Function and control of the fish secondary vascular system, a contrast to mammalian lymphatic systems

The mammalian lymphatic system consists of blind-ending lymphatic vessels, lymph nodes and other tissues, including the spleen, that collect, filter and return fluid that drains from the tissues to the venous system and transport fats and proteins from the gut into the blood system (Alitalo et al., 2005; Wang and Oliver, 2010). In teleost fishes, some circulatory vessels contain clear fluid and have been described as lymphatic vessels (Kampmeier, 1969). Yaniv et al. (Yaniv et al., 2006) and Isogai et al. (Isogai et al., 2009) have also demonstrated that these vessels share some morphological and molecular characteristics with the lymphatic vessels found in other vertebrates, and suggested that these clear vessels in zebrafish could be used as a model for the mammalian lymphatic system. However, several decades ago these same vessels were hypothesized as part of a secondary vascular system (SVS), receiving inflow from the arterial system via arterio-arterial anastomoses and emptying into the venous circulation (Vogel, 1981; Vogel and Claviez, 1981; Vogel, 1985; Steffensen et al., 1986; Steffensen and Lomholt, 1992; Jensen et al., 2009). These vessels are not blind-ending and sometimes include red blood cells (RBCs) (Ishimatsu et al., 1992). Furthermore, there is no evidence that the SVS vessels are draining edematous tissues or are playing a role in lipid transport. If not lymphatic, what then is the role of this secondary vascular system?

The volume of the SVS can range from 10–50% to nearly twice the volume of the primary circulation (Bushnell et al., 1998; Gallaugher and Farrell, 1998; Skov and Steffensen, 2003; Steffensen and Lomholt, 1992) and therefore may serve to buffer osmotic and ionic changes in the primary circulation. Under resting conditions, no red blood cells (RBCs) are present in the SVS. As many of the capillaries of the SVS are superficial, it may be that plasma supplies nutrients to the skin, and gas transfer occurs between water and superficial structures directly across the skin surface (Steffensen and Lomholt, 1992). During exercise the addition of RBCs to the SVS could augment cutaneous O₂ uptake. However, during aquatic hypoxia, there could be cutaneous loss of O₂ across the skin. We hypothesize, therefore, that exercise will be associated with RBC flow into the SVS, and aquatic hypoxia will shut down RBC flow into the SVS.

Glass catfish [Kryptopterus bicirrhis (Valenciennes 1840)] were used to investigate the control mechanisms and functions of the SVS. Glass catfish are transparent, making non-invasive, in vivo observations and experiments possible. Furthermore, substantial baseline observations have been made on the primary vascular system (PVS) and SVS in this species (Steffensen et al., 1986; Steffensen and Lomholt, 1992; Jensen et al., 2009). Our objectives were to: (1) map flow patterns in the SVS using injected fluorescent microspheres and video-microscopy, (2) map SVS flow patterns without injecting microspheres in unanaesthetized fish during rest, exercise and aquatic hypoxia – with and without surface access, (3) use pharmacological methods to further understand how SVS flow is regulated and (4) measure any changes in PVS haematocrit associated with the SVS filling with RBCs.

Fluorescent microspheres were injected into the PVS. The 0.02, 1 and 4 μm microspheres dispersed rapidly throughout both the PVS and SVS with a clear image for the 0.02 and 4 μm microspheres (Fig. 1). The 10 μm fluorescent microspheres gave a strong signal but did not disperse from the dorsal aorta (DA), the original injection site. The RBCs obtained from a resting, unstressed glass catfish have a mean length of 11.5±0.5 μm, width of 7.1±0.12 μm, and a width-to-length ratio of 63.0±0.1% (means ± s.e.m., N=6). When the RBCs increased in volume, the cells became more rounded and less flattened. When exposed to 1 μmol l⁻¹ isoproterenol, RBC width-to-length ratio increased significantly from 63.0 to 69.3±0.7% (P<0.001, paired t-test). We measured several red (10 μm) fluorescent beads and found them spherical with diameters between 10 and 11 μm.

