Fungal Glycolipid Hydrolase Inhibitors and Their Effect on Cryptococcus neoformans

Article abstract of DOI:10.1002/cbic.201600538

Pathogenic fungi kill an estimated 1.3 million people each year. This number is predicted to rise as drug resistance spreads, thus antifungal drugs with novel modes of action are urgently required. Fungal endoglycoceramidase-related proteins 1 and 2 (EGCr

Full text of DOI:10.1002/cbic.201600538

Accepted Article
Title: Fungal glycolipid hydrolase inhibitors and their effect on
Cryptococcus neoformans
Authors: Andres Gonzalez Santana, Christina Tysoe, Guanggan
Hu, James Kronstad, Ethan Goddard-Borger, and Stephen
Withers
This manuscript has been accepted after peer review and appears as an
Accepted Article online prior to editing, proofing, and formal publication
of the final Version of Record (VoR). This work is currently citable by
using the Digital Object Identifier (DOI) given below. The VoR will be
published online in Early View as soon as possible and may be different
to this Accepted Article as a result of editing. Readers should obtain
the VoR from the journal website shown below when it is published
to ensure accuracy of information. The authors are responsible for the
content of this Accepted Article.
To be cited as: ChemBioChem 10.1002/cbic.201600538
Link to VoR: http://dx.doi.org/10.1002/cbic.201600538
A Journal of
www.chembiochem.org

10.1002/cbic.201600538
ChemBioChem
FULL PAPER
Fungal glycolipid hydrolase inhibitors and their effect on
Cryptococcus neoformans
Andres G. Santana[a][d], Christina Tysoe[a][b][d], Guanggan Hu[b][c], James Kronstad[b][c], Ethan D. Goddard
-Borger [a][b], Stephen G. Withers[a][b]
*
outweigh their potential benefit to the infirm[3]. When treatment
fails clinical options are limited. Antimycotics with novel modes of
action are urgently required in the clinic.
Abstract: Pathogenic fungi kill an estimated 1.3 million people each
year. This number is predicted to rise as drug resistance spreads, thus
antifungal drugs with novel modes of action are urgently required.
Fungal endoglycoceramidase-related proteins 1 and 2 (EGCrP-1 and
-2), which hydrolyse glucosylceramide and ergosteryl β-glucoside,
respectively, are important for fungal cell growth and have been
identified as potential targets for drug development. A library of
iminosugar derivatives was screened against EGCrP-1 and -2,
thereby identifying a number of competitive inhibitors with nanomolar
affinities. In addition, a mechanism-based inhibitor was shown to form
a covalent derivative with EGCrP-2. Nine of the inhibitors were
evaluated against Cryptococcus neoformans. Several showed growth
inhibitory activity, but only against a C. neoformans strain lacking the
outer fungal polysaccharide capsule, implying that penetration into the
cell is a significant handicap for these inhibitors. Pro-drug versions of
these inhibitors may address this issue.
The glycolipids glucosylceramide (GC) and steryl β-glucoside
(SBG) play important roles in cell membrane phenomena,
including lipid microdomain and vesicle formation, in both fungi
and metazoans[4]. Perturbation of the biosynthesis and catabolism
of these glycolipids in fungi can have pronounced effects on their
growth[5]. Selective targeting of these pathways in fungi over the
host represents a promising anti-fungal therapeutic strategy and
might be realised by targeting a family of fungal glucosidases
dedicated to processing these glycolipids.
Introduction
Opportunistic fungal pathogens like Cryptococcus neoformans kill
an estimated 1.3 million people per year, a mortality rate
comparable in magnitude to that of tuberculosis and malaria.
Fungal
infections
are
especially
problematic
for
immunocompromised patients because they can be difficult to
diagnose, leading to late interventions, and are frequently
recalcitrant to existing therapies[1]. Current antifungal agents hail
from just a handful of drug classes that target fungal DNA
synthesis, membrane and cell wall assembly, or ergosterol
production[2]. These drugs face a number of issues, including
narrow species specificity, incompatibilities with other drugs
(especially HIV and cancer therapeutics), and toxicities that can
Scheme 1. (A) GC biosynthesis in the ER of C. neoformans involves a quality
control mechanism whereby EGCrP-1 degrades immature GCs. (B) EGCrP-2
degrades ergosteryl β-glucoside in the vacuole of C. neoformans.
[a]
Department of Chemistry, University of British Columbia, 2036 Main
Mall, Vancouver, B.C., CANADA V6T 1Z1
604-822-3402
withers@chem.ubc.ca
Fungal GC, which is important for pathogenicity and virulence, is
quite different from its metazoan counterpart, possessing an extra
unsaturation and a branching methyl group on the sphingosine
backbone as well as alpha-hydroxylation on the N-acyl group
[b]
[c]
Michael Smith Laboratories, University of British Columbia, 2185
East Mall, Vancouver, B.C., CANADA V6T 1Z4
Department of Microbiology and Immunology, University of British
Columbia, Life Sciences Centre, 2350 Health Sciences Mall,
Vancouver, B.C., CANADA V6T 1Z3
[d]
These authors contributed equally to this work.
For internal use, please do not delete. Submitted_Manuscript
This article is protected by copyright. All rights reserved.

