Professional Documents
Culture Documents
artifacts (19-22) and to be incompatible with the other parameters in the enzyme kinetic expressions, namely, the rates of the reactions, metabolite concentrations, and kinetic constants. The values to replace Vmax were therefore calculated for each equation as the only unknown, assuming the other parameters to be correct. These calculated values are called Vvivo. Reaction rates in vivo were determined in a number of ways and were very reliable. The ratesof individual reactions were based on the rateof accumulation of counts in a product knowing the specific radioactivity of the immediate precursor (5, 23, 24). These values were substantiated using a specific radioactivity curve-matching program (TFLUX (25)) to simulate data in which the specific radioactivity of many intermediates in the network was followed as a function of time after exposure to tracer (13, 14) or perturbing (2) levels of [4C]glucose. Asthese analysesinvolved many interdependent metabolites and reactions, they also provided information An intriguing question for biochemists is the extent to regarding compartmented pools. For example, a difference in which enzyme activities, metabolite concentrations, kinetic mechanisms and constantsdetermined i n vitro apply in vivo. the labeling patterns of UDP-glucose and glucose 1-phosphate One approach to this problem is to construct metabolic models required the existence of two pools of glucose 1-phosphate (2) already been established that UDP-glucose was a single that incorporate such in vitro data and then to determine how (it had well the behavior of the model describes metabolism in the pool (14)). Other studies directly demonstrated the location living organism. This in turn can only be judged by the of the two pools of hexose phosphates (13). Knowing total predictive value of the model. Making a model match a given substrate concentrations, these analyses gave the concentradata set does not necessarily indicate that it reflects reality; tions of the metabolite compartments. As most metabolites the model must be sufficiently realistic to have predictive are relatively stable, their total concentration can be detervalue. When predictive value has been demonstrated, it is mined quite accurately. In comparing metabolite and enzyme reasonable to conclude that the model parameters may ap- concentrations in seven different organisms, metabolite concentrations were found to be more consistent, as might be proximate these operative in vivo. values are also relatively consistent In 1967 a simple metabolic model was constructed simulat- expected (26). K,,, and Ki ing the glycogen cycle and saccharide end-product synthesis and areoften similar for a given enzyme isolated from differin Dictyostelium discoideum. Five assumptions (predictions) ent organisms. For example, for the pyruvate dehydrogenase in thismodel were later demonstrated to be correct (1). Over complex in Dictyostelium (27), Ascaris (28), and cauliflower the years, about 40 predictions based on increasingly complex (29), respectively, the K,,, values (mM) for pyruvate were 0.14, models of metabolism in thissystem have been substantiated, 0.18, and 0.20; for CoA,0.008,0.005, and 0.007; for NAD, three unknowingly in otherlaboratories (1-17). The program 0.11, 0.01, and 0.12; the Ki values (mM) for NADH were 0.049, used in the construction of these realistic models is called 0.025, and 0.034; for CoA, 0.024, 0.062, and 0.013. For malate METASIM (18),which contains 27 different enzyme mecha- dehydrogenase in Dictyostelium (30), pig heart (31), mitonisms; the user indicates the appropriate onefor each reaction chondrial bovine heart (32), soluble bovine heart (32), and and supplies the kinetic constants. Other input data are the Bacillus sdtilis (32), respectively, the K , values (mM) for flux relationships and compartmented metabolite concentra- malate were 1.33,0.8,0.99,0.54, and 0.90; for NAD, 0.10,0.20, tions in the metabolic network. Most of the enzymes depicted 0.54, 0.20, and 0.4; for oxaloacetate 0.27, 0.04, 0.04, 0.05, and in the models were isolated from Dictyostelium, purified and 0.06; for NADH, 0.04, 0.02, 0.02, 0.04, and 0.03. The relative characterized using standard in vitro methods. consistency of metabolite concentrations and kinetic conIn these models with demonstrated predictive value, Vmax stants among different organisms suggested that these values values could not be used, as theywere frequently found to be were reliable as input datafor metabolic models. In contrast toreaction rates, kinetic constants, metabolite * This work wassupported by National Institutes of Health Public concentrations and compartments, there are a number of Health Service Grant AG03884. The costs of publication of this article values to enzyme were defrayed in part by the payment of page charges. This article reasons to question the relevance of Vmax must therefore be hereby marked advertisement in accordance with activity in intact cells. Cellular organization is destroyed in the preparation of extracts, membrane-bound enzyme com18 U.S.C. Section 1734 solely to indicate this fact.
