The role of brain-derived neurotrophic factor and the neurotrophin receptor p75NTR in age-related brain atrophy and the transition to Alzheimer’s disease

Introduction

Dementia is a debilitating condition characterised by a deterioration in cognitive functions that mostly affects elderly people (Alzheimer’s association 2017). The number of people with the condition has increased in recent decades and will likely further increase in future as the population ages (Nichols et al. 2019). A common type of dementia is Alzheimer’s disease (AD), characterised by the presence of two proteinaceous deposits in the brain and wide-spread synapse loss (Abner et al. 2017; Mecca et al. 2020). The protein deposits include amyloid plaques composed of the amyloid beta peptide (Aβ) and neurofibrillary tangles (NFTs) composed of hyperphoshorylated tau (Abner et al. 2017). Current treatments for AD temporarily improve cognitive performance by increasing acetylcholine in the synapses of degenerating neurons, but do not stop its progression (Wattmo et al. 2016). Consequently, research into the underlying causes of AD is necessary to identify drug targets with disease-modifying potential. Evidence suggests that pathological levels of soluble Aβ oligomers cause synapse loss early in AD before the onset of other pathologies (He et al. 2019; Moreno et al. 2009; Wang et al. 2017). However, drugs that reduce Aβ have been unsuccessful in clinical trials (Aisen et al. 2011; Egan et al. 2018; Green et al. 2009; Salloway et al. 2014) – possibly because other age-related mechanisms may work synergistically with Aβ. An understanding of these mechanisms may identify drug targets with disease-modifying potential and halt the progression of AD.

The neurotrophins are a family of growth factors necessary for the survival and dendrite formation of specific neuronal populations during brain development (Baydyuk et al. 2013; Rauskolb et al. 2010). The factors may also protect the aged hippocampus, a region involved in learning and memory, by preventing neuronal death and axonal degeneration (von Bohlen und Halbach et al. 2003). Some of the well-known neurotrophins include nerve growth factor (NGF), brain derived neurotrophic factor (BDNF), neurotrophin-3 (NT-3) and neurotrophin-4 (NT-4) (Schulte-Herbruggen et al. 2008). BDNF plays a role in the mechanisms that underpin learning and memory by stabilising excitatory hippocampal synapses and improving their function following stimulation (Kellner et al. 2014; Petkova-Tuffy et al. 2021). Consequently, age-related alterations in BDNF may increase the risk of AD, which initially manifests as memory impairments (Fang et al. 2019; Weinstein et al. 2014). There are many aspects of BDNF that may be considered when assessing its involvement in brain ageing and the transition to AD. Firstly, BDNF transcription is regulated by complex mechanisms within neurons, so age-related changes in these may affect its expression (Palomer et al. 2016). Secondly, BDNF is synthesised as a precursor (proBDNF), which exerts the opposite effect of its mature form (mBDNF), when bound to the neurotrophin receptor, p75NTR (Yang et al. 2014). Age-related imbalances between these two species, or the expression of p75NTR, may facilitate the production of Aβ and/or mediate its toxic effects (Chen et al. 2017; Shen et al. 2019; Yang et al. 2020). Therefore, alterations in the synthesis, processing and signalling of BDNF should be explored when assessing its role in brain ageing and the transition to AD.

This review will explore the evidence on the role of BDNF and p75NTR in brain ageing and the transition to AD. The basic biology of BDNF will be briefly reviewed first before covering the mechanisms of age-related brain atrophy and the transition to AD. This analysis will be critical to examining how age-related alterations in BDNF may increase the risk of the disease. Finally, treatments currently in clinical trials aimed at modifying the course of AD by targeting neurotrophin signalling will be reviewed.

Transcription, processing and biological effects of BDNF

Transcription of BDNF is augmented by neuronal activity in the brain, suggesting it is involved in adaptive responses (Palomer et al. 2016; Singer et al. 2018; Tuvikene et al. 2021). The BDNF gene contains 9 promoters from which transcription may be initiated, each positioned up-stream from a unique 5′ untranslated region (Maynard et al. 2017; Pruunsild et al. 2007). The use of multiple start sites gives rise to various transcripts (transcripts I-IXa), however, because they are spliced to a common coding exon (exon IX), only one protein is translated (Pruunsild et al. 2007). Under basal conditions, transcription is low in cortical and hippocampal neurons, indicating that constitutive BDNF expression is minimal (Maynard et al. 2017; Singer et al. 2018; Tuvikene et al. 2021). By contrast, transcription is increased following depolarisation in cortical neurons and N-methyl D-aspartate (NMDA) stimulation in hippocampal neurons, a stimulus that mimics glutamate neurotransmission (Palomer et al. 2016). Although BDNF transcription is generally stimulus-dependent, the size and duration of the response differs for each transcript. For example, transcripts II and VI increase more than 2-fold 30 min after low-dose NMDA stimulation in hippocampal neurons before returning to basal levels after 60 min (Palomer et al. 2016). By contrast, transcripts I and IV increase only slightly (∼40%) but remain stably expressed thereafter (Palomer et al. 2016). The recent discovery of an enhancer region within the BDNF gene may further complicate this differential transcription (Tuvikene et al. 2021). In cortical neurons, a +3 kb enhancer downstream of exon III was shown to further augment expression of transcripts I–III, but not IV, VI or IXa following depolarisation (Tuvikene et al. 2021). These same effects were not observed in astrocytes, indicating that the adaptive mechanism is specific to neurons (Tuvikene et al. 2021). The differential expression of BDNF transcripts may be explained by their targeting to different dendritic locations for translation. The dendritic complexity of pyramidal neurons from cornu ammonis 1 and 3 (CA1, CA3) of the hippocampus is differently affected by disruptions to specific transcripts in mice (Maynard et al. 2017). Consequently, differential expression of BDNF transcripts may be due to their distinct roles in the formation of hippocampal dendrites.

