Non-invasive imaging of blood cell concentration and blood distribution in zebrafish Danio rerio incubated in hypoxic conditions in vivo

The zebrafish Danio rerio has become an important model organism for the study of vertebrate biology, being well suited for developmental and genetic studies. Large-scale genetic screens have identified hundreds of mutant phenotypes, many of which may serve as models of human disease(Warren and Fishman, 1998; Barut and Zon, 2000). However,studies focussing on the physiology of the developing zebrafish embryo or larvae are scarce, and our understanding of the basic physiology of D. rerio lags far behind our knowledge of the genetics(Burggren and Keller, 1997). Blood circulation in all vertebrates starts early in development, and the first heart beat is typically observed before the heart is completely differentiated (Pelster and Bemis,1991). Despite this early onset of cardiac activity and blood circulation, the physiological function of blood convection has been questioned (Pelster and Burggren,1996; Pelster,1999). In small larvae such as zebrafish, diffusion of oxygen through the body surface alone appears to be sufficient to meet the metabolic needs of the animal (Territo and Burggren,1998; Territo and Altimiras,1998; Pelster,1999; Gielen and Kranenbarg,2002). This demonstrates that coupling between metabolism and convective oxygen transport is not yet established in the early larval stages,and Rombough (2002) suggests that ion- and osmoregulatory functions may require blood flow much earlier in development than metabolism.

Hypoxic conditions are observed in the flowing and stagnant waters that are the natural environment of the tropical zebrafish. The coupling of convective oxygen transport and metabolic activity ensures sufficient oxygen supply to the cells and prevents oxygen shortages at the organ level. Accordingly, in adult animals hypoxia itself acts a stimulus and induces profound changes in cardiac activity and peripheral resistance, and even stimulates erythropoiesis. If this coupling is not yet established in early developmental stages, it could mean that hypoxia does not act as a stimulus in early developmental stages. In a recent study in zebrafish we were able to demonstrate that long before coupling between metabolic requirements and blood flow is established, environmental hypoxia can be sensed and induces stimulation of cardiac activity (Jacob et al., 2002). A reduction in the oxygen-carrying capacity of the blood, however, had no effect on cardiac activity. Thus, hypoxia does exert a signaling effect, even in early larval stages. The aim of the present study was to investigate whether hypoxia, in addition to the modification of cardiac activity shown in an earlier study, would also induce a redistribution of blood and/or stimulate the production of red blood cells in the zebrafish. In addition, the vascular bed of various organs was compared in animals raised under normoxic and hypoxic conditions in order to test the hypothesis that hypoxia stimulates the formation of blood vessels even at early developmental stages. To answer these questions for millimetre-sized zebrafish, we used the recently developed method of digital motion analysis(Schwerte and Pelster, 2000)and extended it so that we could determine the concentration of red bloods cells in a defined volume of blood and also visualize blood distribution within the animal.

Zebrafish Danio rerio Hamilton 1822 larvae were obtained from our own breeding colony. Because of their greater transparency, poorly pigmented mutants of the zebrafish (Albino, Brass) were used. Parent animals used to start the breeding colonies were either obtained from a local supplier or generously provided by Dr Frohnhöfer (Max-Planck Institute for Developmental Biology, Tübingen, Germany) and Ms Loos (University of Konstanz, Germany). Breeding colonies and larvae were kept in small aquaria at 28°C. Animals were fed after swimbladder filling (4-6 d.p.f.) with micro powder food (Zebrafish Management; www.zmsystems.demon.co.uk)and Artemia (from 10 d.p.f.).

An inverted microscope (Zeiss Axiovert 25 CF) was placed on a solid,heavy-weight steel plate to reduce vibration and the illumination set to infrared light (913 nm wavelength) to prevent any light-induced stress reactions in the animals. The microscope was equipped with a 2/3″ CCD camera (Hamamatsu C-2400 without infrared cut-off filter) which, in turn, was connected to the luminance input of a SVHS video recorder (Sony S-9500),remote-controlled via the RS232 serial communication port. Recorded images were digitized by a monochrome frame-grabber card (Imagenation PX-610)with a personal computer (PIII 450 MHz). The depth of view was adjusted to provide images that visualized most of the erythrocytes in one field.

