Discrete Soluble Forms of Middle and High Molecular Weight

Vol. 262, No. 35, Issue of December 15, pp. 17009-17015,1987 Printed in U.S.A.

THEJOURNALOF BIOLOGICAL CHEMISTRY 0 1987 by The American Society for Biochemistry and Molecular Biology, Inc.

Discrete SolubleForms of Middle and High Molecular Weight Neurofilament Proteins in DiluteAqueous Buffers* (Received for publication, June 2, 1987)

Jeffrey A. CohlbergS, HamidHajarian, and Suzanne Sainte-Marie From the Chemistry Department, California State University at Long Beach, Long Beach, California90840

1982; Jones andWilliams, 1982).These C-terminal extensions have been suggested to extend outward as projections from the filament,where they may be involved in interactionswith other cellular components (Metzuals et al., 1981; Geisler et al., 1983, 1984; Hirokawa et al., 1984). Both purified NF-L (Geisler and Weber, 1981; Liem and Hutchison, 1982; Zackroff et al., 1982) and purified NF-M (Gardneret al., 1984) can be induced to form 10-nm filamentsby dialysis against buffers of physiological pH and ionic strength, but only short structures have been seen with NF-H (Gardneret al., 1984). Although until a few years ago a model for intermediate filament structure based on three-stranded coiled coils was favored (Steinert et al., 1982), it is now generally accepted that IF consist of two-stranded coiled coils organized into four-chain units (Woods and Gruen, 1981; Geisler and Weber, 1982; Steinert and Parry, 1985).A detailed model for the organization of polypeptide chains within intact filaments has been proposed (Fraser et al., 1985). Our understanding of the arrangement of the polypeptide chains in IF has been aided by studies on the state of purified IF proteins in aqueous solution. Both desmin (Huiatt et al., 1980; Geisler and Weber, 1982) and vimentin (Ipet al., 1985) havebeenshown t o existastetramersindilute aqueous buffers at slightly alkaline pH, as have complexes containing Neurofilaments, cytoskeletal elements found in neurons, equimolar amounts of Type I and Type I1 keratins (Quinlan comprise one of the five tissue-specific classes of intermediate et al., 1984). A solution study of NF-L showed that it, too, filaments (IF)’(Dah1 and Bignami, 1985).They have a unique existed in solution as a species sedimenting with an sz0,,, of polypeptide composition in that they contain three types of 4.8 S (Lifsics and Williams,1984).Althoughamolecular polypeptidechain(HoffmanandLasek, 1975; Liem et al., weight determination indicated that a trimer was the predom1978), designated NF-L (low), NF-M (middle), and NF-H inant species, a re-evaluation of the data witha more accurate (high) according to their relative molecular weights (Kauf- subunit molecularweight indicatesthatNF-L, too, forms typical IF tetramers in aqueous solution (see “Discussion”). Quinlan et mann et al., 1984). All threepolypeptidesare proteins, containing a central helical domain which is pre- al. (1986) showed that, like other IF proteins, NF-L exists as dicted to form a double-stranded coiled coil, flanked by non- a dimer in3 M guanidine hydrochloride; they observed an s25.w helical headandtaildomains at the amino and carboxyl of2.9 S.These are the only published studies of NF-L in termini. Thetail regions of allthreeproteinsare highly solution, and no solution studies of the other two neurofilacharged because of a high content of lysine and glutamicacid ment proteins have appeared. residues, and the variable size of the proteins is due principally We now report studies on the solution stateof each of the t o variation in the length of these tail regions (Geisler et al., two larger neurofilament proteins, with emphasis on NF-M. 1982-1985). The charged natureof the tailregions also results Surprisingly, under conditions which generallypromote tetrafrom a high degree of phosphorylation (Julien and Mushynski,mer formation, bothof the large neurofilament proteins exist predominantly as species smaller than tetramers. This con* This work was suppored by Research Grant DCB-8502594 from the National Science Foundation and by California State University, clusion is supported by sedimentation velocity and equilibLong Beach. The costs of publication of this article were defrayed in rium studies, which suggest that the predominantspecies of part by the payment of page charges. This article must therefore be both proteins are monomers. Studies on thegradual aggregahereby marked “advertisement” in accordance with 18 U.S.C. Section tion of NF-M upon increasing the ionic strength failed to 1734 solelyto indicate this fact. reveal the formation of tetrameric assembly intermediates. $ To whom correspondence should be addressed. We also report circular dichroism spectra of both protein ’The abbreviations used are: IF, intermediate filarnent(s); NF-L, preparations; the spectrumof NF-M indicated a greater conNF-M, and NF-H, low, middle, and high molecular weight neurofilament proteins, respectively; PMSF, phenylmethylsulfonyl fluoride; tent of a-helix thanwas previously reported.

Neurofilament proteins purified from bovine spinal cord were characterized by sedimentation studies in aqueous buffers. In 10 mM Tris, pH 8, the middle molecular weight neurofilament protein (NF-M) has a sedimentation coefficient, s2,,+,, of 2.6 S. Sedimentation equilibrium data shows considerable nonideality; extrapolation to infinite dilution and correction for the primary charge effect yielda molecular weight of 1.09 X lo’, indicative of a monomeric structure. When the ionic strength was increased, the sedimentation coefficient increased slightly, and the protein began to form larger aggregates. Reconstitution of short intermediate filaments was observed upon dialysis of denatured NF-M versus a reconstitution buffer. A circular dichroism spectrum of NF-Min 10 mM Tris was typical of cr+B proteins. High molecular weight neurofilament protein (NF-H) showed a considerable tendency to aggregate in 10 mM Tris, but a principal species with a sedimentation coefficient of 3.2 S was observed, and sedimentation equilibrium data also suggest a monomeric structure.

