An Analysis of the Effects of pH on Oxygen Binding by Squid (Illex Illecebrosus, Loligo Pealei) Haemocyanin

Among cephalopod haemocyanins, highly pH-dependent molecules with Bohr coefficients (ΔlogP₅₀)/ΔpH) below −1 predominate (Brix et al. 1981, 1989; Houlihan et al. 1982). This observation has cast doubt on whether Bohr and Haldane effects in these and other animals (e.g. some crustaceans, Morris et al. 1985) can support oxygen loading and CO₂ unloading of the blood at the gills and oxygen release and CO₂ loading in the tissues, in the classically accepted manner. Since Bohr and Haldane coefficients are expected to be identical (Wyman, 1964), the Haldane effect would cause high proton consumption during deoxygenation. The support of oxygen release in the tissues would require more protons than would be available from the formation and accumulation of CO₂ (linked to the consumption of transported oxygen and to realistic RQ values). Deoxygenation would cause an alkalosis, oxygen affinity would increase, and further oxygen release would require a large drop in venous ⁠.Consequently, instead of supporting venous unloading, large Bohr and Haldane effects would appear to obstruct the oxygen supply to the tissues.

Brix et al. (1981) proposed that the large Bohr effect acts predominantly to support oxygen uptake at the gills. Venous alkalization, it is suggested, is minimized by oxygen-linked CO₂ binding. According to this mechanism, both O₂ and CO₂ are bound to the haemocyanin at the gills and are released in the tissues (Lykkeboe et al. 1980). Finally, Brix et al. (1989) concluded that Bohr and Haldane shifts in cephalopod blood allow the haemocyanin to contribute to pH stabilization and prevent ‘any interference With the loading of O₂ at the gills’.

Is oxygen-linked CO₂ binding widely distributed among cephalopods? This study is designed to investigate oxygen and CO₂ binding in squid (Loligo pealei and Illex illecebrosus) haemocyanin. The potential role of pH in the oxygenation/ deoxygenation cycle is also addressed. For this purpose, a graphical depiction is chosen (Figs 1A and 2A) which presents changes in oxygen saturation with pH at various values. This presentation illustrates the dependence of haemocyanin oxygen-binding on pH and allows the conditions required for optimum function of the pigment in oxygen transport and delivery to be defined.

Squid (Illex illecebrosus, 300–500 g, Loligo pealei, 200–400 g) were caught by commercial fishermen in St Margarets Bay or close to Herring Cove, Nova Scotia, from October to December 1986. The animals were placed in plastic bags filled with oxygenated sea water at 2–6°C and transported to Halifax. There they were held in running sea water at ambient temperatures of 8–15 °C. At high ambient temperatures they were used as soon as they recovered from transport and handling (after 2–4 h). When ambient temperatures fell below 12 °C, the animals were brought close to the experimental temperature for 12-–24 h before being used.

Squid, some of them cannulated in the vena cava (H. O. Pörtner, D. M. Webber, R. G. Boutilier and R. K. O’Dor, in preparation), were placed in a Beamish-type respirometer filled with approximately 1001 of sea water at 15±0.5°C. Water flowed through the system continuously, some of it being circulated through the animal chamber at approximately 0.1m s⁻¹. After about 1.5 h of recovery from handling, a blood sample was taken, the respirometer was closed and the animal anaesthetized by adding 21 of pure ethanol to the water circulation downstream of the animal. Full anaesthesia (indicated by the cessation of ventilation) was reached after 3–5min. The animal was removed from the respirometer and quickly decapitated. Additional blood samples were collected from the gill hearts and the ventricle. Blood samples from all experimental animals were pooled, frozen and stored at close to −20°C until utilized for in vitro studies of oxygen binding.

