Pulsative venous return from the branchial vessels to the heart of the bivalve Mytilus galloprovincialis supports the constant-volume mechanism Free

The heartbeat is considered to be the major vital sign in animals. The physiology and pathology of the hearts of bivalve molluscs have been studied in investigations of a variety of factors, such as salinity (Sara and De Pirro, 2011), temperature (Bayne et al., 1976; Dong et al., 2021), ocean acidification (Zittier et al., 2015) and ocean contamination (Bakhmet et al., 2009). However, in bivalves such as Mytilus and Anodonta, not only are bradycardia and cardiac arrest caused by hypoxia (Curtis et al., 2000; Seo et al., 2014a; Brand, 1976). For example, Bathymodiolus septemdierum can change heartbeat spontaneously from a regular rhythm to bradycardia, arrhythmia or even a long cardiac arrest, and then return back to a regular rhythm (Fig. S1;Seo, 2017). This particular mussel survived more than 6 weeks after the experiment. In humans, it is very difficult to survive a cardiac arrest lasting more than 10 min because venous return decreases owing to a decrease pressure between the right atrium and the mean venous pressure. In bivalves and gastropods, filling of the heart has been explained by the constant-volume (CV) mechanism (Ramsay, 1952; Krijgsman and Divaris, 1955; Jones, 1983). A schematic diagram of the heart and adjacent vessels of Mytilus galloprovincialis is shown in Fig. 1. The CV mechanism consists of four stages. The ventricle and auricle are in the pericardial cavity (PC) covered by the pericardium. The pericardium is a sac that consists of dense connective tissue, and the outer surface is supported by the shells and the posterior retractor muscles and visceral tissues. Therefore, (1) the total volume of the heart, and thus the outer dimension of the pericardium, is constant during heartbeats. At the end of the diastolic phase (Fig. 1A), the volume of the ventricle is at its maximum. (2) When the ventricular muscle starts contraction (systolic phase; Fig. 1B), the intraventricular pressure increases so that the auriculoventricular (AV) valve (#) shuts and the aortic valve (Av) opens. Thus, the volume of the ventricle decreases. As a result, the pressure in the PC decreases. (3) The intraauricular pressure also decreases to the same pressure as that in the PC, because the auricular wall is a leaky filtration membrane to the haemolymph. (4) The low pressure in the auricle causes a venous return from the afferent oblique vein (AOV), so that the auricle is dilated by filling of the haemolymph. When the ventricular muscle relaxes (diastolic phase; Fig. 1C), the intraventricular pressure decreases, so that the aortic valve closes and the AV valve opens. The haemolymph flows into the ventricle from the auricle via the AV valve and refills the ventricle (Krijgsman and Divaris, 1955; Beninger and Le Pennec, 2006). Several studies have been conducted to test the CV mechanism. Pressure changes in the ventricle, the auricles and the PC were reported for the bivalve Anodonta anatina (Brand, 1972). We also confirmed changes in the volume of the heart and the flow of haemolymph in the heart (Seo et al., 2014a). However, the flow of haemolymph from the interstitial space to the AOV remains an open question (Fig. 1B). In order to get an efficient venous return, the low pressure created by the systole of the ventricle must reach the initial part of the afferent BVs. Also, compliance of the walls along the pathway needs to be low, and it is also expected that there is a pulsative flow synchronized with the ventricular contraction. When used in this context, compliance is a term characterizing vascular wall dynamics. Compliance (C) is defined as the change in volume (ΔV) for a given change in pressure (ΔP): C=ΔV/ΔP (Guyton and Hall, 2006). It can be considered that the venous return to the heart is the same as the stroke volume (SV), the decrease in the pressure in the auricle (ΔP_au) is approximately – SV/C_p, where C_p is the compliance of the pericardium. Because the total volume of the heart surrounded by the pericardium is constant, C_p must be a low value. The ΔP_au conducts to the AOV and decrease in the pressure in the AOV (ΔP_aov) is approximately – SV/C_aov, where C_aov is the compliance of the AOV. The ΔP_aov will decrease when the C_aov is increased, and vice versa. Therefore, the C_aov must be low enough to keep ΔP_ao at a negative value from the end to the beginning of the AOV, which is the driving force of the venous return of the haemolymph. As far as we have found in the literature, no one has studied the pulsative flow in the AOV and BVs caused by the CV mechanism.

