An Ongoing Search for Multitarget Ligands as Potential Agents for Diabetes Mellitus and Its Long-Term Complications: New Insights into (5-Arylidene-4-oxothiazolidin-3-yl)alkanoic Acid Derivatives

Abstract

Background: Diabetes mellitus is a multifactorial disease characterized by complex metabolic dysfunctions and chronic complications induced by hyperglycaemia. The design of multitarget ligands, capable of simultaneously controlling different pathogenic processes, was proposed as a promising approach to identify novel antidiabetic drugs endowed with improved efficacy. Methods: (5-Arylidene-4-oxothiazolidin-3-yl)alkanoic acid derivatives 1a–g and 2a–g were synthesized as potential multitarget antidiabetic agents. They were tested in vitro as inhibitors of both human recombinant AKR1B1 and PTP1B, and kinetic studies and molecular docking simulations for both enzymes were performed. Their effects on cellular glucose uptake, insulin signalling, and mitochondrial potential were assayed in cultures of murine C2C12 myocytes. A lipid accumulation assay was performed in HepG2 liver cells. The effects on high glucose-induced sorbitol accumulation were evaluated in lens HLE and retinal MIO-M1 cells. Results: All compounds displayed excellent AKR1B1 inhibitory activity (IC₅₀ 0.03–0.46 μM 1a–g; IC₅₀ 0.48–6.30 μM 2a–g); 1g and 2e–g also appreciably inhibited PTP1B at micromolar concentrations. Propanoic derivatives 2e–g significantly stimulated glucose uptake in C2C12 myocytes, in an insulin-independent way, reduced lipid accumulation in HepG2 liver cells, and caused hyperpolarization of C2C12 mitochondria at 10 μM concentration. Derivative 2e significantly reduced sorbitol accumulation in both HLE and MIO-M1 cells at a 5 μM concentration. Conclusions: The results reported here provided new insights into the mechanisms of action and structure/activity relationships of 4-thiazolidinone derivatives, underscoring the capability of compounds 2e–g of eliciting insulin-mimetic effects independent of hormone signalling. Among them, compound 2e also proved to inhibit AKR1B1-dependent sorbitol accumulation and, thus, emerged as a promising multitarget agent that can be considered for further investigations.

1. Introduction

The development of several serious pathologies, such as diabetes, cancer, and cardiovascular and neurodegenerative diseases, involves complex etiological processes, simultaneously including diverse pathogenic alterations mediated by multiple targets. For this reason, the management of these diseases by single-targeted drug therapies is often unsatisfactory, and, therefore, drug combinations are required to improve clinical efficacy by controlling different mechanisms implicated in the course of the disease. However, combination therapies may give rise to complications regarding drug–drug interactions, pharmacokinetics, toxicity, and patient compliance. Therefore, in recent years, multitargeted ligands have attracted considerable attention for developing possible alternatives to drug combinations in the hope that they could provide enhanced efficacy, lower risk of adverse reactions, technological advantages, and improved adherence to long-term treatments [1,2,3,4]. In this context, designed multiple ligands (DMLs) are molecules rationally designed to be simultaneously directed at two or more biological targets, which are selected because of their proven implications in the pathogenesis of a certain disease [1]. The discovery of DMLs is generally based on well-established knowledge of the structures of the selected targets and/or of the pharmacophoric features of their ligands; in fact, a feasible medicinal chemistry approach consists of merging the pharmacophoric moieties of ligands directed to each target in a single “hybrid” molecule [1,3,5]. The subsequently performed optimization involves several challenging steps, in which different strategies may be exploited with the aim of balancing the effectiveness of the DML towards the selected biomolecules and minimizing its undesired activities towards irrelevant targets, in addition to modulating its drug-like properties [1,3].

Among the multifactorial diseases that could benefit from the availability of DMLs as novel drugs, diabetes mellitus (DM) represents one of the most serious global health threats. Currently, about 590 million adults are estimated to suffer from DM, and their number is predicted to further increase to more than 850 million by 2050. Type 2 DM (T2DM) accounts for more than 96% of DM cases worldwide, with an alarming increase not only among adults but also among adolescents [6,7,8]. DM is a chronic disease characterized by insulin resistance and/or insufficient insulin secretion, which are responsible for hyperglycaemia. This latter can compromise various cellular functions, thus causing tissue damage and promoting the development of severe complications, such as retinopathy, nephropathy, neuropathies, and cardiovascular pathologies. For these reasons, DM is included among the leading causes of death and disability worldwide, and growing attention is required to improve the management of this disease as well as the prevention of its complications.

In the treatment of DM, the main therapeutic aim is the control of hyperglycaemia, which, however, represents a challenging goal and, in the case of T2DM, often requires drug combinations. On this basis, the search for DMLs that can be directed to control multiple dysfunctions implicated in the development of T2DM is considered an attractive approach to achieve novel therapeutic options.

In the last few years, we have focused on the search for new DMLs as potential antidiabetic agents by investigating 4-thiazolidinone derivatives targeting both protein tyrosine phosphatase 1B (PTP1B) and aldose reductase (AKR1B1), which are enzymes critically implicated in specific signalling alterations underlying the development of T2DM and its chronic complications [9,10,11].

PTP1B exerts pivotal functions in the control of insulin signalling, mainly via the dephosphorylation of specific phosphotyrosine residues of the insulin receptor. Alterations of the action or expression of PTP1B were shown to be critically implicated in the development of T2DM by promoting and sustaining insulin resistance in both the central nervous system and peripheral tissues [12,13]. PTP1B overexpression was also found to be responsible for resistance to leptin, a hormone produced by adipocytes that exerts anorexigenic effects in the hypothalamus and promotes energy expenditure [14,15]. Both insulin and leptin play pivotal roles in the central nervous system by regulating glucidic and energetic homeostasis; their coordinated actions in the hypothalamus are also crucial for metabolism in peripheral tissues [16]. PTP1B inhibition proved to be an effective strategy to improve the cellular signalling of both hormones and, therefore, the search for inhibitors of this enzyme could pave the way for the development of novel therapeutic agents for the management of T2DM and its associated pathologies, such as obesity and metabolic syndrome [17,18,19].

AKR1B1 is crucially involved in the etiopathology of diabetic long-term complications, since it catalyzes the first step of the polyol pathway, i.e., the NADPH-dependent reduction of glucose to sorbitol. Under hyperglycaemic conditions, the markedly increased AKR1B1-mediated metabolism of glucose through the polyol pathway promotes oxidative stress and inflammatory signalling, consequently triggering cellular dysfunctions and tissue damage, which are major causes of the onset and progression of chronic diabetic complications [20,21]. In fact, the inhibition of AKR1B1 proved to be an effective strategy to control hyperglycaemia-induced damage and, thus, to prevent or delay the development of DM-associated complications [20,21,22,23,24]. Among the numerous AKR1B1 inhibitors (ARIs) explored so far, the 4-oxo-2-thioxothiazolidine derivative epalrestat (Figure 1) is marketed in some Asian countries, such as Japan, for treating diabetic neuropathy [22].

Figure 1.

Design and structures of (5-arylidene-4-oxothiazolidin-3-yl)alkanoic acids 1a–g and 2a–g.

Although PTP1B and AKR1B1 belong to different enzyme families and play different roles, some shared structural features can be observed in diverse ligands of these enzymes, thus suggesting that dual inhibitors could be designed. In fact, (5-arylidene-4-oxothiazolidin-3-yl)acetic acid derivatives (I, Figure 1), which were shown to act as potent ARIs [25], and 4-[(5-arylidene-4-oxothiazolidin-3-yl)methyl]benzoic acid derivatives (II, Figure 1), which exhibited significant PTP1B inhibitory properties [26], share a pharmacophoric acidic moiety (in position 3 of the thiazolidinone scaffold), and a hydrophobic portion (on C-5 of the heterocycle). On this basis, according to a knowledge-based multitarget approach, we recently obtained dual AKR1B1/PTP1B inhibitors (III, Figure 1) by merging the pharmacophores of 4-thiazolidinones of series I and II. Among derivative III, several compounds proved to act as dual AKR1B1/PTP1B inhibitors at low micromolar or submicromolar concentrations [9,10,11]. Interestingly, some of them also displayed significant cellular activities, such as insulin-sensitizing/mimetic effects in cultured C2C12 myoblasts as well as the capability of reducing intracellular sorbitol content in cultured human lens epithelial cells [10,11].

Pursuing this ongoing search for DMLs as potential antidiabetic agents, here we report a new series of (5-arylidene-4-oxothiazolidin-3-yl)alkanoic acids (1a–g and 2a–g, Figure 1), synthesized with the aim of acquiring further insights into the role played by the 5-arylidene moiety on the activity profile of this class of DMLs. Starting from previously acquired data, which indicated that an extension of this portion may be beneficial for tuning the dual AKR1B1/PTP1B inhibitory activity, in compounds 1a–g and 2a–g the 5-arylidene moiety was modified by inserting one or two methoxy substituents on the distal phenyl ring and/or elongating the linker chain between the two aromatic rings. These structural modifications led to the identification of compounds capable of eliciting cellular insulin-mimetic effects independent of the signal transduction of the hormone and, simultaneously, to strongly inhibit AKR1B1.

2. Results and Discussion

2.1. Chemistry

(5-Arylidene-4-oxo-2-thioxothiazolidin-3-yl)alkanoic acids 1a–g, 2a–g were prepared by means of a convenient two-step synthesis, as depicted in Scheme 1. The reaction of O-alkylation of 3-hydroxybenzaldehyde or 4-hydroxybenzaldehyde with the appropriate arylyalkyl bromide, in the presence of potassium carbonate, produced arylalkoxy-substituted benzaldehydes 3a–g. Then, the Knoevenagel condensation of aldehydes 3a–g with (4-oxo-2-thioxothiazolidin-3-yl)acetic acid (4) or 3-(4-oxo-2-thioxothiazolidin-3-yl)propanoic acid (5), carried out in refluxing glacial acetic acid, in the presence of sodium acetate, provided compounds 1a–g and 2a–g, respectively (Scheme 1).

Scheme 1.

Synthesis of (5-arylidene-4-oxo-2-thioxothiazolidin-3-yl)acetic acids 1a–g and 3-(5-arylidene-4-oxo-2-thioxothiazolidin-3-yl)propanoic acids 2a–g. Reagents and conditions: (a) arylalkyl bromide, K₂CO₃, anhydrous DMF, 50–70 °C; (b) CH₃COOH, CH₃COONa, Δ.

The structures of compounds 1a–g and 2a–g were assigned on the basis of analytical data and NMR spectroscopy (see Section 3 and Figures S1–S28). Besides standard ¹H and ¹³C NMR 1D spectra, experiments with 2D techniques, such as ¹H homocorrelated COSY and ¹H-¹³C heterocorrelated gHSQCAD, were also performed to accomplish the unambiguous assignment of signals.

In the ¹H NMR spectra, a singlet at 4.52–4.74 ppm, attributable to the methylene group on N-3 of acetic acid derivatives 1a–g, or two coupled triplets at 2.61–2.64 ppm and 4.12–4.23 ppm, due to the resonance of the propanoic chain on N-3 of derivatives 2a–g, were diagnostic. The carbon atoms of these methylene groups gave rise to signals in the range between 42.5 ppm and 47.3 ppm for acetic acid derivatives 1 and between 31.4 ppm and 69.6 ppm for the propanoic analogues 2. Moreover, in the ¹³C NMR spectra of all synthesized compounds 1a–g, 2a–g, other diagnostic signals were due to the resonance of the carbonyl and thiocarbonyl groups of the (4-oxo-2-thioxothiazolidin-3-yl)alkanoic scaffold, in particular, a singlet at 193.6–193.8 ppm, attributable to the thiocarbonyl group in position 2 of the heterocyclic core, and two different singlets in the range between 166.8 ppm and 172.3 ppm, attributable to the carbonyl group in position 4 of the thiazolidinone and the carboxylic group of the alkanoic chain.

All derivatives 1, 2 were obtained only as Z isomers; in fact, their ¹H NMR and ¹³C NMR spectra showed only one set of signals, including a diagnostic singlet in the range 7.69–7.86 ppm of ¹H-NMR spectra, which can be attributed to the resonance of the 5-methylidene proton.

2.2. In Vitro Enzyme Inhibition

The in vitro inhibitory activity of compounds 1a–g and 2a–g was evaluated against both human recombinant PTP1B, by using p-nitrophenylphosphate as substrate, and human recombinant AKR1B1, by using L-idose as substrate. Sodium metavanadate and sorbinil were the respective reference drugs.

Compounds 1a–g and 2a–g behaved as potent ARIs, showing excellent inhibitory ability against AKR1B1 with nanomolar or low micromolar IC₅₀ values (Table 1). Acetic acid derivatives 1a–g showed IC₅₀ values lower than that of sorbinil, in the range between 0.03 μM (1a, 1g) and 0.46 μM (1e). When the acetic chain on N-3 was replaced by the propanoic acid residue (compounds 2a–g), less marked AKR1B1 inhibitory ability was observed (Table 1), with a decrement ranging from very slight (2e vs. 1e 1.5-fold) to significant (2b vs. 1b almost 78-fold); however, all compounds 2a–g produced appreciable AKR1B1 inhibition, with IC₅₀ values between 0.48 μM and 6.30 μM.

Table 1.

In vitro inhibition of human AKR1B1 and PTP1B by compounds 1a–g and 2a–g.

