Differential impact of lymphatic outflow pathways on cerebrospinal fluid homeostasis

Circulation of blood and cerebrospinal fluid (CSF) is pivotal in central nervous system (CNS) homeostasis. The cardiovascular system supplies nutrients and is able to uptake and remove certain waste products from tissue; however, the lymphatic system is additionally required to help recover water that otherwise accumulates in tissue and to remove other waste products through bulk fluid reuptake (Swartz and Fleury, 2007; Wiig and Swartz, 2012; Moore and Bertram, 2018; Oliver et al., 2020). Most tissues are extensively vascularized by the lymphatic system, beginning with blind-ending initial lymphatic capillaries that merge into larger collecting lymphatic vessels (cLVs) with unidirectional valves before reaching the lymphatic ducts and finally draining back into the blood near the subclavian veins (Moore and Bertram, 2018). Depending on the tissue and context, lymph flow is driven by a combination of intrinsic contractile pumping of the cLVs and extrinsic forces such as pressure created by skeletal muscle contraction or positional changes (Moriondo et al., 2016; Moore and Bertram, 2018; Zawieja et al., 2018; Castorena-Gonzalez et al., 2018; Petrova and Koh, 2020). Along the way, lymph will pass through one or more lymph nodes, where the adaptive immune system surveys the lymph for signs of infection, injury, and malignancy (Grant et al., 2020; Petrova and Koh, 2020; Randolph et al., 2017).

Whereas the lymphatic system in many tissues is extensively distributed, the CNS-draining lymphatics are restricted to the edges, underscoring the importance of pre-lymphatic fluid dynamics in conveying waste products and immune-relevant molecules from the parenchyma to CSF and finally to lymphatic vessels at the borders of the CNS. Our current understanding of this process is centered around the concept of the “glymphatic” system (Xie et al., 2013; Iliff et al., 2012; Louveau et al., 2017; Papadopoulos et al., 2020). The glymphatic system is a model in which CSF in the subarachnoid space flows into the parenchyma along arteries, filters through the extracellular space, and effluxes along veins (Iliff et al., 2012; Hablitz and Nedergaard, 2021). The perivenous spaces also facilitate CSF exchange with the dura mater where vessels cross the arachnoid barrier (Smyth et al., 2024). In the dura mater, an expanded array of immune cells is concentrated around the dural venous sinuses and is able to interact with the content of CSF (Papadopoulos et al., 2020; Rustenhoven et al., 2021). Finally, the dura also hosts the meningeal lymphatic vasculature, where fluid is taken up and removed from the system (Louveau et al., 2015; Aspelund et al., 2015; Ahn et al., 2019). The fluid then follows two routes to the deep cervical lymph node (dcLN), including a path adjacent to the jugular vein and a path through the nasopharyngeal lymphatic plexus that merge just before the node (Jacob et al., 2022; Yoon et al., 2024). Fluid outflow also occurs across the cribriform plate—although this specific route is likely less active in humans than rodents (Melin et al., 2020)—as well as along other nerve and vessel exit points, including in the spine (Louveau et al., 2018; Ahn et al., 2019; Ma et al., 2019), ultimately draining through other lymphatic networks. There has been a controversy about the roles of intracranial (i.e., meningeal) and extracranial (i.e., nasal) lymphatics in CSF efflux and CNS homeostasis as well as the connectivity and function of different CSF-draining lymph nodes.

Lymphatic drainage from the CNS has been implicated in Alzheimer’s disease (Da Mesquita et al., 2018, 2021; Zou et al., 2019; Ding et al., 2021), traumatic injury (Bolte et al., 2020; Salvador et al., 2023; Liao et al., 2023), brain tumors (Song et al., 2020; Hu et al., 2020), migraine (Nelson-Maney et al., 2024), craniosynostosis (Ang et al., 2022; Ma et al., 2023), and stroke (Yanev et al., 2020; Boisserand et al., 2024). Here, we focused on the contribution of CSF-draining cranio-cervical lymphatic networks to fluid entry into, circulation within, and efflux from the CNS. We found that outflow to the dcLN was largely driven by extrinsic forces, while active pumping drove superficial cervical lymph node (scLN) drainage. At baseline, drainage to dcLN was most prominent, whereas drainage to scLN is primarily recruited in cases where outflow to dcLN is insufficient. We also found that dual-ligation of both dcLN and scLN pathways has distinct effects on CNS fluid outflow when compared to ligation of either pathway alone, which has implications for modeling CNS lymphatic dysfunction in aging and disease.

