Abstract

   Down syndrome (DS) is the most common chromosomal disorder and a major
   cause of intellectual disability. The genetic etiology of DS is the
   extra copy of chromosome 21 (HSA21)–encoded genes; however, the
   contribution of specific HSA21 genes to DS pathogenesis remains largely
   unknown. Here, we identified ZBTB21, an HSA21-encoded zinc-finger
   protein, as a transcriptional repressor in the regulation of synaptic
   function. We found that normalization of the Zbtb21 gene copy number in
   DS mice corrected deficits in cognitive performance, synaptic function,
   and gene expression. Moreover, we demonstrated that ZBTB21 binds to
   canonical cAMP-response element (CRE) DNA and that its binding to CRE
   could be competitive with CRE-binding factors such as CREB. ZBTB21
   represses CRE-dependent gene expression and results in the negative
   regulation of synaptic plasticity, learning and memory. Together, our
   results identify ZBTB21 as a CRE-binding protein and repressor in
   cAMP-dependent gene regulation, contributing to cognitive defects in
   DS.
     __________________________________________________________________

   ZBTB21 overdosage impairs synaptic function in Down syndrome.

INTRODUCTION

   Down syndrome (DS) is a congenital disorder in which individuals are
   born with extra or partial chromosome 21 (HSA21). The extra number of
   genetic components may result in a variety of health problems,
   including developmental retardation, intellectual disability, and
   increased risk of Alzheimer’s disease ([30]1, [31]2). Although a
   variety of abnormalities may occur in people with DS, most DS
   individuals exhibit moderate to severe cognitive impairment, which
   affects their quality of life. More than 200 genes on chromosome 21 may
   exert direct or indirect effects on DS-related deficits ([32]3, [33]4);
   however, the contribution of specific HSA21 genes to DS pathogenesis
   remains largely unknown.

   Synaptic dysfunction plays a notable role in the intellectual
   disability observed in individuals with DS ([34]5–[35]8). Synaptic
   abnormalities ([36]5) are evident in both prenatal and postnatal DS
   brains ([37]6, [38]9, [39]10), with an observed imbalance in excitatory
   and inhibitory synaptic transmission in the brains of both individuals
   with DS and DS model mice ([40]11, [41]12). Synaptic plasticity,
   essential for memory encoding in the brain, is widely recognized
   ([42]13, [43]14). In addition, defects in hippocampal long-term
   potentiation (LTP) have been detected in DS mouse models ([44]15,
   [45]16).

   Transcriptional regulation is critical for synaptic plasticity and
   memory formation. In particular, the cyclic adenosine
   3′,5′-monophosphate (cAMP)–dependent transcription pathway plays a
   central role in neuronal plasticity and long-term memory formation
   ([46]17–[47]19). The cAMP-dependent pathway has been found to function
   as a molecular switch to convert short- to long-term memory ([48]20,
   [49]21). cAMP response element–binding protein (CREB) is a key factor
   that regulates the transcription of multiple synapse-related genes by
   binding to cAMP-responsive elements (CREs) that occur either as a
   palindrome (TGACGTCA) or a half site (CGTCA/TGACG) in their promoters
   ([50]22). The CREB protein is activated by the phosphorylation of
   protein kinase A at serine 133, further recruiting the coactivator
   CREB-binding protein (CBP) and its homolog p300. The recruitment of
   CBP/p300 is necessary for transcriptional activation. In addition to
   acting as coactivators, CBP and p300 are histone acetyltransferases
   that can also modulate histone acetylation ([51]23). CBP/p300-targeted
   histone acetylation (H3K18 and H3K27) ([52]24) is essential for
   CREB-dependent transcriptional pathways ([53]25). CREB-mediated
   transcription provides a critical mechanism for regulating synaptic
   function, and defective cAMP-dependent transcription in DS mouse brains
   could contribute to DS-associated cognitive deficits ([54]26).

   Zinc finger (ZF) and broad-complex, tramtrack and bric-à-brac (BTB)
   domain–containing protein 21 (ZBTB21 or ZBT21) is a chromosome
   21–encoded protein that may play a role in transcriptional regulation
   via the ZF domain as a DNA binding motif ([55]27, [56]28). ZBTB21 has
   been reported to be involved in a number of physiological and disease
   conditions, including early embryonic development ([57]29), muscle
   development ([58]30), and cancer ([59]31). ZBTB21 is located within the
   DS critical region (DSCR) and is potentially associated with congenital
   heart defects in DS ([60]32). ZBTB21 is highly expressed in brain
   tissues, but its roles in learning and cognition and DS pathogenesis
   have not been elucidated.

   Here, we demonstrated a role for chromosome 21–encoded ZBTB21 in
   DS-associated synaptic dysfunction. Our study showed that ZBTB21
   competes with CRE motif binding with CREB and CBP/P300-mediated histone
   acetylation, leading to CRE-mediated transcriptional inhibition.
   Normalization of the Zbtb21 gene copy number in DS mice corrected
   CRE-dependent gene expression and restored synaptic function.
   Furthermore, conditional overexpression of ZBTB21 in neurons results in
   synaptic and cognitive deficits in Zbtb21 transgenic mice. In summary,
   our study identified ZBTB21 as a previously unknown transcriptional
   repressor of CRE-dependent gene expression and a pathogenic factor that
   disrupts synaptic function and cognitive performance in DS.

RESULTS

ZBTB21 acts as a cAMP response element-dependent transcriptional repressor

   To identify previously unknown CRE-binding proteins, we transfected
   human embryonic kidney (HEK) 293T cells with biotin-labeled,
   CRE-containing double-stranded DNA fragments. Following transfection,
   we performed a pull-down assay followed by quantitative mass
   spectrometry (MS) analysis ([61]Fig. 1A). The top 10 proteins enriched
   in biotin-CRE DNA pull-down are listed in [62]Fig. 1B, and their MS ion
   intensities are shown in fig. S1A. As expected, known CRE-binding
   proteins such as CREB, activating transcription factor 1 (ATF1), and
   cAMP responsive element modulator were identified. The detection of
   c-Jun, JunB, and JunD may be attributed to the leucine zipper region in
   CREB, which is known for its role in forming homodimers or heterodimers
   with other proteins that have basic leucine zipper domains, including
   components of activator protein (AP)-1. Notably, ZBTB21, a previously
   unrecognized CRE-binding protein, was also identified ([63]Fig. 1B).
   The specific interaction between the biotin-labeled CRE DNA fragment
   and ZBTB21 was further confirmed by the observation that ZBTB21 was
   pulled down by biotin-CRE but not by AP-1 ([64]Fig. 1, C and D).

Fig. 1. Identification of ZBTB21 as a suppressor of CRE-dependent gene
transcription.

   [65]Fig. 1.
   [66]Open in a new tab

   (A) Workflow diagram showing CRE-binding protein isolation using a
   biotin-labeled CRE DNA element and analysis by quantitative MS. (B)
   HEK293T cells transfected with biotin-CRE DNA element and treated with
   10 μM forsklin (FSK) for 0, 30, and 60 min; proteins were pulled down,
   followed by MS analysis. (C) Pull-down assay using biotin-DNA element
   in WT and ZBTB21^−/− HEK293T cells, analyzed by immunoblotting for
   ZBTB21 and CREB. (D) DNA sequence comparison of the ZBTB21 locus in the
   WT and ZBTB21^−/− HEK293T cells. (E and F) Effect of ZBTB21
   overexpression on FSK-induced CRE (E) and TNFα-induced NF-κB (F)
   promoter activity. (G and H) Effect of ZBTB21 KO on FSK-induced CRE
   promoter activity at 6 hours (G) and various time points (H). (I)
   Rescue of ZBTB21 KO effect on FSK-induced CRE promoter activity by
   ZBTB21 reexpression. All data represent mean ± SEM. P values were
   determined by two-tailed Student’s t tests. ns, not significant. *P <
   0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. Each sample in the
   experiments had three replicates, and independent experiments were
   performed at least two times. The data presented are representative
   results. DMSO, dimethyl sulfoxide.

