Immunochemistry, 1975, Vol. 12, pp. 849-854. Pergamon Press. Printed in Great Britain
AFFINITY LABELING OF ANTI-DNP, ANTI-DNP-GLYCYLGLYCYLGLYCINE AND ANTI-DNP-p-AMINOB ENZOYLGLUTAMANATE ANTIBODIES WITH 2,4-DINITROPHENYL-¢ -AMINOCAPROYL DIAZOKETONE AND 2,4-DINITROPHENYL-p-AMINOBENZOYL DIAZOKETONE* J. G. LINDEMAN, D. K. WOODARD, M. E. WOEHLER, G. E. CHISM and R. E. L O V I N S t Department of Biochemistry, University of Georgia, Athens, Georgia 30602, U.S.A. (First received 22 February 1975; in revised form 9 May 1975)
Abstract--Specifically purified antibodies to DNP-bovine gamma globulin (DNP-BGG), DNPglycylglycylglycine-BGG (DNP-GIy0 and DNP-p-aminobenzoylglutamate-BGG (DNP-pABG) were affinity labeled with the photoactive label 2,4-dinitrophenyl-E-aminocaproyl diazoketone (DNP-eACDK). The labeling of the antibody preparations as measured by the ratios of heavy to light chains (H:L) labeled were 2-9 and 0-7 for the two anti-DNP antibody samples, 2.2 for the anti-DNP-Gly3 antibody preparation and the H : L labeling ratio for the anti-DNP-p ABG antibodies were 1.1 and 4.0. A sample of anti-DNP-pABG antibodies was also labeled with the label, 2,4-dinitrophenyl-paminobenzoyl diazoketone (DNP-p ABDK). The ratio of heavy to light chain (H:L) labeled resulting from the labeling of anti-DNP-p ABG antibodies with DNP-p ABDK was 1.8. Heavy chains from the DNP-p ABDK labeled anti-DNP-p ABG antibody preparation were treated with nagarse and the peptides containing covalently bound label were isolated and analyzed by Edman degradation and mass spectrometry. The analytical results suggested that the label was primarily localized on histidine and glycine residues. This is in contrast with the results of affinity labeling of anti-DNP antibodies using DNP containing bromoacetyl, fluoroborate and DNP-l-azide labels which label primarily tyrosine, lysine, tyrosine and alanine residues respectively. The labeling differences may reflect differences in molecular size and reactivities of the three labels for different amino acids. INTRODUCTION antibody combining site properties of antibodies to In an earlier study (Woehler et al., 1974) rabbits the three haptenic groups, DNP, DNP-Gly3 and were immunized with 2,4-dinitrophenylated-bovine DNP-pABG, we have covalently labeled angamma globulin (DNP-BGG), bovine gamma globu- tibodies ellicited by BGG conjugates of each of lin conjugated with 2,4-dinitrophenylglycyi- these haptenic groups using the photoactive label glycylglycine (DNP-Gly3-BGG) and bovine gamma 2,4-dinitrophenyl-~-aminocaproyl diazoketone globulin conjugated with 2,4-dinitrophenyl-p- (DNP-¢ ACDK). DNP-p ABG antibodies were also aminobenzoyl glutamic acid (DNP-ABG-BGG) and affinity labeled with the photoactive label 2,4the antibodies to the DNP containing haptenic dinitrophenyl-p-aminobenzoyl diazoketone (DNPdeterminants (i.e. anti-DNP, anti-DNP-Gly3 and p ABDK). The results of this study are presented in anti-DNP-ABG) were specifically purified. These this communication together with the results of the antibodies exhibited a progressive restriction in mass spectral identification of the amino acid temporal and functional heterogeneity with the residues which were labeled in the heavy chains of increasing molecular size of the DNP moiety as the D N P - p A B G antibodies by the label DNPassessed by changes in hapten affinity with time pABDK. during the immune response and by binding heterogeneity. It was observed however that there MATERIALS AND METHODS was not a corresponding restriction in structural Antibody preparation heterogeniety as measured by microisolectric foThe rabbit anti-DNP-BGG (DNP), anti-DNP-glycylcussing. That all of the antibody preparations were glycylglycine-BGG (DNP-GIy3) and anti-DNP-p-aminospecific for the DNP portion of the hapten was benzoylglutamate-BGG (DNP-pABG) antibodies used shown by the almost complete cross-reactivity of in the labeling experiments with DNP-E-aminocaproyl diazoketone were elicited and purified in the manner each of the antibody preparations for all three of described earlier by Woehler et al. (1974). The rabbit antithe haptens. DNP-p-aminobenzoylglutamate (DNP-p ABG) anIn a continuing probe of the nature of the tibodies used in the labeling experiments with DNP-paminobenzoyl diazoketone were isolated by a more * Supported by a grant # AI 10086 from the National recently described procedure (Woehler et al., 1975). Institutes of Health and grant #GB 26555 from the National Science Foundation. Affinity labelling t NIH Research Career Development Awardee and to The photoactive label used in the labeling experiments, whom corresponsence should be addressed. DNP-~-aminocaproyl diazoketone (DNP-eACDK), was
849 IMM VoL 12, No. I1--A
J . G . LINDEMAN et al.
