Loss of Cholinergic Receptor Muscarinic 1 (CHRM1) Protein in the Hippocampus and Temporal Cortex of a Subset of Individuals with Alzheimer’s Disease,Parkinson’s Disease,or Frontotemporal Dementia: Implications for Patient Survival

Abstract

Background:

Dysfunction of cholinergic neurotransmission is a hallmark of Alzheimer’s disease (AD); forming the basis for using acetylcholine (ACh) esterase (AChE) inhibitors to mitigate symptoms of ACh deficiency in AD. The Cholinergic Receptor Muscarinic 1 (CHRM1) is highly expressed in brain regions impaired by AD. Previous analyses of postmortem AD brains revealed unaltered CHRM1 mRNA expression compared to normal brains. However, the CHRM1 protein level in AD and other forms of dementia has not been extensively studied. Reduced expression of CHRM1 in AD patients may explain the limited clinical efficacy of AChE inhibitors.

Objective:

To quantify CHRM1 protein in the postmortem hippocampus and temporal cortex of AD, Parkinson’s disease (PD), and frontotemporal dementia (FTD) patients.

Methods:

Western blotting was performed on postmortem hippocampus (N = 19/73/7/9: unaffected/AD/FTD/PD) and temporal cortex (N = 9/74/27: unaffected/AD/PD) using a validated anti-CHRM1 antibody.

Results:

Quantification based on immunoblotting using a validated anti-CHRM1 antibody revealed a significant loss of CHRM1 protein level (<50%) in the hippocampi (78% AD, 66% PD, and 85% FTD) and temporal cortices (56% AD and 42% PD) of dementia patients. Loss of CHRM1 in the temporal cortex was significantly associated with early death (<65–75 years) for both AD and PD patients.

Conclusion:

Severe reduction of CHRM1 in a subset of AD and PD patients can explain the reported low efficacy of AChE inhibitors as a mitigating treatment for dementia patients. Based on this study, it can be suggested that future research should prioritize therapeutic restoration of CHRM1 protein levels in cholinergic neurons.

Keywords

Alzheimer’s disease β-arrestin Cholinergic Receptor Muscarinic 1 CHRM1 frontotemporal dementia Parkinson’s disease survival

INTRODUCTION

Alzheimer’s disease (AD) is a degenerative brain disease characterized by a progressive decline in memory, cognition, and behavior. Over the years, various descriptive hypotheses have been proposed as causal factors for AD [1]. In the mid-1970 s, comparative analysis of diseased and unaffected human brain tissue-based studies reported substantial neocortical deficits of choline acetyltransferase (ChAT) in AD [2 –4]. ChAT is an enzyme responsible for the synthesis of the neurotransmitter acetylcholine (ACh) [5]. Subsequent studies reported reduced choline uptake by synaptosomes prepared from hippocampal and neocortical neurons derived from AD patients compared to unaffected individuals [6]. Furthermore, it was demonstrated that AD brain slices released less ACh compared to unaffected controls [7]. Loss of cholinergic neurons was observed in the basal forebrain in AD patients compared to unaffected individuals [8]. Cholinergic neurons mainly use ACh for neurotransmission. These observations along with the discovery of the emerging role of ACh in learning and memory [9] led to the establishment of the “cholinergic hypothesis of AD” [10]. It was proposed that structural alterations in cholinergic synapses, loss of specific subtypes of ACh receptors, or death of ACh-generating neurons impaired cholinergic neurotransmission in the cerebral cortex and other areas, significantly contributing to the cognitive decline in AD patients [10]. An outcome of this proposed mechanism of AD was the development of inhibitors of the ACh-hydrolyzing enzyme, acetylcholinesterase (AChE), that could increase synaptic ACh, restoring neuronal communication with nicotinic (nAChR) and muscarinic (mAChR) receptor containing neurons [11]. Based on this strategy several AChE inhibitors were developed for symptomatic treatment of AD. However, the benefit of administration of these compounds was found to be mild, with limited duration of benefit [12]. The exact reason for such failure is unknown but, it likely arises from the near complete loss of cholinergic neuronal function.

ACh signals through two classes of receptors: metabotropic muscarinic receptors (mAChRs) and ionotropic nicotinic receptors (nAChRs) [13]. Muscarinic receptors are G protein-coupled receptors (GPCRs) [14], whereas nicotinic receptors are non-selective, monovalent cation channels [15]. Both mAChRs [16 –18] and nAChRs have been implicated in AD [19 –24]. To understand the limited efficacy of AChE inhibitors for the treatment of AD, this study focused on examining the abundance of mAChRs, specifically a subtype known as the M₁ subtype or CHRM1, in different regions of the AD brain compared to unaffected individuals. Five distinct mAChR subtypes (CHRM1-CHRM5) are known to exist [25]. The distribution of mAChR subtypes in the brain has been mapped using a variety of methods, for example using radioactive ligands [26 –28], in situ hybridization [29], and sub-type selective antibodies [30 –33]. CHRM1 is more abundant in the cerebral cortex and hippocampus compared to other subtypes [34 –36]. Neuropathological studies demonstrated that deposition of neurofibrillary tangles (NFTs), a pathological hallmark of AD, progress in an anatomical hierarchy in aging and AD brains, beginning in the transentorhinal cortices followed by the hippocampus and limbic cortices [37]. The hippocampus, one of the earliest affected brain regions in AD, plays a major role in learning and memory [38]. The transentorhinal cortex is located in the medial temporal lobe [37, 39] and medial temporal atrophy is considered an important metric for the prediction of AD in patients with minor cognitive impairment [40]. These observations raise an important question: does a reduction of CHRM1 in the cortex and hippocampus occur in the majority, or a subset, of AD patients? Severe loss of CHRM1 in a subset of the AD patient cohort would provide a potential explanation for the poor efficacy of AChE inhibitors as an AD treatment strategy.

In this study, we addressed the above-mentioned important biological question by analyzing the abundance of CHRM1 protein in the hippocampus (unaffected: N = 19, AD: N = 73, frontotemporal dementia (FTD): N = 7, and Parkinson’s disease (PD): N = 9) and temporal cortex (unaffected N = 9, AD: N = 74, and PD: N = 27) tissues derived from the brain of unaffected, AD, FTD, and PD individuals, postmortem. In this context, it is important to note the definition of an unaffected individual was based on the neuropathological diagnosis reported by the respective brain repository. This definition may change based on the subsequent neuropathological/molecular examination of the postmortem brains used in this study.

Hippocampus and temporal cortex tissue samples derived from postmortem PD and FTD patients were used to assess the expression of CHRM1 protein in other forms of dementia. PD is generally considered a separate disease entity from AD [41]; however, it has been reported that a cohort of patients displays AD-type cognitive deficits [42]. Thus, it can be hypothesized that a subset of PD patients will also exhibit loss of CHRM1 in the cortex and hippocampus. FTD is a progressive degenerative disease of the frontal and temporal lobes of the brain [43]. AChE inhibitors have not demonstrated clinically significant efficacy in the treatment of FTD patients and it was thought to be due to the lack of cholinergic deficits in FTD [43]. Therefore, we included a small cohort of FTD patients as an additional control group to compare CHRM1 protein levels.

MATERIALS AND METHODS

Postmortem human brain tissues

Frozen (–80°C) and pulverized human hippocampus tissue samples were received postmortem from the National Institutes of Health NeuroBioBank (NIH NBB; Sample request number: 1883), the University of Maryland Brain and Tissue Bank, University of Miami Miller School of Medicine Brain Endowment Bank, Harvard Brain Tissue Resource Center, Sepulveda Research Corporation, and Mount Sinai Brain Bank. Frozen temporal cortex and cerebellum tissue samples were obtained from the University of Florida Neuromedicine Human Brain and Tissue Bank.

Chrm1 knockout mouse

The Chrm1 knockout mouse (Chrm1^-/-) line was provided by Dr. Jurgen Wess, NIH. The brain tissues from Chrm1^-/- and age-matched wild-type mice were provided by Dr. Paul Fernyhough, University of Manitoba.

Western blotting and quantification

Quantitative analysis of protein abundance by using western blotting for a large number of samples involving different experimental conditions and multiple replicates requires careful design to ensure precision and accuracy [44]. The signals derived from the protein bands on a western blot vary with the amount of total protein lysates loaded onto the protein gel. Excessive loading of protein can compromise the linearity of the detection system [45]. Therefore, we calibrated the amount of sample loading within the linearity of the detection system by performing an immunoblot using 20, 40, 60, 80, and 100μg of total protein per well (Supplementary Figure 1A–C) [46]. The optimal total protein loading, determined from the calibration curve, for the detection of CHRM1 (Catalog number: sc-365966; Santa Cruz Biotechnology: SCBT) and ARRB1/2 (Catalog number: sc-74591; SCBT) proteins in the western blotting was 50μg/well, whereas the optimal loading used for detection of the relatively more abundant glyceraldehyde-3-phosphate dehydrogenase (GAPDH; Catalog number: sc-47724; SCBT) protein was 25μg/well (Supplementary Figure 1A–C).

