We report a systemic characterization of endogenous NE-releasing LC neurons, their cortical projections, levels of released NE as well as microglial sensitivity to NE in the brains of 5xFAD mice of 4, 6, and 9 months of age, representing early, mid and late amyloid pathogenesis respectively. We found early loss of NE cortical fibers which correlated with reduced cortical NE levels and at later ages reduced cortical NMN levels. LC neuronal loss was only observed in advanced stages of the disease. Isolating microglia from 5xFAD and control mice revealed a striking loss of microglial mRNA expression of β2AR, the receptor chiefly responsible for translating direct NE signaling to microglia 23,24. This loss was especially profound in microglia directly associated with plaques and occurred even at young ages. The synergistic degeneration of the LC-NE system and microglial responsiveness to NE resulted in impaired microglial dynamics in 5xFAD mice. Interestingly, aging alone, in the absence of amyloid pathology, also diminished microglial responses to endogenous NE. We also demonstrated that pharmacological targeting of β2ARs could alter microglial behavior and attenuate plaque deposition despite the decreases in expression of the receptor on microglia. On the other hand, inhibition of this receptor, or specific deletion of the receptor in microglia, worsened plaque pathology, implicating loss of direct signaling of NE to microglia in detrimental outcomes in AD. Our findings suggest the potential to leverage microglial β2AR signaling for AD disease-modifying therapies (working model in Supplementary Fig. 10).
Degeneration of the endogenous noradrenergic system 5xFAD mice
We describe changes in the noradrenergic system of 5xFAD mice at both early and late stages of amyloid pathology. While reductions in cell number and concomitant increases in soma size occurred only in 9-month-old 5xFAD mice (Fig. 1), cortical TH+ nerve fibers appear to degenerate early at the onset of pathological changes (Fig. 2), and may be the substrate for altered endogenous NE signaling in AD at a time when LC neuronal cell bodies are still unaffected. Our studies agree with previous reports in 5xFAD mice 14, and other amyloidosis models 28,34–37, where LC neuron loss is only reported in older animals, and is preceded by neuronal hypertrophy 14,35. The lack of senile plaques in the LC (Supplementary Fig. 1) suggests that the combination of neuroinflammation (Supplementary Fig. 1) and vulnerability of LC neurons 38, either inherently or through the earlier loss of their projections, are chiefly responsible for LC neuronal loss and hypertrophy. We did not observe further loss of cortical TH+ nerve fibers with age, suggesting that compensatory mechanisms may prevent further degeneration as plaque load and inflammation increase. Interestingly, at 9 months of age, WT mice show similar losses in TH+ fibers, suggesting that aging may affect fiber degeneration as potently as amyloid pathology (Fig. 2). Though Cao et al., also reported gradual loss of TH+ projections in WT mice with aging, at 12 months of age, APP/PS1 mice still exhibited more pronounced degeneration of TH+ fibers 28. This discrepancy is likely due to differences between the timing of amyloidosis in the two models.
Since TH+ projections represent axons of both noradrenergic and dopaminergic neurons, we performed ELISA on flash-frozen cortices to examine levels of total NE produced and its metabolite NMN, which is the product of synaptic COMT-mediated degradation of NE 31,32. Interestingly, we observed an early reduction in total NE, but not NMN in 4-month-old 5xFAD mice, which normalized with age (Supplementary Fig. 2), similar to a previous report in aged APP/PS1 mice 28. NMN levels in older 5xFAD mice were lower than that in WT mice although the effect was not significant. We also observed an overall trend of age-dependent increase in cortical NA content, peaking at 6-month-old, although no statistical comparisons were made because animals of different age groups were harvested on different days. Similar findings of age-related NE elevation have been reported in both mice 39 and rats 40–42. Taken together, our findings suggest that there might be multiple compensatory mechanisms to normalize NE cortical levels at different stages of amyloidosis. For instance, early loss of NE projections in the frontal cortex may lower NE release but compensatory electrophysiological changes at the synapse may facilitate NE release or increase NE sensitivity later in the disease. In fact, increased activity of remaining LC neurons in AD brains 43,44 and increased excitability of LC axons in the frontal cortex during aging 41,42 might compensate for the age-related reductions in cortical NE innervation. BDNF is shown to be crucial for maintenance of cortical NE axonal branching and plasticity during aging 42. However, both protein and mRNA levels of BDNF in the brain decrease in AD compared to healthy aging 45–49, suggesting that BDNF might be an important regulator of synaptic plasticity to influence NE release, which is perturbed in advanced AD pathology.
