Biochem. J. (1995) 312, 725-731 (Printed in Great Britain)
`
`725
`
`Albumin binding of insulins acylated with fatty acids: characterization of the
`ligand-protein interaction and correlation between binding affinity and
`timing of the insulin effect in vivo
`Peter KURTZHALS*, Svend HAVELUND, lb JONASSEN, Benedicte KIEHR, Ulla D. LARSEN, Ulla RIBEL and Jan MARKUSSEN
`Novo Research Institute, Novo Nordisk A/S, Novo Alle, DK-2880 Bagsvaerd, Denmark
`
`Albumin is a multifunctional transport protein that binds a wide
`variety ofendogenous substances and drugs. Insulins with affinity
`for albumin were engineered by acylation of the e-amino group
`of LysB29 with saturated fatty acids containing 10-16 carbon
`atoms. The association constants for binding of the fatty acid
`acylated insulins to human albumin are in the order of 104-105
`M-1. The binding apparently involves both non-polar and ionic
`interactions with the protein. The acylated insulins bind at the
`long-chain fatty acid binding sites, but the binding affinity is
`lower than that of the free fatty acids and depends to a relatively
`small degree on the number of carbon atoms in the fatty acid
`
`chain. Differences in affinity of the acylated insulins for albumin
`are reflected in the relative timing of the blood-glucose-lowering
`effect after subcutaneous injection into rabbits. The acylated
`insulins provide a breakthrough in the search for soluble,
`prolonged-action insulin preparations for basal delivery of
`the hormone to the diabetic patient. We conclude that the bio-
`chemical concept of albumin binding can be applied to protract
`the effect ofinsulin, and suggest that derivatization with albumin-
`binding ligands could be generally applicable to prolong the
`action profile of peptide drugs.
`
`INTRODUCTION
`The therapeutic applicability of a biologically active peptide
`depends on the possibility of delivering it at its site of action with
`a suitable time-profile. Peptide and protein drug delivery is
`associated with several problems [1]. For example, peptides must
`in general be administered by injection because they are sus-
`ceptible to enzymic breakdown and penetrate poorly through
`mucosal membranes. Furthermore, most peptides have a short
`half-life within the circulation and must be gradually released
`into the bloodstream to have a sustained effect. Advanced
`controlled-delivery systems for peptides, such as pumps, lipo-
`somes and microspheres, have not proved successful [2].
`Insulin is a peptide hormone that has been in clinical use for
`decades. To meet the requirement for a constant basal supply of
`the hormone, diabetic patients receive daily subcutaneous injec-
`
`NH2I
`
`NH2'
`
`A-chain
`
`B-chain
`
`1
`
`ILySB29-x
`
`R
`
`Schematic representadlon of the fatty acid acylated insulins
`Figure 1
`R denotes the fatty acid attached by an amide bond to the e-amino group of LySB29. In the
`present study we have prepared derivatives of des-(B30) human insulin (X is deleted) in which
`R has from 10 to 16 carbon atoms, and analogues of human insulin (X is Thr) in which R has
`10 or 14 carbon atoms.
`
`tions of long-acting insulin suspensions [3]. In recent years, much
`effort has been devoted to the development of soluble, long-
`acting insulin analogues with a more reproducible and a more
`prolonged effect than the insulin suspensions [4-6], but no
`analogues have shown improved clinical results.
`Albumin is a multifunctional transport protein that binds
`reversibly a wide variety of endogenous substances and drugs
`[7-10]. Owing to the restricted passage of albumin-drug com-
`plexes across membranes, the pharmacokinetic parameters of
`many drugs can be altered by modification of their affinity for
`albumin [1 1,12]. To test whether albumin binding can be applied
`to protract the effect of peptide drugs, we have engineered insulin
`derivatives with affinity for albumin by acylation of the hormone
`with fatty acids (Figure 1). In the present study we investigate the
`interaction between the fatty acid acylated insulins and human
`serum albumin (HSA), using albumin immobilized on agarose.
`Differences in affinity of acylated insulins for albumin are
`correlated to the relative degree of protraction of the insulins in
`rabbits. The suitability of using albumin binding to protract
`insulin action is discussed.