MES, 4-morpholineethanesulfonicacid; M,, apparent weight-average molecular weight; M,, apparent number average molecular weight; My,, first ideal molecular weight moment (Roark and Yphantis, 1969); EGTA, [ethylenebis(oxyethylenenitrilo)]tetraaceticacid.

EXPERIMENTALPROCEDURES

Preparation of Neurofilament Proteins-Neurofilaments were isolated from bovine spinal cord by the method described for porcine

17009

17010

Neurofilament LargeProteins

spinal cord by Delacourte et al. (1980) as modified by Lifsics and Williams (1984).The crude neurofilaments were dissolved in a buffer containing 6 M urea and subjected to chromatography on Whatman DE32 in 6 M urea at pH 7.5 as described by Lifsics and Williams (1984). Column fractions were analyzed by sodium dodecyl sulfatepolyacrylamide gel electrophoresis by the method of Laemmli (1970). Fractions containing each of the three neurofilament polypeptides were pooled. Final purification was achieved by rechromatography of the partially purified proteins on DE32 in 6 M urea at pH 7.5. For the separation of residual NF-M from partially purified NF-L, cbromatography on hydroxyapatite (Lifsics and Williams, 1984) or singlestranded DNA-cellulose (Nelson et d . , 1982) was also found to be useful. In the latter method both NF-M and degraded NF-L eluted before intact NF-L. Purified proteins dissolved in column buffers containing 6 M urea were frozen rapidly in dry ice-ethanol baths and stored at -80 "C. After thawing, the proteins were incubated with 1%mercaptoethanol prior to overnight dialysis against Buffer A (10 mM Tris-HC1, 6 mM 2-mercaptoethanol, 0.4 mM PMSF, pH 8) or other aqueous buffers. Protein Concentrations-Protein concentrations were determined by measuring the absorbance at 280 nm. An extinction coefficient of 0.374 ml mg" cm" for NF-L was reported by Lifsics and Williams (1984). To determine the extinction coefficients of NF-M and NF-H, solutions were dialyzed against 10 mM Tris, pH 8, and theabsorbances at 280 nm were measured. An aliquot of each solution was subjected to amino acid analysis by hydrolysis in 6 N HCI, derivatization with phenyl isothiocyanate, and separation on an Altex Ultrasphere column in 0.15 M sodium acetate, pH 5.5, a t 50 "C. The yields were corrected for hydrolysis and handling losses by comparison with a standard amino acid mixture which had undergone a parallel bydrolysis and handling protocol. Protein concentrations of the original solutions were calculated from these corrected yields, and values for the extinction coefficients of 0.45 ml mg" cm" for NF-M and 0.28 mlmg"cm" for NF-H were determined. The same solutions were analyzed by measurements of the refractive indices in the analytical ultracentrifuge by the method of Babul and Stellwagen (1969) and by the Coomassie Blue binding method of Bradford (1976), using bovine serum albumin as a standard. The results from the three methods agreed to within 10%of each other. Sedimentation-sedimentation velocity experiments were performed in a Beckman model E analytical ultracentrifuge equipped with a photoelectric scanner, using light with a wavelength of280 nm. Sedimentation equilibrium experiments were performed by the meniscus depletion method of Yphantis (1964), using the Rayleigh interference optical system. The fringe displacements were measured on a Nikon model 6C profile projector. The data were analyzed by a computer program which plotted In c uersus 4, determined the slope at each point from a sliding least squares quadratic fit,and used this slope in the calculation of the apparent weight-average molecular weight. Apparent number-average molecular weights were calculated from Equation 5 of Yphantis (1964). Partial specific volumes were calculated from the amino acid compositions reported by Hogue-Angeletti et al. (1982), with the "consensus" residue volumes tabulated by Perkins (1986). Values of 0.735, 0.729, and 0.729 cm3/g were determined for NF-L, NF-M, and NFH, respectively. These values were then corrected for the occurrence of phosphoserine residues. Various values of the phosphate content of neurofilaments have been reported (Jones and Williams, 1982; Julien and Mushynski, 1982; Wong et QL, 1984; Georges et al., 1986). The last three sets of values agree fairly closely, with average values of 3 phosphates/70 kDa of NF-L, 11 phosphates/l60 kDa of NF-M, and 23 phosphates/210 kDa of NF-H. From the 8 values of alanine, serine, and HP02-4 we estimate a residue P of 0.51 cm3/g for phospboserine. The average phosphate contents listed above then yield corrected 8 values of 0.732, 0.725, and 0.723 cm3/g, respectively for NF-L, NF-M, and NF-H. Reconstitution of Filaments-Neurofilaments or purified proteins which had been stored frozen in buffers containing 6 M urea were dialyzed overnight uersus areconstitution buffer similar to that described by Gardner et al. (1984), which contained 0.01 M MES, 0.2 M NaCI, 6 mM 2-mercaptoethanol, 0.5 mM EGTA, 0.4 mM PMSF, pH 6.5. For some experiments an initial dialysis was performed uersus 6 M guanidine hydrochloride, 10 mM Tris, 6 mM 2-mercaptoethanol, 0.4 mM PMSF, pH 8, followedby dialysis against reconstitution buffer. Ebctron Microscopy-A drop of suspension containing native or reconstituted neurofilaments was applied to a carbon-coated Formvar

in Aqueous Solution grid which had been subjected to glow discharge. After 1 min the liquid was removed by blotting on filter paper, and a drop of 2% uranyl acetate was applied to the grid and left for 10 s. The uranyl acetate treatment was repeated twice. Alternatively, some samples were absorbed onto carbon-coated mica, and the carbon films were floated off the mica, stained with 2% uranyl acetate, andthen transferred to uncoated grids, as described by Lake (1979). The grids were examined in a Siemens model 1A electron microscope operated at 60 kV. Circular Dichroism-Circular dichroism spectra were recorded on a modified Beckman circular dichroism spectrophotometer (Honvitz et al., 1979).Camphorsulfonic acid was used as a standardto calibrate the instrument. A cuvette with an 0.2-mm pathlength was used. For protein concentration determination, an aliquot of the solution withdrawn from the cuvette was subjected to amino acid analysis as described above. Mean residue weights of 122 for both proteins were calculated from the amino acid compositions reported by HogueAngeletti et al. (1982). RESULTS