Oxygen-binding characteristics of squid haemocyanin were studied using a tonometer (Instrumentation Laboratory, model 273, Padorno Dugano, Italy) on whole blood at 15 °C. Samples of pooled blood were equilibrated with humidified gas mixtures of varying (between 0 and 98.6kPa, lmmHg=0.1333kPa, 1 kPa=7.502 mmHg) provided by gas-mixing pumps (type M 303/a-F; Wösthoff, Bochum, FRG). Blood pH was varied by changing the (between 0.09 and 12.0 kPa) or by replacing small volumes (<10 μl per 2 ml of blood) of supernatant plasma after ultracentrifugation (1 h at 120000g; see Morris et al. 1985) with fixed acid (1moll⁻¹ HC1) or base (2moll⁻¹ NaOH), thereby causing a pH shift of approximately 0.3 units in the acid or alkaline direction, respectively. Total CO₂ was analysed in 50 μl blood samples using the gas chromatography method of Lenfant and Aucutt (1966) modified after Boutilier et al. (1985). The pH was measured for each data point by using a microcapillary pH electrode (G299, Radiometer, Copenhagen, Denmark) thermostatted at 15±0.1°C and calibrated with precision phosphate buffers (Radiometer, Copenhagen). 75 μl blood samples were analysed for oxygen content by measuring the increase in after release of bound oxygen in a solution containing 6 g l⁻¹ KCN and 3 g l⁻¹ saponin (Bridges et al. 1979) using a ‘Tucker’ chamber (Tucker, 1967).

Changes in haemocyanin oxygenation (measured as total oxygen minus physically dissolved oxygen) and pH were plotted in a pH/saturation diagram (Figs 1A, 2A). The resulting oxygen-binding curves appear as isobars delineating the change in oxygenation with pH at constant ⁠. The points of intersection of the oxygen isobars with the line of half saturation show at which pH of the isobar is equivalent to logP₅₀ of the pigment. The values of logP₅₀ are shown on the line in the upper part of the diagram. Since, from these intersections, the change in logP₅₀ with pH was found to be constant over the whole range of pH, the Bohr coefficient, ΔlogP₅₀/ΔpH could be evaluated by linear regression analysis.

Proton release or consumption during oxygenation/deoxygenation was calculated in accordance with the linkage equation of Wyman (1964), as used by Siggaard-Andersen (1974) and Brix et al. (1981). The Haldane coefficient (ΔHCO₃⁻ /ΔHcyO₂) was evaluated from the vertical distance between buffer lines (see Figs 5 and 6). The linkage equation is:

pH changes required to bring about deoxygenation can be estimated both from a consideration of Bohr coefficients and from the slopes of the oxygen isobars (in Figs 1A, 2A, 4A). The steepness of these slopes can be quantified from ΔHcyO₂/ ΔpH or ΔS/ΔpH (where S is oxygen saturation) at constant ⁠.

Along each isobar, values of S depend on pH values and the of the isobar. The pH/saturation diagram allows comparison of S, pH and with P₅₀ at S=0.5 and the same pH (=pH₅₀, Fig. 3A). This provides values of S, and logP₅₀ which are needed for the analysis of cooperativity at a specific pH. If this is done in the range of saturation S between 0.4 and 0.6, the analysis allows a close estimate of Hill coefficients (n₅₀) to be made according to (see Fig. 3B):

The pH-dependence of oxygen binding can be described by plotting the change in the of half-saturation (P₅₀) against H⁺ activity on a double logarithmic scale. pH effects on cooperativity can be evaluated in the region of P₅₀ as the change of slope of the oxygen-binding curve in the Hill plot. The Hill coefficient n₅₀ characterizes this change in cooperativity. pH effects for other ranges of oxygen saturation may be analysed in the same fashion. It rapidly becomes clear, however, that these analyses of pH effects focus on only one value of saturation at a time and do not provide a comprehensive illustration of pH effects within the whole range of oxygenation.

Such an illustration is also lacking ln the classical presentation of oxygen-binding curves using a /saturation diagram. For pH-sensitive pigments, the oxygenbinding curve shifts with pH; this can be taken into account by presenting several curves, each valid for one pH value. No pH scale is included in the presentation, with the result that the course of transitions in oxygen binding with changing pH cannot easily be quantified. In highly pH-sensitive pigments such as cephalopod haemocyanin this presentation is also problematical for practical reasons. During the analysis of oxygen binding with changing ⁠, the changes in oxygenation will change pH according to the Haldane coefficient. According to the Bohr coefficient, saturation values are affected by these pH changes and, consequently, the oxygen-binding curve will not be valid for a specific pH. For this reason, buffers have been used in pigment solutions to minimize pH fluctuations (e.g. Miller and Mangum, 1988). The use of buffers, however, is unphysiological and may not always be advisable (as discussed by Brix et al. 1989), for example, if oxygenbinding data in vitro and in vivo are to be compared.

