Proc. Nati. Acad. Sci. USA Vol. 75, No. 1, pp. 395-399, January 1978 Genetics Structural flexibility of isozyme variants: Genetic variants in Drosophila disguised by cofactor and subunit binding (electrophoresis/polymorphism/ dehydrogenase/NAD) GEORGE B. JOHNSON Department of Biology, Washington University, St. Louis, Missouri 63130 Communicated by Peter H. Raven, September 16, 1977

ABSTRACT Wild populations of Drosophila mojavensis cause any factor that tends to stabilize tertiary structure will exhibit considerable conformational variation in the NAD+-free act to obscure the variation from detection by electrophoresis. form of (alcohol:NAD+ oxidoreductase; Thus, that tightly bind cofactors or that aggregate into EC 1.1.1.1). The variation appears genetic, as it does not occur within an inbred strain. The NAD+-bound form of alcohol de- multimers may exhibit fewer conformational classes than might hydrogenase, present in the same individuals, does not exhibit have been detected in the absence of the stabilizing influences the variation, suggesting that the binding of NAD+ acts to sta- of cofactor or subunit binding. bilize conformation. Such cofactor binding to enzymes may thus This paper reports studies on the stabilizing influences of conceal considerable variation. A similar effect is suggested for cofactor binding and of subunit aggregation on shape binding of esterase subunits. in Drosophila. The results suggest that variation in In the decade since electrophoretic variation at enzyme loci was shape may indeed be masked by such stabilizing influences. first reported (1), extensive investigation of insect (2), plant (3), MATERIALS AND METHODS small vertebrate (4), and (5) populations has confirmed the generality of widespread polymorphism. Recent studies Stocks. Stocks of the cactus fly Drosophila mojavensis were suggest even greater levels of variation than were first detected provided by William Heed. Eighteen of the 20 lines investi- (6-8) and raise important questions concerning ultimate gated were established from single-pair matings of flies elec- amounts of variation and its biological significance. trophoretically homozygous for alcohol dehydrogenase (alco- One such question concerns the role of protein shape. In hol:NAD+ oxidoreductase; EC 1.1.1.1); two other lines of studies of the electrophoretic behavior of in gels of multiple founder stocks were >90% one allele. Sixteen of the various pore size, alleles were detected that appeared to differ lines were initially collected at Libertan and San Carlos, Sonora, only in shape, having identical charge (9); in subsequent studies, Mexico; the other four originated from Arizona. Two isolates most electrophoretically distinguishable alleles were shown to of each of the 10 Libertad lines were analyzed. differ in shape as well as in charge (7, 10). Control studies Stocks of D. pseudoobscura were provided by Richard Le- showed that known conformational changes in proteins may wontin, as part of a blind test of the ability of gel sieving analysis be easily detected on electrophoresis and that neither modified to detect genic variation at the esterase-5 locus (12). From 10 subunit exchange rates nor altered hydration shells could be to 20 flies were individually analyzed from each stock. invoked as alternative explanations to conformational differ- Electrophoresis. Individual flies were ground in 75 Al of 50 ence between variants. The shape of particular variants proved mM Tris-HCI buffer, pH 7.3, 5% wt/vol with respect to sucrose, highly reproducible: in analyses of highly inbred lines of Dro- and the homogenate was centrifuged briefly in a capillary sophila, between-individual estimates varied no more than centrifuge in the cold. Ten microliters of this sample was added replicate characterizations of the purified hemoglobin run in to each of six gels of different acrylamide concentrations (4, 5, the same gels. 6, 7, 8, and 9% acrylamide). Beef hemoglobin, horse ferritin, An unusual aspect of these studies of natural polymorphism and bromphenol blue were added to each gel to serve as internal in protein shapes is the anomalous behavior of heterozygotes: standards (13). Multiphasic disc gels were constructed and their low frequency suggests that some of the electrophoreti- Tris/glycine buffers prepared according to Ornstein (14). cally detected variants may reflect post-translational epigenetic Electrophoresis was at 10° and 200 V for an average of 140 phenomena (7, 11). When heterozygotes between differing min. variants of dimeric enzymes do occur, the characteristics of the Alcohol dehydrogenase was assayed in 0.1 M Tris-HCI buffer, hybrid dimer are of particular interest: where the variants differ pH 8.4, with as substrate (15). Assays were stopped in in shape, the hybrid dimer is rarely intermediate in shape, cold 25% propanol/10% acetic acid, and all gels were scanned usually assuming the conformation of one of the variants or, on a Gilford 240 spectrophotometer at 540 nm within 90 min. more rarely, a widely divergent shape (7). This suggests a rather Esterase was assayed by the procedures of Hubby and Lewontin unexpected flexibility in protein shape, one subunit's confor- (16), and all, gels were scanned as before. NAD+, when added mation being significantly altered by the binding of the to grinding and running buffers, was at a final concentration other. of 1 mM. Conformational.flexibility between variants of an enzyme Gel Sieving Analysis. Protein shape for an individual sample has important implications in terms of our ability to detect the was estimated by plotting the logarithm of band mobility rel- variants in surveys of enzyme polymorphism. This is true be- ative to bromphenol blue against % acrylamide and taking the slope of the resulting regression as an estimator of size/shape The costs of publication of this article were defrayed in part by the (7, 9, 12, 13). In these data, experimental error in slope was payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U. S. C. §1734 solely to indicate Abbreviation: AdHj, AdH2, etc., electrophoretic forms of alcohol de- this fact. hydrogenase. 395 Downloaded by guest on October 2, 2021 396 Genetics: Johnson Proc. Natl. Acad. Sci. USA 75 (1978)

