The sites of respiratory gas exchange in the planktonic crustacean daphnia magna: an in vivo study employing blood haemoglobin as an internal oxygen probe

Aerobic energy production depends on the continuous exchange of oxygen and carbon dioxide between the cellular combustion sites of an organism and its environment. With increasing body size, the metazoans of the higher phyla have had to evolve dedicated organs with surface-enlarged thin epithelia mediating the transfer of respiratory gases between the ventilatory and the circulatory systems. Such extensively elaborate structures are usually not present in animals smaller than a few millimetres in length (Graham, 1988; Rombough and Ure, 1991; Rombough, 1998), which is not particularly surprising given their larger surface-to-volume ratios (Krogh, 1941). Thus, it is sometimes difficult to ascertain precisely what the respiratory organ is or, if it is lacking, to determine whether the whole body surface or only parts of it are employed for integumentary respiration.

The present study aims to determine the favoured sites of respiratory gas exchange in the water flea Daphnia magna. The literature on crustacean biology (Gerstaecker, 1866–1879; Giesbrecht, 1921; Storch, 1925; Krumbach, 1926–1927; Flößner, 1972; Villee et al., 1979; Gruner, 1993) presents different views about the sites of respiratory gas exchange in the genus Daphnia, including suggestions of (i) gill breathing, (ii) intestinal respiration and (iii) integumentary respiration.

Assigned to the class Branchiopoda, the genus Daphnia possesses vesicle-like epipodites on its thoracic limbs, which have been repeatedly regarded as gills and have sometimes been termed branchial sacs (e.g. Claus, 1876). This assumption is, so far, consistent with the morphological organization of a typical crustacean, in which the gills derive primarily from evaginations of the limb integument (Barnes, 1969; Gruner, 1993). In contrast to the gills of the advanced crustaceans, however, the epipodites of the genus Daphnia are in no way elaborated with respect to surface area and epithelial thickness (Bernecker, 1909; Fryer, 1991) to enhance the rate of transfer of respiratory gases. Although overlain by a cuticle that is very thin (0.2–0.5 μm) relative to that of the rest of the leg (1–3 μm), the epipodites are lined with an epithelium that is considerably thicker (15–20 μm) than ordinary epithelium (3–5 μm; Kikuchi, 1983). The selective stainability of the epipodites by silver salts or vital stains (Fischel, 1908; Gicklhorn, 1925; Gicklhorn and Keller, 1925a), formerly misinterpreted as characteristic of respiratory epithelia (e.g. Gicklhorn and Süllman, 1931), points in D. magna, as it does in other crustaceans (Panikkar, 1941; Croghan, 1958), to an osmoregulatory function, which has more recently been confirmed ultrastructurally (Kikuchi, 1983). The role of the neck or nuchal organ, in D. magna, a morphological feature restricted to the first instar juvenile (Halcrow, 1982), has to be seen in the same functional context (Potts and Durning, 1980; Halcrow, 1982) rather than linked to respiratory gas exchange (Gicklhorn and Keller, 1925b; Dejdar, 1930).

The striking phenomenon of anal water intake prompted Lereboullet (1850) and later Weismann (1877) to assume that intestinal respiration occurred in the daphniids. Anal water uptake, caused by antiperistaltic movements of the hindgut (Hardy and MacDougall, 1895), was later related to turgor restoration (Fox, 1952; Fryer, 1970). It is also thought to improve the efficiency of food utilization in the intestine (Fryer, 1970).

