A furin-like convertase mediates propeptide cleavage of BACE, the Alzheimer's beta -secretase.

The novel transmembrane aspartic protease BACE (for Beta-site APP Cleaving Enzyme) is the beta-secretase that cleaves amyloid precursor protein to initiate beta-amyloid formation. As such, BACE is a prime therapeutic target for the treatment of Alzheimer's disease. BACE, like other aspartic proteases, has a propeptide domain that is removed to form the mature enzyme. BACE propeptide cleavage occurs at the sequence RLPR downward arrowE, a potential furin recognition motif. Here, we explore the role of furin in BACE propeptide domain processing. BACE propeptide cleavage in cells does not appear to be autocatalytic, since an inactive D93A mutant of BACE is still cleaved appropriately. BACE and furin co-localize within the Golgi apparatus, and propeptide cleavage is inhibited by brefeldin A and monensin, drugs that disrupt trafficking through the Golgi. Treatment of cells with the calcium ionophore, leading to inhibition of calcium-dependent proteases including furin, or transfection with the alpha(1)-antitrypsin variant alpha(1)-PDX, a potent furin inhibitor, dramatically reduces cleavage of the BACE propeptide. Moreover, the BACE propeptide is not processed in the furin-deficient LoVo cell line; however, processing is restored upon furin transfection. Finally, in vitro digestion of recombinant soluble BACE with recombinant furin results in complete cleavage only at the established E46 site. Taken together, our results strongly suggest that furin, or a furin-like proprotein convertase, is responsible for cleaving the BACE propeptide domain to form the mature enzyme.

Characterization of Alzheimer's beta -secretase protein BACE. A pepsin family member with unusual properties.

The cerebral deposition of amyloid beta-peptide is an early and critical feature of Alzheimer's disease. Amyloid beta-peptide is released from the amyloid precursor protein by the sequential action of two proteases, beta-secretase and gamma-secretase, and these proteases are prime targets for therapeutic intervention. We have recently cloned a novel aspartic protease, BACE, with all the known properties of beta-secretase. Here we demonstrate that BACE is an N-glycosylated integral membrane protein that undergoes constitutive N-terminal processing in the Golgi apparatus. We have used a secreted Fc fusion-form of BACE (BACE-IgG) that contains the entire ectodomain for a detailed analysis of posttranslational modifications. This molecule starts at Glu(46) and contains four N-glycosylation sites (Asn(153), Asn(172), Asn(223), and Asn(354)). The six Cys residues in the ectodomain form three intramolecular disulfide linkages (Cys(216)-Cys(420), Cys(278)-Cys(443), and Cys(330)-Cys(380)). Despite the conservation of the active site residues and the 30-37% amino acid homology with known aspartic proteases, the disulfide motif is fundamentally different from that of other aspartic proteases. This difference may affect the substrate specificity of the enzyme. Taken together, both the presence of a transmembrane domain and the unusual disulfide bond structure lead us to conclude that BACE is an atypical pepsin family member.

A comparison of folding techniques in the chemical synthesis of the epidermal growth factor-like domain in neu differentiation factor alpha/beta.

The 52-residue alpha/beta chimera of the epidermal growth factor-like domain in neu differentiation factor (NDFealpha/beta) has been synthesized and folded to form a three disulfide bridge (Cys182-Cys196, Cys190-Cys210, Cys212-Cys221) containing peptide. We investigated two general strategies for the formation of the intramolecular disulfide bridges including, the single-step approach, which used fully deprotected and reduced peptide, and a sequential approach that relied on orthogonal cysteine protection in which specific pairs are excluded from the first oxidation step. Because there are 15 possible disulfide bridge arrangements in a peptide with six cysteines, the one-step approach may not always provide the desired disulfide pairing. Here, we compare the single-step approach with a systematic evaluation of the sequential approach. We employed the acetamidomethyl group to protect each pair of cysteines involved in disulfide bridges, i.e. Cys182 to Cys196, Cys190 to Cys210 and Cys212 to Cys221. This reduced the number of possible disulfide patterns from 15 to three in the first folding step. We compared the efficiencies of folding for each protected pair using RP-HPLC, mapped the disulfide connectivity of the predominant product and then formed the final disulfide from the partially folded intermediate via 12 oxidation. Only the peptide having the Cys182-Cys196 pair blocked with acetamidomethyl forms the desired disulfide isomer (Cys190-Cys210/Cys212-Cys221) as a single homogeneous product. By optimizing both approaches, as well as other steps in the synthesis, we can now rapidly provide large-scale syntheses of NDFealpha/beta and other novel EGF-like peptides.