`
`EXPERIMENTAL
`Materials
`Fatty acid-free (< 0.005 %) and globulin-free HSA purchased
`from Sigma were used throughout this study. Fatty acids of
`analytical grade were from Fluka AG (Switzerland) or Aldrich
`(Germany). Divinylsulphone-activated Sepharose 6B (Mini-Leak
`Low) was obtained from Kem-En-Tec A/S (Copenhagen,
`Denmark). Insulin Protaphane (NPH-insulin) and human insulin
`were supplied by Novo Nordisk A/S (Denmark). Des-(B30)
`human insulin was prepared as previously described [13]. [9,10-
`3H]Myristic acid (33.5 Ci/mmol) was supplied as an ethanolic
`
`Abbreviations used: Boc, t-butyloxycarbonyl; HSA, human serum albumin; NPH-insulin, Neutral Protamine Hagedorn insulin (Insulin Protaphane);
`Tris/TX-100, 0.1 M Tris/HCI, pH 7.4, containing 0.025% (v/v) Triton X-100; RA, relative receptor affinity.
`* To whom correspondence should be addressed.
`
`MPI EXHIBIT 1035 PAGE 1
`
`MPI EXHIBIT 1035 PAGE 1
`
`

`

`726
`
`P. Kurtzhals and others
`
`solution (1 mCi/ml) from DuPont NEN. All other chemicals
`used were of reagent grade or better.
`Protein concentrations were determined by UV absorbance,
`taking molar absorption coefficients of e279 35.7 x 10' M- -cm-'
`for HSA [7] and 6276 6.2 x 10' M-l cm-' for the insulins.
`
`Preparation of fatty acid acylated Insulins
`Human insulin and des-(B30) human insulin were treated with
`di-t-butyldicarbonate in DMSO/triethylamine (20: 1, v/v), and
`the GlyAl, PheBI di-Boc (where Boc represents t-butyloxy-
`carbonyl) insulins were separated from the isomers by reversed-
`phase HPLC. The hydroxysuccinimide esters of fatty acids were
`prepared from fatty acids and N-hydroxysuccinimide by using
`dicyclohexylcarbodiimide in dimethylformamide for ester for-
`mation, and ethanol for recrystallization of the products. The C-
`amino group of LysB29 was selectively acylated by treatment of
`GlyAl, PheBI di-Boc insulin with the fatty acid hydroxy-
`succinimide esters in dimethylformamide/DMSO (1:7, v/v) at
`15 °C, using 20 equivalents of a tertiary amine such as 4-
`methylmorpholine. The protecting groups were removed with
`trifluoroacetic acid and the LysB29-acylated insulins were purified
`by reversed-phase HPLC. The identity of the fatty acid acylated
`insulins was confirmed by mass spectrometry, determining the
`molecular mass of the insulin derivative, of the B-chain after
`treatment with dithiothreitol, and of the B-chain fragment
`containing LysB29 obtained by treatment with Staphylococcus
`aureus protease.
`Mono l25I-(TyrAl4)-labelled insulins were prepared as pre-
`viously described [14].
`
`Immobilization of HSA
`HSA was coupled to the activated agarose matrix (Mini-Leak
`Low) according to the guidelines of the manufacturer. For each
`gram of gel was added 2 ml of 5 % (w/v) HSA and 2 ml of 30 %
`(w/v) PEG 20000 in 0.3 M NaHCO3, pH 8.6. The suspension
`was gently agitated overnight at 23 'C. Excess active groups were
`blocked by treatment with 30 mM ethanolamine at pH 9.0 for
`5 h at 23 'C. The molar content of HSA in the gel was
`0.4-0.6 nmol per mg suction-dried gel as determined by titration
`with HSA in solution as described below. At pH 8.6, amino
`groups and thiol groups can be coupled to the activated matrix,
`whereas hydroxyl groups react at pH > 10 [15]. Because HSA
`contains 59 lysine residues and a single free cysteine [7], the
`coupling is expected to involve primarily c-amino groups of
`lysine residues.
`
`Equilibration of ligands with immobilized HSA
`Immobilized HSA was washed with 2-3 volumes of 0.1 M Tris,
`pH 7.4, on a suction filter and drained until cracks were seen in
`the gel. A portion of the gel was weighed out and suspended in
`0.1 M Tris, pH 7.4, containing 0.025 % (v/v) Triton X-100
`(Tris/TX-100). The concentration of immobilized HSA in the
`suspension was less than 40 mg/ml. The exclusion volume
`provided by the Mini-Leak matrix (less than 4% of the total
`volume) is considered to be insignificant. Portions of the sus-
`pension were pipetted into vials during stirring and combined
`with the radiolabelled ligand and Tris/TX-100 to give a final
`volume of 1.00 ml. When binding was examined in the presence
`of fatty acid or albumin in solution, these components were
`included in the mixture before Tris/TX-100 was added to the
`final volume of 1.00 ml. The concentration of radioactivity in the
`final mixture was 0.01-0.05 ,aCi/ml. The vials were rotated
`at 30 rev./min for 2 h at room temperature and centrifuged for
`
`5 min at 1800 g. Incubation for 2 h was found to be sufficient to
`reach equilibrium.
