291X(03)00116-5

have not been determined by crystal structure or nuclearmagnetic resonance spectroscopy. To develop a theo-

retical model of a protein, a template of a structurally

known protein is needed. The template should ideally

have more than 20% sequence identity to perform direct

homology modelling. When sequence identity is less

than 20% it is preferable to model by fold prediction and

threading. The latter method was chosen for modelling

of Mce1A. The mce1A gene encodes 454 amino acidresidues and the N-terminal signal peptide is predicted

to contain 39 or 40 amino acids, respectively, according

to the Neural Network (NN) and Hidden Markov

Model (HMM) methods provided by the SignalP server.

Flesselles et al. [5] raised a mouse monoclonal anti-

body, TB1–5 76C, against a synthetic 60-mer peptide,

corresponding to residues 106–165 in the N-terminal

part of Mce1A. Recently, this antibody was shown toreact strongly with the recombinant fusion protein

GST–Mce1A, but not with free GST [6]. The epitope

reacting with the monoclonal TB1–5 76C antibody

could be narrowed to the KRRITPKD region, residues

131–138 in Mce1A.

This sequence is contained in a region of Mce1A in-

volved in its facilitation of cell entry and absent in the

corresponding region of Mce2A which does not pro-mote cell entry [7], indicating that promotion of cellular

uptake also might be narrowed down to this sequence.

To provide further evidence on this point, the pur-

pose of this work was to apply available methods for

prediction of the molecular structure of Mce1A (1) to

see whether the structure supports its proposed function

to promote cellular entry and survival, and (2) to lo-

calize and further characterize the TB1–5 76C reactiveepitope in relation to the predicted molecular structure

of Mce1A.

Materials and methods

Sequence retrieval. The Mce1A amino acid sequence was retrieved

from the Sanger Institute using the server http://www.sanger.ac.uk.

Secondary structure prediction. The secondary structure prediction

was done based on PSIPRED (http://bioinf.cs.ucl.ac/psipred/) and

PHD (http://npsa-pbil.ibcp.fr/cgi-bin/npsa_automat.pl?page¼ /NPSA/npsa-phd.html) fold servers.

Primary sequence alignment. Primary sequence alignment was done

using ClustalW multiple sequence alignment (http://searchlaun-

cher.bcm.tmc.edu/multi-align/Options/clustalw.html). The output was

further given to the BOXSHADE server (version 3.21, Hofmann and

Baron) (http://www.ch.embnet.org/software/BOX_form.html).

Structural template identification. Using the BLAST server (http://

www.ncbi.nlm.nih.gov/BLAST), template structures homologous to

Mce1A were sought, but no appreciable homology was found in the

Brookhaven pdb database. The appropriate template structure was

therefore chosen by comparing the results of fold prediction servers

3DPSSM (http://www.sbg.bio.ic.ac.uk/~3dpssm/), GENTHREADER

(http://bioinf.cs.ucl.ac.uk/psiform.html), FFAS (http://www.bioinfor-

matics.burnham-inst.org/FFAS/index.html), UCLA-DOE (http://

fold.doe-mbi.ucla.edu/psiform/), BIOINBGU (http://www.cs.bgu.ac.il/

~bioinbgu/), and FUGUE (http://www-cryst.bioc.cam.ac.uk/~fugue/).

The best template, which came out from these investigations, was

Colicin N (pdb code 1A87) of E. coli.

Molecular modelling. For modelling of Mce1A (Rv 0169), we used

MODELLER [8]. The modelling procedure begins with an alignment

of the sequence to be modelled (target) with related known three-di-

mensional structures (templates). This alignment is the usual input to

the program. The output is a 3D model for the target sequence con-

taining all main chain and side chain non-hydrogen atoms.

The degree of primary sequence identity between target-template

and the fold prediction results indicated that the crystallographic

structure of Colicin N of E. coli would be a good model to be used as

template for Mce1A. Using this template, the residues from 68 to 376

of Mce1A were modelled. The N-terminal 67 residues and the C-ter-

minal 78 residues were not modelled due to lack of homology. A total

of 450 models were generated for Mce1A and the final model was

selected based on stereochemical quality. The best model in accord

with best stereochemical property was subjected to energy minimiza-

tion using the Insight II (Version 2000) software package. Optimisa-

tion related to energy minimization has been computed using the

steepest descent minimization algorithm in the discover module of

Insight II.