Stimulus-driven transcription of BDNF depends on the function of elaborate Ca²⁺ dependent and independent signalling pathways within neurons (Cohen et al. 2018; Palomer et al. 2016). Coupling excitation to the nucleus in hippocampal neurons requires Ca²⁺ influx through the activation of NMDA receptor (NMDAR) or L-type voltage-gated Ca²⁺ channels (LTCCs) (Cohen et al. 2018; Wang et al. 2017). Intracellular Ca²⁺ activates calmodulin (CaM), which in turn, binds to and activates Ca²⁺/calmodulin-dependent kinase II (CaMKII). CaMKII is a holoenzyme that consists of multiple subunits, some of which are catalytic and others autoregulatory (Cook et al. 2021). Although multiple isoforms are expressed in the brain, α and β are relatively highly expressed in the hippocampus of adult wild type (WT) mice, suggesting that they are important for learning and memory (Cook et al. 2018). CaMKIIα couples LTCCs to the phosphorylation of nuclear cAMP-responsive element binding protein (pCREB) in hippocampal neurons (Wang et al. 2017), a transcription factor that binds BDNF promoters I, II, IV and VI (Palomer et al. 2016). The enzyme may also function as a molecular switch in hippocampal neurons by sensing changes in stimulation and initiating synaptic alterations in response (Cook et al. 2021). However, CaMKIIγ appears to be critically involved in the stimulus-driven transcription of BDNF, indicating that CaMKII isoforms may have overlapping functions (Cohen et al. 2018). CaMKIIγ translocates Ca²⁺/CaM across the nucleus in cortical neurons, which in turn, activates nuclear CaMKIV, another CREB kinase, following NMDA stimulation (Cohen et al. 2018). Disrupting this translocation in transgenic mice impairs cognitive performance, suggesting that CaMKIIγ is involved in the mechanisms of learning and memory (Cohen et al. 2018). Stimulus-driven BDNF transcription may depend on more than the function of Ca²⁺/calmodulin-dependent kinases (Impey et al. 2002). Inhibition of MAPK reduces pCREB, but not pCaMKIV, in cortical and hippocampal neurons, suggesting that MAPK phosphorylates CREB independent from Ca²⁺ (Cohen et al. 2018; Impey et al. 2002). MAPK is activated by protein kinase A (PKA) and protein kinase C (PKC), which are stimulated by elevations in cAMP (Roberson et al. 1999). The diverse range of pathways that converge on CREB may be due to differential expression of BDNF transcripts. In hippocampal neurons, CREB requires a binding partner, CREB-binding protein (CBP), to initiate transcription at promoters II and VI following NMDA stimulation (Palomer et al. 2016). However, transcription is initiated at promoters I and IV without CBP, suggesting that some transcripts are not as tightly regulated (Palomer et al. 2016). Therefore, the mechanisms which regulate activity-dependent transcription of BDNF may vary, depending on the transcript.

BDNF secretory vesicles are targeted to the regulated secretory pathway in neurons and released in response to synaptic activity (Balkowiec and Katz 2002; Moro et al. 2020). Like other neurotrophins and growth factors, BDNF is translated as a precursor containing an N-terminal pro-domain that is later removed by proteolytic cleavage (Benicky et al. 2019). The pro-domain is required for protein stability (Mowla et al. 2001), regulated secretion (Chen et al. 2005), and possibly also, for protein folding (Rattenholl et al. 2001). Secretion of BDNF depends on the interaction between sorting motifs within the pro- and mature domains of the precursor and intracellular sorting receptors (Chen et al. 2005; Lou et al. 2005). Sortilin, a vesicle-associated sorting receptor (Nykjaer and Willnow 2012), plays a key role in regulated secretion by binding to the pro-domain and targeting proBDNF to secretory vesicles (Chen et al. 2005). Huntington-associated protein 1 may support this process by forming a complex with Sortilin and increasing the efficiency of proBDNF trafficking (Chen et al. 2005; Yang et al. 2011). Regulated secretion also depends on the interaction between a sorting motif contained within the mature domain and carboxypeptidase E (Lou et al. 2005). BDNF is stored in secretory granules that are released in response to depolarisation, suggesting that it is released following synaptic activity (Balkowiec and Katz, 2002; Moro et al. 2020). Similar to gene transcription, this release depends on elevations in intracellular Ca²⁺.

