SHMT2 Mediates Small-Molecule-Induced Alleviation of Alzheimer Pathology Via the 5'UTR-dependent ADAM10 Translation Initiation

Abstract

It is long been suggested that one-carbon metabolism (OCM) is associated with Alzheimer's disease (AD), whereas the potential mechanisms remain poorly understood. Taking advantage of chemical biology, that mitochondrial serine hydroxymethyltransferase (SHMT2) directly regulated the translation of ADAM metallopeptidase domain 10 (ADAM10), a therapeutic target for AD is reported. That the small-molecule kenpaullone (KEN) promoted ADAM10 translation via the 5' untranslated region (5'UTR) and improved cognitive functions in APP/PS1 mice is found. SHMT2, which is identified as a target gene of KEN and the 5'UTR-interacting RNA binding protein (RBP), mediated KEN-induced ADAM10 translation in vitro and in vivo. SHMT2 controls AD signaling pathways through binding to a large number of RNAs and enhances the 5'UTR activity of ADAM10 by direct interaction with GAGGG motif, whereas this motif affected ribosomal scanning of eukaryotic initiation factor 2 (eIF2) in the 5'UTR. Together, KEN exhibits therapeutic potential for AD by linking OCM with RNA processing, in which the metabolic enzyme SHMT2 "moonlighted" as RBP by binding to GAGGG motif and promoting the 5'UTR-dependent ADAM10 translation initiation.

Keywords: 5′UTR; ADAM10; Alzheimer's disease; Kenpaullone; RNA binding protein; SHMT2; translation inhibitory element.

Figure 1

Kenpaullone (KEN) increases ADAM10 protein levels in human and murine cells. A) Full‐membrane immunoblots of ADAM10 show the immature (im‐ADAM10) and mature (m‐ADAM10) forms at a molecular weight of ≈80 and 60 kDa, respectively, in SH‐SY5Y cells treated with increased concentration of Kenpaullone (KEN). B) Chemical structure of KEN (top), immunoblots (middle), and quantification (bottom) of ADAM10 in SH‐SY5Y cells treated with KEN (0.25–2 µM for 36 h, n = 5). C) Chemical structure of KEN analog alsterpaullone (ALS, top), and immunoblots (middle) and quantification (bottom) of ADAM10 in SH‐SY5Y cells treated with ALS (0.5–10 µM for 36 h, n = 5). D) Chemical structure of another CDK/GSK inhibitor AT7519 (top), and immunoblots (middle) and quantification (bottom) of ADAM10 in SH‐SY5Y cells treated with AT7519 (1 to 20 µM for 36 h, n = 5). E) Time‐course effect of KEN (0.75 µM) on ADAM10 protein levels in SH‐SY5Y cells (n = 5). (F‐H) Immunoblots (top) and quantification (bottom) of ADAM10 treated with KEN (0.25–2 µM for 36 h) in human HEK293 cell line F), murine hippocampal cell line G) and primary cortical neurons (H), respectively (n = 5). (I) Immunofluorescent images (top) and quantification (bottom) of ADAM10 signals (green) in SH‐SY5Y cells treated with KEN (0.75 µM for 36 h). DAPI is a nuclear marker (blue), scale bar = 25 µm (n > 10). (J) Cell viability of SH‐SY5Y cells treated with KEN (0.25 to 2 µM for 36 h) measured by CCK‐8 kit (n = 5). (K) Immunoblots (top) and quantification (bottom) of sAPPα, ADAM10, APP, and BACE1 in SH‐SY5Y‐APP cells treated with KEN (0.75 µM for 36 h, n = 4). (L) Aβ1‐42 levels in conditioned medium of SH‐SY5Y‐APP cells treated with KEN (0.75 µM for 36 h, n = 6). All values were normalized to vehicle control (CTRL) in each experiment. Data are presented as mean ± SEM from three or more independent experiments. n.s: no significant difference. ^* p < 0.05, ^** p < 0.01.