Analysis of the model. The overall stereochemical quality of the

final model for Mce1A was assessed by the program PROCHECK [9].

The resolution was set at 1.5�AA. The final model of Mce1A was verified

through the program VERIFY 3D (http://www.doe-mbi.ucla.edu/

Services/Verify_3D/). The model has been deposited in public data-

bases. The Research Collaboration for Structural Bioinformatics, ID

code is RCSB017713, and the PDB ID code is 1NA9.

Peptide synthesis. Synthetic peptides were obtained from Euro-

gentec, Seraing, Belgium through MedProbe A/S, Oslo, Norway as

peptide amids, blocking the C-terminal carboxyl of the peptide to more

closely mimic the charge environment of the native protein.

ELISA technology. To test for antibody reactivity with synthetic

peptides, Costar polystyrene 96-well plates (Corning, Corning, NY)

were coated with individual peptides by applying 1 lg peptide per wellin 100ll phosphate-buffered saline (PBS), pH 7.4, blocking with 5mg/

ml bovine serum albumin (BSA) in PBS, and washing as described

previously [6]. Detection was with HRP conjugated sheep anti-mouse

immunoglobulin (Amersham LIFE SCIENCE, Amersham, UK). 2,20-

Azino-di-[3-ethylbenzthiazoline-sulfonate] (ABTS) (Sigma Chemical,

St. Louis, MO) was used as the substrate.

Results

Primary sequence characterization and comparison

The secondary structural predictions obtained from

PHD and PSIPRED are given in Fig. 1. The sequence

was then fed to all the fold prediction programs listed in

Materials and methods, and this showed that Colicin N

was the best template. Colicin N is the smallest poreforming colicin known to date, and the crystal structure

of its membrane receptor, the porin OmpF, is already

known. Pore forming colicins are water soluble bacte-

riocins capable of binding to and translocating through

the E. coli cell envelope. After undergoing a transition to

a transmembrane ion channel in the cytoplasmic mem-

brane, they lead to cell death [10].

The sequence structure alignment of Colicin N(template) and Mce1A (target) is shown in Fig. 2 as

done by 3DPSSM sequence structure alignment. The

A. Kumar Das et al. / Biochemical and Biophysical Research Communications 302 (2003) 442–447 443

sequence identity was 19% (59/309). By pairwise align-

ment using ClustalW, there were 16% identical residues,

12% strongly similar residues, and 13% weakly similar

residues.

Quality of the model

Mce1A, as the model (Fig. 3) predicts, is an a=bprotein consisting of two major (a and b) domains,

Fig. 1. Secondary structure prediction of Mce1A using two algorithms, PHD and PSIPRED. C or c is prediction of random coil, E or e is prediction

of extended strand, and H or h is prediction of alpha helix. The upper line represents the results from PHD, the middle line represents the results from

PSIPRED (PSI), and the lower line gives the complete Mce1A sequence (SEQ). Numbering is given as for the whole Mce1A protein. The model of

Mce1A covers residues 68–376.

Fig. 2. Comparison between Colicin N of E. coli and Mce1A of M. tuberculosis by sequence structure alignment based on the 3DPSSM fold pre-

diction server result. The sequence in the upper line corresponds to Colicin N and the lower line corresponds to Mce1A. The numbering starts with

the first amino acid of the model which is amino acid 68 of Mce1A and ends with amino acid 309 of the model, corresponding to amino acid 376 of

Mce1A. The alignment gave 17% identity. Identical residues are marked with an asterisk (*).

444 A. Kumar Das et al. / Biochemical and Biophysical Research Communications 302 (2003) 442–447

connected by a long a helix. The Mce1A model revealed

that the protein contains 12 helices, 9 strands, and 1

turn.

