Research ArticleNeurodegeneration

Pin1 mediates Aβ42-induced dendritic spine loss

See allHide authors and affiliations

Science Signaling  20 Mar 2018:
Vol. 11, Issue 522, eaap8734
DOI: 10.1126/scisignal.aap8734

Pinpointing amyloid’s toxicity

An increase in the amount of amyloid-β (Aβ) in neurons alters calcium signaling and causes synaptic dysfunction and dendritic spine loss, which is believed to cause neurodegeneration and cognitive deficits in Alzheimer’s disease (AD). The decreased activity of Pin1, a protein that structurally alters the function of serine- and threonine-phosphorylated proteins (including amyloid precursor protein and tau in the postsynaptic space of neurons), is also associated with AD. Using mouse models of AD, Stallings et al. found that Pin1 was dephosphorylated and inactivated by the calcium-dependent phosphatase calcineurin, whose activity in postsynaptic neurons was induced by Aβ. Aβ-induced spine loss was prevented by treating mice with the calcineurin inhibitor FK506 (also known as tacrolimus), an immunosuppressant used to reduce the rejection of organ transplants, suggesting that this drug might be repurposed to treat patients with AD.


Early-stage Alzheimer’s disease is characterized by the loss of dendritic spines in the neocortex of the brain. This phenomenon precedes tau pathology, plaque formation, and neurodegeneration and likely contributes to synaptic loss, memory impairment, and behavioral changes in patients. Studies suggest that dendritic spine loss is induced by soluble, multimeric amyloid-β (Aβ42), which, through postsynaptic signaling, activates the protein phosphatase calcineurin. We investigated how calcineurin caused spine pathology and found that the cis-trans prolyl isomerase Pin1 was a critical downstream target of Aβ42-calcineurin signaling. In dendritic spines, Pin1 interacted with and was dephosphorylated by calcineurin, which rapidly suppressed its isomerase activity. Knockout of Pin1 or exposure to Aβ42 induced the loss of mature dendritic spines, which was prevented by exogenous Pin1. The calcineurin inhibitor FK506 blocked dendritic spine loss in Aβ42-treated wild-type cells but had no effect on Pin1-null neurons. These data implicate Pin1 in dendritic spine maintenance and synaptic loss in early Alzheimer’s disease.


Early stages of Alzheimer’s disease (AD) or mild cognitive impairment are characterized by memory loss and behavioral changes. These symptoms likely reflect loss of dendritic spines and their synaptic connections induced by soluble, multimeric amyloid-β (Aβ42) (1, 2). Spine and synaptic loss occurs before neurodegeneration or tau pathology, suggesting that spine preservation strategies could attenuate or prevent AD progression (1). It is not clear how regulators of spine maintenance are negatively affected during the development of AD by increased, soluble Aβ42.

One protein whose dysfunction is implicated in both spine physiology and AD pathogenesis is the peptidyl-prolyl isomerase Pin1 (3). Pin1 binds to and catalyzes the cis-to-trans conversion of amyloid precursor protein (APP), phosphorylated tau (p-tau), and other postsynaptic proteins at phosphoserine (pSer)/pThr-Pro bonds, altering their function and/or catabolism (4). Target binding is mediated through an N-terminal, WW domain, whereas isomerization occurs through a C-terminal enzymatic domain. Ser or Thr phosphorylation, typically by proline-directed kinases, such as mitogen-activated protein kinase (MAPK), protein kinase C (PKC), and cyclin-dependent kinase (CDK) families, accelerates Pin1-mediated isomerization of peptide bond by approximately 1000-fold (5). Under basal conditions, Pin1 is highly active in neuronal cell bodies, nuclei, axonal growth cones, and dendritic spines but can be inhibited by glutamatergic, outside-in signaling that induces phosphorylation of Ser16 within the WW binding domain, preventing target recognition (6). Most relevant to AD, Pin1-mediated isomerization drives APP cleavage toward the α and away from the β/γ secretase pathway, reducing Aβ42 production (7, 8). By binding to and isomerizing p-tau, Pin1 also prevents neurofibrillary tangle formation (4), again suggesting a possible role for its dysfunction in AD pathogenesis.

Consistent with this hypothesis, cortical Pin1 protein content (7) and activity (9) are often reduced or lost in end-stage AD, whereas germline Pin1 knockout (KO) accelerates AD pathology in Tg2576 mice, which overproduce Aβ42 (8). The isomerase activity of Pin1 is lost due to oxidation of Cys113, a critical amino acid within its active site (9). Whether alterations in inside-out signaling, especially at synaptic sites, can affect Pin1 activity or steady-state protein amounts during AD evolution is unknown.