prepared using the procedure described in a previous publication (Cannon et al., 1974). The photoactive label, DNP-p-aminobenzoyl diazoketone (DNP-pABDK) was prepared using essentially the same procedure substituting 2,4-dinitrophenyl-p-aminobenzoic acid for 2,4dinitrophenyl-e-aminocaproic acid. The structure of the DNP-p ABDK label was confirmed from its infrared and mass spectral properties. The major distinguishing characteristic of the infrared spectra of both the caproyl diazoketone and the DNP-pABDK label is the strong diazo stretching frequency at 2110cm-'. Unlike the caproyl diazoketone label however the mass spectrum of the DNP-p ABDK label contained not only characteristic ions at M/e 167, 182, 257, 286, 299 and 303 but also included a strong molecular ion at M/e 328 and an M+-I ion at M/e 327. The mass spectrum of the DNP-eACDK label did not contain a molecular ion but instead exhibited an ion at M/e 279 corresponding to the loss of CH2N2 from the molecule. The antibodies were labeled in 5 mg quantities using a five-molar excess of label. One sample of 100 mg of DNP-pABG antibodies was labeled with DNP-pABDK using a twenty fold molar excess of label. The antibody sample was placed in a small beaker and the label added in dioxane solution (<1% v/v final dioxane concentration), in the dark at 4°C. From this point on, two different procedures were used to label the antibodies. Procedure 1 involved incubating the label-antibody mixture in the dark for at least one hr at 0--4°C, then passing this solution through a Sephadex G-25 column (2.5 x 40 cm) equilibrated and eluted with PBS (phosphate buffered saline solution: 0.01M phosphate; 0.15M sodium chloride; pH 7.4) in the dark. The eluted antibody was pooled and photolysed for 8 hr at a distance of 5 cm from six 15 watt Westinghouse F15T8/BLB high intensity u.v. lamps with the temperature maintained between 0 and 4°C. Procedure 2 differed from procedure 1 in that photolysis preceeded separation of excess label by G-25 chromatography. Antibodies from rabbits A-13 (antiDNP), G-13 (anti-DNP-Gly3), 1-13 and 1-14 (anti-DNPp ABG) and one of two samples of FY-3 (anti-DNP) were labeled using procedure 1, the rest of the samples were labeled using procedure 2. The labeled antibodies were all dialyzed for 24hr against 6 M guanidine hydrochloride-I N propionic acid and then against glass-distilled water until no radioactivity was detected in the final dialysate, and the protein and remaining label quantitated. The solution was subsequently concentrated for reduction and alkylation. Antibodies from rabbit A-13 (anti-DNP) were labeled with non-radioactive DNP-~ ACDK and passed through a BioGel P-6 column (2.5 × 40 cm) equilibrated and eluted with 6 M urea at 4°C. Equilibrium dialysis was performed on this labeled pool after it was dialyzed against glass-distilled water and finally PBS. The hapten used for binding study was tritiated 2,4-dinitrophenyl lysine. Competition experiments were performed in the same manner as the labeling experiments except that the natural hapten (DNP, DNP-GIy3, or DNP-pABG) or DNP-~aminocaproic acid (DNP-eACA) was added to the antibody preparations in a five-molar excess 1 hr before incubation of the antibody preparation with the affinity labeling reagent. Labeling procedure 2 with dialysis followed as above.
