A dural lymphatic vascular system that drains brain interstitial fluid and macromolecules

The brain is enveloped by meningeal linings consisting of three layers: the pia mater tightly attached to the surface of the brain, the avascular arachnoid mater overlying the subarachnoid space, and the vascularized dura mater fused to the cranial bones. To determine whether lymphatic vessels exist within the CNS and surrounding meninges, we analyzed the Prox1-GFP and Vegfr3^+/LacZ reporter mice and whole-mount immunofluorescence preparations of the skull and brain of WT mice against LYVE1, PROX1, PDPN, CCL21, VEGFR3, and PECAM1. To visualize blood vessels, the Prox1-GFP mice were perfused with the fluorescent dye 1,1′-dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine (DiI; Li et al., 2008).

After removing the brain from the skull, no lymphatic vessels were seen on the brain parenchyma or pia mater (not depicted). However, a surprisingly extensive network of lymphatic vessels was observed in the meninges underlying the skull bones (Fig. 1, A–J; and Video 1). In sagittal planes of the inner skull, lymphatic vessels were observed to run down toward the base of the skull along the transverse sinus, the sigmoid sinus, the retroglenoid vein, the rostral rhinal vein, and the major branches of the middle and anterior meningeal arteries (Fig. 1, B and D; and Video 1). In preparations of the superior portions of the skull, the lymphatic vessels were visualized along the superior sagittal sinus, the transverse sinus, the rostral rhinal veins, and the middle meningeal artery (Fig. 1, E and F). A concentration of lymphatic vessels could be observed to exit the skull along the meningeal portions of the pterygopalatine artery, a branch of the internal carotid artery which dives in and out of the skull to give rise to the middle meningeal artery (Fig. 1 I). Lymphatic vessels along the sigmoid sinus and retroglenoid vein exited the skull along the veins (Fig. 1, B and D). In preparations of the base of the skull, lymphatic vessels could be seen in the distal portion of several cranial nerves (optic, trigeminal, glossopharyngeal, vagus, and accessory), exiting the skull along the nerve (Fig. 1, G and H). Lymphatic vessels could be observed also in the dural lining of the cribriform plate, where some vessels passed through the skull into the nasal mucosa (Fig. 1, G and J).

Generally, lymphatic vessels were relatively scarce in the superior portions of the skull, whereas the base of the skull contained a more extensive lymphatic vessel network (Fig. 1 A). Interestingly, only the lymphatic vessels at the base of the skull contained valves, but their distribution was relatively scarce. Valves were separated by long stretches of valveless vessel segments (Fig. 1, B–D).

To determine the localization of these vessels in relation to the meninges, thick skull sections were analyzed. In these preparations, PROX1- and CCL21-positive lymphatic vessels were observed in the meninges underlying the bony parts of the skull, adjacent to the dural blood vasculature (Fig. 1 K).

Whole-mount immunofluorescence staining of the superior sagittal lymphatic vessels showed that, like conventional lymphatic vessels, the dural lymphatic vessels express very low levels of PECAM1 (Fig. 1 L) but high levels of LYVE1, PDPN, VEGFR3, CCL21, and PROX1 (Fig. 1, M–P; Aspelund et al., 2014). Thus, the dural lymphatic vessels are lined by terminally differentiated lymphatic endothelium.

Overall, these data indicated that lymphatic vessels are present in the dura mater of the CNS and drain out of the skull via the foramina of the base of the skull alongside arteries, veins, and cranial nerves. We named these lymphatic vessels on the basis of their venous, arterial, or cranial nerve counterparts. The localization of the vessels suggested a possible role in CSF absorption through the arachnoid mater.

