Close this search box.

Tau depletion in human neurons mitigates Aβ-driven toxicity – Molecular Psychiatry

Generation of the isogenic MAPT−/− iPSC panels

To generate MAPT−/− iPSC lines, two parental iPSC lines from healthy individuals were used with two different targeting strategies as described in Methods – either a single gRNA targeting MAPT Exon 1 in one of the parental iPSC lines, or a pair of gRNAs targeting Exon 4 in the other, to diversify targeting options for achieving successful homozygous CRISPR-Cas9-mediated MAPT knockout. The intended edit was either a double-stranded break towards the end of Exon 1 3-bp upstream from the 3’-end of the gRNA binding region, or a 25-bp deletion towards the beginning of Exon 4 flanked by the pair of gRNAs used (Fig. 1A and Supplementary Table 1). After successive rounds of sequencing validations, we identified two MAPT−/− iPSC clones from ninety-six clones (2.1% successful targeting efficiency) of the Exon 1-targeted line and one MAPT−/− iPSC clone from seventy-two clones (1.4% successful targeting efficiency) of the Exon 4-targeted line. We also identified MAPT+/− (heterozygous knockout) and MAPT+/+ (clones that failed to have their MAPT genetically edited and remained as wild-type) from each parental line to be included in downstream experiments. These isogenic iPSC lines with a range of MAPT genotypes are herein referred to as Exon 1 or Exon 4 isogenic panels.

Fig. 1: Generation of MAPT−/− iPSCs.
figure 1

A Illustrations of gene editing strategies showing the positions of gRNA(s) in Exon 1 and 4 DNA loci. The intended double-stranded break in Exon 1 is indicated by an arrow, whereas the intended 25-bp deletion is indicated between the gRNA pair used to target Exon 4. B Sequencing results of edited MAPT loci, including all combinations of alleles in the Exon 1 isogenic panel. For the Exon 4 isogenic panel, both MAPT+/− and MAPT−/− lines harbour the 25-bp deletion at the same site. C Schematic of the cortical neuron differentiation protocol used throughout this study – iPSC lines were first differentiated concurrently to NPCs, before they were subject to lentiviral transduction for Ngn2 expression (“NPC-Ngn2 protocol”) for maturation either in co-culture with primary rat cortical astrocytes or in monoculture on coated surface.

The Exon 1-targeting strategy produced a range of genetic alterations due to a single DNA double-stranded break. In one of the DNA sequencing-confirmed Exon 1 MAPT−/− lines (MAPT−/− #1), one allele carries a 1-bp insertion whereas the other allele has a single-nucleotide mutation in addition to the insertion. Both genetic alterations resulted in a single-nucleotide frame shift which gave rise to a stop codon within Exon 1 (Fig. 1B; MAPT−/− Alleles #1 and #2). The other Exon 1 MAPT−/− line (MAPT−/− #2) has a single-nucleotide insertion as in MAPT−/− Allele #1 in both alleles. In the Exon 1 MAPT+/− line, one of the alleles sustained a 1-bp deletion (MAPT+/− Allele #1), similarly giving rise to a frame shift and a stop codon within Exon 1. The other allele experienced a 3-bp in-frame insertion, coupled with several mutations around the insertion site to causing changes in five amino acid residues without producing a stop codon (MAPT+/− Allele #2). The Exon 4 MAPT−/− and MAPT+/− lines both carry the same expected 25-bp deletion that led to a frame shift plus a stop codon within Exon 4 (Fig. 1B). A Human OmniExpress v1.2 BeadChip array (Illumina) was used post-editing to check for any gross karyotype abnormalities, and none was detected in all iPSC lines used in this study (Supplementary Fig. 1). We subsequently differentiated the Exon 1 and 4 isogenic panels into Day 30–35 NPCs and Day 50 cortical neurons as described in Supplementary Methods and illustrated in Fig. 1C, before characterising tau depletion in these isogenic panels.