Sedimentation Studieson NF-M in 10 mM Trk-A solution of NF-M in a 6 M urea buffer was dialyzed against Buffer A (10 mM Tris-HC1, 6 mM 2-mercaptoethanol, 0.4 mM PMSF, pH 8) and examined ina sedimentation velocity experiment. A sharp boundary was observed (Fig. 1, A and B) with a sedimentation coefficient of 2.47 S. A small amountof heavier aggregate was also present (Fig. lA),but the 2.5 S species comprised more than 90% of the absorbingmaterial. In other experimentsundersimilarconditions, rangedfrom 2.34 to 2.78 S , with an average value of 2.58 S. A sedimentation equilibrium experiment on the preparation of Fig. 1 gave In c uersus ? plots with distinct downward curvature (Fig. U ) ,which indicated nonideal behavior. Consequently,the reciprocals of theapparentnumber-and weight-average molecular weights, l/Mn and l/M,, as determined by the computer program, were plotted uersus protein concentration (Fig. 2B). Also shown is the reciprocal of the ideal molecular weight moment, l/Myl (Roark and Yphantis, 1969),which is equal to 2/Mn - l/M,. This value should not be affected by nonideality.Both 1/M, and l/Mn were approximately linear functions of concentration up to a concentration of about 1.5 mg/ml, and thevalue of l/Myl was constant uptoabout 1.5 mg/ml andthen increasedslightly. This indicates that the solution contained a single component and that therewas no appreciable equilibrium associationto heavier species. Extrapolation of thethreegraphstoinfinite dilution yielded intercepts close to one another on the vertical value axis. Since thevalues of M, and My,are sensitive to the of the meniscus concentration used in the calculations, the extrapolated value of M , was taken as the correct molecular weight value. In the experiment shown this value was 1.185 X lo5. The results from a secondsample runatdifferent protein concentrationwere virtually perfectly superimposable on the data shown in the figure, and the extrapolated value of M, was 1.157 X lo5. The average of the two experiments is 1.171 X

lo5.

The sedimentationequilibrium behavior of highly charged proteins in dilute buffers is affected by counterion binding. The molecular weight of the protein component can be estimated by performing the calculations outlined by Johnson and Yphantis (1978). The first step is the estimation of the "effective charge" of the protein. From theslope of the graph of l/Mwuersus concentration, one cancalculate a value of 3.1 X for the secondviralcoefficient, B. Assuming that the nonideality is due entirely to the Donnan effect (Roark and Yphantis, 1971) leads to an estimateof -38 for the effective charge of the molecule. Taking as a model for the protein a rod of axial ratio21 (see "Discussion") leads to theconclusion that the excluded volume contribution to B (Tanford, 1961)

17011

Large Neurofilament Proteins inAqueous Solution

1

I

J

4

I

I

1

FIG. 1. Sedimentation velocity of neurofilament proteins. A-E, NF-M at 1.3 mg/ml, rotor speed 52,000 rpm: A , Buffer A, 1.3 mg/ml, 20.9 “C, 48 min after reaching speed; B , same as A , 134 min; C, 25 mM Tris, 19.2 “C, 48 min; L), 50 mM Tris. 19.2 “C. 17 min: E , 100 mM Tris, 20.0 ‘C, 19 min. F, NF-H in Buffer A, 1.2 mJml, 60,000 rpm, 19.2 ‘C, 72 min.

sists of the isoionic protein plus 82 molecules of Tris. The Scatchard component is formed from the macromolecular salt by removing 18.5 molecules (one half the effective charge) of Tris hydrochloride. Using values of 92 cm3 and 142 cm3 for the molar volumes of Tris and Tris hydrochloride, respectively, one can calculate that this Scatchard component has a 5 of 0.723 cm3/g and a mass 6.6% greater than that of the isoionic protein. Analysis of the data with this value for d leads to a value of 1.162 X lo5 for the molecular weight of the -3 0.5 Scatchard component and hence 1-09 x lo6 for the molecular 44 45 46 0.0 1.0 2.0 conc (rnghnl) r2 weight of the isoionic protein. (Using the range of-62 to -102 for the titration charge yields a range of 1.07 to 1.11 X FIG. 2. Sedimentation equilibrium of NF-M in Buffer A. lo5 for the isoionic molecular weight.) Thus, because of the Speed, 20,000 rpm. Temperature, 13.5 “c.Concentration, 1.0 mg/ml. A, plot of In c versus 3. B , plot of reciprocals of apparent molecular effects of counterion binding to the charged protein, the weights versw concentration: filled squares, 1/MW;open squares, 1/ apparent molecular weight at infinite dilution is about 7% M.; filled triangles, l/M,.l. higher than the isoionic molecular weight. Since the polypeptide chain molecular weight of NF-M is is about 1.4 X If thisnumber is subtracted from the about 1.07 X lo5 (Kaufmann et al., 1984), the sedimentation observed value of B, an effective charge of -37/107 kDa is equilibrium results indicate that the 2.6 S species is a moncalculated. Flexible models for the polypeptide conformation omer. give an even smaller value of the excluded volume contribuSedimentation Studies onNF-M at Higher IonicStrengthtion to B. Assuming that the effective charge is 45 f 15% of Since, like other IF proteins, NF-M is capable of forming the actual “titration charge,” a value of 82 5 20 is obtained homopolymer IF, we found it surprising that NF-M did not also exist as a tetramer in dilute alkaline buffers. We, therefor the titration charge. The concentration distribution at sedimentation equilib- fore, sought conditions which might induce tetramer formarium is related to the component defined by Scatchard et al. tion as an intermediate step in the assembly of IF. First, we (1946), as defined in Footnote 4 of the paper by Johnson and raised the salt concentration while maintaining an alkaline Yphantis (1979). In this case the “macromolecular salt” con- pH. Fig. 1C shows the sedimentation velocity pattern observed

