Báo cáo khoa học: 9-Deazaguanine derivatives connected by a linker to difluoromethylene phosphonic acid are slow-binding picomolar inhibitors of trimeric purine nucleoside phosphorylase potx

The inhibition constant at equilibrium 1 mm phosphate concen-tration with calf spleen PNP was shown to be Kieq = 85 ± 13 pm pH 7.0, 25C, whereas the apparent inhibition constant determin

difluoromethylene phosphonic acid are slow-binding

picomolar inhibitors of trimeric purine nucleoside

phosphorylase

Katarzyna Breer1, Ljubica Glavasˇ-Obrovac2, Mirjana Suver2, Sadao Hikishima3, Mariko Hashimoto3, Tsutomu Yokomatsu3, Beata Wielgus-Kutrowska1, Lucyna Magnowska1and Agnieszka Bzowska1

1 Department of Biophysics, Institute of Experimental Physics, Warsaw University, Poland

2 University Hospital Osijek and School of Medicine, University of J J Strossmayer in Osijek, Croatia

3 School of Pharmacy, Tokyo University of Pharmacy and Life Science, Japan

Keywords

9-deazaguanine; multisubstrate analogue

inhibitors; purine nucleoside phosphorylase;

slow-binding inhibitors; tight-binding

inhibitors

Correspondence

A Bzowska, Department of Biophysics,

Institute of Experimental Physics, Warsaw

University, _Zwirki i Wigury 93, 02-089

Warsaw, Poland

Fax: +48 22 554 0771

Tel: +48 22 554 0789

E-mail: abzowska@biogeo.uw.edu.pl

(Received 4 October 2009, revised 14

January 2010, accepted 29 January 2010)

doi:10.1111/j.1742-4658.2010.07598.x

Genetic deficiency of purine nucleoside phosphorylase (PNP; EC 2.4.2.1) activity leads to a severe selective disorder of T-cell function Therefore, potent inhibitors of mammalian PNP are expected to act as selective immunosuppressive agents against, for example, T-cell cancers and some autoimmune diseases 9-(5¢,5¢-difluoro-5¢-phosphonopentyl)-9-deazaguanine (DFPP-DG) was found to be a slow- and tight-binding inhibitor of mamma-lian PNP The inhibition constant at equilibrium (1 mm phosphate concen-tration) with calf spleen PNP was shown to be Kieq = 85 ± 13 pm (pH 7.0,

25C), whereas the apparent inhibition constant determined by classical methods was two orders of magnitude higher (Kiapp = 4.4 ± 0.6 nm) The rate constant for formation of the enzyme⁄ inhibitor reversible complex is (8.4 ± 0.5)· 105m)1Æs)1, which is a value that is too low to be diffusion-controlled The picomolar binding of DFPP-DG was confirmed by fluorimet-ric titration, which led to a dissociation constant of 254 pm (68% confidence interval is 147–389 pm) Stopped-flow experiments, together with the above data, are most consistent with a two-step binding mechanism:

E + I M (EI) M (EI)* The rate constants for reversible enzyme⁄ inhibitor complex formation (EI), and for the conformational change (EI) M (EI)*, are

kon1= (17.46 ± 0.05)· 105m)1Æs)1, koff1= (0.021 ± 0.003) s)1, kon2= (1.22 ± 0.08) s)1 and koff2= (0.024 ± 0.005) s)1, respectively This leads

to inhibition constants for the first (EI) and second (EI)* complexes of

Ki= 12.1 nM (68% confidence interval is 8.7–15.5 nm) and Ki = 237 pm (68% confidence interval is 123–401 pm), respectively At a concentration of

10)4m, DFPP-DG exhibits weak, but statistically significant, inhibition of the growth of cell lines sensible to inhibition of PNP activity, such as human adult T-cell leukaemia and lymphoma (Jurkat, HuT78 and CCRF-CEM) Similar inhibitory activities of the tested compound were noted on the growth

of lymphocytes collected from patients with Hashimoto’s thyroiditis and Graves’ disease The observed weak cytotoxicity may be a result of poor membrane permeability

Abbreviations

6C-DFPP-DG, 9-(5¢,5¢-difluoro-5¢-phosphonoheptyl)-9-deazaguanine; DFPP-DG, 9-(5¢,5¢-difluoro-5¢-phosphonopentyl)-9-deazaguanine;

DFPP-G, 9-(5¢,5¢-difluoro-5¢-phosphonopentyl)-guanine; homo-DFPP-DG, 9-(5¢,5¢-difluoro-5¢-phosphonohexyl)-9-deazaguanine;

MTT, 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl-tetrazolium bromide; PNP, purine nucleoside phosphorylase.