`When 12"I-labelled ligands were used, the total radioactivity in
`each vial (I) and the radioactivity in 500,u of supernatant (21)
`were counted on a Packard Cobra Auto-Gamma instrument
`(Packard, Meriden, U.S.A.). The bound radioactivity (B) was
`determined as B = T-F.
`When 3H-labelled myristic acid was used, 500 ,1 of supernatant
`was transferred to a scintillation counting vial, 10 ml Ultima
`Gold (Packard) was added, and the radioactivity (12F) was counted
`on a Packard Tri-Carb liquid scintillation analyser. In this case,
`the total radioactivity in each vial was estimated from a measure-
`ment of the radioactivity (27) in 500 #1 of a mixture treated as
`described above but in the absence of immobilized albumin.
`No unspecific binding was seen when radiolabelled acylated
`insulins were incubated as described above with Mini-Leak Low
`blocked with ethanolamine. Triton X-100 was included in the
`buffer to prevent non-specific binding of the acylated insulins to
`vials and pipettes. By varying the concentration of the detergent
`between 0 and 0.025% (v/v) it was found that 0.025 % (v/v)
`Triton X-100 prevented non-specific binding without affecting
`the binding to albumin.
`
`Titration of Immobilized HSA for binding of fatty acids
`A weighed amount (about 25 mg/ml) of immobilized HSA
`(HSAimm) was equilibrated with 1.5 nM [9, 10-3H]myristic acid
`and varying concentrations (1-40 ,sM) of HSA in solution
`(HSAfree). As HSAfree is present in large molar excess over [9,10-
`3H]myristic acid, a plot of F/B against [HSAfree] is linear.
`Assuming that the binding constants to HSAImm and HSAfree are
`equal, the slope of the line is 1/[HSAimm]. The molar content of
`HSA per mg of suction-dried gel is calculated from the amount
`of gel in the suspension (about 25 mg/ml) and [HSAimm]. In one
`case, the albumin content of a hydrolysed sample of Mini-Leak
`HSA was determined by amino acid analysis. The amino acid
`analysis showed a 20 % higher albumin content than found by
`titration, indicating that the binding properties of HSA are not
`completely unaffected by immobilization. It has previously been
`shown that a similar coupling of albumin to agarose does not
`alter the primary binding sites for most ligands, including fatty
`acids [16].
`
`Receptor affinity
`Relative affinities of l25I-(TyrAl4)-labelled insulins for the soluble
`insulin receptor were determined by a modification of a previously
`described assay [17]. In brief, the soluble insulin receptor was
`immobilized on Mini-Leak to a concentration of ,uM in the gel.
`Various concentrations (0-20 nM) of immobilized receptor were
`equilibrated with 10 pM of the radiolabelled insulin in a binding
`buffer (pH 7.8) containing 0.1 M Hepes, 0.1 M NaCl, 0.01 M
`MgCl2 and 0.025 % Triton X-100 for 2 h at 23 'C. Bound tracer
`was isolated by centrifugation and the relative receptor affinities
`(RA) were calculated from a plot of bound insulin against
`receptor concentration as described elsewhere [17]. Human
`insulin represents the RA of 100%.
`
`Animal experiments
`Studies in rabbits were performed on non-diabetic, fasted, male
`New Zealand White rabbits, 0.5-3 years of age and weighing
`2.5-3.5 kg, receiving subcutaneous injections ofeither an acylated
`insulin or NPH-insulin. At least 1 h before dosing, the rabbits
`were fixed in pillories. Acylated insulins were given as aqueous
`solutions containing 600 nM (100 units/ml) of insulin. The dose
`
`MPI EXHIBIT 1035 PAGE 2
`
`MPI EXHIBIT 1035 PAGE 2
`
`

`

`Albumin binding of fatty acid acylated insulins
`
`727
`
`8,
`U-
`
`0co
`
`[HSAimm. (pM)
`
`20
`
`Figure 3
`Plots of bound/tree insulin concentration ratios against HSAJ",,
`concentration
`l2Sl-(TyrAl4)-labelled insulins (2.5 nM) were equilibrated with 1-20 ,uM immobilized HSA at
`AL, LysB29-
`23 °C. CJ, Lys829-decanoyl insulin; *, Lys829-decanoyl, des-(B30) insulin;
`dodecanoyl, des-(B30) insulin; 0, LysB29-tetradecanoyl insulin; 0, Lys829-tetradecanoyl, des-
`(B30) insulin.