We studied the in vivo behavior of cLVs afferent to CSF-draining lymph nodes (Fig. 1 A) using high-speed imaging of Prox1-eGFP reporter mice (Jung et al., 2017). We observed spontaneous lymphatic pumping in scLN-afferent cLVs; however, pumping was rarely observed in dcLN-afferent cLVs (Fig. 1, B–F; Fig. S1, A–C; and Video 1). Despite evident drainage of CSF tracer to both sets of lymph nodes, the differences in cLV behavior of the scLN and dcLN afferents suggest underlying differences in the physiological context in which CSF flows into these networks.

In cLVs with less intrinsic pumping behavior, extrinsic tissue-dependent forces play a greater role in generating a pressure gradient (Quick et al., 2009; Moriondo et al., 2015, 2016; Scallan et al., 2016). To gain insight into extrinsic factors that drive CSF outflow into the dcLN, we visualized flow in dcLN afferents with a particulate tracer (1.0 and 0.5 µm fluorescent microspheres) delivered by intracisternal magna (i.c.m.) injection and tracked the movement of these beads within the vessel (Fig. 1 G and Video 2). Flow in dcLN-afferent cLVs was oscillatory, though not in the ∼2–20 min⁻¹ frequency range of lymphatic pumping (Scallan et al., 2016; Davis and Zawieja, 2024), but instead at 137.6 ± 3.3 min⁻¹ (Fig. 1 H and Fig. S1 D), around the typical respiratory rate we observed in anesthetized mice yet below the heart rate (Fig. S1 E). The respiratory rate is also markedly represented in the intracranial pressure (ICP) waveform (Fig. S1, F and G). Collectively, these results suggest that flow to the dcLN is extrinsically driven, coupled with ICP oscillation during the cardiorespiratory cycle. The absence of a lymphatic pumping frequency in particle flow into the dcLN also indicates that upstream lymphatic pumping is not generating the pressure to sustain drainage.

We were also able to identify lymph-flow oscillations in afferent collecting lymphatic by observing the opening and closing of the lymphatic valves in Prox1-eGFP reporter mice without injection of tracer into the CSF. Within dcLN-afferent cLVs, we found significant frequencies in the range of the respiratory rate, but only rare, weak components in the frequency range of lymphatic pumping (Fig. 1, I–K; Fig. S1, H and I; and Video 3). Conversely, valve position in scLN afferents reflects the frequencies from lymphatic pumping but did not contain additional higher peaks in the cardiorespiratory ranges (Fig. 1, C, I, and J; Fig. S1, H and I; and Video 3). This indicates that the flow of CSF into the scLN is driven by contractile pumping and, furthermore, lacks the coupling to ICP oscillations that are present in dcLN outflow.

Although we did not generally observe pumping in dcLN-afferent cLVs, our vessel measurement experiments suggested gradual changes in vessel diameter (Fig. 1 C). These could be caused either indirectly by changes in intravascular pressure or directly by active, tonic changes to regulate flow. To distinguish between these possible explanations, we investigated the behavior of lymphatic muscle cells (LMCs) by imaging calcium-indicator fluorescence in Myh11-CreER^T2 Salsa6F mice, which express tdTomato and GCaMP6f in smooth muscles (Fig. 1, L and M; Fig. S1, J–L; and Video 4). In scLN-afferent cLVs, waves in GCaMP6f fluorescence coincided with vessel contraction during pumping (Fig. 1 N). In dcLN-afferent cLVs, we observed minimal changes in GCaMP6f fluorescence in vessels that maintained constant diameter; however, in a case where gradual dilation was observed, we found that GCaMP6f fluorescence gradually decreased, consistent with active tonic vasodilation. This finding suggests that dcLN-afferent cLV diameter is actively regulated by LMC tone under physiological conditions.