Fe 0

Fe

Hb AdH

Hb AdH

Fe

Hb 0 AdH

FIG. 1. An experiment demonstrating the NAD+-mediated interconversion of AdH3 and AdH5. Electrophoresis was performed on 8% acrylamide disc gels at 100 in Tris/glycine buffer, pH 8.9, at 200 V. Gels were scanned on a Gilford 240 spectrophotometer at 540 nm. Bovine hemoglobin (Hb) and horse spleen ferritin (Fe) were used as internal standards in each gel. uniformally less than 5% of the slope value. Sampling error was of Ad-H activity indeed represent forms interconvertible by estimated by comparison with the hemoglobin standard run NAD+ was verified by four lines of investigation: in the same gels: the distribution for a homogeneous sample of (i) Experiments were done in which sample preparation and an enzyme should be no greater than for the hemoglobin electrophoresis were both carried out in the presence of 1 mM standard. Such comparisons were done by plotting an estimator NAD+, while in parallel analyses of the same lines no exogenous of charge ("mid-Y") on the vertical axis and the corresponding NAD+ was added. In the absence of exogenous NAD+ both slope (KR) of the horizontal axis [the procedure has been de- bands are obtained, but in its presence only the faster of the two scribed in detail (7)]. bands is observed. This suggests that the slower form corre- sponds to AdH5 and the faster one to AdH3. RESULTS (ii) Experiments were done in which endogenous NAD+ was removed by passage through a 5-cm Sephadex-G25 column. Alcohol dehydrogenase In no case was the fast band (AdH3) removed by this treatment, Alcohol dehydrogenase occurs in five interconvertible elec- but the proportion of total alcohol dehydrogenase activity at- trophoretic forms in D. melanogaster (17). The most slowly tributable to the slower band (AdH5) was more than doubled migrating form (AdH5) may be converted into faster forms by by the passage through Sephadex. NAD+ in a manner elegantly investigated by Jacobson et al. (iii) The interconversion of AdH3 and AdH5 by NAD+ was (18). In adult Drosophila only two forms are usually observed verified by performing serial cofactor addition and removal (AdH5 and-AdH3), with the fastest form, ADH1; present in trace experiments on extracts prepared from single individuals. A amounts and forms AdH2 and AdH4 undetectable. NAD+-free typical result is presented in Fig. 1. In all cases the results were AdH5 may be converted NAD+-bound AdH3 in vitro under fully consistant with NADW-mediated conversion. low concentrations of NAD+ (<10 mM); at higher concentra- (iv) In line with a penetrating suggestion by Ornstein (14) tions of NAD+ (>10 mM), significant conversion to AdH1 oc- and Cann and Goad (20), experiments were carried out in curs. Importantly, the NAD+-bound forms of AdH are reported which the sample concentration was varied. Ornstein pointed to be significantly more heat-stable than the NAD+-free form out an interesting property of "stacking" gels such as used in (18). Preliminary investigations have suggested that confor- these studies: in stacking gels band focusing is obtained by a mational change is associated with the binding of NAD+ local electrical field created by the ions and charged proteins (19). and ithe local strength of that field depends upon the amount When AdH was analyzed in D. mojavensis, two bands were of charged material. Because most reversible aggregations (such uniformly obtained which differed substantially in charge as alcohol dehydrogenase with NAD+) dissociate in strong (mid-Y of 0.20 compared to mid-Y of 0.09). That the two bands electric fields, the degree of dissociation (proportion of total Downloaded by guest on October 2, 2021 Genetics: Johnson Proc. Natl. Acad. Sci. USA 75 (1978) 397 Table 1. Effect of sample concentration upon relative proportions of AdH forms Sample concentration, Relative proportion ,41 of fast band (AdH3) 0.1 0.02 0.5 0.04 1.0 0.08 2.0 0.07 5.0 0.12 10.0 0.19 20.0 0.57 40.0 0.58 AdH activity was assayed on 8% acrylamide gels and each gel was scanned on a recording spectrophotometer. Peak areas were calculated as peak height X width at half peak height, and relative proportions -0.05 -0.06 -0.07 were estimated as fractions of total activity peak area. activity in AdH5 form) should be a function of the local field strength and thus of the sample (alcohol dehydrogenase) con- centration. Specifically, Ornstein predicts that increasing .6 0.40 AdH sample concentration will produce increasing proportions of the bound form (AdH3). This was tested in our studies by varying Drosophila alcohol dehydrogenase concentrations over 0.35I two orders of magnitude (Table 1). There is a clear effect of sample concentration upon the relative proportions of the two forms, just as predicted by the hypothesis that the two forms / NAD+-bound * form represent different states of dissociation. 