PEDS Advance Access originally published online on April 18, 2006
Protein Engineering Design and Selection 2006 19(7):291-297; doi:10.1093/protein/gzl011
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
The pharmacokinetics of an albumin-binding Fab (AB.Fab) can be modulated as a function of affinity for albumin
1 Departments of Assay and Automation Technology, Genentech, Inc. South San Francisco, CA 94080, USA 2 Departments of Pharmacokinetic and Pharmacodynamic Sciences, Genentech, Inc. South San Francisco, CA 94080, USA 3 Departments of Antibody Engineering, Genentech, Inc. South San Francisco, CA 94080, USA 4 Departments of Process Sciences, Genentech, Inc. South San Francisco, CA 94080, USA
5To whom correspondence should be addressed. E-mail: msd{at}gene.com
 |
Abstract |
|---|
 Top Abstract IntroductionMethodsResultsDiscussionAcknowledgementsReferences |
|---|
An AB.Fab (albumin-binding Fab) consists of a Fab and a phage-derived albumin-binding peptide. This molecule is capable of binding both antigen and albumin simultaneously. Using a Fab derived from Herceptin® we generated a panel of AB.Fab variants with wide-ranging affinities for albumin. An assay that measured AB.Fab binding to albumin in solution was developed to most accurately reflect the binding affinity for albumin in vivo. Affinity varied depending upon the species of albumin tested. For rat and rabbit albumin, affinities ranged from 0.04 to 2.5 µM. Reduced affinity for albumin correlated with a reduced half-life and higher clearance rates in both species; the beta half-life ranged 6-fold while clearance ranged over 50-fold in rats and 20-fold in rabbits. To estimate the pharmacokinetic properties of an AB.Fab in humans, AB.Fab variants with similar affinities for rat and rabbit albumin were selected. Using their pharmacokinetic parameters and the principles of allometric scaling for albumin, we estimate an approximate beta half-life for an AB.Fab with 0.5 µM affinity for albumin of up to 4 days in humans with a clearance of 76 ml/h. These variants demonstrate the ability to modulate the clearance of a Fab fragment in vivo and help to establish guidelines for pharmacokinetic engineering of molecules through albumin binding.
Keywords: AB.Fab/albumin affinity/albumin-binding Fab/half-life/Herceptin/pharmacokinetics
|
Introduction |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionAcknowledgementsReferences |
|---|
Amplification of the growth factor receptor HER2 in 20–30% of human breast and ovarian cancers is associated with decreased median patient survival time. Significant clinical benefit is gained when Herceptin® (Trastuzumab), a humanized monoclonal antibody, is used for the treatment of these carcinomas (Slamon et al., 1989
; Carter et al., 1992b; Baselga et al., 1996; Slamon et al., 2001; McKeage and Perry 2002; Vogel et al., 2002). The long circulatory half-life of Herceptin® as well as other immunoglobulins stems from their large molecular weight (
150 kD) and the presence of recycling mechanisms in vivo that enable convenient dosing in many therapeutic settings (Ghetie and Ward, 2000).
Diffusion of an IgG into a tumor is slow as a result of its large size and the interstitial pressure present in the tumor (Jain, 1989, 1999; Graff and Wittrup, 2003); yet, its long permanence in serum gradually drives its diffusion into the tumor. This prolonged exposure in both normal and diseased tissue may be less than ideal when an IgG is conjugated with imaging or cytotoxic agents. A shorter half-life and higher rate of diffusion may enhance tumor penetration and limit systemic exposure. Antibody fragments such as scFv, diabodies and Fabs offer rapid tumor penetration and have been explored for these indications; however, these fragments are cleared rapidly and their ability to be retained in the tumor is limited (Wu and Yazaki, 2000; Kashmiri, 2001; Todorovska et al., 2001; Borsi et al., 2002). The ideal targeting molecule that fully penetrates tumors, provides sufficient exposure for good tumor retention, and yet clears rapidly so as to limit normal tissue exposure remains to be realized.
