The level of PTPRT expression decreased in the brains of human AD and model mice
PTPRT is highly expressed in the mammalian brain and regulates STAT3 phosphorylation. Its function in neurodevelopment has been reported. To investigate whether and how alteration in PTPRT signaling occurs in age-related neurodegenerative diseases such as AD remains unclear, first, the RNA-seq data set of human AD brain samples were downloaded from NCBI and analyzed. Surprisingly, Ptprt mRNA expression was significantly downregulated in several brain regions of human AD including PFC, HP, EC, and TC compared to that of controls (Fig. 1a, Figure S1a). Next, using immunohistochemistry (IHC) the levels of PTPRT in age- and sex-matched postmortem human control and AD brain samples were observed. Indeed, a significant decrease in the intensities of PTPRT signals was found in the PFC and hippocampus of AD compared to age-matched control samples (Fig. 1b-c). Western blot data from human AD brain tissues further confirmed a decrease in the protein level of PTPRT full length and two CTF cleavages (Fig. 1d-e).
Next, to assess whether AD mouse models can recapitulate decreased mRNA expression and protein level of PTPRT found in human AD brain, first, RNA-seq data from 3xTg and APP/PS1 mouse brain samples were downloaded from NCBI, and the dynamics of Ptprt expression in APP/PS1, 3xTg, and wild type mice were analyzed. Interestingly, the levels of Ptprt mRNA showed significantly decrease in 1-year-old 3xTg and APP/PS1 mouse brains. Interestingly, the expression of ptprt would reach a peak at 4-month-old, and then decreases gradually with the age growing in both mouse cortex and hippocampus. While wild type, APP/PS1, and 3xTg mice showed similar patterns, the mRNA expression level of Ptprt in both APP/PS1 and 3xTg mice is much lower than that of wild type (Fig. 1f, Figure S1b). We further found that the level of Ptprt expression showed a significant negative correlation with b-amyloid (r2 = 0.1796, P = 0.0064) and pathological burden (r2 = 0.2310, P = 0.001) in APP/PS1 mice (Fig. 1g). Despite a strong age- and Aβ pathology-related decrease in the levels of Ptprt expression, there is no correlation between Ptprt expression and tau pathology in 3xTg (Figure S1c). Next, using immunohistochemistry and Western blot we further verified the decrease in the levels of PTPRT in PFC and CA1 of APP/PS1 mice (Fig. 1h-k).
The level of pSTAT3Y705 increased in the brains of human AD and model mice
Accumulation of pSTAT3Y705 has been reported in the brain of neurodegenerative diseases. To detected whether decreased PTPRT could affect the level of pSTAT3Y705 in AD brains, first, using immunostaining we examined pSTAT3Y705 in the brains of human AD and APP/PS1 mice. Indeed, the intensities of pSTAT3Y705 were significantly upregulated in PFC and hippocampal CA1 of human AD and APP/PS mouse brains compared to control groups (Fig. 2a-b, 2e-f). Western blot data from human and mouse brain tissues further confirmed the decrease in the level of pSTAT3Y705 in AD brains (Fig. 2c-d, 2g-h). Lastly, using primary cortical neurons we observed a downregulated PTPRT and an upregulated pSTAT3Y705 after Ab treatment at DIV14 (Fig. 2i-k). In the late stage of AD, the brain is characterized by serious neurodegeneration and neural loss in the hippocampus, which is the most vulnerable brain area to Ab deposition. This data suggests that the downregulation of PTPRT expression could lead to accumulation of pSTAT3Y705 in the brains of human AD and model mice.