In light of the different driving forces between the dcLN and scLN afferent cLVs, we next investigated how specific physiological variables might modulate outflow through each network. Changes in posture have a substantial effect on ICP, as well as other pressure gradients (Fig. 2, A and B; and Fig. S2 A) (Gergelé and Manet, 2021). To test how posture affects CSF outflow through these routes, we delivered ovalbumin (OVA) tracer to CSF by i.c.m. injection and then measured drainage to scLN and dcLN 1 h later in three groups of mice: prone mice with 30° head elevation, prone mice with 30° head declination, and prone level mice. These postural conditions did not have a significant effect on tracer outflow to the dcLN; however, head-down tilt resulted in a robust increase in tracer drainage to the scLN (Fig. 2, C–E). We also assessed the lumbar lymph nodes but did not find a detectable tracer in any group (Fig. S2, B and C). This is likely due to the net upward movement of spinal subarachnoid CSF and ventricular tracer that has been reported to reach the spine more readily (Dreha-Kulaczewski et al., 2017; Ma et al., 2019). Interestingly, the increased drainage to scLN in the head-down position was accompanied by substantial accumulation of tracer in the nasal compartment of the cranium imaged after tissue clearing and light-sheet microscopy (Fig. 2, F and G), as well as immediately after dissection (Fig. 2, H and I). The tracer in the nasal compartment was found diffusely in the soft tissue of the nasal mucosa. This finding suggests that although drainage to the scLN is not directly coupled to ICP oscillations, it is indirectly driven by changes in mean ICP via pressure-dependent CSF outflow into nasal soft tissue across postural changes.

To better understand the relationship between the dcLN and nasal-to-scLN pathways, we compared mice with intact lymphatic networks to mice with ligation of the afferent cLVs of either the scLN, dcLN, or both (Fig. 3 A). 1 h after OVA tracer injection to the CSF, scLN ligation did not affect tracer accumulation in the dcLN (Fig. 3, B and C); however, dcLN-ligation significantly increased tracer outflow to the scLN (Fig. 3, B and D). As we observed in the 30° head declination condition, increased scLN tracer accumulation in dcLN-ligated mice was also accompanied by increased tracer accumulation in nasal soft tissue, observed by light-sheet imaging of cleared skulls (Fig. 3, E and F) and immediately after dissection (Fig. 3, G and H). Interestingly, however, tracer still accumulated in the nasal compartment in the dual-ligation condition, where the nasal-to-scLN pathway is not active. Although the tracer filled a comparable volume of the nasal soft tissue, the mean intensity was higher in dual-ligated mice than in dcLN-ligated mice. While we expected that scLN ligation would impede CSF-to-nasal fluid efflux, the accumulation of OVA in the nasal compartment suggests that fluid is still able to pass through the compartment.

To understand how dysfunction in one or both lymphatic pathways impacts CSF dynamics, we studied two pathway-independent readouts of CSF fluid homeostasis. First, we measured the transit of CSF tracer to blood. Evans Blue (DB53) behaves as a small molecule tracer (960 Da) in the CSF but is also retained in blood due to binding endogenous albumin (Plá et al., 2022). For the same reason, the free dye is less retained in lymph nodes than large tracers and therefore better represents both lymphatic and non-lymphatic routes of CSF clearance to blood (Clement et al., 2018; Plá et al., 2022). Over the course of 60 min after i.c.m. injection of 4% wt/vol DB53, we assessed outflow to the blood via serial imaging of the femoral vein (Fig. 4, A and B). We found that total tracer efflux was significantly diminished in the dcLN- and dual-ligation conditions (Fig. 4 C and Fig. S3 A), while scLN-ligation did not differ from sham controls. These results suggest that the dcLN pathway through intracranial meningeal initial lymphatics is the major route for CSF-to-blood outflow at baseline, with the nasal pathway able to provide supplemental capacity when required. After ligation of the cervical lymph nodes, CSF efflux to blood is still significantly impaired, suggesting that additional CSF efflux routes (Louveau et al., 2018; Ma et al., 2019; Wang et al., 2020) are unable to completely compensate.

