Development of a process to manufacture PEGylated orally bioavailable insulin

Development of a Process to Manufacture PEGylated Orally Bioavailable Insulin

Partha Hazra, Laxmi Adhikary, Nitesh Dave, Anand Khedkar, H. S. Manjunath,Ramya Anantharaman, and Harish IyerResearch and Development, Biocon Limited, Electronic City, Bangalore 560100, India

DOI 10.1002/btpr.487Published online October 14, 2010 in Wiley Online Library (wileyonlinelibrary.com).

To make insulin orally bioavailable, insulin was modified by covalent attachment (conju-gation) of a short-chain methoxy polyethylene glycol (mPEG) derivative to the e-aminogroup of a specific amino acid residue (LysB29). During the conjugation process, activatedPEG can react with any of the free amino groups, the N-terminal of the B chain (PheB1),the N-terminal of the A chain (GlyA1), and the e-amino group of amino acid (LysB29), result-ing in a heterogeneous mixture of conjugated products. The abundance of the desired prod-uct (Methoxy-PEG3-propionyl—insulin at LysB29:IN-105) in the conjugation reaction can becontrolled by changing the conjugation reaction conditions. Reaction conditions were opti-mized for maximal yield by varying the proportions of protein to mPEG molecule at variousvalues of pH and different salt and solvent concentrations. The desired conjugated molecule(IN-105) was purified to homogeneity using RP-HPLC. The purified product, IN-105, wascrystallized and lyophilized into powder form. The purified product was characterized usingmultiple analytical methods including ESI-TOF and peptide mapping to verify its chemicalstructure. In this work, we report the process development of new modified insulin preparedby covalent conjugation of short chain mPEG to the insulin molecule. The attachment ofPEG to insulin resulted in a conjugated insulin derivative that was biologically active, orallybioavailable and that showed a dose-dependent glucose lowering effect in Type 2 diabetespatients. VVC 2010 American Institute of Chemical Engineers Biotechnol. Prog., 26: 1695–1704, 2010Keywords: PEGylation, conjugation, conjugated insulin, IN-105, crystallization

Introduction

Covalent modification of biological molecules with poly-ethylene glycol (PEGylation) is an established technologythat has been studied extensively.1–4 In a typical amino spe-cific PEGylation reaction, an activated monofunctional PEGmolecule is reacted with one or more free epsilon aminogroups of lysine or the N-terminal amino group. Reports ofPEGylation at other nucleophilic sites such as cysteine, histi-dine, arginine, or tyrosine are also available.5–8 Several PEGconjugated therapeutic proteins have been shown to exhibitclinical superiority compared with the unmodified moleculein terms of increasing circulating half life in vivo, reducingimmunogenicity, increasing solubility, and stability.9–11 Apotential aspect of PEGylation that is being explored in thepharmaceutical industry is the protection of protein drugs bycovalent attachment of PEG molecule to suitable sites of theprotein. This protein–PEG complex can generate a proteinmolecule suitable for oral administration.12–14 Human insulinand its analogues that are used therapeutically contain threepotential amino acid residues containing free primary aminogroups. All three primary amino groups, namely the N-termi-nal (alpha amino groups) of the A and B chains (GlyA1 andPheB1) and the e-amino group of LysB29, may be modified

by conjugation with oligomer, producing a mixture of conju-gated product called PEG insulin mixture.11,15,16 When aparticular conjugate is desired, for example insulin monocon-jugated at LysB29, it can be a considerable challenge to sep-arate it from a complex mixture of conjugates to therequired degree of purity, and the process can prove to beexpensive. Low yields from PEGylation add significantly tothe downstream processing costs in commercial manufactur-ing, and may sometimes offset the potential benefits ofPEGylation.17 Thus, an optimum condition for conjugation isrequired to maximize the productivity of the most desiredproduct at the end of the conjugation reaction and will alsohelp to increase the step yield as well as to reduce the costand effort expended in successive purification steps.

Here, we have reported a systematic approach for themanufacture of the drug substance, and for its scale up pro-duction. Effort was directed toward improving the efficiencyof covalent attachment of activated polyethylene glycol(PEG) to the specific site of the target protein molecule. Wehave described the preparation of an orally bioavailable insu-lin derivative, focusing on optimization of the conjugationreaction generating monoconjugate at LysB29 in predomi-nance, relative to other conjugated products (Figure 1). Theoligomer moiety used for the studies is N-hydroxysuccini-mide activated methoxy-triethylene glycol propionyl moiety,which has a molecular mass of 333 Da. Specifically, thereport relates to a greatly simplified, cost effective, and

Correspondence concerning this article should be addressed toP. Hazra at [email protected].

VVC 2010 American Institute of Chemical Engineers 1695

process scalable method of producing insulin-oligomer con-jugates. Subsequently, the desired molecule was purified todesired purity using RP-HPLC and was then characterized.Purified IN-105 was converted to a tablet form and used fororal administration.

Experimental Procedure

Materials

N-hydroxisuccinimide activated methoxy-triethylene glycol(molecular weight 333 Da) and crystalline recombinanthuman Zn-insulin (Purity 98%) used for these studies wereprepared at Biocon Limited, Bangalore, India. The chroma-tography media used for RP-HPLC was C8 Kromasil, 13lm, 100 A silica (Eka Chemicals, Bohus, Sweden). Aceticacid, trifluoroacetic acid, acetonitrile, zinc chloride, phenol,boric acid, sodium hydroxide, and all other reagents and or-ganic solvents used were analytical grade and were procuredfrom qualified suppliers. D,L-Dithiothreitol (minimum 99% ti-tration) and enzyme Endoproteinase Glu-C were obtainedfrom Sigma Chemical Company, St. Louis, USA.

