Avidin Binding of Radiolabeled Biotin Derivatives*

Three N-acyl derivatives of biotinylethylenediamine were prepared: I, biotinylamidoethyl-3-(3-['261]iodo-4-hydroxypheny1)propionamide; 11, biotinylamido- ethyl-[SH]acetamide; and 111, biotinylamidoethyl-3-(3,5-['a61]diiodo-4-hydroxyphenyl)propionamide. Each compound was combined with a large excess of avidin, yielding 1:l molar complexes. Aside from a small fraction of each complex that dissociated more rapidly, the dissociation half-lives of these complexes were: I, 41 days; 11, 4.4 days; and 111, 148 days. The iodo- (mono or di) hydroxyphenylpropionyl moieties of I and 111, therefore, contribute significantly to the binding strength of these compounds toward avidin. We also formed 4:l complexes of I, 11, and 111 with avidin (compound in excess), each of which exhibited biphasic dissociation, with initial half-lives of 4, 3.2, and 24 days, respectively. Thus, I or especially 111 potentially can be used as a sensitive tracer in quanti- tative studies with avidin.

The avidin-biotin system (Green, 1975) continues to emerge as a useful molecular binding tool in work with biological molecules, as has been reviewed (Wilchek & Bayer, 1984;Bayer & Wilchek, 1980). This system is attractive largely because avidin is a small stable protein that binds four biotin molecules, and biotin is somewhat hydrophilic and easily conjugated onto other molecules giving products that complex with avidin. The avidin-biotin system has been useful, for example, in both qualitative and quantitative studies of membrane receptors (Wilchek and Bayer, 1984) and in increasing the sensitivity of detecting DNA probes (Singer and Ward, 1982).
To facilitate some of the studies and applications of the avidin-biotin system, it would be useful to possess a radioactive derivative of biotin that could be detected with high sensitivity. An lZ5I-labeled Bolton-Hunter derivative is a logical choice due to the commercial availability of the lZ5Ilabeled Bolton-Hunter reagent and the high specific activity Here we present such a derivative of biotin, in which ethylenediamine is used as a spacer group bridging the biotin and 'Z51-labeled Bolton-Hunter moiety. Because the resulting biotin derivative was found to bind unusually strongly to * This investigation was supported by DARPA Contract N00014-844-0254 administered by the Office of Naval Research. This paper is Contribution 334 from the Barnett Institute of Chemical Analysis. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
$ Current address: E. I. duPont de Nemours & Co., New England Nuclear Research Products, Boston, MA.
3 To whom reprint requests should be addressed. of 1251. avidin, we also prepared and similarly tested corresponding [1251]diiodo- acetyl derivatives of biotin.
Chromatography-HPLC was done on a gradient system equipped with a variable wavelength detector. Gamma radiation from ' " I was monitored externally through Tygon tubing by a sodium iodide scintillation crystal and a ratemeter from Ludlum Instruments. TLC was done on a silica gel plate from Analtech. For detection of radioactivity after cutting, plastic-backed silica from Eastman Kodak was used. Solvent A was 1-butanokacetic acidwater (707:10, v/v/v), and solvent B was 1-butano1:acetic acid:water:@-mercaptoethanol (707:102.8, v/v/v/v). Biotin and its derivatives were specifically stained with p-(dimethy1amino)cinnamaldehyde as described (Mc-Cormick and Roth, 1970). Separation of free from bound radioactive biotin derivative in the avidin binding assays was accomplished by spotting 3-5 pl on an ITLC strip (1.2 x 9 cm from Gelman), allowing it to dry for 3-4 min, and then developing it in 0.15 M NaC1, ethanol (95:5, v/v).

