We hypothesized that AD + P is the result of a more severe synaptopathy than seen in AD-P, based on multiple studies that have compared indirect measures of synapse integrity both in vivo and in postmortem tissue, between AD + P and AD-P groups 15. In the current report we provide the first direct comparison of the PSD between AD + P and AD-P subjects, finding severe disruption in the PSD proteome in association with psychosis in AD. Relative to AD-P, PSDs from AD + P subjects had lower levels of regulators of Rho GTPases and other proteins regulating the actin cytoskeleton. The PSD protein alterations in AD + P relative to AD-P were present within different subsets of study subjects and were not accounted for by altered burden of neuropathologies or the greater reduction in excitatory neurons that we have previously reported in AD + P relative to AD-P. Examining the pattern of PSD protein changes in AD + P and AD-P relative to cognitively normal elderly comparison subjects revealed additional insights. While both groups showed highly correlated PSD protein alterations relative to comparison subjects, AD-P subjects, unlike AD + P subjects, did not demonstrate an overall reduction in PSD protein abundance. In fact, the PSD of AD-P subjects was enriched, relative to comparison PSDs, for increased levels of proteins involved in protein isomerization and protein quality control, suggesting novel mechanisms of synaptic resilience in AD-P.
Given the prominent role post-synaptic regulation of the actin cytoskeleton plays in the maintenance and plasticity of dendritic spines 40, it is not surprising that the PSD proteome of AD + P was selective depleted of proteins enriched for functions associated with actin regulation. The identification of impairments of this signaling network provides a novel opportunity for identifying compounds that might have novel, specific therapeutic benefits for individuals with AD + P. We used a novel computational strategy, examining the overlap of gene knockout signatures with our AD + P PSD proteome alterations to identify upstream genes that could be pharmacologically targeted by existing compounds to potentially reverse the PSD proteome alterations in AD + P. The predictions we generated using this approach are subject to several limitations inherent in the available data used to generate them. Most studies reporting the effects of gene knockdown rely on an assessment of the transcriptome, rather than the proteome, to characterize the effects of manipulation of these targets. These knockdown studies are most commonly conducted in cells and tissues of non-neuronal origin, and without exception do not assess the PSD. Although we utilized drug candidate signatures from studies obtained in CNS cells, we again relied on transcriptome, not proteome changes, and did not assess PSDs. As a result, our candidates will require validation in the PSD proteome of an intact neural model system. Additionally, we identified reductions in levels of a broad network of kinases. It is likely that not just PSD protein levels, but their phosphorylation state, are altered in AD + P. Thus, both delineation of the phosphoproteome signature of the AD + P PSD, and the impact of drug candidates upon this signature, remains needed.
Nevertheless, our strategy identified several genes that could represent novel targets for addressing the synaptic impairments in AD + P. For example, Cathepsin K (CTSK), which has been primarily studied for its role in bone metabolism, is expressed in neurons 41. Germline knockout of CTSK leads to upregulation of some synaptic proteins, but is also associated with impaired learning and memory 41. How a more moderate pharmacologic inhibition in adulthood would impact the PSD and cognition remains unknown. Aurora Kinase B, AURKB, has recently been shown to play a role promoting neurite formation during early neuronal development 42 and to promote axonal repair after injury in post-mitotic neurons 43. Actin rod formation in neurons, which is promoted by agonism of the C-C Motif Chemokine Receptor 5, CCR5, has been increasingly recognized as a component of Alzheimer’s neuropathology that induces PSD loss 44,45. Although not specifically linked to inhibition of actin rod formation, the administration of CCR5 antagonists in a mouse model of AD reduced cognitive impairment 46. Likewise, treatment with the CCR5 inhibitor, maraviroc, improved cognition in a small study of individuals with HIV-associated neurocognitive disorder 47.
Additional considerations apply in conceptualizing to prioritize among drug candidates at these targets. Repurposing a drug developed for treating other conditions makes it likely that off-target effects will be present and these may be detrimental. Ultimately any medication used will be applied to frail elderly individuals with multiple comorbidities, so preferred agents would be documented to be well-tolerated in such a population. Some of the medications we identified, e.g., fostamatinib, which was developed for chronic use in populations with serious autoimmune disease, and maraviroc, which was developed for chronic treatment of HIV infection, may well meet the tolerability benchmark 48,49. For other identified agents in earlier stages of development, such as MV-711, further evidence is needed, but it should be noted that development of other CTSK inhibitors, including odanacatib, has been discontinued due to adverse events 50.