Analysis of samples

Samples at each downstream process stage were analyzedusing a reversed phase high-performance liquid chromatogra-phy (RP-HPLC) Agilent 1100 series system (Agilent Tech-nologies, Santa Clara, USA) fitted with an analytical (4.6mm � 250 mm) symmetry C18, 5 lm, 300 A column(Waters Corporation, Milford, MA). A gradient elution wasperformed at a flow rate of 1 mL/min with solvent A (0.1%trifluoroacetic acid in water) and solvent B (acetonitrile).

The column was equilibrated with 25% solvent B for 3 min,followed by a gradient program where solvent B changedfrom 25% to 40% over 15 min.18 All the chromatogramswere monitored at UV 220 nm.

The reaction by-products in the oligomer insulin mixtureresulting from the conjugation reaction were identified usinghigh-pressure liquid chromatography tandem with mass spec-trometry. Molecular masses were analyzed in LC-ESI-MS(Agilent 1100 LCMS) and ESI-Q-TOF mass spectrometer(Q-Star-EXEL from ABJ Sciex).

Liquid chromatography—mass spectrometry

The conjugation reactions were monitored online usingLC-ESI-MS. The mass spectrometer was the HCT UltraPTM discovery system (Bruker Daltonics, Germany) withelectrospray ionizer. The analysis was performed in positiveionization mode with nebulizer gas pressure at 60 psi, dryinggas at 12.0 L min�1 and drying temperature was kept at350�C. The mass scan range was between 400 and 2200 m/z.

Mass spectrometric analysis of intact IN-105 was per-formed on an Autoflex III smart beam matrix assisted laserdesorption ionization time-of flight time-of flight (MALDI-TOF/TOF) mass spectrometer (Bruker Daltonics, Germany)in positive ionization mode, using a-cyano-4-hydroxycin-namic acid (CHCA) as matrix. Operating conditions were asfollows: lens voltage 8.3 kV, ion source 1 ¼ 19.0 kV, ionsource 2 ¼ 16.48 kV, pulsed ion extraction time ¼ 50 ns,matrix suppression ¼ 500 m/z, and detector gain voltage1590. The samples were mixed with matrix solution at aconcentration of 1 lg lL�1. One microliter of sample wasdeposited on the MALDI plate and allowed to air dry atroom temperature. The data were recorded and analyzedusing the Flex control software.

Procedure for conjugation of oligomer to human-insulin(PEGylation reaction)

Purified crystalline human Zn-insulin (35 gm) was solubi-lized in 500 mM borate buffer, pH 8.3. The solution becameclear after stirring for 10–15 min at room temperature. Thissolution was used as stock and different conjugation reactionconditions were prepared (each 5 mL) for screening purposesfrom the stock solution. The first set of conjugation reactionswas carried out at different pH values (pH 5.5–11) at every0.5 pH interval, and the pH was adjusted with acetic acid orwith 2.5 N NaOH. The reactions were performed using afixed molar ratio of insulin to activated oligomer (1:1.6) inthe presence of 200 mM borate buffer. The required amountof oligomer was solubilized in acetonitrile and added to themixture to achieve a final solvent concentration of 28% (V/V). A second set of conjugation reactions were carried outby varying the molar ratio of activated oligomer to insulin ina wide range (1:0.7, 1:0.9, 1:1.2, 1:1.4, 1:1.6, 1:1.8, 1:2,1:2.3). The required amount of oligomer was solubilized inacetonitrile, maintaining the final solvent concentration at28% (V/V). The conjugation reactions were performed at pH10.5 using 200 mM borate buffer. A third set of reactionswere performed with different molar concentrations of boratebuffer ranging from 50 to 300 mM at pH 10.5, using 1:1.6molar ratio of insulin to activated oligomer. The oligomerwas solubilized in acetonitrile at a final solvent concentrationof 28% (V/V). A fourth set of reactions were carried out bydissolving the required amount of activated oligomer in

Figure 1. Primary structure of IN-105.

The molecule is composed of two peptide chains referred to asthe A chain (starts with amino acid G) and B chain (starts withamino acid F). A and B chains are linked together by two di-sulfide bonds, and an additional disulfide is formed within theA chain. The A chain consists of 21 amino acids and the Bchain of 30 amino acids. Methoxy-triethylene glycol propionylmoiety is attached to the lysine residue at 29th position of theB chain. The single letter represents the abbreviated form ofthe corresponding amino acid in the A and B chain as follows:A Chain: G: glycine, I: isoleucine, V: valine, E: glutamic acid,Q: glutamine, C: cysteine, C: cysteine, T: threonine, S: serine,I: isoleucine, C: cysteine, S: serine, L: leucine, Y: tyrosine, Q:glutamine, L: leucine, E: glutamic acid, N: asparagine, Y: tyro-sine, C: cysteine, N: asparagine; B Chain: F: phenylalanine, V:valine, N: asparagine, Q: glutamine, H: histidine, L: leucine, C:cysteine, G: glycine, S: serine, H: histidine, L: leucine, V: va-line, E: glutamic acid, A: alanine, L: leucine, Y: tyrosine, L:leucine, V: valine, C: cysteine, G: glycine, E: glutamic acid, R:arginine, G: glycine, F: phenylalanine, F: phenylalanine, Y: ty-rosine, T: threonine, P: proline, K: lysine, T: threonine.

1696 Biotechnol. Prog., 2010, Vol. 26, No. 6

varying quantity of solvents, and the reaction was performedusing same molar ratio of insulin to activated oligomer(1:1.6), in 200 mM borate buffer at pH 10.5. The requiredamount of oligomer was solubilized in different amounts ofacetonitrile so that percentages of solvent in the reactionmixture were 9%, 16%, 28%, 37%, and 44%. Different typesof protic and aprotic solvents were explored in the fifth setof conjugation reaction. Conjugation reactions were carriedout using the optimum solvent concentration with differenttypes of solvent [I: ethanol, II: acetonitrile, III: dimethylformamide, IV: iso-propyl alcohol, V: ethanol and acetoni-trile mixture (1:1), VI: ethanol and dimethyl formamide mix-ture (1:1)], keeping all the other previous optimizedparameters constant.

