Conformational and Molecular Weight Studies of Tetanus Toxin and Its Major Peptides*

T w o forms of tetanus toxin have been purified from Clostridium tetani cultures, These forms, obtained from filtrate and cellular extracts, were characterized by analytical ultracentrifugation using both conventional and meniscus-depletion sedimentation equilibrium. The molecular weight of filtrate toxin was found to be 128,000 2 3,000, while the extract toxin, which tended to self-associate, appeared somewhat larger, 140,000 -t 5,000. The heavy and light chains were prepared from filtrate toxin, and their molecular weights were estimated to be 87,000 and 48,000, respectively, using polyacrylamide gel electrophoresis in the presence of sodium dodecyl sulfate. The circular dichroic spectra of the extract and filtrate toxins are quite similar between 200-300 nm in-dicating that no major conformational difference exists between the two. The toxins contain both a-helicity and /?-structure. Interestingly, the isolated chains contain appreciable helicity (e.g., the sum of the chain helicities is over 80% of that found in filtrate toxin), but they appear to have relatively low contents of /?-structure. The sum of the spectra of the chains in both the near-and far-ultraviolet does not yield that found for filtrate toxin, although the similarity is far more striking than the difference. The prominent 293.5 nm negative cir- cular dichroic band of tetanus toxin can be assigned to tryptophanyl residues almost exclusively in the heavy chain. The similarity in the magnitude of this band in the separated chain and toxin suggests that the mi-croenvironments and was in the discarded portion of the trailing edge of the toxin after gel filtration on Sephadex G-150. The purified toxin showed no protease activity either by the gelatinase or Azocol assay. Circular Dichroism-CD spectra were obtained at 26 "C using a Cary 60 spectropolarimeter equipped with a CD attachment. Above 250 nm, a path length of 1 cm was used, and protein concentrations varied from 0.6 to 2 mg/ml. Below 250 nm, cells with path lengths of 0.5 and 2 mm were used, and protein concentrations were between 60 p g / d and 0.25 mg/ml. The following molecular weights were used for the toxins and fragments: extract toxin (140,OOO), filtrate toxin (128,000), heavy chain (87,000). and light chain (48,000). The average residue molecular weights were calculated from the amino acid com-positions, and all were found to be 113 except the heavy chain, which was 112. Scans were always taken in duplicate and often in replicate (up to 8 scans). The reported spectra refer to mean values and standard errors were generally within *IO% of the mean at each wavelength. The spectra above 250 nm are reported as molar elliptic- ity and those below 250 nm are given as mean residue ellipticity.

Instead, we will only refer to those areas which directly apply to our results.
The toxin is prepared either by extraction from log phase cells (extract or intracellular toxin) or by purification from culture filtrate of older cultures (filtrate or extracellular toxin). It is generally agreed that the former is a single polypeptide chain while the latter has been cleaved by endogenous proteases and consists of two major peptides connected by disulfide linkage (1)(2)(3)(4)(5)(6)(7)(8)(9). One peptide in the cleaved form is much larger than the other and the two are often referred to as heavy and light chains based on the size difference. We have recently shown that cleavage of tetanus toxin by endogenous proteases can occur near to and on either side of the disulfide bond linking the heavy and light chains (15). Filtrate toxin preparations therefore can contain more than one form of toxin. In one form the light and heavy chains are covdently connected by disulfide linkage, but in the other they are noncovalently held.
We have prepared highly purified extract and filtrate toxins and examined the ultraviolet circular dichroic spectra of each. The purified filtrate toxin, which is the form consisting of covalently connected heavy and light chains, was separated into its heavy and light chain components and these too were examined.
There has been considerable disagreement about the molecular weight of tetanus toxin. The older literature reveals values that are at wide variance with the more acceptable current structural models. Most recent reports using gel filtration, electrophoresis, or the analytical ultracentrifuge give values which vary between 140,000 and 160,000 (1,3,5,10). It is generally known that there is considerable aggregation and fragmentation in those samples studied. Consequently, in addition to the CD spectral analyses, we also report here a reexamination of the molecular weight of both filtrate and extract tetanus toxin.

