Discovery of ‑Acylhydrazone Derivatives as ROCK Inhibitors: A Journey from Virtual Screening and Structure-Based De Novo Design to the Identification of ROCK2 Selective Inhibitors and Beyond.

Introduction
Rho-associated coiled-coil containing
kinases (ROCKs) are serine/threonine
kinases that act as principal effectors of the small GTPase RhoA and
regulate actin–myosin contractility, cell shape, motility,
adhesion, and cytokinesis. ROCK signaling controls cytoskeletal dynamics
through phosphorylation of substrates such as myosin light-chain phosphatase
(MYPT1) and LIM kinases, thereby modulating cell contractility, migration,
and extracellular matrix interactions.
,
Owing to their
central role in these processes, ROCK kinases have emerged as therapeutically
relevant targets, being implicated in cancer progression, fibrosis,
cardiovascular, and neurological disorders.
−

Structurally, ROCK1 and ROCK2 share a highly
conserved N-terminal kinase domain, a central coiled-coil
region that
mediates dimerization and contains the Rho-binding domain, and a C-terminal pleckstrin homology domain that contributes to
autoregulation and membrane association.
,,
High-resolution crystallographic data for ROCK kinases
show that most small-molecule inhibitors bind in the ATP site of the
kinase domain. Available structures indicate that the kinases adopt
a canonical active state (DFG-in/αC-in) and,
in general, small-molecule inhibitors act as type 1 ATP-competitive
inhibitors.
−

Small-molecule inhibitors of Rho-associated kinases (ROCKs)
encompass
diverse chemical classes and have been explored across multiple therapeutic
areas, reflecting the broad physiological and pathological significance
of ROCK signaling. Fasudil, a nonselective ROCK1/2 inhibitor, represents
a historical milestone as the first kinase inhibitor ever approved
for clinical use, preceding even the introduction of tyrosine kinase
inhibitors such as imatinib. It was approved
in Japan (1995) and China (2006) for the treatment of cerebral vasospasm,
,
and was also investigated in other conditions, such as amyotrophic
lateral sclerosis, Parkinson’s
disease, cancer and beyond. In ophthalmology, ripasudil and netarsudil are
approved for the treatment of glaucoma and ocular hypertension. More recently, the selective ROCK2 inhibitor
belumosudil received FDA approval in 2021 for chronic graft-versus-host
disease, marking the first isoform-selective ROCK inhibitor to reach
the clinic. Furthermore, next-generation
inhibitors such as verosudil, sovesudil and zelasudil are currently
in clinical trials for cancer, fibrotic, cardiovascular, neurological
and inflammatory diseases, underscoring the expanding therapeutic
scope of ROCK inhibition.
,
The chemical structure
of the mentioned ROCK inhibitors are shown in Figure
.
There is growing preclinical and translational
evidence linking
ROCK activity to tumor biology, ROCK signaling influences cancer-relevant
processes including cell invasion, migration, angiogenesis, mechanotransduction,
and therapy resistance. Experimental inhibition of ROCK reduces metastatic
behaviors and impairs tumor progression, supporting ROCK as a candidate
therapeutic target in oncology.
,,
Multiple ROCK inhibitors are being evaluated preclinically and in
early clinical studies for oncologic indications or are being repurposed
from vascular/ophthalmic uses toward cancer research.
,

Taken together, ROCK kinases combine a well-defined enzymatic
domain
amenable to small-molecule inhibition with pleiotropic control over
cytoskeletal and microenvironmental mechanisms that drive malignant
progression. These features make ROCK a tractable and biologically
rational target for antimetastatic and combination cancer therapies,
although isoform selectivity, context-dependent effects and potential
on-target toxicities remain key challenges for clinical translation.
In this work, we envisioned the usage of our internal collection
of small-molecule drug candidates, i.e., the LASSBio Chemical Library, in an initial hit identification campaign through
SBVS. Next, based on the identified hits, we applied a de novo design
strategy using the NAH framework as a central core linking the hinger
binding motif with the terminal aromatic rings that occupy the ROCK
affinity pocket. Our approach led to
the identification of potent ROCK1/2 dual inhibitors (12, 17 and 18), as well as ROCK2-selective
inhibitors (21). To our knowledge, this is the first
study reporting NAH derivatives as ROCK inhibitors.

Results and Discussion

Virtual Screening Campaign against ROCK
Through the
recently reported pharmacophoric map for the identification of ROCK
inhibitors in SBVS campaigns, we applied
the same methodology using our in-house chemical library, the LASSBio
Chemical Library. In this way, we were
able to identify 3 hits: LASSBio-1828 (1), LASSBio-1829 (2), and LASSBio-1919 (3). LASSBio-1828 (1) showed 43% inhibition of ROCK1 at
10 μM with an estimated IC50 of 44 μM, while
it showed a slightly less significant inhibition of ROCK2 (27%) at
the same concentration. In contrast, LASSBio-1829 (2),
which is structurally similar to LASSBio-1828 (1), differing
only at the hinge binder motif, showed IC50 values with
a 1-digit micromolar range for both ROCK isoforms (ROCK1 IC50 = 1.74 μM; ROCK2 IC50 = 2.90 μM). LASSBio-1919
(3), on the other hand, showed selectivity for ROCK2
inhibition with an IC50 value of 2.61 μM while showing
23% max inhibition of ROCK1 at 10 μM. In Figure
D, the interaction mode of LASSBio-1829 (2) in ROCK1 is shown, in which the 7-azaindole moiety forms
a bidentate hydrogen bond interaction with the main chain of Met156
and Glu154, located at the hinge region. It is important to note that,
compared to the cocrystallized ligand in the PDB 6E9W, this ligand can
access the affinity pocket, forming a cation-π interaction with
the catalytic Lys105 residue. In contrast to compounds 1 and 2, 3 features a shorter linker and
an isoquinoline hinge-binding motif. Docking and subsequent pose inspection
indicate that 3 can still be accommodated in the ATP-binding
site, with the isoquinoline ring positioned at the hinge region to
maintain a canonical hinge-anchoring interaction pattern (Figure
E). Notably, the
predicted interaction pattern and overall pose of 3 were
essentially analogous to that observed for compound 21, which will be discussed further. The identified hits from LASSBio
Chemical Library do not present conformational freedom to access this
cavity. From our SBVS some major points were depicted: (i) focusing
on hinge binders that can perform bidentate hydrogen bonding interactions,
as the 7-azaindole moiety, seems to be more promising; (ii) the N-acylhydrazone moiety is an interesting linker, but it
needs to be more flexible; (iii) the trimethoxyphenyl moiety resulted
in ROCK2 selectivity.

