A non-human primate model of familial Alzheimer’s disease for translational medicine


 Aging is a primary risk factor of Alzheimer’s disease (AD), with the world-wide number of patients anticipated shortly to exceed 50 million. Despite extensive research efforts, no effective measures are available for its prevention or treatment due in part to a lack of human-like animal models. Here, we describe the generation of three mutant marmoset individuals in which exon 9 of the PSEN1 gene product is deleted (PSEN1-ΔE9); the ΔE9 mutations have been reported to cause early onset familial AD1-5. We used Transcription Activator-Like Effector Nuclease (TALEN) to delete the 3’ splice site of exon 9 in the marmoset PSEN1 gene; to this end, TALEN exhibits high genome-editing efficacy, generates few off-target effects, and produces minimal mosaicism. Indeed, whole genome sequencing and other analyses illustrated an inclusive absence of off-target effects and apparent absence of mosaicism. Fibroblasts obtained from newborns indicated occurrence of full-length presenilin 1 protein (PS1) caused by perturbation of PS1 endoproteolysis as well as an increased ratio of Aβ42/Aβ40 production, a signature of familial AD pathogenesis. To our knowledge, this is the first non-human primate model of familial AD. The model lines will be made available to the research community to facilitate the global fight against AD.


Introduction
AD is the most common neurodegenerative disease that deprives patients and their families of human dignity. The number of individuals with dementia in the world was approximately 50 million in 2019 6 , and AD accounted for 50-70% of all these cases 7 . Clinically, AD is characterized by early memory deficits followed by a decline in other cognitive functions 7,8 . The pathological changes associated with AD development precede the clinical manifestation by approximately 20 years 9 and comprise the following: deposition of amyloid β peptide (Aβ) as extracellular plaques, accumulation of hyperphosphorylated tau as intracellular neurofibrillary tangles (NFTs) and chronic neuroinflammation followed by neurodegeneration mainly in the cerebral cortex and hippocampus 10,11 . The genetic and pathological observations collectively support the A hypothesis depicting that A plays a central role in the AD pathogenesis 12 . Mouse AD models appear to have reached a highly advanced state of development by overcoming overexpression artefacts in recent times [13][14][15][16][17] . The amyloid precursor protein (App) knock-in mice that harbor familial AD mutations with the A sequence humanized 13 reproduced A pathology and neuroinflammation without overexpression of APP. These mice exhibited cognitive dysfunctions as analyzed by Intellicage 18 presumably through vasoconstriction by pericytes 19 , impairment of grid cells 16 and plaque-induced gene network 17 , providing the mechanistic insights into the action of A pathology.
These model mice however never exhibited tau pathology or neurodegeneration even when crossbred with human MAPT knock-in mice, in which the entire Mapt gene had been humanized 15,20 .
The reason for the absence of tauopathy and neurodegeneration in these models remains elusive, but this may simply be because mice live only as long as approximately two years whereas the entire pathological changes in human brain proceed over decades 9 . The discrepancy between mice and humans may also be accounted for by the species differences in genetics, neuroanatomy, immunity and metabolism. In addition, various higher cognitive functions specific to primates, represented by the presence of highly developed prefrontal cortex 21 , are affected in AD. In this respect, outcomes from mouse models are more suitable for application to preclinical studies than directly to clinical studies. We thus came to the conclusion that a non-human primate model is needed for "near"-clinical studies [22][23][24] . The developmental engineering of primates is, however, much more challenging than that of rodents; thus far we have created three mutant primates using 810 oocytes as described below.
Common marmosets (marmosets, Callithrix jacchus) are small non-human primates that belong to the New World Primates. They have been increasingly utilized in neuroscience because of advantages that arose over other research primates 22 (Supplementary Table 1). Marmosets possess genetic backgrounds, physiological functions, brain structures and complex cognitive/social behaviors similar to those of humans; they communicate mainly via visual and auditory measures. In association with AD research, the amino acid sequence of Aβ is identical to that of humans, and the aged wildtype marmosets start accumulating A from 7 years old or even earlier 25,26 . In addition, adolescent marmosets exhibit tau hyperphosphorylatiton, but not NFT formation, in brain that increases with aging 26 . Their life spans in captivity are as long as 10 to 15 years, making them suitable for aging research 27 . Their immune systems and metabolic functions resemble those of humans 27,28 and thus may affect the pathogenic processes related to AD [29][30][31] . Because sleep disorder is an early clinical incident in AD 32 , it is noteworthy that marmosets share with humans the sleep phases composed of rapid eye movement (REM) and non-REM cycles 33 . Among various non-human primate species, the marmoset seems most applicable to genetic manipulation, i.e. generation of designed mutants, for which the high reproductive efficacy is advantageous 22,34 . Furthermore, fecundity characteristics of marmosets, such as the short period of sexual maturity, multiple birth, short gestation interval, are suitable to produce genetically modified disease models. We decided to introduce a pathogenic mutation in the marmoset PSEN1 gene because the majority of familial AD-causing mutations reside in the PSEN1 gene 35 . Typically, deletion mutations in exon 9 [1][2][3][4][5] or point mutations at the 3' splice site (acceptor site) of exon 9 in the PSEN1 gene cause dominantly inherited familial AD. The point mutations instigate exon 9 elimination and S290C modification in the corresponding mRNA at the junction sites of exons 8 and 10 via conversion of alternative splicing [36][37][38][39][40] . We thus set out to generate a marmoset model of AD in which the exon 9 of PSEN1 gene product is deleted using TALEN to produce AD marmoset models because TALEN exhibits high genome-editing efficacy and because TALEN generates few off-target effects and produces little mosaicism 34,41 . To our knowledge, this is the first case, in which TALEN has successfully been applied to artificial exon skipping in non-human primates.