ProBDNF is converted interchangeably by both intracellular and extracellular proteases following stimulation in hippocampal neurons (Pang et al. 2016). Under basal conditions, small amounts of proBDNF are detectable in hippocampal neurons, whereas mBDNF is absent (Pang et al. 2016). Three hours post stimulation, mBDNF becomes abundant, while proBDNF remains unchanged, indicating that stimulus-driven BDNF transcription is followed by increased conversion (Pang et al. 2016). This stimulus-driven conversion likely takes place both intracellularly and extracellularly (Frazzini et al. 2018; Obiang et al. 2011; Wetsel et al. 2013). Intracellular conversion is mediated by furin (Pang et al. 2016), a ubiquitously expressed pro-protein convertase with a diverse range of substrates (Bennett et al. 2000; Dubois et al. 2001; 1995, Kang et al. 2002; Seidah et al. 1996) and also, by pro-convertase 7 (PC7) (Wetsel et al. 2013). Extracellularly, proBDNF is converted by several proteases such as matrix metalloproteinase 9 (MMP-9) and plasmin (Mizoguchi et al. 2011; Obiang et al. 2011). MMP-9 is a zinc-dependent protease involved in tissue remodelling that increases BDNF maturation following excitation in the hippocampus (Mizoguchi et al. 2011). Reducing MMP-9 activity, through sequestration of zinc, impairs learning and memory in WT mice, indicating that it plays an important role in BDNF-related cognitive functions (Frazzini et al. 2018). Another proBDNF protease, plasmin, is converted from its inactive precursor, plasminogen, by tissue plasminogen activator (tPA). tPA is highly expressed in the granule cell layer of the dentate gyrus (DG) and pyramidal neurons of CA1 and CA3 of the hippocampus in adult mice (Stevenson and Lawrence 2018). However, tPA protein is transported to mossy fiber (MF) axons of dentate granule cells in the hilus and the stratum lucidum lamina, where it likely activates plasmin (Stevenson and Lawrence 2018). tPA is required for the early induction of long-term potentiation (LTP) in CA1 of hippocampal slices, but not for ongoing maintenance (Pang et al. 2016). LTP is a form of excitatory plasticity in the hippocampus that leads to a long-lasting increase in synaptic strength (Minichiello et al. 2002). By contrast, furin and/or PC7 are required for ongoing maintenance, but not for initiation (Pang et al. 2016). These findings suggest that intracellular and extracellular proteases interchangeably regulate activity dependent BDNF maturation at different stages of LTP.

ProBDNF may play opposing roles in the brain depending on the relative expression of neurotrophin receptors (Fayard et al. 2005). It has been widely demonstrated using primary neuronal cultures, cell lines and animal models that proBDNF is a neurodegenerative ligand in the CNS (Garad et al. 2021; Liu et al. 2018; Sun et al. 2012; Yang et al. 2014). Heterozygous proBDNF knock-in mice show reduced CA1 hippocampal spines and impaired LTP at Schaffer collateral (SC)-CA1 synapses (Yang et al. 2014). Conversely, in hippocampal neurons, proBDNF facilitates long-term depression (LTD) – a form of synaptic plasticity that shrinks dendritic spines – at SC-CA1 hippocampal synapses (Garad et al. 2021). ProBDNF also collapses neurite outgrowth in primary cortical neurons of the CNS and superior cervical ganglion neurons of the PNS (Sun et al. 2012). In several experimental models, proBDNF mediates these toxic effects through the neurotrophin receptor p75NTR (Liu et al. 2018; Sun et al. 2012). However, it may signal independently from p75NTR at some hippocampal synapses. For example, proBDNF does not require p75NTR to mediate LTD at MF-CA3 synapses, suggesting that it may interact with other receptors in certain cell types (Garad et al. 2021). ProBDNF may also promote neurotrophism in some cells through an interaction with the tropomyosin-related kinase receptor (TrkB) (Fayard et al. 2005). ProBDNF increases neurite outgrowth in the rat pheochromocytoma (PC12) cell line by binding to TrkB and activating the downstream extracellular-regulated kinase (ERK) pathway (Fayard et al. 2005). Therefore, proBDNF mediates neurotoxicity when bound to p75NTR, but may promote neurite outgrowth when bound to TrkB. The relative expression of these receptors in hippocampal neurons, therefore, may determine whether proBDNF weakens or strengthens excitatory synapses.

