How the house sparrow Passer domesticus absorbs glucose

The mechanism(s) by which house sparrows and many other species of birds absorb the majority of their dietary sugars is somewhat of a mystery. All vertebrates that have been studied possess intestinal sugar transporters such as the Na-coupled glucose transporter SGLT1 in the apical membrane and GLUT2 in the basolateral membrane (Karasov and Hume, 1997). In mammals such as the rat, mouse and rabbit, the intestine's absorptive capacity by these mediated mechanisms, as measured in vitro, meets or exceeds dietary carbohydrate intake(Ferraris and Diamond, 1989). In contrast, in many avian species(Karasov and Cork, 1994; Caviedes-Vidal and Karasov,1996; Levey and Cipollini,1996; Afik et al.,1997) the rates of mediated glucose transport measured in vitro were much lower than rates of glucose absorption actually achieved by freely feeding birds.

Pappenheimer and colleagues(Pappenheimer and Reiss, 1987; Pappenheimer, 1990) suggested that the intestine's capacity to absorb glucose by mediated pathways is inadequate to meet the daily intake of glucose and that most absorption occurs by solvent drag across intestinal tight junctions (paracellular absorption),secondary to active sugar and amino acid transport. In in vivoexperiments, a number of wild avian species achieved nearly complete absorption of ingested l-glucose, the stereoisomer of d-glucose that does not interact with the intestine's glucose transporters and can only be absorbed passively(Chang et al., 2004): in nectarivorous rainbow lorikeets, 80%(Karasov and Cork, 1994);granivorous house sparrows, 80%(Caviedes-Vidal and Karasov,1996); and omnivorous yellow-rumped warblers, 91%(Afik et al., 1997). These findings in birds with diverse diet and taxonomic associations suggest that,in birds, passive absorption is important for nutrient intake. However,Schwartz et al. (1995) pointed out that l-glucose might be absorbed at a much slower rate than d-glucose, but over the entire length of the intestine and the extended time of digesta residence in the gut, its absorption could still be fairly complete. Many birds have intestine lengths and digesta residence times that are relatively short, not long(Karasov and Levey, 1990),however, so how could birds exhibit nearly complete glucose absorption with low rates of mediated and passive uptake?

An elegant approach to resolving this issue has been to compare the extent and/or rate of absorption of l-glucose (absorbed only passively) vsd-glucose or its analogue (absorbed actively and passively) simultaneously in intact animals. In laboratory rats, the absorption rate of the nonmetabolizable, actively transported 3-O-methyl-d-glucose apparently exceeded that of l-glucose by about 9:1, implying that most glucose was absorbed actively (Uhing and Kimura,1995). Similar conclusions have been drawn for dogs(Lane et al., 1999) and humans(Fine et al., 1993). In this study, we report the first measurements of this kind in an avian species, and the findings are very different from those in the mammals studied so far.

Based on the earlier findings in birds, we predicted that passive absorption must be important to glucose absorption, and we applied a pharmacokinetic method focusing on appearance of radiolabeled 3-O-methyl-d-glucose (3OMd-glucose) and l-glucose in blood to measure absorption in intact birds. Both of these compounds are metabolically inert, which removes the possibly confounding factor of post-absorptive catabolism from the analysis, but only the former is transported by mediated processes at the brush border of house sparrows (Chang et al., 2004)and at the basolateral membrane (Kimmich,1981; Burant and Bell,1992). If mediated absorption is relatively most important, then this will be apparent in slower absorption of l-glucose over an extended period of time, relative to 3OMD-glucose. Alternatively, if the majority of glucose is absorbed passively, then the absorption of the radiolabeled 3OMd-glucose and l-glucose will be similar in rate and extent (Fig. 1).