Evaluation of TALEN activity
We used the Platinum TALENs designed to target 3' splice site of exon 9 of the marmoset PSEN1 gene (Figure 1a) 42,43 . After introduction of the TALEN mRNAs into the nucleus of marmoset pronuclear stage embryos, we performed surveyor assays and sequencing with the genomic DNA extracted from the developed embryos. We found that 2 out of 3 embryos exhibited deletions at the target sequences including the acceptor site as expected (Figure 1b). To confirm exclusion of exon 9 in the PSEN1 mRNA, we performed RT-PCR and sequenced the corresponding cDNA sequencing using RNAs extracted form 4-cell-stage or single blastomeres of the TALEN-injected marmoset embryos (Figure   1c) 34 . We consequently verified complete exclusion of exon 9 in two out of two 4-cell-stage embryos and 2 out of 5 single blastomere, and 3 were failed to amplify the cDNA (Figure 1d). This indicated that the 3' splice site-deletion resulted in exclusion of exon 9 of the mRNA transcribed from the PSEN1 gene. There was no wild-type sequence in the embryos or single blastomeres in which 3' splice site was destroyed, suggesting that the mutations took place in a biallelic manner. We later noted that homozygous deletion of PSEN1 exon 9 appear to cause embryonic lethality in vivo (see the following).

Generation of marmosets lacking PSEN1 exon 9 in the corresponding mRNA (PSEN1-ΔE9) and apparent absence of mosaicism
The embryos treated with TALEN were transferred into uterus of surrogate mother marmosets to generate PSEN1 exon9 deleted AD model marmosets. However, the developmental rate of the embryos (35%) was significantly lower than that in our previous study 34 , and no pregnant animals were obtained even after 40 zygotic injections. These results implied that the deletion mutations of the PSEN1-ΔE9 in the biallelic status are likely to cause the low embryonic developmental rate and embryonic lethality presumably due to the disruption of Notch signaling 44 . This was somewhat expected because mice deficient in PS1 45,46 or lacking the active site aspartate 47 exhibited embryonic lethality and because the protein domain corresponding to PSEN1 exon 9 is a site for endoproteolysis 48 , which is perturbed by the active site destruction 47 . We thus came to realize that we need to introduce the ΔE9 mutations in a monoallelic manner.
To avoid the biallelism of the PSEN1 gene mutations, we injected TALEN mRNAs into 810 ova and then performed in vitro fertilization on the 578 (71.4%) ova that survived, using wild-type marmoset sperms. After the in vitro fertilization, 218 (37.7%) zygotes were obtained then transferred 154 (70.6%) embryos that developed to above 6-cell stage into uterus of 77 surrogate mothers. Around the 145th day from the ovulation of the surrogate mother, we obtained 6 marmoset neonates by normal delivery or by Caesarean section (Figure 2a, Table 1).
Surveyor assays of the genomic DNA isolated from hair roots of the newborns suggested  50 . These results and speculations indicate a probable absence of mosaicism in the mutant marmosets 34 . Consistently, single cell PCR analyses indicated that two of the mutants, I774gmF and I949gmM, were mosaicism-free: ten out of ten single cells showed heterozygous mutations. However, the other mutant, I943gmM, exhibited an extensive mosaicism (>80%) and was thus excluded from further analyses (Supplementary Figure 1). The sequence analyses of the cDNAs also showed that there were no insertions or deletions in the exons 8 and 10 in the PSEN1 gene ( Figure   2d, Supplementary Figure 2). We thus concluded that we successfully acquired PSEN1-ΔE9 marmoset neonates.