Potent membrane-permeable inhibitors of mammalian

purine nucleoside phosphorylase (PNP; EC 2.4.2.1) are

expected to act as selective immunosuppressive agents

against T-cell cancers, host-versus-graft reaction in

organ transplantation, and against some autoimmune

diseases [1] This is because a genetic lack of PNP

activity leads to a severe selective disorder of T-cell

function with normal or even elevated B-cell function

(humoral immunity), as shown by Giblett et al [2]

PNP catalyzes the reversible phosphorolytic cleavage

of the glycosidic bond of purine nucleosides:

b-purine nucleoside + orthophosphate = purine base +

a-d-pentose-1-phosphate The best inhibitors reported

to date are either transition state analogues,

immucil-lins, which bear features of the proposed transition

state (i.e positive charge on the pentose moiety and

N7 of the base protonated) [3], or multisubstrate

ana-logue inhibitors capable of competing simultaneously

for both the nucleoside and phosphate-binding sites

[4] However, in contrast to immucillins, which show a

pKafor pentose protonation at neutral pH (pK = 6.9)

[5], multisubstrate analogue inhibitors are anions, or

even a mixture of mono- and di-anions at neutral pH,

and, as charged molecules, do not readily penetrate cell

membranes They also have short plasma lifetimes

because of a susceptibility to phosphatases Hence,

they are not promising candidates as in vivo inhibitors

This has stimulated the synthesis of some mimics with

the terminal phosphate being replaced by a

phospho-nate [6] or a difluorometylene phosphophospho-nate [7], which

confer metabolic stability Moreover, some

phospho-nates appear to be capable of slowly traversing the cell

membrane, conceivably via an endocytosis-like process

[8,9]

To logically extend the above findings, we have

synthesized a series of multisubstrate analogue

inhi-bitors of PNP, namely 9-deazaguanine derivatives

connected by a linker to difluoromethylene

phos-phonic acid [10,11] All of these 9-deazaguanine

derivatives are potent inhibitors of calf spleen and

human erythrocyte PNP, with apparent inhibition

constants as low as approximately 5 nm; for example,

for

9-(5¢,5¢-difluoro-5¢-phosphonopentyl)-9-deazagua-nine (DFPP-DG) [10] Up to now, however, only

apparent inhibition constants were reported It should

be noted that, for tight-binding ligands, the inhibitor

concentration usually used in the course of classical

experiments, It, is comparable with the total enzyme

concentration, Et, which is in the nanomolar range,

and under such conditions steady-state assumptions

may not hold

In the present study, we employed such an approach and report the true inhibition constants, and also the dissociation constants for binding, of DFPP-DG and some analogues with trimeric PNPs To examine these analogues as possible candidates for in vivo PNP inhib-itors, we also determined some of their biological properties In particular, cytotoxic activities of

DFPP-DG against human lymphocytes from healthy subjects and patients with autoimmune thyroid diseases (i.e Hashimoto’s thyroiditis and Graves’ disease), as well

as against a panel of human leukaemia and lymphoma cell lines, were determined

Results and Discussion

Apparent inhibition constants Structures of new compounds embraced in the present study are shown in Fig 1 Apparent inhibition con-stants versus two mammalian purine nucleoside phos-phorylases, from calf spleen and human erythrocytes, with 7-methylguanosine (m7Guo) as a variable sub-strate, were determined using methods described previ-ously for other inhibitors of trimeric PNPs [12,13] With fixed concentrations of one substrate (i.e inor-ganic phosphate), apparent inhibition constants (Kiapp) were determined from initial velocity data with variable concentrations of both the inhibitor and the second substrate (m7Guo) Dixon plots displayed a competitive mode of inhibition, as shown in Fig 2 for DFPP-DG and human erythrocyte PNP Data sets were analysed, and apparent inhibition constants calculated, with the use of the weighted least-squares nonlinear regression software leonora [14], as summarized in Table 1 For comparison, inhibitory activities of

9-(5¢,5¢-difluoro-Fig 1 Structure of DFPP-DG and analogues: n = 1, DFPP-DG;

n = 2, homo-DFPP-DG; n = 3, 6C-DFPP-DG (left); and the structure

of immucillin H (right).

5¢-phosphonopentyl)-guanine (DFPP-G) [15] and a

transition state analogue inhibitor, immucillin H [3],

are also included

All compounds were found to be very potent

inhibi-tors of m7Guo phosphorolysis, with apparent

inhibi-tion constants, Kiapp, in the nanomolar range

Inhibition is competitive versus nucleoside (m7Guo),

and the apparent inhibition constants, Kiapp, decrease

with decreasing phosphate (fixed substrate)

concentra-tion (Table 1) This indicates that the inhibitors bind

to both nucleoside- and phosphate-binding sites, and hence act as multisubstrate analogue inhibitors

As predicted by previous structural studies [16], DFPP-DG allows more favourable interactions with the base-binding site of calf spleen and human erythro-cyte PNPs compared to DFPP-G, and therefore yields

a Kiapp lower than observed for DFPP-G (Table 1) However, the effect is not large as a result of enthalpy-entropy compensation The gain in enthalpic contribu-tion to the Gibbs binding energy, when compared with DFPP-G binding, is balanced by an entropic effect [17]