Somewhat paradoxically, hippocampal slices from pure-background, germline Pin1 KO mice display increased, rather than reduced, long-term potentiation (LTP) (10) and show increased, rather than decreased, hippocampal dendritic spine density (6), phenotypes that are opposite than is predicted from the pathology seen in AD or murine AD models. Pure-background Pin1 KO mice are very difficult to breed, with only ~2 to 5% of the expected viable births. These discrepancies suggested that viable KO mice somehow complemented for the loss of Pin1, masking the true effects of gene ablation. This encouraged us to develop Pin1fl/fl mice so that Pin1 loss could be assessed after development in vivo and in vitro. Using floxed mice and neuronal cultures derived from them, we found that Pin1 is required to maintain mature dendritic spines, and that postsynaptic Pin1 is inhibited by soluble, multimeric Aβ42-mediated signaling through calcineurin, and that this event causes spine loss. These results suggest that Pin1 can be inactivated early in the evolution of AD by Aβ42 signaling that directly contributes to spine pathology. The results also suggest that previous data from germline Pin1 KO mice should be carefully reassessed.


Neocortices from AD patients show progressive losses of Pin1 along with synapses and dendritic spines (11) that can be mimicked in cultured neurons (12) or slice preparations (13) treated with multimeric Aβ42. Unexpectedly, germline Pin1 KO cortex tissues exhibit increased dendritic spine density and increased LTP in hippocampal slices (6, 10). To clarify these seemingly contradictory results, we measured total dendritic spine density of primary DIV21 (21 days in vitro) Pin1fl/fl neurons after transfection on DIV7 with Td-Tomato and either nuclear localization signal (NLS)–green fluorescent protein (GFP) (WT) or NLS-GFP-Cre (KO) (fig. S1A). Spine counts were reproducibly 35% lower in Pin1 KO versus WT cultures (3.04 ± 0.39 per 10 μm compared to 4.77 ± 0.25 per 10 μm) (Fig. 1A and fig. S1, B and C), with significant reductions in mature mushroom and stubby spines (fig. S1C). To verify the results seen in genetically modified cells, we exposed cultured neurons to recombinant WW domain fused to the transactivator of transcription (TAT) tag (TAT-WW), which we and others have previously established would enter spines directly, very rapidly, and with minimal toxicity (10, 14, 15). TAT-WW functions as a dominant-negative, preventing endogenous Pin1 from binding to target proteins (16). Spine counts were similarly reduced in WT cells transduced with TAT-WW (fig. S1, D to F) but were restored to normal in KO cells after transduction with recombinant TAT-Pin1 (Fig. 1B and fig. S1, G and H). To control for nonspecific effects of TAT, neurons were also transduced with TAT-GFP. These cells showed no changes in spine counts or spine types from untreated, control neurons (fig. S1, D to F). Therefore, postdevelopmental Pin1 loss or inhibition caused loss of mature dendritic spines that could be rescued by exogenous TAT-Pin1 but not TAT-GFP.

Fig. 1 Pin1 KO causes dendritic spine loss in vitro and vivo.

(A) Total spine counts in Pin1fl/fl neurons after transfection with Td-Tomato and either NLS-GFP [wild-type (WT), black] or NLS-GFP-Cre (KO, gray). Neurons were fixed at DIV21. (B) As described in (A), in DIV21 neurons transduced with TAT-GFP (“G”) or TAT-Pin1 (“P”). (C) Representative images of GFP immunofluorescence in hippocampi of AAV-GFP– or AAV-GFP-Cre–injected 3-month-old Pin1fl/fl mice at 3, 6, 9 and 12 DPI. Sections were counterstained with DAPI. Scale bar, 1000 μm. (D) Total synapse counts at 3, 6, 9, and 12 DPI from 24 mouse hippocampi injected at 3 months old with either AAV-GFP (WT, black) or AAV-GFP-Cre (KO, gray). (E) Imaging of GFP (green) and antibody to synaptophysin (Syn; red) in dendritic spines from hippocampal neurons described in (C). A representative composite Z-stack image of three 0.33-μm-thick step sizes is shown. Scale bar, 10 μm. Data are means ± SEM; n > 100 spines from ≥15 images and ≥3 coverslips per condition were counted. *P < 0.05 by an unpaired t test (A) or Fisher’s least significant difference (LSD) test following two-way analysis of variance (ANOVA) (B and C). NS, not significant.

To determine whether Pin1 was required for spine stability in vivo, we injected adeno-associated virus (AAV)–GFP–Cre or AAV-GFP into the hippocampus of adult Pin1fl/fl mice (Fig. 1C). Synapse numbers in projections from GFP and synaptophysin co-positive CA1 hippocampal neurons were unchanged between the controls and KO mice at 3 days post-injection (DPI) but were significantly decreased (~40%) by 6 DPI in Pin1 KO neurons (Fig. 1, D and E). Laser-captured, microdissected GFP-positive hippocampal cells were analyzed by polymerase chain reaction for recombination. Hippocampi injected with AAV-GFP-Cre showed substantial recombination that was not seen in the thalamus of these animals or in mice injected with AAV-GFP (fig. S2A). Thus, Pin1 loss both in vitro and in vivo caused spine loss.

