miR155, triplicated in Down syndrome, regulates the development of neural stem cells and GABAergic interneurons in Alzheimer's disease mouse and human iPSC models

Abstract

INTRODUCTION

Dysfunctional microRNAs and GABAergic interneurons are features of Alzheimer's disease (AD). The role of neuronal microRNA155 (miR155), elevated in both AD and Down syndrome (DS), remains unknown.

METHODS

We utilized in silico analyses of published databases, MIR155‐deleted and ‐overexpressing human induced pluripotent stem cell (hiPSC)‐derived cells, cortical organoids, and amyloid beta precursor protein (APP)/PS1‐miR155 knockout mouse.

RESULTS

MIR155HG (miR155 host gene) colocalizes with APP in a neuron‐specific, topologically associated domain (TAD) in chromosome 21. In human neural stem cells (NSCs), neurons, and cortical organoids, MIR155 deletion enhanced NSC proliferation and GABAergic interneuron generation. MIR155 overexpression inhibited NSC marker expression and GABAergic interneuron generation. In APP/PS1 mice, miR155 deletion induced the expansion of hippocampal NSCs and increased hippocampal GABAergic interneurons.

DISCUSSION

Our findings, alongside the extensive studies of the role of microglial miR155 in neuroinflammation, reveal previously unrecognized miR155 roles in hippocampal NSC dynamics and GABAergic interneuron development, highlighting miR155 as a therapeutic target.

Highlights

• MIR155HG is linked to APP, DS, and AD.

• MIR155 is a repressor of proliferation in NSCs.

• MIR155 is a negative regulator of GABAergic interneuron development.

• Mir155 positively regulates its mRNA targets NR1F1 and NR2F2 in human neurons.

• Mir155 deficiency and overexpression have a profound impact on neurons, in addition to its known role in microglia.

1. BACKGROUND

MicroRNAs (miRNAs) are short, 21‐ to 23‐nucleotide, single‐stranded, non‐coding RNA molecules. They bind most commonly to the 3'UTR (untranslated region) of their target mRNAs and repress protein production by destabilizing the mRNA, cleavage of the mRNA strand, and translational silencing. ¹ In mammalian brains, miRNA expression profiles are cell‐type‐specific, ² , ³ with distinct functions and roles across different neural subtypes, influencing both brain development and susceptibility to neurological disorders. ⁴ , ⁵ Mir155 is a multifunctional miRNA first described as a central player in the macrophage inflammatory response ⁶ and then extensively studied as a microglial master regulator in neuroinflammation in neurodegenerative conditions, including Alzheimer's disease (AD). ⁷

Recent studies, including some from our laboratory, demonstrated that miR155 is upregulated in the hippocampal neurons of AD patients and in mouse models of AD pathology. ⁸ , ⁹ , ¹⁰ Other studies have focused on the increase of miR155 as part of Trisomy 21 in Down syndrome (DS) patients ¹¹ showing impaired neurogenesis and synaptogenesis during human brain development ¹² that is also manifested in cortical neurons derived from DS patient human induced pluripotent stem cells (hiPSCs) ¹³ , ¹⁴ and in the Ts65Dn mouse model of DS. ¹⁵ The hippocampus plays an essential role in human cognition and memory and is severely affected in AD. ¹⁶ Hippocampal neurogenesis contributes to hippocampal plasticity through the integration of new neurons into existing circuits. ¹⁷ Hippocampal neurogenesis in rodents ¹⁷ involves radial glia‐like neural stem cells (RGL‐NSCs) in the subgranular zone (SGZ) of the hippocampal dentate gyrus (DG), which generate proliferating intermediate progenitor cells (IPCs) and neuroblasts that differentiate into dentate granule neurons. Markers of hippocampal neurogenesis decline sharply in AD. ¹⁸ , ¹⁹ Hippocampal GABA(gamma‐aminobutyric acid)ergic interneurons primarily originate from the ventral telencephalon, and their generation and specification are governed by transcription factors such as NR2F1/2. ²⁰ , ²¹ They represent 10% to 15% of the total neuronal population and serve as major determinants of virtually all aspects of cortical circuit function. ²⁰ Parvalbumin (PV⁺) and somatostatin (SST⁺) subtypes comprise ∼70% of inhibitory GABAergic interneurons. ²⁰ , ²¹ In AD, there are significant reductions in levels of GABA and somatostatin in the cerebrospinal fluid (CSF) and brain. ²² , ²³ , ²⁴ Single‐cell RNA sequencing (RNA‐seq) studies confirmed the selective depletion of PV⁺‐ and SST⁺‐positive GABAergic inhibitory neurons. ²⁵ , ²⁶

To determine whether miR155 overexpression individually contributed to the changes described above, we utilized (1) in silico analyses of published databases to show the relationship in three‐dimensional (3D) chromatin structure between transcription of MIR155HG and APP; (2) newly generated MIR155‐deleted hiPSCs using CRISPR/Cas9 genome editing technology ²⁷ to produce NSCs, cortical and GABAergic neurons, and cortical organoids; (3) lentivirus‐mediated MIR155‐overexpressing hiPSC‐derived NSCs and cortical and GABAergic neurons; and (4) miR155 KO mice to compare and contrast the effects of decreased miR155 in vitro and in vivo. Integration of data from these systems supports a novel role for MIR155 in the regulation of interneuron numbers and phenotypes highly relevant to AD and DS, likely dependent on NR2F1 and NR2F2. Taken together, our findings suggest distinct cell‐type‐specific roles of miR155 in GABAergic interneurons in the pathogenesis of AD. This raises the possibility that therapeutic modulation of miR155 in both neurons and microglia might be beneficial in the treatment and/or prevention of AD, but the cell‐type‐specific valence creates a unique challenge, as altering miR155 expression in either direction or at different stages of disease could be harmful in one cell type while being protective in another. ²⁸

2. METHODS

2.1. Animals

The experimental procedures were conducted in accordance with National Institutes of Health (NIH) guidelines for animal research and were approved by the Institutional Animal Care and Use Committee (IACUC) at the Icahn School of Medicine at Mount Sinai. All mice were on a C57Bl6/J background. APP^KM670/671NL/PSEN1^Δexon9 (= APP/PS1) ²⁹ and miR155 knockout (KO) (miR155^−/− ) ³⁰ mice were obtained from Jackson Laboratories. APP/PS1 mice were crossed with miR155^−/− mice to obtain WT, miR155^+/− , miR155^−/− , APP/PS1, APP/PS1‐miR155^+/− , and APP/PS1‐miR155^−/− mice. Male and female mice were used for immunohistochemistry, transcriptomic analyses, the quantification of target cell numbers, amyloid plaque positive area quantification, and Western blot analyses. If applicable, the use of both sexes is specified in the figure legends, and 4‐ and 8‐month‐old mice were sacrificed by decapitation. One hemisphere was collected and immersion‐fixed in 4% paraformaldehyde (PFA, w/v in phosphate‐buffered saline [PBS]) for immunohistochemistry analysis. The other hemisphere was dissected, and the prefrontal cortex (PFC) and hippocampus were collected for qPCR and transcriptomic analysis. PFC, hippocampi, and cerebral hemispheres were snap‐frozen and stored at −80°C prior to RNA isolation or biochemical analysis.

2.2. Maintenance and culture of hiPSCs

All stem cell work was performed at the New York Stem Cell Foundation Research Institute (NYSCF). The three hiPSC lines used in this study, AJ0083, AJ0094, and AJ0123 (all APOE ε3/ε3 background), were generated as part of the Religious Orders Study/Memory and Aging Project  (ROSMAP) cohort ³¹ using the modified mRNA reprogramming method and were characterized for the typical hiPSC pluripotency, a normal karyotype, and absence of mycoplasma contamination. hiPSCs were maintained in six‐well plates (Corning) in feeder‐free conditions using Geltrex in complete mTeSR1 medium in a humidified incubator (5% CO², 37°C). hiPSCs were fed fresh medium daily and passaged using ReLeSR every 7 to 8 days.

2.3. Generation of isogenic MIR155‐deleted hiPSC lines

We designed a pair of sgRNAs (Figure S1B and Table S3) targeting exon 3 of MIR155 host gene (MIR155HG) to remove a short (about 140 bp) genomic fragment containing pre‐MIR155 (65 bp) using the CRISPR Guide RNA design tool (https://www.benchling.com/crispr).

2.3.1. Electroporation

hiPSCs with 70% to 80% confluence were dissociated by treatment with Accutase and 10 µM ROCK inhibitor for 10 min. After spinning down hiPSCs at 300 × g for 5 min, cells were counted, and two million cells were subjected to electroporation. Electroporation was performed using Nucleofector—Amaxa and Human Stem Cell Nucleofector Kit 1 (Lonza) according to the manufacturer's instructions. In brief, cells were resuspended in 100 µL of reaction buffer (82 µL Nucleofector solution and 18 µL Supplement) from the kit and 25 µL of ribonucleoprotein particles (RNPs) and transferred to a Nucleocuvette. The RNPs were complexes formed by combining 15 µL of sgRNA (100 µM, resolved in pH 7.0 TE buffer) and 10 µL of Cas9 (Alt‐R S.p.Cas9 Nuclease V3 (Integrated DNA Technologies), 20 µM, resolved in pH 7.0 TE buffer) and followed by incubation at room temperature for 20 min. Further, 2 µL of puromycin plasmid (0.5 µg/µL) were added to cell suspension. After nucleofection with the protocol B‐016, cells were transferred to Geltrex‐coated cell culture plates and cultured in complete mTeSR1 medium containing 10% CloneR for 2 days. Puromycin (1.0 µg/mL) was added to the medium for 1 day to select for electroporated hiPSC clones.

RESEARCH IN CONTEXT

1. Systematic review: We recently found that miR155 was upregulated in hippocampal neurons of AD patients compared with age‐matched controls. The authors reviewed previous publications using traditional sources, including PubMed and preprint server databases. The role of miR155 has been extensively studied in immune cells and mononuclear phagocytes, especially microglia in neurodegenerative diseases. There are no peer‐reviewed publications regarding the role of neuronal miR155 in the brain or in the context of AD.

2. Interpretation: Our results reveal previously unrecognized miR155 roles in hippocampal NSC dynamics and GABAergic interneuron development in AD.

3. Future directions: Our findings of a neuronal role for miR155 in AD should provide context for further investigation of underlying molecular mechanisms regarding how miR155 regulates its mRNA targets in the context of AD. These insights should help clarify the role(s) in AD pathogenesis for miR155 and its mRNA targets (especially NR2F1/2) and may have diagnostic and/or therapeutic implications.