Off-target analyses of the mutant PSEN1-ΔE9 marmosets
A number of genome editing-based studies suffer form off-target effects 41 . We analyzed the off-target mutations in the mutant PSEN1-ΔE9 marmosets in two ways. The online Paired Target Finder tool was used to search for candidate off-target sites 34 for calculation an average score at each potential off-target site obtained from the search and identified the top 10 candidate sites (Supplementary Table 2). Genomic DNA extracted from hair roots of PSEN1-ΔE9 marmosets was subjected to PCR then determined the sequence by direct sequencing. None of the PSEN1-ΔE9 marmosets exhibited any mutations in the off-target candidate sites. Second, we performed whole genome sequencing (WGS) of the genomic DNA obtained from the first PSEN1-ΔE9 marmoset (I774gmF) and from her parents at the average coverage over genome of 42.1, 57.4, and 62.2x for the neonate, father, and mother, respectively (Supplementary Table 3). In order to process the WGS data sets, we performed Illumina Dynamic Read Analysis for GENomics (DRAGEN) pipeline (version 3.5.7) 51 for mapping and variant calling process at a default parameter with "CJA1912RKC" marmoset genome reference sequence (GCA_013373975.1). We used this highly permissive parameter in variant calling and did not filter out variants to keep as many variants in off-target candidate sites as possible. By comparing all the variants from the PSEN1-ΔE9 marmoset (I774gmF) with those from her parents, we found 306,492 variants only in the PSEN1-ΔE9 marmoset (I774gmF) (Supplementary Figure 3). None of the 306,492 variants were present within the 10 off-target candidate sites. These analyses further confirmed the absence of the off-target mutations. We thus declare that application of the TALEN technology to generation of the PSEN1-ΔE9 marmosets has fully succeeded.

Endoproteolysis of the mutant marmoset-derived PS1 and secretion of A 42 versus A 40
PS1 serves as a major catalytic subunit for the -secretase complex that generates A from the Cterminal fragment of APP after the limited proteolysis by -secretase 45,47,48,52 . Most, if not all, of the pathogenic mutations found in the PSEN1 gene cause familial AD by elevating the ratio of A 42 /A 40 production 53-56 . The ratio matters because Kim et al. demonstrated that A 42 is causal whereas A 40 is protective in the pathological processes 57 . To biochemically analyze the mutant marmosets in the least invasive manner, we established primary fibroblasts from the wild-type and mutant marmosets edge of the ear lobe. We first examined the effect of exon 9 deletion on the endoproteolytic status of PS1 protein because such deletion impacts on the global PS1 endopeptidase activity 5 . We performed Western blot analysis of fibroblasts derived from wild-type and PSEN1-ΔE9 marmosets using a set of antibodies that recognize the N-and C-termini of PS1. The exon 9 deficiency in the two independent mutants, i.e. I774gmF and I1943gmM, gave rise to a full-length PS1 that was absent in the wild-type fibroblasts (Fig. 3a). Consistently, the quantities of the N-terminal and C-terminal fragments in the mutants were reduced to approximately half the levels compared to those in wild-type cells. These observations indicate that the monoallelic exon 9 deletion caused perturbation of the PS1 endoproteolysis in a heterozygous manner and that the nature of endoproteolysis was autolytic. The deletion also induced an amino acid sequence conversion, S290C, as shown in Supplementary Figure   2 40 .
We then quantified A 42 and A 40 secreted from the wild-type and mutant fibroblasts by enzyme-linked immunosorbent assay (ELISA) for the reasons stated above. The quantity of A 42 was significantly increased in the mutant-derived media while that of A 40 also showed a significant reduction. Consequently, the ratio of Aβ 42 /Aβ 40 production was statistically significantly increased approximately 3-fold (Fig. 3b), a hallmark of familial AD pathogenesis. Similar results had been reported using cultured human fibroblasts and lymphocytes obtained from PSEN1 mutation carries [53][54][55][56] . A 43 58 was undetectable. These observations indicate that the PSEN1-ΔE9 mutations introduced into the marmosets are indeed pathogenic.

