The role of the dodecamer subunit in the dissociation and reassembly of the hexagonal bilayer structure of Lumbricus terrestris hemoglobin.

The dissociation of the 3500-kDa hexagonal bilayer (HBL) hemoglobin (Hb) of Lumbricus terrestris upon exposure to Gdm salts, urea and the heteropolytungstates [SiWO] (SiW), [NaSbWO] (SbW) and [BaAsWO] (AsW) at neutral pH was followed by gel filtration, SDS-polyacrylamide gel electrophoresis, and scanning transmission electron microscopy. Elution curves were fitted to sums of exponentially modified gaussians to represent the peaks due to undissociated oxyHb, D (200 kDa), T+L (50 kDa), and M (25 kDa) (T = disulfide-bonded trimer of chains a-c, M = chain d, and L = linker chains). OxyHb dissociation decreased in the order Gdm•SCN > Gdm•Cl > urea > Gdm•OAc and AsW > SbW > SiW. Scanning transmission electron microscopy mass mapping of D showed 10-nm particles with masses of 200 kDa, suggesting them to be dodecamers (a+b+c)d. OxyHb dissociations in urea and Gdm•Cl and at alkaline pH could be fitted only as sums of 3 exponentials. The time course of D was bell-shaped, indicating it was an intermediate. Dissociations in SiW and upon conversion to metHb showed only two phases. The kinetic heterogeneity may be due to oxyHb structural heterogeneity. Formation of D was spontaneous during HBL reassembly, which was minimal (≤ 10%) without Group IIA cations. During reassembly, maximal (60%) at 10 mM cation, D occurs at constant levels (15%), implying the dodecamer to be an intermediate.

The giant, hexagonal bilayer (HBL) 1 extracellular Hbs and chlorocruorin of annelids and vestimentiferans are ϳ60 S proteins with an acidic isoelectric point, high cooperativity of ox-ygen binding, and a characteristically low iron and heme content, about two thirds of normal (1)(2)(3)(4). They represent in many ways a summit of complexity for structures containing globins (5). The most extensively studied Hb is that of the common North American earthworm Lumbricus terrestris. Although it has been the subject of numerous studies since Svedberg determined its mass by centrifugation in 1933, the molecular architecture of this complex of ϳ180 polypeptide chains remains uncertain in the absence of a crystal structure. An early SDS-PAGE study showed that it consisted of at least six subunits (6), four of which were globins, comprising a monomer subunit M (7) and a disulfide-bonded trimer T (8), the remainder being linkers, chains of 24 -32 kDa. The amino acid sequences of the T and M subunits have been determined (9,10). Although only three linker chains were thought to exist (11), only one of which had been sequenced (12), a recent ESI-mass spectroscopy study provided a detailed inventory of all the constituent polypeptide chains and indicated the existence of four linker chains (13). Here we report the results of a study of the dissociation and reassembly of Lumbricus Hb, which support the role of the dodecamer of globin chains [3Tϩ3M] as a principal intermediate in both processes.

EXPERIMENTAL PROCEDURES
Materials-L. terrestris Hb was prepared as described previously in 0.1 M Tris⅐Cl buffer, pH 7.0, 1 mM EDTA, 2 mM phenylmethanesulfonyl fluoride, from live worms collected around London, Ontario (Carolina Wholesale Bait Co., Canton, NC) (4,6). The concentration of the Hb was determined from the absorbance of the native form at 280 nm or of the cyanmet form at 540 nm (4), employing the respective extinction coefficients, 2.063 Ϯ 0.032 ml⅐mg Ϫ1 ⅐cm Ϫ1 and 0.442 Ϯ 0.013 ml⅐mg Ϫ1 ⅐cm Ϫ1 (13). The Gdm salts were purum grade from Fluka AG (9470 Buchs, Switzerland) and urea was from Sigma. The heteropolytungstate salts KSiW (K 8 Klemperer (14).
Analytical Gel Filtration-Low pressure, isocratic gel filtration was carried out at room temperature (20 Ϯ 2°C) employing an FPLC system (Pharmacia Biotech Inc.) and 1 ϫ 30-cm columns of Superose S12 or S6 (Pharmacia). Flow rate was 0.4 ml/min and the eluate was monitored at 280 nm. A constant amount of protein in a constant sample volume, ϳ800 g/200 l, was loaded each time.