   To determine whether ZBTB21 specifically regulates CRE-mediated gene
   transcription, we analyzed the transcriptional regulatory function of
   ZBTB21 using a CRE reporter. In HEK293T cells expressing CRE
   promoter-luciferase construct, ZBTB21 overexpression markedly reduced
   forskolin (FSK)-induced luciferase activity ([67]Fig. 1E). However,
   ZBTB21 failed to regulate the expression of tumor necrosis factor
   (TNF)–induced nuclear factor κB (NF-κB)–driven luciferase ([68]Fig.
   1F), indicating that ZBTB21 selectively inhibits CRE-dependent
   transcription. Furthermore, the knockout (KO) of ZBTB21 in HEK293T
   cells markedly increased FSK-induced CRE-dependent gene transcription
   ([69]Fig. 1, G and H). This enhancement in gene transcription in ZBTB21
   KO cells was eliminated by restoring ZBTB21 expression ([70]Fig. 1I).
   It needs to be noted that two splicing variants of ZBTB21 exist (fig.
   S1B), and we used the cDNA of the long ZBTB21 variant to restore the
   expression of ZBTB21 ([71]Fig. 1I). This cDNA led to splicing, as
   evidenced by the detection of two bands in Western blotting (fig. S1C).
   When the cDNA of short-form ZBTB21 was used (fig. S1C), it effectively
   suppressed CRE-dependent gene expression (fig. S1D). Although there was
   a noticeable trend indicating reduced cAMP and cAMP-dependent protein
   kinase (PKA) activity in ZBTB21^−/− cells, this difference did not
   achieve statistical significance. Furthermore, our investigations
   revealed that overexpression of ZBTB21 did not alter cAMP levels or PKA
   activity (fig. S1E). Collectively, our findings demonstrate that ZBTB21
   acts to suppress CRE-dependent transcription.

   CREB is the most studied CRE-binding protein, and we determined that
   the interaction of ZBTB21 with CRE did not require the presence of CREB
   (fig. S1F). We subsequently tested whether ZBTB21 could compete with
   the binding of CREB to the CRE element using coexpression and pull-down
   assays and found that ZBTB21 could compete with the binding of CREB or
   ATF1 to the CRE ([72]Fig. 2A and fig. S1G). Similarly, an
   electrophoretic mobility shift assay (EMSA) showed an efficient shift
   of the CRE DNA fragment by CREB, and ZBTB21 dose-dependently reduced
   the shifted CRE DNA ([73]Fig. 2B). However, the binding of ZBTB21 to
   the CRE element could not be detected as a uniform band in the EMSA
   ([74]Fig. 2B).

Fig. 2. ZBTB21 competes with CREB for binding to the CRE and the domains
essential for the suppression of CRE-dependent gene expression.

   [75]Fig. 2.
   [76]Open in a new tab

   (A) HEK293T cells were transfected with HA-ZBTB21 expression plasmid
   (0, 10, or 30 μg) and subsequently transfected with the biotin-CRE DNA
   element. The total cell lysates (input) and biotin-CRE pull-down were
   then analyzed by immunoblotting with antibodies as indicated. (B)
   Lysates of HEK293T cells overexpressing CREB and/or HA-ZBTB21 as
   indicated were mixed with biotin-labeled CRE DNA. EMSA analysis of the
   binding between biotin-CRE and CREB in the presence of different
   amounts of ZBTB21 is shown. The proteins were analyzed by
   immunoblotting with antibodies as indicated. (C) Profile heatmap around
   the TSSs of H3K27ac-modified genes. Read counts were extracted for all
   CUT&Tag-seq experiments within a region spanning ±3 kb around the TSS.
   The blue-to-red gradient indicates high-to-low counts in the
   corresponding region (P value < 0.05, |log[2]fold change| >1), n = 3.
   (D) GO and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses of
   genes containing modified regions of H3K27ac. The top enriched
   biological process terms are displayed. The size of a circle indicates
   the number of enriched genes, and the color reflects the adjusted P
   value (|log[2]fold change| >1), n = 3. (E to G) Mapping ZBTB21
   functional domains involved in transcriptional regulation. (E)
   Schematic depicting the generation of truncated ZBTB21 expression
   constructs. (F) Immunoblot analysis of HA-tagged truncated ZBTB21
   proteins. (G) HEK293T cells were cotransfected with plasmids expressing
   different domains of the ZBTB21 protein and the CRE-Luc reporter
   plasmid. Luciferase activity was measured after FSK stimulation. All
   data represent mean ± SEM. P values were determined by two-tailed
   Student’s t tests. ***P < 0.001. Independent experiments were performed
   at least two times. The data presented are representative results.

   Epigenetic modification of CBP/P300 plays a critical role in
   CRE-mediated transcription. To determine whether ZBTB21 regulates
   CRE-mediated transcription through epigenetic alterations, we infected
   mouse primary neurons with lenti-Zbtb21 (ZBTB21-OE) and analyzed the
   genome-wide DNA binding patterns of H3K27ac through cleavage under
   target and tagmentation sequencing (CUT&Tag-seq) (fig. S1H) ([77]33).
   Our findings revealed that, compared to the control group, the ZBTB21
   overexpression group exhibited a significantly reduced signal intensity
   of H3K27ac at gene peaks located in the transcription start site (TSS)
   regions [P value < 0.05, |log[2] fold change| > 1] ([78]Fig. 2C).
   Furthermore, Gene Ontology (GO) and Kyoto Encyclopedia of Genes and
   Genomes pathway enrichment analysis revealed that genes with altered
   H3K27ac regions highlighted associations with synaptic function,
   learning and memory, and cAMP-mediated signaling pathways ([79]Fig.
   2D).

   To determine the functional domains essential for suppressing
   CRE-mediated transcription, we generated different truncated ZBTB21
   constructs and examined their effects on the expression of the
   CRE-luciferase reporter ([80]Fig. 2, E to G). We found that the ZBTB21
   ΔBTB and ZBTB21 ΔZF domains were less effective at suppressing
   CRE-luciferase activity than the full-length ZBTB21 domain ([81]Fig.
   2G), suggesting that the N-terminal BTB and ZF domains are required for
   the function of ZBTB21 as a CRE transcriptional repressor.

Normalizing the Zbtb21 copy number restores synaptic and cognitive deficits
in Dp16 mice

   To determine whether Zbtb21 triplication contributes to DS
   pathogenesis, we generated wild-type (WT), Zbtb21^+/−, Dp16, and
   Dp16;Zbtb21^+/− mice by breeding Dp16 mice with Zbtb21^+/− mice
   ([82]Fig. 3A). As expected, Zbtb21 mRNA expression in Dp16 mice was
   higher than that in WT mice, and in Dp16;Zbtb21^+/− mice, it was
   comparable to that in WT mice ([83]Fig. 3B). To evaluate the cognitive
   performance of WT, Zbtb21^+/−, Dp16, and Dp16;Zbtb21^+/− mice, we
   performed a series of behavioral tests and found that the learning and
   memory abilities of the Dp16 mice were impaired and that the cognitive
   deficits were largely restored in the Dp16;Zbtb21^+/− mice (fig. S2A
   and [84]Fig. 3, C to E). Notably, Zbtb21^−/− mice exhibited normal
   learning and memory (fig. S2, B to D), suggesting that the effect of
   ZBTB21 on cognitive performance is more likely to be exclusively
   negative.

