Changes in intrathoracic pressure, not arterial pulsations, exert the greatest effect on tracer influx in the spinal cord

After induction of general anaesthesia with 5% isoflurane in oxygen, each animal was positioned supine on a heating pad and maintained under anaesthesia with isoflurane in 0.2 L/min of oxygen. Heart rate, oxygen saturation, respiratory rate, and temperature were continuously monitored by pulse oximetry and rectal thermometer. Rats undergoing two-photon excitation microscopy were pre-medicated with subcutaneous midazolam (1 mg/kg) and buprenorphine (0.05 mg/kg) prior to anaesthesia with isoflurane.

A right transverse inguinal incision was made, and the femoral neurovascular bundle was exposed for cannulation by arterial and central venous polyethylene catheters. This was followed by a midline suprasternal incision. The trachea was exposed for insertion of an endotracheal tube. This was connected to a respiratory circuit, delivering isoflurane in oxygen. The egress tubing was connected in series to a capnometer (Capstar-100, CWE Inc., Ardmore, PA, USA), for continuous end-tidal carbon dioxide (CO₂) monitoring, as well as to a custom-made respiratory circuit pressure manometer. At this point, for the subset of rats where heart rate was modulated, a 3-cm length of the right external jugular vein was dissected out. A custom manufactured atrial pacing wire was inserted into this vein to rest just above the sinoatrial node. The pacing wire was connected to an isolated pulse stimulator (A-M Systems Inc, model 2100) [20].

The animal was then repositioned prone and the respiratory circuit manometer and arterial line were connected to pressure transducers, enabling the continuous measurement of blood pressure and circuit pressure. All vital statistics were recorded continuously for the remainder of the experiment on a data acquisition interface, Power1401 (Cambridge Electronic Device, Cambridge, UK). Arterial blood gas was analysed for pH, partial pressure of oxygen, and partial pressure of CO₂. Each physiological variable—respiration, blood pressure and heart rate—was then manipulated in isolation to investigate its effect on spinal subarachnoid space flow and parenchymal inflow (Fig. 1).

After the desired physiological parameter target was reached and maintained, one of three surgical procedures was performed to investigate fluid transport in the spinal subarachnoid space, and into the spinal cord. CSF tracer movement was assessed using one of the following fluorescent tracers injected into the cisterna magna: ICG (Verdye, Aschheim-Dornach, Germany); 1 μm FluoSpheres™ (13% v/v: ThermoFisher Scientific, Massachusetts, USA), Ovalbumin Alexa-Fluor®-647 conjugate (AFO-647: Life Technologies, Victoria, Australia). CSF tracer movement was imaged using NIR intravital imaging, two-photon microscopy, and ex vivo microscopy.

The redistribution of fluorescent tracer from the cisterna magna to the cervicothoracic spinal cord interstitium was assessed qualitatively and quantitatively. As described above, a 30G needle was used to inject a 10 μL volume of 20 µg/µL AFO-647 tracer into the cisterna magna at a rate of 33 nL/s. The needle was left in situ ensuring CSF leakage did not occur. After 10 min the animal underwent transcardiac perfusion.

All spinal cord axial sections from C2 to T4 were imaged with a Zeiss Axio Imager fluorescence microscope (Carl Zeiss Microimaging GmbH, Germany). In axial sections that had undergone immunohistochemistry, arterioles were positive for RECA-1 and SMA, whereas venules and capillaries were labelled by RECA-1 only (vessels that had a luminal diameter < 6.5 μm were classified as capillaries). Further delineation of vascular structures and the central canal was undertaken with confocal microscopy (LSM 880, Carl Zeiss Microimaging GmbH, Germany).

Physiological vital statistics were compared with two tailed Student’s t-test and values expressed as mean ± standard deviation (SD). Where there were three groups, one-way analysis of variance (ANOVA) was employed and adjusted for multiple comparisons with Tukey post-hoc tests. Fluorescence intensities (integrated densities and mean pixel intensities) were compared using two-way ANOVA and adjusted for multiple comparison using Bonferroni’s post-hoc tests. A p value < 0.05 was considered statistically significant in all analyses. All fluorescence values were expressed as mean ± standard error of the mean (SEM). GraphPad Prism (v7.02, GraphPad Software Inc, California) was used to perform all statistical analysis. Graphs were plotted using GraphPad Prism and MATLAB.

Following intracisternal injection of ICG, the distribution of this fluid tracer was qualitatively and quantitatively assessed in real-time in vivo using a NIR camera (Fig. 2, Additional file 1: Video S1, Additional file 2: Video S2). In all rats, ICG tracer migrated caudally along the lateral aspects of the vertebral canal. Over time, increasing fluorescence was detected at the more caudal spinal levels, and the area of highest intensity shifted caudally.