MATERIALS AND METHODS
Chemicals-All chemicals used in this work were reagent grade unless otherwise stated. Acrylamide and N,N"methylenebisacrylamide were purchased from Eastman. The acrylamide was recrystallized from chloroform. Guanidine hydrochloride was Heico's extreme purity product. Glutathione was purchased from Sigma and thioglycolic acid from Pierce. The apparatus used for preparative gel electrophoresis was a Buchler Poly Prep 200 preparative gel electrophoresis unit. Phenylmethanesulfonyl fluoride was purchased from Sigma. Azocol was obtained from Calbiochem-Behring.
Toxin Production and Purification-C. tetani, Massachusetts strain, was grown in 20-to 40-liter quantities on the medium described by Latham et al. (16). Cultures to produce fdtrate toxin and extract toxin were prepared as reported by others (5, 7). Both forms of the toxin were initially purified by a combination of gel filtration and ion exchange chromatography as described earlier (IO).
Further purification was carried out by preparative gel electrophoresis. Electrophoresis was conducted at 45 mA using an elution buffer flow rate of 12 to 18 ml/h. The buffers consisted of the following: upper electrode buffer (6.32 g of Tris and 3.94 g of glycine/1000 ml), lower electrode buffer (48.4 g of Tris and 200 ml of 1 N HC1/1000 ml), and elution buffer (12.1 g of Tris and 50 ml of 1 N HC1/1000 ml). The resolving gel consisted of 4% polyacrylamide in 60-to 90-ml volume. The stacking gel contained 2.5% polyacrylamide and 4 M urea in a 32ml volume. When reducing conditions were desired 2 mM glutathione and 1 m~ thioglycolic acid were added to the upper electrode and elution buffers, respectively. Elution profiles and further details of this procedure will be published elsewhere. ' The toxin and light chain preparations were promptly concentrated using an Amicon apparatus. Heavy chain fractions were pooled and concentrated by placing them in dialysis tubing which was then packed in solid sucrose.
All samples were dialyzed into 0.2 M sodium phosphate buffers at neutral pH before determining the circular dichroic spectra or subjecting to ultracentrifugation.
Assay for Protease-Proteolytic enzymes were recently described in the growth medium of C. tetani cultures (17). The one most active in cleaving the toxin was reported to be inhibited by phenylmethanesulfonyl fluoride. All toxin preparations were treated with phenylmethanesulfonyl fluoride at both early crude and later purified stages of preparation. Protease activity was monitored in all fractions at follows: a 0.05-ml aliquot of each fraction was added to 1 ml of 5% each stage of purification for gelatinase activity. The assay was as gelatin in 0.15 M NaCl and 5 m~ sodium phosphate buffer, pH 7.0.
After a 4-h incubation at 37 "C the mixture was cooled to 4 "C and examined for liquefaction. Although the protease activity closely follows toxin during ion exchange chromatography, it appeared to be retarded and was in the discarded portion of the trailing edge of the toxin after gel filtration on Sephadex G-150. The purified toxin showed no protease activity either by the gelatinase or Azocol assay. Circular Dichroism-CD spectra were obtained at 26 "C using a Cary 60 spectropolarimeter equipped with a CD attachment. Above 250 nm, a path length of 1 cm was used, and protein concentrations varied from 0.6 to 2 mg/ml. Below 250 nm, cells with path lengths of 0.5 and 2 mm were used, and protein concentrations were between 60 p g / d and 0.25 mg/ml. The following molecular weights were used for the toxins and fragments: extract toxin (140,OOO), filtrate toxin (128,000), heavy chain (87,000). and light chain (48,000). The average residue molecular weights were calculated from the amino acid compositions, and all were found to be 113 except the heavy chain, which was 112. Scans were always taken in duplicate and often in replicate (up to 8 scans). The reported spectra refer to mean values and standard errors were generally within *IO% of the mean at each wavelength. The spectra above 250 nm are reported as molar ellipticity and those below 250 nm are given as mean residue ellipticity. Protein concentrations were based on multiple Lowry analyses (18) using bovine serum albumin as standard (IO, 11). The two forms of toxin were studied in 0.2 M sodium phosphate buffer, pH 6.9, and the constituent chains were in the same buffer containing 2 mM glutathione.
Molecular Weight Determination of Tetanus Toxin-Molecular weights were determined in the analytical ultracentrifuge by sedimentation equilibrium by both conventional (19) and meniscus-depletion (20) techniques. In order to test for homogeneity in these preparations, the Lamm plot 2RT. All ultracentrifuge runs were made at 20 "C with rotor speeds between 7,000 and 9, OOO rpm for conventional runs and near 14,000 rpm for meniscus-depletion runs. In a l l cases the buffer was 0.2 M sodium phosphate, pH 6.9, and solutions of extract toxin also contained 1 mM thioglycolic acid. The values for partial specific volumes were derived from amino acid composition of the toxins using standard values (22). Buffer densities were computed from available data (23). Solution densities were calculated using partial specific volume of the protein, the macromolecular concentration, and buffer densities.
Molecular Weight of Heavy and Light Chains by Sodium Dodecyl Sulfate-Polyacrylamide Gel Electrophoresis-The molecular weight of these two peptides were estimated from their electrophoretic mobilities relative to those of standard proteins according to Weber and Osborn (24).  These data give an M, value of 131,000.