Immunoblotting experiments were performed using 25–50μg of total proteins loaded in each well of the SDS-PAGE gels. The protein lysates were prepared in 1×PBS buffer supplemented with 1% IGEPAL 630, 1×Halt protease, and phosphatase inhibitor cocktail (ThermoFisher Scientific, Cat No: 78441). The protein lysates were denatured in Laemmli buffer containing 2% SDS, 10% glycerol, 0.002% bromophenol blue, and 0.75 M Tris-HCl pH 6.8 supplemented with 100 mM DTT. The lysates were heat-denatured at 55°C for 10 min. We adopted an innovative approach of sample loading to minimize the number of immunoblots required to analyze ~100 tissue samples. We used a triple-wide (33 cm wide and 10 cm height) vertical western blot electrophoresis apparatus (catalog no: MGV-202-33; C.B.S Scientific) to accommodate 52 wells in a single gel. In this gel system, we loaded 24 tissue samples (48 wells) in duplicates along with four wells used for protein molecular weight ladders. A reference control tissue sample (sample identification numbers for hippocampus and temporal cortex tissues: #4456 and A21-006, respectively) was consistently loaded in each batch of immunoblots and was subsequently used as a reference (100%) for the normalization of the protein of interest in different individuals. This also allowed comparison between different immunoblots containing different batches of samples and variable experimental conditions. Following electrophoresis, the gel was excised to accommodate all bands within a desired molecular weight range, for example, 100-40 kDa range for detecting CHRM1 protein, placed on an 8.5 x 13.5 cm 0.2μm pore size nitrocellulose membrane for rapid semi-dry transfer of proteins using Bio-Rad Trans-Blot Turbo Transfer System. The membranes were blocked using EveryBlot Blocking Buffer (Bio-Rad, catalog number 12010947). The immunoblots were detected using chemiluminescence and imaged using a Bio-Rad ChemiDoc MP Imaging System (for the hippocampus samples). The immunoblots for the cortex and cerebellum samples were detected using Odyssey® XF Imaging System (LI-COR Biosciences).

All immunoblots were quantified using ImageJ (version 1.48) Software [47]. Relative quantification of the proteins in the western blot was performed as described previously [48]. In recent years, total cellular protein staining has emerged as a preferred method for normalization of sample loading in quantitative analysis of western blot data [44]. We used total cellular protein staining as a method for the normalization of the target protein of interest that was extensively validated and discussed in our previous publication [48]. Total proteins were stained with either Oriole (Catalog number: 1610495; Bio-Rad) or SYPRO Ruby (Catalog number: 1703125; Bio-Rad) fluorescent stains as per the manufacturer’s instruction and imaged as described previously [48]. The protein-of-interest bands within a single immunoblot were first normalized by dividing the band intensities with the intensities of the corresponding Oriole/SYPRO Ruby-stained protein bands or intensities. Next, the normalized band intensity of the protein-of-interest in the reference sample (X) is converted to 100% by using the formulae (X/X*100). The protein-of-interest bands in the other samples within the same immunoblot were then converted to a percent of the reference sample by using the formulae (Y/X)*100, where “Y” is the normalized band intensity of the protein-of-interest in the target sample. This allows comparison between different immunoblots derived from independent experiments/biological replicates. Besides, this approach overcomes the inherent problems faced when using housekeeping proteins as loading controls, thus permitting the user to obtain truly quantitative western blot data by normalizing bands to the loaded proteins. However, we immunoblotted GAPDH as a housekeeping protein in the temporal cortical tissues as an additional control to validate the loss of CHRM1.

Statistical analysis

Statistical analysis was performed using Prism software version 7.00 (GraphPad Software). In this study, the expression levels of the protein of interest in the unaffected and diseased groups were measured and presented as the percentage of the normalized expression level in a reference unaffected sample (100%). This allowed comparison between different immunoblots derived from independent experiments/biological replicates [48]. An unpaired parametric t-test was used for two group comparisons. When there were 3 or more groups, a one-way ANOVA (randomized) was used [49, 50] followed by Dunnett’s post hoc multiple comparison test to determine differences between the experimental groups [49]. The differences were considered significant with p < 0.05 and throughout the text, if a p-value is≤0.05,≤0.01,≤0.001, or≤0.0001, it was flagged and represented with one, two, three, or four asterisks, respectively. The Pearson correlation test was performed considering that the protein expression and age of death values were sampled from populations that approximately follow a Gaussian distribution. The correlation coefficient (r) ranges from –1 to +1. A perfect correlation has r = 1, r = 0 to 1 indicates that the two variables tend to increase or decrease together, r = 0 indicates that the two variables do not vary together at all, and r = –1 to 0 indicates one variable increases as the other decreases, whereas r = –1 indicates a perfect negative or inverse correlation.

RESULTS

A 70 kilodalton (kDa) N-glycosylated CHRM1 protein is present in the hippocampus and temporal cortex, but not in the cerebellum

We chose an anti-CHRM1 monoclonal antibody raised against an epitope within the amino acids 231–350 of human CHRM1. This sequence is conserved in different mammalian species, for example, human, pig, and mouse. The mAChR subtype specificity of this antibody was previously validated [51]. In this study, we further validated this antibody using three different approaches. The first approach involved the use of a human brain tissue sample that does not express CHRM1 and was considered a negative control. The second approach involved the use of Chrm1^-/- mouse [52] brain tissues. The third approach involved enzymatic removal of the post-translational modifications of CHRM1 to demonstrate a shift in the molecular weight of CHRM1 that was detected by the anti-CHRM1 antibody. We searched the human Genotype-Tissue Expression (GTEx) database, identifying the cerebellum as having a very low CHRM1 mRNA expression value compared to the cortex and hippocampus tissues (Fig. 1A). The median transcripts per million value for CHRM1 in the cerebellum is 0.198, whereas it is 55.18/52.43 and 18.56 in the cortex/frontal cortex and hippocampus, respectively (Fig. 1A). Immunoblotting using the chosen antibody revealed the presence of a ~70 kDa band in the hippocampus and temporal cortex tissues that was absent in the cerebellum as expected based on the GTEx gene expression data (Fig. 1B–F). We reproduced this finding using pig hippocampus, cortex, and cerebellum tissues documenting that this brain region-specific expression pattern of CHRM1 is conserved in another mammal (Fig. 1G–K). Thus, the use of the cerebellum as an additional negative control validated the epitope specificity of the anti-CHRM1 antibody. Next, immunoblotting using wild-type and Chrm1^-/- mouse brain tissues revealed the presence of a p75 anti-CHRM1 positive band in the wild-type cortex and hippocampus (Supplementary Figure 1). The p75 bands were absent in the corresponding brain regions of the Chrm1^-/- mouse (Supplementary Figure 1D,E, black arrow), again validating the epitope specificity of the anti-CHRM1 monoclonal antibody.

Fig. 1

CHRM1 mRNA and protein expression in the cortex, hippocampus, and cerebellum of the human and pig brains. A) Violin plot showing CHRM1 mRNA expression in primary human brain tissues. TPM, transcripts per million. The data used for this analysis was obtained from the GTEx portal on June 6, 2022. Sample size: cerebellar tissues (N = 251), cortex tissues (N = 255), hippocampus (N = 197), and spinal cord (N = 159). Box plots are shown as median, 25th, and 75th percentiles. B, C) Representative images of coronal sections of the human brain showing the regions (yellow oval/red rectangles) used for the collection of the hippocampus (B), temporal cortex, and cerebellum (C), samples, postmortem. cm, centimeter. D, E) Immunoblots showing the abundance of CHRM1 (D) and GAPDH (E) in the hippocampus, temporal cortex, and cerebellum samples derived from different human subjects. The human subject’s identification numbers are written on the top panel. The asterisk indicates a neonatal brain. Blue and red highlights indicate female and male sex. NS, non-specific bands. F) SYPRO Ruby-stained SDS-PAGE gel showing total protein loading corresponding to the immunoblot presented in D and E. G, H) Images showing the areas (black rectangles) used for the collection of the temporal cortex and cerebellum (G), and hippocampus (H), from the domesticated pig brain. I, J) Immunoblots showing the abundance of CHRM1 (I) and GAPDH (J) in the hippocampus, cortex, and cerebellum tissue samples derived from domesticated pig brain. The green bracket indicates the absence of CHRM1 in the cerebellum. K) SYPRO Ruby-stained SDS-PAGE gel showing total protein loading corresponding to the immunoblots presented in I and J. L, M) Diagrammatic representation of the crystal structure of CHRM1 protein showing N-glycan attachments and the strategy for enzymatic removal of the N/O-glycans from the receptor. The crystal structure of CHRM1 was based on AlphaFold prediction [106]. M) Immunoblots showing the presence of CHRM1 and Tubulin Beta 3 Class III (TUBB3) proteins in different glycosidase-treated hippocampal tissue lysates.

