Abstract
The gradual accumulation of amyloid-β (Aβ) is a neuropathologic hallmark of Alzheimer’s disease (AD); playing a key role in disease progression. Aβ is generated by the sequential cleavage of amyloid precursor protein (APP) by β- and γ-secretases, with BACE-1 (β-site APP cleaving enzyme-1) cleavage as the rate limiting step 1–3. CRISPR/Cas9 guided gene-editing is emerging as a promising tool to edit pathogenic mutations and hinder disease progression 4,5,6 However, few studies have applied this technology to neurologic diseases 7–9. Besides technical caveats such as low editing efficiency in brains and limited in vivo validation 7, the canonical approach of ‘mutation-correction’ would only be applicable to the small fraction of neurodegenerative cases that are inherited (i.e. < 10% of AD, Parkinson’s, ALS); with a new strategy needed for every gene. Moreover, feasibility of CRISPR/Cas9 as a therapeutic possibility in sporadic AD has not been explored. Here we introduce a strategy to edit endogenous APP at the extreme C-terminus and reciprocally manipulate the amyloid pathway – attenuating β-cleavage and Aβ, while up-regulating neuroprotective a-cleavage. APP N-terminus, as well as compensatory APP homologues remain intact, and key physiologic parameters remain unaffected. Robust APP-editing is seen in cell lines, cultured neurons, human embryonic stem cells/iPSC-neurons, and mouse brains. Our strategy works by limiting the physical association of APP and BACE-1, and we also delineate the mechanism that abrogates APP/BACE-1 interaction in this setting. Our work offers an innovative ‘cut and silence’ gene-editing strategy that could be a new therapeutic paradigm for AD.
Our broad idea is to rationally edit small segments of wild-type (WT) proteins known to play key roles in the progression of sporadic disease, with the ultimate goal of attenuating their pathologic activity. As endogenous proteins expectedly play physiologic roles as well, it is also important to conserve the normal function of these molecules, as far as possible. Motivated by this idea, we designed sgRNAs targeting the extreme C-terminus of mouse APP, and one of these led to robust APP-editing, as shown in Fig. 1 (see protospacer adjacent motif – PAM – site and genomic target recognized by the sgRNA in Fig. 1a). APP-editing resulted in attenuated staining with an antibody (Y188) recognizing the extreme C-terminus of APP that is distal to the sgRNA-targeting site (neuroblastoma cells shown in Fig. 1b, see Y188 epitope in Fig. 1a). APP-sgRNA also attenuated the production of APP C-terminal fragments (CTFs; Fig. 1c, d - time-course of editing in Fig. 1e). However, an antibody recognizing the APP N-terminus (22C11) showed no differences between control and sgRNA-treated samples (Fig. 1c), suggesting that the editing only affected the short intracellular C-terminus. Genomic deep-sequencing confirmed efficient editing of mouse APP at the expected target (Fig. 1f).
Though the abovementioned TGG PAM is conserved in both mouse and human APP, the upstream targeting sequence differs only by two nucleotides (Fig. 1g). Despite this, the mouse APP-sgRNA was unable to edit human APP, perhaps reflecting the specificity of the CRISPR/Cas9 system; also attested by other groups 10–12. However, a sgRNA specific to the human APP targeting sequence robustly edited APP in HEK293 (Extended Data Fig. 1a-c), as well as in human embryonic stem cells (Extended Data Fig. 1d-f); as determined by immunostaining, western blots and deep-sequencing. APP editing and decreased CTFs was also seen in human iPSCs differentiated to neurons (Fig. 1h, i). While APP cleavage by BACE-1 initiates the cascade of events leading to Aβ deposition, the alternative a-cleavage “anti-amyloidogenic” pathway – mediated by a-secretases – is thought to be protective [reviewed in 13]. Interestingly, sAPPa was up-regulated in iPSC-neurons (Fig. 1j), suggesting activation of the neuroprotective a-cleavage pathway. ELISAs showed corresponding attenuation of Aβ-40/42 in the iPSC-neurons, confirming inhibition of β-cleavage (Fig. 1k).