Optical Spectrophotometry-The absorption spectra over the the 200 -650 nm range were obtained using an OLIS (Bogart, GA) spectrophotometer employing a Hewlett Packard diode array detector or a Hitachi model 2000 spectrophotometer.
Fitting of Elution Profiles-The elution curves were either digitized on a Summagraphics Summasketch MM18 tablet using a Sigma Scan version 3.0 (Jandel Scientific, Corte Madera, CA) or acquired using the Easyest System 8 (Keithley Instruments, Inc., Rochester, NY) and an IBM PC/386 computer. The elution curve was then fitted as a sum of four EMGs, each representing the undissociated Hb (HBL) and peaks D, TϩL, and M, employing least squares minimization (Peak Fit version 2.0, Jandel Scientific). The EMG function is a convolution of a gaussian and a decreasing exponential and is known to represent well the shape of chromatographic elution peaks (16 -18), where a 0 is the amplitude, a 1 is the center, a 2 is the width of the gaussian, and a 3 is the width of the exponential. The EMG is asymmetric with an exponential tail on the right side; the falloff rate of the tail is controlled by the parameter a 3 . The areas of the individual peaks were plotted as percent of total versus time and fitted to sums of exponentials, using PSI-Plot software (Poly Software International, Salt Lake City, UT) employing the Marquardt-Levenburg method. The acceptability of fits was judged by the absence of systematic trends in the plot of residuals with time.
Dissociation of Lumbricus OxyHb-The dissociating agent was dissolved in 0.1 M Tris⅐HCl buffer, pH 7.0, 1 mM EDTA, and Hb stock solution added to obtain the desired concentration, ϳ3.6 mg/ml. The following dissociating agents were employed: urea, Gdm⅐SCN, Gdm⅐Cl, Gdm⅐OAc, SiW, SbW, and AsW. The dissociations of oxyHb at neutral pH and at pH 8.0 and 8.2 in 0.1 M Tris⅐Cl buffer, 1 mM EDTA, were followed by FPLC at neutral pH.
Oxidation of Lumbricus Hb-The oxidation of Lumbricus oxyHb was effected by the addition of potassium ferricyanide (Fisher) or of sodium nitrite (Aldrich) at molar ratios relative to heme, ranging from 1 to 1000, in 0.1 M Tris⅐Cl buffer, pH 7.0, 1 mM EDTA. The conversion of oxyHb to metHb was monitored using optical spectrophotometry over the 450 -650 nm range and was complete in 5-20 min. The metHb solution was then immediately passed over a 1.5 ϫ 20-cm Sephadex G-25 column to remove the oxidizing agent, and the progress of the dissociation measured using FPLC at neutral pH.
Reassembly of HBL Structure from Completely Dissociated OxyHb-Lumbricus Hb (30 -60 mg/ml) was dissociated in the presence of 4 -8 M urea in 0.1 M Tris⅐Cl buffer, pH 7.0, 1 mM EDTA at room temperature. The completeness of the dissociation to TϩLϩM was checked by FPLC, urea was removed by dialysis against 2 ϫ 1.5 liters of Tris⅐Cl buffer for ϳ2 h and the reassociation followed by FPLC in the presence of Group IIA cations Mg 2ϩ , Ca 2ϩ , and Sr 2ϩ over the 0 -50 mM range; the solutions were kept at 7°C. Alternatively, the TϩLϩM fractions were obtained by preparative gel filtration of oxyHb exposed to 4 M urea, pooled, concentrated by pressure filtration using Centricon 10 concentrators (Amicon Division, W. R. Grace & Co., Danvers, MA), and its reassociation followed in the absence and presence of Ca 2ϩ .
STEM Imaging and Mass Measurement of Unstained Protein-The mass measurements were performed using the STEM at the Brookhaven National Laboratory (19). Preparation of the unstained specimens with TMV fibers as internal mass standards was carried out as described by Kapp et al. (20). Negative staining was with 0.5% (w/w) uranyl acetate. The STEM was operated at 40 kev, a dose level Ͻ10 e/0.1 nm 2 and a resolution of ϳ0. 25 nm. An interactive program (19) was used to select the electron micrographs on the basis of clean background and apparent quality of TMV fibers and protein particles; it computes the background and permits the operator to select the TMV segments for internal mass calibration and the particles for mass measurement. At least 3-5 good TMV segments are generally chosen to calculate the internal mass calibration. The individual particles are selected based on clean background around the particles and absence of visible flaws.