We have sought to take advantage of the high concentration and long serum half-life of albumin as means to enhance the half-life of a Fab. Albumin, present in plasma at 600 µM, plays a vital role in vivo by reversibly binding and transporting a wide variety of endogenous substances and drugs (Peters, 1996). Owing to its high concentration, even relatively weak interactions with albumin (affinities in the range of 0.1–100 µM) can play an important role in the in vivo exposure of many organic anions and long chain fatty acids. Fatty acid acylated insulins have been shown to exploit these ligand-binding sites to extend the action of insulin in pigs (Markussen et al., 1996). Similarly, we have demonstrated that the clearance of a Fab fragment can be dramatically decreased through association with albumin (Dennis et al., 2002). Association with albumin was accomplished by fusing an albumin-binding peptide sequence to the light chain of the Fab generating an albumin-binding Fab (AB.Fab) fully capable of binding antigen while bound to albumin. This albumin-binding peptide, identified using peptide phage display, binds with a stoichiometry of 1:1 and at a site distinct from known small molecule ligand-binding sites on albumin. This site is conserved among albumins from multiple species facilitating studies in many different animal models; however, the affinity of the AB.Fab for albumin and the in vivo half-life of albumin is species dependent. This complicates the ability to predict pharmacokinetic properties from one species to another.
The half-life of albumin in different species generally adheres to an allometric scaling based upon animal weight. For example, in mouse, rat, rabbit and humans it has been estimated as 1, 1.9, 5.55 and 19 days, respectively (Sterling, 1957; Reed and Peters, 1984; Stevens et al., 1992; Hatton et al., 1993), and suggests the relationship: albumin half-life (days) = 3.75 * Body Weight (kg)0.368 assuming typical body weights of 0.02, 0.25, 3 and 70 kg, respectively (Davies and Morris, 1993). If a method could be developed to take into account the affinity of an AB.Fab for albumin in one species relative to another, allometric scaling based on albumin could be used to predict the pharmacokinetic properties of an AB.Fab in human. Thus, we sought to design AB.Fab variants with affinities for rat or rabbit albumin that were comparable to the affinity for human albumin. Albumin affinity was modified by shortening the albumin-binding peptide and an assay to measure the dissociation constant (Kd) of the AB.Fab variants for soluble albumin was used in order to accurately assess albumin binding in vivo. The AB.Fab variants demonstrated a direct correlation between albumin affinity and their pharmacokinetic attributes enabling us to define the albumin affinity required to achieve a desired phamacokinetic profile. Taken together with the half-life of albumin among various animal species, we have predicted the terminal half-life and clearance of an AB.Fab in humans.
|
Methods |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionAcknowledgementsReferences |
|---|
Construction of AB.Fab variants
AB.Fab variants 4D5-H, 4D5-H4, 4D5-H8 4D5-H10 and 4D5-H11 were engineered to have a wide range of binding affinities for albumin and were designed based on the binding affinities of peptides SA06, SA26, SA30, SA04 and SA34, respectively, for rabbit albumin (Dennis et al., 2002). AB.Fab variants 4D5-H, 4D5-H4 and 4D5-H8 were constructed by digesting pAK19 (Carter et al., 1992a) with Sal I and Sph I, and replacing this region with an annealed and ligated 4-oligonucleotide cassette. This cassette introduced albumin-binding peptide sequences of varied length to the carboxyl terminal of the Fab heavy chain 4 residues after the last cysteine in the constant domain (i.e. following the sequence: CDKTH, Figure 1). 4D5-H included a linker sequence (GGGS) that was omitted in the other variants. AB.Fab variants 4D5-H10 and 4D5-H11 were constructed by introducing deletions into 4D5-H4 using QuikChange mutagenesis (Stratagene, La Jolla, CA).
|
Purification of AB.Fab variants
AB.Fab variants were expressed in Escherichia coli and secreted into the periplasmic space (Carter et al., 1992a). Frozen cell paste was suspended in 25 mM Tris, 25 mM NaCl, 5mM EDTA, pH 7.1 and homogenized using a Microfluidic Corporation HC 8000 homogenizer. Polyethyleneimine was added to facilitate clarification of the homogenate at up to 0.5% v/v prior to centrifugation.