Conjugation reactions were initiated by adding theoligomer solution to the insulin reaction mixture. The reac-tions were allowed to continue for 1 h at 24�C � 2�C understirring conditions. The reactions were quenched by decreas-ing the pH of the reaction mixture to 4.0 by adding aceticacid. The samples were analyzed using analytical reversedphase HPLC for monitoring the progress of the reactions.

Percentage conversion of the final product, IN-105, otherproduct related impurities, and unreacted insulin which werepresent in the conjugation reaction mixture were calculatedusing the HPLC peak area percentage.

% Conversion to IN-105 ¼% Peak area of IN-105

%peak area of all conjugatedproductsþ%unreacted Insulin

Crystallization of conjugated product

The end-conjugation product, containing IN-105 and theproduct related impurities as well as unreacted insulin, wasacidified with 0.4 N acetic acid till the pH reaches to �4.0.At this stage, 0.2% distilled phenol (V/V) and 6% of ZnCl2(V/V) from 0.3 M stock ZnCl2 solution were added. Finally,pH was raised to 5.0 � 0.1 with 2.5 N NaOH to initiate thenucleation.19,20 The mixture was stirred for an additional 30min at 24�C and then kept at 4�C for 14 h without stirringfor complete crystallization.

The crystal slurry was centrifuged at 15,000g for 30 minin Avanti J-26 XP centrifuge (Beckman Coulter, Fullerton,CA), and the supernatant was analyzed in analytical reversedphase HPLC for determining the crystallization efficiency.The morphology of the crystals was checked undermicroscope.

Purification of IN-105 from the methoxy-PEG insulinmixture and unmodified insulin

IN-105, the major component of the conjugation reactionmixture, was purified from the mixture of monoconjugates,multiconjugates, and the unreacted residual insulin by usingpreparative RP-HPLC.

Reverse phase chromatography trials were carried out inthe AKTA explorer system (GE healthcare, Bio-SciencesAB, Uppasala, Sweden) using 250 mM acetic acid, 50 mMsodium acetate, pH 4.0, (Buffer A), and acetonitrile (BufferB) as organic modifier. Chromatographic conditions wereoptimized using Kromasil C8, 13 lm, 100 A silica mediaprepacked in 4.6 mm � 250 mm column. Large-scale prepa-rative RP-HPLC was carried out in 50 mm � 250 mm

dynamic axial compression (DAC) column, packed manuallyusing the same silica resins. Crystallized reaction mixturewas solubilized in buffer A, filtered through 0.2-lm filterand loaded onto the column. A linear gradient of 20–23% ofBuffer B over 20 CVs was carried out to separate the impur-ities18,21–23 at a flow rate of 200 cm/h. Individual fractionsof flow through as well as gradient elution were collectedand analyzed by C18 Symmetry column. The fractions fromRP-HPLC containing purified IN-105 were pooled, and thefraction pool was taken for crystallization.

Crystallization of purified IN-105

The elution pool, pH 3.9, was diluted by 25% with waterto bring down the acetonitrile concentration for efficientcrystallization. At this stage, 0.4% phenol (V/V) and 8% ofZnCl2 (V/V) from 0.3 M stock solution of ZnCl2 wereadded. Finally, pH was raised to 4.8 � 0.1 with 2.5 NNaOH to initiate the nucleation process. The mixture wasstirred for additional 30 min at 24�C, and finally solutionwas kept at 4�C for 12 h without stirring for completion ofcrystallization and settling of the crystals under gravity.Samples from the crystal slurry were collected at an intervalof 1 h for 12 h. Morphology of the crystals was checkedunder the BH-2 microscope (Olympus, Japan) fitted with acamera. The crystal slurry sample was centrifuged at15,000g for 5 min (Eppendorf, Hamburg, Germany), and thesupernatant was analyzed in the analytical RP-HPLC tomonitor the product loss in the supernatant. The crystalswere washed with purified cold water to remove excess zinc,phenol, and acetonitrile. The crystal cake, collected from thecentrifuge, was lyophilized to dry powder.

Scale up of IN-105 manufacturing

Scale up production of IN-105 was performed in the pilotscale using the optimum conditions for each step. In brief,23.5 gm of purified crystalline human insulin was solubilizedin 200 mM borate buffer, pH 8.3. The pH of the insulin so-lution in borate buffer was adjusted to 10.5 with 2.5 MNaOH. Activated oligomer (2 gm) was solubilized in aceto-nitrile to make 28% (V/V) of solvent concentration in thefinal reaction mixture. The conjugation reaction was initiatedby addition of oligomer solution to the insulin solution. Theconjugation reaction was continued for 1 h at 24�C � 2�Cand was terminated by bringing down the pH of the reactionmixture to 4.0 � 0.1 by adding 0.4 N acetic acid. The acidi-fied reaction mixture, containing IN-105 and the productrelated impurities, was crystallized and centrifuged asdescribed earlier. The crystal pellet was redissolved in 250mM acetic acid, 50 mM sodium acetate, pH 4.0, and loadedonto the C8 100 A, 13-l resin packed in LC-50 column. Thepurified IN-105 was eluted from column using a linear gradi-ent of 20–23% acetonitrile. The fractions were analyzed andonly the fractions which showed greater than 98% purity forIN-105 were pooled together. The purified elution pool wascrystallized and centrifuged as described earlier. The crystalcake was collected from the centrifuge, washed with coldwater, and dried in the lyophilizer.

