Testing the constant-volume hypothesis by magnetic resonance imaging of Mytilus galloprovincialis heart

The physiology of the hearts of bivalve molluscs has been studied for many years (Bayne, 1976; Beninger and Le Pennec, 2006). Filling of the heart was explained by the constant-volume (CV) hypothesis, which was proposed by Ramsay (Ramsay, 1952), and Krijgsman and Divaris (Krijgsman and Divaris, 1955). The CV hypothesis postulated that, (1) the total volume of the heart is always the same, (2) ventricle contraction causes a decrease in pressure in the pericardium, (3) the auricles are dilated by the low pericardial pressure, and (4) the low pressure of the auricles increases venous return from the anterior oblique veins (Krijgsman and Divaris, 1955; Beninger and Le Pennec, 2006). To test this hypothesis, it is necessary to measure (a) the changes in the volume of the ventricle, the auricles and the pericardium during the cardiac cycle (1st and 3rd postulates), (b) the changes in the pressure in the ventricle, the auricles and the pericardium (2nd postulate), and (c) the flow direction and velocity of the haemolymph in the ventricle, the anterior aorta, the auricles and the anterior oblique veins (4th postulate). As far as we could determine, pressure changes in the ventricle, the auricles and the pericardium were confirmed for bivalve (Anodonta anatina) (Brand, 1972) and gastropod (Patella vulgata) (Jones, 1970) species. However, there have been no reports concerning the volume of the heart and the flow of the haemolymph in the heart and adjacent vessels. Indeed, Krijgsman and Divaris (Krijgsman and Divaris, 1955) have discussed the possibility of a reflux of haemolymph to the pericardium through the renopericardial canals, because a backflow would minimize the pressure decrease caused by ventricular contraction. However, for more than 50 years, no one has published on this subject. Therefore, in a future study, we also plan to measure the flow in the pericardium and the renopericardial canals, in addition to the vessels listed above, and further test the CV hypothesis.

Mytilus galloprovincialis Lamarck 1819 was selected as the experimental animal as a lot of related anatomical and physiological information is available (for example Bayne, 1976), and magnetic resonance imaging (MRI) was used as the main experimental tool. MRI is mainly used in the medical field, but MRI studies on invertebrates are gradually increasing (Bock et al., 2001; Herberholz et al., 2004). However, we could not find any MR images of M. galloprovincialis in the literature. We needed to obtain measurements from anatomical MR images of the heart and the adjacent vessels and organs, and to compare these with results obtained by conventional histological methods (Prudie, 1887; Borradaile and Potts, 1935; Yamamoto and Handa, 2013). Furthermore, MRI is a precise and non-invasive technique for the measurement of heart rate, for estimating the volume of the ventricle, the auricles and the pericardium, and for determining the direction and velocity of the flow of haemolymph in the heart and adjacent vessels in living mussels. In order to improve the reliability of our conclusions, we conducted our tests of the CV hypothesis under two cardiac functional activity conditions, one more and the other less active. Among several possible environmental factors that may influence this, such as temperature, salinity, oxygen tension, etc., we noticed reports on the effects of aerial exposure, which is expected at low tide (Bayne, 1976). When M. edulis was exposed in air, the heart rate reduced and occasionally the heart beat was suppressed completely. This suppression of the heart is not a pathological reaction, but rather a physiological reaction, because the heart rate recovered immediately when the mussel was immersed again after several hours (Helm and Trueman, 1967; Coleman and Trueman, 1971). Therefore, the mussels were

examined under two conditions: (1) when immersed in synthetic seawater (immersed condition) and (2) when exposed in air (emersed condition). The results obtained under these two conditions were examined to test the CV hypothesis.

