A Compact Form of Rat Liver Mitochondrial DNA Stabilized by Bound Proteins*

A highly folded, rapidly sedimenting form of rat liver mitochondrial DNA has been released from the organ- elles with BRIJ 58 and sodium deoxycholate in the presence of 0.5 M NaCl and isolated by sedimentation velocity in sucrose gradients. Under these conditions a majority of the mitochondrial DNA labeled in vitro sedimented beyond 39 S, the sedimentation coefficient of a highly purified mitochondrial DNA supercoil, and appeared as a stable, heterogeneous population of spe- cies ranging in s values between 42 S and about 70 S. Under formamide-spreading conditions most of the rapidly sedimenting forms appeared in the electron microscope single genome length at the center in a dense core. an extraordinary structural features along the smooth loops projecting radially from the central core. In sucrose gradients of ethidium the sedimenta- tion of the folded DNA a biphasic fashion At

molecular weight of approximately 9 X 10" (see reviews, Refs. 1 and 2). Upon isolation and purification of mtDNA using a deproteinizing procedure, several forms of circular molecules are obtained. The major species is the covalently closed supercoiled molecule having a sedimentation coefficient of about 39 S (3)(4)(5). The open circular form sedimenting at 27 S (3-5) is present to a lesser extent. Its presence can result either from the introduction of at least one single strand scission into the covalently closed circle (3), or as in the case of mouse cells (6,7), as a result of the insertion of a displacing strand (68) that stabilizes the winding deficiency of the parental duplex, thus considerably reducing the number of supercoils (9). In addition to these species, catenated molecules containing two, or more, interlocked circles are found. Catenated dimers sediment between 36 S and 51 S depending upon whether the constituent monomers are open or closed circles. Catenanes are present at a frequency of between 3 to 9% of the mtDNA molecules from normal animal tissues (10). Using electron microscopy, Nass et al. (11) demonstrated that in thin tissue sections stained with osmium tetroxide, the mtDNA appeared in a compact state as a small, rodlike structure in the matrix. It was apparent from this work that the mitochondrial genome, like that of the bacterial chromosome (12,13), must exist inside the organelle in a packaged, physical state that is far more compact than an isolated supercoil. However, the specific nature of the packaged form and the mechanisms for maintaining the supramolecular state of animal mtDNA in vivo have remained obscure.
Additional early work by Nass (14) and later by Van Tuyle and Kalf (15) provided evidence for the association of animal mtDNA with the mitochondrial membrane. More recently, Albring et al. (16) demonstrated that most of the HeLa circular mtDNA released from the organelles by Triton X-100 in the presence of low salt was associated with a proteinaceous structure which varied in appearance in the electron microscope between a lo-to 20-nm knob and a lOO-to 500-nm membrane-like patch. This stucture was found to be bound near the origin of replication.
In all these studies, however, the presence of an isolated membrane/DNA structure that existed in a compact, packaged state was not observed. Pinon et al. (17) have indicated that some of the DNA released from Xenopus laevis oocyte mitochondria by gentle lysis in the presence of spermidine was presumably associated with proteins forming a complex having a sedimentation coefficient of 58 to 60 S. In the electron microscope these structures appeared as small relaxed circles consisting of globules connected by a thin filament.
The work presented here describes a stable compact form of mitochondrial DNA isolated from rat liver. This DNA structure is visualized as a constrained circle forming a rosette with a dense central core. The structure is maintained in the highly folded state by associated proteins. Evidence is also

Isolation
of a Rapidly Sedimenting Form of mtDNA-Mitochondria were labeled in the DNA in vitro and lysed by treatment with the nonionic detergent BRIJ 58 and the ionic detergent sodium deoxycholate in the presence of 0.5 M NaCl. After removing any undissolved particulate material from the suspension by centrifugation, the cleared lysate was analyzed on a 5 to 20% sucrose density gradient. The radioactivity profile of a typical gradient is shown in Fig. 1. The labeled DNA sedimented as a heterogeneous population of species having s values ranging from about 24 S to greater than 70 S. A majority of the labeled DNA sedimented as several broad zones (bracketed material) beyond 39 S and ranged from about 42 S to more than 70 S. This population of DNA molecules, sedimenting beyond the 39 S position, has been designated ">39 S mtDNA." This rapidly sedimenting material has been examined more carefully to elucidate its structural characteristics.
The broad, but apparently discrete, peaks in the >39 S region of the gradient have been fairly consistent in mean s value in different preparations.