Disulfide assignment of the C-terminal cysteine knot of agouti-related protein (AGRP) by direct sequencing analysis.

We have assigned the disulfide structure of Md-65 agouti-related protein (Md65-AGRP) using differential reduction and alkylation followed by direct sequencing analysis. The mature human AGRP is a single polypeptide chain of 112 amino acid residues, consisting of an N-terminal acidic region and a unique C-terminal cysteine-rich domain. The C-terminal domain, a 48 amino acid peptide named Md65-AGRP, was expressed in Escherichia coil cells and refolded under different conditions from the mature recombinant protein. The disulfide bonds in the cystine knot structure of Md65-AGRP were partially reduced using tris(2-carboxyethyl) phosphine (TCEP) under acidic conditions, followed by alkylation with N-ethylmaleimide (NEM). The procedure generated several isoforms with varying degrees of NEM alkylation. The multiple forms of Md65-AGRP generated by partial reduction and NEM modification were then completely reduced and carboxymethylated to identify unreactive disulfide bonds. Differentially labeled Md65-AGRP were directly sequenced and analyzed by MALDI mass spectrometry. The results confirmed that Md65-AGRP contained the same disulfide structure as that of Md5-AGRP reported previously [Bures, E. J., Hui, J. O., Young, Y. et al. (1998) Biochemistry 37, 12172-12177].

The intermolecular disulfide bridge of human glial cell line-derived neurotrophic factor: its selective reduction and biological activity of the modified protein.

Recombinant human glial cell line-derived neurotrophic factor has been implicated to have therapeutic potential in the treatment of neurodegenerative diseases. The mature protein is a single polypeptide of 134 amino acid residues and functions as a disulfide-linked dimer. Reduction of the protein with dithiothreitol at pH 7.0 and in the absence of denaturant showed that the single intermolecular cystine bridge was reduced preferentially. Direct alkylation of the generated free sulfhydryl group using iodoacetamide or iodoacetate without denaturant was incomplete. Unfolding the protein in 6 M guanidine hydrochloride prior to the modification showed rapid disulfide scrambling. However, the sulfhydryl-modifying reagent N-ethylmaleimide was able to label quantitatively the free cysteinyl residue in the absence of any added chaotropic agent. By a combination of peptide mapping, Edman degradation, and mass spectrometric analysis, the labeled residue was identified to be Cys101, hence verifying the location of the intermolecular disulfide bond. The modified protein behaved as a noncovalent dimer when chromatographed through a Superdex 75 column under nondenaturing conditions and was comparable in biological activity to an unmodified control sample. The results therefore indicate that the intermolecular disulfide bridge of the protein is not essential for its biological function.

Coexpression of G-CSF with an unglycosylated G-CSF receptor mutant results in secretion of a stable complex.

Previously, we have shown that the entire extracellular domain of the granulocyte-colony stimulating factor receptor (sG-CSFr) produced in Chinese hamster ovary (CHO) cells forms a stable complex with its ligand G-CSF, at a stoichiometry of 2:2. A truncated receptor molecule consisting of the cytokine receptor homology domain and N-terminus Ig-like domain (Ig CRH) behaves quite similarly. Both of these forms of the receptor are highly glycosylated. To address the importance of glycosylation toward receptor activity and stability, and possibly obtain nonglycosylated receptor for crystallization, mutations were made to replace four Asn residues which are N-glycosylated in the truncated receptor. Virtually no receptor was recovered from conditioned media of CHO cells transfected with this mutant construct, although a high-level of mRNA coding for receptor was detected; this mRNA was translated as determined by Western blots of cell lysates. These results indicate that the translated product is apparently not secreted from these cells. Cells transfected with mutant receptor cDNA were cotransfected with a cDNA construct expressing G-CSF in which the single O-glycosylation site was eliminated by mutation. Upon fermentation of the cotransfectants, we observed a large amount of receptor-ligand complex in the conditioned media. The purified unglycosylated complex appeared to be of the same binding stoichiometry and approximate binding affinity as that of complex formed by addition of purified ligand and unmutated receptor. These results show that while glycosylation of sG-CSFr is not necessary for ligand binding, it appears to be crucial in folding and export from the cell.

Identification of Asp95 as the site of succinimide formation in recombinant human glial cell line-derived neurotrophic factor.