A rabbit albumin affinity column and a cation exchange resin were used to purify the AB.Fab variants. Rabbit albumin (Sigma, St Louis, MO), coupled to CNBr-activated Sepharose 4B (Amersham Biosciences, Piscataway, NJ) as per manufacturer's guidelines, was used as an affinity matrix for the capture of proteins presenting an albumin-binding peptide. The albumin affinity column selectively purified AB.Fab variants with the proper disulfide bond correctly folded within the albumin peptide. The column was equilibrated, loaded and washed at neutral pH and AB.Fab variants were eluted using 25 mM citrate, 25 mM NaCl, pH 2.8. Elution pools were adjusted to pH 5.5 with 1.5 M Tris base. Elution pools from the albumin affinity column were loaded onto a cation exchange column (SP SFF; Amersham Biosciences) and washed at pH 5.5. AB.Fab variants were eluted using a NaCl gradient, eluting at 50 mM NaCl. The SP SFF elution pools were formulated by ultrafiltration against 50 mM potassium phosphate, pH 6.0, diluted to a final concentration of 10 g/l and filtered using a 0.22 mm cellulose acetate vacuum filter (Corning Inc. Life Sciences, Acton, MA). Yields of AB.Fab variants ranged between 0.4 and 0.6 g AB.Fab/kg of E.coli cell paste.
Direct Binding ELISA
Mouse, rat, rabbit or human albumin (Sigma) were immobilized onto NUNC Maxisorp 96-well plates at 2 µg/ml overnight at 4°C. The plates were blocked with binding buffer (PBS, 0.5% ovalbumin and 0.05% Tween-20) for 1 h at 25°C. AB.Fab variants were serially diluted in binding buffer and added at 100 µl per well to the immobilized albumin for 30 m at 25°C. Unbound AB.Fab variant was removed by washing wells with 0.05% PBS/Tween-20 and bound AB.Fab variant was detected with goat anti-human Fab'2-horseradish peroxidase (HRP) for 1 h at 25°C. Bound HRP was measured with a solution of tetramethylbenzidine/H2O2. After 15 m, the reaction was quenched by the addition of 1 M phosphoric acid. The absorbance at 450 nm was read with a reference wavelength of 650 nm.
Solution Binding ELISA
A fixed concentration of AB.Fab variant (determined in above binding ELISA) was incubated in solution with varying concentrations of albumin. After at least a 2 h incubation at room temperature, 100 µl of the reaction mixture was transferred to an albumin coated ELISA plate to capture unbound (free) AB.Fab variant. The Direct Binding ELISA, described above, was then used to determine the concentration of captured AB.Fab variant.
The fixed concentrations of AB.Fab variant and the starting concentrations of each albumin are listed in Table I. Curves were generated using 8 concentrations of mouse, rat, rabbit or human albumin diluted serially (1:3) from the maximum concentrations indicated in Table I.
|
Pharmacokinetic studies in rat, rabbit and mouse
All pharmacokinetic (PK) studies were conducted according to protocols approved by the Institutional Animal Care and Use Committee at Genentech, Inc. Sprague Dawley rats and BALB-c mice were supplied by Charles River Laboratories (Hollister, CA, USA). New Zealand White (NZW) rabbits were supplied by Myrtle's Rabbitry (Thompson Station, TN, USA).
Rats weighing between 279 and 314 g were given a 5 mg/kg body weight, IV bolus dose of AB.Fab variant 4D5-H, 4D5-H4, 4D5-H8, 4D5-H10 or 4D5-H11 (n = 4 rats/group) via a cannula inserted in the femoral vein. At pre-dose, and over the course of 7 days post-dose, plasma was collected via a cannula inserted in the jugular vein.
Rabbits weighing between 2.9 and 3.5 kg were given a 0.5 mg/kg body weight, IV bolus dose of AB.Fab variant 4D5-H, 4D5-H4, 4D5-H8, 4D5-H10, 4D5-H11 or Fab4D5 (n = 3 rabbits/group) via a catheter inserted in the marginal ear vein. At pre-dose and over the course of 21 days post-dose, serum was collected via an arterial catheter inserted in the contralateral ear.
Mice weighing between 17 and 20 g were given 5 mg/kg body weight IV bolus dose of AB.Fab variant 4D5-H, 4D5-H4 or 4D5-H8 (n = 9 mice/group) via the tail vein. Over the course of 7 days serum was collected in three mice per time point by retro-orbital bleed or cardiac stick.