Compd.	AKR1B1 IC50 (μM) a	PTP1B IC50 (μM) a
1a	0.03 ± 0.01	80.5 ± 1.8
1b	0.04 ± 0.01	91.4 ± 1.4
1c	0.08 ± 0.01	50.8 ± 2.8
1d	0.12 ± 0.02	100.0 ± 0.3
1e	0.46 ± 0.09	>100
1f	0.09 ± 0.01	43.8 ± 1.0
1g	0.03 ± 0.01	27.7 ± 1.3
2a	0.48 ± 0.13	58.0 ± 6.0
2b	3.11 ± 0.73	48.2 ± 4.0
2c	3.73 ± 0.63	59.7 ± 4.4
2d	6.30 ± 1.65	>100
2e	0.69 ± 0.13	12.1 ± 0.2
2f	1.18 ± 0.21	23.9 ± 0.7
2g	1.89 ± 0.24	24.6 ± 0.7
Sorbinil	0.99 ± 0.13	---
Vanadate	---	0.4 ± 0.01

^a Values reported in the table represent IC₅₀ ± SE.

Among derivatives 1a–e, which are methoxy-substituted on the distal phenyl ring of the 5-arylidene moiety, 4-[2-(4-methoxyphenyl)ethoxy]benzylidene-substituted compound 1a was the most potent AKR1B1 inhibitor (IC₅₀ = 0.03 μM); its isomers 1b and 1c were slightly less active (1.3-fold and 2.7-fold, respectively). Moreover, 3,4-dimethoxy-substituted analogue 1d was shown to be 4-fold and 1.5-fold less effective than 1a and 1c, respectively, thus indicating that a methoxy group in the meta position of the distal phenyl ring is less tolerated than in the para position. Moreover, the insertion of a methylene in the linker chain was detrimental, leading to a 3-arylpropoxy derivative, 1e, which was 15-fold less active than its 2-arylethoxy counterpart 1a (Table 1).

A similar trend was observed in propanoic acid analogues 2a–e, with 3,4-dimethoxy substituted analogue 2d, which proved to be the least effective (IC₅₀ = 6.30 μM) among all AKR1B1 inhibitors 1, 2 (Table 1). However, in series 2, it is worth noting that the 4-[3-(4-methoxyphenyl)propoxy]benzylidene-substituted derivative 2e displayed significant AKR1B1 inhibitory activity (IC₅₀ = 0.69 μM), similar to that of 2-(4-methoxyphenyl)ethoxy analogue 2a (IC₅₀ = 0.48 μM).

Moreover, phenoxybutoxy-substituted analogues 1f, 1g, and 2f proved to be from 2- to 51-fold more potent AKR1B1 inhibitors than their phenylbutoxy counterparts, which we investigated previously [11].

Regarding PTP1B, acetic acid derivatives 1a–g were found to be moderately or scarcely active toward the enzyme; compound 1g showed the most appreciable inhibitory effect with an IC₅₀ value of 27.7 μM (Table 1). Among propanoic acid analogues 2, compounds 2e–g were the most interesting PTP1B inhibitors, showing IC₅₀ values in the range between 12.1 μM and 24.6 μM (Table 1). Comparing the PTP1B inhibitory ability of compounds 1, 2 with that of the previously reported analogues [9,10,11] provided further evidence that the 5-arylidene moiety plays a crucial role in determining the PTP1B inhibitory effectiveness of (5-arylidene-4-oxothiazolidin-3-yl)alkanoic acid derivatives. In fact, the insertion of one or two methoxy groups on the distal phenyl ring of the phenylethoxy tail was detrimental, leading to analogues with moderate or marginal effects on PTP1B activity in both acetic and propanoic acid series 1a–d and 2a–d. On the other hand, the 4-[3-(4-methoxyphenyl)propyloxy]benzylidene derivative 2e showed a noteworthy PTP1B inhibitory activity (IC₅₀ = 12.1 μM), also proving to be the most potent PTP1B inhibitor among all compounds 1, 2; its activity toward PTP1B was appreciably higher than that of the previously reported 3-phenylpropyloxy analogue, which was almost inactive at a concentration of 5 μM [11]. Overall, comparing our current and previous data relative to both (5-arylidene-4-oxo-2-thioxothiazolidin-3-yl)acetic acid and 3-(5-arylidene-4-oxo-2-thioxothiazolidin-3-yl)propanoic acid series showed that the elongation of the linker connecting the two phenyl rings of the 5-arylidene moiety can, in general, improve PTP1B inhibitory effectiveness, but also suggests the existence of a cut-off which might correspond with the length of a pentatomic chain.

Out of the newly synthesized alkanoic acid derivatives 1, 2, compound 2e stood out as the most interesting dual AKR1B1/PTP1B inhibitor, because of its significant ability to inhibit both target enzymes at submicromolar and low micromolar concentrations, respectively (Table 1). Compounds 2f and 2g also showed appreciable activity profiles, thus deserving further investigation along with 2e. Analogue 1g was also selected for additional assays, because it showed excellent AKR1B1 inhibitory effectiveness along with PTP1B inhibitory potency similar to that of 2f and 2g.

2.3. Kinetic Studies

Compounds 1g and 2e–g were further characterized to assess their mechanism of action toward AKR1B1. Based on the measured IC₅₀ values, compound 1g was considered a tight-binding inhibitor. The experimental data of the reaction rates measured at different L-idose concentrations in the presence of different inhibitor concentrations are reported in the Supplementary Materials (Figures S29–S32). The secondary plots of ^appV_max and ^appK_M for compounds 2e–g are reported in Figure 2, Figure 3 and Figure 4. For compound 1g, the primary kinetic data (Figure S32) were fitted to the Morrison equation (Figure 5A), and the secondary plot of K_i as a function of L-idose concentration is reported in Figure 5B. For all the selected compounds, the resulting K_i and K′_i values are reported in Table 2. From these values, it turned out that compounds 2f and 2g acted as mixed-type inhibitors, while 2e behaved as an uncompetitive inhibitor. The most potent analogue, 1g, acted as a pure noncompetitive inhibitor, with almost identical K_i and K′_i values (Table 2).

Figure 2.

Kinetic characterization of compound 2e as an AKR1B1 inhibitor. Kinetic measurements reported in Supplementary Figure S29 were used. Panel (A) refers to the 1/^appV_max versus [I] plot. Panel (B) refers to the ^appK_M/^appV_max versus [I] plot. Error bars represent the standard deviations of the mean from at least three independent experiments.

Figure 3.

Kinetic characterization of compound 2f as an AKR1B1 inhibitor. Kinetic measurements reported in Supplementary Figure S30 were used. Panel (A) refers to the 1/^appV_max versus [I] plot. Panel (B) refers to the ^appK_M/^appV_max versus [I] plot. Error bars (when not visible are within the symbol size) represent the standard deviations of the mean from at least three independent experiments.

Figure 4.

Kinetic characterization of compound 2g as an AKR1B1 inhibitor. Kinetic measurements reported in Supplementary Figure S31 were used. Panel (A) refers to the 1/^appV_max versus [I] plot. Panel (B) refers to the ^appK_M/^appV_max versus [I] plot. Error bars (when not visible are within the symbol size) represent the standard deviations of the mean from at least three independent experiments.

Figure 5.

Kinetic characterization of compound 1g as an AKR1B1 inhibitor. Panel (A): experimental kinetic data (see Supplementary Figure S32) were fitted to the Morrison equation (see Section 3). The numbers alongside the symbols refer to the L-idose concentration used in the assay. Error bars (when not visible are within the symbol size) represent the standard deviations of the mean from at least three independent experiments. Panel (B): The apparent inhibition constant (K_i) values were plotted against substrate concentration and fitted by nonlinear regression analysis to Equation (1) (see Section 3). Error bars (when not visible are within the symbol size) represent the standard deviations of the mean from at least three independent experiments.

Table 2.

Values of inhibition constants of selected inhibitors.

Compd.	AKR1B1		PTP1B
	Ki	K′i	Ki	K′i
1g	0.03 ± 0.01	0.03 ± 0.01	3.0	n.d
2e	n.d.	1.18 ± 0.01	10.8 ± 0.3	29.6 ± 3.4
2f	7.75 ± 0.26	2.01 ± 0.01	0.16 ± 0.01	6.9 ± 1.6
2g	1.14 ± 0.06	4.43 ± 0.30	0.03 ± 0.01	2.2 ± 1.0

K_i and K′_i values, expressed in μM, are the mean ± SE; n.d.: not detectable.

To determine the mechanism of PTP1B inhibition by compounds 1g and 2e–g, further kinetic analyses were performed. A dilution assay with the selected compounds showed that PTP1B fails to recover enzymatic activity after extensive dilution in the assay buffer following incubation with compounds 1g, 2f, and 2g (Figure S33). This finding suggests that these compounds might behave as irreversible or slow-binding inhibitors. To clarify this aspect, we analyzed the time course of PTP1B hydrolysis rates after diluting an aliquot of the enzyme into samples containing increasing concentrations of compounds 1g, 2f, and 2g. The data obtained are presented in Figures S34–S36. In all cases, the hydrolysis rate of the control sample appeared to be linear over the observation period, indicating that the steady-state rate was reached immediately. In contrast, in the presence of compounds 1g, 2f, and 2g, a lag time was observed before the full onset of inhibition, suggesting that these compounds act as slow-binding inhibitors. Furthermore, additional tests were carried out using a fixed inhibitor concentration and increasing concentrations of pNPP to evaluate whether the compounds act as competitive or non-competitive inhibitors. In the case of compound 1g, we observed that PTP1B activity remained low at 5 and 10 mM pNPP concentrations, while it strongly increased at 20 mM pNPP (Figure S37). This indicates that a high substrate concentration can overcome the inhibitory effect caused by compound 1g, suggesting that it acts as a competitive PTP1B inhibitor. Conversely, tests with both compounds 2f and 2g showed that the hydrolysis rate of PTP1B did not change with increasing pNPP concentrations, proving that these compounds behave as non-competitive inhibitors (Figures S38 and S39). Moreover, kinetic analyses revealed that the most active PTP1B inhibitor (compound 2e) acts as a reversible, non-competitive mixed-type inhibitor (Figure 6 and Figures S40–S43). As concerns compounds 1g, 2f, and 2g, the determination of Ki values was carried out using the equation derived from the appropriate kinetic models (see Supplementary Materials).

Figure 6.

Kinetic characterization of compound 2e as a PTP1B inhibitor. Panel (A) and Panel (B) report the secondary plots of 1/^appV_max (Panel (A)) and ^appK_M/^appV_max (Panel (B)), obtained from primary plots of kinetic measurements (see Supplementary Materials), as a function of the inhibitor concentration. Error bars (when not visible, they are within the symbol size) represent the standard deviations of the mean from at least three independent experiments.

2.4. Molecular Docking Studies

AKR1B1 kinetic studies revealed that compounds 1g, 2f, and 2g can interact with both the free enzyme and the ES complex, while compound 2e can interact only with the ES complex. To mechanistically understand the binding mode, molecular docking studies were conducted on the AKR1B1 enzyme, targeting both the catalytic binding site and the AKR1B1-idose complex (PDB-ID: 3V36) [27]. The resulting docking poses were analyzed using structure-based 3D pharmacophore models [28].

Molecular docking studies in the AKR1B1-idose complex (Figure 7) revealed ionic interactions with Lys221 for all ligands.

Figure 7.

Compounds 1g and 2e–g docked into the AKR1B1-idose complex (PDB-ID: 3V36) [27]. The 2D and 3D structures of docked compounds (A) 1g, (B) 2e, which is further away from Arg217 due to the methoxy group on the terminal phenyl ring (C) 2f, (D) 2g. Colour codes: grey ribbon—AKR1B1, yellow protein surface—lipophilic surface, blue protein surface—hydrophilic surface.

Compounds 1g, 2f, and 2g demonstrated an additional ionic interaction with Arg217. However, 2e was unable to form this interaction due to the methoxy group on the terminal phenyl ring, which increases the size of the molecule and consequently distances the propionic acid moiety further away from Arg217 compared to the acid moieties of the compounds 1g, 2f, and 2g. Therefore, these three compounds formed a hydrogen bond between the thioxothiazolidinone moiety and the backbone amide of Ala299, whereas 2e formed a lipophilic contact between the 5-benzylidene ring and the side chain of Ala299. The 5-benzylidene ring in all compounds formed lipophilic contacts with Leu301 and Trp219. The terminal rings of 1g, 2f, and 2g established hydrophobic contacts with Val47 and Trp20, whereas 2e only showed a hydrophobic contact to Phe122 due to the methoxy group preventing the terminal phenyl ring from fitting deeper into the binding site. Docking experiments with 1g and 2e–g all resulted in reasonable docking poses, indicating that they fit into the AKR1B1-idose complex, which is in correlation to the kinetic studies (Figure 7, Table 3).

Table 3.

AKR1B1-idose complex protein–ligand interactions with 1g, 2e–g.

