Apolipoprotein M-bound sphingosine-1-phosphate regulates blood–brain barrier paracellular permeability and transcytosis
Abstract
The blood-brain barrier (BBB) is formed by the endothelial cells lining cerebral microvessels, but how blood-borne signaling molecules influence permeability is incompletely understood. We here examined how the apolipoprotein M (apoM)-bound sphingosine 1–phosphate (S1P) signaling pathway affects the BBB in different categories of cerebral microvessels using ApoM deficient mice (Apom-/-). We used two-photon microscopy to monitor BBB permeability of sodium fluorescein (376 Da), Alexa Fluor (643 Da), and fluorescent albumin (45 kDA). We show that BBB permeability to small molecules increases in Apom-/- mice. Vesicle-mediated transfer of albumin in arterioles increased 3 to 10-fold in Apom-/- mice, whereas transcytosis in capillaries and venules remained unchanged. The S1P receptor 1 agonist SEW2871 rapidly normalized paracellular BBB permeability in Apom-/- mice, and inhibited transcytosis in penetrating arterioles, but not in pial arterioles. Thus, apoM-bound S1P maintains low paracellular BBB permeability in all cerebral microvessels and low levels of vesicle-mediated transport in penetrating arterioles.
Introduction
The blood-brain barrier (BBB) allows only certain molecules and cells to pass from the blood circulation to the brain. Molecules cross the BBB via the paracellular route by diffusion between endothelial cells (Tietz and Engelhardt, 2015), or by transcellular route, including vesicular transport across the endothelium (De Bock et al., 2016). The selectivity of the BBB towards small molecules is maintained by tight and adherens junctions between adjoining endothelial cells (Tietz and Engelhardt, 2015), whereas the transport of large molecules is restricted by receptor-mediated interactions (De Bock et al., 2016). Under normal conditions, the BBB maintains low permeability due to the structural barrier maintained by junctional complexes and the low rate of transcytosis (De Bock et al., 2016; Daneman and Prat, 2015; Nico and Ribatti, 2012; Abbott et al., 2010). In disease states, the loss of BBB function leads to increased permeability, which is viewed mainly as the consequence of disrupted junctional complexes at endothelial cell contact sites (Davies, 2002; Engelhardt et al., 2014; Salameh et al., 2016). However, recent evidence suggests equal importance of transcytosis, as a failure in the homeostatic regulation of endothelial trans-cellular transport may aggravate brain pathologies (Knowland et al., 2014; Habgood et al., 2007; Hashizume and Black, 2002; Krueger et al., 2013).
Sphingosine-1-phosphate (S1P) is a bioactive sphingolipid synthesized by all mammalian cells, with blood-circulating S1P produced by vascular endothelial cells, platelets, and erythrocytes (Thuy et al., 2014; Blaho and Hla, 2014; Xiong and Hla, 2014; Prager et al., 2015). S1P is a ligand for five membrane-bound G-protein-coupled receptors (S1PR1-5), and the brain endothelium expresses S1PR1-3 (Prager et al., 2015). The majority (~70%) of S1P in the blood circulation is transported by apolipoprotein M (apoM), a 26 kDa protein mainly associated with high-density lipoprotein (HDL) (Xu and Dahlbäck, 1999), while the remaining S1P fraction is carried primarily by albumin (Ruiz et al., 2017; Christoffersen et al., 2011).
Depending on the carrier (i.e., apoM or albumin), S1P preferentially activates S1PR1 or S1PR2 and S1PR3, which can exert opposite effects (Murata et al., 2000). For example, S1PR1 stimulation inhibits neuroinflammation, whereas S1PR2 and S1PR3 stimulation facilitates pro-inflammatory signaling cascades (Sanchez et al., 2007; Sanchez, 2016; Blaho et al., 2015). The onset and severity of sepsis in humans correlate with reduced apoM-S1P plasma concentrations and loss of BBB integrity (Davies, 2002; Kumaraswamy et al., 2012). In mice, lack of apoM compromises the endothelial barrier in the lungs and brown adipose tissue (Christoffersen et al., 2011; Christensen et al., 2016; Christoffersen et al., 2018), but its effects on the BBB are largely unexplored. First, apoM-deficient mice are less protected against neuroinflammation (Blaho et al., 2015). Second, conditional knock-out of S1PR1 increases the BBB permeability towards small molecules (Yanagida et al., 2017), but no data are available on how the BBB properties change with consistent S1PR1 expression, but during reduced apoM levels.
Our objective was to determine the modulatory role of apoM for the BBB. We used two-photon microscopy to image the brains of apoM-deficient (Apom-/-) mice in vivo in order to characterize changes in BBB permeability in different categories of cerebral microvessels. Lack of apoM increased the BBB permeability to small molecules without affecting the ultrastructural components of junctional complexes and caused substantial increases in adsorptive transcytosis in pial and penetrating arterioles, but not in venules and capillaries. Systemic administration of a selective S1PR1 agonist SEW2871 promptly reversed the increased permeability for the paracellular route, whereas the normal rate of transcytosis was restored only in penetrating arterioles. We suggest that modulation of the apoM signaling axis may be clinically relevant, but heterogeneous responses of distinct vessel types to apoM shortage and S1PR1 agonist treatment should be taken into account in the development of protective strategies in neurological conditions.
Results
Deficiency in apoM increases BBB permeability
To assess the BBB with all structural constituents we used two-photon fluorescence imaging in vivo in transgenic apoM-deficient mice (Apom-/-) mice (Christoffersen et al., 2011), which in contrast to other animal models, e.g. S1pr1iECKO (S1pr1 knock-out) mice (Yanagida et al., 2017), retain expression and function of S1PR1 (Christoffersen et al., 2011). Following preparative surgery (Figure 1a), the brain was imaged in living anesthetized animals through an acute craniotomy over the somatosensory cortex (Figure 1b).

Two-photon imaging of paracellular permeability changes in Apom-/- mice.
(a) Schematic of a mouse after preparative microsurgery for two-photon imaging, with (b) craniotomy over the somatosensory cortex. (c) Three-dimensional reconstruction of the brain microvasculature from a single Z-stack after i.v. injection of NaFluo. Each Z-stack corresponds to a single time point and underwent dimensionality reduction by maximum intensity projection along the Z-axis (depth). (d) Relative fluorescence change over the baseline (first projected Z-stack in the recording). Contours delineate the main cerebral microvasculature. Compared to WT, the Apom-/- animals had increased accumulation of NaFluo in the brain parenchyma over time, which was ameliorated by treatment with S1PR1 agonist SEW2871. (e) Bolus injections of fluorophore were separated by 30 min time-lapse imaging intervals. The panels show fluorescent molecule accumulation in the brain parenchyma for all tested fluorophores at 30 min post-injection. (f) Changes in fluorescence intensity over time in the brain parenchyma and the blood circulation expressed as a fraction of the baseline. (g) Quantitative assessment of the total effect 30 min post-injection by area under the curve (AUC) calculation. Left bar plot: increased paracellular permeability of the BBB in Apom-/- mice towards sodium fluorescein (0.365 kDa) and Alexa Fluor488 (0.643 kDa), but not large molecule FITC-dextran (10 kDa). Right bar plot: no differences in the signal decrease of blood-circulating fluorophores among all experimental groups. (h) The mean arterial blood pressure during imaging was the same in all experimental groups. All data are presented as mean ± SEM. *p<0.05, **p<0.01, ***p<0.001.
First, the integrity of the BBB was assessed after an i.v. bolus injection of sodium fluorescein (NaFluo). NaFluo is a small (0.365 kDa), negatively charged hydrophilic molecule, which determines its preferred route across the BBB via paracellular diffusion, and has been shown to be suitable for characterizing the paracellular leak, independently from vesicular transport across the barrier (Cheng et al., 2010; Hawkins and Egleton, 2006; Kozler and Pokorný, 2003).
The fluorescence signal from blood-circulating fluorophore and brain parenchyma was simultaneously measured for 30 min in a series of Z-stacks of relatively large brain volume (750 μm x 750 μm x 114 μm) that contained pial and penetrating microvessels, and capillaries (Figure 1c). Each Z-stack underwent dimensionality reduction, that is, maximum intensity projection along the Z (depth) axis. The changes in fluorescence in brain parenchyma were not apparent on raw fluorescence data images (Figure 1c), therefore we chose to represent the data as relative increases of fluorescence over the baseline (Figure 1d–e).
The morphology of the vasculature delineated by blood-circulating NaFluo did not differ between WT and Apom-/- mice (Figure 1d). We detected a progressive accumulation of NaFluo in the brain parenchyma in both WT and apoM-deficient mice. However, Apom-/- mice exhibited considerably higher levels of fluorophore accumulation in the brain compared to WT mice already at 15 min after injection, with a clear difference at the 30 min time-point (Figure 1d, Video 1). As apoM is the main carrier for S1P in the blood stream and S1P bound to apoM primarily activates S1PR1 on the endothelial cells (Prager et al., 2015; Murata et al., 2000), we tested whether the increase in BBB permeability due to the lack of apoM was a consequence of the S1PR1 hypostimulation. The Apom-/- mice were i.p. injected 150 min prior to imaging with a selective S1PR1 agonist, SEW2871. Compared to untreated Apom-/-, the SEW2871-treated Apom-/- mice exhibited decreased permeation of NaFluo into the brain (Figure 1d, Video 1), suggesting that the S1PR1 agonist rescued the abnormal Apom-/- BBB phenotype. This result indicates that the apoM deficiency impairs S1PR1 signaling to the degree that causes increased BBB permeability. We next examined whether the BBB also exhibited increased permeability towards other molecules of different sizes and chemical properties. Each animal in the WT, Apom-/-, and Apom-/- SEW2871-treated group was injected at 30 min intervals with fluorophores of decreasing molecular size, starting with 10 kDa FITC-dextran, followed by 0.643 kDa Alexa Fluor 488 and 0.365 kDa NaFluo (Figure 1e). After each injection, we imaged the same brain volume over time, with subsequent dimensionality reduction of the Z-stacks. To compare animal groups, we measured the changes in fluorescence signal over time from ROIs placed in the brain parenchyma and vasculature (n = 7 mice in each group), and accounted for systematic error in measurement of consecutively injected fluorophores (see Methods section, Figure 1—figure supplement 1).
Lack of apoM leads to a reversible increase in BBB paracellular permeability.
Accumulation of NaFluo in brain parenchyma presented as fluorescence increase relative to the baseline. Fluorophore injections occurred 1 min before the first recorded imaging plane.
For both tracer molecules, Alexa Fluor 488 and NaFluo, and in all animal groups, we detected fluorescence increase over time in the parenchyma, and the effect was persistent without the signal reaching saturation (Figure 1f). In contrast, there was no increase in brain fluorescence for FITC-dextran (Figure 1e–f). To better characterize the fluorescence increase over time, we calculated the area under the curve (AUC) of the signal traces, that is, the cumulative fluorescence increase for the duration of 30 min after each fluorophore injection (Figure 1f–g). We observed no differences in BBB permeability between WT and Apom-/- mice for 10 kDa FITC-dextran (WT 0.0086 ± 0.0056 min vs. Apom-/- 0.033 ± 0.023 min, p=0.3068), but the BBB in Apom-/- mice was significantly more permeable to small molecules, such as Alexa Fluor 488 (WT 0.90 ± 0.31 min vs. Apom-/- 2.43 ± 0.61 min, p<0.05) and NaFluo (WT 5.11 ± 1.3 min vs. Apom-/- 27.5 ± 4.9 min, p<0.001) (Figure 1g, Videos 2–3).
WT mice BBB paracellular permeability towards NaFluo, Alexa Fluor 488, and 10 kDa FITC-dx.
Accumulation of fluorophores in brain parenchyma presented as fluorescence increase relative to the baseline. For each time-lapse recording, the respective fluorophore was injected 1 min before the first recorded imaging plane.
Increase in the BBB paracellular permeability in Apom-/- mice towards NaFluo and Alexa Fluor 488, but not 10 kDa FITC-dx.
Accumulation of fluorophores in brain parenchyma presented as fluorescence increase relative to the baseline. For each time-lapse recording, the respective fluorophore was injected 1 min before the first recorded imaging plane.
Treatment of the Apom-/- mice with SEW2871 markedly reduced the leakage of both Alexa Fluor 488 and NaFluo compared to untreated Apom-/- mice (0.75 ± 0.38 vs. 2.43 ± 0.61 min, p<0.05; and 5.45 ± 1.6 min vs. 27.5 ± 4.9 min, p<0.01; Figure 1e–g, Videos 3–4). The reversal of the dysfunctional BBB phenotype occurred relatively fast, and at 150 min post-treatment with SEW2871, the BBB permeability was the same in Apom-/- and WT mice (p=0.6949 and p=0.8778 for Alexa Fluor 488 and NaFluo, respectively; Figure 1g). In addition we performed the same analysis on arbitrary fluorescence units [a.u.], that is non-normalized data (Figure 1—figure supplement 1e–f) and the results were in accord with relative fluorescence increases. We chose to report the data as relative changes for the results to be more easily comparable between different imaging setups.
Apom-/- SEW2781-treated mice exhibit normal BBB paracellular permeability towards NaFluo, Alexa Fluor 488, and 10 kDa FITC-dx.
Accumulation of fluorophores in brain parenchyma presented as fluorescence increase relative to the baseline. For each time-lapse recording, the respective fluorophore was injected 1 min before the first recorded imaging plane.
