Recombinant Human and Rat Ciliary Neurotrophic Factors

Journal of Neurochemistry Raven Press, Ltd., New York 0 199 I International Society for Neurochemistry

Recombinant Human and Rat Ciliary Neurotrophic Factors

Piotr Masiakowski, Haoxing Liu, Czeslaw Radziejewski, *Friedrich Lottspeich, *Walter Oberthuer, Vivien Wong, Ronald M. Lindsay,

Mark E. Furth, and Nikos Panayotatos

Regeneron Pharmaceuticals Inc., Tarrytown, New York, U.S.A.; and *Max-Planck Institute for Biochemistry, Plunegg-Murtinsried, F.R. G.

Abstract: The human ciliary neurotrophic factor (CNTF) gene was identified and cloned, based on homology with the recently cloned rat cDNA. The gene encodes a protein of 200 amino acids, which shares about 80% sequence identity with rat and rabbit CNTF and, like these homologues, lacks an apparent secretion signal sequence. The human CNTF gene, like the rat gene, appears to contain a single intron separating two protein coding exons. An intronless human CNTF gene was constructed by the use of polymerase chain reactions and introduced into vectors designed for expression of foreign proteins in E. coli. The rat CNTF gene was also introduced into similar vectors. Both the human and rat proteins were expressed at exceptionally high levels, at 20-40%

and 60-70% of total protein, respectively. Extraction of the recombinant proteins from inclusion bodies by guanidinium chloride, followed by two column chromatography steps, produced high yields of pure CNTF that supported survival and neurite outgrowth from embryonic chick ciliary neurons in culture. The biological activity of both recombinant proteins was comparable to that of native rat CNTF. Key Words: Genetic engineering-Antibiotic resistance-Polymerase chain reaction-Plasmid-DNA-Protein purification- Inclusion bodies. Masiakowski P. et al. Recombinant human and rat ciliary neurotrophic factors. J. Neurochem. 57, 1003- 1012 (1991).

Several specialized neurotrophic factors are expected to participate in the structural and functional com- plexity of the nervous system (Greene and Shooter, 1980; Snider and Johnson, 1989; Walicke, 1989). One such activity, termed ciliary neurotrophic factor (CNTF), capable of supporting the survival of dissociated chick ciliary ganglion neurons in culture, was identified from several sources (Manthorpe and Varon, 1985), including sciatic nerve (Williams et al., 1984). CNTF was purified to homogeneity from rat and rabbit sciatic nerve, and the corresponding cDNAs were cloned from both species (Lin et al., 1989; Stockli et al., 1989). CNTF does not show significant amino acid sequence homology to other known proteins and, in particular, it is entirely distinct from nerve growth factor (NGF) and the NGF-related factors, brain-derived neurotrophic factor and neurotrophin-3 (Leibrock et al., 1989; Hohn et al., 1990; Maisonpierre et al., 1990). CNTF also differs from these three neurotrophic factors both in its activity and in the absence of a secretory signal sequence.

The effects of CNTF observed in vitro suggest that this factor could play important roles in neuronal development. Its activities in primary cultures include the following: the support of the survival of neurons from ciliary and certain sensory and sympathetic ganglia (Adler et al., 1979; Barbin et al., 1984); the induction of cholinergic differentiation of sympathetic neurons (Ernsberger et al., 1989; Saadat et al., 1989); and the induction of type-2 astrocyte differentiation of bi- potential glial precursor cells (Hughes et al., 1988; Lil- lien et al., 1988). On the other hand, the apparent absence of a known secretory mechanism, the localization of CNTF as a soluble cytoplasmic protein in nerves, and its expression in mammalian cells point to an important pathophysiological role after release from injured tissue (Manthorpe and Varon, 1985; Lin et al., 1989; Stockli et al., 1989). This notion is supported by the recent observation that local application of CNTF prevents lesion-induced death of motor neurons of the rat facial nerve (Sendtner et al., 1990).

To explore the potential involvement of CNTF in

~~

Received August 30, 1990; final revised manuscript received Feb- ruary 12, 1991; accepted February 15, 1991.

Address correspondence and reprint requests to Dr. N. Panayotatos at Regeneron Pharmaceuticals Inc., 777 Old Saw Mill River Rd., Tarrytown, NY 1059 1, U.S.A.

Abbreviations used. CNTF, ciliary neurotrophic factor; DRG, dorsal root ganglia; E, embryonic day; NGF, nerve growth factor: PBS, phosphate-buffered saline; PCR, polymerase chain reaction; TFA, trifluoroacetic acid.

1003

I004 P. MASIAKOWSKI ET AL.

the development and regeneration of the human nervous system, we sought to identify and clone a human gene homologous to rat CNTF. The expression of the product of this gene in bacteria and the characterization of its biological activity, described here, establish the existence of human CNTF. Because the sequence of CNTF shows no apparent consensus sites for post- translational processing or modification, the possible production of the protein using prokaryotic expression vectors seemed particularly attractive. Here, we report the expression of both human and rat CNTFs in E. coli at unusually high levels, and the purification of active CNTF from inclusion bodies by methods suit- able for large-scale production.

MATERIALS AND METHODS Bacterial strain

E. coli W3 1 10lucIqF-, a strain that overproduces the lactose operon repressor, has been used previously (Panayotatos, 1988).

Molecular cloning and sequencing of the human gene Oligodeoxyribonucleotides used as primers for polymerase

chain reactions (PCRs) and sequencing were synthesized in a Cyclone Plus DNA Synthesizer (MilliGen/Biosearch). PCR was performed for 35 cycles using the GeneAmp kit and DNA Thermal Cycler (Perkin-Elmer/Cetus), essentially as recommended by the supplier. When this approach was used to produce radioactive probes, cold dCTP was replaced by 50-100 pCi [ L U - ~ ~ P I ~ C T P (3,000 Ci/mmol; NEN) and the number of cycles was reduced to six. Screening of the human genomic library in EMBL-3 SP6/T7 (Clontech) and isolation and purification of recombinant phage were done according to standard procedures (Ausubel et al., 1989). To rapidly sequence PCR products excised from agarose gels, we used the FASTaq kit (Int. Biotech., Inc.) with DNA in melted 2% NuSieve agarose gel (FMC BioProducts). Sequenase Version 2.0 kit (USB Corp.) was used for DNA sequencing with 10 pg of purified recombinant phage or 4 pg of supercoiled plasmid DNA as a template.