De Novo Design of New N-Acylhydrazone Derivatives
Based on the three identified hits and the information gathered
in the virtual screening campaign, we conducted a de novo design strategy.
This strategy was focused on the NAH framework as a linker, bridging
the hinge binding core with the terminal aromatic system. Thus, allowing
the effective hydrogen bonding interactions with the hinge, while
performing a cation-π interaction with the catalytic lysine
in the affinity pocket of ROCK1/2.
The de novo design approach
was focused on our internal collection of reagents and fragments.
We built a KNIME workflow to filter and
collect carboxylic acids. These were transformed into the corresponding
hydrazides using the RDKit “Two Component Reaction”
node through a SMARTS pattern for the reaction (Figure
A). Next, aldehydes of interest (isoquinoline-5-carbaldehyde
(6), quinoline-4-carbaldehyde (7), 4-(pyridin-4-yl)­benzaldehyde
(8), isonicotinaldehyde (9), 1H-pyrrolo­[2,3-b]­pyridine-5-carbaldehyde (10), 1H-pyrrolo­[2,3-b]­pyridine-3-carbaldehyde (11)) were selected
from our catalog and condensed with the generated hydrazides in a
second reaction using also the RDKit “Two Component Reaction”
node through a SMARTS pattern for the reaction (Figure
A). These resulted in 1282 unique virtual
NAH derivatives that were subsequently filtered based on sp3 fraction (≥0.1) and number of stereo centers (= 0), resulting
in 321 derivatives (Figure
A), which were analyzed by means of chemical space (Figure
B) and exported in
SDF format for three-dimensional geometry optimization in Spartan’24,
followed by molecular docking studies using GOLD’2025.1.0.
In this de novo design approach, we intentionally
explored both
phenyl-hydrazone linkers and “direct” hydrazone linkers
(i.e., hinge binders connected directly to the NAH unit without an
intervening phenyl ring), motivated by the initial hit LASSBio-1919
(3), which features an isoquinoline hinge binder directly
attached to the NAH framework (Figure
A). Importantly, during docking-based prioritization
of the virtual de novo library, the candidates that most consistently
satisfied the key pharmacophoric requirements in the ROCK ATP siteproductive
hinge anchoring together with the expected interaction toward the
affinity pocketwere predominantly those lacking the phenyl
linker.
Keeping in mind our goal, i.e., identifying a small
molecule capable
of binding simultaneously at the hinge region and the affinity pocket,
we selected two promising NAH derivatives from our de novo design
approach. According to our docking protocol, 12 and 13 are both capable of performing the key interactions with
ROCK1 and ROCK2, as a consequence of their higher flexibility, a necessary
feature that was already anticipated in the virtual screening campaign.
As an example, the interaction mode of 12 in ROCK1 and
ROCK2 is shown in Figure
A,B, in which the only difference between 12 and 13 regards the binding mode at the hinge. While 13 forms a monodentate hydrogen bond through the pyridine ring, 12 forms a bidentate hydrogen bond through the 7-azaindole
core. Next, 12 and 13 were synthesized (Scheme
, see the next section
for more details) and evaluated against ROCK1/2 (Figure
C,D, Table
) in order to validate our approach. Compound 13 exhibited IC50 values of 3.39 μM and 3.98
μM against ROCK1 and ROCK2, respectively, whereas 12 was 450- and 280-fold more potent, with IC50 values of
7.5 nM and 14.1 nM for ROCK1 and ROCK2, respectively. Ultimately,
this not only validates our approach but also clearly indicates that
the design of new derivatives should prioritize the bidentate hydrogen
bonding in the hinge while maintaining some level of sp3 fraction
at the NAH framework.
Next, we design a new congeneric series of NAH derivatives
using 12 as a prototype. We applied a combination of
homologation and bioisosteric replacement strategies in order to evaluate the linker size
and the
effect of replacing the 7-azaindole pharmacophore core with other
bicyclic heteroaromatic rings, such as indazole, 5,7-azaindole, and
pyrazole, maintaining the acceptor/donor hydrogen bond features, giving
rise to compounds 14-20. In addition, four
analogues were designed (21-24) based on
molecular hybridization, in which the
common terminal indole moiety was replaced by a trimethoxyphenyl from
LASSBio-1919 (3). This was done in order to analyze the
trimethoxyphenyl group in depth, as an auxophoric group capable of
generating ROCK2 selectivity over ROCK1.

Chemical Synthesis
The designed molecules were obtained
through a linear synthetic route shown in Scheme
. Initially, the commercially available corresponding
carboxylic acids (25, 26, 29 and 30) were converted to their corresponding methyl
ester derivatives using Fischer esterification conditions. Next, the
esters were converted to their corresponding hydrazide derivatives
(27, 28, 31 and 32) by carbonyl substitution reaction in the presence of hydrazine
hydrate. Finally, the aimed compounds (12-24) were obtained using condensation reactions with the corresponding
aldehydes, catalyzed in an acidic medium.
The characterization
of the synthesized NAH derivatives was conducted using1H and 13C Nuclear Magnetic Resonance (NMR) spectroscopy,
high pressure liquid chromatography (HPLC) and high-resolution mass
spectrometry (HRMS). The NMR spectra of all compounds bearing one
or two methylene units between the carbonyl group and the NAH moiety
displayed duplicated signals, as illustrated in Figures S1 and S3, which shows the 1H and 13C spectrum of compound 18, respectively. This
duplication initially raised the question of a possible mixture of
diastereomers or conformers. 1H NMR spectra acquired at
80 °C showed coalescence of duplicated signals (Figure S2), confirming that the duplication arises from conformers.
Moreover, this is likely a consequence of syn- and anti-rotamers around the NAH amide σC–N bond.
Potential energy surface (PES) scans using the ωB97-XD/6–311+G­(d,p)
hybrid functional for evaluation of the amide σC–N bond
of 18 and its superior homologue 14 revealed
an energy barrier of approximately 16 kcal·mol–1 at room temperature in polar solvent, consistent with the presence
of a slow dynamic equilibrium between rotamers (Figure S4), as already reported in previous studies of structure–property
relationship of NAH derivatives.
−

Moreover, the (E) configuration of the NAH moiety is further supported
by the chemical shift of the imine proton and by previous studies
on related bioactive NAH.
,−

SAR of the Congeneric Series of Hydrazones for ROCK Inhibition
The evaluation of NAH derivatives against ROCK1 and ROCK2 revealed
a broad spectrum of activity, ranging from low micromolar to low nanomolar
potency, along with distinct selectivity patterns. Among the derivatives
with a two-methylene linker (12–16), the best hinge-binding scaffolds were identified as the 7-azaindole
(12), as previously noted, and the indazole (14), which exhibited IC50 values of 0.106 μM for ROCK1
and 0.036 μM for ROCK2. In contrast, replacing the 7-azaindole
with a 5,7-azaindole (15) or simplifying the indazole
to a pyrazole ring (16) produced inactive compounds,
highlighting the importance of these specific scaffolds. Reducing
the linker length to a single methylene generated the inferior homologues 17–20, among which the 7-azaindole derivative 17 (IC50 = 0.005 μM for ROCK1 and 0.003 μM
for ROCK2) and the indazole derivative 18 (IC50 = 0.003 μM for ROCK1 and 0.001 μM for ROCK2) emerged
as the most potent compounds of the series. Comparison of 17 and 18 with their superior homologues, i.e., 12 and 14, clearly demonstrates that the one-methylene
linker provides the optimal length to engage the ROCK1 and ROCK2 affinity
pockets, likely by better positioning the scaffold to establish a
cation−π interaction with the catalytic lysine. Conversely,
compounds 19 and 20 were inactive against
both kinases at 1 μM. The trimethoxyphenyl derivatives (21–24) also provided insightful results.
Compound 21 displayed remarkable ROCK2 selectivity, with
an IC50 of 0.051 μM for ROCK2 and 1.143 μM
for ROCK1, showing 21-fold selectivity for ROCK2 over ROCK1. However,
substituting the indazole (21) with a 7-azaindole (22) abolished this selectivity, yielding similar potencies
against both ROCK1 (IC50 = 0.394 μM) and ROCK2 (IC50 = 0.232 μM). Increasing the linker length by one methylene
while maintaining the indazole core afforded compound 23 (IC50 = 0.870 μM for ROCK1 and 0.353 μM for
ROCK2), which showed reduced ROCK2 potency, slightly improved ROCK1
inhibition, and overall loss of selectivity in comparison with 21. Similarly, extending the linker in compound 22 gave rise to compound 24 (IC50 = 1.254 μM
for ROCK1 and 1.077 μM for ROCK2), resulting in decreased potency
for both targets. These findings establish a clear structure–activity
relationship within the NAH series. Derivatives 12, 17, and 18 emerged as the most potent inhibitors,
achieving balanced ROCK1 and ROCK2 inhibition at low nanomolar concentrations,
while compound 21 stood out by exhibiting 21-fold selectivity
for ROCK2. Achieving ROCK2 selectivity is particularly challenging,
given that ROCK1 and ROCK2 share more than 90% sequence identity within
their kinase domains. Although a recent
study suggested that ROCK2 selectivity may arise from interactions
with a hydrophobic pocket accessible only in ROCK2 as a result of
a DFG-out conformation, this conformation
is induced by a shift in the αC-helix that
can occur in ROCK2 but is hindered in ROCK1 due to the residue difference
L210 in ROCK2 versus F194 in ROCK141 (for details, see Figure S5 in the Supporting Information). In this way, ROCK2-selective inhibitors function
as type 2 kinase inhibitors. Taken together,
our results underscore the NAH scaffold as a promising framework for
the development of potent and potentially isoform-selective ROCK inhibitors.
To obtain an initial overview of selectivity, compounds 18 and 21 were profiled at 500 nM against a representative
panel of kinases structurally related to ROCKIKKβ/IKBKB,
JAK1, MLCK/MYLK, PKAcβ, PKCα, and PKG1α (Table
). Importantly, 18 is substantially more potent on ROCK2 (IC50 =
1 nM) than 21 (IC50 = 51 nM); therefore, screening
both compounds at a single fixed concentration (500 nM) means that 18 was evaluated at ∼500-fold its ROCK2 IC50, whereas 21 was evaluated at ∼10-fold. This
difference in on-target potency can influence the apparent selectivity
margin across a small panel when only one concentration is used.