Structural characterization of IN-105

Reduction and Alkylation of IN-105 and ComparativeStudies with Insulin: Insulin and IN-105 were reduced toA- and B-chains, and the masses were measured separately

Biotechnol. Prog., 2010, Vol. 26, No. 6 1697

to identify the chain at which conjugation had taken place.The method was qualified using human insulin as a standard.Human insulin (10 mg) was dissolved in 500 lL of 8 Mguanidine HCl, 0.1 M TRIS, and 1 mM EDTA buffer main-tained at a pH of 9.0. To this, 10 lL of 1 M dithiothreitol(DTT) was added. The contents were mixed and incubatedat 37�C for 2 h and analyzed by liquid chromatography massspectrometry (LC/MS).

Peptide Mapping of IN-105 and Comparative Studies withInsulin: Insulin and IN-105 were digested using Endopro-teinase Glu-C to generate the peptide fragments. Both insulinand purified IN-105 were solubilized in ammonium bicar-bonate buffer, pH 8.2 at a concentration of 1 mg/mL. Tenmicroliter of endoproteinase Glu-C (1 mg/mL) was added to1 mL protein solution, and the reaction mixture was incu-bated at 37�C for 2 h.24,25 Analytical RP-HPLC was used tomonitor the protease reactions, and mass spectral analyseswere performed to identify each chromatographic peak.

Preparation of IN-105 tablet and itspharmacological properties

Product IN-105 of 5 mg, 7.5 mg, 10 mg, and 15 mg wasmixed with sodium caprate, mannitol, polyvinyl pyrrolidone,colloidal silica, magnesium stearate as excipients. The tab-lets, containing different amounts of IN-105, were preparedby direct compression method. The tablets were uncoatedand designed to give[75% drug release in 15 min and wereexpected to release the entire drug in the stomach region.The tablets were tested in a dose ranging study in 20 Type 2diabetes patients. The doses tried were 10 mg (2 tab � 5mg), 15 mg (2 tab � 7.5 mg), 20 mg (2 tab � 10 mg), and30 mg (2 tab � 15 mg). The tablets were administered 20min before the meal. The pharmacokinetics and pharmaco-dynamics were compared against placebo tablets.26

Results

Optimization of conjugation reaction

The conjugation reaction was completed within 40 min.Depending on the conjugation reaction condition, insulinwas converted to a complex mixture of three different typesof mono- (M1, M2, and M3), three different types of di- (D1,D2, and D3), and trace of tri-conjugates (T) (Figure 2,Table 1).

Effect of pH on the Conjugation Reaction: pH played asignificant role in the conjugation reaction. At acidic pH, theconversion of insulin to its different PEGylated forms waspoor. At higher pH, the conversion of PEGylated forms wassignificantly increased. At pH 7, formation of diconjugatedproducts D2 and D3 and monoconjugated product M3 washigh, whereas monoconjugated M1 was found to be maximalat pH 8.0. Diconjugated product D1 was maximal at pH 8.5,whereas the LysB29 site was more reactive under alkalineconditions, and the desired monoconjugated product M2 (IN-105) was steadily increased with increasing pH (Figure 3A).The conversion of insulin to the desired product IN-105 wasabout 78% at pH 11.0.

Effect of Different Molar Ratio of Reactants in theConjugation Reaction: Conjugation reactions were carriedout at pH 10.5 in 200 mM borate buffer at 24�C � 1�Cusing different molar ratio of human recombinant insulinand activated oligomer. It was observed that a largeamount of insulin remains unreacted under lower molar ra-tio of 1:0.7 (insulin:oligomer). With an increase in theamount of activated oligomer, the desired monoconjugatedproduct IN-105 was increased. The diconjugated productD1 increased as the molar ratio of oligomer increased, andhence the overall yield of the product IN-105 in the reac-tion was reduced (Figure 3B). The other components of the

Figure 2. Analytical RP-HPLC chromatogram of conjugation reaction mixture.

Procedure for analytical RP-HPLC run is described in Experimental Procedure.

Table 1. Retention Times and Masses of Conjugate Mixture

Peak Identification Retention Time (min)

Molecular Weight (Da)

Nature of ConjugationCalculated Experimental

Insulin 12.088 5807.7 5808.7 NonconjugatedM1 12.629 6025 6025.2 MonoconjugatedM2 (IN-105) 12.879 6025 6025.6 MonoconjugatedM3 13.961 6025 6025.2 MonoconjugatedD1 13.558 6243 6243.6 DiconjugatedD2 14.539 6243 6243.6 DiconjugatedD3 14.922 6243 6243.2 Diconjugated


conjugated insulin mixture (e.g., M3, D2, and T) did notchange significantly with change in the molar ratio of insu-lin and activated oligomer (data not shown). Thus, an opti-mum molar ratio of insulin and activated oligomer wasrequired to have the maximum yield in the step with theformation of a minimum amount of related impurities,which aided in effective subsequent chromatographyseparation.

Effect of Buffer Molarities: Conjugation reactions werecarried out at pH 10.5 and an insulin:oligomer molar ratio of1:1.6 at varying borate concentrations. It was observed thatincrease of borate buffer strength had a positive effect on thepercentage purity of IN-105 and the yield of conjugationreaction. However, there was no increase in product purityobserved beyond 200 mM of borate buffer concentration(Figure 3C). With an increase in the borate buffer

Figure 3. Relative amount of different conjugated components as well as unconverted insulin in the reaction mixture at the end ofconjugated reaction with variation of different parameters.

The conjugated sample under different experimental conditions were analyzed in the analytical HPLC as described in experimental procedure sec-tion. Different percentage proportions of individual component present in the conjugation mixture were plotted in the Y axis. Different conditions ofindividual parameter were plotted in the X axis of the individual figure (A–E). (A) pH, (B) molar ratios of Insulin and oligomer, (C) salt concentra-tion, (D) solvent concentration, and (E) different types of solvents [I: ethanol, II: acetonitrile, III: dimethyl formamide, IV: iso-propyl alcohol, V:ethanol and acetonitrile mixture (1:1), VI: ethanol and dimethyl formamide mixture (1:1)]. Descriptions of the experiments are available in Materialsand Methods section. The different components are represented as insulin (—l—), M1 (—n—), IN-105 (—~—), M3 (—^—), D1 (—X—), D2(—D—), and D3 (—*—).


concentration, generation of diconjugated product, D1 wasdecreased, whereas unreacted insulin was increased.

