E L S E V I E R Molecular Brain Research 29 (1995) 369-375

MOLECULAR BRAIN

RESEARCH

Short communication

Evidence for the presence of/33-adrenergic receptor mRNA in the human brain

Marianne Rodriguez a, Christine Carillon a, Antoine Coquerel b, G6rard Le Fur a, Pascual Ferrara a, Daniel Caput a,,, David Shire a

a Sanofi Recherche, Lab~ge-lnnopole Voie No. 1 BP 137, 31676 Labkge Cedex, France h Laboratoire de Radioanalyse et Hormonologie, H~pital Charles-NieoUe, 76031 Rouen Cedex, France

Accepted 22 November 1994

Abstract

The /33-adrenergic receptor (AR) is widely distributed in peripheral tissues, but up to now it has not been detected in the central nervous system. By using the polymerase chain reaction (PCR) technique, we found the/33-AR mRNA to be present in all the regions of the human brain we investigated. The quantities found were very low compared to those of the /31-AR and /32-AR mRNAs, being hardly detectable in adult brain. In contrast, the brain of very young infants contained about 100 times more/33-AR mRNA than the adult brain, whereas the amounts of/31-AR and /32-AR transcripts were essentially the same. In addition, using PCR we have cloned a central/33-AR coding region from a human frontal cortex cDNA library and have found it to be identical to the corresponding peripheral sequence.

Keywords: Adrenergic/33 receptor; Human brain; PCR; Quantitative PCR; Cloning; Sequence analysis

/3-adrenergic receptors (/3-AR) mediate the physio- logical effects of the catecholamines by activating adenylyl cyclase via the stimulatory Gs protein. Three subtypes of /3-AR receptors, /31-AR, /32-AR and /~3- AR have been identified on the basis of their phar- macological properties [31], physiological effects [6,23] and genetic structure [7,9,12,16,21,30]. The widespread, if heterogeneous, distribution of /31-AR and /32-AR throughout both the human [14] and rodent [25] brain is well-established.

The /33-AR has been detected in a wide range of peripheral tissues [18], compatible with its putative role in the regulation of fat metabolism and gastrointestinal functions (see Ref. [6] for a review). Although antide- pressant-like effects of the /33-AR selective agonist SR 58611A have been described in a rodent model [28], pointing to the presence of the /33-AR in the brain, no direct evidence for its existence in the central nervous system has been presented up to now. On the contrary, the sensitive RNase protection assay failed to detect /33-AR mRNA in rat cerebral cortex [11] and the

* Corresponding author. Fax: (33) 61 39 86 37.

0169-328X/95/$09.50 © 1995 Elsevier Science B.V. All rights reserved SSDI 0169-328X(94)00274-6

receptor has also been reported not to be expressed in the human brain [8]. We reasoned that if a central /33-AR indeed existed, then it would only be present at very low levels and would, perhaps, be highly localized. Direct localizations using autoradiographic techniques are difficult to undertake because of the absence of suitable radiolabelled ligands specific for this receptor.

We therefore investigated the presence of /33-AR mRNA in various human brain compartments derived from both young infants and adults using PCR and reverse transcriptase (RT) PCR techniques to amplify a supposedly weak signal. Two sets of primers (Table 1) based on the published human /33-AR eDNA se- quence [21] were used in PCR experiments with cDNA isolated from eight commercial lambda eDNA li- braries: human brain stem, temporal cortex, cerebel- lum, hippocampus (all from 2-year-old females), frontal cortex (female, 85 years), occipital cortex (female, 58 years), cauda te /pu tamen (males, 57 and 63 years) and substantia nigra (male and female, 60 years). One primer set, /33-AR(1), consisted of a sense primer for the start of the coding region (primer ls in Fig. la) coupled with an antisense primer at the end of the coding region (primer la), corresponding to the six

