By using immunostaining, we imaged the structure of LVs in both APP/PS1 and WT mice, treated with XST or Vehicle (Veh). Two regions were examined, SSS and TS (Figure 1A), where LVs have been shown to be altered in both AD and during aging (Mesquita et al., 2018). In the SSS region, diameters of LVs in APP/PS1 mice were significantly smaller than those in WT mice (Figure 1B; Lyve-1⁺; Figure 1D; p < 0.01), consistent with previous results on defected LVs in AD brains (Da Mesquita et al., 2018b). The XST-treated APP/PS1 group, compared to the Veh-treated group, had significantly larger LV diameters, a change not observed in WT mice (Figure 1D; p < 0.05). Since LV diameters cannot be reliably measured in the TS region, we measured areas positive for Lyve-1 in TS (% of entire field) (Patel et al., 2019). We found a significant reduction in the Lyve-1⁺ area in the APP/PS1 mice compared to WT mice (Figures 1C,E; p < 0.01), while these areas were significantly larger in XST-treated mice (both APP/PS1 and WT) (Figures 1C,E; p < 0.01).

In addition to LVs, we also measured BVs attached to the meninges. In the SSS region of both APP/PS1 and WT mice (Figure 1B; CD31⁺), we found a significant reduction in the diameters of CD31⁺ BVs in APP/PS1 mice compared to WT mice (Figures 1B,F; p < 0.05), while XST-treated group exhibited significantly larger diameters in both APP/PS1 and WT mice (Figures 1B,F; p < 0.05). In the TS region, there was a significant reduction in the CD31⁺ area in APP/PS1 mice compared to WT (Figures 1C,G; p < 0.01), but no difference was seen between XST- and Veh-treated groups in either WT or APP/PS1 mice. Thus, the impacts of XST on BVs are brain region-dependent.

Previous studies indicated enhanced clearance ability with enlarged LVs (Da Mesquita et al., 2018a; Sun et al., 2018b). To test whether this is the case with XST, we examined Aβ clearance in the brain after XST or Veh treatment by staining Aβ in brain sections containing either SSS or TS region. We found that, in the XST-treated group, Aβ-positive areas are significantly smaller (Figures 2A,B; p < 0.01), suggesting a more effective removal of Aβ from LVs with XST treatment. We also found a significantly smaller Aβ plaque area in the TS region of XST-treated APP/PS1 mice (Figures 2C,D; p < 0.01). Due to the absence of Aβ in WT mouse brains, this analysis was not performed.

Macromolecules in CSF are transported mainly along the dural LVs into lymph nodes of the neck and extracranial systemic circulation (Kisler et al., 2017; Da Mesquita et al., 2018b). It was reported that Aβ is drained from TS into the lymph nodes along the side of the neck (Louveau et al., 2017). We then examined Aβ plaques in regions where LVs merge with BVs in the lymph node (Figure 2E). A significantly smaller Aβ area was seen in the XST-treated compared to Veh-treated APP/PS1 mice (Figures 2E,F; p < 0.01). All together, these results indicate a reduced Aβ plaque size consistent with our previous findings (Huang et al., 2018) and suggest that XST enhances Aβ clearance capacity in APP/PS1 mice.

Another reported consequence of enhanced LV structure/function is an improved inflammatory response shown as higher T-cell density in the SSS region (Kim and Song, 2017). To examine this, we first counted the number of CD3e⁺ T-cells in the SSS and TS regions (Figure 3A). Density of CD3e⁺ cells inside LVs was significantly higher in the APP/PS1 mice compared to WT mice (Figures 3A–C; p < 0.01 for both regions), suggesting elevated immune responses in APP/PS1 mice (Toly-Ndour et al., 2011; Zhu et al., 2019). A significantly higher density of CD3e⁺ cells was seen in APP/PS1 mice treated with XST, compared to the Veh group (Figure 3B; p < 0.01), suggesting a further increase in T-cell density and likely enhanced responses after XST treatment. When the same measurement was repeated in the TS region, we also found a significantly higher density of CD3e⁺ cells in APP/PS1 mice in the XST-treated group compared to the Veh group (Figures 3A,C; p < 0.01). These results suggest that XST may promote immune surveillance. However, the CD3e⁺ cell density in neck lump node (CLNs) was not different between the XST- and Veh-treated groups in the APP/PS1 mice (Figures 3D,E), suggesting that T-cell storage in CLNs is not affected by XST.