We next assessed Pin1 abundance and isomerase activity in soluble fractions from fresh-frozen, frontal cortex tissue from AD patients and from age-matched individuals who were not diagnosed with cognitive deficit (fig. S2B). The abundance of Pin1 and βIII-tubulin protein was consistently reduced in AD samples, whereas the abundance of synaptosomal-associated protein 25 (SNAP25), a commonly used measure of synaptic number (17), was more variable compared to normal, age-matched controls (Fig. 2A) (7). Protease-coupled isomerase assay (18) specific for Pin1 revealed significantly reduced isomerase activity in AD brain lysates compared to controls (Fig. 2B). However, given the amount of immunoreactive protein, Pin1 activity was less than expected in AD samples. These data suggested that Pin1 was somehow inactivated in AD lysates. Pin1 can be oxidized at Cys113, blocking activity (9) or inhibited by outside-in glutamatergic signaling (10). To identify the mechanism, we used mouse synaptoneurosomes (SNs), a highly enriched preparation of pre- and postsynaptic neuronal connections from brain. Isomerase activity was measured in control SNs or after incubation with soluble AD brain lysates or age-matched control lysates. Control brain lysates had no effect on Pin1 activity, but AD lysates significantly suppressed activity (Fig. 2C). We were able to entirely replicate these findings with 100 nM multimerized, soluble Aβ42 (fig. S2C) or with PiB, TAT-WW, or juglone, all established Pin1 inhibitors (Fig. 2D) (12). TAT-W34A, which contains an inactivating mutation of the terminal tryptophan in the WW domain, had no effect (Fig. 2D), demonstrating that Pin1 activity in SNs was not affected by TAT-mediated transduction. These data demonstrate that synaptic Pin1 can be inhibited by soluble, multimeric Aβ42 signaling.

Fig. 2 42 inhibits Pin1 isomerase activity.

(A) Western blotting for Pin1, SNAP25, βIII-tubulin, and GAPDH (glyceraldehyde-3-phosphate dehydrogenase) in soluble fractions of temporal cortex lysates from three Alzheimer’s disease (AD) patients and three age-matched control individuals (not diagnosed with dementia; Con). (B) Pin1 activity assay in equal amounts of soluble protein extract from control and AD patient brain tissues described in (A). (C) Pin1 activity assay with murine SNs that were untreated (red) or treated for 10 min with control brain extract (+HC; black), AD brain extract (+AD; purple), or Aβ42 (green). (D) Pin1 activity assay in SNs that were either untreated (red) or treated with juglone, PiB, TAT-WW, TAT-W34A, or Aβ42. Data are means of n ≥ 8 replicates per treatment. *P < 0.05 by Fisher’s LSD test following two-way ANOVA (B to D).

On the basis of the results, we determined the effects of Aβ42 on dendritic spines in Pin1 KO and WT neurons. As expected, Aβ42 significantly reduced total and stubby spine density of WT neurons to the numbers seen in untreated KO cells (Fig. 3A and fig. S3, A and B). Neither Aβ42 nor TAT-GFP had a significant effect on the total spine density of KO neurons (Fig. 3B and fig. S3, C and D). The spines of WT neurons transduced with excess TAT-Pin1 were fully resistant to Aβ42 (Fig. 3A and fig. S3, A and B), whereas KO cultures treated with TAT-Pin1 were partially sensitive to Aβ42 (Fig. 3B and fig. S3, C and D). Therefore, Pin1 plays a previously unappreciated role in dendritic spine maintenance and its loss phenocopied spine alterations seen in neurons exposed to multimeric Aβ42. These results strongly suggest that Aβ42 signaling causes spine loss by inhibiting Pin1.

Fig. 3 Calcineurin interacts with Pin1 and mediates Aβ42 signaling.

(A and B) Total spine counts for DIV21 WT (A) or DIV21 KO (B) neurons transduced with TAT-GFP (“G”) or TAT-Pin1 (“P”) ± Aβ42 (Aβ). Data are means ± SEM; n > 100 spines from ≥15 images and ≥3 coverslips per condition were counted. *P < 0.05 by Fisher’s LSD test following two-way ANOVA. (C) Western blot of catalytic and regulatory subunits of calcineurin in Pin1 immunoprecipitations from SNs prepared from P28 mice. (D) Calcineurin (CN) activity in SNs that were either untreated (−) or treated with Aβ42 (Aβ) for 10 min before lysis. Data are means ± SEM; n = 3 biological replicates. *P < 0.05 by an unpaired t test. (E) Pin1 activity assay in SNs that were either untreated (SN) or treated with Aβ42, FK506, FK506 + Aβ42, CsA, or CsA + Aβ42 before lysis. Data are means from n ≥ 8 replicates per treatment. *P < 0.05 by Fisher’s LSD test following two-way ANOVA.