As a second approach, we measured ICP and outflow resistance in the subarachnoid CSF at the cisterna magna 1 wk after ligation and did not find significant pressure changes in any condition (Fig. 4 D). This suggests that although CSF efflux and turnover are impaired, normal pressure is maintained at least initially. CSF outflow resistance can also be determined through ICP challenge with a constant-rate infusion of artificial CSF while monitoring the pressure response (Andersson et al., 2008; Eklund et al., 2007; Czosnyka et al., 2012; Xu et al., 2021). At each infusion rate, ICP rises until plateauing (Fig. 4 E). At the resulting stable pressure, CSF outflows at the same rate as it is infused, and a higher relative pressure reflects increased outflow resistance. Both dcLN- and dual ligation led to a significant increase in outflow resistance, while scLN-ligation alone did not (Fig. 4 F). Further, post hoc analysis of individual infusion rates revealed that the outflow resistance of the dual-ligation conditions is consistently elevated above baseline across infusion rates, whereas in the dcLN-ligation condition with intact scLN drainage, elevated outflow resistance only emerges at higher infusion rates (Fig. 4 G and Fig. S3 B). Although outflow resistance is often approximated linearly, it is a composite of multiple lymphatic and non-lymphatic outflow pathways that are individually nonlinear (Davson et al., 1970; Boulton et al., 1998; Andersson et al., 2008; Pollay, 2010). We did observe the ICP change to trend linearly with infusion rate (Fig. S3 C); however, the dcLN-ligated group exhibited normal outflow resistance at lower infusion rates before decompensating and mirroring the dual-ligated group at higher infusion rates (Fig. 4 G and Fig. S3 B). These findings reveal physiologically important distinctions between the consequences of systemic lymphatic deficits and dcLN-network deficits alone.

dcLN-afferent cLVs, as well as other visceral cLVs, have previously been shown to exhibit less intrinsic pumping under a range of ex vivo conditions when compared with peripheral cLVs (Gashev et al., 2004; Davis et al., 2012; Scallan et al., 2016; Gasheva et al., 2017; Zawieja et al., 2018, 2019; Yoon et al., 2024). The distinct responses across ex vivo conditions are partially accounted for by differences in the organization and specialization of LMCs. In internodal cLVs, decreased active pumping is explained by the presence of a favorable pressure gradient where flow occurs spontaneously and pumping is counterproductive (Quick et al., 2009). Since dcLN outflow is driven by extrinsic forces, changes in vessel diameter could result in substantial changes in fluid throughput. Indeed, previous studies have indicated that pharmacological manipulation of dcLN-afferent cLV tone modulates tracer outflow (Yoon et al., 2024). However, it was not clear whether this mechanism was part of normal physiology. Our findings suggest that dcLN-afferent cLV tone is actively regulated in physiological conditions. However, further study is needed to understand how this mechanism is controlled and how it can be targeted to modulate outflow.

This study clarifies the functional roles and regulation of the meningeal and nasal lymphatic networks in CNS fluid homeostasis and clarifies mechanisms linking lymphatic dysfunction to pathology. While the dcLN pathway is dominant in baseline conditions, the scLN pathway is recruited to balance pressure and volume perturbations when the dcLN pathway is insufficient, such as in postural changes. Interestingly, scLN-ligation did not prevent CSF outflow into the nasal compartment. The outflow of CSF tracer into blood was delayed in the dcLN- and dual-ligation conditions, with scLN ligation having little effect on outflow to blood. 1 wk after single-ligation of either dcLN- or scLN-afferent lymphatics, or dual-ligation of both, there was no increase in resting ICP; however, dcLN- or dual-ligation does produce elevated CSF outflow resistance.

In the OVA tracer studies, dcLN-ligation recruited the nasal-to-scLN pathway; however, also ligating the scLN in the dual-ligation condition did not negate CSF outflow into the nasal compartment. This suggests that the efflux of CSF across the cribriform plate—not the functionality of the scLN network—defines the ability of the nasal compartment to absorb excess CSF. This distinction helps reconcile the results of previous studies, which found that surgical obstruction of the cribriform plate increased outflow resistance and elevated resting ICP (Silver et al., 2002; Mollanji et al., 2002; Norwood et al., 2019). For CSF to enter the nasal compartment even with deficient lymphatics, the nasal compartment must be able to either store additional fluid or remove it via an alternate route. Our DB53 tracer studies demonstrate that dcLN ligation impairs CSF transit to blood. From our OVA tracer studies, we hypothesized that the nasal pathway would be recruited; however, the additional ligation of the scLN did not further impede DB53 efflux. This suggests that the blood vasculature in the nasal compartment can absorb fluid even if the larger OVA tracer is retained and accumulates in the nasal tissue. This model is also supported by the CSF outflow resistance data, where the scLN network can compensate for dcLN-ligation, but only to a limited extent. Understanding the relationship and differences between these two watersheds may resolve some discrepancies in the field, such as differing reports of the extent of extracranial lymphatic drainage of CSF, which depends strongly on ICP and other technical and experimental differences.