Except for 9-(5¢,5¢-difluoro-5¢-phosphonohexyl)-9-de-azaguanine (homo-DFPP-DG) versus human erythro-cyte PNP, which exhibits even better binding properties than those observed for DGPP-DG (Kiapp = 5.3 nm compared to 8.1 nm; Table 1), the other derivatives with shorter and longer linkers exhibited weaker inhibitory effects

Time-dependence of inhibition The inhibition constants shown in Table 1 should be treated as apparent values because the reaction rates observed in the presence of DFPP-DG and its ana-logues exhibit some initial inhibition (see the initial velocity experiments), which increases as a function of time (Fig 3, left) This may be a result of low enzyme and inhibitor concentrations (both in the nanomolar range), leading to slow-binding inhibition because the equilibrium may not be attained in the time-scale

of the initial velocity studies [18] Therefore, the

Table 1 Inhibitory properties of DFPP-G, DFPP-DG and their analogues versus calf-spleen and human erythrocyte PNPs, and rates of associ-ation (k) of some of the analogues with calf spleen PNP Kiappis an apparent inhibition constant observed by the classical initial velocity method, whereas Kieq is an equilibrium inhibition constant determined after the slow-binding inhibitor is allowed to equilibrate with the enzyme (see Materials and methods) For classical inhibitory studies, all reactions were carried out in 50 m M Hepes buffer (pH 7.0) at 25 C, with m 7 Guo as variable substrate, in the presence of a fixed concentration of phosphate, as indicated For equilibrium studies, and for deter-mination of the association rate-constant, the enzyme was incubated with 1 m M phosphate and various concentrations of inhibitor and, after

a given time interval (0.5–120 min), activity was determined with 60 l M m7Guo (in 50 m M Hepes buffer, pH 7.0, at 25 C).

Compound

Phosphate concentration [m M ]

Kiapp[n M ] human PNP

Kiapp[n M ] calf PNP

Kieq[p M ] calf PNP

k [ M )1Æs)1]

calf PNP

a Data from Iwanow et al [15] b Poor solubility c From Miles et al [3], with the constant for the first reversible step d From Miles et al [3], with the constant in equilibrium.

Fig 2 Inhibition of human erythrocyte PNP by DFPP-DG m 7 Guo

was a variable substrate s, 8.4 l M; •, 12.8 l M ; h, 25.2 l M ;

, 210 l M

inhibition constant for binding of DFPP-DG to calf

spleen PNP was also determined at equilibrium, as

described in the Materials and methods

The approach to equilibrium was followed by

mea-suring the velocity observed after various times of

incubation (0.5–120 min), and for various inhibitor

concentrations (in the range 0.5–20 nm) steady-state

velocities, vswere determined as shown in Fig 3 From

the set of vs for various inhibitor concentrations, the

inhibition constant at equilibrium, Kieq, was determined

by fitting Eqn (2) to the vs[I]⁄ kcdependence, and was

found to be Kieq= 85 ± 13 pm, and hence two orders

of magnitude lower than the apparent inhibition

con-stant determined in the standard initial velocity

experi-ment, Kiapp = 4.4 nm (see above)

Slow-onset binding, slow binding or binding

limited by diffusion

Time-dependence of inhibition was previously reported

for the transition-state inhibitors, immucillins [3], and

was interpreted as a slow-onset (i.e two-step binding)

mechanism For such a mechanism, binding involves

the rapid formation of the enzyme⁄ inhibitor collision

complex, followed by a slow conformational change,

leading to a more tightly bound enzyme⁄ inhibitor

com-plex: E + I M (EI) M (EI)* However, it should be

noted that the presence of a slow-onset phase,

espe-cially when nanomolar enzyme and ligand

concentra-tions were used, does not unequivocally point to

binding as a two-step mechanism It may simply be the

observation of a process of achieving equilibrium

between ingredients (Figs S1 and S2, data simulated assuming one-step and two-steps mechanisms) The question then arises as to whether the one- or two-step mechanism also applies to binding of DFPP-DG and analogues to trimeric PNPs The data presented in Fig 3 (left) suggest only that equilibrium is reached more rapidly with higher DFPP-DG concentrations, in agreement with both mechanisms The rate of the exponential decay (Eqn 1; see Materials and methods) increases linearly with increasing inhibitor concentra-tion (Fig 3, insert) This is usually considered as an indicator for a mechanism involving two molecules (i.e E + I M E I), and not the conformational change

of the (EI) complex, (EI) M (EI)* However, the simu-lated data according to a two-step mechanism show that linearity may be observed in the case of more complicated binding patterns [19] The rate constant derived form the data shown in the insert to Fig 3 resulted in a value of (8.4 ± 0.5) · 105m)1Æs)1 for complex formation between PNP and DFPP-DG, which is too small to be classified as a diffusion-con-trolled encounter rate, and which is approximately 108

or higher [20] However, to confirm that complex formation is not limited by diffusion, a control experi-ment was performed The reaction mixture containing the enzyme (2.3 nm) and the inhibitor (3.0 nm) was continuously mixed The rate measured in this case did not differ from the rate measured without mixing (Fig S3)