General integumentary respiration seems plausible because of the large surface-to-volume ratio of this millimetre-sized animal and because of its delicate thin-walled integument (Halcrow, 1976; Dahm, 1977). This hypothesis was supported by the finding that the beating rate of the thoracic limbs stays constant in D. magna (Heisey and Porter, 1977; Paul et al., 1997) when the ambient oxygen concentration decreases. If the limb movements serve for ventilation, then the expected response to hypoxia in a water-breather with oxyregulatory capacities would be an enhanced limb beating rate (Randall et al., 1997). In the oxyregulating D. magna, however, systemic responses differ from those of a typical water-breather (Paul et al., 1997). The fact that there is no increase in the limb beating rate need not be regarded as negative proof of ventilatory function. In an attempt to filter out as much food as possible from the ambient medium, when there is little or no food available, planktonic filter feeders such as D. magna exhibit close to maximum limb beating rates (Porter et al., 1982). Elevated food concentrations lower limb beating rate in D. magna and, surprisingly, the expected ‘hyperventilatory’ response can then be evoked by reducing ambient oxygen concentration (R. Pirow and I. Buchen, in preparation), indicating that the limb movements do indeed have a ventilatory function. Respiratory gas exchange should therefore occur within the animal’s filtering chamber, because this region is well irrigated with fresh ambient water during the steady process of filter feeding. Moreover, the oxygen partial pressure was found to be lowered in the medium leaving the filtering chamber (Pirow et al., 1999), which indicates that oxygen is extracted from the feeding current.

It has repeatedly been suggested that the inner wall of the carapace is a major seat of respiratory exchange in the genus Daphnia (Leydig, 1860; Fryer, 1991). Deriving from an integumental fold of the maxillary region (Fryer, 1996), the double-walled carapace consists of two shell valves, which encase the thorax, abdomen and limbs, thus forming the lateral boundaries of the filtering chamber. Taking into consideration the water flow within the filtering chamber (Westheide and Rieger, 1996) and the complex circulatory pattern (Hérouard, 1905; Storch, 1925), it seems very likely that the haemolymph enters into intensive gas exchange with the medium when circulating through the spaces between the inner and outer walls of the carapace shell valves. Utilizing the presence of blood haemoglobin (Hb), a respiratory protein with useful oxygen-sensitive spectral characteristics, a newly developed spectroscopic imaging technique enabled us to test this hypothesis experimentally.

Female water fleas Daphnia magna Straus were cultured under the conditions described previously (Pirow et al., 1999). To induce an increased blood haemoglobin (Hb) concentration, parthenogenetic offspring were raised under conditions of moderate hypoxia (30–40 % air saturation) produced by bubbling nitrogen through the culture medium. According to Kobayashi and Hoshi (1982), such hypoxic conditions result in a sevenfold elevation of blood Hb concentration (basic level 1g l⁻¹ or 0.06 mmol O₂ l⁻¹) in 2.5 mm long adult females. The animals used in our experiments had a body length ranging from 2.8 to 3.0 mm, measured from the anterior part of the head to the posterior edge of the carapace at the base of the apical spine.

The experiments were carried out at 20 °C in a thermostatted perfusion chamber (see Paul et al., 1997) that allowed microscopic observation of single animals. To analyze its spectral characteristics, the animal was immobilized by glueing its apical spine to a 1 cm long synthetic brush-hair (histoacryl adhesive; B. Braun Melsungen AG, Melsungen, Germany; Cowles and Strickler, 1983). The animal was positioned lateral-side down with the opposite side of the brush-hair and the distal part of the ipsilateral second antenna glued onto a coverslip. Owing to the curvature of the carapace shell valve, the carapace came into contact with the coverslip at the level of the base of the middle limbs pairs. The flow of medium around the animal was consequently blocked only at this contact site and was somewhat reduced at points surrounding this area. The coverslip with the tethered animal was placed onto the glass bottom of the perfusion chamber, which was then sealed with a transparent screw-top without touching the animal contralaterally. Experimental animals were perfused from the anterior end with medium of variable oxygen partial pressure (see Pirow et al., 1999). Taking into account the specific systemic adjustments of D. magna in response to changes in (Paul et al., 1997), an acclimation period of at least 10 min preceded the data acquisition at each level, and this was found to be sufficient for the animal to attain a new stable heart rate and Hb oxygen-saturation (R. Pirow, C. Bäumer and R. J. Paul, unpublished data).

