Content uploaded by Robert Root-Bernstein
Author content
All content in this area was uploaded by Robert Root-Bernstein on Jan 11, 2018
Content may be subject to copyright.
J. theor. Biol. (2002) 218, 71–84
doi:10.1006/yjtbi.3056, available online at http://www.idealibrary.com on
Molecular Complementarity III. Peptide Complementarity as a Basis
for Peptide Receptor Evolution: A Bioinformatic Case Study of
Insulin, Glucagon and Gastrin
Rob ert Root-Ber nstei n* w
wDepartment of Physiology,Biophysical Sciences Building,Michigan State University,
East Lansing,MI 48824, U.S.A.
(Received on 11 February 2002, Accepted in revised form on 15 April 2002)
Dwyer has suggested that peptide receptors evolved from self-aggregating peptides so that
peptide receptors should incorporate regions of high homology with the peptide ligand. If
one considers self-aggregation to be a particular manifestation of molecular complementarity
in general, then it is possible to extend Dwyer’s hypothesis to a broader set of peptides:
complementary peptides that bind to each other. In the latter case, one would expect to find
homologous copies of the complementary peptide in the receptor. Thirteen peptides, 10 of
which are not known to self-aggregate (amylin, ACTH, LHRH, angiotensin II, atrial
natriuretic peptide, somatostatin, oxytocin, neurotensin, vasopressin, and substance P), and
three that are known to self-aggregate (insulin, glucagon, and gastrin), were chosen. In
addition to being self-aggregating, insulin and glucagon are also known to bind to each other,
making them a mutually complementary pair. All possible combinations of the 13 peptides
and the extracellular regions of their receptors were investigated using bioinformatic tools (a
total of 325 combinations). Multiple, statistically significant homologies were found for
insulin in the insulin receptor; insulin in the glucagon receptor; glucagon in the glucagon
receptor; glucagon in the insulin receptor; and gastrin in gastrin binding protein and its
receptor. Most of these homologies are in regions or sequences known to contribute to
receptor binding of the respective hormone. These results suggest that the Dwyer hypothesis
for receptor evolution may be generalizable beyond self-aggregating to complementary
peptides. The evolution of receptors may have been driven by small molecule complemen-
tarity augmented by modular evolutionary processes that left a ‘‘molecular paleontology’’
that is still evident in the genome today. This ‘‘paleontology’’ may allow identification of
peptide receptor sites.
r2002 Elsevier Science Ltd. All rights reserved.
Introduction
The origins of receptor–ligand pairs remain an
evolutionary mystery. Have ligands and their
receptors evolved randomly and independently
so that the particular function such a pair plays
is purely up to chance? Or are there guiding
principles that limit the range of possibilities so
that specific ligands and receptors co-evolve
deterministically? Various theories have been
presented. This paper provides a new possibility
based on molecular complementarity and mod-
ular evolutionary processes.
It is possible to imagine that ligands and
receptors evolved independently, finding their
*Corresponding author. Tel.: 517-355-6475-1101; fax:
517-432-1967.
E-mail address: rootbern@msu.edu (R. Root-Bernstein).
0022-5193/02/ $35.00/0 r2002 Elsevier Science Ltd. All rights reserved.
specific functions by chance as random muta-
tions resulted in complementary structures. Such
probabilistic evolution would yield random
assignments of function. For example, the fact
that insulin and glucagon are the two major
hormones regulating sugar metabolism would
be a result of pure chance. There should be no
obvious structural relationship between them or
between their sequences as peptides and the
protein sequences found in their receptors. Any
similarities or structural relationships between
ligands and their receptors should occur at best
by chance.
Random theories of molecular evolution are
not amenable to direct experimental inquiry
since they do not make any specific predictions
to test. Such theories provide, in essence, a null
hypothesis against which to test other possibi-
lities. One alternative is that receptor evolution
has been deterministic in nature. Blalock, for
example, has proposed a deterministic theory of
receptor–ligand evolution based upon his ‘‘mo-
lecular recognition hypothesis’’ (MRH). The
MRH suggests that receptor sequences are
encoded in the 50-30direction by the comple-
mentary or antisense strand of DNA (cDNA)
that corresponds to that of the DNA encoding
the ligand (Carr et al., 1986). While the MRH
theory has generated some apparently confirma-
tory data (Ruiz-Opazo et al., 1995; Jarpe and
Blalock, 1994; Tropsha et al., 1992; Gorcs et al.,
1986), independent tests of MRH have demon-
strated that the proteins isolated using the
hypothesis are not peptide receptors. (Jurzak
et al., 1993; Eberle & Huber, 1991; Kelly et al.,
1990; DeGasparo et al., 1989; Rasmussen &
Hesch, 1987). Moreover, every physicochemical
test of the MRH has demonstrated that there is
no interaction between peptide ligands and their
putative peptide receptor sequences (reviewed in
Root-Bernstein & Holsworth, 1998) and the
presence of such antisense peptide sequences in
peptide receptors has been shown to occur no
more often than would be predicted by chance
(Goldstein & Brutlag, 1989). Thus, the MRH
has little to recommend it at present.
A more compelling directed evolution theory
has been presented by Dwyer, who suggested
that receptor–ligand systems may have evolved
from self-aggregating (or self-complementary)
peptide complexes. He has reported that alpha
bungarotoxin self-aggregates and that its recep-
tor, the acetylcholine receptor, has a peptide
sequence at the bungarotoxin binding site that
mimics the self-aggregating region of the toxin
(Dwyer, 1989, 1998). He argues from this
observation that self-aggregating peptides and
proteins may be the basis from which their
receptors evolved. Dwyer also reports that two
other self-aggregating peptides share homologies
with their receptors: alpha scorpion toxins mimic
regions of the sodium channel proteins they
block, and interleukin 2 (IL-2) shares regions of
homology with the IL-2 receptor (Dwyer, 1989).
I suggest here an extension of Dwyer’s theory
based upon a broader concept of peptide
complementarity. Root-Bernstein & Dillon
(1997) recently suggested that one of the driving
forces of evolution is molecular complementar-
ity. A prediction that follows from the theory
is that molecular functions are not randomly
assigned to molecules by chance, but evolve from
stereospecific chemical interactions. More pre-
cisely, primitive metabolism may have been
regulated by interactions between complemen-
tary compounds within a synthetic pathway so
that the relative concentrations of free vs.
complexed molecules would determine metabolic
rate. Thus, the opposing physiological effects of
insulin and glucagon on blood sugar regulation
are mirrored in the fact that they are molecularly
complementary and bind to one another. It
follows that the receptor–ligand interactions that
we observe in living things today may have
originated as complementary complexes to
which regulatory, transmembrane, and second-
messenger-interactive sequences have been
added.
A detailed proposal for peptide hormone
receptor evolution by means of complementarity
combined with modular swapping is presented
here. This paper provides theoretical and
experimental tests of both the Dwyer and
Root-Bernstein/Dillon (RBD) theories. Dwyer’s
theory predicts that peptide receptors will con-
tain ligand-like sequences within the binding
region, if the peptide is self-complementary. The
RBD theory predicts that if a peptide (A) has a
complementary peptide (B), then that comple-
mentary peptide sequence (B) will appear in the
R. ROOT-BERNSTEIN72
extracellular binding region of the peptide (A)
receptor. These predictions are confirmed for the
cases of gastrin, insulin, and glucagon.
