Обнаружение сфингосин-1-Фосфатных модуляторов-рецепторов в качестве потенциальных CHI3L1 Inhibitors с помощью лигандского виртуального скрининга и моделирования молекулярной динамики.

Introduction

Multiple sclerosis (MS) is an autoimmune
disorder affecting the
central nervous system (CNS), causing chronic inflammation, demyelination,
and neuronal degeneration. Diagnosis of MS relies on identifying characteristic
lesions in at least two distinct CNS areas. Recent studies have shifted focus toward atrophy and cortical lesions
over lesion quantity and relapse rates for predicting disease progression
and cognitive impairment in MS. However, current assessment methods,
mostly subjective, often lack precision, highlighting the need for
personalized treatment strategies based on individual prognosis and
risk assessment. Biomarkers play a crucial
role in predicting disability progression, monitoring disease activity,
and assessing treatment response in MS. Biomarkers are measurable
characteristics indicating normal biological processes, disease development,
or response to treatment. In MS, biomarkers aid in early diagnosis,
improving patient care, and disease management. Among various cerebrospinal fluid (CSF)/serum biomarkers,
Chitinase 3-like 1 (CHI3L1), also known as YKL-40, has recently emerged
as a potential biomarker for MS. The expression
of CHI3L1 has been found to be higher in the CSF of individuals with
MS.

CHI3L1, a glycoprotein expressed
in various tissues and cells such
as macrophages, neutrophils, chondrocytes, and endothelial cells,
plays a role in inflammation, tissue injury, repair, and remodeling
across different tissues. Studies have
shown that elevated levels of CHI3L1 in CSF are associated with a
greater likelihood of conversion from clinically isolated syndrome
(CIS-MS) to clinically definite MS. Additionally,
higher levels of CHI3L1 in CSF are correlated with increased disability
in MS patients. Also, individuals with
higher concentrations of CHI3L1 in CSF have a significantly heightened
risk of relapse, increased MRI activity, and evidence of disease activity.

In humans, CHI3L1 has a molecular weight
of 40 kDa and is encoded
by the CHI3L1 gene. Belonging to the chitolectin
class of glycosyl hydrolase family 18, this protein bears significant
homology to chitotriosidase (CHIT1) and acidic mammalian Chitinase. However, a mutation at a crucial residue Glu140
to Leu140 renders it enzymatically inactive, and as a results, CHI3L1
lacks chitin hydrolase activity. Despite
this, they maintain the ability to bind to chitin oligomers. While
CHI3L1 has emerged as a target in various clinical studies, there
is a limited number of documented inhibitors with confirmed affinity
for CHI3L1. Only a couple of small molecules have been discovered
and used successfully as CHI3L1 inhibitors in different preclinical
models. CHI3L1 binders can inhibit the protein through various mechanisms
of action. They may interfere with CHI3L1-receptor interactions, such
as those with IL-13Rα2, either by disrupting glycosaminoglycan-mediated
binding or through direct interaction with the receptor. Another mode
of action involves small-molecule inhibitors that block the release
of cytokines by CHI3L1 from the extracellular matrix.

In the
recent last years, structure-based drug discovery has led
to the identification of some inhibitors targeting CHI3L1 ( Figure
). In 2021, Lee et
al. discovered that the binding of K284 to the chitin-binding domain
(CBD) of CHI3L1 inhibited its interaction with the interleukin-13
receptor subunit alpha-2 (IL-13Rα2), effectively suppressing
the signal mediated by CHI3L1. Subsequently,
in 2022, another inhibitor of CHI3L1, G721-0282, demonstrated antianxiety
effects by blocking CHI3L1-mediated neuroinflammation. More recently, and during the preparation of
this manuscript, two compounds (including compound 30) exhibiting
nanomolar affinity for CHI3L1 have been identified, and the structure
of CHI3L1 bound to these compounds has been successfully determined (inhibitor 30). As of now, there has not been any clinically established drug known
for its inhibitory effect against CHI3L1 that could be used in patients,
including those with conditions like MS, as an effective therapeutic
strategy.

Chemical structures of known CHI3L1 inhibitors.