Images from T₂-weighted rapid acquisition with relaxation enhancement (T_2w)-MRI of living M. galloprovincialis are shown in Fig. 1. The detailed structures of the paraformaldehyde (PFA)-fixed mussels are shown in Fig. 2. MR images of the heart and adjacent vessels and organs were compared with the results obtained with conventional histological methods (Prudie, 1887; Borradaile and Potts, 1935; Yamamoto and Handa, 2013). As the shell consists of calcium bicarbonate, no MR signal was detected from the shell, and so the MR images consist of soft tissues and the seawater in the shell. The heart is positioned on the dorsal side near the hinge of the shells, and is pierced by the rectum, as shown in a mid-longitudinal image (Fig. 1A,C). The dorsal part of the heart is surrounded by the shell, and the ventral sides are supported by posterior retractor muscles (Fig. 1B) and the intestine (Fig. 1C). The heart consists of a single ventricle, with a pair of auricles, and these are surrounded by the pericardium. The ventricle has a single outlet to the anterior aorta and two inlets from auricles (Fig. 2A,B). The anterior aorta leaves from the anterior end of the ventricle. The aortic valve is positioned at the root of the anterior aorta (Fig. 2J–L), which is similar to that in A. anatina (Brand, 1972), and it prevents backflow from the anterior aorta to the ventricle during diastole of the ventricle. The auriculoventricular (AV) valves are positioned one-third of the way from the anterior end of the ventricle (Fig. 2B,E). The AV valves are dual-flap valves, consisting of anterior and posterior flaps (Fig. 2B,E), which prevent backflow from the ventricle to the auricles during contraction of the ventricle.

The auricles are positioned on the lateral sides of the ventricle. Auricles receive haemolymph from the anterior oblique vein, which is connected at the anterior end of the auricles, and supplies haemolymph to the ventricle via the AV valves (Fig. 1D; Fig. 2B,C,E,G). The posterior ends of the auricles are connected to each other by a communication canal underneath the rectum (Fig. 2C,I). The folding of the auricular wall is also depicted, which is assumed to be the filtration membrane of the haemolymph (Andrews and Jennings, 1993). The pericardium is not a closed cavity. The anterior end of the pericardium is connected with the renopericardial canal, which is connected to the kidney. As shown in a 3D reconstructed image (Fig. 1D), the anterior oblique vein and a renopericardial canal run side by side. Therefore, with our procedure, it was possible to obtain images of both of the vessels with a transverse image sliced around the AV valve (Fig. 2G).

As a result, we could obtain MR images showing the ventricle, the auricles and the pericardium, and their inlets and outlets, and also the valves, which are the essential components for the cardiac function of the mussel.

In the immersed condition, M. galloprovincialis opened their shells in synthetic seawater at 23°C. Heart rate varied from 0.74 to 1.35 Hz, and the mean (±s.e.m.) value was 1.07±0.09 Hz (N=8). These values are similar to the heart rate (25 beats min⁻¹ at 21°C) of resting M. galloprovincialis, as observed by impedance pneumography (Braby and Somero, 2006). In the emersed condition, M. galloprovincialis shut their shells, and heart rate remained unchanged (0.88±0.05 Hz, mean ± s.e.m., N=5; P>0.05).

As shown in supplementary material Movie 1, a cardiac cycle was divided into 10 frames. As we did not obtain electrocardiographic data, the frame at end-diastole was used as a reference for the timing of the cardiac cycle. Frames of end-diastole and end-systole are shown in Fig. 3A, and changes in the area of the ventricle are shown in Fig. 3B. The results obtained from five slices between 2 mm anterior and 2 mm posterior to the AV valve were almost identical, not only in timing but also in area. The mean area at end-systole was 33.5±3.3% of that of end-diastole (N=5). As shown in Fig. 2, the anterior end of the ventricle was fixed by the anterior aorta, and the posterior end was fixed by the rectum and the wall of the pericardium. Indeed, in the living mussels (Fig. 1C), the wall of the pericardium is clearly shown, and it is not likely that it would move during contraction. Therefore, the length of the long axis of the ventricle is constant during the ventricular contraction. Hence, the volume of the ventricle could be estimated from the area of the ventricle in a single slice. It was also assumed that the volume of the rectum does not change during the heart cycle. For convenience, a slice 1 mm posterior to the AV valve was used for estimation of the cardiac volume. The area of the ventricle, auricles and pericardium was used to calculate the volume of the ventricle, the auricles and the pericardium, respectively.