2.3.2. Polymerase chain reaction (PCR), DNA electrophoresis, and DNA Sanger sequencing

After 1‐day puromycin selection, each colony was transferred to one well of a 48‐well plate coated with Geltrex and maintained in complete mTeSR1 medium until the colony grew large enough to be transferred to a six‐well plate for further expansion. After the second transfer, hiPSCs in the original plate were dissociated, and genomic DNA was extracted. Genotyping primers (Figure S1B and Table S3) were used to amplify a DNA fragment containing the MIR155HG gene, and PCR products were subjected to DNA electrophoresis and submitted to GENEWIZ for Sanger sequencing.

2.3.3. Karyotyping

To identify and evaluate the size, shape, and number of chromosomes in hiPSCs, we performed karyotyping (Giemsa banding and fluorescence in situ hybridization [FISH]) after Sanger sequencing. hiPSCs were cultured on Geltrex‐coated six‐well plates in complete mTeSR1 medium until 60% confluence and then sent to the Tumor CytoGenomics laboratory at the Icahn School of Medicine at Mount Sinai for karyotyping.

2.4. Cortical neuron differentiation

hiPSC‐derived cortical neurons were generated as previously described ³² (Figure S2A). For neural differentiation to generate NSCs, hiPSCs were dissociated with Accutase and replated on 12‐well tissue culture plates coated with Geltrex in complete mTeSR1 medium with 10 µM ROCK inhibitor. When cells were 100% confluent, the medium was replaced with Neural Induction Medium (NIM, 50% DMEM/F12 medium, 50% neurobasal medium, 2% B27 minus vitamin A, 1% N2 supplement, 2 mM GlutaMAX, and 100 U/mL penicillin‐streptomycin) (days in vitro 0 [DIV0]) and maintained for 15 days. Inhibitors used in dual‐SMAD inhibition protocol (LDN/SB/XAV, DIV0‐10; XAV, DIV11‐15) included LDN193189 (100 nM), SB431542 (10 µM), and XAV939 (1.0 µM). On DIV15, hiPSC‐derived NSCs were dissociated using Accutase and could be stocked in Synth‐a‐Freeze cryopreservation medium (Thermo Fisher Scientific) in a liquid nitrogen tank. Two wild type (WT) NSC lines derived from NIMH cohorts (553 and 2607) were used in the experiments of Aβ42 prefibrillar and fibrillar oligomer mixture (A11‐OC, Table S1) or OC‐type amyloid beta (Aβ) 42 fibrillar oligomer (OC, Table S1).

For terminal differentiation to generate functional cortical neurons, hiPSC‐derived NSCs (DIV15) were replated at 200,000 to 300,000 cells/cm² in NIM with 10 µM ROCK inhibitor on dried poly‐l‐ornithine (Sigma‐Aldrich) and laminin‐coated six‐well plates or eight‐well slide chambers. Cells were left to adhere overnight. On DIV16, NIM with ROCK inhibitor was replaced with neuron maturation medium (BrainPhys Neuronal Medium, 2% B27 with Vitamin A, 2 mM GlutaMAX, and 100 U/mL penicillin‐streptomycin) supplemented with PD0325901 (10 µM), SU5402 (10 µM), DAPT (10 µM), brain‐derived neurotrophic factor (BDNF) (40 ng/mL), glial cell‐derived neurotrophic factor (GDNF) (40 ng/mL), laminin (1.0 µg/mL), dBcAMP (250 µM), and L‐ascorbic acid (200 µM). During DIV16‐23, neuron maturation media were half‐changed every other day. After DIV24, the medium was replaced with neuronal maintenance media (BrainPhys Neuronal Medium, 2% B27 with Vitamin A, 2 mM GlutaMAX, and 100 U/mL penicillin‐streptomycin) supplemented with BDNF (40 ng/mL), GDNF (40 ng/mL), laminin (1.0 µg/mL), dBcAMP (250 µM), and L‐ascorbic acid (200 µM). After DIV32, hiPSC‐derived cortical neurons were active at this stage, which was confirmed by Ca²⁺ imaging. Between DIV50 and DIV60, hiPSC‐derived cortical neurons were used for electrophysiological recordings.

2.5. Electrophysiological recordings and analysis

For whole‐cell recordings, cortical neurons derived from WT and MIR155‐deleted hiPSC lines between DIV50 and 60 were visualized using an upright Olympus BX51WI microscope equipped with a 40× objective and differential interference contrast optics. Cortical neurons were constantly perfused with 95% O₂/5% CO₂ in BrainPhys medium preheated to 30°C to 31°C. Patch electrodes were filled with internal solutions containing 130 mM K⁺gluconate, 6 mM KCl, 4 mM NaCl, 10 mM Na⁺HEPES, 0.2 mM K⁺EGTA; 0.3 mM GTP, 2 mM Mg²⁺ATP, 0.2 mM cAMP, and 10 mM D‐glucose. The pH and osmolarity of the internal solution were adjusted to resemble physiological conditions (pH 7.3, 290 to 300 mOsmol). Current‐ and voltage‐clamp recordings were carried out using a Multiclamp 700B amplifier (Molecular Devices), digitized with Digidata 1440A digitizer, and processed with a pClamp 10.0 software package (Molecular Devices). For spontaneous EPSC recordings, cortical neurons were held at a chloride reversal potential of −75 mV. Data processing and analysis were performed using ClampFit 10.0 (Molecular Devices) and MATLAB (MathWorks) software. P values to compare neurons alone to neurons with MIR155 deletion were calculated using a two‐tailed, unpaired t‐test.

2.6. Isolation and characterization of lentivirus‐mediated MIR155‐overexpressing hiPSC‐derived NSCs and GABAergic interneurons

Human MIR155 (hsa‐MIR155) precursors and approximately 100‐bp upstream and downstream flanking genomic sequences were PCR amplified and cloned into a self‐inactivated (SIN) lentiviral vector to generate pLV‐miRNA vectors (Figure S2D). The cloning site of pre‐miRNA genomic fragments is within the intron of the human housekeeping gene EF1α promoter region. The rPuro gene product, expressed from the EF1α promoter, is the red fluorescent (mCherry) puromycin‐N‐acetyltransferase, and the transduced cells display red fluorescence at excitation/emission wavelengths of 587/610 nm. The miRNA lentiviral stock is prepared from cotransfecting HEK 293T cells with the pLV‐miRNA plasmid and plasmids expressing Gag‐Pol gene products and the vesicular stomatitis virus envelope G (VSV‐G). The lentiviral supernatants were collected at 48 h after transfection and stored at −70°C. The titer of the virus is generally above 1 × 10⁷ infectious units per mL (IU/mL). The miRNA of interest is delivered into cells by lentiviral transduction.

Dissociated hiPSC (cell line AJ0083)‐derived NSCs were cultured on Geltrex‐coated six‐well or 96‐well plates in NIM containing 10 µM ROCK inhibitor. For lentivirus infection, we used lentivirus diluted in NIM containing 8 µg/mL polybrene (R&D Systems) to directly infect hiPSC‐derived NSCs at a range of MOIs (1 to 20), which was calculated based on the starting virus concentration (2 × 10⁷ PFU/mL) and initial seeding density. To increase transduction efficiency, the plates were spun down at 1000 × g for 30 min at room temperature. After a 3‐day infection, the medium was replaced with fresh NIM containing FGF2 (20 ng/mL), EGF (20 ng/mL), and puromycin (1.0 µg/mL) to select and expand for stably transduced cells. MiRNA qPCR analysis and double immunolabeling with mCherry and markers of NSCs (NESTIN, SOX2, SOX1, and Ki67) were performed for characterization of lentivirus‐mediated MIR155‐overexpressing hiPSC‐derived NSCs (Figure S2E,F). All virus work was approved by the Mount Sinai Institutional Biosafety Committee.

hiPSC‐derived cortical GABAergic interneurons were generated as described ³³ (Figure S2B). Briefly, dissociated hiPSC‐derived NSCs (DIV15) were cultured for 8 days on Geltrex‐coated six‐well plates in NIM containing 10 µM ROCK inhibitor, 1.0 µM purmorphamine, and 5 nM SHH, a treatment condition referred to as “SHH,” which enabled rapid and robust induction of ventral forebrain progenitor populations. For terminal differentiation to generate functional cortical GABAergic interneurons, hiPSC‐derived NSCs (DIV23) were replated at 200,000 to 300,000 cells/cm² in NIM with 10 µM ROCK inhibitor on dried poly‐l‐ornithine (Sigma‐Aldrich) and laminin‐coated six‐well plates or eight‐well slide chambers. Cells were left to adhere overnight, and on DIV24, NIM with ROCK inhibitor was replaced with neuron maturation media (BrainPhys Neuronal Medium, 2% B27 with vitamin A, 2 mM GlutaMAX, and 100 U/mL penicillin‐streptomycin) supplemented with BDNF (40 ng/mL), GDNF (40 ng/mL), laminin (1.0 µg/mL), dBcAMP (250 µM), and L‐ascorbic acid (200 µM). During DIV24‐70, neuron maturation media were half‐changed every other day.

2.7. Cortical organoid differentiation

Cortical organoids were generated from hiPSCs as previously described, with modifications noted below ³⁴ (Figure 5A). For the patterning of neurocortical organoids, hiPSCs cultured on Geltrex were disassociated with Accutase to a single‐cell suspension, to which was added an appropriate volume of mTeSR1 medium containing ROCK inhibitor (10 µM) to obtain 20,000 live cells per 200 µL. Then 200 µL of single‐cell suspension were replated in each well of an ultra‐low‐adherence V‐bottom 96‐well plate (S‐Bio Prime; MS‐9096VZ), and the plate was centrifuged at 1000 × g for 5 min at room temperature. The following day the medium was changed to mTeSR1 medium without ROCK inhibitor. On differentiation day 1, the cells were induced in 200 µL of Organoid Starter Medium with 250 nM LDN193189 and 10 µM SB431542. Organoid Starter Medium was DMEM/F12 containing 1% N2 supplement, insulin (25 µg/mL), MEM non‐essential amino acids (NEAA), Glutamax, β‐mercaptoethanol, and 100 U/mL penicillin‐streptomycin. The same media were used for the next 5 days with daily half‐medium changes, after which the media were changed to Organoid Differentiation Media. Organoid Differentiation Media were Neurobasal‐A medium containing B‐27 supplement without vitamin A, Glutamax, and 100 U/mL penicillin‐streptomycin. From differentiation day 7 to day 25, 20 ng/mL FGF2 and 20 ng/mL EGF were added to the Organoid Differentiation Media. Organoids were cultured in 96‐well plates through day 20 and fed every other day with Organoid Differentiation Media. On day 20, organoids were transferred to ultra‐low‐attachment 24‐well plates (Corning; CLS3473) at a density of one organoid per well and cultured in that form through the remainder of the protocol. Neural differentiation was induced between days 27 and 41 by supplementation of Organoid Differentiation Media with 20 ng/mL BDNF and 20 ng/mL NT3. Half‐medium changes were performed every other day between days 17 and 41.