Fig. 3. Normalizing the Zbtb21 copy number restores synaptic and cognitive
function in Dp16 mice.

   [85]Fig. 3.
   [86]Open in a new tab

   (A) Mouse breeding strategy schematic. (B) Quantitative reverse
   transcription polymerase chain reaction (qRT-PCR) analysis of Zbtb21
   mRNA in hippocampi of 6-month-old WT, Zbtb21^+/−, Dp16, and
   Dp16;Zbtb21^+/− mice (n = 6 each). (C) Time to target platform in MWM
   test for 6-month-old WT (n = 12 mice), Zbtb21^+/− (n = 12 mice), Dp16
   (n = 12 mice), and Dp16;Zbtb21^+/− (n = 11 mice) mice. “*” on day 4:
   difference in time to reach platform, WT versus Dp16. (D) Time in the
   target quadrant during MWM probe test. (E) Percentage of freezing in
   fear conditioning tests. WT (n = 15 mice), Zbtb21^+/− (n = 13 mice),
   Dp16 (n = 14 mice), and Dp16;Zbtb21^+/− (n = 14 mice). (F and G) LTP
   recording in hippocampal CA1 region (n = 6 mice, 11 to 13 slices). (H)
   Input-output curve of basal synaptic transmission in hippocampal CA1 (n
   = 6 mice, 17 to 18 slices). (I and J) Spontaneous excitatory
   postsynaptic currents (sEPSC) and spontaneous inhibitory postsynaptic
   current (sIPSC) recordings in CA1 pyramidal neurons (n = 27 to 29 cells
   from four mice each). (K) Excitatory frequency/inhibitory frequency
   ratio in CA1 pyramidal neurons. (L and M) Golgi staining analysis of
   dendritic spine density in hippocampal CA1 (n = 57 to 62 dendrites from
   four mice each). Scale bar, 5 μm. (N and O) Electron microscopy
   analysis of synaptic structure and postsynaptic density (PSD) in
   hippocampal CA1 (n = 36 to 46 fields from three mice each). All data
   represent mean ± SEM. P values were determined by one-way analysis of
   variance (ANOVA) with Tukey’s post hoc analysis. *P < 0.05, **P < 0.01,
   ***P < 0.001, and ****P < 0.0001.

   Given the key role of CREB in regulating synaptic plasticity, we
   performed electrophysiological recordings in the hippocampal CA1 region
   and found that Dp16 mice showed compromised LTP, and synaptic deficits
   were corrected in Dp16;Zbtb21^+/− mice ([87]Fig. 3, F and G). In
   addition, we noted that Zbtb21 haploinsufficiency in Dp16;Zbtb21^+/−
   mice corrected the abnormalities in basal synaptic transmission
   ([88]Fig. 3H). Moreover, normalizing the Zbtb21 copy number restored
   the frequency of spontaneous excitatory postsynaptic currents (sEPSCs)
   ([89]Fig. 3I) but not the frequency of spontaneous inhibitory
   postsynaptic current (sIPSCs) ([90]Fig. 3J) in the hippocampal CA1
   pyramidal neurons of Dp16 mice. These findings collectively suggest
   that the disruption in the balance between excitation and inhibition
   ([91]Fig. 3K) observed in the hippocampus of Dp16 mice can be
   ameliorated by Zbtb21 haploinsufficiency, leading to the restoration of
   synaptic and cognitive function in Dp16;Zbtb21^+/− mice.

   Consistent with the electrophysiological results, we found that
   dendritic spine density was also restored in the hippocampus of
   Dp16;Zbtb21^+/− mice, as revealed by Golgi staining ([92]Fig. 3, L and
   M) and electron microscopy ([93]Fig. 3, N and O, and fig. S2E). Our
   findings indicate that there are no significant differences in the
   lengths of embryos across the three genotypes, and morphologically, the
   embryos did not exhibit any significant disparities. Furthermore, we
   extended our investigation to assess the brain morphology of these
   mice. Our analysis revealed no notable differences in cortical
   thickness or the area of the hippocampus among Zbtb21^+/+, Zbtb21^+/−,
   and Zbtb21^−/− mice. In addition, our experiments predominantly
   involved Zbtb21^+/− breeding with Dp16 mice, including behavioral and
   electrophysiological analyses. Our results show no significant
   differences between the Zbtb21^+/− mice and their WT littermates. This
   suggests that ZBTB21 haploinsufficiency has minimal impact on mouse
   behavior (fig. S2, F and G) and development (fig. S2, H and I) and
   demonstrated intact synaptic function (fig. S2, B to D). Together,
   these results suggest that ZBTB21 overdosage compromises synaptic
   function in the context of DS.

Neuron-specific ZBTB21 overexpression results in synaptic and cognitive
deficits

   We further investigated the expression of ZBTB21 in various brain
   tissue cells of Dp16 mice. Compared to WT mice, ZBTB21 expression was
   higher in the Dp16 mice across these cell types. Notably, neurons
   exhibiting trisomy displayed the most pronounced increase in
   transcriptional levels (fig. S3, A to D). Furthermore, we analyzed
   ZBTB21 expression in the brains of individuals with DS, using data from
   a previously published study ([94]34). The findings revealed that
   ZBTB21 expression in the brain tissue of DS individuals was markedly
   elevated in excitatory neurons when compared to the control group (fig.
   S3E). We further investigated the expression of Zbtb21 in different
   cell types in the brain of WT mouse, finding it to be abundantly
   expressed in neurons and astrocytes (fig.S3F). To determine whether
   ZBTB21 overexpression in the nervous system leads to DS-like
   neuropathology, we generated Nestin-Cre;Zbtb21^LSL-Tg conditional
   transgenic mice (Nestin-Zbtb21) by breeding Nestin-Cre mice with mice
   carrying the ROSA26-LSL-Zbtb21 allele (Zbtb21^LSL-Tg) ([95]Fig. 4A),
   and quantitative reverse transcription polymerase chain reaction
   (qRT-PCR) confirmed the increased expression of ZBTB21 in the
   hippocampus and in diverse cell types derived from Nestin-Zbtb21
   transgenic mice (fig. S4, A and B). We examined mouse cognitive
   performance and found that, compared with Nestin-Cre mice,
   Nestin-Zbtb21 transgenic mice displayed impaired hippocampus-dependent
   learning and memory in the Morris water maze (MWM), novel object
   recognition (NOR), and fear conditioning (FC) tests ([96]Fig. 4, B to
   D, and fig. S4C). Moreover, defects in hippocampal LTP ([97]Fig. 4, E
   and F) and basal synaptic transmission ([98]Fig. 4G) were also observed
   in Nestin-Zbtb21 mice.

Fig. 4. ZBTB21 overexpression in Nestin-Zbtb1 transgenic mice phenocopies
synaptic and cognitive deficits in DS mice.