Fluorescence intensities at each spinal level were quantitatively sampled at 150 s intervals after injection (Fig. 2). At every time point analysed after injection of tracer, significantly higher fluorescence intensities were measured in spontaneous breathing rats compared with Control animals on positive pressure ventilation (Fig. 2E). Significant difference at C3 was detected 2.5 min after tracer injection (p = 0.003). When hypertensive rats were compared to controls, no statistically significant difference in fluorescence intensity was detected during the 20 min period after intracisternal ICG injection (Fig. 2F). Similarly, the only difference in fluorescence signal intensities between Control and tachycardic rats was observed at 2.5 min (Fig. 2G).

A pulse-like wavefront of tracer fluorescence was detected and quantified in the upper cervical levels. The velocity of this wavefront was significantly higher in spontaneous breathing rats compared with the Control group (Fig. 2). In spontaneously breathing and positive pressure Control, the mean values were 0.9 and 0.4 pixels/s (95% CI 0.04–1.0, p = 0.04 two-tailed t-test), respectively. No significant difference was observed between Control and hypertensive or tachycardic cohorts. In these NIR experiments, we have demonstrated that spontaneous respiration, but not arterial pulsations, is associated with greater displacement of injected CSF tracer.

To further determine the effects of each of these altered physiological parameters on fluid exchange at the microscopic level, intracisternally injected fluorescent particles were tracked in real time in vivo at the junction of the subarachnoid and/or perivascular space. A spinal two-photon intravital imaging technique was developed to characterise the motion of injected microspheres, in relation to pial blood vessels, between T2 and T5. Particle movement was dynamic, with displacement of the 1 µm microspheres appearing comparatively greater in spontaneous breathing animals compared to controls (Additional file 3: Video S3, Additional file 4: Video S4), which may account for the higher pulse wave velocities noted above. Hypertension resulted in greater back and forth movement, but it did not appear that this corresponded to an increase in overall displacement of particles (Additional file 5: Video S5). This may account for the relatively reduced CSF velocities calculated by NIR imaging. In controls, spontaneous breathing, hypertensive, and tachycardic animals, microspheres moved in an oscillatory pattern (Fig. 3). Interestingly, there was no evidence microspheres were compelled to oscillate within the paravascular spaces as noted by others in the brain [15]. This may be due to the relatively large particle sizes or that this phenomenon only occurs deeper in the interstitium.

Therefore, spontaneous respiration was associated with greater cervicothoracic CSF tracer movement than positive pressure mechanical ventilation. Increasing the MAP or heart rate resulted in little effect on subarachnoid space molecular transport. The emerging in vivo data suggests greater displacement and mixing may be responsible for this finding.

In anaesthetised, spontaneously breathing rats, bradypnea of 50–55 breaths/min was observed, with resultant CO₂ retention and respiratory acidosis (pCO₂ 60 ± 6.8 mmHg, pH 7.3 ± 0.05). Hypercarbia is widely believed to induce vasodilatory effects on arteries in the CNS [23]. To eliminate acidosis as a potential confounding variable in the current study we compared our Control animals ventilated at a low respiratory rate, resulting in respiratory acidosis (pCO₂ 63.8 ± 10.6 mmHg, pH 7.3 ± 0.04), and animals ventilated at a respiratory rate of 66 breaths/min (Normal), which normalized the partial pressure of CO₂ and pH (43.4 ± 4.8, 7.4 ± 0.06, respectively). These two controls were used to assess the effects, if any, of blood pH, partial pressure of CO₂ and rate of ventilation on ISF/CSF fluid dynamics. As arteriolar vasodilation was thought to exert the greatest effect on fluid transport at the level of the perivascular space, this Normal cohort, was established only in ex vivo experiments involving intracisternal injection of AFO-647.

Our in vivo studies showed that exposure to negative intrathoracic pressures during spontaneous breathing facilitated the dispersion of tracers within the spinal CSF compartment but did not indicate whether such effects influenced ISF dynamics. A previous study by our group indicated that tracer injected into the spinal grey and white matter had distinct patterns of distribution. Tracer injected into the grey matter was transported radially, while transport occurred along the axonal fibres in the white matter [16]. Given this, we investigated CSF tracer in the whole spinal cord and separately in the grey and white matter.