The theoretical molecular weight of CHRM1 is 51 kDa; however, immunoblotting revealed the appearance of 70–75 kDa bands, which indicates the gain of additional molecular mass that could only be possible due to the post-translational modifications of the protein. Phosphorylations [53], glycosylation [54, 55], and SUMOylation of CHRM1 [56] have been reported. A single phosphorylation event adds 79.9 Da molecular mass. Post-translational addition of a glycan, for example, N-glycolylneuraminic acid (sialic acid) adds 307 Da mass, whereas, the addition of a SUMO protein may cause a 15–17 kDa shift in SDS-PAGE [57]. We treated hippocampal protein lysate with PNGase F or a cocktail of O-glycosidase and neuraminidase to remove either N glycans or O-glycans that may be responsible for the appearance of the high molecular weight form of CHRM1 (Fig. 1L). Immunoblotting revealed a shift of hippocampal p70 CHRM1 to a p60 kDa band in the PNGase F-treated fraction (Fig. 1M) indicating that the p70 CHRM1 in the brain tissues is N-glycosylated. The recognition of the deglycosylated p60 CHRM1 by the anti-CHRM1 antibody indicates further validation of the antibody. In this context, it is important to note that our finding of a 70 kDa N-glycosylated CHRM1 in the human brain has been previously reported [58 –60]. The p70 CHRM1 was also reported in the digitonin soluble fraction of the pig brain and was found glycosylated due to the binding to glycan-binding lectin proteins [58]. Furthermore treatment with endoglycosidase F, an enzyme that removes both complex and high mannose type N-linked carbohydrate chains reduced CHRM1 from 92 kDa to 77 kDa in human glioblastoma cells (1321N1) and from 66 kDa to 45 kDa in hybrid mouse neuroblastoma/rat glioma cells (NG108-15) [59]. Furthermore, Ohara et al. reported that endoglycosidase F treatment of pig CHRM1 protein purified from the cerebrum decreased its apparent molecular weight from 70 kDa to 51 kDa [60], thus supporting our findings that PNGase F treatment also caused a 10 kDa reduction in the apparent molecular weight of CHRM1. The discrepancy between our study and Ohara et al.’s study may be due to the use of different glycosidases. Endoglycosidases F enzymatic activity is less subject to protein conformation limitations compared to PNGase F, thereby providing greater protein deglycosylation [61]. Therefore, it is possible that PNGase F was not effective in removing all glycans from CHRM1; however, the relative shift detected by the anti-CHRM1 protein validated the epitope-specificity of the antibody. Overall, these results validated the specificity of the anti-CHRM1 antibody and indicated that an N-glycosylated form of CHRM1 is abundant in the hippocampus and cortex of the human brain, but not in the cerebellum.

Hippocampal loss of CHRM1 in a subset of individuals originally designated as unaffected individuals by the brain repositories

We collected 19 hippocampus tissue samples (13 from the Mount Sinai Brain Bank and 6 from the University of Maryland Brain Bank) designated as unaffected controls (Supplementary Table 1). Among these samples, one hippocampus tissue sample was collected from an infant (NIH NBB identification number-ID: 4456) with the age of death recorded as 68 days. Though the infant’s cause of death was not reported in the NBB database, it was assumed that a neurodegenerative abnormality at this age would not be expected, therefore, the CHRM1 protein level in the hippocampus of this sample was used as a reference (100%) to quantify all proteins of interest in all other samples.

Quantification based on immunoblotting revealed a severe reduction of hippocampal CHRM1 protein in a subset of unaffected individuals (Fig. 2, Supplementary Figures 2–4). A majority (73%, N = 14 out of 19) of the unaffected individuals with an average age of death recorded as 69.39±17.35 years (mean±SD; range 24–86 years) exhibited low levels (<50%) of hippocampal CHRM1 protein (mean±SD: 26.9±16.4 %) (Fig. 2A, Supplementary Figures 2A and 3A). The remaining five unaffected individuals, including the reference standard: (IDs: 4456, 47819, 5337, 638682, 5346) with an age of death reported as 0.18, 78, 28.4, 64, and 24.9 years, respectively exhibited≥50% mean hippocampal CHRM1 protein levels (100%, 109.9%, 85.9%, 82.1%, 72.2%, respectively) (Fig. 2A, Supplementary Figures 2A and 3A). The reduction of CHRM1 protein within a subset of individuals marked as unaffected may be due to an age-related phenomenon or some of those unaffected individuals may have had an underlying disease condition that was not reported to the NBB. In support of this view, it can be pointed out that within the five unaffected individuals with≥50% hippocampal CHRM1 protein levels, one individual (ID: 47819) had a clinical dementia rating (CDR) [62] score reported as 2 and Braak score reported as II [63], the latter indicates the presence of NFTs. On the other hand, in another unaffected individual (ID: 638682) within the same group, the Braak score was reported as stage I, an indication of NFTs in the hippocampus in the transentorhinal region. Furthermore, within the group of unaffected individuals with < 50% hippocampal CHRM1 protein levels, six individuals (IDs: 57242, 298992, 71029, 70468, 707617, and 82597) were in Braak stage I, one individual (ID: 2482) in Braak stage II, one individual (ID: 29972) in Braak stage III, with another individual in Braak stage VI, indicating different degrees of NFTs accumulation in their hippocampi (Supplementary Table 1). Furthermore, six individuals within this group of 14 unaffected individuals had an Apolipoprotein E (APOE) allelic composition [64] documented as APOE3/4 and one had APOE4/4 allele. The APOE4 allele is associated with an increased risk and earlier age of onset of dementia [65]. Overall, these results highlight the limitation of obtaining neuropathologically normal unaffected postmortem brains through existing brain repositories.

Fig. 2

Abundance of CHRM1 and ARRB1/2 proteins in the hippocampus of unaffected individuals and AD/FTD patients. A, B) Immunoblots showing the relative abundance of CHRM1 (A) and ARRB1/2 (B) in hippocampus tissues derived from postmortem human brains. The cyan arrow indicates that the corresponding immunoblot was incubated with anti-ARRB1/2 without stripping. The sample identification numbers are provided on the top panel of each immunoblot, Male and females are highlighted in red and blue, respectively. The unaffected, AD and FTD samples are color-coded green, pink, and blue, respectively. The PMIs for the individuals are displayed in minutes. C) Oriole-stained gel showing the total protein loading for the samples corresponding to the immunoblots presented in A and B. The red dotted rectangle indicates the differential abundance of a p25 kDa unknown protein in the AD and FTD brains and its absence in the unaffected brain.

ACh binding to the CHRM1 causes G protein activation (Gq/11 subfamily of G proteins) followed by the recruitment of β-arrestins (ARRB1/2), the latter contributes to CHRM1 desensitization, endocytosis, and subsequent degradation or recycling of the receptor back to the plasma membrane [66]. Therefore, we studied the abundance of ARRB1/2 as a measure to determine if the loss of CHRM1 is associated with a concomitant loss of CHRM1 signal transduction components, for example, ARRB1/2. Interestingly, the unaffected individuals (IDs: 1848, 5349, 912489, 57242, 55015, 38736, 29972) with very low CHRM1 protein levels (35.4%, 26.9%, 18.6%, 15%, 12.5%, 5.9%, and 5% respectively) retained more or less normal levels of hippocampal ARRB1/2 (86.9%, 76.1%, 101%, 90.6%, 118.4%, 80.9%, and 80.5%) protein (Figs. 2A–D, and 3A–C Supplementary Figures 2A–D). This indicates that the loss of CHRM1 protein in the unaffected individuals was not associated with an equivalent loss of the CHRM1 downstream signaling component.

Postmortem change of proteins in tissue samples is an important factor for immunoblotting-based quantification of target proteins [67]. Comparison of the protein level alterations in rat brain kept at 23°C for different postmortem times up to 72 h revealed that most differences occurred after 24 h or even after 48 h postmortem [68]. In our study, the mean postmortem interval (PMI) for the hippocampus samples was 10.2 h (SD: 6.4 h; range: minimum 1 h –maximum 27.5 h), and for the temporal cortex samples, was 11.8 h (SD: 8.2 h; range: minimum 2 h –maximum 56.5 h) (Supplementary Tables 1 and 2). Visual comparison of the Oriole-stained total protein profiles of the hippocampus samples exhibited some difference in the banding patterns, for example, a p 25 kDa band was absent in the neonatal hippocampus (ID: 4456), whereas the intensity of this band was reduced in some samples (ID: 5337, 5346, 4668, 4721, 4743) compared to others (ID: 1848, 2901, 3026, and 3291) (Fig. 1C, D). It is difficult to characterize whether the difference in the intensity of the p 25 band was due to protein degradation or developmental stage-specific differential protein expression (neonatal brain versus adult brain), or pathological alteration due to the disease state. However, a comparison of the CHRM1 and ARRB1/2 protein levels with PMI revealed that the alterations in the relative level of these proteins were not correlated with PMIs. For example, the hippocampal samples –1848 and 5337 have the same PMI (540 min) but their mean CHRM1 protein levels are 35.3% and 85.9%. Therefore, this difference cannot be attributed to postmortem protein degradation (Figs. 1A and 3A). On the other hand, the samples: 5349, 5790, 3939, and 6107 have PMIs (minutes) of 960, 1350, 960, and 1440 respectively, whereas their mean CHRM1 protein levels are 26.9%, 61.9%, 67.4%, and 9.7%, respectively (Figs. 1A and 3A). This suggests that CHRM1 protein levels were not affected by PMIs.

Fig. 3

Abundance of CHRM1 and ARRB1/2 in unaffected, AD, FTD, and PD hippocampi. A) Bar graph showing the relative amount (%; mean±SD) of CHRM1 and ARRB1/2 proteins in the unaffected (N = 19), AD (N = 73), FTD (N = 7), and PD (N = 9) hippocampi. Four replicates from two independent western blot experiments were used to generate the table. B, C) Scatter plots showing CHRM1 and ARRB1/2 proteins level (%, mean from A) in unaffected, AD, FTD, and PD hippocampi. The red and blue bars represent the mean and SD in this figure and subsequent figures, respectively. D) Scatter plot showing the CHRM1 proteins level (%, mean from A) in male versus female AD patients. E) Bubble plot showing the hippocampal abundance of CHRM1 (%, Mean from A) combined with the sex and age variables of the AD patients. F) Bubble plot showing the hippocampal abundance of CHRM1 and ARRB1/2 proteins (%, mean from A) combined with the sex and age variables of the AD patients. G) Scatter plot showing the hippocampal CHRM1 protein levels (% mean from A) in different age groups of AD patients.

