Cell-specific expression of APP and GABA_BR1 isoforms in wild-type and the APP NL-G-F mouse model of Alzheimer’s disease

Abstract

Amyloid precursor protein (APP) is a type I transmembrane protein that undergoes proteolytic processing to generate amyloid-β, the main component of amyloid plaques found in brains with Alzheimer’s disease (AD). APP proteolytic processing also generates soluble APP alpha (sAPPα) which modulates synaptic transmission and neurite outgrowth through the γ-aminobutyric acid type B receptor (GABA_BR). sAPPα specifically binds the 1a isoform of GABA_BR subunit 1 (R1a) which contains two sushi domains absent in R1b. Here, we define the cell type-specific expression patterns of APP and GABA_BR1a in the brain and assess their alterations in the APP NL-G-F model of AD. We determined relative RNA levels of R1a and R1b isoforms in oligodendrocytes, microglia, endothelial cells, astrocytes, and neurons in adult mice using two approaches. We developed an isoform-specific RNAseq analysis workflow to probe publicly available datasets. We also isolated these cell types from a single mouse brain and performed RT-qPCR with isoform-specific primers. APP and GABA_BR1 isoforms were differentially expressed among cell types. APP and GABA_BR1a expression were highest in oligodendrocytes. In contrast, GABA_BR1b expression was highest in astrocytes. While microglia expressed relatively low levels of GABA_BR, they showed the greatest change in GABA_BR expression in the APP NL-G-F mouse model, with alterations that were both sex- and age-specific. Together, our findings identify candidate cell types in which future studies may investigate the potential role of APP-GABA_BR signaling in health and AD.

Graphical Abstract

1. Introduction

The amyloid precursor protein (APP) is a type I transmembrane protein expressed in the brain as well as other organs (van der Kant & Goldstein, 2015). APP was discovered in 1987 during a search to uncover the source of the amyloid beta peptide (Aβ) (Goldgaber et al., 1987; Kang et al., 1987; Tanzi et al., 1987), the main component of the amyloid plaques associated with Alzheimer’s disease (AD) (Delport & Hewer, 2022). APP undergoes proteolytic processing to release not only Aβ but also other fragments which have important roles in the brain. APP can be cleaved by either α- or β- secretase followed by γ-secretase. α-secretase cleaves within the Aβ region and thus precludes Aβ generation and produces soluble APP α (sAPPα), P3, and the APP intracellular domain (AICD). β-secretase initiates the amyloidogenic pathway which generates sAPPβ, Aβ, and AICD (Haass et al., 2012). APP can also be cleaved by η-secretase prior to α or β cleavage within the sAPP region to generate sAPPη and additional fragments Aη-α or Aη-β (Hefter et al., 2020). Full-length APP and its fragments are involved in crucial brain functions (Müller et al., 2017). They modulate neurodevelopmental processes such as axonal outgrowth, dendritic arborization, neurogenesis, and neuronal migration (Chau et al., 2023; Müller et al., 2017). In mature neurons, full-length APP and sAPPα regulate synaptic structure, plasticity, and neurotransmission (Hefter et al., 2020; Kruse et al., 2025). Though best known for its neuronal roles, APP is also implicated in glial and endothelial functions(Dunot et al., 2023). APP knockout (KO) mice display hypomyelinated axons (Truong et al., 2019), blunted microglial and astrocytic inflammatory responses (Carrano & Das, 2015), mitochondrial fragmentation in astrocytes (Montagna et al., 2019), and reduced production of endothelial nitric oxide species (d’Uscio et al., 2018). In human stem cell-derived microglia, APP knockdown impaired the function of voltage-gated proton channels implicated in microglial activation (Zhao et al., 2025). Despite these observations, the role of APP in glial cells remains largely undefined and understudied.

Identifying proteins that interact with APP and its fragments has been a key strategy to elucidate its physiological functions (Müller et al., 2017). In 2019, full-length APP (Dinamarca et al., 2019) and its secreted ectodomain sAPPα (Rice et al., 2019) were shown to interact with the γ-amino butyric acid (GABA) type B receptor (GABA_BR). GABA_BR is a class C G-protein coupled receptor (GPCR) and is an obligatory heterodimer (Terunuma, 2018). The GABA_BR1 subunit contains the endogenous agonist binding site, and GABA_BR2 is intracellularly linked to G-proteins. The GABA_BR1 subunit has various isoforms with R1a and R1b being the main two variants responsible for GABA_BR diversity in humans (Terunuma, 2018). The main difference between the two isoforms is the presence of two sushi domains on the N-terminus of R1a but not in R1b due to differential promoter usage (Huang, 2006). These sushi domains are responsible for the preferential localization of the GABA_BR1a subunit to axon terminals (Biermann et al., 2010; Vigot et al., 2006). The intact (full-length APP) (Dinamarca et al., 2019) and shed (sAPPα) ectodomain (Rice et al., 2019) both bind at the N-terminal sushi domain 1 of GABA_BR1a (Rice et al., 2019). The sAPP-GABA_BR interaction was shown to reduce synaptic vesicle release consequently modulating hippocampal synaptic plasticity and neurotransmission (Rice et al., 2019) as well as neurite outgrowth (Barber et al., 2026). Moreover, deletion of APP impaired GABA_BR-mediated presynaptic inhibition and axonal GABA_BR levels (Dinamarca et al., 2019). The downstream mechanism mediating the effects of sAPPα-GABA_BR signaling remains under investigation as studies suggest it may deviate from canonical GABA_BR signaling (Rem et al., 2023).