Methods

Ligand-Based Virtual Screening

The FDA-approved drugs
have been screened to discover new compounds capable of inhibiting
CHI3L1. Four commonly used 2D similarity metrics, namely, Buser, Cosine,
Kulczynski, and Tanimoto, have been employed to prioritize potential
hits. This prioritization was based on
the linear fingerprint method implemented within Schrödinger’s
similarity-based virtual screening protocol. MolPrint2D fingerprints, which encode
atom environment descriptors based on molecular connectivity tables,
were utilized as 2D similarity descriptors. Two CHI3L1 selective inhibitors,
G721-0282 and K284 were chosen as reference structures for the 2D similarity
search.

Docking of the Selected Compounds

The crystal structure
of CHI3L1 bound to inhibitor 30 PDB ID: 8R4X
was used
for docking the selected compounds. This structure was selected due
to being the only X-ray structure available bound to an inhibitor
and featuring a high resolution of 1.54 Å. The molecular docking
calculations and preparation of both protein and ligands were conducted
by using the Schrödinger software package. The protein was
prepared and minimized using the Protein Preparation Wizard, while the chemical structure of selected compounds
was optimized by LigPrep. The compound’s
binding modes were determined through GLIDE docking of each compound
onto the minimized X-ray structure. The grid box was positioned at
the inhibitor binding site, utilizing default parameters for receptor
grid generation. The ligands were docked
using GLIDE extra precision (XP) and Induced-fit docking (IFD).

Molecular Dynamics Simulations

Extended molecular dynamics
(MD) simulations were employed to validate the stability of the docking
poses and investigate the structural and dynamic features of the simulated
systems. To achieve this, the CHI3L1 complexes were constructed using
the protein complexed with fingolimod, siponimod, ponesimod, G721-0282,
and K284 selected based on the most favorable docking scores. Additionally,
to facilitate comparison, the CHI3L1 complexed with compound 30 was also simulated to assess its binding stability.

Simulations were conducted using the AMBER20 package and Amber ff14SB force field for the protein, while the ligands (fingolimod, inhibitor
30, G721-0282, K284 siponimod, and ponesimod) were parametrized with
the general amber force field (GAFF) and
restrained electrostatic potential (RESP) partial atomic charges derived from B3LYP/6-31G­(d) calculations. The
ionizable residues were assigned their standard protonation states
at physiological pH, and ACE and NME capping groups were added to
the protein’s N- and C-termini. The systems were embedded in
an octahedral box filled with TIP3P water molecules, and counterions were included to ensure system neutrality.
The final setup consisted of the CHI3L1 protein, the ligand, approximately
11,000–12,000 water molecules, and a variable number of Cl
ions, resulting in simulated systems containing around 40,000 atoms.
Simulations were carried out in the NPT ensemble for equilibration
and in the NVT ensemble for molecular dynamics (MD) production, utilizing
periodic boundary conditions and Ewald sums (with a grid spacing of
1 Å) to handle long-range electrostatic interactions. All complexes
underwent duplicate simulations.

Initially, the systems were
minimized by refining the positions
of hydrogen atoms in the protein (2000 cycles of steepest descent
followed by 8000 cycles of conjugate gradient), then minimizing the
water molecules (using the same approach), and finally minimizing
the entire system (4000 steepest descent cycles and 1000 conjugate
gradient cycles). Subsequently, the system temperature was gradually
increased from 100 to 300 K over five steps, each lasting 50 ps, using
the NVT ensemble. During this phase, restraints (8 kcal mol –1 Å –2 ) were applied to keep the ligand within
the binding pocket and prevent artificial rearrangements during equilibration.
,
An additional 5 ns step in the NPT ensemble was performed to equilibrate
the system’s density, with the restraints being gradually released
in the following steps. Production MD
simulations were conducted for 500 ns per replica, totaling 5 μs
of simulation time for the ligand-bound CHI3L1 complexes.

Essential Dynamics (ED)

This approach was employed
to identify the key motions associated with structural variations
observed during the MD simulations. ED analysis involves examining
and visualizing movements along individual modes separately, facilitating
the characterization of the dominant collective motions in the simulations.
The method relies on constructing and diagonalizing a positional covariance
matrix to extract the collective deformation modes (eigenvectors),
with the eigenvalues reflecting the relative contribution of each
motion to the overall structural variability of the protein. In this
study, ED analysis was applied to 25,000 snapshots obtained from 500
ns of simulation for each system, focusing exclusively on backbone
atoms. The computations were carried out using the PCAsuite software
(available at http://www.mmb.irbbarcelona.org/software/pcasuite/pcasuite.html)
and integrated into the pyPCcazip tool suite.