All serum and plasma samples were assayed for Fab4D5 or AB.Fab variant concentrations using a HER2 Binding ELISA. Samples were added to microtiter wells coated with HER2 extracellular domain for 2 h. The wells were washed, and goat anti-huFab-HRP was added for 1 h. Unbound HRP was removed by washing and enzyme substrate was added to detect bound HRP. After 15 m, the reaction was quenched by the addition of 1 M phosphoric acid. The absorbance at 450 nm was read with a reference wavelength of 650 nm. Concentrations of AB.Fab were extrapolated by comparison to a standard curve of the dosed molecule.
For rats and rabbits, serum concentration versus time profiles displayed biphasic elimination. For each animal concentration-time profiles were fit to a two compartment model of the form, C(t) = A x exp(–
t) + B x exp(–ßt) using iterative re-weighting to estimate the pharmacokinetic parameters of area under the concentration–time curve (AUC) from time zero extrapolated to infinity, clearance (CL), beta half-life (
), volume of distribution of the central compartment (V1), and steady state volume of distribution (Vss) using WinNonlin software (Pharmacokinetic Model 8, Version 3.2, Pharsight, Inc., Mountain View, CA) and group means were calculated (Rowland and Tozer, 1995).
In mice, a group mean serum concentration versus time profile was determined, producing one estimate for pharmacokinetic parameters for each AB.Fab dosing group. AB.Fab variants 4D5-H and 4D5-H4 were analyzed by a two compartment model as described above. AB.Fab variant 4D5-H8, which displayed a monoexponential serum concentration versus time profile, was fit to a one compartment model of the form, C(t) = Dose/[V1 x exp(–t)], and estimates for the PK parameters of AUC, CL, half-life () and V1 were determined (Pharmacokinetic Model 1, WinNonlin).
|
Results |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionAcknowledgementsReferences |
|---|
Design, construction and purification
AB.Fab variants were engineered to possess a wide range of affinities for albumin based upon our previous investigations of peptides binding to rabbit albumin (Dennis et al., 2002). In an attempt to maintain relative in vivo stability of the AB.Fab variants, we chose to reduce the length of the albumin-binding peptide rather than to alter its amino acid sequence (Fig. 1). In addition we discovered that the linker used between the Fab and the peptide could be deleted without significantly affecting albumin binding.
Despite their differing affinities for rabbit albumin, all of the AB.Fab variants could be rapidly and efficiently purified using a rabbit albumin affinity column. AB.Fab variants were essentially >99% pure by SDS–PAGE following the albumin affinity column, however, an additional cation exchange step was required to remove trace endotoxin and E.coli proteins in order to make the proteins suitable for in vivo studies. AB.Fab variants were deemed monomeric by size exclusion chromatography. An SDS–PAGE analysis of AB.Fab variants 4D5-H, 4D5-H4, 4D5-H8, 4D5-H10 and 4D5-H11 is shown in Fig. 1.
AB.Fab affinities for albumin
Several assays were explored in an effort to accurately determine the affinity of the AB.Fab variants for mouse, rat, rabbit and human albumin. Initially surface plasmon resonance was employed. Direct AB.Fab variant binding to immobilized albumin from the various species resulted in inconsistent kinetic measurements for some of the weaker affinity variants (e.g. AB.Fab variants 4D5-H10 and 4D5-H11); however, the relative rank affinity of the variants could be determined as 4D5-H > 4D5-H8 > 4D5-H4 > 4D5-H10 > 4D5-H11, with 4D5-H showing the highest affinity. This affinity ranking of the AB.Fab variants remained the same whether binding to mouse, rat or rabbit albumin.
As an alternative to surface plasmon resonance, we chose to develop an ELISA method that could accurately assess the dissociation constant (Kd) of the AB.Fab variants to the different species of albumin. Initially we explored binding of the AB.Fab variants to immobilized albumin using a Direct Binding ELISA. The EC50 for each AB.Fab and albumin combination is listed in Table II. Estimates of EC50 were made for AB.Fab variants 4D5-H10 and 4D5-H11 as a result of their lower signals.
|
Potential artifacts arising from the absorption of albumin on plastic could distort the true binding affinities of the AB.Fab variants for albumin in solution (i.e. plasma). To eliminate this possibility, we developed a 2-step ELISA approach first described by Friguet et al. (1985)
and also used for determining the Kd of humanized antibodies to HER2 (Carter et al., 1992a). This assay established a solution phase binding equilibrium followed by detection of unbound (free) AB.Fab using the Direct Binding ELISA.