The above results suggest an increased T-cell density which is expected to reduce inflammation in the brain. We next examined the serum level of IL-1β (a marker of anti-inflammatory response) (Williams et al., 1999). A significantly lower level of IL-1β was seen in APP/PS1 mice compared to WT mice (Figure 3F; p < 0.01) and a significantly higher level in the XST-treated APP/PS1 mice (compared to Veh-treated) (Figure 3F; p < 0.05). Consistently, serum IL-10 level was significantly higher in the XST-treated APP/PS1 mice than in Veh-treated APP/PS1 mice (Figure 3G; p < 0.05). On the other hand, serum IL-6 level (a proinflammation cytokine) was significantly higher in the APP/PS1 mice compared to WT mice (Figure 3H; p < 0.01) and significantly lower in the XST-treated APP/PS1 mice compared to Veh group (Figure 3H; p < 0.05). In addition, we also found a significantly lower IL-6 level in the brain in the XST-treated APP/PS1 mice (Supplementary Figure S2C; p < 0.05) and a significantly higher IL-10 level in the XST-treated APP/PS1 mice than in Veh-treated APP/PS1 mice (Supplementary Figure S2D; p < 0.01). Thus, these results suggest that the inflammation level is significantly reduced by XST in APP/PS1 mice.

Enhanced responsiveness of both microglia and astrocytes has been widely reported in both AD patients and AD mice (Fakhoury, 2018; Kaur et al., 2019). In the APP/PS1 mice, a larger area occupied by astrocytes (GFAP⁺) was found in APP/PS1 mice (compared to WT mice) (Figures 4A,B; p < 0.01), and GFAP⁺ area was not different between XST- and Veh-treated groups (Figures 4A,B). Interestingly, XST-treated WT mice showed a small but significantly smaller GFAP⁺ area (Figures 4A,B; p < 0.05). In the same mice, the areas covered by Iba1⁺ cells were significantly larger in APP/PS1 mice (Figures 4C,E; p < 0.01) and significantly smaller in APP/PS1 and WT mice treated with XST [Figures 4C,E; p < 0.01 (APP/PS1), p < 0.05 (wt)]. In addition, we found a significantly smaller Aβ plaque density in XST-treated APP/PS1 mice (Figures 4A,D; p < 0.05), consistent with a reduced Aβ level in the previous experiments (Figure 2). Together with the results in Figure 3, XST reduces the activation of astrocytes and microglia which may underlie the altered serum cytokine levels and enhanced clearance of Aβ plaque.

In the above analysis, we conducted experiments at one-time point (15 days after XST treatment). To understand the time course of changes in astrocytes and microglia, we analyzed staining of GFAP and Iba1 at 5 days, 15 days, and 1 month in APP/PS1 mouse brains treated with either XST or Veh (Figure 5A). Interestingly, we found a significantly higher density of Iba1 and bigger GFAP-positive area in mice treated with XST for 5 days (Figures 5A,B for Iba1; p < 0.05; Figures 5A,E for GFAP; p < 0.05). In mice treated with XST for 15 days and 1 month, we found a significantly lower density of Iba1 (Figures 5A,C,D, p < 0.05) and the trend towards reduced GFAP-positive area (Figures 5A,F,G) in the cortex (S1). In addition, we also found significantly smaller GFAP-positive and Iba1-positive areas in the hippocampus of mice treated with XST for 15 days (Figures 5H,I, p < 0.01, p < 0.05). Thus, XST treatment leads to an initial increase in the density of both astrocytes and microglia, which is followed by a decrease in density.

