Intracellular rebinding of transition-state analogues provides extended in vivo inhibition lifetimes on human purine nucleoside phosphorylase

Purine nucleoside phosphorylase (PNP) is part of the human purine salvage pathway. Its deficiency triggers apoptosis of activated T-cells, making it a target for T-cell proliferative disorders. Transition-state analogues of PNP bind with picomolar (pm) dissociation constants. Tight-binding PNP inhibitors show exceptionally long lifetimes on the target enzyme. We solve the mechanism of the target residence time by comparing functional off-rates in vitro and in vivo. We report in vitro PNP-inhibitor dissociation rates (t½) from 3 to 31 min for seven Immucillins with dissociation constants of 115 to 6 pm. Treatment of human erythrocytes with DADMe-Immucillin-H (DADMe-ImmH, 22 pm) causes complete inhibition of PNP. Loss of [14C]DADMe-ImmH from erythrocytes during multiple washes is slow and biphasic, resulting from inhibitor release and rebinding to PNP catalytic sites. The slow phase gave a t½ of 84 h. Loss of [14C]DADMe-ImmH from erythrocytes in the presence of excess unlabeled DADMe-ImmH increased to a t½ of 1.6 h by preventing rebinding. Thus, in human erythrocytes, rebinding of DADMe-ImmH is 50-fold more likely than diffusional loss of the inhibitor from the erythrocyte. Humans treated with a single oral dose of DADMe-ImmH in phase 1 clinical trials exhibit regain of PNP activity with a t½ of 59 days, corresponding to the erythropoiesis rate in humans. Thus, the PNP catalytic site recapture of DADMe-ImmH is highly favored in vivo. We conclude that transition-state analogues with picomolar dissociation constants exhibit long lifetimes on their targets in vivo because the probability of the target enzyme recapturing inhibitor molecules is greater than diffusional loss to the extracellular space.