Phosphorylation of Pin1 at Ser16, Ser71, or Ser138 alters activity (16, 19, 20). We immunoprecipitated (IP) Pin1 from SNs after brief treatment (10 min) with Aβ42 and analyzed for interactions and posttranslational modifications (PTMs) by mass spectrometry (MS). Calcineurin subunits were reproducibly associated with Pin1 by MS (fig. S3E) that was confirmed by IP/immunoblotting (Fig. 3C). Aβ42 induced Pin1 dephosphorylation at Ser111, Ser147, and Ser154, sites not previously reported as modified. As seen in primary cortical neurons, dendritic spines, and N2A cells (21, 22), we confirmed that Aβ42 quickly increased calcineurin activity in SNs (Fig. 3D). We therefore asked whether the calcineurin inhibitor FK506 or cyclosporin A (CsA) could attenuate Aβ42-mediated suppression of Pin1 activity in SNs. FK506 at 3 nM [median inhibitory concentration (~IC50)] or CsA at 5 nM (~IC50) completely prevented the suppression of Pin1 isomerase activity by Aβ42 (Fig. 3E). Therefore, calcineurin binds to and likely dephosphorylates Pin1 at Ser111, Ser147, and/or Ser154 in response to Aβ42 signaling, possibly altering Pin1 activity.

Application of FK506 restored spine numbers and complexity in neurons exposed to Aβ42 in vivo (23) and in vitro (24). The dendritic targets of calcineurin that are required for spine maintenance are largely unknown. As expected, FK506 completely rescued spine numbers and maturity in WT cells treated with Aβ42 (Fig. 4A and fig. S4, A and B) but had no effect on spine numbers in Pin1 KO cells irrespective of Aβ42 treatment (Fig. 4A and fig. S4, A and C). Therefore, Aβ42 and calcineurin signaling inhibit Pin1 whose activity is required for spine maintenance. Therapeutics such as FK506 that block calcineurin and preserve Pin1 activity could potentially be effective in attenuating spine losses in early AD.

Fig. 4 Dephosphorylation of Pin1 at Ser111 inhibits Pin1 activity.

(A) Total spine counts for DIV21 WT (black) or KO (gray) neurons ± FK506 (FK) ± Aβ42 (Aβ). (B) Pin1 activity assay in SNs treated with TAT-GFP (GFP), TAT-Pin1-S111D (S111D), or TAT-Pin1-S111A (S111A) ± Aβ42 before lysis. Data are means from n ≥ 8 replicates per treatment. *P < 0.05 by Fisher’s LSD test following two-way ANOVA. (C and D) Total spine counts in DIV21 WT (C) and KO (D) neurons transduced with TAT-GFP (“G”), TAT-Pin1 (“P”), TAT-Pin1-S111A (“A”), or TAT-Pin1-S111D (“D”) ± Aβ42. Data are means ± SEM; n > 100 spines from ≥15 images and ≥3 coverslips per condition were counted. *P < 0.05 by Fisher’s LSD test following two-way ANOVA (A, C, and D).

Ser111 is in very close proximity to the active-site Cys113 (25, 26). Thus, TAT-GFP (TAT control), TAT-Pin1, TAT-Pin1-S111A (phospho-null), or TAT-Pin1-S111D (phospho-mimetic) recombinant proteins were produced and added to SNs or to Pin1 WT and KO neurons, and isomerase activity or spine numbers were determined. TAT-Pin1-S111D–transduced SNs were completely resistant to Aβ42, whereas TAT-Pin1-S111A functioned as a dominant-negative similar to C113A (Fig. 4B and fig. S5A). TAT-GFP had no effect on Pin1 activity or responsiveness to Aβ42 (Fig. 4B and fig. S5A). TAT-Pin1-S111A significantly reduced spine numbers in WT cultures to the same degree as Aβ42 or as seen in untreated KO neurons (Fig. 4, C and D, and fig. S5, B to K). Conversely, TAT-Pin1 or TAT-Pin1-S111D had no effect on control WT cells but significantly increased spine counts in WT neurons treated with Aβ42 or untreated KO neurons (Fig. 4, C and D, and fig. S5, B to K). However, only TAT-Pin1-S111D was reproducibly able to rescue spine counts in KO cells treated with Aβ42 (Fig. 4D and figs. S4C and S5, H to K). In aggregate, these results suggest that Aβ42 induces calcineurin to dephosphorylate Pin1 at Ser111, rendering it inactive. Pin1-S111D is constitutively active, is resistant to Aβ42 signaling and calcineurin-mediated dephosphorylation, and supports spine numbers and complexity in both WT and KO neurons.


One of the earliest pathologic events in evolving AD is the loss of synapses and dendritic spines (27). These changes presumably underlie initial memory and behavioral deficits and occur before neuronal losses or neurofibrillary tangle formation (27). The work presented here demonstrates that a signaling cascade initiated by soluble Aβ42 leads to calcineurin activation and Pin1 inhibition, negatively affecting dendritic spine health. Furthermore, the data suggest that calcineurin inhibitors such as FK506 could attenuate spine loss due to Aβ42 signaling, potentially reducing early AD symptoms. Because spine loss can accelerate neuronal death (25), spine sparing strategies could also slow AD development.