In the present study, using Mytilus galloprovincialis, we investigated haemolymph flow caused by ventricular contraction. First, we investigated the vascular structure from the auricles to the BVs using high-resolution magnetic resonance imaging (MRI) and Haematoxylin and Eosin (HE) staining. Second, we investigated the direction of the haemolymph flow in the heart, the AOV and the BVs, using phase-contrasted flow MRI (PC-MRI). Third, changes in the flow velocity along the cardiac cycle were detected by T₁-weighted MRI and the retrospectively self-gated fast low angle shot sequences (IntraGate) MRI. Then, we conducted an estimate of how far the negative pressure reached from the auricle. Fourth, and finally, we examined the hypothesis of venous return based on the CV mechanism in comparison to the closed circulatory system found in humans.

The specimens of Mytilus galloprovincialis Lamarck 1819 used in this study were collected from the subtidal zone of the shores of Hisanohama in Fukushima, Ishinomaki in Miyagi and Kujyukurihama, Chiba in Japan. At the laboratory, 10–15 mussels were kept in 5 litre baths for each mussel in aerated synthetic seawater (salinity 36‰) at room temperature (20–24°C) (Seo et al., 2014b). The salinity of the seawater was measured by a refractometer (Master-S28α, Atago, Tokyo, Japan), and half of the seawater was replaced every week. A total of 20 mussels were used for the MRI study. The length of the mussels was 37.8±0.5 mm (mean±s.e.m.), respectively. 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’ set by the Physiological Society of Japan, and approved by the Animal Research Councils at Dokkyo University School of Medicine (#840).

The ¹H magnetic resonance (MR) images were obtained with ParaVison operating software (version 5.1), using a microimaging system (AVANCE III, Bruker Biospin, Ettlingen, Germany) with a 7 T vertical superconducting wide-bore magnet (inner diameter of 89 mm: BZH 71/300/107C, Bruker Biospin, Fällanden, Switzerland), and equipped with an active shielded gradient (micro2.5, Bruker Biospin, Ettlingen, Germany) and a 25 mm ¹H linear radiofrequency coil. The mussels were immersed in synthetic seawater without aeration, and the temperature was kept at 20°C. When it was necessary to examine the cardiac cycle precisely, the temperature was set at 15°C in order to decrease the heart rate. The mussel was fixed by a piece of elastic silicone strip in a plastic tube (inner diameter, 22.5 mm) to prevent large motion artifacts during MRI measurements (Seo et al., 2014a).

The motion of the heart and the flow of the haemolymph were imaged by the inflow effect of T₁-weighted gradient-echo imaging (T_1w-MRI) (Bock et al., 2001; Seo et al., 2014b). The typical transverse imaging parameters used for the in vivo T_1w-MRI were as follows: 24×24 mm field of view (FOV) with a 64×32 data matrix and a slice thickness of 1 mm, 10 ms relaxation delay (T_R), 3.5 ms echo-time (T_E), 22.5 deg flip angle (FA) and 1 accumulation. Each MRI measurement session consisted of 128 images obtained every 0.32 s. Sagittal or oblique images were obtained by 48×24 mm FOV with a 128×64 data matrix and a slice thickness of 1 mm. Image intensity is presented as relative intensity compared with that of the seawater.

The heart motion and flow in the BVs were also imaged by retrospectively self-gated fast low angle shot sequences (IntraGate) MRI (Bohning et al., 1990; Bishop et al., 2006). The typical parameters used for the transverse or longitudinal slices were a voxel resolution of 100×100 µm and a slice thickness of 1 mm with a combination of T_R=30 ms, T_E=4 ms and FA=22.5 deg. Data from 200 images were obtained sequentially for 6.5 min and reconstructed into 10 images per cardiac cycle. Image intensity is presented as relative intensity compared with that of the seawater.