However, the degree of labeling of >39 S mtDNA, relative to 39 S and to 27 S DNA (seldom detectable), varied somewhat from preparation to preparation. When the material in the fractions from the lower region of the gradient of Fig. 1 (bracketed  material) was pooled, pelleted by centrifugation, and a portion was resedimented on a second sucrose density gradient ( Fig.  2A), the DNA again sedimented beyond 39 S as a rather broad band composed of species of apparently heterogeneous nature. Virtually no conversion to the 39 S or 27 S forms was observed. Subsequent experiments indicated that >39 S mtDNA could be stored frozen for several months with little or no conversion to slower sedimenting forms. Pancreatic RNase had no effect on the s value, or the shape, of the band of >39 S mtDNA (not shown), but pancreatic DNase degraded it to small fragments that remained at the top of the gradient (Fig. 2B) When another portion of the pooled material was treated by a SDS/phenol purification step, nearly complete conversion to the 39 S form was observed (Fig. XC). These results suggest that the >39 S mtDNA is maintained in a rapidly sedimenting, compact state by tightly bound non-DNA constituents and that the high s value of this form is not a property inherent in the structure of the DNA itself. They also rule out the possibility that the >39 S mtDNA consists primarily of topological multimers, because catenanes and concatenanes would not revert to the monomeric form as a result of a deproteinizing purification step. Summary of Various Lysing Procedures that Yield >39 S mtDNA-In most of the experiments reported here, the mtDNA was released by lysis of the labeled mitochondria with BRIJ 58 and deoxycholate in the presence of 0.5 M NaCl. To demonstrate that the >39 S mtDNA does not arise as a result of this specific lysing procedure, numerous other procedures listed in Table II have been used successfully to  isolate this form. The agents enumerated  in Table IL4 are  either nonionic, or rather gentle ionic detergents such as deoxycholate and Sarkosyl. It was found that when mitochondria were lysed by any of these agents in concentrations of NaCl below 0.3 M, most of the mtDNA was degraded to small fragments ranging in size between 4 S and 10 S. However, in concentrations of NaCl between 0.3 M and 1 M, no degradation was seen, and regardless of the detergent used, essentially identical sucrose density gradient profiles were observed, and >39 S mtDNA was evident. Apparently, these gentle detergents released active nucleases associated with the mitochondrial preparation resulting in the degradation of the exposed mtDNA when the NaCl concentration was below 0.3 M. However, elevated levels of NaCl, known to inhibit mitochondrialassociated nuclease activities (30, 31) allowed the isolation of intact DNA after lysis by these detergents.
It is noteworthy that >39 S mtDNA could be released from the mitochondria by treatment of the intact organelles with pronase. Visible clearing of the suspension revealed that pronase-dependent lysis had occurred. In this case, the yield of total mtDNA was reduced compared to that obtained by detergent lysis. However, the sucrose gradient profile of the released DNA species (Fig. 3)  Aggregate-In spite of the above observation that different lysing agents and procedures could release a major portion of the mtDNA as a rapidly sedimenting form, there was the possibility that the >39 S mtDNA was an artifact resulting from the aggregation of several supercoiled molecules or from the spurious binding of clusters of non-DNA constituents to a single DNA molecule. Such theoretical aggregates could increase the effective mass of the DNA yielding artificial rapidly sedimenting species. To address these possibilities experimentally, highly purified deproteinized mtDNA that sedimented as a sharp peak of 39 S when analyzed alone was mixed into the BRIJ 58/deoxycholate-lysing medium. The mixture was then added to half of a typical preparation of incubated, but unlabeled mitochondria. The other half of the mitochondrial preparation was labeled and lysed with the same medium without exogenous DNA. Fig. 4  the same DNA. Additional increases in dye concentration after the relaxed form is attained result in the introduction of positive superhelices into the molecule and accompanying increases in s value (33).
When nondeproteinized >39 S mtDNA was analyzed in sucrose density gradients containing increasing amounts of free ethidium bromide, a complex pattern emerged (Fig. 5). The apparent single peak seen in the absence of ethidium bromide (Fig. 6A) resolved into two slower sedimenting peaks at 0.4 to 0.7 pg dye per ml (Fig. 5)  At higher dye concentrations, these peaks reconverged into one band again and continued to increase in mean s value until a profile essentially identical with that seen in the dye-free gradient was achieved (Figs. 5 and 6).