A cast of the vascular bed was obtained by accumulation of the shifting vectors of moving erythrocytes from a number of subsequent difference pictures, as described in a previous study(Schwerte and Pelster, 2000). Briefly, by subtracting the two fields of a video frame, any movement that occurred within the 20 ms necessary for the acquisition of one field was visualized. The length of the shifting vectors, generated by this subtraction,represent a direct measurement of the velocity of a moving particle, i.e. an erythrocyte in the vascular system. By accumulation of shifting vectors generated from several consecutive video frames, a complete trace of the routes moved by the erythrocytes was obtained(Fig. 1). Vascular beds of the entire animal can be visualized non-invasively using this method(Schwerte and Pelster, 2000). Typically the difference pictures of about 30 consecutive images were accumulated to obtain a complete cast of the vascular bed.

As shown in Fig. 1, an image showing a complete cast of the vasculature of a body section or an organ permits measurement of the total area covered by these vessels relative to the size of the total area covered by the organ. This value can be used as an indicator for the vascularization of a tissue. Values for the whole tail and the gut were calculated.

Developing this method further, it is also possible to obtain data about the concentration of erythrocytes in a given section of a blood vessel. The grey-scale value of any given pixel, or of a defined number of pixels, in the image generated by digital motion analysis increases linearly from 0 to 255,depending on the number of erythrocytes passing it. Although the depth of the grey scale display on the screen is limited to 8 bit, the actual range for the calculations was extended to 24 bit. Thus the erythrocyte distribution could automatically be recorded in defined blood vessels or in the whole animal. Calibration of the signal was done by correlating the grey scale values of a defined area with the number of erythrocytes passing this area, as counted by the conventional frame-to-frame technique(Schwerte and Pelster, 2000). These recordings were made using a 20× objective.

To evaluate blood redirection, mean values for the number of blood cells passing a specific tissue were compared and taken as an indicator of relative blood flow. A comparison of the concentration of red cells in various sections of the vascular system did not reveal the existence of any significant organ-specific differences in the hematocrit in zebrafish larvae, and our studies were focused on the brain, gut and tail musculature. The mean grey levels of the specific tissues obtained by this method were used to determine relative changes in tissue blood flow.

Using a 40× objective, 4-8 image frames had to be patched in order to obtain a picture from the whole animal. After accumulating sufficient difference pictures in order to obtain a cast of the vascular bed, the blood vessel diameter of defined sections of the vessels was measured and the volume of a defined blood vessel section calculated, assuming a circular cross-section of the vessels.

Subsequently, erythrocytes within the defined section of the vessel were tracked on individual images and a series of images showing individual blood cells was stored to the computer hard drive. Blood cells were automatically detected by their characteristic grey scale value and motion, marked with a red cross and counted. The results were controlled by optical inspection of an image series and misinterpretations were eliminated. This determination was repeated five times for each animal to take possible clustering of blood cells into account. The process of counting red blood cells and the visualization of red blood cell distribution is described in Fig. 2.

Albino zebrafish eggs were incubated under hypoxic conditions(P_O2 8.7 kPa) at 28°C. Oxygen tension was adjusted using a gas flow meter, which prepared a gas mixture of air and nitrogen. The gas was infused into the water (28°C) of sealed 16 liter aquaria through a fine-pored tube. Oxygen tension in the water was controlled twice a day with a calibrated Clark oxygen electrode (Radiometer Copenhagen,Willich, Germany) to ensure stable values.