Influx of AFO-647 tracer into the spinal parenchyma was measured in axial cross-sections from spinal levels C2–T4. There was significantly higher tracer signal within the whole section of the spinal cord, as well as within constituent grey and white matter compartments, in spontaneous breathing rats compared with the positive pressure ventilated controls (p < 0.0001 for both controls) (Fig. 4B–D, refer to figure for p values). There was significantly greater number of fluorescent deposits (representing tracer associated with blood vessels or “perivascular events”) in the grey matter of spontaneous breathing rats compared with controls (Normal p = 0.0004, Control p = 0.01) (Fig. 5C). Moreover, the proportion of the grey matter area occupied by fluorescence was higher in spontaneous breathing rats (p = 0.04) compared with Normal controls (Fig. 5D). On post hoc analysis, more “perivascular events” were found at C2 and C3 in the spontaneous breathing group compared with controls. Thus, differences in AFO-647 transport were the most pronounced at the upper cervical levels.

The overall AFO-647 fluorescence was significantly higher in hypertensive rats compared with the Control group but lower compared with Normal controls. This was observed within the whole cord and the white matter (Fig. 4E, F). Curiously, within the grey matter there was significantly higher AFO-647 signal in the hypertensive group compared with either controls. This trend reached significance at C2 on post hoc analysis (Fig. 4G). These findings were not borne out in adjunctive analyses (Fig. 5). Compared with hypertensive rats, there were significantly more “perivascular events” (p = 0.04) in the Control group and greater percentage of grey matter occupied by tracer (p = 0.0001) in both control cohorts (Fig. 5E, F).

There was higher tracer signal in tachycardic rats compared to the matched Control animals (p < 0.0001), but not the Normal controls (Fig. 4H, I). Within the grey matter, the overall AFO-647 intensity was significantly higher in tachycardic rats than either control groups (p < 0.0001) (Fig. 4J). Adjunctive analysis of “perivascular events”, and the percentage of grey matter occupied by tracer did not reveal any differences between tachycardic and the control groups (Fig. 5G, H).

Multiple tracer experiments have, thus far, consistently demonstrated that spontaneous breathing enhances CSF tracer deposition in the spinal subarachnoid space as well as transport into the interstitium. The findings from tachycardic and hypertensive rats have been less certain with increasing blood pressure and heart rate driving more CSF tracer into the spinal cord than the matched controls. However, this is offset by increasing respiratory rate and return to normocapnia.

Further insights into spinal fluid transport may be gained from qualitative assessment of AFO-647 influx into the interstitium.

In all animals, AFO-647 deposited on the pial surface and around intramedullary and extramedullary blood vessels. A transpial, “parenchymal”, diffusive pattern of tracer transport was also observed (Fig. 6), but this was largely noted in spontaneous breathing rats. In this cohort, the transpial pattern was observed in 33% of spinal levels compared with just 10% in the Normal cohort, and none in the matched Control arm. In hypertensive and tachycardic rats, this transpial deposition of AFO-647 was found in 14% and 19% (respectively) of spinal levels, similar to that detected in the Normal group (Table 1). Note that tracer colocalisation to anatomical structures was noted as either present or absent (after ensuring uniform adjustment of brightness and contrast in every photomicrograph). Further quantitative or statistical analyses were not performed.

Labelling of large extramedullary vessels such as the anterior spinal artery was ubiquitous. In both the white and grey matter, tracer selectively accumulated around arterioles and venules, especially around radially projecting blood vessels (Fig. 6A–D). Pericapillary tracer was detected but was not as abundant as periarterial or perivenular deposition. In the positive pressure ventilated control groups, Normal and Control, tracer was identified around white matter arterioles in 76% and 39% of spinal levels respectively. In spontaneous breathing, hypertensive, and tachycardic rats this was observed in 86%, 62% and 76% of spinal levels, respectively. In the grey matter, periarteriolar tracer fluorescence was detected in 61% and 62% of spinal levels in Normal and Control cohorts, respectively. In spontaneously breathing animals, this phenomenon was observed more frequently (71%), while in both paced and hypertensive rats, this was observed less commonly (33% each).

In both extramedullary and intramedullary arteries and arterioles, confocal microscopy confirmed perivascular deposition of AFO-647 (Fig. 7A, D–G). Distinct layers of tracer were detected immediately external, within, as well as immediately internal to the smooth muscle cells of the tunica media of the anterior spinal artery. This was also observed in the arterioles of the vasocorona and within the grey matter. Tracer deposited immediately around the RECA-1 labelled endothelium of intramedullary venules and capillaries, as well as the extramedullary veins of the ventral median sulcus and the superficial pial network (Fig. 7H, I). Furthermore, at higher magnifications, deposition of AFO-647 in the spinal cord extracellular space highlighted the tortuous nature of the interstitial microarchitecture (Fig. 7I).