The direction and velocity of the haemolymph in the vessels were measured by phase-contrast gradient echo sequences (PC-MRI) (Lotz et al., 2002), using transverse or horizontal slices with a voxel resolution of 100×100 µm and a slice thickness of 1 mm with a combination of T_R=100 ms, T_E=10 ms, FA=45 deg. Eight pairs of velocity-encoding gradients were used, with a strength corresponding to a velocity from −16.75 to 15 mm s⁻¹ with a 3.75 mm s⁻¹ step, and a total image acquisition time of 10 min.

Anatomical information was obtained by three-dimensional MRI. The imaging parameters used for the in vivo three-dimensional T₁-weighted gradient-echo imaging (3D T_1w-MRI) were FOV=46.08×23.04×23.04 mm with a voxel size of 180×180×180 µm, a combination of T_R=50 ms, T_E=3 ms, FA=45 deg, and a total image acquisition time of 27 min 18 s. Three-dimensional T₂-weighted rapid acquisition with relaxation enhancement imaging (3D T_2w-MRI) was also measured with a voxel size of 140×140×140 µm with a combination of T_R=1000 ms, T_E=30 ms, RARE-factor=8 and a total image acquisition time of 136 min. When necessary, to image the kidney the mussels were incubated in 10 µmol l⁻¹ MnCl₂ containing seawater for 1 h. After the in vivo MRI experiments, 3 mussels were fixed with synthetic seawater containing 4% paraformaldehyde (PFA) for 1 week at 5°C. For the PFA-fixed species, high-resolution 3D T_1w-MRI was measured with a voxel size of 90×90×90 µm with a combination of T_R=50 ms, T_E=3.5 ms, FA=45 deg and a total image acquisition time of 7 h 16 min.

For the histological studies, the paraffin sections of the PFA-fixed mussels were prepared using a slice thickness of 10 µm. The sections were stained with HE. Images were detected by microscopes of BZ-X710 with an image-stitching mode by the BZ-X viewer and BZ-X Analyzer software (Keyence, Osaka, Japan) and BX63 with an image-stitching mode by the cellSens imaging software (Olympus, Tokyo, Japan). Low magnitude images were detected by a stereomicroscope SZX16 (Olympus, Tokyo, Japan) with a WRAYCAM-NOA2000 digital camera operated by the MicroStudio software (WRAYMER, Osaka, Japan).

The anatomical structure of the vasculature from the branchial vessels (BVs) to the auricles via the afferent oblique vein (AOV) were examined by HE staining and T_1w-MRI of the mussel and compared with the reported anatomy of Mytilus (Sabatier, 1877; Purdie, 1887; Borradaile and Potts, 1935; Yamamoto and Handa, 2013; Seo et al., 2014a). As shown in the transverse sections at the position of 1 mm on the anterior side of the arteriovenous valve (Fig. 2A), the four BVs are positioned in the axis and the free-end of the outer and inner demibranches of the gill (Fig. 2Aa–d). The AOV connects the heart and the BVs in the axis of the gill (Fig. 2B–D). The BVs consist of the inner lamella of the outer demibranches (BV-ILOD) and the outer lamella of the inner demibranch (BV-OLID), and were positioned in the axis of the gill and in the ventral side of the kidney (Fig. 2A). As far as we have observed by MRI, there are no openings between the two BVs and the kidney. As shown in a sagittal section along the BV-ILOD (Fig. 2C,D), the BV-ILOD runs in the longitudinal direction, and then branches to the dorsal direction and connects to the AOV (Fig. 2C,Di,iii). The BV-OLID also runs beside the BV-ILOD, and the two BVs merge at a point near the auriculoventricular valve (Fig. 2Diii). Then, there is another branch to the AOV (Fig. 2C,D). The majority of the merged BVs divert from the gill axis, passing underneath the posterior adductor muscle forming the posterior longitudinal vein (PLV). Then, the posterior end of the PLV is connected to the marginal sinus in the mantle. A smaller vessel remains in the axis and the free ends of the gill. However, the diameters of the posterior end of the four BVs are very small, and it was not clear whether or not these were connected with the marginal sinus in the mantle. The anterior end of the four BVs is also connected to the marginal sinus in the mantle. The inner diameter of the outer lamella of the outer demibranches (BV-OLOD) and the inner lamella of the inner demibranch (BV-ILID) is larger than those of the BV-ILOD and the BV-OLID. In the posterior portion of the AOV, the pilica canals (PL) are connected between the pallial vessels and the AOV (Fig. 2Di,iii). The AOV has a large internal lumen and is embedded in the dense visceral interstitial tissues in the initial part of AOV (Fig. 2Dii), and the lateral wall is attached to the shell near the heart (Fig. 2Di). The AOV runs in the anterior dorsal direction in parallel to the renopericardial canal (RPC) (Fig. 2B–D). Then, it turns to a posterior direction, towards the auricle (Fig. 2B). These results are summarized in a schematic diagram of the heart and the left side of the AOV and BVs of the gill (Fig. 3A).