FIGURE 5.

miR155 deletion enhanced the ventral patterning and GABAergic interneuron generation in 3D cortical organoids. (A) Schematic for generating cortical organoids from hiPSCs. The culture conditions, medium, and cocktails of growth factors or small molecules are indicated. (B and D) 10‐day cortical organoids derived from WT and MIR155KO Clone 5 or Clone 25 hiPSCs were immunostained with antibodies against SOX1 (red) and PAX6 (green) (B), SOX2 (red) and N‐cadherin (green) (D). Scale bar, 100 µm (B and left panel of D) and 40 µm (right panel of D). (C, E, F) Percentage of PAX6‐ and SOX1‐positive NSCs over total cells (DAPI‐positive) (C), the density (E) and area (F) of neural rosettes in 10‐day cortical organoids. The numbers of PAX6‐ and SOX1‐positive NSCs, total cells (DAPI‐positive), and neural rosettes were counted using ImageJ software. Neural rosettes and the corresponding organoid were outlined by manual tracing, and then the area per rosette or organoid was measured with ImageJ software. (G, H, I) Representative images of neural progenitors (G, NKX2‐1 [ventral, red] and PAX6 [dorsal, green]) and GABAergic interneuron progenitors (H, NKX2‐1 [ventral, red] and OLIG2 [ganglionic eminence, green]) in cortical organoids at week 6 of differentiation. Scale bar, 40 µm. (I) Number of NKX2‐1‐positive cell populations per organoid at week 6 of differentiation. (J and K) Cortical organoids at week 6 of differentiation, which were derived from WT or MIR155KO Clone 5 or 25 hiPSCs, were immunostained with antibodies against TUJ1 (green) and GABA (red, J), along with nuclear counterstaining with DAPI. Scale bar, 40 µm. (K) Percentages of GABA‐positive cells out of total neurons (TUJ1‐positive) in organoids at week 6 of differentiation. (L and M) Representative images of SST‐positive GABAergic interneurons (L, co‐stained with NeuN) in cortical organoids at day 66 of differentiation. Scale bar, 40 µm. (M) Number of SST‐positive cell populations per organoid at day 66 of differentiation. All image analyses of cortical organoids were performed on the images (n = 16) from 16 organoids per batch of organoid culture for a total of three independent batches of organoid cultures; *p < 0.05, ***p < 0.001. Data represent the mean, error bars indicate standard deviation.

To improve the maturation of cortical organoids, beginning on day 42, organoids were cultured in Organoid Maturation Media (Neurobasal‐A medium containing B‐27 supplement with vitamin A, Glutamax, and 100 U/mL penicillin‐streptomycin), and 200 µM vitamin C and 250 µM dbcAMP were added into the every‐other‐day medium changes for 2 months. After day 100, mature cortical organoids were maintained in Organoid Maturation Media with every‐other‐day medium changes until completion of the experiment.

2.8. Immunocytochemistry and immunohistochemistry

Sixteen cortical organoids at day 10 of differentiation per batch of culture were pooled together for whole‐mount staining and initially fixed with 4% ice‐cold PFA for 1 h at 4°C, washed three times in PBS, and penetrated with PBS/0.3% Triton X‐100. For immunocytochemistry (ICC), cells were washed three times with DPBS (1X) and fixed with cold 4% PFA for 15 min at 4°C, followed by three washes for 10 min each with PBS/0.1% Triton X‐100 (PBST). The fixed cells or organoids were blocked with 10% goat or donkey serum/PBST for 1 h at room temperature. The cells or organoids were then incubated with primary antibodies at the appropriate dilutions in 5% goat or donkey serum/PBST at 4°C overnight. The next day, cells or organoids were washed three times with PBST for 10 min each and stained with Alexa Fluor conjugated secondary antibodies (Thermo Fisher Scientific) at 1:400 for 2 h at room temperature in the dark. After a final three washes with PBST for 10 min each and DAPI staining for 5 min, the slides with cultured cells or mounted organoids were coverslipped with ProLong Gold Antifade Mountant (Thermo Fisher Scientific).

Sixteen cortical organoids at day 42 or 66 of differentiation per batch of culture for immunohistochemistry (IHC) were pooled together and initially fixed with 4% ice‐cold PFA for 2 h at 4°C, washed three times in PBS, and equilibrated with 30% sucrose at 4°C overnight. Then organoids were embedded in optimal cutting temperature O.C.T. (Tissue‐Tek) and sectioned at 10 µm using a cryostat, and the sections were mounted onto slides and stored at −20°C. For IHC, mouse brains were collected and fixed in 4% PFA at 4°C overnight. Brains were then washed three times with PBS and placed in sucrose solution (30% w/v in PBS) overnight before being embedded in O.C.T. (Tissue‐Tek). Embedded tissue was sectioned at 30 µm using a cryostat, and free‐floating sections were stored in cryopreservation medium at −20°C until staining. Sections were blocked in blocking solution (1X PBS, 0.3% Triton X‐100 and 10% goat serum) for 1 h at room temperature with gentle shaking. Prior to blocking, heat‐mediated antigen retrieval (NB: immunostaining experiments of organoid sections skipped this step) was performed by placing free‐floating sections in a 1.5‐mL micro centrifuge tube containing 1 mL of citrate buffer solution (pH 8.0) (Sigma) and placing in a preheated temperature block set at 80°C. Tissue was heated for 30 min at 80°C then removed and allowed to cool to room temperature before washing with PBS three times for 10 min each and then proceeding with the blocking step. The sections were then incubated with primary antibodies at appropriate dilutions in blocking solution at 4°C overnight with gentle shaking. The next day, sections were washed three times with PBST for 10 min each and stained with Alexa Fluor conjugated secondary antibodies at 1:400 for 2 h at room temperature with slight shaking in the dark. After secondary staining, sections were washed in PBST three times for 10 min each and mounted on glass slides. After mounting, slides were coverslipped with DAPI‐counterstain mounting medium (Fluoromount, Southern Biotech).

Antibodies used in this study were verified by immunostaining mouse brain sections (data not shown) and are listed in Table S1. DAPI was used to counterstain the nuclei. Stained sections were analyzed with a Leica SP5 DMI confocal microscope (Leica) and Zeiss LSM980 Airyscan 2 confocal microscope (Zeiss). Images were assembled using Adobe Illustrator.

Mouse brain confocal images were thresholded and binarized using the Fiji ImageJ algorithm. The areas of the region of interest (ROI) in mouse hippocampi, DG, granule cell layer (GCL), SGZ, and hilus, were outlined by manual tracing with ImageJ software and were calculated using the “measure” function. The numbers of RGL‐NSCs (Sox1‐, Sox2‐, and/or Gfap‐positive cells) and GABAergic interneurons in ROIs were manually counted, and numbers were normalized to the area of each ROI. Values were expressed as a percentage of the control group.

2.9. Western blotting

Aliquots of protein lysates (30 µg) prepared from mouse brains or hiPSC‐derived cells and homogenized in RIPA buffer containing dissolved Protease Inhibitor Tablets (Thermo Fisher Scientific) were loaded in Criterion XT 4–12% Bis‐Tris gels and transferred onto PVDF membranes (0.45 µm; Millipore, Billerica, MA, USA). The membranes were probed with primary antibodies (Table S1) at 4°C overnight with slight shaking. The next day, membranes were washed three times with TBST for 10 min each and subsequently incubated with anti‐rabbit or anti‐mouse HRP‐conjugated secondary antibodies (1:2000, Catalog No.: PI‐1000 or PI‐2000, Vector Laboratories) at room temperature for 2 h. After washing three times with TBST for 10 min each, membranes were developed with ECL Western blotting substrate (Pierce, Rockford, IL, USA) and were imaged on ChemiDoc MP imaging system (Bio‐Rad). Normalization was achieved using GAPDH antibody (Table S1). The integrated density of immunoreactive bands was measured using Fiji software (ImageJ).

2.10. RNA isolation, qPCR, and miRNA qPCR analysis

Snap‐frozen mouse brain tissues from 4‐ and 8‐month‐old mice and hiPSC‐derived cells were homogenized in QIAzol Lysis Reagent (Qiagen). Total RNA purification was performed with QIAGEN miRNeasy Mini Kit (Valencia, CA, USA) following the manufacturer's guidelines. RNA quantification and quality were assayed using a NanoDrop 2000 (Thermo Fisher Scientific, Wilmington, DE, USA). For reverse transcription, 1.0 µg of total RNA was transcribed into cDNA using a High‐Capacity RNA‐to‐cDNA™ Kit (Applied Biosystems). Quantitative real‐time PCR (qPCR) was performed in a StepOne Plus system (Applied Biosystem) using SYBR Green I Master (a hot start reaction mix, Roche, Germany), and data were normalized to the average of GAPDH and HPRT1 expression. To reduce non‐specific signals, primer sequences were designed using LightCycler Probe Design Software 2.0 (Roche, Germany) such that they spanned splice junctions; the primers are listed in Table S2 and were validated using the pga.mgh.harvard.edu/primerbank/program. Amplifications were performed in 20 µL containing 1.0 µL of each primer (5 µM), 10 µL SYBR Green I Master, 3 µL H₂O and 5 µL of 20‐fold diluted cDNA. Forty PCR cycles were performed with a temperature profile consisting of 95°C for 10 s, 60°C for 15 s, and 72°C for 15 s. The melting curve of each PCR product was determined to ensure that the observed fluorescent signals were only from specific PCR products, which were further verified by DNA Sanger sequencing. After each PCR cycle, the fluorescent signals were detected at 95°C for 5 s and 65°C for 60 s to melt primer dimers. T_ms of all primer dimers used in this study were <76°C.

In this study, TaqMan Small RNA Assays were performed to detect and quantify mature miRNAs. For miRNA reverse transcription (RT), 15 ng of total RNA was transcribed into cDNA in RT Reaction Mix containing RT Primers (5X) (Table S2). MiRNA qPCR was performed in the StepOne Plus system using TaqMan Gene Expression Master Mix containing TaqMan Small RNA Assay (20X) (Table S2) and two‐fold diluted cDNA templates. Data were normalized to the expression of endogenous control RNU49 (small nucleolar RNA, C/D box 49A, U49). The temperature profile of qPCR was described as above but the melting curve step was skipped during miRNA qPCR.

All qPCRs were performed on three biological replicates and repeated using more than three independent samples, and data were plotted as means ± standard deviation (SD).