To confirm that DFPP-DG is a slow-binding inhibi-tor of trimeric PNP, we conducted an experiment with calf spleen PNP and DFPP-DG, using continuous

Fig 3 Left: Time-dependence of inhibition of calf spleen PNP by DFPP-DG PNP (2.3 n M subunits), DFPP-DG (s) 0 n M , (*) 1.0 n M , ()) 2.0 n M or (•) 3.0 n M and only one PNP substrate (1 m M phosphate) Data for several other inhibitor concentrations were collected, but are not shown The insert shows the dependence of the observed rate constants on DFPP-DG concentration, with an exponential decay fitted, leading to an association rate constant of (8.4 ± 0.5) · 10 5

M )1Æs)1 Right: Determination of the inhibition constant at equilibrium, Keq

i , for interaction of DFPP-DG with calf PNP Constants were obtained by fitting equation [2] to the equilibrium velocities, vs, obtained from experi-ments depicted in the upper panel The Kieqvalue obtained from these data is 85 ± 13 p M

monitoring with saturating substrate concentration,

according to Morrison and Walsh [18] This was

previ-ously performed with inosine as a substrate and

immu-cillin as an inhibitor (at pH 7.7) to characterize the

slow-onset binding observed with such transition state

inhibitors [3] Therefore, as a control, we performed

the same experiment with immucillin H, both at the

same pH 7.7 (not 7.0) The data presented in Fig 4

clearly show that the slow-onset phase (i.e the

charac-teristic feature of interaction of immucillins with

trimeric PNPs) is also observed with DFPP-DG, but is

not as well defined In the initial phase of the reaction

(10 min), 12.3 nm of immucillin H does not cause

any inhibition of inosine phosphorolysis, by contrast

to DFPP-DG Ki for the rapidly reversible complex

observed for immucillin H is 41 ± 8 nm (Table 1) [3]

However, over time, immucillin inhibits more and

more strongly, and finally the equilibrium for the

slow-onset step is attained (Fig 4, left) with the equilibrium

dissociation constant for immucillin being

Kieq= 23 ± 5 pm [3] This is not so with DFPP-DG

as an inhibitor In this case, an almost linear

depen-dence of uric acid formation [the final product of the

couple assay for inosine as a PNP substrate) versus

time is observed over the whole course of the

experi-ment (Fig 4, right; but see also below)

In the progress curve method, the inhibitor competes

with a high excess of substrate for the active sites of

the enzyme; therefore, the slow-onset phase of the

reaction may not always be observed [18] This is

shown in the left panel of Fig 4, where, in the case of

immucillin H, a change of pH from 7.7 to 7.0 is such

that equilibrium for the slow-onset phase is not

reached in the course of the experiment Hence, to

dis-tinguish between one-step slow-binding and two-step

slow-onset binding, it is important to fit these two

models to a set of progress curves using software based

on numerically solving systems of differential

equa-tions (e.g dynafit; BioKin, Ltd, Watertown, MA, USA) However, some problems may arise We used dynafit, version 4.0 to simulate sets of progress curves described by both mechanisms Only in the case

of one-step binding were we able to reconstruct param-eters used for simulations (Docs S1 and S2; Figs S1 and S2)

Confirmation of picomolar binding constant by titration experiments

To confirm strong binding of DFPP-DG by calf spleen PNP, the dissociation constant for this complex was determined directly Classical fluorimetric approaches were employed but only provided confirmation that one ligand molecule is bound per enzyme monomer and that binding is strong because the binding curve displayed the typical stoichiometric character, which means that the binding process was rapidly stopped when the ligand concentration added was equal to the PNP subunit concentration (Fig 5) A classical data evaluation (i.e fitting of the well-known Eqn (5) derived under assumptions described in the Materials and methods, separately for each titration, resulted in plots of residuals showing unequivocally that the used model does not properly describe the experimental data (Fig 5, lower panel)

Therefore, an approach using dynafit software was employed Three various models were tested (see Mate-rials and methods): assuming non-identical changes of fluorescence upon binding of the first, second and third ligand molecule but identical affinity to ligand by subunits, then non-identical affinities (allosteric behav-iour) but identical changes of fluorescence and, finally, non-identical changes of fluorescence and affinities The fit based on the assumption that monomers bind the ligand with identical affinities, but with different fluorescent responses, was the most accurate (Fig 6)

Fig 4 Slow-onset binding of immucillin H

by calf spleen PNP (left, pH 7.7, if not

other-wise indicated) and the similar, but less well

defined, slow-onset phase for binding of

DFPP-DG (right, pH 7.7).