‘

17012

Large Neurofilament Proteins Aqueous Solution in

the in 25 mM Tris, pH 8. The principal boundary had a n s20,w of tratedguanidine hydrochlorideconsiderablyimproved 2.95 S,and the absence of a flat plateau indicateda significant yield, length,and overall appearance of thereconstituted amount of aggregation. A sedimentation equilibrium experi- filaments. Filaments formed by the reconstitutionof NF-M areshown ment was performed on this preparation. Instead of a plot clearly shorter andsomewhat with downward curvature, a linear plot of In c versus ? was in Fig. 30. These filaments are obtained, witha slope corresponding toa molecular weight of different in appearancefrom the other preparations but have 1.19 X lo5. It is unlikely that the absence of curvature in the a similar average diameter of 12 nm. In this case there was In cversus 9 plotindicates a completedisappearance of virtually no filament formation if the guanidine hydrochloride nonideal behavior with only a small change inionic strength. step was omitted. We wondered whether the effects of guaMore likely, the downward curvature due to nonideality is nidine hydrochloride treatment would be observed in a sedicompensated by an upward curvature due to theheterogeneity mentation velocity experiment. However,when NF-M disof the preparation. solved in 6 M guanidine hydrochloride was dialyzed against Fig. 1, D and E , shows sedimentation patterns in 50 and Buffer A, the sedimentation velocity pattern was similar to 100 mM Tris at pH8. As the salt concentrationwas raised an that of protein directly dialyzed against Buffer A (data not increasing amount of the protein aggregated to heavier ma- shown). SedimentationStudieson NF-H-We had considerably terial, and the size of the heavier material increased. In 50 mM Tris, the principal boundary had an s20,w of 3.13 S , and a more difficulty in observing any distinct boundaries in sedisecond broader boundary with an average szO,wof 13 S com- mentation velocity experiments on NF-H. The sedimentation of this heavier material patterns showed sharply sloping plateau regions, indicating prised 32% of the protein. The nature that this protein hada great tendency toaggregate in a very is unknown, but it is probably larger than a tetramer (see heterogeneous fashion. In a few experiments we succeeded in “Discussion”). A sedimentation equilibrium experimenton observing a principal sharp boundary similar to that observed this preparation gave a n In cversus ? plot with distinct for NF-M. A sedimentation trace from one such experiment upward curvature due to heterogeneity, with weight-average molecular weights varying from 1.19 x lo5 near the meniscus is shown in Fig. 1F. Scans taken early during the run (not to about 1.74 X lo5 a t a concentration of 1 mg/ml (data not shown) revealed that about 41% of the sedimenting material existed as heterogeneous aggregates. The trace shown, taken shown). In 100 mM Tris, the principal boundary had an szu,wof 3.24 after this aggregated material had sedimented to the bottom S,with 48% of the protein forminga broader 17 S boundary. of the cell, shows that the remaining 59% of the material In both 50 and 100 mM Tris the plateau region beyond the sedimented as a sharp boundary with ansPu+,of about 3.2 S. A sedimentation equilibrium experiment was conducted heavier boundary was flat. These heavier boundaries are not as sharp as one would expect for a fully homogeneous collec- with this preparation. The speed of 16,000 rpm was suffition of oligomeric species, but the patterns do indicate that ciently high to pellet virtually all of the aggregated material the range of sedimentation coefficients for the heavier mate- at the bottom of the solutioncolumn. Analysis of the concenrial is fairly narrow.These heavierspeciesmay represent tration distribution of the remaining material yielded a plot intermediates in the assembly of NF-M into100-nm filaments of In c uersus 12 which was approximately linear, witha slope (see “Discussion”). In other experiments(data not shown) the correspondingto amolecular weight of 1.36 x lo5. This p H was lowered while the buffer concentration was kept at suggests that the3.2 S species present in this preparationwas 5-10 mM. Similar results were obtained; most of the protein a monomer. Circular DichroismSpectra of NF-M andNF-H-S’ mce our sedimented as a sharp boundary with an szu,w value of 2.6 to 3.0 S,and broader boundaries with s2u,w values of 7 to 18 S preparations of NF-M and NF-H had sedimentation properties not observed by others, we examined the circular dichrowere observed. Sedimentationin 100 mM Tris could be influenced by ism spectra of these preparations in 10 mM Tris, pH 8. NFsecondary salt effects, i.e. a n electric field caused by Tris+ H (Fig.4B) had a spectrum virtually identical to that reported shown). ions sedimenting more rapidly than C1- ions could accelerate by Lifsics and Williams (1984), as did NF-L (data not the negatively charged protein molecules (Pedersen, 1958). The spectrum of NF-M (Fig. 4A) is somewhat different in Therefore, sedimentation was also performed in 10 mM Tris shapeandmuchgreaterinmagnitudethanthespectrum as a reported by Lifsics and Williams(1984). (The protein concenbuffer(Buffer A) to which 0.1 M NaClwasadded supporting electrolyte. An szu,w value of 2.92 S was observed tration was determined by amino acid analysis of an aliquot of solution removed from the CD cuvette.) The overall shape for the principal species. Reconstitution of N F - M into IO-nm Filaments-The NF-M of the spectrum issimilar to thatof a+@ proteins (Manavalan preparation used in these experiments was examined for its and Johnson, 1983). Twodifferentpreparations of NF-M capacity t o form reconstituted IF. Fig. 3 shows electron mi- from different batches of spinal cord gave virtually identical crographs of various preparations of native and reconstituted spectra. For oneof the preparations the solutionused for the neurofilaments. Native neurofilaments are shown inFig. 3A. CD spectrum was also examined in a sedimentation velocity a sharp Theseneurofilaments were dissolved in 6 M urea, 0.01 M experiment; about82% of the material sedimented as sodium phosphate, p H 7.5, containing 6 mM 2-mercaptoeth- boundary with anszu.w of 2.34 S. anol and0.4 mM PMSF, and stored frozen for severalmonths. DISCUSSION After thawing anddialysis versus6 M guanidine hydrochloride Molecular Weight of Soluble Forms of Neurofilament Proandthenreconstitution buffer (see“ExperimentalProcedures”), filamentsclose in appearance to the starting materialteins-Lifsics and Williams (1984) reportedtheresults of were obtained (Fig. 3B). When reconstitution was performed sedimentation studies on NF-Lin 10 mM Tris, pH 8.5. They with purified NF-L,thefilaments shown in Fig. 3C were reportedthe existence of a species with a sedimentation obtained. They are similar to the other preparationsmore but coefficient of 4.8 S. Sedimentation equilibrium experiments weblike in appearance. In all three preparations the average showed nonideal behavior with molecular weights “equal to filament diameter was about 12 nm. As observed by others or somewhat greater than2lO,OOO” at infinite dilution. Using (Lifsics and Williams, 1984), a first dialysis against concen- a value for the polypeptide molecular weight of 70,000, they