18 displayed potent off-target inhibition
of PKAcβ
(94%) and PKG1α (98.9%), indicating a limited selectivity within
the AGC subfamily. In contrast, 21 exhibited minimal
inhibition of all tested kinases (<10%) except for PKG1α
(32.9%), suggesting a markedly improved selectivity profile relative
to 18. These results reinforce that the indazole–trimethoxyphenyl
hybrid structure of 21 confers superior discrimination
against the tested kinases, potentially linked to its ability to stabilize
the DFG-out conformation of ROCK2 rather than interacting promiscuously
with other kinases. Additional profiling across a broader kinase panel
would further elucidate the selectivity landscape of this chemotype.

Docking and Molecular Dynamics Simulations of Compounds 18 and
21
Compounds 18 and 21 were investigated
for their interaction modes within ROCK1 (PDB: 6E9W

DFG‑in
) and ROCK2 (PDB: 7JNT

DFG‑in
and 8GDS
DFG‑out
).
,
Two ROCK2 crystal structures were employed, since 8GDS presents
the open hydrophobic pocket caused by a shift of the αC-helix (DFG-out conformation). The aim was to characterize
the binding modes of both compounds and, through molecular dynamics
simulations, to explore the molecular basis underlying the 21-fold
selectivity of 21 for ROCK2 over ROCK1. We anticipated
that 21 would exhibit a better fit in the 8GDS structure
compared to 7JNT and 6E9W.
The binding poses of 18 and 21 in 6E9W and 7JNT followed the expected
pattern: both compounds established bidentate hydrogen bonds with
the hinge region, interacting with the backbone of the corresponding
glutamate and methionine residues (Figure
A–D). In addition to hinge anchoring,
the linker length relative to compound 12 is central
for the alignment between the hinge-binding region and the terminal
aromatic system in this series. 12 (two methylene units)
is already capable of establishing a Lys-centered cation−π
interaction within the affinity pocket; however, shortening the spacer
to one methylene unit in 18 optimizes the geometry of
this interaction and improves the overall fit of the terminal aromatic
group toward the affinity pocket. In contrast, removing the methylene
spacer in 21 (direct NAH connection) shortens the reach
and is not expected to sustain the same cation−π interaction,
which is consistent with the ROCK2 potency differences (Table
). Moreover, unlike derivative 12, the indazole nucleus oriented the carbonyl oxygen of the
NAH moiety toward the catalytic lysine, forming a stable hydrogen
bond. This interaction was observed for both 18 and 21 in ROCK1 (6E9W) and ROCK2 (7JNT). Moreover, compound 18 was capable of forming cation−π interactions
within the affinity pocket framed by the catalytic lysine and the
G-loop.
It is noteworthy that the 8GDS crystal of ROCK2
presents a significant
conformational rearrangement, not only involving the αC-helix
and the DFG motif but also the catalytic lysine, as illustrated in Figure S5. In this conformation, the catalytic
lysine is displaced forward, assisting in the orientation of the cocrystallized
ligand toward the open hydrophobic pocket, while simultaneously closing
the affinity pocket previously formed between the lysine and the G-loop.
At first glance, neither 18 nor 21 displayed
an optimal fit in this structure compared to 7JNT (Figure
E,F). For 18,
the hydrogen bonds appeared less directional, whereas 21 adopted a monodentate hinge interaction, forming a single hydrogen
bond with Met172. Due to the displacement of the catalytic lysine,
hydrogen bonding with this residue was lost, however, the terminal
aromatic subunits of both compounds shifted to occupy the hydrophobic
pocket.
These docking results provided the structural basis
for subsequent
molecular dynamics simulations aimed at evaluating the stability of
the complexes and elucidating the structural determinants responsible
for the enhanced ROCK2 selectivity of compound 21.
To further investigate the structural determinants underlying the
differential inhibition of ROCK1 and ROCK2, molecular dynamics (MD)
simulations were carried out for 18 and 21 complexed with ROCK1 (6E9W
DFG‑in
), ROCK2 (7JNT
DFG‑in
), and ROCK2
(8GDS
DFG‑out
). The simulations
provided insights into the conformational stability of the complexes
and into the ability of each ligand to accommodate and stabilize the
distinct kinase conformations.
Overall RMSD profiles (Figure
A–C) indicated
that all systems reached equilibration
during the simulation time. In ROCK1 (6E9W), the total RMSD values
stabilized between 3–4 Å for the ligand-bound systems
and 2–3 Å for the apo form, suggesting that ligand binding
did not induce significant global rearrangements. Similar behavior
was observed for ROCK2 (7JNT), with RMSD values between 4–5
Å for both 18 and 21, and 2–3 Å for the apo enzyme.
In the 8GDS structure (ROCK2
DFG‑out
), all complexes displayed slightly higher fluctuations (4–5
Å), reflecting the intrinsic flexibility of this open conformation.
RMSF analysis (Figure
D–F) further confirmed that ligand binding did not significantly
alter the mobility pattern of key secondary-structure elements, indicating
that conformational changes were localized rather than global.
Ligand RMSD profiles (Figure
G–I) showed remarkable stability,
remaining
below 2 Å throughout the simulations, except for compound 18 in 8GDS, which required approximately 50 ns to reach equilibration.
This higher fluctuation was accompanied by a conformational transition
from the DFG-out to the DFG-in state, suggesting that 18 does not efficiently stabilize the open hydrophobic pocket characteristic
of the DFG-out conformation. In contrast, 21 maintained
the DFG-out conformation throughout the trajectory, indicating its
greater structural compatibility with this ROCK2-specific state. This
observation provides a dynamic explanation for the experimentally
observed 21-fold selectivity of compound 21 toward ROCK2.
Hydrogen bond analysis with the hinge residues Glu154/Met156 (ROCK1)
and Glu170/Met172 (ROCK2) confirmed the persistence of key interactions
across all systems (Figure
J–L), with distances consistently around 2 Å.
In agreement with the docking results, 21 formed a monodentate
hydrogen bond with Met172 in the DFG-out structure (8GDS), while maintaining
a stable orientation within the binding pocket.
The distance
between the carbonyl oxygen of the N-acylhydrazone
moiety and the catalytic lysine (Lys105 in ROCK1;
Lys121 in ROCK2) remained close to 3 Å in both ROCK1 and ROCK2
(7JNT) complexes (Figure
M,N). However, in the ROCK2 (8GDS) complex with 18, the distance initially increased to approximately 8 Å during
the first 20 ns and then gradually decreased and stabilized near 3
Å (Figure
N).
This shift in distance reflects the conformational rearrangement of
the activation loop, marking the transition from the DFG-out to the
DFG-in state.
The cation−π interactions involving
the indole moiety
of 18 and the catalytic lysine were also analyzed (Figure
O). In ROCK1 (6E9W)
and ROCK2 (7JNT), these distances remained stable around 5 Å
and 4 Å, respectively, corroborating the presence of persistent
electrostatic interactions within the affinity pocket. Conversely,
in the DFG-out structure (8GDS), the distance initially increased
to 10 Å before stabilizing near 4 Å after 20 ns, further
reflecting the structural rearrangement of the catalytic lysine during
the conformational transition.
Taken together, the MD results
demonstrate that 21 not only fits well within the DFG-out
conformation of ROCK2 but
also stabilizes it throughout the simulation, maintaining productive
hinge interactions and consistent orientation of its N-acylhydrazone core. In contrast, 18 fails to preserve
the open hydrophobic pocket, favoring the return to a DFG-in–like
arrangement. These findings support the hypothesis that the selectivity
of 21 for ROCK2 arises from its ability to stabilize
and exploit the DFG-out conformation, a structural feature that is
energetically disfavored or sterically hindered in ROCK1.

PAMPA Assay
Prior to cellular evaluation, compounds 12, 17, 18, and 21,
along with the reference inhibitors fasudil and belumosudil, were
assessed for permeability using the Parallel Artificial Membrane Permeability
Assay (PAMPA). In the PAMPA-GIT model, fasudil (Pe = 3.575 × 10–6 cm/s) and 21 (Pe = 5.562 × 10–6 cm/s) were classified
as highly permeable, corresponding to predicted absorbed fractions
of 74% and 88%, respectively. The remaining compounds were not soluble
under the experimental conditions that mimic the gastrointestinal
barrier. In the PAMPA-BBB model, fasudil (Pe = 1.364
× 10–6 cm/s) and 21 (Pe = 0.537 × 10–6 cm/s) were classified as nonpermeable
(CNS−), while 12 (Pe = 2.453
× 10–6 cm/s) and 17 (Pe = 2.614 × 10–6 cm/s) showed uncertain permeability
(CNS±). Belumosudil and 18 were insoluble under
the experimental conditions, which mimic the blood-brain barrier.