We fitted simultaneously more than one data set, obtained with various protein concentrations, but trea-ted molar fluorescence parameters for different forms

of the PNP⁄ DFPP-DG complexes as the independent adjustable parameters for each curve as a control We obtained comparable values of fluorescent increments upon binding for both curves, which confirms that the used model properly describes the experimental data From this fit, the first dissociation constant, Kd1 (see Materials and methods) was found to be 84.6 pm (68% confidence intervals is 49.3–129.4 pm), which corre-sponds to a classical dissociation constant three-fold higher, Kd= 3Kd1(i.e Kd= 254 pm) (68% confidence intervals is 147–389 pm) This value is somewhat higher than the one obtained from inhibition at equilibrium,

Kieq= 85 ± 13 pm but, according to the 90% confi-dence intervals, the data are in agreement (Tables S1 and S2) It should be recalled that additions of the ligand in the fluorimetric titrations were made every

40 s It could be argued that, as a result of slow bind-ing, equilibrium may not be fully achieved However, the concentrations used for the titrations were a few orders of magnitude higher than in the kinetic approach Furthermore, we did not observe any change

in signal when data were collected for an additional

40 s, which means that the formation of the first com-plex is completed during only 40 s These facts taken together suggest a two-step binding mechanism for DFPP-DG rather than one-step binding Both methods confirm that DFPP-DG binds as strongly as the transi-tion state analogue inhibitors, immucillins

380

400

420

440

460

480

500

520

540

560

–3

–2

–1

0

1

2

3

4

Fig 5 Fluorimetric titration of calf spleen PNP (0.4 l M binding

monomers; see Materials and methods) with DFPP-DG Data show

that binding is stoichiometric and, hence, with a very low

dissocia-tion constant (and much lower than the enzyme concentradissocia-tion) (i.e.

the binding process stops rapidly when the added ligand

concentra-tion is equal to the concentraconcentra-tion of the active binding sites) The

classical approach was employed to analyse the data (see Materials

and methods); however, the residual plot (lower panel) shown

indicates that this method is not correct.

0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14

0.0

0.1

0.2

0.0 0.1 0.2 0.3 0.4 0.5 0.6 380

400 420 440 460 480 500 520 540 560

0.0 0.1 0.2 0.3 0.4 0.5 0.6 –1.5

–1.0 –0.5

–1.3

–1.2

–0.1

0.0 0.5 1.0 0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14

100

105

110

115

120

125

130

135

140

145

150

Fig 6 Fluorimetric titrations of calf spleen PNP with DFPP-DG (upper panels) with resi-dual plots (lower panels) for the best model fitted (see Results and Discussion) Protein concentrations (in terms of binding mono-mers; see Materials and methods) were

0.-1 l M (left) and 0.4 l M (right) Data were analysed simultaneously, using DYNAFIT soft-ware as described in the Materials and met-hods The dissociation constant obtained from this fit is 254 p M (68% confidence int-erval is 147–389 p M ) The molar fluores-cence for protein complexes with one, two and three ligand particles are:

fPL1= 414.4 ± 20.7, fPL2= 800.0 ± 30.8,

fPL3= 1070.3 ± 36.6 AU (left) and

f PL1 = 414.7 ± 4.8, f PL2 = 725.9 ± 6.6,

fPL3= 1002.6 ± 7.5 AU (right).

Stopped-flow measurements

To finally resolve the one- or two-step binding

prob-lem, we conducted a series of stopped-flow experiments

(Fig 7) Kinetic traces were analysed using dynafit

software Various models were considered Data may

be adequately well described by the one-step model,

although the dissociation constant calculated from the

rate constants obtained in this case, kon1= (16.6 ±

0.1)· 105m)1Æs)1, koff1= (0.0013 ± 0.0001) s)1, is an

order of magnitude higher than the values derived

from other methods (koff1⁄ kon1= 783 pm compared to

85 and 254 pm; see above) The two-step model with

similar fluorescence properties of both enzyme-ligand

complexes, (EI) and (EI)* (2.12 AU and 2.15 AU

respectively), gave a slightly lower sum of squares, but

also much better agreement with the results obtained

by other methods Rate constants derived from this model are: kon1= (17.46 ± 0.05)· 105m)1Æs)1,

koff1= (0.021 ± 0.003) s)1 for the (EI) complex and

kon2= (1.22 ± 0.08) s)1, koff2= (0.024 ± 0.005) s)1 for the (EI)* complex, leading to inhibition constants for the (EI) and (EI)* complexes Ki= 12.1 nm (68% confidence interval is 8.7–15.5 nm) and K

i = 237 pm (68% confidence interval is 123–401 pm), respectively The second value is in excellent agreement with the steady-state titration experiments (see above) We conclude that DFPP-DG binding with calf PNP follows a two-step binding model