   [99]Fig. 4.
   [100]Open in a new tab

   (A) Generation of conditional Zbtb21 transgenic mice by crossing
   CAG-Loxp-Stop-Loxp-Zbtb21 (CAG-Zbtb21) mice with Nestin-Cre or
   Camk2a-Cre mice. (B) Time taken to reach the submerged target platform
   during the training phase of MWM test. (C) Time spent in the target
   quadrant during MWM probe test. (D) Percentage of freezing time in fear
   conditioning tests. Four-month-old Nestin-Cre mice (n = 13 mice) and
   Nestin-Zbtb21 mice (n = 10 mice) were used in behavioral tests. (E) LTP
   recordings from the hippocampal CA1 region of 4-month-old Nestin-Cre (n
   = 6 mice, 14 slices) and Nestin-Zbtb21 mice (n = 6 mice, 11 slices).
   (F) fEPSP amplitude of the last 10 min of LTP recording in (E). (G)
   Input-output curve of basal synaptic transmission in the hippocampal
   CA1 region of 4-month-old Nestin-Cre (n = 6 mice, 13 slices) and
   Nestin-Zbtb21 mice (n = 6 mice, 11 slices). (H and I) Golgi staining
   analysis of dendritic spine density in the hippocampal CA1 region of
   mice. Representative images (H) and quantification (I) of dendritic
   spines in 4-month-old Nestin-Cre and Nestin-Zbtb21 mice. Scale bar, 5
   μm. n = 4 mice. (J and K) Electron microscopy analysis of synaptic
   structure and PSD in hippocampal CA1 region. n = 60 fields of view from
   three mice per group were counted. All data represent mean ± SEM. P
   values were determined by two-tailed Student’s t tests. *P < 0.05, **P
   < 0.01, and ****P < 0.0001.

   Furthermore, we performed Golgi staining and found a marked reduction
   in spine density in the hippocampus of Nestin-Zbtb21 mice compared to
   Nestin-Cre mice ([101]Fig. 4, H and I). Similar results were also
   observed in electron microscopy analysis ([102]Fig. 4, J and K). Given
   that Nestin-Cre expresses Cre within neural progenitors and affects
   multiple cell types in the central nervous system (CNS) ([103]35,
   [104]36), to further characterize the contribution of neurons and
   astrocytes to ZBTB21-induced synaptic deficits, we generated
   Camk2a-Cre;Zbtb21^LSL-Tg (Camk2a-Zbtb21) and
   Aldh1l1-ert2-Cre;Zbtb21^LSL-Tg (Aldh1l1-ert2-Zbtb21) transgenic mice.

   We first performed behavioral tests and found that neuron-specific
   overexpression of ZBTB21 markedly impaired cognitive performance in
   Camk2a-Zbtb21 mice. These mice displayed significant deficits in
   hippocampus-dependent learning and memory in the MWM and FC tests.
   However, the impairment was not significant in the NOR test. This
   suggests that the cognitive deficits in Nestin-Zbtb21 mice, which were
   more severe, might be due to a broader expression pattern of ZBTB21
   ([105]Fig. 5, A to C, and fig. S4, D to F). Consistent with the changes
   in behavioral performance, both synaptic function and density were
   compromised in the hippocampus of Camk2a-Zbtb21 mice ([106]Fig. 5, D to
   H). The Aldh1l1-ert2-Zbtb21 mice demonstrated trends toward changes,
   but these were not statistically significant in the MWM test (fig. S5,
   A to E) and LTP recording (figs. S5, F to H). Together, these results
   suggest that neurons represent the major cell type affected by ZBTB21
   overexpression in the DS brain.

Fig. 5. Neuronal ZBTB21 overexpression compromises synaptic and cognitive
function in Camk2a-Zbtb21 transgenic mice.

   [107]Fig. 5.
   [108]Open in a new tab

   (A) Time taken to reach the submerged target platform during the
   training phase of the MWM test. (B) Time spent in the target quadrant
   in the MWM probe test; 4-month-old Camk2a-Cre (n = 20 mice) and
   Camk2a-Zbtb21 mice (n = 14 mice) were used. (C) Percentage of freezing
   time in the fear conditioning tests. Four-month-old Camk2a-Cre (n = 20
   mice) and Camk2a-Zbtb21 mice (n = 14 mice) were used. (D) LTP
   recordings from the hippocampal CA1 region of 4-month-old Camk2a-Cre (n
   = 6 mice, 11 slices) and Camk2a-Zbtb21 mice (n = 6 mice, 11 slices).
   (E) fEPSP amplitude of the last 10 min of LTP recording in (D). (F)
   Input-output curve reflecting basal synaptic transmission in the
   hippocampal CA1 region of 4-month-old Camk2a-Cre (n = 6 mice, 13
   slices) and Camk2a-Zbtb21 mice (n = 6 mice, 12 slices). (G and H) Golgi
   staining (G) and quantification (H) of dendritic spines in the
   hippocampal CA1 region of 4-month-old Camk2a-Cre and Camk2a-Zbtb21
   mice. Scale bar, 5 μm. n = 4 mice. All data represent mean ± SEM. P
   values were determined by two-tailed Student’s t tests. *P < 0.05, **P
   < 0.01, and ***P < 0.001.

ZBTB21 binds to CREs in the promoter regions of genes in neurons

   To elucidate the mechanism by which ZBTB21 contributes to synaptic and
   cognitive impairments, we infected mouse primary neurons with
   lenti-Zbtb21 and analyzed the genome-wide DNA binding profile of Zbtb21
   using CUT&Tag-seq (fig. S1H).

   We identified approximately 4700 potential ZBTB21 target sites in the
   promoter region (defined as ±3 kb from the TSS), using a significance
   threshold of P value < 0.05 and |log[2]fold change| > 1. Notably, among
   the top consensus ZBTB21-binding motifs, three were CRE motifs
   ([109]Fig. 6A). The analysis further revealed that the most significant
   enrichment of ZBTB21 binding occurred within promoter regions, introns,
   and distal intergenic regions ([110]Fig. 6B). Specifically, the genes
   Dcdc2a and Impact, which are known to be involved in learning and
   memory ([111]37–[112]39), were used as examples to demonstrate the
   highly abundant ZBTB21 binding to the CRE motif in their promoters.
   Furthermore, CUT&Tag-seq revealed an enrichment of CRE binding
   ([113]Fig. 6C). Similarly, chromatin immunoprecipitation (ChIP)
   analysis using an anti–hemagglutinin (HA) antibody in
   HA-Zbtb21–overexpressing mouse primary neurons detected a high level of
   ZBTB21 binding to the CRE in the promoter of Crtcl ([114]40, [115]41),
   another learning and memory-related gene ([116]Fig. 6D). Together, our
   data indicate that ZBTB21 suppresses CRE-mediated transcription through
   direct binding with the canonical CRE in the promoter region of target
   genes.

Fig. 6. ZBTB21-induced inhibition of CRE-dependent gene expression is
conferred by its binding to the CRE element.

   [117]Fig. 6.
   [118]Open in a new tab

   (A) Homer software for different peak motif analyses (P value < 0.05,
   |log[2]fold change| > 1). The top six enriched motifs are shown. (B)
   Peak distributions relative to gene features. The majority of the peaks
   fell into promoters and intergenic regions. (C) Primary mouse neurons
   were infected with lenti-HA-Zbtb21, followed by chromatin CUT&TAG-seq
   analysis using an anti-HA antibody. The sequencing results were used to
   quantify the abundance of the CRE motif enriched by HA-ZBTB21. The
   distribution of ZBTB21 binding at the promoter region of Dcdc2a and
   Impact are shown. Red boxes indicate the regions that are enriched for
   ZBTB21 binding. (D) Primary mouse neurons were infected with
   lenti-HA-Zbtb21, followed by ChIP analysis using an anti-HA antibody.
   The sequencing results were used to quantify the abundance of the CRE
   motif enriched by HA-ZBTB1. The distribution of ZBTB21 binding at the
   promoter region of Crtc1 is shown. The red box indicates the region
   that shows enrichment for ZBTB21 binding. Enrichment of ZBTB21 binding
   in the promoter region of Crtc1 in the box was confirmed by qPCR and is
   shown in the right. All data represent mean ± SEM. P values were
   determined by two-tailed Student’s t tests. ****P < 0.0001. 5′UTR, 5′
   untranslated region; IgG, immunoglobulin G.