210
Avidin Binding 211 consistently above 80%. This product was diluted to 50 pCi/ml with 50% ethanol containing 0.01 M @-mercaptoethanol. In the absence of the latter reagent, the product gradually oxidized to a and B sulfoxides? TLC was done on a 250-pm plate using solvent system B which contained @-mercaptoethanol. Initial radiochemical purity of the product was >98% (RF = 0.75) and was still >95% after 30 days of storage at either -10 or 24 "C. Assuming that this compound was both chemically and radiochemically pure as indicated by both HPLC and TLC, then its specific radioactivity was 2200 Ci/mmol, the same as the 1261-labeled Bolton-Hunter reagent. Biotinylurnidoethylacetamide (Nonradioactive ZZ)-Acetylation of biotinylethylenediamine was done by modifying the method for peptides (Riordan and Vallee, 1967). Biotinylethylenediamine HC1 (50 mg, 0.155 mmol) was dissolved in 5 ml of 0.5 M sodium acetate buffer and acetylated with acetic anhydride (158 mg, 1.55 mmol) overnight at 4 'C. The acetylated product was extracted with 1-butanol, dried by rotary evaporation, redissolved in 0.1% acetic acid, and purified by HPLC using a Rainin Dynamax C-18 column (21.4 mm X 25 cm) eluted at 5 ml/min. Injection was made into acetonitrile, 0.1% acetic acid in water (1090, v/v), and this mobile phase was immediately changed to 15235, v/v. The product eluted at 30 min and was lyophilized. It was a single spot by TLC (250 pm) with solvent A ( R F = 0.31) positive by iodine vapor and p-(dimethy1amino)cinnamaldehyde. m.p. = 218-219 "C. Mass spectrum, m/z = 328. NMR (CD,OD, DzO) 6 3.31-3.34 (4H, multiplet, -NHC,CH,N-), 1.98 (3H, singlet, -CH3).
Biotinylumidoethyl-PHIacetarnide (ZZ)-[3H]Acetic anhydride (1 mCi) was condensed at -78 "C in a breakseal tube. Biotinylethylenediamine HCl (644 pg, 2 pmol) in 0.64 ml of 0.5 M sodium acetate buffer was added, and the acetylation proceeded at ambient temperature for 30 min. Ethanol (5 ml) containing 0.05 M @-mercaptoethanol was added, and the reaction mixture was stored at -20 "C until purification. An aliquot (400 pl) was taken to dryness by vacuum and redissolved in 0.1% aqueous acetic acid. HPLC was done using a Waters C-18 column eluted with 0.1% aqueous acetic acidacetonitrile (9010, v/v) at 1 ml/min with detection at 214 nm. Fractions were collected, and aliquots were assayed for radioactivity. The product eluted at 12 min (26 pCi) and was diluted with 1.5 ml of 0.05 M 0mercaptoethanol in ethanol. It co-chromatographed with the corresponding nonradioactive compound (ZZ) on TLC using solvent B ( R F = 0.31). The specific radioactivity of the product was determined to be 24 Ci/mmol by the HABA avidin binding assay (Green, 1970).
After 3 weeks at -20 "C, the purity was still >95% by TLC.
Stoichiometry of Z Binding to Avidin-Into a series of polystyrene test tubes was added 400 pl of 0.05 M potassium phosphate, pH 7.4, 0.05% lysozyme, 0.15 M NaCl (buffer). Equal amounts of avidin were added to each tube. Since a quantitative end point was desired, the protein concentration of the avidin stock solution was measured by UV at 282 nM, t = 24,00O/subunit (Green, 1975), and the biotin binding site concentration was measured by the HABA assay (Green, 1970). Avidin/tube based on UV was 3.45 pmol, and biotin binding sites measured by the HABA assay were 3.40 pmol/tube. Increasing amounts of Z (0.5-30 pCi) were added. After 4 h, the percent bound of Z to avidin was determined by ITLC as described above.