The current study also helps contextualize some of the disparities in the interpretation of certain anatomical findings. Although some lymphatic vessels from the nasal network appear to cross the cribriform plate at areas of arachnoid discontinuity, particularly under neuroinflammatory conditions and thus have constitutive access to CSF (Louveau et al., 2018; Hsu et al., 2022; Spera et al., 2023), the substantial increase in scLN drainage of CSF that occurs in response to physiological challenges likely occurs mostly through bulk outflow into the nasal mucosa and uptake by the much larger nasal lymphatic network.

These data also provide insight into the role of lymphatics in CNS fluid homeostasis. First, CSF outflow resistance is proportional to ICP and the volume of fluid moving through the system (Eklund et al., 2007; Czosnyka et al., 2012). Thus, in the dual-ligation condition, elevated outflow resistance without elevated baseline ICP implies decreased CSF production. Although dcLN ligation also elevated average outflow resistance without elevating baseline ICP, the outflow resistance at low infusion rates is closer to the sham condition, suggesting that CSF production may not be as impaired. Aging has systemic effects on lymphatic physiology, spanning both atrophy of collecting lymphatics and their contractile muscle, as well as regression of meningeal initial lymphatics (Nagai et al., 2011; Da Mesquita et al., 2018). While prior studies primarily used dcLN ligation as a model of lymphatic dysfunction, our current findings suggest that dual ligation may better reflect CNS fluid physiology in aging with the important distinction of decreased, rather than shunted, CSF production and circulation. Indeed, there is experimental evidence of decreased CSF production in aging and neurodegeneration (May et al., 1990; Liu et al., 2020). Moreover, the effects of chronic dual-ligation, and the relationship to pathophysiology observed in aging and disease, is an important area of further study.

Our study provides insight into the fluid pathophysiology in the CNS following lymphatic dysfunction and carries implications for further development of therapeutic strategies targeting lymphatics in neurological disease. Drawing translational inferences from model species to human biology requires care and diligence as the field moves forward. In the mouse, the craniocervical lymphatic system is comparatively simple and provides an excellent system to study intracranial lymphatic uptake alongside perivascular and perineural outflow to peripheral lymphatic uptake. In humans, the head and neck are subdivided into many more lymphatic networks and nodes with substantial anatomical variation between individuals (Pan et al., 2011; Suami, 2017). However, the core features of intracranial and adjacent lymphatic networks identified in model species provide a foundation for future research. Understanding the key variables of lymphatic CNS outflow is also vital to developing clinical assessment tools for CNS lymphatic drainage that can be used diagnostically and perioperatively to minimize surgical injury to lymphatics that are involved in CNS homeostasis (Mehta and Mehta, 2024).

C57Bl/6J (000664; JAX), Myh11-CreER^T2 (#019079; JAX), Salsa6f (#031968; JAX), and Prox1-eGFP (gift from Y.K. Hong [University of Southern California, Los Angeles, CA, USA]) mice were housed in climate-controlled facilities with 12/12-h light-dark cycling. Cages contained five or fewer animals with food and water ad libitum. Unless otherwise noted, experiments were performed at 2–4 mo of age. During surgery, temperature was maintained with a warming pad and thermometer (RT-JR-15; Kent Scientific). Unless otherwise specified, ketamine/xylazine (100/10 mg kg⁻¹) anesthesia was used during surgeries. These experiments and procedures were approved by the Institutional Animal Care and Use Committee at Washington University in St. Louis or the University of Virginia.

I.c.m. injections of tracer were performed through a small (5 mm) vertical incision at the base of the skull. The outermost muscle layer was bisected vertically, and the deeper layers were simply dissected to expose the atlanto-occipital (AO) membrane. Tracer injections consisted of a volume of 5 μl delivered manually over 2 min except where otherwise specified (Ovalbumin-AF594, O34783; Invitrogen; Ovalbumin-AF647, O34784; Invitrogen; 1.0 µm Yellow-Green Fluospheres, F8823; Invitrogen; 0.5 µm Red Fluospheres, F8812; Invitrogen). Rather than a needle, a glass micropipette (1B100-4; Kwik-Fil) was used because we found that the risk of CSF leaking was lower. After the injection was finished, the syringe and micropipette were left in place for 5 min to mitigate tracer reflux when withdrawn. The muscle layers were reapproximated prior to withdrawal as well.