Methods
A bioinformatic approach to testing the RBD
and Dwyer theories was performed using 13
peptides with which this lab is familiar from
previous experimental work: insulin (INS),
glucagon (GLU), somatostatin (SOM), amylin
(AMY), adrenocorticotropic hormone (ACTH),
angiotensin II (AII), luteinizing-hormone-releas-
ing hormone (LHRH), substance P (SP), atrial
natriuretic peptide (ANP), vasopressin (VP),
oxytocin (OX), neurotensin (NT), and gastrin.
Insulin is well known to self-aggregate in both
dimer and hexamer forms (Dathe et al., 1990;
Lougheed et al., 1983). Glucagon self-aggregates
into dimers, trimers and hexamers (Chou &
Fasman, 1975; Johnson et al., 1979; Kim et al.,
2000). Gastrin forms dimers (Attwood et al.,
1974). Thus, insulin, glucagon, and gastrin satisfy
the criteria for testing Dwyer’s theory. Insulin and
glucagon are also mutually complementary, bind-
ing to each other with micromolar affinity (Bloch,
1994; Root-Bernstein & Dobbelstein, 2001). Thus,
the insulin–glucagon pair satisfies the criteria for
the RBD theory. None of the other peptides
studied here are known to bind to themselves or
to each other, and research in our lab has shown
that some definitely do not (Root-Bernstein &
Westall, 1986; Root-Bernstein & Dobbelstein,
2001). Thus, the peptides selected for study
include several instances that should satisfy both
the Dwyer and RBD predictions concerning
peptide (self ) complementary sequences within
the respective receptor binding sites. The selection
also includes a very large number of what are
predicted to be negative controls.
Predictions were tested by analysing the
peptide receptors for the presence of sequences
similar to any of the 13 peptides in the study.
Two glucose transporters (GLUT 2 and GLUT
4) and angiotensin converting enzyme (ACE)
were also added to the ‘‘receptor’’ category as
controls for the glucose-regulating hormones
and for AII. Neurophysin was included in
the peptide receptor category since it was
possible that it might mimic receptor sequences
or ligand sequences, or both. In all cases, human
peptide and receptor sequences were used for the
analysis. Some receptors exist in more than one
form and where these sequence variations were
available in the SWISSPROT protein data base
as of December 2001, all were examined. The
SWISSPROT accession numbers are as follows:
ACTH receptor Q01718; AII receptor Q13725;
angiotensin converting enzyme P22966; atrial
natriuretic peptide receptor A P16066; ANP
receptor B P20594; ANP receptor C P17342;
gastrin binding protein Q16679; gastrin receptor
B P32239; glucagon receptor P47971; glucose
transporter 2 P11168; glucose transporter 4
P14672; insulin receptor P06213; LHRH
(GnRH) receptor P30968; neurophysin 1
P01178; neurophysin 2 P01185; neurotensin
receptor 1 P30989; oxytocin receptor P30559;
somatostatin receptor 1 P30872; somatostatin
receptor 2 P30874; somatostatin receptor 3
P32745; somatostatin receptor 4 P31391; soma-
tostatin receptor 5 P35346; substance P receptor
P25103; vasopressin receptor V1a Q62463;
vasopressin receptor V1b P47901.
Sequence similarities were determined using
LALIGN (Huang & Miller, 1991) using default
parameters for scoring (penalty for the first
residue in a gap, 14; penalty for each addi-
tional residue in a gap, 4). All alignments for
all combinations of the peptides and receptors
were then screened against the receptor structure
to eliminate matches to transmembrane and
intra-cytoplasmic regions, since these are not
likely to be involved in peptide binding.
Sequences were deemed to be significantly
homologous only if they had at least 50%
identity over a sequence of at least ten amino
acids and a raw score of at least 30. Ten amino-
acid-long sequences were chosen in the first place
because they represent a significant proportion
of all of the peptides analysed (between 100 and
25% of the total hormone length) and also
because ten amino acid stretches are approxi-
mately the length recognized by T-cell receptors
and are therefore of some biological significance
(Rudensky et al., 1991). Raw scores are used by
LALIGN to derive a function that calculates the
probability that an alignment with such a score
is likely to occur if any peptide of an equivalent
length were used to search 10, 000 proteins of
CASE STUDY OF INSULIN, GLUCAGON AND GASTRIN 73
equivalent length to the receptor sequence. In
general, raw scores below 20 corresponded to
probabilities of greater than 50% that the
similarity occurred by chance [i.e., p40.50] and
30 corresponded to probabilities that the simi-
larity occurred about 6–10% of the time by
chance [i.e., p¼0.06–0.10]. In occasional cases,
such as the gastrin comparisons, raw scores of 30
corresponded to pvalues of 0.05 or less. Raw
scores of 35 or above in this study always
corresponded to probabilities of less than 1 in 20
that such alignments occur by chance [po0.05].
Raw scores above 50 generally corresponded to
probabilities of less than 1 in 100 that such
alignments occur by chance [po0.01].
Results
Table 1 shows the highest percent identity and
its corresponding raw score for each peptide–
receptor comparison. Of 325 combinations of
peptides and receptors screened, only 11 matches
were found to be of statistical significance (i.e., at
least five homologous amino acids in a minimum
sequence of 10 and po0.05). These statistically
significant homologies are shown in bold, under-
lined large type in Table 1. Two clusters of
significant homologies were found. Insulin-like
sequences are found in both the extracellular
binding regions of both insulin and glucagon
receptors and glucagon-like sequences are found
in the extracellular binding regions of both
glucagon and insulin receptors. Gastrin-like
sequences also appear in the gastrin receptor
and gastrin binding protein.
What Table 1 does not show is that there are,
in most of these cases, not one but multiple,
significant peptide–receptor homologies. These
multiple homologies are displayed in Tables 2–6.
The sheer number of these multiple homologies
within a single receptor sequence is far in excess
of what would be predicted to occur by chance.
A particularly striking observation is that
many of the complementary sequences identified
in Tables 2–5 exist within regions of the
glucagon and insulin receptors that are asso-
ciated with ligand binding. Although the specific
receptor sequences involved in ligand binding to
the glucagon receptor and the insulin receptor
are currently unknown (Burcelin et al., 1996;
Ishida & Nagamatsu, 2001), significant data
exist linking some specific receptor regions or
sequences to the respective binding sites.
Glucagon binding to its receptor has been
localized to the membrane-proximal half of the
amino terminal extension of the receptor, thus
encompassing residues from about 100 to 230
(Buggy et al., 1996). This region includes almost
all of the sequences listed in both Tables 2 and 4
showing the existence of glucagon- and insulin-
like sequences in the glucagon receptor. More
specifically, glucagon binding to its receptor is
blocked by antibodies to amino acids 126–137
and 206–219, suggesting that these sequences are
either in or immediately adjacent to the glucagon
binding site (Unson et al., 1996). Sequence 126–
137 is immediately adjacent to sequence 139–50
listed in Table 2 (glucagon in glucagon receptor)
and 206–219 overlaps two additional sequences
in Table 2: 195–208 and 210–230. Sequence 206–
219 is also immediately adjacent to sequence
220–240 in Table 4 (insulin in glucagon recep-
tor). Thus, there is some probability that one or
more of the (self ) complementary regions listed
in Tables 2 and 4 are of real significance for
understanding the evolution of the glucagon
receptor binding site.