Expression vectors The vector plasmids pCP93, pCPl10, pblk0, and pblkl

have been used in previous studies (Panayotatos, 1988; Pan- ayotatos et al., 1989). The coding sequences for the human and rat CNTF genes were introduced into these plasmids to generate CNTF expression vectors as follows.

Rut CNTF. pRPNl1: The complete rat CNTF gene was obtained as a 622-bp fragment from a cDNA clone (Stockli et al., 1989) by PCR. The 5’ synthetic oligodeoxyribonucleotide primer used to obtain this fragment was designed as for the human gene (see Results), whereas the 3‘ primer cor- responded to the sequence 19 bp beyond the TAG termination codon. This fragment then ligated with a 3,920-bp restriction fragment from pCP93 that had been prepared by restriction with Sull, treatment with S1 nuclease to remove the protruding ends and purification by agarose gel electrophoresis. After transformation in E. coli W3 1 10lucIqF- and selection for antibiotic resistance conferred by the plasmid, colonies carrying the desired molecules were identified by restriction mapping of their plasmids and tested for the presence of a protein of the expected size and for CNTF activity.

Upon further characterization by DNA and protein se-

quencing, a single base pair mutation was found in the CNTF gene in pRPNl1, relative to the original rat cDNA, leading to incorporation of asparagine in place of tyrosine at position 193 of the protein sequence. This mutation, which must have arisen during PCR amplification, was carried over into all other vectors carrying the rat CNTF gene, but was corrected to the original sequence in pRPNl10 (see below). This mutation did not alter the expression levels or the physical and biological properties of the recombinant protein in any detectable manner.

pRPN 12: This plasmid is identical to pRPNl1 except for a mutation in the copy control region (copl) that increases the copy number in host cells approximately fivefold (Pan- ayotatos, 1987); it was constructed by replacing the DNA between the EugI and PvuI sites (clockwise) in pRPN 1 1 with the same sequence from pCPl10 (Panayotatos et al., 1989).

pRPN37: The DNA region between the two h e 1 restriction sites spanning the P-lactamase gene in pRPN 12 (Fig. 4) was replaced with a DNA segment conferring resistance to kanamycin (kanR). In this vector, the kunR gene is under the transcriptional control of its wild-type promoter.

pRPN38: This plasmid is identical to pRPN37, except for a 4-bp mutation that minimizes the strength of the kanR promoter (Panayotatos, 1988).

pRPN 1 10: The DNA region between the unique NheI and Eugl sites in pRPN12 was replaced with a PCR fragment obtained with synthetic primers spanning these sites. The 3‘ primer also spanned the single base mutation in pRPN 12 and carried the correct sequence.

Human CNTF. pRPN32: The complete human CNTF gene was reconstructed as a 6 12-bp DNA fragment from a cDNA clone, as described under Results. This fragment was then ligated with a 3,636-bp restriction fragment from pCP93 that had been obtained by restriction with SulI, treatment with S1 nuclease to remove the protruding ends, digestion with EugI, and purification by agarose gel electrophoresis.

pRPN33, pRPN39, and pRPN40: These plasmids are analogous to pRPN 12, pRPN37, and pRPN38, respectively, and were constructed in the same fashion, using pRPN32 as the parent plasmid. The copy control region between the Eugl and PvuI sites (clockwise) in pRPN32 was replaced with the same sequence from pCPl10 to create pRPN33. Then the P-lactamase coding region between AseI restriction sites in pRPN33 was replaced with the mutated kunR region of pblkl (Panayotatos, 1988) to create pRPN40. Finally, the region between the NdeI and BglII restriction sites of pRPN40 and pblkO (Panayotatos, 1988) was exchanged to create pRPN39, in which the kunR gene is under the control of its wild-type promoter.

Induction of protein synthesis and rapid extraction Cells were shaken in LB broth at 37°C to an ODsgo of 1,

lactose was added to 1% (wt/vol), and incubation was con- tinued for 16-20 h, as described (Panayotatos, 1988).

For crude lysates of “soluble” protein, cell pellets were resuspended in 0.16 ml of phosphate-buffered saline (PBS) ( 1.44 g of Na2HP04 - 2H20, 0.2 g of KH,PO,, 8 g of NaC1, 0.2 g of KCI per liter, final pH 7.2) and 20 pg of lysozyme (2 p l , 10 mg/ml stock) were added. After 30 min on ice, the samples were subjected to three cycles of freezing at -7O”C, followed by thawing at 37”C, and cleared by spinning at top speed in the microcentrifuge at 5 “C for 15 min. The supernatant representing “soluble” protein was diluted fivefold in PBS and assayed for biological activity.

Total cellular protein for gel electrophoresis was prepared

J. Nntrochcm , Vo/ 5 7, Nu. 3, 1991

RECOMBINANT CNTF FROM E. COLI 1005

by resuspending cell pellets from 0.5 ml of culture (OD590 = 2) in 0.16 ml of lysis buffer (100 mM Tris-HC1, 10% glyc- erol, 4% sodium dodecyl sulfate, 1 mM dithiothreitol, 0.5 mg/ml bromophenol blue, pH 6.8) and boiling for 5 min.

Selective extraction/solubilization The method initially described for recovery of recombinant

human leukocyte interferon a2 from inclusion bodies (Thatcher and Panayotatos, 1986) was used, modified as follows. Cells from induced cultures were resuspended and stored frozen at -20°C or lower. Following lysozyme treatment, the viscous suspension was passed through a French Press (SLM-Aminco) at 8,000 psi and centrifuged at 11,000 g. Inclusion bodies present in this pellet were then washed with 2 M guanidinium chloride, solubilized by 8 M guanidinium chloride, and processed, as described (Thatcher and Panayotatos, 1986). After exhaustive dialysis against buffer D (10 mM Tris-HC1, pH 8.0, 5 mM EDTA, 0.1 mM dithiothreitol) and centrifugation at 1 1,000 g, the clear supernatant was passed aseptically through a 0.22-pm Sterifil D-GV filter (Millipore).

Chromatography The filtrate containing solubilized CNTF was applied to a

5 X 10 cm DEAE-Sephacel column (Pharmacia) equilibrated with buffer E (20 mM Tris-HC1, pH 8.0, 0.1 mM EDTA, I mkfdithiothreitol, 25 mMNaCI). After washing with 10 bed volumes of buffer E, 600 ml of a 25-500 mM NaCl linear gradient in buffer E were applied. CNTF eluted at 150-250 mM NaCl.