Effect of Solvent on Conjugation Reaction: Conjugationreactions were carried out with different concentrations ofsolvent and different types of solvent. Molar ratio of insulin-oligo, pH and borate buffer concentration in the reactionwere kept the same as the optimized values in the earlierexperiments. At low-solvent concentrations, the coupling ofoligomer with insulin was much more nonspecific and for-mation of diconjugated products was high, with a minimalamount of unreacted insulin remaining (Figure 3D). Yield ofthe product was maximal at 28% solvent concentration inthe conjugation reaction. Amongst all the tested solvents inthe conjugation reaction, acetonitrile (28%, V/V) was foundto be the most efficient solvent for the conjugation reaction,whereas iso-propyl alcohol was least effective (Figure 3E).

The maximum relative percentage of the product wasobserved to be 65–70% using optimized conditions of all theparameters in the conjugation reaction.

Crystallization of conjugated product

Crystallization after conjugation was observed to be highlydependent on pH. Morphology of the crystals and the effi-ciency of the crystallization varied widely at different pH(data not shown). At low pH (4.8–5.0), product loss in themother liquor was minimum, whereas the clusters of rodshaped crystals (�10 lm size) were visible when observedunder the microscope (Figure 4A). Residual productincreased in the supernatant after centrifugation of the crystalbroth at pH[ 5.0. The supernatant fraction, after centrifuga-tion, was discarded and around 2% product loss wasobserved in the supernatant under optimum crystallizationconditions. The product was purified from unreactedoligomer that was removed in the supernatant fraction aftercentrifugation.

Preparative chromatography separation of IN-105from the conjugated mixture

The conjugation reaction produced a mixture of propionylinsulin conjugated products. IN-105, the major component ofthe conjugation reaction mixture, was purified from the othercomponents of conjugated insulin (monoconjugates; multi-conjugated insulin) and unmodified insulin by using prepara-tive RP-HPLC. Upon analysis of the fractions throughanalytical RP-HPLC, it was observed that most of theunreacted insulin flows through the column in the specifiedcondition. The procedure for both preparative RP-HPLC and

analytical RP-HPLC were described in experimental proce-dure section. The early gradient elution pools in the prepara-tive RP-HPLC run content mixture of M1, D1, and D2. Theproduct M2 (IN-105) started elution at the later part of thegradient elution fractions. The IN-105 enriched fractionswere pooled together and analyzed once again. It wasobserved that the pooled fractions contain IN-105 with [98% purity.

Final crystallization of IN-105

Crystallization of purified IN-105 could be efficiently per-formed directly from the chromatographic elution pool. Mor-phology of the crystals was checked under microscope(40�). No crystal formation was observed below pH 4.5. Ef-ficiency of crystallization increased as the pH of crystalliza-tion increases above pH 4.5 and was found to be maximal atpH 4.8–5.0, using 0.4% (V/V) phenol and 8% ZnCl2 (V/V)of 0.3 M stock ZnCl2 solution. Initiation of needle-shapedcrystals was observed from 1 h of settling. The crystal for-mation completed within 6–8 h of settling (Figure 4B). Theproduct loss in initial time of mother liquor samples wasfound to be very high, but it gradually decreased and ulti-mately stabilized after 8 h under the above mentionedcondition.

Scale up downstream process

The optimized condition of each step for manufacturing ofIN-105 from recombinant human insulin was scaled up eas-ily. Results of step yield and purification along the down-stream process was shown in the purification table (Table 2).An overall yield of 48% was achieved in the downstreampurification process. Although the chromatography purity ofthe starting material was [98%, the purity came down to71% after the conjugation step due to the formation of non-specific conjugated product and unreacted insulin present inthe conjugation reaction mixture. Finally, upon analysis inanalytical RP-HPLC, IN-105 with more than 98% purity wasachieved at the end of purification process.

Characterization of IN-105

The analytical HPLC profile of the conjugation reactionmixture recorded at 220 nm showed peaks corresponding tounreacted insulin, monoconjugated insulin (IN-105), anddiconjugated insulin in Figure 2. The retention times of indi-vidual components and their masses are given in Table 1.Introduction of single PEGylation unit in human insulinresulted in shifting of retention time from 12.1 to 12.9 min

Figure 4. Photo micrograph of conjugation end product (A) and purified IN-105 (B) crystals.


(due to decrease in polarity) and an increase in molecularmass of 218 Da. Table 1 also shows other monoconjugatedinsulin (m/z: 6025) and diconjugated insulin (m/z: 6243. 6)in addition to IN-105 peaks. The molecular weight observedfor IN-105 by electrospray ionization and matrix assistedlaser desorption mass spectra is consistent with calculatedmolecular weight for the formula C267H401N65O82S6; m/z:6025.6 (Figure 5).

Reduction experiments on IN-105 were carried out bydithiothreotol (DTT) to identify the chain at which conjuga-tion occurred. The retention times and masses obtained afterreduction of insulin and IN-105 are given in Table 3.Reduced IN-105 showed peaks for chain A and chain B at13.9 min and 14.2 min, respectively. A comparative analysisof IN-105 with insulin showed an increase in retention timefor chain B from 12.4 to 14.2 min and increase in molecularweight of 218 Da. Identification of site of conjugation wascarried out by peptide mapping experiments using enzymeGLU-C. The summary of peptide mapping data observed forIN-105 in comparison with insulin is shown in Table 4. Thepeptide fragment B22–B30 of IN-105 showed increase inretention time from 23.8 to 36.7 min due to addition of asingle PEG unit. Increase in molecular weight of 218 Daalso confirmed the presence of oligomer in B22–B30 fragmentof IN-105. On the basis of the peptide mapping experiment,we concluded that the position of oligomer in IN-105 is onB-29 Lysine residue. Site of oligomer attachment was furtherconfirmed by performing MSMS on B22-B30 peptide frag-ment of IN-105 (data not shown).

