Engineering methionine γ-lyase from Citrobacter freundii for anticancer activity

Article abstract of DOI:10.1016/j.bbapap.2018.09.011

Methionine deprivation of cancer cells, which are deficient in methionine biosynthesis, has been envisioned as a therapeutic strategy to reduce cancer cell viability. Methionine γ-lyase (MGL), an enzyme that degrades methionine, has been exploited to selectively remove the amino acid from cancer cell environment. In order to increase MGL catalytic activity, we performed sequence and structure conservation analysis of MGLs from various microorganisms. Whereas most of the residues in the active site and at the dimer interface were found to be conserved, residues located in the C-terminal flexible loop, forming a wall of the active site entry channel, were found to be variable. Therefore, we carried out site-saturation mutagenesis at four independent positions of the C-terminal flexible loop, P357, V358, P360 and A366 of MGL from Citrobacter freundii, generating libraries that were screened for activity. Among the active variants, V358Y exhibits a 1.9-fold increase in the catalytic rate and a 3-fold increase in K_M, resulting in a catalytic efficiency similar to wild type MGL. V358Y cytotoxic activity was assessed towards a panel of cancer and nonmalignant cell lines and found to exhibit IC₅₀ lower than the wild type. The comparison of the 3D-structure of V358Y MGL with other MGL available structures indicates that the C-terminal loop is either in an open or closed conformation that does not depend on the amino acid at position 358. Nevertheless, mutations at this position allosterically affects catalysis.

Full text of DOI:10.1016/j.bbapap.2018.09.011

Accepted Manuscript
Engineering methionine γ-lyase from Citrobacter freundii for
anticancer activity
Samanta Raboni, Svetlana Revtovich, Nicola Demitri, Barbara
Giabbai, Paola Storici, Chiara Cocconcelli, Serena Faggiano,
Elena Rosini, Loredano Pollegioni, Serena Galati, Annamaria
Buschini, Elena Morozova, Vitalia Kulikova, Alexey Nikulin, Edi
Gabellieri, Patrizia Cioni, Tatyana Demidkina, Andrea Mozzarelli
PII:
DOI:
Reference:
S1570-9639(18)30165-1
doi:10.1016/j.bbapap.2018.09.011
BBAPAP 40145
To appear in:
BBA - Proteins and Proteomics
Received date:
Revised date:
Accepted date:
19 June 2018
27 August 2018
25 September 2018
Please cite this article as: Samanta Raboni, Svetlana Revtovich, Nicola Demitri, Barbara
Giabbai, Paola Storici, Chiara Cocconcelli, Serena Faggiano, Elena Rosini, Loredano
Pollegioni, Serena Galati, Annamaria Buschini, Elena Morozova, Vitalia Kulikova,
Alexey Nikulin, Edi Gabellieri, Patrizia Cioni, Tatyana Demidkina, Andrea Mozzarelli ,
Engineering methionine γ-lyase from Citrobacter freundii for anticancer activity. Bbapap
(2018), doi:10.1016/j.bbapap.2018.09.011
This is a PDF file of an unedited manuscript that has been accepted for publication. As
a service to our customers we are providing this early version of the manuscript. The
manuscript will undergo copyediting, typesetting, and review of the resulting proof before
it is published in its final form. Please note that during the production process errors may
be discovered which could affect the content, and all legal disclaimers that apply to the
journal pertain.

ACCEPTED MANUSCRIPT
Engineering Methionine -Lyase from Citrobacter freundii
for Anticancer Activity
Samanta Rabonia,b, Svetlana Revtovichc, Nicola Demitrid, Barbara Giabbaid, Paola
Storicid, Chiara Cocconcellia, Serena Faggianoa,b, Elena Rosinie, Loredano
Pollegionie, Serena Galatif, Annamaria Buschinif, Elena Morozovac, Vitalia
Kulikovac, Alexey Nikuling, Edi Gabellierib, Patrizia Cionib, Tatyana Demidkinac*
and Andrea Mozzarellia,b,h*
aDepartment of Food and Drug, University of Parma, Parma, Italy
bInstitute of Biophysics, National Research Council, Pisa, Italy
cEngelhardt Institute of Molecular Biology of the Russian Academy of Sciences, Moscow, Russia
dElettra Synchrotron Trieste, Trieste, Italy
eDepartment of Biotechnology and Life Sciences – University of Insubria, Varese, Italy
fDepartment of Chemistry, Life Sciences and Environmental Sustainability,
University of Parma, Parma, Italy
gInstitute of Protein Research, Russian Academy of Sciences, Pushchino, Russia
hNational Institute of Biostructures and Biosystems, Rome, Italy
*Corresponding Authors:
Andrea Mozzarelli, Department of Food and Drug, University of Parma, Parma, Italy. Email:
andrea.mozzarelli@unipr.it
Tatyana Demidkina, Engelhardt Institute of Molecular Biology of the Russian Academy of
Sciences, Moscow, Russia. Email: tvd@eimb.ru
1

ACCEPTED MANUSCRIPT
RUNNING TITLE: Methionine -lyase engineering
ABBREVIATIONS:
MGL, methionine γ-lyase; PLP, pyridoxal 5’-phosphate; HO-Hxo-DH, D-2-hydroxyisocaproate
dehydrogenase; DNPH, 2,4-dinitrophenylhydrazine; NADH, nicotinamide adenine dinucleotide
reduced form; DTT, D,L-dithiothreitol; LB, Luria–Bertani; MMEPEG, polyethylene glycol
monomethyl ether; DMEM, Dulbecco's Modified Eagle Medium; FBS, fetal bovine serum.
HIGHLIGHTS
 The C-terminal loop of C. freundii MGL is highly variant
 Most mutations at positions P357, V358, P360 and A366 cause enzyme inactivation
 Variant V358Y showed two-fold increase in catalytic rate and KM.
 Comparison of X-ray structures of MGLs and V358Y suggests a conformational adaptation
to mutations
KEYWORDS: Site-saturation mutagenesis; pyridoxal 5’-phosphate dependent enzyme; protein
engineering; methionine depletion; cytotoxicity activity; cancer therapy.
2