2.11. miRNAscope

For miRNAscope (FISH) to detect miR155 expression in hippocampal cells (NSCs and glutamatergic neurons and/or GABAergic interneurons), mouse brains were dissected and snap‐frozen in liquid nitrogen before being embedded in O.C.T. (Tissue‐Tek). miRNAscope was conducted on mouse brain sections using RNAscope Plus smRNA‐RNA HD Reagents Kit (Advanced Cell Diagnostics, Catalog No.: 322785) and miRNAscope Probe‐SR‐mmu‐miR‐155‐5p‐S1 (Advanced Cell Diagnostics, Catalog No.: 887751‐S1) according to the manufacturer's instructions with some modifications. The dehydrated tissue sections were treated with hydrogen peroxide for 10 min at room temperature. After rinsing with water, the sections were treated with protease III for 20 min at 40°C using the HybEZ hybridization system. After washing with water, the sections were incubated with the mmu‐miR‐155‐5p‐S1 probe for 2 h at 40°C using the HybEZ hybridization system. After washing with RNAscope wash buffer, amplifiers 1 to 3 were used following the manufacturer's protocol, and the tissue sections were covered in HRP‐S1 reagent and incubated for 15 min at 40°C. After washing with RNAscope wash buffer, the sections were covered with diluted fluorophore Opal 570 and incubated for 30 min at 40°C. After washing with RNAscope wash buffer, the sections were covered with HRP Blocker and incubated for 15 min at 40°C, followed by RNAscope wash buffer. Finally, slides were coverslipped with DAPI‐counterstain mounting medium (Fluoromount, Southern Biotech).

2.12. RNA sequencing data analysis

RNA isolation was performed on the lysates of WT, MIR155KO Clone 5, and MIR155KO Clone 25 hiPSC‐derived cells, and MIR155‐overexpressing NSCs and cortical neurons, according to the manufacturer's protocol (miRNeasy mini‐kit, Qiagen). Isolated RNA samples were submitted to Novogene Corp. for analysis of RNA quality and integrity. All samples passed quality control, and RNA‐seq libraries were prepared for sequencing using NovaSeq. Paired‐end RNA‐seq FASTQ files for all samples were aligned to the Homo sapiens reference genome (Grch38) using STAR read aligner, and mapped reads were summarized to gene counts using the “subread” function of FeatureCounts. Genes with expression (<1 FPKM) across all samples were filtered from all subsequent analyses. Hierarchical clustering and principal component analysis (PCA) were performed as part of quality control to visualize biological variance between replicates, which were displayed using volcano plots and Venn diagrams. We then performed differential gene expression analysis to compare all cell types with each other on normalized counts with EdgeR. Multiple biological replicates were used for all comparative analyses. Differentially expressed genes (DEGs) were identified at a false discovery rate (FDR) of 0.05. To identify biological pathways that may be differentially dysregulated by MIR155KO or MIR155 overexpression, we performed gene set enrichment analysis (GSEA) on the identified DEGs using Enrichr, GSEA‐MSIGDB, and Ingenuity Pathways Analysis (IPA), where we identified enriched biological pathways, gene ontology sets, and mRNA targets of miRNA at an adjusted p value < 0.05. In heatmaps (Figure 3B,C,F,G and Figures S4A,S4C,S4E,S8I,S8J,S8P), the x‐axis (horizontal) represents different comparative analysis groups (MIR155KO Clone 5 hiPSC‐derived cells vs WT hiPSC‐derived cells, MIR155KO Clone 25 hiPSC‐derived cells vs WT hiPSC‐derived cells, and MIR155‐overexpressing cells vs control cells). Genes are represented along the (vertical) y‐axis. The upper‐right scale bars for color coding display the range of Log₂ fold change, with values greater than 0 indicating upregulation (red) and values less than 0 indicating downregulation (green).

FIGURE 3.

miR155 is a developmental regulator of hiPSC‐derived GABAergic interneurons. (A) GO terms from differential gene expression analysis of WT and MIR155‐deficient hiPSC‐derived NSCs. Genes upregulated in MIR155‐deleted NSCs are associated with top statistically significant terms (red histogram) in four modules; Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis is shown as a bar graph. (B) Heatmap of top statistically significant DEGs encoding GABAergic and glutamatergic synaptic proteins that were upregulated in MIR155‐deleted NSCs (90 genes derived from GO terms in Figure 3A and Figure S3E, typical genes encoding GABA receptors, SLC6A1, GAD1, GAD2, and glutamate receptors are marked in red). Adjusted p value of DEGs shown in heatmap < 0.05. (C) Expression profiles of neural transcription factors (NKX2‐1, LHX6, DLX1, DLX2, OLIG1, OLIG2, …) and genes involved in signaling pathways (NOTCH, WNT, BMP, SHH, …) are shown as a heatmap. Adjusted p value of DEGs shown in heatmap < 0.05. (D) Cortical neurons differentiated from WT and MIR155KO Clone 5 and MIR155KO Clone 25 hiPSCs were stained for TUJ1 at differentiation days 22 and 32. Scale bar, 50 µm. (E) Excitatory postsynaptic current (EPSC) properties were recorded in WT and MIR155‐deleted hiPSC‐derived cortical neurons (n = 13 and 12, respectively) at differentiation day 50 to 60 from two batches of cultures. ***p (= 0.000138) < 0.001; t‐test. (F) Dysregulated expression of GABAergic and glutamatergic signaling components in hiPSC‐derived cortical neurons induced by MIR155 deletion are shown as a heatmap. Adjusted p value of DEGs shown in heatmap < 0.05. (G) Heatmap of downregulated genes, dorsal forebrain PAX6‐derived cortical neuronal transcription factors including PAX6, TBR1, SATB2, and EMX2 (in green), NR2F1 and NR2F2 (in green). Here 59 genes were derived from the top four GO terms of GO_Cellular_Component in (Figure S5D). Adjusted p value of DEGs shown in heatmap < 0.05. (H and I) Control and MIR155‐overexpressing GABAergic interneurons in culture at day 56 were stained with antibodies against TUJ1 (red, H), GABA (green, H), and markers of GABAergic interneuron subtypes, the calcium‐binding protein parvalbumin (PV, undetected) and the neuropeptide somatostatin (SST) (green, H). Scale bar, 50 µm. (I) The percentage of GABA‐ or SST‐positive cells out of total neurons (TUJ1‐positive) in control and MIR155‐overexpressing GABAergic interneurons was determined by cell counting. Data represent the mean, n = 9 (nine independent biological samples from three independent cell cultures). Error bars indicate standard deviation; ***p < 0.001; t‐test.

2.13. ATAC‐sequencing, H3K4me3 ChIP–sequencing, H3K27ac ChIP–sequencing, and Hi‐C‐sequencing data analysis

These external validation datasets were used: ³⁵ , ³⁶ , ³⁷ , ³⁸ MSBB RNA‐seq of post mortem brains (Synapse ID: syn3157743), ATAC‐seq on fluorescence‐activated nuclear sorting (FANS)‐sorted NeuN+/− nuclei from post mortem adult brains (Synapse ID: syn21531502), H3K4me3 and H3K27ac ChIP–seq of neuronal nuclei isolated from adult control brains (Synapse ID: syn25705564), and Hi‐C‐seq on human fetal cortical plate and on FANS‐sorted neuronal (NeuN+) and glial (NeuN−) nuclei isolated from adult prefrontal cortex (Synapse ID: syn26256090 and syn21760712).

ATAC‐seq libraries were generated from neuronal and non‐neuronal nuclei isolated by FANS from frozen post mortem human brain tissue dissections as described. ³⁵ All libraries were processed by an established computational pipeline that performs read mapping (STAR), peak calling (MACS), genotype calling (GATK and KING), and quality control checking. Extensive quality control of ATAC‐seq libraries based on the cell type, sex, and genotype concordance, as well as sample quality metrics and sequencing depth, yielded a total of 636 samples constituting a total of 19.6 billion read pairs with an average of 30.8 million non‐duplicated read pairs per library.

The Hi‐C‐seq data analysis was performed using a standard protocol. ³⁵ , ³⁶ , ³⁸ Hi‐C‐seq data were aligned using the HiC‐Pro strategy. Briefly, paired‐end reads were mapped independently to human genome hg38 using bowtie2 in stringent mode with parameters “—very‐sensitive ‐L 20 —score‐min L,‐0.6,‐0.2 —end‐to‐end.” Then the chimeric reads that failed to align were trimmed after ligation sites (MboI “GATCGATC”) and mapped to the genome. All the aligned reads from two ends were then merged based on read names and mapped to MboI restriction fragments using the hiclib package. After that, self‐circles, dangling ends, PCR duplicates, and genome assembly errors were discarded. Samples of the same cell type were merged. We binned the interaction matrix at different resolutions and corrected it with iterative correction (ICE) for downstream analysis. Chromatin loops were called independently using HICCUPS for two different cell types. First, we converted the filtered interaction files into Juicer format with JuicerTools. Chromatin loops were called using Juicer HICCUPS with bin sizes iterated from 10 to 25 kb in 1‐kb intervals and parameters “‐k VC_SQRT ‐p 1 ‐i 3.” Only reproducible loops were retained, and only the highest resolution of the overlapping loops was used. Topologically associated domains (TADs) were identified with TopDom (version 0.0.2(18)) at 10K resolution and 200‐kb window size. For compartment analysis, first, we calculated the genome‐wide correlation matrix at 200‐kb resolution with only intrachromosomal interactions. Then the first eigenvector of the correlation matrix was obtained, and the sign was corrected to have a positive correlation with guanine‐cytosine content and gene density. The signs of the eigenvector were used to assign the genome into compartments A and B.

2.14. Statistical analysis

No statistical methods were used to predetermine sample size, and the experiments were not randomized. Experimental data were analyzed for significance using Excel. P < 0.05 was considered statistically significant. All experiments were analyzed by ANOVAs followed by Tukey's test, Dunnett's test, or unpaired Student's t‐tests.

2.15. Data availability

The Gene Expression Omnibus (GEO) database accession numbers for the RNA sequencing data reported in this paper are GSE277199 and GSE277200.