in Aqueous Solution

Neurofilament LargeProteins

FIG. 3. Electron microscopy of reconstituted neurofilaments. A, native neurofilaments. B, reconstituted from unfractionated NF protein. C, reconstituted from purified NF-L. D, reconstituted from purified NF-M. Theburs indicate 0.1 pm.

-

.

~

D

2

-3 190 200 210 220 230 240

200 210 220 230 240 250

17013

-3

wavelength (nm)

FIG. 4. Circular dichroism of neurofilament proteins in Buffer A. A , NF-M, 0.14 mg/ml. B, NF-H, 0.58 mg/ml.

interpreted the data as support for a trimeric structure. With the revised polypeptide molecular weight of 6.2 x lo4 (Kaufmann et al., 1984), these dataseem to indicate that the4.8 S species is a tetramer. Ourown sedimentation equilibrium data (not shown)yielded a molecular weightof 2.36 X lo5and thus support this conclusion. The results presented in this study are the first evidence for the existence of discrete soluble forms of the two larger neurofilament proteins, NF-M and NF-H, in aqueous solution. Surprisingly, the results indicate that under conditions of low ionic strength and alkaline pH, where NF-L and other intermediate filament proteins are tetramers, both proteins exist as forms clearly smaller than tetramers. Under these a 2.6 S monomer with onlya slight conditions NF-M exists as tendency toward aggregation, while NF-H has a very great tendency toward heterogeneous aggregation, with a distinct 3.2 S form comprising the principal component only a t low protein concentrations. The molecular weight of 1.09 x lo5determined from the sedimentation equilibrium experiment can be compared to the polypeptide molecular weights of 1.08 X lo5 determined from sedimentation equilibrium6 M guanidine hydrochloride (Kaufmann et al., 1984) and 1.29 X lo5from Ferguson plots of data from SDS-polyacrylamide gel electrophoresis (Scott

et al., 1985). Analytical gel filtration in 6 M guanidine hydrochloride gave values of 1.06 x IO5(Kaufmann etal., 1984) and 1.48 X lo5(Scott etal., 1985). Kaufmann et al. (1984) chose a “proposed value” of 1.07 X lo5.It appears that the value of 1.09 x lo5corresponds to a monomeric species. The molecular mass values reported for the NF-Hpolypeptide range from 112 kDa (sedimentation equilibrium) to 179 kDa (Ferguson plot) to 138 or 189 kDa (gel filtration), with a “proposed value” of 110-140 kDa. Once again the observed value of 1.36 x lo5 indicates a monomeric structure, although it shouldbe noted that the solution used forthe sedimentation equilibrium experiment washeterogeneous in composition, and the great tendency of NF-H to aggregateprecludesa reliable molecular weight determination. Sedimentation Coefficients-The sedimentation coefficient of NF-M increased as the ionic strength of the solvent was increased. The cause of the increase is uncertain, and there are several possible explanations. One type of explanation involves no change in the size or shape of the protein. First, asoutlined by Pedersen (1985), the slower sedimentation coefficient of the counterions compared to the protein leads to a separation of charge and the generation of an electrical field which retards sedimentation, the “primary charge effect”; this effect is eliminated by a sufficient increase in salt concentration. Second, some of the increase in s could be due to the contractionof the ionic atmosphere and a consequent reduction of the size of the hydrodynamic component responsiblefor the frictional coefficient. Third, a portion of the increase could be due to a “secondary salt effect” from the electrostatic field generated by the increased sedimentation of Tris’ relative to C1-. This effect would not be seen when NaClis used asthesupporting electrolyte, andinfact a smallersedimentation coefficientwasobserved when Tris chloride was replaced with NaC1. Other explanations for the ionic strength effect consider the possibility of a change in the protein. The increased ionic strength might have led to decreased charge repulsion in the tail of the polypeptide, causing it to assumea more compact conformation. Alternatively, the protein might be undergoing

