The critical role of polyamines in key processes such as cell growth, differentiation and macromolecular synthesis makes the enzymes involved in their synthesis potential targets in the treatment of certain types of cancer and parasitic diseases. Here we present a study on the inhibition of human and Leishmania donovani ODC (ornithine decarboxylase), the first committed enzyme in the polyamine biosynthesis pathway, by APA (1-amino-oxy-3-aminopropane). The present study shows APA to be a potent inhibitor of both human and L. donovani ODC with a Ki value of around 1.0 nM. We also show that L. donovani ODC binds the substrate, the co-enzyme pyridoxal 5′-phosphate and the irreversible inhibitor α-difluoromethylornithine (a curative agent of West African sleeping sickness) with less affinity than human ODC. We have also determined the three-dimensional structure of human ODC in complex with APA, which revealed the mode of the inhibitor binding to the enzyme. In contrast with earlier reports, the structure showed no indication of oxime formation between APA and PLP (pyridoxal 5′-phosphate). Homology modelling suggests a similar mode of binding of APA to L. donovani ODC. A comparison of the ODC–APA–PLP structure with earlier ODC structures also shows that the protease-sensitive loop (residues 158–168) undergoes a large conformational change and covers the active site of the protein. The understanding of the structural mode of APA binding may constitute the basis for the development of more specific inhibitors of L. donovani ODC.
- drug target
- inhibitor complex
- polyamine synthesis
- tropical parasite
The first and rate-limiting step in the polyamine biosynthetic pathway is the conversion of ornithine into putrescine. This reaction is catalysed by ODC (ornithine decarboxylase), a PLP (pyridoxal 5′-phosphate, vitamin B6)-dependent enzyme, known to form obligate dimers in eukaryotes . The critical role of polyamines in key processes such as cell growth, differentiation and macromolecular synthesis makes the enzymes involved in their synthesis potential targets in the treatment of certain types of cancer [2,3] and parasitic diseases [4–6]. Thus the irreversible ODC inhibitor, DFMO (α-difluoromethylornithine) is currently used for the treatment of patients with advanced West African sleeping sickness, caused by the protozoan parasite Trypanosoma brucei gambiense [6,7]. DFMO also exhibits some activity against other parasitic protozoa, like Plasmodium falciparum, Plasmodium berghei  and Leishmania donovani . Another compound, APA (1-amino-oxy-3-aminopropane), an isosteric analogue of putrescine, is a more potent inhibitor of ODC with inhibition constants in the nM range [10,11]. APA and its derivatives exert antiproliferative effects on various tumour cells and parasites [12–17]. Similar to other amino-oxy analogues, the inhibition of ODC by APA has been suggested to proceed through the formation of an oxime with PLP in the enzyme's active-site [10,18,19].
The X-ray structures of ODC from mouse , T. brucei [21,22] and human  show a dimeric enzyme in which each N-terminal domain is a TIM-like α/β-barrel and each C-terminal domain forms a β-sheet. The two active sites are found at the dimer interface, between the N-terminal domain of one monomer and the C-terminal domain of the other, with Lys69 and Cys360 contributing to each active site from opposite monomers [24,25]. The structures also identified the residues important for PLP binding and interactions with DFMO [21,26]. In the present study, we show the kinetic parameters of inhibition of human and L. donovani ODC by APA, as well as the crystal structure of the complex between APA and human ODC. The structure shows that APA is bound in the substrate-binding pocket in close proximity to PLP. However, no oxime formation is observed between the inhibitor and PLP. Examination of the structure of L. donovani ODC using homology modelling shows a high degree of similarity between the human and parasite enzyme active sites, which may explain the similarity in the kinetic parameters of inhibition of the two enzymes by APA.