If, however, the >39 S mtDNA was first purified by SDS/ phenol extraction prior to such an analysis, the DNA behaved as a single species throughout the ethidium titration curve (Fig. 5). The response to increasing amounts of ethidium bromide was uniform and typically biphasic. In the absence of dye, the deproteinized form sedimented at 39 S, decreased to a minimum of about 27 S at an ethidium bromide concentration of 0.5 pg/ml, and returned to near 39 S in the presence of still higher amounts of dye, Such a response would be expected of a closed circular DNA that had no extraordinary constraints imposed upon its superhelical structure. The observation that both the deproteinized form and the several species of untreated >39 S mtDNA reached s value minima at approximately the same dye concentration (Fig. 5) indicates that all these forms had similar superhelix densities. Perhaps more importantly, the fact that none of the untreated species was able to untwist to the completely relaxed 27 S form indicates that there were additional constraints superimposed upon these molecules such that they were maintained in a form that is more compact than a supercoil.
Electron Microscopic Visualization- Fig.  7 is an electron photomicrograph of >39 S mtDNA that was fixed in 0.1% glutaraldehyde and relaxed in formamide prior to application to the grid. Eighty-two of the 96 molecules examined were seen as compact rosettes that were constrain&d at the center in what appeared to be a rather dense core. The portions of each molecule projecting radially from the central region appeared as relaxed smooth loops. Closer examination of these molecules revealed that the loops were variable in length, and except for several D-loops (arrows), no unusual structural features were obvious along their contour. Moreover, the dense cores did not appear as mere supercoiled regions of the DNA. In some cases their knob-like appearance was suggestive of the presence of non-DNA material at these sites. Estimates of the length of the DNA in the rosettes ranged between 4.68 and 4.99 pm. Acknowledging the difficulties in making these measurements, the contour lengths obtained agreed very well with previously reported values of about 5 pm for the length of single rat liver mtDNA molecules (34, 35). The larger rosette marked (C) in Fig. 7 measured to be about 9.62 pm indicating that two genome length molecules were present in this form. The nature of their attachment to each other, however, was undiscernible from the electron photomicrograph.
Only three of the 96 molecules examined were dimers.
In addition to the tightly constrained rosettes in Fig. 7, one of the molecules shown (B) appears loosely constrained, having little or no evidence of a central, dense core. This molecule may belong to the subpopulation of >39 S mtDNA that was seen previously to be more responsive to ethidium bromide titration (Figs. 5 and 6). Fig. 8 presents an electron photomicrograph of >39 S mtDNA that had been SDS/phenol extracted prior to preparation for electron microscopic examination.
All of these molecules were in the relaxed form, and none possessed a rosette-like appearance. Comparison of the molecules in Figs. 7 and 8 suggests that deproteinization with SDS/phenol, a procedure that should have no effect on the structural characteristics of the DNA per se, probably removes proteins that stabilize the DNA in the tightly folded conformation that gives rise to the >39 S forms. The Association of Proteins with the >39 S mtDNA--To confirm the presence of proteins bound to the rapidly sedimenting DNA and to establish the relative amounts of protein and DNA in the complex, direct chemical quantitative measurement of protein and DNA in resedimented particles from three different preparations revealed a protein to DNA weight ratio ranging between 0.24 and 0.35. Such low values are consistent with the idea that the rapid sedimentation rate of the particle does not result merely from the increased effective mass conferred by the bound proteins and lends credence to the probability that the high s value is predominantly due to its compact folded nature.
SDS-polyacrylamide slab gel analysis of the proteins bound to the >39 S DNA is shown in Fig. 9D. The bound array consists of polypeptides ranging in molecular weight between 150,000 to about 11,000. The most intense band among those from the >39 S mtDNA (i'14~ = 58,000) is undetectable in any of the other submitochondrial fractions (Fig. 9, A, B, and C). This protein's exclusive appearance in the DNA-bound array signals its exceptional specificity for the DNA molecule. In subsequent experiments it was shown that if the DNA was degraded with pancreatic DNase prior to the second sucrose density gradient centrifugation, the fractions of the gradient that correspond to the >39 S region, where the DNA would have sedimented had it not been degraded, contained no detectable protein bands. Thus, any proteins from the mitochondrial lysate not bound to the DNA do not co-sediment as free aggregates in detectable amounts into the middle and lower regions of the gradients.