17014

Large Neurofilament Proteins

in Aqueous Solution

a rapid monomer-dimer equilibrium, in which the concentra- amounts of botha-helixand@-pleated sheet. Lifsics and tion of dimer is negligible a t low ionic strength but becomes Williams (1984) presented spectra similar to those reported significant as the ionic strength is raised. If the equilibrium earlier in terms of the positionsof the troughs and crossovers, is sufficiently rapid, a singleboundary might beobserved with but different in magnitude and in detailed shape. They cona sedimentation coefficient which increases with increasing cluded that NF-L had high contents of both a-helix and @protein concentration (Gilbert, 1955, 1959). Further studies sheet, that NF-M had a high content of @-sheetwith little or would be needed to distinguish these possibilities. In any no a-helix, and that the NF-H spectrum wasdifficult t o event, the value of 2.9 s observed with 0.1 M NaCl should interpret. Our own spectra for NF-L and NF-H were very probably be taken as the"true" value of szo,,,,. similar t o those of Lifsics and Williams (1984), but the specThis value of 2.9 S is clearly not consistent with a tetrameric trum of NF-M, though similar in the positions of troughs and structure. The NF-L tetramer ahas sedimentation coefficient, crossovers, was quite different in its detailed shape and greater s ~ of ~ 4.8 S, , and ~ a~ tetramer of NF-M, with a molecular in magnitude. From theellipticity at 208 nm one can estimate weight 66% greater, would certainly have an even greater avalue of 52% helixfor NF-M (Greenfield andFasman, sedimentation coefficient. (If the two-thirds power depend- 1969). The overall shape of the spectrum resembles that of ence of the s value on molecular weight observed for globular a+p proteins (Mannavalan and Johnson, 1983). The use of proteins is assumed, a tetramer of NF-M would have a sedi- amino acid analysis for our determination of protein concenmentation coefficient of 6.9 S; however, the validity of this tration and the reproducibility of the spectrum on samples relationshipisquestionable for the rodlike neurofilament from two different batches of spinal cord give us confidence oligomers.) that this result is accurate. rule out adimeric The observed s values alonedonot The circulardichroism spectrum thus indicates that NF-M structure. The dimers of other IF polypeptides observed in 3 in aqueous solution has a high content of a-helix. Like other M guanidine hydrochloride hadsedimentation coefficients intermediate filament proteins, NF-M has a central region calculated as s~~,,,, varying from 2.4 S for the dimer of rat which is predicted to form a double-stranded coiled coil in cytokeratin D to 3.4 S for desmin (Quinlan et al., 1986). The solution. The circular dichroism results suggest that this core , , ,2.9 S. The expected s20,w values region does form ahelix in dilute buffers, which is apparently NF-L dimer had an s ~ ~ ,of range from 2.1 to 3.0 S (2.6 S for NF-L). Hence theobserved stable even in the absenceof further interactions leading to a s values for NF-M are not inconsistent with a dimeric struc- dimeric coiled coil. The correlationof the overall shape of the ture. However, the comparison of our data with these results spectrum with that of a + @ proteins suggests that the is difficult, since the sedimentationcoefficients observed in 3 nonhelical ends of the molecule may consist in part of @M guanidine hydrochloride are influenced by additional phys- pleated sheet structures. ical chemical factors.' Furthermore, in 3 M guanidine hydroOur spectrum for NF-H is virtually identical to that rechloride the nonhelical endsof the molecules may have been ported by Lifsics and Williams (1984). Use of the ellipticity unfolded, and greaters values might have been observedunder a t 208 nm leads to a value of 44% helix (Greenfield and less strongly denaturingconditions. It is to be noted that the Fasman, 1969), but it is tobe noted that in the earlier study magnitude of thesedimentation coefficientsobserved by the spectrum could not be reasonably fit with a linear comQuinlan et al. (1986) did not correlate with the polypeptide bination of reference spectra representing various secondary chain molecular weights, and thismay be due toa differential structures (Lifsics and Williams, 1984). sensitivity of the nonhelical ends of the molecules to unfolding Assembly Properties of NF-"When the salt concentrain 3 M guanidine hydrochloride. tion was gradually raised the 2.9 S species, which initially The observed s values do appear tobe more consistent with comprised more than 90% of the total, began to aggregate to a monomeric structure. Calculation of the frictional ratio, f / gradually larger structures. Both the size of the aggregates fo (Oncley, 1941), from an s value of 2.9 S combined with a and the fraction of protein formingaggregates increasedgradmolecular weight of 107,000 and a partial specific volume of ually as the sale concentration was raised. We have no evi0.729 g/cm3 gives a value of 2.6, whereas use of a molecular dence that these large aggregated forms represent true interweight of 214,000 gives the unusuallyhigh value of 4.1. If one mediatesinthe assembly of filaments,althoughtheyare assumes a hydration value of 0.4 g of water/g of protein perhaps lessheterogeneous in size than one would expect (Cantor and Schimmel, 1980), the monomerwould correspond from completely nonspecific aggregation. The sedimentation to a prolate ellipsoid of axial ratio 26 or a cylinder of axial coefficientsobservedfor these aggregates were generally ratio 21, and the dimer toa prolate ellipsoid of axial ratio 74 larger than onemight expect for an NF-M tetramer. We have or a cylinder of axial ratio60. For comparison,f/fo of the NFthus far seen no indication in sedimentation experiments of L tetramer is 2.9 and for the dimer 3.4. Of course, neither the formation of discrete tetrameric assembly intermediates. ellipsoidal nor cylindrical models maybe accurately applicable This suggests that the assembly of NF-M into homopolymer to a helical rod with nonhelical heads andtails. a differentpathwaythanthat Secondary Structure of NF-"The circular dichroism filaments may proceed by followed by other IF proteins. spectrum of NF-M is different from results reported previThe failureof NF-M to form tetramers does not appear to ously. Two sets of circular dichroism spectra have appeared. The results of Delacourte et al. (1982),obtained with proteins be correlated with the inability of the protein to form filaof NF-M have dissolved in 0.5 M NaF, 10 mM sodium phosphate, pH 7.0, mentous structures. Two different preparations been used in these studies; both form a 2.6 S structure in were interpreted toshow that all three proteins had significant dilutebuffersandboth form short pieces of filamentous shown in Fig. 3 0 , * In addition to the correction for solvent viscosity and density, an material upon reconstitution similar to that of native neurofiladditional factorofabout 0.85 could be applied to correctfor the whose diameter of 12 nm is similar to that effects of preferential solvent interactions in 3 M guanidine hydro- aments. A first dialysis against 6 M guanidine hydrochloride chloride (Reisler and Eisenberg, 1969; Reisler et al.,1977). This would was essential for filament formation. In our hands for both yield corrected s20,u,values of 1.6-2.4 S for the various dimers, with a NF-LandNF-Mtheparticular sequence of stepsinthe value of 2.2 S for NF-L. There may be an additional decrease in the renaturation protocol strongly affected the yield of reconstis values observed in guanidine hydrochloride due to secondary salt tutedfilaments.Inthis regard itshould be notedthat a effects (Pedersen, 1958).

