Peptides with more than one 106-amino acid sequence motif are needed to mimic the structural stability of spectrin.

The primary sequence of human erythrocyte spectrin contains repetitive homologous sequence motifs of approximately 106 amino acids with 22 such motifs in the α-subunit and 17 in the β-subunit. These homologous sequence motifs have been proposed to form domains with a triple-helical bundle type structure (Speicher, D. W., and Marchesi, V. T. (1984) Nature 311, 177-180; Parry, D. A. D., Dixon, T. W., and Cohen, C. (1992) Biophys. J. 61, 858-867). In this study, we show that these sequence motifs, while they do form compact proteolytically resistant units, are not completely independent. Peptides composed of two or three such motifs in tandem are substantially more stable than peptides composed of a single motif, as measured by proteolysis or by fluorescence or circular dichroism studies of urea or thermal denaturation. Circular dichroism and infrared spectroscopy measurements also indicate that these larger, more stable peptides exhibit greater secondary structure. In these respects, the peptides with tandem sequence motifs are more similar to intact spectrin than the peptide with a single sequence motif. Thus, we conclude that peptides with more than one sequence motif model spectrin more adequately than the peptides with one sequence motif, and that these sequence motifs are not completely independent domains.

Human erythrocytes contain a dense, two-dimensional network of spectrin and other proteins that provides support to the lipid bilayer and maintains erythrocyte deformability (1). Spectrin, comprising ␣and ␤-subunits, plays a critical role in maintaining the architecture and therefore the integrity of the red cell membrane. Many hereditary hemolytic anemias involve spectrin mutations (2)(3)(4). Thus, it is important to understand the structural properties of spectrin. The bulk (about 90%) of the primary structure of spectrin comprises repetitive homologous units of approximately 106 amino acids in length. Several other proteins, including brain spectrin (fodrin), dystrophin, and ␣-actinin, also have similar repetitive amino acid units in their sequences and are known as the spectrin superfamily (5).
A triple-helical bundle model has been suggested for the 106-amino acid sequence motif in which the three helices are aligned side by side, with the first and third parallel and the intervening second helix antiparallel (6 -8). X-ray diffraction studies of one such sequence motif unit from a non-erythroid spectrin support this model (9). In these x-ray studies, the peptide used was found to form a homodimer containing two triple-helical structures, in which two of the three helices are contributed by one monomer and the remaining helix by the other monomer. It is thought that this peculiar arrangement may be an artifact of crystallization, and the true structure of a single unit may be similar to the earlier suggested triplehelical bundle with a zigzag arrangement of the helices (9,10). This arrangement aligns the amino-and carboxyl-terminal residues at opposite ends of this triple-helical bundle, and thus sequential motifs are thought to be linked in tandem in intact spectrin, producing a very long rod-shaped molecule, approximately 100 nm in length, as seen by electron microscopy (11).
Circular dichroism (CD) (12,13) and Fourier transform infrared (FTIR) 1 (14) studies of intact spectrin corroborate this model, indicating that spectrin consists of a large proportion of ␣-helix. However, little experimental information is available on the molecular structure of intact spectrin, in part due to its large size and its structural flexibility which makes nuclear magnetic resonance or x-ray studies difficult. Recombinant DNA methods have been used to prepare more tractable fragments of spectrin which may serve as models for spectrin, as in the aforementioned x-ray studies. However, the question of which fragment of spectrin to choose as an appropriate model is important. The two factors that need to be considered are phase and length of the fragment.
The issue of appropriate phase results from the fact that the homologous sequences are joined in tandem with no obvious intervening delimiting sequences, making the initial residue of the folded domains unclear. The start of the sequence for an appropriate structural unit may not be assumed to coincide with the start of the homologous sequence motif in the primary sequence, especially since the onset of the repetitive homologous sequence occurs with different phases in the amino acid sequence of different members of the spectrin family of proteins.
We have used limited proteolysis to identify Ser-52 as the first residue of the first structural domain of human erythrocyte ␣-spectrin (15). This places the start of the first structural domain after the onset of the homologous sequence motif (6), indicating that there is a fractional motif at the amino termi-nus in this protein. Proteolysis experiments to delimit compact structural domains have also been done by others on invertebrate spectrin (16), as well as on actinin (17) and dystrophin (18). These studies have shown that the domain boundaries in these spectrin or spectrin-like proteins are similar and may be identified by a pair of highly conserved tryptophan residues at positions 17 and 90 into the domain.