0.30 - i The two bands of alcohol dehydrogenase activity observed in D. mojavensis thus seem quite clearly to be interconvertable by NAD+ and analogous to the AdH5 and AdH3 forms reported earlier. 0.25 The specificity of NAD+ in producing the conversion of AdH5 to AdH3 is not absolute. Jacobson and coworkers report 0 9 1 NAD+-free that acetone concentrations from 10 to 100 mM also promote 0.20 form this conversion, although lower concentrations have no effect. 3:,..... It seems unlikely that acetone plays any significant role in the experiments reported here, however, as exogenous acetone concentrations are uniformly low. The level of acetone con- 0.15 I0 < > taminating commercial NAD+ is about 30 mol/100 mol of NAD+, and the NAD+ concentrations used were 1 mM, so that exogenous acetone concentrations were of the order of 0.3 mM. -0.05 -0.06 -0.07 Endogenous acetone levels in the flies were not measured, but KR it is unlikely that the flies themselves could maintain a physi- FIG. 2. Gel sieving behavior of alcohol dehydrogenase from a ological acetone concentration of the order of 100 mM, as far highly inbred line ofD. melanogaster. The line was inbred by single lower levels would be highly toxic. sib matings for 456 generations (strain Y49. 455, kindly provided by The second line of investigation was to determine whether S. Barker). Each point represents an analysis of one individual (after ref. 12). The vertical axis is Rf at average pore size, mid- Y, and is a or not the binding of NAD+ produces an associated change in measure of charge; the horizontal axis is the retardation coefficient, conformation, as had been suggested earlier for alcohol dehy- KR, taken as the slope of a regression of Rf on pore size, and is a drogenase in D. melanogaster (19). Investigation of alcohol measure of shape. Hb indicates the values obtained for hemoglobin dehydrogenase in 20 individuals of a highly inbred line of D. internal standards run in the same gels. melanogaster produced very little evidence of conformational difference between the NAD+-free and the NAD+-bound NAD+-free form is highly variable. Note particularly the range forms (Fig. 2). Both exhibit similar values of KR. There is a slight Of KR values encompassed by the variation (-0.03 to -0.09 in indication of conformation (KR) heterogeneity within the Fig. 3) as compared to the inbred line (-0.05 to -0.06 in Fig. NAD+-free form, which might correspond to the electropho- 2). Yet every point in the highly variable distribution of the slow retically identical a and b forms described earlier (18). alcohol dehydrogenase of Fig. 3 represents an individual with The third and principle line of investigation was to determine a corresponding NAD+-bound alcohol dehydrogenase (fast) whether or not conformational variation occurs at the alcohol point determined from the same gels, and the distribution of dehydrogenase locus, as reported for many other insect loci (7), NAD+-bound points is not variable (it corresponds precisely and particularly whether or not conformational difference is to the internal standard). Clearly the binding of NAD+ removes altered by NAD+ binding. With this in mind, 30 different lines a significant variance in shape and charge. The implication is of D. mojavensis were individually investigated. The results that form AdH5 may assume a variety of conformations, and can be seen in Fig. 3. While the internal standard hemoglobin that the binding of NAD+ acts to stabilize the enzyme so that (Hb) and the NAD+-bound fast form of alcohol dehydrogenase these differences are no longer seen. The lack of variation in show no more variability than detected in the inbred line, the the inbred strain strongly suggests that the observed variation Downloaded by guest on October 2, 2021 398 Genetics: Johnson Proc. Natl. Acad. Sci. USA 75 (1978)