The meningeal lymphatic system is necessary for effective clearance of macromolecules from the brain ISF and may serve as a common pathway for effective removal of wastes from brain parenchyma (Plog and Nedergaard, 2018; Chen et al., 2020). We have shown that XST elevated CBF in both APP/PS1 and WT mice (Huang et al., 2018), and thus elevated CBF could drive the improved LV structure/function. Since ceftriaxone (Cef) injection reliably increases CBF (Kortela et al., 2017; Holschneider et al., 2020), we first confirmed this in the APP/PS1 mice using two-photon linescan imaging (Huang et al., 2018). We have focused on small BVs (diameters 7–15 μm) due to their reported alteration in AD (Ahn et al., 2018; Bracko et al., 2019) and used medium-sized vessels as control (diameters 40–60 μm). The velocity of blood flow in small BVs was significantly higher in the Cef-injected APP/PS1 mice compared to Veh-injected (Figure 6A, p < 0.01), while no difference was seen in mid-size BVs between these two groups (Figure 6B). We then imaged the structures of LVs and BVs in APP/PS1 mice treated with Cef or Veh. Diameters of LVs in the SSS or TS region were not different between Cef- and Veh-injected groups (Figures 6C,D), while the diameters of BVs in the SSS region were also not different between Cef- and Veh-injected group (Figure 6E). Thus, elevating CBF by Cef injection is not sufficient to alter LV structure in APP/PS1 mice.

To understand whether Aβ plaques could be affected by Cef treatment, we measured their density in the primary somatosensory cortex and found no difference between Cef- and Veh-injected APP/PS1 mice (Figures 6F,G), suggesting that elevating CBF by Cef is not sufficient to modulate plaques, in contract to the significant reduction after XST treatment. Since significant alterations in astrocytes and microglia were seen in APP/PS1 mice after XST treatment (Figure 5), we measured the density of these cells in Cef-treated APP/PS1 mice. Microglia density (Iba1) was significantly lower in APP/PS1 mice treated with Cef for either 5 days or 15 days (Figures 6H,I, p < 0.05), but the area of astrocyte (GFAP) was lower in the 15-day but not 5-day treatment group (Figures 6H,J). Thus, compared to XST, Cef treatment does not increase astrocytes and microglia density and thus may have a distinct impact on inflammatory responses. Taken together, although elevated CBF is seen after either XST or Cef in APP/PS1 mice, changes in LVs and Aβ plaques were only seen after XST treatment, suggesting that elevated CBF is unlikely to underlie these differential changes in APP/PS1 mice with XST.

To further understand what could drive or underlie the observed distinct changes associated with XST or Cef treatment, we examined the levels of aquaporin 4 (AQP4) and glial glutamate transporter 1 (GLT-1) since they play important roles in regulating the exchange between brain parenchyma and CSF (Sharma et al., 2019; Simone et al., 2019). AQP4 could influence astrocytic calcium signaling (Yang et al., 2016) and potassium homeostasis and is associated with the removal of interstitial Aβ and modulation of glutamate-mediated synaptic transmission (Lan et al., 2017; Rasmussen et al., 2018) and AQP4 deficiency may also impair learning and memory, in part, through glutamate transporter-1 (GLT-1) (Lan et al., 2016a; Lan et al., 2016b). In addition, GLT-1 reduction and AQP4 alteration were reported in APP/PS1 mice (Takahashi et al., 2015; Benveniste et al., 2019). We found that XST treatment led to a significantly higher GLT-1 level but not AQP4 level in APP/PS1 mice (Figures 7A,B; p < 0.05). The polarity of AQP4 (see Methods for quantification) was significantly higher in APP/PS1 mice compared to WT mice (Figures 6C,D; p < 0.01), which reflects dysfunction of astrocytes (Lan et al., 2017). AQP4 polarity is significantly lower in the XST-treated than in Veh-treated APP/PS1 mice (Figures 7C,D; p < 0.05), a change that was absent in the WT mice. Similar to XST treatment, a significantly higher GLT-1 protein level but no change in AQP4 level was seen in APP/PS1 mice treated with Cef, compared to the Veh-treated group (Figures 7E,F, p < 0.05). However, AQP4 polarity was unchanged in Cef-injected APP/PS1 mice (Figures 7G,H). Since matrix metalloprotein 9 (MMP-9) has been implicated in targeting AQP4 to astrocyte endfeet (Yan et al., 2016; Amtul et al., 2018), we examined the level of MMP-9 in XST- or Cef-treated APP/PS1 mice. A significantly lower level of MMP-9 was seen only in the XST-treated group (Figures 7I,J, p < 0.01), suggesting reduced AQP4 polarity being mediated by reduced MMP-9 level. Furthermore, MMP-2/MMP-9 is known to regulate lymphangiogenesis (Du et al., 2017) and previous studies have found that VEGF significantly increases lymphangiogenesis (Wong et al., 2016). Moreover, hippocampus is an important brain area affected in AD. Thus, we also measured the protein levels of GLT1, AQP4, and MMP9 in the hippocampus and found changes similar to those at the whole-brain level (Supplementary Figures S2A,B). However, we found VEGF-C level not different between XST- and Veh-treated groups in the serum of APP/PS1 mice (Figure 7K), suggesting that VEGF-C-independent signaling pathway may underlie the XST-induced improvement in the lymphatic system.