ACCEPTED MANUSCRIPT
ABSTRACT
Methionine deprivation of cancer cells, which are deficient in methionine biosynthesis, has been
envisioned as a therapeutic strategy to reduce cancer cell viability. Methionine γ-lyase (MGL), an
enzyme that degrades methionine, has been exploited to selectively remove the amino acid from
cancer cell environment. In order to increase MGL catalytic activity, we performed sequence and
structure conservation analysis of MGLs from various microorganisms. Whereas most of the
residues in the active site and at the dimer interface were found to be conserved, residues located in
the C-terminal flexible loop, forming a wall of the active site entry channel were found to be
variable. Therefore, we carried out site-saturation mutagenesis at four independent positions of the
C-terminal flexible loop, P357, V358, P360 and A366 of MGL from Citrobacter freundii,
generating libraries that were screened for activity. Among the active variants, V358Y exhibits a
1.9-fold increase in the catalytic rate and a 3-fold increase in KM, resulting in a catalytic efficiency
similar to wild type MGL. V358Y cytotoxic activity was assessed towards a panel of cancer and
nonmalignant cell lines and found to exhibit IC50 lower than the wild type. The comparison of the
3D-structure of V358Y MGL with other MGL available structures indicates that the C-terminal
loop is either in an open or closed conformation that does not depend on the amino acid at position
358. Nevertheless, mutations at this position allosterically affects catalysis.
3

ACCEPTED MANUSCRIPT
1. INTRODUCTION
A striking metabolic difference between normal and many human tumor cells is the dependence of
cancer cells on L-methionine, an essential amino acid, for survival and proliferation. Normal cells
do not suffer from methionine deficiency since they possess the ability to grow on homocysteine,
one of methionine metabolic precursor. The high frequency of methionine auxotrophy in cancer
cells suggested that it might be a general metabolic feature of cancer cells, termed “Hoffman effect”
[1], and an important aspect of oncogenic transformation, analogous to the “Warburg effect” for
glucose. Consequently, methionine depletion has gained increased recognition as a tumor specific
therapeutic strategy. This, in turn, has led to the exploitation of methionine-restricted diet and
methionine-degrading enzymes for the depletion of plasma methionine to limit cancer cells
proliferation [2, 3]. The most frequently used enzyme is methionine γ-lyase (EC 4.4.1.11, MGL),
which degrades L-methionine to α-ketobutyrate, ammonia and methanethiol via a ,-elimination
reaction [4]. Several strategies have been adopted to deliver MGL to cancer cells. These include
free and polyethylene glycol-conjugated MGL [5-8], erythrocytes loaded with MGL [9], stealth
PLGA/liposome-containing MGL [10] and gene therapy based on adenoviral vector carrying the
MGL gene [11]. Recent studies proved the efficacy of MGL in a patient-derived orthotopic
xenograft model, suggesting potential clinical development, especially in recalcitrant cancers such
as Ewing’s sarcoma [12-17].
MGL is assigned to the aspartate aminotransferase family of pyridoxal 5’-phosphate (PLP)-
dependent enzymes and exhibits the structural features characteristic of a fold type I [18-21]. The
three-dimensional structure of wild type MGL in the absence and presence of substrate analogues
has been determined [22-29]. MGL is a tetrameric enzyme composed of four identical subunits with
two tightly associated dimers forming the active site at the dimer interface. Each subunit consists of
an N-terminal domain, the PLP-binding domain and the C-terminal domain. PLP is located within a
4

ACCEPTED MANUSCRIPT
rather deep cavity on the dimer surface bound via a Schiff base to K210 (C. freundii numbering)
[23, 24, 27, 30].
Considering the therapeutic potential of MGL as an anticancer agent, our goal was to engineer the
enzyme from C. freundii to improve its catalytic activity by using semi-rational design, known as
site-saturation mutagenesis [31-34]. The integration of structural information and evolution analysis
of MGL guided the selection of candidate residues to be replaced. A parallel effort is aimed at the
development of suitable MGL formulation for in vivo delivery [35].
2. MATERIALS AND METHODS
2.1. Materials - L-Methionine, D,L-homocysteine, S-methyl-L-cysteine, cysteine, nicotinamide
adenine dinucleotide reduced form (NADH), PLP, ampicillin, Luria-Bertani broth, tryptone, and
yeast extract were purchased from Sigma-Aldrich (San Louis, MO). D,L-dithiothreitol (DTT) was
from Applichem (Darmstadt, Germany). DEAE-sepharose was from GE Healthcare Life Sciences
(Pittsburg, PA, USA). The plasmid with the gene of HO-Hxo-DH was a kind gift of K. Muratore
(University of California, Berkeley, Department of Molecular and Cell Biology, Berkeley, CA,
USA). The E. coli strain BL21 (DE3) F- ompT hsdSB gal dcm (DE3) (Merck, Darmstadt,
Germania) was used to express C. freundii MGL and the mutants.
2.2.
Structure and sequence comparison - The 37 structures of MGL available on Protein Data
Bank [36] were downloaded and visually inspected using PyMOL (PyMOL Molecular Graphics
System, Version 1.8 Schrödinger, LLC.) in order to select one crystallographic structure for each
microorganism. The selection took into account chain integrity, the absence of mutations and the
presence of ligands. Among them, six wild type structures were selected for further analysis (PDB
ID: 4OMA, 2O7C, 3ACZ, 4U1T, 1E5F, 5DX5) [28, 29, 37]. A library of 3558 primary sequences
of MGL belonging to several microorganisms was downloaded in FASTA format from RefSeq
Database of NCBI Protein Data bank [38]. Only lyases whose sequence length was between 370
5