Pooled fractions from the DEAE-Sephacel column were adjusted to 5 mM potassium phosphate, pH 5.5, 0.1 mM EDTA, and 1 mMdithiothreito1 (buffer G) and were applied on a ZETA PREP 60-mm SP-Disc (CUNO). After washing with 10 bed volumes of buffer G, a 600-ml 0-1 M NaCl gradient in buffer G was applied. CNTF eluted at 600-700 mM NaCI. Peak fractions were dialyzed against buffer E, sterilized by filtration, and stored at -7OOC.

Peptide analysis Rat CNTF. Recombinant rat CNTF (50 p g ) was subjected

to cleavage by CNBr as described (Stockli et al., 1989). The resulting peptides were separated by reverse-phase HPLC using a C4 column, and the chromatographic pattern was compared with that of CNBr-cleaved rat sciatic nerve CNTF. The peptide previously identified as the most C-terminal was subjected to amino acid analysis. In addition, the N-terminal peptide was identified and subjected to automated gas-phase microsequencing on an Applied Biosystems 477A sequencer.

Human CNTF. Recombinant human CNTF (10 pg) in 1.5 ml of 0.1 % trifluoroacetic acid (TFA)/50% acetonitrile was concentrated in a Speedvac to a final volume of 300 pl. The sample was loaded onto a minicolumn (Vydac C-4,2 14 TPB, 300 A, 10 pm), washed twice with 0.1% TFA/lO% acetonitrile, and eluted with 0.1% TFA/70% acetonitrile. The eluate was concentrated in the Speedvac to approximately 10 pl . The N-terminal peptide sequence was determined by cleavage with CNBr, as described above for rat CNTF. The C-terminal sequence was determined by cleavage with carboxypeptidase Y and P (Boehringer Mannheim, sequencing grade) in 20 p1 oftotal volume at 33°C. Cleavage was camed out at pH 3.79, 5.0, and 6.12 adjusted with sodium citrate (final concentration of 25-50 mM). After incubation for de- fined periods varying between 10 and 65 min, 3 pl of 99% formic acid and 200 pmol of aminoethanol (internal standard) were added, and the samples were applied to the minicolumn,

as above. Cleaved amino acids were eluted, and the column was washed twice with 0.1% TFA/10% acetonitrile. The amino acids were dried and analyzed after derivatization with o-phthaldialdehyde. CNTF was eluted from the column with 0.1% TFA/70% acetonitrile and concentrated to 10 pl, and cleavage was repeated.

Biological activity assays Recombinant CNTF was assayed on explants of chick em-

bryo dorsal root ganglia (DRG) and dissociated cultures of ciliary ganglion neurons as described (Adler and Varon, 1980; Lindsay and Rohrer, 1985). Briefly, DRG were dissected from chick embryos of 10 days incubation [embryonic day 10 (ElO)], and five to six ganglia were explanted in 1 ml of a collagen gel matrix in 35-mm tissue culture dishes. After the gel had set, 1 ml of tissue culture medium F14 (Imperial Labs., U.K.), supplemented with 5% heat-inactivated horse serum (GIBCO), was added before adding CNTF (1-20 p l ) to a final concentration of 0.1-100 ng/ml. In such explant cultures, CNTF activity was determined by assessing the extent of fiber outgrowth in treated cultures compared to con- trols. Fiber outgrowth was scored on a scale of 0 to 5+ by comparing cultures to photographs of a dose-response of explanted DRG to NGF (Lindsay and Rohrer, 1985).

Ciliary ganglia were dissected from E8 chick embryos and incubated for 30 min in 0.25% trypsin (Worthington) in cal- cium- and magnesium-free PBS. The ganglia were then washed three times in F14 medium containing 5% horse serum before being dissociated to single-cell suspension by trituration through the bore of a Pasteur pipette. Enrichment for ciliary ganglion neurons was achieved by plating the cell suspension for 3.5 h in a 60-mm dish, during which non- neuronal cells (fibroblasts and Schwann cells) attached firmly to the plastic while phase-bright neurons remained in suspension. The purified neurons were plated on polyornithine- laminin-coated 35-mm culture dishes at 4,000-5,000 neurons/dish. In such dissociated neuron cultures, CNTF activity was determined by counting at 48 h the percentage of process- bearing neurons in control and CNTF-treated cultures (Lindsay and Rohrer, 1985). In all cases, results were derived from triplicate cultures.

Other methods Conditions for enzymatic reactions, DNA electrophoresis,

and other techniques used in these studies have been described in detail (Panayotatos, 1987).

RESULTS AND DISCUSSION

Cloning of the human CNTF gene Southern blot hybridization of a rat CNTF probe

(Stockli et al., 1989) to human genomic DNA digested with EcoRI revealed a single weakly hybridizing band of about 10 kb, indicating the probable existence of a human CNTF gene in the fragment. In order to obtain a better probe for this gene, we employed a strategy based on PCR. Several pairs of oligodeoxyribonucleotides, corresponding to selected parts of the rat CNTF cDNA sequence (Stockli et ai., 1989), were used as PCR primers with human genomic DNA as a template. One of these pairs (CNTF. 10 and CNTF. 1 1 ; see Fig. 1) successfully primed the amplification of a DNA fragment, H10-11, of the size (270 bp) expected from the rat gene sequence. Sequencing of this amplified

J. Neurochem., Vol. 57, No. 3, 1991

1006 P. MASIAKOWSKI ET AL.

10 20 30 40 50 60

MetAlaPheThrGluHisSerProLeuThrPro TGTTTCCTGGGACAGTTGAGTT?+?+GGG~GGCTTTCACAGAGCATTCACCGCT~CCCT

> CNTF. 23

HisArgArgAspLeuCysSerArgSerIleTrpLeuAlaArgLysIleArgSerAspLeu CACCGTCGGGACCTCTGTAGCCGCTCTATCTGGCTAGCAAGGAAGATTCGTTCAGACCTG

[GTAAGTTGCCTATTT . . . . TGTATCCTCGGCCAGI W

ThrAlaLeuThrGl USerTyrValLysHisGl nGlyLeuASnLySAsnI1 eASnLeuASp ACTGCTCTTACGGAATCCTATGTGAAGCATCAGGGCCTGAACAAGMCATCAACCTGGAC