10.1002/cbic.201600538
ChemBioChem
FULL PAPER
(Scheme 1). These unique features of fungal GC are important for
its function, so much so that fungi have evolved a quality-control
enzyme, EGCrP-1, that degrades ‘immature’ GC in the ER[6].
EGCrP-1 is a GH5 β-glucosidase that hydrolyzes the glycosidic
bond in GC. Disruption of egcrp1 in C. neoformans results in
disregulation of fungal GC biosynthesis and an impaired capsule
formation phenotype, implicating EGCrP-1 as a potential drug
target[6]. EGCrP-2 is a vacuolar homolog of EGCrP-1 with a
broader substrate specificity profile, displaying the ability to
hydrolyze both fungal GC and SBG. Disruption of egcrp2 in C.
neoformans led to growth arrest, dysfunctional cell budding and
abnormal vacuole morphology, making it an even more promising
drug target than EGCrP-1[7].
Scheme 2. The library of iminosugar-based competitive β-glucosidase
inhibitors. EGCrP-1 was screened against compounds 1 to 29. The subset
shown within the boxes were screened against EGCrP-2.
Because EGCrP-1 and -2 share sequence homology and a
common activity on fungal GC, we reasoned that they could be
simultaneously targeted using selective β-glucosidase inhibitors.
Their combined inhibition might then lead to the growth arrest or
death of fungal pathogens. In order to explore this hypothesis, we
conducted a screen against a collection of iminosugar-based
competitive β-glucosidase inhibitors (Scheme 2) that our
laboratory had previously developed as inhibitors of human
lysosomal glucocerebrosidase (GCase), an enzyme that has
similar activity to EGCrP-1 and -2 on GC[8]. Azasugars have
proved to be potent inhibitors of their cognate glycosidases in a
number of cases, with the protonated amine mimicking the
oxocarbenium character derived at the reaction transition state[9].
We were not particularly concerned by off-target inhibition of host
GCase, since Gaucher disease patients with greatly diminished
GCase activity can survive in health for many years, suggesting
that the acute inhibition of host GCase when targeting a fungal
infection is not a great concern.
Results and Discussion
The enzymes used in our screen were expressed in Escherichia
coli. Heterologous expression of C. neoformans and A. fumigatus
EGCrP-1 alone or with an N-terminal fusion partner (MBP, GST,
NusA) was unsuccessful. Fortunately, EGCrP-1 from Rhizopus
oryzae, which has high sequence similarity to the C. neoformans
enzyme, expressed well in E. coli with a simple C-terminal
hexahistidine tag[6]. EGCrP-2 of C. neoformans expressed well in
E. coli using a similar approach.
Our recombinant EGCrP-1 possessed no activity on the
commercial chromogenic β-glucosidase substrates 4-nitrophenyl
β-D-glucopyranoside (PNP-Glc) and 7-methylumbelliferyl β-D-
glucopyranoside (MU-Glc), an issue that was overcome by
employing the more reactive chromogenic substrate 2,4-
dinitrophenyl β-D-glucopyranoside (DNP-Glc), which has
a
leaving group of pKa 4 versus those of pKa 7 - 8 in the other
substrates. Under optimized conditions, the Michaelis-Menten
parameters for cleavage of DNP-Glc by EGCrP-1 were
determined to be: kcat = 0.16±0.01 s-1; Km = 2.8±0.3 mM and
k
cat/Km = 57±5 M-1s-1. The catalytic efficiency of EGCrP-1 for
cleavage of DNP-Glc is approximately 100-fold lower than that
with the native substrate, GC (kcat = (38.3±0.2)×10-3 s-1; Km
=
5.8±0.3 uM; kcat/Km = (6.6±0.4)×103 M-1s-1)[6]. This level of activity
was sufficient for our screen, however a high concentration of 2of
spontaneous hydrolysis of this highly activated substrate. Despite
For internal use, please do not delete. Submitted_Manuscript
This article is protected by copyright. All rights reserved.