Figure 2

KEN‐induced ADAM10 translation is dependent on the 5′UTR. A) Relative mRNA levels of ADAM10 in SH‐SY5Y cells treated with KEN (0.75 µM for 36 h, n = 8). B) Immunoblots (top) and quantification (bottom) of ADAM10 in SH‐SY5Y cells treated with KEN (0.75 µM for 36 h), in the absence or presence of transcription inhibitor actinomycin D (ActD, 0.1 µM for 12 h) or protein synthesis inhibitor cycloheximide (CHX, 10 µM for 8 h), respectively (n = 5). C) Immunoblots (top) and quantification (bottom) of ADAM10 in SH‐SY5Y cells treated with KEN (0.75 µM for 36 h), in the absence or presence of lysosome inhibitor chloroquine (CQ, 100 µM for 6 h) or proteasome inhibitor MG132 (1 µM for 6 h), respectively (n = 5). D) Immunoblots (left) and quantification (right) in SH‐SY5Y cells incubated with KEN (0.75 µM for 36 h) and transiently transfected with CDK5 siRNA (siCDK5) for 48 h (n = 3). E) Immunoblots (left) and quantification (right) in SH‐SY5Y cells incubated with KEN (0.75 µM for 36 h) and transiently transfected with GSK‐3β siRNA (siGSK‐3β) for 48 h (n = 3). F) Immunoblots (left) and quantification (right) of ADAM10 in SH‐SY5Y cells treated with KEN (0.75 µM for 36 h), in the absence or presence of a competitive inhibitor of eIF4E/eIF4G (4EGI1, 50 µM for 24 h, n = 4). G) Immunoblots (left) and quantification (right) of ADAM10 in HEK293 cells transiently transfected with human ADAM10 constructs in which the 5′UTR sequence was deleted (−5′UTR) or included (+ 5′UTR), in the absence (CTRL) or presence of KEN (0.75 µM for 36 h, n = 3). H) Left: the schematic diagram shows that the 5′UTR of ADAM10 is truncated into different fragments and subcloned into a pGL4.17 vector to construct the luciferase reporter plasmids: pGL4.17‐ADAM10‐D, C, E1/E2 and F, respectively. Numbers indicate the relative positions with respect to nucleotides in the 5′UTR. Right: relative luciferase activities in SH‐SY5Y cells transiently transfected with different pGL4.17‐ADAM10 plasmids in the absence (CTRL) or presence of KEN (0.75 µM for 36 h). The tested luciferase activity in each group was normalized to the value of the internal control plasmid pGL4.17. I) Relative luciferase activities in SH‐SY5Y cells transiently transfected with pmirGLO/pGL4.51 plasmid that was without (vector) or with BACE1 5′UTR for 12 h, in the absence (CTRL) or presence of the following KEN (0.75 µM for 36 h) treatment. Data are presented as mean ± SEM from three or more independent experiments. All values were normalized to vehicle control (CTRL) in each experiment. n.s: no significant difference. ^* p < 0.05, ^** p < 0.01.

Figure 3

KEN enhances ADAM10 and rescues cognitive deficits in APP/PS1 mice. A) Schematic diagram represents the time course of the in vivo test of KEN. B) Immunoblots (left) and quantification (right) of APP, sAPPα, ADAM10, Tau396, Tau262, and NrCAM in the hippocampus. Wild‐type (WT) and APP/PS1 mice (female at 12 months) were administered intraperitoneally with vehicle (CRTL) or KEN (7 mg/Kg, every other day) for 2 months, leading to four groups: WT, WT‐KEN, APP/PS1 and APP/PS1‐KEN, respectively (n = 6). C) Immunohistochemical images (left) and quantification (right) of Aβ deposition in the brain of APP/PS1 and APP/PS1‐KEN, scale bar = 500 µm (n = 6). D) The soluble and insoluble Aβ1‐40/1‐42 levels in the hippocampus of APP/PS1 and APP/PS1‐KEN, respectively (n = 5). E–J) Spatial and associative learning and memory performances are assessed by Morris water maze tests E–H) followed by fear conditioning tests (I&J), respectively. E) The heatmap shows in the hidden platform tests, the time (seconds, s) spent in searching for the platform (average escape latency) in different trial days (1–5), is compared among WT, WT‐KEN, APP/PS1 and APP/PS1‐KEN, respectively. n = 9 to 11. F,G) To seek the site where the hidden platform was previously located, the time period of staying (F, staying time) and the number of times crossed the site (G, passing times) are compared. H) Representative movement trajectories of mice in different groups on the sixth day. I,J) The heatmaps show the number of freezing (I, freezing times) and duration of freezing (J, freezing time in seconds) in different groups, and on different trial days (1–3). n = 6–8. All values were normalized to vehicle control (CTRL) in each experiment. Data are presented as mean ± SEM from three or more independent experiments. n.s: no significant difference. ^* p < 0.05, ^** p < 0.01.