In the brain, both GABA_BR1a and GABA_BR1b are expressed in most neuronal subpopulations including glutamatergic, dopaminergic, and GABAergic neurons (Bettler et al., 2004). Neurons utilize canonical GABA_BR signaling pathways to regulate key processes like synaptic transmission and excitatory/inhibitory balance (Terunuma, 2018). GABA_BRs couple to Gα_i/o and Gβγ proteins to inhibit adenylyl cyclase and presynaptic Ca2+ channels to reduce presynaptic release probability and activate postsynaptic GIRK channels to reduce neuronal excitability (Bettler et al., 2004; Syrovatkina et al., 2016; Terunuma, 2018).The inhibitory functions of GABA_BR have been extensively studied in neurons but are not as well characterized in glial cells. Oligodendrocytes express functional GABA_BR1 that can modulate adenylyl cyclase signaling (Luyt et al., 2007). Furthermore, GABA_BR activation using Baclofen, an agonist with no isoform selectivity, regulates oligodendrocyte precursor cell differentiation in vitro (Serrano-Regal, Luengas-Escuza, et al., 2020) and can accelerate remyelination after induced primary demyelination in mouse spinal cord (Serrano-Regal et al., 2022). Studies in astrocytes and microglia show that these glial cells also express GABA_BR1 and its activation using baclofen resulted in a decreased inflammatory response to lipopolysaccharide (LPS) (Charles et al., 2003; Kuhn et al., 2003; Lee et al., 2011). A study using Baclofen in endothelial cells suggests that they also express functional GABA_BRs which can modulate their response to homocysteine toxicity (Shastry et al., 2006). Although some of these studies detected both isoforms of GABA_BR1, it remains unclear whether glial cells exhibit preferential expression of a specific isoform.

GABA_BR levels are altered during AD. In the human brain, GABA_BR1 levels transiently increase in the hippocampus in a Braak-staging-dependent manner suggesting that this change may be compensatory (Iwakiri et al., 2005). Furthermore, studies assessing GABA_BR levels in the APP/PS1 mouse model of AD showed that all subunit levels are reduced in the hippocampus at ages coinciding with cognitive impairment in this model (Iwakiri et al., 2005; Martín-Belmonte, Aguado, Alfaro-Ruíz, Moreno-Martínez, de la Ossa, Martínez-Hernández, Buisson, Früh, et al., 2020; Martín-Belmonte, Aguado, Alfaro-Ruíz, Moreno-Martínez, de la Ossa, Martínez-Hernández, Buisson, Shigemoto, et al., 2020). However, whether these changes are occurring in a specific cell type and whether they are isoform-specific is not yet clear.

In this study, we provide a comprehensive cell type-specific expression profile of APP across oligodendrocytes, microglia, endothelial cells, astrocytes, and neurons sorted from the same mouse brain (Houmam et al., 2025). We further assessed isoform-specific GABA_BR RNA expression in these five cell types using RT-qPCR under baseline conditions as well as in the APP NL-G-F knock-in mouse model of AD. Understanding cell type- and isoform- specific changes in GABA_BR expression across brain cell populations is essential for clarifying how GABA_BR signaling contributes to brain function and AD pathology.

2. Results

2.1. Oligodendrocytes show the highest levels of APP RNA

APP is broadly expressed throughout the brain, yet its relative expression across different cell types and the implications of this expression remain underexplored. We examined the relative RNA expression of APP (gene name App) in two publicly available datasets. First, a scRNAseq dataset generated by Allen et al. (2023) (Allen et al., 2023) in young (4-week-old) and old (90-week-old) mice. App was highly enriched in oligodendrocytes and endothelial cells, followed by neurons (Figure 1A). This was similar in both the young and old mice so data from both ages was combined. Second, we analyzed a publicly available bulk RNAseq dataset generated by Zhang et al. (Zhang et al., 2014) which isolated oligodendrocytes, microglia, endothelial cells, astrocytes, and neurons using a variety of cell-sorting and isolation methods. Consistent with the previous dataset, App was most highly expressed in oligodendrocytes followed by endothelial cells and neurons (Figure 1B). While this dataset included all cell types of interest, samples were prepared using different cell isolation methods and mouse genotypes. Therefore, we performed magnetic-activated cell sorting (MACS) to sequentially isolate oligodendrocytes, microglia, endothelial cells, astrocytes, and neurons from the same mouse brain (Houmam et al., 2025) and probed for RNA expression of APP using RT-qPCR. We used 5mo male and female wild-type (WT) mice. Primers were validated using age-matched APP KO mice (Supplemental figure 1A). The purity and viability of the sorted cell fractions were validated using flow cytometry and RT-qPCR, as reported in Houmam et al., 2025 (Houmam et al., 2025). Our results corroborated those shown in the publicly available RNAseq datasets (Allen et al., 2023; Zhang et al., 2014). App was strongly expressed in oligodendrocytes in both males (1C) and females (1D) followed by endothelial cells and neurons (Figure 1C and D). These results identify oligodendrocytes as a cell type with particularly high APP transcript expression.

Figure 1: APP RNA expression levels in oligodendrocytes, microglia, endothelial cells, astrocytes, and neurons.

(A) Uniform manifold approximation (UMAP) visualization of App expression across cell types generated with Seurat using GEO #GSE207848. B) DESeq2 normalized read counts for APP in oligodendrocytes, microglia, endothelial cells, astrocytes, and neurons using GEO #GSE52564 (Kruskal-Wallis P=0.0011; Dunn’s multiple comparisons: all ns due to small sample size, n=2 per group). (C) Relative APP RNA expression measured by RT-qPCR in sorted cells from male mice (Kruskal-Wallis P=0.0005; Dunn’s multiple comparisons: * = p≤0.05, *** = p≤0.001; n=4 per group). (D) Relative APP expression in sorted cells from female mice (Kruskal-Wallis P=0.0010; Dunn’s multiple comparisons as in B; n=4 per group). Data shown as median ± IQR.