10.1002/cbic.201600538
ChemBioChem
FULL PAPER
having a significantly higher affinity for GC over DNP-Glc, the rate
constant of EGCrP-1 for cleavage of GC is considerably lower.
These rate constants are meagre in comparison to the kcat values
GC substrate has a longer carbon chain. However, this should be
approached with caution as iminosugars with long lipophilic
chains tend to be cytotoxic, possibly due to their surfactant-like
nature and ability to perturb membranes. Nonetheless, the
scaffold of inhibitor 17 has promise, and the lipophilic appendage
could be further optimised to enhance potency.[8]
of other β-glucosidases, such as Agrobacterium β-glucosidase[10]
,
which cleaves DNP-Glc with a kcat of 88 s-1. This is commensurate
with EGCrP-1’s role as a quality-control enzyme: a slow turnover
of GC is required to prevent the depletion of mature GC while still
allowing the hydrolysis of immature forms. On the other hand,
EGCrP-2 showed appreciable catalytic activity with both PNP-Glc
and DNP-Glc substrates, and hydrolysed DNP-Glc ten-fold faster
than did EGCrP-1, with a Km of 0.46±0.1 mM, kcat of 29±2 s-1, and
In an extension to this study, we sought to identify mechanism-
based covalent inhibitors of the enzymes, based upon a class of
previously identified potent inhibitors of human GCase10. In these
studies
a class of 2-deoxy-2-fluoroglucosides bearing a
k
cat/Km = (63±9)×103 M-1s-1.
phosphonoester leaving group was described. The fluorine at C2
destabilises the oxocarbenium ion-like transition states for
formation and hydrolysis of the covalent glycosyl-enzyme
intermediate, thus slowing both glycosylation and deglycosylation
of the enzyme, while the highly activated leaving group ensures
not only that the first glycosylation step is faster than
deglycosylation, but also that selectivity for this particular enzyme
can be optimized by mimicking, to some extent, the lipophilic
ceramide moiety. Inactivation studies were carried out with
EGCrP-2 and the most effective of these reagents so far,
compound 30 (Figure 2) proved to be effective in this case too,
with inactivation parameters of ki = 1.11 min-1; Ki = 11.8 µM; ki/Ki
= 94 mM-1min-1. In fact this inactivator reacted three times more
rapidly with EGCrP-2 than it did with human GCase (ki/Ki = 29
mM-1min-1)[11]. Interestingly, inactivation of EGCrP-1 was less
effective, and did not proceed to completion.
The library of β-glucosidase inhibitors was screened against
EGCrP-1 (Figure 1). The lipophilic iminosugars 10 and 17
provided > 80% inhibition at a concentration of 1 µM and were
selected for further analysis. The Ki values of inhibitors 10 and 17
were thereby determined to be 130 nM and 170 nM, respectively.
The screen against EGCrP-2 provided similar hits (Figure 1),
despite a number of compounds being omitted from the second
screen. The inhibition profile of EGCrP-2 closely mirrors that of
EGCrP-1, with inhibitors 16, 23 and 24 providing more than a 60%
reduction of the enzymatic rate and inhibitor 17 showing more
than 80% inhibition. Kinetic analysis of the best of these revealed
the Ki of inhibitor 17 to be 21 nM when tested against EGCrP-2,
eight times more potent than against EGCrP-1.
Figure 1.Percentage inhibition of EGCrP-1 and -2 at inhibitor concentrations of
1.0 µM.
Figure 2. Mechanism-based inactivation of EGCrP-2 by compound 30. EGCrP-
1 was not inactivated by this compound.
For both enzymes there was a clear trend in their inhibitor
sensitivity, with inhibitors becoming more effective with increasing
lipophilicity of the iminosugar substituent. One could likely
improve the potency of these inhibitors even further by extending
the length of the hydrocarbon chain substituent, since the native
Encouraged by these results, inhibitors 10, 16, 17, 20, 21, 22, 23,
24, and 30 were tested in vivo against C. neoformans. Cap59, a
mutant strain missing the outer fungal capsule, was used in
addition to the wild-type C. neoformans. A growth response assay
For internal use, please do not delete. Submitted_Manuscript
This article is protected by copyright. All rights reserved.

