EP0590059A4 - Production and recovery of recombinant neurotrophins - Google Patents

Abstract

We disclose a process for producing and recovering a recombinant neurotrophin comprising the steps of: (1) culturing a non-animal host cell transformed with a recombinant DNA molecule comprising an expression control sequence operatively linked to a DNA sequence encoding a neurotrophin; (2) solubilizing the neurotrophin in a solution comprising a protease inhibitor in an amount effective to inhibit degradation of the neurotrophin by proteases and a strong denaturing agent, said solution being essentially free of reducing agents; (3) exchanging the strong denaturing agent for a weak denaturing agent; (4) adjusting the solution to comprise a basic amino acid at a concentration effective to maintain the solubility of the neurotrophin in a nondenaturing environment; (5) removing the weak denaturing agent; and (6) purifying the neurotrophin. In a specific embodiment, also disclosed is the bacterial production of recombinant neurotrophin-4.

Description

PRODUCTION AND RECOVERY OF RECOMBINANT NEUROTROPHINS

This application is a continuation-in-part of copending application Serial No. 07/715,185, filed June 12, 19.91, which is incorporated by reference in its entirety.

1. INTRODUCTION This invention relates to the fields of neurobiology, molecular biology and protein biochemistry and, in particular, to neurotrophins and methods of producing and recovering them.

2. BACKGROUND OF THE INVENTION The neurotrophins are a group of proteins involved in the functioning of the nervous system. They stimulate the growth of nerve cells and, during embryonic development, support their survival. To date, researchers have identified four neurotrophins: nerve growth factor (NGF) (Ullrich et al., 1983, Nature 30.3:821-825) , brain derived neurotrophic factor (BDNF) (Leibrock et al., 1989, Nature 341:149-152) , neurotrophic factor 3 (NT-3) (Maisonpierre et al., 1990, Science 247:1446-1451) and neurotrophin-4 (Hallbook, et al., 1991, Neuron .6:845-858). Nerve growth factor is essential for the development and survival of the peripheral and sympathetic nervous system and the cholinergic neurons of the brain. It is currently being tested for use in Alzheimer's disease. BDNF promotes the survival of sensory neurons in the central nervous system and shows promise as a therapy for Parkinson's disease. NT-3 and NT-4 are newly discovered and their biological role is now being investigated. Like NGF, NT-3 seems to act on sensory and sympathetic neurons.

Native human NGF has three subunits: α, λ and y . Only the β subunit has neurotrophic activity. /3-NGF is produced

SUBSTITUTE SHEET as a pre-pro-protein which is secreted from the cell and processed into the mature form. The pre-pro-peptide contains 187 amino acid residues. During processing, the pre- and the pro-sequences are removed, as are two amino acid residues at the C-terminus. This yields a mature protein of 118 amino acid residues. The mature protein has six cysteine residues that form three disulfide bonds. The protein also contains many basic amino acids resulting in a positive charge at neutral pH.

Structurally, the neurotrophins form a family, apparently having evolved from a common ancestral gene. (Hyman et al., 1991, WO 91/03568.) For example, the amino acid sequence of the β polypeptide of human NGF (hNGF) is 90% homologous to mouse NGF and bovine NGF; and the three human neurotrophins, hNGF, hBDNF and hNT-3, share 60% homology at the amino acid level. The six cysteine residues are conserved in all known neurotrophins, implying that the disulfide bonds are important for function.

Natural sources of neurotrophins are limited. The mouse sub axillary gland is rich in NGF but isolation is difficult and the quantities recovered, small (Hyman et al., 1991, WO 91/03568). del la Valle et al. (EP 0 333 574, 1989) have reported the isolation of small mounts of hNGF from human placenta by chromatographic fractionation.

Therefore, researchers have turned to molecular genetics to provide a ready source of biologically active neurotrophins. DNA sequences encoding NGF, BDNF and NT-3 have all been isolated (Ullrich et al., Nature 303:821-825; Hyman et al., WO 91/03568; Hohn et al., WO 91/03569; and Kaisho et al., FEBS Letters 266:187-191) . Researchers have transformed animal and non-animal hosts with these sequences in order to express the neurotrophins.

Kanaya et al. (1989, Gene 83:65-74) fused a DNA sequence encoding mature hNGF to one encoding the yeast α-mating factor pre-pro sequence. Upon expression, the

SUBSTITUTESHEET culture supernatant contained a protein recognized by α-hNGF antibodies. However, the partially purified protein exhibited low neurotrophic activity. Furthermore, the level of expression was low. Kanaya et al. stated that these results may have been due either to the failure of the yeast system to remove the two extra C-terminal amino acid residues during maturation or to a problem with folding the hNGF protein.

Chan et al. (EP 0 370 171, 1990) produced mature hNGF in insect cells. They fused a gene for pre-pro-hNGF to the polyhedrin promoter and leader sequence of baculoviral DNA. They reported the production of 6 μg/ml hNGF, but did not physically characterize the product.

Researchers have also expressed human NGF, BDNF and NT-3 in mammalian expression systems. Bruce and Heinrich (1989, Neurobiology of Aging 10.:89-94) expressed a DNA sequence encoding the complete precursor for hNGF in COS cells and detected hNGF di er in the conditioned medium. However, they could not determine the efficiency at which pre-pro-hNGF was converted to mature hNGF. Kakinuma et al. (EP 0 414 151, 1991) expressed active hNGF in CHO cells. Hyman et al. (WO 91/03568, 1991) expressed hBDNF in CHO cells. Nakahama et al. (EP 0 386 752, 1990) and Hohn et al. (WO 91/03569, 1991) expressed hNT-3 in COS cells.

Mammalian systems provide the most natural environment for the production of mammalian proteins. However, the production of large quantities of proteins in these systems is very expensive. Therefore, there is a need to develop systems that are both less expensive and more productive. One such system is E. coli.

Researchers have attempted to produce neurotrophins in E. coli. but without significant success. Gray and Ullrich (EP 0 121 338, 1984) constructed an expression vector encoding N-methionyl-hNGF and expressed the gene in E. coli. They reported identification of hNGF by

SUBSTITUTESHEET immunodetection on Western blot, but they did not isolate the protein or demonstrate biological activity.

Hu and Neet (1988, Gene:57-65) attempted to express mouse NGF in E. coli. They cloned a DNA sequence encoding a mature mouse NGF in which they replaced serine, the N— terminal amino acid in the native protein, with methionine. They inserted the DNA sequence into a plasmid having a temperature inducible lambda P_L promoter. This system expresses other heterologous proteins at rates of 10% - 25% total cellular protein. They expressed the gene and isolated NGF by ammonium sulfate precipitation followed by dialysis against acetate buffer. However, as tested by bioassay, this system yielded only 0.0005% to 0.1% NGF. The authors speculated that the highly inconsistent and low yields were due to toxicity of NGF to the cells, instability or translational inefficiency of the mRNA, or mismatched disulfide bonds in the refolded, oxidized protein.

Iwai et al. (1986, Chem. Pharm. Bull. 34:4724-4730) reported synthesis of a gene encoding hNGF with codons preferred in E. coli. They expressed the gene directly as N-methionyl-hNGF or as a fusion protein with human growth hormone. Direct expression was only one-fourth as efficient as expression of the fusion protein. They examined the proteins by SDS-PAGE, but did not otherwise isolate them.

Dicou et al. (1989, J. Neurosci. Res. 22:13-19) expressed mouse pre-pro-NGF and a fragment of hNGF as fusion proteins with _/3-galactosidase and found that the addition of protease inhibitors improved yield.

In summary, previous attempts to express neurotrophins in E. coli have resulted in cell death, low levels of protein accumulation, and expression of protein with virtually no biological activity.

SUBSTITUTESHEET 3. SUMMARY OF THE INVENTION This invention provides processes for producing and recovering biologically active recombinant neurotrophins from non-animal host cells. Processes for producing recombinant neurotrophins comprise the step of culturing a host cell transformed with a recombinant DNA molecule comprising an expression control sequence operatively linked to a DNA sequence encoding a neurotrophin. When the neurotrophins are expressed directly into the cytoplasm, the host cells preferably are protease-deficient mutants and, according to the best mode of which e are aware, heat shock regulatory gene mutants. In bacterial hosts, the neurotrophin gene is preferably under the control of a controllable promoter and expression is preferably un-induced or repressed until the cells reach late log phase. The neurotrophins may be also secreted from the cell during production by providing a DNA sequence encoding a signal peptide upstream of the DNA sequence encoding the neurotrophin.

Processes for recovering biologically active recombinant neurotrophins produced by host cell cultures comprise the step of solubilizing the neurotrophin in a solution comprising a strong denaturing agent, the solution being essentially free of reducing agents. According to one embodiment of the invention the solution further comprises a protease inhibitor in an amount effective to inhibit degradation of the neurotrophin by proteases. Other embodiments of the invention further comprise some or all of the following steps: exchanging the strong denaturing agent for a weak denaturing agent; adjusting the solution to comprise a basic amino acid or equivalent at a concentration effective to maintain solubility of the neurotrophin in a non-denaturing environment; purifying the neurotrophin from other molecules in the solution; removing

SUBSTITUTESHEET the weak denaturing agent; and completing the purification in the absence of the denaturing agent.

In a further embodiment of the invention, the neurotrophin molecule expressed in E. coli and recovered in a biologically active form is neurotrophin-4. In a preferred embodiment of the invention directed to expression of biologically active NT-4 in E. coli, the NT-4 is encoded by a nucleic acid molecule comprising a sequence substantially as set forth for hNT-4 in Figure 11 (SEQ ID NO: 22) or may comprise a sequence that is at least about seventy percent homologous to such a sequence.

4. DESCRIPTION OF THE DRAWINGS Figure 1 is a schematic representation of plasmid pRPN133. The solid line represents pBR322-derived DNA sequences with the origin of replication (ORI) and the _/3-lactamase gene (ampicillin resistance) (Ap) indicated. Distinctive features of the plasmid are indicated by boxed regions with arrowheads indicating the direction of transcription or replication. P_L indicates the lambda P_L promoter, rbsl is the wild type promoter and ribosome binding site of phage T7 l.l. The hNGF gene encodes a mature polypeptide in which the second amino acid residue, serine is replaced by threonine. cI857 indicates the heat inactivatable λ repressor gene.