Severe loss of hippocampal CHRM1 in a subset of AD, PD, and FTD patients

Immunoblotting revealed that 21.9% (16/73) of AD patients exhibited≥50% (mean±SEM: 65.6±3.9; N = 16) CHRM1 protein level, whereas 78% of AD patients had < 50% (mean±SEM: 19.7±1.9; N = 57) CHRM1 protein level in the hippocampus (Figs. 2 and 3A, B, Supplementary Figures 2–5). In contrast, the hippocampal ARRB1/2 protein levels (mean±SEM) in these two groups of AD patients were 117.7±10.53 and 72.8±4.2, respectively. Though the ARRB1/2 levels were comparatively reduced in the group of AD hippocampi containing low levels (<50%) of CHRM1 protein compared to the group of AD hippocampi with higher levels (≥50%) of CHRM1, the extent of ARRB1/2 reduction was not similar to the extent of CHRM1 loss (Fig. 3B–E). This is evident in the AD hippocampi samples (IDs: 889449, 6151, 850463, 26968, 29466, BEB18141, 83778, 3026, BEB19036, 42537, 639732, 22142, 80772, 1625, BEB19105, 3291, BEB19144, 5854, 1626, 2901, BEB18129, 514468, 5955, and 6143) with very low (<10%) CHRM1 protein level (9.7%, 9.6%, 9.4%, 9.0%, 8.7%, 8.4%, 8.4%, 8.4%, 8.2%, 8.1%, 6.8%, 6.6%, 6.3%, 6.2%, 5.1%, 4.5%, 4.0%, 3.6%, 3.0%, 2.6%, 2.3%, 2.1%, 2.0%, and 1.6%) but containing relatively higher level (31.7%, 64.6%, 46.7%, 62.4%, 36.5%, 67.7%, 79.6%, 100.2%, 87.3%, 62.2%, 25.9%, 11.6%, 87.7%, 96.8%, 24.8%, 72.1%, 38.5%, 79.4%, 61.6%, 42.8%, 74.4%, 51.5%, 130.0%, and 72.0%) of ARRB1/2 protein (Figs. 2A–D and 3A–C, Supplementary Figure 2A–D). The Pearson correlation coefficient was computed to assess the linear relationship between loss of CHRM1 and ARRB1/2 protein levels in all dementia patients. There was a positive correlation between the two variables, r = 0.45, p < 0.0001. The abundance of the hippocampal CHRM1 protein in the male and female AD patients exhibited no significant (p = 0.61) difference in an unpaired t-test (Fig. 3D, E). Furthermore, when AD patients were grouped based on their age at the time of death (<65;≥65–75;≥75–85;≥85 years) and the mean CHRM1 protein level within these groups were compared by one-way ANOVA, the difference among the groups was not statistically significant (p = 0.22) (Fig. 3G). In addition, a two-tailed Pearson correlation test comparing the age of death with the mean CHRM1 or ARRB1/2 protein levels in the AD patient cohort revealed r = –0.07 and –0.16 (p = 0.54 and 0.16) respectively, indicating no correlation between these variables. This indicates that the loss of hippocampal CHRM1 protein was not correlated with the age of death (survival) of AD patients.

We analyzed the abundance of hippocampal CHRM1 and ARRB1/2 in a smaller cohort of FTD (N = 7) and PD (N = 9) patients. Within the FTD cohort, one individual (ID: 6316) exhibited≥50% hippocampal CHRM1 (mean: 68.4%) whereas the rest of the individuals (IDs: 6220, HBQS, S12404, HBFW, 6107, HBER) exhibited < 50% hippocampal CHRM1 (44.8%, 18.6%, 18.1%, 14.8%, 9.7%, and 2.1%, respectively) protein level (Fig. 2, Supplementary Figures 4 and 5). In contrast, within the PD cohort (N = 9), three individuals (IDs: HBRE, HBFI, and HBGL) exhibited≥50% hippocampal CHRM1 (mean: 80.2%, 71.7%, 63.3%, respectively) whereas the rest of the individuals (IDs: BEB18127, HBEO, HBRH, BEB19010, HBFC, and HBHV) exhibited < 50% hippocampal CHRM1 (42.6%, 36.1%, 29.2%, 23.2%, 17.2%, and 6.1% respectively) protein level (Supplementary Figure 5). The mean ARRB1/2 level (%) in PD patients with≥50% CHRM1 level was 156.8%, 164.1%, and 187%, whereas in the PD patients with < 50% CHRM1 protein level the mean ARRB1/2 level was 157.7%, 142.7%, 71.7%, 165.1%, 55.7%, and 149.3% (Supplementary Figure 5). Overall, these results indicate substantial loss (<50%) of hippocampal CHRM1 in 66% of PD (6 out of 9) and 85% of FTD (6 out of 7) patients. Analysis of the alterations in CHRM1 or ARRB1 protein levels with the PMIs in the combined group of AD, PD, and FTD patients by a two-tailed Pearson correlation test revealed r =–0.05 and –0.15 (p = 0.59 and 0.106), respectively indicating that the reduction in these protein levels is not correlated with PMIs. Analysis of the CHRM1 protein level with the age of death in all dementia patients by a two-tailed Pearson revealed a correlation coefficient revealed r = –0.21, p = 0.02 indicating an increase in age of death was correlated with a decrease in CHRM1 protein level. However, Pearson analysis of ARRB1 protein level and age of death in all dementia patients (r = –0.041, p = 0.66) were not correlated.

Loss of CHRM1 protein in the temporal cortex of a subset of individuals designated as unaffected controls by the respective brain bank

We characterized CHRM1 protein in the temporal cortex of a cohort of 9 individuals provided as unaffected controls based on pathology. An individual with a CERAD (Consortium to Establish a Registry for Alzheimer’s Disease) score of “none” [69] was selected as a reference (100%) to quantify all proteins of interest in the temporal cortex samples collected from unaffected, AD, and FTD individuals (Supplementary Table 2). the CERAD score reflects the abundance (not the spatial-temporal distribution) of neuritic plaques (none, sparse, moderate, and frequent) in 3 areas of the isocortex (frontal, temporal, and parietal) [70]. Within this cohort of unaffected individuals, the mean CHRM1 protein level was found to be very low (1.6%) in the temporal cortex of one individual (ID: A19-036, CERAD: none) with no significant pathological findings (Fig. 4A and Supplementary Table 2). Another individual (ID: 110002, CERAD: none) exhibited 53% mean CHRM1 protein level in the temporal cortex, whereas the rest of the individuals (IDs: A20-031, 110408, A19-020, A16-008, 110001, 110003) exhibited relatively high (177.7%, 188.7%, 147.9%, 126.3%, 116.7%, and 113.8%) cortical CHRM1 level, respectively (Supplementary Figure 4A). The mean GAPDH protein level in these individuals was 103% ±7.4% (mean±SD; range 21.2) (Fig. 4B). This indicates loss of CHRM1 protein in the temporal cortex of a subset of two individuals within a cohort of nine unaffected individuals.

Fig. 4

CHRM1 and GAPDH protein levels in the temporal cortex of unaffected individuals and AD/PD patients. A–D) Immunoblots showing the relative abundance of CHRM1 (A, C) and GAPDH (B, D) in temporal cortex derived from postmortem human brains. The sample identification numbers are provided on the top panel of each immunoblot, Male and females are highlighted in red and blue, respectively. The unaffected, AD and PD samples are color-coded green, yellow, and red, respectively.

Loss of CHRM1 protein in the temporal cortex of a subset of AD patients correlated with their survival

We analyzed CHRM1 protein levels in the temporal cortex tissues derived from 74 AD patients. Immunoblotting revealed that 43.2% (32/74) AD patients exhibited≥50% (mean±SEM: 112.5±14.3; N = 32), whereas 56.7% AD patients exhibited < 50% mean CHRM1 protein level (mean±SEM: 26.9±2.4; N = 42) in the temporal cortex (Figs. 4, 5A, B, 6A and Supplementary Figure 6). In contrast, the mean GAPDH protein levels (mean±SEM) in these two groups of AD patients were relatively similar, 112.5±14.3 and 103.4±1.2, respectively (Fig. 4A–D, Supplementary Figures 5A, B, 6A, C, E, and 6A–D). Based, on these observations, it can be concluded that there is a selective loss of CHRM1 protein compared to GAPDH in a subset of AD patients (Fig. 6E). The abundance of the CHRM1 protein in the male and female AD temporal cortex did not differ significantly (p = 0.26) as revealed by an unpaired t-test (Fig. 6D, F). Interestingly, the mean CHRM1 protein level in AD patients’ temporal cortex compared with the age of death was significantly correlated by a two-tailed Pearson correlation test with r = 0.24 and p = 0.04), indicating that lower CHRM1 values were associated with an earlier age of death. To further elaborate on this observation, we grouped AD patients based on their age (<65;≥65–75;≥75–85;≥85 years), comparing the mean CHRM1 protein levels between the groups by one-way ANOVA, the difference between the groups was statistically significant (p = 0.02) (Fig. 3G). The AD patients who died at the age < 65 years (N = 10) or between 65–75 years (N = 14) exhibited mean CHRM1 protein levels of 36.6% ±7.2 (±SEM) and 25.3±4.9 (±SEM) which is relatively lower compared to the AD patients who died at the age 75–85 years (N = 34) or≥85 years (N = 16) exhibiting mean CHRM1 protein levels as 81.9% ±14.6 (±SEM) and 76.7±15.4 (±SEM), respectively (Fig. 6G). No significant correlation (r = 0.24, p = 0.23) was found between the mean cortical CHRM1 protein level and PMI in AD patients. Overall, this result indicates that loss of CHRM1 below a threshold level (50%) may significantly decrease the survival age of AD patients.