Changes in the volume of the ventricle, the auricles and the pericardium are shown in Fig. 4. In the immersed condition (Fig. 4A; supplementary material Movie 1), the end-diastolic volume (EDV) and the end-systolic volume (ESV) were 50.2±2.4% and 21.3±3.0% (means ± s.e.m., N=4) of the heart volume, respectively. The stroke volume (SV=EDV–ESV) was estimated as 28.9±2.9% (N=4). The ejection fraction (EF=SV/EDV) was 57.5±5.1% (N=4). The systolic period was longer than the diastolic period, with a ratio of ~3:2. The motion of the auricles showed antiphase, compared with that of the ventricle. The volume of the auricles (V_A) reached a maximum at end-systole (22.1±2.2%, N=4) and a minimum at end-diastole (11.7±1.9%, N=4). The volume of the pericardium (V_P) was minimized at end-diastole (38.1±3.1%, N=4) and maximized at end-systole (56.6±3.1%, N=4). In the emersed condition (Fig. 4B), cardiac motion changed dramatically. EDV, ESV, SV and EF decreased to 30.5±1.8%, 20.1±1.0%, 11.0±1.1% and 33.4±3.7%, respectively (N=5). The systolic period was longer than the diastolic period, with a ratio of ~3:2. The motion of the auricles was in phase with that of the ventricle. V_A reached a maximum at end-diastole (14.5±1.8%, N=5) and showed a minimum at end-systole (11.0±1.1%, N=5). As a result, V_P was minimized at end-diastole (55.1±1.6%, N=5) and maximized at end-systole (68.9±0.7%, N=5). Under the immersed condition, the SV was statistically higher than that observed under the emersed condition (P<0.05).

The flow of haemolymph in the transverse section at the AV valve is summarized in Figs 5 and 6. In the image from T₁-weighted gradient-echo imaging (T_1w-MRI), the flow was detected as a higher signal intensity (Fig. 5B), and we could identify vessels, the ventricle and auricles by comparison with an anatomical image obtained by T_2w-MRI (Fig. 5A). The direction and velocity of flow of haemolymph were detected by a phase-contrast gradient-echo (PC)-MRI. Fig. 5C shows the haemolymph flow at 7.5 mm s⁻¹. The flow of haemolymph in the ventricle, the auricles, the anterior oblique vein, the renopericardial canals and the branchial vessels was detected. In Fig. 5D, the anterior flow and posterior flow are shown in red and blue, respectively. Haemolymph in the anterior oblique vein flows back to the auricle from the vein, while haemolymph in the renopericardial canals flows to the kidney from the pericardium. Slow flow (around 2 mm s⁻¹) is shown in Fig. 6A–D. The flow of haemolymph in the pericardium was not clear in Fig. 5, as the anterior flow velocity was around 2 mm s⁻¹ in the pericardium (white arrows in Fig. 6B). In more detail, the flow of haemolymph in the dorsal part of the ventricle was directed to the posterior (yellow arrowhead in Fig. 6C,F), while the flow of haemolymph in the ventral part of the ventricle was directed to the anterior (orange arrowhead in Fig. 6B,F). In a horizontal image of the heart (Fig. 6G), haemolymph flow into the auricle and the ventricle via one of the AV valves is shown as an orange and a yellow arrowhead, respectively (Fig. 6H). As shown in supplementary material Movie 2, at the position of the AV valves, inflow occurred from the AV valves towards the dorsal opposite side.

The flow of haemolymph in the auricles moved in the posterior direction, except for one mussel (data not shown). In that mussel, the inflow via the right AV valve was smaller than that on the left side, suggesting an impairment of the right AV valve. The direction of the haemolymph flow of the right auricle was posterior, and that of the left auricle was anterior. As a result, inflow from the left AV valve increased and may have compensated for the impairment of the right AV valve. Therefore, the haemolymph flow through the communication canal of the auricles might function as a safety circuit to maintain venous return to the ventricle.