Tracers injected into the brain ISF have been shown to translocate into the CSF via the glymphatic system and further into dcLNs (Koh et al., 2005; Iliff et al., 2012; Plog et al., 2015). However, it is unclear how these tracers gain access into the LNs. We hypothesized that the dura mater lymphatic vessels absorb brain ISF and CSF. To test this, we injected an inert 20-kD poly(ethylene glycol) (PEG) conjugate of the bright near-infrared dye IRDye 680 (PEG-IRDye; Proulx et al., 2013) into the brain parenchyma of the Prox1-GFP mice. 2 h after injection, the tracer was observed to exit the brain via paravenous routes for entry into the CSF space (not depicted), as previously reported (Iliff et al., 2012). Lymphatic drainage of brain ISF was confirmed by visualization of an intense signal in the dcLN but not in the superficial cervical LNs (scLNs; Fig. 2, A and B). Preferential drainage into the dcLN ipsilateral to the side of injection was observed (Fig. 2, A and C). When the tracer-filled afferent lymphatic vessels of the dcLNs were followed upstream, the vessels appeared to drain from the base of the skull (Fig. 2, C and D). Inside the skull, some PEG-IRDye filling of dura mater lymphatic vessels was observed only in the basal parts of the skull (Fig. 2, E and F), suggesting uptake by the lymphatic vessels but a quick washout. When the efferent lymphatic vessel of the dcLN was ligated (Fig. 2, G and H), enhanced filling of the dural lymphatic vessels was observed (Fig. 2, I–K). These data suggest that the dura mater lymphatic vessels absorb brain ISF/CSF from the subarachnoid space for transport into downstream dcLNs.

VEGF-C/D signaling via VEGFR3 is a critical regulator of lymphangiogenesis (Secker and Harvey, 2015). To (a) study whether dura mater lymphatic vessels are regulated by VEGFC/D–VEGFR3 signaling and (b) characterize an animal model in which the functional consequences of dura mater lymphatic vessel aplasia can be examined, we investigated the K14-VEGFR3-Ig transgenic (TG) mouse, which has impaired VEGF-C/D–VEGFR3 signaling. These mice express a soluble VEGF-C/D trap protein consisting of the ligand-binding Ig homology domains 1–3 of VEGFR3 fused with the Fc domain of Igγ (Mäkinen et al., 2001). Although the VEGF-C/D trap transgene is expressed in keratinocytes, the circulating protein inhibits lymphangiogenesis in most tissues, and the mice display LN hypoplasia (Mäkinen et al., 2001; Alitalo et al., 2013). Lymphatic vessels were absent from both superior and basal parts of the skull in the TG mice compared with WT littermate mice (Fig. 3, A–F). Surprisingly, the mice displayed absence of only the scLNs but not dcLNs (Fig. 3, G–I; and Fig. 4 C). These data indicate that the dura mater lymphatic vessels are very sensitive to the inhibition of VEGF-C/D signaling and that the K14-VEGFR3-Ig TG mouse is a suitable model for studying the functional consequences of the absence of lymphatic drainage from the brain.

First, we hypothesized that the absence of dura mater lymphatic vessels would impair the clearance of ISF and solutes from the brain. Thus, brain water content and ISF pressures (IFPs) were measured in TG and WT mice. Surprisingly, the IFP (TG vs. WT: 2.50 ± 0.54 vs. 2.53 ± 0.53 mmHg, P = 0.92, n = 6/group) and water content (TG vs. WT: 3.68 ± 0.023 vs. 3.71 ± 0.043 g/g dry weight, P = 0.27, n = 4/group) were not significantly different between the two groups. These results suggest that in physiological conditions, the brain has alternative ways to manage fluid extravasated from the blood vessels.