From preliminary observations on unanaesthetized fish we determined that under resting conditions only plasma flows slowly through the SVS. Flow in the tail section is always towards the head away from the tail and caudal heart. When a fish became stressed (by injection, removing scales, temperature change, excessive handling), RBCs rapidly entered the SVS. We categorized flow in the SVS into four stages (Fig. 2). Stage 0, resting, was when no RBCs could be observed anywhere in the SVS (i.e. only plasma). Stage 1 was when a few non-moving RBCs could be observed in the SVS. Stage 2 was when RBCs could be observed entering the SVS and flow, albeit slow, was noticeable. Stage 3 was if flow was rapid and RBCs were densely packed throughout the SVS.

We observed a total of five mortalities throughout the entire project with no association with any particular treatment or time period.

Hypoxia and exercise both increased flow and number of RBCs in the SVS, and this resulted in a marked reduction in haematocrit in the PVS (P<0.001, ANOVA; Fig. 3). Flow of RBCs into the SVS was more pronounced when fish were permitted to use aquatic surface respiration (ASR) throughout the duration of hypoxia exposure (Fig. 4). Immediately following hypoxia exposure with no ASR, RBCs in the SVS were sparse and flow was minimal or absent (Fig. 4). Without surface access, fish exposed to hypoxic water swam to the bottom of the experimental container, even resting on its side in some instances. This behaviour clearly contrasted with that of the actively swimming hypoxia-exposed fish that were permitted surface access. After 1 h of recovery, there was greater influx of RBCs into the SVS in both hypoxic fish groups. However, flow of RBCs into the SVS was more pronounced in exercised fish when compared with hypoxia-exposed fish (Fig. 4). Exercised fish also exhibited a significant increase in heart rate (f_H) that started to come back to resting by 1 h but was still significantly elevated (P<0.001, repeated measures ANOVA; supplementary material Fig. S1). However, there was no change in f_H in fish exposed to hypoxia regardless of surface access (ASR) (P=0.115, repeated measures ANOVA; supplementary material Fig. S1). Neither hypoxia nor exercise was associated with any change in the length-to-weight ratio [Fulton's condition factor, K=100(W/L³), ANOVA; Fig. 3].

Exposure to the β-agonist isoproterenol (Iso) dramatically increased RBC numbers and flow into the SVS. These changes were associated with an immediate increase in f_H but recovery by 1 h (P<0.001, repeated measures ANOVA; supplementary material Fig. S1). This response was less pronounced when compared with the significant increase in f_H with exercise, which persisted even after 1 h recovery (P<0.001, repeated measures ANOVA; supplementary material Fig. S1). The effects of Iso were ameliorated when exposure was combined with the β-antagonist propranolol (Prop) (Fig. 4), which also significantly decreased f_H (P<0.001, repeated measures ANOVA; supplementary material Fig. S1). Recovery was evident by 1 h. Upon exposure to the α-antagonist phentolamine (PHT), however, RBC flow into the SVS was minimal, and there was no effect on f_H (supplementary material Fig. S1).

There have been several reports of arterio-arterial anastomoses between the SVS and PVS in fish (see Steffensen and Lomholt, 1992) using vascular castes and in vivo microscopic observation. During resting conditions, only plasma flows through the SVS, presumably because inflow diameters are too small to allow RBC entry. Fluorescent microspheres, 0.02, 1 and 4 μm were small enough to pass into the SVS within seconds (Fig. 1), but 10 μm microspheres were not, indicating that input vessels from the PVS to the SVS had a diameter of less than 10 μm. Hypoxia, exercise and stress resulted in varying levels of RBC flow into the SVS. We hypothesized that RBC entry into the SVS would be shut down during aquatic hypoxia to prevent oxygen loss to the water across the skin. This was largely the case when fish were exposed to hypoxia but prevented from accessing the surface to perform ASR. However, fish actively swimming at the surface using ASR during hypoxia exposure exhibited some RBCs and flow into the SVS. Therefore we cannot fully accept our hypothesis that the SVS is not in use during hypoxia. We also did not measure pressure and do not know if the opening of the anastomoses during hypoxia and exercise in the glass catfish was due to increased PVS arterial pressure or relaxation of presumed sphincters around the anastomoses. Stresses such as hypoxia and exercise have been reported to increase circulating catecholamines in fish (Primmett et al., 1986; Thomas and Perry, 1992; Caldwell et al., 2006), and this, combined with our results outlining the effects of adrenergic manipulations indicate that catecholamines are involved in the increased RBC entry into the SVS following hypoxia and exercise. Nitric oxide is also reported to mediate SVS dilation under hypoxia (Jensen et al., 2009) but was not investigated here.