Tracer deposited around the central canal, but the spinal levels over which this was found were often non-contiguous. AFO-647 deposition around the central canal occurred in 47%, 17% and 8% of total spinal levels in spontaneous breathing, Normal, and Control cohorts, respectively. Hypertensive rats exhibited a similar incidence (19%), while tachycardic rats had increased tracer deposition around the central canal (75%) compared with the control groups. On confocal microscopy, tracer distributed heterogeneously within the ependymal layer of the central canal. There was preferential accumulation along the luminal border of ependymal cells, with less fluorescence along the abluminal aspect (Fig. 7C). On higher magnification, there was speckled clumping of tracer on the luminal aspect (Fig. 7C). It was not possible to confirm whether there was tracer uptake by ependymal cells. Occasionally, intraluminal tracer was detected. Serpiginous trails of tracer were also observed between the abluminal aspect of the ependymal layer and nearby subependymal microvessels (Fig. 7B). This suggested a pathway linking subependymal perivascular spaces and the central canal.

It has long been held that cardiac pulsations drive CSF flow in the cranial subarachnoid space [7, 24‐27] and cardiac-gating imaging methods reinforced this belief. There is evidence from small historical investigations that respiratory pressures have a significant effect on spinal CSF [28]. A number of investigators posited that lumbar CSF pressure waves originate from spinal arterial pulsations with some contribution from the cranial vasculature (such as the choroid plexus) [29‐31]. This is supported by at least one canine study involving spinal blocks and aortic occlusions [29]. Most of the data about physiological drivers of CSF flow have been derived from human MRI brain studies, and the role of respiration in spinal CSF dynamics is not clear. In these studies, the dural venous sinuses transmit pressure changes from the thorax and abdomen to the cranial subarachnoid space [32]. Recently, the Dreha-Kulaczewski group [33‐35] demonstrated prominent rostral movements of CSF throughout the spinal canal at the beginning of forced inhalation in human subjects using real-time PC-MRI. Inspiration-induced CSF waves were more pronounced than those associated with cardiac pulsations, which were low amplitude. However, forced expiration resulted in caudal flow from T6 and very low flows rostral to T1–4. The authors hypothesised that a “watershed” point at the level of the heart divided the spinal subarachnoid space into two compartments. This in turn reflects the intimacy between intrathoracic and intraabdominal pressures, and the pressures within epidural venous plexus that are thought to modulate CSF dynamics. A recent study has emphasized this interaction between thoracic and abdominal pressure, reporting that it is the combination and balance of these pressures that govern spinal CSF movements [14].

The findings in our in vivo studies support previous MRI and computational experiments. We posit that the cyclical negative and positive pressure generated by the thoracic and abdominal cavities are transmitted directly to the continuous epidural venous–azygos venous system. The pliant thecal sac is, in turn, directly exposed to this venous network. Deformation of the dura drives CSF movement. Moreover, there is in silico evidence that pulsations in the subarachnoid space are transmitted to the microvasculature [7]. If negative intrathoracic pressures are eliminated, either increased positive pressure drives CSF cranially, or the movement of fluid even at the perivascular level is dampened. This was observed in our NIR and intravital experiments.

We described the first deployment of ICG and the IR800 NIR function on a commercial operating microscope to assess tracer movement in the spinal CSF in vivo in an animal model. This technique established that changes in intrathoracic pressures, not cardiac pulsations, affect the displacement and velocity of CSF redistribution. The intravital two-photon excitation microscopy demonstrated the dynamic nature of fluid movement in the spinal subarachnoid space. The ex vivo experiments confirmed that the dynamic flow of CSF observed in the subarachnoid space of spontaneously breathing animals correlated with quantitatively greater tracer penetration to the deep interstitium. Evidence from all three modalities taken together, suggest that the increased interstitial inflow of tracer is driven by greater displacement and mixing creating greater exchange of solute and fluid between the subarachnoid space and spinal cord, through the pial surface and at perivascular spaces.

In the current study, we attempted to individually manipulate respiration, blood pressure, and heart rate. However, there is increasing evidence that artificial ventilation exerts complex effects on cardiac pre- and afterload, analogous to the administration of vasopressors [36]. Moreover, central venous pressure may be sensitive to ventilatory parameters such as tidal volume [37]. Therefore, it is important to be cognizant that cardiorespiratory forces are inextricably linked, and it may be difficult to examine such physiological drivers in complete isolation. Future studies will need to consider central venous pressure measurements.