The solution phase binding equilibrium, generated in the Solution Binding ELISA, requires that the concentration of AB.Fab variant should be as low as possible yet still provide a sufficient signal to measure free AB.Fab. The minimum concentration of each AB.Fab variant used with each species of albumin was determined by titrating the AB.Fab variant in the Direct Binding ELISA using the corresponding immobilized albumin. This minimum concentration of AB.Fab was then titrated with soluble albumin in the Solution Binding ELISA. The minimum concentration of each AB.Fab variant and the initial albumin concentration used are listed in Table I. The concentration of unbound (free) AB.Fab variant was then determined utilizing the Direct Binding ELISA in which the corresponding albumin was immobilized. In order to determine the Kd-value using Scatchard Analysis (Scatchard, 1947), it is important that the equilibrium between the AB.Fab variant and albumin in solution be reached prior to the determination of un-bound AB.Fab variant.
To verify that the AB.Fab-albumin mixture had reached equilibrium, AB.Fab4D5-H was incubated with rabbit albumin at various times before the mixtures were assayed in the Direct Binding ELISA. Equilibrium was essentially reached after a 2 h incubation. The optimal time needed to capture unbound (free) AB.Fab4D5-H to immobilized albumin was determined by incubating the AB.Fab-albumin mixture with immobilized albumin. The minimum amount of time required to bind the free AB.Fab4D5-H was 15 m; however, 30 m was ultimately used out of convenience.
Under these experimental conditions, the capture of free AB.Fab in the Direct Binding ELISA does not significantly shift the AB.Fab and albumin equilibrium, so the fraction of bound AB.Fab variant (v) is related to the signal measured in the ELISA. Since the concentration of total albumin is 10- to 1000-fold higher than the concentration AB.Fab variant, the concentration of free albumin (a) approximates the total albumin concentration. Thus, the Kd can be determined by plotting the fraction of bound (v) versus v/a (Scatchard, 1947). A comparison of the AB.Fab variant affinities for mouse, rat and rabbit albumin determined by these various assay methods is summarized in Table II along with the affinity of 4D5-H for human albumin. The affinity values obtained for other AB.Fab variants with human albumin were >10 µM and deemed unreliable.
The AB.Fab variants present an even distribution of affinities for rabbit albumin over a 30-fold range from 36 to 1110 nM. Similarly in rat, the affinities ranged over 27-fold from 92 to 2429 nM. In mouse, however, the distribution was different with 4D5-H, 4D5-H4 and 4D5-H8 having very similar affinity and 4D5-H10 and 4D5-H11 displaying very weak affinity for mouse albumin.
Pharmacokinetics of AB.Fab variants in rat, rabbit and mouse
To explore the role of affinity on the ability of albumin to extend the half-life of an AB.Fab, we investigated the PK of the AB.Fab variants in mouse, rat and rabbit. In rats and rabbits, an affinity dependent increase in exposure (AUC) was observed. Group mean PK parameters are listed in Table III. The pharmacokinetic profiles of the AB.Fab variants in rat and rabbit are shown in Fig. 2. In both rats and rabbits, the AB.Fab variants displayed biphasic elimination.
|
|
In rats, a 27-fold increase in albumin-binding affinity resulted in a 50-fold increase in exposure. AUC ranged from 57.0 ± 7.39 to 2880 ± 466 h x µg/ml. In rabbits, a 30-fold increase in albumin-binding affinity among the AB.Fab variants resulted in a 20-fold increase in exposure. AUC among the AB.Fab variants ranged from 24.6 ± 2.74 to 514 ± 48.9 h x µg/ml. AUC for Fab4D5 was 8.75 ± 1.05 h x µg/ml.
Consistent with the changes in AUC, increased affinity of the AB.Fab variants resulted in decreased clearance in rats and rabbits (Table III). In rats, clearance decreased 53-fold, with clearance ranging from 80.2 ± 11.8 to 1.50 ± 0.220 ml/h/kg. In rabbits, clearance of the AB.Fab variants ranged from 20.6 ± 2.24 to 0.841 ± 0.053 ml/h/kg, approximately a 24-fold decrease. By comparison, the clearance of Fab4D5 (60.8 ml/h/kg), with no specific binding affinity for albumin, was 3- to 73-fold faster in rabbits than any of the AB.Fab variants.