A steady-state model of the tricarboxylic acid cycle w a s constructed using a dynamic systems analysis computer program, METASIM. The model was based on radioactive tracer analyses which provided flux relationshipsandcompartmentedmetaboliteconcentrations. Ten of the enzymes modeled were purified and Dictyostelium discoideum. Alcharacterized from though experimentally determined enzyme mechanisms and constantswere used in the model, Vmax values were found to be unreliable, i.e. they did not reflect enzyme activity in vivo.This value was therefore calculated as the only unknown in each enzyme kinetic equation and called Vvivo,to distinguish it from Vmax determined in vitro.
3101
3102
FIG. 1. A steady-state model of the citric acidcyclein D. discoideum. Encircled reactions are those for which the enzyme from D. discoideum has been kinetically characterized. All metabolite pools are intramitochondrial, as defined by Kelly et al. (37). See Experimental Procedures for further details. Glu, glutamate; 2KG, 2-ketoglutarate; SUC, succinate; FUM, fumarate; MAL, malate; OAA, oxaloacetate; CZT, citrate; ASP, aspartate; ALA, alanine; PROT, protein; PYR, pyruvate; ACO, acetyl-coA.
TABLE I1 plexes are disrupted; enzymesites bouna by product or inhibA comparison of calculated enzyme activity in vivo, Vuiuo, and V,, itor in vivo (33, 34) may become freed and available to values for seven enzvmes of the tricarbonvlic acid cvcle substrate on dilution in extracts; unknown inhibitors or acEnzvme V. V; Ref. tivators may be diluted,destroyed, or created; proteolytic inactivation may be initiated by extract preparation (19, 20); mM/min mM/min andfinally,enzymeactivity is frequentlymeasuredunder 1.8 40 Isocitrate DHb 271.0 optimal rather than physiological conditions of pH, temper2-Ketoglutarate DC 4.8 41 7,608.0 196.0 Malate DH 77.8 30 ature, cofactor and protein concentration. As proteins, enMalic enzyme 8.6 39 3.08 zymesmayplayrolesother thanbeing catalysts. Proteins 45.0 Succinate DH 3.15 42 may be used as an energy source, especially under nutritional Citrate synthase 8.23 43 1,071.0 stress. Because of the general vulnerability of most proteins 2.1 27 Pvruvate DC 258.0 to proteolytic attack,excessive enzyme protein concentration Converted to m M mitochondrial volume, using the conversion may beessential to ensure adequate catalytic activity in times relationships: 1 mg of protein = 1.4 mg of dry weight; pmol/min/mg of stress. Catalytically active enzymes have also been found dry weight X 150 = m M packed cell volumein amoebae (44), and cell/ to serve structural roles (35). Regardless ofthe specific events mitochondrial volume is 1/5. DH, dehydrogenase. occurring in uiuo, treating VmaX as the unknown in each e DC, dehydrogenase complex. enzyme kinetic expression and calculating Vvivoresultedin models with predictive value. This approachwastherefore used to construct the model of the tricarboxylic acid cycle. to operate and maintain steady-state levels of cycle intermediates,
~ ~
EXPERIMENTAL PROCEDURES
In this model (Fig. l ) , flux through the cycle was consistent with
the flux of acetyl-coA into the cycle must equal that of oxaloacetate. Therefore a pathway is required which converts excess four- or five-carbon intermediates to acetyl-coA. The pathway for converting tricarboxylic acid cycle intermediates to acetyl-coA involves malic enzyme. In this reaction, malate is decarboxylated to yield pyruvate which is in turn decarboxylated to form acetyl-coA via the pyruvate dehydrogenase complex. This pathway has been shown to occur in Dictyostelium, and both of these enzymes have been purified and kinetically characterized (27, 39). We have purified and characterized seven of the nine enzymes essential to the cycle (Fig. 1 and Table I): 2-ketoglutarate dehydrogenase complex, succinate dehydrogenase, malate dehydrogenase, malic enzyme, pyruvate dehydrogenase complex, citrate synthase, and isocitrate dehydrogenase. These enzymes are listed in Table 2, together with their V,,, values and their calculated V,i, values. Two of the most striking discrepancies between the V,i, and VmaX values are for the enzyme complexes, 2-ketoglutarate dehydrogenase complex and pyruvate dehydrogenase complex. As these complexes are membrane bound, they dissociate during extract preparation, and higher Vvivo compared with VmaX values might have been anticipated. Most isolated preparations of succinate dehydrogenase containa deactivated form of the enzyme bound tightly to oxaloacetate (34), preventing succinate from binding. As purification procedures can