Validation of the isogenic MAPT−/− iPSC panels

We collected cell lysates both for quantitative reverse-transcription PCR (qRT-PCR; Supplementary Methods and Supplementary Table 2) and western blots (Supplementary Methods and Supplementary Table 3) to probe for MAPT transcripts and tau proteins, respectively, in addition to fixing iPSC-derived cortical neurons for immunocytochemistry (ICC). We conducted qRT-PCR on cDNA derived from Day 35 iPSC-derived NPCs across multiple exons spanning the entire MAPT transcript to determine the pattern of MAPT expression in the MAPT−/− lines (Fig. 2A). Overall, a MAPT dose-dependent expression pattern was observed in both Exon 1 and Exon 4 isogenic panels, with marked reductions in transcript levels in the MAPT−/− NPCs, but not a complete elimination.

Fig. 2: Validation of MAPT−/− iPSCs.
figure 2

A qRT-PCR results using cDNA samples from Day 35 NPCs of both isogenic panels. Each bar graph title indicates the primer pairs used for MAPT transcript detection. Exon 4(del) describes the the CRISPR targeted and excised locus. The data points were normalised to the respective MAPT+/+ NPCs for each differentiation. Mean ± SEM and n = three independent cortical neuron differentiation repeats. Kruskal–Wallis with Dunn’s multiple comparison test was used for statistical analysis. B ICC of Day 50 iPSC-derived cortical neurons from both isogenic panels using Tau-12 antibody targeting to probe for total tau, showing representative images (top) and quantifications (bottom). The parameters are Tau-12+ cytoplasm normalised to the total number of nuclei (top row) and Tau-12+ area relative to the total beta-3 tubulin (B3T)+ and MAP2+ areas (bottom row). Scale bar = 50 μm. Mean ± SEM. N = two wells of neurons for the IgG control and four wells of neurons for the positive Tau-12 staining from one differentiation. Two-way ANOVA with Sidak’s multiple comparison test was performed for statistical analysis. C Western blots probing for tau using three different antibodies (Tau-1 – mid-region, Tau-5 – mid-region and Tau-46 – C-terminus) either on Day 30 (NPC) or Day 50 (neuron) of neuronal differentiation for both MAPT−/− isogenic panels. 6 ng of recombinant tau ladder was used and 5 μg of lysate was added per lane (except for the Tau-46 blot where 10 μg of lysate was added). Anti-β-actin blots were used as the expression control housekeeping protein for the lysates. Full blots are shown in Supplementary Fig. 3. D Nanopore long-read sequencing of MAPT transcripts in the Exon 4 isogenic panel Day 35 NPCs. Bar graph of MAPT transcript long-read sequencing depth (normalised across genotypes) focusing on Exon 4 showing lower abundance of MAPT transcripts in the MAPT+/− and MAPT−/− lines as compared to the MAPT+/+ line. A truncated form of Exon 4 is included in a minority of reads in the MAPT+/− and MAPT−/− lines indicated by the red arrows. E Sashimi plot illustrating Exon 4 inclusion in the MAPT+/+ line and skipping in the MAPT+/− and MAPT−/− lines. Splicing patterns supported by fewer than 10% of total reads per sample were filtered for clarity.

We then performed ICC on Day 50 iPSC-derived neurons by probing for total tau using a Tau-12 antibody (Fig. 2B) to further confirm dose-dependent tau depletion in all MAPT−/− neurons from both isogenic panels. For tau depletion characterisation by western blot, antibodies targeting either the mid-region (Tau-1 from amino acid 192 to 204 and Tau-5 from amino acid 218 to 225) or the C-terminus (Tau-46 from amino acid 404–441) of tau did not detect any 0N3R tau isoform which would be the only isoform readily expressed by the iPSC-derived NPCs and cortical neurons on Day 30 and 50, respectively (Fig. 2C).

A band corresponding to the 2N3R tau isoform was also noticeable in the Tau-46 blots across all lines, but the band was considered unspecific. Since tau isoform expression is developmentally regulated [14] i.e., expressing progressively from the shortest isoform in foetus to eventually including the full-length 2N4R tau isoform in adults, it is unlikely that the cells were expressing a more mature tau isoform without expressing the 0N3R isoform first. Furthermore, Tau-46 antibodies are known to present cross-reactivity with microtubule-associated protein 2 (MAP2) as commonly stated by commercial manufacturers, and recently shown biochemically using MAP2 peptides [20]. This likely explains the spurious band present in the Tau-46 blots as both tau and MAP2 are highly expressed in neurons. We have also probed our samples with an antibody against MAP2 and observed a strong signal which is present within the range of tau ladder like the spurious band, suggesting that the band is an isoform/truncated form of MAP2 (Supplementary Fig. 2). The extensive qRT-PCR experiments in Fig. 2A complements this observation with consistent depletion across the whole MAPT transcript, and therefore tau protein, in the edited lines.