Because of uncertainty about normal luminal nutrient concentrations(Ferraris et al., 1990; Pappenheimer, 1993), and how they might influence the comparisons between the radiolabeled compounds, we took measurements under substrate conditions that were both nonsaturating and relatively more saturating for SGLT1, whose half-saturation glucose concentration (apparent K_m) is 5 mmol l^-1(Caviedes-Vidal and Karasov,1996). Also, we point out at the outset that in our comparison,3OMd-glucose is handicapped relative to l-glucose for several reasons. First, the molecular mass of 3OMd-glucose (194.2 Da) is greater than that of l-glucose (180.2 Da), which lowers its diffusion coefficient in water and may decrease its rate of permeation,relative to l-glucose, through the paracellular space, which discriminates according to molecular size(Chediack et al., 2003). Also,the affinity of the glucose transporters for 3OMd-glucose is lower than for d-glucose (Kimmich,1981; Ikeda et al.,1989), so the former is an imperfect substitute for the latter. To evaluate the implications of these differences, and other assumptions, on our conclusions, we utilize a simulation model in conjunction with our results.

House sparrows Passer domesticus L. were captured in the vicinity of the University of Wisconsin-Madison, USA. They were housed in cages (0.80 m×0.60 m×0.60 m) indoors under relatively constant environmental conditions (25.0±0.5°C, relatively humidity 50±9%,photoperiod 12 h:12 h light:dark) and received ad libitum water and commercial bird seed mix. Sparrows were acclimated to laboratory conditions for at least 2 weeks prior to use in experiments. The routine animal care procedures and experimental protocol used in this study were reviewed and approved by the University of Wisconsin Research Animal Resources Center.

Food was removed from birds overnight prior to an experiment, when these diurnal birds would not normally eat, and measurements began 2.5-3 h after lights went on in the morning. When measuring absorption under conditions relatively nonsaturating for SGLT1, seven house sparrows were gavaged with 500μl of a solution containing 200 mmol l^-1 mannitol, 10 μCi[¹⁴C]l-glucose and 20 μCi[³H]3OMd-glucose (American Radiolabeled Chemicals St Louis, MO, USA), and 80 mmol l^-1 NaCl. When measuring absorption under relatively saturating conditions, the mannitol was replaced with 200 mmol l^-1 3OMd-glucose (N=6). To determine elimination rate for absorption calculations, these same individuals were injected into the pectoralis on other days with[¹⁴C]l-glucose and[³H]3OMd-glucose (5×μCi and 10 μCi,respectively, in 250 μl 200 mmol l^-1 mannitol and 80 mmol l^-1 NaCl). The sequence of trials was random, and repeated measures on individuals were performed after intervals of at least 7 days, easily enough time for residual radioactivity to return to background (normally less than 24 h; see Results). The total osmotic pressure of gavage or injection solutions was controlled at 360 mOsm with 5% variation (tested using a vapor pressure osmometer; Wescor vapor 5502, Logan, UT, USA) so that solutions were isosmotic with avian blood, and aliquot samples were saved for radioactivity analysis. Following either gavage or injection, 7-9 blood samples (20 μl each; total volume collected was <10% of blood volume in normal house sparrows; Stangel, 1986) were collected from the brachial vein with heparinized capillary tubes over the next 3-4 h. Immediately after collection, blood was centrifuged (1500 g) for 5 min, plasma was separated and weighed(10-20±0.1 mg), then mixed with 2.0 ml of scintillation fluor(Hionic-Fluor, Packard, USA) and counted in a scintillation spectrometer(Tracor Analytic, MarkIII, USA) for disintegrations per minute (d.p.m.) with automatic external standardization (AES) and background correction(Caviedes-Vidal and Karasov,1996).

In order to test for radiopurity, plasma samples from three different birds gavaged with different radioactive solutions were collected and analyzed along with three aliquot samples from each radioactive solution. Radiopurity was checked by high performance liquid chromatography (HPLC) using an NH₃ column (Alltech, USA) with an acetonitrile:water (85:15%)mobile phase. All ³H activity was associated with 3OMd-glucose in the solution (92%) and in plasma (91%) and all ¹⁴C activity was associated with l-glucose in the solution (97%) and in plasma (93%). We also checked and found in preliminary injection and gavage experiments (not shown) that we obtained similar results for absorption of l- and 3OMd-glucose regardless of whether the ¹⁴C and ³H labels were on l-glucose or 3OMd-glucose.