The issue of length is another matter. The proteolysis work has clearly shown that a single sequence motif folds into a compact structure, but does this structure capture the properties of the corresponding region in intact spectrin, or are larger fragments necessary? In order to assess this question, more detailed examination of the proteins involved are necessary. We have addressed this issue by producing peptides spanning the first three homologous sequence motifs of human erythrocyte spectrin, both in isolation and in tandem, and comparing them with intact spectrin in terms of their stability and spectral properties. We found that peptides containing more than one sequence motif were notably more stable to chemical and thermal denaturation, both thermodynamically and kinetically, and that the properties were more similar to those of spectrin. Our results suggest that a single 106-amino acid unit of ␣-spectrin folds into helical bundles as a stable structural unit but does not fully capture the structural and unfolding properties of intact spectrin. However, peptides encompassing multiple 106-amino acid sequence motifs are much more faithful in capturing the properties of intact spectrin.

EXPERIMENTAL PROCEDURES
Purification of intact spectrin from human red blood cells was carried out as before (19,20).
Briefly, cDNA encoding the NH 2 -terminal region of the ␣-subunit of human erythrocyte spectrin (clone ␣3, a generous gift of Dr. B. G. Forget, Yale University School of Medicine, New Haven, CT) was used as a template for polymerase chain reaction amplification of the spectrin gene fragments. Restriction endonuclease sites were incorporated into the primers to facilitate subcloning manipulations and incorporation into the expression vector, pGEX-2T (Pharmacia Biotech Inc.). For example, the sequence of the 5Ј primer used in the synthesis of Sp␣52-156 was 5Ј-cttggatccTCCTATCACTTAca-3Ј, with uppercase letters for the homologous region encoding for the NH 2 -terminal amino acids SYHL. Lowercase letters denoted non-homologous nucleotides, while those underlined denoted the EcoRI site. The sequence of the 3Ј primer was 5Ј-cgggaattcctaCCGCAGCAACTG-3Ј, with uppercase letters for the homologous region encoding for the COOH-terminal amino acids QLLR. Those underlined denoted the BamHI site. All other peptides were prepared with similar primer designs. Amplified fragments were initially subcloned into a polymerase chain reaction cloning vector (pCRScript; Stratagene, La Jolla, CA) and were then excised with appropriate restriction enzymes (BamHI and EcoRI) and cloned into pGEX-2T.
Fusion proteins isolated from E. coli were treated with either thrombin resin (15) or soluble thrombin (Enzyme Research Laboratories, South Bend, IN) to release the spectrin peptides. The use of soluble thrombin involved incubation with thrombin (1 unit/mg of fusion protein) at 37°C for about 3 h, until digestion was complete as judged by SDS-PAGE. Thrombin was then inactivated by addition of phenylmethylsulfonyl fluoride (Aldrich) to 1 mM.
The rest of the purification procedure was as described previously, except that the buffer pH for the MonoQ FPLC (Pharmacia) purification step was changed to pH 8.1, which moved the elution position for the spectrin peptides to approximately 125 mM NaCl. The spectrin peptides were then further purified, if necessary, by molecular sieve chromatography over a G-75 (Pharmacia) column in 5 mM phosphate buffer containing 150 mM NaCl at pH 7.4 (PBS). Peptide purity was confirmed by SDS-PAGE (12-16%). Peptides were also checked by amino-terminal sequencing (Biocore Facility, University of Notre Dame, South Bend, IN), and mass spectrometry using time of flight/matrix-assisted laser desorption ionization methods (Wistar Mass Spectrometry/Sequencing Facility, Philadelphia, PA).
To determine whether the peptides formed aggregates in PBS solution, the molecular masses of the peptides in solution, at 0.1-1 mg/ml concentrations, were determined at 35°C with an absolute molecular mass determination instrument (PD2000; Precision Detectors, Inc., Amherst, MA). The intensity of light scattered at 15°and 90°from a 630-nm laser, and the refractive index were monitored as the peptide under investigation eluted from a G-75 column. Data were analyzed to give solution molecular mass distribution profiles (22). The intensity of the scattered light is proportional to the molecular mass of the molecules in solution, with the following relationship.