Human glial cell line-derived neurotrophic factor is a single polypeptide of 134 amino acids and functions as a disulfide-linked dimer. Incubation of the protein in pH 5.0 and at 37 degreesC for 1 week showed that 5% of the material was converted to a form that eluted after the major protein peak on a cation-exchange column. The modified component gave an average molecular mass of 30367.0 u (theoretical = 30384.8 u). Within measurement error, this 17.8-u decrease in mass indicated the loss of a water molecule. This observation, together with the protein's behavior on cation-exchange chromatography and the mode of incubation used to generate the modification, was consistent with cyclic imide (succinimide) formation at an aspartyl residue. Hence, only a monomer of the dimeric protein was modified. The modified monomer was purified and subjected to peptic degradation. By a combination of N-terminal analysis and mass spectrometry, the region containing Asp95-Lys96 was identified to be modified. This was further confirmed by carboxypeptidase Y digestion of the modified peptide where the modified region was found to be resistant to further enzymatic degradation. Furthermore, incubation of the modified monomer in pH 8. 5 for 2 h yielded two peaks, in agreement with the succinimide model where the cyclic imide was hydrolyzed into a mixture of isoaspartate and aspartate. Tryptic mapping of the isoaspartyl-containing protein showed that Asp95 was refractory to Edman degradation, confirming it was in the isoaspartate form. Hence, the modification observed was due to succinimide formation at Asp95. This is the first report of succinimide formation at an Asp-Lys linkage.

Human leptin receptor. Determination of disulfide structure and N-glycosylation sites of the extracellular domain.

The leptin receptor (OB-R) is a member of the class I cytokine receptor family and mediates the weight regulatory effects of its ligand through interaction with cytoplasmic kinases. The extracellular domain of this receptor is comprised of two immunoglobulin-like and cytokine-receptor homology domains each and type III fibronectin domains. The extracellular domain of human leptin receptor was expressed in and purified from Chinese hamster ovary cells and was found to contain extensive N-glycosylation (approximately 36% of the total protein). The purified protein had a molecular weight of approximately 145,000 and exhibited ligand binding ability as evidenced by formation of ligand-receptor complex, followed by chemical cross-linking. The determined disulfide motif of the soluble leptin receptor contained several distinct cystine knots as well as 10 free cysteines. The N-glycosylation analysis revealed that Asn624 of the WSXWS motif (residues 622-626) within the C-terminal cytokine receptor homology domain was glycosylated, indicating that this region is solvent-exposed. On the other hand, the N-terminal WSXWS motif was not glycosylated.

Determination of disulfide structure in agouti-related protein (AGRP) by stepwise reduction and alkylation.

The agouti-related protein gene (Agrp) plays an important role in body weight regulation. The mature human protein is a single polypeptide chain of 112 amino acid residues, consisting of an N-terminal acidic region and a unique C-terminal cysteine-rich domain. The disulfide structure of recombinant human AGRP was determined by chemical methods using partial reduction with tris(2-carboxyethyl)phosphine under acidic conditions, followed by direct alkylation with N-ethylmaleimide or fluorescein-5-maleimide. Partial reduction and alkylation provided several forms of AGRP that were modified in a stepwise fashion. The resulting proteins were characterized by peptide mapping, sequence analysis, and mass spectrometry, showing that AGRP contained a highly reducible disulfide bond, C85-C109, followed by less reactive ones, C90-C97, C74-C88, C67-C82, and C81-C99, respectively. The chemically defined disulfide connectivity of the recombinant human AGRP was homologous to that of omega-agatoxin IVB except for an additional disulfide bond, C85-C109.

Interactions between brain-derived neurotrophic factor and the TRKB receptor. Identification of two ligand binding domains in soluble TRKB by affinity separation and chemical cross-linking.

The extracellular domain of the human neurotrophin TRKB receptor expressed in Chinese hamster ovary cells is a highly glycosylated protein, possessing binding ability for brain-derived neurotrophic factor (BDNF). Two distinct ligand binding domains of TRKB were isolated from proteolytic digests of the receptor by affinity separation on immobilized BDNF. One of these domains consists of amino acid residues 103-181 and contains both the third leucine-rich motif and the second cysteine cluster domain. The second domain is close to the second immunoglobulin-like domain (amino acid residues 342-394). Each of these two domains can bind BDNF independently. Disulfide linkages present in the first domain are necessary for BDNF binding, probably because of preservation of the native conformation. To study the second domain in greater detail, a truncated form of TRKB containing the second immunoglobulin-like domain (residues 248-398) was expressed in Escherichia coli. This domain was cross-linked to BDNF through a 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide coupling reaction. Several synthetic peptides corresponding to amino acid residues 343-379 were able to bind immobilized BDNF. Amino acid substitution and cross-linking analysis indicated that amino acids Phe347, Asp354, and Tyr361 are intimately involved in BDNF binding. These results, obtained from a variety of experimental techniques, highlight the importance of two distinct regions of the extracellular domain of the TRKB receptor in binding BDNF.

