Adenosine Kinase Expression in the Frontal Cortex in Schizophrenia

Abstract

The adenosine hypothesis of schizophrenia posits that reduced availability of the neuromodulator adenosine contributes to dysregulation of dopamine and glutamate transmission and the symptoms associated with schizophrenia. It has been proposed that increased expression of the enzyme adenosine kinase (ADK) may drive hypofunction of the adenosine system. While animal models of ADK overexpression support such a role for altered ADK, the expression of ADK in schizophrenia has yet to be examined. In this study, we assayed ADK gene and protein expression in frontocortical tissue from schizophrenia subjects. In the dorsolateral prefrontal cortex (DLPFC), ADK-long and -short splice variant expression was not significantly altered in schizophrenia compared to controls. There was also no significant difference in ADK splice variant expression in the frontal cortex of rats treated chronically with haloperidol-decanoate, in a study to identify the effect of antipsychotics on ADK gene expression. ADK protein expression was not significantly altered in the DLPFC or anterior cingulate cortex (ACC). There was no significant effect of antipsychotic medication on ADK protein expression in the DLPFC or ACC. Overall, our results suggest that increased ADK expression does not contribute to hypofunction of the adenosine system in schizophrenia and that alternative mechanisms are involved in dysregulation of this system in schizophrenia.

Introduction

Schizophrenia is a serious mental illness that affects approximately 1% of the world’s population and costs society billions of dollars each year.^1,2 Schizophrenia is characterized by 3 main type of symptoms, positive (eg, hallucinations, delusions), negative (eg, apathy) and cognitive symptoms (eg, inattention, difficulty completing tasks, memory problems).³ Dysregulation of the dopamine and glutamate neurotransmitter systems are implicated in the onset of these symptoms in schizophrenia.^4,5

The purinergic hypothesis of schizophrenia posits that altered adenosine activity could contribute to the pathophysiology of schizophrenia via its neuromodulatory actions on both the dopamine and glutamate transmitter systems.^6,7 Extensive preclinical work implicates the adenosine system in schizophrenia.^8,9 Knockout of astrocytic A_2A receptor induces schizophrenia-related features including cognitive impairment and altered psychomotor response to MK-801.¹⁰ Receptor knockout is sufficient to disrupt glutamate homeostasis, which is thought to underlie several endophenotypes relevant to schizophrenia.¹⁰ Behavioral pharmacology studies suggest that adenosine receptor agonists have antipsychotic-like effects in models of dopamine hyperfunction and glutamate NMDA receptor hypofunction, while adenosine receptor antagonists may prove to be effective in treating cognitive deficits, see review.⁸ In genetic models, primarily adenosine receptor knockout studies, deficits in adenosine affects dopamine and glutamate transmission and results in severe cognitive impairment, suggesting that imbalance in adenosine may result in endophenotypes associated with schizophrenia.⁸ A number of postmortem and clinical studies also support a role for the adenosine system in schizophrenia. Adenosine receptor A_2A density is significantly increased in the striatum in schizophrenia and may contribute to a hyperdopaminergic state.¹¹ Polymorphisms in adenosine A₁ receptor are implicated in schizophrenia in a Japanese population.¹² Others have reported that a low activity variant of adenosine deaminase, which metabolizes extracellular adenosine to inosine, is found at lower frequencies in schizophrenia patients and likely results in reduced levels of extracellular adenosine.¹³ Furthermore, purine degradation inhibitors and nucleoside transporter blockers, like allopurinol and dipyridamole, have had moderate success in treating symptoms of schizophrenia when administered with antipsychotics.^14–16 In addition to its role as a neuromodulator, adenosine, as the product of ATP hydrolysis, is directly related to energy consumption.¹⁷ Bioenergetic deficits are an established feature of schizophrenia^18,19 but the role of the adenosine system in energy metabolism in schizophrenia has yet to be fully elucidated.

The purinergic hypothesis was refined to suggest a role for hypofunction of the adenosine system in schizophrenia.^8,20,21 Adenosine kinase, a highly conserved ribokinase,²² is an important regulator of extracellular adenosine levels. ADK has a low K_m for adenosine and efficiently removes adenosine by converting it to AMP in a largely astrocyte-based cycle.²³ Small changes in ADK activity can result in significant changes in adenosine levels.²⁴ A mouse model of brain-wide ADK overexpression results in adenosine deficits and produces schizophrenia-relevant behavioral phenotypes, including working memory and attentional deficits that are improved by augmenting adenosine expression via ADK inhibition.^20,21

We hypothesize that, in line with the adenosine hypofunction hypothesis of schizophrenia, ADK is overexpressed in the frontal cortex in schizophrenia. In this study, we assay ADK expression in frontocortical tissue from subjects with schizophrenia. ADK gene expression in rodents treated chronically with haloperidol-decanoate was assayed to determine the effects of antipsychotic medication on ADK expression. In addition, we apply an in silico analysis of ADK gene expression in postmortem RNAseq and microarray studies to determine the effects of antipsychotic medication on ADK in schizophrenia subjects.