MM/GBSA Calculations

The MM/GBSA method designed to
evaluate the free energy difference between two states, typically
representing the bound and unbound forms of solvated molecules. It can also be applied to compare the free energies
of different solvated conformations of a single molecule. In this
study, we employed MM/GBSA scripts integrated within Amber and AmberTools
to automate the binding free energy calculations for the protein–ligand
complex.

Results

Ligand-Based Virtual Screening

In the present study,
we performed ligand-based virtual screening (LBVS) to investigate
the potential of repurposing Food and Drug Administration (FDA)-approved
drugs as inhibitors against CHI3L1. Employing the known CHI3L1 inhibitors,
K284 and G721-0282, as scaffolds, our objective is to identify novel
compounds capable of inhibiting CHI3L1 that can be repurposed in the
clinical setting, thereby offering promising therapeutic approaches
for inflammatory diseases like MS.

The main goal of the ligand-based
virtual screening (LBVS) was to identify potential novel inhibitors
of CHI3L1 by screening FDA-approved drugs chemical library through
computational methodologies. LBVS relies
on the concept that compounds with similar structures tend to exhibit
similar biological activities. The main ligand-based methodologies
involve the use of pharmacophores (abstract representations of the
essential features required for a molecule to exhibit activity), shape-based
similarity, fingerprint similarity, and employing machine learning
techniques using molecular properties and data derived from the aforementioned
methods. In this context, aiming to repurpose
FDA-approved drugs, a virtual screening of this library has been conducted
to identify novel compounds potentially active against CHI3L1 using
2D similarity metrics, namely, Buser, Cosine, Kulczynski, and Tanimoto.
They have been employed to prioritize potential hits. MolPrint2D fingerprints were
utilized as 2D similarity descriptors. Two CHI3L1 selective inhibitors,
G721-0282 and K284, ( Figure
) were chosen as reference structures for the 2D similarity search.

Using fingerprint similarity analysis, a selection was made, identifying
a list of approved drugs with similarity indexes exceeding 0.85 in
relation to both K284 and G721-0282 ( Tables S1 and S2 in the Supporting Information). This list encompasses
various drug families, including kinase inhibitors like cabozantinib,
alpha-1 adrenergic antagonists such as terazosin and dixazosin, and
anti-inflammatory medications such as nabumetone and phenylbutazone.
Notably, fingolimod, a chemically diverse compound with an inhibitory
effect against sphingosine-1-phosphate receptor (S1P), which has been
approved for the treatment of MS, is also present among these drugs. Its similarity index and chemical structure
are reported in Table
. Fingolimod was selected due to its fingerprint similarity to that
of G721-0282. Interestingly, a recent paper describes the reduction
of CHI3L1 levels in CSF of MS patients treated with fingolimod, postulating
that this protein is an efficacy biomarker related with the inflammation
decrease produced by the drug treatment. These findings are aligned with the outcomes from the LBVS conducted
in this study.

1 2D Similarity Index (Buser Matrix)
of Fingolimod Relative to G721-0282, alongside Its Chemical Structure

2D Similarity Index (Buser Matrix)
of Fingolimod Relative to G721-0282, alongside Its Chemical Structure

Molecular Docking and Dynamics Simulations

Docking
studies were conducted to thoroughly understand the binding mode and
inhibition mechanism of fingolimod and the reference compounds G721-0282
and K284. The crystal structure of CHI3L1 (PDB 8R4X ) served as the basis
for docking the selected compound. The most favorable conformation
revealed that fingolimod binds to the pocket similarly to the known
inhibitor described in the X-ray structure (inhibitor 30), with a
docking score of −10.349, slightly higher in comparison to
−9.764 for the inhibitor 30 ( Figure
, orange stick). In terms of inhibitor 30,
it establishes hydrogen bonds with Tyr206 and Asp207, along with hydrophobic
interactions with Trp352, Thr293, and Phe261.

