Abstract
Psychosis is a hallmark symptom of schizophrenia and highly prevalent
in bipolar disorder. Previous work has shown altered gene expression
within subregions of the striatum in subjects with psychosis, but it is
unclear if these alterations differ across subregions. Moreover,
despite known sex differences in the presentation of psychosis (such as
age of onset and disease course), it is unclear if there are sex
differences in gene expression across subregions of the human striatum
in the context of psychosis. Using RNA-sequencing data from human
postmortem nucleus accumbens (NAc), caudate, and putamen, we first
performed differential expression analyses across these striatal
subregions in unaffected (n = 60) and psychosis (n = 36) subjects. For
analysis of sex differences, we used equal numbers of males and females
in each subject group and evaluated sex and psychosis effects within
each brain region. We found that the NAc is the most transcriptionally
unique region compared to the caudate and the putamen in both psychosis
and unaffected subjects. We also found distinct patterns in gene
expression across the three striatal subregions, with an altered
pattern of cilia-related genes in subjects with psychosis. Our
sex-based analyses showed a striking discordant expression pattern,
with opposite effect directions between male and female subjects with
psychosis in all three subregions, including a reversal of sex
differences in immune- and angiogenesis-related pathways. Overall, we
identified regional and sex differences in gene expression across the
human striatum that may underlie sex-specific striatal dysfunction and
symptomatology in psychosis.
Subject terms: Molecular neuroscience, Genomics
Introduction
Psychosis is a hallmark of schizophrenia (SCZ) and a common feature in
type 1 bipolar disorder (BD). It manifests as a severe disconnect from
reality, often involving delusions, hallucinations, or incoherent
speech. Psychosis is intricately linked to dysfunction, particularly
elevated dopamine transmission, in the striatum [[48]1]. The human
striatum is subdivided into three regions with both overlapping and
distinct functions [[49]2]. The ventral striatum is primarily composed
of the nucleus accumbens (NAc), which is involved in limbic regulation
of motivation, reward processing, and goal-directed behavior [[50]2].
Two anatomically and functionally similar subregions compose the dorsal
striatum: the caudate nucleus and putamen. Both play a role in
associative and sensorimotor processing, but the caudate is more
associated with anxiety, working memory, and conscious movement while
the putamen is the epicenter of repetitive behavior and motor control
[[51]3, [52]4]. While these regions have important functional
differences, little is known about differences in transcript expression
across striatal subregions.
In previous work, we identified differential expression of immune- and
inflammation-related genes in subjects with psychosis compared to
unaffected subjects in each striatal subregion [[53]5]. We also
identified region-specific findings, including a decrease in the
expression of mitochondrial transcripts in the NAc of subjects with
psychosis who died at night [[54]5]. While these studies identified
important gene expression differences in subjects with psychosis within
striatal subregions, it is unclear how gene expression is altered
between each of the striatal regions and the pathways that are most
impacted by these differences. Moreover, despite known sex differences
in SCZ and BD, the existence of sex differences in the striatal
transcriptome of subjects with psychosis is currently unknown. Evidence
suggests that females tend to have a later age of onset of SCZ, less
positive symptoms such as psychosis (although findings are mixed)
[[55]6], respond better to antipsychotics, and have an overall better
course of disease [[56]7]. Similar sex differences are observed in
subjects with BD, particularly those with psychosis symptoms [[57]8].
Interestingly, recent studies have identified sex-specific alterations
in gene expression in the human prefrontal cortex of subjects with SCZ
and BD compared to unaffected subjects [[58]8–[59]10]. Additionally, a
recent study found sex-specific expression differences in the caudate
of subjects with SCZ [[60]11]. Though, to our knowledge, no study has
investigated sex differences in gene expression across all three
substriatal regions of subjects with psychosis. Therefore, in the
current study, we aimed to determine how gene expression is altered
across striatal subregions in psychosis and unaffected comparison
subjects, while integrating an exploratory sex-based analysis to gain
insight into how sex contributes to expression differences across the
striatum in subjects with psychosis.
Methods
Ethics approval and consent to participate
Previously, RNA-sequencing data from human postmortem striatum was
generated at the University of Pittsburgh by our lab and used in
separate analyses published in Ketchesin et al., 2021 [[61]12] and 2023
[[62]5] (NCBI GEO accession no. [63]GSE160521 and [64]GSE202537).
Briefly, NAc, caudate, and putamen samples were collected from subjects
with psychosis (n = 36; SCZ/schizoaffective disorder [n = 28] and BD
with psychosis [n = 8] and unaffected subjects (n = 59) obtained
through the University of Pittsburgh Brain Tissue Donation Program and
National Institutes of Health NeuroBioBank, following informed consent
from the next of kin. All methods and procedures were approved by and
conducted in accordance with the guidelines of the University of
Pittsburgh Institutional Review Board for Biomedical Research and
Committee for Oversight of Research and Clinical Training Involving
Decedents.