Meningeal LVs have recently been recognized as an important player in the complex circulation and exchange of soluble contents between CSF and ISF (Elham et al., 2020), and CSF-ISF exchange is a removal mechanism for macromolecules from the brain parenchyma to CSF (Jessen et al., 2015; Plog and Nedergaard, 2018). CSF circulation via the glymphatic system and meningeal LVs has been linked to neurological disorders (Rasmussen et al., 2018), including AD. Da Mesquita et al. (2018b) discovered that meningeal LVs drain macromolecules from the CNS (CSF and ISF) into the CLNs in mice. Aging-associated decrease in paravascular recirculation of CSF and ISF is likely to be responsible for the accumulation of Aβ in the brain parenchyma (Benveniste et al., 2019), and brain-CSF-dural LVs pathway is an alternative route of Aβ recirculation from brain ISF into the CSF sink (Sun et al., 2018b), supported by the presence of Aβ in the CLNs of AD transgenic mice (Karaman et al., 2018). Thus, an impaired meningeal lymphatic drainage of CSF may affect Aβ clearance and exacerbate amyloid burden in AD (He et al., 2017; Da Mesquita et al., 2018a). In addition, mice lacking meningeal LVs show dysfunction in their ability to clear parenchymal macromolecules (Aspelund et al., 2015; Ma et al., 2017). Our current findings of significant Aβ deposition in the meningeal lymphatics of APP/PS1 mice and significantly lower Aβ level in CLNs and meningeal LVs in XST-treated APP/PS1 mice are consistent with these findings. Hence, XST may increase meningeal LVs to remove Aβ from the CSF of APP/PS1 mice. Previous reports have shown that upregulation of dural lymphangiogenesis is associated with significantly reduced senile amyloid plaques in APP/PS1 mice (Karaman et al., 2018; Da Mesquita et al., 2018b), consistent with our finding, whereas we do not have direct evidence for enhanced functions of the LV system or direct evidence for removal of Aβ which is most directly revealed by imaging removal of injected fluorescent Aβ into the brain or real-time monitoring of Aβ clearance before and after XST administration. Due to technical reasons, we were not able to do this reliably. However, both our previous and current results have shown significant reduction in Aβ level or plaque size in the APP/PS1 mouse brains after XST treatment and in reducing levels of both fibrillar and surrounding soluble Aβ (Huang et al., 2018). Thus, we suggest that the improved clearance of Aβ may be related to the improved structures of LVs. In light of the ongoing debate on the functional contribution of LVs (Carare et al., 2020), we need to interpret our results with caution and need more direct examination to confirm the above hypothesis.

The newly discovered meningeal LVs are a novel path for cerebrospinal fluid drainage and represent a more conventional path for immune cells to egress the central nervous system (Louveau et al., 2015; Thomas et al., 2019). We found a significantly higher density of T-cells (CD3e⁺) in meningeal LVs of XST-treated APP/PS1 mice. In addition, IL-10 level was significantly increased and IL-6 level significantly decreased in the XST-treated group, consistent with reducing inflammation by XST (Da Mesquita et al., 2018b; Wang et al., 2019). We found that XST reduces the inflammatory responses in the serum and brains of transgenic AD mice, and this effect may be the result of improvements in LVs. Due to the potential multitarget effect of XST, it is quite possible that the modulation of inflammation is at least partially mediated by targets other than LVs. Reduced activation of microglia by XST may also ameliorate the production of proinflammatory cytokines, nitric oxide, and reactive oxygen species caused by excessive activation of microglia (Dando et al., 2018). It remains to be tested whether the initial elevated immune responses are a trigger for reducing Aβ level which results in subsequently reduced inflammation.