Dissociation of OxyHb by Urea and Guanidinium Salts
Elution Profile of the Dissociation Products- Fig. 1A shows a typical FPLC elution profile of oxyHb dissociated at neutral pH in the presence of 1.5m Gdm⅐Cl. In addition to the undissociated Hb (HBL), three peaks are observed, D, TϩL, and M, at elution volumes corresponding to approximately 200, 60, and 25 kDa. The unreduced SDS-PAGE (inset) shows that the subunit content of undissociated Hb is similar to the native Hb, peak D consists of subunits T (54.5 kDa; chains aϩbϩc) and M (16.6 kDa, chain d), peak TϩL is the envelope of unresolved peaks due to subunit T and the four linker subunits L (L1-L4: 24.1, 24.9, 27.6, and 32.1 kDa), and peak M is the monomer (the masses provided here are from a recent mass spectrometric study of the Hb; Ref. 13).
Fitting of Elution Profiles- Fig. 2A shows a representative fit of an FPLC elution curve with four EMG functions. The elution volumes of the four peaks remained unchanged throughout the course of the dissociation to Ϯ3%, as did the fitted variables a 1 (peak position). Fig. 2B shows a similar fit of an elution curve obtained following reassociation of TϩLϩM in 10 mM Ca 2ϩ .
Dissociation of OxyHb at Zero Time-In these experiments, the oxyHb was loaded on the column right after mixing with the dissociating agent and subjected to FPLC. The time elapsed between mixing and complete penetration of the sample into the column was 110 -120 s. Plots of percent dissociation versus the concentration of the dissociating agent are shown in Fig. 3A. Fig. 3B shows the relative percent of the four peaks as a function of increasing concentrations of Gdm⅐SCN and Gdm⅐OAc. The relative proportion of the dodecamer D is much less in Gdm⅐SCN than in Gdm⅐OAc, the weakest dissociating agent.
Time Course of OxyHb Dissociation in 4 M Urea and the Effect of Ca 2ϩ - Fig. 4A shows the time course of oxyHb dissociation in 4 M urea; although it is almost complete within 2 h, peak D remains constant indicating its stability in 4 M urea. Fig. 4B shows the time course of oxyHb dissociation in 4 M urea in 2.5 mM Ca 2ϩ ; it can be fitted as the sum of two exponentials. However, since substantial dissociation occurs within the dead time of the FPLC method (ϳ2 min), there appear to be at least three separate dissociation processes with apparent t 1 ⁄2 Ϸ Յ1 min, ϳ1 h, and ϳ50 h.   judged by the absence of any trends in the plots of residuals versus time provided at the bottom of each panel. Table I summarizes the amplitudes and kinetic constants determined from the fits.

STEM Imaging and Mass Mapping of Undissociated
OxyHb and Peak D Fig. 6 shows views of unstained, cryolyophilized undissociated oxyHb obtained at 11% (A and C) and 89% (B and D) dissociation, respectively. Fig. 7A shows a histogram of the STEM masses of the complete HBL structures observed at 89% dissociation. Although the mean mass, 3540 Ϯ 260 kDa (n ϭ 120), is similar to the value 3560 Ϯ 130 kDa obtained previously for native Hb (13), the distribution of masses is more asymmetric at the lower end. Fig. 6 (E and F) shows typical views of unstained, cryolyophilized peak D obtained by dissociation in SiW and at pH 8.3, respectively; the observed particles are ϳ10 nm in diameter and histograms of the STEM masses within the range 150 -250 kDa (Fig. 7, B and C) had corresponding mean masses of 200 Ϯ 26 kDa and 195 Ϯ 21 kDa, respectively.