3. RESULTS

3.1. Colocalization of APP and MIR155HG genes in a neuron‐specific chromatin TAD within a regulatory network on human chromosome 21

MIR155 is encoded by the host gene, MIR155HG, also known as the B‐cell Integration Cluster (BIC) gene. It is composed of three exons that span a 13‐kb region on human chromosome 2. ³⁰ MIR155HG is in the obligate DS locus, proximal to the APP gene, which contains 18 exons spanning 290 kb (Figure S1A). Both are overexpressed in DS patients, all of whom develop AD neuropathology. ¹¹ , ¹² We analyzed genome‐wide maps of chromatin accessibility, ATAC‐seq (assay for transposase‐accessible chromatin with high‐throughput sequencing) data from neuronal nuclei isolated from the superior temporal gyrus (STG) and entorhinal cortex (EC) of AD cases and controls from the Mount Sinai Brain Bank AD (MSBB‐AD) cohort. ³⁵ In addition, we analyzed H3‐lysine 4 trimethylation (H3K4me3) and H3‐lysine 27 acetylation (H3K27ac) chromatin immunoprecipitation sequencing (ChIP‐seq) data from PFC neurons of adult control brains. ³⁷ Within MIR155HG, we identified ATAC‐seq peaks in exon 3, H3K4me3 ChIP‐seq peaks in the transcription start site and exon 3 and H3K27ac ChIP‐seq peaks in exon 3 (Figure 1A). These histone markers and chromatin accessibility strongly suggest that MIR155HG and APP, along with other genes proximal to APP that are associated with AD (e.g., MRPL39, JAM2, and GABPA; Figure S1A), ³⁹ are actively transcribed in human neurons (Figure 1A).

FIGURE 1.

Colocalization of APP and MIR155HG genes in a neuron‐specific chromatin TAD within a regulatory network. (A) ATAC‐seq peaks in STG and EC neurons isolated from AD patients and controls, H3K4me3 and H3K27ac ChIP‐seq peaks in PFC neurons of adult control brains visualized using the Integrated Genome Viewer (IGV). The peaks in exon 3 of MIR155HG are marked with a blue rectangle. (B) The APP gene‐centered TADs and compartments on chromosome 21 were identified in neuronal, fetal, and glial nuclei and were visualized using the IGV. Individual TADs were labeled with blue (“2”) or red lines (“1”). Changes in compartment scores were demonstrated with the orange line. (C and D) Chromatin contact matrices from 3D genome browser JuiceBox comparing merged Hi‐C sequencing data in adult neurons (NeuN ⁺ cells, top‐right triangles in panels C and D) versus fetal cortical plate tissue (C, bottom‐left triangle) and adult neurons versus glia (NeuN− cells) (D, bottom‐left triangle) for a 0.7‐Mb locus (chr21:25,500,000‐26,200,000) centered on APP gene at 5‐kb resolution. The Hi‐C color scale ranges normalized contact frequencies within this locus. Compared to fetal (C) and glia (D), the neuron‐specific chromatin contacts were marked with circles. The blue right angles in the upper right of panels (C) and (D) spanning the genomic region between MIR155HG and APP genes represent a neuron‐specific chromatin TAD; however, in fetal (C) or glial nuclei (D), the corresponding genomic region was divided into two chromatin TADs, which are further marked with a red or blue right angle in the lower left of the figure. (E) Comparative analysis (neuron vs fetal and neuron vs glia) of the frequency of chromatin interacting regions of MIR155HG‐JAM2, MIR155HG‐GABPA, MRPL39‐APP, JAM2‐APP, and GABPA‐APP is summarized in the table. Observed value (O), expected value (E), control value (C), expected Control value (EC), (O/E)/(C/EC) = (observed value/expected value)/(control value/expected control value). Compared to control (neuron), the interaction frequency (observed, fetal or glia) significantly decreased. (F) Dynamic expression profiles of JAM2, GABPA, MRPL39, APP, and CYYR1 in cortical neurons derived from wild type (WT), MIR155KO (Clone 5), or MIR155KO (Clone 25) hiPSCs, which were generated from RNA sequencing datasets (seen in Figures S1–S5), are illustrated as a chart. ***Adjusted p value (adj.p val) < 0.001. (G) Schematic illustration of localization of APP and MIR155HG genes in chromatin TADs within regulatory networks. Here chromatin TADs in neuronal (top), fetal (middle), and glial nuclei (bottom) are shown. Blue or red rectangles represent the transition of individual chromatin TADs. Red and green curves indicate the strong and weak interactions among DNA fragments of different genes, respectively.

Chromosomes are partitioned into megabase‐sized compartments, named A and B, which contain active and repressed chromatin, respectively. Compartments are further separated into self‐interacting neighborhoods called TADs, the disruption of which plays an important role in transcriptional alterations in disease. ⁴⁰ The published data of in situ Hi‐C‐seq, a method that comprehensively detects genome‐wide chromatin interactions in the nucleus, were generated from nuclei isolated from human 18 to 24 weeks post‐conception cortical plate tissue (designated “fetal”) and from neurons and glia sorted from adult prefrontal cortex (designated “neuron” and “glia”). ³⁵ , ³⁶ , ³⁸ Compared to those in fetal nuclei, there were dramatic transitions in chromosome compartments around the APP gene in neurons and glia (Figure 1B). We found a neuron‐specific, APP‐centered TAD that included genes proximal to APP, including MIR155HG, MRPL39, JAM2, and GABPA (Figure 1B). Moreover, compared to those in fetal or glial nuclei, there were significant increases in the frequency of chromatin interacting regions in this neuron‐specific TAD, particularly in the interacting regions of MIR155HG‐JAM2, MIR155HG‐GABPA, MRPL39‐APP, JAM2‐APP, and GABPA‐APP (Figure 1C–E). These findings were further supported by MIR155KO RNA‐seq data (Figures S1–S5 and Figure 1F), which showed that MIR155 deletion caused statistically significant changes in the expression of JAM2 and GABPA and by an independent study showing the substantial regulatory impact of the APP gene on the expression of proximal genes on chromosome 21 in hiPSC‐derived DS neurons. ¹³ Taken together, these results showed the colocalization of APP and MIR155HG genes in a neuron‐specific chromatin TAD within a regulatory network (Figure 1G).

3.2. MIR155 is a repressor of proliferation in hiPSC‐derived NSCs

Mature MIR155 is processed from a long non‐coding primary transcript derived from exon 3 ³⁰ (Figure S1B). To obtain insights into the physiological function of MIR155 in neurons and their precursors, we generated isogenic MIR155 KO (MIR155KO) hiPSC lines using CRISPR/Cas9 methodology. ²⁷ The cell lines were individually monoclonalized and validated by PCR, DNA electrophoresis, and Sanger sequencing (Figure S1C–E). Further characterization of MIR155KO clones by immunostaining, qPCR, and G‐band karyotype analysis showed homogeneous expression of the pluripotency markers OCT4, NANOG, SSEA4, and TRA‐1‐60, an absence of random differentiation, and normal karyotypes (Figure S1F–H).

To characterize neural phenotypes caused by MIR155 deletion, MIR155KO hiPSCs, along with WT lines, were differentiated into human NSCs and cortical neurons ³² (Figure S2A). hiPSC‐derived NSCs expressed general markers of NSCs with dorsal forebrain fate, including FOXG1, NESTIN, PAX6, SOX1, SOX2, and Ki67 (Figure 2A). Cortical neurons (TUJ1‐positive) were observed to express TBR1, a marker of layer VI of the developing human cortex, and VGLUT1, a marker of glutamatergic excitatory neurons (Figure S2C). During the differentiation of human NSCs into cortical neurons, qPCR assays of miRNAs revealed that MIR155‐5P expression peaked in NSCs and dropped dramatically in mature neurons (Figure 2B), whereas MIR155‐3P was undetected. In parallel, the expression of neuron‐enriched MIR128, one of the most abundant miRNAs in the adult mouse and human brain, ³ , ⁴¹ increased gradually from NSCs to terminally differentiated neurons (Figure 2B). Our miRNA qPCR results are consistent with published data that revealed enriched MIR155‐5P expression in maturing deep layer neurons and newborn upper layer neurons in human brain. ² MIR155 deletion promoted the expansion of NSCs by doubling their proliferation (Figure 2C). In addition, MIR155‐deleted NSC clones were less cohesive and included more glial‐like progeny than WT. We performed RNA‐seq on samples from WT, MIR155KO Clone 5, and MIR155KO Clone 25 hiPSC‐derived NSCs and mature neurons (Figure S3A,B). As shown in GSEA (Figure 2D,E and Figure S3C,D) and the heatmap (Figure S4A), MIR155‐deleted NSCs upregulated genes associated with GLIOBLASTOMA_PRONEURAL and ASTROCYTE features. ⁴² , ⁴³ This suggests that MIR155‐deleted human NSCs acquired the identity of RGL‐NSCs, including high self‐renewal ability, increased proliferation, multilineage potency, and migration capacity. These findings also revealed the critical link between transcriptomic dysregulation and the significant increase in NSC proliferation caused by MIR155 deletion. Furthermore, MIR155 deletion led to downregulation of genes related to focal adhesion, extracellular matrix, and cytoskeleton (Figures S3F and S4B,C).

FIGURE 2.

miR155 is a repressor of proliferation in hiPSC‐derived NSCs. (A) Representative images of wild type (WT) and MIR155KO Clone 5 and Clone 25 hiPSC‐derived NSCs in cell culture at differentiation day 15. In the study, two isogenic MIR155KO hiPSC clones, Clones 5 and 25, were used in parallel with WT in all experiments. Scale bar, 50 µm. (B) The expression levels of MIR155‐5P, MIR155‐3P, and MIR128 in hiPSCs, hiPSC‐derived NSCs, and cortical neurons (in the absence or presence of MIR155) were determined by miRNA qPCR. (C) NSCs were seeded at identical densities in eight‐well slide chambers and cultured for 4 days and then labeled with markers of NSCs (A) and DAPI. The eight‐well slides were scanned to determine NSC numbers by Agilent BioTek Cytation (four‐well replicates of WT and MIR155KO Clone 5 or Clone 25 hiPSC‐derived NSC culture per each experiment for a total of three independent experiments). The y‐axis refers to relative proliferation speed normalized to WT. **p < 0.01, t‐test. (D and E) GSEA enrichment plots depicting VERHAAK_GLIOBLASTOMA_PRONEURAL (D) and ZHONG_PFC_C3_ASTROCYTE (E)‐associated transcriptional changes in MIR155‐deleted hiPSC‐derived NSCs. Adjusted p value of DEGs for GESA < 0.05A. (F) Canonical miRNA complementary sites. The seed region of miR155 is marked in red, and 6‐nt (6mer), 7‐nt (7mer‐A1 and 7mer‐m8), and 8‐nt (8mer) in 3’UTRs of target mRNAs that match the seed region of miR155 are marked in blue. The bottom panel shows the conserved 8mer site in 3’UTR of human SOX1. (G) The expression levels of SOX1 and SOX2 in control and MIR155‐overexpressing NSCs were determined by qPCR. (H) Representative images of control and MIR155‐overexpressing NSCs co‐stained for NESTIN (green) and SOX1 (red) or SOX2 (red), along with nuclear counterstaining with DAPI. Scale bar, 50 µm. (I) The percentage of SOX1‐ or SOX2‐positive cells out of the total cells (DAPI‐positive) in control and MIR155‐overexpressing NSCs was determined by cell counting. Data represent mean, n = 9 (nine independent biological samples from three independent cell cultures). Error bars indicate standard deviation (SD). **p < 0.01; t‐test. (J) Representative Western blots for SOX1, SOX2, and GAPDH using lysates of control and MIR155‐overexpressing NSCs. (K) Ratios of SOX1 or SOX2 to GAPDH from (J) showing a significantly lower SOX1 protein level in MIR155‐overexpressing NSCs than in control NSCs. Data are reported as mean, n = 9 (nine independent biological samples from three independent experiments). Error bars indicate SD. ***p < 0.001; t‐test. (L) The expression levels of MIR155‐5P and MIR128 in human NSCs treated with Aβ42 prefibrillar and fibrillar oligomer mixture (A11‐OC, 200 nM, Table S1) or OC‐type Aβ42 fibrillar oligomer (OC, 200 nM, Table S1) for 2 days were determined by miRNA qPCR. Two NSC lines derived from NIMH cohorts (553 and 2607) were used in the experiments, and the miRNA qPCR result from NSC line 2607 is shown here. In the qPCR histogram, the y‐axis indicates relative miRNA level normalized to RNU49 (B and L) or relative gene expression normalized to control (G). Data represent mean. Error bars indicate SD. Each data point represents nine technical replicates from three independent experiments; *p < 0.05, **p < 0.01, ***p < 0.001; t‐test.