Mature BDNF stabilises excitatory hippocampal synapses and improves their signalling though its interaction with TrkB following stimulation (Petkova-Tuffy et al. 2021). mBDNF/TrkB signalling plays an important role during the late phase of hippocampal LTP through activation of the MAPK, PI-3 K/Akt and PL-Cγ pathways (Minichiello et al. 2002). These pathways lead to changes in the stability and function of excitatory synapses which collectively improves strength. At the pre-synapse, mBDNF/TrkB signalling increases the release probability of synaptic vesicles, suggesting that it promotes neurotransmitter release (Niculescu et al. 2018; Petkova-Tuffy et al. 2021). At the post-synapse, mBDNF/TrkB signalling promotes structural and neurochemical alterations that stabilise the synapse. These changes may be mediated in part by an interaction with neuroligin 1, a post-synaptic cell adhesion protein that binds pre-synaptic neurexin to stabilise excitatory synapses (Petkova-Tuffy et al. 2021). The protein is reduced in the hippocampus of AD brains, suggesting that it is involved in the pathogenesis of the disease (Dufort-Gervais et al. 2020). mBDNF/TrkB signalling also increases the frequency and amplitude of excitatory post-synaptic currents, indicating that it enhances glutamate neurotransmission (Rauti et al. 2020). Taken together, mBDNF strengthens excitatory hippocampal synapse following synaptic activity in the brain.

The transition from mild cognitive impairment to AD

The transition from mild cognitive impairment (MCI) to AD is characterised by a period of accelerated brain atrophy that exceeds normal ageing (Franke and Gaser 2012; Gaser et al. 2013). Brain volume declines with age in humans and is associated with the progressive cognitive decline that is inevitable with age (Fjell et al. 2014; Veldsman et al. 2021). From 65 years onwards, this decline is more pronounced in the hippocampus – a region involved in learning and memory – suggesting that it may be more vulnerable to ageing (Veldsman et al. 2021). While cognitive decline may be inevitable with age, impaired cognition beyond normal ageing can result in MCI (Abner et al. 2017). MCI is an intermediate stage in the transition to AD defined as an impairment in one or more of the four cognitive domains (Overton et al. 2019). However, MCI is not always a prognosis of dementia as some people with the condition never transition to AD (Fang et al. 2019; Franke and Gaser 2012), while others revert to normal (Abner et al. 2017; Overton et al. 2019). People with stable MCI do not show abnormal brain atrophy and are more likely to have moderate AD neuropathologies (Abner et al. 2017; Franke and Gaser 2012; Gaser et al. 2013; Mi et al. 2017). By contrast, people with progressive MCI undergo a period of rapid brain atrophy before the transition to AD and are more likely to have severe AD neuropathologies (Figure 1) (Abner et al. 2017; Fang et al. 2019; Gaser et al. 2013). The development of biomarkers to distinguish progressive from stable MCI is a desirable outcome to identify people most at risk of AD. However, understanding the molecular basis of this early transition could identify drug targets with the potential to halt or reverse the disease in its later stages.

Figure 1:

Proposed stages in the transition from brain ageing to AD and potential risk factors for progression from one stage to the next.

An age-related neurotrophic imbalance due to impaired activity-dependent mechanisms may increase the risk of synaptic dysfunction, which is an early event in AD (A). However, many people age healthily and only experience mild brain atrophy, so other risk factors such as obesity and diabetes likely contribute to the development of MCI (B). Finally, an over production of Aβ, or a failure to clear it, may be the final stage that triggers synaptic dysfunction, and begins the path towards relentless neurodegeneration (C). Flow diagram created with BioRender.com. MCI, mild cognitive impairment.

Figure 2:

Proposed mechanism of age-related impairments in neurotrophic signalling that may increase the risk of AD in some individuals.

Impaired activity-dependent mechanisms due to ageing may result in a lack of CaMKII responsiveness following stimulation in hippocampal neurons (A). The impaired translocation could lead to delayed, or attenuated, augmentation of BDNF transcription (B). In addition, age-related reductions in the activity of both intracellular and extracellular proteases could impair the rapid maturation of BDNF that normally occurs following excitatory synaptic activity (C). The net result is an increase in proBDNF relative to mBDNF which skews neurotrophic signalling towards LTD and increases the risk of synaptic dysfunction (D). Image created with BioRender.com. BDNF, brain-derived neurotrophic factor; CaM, calmodulin; CaMKII, Ca²⁺/calmodulin-dependent kinase II; CaMKVI, Ca²⁺/calmodulin-dependent kinase IV; CREB, calcium-responsive element binding protein; LTCC, L-type voltage gated Ca²⁺ channel; MMP-9, matrix metalloproteinase 9; NMDAR, N-methyl d-aspartate receptor; TrkB, tropomyosin-related kinase B.