3103
cussion).
RESULTS
The METASIM modelshown in Fig. 1 was constructed based on an original TFLUX analysis of the tricarboxylic acid cycle in D. discoideum (36, 37). This model is highly constrained since many related intermediates were isolated and their interdependent specific radioactivities determined. All single poolsincluded in the original TFLUX modelwere assumed to exist entirely within the mitochondrion. For those metabolites that existed in two or more pools,such as malate, the pool external to the cycle was considered to be cytosolic and was eliminated from the present analyses. In these cases, the mitochondrial pool concentration predicted by the TFLUX model was used rather thantotal metabolite concentration. Theenzyme mechanisms and kinetic constants used in the model are given in Tables 3 and 4. To consider intramitochondrial concentrations and fluxes, mitochondrial volume wasassumed to be 20%of the totalcell volume (46). Cycle flux, determined from net protein degradation, or oxygen consumption (36, 37) was approximately 0.4 mM/min based on total cell volume. Therefore, based on mitochondrial volume, cycle flux is 2 mM/min. Amino acids serve asan energy source inthis organism (3), andthe constant influx of carbon is balanced by the efflux of carbon through the generation of COz. A number of structural constraints exist on the fluxes in a metabolic pathway in steady state when metabolite concentrations and fluxes are constant over time (3, 37). For each metabolite, the requirement that the input flux equals the output flux imposescertain constraints on the relative values of the fluxes (namely that the vector of steady-state fluxes must lie in the null space of the stoichiometric matrix of the metabolic system). In this case, there are 23 net fluxes (described by26 reactions) and 13 variable metabolites; rank analysis of the stoichiometry matrix confirms that there are no moiety conservation relationships among the metabolites, so the dimension of the null space is 10 (23 fluxes - 13 metabolites); that is, there are only 10 independent fluxes, and the remaining 13 can be determined from these. Six of the independent fluxes are the input reactions from protein degradation and were determined directly from experimental measurements (3). That the reaction pairs 9 and 10, 14 and 19, and 18 and 20 have zero net flux at the observed steady state accounts for a further three degrees of freedom. Thus, of the total 26 reactions, 13 can be specified. The 13 independent fluxes wereconsidered to be 9,10,14,16,18-26 (Fig. 1).Table 5 shows the net reaction rates used in the steadystate tricarboxylic acid model compared with similar rates from the TFLUXanalysis. The metabolite concentrations are shown in Table 6 and were the same as thevalues determined from the TFLUX analysis adjusted for mitochondrial volume (36, 37). Table 1gives the enzyme mechanisms and kinetic constants determined in vitro for enzymes purified from D. discoideum. These data were used in the model (Table 3), supplemented (as required by the mechanisms) bysome assumed values
whichcould not be obtained experimentally. A dead-end competitive inhibition mechanism (Table 4) rather than a rapid equilibrium-ordered BiBi mechanism (Table 1) was used for three enzymes (glutamate dehydrogenase, isocitrate dehydrogenase, and citrate synthase) to include product inhibitions which were physiologically significant (40, 43, 48). As discussed earlier, experimentally determined Vmax values were not used, but rather a calculated value called Vvivo (Vl, V2 in Table 3) was used for maximal enzymatic activity in these models. Flux and metabolite concentrations over the time course of the simulation are shown in Table 6. The concentrations of NAD, NADH, and CoAwere fixed time constants, taken from experimental measurements (45), and the fixed input rates (reactions 21-26)were based on the constant rate of net protein degradation and calculated rates of conversion of individual amino acids to specific cycle intermediates (3,37). Over a 10-min simulation period the flux and metabolite concentrations vary less than 0.04%, indicating steady-state conditions. The definition of steady state is discussed in the companion paper.