In both rats and rabbits, the volume of distribution of the central compartment (V1) for all AB.Fab variants approximated serum volume. Also in both rats and rabbits, terminal half-life increased 6-fold as a result of increased affinity among the AB.Fab variants. Terminal half-life of the AB.Fab variants ranged from 4.21 ± 0.151 to 26.9 ± 3.11 h in rats and 11.9 ± 2.61 to 68.5 ± 5.58 h in rabbits.
In summary, there was a direct correlation between AB.Fab variants with a high affinity for albumin and a slower clearance and longer half-life in either rat or rabbit. Interestingly, when the PK of AB.Fab variants 4D5-H, 4D5-H4 and 4D5-H8 were investigated in mouse, all three variants displayed similar clearance. This is not surprising given their similar affinities for mouse albumin (Table II). Further, in mouse and rabbit, all AB.Fab variants tested had a slower clearance than Fab4D5.
Correlation between albumin-binding affinity and clearance
The correlation between albumin binding and beta half-life is illustrated for both rabbits and rats in Fig. 3. The separation of the curves for rabbits and rats illustrates that AB.Fab variants with similar albumin binding must be compared for appropriate allometric scaling. In rabbits AB.Fab4D5-H4 has binding affinity of 444 nM and in rats AB.Fab4D5-H10 has binding affinity of 493 nM similar to the affinity of AB.Fab4D5-H in human of 556 nM (Table II). Utilizing the fact that the pharmacokinetics of albumin strongly correlate with body weight, the PK parameter estimates of clearance and beta half-life for AB.Fab4D5-H4 in rabbits and AB.Fab4D5-H10 in rats as a function of body weight were extrapolated to human (70 kg) and can be described using the following equations:
 |
 |
|
Using the equations above, the predicted clearance and beta half-life of an AB.Fab with a binding affinity of
500 nM in humans is 76 ml/h (1 ml/h/kg) and 4 days, respectively. A more robust estimate that would have included mouse in the allometric scaling is precluded since the binding affinities of the AB.Fab variants for mouse albumin are either 10-fold tighter or 2.5-fold weaker than the range utilized in the above prediction. |
Discussion |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionAcknowledgementsReferences |
|---|
Covalent association with albumin has been achieved through the genetic fusion of rapidly cleared proteins to albumin (Yeh et al., 1992
; Syed et al., 1997; Marques et al., 2001; Smith et al., 2001; Sung et al., 2003), through non-specific chemical modification to attach proteins (Wong et al., 1980) or small molecules (Stehle et al., 1997), and through specific modification using the reactive free cysteine in albumin (Kratz et al., 2000; Smith et al., 2001). Unlike these covalent fusions to albumin, the non-covalent association of the AB.Fab variants with albumin allows their clearance to be modified. The fraction of free unbound AB.Fab in serum can be calculated using the affinity of each variant for albumin, the reported serum concentration of albumin [600 µM (Peters, 1996)] and the following equation:
 |
Thus, it appears that a very small difference (e.g. 0.1%) in the fraction of AB.Fab that is unbound by albumin in vivo will have a profound effect on its rate of clearance. Thus, the affinities of these AB.Fab variants for albumin lie in a range where an increase or decrease in association has a measurable effect on clearance and half-life. The curves in Fig. 3 indicate that further increases in albumin-binding affinity could lead to further increases in half-life since a plateau has not yet been reached, although at some point other clearance mechanisms may come into play.
Albumin binding has been employed previously as a strategy for reducing the clearance of fatty acid acylated insulins (Markussen et al., 1996). A direct correlation between albumin binding and clearance in pigs was observed for nine derivatives over an affinity range from 4 to 70 µM. The AB.Fab variants presented here have affinities for albumin that are 10-fold higher and demonstrate a continued and more dramatic reduction in clearance. Further, we have shown previously that AB.Fab association with albumin does not impair the interaction of the Fab with antigen nor does it compete with any of the physiological known ligands that are carried by albumin in vivo (Dennis et al., 2002). Interestingly, the combined affinity range observed to impact half-life of the AB.Fab variants and the fatty acid acylated insulins is similar to the affinity range of many physiologically relevant molecules that are carried by albumin. For example, many organic anions have affinities of 1–100 µM and long chain fatty acids bind to albumin in the 100 nM range (Peters, 1996). The affinity of these molecules for albumin as well, most probably, plays a role in their clearance.