For spectral imaging, a series of gray-scale images of the specimen was acquired while changing the wavelength of monochromatic illumination. The apparatus (Fig. 1A) consisted of an inverted microscope (Zeiss Axiovert 100, Carl Zeiss, Oberkochen, Germany) combined with a computer-driven scanning-grating monochromator (T.I.L.L. Photonics, Planegg, Germany; 75 W xenon arc lamp, spectral range 260–680 nm, spectral bandwidth 13 nm, response time <2 ms) as an illumination system. Collimated monochromatic light was guided to the microscope via a quartz fibre-optic light guide (1.5 mm in diameter). The illumination wavelength was set by a computer equipped with a D/A converter (DAS1602, Keithley Metrabyte, Taunton, MA, USA). A 16-bit liquid-nitrogen-cooled slow-scan CCD camera (576×384 pixels; LN/CCD-576E, Princeton Instruments, Trenton, NJ, USA) was mounted on the camera adapter of the microscope for image acquisition. Images were digitized by a CCD controller (ST-138, Princeton Instruments) and were transferred to the computer via a high-speed serial interface (430 kHz maximum pixel rate, Princeton Instruments). The low noise and the large dynamic range of the CCD camera allowed the resolution of minor differences in light absorption.

The imaging software WinView and WinSpec (Princeton Instruments) were used for image acquisition and analysis. The built-in C-like programming language was employed to generate macros, which automated image-operation sequences and synchronized image acquisition and the selection of illumination wavelength.

For image acquisition, we used an exposure time of 20 ms and binning in the range 2×2 to 3×3 pixels. Binning had the advantage of reducing acquisition time and saving storage capacity at the expense of spatial resolution. To compensate for CCD dark charge, a background image, taken with the camera shutter closed, was automatically subtracted from the incoming image data. Inhomogeneous illumination of the microscopic field was automatically corrected by the image-acquisition programme. Prior to the wavelength scan, the experimental chamber with the animal inside was placed under the microscope. While changing the illumination wavelength gradually from 400 to 437 nm in 1 nm steps, a stack of 38 images was taken within 4–11 s.

The image stack could be regarded as a three-dimensional data package of intensity values I(x,y,λ) comprising the intensity (I) for each pixel (x,y) of the image as a function of the wavelength (λ). To retrieve the intensity spectrum I_x,y(λ) for a selected x,y position, the image stack was scanned along the wavelength axis (Fig. 1B). The corresponding absorption spectrum A_x,y(λ) was calculated according to Lambert–Beer’s law by taking log₁₀{[I₀(λ)]/[I_x,y(λ)]}, where I₀(λ) is the reference spectrum (Fig. 1B). The value for I₀(λ) was retrieved from a region outside the image of the animal and represented light that had not interacted with the specimen.

Although D. magna is highly transparent, the presence of light-absorbing compounds other than Hb must be taken into consideration. The main light absorber in biological fluids and tissues in the violet and blue parts of the spectrum is the porphyrin system, which is a constitutional part of Hb and cytochromes. Further relevant chromophores in D. magna are algal chlorophylls (light absorption in the range 400–500 nm; Libbert, 1987) in the gut lumen and ingested carotenoids (absorption maxima in the range 450–500 nm; Herring 1968), which can accumulate in the gut wall, the fat cells and the ovaries (Green, 1957). Although carotenoids are transported in the circulatory system, the haemolymph colour in Hb-rich D. magna results from Hb (Herring, 1968). When spectroscopically analyzing haemolymph spaces not obstructed by the organs mentioned above, haemolymph Hb is identifiable if present at sufficiently high concentration (2.5 g l⁻¹; Kobayashi and Takahashi, 1994). This has been successfully achieved in several studies (Fox, 1948; Green, 1956; Hoshi and Yahagi, 1975; Kobayashi and Takahashi, 1994).