Figs. 4A and B show the Ramachandran diagram for

the resulted Mce1A model structure and for the crys-

tallographic Colicin N structure solved to 1.5 and 3.10�AAresolutions, respectively. Analysis of the Ramachandranplot of the Mce1A model shows that 85.3% were in the

core region, 11.7% in the allowed region, 1.5% in the

general region, and 1.5% (Val13ð80Þ, His33ð100Þ,

Asp74ð141Þ, and Lys243ð310Þ) were found in the disallowed

region. They were all located in loops. The same analysis

for the crystallographic structure of Colicin N presented

93.2% in the core region and 6.8% in the allowed region.

The final model of Mce1A was verified through theprogram VERIFY 3D, and more than 90% of the resi-

dues were in the favourable region. Considering the

length of the modelled protein, the quality of the model

is therefore good.

Leucine zipper motifs

After using the ScanProsite program, which allows

one to scan a protein sequence for the occurrence ofpatterns and profiles stored in the Prosite database, it

was revealed that there may be a high probability of the

occurrence of a Leucine zipper motif from the 274th to

the 295th residue in the original Mce1A sequence (i.e.,

206th–227th in the model). The model clearly depicts,

however, that in that region there is a loop as well as a b-turn-b motif. Thus, the probable presence of a leucine

zipper motif is not in accordance with the model struc-ture and is thought to be due to the preponderance of

leucine residues in Mce1A.

Epitope characterization

Previous bioinformatic analysis of the epitope react-

ing with TB1–5 76C indicated that the KRRITPKDregion, corresponding to residues 64–71ð131–138Þ in

Mce1A, was involved [6]. To localize the borders of the

epitope, a series of truncated peptides were made de-

leting one amino acid stepwise from each end of the

NTPKRRITPKDVI peptide. Reactivity in ELISA

Fig. 3. The Mce1A model is shown by stereo view of ribbon diagram

covering residues 68–376 of the complete Mce1A sequence.

Fig. 4. The quality of the Mce1A model is demonstrated by (A) Ra-

machandran plot (/ vs w) and compared with (B) the crystallographicstructure of Colicin N (1A87).

A. Kumar Das et al. / Biochemical and Biophysical Research Communications 302 (2003) 442–447 445

required the presence of K64ð131Þ and D71ð138Þ. Presenceof the NPT residues upstream and VI downstream in-

creased the OD values moderately but was clearly not

required for binding. The reaction pattern of the

KRRITPKD peptide, residues 64–71ð131–138Þ was subse-

quently characterized in amino acid replacement studies

summarized in Table 1. The results indicated that the

core epitope involves the four amino acids TPKD and

that the flanking residues upstream, KRR, also influencebinding. Alanine exchange stepwise along the

NTPKRRITPKD peptide resulted in lack of reactivity at

change of the three T-KD residues, indicating that re-

action with these residues contributed mostly to the

binding energy of this antigen–antibody reaction [11].

The TB1–5 76C monoclonal antibody also reacted with

Mce1A in immunoblotting after native PAGE further

indicating reactivity with a linear epitope in the molecule.

Localization of the TB1–5 76C epitope in the model of

Mce1A

The TB1–5 76C reactive epitope was located at the

surface of the molecule. The side chains of KRRIT,

residues 64–68ð131–135Þ and K, residue 70ð137Þ were ex-

posed at the end of a b strand and the D, residue 71ð138Þ

in a loop region as illustrated in Fig. 5A. The spacefilldiagram (Fig. 5B) shows that the epitope is readily ac-

cessible at the surface. The distance from K64 to D71

was 25�AA and the side chains of R65 and R66 protruded

11�AA apart. The minimal surface area for contact be-

tween the antibody idiotype and the epitope was there-

fore estimated to be about 250–300�AA2. By comparison,

the size of the interface between antibody and antigen

(hen egg white lysozyme) in crystallized complexes hasbeen measured to be from 560 to 844�AA

2[11,12]. These

epitopes are conformational and it is possible that linear

epitopes have a smaller contact area. It is however

necessary to model the antibody together with the an-

tigen to determine the final size of the epitope.

Discussion

Multiple receptors influence mycobacerial entry into

macrophages [13]: complement receptors, mannose re-

ceptors, Fc receptors internalizing IgG-opsonized bac-teria, scavenger receptors, and probably Mce.