The present study tries to improve on these drawbacks by analysing haemocyanin oxygen-binding in whole blood and applying a method of analysis that allows continuous and quantitative monitoring of pH effects. This is simply achieved by plotting saturation and pH (instead of ⁠) and including in the form of oxygen isobars. Figs 1A and 2A show the influence of pH and on the oxygenation of Illex illecebrosus and Loligo pealei haemocyanin in a pH/saturation diagram. This diagram is reminiscent of that presented by Redfield and Goodkind (1929) (also a Cartesian diagram), which showed total CO₂ (not pH) and total O₂ (not bound O₂). In the pH/saturation diagram, each oxygen isobar represents an oxygen-binding curve. At constant ⁠, transitions in oxygenation are depicted as they occur with changes in pH. Oxygen isobars show changing slopes which are linear in the range around P₅₀. Each isobar indicates saturation (S, x̄±S.D.) where it runs parallel to the pH axis (S=1.37±0.07mmol l⁻¹, N=20 for Illex illecebrosus, S=1.38±0.04mmol l⁻¹, N=14 for Loligo pealei).

Modulation of the haemocyanin oxygen-affinity by organic and inorganic cofactors and by temperature fluctuations would lead to changes in the location of the oxygen isobars. Modulation of cephalopod haemocyanins by organic cofactors can probably be excluded (C. R. Bridges, personal communication), while modulation by inorganic cofactors would largely depend on changes in the environment, which are unlikely to occur in the case of pelagic squid. ™

In whole blood, experimental variation of oxygenation always causes pH changes. This must be considered both when /saturation curves are constructed (see above) and when Hill coefficients (n₅₀) are analysed (Fig. 3B). For the latter, saturation values slightly smaller or larger than 0.5 are usually selected. Changes in oxygenation are elicited by varying ⁠. In the range of saturation below 0.5, PH is higher than pH₅₀ owing to proton consumption during deoxygenation. In the range of saturation above 0.5, however, pH is below pH₅₀. This means that the analysis of n₅₀ is affected by the Bohr effect shifting saturation to higher values at higher pH (saturation below 0.5) and to lower values at lower pH (saturation above 0.5). If these saturation values are used for evaluating the Hill coefficient, the real slope of the oxygen-binding curve at pHso is underestimated. The degree to which this error occurs depends on the magnitude of the Bohr effect (Fig. 3B). The error is independent of oxygen affinity. It is especially large with a high oxygen-binding capacity (that is, with high pigment concentrations in the blood) and at low ⁠, since, owing to the low CO₂/bicarbonate buffering, the pH change is large under these conditions. The error increases with increasing cooperativity and, therefore, may prevent the detection of a pH-sensitivity of cooperativity. Again, buffered pigment solutions may be used (DePhillips et al. 1969; Miller, 1985), but the graphical analysis (Fig. 3A) based on Figs 1A, 2A and 4A completely eliminates the introduced error, even for whole blood, since values of S, and logP₅₀ are evaluated for one specific pH.

Figs IB and 2B show the change in cooperativity with pH for the two squid species. Hill coefficients are found to be largest in the range of pH where oxygen isobars exhibit a maximum steepness (Figs IA, 2A). The maximum is more prominent in Illex and occurs at higher and lower pH values in Illex than in Loligo (see Table 2). In both species the largest Hill coefficients are found close to the range of in vivo blood and pH values (based on H. O. Pörtner, D. M. Webber, R. G. Boutilier and R. K. O’Dor, in preparation). The same is likely to be true for Octopus dofleini if pH values published by Lenfant and Johansen (1965) are compared with in vitro studies of buffered haemocyanin solutions (Miller, 1985; Miller and Mangum, 1988; Fig. 4A,B). The values of n₅₀ shown in Fig. 2B are close to the n_max values given by DePhillips et al. (1969) for buffered Loligo pealei haemocyanin solutions at the respective pH.

Analysis of total CO₂ during variations of and in tonometered blood gives the buffer lines shown in the pH/bicarbonate diagrams (Figs 5 and 6). In both species, the position of the buffer lines is different for oxygenated and deoxygenated blood. Non-bicarbonate buffer values (β) are close in the two squid species, that in Illex blood (5.0 mmol l⁻¹ pH unit⁻¹) being slightly smaller than that in Loligo blood (5.8 mmol l⁻¹ pH unit⁻¹). The vertical distance between the buffer lines for oxygenated and deoxygenated blood (Figs 5 and 6) yields the Haldane coefficient (ΔHCO3⁻/ ΔHcyO₂). In contrast to reports for other cephalopods (Brix et al. 1981; Lykkeboe et al. 1980), the Haldane coefficient does not depend on pH.