Molecular Weight
Determinations-Filtrate toxin, with Hz and L chains being held covalently by disulfide linkage, was prepared by preparative gel electrophoresis, without reduction, as described under "Materials and Methods." The purified toxin was immediately dialyzed into 0.2 M sodium phosphate buffer and examined in the analytical ultracentrifuge. Meniscus depletion gave a molecular weight of 128,000, and a plot of h e,. versus r2 was linear over a 60-fold concentration range. When the same material was examined by low speed equilibrium, data analysis gave Mu cell as 127,000 and M, cell as 130,000. An M , plot of In [(dc/dr)/r] Versus rz according to Lamm (21) is shown in Fig. 1 and exemplifies a monodisperse preparation. The value of M, from this plot is 131,000 and agrees well with M, obtained from the meniscus-depletion experiment. We therefore conclude that fresh preparations of the species of filtrate toxin, in which the H and L chains are covalently held gave M, = 128,000 f 3,000.
If these preparations are allowed to stand for 1 week at 4 "C they no longer show such homogeneity but show considerable fragmentation and aggregation. The reason for this change is at present uncertain but may involve residual activity of an endogenous proteaseb). Extract toxin was prepared by preparative gel electrophoresis in the presence of reducing agents as described under "Materials and Methods." Following dialysis into buffer containing 1 m~ thioglycolic acid, the sample was immediately subjected to analytical ultracentrifugation. From two separate meniscus-depletion runs the molecular weight was determined as 138,000 and 143,000 and both gave linear plots of In C, versus r2 over a 60-fold concentration range. From a low speed sedimentation equilibrium run M, was calculated using the initial portion of a plot of In [(dc/dr)/r] versus r2 (Fig. 2) to be 140,000, although clearly the solution is not monodisperse.  Fig. 3 shows a plot of molecular weight versus concentration and demonstrates that within the extract toxin preparation there is considerable aggregation. However, the data shown in Fig. 3 a t concentrations between 0.4 and 1.1 g/100 ml agree with the meniscus-depletion results.

considerable uncertainty, and association is evident at concentrations
The H and L chains were also examined by analytical ultracentrifugation, but the results were unsatisfactory due to aggregation in the absence of denaturing agents. Thus molecular weights were estimated using sodium dodecyl sulfatepolyacrylamide gel electrophoresis. The standard proteins A summary of the molecular weight estimates of the two forms of toxin and the two chains, along with the standard errors, is given in Table I. It is noteworthy that the combined value of 135,000 for the expected molecular weight of the H and L chains of tetanus toxin is in reasonable agreement with the experimental value of 128,000 determined by analytical ultracentrifugation of filtrate toxin.
CD Spectroscopy-The near-UV CD spectra of the two forms of toxin and of the H and L chains are shown in Fig. 4 similar to each other; the approximate 10% difference between the two is within experimental error. Since this spectral region is particularly sensitive to a-helicity and p-structure, it is safe to conclude that the two forms of toxin differ little if any in secondary structure. The CD spectra of the H and L chains (Fig. 5A) are quite different from each other. The H chain is characterized by a spectrum which is much more similar to toxin than that of the L chain. The negative bands and shoulders at 210-211 nm and 219-221 nm in the toxins and H chain are assigned, respectively, to the peptide 7-r* and q-r* transitions associated with the a-helix (25). The L chain bands

TABLE I Summary of molecular weights of the two forms of tetanus toxin and the H and L chains
The data for the two forms of toxin are from analytical ultracentrifugation (low speed sedimentation equilibrium and meniscus depletion), and the results for the constituent chains of fdtrate toxin are estimates from sodium dodecyl sulfate-polyacrylamide gel electrophoresis.