Information concerning the insulin receptor
binding site also suggests that some of the
homologies listed in Tables 3 and 5 play a role
in insulin binding. Insulin binds primarily to the
extracellular alpha subunit of the receptor and
both the cysteine-rich region (sequence 150–360)
and the N-terminal sequence (10–50) have been
implicated more specifically (Yip, 1992). Within
these regions, antibodies to the sequences 205–
316, and even more specifically 241–251, block
insulin binding (Yip et al., 1991). Notably, three
insulin-like sequences appear in the insulin
receptor in the 222–248 and 450–460 regions
listed in Table 5 (insulin in the insulin receptor).
In addition, sequences 157–173 and 284–304 in
Table 3 (glucagon in the insulin receptor) exist
within this broad binding-associated region.
Antibody blocking studies also implicate other
alpha subunit receptor regions as sites of insulin
binding, including 469–592 (Prigent et al., 1990)
and 650–758 (Fabry et al., 1992). The insulin-like
sequence 658–695 in Table 5 spans part of
the latter region. Finally, antibody-blocking
R. ROOT-BERNSTEIN74
Ta b l e 1
Homology scores for peptides vs. peptide receptors*
INS GLUC AMYL SOM ANP AII LHRH SUBP ACTH VP OXY NT GAST
INSULIN REC 60/50 60/57 40/27 40/28 40/29 30/21 30/20 40/26 40/30 30/32 20/26 50/28 30/28
GLUCAGON REC 60/35 50/36 30/25 30/29 30/28 30/24 20/15 30/22 40/22 20/17 30/26 30/22 20/31
GLUCOSE TRANS 2 20/20 40/28 30/24 20/18 30/24 20/13 20/15 20/16 40/26 20/17 30/27 30/15 60/42
GLUCOSE TRANS 4 20/20 40/29 30/18 30/20 30/23 20/13 20/15 20/17 40/30 20/17 40/25 30/20 40/22
SOM REC TYPE 1 40/34 30/19 40/23 40/35 30/30 20/18 20/20 30/27 40/37 30/25 30/25 30/26 30/21
SOM REC TYPE 2 50/35 20/21 30/25 30/25 40/22 20/24 30/36 20/21 30/29 40/24 30/24 40/26 30/23
SOM REC TYPE 3 40/25 30/18 30/24 40/22 40/25 30/23 20/28 20/20 40/43 40/31 30/28 30/20 30/24
SOM REC TYPE 4 40/30 30/22 30/24 50/29 40/30 20/25 30/36 30/22 40/31 30/25 30/23 30/26 30/29
SOM REC TYPE 5 70/31 30/21 30/24 40/21 60/32 20/25 20/26 30/26 40/37 30/25 20/23 30/27 40/29
ANP REC TYPE A 40/34 30/24 30/34 30/16 40/25 30/26 40/32 30/27 40/28 30/28 30/28 40/36 30/24
ANP REC TYPE B 40/28 60/45 40/29 30/26 30/22 30/20 40/30 30/23 40/30 30/21 40/20 30/26 40/31
ANP REC TYPE C 50/36 50/27 40/31 30/27 40/32 30/20 30//30 30/23 40/28 30/20 20/21 30/19 20/23
AII REC TYPE A1 30/23 40/25 40/28 30/21 30/27 30/21 20/26 30/28 30/33 20/16 30/30 30/19 30/23
ACE 40/31 30/28 30/22 20/26 50/31 30/25 20/25 40/30 40/37 30/25 20/26 30/25 30/29
LHRH REC 30/28 40/24 40/26 30/29 40/21 20/12 20/20 30/19 40/22 30/29 30/18 30/21 20/25
SUBSTANCE P REC 30/25 30/22 40/31 30/24 20/16 20/18 40/34 30/26 30/25 30/28 30/21 30/23 30/26
ACTH REC 30/24 50/33 30/31 20/15 30/16 20/15 20/18 20/14 30/21 30//32 20/25 30/26 30/20
VP REC V1A 30/25 30/20 40/25 30/27 40/23 20/21 20/23 20/13 50/30 20/17 20/18 30/19 20/27
VP REC V1B 30/26 20/19 30/22 30/24 20/16 20/12 20/20 20/13 30/28 30/25 20/18 30/23 50/41
OXYTOCIN REC 40/29 30/19 30/23 50/37 30/24 30/26 30/21 20/21 50/25 20/19 20/17 30/22 30/24
NEUROTEN REC 1 40/24 40/30 40/30 50/29 30/20 30/28 30/26 20/20 30/27 30/21 20/24 30/20 50/21
NEUROPHYS 1 40/46 30/21 40/22 30/30 30/27 20/14 20/13 30/24 40/25 20/20 30/22 20/22 40/23
NEUROPHYS 2 40/43 40/23 30/22 40/28 50/28 20/14 20/13 40/31 40/25 20/22 30/22 30/27 40/23
GAST BIND PROT 40/38 40/29 50/46 40/31 40/37 20/18 30/28 50/27 40/28 30/19 20/22 20/23 50/31
GAST/CHOL REC B 30/27 30/30 40/30 20/22 40/27 20/19 30/22 30/21 30/27 20/22 20/20 30/24 40/29
*Homology scores derived from comparing 25 peptide receptors, and associated binding and transport proteins, and enzymes with 13 peptide hormone sequences. All
sequences are from Swiss-Prot protein database and homologies are from LALIGN on the Expasy server. The first number in each pair represents the number of identical
amino acid residues in the most homologous ten amino acid region found by LALIGN, and can therefore be read as a percent identity. Homology scores of 50 and above
were considered significant and are bold and printed in large type. The second number is the significance score (see text) derived by LALIGN for this homologous region.
The significance score represents the likelihood that such homologies occur by chance. The higher the number, the more significant the homology. Scores that are
significant to po0.05 are printed in bold, large type. Those homologies that are significant in terms of both identity and chance occurrence are underlined. Abbreviations
are as follows: INS ¼insulin; INS REC ¼insulin receptor; GLUC ¼glucagon; AMYL ¼amylin peptide; ANP ¼atrial natriuretic peptide; AII ¼angiotensin II;
ACE ¼angiotensin converting enzyme; ACTH ¼adrenocorticotrophic hormone releasing hormone; LHRH ¼luteinizing hormone-releasing hormone; SOM ¼
somatostatin; SUBP ¼substance P; VP ¼vasopressin; OXY ¼oxytocin; NT ¼neurotensin; GAST ¼gastrin; NEUROPHYS ¼neurophysin; CHOL ¼cholecystokinin.
CASE STUDY OF INSULIN, GLUCAGON AND GASTRIN 75
evidence also exists linking the Nterminus of the
beta subunit (amino acids 765–770) to insulin
binding (Prigent et al., 1990). No significant
homologies were found to insulin or glucagon in
this region. Two additional insulin-like regions
in Table 5, as well as several in Table 3, do not
correlate with identified insulin binding regions
of the receptor and may represent additional
sites of insulin interaction.
Structural data do not seem to be available yet
concerning the nature of the gastrin binding site
on its receptor or carrier protein. Perhaps the
homologies listed in Table 6 will help to locate
important gastrin binding regions on these
proteins.
