Abstract
The medial frontal cortex (mFC) and locus ceruleus (LC) are two brain areas that have been implicated in a range of cognitive phenomena, such as attention, memory, and decision-making. Regulators of these brain regions at the molecular level are not well understood but might help to elucidate underlying mechanisms of disorders that present with deficits in these cognitive domains. To probe this, we used chemogenetic stimulation of neurons in the LC with axonal projections to the prelimbic subregion (PrL) of the mFC and subsequent bulk RNA sequencing from the mouse PrL. We found that stimulation of this circuit caused an increase in transcription of a host of genes, including the Apoe gene. To investigate cell type-specific expression of Apoe in the PrL, we used a dual-virus approach to express either the excitatory DREADD receptor hM3Dq in LC neurons with projections to the PrL or a control virus and found that increases in Apoe expression in the PrL following depolarization of LC inputs is enriched in GABAergic neurons in a sex-dependent manner. The results of these experiments yield insights into how Apoe expression affects function in a cortical microcircuit that is important for attention, memory, and decision-making and point to interneuron-specific expression of Apoe as a potential biomarker for circuit function in disorders such as attention-deficit hyperactivity disorder, schizophrenia, and Alzheimer's disease.
Significance Statement
Identifying patterns of gene expression in specific brain circuits is an important first step toward developing treatments for cognitive and behavioral symptoms that rely on those circuits. In this paper, we describe a transcriptome-scale motif in one such circuit—neurons in the locus ceruleus that project to the prelimbic subregion. This circuit has been implicated in attention, memory, and decision-making, and deficits in these cognitive domains are common across many neuropsychiatric disorders. We further explored one of the top differentially expressed genes, Apoe, to identify how it is expressed in distinct cell types following stimulation of this circuit, paving the way for spatially and genetically specific targeting of this gene in disorders that feature dysfunction in this circuit.
Introduction
The anterior cingulate cortex (ACC), located on the medial surface of the frontal lobes in humans, is heavily involved in a variety of cognitive domains, including attention (Davis et al., 2000), memory (Kozlovskiy et al., 2012), and decision-making (Walton and Mars, 2007). For example, the ACC is highly active during sustained attention (MacDonald et al., 2000), as well as conflict monitoring, which is the allocation of attention based on conflicting signals (Carter et al., 1998). ACC activity is also correlated with better performance on tasks that measure working memory (Lenartowicz and McIntosh, 2005) and remote memory (Petersen et al., 1988). ACC activity during these tasks is likely influenced by neuromodulatory systems, such as the locus ceruleus (LC), which supplies cortical norepinephrine (NE) and is reciprocally connected to the ACC in primates (Joshi and Gold, 2022) and rodents (Jodoj et al., 1998; Pudovkina et al., 2001). LC-NE activation influences ACC activity patterns and affects both attention (Dahl et al., 2020) and memory (Murchison et al., 2004; Bahtiyar et al., 2020), implicating the LC as a driver of cortical function in attention and memory-guided behavior. Deficits in ACC and LC function are also present in disorders in which attention and memory deficits are a common cognitive symptom, such as schizophrenia, attention-deficit hyperactivity disorder (ADHD), and major depressive disorder (MDD; Kerns et al., 2005; del Cerro et al., 2020; Shirama et al., 2020), highlighting the importance of understanding how the LC and ACC functionally interact with one another.
Translating research on the primate ACC to rodents has been difficult, in part because identifying regions in the rodent that are homologous to primate cortical areas is not straightforward. Recent research has provided strong evidence that the area commonly referred to as the medial prefrontal cortex (mPFC) in rodents is actually more anatomically homologous with the human cingulate cortex (Laubach et al., 2018; van Heukelum et al., 2020), providing an avenue for comparison of ACC activity between species. In support of this, many studies in rodents have shown that activity in the prelimbic subregion (PrL) of the rodent mPC is strongly related to attention (Birrell and Brown, 2000; Passetti et al., 2002; Fisher et al., 2020) and memory (Brito and Brito, 1990; Delatour and Gisquet-Verrier, 1996; Hallock et al., 2016; Ye et al., 2017). Furthermore, catecholamine signaling in the rodent PrL is involved in attention (Berridge and Spencer, 2016) and memory (Tronel et al., 2004), and drugs that modulate catecholaminergic function often affect memory and attention in rodents (Sherrill et al., 2013; Mar et al., 2017; Caballero-Puntiverio et al., 2019; McDonald et al., 2021), suggesting that brain areas like the LC drive function in cortical areas to promote rodent cognition as well. The PrL and LC functionally interact during rodent versions of attention (Bouret and Sara, 2004; Newman et al., 2008; Hallock et al., 2024) and memory tasks (Giustino et al., 2019), further corroborating the idea that the LC and PrL work together during attention and memory-guided behavior across mammalian species.