1g—Ligand Moiety	Interaction Type	Protein Residues
Phenyl 1	Hydrophobic	Trp20, Val47
Phenyl 2	Hydrophobic	Trp219, Leu301
Oxygen 3	Hydrogen bond acceptor	Ala299
Carboxyl 5	Ionic	Arg217, Lys221
2e—Ligand Moiety	Interaction Type	Protein Residues
Phenyl 1	Hydrophobic	Phe122
Phenyl 2	Hydrophobic	Trp219, Ala299, Leu301
Carboxyl 4	Ionic	Lys221
2f—Ligand Moiety	Interaction Type	Protein Residues
Phenyl 1	Hydrophobic	Trp20, Val47
Phenyl 2	Hydrophobic	Trp219, Leu301
Oxygen 3	Hydrogen bond acceptor	Ala299
Carboxyl 6	Ionic	Arg217, Lys221
2g—Ligand Moiety	Interaction Type	Protein Residues
Phenyl 1	Hydrophobic	Trp20, Val47
Phenyl 2	Hydrophobic	Trp219, Leu301
Oxygen 3	Hydrogen bond acceptor	Ala299
Carboxyl 6	Ionic	Arg217, Lys221

Docking to the catalytic binding site of AKR1B1 (Figure 8) showed that the terminal phenyl rings of compounds 1g and 2e–g all formed hydrophobic contacts in the catalytic binding site with the residues Trp20, Val47, Trp79, and Phe122. The compounds showed hydrophobic contacts between the 5-benzylidene ring and residues Trp219 and Leu301. Compounds 1g, 2f, and 2g additionally interacted with Ala299. Aromatic π-stacking with Trp219 was only observed in mixed-type inhibitors 2f and 2g. This interaction was not present in the compounds 1g and 2e that act as pure non-competitive and uncompetitive inhibitors, respectively (Figure 8, Table 4).

Figure 8.

A 2D and 3D depiction of compounds 1g and 2e–g within the AKR1B1 catalytic binding site (PDB-ID: 3V36) [27]. The 2D and 3D structures of docked compounds (A) 1g, (B) 2e, (C) 2f, and (D) 2g. Colour code: grey ribbon—AKR1B1, yellow protein surface—lipophilic surface, blue protein surface—hydrophilic surface.

Table 4.

AKR1B1 catalytic binding site protein–ligand interactions with 1g, 2e–g.

1g—Ligand Moiety	Interaction Type	Protein Residues
Phenyl 1	Hydrophobic	Trp20, Val47, Trp79, Phe122
Phenyl 2	Hydrophobic	Trp219, Ala299, Leu301
2e—Ligand Moiety	Interaction Type	Protein Residues
Phenyl 1	Hydrophobic	Trp20, Val47, Trp79, Phe122
Phenyl 2	Hydrophobic	Trp219, Leu301
2f—Ligand Moiety	Interaction Type	Protein Residues
Phenyl 1	Hydrophobic	Trp20, Val47, Trp79, Phe122
Phenyl 2	Hydrophobic	Trp219, Ala299, Leu301
Phenyl 2	Aromatic	Trp219
2g—Ligand Moiety	Interaction Type	Protein Residues
Phenyl 1	Hydrophobic	Trp20, Val47, Trp79, Phe122
Phenyl 2	Hydrophobic	Trp219, Ala299, Leu301
Phenyl 2	Aromatic	Trp219
Ethyl 4	Hydrophobic	Leu124

Compounds 1g and 2g differ with respect to their molecular surfaces due to the varying lengths of their acetic acid and propionic acid moieties. The increased length of the propionic acid in 2g allows this compound to form an additional hydrophobic contact with Leu124 that is not present in the docking pose of compound 1g. Thus, a longer acid moiety is advantageous for binding in the catalytic site of AKR1B1 (Figure 9). This observation could explain the behaviour of 2g as a mixed-type inhibitor with a prevalence of the competitive component in comparison to 1g, which acts as a pure non-competitive inhibitor. Therefore, the molecular docking experiments suggest that both the aromatic interaction with Trp219 and the length of the acid moiety could be important for binding within the catalytic binding site of AKR1B1.

Figure 9.

A 3D depiction of the molecular surfaces of compounds 1g and 2g within the AKR1B1 catalytic binding site (PDB-ID: 3V36) [27]. The 3D structures of the molecular surfaces of the docked compounds (A) 1g and (B) 2g, which are closer to the protein and allowed, due to the length of their acetic moieties, to perform an additional hydrophobic interaction with Leu124. Colour code: grey ribbon—AKR1B1, yellow molecule surface—lipophilic surface, blue molecule surface—hydrophilic surface.

PTP1B kinetic studies revealed that compound 2e acts as a mixed-type non-competitive inhibitor, while 2f and 2g act as pure non-competitive inhibitors, and 1g acts as a competitive inhibitor. To obtain insights into the binding mode with the target enzyme, molecular docking experiments were performed on both the PTP1B catalytic binding site (PDB ID: 1Q6T) [29] and the previously described allosteric binding site of PTP1B [9,10,11].

The obtained docking poses of the previously described allosteric binding site of PTP1B showed that all the compounds can establish comparable lipophilic contacts between the terminal phenyl ring and the residues Pro206 and Pro210 (Figure 10). Furthermore, the terminal phenyl rings of compounds 1g, 2f, and 2g form additional hydrophobic contact with the side chain of residue Arg79. Additionally, compounds 1g and 2g showed a hydrophobic contact between the 5-benzylidene ring and the side chain of Lys103. The short acetic acid moiety does not allow the compound 1g to fit deep into the binding pocket and thus shows only ionic interactions with Arg105 and Lys103. In contrast, compounds 2e, 2f, and 2g have a propionic acid moiety and can fit deeper into the binding pocket by forming an additional ionic interaction with Arg169. The absence of this interaction provides insight into the mechanism of action of 1g as a competitive inhibitor and underlines the preference for a propionic acid moiety for binding within the allosteric binding site of PTP1B (Figure 10, Table 5).

Figure 10.

Compounds 1g and 2e–g docked into the PTP1B allosteric binding site (PDB-ID: 1Q6T) [29]. The 2D and 3D structures of docked compounds (A) 1g, which, due to the length of the acetic acid moiety, is unable to perform an ionic interaction with Arg169. (B) 2e, (C) 2f, and (D) 2g. Colour code: light brown ribbon—PTP1B, yellow protein surface—lipophilic surface, blue protein surface—hydrophilic surface.

Table 5.

PTP1B allosteric binding site protein–ligand interactions with 1g, 2e–g.

1g—Ligand Moiety	Interaction Type	Protein Residues
Phenyl 1	Hydrophobic	Arg79, Pro206, Pro210
Phenyl 2	Hydrophobic	Lys103
Carboxyl 5	Ionic	Lys103, Arg105
2e—Ligand Moiety	Interaction Type	Protein Residues
Phenyl 1	Hydrophobic	Pro206, Pro210
Phenyl 2	Hydrophobic	Lys103
Carboxyl 4	Ionic	Lys103, Arg105, Arg169
2f—Ligand Moiety	Interaction Type	Protein Residues
Phenyl 1	Hydrophobic	Arg79, Pro206, Pro210
Carboxyl 5	Ionic	Lys103, Arg105, Arg169
2g—Ligand Moiety	Interaction Type	Protein Residues
Phenyl 1	Hydrophobic	Arg79, Pro206, Pro210
Carboxyl 6	Ionic	Lys103, Arg105, Arg169

The molecular docking studies into the catalytic binding site of PTP1B revealed an ionic interaction of 1g and 2e–g with the side chain of Arg221 (Figure 11). Additionally, the terminal phenyl ring of all compounds formed a hydrophobic contact with the side chain of Met258. Compounds 2f and 2g showed that the propionic acid moiety can form a hydrogen bond with the backbone amide of Phe182.

Figure 11.

Compounds 1g and 2e–g docked into the PTP1B catalytic binding site (PDB-ID: 1Q6T) [29]. The 2D and 3D structures of docked compounds (A) 1g, (B) 2e, (C) 2f, which, due to the location of the terminal phenyl ring, lose their hydrophobic contact with Val49. (D) 2g. Colour code: light brown ribbon—PTP1B, yellow protein surface—lipophilic surface, blue protein surface—hydrophilic surface.

Furthermore, compounds 1g and 2e showed hydrogen bonds from their respective acid moieties to the backbone amides of Cys215 and Ser216, which are not present in either 2f or 2g, showing that an acetic acid moiety or a substitution in the para-position of the terminal phenyl ring is preferred for the achievement of hydrogen bond interactions with these residues. The tests of compound 2f showed that its terminal phenyl ring was situated at a greater distance from Arg24 and was located at the bottom of the binding pocket, with the 5-benzylidene ring being located further away from Val49 due to the para-substitution of the 5-benzylidene ring pushing the terminal phenyl ring further away from the binding site. This resulted in the loss of the hydrophobic contact between the 5-benzylidene ring and the side chain of Val49. The hydrogen bond interactions formed by the acidic moiety with the backbone amides of both Cys215 and Ser216, which were present in compounds 1g and 2e, are crucial factors that affect the activity of the compounds within the catalytic binding site of PTP1B. Consequently, this molecular docking study explains the pure non-competitive inhibitor behaviour of 2f and 2g (Figure 11, Table 6).

Table 6.

PTP1B catalytic binding site protein–ligand interactions with 1g, 2e–g.

1g—Ligand Moiety	Interaction Type	Protein Residues
Phenyl 1	Hydrophobic	Met254
Phenyl 2	Hydrophobic	Val49
Oxygen 4	Hydrogen bond acceptor	Cys215
Carboxyl 5	Ionic	Arg221
Oxygen 6	Hydrogen bond acceptor	Ser216
2e—Ligand Moiety	Interaction Type	Protein Residues
Phenyl 1	Hydrophobic	Met254
Phenyl 2	Hydrophobic	Val49
Oxygen 3	Hydrogen bond acceptor	Cys215
Carboxyl 4	Ionic	Arg221
Oxygen 5	Hydrogen bond acceptor	Ser216
2f—Ligand Moiety	Interaction Type	Protein Residues
Phenyl 1	Hydrophobic	Met254
Oxygen 3	Hydrogen bond acceptor	Phe182
Carboxyl 5	Ionic	Arg221
2g—Ligand Moiety	Interaction Type	Protein Residues
Phenyl 1	Hydrophobic	Met254
Phenyl 2	Hydrophobic	Val49
Oxygen 5	Hydrogen bond acceptor	Phe182
Carboxyl 6	Ionic	Arg221

2.5. In Silico Pharmacokinetic and Toxicity Studies

The assessment of its pharmacokinetic (absorption, distribution, metabolism, elimination) and toxicity profiles (ADMET) is crucial to determine the potential of a bioactive molecule to be further developed as a drug candidate. In the initial phases of drug discovery, in silico models can be valuable tools to anticipate the drug-like properties of active compounds by evaluating certain molecular features, which can be predictive of the in vivo behaviour of a new agent.

Therefore, we carried out an in silico ADMET study of the most promising compounds, 1g, 2e, 2f, and 2g; epalrestat (EPA) was also included as a reference drug, being an orally bioavailable and generally well-tolerated drug, which is structurally related to derivatives 1 and 2. We applied several computational algorithms, available on the SwissADME, PreADMET, and PhaKinPro platforms (available at www.swissadme.ch/index.php, https://preadmet.webservice.bmdrc.org, and https://phakinpro.mml.unc.edu, respectively; accessed on 21–24 January 2025), which incorporate predictive models for physico-chemical properties, drug-likeness, pharmacokinetics, and toxicity [30,31,32,33,34].

It is well-known that the ADMET profile of a bioactive compound is strictly related to certain physico-chemical properties and molecular descriptors, such as molecular weight (MW), polar surface area (PSA), lipophilicity (logP), water solubility, number of rotatable bonds (RB), and number of hydrogen bond donor (HBD) and hydrogen bond acceptor (HBA) groups. The well-accepted and widely applied “rule-of-5” was formulated by Lipinski as a qualitative predictor of absorption/permeability based on the analysis of some of the above-mentioned key properties of a wide library of compounds [32]. According to Lipinski’s rule, poor absorption or permeation is more likely when two or more of the following conditions are not met: MW ≤ 500, calculated logP (clogP) ≤ 5.0, number of HBD ≤ 5, number of HBA ≤ 10. Additional rules were proposed by Veber, based on further structural properties that were shown to increase oral bioavailability in animals; according to Veber’s rules, good oral bioavailability is likely for compounds with RB ≤ 10, PSA ≤ 140 Ǻ, and total hydrogen bonds (HBA + HBD) ≤ 12 [33]. Moreover, Ghose and coll. reported a qualitative and quantitative characterization of known drug libraries, such as the Comprehensive Medicinal Chemistry (CMC) database [34], from which drug-likeness qualifying ranges (covering more than 80% of compounds) were deduced for several parameters, such as clog P (between −0.4 and 5.6), MW (between 160 and 480), total number of atoms (between 20 and 70) [34].

Based on the above-described rules, implemented in both the SwissADME and PreADMET platforms, compounds 1g, 2e, 2f, and 2g were qualified as drug-like compounds. As reported in Table 7, calculated parameters of these compounds met the criteria of Lipinski’s rule without any violations; they also matched Veber’s rule, with only one violation for compounds 2f and 2g (RB > 10). All tested compounds also met Ghose’s rules, whereas their TPSA value (133.46 Ų) was slightly higher than that indicated by Egan (131.6 Ų) as the highest value related to a good passive intestinal absorption [35]. However, the TPSA (topological polar surface area) parameter is calculated by using a fragmental method [36], without considering stereochemistry and molecular flexibility; therefore, the exposed polar surface and its effect on bioavailability might deviate from the in silico prediction.

Table 7.

Values of some calculated molecular descriptors of compounds 1g, 2e, 2f, and 2g ^a.