This method of calculating F is favored because it makes no major assumptions about compartments or kinetics. Fractional absorption estimates how much of the gavaged probe was absorbed into the animal's system.

In the two-compartment model, rate constants and distribution spaces are derived from the constants A, B, α and β. For example, the elimination rate constant of the probe from the apparent central compartment,to which we will refer later, is estimated as (A+B)/[(A/α)+(B/β)]. The rate constants and spaces are used in conjunction with the data from gavage experiments to calculate the cumulative proportion of F that was absorbed at each blood sampling time point (P_t),according to the Loo-Riegelman method(Wagner, 1975; Gibaldi and Perrier, 1982). The true cumulative absorption at any sampling time point (t) is the product of F and P_t, and the apparent absorption rate during two adjacent sampling time points is FΔP/Δt.

In the few cases where a mono-exponential model fit elimination best, or if elimination data were not available because an injection experiment was not performed (4 of 13 birds), we used the mean distribution space measured in other birds and applied the one-compartment model(Wagner, 1975) for calculating F, P and apparent rates. As discussed by Wagner(1975), the two models can yield very similar results under many circumstances, which we also found to be the case (N=9). Also, fractional absorptions did not differ significantly when estimated using a compartmental model versus the compartment-independent method [e.g. for l-glucose respectively,0.73±0.03 (mean ± s.e.m.) vs0.72±0.04; P>0.6, N=9]. Although the two-compartment model is more complex, we retained it when appropriate because, theoretically, the two-compartment model would be better able to detect experimental differences (Loo and Riegelman, 1968; Wagner,1975; Gibaldi and Perrier,1982). If there are some organs and tissues in an organism where blood circulation is not as high as the rest of body, this may lead to a higher concentration of probe in these peripheral compartment(s) compared to the central compartment. Thus these poorly perfused peripheral compartments can serve as another source for probes entering the central compartment,secondary to an absorption site like the small intestine. In comparison to the one-compartment model, a two-compartment model is more conservative in that it provides corrections for the redistribution of probe from these peripheral compartments (Gibaldi and Perrier,1982; Riviere,1999).

Numerical data are presented as means ± s.e.m.(N=number of animals). Although data shown in the figures are mean values, statistical analyses were performed based on data for individuals. Fractional absorption (F) was arcsine square-root transformed before statistical analysis. Results were analyzed by analysis of variance (ANOVA),repeated-measures ANOVA, and Student's t-test (SAS, SAS Institute Inc, Cary, NC, USA). The T- and F-values of these and other ANOVAs are presented in the text with the relevant degrees of freedom as subscripts. Linear regression was by the method of least squares. Nonlinear curve fitting (Gauss-Newton algorithm, SYSTAT; Wilkinson, 1992) was used to fit kinetic data, and kinetic models were compared according to Motulsky and Ransnas (1987). Statistical significance was accepted for P<0.05. One-tailed tests were used for a priori predictions.

When the birds were gavaged with [¹⁴C]l-glucose and[³H]3OMd-glucose simultaneously, the average concentration of both radiolabeled carbohydrates in the plasma peaked within 30 min and then declined exponentially(Fig. 2). At least three patterns were apparent from simple inspection of these plots. First, the time course for absorption of both probes seemed rather similar. Second, within each treatment group, the plasma values of 3OMd-glucose slightly exceeded those of l-glucose. Areas under the curve (AUCs) for 3OMd-glucose were significantly higher than for l-glucose in both treatment groups(Table 1). Third, simply comparing peak concentrations, there seemed to be more of both probes appearing in blood when the measurement was made under more saturating conditions (i.e. 200 mmol l^-1 3OMd-glucose in the gavage solution) than when under relatively nonsaturating conditions (200 mmol l^-1 mannitol replaced 3OMd-glucose). Indeed, the AUCs for both probes were significantly higher when measured under relatively saturating compared with nonsaturating conditions(Table 1). Observations two and three suggest, respectively, that the fractional absorption of 3OMd-glucose exceeded that of l-glucose and that absorption of the probes was enhanced when measured under conditions that are relatively more saturating for SGLT1. Although there is merit in this visual inspection of the patterns in these plots, because it affords the reader a simple and direct way to evaluate the data, less intuitive but more exact calculations of fractional absorption require additional data from the injection/elimination experiments.