ACCEPTED MANUSCRIPT
and 420 amino acids were selected. Redundancy in the organism name or in the amino acid
sequence was avoided using a python script. At the end of this procedure, 1223 sequences were
obtained. In order to reduce the number of sequences, a homology based sequence alignment was
performed using the jackhmmer program included in HMMER3 suite [39]. This iterative protein
search used as query the protein sequences belonging to the six microorganisms whose crystal
structure was present in Protein Data Bank. The protein database was formed by the 1223 sequences
without the sequence under analysis. The first 100 homologous sequences for each of the six query
proteins were merged and the sequences that appeared more than once were removed. A library
containing 354 sequences was generated. The alignment of these sequences was performed with
ClustalO algorithm included in Seaview 4 program [40, 41] available for ubuntu 16.4 operating
system. The residue conservation analysis was performed with ConSurf server [42-44] for each
crystallographic structure. Results were visualized and inspected with PyMOL.
2.3. Site-saturation mutagenesis - Site-saturation mutagenesis was carried out at four
independent amino acid positions by using the QuikChange II XL site-directed mutagenesis kit
(Agilent Technologies, Santa Clara, CA, USA) with the recombinant expression vector pET-mgl
containing the wild type mgl gene [45] as a template, as detailed in [34]. The mutation is introduced
in a single PCR with one pair of complementary degenerate synthetic primer (Table S1). All
primers were synthesized and purified by Eurofins Genomics (Ebersberg, Germany). For all
primers, randomized positions are in boldface with Nꢀ=ꢀA/C/G/T (Table S1). The PCR
amplifications were carried out with PfuUltra High Fidelity DNA polymerase (Agilent
Technologies, Santa Clara, CA, USA). The 50 μl PCR reaction mixture was carried out with 50 ng
template, 2.5 U of PfuUltra High Fidelity DNA polymerase and with dNTPs and primer pair
concentrations according to manufacturer’s protocol. The extension reaction was initiated by pre-
heating the reaction mixture at 94 °C for 1 min; 18 cycles of 94 °C for 50 s, 60 °C for 50 s and 68
°C for 9 min according to the length of template; followed by incubation at 68 °C for 10 minutes.
The template DNA was eliminated by enzymatic digestion with the DpnI restriction enzyme,
6

ACCEPTED MANUSCRIPT
specific for methylated DNA. The PCR products were used to transform BL21(DE3) E. coli cells
and grown on Luria–Bertani (LB) plate containing 100 μg/ml ampicillin overnight at 37 °C. These
clones were used for the screening procedure. For each position, 282 clones were screened and 10
clones were sequenced to verify the mutagenesis efficiency.
2.4. HTP Screening for high activity MGL variants - The mutant libraries obtained from site-
saturation mutagenesis were screened by a rapid colorimetric assay based on the determination of
2-ketobutyryl 2,4-dinitrophenylhydrazone performed with an automated liquid-handler system
(epMotion 5075, Eppendorf, Hamburg, Germany) [34, 46]. Single recombinant E. coli cultures
(1 mL, grown in DeepWell plate, Eppendorf, Hamburg, Germany) were incubated
in
autoinduction medium [47] at 37 °C for 24 h. An aliquot (200 μL) of each culture was transferred to
a 96-well microtiter plate. Cell lysis was performed by adding 50 μL of lysozyme to a final
concentration of 0.5 mg/ml. The catalytic activity was assayed on the crude extract by adding the
substrate (25 mM L-Met final concentration) and incubating the plates at 37 °C for 20 minutes.
Then, 2,4-dinitrophenylhydrazine reagent (1 mM DNPH in 1 M HCl) was added to a final
concentration of 0.3 mM, and incubated at 37 °C for 10 min. The -ketobutyrate is derivatized to
a phenylhydrazone. The color of the phenylhydazone is developed by adding NaOH to a final
concentration of 0.4 M. The absorbance of the mixture was measured at 450 nm by a microtiter
plate reader (Sunrise, Tecan, U.K., Goring-on-Thames, U.K.) and compared to cultures expressing
the wild type MGL. The variants that outperformed the control were selected for further
characterization. Mutations were confirmed by automated DNA sequencing carried out by
Microsynth GmbH (Balgach, Switzerland).
2.5. Enzyme expression and purification - E. coli BL21(DE3) cells containing the plasmid with
the gene of mutant C. freundii MGL were grown and the enzyme was purified according to the
protocol used for wild type [48]. The concentration of homogeneous protein was determined by the
absorbance at 278 nm (0.1%
= 0.8) [49]. Protein purity was assessed by SDS-PAGE under
278
7

ACCEPTED MANUSCRIPT
denaturating conditions and found to be 90%. The protein was stored at -80 °C in buffer 100 mM
potassium phosphate (pH 8.0), 1 mM EDTA, 5 mM DTT, 100 M PLP.
2.6. Activity assay - MGL activity was determined using L-methionine as a substrate by
measuring the rate of α-ketobutyrate production in the coupled reaction with HO-HxoDH
monitoring the decrease of NADH absorption at 340 nm (Δε = 6220 M–1cm–1) at 30 °C. The
reaction mixture contained 100 mM potassium phosphate, pH 8.0, 100 M PLP, 5 mM DTT, 0.2
mM NADH, 70 U HO-HxoDH, and 30 mM L-methionine. One unit of enzyme activity is defined
as the amount of the enzyme that catalyzes the formation of 1.0 mol min-1 of -ketobutyrate at pH
8.0, 30 °C.
2.7. Determination of the steady state parameters of the -elimination reaction and the -
elimination reaction - The steady state parameters of the-elimination reaction were determined
by the rates of -ketobutyrate formation in a coupled reaction with HO-Hxo-DH [50] under the
conditions described above at pH 7.2, 30 °C, by varying the substrate concentration in the reaction
mixture. The steady state parameters of the -elimination reaction were determined by the rates
of pyruvate formation in a coupled reaction with lactate dehydrogenase under the conditions
described above at pH 7.2, 30 °C, by varying the substrate concentration in the reaction mixture.
Data were processed according to the Michaelis–Menten equation. The molecular mass of the
enzyme subunit, 43 kDa, was used in the calculations.
2.8. Circular dichroism - All samples were prepared at a protein concentration of 3 μM
(protomer concentration) in a buffer containing 100 mM potassium phosphate, pH 7.4. CD spectra
were recorded at 25 °C with a Jasco J-715 spectropolarimeter equipped with a Peltier temperature
control unit (Jasco Easton, MD, USA). The far-UV measurements were recorded from 190 to 260
nm using a path length of 1 mm and the spectra obtained were the average of five scans at a scan
rate of 50 nm/min. The buffer contribution was subtracted from the final spectra.
8