The direction of haemolymph flow at >3.75 mm s⁻¹ was detected by PC-MRI. Typical transverse images obtained from the 63 MRI sessions from five mussels, are shown in Fig. 4. The flow of haemolymph in the ventricle, the auricles, the AOV, the BVs, the aorta and some vessels was detected. The anterior and posterior flow are shown in red and blue, respectively. The flow in the four BVs was towards the posterior direction, even after branching to the AOV. At the anterior end of the BVs, flows in the BV-OLOD and the BV-ILID were larger than those of the BV-ILOD and the BV-OLID (Fig. 4D). The flow in the BV-ILOD and the BV-OLID became larger along the posterior axis of the mussel (Fig. 4A,B). Even after the BV-ILOD and the BV-OLID merged into the posterior longitudinal vein (PLV), the flow continued directly in the posterior direction (Fig. 4C,E). We did not detect any flow in the kidney. The flow of haemolymph in the vasculature is summarized in Fig. 3A.

To detect abrupt changes in the haemolymph flow during the heartbeat, flow images were obtained using T_1w-MRI. The haemolymph flow into the slice plane from outside was detected as a higher signal intensity. Typical changes in the MRI image intensity during a regular heartbeat obtained in the 54 MRI sessions using 16 mussels, are shown in Fig. 5A. The heart rate was 18±1.4 (mean±s.d., 11 heartbeat intervals). The image intensity of the ventricle demonstrated two peaks in a single heartbeat, one larger peak appeared in the systolic phase corresponding to ejection of haemolymph from the ventricle, and one smaller peak in the diastolic phase corresponding to filling of haemolymph in the ventricle (ventricle in Fig. 5A). The auricles also presented two peaks, smaller and larger peaks corresponding to filling of auricles and ejection to the ventricle, respectively. The timing of peaks of the AOV, BV-OLID and BV-ILOD were almost synchronized with the peak of the ventricular ejection. We could not detect flow in the BV-ILID and BV-OLOD. In some cases, we detected flow before and after the cardiac arrest (Fig. 5B,C). The heart always stopped at the end of the diastolic period, so the heart stopped after filling the auricles and the ventricle. Therefore, the first beat after the cardiac arrest started from the ventricular contraction (the systolic phase) (Fig. 5B, Movie 1). Since the intensities in the diastolic phase were the same as those during the cardiac arrest, flow in the AOV and BVs must almost stop during the diastolic phase.

Higher spatial resolution images of haemolymph flow in BVs were detected by the IntraGate MRI. Obtained in the 38 MRI sessions using 8 mussels, typical results of transverse images 2 mm anterior from the AV valve are shown in Fig. 6 and Movie 2. Pulsative flow in the four BVs (not only BV-OLID and BV-ILOD, but also BV-ILID and BV-OLOD) were detected. The peak flow in the pair of the BV-OLID and BV-ILOD and the pair of BV-ILID and BV-OLOD are almost synchronized with the peak of the aorta, i.e. the ventricular contraction.