Pin1 can regulate Aβ42 production and p-tau accumulation (3, 7, 8, 28). Discrepancies between spine counts and LTP in germline Pin1 KO and AD model mice that lose Pin1 over time led us to create and study Pin1fl/fl mice. Postnatal Pin1 loss caused a significant decrease in spine density that could be rapidly rescued by the transduction of exogenous Pin1. Corroborating data were seen in Pinfl/fl adult mice after intrahippocampal delivery of AAV-GFP-Cre, confirming that Pin1 is required for dendritic spine maintenance in mature neurons. The differences in spine density phenotype in germline and conditional KO cells suggest that Pin1 has distinct functions in dendritic spine development and spine maintenance. At this point, we do not know whether other observed phenotypes in germline KO mice including impaired numbers and differentiation of neural progenitor cells, growth cone abnormalities, and misorganization of global brain architecture (6, 2931) accurately reflect Pin1’s role in neurodevelopment.

Pin1 activity is sensitive to outside-in glutamatergic signaling (10), but the regulation by Aβ42 signaling has not been previously shown. These effects were observed with multimeric Aβ42, a phenotype similar to the suppressive effects on LTP in slices, dendritic spine density, and neurotoxicity in culture (1, 2, 30). Multimeric, soluble Aβ42 causes a variety of postsynaptic signaling abnormalities including alteration of Ca2+ homeostasis involving both intracellular and extracellular stores (32). Calcineurin is a calcium/calmodulin-dependent protein phosphatase highly expressed in normal brain and aberrantly activated in AD patients (3335), mouse models, and cultured neurons (21, 35, 36). Increased calcineurin activity disrupts synaptic networks, reduces LTP (37), increases long-term depression (38), and decreases memory (34, 3941). On the basis of our data, Aβ42-mediated effects on calcineurin can be modeled and studied in SNs, permitting detailed biochemical analysis.

One of the downstream effects of Aβ42-mediated calcineurin activation is Pin1 inactivation. These effects are likely direct because Pin1 interacts with multiple calcineurin subunits, and the loss of Pin1 activity can be prevented by mechanistically distinct calcineurin inhibitors FK506 and cyclosporine. FK506 binds FKBP12 to form a complex that inhibits calcineurin activity, whereas CsA forms a complex with cyclophilin A to block calcineurin activity (4244). FK506 has been shown to rescue Aβ42-induced dendritic spines loss, calcium dysregulation, hippocampal LTP, and behavioral deficiencies in mice (23, 32, 36, 45). A retrospective study showed that transplant patients receiving FK506 developed markedly less AD than expected (36). Patients immunosuppressed with the mammalian target of rapamycin (mTOR) inhibitor showed the expected incidence of AD, suggesting that calcineurin inhibition, rather than peripheral immunosuppression, was required to reduce AD incidence in transplant patients. Here, FK506 was able to rescue Aβ42-mediated dendritic spine loss only in Pin1-replete cultures. Whether Pin1 is also involved in upstream Aβ42 signaling that activates calcineurin remains to be determined. If so, our data would predict that Ser111-phosphorylated Pin1 would oppose calcineurin activation by Aβ42. Together, these data demonstrate that FK506 rescued Aβ42-mediated dendritic spine loss in a Pin1-dependent manner.

The rapid kinetics of Aβ42-mediated effects are most consistent with changes in PTMs of Pin1. Phosphorylation at Ser16 by PKMζ or PKA (10, 16) or Ser71 by DAPK1 (19) inhibited Pin1 activity, whereas MLK3-mediated Ser138 phosphorylation had the opposite effect (20). Polo-like kinase 1 (PLK1) phosphorylation at Ser65 stabilized Pin1 protein without alteration of isomerization activity (46). The identification of Ser111 phosphorylation as required for activity has not been described. This amino acid is within the catalytic pocket of Pin1 and very close to the active-site Cys113 (26). Phospho-null Pin1-S111A mutants showed dominant-negative activity as previously shown for C113A mutants, whereas phospho-mimetic Pin1-S111D was constitutively active, was fully resistant to Aβ42-mediated inhibition, and supported spine counts irrespective of Aβ42 treatment. On the basis of the work here and the data of others, we propose that the gradual accumulation of multimeric Aβ42 in evolving AD increases calcineurin activity, causing a progressive reduction of Pin1 activity in postsynaptic dendritic spines. Presumably, Pin1 interacts with and is required for the function of protein(s) essential for spine maintenance and complexity. As Pin1 also regulates Aβ42 production (8), we also propose that Pin1 inhibition accelerates Aβ42 production, further suppressing Pin1. The potential for calcineurin inhibition to actively interfere with this feed-forward loop makes it an intriguing therapeutic concept.



Pin1fl/fl mice were created using traditional methods (fig. S2A). Pin1fl/fl mice were maintained on a C57/Bl6 background and were bred by crossing homozygous Pin1fl/fl mice with homozygous Pin1fl/fl mice. Mice were fed a standard chow (Teklad Global 16% Rodent Diet, #2016) ad libitum and were housed in a standard 12-hour light/dark cycle. All animal use was done in accordance with protocols approved by the University of Texas Southwestern (UTSW) Medical Center Institutional Use and Animal Care Committee.

