Glucose homeostasis with infinite

Daisyworld parable?
J H Koeslag, P T Saunders and J A Wessels

gain: further lessons from the

Department of Medical Physiology and Biochemistry, University of Stellenbosch, Tygerberg 7505, South Africa and 'Department of Mathematics, King's College, Strand, London WC2R 2LS, UK (Requests for offprints should be addressed to ) H Koeslag, Department of Medical Physiology and Biochemistry, PO Box 19063, Tygerberg, 7505,
South Africa)

Abstract

major unresolved physiological problem is how the rate of hepatic glucose production is increased to match the increased rate of glucose utilization during exercise without a change in arterial blood glucose level. A homeostat with such capabilities is said to have infinite gain. Daisyworld is an imaginary planet orbiting a variable star. The only life is black and white daisies. Black daisies retain heat, slightly warming the planet; white daisies cool it. When the two types of daisies grow best at slightly different temperatures, variations in solar luminosity (over a wide range) cause the ratio of white:black daisies to vary in a manner that keeps the planetary temperature constant. This model therefore achieves infinite gain by having two opposing but interdependent controllers. Here we suggest that the pancreatic islet \g=a\-and \g=b\-cells might act as black and white daisies. For the analogy to
A

and insulin must not only have opposing effects on the blood sugar concentration, but the secretion of the one has, at some quantum level, to be at the expense of the other. Electrical coupling between heterocellular groups of \g=a\- and \g=b\-cells within the pancreatic islets suggests that this might indeed be the case. \g=a\-Cellactivity must, furthermore, promote secretory activity in other \g=a\-cells;similarly with \g=b\-cells.This is probably mediated via pancreastatin and \g=g\-aminobutyric acid (GABA) which are paracrinically co-secreted with glucagon and insulin, respectively. \g=a\-Cellactivity spreads (at the expense of \g=b\-cellactivity) when the blood glucose level is below set point, while \g=b\-cellactivity progressively replaces \g=a\-cell activity above set point. At set point changes in the ratio of \g=a\:\g=b\-cellactivity are inhibited.

apply, glucagon

Journal of Endocrinology (1997)

187\p=n-\192 154,

Introduction
A major unresolved physiological problem is how the rate of hepatic glucose production is increased to match the increased rate of glucose utilization during exercise with out a change in arterial blood glucose level (Ahlborg 1969, Felig & Wahren 1975, Zinman et al. 1977, Winder et al. 1979, Kjaer et al. 1991, Roy et al. 1991, Coggan et al. 1995). There are two classic answers. The first involves 'feedforward control' (Riggs 1963, Guyton & Hall 1996). This means that the anticipation and performance of muscular exercise, working via the autonomie nervous system, cause the pancreatic islets to secrete less insulin and more glucagon. Anticipatory or feedforward responses characterize many physiological systems. However, feed forward systems consist of open loops. The magnitude of the response to a given stimulus is subject, therefore, to 'guess-work', and is, on its own, not self-adjusting. The

response is,

at

best, therefore only approximately correct or

appropriate.

possibility involves 'integral control' (Milsum 1966). Integral controllers are closed loop, negative feedback systems. They differ, however, from the negative feedback systems described in the standard physi ology textbooks. A standard negative feedback system responds only when a sensor detects an error in the controlled quantity (blood sugar, arterial blood pressure, body temperature etc.). The greater the error the greater the response (Cannon 1960, Riggs 1963, 1970, Milsum 1966, Guyton & Hall 1996). A zero error always produces a zero response. Thus, a standard controller is incapable of totally correcting an error in the face of steady state disturbances, such as the increased rate of glucose utiliz ation during exercise. Instead, a standard controller merely
The second reduces the error that would have occurred if that troller had not been operational (Riggs 1963). If the
con error

(the reciprocal of half; Riggs 1963, 1970, Guyton 1996, Milsum 1966). Homeostasis with infinite gain implies perfect control. An integral controller responds not simply to the error in the controlled quantity, but to the error multiplied by the time it persists. It is, at present, the only known type of homeostat with infinite gain (Milsum 1966). Thus, during exercise, an integral controller would respond with progressively increasing degrees of hypoinsulinaenna and hyperglucagonaemia until normoglycaemia is restored. On reaching normoglycaemia, the hypomsulinaemia and hyperglucagonaemia persist until the high rate of glucose
be
two

is reduced to half of what it would have been in the homeostat's absence, then the homeostatic 'gain' is said to
& Hall

utilization

starts to wane.