To detect flow in the AOV from the BV to the ventricle, an oblique slice was set along the AOV (Fig. 7A) in one mussel. This mussel was incubated in 10 µmol l⁻¹ seawater containing Mn²⁺ so that the kidney was shown at a higher intensity (Wakashin et al., 2018). The AOV was detected as a negative image compared with the kidney and surrounded tissues by the 3D T_1w-MRI (Fig. 7B). In the systolic phase, the AOV was clearly depicted by the inflow of haemolymph from the efferent BVs (Fig. 7C). In the diastolic phase, inflow to the auricle and ventricle was detected (Fig. 7D). Obtained from the 57 MRI sessions using 9 mussels, typical changes in T_1w-MRI image intensity of the AOV in an oblique slice are shown in Fig. 8 and Movie 3. Before the ventricular contraction, the haemolymph remained in the AOV, so that the T_1w-MRI signal of the haemolymph in the AOV was attenuated by the RF pulses (red circle in Fig. 5C). In the arrested heart, the image intensity of the first image represents the full magnetization indicating the recovered image intensity corresponding to M_o sin(FA), where M_o is the equilibrium magnetization of ¹H of the haemolymph. In T_1w-MRI, RF pulses were applied every 10 ms, which was much faster than the T₁ of the haemolymph, so the image intensity decreased proportionally to [1–exp(–T_R/T₁)] / [1–cos(FA)·exp(–T_R/T₁)]. Thus, the image intensity of the retained haemolymph in the ventricle decreased to 40% up to the third image. When haemolymph from the outside of the slice entered the slice, the haemolymph was depicted as a higher signal intensity. Image intensity increases proportionally to [1–exp(–T_R·k·v)], where v is the flow velocity and k is a constant that depends on the flip angle and the T₁ of haemolymph (Seo et al., 2014b). Therefore, any increase in the image intensity in the AOV represents the inflow of haemolymph into the AOV due to the ventricular contraction (Fig. 8Bb). As a result, the flow from the AOV to the ventricle was detected (Fig. 8Bc,d). The flow velocity was calculated as around 3.5 mm s⁻¹ from the peak position of the middle part of the AOV (Fig. 8A, mAOV) and the terminal part of the AOV (Fig. 8A, tAOV). The haemolymph flow was evoked from the efferent BVs, and flow was moved into the ventricle via the auricle by a single heartbeat. We could not detect any significant increase in the image intensity in the area of the pilica canals (Fig. 8A).

The circulation system of M. galloprovincialis was investigated intensively in the last quarter of the 19th century and reviewed by Sabatier (1877), Purdie (1887), Field (1922), Borradaile and Potts (1935) and White (1937). These texts and figures have been referenced or re-referenced even in contemporary textbooks (such as Ruppert et al., 2015). Based on the literature reviewed, we summarized the vasculature related to the venous return to the heart in a schema shown in Fig. 3B (the ‘classic schema’). The haemolymph flow in the gill is summarized as follows: (1) haemolymph from the interstitial space in the viscera and the mantle accumulates in the kidney, then, it flows into the afferent BVs, (2) the haemolymph in the afferent BVs divide into the vessels in the gill filament, (3) the haemolymph is accumulated into the efferent BVs, and then flows into the anterior longitudinal vein (ALV). (4) Finally, the haemolymph flows into the auricle via the AOV (Field, 1922; Borradaile and Potts, 1935; White, 1937).

Based on the results of our investigations of the vascular structure and haemolymph flow obtained in this study, we created a new scheme for the vasculature related to the venous return from the gills to the heart (Fig. 3A). The route of the haemolymph from the AOV to the ventricle via the auricle agrees in both summaries. However, the route before the AOV differs: the major points of interest are (1) the haemolymph flow in the BVs; (2) the structure of the longitudinal vein (LV); and (3) the vascular connectivity of the kidney.