The above data suggest that the APP-sgRNA has reciprocal effects on APP a- and β-cleavage. To determine effects of the sgRNA in a more controlled setting, we engineered a stable H4 neuroglioma cell line expressing single copies of APP and BACE-1, using the FlpIn system (APP/BACEsingle_copy; for details, see Extended Data Fig. 2a and methods). Endogenous APP/BACE-1 levels are negligible in these cells, thus almost all the APP/BACE-1 expression is from the introduced single-copies (see Extended Data Fig. 2b).
We also tagged APP and BACE-1 in these cells to the N- and C-terminal fragments of the Venus fluorescent protein (VN/VC) respectively – as in our previous study 14 – for two reasons. First, fluorescence complementation of APP:VN and BACE-1:VC reflects the physical association of this substrate-enzyme pair 14, allowing us to directly evaluate editing efficiency by monitoring YFP fluorescence. Second, the APP-CTFs generated from APP:VN are easier to identify in Western blots (as they run higher). Indeed, transduction of APP/BACEsingle_copy cells with a lentivirus carrying APP-sgRNA and Cas9 essentially eliminated Venus complementation (Extended Data Fig. 2c, d), confirming robust APP editing. Biochemical analyses using APP antibodies against extra- or intra-cellular epitopes confirmed that the sgRNA specifically inhibited APP β-cleavage (Extended Data Fig. 2e, f). Indeed, in line with the data from iPSC-neurons, the sgRNA had reciprocal effects on sAPPα and Aβ (Fig. 11, m); providing confidence that our gene editing strategy is favorably manipulating the amyloid pathway.
Off-target effects of CRISPR/Cas9 is a potential concern. Towards this, we asked if our mouse and human APP-sgRNA were able to edit the top five computationally predicted off-target sites (Extended Data Fig. 3a; also see Extended Data Table 1). No editing was detected (Extended Data Fig. 3b-e), further attesting specificity. Though APP null mice are viable with minor deficits, there is compensation by the two APP homologues APLP1 and 2 that undergo similar processing as APP (reviewed in 15,16). APLP1 and 2 were not amongst the top 50 predicted off-target sites, as their corresponding sgRNA-target sites were substantially different from APP (see sequences in Extended Data Fig. 3f). For further assurance that our sgRNA was not editing the APP homologues, we performed TIDE off-target analyses 17 on cells carrying the sgRNA. As shown in Extended Data Fig. 3g, TIDE analyses showed no editing of APLP 1/2 by the sgRNA.
APP has physiologic roles in axon growth and signaling [reviewed in 18]. As noted above, the N-terminus of APP – thought to play roles in axon growth and differentiation – is entirely preserved in our setting. However, the C-terminal APP intracellular domain (AICD) has been implicated in gene transcription, though the effect appears to be both physiologic and pathologic 19,20. To examine potential deleterious effects of editing the extreme C-terminus of APP, we turned to cultured hippocampal neurons where various parameters like neurite outgrowth and synaptic structure/function can be confidently evaluated. For these studies, we generated AAV9 viruses carrying the APP-sgRNA and Cas9, tagged with GFP and HA respectively (see vector design in Fig. 2a) that transduced almost all cultured neurons (Fig. 2b and data not shown). In blinded analyses, we found no significant effect of the APP-sgRNA on neurite outgrowth, axon-length, synaptic organization, or neuronal activity (Fig. 2). Although further studies are needed, we note that besides editing only a small segment of APP, our strategy: 1) does not completely block β-cleavage; and 2) does not affect the physiologic processing of APLP 1/2 (that also generate CTFs).