370 M. Rodriguez et al. /Molecular Brain Research 29 (1995) 369-375

codons which follow the unique splice junction in the sequence [12,21,30]. We carried out 35 cycles of PCR using 100 ng of each primer with each cDNA library (200 ng) or genomic DNA (400 ng) in 50 izl total volume containing 10 X Vent buffer (5 tzl), DMSO (5 /zl) dNTPs (200 /xM each) and Vent exo (1 izl) with denaturation at 94°C for 1 min, hybridization at 55°C for 1 min and polymerization at 72°C for 1.7 min. Amplicons of the expected 1.2 kb were obtained from each library (Fig. lb, upper panel). Southern blotting with an internal, specific peroxidase-labeled oligo- nucleotide probe (Table 1) using previously described conditions [2] indicated the authenticity of the ampli- cons (Fig. lb, lower panel). We failed to obtain a 2.3 kb amplicon from genomic DNA in this experiment. How- ever, in an experiment using a sense primer just up- stream of the splice junction (primer 2s in Fig. la), an antisense primer in the 3'-non-translated region (primer 2a, set /33-AR(2)) and the above PCR conditions, a band of 1.5 kb revealed by ethidium bromide staining was obtained from genomic DNA (Fig. lc, upper panel). Southern blotting used the same 5'-peroxidase /33-AR internal probe, but Y-end labeled with [a-32p]dCTP, and followed the described conditions [2], apart from a final wash with 2 x SSC, 0.1% SDS at 50°C (near the Tm), before exposure to X-Omat AR film with an intensifying screen for 6 h at -80°C (Fig. lc, lower panel). Again, from each of the eight cDNA libraries positive bands of the expected 442 bp were obtained on amplification (Fig. lc). It is important to note that these experiments are strictly qualitative, since cDNA libraries are not a fair representation of the corre- sponding RNA population.

Using primer set /33-AR(1), the /33-AR coding re- gion was amplified from the cDNA prepared from a lambda frontal cortex library and cloned. Two clones

were dideoxy sequenced [26]; each contained several random point mutations due to polymerase errors dur- ing amplification (results not shown). However, overall, the sequence was identical to that of the alternatively spliced form of the peripheral /33-AR cDNA contain- ing the additional six amino acids at the 3' extremity of the coding region [12,21,30], recalling that the 2a primer was designed to recognize this sequence. These results indicate both a widespread distribution of /33-AR mRNA throughout the human brain and the identity of this alternatively spliced form with that found in the peripheral regions. We do not know whether other spliced forms exist in the brain.

We were next interested in comparing /31, /32 and /33-AR mRNA levels in brain preparations from the same individuals. RNA was isolated from normal hu- man tissue (pooled donors 1 and 2: female infants, 4 and 6 months old; post-mortem interval before freezing 12 h; donor 3: male, 19 years old; post-mortem interval before freezing; 12 h) using the guanidinium isothio- cyanate procedure [4]. All samples were chromato- graphed on Biogel P10 and exhaustively treated with DNase to eliminate genomic DNA before reverse tran- scription primed by oligo dT. By selecting primers on either side of the splice junction (primer set/33-AR(2)), we were able to verify that the samples were indeed free of DNA contamination (Fig. 2c). The cDNA was also amplified with primer sets /31-AR(1) and /32- AR(1) (Table 1) using the above PCR conditions, but for only 30 cycles. Transcripts for/31-AR (Fig. 2a) and /32-AR (Fig. 2b) were present in all the brain regions tested,/32-AR transcripts showing much more variabil- ity than those for /31-AR. Very low levels of /32-AR mRNA were present in the infant parietal cortex and cerebellum and lower levels were present in the adult temporal and occipital cortex than in the correspond-

Table 1 Sequences of the oligonucleotide primers and probes described in the text The (Q) signifies priming sites found also in the multi-specific internal control RNA. The bracketed figures immediately following the sequences are the 5' positions of each oligonucleotide on the relevant coding strand in the GenEMBL library

mRNA Accession No. Sense primer Antisense primer Amplicon size bp

5' 3' 5' 3'

31-AR(Q) J03019 TCGTCTTCTI'CAACTGGCTG (1162) CACCTGCTATCGGTCTTAAG (1584) 442 31-AR(1) J03019 TGCATCATGGCCTTCGTGTAC (798) CAGCCAGTTGAAGAAGACGA (1162) 384 32-AR(Q) Ml5169 ATCCAGGATAACCTCATCCG (2479) GTTATCGCTAGGCACAGTAC (2762) 303 32-AR(1) Ml5169 GGAAGCCATCAACTGCTATG (2124) CAATCCTGAAATCTGGGCTC (2571) 465 33-AR(Q) X70811 CTGAACTGGCTAGGTTATGC (1116) GGGACTCATTCTGAACAGAG (1410) 314 33-AR(1) X70811 CCACACAAGCTTGCCACCATG CCACTCGGATCCCTAAGAAAC