Fig. 5

CHRM1/GAPDH protein levels in the temporal cortex and cerebellum of unaffected and AD subjects. A–D): Immunoblots showing the relative abundance of CHRM1 (A, C) and GAPDH (B, D) in the temporal cortex (A, B) and cerebellum (C, D) derived from postmortem human brains. The sample identification numbers are provided on the top panel of each immunoblot, Male and females are highlighted in red and blue, respectively. The unaffected and AD samples are color-coded, green, and pink, respectively. The red dotted rectangle indicates the absence of CHRM1 protein in the cerebellum.

Fig. 6

CHRM1 and GAPDH protein levels in the temporal cortex of unaffected, AD, and PD brains. A) Bar graph showing the relative amount (%; mean±SD) of CHRM1 and GAPDH proteins in the temporal cortex tissues derived from unaffected (N = 9), AD (N = 74), and PD (N = 27). Four replicates from two independent western blotting experiments were used to generate the graph (B, C): Scatter plots showing CHRM1 and GAPDH protein levels (%, mean from A) in the temporal cortex. The red and blue bars represent the mean and SD in this figure and subsequent figures, respectively. D) Scatter plot showing the CHRM1 protein levels (%, mean from A) in male versus female AD patients. E) Bubble plot showing the abundance of CHRM1 and GAPDH proteins (%, Mean from A) in the temporal cortex combined with the sex and age variables of the AD patients. The red dotted lines indicate 50% and 94% cut-off values for CHRM1 and GAPDH protein levels, respectively. F) Bubble plot showing the abundance of CHRM1 (%, mean from A) in the temporal cortex combined with the sex and age variables of AD patients. G, H) Scatter plot showing the CHRM1/GAPDH protein levels (% mean from A) in the temporal cortex derived from different age groups of AD (G) and PD (H) patients. p-value by One-Way ANOVA.

Loss of CHRM1 protein in the temporal cortex of a subset of PD patients correlated with their survival

We analyzed CHRM1 and GAPDH abundance in the temporal cortex tissue derived from 27 PD patients (Supplementary Table 2). Within the PD cohort, 19 individuals exhibited≥50% mean CHRM1 (mean±SEM: 110.5±9 %) protein level in the temporal cortex, whereas eight individuals (IDs: A19-019, A19-003, A21-010, A16-011, A21-057, A19-008, A15-022, A15-018) exhibited < 50% mean CHRM1 (40.4, 37.2%, 36.0%, 32.0%, 25.6%, 14.9%, 13.6%, and 2.0%, respectively) protein level in the same brain regions (Figs. 4, 6A, B). In contrast, the mean GAPDH protein levels (mean±SEM) in these two groups of PD patients were relatively similar, 99.6±0.6 and 101.8±1.6, respectively (Fig. 6A, C). Based, on these observations, it can be concluded that there was a selective loss of CHRM1 protein compared to GAPDH in a subset of PD patients (Fig. 6A–C). A significant difference in the abundance of temporal cortex CHRM1 protein was found between the male (N = 22) versus female (N = 5) PD patients as revealed by a nonparametric Mann-Whitney test (p = 0.027); however, the disparity in terms of a relatively smaller number of females compared to the males question the reliability of this findings, therefore, future analysis is warranted with a larger and equal number of both sex samples. Interestingly, when the mean CHRM1 protein level in PD patients’ temporal cortex was compared with their age of death by a two-tailed Pearson correlation test, the correlation coefficient r = 0.44 (p = 0.022) indicates an earlier age of death correlated with lower CHRM1 protein level. To further elaborate on this observation, we grouped PD patients based on their age of death (<65;≥65–75;≥75 years), and the mean CHRM1 protein levels between the groups were compared by one-way ANOVA test, the difference between the groups was found statistically significant (p = 0.0014) (Fig. 6H). The mean CHRM1 protein levels (mean±SEM) in the PD patients who died at the age of < 65 years (N = 5),≥65–75 years (N = 10), and≥75 years (N = 12) were 37.5±15.1%, 76.5±16.9%, and 112.5±11.9% respectively (Fig. 6H). No correlation was found between the mean temporal cortical CHRM1 protein level and PMI in PD patients. Overall, this result indicates loss of CHRM1 in the temporal cortex below a threshold level (50%) significantly decreased the survival of PD patients, and that the extent of the loss was associated with the age at death.

The CHRM1 protein was not detected in the cerebellum of AD patients

Previously, we demonstrated that the CHRM1 protein is not present in the human cerebellum of non-AD patients based upon western blotting of a small piece (<1 g) of the cerebellum of two individuals. Analysis of a larger sample of cerebellum collected from 10 AD patients further confirmed the absence of CHRM1 protein in the cerebellum. Thus, there was no indication that the neurodegenerative condition of AD caused ectopic expression of CHRM1 protein in the AD cerebellum (Fig. 5C, D).

DISCUSSION

In this study, we demonstrated that a ~70 kDa N-glycosylated form of CHRM1 protein is expressed in the hippocampus and cerebral cortex, but not in the cerebellum of mammalian (human and pig) brains. Next, we showed that subsets of AD, PD, and FTD patients exhibited a dramatic reduction (<50%) in the hippocampal CHRM1 protein. The reduced hippocampal CHRM1 greatly exceeded the smaller loss of ARRB1/2 proteins, indicating selective loss of the receptor compared to the components of internalization and downstream signaling. A subset of AD and PD patients exhibited a dramatic (<50%) reduction of the CHRM1 protein level in the temporal cortex, which was correlated with poor patient survival. We also showed that a subset of postmortem brains labeled as unaffected controls exhibited a severe reduction of the CHRM1 protein (<50%) in both hippocampus and temporal cortex. Gene expression variability in the population and its dependence on aging may likely be a contributing factor [71], at least in part. On the other hand, loss of the protein in individuals marked as unaffected may be due to underlying neuropathological conditions. This has been reflected in the neuropathological examination of these “unaffected” postmortem brains indicating underlying disease conditions revealed by the presence of NFT pathology identified by the repositories following their harvest which is indicated by their Braak stages [37]. This indicates the limitations in the prospective identification of unaffected controls through postmortem brain repositories. Overall, the findings from this study indicate that AChE inhibitor-mediated elevation of ACh in dementia brains or the use of cholinomimetic drugs to activate mAChRs may not be an effective treatment option for the cohort of AD, PD, or FTD patients that exhibited severely reduced level of CHRM1 in the temporal cortex and hippocampus, brain regions associated with cognitive function. Furthermore, this study also revealed that the temporal cortical CHRM1 protein level is an indicator of the survival of dementia patients.

A key observation in this study is the dramatic loss (<50%) of CHRM1 protein in the temporal cortex (56% of AD and 42% of PD patients) and hippocampus (78% of AD, 66% of PD, and 85% of FTD patients) in a subset of AD, PD, and FTD patients indicating the involvement of this pathological event in a wide spectrum of dementia patients. Cholinergic neuron loss and subsequent impairment in dopaminergic transmission have been suggested as one of the main factors underlying AD-related psychiatric symptoms [72 –74]. Disruption of central cholinergic transmission has been associated with cognitive decline in PD [75] and FTD, but not as severe as in AD [76 –78]. Therefore, our finding of dysfunctional muscarinic cholinergic signaling in AD, PD, and FTD agrees with the previous studies but adds to the concept of loss of cholinoceptivity by showing a reduction of the CHRM1 protein. Thusly, therapeutic efforts should be directed toward restoring not only cholinergic neurons but also CHRM1 protein expression in dementia patients.

A potential explanation for the reduced CHRM1 protein in the temporal cortex and hippocampus of the AD brain may be the abnormal degradation of the receptor after ACh signaling. Agonist-bound mAChRs are phosphorylated by GPCR kinases (GRKs), which initiate their desensitization through uncoupling from G proteins, promoting ARRB1/2 binding, receptor internalization, and receptor degradation (downregulation) or recycling [79]. Agonist-induced phosphorylation of the third intracellular loop of mAChRs by GRK2 is the initial and essential step for receptor internalization and downregulation [79]. Interestingly, it has been reported that GRK2 phosphorylates residues in the third intracellular loop of CHRM1 in an agonist-dependent manner [53]. Overexpression of GRK2 has been reported during the early stages of damage in aged human and AD patients [80]. The link between GRK2 and AD has been further corroborated by the evidence that GRK2 is a potential Microtubule Associated Protein Tau (MAPT) kinase [81] and is qualitatively associated with NFTs in AD patient brain samples [82]. Therefore, the overactivity of GRK2 in the AD brain may be responsible for CHRM1 downregulation. Another possibility is altered regulation of GRK2 activity by intracellular calcium ion (Ca^2 +) concentrations. GRK2 is inhibited by calmodulin in a Ca^2 +-dependent manner [83]. Recent evidence supports the idea that amyloid-β (Aβ) deposits/oligomers and presenilin 1 and 2 (PSEN1/2) mutations, both linked to AD, destroy brain cells by making them unable to control their internal Ca^2 + concentrations [84 –91]. This formed the molecular basis of the “Calcium Hypothesis” underlying AD pathogenesis [92]. Validation of this hypothesis is beyond the scope of his study, but can be the subject of future studies.