Fingolimod’s binding
mode as identified from docking studies
and MD simulations (rendered in violet sticks) and inhibitor 30 from
the X-ray structure PDB ID 8R4X (rendered in orange sticks) within the cavity (illustrated
as a seafoam green cartoon). Detailed close-up of the ligand binding
pocket highlights essential residues within a 5 Å proximity of
the ligand. The yellow dashed lines denote hydrogen bonds, while the
pink dashed lines indicate salt bridges, the blue dashed lines represent
π–π stacking, and green dashed lines show π–cation
interactions. Two-dimensional diagrams of protein-fingolimod interactions
are presented in Figure S2A in the Supporting
Information.

Fingolimod is deeply accommodated within the hydrophobic
groove,
with binding interactions involving hydrogen bonds to Glu290 and Thr293,
along with multiple π–π stacking interactions with
Trp99, Phe261, and Trp352 ( Figure
, violet stick). The reference compound, G721-0282,
is deeply embedded within the hydrophobic groove, where it interacts
through π–π stacking with Trp99 and Trp352 and
forms a hydrogen bond interaction with N100 ( Figure S1A in the Supporting Information). These interactions are
thought to be essential for the inhibition of CHI3L1, as previous
studies indicate that the compound effectively disrupts the movement
of the flexible Trp99 residue. In addition,
K284 also engages the hydrophobic groove through π–π-stacking
interactions with Trp99 and Trp352, reinforcing its stable positioning
within the binding site. Beyond these aromatic contacts, K284 forms
several hydrogen bonds with key polar and charged residues including
Arg35, Arg293, and Glu290, all of which are located within the CBD
of CHI3L1 ( Figure S1B in the Supporting
Information).

Fingolimod is approved for the treatment of MS
based on its mechanism
of action: modulation of the sphingosine-1-phosphate receptor (S1P).
Additionally, two more drugs, namely, siponimod
,
and ponesimod ( Figure
), chemically diverse to fingolimod but with
documented efficiency against S1P, have been also recently approved
for MS clinical treatment.

Chemical structures of the known S1P inhibitors.

Here, we selected these two compounds to assess
their ability to
bind to the hydrophobic pocket of CHI3L1. The research aims to investigate
if they are CHI3L1 binders and if they interact with key protein residues,
as observed in the binding of inhibitor 30 and fingolimod.

Ponesimod
exhibits docking scores of −11.076 and binds to
the CHI3L1 binding site similar to fingolimod and inhibitor 30. It
forms hydrogen bonds with Arg35 and Thr293 and interacts with the
backbone nitrogen atom of Tyr99, alongside a π–π
stacking interaction with the same residue, resembling fingolimod’s
behavior ( Figure
,
shown as gray stick).

Superposition of fingolimod and ponesimod’s binding
modes
(rendered in violet and gray, respectively) as identified from MD
simulations within the cavity (illustrated as a seafoam green cartoon).
Detailed close-up of the ponesimod’s binding pocket, highlighting
essential residues within a 5 Å proximity of the ligands. The
yellow dashed lines denote hydrogen bonds, while the pink dashed lines
indicate salt bridges, the blue dashed lines represent π–π
stacking, and green dashed lines show π–cation interactions.
Two-dimensional diagrams of protein–ponesimod interactions
are presented in Figure S2B in the Supporting
Information.

Compared with the S1P inhibitors mentioned earlier,
siponimod displays
a distinct binding pattern with a docking score of −12.941.
It establishes salt bridge and hydrogen bond interactions with Asp207.
Additionally, it exhibits π–π stacking interactions
with Trp31, as well as a cation−π interaction with Tyr352
( Figure
, depicted
in salmon stick). Notably, siponimod’s binding mode is distinguished
by forming both the salt bridge and hydrogen bond interactions with
the same residue as inhibitor 30, a feature solely observed in siponimod’s
case.

Siponimod’s binding mode as identified from MD simulations
(rendered in salmon stick) within the cavity (illustrated as a seafoam
green cartoon). Detailed close-up of the ligand binding pocket, highlighting
essential residues within a 5 Å proximity of the ligand. The
pink dashed line indicates salt bridge, the blue dashed lines represent
π–π stacking, and the green dashed line shows π–cation
interactions. Two-dimensional diagrams of protein-siponimod interactions
are presented in Figure S2C in the Supporting
Information.

