CHRFAM7A: A human specific fusion gene, accounts for the translational gap for cholinergic strategies in Alzheimer's disease.

BACKGROUND: Cholinergic neuronal loss is one of the hallmarks of AD related neurodegeneration; however, preclinical promise of α7 nAChR drugs failed to translate into humans. CHRFAM7A, a uniquely human fusion gene, is a negative regulator of α7 nAChR and was unaccounted for in preclinical models.
METHODS: Molecular methods: Function of CHRFAM7A alleles was studied in vitro in two disease relevant phenotypic readouts: electrophysiology and Aβ uptake. Genome edited human induced pluripotent stem cells (iPSC) were used as a model system with the human context. Double blind pharmacogenetic study: We performed double-blind pharmacogenetic analysis on the effect of AChEI therapy based on CHRFAM7A carrier status in two paradigms: response to drug initiation and DMT effect. Mini Mental Status Examination (MMSE) was used as outcome measure. Change in MMSE score from baseline was compared by 2-tailed T-test. Longitudinal analysis of clinical outcome (MMSE) was performed using a fitted general linear model, based on an assumed autoregressive covariance structure. Model independent variables included age, sex, and medication regimen at the time of the first utilized outcome measure (AChEI alone or AChEI plus memantine), APOE4 carrier status (0, 1 or 2 alleles as categorical variables) and CHRFAM7A genotype.
FINDINGS: The direct and inverted alleles have distinct phenotypes. Functional CHRFAM7A allele classifies the population as 25% non-carriers and 75% carriers. Induced pluripotent stem cell (iPSC) models α7 nAChR mediated Aβ neurotoxicity. Pharmacological readout translates into both first exposure (p = 0.037) and disease modifying effect (p = 0.0048) in two double blind pharmacogenetic studies.
INTERPRETATION: CHRFAM7A accounts for the translational gap in cholinergic strategies in AD. Clinical trials not accounting for this uniquely human genetic factor may have rejected drug candidates that would benefit 25% of AD. Reanalyses of the completed trials using this pharmacogenetic paradigm may identify effective therapy.

Evidence before this study

α7 nAChR has been implicated in Alzheimer's disease for decades. Extensive basic science, pharmacological preclinical and clinical studies have been performed. At the time of human translation, virtually all drugs that were effective in mice have demonstrated lack of efficacy in humans, underscoring a robust translational gap. CHRFAM7A, a uniquely human fusion gene, was discovered in 1998 as an example of human specific fusion genes that have occurred since the human-chimpanzee divergence. CHRFAM7A functional studies are sparse, and are lacking with the human context. CHRFAM7A gets incorporated into the α7 nAChR pentamer and changes its phenotype.

Added value of this study

We demonstrate functional readouts for the CHRFAM7A alleles for two phenotypic readouts. For these readouts the inverted allele is a null. This functional genetic knowledge is the key for human translation, as it splits the population 1:3 for non-carriers to carriers of the functional direct allele. To demonstrate the translational gap, we performed two double blind pharmacogenetic studies for both first exposure and disease modifying effect.

Implications of all the available evidence

This is the first proof of concept study that genotype and mechanism specific treatment is feasible in AD. CHRFAM7A non carriers, 25% of the AD population accounting for approximately 1.5 million people with AD in the US, could benefit from α7 nAChR selective drugs. Several compounds have been studied extensively in phase 1 and 2 clinical trials, with valuable data to support and accelerate α7 nAChR targeting drug development. Those efforts need to be continued with trial design incorporating CHRFAM7A pharmacogenetics. More broadly, FDA approved AChEI drugs may have a DMT effect in non-carriers and need to be tested in preclinical AD and aMCI. The neuronal toxicity data suggests that non-carriers are more sensitive to Aβ, thus agents that reduce amyloid burden could be effective in non-carriers. Future basic science experiments that characterize the function of CHRFAM7A in the human context may results in novel targetable pathways for AD based on the endophenotype. As α7 nAChR is implicated in a broad range of biological processes and diseases including cognition, memory, schizophrenia, systemic inflammation, sepsis and chronic pain, mechanistic insights into CHRFAM7A function will have an impact on all of these diseases.

Alt-text: Unlabelled box

Introduction

Cholinergic neuronal loss is the hallmark of AD related neurodegeneration. Among the nicotinic acetylcholine receptors (nAChR), α7 has been hypothesized to account for the selective neuronal vulnerability in early AD. Consequently, α7 nAChR has been an active drug target for decades [1,2]. Preclinical promise of α7 nAChR agonists and allosteric modulators in animal models failed to translate into humans [] and frequently contradicting results emerged in the human context. CHRFAM7A, a human specific fusion gene between CHRNA7 and ULK4 [4], is a putative negative regulator of the α7 nAChR [5,6]. We hypothesized that CHRFAM7A may contribute to the preclinical to clinical translational gap in α7 nAChR targeting therapy.

CHRFAM7A evolved as a result of a series of recombination events including duplications, deletion and inversion in the humanoids [4,]. Three distinct alleles emerged: the ancestral allele lacking the fusion gene (0 copy), the fusion gene in direct orientation (CHRFAM7A), and the fusion gene in inverted orientation characterized by a 2 bp deletion in exon 6 (CHRFAM7AΔ2 bp) []. The inverted gene is expressed at the RNA level [], but protein prediction algorithms suggest that translation is unlikely due to Kozak fragment length []. CHRFAM7A gets incorporated into the α7 nAChR homopentamer based on gene dosage [9,10] and functions as a dominant negative modulator [5,6]. We previously showed that CHRFAM7A decreases α7 nAChR channel open probability and mitigates Aβ1–42 uptake beyond physiological concentrations [11]. Recent data suggests that CHRFAM7A may have a role in the cholinergic anti-inflammatory pathway [12]. In this study we asked the question how CHRFAM7A and CHRFAM7AΔ2 bp modify receptor pentamer function in the human context and whether this could account for the translational gap in α7 nAChR targeting, and more broadly in cholinergic therapies. We used human induced pluripotent stem cells (iPSC) as a model system and translated the readout into human clinical data using double blind pharmacogenetic study design.

Materials and methods

Detailed methods are described in the supplementary information.