PTPRT is a novel substrate of ADAM10- and presenilin 1/g-secretase
Extracellular domain shedding in the juxtamembrane region is necessary for the subsequent cleavage by presenilin 1/g-secretase complex. To determine whether PTPRT could undergo sequential processing, we treated the HEK293 cells overexpressed with C terminal FLAG-tagged PTPRT (intracellular domain) with different compounds. DAPT (γ-secretase inhibitor) could increase the accumulation of the smaller C terminal region fragment (CTF2) while it didn’t affect the bigger C terminal region fragment (CTF1), which is already known to be produced by furin cleavage. PMA, as an a-secretase activator via the PKC pathway, could add to CTF1 accumulation by DAPT inhibition. The treatment with ADAM/matrix metalloproteinase inhibitor GM6001 could decrease the CTF2 accumulation (Fig. 3a-b). Treatment with proteasome inhibitor MG132 solely helped us detect the accumulation of intracellular fragments (PICD), which is a little bit smaller than CTF2 (Red arrow represents PICD). We further examined endogenous PTPRT cleavage in primary neurons with an antibody against the PTPRT C-terminal. As expected, a similar result was found (Fig. 3c-d). ADAM10 and ADAM17 are the two most important α-secretases responsible for the ectodomain shedding. We tested g-secretase cleavage in ADAM10 KO and ADAM17 KO HEK293 cells[34], respectively. As we could see, CTF2 didn’t accumulate significantly after DAPT treatment in ADAM10-KO cells compared with WT cells (Fig. 3e-f). It suggested that ADAM10 rather than ADAM17 is the major secretase for PTPRT shedding. For further verification of PTPRT cleavage by γ-secretase, we transfected wild type (MEF-WT) or Psen1-/-/Psen2-/- double knockout mouse embryonic fibroblasts (MEF-DKO) with mouse Ptprt construct. PTPRT CTF2 was absent in WT cells while present in DKO cells (Fig. 3g-h). As the main component of the γ-secretase complex for intramembrane proteolysis, presenilin 1 is found to cut γ-secretase substrates without the help of presenilin 2. We expressed Psen1 delivered by lentivirus to see if it could recuse the processing. Indeed, PTPRT CTF2 could be detected in Psen1 expressed MEF cells. Furthermore, γ-secretase could cleave PTPRT and generated an intracellular domain (PICD) at 37℃ in plasma membrane separated from Ptprt-overexpressed HEK293 cells; and the processing could also be inhibited by DAPT in which the PICD signals were diminished in contrast to an increased level of CTF2 (Figure S2a-b). Together, those results demonstrate the sequential cleavage of PRPRT by ADAM10 and presenilin1/γ-secretase.
pSTAT3Y705 is dephosphorylated by PTPRT in cancer cells. Nevertheless, Ptprt is preferentially expressed in the central nervous system, thus it is worth investigating whether PTPRT dephosphorylates pSTAT3Y705 in the brain, especially in neurons. First, we examined the dynamics of PTPRT in the developmental mouse brain. Indeed, the full length of PTPRT showed an increasing trend with the postnatal age (Fig. 3i-j). On the contrary, the level of pSTAT3Y705 presented a decreasing tendency. Unlike full length, PTPRT-CTF2, the product of ADAM10 cleavage, also decreases gradually during the postnatal stage, showing efficient cleavage of PTPRT-CTF2 by g-secretase with maturing of the brain. It seems that a low level of PTPRT may be necessary for maintaining a relatively high level of pSTAT3Y705, and g-secretase mediated processing may be involved in it. A high level of pSTAT3Y705 is essential for neurogenesis and development in the early stage, while minimum pSTAT3Y705 is needed for adults. Next, to examine whether PTPRT directly dephosphorylates pSTAT3Y705, using shRNAs against Ptprt mRNA we knocked down its expression in wild-type primary cortical neurons. As expected, knockdown of Ptprt significantly led to an increase in the level of pSTAT3Y705 (Fig. 3k-m), suggesting an important role of PTPRT in dephosphorylation of pSTAT3Y705 in neurons.
Membrane to nuclear translocation of PTPRT intracellular domain dephosphorylates STAT3
Nuclear translocation of the released intracellular domain by γ-secretase plays an important role in the regulation of intramembrane proteolysis-mediated signaling. To examine whether a nuclear localization signal (NLS) in PTPRT is responsible for the nuclear translocation of its intracellular domain, using several different types of online software we analyzed and identified that there was putative NLS (KRRKLAKKQK) at the N terminal of PICD with a high score (http://mleg.cse.sc.edu/seqNLS). To test whether the released PICD translocate to the nucleus via the leading of NLS, we cloned PICDwt and PICDdeltaNLS (without NLS) gene to a vector fused with mCherry, and transfected it to HEK293 cells (Fig. 4a). Unlike mCherry with whole-cell dispersion, PICDwt-mCherry is specifically distributed in the nucleus. The deletion of the putative NLS led to little accumulation of PICD in the nucleus. Due to rapid degradation by the proteasome, it’s hard to demonstrate the presence of PICD in the nucleus directly. As a result, we employed a Gal4-UAS based transcription activation system to confirm the nuclear localization of PICD (Figure S2c-d). The system showed that activation of α-secretase shedding by treatment of PMA increased the transcriptional signaling by more PICD nuclear translocation, while DAPT treatment decreased the PICD mediated nuclear transactivation.