Increased RBC flow into the SVS markedly decreased PVS haematocrit following both hypoxia and exercise in glass catfish. We realize the difficulty of blood collection without admixture in such small fish. However, the procedure used was the same for all fish and we see consistent differences between groups. It is unlikely that admixture occurred in only one group or to a different extent between groups, but this remains a possibility. The fish itself did not change volume, as indicated by no change to body condition factor, a proxy used to indicate changes in body volume. Therefore, water influx is unlikely to play a substantial role in the observed haematocrit change. PVS haemoconcentration due to blood volume reduction (Yamamoto, 1988; Pearson and Stevens, 1991), RBC swelling (Caldwell et al., 2006), and splenic release (Gallaugher et al., 1995) occurred in fish during exercise and hypoxia and may have offset the large decrease in PVS haematocrit, i.e. our measurements may underestimate the effect of increased RBC flow into the SVS on the haematocrit of the PVS. However, the magnitude of the haematocrit reduction we observed indicates that the volume of the SVS is at least equal to, if not larger than that of the PVS in glass catfish. We know of no other report of a reduction in haematocrit during hypoxia or exercise in fish, although Steinhausen et al. (Steinhausen et al., 2008) reported that sockeye salmon exhibited no significant change in haematocrit with exercise. In fact hypoxia and exercise have been reported to increase haematocrit in fish (Gallaugher and Farrell, 1998; Wells, 2009). These differences in the magnitude of the changes in haematocrit with exercise and hypoxia may be related to the volume of the SVS. It has been reported that the secondary vascular system can range from 10–50% to nearly twice the volume of the primary circulation (Steffensen and Lomholt, 1992; Bushnell et al., 1998; Gallaugher and Farrell, 1998; Skov and Steffensen, 2003). The magnitude of other effects, such as splenic release of RBCs, increases in RBC volume and blood volume reduction (see above) may also play a role in differences in the magnitude of the haematocrit change with increased RBCs flow into the SVS of different fish species.

Various hypotheses as to the function of the SVS have been proposed. We suggest that retaining RBCs in a small PVS volume in resting fish will reduce work of the heart and RBC production. Recently, Rasmussen and colleagues investigated the role of the SVS in rainbow trout and concluded that, because the SVS can regulate concentration and distribution of cell types, e.g. immune cells, the SVS probably plays an important role in immune function (Rasmussen et al., 2013). Ishimatsu and colleagues (Ishimatsu et al., 1992) suggested that the SVS plays a role in acid–base regulation and epithelial ion transport, probably via Cl⁻/HCO₃⁻ exchangers, which may also be superficially located. There are large fluid shifts with as much as 20% of PVS plasma volume moving into the tissues during exercise or exposure to hypoxia (Yamamoto et al., 1983; Yamamoto, 1988; Pearson and Stevens, 1991). Thus increased SVS flow and, therefore, more rapid mixing of the SVS and PVS could buffer ionic or osmotic changes in the PVS during exercise or other forms of stress. When only plasma flows through the SVS, the probable function is to deliver nutrients and remove waste (Steffensen and Lomholt, 1992), with gas transfer occurring directly across the skin. Flushing the SVS with RBCs following exercise would enhance O₂ uptake across the skin. The SVS was less flushed with RBCs during aquatic hypoxia (no ASR) than it was when fish were in normoxic water during recovery from hypoxia or both immediately following and upon 1 h recovery from exercise. However, these changes were small and not reflected in differences in the reduction in PVS haematocrit between hypoxic and exercise exposed groups.