Figure 2 shows dose-response curves to recombinant human NGF produced in E. coli by (A) E8 chick embryo dorsal root ganglia (DRG) explants and (B) dissociated E8 DRG.

Figure 3 depicts the nucleotide and amino acid sequences of wild type LamB (Fig. 3A, SEQ ID N0:1) and the synthetic LamB signal sequences, LamBl (Fig. 3B, SEQ ID N0:2), LamB2 (Fig. 3C, SEQ ID N0:3), LamB3 (Fig. 3D, SEQ ID N0:4) and LamB4 (Fig. 3E, SEQ ID N0:5).

Figure 4 depicts the signal sequence processing kinetics of the modified LamB signal sequences, LamBl and

§U!§TITUTESHEET LamB3. We cultured strain BL21/DE3 containing the appropriate LamB-hBDNF plasmid. In this strain, the gene encoding T7 RNA polymerase has been inserted in the chromosome and is under the. transcriptional control of the lac promoter (Studier and Moffatt, J. Mol. Biol. 189:113-30) . Addition of IPTG to the growing cells for 10 minutes allows synthesis of T7 RNA polymerase. Subsequent addition of rifampicin to 0.2 mg/ml blocks transcription by E. coli RNA polymerase but allows transcription by T7 RNA polymerase. This results in selective transcription of the LamB - hBDNF gene which is placed immediately downstream from the T7 late l.l promoter and ribosome binding site of rbs2. Cells were pulsed with ³⁵S-methionine for 30 seconds and then chased with an excess, of cold methionine. The cultures were sampled at the indicated times after chase and the labelled proteins were analyzed by SDS-15% PAGE and fluorography. Processing was determined by densitometric scanning of the precursor and mature forms of LamB-hBDNF.

Figure 5 is a schematic representation of pRPN121. The solid line represents pBR322-derived DNA sequences with the origin of replication (ORI) and the β-lactamase gene (Ap) indicated. Distinctive features of the plasmid are indicated by boxed regions with arrowheads indicating the direction of transcription or replication (ORI) . LacUV5 is the promoter. rbs2 indicates the T7 øl.l promoter and the T7 $1.1 ribosome binding site with minor nucleotide substitutions from the wild type designed to create convenient restriction sites. LamB2 is the signal sequence. hBDNF yc is the structural gene.

Figure 6 depicts a protein profile of Fast S-SEPHAROSE® 1 fractionation of hBDNFmyc. W3110 I^qF- /pRPN121 extract after DEAE chromatography was fractionated as described on a 1.6 cm X 6.5 cm Fast S-SEPHAROSE® column in 7 M urea, 50 mM histidine, pH 5.0, 1 mM EDTA. Fractions as indicated were analyzed by SDS-15% PAGE and proteins

SUBSTITUTE SHEET visualized by Coomassie stain. Fractions 26-30 were pooled.

Figure 7 is a summary of the purification procedure. Strain W3110 I^qF"/pRPN121 was grown, induced, and extracted as described above. Pooled column fractions were analyzed by SDS-PAGE on a 15% acrylamide gel and stained with Coomassie Blue. Lane 1, DEAE; lane 2, Fast S-SEPHAROSE® 1; lane 3, Fast S-SEPHAROSE® 2; lane 4, C4 reverse phase HPLC. Molecular weight standards are indicated.

Figure 8 depicts a dose response curve for purified hBDNFmyc stimulation of E8 chicken embryo DRG neurite outgrowth. The hBDNFmyc activity is compared with recombinant hBDNF purified from a Chinese hamster ovary cell line. Both proteins were purified to greater than 95% by C4 reverse phase HPLC.

Figure 9 depicts a dose response curve for the stimulation of E8 chicken embryo DRG neurite outgrowth by recombinant hBDNF purified from E. coli. The hBDNF was purified to greater than 95% by C4 reverse phase HPLC.

Figure 10 is a schematic representation of pRPN149. The solid line represents pBR322-derived DNA sequences with the origin of replication (ORI) and the /3-lactamase gene (Ap) indicated. Distinctive features of the plasmid are indicated by boxed regions with arrowheads indicating the direction of transcription or replication (ORI) . LacUV5 is the promoter. rbs2 indicates the T7 Φl.l promoter and the T7 Φl.l ribosome binding site with minor nucleotide substitutions from the wild type designed to create convenient restriction sites. LamBl is the signal sequence. hBDNF is the structural gene.

Figure 11 is the DNA sequence of a portion of the isolated human genomic phage clone 7-2 encoding human NT-4 (SEQ ID NO:22; ATCC Accession No:75070) . The predicted hNT-4 protein encoded by the genomic clone 7-2 is represented by the one-letter symbols for amino acids (SEQ

SUBSTITUTESHEET ID NO:23). The boxed region represents the predicted cleavage site of the hNT-4 preprotein. Arrows indicate conserved residues in the presequence. The underlined region (N-R-S) represents a consensus sequence for N-glycosylation. The circled region represents the initiating methionine. The splice acceptor site is located at base pair 461-462 (AG) , representing the 3 '-end of the intron.

Figure 12 is a schematic representation of pRG91.. The solid line represents pBR322-derived DNA sequences with the origin of replication (ORI) and the 3-lactamase gene indicated. Distinctive features of the vector are shown by boxed regions. LacUV5 is the promoter. rbs2 is the phage T7 Φl.l promoter and the T7 Φl.l ribosome binding site. LamB2 is the signal sequence.

Plasmid pRG91 (Regeneron Pharmaceuticals) is a pBR322- based vector designated for the expression of recombinant proteins and their secretion into the periplasmic space of Escherichia coli. The vector consists of the strong, regulated, lacUV5 promoter followed by the phage T7 øl.l promoter and ribosome binding site inserted between the EcoRI and Nrul restriction sites in pBR322. These control elements direct the expression of the LamB2 signal sequence to which recombinant protein gene sequences may be fused. The DNA sequences between the unique Nrul and PvuII restriction sites were deleted, resulting in increased plasmid copy number. This plasmid confers ampicillin (Ap) resistance:

Figure 13 is a dose response curve for hNT-4 expressed in E. coli. Soluble protein extract from an induced culture of strain RFJ26 containing plasmid pRG173 was assayed on E8 dorsal root ganglion explants. The background activity of this assay was 0.2 units.

Figure 14 depicts the effect of hNT-4 on CAT activity. Treatment of motor neuron enriched cultures with a

SUBSTITUTE SHEET partially purified extract from an induced culture of strain RFJ26 containing plasmid pRG173 resulted in a 3.6- fold (at 1:20 dilution) increase in CAT activity after 48 hours as compared to untreated (C-NT) and buffer (C-buffer) controls. The E. coli extract was passed through a Sepharose-S column as disclosed infra prior to treatment of motor neuron enriched cultures.

5. DETAILED DESCRIPTION OF THE INVENTION As used herein, the term "neurotrophin" refers to any naturally occurring member of the neurotrophin family. This includes naturally occurring proteins sharing amino acid sequence homology with any known neurotrophin and conserving the six characteristic cysteine residues. (Some of these proteins may fall structurally into the neurotrophin family, yet may exhibit biological activity other than neurotrophic activity.) The term "neurotrophin" also refers to engineered neurotrophins whose amino acid sequences are derived from or patterned after the naturally occurring neurotrophins. For example, it includes chimeric neurotrophins comprising amino acid sequences from different neurotrophins (e.g., NGF and NT-3) or from the same neurotrophin of different species (e.g., hBNDF and pig BDNF) ; neurotrophins whose genetic sequences have point substitutions, addition or deletion mutations; neurotrophins derived from the pre-pro-sequence (e.g., neurotrophins having the two carboxy-terminal amino acid residues encoded by the codons immediately preceding the stop codon of the native gene ("full length neurotrophin")); and includes fragments of a neurotrophin which exhibit neurotrophic activity (e.g., those whose first six amino acids are altered or deleted) . These examples are descriptive and not meant to limit the definition.

This invention provides processes for producing and recovering biologically active recombinant neurotrophins.

SUBSTITUTE SHEET - li ¬

lt further provides the purified, homogeneous, recombinant neurotrophins made by these processes. A recombinant protein, as used in this specification, is a protein expressed from a recombinant DNA molecule in a host cell transformed with it. A recombinant DNA molecule is a hybrid molecule comprising DNA sequences from different sources that have been joined together. Sambrook et al. (1989, Molecular Cloning: A Laboratory Manuel, Cold Spring Harbor Laboratory Press, Cold Spring Harbor) describes many conventional techniques for recombinant DNA technology.

According to this invention, producing recombinant neurotrophins involves culturing a non-animal host cell transformed with a recombinant DNA molecule having an expression control sequence operatively linked to a DNA sequence encoding a neurotrophin. An expression control sequence is operatively linked to a DNA sequence encoding a polypeptide when the expression control sequence directs and promotes the transcription and translation of the DNA sequence. Culturing a transformed host cell involves incubating the cell in culture conditions appropriate for the growth of the cell and the expression of the DNA sequence.

DNA sequences encoding neurotrophins are available from several sources. The literature discloses DNA sequences for the four known human neurotrophins (i.e., hNGF (Gray et al., EP 0 121 338, 1984), hBDNF (Hyman et al., WO 91/03568), hNT-3 (Hohn et al., WO 91/03569) and Xenopus NT-4 (Hallbook, et al., 1991, Neuron 6:845-858) and for many non-human neurotrophins. Preferably, one refers to these sequences for direction to construct oligonucleotide primers suitable for PCR amplification. Genomic libraries are suitable sources of DNA for PCR templates. Also useful are cDNA libraries of cells known to express neurotrophin mRNA. For example, cDNA libraries of human fetal brain mRNA contain sequences for hBDNF. One

SUBSTITUTESHEET may also screen cDNA and genomic libraries using any of the methods known to the art to identify and isolate DNA sequences encoding neurotrophins. Alternatively, one may construct synthetic or semi-synthetic genes using conventional DNA synthesizers. Undoubtedly, researchers will discover DNA sequences encoding new neurotrophins. The methods of this invention will be useful for producing and recovering these neurotrophins, as well.