Importantly, the decrease in fluorescence signal in vessels (i.e., clearance of the respective fluorophore from the blood circulation) did not differ between WT, Apom-/-and Apom-/- SEW2871-treated mice for each respective fluorophore (Figure 1f–g). In addition, we found no differences in mean arterial blood pressure between animal groups before (pre-imaging) or after the administration of all fluorophores (post-imaging) (Figure 1h). Thus, the increased fluorescence accumulation in the brains of Apom-/- mice and reduced accumulation of fluorophores in the brains of SEW2871-treated Apom-/- mice were not caused by different kinetics of a fluorophore clearance from the blood stream or differences in blood pressure.
These results show that apoM shortage increases the BBB permeability towards small molecules (~0.3–0.7 kDa), and that the effect can be reversed by S1PR1 stimulation.
Deficit in apoM signaling does not alter BBB tight junction ultrastructure
Increases in the BBB permeability towards small molecules may be indicative of defective structural elements that restrict paracellular diffusion across the BBB, e.g. junctional complexes (Davies, 2002; Engelhardt et al., 2014; Salameh et al., 2016). Therefore, we next assessed the ultrastructure of the BBB using transmission electron microscopy (TEM).
We analyzed brain microvessels (<6 µm) cross-sections from WT, Apom-/-, and SEW2871-treated Apom-/- mice. In contrast to capillaries, large vessels were susceptible to perfusion-fixation artifacts (non-uniform distortion of endothelium neighboring large tissue-devoid areas, i.e. vessel lumen), which rendered pial and penetrating vessels not suitable for TEM quantitative assessment of the BBB ultrastructure. In each vessel, we measured the thickness of the endothelium, basement membrane (both n = 104 vessels; four mice in each group), tight junctional complexes, and the endothelial cleft, the percentage coverage of endothelial cell contact sites by junctional complexes (all n = 164, four mice in each group), and the endothelial cell area (n = 104 vessels, four mice in each group) (Figure 2, Figure 2—figure supplement 1). The width, length, and coverage of junctional complexes were the same for all three groups of mice (Figure 2a–b, Figure 2—figure supplement 1a), suggesting that the compromised BBB was not caused by a structural defect of junctions, and that the rescue effect of SEW2871 was not mediated by alteration of the morphology of these structural components. However, we detected an ~ 10% increase in the thickness of the endothelium in S1P-signaling-deficient mice compared to controls (Apom-/- 0.078±0.0026 μm vs. WT 0.071 ± 0.0025 µm; p<0.05) and the rescue group (Apom-/-+SEW2871 0.070 ± 0.0025 μm; p<0.05; Figure 2c). Notably, the stimulation of S1PR1 in Apom-/- mice decreased the endothelial cell area (5.80 ± 0.37 vs. 4.24 ± 0.23 μm2; p<0.001), with a strong trend for the cell area in Apom-/- to be larger than in WT mice (5.80 ± 0.37 vs. 4.78 ± 0.24 μm2; p=0.068; Figure 2d). These changes were relatively small, and they are unlikely to disrupt BBB structural integrity to induce robust permeability increase for small molecules. Altogether, our results suggest that the increased permeability observed in Apom-/- mice is not mediated by ultrastructural changes at the level of the junctional complexes.

Small changes in endothelium cell morphology, but not junctional complexes, accompany increased BBB paracellular permeability in Apom-/-mice.
TEM ultrastructure assessment of the structural components of the BBB revealed no changes in the (a) width or (b) length of junctional complexes, with (c) a small increase in endothelial thickness in Apom-/- mice compared to WT mice and a decrease in SEW2871-treated Apom-/- mice. (d) A small decrease in endothelial cell area was observed upon SEW2871 treatment in Apom-/- mice. Data are presented as interquartile distributions with median (horizontal line) and mean (plus sign). The box whiskers indicate 10–90% percentiles. *p<0.05, ***p<0.001.
ApoM/S1PR1 signaling deficit increases transcellular transport
Next, we examined whether lack of apoM leads to increased transcytosis in the BBB. The animals were injected with bovine serum albumin conjugated to Alexa 488 (BSA-Alexa 488) into the blood stream to monitor vesicular transport across the endothelium (Knowland et al., 2014). Following the injections, the animals were continuously imaged for 120 min in a hyperstack mode with subsequent dimensionality reduction (as in Figure 1c). In all animal groups, and as early as 30 min post-injection, we detected BSA-Alexa 488-containing vesicles at the BBB interface (Figure 3). Compared to WT animals, Apom-/- mice exhibited increased albumin in endothelial cells, indicating that vesicular transport was increased in the absence of apoM. Furthermore, the treatment of Apom-/- animals with SEW2871 reversed the observed effect, suggesting that the same apoM-dependent mechanism underlies the previously observed increase in permeability for small fluorescent tracers, as well as elevated transcytosis.

Increase in BSA-Alexa 488 uptake indicates an increase in transcellular transport in Apom-/- mice.
Two-photon assessment of albumin uptake (green puncta) 60 and 120 min post i.v. injection of BSA-Alexa 488. Arrowheads indicate numerous puncta of BSA-Alexa 488 vesicles at the BBB interface in Apom-/- mice but only sparse labeling in WT and SEW2871-treated animals. pa = pial arteriole; pe = penetrating arteriole; cap = capillaries; pv = pial venule; av = ascending venule.
We detected no signs of fibrinogen entry into the brain parenchyma; and similarly to S1PR1 conditional KO mice (Yanagida et al., 2017), there were no differences in morphology of microglia/macrophages and astrocytes between WT and Apom-/- mice, suggesting lack of gliosis upon exposure to albumin (Figure 3—figure supplement 1). However, we detected in the brain parenchyma the presence of cells that sequestered blood-borne albumin (Figure 3). To identify the cells, we performed a series of immunohistochemical analysis on PFA-fixated brains of Apom-/- mice to investigate for the colocalization of cell-internalized BSA-Alexa488 with the immunofluorescence signal from ionized calcium binding adaptor molecule 1 (Iba-1, microglia/macrophages marker); transmembrane protein 119 (TMEM119, microglia-specific marker); RM0029-11H3 (monoclonal antibody against macrophage-specific epitope) and glial fibrillary acidic protein (GFAP, astroglia marker). First, we ascertained that the fixation protocol did not quench the signal from BSA-Alexa488 (Figure 3—figure supplement 2a). The albumin sequestering cells were identified to be macrophages based on the colocalization of BSA-Alexa488 –positive cells with Iba-1 (microglia/macrophages marker), and further with monoclonal anti-macrophage antibody RM0029-11H3 signal (Figure 3—figure supplement 2b). There was no colocalization of the BSA-Alexa488 signal with the microglia-specific marker (TMEM119) and astroglia (GFAP) (Figure 3—figure supplement 2c). We found only minor signs (single puncta) of BSA-Alexa488 uptake by other cell types (Figure 3—figure supplement 2b–c).
S1P signaling deficit elevates transcytosis only in arterioles
Next, we assessed whether the increase in transcellular transport caused by the S1PR1-signaling deficiency was the same for all vessel types. The number of vesicles was quantified over time for each individual vessel type: pial arterioles, penetrating arterioles, capillaries, post-capillary venules, ascending venules, and pial venules (n = 5 mice). The data are expressed as the vesicle surface density, expressed as the number of vesicles per square micrometer of vessel surface area (Figure 4a). In all experimental groups, the distribution of BSA-Alexa 488-positive vesicles was dependent on vessel type and diameter. At 120 min post-injection, the highest vesicle density in WT mice was observed in capillaries and gradually decreased with vessel diameter, regardless of the vessel type (i.e., arterioles or venules) (Figure 4a). In Apom-/- mice, the deficiency in S1PR1 signaling selectively increased transcytosis in pial and penetrating arterioles, but not in venules and capillaries. The effect was most pronounced in penetrating arterioles, where the vascular vesicle density in Apom-/- mice was 10-fold higher than in WT mice (Apom-/- 0.30 ± 0.071 vs. WT 0.038 ± 0.009 vesicles/100 μm2; p<0.01), and three times higher than for pial arterioles (Apom-/- 0.099±0.0023 vs. WT 0.035 ± 0.0011 vesicles/100 μm2; p<0.05; Figure 4b). The level of transcytosis did not change in Apom-/- mice in the capillaries, venules, or pial veins (p=0.2306, p=0.3562, and p=0.9546, respectively).

Heterogeneous susceptibility of different categories of brain microvessels to increased transcytosis in Apom-/- mice.
(a) Scatter plot showing the distribution of the surface vesicle densities of individual vessels plotted against respective vessel diameters 120 min post BSA-Alexa 488 injection. The blue areas represent the morphological division into distinct vessel types. Data from capillaries were mirrored to preserve the order of vessel diameter change in the continuous brain microvascular tree. (b) Quantitative analysis showed the increase in albumin uptake in Apom-/- mice that was present only in the arterial part of the vascular tree and absent in capillaries and venules. Only penetrating arterioles responded to SEW2871 treatment. Numbers in brackets are the number of vessels across five mice in each group. The data are presented as mean ± SEM. (c) Average traces showing the kinetics of the vesicle density changes for WT, Apom-/-, and Apom-/- SEW2871 treated mice with respect to different vessel types. The data are presented as mean ± SEM. *p<0.05, **p<0.01.
Penetrating, but not pial arterioles respond to SEW 2871 treatment
S1PR1 stimulation with SEW2871 led to normalization of the transcellular transport of albumin in the penetrating arterioles of Apom-/- mice to the level of WT controls (Apom-/-+SEW2871 0.68 ± 0.016 vs. WT 0.038 ± 0.009 vesicles/100 μm2, p=0.1156; Figure 4b). In contrast, SEW2871 treatment did not decrease transcytosis in pial arterioles (Apom-/-+SEW2871 0.84 ± 0.017 vs. Apom-/- 0.99±0.023 vesicles/100 μm2; p=0.4879), and albumin transport was still significantly elevated compared to WT mice (Apom-/-+SEW2871 0.84 ± 0.017 vs. WT 0.35 ± 0.011 vesicles/100 μm2; p<0.05). Lastly, we assessed the kinetics of the vesicular density increase over time with respect to each animal group and vessel type. In all vessel types and animal experimental groups, the increase in BSA-Alexa 488 vesicle surface density was persistent and did not reach saturation at 120 min, indicating that the results were not obscured by overloading the transcellular transport with BSA-Alexa 488 (Figure 4c).
These findings show that transcellular transport is upregulated in apoM-deficient mice, and that there is not only heterogeneity in the vasculature susceptibility towards increased permeability, but also towards which vessels can be rescued by stimulation of the S1PR1 signaling pathway.
Discussion
Understanding the influence of blood-borne signaling molecules on brain endothelial cells is critical for an understanding of the BBB function. We report that lack of apoM increased the BBB permeability for small molecules without changes in size, structure, magnitude, or subcellular localization of tight junction complexes. Furthermore, we report an up to 10–fold increase in adsorptive transcytosis in the endothelial cells of pial and penetrating arterioles, but not in venules and capillaries. Thus, ApoM modulates the flux of small and large molecules to different extents in different vascular segments. Finally, the increased BBB permeability for small molecules and transcytosis-mediated large proteins could be rescued by selective stimulation of the S1PR1. So far, the two-photon assessments of the BBB in vivo are scarce. The presented data is first to show the heterogeneity of the BBB responses to S1PR1 impairment and treatment in different vessel types, thus underlines the necessity of studying the BBB in a living brain with respect to different functional and anatomical regions of the microvascular tree. This may be of particular importance in clinical research as S1PR1 modulation is currently used in FDA-approved therapies in humans (Brinkmann et al., 2010).
From among the exogenous tracers typically used in BBB paracellular permeability studies, we chose three molecules with different sizes, NaFluo and Alexa Fluor 488 that readily diffuse via the BBB, and FITC-dextran that has low BBB permeability (Shi et al., 2014; Kutuzov et al., 2018).
The loss of BBB integrity in Apom-/- mice was observed for small tracer molecules < 0.7 kDa, but no significant leak was observed for 10 kDa dextran, which is in accordance with previous findings in S1PR1–knock-out mice (Yanagida et al., 2017). However, in contrast to Yanagida and colleagues (Yanagida et al., 2017), we show that it is not the complete absence of brain endothelial S1PR1 signaling, but a lack of a specific fraction of S1P circulating in the blood (i.e., apoM-bound S1P) is sufficient to impair the BBB, and occurs despite the presence of active S1PR1 (Christoffersen et al., 2011). Although we cannot exclude that some NaFluo molecules were entrapped in transport vesicles during their formation at the luminal side of the endothelium, in our experiments we did not detect any vesicular fraction of NaFluo, Alexa Fluor 488 and FITC-dx at the BBB. In contrast, albumin-conjugated Alexa 488 showed clear vesicular pattern, which is consistent with transcytosis (De Bock et al., 2016; Minshall et al., 2000; Minshall et al., 2002). Furthermore, NaFluo is a strong target for endothelium efflux pumps, that is multidrug resistance proteins (MRP) that sequester the tracer back to the blood stream (Hawkins et al., 2007). Therefore, we consider that the incidental transcytosis of NaFluo, Alexa Fluor 488 or FITC-dx had negligible contribution to fluorescence increases, and the effect observed in the brain parenchyma occurred due to the increase in paracellular permeability.