The SVS is restricted to and variable amongst the Actinopterygii and related to function rather than phylogenetic position (Skov and Bennett, 2004). A more extensive SVS may be linked to increased scope for activity, which in turn may be associated with enhanced skin O₂ uptake via the SVS during exercise. There is no evidence of an SVS in lungfish (Vogel and Mattheus, 1998), but whether our Silurian vertebrate ancestors had a SVS subsequently lost in the Acanthodii or the SVS evolved in the Actinopterygii is unknown. What is clear is that the organization of circulation and requirements for fluid balance are different in aquatic and terrestrial vertebrates. It follows that the changes in the vertebrate circulation between extant fish and mammals are to be expected. Tissue edema and fluid drainage can be problematic in terrestrial mammals, as they have high blood pressure and are subject to the effects of gravity that come with living in an aerial environment. While there is a paucity of information on the mechanisms underlying tissue fluid balance in fish, fish blood pressures are lower than those of mammals, and living in water eliminates the effects of gravity because the density of the environment is similar to that of their own body. Furthermore, there is no evidence that the SVS plays a role in countering tissue edema and aiding fluid drainage. The SVS is not associated with the celiac, mesenteric or renal arteries, and no association has been made between increased SVS flow and food intake (P. V. Skov and J.F.S., unpublished observations) (Skov and Bennett, 2004). The SVS, unlike the lymphatic system of mammals (Robinson and Mead, 1973; Kayama and Iijima, 1976), is not involved in uptake from the gut. Therefore, it is clear that the SVS is not functioning in the same way as the mammalian lymphatic system. Additionally, the SVS vessels are not blind-ending, and flow originates directly from the primary vascular system. Following hypoxia and exercise, RBCs enter the SVS and flow increases. Thus the presence of molecular markers for the mammalian lymphatic system in the SVS vessels cannot be used as indicators of lymphatic function in fish. In summary, and in contrast to the conclusions of Yaniv et al. (Yaniv et al., 2006), the SVS and the PVS are not distinct vascular systems, and there are circumstances during which the SVS does contain RBCs. We conclude, and agree with recent arguments by Vogel (Vogel, 2010), that the SVS has different form and function from, and is not a suitable model for, the mammalian lymphatic system.

A total of 135 glass catfish (wet mass=0.583±0.019 g, standard length=47.6±0.5 mm; means ± s.e.m.) were transported to the Department of Biology and Chemistry at the City University of Hong Kong from a local supplier and maintained at 22°C in dechlorinated municipal tap water under a natural photoperiod between the months of November 2010 and May 2011. Fish were fed daily to satiation with frozen bloodworms, but food was withheld for 24 h prior to experimentation. All animal care and experimental protocols comply with animal ethics regulations outlined by the Department of Biology and Chemistry at the City University of Hong Kong.

The same protocol was used for collecting blood for all treatments. Fish were euthanized via head trauma, and the caudal fin was severed to bleed the caudal vein and artery into a heparinized microcapillary tube. This technique allowed the glass capillary tube to be placed in immediate contact with the caudal artery and vein, thus ensuring other fluids were not collected. Furthermore, fish capture and blood collections were done quickly (<20 s) to minimize any effects that basic handling stress may have on the integrity of the RBCs. Samples (8–20 μl depending on the size of the fish) were centrifuged at 3500 g for 3 min and used to determine haematocrit.

It was first important to determine the size of RBCs during rest and if size and shape would change if cells swelled during stress. This would permit us to select fluorescent microspheres to inject into the glass catfish PVS and SVS to track distribution and determine flow patterns. To determine RBC size, blood was collected from five fish and smeared onto glass slides with either a drop of Cortland's saline or one of two concentrations of a β-adrenergic agonist, isoproterenol hydrochloride (Iso; cat. no. 15627, Sigma-Aldrich, St Louis, MO, USA), 0.001 mmol l⁻¹ (prepared in Cortland's saline) (Caldwell et al., 2006). The RBCs were imaged using a fully motorized compound microscope (model BX61, Olympus, Tokyo, Japan) with a Cool CCD still image camera (model DP-72, Olympus) with capturing software (Image-Pro Plus, version 6.2, MediaCybernetics, Rockville, MD, USA). For each blood sample, 20 RBCs were randomly selected and measured using the tools integrated in the capturing software. Mean lengths for long and short axes and the ratio between the long and short axes were reported so that the appropriate size for the fluorescent microspheres could be selected.

To determine blood flow patterns, fish were lightly anaesthetized with tricaine methane sulfonate (MS-222; 1:25 diluted stock solution that was 4% w/v made using 1 mol l⁻¹ Tris buffer at pH 9). Each fish was positioned on a moist sponge, left lateral side facing up and continuously moistened using a damp paper towel. Using an Olympus stereoscope (model SZX12), the DA of the fish was located, and a 10 μl Hamilton syringe (model 80330) with a 33 G 0.8 inch PT2 needle (model 7803-05) was positioned ~5 mm posterior to the periphery of the operculum. A 2 μl volume of fluorescent microspheres was slowly (over ~10 s) injected into the PVS of the fish via the DA. Four different-sized microspheres, 0.02 μm (green), 1 μm (blue), 4 μm (green) and 10 μm (red) (F8787, F8815, F8859 and F8834, respectively; Invitrogen, Carlsbad, CA, USA) were used. Prior to injection, aliquots were sonicated for 5 min and then centrifuged at 3000 g for 3 min to reduce agglomeration, as per manufacturer recommendations.