RNA sequencing and data preprocessing
Total RNA from the NAc, caudate, and putamen was extracted and
processed as described previously [[65]5, [66]12]. Briefly, total RNA
was extracted from the striatal samples using both TRIzol (Invitrogen)
and RNeasy Lipid Tissue Mini Kit (Qiagen). Libraries were prepped for
RNA-seq using the TruSeq Stranded Total RNA Sample Preparation Kit
(Illumina). Paired-end dual-indexed sequencing (75 bp) was performed
using the NextSeq 500 platform (Illumina) for an average of ~46 million
paired end reads/sample. RNA-seq count data were transformed to log2
continuous counts per million (cpm) data using the cpm function of the
Bioconductor edgeR package [[67]13, [68]14]. Transcripts were retained
for analysis if log[2](cpm) was greater than 1 in 50% or more of
subjects in at least one brain region. All Y-chromosome genes were
eliminated from analysis.
Differential expression analyses
Due to the limited number of available female subjects, the sample size
is relatively small to comprehensively assess the interactions among
our primary variables of interest (i.e., brain region, disease, and
sex) within a single analytical model. Therefore, we adopted a
stratified analytical approach, comprising two separate differential
expression analyses: one stratified by disease to examine brain region
effects and the other by brain region to examine sex and disease
effects.
Region-based analysis
The differential expression analysis stratified by psychosis and
unaffected cohorts focused on the effects of brain region. To address
the confounding between brain region and batch effects, we utilized the
RUV-4 algorithm from the “ruv” R package [[69]15], designed for
adjusting unwanted variation while preserving the effects of interest
using negative controls – 34 striatum housekeeping genes in our case
selected from a housekeeping gene database [[70]16]. To verify the
adequacy of RUV-4 normalization in eliminating batch effects without
compromising brain region effects, Principal Component Analysis (PCA)
plots were generated before and after RUV-4 normalization for the 34
striatum housekeeping genes and the top 2000 variably expressed genes
using the “PCAtools” R package. Within each cohort, a linear mixed
model was fitted for each gene using the “lme4” R package:
[MATH: Yij=ui+β1×Ibrainij=NAc+β2×I
brainij=caudate+β3×I<
/mi>brainij=putamen+γ1×co
v1+γ2
×cov2+γ3×cov3
+ϵij :MATH]
, where
[MATH: Yij :MATH]
is the expression level for subject
[MATH: i :MATH]
and brain region
[MATH: j :MATH]
.
[MATH: ui
:MATH]
is the random intercept. Up to three covariates
[MATH: (cov1,cov2,cov3) :MATH]
are allowed to be selected by BIC from time of death, RIN age, race,
and sex to further control for other potential confounders. The
p-values returned from the feature selection are biased because of
model differences between each gene. Thus, a permutation test was used
to return corrected empirical p-values. We further corrected for
multiple comparisons using a Benjamini-Hochberg procedure. We
identified two major patterns of transcript expression in the regional
analyses: 1) distinct expression in the NAc relative to caudate and
putamen and 2) patterns of expression across the three striatal
regions. For distinct NAc expression, transcripts were considered
differentially expressed (DE) if there was a main effect of brain
region (q < 0.05), two-region comparison post-hoc q < 0.05, and log[2]
fold change < −0.58 or > 0.58 (fold change ± 1.5 or 50% expression
change). For the expression patterns, transcripts were considered DE if
there was a main effect of brain region (q < 0.05), two-region
comparison post-hoc q < 0.05, and log[2] fold change < −0.26 or > 0.26
(fold change ± 1.2 or 20% expression change). A less stringent fold
change cutoff was used here to create a hierarchy in expression across
the three subregions. Based on these parameters, we gathered the list
of transcripts that were DE in the caudate compared to the putamen, the
caudate compared to NAc, and the NAc compared to putamen. Then, we
individually ordered each transcript by region of highest to lowest
expression. This grouped transcripts that had the highest expression in
one region, second highest in another, and lowest expression in the
final region, all with significant differences between each region.
Sex-based analysis
To evaluate sex-based differences accounting for the limited number of
female subjects, we created pairs of male and female subjects, with
pairs matched for age, race, time of death, post-mortem interval, and
the RNA integrity number (psychosis: n = 10/sex; unaffected:
n = 11/sex). To assess the interplay between disease and sex across
different brain regions, a linear model incorporating an interaction
term i.e.,
[MATH:
Yi=<
/mo>β0+β1×
I(sexi=male)+β2×I(groupi<
/mi>=psychosis)+
β3×I(sexi=male,groupi<
/mi>=psychosis)+
ϵi :MATH]
was performed in each region using the Linear Models for Microarray
Data (LIMMA) approach (from “limma” R package) [[71]14]. Given the
smaller sample size, transcripts were considered differentially
expressed if p < 0.01 and log[2] fold change < −0.26 or > 0.26 (fold
change ± 1.2 or 20% expression change). We examined the interplay of
sex, region, and psychosis by investigating expression patterns from
two different angles: 1) within each sex, what is the effect of
psychosis? 2) within each disease group, what is the effect of sex? In
the former, we compare unaffected males to psychosis males and
unaffected females to psychosis females (Table [72]S2). In the latter,
we compare male unaffected subjects to female unaffected subjects and
male psychosis subjects to female psychosis subjects (Table [73]S2).