The increased permeability along the paracellular pathway occurs most commonly in response to severe damage (Farrell and Shivers, 1984; Houthoff et al., 1982; Lossinsky et al., 1995) and is typically reported to be secondary to the ultrastructural abnormalities of junctional complexes, such as truncated morphology during acute BBB breakdown, for example in stroke, sepsis, and brain tumors (Davies, 2002; Engelhardt et al., 2014; Knowland et al., 2014); and in diabetes (Salameh et al., 2016). S1P has been previously shown as an endothelial barrier-enhancing molecule in the lungs (Schaphorst et al., 2003; Dudek and Garcia, 2001; Garcia et al., 2001) and recently, in the brain (Yanagida et al., 2017; Swendeman et al., 2017). Our data ties apoM shortage-induced increase in BBB permeability with S1PR1. The mechanisms of S1PR1-mediated protective effect are not yet fully understood, but may be linked with Rho superfamily of small GTPases. Stimulation of S1PR1 and S1PR2 activates preferentially Rac and Rho, respectively, whilst S1PR3 stimulation activates both GTPases (Xiong and Hla, 2014; Windh et al., 1999). Rac and Rho modulate cytoskeleton dynamics in various cell types (Burridge and Wennerberg, 2004), including the formation and maintenance of a cortical actin ring that stabilizes intercellular junctions between adjoining endothelial cells (Garcia et al., 2001; McVerry and Garcia, 2004). At physiological concentrations, S1P activates preferentially Rac via S1PR1 over Rho (Garcia et al., 2001; Shikata et al., 2003; Hla, 2003), which causes a redistribution of proteins that facilitate cortical actin ring assembly, and consequently tightening of the barrier (Garcia et al., 2001; Mehta et al., 2005). Inhibition of Rac downstream of S1PR1 leads to actin depolymerization, formation of F-actin stress fibers and increased permeability (Wójciak-Stothard et al., 2001; Singleton et al., 2005), and S1PR2 causes Rho activation that inhibits Rac (Dudek and Garcia, 2001; Hla, 2003). However, the formation and disassembly of actin cortical ring affecting BBB permeability should not be simply viewed as Rac vs. Rho antagonism, as inhibition of Rho, similarly to Rac, can reportedly also increase barrier permeability, although without interrupting the actin ring assembly (Garcia et al., 2001; Etienne-Manneville and Hall, 2002). Our quantitative TEM assessment shows that apoM deficiency did not induce changes in junctional complexes ultrastructure. This suggests that the increase of the BBB permeability in Apom-/- mice was either uncoupled from cortical actin ring remodeling, or that during S1PR1 hypostimulation and ongoing S1PR2-mediated Rho inhibition of Rac, the activity of Rac was still sufficient to maintain the elements of scaffolding complex of cortical actin ring that stabilize junctional complexes.
The small (~10%) increase in endothelium thickness in Apom-/- mice may be due to the impaired S1PR1 -dependent formation and the maintenance of the cortical actin ring; and formation of actin stress fibers facilitated by S1PR2 activation (Garcia et al., 2001; Mehta et al., 2005; Wójciak-Stothard et al., 2001; Singleton et al., 2005). Indirectly, BBB permeability changes may be due to osmotic imbalance and cell swelling, but it is uncertain whether this occurs in the brains of Apom-/- mice. Lack of S1PR1 activation aggravates edema in acute brain and lung injury (Ito, 2019), but here the system may not be challenged enough to induce osmotic swelling, as Apom-/- mice have no signs of lung edema and exhibit normal microvasculature morphology (Christoffersen et al., 2011). Other processes may also influence endothelium thickness, but their relevance in Apom-/- mice is yet to be established. For example, S1PR1 activity counteracts vascular endothelial growth factor (VEGF) -mediated endothelium contraction in angiogenesis (Gaengel et al., 2012), but whether S1PR1 hypostimulation affects VEGF signaling in the adult brain in Apom-/- mice is unknown. In addition, impaired S1PR1 signaling activates neuroinflammatory factors (Blaho and Hla, 2014; Blaho et al., 2015), and for example TNF-alpha has been shown to alter cytoskeleton and hyperplastic properties of endothelial cells (Kang et al., 2008; Stroka and Aranda-Espinoza, 2011). Thus, it cannot be excluded, that the changes in endothelium morphology result from both: direct effect of S1PR1 hypostimulation on endothelial cell, and indirect effects linked with impaired BBB-induced loss of brain homeostasis.
Notably, a significant increase in paracellular permeability without apparent changes in junctional complex morphology has been observed by others following osmotic shock (Farrell and Shivers, 1984) and in claudin-5-deficient mice (Nitta et al., 2003). The loss of BBB integrity in Apom-/- mice appeared to be to a lesser extent than in acute pathology (e.g., traumatic brain injury or ischemic stroke), as the BBB permeability towards 10 kDa dextran did not increase in Apom-/- mice (Habgood et al., 2007; Özen et al., 2018). Yet, even with is morphology intact, the BBB with a persistent increase in permeability may still contribute to brain pathology (Carvey et al., 2009).
Our in vivo results characterize apoM as a blood-borne molecule that can affect the BBB permeability via S1PR1. In inflammatory conditions, elevated concentrations of blood-circulating thrombin activate protease-activated receptor-1 (PAR1), and subsequently barrier-disruptive metalloproteinases (Bogatcheva et al., 2002; Rempe et al., 2016). On the contrary, low thrombin concentrations activate protein C (APC), a substrate for endothelial cell protein C receptor (EPCR), which alters PAR1 signaling to promote the synthesis of S1P via Sphk-1, and facilitates barrier enhancement (Riewald et al., 2002; Feistritzer and Riewald, 2005). In addition, APC has been shown to also induce S1PR1 phosphorylation, which further promotes the S1PR1 signaling branch in the endothelium (Lee et al., 2001). In line with our results in Apom-/- mice, the barrier enhancement by blood-borne thrombin occurred via S1PR1, as silencing of S1PR1 but not e.g. S1PR3 abolished the APC-mediated effect (Feistritzer and Riewald, 2005; Finigan et al., 2005). Except for apoM and thrombin, S1P signaling can be modulated by hyaluronic acid (HA). Depending on the form, i.e. high-molecular weight-HA (HMW-HA) or low-molecular weight (LMW-HA), the complex promotes interaction of a cell-surface glycoprotein CD44 isoforms with either S1PR1 or S1PR1/3, respectively (Singleton et al., 2006). This leads to enhancement of the barrier function via S1PR1 or an increase in BBB permeability via S1PR3 due to transactivation (i.e. phosphorylation) of respective receptors by CD44 isoforms, independently of S1P levels (Singleton et al., 2006). This further corroborates our findings on the regulation of S1P signaling by blood circulating molecules via preferential activation of specific S1P receptors.
However, the analogies should be made with caution, as the mechanistic insights come primarily from in vitro barrier models or non-brain tissue, where endothelial barrier properties may differ from the BBB in vivo. For example, Rac and Rho are known to elicit cell-type and subcellular localization-specific effects (Burridge and Wennerberg, 2004; McVerry and Garcia, 2004; Donati and Bruni, 2006), and their intracellular pathways are cross-linked with numerous regulatory feedback loops, including for example Rho inhibiting Rac and vice versa (Burridge and Wennerberg, 2004; Sander et al., 1999). This adds to the complexity of interpretation of Apom knock-out effects on the BBB and is currently an active area in research.
The hypostimulation of S1PR1 in Apom-/- mice is likely to shift the signaling equilibrium towards cellular pathways that promote activation of Rho (via S1PR2) over Rac (via S1PR1), and the effects of increased BBB permeability are likely to be the net outcome of the failure in homeostatic interplay between receptor-dependent branches of S1P signaling, rather than simply a decrease in Rac activity. Consequently, the observed impairment of the BBB cannot simply be linked to a reduction in the amount of circulating S1P, but more specifically to the lack of apoM-bound S1P in favor of albumin-bound S1P in the blood. This notion is further supported by the result of the selective stimulation of S1PR1, which restored the BBB permeability in Apom-/- mice to the level of healthy controls.
Given that inhibition of S1PR2 tightens the endothelial barrier in lungs by facilitating cortical actin ring assembly and stabilization of tight junctions (Sanchez et al., 2007), and in the brain supports the BBB functional integrity (Kim et al., 2015; Cao et al., 2019), it is plausible that S1PR2 stimulation may alleviate BBB phenotype in Apom-/- mice. It should be noted, however, that S1PR1 expression in the endothelium is higher in the arterial branch of cerebral vasculature compared to venules, whereas S1PR2 expression is higher in venules than in arterioles (Vanlandewijck et al., 2018). It may lead, similar as in SEW2871 treatment, to region-specific effects, where the BBB responses to S1PR2 modulation may differ between distinct types of cerebral microvessels.
Though paracellular permeability restriction is typically viewed as the main functional constituent of the BBB, the low prevalence of transcytosis is equally important for preservation of the brain microenvironment (Knowland et al., 2014; Habgood et al., 2007; Hashizume and Black, 2002; Krueger et al., 2013). Recently, Mfsd2a was identified as a critical protein in the suppression of transcytosis. Similar to Apom-/-, a lack of Mfsd2a leads to an increase in vesicular transport across the endothelium that compromises the BBB without structurally opening the paracellular route (Ben-Zvi et al., 2014) or changing junctional protein composition (Yang et al., 2017). Furthermore, the increase in BBB permeability can occur via both the transcellular and paracellular routes, but at different time points; e.g. the increase in transcytosis rate can precede paracellular leakage in stroke (Knowland et al., 2014), and selective dysregulation of the transcellular pathway can occur while the tight junctions are preserved (Krueger et al., 2013). Thus, the intact morphology of junctional complexes, as shown here, is not necessarily indicative of preserved BBB properties.
Noteworthy, the lack of apoM-S1P lead to an increased rate of transcytosis, though only in pial and penetrating arterioles. The susceptibility of the arterial branch of the microvasculature to apoM shortage may be linked to preferential expression of S1PR1 and APC in pial and penetrating arterioles (Vanlandewijck et al., 2018), thus possibly high reliance in these vessel types on S1PR1 signaling in the maintenance of the BBB. It is unclear, however, why penetrating, but not pial arterioles responded to SEW2871, but anatomical differences may be of importance. For example, S1P signaling promotes the formation of the glycocalyx, which limits the diffusion of hydrophilic and negatively charged substances, and supports the barrier function (Zhang et al., 2016). Glycocalyx in pial arterioles is estimated to be thicker compared to the penetrating arterioles (Yoon et al., 2017), but it is unknown how S1PR1 hypostimulation affects distribution and coverage of glycocalyx layer in Apom-/- mice. In addition, the reconstitution of barrier function may have a different time course in pial and penetrating arterioles. Lastly, in contrast to bolus SEW2871 injection, a sustained stimulation of S1PR1 may be needed to restore BBB function in pial arterioles during ongoing apoM deficit.
High-density lipoprotein (HDL) particles are heterogeneous in size and content of apolipoproteins and lipids. In humans, HDL plasma concentrations correlate with reduced susceptibility to cerebro- and cardiovascular diseases (Rader and Tall, 2012; Hovingh et al., 2015; Fellows et al., 2015), and approximately 5% of HDL contains apoM (Xu and Dahlbäck, 1999; Christoffersen et al., 2006). The beneficial effects of HDL can be, in part, explained by HDL association to apoM/S1P. We show that lack of apoM increases the flux of small molecules across the BBB, similar to previous findings investigating the permeability in lung and adipose tissue (Christoffersen et al., 2011; Christensen et al., 2016; Christoffersen et al., 2018). Stressing the system even further, lack of apoM increases the susceptibility to neuroinflammation (Blaho et al., 2015; Galvani et al., 2015). Mainly this has been interpreted as effects caused by blood-circulating apoM/S1P complex. However, a recent study suggested that porcine brain endothelial cells express and secrete apoM into the brain parenchyma (Kober et al., 2017). This has not been reported in other species but could also play a role in the increased permeability observed in the present study. Except S1P, apoM can potentially bind other ligands, for example retinoic acid (Ahnström et al., 2007) or oxidized phospholipids (Elsøe et al., 2012). Lack of apoM would increase free oxidized phospholipid content, which potentially can also affect the BBB (Singleton et al., 2009). However, reversing the BBB phenotype by SEW2871 suggests that the observed effects are primarily mediated via S1PR1 rather than for example oxidized phospholipids. Noteworthy, with the absence of apoM, S1P may still associate to HDL. A recent study showed that in mice lacking expression of both, apoM and albumin, S1P binds to HDL via novel S1P binding protein apoA4 (Obinata et al., 2019). This may potentially constitute a compensatory mechanism in Apom-/- mice, but whether HDL/apoA4/S1P complex also plays a role in maintaining the BBB is unknown and will need further investigations.
Two-photon imaging allows studying the endothelial barrier in the living brain with all its structural and functional constituents, including blood flow, and at the resolution of single microvessels. However, like any imaging approach, it has also limitations that apply to our study. In contrast to transcellular vesicular transport, the fast diffusion of NaFluo in the brain parenchyma made it not feasible to analyze for differences in paracellular permeability increase between distinct vessel types in Apom-/- mice. Regarding the transcellular permeability assessment, we have only investigated the transcytosis of albumin, which represents a fraction of receptor-mediated transcellular pathways. Lastly, the surgical procedures might also affect the brain physiology, however in our previous works, we observed no significant effects of cranial window preparation and imaging on the neurovascular coupling, neuronal and astrocytic calcium transients, electrophysiological activity and cell morphology, suggesting lack of effects on the brain that could significantly affect the BBB function (Kutuzov et al., 2018; Kucharz and Lauritzen, 2018; Hall et al., 2014). Instead of acute craniotomy, further studies may employ chronic window imaging to address changes in BBB permeability in Apom-/- mice over a longer time.