0.58 Table 2. Gel sieving behavior of alcohol dehydrogenase from D. melanogaster Hb Alcohol Range of variation in KR dehydrogenase variant Fast allele Slow allele 0.54 AdH3 (NAD+) 0.050-0.052 0.045-0.048 AdH5 (no NAD+) 0.045-0.057 0.033-0.045

Forty-three independent alcohol dehydrogenase isolates were ex- 0.22 AdH amined, each previously placed in a common genetic background (G. B. Johnson and D. Hartl, unpublished).

gaster originating from a single population were prepared in this manner by Dan Harti, who kindly provided them to me. 0.18 0 When examined, they revealed the same pattern of variation seen for D. mojavensis (Table 2). Thus the source of the genetic variation in alcohol dehydrogenase is on the same as the alcohol dehydrogenase structural . 0.14 Esterase-5 In assaying for esterase activity in D. pseudoobscura, one also .0~~~~~ obtains two bands in most individuals. Examination of molec- 0.10 ular weights by gel sieving (13) reveals that the universally occurring slow band exhibits a KR value that corresponds to precisely double the molecular weight of the often occurring faster second band (10). The inference is that the slow band is I 0.06 a dimer (heterozygotes show three bands) derived from the corresponding monomer, also active as the fast band. 0.12 0.16 0.04 0.08 The results of gel sieving analysis of esterase in 10 lines of D. -KR pseuboobscura are presented in Fig. 4. Two band systems are FIG. 3. Gel sieving behavior of alcohol dehydrogenase from D. observed, with KR values corresponding to monomer and mojavensis. Each point represents a different individual line of Drosophila. 0, NAD+-bound form; 0, NAD+-free form. The dashed line connects the three bands of an apparent heterozygote. Note that no heterozygote is detectable in the corresponding NAD+-bound band, and that the presumptive hybrid dimer, while intermediate in charge, appears to have a very different shape.