CNTF.21

CNTF.22 >

SerAlaAspGlyMetProValAlaSerThrAspGlnTrpSerGl ULeuThrGluAlaGl u TCTGCGGATGGGATGCCAGTGGCAAGCACTGATCAGTGGAGTGAGCTGACCGA-GAG

. . . . . . . . . . . . . . . . . . . . . . . CNTF .I1

ArgLeuG1 n G l u A s n r . e u G I n A l a T y r A r g T 1 : r P t e H ~ sVa?LeuLeuAlaArgLeuLeu CGACTCCAAGAGAACCTTCMGCTTATCGTACCTTCCATGTTTTGTTGGCCAGGCTCTTA

GI uAspGInGl nVd I HisPteThrProThrGl uGlyP.spPhc HisGlnAlalleHisThr GAAGACCAGUGGTGCATTTTACCCCAACCGAAGGTGACTTCCATCAAGCTATACATACC

LeuLeuLeuGlnVdlAlaAIaPheAlaTyrGl n I l eGl uGl U t e JMet I 1 el.euLeuG1 u CTTCTTCTCCMGTCGCTGCCTTTGCATACCAGATAGAGGAGTTMTGATACTCCTGGM

Tyriys?leP:oArgAsnGl ;AlaAspGly?l~tProI1eAsnValu^!yAs~GlyGly TACAAGATCCCCCGCAATGAGGCTGATGGGATGCCTATTAATGTTGGAGATGGTGGTCTC

< ****..t++*l****t***

CNTF.10

PheGluLysLysLeuTrpily LeJLysValLcuGInSl uLeuSerClnTrp?'hrValArg TTTGAGMGAAGCTGTGCTAUGGTGCTGCAGGAGCTTTCACAGTGGAUGTMGG

Ser I1 eHi SAspLeuArgPheIl eSerSerHi sG1 nThrGl yIl eProAl aArgGlySer TCCATCCATGACCTTCGTTTCATTTCTTCTCATGXGACTGGGATCCCNXACGTGGGAGC

HisTyrIleAlaAsnAsnLysLysMet UTTATATTGCTM~~~TGTAGCAGTTAGTCCCTTCTCTCTTCCTTACTTTC

< *--***-*- CNTF. 24

FIG. 1. Sequence of human CNTF and the PCR primers used in cloning. The DNA sequence of the protein-coding region is shown in bold with the deduced protein sequence above. Solid arrowheads indicate the last nucleotide of exon 1 and the first nucleotide of exon 2; the 5'- and 3-terminal intron sequences are shown in brackets above the arrowheads. The location and 5' to 3' polarity of selected oligodeoxynucleotide primers are shown by arrows. Asterisks denote mismatches. The oligodeoxynucleotide sequences of the primers (5' to 3 ) are as follows: CNTF.23: GCTTTCACAGAGCATTCACCG; CNTF.21: AGGCCCTGATGC- TTCACATAGGATTCCGTAAGAG; CNTF.22: CTCTTACGGAAT- CCTATGTGAAGCATCAGGGCCT; CNTF.24: GAGACGGC- CGTAACTGTTACATTTTCTTGTTGTTAG; CNTF. 10: CCA- AGCTTCTAGAATTCGCAGGCATCCCATCAGCCT; CNTF. 1 1 : GACTCGAGTCGACATCGGAGATGACTGAGGCAGA.

human DNA revealed 82% sequence identity with the corresponding region of rat CNTF. When the radio- labeled H 10- 1 I probe was hybridized to a Southern blot of human and rat genomic DNA digested with EcrtRI, a single strong band of about 10 kb was detected in human DNA and a single weak band of about 4 kb in rat DNA (data not shown). Because the relative intensities reflect the extent of sequence homology, we concluded that the H 10-1 1 fragment was derived from the human CNTF gene and used it as a probe to screen a human genomic library.

One positive phage clone, genC.l, was purified by several rounds of single plaque isolation. The presence in this clone of a DNA sequence coding for the putative N terminus of human CNTF was tested initially by PCR using oligodeoxyribonucleotide primers corre-

sponding to the 5' end of the coding region of the rat CNTF gene and to the internal region of human H 10- 1 1. A single major reaction product was observed, im- plying that the 5' end of the human coding sequence was indeed present in this bacteriophage clone, and that it was sufficiently similar to the rat sequence to permit priming. However, the reaction product had a size of approximately 1.5 kb, i.e., much larger than the size expected for an uninterrupted coding sequence, suggesting that the human CNTF gene is interrupted by at least one intron. It is noteworthy that the rat CNTF gene has been shown to possess a single intron of about 1 kb, localized upstream from the H10-11 region (P. Carroll and H. Thoenen, personal communication).

Partial sequencing of the 1.5-kb fragment and comparison with the coding sequence of rat CNTF revealed that the human CNTF gene has an intron located at exactly the same position as in the rat gene (Fig. 1). This assignment of the intron-exon boundaries was based on the perfect amino acid sequence conservation among rat, rabbit, and human CNTFs in the region immediately surrounding the exon junctions (Fig. 2), and by the presence of the dinucleotides GT and AG at the 5' and 3' ends of the proposed intron sequence, respectively (Fig. l), in agreement with the general rules for RNA splicing (Breathnach et al., 1978).

The sequence of the protein-coding region of the human CNTF gene (represented by bold type in Fig. 1) was determined initially from PCR fragments. The results were confirmed by direct sequencing, carried out with the genC. 1 genomic clone DNA as a template. Optimal alignment of this sequence with the rat CNTF sequence shows that the human gene encodes a protein of the same size (200 amino acid residues) and 84% identity with rat CNTF. The size identity is consistent with the presence of an in-frame TAA stop codon two codons upstream from the initiating codon. When the rabbit CNTF (Lin et al., 1989) is included in this comparison (Fig. 2), all three proteins share 152 (76%) identically placed residues. The rabbit protein is one amino acid shorter than its human and rat counter- parts; an optimal alignment is obtained by placing this deletion at the position corresponding to asparagine 145 in the human sequence, in a region of relatively low sequence conservation. Like the rat (Stockli et al., 1989) and rabbit (Lin et al., 1989) proteins, human CNTF lacks a consensus secretion signal sequence.

Construction of an intronless gene for bacterial expression

Demonstration of the functional identity of the human CNTF gene product required evidence of its biological activity. Because we had found that rat CNTF could be produced efficiently in E. coli (see below), we applied the same approach to the human gene product.