Measurements were made at 3, 5, 7, 12 and 15 d.p.f. For measurement the animals were anaesthetized with a neutralized tricaine solution (100 mg l^-1) at the adjusted P_O2, and embedded in low-melting-point agarose (containing 100 mg l^-1tricaine). For animals older than 11 d.p.f., the gills were sculpted free from the agarose to allow gill ventilation(Rombough, 2002). The animals were covered with a thin layer of water (100-200 μm, containing 100 mg l^-1 tricaine). In the hypoxic groups all media were adjusted to the desired P_O2 and the animal chamber was sealed tight. The sealed animal chamber had a reservoir of 0.7 ml gas with the desired P_O2, enough to maintain the P_O2 during the measurement. This was checked using a miniature oxygen electrode. Oxygen deprivation has been shown to cause general developmental retardation, so the staging was carefully done using the following morphological criteria: yolk sac, animal length, diameter at the position of the heart and vascular bed(Isogai et al., 2001)(Fig. 1). The animals were not fed prior to experiments. To avoid artefacts from differently fed animals(older than 6 d.p.f.), the gut filling was microscopically inspected and found to be similar in all animals prior to data acquisition. The experimental groups were pooled from 2-3 clutches obtained from a group of 15 female with 5 male fish.

For comparison of two means, statistical significance was evaluated by unpaired Student's t-test. For multiple comparisons, one-way analysis of variance (ANOVA) followed by Student—Newman—Keuls multiple comparison test was used. Differences were considered significant at P<0.05.

Hypoxic and normoxic animals were not different in size from 3 to 12 d.p.f. By 15 d.p.f. the body length of animals of the hypoxic group was reduced by 0.2±0.03 mm (N=6) compared to normoxic animals (not shown).

Between 3 and 7 d.p.f. the concentration of red cells in the blood remained quite stable and no differences between normoxic and hypoxic animals were observed (Fig. 3). At 12 and 15 d.p.f., however, the concentration of red blood cells decreased significantly,and at 15 d.p.f. it was reduced to approximately 30% of the value recorded at 7 d.p.f. in control animals. In hypoxic animals the concentration of red blood cells also decreased at 12 and 15 d.p.f., but this decrease was significantly smaller than in normoxic animals.

There were no significant changes in vascularization of the tail musculature and the gut. The basic pattern of the vascular bed was similar in normoxic and hypoxic animals (see Fig. 4). Nevertheless, in hypoxic animals the number of intersegmental anastomosis was not significantly higher in hypoxic animals, and the number of animals showing a caudal vascular tree was higher in hypoxic than in normoxic animals. A statistical analysis of the morphometric data obtained from sections through normoxic and hypoxic incubated animals(Table 1), however, showed no significant differences in the area covered by blood vessels.

Fig. 4 shows typical false-colour-coded images of the vascular cast obtained from 12 and 15 d.p.f. animals raised under either normoxic or hypoxic conditions. It was obvious that in 12 and in 15 d.p.f. animals, blood perfusion was higher in the muscle tissue of hypoxic animals. By contrast, perfusion of the gut was significantly lower only in 12 d.p.f. animals, but not in 15 d.p.f. animals. Brain perfusion was not affected by hypoxia. Fig. 5 summarizes these differences as percentage of control value(100%).

The methods described in the present study are very sensitive to animal movements, so to restrict their movement the animals were mounted in low-melting-point agarose. The gelling temperature of the agarose(26-30°C) is within the physiological temperature range for zebrafish. High viscosity solutions reduce water convection and therefore increase the magnitude of unstirred layers, which may significantly impair gas exchange(Pinder and Feder, 1990; Feder and Booth, 1992). Measurements of oxygen equilibration in a 15.0 mm layer of 1% agarose compared with a layer of stirred water confirmed that the time necessary to equilibrate the agarose was approximately twice as long.

The quality of the automated red blood cell count was ensured by careful visual inspection of detected blood cells. Double labelling as well as underestimation because of blood cell clusters could easily be detected by visual inspection of the image series. Repeated measurements in the same animal were within an error range of 2%.