Restoration of ZBTB21 expression rescues synaptic function by correcting
transcriptional defects in Dp16 mouse neurons

   To gain further insight into the roles of ZBTB21 in DS pathogenesis by
   regulating CRE-mediated transcription, we performed whole-genome RNA
   sequencing (RNA-seq) on primary neurons from DS mice (WT, Dp16, and
   Dp16; Zbtb21^+/−). Our findings revealed that 420 genes, which were
   down-regulated in Dp16 mouse neurons [differentially expressed genes
   (DEGs) with a fold change > 1.2 and a P value < 0.05], showed restored
   expression levels in Dp16;Zbtb21^+/− neurons ([119]Fig. 7A). GO
   enrichment analysis of these genes revealed their involvement in
   pathways related to cAMP-mediated signaling, learning and memory, and
   synaptic function ([120]Fig. 7B). Notably, two-thirds of these 420 DEGs
   contained CRE motifs ([121]Fig. 7C). Within these, 278 CRE-containing
   genes could be further categorized into four functional groups
   ([122]Fig. 7D), with learning/memory and synapse-related pathways being
   significant categories. This suggests that CRE-containing genes are
   potential targets of ZBTB21 regulation, with a third of the non-CRE
   DEGs possibly indirectly regulated by ZBTB21. We further reanalyzed the
   data from a previous study ([123]34) to investigate the competitive
   relationship between ZBTB21 and CREB by examining the correlation
   between the expression levels of specific CRE motif–containing genes
   and ZBTB21 in human DS brain tissues. Our analysis revealed a negative
   correlation between the expression of protocadherin 9, calcium
   dependent secretion activator 2, astrotactin 2 (which are important for
   synaptic function and cognition) and ZBTB21 expression (fig. S6A).
   These findings support the mechanism by which ZBTB21 competitively
   suppresses CREB-mediated transcription. A heatmap of the CRE-containing
   DEGs categorized as synapse-related genes is presented in [124]Fig. 7E.

Fig. 7. Normalizing the Zbtb21 copy number rescues synaptic function by
correcting transcriptional defects in Dp16 mouse neurons.

   [125]Fig. 7.
   [126]Open in a new tab

   (A) Venn diagram depicting up-regulated genes in Dp16;Zbtb21^+/− mice
   relative to Dp16 mice and down-regulated genes in Dp16 mice relative to
   WT mice. The 420 genes in the overlapping region indicate
   down-regulated genes in Dp16 restored by Zbtb21 normalization. (B) GO
   analysis of the overlapping genes in (A). The top enriched biological
   process terms are displayed. The size of a circle indicates the number
   of enriched genes, and the color reflects the adjusted P value. (C) The
   outside circle represents the 420 overlapping genes in (A), and the
   inside circle shows the 278 genes containing CRE motifs among these 420
   DEGs. (D) The central circle represents the 278 CRE-containing DEGs in
   (C), and the petals show the number of CRE-containing DEGs classified
   into different categories. (E) Heatmap depicting the 42 synapse-related
   genes identified in (D). (F) Venn diagram depicting the overlapping CRE
   motif–containing genes from RNA-seq and CUT&TAG-seq. (G) GO analysis of
   the overlapping CRE-containing genes in (F). The top enriched
   biological process terms are displayed. GO enrichment analysis was
   performed under the filtering condition of a q value of 0.05.

   We further investigated the overlap between DEGs from CUT&Tag-seq and
   RNA-seq ([127]Fig. 7F and fig. S6, B and C). We found 122 genes that
   were common to both datasets, 79 of which contained CRE motifs. GO
   enrichment analysis revealed that these overlapping genes were involved
   in pathways related to cAMP-mediated signaling, learning and memory,
   and synaptic function ([128]Fig. 7G).

   Together, these results demonstrate that the restoration of neuronal
   ZBTB21 expression to normal levels in Dp16 mice rescues synaptic and
   cognitive function in this model of DS, at least in part by restoring
   CRE-dependent transcription. The concordant restoration of
   CRE-dependent transcription provides evidence that reducing ZBTB21 acts
   through this mechanism to ameliorate DS-related phenotypes.

DISCUSSION

   While the CREB family of proteins is well-recognized for its pivotal
   role in regulating synaptic function, learning and memory, the
   pathological implications and regulatory mechanism of CRE-dependent
   gene expression remain largely unexplored. This study uncovers that
   overdosage of the chromosome 21–encoded gene ZBTB21 contributes to the
   dysregulation of CRE-dependent gene expression and synaptic dysfunction
   in the DS brain. Our findings suggest that the detrimental impact of
   ZBTB21 overdosage on CRE-dependent gene expression in the DS brain is a
   key contributing factor to synaptic and cognitive deficits (fig. S7).
   This finding is further supported by a previous study that demonstrated
   a close link between human segmental trisomy containing the ZBTB21
   locus and intelligence quotient levels in individuals with DS
   ([129]42).

   Previous research has reported impaired cAMP production in the
   hippocampus of a murine DS model ([130]26). Rolipram, a
   phosphodiesterase inhibitor, has been shown to ameliorate behavioral
   abnormalities in mice with cognitive impairment by elevating neuronal
   cAMP levels ([131]43–[132]45). In addition, gonadotropin-releasing
   hormone (GnRH) therapy, which enhances cAMP levels, has demonstrated
   beneficial effects in individuals with DS and DS model mice
   ([133]46–[134]48). Overcoming the impairment of the cAMP-CREB pathway
   could potentially benefit individuals with DS. Given that ZBTB21
   overdosage suppresses CRE-dependent gene expression, increasing cAMP
   levels may be a potential approach to compensate for the inhibition of
   CRE-dependent gene expression induced by trisomy 21. RCAN1, a DSCR gene
   that plays a crucial role in learning and memory, is known to be
   negatively regulated by proteasomal degradation in a CREB
   transcriptional activation-dependent manner ([135]49, [136]50). The
   suppression of CRE-dependent gene expression by ZBTB21 could form a
   feedforward loop with RCAN1, exacerbating DS-associated cognitive
   deficits.

   The ZBTB21 protein is expressed in multiple cell types within the CNS.
   While we have demonstrated that neuronal ZBTB21 is important for
   learning and memory, its roles in other cell types remain to be further
   characterized. We observed that the defects in Nestin-Zbtb21 mice were
   more severe than those in Camk2a-Zbtb21 mice, suggesting that multiple
   cell types in the CNS may contribute to neurological phenotypes in
   Zbtb21 transgenic mice. Moreover, Nestin-Cre–mediated targeting of
   progenitor cells could result in ZBTB21 overexpression at the embryonic
   stage, which might also explain the stronger phenotype observed in
   Nestin-Zbtb21 mice. Future studies analyzing the effects of conditional
   Zbtb21 deletion in various CNS cell types will be necessary to
   characterize the cell type–specific effects of ZBTB21 perturbation in
   the CNS.