Separation of light and heavy chains of labeled antibody Affinity labeled antibody was completely reduced and alkylated according to the procedure of Haimovich et al. (1972) with minor modifications. The ratio of label present in the light and heavy chain fractions was determined by liquid scintillation counting of the two fractions containing the 3H label after they had been dialyzed exhaustively against glass-distilled water.
Nargarse digestion of aninity labeled heavy chains The method of Thorpe & Singer (1969) was used to completely digest the antibody heavy (H) chains of rabbit 1-14 (DNP-pABG) labeled with DNP-p-aminobenzoyl diazoketone (DNP-p ABDK). To a solution of H chains (= 10 mg/ml in 0.20 M ammonium bicarbonate (NH,HCO3), pH 8.0, nargarse (Enzymes Development Corp., N.Y., ) was added to give a weight ratio of H chain/enzyme of twenty. The mixture was held at 37°C for 20 hr. Insoluble material after that time was removed by centrifugation at 6000 g at 4°12. The supernatant was stored frozen. Purification of labeled peptides Nagarse digests were made of the heavy (H) chains of rabbit 1-14 only. These heavy chains had been affinity labeled with the radioactive ~H-DNP-p ABDK label and were chosen for this part of the study due to the large quantity of antibody isolated from this rabbit. The procedure used (Cannon et al., 1974), employs antidinitrophenyl (DNP) antibodies produced and isolated by a previously described procedure (Woehler et al., 1974). Nagarse digests were fractionated on a BioGel P-2 column (1.5 x 28 cm) equilibrated and eluted with 0.05 M ammonium bicarbonate (NH4CHO3), pH 8.0. Fractions containing radioactivity were pooled and dried by lyophilization. 3H-DNP-pABDK labeled material was separated from unlabeled material by the addition of a molar equivalent amount of anti-DNP antibody followed by 50% ammonium sulfate [(NH,)2SO,] precipitation. The precipitates were centrifuged at 8000 g at 4"C for 20 min and were washed three times with saturated (NH,)2SO, followed by recentrifugation under the above conditions. The precipitates of anti-DNP-antibody bound ~H-DNPABDK-labeled peptides were solubilized in PBS. The labeled peptides and amino acids were displaced from the anti-DNP-antibody by the addition of a four-fold molar excess of 2,4-dinitrophenol (DNP-OH) (0.1 M solution in PBS) to antibody. This mixture was held at room temperature for 10 min, after which an equal volume of saturated (NH,)2SO, was added. Following centrifugation, the supernatant was chromatographed on a BioGel P-2 column (1.5x30cm) equilibrated with glass-distilled water. Fractions containing radioactivity were pooled and dried by rotary evaporation. Residual (NH,)2SO, was removed by extraction with absolute ethanol. Mass spectral analysis of labeled amino acids The peptides containing covalently bound DNPpABDK, which were purified from the nagarse digests were treated with excess methylisothiocyanate in a manner previously described (Fairwell et al., 1970). The resulting derivatized peptide mixture was analyzed on a Du Pont Instrument Model 490 mass spectrometer, using techniques described elsewhere, with the exception that pyridine/water 80 vol/20 20 vol) was used as the reaction solution (Fairwell et al., 1973; Cannon et al., 1974). RESULTS
In t h e s e studies, specifically purified r a b b i t a n t i b o d i e s to D N P - B G G , D N P - G l y 3 - B G G a n d D N P - p A B G - B G G w e r e c o v a l e n t l y labeled with t h e photoactive label (3H)DNP-e-aminocaproyl d i a z o k e t o n e ( D N P - ~ A C D K ) (a) a n d specifically purified r a b b i t a n t i b o d i e s to D N P - p A B G - B G G w e r e labeled w i t h t h e p h o t o a c t i v e label (3H) D N P p - a m i n o b e n z o y l d i a z o k e t o n e ( D N P - p A B D K ) (b). T h e labeling e x p e r i m e n t s w e r e c a r r i e d o u t u s i n g t h e t w o p r o c e d u r e s d e s c r i b e d in M a t e r i a l s a n d M e t h o d s . A n t i b o d i e s to D N P (rabbits # A-13 a n d FY-3), DNP-Gly3 ( r a b b i t # G-3) a n d D N P - A B G (rabbits # I - 1 3 a n d 1-14) w e r e labeled u s i n g
Affinity labeling of anti-DNP antibodies o II
'NO2 0 II
procedure one. The labeling experiments involving competition between label and excess hapten were performed using procedure two. The results of the labeling experiments are tabulated in Table 1. The ratios of H / L chains labeled ranged from 0.70 to 4.0. In attempts to assess the extent of non-specific interaction of label with antibody, an anti-DNP antibody preparation (A-13) was labeled with nonradioactive DNP-~ ACDK. The binding affinity of the labeled antibody for the hapten (3H) DNP-lysine was determined by equilibrium dialysis techniques described elsewhere (Woehler et al., 1974). The binding of DNP-lysine by A-13 (antiDNP) antibody was reduced by 67% after introduction of the D N P - ~ A C D K label. The K0 for the unlabeled antibody was 3-6 x 106 while that for the labeled antibody was 2.2 x 10 6. Plots of bound vs unbound hapten obtained from the equilibrium dialysis experiments with the labeled and unlabeled A-13 antibody showed a 66% decrease in binding site concentration (1.6 x 106 moles for the unlabeled vs 1.1 × l06 moles for labeled antibody) (Fig. 1), suggesting that the label was labeling primarily to the combining site. To further measure the extent of non-specific labeling, a sample of bovine gamma globulin (BGG) Table 1.