To attempt to express the human CNTF gene in bacteria, it was first necessary to remove the intron separating the protein-coding sequences in the genomic


human

rat rabbit

human

rat rabbit

10 20 30 40 50 60 70 80 90 100

MAFTEHSPLTPHRRDLCSR~ IWLRRKIRSDLTALTESYVKHQGLNKNINLDSADGMPVASTDQWSELTEAERLQENLQA~RTFHVLLAR~LEDQQ~FT~ ................................................................................. MAFAEQTPLTLHRRDLCSRSIWLARKIRSDLTA~SYHQG~~INLDSVDGVPVASTDRWSEHTEAERLQENLQAYRTFQGMLTKLLEDQ~VHFTP MAFMEHSALTPHRRELCSRTIWLARKIRSDLTALTESYVKHQGLNKNINLDSMGVPPlASTDQWSELTEAERLQENLQAYRTFHIMLARLLEDQQVHFTP

110 120 130 140 150 160 170 180 190 200

T E G D F H Q A I H T L L L Q V A A F A Y Q I E E ~ I L L E Y K I P ~ A D G M P I ~ G D ~ L F E K K L W G L K ~ Q E L S Q ~ ~ S I H D L ~ I S S H Q T G I P ~ G S H Y I ~ K ~

TEGDFHQAIHTLHLQVSAFAYQLEE~LLEQKIPE~AD~ATVGDGGLFE~L~LKVLQELSQWT~SIHDLRVISSHQMGISALESHYG~KQM AEGDHFQAIHTLLLQVAAFAYQIEELMVLLECNIPPKDADGTPV-IGGDGLFEKKLWGLKVLQELSHWTVRSIHDLRVISCHQTGIPAHGSHYIANDKEH

.......................................................................

FIG. 2. Alignment of amino acid sequences of human, rat, and rabbit CNTFs. The human sequence is from Fig. 1, the rat sequence is from Stackli et al. (1989), and the rabbit sequence is from Lin et al. (1989). Dots indicate residues shared by all three sequences.

clone genC. 1. This was accomplished by using PCR to amplify and join the regions flanking the intron. Two reactions, each with 100 ng of genC. 1 DNA as a template, were set up (Fig. 1): one contained 1 pM CNTF.23 primer and 10 nM CNTF.21 primer, and the other 1 pMCNTF.24 primer and 10 nMCNTF.22 primer. After 10 PCR cycles (each cycle constituting incubation for 1 min at 94"C, 2 min at 50°C, and 2 min at 72"C), the two samples were combined and subjected to another 25 cycles in the DNA Thermal Cycler. Because the internal primers CNTF.21 and CNTF.22 are fully complementary to each other, the products of the first stage PCR reactions can subsequently anneal. Furthermore, in the second stage reaction, the presence of substantially higher concentra- tions of the external primers CNTF.23 and CNTF.24 drives the synthesis of large amounts of the desired full-length product. The internal primers were chosen to bridge the two segments of the coding region, thus leading to the deletion of the intron. The products of the final PCR reaction were analyzed by agarose gel electrophoresis, and only one major band was detected by ethidium bromide staining. The size (about 600 bp) and a partial nucleotide sequence indicated that this band represented a precisely spliced coding region of human CNTF.

The 5' external primer CNTF.23 (Fig. 1) was designed to generate a blunt end coding for the second amino acid (alanine) in the CNTF sequence, so that ligation to the initiating codon provided by the vector would result in translation of the correct reading frame. The 3' external primer CNTF.24 (Fig. 1) provided an EugI restriction site 12 bp beyond the end of the CNTF coding sequence. This primer was also designed so as to replace the naturally occurring TAG termination codon with TAA, which is less subject to translational readthrough in E. coli. The PCR fragment was re- stricted with EagI, and the resulting 6 12-bp fragment was purified by polyacrylamide gel electrophoresis and inserted in pCP93, as described under Materials and Methods.

Identification of active recombinant human CNTF Analysis of protein synthesis by gel electrophoresis

in induced cultures of E. coli W3 1 10lucIqF- carrying one of the candidate plasmids (pRPN32) revealed the

presence of a protein band of an apparent MW of 27,000 that was absent in induced control cultures of bacteria carrying the pCP93 plasmid vector. Rapid protein extracts from these cultures revealed the presence of biologically active CNTF in two assays. Initially, explants of El0 chick DRG were used for rapid and semiquantitative detection of CNTF activity in E. coli lysates. In the presence of CNTF, fiber outgrowth was observed, whereas in the absence of exogenous neurotrophic factor there was little or no outgrowth from control ganglia (Fig. 3A and B). The maximum fiber outgrowth at saturating levels of CNTF was about 50- 75% of that seen with saturating levels of NGF.

A more specific assay measures neuronal survival in dissociated, neuron-enriched cultures of E8 chick ciliary ganglia. After 48 h in control cultures, almost all neurons had degenerated (Fig. 3C), whereas at least 40-50% of the neurons plated in the presence of saturating levels of CNTF survived and elaborated long neurites (Fig. 3D and E). These specific effects on ciliary neurons were observed with crude bacterial lysates diluted to estimated subnanogram amounts of soluble CNTF and with purified recombinant protein (see below). These results establish the existence of a human homologue of the CNTF genes previously identified in other mammals.

High level expression of CNTF genes Previous studies of expression of foreign proteins in

E. coli have identified several parameters that can con- tribute to high level expression and can facilitate the recovery of biologically active product. We have uti- lized expression vectors that employ several important features, including the following: regulation of the lacUV5 promoter by the lactose operon repressor; a strong ribosome binding site from bacteriophage T7; a mutation in the replication control region of the plasmid to increase copy number; and a mutation to limit the expression of the antibiotic resistance protein. We have explored specifically the effects of the latter two features on the production of CNTF in E. coli. The properties of the various vectors employed are summarized in Table 1 and shown schematically in Fig. 4.