   The potential benefits of rolipram or GnRH treatment support the
   restoration of cAMP levels as a possible strategy to treat DS-related
   cognitive deficits. Considering the critical role of ZBTB21 in
   suppressing CRE-dependent gene expression, further investigation into
   the function of ZBTB21 may provide insights into therapeutic strategies
   aimed at preventing or treating DS-related cognitive deficits by
   restoring CRE-dependent gene expression.

MATERIALS AND METHODS

Experimental animals

   Dp16 mice (stock no. 013530), Nestin-Cre mice (stock no. 003771), and
   Camk2a-Cre mice (stock no. 005359) were obtained from the Jackson
   Laboratory (Bar Harbor, Maine, USA). Aldh1l1-ert2-Cre mice were
   generated as described previously by Hu et al. ([137]51). Zbtb21^−/−
   mice were obtained from genetically modified fertilized eggs by
   injecting two strips of single guide RNA targeting the largest exon
   region of Zbtb21 and Cas9 mRNA into fertilized mouse eggs.
   Rosa26-Zbtb21 mice were generated using CRISPR-CAS9 technology. A
   plasmid (Rosa26-CAG-floxp-stop-floxp-Zbtb21) containing the Zbtb21 gene
   and its regulatory elements was integrated into the mouse Rosa26 locus
   through homologous recombination. This plasmid uses the LoxP-Stop-LoxP
   (LSL) cassette, which enables conditional activation of gene expression
   in mammalian cells and animals through Cre-mediated recombination.
   After Cre-mediated recombination, the LSL cassette was excised,
   allowing for the transcription and expression of the gene of
   interest.Camk2a-Zbtb21 mice were generated by crossing Rosa26-Zbtb21
   mice with Camk2a-Cre mice, resulting in neuron-specific overexpression
   of ZBTB21 driven by a ubiquitous CAG promoter. A stop codon cassette
   was flanked by two LoxP sites. The STOP codon was expressed before
   Cre-mediated recombination. After Cre-mediated recombination, the STOP
   codon cassette was excised, and the Zbtb21 gene was expressed instead
   of the STOP codon.Nestin-Zbtb21 mice were generated by breeding
   Rosa26-Zbtb21 mice with Nestin-Cre mice, resulting in the
   overexpression of ZBTB21 specifically in CNS precursor cells. The
   breeding strategy used for the generation of Nestin-Zbtb21 mice was the
   same as that used for Camk2a-Zbtb21 mice.Aldh1l1-ert2-Zbtb21 mice were
   generated by crossing Rosa26-Zbtb21 mice with Aldh1l1-ert2-Cre mice,
   resulting in astrocyte-specific overexpression of ZBTB21. The breeding
   strategy used for the generation of the Aldh1l1-ert2-Zbtb21 mice was
   identical to that used for the Camk2a-Zbtb21 mice.

   All the experimental mice used in our study were bred and maintained on
   a C57BL/6J background. Specifically, only male mice were used for the
   animal experiments mentioned in our work, which were performed
   according to the guidelines of the Institutional Animal Care and Use
   Committee of Xiamen University. The primers used for genotyping are
   shown in table S1.

Cell culture

   Primary neurons were grown and maintained in neurobasal medium (Thermo
   Fisher Scientific, 21103-049) supplemented with B27 (Thermo Fisher
   Scientific, 17504044) and 1% penicillin/streptomycin. For primary
   astrocyte cultures, both hemicortices of mice on postnatal days 1 to 2
   were dissected and chopped into small pieces. The tissue was dispersed
   by pipetting. After centrifugation, the cell pellet was resuspended in
   Dulbecco’s modified Eagle’s medium (DMEM)/F12 culture medium (Thermo
   Fisher Scientific, 10565-018) containing 10% fetal bovine serum (FBS)
   and 1% penicillin/streptomycin. The cells were seeded in culture
   flasks, and the culture medium was changed by half every 3 days. After
   7 days, the mixed culture was mechanically agitated at 180 rpm for 1
   hour. The astrocytes were then dissociated using trypsin-EDTA and
   plated in 60-mm plates for RNA extraction. HEK293T cells were cultured
   in DMEM (Gibco, C11995500BT) supplemented with 10% FBS and 1%
   penicillin/streptomycin. For primary microglial cultures, both
   hemicortices of mice on postnatal days 1 to 2 were dissected and
   chopped into small pieces. The tissue was dispersed by pipetting. After
   centrifugation, the cell pellet was resuspended in DMEM culture medium
   (Gibco, C11995500BT) containing 10% FBS and 1% penicillin/streptomycin.
   The cells were then seeded in culture flasks. Granulocyte-macrophage
   colony-stimulating factor (25 ng ml^−1; R&D Systems, Minneapolis, MN,
   USA; 415-ML-050) was added to the cultures after 3 days. Primary
   microglia were harvested by shaking the flasks at 180 rpm for 1 hour,
   10 to 12 days after plating, and once every 3 days thereafter for RNA
   extraction.

Drug administration

   Tamoxifen (MedChemExpress, HY-13757A) was used to induce
   Cre^ERT2-mediated recombination. Tamoxifen was dissolved in 90% corn
   oil (Sigma-Aldrich, C8267). Adult P56-60 mice were subjected to daily
   intraperitoneal injections of tamoxifen (80 mg/kg) for seven
   consecutive days, as previously described ([138]52). 4-Hydroxy
   tamoxifen (Sigma-Aldrich, H7904-5MG) was used to induce
   CreERT2-mediated recombination in primary cell, the final concentration
   of 4-hydroxy tamoxifen is 5 μM, and incubated for 6 days.

Fear conditioning test

   The experimental mice were placed in a conditioning chamber (Noldus,
   Ugo Basile) and allowed to freely explore for a period of 2 min.
   Following the exploration phase, a 30-s voice prompt was presented as
   the conditioned stimulus (CS). During the last 2 s of the voice prompt,
   the mice received 0.8-mA foot electrical stimulation as an
   unconditioned stimulus (US). The voice prompt accompanied by the
   electric stimulus (CS-US) was repeated every minute for the subsequent
   3 min. For the contextual test, the mice were reintroduced to the same
   chamber 24 hours later and allowed to explore the familiar environment
   for 5 min without any stimulation. For the cued test, the mice were
   placed in a new test room with a different contextual environment for 2
   min, followed by 3 min of voice prompts.

NOR test

   The NOR test was conducted as previously described ([139]52). On day 1,
   each mouse was acclimated to the test chamber for 5 min. On day 2, two
   objects of identical shape and size were positioned diagonally within
   the home cage. Subsequently, the mice were placed in the center of the
   cage and allowed to freely explore the open field environment and
   objects for 10 min to become familiar with them. On day 3, one of the
   objects remained unchanged, while the other object was replaced with a
   block that closely resembled a different shape. The mice were then
   placed in the center of the cage again and allowed to explore freely
   for 10 min. The time spent by the mice investigating the old object and
   the new object was recorded.

MWM test

   The experiment was conducted in a quiet indoor circular pool with a
   diameter of 120 cm, maintaining a constant water temperature of 22° ±
   1°C. A fixed platform (10 cm in diameter) was placed in a preset target
   quadrant and submerged approximately 1 cm beneath the water surface.
   The pool was adorned with black geometric symbols on the diagonal
   walls, which served as cues for the mice to find the platform. During
   the training phase, the mice were released from one of six randomized
   positions, and two trials were performed every day for four to five
   consecutive days. Each mouse was given 60 s to find the hidden
   platform, and if a mouse failed to find the platform within 60 s, then
   it was guided to the platform by the experimenter and left there for 30
   s. During the test phase, the platform was removed, and the mice were
   placed in the water from the predetermined test point. The time taken
   by the mice to reach the platform and the percentage of time spent in
   the quadrants were recorded using a Noldus EthoVision system.