Dissociation Kinetics for Z:Avidin as a 1:l Complex-Into duplicate 12 X 75-mm polystyrene test tubes were added 0.25 ml of 0.05 M potassium phosphate, pH 7.4, 0.15 M NaCl, 0.05% lysozyme, radioiodinated tracer (0.5 pCi, 0.23 pmol), and avidin (2.5 pg, 37 pmol). After incubation for 20 min, biotin (21 pg, 85 nmol) was added, and the temperature was maintained at either 4 or 20 "C. A temperaturecontrolled recirculating water bath (Haake) was used to maintain 20 'C. Free from bound tracer was determined at subsequent time intervals by spotting 4 p1 from each sample on an ITLC strip (1 X 6 cm) and developing in 0.15 M NaCl, ethanol (95:5, v/v). The avidin and any bound Z stayed at the origin while any free Z migrated to the solvent front. The strip was cut in half, and each section was counted in a y counter. Percent bound was calculated as (origin cpm/(origin + solvent front cpm)) X 100.
Dissociation Kinetics for Z:Avidin (4:l)"This was done as described for the 1:l complex with the following changes. An excess of Z (30 pCi, 27 pmol) in 126 pl of buffer (concentrated from the original storage solvent) was added to avidin (0.236 pg, 3.45 pmol of monomer) in 300 pl of buffer and incubated at 20 "C for 7 h. Excess biotin (22 pg, 90 nmol) was added at time zero.
Dissociation Kinetics for ZZ:Auidin (1:l)-ZZ (0.15 pCi, 6.25 nmol) was incubated with excess avidin (1.5 mg, 22 nmol) in 0.5 ml of 0.05 M potassium phosphate buffer, pH 7.4,0.15 M NaCl, 0.05% lysozyme, 0.05% sodium azide for 2 h at ambient temperature. Unlabeled biotin (1.5 mg, 6 pmol) was added, and the solution was kept at 4 or 20 'C. Percent bound was determined as above except 5 pl was spotted, and the solvent front section of the ITLC strip after cutting was soaked in 1.5 ml of ethanol for 1 h to elute the radioactivity. Liquid scintillation fluor (4.5 ml) was added, and the radioactivity was measured (counting efficiency for 3H was 31.9%). To determine the total counts spotted on the ITLC strip, a blank strip (no sample application) was prepared, cut, and soaked in 1.5 ml of ethanol. Five pl of ZZ was added (the same amount as spotted above), followed by the fluor and counting. Percent bound was calculated as (1 -(solvent front cpm/ total cpm)) X 100.
Dissociation Kinetics for ZZZ:Auidin (1:l)-Binding assays were set up as for Z with the following changes. Sodium azide (0.05%, w/v) was present in the buffer. IZZ (0.25 pCi, 0.014 pmol) and avidin (1.25 pg, 18.5 pmol) were incubated for 1 h at 24 "C. The temperature was then controlled at 4 or 20 "C, and biotin (21 pg, 86 mmol) was added Z.
at time zero. Separation of free from bound was done as described for Dissociation Kinetics for ZZZ:Avidin (4:l)"When 14.2 fmol of ZZZ was placed in a polystyrene test tube in the usual phosphate-salinelysozyme buffer, and avidin (14 fmol intended) was added, no binding was observed apparently due to adsorption losses of the avidin. More avidin was added (total 149 fmol intended) until 46 f 0.82% binding of ZZI was achieved. The exact amount of avidin in solution at that point was unknown. After incubating the samples at 24 "C for 20 h, 50 nmol of biotin were added to initiate dissociation, and percent bound was determined as above.
Denaturation of Auidin:III-The above 1:l avidin:III complex (100 pl) was combined with 200 pl of ethanokacetic acid (8020, v/v) and heated to 70-74 "C for 1 h. Percent binding was determined by ITLC as previously described.  for 1 h, and then excess biotin (3.4 pg, 14 nmol) was added. The temperature was kept at 20 'C, and separation of free from bound was done as described for ZZ.

RESULTS
The scheme that we followed to prepare three radiolabeled derivatives of biotin is shown in Fig. 1 To determine the dissociation rate for I complexed to avidin, a large excess of biotin was added (2000-fold molar excess over avidin binding sites) at time zero, and separation of free from avidin-bound I was done by ITLC as a function of time. Free I migrated with the solvent front, and bound I remained at the origin. This technique gave a rapid separation (3-4 min) with low blanks (1-2%; see Footnote a in Table I).