Extent of Labeling" Whole Molecule K. ~(L/M × 10 6) Irradiation After Dialysis
Labeling Ratio (R) H/L
A. Results of DNP-EACDK Labeling of DNP, DNPGLY3 and DNP-p ABG Antibody Preparations. H-1 (anti-DNP) H-1 + DNP-LYS c H-1 + DNP-EACA~ FY-3 (anti-DNP) FY-3 + DNP-LYS c FY-3 + DNP-EACA' G-3 (anti-DNP-GLY3) G-3 + DNP-LYS c G-3 + DNP-EACAc 1-13 (anti-DNP-p ABG) 1-14 (anti-DNP-p ABG)
2.3 --1.3 --4-4 --0.77 0.91
0.83 1.1 0.83 1.0 1.0 1.2 0.87 0.510 0"83 1-43 0-43
2.9 --0.7 --2.2 -4.0 1.1
B. Results of DNP-p ABDK Labeling of DNP'-p ABG Antibody Preparations. 1-14 (anti-DNP-p ABG)
°Moles of label per mole of protein. bKo measured for each antibody using the natural hapten. CAntibody was incubated with five-fold excess of hapten (indicated) prior to incubation with five-fold excess of H3-DNP-~ACDK and irradiation.
3"0-0201.0 6 0
•"~'bX106 Fig. 1. Equilibrium dialysis of A-13, unlabeled (© O) and labeled (A &) with DNP-~ aminocaproyldiazoketone vs the hapten 3H DNP-lysine. l/H, vs l/Hb. Ab* (labeled antibody) concentration is 3.03 x 10 7 M (/x A). was labeled with a five-fold excess 3H-DNPEA C D K using procedure two outlined for the labeling of the antibody preparations. After G-25 chromatography and separation of light and heavy chains there appeared to be no label covalently bound to the BGG protein as measured by scintillation counting of the light and heavy chains (Fig. 2) (Cannon et al., 1974). In competition experiments, antibody preparations previously incubated with a 5 molar excess of their natural hapten (i.e. DNP-Lys, DNP-Gly3 or D N P - p A B G ) were labeled with 3H-DNP-EACDK (5 molar excess) using procedure two described in the experimerital section. The results of these experiments are also included in Table 1. In each case it is seen that almost an equivalent amount of label is incorporated in the presence of hapten. The amount of label incorporated was measured prior to reduction and alkylation and may reflect noncovalently bound label thus accounting for the high concentration of label incorporated in the presence of hapten. A second portion of anti-DNP-pABG antibody (1-14) was labeled with the label D N P - p aminobenzoyl diazoketone. After separation of the light and heavy chains on a Sephadex G-100 column the heavy chain pool was digested with the enzyme nagarse. The labeled nagarse peptides were isolated as described in Materials and Methods and the selective cleavage and identification of the Nterminal amino acid residues of these labeled peptides (isolated from the P-2 column) was accomplished using the modified Edman degradation procedure (Fairwell et al., 1970). The resulting methylthiourea derivatives were thermally cyclized in the mass spectrometer and were identified as methylthiohydantoin derivatives (MTH) by their
J.G. LINDEMAN et al.
