Regulation of the humoral immune response by polyspecific natural autoantibodies

© ELSEVIER Paris 1988

Ann. lnsl. Pasteur/Immunol.

1988, 139, 349-360

REGULATION OF THE HUMORAL IMMUNE RESPONSE BY POLYSPECIFIC NATURAL Ar~rOANTIBODIES W. Mahana, B. Guilbert and S. Avrameas Unit6 d'lmmunocytochimie, Institut Pasteur, 75724 Paris Cedex 15

SUMMARY

Two different BALB/c IgMk polyspecific monoclonal natural autoantibodies E7 and D23 were administered to neonatal BALB/c mice. When adults, these mice were immunized and challenged with calf myosin, BALB/c actin, human transferrin, calf thymus DNA or TNP-coupled bovine serum albumin (TNP/BSA), in complete Freund's adjuvant. The levels of serum antibody were evaluated by enzyme immunoassay. No differences in anti-actin, anti-transferrin and anti-DNA antibody titres were noted between control and antibody-treated mice. However, anti-myosin antibody titres significantly increased in mice treated with either the E7 or D23 antibody, and anti-TNP antibody titres significantly decreased in mice treated with E7 but no: with D23. These differences persisted after antigenic challenge and involved only the IgG response of treated mice. These results suggest that polyspecific natural autoantibodies may be involved in the regulation of the humoral immune response. KEY-WORDS" Autoantibody, Immunoregulation; Humoral response, Polyspecificity, EIA. INTRODUCTION Naturally occurring autoantibodies have been found in normal human and animal sera [1, 4, 5, 6, 11, 15]. These antibodies recognize a variety of constituents, certain haptens and effectors of the immune response. Furthermore,

Submitted March 18, 1988, accepted May 28, 1988. Correspondence to W. Mahana.

350

W. M A H A N A

AND COLL.

hybridomas prepared from physiologically normal mice have been found to produce monocionai antibodies reacting with several apparently antigenicmly dissimilar self and non-self antigens [8, 9]. Several hypotheses have been formulated concerning the biological role of natural autoantibodies. It has been proposed that natural antibodies act as a primary line of defence against pathogenic agents and that they may be involved in the clearaT::ce of catabolic products from the organism, in induction of tolerance and in regulation of the immune response [2, 7, 10]. In the present study, we administered two different IgMk monoclonal polyspecific natural a1~toantibodies to newborn mice. When these mice reached adulthood, they were immunized and challenged with various antigens incorporated in complete Freund's adjuvant and their serum antibody titres were evaluated. The results obtained demonstrate that animals treated with the two ~atural autoantibodies responded differently from control mice to antigenic stimulation, thus suggesting that natural autoantibodies may be involved in the regulation of the hun-ioral immune response.

MATERIALS

AND METHODS

l~Lce. E x p e r i m e n t s w e r e p e r f o r m e d w i t h n e w b o r n B A L B / c m i c e b r e d in t h e c o l o n y o f the Institut Pasteur, Paris.

Antibodies.

The two IgMk natural monoclonal autoantibodies E7 and D23 were obtained by fusion of the non-secreting myeloma cells X63 and SP2/0, respectively, with spleen cells from 12-week old unprimed BALB/c mice [9]. These monoclonal antibodies were found to react with a large panel of antigens, as described by Dighiero et al. [9]. The monoclonai IgM were purified from ascitic fluid either by precipitation at low ionic salt concentration [18] or by gel filtration on S.200 Sephacryl [3]. In addition, E7 was isolated by affinity chromatography on a DNP-lysine Sepharose-4B c~lamn [14].

BSA = bovine serum albumin. DNA/MBSA = deoxyribonucleic acid/methylated BSA. EIA = enzyme immunoassay. PBS = phosphate-buffered saline.

TNPll

= IgMk anti-TNP monoclonal antibody.

TNP2s/BSA, TNP,oo/[3-gal and TNPI3/OVA = trinitrophenylated BSA, [~galactosidaseor ovalbumin.

A UTOANTIBODIES I N H U M O R A L I M M U N O R E G ULA TION

351

One monoclonal IgMk antibody with specificity o ~ y for TNP (TNP11) was obtained by immunizing BALB/c mice with TNP-Ficoll and was a kind gift of Dr D. Jouy (institut Pasteur, Paris, France); it was purified by affinity chromatography on a DNP-lysine-Sepharose-4B column. Antigens.