Advanced theories describe the spectral behaviour of chromophores in turbid tissues (Cheong et al., 1990; Seiyama et al., 1994). However, the transparency of D. magna allowed us to choose a less complex approach. Identification of Hb was based on a spectral comparison of in vivo spectra with reference spectra of oxygenated (oxy-Hb) and deoxygenated (deoxy-Hb) Hb (see below). The regression equation (Fig. 2B; Hb fit) was derived from Beer’s law, which describes a two-component system assuming oxy-Hb and deoxy-Hb to be the only absorbing substances. Incident light can be scattered in D. magna by floating haemolymph cells, muscle bundles or supporting structures such as the carapace or endoskeletal sheets. This effect was taken into account by the scattering factor b (Fig. 2B; Hb fit), which was assumed to be wavelength-independent within the narrow wavelength range selected. A second regression equation (Fig. 2B; Gauss fit) was used to determine the peak wavelength. Regression parameters and correlation coefficients were checked for plausibility (Table 1) to distinguish Hb-characteristic spectra from non-Hb spectra.

After analyzing the absorption spectra at all x,y positions, the results were depicted as an image (see Fig. 3) in which pixel intensity encoded Hb oxygen-saturation (pseudo-colour presentation). Blue represented deoxygenated Hb and red represented fully oxygenated Hb. For those x,y coordinates where no Hb spectra were detectable, the pixel intensity was set to black.

A diluted Hb solution was prepared from haemolymph samples taken from 20 Hb-rich female adults. After amputating the distal half of the second antenna, the oozing haemolymph was aspirated into a pulled-glass capillary tube (0.58 mm i.d.), which was then emptied into 380 μl of ice-cold phosphate buffer (20 mmol l⁻¹, pH 7.0) containing 2.63 mmol l⁻¹ ascorbate. After centrifugation at 10 000 g (10 min, 4 °C), the supernatant was transferred to a flow-through cuvette (138-OS, Hellma, Müllheim/Baden, Germany; 5 mm path length), and absorbance spectra were acquired under normoxic and anoxic conditions at 20 °C using the apparatus described above (Fig. 1A). Oxygen was exhausted enzymatically (Lo et al., 1996) by injecting 10 μl of ascorbate oxidase (10 units; Sigma Chemical Co.) into the cuvette. The absorbance maxima at 414 nm and 427 nm were identified as those of the Soret bands of oxy-Hb and deoxy-Hb, which corresponded to the values of Sugano and Hoshi (1971) for oxy-Hb (414 nm) but not for deoxy-Hb (423 nm). However, comparison with the absorbance spectrum of purified, deoxygenated D. magna Hb (B. Zeis, personal communication) revealed that the peak wavelength of 427 nm was correct. Both the oxy-Hb and deoxy-Hb spectra were employed as templates with which in vivo Hb spectra were compared (Fig. 2B).

Data are expressed as mean values ± standard deviation (S.D.), with N indicating the number of animals examined. Absorption spectra were fitted by linear regression analysis (Hb fit, see Fig. 2B for regression equation) with stepwise variation of the oxygen-saturation coefficient n from 0 to 1.0. To determine peak wavelength, absorption spectra were fitted with a four-parameter Gaussian equation (Gauss fit, Fig. 2B) utilizing the Levenberg–Marquardt algorithm (Press et al., 1992). All regression equations were programmed in the C-like macro language of WinSpec software, which allowed fast off-line analysis of spectral images.

Haemoglobin oxygen-saturation images were analysed further to obtain in vivo oxygen-binding curves for different regions of the circulatory system. Regions of interest were selected, each comprising 200–400 valid Hb oxygen-saturation values. For each region of interest, a frequency distribution was calculated (class width 2 % Hb oxygen-saturation), and the median value was determined. Statistical differences in Hb oxygen-saturation between different haemolymph regions were assessed using a paired one-sided t-test (P<0.05).

Images of Hb oxygen-saturation were taken in lateral views of D. magna at different ambient levels (Fig. 3). Determination of was possible in those haemolymph spaces that were not obstructed by the large antennae, the gut, the eggs or the beating limbs. The rostral head region and the posterior parts of the carapace shell valves showed generally higher values than the other haemolymph spaces under normoxic and hypoxic conditions. Because of posture variations in the abdominal region, could not always be determined in the ventral posterior carapace region (Fig. 3A,B).

