Introduction

The worldwide prevalence of diabetes mellitus has been estimated to 2.8 % in 2000 and is expected to raise by 4.4 – 9 % within the next two decades.^1,2 Careful predictions anticipate 366 million people worldwide suffering from the “diabetes epidemic” in 2030. Even though various therapy options have been developed for the treatment of diabetes mellitus including compounds interfering with hydrocarbon digestion and glucose utilisation, insulines, insulin-releasing agents, and recently drugs compensating insulin resistance, a stringent blood glucose control as maintained under physiological conditions could not be achieved.^3-5 In consequence, in diabetic patients non-physiological hyperglycaemic events are responsible for severe long-term complications including retinopathy, nephropathy, neuropathy, cataract and angiopathy.^1,6,7 Under these elevated glucose levels an increased flux of glucose through the polyol pathway occurs inducing various biochemical imbalances, thereby strongly contributing to the onset of diabetic complications. In particular, the polyol pathway consists of two enzymes: the first and rate-limiting one, aldose reductase (ALR2) catalyses the conversion of glucose to sorbitol using NADPH as reducing cofactor, the second enzyme, sorbitol dehydrogenase, oxidizes sorbitol to fructose NAD⁺-dependently.^8,9 Thus, increased polyol pathway activity is accompanied by generation of osmotic and oxidative stress causing various pathological interferences with cytokine signalling, regulation of apoptosis as well as activation of kinase cascades.¹⁰ For instance, recent experimental observations provided evidence that under increased glucose metabolism via the polyol pathway p38-MAP kinase shows increased activity causing nerve conduction deficits and, thus, leads to neuropathy.¹¹ Secondly, increased protein kinase C activity under elevated polyol pathway flux has been shown to induce smooth muscle cell proliferation of blood vessels being in agreement with atherosclerosis. This also explains estimations that 75 – 80 % of adults with diabetes die from complications of atherosclerosis.^12,13 In addition, extended polyol pathway

activity has been shown to provoke endothelial cell damages by increased oxidative stress and thereby contribute to atherosclerotic complications.¹⁴ Aggravatingly enough, increasing amounts of fructose derived from the polyol pathway contribute to the formation of advanced glycosylated endproducts (AGEs) and thereby lead to pathological changes of proteins functionally affected by covalent modification.^15,16 Altogether, the pathological activity of ALR2 plays a key role in the development of diabetic complications and thereby represents an excellent drug target. In fact, in vitro and in vivo studies suggest a clear benefit of the administration of aldose reductase inhibitors (ARIs) in various model systems exposed to high glucose levels as well as during the therapy of diabetic patients.^7,17-21 Thus, extensive effords have been performed to develop appropriate drug candidates. Most of these inhibitors can be classified according to their negatively charged anchor groups into carboxylate-type or hydantoin-type inhibitors. However, most of the inhibitors evolved from these approaches failed in clinical trials either due to poor bioavailability or selectivity properties.^1,22 It has been argued that carboxylate-type inhibitors are inappropriate due to their more acidic properties compared to hydantoins. Under physiological conditions carboxylate groups will be almost completely ionised impairing their ability to cross biological membranes.²³ Nevertheless, a recently published carboxylate-type ligand, lidorestat, exhibits a favourable pharmacokinetic profile resulting in desirable tissue penetration. Accordingly, even though possessing this ionisable group, ligands can be optimized with respect to sufficient penetration behaviour.²⁴ Besides an appropriate pharmacokinetic profile, potential drug candidates should also possess high selectivity to aldose reductase in comparison to the highly related aldehyde reductase (ALR1) which shares a sequence identity of about 65 %.^24-28 As ALR1 detoxifies various aldehydes derived from oxidative stress including 3-deoxyglucosone and methylglyoxal by conversion to their corresponding non-reactive alcohols,²⁹ its reducing activity is of utmost physiological importance, in particular under increased oxidative conditions as experienced during a diabetic situation.