Purine nucleoside phosphorylase (PNP) is part of the human purine salvage pathway. Its deficiency triggers apoptosis of activated T-cells, making it a target for T-cell proliferative disorders. Transition-state analogues of PNP bind with picomolar (pM) dissociation constants. Tight-binding PNP inhibitors show
exceptionally long lifetimes on the target enzyme. We solve the mechanism of the target residence time by comparing functional off-rates in vitro and in vivo. We report in vitro PNPinhibitor dissociation rates (t1 ⁄2 ) from 3 to 31 min for seven Immucillins with dissociation constants of 115 to 6 pM. Treatment of human erythrocytes with DADMe-Immucillin-H (DADMe-ImmH, 22 pM) causes complete inhibition of PNP. Loss of [ 14 C]DADMe-ImmH from erythrocytes during multiple washes is slow and biphasic, resulting from inhibitor release and rebinding to PNP catalytic sites. The slow phase gave a t1 ⁄ 2 of 84 h. Loss of [ 14 C]DADMe-ImmH from erythrocytes in the presence of excess unlabeled DADMe-ImmH increased to a t1 ⁄ 2 of 1.6 h by preventing rebinding. Thus, in human erythrocytes, rebinding of DADMe-ImmH is 50-fold more likely than diffusional loss of the inhibitor from the erythrocyte. Humans treated with a single oral dose of DADMe-ImmH in phase 1 clinical trials exhibit regain of PNP activity with a t1 ⁄ 2 of 59 days, corresponding to the erythropoiesis rate in humans. Thus, the PNP catalytic site recapture of DADMe-ImmH is highly favored in vivo. We conclude that transition-state analogues with picomolar dissociation constants exhibit long lifetimes on their targets in vivo because the probability of the target enzyme recapturing inhibitor molecules is greater than diffusional loss to the extracellular space.
Analyses of drug efficacy have considered the significance of the dissociation rate constant or off-rate (k off ) and its correla-tion with residence time for drug candidates binding to their biological targets (1)(2)(3). Residence time is an important factor in determining biological efficacy and can be determined from half-life measurements obtained from the inhibitor off-rates (4). Off-rates are most commonly measured in vitro by incubating the enzyme and inhibitor, diluting the complex, and observing the resulting regain of enzymatic activity (5).
Transition-state analogues often exhibit physiological inhibitory effects extending beyond inhibitory expectations extrapolated from in vitro studies. For example, in mice, a single oral dose of the PNP inhibitor DADMe-ImmH ( Fig. 1) gave inhibition in circulating erythrocytes for the life of the cells, whereas k off in solution predicted a much shorter efficacy (4). Extended inhibition on the target suggests a reduced k off rate under physiological conditions, cellular accumulation of inhibitor, covalent inhibition, or frequent rebinding events. Extended inhibitor-target efficacy will occur when the inhibitor has a higher probability of rebinding to the target enzyme than loss through cell membrane diffusion. Rebinding contributions to long-term inhibition have been mathematically treated and are known to contribute to cell receptor-ligand interactions (6 -8).
Purine nucleoside phosphorylase (EC 2.4.2.1; PNP) is a homotrimeric enzyme containing three active sites (9). It is essential in human purine salvage by catalyzing the phosphorolysis of 6-oxypurine nucleosides and 2Ј-deoxynucleosides to purine bases and ␣-D-ribose 1-phosphate in the production of purine bases for salvage and recycling or for urate production (10). The human genetic deficiency of PNP causes the accumulation of 2Ј-deoxyguanosine (dGuo) 4 in the blood and induces T-cell immunodeficiency when activated T-cells convert dGuo to 2Ј-deoxyguanosine triphosphate (dGTP). The unbalanced dGTP pool causes T-cell specific apoptosis, making PNP a chemotherapeutic target for T-cell proliferative disorders (10 -14). The catalytic sites of homotrimeric PNP bind transition-state analogues with negative cooperativity. Inhibitor binding at the first site has the highest affinity and causes complete inhibition of the enzyme. Inhibitor-binding exhibits negative cooperativity, with weaker binding at the second and third sites of the trimer, each by more than an order of magnitude relative to the first site (15).
Immucillin-H (ImmH; Fig. 1) gains access to human leukemia CCRF-CEM cells via the equilibrative nucleoside transporters (ENT1 and ENT2), whereas dGuo uptake depends on ion-dependent concentrative nucleoside transporters (sodiumdependent transporters). Transporters ENT1 and ENT2 are also located in the plasma membranes of erythrocytes (16,17). Here we demonstrate that the ENT1 and ENT2 transporters are involved in the transport of DADMe-ImmH by dipyridamole, a known inhibitor of these transporters.
ImmH was designed from the transition-state structure of bovine PNP and is also a transition-state analogue of human PNP. The second generation analogues, DADMe-ImmH and DADMe-ImmG, were designed specifically from the transition-state structure of human PNP (24). Although bovine and human PNPs are 87% identical, their transition states differ. ImmH binds more tightly to bovine PNP and is severalfold weaker for human PNP. DADMe-ImmH is a chemically simpler molecule with two stereochemical centers and a 22 pM inhibitor of human PNP at 37°C, as reported here. All of these inhibitors share an elevated pK a for N7 in the deazapurine group and a cationic nitrogen to mimic the ribocation charge of the transition state.
These Immucillins demonstrate slow-onset inhibition where the initial enzyme-inhibitor (E-I) complex slowly undergoes a conformation change into the tightly bound complex (E*-I). The initial complex is formed rapidly and is governed by steps k 3 and k 4 , whereas the formation and dissociation of E*-I are governed by slow steps k 5 and k 6 , respectively (Fig. 2). Crystallographic and hydrodynamic analysis of tightly bound PNP-Immucillin complexes show that PNP condenses around the inhibitor with the formation of multiple new hydrogen bonds and ionic interactions (25)(26)(27). PNP catalytic sites are formed by interactions from two subunits, explaining the full inhibition by binding of a single inhibitor molecule to one of the three catalytic sites and the strong negative cooperativity of inhibitor binding.
DADMe-ImmH single-dose oral administration in mice caused blood PNP inhibition in minutes with continued inhibition of erythrocyte PNP for the cell replacement time (11.5 day t1 ⁄ 2 ) despite rapid clearance of drug (4). This difference in offrate constants (DADMe-ImmH release from PNP is t1 ⁄ 2 ϭ 8.3 min at 37°C, as reported here) suggests irreversible tight binding in vivo or frequent rebinding. Here we resolve these possibilities by the systematic measurement of off-rates for PNP transition-state analogues using purified human PNP, human erythrocytes (RBCs), and data from human phase 1 clinical trials. Although inhibitor dissociates from PNP in minutes in vitro, inhibitor rebinding in vivo maintains PNP in an inhibited state. Inhibitor displacement experiments establish the dynamic exchange of the dissociation-association process in cells. Fast binding k on rates and slow dissociation rates and multiple intracellular rebinding events cause the sustained inhibition profile of the Immucillins.