Subcloning of L. donovani and human ODC gene fragments
A full-length L. donovani ODC gene, subcloned into pET-30 Xa/LIC vector (Novagen), and a human ODC cDNA, subcloned into pUC, were amplified by PCR, using the primer pairs: forward: 5′-ATGGGTGATCATGACGTCG-3′ and reverse: 5′-TTATCACTCGCTCACACACCTCA-3′; and forward: 5′-ATGAACAACTTTGGTAATGAAGAGTTTG-3′ and reverse 5′-TACACATTAATACTAGCCGAAGCACAG-3′ respectively. PCR was performed in a 20 μl reaction volume containing 3 ng of the template, 0.5 μM each of forward and reverse primers, and 1× Phusion™ High-Fidelity PCR Master Mix (Finnzymes, Finland). Amplification was performed in a temperature gradient DNA Thermal Cycler under the following conditions: initial denaturation (98°C for 30 s), and then 29 cycles of 98°C for 10 s, 48–72°C for 30 s (annealing) and 72°C for 10 s (extension). A final extension was carried out at 72°C for 10 min. The reaction mixture was further incubated with 0.8 units of Taq DNA Polymerase (Fermentas) in order to introduce sticky ends for subsequent cloning into pEXP5-NT/TOPO® TA vector (Invitrogen). The amplified DNA fragments were cloned into pEXP5-NT/TOPO® TA vector according to the manufacturer's instructions. PCR was used to identify colonies with correctly oriented inserts, which then were sequenced. The constructs without any introduced mutations were used for transformation of BLR (DE3) chemically competent Escherichia coli.
Expression and purification
A single colony of transformed E. coli was grown overnight at 37°C in Luria–Bertani medium containing 100 μg/ml of ampicillin. For L. donovani ODC, the overnight culture was diluted 1.5:100 and incubated for 22 h at 20°C. The expression of L. donovani ODC was very high even without the addition of the inducer, IPTG (isopropyl β-D-thiogalactopyranoside) (leaky expression). For human ODC, the overnight culture was diluted 1.5:100 and incubated at 37°C until D600 reached 0.5. The culture was then induced with 1 mM IPTG and incubated further for 4 h at 37°C. Bacteria cultures were harvested by centrifugation at 3840 g at 4°C for 15 min and the pellet was stored at −80°C until use.
The frozen cell pellet from a 1 litre culture was thawed on ice and resuspended in 10 ml of lysis buffer [50 mM Tris/HCl, pH 8.0, 300 mM NaCl, 1 mM DTT (dithiothreitol), 1 mM PMSF and 0.02% Brij 35] containing 5 mM imidazole. The lysis buffer for L. donovani did not contain imidazole. The cells were sonicated and the lysate was centrifuged at 36000 rev./min for 1 h at 4°C (Beckman 45 Ti rotor). The supernantant was mixed with 2 ml of a 50% (w/v) Ni-NTA (Ni2+-nitrilotriacetate) agarose (Qiagen) and left at 4°C with shaking for 2 h. The agarose was then loaded in a column. For L. donovani ODC the column was washed with lysis buffer containing increasing concentrations of imidazole: 8 mM (15 ml) and 20 mM (4 ml), and the His6-tagged L. donovani ODC was eluted with 8 ml of buffer containing 250 mM imidazole. For human ODC, the column was washed with lysis buffer first containing 20 mM (15 ml) and then 40 mM (4 ml) imidazole, and the His6-tagged human ODC was eluted with 8 ml of buffer containing 250 mM imidazole. The 8 ml of affinity-purified proteins were diluted with 42 ml of buffer A (20 mM Tris/HCl, pH 8.0, 1 mM PMSF and 2 mM DTT) and passed through a Mono Q column (5/50 GL, Amersham). The proteins were eluted by a salt gradient using buffer B (20 mM Tris/HCl, pH 8.0, 1 mM PMSF, 2 mM DTT and 1 M NaCl). The peaks on the chromatogram were analysed by Tris/HCl/SDS/4–20%-(w/v)-PAGE, and the fractions from a single peak containing the protein of interest were concentrated and run on a gel filtration column (Superdex 200 HR 10/30) with 25 mM Hepes (pH 7.2), 2 mM DTT, 0.5 mM EDTA, 0.02% Brij 35 and 1 mM PMSF, with and without 0.02 mM PLP. The gel filtration fractions of each protein were concentrated to approx. 6 mg/ml, aliquoted (50 μl) and stored at −20°C or 4°C.