In sum, the homology data displayed in
Tables 1–6 confirm both the RBD and Dwyer
theories that peptide (self) complementarity
may be involved in peptide receptor evolution.
Insulin and glucagon, as noted above, are both
known to be self-aggregating peptides. Dwyer
would predict that their receptors should have
homologous regions in their bindings sites and
they do. It is also known that insulin and
glucagon are molecular complements that not
only bind directly to one another with micro-
molar affinity, but also induce antibodies that
behave as an idiotype–antiidiotype pair (Root-
Bernstein & Dobbelstein, 2001). Thus, the RBD
theory would predict that there should be an
insulin-like region in the glucagon receptor and
a glucagon-like region in the insulin receptor.
Finally, the homologous sequences should ap-
pear at identified ligand binding regions of the
receptors, and they do.
Even the negative results are significant. The
observation that only a handful of significant
peptide–receptor homologies were found among
the 300 remaining peptide–receptor comparisons
is also consistent with the predictions of both the
RBD and Dwyer theories. There is no evidence
in the scientific literature that any of the peptides
other than insulin, glucagon and gastrin self-
aggregate. There is no evidence that any of the
peptides other than insulin+glucagon, gluca-
gon+somatostatin and possibly LHRH+AII
(Root-Bernstein & Dobbelstein, 2001) bind to
one another. In fact, many of the peptide pairs
investigated here have previously been experi-
mentally tested in our laboratory for binding
using several physicochemical and antibody
techniques, as summarized in Table 7. Except
for the possibility that LHRH and AII are
complementary, the remaining homology search
data are highly consistent with available experi-
mental binding data.
These results provide the basis for suggesting
a possible process by which peptide receptors
evolved and therefore for locating binding
regions for peptide ligands in receptors for
which the binding sites are as yet unknown.
A Theory of Peptide Receptor Evolution
The results presented here suggest that peptide
receptor specificity did not result from random
evolution but evolved from pre-existing peptide
interactions. One must, however, consider all
possible explanations of these observations. One
Ta b l e 2
Significant similarities between glucagon and the
glucagon receptor
GLUCAGON REC 139−50 SFQVMYTVGYSLSL
\ | : | | | |
GLUCAGON 1−15 HSQGTFTSDYSKYLD
| : | | | : | : / : : |
GLUCAGON REC 372−85 HAQGTLRS - AKLFFD
GLUCAGON REC 210−230 LSVSTWLSDGAVAG-CRVAAVF
| : | : | | : : | | |
GLUCAGON 1−22 HSQGTFTSDYSKYLDSRRAQD
F
: | : \ | | \ : \
GLUCAGON REC 195−208 DGLLRTRYSQKIGD
Ta b l e 3
Significant similarities between glucagon and the
insulin receptor
INSULIN REC, α 390−405 E I SGYLK IRRSYALVS
: | | | | | : : | :
GLUCAGON 8−25 SDYSKYLDSRRAQDFVNW
: / / / | | : : : : : : : |
INSULIN REC, α 424−41 NYSFYALDNQNLRQLWDW
INSULIN REC, α 284−304 SFCQDLHHKCKNSRRQGCH-QY
: | : | : | : | | | \ : : : :
GLUCAGON 5−25 TFTSD -YSKYLDSRRAQDFVNW
| | : | : : | | | : : |
INSULIN REC, α 157−173 T ID -WSRI LDSVEDNH IV
R. ROOT-BERNSTEIN76
Ta b l e 5
Significant homologies between insulin and the insulin receptor
INS B CHAIN 39−53 LYLVCGERGFF-YTPK
| : : | : | | | | |
INS REC, αSUBUNIT 450−60 LTTTQG -KLFFHYNPK
INS REC, αSUBUNIT 91−118 FRVYGLESLKDLFPNLTVI RGSRLFFNY
| / | : \ : | : | : : : | : | | | |
INS B CHAIN 25−50 FVNQHLCGSH-LVEALYLVCGERGFF-Y
| | | | : | | | | | :
INS REC, β SUBUNIT 897−916 HLCVSR-KHFALERGCRLRGL
INS B CHAIN 25−62 FVNQHLCGSHLVEALYLVCGERGFFYTPKTRREAEDLQ
| : \ : | : | : | / | \ : | : : : | : | | |
INS REC, α 658−695
FWERQAEDSELFELDYCLKGLKLPSRTWSPPEESEDSQ
INS B CHAIN 30−43 LCGSHLVEALYLVC
: | | | | | |
INS REC, αSUBUNIT 222−248 ICKSHGCTAEGLCCHSEC-LGNCSQPDD
| : : | | | | | | : : : : :
INS A CHAIN 90−110 G - I VEQ- CCTSI CSLYQLENYCN
Ta b l e 4
Significant similarities between insulin and the glucagon receptor
INS A CHAIN 3−20 EQCCTS I CSLYQ-LENYCN
: | | : : / | | | |
GLUCAGON REC 41−59 DQCHHNLSLLPPPTELVCN
(EXTRACELLULAR)
INS A CHAIN 2−20 IVEQCCTS - - I CS LYQLENYC
: | | : : : / : | | |
GLUCAGON REC 220−240 AVAGCRVAAVFMQYGI VANYC
(EXTRACELLULAR-TRANSMEMBRANE)
INS B CHAIN 1−22 FVNQH-LGSH - - LVE A LYLVCGERG
: \ | | : : : : : | | | |
GLUCAGON REC 83−109 HKVQHRFVFKRCGPDGQW-VRGPRG
(EXTRACELLULAR)
INS B CHAIN 1−25 FVNQHLCGSHLVEALYL - - - - - - - VCGERGFFY
: | \ : | / | | | : | | | | | : | | \
GLUCAGON REC 234−267 GIVANY - C- WLLVEGLYLHNLLGLATLPERSFFSLY
(TRANSMEMBRANE+CYTO) | | | : | : / / / | |
INS A CHAIN 1−18 GIVEQC - C -TS ICSLYQLEN
INS C CHAIN 68−81 LGGGPGAGS LQ- -P L A
| | | | | | | : |
GLUCAGON REC 268−283 LG I GWGAPMLFVVPWA
(TRANSMEMBRANE)
INS C CHAIN 72−81 PGAGSLQPLA
| | | : | / | |
GLUCAGON REC 464−473 PLAGGLPRLA
(CYTOPLASMIC)
CASE STUDY OF INSULIN, GLUCAGON AND GASTRIN 77
is that convergent evolution is the cause of the
homologies reported here. Since insulin self-
aggregates, it is possible that proteins evolving to
act as an insulin receptor would be selected for
the same binding properties, and thus the same
general sequences, as insulin itself. The same
argument could be made, mutatis mutandis for
other self-aggregating peptides. Another possible
explanation of these observations is that com-
plementary peptides and receptors select each
other out of a large random pool of ligands and
receptors. The basic problem with both the
convergent evolution and random selection
models is that it is difficult to imagine how these
would become genetically encoded so as to be
inheritable. How does a presumably functionless
peptide encoded by one gene randomly evolve to
interact with a totally unrelated and equally
functionless protein encoded by another gene so
that at some point their interaction increases
the fitness of the organism? Why, in either case,
should both self-binding and complementary
sequences be found in the insulin and glucagon
receptors? Viable answers to these questions may
exist, but they are not evident at present.