Second, we hypothesized that the absence of dura mater lymphatic vessels may impair macromolecule clearance from the brain. To test this, we studied the cerebral clearance of Alexa Fluor 488–conjugated OVA (A488-OVA, ∼45 kD), a macromolecule which retains fluorescent signal during fixation. We recorded cerebral, dcLN, and dura mater lymphatic vessel fluorescence from tissues 2 h after injection into the brain parenchyma of TG and WT littermate mice. Mice were perfusion fixed after sacrifice to prevent outflow of the tracer. Interestingly, the TG mice displayed a significant reduction in the amount of OVA cleared at the 2-h time point after injection (Fig. 4, A and B). Furthermore, a nearly complete abrogation of OVA accumulation was observed in the dcLNs of the TG mice (Fig. 4, C and D). Tracer-filled lymphatic vessels could be observed around the pterygopalatine artery and middle meningeal artery of WT mice, but this was absent in the TG mice (Fig. 4, E and F). To assess other possible causes for the drainage defect, we analyzed glymphatic function and the dcLN capacity for drainage. To this extent, the TG mice did not display qualitative defects in glymphatic function, as indicated by detectable paravascular outflow of the tracer in the subendothelial and perivascular space (Fig. 4, G and H), or a significant reduction in the amount of draining lymphatic vessels in the dcLN (Fig. 4, I and J). We also studied PEG-IRDye transfer from the subarachnoid space into the dcLNs after cisterna magna injection, which was significantly inhibited in the TG mice (Fig. 5). Overall, these data imply that the dura mater lymphatic vessels contribute to the clearance of macromolecules from the brain.

In this study, we report the surprising finding of a lymphatic vessel network in the dura mater of the CNS and show that dura mater lymphatic vessels are lined by fully differentiated PROX1⁺/VEGFR3⁺/LYVE1⁺/PDPN⁺/CCL21⁺/PECAM1^low lymphatic endothelium that is unique in its morphology and scarcity of valves. In the late eighteenth century the Italian anatomist Paolo Mascagni described what he called lymphatic vessels in the meninges and on the surface of the brain, but his finding could never be reproduced (Mascagni and Bellini, 1816; Lukić et al., 2003). The CNS proper has since been considered devoid of lymphatic vasculature. Incidentally, lymphatic vessels were mentioned in an electron microscopic study of the rat dura mater innervation. Furthermore, lymphatic vessels were detected in association with the murine cribriform plate and the human optic nerve (Andres et al., 1987; Gausas et al., 2007; Furukawa et al., 2008). However, the extent of the dura mater lymphatic network, or its role in CSF clearance, has not been realized.

According to the classical textbook model, CSF is produced by the choroid plexus, flows through the ventricles and the subarachnoid space, and is absorbed by arachnoid granulations for transport into the cerebral venous sinuses (Pollay, 2010). However, recent discoveries have established the glymphatic system as a critical regulator of cerebral waste clearance, especially during sleep (Iliff et al., 2012; Xie et al., 2013). In addition to the CSF clearance via arachnoid granulations, several studies have established that a part of brain ISF and CSF is drained into cervical LNs, yet it has been unclear how CSF enters the LNs (Koh et al., 2005; Weller et al., 2009). The observation of CSF tracers in the nasal lymphatic vessels under the cribriform plate has suggested clearance via olfactory nerve sheaths through the cribriform plate (Kida et al., 1993; Koh et al., 2005). Additionally, CSF clearance has been observed to occur along spinal and cranial nerve sheaths with subsequent entry into extracranial lymphatic vessels (Miura et al., 1998; Weller et al., 2009).

Our data indicated filling of the dura mater lymphatic vessels after intraparenchymal injection of the tracer and the lack thereof in the K14-VEGFR3-Ig TG mice. This suggests a model in which a part of the brain ISF, downstream of the glymphatic system, is cleared directly from the subarachnoid space as CSF into the dura mater lymphatic vasculature. Interestingly, we also observed lymphatic vessels draining out of the skull along the dura mater of cranial nerves. Furthermore, we observed lymphatic vessels crossing the cribriform plate, which may explain some of the previous observations. Because of the lack of other known direct anatomical connections between the CSF space and extracranial lymphatic vessels, the dura mater lymphatic vessels are likely to represent the most important CSF source for the extracranial lymph compartment.

