Background
The physiology of CSF and ISF circulation in health and disease has become an area of increasing interest. In the brain it is now recognised that turnover and clearance of these fluids are critical for circulating nutrients, waste removal, as well as regulating intracranial pressure [
1]. Irregularities in the movement of ISF and CSF are now recognised to play an important role in diverse neurological pathologies, including Alzheimer’s disease, multiple sclerosis and, in the spinal cord, syringomyelia [
2‐
4]. Despite this, there is a paucity of knowledge on the anatomical pathways and drivers of this fluid movement. This has hampered attempts to understand the pathophysiology of CSF disorders including syringomyelia, a condition where there is accumulation of fluid within the spinal cord associated with a wide range of pathologies, including spinal cord injury. There may exist a common pathway that results in an imbalance of fluid inflow and outflow, leading to fluid accumulation. In order to grasp what abnormal physiology might entail, normal physiology needs to be better understood.
Perivascular spaces are sites of great importance in CNS fluid exchange. Regional variations in the density of blood vessels may affect the susceptibility of the spinal parenchyma to fluid transport pathologies. This is supported by a previous study that demonstrated that fluorescent tracer in the grey matter was transported along vasculature structures in a radial pattern to the pial surface. In contrast, tracer from the white matter preferentially travelled longitudinally along the spinal cord [
5]. An inefficient fluid and solute exchange system in either white or grey matter may result in accumulation of excitotoxic and oxidative factors and further local tissue damage.
Much of the research in perivascular transport and CNS fluid homeostasis has been performed on the brain, which has generated a growing body of literature. However, the mechanisms of fluid transport in the spinal cord remain largely unexplored. Until recently, there was little information on the pathways of normal spinal fluid outflow [
5]. We have provided new insight into the effects of cyclical intrathoracic pressures on tracer movement from the subarachnoid space into spinal perivascular spaces (at least at the leptomeningeal level) (Liu et al. under review). The aim of the current study was to determine the effects of respiratory and cardiovascular parameters on movement of tracer from the spinal cord parenchyma.
Methods
Male Sprague–Dawley rats 8–12 weeks of age and weighing 280–430 g were used. All procedures were approved by the Animal Ethics Committee at Macquarie University (Animal Research Authority Number: 2016/032) and conducted in accordance with the Australian Code of Practice for the Care and Use of Animals for Scientific Purposes.
Surgical preparation
Animals were placed under general anaesthesia with 5% isoflurane in oxygen, then positioned supine on a heating pad and maintained under anaesthesia with 1.5–2.5% isoflurane in 0.2 L/min of oxygen. Heart rate, oxygen saturation, respiratory rate, and temperature were closely monitored by pulse oximetry (PhysioSuite®, Kent Scientific Corporation, CT, USA) and rectal thermometer connected to a homeothermic heating pad (Harvard Apparatus, Holliston, MA, USA).
The femoral artery and vein were cannulated with polyethylene catheters pre-loaded with heparinised 0.9% saline (5000 IU/L) and attached to a 3-way-tap. The arterial line was connected to a pressure transducer, enabling the continuous measurement of blood pressure. The venous line was used to administer saline and drugs as needed. A 14G endotracheal tube was then inserted into the trachea and secured in place with silk sutures. The endotracheal tube was connected to a respiratory circuit, delivering 1.5–2.5% isoflurane in oxygen. The expiratory tubing was connected to a capnometer (Capstar-100, CWE Inc., Ardmore, PA, USA), to allow end-tidal carbon dioxide (CO
2) monitoring, and to a respiratory circuit pressure manometer made in-house, to measure relative changes in intrathoracic pressure. In the subset of rats where tachycardia was induced, a custom-made atrial pacing wire was inserted into the right external jugular vein. The pacing wire was connected to an isolated pulse stimulator (A-M Systems Inc, model 2100) [
6]. At this point all physiological vital statistics were recorded continuously for the remainder of the experiment on a data acquisition interface, Power1401 (Cambridge Electronic Device, Cambridge, UK) and recorded using Spike2 software (v6. CED Ltd., Cambridge, UK). Arterial blood gas was analysed for pH, partial pressure of oxygen, and partial pressure of CO
2 (VetStat Electrolyte Blood Gas Analyser, IDEXX Laboratories Pty. Ltd, Australia). Respiration, blood pressure and heart rate were then modulated in isolation to investigate the effect of these physiological variables on fluid flow out of the spinal cord.