This is the first report of measurements of SV and EF of the mussel heart. Cardiac output and SV in the invertebrate have been measured by pulsed Doppler flowmeter and techniques based on the Fick equation, such as the thermodilution technique (Stecyk and Farrell, 2002; Jorgensen et al., 1984). These techniques are advantageous because measurements can be obtained for short-term variations of cardiac functions, compared with MRI. However, using MRI, we were able to measure the exact volume of the heart, and it allowed us to estimate not only SV but also the EF. Under the immersed condition, the SV (29% of the heart volume) and EF (58% of EDV) values were much higher than expected, and EF was similar (60%) to that shown in humans (Boron and Boulpaep, 2004). In M. galloprovincialis, the anterior aorta is the only outlet from the ventricle, and the AV valve is positioned one-third of the way from the anterior end of the ventricle (Fig. 2B). The ventricle contracted only on the short axis, as the anterior and posterior ends of the ventricle are anchored at the pericardium wall and the rectum (Fig. 2A). Therefore, it is necessary to pump out haemolymph in the posterior part of the ventricle. Based on the results of the haemolymph flow measurements (Fig. 5 and Fig. 6A–H), the flow of haemolymph in the ventricle was estimated, and is summarized in Fig. 6I,J. The direction of the inflow to the AV valves is not perpendicular to the long axis of the ventricle but, rather, tilted in the posterior dorsal direction. The haemolymph comes from the anterior part of the auricle, which is positioned at the anterior ventral side of the AV valves. Thus, the inflow through the AV valves moves towards the dorsal posterior and the side opposite to that of the ventricle. The streams from the right and left valves may merge together at the posterior end of the ventricle, and are then skewed towards the anterior direction by the inner curvature of the ventricle, and flow in the ventral side of the ventricle to the anterior aorta. These vortices in the haemolymph stream may account for the high EF of the heart, even with a single outlet. Similar vortices in the blood stream were also observed in the human heart (Markl et al., 2011).

The CV hypothesis was tested against the four postulates (Krijgsman and Divaris, 1955; Beninger and Le Pennec, 2006). The first postulate is that the total volume of the heart is always the same. We confirmed that the heart volume is always constant, because the wall of the pericardium was imaged clearly even with non-gated T_2w-MRI (Fig. 1). The second postulate has already been confirmed by pressure changes in the ventricle, the auricles and the pericardium observed for a bivalve (A. anatina) (Brand, 1972) and gastropod (P. vulgata) (Jones, 1970) species. Changes in the volume of the pericardium (Fig. 4) also support a decrease in the pressure in the pericardium during systole of the ventricle. The third postulate is that the auricles are dilated by the low pericardial pressure. In the immersed condition, the volume in the auricles was maximized at end-systole and decreased by 10% at end-diastole. As the flow of haemolymph in the auricles was in the posterior direction (Fig. 6C), the haemolymph in the posterior part of the auricles could not return to the AV valve. At least 10% of the haemolymph should be filtered out to the pericardium, which was supported by the fact that we demonstrated anterior flow of the haemolymph in the pericardium around the auricles (Fig. 6B). As shown in Fig. 2, the wall of the auricles has innumerable involutions and folding. Andrews and Jennings (Andrews and Jennings, 1993) found podocytes in the auricular wall of M. edulis. Podocytes are now accepted as diagnostic of a site of filtration and the formation of urine. Our results also support the concept that the auricular wall is the filtration site. The filtration of the haemolymph is assumed to be driven by pressure caused by contraction of the ventricle. In order to maintain a constant heart volume, filtration of auricles to the pericardium and the output to the kidney from the pericardium via the renopericardial canals should be the same. Otherwise, the heart volume could not be constant. In the emersed condition, however, the volume in the auricles was minimized at end-systole, even though the volume of the pericardium was maximized at end-systole. Therefore, the third postulate was apparently not applicable to the mussel under the emersed condition.

In order to deal with the remaining questions shown above, and also to test the last postulate, that the low pressure of the auricles increases venous return from the anterior oblique veins, we constructed a model as follows (Fig. 7).

1. The ventricle has a single outlet to the anterior aorta and a pair of inlets from the auricles via the AV valves. Changes in the volume of the ventricle (ΔV_V) during a period (t) is the difference in the inflow from the auricles (I_AV) and the outflow to the aorta (I_a):

   

   (1)

   where I_AV and I_a are zero during systole and diastole, respectively.
2. The auricles receive venous return from the anterior oblique veins (I_V). The auricles transfer haemolymph to the ventricle (I_AV), and filtrate haemolymph to the pericardium (I_AP). Changes in the volume of the auricles (ΔV_A) can be expressed as follows:

   

   (2)

   where I_AV is zero during systole, and I_AP was assumed to be zero during diastole.