Building on our previous work (Dennis et al., 2002), we have now demonstrated the prolonged half-life and reduced clearance of two different Fab fragments through their association with albumin by way of an albumin-binding peptide. In contrast to the current study, the albumin-binding peptide sequence was fused to the carboxyl terminus of the light chain of an anti-tissue factor Fab; yet, similar pharmacokinetic parameters were observed (the K10 half-life was reported previously, whereas was calculated from the same data and used in this report). We now fully demonstrate that the enhanced pharmacokinetics of an AB.Fab is a direct function of its affinity for albumin. Further, by utilizing variants with varied affinities for albumin, we are able to estimate a potential elimination half-life for AB.Fab4D5-H in human.
Albumin binding is a common strategy for reducing the clearance of small molecule drugs. This information could be useful in the design of such drugs where, unlike the AB.Fab, the interplay between achieving a prolonged half-life as a result of albumin binding is balanced against a potential loss in function as albumin binding of the small molecule precludes it from binding to its intended target. Once establishing the concentration of free drug required for efficacy, the balance between this concentration and its potential half-life as a function of albumin binding might be estimated from these data.
By extending the half-life of a Fab, an AB.Fab may also have utility in tumor targeting where it has been observed that size and half-life of the targeting agent can have a dramatic affect on tumor delivery and retention (Hu et al., 1996; Wu and Yazaki, 2000). The ability to fine tune the pharmacokinetics of an AB.Fab could prove useful in identifying the ideal properties required for optimum delivery.
|
Footnotes |
|---|
Edited by Paul Carter
|
Acknowledgements |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionAcknowledgementsReferences |
|---|
We would like to thank Francisco Valles for running the fermentations, Mark Vasser, Parkash Jhurani and Peter Ng for oligonucleotide synthesis, Bridget Currell for DNA sequencing, Amy Oldendorp, Michelle Gonzales, Mike Reich, Dongwei Li and Shannon Stainton for assistance with all pharmacokinetic studies, and Janie Pena for her assistance in preparing this manuscript.
|
References |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionAcknowledgementsReferences |
|---|
Baselga J, et al. (1996) J. Clin. Oncol. 14:737–744.
Borsi L, et al. (2002) Int. J. Cancer 102:75–85.[CrossRef][Web of Science][Medline]
Carter P, et al. (1992a) Biotechnology 10:163–167.[CrossRef][Medline]
Carter P., Presta L.G., Gorman C.M., Ridgway J.B.B., Henner D., Wong W.L.T., Rowland A.M., Kotts C., Carver M.E., Shepard M.H. (1992b) Proc. Natl Acad. Sci. USA 89:4285.
Davies B. and Morris T. (1993) Pharm. Res. 10:1093–1095.[CrossRef][Web of Science][Medline]
Dennis M.S., Zhang M., Meng Y.G., Kadkhodayan M., Kirchhofer D., Combs D., Damico L.A., et al. (2002) J. Biol. Chem. 277:35035–35043.
Friguet B., Chaffotte A.F., Djavadi-Ohaniance L., Goldberg M.E. (1985) J. Immunol. Methods 77:305–319.[CrossRef][Web of Science][Medline]
Ghetie V. and Ward E.S. (2000) Annu. Rev. Immunol. 18:739–766.[CrossRef][Web of Science][Medline]
Graff C.P. and Wittrup K.D. (2003) Cancer Res. 63:1288–1296.
Hatton M.W.C., Richardson M., Winocour P.D. (1993) J. Theor. Biol. 161:481–490.[CrossRef][Medline]
Hu S., Shively L., Raubitschek A., Sherman M., Williams L.E., Wong J.Y., Shively J.E., Wu A.M. (1996) Cancer Res. 56:3055–3061.
Jain R.K. (1989) J. Natl. Cancer Inst. 81:570–576.