Modulation of physiological parameters
To examine the effects of changes in respiration, the conditions tested were positive intrathoracic pressure only vs positive and negative intrathoracic pressures. Animals were either allowed to breathe spontaneously, generating both negative and positive intrathoracic pressures while connected to the respiratory circuit, or a neuromuscular blockade was administered (pancuronium bromide 0.8 mg IV induction, 0.4 mg/h IV maintenance, Astra Pharmaceuticals Pty Ltd, Sydney, NSW, Australia) followed by mechanical ventilation using a small animal ventilator (Harvard 7025 Rodent Ventilator, set at a tidal volume of approximately 1.2 mL). In these animals the negative intrathoracic pressure typically generated by natural respiration was eliminated. Pancuronium bromide is known to have vagolytic effects, resulting in hypertension and tachycardia in some animals. To counteract these effects, metoprolol (10–15 mg/kg in 0.9% saline IV) was administered when necessary. The end tidal CO2 was maintained within a physiological range of 3.5–4.5%.
In anaesthetised spontaneously breathing rats (SB), a respiratory rate of 50–55 breaths/ min was observed, with resultant CO
2 retention and respiratory acidosis. The respiratory rate and blood gas profile of the mechanically ventilated controls (MV) was matched to that of their spontaneous breathing counterparts. These animals also served as controls for experiments investigating the effects of heart rate and blood pressure modulation on fluid outflow. Other variables including weight, heart rate, CO
2, circuit pressure and mean arterial pressure (MAP) were recorded. Comparison of these physiological variables is shown in Table.
1.
Table 1
Comparison of physiological variables
Respiration: spontaneous breathing vs Mechanical ventilated controls |
Mass | ns | p = 0.03 |
Mean arterial pressure | ns | ns |
Heart rate | ns | ns |
CO2 | ns | ns |
Blood pressure: high blood pressure vs low blood pressure |
Mass | ns | 0.005 |
Respiratory rate | ns | ns |
Heart rate | 0.0008 | 0.0005 |
CO2 | ns | ns |
Circuit pressure | ns | ns |
Heart rate: high heart rate vs low heart rate |
Mass | ns | 0.006 |
Respiratory rate | ns | ns |
Mean arterial pressure | ns | 0.004 |
CO2 | 0.02 | 0.0003 |
Circuit pressure | ns | ns |
To examine the effects of blood pressure, hypertensive rats were compared to the MV group, which had an approximate MAP of 70 mmHg. Hypertension was induced by an infusion of phenylephrine, used to raise the MAP to a target of 140 mmHg (a ~ 40% increase from base line and approximately double that of controls). To prevent baroreflex compensation, and maintain heart rate, the nicotinic receptor antagonist hexamethonium was administered. All hypertensive animals were mechanically ventilated to relative hypercapnic levels (matching the MV control group).
To examine the effect of heart rate, tachycardic rats were compared to the MV control group, which had an approximate heart rate of 330 beats/min (bpm). A pacing wire (described above) was connected to an isolated pulse generator (A-M Systems Inc, model 2100). A 2 ms pulse duration and an amplitude ~ 1.0 V was used to increase heart rate to ≥ 500 bpm. Blood pressure remained stable for the duration of atrial pacing. All tachycardic animals were mechanically ventilated to a relative hypercapnic level (matching the MV group).