Results
Differential expression across striatal subregions in unaffected and
psychosis subjects
We measured differential expression between the three striatal
subregions separately in unaffected and psychosis subjects (Table
[74]S1). After RUV-seq normalization of expression values, there was a
distinct separation in the principal components clustering the caudate
and the putamen and the NAc in unaffected subjects (Figure [75]S1).
Then, we performed a differential expression analysis. Genes were
considered to have an increased or decreased expression based on a
p < 0.05 and a log[2] fold change < −0.58 or > 0.58 (fold change ± 1.5
or 50% expression change) cutoff. In unaffected subjects, there were
significantly fewer transcripts differentially expressed between the
caudate and putamen (361; q < 0.05) versus the caudate or putamen
compared to the NAc (1836 and 1691, respectively) (Fig. [76]1A; Dataset
[77]S1). Consistent with the regional differences in expression, many
transcripts had significantly higher expression in the NAc, with
similar expression between the caudate and putamen ([N > (C = P)] (574)
or significantly lower expression in the NAc and similar expression in
the caudate and putamen [N < (C = P)] (610) (Fig. [78]1B–D). Unique
GPCR pathways and a distinct alternative splicing pathway (snRNP
pathway) were overrepresented in transcripts that followed an
N > (C = P) pattern, while transcripts that followed an N < (C = P)
pattern had an overrepresentation of stimulus-response and other GPCR
pathways (Fig. [79]1E).
Fig. 1. Differential expression patterns across the nucleus accumbens (N),
caudate (C), and putamen (P) in unaffected subjects.
[80]Fig. 1
[81]Open in a new tab
A Table of the number of transcripts that meet the criteria of being
differentially expressed between two regions based on a cutoff of a
main effect of brain region (q < 0.05), two-region comparison post-hoc
q < 0.05, and a log[2] fold change < −0.58 or > 0.58 (fold change ± 1.5
or 50% expression change). B Table of the number of transcripts that
follow certain expression patterns based on a q-value and log[2] fold
change. For the top two patterns, transcripts are considered
differentially expressed if they meet a cutoff of main effect of brain
region (q < 0.05), two-region comparison post-hoc q < 0.05, and a
log[2]fold change < −0.58 or > 0.58. For the following five expression
patterns (hierarchy in expression across the three striatal
subregions), transcripts are considered differentially expressed if
they meet a cutoff of a main effect of brain region (q < 0.05),
two-region comparison post-hoc q < 0.05, and a log[2] fold change <
−0.26 or > 0.26 (fold change ± 1.2 or 20% expression change). C Heat
maps for two of the expression patterns in which the input is the list
of genes corresponding to the N > (C = P) or N < (C = P) patterns.
Expression values are mapped for each region, where the green column
represents NAc, pink Caudate, and light blue Putamen. Within each
larger column, every individual column represents a subject and every
row represents a transcript. Yellow represents high expression and blue
low expression. D Representative scatterplots of the expression of two
transcripts. The scatterplot on the left follows a N > (C = P) pattern
and the one on the right represents a N < (C = P) pattern. Each
individual dot represents a subject. The mean expression value for all
subjects is indicated by a black line, with error bars indicating the
standard error of the mean. E Heatmaps of GO Biological Process
enrichment using the web-based portal Metascape for the unique NAc
enrichment (left) and expression patterns (right).
We also identified transcripts that followed distinct expression
patterns, with highest expression in one region, intermediate in
another, and lowest in the third (Fig. [82]1B, S2). The caudate > NAc >
putamen pattern (C > N > P) and NAc > caudate > putamen (N > C > P)
pattern were both enriched for ciliary pathways (though the latter’s
association is modest), including the axonemal dynein complex assembly
and extracellular matrix organization (Fig. [83]1E). Overall, this
suggests that ciliary-related pathways have the lowest expression in
the putamen. However, there do seem to be regional differences in
cilia-related pathways across the striatum, as enrichment was strongest
for the C > N > P pattern (Fig. [84]1E).
Differential expression across substriatal regions in psychosis subjects
In psychosis subjects, we found similar patterns in which the number of
transcripts that were differentially expressed between the caudate and
putamen (241) was strikingly lower than the number of transcripts that
were differentially expressed between NAc and caudate (1819) and
between NAc and putamen (1933) (Fig. [85]2A; Dataset [86]S2).