DISCUSSION
In the model of the tricarboxylic acid cycle, flux was based on 0 consumption, a TFLUXanalysis, the net rateof protein degradation, and the rate of conversion of amino acids to specific cycle intermediates. In terms of mitochondrial volume the rate of conversion of malate to pyruvate (reaction 6) was 1.09 mM/min, and therate of pyruvate formation from amino acids (reaction 25) was 0.45 mM/min, giving a combined rate of 1.54 mM/min for reaction 8, which is catalyzed by pyruvate dehydrogenase. This rate was also checked independently in vivo. The pyruvate dehydrogenase reaction in the slime mold, assayed in vitro using [l-4C]pyruvate and quantitating the evolution of 14C02, can be examined with virtually no interference from other enzymatic reactions (49). Pyruvate carboxylase, known to interfere with the pyruvate dehydrogenase assay (38), has not been detected in D.discoideum, and glycolysis has been shown to play only a minor role. Significantly, the pyruvate dehydrogenase reaction assayed with [l-14C]pyruvatein crude extracts was completely dependent upon NAD, coenzyme A, and thiamine pyrophosphate, suggesting that pyruvate was converted to 14C02only via the reaction catalyzed by pyruvate dehydrogenase (27). In all likelihood this is also true of the reaction measured in vivo. The in uivo rate of the reaction catalyzed by pyruvate dehydrogenase was measured by exposing cells to [l-C]alanine and relating the rate of 14C0, evolution to the specific radioactivity of the isolated [l-4C]pyruvate (49). The rate obtained, 1.65 mM/min, is in excellent agreement with the rate obtained using the othertypes of analyses summarized above. Inthe METASIM models of carbohydrate metabolism there is evidence that many of the K,,, values used do reflect K , values operative in vivo. Perturbation studies have been carried out in which Dictyostelium was exposedto glucose, Pi, uracil, or uridine, for example, and compared with the models perturbed by these same metabolites (1, 10, 11).These perturbations resulted in changes in the levels of other metabolites, such as glucose 6-phosphate, UDP-glucose, glycogen and trehalose, in boththe models and Dictyostelium. The similarity of these effects indicate similar substrate-K,,,relationships in themodels and the organism. For example, glucose perturbation results in approximate 3-fold increases in glucose 6phosphate and trehalose levels and less than a 2-fold increase in UDP-glucose levels (1, 11).The increased trehalose level results from the relationships between the K , values of tre-
3104
halose-6-phosphate synthase for glucose 6-phosphate and UDP-glucose and the concentrations of these substrates in the cell and in themodel. Perturbation studies will be carried out with the tricarboxYlic acid cyclemodel and the organism to test the many assumptions and predictions inherent in the model. Constraints on the model will also be examined, to determine if model behavior is affected by changes in enzyme mechanisms, configuration, kinetic constants, andso on.