The BVs consist of four vessels, starting from the anterior part of the marginal sinus (MS). Haemolymph flows in the posterior direction until the end of the BVs (Fig. 4). From the anatomical connectivity (Fig. 2), BVs of the inner lamella of the outer demibranches (BV-ILOD) and the outer lamella of the inner demibranch (BV-OLID) were assigned as the efferent BVs. In the classic schema, the BVs consist of a single afferent BV in the gill axis and the two efferent BVs at the free ends of the demibranches (Fig. 3B). Compared with the classic schema, the position of the afferent and efferent BVs is opposite, and the flow in the gill filaments is in the reverse direction (Fig. 3A,B). Furthermore, this new assignment, demonstrated for the first time in this study, was confirmed by the haemolymph flow direction to the AOV from the BV-ILOD and BV-OLID, and also by the timing of the peak flow (Figs 6 and 7). Even though we could not detect the flow direction in the gill filaments, pulsative flow in the BV-ILID and BV-OLOD was detected, and those were synchronized with the flow of the efferent BVs (Fig. 6). This pair of BVs were connected by the gill filaments. Therefore, the BV-ILID and BV-OLOD must be the afferent BV vessels. In the classic schema, haemolymph come from the kidney into the afferent BV. Then, part of the haemolymph flows in the gill filaments to the BV-ILID and BV-OLOD, and was expected to flow in the anterior direction (Fig. 3B). This is not likely, because the flow in the BV-ILID and BV-OLOD were shown to be directed to the posterior direction in this study (Fig. 4).

Considering the longitudinal vein, we detected the posterior longitudinal vein (PLV) that resulted from the merged BVs of the BV-OLID and BV-ILOD (Fig. 2). However, the direction of the flow of haemolymph was in the posterior direction, not the anterior direction described in the classic schema (Fig. 3B). We could not detect the anterior longitudinal vein (ALV) shown in the classic model (Fig. 3B). Sabatier (1877) assigned the ALV and the afferent BV as corresponding to the BV-ILOD and BV-OLID (fig. 3 in plate XXVIII), respectively. Purdie (1887) also assigned the ALV corresponding to the BV-ILOD (fig. 37). However, Borradaile and Potts (1935; fig. 392) and Field (1922; fig. 148) assigned the BV-ILOD and BV-OLID as the afferent BV and renal vein, respectively. None of these authors presented any evidence of their assignments of the ALV or the afferent BV. Therefore, the ALV in the classic schema might be one of the efferent BVs (either the BV-OLID or the BV-ILOD).

Concerning the vascular connectivity of the kidney, in the classic schema, the kidney receives haemolymph from the gills, the mantle via the pilicate canals (PL), the viscera and the renopericardial canal (RPC), and supplies the haemolymph to the afferent BV (Fig. 3B). However, as far as we have observed in high-resolution 3D T_1w-MR images and light microscopic images of HE-stained slices, the RPC is only the inlet to the kidney and the excretory pore is the only outlet from the kidney (Wakashin et al., 2018). We noted printing errors in the vascular assignments of Fig. 1E in Wakashin et al. (2018). There are no direct openings to the afferent BVs (Wakashin et al., 2018). It is also true that there was no detectable flow in the kidney (Fig. 4). In the Mn²⁺-incubated mussel, the boundary of the kidney is clear and not blurred (Fig. 7A,B). Even after the transient decrease of Mn²⁺ in the kidney, Mn²⁺ was not released or leaked into the surrounded tissues (fig. 7 in Wakashin et al., 2018). Since injected manganese-ethylenediaminetetraacetic acid (MnEDTA) concentrated in the kidney threefold (fig. 6 in Wakashin et al., 2019), it is likely to reabsorb water through the epithelium of the kidney, but there are no direct openings or canals into the kidney except for the RPC, and there are no direct openings or canals out of the kidney except for the excretory pore (Fig. 1; Wakashin et al., 2018).