Visualization of blood vessels for morphometric analysis is based on the movement of red blood cells, and therefore plasma skimming(Schmid-Schönbein, 1988)or layers of erythrocyte-free plasma may cause an underestimation of the vessel diameter. On the basis of model calculations of red cell movements in a capillary system, this plasma layer is estimated to have a thickness of approximately 1 μm or less(Schmid-Schönbein, 1988). To evaluate the accuracy of our method, we compared the blood vessel thickness determined from the accumulated difference images with the diameter measured from a complete video image of the same site assessed by microscopical inspection. The results differed by no more than 2-3%. We therefore conclude that our volumetric analysis based on the movement detection and the red blood cell count provides very accurate and highly reproducible results.

In adult vertebrates, cardiac activity and blood flow to tissues are both mainly determined by metabolic demand. Adult fish and amphibians are typically oxy-regulators, but in the earliest embryonic and larval stages of these vertebrates oxygen uptake appears to decrease with decreasing environmental P_O2, i.e. they are oxyconformers(Hastings and Burggren, 1995). In fact, the first embryonic stages of the zebrafish can survive complete anoxia in a state of suspended animation(Padilla and Roth, 2001), and the Arctic charr can survive several hours of anoxia, with severe metabolic depression and a significantly reduced cardiac activity(Pelster, 1999). Anoxia certainly is an extreme situation for a vertebrate embryo, but mild or chronic hypoxia has also been shown to provoke significant physiological adaptations,ranging from metabolic effects to ventilatory and circulatory adjustments (for a review, see Pelster, 1997),and in zebrafish larvae a stimulation of cardiac activity in response to hypoxia was observed as early as 3 d.p.f. at 28°C(Jacob et al., 2002). Chronic hypoxia during embryonic and larval development may also significantly modify differentiation and growth. In larval amphibians, for example, aquatic hypoxia stimulates growth of respiratory surfaces and enhances the transition from gill to lung respiration (Guimond and Hutchison, 1976; Burggren and Mwalukoma, 1983; Burggren and Just, 1992), but often hypoxia results in retardation of development (Pelster, 1997). In our experiments the first signs of retarded development were observed at 15 d.p.f., and therefore at this age the experiment was terminated in order to prevent a situation where physiological effects induced by hypoxia are mixed up with effects induced by developmental retardation.

In mammals hypoxia usually causes an increase in the oxygen transporting capacity of the blood, i.e. an increase in the number of circulating erythrocytes and in hemoglobin concentration. Similar observations have been reported for embryos of the turtle Pseudemys nelsoni(Kam, 1993; Jacob et al., 2002), whereas in chicken embryos during early development no stimulation of red cell production was observed (Baumann and Meuer,1992). The results of our study clearly show that hypoxia increases red blood cell concentration in zebrafish larvae during the second week after fertilization, but no effect was observed until 7 d.p.f. Up to 7 d.p.f., the concentration of red cells remained fairly constant in normoxic as well as in hypoxic animals. Weinstein et al.(1996) showed that the first presumptive proerythroblast-like cells can be detected in zebrafish by the end of the first day (20 h post-fertilization), and this primitive cohort of cells originates in the intermediate cell mass. By transfusing fluorescently labelled blood cells into 1.5-day-old host embryos, recording the fraction of these labelled cells and comparing the histology of the red cells, these authors concluded that this first cohort of primitive red cells provides the embryo with all, or nearly all, its red blood cells for at least 4 days. A population of new, larger and more adult-appearing erythrocytes became predominant by 10 d.p.f.. The constant concentration of red cells observed until 7 d.p.f. in our experiments is in line with these results, but the severe decrease in red cell count at 12 d.p.f. was unexpected. A possible explanation would be that the growing volume of the vascular bed is accompanied by an increase in plasma volume, but not by an equivalent production of erythrocytes. Accordingly, hematocrit and red cell concentration would reduce. An increased cell volume of the newly produced erythrocytes at constant hematocrit would also cause a decrease in red cell concentration, but this latter explanation does not appear very likely.

In hypoxic animals this decrease in the red cell concentration was significantly less than in normoxic animals. If the decrease in red cell concentration observed between 7 d.p.f. and 12 d.p.f. is caused by an expansion of plasma volume, in hypoxic animals the increase in plasma volume would be smaller than in normoxic animals. Information about the total blood volume, however, is necessary to test this idea.