In vivo functional activity of oral PEG-conjugated insulin

The study results indicated that IN-105 was absorbed in adose-dependent manner with Tmax of about 20 min, and theplasma IN-105 levels came back to baseline in about 80–100min (Figure 6A). The glucodynamic effect in terms of maxi-mum drop in plasma glucose levels from the baseline for dif-ferent doses of IN-105, in comparison with placebo, alsoshowed a linear dose-response relationship (Figure 6B).

Discussion

With the use of prior analytical information, the PEGylationreaction was studied most extensively and it was concludedthat a perfectly optimized reaction condition is required toachieve a high degree of specific binding of PEG derivativeto insulin. A perfect balanced formation amongst M1, M2, M3,D1, D2, D3, and T can be achieved by changing the reactionconditions. Depending on the reaction conditions, N-acylationof an unprotected insulin led to formation of variable individ-ual components in a complex mixture of mono-, di-, and tri-conjugates, as shown in the results section (Figure 2, Table1). Each individual component of the conjugation mixturewas identified by mass spectophotometry and the site ofattachment of oligomer moiety to the insulin was confirmedby peptide mapping and fragmentation studies on B22–B30

fragment observed in IN-105. It was found that oligomer isattached at the B-29 position in IN-105 (Table 4).

The relevant parameters for optimal production of IN-105have been identified in this study. It was observed that molarratio of the PEG molecule and insulin, the pH of the reac-tion, and the buffer compositions are the most important pa-rameters that control the conjugation reaction. However, thereaction time did not influence the production yield strongly.The optimal reaction condition improved the step yield andsimultaneously reduced the product related impurities, whichsimplified the further downstream purification steps. Toachieve high yield in the chromatographic step, it was im-portant to get high-relative abundance of the product (IN-105) in the mixture in the conjugation step without sacrific-ing the step yield.

The number of PEG molecules attached to a proteinincreased nonlinearly if a large excess of activated PEG waspresent. Precautions had been taken in the process to avoidprecipitation in the presence of high excess activated PEG,which also helped in economics of the process. It is obviousthat higher oligomer concentration would definitely lead toformation of di-/tri-conjugated product, and the step yieldwould simultaneously decrease. On the other hand, lowerconcentrations of oligomer would lead to a higher amount ofunreacted insulin molecule, which again hampers the stepyield. Thus, an optimum oligomer concentration is requiredto achieve abundant PEGylation at the desired B-29 site. Itwas observed that a molar ratio 1:1.6 (insulin:oligomer) wasthe best possible combination for achieving the maximumyield. The pH of the conjugation reaction also played a very

Table 2. Purification Table

Step Product ID Amount (gm) Product Purity (%) Step Yield (%) Overall Yield

Starting material Zn-insulin 23.5 98.8Conjugation IN-105 15.4 71 66 66Crystallization 1 IN-105 14.5 71.7 94 62Revered phase HPLC IN-105 11.4 98.1 79 48.5Crystallization 2 IN-105 11.2 98.5 98 48

Figure 5. Electrospray ionization (Top) and matrix assistedlaser desorption ionization mass spectra (Bottom) ofIN-105.


important role. High pH helped in coupling the oligomermoiety in the B-29 (Lys) position of human insulin. It isknown that the reactivity of a functional group is influencedby its location on the protein macromolecules. The heteroge-neity in reactivity has been attributed to differences in pKa

of specific amino acid side chain, as well as conformationaland environmental factors.27 Nucleophilic attack will onlytake place when the pH of the protein solution is near orabove the residue’s pKa.

7,28 LysB29 (pKa � 10.5) is far morereactive at the conjugation reaction pH 10.5 than the otherreactive group PheB1 (pKa \ 7.0) and GlyA1 (pKa � 8.0).Although complete selectivity was not observed, extremeheterogenicity of multiconjugated product was greatlyreduced by proper conjugation optimization processes usedin this study. Similar types of observations were found byLee et al.29 Changes in other conjugation conditions likebuffer molality, solvent concentration, and the effect of dif-ferent types of solvents have been explored extensively.Upon evaluation of the preliminary reaction optimizationcondition, it was observed that there is no marked differen-ces in the reaction profile in the presence of a wide range ofsolvent and buffer concentrations. Thus, as optimizing im-portant parameters (pH and oligomer concentration), a fixedamount of acetonitrile (28%) and 200 mM borate buffer wasused. Once the other important parameters for optimum reac-tion condition were fixed, the role of different solvents atdifferent concentrations and the borate buffer concentrationwas determined. Although those parameters had no drasticeffect on the conjugation step, to achieve the maximum yieldand purity all these parameters were also checked. Theunique folding of a protein chain in the presence of differentreaction conditions can be responsible for masking functionalgroups, and can make them more or less approachable to theoligomer. Because the folding of protein chains is governedby thermodynamic factors, the availability and the reactivityof the target site of the protein can be altered by changingthe reaction conditions. Therefore, the reactivity of PEG de-rivative towards protein depends largely on the structure ofthe protein, as well as specific reaction conditions in whichthe reaction is carried out.30

Introducing the crystallization step right after the conjuga-tion can remove unwanted material like excess oligomer,

solvent, borate, etc. in the supernatant of the process, butthis step was unable to improve the chromatographic purityof the product. Although the starting material for the crys-tallization in this stage was quite impure, a good crystalliza-tion condition was obtained with minimum product loss.The main aim of the crystallization step at the end of theconjugation reaction was to clean up the product beforeprocessing in the chromatographic step; the morphology ofthe crystals was not the prime interest at that stage.Although exploring the optimization of the crystallizationprocedure at this stage, we observed that the individualcrystal morphology of the two different stages was quitesimilar. However, due to the presence of the excess solvent,oligomer, and borate buffer at the end of the conjugationstage, the rod-shaped crystals form clusters. Crystallizationis an important intermediate step in the scale up manufac-turing process of bio-therapeutics. It provides the flexibility