Five positions in the circulatory system of D. magna were selected for a quantitative comparison. The corresponding in vivo oxygen-binding curves showing as a function of ambient were constructed to make the regional differences in Hb oxygen-saturation more apparent (Fig. 4). At normoxia (155.7±2.3 mmHg, 20.75±0.31 kPa, N=5), the highest values were found in the rostral region (position 1, 90.2±4.8 %) and in the carapace lacuna near the posterior part of the shell valves at the base of the apical spine (position 3, 80.4±4.4 %). Position 5 yielded the lowest (24.1±19.6 %), which represented, in principle, the mean value of the Hb oxygen-saturations of two types of overlying haemolymph space at that position: the dorsal lacuna and the carapace lacuna (see Fig. 5). However, owing to differences in lateral extension and therefore in optical path length, the dorsal lacuna may have made a greater contribution to this mixed value. The oxygenation of the haemolymph in the unobstructed carapace lacuna at the level of the brood chamber positioned to the right and dorsally from position 5 was generally higher than that measured at position 5. It is therefore reasonable to assume that the of the dorsal lacuna, which guides the haemolymph coming from the post-abdomen (see Fig. 5), was actually lower than that measured at position 5.

When entering the extended carapace lacuna, the haemolymph spreads out radially along curved paths that become confluent at the median dorsal ridge, where blood flow is directed anteriorly to the heart. Located at the dorsal carapace ridge half-way between position 3 and the heart, position 4 showed a lower (62.8±3.6 %) than position 3. The haemolymph currents in the dorsal lacuna and the carapace lacuna become mixed in the pericardium, from where the haemolymph is aspirated and expelled into the dorsal head region, where an even lower of 57.3±17.6 % was determined (position 2). This hierarchy of levels found among the five selected haemolymph regions persisted at lower ambient oxygen concentrations. A statistical comparison (paired one-sided t-test) at two levels (155.7 and 14.1 mmHg, 20.75 and 1.88 kPa) showed that the in the rostral head region (position 1) was significantly higher than at all other positions (t>3.57, d.f.=4, P<0.05, N=5). Moreover, the in the carapace lacuna near the base of the apical spine (position 3) was significantly higher than at positions 4 and 5 (t>4.5, d.f.=4, P<0.05, N=5).

The in vivo oxygen-binding curves were further used to estimate the ambient oxygen partial pressures (P₅₀) at which the Hb was half-oxygenated using linear interpolation. For positions 1 and 3, which we regarded as regions containing oxygenated haemolymph (see Discussion), P₅₀ was 6 mmHg (0.8 kPa) at position 1 and 15 mmHg (2.0 kPa) at position 3.

The oxygen-saturation of Hb was measured in the circulatory system of the planktonic crustacean Daphnia magna with two-dimensional spatial resolution. In contrast to previous in vivo studies (Fox, 1945; Kobayashi and Tanaka, 1991), this imaging technique provided a view of the level of haemolymph oxygenation of various body regions, which can be used to localize the uptake of oxygen from the environment and the release of oxygen to the tissues. This technique holds further potential for the visualization of rapidly changing oxygen distributions in small transparent animals with a time resolution within the hundred millisecond range (R. Pirow, F. Wollinger, U. Baumeister and R. J. Paul, in preparation).

The level of Hb oxygen-saturation in a section of the circulatory compartment is, in principle, determined by (i) the immediate diffusive loss of oxygen to the tissues, (ii) the immediate diffusive influx of oxygen from the ambient medium, and (iii) the oxygenation level and the flow rate of the haemolymph entering that section. Because the different sections of the open circulatory system are directionally linked by the blood flow pattern, the corresponding Hb oxygen-saturation values can be related to each other to characterize the diffusive exchange processes for an individual section.