ACCEPTED MANUSCRIPT
2.9. Absorbance and fluorescence measurements - Absorption spectra were carried out using a
double beam Cary4000 UV/Vis spectrophotometer (Agilent Technologies, Santa Clara, CA, USA).
Fluorescence measurements were performed using a Fluoromax-3 spectrofluorometer (Horiba Jobin
Yvon, Milan, Italy). Excitation wavelengths of 295 nm, 330 nm and 415 nm were used, with an
excitation and emission slit width of 3-5 nm. All spectra were corrected for buffer contribution.
Measurements were carried out at 20 °C in 100 mM potassium phosphate, pH 7.4, at 13 μM MGL
(protomer concentration).
2.10. Evaluation of in vitro cytotoxicity by MTS assay - The cytotoxic activity was evaluated for
the human prostate (PC-3), mammary gland (MCF7), colon (HT29) adenocarcinoma and human
skin fibroblast (Hs27) cell lines. In the exponential growth phase, cells were seeded at 5 x 104/mL
into 96-well flat bottom microplates. Cells were cultured in DMEM supplemented with 5% FBS,
1% L-glutamine and 1% penicillin/streptomycin at 37 °C and 5% CO2. After 24h from seeding,
wild type and V358Y MGL contained in a sterile PBS solution, pH 7.4, were added to DMEM
medium supplemented with 5% FBS and 0.5 mM PLP in PBS and co-incubated for 72 h. Enzyme
concentrations in the culture medium were in the range 0.0005–0.5 mg/mL. An equal volume of
PBS with 0.5 mM PLP was added to DMEM medium in the control wells. The anti-proliferative
activity was evaluated by a colorimetric assay (CellTiter96R Aqueous One Solution Cell
Proliferation Assay, Promega Corporation, Madison, WI, USA). A small amount of CellTiter96R
AQueous One Solution Cell Proliferation Assay was added directly to culture wells, incubated for 4
h and then the absorbance recorded at 450 nm with a 96-wells plate reader (SPECTRAFluor, Tecan
U.K., Goring-on-Thames, U.K.). Cytotoxicity experiments were carried out in quadruplicate.
2.11. Crystallization and data collection - Crystals of V358Y MGL were obtained at 30°C by the
hanging drop vapor diffusion at the same conditions as described [24] using 35% (w/v)
polyethylene glycol monomethyl ether (MMEPEG) 2000 as the precipitant. Rhombic crystals grew
up to 0.3-0.4 mm dimension within 10 days.
9

ACCEPTED MANUSCRIPT
Diffraction data were collected at the Elettra XRD1 beamline (Trieste, Italy) using Pilatus 2M
(Dectris) detector and processed by XDS program [51]. Crystals belonged to the space group I222
with unit cell parameters a = 56.67 Å, b = 122.87 Å, c = 127.52 Å, α=β=γ=90° and contained one
subunit in an asymmetric unit. Detailed data collection statistics are shown in Table 1.
Table 1. Data collection and refinement statistics. Values in parentheses are for the highest
resolution shell. CC1/2 is the Pearson’s correlation coefficient calculated for Imean by splitting the
data randomly in half by AIMLESS/SCALA [52].
Space group
I222
Unit cell parameters (Å) a = 56.67 Å, b = 122.87 Å, c = 127.52 Å
 = β =  = 90°
Wavelength (Å)
Resolution (Å)
Completeness (%)
Redundancy
1.00
50-1.45 (1.50-1.45)
99.47 (99.7)
2.25 (2.36)
0.026 (0.434)
0.037 (0.614)
7.2 (2.0)
Rmerge (%)
Rmeas (%)
Mean((I)/sd(I))
CC(1/2)
0.999 (0.837)
1, 398
Disordered protein
residues
No. of non-H protein
atoms
3635
383
No. of water atoms
No. of unique reflections 78624
R/ Rfree
0.138/0.174 (0.214/0.266)
3915 (4.98%)
Rfree test set reflections
Mean temperature factor 26.00
B (Å2)
10

ACCEPTED MANUSCRIPT
R.m.s. deviation from
ideal values
Bond lengths (Å)
0.005
0.844
0.079
0.006
Bond angles (°)
Chirality angles (°)
Planar angles (°)
Ramachandran plot
Favoured region (%)
Allowed region (%)
Outlier region (%)
97.64
1.77
0.59 [S190, K210] †
†
The conformation of the amino acid residues at this position is a characteristic feature of the
protein [24].
2.12. X-ray structure determination and refinement - The structure was solved by molecular
replacement using the structure of C. freundii MGL (PDB code 2RFV) by rigid body procedure,
implemented in PHASER of the CCP4 software suite [53]. Flexible loops of the protein and water
molecules were removed from the initial model to exclude model bias during the first round of
refinement. The model was improved using manual rebuilding with COOT [54] and maximum
likelihood refinement using REFMAC5 [55]. An excess of electron density at the sulfur atom of C4
and C115 forced us to change these residues to 3-sulfenoalanine. The final step of the structure
refinement was performed to 1.45 Å using the phenix.refine procedure implemented in the PHENIX
suite [56] with Rwork of 13.8% and Rfree of 17.4% (Table 1). The final model contains 3635 non-
hydrogen atoms including the internal aldimine, two molecules of MMEPEG 2000 and 346 water
molecules. The structure was submitted to the Protein Data Bank with PDB code 6EGR and
validated.
3.
RESULTS AND DISCUSSION
11