DFPP-DG analogues DFPP-DG analogues also bind slowly with trimeric PNPs Moreover, slow binding is not limited to com-pounds with the 9-deazaguanine aglycone because the same slow-binding effect was observed also for

DFPP-G Hence, it appears that the 9-deaza feature is not responsible for the slow-binding phenomenon For DFPP-G, the rate constant for EI complex formation

is (4.5 ± 0.7)· 106m)1Æs)1, and the difference between the apparent and equilibrium inhibition constants is only approximately ten-fold (6.9 nm compared to 0.79 nm; Table 1), and much less pronounced than with DFPP-DG

Cytotoxic activities Tight binding of DFPP-DG to PNP led us to check its possible inhibitory potential on the growth of human normal cells and cell lines derived from haematological malignancies Cells selected for testing were human normal lymphocytes, lymphocytes of patients with autoimmune thyroid diseases, and a panel of lym-phoma and leukaemia cells from B- and T-cells T-cell malignancies have specific biochemical, immunological and clinical features, which separate them from non-T-cell malignancies [21]

DFPP-DG moderately affects growth of several leukaemia and lymphoma cell lines, especially T-cell leukaemias (Jurkat and MOLT), acute lymphoblastic leukaemia (CCRF-CEM) and T-cell lymphoma (HuT78) Some differences were observed between the effects on the growth of tumor cells sensible to inhibition of PNP activity, such as human adult T-cell leukemia and lymphoma (Jurkat, MOLT, HuT78, CCRF-CEM) and other leukaemia and lymphoma cells of B-cell, or non-T- and non-B-cell lineages (K562, Raji, HL-60) However, the effects were detect-able only at the highest concentration applied, 10)4m (Table 2)

Fig 7 Set of stopped-flow kinetic traces obtained after mixing of

PNP with DFPP-DG Concentrations of PNP subunits in the

stopped-slow spectrometer, 0.4 l M (black), 0.2 l M (grey) and

0.1 l M (light grey), and the concentration of DFPP-DG (in l M ) are

given for each trace Data were analysed simultaneously using

DYNAFIT software (see Materials and methods) and the curves fitted

are also shown.

Hashimoto’s thyroiditis and Graves’ disease are

T-cell mediated autoimmune thyroid diseases [22–25]

Regarding the known features of autoimmunity to the

thyroid gland, we expected significant inhibitory effects

of DFPP-DG on lymphocytes collected from patients

suffering from human autoimmune thyroid disorders,

relative to normal lymphocytes DFPP-DG, at 10)4m,

exhibited modest, but statistically significant, inhibitory

effects (almost 30%) on lymphocytes from patients

suffering from Hashimoto’s thyroiditis and Graves’

disease

The reason behind the modest cytotoxic properties of

DFPP-DG observed in vivo, despite its excellent

inhibi-tory properties versus trimeric PNP, lies most probably

in the poor penetration capability of this compound

through cell membranes Some phosphonates appear to

be capable of slowly traversing the cell membrane [8,9]

However, DFPP-DG is a difluorometylene

phospho-nate It is known that fluorination of

alkylphospho-nates yields compounds with properties suitably

resembling phosphate esters [7,26], and, in turn, this

leads to optimized interactions of such analogues with

the phosphate-binding site residues in the PNP active

site [16,27] Because the physical properties of

DFPP-DG are rather similar to those of phosphates, it is not

unusual that this compound is not readily taken up by

the cells To demonstrate this, we plan to mark

DFPP-DG with a fluorescent dye so that we can follow its

entry into cells and its intracellular localization If we confirm that the poor uptake is, in fact, responsible for the mild cytotoxic effects observed, we plan to synthe-size a pro-drug of DFPP-DG Alternatively, we also plan to employ one of the recently developed drug-delivery systems [28,29], to improve the cell penetration

of this excellent PNP inhibitor

Conclusions

DFPP-DG and some analogues show inhibition and dissociation constants versus trimeric purine nucleoside phosphorylases in the picomolar range Similarly to immucillins – transition state analogue inhibitors [3], the compounds described in the present study exhibit slow-onset binding pattern as well Stopped-flow exper-iments together with data obtained by other methods are consistent with two-step binding mechanism, and hence similar to that observed in the case of immucil-lins DFPP-DG shows moderate inhibitory effects on the growth of lymphocytes from patients with human autoimmune thyroid disorders and T-cell leukaemia and lymphoma cells, but only at a concentration of