Discussion
Steiner et al. previously demonstrated that a PSEN1-ΔE9 mutation exerted its pathogenic effect via an amino acid conversion, S290C, rather than perturbed PS1 endoproteolysis 40 . This effect of the amino acid conversion on the Aβ 42 /Aβ 40 ratio is also likely to apply to our observation in the fibroblasts obtained from the mutant marmosets because they also carry the S290C conversion (Supplementary  To our knowledge, this is the first non-human primate model of familial AD ever produced.
Because the wild-type marmosets start accumulating A in brain around the age of 7 or even earlier 25 , because the age-of-onset of PSEN1-ΔE9 humans varies in their 40's 1 2 39 , and because A deposition arises approximately two decades prior to the disease on-set in humans 9 , we anticipate the mutant marmosets to start exhibiting A pathology from 2-3 years old. Relevance of the mechanistic findings obtained using the model mice 16,17,19 shall be confirmed if the mutant marmosets are utilized. Because the marmosets live much longer than mice, they may develop tau pathology and neurodegeneration, which model mice failed to recapitulate. If so, they would become convenient tools to establish the cause-and-effect relationship and to elucidate the mechanisms by which A amyloidosis evokes tau pathology and neurodegeneration. They would first become applicable to primary basic studies being used for humans, including non-invasive imaging analyses by MRI to determine morphometric

Animals
All animal experiments were carried out in accordance with guidelines for animal experimentation from the RIKEN Center for Brain Science and the Central Institute for Experimental Animals (CIEA: 18031A, 19033A). The CIEA standard guidelines are in accordance with the guidelines for the proper conduct of animal experiments determined by the Science Council of Japan. The marmosets used in this study were purchased from CLEA Japan, Inc., Tokyo, Japan. The animals ranged from 2-4.5 yearsold with body weights between 300 and 450 g.

In vitro transcription of TALEN mRNA
We employed a two-step Golden Gate assembly method using the Platinum Gate TALEN Kit (Addgene; cat#1000000043) to construct Platinum TALEN plasmids containing the homodimer-type FokI nuclease domain. The assembled repeat arrays were subsequently inserted into the final destination vector, ptCMV-153/47-VR. TALEN mRNA was synthesized using the mMESSAGE mMACHINE T7 Ultra Transcription Kit (Thermo Fisher Scientific, AM1345). Transcribed mRNA was purified using the MEGAclear Transcription Clean-Up Kit (Thermo Fisher Scientific, AM1908), and Nuclease-Free water (Thermo Fisher Scientific, AM9937) was used for the elution step and subsequent dilution. For microinjection, an aqueous solution was used in which left TALEN and right TALEN mRNAs were mixed in equal amounts so that the final concentration was 8 ng/μl.

Procedures for oocytes collection and in vitro fertilization
Oocyte collection and in vitro fertilization (IVF) was performed as previously described 22 65 . For IVF, semen was collected from healthy male marmosets and diluted to 3.6✕10 6 sperm/ml, after washing with TYH medium (LSI Medience, DR01031). Insemination was performed via the co-incubation of 3.6✕10 4 sperm per oocyte for 10-16 hours at 37°C with 5% CO 2 , 5% O 2 , and 90% N 2 .