`
`1
`
`high affinity. The capacity of albumin for binding the insulins
`apparently exceeds 5 mol/mol, without any indication of satu-
`ration within the tested concentration range. However, the
`interpretation of binding data obtained at insulin concentrations
`above 10 ,uM is difficult owing to extensive insulin self-association
`and to the relatively low solubility of the acylated insulin
`analogues in the applied buffer.
`To quantify and compare the affinities of the fatty acid
`acylated insulins for HSA, '251-labelled insulins were equilibrated
`at 2.5 nM with various concentrations (1-20 ,uM) of immobilized
`HSA. At these experimental conditions less than 0.25 % of the
`albumin is occupied by an insulin molecule, and the first
`association constant, K., for binding of insulin to albumin can be
`estimated by:
`B
`F [HSAimm]
`where B/F is the ratio between bound and free 125I-labelled
`insulin and [HSAimm] is the total concentration of immobilized
`albumin [19]. Plots of B/F against [HSAimm] are linear with
`slopes that estimate K,. Representative plots are shown in Figure
`3. The Ka for binding LysB29_tetradecanoyl, des-(B30) insulin to
`HSA is determined as 2.4 x 105 M-1 by this approach. This value
`seems to be in good agreement with the estimate of K. obtained
`by extrapolation of the Scatchard graph (Figure 2, inset) to the
`intercept with the ordinate axis [20,21]. LysB29-acylated des-(B30)
`insulins have higher affinities for albumin than their full B-chain
`counterparts. The increase in - AG for binding obtained by
`deletion of ThrB30 is about 0.35 kcal/mol for the decanoyl and
`tetradecanoyl derivatives (Figure 4). The location of the C-
`terminal carboxylate group closer to the lipophilic side chain
`may favour formation of an ionic bond to a basic residue at the
`binding site. K. for binding the LysB29 acylated des-(B30)-insulins
`to HSA rises from 0.28 x 105 M-1 to 2.4 x 105 M-1 when the
`
`was 12 nmol of insulin per animal. Blood samples were drawn
`before and 1, 2, 4 and 6 h after injection. Glucose analysis was
`performed by the hexokinase method [18].
`Euglycaemic glucose clamps were carried out in non-diabetic,
`conscious, female pigs, cross-bred from Danish Landrace,
`Yorkshire and Durok, 4-5 months of age and weighing 70-95 kg.
`Before the experiments the pigs were fasted overnight, for 18 h.
`Two catheters were inserted in the jugular veins, one for glucose
`infusion and one for blood sampling. The pigs were free to move
`in their pens during the clamp period. Five pigs received NPH-
`insulin and LysBa9-tetradecanoyl, des-(B30) insulin in random
`order with an interval of 10 days. The dose was 216 nmol of
`insulin. The pigs were kept euglycaemic at their individual
`fasting glucose levels by infusion of a glucose solution (270 g/l)
`at a variable rate. Depending on changes in plasma glucose
`concentration obtained during frequent plasma glucose moni-
`toring, the necessary adjustments of the glucose infusion were
`made empirically. Blood samples were collected in heparinized
`glass tubes every 15 min, plasma was separated, and glucose was
`determined within 1.5 min of blood sampling with a YSI (Yellow
`Springs Instruments) glucose analyser (glucose oxidase method).
`
`RESULTS
`Binding of acylated Insulins to immobilized HSA
`The interaction between albumin and insulins acylated with fatty
`acids at the e-amino group of LysB29 was studied using albumin
`immobilized on agarose. This method was chosen because the
`size of the ligands precludes performing binding studies by
`dialysis. Typical plots for binding of the insulins to immobilized
`HSA are shown in Figure 2. Scatchard plots are not linear, as
`exemplified in the inset to Figure 2, indicating that binding
`occurs at more than one class of sites. The initial part of the
`Scatchard plot is consistent with binding of at least 1 mol of
`LysB29-tetradecanoyl, des-(B30) insulin per mol of HSA with
`
`0.3
`
`E
`
`0.2
`
`5
`
`4
`
`0.1 ~
`
`~
`
`*A
`
`0~~~~~~~
`0
`2 46 8
`[Bound inslSinfl/[HSAimm
`
`1/ .