Next we asked if our sgRNA could edit APP in vivo. Injection of the AAV9s into mouse hippocampi (Fig. 3a) led to efficient transduction of both sgRNA and Cas9 in dentate neurons (86.87 ± 2.83 % neurons carrying the sgRNA also had Cas9 – sampling of 495 neurons from 3 brains; see representative images in Fig. 3b). Immunostaining of transduced neurons with the APP Y188 antibody suggests editing of endogenous APP in vivo (Fig. 3c). To achieve a more widespread expression of the sgRNA and Cas9 in mouse brains – and also evaluate editing by biochemistry – we injected the viruses into the ventricles of neonatal (P0) mice and examined the brain after 2-4 weeks (Fig. 3e). Previous studies have shown that when AAVs are injected into the ventricles of neonatal mice, there is widespread delivery of transgenes into the brain – also called somatic transgenesis 21,22. Indeed, we saw patchy attenuation of APP Y188 staining in cortical regions (Fig. 3f), and also a decrease in CTFs by western blotting (Fig. 3g, h); suggesting that our gRNA can edit APP in vivo.
Finally, we sought to understand the mechanism by which CRISPR-mediated APP editing attenuates the β-cleavage pathway (note that the sgRNA-editing site is distant from the β-cleavage site). Genomic sequencing showed that sgRNA-editing leads to a translational product where the last 36 aa of APP are truncated (Fig. 4a). To evaluate functional consequences of editing, we generated a truncated “APP CRISPR-mimic” construct (APP-659). Using our fluorescence complementation assay 14, we first asked if APP-659 interacted with BACE-1. APP-659/BACE-1 approximation was greatly attenuated (Fig. 4b), along with a decrease in β-CTF generation (Fig. 4c). What is the mechanism underlying this decrease in APP/BACE-1 complementation? Since the sgRNA targeting site is distant from the BACE-1-cleavage site (see Fig. 1a), it seems unlikely that gene editing directly interferes with APP/BACE-1 interaction. Instead, trafficking alterations of the edited APP molecules – eventually leading to decreased APP/BACE-1 approximation – seem more likely. APP is synthesized in the ER→Golgi pathway, and Golgi-derived vesicles carrying APP are transported into axons and dendrites, inserting into the plasma membrane. Subsequently, surface APP is internalized into endosomes, where it is cleaved by BACE-1, and this is thought to initiate β-cleavage 23–5.
Accordingly, we systematically explored various trafficking steps in hippocampal neurons. First, we visualized axonal and dendritic transport of APP-WT and APP-659. Although there were modest changes (Extended Data Fig. 4 and Extended Data Table 2), it seems unlikely that such transport perturbations would lead to the dramatic attenuation of β-cleavage and Aβ-production seen in our experiments. Asking if amino-acid residues within the CRISPR-edited segment of APP might offer clues, we noted that APP residues T668 and Y682-Y687 (the “YENPTY motif”, see Fig. 4d; also present in APLP1/2) in the edited segment have been reported to play a role in Aβ production ¾27. Specifically, APP phosphorylated at T668 preferentially colocalizes with BACE-1 in endosomes 26, and the YENPTY motif mediates APP internalization from the plasma membrane 28. Indeed, the extent of APP/BACE-1 attenuation by the YENPTY mutation strongly resembled the decrease in fluorescence complementation by the APP-659 (“CRISPR-mimic”) construct (Fig. 4e). A prediction from these experiments is that the endocytosis of APP-659 from the cell surface should be attenuated; and indeed, this was the case in neuronal internalization assays (Fig. 4f). Collectively, the data suggest that our gene-editing approach does not have a major effect on post-Golgi trafficking of APP, but attenuates its endocytosis from cell surface, and consequently, its interaction with BACE-1 in endosomes. Since most of the APP α-cleavage is thought to occur at the cell surface 29, this may also explain why the non-amyloidogenic pathway is enhanced by our approach.
Using CRISPR/Cas9 technology, here we provide proof of concept for a gene-editing strategy that can favorably manipulate the amyloid pathway – attenuating β-cleavage and Aβ production, while up-regulating the neuroprotective α-cleavage. APP editing was efficient in a variety of human and mouse cells, neurons, and in vivo. After almost thirty years of controversies and failures, gene therapy of neurologic diseases has turned a page, with striking results in clinical trials 6,30. A key advance has been the development of AAV9-based vectors that can diffusely deliver genes throughout the nervous system 31, and our vision is to leverage these innovations to develop gene-editing therapies for AD. In principle, our ‘cut and silence’ CRISPR-editing approach might also work for attenuating other endogenous pathology-driving proteins in neurodegenerative diseases – such as BACE, tau, presenilins, α-synuclein, and TDP-43 – and may usher in a new way of therapeutically tackling these devastating diseases.
