DRWITT8 . GOODMAN
3592
Vol. 80
TABLE I1 solvent is heptane saturated with water, coniREPORTEDVALUES OF THE DIMERIZATIOS( A s s o c r a ~ ~ o s )1)ared to anhydrous heptane ( d e supra). -Us0 COSSTAXTSOF FATTXACIDS IS A s r r ~ n ~ o rORGANIC -s SOL- relevant is the fact that the constants reported by Pohl, Hobbs and Gross for heptane are adiiiittetlly VESTS, 30"" Acid
Solvent
17 ,l h
Formic Benzene 2 . 3 x 1112 Acetic Benzene 3 . i X IO' Propionic nenzcllc 2 . 8 x 10: Butyric Benzeiie 4,' x;l o ? Stearic Bcnzcuc .i 2 x t i l 2 Formic Heptaiie 2 x 104 Acetic Heptaiie -1 x tii' Data of Pohl, Hobbs :tiid Gross"1O and Mnryol t , H o h h and Gross." * Calculated from their reportetl tiissociation constants, after correcting their constants (given in units of moles/inolar volutne of solvent) to iunits of iiiolc
should also be noted that the estiniatetl \ d u e s o f k d for formic and acetic acids in anhydrous heptane (Table 11) are greater than the values obtained iii this study (Table I).:i2 Part of this difference may reside in the fact that in thc present studies the (32) Reasoning frum the vnlilrs Tor /;,I fur beozellc given i l l ' ~ x l ~ l11 c ~ one would expect I z , ~i n heptane for t h e acids 1on-f.r t h a n uccLir those in T:cble 1) t o l>e &:renter iliaii /:,I f o r :icetic, :iiiil f , > r i i i i heptane.
only estimates, since they were unable to make sufficiently precise measurements in solutions of sufficiently low concentration in heptane. I n conclusion, the magnitude of the values of k,l observed in these studies dcscrves re-emphasis. 111 heptane solution diinerizatioii is significant a t concentrations well below l o p 4 iiiolar, ant1 the dimer is the stable and predominant form o w r a wide range of concentration. The studies of Pauling and Rrockway??and others have shown that thc configuration of such dimers involves a syinirietrical coplanar structure of the two carlxjuyl groups, connected by two hydrogen bonds. Tlic magnitude of k d herein observed is :I testimony to the qrc:it stability ol this structure. Acknowledgment,-I would like to th.inl, 1)r. Iiobert S.Gordon, Jr,, for a great ( l e d of .ialuahle It has long been known that serum :dbuiriin interacts strongly with fatty acid anions. I n 194.1 Kendall found that crystalline albumin was alwnys associated with a sinall amount of free fatty acid that could not be reimn-ed by repeated crystallization.' Since then the nature and effccts of this interaction have been the subjects oi iitiiiicwus investigations. Thus, fatty acid anions 1i:~velieen shown to stabilize serum albumin against tlenaturaand to conil)ei.e tion by urea, guanidinc and effectively with organic dyes for bintliiig sites oil the albumin molecule. ' Electrophoretic studies have demolistrated x t i increase in the iiioliility ill' (1) F. E. Kendali, J . Bioi. C h C m, 1 3 8 , 97 ( l U 4 l i . ( 2 ) P. D. Royer, I;. G. Lum. G A . B ~ L I I o IJI ,. A I , T.nck :md I < , (; Rice, ibid., 162, 151 (l94ii). (3) P . D . BuS-er, G A . B a l l ~ t ~ a nJi l R1. I.uck, ibi.1.. 162, l$lil ( l ! i $ l \ l , (4) G . E . Cogin and I3. 13. Davis. 'THIS J O I : K K A I . , 7 3 , 313: I l O i l ) . ( 6 ) U . Westphal, J . 77 S t e t a ~ n Sd C P r i m , A r i h . n r n < i t # m /e, the fit for higher values of G would be somewhat better if 12 = 6 was used for myrlstate a n d laurate, a n d nt = 4 for stearate a n d palmitate.
__.
v
1;ig. ,4,
were turbid ;titer equilibratioii with coiiceiitratioris highcr than those plotted in Figs. 2 and 5 . Finally, the values of ka’ are highly inaccurate and shoul~lbe considered merely as approximations. 35,36 Relation of Apparent Constants to Intrinsic Constants.-The apparent association constants herein determined differ from the intrinsic association constants for the interaction of serum albumin with fatty acid anions in several ways. In the first place, these experiments were conducted in phosphate buffer, and the possibility (35) T h e only value of ks’ which has a n y accuracy is t h a t for liaoleate, since this was the only ion where values of V greater t h a n 10 were observed (up to 13.5). (36) In addition, the possibility t h a t there is a fourth larger class of still weaker binding sites, consisting of the remaining cationic residues on the albumin molecule, should be mentioned. These sites, if present will never be detectable for t h e long-chain f a t t y acid anions because of the limited solubility of these materials.