Rat CNTF. Effect of copy number: Analysis of protein synthesis in lactose-induced cultures of E. coli W3 1 1 OlucIqF-/pRPN I 1 by gel electrophoresis revealed

J. Neurochem.. Vol. 57, No. 3, 1991


FIG. 3. Biological activity of recombinant human CNTF towards explant cultures of E l 0 chick embryo DRG (A, B) and dissociated. neuron-enriched cultures of E8 chick embryo ciliary ganglia (CG; C-E). A and B are darkfield micrographs of control (A) and CNTF-treated (B. 5 ng/ml) DRG explants after 48 h in culture. C-E are phase-contrast micrographs of control (C) and CNTF-treated (D, 100 pg/ml; E, 5 ng/ml) dissociated cultures of ciliary ganglion neurons. Bar = 400 pm (A, B) and 50 Ftn (C-E).

the relatively weak expression of an approximately 24,000-kDa polypeptide, the anticipated size for rat CNTF. A band of that size was absent in induced control cultures of bacteria carrying the pCP93 vector (Fig. 5). Most remarkably, however, in induced cultures of

TABLE 1. Properties of various vectors

Special ampR or Plasmid Common features' featuresa CNTFb kanRb

pRPNl1 1ac.rbsl.ratCNTF ump.cop+ 1-2 (0.5 pRPN 12 lacxbs 1 .ratCNTF ump.cop1 30-50 1-2 pRPN37 1ac.rbsl.ratCNTF kun0.copI 5-10 5-10 pRPN38 1ac.rbsl.ratCNTF kunl.cop1 50-70 <O. 1

pRPN32 1ac.rbsl.humCNTF amp.cop+ 1-2 <0.5 pRPN33 lac.rbsl.humCNTF ump.copI 10-20 1-2 pRPN39 1ac.rbsl.humCNTF kun0.copI 15-25 5-10 pRPN40 1ac.rbsl.humCNTF kunl.cop1 25-35 10.1

a lac, lacUV5 promoter; rbsl, ribosome binding site; ratCNTF or humCNTF, rat or human CNTF gene; amp, ampicillin resistance gene; kanO or kunl, wild type or mutated kanR genes; cop+ or copl. normal or high copy number plasmid.

As percentage of total cellular protein.

E. coli W31 10lacIqF- hosting pRPN12, which differs from pRPNl1 only by the presence of the copy number mutation copl, the production of CNTF was increased to approximately 30-50% of total cellular protein (Fig. 5). Thus, the fivefold increase in the copy number of pRPN12 relative to pRPNl1 resulted in a 30-50-fold increase in the level of recombinant protein (Table 1). This effect of the cop1 mutation has been documented with other recombinant proteins (Buell and Panayo- tatos, 1987).

Effect of antibiotic resistance: In previous studies on the expression of recombinant proteins in E. coli, it was observed that synthesis of the antibiotic resistance protein encoded by the vector interfered with optimal recombinant protein production. This interference was attributed to competition by the two genes for the limiting synthetic machinery of the cell (Panayotatos, 1988). To test this hypothesis further, and potentially to improve the levels of CNTF production, the P-lactamase gene (ampR) in pRPN 1 2 was replaced with the kanamycin resistance (kanR) gene either under the transcriptional control of its native promoter (pRPN37) or under the control of a weaker mutant promoter (pRPN38).

J Neurochem ~ Vol. 57, No. 3. 1991

RECOMBINANT CNTF FROM E . COLI 1009

lacLN5 A.

A x I rbsl

pRPN 11,12,32,33

Nde I

lacuv5 B.

Ase I & rbsl

Eag I Bgl U

(knnl) pRPN 37,38,39,40

Analysis of protein levels in induced cells hosting pRPN37 indicated that rat CNTF constituted 10-20% of total cellular protein and that the kanR protein was synthesized at approximately one half of that level. In contrast, in induced cells hosting pRPN38, which bears the mutant kanR promoter, CNTF constituted 50-70% of total cellular protein, whereas the kanR protein was undetectable (Fig. 5 and Table 1). The significantly higher expression of CNTF observed with pRPN38 presumably results directly from the decrease in the level of expression of the antibiotic resistance gene (Panayotatos, 1988).

Human CNTF. High level production of human CNTF was achieved using the same strategy as for rat CNTF. The relative levels of expression obtained with these vectors are shown in Fig. 5 and are summarized in Table 1. The maximal levels with each vector were lower than the levels observed for the analogous vectors carrying the rat gene, but the overall pattern was the same; a 30-50-fold increase was again observed with the higher copy number (cop]) plasmids, and the maximal levels were again obtained with the combination

FIG. 5. Comparison of human and rat CNTF expression using various vectors. Total protein from the indicated strains was analyzed on 8-25% polyacrylamide gels stained with Coomassie. Molecular weights of the markers (in thousands) are indicated (M), as well as the positions of bands corresponding to the human CNTF, rat CNTF, ampR, and kanR proteins.

Nde I

FIG. 4. Main features of the expression plasmids. The promoter (IacUVS), the ribosome binding site (rbsl), the CNTF, ampR, and kanR genes, as well as the cop7 (0) and kanl (0) mutations are indicated. The restriction sites Asel, Eagl, Ndel, and Pvul were used in plasmid constructions.

of high copy number and low expression of kanamycin resistance. For the expression of human CNTF, the effect of reduced kanR (pRPN39 versus pRPN40) is somewhat less striking than the effect on rat CNTF expression (pRPN37 versus pRPN38). This may be because competition of the two transcription units only becomes evident when the synthetic machinery of the cell becomes limiting, i.e., at extremely high levels of recombinant protein synthesis.

Purification of rat and human CNTF Like other recombinant proteins expressed at high

levels in E. coli, CNTF was found mostly in insoluble inclusion bodies; 85-90% of the rat protein and 60- 70% of the human protein were resistant to extraction by neutral buffer or 2 M guanidinium chloride. This was turned into advantage by efficiently washing away most of the hydrophilic proteins by these reagents. Subsequently, extraction and solubilization of CNTF and other (mostly hydrophobic) proteins trapped in inclusion bodies were effected with 8 M guanidinium chloride. Finally, slow removal of guanidinium chloride by dialysis led to precipitation of the hydrophobic host proteins and left in solution rat CNTF at better than 95% purity and human CNTF at better than 90% purity (Fig. 6). At this point, a single chromatography step (DEAE-Sephacel) was sufficient to purify recombinant rat CNTF to better than 99%, as determined from the relative intensities of intentionally overloaded polyacrylamide gels (Fig. 6A) and by HPLC analysis (data not shown). In part because of the lower expression levels and the weaker affinity of the human protein for DEAE-Sephacel, it was necessary to add a second chromatography step (Fast S) in order to achieve better than 99% purity (Fig. 6B).

Yield Given the estimated expression levels and assuming

that the total protein content of the wet cell pellet is 10% protein by weight, the theoretical yield would be 60 mg of rat CNTF and 30 mg of human CNTF per gram of wet cell pellet. The actual yield was found to be about 40 mg for rat CNTF and about 6 mg for human CNTF. Most of the loss in the case of the hu-

J. Neurochem., Vol. 57, No. 3, 1991

1010

A.