Jain R.K. (1999) Annu. Rev. Biomed. Eng. 1:241–263.[CrossRef][Web of Science][Medline]
Kashmiri S.V.S. (2001) J. Nucl. Med. 42:1528–1529.
Kratz F., Muller-Driver R., Hofmann I., Drevs J., Unger C. (2000) J. Med. Chem. 43:1253–1256.[CrossRef][Web of Science][Medline]
Markussen J, et al. (1996) Diabetologia 39:281–288.[Web of Science][Medline]
Marques J.A., George J.K., Smith I.J., Bhakta V., Sheffield W.P. (2001) Thromb. Haemost. 86:902–908.[Medline]
McKeage K. and Perry C.M. (2002) Drugs 62:209–243.[CrossRef][Web of Science][Medline]
Peters T. Jr. (1996) All About Albumin (Academic Press Inc., San Diego).
Reed R.G. and Peters T. Jr. (1984) Fed. Proc. 43:1858.
Rowland M. and Tozer T.N. (1995) Clinical Pharmacokinetics: Concepts and Applications (Williams & Wilkins, Media).
Scatchard G. (1947) Ann. N. Y. Acad. Sci. 51:660.[CrossRef]
Slamon D.J., et al. (1989) Science 244:707–712.
Slamon D.J., et al. (2001) N. Engl. J. Med. 344:783–792.
Smith B.J., et al. (2001) Bioconjug. Chem. 12:750–756.[CrossRef][Web of Science][Medline]
Stehle G., Sinn H., Wunder A., Schrenk H.H., Schutt S., Maier-Borst W., Heene D.L. (1997) Anticancer Drugs 8:677–685.[Medline]
Sterling K. (1957) J. Clin. Invest. 30:1228–1237.
Stevens D.K., Eyre R.J., Bull R.J. (1992) Fundam. Appl. Toxicol. 19:336–342.[CrossRef][Web of Science][Medline]
Sung C, et al. (2003) J. Interferon Cytokine Res. 23:25–36.[CrossRef][Medline]
Syed S., Schuyler P., Kulczycky M., Sheffield W.P. (1997) Blood 89:3243–3252.
Todorovska A., Roovers R.C., Dolezal O., Kortt A.A., Hoogenboom H.R., Hudson P.J. (2001) J. Immunol. Methods 248:47–66.[CrossRef][Web of Science][Medline]
Vogel C.L, et al. (2002) J. Clin. Oncol. 20:719–726.
Wong K., Cleland L.G., Poznanski M.J. (1980) Agent Action 10:231–239.[Medline]
Wu A.M. and Yazaki P.J. (2000) Q. J. Nucl. Med. 44:268–83.[Web of Science][Medline]
Yeh P., et al. (1992) Proc. Natl Acad. Sci. USA 89:1904–1908.
Received January 19, 2006; revised March 10, 2006; accepted March 14, 2006.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  A. Jonsson, J. Dogan, N. Herne, L. Abrahmsen, and P.-A. Nygren Engineering of a femtomolar affinity binding protein to human serum albumin Protein Eng. Des. Sel., August 1, 2008; 21(8): 515 - 527. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
L. J. Holt, A. Basran, K. Jones, J. Chorlton, L. S. Jespers, N. D. Brewis, and I. M. Tomlinson Anti-serum albumin domain antibodies for extending the half-lives of short lived drugs Protein Eng. Des. Sel., May 1, 2008; 21(5): 283 - 288. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. Schlapschy, I. Theobald, H. Mack, M. Schottelius, H.-J. Wester, and A. Skerra Fusion of a recombinant antibody fragment with a homo-amino-acid polymer: effects on biophysical properties and prolonged plasma half-life Protein Eng. Des. Sel., June 26, 2007; (2007) gzm020v1. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. Bertschinger, D. Grabulovski, and D. Neri Selection of single domain binding proteins by covalent DNA display Protein Eng. Des. Sel., February 1, 2007; 20(2): 57 - 68. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. S. Dennis, H. Jin, D. Dugger, R. Yang, L. McFarland, A. Ogasawara, S. Williams, M. J. Cole, S. Ross, and R. Schwall Imaging Tumors with an Albumin-Binding Fab, a Novel Tumor-Targeting Agent Cancer Res., January 1, 2007; 67(1): 254 - 261. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||