Changes in microglial NE sensitivity in 5xFAD mice
In parallel with changes in adrenergic neurons, microglia also showed changes in their sensitivity to NE. Microglia isolated from 5xFAD brains showed an amyloid pathology-dependent downregulation of β2AR mRNA levels, in line with single-cell transcriptomic data that shows lower β2AR expression in the DAM cluster 26. Interestingly, plaque-associated microglia showed very low levels of β2AR mRNA at all ages, suggesting that β2AR is downregulated early in the transition to a DAM phenotype. The fact that plaque-distal microglia showed intermediate levels of β2AR expression suggests that amyloid pathology impacts microglial adrenergic function even in the absence of direct interaction of microglia and amyloid plaques. It is worth noting that for all in vivo imaging experiments, we used mice with only one functional copy of CX3CR1, which can alter microglial transcriptomic profiles 50 and exert a complex effect on amyloid pathology progression, slowing down disease onset but facilitating progression 51,52. We thus validated our findings in 5xFAD CX3G/+ mice, showing similar downregulation of β2AR in plaque-associated microglia regardless of age and an age-dependent reduction in β2AR levels in plaque-distal microglia (Fig. 4 vs. Supplementary Fig. 4).
While microglia in young WT mice responded to the decrease in NE levels during anesthesia with increased surveillance as previously reported 23,24 (Fig. 3), we show that microglia in both aged 5xFAD and WT mice no longer respond to anesthesia likely due to the age-related degeneration of NE projection fibers. This is an important result suggesting that adrenergic dysregulation of microglial signaling could contribute to the aging-related dysfunctions of microglia reviewed in 53. If NE production does indeed rise with age (as discussed above), these results suggest this increase is not sufficient to compensate for the loss in NE axonal innervations, potentially due to impaired local NE dissemination in the synaptic and extra-synaptic areas. It will be important to monitor patterns of spatiotemporal NE release in vivo, possibly using multi-photon imaging with robust new NE sensors 54 that can provide more reliable measurements than bulk ELISA-based methods. In line with our observations that both NE projections and microglial β2AR expression decrease with amyloid pathology, 5xFAD microglia, especially those associated with plaques, did not respond to anesthesia starting early in amyloidogenesis (Fig. 3). Surprisingly, despite their striking downregulation of β2AR mRNA expression, 5xFAD microglia were able to respond to direct stimulation of the β2AR by the agonist clenbuterol (Fig. 5), suggesting that the lack of response to anesthesia was not solely due to a loss of this receptor but rather a combination of lower NE release and lower sensitivity to NE of microglia. This also suggests that residual β2AR function in microglia could be targeted therapeutically with pharmacological interventions even in DAM, presenting a possibility for adrenergic therapy later in the disease. Additionally, aged microglia in both 5xFAD and WT mice no longer responded to β2AR stimulation despite the presence of high levels of β2AR mRNA in aged WT microglia (Fig. 4, Fig. 5, and Supplementary Fig. 4). This shows a lack of correspondence in β2AR mRNA levels and microglial sensitivity likely due to β2AR protein trafficking or receptor binding affinity/kinetics. Unfortunately, proteomic assays for the β2AR are unreliable, making it hard to address changes to this receptor at the protein level. It is important to also note that 9 months of age is not generally considered old for WT animals, and the age-related microglia cluster has been shown to downregulate β2AR expression much later at 18 months 25. Thus, it is important to profile both β2AR mRNA and protein expression and address the timescale of changes in their expression patterns.