P. MASIAKO WSKI ET AL.

B.

FIG. 6. Purification of CNTF. A Rat CNTF, 8: Human CNTF. Lysate, total cellular protein; 8 M GuHCI, dialyzed extract; 0.5, 5, 10, and 50, amount of protein in micrograms loaded on each lane after DEAE-Sephacel or Fast S.

man protein is due to “soluble” CNTF extracted and discarded during the preparation of inclusion bodies and the additional purification step.

Protein characterization Recombinant CNTF, both rat and human, prepared

by the above procedures was better than 99% pure. The proteins had very low pyrogenicity (< 1 ng/mg of protein) by the Limulus Amebocyte Lysate test (As- sociates of Cape Cod). Furthermore, as discussed below, they were found to be biologically active at the picom- olar level.

Upon sequencing of the recombinant rat and human CNTF, alanine was recovered as the N-terminal amino acid with initial yields close to the expected theoretical values. These results indicate that the initiating methionine is quantitatively removed in E. coli, as is gen- erally the case when the second residue of a protein is the nonbulky amino acid alanine. By contrast, the N terminus of CNTF purified from rat sciatic nerve is blocked and presumably retains the terminal methionine residue.

At the carboxyl end, the sequence expected for recombinant human CNTF was obtained upon enzymatic C-terminal sequencing with carboxypeptidase (data not shown). By contrast, amino acid composition analysis of a C-terminal peptide obtained after CNBr cleavage of rat CNTF indicated that although it had the expected composition, it lacked a tyrosine residue and had an extra asparagine. DNA sequence analysis revealed a point mutation causing a tyrosine-to-asparagine substitution at codon 193; this apparently arose during copying of a cloned rat CNTF cDNA by PCR to construct the original expression vector pRPN 1 1. By correcting the mutation in pRPNl10 and purifying protein of identical specific activity by the exact same method, we have found that this mutation does not

affect the expression level and/or the properties of the protein in a detectable way (data not shown).

Rat and human CNTFs have the same number of amino acids (calculated MW of 22,780 and 22,700, respectively, after removal of N-terminal methionine) and share extensive parts of their sequences. Yet, on both reducing and nonreducing sodium dodecyl sulfate-polyacrylamide gels, human CNTF migrates somewhat more slowly than rat CNTF (Fig. 6). This difference in mobility between two molecules similar in structure and identical in length may reflect some unusual structural constraint in the human protein.

Biological activity CNTF is Characterized by its capacity to support the

survival of dissociated ciliary neurons of E8 chick embryos. By this criterion, purified recombinant CNTF was found to be as active as the native protein. Figure 7 shows dose-response curves of dissociated, neuron- enriched cultures of E8 chick embryo ciliary ganglia. Purified recombinant rat CNTF was found to be active at 2 pg/ml and saturation was observed at 0.5- 1 .O ng/ ml, with an ECso of 50 pg/ml or 2 pM. By comparison, activity of the protein purified from rat sciatic nerve (Stockli et al., 1989) was detectable at 5 pg/ml and maximal neuronal survival was seen at 1-2 ng/ml, with an ECso of 100 pg/ml or 4 pM (M. Sendtner, personal communication). Thus, the purified recombinant rat CNTF was at least as active as native protein purified from sciatic nerve.

In parallel experiments, purified recombinant human CNTF supported detectable neuronal survival at less than 10 pg/ml and saturation required 5 ng/ml, with an ECso of 200 pg/ml or 8 pM (Fig. 7). By this assay, recombinant human CNTF appears to be four- fold less active towards chick neurons than recombinant rat CNTF. Comparison of the dose-response

J . Neurochem.. Vol. 57, No. 3, 1991


100

rat 80

'̂i 60

20 40L @A'

0 , , . . .- , .. ...., , , ...,.., , , ......, , , ,. ,001 .01 .1 1 10 100

[CNTFl (ng/ml)

FIG. 7. Dose-response of dissociated E8 chick ciliary neuron survival in the presence of recombinant rat or human CNTF. The number of ciliary neurons that survive in the presence of increasing amounts of rat or human CNTF at 48 h in vitro is expressed as a percentage of the number of neurons originally plated. Each ex- perimental point represents the mean of three determinations.

curves in Fig. 7 also shows that the maximal levels of surviving neurons obtained with rat CNTF (65%) are higher than those obtained with human CNTF (43%). Because both the human and rat recombinant proteins were purified by the same method, the lower activity of the human protein towards chick neurons most likely reflects the greater evolutionary distance between these species.

Conclusions The results presented here establish the existence of

a human homologue of the CNTF genes previously identified in other mammals. The gene was cloned from a genomic library, and sequence analysis indicated the presence of a single intron at the same position as observed in the rat gene. The expression in bacteria of active human recombinant CNTF from an intronless gene constructed with PCR technology confirmed the assignment of the intron boundaries.

The level of human and rat CNTF production in E. coli reported here is exceptionally high. This apparently results from a favorable combination of several factors, including the use of a vector plasmid that is maintained stably at a relatively high copy number, a moderately strong promoter, a strong ribosome translational ini- tiation signal, and the minimal synthesis of the antibiotic resistance protein. In addition, recombinant protein production is determined by the physical properties of the protein itself and its stability in the host. In this respect, rat CNTF appears to be particularly amenable to expression in E. coli. Human CNTF is

also expressed very efficiently, although at somewhat lower levels.

It is noteworthy that rat, rabbit (Lin et al., 1990), and human CNTFs all display potent survival activity on embryonic chick ciliary ganglion neurons. However, a related neurotrophic molecule purified from chick sciatic nerve that possibly represents the avian homologue of CNTF is reported to have a severalfold more potent survival activity on the same neuronal population (Eckenstein et al., 1990). The relatively greater activity of this molecule towards chick neurons may well result from the evolutionary divergence of this protein.

The successful production of human CNTF in E. coli should greatly facilitate the exploration of its full range of biological activities and potential therapeutic applications. This may be of particular significance in light of the recent demonstration that CNTF can pro- mote the survival of injured motor neurons in exper- imental animals (Sendtner et al., 1990).

Acknowledgment: We thank Dr. Michael Sendtner for the experiment with native rat CNTF and for valuable discus- sions, Drs. Patrick Carroll, Kurt Stockli, Michael Sendtner, and Hans Thoenen for sharing unpublished results, and Chia- Lin Tien, Elizabeth Radziejewska, Dan Everdeen, Yang Cong, and Chris Cox for technical assistance.