2.2. Oligodendrocytes show the highest GABA_BR1a RNA levels

GABA_BR is an obligate heterodimer composed of subunit GABA_BR1 which contains the binding site for GABA, its endogenous agonist, and subunit GABA_BR2 which is linked intracellularly to G-proteins. Alternative promoter usage can result in two predominant isoforms of the GABA_BR1 subunit referred to as GABA_BR1a and GABA_BR1b. The GABA_BR1a isoform contains two N-terminal sushi repeats not present in subunit GABA_BR1b (Figure 2A and B). Sushi domain 1 is required for APP-GABA_BR binding and therefore this interaction requires the GABA_BR1a isoform to function(Dinamarca et al., 2019; Rice et al., 2019). While both GABA_BR1 and GABA_BR2 subunits have been shown to be expressed in neurons and glial cells, whether these cells express both the GABA_BR1a and 1b isoform equally is still unclear. Therefore, we sought to determine the isoform-specific expression patterns of GABA_BR across cell types.

Figure 2: GABA_BR isoform-specific RNA expression levels in oligodendrocytes,microglia, endothelial cells, astrocytes, and neurons.

(A) Schematic of GABA_BR subunits and isoforms. The GABA_BR1a isoform contains sushi domains (blue) absent in GABA_BR1b (purple). (B) Schematic of GABA_BR1a and GABA_BR1b transcripts: GABA_BR1a contains five unique exons (blue), while GABA_BR1b contains only a short unique sequence at the start of exon 1 (purple). (C–F) DESeq2 normalized read counts from GEO #GSE52564 for GABA_BR1 (C, P=0.0011), GABA_BR2 (D, P=0.0053), GABA_BR1a (E, P=0.0011), and GABA_BR1b (F, P=0.0011) across oligodendrocytes, microglia, endothelial cells, astrocytes, and neurons (Kruskal-Wallis. Dunn’s multiple comparisons all ns due to small sample size, n=2). (G–N) Relative RNA expression measured by RT-qPCR in sorted oligodendrocytes, microglia, endothelial cells, astrocytes, and neurons from male (G–J) and female (K–N) 5mo WT mice. Statistical significance was determined by Kruskal–Wallis test (Males: Gabbr1, P=0.0007; Gabbr2, P=0.0005; 1a, P=0.0013; 1b, P=0.0007; Females: Gabbr1, P=0.0006; Gabbr2, P=0.0028; 1a, P=0.0031; 1b, P=0.0004), followed by Dunn’s multiple comparisons (* = p ≤ 0.05; ** = p ≤ 0.01; *** = p ≤ 0.001). Data shown as median ± IQR; n=4 per group.

Using the publicly available dataset generated by Zhang et al. (Zhang et al., 2014), we probed for the RNA expression of total GABA_BR1 and GABA_BR2 and found that RNA levels of both subunits are highest in astrocytes followed by neurons and oligodendrocytes (Figure 2C and D). To be able to distinguish read counts for the GABA_BR1a and GABA_BR1b isoforms of the GABA_BR1 subunit, we applied an isoform-aware workflow by aligning read counts generated by the study to a manually edited genome file where sequences unique to each isoform transcript were labeled Gabbr1 (1a) and Gabbr1 (1b) (Figure 2B). We found that total GABA_BR1 RNA was highest in astrocytes. However, RNA levels of the GABA_BR1a subunit were highest in oligodendrocytes, followed by neurons, astrocytes, and microglia (Figure 2E). On the other hand, GABA_BR1b RNA was highest in astrocytes, followed by neurons and oligodendrocytes (Figure 2F).

The unique isoform-specific regions of GABA_BR1a and GABA_BR1b are short, GC rich, and located at the 5’ end of the transcripts (Figure 2B). Since these features result in low read counts in RNAseq (Jiang et al., 2011; Oshlack & Wakefield, 2009), which is typically 3’- biased and inefficient at detecting, short, GC-rich sequences, we also employed a more targeted approach. We performed RT-qPCR with commercially available primers to assess expression of total GABA_BR1 and GABA_BR2 and custom-designed primers to assess isoform-specific GABA_BR1a and GABA_BR1b levels (Figure 2B). We validated our custom primers (Table 1) using a GABA_BR1a-specific KO mouse model. The GABA_BR1a primer was specific to the R1a isoform, showing no amplification in GABA_BR1a KO samples (Supplemental figure 1B). As expected, primers targeting the 1b isoform showed no change in expression between the GABA_BR1a KO and WT (Supplemental figure 1C). Total GABA_BR1 RNA levels were highest in astrocytes followed by oligodendrocytes and neurons in both male (Figure 2G) and female (Figure 2K) mice. GABA_BR2 RNA levels were highest in astrocytes followed by neurons in both males (Figure 2H) and females (Figure 2L). Isoform-specific RT-qPCR showed that RNA levels of the GABA_BR1a subunit isoform were highest in oligodendrocytes followed by astrocytes and neurons in males (Figure 2I) and females (Figure 2M), and the GABA_BR1b was highest in astrocytes followed by oligodendrocytes and neurons in male (Figure 2J) and female mice (Figure 2N). The relative RNA levels of GABA_BR subunits and isoforms were consistent in cells isolated from mice aged from 3 to 12mo (Supplemental Figure 2). These findings suggest a cell-type bias, with oligodendrocytes exhibiting higher GABA_BR1a expression and astrocytes exhibiting higher GABA_BR1b expression. Moreover, our findings highlight oligodendrocytes as a cell type where both APP and GABA_BR1a RNA are most highly expressed in the brain.

Table 1:

Designed isoform-specific primer sequences.