	1g	2e	2f	2g	EPA
MW	443.54	457.56	457.56	457.56	319.40
logP	3.84	4.00	4.07	4.10	2.43
HBD	1	1	1	1	1
HBA	5	5	5	5	3
RB	10	10	11	11	4
TPSA (Å2)	133.46	133.46	133.46	133.46	115.00
Lipinski’s rule violation	0	0	0	0	0
Veber’s rule violation	0	0	1	1	0

^a SwissADME http://www.swissadme.ch/index.php.

Based on the set of values of the calculated descriptors, compounds 1g, 2e, 2f, and 2g showed a promising drug-likeness and, therefore, it can be expected that they are suitable for oral administration.

A bioavailability score of 0.56 was estimated by SwissADME for compounds 1g, 2e, 2f, and 2g; according to a study performed by Martin [37], this score was calculated by combining total charge, TPSA, and Lipinski’s filters. A score of 0.56 indicates a probability of 56% that the tested compounds achieve at least 10% oral bioavailability in rats or display measurable Caco-2 permeability.

In vitro human colorectal adenocarcinoma Caco-2 cell permeability is considered a reliable model for the prediction of oral drug absorption. The analysis of data from the above-mentioned programmes (Table 8) suggested appreciable Caco-2 permeability for compounds 1g, 2e, 2f, and 2g, which correlates with the reference EPA. Accordingly, the predicted oral bioavailability (F, which expresses the fraction of an orally administered agent that reaches systemic circulation; F = 1 indicates 100% bioavailability following an intravenous administration) was in the range from 0.5 F to 0.8 F (predicted oral bioavailability from 50% to 80%) for compounds 2e and 2f and higher than 0.8 F (>80%) for compounds 2g and 1g, whereas the value calculated for EPA was lower than 0.5 F (<50%) (Table 8) [30]. Therefore, the compounds under study might be endowed with better oral bioavailability compared to the analogue EPA; this prediction might be related to the higher lipophilicity due to the 5-arylidene moiety of compounds 1, 2, compared with the reference drug.

Table 8.

Predictive ADME profile of compounds 1g, 2e, 2f, and 2g ^a,b.

	1g	2e	2f	2g	EPA
Caco2 cell permeable a,b	Yes	Yes	Yes	Yes	Yes
Oral bioavailability b	>0.8 F	0.5–0.8 F	0.5–0.8 F	>0.8 F	<0.5 F
Plasma protein binder a,b	Yes	Weak	Yes	Yes	Yes
Hepatic stability at 60 min b	>50%	>50%	>50%	>50%	>50%
CYP2D6 substrate a	No	No	No	No	No
CYP3A4 substrate a	Weakly	No	Weakly	Weakly	Yes
CYP2C19 inhibition a	No	No	No	No	No
CYP2C9 inhibition a	Yes	No	No	No	No
CYP2D6 inhibition a	No	No	No	No	No
CYP3A4 inhibition a	No	Yes	Yes	No	No

^a PreADMET (https://preadmet.webservice.bmdrc.org). ^b PhaKinPro (https://phakinpro.mml.unc.edu).

Moreover, it is worth noting that out of the compounds 1g, 2e, 2f, and 2g, none was predicted to act as a substrate for glycoprotein P (P-gp); since P-gp functions as an efflux pump, this feature could contribute to a good bioavailability prediction.

Compounds 1g, 2e, 2f, and 2g displayed moderate water solubility, with Log S values (calculated by using an implementation of the ESOL model) [31] ranging between −5.33 and −5.28. Their capability to cross the blood–brain barrier as well as their skin permeability were predicted to be scarce.

The prediction of interaction of potential drugs with cytochromes P450 (CYP) could also be interesting, since these isoenzymes play key roles in the metabolism of xenobiotics. Compounds 1g, 2e, 2f, and 2g, as well as EPA, appeared not to be metabolized by the isoform CYP2D6, whereas 1g, 2f, and 2g might be weak substrates of CYP3A4 (Table 8); accordingly, the calculated hepatic stability at 60 min was >50% for all compounds under consideration. Moreover, the inhibition of major CYP isoforms, such as CYP2C19, CYP2C9, CYP2D6, and CYP3A4, could cause pharmacokinetic drug–drug interactions. Interestingly, compounds 1g, 2e, 2f, and 2g were generally predicted not to be good CYP inhibitors; however, isoform CYP3A4 might be inhibited by 2e and 2f, whereas analogue 1g may inhibit CYP2C9 (Table 8).

The in silico models implemented in the PreADMET software predicted no carcinogenicity risk at 2 years in rat and mouse models, as well as very low acute toxicity in both animal and plant organisms for all tested compounds, analogous to the reference EPA (Table 9). A medium risk of hERG inhibition was calculated, which might be related to potential cardiotoxicity.

Table 9.

Predictive toxicity profile of compounds 1g, 2e, 2f, and 2g ^a.

	1g	2e	2f	2g	EPA
2-year carcinogenicity in mice	NC	NC	NC	NC	NC
2-year carcinogenicity in rats	NC	NC	NC	NC	NC
Acute algae toxicity	0.0069	0.0060	0.0060	0.0053	0.0280
Acute daphnia toxicity	0.0058	0.0039	0.0041	0.0044	0.0329
Acute fish toxicity	0.00023	0.00015	0.00016	0.00016	0.0027
hERG inhibition risk	Medium	Medium	Medium	Medium	Medium

^a PreADMET (https://preadmet.webservice.bmdrc.org); NC = non-carcinogen.

Overall, the predicted drug-likeness and ADMET profiles of the selected compounds 1g, 2e, 2f, and 2g were promising, encouraging the continuation of the research on these enzyme inhibitors.

2.6. Ex Vivo Assays

To evaluate the ability of selected compounds to increase cellular sensitivity to insulin, tests were carried out using C2C12 cells. First, we evaluated the cytotoxicity of compounds 1g, 2e, 2f, and 2g by incubating C2C12 myoblasts with increasing concentrations of these compounds for 24 h. After this time, cell viability was determined by means of an MTT assay. We observed that all compounds were well tolerated by C2C12 cells, as a significant reduction in cell viability was observed only at the highest doses (Figure S47).

Then, we evaluated the capability of compounds 1g, 2e, 2f, and 2g to stimulate glucose uptake. C2C12 cells were treated with 10 nM insulin or 10 μM of each compound for 30 min and then incubated with 2-NBDG for 3 h. After this time, cells were detached and analyzed using a flow cytometry apparatus (Figure 12).

Figure 12.

Glucose uptake assay. C2C12 cells grown on a p35 well were starved for 24 h and then treated for 30 min with starvation medium containing 10 nM insulin, and/or 10 μM of compounds 1g, 2e–g. After this period, cells were washed and incubated for 3 h with fresh medium containing 40 μM of fluorescent glucose. The cells were then washed, detached, and analyzed using a flow cytometer. For each test, at least 10,000 events were acquired. Data were normalized to the control test. Data reported in the figure represent the mean value ± SD (n = 3). The t-test was used to assess the statistical significance between the control and other groups. * p < 0.05; ** p < 0.01.

We found that compounds 2e and 2f strongly stimulated the glucose uptake when administered alone, while compound 2g showed a weaker glucose uptake-stimulating activity. No differences in the intracellular glucose levels were found when cells were treated with a combination of compounds 2e–f and insulin, thus suggesting that these compounds could act as insulin-mimetic agents. Finally, compound 1g was not able to improve the cellular glucose uptake (Figure 12).

Moreover, the ability of the selected compounds to improve the insulin signalling pathway was evaluated by monitoring the activation of the kinase Akt in the C2C12 cell line. (Figure 13). We found that none of the analyzed compounds could increase the phosphorylation status of the Akt kinase, either when administered alone (at 10 mM concentration) or in combination with insulin (Figure 13).

Figure 13.

Effect of compounds 1g, 2e–g on the insulin signalling pathway. C2C12 cells were starved for 24 h and then stimulated for 30 min with 10 nM insulin, either with 10 μM of each compound 1g, 2e–g, or with a combination of insulin + compound. After 30 min, cells were lysed and protein extracts analyzed to evaluate the phosphorylation levels of kinase Akt by Western blot (figures on the left). Each test was carried out in triplicate. Data reported in the graphs on the right represent the mean value ± SD. All data were normalized with respect to the control sample.

Therefore, it can be inferred that the tested compounds can stimulate glucose uptake, as observed in C2C12 cells, by an insulin-independent mechanism. Recent studies reported that some 2,4-thiazolidinediones acutely inhibit mitochondrial pyruvate carrier, impairing cellular respiration and mitochondrial ATP production [38]. Based on this evidence, we wondered if 4-thiazolidinone derivatives 2e–g could act in a similar manner. To address this question, we evaluated the mitochondrial potential in C2C12 cells acutely treated with compounds 2e–g (Figure 14).

Figure 14.

Evaluation of mitochondrial membrane potential with the TMRE probe. C2C12 cells were starved for 24 h and then incubated for 30 min with 10 μM compounds 1g, 2e–g, with 20 μM UK5099 or with 10 mM methylpyruvate. Then, cells were washed with PBS and incubated for a further 30 min with fresh starvation medium containing 0.2 μM TMRE. After, cells were analyzed using a flow cytometer. All tests were carried out in triplicate, and for each experiment, 10,000 events were analyzed. Data were analyzed using FlowJo v.10 (FlowJo) and normalized with respect to control cells. Data reported in the figure represent the mean value ± SD. The differences between the experimental and control groups were compared using the Student’s t-test. * p < 0.05.

Data reported in Figure 14 show that the fluorescence of the TMRE probe incorporated into the mitochondria of C2C12 cells treated with compounds 2e–g was significantly higher than that of the control cells and very similar to that observed after incubation of muscle cells with methyl pyruvate, a membrane-permeable ester of pyruvate. Instead, fluorescence levels like those of the control cells were observed following treatment with compound 1g, whereas treatment with UK5099—a well-known inhibitor of the mitochondrial pyruvate carrier—resulted in a decrease in the mitochondrial membrane potential. This latter result was expected, as a previous study demonstrated that UK5099 impairs oxygen consumption when administered to C2C12 myoblasts [38]. The results of the TMRE fluorescence test suggest that the increased glucose uptake triggered by compounds 2e–g could be related to the hyperpolarization of cellular mitochondria, by mimicking the effects observed following methyl pyruvate administration. A possible hypothesis is that compounds 2e–g may target the mitochondrial pyruvate transporter [39], by stabilizing it in a conformation that facilitates the translocation of pyruvate into the mitochondria, thereby also enhancing cellular glucose uptake through a sort of “carryover effect.” Based on this hypothesis, we might expect that the mitochondria of myoblasts treated with compounds 2e–g can produce greater amounts of reduced coenzymes (i.e., NADH and FADH₂, produced by the Krebs cycle), whose subsequent oxidation via mitochondrial complexes I and II promotes an increase in mitochondrial membrane potential. However, if this assumption is correct, we would also expect that, as a consequence of the increased pyruvate flux into the mitochondria, acetyl-CoA synthesis would also increase, leading to the activation of pyruvate carboxylase, an enzyme responsible for oxaloacetate synthesis. Under these conditions, acetyl-CoA molecules derived from other metabolic pathways, such as β-oxidation, could be metabolized, resulting in a higher rate of fatty acid oxidation.

To confirm this hypothesis, we evaluated the ability of liver HepG2 cells, overloaded with fatty acids and then treated with compounds 2e–g, to reduce the content of lipid droplets. We found that liver cells treated with compounds 2e–g (10 μM) showed a significant reduction in lipid content with respect to untreated cells, therefore confirming that these compounds can boost fatty acid degradation, thus preventing lipid accumulation in hepatocytes (Figure 15 and Figure S48).

Figure 15.

Lipid accumulation assay. HepG2 liver cells were treated with 0.4 mM oleic acid (OA) for 24 h, and after were treated with 10 μM compounds 2e–g for a further 72 h. Then, the intracellular lipid content was determined by using the Oil Red O dye. All tests were carried out in quadruplicate. Data obtained were normalized with respect to protein content and reported as a percentage of the control test. The data shown in the figure represent the mean values ± SD. ** p < 0.01.

Taken together, these results suggest that 4-thiazolidinone derivatives 2e–g might interact with the mitochondrial pyruvate transporter differently from analogous 2,4-thiazolidinediones by promoting pyruvate transport rather than inhibiting it.

Furthermore, human lens epithelial B3 line (HLE) cells and human retinal glial line (MIO-M1) cells were used as cellular models for the onset of high-glucose-induced cataract and retinopathy, respectively. Dual AKR1B1/PTP1B inhibitors 2e and 2f showed interesting effects in C2C12 myoblasts by significantly improving glucose uptake; therefore, it was also worth assessing their effect on high glucose-induced sorbitol accumulation. Compound 1g was also included in these assays as the most potent AKR1B1 inhibitor among compounds 1a–g, 2a–g (Table 1).

The effect of compounds 1g, 2e, and 2f on cell viability was tested. No toxicity was observed upon 48 h incubation of HLE cells in the presence of inhibitor concentrations of up to 5 μM (Figure S49). In the case of MIO-M1 cells, similar results were obtained, except for 1g, which caused significant toxicity at a concentration of 5 μM (Figure S49). Thus, in further experiments, 1g was used at a lower concentration (i.e., 1 μM), which did not affect the MIO-M1 cells’ viability (Figure S49). The ability of selected compounds to impair the sorbitol accumulation observed in HLE and MIO-M1 cells exposed to high glucose was evaluated, using Sorbinil as a reference inhibitor (Figure 16). Both cell lines were sensitive to compound 2e, which was significantly effective in impairing sorbitol accumulation in both cell lines in a comparable manner, but exhibited complete insensitivity to compound 2f. Compound 1g significantly impaired sorbitol accumulation in HLE cells, but the lower concentration tested, due to its observed toxicity in MIO-M1 cells, did not allow any effect to be highlighted in this latter cell line. Finally, differences between the sensitivity of two cell lines to the reference drug were observed, since sorbinil was more effective in recovering the normal sorbitol content in MIO-M1 with respect to HLE cells (Figure 16).