Muscle mouse ~ctin [21] and calf myosin [25] were prepared according to described methods. Purified human transfendn was purchased from Behringwerke A.G. (Marburg, FRO. Calf thymus DNA and bovine serum albumin (BSA) were from the Sigma Chemical Co. (St. Louis, Mo.). dsDNA and ssDNA were complexed to methylated BSA according to procedures previously described [19]. Trinitrophenylation of BSA (TNP25/BSA), ~-galactosidase (TNP100/~-gal) and ovalbumin (TNPI3/OVA) were performed as previously described [16]. Treatment of mice with monodonai antibodies.

Four groups of newborn mice were immunized intraperitoneally with either sterile phosphate-buffered saline (PBS) or with E7, D23 or TNP11, also in PBS. Animals were injected as outlined in table I.

TABLE I. - - Injection protocol in newborn mice. Injection number

Days after birth

Diluent volume (PBS)

Quantity of injected ]gM (E7 or D23 or TNP11)

1 2 3 4 5 6 7 8 9 10 11 12

0 2 4 6 8 I0 12 14 16 18 20 22

50 ~.I 50 btl 50 I.d 50 p.l 50 Izl 50 ~tl 100 ~ 100 100 ~1 100 pd 100 I~l 100 ~1

20 V-g 20 ~g 20 t~g 20 Izg 5O izg 50 Izg 100 ~tg 100 ~tg 100 ~tg 100 ~tg 100 ~tg 100 ~tg

Animal immunization.

Two-month old treated and control animals were bled and injected intramuscularly five days later with each of the antigens suspended in complete Freund's adjuvant (100 lzg for all the antigens with the exception of TNP2 -BSA, which was used at 25 ~g)° ~ r u m was drawn twenty-one days after i m m ~ o n . The mice were boosted on day 30 with the antigen as above. Serum samples were collected on day 40.

352

IV. M A H A N A A N D COLL.

Screening of serum antibody activity by enzyme immunoassay (EIA). Serial dilutions of mouse serum in PBS containing 0.I °70 Tween 20 and 0.5 o70 gelatin were incubated on Luxlon microtitre plates (CEB, Nemours, France) coated with the various antigens and processed as described in d~v,~1 elsewhere [12].

RESULTS The multiple specificities of the two IgMk natural monoclonal autoantibodies E7 and D23 have been described in detail elsewhere [9, 17]. The dissociation constants of E7, D23 and TNP11, ~¢hich reacts only with TNP, have been reported [22]. The immunizing antigens were selected because of their reactivity with E7 and D23 and for the following additional characteristics. BALB/c mouse ac÷~, .... ~,~ r~lf myn~in were chosen A~ ~elf a.~"l nc ~-.~elfintracytoplasmic constituents; DNA (complexed to methylated BSA) was chosen because it is an intranuclear antigen which can be considered either as self or as a sequestered <~non-self>> antigen; human transferrin because it represents a humoral nonself antigen; and TNP/BSA because almost all murine natural autoantibodies react with TNP. In a first series of experiments, 64, 60 and 67 mice were injected with E7, TNP11 and PBS, respectively, as outlined in table I. At two months after birth, serum samples in a 1/200 dilution were tested by EIA using plates coated with either actin, myosin, DNA, transferrin or TNP2s/BSA. No differences in natural antibody titres to these antigens were noted among the three groups of mice. Five days after blood sampling, the 3 groups of mice were each divided into three subgroups and injected with complete Freund's adjuvant containing calf myosin, BALB/c actin or E~NA/MBSA. Sera were collected 21 days later and examined by EIA using plat,,:s coated with the same antigens as above. It was found that the IgM and IgG ar~tibody amounts for all antigens examined with the exception of myosin were net sigl~ficantly different among the various subgroups. Mice treated with E7 and immunized with myosin had somewhat lower amounts of IgM anti-myosin antibodies than mice injected with PBS and then immunized with myosin (fig. 1). In contrast, in the same mice, the amounts of IgG antibody to myosin were increased compared to control mice (fig. 1). This difference was significant (p 0.05 Student's test) and it was found to persist after a second immunization 30 days after the first (fig. 1). This difference was also noted after a third (day 60) or fourth (day 90) administration of antigen, but a statistical analysis could not be carried out. Mice treated with the TNP11 antibody and then immunized with the various antigens had antibody titres similar to those found in PBS-treated animals. In order to confirm the above results, a second series of experiments was performed with mice treated with either PBS (20 mice) or E7 antibody (22 mice). Each group of animals was divided into two subgroups; the first