This then

reverses

integral

However, the mechanism for such integral control is unknown. It is striking, however, that most of the body's
effectors

the process.

controllers operate via

(hormones

counterregulatory pairs

of

LUMINOSITY

or

nerves).
temperature response (heavy black lines) to changes on a Daisyworld with microclimates. The thin line indicates the temperature changes on the bare planet. Temperature changes on the planet with black and white daisies are resisted over a wide range of solar luminosities by changes in the proportion of black:white daisies on the planet (see text).

The

Daisyworld parable

Figure

1 The

in solar

luminosity

The sun is probably about 30% warmer 3000 million years ago when life first

planet (Margulis

and carbon dioxide (Lovelock 1972). The Earth's sur face temperature would consequently rise to 8090 C, rendering the planet uninhabitable. To illustrate how even a simple biota might regulate the surface temperature of a planet, Watson & Lovelock (1983) devised an imaginary planet called Daisyworld. It orbits a star very much like our own sun. Daisies are the only form of life. There are only two species, one white and one black. Otherwise they are identical. Both grow best at 22-5 C, and neither can grow at temperatures below 5 C or above 40 C (Watson & Lovelock 1983, Saunders 1994). The temperature on Daisyworld depends on solar luminosity and the planet's albedo (or reflectivity). Black daisies have a lesser albedo and so absorb more heat than bare earth. Hence for a given solar luminosity a planet covered in black daisies will be warmer than a lifeless one. Conversely, a planet covered in white daisies will be cooler than its dead counterpart. If both daisies are present, it is warmer where there are black daisies and cooler where there are white ones. It turns out that these very ordinary properties have remarkable consequences. Consider a planet which is gradually warming under the influence of an evolving sun. When the temperature reaches 5 C daisies begin to

Earth's surface very much more than about 5 C from its present level. This is almost certainly due to the metabolic activities of its biota. Thus, if life were suddenly to disappear today, the amount of atmospheric oxygen and nitrogen would decline to trace levels in an atmosphere of water vapour

appeared on this Sagan 1995). Yet, remarkably, the temperature has probably never varied by
&

today than it was

encouraging more black daisies and, now, even a few white daisies to germinate. The process stops when the planet is covered with flowers. The ratio of black:white daisies, that have germinated, ensures, surprisingly, that the temperature on Daisyworld is now just above 22-5 C
life does not disappear. Instead, black daisies replace the white daisies that have germinated, lowering the planet's albedo, and raising the overall temperature slightly. With an increase in solar luminosity the opposite happens. This is because a change in solar luminosity causes an initial change in planetary temperature. With a rise in tempera ture, for instance, the black daises, which already live in microclimates somewhat above 22-5 C, are disadvantaged, compared with the white daisies, which live in microclimates slightly below optimum, and, therefore, benefit from the rise in temperature. Thus, white daisies replace the black daisies, and will continue to do so until the overall temperature returns to approximately 22-5 C. Over a comparatively wide range of solar luminosities, which would otherwise cause a planetary temperature change of 50 "C, there is, when daisies are present, temperature homeostasis (Fig. 1).

The white ones cool their immediate environ below 5 C and die. The black ones, however, warm their surroundings, encouraging more daisies to germinate. This makes the planet progressively warmer,

germinate.
ment to

(Saunders 1994). If, at this point, solar luminosity were to decrease again,

The Watson 8c Lovelock model

wide range of solar luminosities. The homeostatic gain is infinite. The reason is easy to see. At the cross-over point, say at 22-5 C, both species grow equally well. If the temperature rises even minutely to, say, 22-6 C the white daisies grow faster than the black ones. White daisies therefore replace the black ones, and continue to do so for as long as the temperature on Daisyworld remains above 22-5 C. But, in spreading, the white daisies increase the planet's albedo, causing it eventually to cool to 22-5 C. At this point the black and white daisies grow equally well again, and the replacement process stops. Thus, seen from space, the planet responds to changes in solar luminosity by changing colour from black (at low solar luminosities) to white (at high solar luminosities), keeping its surface temperature robustly at 22-5 C. We believe that the physiology of the insulimglucagon counterregulatory pair of hormones is strongly reminiscent of white and black daisies.
curves, over a

primarily an ecological one, can be simplified to make it more generally applicable. The microclimates are un necessary if winds equalize the temperature over the whole planet, and if black daisies grow optimally at a slightly lower temperature (e.g. 20 C) than the white ones do (e.g. 25 C). Then, irrespective of the shapes of the two growth curves, the temperature on Daisyworld stabilizes at exactly the cross-over point of the two growth