In specific, the invention relates generally to production of NT-4 or a derivative or fragment thereof by growing a recombinant bacterium containing a nucleic acid encoding NT-4 or derivative or fragment under conditions such that the NT-4 or derivative or fragment thereof is expressed by the bacterium, and recovering the produced NT- 4 or NT-4 derivative or fragment. In a preferred embodiment, the NT-4 is human NT-4. In another embodiment, an NT-4 derivative which is a chimeric or fusion protein is produced. In a most preferred embodiment, the produced NT-4, or NT-4 derivative or fragment is biologically active, i.e., capable of exhibiting one or more of the known functional activities of NT-4, as assayed by any methods known in the art or taught herein (e.g., in vitro assays of the ability to promote outgrowth in E8 DRG explants, the ability to stimulate CAT activity in purified motor neuron cultures; see Section 9 , infra) .

In a specific embodiment of the invention, sequences encoding human NT-4 are expressed in an E. coli expression system and a purification scheme as disclosed infra is used to produce useful amounts of human NT-4. The nucleic acid encoding human NT-4 which is thus expressed can be that contained in nucleic acid pRG173 (ATCC Accession Number 75131) or HG7-2 (ATCC Accession Number 75070) or shown in Figure 11 (SEQ ID NO:22), or isolated by any methods known in the art, or as follows: Mixtures of 5' and 3' oligonucleotides representing all possible codons

SUBSTITUTE SHEET corresponding to known NT-4 sequences or to conserved amino acid sequences from known neurotrophins are utilized as primers in the polymerase chain reaction (PCR) . Primary and secondary PCR amplification reactions of human (or other mammalian) cDNA or genomic libraries result in the isolation of a PCR product that can be utilized as ³²P-labelled probes to isolate a full length cDNA or genomic clone encoding NT-4. The term "human neurotrophin-4", as used herein should be understood as meaning any human homologue of the Xenopus NT-4 (Hallbook, et al., 1991, Neuron 6:845-858) , including a distinct yet homologous (e.g., at least about seventy percent homology) neurotrophin molecule.

The literature discloses a variety of expression control sequences useful for expressing DNA sequences in transformed non-animal hosts. These include, among others, in bacteria, the lac system, the trp system, the TAC system, the TRC system, the lambda P_L promoter, the T7 late promoters, and the control regions of the fd coat protein; and in yeast, the phosphoglycerate kinase promoter, the Gal 4 promoter, the His promoter, the alcohol dehydrogenase promoter, the alkaline phosphatase promoter and the ct- ating factor promoter. Controllable expression control sequences are preferable and, among these, a temperature inducible lambda P_L promoter, the lacUV5 promoter and the T7 Φl.l promoter are most preferable for expression in E. coli.

The literature also discloses a variety of expression vector/host systems suitable for bacterial and fungal hosts, including plasmids, bacteriophages, cosmids and derivatives of them. Examples of these systems are, for bacteria, col Fl, pCRl, pBR322, pMB9, RP4, phage lambda, M13 and filamentous viruses; and, for yeast, 2μ. Preferred plasmids in bacteria are stable in the host cell and present at medium copy numbers of 20-200 copies per cell.

SUBSTITUTE SHEET We used plasmids derived from pBR322, such as those described by Panayotatos (1987, Engineering an Efficient Expression System, in: Plasmids-A Practical Approach, ed. Herdy, K. , IRL Press, Oxford/Washington, D.C.). In particular, we prefer plasmids of the RPN class, developed by Regeneron Pharmaceuticals, Inc. and described more fully below.

The art is also familiar with many non-animal hosts useful to express heterologous proteins, including E. coli. Bacillus, Strepto vces, Saccharomyces and Pichia pastoris. We prefer E. coli.

Culturing transformed host cells results in the expression of neurotrophins. We have found that neurotrophins expressed in bacteria are subject to degradation by intracellular proteases, especially those induced as part of the "heat shock" response to foreign proteins (Goff et al., 1984, Proc. Natl. Acad. Sci. USA .81:6647-6651). Therefore, we prefer to use protease-deficient mutants as host cells.

The expression of the heat shock genes is regulated by heat shock regulatory genes, such as the htpR gene of E. coli. Mutants of the HtpR gene are deficient in expression of heat shock protease genes, as well as other genes that contribute to the heat shock response. We have found that expressing recombinant neurotrophins in heat shock regulatory gene mutants, especially htpR^* mutants, significantly improves yield. HtpR' Ion' double mutants are particularly useful. We prefer the E. coli strain LC137, an htpR" Ion' mutant. This strain is available from Prof. Alfred Goldberg, Harvard University. U.S. patent 4,758,512 (Goldberg et al.) describes other suitable strains. Another preferred E. coli strain is RFJ26.

Neurotrophins are toxic to bacterial cells that express them. Therefore, in order to maximize yield, we prefer to induce neurotrophin expression only in dense

SUBSTITUTE SHEET bacterial cultures, for example, cultures in late log phase. We have used an inducible system based on the lambda P_L promoter and the cI857 temperature sensitive repressor. We typically induce production of neurotrophins by shifting growth temperature to 42°C for 30 minutes and continuing incubation for 3-20 hours at 38°-42°C. Inducing expression for more than 16 hours in actively growing cells eventually causes cell death.

The inability of E. coli to accumulate neurotrophins may be the result of one or more properties of the neurotrophin gene or protein. The structure of the neurotrophin mRNA, particularly the structure proximal to the translation start point, may prevent efficient translation. Alternatively, the neurotrophin may prevent its own synthesis by interacting directly with its mRNA, or it may interact directly or indirectly with some component of the DNA replication/transcription/translation machinery of E. coli.

According to another embodiment of our invention, we secrete the neurotrophin into the periplasmic space of E. coli rather than express it intracellularly. We accomplish this by fusing a signal sequence to a mature neurotrophin. A signal sequence gene fused to the 5* end of the neurotrophin gene may provide a nucleotide sequence proximal to the translation startpoint that is more conducive to efficient translation, thus resulting in higher levels of neurotrophin accumulation. In addition, sequestering the neurotrophin in the periplasmic space prevents it from interfering with any cytosolic component necessary for protein synthesis. It also protects it from attack by cytosolic proteases. It is also possible that the secretion of a mature neurotrophin into the periplasmic space may provide an environment more conducive to the proper folding of the protein.

SUBSTITUTESHEET A signal sequence is provided by constructing a recombinant DNA molecule in which the DNA sequence encoding the neurotrophin comprises, from 5* to 3' , a fused gene encoding a signal or leader sequence appropriate to the host cell which is in-frame with a DNA sequence for the neurotrophin. The literature describes several signal sequences useful in such constructions. For example, LamB, O pA and PhoA are useful in E. coli. (Denefle et al., 1985, Gene .85:499-510; Wong et al., 1988, Gene 658.:193-203) . We prefer LamB and, in particular, modified LamB signal sequences that improve the translational efficiency of the LamB mRNA. We constructed genes for modified LamB signal sequences in the following manner. We made degenerate substitutions to the third nucleotide of several codons of LamB, replacing G or C with A or T. These substitutions do not change the amino acid sequence of the LamB signal peptide, but do decrease the potential number of hydrogen bonds in any secondary structure. This reduces the stability of possible secondary structures involving this region of LamB mRNA. We also introduced codon changes based on codon usage models, to more nearly approximate codons used most frequently by E. coli.

We also modified the LamB signal sequence to improve efficiency of processing of the LamB precursor protein into mature protein. Native LamB has a hydrophobic core of 10 amino acid residues. Mutational analysis of several E. coli signal sequences suggests that the length of the hydrophobic core region can have a strong effect on signal sequence activity. We have found that increasing the length of the hydrophobic region by the addition of up to ten hydrophobic amino acid residues improves the efficiency of processing LamB fusion precursor polypeptides. Fewer than six is preferable and four is most preferable. The choice of hydrophobic amino acids added is not critical, nor is the precise location at which they are added to the

SUBSTITUTESHEET hydrophobic core region. However, we prefer to add the tetra-peptide Leu-Ala-Val-Leu ("LAVL") (SEQ ID NO:6) at a convenient restriction site near the N-terminal end of the hydrophobic region. We describe particular genes for modified LamB signal sequences in Example 7.

In accordance with our invention, the recombinant neurotrophins are released from the culture of host cells by harvesting the cells, lysing them, centrifuging the lysate and collecting the lysate pellet. One typically harvests the cells by centrifugation. To inhibit degradation of the neurotrophin after harvesting the cells, one preferably resuspends them in 50 mM EDTA. Then we either use the cells directly or freeze them for later use. The art is familiar with many techniques to lyse cells including enzymatic (e.g., lysozy e) , chemical (e.g., alkali, SDS) and mechanical (e.g., French press, hydrodynamic shear) . Mechanical means are preferred because there is less risk of harm to the neurotrophins. We prefer to lyse the cells by passing them through a French press or a STANSTED® cell disrupter at 10,000 psi. We obtain the lysate pellet by centrifugation, at about 15,000 x g.

This invention further provides processes for recovering the recombinant neurotrophins produced in cell cultures. These processes overcome the problems associated with conventional methods of recovering recombinant proteins that have been applied to neurotrophins.

Native neurotrophins are soluble in neutral buffers. However, recombinant neurotrophins from E. coli behave as insoluble proteins. Recombinant neurotrophins have been detected in the cytosolic fraction and, when exported, have been recovered from the periplasmic space using standard techniques such as osmotic shock, spheroplasting, or freeze-thaw (Bochner et al., U.S patent 4,680,262).

SUBSTITUTE SHEET However, they are isolable only as a small fraction of the total neurotrophins in the cell.

Many mammalian proteins expressed in bacteria are insoluble. Produced in a foreign ionic environment, they do not fold correctly and expose normally hidden hydrophobic regions. Consequently, the proteins form aggregates and precipitate as inclusion bodies. The literature suggests that the cysteine residues of these proteins are at least partially reduced or mismatched (Tsuji et al., 1987, Biochemistry 26:3129-3134) .

The literature describes standard techniques for purifying recombinant proteins from inclusion bodies (see, e.g., Builder et al. , U.S. patent 4,620,948; Hershenson et al., U.S. patent 4,961,969; and Hung et al. , U.S patent 4,734,362). These techniques involve dissolving the protein in a solution comprising a strong denaturant and a reducing agent. After exchanging the strong denaturant for a weak denaturant, the proteins are renatured in a neutral solution and oxidized to form the correct disulfide bonds. The majority of the recombinant neurotrophin produced by these methods is biologically inactive. We believe that this is due, in part, to the complexity of the neurotrophins¹ tertiary structure. After denaturing the protein and reducing the sulfhydryl groups, the polypeptide does not refold into the proper configuration for the correct disulfide bonding. Furthermore, our results suggest that recombinant neurotrophins are present in aggregates that are not typical inclusion bodies.