DEWITTS. GOODMAN
3896
Vol. 80
1
0
140
75
A---A
Stearate
O---O
Palmitate
I
i
li
50 1
-
-
v
V. Fig. 6.
Fig. 5.
exists that phosphate ions can compete with fatty acid ions for the same binding sites. Scatchard has shown31that when several small ions (A,, B, D, . . . etc.) compete for binding sites on a protein, the
the experiniental technique employed iii these studies requires the presence of a buffer, for the values of the partition ratio are highly sensitive to pH changes. A second difference between the apparent and the TABLEI intrinsic association constants derives from the fact THEAPPARENT ASSOCIATION CONSTANTS FOR THE INTERAC- that the fatty acid anions probably bind to cationic TION OF HUMAN SERUMALBUMIN WITH FATTY ACIDANIONS sites on the albumin molecule and that the availAT pH 7.45, IONICSTRENGTH 0.160, 23', WITH THREE ability of such cationic sites will depend on their SITES,nl = 2, n2 = 5 AKD n3 = 20 equilibrium with hydrogen ions. Saroff CLASSESOF BINDING has Fatty acid anion ki ' k2' k3' shown that when this is the case the apparent as6 X lo2 Laurate 1.6 x io6 2 . 4 x 105 sociation constant should be corrected by the exMyristatc 4 . 0 X lo6 1 . 4 X lo6 2 x 102 pression 6.0 X 8.0 x 1.1 x 1.3 X
Palmitate Stearate Oleate Linoleate
lo7 107 108
lo7
3 . 0 X lo6 8 . 0 x 105 4 . 0 x 106 2 . 5 X lo6
1x 1 x 1 x 2.5 X
103 103 103 lo3
apparent association constant k'Ai is related to the association constant in the absence of such competition kAi by the equation k'Ai
k2\i
1 f ksiCB
+
kDiCD
+ ...
(2)
The difference between k'Ai and kAi might be quite large if competition actually occurred in thc present studies. Thus if one assumed that phosphate ions compete with fatty acid ions and that phosphate binding is similar in magnitude to chloride binding the application of equation 2 would indicate that k1 is roughly 100 X kl' and kz roughly 5 X k ~ ' . ~ ?Whether such competition occurs is, of course, not known, and no attempt has therefore been made t o correct the apparent association constants (Table I) for competition with buffer ions. It should be pointed out parenthetically that (37) This rough estimate neglects consideration of electrostatic effects, in using Scatchard's intrinsic constants for chloride binding,al and should be considered only as indicating the possible order of magnitude of the effects involved. If electrostatic effects are considered the correction factors will be smaller.
(3)
where kAi is the corrected apparent association constant, kHi is the association constant for E%+ binding to the group to which A also binds, and CH is the concentration of hydrogen ions. Equation 3 is of such form that this correction factor is significant only when kHiGH is very small (ie., is one or less). Thus, for these studies at pH 7.45, if the cationic sites to which fatty acid ions bind are the €-amino groups of lysine (PKOH9.S)39 this correction factor is not significant (less than 1%). If, however, the important binding sites are the ionized imidazolium groups of histidine (PK'H 6.9) the correction factor would be significant, for in this case kAi would be of the order of 4 X k'Ai, and, in fact, the values of n might be sensitive to pH changes. Since the nature of the sites to which fatty acid ions bind is not known, whether or not this correction is applicable to the present studies cannot be decided a t this time. A final difference between the apparent and intrinsic association constants arises from electrostatic interactions between the fatty acid anions and (38) H. A. Saroff, J . P h y s . Chem., 61, 1364 (1957). (39) C.Tanford, S. A. Swanson and W. S.Shore, THIS JOURNIL,77, 6414 (1955).