Although specific brain circuits, like the one between the LC and PrL, have been implicated in a range of cognitive processes, little research has been done to understand the molecular drivers of function in these circuits. Gene expression in cortical tissue is heterogeneous between brain regions and cell types in both humans (Darmanis et al., 2015; Maynard et al., 2021) and rodents (Zeisel et al., 2015), and these patterns of gene expression change in response to behavioral experience (Cho et al., 2016; Mukherjee et al., 2018), suggesting that unique cell type-specific patterns of gene expression could be used as markers of circuit function. These results also raise the possibility that cells embedded in specific circuits could be noninvasively accessed by manipulating genes selectively expressed in those cells, paving the way for gene-targeted treatments for cognitive symptoms in neuropsychiatric disorders, such as deficits in attention, memory, and decision-making. In order to uncover gene expression motifs in one such circuit, we used chemogenetic targeting to activate neurons in the LC that send direct axonal projections to the PrL in mice. We subsequently used bulk RNA sequencing and single-molecule in situ hybridization to look at cell type-specific transcription in the PrL in response to stimulation of LC inputs. We found that, among other genes, the gene encoding for the apolipoprotein E protein (Apoe) was enriched in PrL tissue following circuit activation. This increase in Apoe transcription was driven by expression in GABAergic neurons, and this effect was more robust in female mice.
Materials and Methods
Subjects
For the RNA sequencing experiment, we used a cohort of eight male (four experimental and four control) wild-type c57bl/6j mice (Jackson strain 000664). For RNAscope experiments in the PrL, we used a cohort of 12 female (6 experimental and 6 control) and 12 male (6 experimental and 6 control) wild-type c57bl/6j mice (24 total mice). Mice were group housed (3–5 animals per cage) with ad libitum access to both food and water. The colony room was temperature and humidity controlled on a 12 h light/dark cycle. All experiments were performed during the light cycle. At time of surgery, animals were roughly 90–120 d of age. All procedures were in accordance with the Institutional Animal Care and Use Committees of The Lieber Institute for Brain Development (mice used in the RNA sequencing experiment) and Lafayette College (mice used in RNAscope experiments).
Surgical and extraction procedures
For all experiments, mice were anesthetized with isoflurane (1–2.5% oxygen) and then placed into a stereotaxic frame (Kopf Instruments). An incision was made along the midline of the scalp, the skull was leveled, and bregma was identified. Holes were drilled in the skull above the target brain regions (LC, −5.4 mm AP from the bregma, ±0.9 mm ML from the midline; PrL, +1.7 mm AP from the bregma, ±0.3 mm ML from the midline), and an automated infusion pump (World Precision Instruments) was used to inject the viruses at 3 nl/s for a total volume of 600 nl/hemisphere in the PrL (1.7 mm ventral to the surface of the brain) and 300 nl/hemisphere in the LC (3.0 mm ventral to the surface of the brain). Measurements for these brain regions were obtained from the Paxinos and Franklin mouse brain atlas (Paxinos and Franklin, 2019). For the RNA sequencing experiment, a retrograde virus encoding Cre-recombinase (AAVrg-hSyn-Cre; Addgene, catalog #105553-AAVrg) was injected into the PrL, and a virus coding for the Cre-dependent expression of an excitatory DREADD receptor (AAV8-hSyn-DIO-hM3Dq-mCherry; Addgene, catalog # 44361-AAV8) was injected into the LC (Fig. 1A). For RNAscope experiments, AAVrg-hSyn-Cre was injected into the PrL, and either a virus coding for the Cre-dependent expression of an excitatory DREADD receptor (AAV8-hSyn-DIO-hM3Dq-mCherry; experimental group) or a control virus (AAV8-hSyn-DIO-mCherry; control group; Addgene, catalog #114472-AAV8) was injected into the LC (Fig. 2A). For the RNA sequencing experiment, we waited 4–5 weeks for the virus to infect a sufficient number of cells and then gave intraperitoneal injections of either clozapine-N-oxide (CNO; 2.5 mg/kg; experimental group; Tocris Bioscience, catalog #4936) or vehicle [filtered 1× phosphate-buffered saline (PBS); control group] into the mice. Vehicle injections were matched to the volume of CNO injections in experimental mice. Because CNO exerts its peak effects after ∼30 min, and peak expression of many immediate early genes (IEGs) is typically seen ∼90 min after stimulation (Barry and Commins, 2017), we killed the mice 120 min following injections. Following cervical dislocation, the brains of the mice were extracted, the medial wall of the PrL from both hemispheres was dissected with a brain block and razor blades on wet ice, and each hemisphere was flash-frozen in 2-methylbutane and stored in a 1.5 ml Eppendorf centrifuge tube at −80°C. For RNAscope experiments, we again waited 4–5 weeks for the virus to infect a sufficient number of cells, gave all mice CNO injections (2.5 mg/kg, i.p.), and killed the mice 120 min following injections. We then extracted the brains, flash-froze them in 2-methylbutane, and stored them at −80°C.
