Process for the purification of recombinant albumin

Abstract

A process is provided for the preparation of a highly pure albumin solution the process comprising subjecting albumin (preferably expressed and secreted by transformed yeast) to a series of chromatographic steps. Preferably, the process comprises the steps of positive mode cation exchange chromatography, positive mode anion exchange chromatography, positive mode affinity chromatography, negative mode affinity chromatography (preferably using immobilized aminophenylboronic acid), negative mode cation exchange chromatography, and negative or positive mode anion exchange chromatography. A process for reducing the level of nickel in an albumin solution is also disclosed, as is a recombinant albumin coding sequence comprising two or more in-frame translation stop codons. Also disclosed is a process for producing recombinant albumin, the process comprising culturing a fungal cell expressing a recombinant albumin coding sequence, wherein the cell has a reduced capacity of mannosylation of the recombinantly-expressed albumin.

Claims

1. A process for producing recombinant albumin, the process comprising culturing a fungal cell expressing a recombinant albumin coding sequence and obtaining the albumin, wherein the cell has a genetic modification which causes the cell to have at least a reduced capacity of mannosylation of the recombinantly-expressed albumin and wherein the culture medium is at least 1,000 L and is of pH 5.3-6.8.

2. The process according to claim 1 wherein said modification(s) comprises any suppression, substitution, deletion, addition, disruption and/or mutational insertion.

3. The process according to claim 2 wherein said modification(s) are stably-inherited and/or are non-reverting and/or are non-leaky.

4. The process according to claim 1 wherein said modification(s) are located in a coding region of a gene or in a region involved in the expression of a gene.

5. The process according to claim 4 wherein the gene is a PMT gene.

6. The process according to claim 1 wherein the fungal cell is cultured in a culture medium of at least 5,000 L.

7. The process according to claim 1 wherein the fungal cell is cultured at pH 6.2-6.7.

8. The process according to claim 1 wherein the fungal cell is cultured at pH 5.3-5.9.

9. The process according to claim 1, wherein the fungal cell is a yeast cell.

10. The process according to claim 9, wherein the yeast cell is a Saccharomyces cell.

11. The process according to claim 9, wherein the yeast cell is a Saccharomyces cerevisiae cell.

12. The process according to claim 9, wherein the yeast cell is a Pichia or Kluyveromyces cell.

13. The process according to claim 5, wherein the gene is PMT1.

14. The process according to claim 1, wherein the fungal cell is cultured in a culture medium of at least 7,500 L.

15. The process according to claim 1, wherein the fungal cell is cultured at pH 6.3-6.5.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

(1) The invention is further illustrated in the appended drawings in which:

(2) FIGS. 1 to 7 respectively show the construction of plasmids pAYE309, pAYE440, pAYE438, pDB2241, pDB2242, pDB2243 and pDB2244; and

(3) FIG. 8 shows electrospray mass spectrometry of conA-binding rHA fraction from rHA prepared according to the invention.

(4) FIG. 9 shows the effect of pH and time on nickel removal from rHA by Chelex.

(5) FIGS. 10 (SEQ ID NO: 1) and 11 (SEQ ID NO: 2) represent two DNA sequences with homology to the protein encoding region Saccharomyces cerevisiae PMT1.

(6) FIGS. 12 to 15 (SEQ ID NO: 3, SEQ ID NO: 4, SEQ ID NO: 5, and SEQ ID NO: 6, respectively) represent four DNA sequences with homology to the protein encoding region Saccharomyces cerevisiae PMT7.

(7) FIGS. 16 (SEQ ID NO: 7) and 17 (SEQ ID NO: 8) represent two DNA sequences with homology to the protein encoding region Saccharomyces cerevisiae PMT5.

DETAILED DESCRIPTION OF PREFERRED EMBODIMENTS OF THE INVENTION

(8) Whereas the processes of the present invention can be utilised to obtain highly purified albumin from an impure albumin solution from a number of sources, such as serum, it is particularly applicable to purifying recombinant human albumin (rHA). The albumin produced in accordance with the invention may be any mammalian albumin, such as rat, bovine or ovine albumin, but is preferably human albumin.

(9) DNA encoding albumin may be expressed in a suitable host to produce albumin. Thus, DNA may be used in accordance with known techniques, appropriately modified in view of the teachings contained herein, to construct an expression vector, which is then used to transform an appropriate host cell for the expression and production of albumin.

(10) The DNA encoding the albumin may be joined to a wide variety of other DNA sequences for introduction into an appropriate host. The companion DNA will depend upon the nature of the host, the manner of the introduction of the DNA into the host, and whether episomal maintenance or integration is desired.

(11) Generally, the DNA is inserted into an expression vector, such as a plasmid, in proper orientation and correct reading frame for expression. If necessary, the DNA may be linked to the appropriate transcriptional and translational regulatory control nucleotide sequences recognised by the desired host, although such controls are generally available in the expression vector. It is beneficial to incorporate more than one DNA sequence encoding a translational stop codon, such as UAA, UAG or UGA, in order to minimise translational read-through and thus avoid the production of elongated, non-natural fusion proteins. A DNA sequence encoding the translation stop codon UAA is preferred. The vector is then introduced into the host through standard techniques, followed by selection for transformed host cells. Host cells so transformed are then cultured for a sufficient time and under appropriate conditions known to those skilled in the art, and in view of the teachings disclosed herein, to permit the expression of the albumin, which can then be recovered.

(12) Many expression systems are known, including bacteria (for example E. coli and Bacillus subtilis), yeasts (for example Saccharomyces cerevisiae, Pichia pastoris and Kluyveromyces lactis), filamentous fungi (for example Aspergillus), plant cells, animal cells and insect cells. The preferred micro-organisms are the yeasts Saccharomyces cerevisiae, Kluyveromyces lactis and Pichia pastoris. It is particularly advantageous to use a yeast deficient in one or more protein mannosyl transferases involved in O-glycosylation of proteins, for instance by disruption of the gene coding sequence.

(13) The albumin protein sequence does not contain any sites for N-linked glycosylation and has not been reported to be modified, in nature, by O-linked glycosylation. However, it has been found that rHA produced in a number of yeast species can be modified by O-linked glycosylation, generally involving mannose. The mannosylated albumin is able to bind to the lectin Concanavalin A. The amount of mannosylated albumin produced by the yeast can be reduced by using a yeast strain deficient in one or more of the PMT genes (WO 94/04687).

(14) The most convenient way of achieving this is to create a yeast which has a defect in its genome such that a reduced level of one of the Pmt proteins is produced. For example, there may be a deletion, insertion or transposition in the coding sequence or the regulatory regions (or in another gene regulating the expression of one of the PMT genes) such that little or no Pmt protein is produced. Alternatively, the yeast could be transformed to produce an anti-Pmt agent, such as an anti-Pmt antibody.

(15) To modify one of the PMT genes so that a reduced level of Pmt protein is produced, site-directed mutagenesis or other known techniques can be employed to create single or multiple mutations, such as replacements, insertions, deletions, and transpositions, as described in Botstein and Shortle. Strategies and Applications of In Vitro Mutagenesis, Science, 229: 193-210 (1985), which is incorporated herein by reference. Suitable mutations include chain termination mutations (clearly stop codons introduced near the 3 end might have insufficient effect on the gene product to be of benefit; the person skilled in the art will readily be able to create a mutation in, say, the 5 three quarters of the coding sequence), point mutations that alter the reading frame, small to large deletions of coding sequence, mutations in the promoter or terminator that affect gene expression and mutations that de-stabilise the mRNA. Specific mutations can be introduced by an extension of the gene disruption technique known as gene transplacement (Winston, F. et al (1983) Methods Enzymol. 101, 211-228).

(16) Generally, one uses a selectable marker to disrupt a gene sequence, but this need not be the case, particularly if one can detect the disruption event phenotypically. In many instances the insertion of the intervening sequence will be such that a stop codon is present in frame with the Pmt sequence and the inserted coding sequence is not translated. Alternatively, the inserted sequence may be in a different reading frame to Pmt.

(17) The gene may have one or more portions (optionally including regulatory regions, up to the whole gene) excised or inverted, or it may have a portion inserted, in order to result in reduced production of protein from one of the PMT loci and/or in the production of protein from one of the PMT loci having a reduced level of activity.

(18) The PMT genes of Saccharomyces cerevisiae encode a family of seven (PMT1-PMT7) protein O-mannosyltransferases which vary in their specificity. These proteins are also known as dolichol phosphate-D-mannose: protein transferases, dolichyl-phosphate-D-mannose:protein O-D-mannosyttransferases or phosphomannose transferases (Gentzsch and Tanner, EMBO 15, 5752-5757, 1996, and references included therein). This family of integral membrane enzymes catalyses the transfer of mannose, in the form of dolichyl phosphate mannose, onto the hydroxyl group of serine or threonine within the polypeptide chain, described by the following reaction:

(19) ##STR00001##

(20) The available evidence suggests that the synthesis of dolichyl phosphate mannose and the subsequent transfer of mannose to the protein occurs in the endoplasmic reticulum.