To mimic miR155 upregulation in hippocampal neurons of AD and DS patients and mouse models, ⁸ , ¹⁰ , ¹¹ we established and characterized lentivirus‐mediated MIR155‐overexpressing human NSCs (Figure S2D–F). SOX1 is a NSC marker, ⁴⁴ and three independent programs that invoke evolutionary conservation (i.e., TargetScan, ⁴⁵ miRDB, ⁴⁶ and PicTar ⁴⁷ ) predict that its mRNA is a miR155 target due to a conserved octameric site complementary to the MIR155 seed sequence in the 3’UTR of SOX1 (Figure 2F). Furthermore, MIR155 binding to the 3’UTR of SOX1 is found in a database of experimentally validated mRNA targets of miR15. ⁴⁸ We found that the SOX1 transcript and protein levels were downregulated in MIR155‐overexpressing NSCs by about two‐fold (Figure 2G–K). RNA‐seq on MIR155‐overexpressing human NSCs and cortical neurons (Figure S7A,B) showed that SOX1 was among the top five downregulated DEGs (Figure S7C), further confirming that SOX1 mRNA is a target of MIR155. To investigate the proliferation capacity and the survival rate of hiPSC‐derived NSCs upon their treatment with Aβ42 oligomers and/or fibrils in order to simulate the presence of this toxic peptide in AD brain, we treated human NSCs with Aβ42 prefibrillar and fibrillar oligomer mixture (A11‐OC) or OC‐type Aβ42 fibrillar oligomers (OC) and found that A11‐OC treatment induced an increase in MIR155‐5P expression (Figure 2L), without significant changes in NSC proliferation or survival (data not shown). MIR155 overexpression led to a marked upregulation of genes related to focal adhesion, extracellular matrix, and cytoskeletal organization (Figures S4D and S7D), showing a pattern opposite to that observed with MIR155 deletion (Figure S4E). Lastly, a proliferation deficit has been consistently observed in NSCs differentiated from hiPSCs derived from DS patients where MIR155HG gene is overdosed. ⁴⁹ Taken together, these data suggest that MIR155 acts as a negative regulator of NSC proliferation.

3.3. MIR155 is a negative regulator of hiPSC‐derived GABAergic interneuron development

Our transcriptomic analysis (Figure S3C,D) highlighted upregulation of GABAergic synaptic gene expression in MIR155‐deleted hiPSC‐derived NSC cultures, revealing a previously unknown link between MIR155 and inhibitory interneuron development. Gene ontology enrichment analysis (Figure 3A and Figure S3E) and heatmap visualization (Figure 3B) demonstrated increased expression of genes encoding GABAergic and glutamatergic synaptic proteins following MIR155 deletion in hiPSC‐derived NSCs. These genes include GABA_A receptors (GABRA1, GABRB3, GABBR1, and GABRD), SLC6A1 (GAT1, GABA transporter 1), glutamate decarboxylases (GAD1 and GAD2), and ionotropic glutamate receptor subunits (GRIA2, GRIK2, GRIK3, GRIK5, and GRIN3A). RNA‐seq analysis showed that MIR155 deletion also induced changes in the expression of neural transcription factors and genes involved in signaling pathways (NOTCH, WNT, BMP, SHH) (Figure 3C) crucial for self‐renewal, proliferation, and neural differentiation. For example, there is upregulation of the SHH‐related gene PTCH2, which plays key roles in cell proliferation and induction of ventral patterning during neural development. ⁵⁰ Multiple transcription factors controlling neurogenesis, cell commitment, and maturation of cortical GABAergic interneurons, originating from the ventral telencephalon, ²¹ were significantly upregulated, including NKX2‐1, LHX6, DLX1, DLX2, OLIG1, and OLIG2 ²⁰ , ²¹ (Figure 3C), indicating that MIR155 may negatively regulate GABAergic interneuron development.

To further investigate the potential glutamatergic and/or GABAergic phenotype associated with MIR155 deletion, we differentiated hiPSCs toward cortical neurons, using a well‐established protocol that primarily generates dorsal glutamatergic excitatory neurons, along with a smaller proportion of interneurons ³² (Figure S2A,C). MIR155‐deleted neurons exhibited enhanced neurite growth (Figure 3D), consistent with another report. ⁵¹ Electrophysiological analysis showed that MIR155 deletion induced a significant increase in the frequency of spontaneous synaptic events in hiPSC‐derived cortical neurons, without affecting neuronal firing properties (Figure 3E and Figure S2G–H). Transcriptomic analysis (Figure S5A–D) comparing MIR155‐deleted versus WT neuronal cultures revealed a robust induction of GABAergic interneuron markers, including GAD1, GAD2, SLC6A1, and SST (Figure 3F), showing overlap with the genes upregulated in MIR155‐deleted NSCs, described earlier. These markers were likely induced at the expense of dorsal forebrain markers such as EMX2, PAX6, TBR1, and SATB2 (Figure 3G). MIR155 deletion induced downregulation of NR2F1 and NR2F2 (Figure 3G).

To mimic miR155 upregulation in hippocampal neurons in AD ¹⁰ and DS patients, ¹¹ we generated MIR155‐overexpressing hiPSC‐derived cortical GABAergic interneurons as described ³³ (Figure S2B and S2I–J). Only 37 DEGs were identified in MIR155‐overexpressing cortical glutamatergic neurons (Figures. S2D–F and S7E), raising the possibility that the transcriptomic effects of MIR155 upregulation may be cell‐type specific. We found that MIR155 overexpression resulted specifically in a significant decrease in the percentage of SST⁺‐ cortical GABAergic interneurons (Figure 3H,I). These data are consistent with those derived from systems containing the entire triplication ⁵² , ⁵³ and therefore identify miR155 as a causative gene in DS. GABA receptor genes (GABRA1, GABRA4, and GABRB2) and NR2F2 are included in the list of mRNA targets of MIR155 (Figure 4A) according to three independent prediction tools. ⁴⁵ , ⁴⁶ , ⁴⁷ Gene expression analysis by qPCR showed that MIR155 overexpression induced significant changes in the expression of its mRNA targets, downregulating GABRA1 and GABRA4, but upregulating NR2F2 and NR2F family homolog NR2F1 (Figure 4B). Collectively, our studies showed downregulation of NR2F1 and NR2F2 expressions in MIR155‐deleted hiPSC‐derived cortical neurons (Figure 3G) and upregulation of NR2F1 and NR2F2 expression in MIR155‐overexpressing NSCs (Figure S7C), in MIR155‐overexpressing cortical neurons (Figure S7E), and in MIR155‐overexpressing GABAergic interneurons (Figure 4B). During the transition from NSCs (WT, control, or MIR155‐overexpressing) to cortical neurons, MIR155‐5P expression dropped dramatically (Figure 2B and Figure S2F), and the expression of NR2F1 and NR2F2 decreased synchronously (Figures S6A–D and S7F–J and Figure 4C). A bioinformatics screen using miRBASE identified 106 target genes with the ARE (AU‐rich element) motifs in the highly conserved 3’UTRs of their mRNAs complementary to the seed regions of miR155, including NR2F1 and NR2F2 (Figure 4D,E). These results support a critical role of MIR155 in positively regulating the expression of NR2F1 and NR2F2, possibly by stabilizing their mRNAs through MIR155 binding with the AREs in their 3’UTRs. ⁵⁴ In summary, these results point to MIR155 as an important regulator of the development of GABAergic interneurons or their specific subtypes and of NR2F1 and NR2F2 expressions.

FIGURE 4.

miR155 positively regulates expression of its mRNA targets, NR2F1 and NR2F2. (A) Canonical miRNA complementary sites. The seed region of miR155 is marked in red, and 7‐nt (7mer‐A1 and 7mer‐m8) and 8‐nt (8mer) in 3’UTRs of target mRNAs, which match the seed region of miR155, are marked in blue. The bottom panel shows the conserved sites in 3’UTRs of the mRNA targets of MIR155, NR2F2, GABRA1, GABRA4, and GABRB2. (B) The expression levels of selected genes in control and MIR155‐overexpressing GABAergic interneurons were determined by qPCR. In the qPCR histogram, the y‐axis indicates relative gene expression level normalized to control. Data represent mean. Error bars indicate standard deviation. Each data point represents nine technical replicates from three independent experiments; *p < 0.05, **p < 0.01; t‐test. (C) Dynamic expression profiles of NR2F1 and NR2F2 during the transition from NSCs (WT, control, or MIR155‐overexpressing) to cortical neurons are illustrated in chart. ***Adjusted p value < 0.001. (D) Intersections of downregulated (green) DEGs derived from comparison of MIR155‐deleted (Clone 5 or 25) versus WT human cortical neurons, the upregulated (red) DEGs derived from the comparison of MIR155‐overexpressing versus control NSCs and predicted mRNA targets with AREs in their 3’UTRs are illustrated by a Venn diagram. (E) The AREs and ARE motifs (AUUA and AUUUA) in the 3’UTRs of mRNAs of human NR2F1 and NR2F2 are marked in green. The seed sequences (complementary to ARE motifs) in MIR‐155‐5P are marked in red.