Primer	Component	Primer sequence (5’-3’)	Nucleotide numbers	Location	Amplicon size
GABABR1a	Forward	TGGAGCCTGGATTCCTGTG	191-210	Exon 1	222bp
	Reverse	TGGCAACCTTCCGAGGTG	396-413	Exon 2-3
	Probe	TGCTGCTGGTGCCTCTCTTCCTC	335-357	Exon 2
GABABR1b	Forward	TGCCTCTTCTGCTGGTGATG	264-283	Exon 1	275bp
	Reverse	TGGGTCGCACTTGCTGTC	521-538	Exon 1-2
	Probe	CCTCCTCAGAACGGCGTGCAGT	360-381	Exon 1

2.3. APP KO does not affect GABA_BR1a transcript levels

We tested whether APP KO may affect transcript levels of GABA_BR1. Using APP KO mice (Supplemental figure 1A), we assessed overall levels of GABA_BR1 as well as GABA_BR1a and GABA_BR1b. APP KO did not affect levels of GABA_BR1a in any of the cell types (Supplemental Figure 1D). Intriguingly, GABA_BR1b was increased in APP KO astrocytes (Supplemental Figure 1E). This may reflect an astrocytic compensatory response to APP deletion, although the underlying mechanism remains unclear. We also tested whether GABA_BR1a KO may alter RNA levels of APP. APP RNA levels were unchanged in all five cell types in GABA_BR1a KO mice (Supplemental Figure 1F).

2.4. Microglia show sex- and age- specific alterations in GABA_BR levels in the APP NL-G-F mouse model of AD

Alterations in GABA_BR levels in both patients with AD and mouse models of AD have been reported (Iwakiri et al., 2005; Martín-Belmonte, Aguado, Alfaro-Ruíz, Moreno-Martínez, de la Ossa, Martínez-Hernández, Buisson, Früh, et al., 2020; Martín-Belmonte, Aguado, Alfaro-Ruíz, Moreno-Martínez, de la Ossa, Martínez-Hernández, Buisson, Shigemoto, et al., 2020; Salazar et al., 2021). However, it is unclear whether these changes can be attributed to a specific cell type, and whether both isoforms of GABA_BR1 are affected. We used the APP NL-G-F mouse model of AD to assess cell-specific changes in GABA_BR expression in response to amyloid pathology. The APP NL-G-F mice carry knock-in mutations in the endogenous App gene (Swedish, Arctic, Iberian) leading to elevated Aβ42 production, progressive plaque deposition detectable around 2-3mo and robust by 6mo, and cognitive decline by 6mo with maximum impairment at 12mo (Saito et al., 2014). We selected ages 3, 6, and 12mo to capture these milestones and used both male and female mice to detect sex-specific effects. We isolated oligodendrocytes, microglia, endothelial cells, astrocytes, and neurons from the same mouse brain (Houmam et al., 2025) and probed for RNA expression of GABA_BR using RT-qPCR with isoform-specific primers. The pre-sorted input samples show an overall decrease in total GABA_BR1, both GABA_BR1 isoforms, and GABA_BR2 only in male mice (Figure 3A). We observed the most pronounced change in microglia, with a 3-fold increase in GABA_BR1 expression at the 3mo age time-point specifically in female mice (3A–C). This change was not isoform-specific as both the R1a and R1b isoforms were increased (Figure 3A, 3C) and was not observed in male mice (Figure 3A, 3B). We probed for markers of homeostatic and disease-associated microglia (DAM), a subset that arises during neurodegeneration and other pathological conditions. Only female microglia showed a significant decrease in homeostatic markers in the NL-G-F mice starting at 3mo (Figure 4 A–B), whereas both male and female microglia showed a gradual increase in the expression of DAM markers in the NL-G-F mice with age (Figure 4 C–D). Oligodendrocytes showed no statistically significant changes (Figure 3A). There was a 0.5-fold decrease in GABA_BR2 expression in endothelial cells from male mice only (Figure 3A). On the other hand, astrocytes only show a transient 1.5-fold increase in GABA_BR1a expression at the 6mo age time point in female mice (Figure 3A). Lastly, neurons show a 1.5-fold increase in GABA_BR1a in female 12mo mice while in males, there is a 0.5-fold decrease at the 3mo time point in GABA_BR1b (Figure 3A). Microglia were therefore the cell type that showed the greatest magnitude of change in GABA_BR in the APP NL-G-F model.

Figure 3: Cell type-specific expression of GABA_BR subunits and isoforms in WT and APP NL-G-F mice.

(A) Heatmap showing RT-qPCR RNA expression levels of GABA_BR1, GABA_BR2, GABA_BR1a, and GABA_BR1b across five sorted cell types from WT and APP NL-G-F mice at 3,6, and 12mo from male and female mice. Scale corresponds to log2 fold change. Relative RNA expression of GABA_BR subunits and isoforms measured by RT-qPCR in male (B) and female (C) microglia isolated from WT and APP NL-G-F mice ages 3, 6, and 12mo. (Two-way ANOVA with factors genotype and age with Sidak’s multiple comparisons * = p<0.05, ** = p<0.01, *** = p<0.001, **** = p<0.0001, n=3 per group). Data shown as median ± IQR.

Figure 4: Homeostatic and DAM-associated marker RNA levels in microglia from WT and APP NL-G-F mice.

Relative RNA expression levels of homeostatic and DAM-associated markers in male (C,D) and female microglia (E,F). (Two-way ANOVA with factors genotype and age with Sidak’s multiple comparisons * = p<0.05, ** = p<0.01, *** = p<0.001, **** = p<0.0001, n=3 per group). Data shown as median ± IQR.