(1983), which

is

junctions

make functional syncytia of small groups of homologous and heterologous islet cells (Meda et al. 1982). Each islet probably consists of a collection of such syncytial

Pancreatic islets

The human islets of Langerhans contain glucagon secret ing -cells, insulin secreting -cells, and somatostatin secreting D-cells. The -cells are situated in the core of a typical islet. The glucagon-secreting -cells form a periph
three cells in thickness (Unger et al. 1978). of -cells (together with D-cells) project into the central portions of the islets along capillary axes, forming lobulations in which the a- and D-cells remain peripheral to the centrally assembled -cells. The D-cells

eral rim
In

one to

humans,

rows

et al. 1994). The mechanisms for the stimulus-secretion coupling in the glucagon producing -cells are unclear. The rate of glucose metabolism is only about 1520% ofthat observed in the -cell (Rorsman et al. 1991). Furthermore, although in vitro studies have indicated that -cells are to some extent glucose sensitive (Sumida et al. 1994), it is unclear whether variations in glucose concentration in the physio logical range have a direct effect on -cells in vivo (Asplin et al. 1981, Rorsman & Helman 1988, Rorsman et al. 1991). Indeed, the -cells of patients with insulindependent diabetes mellitus continue to secrete glucagon at normal resting rates despite severe hyperglycaemia (Unger et al. 1970, Raskin et al. 1976, Powell et al. 1993). These patients also do not increase glucagon secretion during hypoglycaemia (Asplin et al. 1981, Bolli et al. 1983, Powell et al. 1993). The -cells therefore probably receive their glycaemic cues primarily from the -cells, presum ably via electrical coupling (via gap junctions) between these cells (Orci et al. 1975, Michaels & Sheridan 1981, Meda et al. 1982, 1986).

The -cells are unequivocally glucose sensitive. The mechanism may be summarized as follows (Ashcroft et al. 1994, Dunne et al. 1994). High plasma glucose concen trations lead to high rates of glycolysis. Some product of this metabolism, now generally accepted to be ATP, channels brings about closure of the ATP-sensitive in the plasma membrane, causing depolarization. This leads to the opening of voltage-dependent Ca" channels, and increases in the intracellular free Ca" concentration. This initiates exocytosis of insulin-containing secretory vesicles (Prentki & Matschinsky 1987, Ashcroft et al. 1994, Berggren & Larsson 1994, Dunne et al. 1994). Insulin secretion, for any given -cell, appears to be all-or-none

aggregates.

(Pipeleers

situated between the outer mantle of -cells and the core of -cells. In addition, there are heterocellular regions in each islet, with -, - and D-cells in close proximity
are

Insulin and glucagon responses hypoglycaemic clamps

to

hyper-

and

A characteristic feature of the islet cells are membrane specializations that may involve intercellular relationships. These specializations consist of tight junctions, desmosomes and gap junctions (Orci et al. 1973, 1975). Gap junctions, through which molecules of less than 1000 Da can move from the cytoplasm of one cell to that of an adjacent cell without entering the intercellular space, have been found between a- and -cells (Orci & Unger 1975, Orci et al. 1975, Orci 1976, Meda et al. 1982), as well as between D-cells and a- or -cells (Unger et al. 1978, Michaels & Sheridan 1981, Meda et al. 1982). Thus, gap

(Orci

&

Unger 1975).