Surgical procedures for investigation of tracer efflux from spinal cord
Once the desired physiological parameter was manipulated, the efflux of spinal ISF was assessed by analysing the rostrocaudal distribution and clearance of fluorescent ovalbumin (AFO-647) injected into the spinal parenchyma. The spinal grey and white matter were investigated separately. Animals were placed in the prone position, and following muscle dissection to expose the bony anatomy, a right sided hemilaminectomy at T1 was performed. A Hamilton syringe (Hamilton Company, Reno, USA) fitted with a 34G needle and positioned in a stereotaxic frame was inserted into the spinal parenchyma at a position 0.5 and 1 mm lateral to the dorsal midline vein, for the grey and white matter respectively [
5]. To prevent CSF leak the dura was punctured in a single pass and cyanoacrylate glue was applied around the puncture site. A 500 nL bolus of fluorescent ovalbumin tracer, Ovalbumin Alexa-Fluor®-647 conjugate (AFO-647, Life Technologies, Victoria, Australia) was injected at a rate of 2 nL/s. The point of injection was confirmed by the presence of fluorescent microspheres FluoSpheres™ (ThermoFisher Scientific, Massachusetts, USA) (Fig.
6a). The needle was left in situ for the duration of the tracer experiment (180 min), after which the animal underwent transcardiac perfusion with 4% paraformaldehyde.
Tissue processing and immunohistochemistry
The brain and spinal cord were dissected en bloc for macroscopic fluorescent imaging. The spinal cord was then segmented from C2–T4 after post-fixation and cryoprotection. Each spinal segment was embedded in Optimal Cutting Temperature Compound (Tissue-Plus™ O.C.T. Compound, Thermo Fisher Scientific, Massachusetts, USA) and frozen on dry ice. Axial sections were cut on a cryostat at 40 μm thickness and mounted onto glass slides for immunohistochemistry. To label the endothelium, slides were incubated with the primary Rat Endothelial Cell Antibody (RECA-1, Abcam, Cambridge, UK) in 4% Normal Donkey Serum (NDS), followed by the secondary antibody, anti-mouse IgG Alexa-Fluor®-488 (Molecular Probes, Life Technologies, New York, USA). Smooth muscle cells were then labelled by anti-actin, α-smooth muscle-Cy3™ antibody (SMA, Sigma-Aldrich, St. Louis, Montana). Slides were cover-slipped with fluorescent mounting medium (DAKO, NSW, Australia).
Image acquisition
The rostrocaudal macroscopic distribution of AFO-647 along the neuraxis was captured with white-light and a single fluorescent channel (excitation wavelength 630 nm and emission wavelength 700 nm, exposure time of 4 s) using the small animal optical imaging system MS FX PRO (Bruker UK Ltd.). Images were taken from both the dorsal and ventral directions.
Spinal cord axial sections from C2–T4 were imaged with a Zeiss Axio Imager fluorescence microscope (Carl Zeiss Microimaging GmbH, Germany). Immunohistochemistry was used to image the spinal vasculature and identify vessel types. Arterioles were identified as vessels positive for RECA-1 and SMA, whereas venules and capillaries were labelled by RECA-1 only. Blood vessels that had a luminal diameter < 6.5 μm were classified as capillaries. Confocal microscopy (LSM 880, Carl Zeiss Microimaging GmbH, Germany) was used to further characterise vascular structures and the central canal.
Image processing and analysis
In all macroscopic fluorescence images, the white-light image was used to delineate regions of interest at each spinal level from C2-T4. The tracer signal was then measured from the fluorescent image at each of these spinal levels on the dorsal and ventral surfaces. In the microscopic axial sections, the integrated density of the CSF tracer (mean pixel density multiplied by area) was calculated. The whole spinal cord, white matter, and grey matter were identified as regions of interest using the manual tracing tool in ImageJ, carefully excluding the dura and nerve roots. At least three sections were analysed per level from C2-T4, and averaged. All analysis was carried out using ImageJ software, version 1.46r [
7].