Acknowledgments-We would like to thank Dr. David Fell for his considerations of our problems in meeting the requirements for the steady state in these models and kindly analyzing our pathway for us. We would also extend thanks to the organizers of the symposium on metabolic control held in El Ciocco, Italy, in 1989 for providing us with a forum for presenting these ideas and interacting with other researchers in the field. REFERENCES 1. Wright, B. E., and Kelly, P. J. (1981) Curr. Top. Cell. Regul. 1 9 , 103-158 2. Wright, B. E., and Reimers, J. M. (1988) J. Biol. Chem. 2 6 3 , 14906-14912 3. Wright, B. E., and Butler, M. H. (1987) in Evolution and Longeuity in Animals (Woodhead, A. D., and Thomson, K. H., eds) pp. 111-122, Plenum Publishing Corp., New York 4. Wright, B. E., and Park, D. J. M. (1975) J. Biol. Chem. 2 5 0 , 2219-2226 5. Sargent, D., and Wright, B. E. (1971) J. Biol. Chem. 246,53405344 6. Wilson, J. B., and Rutherford, C. L. (1978) J. Cell Physiol. 9 4 , 37-46 7. Wright, B. E. (1968) J. Cell Physiol. 7 2 , (Suppl. l ) , 145-160 8. Wright, B.E., and Dahlberg, D. (1967) Biochemistry 6 , 20742079 9. Wright, B. E., and Marshall, R. (1971) J. Biol. Chem. 246,53355339 10. Wright, B. E., Tai, A., and Killick, K. A. (1977) Eur. J. Biochem. 74,217-225 11. Wright, B. E., Tai, A., Killick, K. A., and Thomas, D. A. (1979) Arch. Biochem. Biophys. 192,489-499 12. Wright, B. E. (1984) J. Theor. Biol. 110, 445-460 13. Chiew, Y. Y., Reimers, J. M., and Wright, B. E. (1985) J. Biol. Chem. 260,15352-15331 14. Wright, B. E., Thomas, D. A., and Ingalls, D. A. (1982) J. Biol. Chem. 267,7587-7594 15. Emyanitoff, R. G., and Wright, B. E. (1979) J. Bucteriol. 140, 1008-1012 16. Dimond, R. L.,and Loomis,W.F. (1976) J. Biol. Chem. 2 6 1 , 2680-2687 17. Dimond, R. L., Farnsworth, P. A., and Loomis, W. F.(1976) Deu. Biol. 50,169-181 18. Park, D. J. M., and Wright, B. E. (1973) Comp. Prog. Biomed. 3 , 10-26 19. Wright, B. E. (1960) Proc. Natl. Acud. Sci. U. S. A. 4 6 , 798-803
3105
I.
THE BASIS FOR HOOEL CONSTRUCTION H. B u t l e r and K a t h y R . Albe TABLE 4 RATE' EQUATIONS USED I N TCA MIDEL
Barbara E. W i g h t ,M a r g a r e t TABLE 1 ENZYME MECHANISMS AND KINETIC CONSTANTS DETERWINED LH KLEQ ENZYME REACTION 1 2
K. (nM) GluDH
2KG 2-KGDC NAD CoA Glu-2.0 NAD = 0.2
K,
(mnl
--
--
50
Pong
41
U n i n o l e c u l a r Mars A c t i o n Rate K[A] Aspartate -,Oxaloacetate I (18): ENZYMES MODELED: C I t r a t e 1 -> I s o c ? t r a t e (12)': O x a l o a c e t a t e 1 -> A s p a r t a t e (20); A l a n i n e - > Pyruvate ( 7 ) : Glutamate ~> s u c c i n a t e 1 ( 1 4 ) ; Succinate 1 - > Glutamate ( 1 9 ) : O x a l o a c e t a t e 1 - > O x a l o a c e t a t e 2 (17):Aspartate -> O x a l o a c e t a t e2 ( I O ) ; O x a l o a c e t a t e2 - > A s p a r t a t e ( 9 ) . Dead-end C o r n p e t l t l v e l n h i b l t l o n P l n g Pong B i B i Rate VI[A][B)/(KB[A]+WI(l.O+[~NHl/K~)[B~+[A][B]) ENZYMES MODELED: Glutamate Dehydrogenase ( I ) ; C l t r s t e Synthase (11); I P O C i t l a t e Dehydrogenase (13). A l l o r t e n cM i c h a e l l r - M e n t e n Rate Vl[Al[E]/(WI+[Al)(KEPi[E]) ENZYMES MODELED: M a l i c Enzyme ( 6 ) .