Purified preparations of human and L. donovani ODC were used for kinetic analyses. The ODC activity was assayed at 37°C by measuring the release of 14CO2 from L-[1-14C]ornithine in 0.1 M Tris/HCl (pH 7.5) containing 2.5 mM DTT, 0.1 mM EDTA and 0.1% Tween 80 . Determination of the Km/Vmax for the substrate and Kd for the coenzyme was achieved by varying the concentrations of L-ornithine and PLP respectively in the assay.
The Ki of APA was determined by measuring its effects at various concentrations on the apparent Km for the substrate. The kinetics of time-dependent, irreversible inhibition of ODC by DFMO was determined by incubating the enzyme with various concentrations of DFMO for different times, whereupon remaining ODC activity was measured during 5 min. The results of the kinetics analysis are shown in Table 1.
Crystallization and data collection
Attempts to crystallize the human ODC–APA complex using the conditions reported previously  did not succeed. A screen was performed in the search for new crystallization conditions. The protein solution used in the screens contained 6 mg/ml protein (in 25 mM Hepes, pH 7.2, 2 mM DTT, 0.5 mM EDTA, 0.02% Brij 35 and 1 mM PMSF) and was pre-incubated with 2 mM APA for 10 min at room temperature (22°C) before setting up crystallization drops. Initial conditions were found using the Index Screen (HR2-144; Hampton Research) by mixing equal volumes of protein and well solution in sitting drops (500 μl of well solution) at 15°C. Needle clusters appeared after 2 days in 25% (v/v) PEG3350, 0.2 M ammonium acetate and 0.1 M Bis-Tris (pH 6.5). Further optimization of this condition yielded fairly single crystals in 18% (v/v) PEG3350, 0.2 M ammonium acetate, 0.1 M Bis-Tris, pH 6.5, and 3 mg/ml protein concentration. Data were collected to 3.0 Å (1 Å=0.1 nm) at beamline I911-3, Max Lab, Lund. The crystals belonged to space group P212121, with cell dimensions a=60.5, b=104.8, c=137.4, which are similar to the space group and cell constants of the crystals of the earlier structure (a=61.7, b=107.4, c=139.7; ). The structure was solved by molecular replacement using the co-ordinates of the native human ODC (PDB code 1D7K) with all water molecules removed.
For the human ODC–PLP–APA complex, 6 mg/ml protein, gel-filtered in a buffer containing 0.02 mM PLP, was pre-incubated with 2 mM APA for 10 min at room temperature. Equal volumes of protein and well solution containing 25% (v/v) PEG3350, 0.2 M ammonium acetate and 0.1 M Bis-Tris (pH 6.5) were mixed in sitting drops at 15°C. Needle clusters appeared after 2 days. However, attempts to get single crystals from these clusters were not successful. Subsequently, the buffer in the well solution was changed to Mes (pH 6.5), and Hampton additive screens I, II and III were used in the search for possible additives that could improve the crystal quality. The crystallization buffer, which yielded new large and single crystals, contained 0.1 mM Mes (pH 6.5) and 0.3% 1, 5-diaminopentane 2× HCl as additive (additive # 12 of Hampton additive screen 11, cadaverine). Data to 1.9 Å (where 1 Å=0.1 nm) were collected from these crystals at beamline I911-2, Max Lab Lund, Sweden. Only 1:1 (2 μl sitting drops over 500 μl of well solution) drops gave crystals. The crystals belonged to space group C2 with cell dimensions a=108.0, b=87.1, c=130.1, β=91.0 and one dimer in the asymmetric unit.