Inhibition studies
Inhibition constants (K i and K i *) were determined for seven PNP transition-state analogues at 25 and 37°C (Fig. 3). The data obtained from the assays were fit to Equation 1 (Fig. 4).
Inhibition constants were calculated from Equation 1 for each inhibitor (Table 1). All inhibitors show slow-onset tightbinding inhibition with dissociation constants in the picomolar range. At both temperatures, SerMe-ImmG was the tightest, whereas ImmH was the weakest inhibitor. The K i values are more sensitive to temperature than K i * values, suggesting that the more open, Michaelis-like complexes governing K i are more temperature-sensitive than the more constrained ensem-

Biological efficacy of transition-state analogues
bles with bound transition-state analogues. Table 1 provides a direct temperature comparison for the four generations of Immucillins, significant here for comparing in vitro and in vivo analysis.

Inhibitor dissociation rates from PNP
Initial off-rates and inhibitor half-lives for each PNP inhibitor were measured at 37°C by diluting a stoichiometric enzyme-inhibitor mixture and observing the partial recovery of PNP catalytic activity (Equation 2; Table 2; Fig. 5). DATMe-ImmG exhibited the fastest release rate followed by SerMe-ImmH, SerMe-ImmG, ImmH, DADMe-ImmG, DADMe-ImmH, and DATMe-ImmH, roughly in the order of their K i * values. For these inhibitors, the t1 ⁄ 2 for partial regain of catalytic activity exhibited a range of 3 to 31 min. Even with a 1000-fold dilution and high substrate concentration in the assay mixtures, only partial regain of activity was observed, as expected for inhibitors in the pM range for K i *.
Temperature dependence for the K i * dissociation constant is contrary to the usual pattern for enzyme-inhibitor complexes, where affinity increases at lower temperatures. For all of the inhibitors, the values of K i * indicate equal or tighter binding at 37°C. This pattern has been observed before for 5Ј-methylthioadenosine phosphorylase and has been interpreted to indicate catalytic site dynamic motion contributing to favorable interaction with transition-state analogues (28).

Off-rate without added inhibitor
Excess [ 14 C]DADMe-ImmH was incubated with RBCs in nutrient media to permit binding to cellular PNP. Extensive washing of the labeled cells before initiation of the experiment removed excess unbound inhibitor. At each time point over a period of 48 h, cells were subject to additional washes, maintaining the external concentration of [ 14 C]DADMe-ImmH near zero. Under these conditions, the amount of [ 14 C]DADMe-ImmH retained in the cell is in equilibrium with diffusive recapture or is non-exchangeable.
The amount of [ 14 C]DADMe-ImmH decreased over the early incubation period (0 -2 h) with a t1 ⁄ 2 of 48 min for 2 ⁄ 3 of the inhibitor loss (Fig. 6). The initial concentration of [ 14 C]DADMe-ImmH in erythrocytes was estimated to be 3.8 M based on the specific radioactivity of the inhibitor. After the initial loss, a slower loss over the following 48 h gave a t1 ⁄ 2 of 84 h for the last 1 ⁄ 3 of the [ 14 C]DADMe-ImmH. The concentration of [ 14 C]DADMe-ImmH in the erythrocytes remained at ϳ1.2 M at 48 h. The initial loss rate constant was 8.7-fold slower than the in vitro off-rate for DADMe-ImmH, and the final rate constant was 916-fold slower than observed off-rate studies. Human PNP is trimeric with the first site binding tightly to transition-state analogues and sites two and three with reduced affinity. Loss of [ 14 C]DADMe-ImmH in the initial phase is interpreted to be from the two more weakly binding sites and the slow phase loss from the tightly bound final, inhibitory catalytic site. Quantitation of the bound [ 14 C]DADMe-ImmH in    Table 1.