Our method provides a two-dimensional projection of the complete cast of the vascular system. A comparison of projection areas revealed no significant differences between the experimental and the control groups. Assuming a circular cross-section and similar size of vessels, this result means that there was no significant difference in the total volume of the vascular bed in the two groups. However, the acquisition of images for the visualization of the blood redistribution as compared to the acquisition of images to create a complete vascular cast required a compromise between acquisition speed and spatial resolution. Although the resolution of our images was sufficient to show changes in blood distribution, small changes in the diameter of blood vessels, and therefore changes in total blood volume, may not be detectable. For the exact determination of the total blood volume, further studies using a more refined method with higher spatial resolution are needed.

Another explanation for this observation would be that hypoxia stimulates erythropoiesis in zebrafish larvae, as it does in fetal mammals(Richardson and Bocking, 1998)or in embryos of the turtle Pseudemys nelsoni(Kam, 1993; Jacob et al., 2002). Recent molecular studies revealed that hematopoiesis in zebrafish larvae is stimulated by VEGF, and VEGF is also known to drive angiogenesis(Liang et al., 2001). The differential regulation of these processes appears to include a regulatory loop by which VEGF controls survival of hematopoietic stem cells in mice(Gerber et al., 2002). Erythropoietin (EPO), also well known to be involved in hypoxia-induced erythropoiesis, has been shown to mediate hypoxia-induced VEGF expression in rats (Liu et al., 1995).

A redistribution of blood in response to different metabolic demands is a well-established physiological adaptation in many species. This is the first study to visualize a `source and sink' pattern of blood distribution in zebrafish larvae in response to oxygen deprivation. It is obvious that under hypoxic conditions blood is driven from the gut (source) to the muscles in the tail (sink), but until 7 d.p.f. no changes in blood distribution were observed. In a previous study we observed a hypoxic stimulation of cardiac activity as early as 3 d.p.f., which demonstrates that at about 1 day after hatching oxygen receptors are present that respond to hypoxia, and this information can be translated into a signal generating an increase in heart rate and in cardiac output (Jacob et al.,2002). Thus, the lack of oxygen can be identified by the larvae,but until 7 d.p.f. this information is not used to induce a change in blood distribution and in erythrocyte production. At 12 d.p.f., however, hypoxia induces a significant increase in the perfusion of the muscle tissue, and the so-called red-layer of muscle has been implicated in the uptake of oxygen in early larvae and at about the time of hatching(El-Fiky and Wieser, 1988). Possibly blood is redistributed towards the muscle tissue in order to enhance oxygen uptake through the body surface under hypoxic conditions. Under normoxic conditions, the oxygen requirements of zebrafish larvae apparently are met by bulk diffusion until 12-14 d.p.f.(Jacob et al., 2002), and this is the time when the gills with secondary lamellae are developed(Rombough, 2002). Also at this time the site of oxygen uptake is shifted towards the gills, and convective oxygen transport becomes a necessity for transporting oxygen from the gills to the tissues. This implies that blood flow is not necessary to sustain aerobic metabolism in normoxic larvae until about 12 d.p.f., but under hypoxic conditions zebrafish larvae are obviously able to stimulate cardiac activity and to redirect blood flow in order to increase perfusion of the muscle tissue. Following the hypothesis that the red layer of muscle may be implicated in the uptake of oxygen in early larvae(El-Fiky and Wieser, 1988),this would suggest that oxygen uptake through the red layer may be enhanced,which, given the small cross-sectional area of the larvae, is essentially cutaneous respiration in the area of the red muscle tissue. To test this idea,we are currently developing a method to visualize changes in hemoglobin oxygenation in vivo by recording changes in the absorption spectrum of the hemoglobin.