Figure 4

SHMT2 is a target gene of KEN and the 5′UTR‐interacting RBP. A) Volcano plots of differentially expressed genes in SH‐SY5Y‐APP cells induced by KEN (0.75 µM for 36 h). SHMT2, along with other mitochondrial genes associated with one‐carbon metabolism including ALDH1L2, MTHFD2, and MTHFD1L, is significantly increased by KEN. B) KEGG pathways were analyzed from the significantly up‐regulated genes by KEN. C) A simplified PPI network shows SHMT2 interacts with other mitochondrial genes upregulated by KEN. PPI‐network was built according to KEGG database and constructed by Cytoscape. Com: complex. D) Whole lysates of SH‐SY5Y cells treated with or without KEN (0.75 µM for 36 h) were mixed with BrU‐labeled 5′UTR and anti‐BrU conjugated beads, the RNA‐proteins recognized by BrU antibody were subjected to SDS/PAGE, and silver‐stained protein bands were used for LC/MS‐MS analysis. E) A simplified PPI‐network based on STRING online‐analysis shows that SHMT2 is associated with the classical RBPs including YBX1 and CSDE1, and ribosome complex proteins. F) Immunoblots of YBX1, CSDE1, and SHMT2 by the 5′UTR‐targeted RNA‐pulldown assay. G) Immunoblots (top) and quantification (bottom) of SHMT2 protein in SH‐SY5Y cells treated with KEN at indicated concentrations for 36 h (n = 3). (H) Immunoblots (top) and quantification (bottom) of ADAM10 in HEK293 cells incubated with KEN (0.75 µM for 36 h, n = 4) and transiently transfected with SHMT2 siRNA (siSHMT2) for 48 h. (I) Immunoblots (top) and quantification (bottom) of SHMT2 and ADAM10 in SH‐SY5Y cells treated with KEN (0.75 µM for 36 h), in the absence or presence of SHMT2 inhibitor Glycyrrhetinic acid (GA, 100 µM for 48 h) or the 5′UTR‐dependent ADAM10 enhancer Cosmosiin (Cos, 5 µM for 36 h), respectively (n = 3). All values were normalized to vehicle control (CTRL) in each experiment. Data are presented as mean ± SEM from three or more independent experiments. n.s: no significant difference. ^* p < 0.05, ^** p < 0.01.

Figure 5

SHMT2 knockdown attenuates the effect of KEN on ADAM10 and cognitive function in APP/PS1 mice. A) Immunoblots (left) and quantification (right) of APP, ADAM10, sAPPα and SHMT2 in the hippocampus of APP/PS1 mice (male at 6‐month) injected with AAV vehicle (Vehicle), AAV‐SHMT2 shRNA in the absence (sh‐SHMT2‐CTRL) or presence of KEN (sh‐SHMT2‐KEN), n = 6 in each group. B) Immunofluorescence images (left) and quantifications (right) of Aβ load (6E10, red) in the hippocampus of APP/PS1 mice of three different groups. NeuN (green) is a marker of neuron, n = 6 in each, scale bar = 100 µm. C) Aβ1‐40/1‐42 levels measured by ELISA in the hippocampus of Vehicle, sh‐SHMT2‐CTRL and sh‐SHMT2‐KEN, respectively. D–G) Cognitive performances assessed by Morris water maze test, showing the escape latency D), staying times E) and passing times F), and the representative movement trajectories G) in different groups indicated, n = 9–10 in each. H–K) Novel object recognition tests show the total investigation time in two different object regions (H), the exploration time in the novel (the round one) object region I), the average discrimination index (DI, J), and the representative movement trajectories K) in APP/PS1 mice of three groups indicated, n = 7–8 in each. All values were normalized to vehicle control (CTRL) in each experiment. Data are presented as mean ± SEM from three or more independent experiments. n.s: no significant difference. ^* p < 0.05, ^** p < 0.01.