The importance of understanding the mechanisms of brain waste management are highlighted in patients suffering from Alzheimer’s disease and other neurodegenerative diseases characterized by the pathological accumulation of misfolded proteins, such as amyloid β, into the brain parenchyma (Deane et al., 2008; Huang and Mucke, 2012). In other tissues, lymphatic vessels are critical for the absorption of macromolecules (Tammela and Alitalo, 2010). In the brain under physiological conditions, a major part of the cerebral amyloid β is removed by the transvascular route (Zlokovic, 2011; Zhao et al., 2015). However, recent evidence suggests that the glymphatic system may also be key in amyloid β clearance (Iliff et al., 2012). The present data show that the absence of dura mater lymphatic drainage results in inhibited clearance of OVA from the brain interstitium, suggesting that dura mater lymphatic vessels are critical for the absorption of macromolecules from the brain ISF and CSF.

Importantly, these findings open new avenues for research. Several other potential roles of dura mater lymphatic vessels can be envisioned, such as in the trafficking of cerebral immune cells, in antigen presentation in the dcLNs, and in the clearance of brain edema. These data may also explain why primary brain tumors can rarely metastasize into cervical LNs (Mondin et al., 2010). Interestingly, surgical removal of the dcLN results in cognitive impairment in mice (Radjavi et al., 2014), and ligation of the deep cervical lymphatic vessels has been reported to aggravate cerebral ischemia after stroke by increasing brain water edema and infarction volume in rats (Si et al., 2006). Further studies should be conducted to define the full contribution of dura mater lymphatic vasculature in CNS homeostasis and disease.

The study was approved by the Committee for Animal Experiments of the District of Southern Finland.

The K14-VEGFR3_1-3-Ig (FVB/N and C57BL/6J backgrounds; Mäkinen et al., 2001), Prox1-GFP (C57BL/6J albino background; Choi et al., 2011), and Vegfr3^+/LacZ (FVB/N background; Dumont et al., 1998) mouse lines have been published previously. WT littermate mice were used as controls. For WT analysis, C57BL/6J mice were used. For tissue analysis, mice were given a lethal dose of ketamine and xylazine and perfusion fixed through the left ventricle with ice-cold 1% paraformaldehyde (PFA) after puncture of the right auricle. The tissues were immediately immersed in 4% ice-cold PFA and postfixed overnight at 4°C, washed in PBS, and processed for staining. Prox1-GFP mouse tissues were freshly imaged without fixation.

For whole-mount staining of the skull bones for laser-scanning confocal microscopy, the fixed skulls were dissected and underwent a mild decalcification with 0.5 M EDTA, pH 7.4, overnight at 4°C. For whole-mount staining of the basal skull for fluorescent stereomicroscopy, no decalcification was performed. After washes with PBS, the tissues were permeabilized in 0.3% Triton X-100 in PBS (PBS-TX) and blocked in 5% donkey serum/2% bovine serum albumin/0.3% PBS-TX. Primary antibodies were added to the blocking buffer and incubated with the tissue overnight at room temperature (RT). After washes in PBS-TX, the tissues were incubated with fluorophore-conjugated secondary antibodies in PBS-TX overnight at RT, followed by washing in PBS-TX. After postfixation in 1% PFA, the superior portions of the skull were washed with PBS and mounted in Mowiol 4-88 mounting medium (Sigma-Aldrich) containing 1,4-diazabicyclo[2.2.2]octane (DABCO; Sigma-Aldrich) and sealed with Cytoseal (Thermo Fisher Scientific). Clothespins were used to hold the coverslip and the microscopic slide together before the Cytoseal and Mowiol hardened. Tissues for fluorescent stereomicroscopy were stored in PBS and imaged immediately.