(21) It is clear that the enzymes of this family have different substrate (protein) specificities (Gentzsch and Tanner (1997) Glycobiology 7, 481-486). Five of seven test proteins were substrates for Pmt1p and Pmt2p, the products of the PMT1 and PMT2 genes respectively, as shown by their under-glycosylation in pmt1 or pmt2 mutant Saccharomyces cerevisiae strains. Another two test proteins were apparently unaffected by either PMT1 or PMT2 mutations, but were under-glycosylated in a pmt4 mutant strain.

(22) The 92 kD Pmt1p protein O-mannosyltransferase enzyme has been purified to homogeneity from solubilised Saccharomyces cerevisiae membranes (Strahl-Bolsinger and Tanner (1991) Eur. J. Biochem. 196, 185-190). The gene encoding for the Pmt1p (PMT1) has been cloned and sequenced. The gene is located on chromosome IV and encodes a single polypeptide with a primary sequence of 817 amino acids (Strahl-Bolsinger et al (1993) P.N.A.S. USA 90, 8164-8168). The sequence information of PMT1 (and other PMT genes) may be used for the identification of related mannosyltransferases encoding genes in Saccharomyces cerevisiae.

(23) The sequences shown in FIGS. 10 and 11 are homologous with the protein encoding region Saccharomyces cerevisiae PMT1, the sequences shown in FIGS. 12 to 15 are homologous with the protein encoding region Saccharomyces cerevisiae PMT7 and the sequences shown in FIGS. 16 to 17 are homologous with the protein encoding region Saccharomyces cerevisiae PMT5. Persons skilled in the art will appreciate that any one of these sequences may be used to identify (or disrupt) a Saccharomyces cerevisiae mannosyltransferase gene. It will be appreciated that fragments of the sequences represented in FIGS. 10 to 17 may similarly be used, as may sequences which are homologous with the sequences represented in FIGS. 10 to 17 and the fragments thereof. Techniques for generating homologous sequences are well known in the art.

(24) It should be appreciated that by a homologous sequence, we include sequences having at least 70%, 80%, 90%, 95%, or 98% homology with a sequence shown in any one of FIGS. 10 to 17, or with a fragment of a sequence shown in any one of FIGS. 10 to 17.

(25) Percent homology can be determined by, for example, by comparing sequence information using the GAP computer program, version 6.0 described by Devereux et al. (Nucl. Acids res. 12:387, 1984) and available from the University of Wisconsin Genetics Computer Group (UWGCG). The GAP program utilises the alignment method of Neddleman and Wunsch (J. Mol. Biol. 48:443, 1970), as revised by Smith and Waterman (Adv. Appl. Math 2.482. 1981). The preferred default parameters for the GAP program include: (1) a unary comparison matrix (containing a value of 1 for identities and 0 for non-identities) for nucleotides, and the weighted comparison matrix of Bribskov and Burgess, Nucl. Acids Res. 14:6745, 1986 as described by Schwarts and Dayhoff, eds, Atlas of Protein Sequence and Structure, National Biomedical Research Foundation, pp 353-358, 1979; (2) a penalty of 3.0 for each gap and an additional 0.10 penalty for each symbol in each gap; and (3) no penalty for end gaps.

(26) If a yeast other than S. cerevisiae is used, disruption of one or more of the genes equivalent to the PMT genes of S. cerevisiae is also beneficial, eg in Pichia pastoris or Kluyveromyces lactis. The sequence of PMT1 (or any other PMT gene) isolated from S. cerevisiae may be used for the identification or disruption of genes encoding similar enzymatic activities in other fungal species. The cloning of the PMT1 homolog of Kluyveromyces lactis is described in WO 94/04687.

(27) If a yeast other than S. cerevisiae is used, the sequences represented in FIGS. 10 to 17 may also be used to identify (or disrupt) a gene equivalent to a S. cerevisiae PMT gene. Persons skilled in the art will appreciate that fragments of the sequences represented in FIGS. 10 to 17 may similarly be used, as may sequences which are homologous with the sequences represented in FIGS. 10 to 17 and the fragments thereof.

(28) Methods for carrying out gene disruptions are described in the literature, an example of which is described by Boehm et al. (Boehm, T., Pirie-Shepherd, S., Trinh, L., Shiloach, J. and Folkman, J. 1999) Yeast 15 563-572) which describes the use of the Saccharomyces cerevisiae SUC2 gene as a marker flanked by Pichia pastoris DNA specific to the target gene. In the example of Pichia pastoris disruption, the SUC2 DNA sequence could be inserted at a position within any of the DNA sequences represented in FIGS. 10 to 17.

(29) The yeast will advantageously have a deletion of the HSP150 and/or YAP3 genes as taught respectively in WO 95/33833 and WO 95/23857.

(30) In a preferred embodiment the yeast is transformed with an expression plasmid based on the Saccharomyces cerevisiae 2 m plasmid. At the time of transforming the yeast, the plasmid contains bacterial replication and selection sequences, which are excised, following transformation, by an internal recombination event in accordance with the teaching of EP 286 424. The plasmid also contains an expression cassette comprising: a yeast promoter (eg the Saccharomyces cerevisiae PRB1 promoter), as taught in EP 431 880; a sequence encoding a secretion leader which, for example, comprises most of the natural HSA secretion leader, plus a small portion of the S. cerevisiae -mating factor secretion leader as taught in WO 90/01063; the HSA coding sequence, obtainable by known methods for isolating cDNA corresponding to human genes, and also disclosed in, for example, EP 73 646 and EP 286 424; and a transcription terminator, preferably the terminator from Saccharomyces ADH1, as taught in EP 60 057. Preferably, the vector incorporates at least two translation stop codons.

(31) The choice of various elements of the plasmid described above is not thought to be directly relevant to the purity of the albumin product obtained, although the elements may contribute to an improved yield of product. A preferred embodiment of the fermentation and purification process is described in Example 1.

EXAMPLE 1

(32) The cloning strategy for construction of the albumin-producing micro-organism was as disclosed in EP 431 880 except that the 3 end of the albumin coding sequences and its junction with the ADH1 transcription termination sequence were altered such that the ADH coding sequence was eliminated and such that two consecutive in-frame translation stop codons were present, followed by a third stop codon downstream, as follows:

(33) TABLE-US-00001 ...LGLstopstopAstop ...TTAGGCTTATAATAAGCTTAA... .

(34) This was achieved by modification of the ADH1 terminator from plasmid pAYE309, described in EP 431 880, by PCR mutagenesis using two single stranded oligonucleotides, JMADH1 and JMADH2 with the sequences:

(35) TABLE-US-00002 JMADH1 HindIII _______ 5-GCATAAGCTTTGGACTTCTTCGCCAGAGGTTTGGTCAAG-3 JMADH2 NotIBamHI ___________ 3-TGGACAACATTAGCAAGAAGGTGTGCCTAGCGCCGGCGCCTAG GTACG-5

(36) The PCR conditions were 25 cycles of 94 C. for 60 seconds, 37 C. for 120 seconds and 72 C. for 180 seconds. The 0.48 kb PCR product was digested with both HindIII and BamHI and ligated into plasmid pBST+, described in WO 97/24445, similarly digested with HindIII and BamHI, to create plasmid pAYE440 (FIG. 2). The ADH1 terminator was further modified by PCR mutagenesis using two single stranded oligonucleotides, AT19R and the universal 40 primer with the sequences:

(37) TABLE-US-00003 AT19R HindIII ______ 5-AGTCCAAGCTTAATTCTTATGATTTATGAT-3 40 3-CAGCACTGACCCTTTTG-5.

(38) The PCR conditions were 25 cycles of 94 C. for 30 seconds, 50 C. for 40 seconds and 72 C. for 50 seconds and then one cycle of 72 C. for 10 minutes, using the ADH1 terminator in pAYE440 as a template (FIG. 2). The machine used was a Perkin Elmer GeneAmp PCR System 9600. A product of the correct size, approximately 0.33 kb, was obtained and digested with both HindIII and BamHI. Plasmid pAYE309, described in EP 431 880, was digested with NotI and HindIII and the 0.84 kb DNA fragment containing the PRBI promoter fragment and part of the HSA/MF-1 leader sequence (WO 90/01063) employed to direct secretion of mature HSA, was ligated into NotI and HindIII digested pBST+, described in WO 97/24445, to generate plasmid pAYE438 (FIG. 3). The recipient plasmid pAYE438 was digested with HindIII and BamHI and the modified ADH1 terminator was successfully cloned into this vector to generate plasmid pDB2241 (FIG. 4). This plasmid contains the pBST+(WO 97/24445) backbone, the PRB1 promoter and the modified ADH1 terminator.

(39) To facilitate the introduction of two translation stop codons at the end of the HSA coding region and create the required HindIII site, the 3N end of the HSA coding region was altered.

(40) The double stranded oligonucleotide linker, AT21/AT22 was ligated into AflII/HindIII cut pDB2241 and comprised an AflII site at its 5N end, a stuffer region and then the Bsu361 to HindIII sequence of the HSA coding DNA, but with the addition of an extra TAA translation stop codon. Clones with the linker inserted were checked by DNA sequencing and the correct plasmid designated pDB2242 (FIG. 5).