3. The pericardium receives filtrate from the auricles (I_AP), and transfers haemolymph to the kidney (I_PK). Changes in the volume of the pericardium (ΔV_P) can be expressed as follows:

   

   (3)

   where I_PK and I_AP were assumed to be zero during systole and the diastole, respectively. Values for ΔV_A, ΔV_V and ΔV_P in the immersed and emersed conditions were used, and the ratio of the systolic period and the diastolic period was set to 3:2. The results of the calculations are summarized in Table 1.

With regard to venous return, in the immersed condition, venous return to the auricles (I_V) was almost constant during the cardiac cycle. Therefore, the low pressure of the auricles during the systolic period could maintain venous return to the auricles. However, in the emersed condition, the cardiac output was reduced to 1/3 that of the immersed condition, and systolic I_V decreased to 1/2 diastolic I_V, because ventricular contraction could not produce enough pressure difference to maintain venous return during the systolic period. In summary, in the emersed condition, because of the increase in the auricular volume during systole, which causes an increase in auricular pressure, the venous return is reduced. In the immersed condition, it is likely that the contraction of ventricle and the associated decrease of pressure in the pericardium maintains the venous return during the systolic period. Therefore, there is no direct relationship between the increase in volume of the auricles and the venous return, which was proposed by the last postulate of the CV hypothesis.

This is the first study to provide an estimation of the filtration rate of haemolymph from the auricles to the pericardium (I_AP). In the immersed condition, I_AP was 2/3 of the cardiac output (I_a) during systole. Under the emersed condition, I_AP was 80% of that observed in the immersed condition, even though I_a decreased to 1/3 that of the immersed condition. This is probably due to the increase in the auricular volume during systole, which causes an increase in auricular pressure, so the filtration rate could be maintained even though there was a smaller decrease of the pressure in the pericardium during systole. In order to maintain the constant volume of the heart, the same amount of haemolymph should be output via the renopericardial canals. Circulation in the renopericardial canals and kidney has attracted little attention (Borradaile and Potts, 1935; Beninger and Le Pennec, 2006). However, the flow of haemolymph in the renopericardial canals was 2/3 of the aortic output in the immersed condition, and was maintained at the same level under the emersed condition. The kidney and adjacent tissues may prefer a constant supply of haemolymph, as the human kidney does (Boron and Boulpaep, 2004). Therefore, if the ventricle is considered as a high pressure chamber, like the left ventricle in humans, the pericardium may function as a low pressure chamber, like the right ventricle in humans, to maintain homeostasis of the hemolymph. In this simulation, we assumed that output occurred during the diastolic period, as dilatation of the ventricle increases the pressure in the pericardium (P_P), and the increase in pressure causes output to the kidney via the renopericardial canals. We also assumed that back flow in the renopericardial canals during diastole was negligible, because we only detected forward flow to the kidney (Fig. 5D). However, we could not find any valve structure along the renopericardial canals. In order to confirm this speculation, we plan to investigate circulation in the kidney and related tissues in future studies.

In summary, we applied the MRI technique to test the CV hypothesis, and concluded that minor modifications to the CV hypothesis are required. In addition to two postulates of the CV hypothesis: (1) the total volume of the heart is always the same, and (2) ventricle contraction causes a decrease in the pressure in the pericardium, we modified two postulates: (3) the low pericardial pressure maintains venous return from the anterior oblique vein to the auricle, and (4) the pressure difference between the auricle and the pericardium drives filtration of haemolymph through the wall of the auricles. We also added a new postulate: (5) the dilatation of the ventricle is associated with the output of the haemolymph to the kidney via the renopericardial canals.

In addition, the MRI technique promises a massive step forward in our approach to related measurements and our understanding of cardiac functions in the mussel.