Immediately upon injection, fish were transferred to a plastic Petri dish filled with ~20 ml water without anaesthetic, enough to completely cover the fish, which was not restrained but remained calm in the Petri dish. We observed the fish using fluorescent light (X-Cite series 120Q EXFO using GFP/UV/RFP filters, Lumen Dynamics Group, Ontario, Canada) and videos of the injection site were recorded to trace distribution of the microspheres (video camera: Panasonic GPUS932HE; software: AVer Mediacenter version 1.7.3, H727 PCle Hybrid DVBT HDMI) until no further changes could be observed.

Fish were transferred from holding tanks using dip nets and held individually, one per 500 ml glass beaker, in continuously aerated, clean water maintained at 22°C for 2 h before experiments commenced. During this time, all beakers were placed in white polystyrene containers with lids so fish would stay light in colour; dark background results in melanophore dispersal mediated by cholinergic, muscarinic-type receptors (Fujii et al., 1982), making fish less transparent and harder to visualize. This was also done to ensure that fish were in an unstressed state so that valid resting conditions could be observed (see below). Additionally, because it was determined from earlier observations that anaesthetic promoted RBC distribution into the SVS, fish were not anaesthetized for Series 2 experiments so that other regulatory mechanisms could be revealed.

These fish remain quiet when handled carefully and gently, but stressed glass catfish have RBCs in the SVS. If we transferred a fish to the Petri plate for observation without RBCs entering the SVS, we referred to this fish as ‘resting’. Because of subsequent experimental design we also determined that multiple transfers, if done gently, did not promote RBC entry into the SVS. For resting measurements, unanaesthetized individual fish (N=28) were placed in plastic Petri plates containing 20 ml of well-aerated, clean water and examined under a dissecting microscope using white light, similarly to how fish were observed using fluorescent light, as described above. The caudal fin region of the fish was examined for the presence of RBCs in the SVS. Next, by gently tilting the Petri plate to the side, the fish was positioned (ventral side up) against the side wall of the Petri plate. Because of the transparent nature of this species, the heart was also visible to the naked eye. Therefore, the image of the beating heart was projected live onto a computer monitor and f_H was visually assessed and recorded as the number of beats per minute (using 20 s assessments). After each count, the caudal fin and posterior ventral regions of the fish were re-examined for presence or absence of RBCs in the SVS. This was replicated three times in order to achieve a mean f_H per fish and form a consensus as to the status of RBCs in the SVS. Videos were recorded for each individual.

Blood was collected from another group of resting fish (N=10) and used to determine resting haematocrit. Fish were then gently patted dry, weighed (g), measured (standard length, mm), and length-to-weight ratio [Fulton's condition factor, K=100(W/L³)] was calculated.

Hypoxic conditions were created in a 30-litre white polystyrene container either allowing the fish to use ASR or directly covering the surface with a clear Perspex sheet to prevent the fish from using ASR. Nitrogen was bubbled into the water, which reduced water oxygen levels (P_w,O₂) to 20 mmHg. Hypoxic conditions were continuously monitored using a pre-calibrated fibre optic O₂ sensor (PreSens, Precision Sensing, Regensburg, Germany) (tip diameter 50–140 μm) amplified using an Oxy-micro oxygen meter and signal amplifier (PreSens, Precision Sensing). Data were collected throughout the duration of each experiment at a sampling rate of 1 Hz, and integrated with the manufacturer's software packages for PC Windows. All data were saved as text files and analyzed using Acqknowledge data acquisition software (version 3.7.3, BIOPAC Systems, Inc., Goleta, CA, USA).