REFERENCES

Adler R. and Varon S. (1980) Cholinergic neuronotrophic factors. V. Segregation of survival- and neurite-promoting activities in heart-conditioned media. Bruin Res. 188, 437-448.

Adler R., Landa K. B., Manthorpe M., and Varon S. (1979) Cholin- ergic neuronotrophic factors: intraocular distribution of trophic activity for ciliary neurons. Science 204, 1434-1436.

Ausubel F. M., Brent R., Kingston R. E., Moore D. D., Seidman J. G., Smith J. A., and Struhl K. (1989) Current Protocols in Molecular Biology. John Wiley and Sons, New York.

Barbin G., Manthorpe M., and Varon S. (1984) Purification of the chick eye ciliary neuronotrophic factor. J. Neurochem. 43, 1468- 1478.

Breathnach R., Benoist C., OHare K., Gannon F., and Chambon P. ( 1 978) Ovalbumin gene: evidence for a leader sequence in mRNA and DNA sequences at the exon-intron boundaries. Proc. Natl. Acad. Sci. USA 75,4853-4857.

Buell G. and Panayotatos N. (1986) Mechanism and practice, in From Gene to Protein: Steps Dictating the Maximal Levels of Gene Expression (Reznikoff W. S. and Gold L., eds) Butterworth, Stoneham, Massachusetts.

Eckenstein F. P., Esch F., Holbert T., Blacher R. W., and Nishi R. (1990) Purification and characterization of a trophic factor for embryonic peripheral neurons: comparison with fibroblast growth factors. Neuron 4, 623-63 1.

Ernsberger U., Sendtner M., and Rohrer H. (1989) Proliferation and differentiation of embryonic chick sympathetic neurons: effects of ciliary neurotrophic factor. Neuron 2, 1275-1284.

Greene L. A. and Shooter E. M. (1980) The nerve growth factor: biochemistry, synthesis, and mechanism of action. Annu. Rev. Neurosci. 3, 353-402.

Hohn A,, Leibrock J. , Bailey K., and Barde Y.-A. (1990) Identification and characterization of a novel member of the nerve growth factorbrain derived neurotrophic factor family. Nature 344,339- 341.

Hughes S. M., Lillien L. E., Raff M. C., Rohrer H., and Sendtner M.

J. Neurochem.. Vol. 57, No. 3, 1991


( 1988) Ciliary neurotrophic factor induces type-2 astrocyte differentiation in culture. Nature 335, 70-73.

Leibrock J., Lottspeich F., Hohn A., Hofer M., Hengerer B., Masi- akowski P., Thoenen H., and Barde Y.-A. (1989) Molecular cloning and expression of brain-derived neurotrophic factor. Nature 341, 149-152.

Lillien L. E., Sendtner M., Rohrer H., Hughes S. M., and Raff M. C. (1988) Type-2 astrocyte development in rat brain cultures is initiated by a CNTF-like protein produced by type-I astrocytes. Neuron 1,485-494.

Lin L.-F. H., Mismer D., Lile J. D., Armes L. G., Butler E. T. 111, Vannice J. L., and Collins F. (1989) Purification, cloning and expression of ciliary neurotrophic factor (CNTF). Science 246,

Lin L.-F. H., Armes L. G., Sornmer A., Smith D. J., and Collins F. ( 1990) Isolation and characterization of ciliary neurotrophic factor from rabbit sciatic nerves. J. Biol. Chem. 265, 8942-8947.

Lindsay R. M. and Rohrer H. (1985) Placodal sensory neurons in culture: nodose ganglion neurons are unresponsive to NGF, lack NGF receptors but are supported by a liver-derived neurotrophic factor. Dev. Biol. 112, 30-48.

Maisonpierre P. C., Belluscio L., Squint0 S., Ip N. Y., Furth M. E., Lindsay R. M., and Yancopoulos G. D. (1990) Neurotrophin- 3: a neurotrophic factor related to NGF and BDNF. Science

Manthorpe M. and Varon S. (1985) Regulation of neuronal survival and neuritic growth in the avian ciliary ganglion by trophic factors, in Growth and Maturation Factors, Vol. 3 (Guroff G., ed), pp. 77-1 17. John Wiley and Sons, New York.

1023- 1025.

247, 1446- I45 1.

Panayotatos N. (1987) Engineering an efficient expression system, in Plasmids: A Practical Approach (Hardy K. G., ed), pp. 163- 176. IRL Press, Oxford.

Panayotatos N. (1988) Recombinant protein production with minimal-antibiotic resistance vectors. Gene 74, 357-363.

Panayotatos N., Fontaine A,, and Backman S. (1989) Biosynthesis of a repressor/nuclease hybrid protein. J. Biol. Chem. 264, 15066-15069.

Saadat S., Sendtner M., and Rohrer H. (1989) Ciliary neurotrophic factor induces cholinergic differentiation of rat sympathetic neurons in culture. J. Cell Biol. 108, 1807- 18 16.

Sendtner M., Kreutzberg G. W., and Thoenen H. ( I 990) Ciliary neurotrophic factor prevents the degeneration of motor neurons after axotomy. Nature 345, 440-441.

Snider W. D. and Johnson E. M. (1989) Neurotrophic molecules. Ann. Neurol. 26, 489-506.

Stockli K. A., Lottspeich F., Sendtner M., Masiakowski P., Carroll P., Gotz R., Lindholm D., and Thoenen H. (1989) Molecular cloning, expression and regional distribution of rat ciliary neurotrophic factor. Nature 342, 920-923.

Thatcher D. R. and Panayotatos N. ( 1 986) Purification of recombinant IFN-a2. Methods Enzymol. 119, 166-177.

Walicke P. A. (1989) Novel neurotrophic factors, receptors, and on- cogenes. Annu. Rev. Neurosci. 12, 103-126.

Williams L. R., Manthorpe M., Barbin G., Nieto-Sampedro M., Cot- man C. W., and Varon S. (1984) High ciliary neuronotrophic specific activity in rat peripheral nerve. Int. J . Dev. Neurosci. 2, 177-180.

J Neirrochem., Vol. 57. No. 3, 1991

Recombinant Human and Rat Ciliary Neurotrophic Factors

Documents

Transcript of Recombinant Human and Rat Ciliary Neurotrophic Factors