- 30 //
24 28 Tube No.
Fig. 2. Control experiment, labeling of BGG with DNP-~-aminocaproyl diazoketone followed by separation of BGG and excess label by Sephadex G-25 chromatography. (1) A28o (O O), (2) CPM/ml ( - - - - ) , (3) Fraction volume = 8.88 ml/tube.
characteristic mass spectra produced for each amino acid residue (Fairwell et al., 1970). Using this procedure the amino acids containing the covalently bound affinity label were isolated and tentatively identified. Since the DNP-p aminobenzoyldiazoketone reacts via the ketene or ketocarbene intermediates as shown in Fig. 3, the reactive species when covalently bound to the side chain of the amino acid will add 300 a.m.u, to the apparent mol. of the amino acid methyl thiohydantoin (MTH). Consequently the mass spectra of the DNP-pABDK-Iabeled amino acid MTH's should contain molecular ions at masses which correspond to the mass of the MTH plus 300 a.m.u. Fragmentation ions from the labeled amino acid MTH's are also expected to be present. The amino acids containing the covalently bound label were isolated on the fourth and fifth steps of the sequential degradation of the peptide mixture. In the mass spectra of the fourth residue, ions at M/e 430, 402 and 401 a.m.u, were detected (Fig. 4).
These were presumed to be due to DNP-p ABDKlabeled-glycine since they correspond to the molecular ion and major fragment ion of glycine MTH (M/e 130, 102 and 101) plus 300 a.m.u. The spectrum of the next residue containing ions at M/e 510, 508 and 430 a.m.u. (Fig. 5) suggested the presence of DNP-pABDK-Iabeled-MTH histidine residues again based on ions due to the characteristic histidine MTH ions at M/e 130 and 210 plus 300 due to the label. The ion at M/e 508 is probably an M+-2 fragment which is analogous to the M/e 208 found in the spectrum of histidine MTH. The ions present in the high mol. wt region corresponding to the labeled histidine and glycine MTH's were free of contaminating ions and were not present in spectra corresponding to a blank reaction. In addition, the lower mol. wt region of the spectra contained all of the fragment ions for the same (histidine and glycine) amino acids. The representative mass spectra of these residues are shown in Figs. 4 and 5.
2 v o
'NOz ~lff rearrangement (c)
Fig. 3. The sequence of reactions which occur upon photolysis of the photoactive label DNP-p-aminobenzoyl diazoketone, a. diazoketone; b, ketocarbene generated by high intensity light; c, ketene spontaneously formed from the ketocarbene via a Wolff rearrangement.
Affinity labeling of anti-DNP antibodies
(M/e 4 0 ~
(M/e 102 ) 25
Fig. 4. Mass spectrum of the methyl thiohydantoin isolated from step 4 to the Edman degradation of the labeled containing peptides isolated from nagarse digestion of the heavy chain of anti-DNP-p ABG antibody labeled with DNP-p-aminobenzoyl diazoketone.
75 (M/e 430} ~" (M/e 510)
(M/e 210) g
Fig. 5. Mass spectrum of the methyl thiohydantoin isolated from step 5 of the Edman degradation of the labeled containing peptides isolated from nagarse digestion of the DNP-p antibody heavy chains labeled with DNP-p-aminobenzoyl diazoketone. DISCUSSION
The heavy to light chain labeling ratios (H : L) of 3:1 for H - l , 0-7:1 for FY-3, 2.2:1 for G-3, 4:1 for 1-13 and 1:1 for 1-14 are in accord with the 'characteristic' H : L ratios of 2--4:1 repeatedly observed for the affinity labeling of ' D N P ' antibodies (Franek, 1971; Good et al., 1968; Thorpe & Singer, 1969; Press et al., 1971). These ratios may represent averages of a number of different orientations of the labels in the DNP antibody combining site population hence labeling both light and heavy chains of the same combining site or alternatively may represent an average of individually homogeneous sites which have been labeled by the affinity reagent exclusively on either the heavy or light chain. The results of labeling experiments using monoclonal mouse myeloma proteins having DNP specificity would tend to support the latter alternative (Singer et al., 1971; Haimovich et al., 1972; Goetzl & Metzger, 1970; Martin et al., 1972). The labeling of mouse MOPC-315 protein with a bromoacetyl label and mouse HPC-3 protein with a diazoniumfluoroborate label resulted in H : L ratios in both cases of 10:1 while the labeling of mouse