Mytilus galloprovincialis were collected on the shore of Kujyukurihama, Chiba, Japan, on 7 July 2012. The mussels were immersed in natural seawater, cooled by ice, and then transported to the laboratory by car within a time period of 3 h. Mussels were also supplied by Hamasui Co. Ltd (Hatsukaichi, Hiroshima, Japan). These mussels were collected from a subtidal area and cultivated using a floating suspended culture off the shore of Miyajima, Hiroshima on 30 October 2012 and 31 March 2013. After collection, the mussels were immersed in natural seawater, cooled by ice, and then transported to the laboratory by a refrigerated transport service, maintained at 10°C, within a time period of 16 h (Cool Ta-Q-BIN, Yamato Transport Co. Ltd, Tokyo, Japan). At the laboratory, the mussels were housed in aerated natural seawater (50 l) in a 90 l bath at room temperature (20–25°C). For convenience, a small number of mussels (around 10) were kept in aerated synthetic seawater (4 l) in a 5 l bath. The natural seawater was taken from the sea near Hachijo-jima (Hachijo, Tokyo, Japan), and the synthetic seawater (salinity 3.6%) was made by dissolving a synthetic seawater mixture supplied by Matsuda Inc. (Daito, Osaka, Japan) with distilled water. Half of the volume of the seawater was exchanged every week. Mussels were fed with Chaetoceros calcitrans (WDB Environmental and Biological Research Institute, Kaifu, Tokushima, Japan), and 2.5×10⁶ cells l⁻¹ seawater in the bath were applied at intervals of 1–2 days. The experiments were conducted from 1 to 10 weeks after sampling. A total of 19 mussels were used in this study. The length, height and width of mussels were 34.1±0.4, 20.3±0.3 and 13.0±0.3 mm (means ± s.e.m.), respectively. In order to obtain high spatial resolution MRI, four mussels were fixed with synthetic seawater containing 4% PFA for 1 week at 5°C. All of the animal experiments conducted in this study were carried out under the rules and regulations of the Guiding Principles for the Care and Use of Animals, as approved by the Council of the Physiological Society of Japan.

The ¹H MR images were obtained with ParaVison operating software (version 5.1), using a 7T microimaging system (AVANCE III, Bruker Biospin, Ettlingen, Baden-Württemberg, Germany) equipped with an active shielded gradient (micro2.5) and a 25 mm ¹H birdcage radiofrequency coil. The mussels were placed in a plastic tube (barrel of a 30 ml syringe with inner diameter of 22.5 mm; Henke-Sass, Wolf GmbH, Tuttlingen, Baden-Württemberg, Germany) (Fig. 8A). Each mussel was positioned in place using a piece of elastic silicone strip (0.5–1.5 mm thickness and 5–10 mm length), which was inserted at the hinge position of the shell. The elasticity of the silicone strip and the byssus filaments at the ventral edge of the mussel allowed the mussels to open or close their shells. When necessary, two pieces of silicone tubing were cut in half (9 mm outer diameter) and placed beside the shells to immobilize the mussels. The mussels were immersed in 12 ml of synthetic seawater without aeration. When necessary, seawater was drawn from the outlet in the bottom of the tube, and then the mussel was emersed. To prevent drying of the mussels, a piece of wet paper was put on the top of the tube. The temperature of the seawater and the ambient air near the mussels was monitored during the MRI experiments using a fluorescence thermometer (AMOS FX-8000-210, Anritsu Meter, Tokyo, Japan), and the temperature was adjusted to 23±0.3°C (peak-to-peak variance) using a temperature control unit (BCU20, Bruker Biospin, Ettlingen, Baden-Württemberg, Germany). In M. californianus, Bayne et al. (Bayne et al., 1976) reported that the heart rate and oxygen tension of the fluid in the mantle cavity stay almost constant for 1–6 h after the start of aerial exposure. Thus, we usually waited for 30–60 min before starting the MRI experiments. Then, during the setting of the slice position, the orientation of mussels and adjustment of the MRI parameters (30 min), the beating of heart and the flow in the vessels were checked at 10 or 20 min intervals. If the heart beat was suppressed completely, we waited up to 2 h for recovery of the heart beat and flow in the vessels. If not, we abandoned the MRI experiment, and replaced the mussel with another one. The series of MRI experiments consisted of (1) detection of the flow of haemolymph in the heart and vessels (2.5 min), (2) measurement of heart rate (5 min), (3) measurement of the motion and volume of the heart (20 min), and/or (4) measurement of the direction and velocity of the flow of haemolymph (10 min). When necessary, sets of measurements from steps 1 to 4 were repeated. The measurements were usually finished within 6 h. In a separate experiment, 3D T_1w-MRI (20–40 min) and T_2w-MRI (136 min) were measured in order to obtain anatomical information.