Large Neurofilament Proteins number of publications which appeared before the successful experiments of Gardner et al. (1984) reported that NF-Mwas unable to form filaments by itself (Geisler and Weber, 1981; Liem and Hutchison, 1982; Zackroff et al., 1982). Since we have no quantitative information on the yield of reconstituted filaments,we cannot exclude the possibility that the 2.6 S species is incompetent for assembly and that filament assembly arises from that small fraction of material whichdoes notremain monomeric.However, thegradual aggregation of the bulk of the protein to larger forms upon the addition of salt suggests that the bulkof the protein does have a capacity to assemble. Furthermore, we have recently demonstrated that under defined conditions NF-M and NFL form hybrid oligomers i n s ~ l u t i o n . ~ preparation The of NFM used in these studies was capable of forming sucholigomers in high yield. Concluding Remarks-Further work needs to be done in several areas. The soluble form of NF-M can be examinedby electron microscopy. Preliminary studiesof negatively stained preparations have failed to yield any consistent readily interpretable images, and further work should employ different as rotary shadowing or sample preparation methods, such deep etch. The assemblyprocess canalsobe followed by electron microscopy t o visualize thechangesoccurringin solutions of NF-M upon the additionof salt or a decrease in pH. Finally, the role of phosphorylation in the behavior of NF-M needs to be examined. Although Wong et al. (1984) indicated that phosphorylation of NF-M is necessary for its incorporation into reconstituted filaments along with NF-L, a subsequent study (Georges et al., 1986) indicated that dephosphorylation of a mixture of the three neurofilament proteins had no effect on the ability of the three proteins to be incorporated into reconstituted filaments. It will be interesting to examine the effect of the phosphorylation stateof NFM on the solution behaviorof the protein. Acknowledgments-We wish to thank Kevin Burke of the Biochemistry Department at the University of California at Irvine for performing the amino acid analyses, Dr. Joseph Horwitz of the Jules Stein Eye Institute at the University of California at Los Angeles (UCLA) for running the circular dichroism spectra, Rye Hefley and Tai Do for writing and modifying the computer program used for analysis of the sedimentation equilibrium data, Dr. David Yphantis for advice on the analysis of the charge effect, Dr. James Lake of UCLA and Dr. Andrew Z. Mason for advice on the electron microscopy, and Dorothy J. Caughey for technical assistance. REFERENCES Babul, J., and Stellwagen, E. (1969) Anal. Biochem. 2 8 , 216-221 Bradford, M. M. (1976) Anal. Biochem. 7 2 , 248-254 Cantor, C.R., and Schimmel, P. R. (1980) Biophysical Chemistry, Part II. Techniques for Structure and Function, pp. 550-555, W. H. Freeman and Co., New York Dahl, D., and Bignami, A. (1985) Cell Muscle Motil. 6 , 75-96 Delacourte, A,, Filliatreau, G., Boutteau, F., Biserte, G., and Schrevel, J. (1980) Biochem. J . 1 9 1 , 543-546 Delacourte, A., Dousti, D., and Loucheux-Lefebvre, M.-H. (1982) Biochim. Biophys. Acta 7 0 9 , 99-104 Fraser, R. D. B., MacRae, T. P., Suzuki, E., Parry, D. A. D., Trajstman, A. C., and Lucas, I. (1985) Znt. J. Biol. Mucromol. 7 , 258-274 Gardner, E. E., Dahl, D., and Bignami, A. (1984) J. Neurosci. Res. 1 1 , 145-155 Geisler, N., and Weber, K. (1981) J . Mol. Biol. 1 5 1 , 565-571 Geisler, N., and Weber, K. (1982) EMBO J. 1, 1649-1656 Geisler, N., Plessmann, U., and Weber, K. (1982) Nature 2 9 6 , 448450 H. Hajarian, F. Mogadam, and J. Cohlberg, unpublished experiments.