ALR2 (EC 1.1.1.21) is a 36 kDa (β/α)8-TIM-barrel shaped aldo-keto reductase with the active site located at the C-terminal region of the enzyme.23,25,26,30,31 The deeply buried substrate binding pocket comprises residues presumably participating in the catalytic mechanism (Tyr 48, Lys 77, His 110). Furthermore, the nicotinamide moiety of NADP⁺ and Trp 111 interact with the head group of most described ligands. Additionally, hydrophobic contacts can be formed by the side-chains Trp 20, Val 47, Trp 79, and Trp

219. This catalytic site is usually addressed by hydrophilic, negatively charged building blocks. Ligands decorated at the opposing terminal end with appropriate hydrophobic groups exhibit to varying degree the ability to induce an opening of distinct pockets.

They are referred to as “specificity pockets” formed in consequence of different rotameric states of Ala 299, Leu 300 and Phe 122 at the solvent-exposed face and the side chain of Trp 111 facing the center of the TIM-barrel.

ALR2 converts various aldehydes (including glucose under diabetic conditions) to their corresponding alcohols using NADPH as reducing cofactor. Even though the exact mechanism is currently under discussion, NADPH donates a hydride ion to the carbonyl carbon of the aldehyde. Most likely, this step is followed by a subsequent transfer of a proton from one of the neighbouring acidic protein residues to the intermediately formed substrate anion.^32,33

An ultra-high resolution structure of ALR2 in complex with the carboxylate-type inhibitor IDD 594 (analogue of 3, Fig. 2.1 with a thioamide instead of an amide group and a chlorine atom replaced by fluorine) has been refined to a resolution of 0.66 Å.

The crystal structure provides evidence for the protonation states of the titratable and catalytically relevant residues involved in inhibitor binding.²⁵

Four successful in-silico screening studies have been reported on ALR2 up to now to find novel lead compounds.^34-37 Recently, a virtual screening study has been performed in our laboratory based on the coordinates and protonation states observed in the ultra-high resolution crystal structure. This computer screening resulted in six new carboxylate-type leads, among them two ligands in the low- and submicromolar affinity range.³⁷ Both ligands contain a nitro-substituted terminal moiety linked to a 5-membered heterocycle connected via an alkyl spacer to the carboxylic head group (1, 2, Fig. 2.1). In all screening studies reported so far on ALR2, binding affinity of identified hits was discussed with respect to binding geometries obtained from docking with or without subsequent force-field minimization or molecular dynamics simulations.

However, as could be demonstrated in several cases, surprising differences between the docking predictions and the subsequently determined crystal structures have been reported.^38-40 Accordingly, it is highly advisable to determine the crystal structure of virtual screening hits in complex with the target protein prior to embarking into a synthesis program with the goal to optimize binding properties. Furthermore, detailed insight into the thermodynamic driving forces of inhibitor binding can be very supportive to select the best screening hit for a synthesis follow-up program.⁴¹

N O N

O OH

O₂N ^O _{N N}

O S

O O₂N OH

NH

O O OH

O

Br F Cl

NH

O O OH

O

Cl NO₂

NH

O O OH

O

F S

N F

F F CH₃

O NH

S

CH₃ CH₃ O

O NH

O OH

N

N O N

O OH

N O N

O Cl OH

1 2

3 (IDD 388) 4 (IDD 393)

5 (IDD 552) 6 (IDD 384)

7 8

Figure 2.1. Chemical formulae of inhibitors discussed in this study. Ligands 1 and 2 have been identified by virtual screening. 3: IDD 388, 4: IDD 393, 5: IDD 552, 6: IDD 384. Ligands 7 and 8were selected as analogues to 1 in order to evaluate the influence of the nitro substituent at the terminal aromatic moiety.

In the present study, we report on the crystallographically determined binding modes of 1 and 2 in complex with ALR2 together with the thermodynamic driving forces responsible for binding as available from isothermal titration calorimetry (ITC).

Furthermore, we identified initial structure-activity relationships of both leads by facing their properties to two inhibitors of the IDD series (3, 4, Fig. 2.1, respectively).