Materials and Methods

Subjects

Postmortem tissue from the dorsolateral prefrontal cortex (DLPFC) and anterior cingulate cortex (ACC) of schizophrenia subjects and controls was obtained from 3 brain banks. DLPFC was obtained from the Maryland Brain Collection (MBC) and the NIH Human Brain Collection Core (HBCC). ACC was obtained from the Bronx-Mt. Sinai NIH Brain and Tissue Repository (NBTR). The tissue was acquired with consent from next of kin with IRB approval protocols. Subjects were matched for age, sex, postmortem interval (PMI), and pH (Table 1). Two independent psychiatrists established DSM-IV diagnoses based on available medical records, autopsy reports, and interviews with family members using the Structured Clinical Interview for DSM-IIIR and DSM-IV. Subjects were excluded if they had a history of substance abuse, death by suicide, or were comatose for more than 6 hours before death. If subjects were on medication during the last 1–6 months of life, then antipsychotic medication status was determined to be “on”. Brains were stored at −80°C until dissected. Tissue from the MBC and NBTR were obtained as blocks of tissue approximately 1 cm³. Tissue from the NIH was provided as 14 µm thick sections on glass slides.

Table 1.

Subject Demographics

	DLPFC				ACC
Tissue Bank	MBC		HBCC		NBTR
Diagnosis	CTL	SCZ	CTL	SCZ	CTL	SCZ
N	20	20	46	26	12	12
Sex	3F/17M	3F/17M	17F/29M	13F/13M	6F/6M	6F/6M
Age	43.9 ± 9.1	44.9 ±10.5	42.2 ± 14.8	58 .2± 16.8	74.7 ± 10.6	75.0 ± 11.7
PMI (h)	12.9 ± 4.1	13.6 ± 5.6	35.3 ± 15.6	42.7 ± 21.8	9.1 ± 6.2	10.7 ± 4.6
pH	6.6 ± 0.3	6.6 ± 0.3	6.5 ± 0.3	6.4 ± 0.3	6.4 ± 0.2	6.2 ± 0.3
RIN	5.5 ± 1	6.1 ± 1	-	-	-	-
Medication	-	17/1/2	-	16/10/0	-	4/6/2

Note: CTL, control subjects; SCZ schizophrenia subjects; F, female; M, male; RIN RNA integrity number; DLPFC, dorsolateral prefrontal cortex; ACC, anterior cingulate cortex; MBC, Maryland Brain Collection; N, number of subjects; NBTR Bronx-Mt. Sinai NIH Brain and Tissue Repository, NIH Human Brain Collection Core (HBCC); PMI postmortem interval. Data presented as mean ± SD. Medication status is displayed as subjects on/off/unknown antipsychotics.

Animal Studies

All rat studies were performed in accordance to the IACUC guidelines at the University of Alabama at Birmingham. Adult male Sprague-Dawley rats (250 g) were pair-housed and maintained on a 12-hour light/dark cycle with ad libitum access to food and water. Rats received intramuscular injections of 28.5 mg kg⁻¹ haloperidol-decanoate or vehicle (sesame oil) once every 3 weeks for 9 months.^25,26 Brains were removed and stored at −80°C.

Sample Preparation

DLPFC and ACC tissue blocks were cryo-sectioned as necessary and mounted on 1 × 3 inch superfrost plus glass slides (Superfrost Plus glass slides, Fisher Scientific). Tissue sections (14 µm) were scraped from glass slides and homogenized in 60 µl of mammalian protein extraction reagent (MPERS) (#78501, ThermoFisher Scientific) with protease and phosphatase inhibitor (#78440, ThermoFisher Scientific). Total protein concentration was determined for each sample using a bicinchoninic acid assay.²⁷