I /I 0 ϭ 2 2 ͑n͑␦n/␦c͒͒ 2 /N A 4 r 2 ϫ ͑1 ϩ cos 2 ͒M ϫ c (Eq. 1) I 0 and I are the intensities of light with wavelength scattered at angles 0 and degrees; n is the refractive index of the pure solvent; ␦n/␦c is the change in refractive index of solutions of the molecule of interest with respect to its concentration, c; N A is Avogadro's number; r is the distance between scattering center and the detector; and M is the molecular mass (22). With the concentration c determined from the refractive index signal (c ϭ ⌬n ϫ ␦c/␦n), the molecular mass was calculated from I /I 0 . The above relation assumes ideal solution behavior. For real solutions, this exact proportionality breaks down, and accurate determinations necessitate extrapolation to c ϭ 0. The use of light scattering in conjunction with chromatography facilitates this extrapolation, as light scattering and refractive index signals are monitored during the elution of the molecules from the column, during which the concentration varied. The precise protein concentrations were determined with absorbance values at 280 nm (A 280 ), using extinction coefficients determined from the primary sequence by ProtParam. 2 Due to the large size of the spectrin molecule, absorbance measurements of spectrin solutions were complicated by scattering. In order to compensate for this effect, the contribution of absorption due to scattering was estimated by both the absorbance at 320 and 350 nm (where the absence of electronic chromophores was presumed) and the dependence of scattering intensity on wavelength (23). This usually resulted in about a 3% correction.
Elastase digestion was performed as described previously (15). Briefly, the peptides, at approximately 1 mg/ml in 50 mM Tris at pH 8.7, were incubated with elastase at 20°C (mass ratio of peptide to elastase was about 1:250), and aliquots were analyzed by SDS-PAGE at 0.5-h intervals for 2 h and then every hour for the next 5 h. Stained gels were dried and digitized with a Scanman II scanner (Logitech, Freemont, CA) at a resolution of 400 dpi. The protein content of individual bands in each lane was determined by integration of digitized images of band intensities, using the SigmaGel program (Jandel Scientific Software, San Raphael, CA). Time constants for the disappearance of each peptide ( g ) were determined by fitting the initial band intensity (I 0 ) and the remaining band intensity (I), at each time point (t), to a simple exponential decay curve, I/I 0 ϭ exp(Ϫt/ g ). Molecular masses of the initial peptides, as well as the degradation products were measured by comparison with standards of known molecular mass.
FTIR spectra were collected as described previously (14). Briefly, samples at ϳ30 mg/ml in 5 mM sodium phosphate buffer at pH 7.4 were used. Spectra were collected using a Bio-Rad FTS 410 FTIR spectrophotometer and a cell path length of 6 m. An interferogram with a size of 4,600 points, corresponding to the 400 -4,000 cm Ϫ1 region, was acquired for 1,024 scans. Buffer and atmosphere (for water vapor) blanks were subtracted from each sample to yield a flat base line in the 1,750 -2,000 cm Ϫ1 region. Center frequencies of the amide I bands were determined by second derivative maxima.
CD measurements on samples at about 1.5 mM residues (A 280 ϳ 0.2) were obtained with a JASCO 710 CD spectrometer, using a thermostated (water jacketed) cell with 0.1-cm path length. Under our buffer conditions (PBS), spectra were limited to wavelengths greater than 190 nm. Spectra were obtained at 0.5 nm resolution from 190 to 300 nm, using subtraction of a buffer blank to smooth the base line. The molar ellipticity at 222 nm, ⌽ 222 (degree cm 2 dmol Ϫ1 ), was calculated. In determining ␣-helical contents for each peptide from CD data, the more usual deconvolution procedures that yield percentage compositions of ␣-helix, ␤-sheet, turns, and random coil (24,25) were not used. These methods rely on a basis set of proteins composed of known percentages of the secondary structural elements and most accurately predict structures of proteins with similar structural elements. Proteins containing structures of coiled coils were generally not included in the basis set. In addition, the specific alignment of the ␣-helices with respect to each other in spectrin is not random (being exclusively parallel or antiparallel), a situation that is known to distort CD spectra (24), further complicating the analysis. Thus, we followed the method used for dystrophin, which has a structure similar to spectrin, by using a value of ⌽ 222 ϭ Ϫ36,000 degrees cm 2 dmol Ϫ1 to represent peptides with 100% ␣-helical structure (18).