Based on previous results by Chitale et al. [7] and

sequence alignments of Mce1A and Mce2A it is con-

ceivable that the epitope for TB1–5 76C represents a

ligand reponsible for mammalian cell entry. The crystal

structure of Colicin N suggested a model of toxicity [10]

Table 1

Activity of monoclonal antibody TB1–5 76C in ELISA with synthetic peptide on the solid phase after replacement of individual amino acid residues

Kind of replacement K R R I T P K D

Basic charge R +++ K +++ K +++ K+++ K ) K) R +++ K )Acidic charge D + D ) D ) D +++ D ) D ) D ) E +

Polar uncharged Q +++ Q +++ Q + S +++ S +++ G ) Q ) N +

Non-polar hydrophobic M +++ M +++ M ++ V +++ V ) V ) M ) V )Alanine A +++ A ++ A ++ A +++ A ) A ++ A ) A )

Activity with control peptide NPTKRRITPKD, adjusted OD value 1.000 in ELISA. OD values> 0.600 recorded as +++, 0.400–0.250 as ++,

0.200–0.100 as +, and <0.050 as ).

Fig. 5. (A) Stereo view of ribbon diagram of the N-terminal domain of

the Mce1A model. The side chains of the epitope (TPKD) of Mce1A

reacting with monoclonal antibody TB1–5 76C are visualized using

electron clouds. (B) Space fill diagram of the Mce1A model. The side

chains of the epitope (TPKD) are colored cyan.

446 A. Kumar Das et al. / Biochemical and Biophysical Research Communications 302 (2003) 442–447

by undergoing a transition to a TM ion channel in thecytoplasmic membrane of the target bacillus leading to

cellular death. This protein can also be exploited to

predict the structure–function relationship of Mce1A.

The b domain with predominant b-sheets helps in poreformation in all the proteins belonging to the Colicin

family, and may be responsible for cell entry in Mce1A.

This idea may be emphasized from the fact that bac-

terial porins perforate the lipid bilayer as a b barrel[14].

The molecular modelling of Mce1A showed that the

epitope is exposed at the surface of the b domain of

Mce1A and comparison with what is known about

structure and function of Colicin N supports the hy-

pothesis that the epitope may be involved in mammalian

cell entry.

Colicins are bacteriocins of Gram negative bacteria.Mycobacteria have more in common with Gram posi-

tive bacteria, having a substantial lipid rich cell wall

responsible for their acid fast properties. However,

mycobacteria also have an outer permeability barrier

[15], which provide features of Gram negative bacteria.

The essential question is therefore whether the Mce

proteins could be bacteriocins or whether they are in

fact primarily virulence factors promoting mammaliancell entry. Mce genes have been found in many myco-

bacterial species [16] including apathogenic species such

as Mycobacterium smegmatis. This may be used as an

argument for Mce primarily being bacteriocins. How-

ever, evolution of pathogenic mycobacteria may have

utilized Mce to generate functionality vs eukaryotic or

mammalian cells by developing ligands to target recep-

tors on host cells. No receptor for Mce1A is known todate, but the original experimental data [1] combined

with the new data on the B-cell epitope and its locali-

zation on the Mce1A model suggest that there may be a

receptor for Mce1A on host cells.

Alignment of the epitope sequence, KRRITPKD,

with the corresponding parts in Mce2A, 3A, and 4A

shows very little conservation of this part of the

Mce1A. This is also the case when comparing the epi-tope sequence with the Mce1A homologs of Mycobac-

terium avium, M. smegmatis, and Mycobacterium leprae

(see alignment of these proteins in Fig. 1 in [16]). These

observations indicate that the sequence, KRRITPKD,

carries a unique functionality of M. tuberculosis

Mce1A.

Acknowledgments

This work was supported by grants from the Indo-Norwegian In-

stitutional Cooperation program supported by the Norwegian Gov-

ernment. We thank Gunni Ulvund for the excellent technical assistance

and Amit Luthra who carried out the energy minimization.

References

[1] S. Arruda, G. Bomfim, R. Knights, T. Huima-Byron, L.W. Riley,

Cloning of an M. tuberculosis DNA fragment associated with

entry and survival inside cells, Science 261 (1993) 1454–1457.

[2] S.T. Cole, R. Brosch, J. Parkhill, T. Garnier, C. Churcher, D.