RNA sequencing reveals enrichment of genes in PrL tissue following stimulation of LC inputs. A, Schematic showing viral injection strategy for chemogenetic activation of prelimbic cortex (PrL)-projecting neurons in the locus ceruleus (LC). B, Experimental design for identification of enriched gene sets in the PrL following chemogenetic activation of PrL-projecting LC neurons. C, Volcano plot showing genes that were enriched in PrL tissue in the experimental (CNO-receiving) group of mice (red dots) and genes that were enriched in PrL tissue in the control (saline-receiving) group of mice (blue dots) based on differential expression analysis. The top 10 differentially expressed genes (based on FDR-adjusted p value) are labeled. D, Dot plot showing gene ontology terms that contained differentially expressed gene sets for experimental mice (left column) and control mice (right column). The top 10 gene ontology terms by FDR-adjusted p value are listed on the y-axis.
Figure 1-1
Table showing log fold changes, p-values, and FDR-adjusted p-values for all differentially-expressed genes in the bulk RNA-sequencing experiment. Download Figure 1-1, XLS file.
Figure 1-2
Table showing p-values, FDR-adjusted-values, and core genes (in ENSEMBL notation) for all gene ontology (GO) terms from the bulk RNA-sequencing experiment. Download Figure 1-2, XLS file.
Verification of DREADD-mediated activation of LC neurons. A, Schematic showing viral injection strategy for activation of PrL-projecting neurons in the LC for RNAscope experiments. B, Experimental design. C, Confocal image showing tyrosine hydroxylase (Th) expression in the locus ceruleus (LC). D, Confocal image at 40× magnification showing expression of Fos, Th, and Cre in LC tissue. E, In mice that had the DREADD receptor, Fos expression was increased in Cre+ cells (left panel—t(7.786) = 3.2078, p = 0.01292), Th + cells (middle panel—t(7.6345) = 3.8082, p = 0.005647), and cells that were both Th and Cre+ (right panel—t(9.9693) = 2.3047, p = 0.04398) compared with mCherry controls. Error bars = SEM.
Bulk RNA sequencing
Total RNA was isolated and extracted from the tissue using TRIzol (Life Technologies), purified using RNeasy minicolumns (Qiagen; catalog #74104), and quantified using a NanoDrop spectrophotometer (Agilent Technologies). The Nextera XT DNA Library Preparation Kit was used to generate sequencing libraries according to manufacturer instructions, and samples were sequenced on a HiSeq 2000 (Illumina). Reads were aligned to the mm10 genome using the HISAT2 splice-aware aligner (D. Kim et al., 2015), and alignments overlapping genes were counted using featureCounts version 1.5.0-p3 relative to Gencode version M11. Differential expression analyses were performed on gene counts using the voom approach (Law et al., 2014) in the limma R/Bioconductor package (Ritchie et al., 2015) using weighted trimmed means normalization factors with condition (CNO vs saline) as the main outcome of interest and adjusting for the exonic mapping rate. Multiple testing correction was performed using the Benjamini–Hochberg approach to control for the false discovery rate (FDR). Gene ontology (GO) analyses were performed using ENSEMBL gene IDs with the clusterProfiler R Bioconductor package (Ashburner et al., 2000; Yu et al., 2012).
Single-molecule fluorescent in situ hybridization
To perform single-molecule in situ hybridization (RNAscope) on LC and PrL tissue, we took coronal sections of the PrL and LC (16 μm) on a cryostat (Leica), mounted them onto slides, and performed the RNAscope protocol using the fluorescent multiplex V2 kit from ACDBio (catalog #323110). Specifically, tissue sections were briefly fixed with a 10% neutral buffered formalin solution at room temperature and subjected to serious dehydration with ethanol. We then pretreated the sections with protease IV and hydrogen peroxide and incubated the slides at 40°C with a combination of three probes. For PrL tissue, these probes were as follows: channel 1, Gfap; channel 2, Rbfox3; channel 3, Apoe for Experiment 1; channel 1, Slc17a7; channel 2, Gad1; channel 3, Apoe for Experiment 2; and channel 1, Sst; channel 2, Pvalb; channel 3, Apoe for Experiment 3. For LC tissue, these probes were as follows: channel 1, Fos; channel 2, Th; channel 3, Cre. After incubation, we applied amplification buffers for each channel and opal dyes (520 nm for channel 1; 570 nm for channel 2; 690 nm for channel 3; Akoya Biosciences, catalog #FP1487001KT, FP1488001KT, and FP1497001KT) in order to fluorescently label each transcript. Lastly, we stained the sections with DAPI to demarcate the nuclei of the cells. We then took z-stacked images of the PrL (four sections per slide, one image per section, four images per mouse total) and LC slides (one section per slide, one image per section, one image per mouse total) using a Zeiss LSM800 confocal microscope. For analysis, we used a MATLAB program to quantify transcript expression in each image (dotdotdot; Maynard et al., 2020). Specifically, we used the “CellSegm” toolbox to perform nuclear segmentation in x, y, and z-dimensions to define regions of interest (ROIs) based on DAPI expression and watershed analysis to identify distinct transcripts (individual dots) in each of the three microscope channels corresponding to an opal dye. We then colocalized each identified transcript with an identified nucleus (ROI). Transcripts that were not classified by the program as being colocalized with a nucleus were not used for analysis. Background noise, which could result from bleed-through from adjacent wavelengths in the gene channels, was eliminated using the “imhmin” function. This function suppresses all of the minima in the grayscale image whose depth is less than the standard deviation of the image. We used a cutoff of five transcripts to categorize a cell as either Rbfox3 or Gfap-expressing, Gad1 or Slc17a7-expressing, and Pvalb or Sst-expressing, respectively.