Prior to data collection, crystals were transferred to 25% (v/v) glycerol for a few seconds before flash freezing in a boiled-off liquid nitrogen stream at 100 K. Data were processed using the XDS package . The data quality was checked with the program TRUNCATE . A summary of the data statistics is shown in Table 2.
Structure determination and refinement
Molecular replacement using the program PHASER  and the co-ordinates of subunit A of the native human ODC structure with solvent molecules removed (PDB ID: 1D7K; ) were used in the search for two molecules per asymmetric unit. Coot  and refmac5  were used for model building and refinement respectively. Non-crystallographic symmetry with tight main chain and loose side chain restraints was used in the refinement of the ODC–APA complex. The library files of the ligands were generated by the PRODRG2 server . Model quality was checked with PROCHECK  and the validation menu available in Coot .
Kinetic analysis of ODC
Table 1 shows that the Km value for the substrate L-ornithine was about 5-fold higher for L. donovani ODC than for the human enzyme. However, the kcat values did not differ markedly between the two enzymes. In fact a higher value was obtained for L. donovani ODC compared with that of the human enzyme, which most probably only reflects minor differences in the quality of the preparations. Nevertheless, the catalytic efficiency value (kcat/Km) was still higher for the human enzyme compared with L. donovani ODC (Table 1). The apparent dissociation constants for PLP were 0.54 μM and 0.05 μM for the L. donovani and human ODCs respectively (Table 1). Both enzymes were effectively inhibited in vitro by DFMO in a time- and concentration-dependent manner (results not shown). The inhibition followed first-order kinetics, and both enzymes exhibited a 5 min half life (t1/2) of activity at infinite DFMO concentration (Table 1). However, the apparent dissociation constant (Kd) for DFMO was 5-fold higher for L. donovani ODC compared with that of the human ODC (Table 1). The ODC inhibitor APA proved to be a highly potent inhibitor of L. donovani ODC with a Ki of 1.0 nM, which actually was somewhat lower than that found for the human enzyme (Table 1).
The structure of the human ODC–APA–PLP complex
During the refinement of the ODC–APA–PLP complex residues one–six could be built into the electron density of both subunits of the dimer. In addition, nine residues after the N-terminal methionine were built for subunit A and eight residues for subunit B. These extra residues are part of the His6-tag present in the cloning vector, and together they form an additional 15-residues-long helix, which is absent in the 1D7K structure used in the molecular replacement search. Residues from this helix are involved in crystal contacts, which probably explain the different space group, as compared with the 1D7K structure. The protease-sensitive loop (residues 158–168; [23,35]), which in the 1D7 K structure had high temperature factors (60–80 Å2), has well defined density in the present model with temperature factors refined to around 20 Å2. The loop also appears to adopt a different conformation, compared with that in the apo-structure, and moves towards the core of the α/β-barrel, closer to the active-site pocket (Figures 1A and 1B). In this position, the loop seems to contribute to the stabilization of substrate binding. The loop next to the protease-sensitive loop (residues 197–205) also moves towards the active site. However, residues 298–310 and 423–461 still could not be modelled. The final refined dimer has a total of 939 amino acid residues, which include 861 ODC residues and 17 residues from the His6-tag. In addition, there were two PLP molecules, two APA molecules, two cadaverine molecules and three acetate groups. The crystallographic R and Rfree had values of 18.4% and 21.3% respectively (Table 2).