Ethical statement, skin biopsy and genotyping

The Institutional Review Board approved the study. The informed consents were obtained from the donors. Subjects requiring legally authorized representatives were excluded from the study.

iPSC cell culture

iPSCs were grown on irradiated mouse embryonic fibroblasts in DMEM/F12-Glutamax medium supplemented with 10% KnockOut Serum Replacement, 1% Non-essential Amino Acids (NEAA), and 0.1% 2-Mercaptoethanol (all Thermo Fisher). Cells were maintained at 37 °C/5% CO2 and subcultured every 4–6 days using Dispase (Thermo Fisher).

MGE progenitors differentiation, culture, and transfection

Neuronal differentiation of iPSC towards Medial Ganglionic Eminence (MGE) progenitors was carried out as described previously[13,14]. MGE cell culturing is described in details in Supplementary materials. MGE progenitors were transfected with either pcDNA3.1-CHRFAM7A-mCherry, pcDNA3.1-CHRFAM7ADelta2bp-mCherry (both a gift from Henry Lester (Addgene plasmid # 62,635, Addgene plasmid # 62,638, respectively[15]) or with pcDNA3.3-mCherry (a gift from Derrick Rossi (Addgene plasmid # 26,823 [16],) constructs according to Ma et al. [17].

HEK 293 cell culture and transfection

Human embryonic kidney (HEK) 293 cells were maintained in Dulbecco's minimal essential medium (DMEM) supplemented with 10% fetal bovine serum and 1% penicillin–streptomycin, pH 7.4. Human α7 nAChRs and CHRFAM7A were expressed in HEK 293 cells by transient transfection (CaPO4 precipitation method) of these cDNA in ratios of 4:1 or 1:4. To aid surface expression of α7, we co-transfected intracellular chaperones Ric-3 (resistance to inhibitors of cholinesterase 3 [18]) and NACHO (transmembrane protein TMEM35A [19]) in 1:1:1 ratio in all experiments.

Electrophysiology

Whole cell and single-channel currents were recorded in the cell-attached patch configuration as described previously [20] (Supplementary data, Methods). Kinetic analyses of single channel currents were performed by using QuB [21]. Single channel currents were idealized by segmental k-means algorithm (SKM). nPo was estimated by dividing the cumulative open probability by the number of channels in the patch (maximum number of overlaps of open current levels in the data) as follows:

Generation of isogenic iPSC lines, electroporation, and colony selection

TALEN-mediated KI of direct CHRFAM7A (UB068i_KI) into the AAVS1 safe harbor site of the constitutively expressed gene PPP1R12C on chr19 (position 19q13.42) of the 0 copy ancestral line (UB068) was performed. The expression vector (GeneCopoeia (Cat. No.: DC-DON-SH01 - AAVS1 donor vector)) harbours puromycin resistance gene for selection and fluorescent signal (GFP) for visualization. Individual GFP expressing colonies were picked, PCR screened for the insertion, and passaged. Breakpoint specific TaqMan assay and RT-qPCR were used to quantify copy number and expression level of the inserted gene. Whole genome sequencing was performed to confirm transgene insertion and exclude off target mutations.

Amyloid beta uptake by fluorescent microscopy

Unlabelled and fluorescently labelled Aβ1–42 (both from AnaSpec) were reconstituted according to the manufacturer's protocol. The species of Aβ1–42 was confirmed by measuring concentration and absorbance at 280 nm and using Beer's Law (extinction coefficient = absorbance/concentration/path length). We derived the extinction coefficient at 1280M-1 cm-1 corresponding to the 42 amino- acid peptide with MW 4514. Aβ1–42 was diluted in 1% NH4OH and PBS was added to reach the concentration of 1 mg/ml, briefly sonicated and either used immediately or aliquoted and kept at −20 °C.

Amyloid beta uptake was performed using fluorescently labelled Aβ1–42 as described previously [22]. Uptake was pathologically characterized by a blinded pathologist and two independent raters for positive cells, grade (no uptake, low grade, high grade) and intensity measured by ImageJ plugin. α7 nAChR mediated endocytosis was characterized by co-localization of CHRNA7, acidic compartment (LysoTracker) and fluorescently labelled Aβ1–42 (HiLyteFluor555).

Flow cytometry

MGE progenitors grown on 12 well plates (400,000 cells/well) were treated with various concentrations of HiLyteFluor 488- Aβ1–42 or Aβ1–42 HiLyteFluor 555- Aβ1–42 (1 nM to 500 nM) for 18 h. The cells were detached with Accutase, pelleted by centrifugation at 200xg for 5 min, and washed three times with PBS. Flow cytometry was performed using LSRII-Fortessa with FACS DIVA (BD Biosciences). Data were analysed using FlowJo software (https://www.flowjo.com/).

ELISA

After Aβ1–42 treatment, total cell lysates were collected and stored at −80 °C. Concentration of Aβ1–42 in the cell lysates was estimated using a human specific high sensitivity Aβ1–42 ELISA kit (Thermo Fisher) according to the manufacturer's protocol.

Lactate dehydrogenase (LDH) cytotoxicity assay

The LDH assay was carried out according to the manufacturer's protocol. Briefly, MGE progenitors derived from the UB068, UB068_CHRFAM7A, and the UB052 lines were seeded in 96‐well plastic plates (20,000 cells/well). The cells were cultured for 24 h prior to treatment with either AChEIs (50 μM donepezil, 50 μM rivastigmine) or a specific α7 nAChR agonist, encenicline (50 μM,) followed by 48 h treatment with Aβ 1–42 (7,5 μM). LDH release was detected spectrophotometrically. Absorbance was measured at 450 nm and 600 nm using BioTek microtiter plate reader.

ApoTox-Glo Cytotoxicity and Caspase-GLO 3/7 Assay

ApoTox-Glo (Promega) Assay to measure cytotoxicity and caspase activity was done according to the manufacturer's protocol. Live and dead cell protease activities were measured using fluorogenic cell-permeant (live) and cell-impermeant (dead) peptide substrate. Fluorescence was measured at 400Ex/505Em for viability and 485Ex/520Em for Cytotoxicity. Caspase 3/7 activity was measured by luminescence after adding Caspase-Glo 3/7 reagent to all wells.