Cell Viability Assay
Compounds 12, 17, 18, and 21 were selected for
in vitro functional evaluation to confirm ROCK engagement at the cellular
level. Initially, cell viability was assessed after 24 h exposure
to tested compounds at a concentration of 30 μM, using 1% DMSO
as the vehicle. Fasudil and belumosudil were included as reference
inhibitors. Results were normalized to the 1% DMSO control (Figure
). At the tested
concentration, all compounds reduced cell viability by less than 30%,
with 17 showing the strongest effect (28.8% inhibition).

Wound Healing Assay
The effects of the selected compounds
on cell migration were evaluated using a wound healing (scratch) assay
with the MDA-MB-231 triple-negative breast cancer cell line. This
cell line is known for its high aggressiveness and metastatic potential
and exhibits overexpression of both ROCK isoforms, with ROCK2 levels
exceeding those of ROCK1. Considering
that cell migration and invasion are crucial processes in cancer metastasis,
and that the Rho/ROCK signaling pathway plays a central role in their
regulation, we examined the effects of
molecules developed in our research group as potential ROCK1/2 inhibitors
on this cellular model.
Cells were treated with four concentrations
(0.1, 10, 30, and 100 μM) of the test compounds 12, 17, 18, and 21, as well
as the reference inhibitors fasudil and belumosudil, using 1% DMSO
as a vehicle control. 12 and its inferior homologue 17, quantification of wound closure was not possible at 100
μM due to compound precipitation (Figure
A,B, respectively). For 12,
precipitation was also observed at 30 μM. As shown in Figure
, both compounds
inhibited migration from 10 μM. 12 showed similar
inhibition at 10 and 30 μM, whereas 17 demonstrated
enhanced inhibition at 30 μM, likely due to improved solubility
at this concentration. For both compounds, changes in cell morphology
became apparent above 10 μM (Figures S7 and S8, Supporting Information)
compared with the DMSO control.
According to Figure
C,G, 18 reduced wound closure to less
than 50% at 10
μM and to below 25% at 30 μM, showing similar inhibition
at 100 μM. Although precipitation occurred at 100 μM,
no visible precipitate was detected at 30 μM. Morphological
alterations of the cells were evident from 30 μM onward (Figure S9, Supporting Information). 21 exhibited a pronounced effect at lower concentrations:
as early as 0.1 μM, a significant reduction in migration was
observed (p < 0.05) compared with the control.
At 10 μM, the wound closure area was reduced by ≤25%
(Figure
D,G). At 30
μM, fewer crystals were visible in the well, whereas at 100
μM, compound precipitation was clearly observed (Figure S10, Supporting Information). Morphological changes were also detected at concentrations above
10 μM.
As shown in Figure
E,G, fasudil reduced cell migration by approximately
50% after 24
h at 30 μM. Microscopic analysis revealed that at 30 μM,
the wound area remained largely unfilled compared with lower concentrations,
and at 100 μM, wound closure was minimal. In addition, morphological
changes in the cells were evident at 100 μM (Figure S11), suggesting that cell viability may be compromised
at higher concentrations.
In Figure
F,G,
the results for belumosudil are shown. Migration inhibition was observed
from 10 μM onward, with the filled area reduced to less than
50% at both 10 and 30 μM, with no significant difference between
these two concentrations. At 100 μM, the assay could not be
quantified due to poor solubility of the compound, and amorphous precipitates
were already visible at 30 μM (Figure S12). Although both fasudil and belumosudil act as ROCK1/2 inhibitors,
belumosudil inhibited migration at lower concentrations. This behavior
aligns with its reported selectivity for ROCK2, whereas fasudil displays
minimal isoform preference (>2-fold), despite showing similar potency
toward ROCK2. Differences in cell permeability
and compound solubility may also contribute to these results.
Notwithstanding, because our selectivity assessment was restricted
to a limited set of structurally related kinases, we cannot exclude
that part of the migration phenotypeespecially at higher concentrations
where morphological alterations were observedmay involve off-target
contributions. While the overall behavior is consistent with ROCK
pathway modulation and aligns with the effects of the reference inhibitors
(fasudil and, particularly, the ROCK2-selective belumosudil), a broader
kinase profiling panel and orthogonal target-engagement readouts in
cells will be important in future studies to more conclusively disentangle
on-target ROCK inhibition from potential off-target mechanisms.
Taken together, the four identified NAH derivatives (12, 17, 18, and 21), showed
stronger inhibition of cell migration than fasudil and comparable
effects to belumosudil in the MDA-MB-231 cell line. These findings
support their potential as promising leads to modulate the Rho/ROCK
signaling pathway in highly invasive cancers, while highlighting the
need for expanded selectivity profiling and orthogonal cellular target-engagement
studies to further substantiate the mechanism.

Conclusion
In this study, a rational and integrative
strategy combining structure-based
virtual screening, structure-based de novo design, molecular modeling,
and experimental validation enabled the discovery of potent and selective
ROCK inhibitors within the NAH chemical space. Starting from the LASSBio
Chemical Library, the approach led to the identification of LASSBio-1829
(2) and LASSBio-1919 (3) as initial hits,
which guided the structure-based de novo design approach of new derivatives
featuring optimized hinge binders and flexible linkers, followed by
experimental validation of two selected NAH derivatives and SAR exploration.
Among the synthesized compounds, derivatives 12, 17, and 18 exhibited balanced low nanomolar inhibition
of both ROCK1 and ROCK2, whereas 21 demonstrated remarkable
ROCK2 selectivity (21-fold), a rare and therapeutically valuable property.
In addition, we acknowledge that a certain degree of structural similarity
at the hinge-anchoring region is a general feature of ROCK inhibitors,
as can be observed in Figure
, reflecting the conserved hydrogen-bonding requirements for
productive ATP-site recognition.
Assays against structurally
related kinases revealed that 21 inhibited less than
10% of the evaluated off-targets at
500 nMexcept for PKG1α (32.9%)while 18 strongly inhibited PKAcβ (94%) and PKG1α (98.9%), indicating
that the indazole–trimethoxyphenyl hybrid scaffold of 21 confers substantial selectivity for ROCK2 over other related
kinases.
Molecular docking and molecular dynamics simulations
revealed that 21 preferentially stabilizes the DFG-out
conformation of ROCK2,
providing a mechanistic explanation for its isoform selectivity. In
the PAMPA assay, only 21 showed measurable passive permeability
under our experimental conditions, supporting its biopharmaceutical
profile among the tested derivatives. In cell-based assays, the evaluated
compounds significantly reduced migration of MDA-MB-231 triple-negative
breast cancer cells at micromolar concentrations, providing a functional
phenotype consistent with modulation of migration-related pathways.
However, these functional assays do not constitute direct evidence
of cellular ROCK1/2 target engagement, and off-target contributions
cannot be excluded given the limited kinase selectivity profiling
performed to date. Future studies incorporating orthogonal cellular
target-engagement readouts and broader kinase profiling will be important
to more conclusively attribute the observed phenotype to on-target
ROCK inhibition.
Overall, our findings establish N-acylhydrazone
as a versatile and synthetically accessible scaffold for the development
of next-generation ROCK inhibitors, highlighting key structural determinants
for potency and isoform selectivity. These results not only expand
the chemical space of ROCK modulators but also open new opportunities
for the design of antimetastatic and antifibrotic agents targeting
ROCK2 with improved efficacy and safety.
Importantly, we acknowledge
a potential limitation of the N-acylhydrazone scaffold
related to chemical stability.
Acylhydrazones can be susceptible to hydrolysis under aqueous/physiological
conditions, which may regenerate the corresponding aldehyde and hydrazide,
depending on factors such as pH, substituent electronics, and the
local environment. While the present work focuses on potency, selectivity,
and cellular activity, future studies will explicitly address this
liability through systematic stability profiling in buffered media
and biological matrices (e.g., plasma) and by monitoring potential
degradation products. Such data will guide further optimization efforts,
including stabilizing structural modifications or the exploration
of bioisosteric alternatives when needed.

Experimental Session

Chemistry

General Information
The melting points of the compounds
were determined using a Quimis 340 apparatus and are corrected. 1H NMR spectra were determined in dimethyl sulfoxide and the
solvent was used as an internal standard using a Varian 500-MR at
500 MHz or a Varian 400-MR at 400 MHz. 13C NMR spectra
were resolved using the same solvent and spectrometers at 125 or 100
MHz. The progress of all reactions was monitored through thin-layer
chromatography performed on 2.0 × 6.0 cm aluminum sheets precoated
with silica gel 60 (HF-254, Merck) to a thickness of 0.25 mm. The
developed chromatograms were analyzed under ultraviolet light (254–365
nm). The reagents and solvents were purchased from commercial suppliers
and used as received. Analytical HPLC was performed for compound purity
determinations using a Shimadzu LC- 20AD with a Kromasil 10 C18 column
(4.6 mm × 259 mm) and a Shimadzu SPD-M20A detector. The solvent
system used for the HPLC analyses was (A) methanol/water at 7:3 or
(B) acetonitrile/water 5:5. The isocratic HPLC mode was used, and
the flow rate was 1.0 mL/min. The purity of the compounds was higher
than 95%. High-resolution mass spectrometric analyses were performed
on a Q Exactive hybrid quadrupole–Orbitrap mass spectrometer
(Thermo Fisher Scientific, Bremen, Germany) equipped with an electrospray
ionization (ESI) source. Samples were prepared by adding methanol
with 0.1% formic acid, resulting in a final concentration of 10 μg/mL.
The samples were analyzed by direct infusion at a flow rate of 10
μL/min. Spectra were acquired over the m/z range 100–1500 at a resolving power of 70,000 (fwhm
at m/z = 200).