The PTPRT intracellular fragment contains two phosphatase domains, so it was very interesting to see whether the released ICD after γ-secretase cleavage still takes the dephosphorylation activity. pSTAT3Y705 is the most important substrate of PTPRT, and PTPRT regulates cancer behavior by controlling STAT3 transcriptional activity in cancer cells. On the other hand, STAT3 plays a role in regulating neurodevelopment and neurodegeneration. We transfected N2a cells with PICD, and our data demonstrated that PICD decreased the total level of pSTAT3Y705 in cell lysate (Fig. 4b). As phosphorylated STAT3 at Y705 has the potential to form a dimer and translocate to the nucleus and then acts as a transcriptional factor. We separated the nuclear component from the cytoplasm and found that PICD could significantly decrease nuclear pSTAT3Y705 (Fig. 4c), indicating that nucleus localized PICD could dephosphorylate the nuclear pSTAT3Y705 more efficiently and directly.
PICD regulates gene expression
Cleavage and nuclear import have been described for several membrane receptors, yet this mechanism has previously not been implicated in PTPRT signaling. We, therefore, investigated the possibility that nuclear translocation of the PICD could alter gene transcription beyond its regulatory role in the dephosphorylation of pSTAT3Y705. To test this hypothesis, we generated inducible isogenic N2a cell lines stably expressing the mCherry‐tagged PICD as well as mCherry control and subsequently performed RNA deep sequencing (RNA‐seq). We then determined the global transcriptome changes of cells expressing the PICD versus cells expressing mCherry as control. We compute the sum of expression of all genes located in each chromosome. And the differences in gene expression between the control vector and the ICD group were presented (Fig. 5a). If Delta > 0, the sum of gene expression is up-regulated, otherwise, the sum of gene expression is down-regulated. The expression of chromosome 15 has the largest down-regulated expression between the control vector and the ICD group (-27.7%). The largest differential expression location of chromosome 15 is NC_000081.6:47539988-47540137, which can not be annotated to a defined gene. This location contributes to the main difference in chromosome 15 (the delta expression of chromosome 15 is 42659.0-58997.5=-16338.5, the delta expression of NC_000081.6:47539988-47540137 is 16273-32655=-16382). The significant difference between genes was defined with p-value <0.05 and the absolute value of log2Foldchange more than 1 (Fig. 5b). Red points represent significantly down-regulated genes with log2Foldchange < -1, green points represent significantly up-regulated genes with log2Foldchange >1. The grey points are genes without significant difference between the control vector and the ICD group. Down-regulated and up-regulated genes by PICD expression were presented in the heatmap (Figure S3a-b). The Ptprt gene is significantly up-regulated in PICD group. There are 128 genes with a significant difference between the control vector and PICD, which contains 62 down-regulated and 66 up-regulated genes.
Since the Ptprt gene is located on chromosome 2. We selected the biological process and molecular functions that contain the Ptprt gene (Fig. 5c, Figure S3c). All genes on chromosome 2 to the GO database and KEGG database were then annotated (Fig. 5d, Figure S3d). Indeed, most of those biological processes and molecular functions are related to dephosphorylation, phosphatase, or cell adhesion. The annotation of the Ptptr gene in the KEGG database is “receptor-type tyrosine-protein phosphatase T”. There are no genes that have the same KEGG annotation as the Ptprt gene. We then annotated all genes on 21 chromosomes (19+XY) to GO and KEGG databases. There are no genes that have the same KEGG annotation as the Ptprt gene too. But 978 genes have the same GO annotations with the Ptprt gene. Amongst, there is only one gene that has significantly differential expression between the control vector and the ICD group: SMAGP on chromosome 15. The Smagp gene is located on chromosome 15, which has the largest differential expression. We selected the biological process and molecular functions that contain Smagp gene (Fig. 5e, Figure S3e). All genes on chromosome 15 to the GO database and KEGG database were annotated (Fig. 5f, Figure S3f). Indeed, these biological processes and molecular functions are related to cell adhesion and recognition.
PICD prevents Aβ deposition and pathology in the hippocampus of APP/PS1 mice
AAV delivery of the Ptprt gene to the brain could be an ideal way to prevent these molecular events in vivo. However, the Ptprt-coding sequence is about 4.4 kb, together with the promoter, ITR, and other regulatory sequences the whole size will exceed the maximum accommodation. PICD with half of the full size makes it a good fit for AAV-packaging. Using microinjection of viral particles containing PICD into the hippocampus of 9-month-old mice we observed its effect on AD pathologies in vivo. The efficiency of PICD viral infection was firstly verified by immunohistochemistry and Western blot analysis. Notably, PICD significantly decreased pSTAT3Y705 levels in the hippocampus of APP/PS1 mice (Fig. 6a-c).