2.5. Cell-specific APP expression is largely maintained in the APP NL-G-F mouse model of AD

To determine whether cell-specific APP expression patterns are altered in the APP NL-G-F model, APP transcript levels were quantified in enriched cell populations isolated from 3-month-old wild-type and APP NL-G-F mice (Supplemental Figure 3). The relative distribution of APP expression across cell types was largely preserved in APP NL-G-F mice and remained consistent with that observed in wild-type animals, with oligodendrocytes exhibiting the highest APP expression (Supplemental Figure 3A). Notably, microglial APP expression was not significantly altered in APP NL-G-F mice despite the observed changes in GABA_BR1 expression, suggesting that the microglial GABA_BR1 phenotype is unlikely to be driven by changes in APP transcript abundance. APP expression appeared lower in oligodendrocytes from APP NL-G-F mice; however, this difference did not reach statistical significance (Supplemental Figure 3B).

3. Discussion

While APP expression has been documented in numerous brain cell types, the extent and functional implications of its expression across different non-neuronal cell types is still underexplored. Our results confirm that APP is expressed not only in neurons but also in oligodendrocytes, microglia, endothelial cells, and astrocytes. We show that APP is most highly expressed in oligodendrocytes, a cell type in which it remains largely understudied. APP and its cleavage fragments appear to modulate oligodendrocyte-led processes such as myelination and re-myelination of axons (Truong et al., 2019; Zota et al., 2024). Promoting sAPPα production can restore damaged myelin and protect against further demyelination by increasing myelin basic protein and mature oligodendrocyte levels in the cerebellum (Llufriu-Dabén et al., 2018). Oligodendrocytes may play a critical role in AD pathology, as myelin disruption has been identified as a feature of the disease (Papuc & Rejdak, 2018).Moreover, oligodendrocytes have recently emerged as a significant source of Aβ pathology in the brain (Hida et al., 2025; Ishii et al., 2024; Rajani et al., 2024; Skaper et al., 2009), further underlining their potential involvement in AD and supporting the biological relevance of the high APP transcript expression observed in oligodendrocytes in the present study. We also show that endothelial cells express the second-highest levels of APP among brain cell types. This aligns with earlier work showing that APP is expressed in endothelial cells where it contributes to vascular integrity and nitric oxide signaling (D’Uscio et al., 2016; d’Uscio et al., 2018; d’Uscio & Katusic, 2021).

Similar to APP, GABA_BR is well characterized in neurons but largely understudied in glial cells. While GABA_BR is known for its primary role in modulating synaptic inhibition in neurons, its expression is not limited to neurons (Kuhn et al., 2003; Lee et al., 2011; Luyt et al., 2007). We validate this by showing that overall expression levels of GABA_BR1 and GABA_BR2 in the brain are highest in astrocytes. However, when we assessed isoform-specific expression of the GABA_BR1 subunit, we found that RNA levels of the GABA_BR1a subunit are highest in oligodendrocytes while the GABA_BR1b subunit is highest in astrocytes.. The high expression of both APP and GABA_BR1a in oligodendrocytes suggests that APP-mediated GABABR1a signaling could be relevant in this cell type, although functional studies will be required to determine whether such interactions occur in oligodendrocytes. Importantly, functional GABA_BR expression has been demonstrated in oligodendrocyte lineage cells and is regulated during OPC differentiation to myelinating oligodendrocytes (Serrano-Regal, Bayón-Cordero, et al., 2020; Serrano-Regal, Luengas-Escuza, et al., 2020). Furthermore, GABA_BR1 activation using Baclofen is reported to promote remyelination in the brain (Serrano-Regal et al., 2022). These findings suggest that the high GABA_BR transcript levels observed in oligodendrocytes are likely associated with biologically meaningful receptor expression and function. It remains unclear whether these GABA_BR-mediated oligodendrocyte functions are through one or both isoforms, given that Baclofen is not an isoform-specific agonist. Our expression data indicate that oligodendrocytes express relatively high levels of GABA_BR1a RNA, however, functional studies will be required to determine whether this translates into preferential GABA_BR1a-mediated signaling. An isoform specific interactor of GABA_BR1 such as sAPP, therefore, provides a highly useful tool to tease apart R1a-specific effects of GABA_BR activation (Barber et al., 2025).

Our isoform-specific results show that the GABA_BR1b subunit isoform is enriched in astrocytes. In addition, astrocytes were the only cell type to upregulate expression of GABA_BR1b in the APP KO mouse model. GABA_BRs have been shown to be present on astrocytes where they can modulate their inflammatory response to stimuli such as lipopolysaccharide (Lee et al., 2011). Stimulation of GABA_BRs in astrocytes regulates adenylyl cyclase activity (Oka et al., 2006) and mediates calcium signaling in astrocytes which supports interneuron activity (Perea et al., 2016). The enrichment of GABA_BR1b transcripts in astrocytes raises the possibility that some astrocytic GABA_BR functions may preferentially involve this isoform, although functional studies will be required to test this hypothesis.

Our results show that the expression pattern of GABA_BR2 RNA follows that of overall GABA_BR1 closely in the five cell types. However, when assessing isoform-specific expression of GABA_BR1, we observed that GABA_BR2 relative levels match those of GABA_BR1b but not GABA_BR1a. Specifically, oligodendrocytes seem to express significantly high levels of GABA_BR1a but are among the lowest expressing cell types for GABA_BR2. Current consensus in the field indicates that GABA_BRs are obligate heterodimers of GABA_BR1 and GABA_BR2 subunits, with only a few reports challenging this model(Bettler et al., 2004). When expressed alone, GABA_BR1 has been shown to illicit smaller responses suggesting it could be at least partially functional alone or in combination with an unknown partner (Bettler et al., 2004; Kaupmann et al., 1997). In oligodendrocytes, GABA_BR1 levels may be regulated by differentiation stage where the ratio of GABA_BR1 to GABA_BR2 is thought to be changed (Serrano-Regal, Luengas-Escuza, et al., 2020). Cell-type-specific studies of GABA_BR signaling could reveal whether R1/R2 heterodimerization is consistent across all brain cells and whether both subunits are required for sAPP-GABA_BR signaling.