In summary, apoM is critical for maintaining low paracellular and transcellular permeability at the BBB. The BBB is heterogeneous in the susceptibility towards apoM shortage and in the potential for rescue mediated by the S1PR1 signaling pathway. The pronounced increase in transcytosis in Apom-/- mice stresses the importance of considering the development of protective strategies for this pathway to regulate the immediate microenvironment of brain cells and reduce secondary neural damage in both acute and chronic neurological conditions.
Materials and methods
Animals
All animal procedures were approved by the Danish National Ethics Committee and were in appliance with the ARRIVE guidelines. We used 14 to 20 week-old (22–28 g) female transgenic C57bl/6 apoM-deficient mice (Apom-/-) (Christoffersen et al., 2011), and age- and gender-matched wild-type (WT) mice as the control group.
Anesthesia
Request a detailed protocolAnesthesia was induced by a single bolus intraperitoneal (i.p.) xylazine (10 mg/kg body weight [BW]) and ketamine (60 mg/kg BW) injection. During all steps of the surgical preparation, the anesthesia was maintained by i.p. injections of ketamine (30 mg/kg BW) at 25 min intervals. Thirty minutes before imaging, the anesthesia was changed to continuous intravenous (i.v.) infusion of alpha-chloralose (50 µg/hour/g BW).
Animal preparation for two-photon imaging
Request a detailed protocolAnimals were operated on as described previously (Kucharz and Lauritzen, 2018). Briefly, a tracheotomy was performed for mechanical respiration (MiniVent Type 845, Harvard Apparatus) and to monitor exhaled CO2 (MicroCapstar End-tidal CO2 Analyzer, CWE). Two catheters were inserted: the first into the left femoral artery to monitor arterial blood pressure (Pressure Monitor BP-1, World Precision Instruments) and inject compounds, and the second into the femoral vein for infusion of the anesthetic.
Next, a craniotomy was performed. The exposed skull was attached to a custom-made metal head bar with cyanoacrylate gel (Loctite Adhesives) and mounted onto the microscope imaging stage insert. The opening was made using a diamond dental drill (at 7000 rpm) over the somatosensory cortex (4 mm diameter; coordinates: 3 mm lateral, 0.5 mm posterior to bregma). During the procedure, the craniotomy and drill were cooled to room temperature repeatedly with artificial cerebrospinal fluid (aCSF; in mM: NaCl 120, KCl 2.8, Na2HPO4 1, MgCl2 0.876, NaHCO3 22, CaCl2 1.45, glucose 2.55, at 37°C, pH = 7.4) to prevent brain damage from excessive heating of the drill. The dura was removed and a drop of 0.75% agarose (type III-A, Sigma-Aldrich) applied onto the craniotomy. The opening was then secured with an imaging coverslip (Menzel-Gläser) and kept under aCSF to preserve humidity. To ensure physiological O2 and CO2 blood concentrations, prior to imaging, a 50 µl blood sample was collected via the arterial catheter (ABL blood gas analyzer, Radiometer), and the respiration rate and volume adjusted if necessary.
Fluorescent probes
Request a detailed protocolFITC-dextran (MW 10 kDa, 0.5%, 0.05 ml, Sigma-Aldrich), Alexa Fluor 488 (MW 0.643 kDa, 0.025%, 0.04 ml, Invitrogen), and sodium fluorescein (NaFluo, MW 0.365 kDa, 1%, 0.04 ml, Sigma-Aldrich) were each dissolved in sterile saline and administered via the arterial femoral catheter as single bolus injections. Bovine serum albumin (BSA) Alexa Fluor 488-conjugate (BSA-Alexa 488, Invitrogen) was administered via the arterial femoral catheter as a single bolus injection (0.05 ml 1%), together with TRITC-dextran (MW 65 kDa, 0.05 ml 1%, Sigma-Aldrich) for delineation of the vasculature and to ascertain that vessels were not damaged following the cranial window microsurgery that could lead to false-positive results.
Two-photon imaging setup
Request a detailed protocolTwo-photon imaging was performed using an SP5 upright laser scanning microscope (Leica Microsystems) equipped with a MaiTai Ti:Sapphire laser (Millennia Pro, Spectra-Physics) using the 20 × 1.0 NA water-immersion objective. The emitted light was split by a FITC/TRITC filter and collected by separate multi-alkali detectors after 525–560 nm and 560–625 nm bandpass filters (Leica Microsystems).
Monitoring paracellular permeability
Request a detailed protocolThe excitation wavelength for FITC-dextran, Alexa Fluor 488, and NaFluo was 850 nm (14 mWatt output power at the sample). Data were collected as 16-bit image hyperstacks at 512 × 512 pixel resolution (775 μm x 775 μm) in bidirectional scanning mode with a Z-step size of 5 μm (depth span = 114 μm) and 1 min time interval between Z-stacks. The total recording time for each fluorophore was 30 min. The recordings were exported to ImageJ (NIH, version 1.48u4) and Z-projected using maximal signal intensity projection. Regions of interest (ROIs) were chosen throughout the brain parenchyma at all locations deprived of vasculature for the measurement of fluorescent intensity. An increase in fluorescent intensity over the baseline (first projected Z-stack in the time-lapse recording) in the parenchyma indicated crossing of the blood-circulating fluorophore into the brain.
FITC-dextran, Alexa Fluor 488, and NaFluo were consecutively injected in 30 min intervals into the same animal. We did not wait for the fluorophores clearance, as the time of bloodstream clearance differs between fluorophores, and varies between animals. Consequently, this could obscure characterization of the biological effects of SEW2871 on the BBB, as the assessment of permeability in each mouse for each respective tracer would be performed at different time after the drug administration. SEW2871 has a relatively short half-time in the blood plasma with its biological effects peaking at 1 hr, and returning to baseline 4.5 hr after administration (as determined by, for example Akt phosphorylation)(Jo et al., 2005). Instead, we chose to standardize the time of imaging after SEW2871 injection, that is FITC-dx, Alexa Fluor 488, and NaFluo were always imaged 2.5 hr, 3 hr, 3.5 hr after SEW2871 administration, respectively. In our calculations we estimated the contribution of a preceding fluorescence accumulation to a subsequent fluorophore(s) signal measurement. For each mouse, we subtracted the constant autofluorescence background from FITC-dx, Alexa Fluor 488, and NaFluo trace and performed an exponential fit to the FITC-dx trace, and the fit was extrapolated to subsequently imaged Alexa Fluor 488 and NaFluo (Figure 1—figure supplement 1a). We chose the exponential fit to account for the largest systematic error in our experiment, that is the largest theoretical contribution of a preceding tracer to a subsequent tracer signal. The fluorescence is additive, therefore we corrected Alexa Fluor 488 and NaFluo traces by subtracting FITC-dx extrapolated fit values (Figure 1—figure supplement 1a) and effectively, uncoupled FITC-dx kinetics from Alexa Fluor 488 and NaFluo. Next, to the corrected trace of Alexa Fluor 488 we performed an exponential fit and extrapolated it to the time of NaFluo imaging. Similarly as before, we subtracted Alexa Fluor 488 extrapolated fit from NaFluo trace (Figure 1—figure supplement 1b). The same approach, but with exponential decay fit, was used to account for the baseline fluorescence changes in the bloodstream (Figure 1—figure supplement 1c–d).
Monitoring transcellular transport
Request a detailed protocolThe animals were injected intravenously with BSA-Alexa 488. The excitation wavelength was 870 nm (17 mWatt output power at the sample). Data were collected as 16-bit hyperstacks at 2048 × 2048 pixel resolution (386.66 μm x 386.66 μm) in bidirectional scanning mode and triple frame averaging with a Z-step size of 2.50 μm, depth span 144 μm, and 7.5 min interval between Z-stacks for a total recording time of 120 min. The data were exported to ImageJ and Z-projected using maximal signal intensity projection. Based on morphology and location, the vessels were classified as pial arterioles or venules, penetrating arterioles, post-capillary venules, ascending venules, or capillaries. To determine the vesicle surface density, the vessel surface area was calculated based on the diameter and length of the vessel from which the BSA-Alexa 488 signal was counted.
Transmission electron microscopy
Request a detailed protocolThe mice were anesthetized with bolus i.p. injections of xylazine (10 mg/kg BW) and ketamine (60 mg/kg BW), then perfused with Karnovsky fixation (2% glutaraldehyde and 4% paraformaldehyde). Brains were harvested and the cortex cut into 1 mm cubes and left for further fixation in 2% glutaraldehyde in 0.05 M phosphate buffer for 7 days. For transmission electron microscopy (TEM) staining, the samples were washed three times for 15–20 min in 0.12 M cacodylate buffer, followed by post-fixation in 1% osmium tetroxide (OsO4) and 0.05 M (1.5%) potassium ferricyanide (K3FeCn6) in 0.12 M cacodylate buffer for 1 hr at room temperature (RT). Next, the samples were washed three times for 15–20 min in ddH2O and then dehydrated in an ethanol gradient (70%, 96%, and 3 × 15 min in absolute ethanol). Infiltration was performed with propylene oxide twice for 15 min, followed by incubation for 20–40 min in a resin epon/propylene oxide (Bie and Berntsen) gradient (1:3, 1:1, 3:1) and pure resin for 2 hr. Samples were embedded in resin at 60°C for 24 hr. Sections were cut at a thickness of 70 nm (Leica EM UC7), placed on copper grids (Leica), and stained with lead citrate (Leica Ultrostain II, Leica Microsystems) for 3 min and uranyl acetate (Leica Ultrostain I, Leica Microsystems) for 10 min using an automatic contrasting system (Leica EM AC20, Leica Microsystems). Imaging was performed using a CM100 transmission electron microscope (Phillips/FEI, ITEM software) at 9700x – 17500x and 46000-60000x magnification. Two samples from each mouse were imaged, and 20–30 capillaries from each sample were chosen randomly. ImageJ was used for the measurement of BBB morphology.
Immunohistochemistry
Request a detailed protocolThe isoflurane-anesthetized mice were transcardially perfused at the rate of 10 ml/min for 1 min with saline and then for 7 min with ice-cold 4% paraformaldehyde (PFA). The brains were placed in 4% PFA for 24 hr at 4°C for post-fixation, then in 25% sucrose in PBS for ~ 24 hr at 4°C. The brains were frozen in dry ice and cut into a series of 35 μm thick coronal sections (Microm HM450 microtome, Thermo Fisher Scientific).
For fibrinogen staining, we performed antigen retrieval in Tris/EDTA buffer (pH 9.0) for 30 min at 80°C, with a quenching step in 3% H2O2 and 10% methanol in Tris-buffered saline (TBS) for 20 min. The sections were washed 3x in TBS and pre-incubated for 1 hr at the RT with 5% normal goat serum in 0.25% triton-X in TBS (TBS-T). Subsequently, the sections were incubated o/n at RT with anti-fibrinogen primary antibody (1:2000, ab227063, Abcam) in 5% normal goat serum TBS-T solution. Following that, the tissue sections were rinsed 3 × 10 min in TBS-T and incubated in 1% BSA TBS-T with biotinylated secondary goat-anti-rabbit antibody (1:200; Vector Laboratories Inc). Next, the sections were washed 3x in TBS-T for 10 min, then incubated with an avidin-biotin-peroxidase complex solution for 1 hr at RT, and washed again 3 × 10 min with TBS-T. The staining was visualized using 3,3’-diaminobenzidine (DAB) and 0.01% H2O2 according to manufacturer’s instructions. Lastly, the sections were mounted on chromatin-gelatin coated glass slides, stained in 0.1% cresyl violet (CV) solution for 3 min, dehydrated in increasing alcohol solutions (70%, 95%, 100%), cleared in xylene and using a DPX mountant (Sigma-Aldrich), the slides were coverslipped for imaging.
Heat-mediated antigen retrieval was performed on the free-floating brain sections using sodium citrate buffer (pH6) for anti-TMEM119 staining, and EDTA buffer (pH 8.0) at 80°C for 30 min for anti-macrophage staining (RM0029-11H3). The sections were rinsed 3 × 10 min in 0.1 M potassium phosphate-buffered saline (KPBS), then placed for 1 hr in KPBS 0.05% Triton-X (KPBS-T) containing 5% normal goat serum. Next, the sections were incubated o/n at RT with primary antibodies in 5% normal goat serum solution in KPBS-T. The primary antibodies used were: anti-macrophage (1:200; RM0029-11H3; rabbit; Abcam cat# ab56297), recombinant anti-TMEM119 (1:200; rabbit; Abcam cat# ab209064), anti-Iba1 (1:1000; rabbit; Wako cat# 019–19741) and anti-GFAP (1:500; rabbit; DAKO cat# Z033429-2). The sections were rinsed 3 × 10 min in KPBS-T and then incubated with Alexa647-conjugated secondary antibody (1:200, goat-anti-rabbit, Invitrogen) in KPBS-T for 2 hr at RT. Following that, the sections were rinsed 3x for 10 min with KPBS, and incubated with DAPI (4′,6-diamidino-2-phenylindole) solution (Thermo Fisher Scientific) for 10 min. Lastly, the sections were mounted on gelatin-coated slides and secured with coverslips using an anti-fading medium (SlowFade Diamond Antifade Mountant, S36963, Thermo Fisher Scientific) for fluorescence confocal imaging.