Figure 16.

Effect of selected compounds on high-glucose-induced sorbitol accumulation. HLE (Panel A) and MIO-M1 (Panel B) cells were cultured for 48 h in the proper medium (see Section 3), (control), in the presence of 0.05% DMSO. HG refers to cells incubated as above in the presence of 75 mM D-glucose alone or in the presence of the indicated inhibitor concentrations. Sorbitol content for control cells accounted for 28.5 ± 3.2 and 33 ± 6.8 nmol/mg protein for HLE and MIO-M1 cells, respectively. Values are reported as the mean ± SEM of five and three independent measurements for HLE and MIO-M1 cells, respectively. Statistical analysis was performed using one-way Anova with the Dunnet post hoc test. Significance was evaluated with respect to HG (**: p < 0.01; ****: p < 0.0001).

3. Materials and Methods

3.1. Chemistry

Melting points were determined using a Kofler hot-stage apparatus and are reported without correction. Thin-layer chromatography (TLC) was performed on silica gel plates (Merck F254, Darmstadt, Germany), and Rf values were obtained employing appropriate mixtures of diethyl ether and n-hexane as eluents. Elemental analyses (C, H, N) were carried out on a Carlo Erba model 1106 elemental analyzer (Milan, Italy), with results within ±0.4% of the theoretical values. ¹H and ¹³C NMR spectra were recorded on a Varian spectrometer (Palo Alto, CA, USA) operating at 500 MHz (499.74 MHz for ¹H and 125.73 MHz for ¹³C). In addition to standard one-dimensional ¹H and ¹³C NMR experiments, ¹H–¹H COSY and ¹H–¹³C gHSQCAD correlation experiments were performed. Chemical shifts (δ) are reported in ppm and coupling constants (J) in Hz. Spectra were phased and baseline corrected as required. CDCl₃ or DMSO-d₆ served as internal references for both ¹H and ¹³C spectra. Exchangeable protons were confirmed by D₂O addition. Unless otherwise specified, all reagents were purchased from commercial suppliers and used without further purification. The purity of the synthesized compounds was confirmed to be ≥95% by elemental analysis.

3.1.1. General Procedure for the Synthesis of Arylalkoxy Benzaldehydes 3a–g

Aryl/aryloxyalkyl bromide (26.6 mmol) was added, in small portions at regular intervals of a few minutes, to a mixture of 3-hydroxybenzaldehyde or 4-hydroxybenzaldehyde (2 g, 16.4 mmol), and potassium carbonate (9.06 g, 65.6 mmol) in anhydrous DMF (75 mL); the mixture was stirred at 50–70 °C for 2–4 h. Then water (20 mL) was added, and the mixture was acidified to pH 5–6 and extracted with ethyl acetate. The organic layer was washed with H₂O (10 × 100 mL), dried with anhydrous Na₂SO_4, and the solvent was removed under reduced pressure. The crude mixture was separated by silica gel column chromatography, using a mixture of petroleum ether:diethyl ether = 9:1 or cyclohexane:diethyl ether = 8:2 as eluant, to provide pure aldehydes 3a–g.

4-[2-(4-Methoxyphenyl)ethoxy]benzaldehyde (3a)

Yield 61%; yellow oil; ¹H NMR (CDCl₃): δ 3.09 (t J = 7.05 Hz, 2H, CH₂); 3.82 (s, 3H, OCH₃); 4.23 (t J = 7.05 Hz, 2H, OCH₂); 6.89–6.91 (m, 4H, arom); 7.02 (m, 2H, arom); 7.84 (m, 2H, arom); 9.88 (s, 1H, CHO). ¹³C NMR (CDCl₃): δ 34.8, 67.0 (CH₂); 55.4 (OCH₃); 114.0, 114.6, 128.1, 129.0, 130.1, 131.5, 157.5, 164.9 (CH arom, Cq arom); 190.9 (C=O).

3-[2-(4-Methoxyphenyl)ethoxy]benzaldehyde (3b)

Yield 36%; yellow oil; ¹H NMR (CDCl₃): δ 3.08 (t J = 7.0 Hz, 2H, CH₂); 3.81 (s, 3H, OCH₃); 4.21 (t J = 7.0 Hz, 2H, OCH₂); 6.87–6.90 (m, 3H, arom); 7.22–7.24 (m, 2H, arom); 7.40–7.45 (m, 3H, arom); 9.97 (s, 1H, CHO). ¹³C NMR (CDCl₃): δ 35.2, 67.0 (CH₂); 55.7 (OCH₃); 114.0, 116.0, 119.9, 121.3, 129.8, 129.9, 130.2, 137.4, 157.6, 165.1 (CH arom, Cq arom); 191.6 (C=O).

4-[2-(3-Methoxyphenyl)ethoxy]benzaldehyde (3c)

Yield 23%; yellow oil; ¹H NMR (CDCl₃): δ 3.13 (t J = 7.0 Hz, 2H, CH₂); 3.83 (s, 3H, OCH₃); 4.27 (t J = 7.0 Hz, 2H, OCH₂); 7.03–7.05 (m, 3H, arom); 7.26–7.29 (m, 2H, arom); 7.81–7.86 (m, 3H, arom); 9.86 (s, 1H, CHO). ¹³C NMR (CDCl₃): δ 35.3, 67.0 (CH₂); 55.5 (OCH₃); 111.2, 114.0, 114.6, 120.2, 128.3, 129.6, 131.5, 138.1, 159.8, 165.0 (CH arom, Cq arom); 190.8 (C=O).

4-[2-(3,4-Dimethoxyphenyl)ethoxy]benzaldehyde (3d)

Yield 45%; pale yellow oil; ¹H NMR (CDCl₃): δ 3.03 (t J = 6.95 Hz, 2H, CH₂); 3.83 and 3.84 (2s, 6H, 2 OCH₃); 4.19 (t J = 6.95 Hz, 2H, OCH₂); 6.77–6.80 (m, 3H, arom); 6.94–6.97 (m, 2H, arom); 7.76–7.78 (m, 2H, arom); 9.81 (s, 1H, CHO). ¹³C NMR (CDCl₃): δ 35.0, 66.8 (CH₂); 55.9 (OCH₃); 112.2, 112.8, 114.5, 121.8, 128.3, 130.6, 131.7, 146.5, 149.2, 165.0 (CH arom, Cq arom); 190.7 (C=O).

4-[3-(4-Methoxyphenyl)propoxy]benzaldehyde (3e)

Yield 92%; pale yellow solid; m.p. 65 °C; ¹H NMR (CDCl₃): δ 2.10–2.14 (m, 2H, CH₂); 2.78 (t J = 7.5 Hz, 2H, CH₂); 3.80 (s, 3H, OCH₃); 4.04 (t J = 6.3 Hz, 2H, OCH₂); 6.85 (m, 2H, arom); 7.00 (m, 2H, arom); 7.13 (m, 2H, arom); 7.83 (m, 2H, arom); 9.89 (s, 1H, CHO). ¹³C NMR (CDCl₃): δ 30.8, 31.0, 67.2 (CH₂); 55.3 (OCH₃); 113.9, 114.8, 129.4, 129.8, 132.0, 133.1, 157.9, 164.1 (CH arom, Cq arom); 190.8 (C=O).

4-(4-Phenoxybutoxy)benzaldehyde (3f)

Yield 87%; white solid; m.p. 109–110 °C; ¹H NMR (CDCl₃): δ 1.96–2.09 (m, 4H, 2 CH₂); 4.06 (t J = 5.8 Hz, 2H, OCH₂); 4.14 (t J = 5.95 Hz, 2H, OCH₂); 6.91–6.96 (m, 3H, arom); 7.01 (m, 2H, arom); 7.28–7.31 (m, 2H, arom); 7.84 (m, 2H, arom); 9.90 (s, 1H, CHO). ¹³C NMR (CDCl₃): δ 25.9, 27.9, 67.2, 67.9 (CH₂); 114.4, 114.7, 120.7, 129.4, 129.5, 132.0, 158.8, 159.6 (CH arom, Cq arom); 190.8 (C=O).

3-(4-Phenoxybutoxy)benzaldehyde (3g)

Yield 63%; white solid; m.p. 98–99 °C; ¹H NMR (CDCl₃): δ 2.00–2.07 (m, 4H, 2 CH₂); 4.06 (t J = 5.95 Hz, 2H, OCH₂); 4.12 (t J = 6.0 Hz, 2H, OCH₂); 6.91–6.95 (m, 3H, arom); 7.28–7.31 (m, 3H, arom); 7.40–7.47 (m, 3H, arom); 9.98 (s, 1H, CHO). ¹³C NMR (CDCl₃): δ 25.9, 26.0, 67.2, 67.8 (CH₂); 112.8, 114.5, 120.7, 121.9, 123.4, 129.5, 130.0, 137.8, 158.9, 159.6 (CH arom, Cq arom); 192.2 (C=O).

3.1.2. General Procedure for the Synthesis of (5-Arylidene-4-oxo-2-thioxothiazolidin-3-yl)alkanoic Acids 1a–g and 2a–g

A mixture of (4-oxo-2-thioxothiazolidin-3-yl)acetic acid (4) (0.40 g, 2.09 mmol) or 3-(4-oxo-2-thioxothiazolidin-3-yl)propanoic acid (5) (0.43 g, 2.09 mmol) with the appropriate aldehyde 3a–g (2.09 mmol) and sodium acetate (2.13 g, 15.7 mmol), in glacial acetic acid (10 mL), was heated to 100 °C for 4–5 h and subsequently maintained at room temperature overnight. The mixture was poured into H₂O, and the obtained crude precipitate was filtered off, then washed with H₂O and recrystallized from methanol to give pure acids 1a–g and 2a–g.

[(5Z)-5-({4-[2-(4-Methoxyphenyl)ethoxy]phenyl}methylidene)-4-oxo-2-thioxothiazolidin-3-yl]acetic Acid (1a)

Yield 46%; yellow solid; mp 281–284 °C; ¹H NMR (DMSO-d₆): δ 2.99 (t J = 6.9 Hz, 2H, CH₂); 3.72 (s, 3H, OCH₃); 4.24 (t J = 6.9 Hz, 2H, OCH₂); 4.52 (s, 2H, NCH₂); 6.87 (m, 2H, arom); 7.12 (m, 2H, arom); 7.24 (m, 2H, arom); 7.61 (m, 2H, arom), 7.78 (s, 1H, CH methylidene). ¹³C NMR (DMSO-d₆): δ 34.4, 47.2, 69.4 (CH₂); 55.6 (OCH₃); 114.4, 116.2, 126.1, 130.4, 130.5, 133.5, 158.5, 161.3 (CH arom, Cq arom); 119.7 (C-5); 133.7 (CH methylidene); 167.3, 167.4 (C=O); 193.6 (C=S). Anal. (C₂₁H₁₉NO₅S₂) calcd: C 58.73; H 4.46; N 3.26; found: C 58.54; H 4.61; N 3.36.

[(5Z)-5-({3-[2-(4-Methoxyphenyl)ethoxy]phenyl}methylidene)-4-oxo-2-thioxothiazolidin-3-yl]acetic Acid (1b)

Yield 35%; yellow solid; mp 213–215 °C; ¹H NMR (DMSO-d₆): δ 2.99 (t J = 6.9 Hz, 2H, CH₂); 3.72 (s, 3H, OCH₃); 4.21 (t J = 6.9 Hz, 2H, OCH₂); 4.54 (s, 2H, NCH₂); 6.87 (m, 2H, arom); 7.09 (dd J = 8.25 and 2.15 Hz, 1H, arom); 7.19–7.26 (m, 4H, arom); 7.46 (dd J = 7.95 Hz, 1H, arom), 7.81 (s, 1H, CH methylidene). ¹³C NMR (DMSO-d₆): δ 34.5, 47.1, 69.2 (CH₂); 55.6 (OCH₃); 114.3, 116.9, 123.0, 123.2, 130.5, 131.2, 134.9, 158.4, 159.5 (CH arom, Cq arom); 118.2 (C-5); 133.7 (CH methylidene); 167.1, 167.7 (C=O); 193.7 (C=S). Anal. (C₂₁H₁₉NO₅S₂) calcd: C 58.73; H 4.46; N 3.26; found: C 58.80; H 4.40; N 3.17.

[(5Z)-5-({4-[2-(3-Methoxyphenyl)ethoxy]phenyl}methylidene)-4-oxo-2-thioxothiazolidin-3-yl]acetic Acid (1c)

Yield 42%; yellow solid; m.p. 207–210 °C; ¹H NMR (DMSO-d₆): δ 3.01 (t J = 6.7 Hz, 2H, CH₂); 3.79 (s, 3H, OCH₃); 4.27 (t J = 6.7 Hz, 2H, OCH₂); 4.71 (s, 2H, NCH₂); 6.77 (dd J = 8.2 and 2.45 Hz, 1H, arom); 6.86–6.88 (m, 2H, arom); 7.09 (m, 2H, arom); 7.20 (dd J = 8 Hz, 1H, arom); 7.58 (m, 2H, arom), 7.78 (s, 1H, CH methylidene). ¹³C NMR (DMSO-d₆): δ 35.6, 45.9, 69.4 (CH₂); 55.8 (OCH₃); 112.7, 115.5, 116.6, 122.1, 126.2, 130.3, 135.0, 140.6, 160.2, 161.9 (CH arom, Cq arom); 119.4 (C-5); 134.0 (CH methylidene); 167.4, 168.3 (C=O); 194.0 (C=S). Anal. (C₂₁H₁₉NO₅S₂) calcd: C 58.73; H 4.46; N 3.26; found: C 58.91; H 4.35; N 3.42.