10.1002/cbic.201600538
ChemBioChem
FULL PAPER
was performed with inhibitors 10 to 24, incubating the fungi in the
presence of 100 µM inhibitor for 72 hours (Figure 3). Compounds
10, 16 and 17, the iminosugars with long lipophilic side chains,
had noticeable impact on growth in the cap59 strain with a 40–
50% decrease in growth over the course of the assay.
Compounds 20 and 21, possessing shorter side chain
appendages, also exhibited some growth inhibition on the cap59
strain, but to a lesser extent, while 22, 23 and 24 showed no
inhibition. None of the inhibitors afforded a significant decrease in
the growth rate of the wild-type strain.
Figure 4. The sensitivities of wild-type (KN99a) and acapsular (cap59) C.
neoformans strains to EGCrP-1/2 inhibitors 10 and 17. Cultures were incubated
with inhibitors for 48 h at 30°C and optical density measured and normalized
relative to untreated cultures.
Conclusions
This work sought to evaluate EGCrP-1 and -2 as potential targets
for the development of novel antimycotics. Screening against our
library of β-glucosidase targeted iminosugar inhibitors allowed us
to identify inhibitors of these enzymes with nanomolar-range
inhibition constants. While inhibition of EGCrP-1 and -2 with these
chemical tools does appear to slow C. neoformans growth, as
seen in the activity towards the cap59 strain, it did not completely
arrest fungal growth, even at high compound concentrations.
Unfortunately, these compounds were inactive on wild-type C.
neoformans, suggesting that they may be unable to penetrate the
capsule to achieve a sufficient intracellular concentration to inhibit
EGCrP-1 and -2. Nonetheless, the bioavailability of these
inhibitors may yet be enhanced using traditional ‘pro-drugging’
strategies. These would serve as very useful chemical probes for
studying fungal glyocolipid biology.
Figure 3. The growth curves of wild-type (KN99a) and acapsular (cap59) C.
neoformans strains over 72 h in YPD media supplemented with EGCrP1/2
inhibitors (100 µM) at 30°C in the presence of compounds (A) 10, (B) 16, (C)
17, (D) 20, (E) 21, (F) 22, (G) 23 and (H) 24.
A dose response assay was conducted with inhibitors 10, 17 and
30. The results mirrored those seen in the growth response
curves: while there was a significant impact on the cap59 strain,
the wild-type strain was largely unaffected even at higher inhibitor
concentrations (Figure 4). Surprisingly, the mechanism-based
inhibitor 30 showed no activity on the cap59 or wild-type strains.
The lack of significant inhibitory activity in the wild-type strain
suggests that the fungal capsule impedes up-take of these
compounds.
Experimental Section
Enzyme expression. Codon-harmonized genes encoding R. oryzae
EGCrP-1 and C. neoformans EGCrP-2 were purchased from BioBasic Inc.
These genes was ligated into a pET29b expression vector, which was then
transformed into electrocompetent BL21*(DE3) E. coli. Expression was
carried out in LBE-5052 auto-induction media supplemented with 50 μg ml-
1 kanamycin at 25 °C for 24 hours. Cells were harvested by centrifugation
(17,000×g, 20 min, 4 °C) and resuspended in binding buffer (50 mM NaPi,
500 mM NaCl, 5 mM imidazole, pH 7.5) containing EDTA-free protease
inhibitor cocktail and lysozyme (0.1 mg ml-1) by incubating at 4 °C for 30
min. Benzonase (1 μl) was added to the mixture then lysis was effected by
sonication. The lysate was centrifuged (17,000×g, 20 min, 4 °C) and the
supernatant collected. The supernatant was filtered (0.45 μm) and then
subjected to immobilized metal-ion affinity chromatography (elution buffer:
50 mM NaPi, 500 mM NaCl, 500 mM imidazole, pH 7.5). The protein
obtained was estimated to be >95% pure by Coomassie-stained SDS-
PAGE. Protein concentration was determined by Bradford assay.
Recombinant expression of soluble hexahistidine-tagged RoEGCrP-1 was
achieved in high yield of >50 mg l-1. Recombinant expression of soluble
hexahistidine-tagged EGCrP-2 was achieved in an average yield of 20 mg
l-1.
For internal use, please do not delete. Submitted_Manuscript
This article is protected by copyright. All rights reserved.