Glial cell line-derived neurotrophic factor: selective reduction of the intermolecular disulfide linkage and characterization of its disulfide structure.

Glial cell line-derived neurotrophic factor is a protein known to enhance the survival of dopaminergic neurons against several neurotoxins. It has been shown to have therapeutic potential in the treatment of Parkinson's disease and other neurodegenerative diseases. We have determined the inter- and intramolecular disulfide linkages of the dimeric molecule by a combination of direct peptide analysis and peptide analysis after either partial reduction or partial oxidation of the protein. Under an acidic condition, the interchain disulfide bond was selectively cleaved with tris(2-carboxyethyl)phosphine, revealing that Cys101 was involved in the intermolecular disulfide linkage. Three other disulfides, Cys68-Cys131, Cys72-Cys133, and Cys41-Cys102, were identified as intramolecular linkages. The determined disulfide structure is highly homologous to that of transforming growth factor beta 2. Since one intramolecular disulfide points through a ring consisting of eight amino acid residues based on the similarity with transforming growth factor beta 2, the disulfide-linked peptides were not purified by conventional methods. Only the peptides from an N-terminal region (residues -1 to 37) were liberated by proteolytic treatment with trypsin or endoproteinase Lys-C, resulting in a stable cystine-knot protein.

Disulfide structure and N-glycosylation sites of an extracellular domain of granulocyte-colony stimulating factor receptor.

An extracellular domain containing 603 amino acid residues of human granulocyte-colony stimulating factor receptor was expressed in Chinese hamster ovary cells. The affinity-purified material has previously been shown to dimerize when combined with the ligand. In this paper we have characterized the primary structure of this active receptor. Laser desorption mass spectrometry of the purified receptor showed a broad peak at a molecular weight of 84,000, ranging from 77,000 to 91,000. The molecular weight heterogeneity is due to glycosylation. Since the molecular weight based on the amino acid sequence is 67,322, by subtraction the carbohydrate content is approximately 17,000. Disulfide structure of the receptor was determined by peptide mapping in the absence and presence of reducing agent. Sequence and mass spectral analyses of these peptides showed the receptor to contain eight disulfide bonds and three free cysteines. These disulfide bonds are consistent with the known domain motifs of the receptor in that no interdomain disulfides were present. One of the three free cysteines is reactive with alkylating agents, while the others are less reactive, probably being buried in the interior of the molecule. Blocking the free cysteines did not affect the ligand binding. Carbohydrate moieties are somewhat evenly spaced throughout the molecule, at eight different N-glycosylation sites, some of which show heterogeneity in their compositions. Glycosylation seems necessary for stabilizing the molecule against disulfide-linked oligomerization of the receptor, indicating that the free cysteine residues become reactive for oxidation and disulfide exchange upon deglycosylation.

Human neurotrophin-3: a one-step peptide mapping method and complete disulfide characterization of the recombinant protein.

Human neurotrophin-3 (NT-3) is a member of the nerve growth factor (NGF) family of neurotrophic factors, and the recombinant protein is being developed as a therapeutic for neurodegenerative diseases. The final product purity and lot-to-lot variation are monitored routinely by peptide mapping. However, only the N-terminal region of NT-3 was susceptible to proteolysis under native conditions. Complete digestion required that the protein be chemically modified by reduction and S-alkylation prior to proteolysis. Complete proteolytic degradation of the protein was achieved simply by an initial denaturation of NT-3 in 6 M guanidinium chloride (pH6) for 2 hr at 37 degrees C, followed by a tenfold dilution with the digestion buffer (0.1 M Tris-HCl, 1 mM CaCl2 at pH 7.0) and immediate addition of chymotrypsin at 1% by weight. Direct comparison of the peptide map with an identical aliquot that had been reduced and alkylated also allowed the establishment of the cystine linkages present in NT-3: Cys14 to Cys79, Cys57 to Cys108, and Cys67 to Cys110. This disulfide structure is homologous to the NGF family of neurotrophic factors.

Extracellular domain of granulocyte-colony stimulating factor receptor. Interaction with its ligand and identification of a domain in close proximity of ligand-binding region.