Synthesis of the N-Acylhydrazone Derivatives
1, 2 and 3
Compounds 1, 2, and 3 were synthesized as previously described.
−

The structures
of all compounds were confirmed by 1H NMR spectroscopy,
and the spectra were consistent with the data previously reported.
HPLC analyses indicated that all compounds exhibited purity higher
than 95%.

Procedure for the Synthesis of Hydrazide Intermediates (27,
28, 31 and 32)
The corresponding carboxylic acid (1.0 equiv.,
2 g) was dissolved in 200 mL of methanol (MeOH) in a 500 mL round-bottom
flask. Subsequently, 1.0 mL of a 1 M H2SO4 solution
in MeOH was added, and the reaction mixture was refluxed for 12 h
under magnetic stirring. After completion, hydrazine monohydrate (10.0
equiv., 100%) was added to the same flask. The reaction mixture was
then refluxed for an additional 2 h under magnetic stirring. The solvent
was removed under reduced pressure and the resulting solid was resuspended
in n-hexane, filtered under vacuum, and washed with
diethyl ether to afford the desired hydrazide.

2-(1H-Indol-3-yl)­acetohydrazide (27)
The title compound
was obtained as a white solid in 55% yield (MP: 125–128 °C). 1H NMR (500 MHz, DMSO-d
6) δ
10.64 (s, 1H), 8.82 (s, 1H), 7.57 (d, J = 7.9 Hz,
1H), 7.34 (d, J = 8.1 Hz, 1H), 7.17 (d, J = 2.2 Hz, 1H), 7.09 – 7.04 (m, 1H), 7.00 – 6.96 (m,
1H), 4.12 (s, 2H), 3.49 (s, 2H).

3-(1H-Indol-3-yl)­propanehydrazide (28)
The title compound
was obtained as a white solid in 59% yield (MP: 118–121 °C). 1H NMR (500 MHz, DMSO-d
6) δ
10.52 (s, 1H), 8.76 (s, 1H), 7.52 (d, J = 7.9 Hz,
1H), 7.33 (d, J = 8.1 Hz, 1H), 7.08 – 7.04
(m, 2H), 6.99 – 6.95 (m, 1H), 4.09 (s, 2H), 2.95 (t, J = 7.7 Hz, 2H), 2.42 (t, J = 7.7 Hz, 2H).

3,4,5-Trimethoxybenzohydrazide (31)
The title compound
was obtained as a white solid in 93% yield (MP: 160–163 °C). 1H NMR (500 MHz, DMSO-d
6) δ
9.50 (s, 1H), 7.17 (s, 2H), 4.37 (s, 2H), 3.83 (s, 6H), 3.74 (s, 3H).

2-(3,4,5-Trimethoxyphenyl)­acetohydrazide (32)
The title
compound was obtained as a white solid in 94% yield (MP: 120–123
°C). 1H NMR (500 MHz, DMSO-d
6) δ 8.90 (s, 1H), 6.58 (s, 2H), 4.13 (s, 2H), 3.77 (s,
6H), 3.67 (s, 3H), 3.30 (s, 2H).

Procedure for the Synthesis of the N-Acylhydrazone
Derivatives 12–24
The corresponding hydrazide (1.0
mmol) was dissolved in 10 mL of absolute ethanol in a 50 mL round-bottom
flask. The appropriate aldehyde (1.0 mmol) and two drops of 37% HCl
solution were then added, and the reaction mixture was refluxed for
4 h. Reaction progress was monitored by TLC (eluent: dichloromethane:methanol,
9.5:0.5 v/v). Upon completion, the reaction mixture was cooled to
0 °C, and the resulting solid was collected by vacuum filtration.
When necessary, the crude products were further purified by recrystallization
from ethanol/water.

(E)-N′-((1H-Pyrrolo­[2,3-b]­pyridin-3-yl)­methylene)-3-(1H-Indol-3-yl)­propanehydrazide (12)
The title compound was obtained as a yellow solid
in 70% yield (MP: 245–248 °C). 1H NMR (500
MHz, DMSO-d
6) δ 12.04 (s, 1H), 11.15,
11.06 (s, 1H), 10.80 (s, 1H), 8.51, 8.36 (dt, J =
7.2, 3.6 Hz, 1H), 8.30, 8.16 (s, 1H), 8.30 – 8.27 (m, 1H),
7.90 (d, J = 3.9 Hz, 1H), 7.58 (d, J = 7.8 Hz, 1H), 7.34 (dd, J = 8.1, 2.6 Hz, 1H),
7.21 – 7.14, 7.10 – 7.04 (m, 3H), 6.98 (m, 1H), 3.07
(s, 2H), 3.03 (t, J = 7.6 Hz, 1H), 2.58 (t, J = 7.6 Hz, 1H). 13C NMR (125 MHz, DMSO-d
6) δ 173.5, 171.4, 167.8, 149.3, 144.0,
143.9, 142.5, 139.6, 139.5, 136.3, 136.3, 130.1, 129.7, 129.5, 127.1,
127.1, 122.3, 122.2, 121.0, 118.4, 118.3, 118.2, 116.9, 116.8, 116.7,
116.5, 114.1, 113.7, 111.4, 111.4, 110.6, 110.5, 35.2, 32.9, 20.9,
20.2. HRMS calculated for [M + H]+ = 332.15059, found =
332.15027. HPLC purity of 96.3%, RT = 2.709 min (λ = 254 nm).

(E)-3-(1H-Indol-3-yl)-N′-(Pyridin-4-ylmethylene)­propanehydrazide
(13)
The title compound was obtained as a light
yellow solid in 69% yield (MP: 228–230 °C). 1H NMR (500 MHz, DMSO-d
6) δ 11.66,
11.56 (s, 1H), 10.78 (s, 1H), 8.62 (d, J = 5.5 Hz,
1H), 8.58 (d, J = 5.5 Hz, 1H), 8.14, 7.95 (s, 1H),
7.61 (d, J = 5.5 Hz, 1H), 7.57 – 7.53 (m,
2H), 7.36 – 7.31 (m, 1H), 7.17 – 7.12 (m, 1H), 7.06
(dd, J = 7.6, 7.4 Hz, 1H), 6.98 (dd, J = 7.6, 7.2 Hz, 1H), 3.04 – 3.00 (m, 3H), 2.62 (t, J = 7.6 Hz, 1H). 13C NMR (125 MHz, DMSO-d
6) δ 174.6, 168.9, 150.2, 143.4, 141.6,
141.5, 140.1, 136.3, 127.1, 127.0, 122.4, 122.3, 121.0, 121.0, 120.7,
118.4, 118.3, 118.3, 113.7, 113.5, 111.4, 35.2, 32.9, 20.6, 20.1.
HRMS calculated for [M + H]+ = 293.13969, found = 293.13947.
HPLC purity of 99.5%, RT = 4.581 min (λ = 254 nm).

(E)-N′-((1H-Indazol-5-yl)­methylene)-3-(1H-Indol-3-yl)­propanehydrazide
(14)
The title compound was obtained as a white
solid in 75% yield (MP: 245–247 °C). 1H NMR
(500 MHz, DMSO-d
6) δ 13.24, 13.21
(s, 1H), 11.30, 11.18 (s, 1H), 10.77 (s, 1H), 8.24, 8.09 (s, 1H),
8.13, 8.11 (s, 1H), 8.09 (s, 1H), 7.97, 7.92 (s, 1H), 7.81 (dd, J = 8.8, 1.3 Hz, 1H), 7.75 (dd, J = 8.8,
1.3 Hz, 1H), 7.60 – 7.53 (m, 2H), 7.33 (d, J = 7.8 Hz, 1H), 7.17, 7.13 (d, J = 2.1 Hz, 1H),
7.07 (dd, J = 7.5, 7.4 Hz, 1H), 6.99 (dd, J = 7.5, 7.4 Hz, 1H), 3.06 – 2.98 (m, 3H), 2.61 –
2.55 (m, 1H). 13C NMR (125 MHz, DMSO-d
6) δ 173.8, 168.2, 166.7, 166.1, 146.6, 143.4, 141.9,
136.3, 136.2, 134.4, 134.3, 127.1, 127.1, 127.1, 127.0, 123.4, 122.9,
122.3, 122.2, 121.6, 121.6, 120.9, 120.9, 118.3, 118.3, 118.2, 113.9,
113.6, 111.4, 110.8, 110.8, 40.0, 39.9, 39.7, 39.5, 39.4, 39.2, 39.0,
35.2, 33.0, 20.8, 20.1. HRMS calculated for [M-H]− = 330.13604, found = 330.13605. HPLC purity of 99.2%, RT = 10.170
min (λ = 254 nm).