Statistical analysis
For RNAscope results, we used multiple two-way analyses of variance (ANOVAs) to compare gene expression between sex and experimental condition (DREADD vs mCherry controls). We also used Welch's two-sample t tests to compare gene expression in the LC between DREADD and mCherry controls. An alpha level of 0.05 was used to determine statistical significance. All tests were performed with the R programming language (“aov” and “t.test” functions).
Results
RNA sequencing
We identified 96 genes that were differentially expressed in PrL tissue between our CNO and saline groups (genes that had an FDR-adjusted p value < 0.05). Of these genes, 82 were enriched in the saline group, and 14 were enriched in the CNO group (Fig. 1C). Gene ontology (GO) analysis revealed that genes enriched in the CNO group are involved in circadian rhythms, axonogenesis, and neurotrophin signaling, suggesting that stimulation of LC inputs to the PrL induces signaling pathways involved in plasticity (Fig. 1D). Of the genes that were enriched in the CNO group, the Apoe gene (log2FoldChange = 0.34, FDR-adjusted p value = 0.0342) stood out due to its involvement in attention and memory in humans (Parasuraman et al., 2002; Jochemsen et al., 2012; Rusted et al., 2013). Although there were other genes that were more significantly upregulated in our geneset, none of these genes had obvious links to cognition or disorders of the nervous system, making Apoe an attractive candidate to study further. A complete list of differentially expressed genes is available in the extended data (Extended Data Fig. 1-1).
Single-molecule in situ hybridization
In order to verify our RNA sequencing results, and determine cell type-specific expression of Apoe in PrL tissue following depolarization of LC inputs, we replicated our RNA sequencing experiment with several important changes. First, we used both male and female mice in order to determine whether sex differences in Apoe expression levels might be present, as previous research has demonstrated sex differences in mouse LC anatomy and function (Bangasser et al., 2011, 2016). Second, we employed an experimental design in which all mice received CNO injections, as the presence of CNO alone could have accounted for the increase in Apoe transcription that we observed in our RNA sequencing data.
Firstly, to verify that our viral targeting strategy increased neuronal activity in LC neurons, we took coronal sections of the LC and quantified Fos expression in neurons expressing tyrosine hydroxylase (Th-expressing neurons) and neurons expressing Cre-recombinase (Cre-expressing neurons; Fig. 2C,D). Due to the difficulties inherent in slicing LC tissue, we were only able to obtain usable slices from 16 total mice (12 male, 6 in the experimental group, 6 in the control group; 4 female, 2 in the experimental group, 2 in the control group). We found evidence that Th- and Cre-expressing neurons in the LC were robustly activated following CNO injections in the experimental group, as Fos transcription was increased in Cre-expressing (t(7.786) = 3.2078; p = 0.01292), Th-expressing (t(7.6345) = 3.8082; p = 0.005647), and both Cre- and Th-expressing (t(9.9693) = 2.3047; p = 0.04398) neurons in the LC of experimental mice compared with mice in the control group (Welch's t tests; Fig. 2E). Although the small number of female mice in our study precluded analysis of sex differences for this particular experiment, when data were graphed separately for males and females, we saw that Fos expression was increased in LC neurons of the experimental group in both sexes.
We next looked for evidence of Apoe expression in astrocytes and neurons in the PrL following chemogenetic stimulation of LC inputs (Fig. 3B) and found that neither number of Apoe transcripts nor percentage of cells expressing Apoe significantly differed between group or sex in astrocytes (Gfap-expressing cells; Fig. 3C) but that both number of transcripts and percentage of cells expressing Apoe were significantly increased in putative neurons (Rbfox3-expressing cells; F(1,84) = 10.498, p = 0.00171, main effect of group for number of Apoe transcripts; F(1,84) = 40.078, p = 1.14 × 10−8, main effect of condition for percentage of neurons expressing Apoe). Additionally, the number of Apoe transcripts expressed in Rbfox3-expressing cells was significantly higher in females, compared with males (F(1,84) = 15.777, p = 0.00015, main effect of sex; Fig. 3D).