Off-rate with added inhibitor
Erythrocytes labeled with [ 14 C]DADMe-ImmH and washed as above were incubated in the presence of excess (300 M) unlabeled DADMe-ImmH (Fig. 7). Microscopic release of labeled inhibitor is prevented from rebinding by competition from the excess unlabeled inhibitor. The experiment determines if the tightly bound inhibitor (Fig. 6) is in dynamic exchange. An increased release rate for [ 14 C]DADMe-ImmH would be expected if the PNP-bound inhibitor is exchanging and subject to diffusive recapture by PNP. The concentration of erythrocyte [ 14 C]DADMe-ImmH decreased from an initial concentration of 3.5 M to a concentration of 0.61 M after 130 min of exchange (Fig. 7). Without excess DADMe-ImmH, the initial phase had a t1 ⁄ 2 of 48 min and the slower phase had a t1 ⁄ 2 of 84 h. With excess DADMe-ImmH, the initial phase gave a t1 ⁄ 2 of 9.1 min, and the slower phase gave a t1 ⁄ 2 of 1.65 h.

PNP catalytic activity studies
From the amount of bound [ 14 C]DADMe-ImmH (Figs. 6 and 7) the PNP subunit concentration was estimated to be 3.8 -4.5 M in erythrocytes. A direct titration of PNP catalytic activity was made in extracts equivalent to 22.5% hematocrit (Fig. 8). Extracts were incubated with varying concentrations of DADMe-ImmH, and the residual PNP activity was analyzed. The concentration of PNP from a catalytic site titration was 0.40 Ϯ 0.01 M (Fig. 8). Extrapolation to 100% hematocrit indicates a 5.3 Ϯ 0.1 M subunits and 1.8 Ϯ 0.1 M trimer.

[ 14 C]DADMe-ImmH uptake
[ 14 C]DADMe-ImmH uptake experiments were performed to determine the role of ENT1 and ENT2 transporters in DADMe-ImmH uptake. Dipyridamole, a known inhibitor of these transporters (29), was tested for its effects at 10 M (Fig. 9). In the absence of dipyridamole and with 2 M

DADMe-ImmH in phase 1 human trials
Analysis of DADMe-ImmH inhibitor rebinding together with the results from a human phase 1 dose-ranging study reveals a mechanism where the initial blood level of DADMe-ImmH rapidly inhibits PNP of RBCs (Fig. 10). In the following 72 h, the clearance of DADMe-ImmH causes blood levels to decline below 10 ng/ml. However, even at 504 h (21 days), the RBC PNP remains strongly inhibited. Slow regain of PNP activity in blood samples occurs with a t1 ⁄ 2 of 59 days after single oral doses of 0.25 to 3 mg/kg. Human erythrocyte lifetime is 120 days. Regain of activity by hematopoiesis alone would, therefore, be expected to give a t1 ⁄ 2 of 60 days. This PNP activity analysis demonstrates that the dominant factor in the regain of PNP catalytic activity is new cell regeneration rather than loss of DADMe-ImmH from the PNP target. To achieve this durable inhibition in vivo, DADMe-ImmH rebinding to PNP inside RBCs occurs with greater efficiency than in the isolated cell systems (e.g. Fig. 6). With isolated erythrocytes, inhibitor rebinding is 50-fold more likely than diffusional loss of DADMe-ImmH from cells. In human trials with DADMe-ImmH, rebinding of the inhibitor is significantly more efficient. With an intrinsic release rate of 1.65 h (Fig. 7) and no significant

Biological efficacy of transition-state analogues
loss of inhibitor over 504 h (21 days) in human trials, the ratio of rebinding to diffusional loss is Ͼ300-fold. The regain of PNP catalytic activity is equivalent to the rate of RBC replacement. Thus, near-stoichiometric DADMe-ImmH rebinding is required to give the extended PNP inhibition kinetics seen in humans (Fig. 10).