On the other hand, the redirection of blood may not be related to cutaneous respiration, but simply reflect a maturation of control systems (e.g. hypoxic vasodilation and hypoxic vasoconstriction). The presence ofα-adrenergically controlled precapillary sphincters in the intersegmental muscle tissue of zebrafish larvae at 8 d.p.f. has already been shown (Schwerte and Pelster,2000). Furthermore, a general vasodilation in that tissue may be caused by nitric oxide, as demonstrated by Fritsche et al.(2000) to already occur by 5 d.p.f. Thus, hormonal control mechanisms contributing to a redistribution of blood flow are certainly established at this time of development.

The increase in muscle tissue perfusion could also be a consequence of an increased cardiac output, but this would not explain the decrease in gut perfusion simultaneously recorded in our experiments. In a previous study,Jacob et al. (2002) observed an increase in cardiac output by 20-30% under hypoxic conditions, but in our experiments the effect on cardiac activity was even smaller, so that a change in cardiac activity alone cannot explain the redistribution of blood measured in our experiments.

Another possibility is that the vascular volume in muscle tissue increased during hypoxia. As already mentioned, however, the two-dimensional projections of the complete cast of the vascular system did not reveal an increased vascular volume.

Yue and Tomanek (1999)demonstrated that coronary vessels from cultured 6-day-old quail embryo grow faster under hypoxic conditions, while hyperoxia induced a delayed angiogenesis. By contrast, chorioallantoic membrane capillarization of chicken embryos has been shown to increase during hypoxia(Dusseau and Hutchins, 1988; Hudlicka et al., 1992). In larval amphibians, aquatic hypoxia stimulates growth of respiratory surfaces and enhances the transition from gill respiration to lung respiration(Mwalukoma and Burggren,1983).

In our study, however, no significant changes in the vascular bed were observed. There appeared to be quite a high interindividual variation in the expression of small blood vessels like the intersegmental anastomsis or the caudal vascular tree, and we cannot exclude the possibility that minor changes in the expression of these vessels did occur in hypoxic animals. The overall morphometric analysis, however, did not show any significant differences between control and hypoxic animals. Compared with the development of amphibians such as Xenopus or Rana, or the development of salmonid larvae, zebrafish development is very rapid, and it may be possible that a rearrangement of the vascular bed can only occur in later developmental stages. On the other hand, molecular signals involved in the formation of blood vessels have been identified in early embryonic stages. The molecular mechanisms that lead to the extremely regular pattern in the zebrafish trunk have been shown by Childs et al.(2002), and VEGF up- and down-regulation seems to be involved in this process. Whole-mount in situ hybridization of zebrafish embryos indicated that strong expression of VEGF had already occurred at 18 h post fertilization(Weinstein et al., 1996; Liang et al., 1998; Tan et al., 2001).

Under normoxic conditions, oxygen supply via diffusion seems to be sufficient to meet the metabolic demand up to 12-14 d.p.f. This changes during hypoxia, however, because the reduced oxygen gradient cuts down diffusion of oxygen to the tissues. The data available so far show that in this situation convective transport can be enhanced by increasing blood flow(Jacob et al., 2002), and the oxygen carrying capacity of the blood can be increased. In addition, blood flow can be redirected towards a potential site of oxygen uptake. Thus,although under normoxic conditions convective transport is not necessary, the larvae can use the circulatory system as a backup system to augment oxygen distribution and oxygen supply to the tissues, which is strong evidence that the cardiovascular system can operate as a convective transport system for oxygen much earlier then required under normal circumstances. This situation was called `prosynchronotropy' by Burggren and Fritsche(1995). A prosynchronotropic development of convective oxygen transport appears to be very useful, because it creates a safety belt for a situation where bulk oxygen diffusion alone would not be sufficient to ensure oxygen supply to all tissues, and thus widens the range of environmental conditions in which the larvae can survive. Angiogenesis, however, seems not to play a key role in facilitating oxygen uptake. It could be that in the small zebrafish larvae the diffusion distances between tissue capillaries and the cells are small enough to permit an optimal supply with oxygen, so that additional blood vessels would not yet enhance oxygen transport to the cells.

The present study was supported by the Fonds zur Förderung der wissenschaftlichen Forschung (FWF P14976-BIO and FWF P16272-B06).