Blastomere splitting
TALEN-mRNA-injected embryos that had reached the 4-cell stage and above were directed to blastomere splitting as previously described 34 . Briefly, the embryo specimens were treated with acidified Tyrode's solution (Origio, 10605000) to lyse the zona pellucida. Naked embryos were transferred into Embryo Biopsy Medium (Origio, 10620010). After several minutes, embryos that exhibited weak adhesion between blastomeres were split into single blastomeres using glass capillaries and washed in phosphate-buffered saline (PBS) (-) drops immediately. The blastomere was transferred to a test tube with a small amount of PBS and used for subsequent analyses.

Embryo transfer
The embryos that grown to 6-cell-stage or higher were transferred to the uteri of surrogate mothers as previously described 22,64 . The method for obtaining newborns was based on natural delivery, but if delivery was judged to be difficult, a Caesarean section was performed on day 145 after embryo transfer 34 .

PCR for genotyping
Genomic DNA was extracted from marmoset cord blood and hair roots using QIAamp DNA Micro Kit (Qiagen, 56304) according to the manufacturer's instructions. Embryos and blastomeres were used as templates without any genome extraction processing. PCR was performed using specific primers followed by subsequent sequencing or surveyor assay (Supplementary Table 4

Sequencing analyses
To confirm the sequences of the TALEN target sites, PCR products were subcloned into the Zero Blunt PCR Cloning Kit (Thermo fisher scientific, 44-0302) according to manufacturer's instructions. The plasmids obtained were transfomed with ECOS Competent E. coli DH5α (Nippon gene, 316-06233) and cultured on LB agar containing Kanamycin. After amplifying the plasmid using the emerged single colony, it was reacted using M13 primer and Big Dye terminator v3.1 cycle sequence kit (Thermo fisher scientific, 4337455), and gene sequences were analyzed by 3130 Genetic Analyzer (Thermo fisher scientific).

Surveyor assay
The SURVEYOR mutation detection kit (IDT, 706025), one of the mutation analysis methods that identifies mismatches in double-strand DNA, was used to detect genetic modifications at the TALEN target site. Briefly, 8 µl of the PCR product described above. For the analysis of biallelic mutations in the PSEN1 gene, a sample in which 4 µl of the PCR product was mixed with an equal amount of the PCR product prepared using wild-type PSEN1 as a template was also prepared. The sample digested with Nuclease was electrophoresed on Novex TBE Gels, 10%, 12 well (Thermo fisher scientific, EC62752BOX) and then subjected to nucleic acid staining with GelRed (Fujifilm Wako Chemicals, 518-24031).

Single-cell PCR
To establish the marmoset primary fibroblast cells, approximately 3 mm 3 was excised from the edge of the earlobe of marmosets. Tissue from each animal was cultured at 37 °C and 5% CO 2 with Dulbecco's modified Eagle's medium + GlutaMAX (Thermo Fisher Scientific, 10566-024) containing 10% FBS (Biowest) and a penicillin-streptomycin cocktail (Thermo Fisher Scientific, 15240-062).
Cultured fibroblast cells were harvested and washed in PBS (-) following which single cells were isolated one by one with a glass capillary tube and placed into individual sampling tubes. Direct PCR and sequencing were then performed, with the PCR conditions used being the same as those used for the genotyping PCR. The electrophoresis of PCR products was performed with 10%TBE Gels (Thermo Fisher, EC62752BOX), followed by nucleic acid staining with GelRed.