Mice were euthanized with a lethal dose of pentobarbital (Euthasol) and transcardially perfused with heparinized saline. For postmortem analysis, tissues were fixed in 4% paraformaldehyde for 12–24 h. For sectioning, tissues were cryoprotected in a series of sucrose solutions in PBS ending at 30%. Sectioning was done via cryostat (CM3050s; Leica). Brain sections were stained free-floating whereas lymph nodes were stained on gelatin-coated slides. In either case, tissues were washed with PBS, permeabilized with 0.1% Triton-X100 (BP151-500; Thermo Fisher Scientific) in PBS, and blocked with 1% wt/vol BSA (BAH66; Equitech-Bio), and then stained with the indicated antibodies (Lyve1 mAb ALY7 eFluor 570, 41044380; Invitrogen; and eFluor 660, 50044380; Invitrogen) at a dilution of 1:200 in the blocking solution and optionally counterstained with 0.5 μg ml⁻¹ DAPI (D9542; Sigma-Aldrich). Slides were mounted with FluorSave (345789; Millipore) or Prolong Gold (P36930; Thermo Fisher Scientific). For tissue clearing, after fixation, osseous tissues were decalcified with Morse’s Solution (10% sodium citrate [W302600; Sigma-Aldrich] and 20% formic acid [AC410775000; Thermo Fisher Scientific]) in dH₂O. Tissues were decolorized using 25% N,N,N′,N′-Tetrakis(2-Hydroxypropyl)ethylenediamine (122262; Sigma-Aldrich) and then dehydrated with 20/40/60/80/100% methanol and further delipidated with dichloromethane (270997; Sigma-Aldrich). The clearing was done in dibenzyl ether (33630; Sigma-Aldrich), and then samples were incubated in ethyl cinnamate (W243019; Sigma-Aldrich) to equilibrate with the imaging medium.

ICP was measured at the cisterna magna. The sensor (75-0706; Harvard Apparatus) was positioned inside a 1-mm pulled micropipette (21-164-2H; Thermo Fisher Scientific) filled with artificial CSF (aCSF; 59-7316; Harvard Apparatus) in a closed system through several Tuohy-Borst adapters (Qosina) and inelastic tubing (RPT040; Braintree Scientific) with minimal dead volume. The sensor was purged of air and held by a mouse stereotaxic frame via a custom adapter. After a similar preparation to the i.c.m. injection procedure, the AO membrane was wetted with aCSF. The tip of the micropipette was advanced up to (but not touching) the AO membrane, where the acquisition system (Harvard Apparatus, FISO-LS) was initiated in the FISO Evolution software (v2.2.0.0) at 100 Hz with a low-pass filter of 30 Hz. After ∼15 s, the micropipette was advanced into the cisterna magna while observing to ensure that the brain was not damaged, and bleeding or CSF leaking did not occur. Subsequently, the trace was validated for the presence of cardiorespiratory oscillation in ICP as well as stability of the ICP trend, indicating correct placement without CSF leakage. 60 s was typically acquired for ICP measurement after 15-s baseline acquisition.

Constant-rate infusion experiments followed the ICP procedure; however, after the acquisition of the baseline ICP measurement, a syringe pump (130; KDS Legato) was used to deliver a preprogrammed constant-rate infusion program via a 250 μl syringe (7639-01; Hamilton). By running the program in an open atmosphere, we verified that the apparatus did not have significant resistance or compliance of its own. Initial studies were used to determine typical infusion times until sufficient ICP stabilization, leading to a per-step duration of 3,600 s. After the baseline, constant-rate infusion was performed at 0.5, 1, 2, 4, and 8 μl min⁻¹. In preliminary studies, we found that this range was less likely to result in spontaneous leakage or acute complications. Throughout the procedure, animals were monitored for dyspnea, tachypnea, bleeding, CSF leakage, and other indicators. Animals were euthanized after the acquisition.

Additional vital monitoring (ECG, respiration, heart rate, respiratory rate, and temperature) was used, including simultaneously with imaging or ICP measurement (75-1500; Harvard Apparatus). Vital sign readings were synchronized (during imaging or ICP measurement) to the computer clock through the analog outputs using a DAQ (NI USB-6001).