Confocal imaging
Request a detailed protocolConfocal imaging was performed using DMi8 inverted fluorescence microscope (Leica Microsystems) equipped with a 40 × 1.3 NA oil-immersion objective. DAPI, BSA-Alexa488, and Alexa647 were excited with 405 nm, 488 nm, and 638 nm laser diodes, respectively. The imaging was performed in sequential mode (each channel separately), to avoid fluorescence bleed-through between the channels. The emitted light was collected by a hybrid detector after: 410–450 nm (for DAPI), 510–550 nm (for BSA-Alexa488), and 650–750 nm (for Alexa647–conjugated antibodies) bandpass.
Bright-field imaging of fibrinogen/cresyl violet staining was performed using BX53 upright transmitted light microscope (Olympus) equipped with a 60 × 1.4 NA oil-immersion objective.
All images were exported to ImageJ (NIH, version 1.48u4) for further analysis. For comparisons between WT and Apom-/- mice, the images were collected using the same microscope settings, and Z-projected using maximal signal intensity projection from an equally thick section (=10 µm). The WT vs. Apom-/- mice imaging data on the figures are presented in the fluorescence signal intensity scale.
Drug administration
Request a detailed protocolSEW2871 (Sigma-Aldrich) was diluted in PBS (to 0.1 mg/ml in 1% DMSO) and administered as a single i.p. bolus injection (10 μg/g BW) 150 min prior to two-photon imaging or perfusion fixation.
Statistical analysis
Request a detailed protocolStatistical analysis was performed using Prism 7 (v7.0b, Graphpad) and R software (v3.5.1). The data were tested for normal distribution with Pearson’s normality test and either an unpaired two-tailed Student’s t-test with Welch’s correction (normally distributed data) or Mann-Whitney test (non-normally distributed data) was used. The TEM data underwent logarithmic transformation and was evaluated using one-way analysis of variance (ANOVA) followed by Tukey’s multiple comparison test. The sample size has been selected based on our previous experiments using two-photon in vivo microscopy (Kucharz and Lauritzen, 2018), in vivo electrophysiology (Kucharz et al., 2017) and TEM (Christoffersen et al., 2011). The data consisted only of biological replicates, the number of independent biological replicates (N) is provided in the results section and/or on figures. All analyses were performed in randomized order, with TEM analysis performed blinded to experimental conditions. No animals or data outliers have been excluded from analysis.
Data availability
Dataset has been deposited at Dryad Digital Repository: https://doi.org/10.5061/dryad.qg04542.
-
Dryad Digital RepositoryData from: Apolipoprotein M-bound sphingosine-1-phosphate regulates blood-brain barrier paracellular permeability and transcytosis.https://doi.org/10.5061/dryad.qg04542
References
-
Structure and function of the blood-brain barrierNeurobiology of Disease 37:13–25.https://doi.org/10.1016/j.nbd.2009.07.030
-
Hydrophobic ligand binding properties of the human lipocalin apolipoprotein MJournal of Lipid Research 48:1754–1762.https://doi.org/10.1194/jlr.M700103-JLR200
-
An update on the biology of sphingosine 1-phosphate receptorsJournal of Lipid Research 55:1596–1608.https://doi.org/10.1194/jlr.R046300
-
Molecular mechanisms of thrombin-induced endothelial cell permeabilityBiochemistry. Biokhimiia 67:75–84.https://doi.org/10.1023/a:1013904231324
-
Fingolimod (FTY720): discovery and development of an oral drug to treat multiple sclerosisNature Reviews Drug Discovery 9:883–897.https://doi.org/10.1038/nrd3248
-
The blood-brain barrier in neurodegenerative disease: a rhetorical perspectiveJournal of Neurochemistry 111:291–314.https://doi.org/10.1111/j.1471-4159.2009.06319.x
-
Isolation and characterization of human apolipoprotein M-containing lipoproteinsJournal of Lipid Research 47:1833–1843.https://doi.org/10.1194/jlr.M600055-JLR200
-
The blood-brain barrierCold Spring Harbor Perspectives in Biology 7:a020412.https://doi.org/10.1101/cshperspect.a020412
-
Blood-brain barrier breakdown in septic encephalopathy and brain tumoursJournal of Anatomy 200:639–646.https://doi.org/10.1046/j.1469-7580.2002.00065.x
-
Sphingosine 1-phosphate regulates cytoskeleton dynamics: implications in its biological responseBiochimica Et Biophysica Acta (BBA) - Biomembranes 1758:2037–2048.https://doi.org/10.1016/j.bbamem.2006.06.015
-
Cytoskeletal regulation of pulmonary vascular permeabilityJournal of Applied Physiology 91:1487–1500.https://doi.org/10.1152/jappl.2001.91.4.1487
-
Cell-specific blood-brain barrier regulation in health and disease: a focus on hypoxiaBritish Journal of Pharmacology 171:1210–1230.https://doi.org/10.1111/bph.12489
-
Capillary junctions of the rat are not affected by osmotic opening of the blood-brain barrierActa Neuropathologica 63:179–189.https://doi.org/10.1007/BF00685243
-
Protective associations of HDL with blood-brain barrier injury in multiple sclerosis patientsJournal of Lipid Research 56:2010–2018.https://doi.org/10.1194/jlr.M060970
-
Activated protein C mediates novel lung endothelial barrier enhancement: role of sphingosine 1-phosphate receptor transactivationThe Journal of Biological Chemistry 280:17286–17293.https://doi.org/10.1074/jbc.M412427200
-
Sphingosine 1-phosphate promotes endothelial cell barrier integrity by Edg-dependent cytoskeletal rearrangementJournal of Clinical Investigation 108:689–701.https://doi.org/10.1172/JCI12450
-
Changes in blood-brain barrier permeability to large and small molecules following traumatic brain injury in miceEuropean Journal of Neuroscience 25:231–238.https://doi.org/10.1111/j.1460-9568.2006.05275.x
-
Increased endothelial vesicular transport correlates with increased blood-tumor barrier permeability induced by bradykinin and leukotriene C4Journal of Neuropathology & Experimental Neurology 61:725–735.https://doi.org/10.1093/jnen/61.8.725
-
Fluorescence imaging of blood-brain barrier disruptionJournal of Neuroscience Methods 151:262–267.https://doi.org/10.1016/j.jneumeth.2005.08.006
-
Signaling and biological actions of sphingosine 1-phosphatePharmacological Research 47:401–407.https://doi.org/10.1016/S1043-6618(03)00046-X
-
HDL re-examinedCurrent Opinion in Lipidology 26:127–132.https://doi.org/10.1097/MOL.0000000000000161
-
Implications of cerebrovascular ATP-binding cassette transporter G1 (ABCG1) and apolipoprotein M in cholesterol transport at the blood-brain barrierBiochimica Et Biophysica Acta (BBA) - Molecular and Cell Biology of Lipids 1862:573–588.https://doi.org/10.1016/j.bbalip.2017.03.003
-
Altered blood-brain barrier permeability and its effect on the distribution of Evans blue and sodium fluorescein in the rat brain applied by intracarotid injectionPhysiological Research 52:607–614.
-
PSD-95 uncoupling from NMDA receptors by tat- N-dimer ameliorates neuronal depolarization in cortical spreading depressionJournal of Cerebral Blood Flow & Metabolism 37:1820–1828.https://doi.org/10.1177/0271678X16645595
-
Endothelial cell barrier regulation by sphingosine 1-phosphateJournal of Cellular Biochemistry 92:1075–1085.https://doi.org/10.1002/jcb.20088
-
Sphingosine 1-phosphate-induced mobilization of intracellular Ca2+ mediates rac activation and adherens junction assembly in endothelial cellsJournal of Biological Chemistry 280:17320–17328.https://doi.org/10.1074/jbc.M411674200
-
Endothelial cell-surface gp60 activates vesicle formation and trafficking via G(i)-coupled src kinase signaling pathwayThe Journal of Cell Biology 150:1057–1070.https://doi.org/10.1083/jcb.150.5.1057
-
Vesicle formation and trafficking in endothelial cells and regulation of endothelial barrier functionHistochemistry and Cell Biology 117:105–112.https://doi.org/10.1007/s00418-001-0367-x
-
Morphofunctional aspects of the blood-brain barrierCurrent Drug Metabolism 13:50–60.https://doi.org/10.2174/138920012798356970
-
Size-selective loosening of the blood-brain barrier in claudin-5-deficient miceThe Journal of Cell Biology 161:653–660.https://doi.org/10.1083/jcb.200302070
-
Identification of ApoA4 as a sphingosine 1-phosphate chaperone in ApoM- and albumin-deficient miceJournal of Lipid Research 60:1912–1921.https://doi.org/10.1194/jlr.RA119000277
-
Sphingosine 1-phosphate signaling at the blood-brain barrierTrends in Molecular Medicine 21:354–363.https://doi.org/10.1016/j.molmed.2015.03.006
-
The not-so-simple HDL story: is it time to revise the HDL cholesterol hypothesis?Nature Medicine 18:1344–1346.https://doi.org/10.1038/nm.2937
-
Matrix metalloproteinases in the brain and blood-brain barrier: versatile breakers and makersJournal of Cerebral Blood Flow & Metabolism 36:1481–1507.https://doi.org/10.1177/0271678X16655551
-
High-Density Lipoprotein-Associated apolipoprotein M limits endothelial inflammation by delivering Sphingosine-1-Phosphate to the Sphingosine-1-Phosphate receptor 1Arteriosclerosis, Thrombosis, and Vascular Biology 37:118–129.https://doi.org/10.1161/ATVBAHA.116.308435
-
Blood-Brain barrier disruption and neurovascular unit dysfunction in diabetic mice: protection with the mitochondrial carbonic anhydrase inhibitor topiramateJournal of Pharmacology and Experimental Therapeutics 359:452–459.https://doi.org/10.1124/jpet.116.237057
-
Induction of vascular permeability by the sphingosine-1-phosphate receptor-2 (S1P2R) and its downstream effectors ROCK and PTENArteriosclerosis, Thrombosis, and Vascular Biology 27:1312–1318.https://doi.org/10.1161/ATVBAHA.107.143735
-
Sphingosine-1-Phosphate signaling in endothelial disordersCurrent Atherosclerosis Reports 18:31.https://doi.org/10.1007/s11883-016-0586-1
-
Rac downregulates rho activity: reciprocal balance between both GTPases determines cellular morphology and migratory behaviorThe Journal of Cell Biology 147:1009–1022.https://doi.org/10.1083/jcb.147.5.1009
-
Role of sphingosine-1 phosphate in the enhancement of endothelial barrier integrity by platelet-released productsAmerican Journal of Physiology-Lung Cellular and Molecular Physiology 285:L258–L267.https://doi.org/10.1152/ajplung.00311.2002
-
Quantification of blood-brain barrier solute permeability and brain transport by multiphoton microscopyJournal of Biomechanical Engineering 136:031005.https://doi.org/10.1115/1.4025892
-
S1P induces FA remodeling in human pulmonary endothelial cells: role of rac, GIT1, FAK, and paxillinJournal of Applied Physiology 94:1193–1203.https://doi.org/10.1152/japplphysiol.00690.2002
-
Transactivation of sphingosine 1-phosphate receptors is essential for vascular barrier regulation. Novel role for hyaluronan and CD44 receptor familyThe Journal of Biological Chemistry 281:34381–34393.https://doi.org/10.1074/jbc.M603680200
-
An engineered S1P chaperone attenuates hypertension and ischemic injuryScience Signaling 10:eaal2722.https://doi.org/10.1126/scisignal.aal2722
-
Sphingosine 1-phosphate in blood: function, metabolism, and fateCellular Physiology and Biochemistry 34:158–171.https://doi.org/10.1159/000362992
-
Brain barriers: crosstalk between complex tight junctions and adherens junctionsThe Journal of Cell Biology 209:493–506.https://doi.org/10.1083/jcb.201412147
-
Differential coupling of the sphingosine 1-phosphate receptors Edg-1, Edg-3, and H218/Edg-5 to the G(i), G(q), and G(12) families of heterotrimeric G proteinsThe Journal of Biological Chemistry 274:27351–27358.https://doi.org/10.1074/jbc.274.39.27351
-
Rho and rac but not Cdc42 regulate endothelial cell permeabilityJournal of Cell Science 114:1343–1355.
-
S1P control of endothelial integrityCurrent Topics in Microbiology and Immunology 378:85–105.https://doi.org/10.1007/978-3-319-05879-5_4
-
A novel human apolipoprotein (apoM)Journal of Biological Chemistry 274:31286–31290.https://doi.org/10.1074/jbc.274.44.31286
-
In vivo imaging of the cerebral endothelial glycocalyx in miceJournal of Vascular Research 54:59–67.https://doi.org/10.1159/000457799
Decision letter
-
Gary L WestbrookSenior and Reviewing Editor; Oregon Health and Science University, United States
-
Timothy HlaReviewer; Harvard Medical School, United States
In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.
Acceptance summary:
The authors examined the blood-brain barrier (BBB) function of WT and ApoM KO mice, and show using two-photon microscopy that ApoM KO mice exhibit a BBB defect for small molecules. Although this was expected to some degree from prior work, the present manuscript provides a beautiful demonstration using a state-of-the-art live animal imaging. The data are convincing and demonstrate that ApoM / S1P signaling via endothelial S1PR1 is important in controlling the BBB function of small molecular weight compounds.
Decision letter after peer review:
Thank you for submitting your article "Apolipoprotein M-bound sphingosine-1-phosphate regulates blood-brain barrier paracellular permeability and transcytosis" for consideration by eLife. Your article has been reviewed by Gary Westbrook as the Senior Editor, a Reviewing Editor, and three reviewers. The following individuals involved in review of your submission have agreed to reveal their identity: Timothy Hla (Reviewer #2). The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.