We have discovered a process to recover biologically active recombinant neurotrophins from the insoluble host cell fraction. Essentially, the process involves solubilizing the neurotrophin in a strong denaturing agent while maintaining the correct oxidation state of the protein by avoiding the use of reducing agents. Reducing the disulfide bonds during purification destroys the

SUBSTITUTE SHEET activity of purified neurotrophin polypeptides. Because the literature (e.g., Tsuji et al. 1987, Biochemistry 2J5:3129-3134) teaches that insoluble recombinant proteins in inclusion bodies should have incorrectly formed or reduced disulfide bonds, this is an unexpected result. To further recover the neurotrophins, the strong denaturing agent is replaced with a weaker one. We have also discovered, however, that the addition of a basic amino acid, such as histidine, or its equivalent, helps keep the neurotrophin soluble during the renaturation process that accompanies the removal of the weak denaturant (Prior et al., 1990, WO 90/06764).

These discoveries suggest to us a new model for the behavior of recombinant neurotrophins. We believe that the majority of the recombinant neurotrophin polypeptides produced by the host cell fold properly and form the correct disulfide bonds. Furthermore, since neurotrophins have a high pH and are positively charged at neutral pH, we believe that they form associations with negatively charged molecules in the cell, for example DNA, RNA and other proteins. Consequently they become insoluble at neutral pH. Basic amino acids help to dissociate them.

More specifically, our process for recovering recombinant neurotrophins produced by a host cell culture comprises the step of denaturing the neurotrophin by dissolving it in a solution comprising a strong denaturing agent which solution is also essentially free of reducing agents. Denaturing agents as used herein refer to compounds which, in aqueous solution, reversibly unfold dissolved proteins by at least partially eliminating tertiary and secondary structure through the disruption of hydrogen bonds or alteration of the thermodynamic surroundings of the protein. Strong denaturing solutions include guanidinium salts (e.g., guanidinium hydrochloride) and alkali metal thiocyanates (e.g., sodium thiocyanate) at

SUBSTITUTE SHEET concentrations of 4 M - 9 M, and urea at 7 M - 9 M. The preferred denaturing solutions are 7 M - 9 M guanicinium HCl. Preferred conditions are pH 7.0 - pH 9.0 and room temperature. Most preferred is 8 M guanidinium HCl, pH 8.0.

One must also maintain the correct oxidation state of the sulfur atoms in the cysteine residues during this step or risk losing the ability to recover any biologically active neurotrophins. Therefore, the solution must be essentially free of reducing agents. A solution essentially free of reducing agents is one in which a protein's disulfide bonds are maintained. Addition of even small amounts of disulfide reducing agents, such as jS-mercaptoethanol or dithiothreitol, during the practice of this invention will destroy the activity of the purified neurotrophin.

Recombinant neurotrophins apparently are subject to degradation by metalloproteinases. Therefore, preferably, one recovers neurotrophins in solutions comprising metalloproteinase inhibitors. We prefer heavy metal chelators, such as EDTA. The concentration of EDTA may range from a minimum of at least 1 mM to a maximum of about 200 mM. Concentrations of 5 mM - 80 mM are preferable and 50 mM is most preferable. After chelated heavy metal ions have been dialyzed from the solution, EDTA may be reduced or eliminated. Langley et al., 1990, EP 0 398 753, describes peptides useful as metalloproteinase inhibitors. The recovery process may also comprise one or more of the following steps. After solubilizing the neurotrophin, one exchanges the strong denaturing agent in the solution for a weak denaturing agent. Weak denaturing solutions include urea at 4 M - 9 M and, preferably, 6 M - 8 M, pH 7.0 - pH 9.0, at room temperature. The most preferable weak denaturing solution is 7 M urea, 50 mM Tris-HCl, 10 mM NaCl, 5 mM EDTA, pH 8.0. Dialysis is the preferred method

SUBSTITUTESHEET of exchange. If the strong denaturing agent is 7.0 M - 9.0 M urea, the weak denaturing agent of this step is preferably urea of lower concentration.

In order to renature the neurotrophin, one removes the weak denaturing agent from the solution. Dialysis or diafiltration against a non-denaturing solution is the preferred method of removal. We prefer 0 M to 1 M NaCl as a non-denaturing solution.

Another step in the recovery process is purifying the recombinant neurotrophin from other contaminants in the solution. Any of the typical protein isolation techniques known to the art may be used. We prefer a three step isolation involving cation exchange on S-SEPHAROSE®, anion exchange on DEAE-SEPHAROSE® and reverse phase high pressure liquid chromatography (HPLC) . The isolation step can begin at any stage after solubilization of the neurotrophin and may continue through the denaturing agent exchange and removal steps. Preferably, we begin purification at the urea phase, because this agent does not interfere with ion exchange chromatography and because partial purification made the neurotrophin easier to dissolve in the non-denaturing solution.

We have found that unless the neurotrophin is relatively purified it will precipitate out of the solution when one removes the weak denaturing agent. Solubility depends upon degree of purity of the neurotrophin as well as the character of the other contaminants in solution. For example, negatively charged molecules such as DNA will interfere with solubility of even very pure neurotrophin. Therefore, we maintain the solubility of the neurotrophin in the non-denaturing solution by adjusting the solution to comprise a basic amino acid or an equivalent, such as indole acetic acid. The concentration of this species should be effective to maintain the neurotrophin in solution. Solutions of basic amino acids include

SUBSTITUTE SHEET histidine, lysine and arginine or their salts at concentrations above 10 mM. Histidine in concentrations of 20 mM - 500 mM is useful. We most prefer histidine in concentrations of 20 mM - 100 mM. If the neurotrophin in the denaturing solution has been sufficiently purified before removal of the denaturant, this step may be eliminated.

Results indicate that our method produces uniform neurotrophin molecules, consistently having the same N-terminal amino acid. In contrast we have found that expression of neurotrophins in the mammalian CHO cell system produces a mixture of neurotrophin molecules with varying N-terminal amino acids. Therefore, the recombinant neurotrophins produced by our process appear to be unique.

The processes of this invention are also useful for recovering other proteins having biochemical properties similar to the neurotrophins. That is, the processes are useful for recovering proteins that fold correctly in bacteria and form the proper disulfide bonds, but, once denatured and reduced in conventional recovery techniques, are very difficult to renature with proper disulfide bonds. This includes proteins with pH greater than about 9.0 and having at least two disulfide bonds. (Candidate molecules include secretory leukocyte protease inhibitor (Miller et al., 1989, J. Bacteriology 171:2166-2172) and full length recombinant CD4 (Fisher et al. , 1989 WO 89101940) .)

As described in an example section infra, the present invention discloses the expression of biologically active human neurotrophin-4. The human NT-4 DNA sequence was subcloned into the DNA plasmid vector pRG91, resulting in pRG173. This hNT-4 containing plasmid was transformed into E. coli strain RFJ26, and methods described in the instant specification were utilized to recover biologically active NT-4 from the culture system. However, applicants are not to be limited to such a specific embodiment. For example,

SUBSTITUTESHEET any nucleic acid sequence substantially homologous to the region of HG7-2 encoding human NT-4 can be utilized to construct any number of DNA plasmid expression vectors as described throughout the specification or known to the skilled artisan, which in turn can be utilized to transform any number of E. coli bacterial strains in order to produce useful amounts of biologically active NT-4.

In order that this invention may be better understood, the following examples are set forth. These examples are for the purposes of illustration only, and are not to be construed as limiting the scope of the invention in any manner.

6. EXAMPLE: PRODUCTION AND RECOVERY OF

RECOMBINANT hNGF FROM E. COLI

A DNA sequence carrying the mature human NGF (hNGF) gene was amplified from human genomic DNA using known human

NGF sequences (Nancy Ip; Regeneron Pharmaceuticals, Inc.) by PCR amplification. We used the oligo-deoxynucleotide primers PAN-20 (5'-AAGCGGTCGA CATCTOATCC AATCTTCCAT

AGAGGTGAAT TCTCAGTA-3') (SEQ ID NO:7) and EVD-3

(5'-GGCAGGCGGC CGTCATCTCA CAGCCTTCCT GCTG-3 ') (SEQ ID

NO:8). These primers were designed to incorporate a Sail restriction site at the 5* end and an Eaσl restriction site at the 3' end of the human NGF gene. In order to incorporate a Sail site at the 5' end with PAN-20, it was necessary to change the second amino acid from the amino terminal sequence of the mature hNGF protein from serine to the structurally similar amino acid threonine. In view of the limited sequence identity of the three known neurotrophins in their eight N-terminal amino acids, such conservative amino acid changes should not affect activity.

Furthermore, the identity of the amino acid residue in this position is not preserved in NGF molecules from other species (Ibanez et al., 1990, EMBO J. 9:1477-1483). In addition, PAN20 was designed so as to lower the G+C content

SUBSTITUTE SHEET of this region of the NGF gene in a manner that does not affect the sequence of the resulting protein.

The amplified DNA fragment was digested with the restriction endonucleases Sail and Ea l and ligated between the Sail and EagI sites of pRPN50, (Regeneron Pharmaceuticals, Inc.), resulting in plasmid pRPN102. pRPN50 is derived from expression plasmid pNKS97 (Panayotatos, 1987, Engineering an Efficient Expression System, In: Plasmids-A Practical Approach, ed. Herdy, K. , IRL Press, Oxford/Washington, D.C.) by the insertion of the lambda P_L promoter" at the promotor insertion site.

The cI857 lambda P_L repressor gene was incorporated into pRPN102 to facilitate repression at 30°C and to allow derepression by heat shock at 42°C. This DNA sequence was PCR amplified using the primers EVD-26 (5•-CCATTATCGC _. GACCAGAGGT-3•) (SEQ ID NO:9) and EVD-27 (5'-TCTTGCTCGC GAGTTATCAG CTATGCG-3') (SEQ ID NO:10) that generate a Nrul restriction site at each end. This 821 bp fragment extends 70 bp upstream from the cI857 coding sequence in a region that contains the PRM constitutive promoter as well as 38 bp beyond the cI857 termination codon. The PCR amplified cI857 fragment was digested with Nrul and ligated into the Nrul site of pRPN102. Candidates from this ligation were screened to find a plasmid with the cI857 insert that had the same transcriptional direction as the hNGF gene, resulting in plasmid pRPN133 (Fig. 1) . Plasmid pRPN133 has been deposited with the ATCC and has been assigned Accession No: 75029.