In the injection/elimination trials the probes equilibrated rapidly,because the first blood samples at 7 or 10 min had the highest probe concentrations (i.e. no extended lag for distribution from pectoralis to blood; Fig. 3). Birds in the two treatment groups did not differ significantly from each other in AUCs(Table 1) or parameters fitted to the bi-exponential elimination model(Table 2) (all P>0.2). However, 3OMd-glucose was eliminated more slowly from the apparent central compartment than was l-glucose(respectively, 0.016±0.001 vs 0.028±0.002 min^-1, paired T₈=14.5, P<0.001),with the result that the post-injection AUC was larger for the former than for the latter (Table 1). This difference, which may relate to kidney reabsorption and hence longer retention of [³H]3OMd-glucose relative to[¹⁴C]l-glucose, underscores why the extent and rate of absorption of the two probes cannot be compared simply from comparison of plasma probe concentrations and their AUCs after gavage (i.e. Fig. 2). But, the qualitative conclusions drawn from inspection of Fig. 2 were borne out by the more exact calculations of F that take into account the probes' respective elimination kinetics. Fractional absorptions were significantly higher for 3OMd-glucose than for l-glucose when measured simultaneously under relatively nonsaturating conditions (respectively, 0.71±0.02 vs0.66±0.03; paired T₆=2.53, P=0.045),although under more saturating conditions the two did not differ significantly(respectively, 0.8±0.05 vs 0.79±0.04; paired T₅=0.8, P>0.4). Statistical comparison of these values across the treatments showed that fractional absorption was significantly enhanced for [¹⁴C]l-glucose(T₈=2.42, P=0.042), but not for the[³H]3OMd-glucose (T₅=1.5, P>0.3), when measured under conditions that were relatively more saturating for SGLT1.

Information on the apparent rates of probe absorption was derived from plots of fractional absorption as a function of time(Fig. 4A,B). Apparent absorption rates declined with increasing time since gavage, and the rates were compared within the trials using a repeated-measures ANOVA. Under conditions relatively nonsaturating for SGLT1(Fig. 4A), the least-squares adjusted mean absorption rate for 3OMd-glucose (1.91±0.15%absorbed min^-1) significantly exceeded that for l-glucose absorption (1.63±0.14% absorbed min^-1)by 17% (F_1,62=4.01, P=0.049). Under relatively more saturating conditions (Fig. 4B), the apparent absorption rates of the two probes did not differ significantly (respectively, 2.59±0.38 vs2.67±0.42% absorbed min^-1; F_1,45=0.1, P>0.7). Statistical comparison of these values across the treatments showed that apparent rate of absorption was significantly enhanced for the l-glucose probe by 57% (F_1,52=24, P<0.01) and for the 3OMd-glucose probe by 36%(F_1,55=8.28, P<0.001), when measured under conditions that were relatively more saturating for SGLT1.

Assuming that absorption of l-glucose is a proxy for passive absorption of 3OMd-glucose, whereas the absorption of 3OMd-glucose represents the sum of passive + mediated absorption,the ratio of the apparent absorption rates (L/D) indicates the proportion of 3OMD-glucose absorption that occurs via the passive pathway. Accordingly, we plotted this ratio over the course of the trials(Fig. 5). The ratio always exceeded 0.7, with no significant difference between treatments. This suggests that more than 70% of 3OMd-glucose absorption probably occurred via the passive pathway, regardless of whether the measures were made under conditions relatively nonsaturating or relatively saturating for SGLT1. However, the following discussion revisits a number of assumptions upon which this conclusion is based.