Preparation of mouse brain sections

   After anesthesia, the mouse brain was removed and then kept in
   precooled oxygen-saturated (95% O[2] and 5% CO[2]) high-glucose buffer
   [64 mM NaCl, 2.5 mM KCl, 1.25 mM NaH[2]PO[4], 10 mM MgSO[4], 26 mM
   NaHCO[3], 10 mM d-glucose, 120 mM sucrose, and 0.5 mM CaCl[2] (pH 7.4);
   osmotic pressure, 290 to 320 milliosmol (mOsm)]. The 400-μm-thick
   slices were obtained using a vibrator (VT1200S, Leica, Germany). Once
   the brain sections were acquired, the slices were promptly transferred
   to oxygen-saturated artificial cerebrospinal fluid [ACSF; 120 mM NaCl,
   3.5 mM KCl, 1.25 mM NaH[2]PO[4], 1.3 mM MgSO[4], 26 mM NaHCO[3], 10 mM
   d-glucose, and 2.5 mM CaCl[2] (pH 7.4); osmotic pressure, 290 to 320
   mOsm] at a constant temperature of 32°C and incubated for 1 hour,
   followed by an additional 1-hour incubation at room temperature. For
   electrophysiological recording, the slices were transferred to a
   recording chamber, where they were continuously perfused with
   oxygen-saturated (95% O[2] and 5% CO[2]) standard ACSF at 32°C.

Electrophysiological recording

   Electrophysiological recording was performed using a MultiClamp 700B
   amplifier (MD/Axon US). The acquired data were subsequently analyzed
   using pClamp software (version 10.6, Molecular Devices).

Input-output recording

   The field excitatory postsynaptic potential (fEPSP) was evoked by the
   stimulation electrode as mentioned. Once the amplitude of the fEPSP
   reached a stable level and remained consistent for a minimum of 10 min,
   a series of electrical stimulations was applied, gradually increasing
   in intensity. The intensities used were 100, 200, 300, 400, 500, 600,
   700, and 800 μA. The amplitude of the fEPSP corresponding to each
   stimulus intensity was recorded. This collection of data resulted in an
   input-output (I-O) curve, reflecting the response capacity of the
   synapses under different stimulus intensities.

LTP recording

   Responses were elicited by electrically stimulating the Schaffer
   collateral pathway using a concentric electrode. fEPSPs were recorded
   in the hippocampal CA1 region using a patch-type pipette filled with
   ACSF (Re = 2 to 3 megohm). The baseline stimulus intensity was adjusted
   to evoke fEPSPs with amplitudes at approximately 30% of the maximum
   response obtained from the input-output function. The interval between
   each electrical stimulation was set to 20 s, and the electrode
   resistance ranged from 1 to 3 megohm. Stable baseline recordings were
   obtained for at least 20 min before the high-frequency stimulation
   (HFS) used to induce hippocampal LTP. The HFS consisted of two trains
   of stimuli delivered at 100 Hz, with an interval of 30 s. Following the
   HFS, a 60-min recording of the baseline protocol was performed.

Golgi staining

   Golgi staining was performed using an FD Rapid Golgi Stain Kit (FD
   Neuro Technologies) following the manufacturer’s instructions.
   Subsequently, Z-stack images were acquired with a Leica SP8 confocal
   microscope, and the density of the dendritic spines was analyzed using
   National Institutes of Health ImageJ software.

CUT&Tag assay

   The CUT&Tag assay was performed using the Hyperactive Universal CUT&Tag
   Assay Kit for Illumina (Vazyme) according to the manufacturer’s
   instructions. Briefly, the membranes of 2 × 10^5 primary neurons were
   permeabilized with concanavalin A–coated magnetic beads and a non-ionic
   detergent (digitonin). Subsequently, cells were incubated with primary
   antibodies, either H3K27ac (1:500; Cell Signaling Technology, catalog
   no. 8173) or anti-HA (1:100; Santa Cruz Biotechnology, sc-7392 X), at
   4°C overnight. Through the binding of the primary antibody to the
   target protein, corresponding secondary antibody, and Protein A/G, Tn5
   transposases fused with Protein A/G are precisely targeted to cut the
   DNA sequence near the target protein. During the DNA cleavage by the
   transposase, adapter sequences are added to both ends of the cut
   fragment, which can be directly amplified by PCR to form a library for
   high-throughput sequencing. Libraries were constructed using the TD202
   TruePrep Index Kit V2 for Illumina (Vazyme). Sequencing was performed
   using Illumina NovaSeq 6000 system. Peak calling was performed using
   sparse enrichment analysis, and subsequent data analysis was conducted
   using HOMER, a software suite for CUT&TAG analysis.

Quantitative MS

   Biotin-conjugated double-stranded DNA was prepared and introduced into
   HEK293T cells as previously described ([140]53). Subsequently, the
   cells were stimulated to facilitate the binding of transcription
   factors to the DNA. Streptavidin agarose was used to precipitate the
   DNA. The binding proteins were denatured from the eluted DNA using SDS.
   The denatured proteins were then processed and subjected to MS analysis
   on a TripleTOF 5600 (Sciex) mass spectrometer coupled to a NanoLC Ultra
   2D Plus (Eksigent) high-performance liquid chromatography system.

RNA isolation and quantitative reverse transcription PCR

   Total RNA was extracted using the RNAiso Plus reagent (TaKaRa, Japan;
   D9109), and 3 μg of total RNA was reverse-transcribed into cDNA using
   M-MLV Reverse Transcriptase (RNase H^+) (The Beijing Genomics
   Institute, China; EM-004). Real-time PCR was performed on a LightCycler
   480 System (Roche, Mannheim, Germany) using ChamQ Universal SYBR qPCR
   Master Mix (Vazyme, China; Q711-02). Data analysis was performed using
   the 2-^△△Ct method after normalization to the ActB internal control.
   The primers used in RT-PCR are shown in table S1.

RNA sequencing

   RNA-seq was performed using the Illumina NovaSeq 6000 system. The raw
   RNA-seq reads were aligned to the Ensembl mouse reference genome
   GRCm39. The raw gene counts were processed using the DESeq2 package
   (version 1.20.0) clusterProfiler (version 3.8.1, pvalueCutoff = 0.05).
   GO enrichment analysis of DEGs was performed in R using the package
   clusterProfiler (version 3.8.1, qvalueCutoff = 0.05).

DNA pull-down assays

Intracellular assay

   Biotin-labeled double-stranded DNA (Random: CAGGTATC; CRE1: TGACGTCA;
   CRE2: ATGACGTA; AP-1: TGACTCA) was transfected into HEK293T cells.
   After 24 hours, the cells were lysed in lysis buffer [recipe of lysis
   buffer can be found in ([141]54)]. The lysates were incubated on ice
   for 30 min and then centrifuged at 20,000g for 30 min at 4°C. The
   supernatant was precipitated with NeutrAvidin Agarose Resins (Thermo
   Fisher Scientific) at 4°C for 30 min. The resins were then washed four
   times with lysis buffer, and the DNA binding proteins were denatured
   and eluted using 2% SDS. The eluted proteins were prepared using a
   standard protocol for MS or Western blot analysis.