We transformed the E. coli htpR" Rts Ion^" mutant strain, LC137, with RPN133 to yield pRPN133/LC137.

In a 2 liter flask, 400 mL of LB medium supplemented with 100 mg/L a picillin was inoculated with pRPN133 LC137 and incubated at 28°C with shaking. At late logarithmic growth phase (OD₅₉₀ = 1.0 - 2.0 under these conditions), hNGF synthesis was induced by adding to the overnight culture

SUBSTITUTESHEET 400 L of LB media preheated to 55°C and continuing incubation with shaking at 42°C for 5 hours.

Cells were harvested by centrifugation and the cell pellets stored at -20°C. Cell pellets (1.0 - 2.5 g) were thawed, resuspended in 20.0 mL buffer A (100 mM Tris-HCl, 50 mM EDTA, pH 8.0), passed through a STANSTED® cell disrupter at 10,000 psi, and centrifuged at 23,000 x g for 15 minutes at 4°C. The pellet was washed twice with 10.0 mL of 2 M guanidinium-HCl, 5 mM EDTA, pH 8.0.

To solubilize the recombinant hNGF, we resuspended the pellet in 40 mL of 8 M guanidinium-HCl, 5 mM EDTA, pH 8.0 using a Potter homogenizer. Then we centrifuged the suspension at 23,000 x g for 15 minutes.

The supernatant was dialyzed against 2 liters buffer B (7 M urea, 100 mM histidine, 0.1 mM EDTA, pH 6.0) with two buffer changes at 25°C for 20 hours. The dialysate was centrifuged at 23,000 x g for 15 minutes at 4°C.

The supernatant was applied to an S-SEPHAROSE® column (2.5 cm x 6.0 cm d x h) equilibrated in buffer B and washed to baseline absorbance at 280 nm with the same buffer. Proteins were eluted with a gradient of 300 L of 0.0 M - 1.0 M NaCl in buffer B. Fractions containing hNGF were pooled.

The pooled fractions were dialysed against 100-fold excess volume buffer C (7 M urea, 50 mM Tris-HCl, 0.1 mM EDTA, pH 8.5) at 25°C for 20 hours.

The dialysate was applied to a DEAE-SEPHAROSE column (2.5 x 7.5 cm d x h) equilibrated in buffer C at a flow rate of 2.0 mL per minute. The flow-through fractions containing the hNGF were pooled and dialyzed twice against 40 volumes of 100 mM histidine, 0.1 mM EDTA, pH 6.0 at 4°C overnight. The dialysate was centrifuged at 23,000 x g for 15 minutes at 4°C. The dialysate was then applied to a TOYOPEARL CM 6505® (weak cation exchange) column (1.0 x 5.0 cm) equilibrated with 100 mM histidine, 0.1 mM EDTA, pH 6.0

SUBSTITUTE SHEET and eluted with a 0.0 M to 1.0 M NaCl gradient. Fractions containing the hNGF were pooled and filtered through MILLIPORE GV® filter and stored at 4°C. The amino terminal sequence of the purified protein was confirmed by direct sequencing."

The biological activity of the purified protein was tested for neurite outgrowth in E8 explanted and dissociated dorsal root ganglia (Fig. 2A and 2B) (Lindsay et al. , 1985, Dev. Biol. 112:319-328) . By this criteria the recombinant hNGF purified from E. coli by this method was found to be as active as NGF purified from mouse salivary gland.

7. EXAMPLE: PRODUCTION AND RECOVERY OF RECOMBINANT hBDNF and hBDNFmyc FROM E. COLI

7.1. SIGNAL SEQUENCES BASED ON LAMB

We constructed signal sequences based on LamB to facilitate both the synthesis and secretion of neurotrophins in E. coli. The LamB signal sequence (Fig.

3A, SEQ ID N0:1) is a naturally occurring E. coli signal sequence which was selected for the construction of a series of secretion vectors based on the pRPN series of expression vectors developed at Regeneron Pharmaceuticals,

Inc. Secretion vectors were constructed by cloning synthetic DNA fragments encoding the LamB signal sequence into pRPN09 or pRPN16. These plasmids derive from expression vector pNKS97 (Panayotatos, 1987, Engineering an

Efficient Expression System, In: Plasmids-A Practical

Approach, ed. Herdy, K. , IRL Press, Oxford/Washington,

D.C.) into which have been inserted a lacUV5 promoter.

LamB was inserted into the structural gene insert site such that expression of LamB is under control of the lacUV5 or

T7 Φl.l promoter.

The synthetic DNA fragment, LamBl (Fig. 3B, SEQ ID

NO:2), encodes 25 amino acids identical to the wildtype

SUBSTITUTESHEET LamB signal sequence. However, we altered the nucleotide sequence relative to the wild type nucleotide sequence for the purpose of constructing unique restriction enzyme sites and for the maximization of translation efficiency. LamBl also has seven degenerate nucleotide substitutions replacing C or G with A or T. This reduces the stability of possible secondary structure of mRNA. LamBl also has several new restriction sites.

We also created LamB2 (Fig. 3C, SEQ ID NO:3) as a modification of LamBl. To make LamB2 we made nucleotide changes to the 3' end of LamBl by PCR amplification. These changes introduced an Eaσl restriction site which facilitates the cloning of blunt end DNA fragments.

The fusion of mature hBDNF to LamBl results in efficient synthesis of the fusion protein in E. coli. The authenticity of the synthesized product was confirmed by selective synthesis in a DNA-dependent coupled transcription-translation cell-free protein synthesis system, by the selective synthesis of the product using a T7 RNA polymerase expression system in E. coli. and by the synthesis of the product with a C-terminal myc tag allowing for identification of the chimera with a myc-specific monoclonal antibody. Either one of these fusion proteins synthesized in E. coli was processed to mature hBDNF as evidenced by its mobility on SDS-PAGE. The level of expression obtained with LamBl or LamB2 results in accumulation of hBDNF from about 1% to 10% of total cell protein.

We modified LamBl to increase the efficiency of processing. Four amino acids (Leu-Ala-Val-Leu, or LAVL) were inserted into the Nhel site of LamBl to yield LamB3 (Fig. 3D, SEQ ID N0:4). The same insertion was made in LamB2 to yield LamB4 (Fig. 3E, SEQ ID N0:5). This insertion results in extension of the hydrophobic core region of LamBl and LamB2 from 10 to 14 amino acids. It

SUBSTITUTESHEET results in a 5-fold increase in the rate of processing of the hBDNF fusion protein into mature hBDNF (Fig. 4) .

LamB3- and LamB4-hBDNF molecules that are exported into the periplasm are processed into mature hBDNF more rapidly than wild type LamB fusions. However, fewer molecules are exported, so the net amount of mature hBDNF in this case is not increased. In any case the increased translocation efficiency of LamB3 and LamB4 should result in improved yields of other proteins.

The LamBl signal sequence fragment was constructed as two complementary synthetic oligonucleotides (LamB80, 5'-ATGATCACAC TGCGTAAGCT TCCGCCTAGCT GTAGCAGTAG CAGCAGGTGT AATGTCTGCA CAGGCCATGG CCCGGGATCC-3 • (SEQ ID NO:11) and LamB88, 5^«-CTAGGGATCC CGGGCCATGG CCTGTGCAGA CATTACACCT GCTGCTACTG CTACAGCTAG CGGAAGCTTA CGCAGTGTGA TCATCATG-3' (SEQ ID N0:12)), designed so as to generate upon annealing protruding ends corresponding to those of the Spel and SphI restriction sites. This DNA fragment was ligated into the SphI and Spel restriction sites of pRPN16 (Regeneron Pharmaceuticals, Inc.) resulting in plasmid pRPN52.

This plasmid was subsequently modified, for the purpose of creating additional restriction sites, by the insertion of a synthetic DNA fragment, consisting of the annealed product of the complementary LamB2A, (5'-CATGGCCAGT CGGCCGAG-3' (SEQ ID NO:13)) and LamB2B (5^«-GATCCTCGGC CGACTGGC-3• (SEQ ID NO:14)) oligonucleotides, between the unique Ncol and BamHI restriction sites in the LamBl signal sequence in pRPN52. The resulting plasmid, pRPN88, contains the LamB2 signal sequence with unique restriction sites at the signal peptidase recognition sequence to facilitate fusion of the signal sequence to any other DNA sequence. The LamB2 signal sequence in pRPN88 is under transcriptional control of the lacUV5 or the T7 Φl.l promoter and translational control of the T7 Φl.l ribosome binding site.

SUBSTITUTE SHEET 7.2. PRODUCTION AND RECOVERY OF RECOMBINANT hBDNFmyc

Human BDNFmyc is a protein comprising, from N- to C-terminus, mature hBDNF fused to an antigenic "tag." The tag is a peptide having the amino acid sequence EQKLISEEDL (SEQ ID NO:15) These ten amino acid residues derive from the human c-myc proto-oncogene. Antibodies recognizing this tag are useful for identifying hBDNFmyc in a sample (Evan et al., Mol. Cell. Biol. 5_:3610-3616; see also, S. Squinto et al. , "Assay Systems for Detecting Neurotrophic Activity," U.S. application 07/5321,283, incorporated herein by reference) .

To construct a fusion protein comprising the LamB2 signal sequence and the hBDNFmyc protein, the hBDNFmyc DNA sequence was PCR amplified from pCDM8-hBDNFmyc (Regeneron Pharmaceuticals, Inc.) using oligonucleotide primers N8-hBDNF (5'-CCCACTCTGA CCCTGCCCGC CGAGGG-3 ' (SEQ ID NO:16)) and C2-hBDNFmyc (5'-GCTATGCGGC CGCTACAGAT CCTCCTC-3¹ (SEQ ID NO:17). The amplified DNA fragment was digested with Eaσl and cloned into the Ball and EagI sites of the LamB2 sequence in pRPN88, resulting in plasmid pRPN98.