There is considerable evidence that synapse loss early in AD is a consequence of Aβ-induced synaptic dysfunction (He et al. 2019; Lauren et al. 2009; Moreno et al. 2009; Shankar et al. 2007). Reductions in synaptic proteins, dendritic spine density and synapse loss are well established pathologies in post-mortem AD brains (Mi et al. 2017; Proctor et al. 2010; Scheff et al. 2007; Sze et al. 1997). These pathologies occur early in the disease and correlate with cognitive performance suggesting that synapse loss accurately reflects disease progression (Mecca et al. 2020; Sze et al. 1997). Synapse loss early in AD is likely a consequence of synaptic dysfunction mediated by abnormally high levels of Aβ oligomers (Figure 1) (He et al. 2019; Moreno et al. 2009; Shankar et al. 2007). Aβ oligomers are intermediates between peptide monomers and fibrils that culminate in insoluble plaques. Although Aβ oligomers increase synaptic transmission at physiological levels, pathological levels disrupt synaptic transmission and impair hippocampal LTP (He et al. 2019; Lauren et al. 2009; Moreno et al. 2009; Wang et al. 2017). This synaptic dysfunction precedes synapse loss in AD transgenic mice, suggesting that an overproduction of Aβ, or a failure to clear it, is an early event in AD (He et al. 2019). To identify drug targets for AD, therefore, it is critical to understand the molecular mechanisms that cause Aβ to be overproduced or mediate its toxic effects (Figure 1).

Alterations in BDNF transcription and expression during brain ageing and the transition to AD

Age-related impairments in activity dependent BDNF transcription may underpin excessive cognitive decline in rodents (Hongpaisan et al. 2013). BDNF transcription is reduced in the hippocampus of aged rats and in the cortex of aged mice, suggesting that alterations in its stimulus-driven expression occur with age (Belviranli and Okudan 2018; Caughey et al. 2017). Consistent with this interpretation are findings of reduced pCREB, pTrkB and PKCε expression in the aged rodent hippocampus (Calabrese et al. 2013; Hattiangady et al. 2005; Hongpaisan et al. 2013; Smiljanic et al. 2015). Age-related reductions in BDNF transcription may be due to impairments in the mechanisms that couple excitation to the nucleus of hippocampal neurons. The augmented recruitment of CBP to BDNF promoters II and VI following NMDA stimulation is lost in the hippocampus of aged mice (Palomer et al. 2016). This impairment may result from a lack of CaMKIIα activation following Ca²⁺ influx, indicating that stimulus driven BDNF transcription is compromised (Palomer et al. 2016). However, impairments in activity-dependent mechanisms may not occur in all aged animals, which could explain why some are resistant to cognitive decline (Hongpaisan et al. 2013). For example, ERK activation and PKCε expression are reduced in cognitively impaired aged rats, but not aged rats with normal cognition (Hongpaisan et al. 2013; Nagahara et al. 2009). These impairments are reversed by increasing PKCε activity, suggesting that they are caused by a functional loss in activity dependent mechanisms (Hongpaisan et al. 2013). Taken together, impairments in the stimulus-driven transcription of BDNF could explain cognitive impairment in aged rodents.

Transcription of BDNF is reduced in MCI before the transition to AD in humans, suggesting an early loss in activity-dependent mechanisms (Ginsberg et al. 2019). BDNF transcription is reduced in CA1 pyramidal neurons and the hippocampus in MCI before reverting to normal in AD (Ginsberg et al. 2019). This early reduction predicts neuritic plaques in CA1 pyramidal neurons, but not diffuse plaques or NFTs, suggesting that down regulation of BDNF may precede some AD pathologies (Ginsberg et al. 2019). TrkB transcription, on the other hand, is reduced from MCI onwards and correlates with cognitive performance, suggesting that it relates with disease severity (Ginsberg et al. 2019). The early down-regulation of BDNF may be due, in part, to alterations in the subcellular localisation of activated CaMKII (pCaMKII) (Reese et al. 2011). pCaMKII increasingly translocates from dendrites to the soma of granule cells in the DG and CA3 pyramidal neurons in the hippocampus from MCI onwards (Reese et al. 2011). This translocation correlates with the mini-mental state examination (MMSE) scores, suggesting that Ca²⁺ dependent signalling pathway are disrupted during the progression of AD (Reese et al. 2011). Further research in this area may identify other mechanisms involved in the down regulation of BDNF in MCI and its relationship to the progression of AD.