GCTCCGTGGCCTCAC TCCCCAAGAAG GCGCAGCCCAGGCTTTGCCA AACACTAAGGTCTCATCAGGG AC (1299) TTTG

/33-AR(2) X70811

mRNA Accession No. Internal probe

~I-AR J03019 ~2-AR M15169 ~3-AR X70811

(ATG = 126) (1333) 1256

5 r 3 t

TCGCAGCTGTCGATCTTCTT (852) CAATCTTCTGGAGCTGCCTT (2268) A A G A G C A T C A C G A G A G A G G (771)

(1717) 441

M. Rodriguez et al. / Molecular Brain Research 29 (1995) 369-375 371

a

b

I

ATG

intron

l a * 2a polyA

- - 1.2

C

1.5

0.4

Fig. 1. PCR of human brain cDNA isolated from commercial AZap libraries (Stratagene). Panel a shows the positioning of two sets o f /33 -AR primers on the f l3-AR cDNA (not to scale), l s and l a encompass the entire 1.2 kb coding region (ATG to the stop codon *), incorporate restriction sites for cloning and, for ls, a Kozak consensus sequence [17]. The second set comprises 2s and 2a. The upper panels b and c show ethidium bromide stained 1% agarose gels of the amplified fragments. The lower panels are Southern blots probed with either a 5'-linked horse radish peroxidase internal probe (panel lb) as previously described [2] or an [a-a2p]-labelled internal probe (panel lc) as described in the text.

372 M. Rodriguez et al. / Molecular Brain Research 29 (1995) 369-375

a

---442bp

b

~465bp

C

--441bp

..... . . . . . . . . . . . . . . . . . . . . . . . .

Fig. 2. RT-PCR of pooled total RNA isolated from brain sections of two infants (i) and of total RNA from brain sections of a single adult (a). The upper panels show ethidium bromide stained 1% agarose gels of the amplified fragments, using the primers in Table 1, panel a: fll-AR(1), panel b: fl2-AR(1) and panel c: /33-AR(2). The lower panels are Southern blots hybridized with Y-linked horse radish peroxidase internal specific probes as previously described [2].

M. Rodriguez et al. / Molecular Brain Research 29 (1995) 369-375 373

ing infant tissues. Similarly, lower levels of /33-AR transcripts were found in the adult brain tissues than in those of the infant. Age-related changes in quantity and pattern of distribution of the /31-AR and /32-AR have been reported in the cat [27], mouse [10] and human [15] brains and this, therefore, also applies to the central /33-AR.

We next attempted to measure the relative /31//32//33-AR mRNA levels in the total RNA from the pooled 4 adult brain regions and the pooled 6 infant regions. Since PCR in itself is not a quantitative method, we quantitated the transcripts using an RT- PCR technique which involved spiking the cellular RNA with a multi-specific internal control [20]. The control RNA contained a tandem array of priming sites for each of the/3-AR, identical to those on the cellular targets, only the length of the fragment to be amplified differed (Table 1), so allowing easy gel separation of the two amplicons. As before [20], after spiking each of the total RNA samples (300 ng) with known quantities of the internal standard (see legend to Fig. 3), the RNA mixture was co-reverse transcribed using oligo dT priming, then separated into aliquots. A specific /3-AR primer pair (labeled Q in Table 1) was added to each aliquot and PCR was carried out in the presence of [a-a2p]dCTP. The amplicons were separated by polyacrylamide gel electrophoresis (Fig. 3) and quanti- tated as previously described [20]. Overall, /33-AR mRNA levels were lower by a factor of 100 in the adult compared to the infant, whereas /31-AR mRNA levels were about 2-fold lower and /32-AR mRNA remained unchanged.