One possibility is that XST increases the proliferation or activity of endothelial cells lining up the LVs. An alternative possibility is the improved LVs as an indirect effect of enhanced CBF. Many signal pathways are known to proliferate lymphatic endothelial cells and improve the lymphatic system function, including growth factors (VEGF, FGF, IGF, etc.) (Ren et al., 2017), Notch1/Akt (Blesinger et al., 2018; Sun et al., 2018a), MAPK-ERK1/2, and Wnt/β-catenin pathway (Zhao et al., 2017; Cong et al., 2018). However, we did not find the VEGF-C level altered by XST treatment. One major in vivo effect of XST is elevated CBF in both AD and WT mice (Huang et al., 2018), and this efficacy has been the main rationale for XST’s medical use in China (Du et al., 2015; Huang et al., 2018). Interestingly, Cef injection did not modulate LV structure and function albeit it significantly enhances CBF (Holschneider et al., 2020). Hence, despite the numerous and complex interactions between CBF and LVs (Kisler et al., 2017), directly enhancing CBF may not affect LVs, suggesting segregation of these two important fluid exchange systems in the brain. This is important not only for understanding the neurodegenerative diseases but also for development of successful therapeutic drugs. In addition, although both XST and Cef treatment lead to reduced inflammation in AD mice after 15 days, their paths are different: XST results in an initial activation of microglia followed by a decrease while Cef only leads to a significant decrease. Interestingly, XST results in a significant reduction in amyloid plaque level while Cef does not (Zumkehr et al., 2015; Fan et al., 2018).

Matrix metalloproteinase-9 (MMP-9) has been implicated in molecular mechanisms of AD (Kaminari et al., 2018), and MMP-9 expression plays an important role in excessive Aβ deposition (Ning et al., 2014). In addition, MMP-2/-9-mediated dystroglycan cleavage in astrocytes is likely to be crucial for maintaining the polarization of AQP4 (Yan et al., 2016; Zhang et al., 2018). The high expression of MMP-9 in the APP/PS1 mice (Kaminari et al., 2018; García-González et al., 2019) likely drives AQP4 to accumulate around amyloid plaques, hence showing high polarity. Glymphatic clearance mechanisms were strongly associated with AQP4 (Yin et al., 2018), and deletion of AQP4 in AD transgenic mice resulted in increased Aβ burden and exacerbated cognitive impairment (Rasmussen et al., 2018). Thus, reduced MMP expression by XST may contribute to the low polarity of AQP4, while Cef (which does not affect MMP-9 level) does not have a similar effect. GLT-1 (EAAT2 in human), the dominant glutamate transporter in the cerebral cortex and hippocampus, is significantly reduced in AD patients and mice (Pereira et al., 2017; Sharma et al., 2019), and increased GLT-1 level by Cef treatment was associated with improved cognitive ability in APP/PS1 mice (Fan et al., 2018). Thus, it is possible that the observed improvement in cognitive functions in XST- or Cef-treated AD mice is at least partially mediated by increased GLT-1 level.

Here, we provided substantial evidence for a widely used Chinese herbal medicine with the potential of improving LVs in AD model mice. Although significant changes have been observed on XST’s impact on BVs and LVs compared to the Veh-treated group, most of the improvements fell short of reversing to the level in the control/WT mice. This could be due to the following: 1) The treatment dose and duration which are not optimal. We have only tested one dose of XST which has been shown in improving cognitive functions and ameliorated AD pathology in APP/PS1 mice (Huang et al., 2018). It is possible that longer treatment may lead to better improvement. 2) The age of the AD mice. We have used AD mice at 11 months old which have developed prominent pathology and impaired functions. Although these mice are more likely to resemble AD patients when they receive medical treatment, the room for improvement is likely to be small. Future studies may test whether dosing XST in younger AD mice or with less advanced AD pathology may result in larger improvement. 3) Limited efficacy of XST as a drug on LV functions. Like other natural products, XST is not expected to exhibit high potency without significant chemical modification on a given biological function, and identification of key components in XST for improved efficacy and testing potential therapeutic path forward may be of interest.