10.1002/cbic.201600538
ChemBioChem
FULL PAPER
Keywords: Antimycotics • glycosidase inhibitors •
glucosylceramide • iminosugar • Cryptococcus neoformans
Chemical synthesis. Full details of the synthesis and characterisation of
the library of azasugars were published previously 10. Benzyl benzyl-(2-
deoxy-2-fluoro-β-D-glucopyranosyl) phosphonate (30) was synthesized
according to protocols previously reported.[11] Briefly, acetylated 2-deoxy-
2-fluoroglucopyranosyl bromide was coupled to the corresponding benzyl
phosphonic acid under Koenigs-Knorr conditions promoted by silver
carbonate. The resulting product was subsequently de-O-acetylated with
catalytic sodium methoxide to yield compound 30 as a 1:1 mixture of
epimers at the phosphorus centre.
Inhibitor screen. A Michaelis-Menten plot was constructed and kcat and
Km values were determined for the hydrolysis of 2,4-dinitrophenyl
glucoside by EGCrP-1 and -2. Reaction rates were measured by
monitoring the net change in absorbance at 400 nm. The inhibitor screen
was carried out in a Synergy HI plate reader. For analysis of RoEGCrP-1,
100 nM of enzyme, 3 mM DNPGlc (Km), and 1 µM of inhibitor were used
in each reaction. The screen was run in duplicate and results were verified
on a Cary 300 spectrophotometer. In the screen against CnEGCrP-2, 100
nM of enzyme, 0.4 mM DNPGlc (Km), and 1 µM of inhibitor were used in
each reaction. The screen was run in duplicate. Compounds that resulted
in >80% inhibition were selected for Ki determination. Ki determinations
were performed using two to four different substrate concentrations. For
each inhibitor, a range of four to six inhibitor concentrations bracketing the
Ki value was used for each substrate concentration. Dixon plots (1/v vs [I])
were constructed to validate the use of a competitive inhibition model and
to assess the fit of the rest of the data. A competitive inhibition model was
fit to the data by using non-linear regression analysis with Grafit 7.0.
Time-dependent inhibition. The kinetic parameters for the inactivation of
EGCrP-2 were determined as follows: samples of the enzyme were
incubated in buffer in the presence of a range of concentrations of the
inhibitor. Aliquots (20 μl) of these inactivation mixtures were removed at
time intervals and diluted into assay cells containing a larger volume (180
μl) of substrate (at saturating concentrations). This effectively stops the
inactivation both by dilution of the inactivator and by competition with an
excess of substrate. The residual enzymatic activity was determined from
the rate of hydrolysis of the substrate, which is directly proportional to the
amount of active enzyme. The process was monitored until the enzyme
was 80–90% inactivated. Pseudo-first order rate constants (kobs) for each
inactivator concentration were calculated by fitting plots of the residual
activity versus time to a single exponential equation using GraFit 7.0.
These values of kobs were then fit to the equation below describing the
inactivation process to obtain values of ki and Ki:
kobs = ki[I] / ( Ki + [I])
C. neoformans growth response assay. Two C. neoformans strains,
wild-type (KN99a) and the acapsular mutant strain cap59, were assessed
for their sensitivity to the EGCrP-1 and -2 inhibitors. To perform the assays,
the fungal cells were seeded in micro-titre wells in YNB media at 1×106
cells/ml (~0.08 optical density at 600 nm [OD600]) in triplicate with or without
100 µM inhibitor and incubated at 30°C. For each strain, the OD600 was
measured every 2 h after initiation of the experiments for 72 h.
C. neoformans dose response assay. The dose response of two C.
neoformans strains, wild-type (KN99a) and the acapsular mutant strain
cap59, to compounds 10 and 17 was assessed. To perform the assays,
the fungal cells were seeded in micro-titre wells in YNB media at 1×106
cells/ml (OD600 ≈ 0.08) in triplicate with or without inhibitor and incubated
for 48 h at 30°C. For each concentration of the drug and for each strain,
the increase in OD600 for treated samples was normalized against the
untreated control to determine the percentage growth inhibition.
Acknowledgements
We thank the Canadian Institutes for Health Research (CIHR) and
the European Molecular Biology Organisation (EMBO) for
funding.
For internal use, please do not delete. Submitted_Manuscript
This article is protected by copyright. All rights reserved.