Urea denaturation of the peptides was measured by fluorescence spectrometry, using a Hitachi F-2000 fluorescence spectrophotometer at 20°C with a 1-cm path length cuvette. Samples with A 280 ϳ 0.02 in PBS were incubated with urea at different concentrations (0 -6 M) and allowed to equilibrate overnight (16 -20 h). Using an excitation wavelength of 278 nm, emission spectra from 315 to 385 nm for each sample were collected. The values of the mean emission wavelength, with mean defined as shown in Equation 2 (26,27), were calculated from values of intensity at each wavelength (I ) for each spectrum and used to monitor unfolding.
This parameter is a more suitable measure of unfolding than either simple intensity measurements at a particular wavelength (28), or a ratio of intensities at two wavelengths (29). Since mean incorporates the information in the whole emission spectra, it is less noisy and more sensitive to changes in any region of the spectra than those obtained with other generally used methods. The use of this parameter also eliminates the need to single out any particular wavelength for intensity measurements. This is important in multi-step transitions, as certain wavelengths may emphasize one transition over the others and give an incomplete view of the unfolding transition. Most of the mean denaturation data were fit to a single-transition two-state model, in which the free energy change of unfolding (⌬G) was linearly proportional to urea concentration (U), according to the following equation. To simplify the fitted equation, ⌬GЈ was used and was defined as ⌬G/RT.
mean,0 is the mean value without urea and mean,U is the mean value at each urea concentration, U. Parameters to be fitted from the experimental points were ⌬GЈ 0 , [urea] mid , mean,0 , and ⌬ mean . Some data clearly showed two transitions, implying the presence of three states (native, intermediate, and denatured states). These data were then fitted to a three-state model, which was extended from the two-state model.
In order to determine if the samples were indeed at equilibrium, rates at which peptides were denatured by 4 M urea (a minimal urea concentration that sufficiently unfolded all of our peptides) were examined. In these experiments, a 200-l aliquot of protein sample was rapidly added, at 20°C with stirring, to 1,800 l of 4.44 M urea in PBS to yield a final urea concentration of 4 M. Emission spectra were taken as above at intervals, from 1 min to 250 min. The data were fit to a multi-exponential decay curve. The choice of single, bi-, or higher exponential was made by examining the 2 values for the fitted curves. The model with the lowest 2 value was chosen, and in all cases was a bi-exponential decay.
Fluorescence time constants ( f1 and f2 ) were obtained from these curve fits.
Denaturation by urea was also monitored by CD. In this experiment, peptide samples with a final A 280 ϳ 0.4 were made up to various concentrations of urea in the range 0 -6 M and allowed to equilibrate overnight. ⌽ 222 was measured for each sample at 20°C. Thermal denaturation was examined by scanning the temperature upward from 5 to 75°C, at 1°C/min, and monitoring ⌽ 222 . The temperature of the maximum value of d⌽ 222 /dT for each peptide was reported as T m .

RESULTS
All six spectrin peptides were purified to a high degree of purity. For example, gel scan analysis of SDS-PAGE of Sp␣52-156, Sp␣52-262, and Sp␣52-368 (lanes 2 in Fig. 2, A-C) demonstrated that these peptides were at least 90% pure. Occasionally, trace amounts of glutathione S-transferase at ϳ27 kDa were also detected in peptide samples. However, most preparations provided samples with about 99% purity. Molecular masses as estimated from electrophoretic mobilities were ϳ11 kDa for the three peptides containing 105/106 amino acids (Sp␣52-156, Sp␣157-262, and Sp␣263-368), ϳ29 kDa for peptides with two domains (Sp␣52-262 and Sp␣157-368), and ϳ37 kDa for Sp␣52-368. NH 2 -terminal sequencing of all six peptides confirmed the identities of the first 2 residues of each construct as GS (remaining from the thrombin cleavage site), followed by the appropriate residues of spectrin. Time of flight/ matrix-assisted laser desorption ionization mass spectral analysis of Sp␣52-156, Sp␣52-262, and Sp␣52-368 demonstrated that the peptide masses were 12.64, 24.86, and 37.05 kDa, respectively, and were within 0.4% of the theoretical masses (12.65, 24.87, and 36.92 kDa, respectively).