ACCEPTED MANUSCRIPT
3.1. Identification of “hot spots” for MGL engineering by phylogenetic analysis - The
integration of sequence and structure-based analysis in protein engineering is a powerful approach
for the identification of residues affecting the functional properties. The positions more prone to
accommodate diversity are good targets for site-saturation mutagenesis. Such a design strategy
possesses the advantage of an increased probability of beneficial mutations combined with a
significant reduction of the library size. The degree of amino acid variation was performed using the
NCBI protein databank, filtered for MGL-containing organisms available in Refseq database. Only
lyases were selected and the chain length cut-off was set between 370 and 420 amino acid. These
sequences were parsed to remove redundancy. The resulting 1223 sequences were used to perform a
homologues search using as query each of the six MGL sequences from microorganisms whose 3D
structures are deposited in the Protein Data Bank: Citrobacter freundii (UniProtKB AC: Q84AR1),
Pseudomonas putida (UniProtKB AC: P13254), Entamoeba histolytica (UniProtKB AC: Q86D28),
Trichomonas vaginalis (UniProtKB AC: O15564), Micromonospora echinospora (UniProtKB AC:
Q8KNG3) and Clostridium sporogenes (UniProtKB AC: J7TA22). Following filtering to remove
duplicates, 354 different protein sequences were obtained, aligned and used to perform the residue
conservation analysis with Consurf web server. The projection of the conservation degree onto the
molecular structure of MGL revealed patches of highly conserved residues localized in regions with
relevant biological function such as the active site and the dimer interface (Figure S1). Not
unexpectedly, the Consurf analysis shows that the residues building the active site are invariant
(Figure 1a). The percentage of residue conservation for amino acids positioned within a 8 Å sphere
centered on the PLP is reported in Table 2. Furthermore, also residues at the dimer interface show
high Consurf conservation scores (Figure 1b, Table 2) because hydrogen bonds and electrostatic
interactions stabilize the dimeric unit.
12

ACCEPTED MANUSCRIPT
(a)
(b)
13

ACCEPTED MANUSCRIPT
Figure 1. (a) Residues of C. freundii MGL active site in a 8 Å shell from PLP (Consurf coloring –
PLP in yellow). Asterisks mark residues belonging to the second subunit of the active dimer. (b)
Residues of C. freundii MGL protomer involved in the dimer interface (Consurf coloring – PLP in
yellow).
Table 2. Amino acid conservation analysis of MGL. The percentage was obtained from the Consurf
analysis of 354 proteins.
Active site
Y58
%
%
%
100
100
100
100
99.7
98.8
99.1
E156
N160
D185
T187
F188
H206
S207
100
100
100
100
50.6
99.1
100
T209
K210
S339
L340
R374
99.7
100
100
100
100
R60
G88
Y113
G114
C115
T116
14

ACCEPTED MANUSCRIPT
Dimer interface
%
%
%
Q33
S35
T36
F37
F39
F49
I57
69.2
68.6
99.7
84.1
99.7
99.4
57.1
100
100
S87
99.7
100
98.8
99.7
100
98.3
100
57
D240
T242
G243
R256
V338
S339
S352
M353
T254
98.3
96.3
100
100
99.7
100
95.8
100
100
%
Y113
C115
T209
K210
G216
D217
V218
K239
Y58
R60
98.3
%
Active site channel %
H349
P350
A351
S352
M353
T354
99.7
H355
S356
P357
V358
A359
P360
100
93.2
46.9
11.9
4.2
E361
E362
R363
L364
K365
A366
83
95.7
99.7
95.8
100
100
88.7
91.8
18.1
16.1
45.5
54.7
Differently from the active site and the dimer interface, the amino acids at the protein surface,
exposed to the solvent, show a higher degree of variability (Figure 2). In particular, a stretch of
amino acids following the highly conserved sequence of helix 16 (amino acids 349 – 356) is
hypervariant (Figure 2, Table 2). Consurf results show that position 357 harbors a proline in 47% of
the sequences, threonine in 25% and serine in 15%. At position 358, endowed with an intermediate
conservation score, -0.586, valine is present in 12% of MGLs, including C. freundii, tyrosine in
80%, isoleucine in 3% and methionine in 2%. Proline at position 360 is substituted by lysine (10%),
glutamic acid (6%), alanine (15%) and arginine (8%). Instead, residue 366 is alanine in 45% of
MGLs with the main variations being histidine (37%) and methionine (3%), serine (7%) and
tyrosine (4%). These very variable residues flank the entrance of the channel leading to the active
15

ACCEPTED MANUSCRIPT
site of MGL. In the attempt to engineer enzymes, the active site is usually regarded as the obvious
starting point. However, several studies proved that substitution of residues located in distal region
of the protein matrix might be a successful strategy to modulate the enzyme catalytic activity, since
their impact on function may be assessed without disturbing the enzyme-substrate interface and
without disrupting the active site architecture [57, 58]. Thus, these four non-conserved amino acids
located in the active site access channel were selected as target residues in the search for variants
with improved catalytic efficiency of MGL.
(a)
16

ACCEPTED MANUSCRIPT
(b)
Figure 2. (a) Residues delimiting the access to the binding cavity (Consurf coloring of side chains –
PLP in yellow). Residues of the neighboring subunit are marked with asterisks. (b) Sequence
conservation of amino acids 349-366 of MGL generated by Weblogo plot [59]. Data for this logo
consist of 354 sequences. The overall height of each stack is proportional to the sequence
conservation at each position measured in bits. The relative size of each letter indicates the
frequency in the sequences. Letters are ordered from the most to the least frequent amino acid.
Hydrophilic amino residues are colored in blue; neutral are in green and hydrophobic in black.
Construction of site-saturation mutagenesis libraries - Site-saturation mutagenesis experiments
were independently performed at positions 357, 358, 360 and 366 of C. freundii MGL. The ,-
17