After i.c.m. injection of tracer, animals were placed securely but without anatomical strain on an inclined/declined platform (based on Thorlabs AP180). ICP measurement in this model used different reference pressures per position due to the changing angle of the apparatus water column.

Ligation was performed as previously described (Da Mesquita et al., 2018; Louveau et al., 2018). In short, a midline incision was made 5 mm above the clavicle. For scLN-ligation, all of the medial vessels were ligated with a single filament of nylon, non-resorbable suture thread (Living Systems Instrumentation THR-G). For dcLN-ligation, the sternocleidomastoid muscles were retracted laterally from the salivary gland and then the afferent collecting lymphatics were carefully separated from the fascia. Ligations were placed ∼0.5 mm afferent of the node. For dcLN collecting vessels, typically after the medial and lateral cLVs merge, however, some anatomical variations necessitate ligation of each separately. Vessels were observed to ensure distension above and distal collapse of the vessel. Throughout the surgery, the tissue was protected with warmed 0.9% saline. The mice were then sutured and administered analgesia and prophylactic antibiotics. Mice were then monitored until fully recovered and ambulatory.

10 min after completion of i.c.m injection of 4% wt/vol Evans Blue/DB53 (E2129; Sigma-Aldrich), the left femoral vein was exposed from just distal of the inguinal crease ∼5 mm distally. Warmed 0.9% saline was applied topically to the site and then a small circular coverslip was placed to standardize acquisition and prevent drying of the area. Imaging was collected at 2 frames min⁻¹ for 60 min starting 15 min after the injection was initiated.

Live imaging was performed using a fluorescent stereomicroscope (M205FA; Leica). For imaging of cervical collecting lymphatics and draining beads, animals were positioned supine with precut foam to provide additional support. The head was immobilized against the foam by looping a length of monofilament suture just behind the upper incisors. The foam and tension preserved a slight flexion of the chin toward the chest, avoiding hyperextension of the neck. The forelimbs were also immobilized to the foam using medical tape ∼10–15° ventral and posterior from perpendicular to the core. This positioning is extremely important to favorable and consistent positioning of the relevant tissue planes and anatomy as well as preventing tension and compression on the cervical cardiovascular and lymphatic systems. After site preparation, a 1–2 cm midline incision was made and the skin was carefully dissected bluntly from the underlying tissue. The scLNs and their afferent collecting vessels were then found near the midline at the caudal extent of the salivary gland. To access the dcLN and their collecting vessels, the salivary gland was carefully dissected from the sternocleidomastoid muscles and external jugular vein. This blunt dissection does not require any ligation or cautery of vessels, and bleeding or leakage of lymph should not occur. To prepare the imaging field on a given side, the sternocleidomastoid and omohyoid muscles were gently retracted laterally while the salivary gland was retracted gently towards the midline. Once the dcLN was visualized and afferents were located, traction–countertraction was used to ensure that the retractors, anatomical positioning, and fascial planes were not compromising the blood and lymphatic vasculature. Any discernable damage to cervical collecting lymphatics or nodes, any bleeding beyond tissue weeping, or inability to visualize loose, patent, anatomical lymphatics at any of the sites was used as an endpoint and exclusion criteria.

Eight-bit uniform binning and cropping with minimal exposure were used to optimize acquisition frequency. For particle velocimetry, contractility, and valve imaging, the system was optimized for frame rate with a minimum of 20 Hz. For calcium imaging, exposure time was increased resulting in acquisition at ∼8 Hz.

Lymph nodes were sectioned at 20 µm (deep, superficial, and lumbar combined) and split alternating between two slides. Sections were imaged either by confocal imaging (Leica Stellaris SP8) or widefield epifluorescence (VS200; Olympus). The average resulting number of sections imaged was ∼25 for dcLNs and ∼60 for scLNs due to their respective size. Imaging of cleared samples was collected with light-sheet microscopy (LaVision UltraMicroscope II). Multistack images were stitched with Terastitcher (Bria and Iannello, 2012) and visualized with Napari (Sofroniew et al., 2022).