Originally the assay buffer contained bovine serum albumin as a protein carrier, but a poor precision prompted a switch to 0.05% lysozyme which overcame this problem.
The dissociation rate and corresponding half-life (tLh) of I complexed to avidin as a function of temperature is shown in Fig. 2 and in Table I. For a 1:l complex of I with avidin obtained by incubating I with a 160-fold molar excess of avidin (640 molar excess of avidin binding sites) the dissociation is basically monophasic at both 20 and 4 "C, pH 7.4. (There is a small fraction of more rapid or "anomalous" dissociation at the outset that is discussed below.) The halflives for the dissociations at these two temperatures are 41 and 380 days, respectively. At 20 "C, pH 8.5, the half-life is 18 days (Table I).
Biphasic dissociation is seen when a 4:l complex of I to avidin is formed (1.3 molar excess of I over avidin binding sites) at 20 "C, pH 7.4 (Fig. 2). The initial dissociation rate for the 4:l complex corresponds to a half-life of 4 days, which is 10 times faster than for the corresponding 1:l complex. The second slower rate matches that of the 1:l complex, and extrapolating this latter rate back to the ordinate for the 4:1 complex gives an intercept at 50% binding. This demonstrates that both the 1:l and 2 1 complexes of I to avidin share the same dissociation half-lives. Presumably the residual sites on avidin are fully occupied by biotin shortly after its addition in these experiments.
Compound 11-To better understand the contribution of the '251-labeled Bolton-Hunter group to the tight binding of I to avidin (in terms of dissociation), we prepared and similarly tested the corresponding r3H]acetyl compound, structure 1 1 in Fig. 1. The dissociation half-lives at pH 7.4 of 1 1 as a 1:l complex with avidin at 20 and 4 "C are 4.4 and 51 days, respectively, as shown in Fig. 3 and cited in Table I. Thus, 1 1 dissociates nearly 10 times faster than does I at both of the corresponding temperatures, demonstrating that the 4-hydroxy-3-iodobenzyl moiety of I contributes significantly to the binding of I to avidin. Two other conditions were evaluated as well, as indicated in Table I; the half-life was 3.2 days at 24 "C, pH 7.4, and 4.8 days at 20 "C, pH 8.5.
Shown as an inset in Fig. 3 is the initial dissociation of 1 1 from avidin when a 4:l complex is formed. The half-life of 3.2 days for the latter complex is only slightly shorter than that of the corresponding 1:l complex. This contrasts with the 10fold difference in the dissociation rates for the analogous complexes between avidin and I.  8 (26)' "The radiolabeled biotin derivative or biotin plus avidin were preincubated at room temperature from 0.3 to 2 h prior to setting the incubation temperature and adding excess biotin at time zero for the 1:l complexes where a large excess of avidin was present. For the 4 1 complexes, preincubation was overnight. Percent binding refers to B/ Bo X 100, where B is the fraction of total counts (origin and solvent front) found in the origin half of the ITLC strip, and Bo is the same counts when no biotin is added to initiate dissociation. In this calculation there was no correction for the blank in which 1-2% of the tracer is found in the origin half of the ITLC strip in the absence of avidin. All of the biotin derivatives gave an initial B/Bo X 100 = 99-100% (obtained immediately after addition of excess nonradioactive biotin).
bFor the 4 1 complexes, the biotin derivative is present in a significant molar excess (see "Experimental Procedures") over the four binding sites on avidin.
'For the 1:l complexes avidin is present in a significant molar excess (see "Experimental Procedures") over the amount of the biotin derivative.
Anomalous dissociation was calculated as qr where q = initial value of B/Bo X 100 (see Footnote a ) and r = the extrapolated value for BIB, X 100 on the y axis using the predominant slope (later part, least squares linear regression analysis of the subsequent data points) of the dissociation curve. e Dissociation half-life in days for the initial anomalous dissociation. substituted phenolic moiety of I, compound ZZZ was synthesized and tested. In this latter compound, the structure of which is shown in Fig. 1, a second iodine atom is present, replacing an ortho hydrogen on Z.