in AdH5 is genetic, rather than physiological. If the variation .~~ ~ ~ ~ ~ ~ reflected the fact that subunits not bound to cofactors are more susceptible to proteolysis or partial chain unfolding (22), then we~~~~ct the same variation should occur in the inbred strain. These results do not, however, rule out the hypothesis that the binding of NAD+ to alcohol dehydrogenase protects the inzyme from modification by a protease or other enzyme and that it is this second modifying locus that is genetically poly- morphic rather than the alcohol dehydrogenase locus. Such a hypothesis would necessitate that the protease in question be locus-specific [otherwise why do some loci show many such 010* variants and others few (7)?], and that it be polymorphic for alleles producing discreetly different modifications (each variant strain of Fig. 3 is homogeneous, differing from the others, rather than each containing all of the potential varia- tion). Such a polymorphic protease would be difficult to detect, as it would entail genetic variation specific to alcohol dehy- drogenase which produces specific phenotypic states in the C3<~~~~~~~~~3 enzyme. To determine certainly whether the phenotypic states of alcohol dehydrogenase reflect differences in the primary structure of the enzyme, as opposed to such second-site modi- fication, requires amino acid sequence data on each of the variants, a major experimental undertaking. This uncertainty, shared by most published allozyme survey data, may however be evaluated in a somewhat less rigorous manner by examining alcohol dehydrogenase variants in a common genetic back- KR ground. The alcohol dehydrogenase locus of a wild-caught fly FIG. 4. Gel sieving behavior of esterase-5 from 10 lines of D. niay be transferred by appropriate crosses into a standard lab- pseudoobscura. The vertical axes represent mid-Y, the horizontal oratory strain. Forty-three independent lines of D. melano- axes, KR, as in Fig. 2. Downloaded by guest on October 2, 2021 Genetics: Johnson Proc. Natl. Acad. Sci. USA 75 (1978) 399 dimer. The results for the dimer are quite uniform in all- 10 the disappearance of the differences in thermal stability. Thus, lines. The monomer, however, appears quite different from one this form of allelic variation, commonly studied in terms of an line to another! Indeed, in one line there appears to be hetero- operational definition of change in function, may be charac- genity within the line as well. These data thus also suggest that terizable structurally in a relatively simple and straightforward binding, in this case to another subunit, is associated with re- manner. moval of variance in shape. Studies have been initiated in our laboratory with purified alcohol dehydrogenase protein. This permits explicit testing of DISCUSSION several predictions of the hypothesis presented here. For alcohol dehydrogenase, reproducible conformational differences should The results presented here raise the interesting possibility that be observed between NAD+-free enzyme protein purified from naturally occurring variants may not only differ in shape, but different lines and these differences should subsequently dis- may also exhibit some flexibility in shape, so that shape may be appear upon the addition of NAD+ to the purified preparations. modified by molecular interactions. While often discussed in For esterase-5, the test is more critical, as there remains the very terms of minor allosteric adjustments, such variation has not real possibility that the monomeric protein represents a second, been considered in a genetic context. It has important conse- independently varying locus. This can only be tested by puri- quences. fying the monomers, verifying conformational differences (i) The data of Figs. 3 and 4 indicate that the ability to detect between lines, and looking to see if resulting dimers are ho- variation, particularly in shape, may depend rather importantly mogeneous. In parallel, one might dissociate purified dimers on how samples are treated and on the physiological state of the of similar shape in high salt to see if resulting monomers differ organism. Where binding of cofactors or subunits serves to between lines. stabilize conformation into a common form, variants may not be detected in the bound state; where binding serves to ac- 1. Lewontin, R. & Hubby, J. (1966) Genetics 54,595-609. centuate conformational difference, as in some heterozygote 2. Powell, J. (1975) Evol. Biol. 8, 79-119. hybrid dimers (7), then quite novel electrophoretic behavior 3. Gottlieb, L. (1977) Ann. Mo. Bot. Gard. 64, 161-185. In in Droso- 4. Selander, R. & Johnson, W. (1973) Ann. Rev. Ecol. Syst. 4, may be observed. alcohol dehydrogenase studies 75-91. phila, for example, the two bands are commonly treated as an 5. Harris, H. & Hopkinson, D. (1972) Ann. Hum. Genet. 36, annoying artifact and dealt with in genetic studies by the ad- 9-20. dition of exogenous NAD+ to grinding and electrophoresis 6. Bernstein, S., Throckmorton L. & Hubby, J. (1973) Proc. Natl. buffers, so that all alcohol dehydrogenase exists as the NAD+- Acad. Sci. USA 70,3928-3931. bound form. Fig. 3 illustrates how this procedure serves to 7. Johnson, G. (1977) Biochem. Genet. 15,665-693. conceal the very variation that is the subject of investigation. 8. Singh, R., Lewontin, R. & Felton, A. (1976) Genetics 84, 609- The physiological dependence of electrophoretic behavior 629. suggests that surveys that attempt to measure ultimate numbers 9. Johnson, G. (1976) Genetics 83,149-167. of variants, or to compare the degree of homolology or "genetic 10. Johnson, G. (1977) in Measuring Selection in Natural Popula- tions, eds. Christiansen, F., Fenchel, T. & Nielson, 0. (Springer, distance" between groups (22), should be conducted with great Berlin), pp. 223-244. care to experimental detail. 11. Watt, W. (1977) Genetics 87, 177-195. (ii) Because NAD+ commonly binds alcohol dehydrogenase 12. Johnson, G. (1977) Genetics 87, 139-157. prior to substrate binding, and NAD+-bound enzyme does not 13. Johnson, G. (1975) Biochem. Genet. 13,833-847. exhibit conformational variation, these results suggest that the 14. Ornstein, I. (1964) Ann. N.Y. Acad. Sc. 121, 321-349. alcohol dehydrogenase variants may not differ substantially in 15. Shaw, C. & Prasad, R. (1970) Biochem. Genet. 4, 297-320. their binding affinity for substrate, while they may exhibit quite 16. Hubby, J. & Lewontin, R. (1966) Genetics 54,577-594. marked differences in their affinity for NAD+. 17. Jacobson, K. (1968) Science 159, 324-325. (iii) The fact that NAD+-bound alcohol dehydrogenase is 18. Jacobson, K., Murphy, J., Knopp, J. & Ortiz, J. (1972) Arch. to more heat-stable than Biochem. Biophys. 149,22-35. reported be NAD+-free enzyme 19. Knopp, J. & Jacobson, K. (1972) Arch. Biochem. Biophys. 149, suggests that there may be a correspondence between thermal 36-41. stability variants detected at this locus in Drosophila (18) and 20. Cann, J. & Goad, W. (1968) Ann. N.Y. Acad. Sci. 151, 638- the conformational variants reported in this study. Preliminary 662. results indicate that corresponding differences in thermal sta- 21. Thatcher, D. (1977) Biochem. J., 163, 317-323. bility do exist, and that addition of exogenous NAD+ results in 22. Johnson, G. (1977) Annu. Rev. Ecol. Syst. 8, 309-342. Downloaded by guest on October 2, 2021