ACCEPTED MANUSCRIPT
elimination activity of the generated variant libraries was evaluated by high-throughput screening
on a 96-well microtiter plate using an automated liquid-handler system. For each position, 282
clones were screened, a number that gives a probability of approximately 96% of obtaining clones
possessing each single amino acid at the investigated position. The DNPH discontinuous
colorimetric activity assay, described in Material and Methods, allowed selecting clones with the
highest catalytic activity in the bacterial cells lysate. Ten variants with beneficial substitutions were
identified by the screening procedure. Nevertheless, useful information on the substitutability, i.e.
the ability to accept sequence changes without loss of function, are obtained from the percentage of
inactive clones. When performing site-saturation mutagenesis at position 357 and 360, about one
third of clones lost their function indicated by an absorbance intensity in the colorimetric assay less
or equal to the 10% of the wild type signal. At position 358, almost half of the clones were inactive.
Upon substitution at position 366, only one tenth of clones were inactive. Altogether, these data
show that MGL is recalcitrant to random codon changes despite the mobility and exposure to the
solvent of this loop. Because the mutations at positions 357, 358, 360, and 366 are in a mobile loop
not involved in the dimer-of-dimers assembly, it is unlikely that loss of activity is caused by
tetramer misassembly but may be plausibly due to protein misfolding. In turn, this finding supports
the notion of a tight link between distal portions of the protein matrix and the catalytic core.
Interestingly, among the novel sequence combination generated by site-saturation mutagenesis in C.
freundii MGL, some residues have been already naturally selected and are present in protein
homologous sequences. In particular, the randomly selected Tyr exhibits a high frequency of
occurrence at position equivalent to 358 in other species, especially those endowed with high
catalytic activity (see below). In addition, the variant P360K identified by the library screening
introduces in C. freundii the amino acid present at the equivalent position in T. vaginalis, C. novyi
and C. tetani, while Y366 is the amino acid present in P. putida at the equivalent position.
All ten variants identified by the screening procedure were expressed in BL21(DE3) E. coli cells
and purified. For all, the expression yield was similar to the wild type enzyme, i.e. about 60 mg/L.
18

ACCEPTED MANUSCRIPT
An increase in the specific activity for the ,-elimination reaction of L-Met, using the D-2-
hydroxyisocaproate dehydrogenase (HO-Hxo-DH) coupled assay, was observed for seven of the ten
purified variants (Figure S2). The kinetic parameters (kcat and KM values) were determined for the
most active variants of MGL at each position and compared to the values of the wild type (Table 3).
A 1.9-fold increase in kcat was measured for V358Y, while all other variants showed comparable
kcat with respect to the wild type. All variants showed an increased KM resulting in a decreased
catalytic efficiency (kcat /KM). This finding further supports a long distance effect brought about by
single amino acid replacement of channel-lining residues possibly via alteration of the
conformational landscape.
Table 3. Kinetic parameters of MGL mutants in the ,-elimination reaction with L-methionine.
The rate of α-ketobutyric acid production was measured in a coupled assay at 30 °C. The reaction
mixture contained 100 mM potassium phosphate, 100 M PLP, 5 mM DTT, 0.2 mM NADH, pH
7.2.
MGL
kcat (s-1)
7.9 ±0.3
9.0±0.3
15.1±0.6
7.4±0.4
8.0±0.7
KM (mM)
0.8±0.1
3.7±0.4
2.7±0.3
1.6±0.3
5.5±1.3
kcat/KM (M-1s-1)
9.7x103
wild type
P357I
2.4x103
V358Y
P360Q
A366Y
5.6x103
4.6x103
1.5x103
A more detailed characterization of the catalytic properties was carried out for MGL V358Y, the
variant that showed a higher catalytic rate. Specifically, the kinetic parameters were determined for
19

ACCEPTED MANUSCRIPT
C. freundii wild type MGL and MGL V358Y using different substrates that undergo the ,-
elimination (methionine and homocysteine) and/or ,-elimination (S-methyl-L-cysteine and
cysteine) (Table 4).
Table 4. Kinetic parameters of C. freundii MGL V358Y in the ,-elimination and ,-elimination
reaction with different substrates. The rate of α-ketobutyric acid or pyruvate production was
measured in a coupled assay, at 30 °C, for the ,-elimination and ,-elimination, respectively.
The reaction mixture contained 100 mM potassium phosphate, 100 M PLP, 5 to 10 mM DTT, 0.2
mM NADH, pH 7.2.
C. freundii MGL wild type
C. freundii MGL V358Y
kcat (s-1)
7.9±0.3
KM (mM) kcat/KM (M-1s-1) kcat (s-1) KM (mM)
kcat/KM (M-1s-1)
5.6x103
L-methionine
0.8±0.1
9.7x103
15.1±0.6 2.7±0.3
D,L-homocysteine
3.5±0.2
2.2±0.1
1.1±0.2
0.8±0.2
3.2x103
2.7x103
7.8±0.5
5.1±0.1
2.2±0.4
3.3±0.3
3.5x103
1.5x103
S-methyl-L-
cysteine
cysteine
1.5±0.1
0.3±0.1
5.0x103
2.1±0.1
0.6±0.1
3.5x103
For the wild type MGL, the catalytic parameters are similar to those previously determined using a
discontinuous assay for the -elimination and a continuous assay for the -elimination [48]. The
variant V358Y exhibits consistently a two-fold increase in the catalytic rate for all analyzed
20

ACCEPTED MANUSCRIPT
substrates as well as a two- to three-fold increase in the KM resulting in catalytic efficiencies equal
or lower than the wild type enzyme. This finding suggests that the KM values might not represent
the affinity between substrate and enzyme but rather reflect the increase in kcat. Under this
assumption, KM, that following the Michaelis-Menten model is equal to:
KM = (kcat + k-1)/k1
eq. 1
where k1 and k-1 are the rate constants for the formation and disappearance of the enzyme-substrate
complex, can be approximated to
KM = kcat/k1
eq. 2
according to the model proposed by Van Slyke-Cullen for irreversible reactions [60]. The rate
constant in eq. 2 is, thus, an “apparent” bimolecular association rate constant, mathematically
equivalent to the specificity constant [61]. It is worthy to point out that the V358Y mutant displays
similar effects, i.e. simultaneous increase in kcat and KM on different substrates and with different
catalytic mechanisms ( elimination and  elimination). The increase in kcat caused by V358Y
mutation is not easily explained based on differences in the three dimensional structure or in the
conformational flexibility (see below). Thus, one may assume that the higher kcat value for the
variant enzyme is due to an acceleration of a specific step(s) of the reaction after aminocrotonate
formation or in products release. Earlier, we suggested [29] that a movement of C-terminal region
of C. freundii MGL may assist the egress of the products of the physiological reaction. Therefore,
mutations that affect loop mobility or disorder-to-order rearrangement alter the conformational
landscape and the catalytic parameters.
3.2. Spectroscopic characterization of V358Y MGL - Circular dichroism spectra were recorded
for V358Y and compared to wild type MGL (Figure 3a). The spectrum shows two minima at 208
21