10.1002/cbic.201600538
ChemBioChem
FULL PAPER
[1]
B. J. Park, K. A. Wannemuehler, B. J. Marston, N. Govender, P. G.
[7]
[8]
[9]
T. Watanabe, T. Ito, H. M. Goda, Y. Ishibashi, T. Miyamoto, K. Ikeda, R.
Taguchi, N. Okino, M. Ito, J Biol Chem 2015, 290(2), 1005-1019.
E. D. Goddard-Borger, M. B. Tropak, S. Yonekawa, C. Tysoe, D. J.
Mahuran, S. G. Withers, J Med Chem 2012, 55(6), 2737-2745.
a) G. Horne, F. X. Wilson, J. Tinsley, D. H. Williams, R. Storer, Drug
Discov Today 2011, 16(3-4), 107-118; b) R. Lahiri, A. A. Ansari, Y. D.
Vankar, Chem Soc Rev 2013, 42(12), 5102-5118; c) A. E. Stütz, T. M.
Wrodnigg, Adv Carbohydr Chem Biochem 2011, 66, 187-298; d) E. M.
Sánchez-Fernández, J. M. García Fernández, C. O. Mellet, Chem
Commun (Camb) 2016, 52(32), 5497-5515.
Pappas, T. M. Chiller, AIDS 2009, 23(4), 525-530.
[2]
a) M. T. Pasko, S. C. Piscitelli, A. D. Van Slooten, DICP 1990, 24(9),
860-867; b) M. Baginski, J. Czub, Curr Drug Metab 2009, 10(5), 459-
469; c) A. Vermes, H. J. Guchelaar, J. Dankert, J Antimicrob
Chemother 2000, 46(2), 171-179.
[3]
[4]
a) M. Cuenca-Estrella, Rev Esp Quimioter 2010, 23(4), 169-176; b) C.
A. Kauffman, Proc Am Thorac Soc 2006, 3(1), 35-40; c) H. Vanden
Bossche, Rev Iberoam Micol 1997, 14(2), 44-49.
a) M. R. Pinto, M. L. Rodrigues, L. R. Travassos, R. M. Haido, R. Wait,
E. Barreto-Bergter, Glycobiology 2002, 12(4), 251-260; b) G. van Meer,
J. Wolthoorn, S. Degroote, Philos Trans R Soc Lond B Biol Sci 2003,
358(1433), 869-873.
[10] J. B. Kempton, S. G. Withers, Biochemistry 1992, 31(41), 9961-9969.
[11] B. P. Rempel, M. B. Tropak, D. J. Mahuran, S. G. Withers, Angew
Chem Int Ed Engl 2011, 50(44), 10381-10383.
[5]
[6]
S. M. Noble, S. French, L. A. Kohn, V. Chen, A. D. Johnson, Nat Genet
2010, 42(7), 590-598.
Y. Ishibashi, K. Ikeda, K. Sakaguchi, N. Okino, R. Taguchi, M. Ito, J Biol
Chem 2012, 287(1), 368-381.
For internal use, please do not delete. Submitted_Manuscript
This article is protected by copyright. All rights reserved.

10.1002/cbic.201600538
ChemBioChem
FULL PAPER
Entry for the Table of Contents (Please choose one layout)
FULL PAPER
A. G. Santana, C. Tysoe, G. Hu, J.
Kronstad, E. D. Goddard -Borger, S. G.
Withers*
Page No. – Page No.
A library of iminosugar derivatives was tested against fungal enzymes
EGCrP-1 and 2 allowing for the identification of nanomolar-range
inhibitors. Several inhibitors were tested against Cryptococcus
neoformans and showed growth inhibitory activity towards the
acapsular strain.
Fungal glycolipid hydrolase inhibitors
and their effect on Cryptococcus
neoformans.
For internal use, please do not delete. Submitted_Manuscript
This article is protected by copyright. All rights reserved.