Additionally, the principal component analysis had a similar clustering
pattern compared to the unaffected subjects (Figure [87]S3). As seen in
the unaffected subjects, there is a relatively large list of
transcripts that follow the N > (C = P) (722) and N < (C = P) (579)
patterns (Fig. [88]2B–D).
Fig. 2. Differential expression patterns across the nucleus accumbens (N),
caudate (C), and putamen (P) in subjects with psychosis.
[89]Fig. 2
[90]Open in a new tab
A Table of the number of transcripts that meet the criteria of being
differentially expressed between two regions based on a cutoff of a
main effect of brain region (q < 0.05), two-region comparison post-hoc
q < 0.05, and a log2 fold change < −0.58 or > 0.58 (fold change ± 1.5
or 50% expression change). B Table of the number of transcripts that
follow certain expression patterns based on a q-value and log2 fold
change. For the top two patterns, transcripts are considered
differentially expressed if they meet a cutoff of a main effect of
brain region (q < 0.05), two-region comparison post-hoc q < 0.05, and a
log2 fold change < −0.58 or > 0.58. For the following five expression
patterns (hierarchy in expression across the three striatal
subregions), transcripts are considered differentially expressed if
they meet a cutoff of a main effect of brain region (q < 0.05),
two-region comparison post-hoc q < 0.05, and a log2 fold change < −0.26
or > 0.26 (fold change ± 1.2 or 20% expression change). C Heat maps for
two of the expression patterns in which the input is the list of genes
corresponding to the N > (C = P) or N < (C = P) patterns. D
Representative scatterplots of the expression of two transcripts. The
scatterplot on the left follows a N > (C = P) pattern and the one on
the right represents a N < (C = P) pattern. E Heatmaps of GO Biological
Process enrichment using the web-based portal Metascape for the unique
NAc enrichment (left) and expression patterns (right).
Pathway analysis revealed that the NAc is transcriptionally unique
across regions in both psychosis and unaffected subjects, and similar
pathways were enriched in psychosis subjects and unaffected subjects.
In psychosis subjects, stimulus-response- and GPCR-associated
transcripts follow an N < (C = P) expression pattern, and in the
N > (C = P) expression pattern, there is a unique enrichment of
alternative splicing pathways (snRNP) (Fig. [91]2E). Venn diagrams and
circos plots further point to similarities in the genes DE in the NAc
but not in the caudate nor putamen in unaffected and psychosis
subjects, with a high overlap in both the number of transcripts and the
pathways (Figure [92]S5). Additionally, when examining pathway
enrichment, there is a large degree of overlap between psychosis and
unaffected subjects, including GPCR and synaptic signaling related
pathways in the N < (C = P) pattern and alternative splicing related
pathways for the N > (C = P) pattern (Figure [93]S5).
As with the unaffected subjects, we identified expression patterns in
psychosis subjects (Fig. [94]2B, S4). While ciliary pathways are mostly
strongly overrepresented in transcripts with the C > N > P pattern in
unaffected subjects, those pathways are most enriched in the N > C > P
expression pattern in subjects with psychosis (Fig. [95]2E). Upon
further examination, we found that ciliary-related genes that follow a
C > N > P expression pattern in unaffected subjects followed a
(C = N) > P pattern in psychosis subjects and ciliary-related genes
that followed a N > C > P pattern in psychosis subjects followed a
(C = N) > P pattern in unaffected subjects (Figure [96]S6A-B),
suggesting a potential reorganization of cilia-related signaling across
the striatum in psychosis.
Effect of psychosis across sex
We examined the interplay of sex, region, and psychosis by
investigating expression patterns from two different angles: 1) within
each sex, what is the effect of psychosis? 2) within each disease
group, what is the effect of sex? In the former, we compared males with
psychosis to unaffected males and females with psychosis to unaffected
females (Table [97]S2). In the latter, we compared male unaffected
subjects to female unaffected subjects and male subjects with psychosis
to female subjects with psychosis (Table [98]S2).
To examine the effect of psychosis in each sex, we created lists of
genes with increased or decreased expression in psychosis relative to
unaffected subjects; analyses were performed separately in each
striatal region. Genes were considered to have an increased or
decreased expression based on a p < 0.01 and a log[2] fold change <
−0.26 or > 0.26 (fold change ± 1.2 or 20% expression change) cutoff.
The effect of psychosis was especially pronounced in females,
particularly in the NAc, where >300 genes had a higher or lower
expression in psychosis compared to unaffected subjects (Fig. [99]3A;
Table [100]S4; Dataset [101]S3). Next, we generated circos plots to
examine overlap in genes differentially expressed within each sex
across striatal regions (Fig. [102]3B). In males, there was sparse gene
and pathway overlap across striatal regions. In contrast, the effect of
psychosis in females was more consistent across regions. These findings
were reflected in pathway enrichment analysis (Fig. [103]3C). In males,
there were some interesting pathways, including lower expression of
various immune-related pathways in the putamen and caudate, but there
was little overlap in pathways across regions. But in females there was
a striking enrichment for angiogenesis and immune- and
inflammation-related pathways for transcripts with a higher expression
in psychosis, with overlap across the three striatal regions (Fig.