Summary:
As you will see from the original text of the reviews below, there were substantial differences in the reviewers' judgment on the potential impact of this work. We discussed this topic among the reviewers. There was consensus that the work was of high quality using a sensitive technique, but the following issues must be addressed in the revision along with a point-by-point response to ALL reviewers' comments. As you will see some of these points will likely require additional data or analysis.
1) There were places where better quantification is necessary (see point 1 of reviewer 1).
2) A more rigorous discussion of the limitations of the study and how it advanced what is already known is needed (see point 2 of reviewer 1).
3) A discussion of alternative hypotheses, e.g. other effects of germline apoM deletions or compensation by other HDL components should be included.
4) It is important to demonstrate whether there was a glial/microglia reaction following entry of plasma proteins (fibrin stain) into the brain parenchyma. (see point of reviewer 3).
5) Do the authors have data on EM analysis of arterioles and arteries (see point 3 of reviewer 1)?
Reviewer #1:
Sphingosine-1-phosphate (S1P) is a bioactive sphingolipid present in blood, where it is typically bound to either circulating apolipoprotein M (apoM; about 70%) or albumin (about 30%). S1P will preferentially activate different S1P G-protein coupled receptors depending on its binding partner in blood, and apoM-bound S1P preferentially binds to cell surface S1PR1 receptors. Here the authors studied the effect of apoM deficiency on blood brain barrier (BBB) permeability using in vivo two-photon microscopy of extravasated tracers and EM analysis of endothelial tight junctions. Knockout of apoM (apoM-/-) in mice induced increased extravasation of small dye molecules, which could be ameliorated by treatment with S1PR1 agonist SEW2871. EM imaging indicated no changes in endothelial junction length or width in capillaries, but in vivo two-photon microscopy revealed increased numbers of extra-vascular "vesicles" along the arterioles and arteries in the brains of apoM-/- mice, which could be reduced only in the penetrating arterioles with application of S1PR1 agonist SEW2871.
This manuscript presents interesting observations in an understudied subject. The manuscript would be improved by addressing the following:
Essential revisions:
1) The experiments in FigureFigure 1 show increased leakage of the small molecule NaFluo in apoM-/- mice (without quantification). The conclusion in the text indicating that this effect is due to paracellular permeability (subsection “ApoM / S1PR1 signaling deficit increases transcellular transport”) is premature. At this point, all that has been shown is a descriptive illustration of enhanced leakage, not "paracellular permeability." Also, from subsequent data presented in the manuscript, it would seem that transcellular transport would play a role in increasing NaFluo into the brain. This portion of the manuscript (subsection “ApoM / S1PR1 signaling deficit increases transcellular transport”) should be revised, quantified and/or possibly combined with the data in Figure 2.
2) These findings seem to be consistent with previous reports showing that activation of S1PR1 reduces endothelial permeability through rearrangement of cellular cytoskeleton and that S1P leads to barrier enhancement via S1PR1 (see for example McVerry and Garcia, 2004; Burridge and Wennerberger, 2004; Singleton.et al., 2005). Additionally, activated protein C enhances endothelial barriers via EPCR-dependent PAR1-mediated activation of Sphk-1 which generates sphingosine-1-phosphate S1P acting on S1PR1 (Feistritze and Riewald, 2005; Finigan, 2005). Please discuss how the authors observations fit with this previous literature.
3) The EM analysis of endothelial junctions is very nicely done, but performed only at the capillary level. Since the authors found that arterioles and arteries show the most pronounced changes in vesicular transport between genotypes with in vivo imaging, should these vessels be also checked for endothelial junction integrity in apoM-/- mice?
4) In the experiments presented in Figure 2, three tracers with similar fluorescence spectra are sequentially injected into the vasculature. Injection of one tracer after another of the same color has the potential to confound data of the following tracer injections. Have the authors checked that previous tracers injected have cleared before injecting the next tracer in these experiments? Is a baseline taken before each subsequent tracer injection?
5) In the BSA-Alexa 488 transcellular experiments illustrating vesicles shown in Figure 4, could the authors confirm the cell type taking up the BSA in the brain with histology?
6) Additionally, the BSA-Alexa 488 transcellular data would be strengthened by adding a conformation of the observations using a different method, e.g. immunostaining or EM imaging.
7) Why were the authors able to see the larger BSA-Alexa 488 tracer cross the BBB, but not the FITC-dextran?
8) The finding that the endothelial thickness is increased with apoM knockout is interesting. What might be the reason for this? Swelling or something else? Please discuss.
Reviewer #2:
Janiurek et al., studied the BBB function of WT and ApoM KO mice. They show, using two-photon microscopy, that ApoM KO mice exhibit a BBB defect in small molecules. This was expected since we know from studies of Yanagida et al., that S1PR1 ECKO also show this phenotype. The present manuscript provides a beautiful demonstration of this using state of the art, live animal imaging studies. The data are convincing and demonstrate that ApoM / S1P signaling via the endothelial S1PR1 is important in controlling the BBB function of small molecular weight compounds. The authors also examined the endothelial morphology and junctions as well as transcytosis in various vessel subtypes. They report different changes in transcytosis in WT and ApoM KO in penetrating arterioles vs capillaries.
Overall, the study is well done and impressive. Some of the results are expected since we already know that S1PR1 regulates the same function in the BBB. It would be useful if the authors were to extend the discussion they had with S1PR2. For example, does S1PR2 inhibitor inhibit the increased permeability seen in ApoM KO mice? Since they have the techniques to examine this issue, this would add much to the field.
There are some minor errors, such as G proteins downstream of S1PR1 and S1PR2 and not GPCRs. Also, a better discussion of HDL/ ApoM function and BBB would be of interest.
Reviewer #3:
The following observations are already published and generally acknowledged:
1) S1P is required for vascular health.
2) S1P signaling maintains BBB integrity.
3) apoM transports S1P to diverse vascular beds to enhance signaling.
4) S1P1 signaling at the BBB endothelium is mediated at least in part by S1P1.
The advance presented in this paper is that apoM-associated S1P is also the primary signaling moiety for the BBB. Evidence favoring a role for S1P1 is provided indirectly by acute systemic (i.p.) administration of a selective agonist at S1P1 but not by genetic manipulation of endothelial cell S1P1. Notably the role of S1P1 at the BBB has also been published more than once (point #4 above).
The data are solid. However, the new information is incremental and will primarily be of interests to specialists in this field. The experimental design is quite limited, showing only that ApoM-KO mice show increased paracellular permeability to small-molecule tracers without evident junctional complex disruption, although endothelial thickness is increased. Both pial and penetrating arterioles demonstrate increased transcytosis of labeled BSA. All abnormalities are rescued by acute systemic administration of a small molecule compound selective S1P1 agonist except for pial arteriole transcytosis of BSA.
Essential revisions:
1) In apoM mice BBB transcytosis of albumin (and other plasma proteins) with parenchymal exposure is reported. Given that plasma proteins in brain parenchyma are potently stimulatory and should induce chronic neuroinflammation or tachyphylaxis what is the status of glial elements (microglia, astrocytes, OPCs)?
2) What is the rationale for S1P1 agonist rescuing phenotype of penetrating but not pial arterioles?
3) One conclusion of the study is that manipulating apoM/S1P might show therapeutic benefit. In that regard, how can one interpret the finding that S1P1 agonist rescues the apoM-deficient phenotype without need for apoM?
https://doi.org/10.7554/eLife.49405.sa1Author response
Summary:
As you will see from the original text of the reviews below, there were substantial differences in the reviewers' judgment on the potential impact of this work. We discussed this topic among the reviewers. There was consensus that the work was of high quality using a sensitive technique, but the following issues must be addressed in the revision along with a point-by-point response to ALL reviewers' comments. As you will see some of these points will likely require additional data or analysis.
1) There were places where better quantification is necessary (see point 1 of reviewer 1).
We have now improved our quantification method (see: reviewer 1 comment #4).
2) A more rigorous discussion of the limitations of the study and how it advanced what is already known is needed (see point 2 of reviewer 1).
We have now better related the study to the current knowledge on the topic; underlined new findings throughout the Discussion section, performed additional internal controls (see: e.g. reviewer 1 comment #7); and included a new paragraph on methods limitations (see: Authors comment#1 at the end of the document).
3) A discussion of alternative hypotheses, e.g. other effects of germline apoM deletions or compensation by other HDL components should be included.
This has been now discussed and added to the manuscript (see: reviewer 2 comment #3).
4) It is important to demonstrate whether there was a glial/microglia reaction following entry of plasma proteins (fibrin stain) into the brain parenchyma. (see point of reviewer 3)
We have performed a series of immunostaining analyses that address fibrin content in brain parenchyma and gliosis (see: reviewer 3 comment #1). In addition, as requested, we have identified the parenchymal cells that sequestered albumin in Apom-/- mice (see:reviewer 1 comment #5).
5) Do the authors have data on EM analysis of arterioles and arteries (see point 3 of reviewer 1)?
Unfortunately, quantitative TEM analysis of large vessels is problematic and has not been performed (explained in reviewer 1 comment #3).
Reviewer #1:
Sphingosine-1-phosphate (S1P) is a bioactive sphingolipid present in blood, where it is typically bound to either circulating apolipoprotein M (apoM; about 70%) or albumin (about 30%). S1P will preferentially activate different S1P G-protein coupled receptors depending on its binding partner in blood, and apoM-bound S1P preferentially binds to cell surface S1PR1 receptors. Here the authors studied the effect of apoM deficiency on blood brain barrier (BBB) permeability using in vivo two-photon microscopy of extravasated tracers and EM analysis of endothelial tight junctions. Knockout of apoM (apoM-/-) in mice induced increased extravasation of small dye molecules, which could be ameliorated by treatment with S1PR1 agonist SEW2871. EM imaging indicated no changes in endothelial junction length or width in capillaries, but in vivo two-photon microscopy revealed increased numbers of extra-vascular "vesicles" along the arterioles and arteries in the brains of apoM-/- mice, which could be reduced only in the penetrating arterioles with application of S1PR1 agonist SEW2871.
This manuscript presents interesting observations in an understudied subject. The manuscript would be improved by addressing the following:
Essential revisions:
1) The experiments in Figure 1 show increased leakage of the small molecule NaFluo in apoM-/- mice (without quantification). The conclusion in the text indicating that this effect is due to paracellular permeability (subsection “ApoM / S1PR1 signaling deficit increases transcellular transport”) is premature. At this point, all that has been shown is a descriptive illustration of enhanced leakage, not "paracellular permeability."
The purpose of Figure 1 was to introduce to the experimental set-up and data collection process and to demonstrate the time-course of fluorescence accumulation in the brain. Quantifications were shown in Figure 2, as it contained the data from all genotypes and all tested fluorophores. With regard to the properties of NaFluo and paracellular permeability:
NaFluo biochemical properties favor paracellular route over transcellular diffusion or vesicular transport of the molecule across the BBB because:
a) NaFluo is a small (0.365 kDa) hydrophilic molecule, which effectively impedes its transcellular diffusion via the hydrophobic compartment of the lipid endothelial cell membrane.
b) NaFluo is negatively charged, and the electrostatic forces repel the molecule from the negatively charged glycocalyx layer and endothelium cell membrane.
c) In case a small fraction of NaFluo permeates endothelial cell membrane, NaFluo is a strong target for endothelium efflux pumps, i.e. multidrug resistance proteins (MRP) that sequester the tracer back to the blood stream (Hawkins et al., 2007).
Therefore, NaFluo is considered to be suitable for characterizing the BBB paracellular route, independently from transcellular diffusion and vesicular transport across the barrier (Cheng et al., 2010). NaFluo and has been previously used in comparative studies of paracellular and transcellular permeability, e.g. by injections of NaFluo (paracellular route assessment) with albumin-conjugated dyes (transcellular transport assessment) (Hawkins and Egleton, 2006; Kozler and Pokorny, 2003). In addition, our recent work shows that NaFluo (as well as Alexa Fluor 488 and FITC-dextrans) traverse the BBB in vivo with diffusion coefficients and fluorescence distribution that cannot be explained by vesicular transport, but is consistent with mechanisms of passive diffusion of dyes along the concentration gradient across the BBB (Kutuzov et al., 2018).
Also, from subsequent data presented in the manuscript, it would seem that transcellular transport would play a role in increasing NaFluo into the brain.
Although we cannot exclude that some NaFluo molecules can be entrapped in transport vesicle during its formation at the luminal side of the endothelium, in our experiments we did not detect any vesicular fraction of NaFluo, Alexa Fluor 488 and FITC-dx at the BBB. In contrast, albumin-conjugated Alexa488 showed a clear vesicular pattern, which is consistent with transcytosis (De Bock et al., 2016; Minshall et al., 2002; Minshall et al., 2000). Therefore, we consider that the incidental transcytosis of NaFluo, Alexa Fluor 488 or FITC-dx would have a negligible contribution to fluorescence increase, and the effect observed in the brain parenchyma occurs due to the increase in paracellular permeability (See also reviewer 1 comment #7).
This portion of the manuscript (subsection “ApoM / S1PR1 signaling deficit increases transcellular transport”) should be revised;
We have now added the relevant comments above to the Materials and methods and to the Discussion section.
quantified;
We improved our analysis method by accounting for baseline changes and systematic error during sequential injections of fluorophores (see reviewer 1 comment #4).
and/or possibly combined with the data in Figure 2.
We have now combined the data from Figure 1 and Figure 2 into a single figure.