Quantitative Polymerase Chain Reaction

RNA was isolated using the RNeasy Minikit (Qiagen) according to the manufacturer’s instructions. Complementary DNA (cDNA) was made using a High-Capacity cDNA Reverse Transcription Kit (4368814, Applied Biosystems, ThermoFisher Scientific). SYBR-Green quantitative real-time PCR reactions were performed in duplicate using 96-well plates (MicroAmp Fast Optical 96-well reaction plate, Applied Biosystems, ThermoFisher Scientific) on a StepOne Plus machine (Applied Biosystems, ThermoFisher Scientific). For each reaction, 3 μl of cDNA (diluted 1:3) was placed in a 20 μl reaction containing 10 μl of SYBR-Green PowerUp Master Mix (Applied Biosystems, ThermoFisher Scientific) and 3pmol of each primer (Invitrogen, ThermoFisher Scientific). PCR reactions occurred at 95°C for 3 minutes (3 cycles), 95°C for 15 seconds (1 cycle), and 59°C for 60 seconds (50 cycles). cDNA was omitted in the non-template control samples. No RT controls contained cDNA made without reverse transcriptase. Primer specificity was tested by running PCR product on a 1% agarose gel and submitting for Sanger sequencing or as previously described.²⁸ The primer sequences were as follows: Human: ADK Long (ADK-L)

F: CCTTCCCTCCAATCAGCACG; R: AGCAGGTACCACAGCAACTG.

ADK Short (ADK-S) F: TGGGCTGTA GAGCCAAAGTG; R: GCCCAAAAAGCTGAA GGTGG. β-actin F: AGTACTCCGTGTGGATCGGC; R: GCTGATCCACATCTGCTGGA. B2M F: GTGGGA TCGAGACATGTAAGC; R: AGCAAGCAAGCAGAA TTTGGAAT. Rat: ADK Long (ADK-L) F: CCAGAAGCGCTGAGTGAAAAT; R: GTCTTCGG CCAAGATCTGGT. ADK Short (ADK-S) F: ATGACGTCCACCAGTGAAAAT; R: GTCTTCGG CCAAGATCTGGT. CyclophilinA (PPIA) F: CTGCTTCGAGCTGTTTGCAG; R: GACCACAT GCTTGCCATCC. 18S F: CGCCGCTAGA GGTGAAATTC; R: TTGGCAAATGCTTTCGCTC.

Immunoblotting

Twenty-five micrograms of total protein in sample buffer (6X solution: 4.5% SDS, 15% β-mercaptoethanol, 0.018% Bromophenol blue, and 36% glycerol in 170 mM Tris-HCl pH 6.8) were loaded into 10 well pre-cast 4%–12% Bis-Tris gel (NuPAGE Invitrogen, ThermoFisher Scientific) and run at 180V for 1 hour in 1X MES buffer (Invitrogen, ThermoFisher Scientific). Protein loading amount was selected based off the linear part of the standard curve when running test strips. Following semidry transfer (Bio-Rad) to polyvinylidene fluoride (PVDF) membrane (Bio-Rad) at 20V for 45 minutes, membranes were blocked for 1 hour at room temperature in 5% milk in 1X tris-buffered saline with tween 20 (TBS-T) or Licor blocking buffer in 1X phosphate-buffered saline (PBS). Membranes were incubated in primary antibody mouse anti-ADK (1:1000, sc-365470, Santa Cruz Biotechnology), rabbit anti-ADK (1:4000, kindly provided by D. Boison²⁹), mouse anti-glial fibrillary acidic protein (GFAP) (1:1000, MAB360 EMD Millipore) or housekeeping controls mouse anti-valosin containing protein (VCP; 1:5000, ab11433, Abcam) or rabbit anti-beta-tubulin (1:500, ab18207, Abcam) overnight at 4°C. Membranes were then washed 3X in TBS-T and incubated with goat anti-mouse (1:5000, #68070, Licor) or donkey anti-rabbit (1 in 5000, #68073, Licor) secondary antibodies for 1 hour at room temperature. Membranes were then washed 3X in TBS-T and imaged on a Licor Odyssey laser based imaging system. Intensity values with all-segment median intra-lane background subtraction were measured for each band using Image Studio 4.0 software. Each band was normalized to intensity value of the calibrator sample (pool of all samples) present on each blot as an internal control. All samples were run in duplicate.