[(5Z)-5-({4-[2-(3,4-Dimethoxyphenyl)ethoxy]phenyl}methylidene)-4-oxo-2-thioxothiazolidin-3-yl]acetic Acid (1d)

Yield 37%; yellow solid; mp 230–233 °C; ¹H NMR (DMSO-d₆): δ 2.98 (t J = 6.85 Hz, 2H, CH₂); 3.70, 3.73 (2s, 6H, 2 OCH₃); 4.25 (t J = 6.85 Hz, 2H, OCH₂); 4.72 (s, 2H, NCH₂); 6.81 (d J = 8.2 Hz, 1H, arom); 6.86 (d J = 8.2 Hz, 1H, arom); 6.93 (s, 1H, arom); 7.12 (m, 2H, arom); 7.61 (m, 2H, arom); 7.82 (s, 1H, CH methylidene). ¹³C NMR (DMSO-d₆): δ 35.0, 45.7, 69.5 (CH₂); 56.2, 56.3 (OCH₃); 112.6, 113.6, 116.4, 121.6, 126.0, 131.1, 134.9, 148.1, 149.3, 161.7 (CH arom, Cq arom); 119.2 (C-5); 133.9 (CH methylidene); 167.2, 168.1 (C=O); 193.8 (C=S). Anal. (C₂₂H₂₁NO₆S₂) calcd: C 57.50; H 4.61; N 3.05; found: C 57.67; H 4.72; N 2.98.

[(5Z)-5-({4-[3-(4-Methoxyphenyl)propoxy]phenyl}methylidene)-4-oxo-2-thioxothiazolidin-3-yl]acetic Acid (1e)

Yield 37%; yellow solid; mp 238–242 °C; ¹H NMR (DMSO-d₆): δ 2.00 (m, 2H, CH₂); 2.68 (t J = 7.5 Hz, 2H, CH₂C₆H₅); 3.71 (s, 3H, OCH₃); 4.04 (t J = 6.2 Hz, 2H, OCH₂); 4.52 (s, 2H, NCH₂); 6.84 (m, 2H, arom); 7.13–7.15 (m, 4H, arom); 7.61 (m, 2H, arom), 7.79 (s, 1H, CH methylidene). ¹³C NMR (DMSO-d₆): δ 30.9, 31.0, 47.3, 67.7 (CH₂); 55.5 (OCH₃); 114.3, 116.2, 126.1, 129.8, 133.5, 133.6, 158.0, 161.5 (CH arom, Cq arom); 119.6 (C-5); 133.7 (CH methylidene); 167.3, 167.4, (C=O); 193.6 (C=S). Anal. (C₂₂H₂₁NO₅S₂) calcd: C 59.58; H 4.77; N 3.16; found: C 59.70; H 4.81; N 3.09.

2-[(5Z)-5-{[4-(4-Phenoxybutoxy)phenyl]methylidene}-4-oxo-2-thioxothiazolidin-3-yl]acetic acid (1f)

Yield 62%; yellow solid; mp 217–220 °C; ¹H NMR (DMSO-d₆): δ 1.88 (m, 4H, 2 CH₂); 4.03 (t J = 5.5 Hz, 2H, OCH₂); 4.13 (t J = 5.6 Hz, 2H, OCH₂); 4.52 (s, 2H, NCH₂); 6.91–6.94 (m, 3H, arom); 7.13 (m, 2H, arom); 7.26–7.29 (m, 2H, arom); 7.62 (m, 2H, arom), 7.79 (s, 1H, CH methylidene). ¹³C NMR (DMSO-d₆): δ 25.8, 25.9, 42.5, 67.5, 68.2 (CH₂); 115.0, 116.3, 121.0, 126.1, 129.9, 130.1, 159.1, 161.5 (CH arom, Cq arom); 119.6 (C-5); 133.6 (CH methylidene); 167.3, 167.4 (C=O); 193.6 (C=S). Anal. (C₂₂H₂₁NO₅S₂) calcd: C 59.58; H 4.77; N 3.16; found: C 59.40; H 4.62; N 3.27.

2-[(5Z)-5-{[3-(4-Phenoxybutoxy)phenyl]methylidene}-4-oxo-2-thioxothiazolidin-3-yl]acetic Acid (1g)

Yield 62%; yellow solid; mp 215–218 °C; ¹H NMR (DMSO-d₆): δ 1.89–1.90 (m, 4H, 2 CH₂); 4.03 (t J = 5.9 Hz, 2H, OCH₂); 4.11 (t J = 5.85 Hz, 2H, OCH₂); 4.74 (s, 2H, NCH₂); 6.90–6.94 (m, 3H, arom); 7.12 (dd J = 8.45 and 2.3 Hz, 1H, arom); 7.21–7.22 (m, 2H, arom); 7.26–7.29 (m, 2H, arom), 7.47 (dd J = 8.25 Hz, 1H, arom), 7.86 (s, 1H, CH methylidene). ¹³C NMR (DMSO-d₆): δ 25.9, 26.0, 45.6, 67.5, 68.1 (CH₂); 115.0, 116.9, 121.0, 122.7, 123.1, 130.0, 131.2, 134.7, 159.2, 159.7 (CH arom, Cq arom); 118.3 (C-5); 134.5 (CH methylidene); 166.8, 167.8 (C=O); 193.8 (C=S). Anal. (C₂₂H₂₁NO₅S₂) calcd: C 59.58; H 4.77; N 3.16; found: C 59.61; H 4.85; N 3.08.

3-[(5Z)-5-({4-[2-(4-Methoxyphenyl)ethoxy]phenyl}methylidene)-4-oxo-2-thioxothiazolidin-3-yl]propanoic Acid (2a)

Yield 38%; yellow solid; mp 202–205 °C; ¹H NMR (DMSO-d₆): δ 2.62 (t J = 7.85 Hz, 2H, CH₂); 2.99 (t J = 6.8 Hz, 2H, CH₂); 3.72 (s, 3H, OCH₃); 4.20–4.23 (m, 4H, 2 CH₂); 6.87 (m, 2H, arom); 7.11 (m, 2H, arom); 7.23 (m, 2H, arom); 7.59 (m, 2H, arom), 7.77 (s, 1H, CH methylidene), 12.49 (br s, 1H, COOH). ¹³C NMR (DMSO-d₆): δ 31.6, 34.6, 40.7, 69.5 (CH₂); 55.7 (OCH₃); 114.5, 116.4, 126.2, 130.6, 130.7, 134.0, 158.6, 161.5 (CH arom, Cq arom); 119.7 (C-5); 133.7 (CH methylidene); 167.5, 172.5 (C=O); 193.8 (C=S). Anal. (C₂₂H₂₁NO₅S₂) calcd: C 59.58; H 4.77; N 3.16; found: C 59.74; H 4.70; N 3.00.

3-[(5Z)-5-({3-[2-(4-Methoxyphenyl)ethoxy]phenyl}methylidene)-4-oxo-2-thioxothiazolidin-3-yl]propanoic Acid (2b)

Yield 47%; orange solid; mp 132–135 °C; ¹H NMR (DMSO-d₆): δ 2.63 (t J = 7.8 Hz, 2H, CH₂); 2.99 (t J = 6.8 Hz, 2H, CH₂); 3.69 (s, 3H, OCH₃); 4.14–4.21 (m, 4H, 2 CH₂); 6.84 (m, 2H, arom); 7.03 (dd J = 8.3 and 2.45 Hz, 1H, arom); 7.08–7.12 (m, 2H, arom); 7.21 (m, 2H, arom); 7.41 (dd J = 8 Hz, 1H, arom); 7.69 (s, 1H, CH methylidene); 12.53 (br s, 1H, COOH). ¹³C NMR (DMSO-d₆): δ 31.7, 34.8, 40.9, 69.6 (CH₂); 55.9 (OCH₃); 114.7, 117.0, 123.5, 123.6, 130.9, 131.0, 131.7, 135.2, 158.8, 159.8 (CH arom, Cq arom); 118.5 (C-5); 133.9 (CH methylidene); 167.6, 172.8 (C=O); 194.1 (C=S). Anal. (C₂₂H₂₁NO₅S₂) calcd: C 59.58; H 4.77; N 3.16; found: C 59.41; H 4.90; N 3.34.

3-[(5Z)-5-({4-[2-(3-Methoxyphenyl)ethoxy]phenyl}methylidene)-4-oxo-2-thioxothiazolidin-3-yl]propanoic Acid (2c)

Yield 56%; yellow solid; m.p. 169–172 °C; ¹H NMR (DMSO-d₆): δ 2.61 (t J = 7.85 Hz, 2H, CH₂); 3.00 (t J = 6.7 Hz, 2H, CH₂COOH); 3.71 (s, 3H, OCH₃); 4.21 (t J = 7.85 Hz, 2H, OCH₂); 4.26 (t J = 6.7 Hz, 2H, NCH₂); 6.77 (m, 1H, arom); 6.86–6.88 (m, 2H, arom); 7.09 (m, 2H, arom); 7.20 (dd J = 8 Hz, 1H, arom); 7.55 (m, 2H, arom); 7.71 (s, 1H, CH methylidene); 12.51 (br s, 1H, COOH). ¹³C NMR (DMSO-d₆): δ 31.7, 35.6, 40.8, 69.3 (CH₂); 55.8 (OCH₃); 112.7, 115.5, 116.5, 122.1, 126.3, 130.3, 133.8, 140.6, 160.1, 161.6 (CH arom, Cq arom); 119.8 (C-5); 134.1 (CH methylidene); 167.7, 172.7 (C=O); 194.0 (C=S). Anal. (C₂₂H₂₁NO₅S₂) calcd: C 59.58; H 4.77; N 3.16; found: C 59.44; H 4.89; N 3.14.

3-[(5Z)-5-({4-[2-(3,4-Dimethoxyphenyl)ethoxy]phenyl}methylidene)-4-oxo-2-thioxothiazolidin-3-yl]propanoic Acid (2d)

Yield 30%; yellow solid; mp 204–207 °C; ¹H NMR (DMSO-d₆): δ 2.61 (t J = 7.5 Hz, 2H, CH₂COOH); 2.97 (t J = 6.8 Hz, 2H, CH₂); 3.70, 3.73 (2s, 6H, 2 OCH₃); 4.20–4.26 (m, 4H, NCH₂ and OCH₂); 6.81 (dd J = 8.15 and 1.7 Hz, 1H, arom); 6.86 (d J = 8.15 Hz, 1H, arom); 6.92 (d J = 1.7 Hz, 1H, arom); 7.10 (m, 2H, arom); 7.57 (m, 2H, arom); 7.74 (s, 1H, CH methylidene). ¹³C NMR (DMSO-d₆): δ 31.6, 35.1, 40.6, 69.6 (CH₂); 56.2, 56.3 (OCH₃); 112.6, 113.6, 116.4, 121.6, 126.2, 131.2, 133.8, 148.2, 149.4, 161.6 (CH arom, Cq arom); 119.7 (C-5); 134.0 (CH methylidene); 167.5, 172.6 (C=O); 193.9 (C=S). Anal. (C₂₃H₂₃NO₆S₂) calcd: C 58.33; H 4.90; N 2.96; found: C 59.45; H 5.02; N 3.04.

3-[(5Z)-5-({4-[3-(4-Methoxyphenyl)propoxy]phenyl}methylidene)-4-oxo-2-thioxothiazolidin-3-yl]propanoic Acid (2e)

Yield 34%; yellow solid; mp 178–181 °C; ¹H NMR (DMSO-d₆): δ 1.99 (m, 2H, CH₂); 2.62 (t J = 7.85 Hz 2H, CH₂COOH); 2.66 (t J = 7.95 Hz, 2H, CH₂C₆H₅); 3.70 (s, 3H, OCH₃); 4.02 (t J = 6.35 Hz, 2H, OCH₂); 4.22 (t J = 7.85 Hz, 2H, NCH₂); 6.83 (m, 2H, arom); 7.09 (m, 2H, arom); 7.12 (m, 2H, arom); 7.57 (m, 2H, arom); 7.74 (s, 1H, CH methylidene). ¹³C NMR (DMSO-d₆): δ 31.1, 31.2, 31.5, 40.6, 67.9 (CH₂); 55.7 (OCH₃); 114.5, 116.3, 126.1, 130.0, 133.6, 133.7, 158.2, 161.7 (CH arom, Cq arom); 119.6 (C-5); 134.0 (CH methylidene); 167.5, 172.5 (C=O); 193.8 (C=S). Anal. (C₂₃H₂₃NO₅S₂) calcd: C 60.38; H 5.07; N 3.06; found: C 60.55; H 5.18; N 2.95.