When the endocrine pancreas is exposed to a hyperglycaemic clamp (e.g. 11 or 16 mmol glucose/1), there is an initial spike of insulin release lasting for less than 10 min. This is followed by a slow continuous rise in insulin secretion, over the next 2 h, to very high levels (Fig. 2) (Grodsky 1972, Gench et al. 197'4, Bolaffi et al. 1986, Tsuchiyama et al. 1992). A hypoglycaemic clamp (1 mmol/1) produces a similar spike in glucagon secretion, followed by a gradual rise in glucagon production (Weir et al. 1974). The mathematical characteristics of the acute response led Grodsky (1972) to postulate that -cells operate as on-off units with differing blood glucose thresholds. Direct evidence for such a model has recently

neighbouring
>5mM Glucose

aB
GABA

Ab
< 5mM Glucose
'

pancreastatin
neighbouring

Time from start of

hyperglycaemic clamp (mins)

Figure 2 The results of an archetypical hyperglycaemic clamp experiment (Grodsky 1972, Cerich et al. 1974, Bolaffi et al. 1986, Tsuchiyama et al. 1992). If the isolated pancteas is perfused with blood containing a constant 15 mmol glucose per litre, then there is an immediate spike of insulin release, followed by a prolonged 'second phase' of progressively increasing rates of insulin release for more than 2 h. This 'second phase' is characteristic of an

Ab

3 The proposed model of pancreatic islet function. The functional units of the endocrine pancreas are suggested to be

Figure

a-

integral response, indicating that the different -cell thresholds (Pipeleers 1992, Pipeleers et al. 1994) are functions of time x hyperglycaemic stimulus. This result would also typify the white daisy response to a heat clamp experiment on Daisyworld.

sub-islet heterologous syncytial aggregates of electrically coupled and -cells, which operate as flip-flop mechanisms. They are either in the Ab (glucagon secreting) or aB (insulin secreting) mode. Ab->aB and aB->Ab transitions occur spontaneously, but are influenced by paracrine secretions from neighbouring units (see text). The relative effectivenesses of the two types of paracrine secretion are determined by the blood sugar concentration.

been demonstrated (Schuit et al. 1988, Pipeleers 1992, Hellman et al. 1994, Pipeleers et al. 1994). However, the slowly rising 'second phase' of insulin release in response to a hyperglycaemic clamp cannot be due to simple differences in -cell glucose thresholds, as suggested by Grodsky (1972), Pipeleers (1992) and Pipeleers et al. (1994). Instead, the evidence suggests that high steady state blood glucose concentrations recruit more and more active -cells with time. The -cell 'thresholds' are therefore, somehow, for different integrals (stimulus magnitude x time) of the stimulus with respect to time. If -cells operate as on-off units, and are functionally linked to -cells, then the two probably operate as a flip-flop mechanism, which secretes either insulin (in the aB mode) or glucagon (in the Ab mode). Both are secreted maximally or not at all (Schuit et al. 1988, Pipeleers 1992, Hellman et al. 1994, Pipeleers et al. 1994). Only under special, near-normoglycaemic circumstances would the system be modulated (via autonomie nerves, the gut and adrenal hormones) to secrete both hormones, or neither.

Glucose homeostasis

During the recent past it chromogranin-A (CgA) and

has become evident that its proteolytic products are costored and cosecreted with most peptide hormones, including insulin and glucagon (Cohn et al. 1984, Efendi c et al. 1987, Eiden 1987, Greeley et al. 1989, Huttner et al.

1991, Helle & Angeletti 1994, O'Connor et al. 1994, Huttner & Natoli 1995). Pancreastatin, the proteolytic fragment of CgA found in pancreatic -cells (Schmidt et al. 1988, Winkler & Fischer-Colbrie 1992), is a powerful paracrine inhibitor of insulin secretion, -amino butyric acid (GABA) has been shown to colocalize with insulin in the -cells (Gerber & Hare 1980, Garry et al. 1986) and has been shown to inhibit -cell activity (Rorsman et al. 1991). CgA and CgB have similarly been localized in -cells (Winkler & Fischer-Colbrie 1992). An as yet uncharacterized CgA or CgB fragment might therefore also paracrinically inhibit glucagon secretion. The existence of these strongly inhibitory paracrine secretions, and the arrangement of a- and -cells into syncytial aggregates, suggest a novel, Daisyworld-like form of blood glucose homeostasis. Thus, we propose that the individual syncytial aggre gates flip-flop spontaneously between the insulin (aB) and glucagon (Ab) secreting modes (Fig. 3). Aggregates which are in the Ab (glucagon-secreting) mode co-release pan creastatin, which promotes aB>Ab transitions among their immediate neighbours. Ab units (like black daisies) therefore tend to proliferate locally. aB units, which cosecrete GABA with insulin, promote Ab>aB tran sitions. They too proliferate locally (like white daisies). At a blood sugar concentration of 5 mmol/1 these proliferative tendencies are equal (Fig. 3). Thus, although individual aggregates are flipping randomly from one mode to the other, there is no overall change in the Ab:aB ratio. Both