`
`7
`
`6
`
`2
`
`E E
`
`CD)
`
`-:
`
`0
`
`0
`io-9
`
`io-8
`
`10-7
`
`10-6
`[Freel(M)
`
`i0-5
`
`10-4
`
`10-3
`
`Binding of fatty acid acylated insulins to immobilized HSA
`Figure 2
`Various concentrations (0-150,M) of 1251_(TyrAl4)-labelled insulins were equilibrated with
`5 1sM of immobilized HSA at 23 °C. *, Lys829-decanoyl, des-(B30) insulin; A, LysB29-
`dodecanoyl, des-(B30) insulin; 0, Lys829-tetradecanoyl, des-(B30) insulin. Inset: Scatchard
`plot of the data for LysB29-tetradecanoyl, des-(B30) insulin.
`
`MPI EXHIBIT 1035 PAGE 3
`
`MPI EXHIBIT 1035 PAGE 3
`
`

`

`728
`
`P. Kurtzhals and others
`
`0~~~~~~~~~~~~~~~~~~~
`
`0
`
`13
`
`= 12
`
`12
`14
`Number of carbon atoms
`
`16
`
`103/T(K-')
`
`0
`
`0 1
`
`0
`
`13
`
`12
`
`C 11
`
`10
`
`9
`
`Figure 4
`Correlation between K for binding of fatty acid acylated Insulins
`to Immobilized HSA and the numLer of carbon atoms In the fatty acid side
`chain
`Ka was estimated as the slope of linear plots of B/Fagainst [HSAImmI as shown in Figure 3.
`0, Lys829-acylated, des-(B30) insulins; 0, LysB"9-acylated insulins with full B-chain. The
`points shown are the means of at least two determinations.
`
`Van't Hoff plot for the binding of LysB29-tetradecanoyl, des-(B30)
`Figure 6
`insulin to immobilized HSA
`5ll-(TyrAl4)-labelled LysB29-tetradecanoyl, des-(B30) insulin (2.5 nM) was equilibrated with
`1-10,uM immobilized HSA at various temperatures (4-37 °C). The buffer was 5 mM Tris,
`pH 7.4, containing 100 mM NaCI and 0.025% (v/v) Triton X-100. Ka was estimated as the slope
`of linear plots of B/Fagainst [HSAimm] as shown in Figure 3. The points shown are the means
`of four determinations.
`
`100
`
`X 80
`
`_
`~0
`
`60
`
`20
`
`20
`
`0.1
`10
`1
`[Fatty acidJ/[HSAimmJ
`
`100
`
`Displacement of Lysln-tetradecanoyl, des-(B30) Insulin from
`Figure 5
`immobilized HSA with lauric acid
`51-(TyrAl4)-labelled Lys?9-tetradecanoyl, des-(B30) insulin (2.5 nM) was equilibrated with
`10.1 ,uM Mini-Leak HSA in the presence of 0.6 ,uM to 2.5 mM lauric acid at 23 IC.
`
`number of carbon atoms in the acyl chain is increased from 10 to
`14, reflecting that non-polar interactions contribute to the
`binding. The gain in -AG for binding obtained by extension of
`the fatty acid chain with one methylene group is 0.2-0.5 kcal/mol
`(Figure 4), which is in the region of one-half that reported for the
`long-chain free fatty acids [22]. Extension of the chain length to
`16 carbon atoms does not increase the binding affinity any
`further. We note that human insulin per se does not bind with
`detectable affinity to immobilized HSA (Ka < 103 M-1).
`
`Competition with fatty acids
`acids on the binding of LysB29_
`The influence of fatty
`tetradecanoyl, des-(B30) insulin to HSA is shown in Figure 5. A
`significant displacement of the insulin occurs in the presence of
`more than 1 mole of lauric acid per mol of HSA. Thus the trace
`amount of radiolabelled LysB29-tetradecanoyl, des-(B30) insulin
`binds with lower affinity when the first binding site for fatty acid
`
`is occupied. A 50 % displacement is obtained at a fatty acid to
`HSAimm concentration ratio of about 4, and the insulin derivative
`is quantitatively displaced as this ratio approaches 100. The
`results suggest that the fatty acid acylated insulin competes with
`long-chain fatty acids for binding at common sites. Alternatively
`the displacement of the acylated insulins by laurate might be due
`to a conformational change.