ACCEPTED MANUSCRIPT
and 222 nm, typical of proteins with a high helical content. Estimation of the percentages of protein
secondary structure calculated with K2D (Dichroweb) from far-UV circular dichroism spectra of
the variant enzyme indicates a small change in secondary structure content. The UV-visible
absorption spectrum of V358Y is similar to that of the wild type enzyme (Figure 3b). It is
characterized by the typical protein band at 280 nm and a prominent band centered at 420 nm,
assigned to the ketoenamine form of PLP. The band at 330 nm, attributed to the enolimine tautomer
of the Schiff base of PLP [49], is less pronounced suggesting a shift of the enolimine-ketoenamine
equilibrium towards the latter. The protein fluorescence emission spectra of MGL obtained upon
excitation at 295 and 415 nm are reported in Figure 3c-d. The tryptophan fluorescence spectrum
exhibits an emission maximum slightly blue shifted and less intense with respect to the wild type
(Figure 3c). As previously reported [35], this emission originates from the single Trp252 located in
a partially apolar environment. The blue shift might be associated to a small conformational change,
as hinted by the circular dichroism data. The coenzyme fluorescence spectrum upon excitation at
330 nm, the wavelength where the PLP enolimine tautomer absorbs, exhibits a slightly blue shifted
and less intense band (data not shown). When MGL is excited at 415 nm, the wavelength where the
ketoenamine tautomer of PLP absorbs, an emission band centered at 494 nm is observed (Figure
3d), more intense for the mutant with respect to wild type. Overall, the spectroscopic properties of
V358Y are slightly altered by the amino acid substitution, indicating that in spite of the distant
positioning of tyrosine from the PLP binding site the cofactor microenvironment and the active site
architecture are slightly perturbed.
22

ACCEPTED MANUSCRIPT
Figure 3. Comparison of the spectroscopic properties of V358Y (solid line) and wild type (dashed
line) MGL. (a) CD spectra in 100 mM potassium phosphate pH 7.4. (b) Absorption spectra. (c)
Fluorescence emission spectrum upon excitation at 295 nm (slit 3 nm). (d) Fluorescence emission
spectrum upon excitation at 415 nm (slit 5 nm).
3.3. Cytotoxicity of V358Y MGL - We quantified the antiproliferative effect of V358Y MGL on
normal and cancer cells by MTS cytotoxicity assay. Specifically, we tested prostate (PC-3), breast
(MCF7), and colon adenocarcinoma (HT29) cell lines. Healthy dermis fibroblasts (Hs27) were used
as a control. V358Y MGL selectively inhibits cancer cell growth with a half-maximal inhibitory
concentration (IC50) almost two-fold lower than the wild type MGL (Table 5). This finding is quite
encouraging, even if the decrease in IC50 may not be a sole consequence of the increased kcat since
23

ACCEPTED MANUSCRIPT
other factors, such as enzyme stability, coenzyme affinity, resistance to oxidation and proteolysis,
might play a role. Furthermore, a full understanding of the mechanisms triggered by methionine
depletion has not yet been reached. Upon methionine deprivation, cancer cell are reported to arrest
in the S/G2 phase of the cell cycle and to be susceptible to spontaneous death and hypersensitive to
chemotherapeutic agents [62, 63]. Based on our preliminary experiments, the mechanism of cell
death is not via apoptosis.
Table 5. Comparison of cytotoxicity activity (IC50) of V358Y and wild type C. freundii MGL on
normal and cancer cell lines.
Prostate
Breast
Colon
Fibroblasts (Hs27)
adenocarcinoma adenocarcinoma
adenocarcinoma
(HT29)
(PC3)
(MCF7)
wild type
V358Y
37±7 g/ml
22±3 g/ml
33±1 g/ml
17±1 g/ml
34±2 g/ml
17±1 g/ml
>250 g/ml
>250 g/ml
3.4. X-ray structure of V358Y MGL - The crystal structure of the V358Y variant was solved at
1.45 Å resolution. The structural features of V358Y holoenzyme are similar to those of wild type
holoenzyme and of the complexes of the wild type enzyme with inhibitors [27, 29] and substrates
[25]. The crystals exhibit the same space group of symmetry I222 and contain one protomer in an
asymmetric unit (Table 1). As for the wild type enzyme, V358Y is a tetramer, which consists of two
catalytic dimers with the active sites formed by residues from two neighboring subunits [23] (Figure
24