The orientation of the planes (Fig. 8B) was defined using the schema described by Norton and Jones (Norton and Jones, 1992). The longitudinal plane was defined as being parallel to the plane defined by the free edges of the bivalves and the hinge. The horizontal plane was defined as being perpendicular to the longitudinal plane and parallel to the basal line of the pericardium, because efferent branchial vessels run parallel to the basal line of the pericardium. The transverse plane was defined as being perpendicular to the longitudinal and horizontal planes. As a convenience, the slice position of the transverse image was measured from the position of the AV valve.

The heart rate was measured using the motion ghost of the anterior artery or the branchial vessels (Xiang and Henkelman, 1993). Typical parameters for T_1w-MRI were as follows: 25.4×25.4 mm field of view (FOV) with a voxel size of 100×100 μm, 1 mm slice thickness, 4 ms echo-time (T_E), 256 phase encodings (N_p), one accumulation. A sinc-shaped pulse (duration 2 ms, flip angle 22.5 deg) was used for excitation. Transverse T_1w-MRI of vessels was used to obtain the relaxation delay (T_R), which ranged from 150 to 450 ms. The pulse motion of the haemolymph in the vessel caused periodic motion ghosts in the phase direction of the image. The details of the calculations of the heart rate, and the preliminary experiments conducted using the mouse, are shown in supplementary material Fig. S1.

The heart motion was imaged by retrospectively self-gated fast low angle shot sequences (IntraGate) (Bohning et al., 1990; Bishop et al., 2006), using a transverse or longitudinal slice with a voxel resolution of 100×100 μm and a slice thickness of 1 mm with a combination of T_R/T_E/flip angle=30 ms/6 ms/22.5 deg. Data from 300 images were obtained sequentially for 20 min, and reconstructed into 10 images per cardiac cycle.

The flow of the haemolymph in the vessels was imaged by the in-flow effect of T_1w-MRI (Bock et al., 2001), and the direction and velocity were measured by phase-contrast gradient echo sequences (PC-MRI) (Lotz et al., 2002), using a transverse or horizontal slice with a voxel resolution of 100×100 μm and a slice thickness of 1 mm with a combination of T_R/T_E/flip angle=100 ms/10 ms/45 deg. Eight pairs of velocity encoding gradients were used, with a strength corresponding to a velocity from −22.5 to 30 mm s⁻¹ with a 7.5 mm s⁻¹ step, and a total image acquisition time of 10 min. When necessary, a slower velocity (1.9–5 mm s⁻¹) could be detected using a higher velocity encoding gradient. As a convenience, the flow direction was defined and presented as shown in Fig. 8C.

Anatomical information was obtained by 3D MRI. The imaging parameters used for in vivo 3D T_1w-MRI were FOV=35.84×35.84×17.92 mm with a voxel size of 140×140×140 μm, a combination of T_R/T_E/flip angle=25 ms/2.5 ms/22.5 deg, and a total image acquisition time of 20–40 min. 3D T_2w-MRI was also measured with a voxel size of 140×140×140 μm with a combination of TR/TE/RARE-factor=1000 ms/30 ms/8, and a total image acquisition time of 136 min. For the PFA-fixed mussels, high-resolution 3D T_2w-MRI was measured with a voxel size of 55×55×55 μm with a combination of TR/TE/RARE-factor=1500 ms/37.5 ms/8, and a total image acquisition time of 18 h 46 min.

The authors would like to thank A. Kinjo and Prof. K. Inoue (AORI, UT) for supplying mussels and providing helpful comments. We would also like to thank R. Murai and H. Okawara (NIPS) for their technical assistance, as well as Prof. S. Kojima (AORI, UT) for his helpful suggestions and encouragement to E.S.