10)4m Because DFPP-DG is a phosphonate and car-ries a negative charge, the inefficient transport of the inhibitor into cells is most probably responsible for the mild cytotoxic effects Although some phosphonates appear to be capable of slowly traversing the cell mem-brane, conceivably via an endocytosis-like process, this

is not likely the case with DFPP-DG For that reason, future studies will be directed toward the synthesis of a pro-drug of DFPP-DG to improve its cell penetration The problem of the poor uptake of the compound by cells may, in principle, also be overcome by use of one

of the recently developed drug-delivery systems [28,29] One of these approaches is based on use of the cross-linked cationic polymer network (Nanogel) for intra-cellular delivery of negatively charged drugs, and shown to be successful with the cytotoxic 5¢-phosphate

of 5-fluoroadnenosine arabinoside, fludarabine [30], and 5¢-triphosphates of cytarabine (araCTP), gemcita-bine (dFdCTP) and floxuridine (FdUTP) [31] We also plan to mark DFPP-DG with a fluorescence dye to follow its entry into cells and its intracellular localization in an effort to explain the observed mild cytotoxic effects

Materials and methods

Reagents

Commercially available PNP from calf spleen (Sigma,

St Louis, MO, USA), as a suspension in 3.2 m ammonium

Table 2 Cytotoxic effects of DFPP-DG towards various cell types.

Exponentially growing cells were treated with different concentration

of DFPP-DG for 72 h periods Cytotoxicity was analysed by the MTT

survival assay All experiments were performed at least three times.

Cell lines: acute lymphoblastic leukemia (CCRF-CEM), T-cell leukemia

(Jurkat and MOLT-4), T-cell lymphoma (HuT78), acute myeloid

leuke-mia (HL-60), Burkitt’s lymphoma (RAJI) and chronic myeloid

leukemia in blasts crisis (K562) Human blood lymphocytes from

healthy donors, from patients with Graves’ disease and from patients

with Hashimoto’s disease –, no effect *Statistically significant

change (P < 0.05).

Cell line

Percentage inhibition DFPP-DG concentration

10)7M 10)6M 10)5M 10)4M

sulphate, with specific activity versus inosine of

[12] Lyophylized human erythrocyte PNP (from Sigma)

was dissolved in 20 mm Hepes buffer (pH 7.0) (0.5 mg in

100 lL of buffer) The specific activity of this enzyme

chemicals were products obtained from Sigma or Fluka

(Buchs, Switzerland) Xanthine oxidase from buttermilk, a

prepared as previously described [10,11] All solutions were

prepared with high-quality MilliQ water (Millipore,

Billeri-ca, MA, USA)

Concentrations of all substrates and inhibitors were

determined spectrophotometrically using the extinction

linker [9-(5¢,5¢-difluoro-5¢-phosphono-heptyl)-9-deazaguanine

immucillin H [3]

Enzyme concentrations were determined from the

[32] In calculations, the theoretical molecular mass of one

monomer of the calf spleen enzyme, based on its amino

acid sequence, was used; molecular mass = 32 093 Da [33]

(SwissProt entry P55859) Molar concentrations are given

in all experiments in terms of enzyme monomers

Instrumentation

Kinetic and spectrophotometric measurements were carried

out on a Uvikon 930 (Kontron, Vienna, Austria)

spectro-photometer fitted with a thermostatically controlled cell

compartment, using 10, 5, 2 or 1 mm path-length quartz

cuvettes (Hellma, Mullheim, Germany) A Beckman model

F300 pH-meter (Beckman Coulter, Fullerton, CA, USA)

equipped with a combined semi-microelectrode and

temper-ature sensor, was used for pH determination

Fluorescence data were recorded on a Perkin-Elmer

solu-tion

Stopped-flow kinetic measurements were run on a

SX.18MV stopped-flow reaction analyzer from Applied

Photophysics Ltd (Leatherhead, UK) The dead time of the

instrument was 1.2 ms

Cornelius, OR, USA) was used for cell culturing and an

ELISA plate reader (Stat fax 2100; Pharmacia Biotech,

Uppsala Sweden) for absorbance measurement in the

cyto-toxic activity measurements

Standard enzymatic procedures

Kinetic studies, if not otherwise indicated, were conducted

phosphate buffer for determination of inhibition constants, and in 50 mm phosphate buffer for determination of the enzyme specific activity

One unit of PNP is defined as the amount of enzyme that causes phosphorolysis of 1 lmol of inosine to hypoxanthine and ribose-1-phosphate per minute under standard

phosphate buffer, pH 7.0) The standard coupled xanthine oxidase procedure [32] was used in which hypoxanthine, liberated in the PNP catalysed reaction, is oxidized to uric acid by xanthine oxidase The observation wavelength was

m)1Æcm)1[12]

PNP is known for its nonhyperbolic kinetics Deviations from the classical Michaelis–Menten kinetics depend on the nucleoside substrate and concentration of the co-substrate, phosphate [12] Therefore, inhibition type and inhibition constants were determined, if not otherwise indicated, using

for this substrate, the classical Michaelis–Menten [34] equa-tion is sufficient for data analysis [12]

spectrophotomet-rically by a direct method [35] The observation wavelength,

between extinction coefficients of nucleoside substrate,

cationic and zwitterionic forms of m7Guo [12,35]