Although the trunk, the limbs and the ventral half of the carapace in D. magna were only partly accessible using our technique, the data obtained from the remaining body regions can provide indications of potential sites of oxygen uptake when the haemolymph and flow patterns in the medium are taken into consideration. The haemolymph of the first four limb pairs passes into the carapace lacuna (Hérouard, 1905; Fig. 5), where it radiates along curved paths. In the ventral half of the carapace valves, the haemolymph moves in a posterior direction while coming into close contact with the two branches of flow of medium inside the filtering chamber (M. Gophen, personal communication; Kohlhage, 1994: cited in Westheide and Rieger, 1996). The flow of medium and haemolymph are partly concurrent and partly in cross-current orientation to each other. Following the curvature of the ventral-posterior carapace margin, the haemolymph reaches the median dorsal ridge of the carapace at the base of the apical spine (position 3, Fig. 4), where we found higher Hb oxygen-saturation values than at positions located downstream of that region (see Figs 4, 5). The 50 % oxygenation of Hb occurred at an ambient oxygen partial pressure (P₅₀) of 15 mmHg (2.0 kPa). Kobayashi and Tanaka (1991) and Kobayashi and Takahashi (1994) measured Hb oxygen-saturation in the dorsal carapace region of hypoxia-acclimated 2.5–2.8 mm long D. magna and reported a P₅₀ of 15–17 mmHg (2.0–2.3 kPa), which is consistent with our data.

The high oxygenation level in the posterior part of the carapace suggests that influx of oxygen into the haemolymph occurred during flow through the ventral half of the carapace. This conclusion is supported by the recent finding that D. magna extracts ambient oxygen from the feeding current (Pirow et al., 1999): hypoxia-acclimated females showed a of 3.2 mmHg (0.43 kPa) in the exhalant part of the feeding current under hypoxic conditions (16.2 mmHg or 2.16 kPa). Similar hypoxic conditions (15 mmHg or 2.0 kPa) were found to effect 50 % oxygenation of Hb in the posterior part of the carapace (this study). As the haemolymph in that body region must, in principle, be lower than the of the exhalant part of the feeding current, thus ensuring the diffusive transfer of oxygen from the medium to the blood, 50 % oxygenation of Hb should occur at a haemolymph below 3.2 mmHg (0.43 kPa). This inference is reasonable when referring to the data of Kobayashi et al. (1988), who analyzed the in vitro oxygen-binding characteristics of purified D. magna Hb (0.1 mol l⁻¹ phosphate buffer, 20 °C) and reported 50 % oxygenation of Hb at 1.1 mmHg (0.15 kPa) for hypoxia-acclimated (Hb-rich) animals.

Taking into account Fick’s first law of diffusion, it becomes clear that the conditions for gas exchange across the inner carapace wall are indeed favourable. The arthropod integument consists of an outer cuticular and an inner epithelial layer (Gruner, 1993), whose diffusional resistances to oxygen differ by a factor of approximately 10 (Krogh, 1919). The cuticular layer therefore represents a crucial diffusional barrier. D. magna shows no marked cuticular thickening (Halcrow, 1976; Dahm, 1977), and the whole integumental covering has been found to be permeable to oxygen (R. Pirow, F. Wollinger, U. Baumeister and R. J. Paul, in preparation). The inner wall of the double-walled carapace is covered by a delicate cuticle (<1.0 μm), which is several times thinner than the cuticle on other parts of the body (Dahm, 1977) thus facilitating gas transfer across the inner carapace wall.

In addition to the solid integumental barrier, fluid boundary layers represent another obstacle for diffusive gas transport. Millimetre-sized animals live in an environment in which their fluid dynamics can be in the range of low Reynolds numbers where viscous forces are dominant (Koehl and Strickler, 1981; Gerritsen et al., 1988). As a consequence, these organisms are barely able to shed the surrounding viscous water layers. The thickness of the boundary layer depends on the velocity of the medium relative to the body surface. In Daphnia, reduced boundary layers should occur inside the filtering chamber, where the medium drains through in a jerky manner accelerated to velocities of 10–15 mm s⁻¹ (2–3 mm D. magna/pulex; Gerritsen et al., 1988). However, similar velocities can occur during swimming, which directly affect body surfaces other than those inside the filtering chamber. The range of variation is large, from mean swimming speeds of 9–15 mm s⁻¹ (2.8 mm for D. magna, 20 °C: Kobayashi and Gonoi, 1985; Fryer, 1991) to sinking speeds of 3 mm s⁻¹ (1.5 mm D. pulex at 25 °C; Gorski and Dodson, 1996), to more stagnant conditions when the animal rests on the bottom with the carapace contacting the substratum. The two latter situations, as well as conditions in which the animal maintains its position in the water column using strokes of the large antennae, are comparable with our experimental situation.