3.4. MIR155 deletion enhanced the ventral patterning and the induction of GABAergic interneurons in 3D cortical organoids

We next determined whether the phenotypes of MIR155 deletion in 2D hiPSC‐derived cell cultures could be recapitulated in more complex 3D human cortical organoids, prepared using a modified version of an established protocol (Figure 5A). ³⁴ We have demonstrated that cortical organoids recapitulate regional organization and cortical cell types present in the developing human brain, when compared to standard cerebral organoids. ⁵⁵ , ⁵⁶ Cortical organoids at early stages (around 10 days) formed an abundance of large rosette‐like neuroepithelia, which expressed SOX1, PAX6, SOX2, and N‐cadherin (Figure 5B,D) and surrounded a fluid‐filled cavity resembling a ventricle with characteristic apical localization of neuron specific N‐cadherin expression (also called neural rosette, Figure 5D). Compared to the organoids derived from WT hiPSCs, MIR155KO organoids contained many more PAX6‐ and SOX1‐positive NSCs (Figure 5B,C), and the density of SOX2‐ and N‐cadherin‐positive neural rosettes was significantly higher (Figure 5D,E). The size of neural rosettes in MIR155‐deleted hiPSC‐derived organoids was slightly smaller (Figure 5D,F).

The organoids ultimately developed into concentric multilayer structures composed of neural progenitor cells (NPCs) and lower (CTIP2⁺) and upper (SATB2⁺) cortical layer neurons (data not shown). At 6 weeks, the number of ventral NPCs (NKX2‐1⁺) in MIR155‐deleted hiPSC‐derived organoids had increased significantly compared to WT organoids (Figure 5G,I), consistent with the MIR155‐deletion phenotype observed in 2D hiPSC‐derived NSC cultures (Figure 3C). Given that NKX2‐1 is a marker of ventral prosencephalic progenitor populations, ⁵⁷ our results suggest that MIR155 works as a repressor of ventral patterning during brain development. The NKX2‐1 domain in the human embryonic ventral forebrain is further subdivided into an OLIG2‐negative preoptic‐area anlage and an OLIG2⁺ ganglionic eminence where cortical and hippocampal GABAergic interneurons arise. ²⁰ , ³³ Thus, to confirm a link between MIR155 and GABAergic interneuron development, we quantified the number of OLIG2⁺ progenitors and found that nearly 40% of the NKX2‐1‐positive cells in the organoids co‐expressed OLIG2 (Figure 5G,H). Furthermore, MIR155‐deleted organoids contained significantly more GABA⁺/TUJ1⁺ interneurons than WT organoids (Figure 5J,K). In a more mature stage (day 66 of differentiation), cortical organoids derived from WT hiPSCs contained few SST⁺‐ GABAergic interneuron subtypes; however, the number of SST⁺‐ GABAergic interneurons increased significantly in MIR155‐deleted hiPSC‐derived organoids (Figure 5L,M). PV⁺‐GABAergic interneurons were not detected, likely because of their later development during postnatal and adolescent stages, a time frame difficult to capture in hiPSC‐derived models. ²⁰ , ²¹ , ⁵⁸ Taken together, these results indicate that MIR155 deletion enhanced the ventral patterning and the induction of GABAergic interneurons in cortical organoids.

3.5. Identification of mRNA targets of MIR155 in NSCs and cortical neurons

To identify additional potential targets of MIR155 and a possible link between gene expression changes in its mRNA targets and developmental regulation of NSCs and GABAergic interneurons, we used three computational miRNA target prediction tools ⁴⁵ , ⁴⁶ , ⁴⁷ and combined the results with a published database of experimentally validated mRNA targets of miR155 ⁴⁸ (Figure S8A). Intersection of the gene lists from multiple prediction tools yielded 14 targets in hiPSC‐derived NSCs (Figure S8B–D), including GABRA1, TBR1 (a genetic determinant for NSC differentiation ⁵⁹ ), FGF14 (a regulator of GABAergic inhibitory synaptic transmission ⁶⁰ ), and 69 targets in cortical neurons (Figure S8E–H), including NR2F2, SOX1, a NSC marker, ⁴⁴ SATB2, and PROX1, a developmental determinant for hippocampal NSCs. ⁶¹ Additionally, we highlighted a short list of putative candidate genes for further investigation (Figure S8I,J). A significant proportion of well‐studied mRNA targets of MIR155 was downregulated in MIR155‐deficient hiPSC‐derived NSCs and cortical neurons (e.g., NR2F2), indicating the involvement of other transcriptional regulators in the expression of DEGs identified in our RNA sequencing datasets. By performing ChIP Enrichment Analysis (ChEA), we observed an involvement of epigenetic regulators in the expression of DEGs, including the SUZ12‐JARID2‐EZH2 complex (Figure S8K–N), where JARID2 is a verified mRNA target of miR15. ⁶² Loss‐of‐function mutations in JARID2 gene cause a neurodevelopmental syndrome, characterized by developmental delay, cognitive impairment, and autistic features. ⁶³ We detected dynamic changes in the expression of several validated mRNA targets of MIR155, including MEIS1, CEBPB, and CLDN1. These findings indicate that MIR155‐mRNA targets play key roles during the differentiation of NSCs into cortical neurons (Figures S6A–D, S7F–J, and S8O–P).

3.6. Deletion of miR155 induced the expansion of RGL‐NSCs in the SGZ and the hilus and their ectopic localization in the granule cell layer

Based on our work demonstrating mitigation of parts of the phenotype in APP^KM670/671NL/PSEN1^Δexon9 (APP/PS1) ²⁹ double transgenic mice, ⁹ , ¹⁰ we hypothesized that miR155 may have similar functions in adult mouse hippocampal NSCs to those in hiPSC‐derived NSCs. Using a miRNAscope assay, we observed a significant enrichment of miR155 expression in NSCs in the DG and glutamatergic neurons and/or GABAergic interneurons in CA1 (Figure 6A). In WT mice, the number of Ki67‐ and Dcx (Doublecortin)‐positive cells in the DG declined from 1 to 7 months. ⁶⁴ In APP/PS1 mice, hippocampal amyloidosis becomes detectable by 4 months. ⁹ , ¹⁰ , ²⁹ We assessed whether miR155 deletion influenced the proliferation of RGL‐NSCs. Consistent with in vitro results, APP/PS1‐miR155KO mice showed an increase in Sox1‐positive NSCs in both the SGZ and the hilus (Figure 6B,C) compared to APP/PS1 mice expressing miR155. In 4‐month‐old APP/PS1 mice, miR155 deletion induced an increase in the number of Sox1‐ and Gfap‐positive RGL‐NSCs, which lined up along the SGZ and were recognized by their characteristic bipolar or unipolar processes ⁶⁵ (Figure 6B,D). However, there was no change in the overall number of Dcx‐positive immature neurons in the SGZ and GCL of the DG in hippocampi in 4‐month‐old APP/PS1‐miR155KO mice compared with APP/PS1 mice (Figure S9A–C).

FIGURE 6.

miR155 deletion induced the expansion of RGL‐NSCs in SGZ and hilus and their ectopic localization in GCL. (A) miRNAscope of miR155 (red) combined with a nuclear counterstain (DAPI) showing miR155 expression in hippocampal DG and CA1 in 4‐month‐old APP/PS1 mice. Scale bar, 20 µm. (B) Representative images of miR155 deletion‐induced expansion in Gfap (green)‐ and Sox1 (red)‐positive RGL‐NSCs in the SGZ and hilus of APP/PS1 mice at 4 months of age. Scale bar, 50 µm. (C) Quantification of total Sox1‐positive cells in the SGZ and hilus of APP/PS1 and APP/PS1‐miR155KO mice at 4 months of age. (D) Quantification of Sox1‐ and Gfap‐positive cells in the SGZ of 4‐month‐old APP/PS1 and APP/PS1‐miR155KO mice. (E) Representative images of Gfap (green)‐ and Sox2 (red)‐positive RGL‐NSCs in the hippocampal GCL of APP/PS1 and APP/PS1‐miR155KO mice at 4 months of age showing disruption of miR155‐induced their ectopic localization in the GCL (marked with white curves). Scale bars, 50 µm. (F) Quantification of ectopically located Gfap‐ and Sox2‐positive cells in the GCL of 4‐month‐old APP/PS1 and APP/PS1‐miR155KO mice. The total numbers of positive cells in the SGZ, the SGZ and hilus, or the GCL were counted in sections from nine different animals (five female and four male) per group and normalized by the area size. Data presented as means ± SD; ***p < 0.001; t‐test.

Given our in vitro observation that MIR155 regulates genes associated with focal adhesion, extracellular matrix, and cytoskeleton (Figures S1–S7), we next investigated whether miR155 deletion would lead to abnormal migration of hippocampal NSCs. While RGL‐NSCs were present in an organized fashion along the SGZ in 4‐month‐old APP/PS1 mice, APP/PS1‐miR155KO mice showed more Gfap⁺/Sox2⁺ RGL‐NSCs penetrating the GCL, with a two‐fold increase in Gfap⁺/Sox2⁺ RGL‐NSCs ectopically located in the GCL (Figure 6E,F). These results suggest that in APP/PS1 mice, miR155 deletion induces the expansion of RGL‐NSCs in the SGZ and the hilus as well as their migration into the GCL.

3.7. Deletion of miR155 led to an increase in GABAergic interneuron number in the hippocampi of 8‐month‐old APP/PS1 mice

Glutamatergic granule cells are the sole neuronal subtype generated from a pool of adult NSCs in the DG of the mature mammalian hippocampus. ¹⁷ The integration of these new neurons into functional hippocampal circuits contributes to learning and memory. ¹⁷ Our in vitro studies with hiPSC cultures indicated that MIR155 deletion caused induction of GABAergic interneurons (Figures 3 and 5). Using NeuN and GABA immunofluorescence, we found that in 8‐month‐old APP/PS1 mice, miR155 deletion increased the number of GABA‐positive cells in the hippocampal DG, without changing the total number of GABA‐positive cells in the whole hippocampus (Figure 7A,B and Figure S9D,E). Given their ability to control microcircuits and reliably fire at high frequencies, PV⁺ neurons are at the intersection of network synchrony, network oscillations, and memory processing. ⁵⁸ In AD patients and mouse models, impairments in PV⁺ interneurons have been linked to alterations in network hypersynchrony, gamma and theta oscillations, and cognitive deficits. ²⁴ , ⁶⁶ , ⁶⁷ In 8‐month‐old APP/PS1 mice, miR155 deletion caused a significant increase in PV⁺ cells in the hippocampi, most prominently in the CA1 region, with no change in the number of Sst⁺ cells (Figure 7C,D and Figure S9F). These results align well with our previous findings that constitutive absence of miR155 prevented impaired performance in Barnes maze learning behavior, ⁹ , ¹⁰ implying that PV⁺ interneurons are primarily responsible for the beneficial effect. MIR155 positively regulates its mRNA targets, NR2F1 and NR2F2, in vitro (Figure 4), and we found that in APP/PS1 mice, miR155 deletion significantly downregulated the expression of Nr2f1 (Figure 7E–H), further confirming miR155 as a positive regulator of expression of Nr2f1.