(E)-N′-((7H-Pyrrolo­[2,3-d]­pyrimidin-5-yl)­methylene)-3-(1H-Indol-3-yl)­propanehydrazide (15)
The title compound was obtained as a white solid
in 61% yield (MP: 233–236 °C). 1H NMR (500
MHz, DMSO-d
6) δ 12.50, 12.47 (s,
1H), 11.33, 11.14 (s, 1H), 10.76 (s, 1H), 9.46, 9.31 (s, 1H), 8.84
(s, 1H), 8.30, 8.15 (s, 1H), 7.97, 7.94 (s, 1H), 7.55 (dd, J = 7.5, 5.0 Hz, 1H), 7.33 (dd, J = 8.0,
2.3 Hz, 1H), 7.17 – 7.10 (m, 1H), 7.08 – 7.03 (m, 1H),
7.00 – 6.92 (m, 1H), 3.06 (s, 2H), 3.02, 2.59 (t, J = 7.6 Hz, 2H). 13C NMR (125 MHz, DMSO-d
6) δ 174.0, 168.6, 152.6, 152.3, 150.7, 150.2, 141.7,
138.8, 136.5, 136.5, 131.0, 130.8, 127.3, 127.2, 122.5, 122.4, 121.3,
121.3, 118.6, 118.6, 118.5, 118.5, 114.2, 113.8, 111.7, 111.3, 111.1,
35.3, 33.2, 21.0, 20.3. HRMS calculated for [M+Na+]+ = 355.12778, found = 355.13794. HPLC purity of 97.0%, RT
= 3.403 min (λ = 254 nm).

(E)-N′-((1H-Pyrazol-4-yl)­methylene)-3-(1H-Indol-3-yl)­propanehydrazide
(16)
The title compound was obtained as a white
solid in 72% yield (MP: 210–213 °C). 1H NMR
(500 MHz, DMSO-d
6) δ 11.11, 10.98
(s, 1H), 10.75 (s, 1H), 8.08, 7.94 (s, 1H), 7.92 (s, 1H), 7.89 (s,
1H), 7.55 (t, J = 8.0 Hz, 1H), 7.33 (d, J = 8.0 Hz, 1H), 7.13, 7.10 (d, J = 1.8 Hz, 1H),
7.06 (t, J = 7.5 Hz, 1H), 6.97 (dd, J = 7.5, 7.4 Hz, 1H), 3.00 – 2.89 (m, 3H), 2.53 (t, J = 7.8 Hz, 1H). 13C NMR (125 MHz, DMSO-d
6) δ 173.7, 168.1, 140.1, 136.9, 136.4,
136.4, 127.2, 127.1, 122.4, 121.1, 121.1, 118.5, 118.4, 118.4, 118.3,
117.1, 117.1, 114.0, 113.7, 111.5, 35.2, 33.0, 20.9, 20.2. HRMS calculated
for [M+Cl–]− = 316.09796, found
= 316.09680. HPLC purity of 96.1%, RT = 2.744 min (λ = 254 nm).

(E)-N′-((1H-Pyrrolo­[2,3-b]­pyridin-3-yl)­methylene)-2-(1H-Indol-3-yl)­acetohydrazide (17)
The title compound was obtained as a white solid in 65%
yield (MP: 266–268 °C). 1H NMR (500 MHz, DMSO-d
6) δ 11.98 (s, 1H), 11.38, 11.00 (s, 1H),
10.86, 10.80 (s, 1H), 8.51, 8.44 (d, J = 7.8 Hz,
1H), 8.34, 8.26 (s, 1H), 8.26, 8.18 (s, 1H), 7.84 (d, J = 7.1 Hz, 1H), 7.60 (t, J = 8.5 Hz, 1H), 7.36,
7.33 (d, J = 8.5 Hz, 1H), 7.26, 7.20 (s, 1H), 7.17
– 7.10 (m, 1H), 7.06 (m, 1H), 6.98, 6.94 (dd, J = 7.8, 7.1 Hz, 1H), 4.09 (s, 1H), 3.63 (s, 1H). 13C NMR
(125 MHz, DMSO-d
6) δ 173.0, 167.6,
149.5, 149.5, 144.5, 143.7, 140.5, 136.6, 136.5, 130.8, 130.7, 130.6,
130.3, 127.8, 127.6, 124.4, 124.1, 121.6, 121.6, 119.2, 119.1, 119.0,
119.0, 117.4, 117.1, 117.0, 111.9, 110.9, 110.9, 108.9, 108.8, 32.1,
29.6. HRMS calculated for [M-H]− = 316.12039, found
= 316.12021. HPLC purity of 95.2%, RT = 4.210 min (λ = 254 nm).

(E)-N′-((1H-Indazol-5-yl)­methylene)-2-(1H-Indol-3-yl)­acetohydrazide
(18)
The title compound was obtained as a white
solid in 59% yield (MP: 266–270 °C). 1H NMR
(500 MHz, DMSO-d
6) δ 13.24 (s, 1H),
11.46, 11.20 (s, 1H), 10.92, 10.87 (s, 1H), 8.32, 8.11 (s, 1H), 8.13
(s, 1H), 7.97 (s, 1H), 7.88, 7.80 (d, J = 8.8 Hz,
1H), 7.60, 7.57 (d, J = 8.5 Hz, 2H), 7.38 –
7.32 (m, 1H), 7.25 (d, J = 6.2 Hz, 1H), 7.10 –
7.04 (m, 1H), 7.01 – 7.94 (m, 1H), 4.08 (s, 1H), 3.64 (s, 1H). 13C NMR (125 MHz, DMSO-d
6) δ
172.5, 167.0, 147.0, 143.4, 140.5, 140.4, 136.1, 136.0, 134.4, 127.5,
127.2, 127.1, 123.9, 123.7, 123.6, 122.9, 122.9, 121.6, 121.2, 121.0,
120.9, 118.8, 118.7, 118.4, 118.3, 111.4, 111.3, 110.9, 110.8, 108.3,
31.7, 29.1. HRMS calculated for [M + H]+ = 318.13494, found
= 318.13477. HPLC purity of 99.8%, RT = 3.624 min (λ = 254 nm).

(E)-N′-((7H-Pyrrolo­[2,3-d]­pyrimidin-5-yl)­methylene)-2-(1H-Indol-3-yl)­acetohydrazide (19)
The title compound was obtained as a white solid
in 60% yield (MP: 283–285 °C). 1H NMR (500
MHz, DMSO-d
6) δ 12.48, 12.46 (s,
1H), 11.46, 11.20 (s, 1H), 10.90, 10.84 (s, 1H), 9.44, 9.40 (s, 1H),
8.82 (d, J = 9.0 Hz, 1H), 8.38, 8.19 (s, 1H), 7.99
(dd, J = 7.2, 2.0 Hz, 1H), 7.61, 7.57 (d, J = 7.9 Hz, 1H), 7.34 (dd, J = 9.0, 8.6
Hz, 1H), 7.26, 7.21 (d, J = 1.5 Hz, 1H), 7.09 –
7.03 (m, 1H), 6.99, 6.93 (dd, J = 7.5, 7.2 Hz, 1H),
4.12, 3.64 (s, 2H). 13C NMR (DMSO) δ 172.5, 167.1,
152.5, 152.5, 152.4, 150.8, 150.3, 141.9, 138.6, 136.3, 136.2, 130.8,
130.7, 127.6, 127.3, 124.1, 123.9, 121.2, 121.1, 118.9, 118.8, 118.6,
118.6, 115.2, 111.5, 111.1, 111.0, 108.4, 31.8, 29.2. HRMS calculated
for [M + H]+ = 319.13019, found = 319.13034. HPLC purity
of 95.9%, RT = 4.973 min (λ = 254 nm).