A UTOANTIBODIES I N I I U M O R A L I M M U N O R E G U L A l I O N

353

2

1.000

IGM

IGM

.750

.500

,-

.250

r~ A "o

C .m

J~ >, ,1o o .o

B

C

A

B

1.250

1.000

1

t°

2 IGG

IGG C

,,¢

.750

.500

.250

A

B

C

A

B

C

FIG. 1. - - Serum antibody amounts (1/200 dilution) of mice treated with E7 antibody (blank columns) or PBS (hatched columns) and immunized with DNA complexed or methylated bovine serum albumin (A), B A L B / c actin (B) or calf myosin (C).

Sera were tested for IgM and IgG isotypes after primary immunization (1) and after secondary challenge (2). Values given in each group correspond to the mean valae of 20 mice sera tested individually.

was immunized with calf myosin and the second with human transferrin. Sera were collected 21 days after antigen administration, and on day 30 the animals were challenged a second time with the same antigens and sera were collected l0 days later. The m o u n t s of anti-myosin and anti-transferrin antibodies were evaluated by EIA. The sera of mice treated with E7 antibody after a primary immunization with myosin and even more so after a second injection contained significantly higher amounts of IgG anti-myosin antibodies (p = 0.05)

354

W. M A H A N A A N D COLL.

1.500

-

:

1.250

-

< w

1.000

--

.750

--

.500

-

.250

--

e.

t~ c

c m JD >, "0 0

.a t-

I

1/400

1

1151200

1112800

~/3206

I/8G~

Serum d i l u t i o n s

FIG. 2. - - Serum IgG antibody amounts of mice treated with E7 antibody ( ~ ) or PBS (©)

and immunized with calf myosin after primary (1) or secondary (2) challenge.

Values given are as in figure 1.

than did the sera of mice treated with PBS (fig. 2). In contrast, no differences in IgM or IgG anti-transferrin antibody levels were observed between the two subgroups of mice immunized with transferrin. In order to test the effe,:t of the D23 antibody, experiments similar to those described above were carried out. Mice were treated with either D23 antibody (47 mice) or with PBS (50 mice) subdivided into three groups each and immunized twficewith either myosin (17 mice), DNA/MBSA (15 mice) or transferfin (15 mic3). Figure 3 shows the amounts of IgG antibody found in mice injected with D23 or PBS and subsequently immunized with myosin. The sera of mice trea[ed with D23 contained higher amounts of IgG antimyosin antibody than did the sera of mice treated with PBS. This difference was statistically significant (p = 0.05 Student's test) after the first immunization at serum dilutions higher than 1/100, and it persisted and even increased after the second administration of antigen. In contrast, the group of mice treated w~th either D23 or PBS and then immunized with transferrin or DNA prodaced equivalent amounts of the corresponding antibodies. Similarly, animals treated with TNP 11 (60 mice) and then immunized with the various antigens had antibody titres on the same order of magnitude as those found in PBStreated mice.

AUTOANTI&3DIES

IN HUMORAL

355

IMMUNOREGULATION

/

1500-

1.250-•

g g

:

1.ooo -

.750-

"0 e.O

~,

.500 -

0

= :

250

z 1/2700

i 1/900

i

i 'llll'"

I

1/300

11100

Serum

I

I

I.~27000

I IIII

i

1/9000

i

I

I

I Illtj

1/3000

111000

dilutions

FIG. 3. -- Serum IgG antibody amounts o f mice treated with D23 antibody ('~) or PBS (©) and immunized with calf myosin after primary (1) or secondary (2) challenge.

Valuesgivenin each group correspond to the mean value of 17 mouse sera tested individually.