The reaction mixture for the direct method and for the coupled method had a 1 mL volume in a 10 mm path-length

both substrates of the phosphorolytic reaction (phosphate buffer of the same pH as the main buffer, and a nucleoside,

inhibi-tion studies, an inhibitor was included in the reacinhibi-tion mixture The reaction was started by the addition of PNP Initial rate procedures were employed in all kinetic studies

In the case of inhibition studies, for each combination of the initial substrate concentration, co, and the inhibitor concen-tration [I], the rates were determined at least twice The initial

con-trolling the spectrophotometer Linear regression software (Kontron, Vienna, Austria) was used for determination of slopes, with their standard errors, of absorbance versus time

Time-dependence of inhibition: progress curves

Time-dependence of inhibition was measured using two approaches In the first, inosine was the substrate and

continuous monitoring of uric acid formation was used to

measure the progress curve, as described by Miles et al [3]

Briefly the enzyme (1.3 nm subunits) was added to the

7.7) containing an excess of both substrates (0.71 mm

ino-sine, 50 mm phosphate buffer, pH 7.7) and various

inhibi-tor concentrations Formation of uric acid, the final

product of the coupled PNP and xanthine oxidase reaction

[32], was monitored at 300 nm

Time-dependence of inhibition: initial velocity

In the progress curve approach, the inhibitor competes with

the substrate for the active sites of the enzyme With a high

excess of substrate, the slow-onset phase of the reaction

may not always be observed [18] Therefore, the initial

velocity method was also used In this approach, the

enzyme (2.3 nm) and inhibitor (concentration range 0.5Ờ

(pH 7.0) and 1 mm phosphate buffer (pH 7.0) The total

volume was 1.2 mL After a given time interval, t (0.5Ờ

(2000 lm), was mixed with 0.97 mL of the incubated

60 lm, with all other concentrations changed by only 3%,

to allow treatment equal to the initial values The initial

velocities observed after various incubation times for each

inhibitor concentration, vo(t, [I]), were measured

For each inhibitor concentration, the velocity at

vs[I]; steady-state velocity observed in the presence of

inhib-itor at [I] concentration) was determined This was achieved

by fitting the one-phase exponential decay to each set of

velocities observed with various [I], vo(t, [I]):

velocity obtained at time t = 0 (i.e no incubation) with a

absence of inhibitor [i.e vo(0; [0]) = kc] It was also found

that 120 min of incubation has no influence on enzyme

deter-mined as previously described [12], and the value obtained,

determined from Eqn (18), as reported previously for

immucillins [3]:

msơI=kcỬ ơS= K mđ1 ợ ơI=Kieqợ ơS

đ2ỡ

Fluorimetric titrations

described previously [27] but the protein was not diluted during experiments because the ligand stock used for titra-tions was prepared in the buffer and protein solution corre-sponding to their concentrations in a cuvette Experiments were performed in 20 mm Hepes buffer (pH 7.0), in the

concentrations were either 0.2 or 0.8 lm, as determined from UV absorption PNP specific activity was

0.4 lm binding monomers because the activity of the fully

previ-ously [27] The rest of the protein is inactive PNP, which,

as shown previously, does not interfere with binding of ligands by the active enzyme [12,27] Additions of ligand were made every 40 s

The protein-ligand binding model for the trimeric pro-tein, assuming a one-step process for each binding site, is:

Pợ L ,ka1

k d1

PL1 Kd1Ử kd1=ka1

PL1ợ L ,ka2

k d2 PL2 Kd2Ử kd2=ka2 đ3ỡ

PL2ợ L ,ka3

kd3PL3 Kd3Ử kd3=ka3

At any given time, the fluorescence of the solution may

be represented as the sum of the fluorescence of the various molecular species present in the mixture, free trimeric pro-tein, P, free ligand, L, and trimeric protein complexed with one, two or three ligand molecules (PL1, PL2, PL3): FluorescenceỬ ơP fPợ ơL fLợ ơPL1 fPL1ợ ơPL2 fPL2ợ ơPL3 fPL3

đ4ỡ

approaches The classical approach assumed that ligand binds to all three subunits of the trimeric PNP molecule independently and is described by a single dissociation con-stant, Kd; hence, the appropriate equation is [36]:

coeffi-cients of free PNP subunit, free ligand and PNP subunit complexed with the ligand, respectively, [L] is the total con-centration of the ligand, F([L]) is the fluorescence intensity

FđơLỡ Ử F0 đfEợ fL fELỡ [L]

2 ợ

ơEact

Kd

2 

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi đơL  ơEact ợ Kdỡ2ợ 4ơEactKd q

2

0

@

1

A ợ ơLfL đ5ỡ