Statistical analysis
Physiological vital statistics were compared with two tailed Student’s t-test. Fluorescence intensities (integrated densities and mean pixel densities) were compared using two-way analysis of variance (ANOVA) and adjusted for multiple comparison using Bonferroni’s post-hoc tests. A p value ≤ 0.05 was considered statistically significant. All fluorescence values were expressed as mean ± standard error of the mean (SEM). All physiological parameter values were expressed as mean ± standard deviation (SD). GraphPad Prism (v7.02, GraphPad Software Inc, California) was used to perform all statistical analysis.
Discussion
Hypertension and tachycardia appear to increase tracer efflux from the cord into the subarachnoid space. Intrathoracic pressure had less of an effect on the distribution of tracer away from the injection site (Figs.
4,
5). However, there was increased tracer signal macroscopically in the craniocaudal axis in spontaneous breathing rats compared with mechanically ventilated animals. This could be explained by the fact that, unlike the spinal dural sac which is directly exposed to shifts in epidural venous plexus pressures (brought about by changes in intrathoracic pressure) and readily enlarges or collapses in response to these pressure gradients [
9], the spinal cord itself may be protected from these fluctuating pressure changes. The amount of solute egress from the cord is therefore minimally affected by changes in respiration. However, any tracer that is drained into the subarachnoid space from the cord parenchyma is mixed more efficaciously and distributed further along the spinal subarachnoid space, possibly resulting in higher fluorescence intensities on the surface of the spinal cord. Moreover, the amount of tracer that dissipated into the subarachnoid space may be too small for a difference in tracer signal (between the spontaneous breathing and control groups) to be detected when it is subsequently redistributed back into cord parenchyma at remote spinal levels.
In our previous study of molecular transport from the rat spinal cord, tracer redistribution was assessed at 20 and 60 min after delivery into the parenchyma [
5]. Fluorescence intensity did change over time, suggesting that tracer spread was not artefactual due to the tracer infusion, but indicative of solute transport. Nevertheless, interstitial spread was limited to within one level caudal and rostral to the injection point. Here, we have demonstrated that even 180 min after delivery (the upper limit of length of time rats could be kept haemodynamically stable under altered physiology), endogenous tracer movement remained limited to adjacent spinal levels. Tracer injected into the grey matter demonstrated radial redistribution outwards from the injection point. Delivered into the white matter, AFO-647 was largely confined within the parallel myelinated tracts. This is consistent with previous distribution patterns with isotropic distribution in the grey matter and anisotropic distribution in the white matter [
5]. These results suggest that diffusion governs endogenous spinal solute and fluid transport, although advanced techniques such as integrated optical imaging and real-time iontophoresis [
10] will be required to confirm this. There is mounting in silico evidence that convective flow, as suggested by the glymphatic theory, is implausible in the extracellular space of the CNS [
11]. Concerns about the interpretation of ex vivo preparations in efflux studies should be addressed by real-time in vivo imaging of injected extracellular space tracers. Arbel-Ornath et al
. [
12] employed two-photon intravital microscopy to track fluorescent dextrans injected into murine brain. A rapid co-localisation to the arterial basement membrane of the perivascular space was reported. No tracers were detected around venous structures. Biexponential reduction of tracer over 30 min was interpreted by the investigators as evidence of bulk flow in the perivascular space. However, in these experiments the dura was opened, compromising the hydraulic integrity of the system. No equivalent investigation has been undertaken in the spine.
Whether CNS solutes are cleared primarily into lymphatics of the dura and large blood vessels, or into the subarachnoid space is still unclear [
13,
14]. There are likely differences among animal species and variations with age and pathological conditions. Substantial accumulation of tracer around the pia, with subsequent perivascular re-entry of tracer into the parenchyma at spinal levels distant to the injection was not observed in our previous study, up to 60 min after injection [
5]. The results of the current study, however, indicate that transport of tracer from the spinal cord is at least partially to the subarachnoid space.