In our recent genome-wide association meta-analysis of over 12,000 AD subjects with or without psychosis, the most significant genetic locus associated with psychosis in AD was contained in a gene encoding ectonucleotide pyrophosphatase/phosphodiesterase 6 (ENPP6) 11. We now report that ENPP6 is among the 461 PSD proteins with nominally reduced abundance in AD + P relative to AD-P. The expression of ENPP6 is selective for the oligodendrocyte lineage 51, peaking as oligodendrocyte precursor cells (OPCs) differentiate into mature oligodendrocytes 52. It has been established for some time that presynaptic terminals of excitatory neurons containing synaptic vesicles and release machinery align at electron-dense synaptic clefts abutting the processes of OPCs, and that the latter closely resemble the PSDs of dendritic spines, containing a postsynaptic membrane specialization and glutamate receptors (reviewed in53). Though not established, it is likely that these OPC PSDs are co-isolated with neuronal PSDs in our enrichment procedure. Others have demonstrated that synaptic input from glutamatergic neurons onto AMPA and NMDA receptors within the “PSDs” of oligodendrocytes regulate OPC proliferation and differentiation of OPCs to mature oligodendrocytes 54,55, a process during which ENPP6 expression is at its peak, and which is required for learning complex tasks 52. Our finding of reduced ENPP6 abundance in the PSDs of AD + P would therefore suggest that activity dependent maturation of OPCs is impaired in AD + P relative to AD-P, and may contribute to the greater cognitive impairment in AD + P. Whether, and by what mechanism, this may relate to the genetic association of ENPP6 with AD + P risk remains an open question.
We additionally identified an enrichment of genes encoding actin binding proteins, SYNPO, SYNE1, ALDOA, and VPS16, when we examined the overlap of the AD + P PSD proteome signature with genes nominated by our GWAS of AD + P. Two of these, SYNPO and SYNE1, are implicated in processes serving dendritic spine plasticity and maintenance, providing a link between the genetics of AD + P and its biologic signature of post-synaptic deficits. SYNPO encodes the protein synaptopodin, a dendritic spine protein that contributes to maintaining the long-term stability of large dendritic spines 56. SYNE1 encodes multiple actin-binding proteins via extensive alternative splicing. One of these proteins, candidate plasticity gene 2 (CPG2), which is transcribed from the region of SYNE1 associated with risk for bipolar disorder 57, serves as a bridge between the dendritic spine endocytic machinery and the actin cytoskeleton necessary for activity dependent endocytosis of glutamate receptors 58. ALDOA encodes the glycolytic enzyme Aldolase, Fructose-Bisphosphate A protein. While shown to promote lung cancer metastasis by interacting with gamma-actin 59, its role in neuronal function remans obscure. Of interest, however, given our computational predictions, ALDOA protein is found in actin rods 60. VPS16 encodes the VPS16 Core Subunit of CORVET and HOPS Complexes protein, which contributes to autophagosome-lysosome fusion. Mutations in VPS16 cause dystonia syndromes, though the functions of VPS16 in neurons, and the functional consequences of the identified mutations for neuronal and postsynaptic function are not currently established 61.
Individuals with AD-P showed a relative resilience to PSD protein loss, congruent with the widely replicated clinical observations that individuals with AD-P have slower cognitive and functional decline that individuals with AD + P. Our examination of where the PSD proteins that differentiated AD-P individuals from elderly cognitively normal subjects diverged from those of AD + P provides some potential insight into the source of this resilience. PSDs of AD-P subjects were enriched for upregulation of protein isomerases. Knowledge of how peptidyl-prolyl isomerases 62 and protein disulfide isomerases 63 may contribute to prevent tau fibril formation and the generation of Aβ is rapidly increasing. Many of the upregulated isomerases we identified have evidence that they may confer such protection (e.g. PP1A, PTPA, PDIA3) 62,63. FKBP4 provides a more nuanced picture, as it may enhance tau fibrillization while protecting against Aβ generation 62. For other upregulated isomerases we identified (GPI, PGAM1, MMUT, PTPA, PDIA3, FKBP4), no role in these processes is clearly identified.
We report the first study to directly interrogate the PSD in AD + P. Levels of most PSD proteins were substantially reduced in AD + P relative to both AD-P and comparison subjects. Examination of the PSD proteome signature of AD + P, relative to AD-P, subjects revealed lower levels of regulators of Rho GTPases, of multiple kinases, and of other regulators of the actin cytoskeleton, and served to nominate several novel potential pharmacotherapies for the treatment of AD + P. Future studies are required to further assess the ability of these drugs to impact the PSD proteome signature in animal and/or cellular models of Alzheimer disease.