The dissociation of a 1:l complex of avidin and ZZZ is shown in Fig. 4. Since III was doubly labeled with lZ5I, it was necessary to account for the formation of radioactive impurities arising from decay catastrophe of IZI as a function of time. This was done both experimentally (TLC) and by cal- culation, giving, within experimental error, equivalent results as shown by the upper line and points in Fig. 4.
As seen in Fig. 4, and summarized in Table I, ZZZ complexed t o avidin at 20 "C, pH 7.4, dissociates with a half-life of 148 days after correcting for decay. Thus, ZZZ, in this sense, binds 3.6 times stronger to avidin than does Z. The half-life was unchanged at pH 8.5 (150 days) but decreased slightly at pH 5.0 (113 days).
At 4 "C the dissociation half-life of ZZZ was too slow to measure accurately, given the complication of decay catastrophe. Nevertheless, the relative slopes of the 4 "C dissociation line and the decay catastrophe line for ZZZ by least squares regression analysis suggest a dissociation half-life of 6.4 years. Assuming that dissociation half-lives for ZZZ at 20 and 4 "C have the same ratio as those of Z at these two temperatures, then the half-life for ZZZ at 4 "C is approximately 3.7 years.
We confirmed that neither ZZZ nor any of its radioactive decay products covalently bound to avidin during this experiment by extracting all of the counts from a 60-day-old complex of avidin and ZZI with ethanolic acetic acid.
We also measured the dissociation half-life of a 4:l complex of ZIZ and avidin. Once again only an estimate could be made because of experimental difficulties. The problem in this case was that the intense radioactivity and small amount of ZZZ limited the amount of avidin that could be used. As seen in the inset of Fig. 4, there is evidence for biphasic kinetics, and the initial shorter half-life after correcting for decay appears to be about 24 days.  Table I). Thus, there is an initial more rapid dissociation of some of the biotin derivative complexed to avidin when biotin is added.
Biotin-To provide a reference point for our dissociation half-life measurements, we measured the dissociation of a 1:l complex of avidin and biotin using [3H]biotin. As seen in Fig.  5 and listed in Table I, the half-life for this complex under our typical conditions (pH 7.4,20 "C) is 473 days. As with our biotin derivatives, anomalous dissociation (8.0% of the complex) is observed at the outset, although it is much slower (tH = 26 days) than for these derivatives.

DISCUSSION
The very slow dissociation of I complexed 1:l with avidin (t% = 6 weeks at 20 "C, 1.04 year at 4 "C) was not anticipated. Others generally have encountered much weaker binding of monosubstituted biotin derivatives to avidin in the presence of excess biotin. The longest half-life observed by Chignell et al. (1975) for a biotinyl-spin label conjugate (4-biotinamido-2,2,6,6-tetramethyl-l-piperidinyloxy) was 15.5 h at 25 "C. The 1:l complex of N".B'-biotinylamido-insulin with succinylavidin at 25 "C had a half-life of 2.6 h (Finn and Hofmann, 1985).
However, more in line with our result, [biocytin] 25ACTH1_26amide binds strongly to succinylavidin; the halflife is 20 days for dissociation of the 1:l complex at room temperature (Romovacek et al., 1983). This enhanced binding was attributed by the authors to the small size and flexibility of this biotinyl peptide. Supporting this concept, N"*B'-[6-(biotinylamido)hexanoyl]insulin, having a flexible alkyl spacer between the biotin and insulin, has a half-life at 25 "C of 76 days as a 1:l complex (Finn and Hofmann, 1985). Nevertheless, most of the other biotin conjugates cited above are also small and flexible, and yet their complexes with avidin dissociate rapidly. Thus, additional factors may play a role in the strong binding of [bio~ytin]~~ACTH~-~~amide to avidin (see below).