Table 3. Retention Times and Masses of Reduced Insulin and IN-105

Intact Molecule Chain A Chain B

RT (min)

Molecular Weight (Da)

RT (min) Molecular Weight (Da) RT (min) Molecular Weight (Da)ESI-MS MALDI-MS

Insulin 9.6 5808.2 5807.9 13.9 2384.3 12.4 3430.7IN-105 10.8 6026.5 6025.6 13.9 2384.2 14.2 3648.8

Table 4. Comparison of Retention Times and Masses Glu-C

Digested Fragments of Insulin and IN-105

Fragment (s)Retentiontime (min)

MolecularMass (Da)

Intact Insulin 41.7 5808.3IN-105 42.4 6025.6

A1-A4 Insulin 13.1 417.2IN-105 13.1 417.2

B22-B30 Insulin 23.8 1116.7IN-105 36.7 1334.8

B14-B21; A18-A21 Insulin 25.9 1377.6IN-105 25.9 1377.6

B1-B13; A5-A17 Insulin 38.5 2969.1IN-105 38.5 2969.1

Figure 6. Plasma insulin profiles (A) after treatment with IN-105 tablets and (B) the net change in glucose frombaseline.

The IN-105 expressed in terms of plasma total insulin is pro-portional to the dose. The net change in glucose from baseline(B) is proportional to the IN-105 exposure and glucose excur-sions are lower as compared with placebo. The data areexpressed as mean � SE (open circles) placebo, (open squares)10 mg IN-105, (open inverse triangle) 15 mg IN-105, (open tri-angle) 20 mg IN-105, and (open quadrangle) 30 mg IN-105(figure adapted from the Ref. 26)


to store the product in the downstream flow and helps indissolving the crystal in a desired buffer that is suitable forfurther purification steps. The crystallization step was alsocapable of improving the quality of the product in the suc-cessive purification steps. A single RP-HPLC step was ableto improve the purity of the product from 72% to 98% witha yield of �80%, which is ideal to have in the downstreamprocess of a commercially viable therapeutic protein. Opti-mization of conditions in preparative RP-HPLC was per-formed in the 4.6 mm � 250 mm column, but thechromatographic separation in the large scale was success-fully achieved in the LC50 column.

Bulk crystallization was performed directly from the chro-matographic elution pool instead from the lyophilized puri-fied material to ease the downstream process. Although thereverse phase preparative chromatography was performed inthe presence of sodium acetate and acetonitrile under acidicconditions, crystallization of purified IN-105 directly fromthe elution pool could be managed efficiently with high re-covery (Table 2). The microscopic view of the crystal shapeof purified IN-105 opens up the scope of further studies toevaluate the crystal structure of single product crystal.

Finally, tablets of different strength were prepared fromthe crystalline IN-105 material and administered orally toType 2 diabetic patients. Upon oral administration of IN-105, plasma insulin levels peaked at �20 min, suggestingthat the absorption was happening in the upper GI tract. Theglucodynamic effect in terms of the maximum drop inplasma glucose levels after administration of 10, 15, 20, and30 mg doses of IN-105 increased as a function of dose andwas significant when compared with placebo (Figure 6).

The choice of N-hydroxysuccinamide activated methoxy-propionyl moiety of PEG molecule with low molecularweight, and its PEGylation site was the outcome of a sys-tematic screening process in which various lengths andchemistries of PEGs were tried and tested in laboratory.We found that 333 Da PEG retained the complete pharma-cological activity as compared to insulin. It is reported thatbiological activity of PEG mono-substituted insulin is simi-lar to native insulin when the molecular weight of PEGused for PEGylation is low.11,15 The LysB29 site of insulinwas selected for PEGylation with an aim of successful oraldelivery of the conjugated drug. There are reports ofincrease in physicochemical stability of the PEGylated in-sulin if the PEG moiety is attached to PheB1.

11,15 In thesereports, only the physical stability of the molecules havebeen studied, which is of less relevance from an oral deliv-ery point of view. We believe that there are other attributes,which are more important for successful oral delivery. Forexample, resistance to degradation by digestive enzymes inGIT, and the solubility of the molecule combined with sta-bility during the tableting process are all major concerns.Coupling of the PEG moiety to lysine at B-29 position ofinsulin seems to protect insulin from digestive enzymes inin vitro experiments, and the half-life was found to be 2.5times longer than insulin. This could be attributed to stearichindrance created by alkyl PEG moiety. Additionally, thesolubility of LysB29 modified analogue was about fivetimes higher than insulin itself. In addition, there are otheradvantages to selecting LysB29 for modification site likereduced mitogenicity. Because B-22 through B-30 aminoacids are involved in IGF receptor binding, PEGylation atB-29 causes stearic hindrance, resulting in lower IGFRbinding.

It can be concluded in this study that the final IN-105product can be made through a scalable process, the site spe-cific attachment of alkyl PEG to insulin can create orallyabsorbed bioactive conjugated insulin that has a dose-de-pendent glucose lowering effect in Type 2 diabetes patients,and hence, such PEG-insulin conjugates represent newpotential candidates for orally administered insulinpreparations.

Acknowledgments

The authors thank Dr. Vivekanandan Kannan for his valuablecomments on the manuscript.

Literature Cited

1. Zalipsky S. Chemistry of polyethylene glycol conjugates withbiologically active molecules. Adv Drug Deliv Rev. 1995;16:157–182.

2. Abuchowski A, van Es T, Palczuk NC, Davis FF. Alteration ofimmunological properties of bovine serum albumin by covalentattachment of polyethylene glycol. J Biol Chem. 1997;252:3578–3581.