Low BDNF protein levels increase the risk of AD in humans, suggesting that a loss of neurotrophic support occurs early in the disease (Fang et al. 2019; Weinstein et al. 2014). Post-mortem analysis shows reduced BDNF protein levels in brain regions affected by AD, indicating that a loss of neurotrophic support occurs in the disease along with other neuropathologies (Jiao et al. 2016; Lee et al. 2005; Narisawa-Saito et al. 1996). However, these analyses cannot establish whether reduced BDNF protein levels precede the onset of the disease. Prospective studies analyse serum from ageing humans with normal cognition and MCI to determine if reductions in BDNF relate to cognitive impairment and/or the risk of AD (Fang et al. 2019; Weinstein et al. 2014). However, while BDNF may cross the blood–brain barrier, levels in the periphery may not reflect expression in the brain, which limits interpretation of the data. BDNF is expressed by perivascular adipose tissue surrounding arteries into circulation, where it binds receptors on smooth muscle cells (Zierold et al. 2021). This peripheral release of BDNF may mask reductions in the brain, potentially limiting the detection of abnormally low levels. Nevertheless, serum BDNF levels positively correlate with cerebrospinal fluid Aβ42 levels and MMSE scores suggesting that they are related to AD pathologies (Mori et al. 2021). Elderly people (>65 years) with low serum BDNF are more likely to experience worsening cognitive performance than people with higher levels (Fujiwara et al. 2021). In people with MCI, low serum BDNF increases the risk of transitioning to AD, indicating that a loss of neurotrophic support precedes onset of the disease (Fang et al. 2019; Weinstein et al. 2014). These results are consistent with the down regulation of hippocampal BDNF transcription that occurs in MCI (Ginsberg et al. 2019). Taken together, an age-related reduction in BDNF, which may be more severe in some individuals, could increase the risk of AD in humans (Figure 2).

ProBDNF is increased in the aged rodent hippocampus, but it is less clear how mBDNF changes with brain ageing (Buhusi et al. 2017; Perovic et al. 2013; Smiljanic et al. 2015; Wong et al. 2021). It is difficult to establish how total BDNF levels change in the aged rodent hippocampus given the conflicting results of earlier studies (Croll et al. 1998; Hattiangady et al. 2005; Katoh-Semba et al. 1998). Some show that BDNF remains stable (Croll et al. 1998; Silhol et al. 2005), while others show that it increases (Katoh-Semba et al. 1998) and yet others show that it declines (Hattiangady et al. 2005). Nevertheless, with the development of pro and mature BDNF antibodies (Nagappan et al. 2009), most studies now measure each isoform independently (Buhusi et al. 2017; Calabrese et al. 2013; Perovic et al. 2013; Smiljanic et al. 2015; Wong et al. 2021). Increased proBDNF levels have consistently been shown in the aged hippocampus of rodents (Buhusi et al. 2017; Perovic et al. 2013; Smiljanic et al. 2015; Wong et al. 2021), with only one study showing they remain stable (Calabrese et al. 2013). By contrast, proBDNF remains unchanged in the cortex, indicating that these changes are specific to the hippocampus (Table 1). The results for mBDNF are less clear as findings of age-related decreases (Calabrese et al. 2013; Smiljanic et al. 2015), and no changes (Buhusi et al. 2017; Perovic et al. 2013; Wong et al. 2021) have both been reported. Given that maturation of BDNF is dependent on proteolytic processing, it is surprising that age-related changes in the pro/mature BDNF ratio have not been more extensively investigated. Our lab group recently found that proBDNF increases relative to mBDNF with increasing age in the pre-frontal cortex and brain stem, but not in the hippocampus of WT mice (unpublished data). These results suggest that proteolytic processing declines with normal brain ageing, possibly due to age-related decreases in enzymatic activity (Figure 2).

There is limited evidence on how proBDNF relates to AD in humans, but it may be associated with tau pathology (Bharani et al. 2020). The involvement of pro and mature BDNF in the transition to AD is complicated by a technical issue with the ELISA kits used for their detection (Polacchini et al. 2015). The capture antibodies for mBDNF cross-react with proBDNF in some, but not all kits (Lim et al. 2015; Polacchini et al. 2015). However, with recent improvements in the specificity of capture antibodies, both isoforms can usually be reliably measured (Lim et al. 2015; Mizoguchi et al. 2020). Serum proBDNF levels do not correlate with memory scores – as measured by the Rivermead behavioural memory test – in elderly subjects with normal cognition (Mizoguchi et al. 2020). However, the MMSE is often used to detect dementia because performance on this task reflects brain volume (Dinomais et al. 2016). Consequently, it is currently unknown how proBDNF relates to age-related cognitive decline and the transition to AD. Nevertheless, serum proBDNF levels correlate with phospho-tau staining in CA1 of the hippocampus in AD subjects, indicating that it may be related to tau pathology (Bharani et al. 2020). Further studies on how proBDNF changes with age in humans may provide better insights into the molecular basis of the transition to AD.

Age-related increases in hippocampal proBDNF may increase the risk of AD by driving Aβ production and aggregation (Chen et al. 2017). ProBDNF is highly expressed during post-natal brain development and the dominant isoform until early adulthood, when it declines relative to mBDNF (Yang et al. 2014). The high proBDNF levels are accompanied by high expression of p75NTR (Yang et al. 2014), the receptor that mediates its toxic effects (Chen et al. 2017; Sun et al. 2012). Many studies investigate the neurotoxic effects of exogenous proBDNF using neonatal neurons from mice or cell lines (Fleitas et al. 2018; Liu et al. 2018; Sun et al. 2012). However, given that proBDNF is increased in the aged rodent hippocampus (Buhusi et al. 2017; Perovic et al. 2013; Smiljanic et al. 2015), it is critical to understand how endogenous proBDNF contributes to brain ageing and the transition to AD. Endogenous proBDNF facilitates the production of toxic Aβ species in AD transgenic mice indicating that it may have a role in amyloid pathology (Chen et al. 2016). Further research in this area may establish a better understanding of how age-related changes in proBDNF contribute to AD.