Enrichment of Apoe in PrL neurons following stimulation of LC inputs. A, Experimental design. B, Confocal image showing expression of Gfap, Rbfox3, and Apoe in PrL tissue. C, Left panel, Graph showing average number of Apoe transcripts per astrocyte (Gfap+ cell) in the PrL. Right panel, Graph showing percentage of astrocytes (Gfap+ cells) that coexpress Apoe in the PrL. No significant differences in Apoe expression in astrocytes were observed between conditions or sexes. D, Left panel, Graph showing average number of Apoe transcripts per neuron (Rbfox3+ cell) in the PrL. Right panel, Graph showing percentage of neurons (Rbfox3+ cells) that coexpress Apoe in the PrL. Apoe expression significantly increased in neurons following LC depolarization (F(1,84) = 15.777, p = 0.00015, main effect of sex, F(1,84) = 10.498, p = 0.00171, main effect of condition for number of Apoe transcripts per Rbfox3 + cell, and F(1,84) = 40.078, p = 1.14 × 10−8, main effect of condition for % Rbfox3+ cells coexpressing Apoe).
To determine whether neuronal increases in Apoe were predominantly in excitatory (glutamatergic) or inhibitory (GABAergic) neurons, we next looked at Apoe expression in Slc17a7-expressing cells (Slc17a7 encodes the glutamate transporter protein) and Gad1-expressing cells in PrL tissue (Fig. 4A). We found that neither number of Apoe transcripts nor percentage of cells expressing Apoe significantly differed between group or sex in putative excitatory neurons (Fig. 4B) but were both significantly increased in putative inhibitory neurons in the experimental group compared with the control group, irrespective of biological sex (F(1,84) = 9.154, p = 0.00329, main effect of group for number of Apoe transcripts; F(1,84) = 12.898, p = 0.000553, main effect of group for percentage of Gad1-expressing neurons coexpressing Apoe; Fig. 4C).
A, Enrichment of Apoe in GABAergic PrL neurons following stimulation of LC inputs. Confocal image showing expression of Slc17a7, Gad1, and Apoe in PrL tissue. B, Left panel, Graph showing average number of Apoe transcripts per glutamatergic neuron (Slc17a7+ cell) in the PrL. Right panel, Graph showing percentage of glutamatergic neurons (Slc17a7+ cells) that coexpress Apoe in the PrL. No significant differences in Apoe expression in glutamatergic cells were observed between conditions or sexes. C, Left panel, Graph showing average number of Apoe transcripts per GABAergic neuron (Gad1+ cell) in the PrL. Right panel, Graph showing percentage of GABAergic neurons (Gad1+ cells) that coexpress Apoe in the PrL. Apoe expression significantly increased in GABAergic neurons following LC depolarization (F(1,84) = 9.154, p = 0.00329, main effect of condition for number of Apoe transcripts per Gad1+ cell, F(1,84) = 12.898, p = 0.000553, main effect of condition for % Gad1+ cells coexpressing Apoe).
Finally, we looked at Apoe expression in two different subtypes of inhibitory neuron—parvalbumin interneurons (Pvalb-expressing cells) and somatostatin interneurons (Sst-expressing cells; Fig. 5A). We found no evidence of increased Apoe expression in putative parvalbumin interneurons in the experimental group—however, we did find an increased number of Apoe transcripts in Pvalb-expressing cells in females compared with males (F(1,86) = 12.961, p = 0.000531, main effect of sex), as well as an increase in the percentage of Pvalb-expressing cells that coexpressed Apoe in females compared with males (F(1,86) = 5.536, p = 0.0209; Fig. 5B), suggesting that there are baseline differences in Apoe expression in cortical parvalbumin interneurons between sexes in mice. Apoe expression was, however, increased in Sst-expressing cells in the experimental group (F(1,86) = 11.088, p = 0.001281, main effect of group for number of Apoe transcripts). The increase in Apoe expression in putative somatostatin interneurons was partially sex dependent—the number of Apoe transcripts in Sst-expressing cells was also higher in females compared with males (F(1,86) = 13.425, p = 0.000429, main effect of sex for number of Apoe transcripts), and analysis of the percentage of Sst-expressing cells that coexpressed Apoe revealed a sex × group interaction (F(1,86) = 7.476, p = 0.00759), with an increase in the female experimental group specifically.