We found, however, that by extending the western blot exposure time as well as saturating the chemiluminescence signals detected for the MAPT+/+ lines, a non-canonical band with a molecular weight lower than that of the 0N3R (smallest) isoform became detectable in both the MAPT+/− and MAPT−/− lines from the Exon 4 panel especially in the relatively more mature Day 50 neurons (Supplementary Fig. 3). As this tau-immunoreactive band was not present in the MAPT+/+ line, we reasoned that it may represent a non-canonical protein product of the MAPT gene that was generated in response to the out-of-frame deletion in Exon 4. Moreover, we noticed a faint band at the molecular weight corresponding to the 0N3R isoform position in the Exon 1 MAPT−/− #1 specifically in the Day 50 neurons (Tau-1 blot) that was undetectable in the original blot with the settings required to detect physiological levels of tau expression in the MAPT+/+ neurons. We then further tested tau expression in Day 50 iPSC-derived cortical neurons from both isogenic panels by subjecting the cell lysates for immunoprecipitation-mass spectrometry (IP-MS) analysis. The IP-MS results indicated that tau peptides were indeed detectable in all MAPT−/− lines (Supplementary Fig. 4A) but the MS signal intensity (an estimate for quantity) was very low and significantly less than that in the MAPT+/+ neurons at 1%, 0.06% and 11% for the Exon 1 MAPT−/− #1, Exon 1 MAPT−/− #2 and Exon 4 MAPT−/− neurons, respectively (Supplementary Fig. 4B). This IP-MS experiment was unable to identify the specific non-canonical tau peptide observed in the Exon 4 MAPT+/− and MAPT−/− lines, thus it is unclear if those tau peptides are functional. We therefore concluded that these MAPT−/− lines retain residual tau expression, but the expression levels are extremely low for the Exon 1 panel with an estimated ≤1% tau expression, an almost total depletion, whereas the Exon 4 MAPT−/− neurons expressed a non-canonical tau peptide level at approximately 11% relative to the MAPT+/+ neurons.

The qRT-PCR and ICC results from Fig. 2A, B, respectively, also provide indications on the residual tau expression in the Exon 1 lines and non-canonical tau peptide in the Exon 4 lines. We observed that there is a subtle but statistically significant increase in Tau-12+ neurons relative to the IgG control background by ICC in the Exon 1 MAPT−/− #1 line, but not the case in the overall Tau-12+ area across neurons, suggesting that the ≤1% tau expression detected from mass spectrometry comes from a minority of neurons (in approximately 1.5% of neurons relative to the IgG control). The Exon 4 MAPT−/− line displays higher overall Tau-12 signal as compared to the IgG control that is consistent with the western blot and mass spectrometry data. The qRT-PCR results demonstrate that the non-canonical peptide detected in the Exon 4 MAPT−/− line is not a result of an N- or C-terminus truncated transcript due to the targeted locus in Exon 4 because there is a detectable level across the entire transcript, but more likely a result of exon skipping since both N-terminal and C-terminal exons are detectable and the non-canonical peptide is only ~5 kD smaller than the 0N3R tau isoform (Supplementary Fig. 3). This is consistent with the mass spectrometry data demonstrating the presence oof both N-terminus and C-terminus peptides (Supplementary Fig. 4A).

To further elucidate the identity of the non-canonical transcript, we conducted an RT-PCR experiment using the cDNA templates from the Exon 4 isogenic panel NPCs with a forward and reverse primer spanning between Exon 0 and 11, respectively. The PCR products derived from the MAPT+/+ line resulted in a single band just below the 800 bp marker, as expected (Supplementary Fig. 5). However, three separate bands were observed in the MAPT+/− line with the top band corresponding to the MAPT+/+ band. The bottom band is situated just above the 700 bp marker, while the middle band is estimated to be less than 50 bp smaller than the top band.