Microglial β2AR signaling attenuates amyloid pathology
The role of NE as a potent anti-inflammatory agent has been extensively studied in various rodent models of AD where ablation of LC neurons pharmacologically 9,10,55, eliminating their NE synthesis and release 56,57, and blocking NE actions with β blockers 11,12 all exacerbate plaque load and increase levels of inflammatory cytokines. Here, we propose a mechanism through which NE modulates AD pathology. Specific genetic deletion of microglial β2AR exacerbated amyloid pathology, especially plaque-associated neuritic damage, although the effect only reached statistical significance in females (Fig. 6; Supplementary Fig. 5). Pharmacological β2AR inhibition showed similar results in female mice (Fig. 6; Supplementary Fig. 5). Perhaps surprisingly, given the low mRNA expression of β2AR in 5xFAD microglia, we demonstrated that chronic treatment with β2AR agonist was protective, resulting in trends towards less plaque load and neuritic damage in both sexes (Fig. 7; Supplementary Fig. 7). This may be due to the different regulation of β2AR protein as compared to mRNA in microglia as described above, but this also raises the possibility that β2AR stimulation later in the disease could provide therapeutic benefit. We showed that extending treatment time from 1 to 2 months conferred significantly more beneficial impact on AD pathology in males but not females, highlighting the importance of defining duration and timing of β2AR stimulation in different sexes for optimal disease-modifying effects.
Our results open new avenues for future studies to investigate microglial β2AR downstream signaling pathways in disease contexts. β2AR and P2RY12 signaling have been proposed to act in a push-pull system and their balance is crucial for the myriad functions of microglia, evidenced in recent research showing their opposing effects on microglial morphology, motility and chemotaxis 23,24,58,59. However, both β2AR and P2RY12 expression significantly decrease with amyloid pathology, making it difficult to parse their respective roles in mediating downstream intracellular G-protein signaling pathways. Microglia make contacts with both inhibitory and excitatory synapses to control their excitability and plasticity 59–61, thus, it is important to characterize microglia-neuron interaction that might be altered in AD in response to diminishing NE or purinergic signaling.
It should also be noted that these findings do not exclude the possibility that the NE-β2AR signaling pathway can also indirectly influence microglia functions to modulate AD pathology. We did not rule out the possibility that peripheral effects of clenbuterol might contribute to modulate microglia function. In the CNS, although expressed at much lower levels compared to microglia 22, astrocytic β2ARs are necessary for hippocampal long-term memory formation and consolidation 62–64, the deterioration of which is a prominent clinical manifestation of AD. Astrocyte and microglia activity are intimately linked in regulating brain development, synaptic transmission, and glial functional states 65, thus, it is possible that β2AR signaling triggers a positive-feedback loop shifting astrocyte and microglia to a reactive state. Nonetheless, indirect signaling could positively contribute to the potential therapeutic benefits of chronic β2AR stimulation.
A vast literature describes sex-specific signaling in AD and its mouse models, making it possible that microglial signaling through the β2AR may lead to different outcomes in males and females. However, we believe that the smaller effects that did not reach significance in male mice after genetic microglial β2AR deletion may be due to the inherent sparse Aβ plaque deposits at this age in males 29,30 in combination with the use of CX3CR1 haploinsufficient animals, in which the onset of senile plaque formation is delayed 51,52, making it harder to quantify changes in pathology. This would implicate disease stage, rather than sex in the different results obtained in the two sexes. Unfortunately, we could not replicate these results pharmacologically, because male mice did not tolerate long-term osmotic pump implantation. This sex-specific vulnerability to surgery may be interesting and should be explored in the future. Pharmacological stimulation of β2AR receptors through daily injections caused similar changes in male and female mice, suggesting that β2AR signaling can be harnessed for therapy in both sexes. However, sex-specific differences in microglial β2AR signaling in AD should continue to be explored.