Sex-specific enrichment of Apoe in GABAergic neuron subtypes following stimulation of LC inputs. A, Confocal image showing expression of Pvalb, Sst, and Apoe in PrL tissue. B, Left panel, Graph showing average number of Apoe transcripts per parvalbumin interneuron (Pvalb+ cell) in the PrL. Right panel, Graph showing percentage of parvalbumin interneurons (Pvalb+ cells) that coexpress Apoe in the PrL. Apoe expression in parvalbumin interneurons was significantly higher in females (F(1,86) = 12.961, p = 0.000531, main effect of sex for number of Apoe transcripts per Pvalb+ cell, F(1,86) = 5.536, p = 0.0209, main effect of sex for % Pvalb+ cells coexpressing Apoe). C, Left panel, Graph showing average number of Apoe transcripts per somatostatin interneuron (Sst+ cell) in the PrL. Right panel, Graph showing percentage of somatostatin interneurons (Sst+ cells) that coexpress Apoe in the PrL. Apoe expression significantly increased in somatostatin interneurons following LC depolarization (F(1,86) = 11.088, p = 0.001281, main effect of condition, F(1,86) = 13.425, p = 0.000429, main effect of sex for number of Apoe transcripts per Sst+ cell, F(1,86) = 7.476, p = 0.00759, condition × sex interaction for % Sst+ cells coexpressing Apoe).
Discussion
We find that chemogenetic activation of a circuit that is involved in attention (Hallock et al., 2024) induces cortical expression of genes in mice that could be used for molecular identifiers of function in this circuit. Specifically, we find that one of these genes (Apoe) is selectively enriched in GABAergic neurons following circuit activation. These results provide insight into how distinct anatomical inputs might regulate the transcriptome of cortical regions that are involved in a variety of behaviors, such as the prelimbic subregion of the frontal cortex. The design employed in this study compliments research demonstrating behaviorally induced (Cho et al., 2016; Mukherjee et al., 2018) and stimulation-induced (Nelson et al., 2023; Bach et al., 2024) changes in gene expression in rodents. The identification of genes that are upregulated upon activation of brainstem→cortical circuits might serve as a starting point for the development of targeted treatments for cognitive deficits in disorders in which these circuits are dysfunctional, such as ADHD, schizophrenia, and MDD. Interestingly, we also found that a sizable number (82) of genes had reduced transcription in the PrL following stimulation of PrL-projecting neurons in the LC. Gene ontology analysis revealed that these genes are involved in ornithine decarboxylase inhibitor activity, negative regulation of polyamine transmembrane transport, and phosphatase complexes (for a complete list of gene ontology terms and core gene sets, see Extended Data Fig. 1-2). At the individual level, there was no discernable functional relationship between top differentially expressed genes that were downregulated in the CNO group. Some of these genes are implicated in animal models of stress and depressive symptoms (Gm27177 and Krt80, for example; Dai et al., 2022; Lanshakov et al., 2024). Others are involved in olfaction (Bpifb9a; Kuntová et al., 2018) and abnormal tau phosphorylation in animal models of Alzheimer's disease (AD; Hs3st2; Sepulveda-Diaz et al., 2015). LC efferents modulate inhibitory transmission in cortex (Toussay et al., 2013), and a large number of cortical interneurons express ɑ1, β1, and β2 adrenergic receptors (Papay et al., 2006; Santana et al., 2013), raising the possibility that the decrease in transcription of some genes in our study is due to elevated interneuron activity following LC→PrL stimulation. Although we were not able to verify that this was the case, our observation that Apoe transcription increased selectively in interneurons is in line with this hypothesis. Correspondingly, we also did not see increased transcription of any canonical immediate early genes (IEGs), such as Fos, Arc, or Npas4. This raises the possibility that our stimulation protocol was not sufficient to elicit activation of PrL-projecting LC neurons or detect the presence of IEGs in PrL tissue. Several lines of evidence argue against this interpretation, however. First, results shown in Figure 2 demonstrate that Fos transcription was heavily increased in these neurons following CNO administration in our DREADD-expressing mice, showing that these neurons were indeed activated with our CNO dose and timeline. Second, previous studies using the same timeline (30 min for CNO and 90 min for IEG expression, 120 min total) and experimental protocol in PrL-projecting hippocampal neurons found robust IEG expression in PrL tissue (Hallock et al., 2020), suggesting that the lack of IEG expression found in the current study is due to differences in the neural circuitry under investigation.
The APOE gene encodes for a polymorphic protein that is implicated in processes such as neurogenesis, plasticity, and neuronal repair (Rusted et al., 2013). APOE has four allelic variants in humans, with the fourth allele (−e4) being heavily linked to the development of AD. Patients with AD exhibit many facets of attentional and memory deficits; for example, selective attention, or the ability to attend to a particular stimulus in the environment, and executive control of attention, or coordinating goal-directed activities, are both impaired in AD (Parasuraman et al., 2002). APOE genotype, even in the absence of AD pathology, modulates a range of processes, including visuospatial attention and working memory (Espeseth et al., 2006). Interestingly, deficits in attention, specifically the shifting of visuospatial attention, have been found in APOE −e4 allele carriers who do not have dementia (Greenwood et al., 2000).