Study cohorts

The University at Buffalo Alzheimer's Disease and Memory Disorders Center clinical research cohort is an observational study with genetic component [23,24]. Inclusion criteria included consensus diagnosis of aMCI or AD, documentation of time of initiation of AChEI therapy, baseline MMSE within 3 months prior to AChEI initiation, at least 1 follow up assessment within 3–7 months after AChEI initiation, age >55, male or female, any race. Memantine was allowed on stable doses between baseline and follow up MMSE. Exclusion criteria included other brain or systemic disease or medication use that could account for the cognitive impairment. All individuals who met IC/EC were enrolled.

The Texas Alzheimer Research and Care Consortium (TARCC) is a longitudinal multicentre observational study [25]. Inclusion criteria was defined as mild-moderate AD, at least 1 follow up assessment with at least 1year interval, age >55, male or female, any race and actively being treated with one of the two AChEI (donepezil, rivastigmine). Memantine was allowed as it is standard of care in moderate stage AD and was built into the statistical model. Exclusion criteria from the TARCC cohort included other brain or systemic disease or medication use that could account for the cognitive impairment. All individuals who met IC/EC were enrolled.

Statistical analysis

Experiments in the iPSC model system were performed in triplicates or quadruplicates. Values are expressed as means ± S.D. or ± SEM, as indicated in figure legends. Statistical significance was determined by an unpaired Student's t-test (two-tailed). For cumulative distribution comparison we used two-sided two sample Kolmogorov–Smirnov test. P values less than 0.05 were deemed statistically significant.

Clinical data analysis: Power calculation was not performed due to the exploratory nature of this pharmacogenetic study thus sample size was not predetermined. Significance was set at p < 0.05. Change in MMSE score from baseline was performed by 2-tailed T-test. The distributional assumption for the T-test was examined by QQ plot, histogram along with Shapiro-Wilk normality test. Longitudinal analysis of clinical outcome (MMSE) was performed using a fitted general linear model, based on an assumed autoregressive covariance structure. Model independent variables included age, sex, medication regimen at the time of the first utilized outcome measure (AChEI alone or AChEI plus memantine), APOE4 carrier status (0, 1 or 2 alleles as categorical variables) and CHRFAM7A genotype (carrier versus non-carrier of the functional direct allele) and interaction terms based on CHRFAM7A genotype with medication regimen and drug exposure. Type 3 test of model effects were performed in conjunction with a 0.05 nominal significance level. SAS (Cary, NC) version 9.4 statistical software was used for all analyses.

Results

CHRFAM7AΔ2 bp is a null allele for two Alzheimer's Disease (Ad) relevant phenotypes: α 7 nAChR channel opening and α7 nAChR mediated Aβ1–42 uptake

In order to define the pharmacogenetic analysis paradigm we set out to elucidate the functional role of the CHRFAM7A and CHRFAM7AΔ2 bp alleles in two phenotypic readouts: electrophysiological studies using single channel patch-clamp on transfected Human embryonic kidney (HEK) 293 cells, and Aβ1–42 uptake in medial ganglionic eminence (MGE) progenitors, the precursors of basal forebrain cholinergic neurons (BFCN).

In electrophysiological studies, transfection of HEK293 cells with CHRNA7 (pcDNA3.1-CHRNA7-GFP) was used as a 0 copy model. Co-transfection of CHRNA7 with CHRFAM7A (pcDNA3.1-CHRFAM7A-GFP) (direct model) or CHRNA7 with CHRFAM7AΔ2bp (pcDNA3.1-CHRFAM7AΔ2bp-GFP) (inverted model), both in ratios 1:4, as well as transfection of CHRNA7 (0 copy model) all detected PNU 120,596 (positive allosteric modulator of α7nAChR) responsive channel opening, suggesting α7 nAChR specificity (Fig. 1a–d). The 0 copy model and the inverted model demonstrated similar activation patterns, response to PNU 120,596 and overlapping distributions of channel open probability (Fig. 1a, c, d). In contrast, CHRFAM7A transfected cells (direct model) had a decreased channel open probability, similar to the previously reported UB052 iPSC line, carrier of CHRFAM7A [11]. To aid surface expression of α7, co-transfection with intracellular chaperones Ric-3 [18] and NACHO [19] in 1:1:1 ratio was applied in all experiments.