`
`Temperature dependence and thermodynamic parameters
`The temperature dependence of the association constant for
`binding of LysB29-tetradecanoyl, des-(B30) insulin to HSA is
`shown in Figure 6. Ka decreases with increasing temperature in
`the range 4-37 °C, in agreement with previous results for the
`interaction between HSA and fatty acids [21]. The Ka for binding
`of this insulin analogue to HSA at 37 °C is 1 x 105 M-1. Hence at
`an HSA concentration of 0.6 mM, corresponding to the albumin
`level in human plasma, the bound fraction is 98.4 %. The almost
`linear correlation between 1/ T and lnKa allows us to estimate the
`enthalpy change that accompanies binding from the van't Hoff
`equation, InKa = AH '/RT+ constant [23]. The change in free
`energy and entropy at 298 K can be obtained from the equations
`AG0 = - RlnnKa and AG ° = AH '- TAS °. The estimated
`values for AG 0, AH0 and TAS 0 are -29.5, - 19.4 and
`10.1 kJ/mol, respectively, indicating that the association is driven
`by favourable changes in both enthalpy and entropy.
`
`Effect profiles in vivo
`The effect profiles of the fatty acid acylated insulins after
`subcutaneous injection into rabbits are shown in Figure 7(a).
`NPH-insulin, which is the most generally used long-acting insulin
`suspension in the treatment of diabetes, was given as a reference.
`The fall in blood glucose within the first hour after injection
`shows that the acylated insulins have a significantly slower onset
`of action than NPH-insulin. Furthermore, the time until maximal
`effect is generally longer for the acylated insulins than for NPH-
`insulin. However, this time cannot be determined for all insulins,
`because the blood glucose was not followed beyond 6 h. As
`shown in Figure 7b, the initial fall in blood glucose depends on
`
`MPI EXHIBIT 1035 PAGE 4
`
`MPI EXHIBIT 1035 PAGE 4
`
`

`

`Albumin binding of fatty acid acylated insulins
`
`729
`
`8
`
`7 6
`
`%,,4Th~
`
`11
`
`2
`
`y{9kbN1
`
`8 7
`
`E6
`(D5
`on4
`
`D30 2
`
`0 6
`
`0
`0
`
`4
`
`8
`
`12
`
`16
`
`20
`
`24
`
`0
`
`4
`
`8
`
`12
`
`16
`
`20
`
`24
`
`0
`^ 5
`t. =
`c E 40
`
`L.
`
`en
`
`4- m 30
`'A E 20
`3-10
`0
`
`Figure 8
`pigs
`
`u
`n
`
`4
`A
`
`a
`
`uz
`-o
`ir,
`19
`Time (h)
`
`zu
`in
`
`Z4
`IA
`A
`
`u
`n
`
`4
`A
`
`*1
`in
`
`zu
`<X
`
`Z4
`IA
`
`a
`
`119
`Time (h)
`
`Euglycaemic glucose clamp after subcutaneous injection Into
`
`Blood glucose levels (means + S.D.; top) and glucose infusion rates (means ± S.E.M., n = 5;
`bottom)
`during euglycaemic glucose clamp after subcutaneous injection
`of 216 nmol
`NPH-insulin (left) and Lys829-tetradecanoyl, des-(B30) insulin (right) into pigs.
`
`for at least 18 h. However, LysB29-tetradecanoyl, des-(B30)
`insulin showed a more protracted glucose utilization profile than
`NPH-insulin, the times to peak effect being 6.4 + 2.2 h and
`3.4 + 0.2 h, respectively.
`
`DISCUSSION
`Derivatization of the insulin molecule with albumin-binding
`ligands provides a new approach to protracting the blood-
`glucose-lowering effect after subcutaneous injection of the hor-
`mone. Albumin is the most abundant protein in the extracellular
`fluid. The concentration of HSA in human plasma is about
`0.6 mM [25], whereas the level in subcutaneous interstitial fluid
`is about 60 % of that in plasma [12]. The protein binds a wide
`variety of endogenous substances and drugs with binding
`constants that are typically in the order of 104_106 M-1 for
`organic anions, and about 108 M-1 for long-chain fatty acids
`[7-10]. We have obtained insulins with affinities of 104-105 M-1
`for HSA by acylation of the e-amino group of LysB29 with
`saturated fatty acids containing 10 to 16 carbon atoms.
`
`Binding mechanism
`The binding of aromatic anions and fatty acids to albumin
`occurs by a combination of hydrophobic and ionic interactions
`[26-30]. In accord with this general mechanism, the C-terminal
`carboxylate group and the acyl side chain of the fatty acid
`acylated insulins seem to take part in the interaction with HSA.
`Thus the interaction between albumin and fatty acid acylated
`des-(B30) insulins is believed to involve the fatty acid carbon
`chain and the side chain (1 nitrogen plus 4 carbon atoms), the a-
`carbon and the carboxylate group of LysB29. The binding site
`apparently cannot adapt sufficiently for further non-polar inter-
`actions with the ligand when the number of carbon atoms in the
`
`010
`
`2.0
`
`0 0
`
`0
`
`1
`
`2
`
`3
`Time (h)
`
`4
`
`5
`
`6
`
`2.0
`
`(b)
`
`2.