MOPC-315 protein with the same labels produced labeling H : L ratios of 0.05. These labeling patterns suggest that in an individually homogeneous combining site the label does not assume a variety of orientations labeling both light and heavy chains but rather assumes a discreet orientation labeling primarily either the heavy or light chain. In view of the labeling results of mouse myeloma proteins, the reproducibly characteristic H : L labeling ratio of 1-4:1 observed both in these studies as well as in other investigations may reflect the production of antibody populations in response to a dinitrophenylated antigen containing population similarities which are reflected in the reproducible H : L labeling ratios. It is interesting to note that the labeling ratios of the three antibody preparations used in this study did not differ substantially with one another and with the observed H : L ratio for another DNP antibodies even though the DNP haptenic groups in the antigens used to ellicit the antibody response were in different molecular and conformational environments. While it has been observed that the functional and temporal heterogeneity of the antibody preparations to DNP-BGG, DNP-Gly3 BGG and DNP-p ABG-BGG decrease with increasing hapten molecular size, the structural heterogei~y and isoelectric profiles are essentially the same for all three antibody preparations (Woehler et al., 1974). This similarity in structural and isoelectric properties parallels the similarities in labeling profiles again suggesting that there are similarities in the composition of the antibody population ellicited in each case. The localization of the D N P - p A B D K label on histidine and glycine is in accord with the results obtained by labeling anti-DNP antibodies with DNP-EACDK in which the covalently labeled amino acids histidine, alanine and phenylalanine were identified (Cannon et al., 1974). These results are similar to those obtained from the labeling of rabbit anti-2-nitro-4-azidophenyl (NAP) antibodies with E-4-azido-2-nitrophenyl-l-lysine in which cysteine and alanine were labeled (Press et al; 1971). When DNP substituted bromoacetyl labels
J. G. LINDEMAN et al.
(Haimovich et al., 1972) diazoniumfluoroborate (Goetzl & Metzger, 1970; Franek, 1971 ; Good et al., 1968) and DNP-l-azide labels (Richards et al., 1974) were used the predominant amino acid reportedly labeled was tyrosine. These differences may reflect size differences and differences in reactivity of the labels toward specific amino acids rather than major differences in the combined compositions of the combining sites in the different antibody populations. Because of the heterogeneity of the antibody populations studied, these alternatives are difficult to distinguish at present. Work is presently under way, however, to further study the structural features of the combining regions of antibodies to the haptens DNP, DNP-GIy3 and D N P - p A B G using preparative isoelectrofocussing to provide antibodies of restricted structural heterogeneity.
Cannon L. E., Woodard K. K., Woehler, M. E. & Lovins R. E. (1974) Immunology 26, 1183.
Fairwell T., Barnes W. T., Richards F. F. & Lovins R. E. (1970) Biochemistry 9, 2260. Fairwell T., Ellis S. R. & Lovins R. E. (1973) Analyt. Biochem. 53, 115. Franek F. (1971) Eur. J. Biochem. 19, 176. Goetzl E. J. & Metzger H. (1970) Biochemistry 9, 1267, 3862. Good A. H., Ovary Z. & Singers S. J. (1968) Biochemistry 7, 1304. Haimovich J., Eisen H. N., Hurwitz E. & Givol D. (1972) Biochemistry 11, 2389. Martin N., Warner N. L. & Singer S. J. (1972) Biochemistry 11, 4999. Press E. M., Fleet G. W. J. & Fisher C. E. (1971) in Progress in Immunology (Edited by Amos B.), p. 233. Academic Press, New York, N.Y. Richards F. F., Lifter J., Hew C-L., Yoshioka M. & Konigsberg W. H. (1974) Biochemistry 13, 3572. Singer S. J., Martin N. & Thorpe N. O. (1971) Ann. N. Y. Acad. Sci. 190, 342. Thorpe N. O. & Singer S. J. (1969) Biochemistry 8, 4523. Woehler M. E., Cannon L. E., Clark P. D. & Lovins R. E. (1974) Immunology 26, 957. Woehler M. E., Chism G. E., Lindeman J. G. & Lovins R. E. (1975), J. Immun. Meth. 6, 301.