Unfortunately, upon comparing relative/31-AR//33- AR mRNA quantities in SK-N-MC cells, a human neuroblastoma cell line expressing /31-AR and /33-AR transcripts [8], both by Northern analysis and by quan- titative PCR (unpublished results), it was clear that the PCR method underestimated /31-AR mRNA by a fac- tor of 15. After experimentation, we attributed the discrepancy to a failure to completely denature the /31-AR cDNA during the PCR, this unusual stability being due to a high G / C content (74%). Raising the denaturing temperature to 96°C in the presence of a

temperature-stable DNA polymerase only improved results slightly. Other investigators have reported diffi- culties with PCR of /31-AR cDNA [3,18,24], and, al- though we found that the presence of 10% DMSO was necessary for amplification, we also found it was insuf- ficient to fully denature the cDNA. This failure meant that it was impossible to compare /31-AR and /33-AR mRNA levels directly by RT-PCR.

Nevertheless, fl2-AR and /33-AR cDNA amplifica- tions were not enhanced by high denaturation temper- atures and high DMSO concentrations, consistent with their lower C / G contents. We were able, therefore, to compare the relative /32//33 mRNA contents in the two pooled brain tissue RNAs. Brain tissues from the 4- and 6-month-old infants contained about 50-fold less /33-AR mRNA than /32-AR mRNA, the difference rising to at least 3000-fold in the adult, where /33-AR mRNA was too rare to be measurable (Fig. 3). Since in the adult cortex and substantia nigra the/31-AR is 2-4 times more abundant than the /32-AR at the protein level [14], it can be concluded that the/33-AR is indeed very rare in overall terms. The very low mRNA levels for the /33-AR would explain why its presence has hitherto escaped detection [8,11]. However, it must be emphasized that the infant brains contained about 10(I times more /33-AR mRNA than the adult brains. Sev- eral questions arise from these findings, particularly in view of the possible connection between the elevated catecholamine concentrations found in the blood of newborn babies, attributed to birth stress [19], and the up-regulation of /33-AR expression by high agonist concentrations [29]. Do newborn baby brains contain high levels of the /33-AR and, if so, in which brain compartments? Are the levels we find here normal or are they related to the fact that the two infants were cot-death victims and were possibly submitted to stress before death? Further investigations are clearly mer- ited.

In the present investigation we have simply deter- mined the presence of/33-AR mRNA in certain brain compartments and the next step is obviously to investi- gate whether certain structures or cell types in the brain are particularly rich in /33-AR and what, if any,

131-AR 132-AR 133-AR adult infant adult infant adult infant

1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5

Fig. 3. Phosphorlmage of RT-PCR quantitation of /3-AR mRNAs in human brain. Amplicons were separated by 10% PAGE [20]. The amount of internal standard added was lanes 1:12.5 fg; lanes 2 :6 fg; lanes 3 :3 fg; lanes 4:1.5 fg; lanes 6:0.75 fg. The upper amplicons are those from the cellular RNA (sizes given in Table 1), the lower amplicons are from the standard RNA (204 bp).

374 M. Rodriguez et al. / Molecular Brain Research 29 (1995) 369-375

its central role is. The distribution of the fl-AR is of considerable medical and therapeutic importance since these receptors have been directly implicated in a number of mental disorders. In particular, several of these afflictions correlate either with imbalances in relative fl-AR quantities or to overall fl-AR levels. The involvement of noradrenaline in schizophrenia has been suggested [13] and altered noradrenergic input may account for the changes in the 131-AR/fl2-AR balance shown to occur in the hippocampus of schizophrenic cases [14]. These results were obtained from autoradiographic studies using displacement of [125I]iodopindolol by selective fll-AR and fl2-AR an- tagonists. Binding studies with [lzsI]iodopindolol on particulate membrane preparations from Alzheimer's disease subjects again indicated selective, regiospecific changes in the fll-AR/fl2-AR balance [15]. Depres- sion in suicide victims has been associated with deficits in fl-AR sites, largely in cortical areas, quantitated by saturation binding with [3H]CGP 12177A [5]. Similar studies to investigate a possible central role of the fl3-AR are hampered by the absence of suitable radio- labeled ligands specific for this atypical receptor. In- deed, it is only recently that a systematic study of fl-AR agonists and antagonists has been undertaken on recombinant human receptors in CHO cells in order to define the complicated binding characteristics and pharmacological properties of these ligands [1]. It is significant that many fll-AR and fl2-AR antagonists are /33-AR agonists [1]. There are studies which show that /3-AR blockers may improve psychosis [22], whereas fl-AR agonists may be useful in treating de- pression [28]. It is, therefore, clearly of importance for rational drug design to establish precisely the distribu- tion and the regulation of each of the /3-AR subtypes in the central nervous system.