Raw image analysis was conducted in FIJI/ImageJ (v2.14.10/1.54f) (Schneider et al., 2012; Rueden et al., 2017; Schindelin et al., 2012) using CLIj(2) GPU-acceleration where possible (Haase et al., 2020). Nonstandard image formats were converted to tiff stacks with the Bio-Formats plugin (Linkert et al., 2010). Time series were aligned via the descriptor-based registration package using rigid transformation (Preibisch et al., 2010). Bead tracking was performed using the TrackMate (Tinevez et al., 2017; Ershov et al., 2022) package. Automated tracking was used with initial conditions based on brief optimization, however, due to bead density and velocity, a comprehensive manual review was conducted. For tracer measurements, segmentation was automated with manual review using the DAPI counterstaining, and subsequent measurement was automated. Calcium imaging series were registered and the GCaMP6f mean over the background was measured in a 100 µm circular ROI. After the calculation of ∆F/F₀ by subtracting and dividing the mean per sample, a 15-frame (∼2 s) rolling mean was applied to smooth higher-frequency noise. After registration, the diameter was calculated by thresholding XT transpositions across a section of the vessel. DB53 imaging was measured with a small circular ROI centrally and in focus. DB53 imaging data were centered at the starting value. For outflow resistance measurement, the plateau mean ICP at each infusion rate was calculated and interpolated against the infusion rate. We also compared each group within each infusion rate with a two-way ANOVA and Tukey’s HSD post hoc testing.

Experiments include or are representative of two or more replicates. Group sizes, “n,” indicate the total number of animals, but each data point may represent the average of more than one measurement of that animal. We included both male and female animals. Group distribution, data collection, and analysis were conducted blindly and/or uniformly automated. Data were analyzed in R (4.3.1) with RStudio (2024.04.0+735) using the Tidyverse framework (Wickham et al., 2019). Data is presented as mean ± SEM. Spectral analysis was performed using Lomb-Scargle periodograms in the “spectral” R package. For high-frequency imaging, the actual time intervals of frames were lifted from metadata as they significantly diverged from the stated/nominal times. ICP was calculated as the mean difference between the reference measurement and the stable intracranial measurement. For unpaired comparisons of two groups, a two-tailed unpaired Student’s t test was used. Three or more groups with paired measurements were tested by mixed ANOVA with post hoc simple main effects ANOVA and pairwise t tests using Holm’s method. Comparisons of three or more groups were initially tested via ANOVA, with Tukey’s HSD test for post hoc comparisons. Particularly with the ligation conditions, we also analyzed the groups as a matrix of each ligation site, where the conclusions were similar to treating each group separately.

Fig. S1 provides additional quantitative analysis of the experiments of Fig. 1 including non-normalized analysis of cLV diameter and pumping as well as mean diameter, the spectral analysis plots of dcLN-afferent particle tracking measurements, heart and respiratory rate measurements, analysis of ICP waveforms, spectral analysis plots of lymphatic-valve opening, and frequency and amplitude plots of the LMC GCaMP data. Fig S2 includes ICP traces from the different postural conditions and tracer outflow data into the lumbar lymph nodes across postural conditions from Fig. 2. Fig. S3 provides analysis of the final tracer intensity in blood across ligation conditions, additional post hoc statistical analysis from Fig. 4 G, and a plot of the stable ICP versus infusion rate related to the data in Fig. 4. Video 1 shows representative imaging from the cLV contractility data in Fig. 1 (B–F) and Fig. S1 (A–C). Video 2 shows representative imaging of particle flow in cLVs afferent to the dcLN relating to Fig. 1 (G and H) and Fig. S1 D. Video 3 shows representative imaging of lymphatic valves relating to Fig. 1 (I–K) and Fig. S1 (H and I). Video 4 shows representative imaging of GCaMP6f in LMCs relating to Fig. 1 (L–N) and Fig. S1 (J–L).

Data and code are available upon request.

We thank all members of the Kipnis lab for their continued discussions to push the current study forward.

This work was supported by the Cure Alzheimer’s Fund (BEE consortium) and the National Institutes of Health/National Center for Complementary and Integrative Health R01AT011419.

Author contributions: Z. Papadopoulos: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Resources, Software, Validation, Visualization, Writing - original draft, Writing - review & editing, L.C.D. Smyth: Investigation, Methodology, Writing - review & editing, I. Smirnov: Investigation, Methodology, Resources, Supervision, D.A. Gibson: Writing - review & editing, J. Herz: Resources, Visualization, J. Kipnis: Conceptualization, Funding acquisition, Supervision, Writing - original draft, Writing - review & editing.