Harris, S.V. Gordon, K. Eiglmeier, S. Gas, C.E. Barry III, F.

Tekaia, K. Badcock, D. Basham, D. Brown, T. Chillingworth, R.

Connor, R. Davies, K. Devlin, T. Feltwell, S. Gentles, N. Hamlin,

S. Holroyd, T. Hornsby, K. Jagels, B.G. Barrell, Deciphering the

biology ofMycobacterium tuberculosis from the complete genome

sequence, Nature 393 (1998) 537–544.

[3] F. Tekaia, S.V. Gordon, T. Garnier, R. Brosch, B.G. Barrell, S.T.

Cole, Analysis of the proteome of Mycobacterium tuberculosis in

silico, Tuber. Lung Dis. 79 (1999) 329–342.

[4] A. Sali, J. Kuriyan, Challenges at the frontiers of structural

biology, Trends Cell Biol. 9 (1999) M20–M24.

[5] B. Flesselles, N.N. Anand, J. Remani, S.M. Loosmore, M.H.

Klein, Disruption of the mycobacterial cell entry gene of

Mycobacterium bovis BCG results in a mutant that exhibits a

reduced invasiveness for epithelial cells, FEMS Microbiol. Lett.

177 (1999) 237–242.

[6] M. Harboe, A. Christensen, S. Ahmad, G. Ulvund, R.E.

Harkness, A.S. Mustafa, H.G. Wiker, Cross-reaction between

mammalian cell entry (Mce) proteins of Mycobacterium tubercu-

losis, Scand. J. Immunol. 56 (2002) 580–587.

[7] S. Chitale, S. Ehrt, I. Kawamura, T. Fujimura, N. Shimono, N.

Anand, S. Lu, L. Cohen-Gould, L.W. Riley, Recombinant

Mycobacterium tuberculosis protein associated with mammalian

cell entry, Cell Microbiol. 3 (2001) 247–254.

[8] R. Sanchez, A. Sali, Comparative protein structure modeling.

Introduction and practical examples with modeller, Methods Mol.

Biol. 143 (2000) 97–129.

[9] R.A. Laskowski, J.A. Rullmann, M.W. MacArthur, R. Kaptein,

J.M. Thornton, AQUA and PROCHECK-NMR: programs for

checking the quality of protein structures solved by NMR, J.

Biomol. NMR 8 (1996) 477–486.

[10] I.R. Vetter, M.W. Parker, A.D. Tucker, J.H. Lakey, F. Pattus, D.

Tsernoglou, Crystal structure of a Colicin N fragment suggests a

model for toxicity, Structure 6 (1998) 863–874.

[11] D.C. Benjamin, S.S. Perdue, Site-directed mutagenesis in epitope

mapping, Methods 9 (1996) 508–515.

[12] D.R. Davies, G.H. Cohen, Interactions of protein antigens with

antibodies, Proc. Natl. Acad. Sci. USA 93 (1996) 7–12.

[13] S.H. El Etr, J.D. Cirillo, Entry mechanisms of mycobacteria,

Front Biosci. 6 (2001) D737–D747.

[14] C. Lesieur, B. Vecsey-Semjen, L. Abrami, M. Fivaz, v.d.G. Gisou,

Membrane insertion: the strategies of toxins (review), Mol.

Membr. Biol. 14 (1997) 45–64.

[15] L.R. Camacho, P. Constant, C. Raynaud, M.A. Laneelle, J.A.

Triccas, B. Gicquel, M. Daffe, C. Guilhot, Analysis of the

phthiocerol dimycocerosate locus of Mycobacterium tuberculosis.

Evidence that this lipid is involved in the cell wall permeability

barrier, J. Biol. Chem. 276 (2001) 19845–19854.

[16] Y. Haile, D.A. Caugant, G. Bjune, H.G. Wiker, Mycobacterium

tuberculosis mammalian cell entry operon (mce) homologs

in Mycobacterium other than tuberculosis (MOTT), FEMS

Immunol. Med. Microbiol. 33 (2002) 125–132.

A. Kumar Das et al. / Biochemical and Biophysical Research Communications 302 (2003) 442–447 447