(41) TABLE-US-00004 LinkerAT21/22 AT21 AflIIBsu36IHindIII ________________________ TTAAGAGTCCAAGCCTTAGGCTTATAATA CTCAGGTTCGGAATCCGAATATTATTCGA ALGLStopStop

(42) To create the final rHA expression cassette the AflII/Bsu361 fragment of pAYE309 (FIG. 1) was ligated into AflII/Bsu361 digested pDB2242, making plasmid pDB2243 (FIG. 6). Finally, the rHA expression disintegration vector was made by ligating the NotI expression cassette from pDB2243 into NotI cut pSAC35 (Sleep et al, 1991, Bio/Technology 9, 183-187 and EP 431 880) to generate the plasmid pDB2244 (FIG. 7) in which the direction of rHA transcription is in the same orientation as that of the LEU2 gene.

(43) The plasmid pDB2244 is therefore derived from the disintegration vector pSAC3 (Chinery and Hinchliffe (1989) Current Genetics 16, 21-25) and comprises the whole of the 2 m plasmid, the LEU2 gene to complement the host leu2 mutations, the expression cassette in which the PRB1 promoter drives expression of the HSA sequence and the bacterial plasmid pUC9. The latter is excised from the plasmid by the S. cerevisiae 2 m FLP recombinase system such that no bacterial DNA is present in the organism used for production of rHA (Chinery and Hinchliffe, op cit.).

(44) The expression vector utilises the S. cerevisiae PRB1 promoter and ADH1 transcription terminator to control expression and the HSA/MF-1 leader sequence (WO 90/01063) to direct secretion of mature HSA.

(45) The plasmid pDB2244 was introduced into a Saccharomyces cerevisiae strain which was leu2, yap3, hsp150, pmt1 [cir.sup.] by the method described by Hinnen et al, (1978) P.N.A.S. 75, 1929. The pmt1 mutation may be achieved by the method of WO 94/04687.

(46) Transformants were selected on a buffered minimal medium (0.15% (w/v) yeast nitrogen base without amino acids and ammonium sulphate (Difco), 0.5% (w/v) ammonium sulphate, 0.1M citric acid/Na.sub.2HPO.sub.4.12H.sub.2O pH6.5, 2% (w/v) sucrose)) lacking leucine. When transformants were grown for 72 hours at 30 C., 200 rpm in 50 ml flasks containing either 10 ml of complex (YEP, 1% (w/v) yeast extract, 2% (w/v) bactopeptone and 2% (w/v) sucrose), or buffered minimal medium liquid medium, rHA could be detected in the cell free culture supernatant by SDS-polyacrylamide gel electrophoresis and/or by rocket gel immunoelectrophoresis.

(47) A stock master cell culture in buffered minimal medium is used to prepare running stocks (working cell bank) of process yeast suitable for the preparation of shake flask cultures by freezing aliquots of the culture in the presence of 20% (w/v) trehalose.

(48) The fermentation was essentially the same as is described in WO 96/37515 and U.S. Pat. No. 5,728,553, both of which are incorporated herein by reference, except for the following differences:

(49) Seed Fermentation

(50) After the medium for rHA production has been added to the seed fermenter vessel, the operating temperature of 30 C. is set, as well as the minimum stirrer speed set to achieve homogeneity and so avoid gradients of nutrients such as oxygen or carbon. The initial pH is adjusted with ammonia solution (specific gravity 0.901) using a pH controller set at 6.40; controlled at 6.400.10.

(51) Alternatively, pH is maintained in the range of 5.50 to 5.90, with the lower control set point being 5.50. The initial pH may be adjusted with ammonia (eg aqueous ammonia specific gravity 0.880). This lower fermentation pH results in an enhanced mass spectrometry profile of the rHA.

(52) It is preferable for the initial pH to be near the top of the aforementioned ranges to facilitate observation of early metabolism, since a decline in pH is the first sign of growth detectable by on-line instruments.

(53) Particularly for strains with a deficiency in one or more of the PMT genes, it has been found to be beneficial for the fermentation to be conducted at a higher pH than is normally required. Thus, rather than control the pH at approximately 5.5, it is beneficial to have a control set point between pH6.20 and pH6.70, preferably between pH6.3 and 6.5. At such a higher pH, the quality of the centrate is significantly improved due to reduced cell lysis. Cell lysis results in cell debris remaining in suspension following a centrifugation step of the fermentation which is sufficient only to remove all whole cells from the supernatant. This is demonstrated in Table 1, where a significant reduction in the wet weight content of a culture supernatant is shown when the yeast is cultured in the pH range 6.3 to 6.5 compared to pH5.5.

(54) TABLE-US-00005 TABLE 1 Relationship between centrate quality and fermentation pH in seed fermenter vessel. Values in parentheses are standard deviation and number of samples. Wet Weight Content of Supernatant Fermentation pH (g .Math. L.sup.1) 5.5 9.9 (2.4, 6) 6.3-6.5 3.4 (1.0, 13)

(55) 2M H.sub.2SO.sub.4 is also used as a pH corrective agent. Sucrose to 20 g.Math.L.sup.1, MW10 batch vitamins, and Breox FMT30 antifoam to 0.04 g.Math.L.sup.1 are added to the vessel.

(56) Sterile filtered air is introduced into the vessel at 0.5 v/v/m (ie 0.5 litre non-compressed air per litre of medium per minute), the medium is inoculated to >10 mg cell dry weight L.sup.1 from axenic shake flask culture and a supervisory computer control system is initiated. The expected batch phase is 6210 h from an inoculum concentration of 12 mg.Math.L.sup.1. MW10 feed must be connected before the end of the batch phase (volume equal to batch volume).

(57) Features of the fermentation control algorithm include: the end of batch phase being signalled by dissolved oxygen tension (DOT) increase of >15% in 30 min; the feed being initiated at 0.05 ml per litre batch medium; the substrate feed rate being determined according to the formula, SF=SF.sub.oe.sup.k, wherein SF is substrate feed rate (mL.Math.min.sup.1); SF.sub.0 is initial substrate feed rate (mL.Math.min.sup.1), L is specific growth rate (for example 0.06 h.sup.1), and k is a counter variable started at 0 and increased by 0.0167 once every 1 min if all conditions are met.; and the substrate feed rate (via manipulation of k) being reduced in response to DOT<15% and/or respiratory quotient (RQ)1.2.

(58) The feed is stopped if the pH<6.2 or if the temperature <29.0 C. or >31.0 C. This may also be done automatically through the control algorithm. The SF is reduced if the average RQ>1.13 over a 2 h period, or if there is evidence of ethanol or acetate accumulation.

(59) Agitation is increased to maintain DOT>20% air saturation. Once the feed is started, the concentration of Breox FMT30 is increased to 0.3 g.Math.L.sup.1 (calculated on final volume). The expected feed phase duration is 6517 h, dependent upon transfer limitations of the vessel.

(60) The air flow is increased through the fermentation to maintain the values of oxygen uptake rate (OUR) and carbon dioxide evolution rate (CER), at levels sufficient to provide accurate gas analysis. The air flow rate of the fermentation is nominally 1 v/v/m. Daily checks are performed to determine purity of culture and CDW. Appropriate samples are retained. At the end of the feed, the culture is transferred to a production vessel.

(61) Production Fermentation

(62) The production fermenter is inoculated at 0.25-1.00 g.Math.cdw.Math.L.sup.1. The initial pH is adjusted with ammonia solution (SG 0.901) using a pH controller set at pH6.40; controlled at 6.400.10.

(63) Alternatively, pH is maintained in the range of 5.50 to 5.90, with the lower control set point being 5.50. The initial pH may be adjusted with ammonia (eg aqueous ammonia specific gravity 0.880). This lower fermenation pH results in an enhanced mass spectrometry profile of the rHA.

(64) It is preferable for the initial pH to be near the top of the aforementioned ranges to facilitate observation of early metabolism, since a decline in pH is the first sign of growth detectable by on-line instruments.

(65) Particularly for strains with a deficiency in one or more of the PMT genes, it has been found to be beneficial for the fermentation to be conducted at a higher pH than is normally required. Thus, rather than control the pH at approximately 5.5, it is beneficial to have a control set point between pH6.20 and pH6.70, preferably between pH6.3 and 6.5. At such a higher pH, the quality of the centrate is significantly improved due to reduced cell lysis. Cell lysis results in cell debris remaining in suspension following a centrifugation step of the fermentation which is sufficient only to remove all whole cells from the supernatant. This is demonstrated in Table 2, where a significant reduction in the wet weight content of a culture supernatant is shown when the yeast is cultured at pH 6.5 compared to pH 5.5.

(66) TABLE-US-00006 TABLE 2 Relationship between centrate quality and fermentation pH in production vessel. Wet Weight Content of Supernatant Fermentation pH (g .Math. L.sup.1) 5.5 36.3 6.5 4.7

(67) 2M H.sub.2SO.sub.4 is also used as a pH corrective agent. Sucrose to 20 g.Math.L.sup.1, MW10 batch vitamins, and Breox FMT30 antifoam to 0.04 g.Math.L.sup.1 are added to the vessel.