In the last series of experiments, groups of newborn mice received either E7 (18 mice), D23 (20 mice), T N P l l (20 mice) or PBS (17 mice). When they reached adulthood, they were immunized twice with TNP~5/BSA. No differences were noted in the amounts of IgM anti-TNP antibody among the variou~ mice. In contrast, it was noted that mice treated with either the E7 or TNP11 antibodies had significantly lower amounts (p 0.05 at a 1/400 dilution) of IgG anti-TNP antibodies (fig. 4 A) than did mice treated with D23 or PBS. This difference persisted after a second immunization and was observed whether plates coated with TNPI3/OVA (fig. 4 B), TNP100/[~-gal or TNP2~/BSA were used (data not shown).

DISCUSSION The results obtained show that the injection of rather small amounts of two natural autoantibodies into newborn mice can alter their subsequent adult immune response. Such an alteration, when it occurs, appears to be of a permanent nature because it is observed not only after primary antigen administration but also after secondary challenge. It is noteworthy that only the IgG immune response appears to be affected by such treatment.

356

W. M A H A N A A N D COLL.

A

1.000 I .750

_

.500 -.250

_

i: :3

<

I

I

I

I

C

lJi I

I

I

t

I

I

I

I Ili

B

!=

"0 0 .12

1.000

- -

C

.750

--

.500 -.250 -

I

1110800

I

I

1/3600

I

I I II

I

1/1200 Serum

I

I

I

I

I I I I l

I1400

dilutions

FIG. 4. - - Serum IgG antibody amounts o f mice treated with E7 antibody (<~), T N P l l antibody ( ~ , 1)23 antibody ( 0 ) or PBS (C)), and immunized one (,4) or twice (B) with TNP/BSA. Plates coated with TNP-OVA were used. Values given in each group correspond to the mean value of (17 and 20) mouse sera tested individually.

A UTOANTIBODIES IN HUMORAL

IMMUNOREGULA TION

357

These results indicate that natural autoantibodies may play a regulatory role in the humoral immune response. With the two natural autoantibodies examined, such regulation resulted in an amplification of the anti-myosin response, whereas no modification was noted in mice injected either with the induced IgM ( T N P l l ) anti-TNP antibody or with PBS. In contrast, only the two natural autoantibodies as well as TNPI 1 resulted in a diminution of the anti-TNP response. The modifications, although statistically significant, were moderate in all cases. At least two explanations can be advanced for this pher~omenon. Firstly, the amount of injected natural autoantibody was low compared to the circulating concentrations normally present in mice. The second possibility is that natural autoantibodies, because of their multiple reactivities, constitute a network that is difficult to perturb. Indeed, the two IgM natural autoantibodies E7 and D23 used in the present work, as well as almost all the other natural autoantibodies examined, react with practically the same panel of antigens, although to a different extent, and share common idiotypic determinants. These common characteristics make it difficult to assess the mechanisms by which E7 and D23 exert their immunoregulatory action. However, the differential effects on the anti-myosin versus the anti-TNP response suggest somewhat varied mechamsms. Recently, it has been shown that a specific murine monoc!ona! igM antiactin antibody was able to increase cellular metabolism and this activity was abolished after passage of the antibody preparation on an actin immunoadsorbent [20]. Both the monoclonal polyspecific natural autoantibodies used in this study possess a pronounced reacth4ty for actin. It is possible, therefore, that natural autoantibodies, by reacting with structures resembling actin (and possibly other cytoskeletal proteins) exposed on the surface of lymphocytes, may change the metabolism of these cells and lead to a modified immune response. Recently, it has been demonstrated that autoantibody-producing cells are present in embryos and newborn mice. The immunoglobulins synthesized by these cells possess a higher idiotypic connectivity among them, have multiple antibody and idiotypic specificities and their injection into mice results in a modification of the immune response to phosphorylcholine and/or to dextran [23, 24]. It has been reported that E7 and D23 antibodies possess common idiotypic determinants which they share with almost all other routine polyspecific natural autoantibodies. Furthermore, it has been shown that the intera~ions of the antiidiotypic ~ntibody with E7 and with poiyclonal murine anti-myosin IgG were s p e c i i ! ~ y ~n~bited by myosin [17]. From these observations, it can be inferred : ;a~ ; ~,e early establishment of the idiotypic network was disturbed by the in~e~::~l~of the monoclonal autoantibodies leading to an altered anti-myosin re'~:.~ nse. Similarly, the h'nportant role of antibody affinity in the suppression of the humoral immune response has already been stressed [13]. The difference in the anti-TNP response observed among mice injected with either the E7 or the TNP11 antibody and mice injected with D23 probably reflects differences

358

W. M A H A N A A N D COLL.

in the affinity of these antibodies for TNP [22]. Thus, the affiP2ties of the natural autoantibody E7 and that of the induced a,ltibody TNPI 1 either for free TNP or for TNP-protein-conjugates are similar and both are at least one order of magnitude higher than those of D23.