An extracellular domain of human granulocyte-colony stimulating factor (G-CSF) receptor was expressed in and purified from Chinese hamster ovary cells. Complex formation between G-CSF and the receptor was studied by size exclusion chromatography, followed by chemical cross-linking. The receptor-ligand complex contained an equimolar ratio of each protein. Crosslinking experiments using disucciniimide suberate revealed that the native complex contained at least two types of cross-linked complexes; one form contained one or two G-CSF molecules per receptor molecule, whereas another form contained one or two G-CSF per two receptor molecules. The tryptic peptide map of the cross-linked complex provided a unique peptide peak which was not found in a peptide map of the original protein. Sequence analysis and mass spectrometry of the peptide indicated that two peptides were covalently linked by cross-linker, one peptide from G-CSF and the other from the receptor. In the cross-linked peptide, Lys-242 of the receptor cross-linked the amino terminal Met of G-CSF through the cross-linker. It was also shown that the N-terminal Met of G-CSF was readily acetylated in the receptor-ligand complex, indicating that it was not directly involved in receptor binding. The results show that the N-terminal Met of G-CSF is located at a distance of approximately 11 A from a reactive Lys-242 of the receptor in the ligand-receptor complex.

Extracellular domain of neurotrophin receptor trkB: disulfide structure, N-glycosylation sites, and ligand binding.

An extracellular domain of a human neurotrophin receptor trkB was expressed in Chinese hamster ovary cells and isolated as a glycoprotein possessing binding activity for brain-derived neurotrophic factor. The extracellular domain contains 398 amino acids and has a molecular weight of 60.6 kDa according to laser desorption mass spectrometry, indicating that the extracellular domain of trkB contains 33.3% carbohydrate moieties. Six disulfide linkages were determined to be Cys1-Cys7, Cys5-Cys14, Cys121-Cys145, Cys123-Cys163, Cys187-Cys235, and Cys271-Cys314, respectively. Cys300 was detected as a free sulfhydryl residue. Cysteine clusters 1 and 2 located in the N-terminal domain possess a similar type of disulfide structure and two other disulfide bonds in the C-terminal region are homologous to that of the Ig-like C2 domain. Among 12 potential N-linked glycosylation sites proposed in the soluble domain of trkB, 10 sites are actually glycosylated.

Purification and identification of brain-derived neurotrophic factor from human serum.

Brain-derived neurotrophic factor (BDNF), a 27-kDa noncovalently linked homodimer with subunits of approximately 13.5 kDa as viewed by SDS-PAGE, is thought to be primarily produced in the central nervous system. We report here the isolation of BDNF from pooled normal human sera, using a two-step purification process followed by SDS-PAGE, transfer to a polyvinylidene difluoride membrane, and subsequent identification of the protein by sequence analysis of the appropriate band(s) from the membrane. The level of BDNF in pooled human sera was estimated to be approximately 15 ng/ml as determined by an enzyme-linked immunosorbant assay. The average for six individuals was 18.9 +/- 5.7 ng/ml. There is an approximately 200-fold increase in the levels of BDNF in serum relative to plasma. Results from experiments using differential centrifugation suggest that the source of this increase is due to release from platelets. The presence of high levels of BDNF in serum suggests a role for this neurotrophin either in nerve repair at sites of injured tissue or in nonneuronal functions.

Analysis of translational termination of recombinant human methionyl-neurotrophin 3 in Escherichia coli.

A highly efficient UGA stop codon readthrough event during the synthesis of human neurotrophin 3 in E. coli is described. The incorporation of a Trp residue at the UGA stop codon is confirmed combining both the chemical analyses and the molecular and genetic data in this report. The 3' adjacent nuleotide to the UGA stop codon plays a crucial role in determining the readthrough efficiency in the order of A > G > C > U. The replacement of UGA with UAA or UAG totally abolished this readthrough phenomenon and the use of StpR host cells also prevented the occurrence of UGA readthrough. Gene dosage (or plasmid copy number) effect was not indicated in this event; however, the titration of RF-2 by mRNA transcripts under over-expression conditions might explain why tRNAtrp competes so well with RF-2 for UGA. Another apparently less produced readthrough product resulting from a transcript with no stop codon is also recorded, and the addition of a second in-frame stop codon increased the amount of the observed readthrough product.

Formation of heterodimers from three neurotrophins, nerve growth factor, neurotrophin-3, and brain-derived neurotrophic factor.