Aug. 5, 1958
INTERACTION OF HUMAN SERUM ALBUMIN WITH FATTY ACIDANIONS
3897
is only part of the explanation, however, for Scatchard, et al., using resin deionized bovine mercaptalbumin (which contained slightly more than one mole of long-chain fatty acid per mole of k k , = kOA,e-22LZ+h (4) albumin)1gfound that a number of small anions where k.il is the association constant corrected for (Cl-, I-, SCN and CClzCOO-) bound to a single effects of competition, et al. (discussed above), site much more tightly than to the rest and have reported the intrinsic association constant for koA, is the intrinsic association constant, zp is the net charge of the protein, ZA is the charge of A and trichloroacetate to this first binding site to be w is the usual Debye-Huckel parameter. This 4.6 X 104.12 The values of k’I observed in the theory recently has been modified by Tanford and present study are two or three orders of magnitude K i r k ~ o o dusing ~ ~ a model in which the charges on greater than this, and the values of the intrinsic the protein are taken to be discrete unit charges association constants would be even greater (see located a t fixed positions and by- S a r ~ f fusing ~ ~ a above discussion). It is hence apparent that the model in which hydrogen bonding is assumed to non-polar side chains of the fatty acids herein occur between carboxylate and cationic nitrogen studied participate significantly in the binding groups on the albumin molecule. Electrostatic process. That the non-polar portions of an ion corrections have not been applied t o the results of contribute t o its interaction with serum albumin the present studies for two reasons. First, in has, of course, been known for some time. Thus order to apply the proper correction one would have Teresi and Luckgfound that binding increased with to know whether fatty acid ions compete with chain length for the fatty acid anions acetate phosphate ions and, in fact, whether they compete through caprylate, and Karush and Sonenberg4? with HP04= or H2P04- ions. If no competition reported that the binding of alkyl sulfates was inoccurs, the binding of a fatty acid ion would change creased by an increase in the size of the non-polar zp by -1; if, however, a fatty acid ion displaces a part of the molecule. This effect of the non-polar H z P 0 4 - ion, this would result in no net change of portion of an ion usually has been ascribed to a z,, whereas if it displaces a HPO,= ion zp would be summation of a number of short range non-specific changed by + l . Second, in order to estimate zp van der Waals interactions between the non-polar correctly the binding of phosphate ions under these portions of the binding ion and the non-polar side conditions would have to be known. The proper chains of the albumin molecule. The fact that application of electrostatic corrections must hence the values of k’l, for the honlologous saturated fatty await further knowledge about phosphate binding acids laurate through stearate, increase with increasing chain length is qualitatively consistent and about its competition with fatty acid binding. In conclusion it should be restated that the as- with this explanation. Closer inspection of thesociation constants listed in Table I are the ap- actual values of k’l listed in Table I, however, sugparent constants for the conditions employed in gests that a great deal of structural specificity is these studies and might be quite different from the involved. This is, in the first place, suggested by intrinsic constants for fatty acid ion binding. I t the magnitude of the values of k ’ l . More convincshould also be noted that the correction factors are ing, however, is the observation of the very difsuch that the intrinsic constants will certainly be ferent effect on k’l of increasing the length of each larger than the apparent constants. Finally, it fatty acid by two methylene units. Thus k’l for should be noted that thc conditions of these studies the 14Cacid (myristate) is 2.5 X k ’ for ~ the 12C acid (GH 7.43, ionic strength 0.16) are almost the same (laurate), whereas k’1 for the 1GC acid (palmitate) as those which obtain in blood plasma of mammal- is 15 X k1 for the 14C acid, and k’l for the 18C acid ian organisms41and that the apparent association (stearate) isonly 1.3 X k’lfor the 1GCacid. It would constants can therefore be applied to in ‘~livostudics hence appear that the two binding sites in the first class are rather specifically constructed so as to be involving fatty acid-serum albumin interactions. Relation of Fatty Acid Structure to Binding Con- able to bind a 16C or an 18C fatty acid much more stants.-The most striking result of these studies tightly than one of chain length snialler than 16C. is the observation of the extremely high affinity Furthermore, adding one double bond to thc 1SC of the serum albumin molecule for the binding of acid ( k ,oleate) results in an increase in k’l of up to two long-chain fatty acid anions. The values almost 407,, whereas adding a second double bond of k’l listed in Table I are considerably larger than (linoleate) decreases k’, by a factor of almost 10 the values of the association constants previously (as cf. oleate). 1Vhether the oleate ion contains reported for the interaction of serum albumin with the optimal structure for interacting with the any sinall organic or inorganic ions. Part of this first class of binding sites cannot be decided without difference lies in the fdct that almost all previously studying other structures (particularly fatty acids reported binding studies were performed with longer than ISC), but with the information availalbumin preparations which had not been specially able one could reasonably speculate that this might treated to remove tightly bound fatty acid. Such indeed be the case. I n contrast to k’l the values of k’z listed in Table I albumin preparatioiis undoubtedly Contained some tightly bound fatty acid. and it might well be that show a much smaller variation for the six fatty the first two sites were i l l this way obscured. This acids studied ’rhus the values of k’z for palmitate, oleate and linoleate are all very close to each other, (-10) C l a n f o r d d n d J G hirkiruod ] H I S J U U R V A L 79,2333 (1’357) (11) Tl’ith the exception, t h a t c h l u r d e and hicdrhouate, nut phosand, in fact, the differences between these three
the albumin molecules. Using a simple spherical protein model with a uniform distribution of charge on its surface, Scatchard30has shown that
phdte are the major anion5 of pldsmd dud that
37’ not 2 3 O .