Figure 6

SHMT2‐targeted RNAs are involved in AD and the cellular function of KEN. A) Motif analysis of SHMT2‐targeted genes. Left, the Venn diagram shows SHMT2 recruits a total of 6266 and 6874 transcripts in control (CTRL) and KEN‐treated cells, respectively, in which 5226 genes are shared. Right, the most significant binding motifs recognized by SHMT2 are common in both CTRL and KEN, using MEME program. The horizontal axis denotes the base position in the corresponding motif, and the vertical axis shows the bit score with an E‐value of 4.9e‐004 (top) and 2.5e‐003 (bottom), respectively. B,C) KEGG analyses of SHMT2‐targeted RNAs show the most significantly enriched pathways, which are either shared by control (CTRL) and KEN (B), or specifically regulated by KEN relative to CTRL (C). D) A fraction of SHMT2‐targeted mRNAs affected by KEN are also included in KEN‐induced DEGs. Left, the Venn diagram shows 112 genes that are overlapped between SHMT2‐targeted mRNAs and the DEGs induced by KEN. Right, GO analysis of these overlapped genes.

Figure 7

SHMT2 binds to GAGGG motif and controls the 5′UTR activity. A) Schematic diagram shows the ADAM10 5′UTR (Human) sequence that contains GAGGG or GAAAG at indicated positions (top). Different fragment sequences are numbered corresponding to positions in full‐length 5′UTR, in which GAGGG or GAAAG motif is marked red (bottom). Individual RNA sequence was labeled with biotin at the 5′ends. (B) RNA EMSA images show the binding of human SUMO‐tagged recombinant SHMT2 to different fragments of human 5′UTR. RNA‐protein complex (RPC) is only presented in fragments containing GAGGG or GAAAG motif. C) The numbered fragment sequences where GAGGG or GAAAG motif was mutated as indicated (top); and RNA EMSA images show that SHMT2 does not bind to the mutated fragments (bottom). (D) RNA EMSA images show human SHMT2 binding to 5′UTR in the presence of a 100‐fold excess of non‐labeled RNA probe that contains the same numbered sequence. E) The schematic diagram shows the murine ADAM10 5′UTR sequence that contains GAGGG; the numbered sequences with or without GAGGG are shown (top). Murine SUMO‐tagged recombinant SHMT2 (SUMO‐SHMT2) only binds to the numbered sequence that contains GAGGG as shown by RPC (bottom). SUMO: small ubiquitin‐like modifier. Each experiment was repeated for at least three times. F) Relative luciferase activity of the 5′UTR of human ADAM10 in HEK‐293 cells cotransfected with interfering RNA of CTRL (si‐CTRL) or SHMT2 (si‐SHMT2), in the presence or absence of KEN (1 µM for 36 h), n = 6. All values were normalized to vehicle control (CTRL) in each experiment. Data are presented as mean ± SEM from three or more independent experiments. n.s: no significant difference. ^* p < 0.05, ^s p < 0.01.

Figure 8

GAGGG mutation relieves the inhibition of the 5′UTR activity by eIF2 knockdown. (A&B) Predicted secondary structures of the 5′ UTR of human ADAM10 (NM_0 01110), in which GAGGG motif at 26 and 56 nt A) is mutated to GAUUG B). The colored bars (bottom) denote base paring probability, with the corresponding free energy values of the entire RNA structure. C) Relative luciferase activities of the 5′UTR that contain either GAGGG or GAUUG (mut‐GAGGG) in HEK293 cells in the absence (CTRL) or presence of KEN (1 µM for 36 h). (D) Relative luciferase activities of the 5′UTR with GAGGG or mut‐GAGGG in HEK293 cells transiently co‐transfected with interfering RNA of CTRL (si‐CTRL), eIF2S1 (si‐eIF2S1), or eIF2S2 (si‐eIF2S2), in the absence (CTRL) or presence of KEN (1 µM for 36 h), respectively. Data are presented as mean ± SEM. ^* p < 0.05, **p < 0.01 (ANOVA, n = 6). (E) A schematic model shows that KEN increases SHMT2 expression and the resulting ADAM10 translation, thus inhibiting Aβ generation from APP (top). In the 5′UTR of ADAM10, GAGGG motif functions as TIE by intrinsic base‐paring, and SHMT2 that directly binds to this motif might facilitate structural alteration of the 5′UTR, allowing an easy read‐through process of eIF2 complex, and the enhanced translation efficiency (bottom). All values were normalized to vehicle control (CTRL) in each experiment. Data are presented as mean ± SEM from three or more independent experiments. ^* p < 0.05, ^** p < 0.01.