Ab>aB transitions then continue for more than 2 h, producing progressively higher and higher rates of insulin secretion (Fig. 2). The process stops when all the units are presumably in the aB mode. In summary, therefore, we propose that glucose homeostasis, with infinite gain, is critically dependent on a pair of counterregulatory hormones (which is superfluous in a standard homeostat). Our model also accords, for the first time, full physiological significance to the puzzling syncytial connections between sub-islet sets of a- and -cells, as well as to the presence of pancreastatin and other inhibitory paracrine secretions in the pancreatic islets. Standard homeostats benefit from neither of these features. References
1969 Blood glucose during prolonged physical exercise in Frvarsmedicin 3 (Suppl 1) 8599. Ashcroft FM, Proks P, Smith PA, mml C, Bokvist & Rorsman 1994 Stimulus-secretion coupling in pancreatic cells. Journal of Cellular Biochemistry 55S 5465. Asplin CM, Paquette TL & Palmer JP 1981 In vivo inhibition of glucagon secretion by paracnne beta cell activity in man. Journal of Clinical Investigation 68 314-318. Berggren P-O & Larsson O 1994 Ca2+ and pancreatic B-cell function. Biochemical Society Transactions 22 12-18. Bolaffi JL, Heldt A, Lewis LD & Grodsky GM 1986 The third phase of in vitro insulin secretion: Evidence for glucose insensitivity. Diabetes 35 370-373. Bolli G, DeFeo P, Compagnucci R, Cartechini MG, Angeletti G, Santeusanio F, Brunetti P oc Gerich JE 1983 Abnormal glucose counterregulation in insulin dependent diabetes mellitus. Interaction of anti-insulin antibodies and impaired glucagon and epinephrine secretion. Diabetes 32 134-141. Cannon WB 1960 The Wisdom of the Body. New York: WW Norton. Coggan AR, Raguso CA, WiUiams BD, Sidossis LS & Gastaldelli A 1995 Glucose kinetics during high intensity exercise in
man.

glucagon are therefore present in the blood of normoglycaemic persons. (In contrast, a standard homeo stat would require neither hormone to be present.) A fall in the blood glucose concentration, due to a sudden increase in the rate of glucose utilization, would, according to our hypothesis, simply potentiate pancreastatin's action. This would increase the probability of aB>Ab flips, thereby progressively increasing the proportion of units in the Ab mode. The glucagoninsulin ratio in the blood will therefore increase with time, as is typical of exercise (Wahren et al. 1973, Koeslag et al. 1980, 1982). Only when normoglycaemia has been completely restored will the bias towards aBAb transitions cease. Normoglycaemia in the presence of the new, high Ab:aB ratio persists for as long as there is an increased rate of glucose utilization. This is glucose homeostasis with infinite gain. An increase in the blood glucose concentration has the opposite effect. There is now a bias towards Ab>aB transitions which continues until normoglycaemia has been restored. During hyperglycaemic clamp experiments the blood glucose concentration cannot be corrected.