[104]3C). Notably, the NAc was unique in that mitochondrial-related
transcripts (e.g., cellular respiration) had a lower expression in
female subjects with psychosis.
Fig. 3. Effects of psychosis in each sex.
[105]Fig. 3
[106]Open in a new tab
A Table of the number of transcripts that had higher or lower
expression in psychosis in each sex, stratified by striatal subregion.
B Circos plots representing the overlap of transcripts (purple) or
biological processes (blue) generated from the list of transcripts in
Table (A). C Heatmaps of GO Biological Process enrichment from
Metascape for males (top) and females (bottom). D RRHO plots that
visualize the overlap between two lists of transcripts ranked by their
p-values and effect size (with warmer colors indicating a high degree
of overlap). Each quadrant visualizes the overlap between different
sets of transcripts in the NAc (left), caudate (middle), or putamen
(right). In the RRHO interpretation legend, the arrows refer to changes
in male or female subjects with psychosis.
After finding that there was immune pathway enrichment in females with
psychosis and less enrichment of similar pathways in males with
psychosis compared to unaffected subjects, we wanted to further
investigate potential discordance of genes between groups. We
investigated whether this pattern of opposite effect directions was
consistent in gene expression regardless of threshold using a
threshold-free rank-rank hypergeometric (RRHO) analysis. RRHO plots
indicated a discordant expression pattern of psychosis effect between
male and female subjects (Fig. [107]3D). Strikingly, the genes with a
higher expression in females with psychosis have a lower expression in
males and vice versa across all substriatal regions.
Effect of sex across conditions
To put the discordant sex-specific effects of psychosis into context,
we next examined whether genes with sex differences in unaffected
subjects exhibit similar sex differences in subjects with psychosis. To
investigate this, in each substriatal region we first determined which
genes exhibited a sex difference in unaffected subjects, but not in
subjects with psychosis (“loss of sex difference”) and which genes did
not exhibit a sex difference in unaffected subjects but did have a sex
difference in subjects with psychosis (“gain of sex difference”).
Overall, while there were some genes that lost a sex difference in
psychosis compared to unaffected subjects, many more gain a sex
difference in psychosis, particularly driven by an increase in gene
expression in female subjects with psychosis across all striatal
subregions (Fig. [108]4A; Table [109]S4; Dataset [110]S4).
Fig. 4. Gain and loss of sex differences in psychosis.
[111]Fig. 4
[112]Open in a new tab
A Table of the number of transcripts that lost or gained a sex
difference in psychosis, separated by striatal subregion. For
transcripts that lost a sex difference, M > F (unaffected) are the
transcripts that had higher expression in male compared to female
unaffected subjects, but no significant change in psychosis subjects.
F > M (unaffected) conversely are those transcripts that had a higher
expression in females compared to males in unaffected subjects, but no
significant changes in psychosis subjects. For transcripts that gain a
sex difference in psychosis and are in the M > F (psychosis) column,
they did not have a sex difference in unaffected subjects but had a
higher expression in male compared to female psychosis subjects. The
same is true of female subjects with psychosis in the F > M (psychosis)
category. B Circos plots representing the overlap of transcripts
(purple) or pathways (blue) between the list of transcripts derived
from the table in part (A). C GO Biological Process enrichment from the
genes in Table (A) for loss (top) or gain (bottom) of sex differences.
D RRHO plots that map the overlap between two lists of transcripts
ranked by their p-values and effect size (with warmer colors indicating
a high degree of overlap). Each quadrant visualizes the overlap between
different sets of transcripts in the NAc (left), caudate (middle), or
putamen (right). In the RRHO interpretation legend, the arrows refer to
changes in male control or psychosis subjects.
Interestingly, there was little overlap in genes and pathways that lost
a sex difference in psychosis compared to unaffected controls between
regions but a high degree of overlap between regions in genes and
pathways that gained a sex difference in psychosis compared to
unaffected subjects (Fig. [113]4B). This suggests that sex differences
in gene expression are more consistent across regions in subjects with
psychosis. Pathway analyses revealed immune- and angiogenesis-related
transcripts in the putamen exhibited a loss of sex difference in
psychosis compared to unaffected subjects driven by alterations in gene
expression in males with little overlap across regions (Fig. [114]4C).
For transcripts that gained a sex difference in psychosis compared to
unaffected subjects, there was enrichment in myelination-associated
transcripts driven by alterations in gene expression in males and an
enrichment in angiogenesis-related pathways driven by differences in
gene expression in females (i.e., tube morphogenesis) across all
striatal subregions (Fig. [115]4C).