[1] Blin, N., Camoin, L., Maigret, B. and Strosberg, A.D., Struc- tural and conformational features determining selective signal transduction in the/33-adrenergic receptor, Mol. Pharmacol., 44 (1993) 1094-1104.

[2] Bouaboula, M., Legoux, P., Pess6gu6, B., Delpech, B., Dumont, X., Piechaczyk, M., Casellas, P. and Shire, D., Quantitation of cytokine gene expression using a polymerase chain reaction method involving co-amplification with an internal multi-specific control, J. Biol. Chem., 267 (1992) 21830-21838.

[3] Bristow, M.R., Minobe, W.A., Raynolds, M.V., Port, J.D., Ras- mussen, R., Ray, P.E. and Feldman, A.M., Reduced /31 recep- tor messenger RNA abundance in the failing human heart, J. Clin. Incest., 92 (1993) 2737-2745.

[4] Chomczynski, P. and Sacchi, N., Single-step method of RNA isolation by acid guanidium thiocyanate-phenol-chloroform ex- traction, Anal. Biochem., 162 (1987) 156-159.

[5] De Paermentier, F., Cheetham, S.C., Crompton, M.R., Katona, C.L. and Horton, R.W., Brain beta-adrenoceptor binding sites in antidepressant-free depressed suicide victims, Brain Res., 525 (1990) 71-77.

[6] Emorine, L., Blin, N. and Strosberg, A.D., The human /33-

adrenoceptor: the search for a physiological function, Trends Pharmacol. Sci., 15 (1994) 3-7.

[7] Emorine, LJ., Marullo, S., Briend-Sutren, M.-M., Patey, G., Tate, K., Delavier-Klutchko, C. and Strosberg, A.D., Molecular characterization of the human beta3-adrenergic receptor, Sci- ence, 245 (1989) 1118-1121.

[8] Esbenshade, T.A., Han, C., Theroux, T.L., Granneman, J.G. and Minneman, K.P., Coexisting /31- and atypical /3-adrenergic receptors cause redundant increases in cyclic AMP in human neuroblastoma cells, Mol. Pharmacol., 42 (1992) 753-759.

[9] Frielle, T., Collins, S., Daniel, K.W., Caron, M.G., Lefkowitz, R.J. and Kobilka, B.K., Cloning of the cDNA for the human /31-adrenergic receptor, Proc. Natl. Acad. Sci. USA, 84 (1987) 7920-7924.

[10] Goffinet, A.M., Hemmendinger, L.M. and Caviness, V.S., Auto- radiographic study of beta-1 adrenergic receptor development in the mouse forebrain, Brain Res., 389 (1986) 187-191.

[11] Granneman, J.G., Lahners, K.N. and Chaudhry, A., Molecular cloning and expression of the rat /33-adrenergic receptor, Mol. Pharmacol., 40 (1991) 895-899.

[12] Granneman, J.G., Lahners, K.N. and Rao, D.D., Rodent and human /33-adrenergic receptor genes contain an intron within the protein-coding block, Mol. Pharmacol., 42 (1992) 964-970.

[13] Hornykiewicz, O., Brain catecholamines in schizophrenia - a good case for noradrenaline, Nature, 299 (1982) 484-485.

[14] Joyce, J.N., Lexow, N., Kim, S.J., Artymyshyn, R., Senzon, S., Lawerence, D., Cassanova, M.F., Kleinman, J.E., Bird, E.D. and Winokur, A., Distribution of beta-adrenergic receptor subtypes in human post-mortem brain: alterations in limbic regions of schizophrenics, Synapse, 10 (1992) 228-246.

[15] Kalaria, R.N., Andorn, A.C., Tabaton, M., Whitehouse, P.J., Harik, S.I. and Unnerstall, J.R., Adrenergic receptors in aging and Alzheimer's disease: increased beta 2-receptors in pre- frontal cortex and hippocampus, J. Neurochem., 53 (1989) 1772- 1781.

[16] Kobilka, B.K., Frielle, T., Dohlman, H.G., Bolanowski, M.A., Dixon, R.A.F., Keller, P., Caron, M.G. and Lefkowitz, R.J., Delineation of the intronless nature of the genes for the human and hamster /32-adrenergic receptor and their putative pro- moter regions, J. Biol. Chem., 262 (1987) 7321-7327.