3-[(5Z)-5-{[4-(4-Phenoxybutoxy)phenyl]methylidene}-4-oxo-2-thioxothiazolidin-3-yl]propanoic Acid (2f)

Yield 35%; yellow solid; mp 167–170 °C; ¹H NMR (DMSO-d₆): δ 1.89–1.92 (m, 4H, 2 CH₂); 2.64 (t J = 7.4 Hz, 2H, CH₂COOH); 4.03 (t J = 5.95 Hz, 2H, OCH₂); 4.13 (t J = 5.8 Hz, 2H, OCH₂); 4.23 (t J = 7.4 Hz, 2H, NCH₂); 6.92–6.94 (m, 3H, arom); 7.11 (m, 2H, arom); 7.27–7.30 (m, 2H, arom); 7.62 (m, 2H, arom), 7.79 (s, 1H, CH methylidene), 12.42 (br s, 1H, COOH). ¹³C NMR (DMSO-d₆): δ 21.6, 25.8, 25.9, 31.4, 67.5, 68.2 (CH₂); 115.0, 116.2, 121.0, 126.0, 130.0, 133.5, 159.1, 161.5 (CH arom, Cq arom); 119.5 (C-5); 133.8 (CH methylidene); 167.3, 172.3 (C=O); 193.6 (C=S). Anal. (C₂₃H₂₃NO₅S₂) calcd: C 60.38; H 5.07; N 3.06; found: C 60.34; H 4.96; N 3.14.

3-[(5Z)-5-{[3-(4-Phenoxybutoxy)phenyl]methylidene}-4-oxo-2-thioxothiazolidin-3-yl]propanoic Acid (2g)

Yield 63%; yellow solid; mp 163–165 °C; ¹H NMR (DMSO-d₆): δ 1.89 (m, 4H, 2 CH₂); 2.64 (t J = 7.4 Hz, 2H, CH₂COOH); 4.03 (t J = 5.4 Hz, 2H, OCH₂); 4.10 (t J = 5.5 Hz, 2H, OCH₂); 4.23 (t J = 7.4 Hz, 2H, NCH₂); 6.90–6.94 (m, 3H, arom); 7.09 (m, 1H, arom); 7.18 (m, 2H, arom); 7.25–7.27 (m, 2H, arom); 7.46 (m, 1H, arom); 7.77 (s, 1H, CH methylidene), 12.53 (br s, 1H, COOH). ¹³C NMR (DMSO-d₆): δ 25.9, 26.0, 31.4, 40.7, 67.5, 68.0 (CH₂); 115.0, 116.7, 120.9, 122.9, 123.2, 129.9, 130.1, 131.2, 134.8, 159.3 (CH arom, Cq arom); 118.1 (C-5); 133.4 (CH methylidene); 167.2, 172.3 (C=O); 193.8 (C=S). Anal. (C₂₃H₂₃NO₅S₂) calcd: C 60.38; H 5.07; N 3.06; found: C 60.49; H 5.11; N 3.09.

3.2. Molecular Docking

X-ray crystal structure selection and preparation of the AKR1B1-idose complex (PDB-ID: 3V36) [27] and PTP1B (PDB-ID: 1Q6T) [29] were performed as described in our previous studies [9,10,11]. Molecular docking studies with GOLD (version 2020) [40] were carried out in the catalytic sites of PTP1B and AKR1B1, in the AKR1B1-idose complex, and in the allosteric binding pocket of PTP1B, as we previously reported [9,10,11]. A total of 25 docking poses were generated with default settings using the CHEMPLP scoring function [41]. The docking poses were minimized in LigandScout 4.4.3 using the MMFF94 force field [42]. The 3D pharmacophores were generated using LigandScout 4.4.3, and 3D depictions of protein–ligand complexes were created in MOE version 2022.2 (Chemical Computing Group, Montreal, QC, Canada).

3.3. In Silico ADME and Toxicity Predictions

The in silico ADMET evaluation for selected compounds 1g, 2e, 2f, and 2g was performed by using the SwissADME, PreADMET, and PhaKinPro web tools (available at www.swissadme.ch/index.php, https://preadmet.webservice.bmdrc.org, and https://phakinpro.mml.unc.edu, respectively; accessed on 21–24 January 2025), which incorporate predictive models for physico-chemical properties, drug-likeness, pharmacokinetics, and toxicity [30,31,32,33,34,37].

3.4. Assays, Expression, and Purification of AKR1B1

The activity of AKR1B1 was determined by following the decrease in absorbance at 340 nm due to the NADPH oxidation. The standard assay mixture contained 0.25 M sodium phosphate buffer, pH 6.8, 0.5 mM EDTA, 0.18 mM NADPH, 2.4 M ammonium sulphate, and 4.7 mM D,L-glyceraldehyde. One unit of enzyme activity refers to the amount of enzyme that catalyzes the conversion of 1 µmol of substrate under standard conditions. The expression and purification of human recombinant AKR1B1 was performed as previously described [43]; the purified enzyme has a specific activity of 5 U/mg.

3.4.1. Inhibition Studies on AKR1B1

IC₅₀ (concentration of compound necessary to determine 50% inhibition of enzyme activity) values were determined in the assay conditions described above, in the presence of 2 mM L-idose as substrate. In each assay, 10 mU of purified AKR1B1 (corresponding to 83 nM in the assay mixture, calculated based on AKR1B1 molecular weight of 34 Kda) were used. All compounds were dissolved in DMSO. Since this solvent exerts an inhibitory effect on AKR1B1 [43], its concentration in all assays was kept constant at 0.5% (v/v). For each compound, at least five different concentrations of inhibitor, each tested at least in triplicate, were analyzed. IC₅₀ values were determined by nonlinear regression analysis using Prism GraphPad 9.5.

3.4.2. Evaluation of AKR1B1 Inhibition Mechanisms

Reaction rates were measured using at least five different L-idose concentrations in the absence and presence of at least four different inhibitor concentrations. ^appK_M and ^appV_max were determined by nonlinear regression analysis applying the model “enzyme kinetic velocity vs. [substrate]/Michaelis–Menten equation” using Prism GraphPad 9.5. The inhibition constants K_i (apparent dissociation constant of the EI complex) and K_i′ (apparent dissociation constant of the ESI complex) were determined from secondary plots of 1/^appV_max and ^appK_M/^appV_max as a function of inhibitor concentration, respectively. In the case of tight-binding inhibitors (i.e., IC₅₀ values of the same order of magnitude as [E]), the residual activity measurements obtained at each substrate concentration were fitted by nonlinear regression analysis to the Morrison equation [44] for ^appK_i evaluation, applying the model enzyme kinetics-inhibition/tight inhibition Morrison equation using Prism GraphPad 9.5. K_i and K_i′ values were obtained from the ^appK_i measured at different substrate concentrations according to the following equation:

$${}^{\mathit{app}}\mathit{K}_i = {\displaystyle \frac{\left[S\right]+K_M}{\frac{K_M}{K_i}+\frac{\left[S\right]}{K_i^{\prime }}}}$$ (1)

relative to a general case of a tight-binding non-competitive inhibition model. The K_M value for L-idose used in Equation (1) was 4.0 mM [43]. The reversibility of the inhibitory action was evaluated by measuring the AKR1B1 activity after extensive dialysis (Amicon Ultrafiltration membrane (Sigma-Aldrich, St. Louis, MO, USA), 10 KDa cut off) of an enzyme sample (10 mU) previously treated with an inhibitor concentration able to reduce the enzyme activity to at least 75% of the control value. The recovered activity was compared to that measured for an AKR1B1 sample subjected to the same treatment in the absence of an inhibitor.

3.5. Assays, Expression, and Purification of PTP1B

The recombinant human PTP1B was obtained as previously described [11]. Briefly, the sequence codifying PTP1B (1-302 aa) was cloned in bacterial expression vector pNic28 in frame with a poly-His (6xHis) sequence, which was then used to transform the E. coli BL21 bacterial strain. The bacteria were grown and then incubated with IPTG (50 µg/L) to induce the expression of a fusion protein. The recombinant PTP1B was purified from bacterial lysate by affinity chromatography using a column packed with a Ni-NTA Agarose resin (Termo Fischer Scientific, Waltham, MA, USA). The fractions containing PTP1B were collected, concentrated by using centrifuge concentrators (Millipore, Burlington, MA, USA). After, the enzyme preparation was loaded on a Superdex G-75 column (Cytiva, Marlborough, MA, USA) and eluted in 50 mM Tris-HCl buffer, containing 150 mM NaCl and 0.5 mM mercaptohetanol. The purity of protein preparations was analyzed by SDS–PAGE according to Laemmli [45]. The solutions containing the proteins were aliquoted in 500 µL fractions and stored at −80 °C.

3.5.1. Inhibition Studies on PTP1B

To determine the activity of PTP1B, enzymatic assays were carried out using the synthetic substrate pNPP, which was dissolved in 0.075 M β,β-dimethylglutarate buffer (pH 7.0). Each assay was initiated by adding an aliquot of the enzyme to the assay solution and then stopped after an appropriate time using a NaOH solution. The amount of pNP released was determined spectrophotometrically by reading each sample at 400 nm [46].

3.5.2. Evaluation of PTP1B Inhibition Mechanisms

Detailed kinetic analyses were carried out to evaluate the mechanism of compounds 1g, 2e–g on PTP1B. To evaluate if compounds acted as reversible or irreversible inhibitors, a dilution assay was carried out. Briefly, an aliquot of PTP1B was incubated for 60 min at 37 °C in the presence of a saturating amount of each PTP1B inhibitor. Then, the residual activity of the enzyme was determined by diluting an aliquot of the enzyme–inhibitor solution in 2 mL of the assay buffer containing 2.5 mM of substrate and measuring the amount of pNP released using a spectrophotometer.

The mechanism of action of compound 2e was determined by measuring the dependence of K_M and V_max from the inhibitor concentration. Data obtained were fitted using both the Michaelis–Menten and Lineweaver–Burk equations. The K_i value was calculated by fitting experimental data with the appropriate equations.

3.5.3. Continuous PTP1B Inhibition Assays

Continuous assay tests were carried out with a final volume of 1.5 mL. Each assay sample contained a buffer of pH 7.0 and substrate (5 mM pNPP). The reactions were started by mixing the enzyme into a solution. In situ pNP release was monitored by reading the absorbance of samples at 405 nm for 40 min at 25 °C. The absorbance values were converted into the pNP concentration using the ε_mM value of pNPP (18 mM⁻¹cm⁻¹). Data were then fitted to Equation (S1) (see Supplementary Materials) to determine the apparent first-order rate constant for establishment of the enzyme–inhibitor equilibrium (k_a). Further continuous tests were carried out using a fixed inhibitor concentration and increasing substrate concentrations to determine the competitive or non-competitive nature of the inhibitors. The K_i and K′_i values were determined using an appropriate equation derived from kinetic models.

3.6. Cell Cultures and Assays

Human lens epithelial B3 line (HLE) cells were purchased from American Type Culture Collection (Rockville, MD, USA) and cultured at 37 °C in a humidified atmosphere, in the presence of 5% CO₂, in an Eagle’s modified minimum essential medium (MEM) containing 5.5 mM D-glucose, supplemented with 20% (v/v) FBS, 50 mU/mL penicillin/streptomycin, 2 mM glutamine and 100 μg/mL hygromycin. Cells were grown in the indicated medium until they were 70% confluent. Before experiments, cells were incubated for 24 h in MEM containing 0.5% FBS, 50 μg/mL gentamicin, 2 mM glutamine, and 100 μg/mL hygromycin. Cells at passages 15–20 (30–40 residual doublings), plated at a density of 20,000 cells/cm², were used for experiments.

The human Muller glial line (MIO-M1) cells were a gift from Prof. Massimo Dal Monte (University of Pisa). MIO-M1 cells were cultured at 37 °C in the presence of 5% CO₂, in a humidified atmosphere, in a Dulbecco’s Modified Eagle Medium (DMEM) containing 5.5 mM D-glucose supplemented with 20% (v/v) FBS, 50 mU/mL penicillin/streptomycin, 2 mM glutamine, and 1% non-essential amino acid solution. Cells were grown in the indicated medium until they were 80% confluent. Before experiments, cells were incubated for 24 h in DMEM containing 0.5% FBS, 50 mU/mL penicillin/streptomycin, 2 mM glutamine, and 1% non-essential amino acid. Cells at passages 14–19, plated at a density of 18,000 cells/cm², were used for experiments.

C2C12 and HepG2 cells were purchased by ATCC and grown at 37 °C in 5% CO₂ and a humidified atmosphere in the presence of high glucose DMEM (Euroclone, Pero, Italy) supplemented with 10% FBS (Euroclone), 50 mU/mL penicillin/streptomycin, and 2 mM glutamine (Merck, Darmstadt, Germany).

3.6.1. Measurements of Cell Viability

HLE, MIO-M1, C2C12, and HepG2 cells’ viability was assessed after 48 h incubation at 37 °C in the proper culture medium (see above), supplemented with 0.05% DMSO alone or in the presence of the selected compounds or the reference drug. Then, cells were washed with phosphate-buffered solution (PBS, Merck Millipore, Darmstadt, Germany) and incubated for 1 h with 0.5 mg/mL of thiazolyl blue tetrazolium bromide (MTT, Merck Millipore) dissolved in a phosphate-buffered solution pH 7.4. The formazane crystals were then dissolved by the addition of 0.04 N HCl in isopropanol, and the absorbance of the solution at 563 nm was measured using a microplate reader [47].