Dissociation of OxyHb by Heteropolytungstates, at Alkaline pH and upon Conversion to MetHb
The complex heteropolytungstate anions SiW 8Ϫ , SbW 18Ϫ , and AsW 27Ϫ are known to form 1:1 complexes with metMb at neutral pH with association constants in the 10 5 to 10 6 M Ϫ1 range and concomitant formation of hemichrome type visible absorption spectra (21). All three dissociate oxyHb; a typical elution curve is shown in Fig. 1B. Fig. 8 (A and B) shows the time courses of dissociation in 4.12 mM and 12.4 mM SiW, together with the fits to sums of two exponentials.
It is well known that HBL Hbs dissociate at ՆpH 8 (7, 22). Fig. 8 (C and D) shows the time courses of dissociation at pH 8.0 and 8.2. Again, it is evident that a third, rapid phase occurs within the dead time of the FPLC (ϳ2 min). Thus, there appear to be three dissociation processes with t 1 ⁄2 Ϸ Յ1 min, 2-22 h, and 50 -1200 h.
An early observation by Ascoli et al. (23) suggested that oxidation of earthworm Hb led to the dissociation of its quaternary structure. We reinvestigated this phenomenon because Lumbricus oxyHb was slowly altered to the met form during the dissociations in urea and Gdm⅐Cl. Fig. 8 (E and F) shows the time courses of dissociation following the conversion of oxyHb to metHb and the removal of oxidant by gel filtration.
The fitted parameters for all the dissociations are provided in Table I.  Fig. 6, panels B, E, and F, respectively. Fig. 9 shows some representative results obtained with the reassembly of HBL structures from completely dissociated oxyHb. In the absence of Group IIA cations, reassembly of was limited, generally much less than 10%. However, in the first 24 h, there is a spontaneous formation of dodecamer as illustrated in Fig. 9A. The same result is also observed in Fig. 9B, which depicts the time course of reassembly to HBL in 5 mM Mg 2ϩ . Fig. 9C illustrates the effect of cation concentration on the extent of HBL reassembly, and Fig. 9D shows that although there may be differences in the extent of reassembly achieved initially, the final [HBL] is remarkably similar for all three cations after 200 h.

Reassembly of HBL Structure
Reassociation, starting with peaks TϩL and M isolated by gel filtration of oxyHb dissociated in 4 M urea, shows that a spontaneous reassociation of T and M to about 20% D had occurred within ϳ6 h prior to the first FPLC (Fig. 10), even though reassembly to the HBL was almost nonexistent (ϳ1%). Fig. 10 also shows the reassembly time courses in 2.5 mM and 10 mM Ca 2ϩ ; although the relative contents of T and M declined steadily, the level of peak D remained fairly constant at 10 -15%. STEM images of unstained HBL[TϩLϩM] are indistinguishable from those of native Hb, and the mass distributions are similar to those determined for native Hb (13). The time courses of HBL reassembly could be fitted reasonably well with a single asymptotic exponential.

A Dodecamer [3Tϩ3M] Is Observed in All Dissociations of
Lumbricus OxyHb-The dissociation of the HBL structure at neutral pH by Gdm salts and heteropolytungstate anions and at mildly alkaline pH (Figs. 1 and 3) provide remarkably similar pictures; a ϳ200-kDa dodecamer D ([3Tϩ3M]), deficient in linker subunits, is always formed in addition to the M, T, and L subunits. In particular, dissociation in the weakest dissociating agent, namely Gdm⅐OAc (Fig. 3), shows that D accounts for about half of the initial dissociation products. The time course of dissociation in 4 M urea (Fig. 4A) also shows that D accounts for 40 -50% of the dissociation products. In addition, it appears that D is fairly stable in the presence of 4 M urea, in agreement with earlier findings (24). Fig. 3 summarizes the effect of urea and several Gdm salts on the dissociation of Lumbricus oxyHb determined by FPLC at zero time. The order of decreasing effectiveness is Gdm⅐SCN Ͼ Gdm⅐Cl Ͼ urea Ͼ Gdm⅐OAc, with the order of the anions in line with the well known Hoffmeister series (25,26).
The order of increasing effectiveness of the three heteropolytungstates, SiW Ͻ SbW Ͻ AsW, appears to be correlated with their total charge and mass, Ϫ8 (3239 Da), Ϫ18 (7178 Da), and Ϫ27 (11,732 Da), respectively, and not with the surface charge density. Although SiW is spherical, SbW is a trigonal pyramid, and AsW is a parallelliped, the charge per unit area is approximately the same: Ϫ1.8, Ϫ2.0, and Ϫ2.1/100 Å 2 , respectively (27).