`
`(U
`
`0
`
`1.5
`
`1.0
`
`0.5
`
`0
`
`0.2
`
`0.4
`
`T
`
`0.6
`Relative K.
`
`0.8
`
`1.0
`
`1.2
`
`Blood-glucose-lowering effect of fatty acid acylated insulins after
`Figure 7
`subcutaneous Injection Into rabbits
`(a) O, LysB29-decanoyl insulin (n = 24); *, LysB29-decanoyl, des-(B30) insulin (n = 36);
`Al LysB29-undecanoyl, des-(B30) insulin (n = 12); A, LysB29-dodecanoyl, des-(B30) insulin
`(n = 18); 0, LysB29-tridecanoyl, des-(B30) insulin (n = 4); 0, LysB29-tetradecanoyl, des-
`(B30) insulin (n = 6). The heavy line shows the effect of NPH-insulin (n = 34). Changes in
`blood glucose are given as means + S.E.M. (b) Correlation between the relative affinities of the
`acylated insulins for HSA and the decrease in blood glucose 1 h after injection. LysB29-
`tetradecanoyl, des-(B30) insulin represents a relative affinity of 1.0.
`
`the affinity of the acylated insulin for albumin, suggesting that
`the protracted effect of the acylated insulins is due to albumin
`binding.
`The receptor affinity relative to human insulin for LysB29_
`decanoyl insulin; LysB29-decanoyl, des-(B30) insulin; LysB29_
`dodecanoyl, des-(B30) insulin; and LysB29-tetradecanoyl, des-
`(B30) insulin was found to be 76%, 76%, 54% and 46%
`respectively; the RAs for LysB29-undecanoyl, des-(B30) insulin
`and LysB29-tridecanoyl, des-(B30) insulin were not determined.
`It has previously been shown that insulin analogues with
`20-300o% receptor affinity have the same biological potency in
`vivo [24]. To preclude the possibility that the diminished action
`on blood glucose is related to a decreased bioefficacy of the
`acylated insulins and to study the effect profile over a longer time
`course, we performed a 24 h euglycaemic glucose clamp after
`subcutaneous injection of 216 nmol LysB29-tetradecanoyl, des-
`(B30) insulin and NPH-insulin into pigs. The resulting profiles
`are shown in Figure 8. The cumulative glucose infusion (0-24 h)
`was the same after injection of LysB29-tetradecanoyl, des-(B30)
`insulin (2.5 + 1.3 mol) and NPH-insulin (2.6 + 0.6 mol), indi-
`cating that the two insulin preparations are equipotent in vivo.
`Both preparations resulted in a significant glucose consumption
`
`MPI EXHIBIT 1035 PAGE 5
`
`MPI EXHIBIT 1035 PAGE 5
`
`

`

`730
`
`P. Kurtzhals and others
`
`fatty acid at LysBa9 increases beyond 14. Alternatively, ac-
`commodation of the C16 chain in the binding cavity might
`impede the favourable ionic interaction at the binding site. The
`Ka for binding of fatty acids to HSA similarly tends towards an
`upper limit, as an increase in carbon chain length from 16 to 18
`carbon atoms results in a relatively small increase in Ka [22].
`Studies with fatty acids containing a higher number of carbon
`atoms were not reported. The thermodynamic data for binding
`of LysB29-tetradecanoyl, des-(B30) insulin to HSA are compatible
`with the involvement of both ionic and hydrophobic interactions
`in binding of the ligand [21]. However, conformational changes
`in the protein can be expected to accompany ligand binding [10],
`and the interpretation of entropy and enthalpy changes is
`therefore speculative.
`
`Binding site
`The binding of a large number of aromatic anions to HSA seems
`to take place in two distinct binding pockets located in sub-
`domains IIA and IIIA of the protein, respectively [26]. The
`binding sites for fatty acids are less well defined, but it seems that
`there are three primary binding sites for long-chain fatty acids,
`possibly located in subdomains IB, IIIA and IIIB [10,31],
`respectively, whereas a large number of fatty acid anions are
`bound with lower affinity [22,32]. The displacement of LysB29_
`tetradecanoyl, des-(B30) insulin from albumin after addition of
`one or more lauric acid equivalents to HSA provides strong
`evidence that the high-affinity binding of fatty acid acylated
`insulins occurs at a primary long-chain fatty acid binding site.