One also can produce a DNA sequence encoding hBDNFmyc by fusing a DNA sequence encoding mature hBDNF to one encoding the ten-amino acid myc tag. DNA sequences encoding mature hBDNF are isolated by PCR amplification as described. (Hyman et al., 1991, WO 091/01568.) A double stranded DNA sequence encoding the ten-amino acid myc peptide tag may be chemically synthesized. For the purposes of this Example, the hybrid hBDNFmyc DNA sequence is then provided with Ball and EagI restriction sites at the 5' and 3' ends, respectively.

It should be noted that there are two Ball sites in pRPN88 but the site in the pBR322-derived sequences is about 50 times less sensitive to Ball cleavage than the site in the LamB2 signal sequence due to dcm methylation

SUBSTITUTE SHEET (New England Biolabs) . In pRPN98 the LamB2 signal sequence is fused to the mature part of the hBDNFmyc protein so that cleavage at the signal peptidase recognition sequence should yield hBDNFmyc protein starting at the histidine residue at +1 relative to the pro-protein processing site (Leibrock, 1989, Nature 341:149-152).

The DNA sequences between the unique Nrul and PyuII sites in pRPN98 were deleted to remove sequences that control copy number of the plasmid (Twigg and Sherratt, 1980, Nature 283:216-218). The resulting plasmid, pRPN121 (Fig. 5) , has a copy number about 3-fold higher than the parental plasmid. Plasmid pRPN 121 has been deposited with the ATCC and been assigned Accession No: 75028.

We transformed E. coli W3110 I^qF' with pRPN121. to produce W3110 IF7pRPN121.

In a 2 L flask, 500 ml of LB medium was inoculated with W3110 I^qF"/pRPN121 and grown to late log with shaking at 37°C (0D₅₉₀ approximately 1.0), after which lactose was added to a final concentration of 1% (w/v) and the culture aerated overnight at 37°C. Cells were harvested by centrifugation, washed once in 0.2 M Tris-HCl, pH 8.0 and the cell pellet frozen at -70°C.

The cell pellet (approximately 5 g) was thawed and resuspended in 20 mL 0.2 M Tris-HCl, pH 8.0, 1 mM CaCl,, and 25 units micrococcal nuclease (Boehringer Mannheim) . The cell suspension was passed through a French pressure cell at 8000 psi and then centrifuged at 15,000 rpm for 15 minutes in a SA600 rotor at 4°C. The pellet was resuspended in 30 mL of 0.2 M Tris-HCl, pH 8.0, 10 mM EDTA, 2% Triton X-100 and gently rocked at room temperature for one hour then centrifuged at 15,000 rpm for 15 minutes in a SA600® rotor at 4°C. The pellet was washed twice in 20 mL of 2 M guanidinium-HCl. The pellet was resuspended in 10 mL of 8 M guanidinium-HCl, 10 mM Tris-HCl, pH 8.5, 10 mM

SUBSTITUTE SHEET NaCl, 1 mM EDTA using a Potter homogenizer. The extract was brought to 20 mL with the same buffer.

The extract was dialyzed overnight at room temperature against 100 X volume 7 M urea, 50 mM Tris-HCl, pH 8.5, 10 mM NaCl, 1 mM EDTA.

The dialysate was applied to a DEAE ZETA-PREP® disk (Cuno, Inc., Meriden, CT) equilibrated in 7 M urea, 50 mM Tris-HCl, pH 8.5, l mM EDTA, at a flow rate of 3 mL/minute and washed to baseline with the same buffer.

The flow-through fractions were brought to 50 mM histidine, pH 5.0 then dialyzed overnight at 4°C against 100 x volume 7 M urea, 50 mM histidine, pH 5.0, l mM EDTA. The dialysate was applied to a 1.6 cm x 6.5 cm column of S-SEPHAROSE® equilibrated with 7 M urea, 50 mM histidine, pH 5.0, 1 mM EDTA, at a flow rate of 1 mL/minute, and washed to baseline with the same buffer. Proteins were eluted with a NaCl gradient from 0.0 M - 1.0 M in 200 ml. Fractions containing hBDNFmyc were collected (Fig. 6) .

Fractions containing hBDNFmyc were dialyzed against 200 X volume 50 mM histidine, pH 5.0, 50 mM NaCl, l mM EDTA. The dialysate was applied to a 1.6 cm X 1.5 cm column of S-SEPHAROSE® equilibrated with 50 mM histidine, pH 5.0, 50 mM NaCl, 1 mM EDTA, at a flow rate of 1 mL/minute and washed to baseline with the same buffer. Proteins were eluted with a NaCl gradient from 0.0 M - 1.0 M in 200 ml. The fractions containing hBDNFmyc were collected and pooled.

The hBDNYmyc in this sample (approximately 85% pure) was applied directly to a 0.45 cm X 5.0 cm C4 reverse phase column (VYDAC®) and eluted with a gradient of 0-67% acetonitrile in 0.1% trifluoroacetic acid in 80 mL at a flow rate of 0.75 mL/minute. The peak fraction was greater than 95% pure (Fig. 7) .

Analysis of the N-terminal amino acid sequence confirmed the purified protein to be authentic hBDNFmyc

SUBSTITUTE SHEET with a homogeneous N-terminus (Leibrock, 1989, Nature 341:149-152) .

The purified hBDNFmyc was biologically active in promoting neurite outgrowth from dorsal root ganglion (DRG) explants and nodose ganglion explants from E8 chicken embryos, as well promoting survival and dendritic outgrowth of dissociated neurons from E8 chicken DRG's (Fig. 8) . This activity is comparable to recombinant human BDNF purified from mammalian cell extracts on DRG explant assays. Material assayed corresponds to Fig. 7, lane 4.

7.3. PRODUCTION AND RECOVERY OF RECOMBINANT hBDNF

The process described for the production and recovery of hBDNFmyc was also used to produce and recover recombinant mature full length hBDNF, and yielded protein with similar bioactivity (Fig. 9) . The construction of plasmid pRPN149 (Fig. 10) , which expresses a LamBl-hBDNF fusion protein, is analogous to the construction of pRPN121. The synthetic LamBl DNA fragment, described above, was cloned into the SphI and Spel restriction sites of pRPN09 (Regeneron Pharmaceuticals, Inc.) resulting in plasmid pRPN31. The mature hBDNF DNA sequence was PCR amplified from pCDM8-hBDNF (Regeneron Pharmaceuticals,

Inc.) using oligonucleotide primers Nl-hBDNF (5'-ACTCTGACCC

TGCCCGCCGA GGGGAGCTG-3') (SEQ ID NO:18) and Cl-hBDNF

(5'-GCGCGGATCC CTATCTTCCC CTTTTAATGG TCAATGTAC-3') (SEQ ID

NO:19) This DNA fragment was cloned into Smal and BamHI sites of pRPN31 resulting in plasmid pRPN34. The

Hindlll-BamHI fragment including the LamBl-hBDNF fusion gene was subsequently cloned into the Hindlll-BamHI restriction sites of pRPN52 (described above) resulting in plasmid pRPN66. The NruI-PvuII deletion of pBR322 sequences resulting in higher copy number (described above) was made in pRPN66 resulting in plasmid pRPN149. In this plasmid, expression of LamBl-hBDNF is under control of

SUBSTITUTESHEET lacUVS and the T7 Φl.l promoter and ribosome binding site. Plasmid pRPN149 has been deposited with the ATCC and has been assigned Accession No: 75027.

We transformed E. coli I^qF-W3110 with pRPN149. Then we produced and recovered recombinant hBDNF by the procedure described in Example 7.

8. EXAMPLE: PRODUCTION AND RECOVERY OF RECOMBINANT hNT-3

Human NT-3 is produced and recovered by processes similar to those we described for hNGF and hBDNF.

A DNA sequence encoding mature full length hNT-3 is PCR amplified from a cDNA library from human brain cells (Hohn et al., 1991, WO 91/03569, incorporated herein by reference). The oligonucleotide primers EVD-45 (5'- CCTATGCAGA GCATAAGAGT CACCGAGGA-3 ^») (SEQ ID NO:20) and EVD-7 (δ'-GTAAGGGCGG CCGAAGTTTA ATAAATAAAG GTC-3 ^») (SEQ ID NO:21) are used. These primers are derived from the DNA sequence for rat NT-3 (Maisonpierre et al. , Science 247:1446-1451) . The sense primer is nearly identical to the human NT-3 sequence. The antisense primer hybridizes approximately one hundred base pairs downstream of the termination codon of the human gene.

This DNA fragment has a C-terminal EagI restriction site suitable for insertion into Ball and EagI sites of pRPN88, where it would replace the hBDNF DNA sequence. The resulting plasmid is used to transform E. coli I^qF-W3110. Then recombinant hNT-3 is produced and recovered as in Example 7. We expect recombinant hNT-3 purified from E. coli to exhibit neurotrophin activity similar to that described by Hohn et al., 1991, (WO 91/03569).

SUBSTITUTE SHEET 9. EXAMPLE: PRODUCTION AND RECOVERY OF RECOMBINANT hNT-4 FROM E. COLI

9.1. CONSTRUCTION OF PRG173

The DNA sequence encoding the putative mature region of the human NT-4 (hNT-4) gene was amplified by PCR from NT-4 HG7-2 DNA, using the N1-NT4

(5'-CCGGGGTCTCTGAAACTGCACCAGCGAGTCG-3 ') [SEQ ID NO:23] and C1-NT4 (5^«-GGTGCAGTTTCAGAGACCCCCATACGCCGGCTGCGGTTGGC-3 ') [SEQ ID NO:24] oligonucleotides as primers. The C1-NT4 oligonucleotide generates an EagI restriction site 3* to the NT-4 gene. The PCR generated fragment was digested with EagI and cloned into MscI-EagI digested pRG91. The resulting plasmid, pRG173, consisted of the LamB signal sequence fused to the mature region of hNT-4 (the glycine at amino acid residue 81 in the HG7-2 translated sequence) under the transcriptional control of the lacUVS and T7 øl.l promoters and the translational control of the T7 øl.l ribosome binding site. This plasmid also possessed the ropl deletion which increases plasmid copy number. The construction was confirmed by restriction enzyme analysis of purified plasmid DNA, DNA sequence analysis of the purified plasmid, in vitro synthesis of a protein of the approximate size estimated to be encoded by pRG173, and in vivo synthesis of a protein of the appropriate size estimated to be encoded by pRG173 and possessing neurite outgrowth stimulating activity (see discussion, supra) , and having the appropriate N-terminus as determined by amino acid sequencing (GVSETAPAE [SEQ ID NO:25]).