RrkStrM~ RI~GULATION DE LA RI~PONSE HUMORALE PAR LES AUTOANTICORPS NATURELS POLYSPI~CIFIQUES

Deux anticorps naturels polysp~cifiques monoclonaux IgMk (E7 et D23), provenant de la souris BALB/c normale non immunis6e, ont ~t6 administr~s aux souris BALB/c nouveau-n~es. Ces souris devenues adultes ont ~t~ immunis6es (avec injection de rappel) par la myosine de veau, l'actine de BALB/c, la transferrine humaine, I'ADN du thymus de veau ou le trinitroph6nyl (TNP) coupl~ a la s~ralbmnine bovine (TNP/BSA), en presence de l'adjuvant de Freund. Le titre des antico~s dans le s~rum des souris a ~t~ ~valu~ par dosage immunoenzymatique. Nous n'avons pas trouv~ de difference ell ce qui concerne le titre des anticorps anti-actine, anti-transferrine et anti-ADN. Cependant, le titre des anticorps anti-myosine a 6t6 significativement augment~ chez les souris trait6es avec les deux anticorps E7 et D23 et le titre des anticorps anti-TNP a ~t~ significativement diminu6 chez les souris trait~es avec E7 mais ne l'a pas ~t~ pour celles trait~es avec D23. Ces diff6rences persistent apr~s l'injection de rappel et affectent seulement la r6ponse IgG chez les souris trait~es. Ces r~sultats sugg~rent que les autoanticorps naturels polysp6cifiques pourraient jouer un r61e daL3 la r6gulation de la r~ponse immunitaire humorale. MOTS-CL~S: Autoanticorps, Immunor6gulation; R6ponse humorale, Polysp6cificit6, Dosage immunoenzymatique.

REFERENCES

[1] AvP~m~s, S., GmLBERT, B. & DIoHm~o, (3., Natural antibodies against tubulin, actin, myoglobin, thyroglobulin, fetuin, albumin and transferrin are present in normal human sera and monoclonal immunoglobul~ns from multiple myeloma and Waldenstr6m's macroglobulinemia may express similar antibody specificities. Ann. lmmunol. (Inst. Pasteur), 1981, 132 C, 231-236. [2] AVRAM~AS,S., DIOHmRO,G., LYMn~m,P. & GUrLBERT,B., studies on natural antibodies and autoantibodies. Ann. Immunoi. (Inst. Pasteur), 1983, 134 D, 103-113.