RRHO analysis revealed a discordant expression pattern between
transcripts that show a sex difference in psychosis and unaffected
subjects, an effect that was strongest in the NAc and putamen. This
suggests that transcripts that have a higher expression in male
unaffected subjects have a lower expression in male subjects with
psychosis and vice versa.
Reversal of sex differences in psychosis
Given the striking discordance revealed by the RRHO analyses for both
the effect of psychosis across sex and the effect of sex across
psychosis and unaffected subjects, we next sought to integrate these
analyses to further investigate the consistency of this discordant
expression pattern across striatal regions. To do so, we combined
expression data across regions and performed two adaptively weighted
Fisher’s method (AW-Fisher) meta-analyses to generate a list of genes
with an opposite effect of psychosis across sex and a list of genes
with an opposite effect of sex across conditions (Fig. [116]5; Table
[117]S5; Dataset [118]S5).
Fig. 5. Reversal of sex differences in psychosis.
[119]Fig. 5
[120]Open in a new tab
A, B Heat maps of biological process enrichment from AW-Fisher
meta-analyses. Two meta-analyses were performed, one for the effects of
psychosis in each sex (psychosis vs control in each sex; left column of
each heat map) and one for sex differences (male vs female in each
disease group; right column of each heatmap). Heat maps in A and B show
the overlap in biological processes between the meta-analyses, with
distinct patterns for each heat map. C Scatterplots of a representative
gene (IL1R1) in the caudate from the M ↓ F↑ psychosis disease
meta-analysis in heat map A. This gene is plotted a different way in
the right panel with male and female grouped together to show the
reversal in sex differences. D Scatterplots of a representative gene
(TMEM52B) in the caudate from the M ↑ F↓ psychosis disease
meta-analysis in heat map B. This gene is plotted a different way in
the right panel with male and female grouped together to show the
reversal in sex differences.
In total, there were 643 genes with an opposite effect of psychosis
across sex (Table [121]S5). As in, they either had lower expression in
male subjects with psychosis compared to male unaffected subjects and
higher expression in female subjects with psychosis compared to female
unaffected subjects (339) or higher expression in male subjects with
psychosis compared to male unaffected subjects and lower expression in
females subjects with psychosis compared to female unaffected subjects
(304). We found 745 genes with an opposite effect of sex across
conditions. One pattern consisted of genes that had higher expression
in male unaffected subjects compared to female unaffected subjects and
higher expression in female subjects with psychosis compared to male
subjects with psychosis (389). The other pattern consisted of genes
with the opposite direction, higher expression in female unaffected
subjects compared to male unaffected subjects and higher expression in
male subjects with psychosis compared to female subjects with psychosis
(356; Table [122]S5).
We were interested to see if there was any overlap between the genes
that show an opposite effect of psychosis across sex and the genes that
show an opposite effect of sex across conditions, which would indicate
a reversal of sex differences for that gene in the striatum. Notably,
there was a considerable degree of overlap in the number of genes with
opposite effect sizes between the psychosis and sex meta-analyses
(Table [123]S6), leading us to investigate overlap at the pathway level
(Fig. [124]5A, B). Pathway analyses for the first psychosis effect
pattern (Fig. [125]5A; lower in males, higher in females with
psychosis) showed an enrichment for immune-, angiogenesis-, and
vasculogenesis-related pathways, consistent with pathway enrichment
observed in Fig. [126]3C. Interestingly, pathway analyses for the
corresponding sex effect pattern (enriched in male controls and female
psychosis) (Fig. [127]5A) showed a high degree of overlap in pathway
enrichment. Pathway analyses for the second psychosis effect pattern
(Fig. [128]5B; higher in males, lower in females with psychosis)
revealed enrichment for pathways such as inorganic ion transmembrane
transport, as well as various mitochondrial and cytoskeletal related
pathways. Remarkably, pathway analyses for the corresponding sex effect
pattern (enriched in female controls and male psychosis) also showed
considerable overlap with the psychosis effect pattern. Scatterplots of
individual representative genes from each of the expression patterns
(Fig. [129]5C, D, S7), along with the overlapping pathways, suggest
that the opposing psychosis effect expression patterns was driven by a
reversal in sex differences.
Discussion
There has been extensive research that has pointed to the striatum
being implicated in psychosis. Previous work from our group identified
an upregulation in immune- and inflammation-regulated genes within
striatal subregions in psychosis when comparing subjects with psychosis
to unaffected subjects [[130]5]. While these findings highlight
critical gene expression differences within individual striatal
subregions, it remains unclear how gene expression differs between
these regions in both psychosis and unaffected groups. The current
study sought to address this through exploring region- and sex-based
gene expression differences in human postmortem brain tissue of
psychosis and unaffected subjects. We used previously generated
RNA-sequencing data [[131]5, [132]12] from the NAc, caudate, and
putamen from the same subjects and performed two differential
expression analyses: one stratified by condition (psychosis or
unaffected) to examine expression changes across striatal regions, and
one stratified by brain region to examine sex and disease effects.