Fig. 1

Inverted CHRFAM7AD2 Bp is a null allele

Single channel traces and open dwell time for wild type α7, with or without the presence of CHRFAM7A in HEK 293 cells. Recordings were made at cell-attached patch configuration and currents were elicited by ACh and PNU 120,596, on HEK 293 cells transfected with α7nAChR alone (a), α7nAChR and CHRFAM7A direct (b); α7nAChR and CHRFAM7A inverted (c) cDNAs in 1:4. Corresponding histograms of channel open time in logarithmic time axis are shown for α7nAChR alone (gray), in the presence of CHRFAM7A (red) and CHRFAM7AΔ2bp (blue) (n = 5 from 32 independent cultures). e, Transfection of UB068 (0 copy) with CHRFAM7A causes a decrease in Aβ1–42 uptake in a concentration-dependent manner compared to transfection with empty vector (EV) analysed as mean fluorescent intensity by flow cytometry. In contrast, transfection with CHRFAM7AΔ2bp does not affect Aβ1–42 uptake via the a7nAChR. Data are presented as mean ± SD. * - P < 0.05 - difference between Aβ1–42 uptake in CHRFAM7A transfected cells compared to EV-transfected cells (T-test) at each given Aβ1–42 concentration (n = 3 from 3 independent cultures). f, Representative histograms of total gated events (∼ 10,000 cells were analysed by flow cytometry) for median samples are presented. Gated on a uniform, normally distributed FSC-A/SSC-A population with doublet discrimination through secondary FSC-A/FSC-H analysis (n = 3 from 2 independent cultures). g, Overexpression of CHRFAM7A and CHRFAM7AΔ2bp in UB068 using EV as transfection control indicates that the inverted allele is a null demonstrating similar Aβ1–42 uptake as EV quantified by cells counts (at least 50 cells from 5 fields/condition in 4 independent cultures were counted by a blinded pathologist and two independent raters). h, Quantification of Aβ1–42 in total cell lysate of MGE progenitors derived from UB068 line and transfected with CHRFAM7A, CHRFAM7AΔ2bp, and EV constructs. The cells were incubated with indicated concentration of Aβ1–42 for 24 h and the amount of Aβ1–42 was detected by ELISA. Data are presented as mean ± SD. * - P < 0.05, ** - P < 0.01 - difference between Aβ1–42 uptake in CHRFAM7A transfected cells compared to EV-transfected cells (T-test) at each given Aβ1–42 concentration (n = 3 from 3 independent cultures). i, Representative confocal images show localization of Aβ1–42 (yellow), LysoTracker Deep Red (magenta), and CHRNA7 (blue) in MGE progenitors derived from UB068 line (upper panel). Nuclei were stained with DAPI (cyan). Cellular uptake of Aβ1–42 overlaps with LysoTracker Deep Red and CHRNA7 (lower panel). Co-localization of Aβ1–42 / LysoTracker and CHRNA7/LysoTracker is indicated by white arrows. Scale bar, 10 µm. The cells were incubated for 24 h with Aβ1–42 (100 nM), washed and incubated for an additional 5 h with LysoTrackerDeep Red (n = 5 from 3 independent cultures). j, Visual grading demonstrates qualitative differences in Aβ1–42 uptake: CHRFAM7AΔ2bp behaves as a null demonstrating similar Aβ1–42 grade distribution as EV; CHRFAM7A prevents the formation of larger Aβ1–42 deposits (high grade) at 100 nM of Aβ1–42 concentration for 24 h. Data are presented as mean ± SD. * - P < 0.05 - difference between high grade Aβ1–42 uptake in CHRFAM7A transfected cells compared to EV-transfected cells, T-test. Inset: representative images show low, high, and no Aβ1–42 uptake by MGE progenitors overexpressing CHRFAM7A, CHRFAM7AΔ2bp or EV (at least 50 cells from 5 fields/condition in 4 independent cultures were counted by a blinded pathologist and two independent raters). k, Grading was validated by semi-automated intensity measure using ImageJ. Distribution curves of Aβ1–42 uptake in MGE progenitors overexpressing CHRFAM7A, CHRFAM7AΔ2bp or EV are presented. Aβ1–42 uptake in MGE progenitors overexpressing CHRFAM7A was significantly lower than in those overexpressing CHRFAM7AΔ2bp (P = 0.037; two-sided p-values from two sample Kolmogorov–Smirnov test).

As the α7 nAChR binds Aβ1–42 with high affinity, we performed genotype phenotype correlation for Aβ1–42 uptake. Previously, we showed that CHRFAM7A mitigates Aβ1–42 uptake, but the functional role of the CHRFAM7AΔ2 bp allele remained unknown. To address this question, UB068 iPSCs[11] harbouring two ancestral alleles (0 copy) were differentiated into MGE progenitors[26,14] and transfected with an empty vector, EV (pcDNA3.3-mCherry), CHRFAM7A (pcDNA3.1-CHRFAM7A-mCherry) and CHRFAM7AΔ2 bp (CHRFAM7ADelta2bp-mCherry)[17,11]. Expression of CHRFAM7A and CHRFAM7AΔ2 bp was confirmed by breakpoint specific RT-qPCR (Supplementary Fig. 1 a). Overexpression of either construct did not affect α7 nAChR expression and localization (Supplementary Fig. 1 a-d).

Concentration dependent Aβ1–42 uptake was quantified by flow cytometry (Fig. 1e,f) and cell counts (Fig. 1g). Fluorescent-Aβ1–42 uptake between concentrations of 10 nM to 250 nM demonstrated similar dose response curves in the EV and CHRFAM7AΔ2 bp transfected MGE progenitors (Fig. 1e - g). In contrast, CHRFAM7A transfected cells showed a mitigated Aβ1–42 uptake compared to EV beyond 50 nM (p < 0.05). Furthermore, ELISA revealed a significantly lower amount of Aβ1–42 in total cell lysate from CHRFAM7A transfected MGE progenitors compared to the ones transfected with EV (Fig. 1h). Areas of co-localization of Aβ1–42 with the acidic compartment (LysoTracker) were detected by confocal microscopy consistent with lysosomal engagement (Fig. 1i). CHRNA7 expression was broader and uniform along the axons and demonstrated spots of co-localization with Aβ1–42 and the acidic compartment. Fluorescent Aβ1–42 uptake at 100 nM was qualitatively assessed and further quantified by microscopy using grading and cell counts (Fig. 1j,k). Grading was validated by semiautomatic intensity measures (Fig. 1k). Differential uptake by MGE progenitors was confirmed by a comparing cumulative intensity distribution between CHRFAM7A and CHRFAM7AΔ2 bp transfected MGE progenitors (p = 0.037; Kolmogorov–Smirnov test) (Fig. 1k). CHRNA7 and CHRFAM7A gene expression levels were unaffected by Aβ1–42 treatment (Supplementary Fig. 1 e,f). While CHRFAM7A is an intrinsic dominant negative modulator, the inverted CHRFAM7AΔ2 bp functions as a null allele in regard of α7 nAChR channel opening and α7 nAChR mediated Aβ1–42 uptake.

Human iPSC models animal and human outcomes implicating CHRFAM7A as the translational gap

To study the effect of CHRFAM7A on drug response in the human context, we genome edited UB068 (CHRFAM7A null) iPSC line to express CHRFAM7A using TALENS mediated insertion of CHRFAM7A into the AVVS1 safe harbor site on 19q13.42 under the constitutively active CMV promoter (Fig. 2a). GFP was inserted in tandem under EF1a promoter to facilitate selection of successfully edited CHRFAM7A harbouring cells (Fig. 2b). The insertion was confirmed by PCR (Fig. 2c) and whole genome sequencing (WGS) (Supplementary Fig. 2 a, b). All lines were characterized for the expression of pluripotency markers at the gene (NANOG, OCT3/4, SOX2) and protein (TRA 1–60, Oct-4, Sox2) levels (Fig. 2b, Supplementary Fig. 2 c, d). Differentiation into three germ layers was assessed by the TaqMan® hPSC Scorecard™ Assay (Supplementary Fig. 2 e). WGS excluded aneuploidy, intragenic indels, and large chromosomal duplications or deletions (Supplementary Fig. 2 a, b). WGS showed a limited number of likely irrelevant de novo non-homonymous variants in ENST00000379458, MUC3A-201, ENST00000456556 and ANKRD36C in isogenic colony 8 and in ENST00000379458 and MUC3A-201 in isogenic colony 12 (Supplementary Tables S5 and S6). These non-homonymous changes have not been established as pathogenic. CHRFAM7A gene expression was consistent with the genotypes (Fig 2d). CHRFAM7A translation was confirmed by immunoblot analysis using CHRFAM7A N-terminus specific antibody (generous gift from Andrew Baird; Fig. 2e). CHRFAM7A did not alter gene expression of the α7 nAChR (Fig 2. d). The membrane localized signal detected with the loop antibody represents both CHRNA7 and CHRFAM7A proteins (Fig. 2f).