Structural Analysis of CHI3L1 Complexes

CHI3L1 inhibition
can be achieved by binding inhibitors to its oligomer binding pocket.
Small molecules that fit into the hydrophobic groove may either directly
block the enzyme or influence, through allosteric mechanisms, the
interaction between the CHI3L1 complex and other proteins such as
IL-13Rα2. To observe the protein’s dynamic response to
inhibitor binding, we performed multiple molecular dynamics (MD) simulations,
analyzing trajectory files by measuring the root-mean-square deviation
(RMSD) and root-mean-square fluctuation (RMSF) of the protein backbone
throughout the simulations. The RMSD analysis of CHI3L1 protein complexes
with different compounds over a 500 ns MD simulation reveals a generally
consistent stability pattern across all systems ( Figure
). All of the complexes exhibit
RMSD values that stabilize below 2 Å, indicating stable protein–ligand
interactions and minimal structural deviation over time. While minor
fluctuations are observed, particularly in the CHI3L1-siponimod complex,
these variations do not suggest significant instability but rather
reflect the dynamic nature of molecular interactions. Overall, the
results demonstrate comparable binding behaviors across all of the
compounds, reinforcing their potential to form reliable and stable
interactions with CHI3L1. This consistency underscores the suitability
of these ligands for further analysis and potential application.

RMSD of
CHI3L1 protein complexes with different compounds in 500
ns MD simulations. The RMSD values for inhibitor 30 (blue), fingolimod
(brown), G721-0282 (violet), ponesimod (cyan), siponimod (orange),
and K284 (green) indicate stable protein–ligand interactions,
with all complexes stabilizing below 2 Å.

The RMSF analysis indicates that the complexes
display comparable
fluctuation profiles with siponimod and ponesimod exhibiting the highest
levels of flexibility. While most residues exhibit fluctuations below
2Å, specific regions, including residues 200–220, 230–242,
and 261–281, show increased fluctuations, suggesting greater
flexibility in these segments ( Figure
A). To further elucidate the effects of inhibitor binding
on CHI3L1 dynamics, ED analysis was performed to assess the influence
of ligand binding on the principal motions of the protein backbone.
The ED results revealed that inhibitor binding induces conformational
changes in CHI3L1 complexes. Among the studied inhibitors, inhibitor
30 exhibited the lowest backbone motion, suggesting that its binding
reduces protein flexibility, thereby indicating its potential as a
potent inhibitor. In contrast, as observed in the RMSF analysis, other
ligands, such as siponimod, demonstrated flexible regions corresponding
to the CBD ( Figure
B, orange: CHI3L1-siponimod complex; blue: CHI3L1-Inhibitor 30 complex),
implying comparatively lower stability of these ligands within the
binding site (the ED analyses of fingolimod, G721-0282, K284, and
ponesimod, along with the contributions of the essential motions,
are presented in Figure S3 and Table S3 in the Supporting Information).

(A) RMSF (Å) averaged across two
independent replicas for
all simulated systems with highlighted regions corresponding to the
CBD. (B) ED analysis of 500 ns MD simulations for CHI3L1-siponimod
and CHI3L1-inhibitor 30, represented in orange and blue, respectively.
The first principal motion of the Cα atoms is depicted.

The MM/GBSA binding free energy results for six
compoundsinhibitor
30, fingolimod, G721-0282, K284, ponesimod, and siponimodbound
to CHI3L1 are displayed in Figure
. Among the compounds, inhibitor 30 demonstrates the
most stable interaction, with a binding free energy of −53.9
kcal/mol, suggesting its high affinity for the binding site. Fingolimod
follows as the second most stable ligand, with a binding free energy
of −44.2 kcal/mol. G721-0282, K284, and ponesimod show slightly
lower stabilities, with values of −40.1, −42.3, and
−43 kcal/mol, respectively. Siponimod exhibits the least stability,
with a binding free energy of −36.2 kcal/mol.

MM-GBSA binding free
energy results for the six compoundsinhibitor
30 (blue), fingolimod (brown), G721-0282 (violet), K284 (green), ponesimod
(cyan), and siponimod (orange)bound to the CHI3L1 protein.