(E)-N′-((1H-Pyrazol-4-yl)­methylene)-2-(1H-Indol-3-yl)­acetohydrazide
(20)
The title compound was obtained as a white
solid in 70% yield (MP: 220–223 °C). 1H NMR
(500 MHz, DMSO-d
6) δ 11.35, 10.94
(s, 1H), 10.86, 10.81 (s, 1H), 8.14, 7.95 (s, 1H), 7.93 (s, 2H), 7.59
– 7.55 (m, 1H), 7.36 – 7.30 (m, 1H), 7.24 – 7.18
(m, 1H), 7.09 – 7.03 (m, 1H), 7.01 – 6.91 (m, 1H), 3.97,
3.58 (s, 2H). 13C NMR (125 MHz, DMSO-d
6) δ 172.9, 167.5, 140.8, 137.1, 136.4, 136.3, 127.7,
127.4, 124.4, 124.3, 121.5, 121.3, 119.1, 119.0, 118.9, 118.7, 117.4,
117.2, 111.8, 111.7, 108.6, 31.8, 29.4. HRMS calculated for [M + H]+ = 268.11929, found = 268.11914. HPLC purity of 97.3%, RT
= 5.100 min (λ = 254 nm).

(E)-N′-((1H-Indazol-5-yl)­methylene)-3,4,5-Trimethoxybenzohydrazide
(21)
The title compound was obtained as a white
solid in 61% yield (MP: 232–235 °C). 1H NMR
(500 MHz, DMSO-d
6) δ 13.28 (s, 1H),
11.66 (s, 1H), 8.56 (s, 1H), 8.16 (s, 1H), 8.03 (s, 1H), 7.89 (d, J = 8.6 Hz, 1H), 7.62 (d, J = 8.6 Hz, 1H),
7.25 (s, 2H), 3.87 (s, 6H), 3.73 (s, 3H). 13C NMR (125
MHz, DMSO-d
6) δ 162.5, 152.8, 148.7,
140.6, 140.4, 134.5, 128.7, 127.1, 123.8, 123.0, 121.9, 111.0, 105.2,
60.2. HRMS calculated for [M + H]+ = 355.14008, found =
355.13974. HPLC purity of 99.8%, RT = 3.801 min (λ = 254 nm).

(E)-N′-((1H-Pyrrolo­[2,3-b]­pyridin-3-yl)­methylene)-3,4,5-Trimethoxybenzohydrazide (22)
The title compound was obtained as a white solid in 64%
yield (MP: 254–257 °C). 1H NMR (500 MHz, DMSO-d
6) δ 12.07 (s, 1H), 11.54 (s, 1H), 8.62
– 8.55 (m, 2H), 8.30 (d, J = 3.2 Hz, 1H),
7.93 (s, 1H), 7.22 (s, 3H), 3.85 (s, 6H), 3.71 (s, 3H). 13C NMR (125 MHz, DMSO-d
6) δ 182.1,
162.7, 153.1, 153.0, 149.5, 145.0, 144.4, 140.5, 130.7, 130.6, 129.2,
117.3, 117.0, 110.9, 105.3, 60.5. HRMS calculated for [M + H]+ = 355.14008, found = 355.13968. HPLC purity of 98.8%, RT
= 3.925 min (λ = 254 nm).

(E)-N′-((1H-Indazol-5-yl)­methylene)-2-(3,4,5-Trimethoxyphenyl)­acetohydrazide
(23)
The title compound was obtained as a white
solid in 65% yield (MP: 211–213 °C). 1H NMR
(500 MHz, DMSO-d
6) δ 13.25 (s, 1H),
11.49, 11.25 (s, 1H), 8.29, 8.09 (s, 1H), 8.13 (s, 1H), 7.99 (s, 1H),
7.89, 7.81 (d, J = 8.8 Hz, 1H), 7.58 (dd, J = 7.6, 7.5 Hz, 1H), 6.66, 6.63 (s, 2H), 3.92, 3.62 (s,
2H), 3.76 (s, 3H), 3.70, 3.60 (s, 6H). 13C NMR (125 MHz,
DMSO-d
6) δ 172.2, 152.7, 147.5,
143.8, 134.5, 131.5, 123.6, 123.0, 121.9, 121.4, 111.0, 106.8, 106.5,
60.1, 56.0, 55.8. HRMS calculated for [M + H]+ = 369.15573,
found = 369.15558. HPLC purity of 96.3%, RT = 2.709 min (λ =
254 nm).

(E)-N′-((1H-Pyrrolo­[2,3-b]­pyridin-3-yl)­methylene)-2-(3,4,5-Trimethoxyphenyl)­acetohydrazide
(24)
The title compound was obtained as a white
solid in 68% yield (MP: 253–256 °C). 1H NMR
(500 MHz, DMSO-d
6) δ 12.05 (s, 1H),
11.28, 11.13 (s, 1H), 8.51, 8.47 (d, J = 7.8 Hz,
1H), 8.35, 8.16 (s, 1H), 8.31 – 8.27 (m, 1H), 7.92 (d, J = 2.4 Hz, 1H), 7.22 – 7.16 (m, 1H), 6.67, 6.65
(s, 2H), 3.96, 3.45 (s, 2H), 3.77, 3.63 (s, 3H), 3.65, 3.60 (s, 6H). 13C NMR (125 MHz, DMSO-d
6) δ
171.5, 165.8, 152.7, 152.6, 149.4, 149.3, 144.0, 143.2, 140.0, 136.0,
131.6, 130.4, 130.3, 130.0, 129.7, 116.9, 116.6, 116.4, 110.5, 110.4,
106.7, 106.4, 60.0, 60.0, 55.8, 55.7. HRMS calculated for [M + H]+ = 369.15573, found = 369.15442. HPLC purity of 99.0%, RT
= 3.550 min (λ = 254 nm).

SBVS, De Novo Design Campaign and Chemical Space Analysis
The SBVS procedure using the LASSBio Chemical Library was done as previously reported. For the design of the virtual compound library
used in the De Novo design campaign, a workflow was constructed in
KNIME. Our internal collection of reagents and fragments was first
imported and converted from SMILES strings to RDKit molecules using
the “RDKit From Molecule” node. Next, only those containing
carboxylic acid functional groups were selected using the “RDKit
Functional Group Filter” node. The resulting molecules were
then filtered to remove salts and duplicates. Subsequently, the filtered
structures were combined with hydrazine to generate the corresponding
virtual hydrazides using the “RDKit Two Component Reaction”
node with the SMARTS pattern “[#6:1](=[#8:2])­[O:3].[#7:4][#7:5]≫[#6:1](=[#8:2])[#7:4][#7:5]”.
The resulting hydrazides were then reacted with the aldehydes of interestisoquinoline-5-carbaldehyde (6), quinoline-4-carbaldehyde (7), 4-(pyridin-4-yl)­benzaldehyde (8), isonicotinaldehyde (9), 1H-pyrrolo­[2,3-b]­pyridine-5-carbaldehyde (10), and 1H-pyrrolo­[2,3-b]­pyridine-3-carbaldehyde (11)again using the “RDKit Two Component Reaction”
node with the SMARTS pattern “[#6:1]­(=O)­[N:2]­[N:3].[#6:4]­[C:5]=O≫[#6:1]­(=O)­[N:2]/N
= [C:5]/[#6:4]”. Descriptors were calculated for the resulting
hydrazones using the “RDKit Descriptor Calculation”
node, and the molecules were filtered based on fraction sp3 (≥0.1) and number of stereogenic centers (= 0). Molecular
fingerprints (ECFP4, 1024 bits) were then computed using the “RDKit
Fingerprint” node. The resulting bit vectors were expanded
and standardized by Z-score normalization, followed by principal component
analysis (PCA) to reduce the chemical space to two dimensions. A min–max
normalization (0.0–1.0) was subsequently applied. Finally,
the “2D/3D Scatterplot” node was used
to visualize the chemical space, with compounds color-coded according
to their fraction sp3 values. The small molecules were
then processed by adding hydrogen atoms to their structures using
the “RDKit Add Hs” node, followed by 3D coordinate generation
with “RDKit Generate Coords.” The resulting geometries
were optimized using the MMFF94 force field via the “RDKit
Optimize Geometry” node. The optimized structures were exported
in SDF format and subsequently reoptimized in Spartan’24 using
the semiempirical PM6 method. The ionization states of all molecules
were manually verified prior to performing the molecular docking studies
against ROCK, as previously described.