til
azuo temperature
IS
(42) F. Karush a n d M Soneuberg, THISJ O U R N A L , 71, 13G9 (1919)
constants are within the estimated error in the determiiiations of k'2. k'? for myristate is approximately half this value, and k'z for laurate is approximately one-tenth this value. .-llso striking is the observation that k':! for stearate is less than k'? for palmitate. The estimated error in k r qis greatest for stearate, but it would appear that this difference is well outside this error. These results suggest that the second class of binding sites contains less structural specificity than the first class and that the structural specificity of this second class is different from that of the first class. No comments can be made about the possible structural specificity of the third, weakest class of binding sites because of the large inaccuracy in the estimations of k'a. Finally, brief mention should be made of the question of whether the same sites which bind fatty acid anions are those which participate in the binding of a great many other anions to serum albumin. Scatchard, et ~ l . , ~found ' that the binding of chloride, iodide, thiocyanate and trichloroacetate ions could be expressed in terms of three classes of binding sites, with ?zl = 1, n:! = 8 and n3 = 20; they also reported that their results could be fitted, though not quite so well, by .iz1 = 2, n2 = 8 and n3 = 1s. The similarity between these values of 721 and 112 and those observed i n the present study is quite suggestive, although any comparison is limited by differences in the protein preparations employed (see above). In addition, the studies of Karush on the competitive bindir1.g of p-(2-hyciroxy-S-n?eth~I~~~ienq.iazo)-ben dodecyl sulfate ions to crystallized boi-ine seruiri albumin (in phosphate buffer !>I3 7.01, showed that one site, or possibly two, bouricl the latter tightly but not the former and that there were trjjo other groups, one with -1 or 3 sites and one with about 17 sites, which bound both ions. 32,.13 Unpublished studies from this also have showr~ that methyl orange anions bind t o two groups of sites on fatty-acid-free huinan seruin albutniii (in phosphate buffer pH 7.45) :t-ith 721 = I . 1 2 . =: I S , k r l = 1.3 X 1oj and k'z = 2.2 X 10:. The binding of methyl orange anions YRS not altered by the addition of 1.8 moles of loiig-chain fatty w i d atlions to the albumin, indicatitig that the first two sitcs which bind fatty acid ions so tightly itre i l o t x v : t i l able for the binding of iii~tliyloraii;re :inions. One iiiay heiice reasoti:tbly r;i".cul:ttc that t!ic sccolld atid third classes o f sitcs which b i i i t i ia are also those to which : i i n t i \ ~ 1jthc.ra i i i c r i but that because o f the gre:ttcr structur: of the first class of sitcs, i t is un:tv;iil:il)lc for tile ( 4 8 ) F. Kariish. 'frirs J o r r n s . & Wi) D. S Goo
binding of relatively bulky ions like the dye ions. The first class of sites, however, might well be available for the binding of small simple ions (e.g., C1-, I-, SCN-)and ions of structure similar to that of long-chain fatty acids (e.g., dodecyl sulfate ions). Implications for Meta.bolic Studies.-Recent studies of the unesterified fatty acids present in human blood plasma have indicated that these fatty acids are of great metabolic significance, in representing a transport form of lipid readily available as substrate for energy production. This conclusion is derived from studies of the arteriovenous differences, i 2 the rapid turnover rates46and the responsiveness of the unesterified fatty acid fraction to changes of dietary status and hormonal administration. l 1 - I 3 The normal level of these fa.tty acids in humans is of the order of 5 X 10-1 mole per liter of plasma.11m13This level represents a i; of fatty acid t o albumin of slightly less than 1 . Preliminary studies of the composition of this fatty acid fraction indicate that i t consists of approximately SO%', of oleic, palmitic, stearic and linoleic acids (in decreasing amounts) , 5-1O% palinitoleic acid and an isomer of oleic acid, 5-107, pdyethenoid 20C acids and .?yoof 11C to 1OC acids. ifi Eighi-y-five to ninety per cent, of these acids are hence represented by the acids whose bindi:lg to albumin has been herein studied. The values oi' kl' are such that for F of 1 the concentration o l free (unbound) fatty acid anion is approximately 10 -9 inole/hter. In normal human plaslna, thereFore, less than 0.017, of the unesterified fatty acids present are free in solution.47 I t thus appears tllat albumin is particularly well constructed to serve as a transport vehicle for fatty acids and that the albuuiiri-fatty acid coinplex inust be thought ol :.is an entity with comiderable biological significance. Acknowledgment.-I wish to thank Dr. Robert S.Gordon, Jr. for ;L great tlcal of valunl)lc ntlvica iiIlf.1 assista11cc. 1'> F.. L H I : s D A , LID.