The role of p75NTR in age-related brain atrophy and the transition to AD

p75NTR may exacerbate AD by facilitating Aβ production and mediating its toxic effects, but it’s role in age-related brain atrophy is not fully understood (Wong et al. 2021; Xie et al. 2019). p75NTR is widely expressed throughout the developing brain (Hallbook et al. 1990), but then becomes refined to specific brain regions during adulthood (Springer et al. 1990). These regions include cholinergic neurons of the basal forebrain (BFCNs) and neurons and astrocytes of hippocampal CA1, amongst others (Wong et al. 2021). BFCNs develop long projections that extend into a range of cortical areas (Li et al. 2018) such as the entorhinal cortex (McGaughy et al. 2005), amygdala (Jiang et al. 2016) and visual cortex (Pinto et al. 2013). The resulting cholinergic circuity regulates critical brain functions including working memory (McGaughy et al. 2005), visual discrimination (Pinto et al. 2013) and attention. BFCNs are critically important to AD because they degenerate early in the disease and are amongst the worst affected neuronal population (Schmitz et al. 2016). Moreover, their early degeneration predicts the spread of AD neuropathologies, suggesting that the disease originates in BFCNs before emanating to other regions (Schmitz et al. 2016). p75NTR has long been proposed as a mediator of amyloid pathology because it is expressed by BFCNs and facilitates the production of Aβ (Coulson 2006, 2009; Zeng et al. 2011). In addition, p75NTR mediates the down-stream toxic effects of Aβ, such as tau pathology, so alterations in its expression or signalling facilitate the spread of AD neuropathologies (Shen et al. 2019; Yang et al. 2020). However, age-related alterations in the expression of p75NTR and its involvement in the degeneration of BFCNs or other brain regions remain unclear. Although recent research has aimed to bridge this gap in the literature, further work may be required.

Evidence from transgenic mice indicates that p75NTR eliminates developing BFCNs but does not affect their age-related degeneration (Boskovic et al. 2014; Busch et al. 2017; Dokter et al. 2015). p75KO mice are a useful tool for investigating the effects of p75NTR in developing BFCNs and on their survival during brain ageing (Greferath et al. 2000). The original models, p75NTR^exonIII, lacked expression of full-length p75NTR, but maintained expression of a truncated isoform, neurotrophin receptor homolog 2 (NRH2) (Kanning et al. 2003). Using these mice, it was shown that p75NTR eliminates BFCNs during development and that these negative effects are long-lasting (Yeo et al. 1997). However, others found, using genetically-matched controls that p75NTR reduces the size of BFCNs but not their number, indicating that it may cause atrophy but not death (Greferath et al. 2000). Despite these contrasting results, the discovery that NRH2 can modulate Trk signalling – potentially countering p75NTR-mediated apoptosis (Murray et al. 2004) – necessitated the development of a new model. The resulting complete knock outs, p75NTR^exonIV, lacked both p75NTR receptor protein and its truncated isoform (von Schack et al. 2001). Adult p75NTR^exonIV mice show an increased number of BFCNs (Naumann et al. 2002), increased cholinergic fiber density in the amygdala (Busch et al. 2017) and increased hippocampal volume (Dokter et al. 2015; Poser et al. 2015). These same findings were also shown in a conditional p75KO mouse model, suggesting that p75NTR eliminates BFCNs during development (Boskovic et al. 2014). However, the same increases are not observed in aged mice relative to adult, suggesting that the receptor is not involved in age-related BFCN degeneration (Boskovic et al. 2014; Dokter et al. 2015; Poser et al. 2015). A limitation of knock-out mice is that the effects of p75NTR up-regulation during adulthood cannot be investigated. Modulating p75NTR signalling in adulthood or old age, therefore, may provide insights into age-related BFCN degeneration.

p75NTR signalling mediates BFCN degeneration in the aged rodent brain, suggesting that it may contribute to age-related brain atrophy (Xie et al. 2019). Modulation of p75NTR signalling in old age mice (21 months) preserves the size of BFCNs and neurite density of their hippocampal CA1 targets (Xie et al. 2019). These same effects are not observed in moderately aged mice (17 months), indicating that the aged brain is susceptible to p75NTR-mediated neurodegeneration (Xie et al. 2019). The neurodegenerative effects of p75NTR on BFCNs may be mediated by neurotrophic imbalances in the brain regions they innervate. p75NTR triggers apoptosis when bound to all neurotrophins, so alterations in BDNF may only partly explain their degeneration (Friedman 2000). For example, disruptions to the maturation of proNGF, which is retrogradely trafficked to BFCN cell bodies, in the frontal cortex of adult rats causes BFCNs to degenerate (Allard et al. 2018). Further research in this area may lead to a better understanding of the involvement of neurotrophins on the age-related degeneration of BFCNs.