Docking and MD Simulations of 18 and 21
Docking analysis
of the small molecules into ROCK1 (PDB: 6E9W) and ROCK2
(PDB: 7JNT) was carried out following the procedure previously
reported. For the docking studies in
ROCK2 (PDB: 8GDS), the software GOLD 2025.1.0 was employed.
Hydrogen atoms were added to the protein structure, and the cocrystallized
ligand was extracted and used as a reference for defining the binding
site, which included all amino acid residues within 8 Å of the
ligand. Docking calculations were performed in the absence of water
molecules using the GoldScore scoring function and the genetic algorithm
with default parameters. A pharmacophore constraint was applied, consisting
of two pharmacophore points: (i) Hydrogen bond acceptorradius:
0.7 Å, constraint weight: 10; coordinates: x = 34.7400, y = 4.8230, z = −70.2390.
(ii) Hydrogen bond donorradius: 1.0 Å, constraint weight:
10; coordinates: x = 32.8370, y =
5.3580, z = −71.2780. MD simulations were
performed using the FlarePro+ software (v10.0.1, Cresset, Litlington,
Cambridgeshire, UK). Prior to MD, missing loops and gaps in the protein
structures were reconstructed using the loop modeling tool implemented
in FlarePro+, and the N- and C-terminal
residues were capped with acetyl (ACE) and N-methylamide
(NME) groups, respectively, to avoid artificial terminal charges.
The resulting protein–ligand complexes were subsequently subjected
to MD simulations using FlarePro+ with the OpenMM engine. MD simulations
were performed employing the Open Force Field 2.2.0 for the ligands
and the AMBER force field for the proteins. The systems were solvated
using the TIP3P explicit water model within a truncated octahedral
box, and AM1-BCC charges were assigned to the ligands. Following energy
minimization, the systems underwent 200 ps of equilibration, followed
by 100 ns of production simulation using a 4 fs integration time step
under standard temperature and pressure conditions. Upon completion,
the simulation trajectories were analyzed for RMSD and RMSF variations,
as well as protein–ligand interaction profiles throughout the
trajectories.

Biochemical Assays for Kinase Inhibition
All kinase
inhibition assays were performed at Reaction Biology Corporation (Malvern,
PA, USA) using the radiometric HotSpot kinase assay platform. IC50 values were determined by curve fitting in the 10-dose IC50 mode, with assays conducted in duplicate using a 3-fold
serial dilution factor. Assays were performed at a fixed ATP concentration
of 1 μM, which is below the reported Km­(ATP) for ROCK kinases
(∼10 μM). This sub-Km ATP condition was intentionally
selected so that the measured IC50 values more closely
approximate K
i values for ATP-competitive
inhibition, consistent with the Cheng–Prusoff relationship K

i
= IC
50/(1+[ATP]/Km
). Because K
i is less dependent on the
ATP concentration and is often more directly comparable to theoretical/computational
predictions, we preferred using ATP concentrations below Km in the
biochemical assays.

Cell Line and Cell Culture
The tumor cell line used
in this study was human triple-negative breast adenocarcinoma MDA-MB-231,
which lacks estrogen receptor (ER) and progesterone receptor (PR)
expression, and amplification of human epidermal growth factor receptor
2 (HER2), and has both ROCK isoform overexpression, where ROCK2 is
higher than ROCK1. Cell wells cultured
with Dulbecco’s Modified Eagle’s Medium – High
glucose (DMEM medium) from Sigma-Aldrich, cat D1152, supplemented
with 10% of Fetal bovine serum (FBS) from Gibco, cat 12657–029,
in a 75 cm2 culture flask and maintained at 37 °C
with CO2 5% (Thermo Scientific, Series 8000 WJ, CO2 incubator). Immortalized cells were obtained from Rio de
Janeiro Cell Bank (BCRJ).

Cell Viability Assay
The 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl-tetrazolium
bromide (MTT) colorimetric assay
,
was performed
at 30 μM final concentration per well, with 24 h incubation
time. First, were seeded in 96-well plates, MDA-MB-231 cells at 2
× 105 cells/mL concentration with supplemented medium
and incubated. After 24 h, the controls and compounds that were previously
dissolved in DMSO were added, considering a final concentration of
DMSO 1% per well. After 24 h, the plate was centrifuged at 440 ×
g for 10 min at 4 °C (Universal centrifuge 320R Hettich), and
110 μL were discarded to add 10 μL of MTT (5 mg/mL in
PBS) to each well. Then, they were incubated for 3.5 h at CO2 incubator. After that, 100 μL of SDS-HCl was added to dissolve
the crystals of formazan, the plate was protected from light and placed
for a minimum 24 h at room temperature. Finally, the absorption was
read with a Molecular Devices Spectramax M5 plate reader at 595 nm.
The data from 3 independent experiments tested in triplicate were
analyzed using GraphPad Prism 8.0, where data are represented as mean
± SD (95% confidence interval).

Cell-Based Scratch Assay
To evaluate the capacity of
migration of the MDA-MB-231 cells after treatments with the controls
and compounds of interest, the cell-based scratch assay was performed.
,
First, 6 × 105 cell/mL were seeded using 24-well
plates with supplement DMEM medium for 24 h to reach 90 – 100%
confluence. A scratch was realized using a 200 μL tip in each
well, followed by 3× washed with PBS to remove nonadherent cells.
After that, the compounds in final concentration per well (0.1 μM,
10 μM, 30 μM, and 100 μM with DMSO 1%) were added
to pure DMEM medium with Arabinose (AraC) at a final concentration
of 10–5 M (antimitotic reagent to inhibit proliferation).
Images at 0 and 24 h were taken with a Microscope Cam 5 (Prime Cam)
camera attached to a Nikon Eclipse TS2 microscope with a 4× objective
lens. The scratch area was analyzed using image J and quantified using
GraphPad Prism 8.0, from 3 different experiments tested in duplicate,
where data are represented as mean ± SD (95% confidence interval).

PAMPA-GIT Assays
Stock solutions (10 mM) of each test
and control compound were prepared in DMSO. An aliquot of 50 μL
of each solution was diluted with 4950 μL of PBS (10 mM, pH
6.6) in a 5 mL glass vial, filtered through a 0.45 μm PVDF membrane,
and reserved. Acceptor plate wells were filled with 180 μL of
PBS (pH 7.4):DMSO (95:5, v/v), while donor plate wells received 5
μL of a soybean L-α-phosphatidylcholine solution (20 mg/mL
in dodecane). After 5 min, 180 μL of the compound solution was
added in triplicate to the donor plate wells. The donor plate was
then carefully placed over the acceptor plate to form a “sandwich”
system, which was incubated under gentle agitation (50 rpm) for 8
h at room temperature (≈25 °C) in a closed container containing
10 mL of PBS (pH 7.4). Following incubation, the donor plate was removed,
and the contents of the acceptor plate were transferred to a UV microplate
for absorbance measurement using a SpectraMax reader (Molecular Devices)
at the wavelength predetermined for each compound. A blank was prepared
using 180 μL of PBS (pH 7.4):DMSO (95:5, v/v). The permeability
result for PAMPA-GIT classifies compounds according to the percentage
of absorbed fraction (Fa%), as high permeability (70–100%),
medium permeability (30–69%), or low permeability.
,

PAMPA-BBB Assays
Each test and control compound (1
mg) was dissolved in 1 mL of ethanol. In a 5 mL glass vial, 1400 μL
of ethanol and 3.5 mL of PBS (pH 7.4) were combined, followed by the
addition of 100 μL of the ethanolic compound solution. The resulting
mixture was filtered through a 0.45 μm PVDF membrane and reserved.
Acceptor plate wells were filled with 180 μL of PBS (pH 7.4):ethanol
(70:30, v/v), while donor plate wells received 5 μL of a porcine
brain lipid solution (20 mg/mL in dodecane). After 5 min, 180 μL
of the compound solution was added in triplicate to the donor plate
wells. The donor plate was then carefully placed over the acceptor
plate to form a “sandwich” system, which was kept undisturbed
for 2 h and 45 min at room temperature (≈25 °C) in a closed
container containing 10 mL of PBS (pH 7.4). Following incubation,
the donor plate was removed, and the contents of the acceptor plate
were transferred to a UV microplate for absorbance measurement using
a SpectraMax microplate reader (Molecular Devices) at the wavelength
predetermined for each compound. A blank was prepared using 180 μL
of PBS (pH 7.4):ethanol (70:30, v/v). The ranges used to predict the
permeability of the controls used and the compounds evaluated are
a) CNS+ (prediction of high permeability through the BBB) Pe (10–6
cm·s–1) ≥ 4.0; b) CNS- (prediction of
low permeability through the BBB) Pe (10–6 cm·s–1) < 2.0; c) CNS ± (method uncertainty range) 2.0 ≤
Pe (10–6 cm·s–1) ≤ 4.0.

Supplementary Material