Lastly, we show that while relative baseline GABA_BR in microglia were low, microglia were the most dynamic in regulating their expression of GABA_BR in the APP NL-G-F mouse model of AD. Microglial dysfunction is a key driver of AD progression. Under physiological conditions, microglia contribute to brain homeostasis through many crucial roles such as synaptic pruning (Cornell et al., 2021), secretion of neurogenic factor (Diaz-Aparicio et al., 2020), support axonal myelination (McNamara et al., 2022), and detection and phagocytosis of antigens(Wright-Jin & Gutmann, 2019). In AD, microglia display impaired overall phagocytosis and clearance of Aβ, a chronic activation state, and altered inflammatory signaling (Azmal et al., 2025; Miao et al., 2023). Microglia express GABA_BRs and their activation reduces cytokine release in response to inflammatory stimuli (Kuhn et al., 2003). Also, GABA_BR1 KO in microglia results in altered inhibitory connectivity and hyperactivity(Favuzzi et al., 2021), a phenotype that has been linked to early AD pathogenesis (Targa Dias Anastacio et al., 2022). Our results show that RNA levels of both isoforms of GABA_BR1 were increased at the 3mo age time point in female microglia only. This age coincides with the beginning of plaque deposition and gliosis in APP NL-G-F mice but precedes synaptic loss and cognitive impairment (Saito et al., 2014). This suggests that an increase in GABA_BR expression may be an early change in the female microglial response to disease. We show that in this model, female microglia begin downregulating homeostatic markers earlier than males. We therefore speculate that the increase in GABA_BR expression at 3mo may reflect a compensatory attempt to limit inflammation, which later fails as expression levels return to baseline. Sex-specific differences in microglia are known to influence neuroinflammation (Lynch, 2022). For example, microglia from female mice are more likely to adopt a disease-associated phenotype (Ocañas et al., 2023). Furthermore, AD is more prevalent in females, and there are sex-specific differences in disease presentation (Guo et al., 2022). Interestingly, input samples demonstrated an overall reduction in GABA_BR subunit expression in male APP NL-G-F mice. Although the mechanisms underlying this sex-specific effect remain unclear, this observation may reflect broader disease-associated transcriptional or cellular composition changes occurring prior to cell-specific enrichment and warrants further investigation. The relative pattern of APP expression across cell types appeared to be maintained in APP NL-G-F mice, and microglial APP expression was not altered despite changes in GABA_BR1 expression. These findings indicate that the observed differences in microglial GABA_BR1 expression are unlikely to be explained solely by changes in APP transcript abundance. While APP expression showed a non-significant trend toward reduction in oligodendrocytes, larger studies will be needed to determine whether subtle disease-associated changes in oligodendrocyte APP expression occur in this model.

To summarize, using both analysis of publicly available datasets (Allen et al., 2023; Zhang et al., 2014) and RT-qPCR of cell-sorted mouse brains, we identify oligodendrocytes as a cell type with high expression of both APP and GABA_BR1a RNA, highlighting them as a novel potential cellular site to investigate the APP-GABA_BR interaction. In parallel, we report sex-specific regulation of GABA_BRs in microglia in the APP NL-G-F mouse model of AD³⁵. Because the present study was limited to transcript-level analyses, the proposed functional relevance of APP-GABA_BR interactions in oligodendrocytes and astrocytes will require future protein-level and functional validation. Future studies can build on this work by testing the functional impact of APP-GABA_BR signaling in oligodendrocytes on key processes such as myelination of neurons and investigating the mechanism behind microglial sex-specific differences in GABA_BR regulation and its role in AD pathogenesis.

Limitations of the study

This study provides insight into cell-specific expression of APP and GABA_BR subunits and isoforms at the transcript level in the brain. Nevertheless, transcriptional changes may not directly correspond to protein abundance and function due to additional post-transcriptional regulation. It is therefore critical that future studies evaluate APP and GABA_BR expression at the protein level to provide a comprehensive understanding. Additionally, the relatively small sample sizes used may limit statistical power and the detection of more subtle transcriptional changes. We use the APP NL-G-F mouse model of AD which captures amyloid-driven pathology but does not encompass other features of the disease. Furthermore, changes in the expression of GABA_BR in AD may not reflect changes in the activity of APP-GABA_BR signaling. Additional studies in other models as well as human samples as well as functional assays are needed to allow extrapolation to the full complexity of AD.

4. Methods

4.2. Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Heather C. Rice (heather-rice@omrf.org).

Materials availability

This study did not generate new unique reagents. All materials used can be purchased from the manufacturers.

4.3. Experimental model and study participant details

All animal procedures were approved (protocol 23-78) by the Institutional Animal Care and Use Committee of the Oklahoma Medical Research Foundation and performed in accordance with the National Institutes of Health (NIH) Guide for the Care and Use of Laboratory Animals. This study used C57Bl6 (Jax strain #000664) male and female mice aged 3-12mo. Age-matched male GABA_BR1a KO and APP KO (Jax strain # 004133) and male and female APP NL-G-F (Saito et al., 2014) mice were used. GABA_BR1a KO mice were generated by the VIB-KU Leuven Center for Brain & Disease Research Mouse Expertise Unit, with support from VIB Discovery Sciences. GABA_BR1a KO mice were provided by Joris de Wit, VIB-KU Leuven Center for Brain & Disease Research, Leuven, Belgium. The APP NL-G-F mice were generated by Dr. Takaomi Saido (Saito et al., 2014) and provided by the RIKEN BRC through the National BioResource Project of the MEXT, Japan.