in Aqueous Solution

17015

Geisler, N., Kaufmann, E., Fischer, S., Plessmann, U., and Weber, K. (1983) EMBO J. 2 , 1295-1302 Geisler, N., Fischer, S., Vandekerckhove, J., Plessmann, U., and Weber, K. (1984) EMBO J. 3 , 2701-2706 Geisler, N., Fischer, S., Vandekerckhove, J., Van Damme, J., Plessmann, U., and Weber, K. (1985) EMBO J. 4, 57-63 Georges, E., Lefebvre, S., and Mushynski, W. E. (1986)J. Neurochem. 47,477-483 Gilbert, G. A, (1955) Discussions Faraday Svc. 20,68-71 Gilbert, G.A. (1959) Proc. R. SOC.Land. B. Bid. Sci. A - 2 5 0 , 377388 Greenfield, N., and Fasman, G. D. (1969) Biochemistry 1 0 , 41084116 Hirokawa, N., Glicksman, M.A., and Willard, M. B. (1984) J. Cell Bid. 9 8 , 1523-1536 Hoffman, P. N., and Lasek, R. J. (1975) Int. Reu. Cytol. 66,351-366 Hogue-Angeletti, R. A., Wu, H.-L., and Schlaepfer, W. W. (1982) J. Neurochem. 3 8 , 116-120 Horwitz, J., Bullard, B., and Mercola, B. (1979) J. Biol. Chem. 2 5 4 , 350-355 Huiatt, T. W., Robson, R.M., Arakawa, N., and Stromer, M. H. (1980) J. Biol. Chem. 2 5 5 , 6981-6989 Ip, W., Hartzer, M. K., Pang, Y.-Y. S., and Robson, R. M. (1985) J. Mol. Biol. 1 8 3 , 365-375 Johnson, M. L., and Yphantis, D.A. (1978) Biochemistry 17, 14481455 Jones, S. M., and Williams, R. C., Jr. (1982) J . Bid. Chem. 2 5 7 , 9902-9905 Julien, J.-P.,and Mushynski, W. E. (1982) J . Biol. Chem. 2 5 7 , 10467-10470 Kaufmann, E., Geisler, N., and Weber, K. (1984) FEBS Lett. 1 7 0 , 81-84 Laemmli, U. K. (1970) Nature 2 7 7 , 680-685 Lake, J . A. (1979) Methods Enzymol. 6 1 , 250-257 Liem, R. K. H., and Hutchison, S. B. (1982) Bivchemistry 2 1 , 32213226 Liem, R. K. H., Yen, S.-H., Solomon, G. D., and Shelanski, M.L. (1978) J. Cell Bid. 7 9 , 637-645 Lifsics, M. R., and Williams, R. C., Jr. (1984) Biochemistry 23,28662875 Manavalan, P., and Johnson, W. C., Jr. (1983) Nature 305,831-832 Metzuals, J., Montpetit, V., and Clapin, D. F. (1981) Cell Tissue Res. 2 14,455-482 Nelson, W. J., Vorgias, C. E., and Traub, P. (1982) Biochem. Biophys. Res. Commun. 106,1141-1147 Oncley, J. L. (1941) Ann. N. Y. Acad. Sci. 4 1 , 121-150 Pedersen, K. 0.(1958) J. Phys. Chem. 6 2 , 1282-1290 Perkins, S. J. (1986) Eur. J . Biochem. 1 5 7 , 169-180 Quinlan, R. A., Cohlberg, J. A., Schiller, D.L., Hatzfeld, M., and Franke, W. W. (1984) J. Mol. Biol. 178,365-388 Quinlan, R. A., Hatzfeld, M., Franke, W. W., Lustig, A., Schulthess, T., and Engel, J . (1986) J . Mol. Biol. 1 9 2 , 337-349 Reisler, E., and Eisenberg, H. (1969) Biochemistry 8 , 4572-4578 Reisler, E., Haik, Y., and Eisenherg, H. (1977) Biochemistry 16,197203 Roark, D. E., and Yphantis, D. A. (1969) Ann. N . Y. Acad. Sci. 1 6 4 , 245-278 Roark, D. E., and Yphantis, D. A. (1971) Biochemistry 10,3241-3249 Scatchard, G., Batchelder, A. C., and Brown, A. (1946) J. Am. Chem. SOC.68,2320-2329 Scott, D., Smith, K. E., OBrien, B. J., and Angelides, K. J. (1985) J. Biol. Chem. 2 6 0 , 10736-10747 Steinert, P. M., and Parry, D.A.D. (1985) Annu. Reu. Cell Bid. 1, 41-65 Steinert, P. M., Idler, W.W., Aynardi-Whitman, M., Zackroff,F., and Goldman, R. D.(1982) Cold Spring Harbor Symp. Qmnt. Biol. 46,413-429 Tanford, C. (1961) Physical Chemistry of Macromolecules, p. 196, John Wiley and Sons, New York Wong, J., Hutchison, S. B., and Liem, R. H. K. (1984) J . Bid. Chem. 259, 10867-10874 Woods, E. F., and Gruen, L. C . (1981) Austr. J. Bid. Sci. 3 4 , 315326 Yphantis, D. A. (1964) Biochemistry 3 , 297-317 Zackroff, R.V., Idler, W. W., Steinert, P. M., and Goldman, R. D. (1982) Prvc. Natl. Acad. Sci. U. S. A. 7 9 , 754-757