(68) The initial substrate feed rate is determined according to the formula:

(69) ${\mathrm{SF}}_o=\frac{1000\left[\mathrm{CDW}\right]V_{\mathrm{batch}}}{60Y_{x\text{/}s}\left[\mathrm{sucrose}\right]}$
wherein SF.sub.0 is initial substrate feed rate, p is specific growth rate (for example 0.06 h.sup.1), V.sub.batch is batch volume (L), Y.sub.x/s is cell yield (g.Math.L.sup.1), [sucrose] is sucrose concentration (g.Math.L.sup.1) and [CDW] is cell dry weight concentration (g.Math.L.sup.1). The substrate feed rate is determined according to the formula, SF=SF.sub.oe.sup.k, wherein SF is substrate feed rate (mL.Math.min.sup.1); SF.sub.0 is initial substrate feed rate (mL.Math.min.sup.1), is specific growth rate (h.sup.1) (for example 0.06 h.sup.1), and k is a counter variable started at 0 and increased by 0.0167 once every 1 min if all conditions are met. A number of conditions are constantly reviewed during the fermentation, and used to adjust SF via manipulation of k; SF is reduced in response to DOT<15% and/or respiratory quotient (RQ)>1.2. The feed is stopped if the pH<6.2 or if the temperature <29.0 C. or >31.0 C. This may also be done automatically through the control algorithm. The SF is reduced if the average RQ>1.13 over a 2 h period, or if there is evidence of ethanol or acetate accumulation.

(70) Agitation increased to maintain DOT20% air saturation, and maintained at a maximum once attained in order to facilitate mixing. Once the feed is started and the culture is under carbon limitation, the concentration of Breox FMT30 is increased to 0.2-0.32 g.Math.L.sup.1 (calculated on final volume). The expected feed phase duration is dependant upon transfer limitations of the vessel, typically 90-120 h at the 8,000 L scale.

(71) The air flow is increased incrementally through the fermentation to maintain the values of oxygen uptake rate (OUR) and carbon dioxide evolution rate (CER), at levels sufficient to provide accurate gas analysis. The vessel is overpressured as necessary to enhance OTR. The air flow rate of the fermentation is nominally 1 v/v/m. Daily checks can be performed to determine purity of culture and CDW, and appropriate samples are retained.

(72) The culture is held for downstream processing at the end of the feed.

(73) Hold of Production Culture

(74) The production culture may be held under appropriate conditions to enable batch processing of the culture. The hold time should be kept to a minimum, but can be extended up to 48 hours and beyond if necessary (eg to 5 days). It will be appreciated that, under conditions of batch processing, the constraints of hold time as expressed herein apply to the final portion of the culture to be processed.

(75) The centrate from the fermentation, or an impure albumin solution from any other source (such as plasma), is prepared, or conditioned, for chromatography on a cation exchange matrix while protecting the rHA from polymerisation and protease activity. Preferably, sodium octanoate is added (Chromatography Solution 14 (CS14)Table 3) to a final concentration of 1-10 mM, for example approximately 5 mM. The pH is adjusted with acetic acid to pH4.3-4.8, preferably 4.500.1 (most preferably 0.05) and the conductivity is checked to be <5.5 mScm.sup.1.

(76) Chromatography

(77) All operations can be carried out at ambient temperature (205 C.). The albumin loads (g/L) for the chromatography columns are determined from titres of albumin (g/L) by either SDS-PAGE (at the first step) or GP-HPLC (for all other columns). The progress of each step is monitored by measuring UV absorbance on line, for example at 254 or 280 nm.

(78) In a particularly preferred embodiment of the present invention the purification process comprises the following steps: cation exchange chromatography (SP-FF); anion exchange chromatography (DE-FF); affinity chromatography (DBA); ultrafiltration and diafiltration; a second affinity chromatography step (PBA); ultrafiltration and diafiltration; a second cation exchange chromatography step (SP-FF2); and a second anion exchange chromatography step (DE-FF2). Preferably, these purification processes are followed by final ultrafiltration/diafiltration followed by a formulation step, and/or placing of the solution into a final container.

(79) The sequence of chromatographic steps as described here is novel and inventive in a number of aspects. The use of an aminophenylboronate (PBA) matrix with an improved buffer, as described herein, and a small load volume has been shown to give increased yeast antigen clearance, as measured by ELISA (about 4-20 fold). The buffer used with the aminophenylboronate matrix was unexpectedly found to be particularly beneficial, and it represents the result of intensive trials of a plethora of buffers of widely ranging constituents and properties. The buffer provides a significantly increased clearance of yeast antigens, when compared with the buffer used in the PBA chromatography step of WO 96/37515.

(80) Loading the aminophenylboronate matrix with a highly concentrated albumin solution, for example 10010 g.Math.L.sup.1, means that improved resolution of the rHA and yeast antigens can be achieved because of the smaller load volume.

(81) WO 96/37515 includes a S200 gel permeation step after a first affinity chromatography step. The gel filtration step purified the albumin with respect to yeast antigens, pigment and dimerised albumin. We have found that this step is no longer necessary because of the improvements we have made to the aminophenylboronate affinity step and the introduction of additional cation and anion exchange steps.

(82) Following the aminophenylboronate affinity step it is preferred that the albumin is concentrated and diafiltered for a negative mode cation exchange step. We have found that the combination of this diafiltration step and cation exchange step substantially reduces the relative concentration of nickel ions. In particular, exposing rHA to a low pH is effective in reducing nickel ion levels. Consequently, albumin purified according to the present invention has a surprisingly low nickel ion content (less than 100 ng/g of albumin).

(83) The negative mode cation exchange step, as described herein, is used to remove Concanavalin A binding material (cbm) which is a small amount of modified rHA, thought to be glycosylated. The negative mode cation exchange step has been found to reduce the cbm content produced by recombinant pmt1-mutant Saccharomyces cerevisiae by approximately 1.3-fold. A greater effect is achieved with rHA derived from non-pmt1 mutants (2-3 fold clearance).

(84) In comparison with other commercial yeasts, Saccharomyces cerevisiae produces a relatively low level of modified rHA. Accordingly, the negative mode cation exchange step and the use of cells with a deficiency in one or more of the PMT genes may be of even greater importance if the rHA is produced by a recombinant host other than Saccharomyces cerevisiae.

(85) The chromatography solutions used during the purification of albumin are detailed in Table 3. Because of the very large scale manufacture of albumin, and the relatively low cost of the product, these buffer salts are the most suitable for the process as they are available in a highly pure form at industrial scale and are low cost compared to other commonly used buffers such as Tris, HEPES or MOPS. Alternative buffers could be used in place of the ones used in Table 3, for example buffers of a similar pK.sub.a (eg malate for acetate), but in most instances cost and availability at large scale rule out their use. Alternative salt forms can be used provided they are soluble, available at industrial scale and low cost.

(86) Chromatography can be performed using either axial flow columns, such as those available from Pharmacia, or using radial flow columns, such as those available from Sepragen. A fluidised bed may be used, at least for the first step.

(87) The buffer solutions can be prepared at the concentrations described below, or concentrated stock solutions can be prepared and mixed or diluted on-line for immediate use.

(88) Cation Exchange Chromatography

(89) Albumin is concentrated and purified with respect to at least yeast proteins (if the albumin is rHA from a yeast fermentation) and other antigens, low molecular weight contaminants and pigmented compounds by cation exchange chromatography. The method uses a commercial cation exchange matrix such SP-Sepharose FF, SP-Spherosil, CM-Sepharose FF, CM-Cellulose, SE-Cellulose or S-Spherodex. Preferably, the matrix is SP-Sepharose FF (Pharmacia) which, if used in an axial flow column, may be at a bed height of 5 to 25 cm, preferably 10 to 15 cm, for example 12.5 cm. If a radial flow-type column is used, a suitable bed flow path length is 11.01.0 cm. A column loading of 10 to 50 g albumin/L, preferably 4010 g albumin/L, of matrix is suitable. The matrix is equilibrated with a buffer to remove the alkali storage solution; preferably the buffer should be strong enough to reduce the pH to approximately pH6.0. A buffer such as CS01 is used to remove storage solution CS07 from the column; however, any buffer with a pH<6.0 could be used. Equilibration is judged to be complete when the pH of the column effluent is approximately pH6.0.