`Alternatively, binding of the acylated insulins might occur at a
`secondary fatty acid site, in which case the influence on insulin-
`binding of the first lauric acid equivalents could be ascribed to a
`fatty acid-induced conformational change in these binding sites.
`However, the first two long-chain fatty acids are generally
`believed to have little effect on binding at other sites of the
`albumin molecule [8,27,33,34].
`The association constants for binding of the acylated insulins
`at the fatty acid binding sites of HSA are 1-3 orders of magnitude
`smaller than the first binding constants for the attached fatty
`acids themselves [22]. Furthermore, the Ka for binding of the
`fatty acid acylated insulins depends to a relatively small degree
`on the number of carbon atoms in the fatty acid side chain. Thus
`the changes in ligand structure caused by attachment of the free
`fatty acids to insulin have a significant influence on the binding
`properties. First, the binding of acylated insulins to albumin may
`be sterically hindered. Secondly, the presence of an amide
`function within the ligand may be unfavourable for binding
`owing to the polarity and diverging geometry of the amide bond
`relative to a carbon-carbon bond.
`
`Protraction of insulin action by albumin binding
`Insulin is secreted into the blood from the pancreatic beta-cells at
`a low basal rate in the fasting state and at a higher rate in
`response to the postprandial increase in the blood glucose level.
`To mimic the normal pattern of insulin release, a combination of
`rapid-acting and long-acting insulin preparations is used in the
`most intensive treatment of diabetes [3]. At present, a neutral
`solution of human insulin provides the rapid-acting component,
`whereas a prolonged action is obtained by injection of insulin
`suspensions of crystals with protamine or zinc. It has recently
`been convincingly shown that the onset of diabetic late compli-
`cations can be delayed by intensive treatment with insulin [35].
`To permit a tighter control of the blood glucose, it is desirable to
`develop rapid-acting insulins with a briefer effect than human
`
`insulin [36] and long-acting insulins with a smoother and more
`reproducible effect profile than the insulin suspensions [37]. The
`acylated insulins provide a breakthrough in the search for soluble,
`prolonged-action insulins for basal delivery of the hormone to
`diabetic patients. The mechanism of protraction is probably
`binding to albumin in the subcutaneous tissue, resulting in a
`lower absorption rate of the acylated insulins than of human
`insulin. Binding to albumin in plasma may increase the plasma
`half-life of the acylated insulins relative to that of human insulin
`and contribute to prolong the action profiles. The selection of
`a specific insulin derivative for drug development based on
`studies in a pig model will be discussed elsewhere.
`As the acylated insulins compete with fatty acids for binding
`to human albumin, the plasma level of free fatty acids may
`influence the effect profile of the acylated insulins. In normal
`subjects, the molar ratio of free fatty acids to albumin varies
`between 0.5 and 1.0, depending on the nutritional state [38].
`Provided that albumin is present in large excess over insulin,
`which is the case in vivo, the binding of acylated insulins is only
`affected to a minor degree by fluctuations in the free fatty acid
`level within this range. A transient high level of free fatty acids
`is expected to have a minor effect on the rate by which insulin is
`released into the bloodstream, because the absorption of un-
`bound insulin occurs relatively slowly, i.e. with a half-time of
`1-2 h [36].
`Derivatization with albumin-binding ligands as described here
`may provide a general approach to prolonging the effect of
`peptide drugs. Owing to the linkage between binding affinity and
`degree of protraction, the action profile can be engineered by
`modification of the ligand structure. The knowledge of the crystal
`structure of human serum albumin and the possibility of per-
`forming X-ray analysis of albumin bound with ligands such as
`fatty acids [10,26] are likely to prove useful in the future design
`of peptide derivatives with an affinity for specific albumin-
`binding sites.
`
`We thank Yvonne B. Madsen, Lene G. Andersen, Birgit D. Spon, Kate Muggler, Lene
`Drube, Lone S0rensen, Mette Dali and Erna Willert for excellent technical assistance;
`Dr. Per F. Nielsen for mass spectrometric analysis; and Dr. Asser S. Andersen and
`Dr. Thomas Kjeldsen for comments on the manuscript.
`
`2
`
`3
`
`4
`
`5
`6
`7
`8
`
`9
`10
`11
`
`12
`
`REFERENCES
`Lee, V. H. L. (1990) in Peptide and Protein Drug Delivery (Lee, V. H. L., ed.), pp.
`1
`1-56, Marcel Dekker, New York
`Banerjee, P. S., Hosny, E. A. and Robinson, J. R. (1990) in Peptide and Protein Drug
`Delivery (Lee, V. H. L., ed.),