Effect of Ca 2ϩ on Urea Dissociation of OxyHb-Ca 2ϩ exerts a markedly protective effect on the quaternary structure of oxyHb in the presence of 4 M urea (Fig. 4). Although dissociation is almost complete (ϳ95%) after 2 h in 4 M urea (Fig. 4A), even 2.5 mM Ca 2ϩ reduces dissociation to ϳ75% after 144 h in 4 M urea (Fig. 4B). The maximum protective effect is reached at [Ca 2ϩ ] Ϸ 10 mM (Fig. 4C). Alkaline earth (Group IIA) cations are known to stabilize the HBL structure of annelid Hbs with respect to dissociation at alkaline pH (22,28,29), at acid pH (30,31), as well as thermal unfolding and autoxidation (32). In some cases, such as Amphitrite Hb (33) and Myxicola chlorocruorin (34), Ca 2ϩ is necessary for maintaining the HBL structure even at neutral pH.
The Kinetic Heterogeneity of Lumbricus OxyHb Dissociation-Our results show that dissociation of oxyHb followed by FPLC over several weeks is not accompanied by alteration in the properties of either the starting material or the products. 1) The elution volumes of the undissociated Hb (HBL) and of the products of its dissociation (peaks D, TϩL, and M) remain unaltered.
2) The subunit compositions of all the peaks as judged by SDS-PAGE remain unchanged. 3) The STEM images of the HBL peak at an early (10%) and a late stage of dissoci-ation (89%) indicate no major alterations in dimensions (Fig. 6,  A-D). Furthermore, the STEM mass distribution at 89% dissociation (Fig. 7A) compared to that of the native Hb (13) exhibits only a slight asymmetry at the low end, probably due to the presence of a relatively small number of "deficient" HBLs, missing 1 ⁄6 and 2 ⁄6 of the HBL structure that can be observed in Fig. 6 (A-D).
The time courses of oxyHb dissociation in 1.75 m urea (Fig.  5A) and 1.22 m Gdm⅐Cl at neutral pH can be satisfactorily represented as the sum of three first-order processes with t 1 ⁄2 Ϸ 1-2 h, 30 -50 h, and 400 -500 h (Table I). Fig. 4 also shows that there are at least three processes occurring in the dissociation of oxyHb in 4 M urea in the absence and presence of Ca 2ϩ .
Three first-order processes are also observed in oxyHb dissociation at alkaline pH, t 1 ⁄2 Ϸ Յ1 min, 2-20 h, and 50 -1200 h (Table I). OxyHb dissociation in the presence of SiW (Fig. 8, A and B) can be fitted with two first-order processes, t 1 ⁄2 Ϸ 10 -40 h and 400-1300 h ( Table I). The latter values correspond roughly to the t 1 ⁄2 for the two slower dissociation processes in urea and Gdm⅐Cl and at alkaline pH.
Two points must be considered before discussing possible mechanisms for the dissociation of Lumbricus oxyHb. 1) Whether slow oxidation of oxyHb to metHb could be responsible for one of the dissociation processes observed. MetHb dissociation (Fig. 8, E and F, and Table I) consists of two phases: a small (ϳ10%) initial dissociation (t 1 ⁄2 ϳ 2 h), followed by a dissociation that is slower by more than 1 order of magnitude than the slowest phase of the oxyHb dissociations (t 1 ⁄2 Ϸ 13,000 -35,000 h versus 50 -1300 h). Hence, dissociation due to metHb formation can be neglected. 2) Can the dissociation of the oxyHb be accompanied by a partial disruption of the tertiary and secondary structures of the globin subunits? It is known that myoglobin does not evince any conformational alterations at urea concentrations less than 5 M (35,36). Hence, it is unlikely that 4 M urea affects either the M or the disulfidebonded T subunit.