Previously, Hadaczek et al. [
15] showed adrenaline-induced hypertension and tachycardia promoted apparent bulk flow of different sized macromolecules through the extracellular space of rat brain. In a murine model of Alzheimer’s Disease exposed to cerebral hypoperfusion (and presumably low blood pressure), mural Aβ was observed to accumulate in leptomeningeal vessels, reflecting reduced solute drainage [
16]. The results from our study support a similar role of blood pressure and heart rate in the spinal cord, with increases to either physiological value promoting tracer movement. Mathematical modelling also supports cardiac pulsations driving this flow [
17]. In contrast, another study investigating CSF tracer influx into the brain reported that decreased heart rate correlated with improved molecular clearance from mouse brain [
18]. It is interesting to note that a recent study by our group looking at the same physiological factors found that the alternating positive and negative intrathoracic pressures that occur during spontaneous breathing had a greater effect on CSF flow from the subarachnoid space into the spinal cord when compared with mechanically ventilated controls with continuous positive intrathoracic pressure. It is possible that the physiological factors that govern tracer influx exert different effects on tracer efflux.
In the current experiments, tracer accumulated in distinct layers both internal and external to the smooth muscle layer of intramedullary and extramedullary arterioles and arteries (Fig.
7). Around veins, venules, and capillaries, AFO-647 was deposited in close proximity to the endothelium. Thus, all blood vessel types have been implicated in spinal outflow, similar to findings from recent spinal inflow studies [
8]. These findings recapitulate results from earlier work on molecular transport pathways in the normal spinal cord [
5]. Dedicated ultrastructural studies will be required to clarify the precise anatomical relationships of the tracer and the various compartments of the perivascular space. However, these results are supportive of the role of vascular basement membrane(s), as well as the compartments between the glia limitans and pial sheath, or the adventitia, in mediating solute clearance. It should be noted that findings from ex vivo preparations should be cautiously interpreted as there is evidence that tracer deposition in some regions may be artefactual, induced either by changes that occur upon death or during the process of perfusion and fixation, which can substantially alter the structure of the perivascular spaces and tissues [
13,
19]. Nevertheless, it is intriguing that pathways of influx and efflux appear shared. This raises the possibility that there is bidirectional, to-and-fro mixing of fluid in the perivascular space that is able to rapidly distribute solutes. The direction of tracer redistribution depends, therefore, on whether it is injected into the subarachnoid space (inwards, towards the central canal) or into the parenchyma (outwards, towards the pial surface) [
20,
21]. It is possible that during inspiration, associated with large magnitude CSF pulse waves in the subarachnoid space, CSF is driven into the spinal cord parenchyma along perivenous and periarterial spaces. Within the spinal cord however, increased blood vessel pulsatility occurring with increases in blood pressure and heart rate, drive more fluid transport through the interstitium, enhancing efflux of solutes and fluid from the spinal cord into the subarachnoid space. In the current study, higher MAP/pulse pressure and heart rates were associated with greater endogenous tracer deposition remote from the delivery site, indicating greater efflux locally from the point of injection. Arterial pulsations, therefore, promote molecular transport in the spinal cord.
The role of the central canal in CSF and molecular transport is largely unknown. Milhorat and colleagues [
22] had previously ascribed the central canal with a “sink” function for solute and metabolites. Spontaneous breathing, tachycardia and hypertension did not appear to have major effects on drainage into this compartment from the extracellular space. More than two decades ago, Cifuentes et al
. [
23] proposed the possibility of bidirectional fluid transport between the central canal and the subarachnoid space via perivascular spaces, particularly via peri-arterial pathways of the central branches of the anterior spinal artery, consistent with our findings. Contiguous bands of tracer were detected between the subependymal microvasculature and the central canal ependymal cells. To confirm and further examine this putative pathway, the next step is to fluorescently label ependymal cells (such as with F-actin [
24]) in future intravital studies of influx and efflux. Tanycytes and complex ependymal basement membranes (labyrinths) are thought to subserve this putative connection between CSF, ISF and the central canal [
25]. The role of the central canal in fluid exchange may not be as important in humans as there is progressive atresia of this structure with age [
26].