Disease-modifying therapies aimed at restoring the neurotrophic imbalance in AD

Restoring the neurotrophic imbalance is one strategy amongst many aimed at regenerating lost synapses and halting AD progression (Cummings et al. 2020). The number of drugs targeting synaptic plasticity/neuroprotection in clinical trials increased from 2016 to 2020, demonstrating that neurorestoration is an increasingly popular therapeutic approach (Cummings et al. 2020). By contrast, drugs targeting amyloid pathology declined over the same period, indicating that a shift in therapeutic strategies has occurred (Cummings et al. 2020). Amongst neurotrophin drug candidates for AD is an analog of NGF which modulates p75NTR signalling (Simmons et al. 2014; Xie et al. 2019). This drug reduces Aβ-induced synapse loss in AD transgenic mice, suggesting that it has the potential to halt AD progression (Yang et al. 2020). Another approach is to alter p75NTR signalling indirectly by scavenging neurodegenerative ligands, such as Aβ and pro-neurotrophins, with the ectodomain of the receptor (Yao et al. 2015). A chimeric fusion protein of p75ECD and IgG (p75ECD-Fc) reduces amyloid plaque load and improves cognitive performance in AD transgenic mice, suggesting that it alters the course of the disease (Wang et al. 2016; Yao et al. 2015). However, modulation of p75NTR signalling is only one strategy that may potentially restore the neurotrophic imbalance (Figure 3). Another strategy involves promoting neurotrophic signalling by indirectly stimulating the expression of BDNF and regenerating lost synapses (Figure 3). Activation of the hippocampal-specific PKCε with Bryostatin-1 effectively restores BDNF levels, counters Aβ production and regenerates lost synapses in AD transgenic mouse models (Hongpaisan et al. 2011). This drug was shown to improve cognitive performance after completion of dosing in a clinical trial, indicating that it may promote synaptogenesis (Farlow et al. 2019). Further research on the involvement of the neurotrophic imbalance in brain ageing and the risk for AD may identify other therapeutic targets.

Figure 3:

Some therapeutic approaches aimed at restoring the neurotrophic imbalance that may modify the course of AD pathology.

One approach that may alter the course of AD progression involves countering p75NTR-mediated neurodegenerative signalling through either direct or indirect modulation of the receptor (A, B). Indirect modulation involves scavenging neurodegenerative ligands to prevent further amyloid deposition and p75NTR-mediated tau hyperphosphorylation (A). Direct modulation involves using compounds to directly interact with the receptor and alter its signalling (B). Another approach is to stimulate BDNF transcription by increasing the activity of kinases that regulate its activity-dependent transcription (C). This approach may lead to the stabilisation of excitatory synapses and improve their function, thereby slowing or halting AD-related synapse loss. Image created with BioRender.com. BDNF, brain-derived neurotrophic factor; CaM, Calmodulin; CaMKII, Ca²⁺/calmodulin-dependent kinase II; CaMKIV, Ca²⁺/calmodulin-dependent kinase IV; cAMP, cyclic adeno monophosphate; CREB, calcium responsive element binding protein; LTCC, L-type voltage-gated Ca²⁺ channel; MAPK, mitogen-activated protein kinase; NMDAR, N-methyl-d-aspartate receptor; p75ECD-Fc, fusion protein of p75NTR and human IgG-Fc; P, phosphorylation; PKA, protein kinase A; PKCε, protein kinase C epsilon.

Discussion/conclusion

This review has critically evaluated current evidence on the role of BDNF and p75NTR in brain ageing and the transition to AD. As outlined, research on how these neurotrophins strengthen and collapse dendritic spines respectively has led to the development of drugs currently in clinical trials. However, gaps in the literature still exist and, if filled, may lead to the identification of further drug targets. These gaps primarily relate to proBDNF, though the role of p75NTR in brain ageing is limited, but can now be investigated with the advent of molecules that modulate its function (Xie et al. 2019). A key question regarding proBDNF is how it relates to brain ageing in humans and whether it can predict the transition to AD. This question is technically feasible now that ELISA kits specific to BDNF isoforms have been developed (Bharani et al. 2020; Mizoguchi et al. 2020). Given that proBDNF mediates hippocampal LTD (Garad et al. 2021), age-related increases in its expression would imply an increased risk of synaptic dysfunction. However, precise experiments in mouse models of AD may be required to make this link. In summary, BDNF and p75NTR are promising therapeutic targets with the potential to modify, or reverse, the course of AD pathology.