For cryosections of the skull, the fixed tissues underwent decalcification with 0.5 M EDTA, pH 7.4, for 72 h, immersion into 20% sucrose and 2% polyvinylpyrrolidone (PVP) for 24 h at 4°C, embedding in OCT compound (Tissue-Tek), and freezing for storage at −80°C. For other cryosections, the fixed tissues were immersed into 25% sucrose and embedded as above. Tissues were cut into 10–100-µm sections using a cryostat (Microm HM 550; Thermo Fisher Scientific). The sections were air-dried, encircled with a pap-pen, rehydrated in PBS, and blocked with 3% BSA in PBS-TX at RT. After primary antibody incubation at 4°C in 3% BSA in PBS overnight, the sections were washed with PBS and incubated for 2–3 h with the appropriate fluorophore-conjugated secondary antibody conjugates and 3% BSA in PBS. After washes with 0.1% PBS-TX, the sections were mounted with Vectashield mounting medium containing DAPI (Vector Laboratories). For X-gal staining of skulls from LacZ reporter mice, the tissues were fixed with 0.2% glutaraldehyde and stained by the β-galactosidase substrate X-Gal (Promega) with a published protocol (Karkkainen et al., 2004).

Blood vessels of Prox1-GFP mice were labeled by cardiac perfusion using a fluorescent lipophilic carbocyanine dye DiI as previously described (Li et al., 2008). DiI incorporated into the EC membranes was visualized with fluorescent stereomicroscopy.

The following primary antibodies were used for immunostaining of mouse tissues: rabbit anti–mouse PROX1 (1:200; Petrova et al., 2008), goat anti–human PROX1 (1:500; R&D Systems), polyclonal goat anti–mouse VEGFR3 (1:50; R&D Systems), rat anti–mouse PECAM-1 (1:500; clone MEC 13.3; BD), polyclonal rabbit anti–mouse LYVE1 (1:1,000; Karkkainen et al., 2004), goat anti-CCL21 (1:500; R&D Systems), IgG fraction of rabbit polyclonal anti–mouse PDPN (1:500; provided by D. Kerjaschki, University of Vienna, Vienna, Austria; Wick et al., 2007), and rat anti–mouse endomucin (EMCN; 1:500; clone V.7C7; Santa Cruz Biotechnology, Inc.). The primary antibodies were detected with the appropriate Alexa Fluor 488, 568/594, or 633/647 secondary antibody conjugates (1:500; Molecular Probes/Invitrogen).

To evaluate the outflow pathways and clearance of tracers from the brain interstitium, mice were anesthetized with a mixture of 80 mg/kg ketamine (Ketalar; Pfizer) and 6 mg/kg xylazine (Rompun vet; KVP Pharma + Veterinär Produke GmbH) or 160 mg/kg ketamine (Narketan; Vétoquinol) and 0.4 mg/kg medetomidine (Domitor; Orion Pharma) and placed into a stereotactic device. A midline skin incision was made to reveal the skull bone, which was thinned with a dental drill ∼2 mm lateral and 2.5 mm caudal to the bregma. 4 mg/ml A488-OVA (O-34781; Molecular Probes) or 20 µM of 20-kD PEG-IRDye (provided by P. Luciani and J.-C. Leroux, ETH Zurich, Zurich, Switzerland; Proulx et al., 2013) was injected into a 2-mm depth from the bregma in 0.5 µl with a 34-G Hamilton needle at a 0.1 µl/min rate over 5 min with a syringe pump (Harvard Apparatus). After the indicated time, the mice were sacrificed with a lethal dose of anesthesia. For visualization of PEG-IRDye, tissues were immediately imaged ex vivo. For visualization of A488-OVA, mice were perfusion fixed, and the entire head and neck with the cervical LNs were immersed in 4% ice-cold PFA and further postfixed overnight at 4°C with constant shaking. Fixed tissues were then washed in PBS and processed for imaging or staining as described above.

To evaluate the PEG-IRDye clearance from the subarachnoid space, mice were anesthetized with a mixture of 160 mg/kg ketamine (Narketan) and 0.4 mg/kg medetomidine (Domitor) and placed into a stereotactic frame. The neck muscles were bluntly dissected through a small midline incision to reveal the dura mater overlying the cisterna magna. 20 µM of 20-kD PEG-IRDye was injected into the subarachnoid space in 10 µl with a 34-G Hamilton needle at a rate of 2 µl/min over 5 min with a syringe pump.