A comprehensive model of spinal fluid transport that consolidates the findings from these experiments, as well as data from previous laboratory and computational studies, may still be out of reach. There is, however, evidence that the same anatomical pathways subserve influx and efflux, so it is reasonable to deduce that the two processes occur simultaneously and may be subject to the same physiological drivers. We have provided evidence here that tachycardia and hypertension enhance movement of tracers injected into the spinal cord parenchyma. However, this has not necessarily resulted in increased overall solute drainage. An interplay of factors—such as the width of the perivascular space as dictated by the phase difference of the arterial wave with CSF pulse wave [
27], the stiffness of the arterial wall, and the opposing forces of influx and efflux which appear to occur along similar pathways—may ultimately determine the net direction and magnitude of fluid and solute exchange as the extremes of physiology are approached. Moreover, the role of the central canal in mediating intramedullary drainage, and local microanatomical geometries in the subarachnoid space at different spinal levels add further layers of complexity. Simple tracer experiments in animal models, while invaluable in our attempts to elucidate the basic mechanisms of fluid transport, cannot capture fully the driving forces governing a dynamic process.
The perivascular space has been recognized as a site of paramount importance in mediating spinal fluid exchange. After injection of AFO-647 into the spinal parenchyma, there was selective accumulation of tracer around radially projecting intramedullary as well as extramedullary blood vessels. There was a contrasting lack of subpial fluorescence, suggesting that solutes could access preferential routes for efflux from the interstitium. With time, it was apparent that tracer drained at least partially into the spinal subarachnoid space, redistributing over the pial surface distant to the injection site. Moreover, higher MAP/pulse pressure and heart rates were associated with more endogenous tracer deposition remote from the delivery site, indicating greater transport locally from the point of injection. Arterial pulsations, therefore, promote spinal efflux.
Limitations
Caveats of the techniques used in this study need to be highlighted. Firstly, we cannot exclude the possibility that injecting 500 nL of tracer into the spinal parenchyma could cause overloading of the system. To minimise this risk, we have chosen a volume that is comparable to or lower than similar studies carried out in mouse brain [
2,
28‐
31]. Secondly, although the complex experimental techniques were designed to modulate a single physiological parameter and measures were taken to maintain all other variables, it is possible that sympathomimetic medications such as phenylephrine can induce vasoconstriction or disturbance of spinal autoregulation. These alterations are difficult to quantify. In the current study measures were taken to mitigate changes to the integrity of the intrathecal sac. The needle was inserted with a single-pass (the dura was not opened prior to needle insertion) and cyanoacrylate glue was placed around the needle. The needle was also left in place for the duration of the experiment prior to perfusion. However, it has to be acknowledged that piercing the dura may impact the pressure within the dural sac and withdrawing the needle at the completion of the experiment just prior to perfusion-fixation would cause CSF leak and altered fluid flow within the spinal cord itself. As mentioned previously, another limitation of ex vivo experiments are the possible post-mortem ultrastructural changes that occur after cardiorespiratory arrest and fixation with aldehydes that may influence the location of the tracer and may not represent the distribution in living animals. While intravital imaging has allowed for some of these flaws to be overcome, there are still unresolved technical challenges in spinal in vivo imaging that prevent analysis of fluid and solute transport at the microscopic level. The much deeper field of visualisation is just one of these barriers.
Clinical implications
There are no effective medical treatments for the deleterious secondary effects of spinal cord injury or the subsequent formation of fluid-filled cysts (syringomyelia). These effects are thought to be mediated by the release of excitotoxic factors that further potentiate local damage after primary injury [
32]. Understanding the factors that drive fluid efflux from the spinal cord may provide an opportunity to enhance the removal of excess fluid, and harmful proteins, amino acids and other molecules.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.