APOE allele type impacts facets of attention and memory across the lifespan as well; younger individuals who are carriers of the −e4 allele perform better on measures of sustained and covert attention when compared with younger individuals who are not carriers of the −e4 allele (Rusted et al., 2013). Interestingly, advantages in performance noted early in life for −e4 carriers might turn into disadvantages later in life. Filippini et al. (2011) looked at blood oxygen level-dependent (BOLD) signals in both older and younger carriers, and noncarriers, of the −e4 allele and found that older −e4 individuals exhibited significantly increased activation of several brain regions relative to younger −e4 carriers, an effect that was not observed between young and old −e4 carriers. Filippini et al. (2011) proposed that the −e4 allele might impact age-related compensatory processes. Similarly, Jochemsen et al. (2012) found that younger −e4 carriers had better memory recall compared with participants with other APOE genotypes, while older −e4 carriers had worse recall. It is possible that the age- and allele-related decline in task performance observed with relation to APOE is related to neurophysiological changes, such as an increase in amyloid beta deposition, which is a hallmark of AD (Filippini et al., 2011). The −e4 allele has the least amount of lipidation out of any of the APOE alleles (Flowers and Rebeck, 2020), which means that the proteins encoded by the APOE −e4 allele are more likely to aggregate and become toxic to neurons.
In order to fully understand the function of APOE as it relates to neural circuits that are involved in cognitive aspects of neuropsychiatric and neurodegenerative disorders, such as attention and memory, it is necessary to know where APOE is being expressed within those circuits. LC neurons degenerate over the progression of AD and have been identified as one of the first locations in the brain to accumulate tau protein, a pathological hallmark of AD (Dahl et al., 2020; James et al., 2020). Alterations in LC-NE system function have also been implicated in schizophrenia (Suttkus et al., 2021) and ADHD (Ressler and Nemeroff, 2001). Ideally, if APOE is used as a potential biomarker for LC→frontal cortex function, it could help shed light on how and when this circuit might malfunction in patients with these disorders. In a healthy human brain, APOE is expressed primarily in astrocytes (Flowers and Rebeck, 2020). Astrocytes provide neurons with cholesterol, and APOE is the predominant carrier for cholesterol transport to these neurons (D. Li et al., 2022). Similar to humans, Apoe is primarily expressed in astrocytes in wild-type rodent brains (Raber et al., 1998). However, Q. Xu et al. (2006) found that upon excitotoxic injury, hippocampal neurons in the mouse brain will express Apoe. Q. Xu et al. (2006) also proposed that, depending on where Apoe is being expressed, it could serve different functions; for example, in damaged neurons, Apoe might be involved in mitochondrial dysfunction and neurofibrillary tangle formation, and in astrocytes, it may be involved in the formation of amyloid plaques. Other papers have demonstrated that APOE is expressed in human (P. T. Xu et al., 1999) and rodent (Boschert et al., 1999; Harris et al., 2004) neurons, suggesting that the APOE protein functions in both astrocytes and neurons to regulate circuit function. Elucidating where Apoe is being expressed in brain regions that are involved in attention and memory may shed light on what Apoe's role is in those circuits and which metabolic products or molecules related to neuropsychiatric conditions Apoe is involved in producing. Our finding that activation of one such circuit induces upregulation of Apoe in neurons, and not astrocytes, is a step toward this goal.
We found that there were no statistically significant differences in the average number of Apoe transcripts per putative excitatory (Slc17a7-expressing) neuron, as well as the proportion of Slc17a7-expressing neurons that coexpressed Apoe between DREADD-expressing and control mice. These results suggest that excitatory neurons are therefore not responsible for the upregulation of Apoe found upon activation of LC→PrL projection neurons. We instead observed that a significantly greater average number of inhibitory neurons coexpressed Apoe in DREADD-expressing mice and that the average number of Apoe transcripts in Gad1-expressing cells was significantly higher in DREADD-expressing mice compared with controls. We therefore conclude that inhibitory neurons are responsible for the upregulation of Apoe following depolarization of PrL-projecting neurons in the LC. Inhibitory neurons are involved in sculpting local networks; they mostly lack long range projections and instead influence the activity of local circuitry (Moore, 1993). The major neurotransmitter used by inhibitory neurons is gamma-aminobutyric acid (GABA). In the present study, we used Gad1 as a general marker for inhibitory neurons. The Gad1 gene encodes for glutamic acid decarboxylase (GAD), which is an enzyme that catalyzes the decarboxylation of glutamate in GABA synthesis (Bu et al., 1992).