(90) TABLE-US-00007 TABLE 3 Chromatography solutions for the purification of albumin Solution Concentration Conductivity No. Name Constituent (g .Math. L.sup.1) pH (mS .Math. cm.sup.1) CS01 SP-FF CH.sub.3COOH 1.85 5.45-5.65 1.9-2.2 Equilibrant/Wash3/ NaOH (27% 4.00 DE-FF Equilibrant (w/w)) CS02 SP-FF Wash 1 CH.sub.3COOH 3.00 3.9-4.1 0.6-0.8 NaOH (27% 1.19 (w/w)) CS03 SP-FF Wash 2 CH.sub.3COOH 1.62 3.9-4.1 125-165 NaOH (27% 1.19 (w/w)) NaCl 117 CS04 SP-FF Eluent/ CH.sub.3COOH 5.13 5.4-5.6 5.0-6.0 DE-FF Pre- NaOH (27% 11.5 Equilibrant (w/w)) Octanoic Acid 0.721 CS05 Salt Clean NaCl 58.4 5-9 75-95 Polysorbate 80 5.00 CS06 0.5M NaOH NaOH (27% 74.1 >12 80-120 (w/w)) CS07 20 mM NaOH NaOH (27% 2.96 >12 3.5-5.5 (w/w)) CS08 DE-FF Wash K.sub.2B.sub.4O.sub.74H.sub.2O 4.80 9.0-9.4 2.5-3.5 CS09 DE-FF Eluent K.sub.2B.sub.4O.sub.74H.sub.2O 33.6 9.2-9.5 15.0-18.0 CS10 DBA CH.sub.3COONH.sub.4 19.3 8.7-9.1 18-22 Equilibrant/Wash NaOH (27% 5.93 (w/w)) CS11 DBA Eluent NaCl 117 6.7-7.1 125-165 NaOH (27% 14.1 (w/w)) H.sub.3PO.sub.4 5.79 (85%(w/w)) CS14 2M Sodium NaOH (27% 281 7.8-8.4 Octanoate (w/w)) Octanoic Acid 288 CS15 Acetic Acid CH.sub.3COOH 1045 CS17 DE-FF2 CH.sub.3COOH 1.50 4.5-4.7 0.85-1.05 Equilibration/Wash NaOH (27% w/w) 1.66 CS18 Positive-mode NaH.sub.2PO.sub.42H2O 8.58 6.8-7.0 5.5-6.5 DE-FF2 Elution NaOH (27% w/w) 4.07 CS19 SP-FF2 CH.sub.3COOH 1.80 5.2-5.4 1.8-2.1 Equilibration/Wash NaOH (27% w/w) 3.52 CS20 PBA Glycine 7.51 8.3-8.6 18-22 Equilibration/Wash NaCl 5.84 NaOH (27% w/w) 0.95 CaCl.sub.22H.sub.2O 7.35 CS21 20% (w/w) Acetic CH.sub.3COOH 205 1.9-2.2 1.8-2.0 Acid H.sub.2O 820 CS22 Final pH Adjustment Na.sub.2HPO.sub.4 71.0 11.2-11.4 43-49 NaOH (27% w/w) 37.0 EXO4 Final pH adjustment NaOH (47% w/w) 42.6 12 80-120 alkali H.sub.2O 970 EXO5 Final pH adjustment HCl (37% w/w) 19.7 1.5 60-90 acid H.sub.2O 982 All weighings are 2%, for this particular example.

(91) The centrate from a fermentation is prepared, or conditioned, for chromatography on a cation exchange matrix while protecting the rHA from polymerisation and protease activity. However if the yeast strain is not deficient in the proteases that degrade rHA at the pH required to purify the rHA then the culture supernatant should be pasteurised, for example by a heat treatment of 50-70 C. for 30 minutes to 5 hours, as detailed in WO 94/03636. Typically 1-10 mM sodium octanoate is sufficient to protect the rHA from heat denaturation and 30 seconds up to 10 minutes at temperatures of 60-80 C. adequate to inactivate the proteases in a batch or flowthrough procedure. Pasteurisation may also be desirable if HSA is used.

(92) The conditioned centrate is then loaded onto the column at a flow rate of, for example, 0.07-0.75 bed volumes/min, preferably 0.3-0.6 bed volumes/min, in this example 0.5 bed volumes/min, and then the column is washed with one or more solutions to remove residual contaminants. The column is washed first with, for instance, eight volumes of 10-100 mM, preferably 30-70 mM, for example 50 mM acetate, pH3.9-4.1, 0.6-0.8 mS.Math.cm.sup.1 (CS02). The column is then washed with four volumes of a high salt buffer containing 1-3M NaCl, preferably 2M NaCl, in sodium acetate buffer (for example 10-50 mM sodium acetate, preferably about 27 mM, pH3.5-4.5, preferably pH4.0 (CS03) and then ten volumes of CS01. The albumin is eluted with, and collected in an acetate/octanoate buffer (for example 40-120, preferably 60-100, eg 85 mM acetate, and 2-50 mM, preferably, 2-20 mM, eg 5 mM octanoate, as in CS04). The collection of albumin starts when the UV signal rises above 0.6 A.sub.254/cm, and collection continues until the UV signal falls below 0.36 A.sub.254/cm. The column is then cleaned using 0.25-3.0M NaCl and 0.05-2% detergent (CS05) and then 0.1-1.0M NaOH(CS06), then stored in diluted (10-50 mM) NaOH(CS07). In this example, the flow rate for the equilibration, loading and wash steps is 0.5 bed volumes per minute. For elution of the albumin, a flow rate of 0.04-0.6 bed vol/min, preferably 0.15-0.35, in this example 0.25 bed vol/min is used.

(93) Anion Exchange chromatography

(94) The eluate from the cation exchanger is then diluted to below 10 mS.Math.cm.sup.1, preferably less than 5 mS.Math.cm.sup.1, especially below 2.5 mS.Math.cm.sup.1 and then loaded onto an anion exchange resin such as QMA-Spherosil, DEAE-Spherodex, Q-Hyper D, DEAE-cellulose, QAE-cellulose, or TMAE, DMAE, or DEAE Fractogel. Preferably, the matrix is the commercial anion exchange matrix DEAE Sepharose-FF (Pharmacia), bed flow path length of 11.01.0 cm, pre-equilibrated with the cation elution buffer (CS04) and then equilibrated with three column volumes of CS01. The albumin is loaded onto the matrix at 3010 g monomeric albumin per litre of matrix and then the matrix is washed with dilute tetraborate buffer, for example 15-25 mM potassium tetraborate or sodium tetraborate (CS08), which has the effect of raising the pH to approximately 9.2, and then the albumin is eluted with a more concentrated tetraborate buffer (for example 80-150 mM potassium tetraborate, preferably 110 mM potassium tetraborate (CS09)). The matrix is cleaned with salt/detergent (CS05) and then NaOH(CS06) before storage in dilute NaOH(CS07). The eluate from the anion exchange matrix is then loaded onto an affinity matrix.

(95) Affinity Chromatography

(96) This step further purifies the rHA with respect to a 45 kDa N-terminal albumin fragment, yeast antigens and pigment. The affinity matrix may comprise any Cibacron Blue type of dye which binds albumin, for example Reactive Blue 2, Procion Blue HB, Blue Sepharose, Blue Trisacryl and other anthraquinone-type compounds. Preferably, the matrix is the Delta Blue Matrix (DBA), prepared as described in WO 96/37515.

(97) The method uses DBA at a bed flow path length of 11.01.0 cm. The DBA is equilibrated in ammonium acetate buffer (100-300 mM, preferably 200-275 mM, for example 250 mM as in CS10) and the albumin applied at 7.0-14.0 g/L, preferably 8.0-12.0 g/L, in this example 10.01.0 g/L. Equilibration, load and wash steps are performed at flow rates of 0.05-0.30 bed vol/min, preferably 0.15-0.27, in this example 0.25 bed vol/min. All other steps are performed at 0.20 bed vol/min. When loading is complete, the column is washed to remove contaminants with 1-5 volumes of ammonium acetate buffer 10-30 mS cm.sup.1, preferably 15-25 mS cm.sup.1, for example CS10, preferably 5 column volumes. The albumin is eluted with a strong salt and phosphate solution (1.0-3.0M NaCl, for example 1.5-2.5M NaCl or 2.0M NaCl, and 5-100 mM, eg 50 mM phosphate, as in CS11. The column is then cleaned using CS06 and stored in CS07.

(98) The eluate from the DBA column is then concentrated and diafiltered in preparation for purification using phenyl boronate agarose (PBA) chromatography. DBA ultrafiltration can be performed with any ultrafiltration membrane used in protein concentration with a nominal molecular weight cut off of 30,000 or less, preferably a polyethersulphone type membrane (eg Filtron Omega series) of 10,000 nominal molecular weight cut off. DBA eluate is concentrated and then diafiltered at 100 g rHA.Math.L.sup.1 against at least 5 volumes of water followed by at least 5 volumes of CS20. At the end of diafiltration, the retentate may be further concentrated if required and the equipment washed out with CS20 to increase step recovery. The concentration of the final retentate should be in the range 20-120 g rHA.Math.L, preferably 70-120 g.Math.L.sup.1, or as in this example 10010 g rHA.Math.L. After use, the membranes are treated by flushing out residual protein with water, cleaning with CS06 and storage in CS07.

(99) PBA is an affinity step to remove glycoconjugates, such as glycoproteins, glycolipids and poly-, oligo- and monosaccharides, and utilises immobilised aminophenylboronic acid as the ligand. The aminophenylboronic acid is covalently coupled via a spacer to an insoluble matrix such as polyacrylamide, agarose, cellulosic or organic polymers. U.S. Pat. No. 4,562,251 (incorporated herein by reference) describes suitable methods for making diborotriazine or monoborotriazine agarose: (1) triazine is O-linked to agarose first and then linked with 3-aminophenylboronic acid (APBA) in a second reaction. (2) Triazine is reacted with APBA first to produce either mono or diborotriazine. These are then O-linked via the free chlorine on the triazine to the ONa activated agarose to produce either mono or disubstituted agarose.

(100) An earlier patent, U.S. Pat. No. 4,269,605, contemplates a variety of activation methods, including epichlorohydrin activation of agarose, preferred herein. Commercially available matrices include Amicon's PBA30 and Sigma's acrylic beaded aminophenylboronate.

(101) It has been found to be particularly beneficial to use a buffer containing glycine (10-500 mM, for example 25-200 mM, preferably 50-150 mM, in this example 100 mM), NaCl (0-500 mM, for example 25-200 mM, preferably 50-150 mM, in this example 100 mM) and CaCl.sub.2 (5-250 mM, preferably 10-100 mM, in this example 50 mM), pH8.0-9.5, preferably, pH 8.0-9.0, in this example pH8.5 (CS20).