Another major finding in this study is the positive correlation between severely (<50%) reduced temporal cortical CHRM1 protein levels and early death (<65–75 years) of AD and PD patients. In contrast, hippocampal CHRM1 loss was not associated with the age of death of AD patients. The exact reason for this difference is not known; however, it may depend on how CHRM1 loss in the hippocampus versus cortex affects the cognitive and other behavioral functions that are directly related to the survival of the individuals. In this context, it is important to note that Chrm1 knockout mice did not impair hippocampus-dependent learning and memory tasks, but displayed significant impairments in nonmatching-to-sample working memory and consolidation compared to wild-type littermates [93, 94]. Based upon these observations, it was concluded that Chrm1 may not be essential for memory formation or initial stability of memory in the hippocampus, but is most likely involved in processes requiring interactions between the cerebral cortex and hippocampus [94]. It is generally considered that consolidation is the process by which hippocampus guides the reorganization of the information stored in the neocortex such that it eventually becomes independent of the hippocampus [95]. The idea is that gradual changes in the neocortex, beginning at the time of learning, establish stable long-term memory by increasing the neuronal complexity, distribution, and connectivity among multiple cortical regions [95]. Therefore, based on CHRM1 knockout mouse-based observations, it may be too speculative to suggest that the loss of CHRM1 in the cortical neurons may have a larger impact on memory consolidation compared to hippocampal loss affecting both working memory and long-term declarative memory during AD pathogenesis associated with survival.

The significance of this study is that it highlights the need for choosing the right patient for the clinical use of drugs targeted to CHRM1. Our findings indicate that the CHRM1 agonist or AChE inhibitors drugs would work best in a cohort of patients with moderate to normal levels of CHRM1 protein in the cortex and hippocampus. Therefore, future research should focus on the retrospective identification of AD patients with a threshold cortical CHRM1 protein level. Identification of peripheral or cerebrospinal fluid (CSF)-based biomarkers reflecting the status of CHRM1 protein level in the brain would be an ideal step towards such an initiative, for example, cyclic guanosine monophosphate (cGMP) level could be used as a biomarker because reduced CSF cGMP level has been associated with the severity of dementia [96]. Furthermore, cholinergic signaling-mediated regulation of neuronal cGMP has been demonstrated by showing that carbachol (agonist)-induced increase in cGMP was blocked both by atropine, a mAChR antagonist and by LY83583, a guanylyl cyclase inhibitor [97]. Over the past decades, CHRM1 agonists emerged as an important tool for AD treatment because of the observations that impaired activation of CHRM1 affects several major AD hallmarks, including cognitive dysfunction, NFTs, and Aβ pathologies [98 –100]. CHRM1 activation decreases MAPT phosphorylation via activation of protein kinase C and inhibition of glycogen synthase kinase beta [100, 101]. Therefore, future studies must focus on the therapeutic enhancement of CHRM1 mRNA expression to augment CHRM1 protein levels in dementia patients. In this context, it is important to note that the CHRM1 promoter harbors consensus regulatory elements of different transcription factors including Nuclear Factor Kappa B (NF-κB) [102]. NF-κB signaling modulates CHRM3 expression [103]. These observations provide a connection between two major hypotheses of AD, the “neuroinflammation hypothesis” and the “cholinergic hypothesis” through NF-κB-mediated regulation of cholinergic signaling [104]. Another approach to restoring the loss of CHRM1 in dementia brains would be to understand the role of GRKs underlying CHRM1 downregulation and therapeutic inhibition of this process. A potential link between dysregulation of GRKs and AD has been reported in the existing literature [80, 105]. Overall, based on our findings, the above discussion sets a direction for future research and therapeutic strategy for the identification of drug targets for AD and PD.

Footnotes

ACKNOWLEDGMENTS

We thank Dr. Stefan Prokop, University of Florida (UF) for providing human brain tissue samples through the UF Neuromedicine Human Brain and Tissue Bank, Florida, US; Dr. Olivia Spicer, NIH NBB, US, for helping us obtaining the hippocampal samples from the NIH NeuroBioBank; Stanley Kim, Alzo Biosciences Incorporation, San Diego, US, for financial support to conduct the research; Dr. Paul Fernyhough, University of Manitoba, Canada, for providing the Chrm1^-/- knockout mouse brain tissues; and Paulina Morelund for assisting in immunoblot quantification. Funding for this research was provided by Alzo Biosciences Inc, San Diego, California to MGS. Infrastructure support was provided by St. Boniface Hospital Albrechtsen Research Centre, Canada, and Nova Southeastern University, USA.

Authors’ disclosures available online ().

The supplementary material is available in the electronic version of this article: .

References

Liu

P-P

, Xie

, Meng

X-Y

, Kang

J-S

(2019) History and progress of hypotheses and clinical trials for Alzheimer’s disease. Signal Transduct Target Ther 4, 29.

Perry

, Gibson

, Blessed

, Perry

, Tomlinson

(1977) Neurotransmitter enzyme abnormalities in senile dementia. Choline acetyltransferase and glutamic acid decarboxylase activities in necropsy brain tissue. J Neurol Sci 34, 247–265.

Bowen

, Smith

, White

, Davison

(1976) Neurotransmitter-related enzymes and indices of hypoxia in senile dementia and other abiotrophies. Brain 99, 459–496.

Davies

, Maloney

(1976) Selective loss of central cholinergic neurons in Alzheimer’s disease. Lancet 2, 1403.

Loewi

, Navratil

(1926) Über humorale Übertragbarkeit der Herznervenwirkung. Pflugers Arch 214, 678–688.

Rylett

, Ball

, Colhoun

(1983) Evidence for high affinity choline transport in synaptosomes prepared from hippocampus and neocortex of patients with Alzheimer’s disease. Brain Res 289, 169–175.

Nilsson

, Nordberg

, Hardy

, Wester

, Winblad

(1986) Physostigmine restores3H-acetylcholine efflux from Alzheimer brain slices to normal level. J Neural Transm 67, 275–285.

Whitehouse

, Price

, Struble

, Clark

, Coyle

, DeLong

(1982) Alzheimer’s disease and senile dementia: Loss of neurons in the basal forebrain. Science 215, 1237–1239.

Drachman

, Leavitt

(1974) Human memory and the cholinergic system. A relationship to aging? . Arch Neurol 30, 113–121.

10.

Bartus

, Dean

, Beer

, Lippa

(1982) The cholinergic hypothesis of geriatric memory dysfunction. Science 217, 408–414.

11.

Martinez

, Castro

(2006) Novel cholinesterase inhibitors as future effective drugs for the treatment of Alzheimer’s disease. Expert Opin Investig Drugs 15, 1–12.

12.

Marucci

, Buccioni

, Ben

, Lambertucci

, Volpini

, Amenta

(2021) Efficacy of acetylcholinesterase inhibitors in Alzheimer’s disease. Neuropharmacology 190, 108352.

13.

Picciotto

, Caldarone

, King

, Zachariou

(2000) Nicotinic receptors in the brain. Links between molecular biology and behavior. Neuropsychopharmacology 22, 451–465.

14.

Wess

(2003) Novel insights into muscarinic acetylcholine receptor function using gene targeting technology. Trends Pharmacol Sci 24, 414–420.

15.

Picciotto

, Caldarone

, Brunzell

, Zachariou

, Stevens

, King

(2001) Neuronal nicotinic acetylcholine receptor subunit knockout mice: Physiological and behavioral phenotypes and possible clinical implications. Pharmacol Ther 92, 89–108.

16.

Jiang

, Li

, Zhang

, Zhao

, Bu

, Xu

, Zhang

Y-W

(2014) M1 muscarinic acetylcholine receptor in Alzheimer’s disease. Neurosci Bull 30, 295–307.

17.

Konar

, Gupta

, Shukla

, Maloney

, Khanna

, Wadhwa

, Lahiri

, Thakur

(2019) M1 muscarinic receptor is a key target of neuroprotection, neuroregeneration and memory recovery by i-Extract from Withania somnifera. Sci Rep 9, 13990.

18.

Lebois

, Schroeder

, Esparza

, Bridges

, Lindsley

, Conn

, Brody

, Daniels

, Levey

(2017) Disease-modifying effects of M1 muscarinic acetylcholine receptor activation in an Alzheimer’s disease mouse model. ACS Chem Neurosci 8, 1177–1187.

19.

Oddo

, LaFerla

(2006) The role of nicotinic acetylcholine receptors in Alzheimer’s disease. J Physiol Paris 99, 172–179.

20.

Lombardo

, Maskos

(2015) Role of the nicotinic acetylcholine receptor in Alzheimer’s disease pathology and treatment. Neuropharmacology 96, 255–262.

21.

, Ishikawa

, Zhang

, Hashimoto

(2010) Brain imaging of nicotinic receptors in Alzheimer’s disease. Int J Alzheimers Dis 2010, 548913.

22.

Lasala

, Fabiani

, Corradi

, Antollini

, Bouzat

(2019) Molecular modulation of human α7 nicotinic receptor by amyloid-β peptides. Front Cell Neurosci 13, 37.

23.

Liu

, Wu

(2006) Neuronal nicotinic acetylcholine receptors serve as sensitive targets that mediate β-amyloid neurotoxicity. Acta Pharmacol Sin 27, 1277–1286.

24.

Shimohama

, Kawamata

(2018) Roles of nicotinic acetylcholine receptors in the pathology and treatment of Alzheimer’s and Parkinson’s diseases. In Nicotinic Acetylcholine Receptor Signaling in Neuroprotection, Akaike A, Shimohama S, Misu Y, eds. Springer Singapore, Singapore, pp. 137–158.

25.