4.4. Method details

Magnetic cell-sorting of whole mouse brain

The cell-sorted fractions were prepared as described in STAR Protocol Houmam et al. (2025) (Houmam et al., 2025). Briefly, mice were anesthetized with isoflurane and euthanized by cervical dislocation and decapitation. The brain was dissected then rinsed in ice-cold D-PBS with calcium, magnesium, glucose, and pyruvate (Thermo Fisher Scientific 14287080). A cut was made to remove the cerebellum, and the remaining cerebrum (forebrain and midbrain) was cut into small cubes on a cold block. The tissue was then transferred to a gentleMACS C-tube containing 5mL of dissociation buffer (papain in EBSS (Worthington LK003150) supplemented with trehalose (MilliporeSigma T9531-10G), AP-5 (MilliporeSigma 165304-5MG), Kynurenic acid (MilliporeSigma K3375-1G), and DNAse (Worthington LK003150)). Actinomycin-D (MilliporeSigma A1410-2MG), anisomycin (MilliporeSigma A9789-5MG), and triptolide (MilliporeSigma T3652-1MG) were added to inhibit ex-vivo activation of glial cells (Ocañas et al., 2022). After dissociation, debris was removed using a gradient spin as in the Worthington dissociation kit, and the solution was strained through a 70um filter to prepare a single-cell suspension in DPBS with 0.5% BSA (Fisher BP1600-100). This was used as input in subsequent RT-qPCR analyses. Myelin removal beads (Miltenyi 130-096-733) were used to separate the suspension into a myelin-positive fraction from which oligodendrocytes were isolated using anti-O4 beads (Miltenyi 130-096-670), and a myelin-negative fraction. From the negative fraction we sequentially isolated microglia using anti-CD11b beads (Miltenyi 130-126-725), endothelial cells using anti-CD31 beads (Miltenyi 130-097-418), astrocytes using anti-ACSA-2 beads (Miltenyi 130-097-678), and neurons using negative selection with biotin tagged beads (Miltenyi 130-115-389). The cell-sorted fractions were validated using flow cytometry for cell-specific markers (Supplemental Figure 4)(Houmam et al., 2025)

RNA extraction and RT-qPCR

RNA was isolated using the Qiagen RNeasy Mini Kit (QIAGEN 74104) following the kit instructions, then quantified using a Nanodrop 1000 spectrophotometer. RNA quality was assessed using a 4150 Tapestation analyzer. RNA was converted to cDNA using the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems 4374966) in a BioRad T100 thermal cycler. RT-qPCR was performed in 384-well plates using Taqman primers and Gene Expression Master Mix (Applied Biosystems 4369016) following the manufacturer instructions. Commercially available taqman primers were used to probe for APP (Mm01344172_m1), GABA_BR1(Mm00444578_m1) and GABA_BR2 (Mm01352554_m1). Primers used to probe for GABA_BR1a and GABA_BR1b transcripts (Gabbr1 NCBI gene 54393) were designed using the ENSMUST00000025338.16 and ENSMUST00000172792.8 sequences respectively (Table 1).

4.5. Quantification and statistical analysis

RNA-seq analysis of publicly available datasets

The publicly available scRNAseq dataset analyzed in this study was generated by Allen et al. (2023)(Allen et al., 2023) and is available on GEO under the accession number: GSE207848. The annotated dataset is accessible on the CELLxGENE database as a .h5ad AnnData file which was generated in Python using standard methods of the package Scanpy by Allen et al. (2023) (Allen et al., 2023). The original data set is comprised of 20929 genes, 4 mice, and 79667 cells of 13 annotated cell types. The .h5ad file was used to create a Seurat object in R. The data was then log normalized and scaled using the functions NormalizeData and ScaleData in the package Seurat at default settings. We subset the data to the cell types of interest based on the Allen et al. annotations: oligodendrocytes, microglia, endothelial cells, astrocytes, and neurons, as well as the two ages analysed in this study: 4 weeks-old and 90weeks-old. Seurat’s FeaturePlot function was used to visualize App expression on a UMAP embedding of single cells across five cell types. The VlnPlot function was used to further visualize the distribution of App expression across cell types in both the 4- and 90- week-old ages.

The publicly available bulk RNAseq dataset analyzed in this study was generated by Zhang et al. (2014) (Zhang et al., 2014) (brainrnaseq.org) and is available on GEO under the accession number: GSE52564. To probe expression of GABA_BR1 (ensembl ID: ENSMUSG00000024462) and GABA_BR2 (ensembl ID: ENSMUSG00000039809), sequence alignment map (.sam) files were generated using the STAR aligner (2.7.11b) to align reads to the Mus musculus GRCm39 reference genome. To probe expression of GABA_BR1a and GABA_BR1b isoforms, we edited the GRCm39 reference genome file to annotate the transcript regions unique to each isoform as Gabbr1 (1a) and Gabbr1 (1b) respectively. The start and end of the isoform-specific exon regions were determined based on the sequences of the transcript ID ENSMUST00000025338.16 (Gabbr1-201) for Gabbr1 (1a) and ENSMUST00000172792.8 (Gabbr1-203) for Gabbr1 (1b). After each alignment, .sam files were converted to binary alignment map files (.bam) using samtools (1.18). Read counts were then generated using the featureCounts command in the subread package (2.1.1) and normalized using DESeq2 (1.42.1) in R.