PLP- and APA-binding
PLP- and APA-binding is shown in Figures 2(A) and 2(B). A superposition of this complex with the 1D7K structure gives root-mean-square deviation of 0.78 Å involving 3128 atoms. The present complex shows that PLP binds in a position similar to that reported earlier for mouse and T. brucei ODC [20,21], with some differences in the binding mode (Figure 2A). Thus the Schiff base link to the co-factor from Lys69 is absent. Unlike the mouse ODC structure, the Nζ group of the invariant Lys69 is turned away from O4, making a hydrogen bond with the side chain of Asp88 and two solvent molecules. The O4 group of PLP is rotated and is hydrogen bonded with the 1-amino-oxy group of the inhibitor, APA. Nϵ2 of His197 makes a hydrogen bond with OP4 of PLP, and OP1 makes a water-mediated hydrogen bond with Nϵ of Arg277. In contrast, in the complex of mouse ODC with PLP there is a direct interaction between OP1 and Arg277. The pyridine nitrogen N1 interacts with Asp274, in a similar manner to the mouse structure, although the distance to the side chain of Asp88 is slightly shorter (3.1 Å and 3.7 Å for the human and mouse structures respectively). The side chain of Cys360 is rotated by approx. 180° towards the bound inhibitor, occupying a position of 3.7 Å from the oxygen of the hydroxylamine moiety of APA.
The efficiency and specificity of hydroxylamine-containing inhibitors of PLP- or pyruvate-dependent enzymes has been attributed to the properly positioned functional groups in the alkyl fragment, securing structural similarity with the substrate or product of the enzymatic reaction [19,21,36]. In the present complex, APA essentially occupies the position of the substrate in the active site (Figures 2B and 2C), with only some slight differences in the positions of the atoms. Its positively charged 3-amino group, suggested previously to anchor the molecule to the enzyme , interacts with Oδ1 of Asp332, the backbone carbonyl of Tyr331 (from subunit A) and a solvent molecule. This solvent molecule in turn makes hydrogen bonds to Asp361 and Tyr323 (from subunit B). The complex also shows that all the other residues reported previously to be involved in substrate binding through hydrogen bonding or hydrophobic interactions, are also involved in APA binding [21,22,37,38]. As mentioned above, the 1-amino-oxy moiety of APA is at a hydrogen bonding distance from O4 of PLP. In the A subunit, this distance is around 3.0 Å, somewhat longer than the corresponding distance in the B subunit, which is 2.7 Å. The B-subunit-binding site also has a contribution from the side chain of Cys164, which makes a water-mediated hydrogen bond to the 3-amino group of APA. The methylene groups of APA are involved in hydrophobic interactions with Tyr389 (subunit A) and Tyr323 (subunit B). Some residues from the protease-sensitive loop lie within 5 Å from APA. There are also three crystallographic water molecules in the binding site of each subunit (Figure 2C). This is in contrast with the complex of T. brucei ODC with DFMO, which had six water molecules . In the present complex, these water molecules are displaced by the protease-sensitive loop, which was, as mentioned above, disordered in the T. brucei DFMO complex. The new position of the loop also results in a closer fit of APA to the binding pocket, which would add to its high affinity. This could be contrary to the hydroxylamine-derived inhibitors of mitochondrial aspartate aminotransferase whose binding has been proposed to be driven by the high reactivity of the amino-oxy moiety, rather than by high degree of complementarity to the binding site .
Interestingly, in the absence of PLP, APA binds at approximately 7 Å from its position in the complex with PLP, and it partially occupies the site of the PO4 group of PLP (Figure 3). However, this binding has probably no physiological relevance.
Cadaverine, which was used as an additive in the crystallization, was bound with an extended conformation inside a surface-exposed pocket between the N- and C-terminal domains of the molecule in close vicinity to the active site (Figure 4). In subunit A of the protein, the 1-amino group forms water-mediated hydrogen bonds with the backbone carbonyl groups of Gly237 and Cys202, whereas the 5-amino group makes a bond to the side chain of Ser241. Other potential interactions within a radius of 3.5–4 Å involve Gly201, Val244, Pro239, Leu246, Arg277 and Asn385. It is also seen from Figure 4 that some residues from the protease-sensitive loop are within approx. 5 Å from cadaverine. The closest distance to APA and PLP is 10 Å and 7.3 Å respectively, which is rather high for any direct interaction to take place. On the other hand the closest distance to Arg277, which is crucial for ODC activity, is only 3.5 Å. Although cadaverine binding in this structure is most probably unspecific, the presence of an invariant residue in combination with some variable residues within the binding pocket may be used in the design of new species-specific inhibitors of ODC. The unspecific binding of cadaverine in this structure is illustrated by the fact that its position in subunit B is slightly different. It is closer to the surface of the molecule and has a much smaller number of interactions with the protein.