A comparison of measured and calculated apparent bicarbonate levels allows us to determine whether CO₂ was specifically bound to the haemocyanin. For this purpose values of CO₂ solubility and pK‴calculated according to Heisler (1986) were used. (This calculation procedure had previously been demonstrated to yield valid results in the body fluid of another marine invertebrate, Sipunculus nudus, H. O. Portner and N. Heisler, unpublished results.) Calculated and measured amounts of apparent bicarbonate agree in both oxygenated and deoxygenated squid blood. Consequently, total CO₂ values include only dissolved CO₂ and apparent bicarbonate levels. Oxygen-linked CO₂ binding, therefore, is not involved in haemocyanin function in Illex illecebrosus or Loligo pealei (or Octopus dofleini, Miller and Mangum, 1988).

The Bohr coefficients obtained by linear regression analysis from Figs 1A and 2A are ΔlogP₅₀/ΔpH=−1.07, r²=0.994 for Illex illecebrosus and ΔlogP₅₀/ΔpH=−1.15, r²=1.0 for Loligopealei. In the pH range between 7 and 8 the linear relationship between logP₅₀ and pH was not affected by the method used to alter pH [titration with CO₂, fixed acid (HC1) or fixed base (NaOH)]. This suggests that CO₂ or fixed acid have no specific influence on the change in oxygen affinity with pH; that is, CO₂ and fixed acid Bohr effects are identical. The values for both species are close to the one found by Brix et al. (1981) in Loligo forbesi and by Brix et al. (1989) in Todarodes sagittatus. In two other squid species (Loligo vulgaris and Architeuthis mariachis) Brix et al. (1989) found Bohr coefficients larger than −1, with the one in the mediterranean species Loligo vulgaris being as much as −0.4 at 20°C. The value found for Loligo pealei is larger than that (−1.8) formerly cited (Lykkeboe et al. 1980; Brix et al. 1989) but is close to the value of −1.3 (23°C) which can be recalculated from original data given for Loligo pealei by Redfield and Goodkind (1929, see Table 1). This also agrees with a value of — 1.16 recalculated from the data of DePhillips et al. (1969) for the pH range between 6.9 and 7.5 (23°C).

Bohr and Haldane coefficients are slightly smaller in Loligo than in Illex blood. For each species, the numerical values of Bohr and Haldane coefficients are virtually identical (cf. Figs 5 and 6). Oxygen-linked CO₂ binding would have caused the (experimental) Haldane coefficient to differ from the numerical value of the Bohr coefficient. The finding of an agreement is in accordance with the linkage equation (Wyman, 1964) and provides additional evidence that O₂-linked CO₂ binding does not contribute to haemocyanin function in the two squid species.

Starting from an initial in the arterial blood, the pigment releases oxygen on its way through the tissues, thus preserving tissue ⁠. In this way, oxygen gradients are kept high to allow for a high diffusive oxygen flow. By depicting the change in oxygenation at a given ⁠, the pH/saturation diagrams (Figs IA, 2A) elucidate this buffer function of the pigment. They demonstrate how pH must change to maintain a high venous ⁠. The slope of the resulting isobar indicates its sensitivity to pH. Obviously, this kind of pH sensitivity is different in different ranges of pH and and cannot be determined by an analysis of the Bohr coefficient. In fact, the magnitude of the Bohr coefficient does not change with pH (see above).

More likely, as suggested by the observation that the maximum cooperativity corresponds to the maximum steepness of the isobars, the change in pH sensitivity may be explained by a change in cooperativity with pH. In this case the oxygen isobar, the steepness of which is given by the ratio ΔHcyO₂/ΔpH or ΔS/ΔpH would vividly illustrate and quantify the physiological meaning of a sensitivity of cooperativity to pH (see below). The linear slope is specific for a combination of P₅₀ and pH₅₀ values and is valid for up to 80–90 % of the range of saturation

(Table 2).

Fig. 7 demonstrates how pH changes may influence oxygen binding ln the living animal. Lines B and C reflect two possible pH changes between arterial and venous blood. Line C results from the large Haldane effect, if oxygen is provided only via the blood (not through the skin, see below) and if RQ values below 1 prevail (Table 3). The alkalosis causes venous to fall more than would be observed with a pH-insensitive pigment.