Possible Mechanisms of OxyHb Dissociation-Several simultaneous dissociations of a HBL structure can be envisaged (  (1) and (3) to TϩM. STEM images of Hb at 11% and 89% dissociation (Fig. 6, A-D) show the presence of deficient HBLs, partially dissociated Hb particles lacking 1 ⁄6 and 2 ⁄6 of the HBL structure (represented schematically in Fig. 11B). We do not know whether these deficient HBLs are intermediates or not. The time course of the appearance of D, which reaches a maximum in the the initial 10% of the dissociation and decreases thereafter (Fig. 5B), is consonant with the formation of deficient HBLs with concomitant formation only of D (Fig. 11B), being the initial stage of HBL dissociation. The latter is reminiscent of the time course of formation of an intermediate B in a simple set of consecutive first-order reactions A 3 B 3 C (37). However, this simple scheme does not fit our results, since: 1) oxyHb dissociation can not be represented by a single exponential and 2) the time courses of TϩL and M appearances (Fig. 5, C and D) do not exhibit an induction period and consequently, an inflection point in their curves, as does the appearance of the final product C in the foregoing model. The latter result suggests that processes (1)-(3) occur simultaneously.
The dissociation of the dodecamer in the presence of urea and Gdm⅐Cl requires two exponentials for a satisfactory fit with t 1 ⁄2 Ϸ 100 -200 h and 2700 -5000 h. 2 The slower process has a t 1 ⁄2 close to that determined for the dissociation of the metdodecamer, which is about an order of magnitude faster than the dissociation of the metHb. It is likely that oxydodecamer dissociation (t 1 ⁄2 Ϸ 100 -200 h) occurs mostly in the later stages of oxyHb dissociation, following the accretion of peak D observed in the first 50 -300 h (Fig. 5B).
There seem to be two simple explanations for the kinetic heterogeneity of oxyHb dissociation. 1) Since peak HBL, whose area is a measure of undissociated HBL structures, contains "complete" HBLs as well as the deficient HBLs lacking 1 ⁄6 and 2 ⁄6 of the structure, one explanation is that the observed three first-order processes reflect the dissociation of the complete and deficient HBLs. However, the STEM appearance and STEM mass distributions at a late stage of dissociation (Figs. 6, B and C, and 7A) indicate the presence of limited numbers of deficient HBLs. 2) Another possibility is that the native Hb consists of three unequal populations of HBL structures differing in their stabilities toward dissociation, each population of HBLs exhibiting its own rate of dissociation in the presence of a given concentration of the dissociating agent. In this view, the deficient HBLs are likely intermediates in the overall dissociations. Our results suggest that the initial, rapid oxyHb dissociation with t 1 ⁄2 of Յ1 min at alkaline pH and 1-2 h in 1.75 m urea and 1.22 m Gdm⅐Cl, which is not observed in the case of SiW, may be related to the ease of penetration into the Hb interior. The penetration of OH Ϫ and its reaction, e.g. with salt bridges stabilizing some intersubunit contacts, should be much more rapid than the penetration by urea or Gdm⅐Cl and their binding to enough peptide groups and/or side-chain groups of the different subunits to effect a similar destabilization. This notion is consistent with the probable inability of the heteropolytungstates to penetrate into the Hb interior and the consequent occurrence of only two first-order processes (t 1 ⁄2 Ϸ 10 -40 h and 400-1300 h), comparable to the two slower processes observed in urea and Gdm⅐Cl and at pH 8.0 (t 1 ⁄2 Ϸ 22-53 h and 400-1200 h, Table I).
Role of the Dodecamer in HBL Structure Reassembly-Our results (Figs. 9 and 10) demonstrate that dodecamers are formed both in the presence and absence of the linker subunits L and in the absence and presence of Group IIA cations. Likewise, in the presence of Mg 2ϩ and Sr 2ϩ (Fig. 9B), but not Ca 2ϩ , there is an increase in D which occurs in the first 24 h, prior to the formation of any significant amount of HBL. These facts suggests that formation of the dodecamer precedes that of HBL and that the dodecamer is an obligatory intermediate in the reassembly of the HBL structure (Fig. 11C). In the presence of Ca 2ϩ (Fig. 10), the formation of HBL is accompanied by only a small decrease in D, the latter remaining at a fairly constant level between 10 and 15%. At optimum concentrations of Mg 2ϩ , 2 P. K. Sharma and S. N. Vinogradov, unpublished observations.