[17] Kozak, M., Compilation and analysis of sequences upstream from the translational start site in eukaryotic mRNAs, Nucleic Acids Res., 12 (1984) 857-872.

[18] Krief, S., L6nnqvist, F., Raimbault, S., Baude, B., Van Spron- sen, A., Arner, P., Strosberg, A.D., Ricquier, D. and Emorine, L.J., Tissue distribution of /33-adrenergic receptor mRNA in man, Z Clin. Invest., 91 (1993) 344-349.

[19] Lagerkrantz, H. and Bistolette, P., Catecholamine release in the newborn infant at birth, Pediatric. Res., 11 (1973) 889-893.

[20] Legoux, P., Minty, C., Delpech, B., Minty, A.J. and Shire, D., Simultaneous quantitation of cytokine mRNAs in interleukin-1/3 stimulated U373 human astrocytoma cells by a polymerisation chain reaction method involving co-amplification with an inter- nal multi-specific control, Eur. Cytokine Netw., 3 (1992) 553-563.

[21] Lelias, J.M., Kaghad, M., Rodriguez, M., Chalon, P., Bonnin, J., Dupre, I., Delpech, B., Bensaid, M., Le Fur, G., Ferrara, P. and Caput, D., Molecular cloning of a human /33-adrenergic recep- tor cDNA, FEBS Lett., 324 (1993) 127-130.

[22] Lipinski, J.F., Keck, P.E. and McElroy, S.L., /3-Adrenergic antagonists in psychosis: is improvement due to treatment of neuroleptic-induced akathisia?, J. Clin. Psychopharmacol., 8 (1988) 409-416.

[23] Lohse, M.J., Strasser, R.H. and Helmreich, E.J.M., The /3- adrenoceptors. In F. Hucho (Ed.), Neurotransmitter Receptors, Elsevier, Amsterdam, 1993, pp. 137-180.

[24] L6nnqvist, F., Krief, S., Strosberg, A.D., Nyberg, B., Emorine,

M. Rodriguez et al. / Molecular Brain Research 29 (1995) 369-375 375

L.J. and Arner, P., Evidence for a functional fl3-adrenoceptor in man, Br. J. Pharmacol., 110 (1993) 929-936.

[25] Rainbow, T.C., Parsons, B. and Wolfe, B.B., Quantitative auto- radiography of ill- and fi2-adrenergic receptors in rat brain, Proc. Natl. Acad. Sci. USA, 81 (1984) 1585-1589.

[26] Sanger, F.S.. Nicklen, A.R. and Coulson, A.R., DNA sequenc- ing with chain termination inhibitors, Proc. Natl. Acad. Sci. USA, 74 (1977) 5463-5467.

[27] Shaw, C.. Needler, M.C., Wilkinson, M., Aoki, C. and Cynader, M., Alterations in receptor number, affinity and laminar distri- bution in cat visual cortex during the critical period, Prog. Neuropsychopharmacol. Biol. Psychiatry, 8 (1984) 627-634.

[28] Simiand, J., Keane, P.E., Guitard, J., Langlois, X., Gonalons, N., Martin, P., Bianchetti, A., Le Fur, G. and Soubri6, P., Antidepressant profile in rodents of SR 58611A, a new selective

agonist for atypical fl-adrenoceptors, Eur. J. PharmacoL, 219 (1992) 193-201.

[29] Thomas, R.F., Holt, B.D., Schwimm, D.A. and Liggett, S.B., Long-term agonist exposure induces up-regulation of beta 3- adrenergic receptor expression via multiple cAMP response elements, Proc. Natl. Acad. Sci. USA, 89 (1992) 4490-4494.

[30] Van Spronsen, A., Nahmias, C., Krief. S., Briend-Sutren, M.-M., Strosberg, A.D. and Emorine, L.J., The promoter and intron/exon structure of the human and mouse /33-adrenergic- receptor genes, Eur. J. Biochem., 213 (1993) I 117-1124.

[31] Zaagsma, J. and Hollenga, C., Distribution and function of atypical /33-adrenoceptors. In E. Szabadi and C.M. Bradshaw (Eds.), Adrenoceptors: Structure, Mechanisms, Function. Birkh~iuser Verlag, Basel, 1991, pp. 47-58.