9.2. PURIFICATION OF HUMAN NT-4 EXPRESSED IN E. COLI RFJ26/PRG173

A 5 ml culture of pRG173 transformed in E. coli strain RFJ26 was grown overnight in LB medium, supplemented with 100 mg/L ampicillin at 37°C with aeration. The overnight culture was diluted into 500 ml LB and grown to OD₅₉₀ = 5.3

SUBSTITUTE SHEET at which time the expression of hNT-4 was induced by the addition of lactose to 1% (w/v) and the culture was grown overnight with aeration. Cells were then collected by centrifugation and frozen at 70°C. The cell pellet (7.2 grams) was thawed and resuspended in 73 ml of 200 mM Tris-HCl, pH 8.0, 50 mM EDTA and lysed by 3 sequential passes through the STANSTED® cell disrupter. The lysate was then centrifuged at 26,000 x g for 10 minutes, yielding 83 ml of supernatant containing the soluble fraction. The soluble fraction was dialyzed against 50 mM Tris-HCl, pH 8.5 at 4°C for 5 hrs then diluted 10-fold with 20 mM MES, pH 6.0 and loaded on a Fast S-Sepharose column equilibrated with 20 mM MES, pH 6.0 and eluted with 1 M NaCl in 20 mM MES, pH 6.0. Recombinant human.NT-4 protein that stimulated E8 DRG outgrowth was recovered in the 1 M NaCl wash.

The insoluble fraction was resuspended and homogenized in 83 ml 8 M guanidinium-HCl, 50 mM Tris-HCl, pH 8.5, 10 mM NaCl, ImM EDTA. The insoluble fraction was dialyzed against 7 M urea, 50 mM Tris-Cl, pH 8.5, and loaded onto a Fast DEAE-Sepharose column equilibrated in 7 M urea, 50 mM Tris-Cl, pH 8.5. The breakthrough fractions were collected and dialyzed against 7 M urea, 100 mM histidine, pH 5.5, 1 mM EDTA and loaded on a Fast S-Sepharose column equilibrated in 7 M urea, 100 mM histidine, pH 5.5, 1 mM EDTA, hNT-4 was eluted with a 0-1M NaCl gradient in the same buffer. Fractions containing hNT-4 were pooled and dialyzed against 100 mM histidine, pH 5.5, 1 mM EDTA. Approximately 95% of the hNT-4 protein fractionated with the insoluble material.

9.3. PREPARATION OF ENRICHED MOTOR NEURON CULTURES

Embryos (E14) from Sprague-Dawley rats (HSD or Zivic- Miller) were used for these experiments. Pregnant rats

SUBSTITUTE SHEET were sacrificed by carbon dioxide asphyxiation, and embryos were rapidly removed and placed in ice-cold medium for further dissection. Spinal cords were removed aseptically from rat embryos of 14 days gestation. The spinal cord was severed caudal to the bulb (at the level of the first dorsal root ganglion) , freed of sensory ganglia and adhering meninges. The cord was then subdivided into ventral and mediodorsal segments for separate cultures. The ventral spinal cord tissues were diced into small pieces and incubated in 0.1% trypsin (GIBCO) and 0.01% deoxyribonuclease type 1 (Sigma) in PBS at 37°C for 20 minutes. Trypsin solution was then removed, rinsed and replaced with medium consisting of 45% Eagle's minimum essential medium (MEM), 45% Ham's nutrient mixture F12 (F12) , 5% heat inactivated fetal calf serum (GIBCO) , 5% heat inactivated horse serum (GIBCO) , glutamine (2 mM) , penicillin G (0.5 U/ml) , and streptomycin (0.5 ug/ l) . The tissue was then mechanically dissociated by gentle trituration through a Pasteur pipet, and the supernatants were pooled and filtered through a nylon fiber (Nitex, Tetko; 40 μm) . The filtered cell suspension were then subjected to a modification of the fraction procedure described by Schnaar and Schaffner (1981, J. Neurosci. 1:204-217). All steps were carried out at 4°C. Metrizamide was dissolved in F12:MEM (1:1) medium, and a discontinuous gradient was established which consisted of a 18% metrizamide cushion (0.5 ml), 3ml of 17% metrizamide, 3ml of 12% metrizamide, and 3ml of 8% metrizamide was prepared. The filtered ventral spinal cord cell suspension (2.5ml) obtained as described above was layered over the step gradient, the tube was centrifuged at 2500g for 15 minutes using a swing-out rotor (Sorvall HB4) . Centrifugation resulted in three layers of cells: fraction I (at 0-8% interface) , fraction II (at 8-12% interface) , and fraction III (at 12-17% interface) . The cells from

SUBSTITUTESHEET each interface were removed in a small volume (about lml) , rinsed twice with serum-free defined medium consisting of 50% F12 and 50% MEM, supplemented with glutamine (2 mM) , insulin (5 ug/ml) , transferrin (lOOug/ l) , progesterone (20 nM) , putrescine (100 uM) , and sodium selenite (30nM) (Bottenstein and Sato, 1979, PNAS 76:514-517). Viable cell count was obtained by hemocytometer counting in the presence of trypan blue. Fractionated ventral spinal cord cells (enriched with motorneurons) were then plated at a density of 100,000 cells/cm² in 6 mm wells precoated with poly-L-ornithine (Sigma: 10 μg/ml) and laminin (GIBCO: 10 μg/ml) . Treatment with NT-4 was given on the day of plating. Cultures were maintained in serum-free defined medium at 37°C in 95% air/5% C0 atmosphere at nearly 100% relative humidity. On day 2 (48 hours) , cells were harvested for measurements of choline acetyltransferase (CAT; Fonnum, 1975, J. Neurochem. 24:407-409.

9.4. RESULTS

The soluble fraction from the purification scheme of an induced overnight culture of pRG173/RFJ26 was assayed on E8 DRG explants. Figure 13 shows a dose-dependent stimulation of neurite outgrowth from the E8 DRG explants.

The insoluble fraction was assayed on E8 dorsal root ganglia (DRG) explants and showed no bioactivity.

The recombinant human NT-4 protein recovered from the 1 m NaCl wash of the soluble fraction was assayed for activity in dissociated motor neuron cultures prepared as described supra at Example Section 9.3, and in other assays. Addition of recombinant human NT-4 at a 1:20 dilution resulted in a 3.6 fold increase in choline acetyltransferase activity in the motor neuron enriched culture 48 hours after treatment. The 3.6 fold increase was measured in relation to untreated (C-NT) and buffer (C- Buffer) controls (Figure 14) .

SUBSTITUTE SHEET 9.5. DISCUSSION Subcloning the NT-4 coding region of bacteriophage HG7-2 downstream from the LAC UV5 and T701.1 promoters, the T701.1 ribosome binding site and the LamB signal sequence resulted in plasmid pRG173. Transformation of pRG173 into the E. coli strain RFJ26 and induction of large scale cultures of pRG173/RFJ26 resulted in the expression of a biologically active form of recombinant human NT-4. The ability to express a biologically active form of human NT-4 in an in vitro prokaryotic expression system substantially increases the ease at which the production of human recombinant NT-4, peptides or derivatives thereof may be scaled up for both therapeutic and diagnostic applications.

The present example thus teaches that a DNA sequence encoding a human NT-4 is amenable to transcription and translation in a prokaryotic system such that human NT-4 is expressed and the biologically active form is amenable to purification schemes such that the activity remains subsequent to purification.

10. DEPOSIT OF MICROORGANISMS The recombinant DNA molecules pRPN133 (hNGF) , pRPN121 (hBDNFmyc) and pRPN149 (hBDNF) were deposited on June 7 , 1991, pRG173 (NT-4) on October 28, 1991 and the recombinant bacteriophage HG7-2 on August 22, 1991 with the American Type Culture Collection, 12301 Parklawn Drive, Rockville, Maryland 20852, under the Budapest Treaty on the International Recognition of the Deposit of Microorganisms for the purposes of Patent Procedure, and assigned the indicated accession number.

SUBSTITUTE SHEET RECOMBINANT DNA MOLECULE ATCC ACCESSION NO. pRPN133 (hNGF) 75029 pRPN121 (hBDNFmyc) 75028 pRPN149 (hBDNF) 75027 pRG173 (hNT-4) 75131

HG7-2 (human NT-4 genomic clone) 75070 The present invention is not to be limited in scope by the deposited microorganisms or the specific embodiments described herein. Indeed, various modifications of the invention in addition to those described herein will become apparent to those skilled in the art from the foregoing description and accompanying figures. Such modifications are intended to fall within the scope of the appended claims.

Various publications have been cited herein which are incorporated by reference in their entireties.

While we have hereinbefore described a number of embodiments of this invention, it is apparent that our basic embodiments can be altered to provide other embodiments which utilize the processes and compositions of this invention. Therefore, it will be appreciated that the scope of this invention includes all alternative embodiments and variations which are defined in the foregoing specification and by the claims appended hereto; and the invention is not to be limited by the specific embodiments which have been presented herein by way of example.

SUBSTITUTE SHEET

Claims

WE CLAIM:

1. A process for recovering a recombinant neurotrophin produced in a non-animal host cell culture comprising the step of solubilizing the neurotrophin in a solution comprising a strong denaturing agent, said solution being essentially free of reducing agents.

2. The process of claim 1 wherein the strong denaturing agent is a 4 M - 9 M guanidinium salt, a .4 M - 9 M alkali metal thiocyanate or 7 M - 9 M urea.

3. The process of claim 2 wherein the strong denaturing agent is 7 M - 9 M guanidinium HCl.

4. The process of claim 3 wherein the strong denaturing agent is 8 M guanidinium HCl.

5. The process of claim 1 further comprising the step of exchanging the strong denaturing agent for a weak denaturing agent.

6. The process of claim 5 wherein the weak denaturing agent is 4 M - 9 M urea.

7. The process of claim 6 wherein the strong denaturing agent is 7 M - 9 M guanidinium HCl and the weak denaturing agent is 6 M - 8 M urea.