In contrast, it is easy to imagine a modular
mechanism by which receptors could evolve
from self-binding and complementary peptide
sequences attached to standardized transmem-
brane spanning sequences and intracellular
second messenger regions. This modular theory
of receptor evolution is based upon Herb
Simon’s principle that the formation of any
stable subsystem dramatically increases the
Ta b l e 7
Summary of experimental peptide complementarity studies*wz
INS INS REC GLUC AMYL AII ACTH LHRH SOM SUBP TRH THY
pi pi pi pi pi pi pi pi pi pi pi
INS +n++ nnnn
INS REC nn+ nnnnnnnn
GLUC +nnn+ n
AMYL nnnnnnnn
AII n +nn
ACTH nnnn
LHRH nnn
SOM nnnn
SUBP n n
TRH nn
THY n
*All immunological data summarized from Root-Bernstein & Dobbelstein (2001); physicochemical data from Root-
Bernstein & Westall (1986); and Root-Bernstein (unpublished data).
w(p) Physicochemical technique [pH titration (indirect measure of binding), or NMR and/or UV spectrophotometry
(direct measures of binding)] (Root-Berntein & Dobbelstein, 2001; Root-Bernstein & Westall, 1986);
zBoth physicochemical (p) and immunological (i) techniques were used as indicated to test for complementarity between
the peptides. Abbreviations for the peptides are as in Table 1 with the following additions: TRH ¼thryotropin releasing
hormone; THY ¼thyroglobulin.
(i) immunological technique [double antibody diffusion (a modification of Ouchterlony immunodiffusion (see Root-
Bernstein & Dobbelstein, 2001) and/or double antibody ELISA, in which one antibody is plated and the amount of the
second antibody binding to it is measured (ibid.))]; (+) evidence of complementarity; () no evidence of complementarity;
(n) not done.
Ta b l e 6
Similarities between gastrin and its binding
proteins
GASTRIN 1−9 GPWLEEEEE
| | | : | |
GASTRIN BINDING PROTEIN 226−234 GPG LKPPEE
GASTRIN 3−11 LEE EEEAYG
| : : | | | |
GASTRIN BINDING PROTEIN 732−41 LKKYE AAYG
GASTRIN 1−13 GPWLEEEEEAYGW
| | / : : | | |
GASTRIN RECEP FRAG. 274−86 GPREQNLGEAELW
R. ROOT-BERNSTEIN78
probability that a complex system will form
incorporating that subsystem (Simon, 1981) com-
bined with Root-Bernstein and Dillon’s recogni-
tion that complementarity serves to create stable
subsystems (Root-Bernstein & Dillon, 1997).
The theory of receptor evolution by comple-
mentarity proposes that clearcut selective
advantages existed for each element in the
evolutionary process from subunits through the
prototype receptor. Such a complementarity-
based modular model of evolution might occur
as follows. Assume that in a protobiological
world, peptides are encoded by short regions of
polynucleotides and that some type of feedback,
involving either direct chemical interaction or
selection for the translated system, exists be-
tween these protogenes and the peptides they
encode. In such a system, selection of the peptide
(phenotype) would confer a selective advantage
on its encoding gene (genotype). Assume that
some peptides bind selectively to other peptides
[as has in fact been demonstrated experimentally
(Root-Bernstein & Westall, 1984a, b; Root-
Bernstein, 1987; Root-Bernstein & Holsworth,
1998; Schenk et al., 1991)] resulting in a
primitive form of metabolic regulation as de-
scribed previously by Root-Bernstein & Dillon
(1997). In such a system, the ‘‘selection’’ of, say,
insulin and glucagon to regulate glucose meta-
bolism would not be by chance, but as a result of
their complementarity. Thus, whether insulin
regulates its activity by self-aggregation or by
binding to its antagonist, glucagonFor by both
mechanismsFis irrelevant to the fact that such
interactions would have been functional and
therefore advantageous to the primitive organ-
ism. Moreover, such a (self ) complementary
system would be highly buffered and unusually
stable. Peptide–peptide interactions would buf-
fer the peptides against degradative processes
resulting in non-random survival of aggregating
peptides (Root-Bernstein & Dillon, 1997). Pep-
tide-gene interactions, or selection of a system
using aggregating peptides, would result in non-
random survival of the genes encoding peptide
pairs. More gene sequences encoding peptides
that interact with other peptides would survive
than gene sequences encoding non-interacting
peptides and the distribution of gene frequencies
would become skewed to favor complementary
networks of peptides and genes. One would
therefore expect to find an unusually large
proportion of genes encoding self-binding and
complementary peptide sequences, as compared
with chance, among the constituents of living
systems. This theory helps to explain a
phenomenon well-known but rarely published
by working biochemists: almost everything in a
biological system seems to bind to some extent to
almost everything else.
One consequence of skewed gene frequencies
would be gene duplication of sequences encoding
(self ) complementary peptides. Gene duplication
is a well-recognized mechanism driving evolu-
tion. Many proteins are known to have evolved
by stringing together multiple copies of peptide
regions found within them and by mixing or
swapping such regions (Dwyer, 1998; Gilbert,
1978, 1987; Doolittle, 1978). Such duplication
and swapping mechanisms would allow a variety
of complementary peptides to become associated
to produce a wide array of novel functional
polypeptides based on a limited repertoire of
component subunits.
Forms of complementarity other than pep-
tide–peptide complementarity would also have
been at work during prebiotic evolution that
would have led to the predominance of other
types of peptides that could participate in
modular processes of variation. Some peptides
would have been selected non-randomly for their
characteristic of incorporating into lipid mem-
branes. This relatively non-specific form of
complementarity would once again buffer these
peptides (and by selection, their genes) against
degradation, giving them a selective advantage
over non-complementary peptides and their
genes. Transmembrane regions of proteins are,
in fact, very highly conserved (Kihara &
Kanehisa, 2000; Mitaku et al., 1999) so that it
is possible to predict whether any particular
sequence will incorporate into a membrane with
extremely high accuracy (Krogh et al., 2001;
Gromiha, 1999; Mitaku et al., 1999; Gromiha &
Ponnuswamy, 1993). The severe restrictions on
beta chain and alpha helical peptide sequences
that are lipophilic, and the high degree of
homology between known sequences, argues
for a high degree of selection early in evolution
and relatively little variation since.
CASE STUDY OF INSULIN, GLUCAGON AND GASTRIN 79
Other peptide sequences would have been
selected for a different sort of chemical
complementarity: their ability to catalyse specific
reactions, bind up ions, or to perform similar
chemical functions that were advantageous to
protobiological systems. Thus, as a result of all
of the various manifestations of molecular
complementarity, a range of genes encoding
modular peptide subunits of different classes of
molecular interaction would have been selected
non-randomly: peptides that bound to other
peptides; peptides that bound to lipid mem-
branes; and peptides that bound to other
chemicals catalysing reactions with them
(Fig. 1). These peptides and their encoding genes
could have been strung together in many
possible ways, some of which would have yielded
protoreceptor sequences.
Modular evolutionary processes could pro-
duce protoreceptors as follows. A gene (or its
RNA) encoding a self-binding or complemen-
tary peptide, a gene (or RNA) encoding a lipid-
incorporating peptide, and a gene (or RNA)
encoding an ion-transferring or catalytic peptide
could be strung together to yield a polypeptide.