Mestre et al. [15] employed two-photon intravital microscopy and particle tracking velocimetry on fluorescent microspheres injected into the cisterna magna of mice, to derive pial surface perivascular CSF flow velocities. By synchronising the electrocardiogram and respiratory waves to the velocity of the tracked microspheres, they demonstrated that the microsphere motion was yoked to the cardiovascular pulsations, respiration had less effect on CSF flow. They thus concluded that the displacement of the arterial walls acted as a “perivascular pump” driving CSF in the same direction as arterial blood flow. It has since been suggested that the arterial wall pulsations reported would be too small to drive net flow of CSF [38]. The authors then acutely raised the MAP of the mice with Angiotensin II and found that the particle velocity reduced, concluding that the increased arterial stiffness, induced by hypertension, diminished the efficacy of the “pump”. It should be noted however, that the authors did not evaluate the effect of cardiac-pulsation-related CSF changes in the subarachnoid space. Moreover, the blood vessels studied were not proven to be penetrating arterioles, but may have been pial arterioles, which likely confounds their results. Bedussi et al. [39] performed intravital confocal microscopy to track CSF infused microspheres on the surface of mouse brains. They also found that paravascular CSF flow is driven primarily by cardiac pulsations. They found no microspheres below the subarachnoid space, but unlike Mestre et al.’s findings, particles were present around both arteries and veins, albeit to a lesser degree. This group’s interpretation of the pulsatile nature of the bead movement was that arterial pulsations generated mixing in the penetrating vessel perivascular space. In a mathematical model, Bilston et al. [40] showed that a mismatch in relative timing in arrival of a CSF pressure wave in the spinal subarachnoid space and an arterial pulse could theoretically drive perivascular fluid inflow along a small artery entering at a right angle to the spinal cord. This has been subsequently shown to be a feasible mechanism in the spinal cord [41] but not the brain [42] due to differences in the cranial and spinal pressure pulse shapes.

Influx of tracer into the spinal interstitium involves both transpial as well as perivascular routes. Spontaneous breathing, but not hypertension and tachycardia, was strongly associated with transpial transport of tracer from the subarachnoid space into the spinal cord. Transpial migration could account for the significantly higher fluorescence signal in the white matter where there is a paucity of arterioles. In spontaneously breathing rats, microspheres in the CSF had larger amplitude oscillations and greater displacement (demonstrated in intravital studies, see Fig. 3, Additional file 3: Video S3, Additional file 4: Video S4, Additional file 5: Video S5, Additional file 6: Video S6). This supported the corresponding in vivo NIR studies that found that the tracer wavefront moved at a higher velocity in the cervical subarachnoid space. This increased displacement of tracers likely corresponds to greater mixing in the subarachnoid space and a greater ability to penetrate the pial boundary. The increase in overall fluorescence and “perivacular events” in the grey matter also suggests that the intrathoracic pressure fluxes that occur in spontaneous breathing are responsible for driving fluid along perivascular spaces, possibly via increased mixing on the surface of the spinal cord (Fig. 8).

In our intravital experiments, hypertension induced greater pulsatile “to and fro” movement but reduced net flow compared to spontaneously breathing animals. This correlated with the slower cranio-caudal movement of tracer in the spinal subarachnoid space with NIR imaging. A similar result was evident in tachycardic animals. The ex vivo studies indicated that this reduced mixing in the subarachnoid space translates to a reduction in transpial deposition of tracer (Fig. 4F). Interestingly, although in both hypertensive and tachycardic rats there was increased tracer fluorescence in the grey matter (Fig. 4G, J), this was not the result of an increase in “perivascular events” (Fig. 5E, G). It is possible that elevating the pulse rate and pressure does not result in an overall increase in perivascular tracer transport, but more tracer from the perivascular space is driven into the surrounding extracellular space. This would account for the increased fluorescence observed within the grey matter, as it is highly vascularized, supplied by the central branch of the anterior spinal artery. Indeed, tracer signal was relatively reduced within the white matter, which has comparatively fewer penetrating vessels. Although blood pressure and tachycardia appear to be involved in fluid transport, there is still inconsistent evidence from the literature and our experiments on the precise effects of hypertension and tachycardia on CNS fluid influx.

In ex vivo samples, AFO-647 selectively coalesced around blood vessels of all types. This confirmed the perivascular space as a privileged pathway for fluid and solute transport. Lam et al. [43] observed deposition of intracisternally injected AFO-647 tracer around spinal intramedullary arterioles, venules, and capillaries. Wei et al. [44] also observed preferential accumulation of cadaverine tracer around numerous blood vessels but did not clarify whether venules were present. Our findings do not support the proposed routes of fluid exchange in the popular glymphatics theory, wherein fluid and solute gain peri-arterial access to the CNS, flow through the extracellular space to collect around major draining veins and exit the interstitium via peri-venular pathways [17, 45, 46].