We reasoned that the top and middle bands from the MAPT+/− line represent its two alleles (i.e. one of the MAPT loci has 25 bp excised from Exon 4). The bottom band, on the other hand, may represent the non-canonical transcript that is <100 bp smaller than the MAPT+/+ transcript due to Exon 4 targeting. This interpretation is confirmed by the MAPT−/− line where only the bottom two bands are present, indicating that both the Exon 4-edited and the non-canonical transcripts are present. Moreover, the non-canonical transcript (bottom) in the MAPT−/− appears to exhibit a stronger signal as compared to that in the MAPT+/− line to suggest that it is indeed caused by Exon 4 targeting. The small <100 bp difference of the non-canonical transcript is consistent with the ~5 kD difference observed on the western blot to suggest that exon skipping is likely.

Lastly, we performed Nanopore long-read sequencing on the MAPT transcripts from the Day 35 Exon 4 isogenic panel NPCs with the goal of identifying the specific non-canonical transcript seen in the Exon 4 MAPT+/− and MAPT−/− lines. A primer targeting Exon 11 was used to reverse transcribe the MAPT transcripts from the 3’ end for sequencing (Supplementary Methods). We found that Exon 4 is included in every origin read in the Exon 4 MAPT+/+ NPCs, while there is a marked reduction in Exon 4 read depth in the MAPT+/− and MAPT−/− NPCs (Fig. 2D). In the MAPT+/− NPCs, some reads span Exon 4 while others skip Exon 4 entirely or cover only part of Exon 4 until the 3’ end of the 25 bp deletion site before continuing to Exon 1 (Fig. 2D, E). Exons 3 and 2 are not incorporated into the MAPT transcripts suggesting that only 0 N tau isoforms are expressed. Among the small number of MAPT transcript reads detected in the MAPT−/− NPCs, however, none of the sequencing reads spans Exon 4 completely with most reads skipping Exon 4 while several cover part of Exon 4 before continuing to Exon 1 as in the case for the MAPT+/− line. The long-read sequencing results confirm that the non-canonical band seen in the western blots (Supplementary Fig. 3) and the peptides detected in the IP-MS experiment (Supplementary Fig. 4) are products of Exon 4 skipping due to Exon 4 targeting in the edited lines, consistent with the observation from the RT-PCR experiment (Supplementary Fig. 5).

Tau depletion protects neurons from AD brain-derived Aβ-driven hyperactivity

The Exon 1 and Exon 4 isogenic panels were then used for downstream experiments detailed in this study (Supplementary Table 4). We began by examining relatively more sensitive phenotypes such as neuronal activity and synapse loss, to axonal transport of mitochondria that involves a shorter time scale before examining relatively more severe cellular phenotypes such as neurite outgrowth impairment and neurodegeneration. To address if tau depletion affects neuronal activity, we differentiated the MAPT+/+ and MAPT−/− lines from both Exon 1 (MAPT−/− #1) and Exon 4 isogenic panels into cortical neurons (Supplementary Fig. 6) on MEA multi-well plates which have sixteen electrodes embedded at the bottom of the plates per well for extracellular field potential detection (Fig. 3A). The MAPT−/− neurons from both Exon 1 and 4 panels exhibited marked impairments in neuronal activity across all parameters quantified i.e., showing reduced firing strength, frequency, synchronicity within the neuronal network and network firing.

Fig. 3: MAPT−/− iPSC-derived cortical neurons demonstrate reductions in neuronal activity and protection from Aβ-driven hyperactivity.
figure 3

A Representative raster plots showing individual neuronal activity spikes for each of the sixteen electrodes (row) in each MEA well over 2 min for both Day 90 MAPT+/+ and MAPT−/− neurons (MAPT−/− #1 from the Exon 1 panel) from each isogenic panel at baseline; Quantification of baseline neuronal activity parameters measured by MEA assays on Day 90–100 iPSC-derived cortical neurons from both MAPT−/− isogenic panels. Mean ± SEM. n = 53–55 (Exon 1 MAPT+/+) or 51–53 (Exon 1 MAPT−/−) wells across three independent neuronal differentiation repeats; 101–105 (Exon 4 MAPT+/+) or 103–114 (Exon 4 MAPT−/−) wells across six independent neuronal differentiation repeats. Some wells did not achieve the threshold needed to register network activities. Two-tailed Mann-Whitney test was used for statistical analysis. B Representative raster plots showing individual neuronal activity spikes for each of the sixteen electrodes (row) in each MEA well over 2 min for both Day 90 Exon 4 MAPT+/+ and MAPT−/− neurons treated with either AD brain homogenate or Aβ-immunodepleted (ID) AD brain homogenate at 25% v/v in the neuronal media for 5 days; Quantification of neuronal activity parameters measured by MEA assays over 5 days on Day 90–93 Exon 4 MAPT+/+ and MAPT−/− iPSC-derived cortical neurons treated with either AD brain homogenate or Aβ-ID homogenate at 25% v/v in the neuronal media. All datapoints were normalised to the baseline recording pre-treatment, and for each time point relative to the wells subject to aCSF (vehicle) control treatment. Mean ± SEM. n = 7–14 (MAPT+/+ID), 11–14 (MAPT+/+AD) or 5–16 (MAPT−/− ID and AD) wells across three independent neuronal differentiation repeats. Two-way ANOVA with Dunnett’s multiple comparison correction was used for statistical analysis compared against the MAPT+/+AD wells 5 days post-treatment.