Fig. 2

Human iPSC models animal and human outcomes implicating CHRFAM7A as the translational gap

UB068 isogenic line (UB068i_CHRFAM7A) expressing CHRFAM7A in the AAVS1 safe harbor site. a, Schematic of the AAVS1-CHRFAM7A construct. 4 colonies from 2 independent electroporation experiments are being characterized. b, Representative image of TRA1–60 live staining of UB068i_CHRFAM7A line, passage 3. Scale bar, 200 µm. c, PCR verification of insertion: 1- PCR with primers for AAVS1–5′ homology arm, 2- PCR with primers for AAVS1–3′ homology arm, 3-PCR with primers spanning AAVS1-CHRFAM7A, 4-PCR with primers specific for CHRFAM7A d) RT-qPCR shows that overexpression of CHRFAM7A gene does not alter expression levels of CHRNA7 in the genome edited line (n = 3 from 3 independent cultures). e) Immunoblot analysis using an antibody directed to the unique N-terminus sequence of CHRFAM7A (gift from Andrew Baird) confirms its expression in the UB068_CHRFAM7A line; positive control the previously reported UB052 CHRFAM7A carrier line UB052 (2 independent cultures). f) Representative confocal images demonstrate that the presence of CHRFAM7A does not affect membrane localization of CHRNA7 (n = 10 cells Z-stack images from 2 independent cultures). Scale bar, 10 µm. Compared to UB068, the presence of CHRFAM7A (both in UB052 and UB068_CHRFAM7A) causes a significant decrease in Aβ1–42 uptake detected by cell counts (n = 3 from 3 independent cultures) (g, h) and flow cytometry (i) (Triplicates from 2 independent cultures). For h-i data are presented as mean ± SD. * - P < 0.05 ** - P < 0.001 - difference between Aβ1–42 uptake in UB068_CHRFAM7A line compared to UB068 (T-test) at each given Aβ1–42 concentration. Aβ1–42-induced cytotoxicity increases in a concentration-dependent manner measured by LDH release (j) in MGE progenitors differentiated from UB068, UB068_CHRFAM7A, and UB052 lines (n = 3 from 3 independent cultures). Data are presented as mean ± SD. * - P < 0.05 - difference between Aβ1–42 uptake in the UB068_CHRFAM7A line compared to UB068 (T-test) at each given Aβ1–42 concentration. k), 24 h pre-treatment with donepezil (Don) decreases Aβ1–42 uptake by MGE progenitors derived from UB068, but not from UB068_CHRFAM7A or UB052 lines. Data are presented as mean ± SD. * - P < 0.05 - difference in Aβ1–42 uptake between the cells pre-treated with donepezil (Don, 50 µM) followed by Aβ1–42 (250 nM, 24 h) compared to Aβ1–42 alone. Cytotoxic effect of Aβ1–42 (7,5 μM, 48 h) on MGE progenitors derived from the three lines was reduced by the pre-treatment with donepezil (Don), rivastigmine (Riv) and encenicline (Enc) only in UB068 line (null), while it remained unchanged in UB068_CHRFAM7A and UB052 as detected by LDH cytotoxicity assay (l), CytoTox-Glo cytotoxicity assay (m) and Caspase-Glo 3/7 assay (n). o) RT-qPCR of BAX, apoptosis activator, corresponds to the CAS3/7 readout suggesting decreased apoptosis in non-carriers compared to carrier lines (UB052, UB068_CHRFAM7A) with donepezil (Don) treatment. In the carrier lines donepezil increase BAX expression. For panels l - o data are presented as mean ± SD. * - P < 0.05 ** - P < 0.001 *** - P < 0.0001 - difference between non-treatment control and Aβ1–42 and between Aβ1–42 and drug pre-treatment. # - P < 0.05 - difference between non-treatment controls in UB052 and UB068_CHRFAM7A compared to UB068 (T-test). All results presented in panels l - o represent average from 4 independent experiments.

To model selective neuronal vulnerability, iPSCs were differentiated into MGE progenitors, the precursors of BFCNs and GABA interneurons[26,14]. MGE progenitors generated from UB068_CHRFAM7A line were evaluated for the presence of the specific gene and protein markers (Supplementary Fig. 3 a, b). As a nascent CHRFAM7A carrier, UB052 was incorporated into the panel [11]. In all three lines, fluorescent Aβ1–42 uptake assessed by cell counts showed genotype specific dose response curves (Fig. 2g,h). Similar results were obtained by flow cytometry (Fig. 2i). In UB068, Aβ1–42 uptake was linear with dose, while in UB068_CHRFAM7A and in carrier (UB052) lines mitigated Aβ1–42 uptake was observed (Fig. 2g–i). CHRNA7 and CHRFAM7A expression remained unchanged upon dose-dependent Aβ1–42 uptake (Supplementary Fig.1. g, h). Neuronal toxicity measured by LDH release demonstrated congruent dose-response with Aβ1–42 uptake (Fig. 2j). Next, we tested the response of MGE progenitors to two FDA approved Acetylcholine Esterase Inhibitors, AChEIs (donepezil, rivastigmine) and encenicline, a selective α7 nAChR agonist. 24 h exposure to ACHEIs proved not to be toxic for the MGE cultures (Supplementary Fig. 4 a) and did not change the expression of CHRNA7 (Supplementary Fig. 4 b). In UB068, AChEIs and encenicline decreased Aβ1–42 uptake (Fig. 2k), Aβ1–42 induced cytotoxicity measured by LDH release (Fig. 2l) and by protease activity using bis-AAF-R110 Substrate (Fig. 2m), and apoptosis measured by Caspase3/7 activity (ApoTox-Glo Triplex assay, Promega) (Fig. 2n). Furthermore, expression of apoptosis regulator BAX correlated with Caspase3/7 activity results (Fig. 2o). UB068 (non-carrier) is genetically equivalent to the preclinical models used in drug selection. In the CHRFAM7A carrier models (UB068_CHRFAM7A and UB052) neurotoxicity was unchanged by AChEIs or encenicline (Fig 2l-n), and donepezil treatment resulted in an increase in apoptosis (Fig. 2n, o). These results suggest that non-carriers of the functional direct allele should benefit from α7 nAChR targeting therapies.