3
5
SOH
Suc
MDH
--0.31
42
30
Iro Ordered BI B i
6 8
ME
PDC
39
21
--
M u l tP i ri in te g 0
u n iu o i Rate V I x V2([Al-[Pl/KEP)/(KP x V2 + V2[AI + VI[PI/KEP) Succinate Oehydrogenare (3); Fumarase ( 4 ) . ENZYHES MOOELED:
Pong w i t h
I1
13
CS
Equilibrium Ordered
CoA Bi = Bi 0.11
43 40
IracDH
lrac 0 13 NADH = 0 34
Asp ZKG
2KG
RE aq pm i dl l bO r lrudm ered BB i i
Ping Pong
81 B i
I5
AlpTA AlaTA
16Ala
P i n g Pang B1 81
Malate Dehydrogenase
(5)
n,
' O r l g i n l l e s t i m a t e of K . was 0.1 mfl Which was used i n t h e a n a l y z e d m a d e l . ' S e e a l s o : Komrnunlecki. P R ( 1 9 7 7 ) Ph.0. D l r r e r t a t i a n , Univelslty of MdslaChuIetts at Amherrt. LEGEND: A b b r e v i a t l a n r GluDH glutamate dehydrogenase, Z~KGDC: 2 - k e t o g l u t a r a t e dehydrogenase; ME: m 1 1 c enzyme; MOH: m a l a t e dehydrogenase complex; SOH: succinate dehydrogenase; PDC. pyruvate dehydrogenase complex; CS:cltrate Synthase; IrocOH: 1lOCltrdte dehydrogenase, AlaTA: a l a n ~ n e t r m s m l n a i e : ArpTA. a s p a r t a t e tranpamlnlse.
Rate = V I X vZ([A][B] - [PI[PI/KEP)/DENOM KB x V2[A1 + WI x VZ I B ) + V2[Al[B] + KP x VI[PI/KEQ + KP x where OENOM VI[PI/KEP + VI[PI[PI/KEP + KP x V I I A I [ P I / ( K I A x KEPI + WI x V2[BI[Pl/KIP ENZYHES MOOELED: Aspartate Transaminare ( I S ) ; Alanine Transaninare (16).
CI.
' Taken from Segel, I.H. (1975) Enzyme K l n e t i c r :B e h a v i o r and a n a l y s i s o f r a p r de q u i l i b r i u m and r t e a d y ~ r t a t eenzyme r y i t e m r . John Y l l e y 8 Sans, New York.
Flgure I .
NET REACTION RATES DETERMINED FROM TRACER EXPERIMENTS COHPAREO TO STEADY-STATE METASIH MODEL
TFLUX Model'
2.40 2.85 2.60 1.90 1.40 I.90
I4 I8
+ +
2.10
19 20 0 0
METASIM MOOEL InH/Mi"l 0.15 2.41 2.77 2.85 1.76 1.09 2.00 2.00 0 0
' H e t a b o l l t ep o o ld e r r g n a t l o n ra r eC o n l i f t e n tw t h Flgure 1. ' see reference 37; valuer bared on i S I v m e dm l t o c h o n d r l a l volume. TABLE 6: FLUX AND METABOLITE CONCENTRATIONS I N STEADY-STATE MODEL AT FIVE MINUTES
a 1 . (36, 3 7 ) far m l t o c h o n d r l a l