Three neurotrophic factors, brain-derived neurotrophic factor (BDNF), neurotrophin-3 (NT-3), and nerve growth factor (NGF) form noncovalent homodimers in solution. Since they are highly homologous proteins, it seemed probable that two monomers of these proteins might associate together to form a heterodimer. This was tested by denaturing the two different proteins together in 6 M guanidine HCl and refolding them in phosphate-buffered saline. When the refolded mixture of BDNF and NT-3 was subjected to Mono S cation exchange chromatography, a new peak was observed eluting between NT-3 and BDNF, which accounted for about 30% of the protein used. This new protein species migrated as a single band upon native gel electrophoresis with mobility between that of the NT-3 homodimer and the BDNF homodimer, indicating that a complex had been formed. Sedimentation equilibrium data show that the dissociation constant of this heterodimer is < 3 x 10(-10) M. The heterodimer was stable upon incubation at 37 degrees C in phosphate-buffered saline over 11 days. Having determined that the heterodimer is highly stable, it was subjected to various biological assays. Autophosphorylation assay using TrkB receptor showed that the heterodimer is indistinguishable from the BDNF or NT-3 homodimer in the ability to induce phosphorylation of the receptor. It was also indistinguishable from the homodimers in the neurotrophic activity using chick dorsal root ganglion explant. In the sympathetic neuron survival assay, the heterodimer behaved more similarly to NT-3, whereas in the dopamine uptake assay, it was intermediate between the two homodimers. In addition, the heterodimer was shown to be retrogradely transported in the dorsal root ganglion neurons. A heterodimer between NGF and BDNF is formed but much less effectively than the NT-3.BDNF heterodimer, and it is not stable even at 4 degrees C. These results indicate that BDNF and NT-3 have an intersubunit contact surface for dimerization resembling each other's but different from the contact surface of NGF.

Characterization of disulfide linkages in platelet-derived growth factor AA.

Intermolecular and intramolecular disulfide linkages of recombinant human platelet-derived growth factor A chain dimer were determined by chemical methods including selective reduction-alkylation, peptide isolation, or detection of diphenylthiohydantoin derivative of cystine from Edman reactions. Cys-37 and Cys-46 were selectively reduced with reducing agents under native conditions and revealed to be involved in intermolecular bridges. Other disulfide linkages including Cys-10-Cys-54, Cys-43-Cys-91, and Cys-47-Cys-93 form intramolecular bridges. The disulfide structure is homologous to that of platelet-derived growth factor B chain dimer.

Formation of mitogenically active PDGF-B dimer does not require interchain disulfide bonds.

Platelet-derived growth factor (PDGF), a major mitogen for mesenchymal cells, is a disulfide-bonded dimer of two subunit polypeptides named A and B. All of the three possible dimeric forms, i.e. AA, BB, and AB, exist in nature. The dimeric structure has been presumed to be necessary for biological activity, since reduction of the dimer results in loss of activity and simultaneous conversion to monomeric form as determined by SDS-gel electrophoresis. However, reduction of the native molecule destroys intrachain, as well as interchain, disulfide bonds, and it is possible that the former rather than the latter are critical for proper conformation of the active protein. We show here that PDGF-B polypeptides in which all 8 cysteines or the 2nd, 4th, 5th, and 8th cysteines have been mutated to serines fail to form covalent dimers and possess dramatically less mitogenic activity than native PDGF-BB. Another mutant, PDGF-B(C2,4S), in which just the 2 cysteines involved in interchain disulfides were converted to serine, ran as a monomer on SDS-polyacrylamide gels as expected. Somewhat unexpectedly, however, the mitogenic activity of the PDGF-B(C2,4S) analog was similar to the activity of wild-type PDGF-BB disulfide-bonded dimer under physiological conditions. The activity of the analog was more sensitive to the effect of low pH than was the activity of wild-type PDGF-BB. Molecular weight analysis utilizing light scattering and sedimentation equilibrium demonstrated that the PDGF-B(C2,4S) analog exists as a noncovalent dimer at pH 4-7 but dissociates to a monomer at pH 2.5. Disulfide analysis of the mutant protein demonstrated that the intrachain disulfide bonds are the same as those formed in wild-type PDGF-BB homodimers. We conclude that proper formation of intrachain disulfide bonds is critical to maintaining the correct conformation of PDGF monomers, but that appropriately folded monomers can associate into active noncovalent dimers in the absence of interchain disulfide bonds. Interchain disulfide bonds thus appear to increase the stability of the PDGF dimer rather than being crucial to its existence.