In summary, judging by the new results of the flow in the BVs, the schema of the vasculature from the gills to the heart is simpler than that shown in the classic schema. The routes of the haemolymph to the heart are as follows (Fig. 3A): (1) the pair of afferent BVs take haemolymph from the anterior part of the MS, and the haemolymph passing through the gill filaments. (2) Then, the haemolymph passed through the gill filaments joins the efferent BVs. (3) The efferent BVs, joined with a number of pilicate canals (PL), may take haemolymph from the mantle, even though we could not detect any pulsative flow in the PL. (4) The pair of the efferent BVs also take haemolymph from the anterior part of the MS. Then, most of haemolymph returns to the auricle via the AOV. Part of the haemolymph in the efferent BVs is not passed through the gill filaments, and part of the haemolymph in the PLV and the afferent BVs could not return to the heart, but flowed to the posterior part of the MS. Therefore, as Sabatier mentioned, the gill circulation system is imperfect as it allows part of the haemolymph to return to the heart without passing through the gills (Sabatier, 1877; White, 1937), and also a part of the haemolymph passing through the gills will shunt to the MS.

Based on the CV mechanism (Ramsay, 1952; Krijgsman and Divaris, 1955; Jones, 1983; Beninger and Le Pennec, 2006), haemolymph flow in the AOV has to be synchronized with the ventricular contraction (the systolic phase) (Fig. 1B). Indeed, as shown in Fig. 5A, the peak positions of the AOV and the efferent BVs are almost the same as that of the ventricle. Because the flow in the AOV and the efferent BVs are also evoked by the first ventricular contraction after the cardiac arrest (Fig. 5B,C), this pulsative flow must be caused by the ventricular contraction, not by the contraction of the BVs.

We also found that the flow in the afferent BVs is also pulsative flow and synchronized with the ventricular contraction. It would be safe to say that this propagation of the pulsative flow corresponds to that of the pulse wave. The pulse wave propagates much faster than the flow wave that is presented in Fig. 8. The pulse wave propagation is faster in the lower compliance and smaller vessels, and the pulse wave is attenuated in higher compliance vessels (Glasser et al., 1997). Therefore, vessels in the gill filaments should be stiff enough to conduct the pressure wave caused by the ventricular contraction. As a result, the afferent BVs could receive the haemolymph from the marginal sinus.

Based on the results obtained in this study, the direction of the haemolymph flow in the gill filaments was expected to be the reverse direction compared with the classic schema (Fig. 3). In general, when a narrow duct connects perpendicular between a pair of large ducts, the direction of the flow in the narrow duct depends on the pressure difference of the two large ducts. In the case of the afferent BVs and the efferent BVs, the initial pressure is the same as that in the anterior part of the MS. Because the flow of the haemolymph in all of the BVs is directly to the posterior direction (Fig. 4), pressures at the posterior ends of the BVs have to be lower than that of the anterior part of the MS. The pressure at the posterior ends of the afferent BVs might be dependent on the interstitial pressure in the posterior side of the mussel or that of the posterior part of the MS. The efferent BVs are connected to the AOV, and the pressure in the AOV during the ventricular systole is lowered by the negative pressure in the auricle, which is almost the same as the pressure in the pericardial cavity (Fig. 1B). Thus, the mean pressure in the efferent BVs is lower than that of the afferent BVs. Therefore, the haemolymph in the gill filaments should flow from the afferent BVs to the efferent BVs.

One remaining question is the flow in the posterior longitudinal vein (PLV). In the classic schema (Fig. 3B), it was speculated that the flow direction was towards the anterior direction (Field, 1922; Borradaile and Potts, 1935), and the lower pressure in the AOV should drive the haemolymph in the PLV to the anterior direction. However, we observed flow towards the posterior direction to the marginal sinus (Fig. 4). One limitation we discovered in this study is that we have no explanation regarding how the posterior directed flow is driven in the PLV. A future study is required to clarify this subject.