Background
Oral administration of a single dose of DADMe-ImmH to mice caused rapid inhibition of blood PNP (t1 ⁄ 2 ϭ 10 min) and an in vivo t1 ⁄ 2 of 11.5 days for regain of catalytic activity (4). As the lifespan of mouse erythrocytes is ϳ25 days, regain of PNP activity can be attributed to erythropoiesis rather than diffusional loss of DADMe-ImmH. Thus, DADMe-ImmH causes apparently irreversible binding to PNP in vivo or is efficiently rebound over multiple release and recapture cycles. A similar pattern was found for inhibition of human blood PNP in phase 1 clinical trials (Fig. 10). Single oral doses of DADMe-ImmH (BCX-4208 in clinical trials) caused elevated blood levels peaking at 4 h and returning to near-baseline levels by 72 h (Fig. 10,  left panel). However, inhibition of blood PNP extended well beyond 72 h, with an in vivo t1 ⁄ 2 of 59 days, consistent with the 120-day lifetime of human erythrocytes.
Here, we compare the microscopic dissociation rates for human PNP-Immucillin complexes in vitro and in vivo inside human erythrocytes. Comparison of release and exchange rates establishes that DADMe-ImmH is being released and rapidly rebound in human erythrocytes. The relative rates permit a mechanistic analysis of the frequency of release and rebinding. Comparison of seven Immucillins provides parameters for evaluating the relative efficiency of these compounds. ImmH has been approved for use against recurrent or resistant peripheral T cell lymphoma in Japan as Mundesine. DADMe-ImmH has completed phase 2 clinical trials for treatment of gout (23), and comparative kinetics will be useful in considering other members of the Immucillin family for pharmaceutical potential. For example, DADMe-ImmG clears Plasmodium falciparum from an Aotus primate model (22). Autoimmune disorders based on auto-antigen T cell activity are also expected to respond to inhibitors of PNP activity (13).

Immucillin affinity and release rates
Slow-onset, tight-binding inhibition is exhibited by the Immucillins described here, a characteristic of transition-state analogues (Figs. 3 and 4). Dissociation constants (K i *) for the Immucillins vary from 115 to 6 pM at 37°C under the conditions of our assays ( Table 1). The unusual temperature dependence with equal or higher inhibitor affinity at elevated temperature has been seen before with transition-state analogues (28). One possible explanation is that increased dynamic motion at the catalytic site at increased temperature permits inhibitors to more closely mimic the dynamic interactions leading to the transition-state (27). Most in vitro inhibitor release rates (t1 ⁄ 2 ) from PNP-Immucillin complexes scaled appropriately with the K i * values, with a fast value of t1 ⁄ 2 ϭ 11 min for ImmH (115 pM) and the slowest t1 ⁄ 2 ϭ 31 min for DATMe-ImmG (6 pM). A notable exception is provided by DATMe-ImmH. With a K i * value of 43 pM, a slow t1 ⁄ 2 inhibitor release time would be expected, but it was near 3 min ( Table 2). The more flexible chemical scaffolds of the DATMe-and SerMe-Immucillins are likely to promote rapid escape from the PNP catalytic sites. The slow release of DATMe-ImmG, SerMe-ImmH, and SerMe-ImmG suggest theses three inhibitors will have the most powerful in vivo action as PNP inhibitors.

[ 14 C]DADMe-ImmH efflux from erythrocytes
The PNP activity of human erythrocytes is readily inhibited by [ 14 C]DADMe-ImmH in vivo (Fig. 10). The kinetics of [ 14 C]DADMe-ImmH release from erythrocytes resolves tight binding from covalent interaction and provides rebinding information. Human PNP is a homotrimer, where filling one of three catalytic sites causes complete inhibition. Binding at the first site gives the kinetic constants from steady-state and slowonset analysis (Tables 1 and 2; Fig. 5). The second and third sites also fill with inhibitor but with negative cooperativity, each at least an order of magnitude weaker than the first (15). ITC titration of human PNP with DADMe-ImmH at 37°C gave an enthalpy of Ϫ21.5 kcal/mol for the first site and an average value of Ϫ12.0 kcal/mol for the second and third sites (15). With a 4Ј-F analogue of DADMe-ImmH, binding at sites 1, 2, and 3 were 2, 64, and 667 nM (15). Incubation of erythrocytes with excess [ 14 C]DADMe-ImmH loads all sites. Incubation