Population frequency of the CHRFAM7A alleles

Based on the functional readouts, the population frequency of the direct, functional CHRFAM7A carriers and non-carriers is the key to understanding the α7 nAChR translational gap. To establish population frequencies of the dosage and alleles, we genotyped 1174 samples from the Texas Alzheimer Research and Care Consortium, Baylor College of Medicine and NCRAD (supplementary information). Genotyping of CHRFAM7A dosage and orientation was performed using breakpoint specific TaqMan assay to decipher copy number, and capillary sequencing to determine orientation of the alleles. Frequency of CN dosage (0, 1, 2 and 3), and alleles (ancestral, direct and inverted) are depicted in Table 1. Non-carriers of the functional direct allele comprise 25% of the human population, while carriers of the direct CHRFAM7A functional allele is present in 75% of the population. Normal aged controls and AD subjects have similar allele frequencies, suggesting that CHRFAM7A is not associated with the disease phenotype, rather it has a pharmacogenetic role for α7 nAChR targeting therapies.

Table 1

Population frequency of the CHRFAM7A dosage and alelles.

	AD (N = 490)		NC (N = 657)
CN frequencies	N	%	N	%
0	5	0.01	6	0.01
1	98	0.20	115	0.18
2	378	0.77	526	0.80
3	9	0.02	10	0.02
Allele frequencies
Ancestral	106	0.11	123	0.09
Direct	483	0.49	639	0.49
Inverted	391	0.40	552	0.42
Carrier frequency
Carrier	362	0.74	476	0.73
Non-carrier	128	0.26	181	0.28

CHRFAM7A non-carriers respond to ACHEI translating from preclinical results

Next we sought to test the α7 nAChR translational gap hypothesis in human data. Upon reviewing publicly available data on the α7 nAChR clinical trials [2,27], it became apparent that most trials were designed to test the receptor mediated cognitive enhancement as the primary outcome measure in relatively short trials of 3–6 months duration [2] and clinical trial data with DNA from the same individuals were not available. As an alternative approach, although likely less sensitive, exposure to AChEI (increasing ACh, agonist of the α7nAChR) and using CHRFAM7A genotype as an α7nAChR specific readout were used as a proof of principal study. Response to initiation of therapy and delayed disease modifying effect werestudied in two cohorts; the UB cohort had information on treatment initiation that was used to assess short term response, while the TARCC cohort had up to 7 years follow up that measured disease modifying treatment (DMT) effect.

For the initiation of treatment response pharmacogenetic analysis we used the UB ADMDC clinical cohort. The study fulfilled double blind criteria as at the time of assessment neither subject nor investigator was aware of the genotype. In these subjects the date of first prescription was systematically documented mimicking synchronous initiation of treatment. 63 subjects met inclusion criteria (supplementary information). 8 subjects were excluded for missing DNA sample. Demographic information is summarized in Table S3. Over 90% of the cohort was Caucasian. All subjects were on maximum doses of an AChEI, 96% on donepezil. Change in MMSE from baseline (pre-exposure) within 4–7 months of treatment initiation were compared by two-tailed T-test. Non-carriers of direct CHRFAM7A allele had superior response to the AChEI as predicted from the iPSC model (p = 0.037) (Fig. 3a). APOE4 carrier status had no effect (Fig. 3b). Individual level data supports the responder rates (Fig. 3c, d).

Fig. 3

CHRFAM7A non-carriers have a superior response to first exposure of AChEI

Pre- and post-exposure double blind pharmacogenetic study using MMSE as primary outcome measure. Response to initiation of ChEI therapy in 55 subjects. (a) CHRFAM7A non-carriers demonstrate superior response to first exposure when compared to carriers (p = 0.037; two-tailed T-test). (b) APOE4 carrier status does not affect first exposure response to AChEIs. Individual level pre- and post-exposure data in CHRFAM7A carriers (c) and non-carriers (d).

We used the TARCC longitudinal dataset to perform the DMT pharmacogenetic analysis. 345 subjects met inclusion criteria (supplementary information). Demographic information is summarized in Table S4. Over 90% of the cohort was Caucasian. 34.87% of the subjects were on maximum doses of an AChEI (mild AD) and 65.13% on AChEI and memantine at the first assessment. We used general linear model including age, sex, medication regimen at the time of visit 1 (AChEI alone or AChEI plus memantine), APOE4 carrier status (0, 1 or 2 alleles as categorical variables) and CHRFAM7A genotype (carrier versus non-carrier of the functional direct allele) and interaction terms including visit*CHRFAM7A genotype and first treatment*CHRFAM7A genotype. We found that CHRFAM7A non-carriers of the functional direct orientation allele had a statistically significant benefit from exposure to AChEI after controlling for the other variables (Fig 4a, Table 2) over the 7-year observation period. The effect was 5 points difference on the MMSE and the separation of the curves started after 3 years of follow-up. The delay in separation of the curves is likely due to the non-synchronized drug initiation. Individuals with an upslope in the first year likely were just started on an AChEI, while others were enrolled after years of drug exposure. The desynchronized drug initiation may wash out the earlier signal. However, the percentage of subjects meeting the 5-point decline cut-off at 1, 2, 3 and 4 years of follow up demonstrates the difference in carrier and non-carrier groups at those earlier time points. APOE4 carrier status (0, 1 or 2 alleles) did not demonstrate a treatment response effect (Fig 4b). Individual level data and demographics are shown in Fig. 4c, d and Table S4. CHRFAM7A carriers had minimal benefit if any compared to the natural history of AD cognitive decline. The non-carrier individuals demonstrate the outcome predicted from preclinical data. These data suggest a disease modifying effect in a genotype specific manner.