Next, to assess whether increased pSTAT3Y705 induced by loss of the intracellular domain of PTPRT contributes to Alzheimer’s pathogenesis as well as neurodegeneration of AD, we decided to examine the effect of the intracellular domain of PTPRT on Aβ deposition in vitro and in vivo. First, primary cortical neurons from wild type and APP/PS1 mouse embryos (E16.5) were infected with lentiviral PICD at DIV 5. Neurons were collected at DIV 14, and immunostained with pSTAT3Y705 and 6E10 antibodies. Indeed, PICD-expressing neurons showed a significant decrease in pSTAT3Y705 and 6E10 signals (Fig. 6d-e). Next, using the histological and Western blot we examined the effect of overexpression of PICD on Ab deposition and pathology. As expect, the induction of PICD led to a significantly reduced 6E10 area and decreased cleaved caspase 3 in the hippocampus of APP/PS1 mice compared to controls of the viral vector (Fig. 6b-c, 6f-g).
PICD prevents synaptic dysfunction and behavioral deficits in APP/PS1 mice
To investigate whether loss of PICD affects the basal synaptic transmission, we generated input/output (I/O) curves by measuring field-excitatory by stimulation of the Schaffer collaterals at increasing stimulus intensities at 12-month-old mice. APP/PS1 mice with microinjection of PICD viral-like particles into the hippocampus exhibited bigger fEPSP slopes and amplitudes at all stimulus intensities tested and had significantly increased maximum fEPSPs relative to wild type mice (Fig. 7a). The deficits of long-term potentiation (LTP) in AD mice have been well documented before[35, 36]. We, therefore, investigated LTP in12-month-old wild type and APP/PS1 mice. Due to impaired LTP in the APP/PS1 mice, and the smaller absolute fEPSP may account for the LTP deficits. Next, we increased the stimulus intensity of the APP/PS1 mice to match baseline fEPSP magnitudes to those of wild type mice. The stimulus intensity used to elicit LTP in the wild type sections was approximately 30% of the maximum fEPSP slopes, equating to a value of ~0.45mV/ms, which is below the max value of the wild type mice. LTP magnitudes in these experiments did not significantly differ in the percentage of potentiation when these stronger stimulus intensities were used, indicating that reduced basal transmission does not likely account for the deficits in LTP in the wild type mice. As with the APP/PS1 mice, raising baseline fEPSPs to wild type levels did not result in significantly different LTP, and thus, the data were pooled. Meanwhile, we also lowered the baseline fEPSP of the wild type mice to match APP/PS1 mice and found no difference in potentiation as compared with the normal LTP protocol. All the data were pooled from the APP/PS1 experiments, where baseline fEPSPs were matched to wild type levels, with the experiments where the normal LTP protocol was used. We also pooled the data from the wild type experiments, where baseline fEPSPs were lowered to APP/PS1 levels, with data from the normal LTP protocol. In consistent with reports from other studies, our pool data showed that the reduced LTP in the APP/PS1 as compared to wild-type mice. Indeed, LTP in the APP/PS1 mice with injected PICD viral-like particles was significantly improved despite also having weaker fEPSPs relative to wild type controls (Fig. 7b).
Next, the learning and spatial memory was further tested using the Morris water maze. The test was conducted in wild type and APP/PS1 mice after one month of microinjection of AAV-PICD into the hippocampus at the age of 9 months. Indeed, In contrast to wild type mice, the total time of exploration in the target quadrant shows that the swimming time of APP/PS1 mice in the target quadrant is significantly reduced (P = 0.0366, n = 6 mice). Nevertheless, PICD is overexpressed group can increase the swimming time of APP/PS1 mice in the target quadrant (P = 0.0069, n = 6 mice) (Fig. 7c-d). Meanwhile, while the latency of APP/PS1 mice crossing the platform in AAV-control group was significantly increased (P = 0.0367, n = 6 mice), the latency of mice crossing the platform in the overexpression PICD group was significantly reduced (P = 0.0236, n = 6 mice) (Fig. 7e-f). This date suggests that the induction of PICD in the hippocampus of APP/PS1 mice can significantly restore the learning ability and spatial learning memory compared to the control groups.