First, we found that the NAc is the most transcriptionally unique
region in both psychosis and unaffected subjects. The transcripts that
were uniquely differentially expressed in the NAc in both of these
cohorts were very similar, with a marked increase in the expression of
alternative splicing-related pathways. In this case, the same set of
splicing-related genes in each condition had higher expression in the
NAc compared to the other regions: RNU5A-1, RNU5B-1, RNU4-2, RNU4-1,
RNU4ATAC, and RNU6ATAC. Furthermore, in both conditions noncoding RNAs
such as snoRNAs have higher expression in the NAc compared to caudate
and putamen and GPCR- and stimulus response-related pathways have a
lower expression. However, it is important to note that despite the
pathways being similar, there are some differences at the gene level.
For example, some GPCR-related genes that have lower expression in the
NAc in unaffected subjects, including ADORA2B, GPR37, TDRN, and VIP, do
not have lower expression in the NAc in psychosis subjects compared to
other substriatal regions. Some unique transcripts with lower
expression in the NAc compared to the caudate or putamen of psychosis
subjects include EDNRA, GLP1R, OPRD1, and TNNT1. Overall, the patterns
of expression and pathway enrichment are highly similar between groups,
suggesting that transcriptional markers of strong regional differences
are intact in subjects with psychosis.
We also identified a smaller set of transcripts that showed
hierarchical patterns of expression, in which a transcript is most
expressed in one region, second most in another, and least expressed in
the third. When examining the pathways associated with transcript
patterns, we found a notable difference between the expression patterns
in ciliary-related pathways in unaffected and psychosis subjects. In
unaffected subjects, ciliary-related transcript expression most
prominently follows a C > N > P pattern and in psychosis it follows a
N > C > P pattern. Cilia are cell surface projections that are
antenna-like cellular sensors that aid in cellular motility, detecting
mechanical or chemical stimuli, and regulating cell differentiation,
migration, and growth [[133]17]. In our data, many of the genes that
show an altered expression pattern, are involved in the function and
development of a part of cilia called dynein [[134]18]. Dynein in the
brain is crucial for axonal transport, and thus synaptic maintenance
and plasticity [[135]19, [136]20]. Our data suggest regional
differences in ciliary and axonal transport-related signaling across
the striatum, with altered regional prioritization in psychosis. Future
studies are necessary to determine the functional consequences of these
differences, as they may represent potential therapeutic targets. While
the role of cilia in psychosis remains understudied, many genes
implicated in ciliopathies are also associated with cognitive deficits.
Additionally, cilia have been linked to dopamine signaling in the
brain, suggesting that their disruption may contribute to altered
reward processing in psychosis [[137]21–[138]23]. Notably, 42% and 17%
of brain-expressed ciliary genes are differentially expressed in SCZ
and BD, respectively [[139]21].
Moving forward, we sought to further investigate the role of sex under
normal conditions and in psychosis. First, we established that there
are sex differences in the striatum of unaffected subjects, such as
higher expression of immune-, inflammation-, and angiogenesis-related
genes in males compared to females, which is consistent with previous
studies [[140]24]. Next, we identified sex differences in gene
expression across the striatum comparing psychosis and unaffected
subjects. Consistent with prior findings [[141]11], we found that the
effect of psychosis was particularly pronounced in the striatum of
female subjects. Notably, we found a profound decrease in the
expression of mitochondria-related transcripts in the NAc in female,
but not male subjects with psychosis. Previous studies have noted that
oxidative stress is implicated in SCZ pathogenesis and have
specifically found that the downregulation of NDUFV1 and CYC1
contributes to increased reactive oxygen species and thus increased
oxidative stress [[142]25]. Both these genes had a lower expression in
female subjects with psychosis compared to unaffected female subjects
in our study. Interestingly, increased estrogen in females, a known
regulator of mitochondrially expressed genes as well as nuclear-encoded
genes that impact mitochondrial function, and a hypothesized protector
in SCZ pathophysiology [[143]26], may also play a role in female
specific changes in gene expression with psychosis [[144]27]. This is
particularly important to consider because mitochondrial dysfunction
and oxidative stress have been repeatedly linked to psychosis and are
hypothesized to contribute to its neurobiological mechanisms [[145]28].
Future animal studies will be important for determining the mechanisms
that underly these sex-specific gene expression changes and how these
differences contribute to behavior.