L. donovani ODC has been cloned previously , and its kinetic parameters were subsequently analysed together with those of the mouse and T. brucei enzymes [40,41]. The Km and kcat reported here for L. donovani ODC are similar to the values reported earlier by Osterman et al. . Also the Km value for the human ODC is similar to that reported for the mouse enzyme, although the kcat is somewhat lower . Our results clearly confirm that L. donovani ODC has kinetic properties that are distinct from those of the human ODC. Thus the binding affinity for L-ornithine and PLP is 5- and 11-fold lower respectively for L. donovani compared with the human enzyme. In addition, the apparent dissociation constant (Kd) for DFMO was 5-fold higher for L. donovani. Since DFMO binding mimics substrate binding to ODC, the difference in Kd is in good agreement with the observed difference in affinity for L-ornithine. These results, together with the observation that L. donovani ODC was not able to form cross-species heterodimers with the T. brucei and mouse enzymes , indicate that L. donovani ODC has a potential of being targeted for drug development against leishmaniasis.
The present study shows that L. donovani and human ODC are strongly inhibited by APA, with Ki values around 1.0 nM. Comparable results were reported for the mouse and P. falciparum ODC, which had Ki values of 3.2 nM and 3 nM respectively [13,15]. Recent results also show that treatment with APA and its derivatives reduces the intracellular polyamine concentrations, with a concomitant antiproliferative effect in cultured cancer cells . It also shows clear anti-parasitic effect in P. falciparum and L. donovani cells [13,15,17].
The first step in the reaction mechanism of ODC is expected to be the formation of a Schiff base between L-ornithine and PLP, and the loss of the bond between Lys69 and PLP . The aminooxy group of APA is unprotonated at physiological pH and is believed to form an oxime with PLP , which would mimic the Schiff base formed between the substrate and PLP [22,36,37,43,44]. The structure of the complex between APA and human ODC clearly shows that APA competes with L-ornithine for the substrate-binding site, although the Lys69-PLP Schiff base link is broken and no oxime was formed between APA and PLP. APA is also bound in a bent conformation, with its amino-oxy group pointing away from PLP. One possible explanation for the absence of the oxime formation could be the presence of cadaverine in close vicinity to the active site. Cadaverine may affect the position of PLP, which appears to be slightly different in the present complex, as compared with the mouse ODC–PLP complex. It should be noted that the cadaverine-binding site found between the two domains in the present study is different from the site occupied by geneticin in complex with T. brucei ODC. Geneticin (Ki 5–8 mM) used at a concentration of 200 mM in the crystallization process was able to sufficiently inhibit the slow decarboxylation of D-ornithine by ODC, allowing the observation of D-ornithine bound in the active site .
Further understanding of the sequences of human and L. donovani enzymes as well as the role of the N-terminal extension in L. donovani ODC may help in future development of new drugs against leishmaniasis.
This work was supported by a grant from the Research School in Pharmaceutical Sciences at Lund University (FLÄK) to S. A.-K., a SIDA (Swedish International Development Agency)-SAREC (Department for Research Co-operation) grant, and grants from the Swedish Research Council, the J. C. Kempe Memorial Foundation and the Royal Physiographic Society in Lund.
Abbreviations: APA, 1-amino-oxy-3-aminopropane; DFMO, α-difluoromethylornithine; DTT, dithiothreitol; IPTG, isopropyl β-D-thiogalactopyranoside; ODC, ornithine decarboxylase; PLP, pyridoxal 5′-phosphate
- © 2007 Biochemical Society