Biological efficacy of transition-state analogues
with multiple washing first removes inhibitor from the weaker sites (sites 2 and 3) for 2 ⁄ 3 loss of [ 14 C]DADMe-ImmH with a t1 ⁄ 2 of 48 min (Fig. 6). This is slow compared with the 8.3 min t1 ⁄ 2 for activity recovery from [ 14 C]DADMe-ImmH binding to the first tight site for PNP (Table 2). Even for PNP weak sites two and three, these are rebinding [ 14 C]DADMe-ImmH approximately six times for each inhibitor molecule diffusing out of the erythrocytes.
The pattern of slow inhibitor loss from the first site for [ 14 C]DADMe-ImmH wash-out experiments from RBCs gave a t1 ⁄2 of 84 h (Fig. 6). The addition of excess unlabeled DADMe-ImmH prevented [ 14 C]DADMe-ImmH rebinding and increased the t1 ⁄2 of 84 h to a t1 ⁄2 of 1.65 h, a 51-fold increase in inhibitor loss rate (Fig. 7). Thus, the inhibitor is in rapid microscopic exchange inside the erythrocytes and rebinds at least 50 times more frequently in the absence of excess DADMe-ImmH than in its presence.

[ 14 C]DADMe-ImmH transport
Transport of ImmH into human CCRF-CEM cells occurs on ENT1 and ENT2 equilibrative nucleoside transporters, also present in erythrocytes and sensitive to dipyridamole inhibition (29). Human RBCs transported [ 14 C]DADMe-ImmH in a dipyridamole-sensitive mechanism, consistent with the same uptake mechanism.

Extended PNP inhibition: Extrapolation to human trials
The long-term biological efficacy of DADMe-ImmH was first seen in mouse experiments, where the inhibitor time on the PNP target approximated the lifetime of mouse erythrocytes. This inhibition characteristic was termed "the ultimate goal" in inhibitor design. We define this interaction as an orally available drug that inhibits the target enzyme for the lifetime of the cell without the complications of covalent attachment. Work here extends the inhibitor "ultimate goal" status to the DADMe-ImmH inhibition of PNP in human blood. The rate of PNP activity regain following DADMe-ImmH treatment depends on the rate of cell replacement. Therefore, tissues with more rapid cell replacement rates are expected to recover PNP activity more rapidly than RBCs. The human clinical trials demonstrated that once a day oral dosing provides adequate whole body inhibition of PNP.

Conclusions
Transition-state analogue inhibitors of human PNP bind with picomolar dissociation constants and cause slow-onset tight-binding inhibition. Residence times of these inhibitors on the PNP target are on the timescale of minutes with in vitro enzyme-inhibitor complexes. The in vivo efficacy in isolated RBCs is much greater than predicted by target residence times. Labeled inhibitor-exchange experiments in isolated RBCs demonstrate that the long inhibitor residence time on the PNP target is caused by inhibitor rebinding to the target more frequently than inhibitor loss from the cell. Inhibitor efficacy in human phase 1 clinical trials indicate that inhibitor rebinding is more effective in vivo, with no significant PNP activity regain from circulating RBCs over their 120-day lifetime in humans.