In the following sections we discuss our two main empirical results. First,when [¹⁴C]l-glucose and[³H]3OMd-glucose were gavaged simultaneously their extent and rate of absorption were very similar, never differing by even 20%. Second, the apparent rates of absorption increased when the measurements were made under conditions relatively more saturating for the brush border glucose transporter SGLT1. Part of our discussion includes a simulation that tests major assumptions important for interpreting the physiological significance of the first major result, and the simulation also aids in interpreting the second major result from a mechanistic perspective.

If 3OMd-glucose is mainly actively absorbed by the intestine,then a much higher rate of absorption of[³H]3OMd-glucose than of[¹⁴C]l-glucose is predicted. For example, if 95% of the instantaneous glucose absorption is active, as suggested in some studies(Uhing and Kimura, 1995; Fine et al., 1993; Lane et al., 1999), and if 3OMd-glucose is a good proxy for D-glucose, then the ratio of the absorption rate for l-glucose to 3OMd-glucose should be about 0.05. However, we found similar fractional absorption and apparent rates of absorption for l- and 3OMd-glucose in house sparrows(Fig. 4) and the ratio never fell below 0.7 (Fig. 5),indicating >70% passive absorption of glucose.

Schwartz et al. (1995) made the point that simply comparing F for 3OMd- and l-glucose might be misleading. Suppose that 3OMd-glucose is absorbed at a high rate in the proximal portion of the intestine, whereas l-glucose is absorbed at a very slow rate. The fractional absorption of l-glucose could still be fairly complete if its slow absorption occurred over the entire length of the intestine and over the entire time of digesta residence. We do not think that this explanation applies to the house sparrows. Absorption of l-glucose did not seem prolonged compared with that for d-glucose (insets in Fig. 4). 3OMd- and l-glucose had apparent absorption rates similar to each other throughout all the sampling time points(Fig. 4). When measurements were done under conditions relatively nonsaturating for SGLT1, even for the largest difference at 10 min after gavage the apparent absorption rate of 3OMd-glucose was only 1.32-times higher than that of l-glucose. When measurements were done under more saturating conditions, the fractional absorption of l-glucose was the same as that for 30Md-glucose, the apparent absorption rate of 3OMd- and l-glucose followed each other closely throughout the course of experiment, and the mean apparent absorption rates were not statistically different from each other either.

In the Introduction we suggested that in the comparison of 3OMd-and l-glucose, the former was handicapped relative to the latter for several reasons. First, we previously found that the fractional absorption of carbohydrates absorbed passively by the house sparrow declines with increasing molecular mass (MW) of the probe, by an average of -0.002 Da^-1 (Chediack et al.,2003). The reasons for this probably include the fact that diffusion coefficients decline with increasing MW^1/2, and the paracellular space discriminates according to molecular size, much like a sieve (Chang et al., 1975; Friedman, 1987). Considering the 14 Da difference in the molecular mass of l-glucose and 3OMd-glucose, the direct comparison of their fractional absorptions(averaging, respectively, 0.73 and 0.76) might be adjusted by increasing the value for 3OMd-glucose by 0.03 (i.e. the product of 14 Da and 0.002 Da^-1). This would not change the overall conclusion that their extent and rate of absorption are very similar when measured simultaneously.

The affinities of both the brush border and basolateral glucose transporter(s) are lower for 3OMd-glucose than for d-glucose (Kimmich,1981; Ikeda et al.,1989), so the former is an imperfect substitute for the latter. To evaluate the implications of this difference for our conclusions, we utilized a simulation model of absorption by mediated and passive pathways that we discuss in the next section. Thinking about how the[³H]3OMd-glucose interacted with a glucose transporter also helped to highlight what was notable about our second major empirical finding - that the rates of absorption increased, even for[³H]3OMd-glucose, when the measurements were done under conditions relatively more saturating for the glucose transporter.