Abrams

, Andersson

K-E

, Buccafusco

, Chapple

, de Groat

, Fryer

, Kay

, Laties

, Nathanson

, Pasricha

, Wein

(2006) Muscarinic receptors: Their distribution and function in body systems, and the implications for treating overactive bladder. Br J Pharmacol 148, 565–578.

26.

Smith

, Annis

, Ehlert

, Leslie

(1991) N-[3H]methylscopolamine labeling of non-M1, non-M2 muscarinic receptor binding sites in rat brain. J Pharmacol Exp Ther 256, 1173–1181.

27.

Pepeu

, Pepeu

(1984) The muscarinic receptors. In Principles and Methods in Receptor Binding, Cattabeni F, Nicosia S, eds. Springer US, Boston, MA, pp. 65–77.

28.

Waelbroeck

, Tastenoy

, Camus

, Christophe

(1990) Binding of selective antagonists to four muscarinic receptors (M1 to M4) in rat forebrain. Mol Pharmacol 38, 267–273.

29.

Buckley

, Bonner

, Brann

(1988) Localization of a family of muscarinic receptor mRNAs in rat brain. J Neurosci 8, 4646–4652.

30.

Levey

, Kitt

, Simonds

, Price

, Brann

(1991) Identification and localization of muscarinic acetylcholine receptor proteins in brain with subtype-specific antibodies. J Neurosci 11, 3218–3226.

31.

, Yasuda

, Wall

, Wellstein

, Wolfe

(1991) Distribution of m2 muscarinic receptors in rat brain using antisera selective for m2 receptors. Mol Pharmacol 40, 28–35.

32.

Yasuda

, Ciesla

, Flores

, Wall

, Li

, Satkus

, Weisstein

, Spagnola

, Wolfe

(1993) Development of antisera selective for m4 and m5 muscarinic cholinergic receptors: Distribution of m4 and m5 receptors in rat brain. Mol Pharmacol 43, 149–157.

33.

Wall

, Yasuda

, Hory

, Flagg

, Martin

, Ginns

, Wolfe

(1991) Production of antisera selective for m1 muscarinic receptorsusing fusion proteins: Distribution of m1 receptors in rat brain. Mol Pharmacol 39, 643–649.

34.

Ehlert

, Tran

(1990) Regional distribution of M1, M2 and non-M1, non-M2 subtypes of muscarinic binding sites in rat brain. J Pharmacol Exp Ther 255, 1148–1157.

35.

Giraldo

, Hammer

, Ladinsky

(1987) Distribution of muscarinic receptor subtypes in rat brain as determined in binding studies with AF-DX 116 and pirenzepine. Life Sci 40, 833–840.

36.

Levey

(1996) Muscarinic acetylcholine receptor expression in memory circuits: Implications for treatment of Alzheimer disease. Proc Natl Acad Sci U S A 93, 13541–13546.

37.

Braak

, Braak

(1991) Neuropathological stageing of Alzheimer-related changes. Acta Neuropathol 82, 239–259.

38.

Halliday

(2017) Pathology and hippocampal atrophy in Alzheimer’s disease. Lancet Neurol 16, 862–864.

39.

Serrano-Pozo

, Frosch

, Masliah

, Hyman

(2011) Neuropathological alterations in Alzheimer disease. Cold Spring Harb Perspect Med 1, a006189.

40.

Visser

, Verhey

FRJ

, Hofman

PAM

, Scheltens

, Jolles

(2002) Medial temporal lobe atrophy predicts Alzheimer’s disease in patients with minor cognitive impairment. J Neurol Neurosurg Psychiatry 72, 491–497.

41.

Perl

, Olanow

, Calne

(1998) Alzheimer’s disease and Parkinson’s disease: Distinct entities or extremes of a spectrum of neurodegeneration? Ann Neurol 44, S19–31.

42.

Fang

, Lv

, Mao

, Dong

, Liu

(2020) Cognition deficits in Parkinson’s disease: Mechanisms and treatment. Parkinsons Dis 2020, 2076942–2076942.

43.

Young

, Lavakumar

, Tampi

, Balachandran

, Tampi

(2018) Frontotemporal dementia: Latest evidence and clinical implications. Ther Adv Psychopharmacol 8, 33–48.

44.

Pillai-Kastoori

, Schutz-Geschwender

, Harford

(2020) A systematic approach to quantitative Western blot analysis. Anal Biochem 593, 113608.

45.

McDonough

, Veiras

, Minas

, Ralph

(2015) Considerations when quantitating protein abundance by immunoblot. Am J Physiol Cell Physiol 308, C426–C433.

46.

Charette

, Lambert

, Nadeau

, Landry

(2010) Protein quantification by chemiluminescent western blotting: Elimination of the antibody factor by dilution series and calibration curve. J Immunol Methods 353, 148–150.

47.

Hartig

(2013) Basic image analysis and manipulation in ImageJ. Curr Protoc Mol Biol Chapter 14, Unit14.15.

48.

Sabbir

, Taylor

, Zahradka

(2020) Hypomorphic CAMKK2 in EA.hy926 endothelial cells causes abnormal transferrin trafficking, iron homeostasis and glucose metabolism. Biochim Biophys Acta Mol Cell Res 1867, 118763.

49.

Dunn

(1964) Multiple comparisons using rank sums. Technometrics 6, 241–252.

50.

Siegel

S ed.

(1956) Nonparametric statistics for the behavioral sciences, McGraw-Hill, New York.

51.

Sabbir

, Calcutt

, Fernyhough

(2018) Muscarinic acetylcholine type 1 receptor activity constrains neurite outgrowth by inhibiting microtubule polymerization and mitochondrial trafficking in adult sensory neurons. Front Neurosci 12, 402.

52.

Sabbir

, Fernyhough

(2018) Muscarinic receptor antagonists activate ERK-CREB signaling to augment neurite outgrowth of adult sensory neurons. Neuropharmacology 143, 268–281.

53.

Haga

, Kameyama

, Haga

, Kikkawa

, Shiozaki

, Uchiyama

(1996) Phosphorylation of human m1 muscarinic acetylcholine receptors by G protein-coupled receptor kinase 2 and protein kinase C. J Biol Chem 271, 2776–2782.

54.

Peterson

, Rosenbaum

, Broderick

, Schimerlik

(1986) Physical properties of the purified cardiac muscarinic acetylcholine receptor. Biochemistry 25, 3189–3202.

55.

Thal

, Sun

, Feng

, Nawaratne

, Leach

, Felder

, Bures

, Evans

, Weis

, Bachhawat

, Kobilka

, Sexton

, Kobilka

, Christopoulos

(2016) Crystal structures of the M1 and M4 muscarinic acetylcholine receptors. Nature 531, 335–340.

56.

, Tan

, Li

, Cui

, Qiu

, Wang

, Zhang

, Li

, Zhu

, Zhou

, Chen

(2019) Direct SUMOylation of M1 muscarinic acetylcholine receptor increases its ligand-binding affinity and signal transduction. FASEB J 33, 3237–3251.

57.

Hilgarth

, Sarge

(2005) Detection of sumoylated proteins. Methods Mol Biol 301, 329–338.

58.

Haga

, Haga

(1985) Purification of the muscarinic acetylcholine receptor from porcine brain. J Biol Chem 260, 7927–7935.

59.

Liang

, Martin

, Harden

(1987) [3H]propylbenzilylcholine mustard-labeling of muscarinic cholinergic receptors that selectively couple to phospholipase C or adenylate cyclase in two cultured cell lines. Mol Pharmacol 32, 443–449.

60.

Ohara

, Uchiyama

, Ohara

, Haga

, Ichiyama

(1990) Interaction of deglycosylated muscarinic receptors with ligands and G proteins. Eur J Pharmacol 189, 341–346.

61.

Tarentino

, Trimble

, Plummer

(1989) Chapter 5 Enzymatic approaches for studying the structure, synthesis, and processing of glycoproteins. In Methods in Cell Biology, Tartakoff AM, ed. Academic Press, pp. 111–139.

62.

Hughes

, Berg

, Danziger

, Coben

, Martin

(1982) A new clinical scale for the staging of dementia. Br J Psychiatry 140, 566–572.

63.

Braak

, Alafuzoff

, Arzberger

, Kretzschmar

, Del Tredici

(2006) Staging of Alzheimer disease-associated neurofibrillarypathology using paraffin sections and immunocytochemistry. ActaNeuropathol 112, 389–404.

64.

Reiman

, Arboleda-Velasquez

, Quiroz

, Huentelman

, Beach

, Caselli

, Chen

, Su

, Myers

, Hardy

, et al.(2020) Exceptionally low likelihood of Alzheimer’s dementia in APOE2homozygotes from a 5,000-person neuropathological study. NatCommun 11, 667.

65.

P-T

, Li

Y-J

, Qin

X-J

, Kroner

, Green-Odlum

, Xu

, Wang

T-Y

, Schmechel

, Hulette

, Ervin

, Hauser

, Haines

, Pericak-Vance

, Gilbert

(2007) A SAGE study of apolipoprotein E3/3, E3/4 andE4/4 allele-specific gene expression in hippocampus in Alzheimerdisease. Mol Cell Neurosci 36, 313–331.

66.

van der Westhuizen

, Choy

KHC

, Valant

, McKenzie-Nickson

, Bradley

, Tobin

, Sexton

, Christopoulos

(2021) Fine tuningmuscarinic acetylcholine receptor signaling through allostery andbias. Front Pharmacol 11, 606656.

67.