ACCEPTED MANUSCRIPT
S3a). Flexible regions previously identified in wild type MGLs are also found in the structure of
V358Y: the long flexible N-terminal loop (residues 47-63) and two C-terminal regions (residues
160-163 and 353-368) (Figure S3b) [27]. These regions are located in the proximity of the active
sites. With the exclusion of these flexible regions, the overall fold of the polypeptide chains of
V358Y and wild type MGL (PDB code 2RFV) is almost identical (root-mean-square deviations of
main chain atoms are 0.14 Å2). The positions of amino acids forming the active site as well as the
PLP site are the same in V358Y and in previously determined structures of C. freundii MGL.
However, four active site residues have alternative positions in V358Y. The displacement of three
of them is the result of the interconnected movement of residues N160, F188 and R374 (Figure S4).
N160 is highly conserved in MGL from different sources (Table 2) [48] and is structurally invariant
in a number of PLP-dependent enzymes of type I fold [18]. Movement of N160 is important for
effective substrate binding and abstraction of the Cα-proton from the external aldimine [27]. The
position of R374 side chain also varies (Figure S3). This flexibility allows R374 to "catch" and
orient the substrate at the active sites. Due to the movement of N160 and/or R374, the ring of F188
is forced to shift downward towards the hydrophobic region formed by L163, M189 and I313.
F188Y substitution in C. freundii MGL lowers the ,-elimination activity (Demidkina,
unpublished results). Thus, it is likely that a tyrosine at position 188 may limit mobility of N160
and preclude its involvement in catalysis. Noteworthy, tyrosine at the equivalent position in P.
putida MGL is located near E156 and N160 and occupies a slightly different position. C115 is the
fourth residue showing two conformations in V358Y. This residue is conserved in most
heterologous MGLs. It provides efficient L-methionine binding [28] and assists ,-elimination [4,
64]. It is an “accessible” cysteine, which can be modified by allicine to S-allylmercaptocysteine
[64]. In V358Y crystal structure, C115 is oxidized to 3-sulfenoalanine. Since V358Y efficiently
catalyzes L-methionine decomposition, it is plausible that C115 oxidation occurs during
crystallization or is the result of the damage induced by X-ray radiation during data collection at the
25

ACCEPTED MANUSCRIPT
synchrotron. We previously observed 3-sulfenoalanine at 115 position in the structure of wild type
MGL with external enolimine with ammonia (PDB code 3MKJ).
3.5. Conformational effect of V358Y substitution - Residue 358 belongs to the unstructured part
of the C-terminal highly flexible region 353-368. Comparison of MGLs structures from different
sources confirms flexibility of this region (Figure 4a). This region makes no contacts with the rest
of the protein matrix, does not interact with the other wall of the channel (residues 111-114) and is
exposed to the solvent (Figure S3a). Despite flexibility, the 353-368 mobile region occupies a
similar position in all the determined X-ray structures of C. freundii MGL. In this region, the largest
deviation (~2 Å with respect to wild type holoenzyme) is calculated for V358Y and S339A (PDB
code 5D5S) structures (Figure S3c).
26

ACCEPTED MANUSCRIPT
Figure 4. Superposition of the Cα traces of MGLs from various species. C. sporogenes MGL (PDB
code 5DX5) chain A is in red and chain B is in brown. C. freundii MGL (PDB code 2RFV) is in
cyan and V358Y MGL (PDB code 6EGR) is in black. T. vaginalis MGL is in green (PDB code
1E5F) and is in gray (PDB code 1PFF). E. histolytica MGL (PDB code 3ACZ) is in yellow and M.
echinospora MGL (PDB codes 4U1T) is in magenta. P. putida MGL is in blue (PDB code 2O7C)
and is in white (PDB code 5X2X). (a) Overall structure of MGLs. (b) C-terminal mobile region, I is
marked for “open” and II for “closed” conformation of active site. (c) Superposition of V/Y358 of
C. freundii MGL and corresponding tyrosine in the structure of P. putida and C. sporogenes.
27

ACCEPTED MANUSCRIPT
Depending on the position of the 353-368 loop, MGLs structures can be classified with either an
“open” or “closed” access to the active site (Figure 4b). All structures of C. freundii MGL belong to
group I (“open” conformation) with the region 353-368 freely exposed to the solvent. Y358 points
towards the solvent and has poor electron density (about 0.6σ for part of the side chain). F49* from
the neighboring subunit of the catalytic dimer is the nearest residue at a distance of about 7Å
(Figure 5a). Besides C. freundii MGL, most structures of P. putida MGL, one structure of T.
vaginalis MGL (PDB code 1PFF) and C. sporogenes MGL (chain A, PDB code 5DX5) belong to
this group. Structural alignment demonstrates that the orientation of V358 in C. freundii wild type
MGL, Y358 in C. freundii V358Y MGL, Y359 in P. putida MGL and Y360 in C. sporogenes MGL
(chain A) is similar (Figure 4c). In the structures of M. echinospora MGL (PDB codes 4U1T,
4U2H) this region is not detected in most of the subunits in the asymmetric unit, which contains
eight protomers, whereas in other subunits (PDB codes 4U1T, chains B, F), the position is
analogous to those in the structures of group I MGLs. In the “closed” conformation (group II), the
region 353-368 moves towards the active site and the opposite wall of the channel and shields from
the solvent the active site residues, such as Y113 and C115, which play key roles in the catalysis
[28, 64]. The “closed” conformation is observed in C. sporogenes MGL (chain B), in two P. putida
MGL structures (PDB codes 5X2X and 5X2Z), and in the structures of E. histolytica and T.
vaginalis MGLs (PDB codes 3ACZ and 1E5F) (Figure 5b). It is noteworthy that similarly to C.
freundii wild type MGL, the enzymes from E. histolytica and T. vaginalis exhibit valine at the
corresponding position. Overall, the structural analysis does not allow concluding that the open or
closed conformations are associated with the presence of a defined residue (i.e. a valine or a
tyrosine) and with an effect on the catalytic activity (Figure S5, Table 6). However, the single
mutation at position 358 is able to affect catalytic activity likely via a conformational
rearrangement.
28

ACCEPTED MANUSCRIPT
Figure 5. (a) Position of Y358 in the C. freundii V358Y MGL. Residues of the neighboring subunit
are marked with asterisks. (b) Superposition of the C. sporogenes MGL (chain B; PDB code 5DX5;
brown) and T. vaginalis MGL (PDB code 1E5F; green).
The V358Y substitution also changes the hydrophobicity profile of the substrate access pathway
potentially affecting interactions that occur as the substrate or product navigate through the protein
matrix. (Figure 6a-b). This, in turn, might affect the transit through the access channel, but a clear
relationship between the polarity of the active site entry channel and the kinetic parameters of
MGLs from different sources (especially KM values) remains elusive (Table 6).
29