In order to characterize absorption of d-, l- and 3OMd-glucose in intact animals, we constructed a model for absorption at the apical membrane using idealized simple reactor theory,following the approach of Penry and Jumars(1987), who proposed that tubular organs such as the intestine are analogous to plug-flow reactors in chemical engineering. In a plug-flow reactor, material is mixed instantaneously and continuously in the radial direction without appreciable axial mixing, and items leave the tube in the same order that they entered. At low and high substrate concentrations, we simulated several possible patterns of 3OMd-glucose absorption, including expected patterns if mediated absorption were dominant, but with different apparent affinities between the transporter and its substrate, and an expected pattern in the event that passive absorption were dominant. We used the results as a check on our conclusions based on empirical results.

Paracellular absorption of hydrosoluble molecules across small intestinal mucosal epithelium is best described in the context of the Kedem-Katchalsky equation (Kedem and Katchalsky,1958), which includes the contribution of both diffusion and solvent drag to the flux of solutes through porous epithelia. K_a is thus a lumped first-order coefficient of both diffusion and solvent drag flux, which both occur in linear proportion to substrate concentration C_t.

Here, C_DPt is the 3OMd-glucose probe concentration (nmol μl^-1) at time t at the absorption site. Luminal unlabeled 3OMd-glucose concentration at time t (C_Ct) can be described by Equation 4 with any given initial concentration at time zero.

In the simulation, the reaction or absorption rate at time t is calculated by formulae as described above with any given initial substrate concentrations. Assuming that the volume of gut contents does not change with time and that the absorption rates stay unchanged from time t to time t+1, the absorption rate times the intestine luminal volume approximates how much substrate is absorbed during the time interval. By subtracting this value from the initial total substrate content at time t, the initial substrate concentration at time t+1 can be calculated from the remainder. With this initial substrate concentration at time t+1, the subsequent absorption rate and the amount of absorption during the next time interval can also be calculated following the same rule. Plotting the values gathered from a simulation run from t=0 to t=50 min, when most absorption had occurred in theory and in fact(Fig. 4A,B), we estimated the relative rate and extent of [¹⁴C]l-glucose and[³H]3OMd-glucose absorption. The selection of initial parameter values is described in the legend of Fig. 6.

The simulated cumulative absorption (Figs 6A, 7A) and apparent rate of absorption (Figs 6B, 7B) of radiolabeled 3OMd-glucose are considerably greater than that of radiolabeled l-glucose in all the situations modeled (i.e. Figs 6Ai-iii,Bi-iii, 7Ai-iii,Bi-iii), except when the apparent passive permeability coefficient was increased (compare Figs 6Aiv,Biv, 7Aiv,Biv with all other panels). Adjusting upward the apparent K_m for mediated 3OMd-glucose absorption (Figs 6Aii, 7Bii) had little effect on this comparison. The predicted rates of probe absorption were in most cases depressed in the presence of high luminal d-glucose concentration(c.f. Figs 6B, 7B), as might be expected because the labeled probe competes with non-labeled substrate for uptake from the lumen. The one situation in which high luminal substrate concentration had little effect on predicted probe absorption rate was when the apparent passive permeability coefficient was increased, and this makes sense because most of the predicted probe absorption is nonmediated and thus not competitively inhibited.

The simulation results support our interpretation of the empirical results,and they add one new insight. First, simulation results are consistent with the empirical findings that absorption by intact house sparrows was not dominated by the mediated pathway. The simulations indicate that when passive absorption dominates, then absorption of 3OMd- and l-glucose will be similar in rate and extent (see especially Figs 6Aiv,Biv, 7Aiv,Biv), which is what we observed empirically (Fig. 4). In contrast, the simulations indicate that if mediated absorption had dominated, then the absorption of 3OMd-glucose should have been more rapid and extensive than that of the l-glucose, even despite the fact that the 3OMd-glucose has a lower affinity than d-glucose for the brush border glucose transporter (see especially Figs 6Aiii,Biii, 7Aiii,Biii), but this was not observed empirically.