Biological efficacy of transition-state analogues Protein expression and purification
Human PNP was expressed and purified as previously described with slight variations (4). A starter culture containing ampicillin (15 ml) was added to 1 liter of LB broth along with ampicillin stock to 100 g/ml. Expression was induced using isopropyl ␤-D-1-thiogalactopyranoside (IPTG, 1 mM). Freshly harvested cells (ϳ20 g) were treated with a few mg of DNase (powdered form), and the cells were disrupted by sonication. Protein was purified by AKTA FPLC with a nickel-nitrilotriacetic acid column of 5 ml. The column was loaded with 42 ml of extract. It was washed with 10 mM imidazole, 300 mM NaCl, and 50 mM potassium phosphate (pH 8.0). Elution was effected by a gradient (1 ml/min) from wash buffer to 150 mM imidazole in the same buffer over 30 min followed by a second gradient from 150 to 500 mM imidazole over 10 min. Human PNP was eluted at 150 -250 mM imidazole. The protein was dialyzed against 100 mM NaCl and 50 mM potassium phosphate (pH 8.0). Concentrated protein was stored frozen at Ϫ80°C. Hypoxanthine co-purifying with the PNP was removed by dialysis against the same buffer containing 0.5% (w/v) powdered charcoal.

Kinetic inhibition studies
Kinetic parameters for PNP inhibition by Immucillins were measured in a mixture of 1 ml containing 60 milliunits of xanthine oxidase, 1 nM enzyme, 50 mM phosphate buffer (pH 7.5), and inosine (1 mM or 10 mM depending on the inhibitor), with concentrations of inhibitor from 0.1 nM to 316 nM. The reaction was monitored using a CARY 300 Bio UV spectrophotometer running the reactions at 25 or 37°C and monitoring the progression at 293 nm for 1 h. To determine K i and K i *, the competitive inhibition equation was used (Equation 1), where V 0 Ј is the initial reaction rate in the presence of inhibitor, V 0 is the initial reaction rate in the absence of inhibitor, [I] is the inhibitor concentration, and [A] is the substrate concentration.
In addition to this equation, Equation 2 below was used to calculate the amount of free inhibitor concentration (for use in Equation 1), where [I] is the total inhibitor concentration, [E] t is the total enzyme concentration, and [I]Ј is the amount of free inhibitor. For each inhibitor, dissociation constants were determined for the initial reaction rates (K i ) and from reaction rates after the system had equilibrated (K i *). Experimental data were fit to the equations by the use of GraphPad Prism 7.

Off-rates from purified human PNP
Purified human PNP (0.5 M) and the inhibitors (0.5 M) in 50 mM phosphate buffer (pH 7.5) (total volume of 50 l) were incubated at 37°C for 1 h. The mixture was diluted 1:1000 using 50 mM phosphate buffer and assayed in the presence of inosine (2 mM) and 60 milliunits of xanthine oxidase. Product formation was observed at 293 nm for 3.5-5 h. Controls were included where enzyme or inhibitor were absent. The data were fit to Equation 3, where P is product formation, k is rate constant for inhibitor release, and v 0 and v s are the initial and steady-state rates, respectively. The half-life for the PNP-inhibitor complex was calculated from Equation 4,

Inhibitor release from isolated red blood cells
Freshly collected blood was washed 3 times by dilution into equal volumes of PBS and centrifugation for 3 min at 400 ϫ g. Pelleted erythrocytes (RBCs) were suspended in 1640 RPMI media to the desired hematocrit before use. Inhibitor [ 14 C]DADMe-ImmH (4 M) and washed RBCs (100 l, 50% hematocrit) were incubated (15 min at 37°C) and centrifuged at 3000 ϫ g for 2 min, and the cells were washed with medium twice to eliminate any excess inhibitor. The RBCs were suspended in 450 l of medium (to give 500 l) and incubated for 5 min. The supernatant along with 200 l of a second wash was collected for analysis. Cells were suspended in 450 l of medium. A sample (10 l) was compared with the collected supernatant. This analysis gave the amount of inhibitor in the supernatant and RBCs at time 0. The sample was incubated, and the analysis was repeated for time points 0.5, 1, 1. 5,2,4,6,8,16,24,32,40, and 48 h. The washed RBC pellets were lysed by the addition of 90 l of water, and samples of the RBCs and supernatants were transferred to scintillation vials where 10-ml scintillation fluid was added. The radioactivity in each sample was counted in a liquid scintillation analyzer Tri-Carb 2910 TR. The data were analyzed using Equation 3, where CPM replaced P in the formulation. In this formulation, v 0 and v s are the initial and steady-state release rates (CPM/min), and k is the rate constant.