Laser-scanning confocal micrographs of the fluorescently labeled samples were acquired at high resolution with an LSM 780 microscope (Carl Zeiss; air objectives 10× Plan-Apochromat with NA 0.45 and 20× Plan-Apochromat with NA 0.80, oil objective 40× Plan-Neofluar with NA 1.3 and oil objective 63× Plan-Apochromat with NA 1.4) using multichannel scanning in frame. The ZEN 2010 software (Carl Zeiss) was used for image acquisition.

Fluorescent stereomicrographs of labeled samples were obtained with an Axio Zoom.V16 fluorescence stereo zoom microscope (Carl Zeiss) equipped with an ORCA-Flash 4.0 digital sCMOS camera (Hamamatsu Photonics) or an OptiMOS sCMOS camera (QImaging). The ZEN 2012 software (Carl Zeiss) was used for image acquisition. Brain epifluorescence images were obtained with a Caliper IVIS Kinetic imaging system (PerkinElmer) equipped with a XFO-6 (Dolan-Jenner Fiber-Lite PL-900 Illuminator; quartz halogen lamp) for fluorescent imaging and an iXon⁺ 888 EMCCD camera (Andor Technology). Images were processed and region of interest efficiencies were calculated with the Living Image 3.2 software. Image brightness and contrast were adjusted using ImageJ (National Institutes of Health) or Photoshop (Adobe) software. Quantitative analysis of the micrographs was performed using the ImageJ software.

Mice were sacrificed with carbon dioxide. The brains were removed from the skull and placed on a preweighed piece of aluminum foil and immediately weighed to obtain the wet weight. The dry weight was recorded after dehydration for 5 d in an 80°C oven. Water content was calculated as (wet weight − dry weight)/dry weight. The data shown in the text is representative of two independent experiments.

Mice were anesthetized with a mixture of 120 mg/kg ketamine (Ketalar) and 0.24 mg/kg medetomidine (Domitor) in saline given s.c. and placed into a stereotactic device. Using a dental drill (2 mm OD), the skull bone was thinned at a site 2 mm caudal and lateral to the bregma. Brain ISF pressure was measured with micropipettes, tip diameter 2–4 µm, as described in detail previously (Wiig and Reed, 1983). Pipettes were inserted through an intact dura, and pressures were recorded 150–300 µm into brain tissue. The data shown in the text is representative of three independent experiments.

All values are expressed as mean ± SD. Quantitative data were compared between different groups by two-sample (unpaired Student’s) two-tailed t test assuming equal variance. Two-way ANOVA followed by Šídák’s post-hoc test was used for multiple comparisons. Differences were considered statistically significant at P < 0.05.

Video 1 shows dura mater lymphatic vessels in the lateral aspects of the interior portions of the skull in sagittal plane in the Prox1-GFP mouse.

We thank Drs. Paola Luciani and Jean-Christophe Leroux for the PEG-IRDye; Dr. Dontscho Kerjaschki for the antibodies against PDPN; Kirsi Lintula, Riitta Kauppinen, Jarmo Koponen, and Tapio Tainola (University of Helsinki, Helsinki, Finland) for technical assistance; the Biomedicum Imaging Unit for help with imaging; and the staff of the University of Helsinki Laboratory Animal Centre for technical assistance with the mouse work.

This study was supported by grants from the Academy of Finland, the Sigrid Juselius Foundation, the European Research Council (ERC-2010-AdG-268804), the Swiss National Science Foundation (310030B_147087), VESSEL-Marie Curie multipartner Initial Training Network for Vascular Biology (EU FP7-PEOPLE-2012-ITN), the Leducq Transatlantic Network of Excellence on Lymph Vessels in Obesity and Cardiovascular Disease (11CVD03; to K. Alitalo and M. Detmar), and the Research Council of Norway (222278/F20).

The authors declare no competing financial interests.