Solution molecular masses obtained from 15°and 90°light scattering data for Sp␣52-156, Sp␣52-262, and Sp␣52-368 in PBS were 10.4, 20.4, and 31.1 kDa, respectively (Fig. 1). These results clearly showed that these peptides, at concentrations of 0.1-1 mg/ml in PBS, existed in monomeric form. That these light scattering measurements were able to detect oligomers of similar peptides was shown by measurements of the solution molecular mass of a protein encoding the first 446 amino acids of the spectrin sequence (Sp␣1-446). This peptide exhibited substantial amounts of dimer and higher order oligomers, indicated by the arrows in Fig. 1.
With prolonged incubation, multiple-domain peptides (Sp␣52-262, Sp␣157-368, and Sp␣52-368) yielded digestion products of about 100 -130 amino acid residues (12-15 kDa), corresponding to fragments presumably containing a single sequence motif of 106 amino acids. The enzyme cleavage sites for the NH 2 -terminal portion (residues 1-167) of spectrin have been given previously (15). The specific molecular masses measured for these products in Sp␣52-262 samples were 14.5 kDa (appeared after 0.5 h of incubation, Fig. 2B) and 12.9 kDa (Fig. 2B, lanes 7-11). Sp␣52-368 (37.2 kDa; Fig. 2C, lane 2) was first digested to 30.2-and 27.8-kDa fragments (Fig. 2C, lanes  4 -11), corresponding to fragments containing two sequence motifs, and then to 14.2 kDa (Fig. 2C, lanes 6 -11), fragments containing a single sequence motif. In each case, bands near but slightly larger than the expected size (14.5-and 12.9-kDa bands during the digestion of Sp␣52-262, when a single motif of about 11-kDa band was expected; and 30.2-kDa band for Sp␣52-368, with two motifs of about 29-kDa band expected) probably reflects site preference specificity of elastase. In a previous study, we identified a similar product as being a whole domain plus the third helix of the previous domain (15). However, these larger fragments were not as stable as the smaller fragments and were converted to the latter products over time. The fact that these products accumulated, while no products smaller than the expected size accumulated, emphasizes the intrinsic stability of these fragments.
CD spectra of all peptides exhibited characteristics of high ␣-helical content. A typical spectrum of Sp␣52-262 in PBS at 20°C (Fig. 3A, solid line)  to helical contents of 51-57% for peptides with single domain, 71-83% for peptides with two domains, and 76% for the peptide with three domains (Table I). When treated similarly, the ⌽ 222 value obtained for intact spectrin dimer corresponded to a 64% helical content. This is consistent with a value of 69% as measured by vibrational CD (14), or 68% as measured by CD (30).
The FTIR central frequency of the main component of the amide I band (assigned to the ␣-helix of spectrin, see Ref. 14) was about 1652 or 1653 cm Ϫ1 for peptides with single domain, and 1650 or 1651 cm Ϫ1 for peptides with two or three domains ( Table I). The value for spectrin was 1650 cm Ϫ1 (14). Although these values were all indicative of ␣-helix structure, the frequency shifts reflected changes in hydrogen bond strength within helices, with lower frequencies corresponding to stronger or more extensive H-bonding networks (14).
In summary, both FTIR and CD data revealed that the spectrin peptides exhibited high helical contents, suggesting that these recombinant peptides were well folded, presumably as triple-helical bundles for each 105/106 amino acid residues. These results are in good agreement with the elastase digestion results. However, the FTIR and CD measurements also indicated that peptides with single domain were slightly less well folded than peptides with multiple domains or intact spectrin.