insulin and

endurance-trained and untrained humans. Journal of Applied Physiology 78 1203-1207. Cohn DV, EltingJJ, Frick M & Eide R 1984 Selective localization of the parathyroid secretory protein I/adrenal medulla chromogranin A protein family in a wide variety of endocrine cells of the rat. Endocrinology 114 1963-1974. Dunne MJ, Harding EA, JaggarJH & Squires PE 1994 Ion channels and the molecular control of insulin secretion. Biochemical Society Transactions 22 612. Efendic S, Tatemte , Mutt V, Quan C, Chang D & stenson C-G 1987 Pancreastatin and islet hormone release. Proceedings of the National Academy of Sciences of the USA 84 7257-7260. Eiden E 1987 Is chromogranin a prohormone? Nature 325 301. Felig & Wahren J 1975 Fuel homeostasis in exercise. New England Journal of Medicine 293 1078-1084. Garry DJ, Sorenson RL, Eide RP, Maley BE & Madsen A 1986 Immunohistochemical colocalization of GABA and insulin in -cells of rat islets. Diabetes 35 1090-1096. GerberJC & Hare TA 1980 GABA in peripheral tissues: presence and actions in pancreatic function. Brain Research Bulletin 5 (Suppl 2) 341-346. GerichJE, Charles MA & Grodsky GM 1974 Characterization of the effects of arginine and glucose on glucagon and insulin release from the perfused rat pancreas. Journal of Clinical Investigation 54 833-841. Greeley GH, Thompson JC, Ishizuka J, Cooper CW, Levine MA, Gorr S-U & Cohn DV 1989 Inhibition of glucose-stimulated insulin release in the perfused rat pancreas by parathyroid secretory protein-I (chromogranin-A). Endocrinology 124 1235-1238. Grodsky GM 1972 A threshold distribution hypothesis for the package storage ot insulin and its mathematical modeling. Journal oj Clinical Investigation 51 2047-2059. Guyton AC & Hall JE 1996. Textbook of Medical Physiology. Philadelphia: WB Saunders. Helle KB & Angeletti RH 1994 Chromogranin A: a multipurpose prohormone? Acta Physiologica Scandinavica 152 1-10. Hellman B, Gylfe E, Grapengiesser E, Lund PE, Berts A, Tengholm A, Pipeleers DG & Ling 1994 Glucose induces oscillatory Ca2+ signalling and insulin release in human pancreatic beta cells. Diabetologia 37 (Suppl 2) S11-S20. Huttner WB & Natori S 1995 Helper proteins for neuroendocrine secretion. Current Biology 5 242-245. Huttner WB, Gerdes H-H & Rosa 1991 The granili (chromogranin/secretogranin) family. Trends in Biochemical Sciences
16 27-30.

Ahlborg

Kiens B, Hargreaves M & Richter EA 1991 Influence of active muscle mass on glucose homeostasis during exercise in humans. Journal of Applied Physiology 71 552557. Koeslag JH, Noakes TD & Sloan AW 1980 Postexercise ketosis. Journal of Physiology 301 79-90. Koeslag JH, Noakes TD & Sloan AW 1982 The effects of alamne, glucose and starch ingestion on the ketosis produced by exercise and starvation. Journal of Physiology 325 363376. Lovelock JE 1972 Gaia as seen through the atmosphere. Atmospheric Environment 6 579580. Margulis L & Sagan D 1995. What is life? London, Weidenfeld & Nicolson. Meda , Kohen E, Kohen C, Rabinovitch A & Orci L 1982 Direct communication of homologous and heterologous endocrine islet cells in culture. Journal of Cell Biology 92 221-226. Meda , Santos RM & Atwater I 1986 Direct identification of electrophysiologically monitored cells with intact mouse islets of Langerhans. Diabetes 35 232236. Michaels KL 8c Sheridan JD 1981 Islet of Langerhans dye coupling among immunocytochemically distinct cell types. Science 214 801-803. Milsum J H 1966 Biological Control Systems Analysis. New York: McGraw-Hill.