These results are consistent with previous analyses,
which further
highlight the stabilizing effects of inhibitor 30 and fingolimod on
the protein–ligand complex. The agreement between MM/GBSA and
other structural analysis findings emphasizes the reliability of the
computational simulations, confirming that inhibitor 30 is the most
stable ligand, followed by fingolimod and ponesimod.

Discussion and Conclusions

Multiple sclerosis is a
chronic and multifactorial disease characterized
by inflammation and demyelination in the central nervous system. Its
complex pathophysiology and heterogeneous clinical presentations require
the development of innovative therapeutic strategies that can address
the diverse needs of patients. Biomarkers
have become essential in advancing MS diagnostics, improving patient
classification, guiding therapeutic decisions, and refining disease
management. Among these, Chitinase-3-like protein 1 has emerged as
a particularly promising prognostic biomarker, highlighting its potential
as a therapeutic target in MS. Its upregulation
in MS patients, along with its reduction following treatment, supports
its active involvement in disease mechanisms rather than its being
a passive marker. CHI3L1 plays a role in driving neuroinflammation,
glial activation, and tissue remodeling, key processes in MS pathology.
Targeting CHI3L1 aims to disrupt these pathways by inhibiting its
interaction with receptors, such as IL-13Rα2, thereby reducing
downstream inflammatory signaling. Although CHI3L1 inhibitors may
not directly suppress its expression, they block its function and
may help interrupt the feedback mechanisms that sustain its elevated
levels.

However, despite the increasing interest in CHI3L1,
there is a
notable lack of small-molecule inhibitors with a confirmed affinity
for this protein, with only a limited number reported in preclinical
models. This gap underscores the need for further research and development
of novel CHI3L1 inhibitors. This study aimed to address this challenge
by employing ligand-based virtual screening to identify potential inhibitors of CHI3L1 from a library of FDA-approved
drugs. This approach leverages the safety profiles and established
pharmacokinetics of existing drugs to streamline the drug discovery
process, bypassing the lengthy and expensive stages of de novo compound
development. Using fingerprint similarity methods and two selective
CHI3L1 inhibitors, G721-0282 and K284, as reference structures for our 2D similarity
search, we identified fingolimod, an FDA-approved sphingosine-1-phosphate
(S1P) receptor modulator, as a potential CHI3L1 inhibitor. Notably,
clinical evidence suggests that fingolimod treatment reduces CHI3L1
levels in the cerebrospinal fluid of MS patients, further validating its relevance as a candidate for CHI3L1-targeted
therapy.

Although fingolimod is known to exert its approved
pharmacological
action via its phosphorylated form, previous pharmacokinetic studies
have shown that both phosphorylated and nonphosphorylated forms coexist
in the blood.
,
These findings support the plausibility
that the parent compound may be available at sufficient levels in
vivoparticularly in the CNSto interact with CHI3L1,
as suggested by our computational results.

To expand on these
findings, two additional S1P modulators, siponimod
and ponesimod, were included in molecular docking and molecular dynamics
(MD) simulations alongside fingolimod and a reference CHI3L1 inhibitor
(inhibitor 30) derived from the X-ray crystal structure of CHI3L1
(PDB ID: 8R4X ).

MD simulations provided critical
insights into the stability and
dynamics of the protein–ligand interactions over a 500 ns time
scale, supporting the outcomes of the docking studies. Inhibitor 30
exhibited the most stable binding profile, maintaining strong and
persistent interactions with key residues, such as Trp99, Trp352,
and Trp31 within the hydrophobic groove of CHI3L1. These interactions,
including hydrogen bonds, salt bridges, and stacking interactions,
constrained the local conformational mobility of the protein, as evidenced
by the reduced root RMSF values for these residues. Fingolimod and
ponesimod showed intermediate stability, forming frequent interactions
with key binding site residues, which aligned with the docking predictions.
Although their RMSD traces were slightly higher than those of inhibitor
30, they demonstrated stable binding conformations that effectively
restricted the global protein motions, as shown by the ED analysis.