We then asked whether tau lowering can protect human iPSC-derived cortical neurons from exogenous toxic insults such as Aβ by treating neurons with AD brain homogenate, or with AD brain homogenate immunodepleted for Aβ (Aβ-ID) and assaying neuronal activity by MEA (Fig. 3B). The MAPT+/+ line demonstrated AD brain homogenate-driven hyperactivity over time both in terms of single-electrode neuronal firing and network firing frequencies, whereas neuronal firing amplitude and synchrony remained unaffected. This hyperactivity phenotype was not present after treatment by Aβ-ID, or after tau depletion, suggesting that the hyperactivity phenotype was specifically Aβ-driven and tau-dependent.

On the synapse level, tau depletion did not affect synaptic density in the iPSC-derived cortical neurons (Supplementary Fig. 7A and 7B). Bulk AD brain homogenate treatment led to a 20–30% synapse loss in the iPSC-derived cortical neurons regardless of their MAPT genotypes or the presence of Aβ (Supplementary Fig. 7C; Exon 4). We then performed extraction and concentration of Aβ from the AD brain homogenate as detailed in Supplementary Methods and showed that the treatment of AD brain-derived Aβ resulted in approximately 10% synapse loss in the MAPT+/+ neurons but not in the MAPT+/− and MAPT−/− neurons (Supplementary Fig. 7D; Exon 4). This indicates that AD brain-derived Aβ-driven synapse loss is tau-dependent, but that there are other soluble factors present in the AD brain homogenate that can result in synapse loss. Both bulk AD brain homogenate and brain-derived Aβ treatments were insufficient to cause synapse loss in the Exon 1 isogenic panel (Supplementary Fig. 7C, D; Exon 1).

Tau depletion mitigates Aβ-driven deficit in retrograde axonal transport of mitochondria

Since tau is mainly localised in the axons as a microtubule-binding protein [21], we next asked whether tau depletion interferes with axonal transport of mitochondria. The iPSC-derived cortical neurons were plated in one side of microfluidic chambers for live imaging of mitochondrial movement along axons with clear directionality as detailed in Methods (Fig. 4A). We did not observe any changes in the ratio (to stationary mitochondria), speed and displacement of motile mitochondria in the MAPT−/− neurons at baseline (Supplementary Fig. 8A).

Fig. 4: Aβ-driven retrograde impairment of axonal transport of mitochondria is absent in MAPT−/− iPSC-derived cortical neurons.
figure 4

A Schematic of the experiments designed to measure axonal transport of mitochondria using microfluidic chambers. B Quantification of ratio of motile mitochondria (motile to stationary) in Day 70–95 iPSC-derived cortical neurons from the Exon 4 isogenic panel with or without directionality over 150 s of live imaging. The neurons were treated with either 2 μM scrambled Aβ1–42 or Aβ1–42 oligomers for 1 h before imaging. Mean ± SEM. n = 8 (MAPT + /+ scrambled Aβ1–42), 13 (MAPT + / + Aβ1–42), 6–9 (MAPT−/− scrambled Aβ1–42) and 12–14 (MAPT−/−1–42) microfluidic chambers measured across five independent neuronal differentiation repeats. Two-tailed Mann-Whitney test was used for statistical analysis.