EXPERIMENTAL PROCEDURES
Constructs, antibodies and reagents
For transient co-expression of CRISPR/Cas9 components, APP gRNA nucleotides were synthesized and cloned into pU6-(Bbs1)_CBh-Cas9-T2A-mCherry vector at Bbs1 site. For viral transduction, a dual vector system was used to deliver CRISPR/Cas9 components using AAV9 32. For making the AAV9 vectors, the APP gRNA was cloned into pAAV9-U6sgRNA(SapI)_hSyn-GFP-KASH-bGH vector at Sap1 site. The CRISPR/Cas9 stable cell lines were generated by lentivirus infection as follows. The APP gRNA was cloned into lentiCRISPR v2 vector at Bbs1 site to produce lentivirus 33. For making APP deletions and relevant constructs, the human APP659 truncation was PCR amplified and cloned at Hind3 and Sac2 sites of pVN to generate pAPP659:VN. The BBS-APP659 was PCR amplified and cloned into pBBS-APP:GFP at Hind3 and Sac2, replacing BBS-APP, to generate pBBS-APP659:GFP. The pBBS-APPYENPTY:GFP was generated by site directed mutagenesis from pBBS-APP:GFP. The pAPPT668A:VN and pAPPT668A+YENPTY:VN were generated by site directed mutagenesis from pAPP:VN and pAPPYENPTY:VN. Antibodies used were as follows: APP Y188 (ab32136; Abcam), APP 22C11 (MAB348; Millipore), APP 6E10 (803001; BioLegend), BACE-1 (MAB931; R&D), GAPDH (MA5-15738, ThermoFisher), GFP (ab290, Abcam), GFP (A10262, Invitrogen), HA (901513, BioLegend), VAMP2 (104211, Synaptic Systems). Reagents were as follows: a-bungarotoxin Alexa-594 conjugate (Life Technologies), Tubocurarine chloride (Sigma), γ-secretase inhibitor BMS-299897 (Sigma), Rho Kinase (ROCK)-inhibitor H-1152P (Calbiochem) and Dynasore (Sigma).
Cell cultures, transfections, viral production/infections, and biochemistry
HEK293 and neuro2a cells (ATCC) were maintained in DMEM with 10% FBS. Cells were transfected with Lipofectamine 2000 and collected 5 days after transfection for biochemical and immunostaining analysis. Primary hippocampal neurons were obtained from postnatal (P0-P1) CD1 mice (either sex), and transiently transfected using Lipofectamine 2000 or Amaxa 4D system (Lonza). Dissociated neurons were plated at a density of 30,000 cells/cm2 on poly-D-lysine-coated glass-bottom culture dishes (Mattek) and maintained in Neurobasal/B27 medium with 5% CO2. For APP/BACE-1 interaction, APP internalization and APP transport studies, DIV 7 neurons were cultured for ~18-20 h after transfection. For spine density analysis, DIV7 neurons were transfected with soluble markers and cultured for 7 d before imaging. For testing the effect of CRISPR/Cas9 on neuronal development, neurons were electroporated with the respective constructs before plating using an Amaxa 4D-Nucleofector system with the P3 Primary Cell 4D-Nucleofector X kit S and program CL-133.