Siponimod displayed a distinct binding orientation compared with
the other compounds. While its RMSD values were slightly higher, indicating
greater conformational flexibility in the binding pocket, siponimod
consistently formed critical interactions with different residues,
such as Asp207. These interactions align with those observed in the
cocrystallized CHI3L1–inhibitor complex, suggesting that siponimod’s
binding may require subtle conformational adjustments to achieve optimal
stability. Despite its unique binding profile, siponimod’s
ability to interact with essential residues highlights its potential
as a CHI3L1 inhibitor, warranting further investigation. Binding free
energy calculations further validated these findings, ranking inhibitor
30 as the most stable binder, followed by fingolimod, ponesimod, G721-282,
K284, and siponimod. These results underscore the consistency between
docking predictions and MD simulation outcomes, reinforcing the potential
of repurposing S1P modulators for CHI3L1-targeted therapy in MS. Importantly,
our findings also highlight the ability of the CHI3L1 binding pocket
to accommodate ligands with varying protonation states and charge
distributions. Positively charged molecules such as fingolimod and
inhibitor 30 engaged the protein through salt bridges and hydrogen
bonds with complementary charged and polar residues. In contrast,
neutral compounds like ponesimod, G721-0282, and K284 adopted different
binding strategies, relying primarily on hydrophobic contacts and
π–π stacking interactions, particularly with aromatic
residues such as Trp99, Trp352, and Trp31 (the superimposed binding
modes of inhibitor 30, G721-0282, and fingolimod are shown in Figure S4 of the Supporting Information). The
zwitterionic nature of siponimod enabled both electrostatic and polar
interactions with residues like Asp207. This diversity in binding
modes suggests that CHI3L1 possesses a chemically versatile binding
site capable of stabilizing structurally and electrostatically diverse
ligands. Such flexibility enhances the feasibility of targeting CHI3L1
with a broader range of small molecules and supports the rationale
for ligand-based screening approaches that include chemically diverse
scaffolds.

In addition to the S1P receptor modulators, the LBVS
approach identified
several other FDA-approved drugs with distinct mechanisms of action,
including kinase inhibitors, alpha-1 adrenergic antagonists, and anti-inflammatory
agents, which shared structural similarities with the reference inhibitors.
These findings expand the pool of potential CHI3L1 inhibitors, which
presents new opportunities for further exploration. However, additional
in vitro and in vivo studies are essential to confirm the binding
mechanisms, therapeutic efficacy, and safety profiles of these candidates.
Such investigations will provide a foundation for developing novel
treatment targeting CHI3L1, addressing the unmet therapeutic needs
in MS, and potentially other inflammatory diseases.

In conclusion,
this study highlights the potential of repurposing
FDA-approved drugs, particularly S1P receptor modulators such as fingolimod,
ponesimod, and siponimod, as inhibitors of CHI3L1, a promising therapeutic
target in MS. By leveraging ligand-based virtual screening, molecular
docking, and molecular dynamics simulations, we identified and validated
the stability and binding affinity of these compounds for CHI3L1,
with results aligning consistently across computational approaches.
Fingolimod’s known ability to reduce CHI3L1 levels in clinical
settings further supports its potential utility in modulating CHI3L1-driven
inflammatory pathways.

Beyond S1P modulators, our findings suggest
a broader chemical
landscape of FDA-approved drugs with structural similarities to known
CHI3L1 inhibitors, warranting further experimental evaluation. These
results not only underscore the value of computational tools in accelerating
drug discovery but also open new avenues for therapeutic development
targeting CHI3L1, addressing critical unmet needs in MS treatment
and potentially benefiting other inflammatory conditions. To strengthen
these computational predictions, it is essential to test the identified
compounds in vitro to confirm their binding to CHI3L1
and to assess their functional inhibitory activity. Such validation
will be crucial to establish their therapeutic relevance and support
progression toward preclinical development.

Supplementary Material

Acknowledgments

Funding from Instituto de Salud Carlos
III (CIBERNED,
grants no. CB18/05/00040 to A.M. and CB22/05/00047 to M.C.) is acknowledged.

The structures
representing the binding mode of all compounds can be found at 10.5281/zenodo.14900502

The Supporting Information
is
available free of charge at https://pubs.acs.org/doi/10.1021/acsomega.5c01968 .

List of FDA-approved drugs selected using fingerprint
similarity, along with 2D diagrams of protein–ligand interactions,
ED analyses, and SMILE of all the compounds ( PDF )

The authors
declare no competing financial interest.

References

Associated Data

Supplementary Materials

Data Availability Statement

The structures
representing the binding mode of all compounds can be found at 10.5281/zenodo.14900502