In Silico Analysis

“Lookup studies” were performed to examine ADK expression in publically available postmortem schizophrenia gene expression databases (figure 4A). The Stanley Medical Research Institute (SMRI) Online Genomics Database multi-study microarray repository was analyzed in silico^30,31 for ADK gene expression. The fold change and p-value for each search criteria: schizophrenia lifetime antipsychotics (within SCZ), heavy alcohol use (within SCZ), heavy drug use (within SCZ) and male and female sex effect (all subjects), is reported in figure 4B. Microarray studies from the DLPFC and ACC from postmortem schizophrenia were searched.³² Gene expression of ADK in postmortem schizophrenia microarray meta-analysis and RNAseq replication studies from the frontal cortex were also searched.³³

Fig. 4.

(A) In silico work flow of “lookup studies” of ADK expression in postmortem schizophrenia databases and bioinformatic analysis of ADK signatures generated in iLINCS. (B) “Lookup” studies of ADK gene expression in publically available postmortem schizophrenia datasets. The SMRI dataset reports ADK gene expression in subjects on and off antipsychotic medication. (C) Pathways associated with ADK KD and ADK OE gene signatures. ADK KD is associated with metabolism-related pathways. ADK OE is associated with immune-related pathways. (D) Heat map of ADK KD and ADK OE top 50 “lookup” study. The top 50 genes that compose the ADK KD and ADK OE gene signatures were searched and a heat map representing changes in gene expression in postmortem schizophrenia datasets was created in Kaleidoscope. ACC, Anterior cingulate cortex; ADK, adenosine kinase; CTL, control; DLPFC, dorsolateral prefrontal cortex; FC, fold change; KD, knockdown; OE, overexpression; SCZ, schizophrenia.

Using the integrative LINCS (iLINCS; http://ilincs.org) genomic data portal, we retrieved knockdown (ADK KD) and overexpression (ADK OE) signatures for ADK (figure 4A), as described previously.³⁴ The signatures were obtained from HEPG2 cell lines. ADK consensus gene knockdown (signature ID: LINCSKD_14594) and ADK overexpression (signature ID: LINCSOE_11634, perturbagen ID: ccsbBroad304_05779) signatures are comprised of the transcriptional changes of 978 landmark genes (L1000 genes) when ADK is perturbed. Using the Enrichr (http://amp.pharm.mssm.edu/Enrichr/) platform, the top 50 most significantly altered genes (P < .05) from each signature were interrogated to assess the biological pathways (Reactome 2016) underlying the transcriptional changes observed following ADK KD and ADK OE (figure 4C). The top 50 genes from the ADK KD and ADK OE signatures were also examined in the “lookup studies” workflow using an RShiny package developed in-house, Kaleidoscope (https://kalganem.shinyapps.io/BrainDatabases/). A heat map of expression changes in these genes in postmortem schizophrenia datasets was generated (figure 4D).

Data Analysis

Data sets were tested for normal distribution using D’Agostino and Pearson omnibus normality test and homogeneity of variance using F-test. Outliers greater than 2 SDs from the mean were excluded. Data were log transformed (human protein studies). Regression analyses were performed to detect association between protein/transcript expression and age, pH and PMI (protein studies) or RIN value (gene expression analysis). Analysis of covariance (ANCOVA) was applied if a significant association was found. Student’s t-test (parametric) or Mann-Whitney test (nonparametric) were applied if no significant associations were found. Post hoc power analysis of our data (1-β) was calculated for RNA data in the MBC cohort (mean 0.16, range 0.08–0.23) and protein data in the DLPFC in the MBC cohort (0.06) and HBCC cohort (0.97) and in the ACC in the NBTR cohort (0.84). Alpha = 0.05 for all statistical tests. Data were analyzed using Statistica 13.0 (Statsoft) and Graphpad Prism 6, (GraphPad Software, www.graphpad.com).

Results

We measured gene and protein expression of ADK in frontocortical regions in schizophrenia and control subjects. There was no significant difference in gene expression of ADK-L (figure 1A, t₍₂₈₎ = 1.2, P = .24), ADK-S (figure 1B, t₍₃₀₎ = 0.185, P = .86) or in the ratio of ADK-L to ADK-S mRNA expression in the DLPFC (MBC cohort) in schizophrenia subjects compared to controls (figure 1C, t₍₂₇₎ = 1.8, P = .078). Gene expression of GFAP, a marker of reactive astrocytes and astrogliosis, was significantly reduced in schizophrenia (figure 1D, t₍₂₈₎ = 2.3, P = .028). We did not detect any significant associations between expression of ADK-L, ADK-S or GFAP and age, PMI or RIN but there was a significant interaction effect (P < .05) for each regression analysis.

Fig. 1.