Further comparisons of the DNA-bound polypeptides with those of the other submitochondrial fractions revealed that the most consistent co-migration existed with the bands obtained from the inner membrane, particularly in the molecular weight range below 30,000. It is clear, however, that not all of the major inner membrane proteins were represented in the  Procedures." Except where 2 M NaCl was used, samples of less than 10 pl of the DNA solution were diluted to 100 ~1 with 0.15 M STE buffer. The diluted samples were treated as indicated in the table and analyzed along with untreated controls on 5-ml gradients of 5 to 20% sucrose in 0.5 M STE buffer. The gradients were centrifuged for 90 min at 50,006 rpm in a Beckman SW 50.1 rotor, and IO-drop fractions were collected directly onto filter paper discs and counted as described under "Experimental Procedures." Where 2 M NaCl was used, both the DNA dilution buffer and the sucrose gradients contained 2 M NaCl. A. Treatments that have no effect on the high s value >39 S array. Furthermore, the relative intensities of the coelectrophoresing DNA-bound bands were different from the relative intensities of the corresponding proteins seen in the total inner membrane profile. Thus, if the lower molecular weight polypeptides associated with the >39 S mtDNA are in fact derived from the inner mitochondrial membrane, then only a specific portion, rather than a large random fragment, of this membrane remains bound to the DNA under these isolation conditions.
The Stabilizing Features of the >39 S &DNA- Table   IIIA contains an extensive list of treatments that were used to probe the structural integrity and stabilizing features of the >39 S mtDNA. Each treatment was chosen on the basis of its probable efficacy in altering protein structure in order to establish the minimal perturbation required to disrupt the DNA-protein interaction and thus unfold the >39 S particle to the 39 S form. It was surprising that none of the treatments listed in Table IIIA  of the particle. Thus, in order to elucidate any role played by the bound proteins specifically in the maintenance of the folded structure, it was necessary to determine whether the enzymatic proteolyses used were digesting any of the proteins associated with the DNA, and if so, which proteins, or portions of proteins, remained bound to stabilize the rapidly sedimenting form. To this end, half of a preparation of >39 S mtDNA was treated with proteinase K prior to the second sucrose gradient purification step. The polypeptides remaining associated with the untreated and digested complexes are shown in Fig. 10. The pattern given by the untreated sample (Fig.  10B) was consistent with previous preparations (compare with Fig. 9D). In the proteinase K-treated sample (Fig. lOA), however, all of the high molecular weight proteins were missing, but one main low molecular weight polypeptide (Mr = 10,500) and a faint band (Mr = 15,500) were detectable under these  conditions. In other experiments a second faint band with a molecular weight of about 18,000 was sometimes seen. Each of these bands appeared to be a discrete partially digested remnant of proteolysis, because none coelectrophoresed exactly with bands from the untreated sample. It was, therefore, apparent that one or more of these tightly bound remnants played a critical role in the maintenance of the compact supramolecular structure of the DNA. Table IIIB presents a listing of the rather severe treatments that were required to unfold the compact structure to the 39 S form. Each of these treatments presumably acts to alter the structural integrity of the stabilizing proteins such that they are no longer able to maintain the back folding of the DNA domains. Although the requirement for such severe treatments is indicative of the extreme stability of the complex, it is clear that covalent modification is not necessary for conversion to the 39 S form. The final entry in the table indicating the efficacy of digestion of the complex with proteinase K in the presence of dithiothreitol at 65°C confirms that proteins do indeed play a central role in the maintenance of the compact structure. Furthermore, as seen in Fig. 11, if any of these three conditions (i.e. digestion with the proteinase, the presence of a disulfide reductant, or elevated temperature) is not met, then unfolding to the 39 S form does not occur. Thus, it can be deduced that portions of the proteins that are crucial to the maintenance of the compact structure can be classified as "tight" (36) polypeptides known to be insensitive to proteolysis except in the presence of a disulfide reductant, and in this case at elevated temperatures. DISCUSSION The results presented here describe a folded, compact form of rat liver mtDNA isolated on the basis of its high sedimentation coefficient. The compact appearance of the isolated structures in the electron microscope is consistent with previously reported descriptions of mtDNA examined in situ (11). The isolated >39 S mtDNA is, therefore, regarded as a physical form of the DNA that closely approximates the in uiuo state of the mitochondrial genome. The high sedimentation value of this particle would suggest that the superhelices of the DNA are further folded intramolecularly resulting in the formation of the compact structure. Direct chemical and electrophoretic analyses reveal that the >39 S mtDNA is associated with tightly bound proteins. Although most of these proteins can be removed by proteolysis without affecting the compact structure of the DNA, portions of some of these proteins are resistant to proteolysis, and these remnants are apparently crucial to the maintenance of the folded form. Furthermore, the observation that dithiothreitol increases the sensitivity of the DNA/protein complex to proteinase K digestion suggests that the stabilizing proteins themselves may be maintained in a tight functional conformation by disulfide bonds. This implication that proteins containing cystine residues may play a role in the maintenance of the structure of mtDNA is noteworthy in light of the previous report (37) that the incorporation of DNA precursors into rat liver mtDNA is inhibited by sulfhydryl compounds, and any incorporation that does occur under these conditions is found only in small fragments that are unassociated with the higher molecular weight parental molecules.