FIGURE 7.

miR155 deletion caused a significant increase in GABAergic interneurons in hippocampi of 8‐month‐old APP/PS1 mice. (A and B) Representative images of hippocampal sections of APP/PS1 and APP/PS1‐miR155KO mice at 8 months of age double‐immunolabeled for NeuN (green) and GABA (red), along with nuclear counterstaining with DAPI (A). Scale bar, 50 µm. (B) Stereological quantification of GABA‐immunoreactive cell number in hippocampal DG (marked with white curves) of 8‐month‐old APP/PS1 and APP/PS1‐miR155KO mice (n = 9 mice/group); ***p < 0.001; t‐test. (C and D) Representative images of hippocampal sections of APP/PS1 and APP/PS1‐miR155KO mice at 8 months of age double‐immunolabeled for NeuN (green) and parvalbumin (PV) (red, C), along with nuclear counterstaining with DAPI. Scale bar, 100 µm. The images (DG and CA1) in the lower right are higher‐magnification images from the boxed area on the left. (D) Stereological quantification of PV⁺‐ or Sst⁺‐ (Figure S9F) immunoreactive cell numbers in hippocampi of 8‐month‐old APP/PS1 and APP/PS1‐miR155KO mice (n = 9 mice/group); ***p < 0.001; t‐test. (E and F) Representative images of brain sections of 8‐month‐old APP/PS1 and APP/PS1‐miR155KO mice double‐immunolabeled for NeuN (green) and Nr2f1 (red, E), along with nuclear counterstaining with DAPI. Scale bar, 100 µm. The two images (DG and CA1) in the lower right are higher‐magnification images from the boxed area on the left. The neuronal expression of Nr2f1 in cortex is shown in the higher‐magnification images to the right. (F) Quantification of integrated density of Nr2f1 immunofluorescence in brains of 8‐month‐old APP/PS1 and APP/PS1‐miR155KO mice (n = 9 mice/group); ***p < 0.001; t‐test. (G) Representative Western blots for Nr2f1 and Gapdh using lysates of the prefrontal cortex of 8‐month‐old APP/PS1 and APP/PS1‐miR155KO mice. (H) The ratio of Nr2f1 to Gapdh from (G) showing a significantly lower Nr2f1 protein level in APP/PS1‐miR155KO than in APP/PS1 mice. Data are reported as the mean; n = 6 mice/group. Error bars indicate standard deviation; ***p < 0.001; t‐test.

4. DISCUSSION

In this study, we utilized an integrated analysis of published epigenetic datasets, in vitro hiPSC technology, and in vivo mouse models of AD pathology to investigate the neuronal expression and function of MIR155. We identified a novel, cell‐type‐specific role for miR155 in GABAergic interneuron development within the hippocampus of AD model mice, uniquely demonstrating a well‐established link of the DS gene, MIR155HG, to APP and AD. ⁶⁸ We demonstrated that expression of miR155 is dynamically regulated during the transition of NSCs into terminally differentiated cortical neurons, with a sharp increase in NSCs followed by a decline in mature cortical neurons. These findings are consistent with the notion that expression of miR155 is developmentally and cell cycle‐regulated in dividing cells, as shown in other systems. ⁶⁹ Importantly, increased miR155 expression has been observed in hippocampal neurons in tissues from patients with AD ⁸ , ¹⁰ and DS ¹¹ or traumatic brain injury ⁷⁰ and is now also supported by the results of ATAC‐seq, H3K4me3 ChIP‐seq, and H3K27ac ChIP‐seq on neuronal nuclei isolated from frozen post mortem human brains. We conclude that miR155 dysregulation in hippocampal NSCs and GABAergic interneurons contributes to AD pathology through mechanisms independent of miR155 actions in microglia. ⁷ Notably, the AD‐associated cell‐type‐specific changes in miR155 exhibit opposing effects, with the role of microglial miR155 contrasting with its function in NSCs and GABAergic interneurons.

In both hiPSC‐derived and mouse hippocampal NSCs in APP/PS1 mice, ²⁹ miR155 deletion enhanced NSC proliferation, whereas its overexpression downregulated NSC markers. A proliferation deficit has been consistently observed in NSCs differentiated from hiPSCs derived from DS patients, ⁴⁹ implying that miR155 is pathogenic for the characteristic Trisomy 21 phenotype. Our results are also supported by a study showing that disruption of miR155 in mice leads to reversal of inflammation‐induced decrease in NSC proliferation and neural differentiation in the DG and that nestin‐specific elevation of miR155 reduced immature neuron survival and induced ectopic localization of RGL‐NSCs in the DG. ⁷¹ In MIR155‐deleted hiPSC‐derived NSC cultures, we observed an induction of GABAergic synaptic gene expression. This induction was accompanied by the expression of key transcription factors involved in GABAergic interneuron development, including NKX2‐1, LHX6, DLX1, DLX2, OLIG1, and OLIG2. ²⁰ , ²¹ In cortical organoids, MIR155 deletion induced a significant increase in the expression of the transcription factor NKX2‐1, a molecular determinant of ventral brain development. ⁵⁷ Furthermore, MIR155 deletion significantly enhanced the generation of GABAergic interneurons in the cortical organoids. These results indicate that MIR155 acts as a repressive regulator of ventral patterning and of GABAergic interneuron induction during brain development, contrary to the actions of Sonic Hedgehog (Shh). ⁵⁰ Existing well‐validated, commercially available miR155 inhibitors may be promising tools to induce the ventral patterning and the generation of GABAergic interneurons in brain organoids.

In hiPSC‐derived cortical neurons, MIR155 deletion enhanced neurite growth, caused a robust induction of GABAergic interneurons, and downregulated NR2F1 and NR2F. ²⁰ , ²¹ Double knockdown of NR2F1 and NR2F2 in the neurosphere and the developing mouse forebrain caused sustained neurogenesis and the prolonged generation of early‐born neurons. ⁷² Moreover, loss of NR2F1 caused an imbalance of excitatory/inhibitory neuron differentiation (overproduced GABAergic inhibitory interneurons and underproduced glutamatergic excitatory neurons) ⁷³ and altered percentages of different cortical interneuron subtypes. ⁷² All these indicate an underlying link between MIR155 deletion‐induced neuronal phenotypes and downregulation of NR2F1 and NR2F2. In brain development, NR2F1 is crucial for establishing cortical patterning and hippocampal development, ⁷² and NR2F2 plays a vital role in the development of the amygdala and hippocampus. ⁷² Our results further demonstrate that the expression of both NR2F1 and NR2F2 is positively regulated by MIR155. Besides miR155, only very few miRNAs have been implicated in positively regulating the expression of their target mRNAs under specific cellular conditions. ⁵⁴ , ⁷⁴ , ⁷⁵ Thus, our studies and others underscore the importance of the miR155‐NR2F1/NR2F2 pathway in the development and function of GABAergic interneurons.

Our observations in cortical organoids and in APP/PS1 mice that miR155 deletion enhanced the generation of GABAergic interneurons, especially the generation of PV⁺‐ hippocampal GABAergic interneurons, is the first demonstration of a functional role for miR155 in the development and function of GABAergic interneurons. PV⁺‐ interneurons control network synchrony and the generation of functional oscillations, ⁵⁸ and their impairments can cause network abnormalities and memory deficits in AD patients and in mouse models of AD. ²⁴ Over the past decade, many studies have explored the restoration of PV⁺ interneuron function to address AD pathology and cognitive impairment in AD mouse models. ²⁴ , ⁶⁶ , ⁶⁷ These studies used various approaches, including chemogenetics, optogenetics, genetic manipulations, and interneuron transplants. We reported that constitutive absence of miR155 in APP/PS1 mice prevented impaired learning behavior performance in the Barnes maze test, suggesting a beneficial effect of miR155 deletion on learning behavior in the presence of early‐onset AD APP and PSEN1 mutations and amyloidopathy. ⁹ , ¹⁰ In the same study, field electrophysiology in hippocampal slices showed that miR155 deletion partially rescued synaptic plasticity in 10‐month‐old APP/PS1 mice despite inducing synaptic dysfunction in WT mice. ¹⁰ The present study supports the notion that using miR155 inhibitors to improve PV⁺ interneuron function is a promising therapeutic strategy for network dysfunction in AD. This pathway may be particularly relevant to AD in DS patients, in whom MIR155 upregulation is genetically programmed, and defects in the generation and maturation of GABAergic interneurons has been demonstrated in both differentiated hiPSCs derived from DS patients ⁵² and post mortem brain tissues from elderly DS patients. ⁵³ Taken together, our findings significantly extend the classical paradigm for the cell‐type‐specific role of miR155 in the pathogenesis of AD and other neurodegenerative diseases, in which miR155 has, until now, been studied solely as a master regulator of microglia and neuroinflammation.

AUTHOR CONTRIBUTIONS

X.D.Z. and M.E.E. conceived the project, designed experiments, and developed experimental protocols, tools, and reagents. S.N., S.G., and M.E.E. supervised the study. J.V.H.M. and A.A. prepared the mouse brain sections and performed immunostaining. M.B. performed components of the computational analysis. P.F.D. performed ATAC sequencing and Hi‐C sequencing data analysis. I.K. carried out electrophysiological recordings and analysis. V.F. provided support on hiPSC‐related experiments. A.Q.L. provided NSC lines derived from NIMH cohorts (553 and 2607) and technical support. X.D.Z. performed all the remaining experiments and computational analyses. X.D.Z. wrote the first draft of the manuscript. J.V.H.M., V.F., S.N., S.G., and M.E.E. reviewed and edited the manuscript. All authors read and approved the paper.

CONFLICT OF INTEREST STATEMENT

Dr. Sam Gandy is a co‐founder of Recuerdo Pharmaceuticals. He has served as a consultant in the past for J&J, Diagenic, and Pfizer, and he currently consults for Cognito Therapeutics, GLG Group, SVB Securities, Guidepoint, Third Bridge, MEDACORP, Altpep, Vigil Neurosciences, Eisai, Memory Garden, the Bell Law Firm, Alzheon, Tolion Brain Health, and Rylands Garth Legal Services. He has received research support from Warner‐Lambert, Pfizer, Baxter, and Avid. The other authors have nothing to disclose. Author disclosures are available in the Supporting Information.

CONSENT STATEMENT

No human subjects were included in the study, and the consent was not necessary.

Supporting information

Supporting Information: alz71684‐sup‐0001‐SuppMat.pdf

Supporting Information: alz71684‐sup‐0002‐ICMJE.pdf