Biological efficacy of transition-state analogues PNP inhibitor off-rates from isolated human RBCs
Erythrocyte cellular release of [ 14 C]DADMe-ImmH was measured by multiple washes of erythrocytes previously equilibrated with the inhibitor. Cell labeling with radiolabeled [ 14 C]DADMe-ImmH permitted monitoring inhibitor content in RBCs and release to the extracellular media.

[ 14 C]DADMe-ImmH release with excess unlabeled inhibitor
[ 14 C]DADMe-ImmH release from human RBCs was measured in the presence of excess unlabeled inhibitor. The procedure was the same as described above except the resuspension of the RBCs was in RPMI medium containing 300 or 3000 M unlabeled DADMe-ImmH followed by incubation at 37°C. Immediately after resuspension, a 50-l sample was collected and centrifuged at 3000 ϫ g for 0.5 min, and the supernatant was removed. This procedure was repeated for: 1,4,8,16,32,48,64,96, and 128 min. Those samples were then set aside. The RBC pellets for each time point were lysed with 90 l of water. The RBC lysates and supernatant samples were transferred to scintillation vials, and the radioactivity was measured and analyzed (Equation 3).

[ 14 C]DADMe-ImmH Uptake
Whole human blood (150 l) was washed with PBS (850 l) and centrifuged at 400 ϫ g for 2 min, removing the supernatant. The isolated RBCs (75 l) were suspended in 675 l of 1640 RPMI media (total volume of 750 l) and divided into two 300-l samples to examine the uptake of [ 14 C]DADMe-ImmH with and without 10 M dipyridamole. The mixtures (343 l) were incubated at 37°C for 10 min with or without dipyridamole, after which [ 14 C]DADMe-ImmH (100 M, 7 l) was added to each tube (2 M final). Samples of 7 l were taken to determine the total counts in each mixture. Subsequent samples (35 l) were taken at the desired times and added to 1 ml of ice-cold PBS to quench transport. The cold samples were centrifuged (2000 ϫ g for 2 min), discarding the supernatant. Water (90 l) was added to lyse the cells, and samples were counted as described above (Fig. 9).
Plasma BCX4208 was determined by a validated LC/MS/MS Method. BCX4208 was extracted from plasma using a Waters Oasis MCX solid phase extraction cartridge on a Zymark RapidTrace Work station. Chromatography was performed isocratically using a Zorbax SB C-3 column with isocratic elution using a mobile phase of 5% methanol in 0.1% acetic acid and methanol (97:3). An Agilent 1100 HPLC system was used for the chromatography. Column effluent was analyzed by positive ion multiple reaction monitoring 265 m/z and 147.9 m/z using a PE Sciex API 2000 MS/MS equipped with Turbo Ion Spray in positive ion mode. The concentration of BCX-4208 was then determined by weighted (1/x) quadratic regression analysis of peak areas produced from the standard curve spanning 5 to 2000 ng/ml.
Oral dosing with BCX4208 resulted in a dose-dependent increase of C max for the doses up to 2.5 mg/kg, cohort 6, with possible plateau for C max at the 2.5 mg/kg dose. C max was achieved 3-7 h after oral administration of BCX4208. Overall, C max and the area under the curve (AUC) increased in a doseproportional manner.

Erythrocyte PNP inhibition in vivo
Inhibition of PNP enzyme activity was measured ex vivo in erythrocytes taken from subjects in all dose groups (0.25-3.0 mg/kg) in the phase I study. PNP enzyme activity in erythrocyte extracts was measured by the conversion of inosine to hypoxanthine using a spectrophotometric assay. A Cary 3 Spectrophotometer, Varian Model 1001206, equipped with Cary WinUV software was used for these studies.
Oral administration of BCX4208 resulted in rapid inhibition of erythrocyte PNP activity. Maximum mean PNP inhibitory activity was Ն80% in all dose groups. Inhibition was rapid and maintained at the elevated levels for at least 21 days after the last dose. The data were used to show a dosedependent decrease in the time to maximum PNP inhibition and the rates of recovery with the slopes almost identical across the dose groups.