At the end of this study, we focused on the driving force of the venous return and compared the open circulatory system with the closed circulatory system. The cardiovascular systems and the intravascular and intra-sinus pressures are summarized in Fig. 9. In the human cardiovascular system (Fig. 9A) (Guyton and Hall, 2006), the heart system consists of a combination of two tanks and a push pump. Seventy percent of the blood pools in the vein (the large tank), and the blood return to the right atrium (the small tank) via the vena cava (VC). Since there is a large compliance in the capillaries, the arterial pulse wave is attenuated completely. Thus, the venous return flow is almost continuous during the cardiac cycle, so that the blood is filled in the ventricle at the end of the diastole. When the ventricular muscle starts to contract, the AV valve is forced to close by the ventricular pressure. Then, the pressure increases until the opening of the aortic valve, and blood is pushed out to the aorta. Therefore, the ventricle is the push pump. As shown in Fig. 9C, the AV valve is closed when the ventricular pressure becomes lower than the aortic pressure, and this keeps the mean arterial pressure at a high level. The arterial pressure is decreased in the arteriole and reaches a value around 40 mmHg at the start of the capillary. Then, it continues to decrease down to 10 mmHg at the end of the capillary. The venous return is driven by the pressure difference between the mean systemic venous pressure and the right atrial pressure (P_VR∼5 mmHg). The P_VR is kept constant as far as the heart beats at the sinus rhythm. When the heartbeats stop, the atrial pressure increases. After a long cardiac arrest, all of the intravascular pressures become the same (dotted line in Fig. 9C), and P_VR becomes zero so that there is no flow in any vessels. Therefore, even when applying cardiopulmonary resuscitation, even when the heartbeat starts again, the venous return is minimum due to the absence of P_VR so that the cardiac output stays low. It is known that the ventricle could attract the venous return during the dilatation. However, because of the high compliance of the walls of the atrium and veins, it is hard to draw the blood to the heart from the veins (Holtz, 1996). In other words, human ventricular output is controlled by the amount of the venous return (Frank–Starling law) (Guyton and Hall, 2006), while the ventricular output cannot control the venous return directly.

In the cardiovascular system of the mussel (Fig. 9B), the heart system consists of a suction pump and a push pump. After a long cardiac arrest, the pressure in the vessels, sinuses and the heart are the same, so that there is no flow in the vessels. Since the ventricle stopped at the end of diastolic phase (Fig. 5B), the ventricle could eject the normal cardiac output by the first ventricular contraction after the cardiac arrest. The AV valve and the aortic valve prevent the backflow of the haemolymph during and after the ventricular contraction. Thus, the ventricle is a force pump, essentially the same as the human ventricle. The haemolymph in the arteries is distributed in the sinuses in the whole body, and the interstitial pressure is ∼0.5 cm H₂O (Brand, 1972). Since there is no pressure difference between the sinus and the auricle, the mussel has to draw the haemolymph from the sinus. Based on the CV mechanism, when the ventricular muscle contracts, the pressure in the auricle is decreased by 1 cm H₂O (Fig. 9D). Owing to the low compliance of the AOV and BVs, this negative pressure can drive venous return from the sinus to the heart. Therefore, the mussel auricles, including the AOV and the BVs, is the suction pump to get the haemolymph from the sinus via the BVs to fill up the ventricle. In addition, the output of the push pump (the ventricle) and the suction pump (the auricle) are linked by the pressure in the pericardial cavity (the CV mechanism). Therefore, the venous return is controlled by the ventricular output.

In conclusion, the heart of the mussel consists of a suction pump and a force pump. The CV mechanism and the low compliance of the AOV, the BVs and gill filaments allows this system to work as a suction pump to fill up the ventricle. Therefore, even after a long cardiac arrest, the mussel can restart the ventricular output and the venous return instantaneously.

The authors express their sincere thanks to Dr T. Okutani for providing helpful comments. We also express our thanks to Dr Y. Kamei, Ms M. Saida, Ms C. Ichikawa and Ms M. Asao (the Optics and Imaging Facility, Trans-Scale Biology Center, National Institute for Basic Biology) for their technical support related to the light microscopies. We also thank Dr D. Gross, Dr V. Lehman and Dr T. Oerther (Bruker Biospin), as well as Ms Y. Imaizumi-Ohashi and Ms M. Yokoi-Hayakawa (DSUM) for their technical assistance. Some of the experiments were conducted in Dokkyo Medical University School of Medicine, Tochigi, Japan.