A UTOANTIBODIES I N H U M O R A L I M M U N O R E G U L A TION

359

[3] BOUVET,J.P., PraEs, R. & PILLOT,J., A modified gel filtration technique producing an unusual exclusion volume of IgM: a simple way for preparing monodonal IgM. J. Immunol. Methods, 1984, 66, 299-305. [4] DE MAEYER-GUIGNARD,J., CACHARD-THOMAS,A. & DE MA~VER,E., Naturally occurring anti-interferon antibodies in Lou/C rats. J. Immunol., 1984, 133, 775-778. [5] DE MA~VER-GuIGNARO,J. 8' DE MAEYER,E., Natural antibodies to interferon-a and interferon-t~ are a common feature of inbred mouse strains. J. Immunol., 1986, 136, 17~8-1711. [6] DIGnIERO, G., GUILBERT,B. & AVRAMEAS,S., Naturally occurring antibodies against nine common antigens in human sera. - - II. High incidence of monoclonai Ig exhibiting antibody activity against actin and tubulin and sharing antibody specificities with natural antibodies. J. Immunol., 1982, 128, 2788-2792. [7] DIGrIIERO, G., LYMBERI, P., GUILBERT, B., TERm'NCK, T. & AVRAMEAS,S., Natural autoantibodies constitute a substantial part of normal circulating immunoglobulins. Ann. N. Y. Acad. Sci., 1986, 475, 135-145. [8] DIGHIERO,G., LYMBERI,P., HOLMBERG,D., LUNDQUIST,I., COUTINHO,A. & AVRAMEAS, S., High frequency of natural autoantibodies in normal newborn mice. J. lmmunol., 1985, 134, 765-771. [9] DIOHIERO,G., LYMBERI,P., MAZn~, J.C., ROUVRE, S., BUTLER-BROWNE,G.S., WHALEN, R.G. & A v e , S., Murine hybridomas secreting natural monoclonal antibodies reacting with self antigens. J. Immunol., 1983, 131, 2267-2272. [i0] GP..ABAR,P., Autoantibodies and the physiological role of im~unoglobulins. Immunol. Today, 1983, 4, 337-340. [11] GUILBERT,B., DIGHIERO,G. & AVRAMEAS,S., Naturally occurring antibodies • against nine common antigens in normal human sera. - - I. Detection, isolation, and characterization. J. Immunol., 1982, 128, 2779-2787. [12] GUILBERT, B., MAHANA, W., GILBERT, M., MAZII~, J.C. & AVRAMEAS, S., Presence of natural autoantibodies in hyperimmunized mice. Immunology, 1985, 56, 401-408. [13] HEYMAN, B. & WIGZELL, H., Immunoregulation by monoclonal sheep erythrocyte-specific IgG antibodies: suppression is correlated to level of antigen binding and not to isotype. J. Iraraunol., 1984, 132, 1136-1143. [14] JAFFE,B.M., EISEN,I-I.M., SIMMS,E.S. & POTTER,M., Myeloma proteins with anti-hapten antibody activity E-2-4-dinitrophenyl binding by the protein produced by mouse plasmocytoma MOPC 460. J. Immunol., 1969, 103, 872-877. [15] KARSENTI,E., GUILBERT, B., BORNENS,M. & AVRAMEAS,S., Antibodies to tubulin in normal non-immunized animals. Proc. nat./tcad. Sci. (Wash.), 1977, 74, 3997-4001. [16] LITTLE, J.R. & EISEN, H.N., Preparation and characterization of antibodies specific for the 2-4-6-trinitrophenyl group. Biochemistry, 1966, 5, 3385-3389. [17] LYMaERI,P., DIOHIERO, G., T~RNYNCt:,T. & AVRAMEAS,S., A high incidence of cross-reactive idiotypes among murine natural autoantibodies. Europ. J. Iramunol., 1985, 15, 702-707. [18] PARHAM,P., ANDRt~LEWlCZ,M.J., BRODSKY,F.M., HOLMES,N.J. & WAYS,J.P., Monoclonal antibodies: purification, fragmentation and application to structurai and functional studies of class. - - I. MHC antigens. J. Immunol. Methods, 1982, 53, 133-173. [19] PLESCIA, O.J., BRAUN, W. & PALCZUK,N.C., Production of antibodies to denatured deoxyribonucleic acid (DNA). Biochemistry, 1964, 52, 279-286.

[20] ROSENBLATI',H.M., PARIKH,N., MCCLURE, J.E., MEZA, I., Hwo, S., BRYAN, J. & SI-mAR~R,W.T., Mitogen-like monoclonal anti-actin antibodies. J. Iramunol., 1985, 13S, 995-1000.

360

W. M A H A N A A N D COLL.

[21] SPUDISH,J.A. & WATT, S., The regulation of rabbit skeletal muscle contrac-

tion. J. biol. Chem., 1971, 246, 4866-4870. [22] TEm~YNCK,T. & AWAMEAS,S., Murine natural monoclonal autoantibodies: a study of their polyspecificities and their affinities. Immunol. Rev., 1986, 94, 99-112. [23] VAKIL,M. & KEARNEY,J.F., F fictional characterization of monoctonal autoantiidiotype antibodies isola. ~d from the early B-cell repertoire of BALB/c mice. Europ. J. Immunol., i986, 16, 1151-1158. [24] VAKIL, M., SAUTER, H., PAIGE, C. • KEARMEY,J.F., In vivo suppression of perinatal multispecific B cells results in a distortion of the adult B-cell repertoire. Europ. d. ImmunoL, 1986, 16, 1159-1165. [25] WHAL~N, R~G., BU~,LFR-BRow~E,G.S. & GROS, F., Identification of a novel form of myosin light chain present in embryonic muscle tissue and cultured muscle cells. J. tool. Biol., 1978, 126, 415-420.