8. The process of claim 7 wherein the strong denaturing agent is 8 M guanidinium HCl and the weak denaturing agent is 7 M urea.

9. The process of claim 1 wherein the solution further comprises a metalloproteinase inhibitor in an

SUBSTITUTESHEET amount effective to inhibit degradation of the neurotrophin by proteases.

10. The process of claim 2 wherein the solution further comprises a metalloproteinase inhibitor in an amount effective to inhibit degradation of the neurotrophin by proteases.

11. The process of claim 10 wherein the solution further comprises 1 mM - 200 mM EDTA.

12. The process of claim 3 wherein the solution further comprises 5 mM - 80 mM EDTA.

13. The process of claim 4 wherein the solution further comprises 50 mM EDTA.

14. The process of claim 7 wherein the solution further comprises 5 mm - 80 mm EDTA.

15. The process of claim 8 wherein the solution further comprises 50 mm EDTA.

16. The process of claim 5 further comprising the steps of (1) adjusting the solution to comprise a basic amino acid or indole acetic acid at a concentration effective to maintain the solubility of the neurotrophin in a non-denaturing environment and (2) removing the weak denaturing agent.

17. The process of claim 6 further comprising the steps of (1) adjusting the solution to comprise a basic amino acid or indole acetic acid at a concentration effective to maintain the solubility of the neurotrophin in

SUBSTITUTESHEET a non-denaturing environment and (2) removing the weak denaturing agent.

18. The process of claim 6 further comprising the steps of (1) adjusting the solution to comprise 20 mM - 500 mM histidine and (2) removing the weak denaturing agent.

19. The process of claim 7 further comprising the steps of (1) adjusting the solution to comprise 20 mM - 100 mM histidine and (2) removing the weak denaturing agent.

20. The process of claim 8 further comprising the steps of (1) adjusting the solution to comprise 20 mM - 100 mM histidine and (2) removing the weak denaturing agent.

21. The process of claim 14 further comprising the steps of (1) adjusting the solution to comprise 20 mM - 100 mM histidine and (2) removing the weak denaturing agent.

22. The process of claim 15 further comprising the steps of (1) adjusting the solution to comprise 20 mM - 100 mM histidine and (2) removing the weak denaturing agent.

23. The process of claim 3, 4, 7, 8, 12, 13, 14, 15, 19, 20, 21 or 22 for recovering recombinant hNGF, hBDNF or hNT-3.

24. The process of claim 23, for recovering recombinant hNGF having threonine substituted for serine as

SUBSTITUTE SHEET the second amino acid residue, full length hBDNF and full length hNT-3.

25. A process for producing a recombinant neurotrophin comprising the step of culturing a nonanimal host cell transformed with a recombinant DNA molecule comprising a controllable expression control sequence operatively linked to a DNA sequence encoding a neurotrophin wherein said host cell is a heat shock regulatory gene mutant.

26. The process of claim 25 wherein the host cell is an E. coli HtpR¹⁵ Ion' mutant strain.

27. The process of claim 26 wherein the host cell is an E. coli HtpR*³ Ion' mutant strain.

28. The process of claim 27 wherein the host cell is E. coli strain LC137.

29. The process of claim 26 wherein the expression control sequence comprises a λ P_L promoter.

30. The process of claim 27 wherein the expression control sequence comprises a λ P promoter.

31. The process of claim 30 for producing hNGF.

32. The process of claim 31 for producing an hNGF in which threonine replaces serine as the second amino acid residue.

33. The process of claim 25 comprising the step of culturing E. coli strain LC137 transformed with pRPN133 (hNGF)

SUBSTITUTESHEET 34. The process of any of claims 26-30 for producing hNGF, hBDNF or hNT-3.

35. A process for producing a recombinant neurotrophin comprising the step of culturing a non-animal host cell transformed with a recombinant DNA sequence comprising an expression control sequence operatively linked to a DNA sequence comprising a fused gene encoding, from 5' to 3' , a signal sequence in-frame with the neurotrophin.

36. The process of claim 35 wherein the host cell is E. coli and the signal sequence is LamB or a modified LamB.

37. The process of claim 36 wherein the expression control sequence comprises a lacUVS promoter and a T7 Φl.l ribosome binding site.

38. The process of claim 37 wherein the signal sequence is LamBl (SEQ ID NO:2), LamB2 (SEQ ID NO:3), LamB3 (SEQ ID NO:4) or LamB4 (SEQ ID NO:5).

39. The process of claim 38 for producing full length hBDNF.

40. The process of claim 39 wherein the signal sequence is LamBl (SEQ ID NO:2) .

41. The process of claim 38 for producing hBDNFmyc.

42. The process of claim 41 wherein the signal sequence is LamB2 (SEQ ID NO:3) or LamB4 (SEQ ID NO:5). 43. The process of claim 35 comprising the step of culturing E. coli strain I^qF^'W1330 transformed with pRPN149(hBNDF) .

44. The process of any of claims 35-38 for producing hNGF, hBDNF or hNT-3.

45. A process for recovering a recombinant protein produced by a non-animal host cell culture, said protein having a pH of at least 9.0 and having at least two disulfide bonds, comprising the steps of solubilizing the protein in a solution comprising strong denaturing agent, said solution being essentially free of a reducing agent.

46. Plasmid pRPN133(hNGF) .

47. Plasmid pRPN121(hBDNFmyc) .

48. Plasmid pRPN149(hBDNF) .

49. A purified, homogeneous neurotrophin recovered by the process of claim 1.

50. A purified, homogeneous neurotrophin recovered by the process of claim 5.

51. A purified, homogeneous neurotrophin recovered by the process of claim 9.

52. A purified, homogeneous neurotrophin recovered by the process of claim 10.

53. A purified, homogeneous neurotrophin recovered by the process of claim 16.

SUBSTITUTESHEET 54. A purified, homogeneous neurotrophin recovered by the process of claim 17.

55. The purified, homogeneous neurotrophin of any of claims 49-54 which is hNGF, hBDNF or hNT-3.

56. The purified, homogeneous neurotrophin of claim 55 which is an hNGF in which threonine replaces serine as the second amino acid residue.

57. The purified, homogeneous neurotrophin of claim 55 which is a full length neurotrophin.

58. A gene for a modified LamB signal sequence gene comprising a DNA sequence encoding a LamB signal sequence having at least one degenerate substitution of A or T for G or C.

59. The gene of claim 58 having 7 degenerate substitutions.

60. The gene of claim 59 which is LamBl (SEQ ID NO:2) or LamB2 (SEQ ID N0:3).

61. A gene for a modified LamB signal sequence gene comprising a DNA sequence encoding a LamB signal sequence having a hydrophobic core region of 11-20 amino acid residues.

62. The gene of claim 61 comprising a core region of 14 amino acid residues.

63. The gene 6f claim 62 wherein the core region comprises the peptide LAVL (SEQ ID NO:6).

SUBSTITUTE SHEET 64. The gene of claim 63 which is LamB3 (SEQ ID N0:4) or LamB4 (SEQ ID NO:5).

65. The process of claim 3 for recovering recombinant neurotrophin—4.

66. The process of claim 65 for recovering human neurotrophin-4.

67. The process of claim 66 for recovering human neurotrophin—4 encoded by a DNA sequence as contained in bacteriophage HG7-2 as deposited with the ATCC and assigned accession number 75070, and illustrated in Figure 11 (SEQ ID NO:22).

68. The process of claim 66 for recovering human neurotrophin—4 encoded by a DNA sequence as contained in DNA plasmid pRG173 as deposited with the ATCC and assigned accession number 75131, and illustrated in Figure 12.

69. The process of claim 25 for producing neurotrophin—4.

70. The process of claim 69 for producing human neurotrophin-4.

71. The process of claim 70 wherein the human neurotrophin—4 is encoded by a DNA sequence as contained in bacteriophage HG7—2, as deposited with the ATCC and assigned accession number 75070 and illustrated in

Figure 12 (SEQ ID NO:22) .

72. The process of claim 70 wherein the human neurotrophin—4 is encoded by a DNA sequence as contained in

SUBSTITUTE SHEET E. coli plasmid vector, pRG173, as deposited with the ATCC and assigned accession number 75131 and as illustrated in Figure 12.

73. The process of claim 25 comprising the step of culturing E^ coli strain RFJ26 transformed with

E. coli plasmid vector, pRG173, as deposited with the ATCC and assigned accession number 75131, having the sequence as illustrated in Figure 12.

74. The process of claim 35 for producing neurotrophin—4.

75. The process of claim 74 for producing human neurotrophin—4.

76. The process of claim 75 wherein the human neurotrophin—4 is encoded by a DNA sequence as contained in bacteriophage HG7—2, as deposited with the ATCC and assigned accession number 75070 and illustrated in

Figure 12.

77. The process of claim 75 wherein the human neurotrophin—4 is encoded by a DNA sequence as contained in E. coli plasmid vector, pRG173, as deposited with the ATCC and assigned accession number 75131, as illustrated in Figure 12.

78. The process of claim 35 comprising the step of culturing E_j_ coli strain RFJ26 transformed with

E. coli plasmid vector, pRG173, as deposited with the ATCC and assigned accession number 75131, having the sequence as illustrated in Figure 12.

79. The DNA plasmid vector pRPN91.

SUBSTITUTE SHEET 80. The DNA plasmid pRG173 as deposited with the ATCC and assigned accession number 75131, having the sequence as illustrated in Figure 12.

81. A purified, homogeneous neurotrophin—4 recovered by the process of claim 49.

82. The purified homogenous neurotrophin-4 of claim 81 which is human neurotrophin—4.

83. The purified, homogeneous neurotrophin-4 of claim 82 encoded by a DNA sequence as contained in the bacteriophage HG7—2, as deposited with the ATCC and assigned accession number 75070, and illustrated in Figure 11 (SEQ ID NO:22).

84. The purified, homogeneous neurotrophin-4 of claim 82 encoded by a DNA sequence as contained in

E. coli plasmid vector, pRG173, as deposited with the ATCC and assigned accession number 75131, as illustrated in Figure 12.

85. A process for producing human NT-4 comprising growing a bacterium containing a recombinant nucleic acid encoding human NT-4 under conditions such that the human NT-4 is expressed by the bacterium, and recovering the produced human NT-4.

86. The process according to claim 85 in which the human NT-4 is biologically active.

SUBSTITUTE SHEET