Additionally, we saw a lower expression of immune pathways in male
subjects with psychosis and a higher expression of immune and
angiogenesis pathways in female subjects with psychosis. We sought to
examine whether these gene expression patterns were influenced by a
gain or loss of sex differences. We found that some immune- and
angiogenesis-related pathways had a high expression in unaffected males
compared to unaffected females but did not differ between sexes in
psychosis subjects. Other immune- and angiogenesis-related pathways,
though, were not differentially expressed in unaffected subjects but
had higher expression in female psychosis subjects compared to male
psychosis subjects. Consistent with the pathway findings, RRHO analyses
showed a striking discordant expression pattern with genes with a
higher expression in female subjects with psychosis having a lower
expression in males with psychosis and vice versa in all substriatal
regions. Noticing similar discordant effects in immune- and
angiogenesis-related pathways, we performed a meta-analysis to assess
the consistency of differential expression patterns across all
substriatal regions.
Markedly, our integrated meta-analysis identified over 600 genes with
opposite effect directions. Genes involved in immune and angiogenesis
pathways such as VEGFA, various tumor necrosis factor receptors
(TNFRs), and IL1R1 have an expression pattern in which unaffected male
expression is higher than expression in unaffected females but in
psychosis, it is higher in females compared to males. This may point to
neuroinflammation, which effects 30–50% of people with SCZ [[146]29]
and is correlated with lower cognitive processing [[147]30], as a
potential contributor to sex-specific differences in psychotic
symptoms. Briefly, both TNFα and IL1 activate and are activated by the
NF-κB pathway (a pathway in which we find the reversal of sex
differences in genes such as RELA and NFKB2), which promotes
inflammation [[148]31, [149]32]. Inflammation can lead to or be
partially caused by hypoxic-like conditions [[150]33]. Response to
hypoxia is one of the pathways with a higher expression in females with
psychosis, and included in that response is the activation of
angiogenesis pathways [[151]34], which includes the higher expression
of genes like VEGFA [[152]35]. Additionally, hypoxia is yet another
regulator of NF-κB [[153]36, [154]37]. Importantly, these genes are
involved both in the development of inflammation and in the response to
inflammation. The fact that we see a reversal of sex differences in
these pathways and genes linked to inflammation implicates striatal
inflammation as a potential contributor of sex differences.
There are other genes in our data, however, such as CD4, TGFβ, and IL18
that are known to be upregulated in SCZ [[155]29], that do not undergo
a reversal of sex difference and only lose a sex difference driven by
alterations in gene expression in males. These are genes that also
interact with the NF-κB pathway and are related to inflammation
[[156]32, [157]38, [158]39]. This speaks to the complexity of the sex
differences present in the pathway and transcriptomic landscape of
psychosis in the striatum. Regardless, it is interesting to note the
recurring presence of immune-, angiogenesis-, and inflammation-pathways
in our analysis that show sex differences in subjects with psychosis.
This is significant when considering how these pathways are involved in
aspects of psychosis known to have sex differences such as treatment
response [[159]40], cognitive function [[160]30], and psychosis
symptomology [[161]41].
These sex differences may be particularly relevant in treating
psychosis, as males and females respond differently to antipsychotic
drugs, with females tending to have better treatment outcomes but more
adverse side effects [[162]42]. Additionally, while many antipsychotics
have anti-inflammatory effects [[163]43], clozapine, used for
treatment-resistant schizophrenia, has been associated with
pro-inflammatory responses [[164]44, [165]45]. Recognizing sex
differences in immune and inflammatory pathways, along with the impact
of treatment on these pathways, may not only inform individualized
treatment strategies but also potentially highlights the importance of
monitoring inflammation levels during treatment.
One limitation of our study was the relatively small sample size for
our sex-based analyses where we were limited by the number of female
subjects. Therefore, these analyses are exploratory and use p-value
cutoffs (p < 0.01) as statistical thresholds, similar to what we have
reported previously [[166]5, [167]12]. Another limitation was our
inability to test for the effects of antipsychotics due to the small
sample size and lack of subjects with psychosis not on antipsychotics.
Future studies with larger sample sizes should consider antipsychotic
treatment history.
In conclusion, we have analyzed gene expression differences in the
striatum of unaffected and psychosis subjects to further elucidate the
role of this region in psychosis (summary of findings in Figure
[168]S8). We found that the NAc was the most transcriptionally unique
region with similar patterns in unaffected and psychosis subjects.
After examining substriatal region differences, we performed an
exploratory analysis of sex-based differences stratified by region. We
found an interesting reversal of sex differences in immune-,
inflammation-, and angiogenesis-related pathways. These findings may
have implications for disease pathology, but future studies will be
paramount in confirming these results in larger cohorts. The
implications of these findings should likewise be studied in cell
culture and animal models to determine how these gene expression
alterations affect cellular function and behavior.
Supplementary information
[169]Supplemental Material^ (23.7KB, docx)
[170]Supplemental Material^ (13.4MB, pdf)
[171]Data Set 1^ (4.1MB, xlsx)
[172]Data Set 2^ (4.3MB, xlsx)
[173]Data Set 3^ (9.3MB, xlsx)
[174]Data Set 4^ (8.1MB, xlsx)
[175]Data Set 5^ (196.2KB, xlsx)
Acknowledgements