Extracellular assay

   The ZBTB21 overexpression plasmid was transfected into HEK293T cells.
   After 24 to 36 hours, the HEK293T cells were lysed in lysis buffer. The
   lysates were incubated on ice for 30 min and then centrifuged at
   20,000g for 30 min at 4°C. Biotin-conjugated double-stranded DNA was
   added to the supernatant and rotated for 2 hours. NeutrAvidin Agarose
   Resins (Thermo Fisher Scientific) were then added to the supernatant
   and rotated for 30 min. The resins were washed four times with lysis
   buffer, and the DNA binding proteins were denatured and eluted using 2×
   sample buffer. The eluted proteins were prepared using a standard
   protocol for Western blot analysis.

ZBTB21 KO cell line

   The targeting sequence in the guide RNA (gRNA) vector for human ZBTB21
   was 5′-catcgacaatgaccgtctg-3′. First, the gRNA with the Cas9 expression
   plasmid was cotransfected into WT HEK293T cells. The transfected cells
   were then incubated with blasticidin to facilitate selection. After
   selection, single-cell clones were isolated from the pool of
   transfected cells using a limiting dilution cloning technique in
   96-well plates. Each well contained only a single cell to ensure
   clonality. The isolated clones were subsequently screened for the KO
   genotype by Sanger sequencing.

Electrophoretic mobility shift assay

   EMSAs were performed as previously described ([142]55) using the
   LightShift Chemiluminescent EMSA Kit (Pierce, 20148). Briefly, protein
   extraction from overexpressed HEK293T cells was achieved using the
   repeated freeze-thaw method, followed by the standard EMSA protocol
   using the EMSA Kit.

Chromatin immunoprecipitation

   ChIP was conducted as previously described ([143]56). Briefly, primary
   neurons infected with lentivirus were fixed for 10 min at room
   temperature with phosphate-buffered saline (PBS) containing 1%
   formaldehyde (Thermo Fisher Scientific, 28908), and the reaction was
   stopped by adding 125 mM glycine for 5 min. Subsequently, the cells
   were washed twice with ice-cold PBS. Fixed cells were then collected by
   centrifugation at 12,000 rpm for 30 s and resuspended in lysis buffer.
   DNA fragmentation was achieved by sonication using a Bioruptor Plus
   (Diagenode), resulting in sheared DNA fragments ranging from 150 to 500
   base pairs. The sonicated cell lysate was diluted 10-fold in ChIP
   dilution buffer and incubated with HA antibody (Abcam, ab9110) and
   isotype immunoglobulin G (Sino Biological, CR1) for 4 to 8 hours at
   4°C, followed by a 2-hour incubation with magnetic beads coated with
   protein G. After thorough washing and elution, the protein-DNA
   crosslinking complex was decrosslinked in TE (pH8.0 1M Tris-HCL; 0.5M
   EDTA)-SDS buffer by overnight heating at 65°C. Immuno-precipitated DNA
   was purified using DNA recovery columns (Qiagen, 28104) and
   subsequently analyzed by qPCR using CFX96 (Bio-Rad).

Electron microscopy analysis

   Tissue collection and transmission electron microscopy (TEM) were
   performed as previously described ([144]57). Briefly, mice were deeply
   anesthetized and perfused with ice-cold PBS. The hippocampus was
   dissected and fixed with 2.5% glutaraldehyde at 4°C for 2.5 hours.
   Afterward, the tissue was washed three times with PBS. Fixed tissue
   processing and TEM were performed according to the standard electron
   microscopy protocol.

Lentiviral packaging

   Lentiviral packaging was conducted as previously described ([145]58).
   Briefly, HEK293T cells were transfected with expression vectors
   containing Zbtb21 and lentiviral packing plasmids (PMDL/REV/VSVG) using
   the calcium phosphate precipitation method. After 8 to 12 hours, the
   cell culture medium was replaced, and the virus-containing medium was
   collected 36 hours later.

cAMP concentration assay

   The cAMP concentration assay was performed following the protocol
   described previously ([146]59), using the cAMP-Glo Max Assay (V1681,
   Promega). The standard assay protocol was followed using the
   abovementioned cAMP levels assay kit.

PKA activity assay

   The PKA activity assay was performed using the PKA Colorimetric
   Activity Kit (EIAPKA, Invitrogen), following the standard assay
   protocol provided with the abovementioned reagent kit. Briefly, the
   293T cells in a 60-mm dish were lysed using 1 ml of lysis buffer, and
   the lysate was then diluted 20-fold for activity testing.

Immunoblot analysis

   Immunoblot analysis was performed as previously described ([147]53).
   Antibodies were used as follows: anti-ZBTB21 (1:500; Proteintech,
   20132-1-AP), anti-CREB (1:1000; Abclonal, A11989), anti-HA (1:5000;
   Biochee, A2022-02L), anti-ACTIN (1:5000; Affinity, T0022), and
   anti-TUBULIN (1:5000; Abmart, M20005M).

Nissl staining

   Nissl staining was performed following the protocol as described
   previously ([148]52). Briefly, frozen sections were stained with Nissl
   staining solution (Beyotime) for 10 min. The staining process was
   terminated with water, and the slides were decolorized using ethanol,
   followed by clearing with xylene for imaging observation.

Imaging Zbtb21 mRNA in brain tissue using hybridization chain reaction

   The hybridization chain reaction technique was adapted from previous
   work by Pierce et al. ([149]60), with minor modifications. Briefly,
   15-μm-thick paraformaldehyde-fixed brain tissue cryosections were
   prepared. Following washes with PBS, the sections were incubated in 70%
   ethanol at −20°C overnight. This was followed by two washes with 2×
   SSC, and the sections were prehybridized with 30% probe hybridization
   buffer at 37°C for 30 min. Subsequently, hybridization was performed
   overnight at 37°C with a final concentration of 4 nM detection probe in
   30% probe hybridization buffer. Afterward, the sections were washed
   with 30% probe wash buffer preheated to 37°C, followed by two washes
   with 5× SSC-Tween buffer. The sections were then preamplified for 30
   min using amplification buffer. The fluorescent probe was annealed and
   diluted in amplification buffer to a final concentration of 40 nM,
   followed by overnight incubation. Last, the sections were stained with
   Hoechst to visualize the nuclei, rinsed once, and mounted for
   observation. The detection probes were designed to target 23 pairs of
   target sites specific to mouse Zbtb21 mRNA. The B5 probe set was
   selected for amplification, and Alexa Fluor 568 fluorescence was used
   for labeling.

Quantification and statistical analysis

   GraphPad Prism 8.0 software was used to analyze the data. Statistical
   differences between the two groups were assessed using paired or
   unpaired t tests. Differences among multiple groups were assessed using
   one-way analysis of variance (ANOVA), followed by Bonferroni’s post hoc
   analysis. The statistical significance of the differences between
   groups was determined in accordance with the following criteria: not
   significant, P > 0.05; *P < 0.05; **P < 0.01; ***P < 0.001; and ****P <
   0.0001.

Study approval

   Animal experiments were reviewed, approved, and approved by the
   Laboratory Animal Management and Ethics Committee of Xiamen University
   (approval number XMULAC20210070) and were performed in strict
   accordance with good animal practices as defined by Xiamen University
   Laboratory Animal Center (Xiamen, China). The mice used in this study
   were housed under specific pathogen–free conditions with a 12-hour
   light/dark cycle and access to food and water ad libitum at the Xiamen
   University Laboratory Animal Center.

Acknowledgments