Tetanus Toxin and
Derived Peptides are indicative of both a-helicity and p-structure. The sum of the H and L chain spectra (with corrections made for the molecular weight differences) is shown in Fig. 5C, and the inset shows the CD difference spectrum of filtrate toxin minus the sum of the consistent chain spectra. The major band in the difference spectrum occurs at 217.5 nm, and this suggests that p-structure is preferentially lost when the H and L chains are formed from filtrate toxin. The smaller band at about 225 nm may reflect a small loss in a-helicity when the chains are formed from toxin, and there may also be a contribution from aromatics and &turns in this region (26-28).
Using highly accurate CD data for proteins of known crystallographic structure (28), we attempted to estimate the amounts of a-helicity, p-structure, and p-turns in the toxins and chains. With no constraints imposed on the fraction of residues in either helical, p-structure, p-turns, or aperiodic forms, the helix fraction was estimated to be between 0.23 and 0.24 in the two forms of toxin. This is in agreement with our earlier studies (10-12) using a somewhat different reference set of spectra, although in the present case a poor sums test was obtained. By constraining the sum of helical fraction, pstructure fraction, and remainder (i.e., aperiodic plus p-turns) to unity, the toxins and the H and L chains had respective apparent helical contents of 35 * 1% (i.e., the helicity of both forms of toxin were within 1% of each other), 30%, and 24%. The respective p-contents were 30 It 1%, t l % , and 2%. It is of interest that the sum of a-helicity in the H and L chains, corrected for their respective molecular weights, yields a combined helicity which is over 80% of that found in filtrate toxin.

DISCUSSION
These experiments reveal that the molecular weights of both nicked filtrate and unnicked extract toxins are somewhat less than the 150,000 which have been reported (2, 10). The higher values in earlier reports are most likely the result of aggregation in the samples which was occasionally acknowledged (10,14). Also, the use of mixtures of filtrate toxins containing both covalently and noncovalently joined heavy and light chains could lead to overestimates of the molecular weight. Moreover, it is apparent that such studies must be completed soon after purification since there is sigdicant fragmentation and aggregation within 1 week of storage at 4 "C. Efforts to examine the two major peptides in the analytical ultracentrifuge have been unsuccessful due to fragmentation and aggregation, and the fragmentation was accelerated when the determinations were attempted in 6 M guanidine hydrochloride. The reason for this sample deterioration is unknown but one possible explanation may be residual protease activity which is accelerated when the peptides are unfolded in the presence of the denaturant. We emphasize, however, that using sodium dodecyl sulfate-polyacrylamide gel electrophoresis the estimates for the molecular weight of the two chains agree well with that expected for intact toxin.
The CD results are particularly illuminating and indicate that any conformational differences between the two forms of toxin must be subtle. The near-UV CD spectra of toxin and chains strongly suggest that the major tryptophanyl band in toxin can be attributed almost exclusively to the H chain. This is in excellent agreement with amino acid composition studies.' Also, our finding that the sum of the spectra of the chains differs from that of toxin indicates that the microenvironments of at least some of the aromatic groups are not the same. This could arise from conformational alternations accompanying chain separation.
The far-UV CD data suggest that much of the a-helicity in toxin can be ascribed to the H chain. Earlier work with another toxin fragment, Fragment I1 which seems equivalent to Fragment C of Helting and Zwisler (7), suggested a predominance of p-structure (ll), although the origin of the structure ( i e . , intramolecular or intermolecular) was not defined. It was concluded earlier that Fragment C was located in the NH2-terminal portion of the heavy chain (7). However, subsequent immunological (29) and NH2-terminal analyses (30) of these fragments now indicate that Fragment C, and consequently our Fragment 11, are located in the COOH-terminal portion of the H chain. Therefore, a-helical structure in the H chain may be confined mostly to the NHn-terminal portion. Lastly, we wish to emphasize that the separated H and L chains appear to retain much of the helical structure that exists in toxin suggesting the possible occurrence of quasi-independent structural domains that may interact via p-structure since this seems to be preferentially lost in the isolated chains.