When the modular process linked the protogenes
[or ‘‘transposable exons’’, as Dwyer (1997)
has called them] in the right order, then a
functional protoreceptor would have resulted
(Fig. 2). This protoreceptor could have been
activated by binding interactions involving
extracellular and intracellular portions of the
polypeptide (Fig. 3).
Protoreceptors would be a selective advantage
for cells because they would permit a separation
of the information carrying function of peptides
from their metabolic functions, as well as
providing a more sensitive mode of functional
control and homeostatic control than can be
achieved by direct peptide–peptide interactions.
Thus, modular-based polypeptides would ac-
quire functions impossible for individual pep-
tides (emergent properties). Since some peptides,
such as insulin and glucagon, are large enough to
engage in multiple peptide interactions, they
might bind to more than one protoreceptor
Fig. 1. Various forms of molecular complementarity result in complexes that have functional and/or structural stability.
The combining of three fundamental types of peptides either at a genetic or post-translational level may have been required
to produce the first protoreceptors: a peptide capable of complementary interactions (A); a peptide that could interpolate
into lipid membranes (B); and a peptide capable of carrying out or participating in a chemical reaction (C). The complexes
each peptide was capable of forming would have buffered the components against degradation and/or resulted in novel
functional properties that had survival value. Thus, peptides involved in such complexes, and the genes encoding them,
would have dominated early evolution and protein sequences based upon combinations of their sequences should be
unusually common in modern proteins. (D) A complex of two peptides (or a peptide and another type of compound). (E) A
complex of a peptide (twisting tube) with a lipid bilayer (ladder) as might be found in a cell membrane. (F) A complex of a
peptide that when folded appropriately can catalyse a chemical reaction involving transformation (arrow) of one compound
(teardrop) into another (ball).
R. ROOT-BERNSTEIN80
simultaneously, creating concerted protorecep-
tor interactions. Such multiple interactions
might eventually have resulted in selection for
multi-transmembrane loops defining ligand
binding sites and the multiple transmembrane
structures of the receptors found so frequently
in higher organisms today. The evolution of
protoreceptors and their more complex conju-
gates would have been fostered by the fact that
the molecular complementarity of the subunits
themselves would have conferred functionality
on the modular constituents at each stage of the
evolutionary process from the origins of meta-
bolism through the diversification of receptor
subtypes.
Implications of the Theory for Receptors
in General
This modular theory of peptide receptors from
complementary subunits has many testable
implications that stem from the observation that
a sort of molecular paleontology exists within
such systems (Root-Bernstein & Dillon, 1997).
This molecular paleontology is capable of
yielding insights into how the chemistry of the
past shaped the present physiology and pharma-
cology of life.
First, there may exist in the binding regions of
many (if not all) peptide receptors, complemen-
tary peptide sequences that not only represent
the origins of the receptor, but which may also
be mimics of fundamental regulatory peptides.
Some of these mimics may have homologues
among known or yet-to-be-discovered regula-
tory peptides, hormones or neurotransmitters.
Whether natural homologues exist or not,
identification of such receptor-embedded peptide
complements may have practical pharmaceutical
benefits.
Second, it has been demonstrated that some
monoamines will bind with high specificity and
moderate affinity to specific peptides. Three
examples are the binding of catecholamines to
enkephalins (Root-Bernstein, 1987); serotonin to
adrenocorticotrophic hormone (Root-Bernstein
& Westall, 1984a); and dopamine to neurotensin
(Schenk et al., 1991). Since ions may also
intercalate between aromatic residues (Kumpf
& Dougherty, 1993), peptides containing pairs of
aromatic side chains [or ‘‘molecular sandwiches’’
(Root-Bernstein, 1984)] may also have specificity
for particular ions. These observations suggest
that receptors for non-peptide ligands, including
Fig. 2. A combination of genes or the post-transcrip-
tional elision of the mRNAs encoding the peptide elements
described in Fig. 1 could lead to a synthetic polypeptide
(integrated gene product) that could function as a proto-
receptor. A peptide sequence capable of binding a ligand
could be expressed on the exterior of a lipid membrane
(left); a reaction-catalysing peptide could be displayed on
the interior of a lipid membrane (right); and a lipophilic
peptide could span the lipid bilayer, linking the binding and
catalysing regions of the polypeptide. This figure illustrates
a configuration that such a protoreceptor might display
when no ligand is bound. The ligand-binding region might
self-associate with part of the transmembrane region,
pulling it partially out of the membrane and drawing an
amphoteric part of the catalytic region into the membrane,
thereby hiding part of the catalytic region, inactivating it.
Fig. 3. A possible mechanism by which a protoreceptor
might be activated by binding of a ligand. Binding of the
ligand dissociates the binding region from its self-associa-
tion with the lipid-spanning region of the polypeptide,
allowing it to reposition itself into the lipid bilayer. The
catalytic portion of the polypeptide in the interior of the
membrane is revealed, allowing the substrate to bind and be
reacted upon.
CASE STUDY OF INSULIN, GLUCAGON AND GASTRIN 81
monoamines, steroids, ions and other classes of
compounds, might also have origins in peptides
with affinity for those ligands. Mulchahey et al.
(1999) have made a similar suggestion for
extending the molecular recognition hypothesis
to non-peptides, which may also be compatible
with the modular complementary theory pre-
sented here. Thus, the modular mechanism
of evolution suggested here for complementary
peptides appears to be generalizable to mono-
amine and other non-peptide receptor systems.
Evolution appears to have utilized a limited
number of basic modules for transmembrane
insertion of protoreceptors as well as a limited
number of second messenger functions. The
greatest diversity of structures and sequences is
in the extracellular, ligand-binding regions of
receptorsFan observation that the current
theory can explain without difficulty.
A third implication of the modular-comple-
mentary theory concerns the evolution of
physiological control systems. Insulin and glu-
cagon are metabolic antagonists. The fact that
each of their receptors contains a binding site for
the other suggests not only a strong case for co-
evolution of not only the peptide–receptor pair
but the two peptides and of their two receptors
as well. Thus, modular co-evolution may be
extended to entire metabolic systems, not just
pairs of molecules.
Fourth, the existence of insulin-like regions in
the glucagon receptor and glucagon-like regions
in the insulin receptor suggests that each
hormone may modulate the receptor affinity
for its complementary ligand. Circulating insulin
levels may, in other words, regulate glucagon
receptor affinity and glucagon levels may reg-
ulate insulin–receptor affinity. Yamauchi &
Hashizume (1986) have reported that glucagon
does alter insulin binding to its receptor. They
suggest that insulin receptor affinity is mediated
by glucagon via intracellular second messengers
such as cAMP, but their data do not preclude a
direct peptide–peptide binding mechanism. Such
peptide hormone binding interactions may be
more general than this single case. Ruiz-Opazo
et al. (1995) have reported a receptor protein
that contains both angiotensin and vasopressin
affinities. Perhaps angiotensin and vasopressin
therefore co-evolved in ways that may reveal an
underlying chemical basis. Should such dual
receptor functions be widespread, then under-
standing the evolution of peptide ligand–recep-
tor systems may yield surprising insights into the
regulation and pharmacology of such systems
as well.
Experimental Tests of the Theory
A number of experimental tests that have
practical as well as theoretical importance are
suggested by the findings reported here. If the
theory is generalizable to the evolution of all
peptide receptors, then it should be possible to
find within every peptide receptor one or more
sequences that are complementary to the ligand.