For western blotting and electrophysiology, DIV7 cultured neurons were infected with either AAV9-APP gRNA-GFP (2.24x1013 Vg/ml) and AAV9-Cas9 (2.4x1014 Vg/ml), or AAV9-GFP (2.58x1013 Vg/ml) and AAV9-Cas9 at a multiplicity of infection (MOI) of 1.5x105. Neurons were analyzed 7 days post-infection. Lentivirus was produced from HEK293FT cells as described 34. Briefly, HEK293FT cells (Life Technologies) were maintained in DMEM with 10% FBS, 0.1mM NEAA, 1 mM sodium pyruvate and 2mM Glutamine. Cells were transfected with lentiviral-target and helper plasmids at 80-90% confluency. 2 days after transfection, the supernatant was collected and filtered with 0.45 μm filter. For experiments with hESCs, cells were cultured on a Matrigel substrate (BD Biosciences) and fed daily with TeSR-E8 culture media (StemCell Technologies). When the cells were around 60-70% confluent, they were infected with a 50/50 mixture of TeSR-E8 (with 1.0 μM H-1152P) and lentivirus supernatant. After 24 h, the virus was removed, and the cells were fed for 2 days (to recover). After 3 days, cells were treated with 0.33 μg/mL of puromycin for 72 h to select for virally-integrated hESCs. For HEK and neuro2a cell lines, cells were infected with the lentivirus carrying APP-sgRNA and Cas9 for 24 h. And then cells were fed for 1 day to recover. After 2 days, cells were treated with 1 μg/mL of puromycin for 72 h to select for virally-integrated cells.
Human NPCs were generated as has been described previously 35, using manual rosette selection and Matrigel (Corning) to maintain them. Concentrated lentiviruses express control-sgRNA or APP-sgRNA were made as described previously 36, using Lenti-X concentrator (Clontech). The NPCs were transduced with either control-sgRNA or APP-sgRNA after Accutase splitting, and were submitted to puromycin selection the subsequent day. Polyclonal lines were expanded, and treated with puromycin for 5 more days before banking. Neuronal differentiations were carried out by plating 165,000 cells/12 well-well in N2/B27 media (DMEM/F12 base) supplemented with BDNF (20 ng/mL; R&D) and laminin (1 ug/mL; Trevigen).
For biochemistry, cell lysates were prepared in PBS + 0.15% Triton X-100 or RIPA supplemented with protease inhibitor cocktail, pH 7.4. After centrifuging at 12,000 g for 15 min at 4 °C, supernatants were quantified and resolved by SDS-PAGE for western blot analysis. For sAPPa detection, cell culture medium was collected and centrifugated at 2,000 g for 15 min at RT. The supernatants were resolved by SDS-PAGE for western blot analysis; band intensities were measured by ImageJ. Human Aβ40 and Aβ42 were detected using kits, according to the manufacturer’s instructions (Thermo KHB3481 and KHB3544). Briefly, supernatants from H4single copy cells or human iPSC derived neurons were collected and diluted (x5 for H4 and x2 for iPSC-neuron). The diluted supernatants and the human Aβ40/42 detection antibodies were then added into well, and incubated for 3 h at RT with shaking. After washing (x4), the anti-Rabbit IgG HRP solution was added and incubated for 30 min at RT. The stabilized Chromogen was added after washing (x4), and incubated for another 30 min at RT in the dark. After addition of stop solution, absorbance at 450 nm was read using a luminescence microplate reader.
Developing a single-copy, stable APP/BACE-1 cell line
H4 tetOff FlpIn empty clone was maintained in OptiMEM with 10% FBS, 200 μg/mL G418 and 300 μg/mL Zeocin. To generate an APP:VN/BACE-1:VC stable cell line carrying single copies of APP and BACE-1, the expressing plasmid and pOG44 plasmids were transfected with Lipofectamine 2000. 2 days after transfection, cells were selected with 200 μg/mL hygromycin B and 200 μg/mL G418 for 1 week. A monoclonal cell line with stable expression was selected by cell sorting, based on fluorescence-complementation of the tagged VN/VC fragments. H4 stable cell lines were then infected with the lentivirus carrying APP-gRNA and Cas9, as described above. After 24 h, the virus was removed, and cells were fed for 1 day to recover. After 2 days, cells were treated with 0.7 μg/mL of puromycin for 72 h to select for virally-integrated cells.