We postulated that I binds strongly to avidin not only because I avoids repulsive interactions with avidin that some other biotinylated substances have encountered, but because the Bolton-Hunter group (3-iodo-4-hydroxyphenylpropionyl) of I interacts favorably with avidin. In particular, it was attractive to consider a hydrophobic interaction between the latter group and avidin given the structural characteristics of this group. Chignell et al. (1975) proposed that the biotin sites on avidin are located in a hydrophobic cleft. This was based on their observation, using ESR, that nonpolar spin labels conjugated to biotin are highly immobilized when these conjugates are complexed with avidin. The secondary structure of avidin is suggestive of such a cleft (Honzatko and Williams, 1982).
Our observation that 11 binds to avidin with a half-life 10fold shorter than that of I confirms that the Bolton-Hunter group of I helps to anchor I to avidin. Whether or not this is due to hydrophobic binding is another matter, however. For example, the Bolton-Hunter group contains an ionizable phenolic OH that might form a hydrogen or salt bond with avidin. The pK, of this OH is anticipated to be near 8.8 (Rogoeczi, 1984;Mayberry et al. 1965). II binds just as strongly to avidin at pH 8.5 as at pH 7.4, showing that the overall structure of avidin in this complex is not perturbed by this change in pH. However, I dissociates twice as fast at the higher pH. Thus, the weaker binding to avidin of I at pH 8.5 versus 7.4 is probably due to a specific unfavorable interaction between these species that does not develop between 1 1 and avidin with this increase in pH. Apparently either the ionization of I or of a side chain on avidin near the Bolton-Hunter group of I is responsible.
To probe this in more detail, we also determined the binding strength of avidin for 111, a diiodo version of I anticipated to have a phenolic pK, near 6.8-7.1 (Rogoeczi, 1984;Mayberry et al., 1965). At pH 7.4 the binding of 111 to avidin (t, = 148 days) is not only 3.6-fold stronger than that of I but is unchanged when the pH is increased to 8.5. Decreasing the pH from 7.4 to 5.0, which passes through the phenolic pK, of 111, slightly weakens the binding of 111 to avidin (the tLh falls from 148 to 113 days). This latter result is probably due to a general effect of pH on the avidin binding site rather than protonation of 111, since avidin also binds biotin 2-fold weaker at pH 5 uersus 7 (Green, 1975).
The simplest explanation for the differences in the dissociation half-lives for the binding of I, 1 1 , and 1 1 1 to avidin, including effects of pH, is that the Bolton-Hunter group, monoiodo or diiodo, basically makes a favorable contact with an accessory hydrophobic binding site on the avidin. The phenolic hydroxyl does not play much of a role unless it is ionized and unshielded as in I at pH 8.5, where it slightly weakens the hydrophobic interaction.
Model building shows that the distance between the biotin moiety and the iodine atom(s) of I or 111 can easily be made to match the corresponding distance between the biotin moiety and the side chain of Val-22 or Tyr-23 in [bio~ytin]~~ACTH,_~~amide (see above). Thus, we speculate that the exceptional affinity, cited above, of [biocytin] 25ACTH,_25amide for avidin is due, in part, to a hydrophobic interaction between avidin and one or more of the nonpolar side chains (e.g. Val-22, Tyr-23, Pro-24) on this peptide. Also the alkyl chain used as a spacer to give Na~B'-[6-(biotinylamido)hexanoyl]insulin a high affinity for avidin (Finn and Hofmann, 1985), or the N-terminal phenylalanine or adjacent valine of this insulin B chain, might similarly make a favorably hydrophobic contact with the avidin.
The dissociation of our 1:l complexes always begins with an anomalous relatively rapid phase, whose magnitude (percent of avidin sites responsible) varies from 1 to 18%, as shown in Table I. Eight percent of unmodified biotin undergoes this type of dissociation with a half-life of about 26 days, reminiscent of a "relatively rapid" dissociation (tH = 8 days) of 4% of the biotin observed previously by Green (1963). He suggested that it might be due to either impure ['"Clbiotin or to the presence of some denatured avidin. Our biotin derivatives and [3H]biotin were all repurified by HPLC and checked for purity by TLC immediately prior to our dissociation experiments, so our results are not explained by an impure tracer. Also, the magnitude of the anomalous dissociation varies significantly among the biotin derivatives, making it unlikely that it arises from some denatured avidin.