Second, the simulations help us interpret a second empirical finding - that when the measurements were done under conditions relatively more saturating for the glucose transporter, the rates of absorption increased, even for[³H]3OMd-glucose (c.f. Fig. 4A,B). In contrast to this empirical finding, the simulation indicated that if most[³H]3OMd-glucose uptake were mediated, then its absorption would have been inhibited by unlabeled substrate in the lumen rather than accelerated. However, studies in mammals(Pappenheimer and Reiss, 1987; Madara and Pappenheimer, 1987; Pappenheimer, 1987; Pappenheimer and Volpp, 1992; Sadowski and Meddings, 1993; See and Bass, 1993; Turner and Madara, 1995; Chang et al., 2004) and in birds (Chediack et al., 2003)have shown that hydrophilic molecules cross the intestinal mucosa by diffusion and/or solvent drag via a paracellular pathway, and that the permeability of this pathway is enhanced when Na⁺-coupled nutrient transport occurs. The mechanism(s) for acceleration by luminal glucose is not known in house sparrows, but might be increased solvent drag and/or cytoskeletal contractions (Madara and Pappenheimer, 1987; Pappenheimer, 1987; Madara et al., 1986, 1988) or protein strand alterations that alter the tight junction effective pore size(Pappenheimer and Reiss,1987). This feature was not included in our simple simulation. Mediated transport by SGLT1 seems to play a role in the acceleration of passive absorption, because when the former was partially blocked by phloridzin, the specific SGLT1 inhibitor, in vivo passive absorption was reduced by about a third in laboratory rats(Fasulo et al., 2001).

Our results underscore how, in some animals, the glucose absorptive capacity of the small intestine cannot be accurately estimated only on the basis of in vitro measurements of mediated glucose uptake. The quantitative determination of physiological capacity in relation to load has been called a new research frontier(Diamond, 1993), and our research helps establish important limits on scientists' attempts to match the capacity to absorb nutrients determined on the basis of in vitro data with nutrient intake seen in vivo.

From an evolutionary perspective, one can argue that there are both costs and benefits to intestinal permeability to hydrosoluble biochemicals. A possible cost is that a high intestinal permeability that permits passive absorption might be less selective than a carrier-mediated system for nutrient absorption and might permit toxins to be absorbed from plant and animal material in the intestinal lumen (Diamond,1991). But it has also been suggested(Pappenheimer, 1993) that passive absorption may confer a selective advantage because it requires little energy and provides a mechanism whereby rate of absorption is matched to rate of hydrolysis. Opposing costs and benefits could lead to variation among species in intestinal permeability to hydrosoluble biochemicals. This could explain why studies of some species indicate relatively low reliance on passive absorption (Fine et al.,1993; Uhing and Kimura,1995; Schwartz et al.,1995; Lane et al.,1999) and others such as ours indicate relatively high reliance(Pappenheimer, 1990, 1998; Karasov and Cork, 1994; Karasov et al., 1996; Levey and Cipollini, 1996). For the latter type, vulnerability to hydrophilic toxins could be an important ecological driving force, constraining food exploratory behavior, limiting the breadth of the dietary niche, and selecting for compensatory behaviors such as searching for and ingesting specific substances that inhibit hydrophilic toxin absorption (Diamond et al.,1999).

To continue to advance knowledge on the nutritional, pharmacological, and ecological significance of paracellular absorption we need to determine in more species the relative importance of passive absorption with whole-animal studies like the present one and test different probes and different putative modulators of the paracellular pathway.

We thank Dr Peter A. Jumars at the University of Maine for providing help in the modeling exercise, and Arijana Barun, Bruce Darken and Michele Skopec for their help during the experiments. The work was supported by NSF IBN-9723793, IBN-0216709 and the Max McGraw Wildlife Foundation to W. H. Karasov.