Immunofluorescence, microscopy/image analysis, APP trafficking and endocytosis assays
For immunostaining of endogenous APP or VAMP2, cells were fixed in 4% PFA/sucrose solution in PBS for 10 min at room temperature (RT), extracted in PBS containing 0.2% Triton X-100 for 10 min at RT, blocked for 2 h at RT in 1% bovine serum albumin and 5% FBS, and then incubated with rabbit anti-APP (1:200) or mouse anti-VAMP2 (1:1000) diluted in blocking buffer for 2 h at RT. After removal of primary antibody, cells were blocked for 30 min at RT, incubated with goat anti–rabbit (Alexa Fluor 488) or goat anti–mouse (Alexa Fluor 594) secondary antibody at 1:1000 dilution for 1 h at RT and then mounted for imaging. z-stack images (0.339 μm z-step) were acquired using an inverted epifluorescence microscope (Eclipse Ti-E) equipped with CFI S Fluor VC 40× NA 1.30 (Nikon). An electron-multiplying charge-coupled device camera (QuantEM: 512SC; Photometrics) and LED illuminator (SPECTRA X; Lumencor) were used for all image acquisition. The system was controlled by Elements software (NIS Elements Advanced Research). z-stacks were subjected to a maximum intensity projection. For APP Y188 staining, the average intensity of single cell body (neuro2A, HEK293 and neurons) or the whole colony (hESCs) was quantified. All the images were analyzed in Metamorph and ImageJ.
Spine density experiments were done as described previously 37. Briefly, DIV 7 neurons were transfected with desired constructs for 7 days, and secondary dendrites were selected for imaging. z-stack images were captured using a 100x objective (0.2 μm z-step) and subjected to a maximum intensity projection for analysis. For the APP/BACE-1 complementation assay, DIV 7 neurons were transfected with desired constructs for ~15-18 h and fixed. z-stack images were captured using a 40x objective (0.339 μm z-step) and subjected to a maximum intensity projection. The average intensity within cell bodies was quantified.
For trafficking studies in axons and dendrites, imaging parameters were set at 1 frame/s and total 200 frames. Kymographs were generated in MetaMorph, and segmental tracks were traced on the kymographs using a line tool. The resultant velocity (distance/time) and run length data were obtained for each track, frequencies of particle movements were calculated by dividing the number of individual particles moving in a given direction by the total number of analyzed particles in the kymograph, and numbers of particles per minute were calculated by dividing the number of particles moving in a given direction by the total imaging time.
APP endocytosis assay was done as described previously 38. Cells expressing APP-GFP or APP659-GFP were starved with serum-free medium for 30 min and incubated with anti-APP (22C11) in complete medium with 10 mM HEPES for 10 min. And then, cells were fixed, permeablized and immunostained for 22C11. The mean intensity of 22C11 along plasma membrane was calculated by dividing the total intensity along plasma membrane (= intensity of whole cell – intensity of cytoplasm) with area of plasma membrane (= area of whole cell – area of cytoplasm). The ratio of mean intensities between plasma membrane and cytoplasm was quantified.
Stereotactic injection of AAV9s into the mouse brain and histology
In vivo injection and immunofluorescence staining was done as described previously 39. Briefly, 1.5๑1 of 1:2 AAV9 mixture of AAV9-APP gRNA-GFP (or AAV9-GFP) and AAV9-Cas9 was injected into the dentate gyrus (-2.0, ±1.6, -1.9) of 8-week old male C57BL/6 mice (either sex). 2-weeks after surgery, the mice were sacrificed by trans-cardiac perfusion of saline, followed by 4% PFA. The brains were dissected, post-fixed with 4% PFA overnight, immersed in 30% sucrose until saturation, and sectioned at 40 μm. Sections were immunostained with following antibodies: mouse anti-HA (1:1000, BioLegend, clone 16B12), chicken anti-GFP (1:1000, Invitrogen, polyclonal) and rabbit anti-APP (1:200, Abcam, clone Y188). Images were acquired using Zeiss LSM800 confocal microscope. Average intensities of APP staining in cell bodies was quantified using Metamorph.