SN preparation

Pin1fl/fl mice in a C57/Bl6 background were used for all SN preparations per the method described in (12). Briefly, mouse cortices were harvested from mice that were heavily sedated with 2,2,2-tribromoethanol (Avertin). Using a 5-ml homogenizer, two brains were processed at a time with 10 strokes of pestle A and then pestle B. Combined homogenate was then prepared per the method described in (10). SN preps were aliquoted and treated with juglone, PiB, CsA, concatamerized Aβ42, FK506, or TAT proteins purified per (10). The reactions were stopped with a final concentration of 1% Triton X-100. Samples were snap-frozen and briefly stored at −80°C until assayed.

Neuronal culture

Cortices from Pin1fl/fl P1 pups were dissected, roughly chopped with a razor blade, and then placed in a 1:1 solution of 0.25% trypsin–Dulbecco’s phosphate-buffered saline (DPBS) with 10 U of deoxyribonuclease per milliliter of dissection solution. After 30 min, the digestion was stopped by the addition of minimum essential medium with 20% fetal calf serum. Cells were centrifuged for 5 min at 1000 rpm, supernatant was aspirated, and the cells were washed again with the stop solution. The medium was aspirated, and the cell pellet was resuspended in 2 ml of Neurobasal-A supplemented with B27, l-glutamine, penicillin, streptomycin, and 5% fetal calf serum. Cells were then filtered through 70- and 40-μm filters and then plated on coverslips coated with poly-d-lysine. At DIV7, cultures were transfected with Td-Tomato and NLS-GFP (wild-type) plasmids or Td-Tomato and NLS-GFP-Cre (KO) plasmids using a calcium phosphate method (47). The Td-Tomato and NLS-GFP-Cre plasmids were used as previously published (48).

DIV21 cultures were treated with 100 nM TAT-Pin1 (WT), TAT-Pin1-S111A, TAT-Pin1-S111D, TAT-WW, TAT-W34A, or TAT-GFP for 3 hours. Aβ42 (100 nM) or vehicle was added to the culture for an additional hour. FK506 (5 nM) was added 15 min before Aβ42. Cells were fixed in 4% paraformaldehyde (PFA) and 4% sucrose and analyzed for spine morphology. A Z stack was captured using a 100× objective on a Zeiss Axiovert LSM510 confocal microscope or a 63× objective with 2.73 zoom on a Leica DM6000B confocal microscope. An entire experimental data set was analyzed on one microscope. At least 15 Z stacks per group were analyzed using NeuronStudio (49), with the modifications described (48). Briefly, dendrites were traced, and the build spine function was used to identify spines. Computer-generated spine classifications were then manually reviewed by the user, and data were exported to Excel and GraphPad Prism, where statistical analysis was performed. Graphing of data was performed in GraphPad Prism v7.01. For image presentation, representative confocal Z stacks were opened in ImageJ, converted to a Z project with max intensity, and saved as a .tif file. Images were then rotated and cropped in Photoshop for insertion in the figure.

Pin1 isomerase activity assays

Pin1 substrate (1 mg) (Suc-Ala-Glu-Pro-Phe-pNA, Peptides International) was suspended in 15 μl of DMSO, then diluted 1:100 in 1 M lithium chloride in trifluoroethanol, and allowed to incubate a minimum of 30 min at room temperature (RT). Chymotrypsin was suspended at 100 mg per 1.67 ml of 0.001 N HCl. Final concentration of assay buffer was 50 mM Hepes (pH 7), 100 mM NaCl (pH 7), 2 mM dithiothreitol (DTT), bovine serum albumin (BSA; 0.04 mg/ml), and chymotrypsin (0.03 mg/ml). Frozen SN lysates were thawed on ice. Analysis was performed on a Jasco V630 spectrophotometer with a multiwell accessory at 390 nm. To perform the peptidyl-prolyl cis-trans isomerase (PPIase) assay on SN preps, 100 μl of SN lysate was premixed with 1 ml of assay buffer and incubated at least 30 min in a 25°C heat block. SN lysate (110 μl) was pipetted into the eight wells in the multiwell accessory unit and individually assessed. Each well was blanked while containing the original sample, and then 5 μl of Pin1 substrate was added and the sample was immediately analyzed every 0.1 s for 60 s. The human brain samples were run with 2 μg per assay. Data were exported to GraphPad Prism v7.01 for graphing and analysis. All isomerase assays were run with a BSA-only control to assess spontaneous isomerization of the substrate. The BSA values were subtracted from the test samples using baseline subtraction. For graphing purposes, the samples were then normalized so that the BSA was set to zero and the SN samples were set to 100%.

Human brain samples and Western blots

Soluble fraction from human brain samples was prepared according to the method of Shankar et al. (2, 50). Human brain lysate (35 μg) was run on an Any kD gel (456-9034, Bio-Rad) and transferred to nitrocellulose, blotted, and scanned using Li-Cor Odyssey and analyzed using Image Studio Lite. Antibodies used were Pin1 (SC-46660, Santa Cruz), SNAP25 (ab24737, Abcam), βIII-tubulin (801202, BioLegend), and GAPDH (60004-1-Ig, Proteintech).