We subsequently investigated whether tau depletion mitigates neuronal response to Aβ in terms of impairments to axonal transport of mitochondria. Recombinant Aβ1–42 oligomers were used as a source of a more acute and toxic Aβ insult to the neurons (Supplementary Methods and Supplementary Fig. 9), as previously reported in mouse primary hippocampal neurons in vitro [8]. After one hour of exposure to Aβ1–42 oligomers, there were fewer motile mitochondria in the MAPT+/+ neurons specifically in the retrograde direction while the Aβ-driven reduction in the number of mitochondria moving towards the soma was mitigated in the MAPT−/− neurons (Fig. 4B). This suggests that the neuronal response to exogenous Aβ insult in axonal transport, at least for mitochondria as a cargo, is tau dependent. The treatment of exogenous Aβ1–42 oligomers did not result in any changes in either speed or displacement of motile mitochondria along axons in the iPSC-derived cortical neurons regardless of their MAPT genotypes (Supplementary Fig. 8B). We also found that there were no differences in mitochondrial membrane potential within the Exon 1 and 4 isogenic panels at basal level, nor in response to exogenous Aβ1–42 oligomers, indicating that the tau-dependent phenotype in axonal transport of mitochondria observed was unrelated to mitochondrial function (Supplementary Fig. 8C, D).

Tau depletion does not consistently result in neurite outgrowth impairments

We went on to investigate whether tau lowering can result in changes in neuronal morphology. To address this question, we transduced a subset of iPSC-derived cortical neurons with vectors expressing GFP alongside NGN2 and tracked their individual neurite outgrowth with live imaging over time (Supplementary Fig. 10A). Both MAPT−/− (#1 line) neurons from the Exon 1 isogenic panel and MAPT−/− from the Exon 4 isogenic panel exhibited neurite outgrowth impairment with shorter neurite and axonal lengths, as well as lower ramification index (levels of branching per root from soma) as compared to the respective MAPT+/+ lines without alterations in neurite branch length (Supplementary Fig. 10B, C). However, the MAPT−/− (#2 line) neurons from the Exon 1 isogenic panel did not suffer from any neurite outgrowth impairment and surprisingly demonstrated a reduction in neurite branch length as compared to the MAPT+/+ neurons suggesting that tau depletion alone is inadequate to consistently result in neurite outgrowth impairments.

Tau depletion protects neurons from Aβ-driven neurodegeneration

Finally, we asked whether tau lowering can mitigate Aβ-driven neurodegeneration in human iPSC-derived cortical neurons. We again used recombinant Aβ1–42 oligomers as a more acute and toxic source of Aβ to induce neurodegeneration in the iPSC-derived cortical neurons and measured the percentage of cleaved caspase 3-positive (CC3+) neurons as a readout for cell death (Fig. 5A). In both Exon 1 and 4 isogenic panels, there was a substantial increase in Aβ1–42 oligomer-driven neurodegeneration in the MAPT+/+ lines (Fig. 5B). This observation is supported by a different cell viability measurement using adenylate kinase (AK) (Supplementary Fig. 11). Tau depletion was effective in mitigating Aβ1–42 oligomer-driven neurodegeneration in both MAPT+/− and MAPT−/− neurons, a finding that was consistent in both isogenic panels, indicating that this phenotype is tau-dependent and, crucially, that partial tau reduction was also sufficient to mitigate Aβ1–42 oligomer-driven neurodegeneration.

Fig. 5: Aβ-driven neurodegeneration is absent in MAPT−/− iPSC-derived cortical neurons.
figure 5

A Representative immunofluorescence images of Day 79–83 (Exon 1 isogenic panel) and Day 79–86 (Exon 4 isogenic panel) iPSC-derived cortical neurons treated with either 10 μM scrambled Aβ1–42 or Aβ1–42 oligomers for 5 days. The neurons were probed with antibodies against human nuclei (green) and cleaved caspase 3 (CC3; yellow) which was used as the marker for cell death. Scale bar = 100 μm. B Quantification of relative CC3+ neuron count post-treatment with either 10 μM scrambled Aβ1–42 or Aβ1–42 oligomers for 5 days. Mean ± SEM. n = three (Exon 1) or four (Exon 4) independent neuronal differentiation repeats. Two-way ANOVA with Šídák’s multiple comparison correction was used for statistical analysis compared against the scrambled Aβ1–42 treatment.