Intracerebroventricular injections and histology
All animal procedures were approved by the Mayo Institutional Animal Care and Use Committee and are in accordance with the NIH Guide for Care and Use of Laboratory animals. Free hand bilateral intracerebroventricular (ICV) injections were performed as previously described 40 in C57BL/6J mouse pups. On post-natal day 0, newborn pups were briefly cryoanesthetized on ice until no movement was observed. A 30-gauge needle attached to a 10 μl syringe (Hamilton) was used to pierce the skull of the pups just posterior to bregma and 2 mm lateral to the midline. The needle was held at a depth of approximately 2 millimeters, and 2 μl of a mixture of AAV9 viruses (ratio 1:2 of AAV9--APP gRNA-GFP or AAV9-GFP+ AAV9-Cas9) were injected into each cerebral ventricle. After 5 minutes of recovery on a heat pad, the pups were returned into their home cages. Mice were sacrificed 15 days after viral injection. Animals were deeply anesthetized with sodium pentobarbital prior to transcardial perfusion with phosphate buffered saline (PBS), and the brain was removed and bisected along the midline. The left hemisphere was drop-fixed in 10% neutral buffered formalin (Fisher Scientific, Waltham, MA) overnight at 4°C for histology, whereas the right hemisphere of each brain was snap-frozen and homogenized for biochemical analysis. Formalin fixed brains were embedded in paraffin wax, sectioned in a sagittal plane at 5-micron thickness, and mounted on glass slides. Tissue sections were then deparaffinized in xylene and rehydrated. Antigen retrieval was performed by steaming in distilled water for 30 min, followed by permeabilization with 0.5% Triton-X, and blocking with 5% goat serum for 1 hour. Sagittal sections were then incubated with primary anti-GFP antibody (1:250, Aves, chicken polyclonal) and anti-APP antibody (1:200, Abcam, clone Y188) overnight at 4°C. Sections were incubated with the secondary antibodies Alexa 488-goat anti-chicken and Alexa 568-goat anti rabbit (1:500, Invitrogen) for 2h at room temperature. Sections were washed and briefly dipped into 0.3% Sudan Black in 70% ethanol prior to mounting.
Electrophysiology
A coverslip with cultured cells at a density of 60,000 cells/cm2 was placed in a continuously perfused bath, viewed under IR-DIC optics and whole-cell voltage clamp recordings were performed (-70 mV, room temp.). The extracellular solution consisted of (in mM): 145 NaCl, 2.5 KCl, 1 MgCl2, 2 CaCl2, 10 HEPES and 10 dextrose, adjusted to 7.3 pH with NaOH and 320 mOsm with sucrose. Whole-cell recordings were made with pipette solutions consisting of (in mM) 140 KCl, 10 EGTA, 10 HEPES, 2 Mg2ATP and 20 phosphocreatine, adjusted to pH 7.3 with KOH and 315 mOsm with sucrose. Excitatory synaptic events were isolated by adding 10 μM bicuculline to block GABA (subscript A) receptors. Miniature synaptic events were isolated by adding 100 nM tetrodotoxin to prevent action potentials. mEPSCs were detected using the template-matching algorithm in Axograph X, with a template that had 0.5 ms rise time and 5 ms decay. Statistics were computed using the Statistics Toolbox of Matlab.