One method for discovering such complementary
peptides would be to synthesize a peptide library
of overlapping sequences mapping the extracel-
lular regions of peptide receptors and then screen
this library for ligand affinity. (Such experiments
would not only test the current theory, but might
also yield leads to novel pharmaceutical antago-
nists to the ligand peptide.)
A second test of the theory is that every
biologically active peptide is likely to have a
natural antagonist to which it binds. One of the
predictions made by the RBD theory that relates
to this principle is that it should therefore be
possible to discover such complements in one of
two ways. First, peptides that are known to have
antagonistic or modulatory effects should be
tested for their ability to bind to one another. It
has been shown, for instance, that bovine pineal
anti-reproductive tripeptide, which antagonizes
LHRH activity, binds to it with micromolar
affinity (Root-Bernstein & Westall, 1986). Sec-
ondly, it should be possible to immobilize a
known peptide ligand and use it to ‘‘fish’’ out its
complement(s) from a homogenate of naturally
occurring peptides. For example, glucagon was
discovered as a ‘‘contaminant’’ of crystallized
insulin (Bloch, 1994), suggesting how important
such an approach might be for discovering
antagonistic or modulatory drugs. In fact, such
complexation appears to have been observed
frequently, but ignored, by the pioneers of
peptide hormone research such as Guillemin
and Schally, who reported that an unknown
R. ROOT-BERNSTEIN82
peptide of constant molecular weight and amino
acid composition bedeviled attempts to purify
LHRH (Root-Bernstein & Westall, 1986; Schally
et al., 1966; Guillemin et al., 1965). (Unfortu-
nately, this ‘‘contaminant’’ was discarded with-
out further characterization shortly after LHRH
was finally isolated and no samples remain to be
further characterized.) Peptide pairs isolated in
this manner may provide important clues as to
the evolution of metabolic control mechanisms
that are now reified as ligand–receptor complexes.
Conclusions
In sum, the fact that the receptors of
complementary and self-complementary pep-
tides often have multiple copies of the ligand in
the binding region of the receptor is a striking
observation. It suggests that receptors and
ligands do not, at least in some cases, evolve
independently or randomly, but by some sort of
concerted mechanism. Such a mechanism is most
compatible with a modular form of evolution
that takes advantage of several different forms of
complementary interactions including peptide–
peptide binding, peptide–membrane binding,
and peptide–substrate binding. This modular,
complementarity-based mechanism has charac-
teristics that make it generalizable to the
evolution of all types of receptors. Such a
mechanism, if correct, leaves a particular type
of molecular paleontology that is accessible to
analysis by both bioinformatic and experimental
exploration and which may have important
physiological and pharmaceutical implications.
REFERENCES
Attwood, D., Florence, A. T., Grieg, R. & Smail, G. A.
(1974). Some solution properties of pentagastrin and
angiotensin: aggregation of pentagastrin. J. Pharm.
Pharmacol.26, 847–853.
Bloch, C. (1994). The importance of being contaminated.
In: Blondes in Venetian Paintings, the Nine-Banded
Armadillo, and Other Essays in Biochemistry, pp. 65–91.
New Haven: Yale University Press.
Buggy, J., Rossomando, A., MacDouga ll, M., Mie rz,
D., Wunde rligh,D.&Yo o - Wa r r e n , H. (1996). Human
glucagon receptor monoclonal antibodies: antagonism of
glucagon action and use in receptor characterization.
Horm. Metab. Res.28, 215–219.
Burcelin, R., Katz,E.B.&Charron, M. J. (1996).
Molecular and cellular aspects of the glucagon receptor:
role in diabetes and metabolism. Diabetes Metab.22,
373–396.
Carr,D.J.,Bost,K.L.&Blalock, J. E. (1986). An
antibody to a peptide specified by a RNA that is
complementary to gamma-endorphin RNA recognizes
an opiate receptor. J. Neuroimmunol.12, 329–337.
Chou,P.Y.&Fa s m a n , G. D. (1975). The conformation of
glucagon: predictions and consequences. Biochemistry 14,
2536–2541.
Dat h e, M., Ga st, K., Zirwer, D., Welfle,H.&Mehlis,
B. (1990). Insulin aggregation in solution. Int. J. Peptide
Protein Res.36, 344–349.
Degasparo, M. S., Wh ite brea d, S., Einsle,K.&Heus-
ser, C. (1989). Are the antibodies to a peptide
complementary to angiotensin II useful to isolate the
angiotensin II receptor? Biochem. J.261, 310–311.
Doolittle, W. F. (1978). Genes in pieces: were they ever
together? Nature 272, 581–582.
Dwyer, D. S. (1989). Amino acid sequence homology
between ligands and their receptors: potential identifica-
tion of binding sites. Life Sci.45, 421–429.
Dwyer, D. S. (1998). Assembly of exons from unitary
transposable genetic elements: implications for the
evolution of protein–protein interactions. J. theor. Biol.
194, 11–27.
Eberle,A.N.&Hu ber, M. (1991). Antisense peptides of
ACTH and MSH: tools for receptor isolation? In:
Antisense Nucleic Acids and Proteins,Fundamentals and
Applications. (Mol, N. M. J. & Krol, R. V.-D., eds),
pp. 187–203. New York: Marcel Dekker, Inc.
Fa b r y , M., Scha efer, E., Ellis, L., Koj ro, E., Fa h r e n -
holz,F.&Brandenburg, D. (1992). Detection of a
new hormone contact site within the insulin receptor
ectodomain by the use of a novel photoreactive insulin.
J. Biol. Chem.267, 8950–8956.
Gilbert, W. (1978). Why genes in pieces? Nature 271, 501.
Gilbert, W. (1987). The exon theory of genes. Cold Spring
Harbor Symp. Quant. Biol.LII, 901–905.
Goldstein,A.&Brutlag, D. L. (1989). Is there a
relationship between DNA sequences encoding peptide
ligands and their receptors? Proc. Natl Acad. Sci. U.S.A.
86, 42–45.
Gorcs, T. J., Gottschall, P. E., Coy ,D.H.&Arimura,
A. (1986). Possible recognition of the GnRH receptor
by an antiserum against peptide encoded by nucleotide
sequences complementary to mRNA of a GnRH
precursor peptide. Peptides 7, 1137–1145.
Gromiha, M. M. (1999). A simple method for predicting
transmembrane alpha helices with better accuracy.
Protein Eng.12, 557–561.
Gromiha,M.M.&Ponnuswamy, P. K. (1993). Predic-
tion of transmembrane beta-strands from hydrophobic
characteristics of proteins. Int. J. Peptide Protein Res.42,
420–431.
Guillemin, R., Sak iz,E.&Ward, D. N. (1965). Further
purification of TSH-releasing factor (TRF) from sheep
hypothalamic tissues, with observations on the amino
acid composition. Proc. Soc. Exp. Biol. Med.118,
1132–1137.
Huang,X.&Miller, W. (1991). A time-efficient, linear-
space local similarity algorithm. Adv. Appl. Math.12,
337–357.
Ishida,H.&Nag am ats u, S. (2001). Chemical structure,
active sites and receptor binding of insulin molecule
CASE STUDY OF INSULIN, GLUCAGON AND GASTRIN 83
and its signal transduction. Nippon Rinsho 59,
2109–2116.