(102) The PBA column uses a flow path length of 11.01.0 cm and is pre-equilibrated with the buffer as described above, eg CS20. The column is loaded at less than 1 column volume, preferably less than 0.5 column volumes, in this example 50.35 column volumes. The PBA is run as a negative step and therefore the albumin is collected in the flow through and wash from the column. All chromatographic steps can be performed at flow rates of 0.005-0.3 bed vol./min. Preferably the equilibration and cleaning of the column are carried out at a higher flow rate, eg 0.19 bed vol./min, than the load and collection of the albumin solution, which is preferably carried out at a flow rate of 0.01-0.05, preferably 0.025 bed vol./min. The column is then cleaned with salt (CS03), borate buffer (CS09), NaOH(CS06) and then stored in dilute NaOH(CS07).

(103) Following PBA chromatography the albumin solution is concentrated and diafiltered to prepare for a negative mode cation exchange step. The combination of this diafiltration step and the negative mode cation exchange chromatography substantially reduces the relative concentration of nickel ions.

(104) PBA ultrafiltration can be performed with any ultrafiltration membrane used in protein concentration with a nominal molecular weight cut off of 30,000 or less, preferably a polyethersulphone type membrane (eg Filtron Omega series) of 10,000 nominal molecular weight cut off. The collected PBA Flow Through is adjusted to pH5.30.5 with CS21, concentrated and then diafiltered at 100 g rHA.Math.L.sup.1 against at least 7 volumes of CS19. At the end of diafiltration, the equipment is washed out with CS19 and further CS19 added as required to give a retentate concentration of 5010 g rHA.Math.L.sup.1. Finally, sodium octanoate is added to give a final concentration of approximately 2-15 preferably 5-10, more preferably 6-9, and in this example 6 mM, eg CS14 is added to 3 mL.Math.L.sup.1. After use, the membranes are treated by flushing out residual protein with water, cleaning with CS06 and storage in CS07.

(105) The albumin solution is then subjected to a second cation exchange step using, for instance, SP-FF Sepharose (Pharmacia), this time in the negative mode, ie the albumin passes through the matrix, rather than being retained. The conditions are such that mannosylated albumin binds to the matrix. The buffer is preferably a sodium acetate buffer (5-110 mM acetate, preferably 10-50 mM, in this example 30 mM), pH 5.2-5.4, CS19). Other buffers which can buffer in the appropriate range may be used, such as a citrate phosphate buffer. Suitably, the buffer has a conductivity of about 2 mS.Math.cm.sup.1. The column has a flow path length of 11.01.0 cm, with the albumin loaded to 10-250 g.Math.L.sup.1 preferably 20-70 g.Math.L.sup.1 and in this example 5015 g or 5010 g.Math.L.sup.1 matrix. Since this is a negative step, the albumin is collected in the flow through and wash.

(106) Following this cation exchange step, the albumin is subject to negative mode anion exchange chromatography. This step removes yeast antigens as measured by ELISA and Western blot. The collected flow through and wash from the second cation exchange step is adjusted to pH4.600.10 with CS21, diluted to 1.050.1 mS.Math.cm.sup.1 with water and the rHA purified using the following conditions. The step uses an anion exchange matrix such as DE-FF Sepharose (Pharmacia) at a flow path length of 11.01.0 cm and the albumin is loaded to 50-250 g.Math.L.sup.1, preferably 15050 g.Math.L.sup.1 matrix. Since this is a negative step, the albumin is collected in the flow through and wash. The pH of the Flow Through and Wash is then adjusted to 7.00.1 with CS22.

(107) Alternatively, as described in Example 9, pH-adjustment may occur in the Final UF feed vessel instead of being performed on the DEAE flow through and wash.

(108) While Example 1 has been illustrated with reference to a pmt1 mutant, it should be appreciated that the purification process of the present invention is equally applicable to host cells which are not mutant at this locus, or indeed which are not mutant at any other pmt locus.

EXAMPLE 2

(109) Two assays were used to investigate centrate quality. The poorer the centrate quality the worse the robustness of the yeast cells.

(110) The two assays were:

(111) 1. Determination of the absorbance of centrate at 600 nm (A.sub.600).

(112) 2. Determination of the wet weight of particles in the centrate (WW).

(113) In both the assays, the higher the value the poorer the centrate quality.

(114) The centrate quality of three different yeast strains under two different pH conditions grown in fed-batch fermentation were compared.

(115) TABLE-US-00008 TABLE 4 A600 and WW values for three different rHA producing strains in fed-batch fermentation grown at two different pH values. In the first column the specific gene deletions are indicated. Values in parentheses are standard deviation and number of samples. Specific gene deletions A.sub.600 WW (g .Math. L.sup.1 centrate) Grown at pH 5.5 pmt1-/hsp150-/yap3- 1.39 (0.52, 24) 12.4 (4.9, 23) hsp150-/yap3- 1.11 (0.62, 9) 9.1 (2.9, 7) yap3- 0.58 (0.34, 10) 3.9 (2.0, 10) Grown at pH 6.4 or 6.5 pmt1-/hsp150-/yap3- 0.41 (0.17, 6) 2.6 (0.8, 6) hsp150-/yap3- 0.47 (0.19, 8) 4.6 (1.4, 7) yap3- 0.41 (0.08, 6) 2.1 (0.8, 6)

(116) From the Table above it can be concluded that at pH 5.5, the multiply-gene deleted strains yield an inferior centrate, whereas at pH6.4 or 6.5, the deleterious effect of these further gene deletions is avoided.

EXAMPLE 3

(117) This example was performed in the same manner described in Example 1, but utilised a strain which is not pmt1 mutant. This strain was also grown at two different pH control values, and the wet weight content of the centrate determined as described in Example 1. The benefit of growth at the elevated pH control point is also seen for this strain of yeast; demonstrated in Table 5, where a significant reduction in the wet weight content of a culture supernatant is shown when the yeast is cultured in the pH range 6.3 to 6.5 compared to pH5.5.

(118) TABLE-US-00009 TABLE 5 Relationship between centrate quality and fermentation pH for non-pmt1 strain. Values in parentheses are standard deviation and number of samples. Wet Weight Content of Supernatant Fermentation pH (g .Math. L.sup.1) 5.5 10.0 (2.3, 4) 6.3-6.5 4.6 (1.4, 7)

(119) Thus, rather than control the pH at approximately 5.5, it is beneficial to have a control set point between pH6.20 and pH6.70, preferably between pH6.3 and 6.5. At such a higher pH, the quality of the centrate is significantly improved due to reduced cell lysis

EXAMPLE 4

(120) This Example was performed in a similar manner as described in Example 1, with the following differences. The yeast Pichia pastoris, strain GS115 (Invitrogen) was grown using the same conditions and medium as described above, but using a pH controller set at 5.90; controlled at 5.90+0.20, a specific growth rate of 0.10 h.sup.1 with glucose as a carbon source. The batch phase duration was 28 h, and the feed phase duration was 42 h. Recombinant human albumin was added once the feed phase had commenced, providing a final concentration of 3.8 g rHA.Math.L.sup.1 culture at the end of the fermentation. The rHA used to spike the Pichia culture had been purified but not in accordance with the purification process of the invention.

(121) The rHA from the Pichia fed-batch culture medium was then purified in accordance with the purification process described in Example 1.

EXAMPLE 5

(122) This Example describes the analysis of rHA purified from Pichia culture media as described in Example 4.

(123) Immunoassay Data

(124) Immunoassays were performed on: (i) the rHA purified from the Pichia culture media; (ii) the rHA used to spike the culture media; and (iii) on albumin produced by Saccharomyces cerevisiae which had purified in accordance with the present invention.

(125) Western Blot Summary

(126) TABLE-US-00010 Antibody Batch Number Ig9601 Gel Type 4-12% SDSNR NOVEX GELS Milk Type UHT Exposure Time 20 seconds

(127) Ig9601 was raised against a non-albumin producing Saccharomyces cerevisiae strain and thus can be used to detect yeast antigens.

(128) The western blot showed that the yeast antigen profile of the albumin derived from the Pichia culture medium contained fewer and less intense bands than the material used to spike the Pichia fermentation. The Pichia-derived albumin yeast antigen profile was very similar to the Saccharomyces-derived profile.

(129) EUSA Blot Summary

(130) Yeast antigen impurities in the albumin purified from the Pichia culture medium and for the albumin used to spike the Pichia medium were quantified by ELISA using Ig9601.

(131) The yeast antigen content of the albumin purified from the Pichia culture medium was below the detectable limit of the assay (approximately 0.004 g.Math.g.sup.1), and the antigen content for the albumin used to spike the Pichia medium was 0.62 g.Math.g.sup.1.

(132) Con A Binding Material

(133) The Con A assay described in Example 9 was performed on albumin purified from the Pichia culture medium and for the albumin used to spike the Pichia medium. The content of Con A binding material for the former was 0.22% (w/w) and for the latter it was 0.57% (w/w).

(134) The level of Con A binding material in the albumin purified from the Pichia culture medium is similar to that of albumin purified from Saccharomyces cerevisiae in accordance with the invention (see Table 6), when the latter is not produced from a pmt1 mutant.

(135) The purity analyses confirm that the process of the invention can be successfully used to purify albumin from yeast other than Saccharomyces cerevisiae (eg Pichia) and that albumin of similar purity to that purified from Saccharomyces cerevisiae can be obtained.