Zissler

, Stoiber

, Steinbacher

, Geissenberger

, Monticelli

, Pittner

(2020) Postmortem protein degradation as a tool toestimate the PMI: A systematic review. Diagnostics (Basel) 10, 1014.

68.

Fountoulakis

, Hardmeier

, Höger

, Lubec

(2001) Postmortem changes in the level of brain proteins. Exp Neurol 167, 86–94.

69.

Morris

, Heyman

, Mohs

, Hughes

, van Belle

, Fillenbaum

, Mellits

, Clark

(1989) The Consortium to Establish a Registry for Alzheimer’s Disease (CERAD). Part I. Clinical and neuropsychological assessment of Alzheimer’s disease. Neurology 39, 1159–1165.

70.

Serrano-Pozo

, Qian

, Muzikansky

, Monsell

, Montine

, Frosch

, Betensky

, Hyman

(2016) Thal amyloid stages do not significantly impact the correlation between neuropathological change and cognition in the Alzheimer disease continuum. J Neuropathol Exp Neurol 75, 516–526.

71.

Bashkeel

, Perkins

, Kærn

, Lee

(2019) Human geneexpression variability and its dependence on methylation and aging. BMC Genomics 20, 941.

72.

Ferreira-Vieira

, Guimaraes

, Silva

, Ribeiro

(2016) Alzheimer’s disease: Targeting the cholinergic system. Curr Neuropharmacol 14, 101–115.

73.

Rosenberg

, Nowrangi

, Lyketsos

(2015) Neuropsychiatric symptoms in Alzheimer’s disease: What might be associated brain circuits? . Mol Aspects Med 43-44, 25–37.

74.

Martorana

, Koch

(2014) Is dopamine involved in Alzheimer’s disease? Front Aging Neurosci 6, 252.

75.

Pasquini

, Brooks

, Pavese

(2021) The cholinergic brain in Parkinson’s disease. Mov Disord Clin Pract 8, 1012–1026.

76.

Wood

, Etienne

, Lal

, Nair

, Finlayson

, Gauthier

, Palo

, Haltia

, Paetau

, Bird

(1983) A post-mortem comparison of the cortical cholinergic system in Alzheimer’s disease and Pick’s disease. J Neurol Sci 62, 211–217.

77.

Sparks

, Markesbery

(1991) Altered serotonergic and cholinergic synaptic markers in Pick’s disease. Arch Neurol 48, 796–799.

78.

Hansen

, Deteresa

, Tobias

, Alford

, Terry

(1988) Neocortical morphometry and cholinergic neurochemistry in Pick’s disease. Am J Pathol 131, 507–518.

79.

Haga

(2013) Molecular properties of muscarinic acetylcholine receptors. Proc Jpn Acad Ser B Phys Biol Sci 89, 226–256.

80.

Obrenovich

, Palacios

, Gasimov

, Leszek

, Aliev

(2009) The GRK2 overexpression is a primary hallmark of mitochondrial lesions during early Alzheimer disease. Cardiovasc Psychiatry Neurol 2009, 327360.

81.

, Wei

, Huang

, Chen

, Wang

, Wu

, Zhang

, Pei

, Wang

(2020) Aβ monomer induces phosphorylation of Tau at Ser-214 through β2AR-PKA-JNK signaling pathway. FASEB J 34, 5092–5105.

82.

Takahashi

, Uchikado

, Caprotti

, Weidenheim

, Dickson

, Ksiezak-Reding

, Pasinetti

(2006) Identification of G-protein coupled receptor kinase 2 in paired helical filaments and neurofibrillary tangles. J Neuropathol Exp Neurol 65, 1157–1169.

83.

Haga

, Tsuga

, Haga

(1997) Ca2+-dependent inhibition of G protein-coupled receptor kinase 2 by calmodulin. Biochemistry 36, 1315–1321.

84.

Marx

(2007) Alzheimer’s disease. Fresh evidence points to an old suspect: Calcium. Science 318, 384–385.

85.

Thornton

, Bright

, Sastre

, Muckett

, Carling

(2011) AMP-activated protein kinase (AMPK) is a tau kinase, activated in response to amyloid beta-peptide exposure. Biochem J 434, 503–512.

86.

Mairet-Coello

, Courchet

, Pieraut

, Courchet

, Maximov

, Polleux

(2013) The CAMKK2-AMPK kinase pathway mediates the synaptotoxic effects of Abeta oligomers through Tau phosphorylation. Neuron 78, 94–108.

87.

Ito

, Oka

, Etcheberrigaray

, Nelson

, McPhie

, Tofel-Grehl

, Gibson

, Alkon

(1994) Internal Ca2+ mobilization is altered in fibroblasts from patients with Alzheimer disease. Proc Natl Acad Sci U S A 91, 534–538.

88.

Cheung

, Shineman

, Muller

, Cardenas

, Mei

, Yang

, Tomita

, Iwatsubo

, Lee

, Foskett

(2008) Mechanism of Ca2+ disruption in Alzheimer’s disease by presenilin regulation of InsP3 receptor channel gating. Neuron 58, 871–883.

89.

Green

, LaFerla

(2008) Linking calcium to Aβ and Alzheimer’s disease. Neuron 59, 190–194.

90.

Demuro

, Parker

, Stutzmann

(2010) Calcium signaling and amyloid toxicity in Alzheimer disease. J Biol Chem 285, 12463–12468.

91.

Bezprozvanny

(2009) Calcium signaling and neurodegenerative diseases. Trends Mol Med 15, 89–100.

92.

Alzheimer’s Association Calcium Hypothesis Workgrou (2017) Calcium Hypothesis of Alzheimer’s disease and brain aging: A framework for integrating new evidence into a comprehensive theory of pathogenesis. , -. Alzheimers Dement 13, 178–182e117.

93.

Miyakawa

, Yamada

, Duttaroy

, Wess

(2001) Hyperactivity and intact hippocampus-dependent learning in mice lacking the M1 muscarinic acetylcholine receptor. J Neurosci 21, 5239–5250.

94.

Anagnostaras

, Murphy

, Hamilton

, Mitchell

, Rahnama

, Nathanson

, Silva

(2003) Selective cognitive dysfunction in acetylcholine M1 muscarinic receptor mutant mice. Nat Neurosci 6, 51–58.

95.

Squire

, Genzel

, Wixted

, Morris

(2015) Memory consolidation. Cold Spring Harb Perspect Biol 7, a021766.

96.

Hesse

, Lausser

, Gummert

, Schmid

, Wahler

, Schnack

, Kroker

, Otto

, Tumani

, Kestler

, Rosenbrock

, von Arnim

CAF

(2017) Reduced cGMP levels in CSF of AD patients correlate with severity of dementia and current depression. Alzheimers Res Ther 9, 17.

97.

Liu

, Ding

, Faiman

, Gillette

(1997) Coupling of muscarinic cholinergic receptors and cGMP in nocturnal regulation of the suprachiasmatic circadian clock. J Neurosci 17, 659–666.

98.

Fisher

(2008) M1 muscarinic agonists target major hallmarks of Alzheimer’s disease–the pivotal role of brain M1 receptors. Neurodegener Dis 5, 237–240.

99.

Fisher

, Brandeis

, Bar-Ner

, Kliger-Spatz

, Natan

, Sonego

, Marcovitch

, Pittel

(2002) AF150(S) and AF267B: M1 muscarinic agonists as innovative therapies for Alzheimer’s disease. J Mol Neurosci 19, 145–153.

100.

Caccamo

, Oddo

, Billings

, Green

, Martinez-Coria

, Fisher

, LaFerla

(2006) M1 receptors play a central role in modulating AD-like pathology in transgenic mice. Neuron 49, 671–682.

101.

Forlenza

, Spink

, Dayanandan

, Anderton

, Olesen

, Lovestone

(2000) Muscarinic agonists reduce tau phosphorylation in non-neuronal cells via GSK-3beta inhibition and in neurons. J Neural Transm (Vienna) 107, 1201–1212.

102.

Pepitoni

, Buckley

, Wood

(1997) Structure of the m1 muscarinic acetylcholine receptor gene and its promoter. J Biol Chem 272, 17112–17117.

103.

Paula

FMM

, Barbosa

HCL

, Carneiro

, Persaud

, Gagliardino

, Boschero

, Souza

KLA

(2010) Requirement of NF-kappaB signalling pathway for modulation of the cholinergic muscarinic M3 receptor expression by INGAP-PP in insulin-producing cells. Eur J Pharmacol 642, 37–46.

104.

Hwang C

, Choi

, Park

, Hong

(2019) NF-κB as a key mediator of brain inflammation in Alzheimer’s disease. CNS Neurol Disord Drug Targets 18, 3–10.

105.

Guimarães

, Swanson

, Kofler

, Thathiah

(2021) G protein-coupled receptor kinases are associated with Alzheimer’s disease pathology. Neuropathol Appl Neurobiol 47, 942–957.

106.

Jumper

, Evans

, Pritzel

, Green

, Figurnov

, Ronneberger

, Tunyasuvunakool

, Bates

, zídek

, Potapenko

, Bridgland

, Meyer

, Kohl

SAA

, Ballard

, Cowie

, Romera-Paredes

, Nikolov

, Jain

, Adler

, Back

, Petersen

, Reiman

, Clancy

, Zielinski

, Steinegger

, Pacholska

, Berghammer

, Bodenstein

, Silver

, Vinyals

, Senior

, Kavukcuoglu

, Kohli

, Hassabis

(2021) Highly accurate protein structure prediction with AlphaFold. Nature 596, 583–589.