The ineffectiveness of pancreatic RNase in converting the >39 S particle to a slower sedimenting form indicates that nascent RNA such as that present in transcription complexes (38) does not contribute to the rapid sedimentation rate. This observation also suggests that RNA may play no role in the stabilization of the compactness of the rapidly sedimenting form. This would, therefore, be a major difference between the mitochondrial folded genome and the Escherichia coli nucleoid in which both RNA (13,39,40) and protein (40) have been implicated in the maintenance of the tertiary folding. There remains the possibility, however, that RNA may still be a stabilizing component of the >39 S mtDNA, but it may be so highly sequestered in the interstices of the dense DNA core that it is inaccessible to enzymic degradation.
From the work reviewed by Borst and Kroon (41) and by Nass (1) it can be calculated that rat liver contains between 2 to 10 mtDNA molecules per mitochondrion. The specific organization of the several copies of mtDNA inside each organelle is at present unknown. It is possible that several of these circular molecules exist naturally as dimeric or oligomerit clusters that are bound to each other by some non-DNA liason such as a common membrane binding site (14)(15)(16) or other macromolecular coupler. Because of the gentle nature of our isolation procedure, it was postulated that >39 S mtDNA may be the isolated form of such a natural cluster. In the electron microscope analysis, however, no large clusters of molecules were seen. Most of the rosette-like structures were of single genome size. The few larger structures seen were roughly 10 pm in length, and it was assumed that these dimers were interlocked circles because they were present at a frequency well within the range noted for normal rat liver mtDNA isolated by more conventional methods (10). Thus, this evidence suggests that >39 S mtDNA is an isolated packaged form of individual mtDNA molecules.
The biphasic response of the compact DNA to ethidium bromide titration shows that superhelical domains are present in the structure. The reason why there were several species in the untreated >39 S mtDNA that responded differently to ethidium bromide is unclear at this point. Several plausible explanations exist. It is possible that the constraints on the DNA were less stringent on certain molecules in the population. These "looser" structures may have been more easily disrupted and unfolded by the dye than those that were more tightly contained. For example, it is conceivable that molecules actively transcribing RNA may undergo a relaxation mechanism resulting in a loosening of the folded structure to allow RNA synthesis to occur (38). In addition, the compact structure of molecules that are in the process of nascent strand expansion during replication of the DNA must be unraveled and unfolded in order to achieve separation of the parental strands in the replicating regions. The validity of either of these possibilities has yet to be confirmed.
The results indicating that some of the DNA-bound proteins may be derived from the mitochondrial membranes are consistent with other reports (15,16) of the isolation of DNA/ membrane complexes from animal mitochondria using similar gentle lysis methods. However, the observation that the major protein associated with the >39 S mtDNA is undetectable in any of the other submitochondrial fractions clearly demonstrates the specificity of this protein for binding the DNA and suggests that it is either a nonmembrane protein or present in the membranes in very low concentration.
The compact DNA structure from rat liver mitochondria described here contrasts markedly to the chromosome-like structure isolated by Pinon et al. (17) from X. Zaevis oocyte mitochondria.
Although these workers used the same lysing detergents, the lysing medium contained 0.15 M NaCl and the polyamine spermidine. Under those conditions a rapidly sedimenting form of the mtDNA was isolated. This structure was seen in the electron microscope as a 1.54~pm open circle having no apparent central dense core and exhibiting a thin fibril studded with globules presumed to be proteins. Their estimates of the protein to DNA ratio were about 3 times higher than those calculated here for >39 S mtDNA, and their structure was sensitive to treatment with 1% Sarkosyl. The reason for these major differences between the two similarly isolated structures is not yet clear and must await further studies.