42 preparation and concatamerization

Human Aβ42 was purchased from Peninsula Laboratories or Bachem America and diluted to 100 μM in 200 mM Hepes (pH 8.0). The solution was gently agitated for 48 hours at RT, aliquoted, and stored at −80°C. The resultant Aβ42 was analyzed by Western blot for concatamerization (fig. S2C) (50). Human amyloid B (82E1, Immuno-Biological Laboratories) and β-amyloid 17–24 (4G8, Covance) were used to detect oligomers.

XL bead preparation

Pin1 antibody (MAB2294, R&D Systems) or normal mouse immunoglobulin G (IgG) (SC-2025, Santa Cruz) was immobilized on Protein G Sepharose Fast Flow Beads (P3296, Sigma-Aldrich). Beads were equilibrated in PBS for 1 hour with multiple buffer changes. Antibody was added at 4 μg per 20-μl bead slurry in PBS and allowed to bind overnight at 4°C. Beads were washed three times with fresh 0.2 M sodium borate. Freshly made dimethyl pimelimidate dihydrochloride was then used at 5.5 mg/ml in 0.2 M sodium borate for 40 min at RT to cross-link the antibody to the beads. Beads were rinsed and incubated for 2 hours at RT with 0.2 M ethanolamine. Beads were washed three times with PBS and stored up to 4 weeks.

Lysate preparation

SN lysates were made as described above except no Triton X-100 was added after treatment. Samples were instead pelleted at 5000 rpm for 7.5 min at 4°C, at which point pellets were washed with chilled DPBS with calcium and magnesium (21-0303-CV, Corning). Pellets were then snap-frozen and subsequently stored at −80°C. Pellet pairs from several preparations were thawed on ice and combined with IP buffer: 1× stimulation buffer [10 mM tris (pH 7), 0.5 mM Na2HPO4, 0.4 mM KH2PO4, 4 mM NaHCO3, 80 mM NaCl, and 0.25% Triton X-100] plus Halt Protease and Phosphatase Inhibitor (1861280, Thermo Fisher Scientific). Samples underwent three freeze/thaw cycles followed by a 30-min incubation on ice with occasional mixing. Samples were then centrifuged (15,000 rpm, 15 min, 4°C). The supernatant was kept, whereas pellets were suspended in additional IP buffer and the spin was repeated. Supernatants from both spins were combined, and concentration was measured with the Pierce BCA Protein Assay Kit (23227, Thermo Fisher Scientific).

Immunoprecipitation and immunoblotting

Freshly made lysates were precleared with normal mouse IgG-XL beads for 2 hours at 4°C. After a quick spin, the precleared lysates were transferred to new tubes and were incubated overnight with Pin1-XL beads. Beads were washed three times with IP buffer and stored in 2× SDS-polyacrylamide gel electrophoresis loading buffer. MS samples were run on an Any kD gel (456-9034, Bio-Rad) and underwent a Pierce Silver Stain for Mass Spectrometry kit (24600, Thermo Fisher Scientific). A visible band around the size of Pin1 (18 kDa) was excised and sent for PTM analysis, or the entire sample was run 10 mm into a protein gel and excised for analysis of Pin1-interacting proteins by the UTSW Proteomics Core. Protein gel pieces were reduced and alkylated with DTT (20 mM) and iodoacetamide (27.5 mM). A solution of trypsin (0.1 μg/μl) in 50 mM triethylammonium bicarbonate (TEAB) was added to completely cover the gel and allowed to sit on ice, and then 50 μl of 50 mM TEAB was added and the gel pieces were digested overnight (Pierce). After solid-phase extraction cleanup with an Oasis HLB μElution plate (Waters), the resulting peptides were reconstituted in 10 μl of 2% (v/v) acetonitrile (ACN) and 0.1% trifluoroacetic acid in water. Five microliters was injected onto an Orbitrap Fusion Lumos mass spectrometer (Thermo Fisher Scientific) coupled to an UltiMate 3000 RSLCnano liquid chromatography systems (Dionex). Samples were injected onto a 75–μm–inside diameter, 50-cm-long EASY-Spray column (Thermo Fisher Scientific) and eluted with a gradient from 1 to 28% buffer B over 60 min. Buffer A contained 2% (v/v) ACN and 0.1% formic acid in water, and buffer B contained 80% (v/v) ACN, 10% (v/v) trifluoroethanol, and 0.1% formic acid in water. The mass spectrometer operated in positive ion mode with a source voltage of 2.4 kV and an ion transfer tube temperature of 275°C. MS scans were acquired at 120,000 resolution in the Orbitrap, and up to 10 MS/MS spectra were obtained in the ion trap for each full spectrum acquired using higher-energy collisional dissociation for ions with charges 2 to 7. Dynamic exclusion was set for 25 s after an ion was selected for fragmentation.

Raw MS data files were converted to a peak list format and analyzed using the central proteomics facilities pipeline (CPFP), version 2.0.3 (51, 52). Peptide identification was performed using the X!Tandem (53) and open MS search algorithm (54) search engines against the mouse protein database from UniProt, with common contaminants and reversed decoy sequences appended (55). Fragment and precursor tolerances of 20 parts per million and 0.6 Da were specified, and three missed cleavages were allowed. Carbamidomethylation of Cys was set as a fixed modification, and oxidation of Met was set as a variable modification. An additional requirement of two unique peptide sequences per protein was used for protein identification. For phosphorylation analysis, phosphorylation of Ser, Thr, and Tyr was also set as variable modifications, and phosphorylation sites were localized using ModeLS PTM Localization 6 and confirmed by manual interpretation.