Line B is in accordance with a buffer function of the pigment and follows the isobar for =7.9kPa. This isobar exhibits a steeper slope than the isobar of = 12.9 kPa. 12.9 kPa may be close to the arterial oxygen tension (Redfield and Goodkind, 1929). If, in the model, 7.9 kPa is the venous ⁠, this isobar would be crucial for the process of oxygen release. The change from 12.9 to 7.9 kPa would represent an arteriovenous oxygen gradient and would illustrate how the pH sensitivity of the pigment could increase between arterial and venous blood and, thereby, minimize the required arteriovenous pH gradient. The pH change of the blood as it passes through the tissue can be quantitatively predicted as deoxygenation proceeds. Such a pH change has not been confirmed in squid since the early study of Redfield and Goodkind (1929). These authors, however, analysed blood sampled from the branchial hearts of restrained (nailed down) animals (Loligo pealei, for the recalculated pH values see Table 1), an experimental situation that may yield different results from those to be observed in unrestrained animals.

This discussion and the observation that pH sensitivity is maximized in the range of in vivo pH suggest that deoxygenation in vivo may be caused predominantly by pH changes rather than by fluctuations. If the pigment actually represents a buffer system for ⁠, the evolution of pH sensitivity clearly renders deoxygenation without concomitant acidification disadvantageous (Fig. 7, line C). Cooperativity and the Bohr effect may, therefore, be coordinated to optimize pH sensitivity with respect to pH fluctuations occurring ln the animal. If pH sensitivity by cooperativity is maximized in the range of in vivo values, this would suggest that cooperativity compensates for the high proton need during deoxygenation; that is, the high Bohr and Haldane effects. Not much acidification in excess of proton consumption is needed to support venous deoxygenation.

A pH drop in arterial blood would initially follow the direction of line A in Fig. 7. High oxygenation (S=1) is maintained. 5 is not maintained, however, when the arterial isobar is deflected in a downward direction. If the 12.9 kPa oxygen isobar reflects the dependence of arterial oxygenation on pH, haemocyanin saturation can no longer be maintained 0.1–0.15 units below arterial pH. Thus, as already outlined by Redfield and Goodkind (1929), an acid shift in arterial pH would be lethal to the animals.

In the range of pH down to below 6.7–6.8 in Illex blood (not shown in Figs 1 and 7), the sensitivity to pH leads to a further decrease in saturation. Even at values of 93.3 kPa, which are far above the physiological range, saturation values remain below those observed at higher pH (S=1). Maximum saturation observed at these high values and at low pH is equivalent to S=0.8 at pH 6.7 or S=0.6 at pH6.3. As a corollary, the scope of oxygenation is reduced. It remains to be investigated whether this effect (which was also observed in Octopus blood by Miller and Mangum, 1988) is of any physiological significance, since it occurs outside the range of pH and values measured in cephalopods in vivo.

The slopes of the oxygen isobars are less steep at higher and at lower pH. In the range of low venous pH (e.g. during muscular activity, following line D, Fig. 7, in the acid direction), there is an increasing amount of bound oxygen which is resistant to further pH-dependent deoxygenation and which can only be depleted by the lowering of (Figs 1A, 2A, 4A). This conclusion is based on the observation that, at low saturation, the slope of each isobar decreases with falling pH until it runs almost parallel to the pH axis. With increasing ⁠, this occurs at increasingly higher values of saturation. This may indicate a venous reserve that is protected against excessive pH changes. Such a pH-insensitive reserve would diminish any overshoot in caused by large pH variations and may help to maintain an equal distribution of oxygen throughout a tissue. Excessively high venous values would, for example, reduce the O₂ gradient from the environment to the tissue and, thus, reduce the amount of oxygen taken up via the skin (see below).

Under conditions of low arterial and high pH (e.g. during hyperventilation under hypoxia, following line A, Fig. 7, in the alkaline direction) a reduction in pH sensitivity may also support a more equal distribution of the oxygen reserves to the tissue. This is only valid, however, if venous pH remains largely unaffected by hyperventilation. Progressive alkalization of arterial blood would cause an increase in haemocyanin oxygen-affinity, thus compensating for a drop in ambient ⁠. However, an increase in venous pH would cause very low venous values, because cooperativity would be reduced and there would be less compensation for the large Bohr and Haldane effects. Therefore, an increase in the arteriovenous pH gradient would be required for the maintenance of pigment function.