Kjaer M,

O'Connor DT, Wu H, Gill 13M, Rozansky DJ, Tang , Manata SK, Mahata M, Eskeland NL, Videen JS, Zhang X, Takiyyuddin MA & Parmer RJ 1994 Hormone storage vesicle proteins. Transcription basis of the widespread neuroendocrine expression of chromogranin A, and evidence of its biological actions, intracellular and extracellular. Annals of the New York Academy of Sciences 733
236-245. Orci L 1976 The microanatomy of the islets of Langerhans. Metabolism 25 (Suppl) 1303-1313. Orci L & Unger RH 1975 Hypothesis: Functional subdivisions of the islets of Langerhans and the possible role of the insular D-cells.
Lancet! 1243-1244. Orci L, Unger RH & Renold AE 1973 Structural coupling between pancreatic islet cells. Experientia 29 1015-1018. Orci L, Malaisse-Lagae F, Ravazzola M, Rouiller C, Renold AE, Perrelet A & Unger RH 1975 A morphological basis for intercellular communication between A- and B-cells in the endocrine pancreas. Journal of Clinical Investigation 56 10661070. Pipeleers DG 1992 Heterogeneity in pancreatic B-cell population. Diabetes 41 777-781. Pipeleers D, Kiekens R, Ling Z, Wilikens A & Schuit F 1994 Physiologic relevance of heterogeneity in the pancreatic beta-cell population. Diabetologia 37 (Suppl 2) S57-S64. Powell AM, Sherwin RS & Shulman GI 1993 Impaired hormonal responses to hypoglycaemia in spontaneously diabetic and recuiTentlv hypoglycmie rats, oiinnil <">/ Clinical Investigation l)2 2667-2674. Prentki M & Matschinsky FM 1987 Ca2+, cAMP, and phosphohpidderived messengers in coupling mechanisms in insulin secretion. Physiological Revieivs 67 1185-1248. Raskin P, Aydin I & Unger RH 1976 Effect of insulin on the exaggerated glucagon response to arginine stimulation in diabetes Diabetes 25 227-229. Riggs DS 1963 The Mathematical Approach to Physiological Problems. Baltimore: Williams & Wilkins. Riggs DS 1970 Control Theory and Physiological Feedback Mechanisms. Baltimore: Williams & Wilkins. Rorsman & Hellman 1988 Voltage-activated currents in guinea pig pancreatic a2 cells: Evidence for Ca~+-dependent action potentials. Journal of General Physiology 91 223242. Rorsman P, Ashcroft FM & Berggren P-O 1991 Regulation of glucagon release from pancreatic -cells. Biochemical Pharmacology 41 1783-1790. RoyJY, Bongbl J, Cardin S, Bnsson GR & Lavoie J-M 1991 Effects of supramaximal exercise on blood glucose levels during subsequent exercise. European Journal of Applied Physiology 63 4851.

Saunders PT 1994 Evolution without natural selection: Further implications of the Daisyworld parable. Journal of Theoretical Biology 166 365-373. Schmidt WE, Siegel EG, Lamberts R, Gallwitz & Creutzfeldt W 1988 Pancreastatin: molecular and immunochemical characterization of a novel peptide in porcine and human tissues. Endocrinology 123 1395-1404. Schuit FC, In't Veld PA & Pipeleers DG 1988 Glucose stimulates proinsulin biosynthesis by a dose dependent recruitment of pancreatic beta cells. Proceedings of the National Academy of Sciences,

mellitus.

USA 85 3865-3869. Sumida Y, Shima T, Shirayama K, Misaki M & Miyaji 1994 Effects of hexoses and their derivatives on glucagon secretion from isolated perfused rat pancreas. Hormone and Metabolism Research 26 222225. Tsuchiyama S, Tanigawa & Kato Y 1992 Impaired glucose priming of insulin from perfused pancreas in aged female rats. Proceedings of the Society for Experimental Biology and Medicine 201 5458. Unger RH, Aguilar-Parada E, Mller WA & Eisentraut AM 1970 Studies of pancreatic alpha cell function in normal and diabetic subjects. Journal of Clinical Investigation 49 837848. Unger RH, Dobbs RE & Orci L 1978 Insulin, glucagon, and somatostatin secretion in the regulation of metabolism. Annual Reviews of Physiology 40 307-343. Wahren J, Felig P, Hendler R & Ahlborg G 1973 Glucose and amino acid metabolism during recovery from exercise. Journal of Applied Physiology 34 838-845. Watson AJ & Lovelock JE 1983 Biological homeostasis of the global environment: the parable of Daisyworld. Tellus 35B 284289. Weir GC, Knowlton SD & Martin DB 1974 Glucagon secretion from the perfused rat pancreas. Studies with glucose and catecholamines. Journal of Clinical Investigation 54 14031412. Winder WW, Hickson RC, Hagberg JM, Ehsani AA & McLane JA 1979 Training-induced changes in hormonal and metabolic responses to submaximal exercise. Journal of Applied Physiology 46
766-771.

Winkler H & Fischer-Colbrie R 1992 The chromogranins A and B: the first 25 years and future perspectives. Neuroscience 49 497528. Zinman B, Murray FT, Vranic M, Albisser AM, Leibel BS, McClean PA & Marliss EB 1977 Glucoregulation during moderate exercise in insulin treated diabetics. Journal of Clinical Endocrinology and Metabolism 45 641-652.

Accepted

Received 29 July 1996 Revised manuscript received 5 November 1997

22

January 1997