T7 Endonuclease 1 Assay and off-target analyses
Genomic PCR was performed around each sgRNA target, and related off-target sites, following the manufacturer’s instruction (using AccuPrime HiFi Taq using 500ng of genomic DNA). Products were then purified using Wizard SV Gel and PCR Clear-Up System (Promega), and quantified using a Qubit 2.0 (Thermo Fischer). T7E1 assay was performed according to manufacturer’s instructions (New England Biolabs). Briefly, 200ng of genomic PCR was combined with 2μL of NEBuffer 2 (New England Biolabs) and diluted to 19μL. Products were then hybridized by denaturing at 95 °C for 5 minutes then ramped down to 85 °C at -2°C/second. This was followed by a second decrease to 25°C at -0.1°C/second. To hybridized product, 1 μL T7E1 (M0302, New England Biolabs) was added and mixed well followed by incubation at 37°C for 15 minutes. Reaction was stopped by adding 1.5 μL of 0.25M EDTA. Products were analyzed on a 3% agarose gel and quantified using a Gel Doc XR system (BioRad). Off-target sites were identified and scored using Benchling (www.benchling.com). The top 5 off-target sites – chosen on the basis of raw score and irrespective of being in a coding region – were identified and analyzed using T7E1 assay as previously described. For TIDE (Tracking of Indels by DEcomposition) analyses (Brinkman et al., 2014), PCR was performed on genomic DNA using Accuprime Taq HiFi (Thermo Fischer) according to manufacture specifications. Briefly, reactions were cycled at 2 min at 94°C followed by 35 cycles of 98°C for 30 seconds, 58°C for 30 seconds, and 68°C for 2 minutes 30 seconds and a final extension phase of 68°C for 10 minutes. Products were then subjected to Sanger Sequencing and analyzed using the TIDE platform (https://tide.nki.nl/). The primers used for TIDE analyses are listed in Supp. Table 1.
Deep Sequencing Sample Preparation and data analysis
Genomic PCR was performed using AccuPrime HiFi Taq (Life Technologies) following manufacturer’s instructions. About 200-500 ng of genomic DNA was used for each PCR reaction. Products were then purified using AMPure XP magnetic bead purification kit (Beckman Coulter) and quantified using a Nanodrop2000. Individual samples were pooled and run on an Illumina HiSeq2500 High Throughput at a run length of 2x125bp. A custom python script was developed to perform sequence analysis. For each sample, sequences with frequency of less than 100 reads were filtered from the data. Sequences in which the reads matched with primer and reverse complement subsequences classified as target sequences. These sequences were then aligned with corresponding wildtype sequence using global pairwise sequence alignment. Sequences that were misaligned through gaps or insertions around the expected cut site were classified as NHEJ events. The frequency, length, and position of matches, insertions, deletions, and mismatches were all tracked in the resulting aligned sequences.
Statistical analysis
Statistical analysis was performed and plotted using Prism software. Student’s t-test (unpaired) or one-way ANOVA test was used to compare two or more groups respectively. A P-value <0.05 was considered significant.
Accession Numbers
Raw reads from sequencing will be available at NCBI Bioproject PRJNA417829 upon publication.
AUTHOR CONTRIBUTIONS
J. Sun and S. Roy designed the overall experiments, analyzed the data and assembled the manuscript; J. Sun performed most of the experiments. J. Carlson-Stevermer and K. Saha did most of the genomic analyses and analyzed the data. U. Das (UCSD) generated some constructs and performed live cell trafficking experiments in cultured neurons. M. Shen and X. Zhao (UW-Madison) planned/executed the hippocampal injection experiments; and M. Delenclos and P. McLean (Mayo Clinic, Jacksonville) planned/executed the intracerebroventricular injection experiments. A. Snead and A. Sproul (Columbia) planned and executed the iPSC experiments. L. Wang and J. Loi helped with cell culture and some data analyses. A. Petersen, M. Stockton, and A. Bhattacharyya planned/executed experiments with human embryonic stem cells; M. Jones did electrophysiology in cultured neurons. The overall idea was conceived by S. Roy.
DECLARATION OF INTERESTS
S. Roy and J. Sun have filed patent applications related to this work. The other authors have no competing interests.
ACKNOWLEDGEMENTS
Methods are reported in the Supplement. Raw reads from sequencing will be available at NCBI Bioproject PRJNA417829 upon publication. We thank Sue Yeon Yi (UW-Madison) for help with constructs; and Karen Jansen-West and Lillian Daughrity (Mayo Clinic, Jacksonville) for AAV packaging and purification. This work was supported by NIH grants (R01AG048218 and R21 AG052404), and a “UW2020 grant” from the University of Wisconsin-Madison to SR.