EXAMPLE 6

(136) In Example 1 a negative mode anion exchange chromatography step (DE-FF2) followed the second cation exchange chromatography step (SP-FF2). In an alternative purification process the second cation exchange chromatography step may be followed by a positive mode anion exchange chromatography step.

(137) From the SP-FF2 eluate at pH5.3 approx. the pH needs to be increased to pH7. There are two means detailed below, pH adjustment and diafiltration. The latter appeared to give a better quality product.

(138) DE-FF2 (A)

(139) SP-FF2 flow through and washings were pH adjusted to pH 7.0 with 0.5 M disodium hydrogen orthophosphate. This material was loaded onto a DEAE under standard positive conditions to give a matrix loading of 40 g rHA.Math.L.sup.1 matrix, the pH and conductivity of the load were 7.0 and 1.29 mS.Math.cm.sup.1 respectively.

(140) DE-FF2(B)

(141) SP-FF2 flow through and washings were diafiltered vs. 10 vol. 10 mM sodium phosphate pH 7.0, concentrated and diluted with buffer to 50 g.Math.L.sup.1 and loaded onto a DEAE under standard positive conditions. The pH and conductivity of the load was 7.0 and 1.43 mS.Math.cm.sup.1 respectively.

(142) The albumin from DE-FF2A/DE-FF2B is suitably eluted by a 45-55 mM sodium phosphate buffer (pH7.0).

EXAMPLE 7

(143) The kinetics of nickel removal from rHA by treatment with low pH were investigated (see FIG. 9). The results showed that between pH 4 and 4.5 both the rate and extent of nickel removal were independent of pH, but that at pH 5 the rate of removal slowed slightly. Both the rate and extent of nickel removal decreased with increasing pH across the range 5-6.5 with little or no removal above pH 6.5.

EXAMPLE 8

(144) Purification of human serum albumin from a sample of cryo-poor plasma paste (Centeon Pharma GmbH) was achieved using the purification process detailed in Example 1.

(145) Recoveries of HSA at each chromatography step were predominantly comparable to the anticipated rHA recovery at the same stage, with the exception of the PBA column. Here, the recoveries were much lower than expected which may have been due to removal of glycated albumin.

EXAMPLE 9

(146) This Example illustrates the concentration, diafiltration and formulation of the highly purified rHA into a suitable product, in this instance 20% (w/v) albumin. This procedure is carried out in two stages, namely final ultrafiltration (UF) and Formulation.

(147) Final UF reduces nickel concentration by diafiltration at low pH and presents rHA in a defined aqueous environment, using water of an appropriate grade.

(148) Final UF begins with transfer of DEAE flow through and wash to the Final UF feed vessel. As described below, the albumin is then concentrated, diafiltered pH adjusted to pH7.0 and further concentrated.

(149) If DE-FF2 is run in positive mode, the DE-FF2 eluate may be used instead of, or in addition to, the DEAE flow through and wash.

(150) Following transfer of the DE-FF2 flow through and wash (or eluate if DE-FF2 is run in positive mode), the rHA-containing process stream is sequentially subjected to primary concentration, diafiltration and secondary concentration phases in an ultrafiltration system fitted with cellulosic membranes with a nominal molecular weight cut off limit of 10,000. The initial concentration step increases the rHA concentration to approximately 100 g.Math.L.sup.1 and is immediately followed by the continuous diafiltration phase where the rHA is diafiltered against at least 5, preferably at least 7 retentate volume equivalents of water-for-injection, preferably a 50 mM salt solution to remove ammonia. Following diafiltration the pH is adjusted to 7.0 and, the secondary concentration phase further increases the rHA concentration to 275-325 g.Math.L.sup.1. At the end of UF the retentate is transferred to the bulk product formulation vessel.

(151) Instead of pH-adjustment being performed on the DEAE flow through and wash, pH adjustment may occur in the Final UF feed vessel, preferably between the diafiltration process and the secondary concentration phase. Preferably, the diafiltration retentate is adjusted to pH 70.1 with EX04. If the pH exceeds 7.1 but remains <pH8.5 then the pH can be decreased with EX05.

(152) The formulation step produces rHA in an appropriate chemical environment and at an appropriate concentration suitable for bulk product sterile filtration and filling. The transferred Final UF retentate is analysed to determine concentrations of albumin, sodium and octanoate. These quantities are taken into account and any necessary further amounts of stock sodium chloride and sodium octanoate excipient solutions and appropriate grade water added to achieve the bulk formulation specification. The final albumin concentration may be 150-250 g.Math.L.sup.1 or 235-265 g.Math.L.sup.1, with a sodium concentration of 130-160 mM. Any other feasible albumin concentration may be made, however, with, for example, a minimum concentration of at least 4% (w/v), preferably 4-25% (w/v). Formulation is complete following addition of appropriate conventional pharmaceutically acceptable excipients, such as polysorbate 80 or those specified in the US Pharmacopoeia for human albumin, and diluting water.

(153) A final concentration of 0.08 mmoles sodium octanoate per gram of albumin may be desirable. The product is sterile and non-pyrogenic. There may be up to 1% dimeric albumin but no larger polymers or aggregates are detectable.

EXAMPLE 10

(154) This Example illustrates the analysis that is carried out to establish the purity of albumin purified in accordance with the present invention. Unless stated otherwise, all of the assays are performed on albumin which has been purified according to Example 1 and formulated according to Example 9.

(155) Glycation of rHA

(156) A microassay for glycated protein has shown that rHA purified in accordance with the invention is not substantially modified by non-enzymic glycosylation (glycation). The microassay measures the stable Amadori product (AP) form of glycated protein, by oxidation of the C-1 hydroxyl groups of AP with periodate. The formaldehyde released by periodate oxidation is quantitated by conversion to a chromophore, diacetyldihydrolutidine (DDL), by reaction with acetylacetone in ammonia. DDL is then detected colorimetrically. The samples were assayed after desalting using a Pharmacia PD-10 (G25 Sephadex) column and then the albumin in the samples was re-quantitated by the Bradford method and 10 mg albumin was assayed. A fructose positive control was included, and the absorbances were read on a Shimadzu UV 2101 spectrophotometer at 412 nm. For every mole of hexose one mole of Amadori product is formed.

(157) TABLE-US-00011 Moles Amadori Sample Product/Moles Albumin A 0.79 B 0.76 C 0.41 D 0.48 E 0.46 F 0.22 G 0.41 H 0.37 I 0.54 J 0.76 K 0.84 L 0.50 M 0.43 N 0.59 O 0.41 P 0.18 Q 0.24 R 0.04

(158) Samples A-Q are commercially available HSA products from US, Europe and Japan (mean=0.490.20). Sample R is rHA purified according to the invention.

(159) Analysis of C-terminus

(160) An important aspect of the quality control of recombinant proteins is the conformation and stability of the pre-determined primary structure. Analysis of the C-terminal tryptic peptide in commercially available HSA and rHA purified according to the invention by N-terminal sequencing and FAB mass spectometry indicated the presence of a truncated peptide, lacking the C-terminal leucine in HSA. The Des-Leu C-terminal tryptic peptide was detected in commercial HSA at approximately 5-10% (not quantitative), but could not be detected in the rHA of the invention, even after 6 months at 30 C. The Des-Leu peptide could not be detected in the HSA 12 weeks at 30 C., and the peak for the full length C-terminal peptide was very diminished compared to the other samples, indicating that perhaps this had undergone further C-terminal degradation.

(161) These results indicate that the rHA, purified in accordance with the invention, has a stable and full length carboxy-terminus, whereas HSA previously available from commercial sources appears to be heterogeneous by comparison.

(162) Nickel Ion Content of rHA Prepared According to the Invention

(163) Measuring Instrument:

(164) SIMAA 6000, Perkin Elmer Furnace: CTT (Constant Temperature Tube) using detection at 232 nm, 2470 C.

(165) Calibration:

(166) The method is based on a three-point calibration (18/30/60 1 g/L standard solutions from Perkin Elmer). After the calibration, a blank of purified water is measured. The control standard is measured after the blank and at the end of each test series (Ni-Standard 20 1 g/L, certified standard from Perkin Elmer).

(167) Sample Preparation:

(168) Each assay is the result of a determination in duplicate which also valid for the calibration and the control standard. Depending on the expected Ni concentration, the sample is diluted in an appropriate ratio to work with a Ni-concentration that is within the calibration range. Samples with a protein concentration of 10% or more have to be diluted at least 1:5 in any case. Dilution is with purified water.

(169) Rinsing solution for the sample capillary: 2 L purified water mixed with 0.5 mL Triton100. Each test series includes a system suitability test.

(170) Requirements: 1. Correlation coefficient of the calibration at least 0.99000. If not, the calibration has to be repeated one time. If the calibration does not comply with the requirement a second time, an error analysis has to be carried out. 2. Characteristic mass measured with the 30 1 g/L-Standard may not exceed the theoretical value of 20 pg/0.0044 A-s by more than 20 percent.

(171) Characteristic Mass m.sub.0:

(172) That amount of the analyte in picogram (pg) that contributes an absorption of 1 percent. An absorption of 1 percent corresponds to 0.0044 A-s (ampere seconds).