Previous research suggests that there is increased susceptibility of GABAergic interneurons to APOE −e4-related pathology (Najm et al., 2019). There exists an interesting connection between APOE allele type and GABAergic inhibitory activity, in that APOE −e4 has been associated with hyperactivity of brain function, measured through BOLD signals, in young carriers (Filippini et al., 2011). It is possible that this reflects a difference in GABAergic inhibitory neuronal function, especially given the association between APOE −e4 and subclinical epileptiform activity that can occur in patients without dementia when experiencing stress (Palop and Mucke, 2009; Andrews-Zwilling et al., 2010). Dysfunction in GABAergic neurons is also a hallmark of many neuropsychiatric disorders, including schizophrenia (Gonzalez-Burgos et al., 2010; Cohen et al., 2015), ADHD (Edden et al., 2012; Ferranti et al., 2024), and MDD (Luscher et al., 2011). At a molecular level, APOE −e4 undergoes proteolytic cleavage in neurons, which generates neurotoxic fragments (Brecht et al., 2004). These fragments eventually lead to tau phosphorylation, a key component, alongside amyloid beta plaque formation, of AD (Brecht et al., 2004; Najm et al., 2019). G. Li et al. (2009) found that levels of neurotoxic Apoe fragments and tau phosphorylation were elevated in the hippocampal neurons of mice that expressed APOE −e4. They also found decreased levels of GABAergic interneuron survival in these mice. This finding demonstrates that GABAergic interneurons, specifically in brain regions like the hippocampus, are susceptible to APOE −e4-derived neurotoxic fragments and tau pathology (Andrews-Zwilling et al., 2010). Both factors contribute to decreased survival of GABAergic interneurons, which also leads to learning and memory deficit characteristic of AD (Knoferle et al., 2014).
Because Gad1 is a general marker for inhibitory neurons, it cannot delineate which subtype of interneuron, if any, is responsible for the upregulation of Apoe found in this experiment. To investigate this, we examined Apoe expression in two major subtypes of inhibitory neuron: parvalbumin interneurons and somatostatin interneurons. Cortical parvalbumin interneurons mainly provide perisomatic inhibition to excitatory pyramidal neurons, while somatostatin interneurons mainly target pyramidal neuron dendrites (Hangya et al., 2014). We find that enrichment of Apoe is primarily observed in putative somatostatin interneurons following activation of PrL-projecting LC neurons. Furthermore, this effect is much stronger in female mice, revealing that cell type-specific expression of Apoe is sex dependent. Somatostatin interneurons in the cortex are widely implicated in a range of behaviors that are affected in neuropsychiatric disorders, such as working memory (D. Kim et al., 2016; Abbas et al., 2018), attention (Urban-Ciecko et al., 2018), and learning (Adler et al., 2019). Apoe may therefore act as a molecular regulator of function in these neurons during cognitive processing. Interestingly, previous research has demonstrated that the distribution of somatostatin interneurons in several brain regions is sexually dimorphic (Y. Kim et al., 2017). Deficits in somatostatin gene production in the cingulate cortex of patients with MDD is also more pronounced in female patients (Tripp et al., 2011; Seney et al., 2015), indicating that sex differences in somatostatin interneuron function may contribute to differences in the onset and severity of neuropsychiatric disorders between biological males and females. Further research is needed to uncover which interneuron subtypes underlie increases in Apoe expression observed in male mice in our study.
Taken together, our findings provide an interesting link between inhibitory neurons and activity-induced Apoe expression in the prelimbic subregion of the medial frontal cortex. Although the research surrounding Apoe and GABAergic interneurons has traditionally focused on learning and memory deficits and allele-specific effects, it is possible that a connection exists between Apoe, GABAergic interneuron function, and cognitive deficits more broadly. Loss of GABAergic interneurons in APOE −e4 carriers is associated with learning and memory deficits, but these occur after attentional deficits arise during the progression of AD. Therefore, it is possible that APOE-related GABAergic interneuron loss is occurring in brain regions that are involved in attention and memory, such as the PrL and the LC. Our results suggest that cell type-specific regulation of Apoe might be useful as a diagnostic biomarker of circuit function in neuropsychiatric and neurodegenerative disorders. A crucial next step would be to examine APOE transcription in cingulate cortex interneurons in patients with disorders that present with attention and memory symptoms, such as AD, schizophrenia, and ADHD. Importantly, although the structure of the mouse Apoe gene and human APOE gene are different, a basic understanding of how and where Apoe is expressed in specific circuits is a foundational step toward illuminating how human alleles might impact cellular function to affect cognitive domains such as attention and memory.
Footnotes
A.E.J. is currently an employee and shareholder of Neumora Therapeutics, which is unrelated to the contents of this manuscript. All other authors declare no competing financial interests.
We thank Aimee Ormond, Deveren Manley, and Amy Badillo for animal care assistance. H.L.H. acknowledges support from a Brain and Behavior Research Foundation (BBRF) Young Investigator Award, K.M. acknowledges support from National Institute of Mental Health (NIMH) R01MH105592, and both H.L.H. and K.M. acknowledge support from NIMH R21MH130066.
This is an open-access article distributed under the terms of the Creative Commons Attribution 4.0 International license, which permits unrestricted use, distribution and reproduction in any medium provided that the original work is properly attributed.
References
In this issue