Western blotting–IP interaction samples were run on an Any kD gel and transferred to nitrocellulose membrane using Trans-Blot Turbo technology. Blots were blocked in 5% BSA in tris-buffered saline. Membranes were scanned using Li-Cor Odyssey. Antibodies to Pin1 (MAB2294) and Ppp3r1 (AF1348) were purchased from R&D Systems, antibody to Ppp3ca (A300-908A-M) was from Bethyl Laboratories, and antibody to Ppp3cb (PA5-15581) was from Thermo Fisher Scientific.

Calcineurin activity assay

Lysates were made the same day as the Calcineurin Cellular Activity Assay Kit, Colorimetric (207007, Millipore) was performed per the company protocol. Plates were read at A620 (absorbance at 620 nm) by a Tecan plate reader Spark 10.

Hippocampal AAV1 stereotactic injection

AAV1.CMV.PI.eGFP.WPRE.bGH and AAV1.CMV.HI.eGFP-Cre.WPRE.SV40 were purchased from the University of Pennsylvania vector core and injected bilaterally into the CA1 of 8-week-old Pin1fl/fl mice in the UTSW Neuro-Models Facility. In total, 24 mice were used with 3 mice per condition per time point. Mice were sacrificed at 3, 6, 9, or 12 DPI. Mice were anesthetized and perfused with 4% PFA. Brains were postfixed in PFA and then transferred through a series of sucrose gradients. Sections (20 μm) were then cut with a vibratome, stained with 4′,6-diamidino-2-phenylindole (DAPI), and mounted. They were then examined for GFP in the hippocampus. For synapse counts in AAV-injected mice, sections were immunostained for synaptophysin (17785-1-AP, Proteintech) and stained with DAPI. The DAPI staining was used to localize to the CA1 region of the hippocampus. Images were then taken in the apical distal region of the CA1 for synapse analysis. Confocal 0.33-μm Z stacks of a least six sections were taken, spines were counted, and colocalization of GFP and synaptophysin was analyzed with ImageJ using the NeuronJ and SynaptJ plugins. Laser capture microdissection was used to isolate tissue from the brains of AAV-injected mice. Regions of infected (green cells) from the hippocampus and uninfected (green negative) from the thalamus were used for DNA genotyping analysis.

Statistical analysis

Experiments were replicated at least three times. Data are means ± SEM. Statistical analysis was performed with Prism (GraphPad) software. Statistical differences between means were analyzed by either an unpaired t test or Fisher’s LSD test following two-way ANOVA. Statistical significance was defined as P < 0.05.


Fig. S1. Pin1 loss causes spine loss.

Fig. S2. Pin1 successfully recombined in mouse hippocampus after AAV-GFP-Cre.

Fig. S3. Pin1 reconstitution restores mature spine counts in the presence of Aβ42.

Fig. S4. FK506 restoration of spines requires Pin1.

Fig. S5. Pin1-S111D restores spine counts in wild-type or knockout cells treated with Aβ42.


Acknowledgments: We thank the UTSW Neuropathology Core of the Alzheimer Disease Center (P30 AG12300) for the brain tissue; the UTSW Proteomics Core (CPRIT RP120613) for the MS analysis; the UTSW Live Cell Imaging Facility, a Shared Resource of the Harold C. Simmons Cancer Center (supported in part by National Cancer Institute Cancer Center Support Grant 1P30 CA142543-01), the UTSW Whole Brain Microscopy Facility (supported by the Texas Institute for Brain Injury and Repair), and the Neuro-Models Facility Core (supported by the Texas Institute for Brain Injury and Repair and the Haggerty Center for Research on Brain Injury and Repair in Stroke) for technical assistance; and members of the UTSW AD research group for their comments and suggestions. We would like to specially thank J. Shelton of UTSW Histo-Pathology Core for his assistance with laser microdissection of cells in the brain. Funding: This work was supported by the Senator Betty and Dr. Andy Andujar Endowment and the NIH (P01HL88594 to J.S.M. and R01AG055577 to I.B. and E.T.K.). Author contributions: N.R.S., M.A.O., E.T.K., I.B., and J.S.M. designed the project. N.R.S. and M.A.O. performed the experiments and analyzed the data along with J.S.M. J.H. generated and maintained the Pinfl/fl mice. N.R.S., M.A.O., and J.S.M. wrote the manuscript with contributions from E.T.K. and I.B. All authors commented on and approved the final version. Competing interests: I.B. is a consultant for Teva Pharmaceutical Industries. The other authors declare that they have no competing interests. Data and materials availability: The MS data have been deposited to MassIVE data repository (, accession ID MSV000082066. All other data necessary for interpretation have been shown in the article or in the Supplementary Materials.
View Abstract

Navigate This Article