(173) $m_0=\frac{\mathrm{volume}\mathrm{Standard}\left(\mathrm{mL}\right)*\mathrm{concentration}\left(\mathrm{mg}\text{/}L\right)*0.0044A\text{-}s}{\mathrm{absorption}\mathrm{sample}*\mathrm{absorption}\mathrm{blank}}$ 3. The measured concentration of the control standard has to be within the confidence range (2 s/3 s criterion).

(174) Calculation:

(175) The measuring instrument calculates the result according to the following term:

(176) $\mathrm{Result}\left(xg\mathrm{Ni}\text{/}L\right)=\left(\frac{A1}{\mathrm{slope}}\frac{A2}{\mathrm{slope}}\right)\text{:}2*V$

(177) A: absorption

(178) slope: slope of the calibration curve (linear regression)

(179) V: dilution

(180) A modifier is not used.

(181) TABLE-US-00012 [Nickel]/[rHA] (g/g) Sample Batch 1 Batch 2 PBA load 0.73 0.74 PBA flow through and wash 0.41 0.43 SP-FF2 load 0.06 0.06 SP-FF2 flow through and wash <0.03 <0.03 DE-FF2 flow through and wash 0.14 0.28
Analysis of Medium and Long Chain Fatty Acids

(182) The fatty acids profiles of albumin according to the invention and commercially available HSA were analysed by acidic solvent extraction and gas chromatography of the free fatty acids using a C17:0 internal standard. No abnormal fatty acids have been detected in the albumin of the invention by this method although the profiles for the rHA and HSA showed significant differences. As expected, both showed large amounts of the added stabiliser, octanoate (C8:0). Apart from this, commercial HSA was characterised by predominantly C16:0, C16:1, C18:0, C18:1 and C18:2 whilst the albumin of the invention contained mainly C10:0 and C12:0 and occasionally C14:0. Further experiments showed that the levels of C10:0 and C12:0 in rHA final product correlated with the levels of these contaminants in the octanoate used for the latter stages of the purification process.

(183) Preferably, the total level of C18 fatty acids does not exceed 1.0% (mole/mole) of the level of octanoate, and preferably does not exceed 0.5% of that level. Moreover, in the albumin of the invention, the level of C18:2, C18:3 and C20 fatty acids is generally undetectable. In commercial HSA, there may typically be about 0.4 moles C18 fatty acids per mole of albumin. In the product of the invention, there are typically no detectable C20 fatty acids and only about 0.02 moles C18 fatty acids per mole of albumin.

(184) SDS Reducing Polyacrylamide Gel Electrophoresis

(185) This assay was performed as described in WO 96/37515. The assay showed that rHA of the invention consists of a single polypeptide chain which when treated with a reducing agent (-mercaptoethanol) migrates as a single band (monomer) on SDS reducing polyacrylamide electrophoresis (PAGE) which indicated that the proportion of albumin present as a monomer is at least 99.9%.

(186) Gel Permeation High Pressure Liquid Chromatography

(187) 25 l of a 10 mg/ml solution of albumin purified in accordance with the invention which had been formulated to 25% w/v was injected onto a TSK3000SWXL column on a Shimadzu LC6A HPLC and found to contain less than 0.1% polymeric albumin. This result indicates that the formulation as described herein has no detrimental effect on the polymer/aggregate content of the purified albumin.

(188) Two Dimensional Gel Electrophoresis

(189) 2 g rHA of albumin prepared by the process of the invention was subject to two-dimensional electrophoresis using a Millipore Investigator system. The separation in the first dimension was a pH 3-10 isoelectric focusing gel and was followed by a 10% polyacrylamide/SDS gel in the second dimension. On staining of the gel with Coomassie Blue, only one spot was visible, indicating the presence of only one protein species.

(190) Mannosylated Albumin/Con A Assay

(191) Concanavalin A (Con A) binds molecules which contain -D-mannopyranosyl, -D-glucopyranosyl and sterically related residues. In the Con A assay, Con A Sepharose (Pharmacia, Cat. No. 17-0440-01) affinity chromatography of recombinant Human Albumin (rHA) and/or Human Serum Albumin (HSA) is used to determine the content of mannosylated albumin.

(192) Recombinant human albumin (rHA) is diluted to 5% (w/v) rHA with 145 mM sodium chloride then 1:1 with Con A dilution buffer (200 mM sodium acetate, 85 mM sodium chloride, 2 mM magnesium chloride, 2 mM manganese chloride, 2 mM calcium chloride pH5.5). 100 mg rHA is then loaded onto an equilibrated 2 mL Con A Sepharose column which is then washed (54 mL) with Con A equilibration buffer (100 mM sodium acetate, 100 mM sodium chloride, 1 mM magnesium chloride, 1 mM manganese chloride, 1 mM calcium chloride pH5.5). The column is eluted with 6 mL Con A elution buffer (100 mM sodium acetate, 100 mM sodium chloride, 0.5M methyl--D-mannopyranoside pH5.5).

(193) Monomeric albumin in the Con A load (diluted to about 0.1 mg.Math.mL.sup.1) and eluate (assayed neat) are quantified by GP.HPLC using a 0-0.2 mg.Math.mL.sup.1 rHA standard curve and the Con A binding albumin monomer recovered in the eluate is expressed as a percentage of the load.

(194) TABLE-US-00013 TABLE 6 Clearance of conA-binding rHA through the process. Batches 1-4 are derived from a pmt1 mutant, whereas batch 5 is derived from a non-mutant strain. ConA-binding rHA (% of load) Batch 1 Batch 2 Batch 3 Batch 4 Batch 5 PBA 0.14 PBA 0.16 PBA 0.15 PBA 0.13 PBA 0.55 FT&W FT&W FT&W FT&W FT&W SP-FF2 0.10 SP-FF2 0.12 SP-FF2 0.14 SP-FF2 0.09 SP-FF2 0.32 FT&W FT&W FT&W FT&W FT&W Final 0.10 Final 0.11 Final 0.12 Final 0.07 Final 0.28 Product Product Product Product Product (FT & W = Flow Through & Washings)

(195) ConA-binding rHA was further analysed by electrospray mass spectrometry (FIG. 8). This indicated that, in addition to a reduction in the amount of conA-binding rHA, the extent of modification of the conA-binding rHA was reduced.

(196) Analysis of Colour

(197) The absorbance of a 5% (w/v) solution of the final product in a 1 cm cuvette was measured at 350 nm, 403 nm and 500 nm and calculated in terms of absorbances per gram of albumin/litre per cm pathlength (ie ABS.Math.L.Math.g.sup.1.Math.cm.sup.1). The albumin of the invention has the following values:

(198) TABLE-US-00014 Wavelength Mean absorbance (n = 4 batches) (nm) (L .Math. g.sup.1 .Math. cm.sup.1) 350 5.75 10.sup.3 403 1.7 10.sup.3 500 0.4 10.sup.3

(199) Generally, the albumin of the invention does not exceed respective absorbances of 8.010.sup.3, 3.010.sup.3 and 0.7510.sup.3 at the said three wavelengths.

(200) Assays of a number of commercially available HSA preparations revealed higher absorbances at these wavelengths (see Table 7).

(201) TABLE-US-00015 TABLE 7 Absorbance (L .Math. g.sup.1 .Math. cm.sup.1) of prior art HSA preparations SAMPLE A.sub.350 A.sub.403 A.sub.500 1 9.95 10.sup.3 4.10 10.sup.3 0.8 10.sup.3 2 9.25 10.sup.3 5.36 10.sup.3 1.1 10.sup.3 3 7.40 10.sup.3 3.26 10.sup.3 0.6 10.sup.3 4 7.20 10.sup.3 3.60 10.sup.3 0.6 10.sup.3 5 8.68 10.sup.3 4.08 10.sup.3 0.8 10.sup.3 6 11.45 10.sup.3 6.26 10.sup.3 1.2 10.sup.3 7 7.20 10.sup.3 3.70 10.sup.3 0.8 10.sup.3 8 6.82 10.sup.3 4.78 10.sup.3 1.8 10.sup.3
Endotoxin

(202) A solution of drug product is assayed using Limulus amoebocyte lysate by kinetic turbidimetric measurement at 340 nm, at a temperature of 36.5-37.5 C. using an automatic endotoxin detection system (eg LAL 5000E). A standard curve is constructed from known concentrations of a standard endotoxin preparation, negative controls and test material solution spiked with a known quantity of standard endotoxin are also included in the assay. The change in turbidity of the reaction mixture is measured over time and a log-log regression. Any endotoxin in the drug product is quantified against the standard curve and recovery of the endotoxin spike is confirmed. No endotoxin was detected.

(203) Free Thiol

(204) Ellman's Reagent, 5,5-Dithiobis-(2-Nitrobenzoate) (DTNB) is a specific means of detecting free sulfydryl groups such as cys-SH (Cys-residue 34 in the case of rHA). The reaction releases the 5 thio-2-nitrobenzoate ion TNB.sup.2 which has an absorption maximum at 412 nm. By measuring the increase in absorbance at 412 nm and dividing by the molar extinction coefficient of the TNB.sup.2 ion at 412 nm, the free sulfydryl content of rHA can be calculated.

(205) TABLE-US-00016 Sample mol .Math. mol.sup.1 A 0.82 B 0.77 C 0.77 D 0.85 E 0.90