3. Veronese FM, Harris JM. Peptide and protein PEGylation. III.Advances in chemistry and clinical application. Adv Drug DelivRev. 2008;60:1–2.

4. Gaberc-Porekar V, Zore I, Podobnik B, Menart V. Obstaclesand pitfalls in the PEGylation of therapeutic proteins. CurrOpin Drug Discov Devel. 2008;11:242–250.

5. Veronese FM, Pasut G. PEGylation, successful approach todrug delivery. Drug Discov Today. 2005;21:1451–1458.

6. Veronese FM. Peptide and protein PEGylation: a review ofproblems and solutions. Biomaterials. 2001;22:405–417.

7. Robert MJ, Bentley MD, Harris JM. Chemistry for peptide andprotein PEGylation. Adv Drug Deliv Rev. 2002;54:459–476.

8. Tom I, Lee V, Dumas M, Madanat M, Ouyang J, Severs J,Andersen J, Buxton JM, Whelan JP, Pan CQ. Reproducible pro-duction of a PEGylated dual-acting peptide for diabetes. AAPSJ. 2007;9:227–234.

9. Tsubery H, Mirronchik M, Fridkin M, Shechter Y. Prolongingthe action of protein and peptide drugs by a novel approach ofreversible polyethylene glycol modification. J Biol Chem. 2004;279:38118–38124.

10. Fuerteges F, Abuchowski A. The clinical efficacy of poly(ethyl-ene glycol)-modified proteins. J Controlled Release. 1990;11:139–148.

11. Uchio T, Baudys M, Liu F, Song SC, Kim SW. Site-specific in-sulin conjugation with enhanced stability and extended actionprofile. Adv Drug Deliv Rev. 1999;35:289–306.

12. Miller MA, Malkar NB, Severynse-Steven D, Yarbrough KG,Mark JB, Dugdell RE, Puskas ME, Krishnan R, Kenneth DJ.Ampiphilic conjugates of human brain natriuretic peptidedesigned for oral delivery: in vitro activity screening. BioconjugChem. 2006;17:267–274.

13. Chae SY, Jin C-H, Shin HJ, Youn YS, Lee S. Preparation, char-acterization, and application of biotinylation and biotin-PEGy-lated glucagon-like peptide-1 analogues for enhanced oraldelivery. Bioconjug Chem. 2008;19:334–341.

14. Iyer H, Khedkar K, Verma M. Oral insulin—a review of currentstatus. Diabetes Obes Metab. 2010;12:179–185.

15. Hinds K, Koh JJ, Joss L, Liu F, Baudys M, Kim SW. Synthesisand characterization of poly(ethylene glycol)-insulin conjugates.Bioconjug Chem. 2000;11:195–201.

16. Conan JF. Size exclusion reaction chromatography (SERC): anew technique for protein PEGylation. Biotechnol Bioeng. 2003;82:200–206.

17. Stainer DF. Cocrystallization of proinsulin and insulin. Nature.1973;243:528–530.

18. Dave N, Hazra P, Khedkar K, Manjunath HS, Iyer H, Suryanar-ayan S. Process and purification for manufacture of a modifiedinsulin intended for oral delivery. J Chromatogr. 2008;1177:282–286.


19. Bromberg L, Julia R-S, Scott T. Insulin particle formation insupersaturated aqueous solutions of poly(ethylene glycol). Bio-phys J. 2005;89:3424–3433.

20. Welinder BS, Sørensen HH, Bruno H. Reversed-phase high-per-formance liquid chromatography of insulin. Resolution and re-covery in relation to column geometry and buffer composition.J Chromatogr. 1986;361:357–367.

21. Welinder BS, Linde S. In CRC Handbook of HPLC for theSeparation of Amino Acids, Peptides and Proteins, Vol. II,Hancock WS Ed, Boca Raton, FL: CRC Press, 1984;357–362.

22. Mandrup G. In-process control of insulin production by high-performance liquid chromatography. Trends Anal Chem. 1992;11:389–395.

23. Lee H, Park TG. A novel method for identifying PEGylationsites of protein using biotinylated PEG derivatives. J PharmSci. 2003;92:97–103.

24. Michael JR, Milton JH. Attachment of degradable poly(ethyleneglycol) to protein has the potential to increase therapeutic effi-cacy. J Pharm Sci. 1998;87:1440–1445.

25. Gordon SJ. Development of oral insulin: progress and currentstatus. Diabetes Metab Res Rev. 2002;18:29–37.

26. Khedkar A, Iyer H, Anand A, Verma1 M, Krishnamurthy S,Savale S, Atignal A. A dose range finding study of novel oral

insulin (IN-105) under fed conditions in type 2 diabetes mellitussubjects. Diabetes Obes Metab. 2010;12:659–664.

27. Hooftman G, Herman S, Schacht S. Polyethylene glycols withreactive endgroups. II. Practical consideration for the prepara-tion of protein-PEG conjugates. J Bioact Compat Polym. 1996;11:135–159.

28. David CW, Voloch M, Lee S, Liu Y-H, Susan C-C, Collette C,Pramanik B. Carboxyalkylated histidine is a pH dependent prod-uct of PEGylation with SC-PEG. Pharm Res. 2001;18:354–1360.

29. Lee S-H, Lee S, Youn YS, Na DH, Chae SY, Byun Y, Lee KC.Synthesis, characterization, and pharmacokinetic studies ofPEGylated glucagon-like peptide-1. Bioconjug Chem. 2005;16:377–382.

30. Katre NV, Knauf MJ, Lorivol WJ. The conjugation of proteinswith polyethylene glycol and other polymers: altering propertiesof proteins to enhance their therapeutic potential. Adv DrugDeliv Rev. 1993;10:91–114.

Manuscript received Oct. 19, 2009, and revision received July 5, 2010.


Documents

Development of a process to manufacture PEGylated orally bioavailable insulin