Statistical analysis

Statistical analysis of relative gene expression levels compared across cell types within the same genotype was performed in Graphpad Prism using a Kruskal-Wallis test with Dunn’s multiple corrections with n=4 where each sample is an independent mouse. Analysis of RT-qPCR data comparing two genotypes was performed using Two-way Anova with factors genotype and cell type, followed by Sidak’s multiple comparisons with n=3 independent mice per group. Bulk RNAseq data was also analyzed in Graphpad Prism using a Kruskal-Wallis test with Dunn’s multiple corrections with n=2. scRNAseq analysis was performed using standard functions in the package Seurat in R.

Supplementary Material

1

Supplemental Figure 1: Cell-type specific expression levels of APP and GABA_BR1 in knockout mouse models Relative RNA expression measured by RT-qPCR in sorted oligodendrocytes, microglia, endothelial cells, astrocytes, and neurons. (A) APP expression in WT (black) and APP KO (orange) mice. (B) GABABR1a expression in WT (black) and GABA_BR1a KO (blue) mice. (C) GABA_BR1b expression in WT and GABA_BR1a KO mice. (D) GABA_BR1a expression in WT and APP KO mice. (E) GABA_BR1b expression in WT and APP KO mice. (F) APP expression in WT and GABA_BR1a KO mice. All mice were 5mo and male. WT data is the same as male data in Figure 2G–J. Two-way ANOVA with factors genotype and cell type was performed, followed by Sidak’s multiple comparisons between genotypes within each cell type (** = p≤0.01, ** = p≤0.0001). n=4 per group). Data presented as median ± IQR.

2

Supplemental figure 2: Cell type-specific expression of GABA_BR subunits and isoforms in WT mice across age Relative RNA expression of GABA_BR subunits and isoforms measured by RT-qPCR in sorted oligodendrocytes, microglia, endothelial cells, astrocytes, and neurons from male and female WT 3mo (A), 6mo (B), and 12mo (C) mice (Kruskal-Wallis with Dunn’s multiple comparisons * = p<0.05, ** = p<0.01, *** = p<0.001, **** = p<0.0001, n=3 per group). Data is the same as WT in Figure 3, here normalized to input. Data shown as median ± IQR.

3

Supplemental Figure 3: Cell type-specific expression of APP in WT and APP NL-G-F 3mo mice (A) Relative APP RNA expression across enriched cell types isolated from 3mo male (green) and female (pink) WT (no fill) and NL-G-F mice (solid fill) (Kruskal-Wallis with Dunn’s multiple comparisons * = p<0.05, ** = p<0.01, n=3 per group. Data shown as median ± IQR.). (B) Heatmap showing RT-qPCR RNA expression levels of APP across five sorted cell types from WT and APP NL-G-F mice at 3mo from male and female mice. Scale corresponds to log2 fold change (Mann-Whitney test, all comparisons ns, n=3 per group).

4

Supplemental figure 4: Flow cytometry validation of cell-sorted fractions (A) Flow cytometry gating strategy and representative scatter plots showing enrichment of cell markers (O4 as a marker of oligodendrocytes, CD11b for microglia, CD31/CD34 for endothelial cells, and ACSA2 for astrocytes). (B) Summary table showing percent cells in each fraction stained by cell type-specific antibody using flow cytometry. Data displayed as average ± SD.

4.1.

Key resource table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Critical commercial assays
RNeasy kit	QIAGEN	74104
Applied Biosystems High-Capacity cDNA Reverse Transcription Kit with RNAse Inhibitor	Applied Biosystems	4374966
TaqMan Gene Expression Master Mix	Applied Biosystems	4369016
Experimental models: Organisms/strains
C57BL/6J mice	Jackson Laboratory	Strain #000664
APP KO mice	Jackson Laboratory	Strain #004133
GABABR1a KO mice	VIB-KU Leuven Center for Brain & Disease Research, Leuven, Belgium	N/A
APP NL-G-F mice	RIKEN BRC, National BioResource Project of the MEXT, Japan	N/A
Oligonucleotides
GABABR1TaqMan qPCR primer	Thermo Fisher Scientific	Mm00444578_m1
GABABR2 Taqman qPCR primer	Thermo Fisher Scientific	Mm01352554_m1
Tmem119 Taqman qPCR primer	Thermo Fisher Scientific	Mm00525305_m1
P2ry12 Taqman qPCR primer	Thermo Fisher Scientific	Mm00446026_m1
Clec7a Taqman qPCR primer	Thermo Fisher Scientific	Mm01183349_m1
Apoe Taqman qPCR primer	Thermo Fisher Scientific	Mm01307193_g1
Lpl Taqman qPCR primer	Thermo Fisher Scientific	Mm00434764_m1
Cst7 Taqman qPCR primer	Thermo Fisher Scientific	Mm00438351_m1
Axl Taqman qPCR primer	Thermo Fisher Scientific	Mm00437221_m1
β-actin TaqMan qPCR primer	Thermo Fisher Scientific	Mm02619580_g1
Software and algorithms
GraphPad Prism 10	GraphPad	graphpad.com
RStudio	Posit PBC	posit.co/download/rstudio-desktop/
MobaXterm	Mobatek	mobaxterm.mobatek.net/
Other
BIORAD T100 Thermal Cycler	Bio-Rad	N/A
Agilent Tapestation 4150	Agilent	N/A
QuantStudio 5	Applied Biosystems	N/A

Highlights.

• APP RNA is enriched in oligodendrocytes in mouse brain.

• GABA_BR1a RNA levels are highest in oligodendrocytes.

• GABA_BR1b RNA levels are highest in astrocytes.

• Microglial GABA_BR1 expression is increased in female 3-month-old APP NL-G-F mice.

Data and code availability

This paper did not generate any new datasets or code. The data that support the findings of this study are available from the corresponding author upon reasonable request. The publicly available RNAseq datasets analysed in this paper are accessible under the following GEO accession numbers: and GSE52564.