We used four groups of fish, for f_H and RBC observation in hypoxia either permitted (group 1, N=7) or not permitted to access the surface (group 2, N=8), and for blood sampling immediately (group 3, N=10) or 1 h (group 4, N=10) following exposure to hypoxia with no access to the surface. All fish were observed (as described above) prior to hypoxia experiments to ensure that they were exhibiting resting conditions in terms of f_H and the absence of RBCs in SVS. Hypoxia exposure (P_w,O₂=20.3±0.1 mmHg) lasted for 5 min during which time they were either permitted surface access where they were observed performing ASR or not permitted surface access due to a clear Perspex lid on the surface of the water. For groups 1 and 2, following hypoxia exposure, fish were immediately (within 10 s) observed and data collected for f_H and the status of RBCs in SVS. Fish were then allowed to recover for 1 h (1 h recovery) in well-aerated, clean water and monitored again. We were not able to monitor all fish immediately and so the N-values for the immediate and after 1 h time points are different. Fish were permitted to recover overnight (12 h), after which time survival was assessed and fish were returned to holding facilities and not used for further experimentation. For this and all subsequent experiments, fish that did not survive 12 h following treatment were not included in analyses. Fish from group 3 were exposed to hypoxia (no ASR), and after 5 min removed from treatment. Blood was immediately collected from each fish to determine haematocrit, and fish were weighed and measured as described above. Fish from group 4 were exposed to hypoxia (no ASR) in the same manner, and after 5 min removed from treatment, and allowed 1 h recovery in well-aerated, clean water. Then blood was sampled from each fish to determine haematocrit and fish were weighed and measured as described above.

We used three groups of fish, for f_H and RBC observation immediately and then 1 h following exercise (group 1, N=10) and for blood sampling immediately (group 2, N=9) or 1 h (group 4, N=11) following exercise. First, all fish were observed (as described above) prior to exercise to ensure resting conditions. Exercise was simulated by chasing individual fish for 30 s with a plastic probe in a beaker of well-aerated, clean water. Following this, fish were permitted recovery for 30 s during which time they were transferred to either the microscope, blood sampling station, or well-aerated, clean water for 1 h recovery. Blood was sampled to determine haematocrit and fish were weighed and measured as described above. The fish that were observed for f_H and RBCs in the SVS were then permitted to recover overnight (12 h), after which time survival was assessed and fish were returned to holding facilities and not used for further experimentation.

All fish for the pharmacology experiments were observed (as described above) prior to treatments to ensure resting conditions. Then, each fish was exposed to one of the following treatments in well-aerated, clean water for 5 min (N=7–12 for each group): β-agonist Iso (Sigma-Aldrich 15627) at a final concentration of 1 mmol l⁻¹ (Caldwell et al., 2006; Rummer et al., 2010; Rummer and Brauner, 2011), β-antagonist Prop (Sigma-Aldrich P0884) at a final concentration of 2 mmol l⁻¹ (Caldwell et al., 2006; Rummer et al., 2010; Rummer and Brauner, 2011), or α-antagonist PHT (Sigma-Aldrich P7547) at a final concentration of 0.01 mmol l⁻¹ (Namoto and Yamada, 1987; Kågström and Holmgren, 1997). Immediately following 5 min, exposed fish were observed as described for resting conditions and transferred back to well-aerated, clean water. In the case of Prop, treatment fish were observed and subsequently exposed to the agonist (Iso) for 5 min, observed (Prop+Iso) and then transferred back to well-aerated, clean water. Following either single or double (sequential) exposure to drugs, fish were allowed 1 h recovery in well-aerated, clean water. Following 1 h recovery, measurements were taken for Iso and Prop+Iso treated fish and observations made as described above. Finally, fish were permitted to recover overnight (12 h), after which time survival was assessed and fish were returned to holding facilities and not used for further experimentation. All drugs were dissolved in well-aerated, clean water from the same source used for maintaining fish in the laboratory. All solutions were made fresh for each experiment.

Data are presented as means ± s.e.m. unless otherwise stated. All P_w,O₂ data were saved as text files and analyzed using Acqknowledge data acquisition software. All data were compared statistically between treatments, and statistical differences were detected via paired t-test, ANOVA or repeated measures ANOVA and, when necessary, a post hoc Holm-Šídák multiple comparisons test. All statistical analyses were conducted using SigmaStat for Windows 3.5.0.54 (Systat Software, Inc.), and all analyses were interpreted using α<0.05 to determine statistical significance.

The authors thank Alice Chan Ka Yee, Alex Lin and the animal care staff employed by the City University of Hong Kong for technical assistance throughout this project as well as two anonymous reviewers for their constructive and helpful comments on this manuscript.