Fig. 4

CHRFAM7A non-carriers have a DMT benefit from AChEI

Longitudinal double blind pharmacogenetic study using MMSE as primary outcome measure demonstrating DMT effect in the TARCC cohort over a 6-year observation period (p = 0.0048). Mean plot shows the mean MMSE scores across all visits grouped by CHRFAM7A genotype (a) and by number of APOE4 alleles (b). Attrition is depicted below the curves. Individual profile plots demonstrate the MMSE score for each individual across all visits in CHRFAM7A non-carries (c) and carriers (d).

Table 2

Type 3 Tests of Fixed Effects for MMSE as outcome.

Effect	Num DF	Den DF	F Value	Pr > F
visit	7	805	49.08	<0.0001
age	1	298	4.78	0.0295
gender	1	298	7.84	0.0054
CHRFAM7A	1	298	8.08	0.0048
ApoE	2	298	1.25	0.2868
firsttrt	1	298	16.96	<0.0001
CHRFAM7A*visit	7	805	2.23	0.0304
CHRFAM7A*firsttrt	1	298	0.54	0.4611

Discussion

The α7 nAChR has been a promising target for diseases affecting cognition and higher cortical functions; however, the effect observed in animal models failed to translate into human clinical trials identifying a translational gap. CHRFAM7A is a human specific fusion gene between CHRNA7 and FAM7A/ULK4, and as it is not present in any other species the CHRFAM7A effect was not accounted for in preclinical studies. We hypothesized that CHRFAM7A may account for this translational gap and understanding its function may offer novel approaches to explore α7 nAChR as a drug target in a genotype specific manner. From previous experimental and in silico work the inverted allele was predicted to be non-translated, thus non-functional [7]. To experimentally validate the prediction, we performed functional characterization of the inverted and direct alleles using two disease relevant phenotypes, electrophysiology and Aβ1–42 uptake with associated neuronal toxicity using iPSC model. CHRFAM7AΔ2 bp allele behaved as a functional null for channel function, while CHRFAM7A decreased channel open time indicating dominant negative effect. In the presence of CHRFAM7A allele, Aβ1–42 uptake was mitigated at post-physiological levels [11] while CHRFAM7AΔ2 bp showed similar dose response as the null line. Both neuronal toxicity and Aβ1–42 uptake were dose-dependent and were mitigated by CHRFAM7A, but not by CHRFAM7AΔ2 bp. The null alleles (ancestral and inverted) are equivalent to the preclinical animal models.

To study the effect of CHRFAM7A on drug response in the human context, we used a panel of iPSC lines including non-carrier (UB068), nascent CHRFAM7A carrier (UB052); and we genome edited UB068 (CHRFAM7A null) iPSC line to express CHRFAM7A for isogenic control. To model selective neuronal vulnerability, iPSCs were differentiated into MGE progenitors, the precursors of BFCNs and GABA interneurons [26,14]. In UB068, Aβ1–42 uptake was linear with dose, while in UB068_CHRFAM7A and in carrier (UB052) lines Aβ1–42 uptake was mitigated. Parallel to uptake, similar dose response was detected for cytotoxicity measured by LDH release. AChEIs and encenicline, an α7 nAChR agonist, demonstrated a beneficial effect on Aβ1–42-induced cytotoxicity in the non-carrier line in contrast with the carrier lines. Change in apoptosis appeared to be the driving cell death mechanism for the detected difference. To establish population frequencies of the alleles, we genotyped 1147 samples using breakpoint specific TaqMan assay to decipher copy number, and capillary sequencing to determine orientation of the alleles. Locus specific genotyping is required to detect CHRFAM7A as the only unique sequence is the breakpoint sequence. The inversion is even more challenging to detect in general, however, in this case the Δ2 bp deletion allows a high throughput assay. These two characteristics, being a fusion gene and the inversion event, make CHRFAM7A elusive on SNP and CNV microarrays and even WGS underperforms due to miss-mapping of the short reads. Locus specific genotyping using sensitive technology and functional characterization established that non-carriers of CHRFAM7A compose 25% of the population in Caucasians, similar to responder rates in AChEI clinical trials [28]. As preclinical development is carried out largely in nonhuman models, drugs screened for α7 nAChR as a target, without accounting for CHRFAM7A, will likely benefit only 25% of the population; for the 75% we need models that harbor CHRFAM7A. The 25%−75% split washes out the effect in the 25%. Previous genotyping efforts reported markedly different allele frequencies [8,29] in African-American, Caucasian and Hispanic populations suggesting that clinical trial admixture, or lack of it, may also affect efficacy readouts.

As α7 nAChR clinical trial data with DNA samples were not available, we used exposure to AChEI (increasing ACh, an agonist of the α7 nAChR) and CHRFAM7A genotype as an α7 nAChR specific readout as a proof of principal study. We performed double-blind pharmacogenetic analysis on the effect of AChEI therapy based on CHRFAM7A carrier status in two paradigms: response to drug initiation and DMT effect. Both paradigms demonstrated the benefit for the CHRFAM7A non-carriers consistent with the prediction. In the DMT analysis to align with the clinical trials that led to FDA approval of the three AChEI drugs, we included mild-moderate AD subjects. According to practice standards, the majority of subjects were on maximum doses of an AChEI (mild AD) or AChEI and memantine (moderate AD) and continued on the same regimen throughout the study or memantine was added when moderate stage was reached. We found that as predicted, CHRFAM7A non-carriers of the functional direct orientation allele had a statistically significant benefit from exposure to AChEI after controlling for the other variables. Of note, the study is double blind as neither subjects nor rater had knowledge of the genotype at the time of assessments. The attrition rate is high; however, it is expected in a study of this length in this vulnerable population. CHRFAM7A carriers had minimal benefit, if any, compared to the natural history of AD cognitive decline. These findings suggest that CHRFAM7A may account for the translational gap in α7 nAChR, and more broadly cholinergic therapies. The non-carrier individuals demonstrate the outcome predicted from preclinical data and the frequency of the non-carriers in the AD population is consistent with the responder rates from the AChEI trials. CHRFAM7A carriers likely require different small molecules screened in a model that harbours CHRFAM7A.