The conclusion from the foregoing discussion is that a decrease in pH of the blood as it passes from the arterial to the venous side would be consistent with a buffering function of the haemocyanin (Fig. 7, line B). Under aerobic steadystate conditions venous pH can only decrease by the build-up of an arteriovenous gradient. However, such a build-up is obstructed by the large Haldane effect. To compensate for this, a mechanism of oxygen-linked, pH-independent CO₂ binding has been proposed (Lykkeboe et al. 1980; Brix et al. 1981; see Introduction). According to this mechanism, both O₂ and CO₂ are transported to the tissues and both the CO₂ produced by metabolism and the CO₂ released during deoxygenation lead to the build-up of the tissue.

Alternatively, the uptake of significant amounts of oxygen via the skin could lead to additional CO₂ formation by metabolism. Such oxygen uptake can occur, especially in regions of the body that are densely vascularized and close to the body surface, such as the outer layers of the mantle musculature (Bone et al. 1981). An elevated metabolic rate would cause to increase in the tissue and, therefore, in the venous blood. This process could provide the excess protons needed to acidify the blood and to compensate for Bohr coefficients smaller than −1.

An increase in the amount of CO₂ being transported by the blood would be detectable if the respiratory quotient were measured (a) for the blood going to and returning from the tissues, and (b) for the whole animal. High RQ values of 1.1 were obtained by Redfield and Goodkind (1929) from measurements of oxygen and CO₂ transport in Loligo pealei blood. Similar RQ values were also obtained from blood gas transport data in unrestrained, cannulated Octopus dofleini (Johansen and Lenfant, 1966). For the whole animal, however, RQ values greater than 1 are exceptional under aerobic steady-state conditions and, if they occur, they are thought to be related to high rates of lipid biosynthesis from carbohydrates. This is improbable in squid, where protein and amino acid catabolism are likely to predominate (Hoeger et al. 1987). The observation that blood RQ values exceed those expected for the whole animal supports the conclusion that additional CO₂ is being produced and released into the blood. The additional CO₂ cannot originate from oxygen-linked CO₂ binding, since this process does not influence the net transport of CO₂ from the tissues to the gills. It is therefore very likely that oxygen uptake via the skin is responsible for the arteriovenous gradient and may allow for a decrease of pH in venous blood below that of arterial blood. Further evidence for oxygen uptake via the skin is discussed by O’Dor et al. (1990).

In conclusion, the graphical presentation selected for Figs 1A, 2A and 4A allows Bohr and Hill coefficients to be analysed. By integrating the Bohr effect and the changes in cooperativity with pH, the diagram illustrates the physiological implications of pH sensitivity for both oxygen affinity and cooperativity. In squid haemocyanin, cooperativity probably helps to compensate for the negative influence of a high proton uptake during deoxygenation (Haldane effect) on venous ⁠.

The diagram demonstrates the in vitro properties of squid haemocyanin and also appears to be suitable for a presentation of arteriovenous changes in pH and in vivo (Fig. 7). As in a pH/bicarbonate/ diagram, the whole range of combinations of pH, and oxygen saturation values is covered. The analysis of in vitro data indicates how pH changes can play an important role in squid haemocyanin oxygenation/deoxygenation cycles. The question of whether the predictions of Figs 1A, 2B and 7 can actually hold in vivo in squid is the subject of further analysis (H. O. Pörtner, D. M. Webber, R. G. Boutilier and R. K. O’Dor, in preparation). The extent to which the graphical presentation is generally applicable to depict the properties and function of other pH-dependent respiratory pigments in vitro and in vivo remains to be investigated.

This study was supported by Deutsche Forschungsgemeinschaft Po 278/1-1, by NSERC operating and equipment grants to R. G. Boutilier and R. K. O’Dor, and by NSERC infrastructure support for the Aquatron Laboratory at Dalhousie University. The help of D. M. Webber, R. K. O’Dor and N. Balch in providing the animals is gratefully acknowledged. The author wishes to thank R. Glosch for help with the artwork, H. Decker and C. R. Bridges for helpful discussions and C. P. Mangum for helpful comments on a first draft of the paper.