RT-PCR
Total RNA was extracted from marmoset embryos, blastomeres, and hair roots using Nucleospin RNA plus XS (Takara, U0990B) according to the manufacturer's instructions. cDNA was synthesized using ReverTra Ace -α-(Toyobo, FSK-101F). In the reaction, the sample was divided into two: one received Reverse Transcriptase, and the other received the same amount of water as a control to evaluate genomic DNA contamination. PCR was performed the primers designed on exons (Supplementary Table 4

Whole-genome sequencing (WGS)
To prepare the genomic DNA for WGS, the edge of the ear lobe (approximately 3 mm square) was cut out from the marmosets as a lowly invasive body tissue collection. Genomic DNA extraction from the ear tissue was performed using the QIAamp Micro DNA kit (Qiagen, 56304), according to the manufacturer's instructions. Briefly, the tissue was incubated in lysis buffer and Proteinase K overnight at 56°C, then genomic DNA was collected by column purification with low-TE (Tris-Hcl 10 mM,  Table 3). We used this highly permissive parameter in variant calling and did not filter out variants to keep as many variants in off-target regions as possible. By comparing all the variants from the mutant PSEN1 (I774gmF) marmoset with those from her parents, 306,492 variants were found only in the mutant PSEN1 marmoset without any filtering.

Off-target analysis by Paired Target Finder Tool
Paired Target Finder (https://tale-nt.cac.cornell.edu/node/add/talef-off-paired) was used to search for potential off-target sites 66 . As the reference, CJA1912RKC (GenBank assembly accession: GCA_013373975.1), which is the whole genome sequence of marmoset, was set in the search database, and the RVD sequence of TALEN was provided as follows; RVD Sequence 1, NI HD HD NG NN NN NI NI NG NG NG NG NN NG HD NG NG; RVD Sequence 2, NI NG NG HD NI HD HD NI NI HD HD NI NG NI HD HD NI. The minimum and maximum lengths of the spacers were set to 10 and 30, respectively. Other parameters were set as recommended. Genomic DNA extracted from marmoset hair roots using QIAamp DNA Micro Kit (Qiagen, 56304) was subjected to PCR with specific primers followed by direct sequencing (Supplementary Table 5

Western blot analysis
We established primary fibroblast lines from ear skin tissues of the wild-type and PSEN1-ΔE9 marmosets, i.e. I774gmF and I943gmM. We performed Western blotting as previously described 67 .
Briefly, cells were lysed and subjected to subcellular fractionation using ProteoExtract Subcellular Proteome Extraction Kit (Merck, #539790) according to the manufacturer's instructions. We loaded the membrane fractions onto polyacrylamide gels, electrophoresed them, blotted them onto PVDF membranes (Merck Millipore), which were then treated with the ECL Prime blocking agent (GE Healthcare) and incubated with the primary antibodies raised against the PS1 N-terminal (G1Nr5) and C-terminal (G1L3) domains 68 69 . We used an antibody to sodium-potassium ATPase (EP1845Y, Abcam, ab76020) as a loading control of the membrane fraction.

Enzyme-linked immunosorbent assay (ELISA)
Aβ 40 and Aβ 42 was quantified using ELISA as previously described 59 . Briefly, culture media, collected from the primary skin fibroblasts at 72 hours after incubation, were mixed with 11 times the volume of 6M guanidine-HCl. We then quantified Aβ 40 and Aβ 42 using Aβ ELISA kit (Wako, 294-62501 and 14 294-62601), according to the manufacturer's instructions. Quantification of A 43 was performed as previously described 58 .

Statistic analysis
All data are shown as the mean ± s.e.m. For comparison between two groups, data were analyzed by Student's t-test. All the data were collected and processed in a randomized and blinded manner.

Conflicts of interest
The authors declare no conflicts of interest in the present study.  DNAs extracted from umbilical cord blood and hair roots were subjected to surveyor assays. Samples, in which 4 µl of the PCR product were mixed in equal amounts with the PCR product from wild-type PSEN1 DNA as a template to detect biallelic mutations in the PSEN1 gene, are designated as +N. c) Sequences of the TALEN-targeted site in PSEN1 gene. the underlined part corresponds to the TALEN recognition sequence: the red letters to the acceptor site, and the dashes to the missing bases. d) Result of sequence analyses of the neonates. RT-PCR products from hair roots obtained from 6 obtained neonates were subcloned and sequenced. The first column shows the marmoset identities, the second the genotypes, the 3 rd the percentage of wild-type cDNA, and the last percentage of mutated cDNA.  154 (70.6) 77 7 (9.1) 2 (28.6) 6 3 3