Mechanistic insights from the in vitro work suggests competitive binding between ACh and Aβ1–42 to the α7 nAChR as both agonist and antagonist mitigates Aβ uptake. The α7 nAChR binds Aβ with high affinity [30] and internalizes Aβ into the endosomes/lysosomes and mitochondria [31,32]. Uptake of Aβ phosphorylates p38, induces apoptosis [31] and α7 nAChR agonists mitigate the Aβ-induced apoptosis in animal models [33]. Further studies are needed to elucidate the mechanistic link between receptor structure and Aβ binding with specifically designed studies incorporating genetic and pharmacological paradigms.

While the presented work is focusing on receptor function, another possible mechanism needs to be considered: the control of neuroinflammation by donepezil through the cholinergic anti-inflammatory pathway. The clinical readout is from chronic drug exposure, which raises the possibility of upregulation of CHRNA7 as demonstrated in rat models after long-term exposure to donepezil, 4 days [34,35]. Interestingly, donepezil treatment decreases LPS induced neuroinflammation in rats [36]. These data suggest that donepezil has an anti-inflammatory effect by upregulating CHRNA7 and, as a result, enhancing the α7 nAChR mediated cholinergic anti-inflammatory pathway. In humans, the presence of CHRFAM7A refines the neuroinflammatory response. In human primary macrophages, donepezil upregulates both CHRNA7 and CHRFAM7A which likely increases the anti-inflammatory tone [12]. While upon LPS stimulation CHRFAM7A is downregulated [37], concomitant donepezil and LPS exposure synergistically upregulates CHRNA7 and extinguishes the opposite effects on CHRFAM7A [12]. As a result, in the absence of stimulation by Pathogen-associated molecular pattern molecules (PAMPs) and perhaps Damage-associated molecular patterns (DAMPs) an anti-inflammatory milieu is set by donepezil [12] . Once an immune trigger in present, such as LPS or Aβ, CHRFAM7A refines the inflammatory tone to activate the innate immune response [11,36]. These immune regulatory differences between rat (non-carriers of CHRFAM7A) and humans (studied lines are CHRFAM7A carriers) could account for the DMT effect in non-carriers by controlling Aβ (DAMP) triggered neuroinflammation.

While randomized, double blind, placebo controlled studies are needed, these findings have several important implications. First, this is the first proof of concept study that genotype and mechanism specific treatment is feasible in AD. Second, CHRFAM7A non-carriers, 25% of the AD population accounting for approximately 1.5 million people with AD in the US could benefit from α7 nAChR selective drugs. Several compounds have been extensively studied in phase 1 and 2 clinical trials with valuable data to support and accelerate α7 nAChR targeting drug development. Those efforts need to be continued with trial design incorporating CHRFAM7A pharmacogenetics. Third, FDA approved AD drugs (AChEIs) may have a DMT effect in non-carriers and need to be tested in preclinical AD and aMCI. Fourth, 75% of patients have an intrinsic modulator for α7 nAChR, thus compounds need to be screened for this group in the relevant model harbouring CHRFAM7A. Fifth, agents that reduce amyloid burden could be more effective in non-carriers as the same Aβ dose reduction has a larger effect based on the uptake and toxicity dose response-curves. The vast amount clinical data on amyloid reducing agents, including monoclonal antibodies and BACE inhibitors, should be reanalysed based on this concept, as they may have had an effect in 25% of patients.

There are several limitations to our study that invite caution and further work. The clinical studies presented build on observational cohorts, however they do fulfill the double-blind criteria as assessments were performed prior to genotyping thus both subjects and raters were blind to the genotype. Although placebo was not used in its traditional sense, the predicted lack of efficacy in the CHRFAM7A carriers serves as a predefined comparison group. Attrition rate was high in the DMT study, inviting for caution, although the individual plots do not suggest a systemic effect. The delay in the observed difference could be the results of non-synchronous initiation of therapy, as prior to the time of enrolment subjects were treated for variable periods of time. This non-synchronous start of drug exposure likely reduces power thus works against the findings. The clinical data needs validation in additional datasets and placebo controlled, randomized, double blind studies are needed to detect the DMT effect of α7 nAChR targeting drugs, and more broadly cholinergic strategies. Furthermore, the preclinical iPSC model is from a limited number of representative individual for carriers and non-carriers. A larger panel of iPSC is needed to serve as model system for small molecular screen with population relevance.

Although much work is needed, these observations open new opportunities to explore completed clinical trials. For example, post hoc reanalysis of the AChEI trials in aMCI based on CHRFAM7A carrier status could detect a signal in the non-carriers in delaying conversion, thus offer an early disease modifying effect. In addition, as CHRFAM7A non-carriers have an Aβ1–42 dose-dependent neuronal vulnerability they are most likely to benefit from Aβ1–42 reduction, thus post hoc reanalysis of the plethora of anti-amyloid therapies may in fact find an effective agent for 25% of AD. For the 75% of carriers, we need relevant preclinical models to screen compounds and basic science to understand the impact of CHRFAM7A on AD pathogenesis. This pharmacogenetic paradigm may apply to other treatment strategies targeting α7 nAChR.

Declaration of interests

The authors declare no conflict of interest.

Author contributions

K.S. conceptualized and designed the study, both the laboratory experiments and clinical aspects, led the wet lab experimental design and data interpretation, performed the first exposure pharmacogenetic study and wrote the first draft of the manuscript. I.I. performed the iPSC experiments, contributed to data interpretation and worked extensively on the manuscript. B.B. contributed the genome editing of the iPSC line. Z.C. performed the statistical analysis for both pharmacogenetic studies. A.O. contributed with scoring in the wet lab experiments and ascertaining the UB clinical cohort. D.I.V performed the patch clamp experiments and contributed the electrophysiological sections of the manuscript. L.C. performed data capture and analysis of the flow cytometry experiment, and wrote those sections in the manuscript. N.S. developed the grading system and was the blinded pathologist for scoring and contributed those sections of the manuscript. J.S.R. performed the data management and extraction from the longitudinal TARCC database. V.P. contributed to the clinical data of the TARCC cohort and participated in the clinical data analysis. R.H.B. contributed clinical data collection, neuropsychological review of the UB cohort and revised the manuscript. A.A. designed the electrophysiological experiments and contributed to those sections of the manuscript. G.W. designed and managed the clinical data analysis for both pharmacogenetic studies and contributed those sections to the manuscript.