CD spectra of our peptides at 75°C were virtually featureless, implying that the peptides were extensively denatured from their native ␣-helical conformations at this temperature. The spectrum of Sp␣52-262 is shown as an example in Fig. 3A  (dotted line). Thermal denaturation as monitored by CD showed that T m values for single-domain peptides were 40, 48, and 50°C and for two-domain peptides were 50 and 56°C. The T m value was 54°C for the three-domain peptide (Table I). Thus thermal denaturation results suggested that peptides with a single domain were less stable than peptides with multiple domains. Spectrin denatured at T m ϭ 47°C, indicating that at least portions of this molecule were in fact more sensitive to thermal stress.
CD spectra of all peptides in the presence of 6 M urea at 20°C were also virtually featureless, again implying that the peptides were extensively denatured from their native ␣-helical conformations by urea treatments. The spectrum of a Sp␣52-262 sample is shown as example in Fig. 3A (dashed line) (Fig. 3B, Table I). Again, these values indicated that single-domain peptides (Sp␣52-156, Sp␣157-262, and Sp␣263-368) were the easiest to be unfolded by urea, followed by two-domain peptides (Sp␣52-262 and Sp␣157-368). The three-domain peptide (Sp␣52-368) was the most stable peptide with respect to urea unfolding. It was also interesting to note that the unfolding of the larger peptides was more cooperative than the unfolding of the smaller peptides (Fig. 3B), suggesting interactions between domains in the larger peptides.
Fluorescence urea denaturation data of all peptides except Sp␣52-262 (for example see data for Sp␣52-156, Sp␣52-368, and spectrin in Fig. 4) were all well fitted by the two-state, single-transition model used for the CD denaturation experiments. The mean values for [urea] mid ranged from 0.04 -2.5 M for single-domain peptides, 2.8 -3.2 M for two-domain peptides, but were 3.6 M for Sp␣52-368 and 4.1 M for spectrin samples ( Table I). The wavelength with maximum intensity ( max ) of Sp␣52-368, for example, moved from 345 nm to 351 nm (redshifted) in the presence of 6 M urea, indicating that the fluorophores (tryptophan residues) moved from a relatively hydrophobic, solvent-inaccessible environment to a more polar, solvent-exposed environment. This is consistent with published work showing that tryptophan residues in similar spectrin fragments are in a shielded hydrophobic environment and appear to form a folding nucleation site (31). The general close correlation of the values of [urea] mid measured by both CD and fluorescence techniques indicated that the two processes (unfolding of helical structures monitored by CD, and the release of buried tryptophan residues to more exposed solvent environment monitored by fluorescence) were generally linked.
Sp␣52-262 exhibited a unique, second transition at a lower urea concentration in fluorescence studies (Fig. 4). The data were analyzed with three-state model to include an intermediate state.
The mean value for [urea] mid for the first transition was about 0.4 M. The second transition was at about 3.2 M (Fig. 4), a value similar to those of peptides that exhibited only a single transition. The low urea transition was most evident in the blue end of the emission spectra. The change was primarily one of quenching, with only a small red shift evident ( max moved from 347 to 353 nm). The high urea concentration transition resulted in a larger red shift ( max moved from 353 to 362 nm) with a slight intensity increase. These results show the merit of using mean (Fig. 4) as the parameter reflecting unfolding. If a single wavelength were used, in addition to producing noisier data, only one of the transitions might be detected without significant contribution of the other. For example, a transition at urea concentration of about 0.4 M was easily seen when the fluorescence intensity at 320 nm was examined, but there was little indication that another transition occurred at higher urea concentrations. Conversely, when the intensity at 370 nm was examined, a transition near 3.2 M was observed, with no sign of the transition at urea concentration less than 1 M.