3.6.2. Assessment of Insulin Signalling via SDS-PAGE and Western Blot Analysis

To evaluate the impact of tested compounds on the insulin signalling pathway, we analyzed the phosphorylation status of the kinase Akt using the Western blot method. Briefly, C2C12 cells were seeded in 35 mm dishes and grown until they reached 80% confluence. The cells were then washed with PBS and incubated in starvation medium for 24 h. After starvation, the cells were stimulated with 10 nM insulin or with 10 μM compound for 30 min and subsequently lysed in 150 μL of 1× Laemmli sample buffer. Before lysis, the cells were washed with cold (4 °C) PBS. Protein lysates were collected, transferred to 1.5 mL conical Eppendorf tubes, and boiled for 5 min to ensure complete protein denaturation. The boiled samples were placed on ice for 15 min and then stored at −20 °C. The following day, protein samples were analyzed by SDS-PAGE, loading 10 μL of each sample onto 4–20% Mini-PROTEAN^® TGX™ Precast Protein Gels (Bio-Rad, Hercules, CA, USA). After approximately 1 h of electrophoresis, the run was stopped, the gels were removed, and the separated proteins were transferred onto PVDF membranes (Immun-Blot PVDF membrane, Bio-Rad) using a Trans-Blot Turbo system (Bio-Rad). Following transfer, the membranes were incubated for 30 min with a blocking solution consisting of 5% BSA in PBS containing 0.1% Tween 20.

To evaluate both the phosphorylation status and total levels of Akt, the membranes were incubated overnight with primary antibodies recognizing phosphorylated Akt (9271S, Phospho-Akt (Ser473) Antibody, Cell Signalling Technology, Danvers, MA, USA) and total Akt (9272S, Akt Antibody, Cell Signalling Technology). The next day, the membranes were extensively washed with PBS containing 0.05% Tween 20 to remove unbound antibodies and then incubated for 1 h at 4 °C with HRP-conjugated secondary antibodies.

After incubation with secondary antibodies, the membranes were washed again with PBS containing 0.05% Tween 20 and analyzed using a luminometer (Imager 600 Luminescence Image Analyzer, GE Amersham, Little Chalfont, UK). Before detection, the PVDF membranes were incubated with 2 mL of ECL START Western blotting Detection Reagent. The intensity of each band was determined using the Amersham Imager 600 Analysis Software. Each test was carried out in triplicate. The data obtained were normalized with respect to the control samples.

3.6.3. Determination of Sorbitol Content

HLE and MIO-M1 cells were washed with PBS, supplemented with 1 mM PMSF, harvested, and lysed by means of three freezing/thawing cycles. The supernatant obtained after centrifugation at 10,000× g at 4 °C per 30 min (crude cell extract) was supplemented with 4 M ice-cold perchloric acid. Samples were then neutralized using 5 M KOH, and the sorbitol content was measured as described [48]. Protein content was measured in the crude extracts according to Bradford [49].

3.6.4. Glucose Uptake Assay

C2C12 cells seeded in 24-well plates were serum-starved for 24 h and then stimulated with 10 nM insulin, and/or with 10 μM of compounds 1g, 2e–g for 30 min. After this time, the starvation medium was removed, the cells were washed with PBS, and they were then incubated with 1 mL starvation medium containing 40 μM of 2-NBDG. After 3 h, the medium was removed, and the cells were washed with fresh PBS and trypsinized. Detached cells were collected by centrifugation at 1000 rpm, washed with PBS, and analyzed using a FACS Canto II flow cytometer (BD Biosciences, Franklin Lakes, NJ, USA). All tests were carried out in quadruplicate.

3.6.5. Evaluation of the Mitochondrial Membrane Potential

C2C12 cells were seeded on 6-well plates (50,000 cells for each well) and incubated at 37 °C. Then, cells were treated for 30 min with a growth medium containing 1 μM FCCP or 25 μM of compounds 1g, 2e–g. After incubation, cells were tripsinized, collected, and analyzed using a flow cytometer (BD FACS Canto II Analyzer, BD Bioscience) and the FlowJo software (v.10). For each test, 10,000 events were acquired.

3.6.6. Measurement of Lipid Accumulation

Oil Red O stock solution was prepared by dissolving 0.2 g of dye in 40 mL of 2-propanol (0.5% w/v). The working solution was obtained by diluting the stock solution in distilled water (2:3 Oil Red/water). Before use, the working solution was filtered through 0.22 μm syringe filters and stored at room temperature. HepG2 cells (70,000 cells per well) were seeded in 24-well plates and incubated at 37 °C in a 5% CO₂ atmosphere. The following day, the cells were washed with PBS and incubated with 2 mL of fresh starvation medium, containing or not containing 0.4 mM oleic acid, then stored again at 37 °C. After 24 h, the medium containing oleic acid was removed, and the cells were washed with PBS. Then, 2 mL of starvation medium containing 25 μM of compound 1g or 2e–g was added to the plate, which was subsequently incubated at 37 °C for an additional 72 h. After this period, the cells were washed with PBS and then analyzed to evaluate the lipid content using the Oil Red O staining method. For this purpose, the medium was removed and replaced with 250 μL of 60% 2-propanol solution. After a 5 min incubation, the isopropanol was withdrawn and replaced with 500 μL of Oil Red O working solution. After 30 min, the Oil Red O solution was removed, and the plates were washed five times with distilled water. Then, 500 μL of 100% 2-propanol was added to each well to aid in the solubilization of the dye bound to intracellular lipids. After a 10 min incubation, the absorbance of the solutions was determined using the iMark™ microplate reader (Bio-Rad), measuring the absorbance at 510 nm. Each experiment was carried out in quadruplicate. After reading, each plate was washed with distilled water to completely remove the Oil Red O dye. Next, 300 μL of Bradford reagent (Sigma Aldrich) was added to each well. The plates were incubated at room temperature for 10 min, then washed with water. The plates were scanned, and the wells were analyzed with ImageJ 1.54g to quantify the protein content in each well. Finally, the data from the Oil Red O test were normalized to the protein content. All data were normalized relative to the control experiment.

3.6.7. Statistical Analysis of Data

When required, statistical analysis was performed by using the Unpaired Student’s t-test or one-way Anova with the Dunnet post hoc test, as specifically detailed in each figure legend.

4. Conclusions

Continuing an ongoing search for potential multitarget antidiabetic agents, a new series of (5-arylidene-4-oxothiazolidin-3-yl)alkanoic acids (1a–g, 2a–g) were synthesized and evaluated, with the main aim of investigating the relationships between their activity and a more extended 5-arylidene moiety, which was modified by means of the introduction of methoxy substituents on the distal phenyl ring and/or elongation of the linker chain between the two aromatic rings. Regarding the AKR1B1/PTP1B inhibitory activity, the newly synthesized compounds 1a–g and 2a–g showed excellent inhibitory activity toward AKR1B1, along with less potent effects against PTP1B. Compounds 1g and 2e–g displayed an appreciable capability to inhibit both target enzymes; however, compared with the previously reported analogues [9,10,11], they did not lead to any substantial improvement in inhibitory potency. Surprisingly, despite their scarce capability to improve insulin signalling in murine myocytes C2C12 by inhibiting PTP1B, compounds 2e, 2f, and, to a lesser extent, 2g markedly stimulated cellular glucose uptake in the same cells; no appreciable differences were observed when the cells were treated with each compound alone or in combination with insulin, clearly indicating an insulin-mimetic activity. It could be inferred that, at the concentration (10 µM) used in the ex vivo assays, which is very close to or lower than the observed IC₅₀ values, the effect of PTP1B inhibition on improving cellular sensitivity to insulin may be insignificant or not detectable. Therefore, compounds 2e–g must be capable of stimulating cellular glucose uptake in an insulin-independent manner. Assays performed in C2C12 myocytes and HepG2 liver cells indicated that 3-(5-arylidene-4-oxo-2-thioxothiazolidin-3-yl)propanoic acid derivatives 2e–g can produce an appreciable increase in mitochondrial potential and alter pyruvate metabolism, suggesting that the observed significant increase in both glucose uptake and fatty acid degradation might be related to these mitochondrial activities. Compound 2e was also able to significantly impair sorbitol accumulation in both human lens epithelial and retinal glial cells, by strongly inhibiting AKR1B1; in the same assay, another potent AKR1B1 inhibitor, acetic acid derivative 1g, showed excellent activity in lens epithelial cells, reaching activity levels like that of the reference drug sorbinil.

The finding that compounds 2e–g could act as insulin-mimetic agents through insulin-independent mechanisms is significant and, therefore, could motivate further studies. Moreover, it could be worth further investigating the relationships of these mechanisms with certain structural features of 4-thiazolidinone derivatives, since analogous 2,4-thiazolidinediones were shown to act differently on pyruvate metabolism [38]. The substitution pattern of the 5-arylidene moiety might play a central role by affecting these concurrent mechanisms and modulating the multifaceted antidiabetic potential of this class of compounds. Moreover, the impact of combined effects elicited by alterations of pyruvate mitochondrial pathways on such a complex disease as DM has not yet been sufficiently studied, and thus the compounds here reported might provide tools for further investigations.

Finally, it is worth noting that thiazolidine derivatives have often been the object of debate on their suitability as starting hits in the process of drug discovery, especially due to concerns regarding their metabolic and toxicity profiles. However, it was demonstrated that the in vivo behaviour of these derivatives can be effectively modulated by a substitution pattern of the heterocyclic core, since it depends on the entire molecular entity, and, therefore, thiazolidine and other five-membered heterocycle derivatives should not be subject to inappropriate negative bias [50,51]. Interestingly, the insertion of the pharmacophoric polar carboxylic chain on N-3 of the thiazolidinone scaffold in compounds 1a–g and 2a–g, as well as in the analogues that we previously reported [9,10,11,25,26], can be responsible for decreased lipophilicity and increased metabolic stability; therefore, these thiazolidinone derivatives could be less prone to producing potentially toxic metabolites [50]. Overall, our structure–activity relationship studies of (5-arylidene-4-oxothiazolidin-3-yl)alkanoic acid derivatives as inhibitors of AKR1B1 and PTP1B demonstrated that the substitution pattern in both positions 3 and 5 of the thiazolidinone scaffold can deeply influence the inhibitory potency as well as the mechanism of action of these compounds. Compelling evidence indicates that substituted 4-thiazolidinones, obtained by means of appropriately designed derivatization [9,10,11,25,26,50,51], can be attractive compounds in drug design and development, thus encouraging us to further investigate this class of enzyme inhibitors and their multifarious activity profiles.

Supplementary Materials

The following supporting information can be downloaded at: https://www.mdpi.com/article/10.3390/ph18121863/s1, Figures S1–S28: ¹H NMR and ¹³C NMR spectra of compounds 1a–g and 2a–g; Figures S29–S32: Rate measurements of the AKR1B1 dependent reduction in L-idose in the presence of compounds 2e, 2f, 2g and 1g; Figure S33: Dilution assay of PTP1B with selected compounds 1g, 2e–g; Figures S34–S36: Continuous inhibition of PTP1B by compounds 1g, 2f and 2g at pH 7.0 and 25 °C; Figures S37–S39: Continuous inhibition of PTP1B by compound 1g, 2f and 2g in the presence of increasing concentrations of substrate; Figure S40: Dependence of Km and Vmax from concentration of compound 2e; Figure S41: Dependence of Km from the concentration of compound 2e; Figure S42: Dependence of Vmax from the concentration of compound 2e; Figure S43: Lineweaver-Burk plot of compound 2e; Figure S44: k_a secondary plot relative to compound 1g; Figure S45: v₀ and k_a secondary plots relative to compound 2f; Figure S46: v₀ and k_a secondary plots relative to compound 2g; Figure S47: Effect of compounds 1g, 2e–g on cell viability (C2C12 cells); Figure S48: Lipid accumulation assay; Figure S49: Effect of compounds 1g, 2e–g on cell viability (MIO-M1 and HLE cells); Scheme S1: Mechanism of action of compound 1g (competitive slow-binding model); Scheme S2: Mechanism of action of compound 2f. References in Supplementary Materials can be found in [52,53].

Author Contributions

Conceptualization, R.M. (Rosanna Maccari), R.O., G.W., P.P., and A.D.C.; methodology, R.M. (Rosanna Maccari), G.W., P.P., and A.D.C.; validation, R.M. (Rosanna Maccari), R.O., G.W., P.P., and A.D.C.; formal analysis, V.T.; R.M. (Roberta Moschini), F.B., F.F., and G.S.; investigation, R.O., V.T., R.M. (Roberta Moschini), F.B., F.F., F.I., G.S., and R.S.; data curation, V.T., R.M. (Roberta Moschini), F.B., F.F., F.I., G.S., and R.S.; writing—original draft preparation, R.M. (Rosanna Maccari), R.O., and V.T.; writing—review and editing, R.M. (Rosanna Maccari), G.W., P.P., and A.D.C.; visualization, R.M. (Rosanna Maccari), R.O., and V.T.; supervision, R.M. (Rosanna Maccari), P.P., and A.D.C.; funding acquisition, R.M. (Rosanna Maccari), P.P., and A.D.C. All authors have read and agreed to the published version of the manuscript.

Data Availability Statement

The original contributions presented in this study are included in the article and Supplementary Materials. Further inquiries can be directed to the corresponding author.

Conflicts of Interest

The authors declare no conflicts of interest.

Funding Statement

This work was funded in part by the University of Messina (FFABR UNIME 2023), University of Florence (“Fondi di Ateneo 2024”), and University of Pisa (“Fondi di Ateneo 2024”).