The maintenance of a large difference in is decisive for rapid diffusion of oxygen from the medium to the blood. In the filtering chamber, this requirement is fulfilled by the rapid renewal rate of the medium (27–87 μl min⁻¹; Pirow et al., 1999), which is several times higher than the perfusion rate (3–4 μl min⁻¹; Paul et al., 1997). The high ventilation-to-perfusion ratio favours the oxygenation of the haemolymph. In addition, the presence of Hb enhances oxygen transfer from medium to haemolymph. As long as the Hb is not fully loaded with oxygen, there is only a slight change in blood during oxygenation, so that the driving force for diffusion can be largely maintained.

The presence of the highest oxygenation levels in the rostral region, with a P₅₀ of 6 mmHg (0.8 kPa), was surprising and implies a thin-walled rostral integument with a low diffusive resistance. More important in this context, however, is the shape of the rostrum, which is flattened laterally. The diffusion distance from the integument to the centre of the haemolymph space is consequently very short, thus permitting more rapid penetration of oxygen by diffusion. External diffusive boundary layers adjacent to the rostral integument should be reduced in free-moving animals, which are propelled forward in a jerky manner by powerful strokes of the second antennae. The conditions in our experiment were comparable insofar as tethered animals were perfused from the front. Whether there is a substantial contribution to total oxygen uptake is questionable. Nevertheless, this bypass would provide additional oxygen for sensory and central nervous structures located in the rostral head region (Claus, 1876), which could be of advantage during severe hypoxia when the convective transport system fails to supply enough oxygen to that location.

The water flea Daphnia magna shows a variety of adaptations that have arisen from a filter-feeding life in a variable aquatic environment. In addition to what is known about the adjustments occurring at the level of the respiratory pigment (Kobayashi et al., 1988, 1990) and at the metabolic and systemic levels (Paul et al., 1997, 1998), this study discloses additional adaptations relevant to respiratory gas transport. The enlarged shell valves of the carapace are not only a constitutional part of the filter apparatus, forming the lateral boundaries of the filtering chamber, but their special structure allows the thin inner wall to be employed for oxygen uptake while the thickened outer wall provides stability for the carapace. The special shape of the rostrum can make ambient oxygen easily accessible to the central nervous structures, which could be important in sustaining vital control functions during hypoxic conditions.

 
• a

  absorption-related coefficient (haemoglobin concentration, optical path length) (Hb fit)

 
• b

  scattering coefficient (Hb fit)

 
• c, d, g

  Gauss fit coefficients

 
• A

  absorbance

 
• A_x,y(λ)

  absorption spectrum at x,y position

 
• H_o(λ)

  absorption spectra of oxy-haemoglobin and

 
• H_d(λ)

  deoxy-haemoglobin

 
• H_x(λ)

  haemoglobin absorption spectra of unknown oxygen saturation

 
• I

  intensity

 
• I₀(λ)

  reference spectrum

 
• I_x,y(λ)

  intensity spectrum at the x,y position of the image

 
• n

  haemoglobin oxygen-saturation coefficient

 
• P₅₀

  oxygen partial pressure at which haemoglobin is half-oxygenated (mmHg, kPa)

 
• 

  oxygen partial pressure (mmHg, kPa)

 
• $SO2$

  haemoglobin oxygen-saturation (%)

 
• x, y

  Cartesian coordinates

 
• λ

  wavelength (nm)

 
• λ₀

  peak wavelength (nm)

The technical assistance of Ina Buchen is gratefully acknowledged. We thank Martina Fasel for the excellent three-dimensional drawings and G. Sundermann and J. Lange for allowing us to inspect their scanning electron microscope images of water fleas. We are especially grateful to Alan Rietman Knauth for the linguistic and stylistic improvements to the manuscript. Supported by the Deutsche Forschungsgemeinschaft (Pa 308/7-1).