Ja r p e ,J.A.&Blalo ck, J. E. (1994). Complementary
peptides: applications of the molecular recognition theory
to peptide and protein purification and design. In:
Peptides:Design, Synthesis and Biological Activity
(Basava, C. & Anantharamaiah, G. M., eds). Boston:
Birkhauser.
Joh nson, R. E., Hruby,V.J.&Ru pl ey , J. A. (1979).
Thermodynamics of glucagon aggregation. Biochemistry
18, 1176–1179.
Jurzak, M., Pav o ,I.&Fa h r e n h o l z , F. (1993). Lack of
interaction of vasopressin with its antisense peptides: a
functional and immunological study. J. Receptor Res.13,
881–902.
Kelly, J. M., Tr i n d e r , D., Phillips, P. A., Castley,
D. J., Kemp, B., Mose r,V.&John son, E. I. (1990).
Vasopressin antisense peptide interactions with the V1
receptor. Peptides 11, 857–862.
Kihara,D.&Kanehisa, M. (2000). Tandem clusters of
membrane proteins in complete genome sequences.
Genome Res.10, 731–743.
Kim, D. Y., Lee, J., Sar swat,V.&Pa r k , Y. H. (2000).
Glucagon-induced self-association of recombinant pro-
teins in Escherichia coli and affinity purification using a
fragment of glucagon receptor. Biotechnol Bioeng.69,
418–428.
Ku m pf ,R.A.&Dougherty, D. A. (1993). A mechanism
for ion selectivity in potassium channels: com-
putational studies of cation-pi interactions. Science 261,
1708–1710.
Krogh, A., Larss on, B., Vo n h i e j n e G., & Sonnham-
mer, E. L. (2001). Predicting transmembrane protein
topology with a hidden Markov model: application to
complete genomes. J. Mol. Biol.305, 567–580.
Lougheed, W. D., Albisser, A. M., Martindale,
H. M., Chow,J.C.&Clement, J. R. (1983).
Physical stability of insulin formulations. Diabetes 32,
424–432.
Mitaku, S., Ono, M., Hrokawa, T., Boon-chieng,S.
&Sonoyama, M. (1999). Proportion of membrane
proteins in proteomes of 15 single-cell organisms
analyzed by the SOSUI prediction system. Biophys.
Chem.82, 165–171.
Mulchahey, J. J., Nagy,G.&Neill, J. D. (1999). A
molecular recognition hypothesis for nonpeptides:
Na+K+ATPase and endogenous digitalis-like peptides.
Cell Mol. Life Sci.55, 653–663.
Prigent, S. A., Stanley,K.K.&Siddle, K. (1990).
Identification of epitopes on the human insulin
receptor reacting with rabbit polyclonal antisera and
mouse monoclonal antibodies. J. Biol. Chem.265,
9970–9977.
Rassmussen,U.B.&Hesch, H. D. (1987). On antisense
peptides: the parathyroid hormone as an experimental
example and a critical theoretical view. Biochem. Biophys.
Res. Commun. 149, 930–938.
Root-B ernstei n, R. S. (1984). ‘‘Molecular sandwiches’’ as
a basis for the structural and functional similarities of
interferons, MSH, ACTH, LHRH, myelin basic protein,
and albumins. FEBS Lett.168, 208–212.
Root-B ernstei n, R. S. (1987). Catecholamines bind to
enkaphalins, morphiceptin and morphine. Brain Res.
Bull.18, 509–532.
Root-B ernstei n,R.S.&Dillon, P. F. (1997). Molecular
complementarity I. The complementarity theory of life
and its origins. J. theor. Biol. 188, 447–479.
Root-B ernstei n,R.S.&Dobbelstein, C. (2001). Insulin
binds to glucagon forming a complex that is hyper-
antigenic and inducing complementary antibodies having
an idiotype–antiidiotype relationship. Autoimmunity 33,
153–169.
Root-B ernstei n,R.S.&Holsw orth, D. D. (1998).
Antisense peptides: a critical mini-review. J. theor. Biol.
190, 107–119.
Root-B ernstei n,R.S.&We stal l, F. C. (1984a).
Serotonin binding sites I. Structures of sites on myelin
basic protein, LHRH, MSH, ACTH, interferon, serum
albumin, ovalbumin and red pigment concentrating
hormone. Brain Res. Bull.12, 425–436.
Root-B ernstei n,R.S.&Wes tall , F. C. (1984b).
Fibrinopeptide A binds gly pro-arg-pro. Proc. Natl Acad.
Sci. U.S.A. 81, 4339–4342.
Root-B ernstei n,R.S.&Westa ll, F. C. (1986). Bovine
pineal antireproductive tripeptide binds to luteinizing
hormone-releasting hormone: a model for peptide mod-
ulation by sequence specific peptide interactions? Brain
Res. Bull.17, 519–528.
Ru de n s ky , A. Y., Pr esto n-Huri burt, P., Hoag , S.-C.,
Balrow,A.&Ja n e way , C. A. (1991). Sequence analysis
of peptides bound to MHC class II molecules. Nature
353, 622–627.
Ru iz - O paz o , N., Akimoto,K.&Herrera, V. L. (1995).
Identification of a novel dual angiotensin II/vasopressin
receptor on the basis of molecular recognition theory.
Nat. Med. 1, 1074–1081.
Schally, A. V., Bowers,C.Y.&Redding, T. W. (1966).
Purification of thyrotropic hormone-releasing factor for
bovine hypothalamus. Endocrinology 78, 726–732.
Schenk, J. O., Mo rocco ,M.T.&Ziemba, V. A. (1991).
Interactions between the argininyl moieties of neuroten-
sin and the catechol protons of dopamine. J. Neurochem.
57, 1787–1795.
Simon, H. (1981). The Sciences of the Artificial. Cambridge,
MA: M.I.T. Press.
Tr o p s h a , A., Kizer,J.S.&Chaiken, I. M. (1992).
Making sense from antisense: a review of experimental
data and developing ideas on sense–antisense peptide
recognition. J. Mol. Recogn.5, 43–54.
Uns on , C. G., Cypess, A. M., Wu, C. R., Goldsmith,
P. K., Merrifield,R.B.&Sak mar, T. P. (1996).
Antibodies against specific extracellular epitopes of the
glucagon receptor block glucagon binding. Proc. Natl
Acad. Sci. U.S.A. 93, 310–315.
Wa l t e r , R. (ed). (1975). Neurophysins: Carriers of Peptide
Hormones. Ann. N. Y. Acad. Sci. 248, passim.
Ya mauch i,K.&Hashizume, K. (1986). Glucagon alters
insulin binding to isolated rat epididymal adipocytes:
possible role of adenosine 30-50-monophosphate
in modification of insulin action. Endocrinology 119,
218–223.
Yi p , C. C. (1992). The insulin-binding domain of insulin rece-
ptor is encoded by exon 2 and 3.J.CellBiochem.48, 19–25.
Yi p , C. C., Grunfeld,C.&Goldfine, I. D. (1991).
Identification and characterization of the ligand-binding
domain of insulin receptor by the use of an anti-peptide
antiserum against amino acid sequence 241–251 of the
alpha subunit. Biochemistry 30, 695–701.
R. ROOT-BERNSTEIN84