2) These findings seem to be consistent with previous reports showing that activation of S1PR1 reduces endothelial permeability through rearrangement of cellular cytoskeleton and that S1P leads to barrier enhancement via S1PR1 (see for example McVerry and Garcia, 2004; Burridge and Wennerberger, 2004; Singleton et al., 2005).
We did not perform imaging of cytoskeleton in our experiments therefore we did not speculate on Apom knock-out effects on cytoskeleton dynamics and BBB permeability. We agree, however, that the cytoskeleton dynamics plays an important role in modulating endothelial barriers, and have now included the following in the Discussion section:
“S1P has been previously shown as an endothelial barrier-enhancing molecule in the lungs (Dudek and Garcia, 2001; Garcia et al., 2001; Schaphorst et al., 2003) and recently, in the brain (Swendeman et al., 2017; Yanagida et al., 2017). Our data ties apoM shortage-induced increase in BBB permeability with S1PR1. The mechanisms of S1PR1-mediated protective effect are not yet fully understood, but may be linked with Rho superfamily of small GTPases. Stimulation of S1PR1 and S1PR2 activates preferentially Rac and Rho, respectively, whilst S1PR3 stimulation activates both GTPases (Windh et al., 1999; Xiong and Hla, 2014). Rac and Rho modulate cytoskeleton dynamics in various cell types (Burridge and Wennerberg, 2004), including the formation and maintenance of a cortical actin ring that stabilizes intercellular junctions between adjoining endothelial cells (Garcia et al., 2001; McVerry and Garcia, 2004). At physiological concentrations, S1P activates preferentially Rac via S1PR1 over Rho (Garcia et al., 2001; Hla, 2003; Shikata et al., 2003), which causes a redistribution of proteins that facilitate cortical actin ring assembly, and consequently tightening of the barrier (Garcia et al., 2001; Mehta et al., 2005). Inhibition of Rac downstream of S1PR1 leads to actin depolymerization, formation of F-actin stress fibers and increased permeability (Singleton et al., 2005; Wojciak-Stothard et al., 2001), and S1PR2 causes Rho activation that inhibits Rac (Dudek and Garcia, 2001; Hla, 2003). However, the formation and disassembly of actin cortical ring affecting BBB permeability should not be simply viewed as Rac vs. Rho antagonism, as inhibition of Rho, similarly to Rac, can reportedly also increase barrier permeability, although without interrupting the actin ring assembly (Etienne-Manneville and Hall, 2002; Garcia et al., 2001). Our quantitative TEM assessment shows that apoM deficiency did not induce changes in junctional complexes ultrastructure. This suggests that the increase of the BBB permeability in Apom-/- mice was either uncoupled from cortical actin ring remodeling, or that during S1PR1 hypostimulation and ongoing S1PR2-mediated Rho inhibition of Rac, Rac activity was still sufficient to maintain the elements of scaffolding complex of cortical actin ring that stabilize junctional complexes.”
Additionally, activated protein C enhances endothelial barriers via EPCR-dependent PAR1-mediated activation of Sphk-1 which generates sphingosine-1-phosphate S1P acting on S1PR1 (Feistritze and Riewald, 2005; Finigan, 2005). Please discuss how the authors observations fit with this previous literature.
We have now included the following in the Discussion section:
“Our in vivo results characterize apoM as a blood-borne molecule that can affect the BBB permeability via S1PR1. In inflammatory conditions, elevated concentrations of blood-circulating thrombin activate protease-activated receptor-1 (PAR1), and subsequently barrier-disruptive metalloproteinases (Bogatcheva et al., 2002; Rempe et al., 2016). On the contrary, low thrombin concentrations activate protein C (APC), a substrate for endothelial cell protein C receptor (EPCR), which alters PAR1 signaling to promote the synthesis of S1P via Sphk-1, and facilitates barrier enhancement (Feistritzer and Riewald, 2005; Riewald et al., 2002). In addition, APC has been shown to also induce S1PR1 phosphorylation, which further promotes the S1PR1 signaling branch in the endothelium (Lee et al., 2001). In line with our results in Apom-/- mice, the barrier enhancement by blood-borne thrombin occurred via S1PR1, as silencing of S1PR1 but not e.g. S1PR3 abolished the APC-mediated effect (Feistritzer and Riewald, 2005; Finigan et al., 2005). Except for apoM and thrombin, S1P signaling can be modulated by hyaluronic acid (HA). Depending on the form, i.e. high-molecular weight-HA (HMW-HA) or low-molecular weight (LMW-HA), the complex promotes interaction of a cell-surface glycoprotein CD44 isoforms with S1PR1 or S1PR1/3, respectively (Singleton et al., 2006). This leads to enhancement of the barrier function via S1PR1 or an increase in BBB permeability via S1PR3 due to transactivation (i.e. phosphorylation) of respective receptors by CD44 isoforms, independently of S1P levels (Singleton et al., 2006). This further corroborates our findings on the regulation of S1P signaling by blood circulating molecules via preferential activation of specific S1P receptors.
However, the analogies should be made with caution, as the mechanistic insights come primarily from in vitro barrier models or non-brain tissue, where endothelial barrier properties may differ from the BBB in vivo. For example, Rac and Rho are known to elicit cell-type and subcellular localization-specific effects (Burridge and Wennerberg, 2004; Donati and Bruni, 2006; McVerry and Garcia, 2004), and their intracellular pathways are cross-linked with numerous regulatory feedback loops, including e.g. Rho inhibiting Rac and vice versa (Burridge and Wennerberg, 2004; Sander et al., 1999). This adds to the complexity of interpretation of Apom knock-out effects on the BBB and is currently an active area in research.”
3) The EM analysis of endothelial junctions is very nicely done, but performed only at the capillary level. Since the authors found that arterioles and arteries show the most pronounced changes in vesicular transport between genotypes with in vivo imaging, should these vessels be also checked for endothelial junction integrity in apoM-/- mice?
We have aimed to characterize endothelial junction ultrastructure in large vessels (ø 25-120 µm) together with capillaries. However, in contrast to capillaries, large vessels were susceptible to perfusion-fixation artifacts (non-uniform distortion of endothelium neighboring large tissue-devoid areas, i.e. vessel lumen), which rendered pial and penetrating vessels not suitable for TEM quantitative assessment of length, thickness, coverage of tight junctions and the area of the endothelium. This information has been now added to the Results section.
4) In the experiments presented in Figure 2, three tracers with similar fluorescence spectra are sequentially injected into the vasculature. Injection of one tracer after another of the same color has the potential to confound data of the following tracer injections. Have the authors checked that previous tracers injected have cleared before injecting the next tracer in these experiments? Is a baseline taken before each subsequent tracer injection?
The time of bloodstream clearance differs between fluorophores and varies between animals. Waiting for tracer clearance would introduce a large systematic error where the time of clearance of a preceding dye affects the timepoint of imaging of a consecutive dye(s). Consequently, this could obscure characterization of the biological effects of SEW2871 on the BBB, as the assessment of permeability in each mouse for each respective tracer would be performed at different time after the drug administration. SEW2871 has a relatively short half-time in the blood plasma with its biological effects peaking at 1 hour, and returning to baseline 4.5 hours after administration (as determined by e.g. Akt phosphorylation)(Jo et al., 2005). Instead, we chose to standardize the time of imaging, i.e. FITC-dx, Alexa Fluor 488 and NaFluo were always imaged 2.5 hours, 3 hours, 3.5 hours after SEW2871 administration, respectively. While this approach is also susceptible to a systematic error that can originate from consecutively injected dyes accumulation, we have now revised our analysis to account for this possibility.
In our revised model, we estimate the contribution of a preceding fluorescence accumulation to subsequent fluorophore(s) signal measurement. We have previously assumed that the fluorescence in the brain parenchyma of a preceding fluorophore(s) did not have a significant effect on the signal on subsequent fluorophore(s). This was based on a preliminary assessment of the data, where the increases of preceding fluorophores were within the standard error of mean changes of a subsequent fluorophore (Author response image 1).

Raw fluorescence signal collected from brain parenchyma after sequential injections of FITC-dx, Alexa Fluor 488 and NaFluo.
A.U. = arbitrary fluorescence unit (detector counts). The time between data points of last frame in Z-stack of preceding fluorophore and time of last Z-stack of subsequent fluorophore was 2 min (1 min of injection + 1 min of Z-stack collection time). n=7 animals in each experimental.
Now, for each mouse, we subtracted the constant autofluorescence background from FITC-dx, Alexa Fluor 488 and NaFluo traces and performed an exponential fit to the FITC-dx trace that was extrapolated to subsequently imaged Alexa Fluor 488 and NaFluo (Author response image 2a). We have deliberately chosen the exponential fit to account for the largest systematic error in our experiment, i.e. the largest theoretical contribution of a preceding tracer to a subsequent tracer signal. The fluorescence is additive, therefore we corrected Alexa Fluor 488 and NaFluo traces by subtracting FITC-dx extrapolated fit values (Author response image 2a) and effectively, uncoupled FITC-dx kinetics from Alexa Fluor 488 and NaFluo. Next, to the corrected trace of Alexa Fluor 488 we performed an exponential fit and extrapolated it to the time of NaFluo imaging. Similarly, as before, we subtracted Alexa Fluor 488 extrapolated fit from NaFluo trace (Author response image 2b).

Correcting for the preceding fluorophore signal contribution to the subsequent fluorophore signal measurement.
(a) Traces from three consecutive injections of fluorophores with a curve fit to FITC-dx trace. The fit is extrapolated and subtracted from Alexa Fluor 488 and NaFluo traces. (b) Second fit is performed to a corrected Alexa Fluor 488 trace, extrapolated and subtracted from NaFluo trace. The model is applicable for both, relatively low (blue traces, WT animal) and relatively high (red traces, apoM-/- mouse) fluorescence changes.
The same approach was used to account for baseline fluorescence changes in the bloodstream (Figure 1—figure supplement1C-D).
We have performed area under curve (AUC) analysis on both types of data, i.e. arbitrary fluorescence units [a.u.] (Figure 1—figure supplement1E-F) and relative change in fluorescence [F/F0] (Figure 1). The results were consistent with our previous notion, i.e. there was no significant leakage of FITC-dx, but an increase of BBB permeability for Alexa Fluor 488 and NaFluo tracers in Apom-/- mice, and that the effect could be reversed by SEW2871 treatment.
We chose to report the data as relative changes for the results to be compatible between different imaging setups. The revised model has been now described in the Materials and methods section with a new supplementary figure (Figure 1—figure supplement 1).
5) In the BSA-Alexa 488 transcellular experiments illustrating vesicles shown in Figure 4, could the authors confirm the cell type taking up the BSA in the brain with histology?
We have performed a series of immunostaining analysis on perfusion-fixated brains of
Apom-/- mice. We tested for colocalization of cell-internalized BSA-Alexa488 in the brain parenchyma with expression of Iba-1 (microglia/macrophages); TMEM119 (microglia-specific); RM0029-11H3 (macrophage-specific) and GFAP (astrocytes). We have successfully identified the BSA-Alexa488 sequestering cells to be macrophages (Author response image 3), and found only minor signs (single puncta) of BSA-Alexa488 uptake by other cell types.

BSA-Alexa 488 uptake by Iba-1 and RM0029-11H3 positive cells, i.e. macrophages.
These results, including additional immunostaining assays, have been added now to the manuscript subsection “ApoM / S1PR1 signaling deficit increases transcellular transport”; (Figure 3—figure supplement 2).
6) Additionally, the BSA-Alexa 488 transcellular data would be strengthened by adding a conformation of the observations using a different method, e.g. immunostaining or EM imaging.
Albumin-based assay is a well-established method to monitor vesicular transcytosis and has been used previously to describe and discern BBB transcellular and paracellular BBB permeability, e.g. in (Knowland et al., 2014). Here, we imaged exclusively the fraction of albumin that circulates in the bloodstream, associates to, and crosses the BBB. Recent evidence suggests that albumin can also be produced in CNS by microglia during neuroinflammatory responses (Ahn et al., 2008). In contrast to blood-borne BSA-Alexa488, it is possible that albumin immunostaining would reveal the total albumin content in the brain, including albumin produced in the parenchyma, thus could potentially obscure the assessment of transcellular BBB permeability.
7) Why were the authors able to see the larger BSA-Alexa 488 tracer cross the BBB, but not the FITC-dextran?
In contrast to FITC-dextrans, transport of albumin (~70kDa) depends on specific interactions of the protein with endothelium. It is dissociated from the diffusion of small molecules via the paracellular route since the space between adjoining endothelial cells is not sufficiently wide to permit albumin leakage. Transport of albumin occurs either by fluid-phase absorption or by selective receptor-mediated transcytosis that involves albumin-docking glycoprotein gp60 and caveolin-1 -mediated formation of albumin-transporting vesicles (De Bock et al., 2016; Mehta and Malik, 2006; Minshall et al., 2002; Minshall et al., 2000; Xiong and Hla, 2014). As discussed in reviewer 1 comment #1b, it is theoretically possible that a fraction of dextran-conjugated fluorophores may be internalized by albumin transport vesicle and cross the BBB, however our data does not support it (Author response image 4).

Additional internal control: the graph shows the fluorescence intensity profile plot of n=10 albumin-containing vesicles.
There was no enrichment of TRITC-dx signal (red) in BSA-Alexa488 –containing vesicles (green) at the BBB.
8) The finding that the endothelial thickness is increased with apoM knockout is interesting. What might be the reason for this? Swelling or something else? Please discuss.