The possibility that the low urea transition was an artifact of incomplete equilibration was eliminated by examining the rate at which these peptides reached equilibrium in the presence of 4 M urea. It was found that the Sp␣52-156 samples reached their final denatured states very rapidly (within 1 min at 20°C). In other samples (Sp␣52-262, Sp␣52-368, and spectrin), the data were fitted best with a bi-exponential decay, with a fast phase and a slow phase. The fast phase with f1 Ͻ 0.02 h (ϳ1 min) was poorly defined due to the paucity of data at this time scale and will not be further considered. The time constant for the slow phase ( f2 ) was 0.6 h for Sp␣52-262, 1.3 h for Sp␣52-368, and 1.9 h for spectrin. Clearly our incubation time of 16 -17 h was sufficient for our samples to reach thermodynamic equilibrium in the denaturation studies. The additional possibility that these secondary transitions reflect oligomerization equilibria was considered, since some of these spectrin recombinant peptides appear to undergo self association (9,32). However, the observed solution masses for our peptides clearly showed that they existed as monomers in solution under the experimental conditions used for CD and fluorescence measurements. DISCUSSION Common assumptions made in the study of proteins containing repetitive sequence motifs, such as spectrin, are that each repetitive sequence motif folds into a structural domain, and that these domains are similar to each other and constitute classical independent structural units. Thus, when structures of parent proteins are difficult to obtain, peptides of a single sequence motif of the parent protein are often used to study the structure of the parent protein. However, that a polypeptide can form a stable structure is a necessary, but not sufficient, cause for it to be considered as an individually folded, independent structural unit. Independence of folding and stability need to be demonstrated experimentally. Such demonstrations entail examining the properties of the individual domains in isolation, which must be substantially similar to those of the structures when present in the parent protein. In the case of proteins containing spectrin-like sequence motifs, the hypothesis that each sequence motif folds into an independent structural unit has been supported by two main lines of evidence: resistance to proteolysis, which pertains to the first folding criteria, and spectroscopic (CD and fluorescence) evidence, which pertains to the latter phenomenological criteria.
Proteolytic resistance has been widely used and has been employed to study fragments containing different repetitive sequences of various spectrins (15,16), of actinin (17), and of dystrophin (18). The basis of this approach is that a polypeptide that is a complete domain will be able to fold into a compact structure. Since proteolytic sensitivity is correlated to the flexibility (33) and accessibility (34) of peptide structures, these putative domains will be resistant to proteolysis while intervening regions will be susceptible. In all of the above studies, polypeptides expressed became resistant to proteolysis only when they spanned a complete sequence motif and when they started at a specific point within the sequence motif (i.e. they were expressed in the proper phase). Our earlier work showed that a properly phased spectrin sequence motif, containing residues 52-156, was resistant to proteolysis, whereas a fragment containing an alternative phasing, containing residues 49 -155, for example, was much less (Ͼ10-fold) resistant to proteolysis (15). An additional incomplete repetitive sequence motif on either end of a complete sequence motif was rapidly digested, showing that while each complete sequence motif could form a compact structure, fractional sequence motifs could not. Gratzer and co-workers (18) expressed fragments of dystrophin and demonstrated that properly phased single sequence motifs exhibited a high proportion of ␣-helix as measured by CD, while improperly phased sequence motifs, differing by only a few amino acids, exhibited much lower helicity. These experiments clearly showed that these single spectrinlike sequence motifs could fold into a compact structure, which might be true domains. However, one hint that these structures exhibit some cooperativity comes from the observation that when two sequence motifs of actinin are expressed in tandem, they are more resistant to proteolysis than either of the isolated constituent motifs, a terminal fractional motif, or improperly phased motifs (17).
We found that fragments containing one (first, second, or third), two (first two or second and third), and three (first three) sequence motifs of ␣-spectrin were all relatively stable to proteolysis, but eventually succumbed to digestion and produced fragments of about 11-12 kDa, the size of a single sequence motif. These results suggest that, although the single sequence motifs in these multi-motif peptides form some sort of selfcontained and protease-resistant structure, their structures are not completely independent, with Sp␣52-156, Sp␣157-262, and Sp␣263-368 being the least resistant to elastase, followed by Sp␣52-262 and Sp␣157-368, and finally by Sp␣52-368.