We have now included the following section to the discussion section:
“The small (~10%) increase in endothelium thickness in Apom-/- mice may be due to the impaired S1PR1 -dependent formation and the maintenance of the cortical actin ring; and formation of actin stress fibers facilitated by S1PR2 activation (Garcia et al., 2001; Mehta et al., 2005; Singleton et al., 2005; Wojciak-Stothard et al., 2001). Indirectly, BBB permeability changes may be due to osmotic imbalance and cell swelling, but it is uncertain whether this occurs in the brains of Apom-/- mice. Lack of S1PR1 activation aggravates edema in acute brain and lung injury (Ito et al., 2019), but here the system may not be challenged enough to induce osmotic swelling, as Apom-/- mice have no signs of lung edema and exhibit normal microvasculature morphology (Christoffersen et al., 2011). Other processes may also influence endothelium thickness, but their relevance in Apom-/- mice is yet to be established. For example, S1PR1 activity counteracts vascular endothelial growth factor (VEGF) -mediated endothelium contraction in angiogenesis (Gaengel et al., 2012), but whether S1PR1 hypostimulation affects VEGF signaling in the adult brain in Apom-/- mice is unknown. In addition, impaired S1PR1 signaling activates neuroinflammatory factors (Blaho et al., 2015; Blaho and Hla, 2014), and e.g. TNF-α has been shown to alter cytoskeleton and hyperplastic properties of endothelial cells (Kang et al., 2008; Stroka and Aranda-Espinoza, 2011). Thus, it cannot be excluded, that the changes in endothelium morphology result from both: direct effect of S1PR1 hypostimulation on endothelial cell, and indirect effects linked with impaired BBB-induced loss of brain homeostasis.”
Reviewer #2:
Janiurek et al., studied the BBB function of WT and ApoM KO mice. They show, using two-photon microscopy, that ApoM KO mice exhibit a BBB defect in small molecules. This was expected since we know from studies of Yanagida et al., that S1PR1 ECKO also show this phenotype. The present manuscript provides a beautiful demonstration of this using state of the art, live animal imaging studies. The data are convincing and demonstrate that ApoM / S1P signaling via the endothelial S1PR1 is important in controlling the BBB function of small molecular weight compounds. The authors also examined the endothelial morphology and junctions as well as transcytosis in various vessel subtypes. They report different changes in transcytosis in WT and ApoM KO in penetrating arterioles vs capillaries.
Overall, the study is well done and impressive. Some of the results are expected since we already know that S1PR1 regulates the same function in the BBB.
We would like to thank the Referee for the nice comment. We have now better addressed the novelty of the study in the manuscript, underlining, among others, our methodological approach and vessel-specific effects. We would also like to note that while Yanagida and collegues addressed the role of S1PR1 in the maintenance of the BBB, our study is focused on the upstream regulator of the S1PR1, i.e. apoM, which, as we show, exhibits its actions on the BBB via S1PR1. In contrast to S1PR1ECKO, Apom-/- mice have still functioning S1PR1, thus S1PR1 knock-out is not necessary to induce the incease in BBB permeability, as even hypostimulation of the S1PR1 is sufficient to cause the loss of the barrier function, which (to our knowledge) has not been previously shown.
It would be useful if the authors were to extend the discussion they had with S1PR2. For example, does S1PR2 inhibitor inhibit the increased permeability seen in ApoM KO mice? Since they have the techniques to examine this issue, this would add much to the field.
We did not perform experiments that involve S1PR2 inhibition. However, we would like to thank the referee for the suggestion, as modulating S1PR2 may be an excellent point of entry for a follow-up study. We have now added the following text to the manuscript that continues the discussion (Discussion section).
“The hypostimulation of S1PR1 in Apom-/- mice is likely to shift the signaling equilibrium towards cellular pathways that promote activation of Rho (via S1PR2) over Rac (via S1PR1), and the effects of increased BBB permeability are likely to be the net outcome of the failure in homeostatic interplay between receptor-dependent branches of S1P signaling, rather than simply a decrease in Rac activity. Consequently, the observed impairment of the BBB cannot simply be linked to a reduction in the amount of circulating S1P, but more specifically to the lack of apoM-bound S1P in favor of albumin-bound S1P in the blood. This notion is further supported by the result of the selective stimulation of S1PR1, which restored the BBB permeability in Apom-/- mice to the level of healthy controls. Given that inhibition of S1PR2 tightens the endothelial barrier in lungs by facilitating cortical actin ring assembly and stabilization of tight junctions (Sanchez et al., 2007), and in the brain supports the BBB functional integrity (Cao et al., 2019; Kim et al., 2015), it is plausible that S1PR2 stimulation may alleviate BBB phenotype in Apom-/- mice. It should be noted, however, that S1PR1 expression in the endothelium is higher in the arterial branch of cerebral vasculature compared to venules, whereas S1PR2 expression is higher in venules than in arterioles (Vanlandewijck et al., 2018). It may lead, similar as in SEW2871 treatment, to region-specific effects, where the BBB responses to S1PR2 modulation may differ between distinct types of cerebral microvessels.”
There are some errors, such as G proteins downstream of S1PR1 and S1PR2 and not GPCRs.
We thank for pointing it out and we believe we have corrected now all the minor errors.
Also, a better discussion of HDL/ ApoM function and BBB would be of interest.
We have now added the following section to the manuscript discussion section.
“High-density lipoprotein (HDL) particles are heterogeneous in size and content of apolipoproteins and lipids. In humans, HDL plasma concentrations correlate with reduced susceptibility to cerebro- and cardiovascular diseases (Fellows et al., 2015; Hovingh et al., 2015; Rader and Tall, 2012), and approximately 5% of HDL contains apoM (Christoffersen et al., 2006; Xu and Dahlback, 1999). The beneficial effects of HDL can be, in part, explained by HDL association to apoM/S1P. We show that lack of apoM increases the flux of small molecules across the BBB, similar to previous findings investigating the permeability in lung and adipose tissue (Christensen et al., 2016; Christoffersen et al., 2018; Christoffersen et al., 2011). Stressing the system even further, lack of apoM increases the susceptibility to neuroinflammation (Blaho et al., 2015; Galvani et al., 2015). Mainly this has been interpreted as effects caused by blood-circulating apoM/S1P complex. However, a recent study suggested that porcine brain endothelial cells express and secrete apoM into the brain parenchyma (Kober et al., 2017). This has not been reported in other species but could also play a role in the increased permeability observed in the present study. Except S1P, apoM can potentially bind other ligands, e.g. retinoic acid (Ahnstrom et al., 2007) or oxidized phospholipids (Elsoe et al., 2012). Lacking apoM would increase free oxidized phospholipid content, which potentially can also affect the BBB (Singleton et al., 2009). However, reversing the BBB phenotype by SEW2871 suggests that the observed effects are primarily mediated via S1PR1 rather than e.g. oxidized phospholipids. Noteworthy, with the absence of apoM, S1P may still associate to HDL. A recent study showed that in mice lacking expression of both, apoM and albumin, S1P binds to HDL via novel S1P binding protein apoA4 (Obinata et al., 2019). This may potentially constitute a compensatory mechanism in Apom-/- mice, but whether HDL/apoA4/S1P complex also plays a role in maintaining the BBB is unknown and will need further investigations.”
Reviewer #3:
Essential revisions:
1) In apoM mice BBB transcytosis of albumin (and other plasma proteins) with parenchymal exposure is reported. Given that plasma proteins in brain parenchyma are potently stimulatory and should induce chronic neuroinflammation or tachyphylaxis what is the status of glial elements (microglia, astrocytes, OPCs)?
S1pr1 conditional KO mice exhibit no signs of reactive astrogliosis (Yanagida et al., 2017), despite the increase in pro-inflammatory markers (Galvani et al., 2015). In contrast to S1pr1 conditional KO, in Apom-/-mice the S1PR1 signaling pathway is not abolished. On the other hand, Apom-/-mice have impaired S1PR1 signaling during embryonic development, unlikely to S1pr1 conditional KO. We have now assessed the fibrinogen content in the brain parenchyma; and morphological features of microglia/macrophages (Iba-1) and astrocytes (GFAP).
We detected no signs of fibrinogen entry into the brain parenchyma; and similarly to S1pr1 conditional KO (Yanagida et al., 2017), the morphology of microglia/macrophages and astrocytes was similar between WT and Apom-/-mice. This data suggests a lack of gliosis and has been now added to the manuscript subsection “ApoM / S1PR1 signaling deficit increases transcellular transport”; Figure 3—figure supplement 1).
2) What is the rationale for S1P1 agonist rescuing phenotype of penetrating but not pial arterioles?
This section has been now added to Discussion section.
“The susceptibility of the arterial branch of the microvasculature to apoM shortage may be linked to preferential expression of S1PR1 and APC in pial and penetrating arterioles (Vanlandewijck et al., 2018), thus possibly high reliance in these vessel types on S1PR1 signaling in the maintenance of the BBB. It is unclear, however, why penetrating, but not pial arterioles responded to SEW2871, but anatomical differences may be of importance. For example, S1P signaling promotes the formation of the glycocalyx, which limits the diffusion of hydrophilic and negatively charged substances, and supports the barrier function (Zhang et al., 2016). Glycocalyx in pial arterioles is estimated to be thicker compared to the penetrating arterioles (Yoon et al., 2017), but it is unknown how S1PR1 hypostimulation affects distribution and coverage of glycocalyx layer in Apom-/- mice. In addition, the reconstitution of barrier function may have a different time course in pial and penetrating arterioles. Lastly, in contrast to bolus SEW2871 injection, a sustained stimulation of S1PR1 may be needed to restore BBB function in pial arterioles during ongoing apoM deficit.”
3) One conclusion of the study is that manipulating apoM/S1P might show therapeutic benefit. In that regard, how can one interpret the finding that S1P1 agonist rescues the apoM-deficient phenotype without need for apoM?
We used the term “phenotype” to describe the loss of BBB functional integrity (as in “BBB phenotype”), not all systemic effects that may be caused by apoM shortage. Regarding SEW2871, the therapeutic effect with the absence of apoM can be explained by the fact that in contrast to S1P, SEW2871 does not require apoM as a carrier in the bloodstream, and does not require apoM to activate S1PR1.
https://doi.org/10.7554/eLife.49405.sa2Article and author information
Author details
Funding
Lundbeck Foundation (RIBBDD Initiative R302-2018-2266)
- Martin Lauritzen
Det Frie Forskningsråd (0602-01965B)
- Martin Lauritzen
Nordea Foundation (Grant to the Center for Healthy Aging 114995)
- Martin Lauritzen
The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.
Acknowledgements
We thank the Core Facility for Integrated Microscopy at the University of Copenhagen, especially Klaus Qvortrup, for sample preparation for TEM. We thank Micael Lønstrup and Nikolay Kutuzov from the Department of Neuroscience at the University of Copenhagen for the assistance in the animal surgeries and comments on the manuscript, respectively. This work was supported by the Lundbeck Foundation, the Danish Council for Independent Research (Medical Sciences), and a Nordea Foundation Grant to the Center for Healthy Aging.
Ethics
Animal experimentation: All animal procedures were approved by the Danish National Ethics Committee (permit no. 2019-15-0201-01655) and were in appliance with the ARRIVE guidelines.
Senior and Reviewing Editor
- Gary L Westbrook, Oregon Health and Science University, United States
Reviewer
- Timothy Hla, Harvard Medical School, United States
Publication history
- Received: June 17, 2019
- Accepted: November 1, 2019
- Version of Record published: November 25, 2019 (version 1)
Copyright
© 2019, Mathiesen Janiurek et al.
This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.
Metrics
-
- 2,876
- Page views
-
- 484
- Downloads
-
- 33
- Citations
Article citation count generated by polling the highest count across the following sources: Scopus, Crossref, PubMed Central.
Download links
Downloads (link to download the article as PDF)
Open citations (links to open the citations from this article in various online reference manager services)
Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)
Further reading
-
- Neuroscience
Sensory responses of cortical neurons are more discriminable when evoked on a baseline of desynchronized spontaneous activity, but cortical desynchronization has not generally been associated with more accurate perceptual decisions. Here, we show that mice perform more accurate auditory judgments when activity in the auditory cortex is elevated and desynchronized before stimulus onset, but only if the previous trial was an error, and that this relationship is occluded if previous outcome is ignored. We confirmed that the outcome-dependent effect of brain state on performance is neither due to idiosyncratic associations between the slow components of either signal, nor to the existence of specific cortical states evident only after errors. Instead, errors appear to gate the effect of cortical state fluctuations on discrimination accuracy. Neither facial movements nor pupil size during the baseline were associated with accuracy, but they were predictive of measures of responsivity, such as the probability of not responding to the stimulus or of responding prematurely. These results suggest that the functional role of cortical state on behavior is dynamic and constantly regulated by performance monitoring systems.
-
- Neuroscience
Successive auditory inputs are rarely independent, their relationships ranging from local transitions between elements to hierarchical and nested representations. In many situations, humans retrieve these dependencies even from limited datasets. However, this learning at multiple scale levels is poorly understood. Here, we used the formalism proposed by network science to study the representation of local and higher-order structures and their interaction in auditory sequences. We show that human adults exhibited biases in their perception of local transitions between elements, which made them sensitive to high-order network structures such as communities. This behavior is consistent with the creation of a parsimonious simplified model from the evidence they receive, achieved by pruning and completing relationships between network elements. This observation suggests that the brain does not rely on exact memories but on a parsimonious representation of the world. Moreover, this bias can be analytically modeled by a memory/efficiency trade-off. This model correctly accounts for previous findings, including local transition probabilities as well as high-order network structures, unifying sequence learning across scales. We finally propose putative brain implementations of such bias.