Our structural integrity studies indicate a similar stability series. All data, including the FTIR, CD, and/or fluorescence FIG. 4. Fluorescence-monitored denaturation. Fluorescence studies of urea denaturation of Sp␣52-156, Sp␣52-262, Sp␣52-368, and spectrin samples. Urea-mediated denaturation was monitored by fluorescence for Sp␣52-156 (f), Sp␣52-262 (å), Sp␣52-368 (ࡗ), and spectrin (E). The mean emission wavelength mean at each urea concentration was used. All proteins show a red shift upon denaturation. Sp␣52-156 started from a longer wavelength ( mean ϳ 350 nm) than the others ( mean ϳ 344 nm). This may indicate a lesser degree of compaction in this peptide. Sp␣52-262 shows two transitions, one at high urea concentration to a denatured state, and one at low urea concentration between a seemingly completely folded state, with a mean ϳ 345 nm, and an incompletely folded intermediate with a mean ϳ 349 nm, similar to that of native Sp␣52-156. spectra, indicated that the peptides with a single domain (Sp␣52-156, Sp␣157-262, and Sp␣263-368) were folded into compact structures that contained substantial ␣-helix (about 55%). However, the peptides with multiple domains (Sp␣52-262, Sp␣157-368, and Sp␣52-368) exhibited higher helical contents (about 75% or more). FTIR results indicated that single-domain peptides exhibited frequency value at 1652-1653 cm Ϫ1 , slightly shifted from the values seen for the other peptides (1650 -1651 cm Ϫ1 ). Since the exact position of this band within the amide I region has been correlated to hydrogen bond strength, this indicates that the hydrogen bonds in single-domain peptides are slightly weaker than those in the other multiple-domain peptides, consistent with a slightly expanded structure.
The amount of denaturant needed to disrupt these structures also increased along the same stability series, with Sp␣52-368 being the most stable and Sp␣52-156, Sp␣157-262, and Sp␣263-368 being the least stable ones. This stabilization was also seen kinetically with rates of denaturation being much slower for the larger peptides, as well as with thermal denaturation, in which the T m values measured were greater for the larger peptides. In all measurements involving denaturation, Sp␣52-368 was most similar to intact spectrin and the single-domain peptides least similar.
Thus, our data suggest that the spectrin repetitive sequence motifs are not classical independent structural domains, but rather are interactive stabilizing substructures. This may have some bearing on the functional properties of spectrin and, in particular, its role in some genetic diseases, such as hereditary elliptocytosis and hereditary pyropoikilocytosis (35). These diseases are often associated with spectrin mutations, which result in reduced levels of spectrin in the membrane skeleton (36,37). Erythrocyte membrane integrity relies on a number of interactions of spectrin with itself and with other membrane proteins, as well as with lipids, but interactions especially germane to this study are the head-to-head interactions between the ␣␤ spectrin dimers to form the (␣␤) 2 spectrin tetramers. Since the dimer-dimer interaction is primarily thought to be the result of the complementary association of sequences in the first 50 residues of ␣-spectrin with a region at the COOH terminus of ␤-spectrin to form a stable hetero-polypeptide triple ␣-helical bundle (38), it is easy to see how mutations at the terminal ends involved in this interaction may disrupt this process and lead to decreased interaction. Many such mutations are known, and indeed these altered spectrin molecules show impaired self-association (39). It is less easy to see how mutations distal to regions involved in direct interaction, and as far as four sequence motifs (Ͼ400 amino acids) away could disrupt dimer-dimer self association; however, such mutations are known. For instance, a spectrin variant, Alexandria, with a deletion of a histidine at position 469 in the ␣-spectrin chain, exhibits decreased tetramerization (as measured by perturbation of the dimer-tetramer equilibrium in 4°C extracts of erythrocyte membranes) (40). Similarly, a hereditary elliptocytosis spectrin variant, Sfax, resulting in the splicing out of codons 363-371 in helix three of the fourth sequence motif (41), also shows increased levels of spectrin dimers and by inference reduced dimer-dimer association. The conclusion to be drawn from these unusual spectrin mutations in hereditary elliptocytosis patients is that mutations distal to the dimer self-association site may have their effects propagated through the intervening domains to the association site by mechanisms that would imply communication and thus cooperativity between the structural units that are formed by these domains. Thus, clinical observations corroborate our experimental evidence based on proteolytic and denaturant stability, implying that the spectrin-like sequence motif, while forming structures with limited independence of function, are not truly independent domains. Our data also demonstrate that the structural properties of peptides of single sequence motifs differ from those of spectrin molecules. There appears to be substantial communication between structural domains containing these sequence motifs, and in order to adequately mimic the properties of spectrin, several tandem domains must be considered.