Relative mRNA expression of adenosine kinase (ADK) splice variants. There was no significant change in the expression of the (A) long (ADK-L) or the (B) short (ADK-S) variants of ADK in schizophrenia compared to controls. (C) There was also no significant difference in the ratio of ADK-S expression to ADK-L expression. (D) There was a significant decrease in GFAP mRNA expression in schizophrenia. Data presented as mean ± SEM. n = 14–16 per group control and schizophrenia.

There was no significant difference in ADK-S protein expression in schizophrenia subjects compared to controls in the DLPFC from the same tissue set (MBC cohort) that gene expression was measured in (figure 2A, t₍₃₃₎ = 0.25, P = .81). In addition, we measured ADK-S expression in DLPFC tissue obtained from a different brain bank (HBCC cohort) and found no significant change in expression (figure 2B, MWU = 440, P = .08). ADK-S protein expression was also examined in the ACC (NBTR cohort) and no significant change in expression was found after controlling for pH (figure 2C, F_(1,21) = 1.09, P = .31). The ADK-L isoform is not the primary brain isoform and it could not be reliably measured in all tissue sets. However, no change in ADK-L was detected in a subset of subjects in the MBC cohort (MWU = 125, P = 0.94, n = 15–17 per group; data not shown). Total GFAP protein levels (all multimer bands) were not significantly altered (figure 2D, MWU = 129, P = .61) in the DLPFC (MBC cohort). There were no significant associations between expression of ADK-S and age or PMI or GFAP and age, pH and PMI in any cohort. Representative immunoblots are shown in figure 2E and 2F and supplementary figure 1.

Fig. 2.

Relative ADK protein expression. (A) ADK protein expression was not significantly altered in schizophrenia in the DLPFC (2A-B; MBC; n = 17–18 per group and HBCC; n = 26–46 per group) or in the ACC (2C; NBTR; n = 12 per group). (D) GFAP protein expression was not significantly altered in schizophrenia in the DLPFC (MBC, n = 16 per group). (E) Representative ADK and (F) GFAP immunoblot images. Control and schizophrenia subjects were run in duplicate. ACC, anterior cingulate cortex; ADK, adenosine kinase; DLPFC, dorsolateral prefrontal cortex; HBCC, NIH Human Brain Collection Core; MBC, Maryland Brain Collection; NBTR, Bronx-Mt. Sinai NIH Brain and Tissue Repository.

In silico analysis of microarray and RNAseq databases of postmortem frontocortical expression of ADK in schizophrenia found that, in line with our results, ADK gene expression is largely unchanged in illness (figure 4A). The SMRI database reports the effects of antipsychotic medication on ADK gene expression in postmortem frontal cortex in schizophrenia and found ADK was significantly increased (1.08 FC, P < .001).

Thus, to determine the effects of antipsychotic medication on ADK splice variant expression, we assayed changes in gene expression of ADK-S and ADK-L in the frontal cortex of rats treated chronically with haloperidol-decanoate compared to vehicle-treated controls. There was no significant difference in expression of ADK-L (figure 3A, Student’s t-test P = 0.9) ADK-S (figure 3B, Student’s t-test P = .8) or the ratio of ADK-S/ADK-L (figure 3C, Student’s t-test P = .9) in antipsychotic-treated animals compared to vehicle-treated animals.

Fig. 3.

In rats treated chronically with haloperidol-decanoate, there was no significant difference in expression of (A) ADK-L or (B) ADK-S. There was no significant difference in (C) the ratio of ADK-S expression to ADK-L expression following chronic haloperidol administration. ADK protein expression was not significantly altered in schizophrenia subjects on antipsychotic medication compared to those off medication in the DLPFC (D) or ACC (E). Subjects whose medication status at time of death was unknown were excluded from analysis. Data presented as mean +/- SEM. N = 10–16 per group HBCC, n = 4–6 per group NBTR, n = 10 per group vehicle and haloperidol-decanoate. ACC anterior cingulate cortex, ADK adenosine kinase, DLPFC dorsolateral prefrontal cortex, HBCC Human Brain Collection Core, MBC Maryland Brain Collection.

In addition, we compared ADK-S protein expression in schizophrenia subjects who were “on” medication to those who were “off” medication at time of death in the DLPFC (HBCC cohort) and ACC (NBTR cohort). There were insufficient subjects “off” medication in the DLFPC MBC cohort to conduct meaningful statistical analysis. ADK-S expression was not significantly altered in the DLPFC (figure 3D, HBCC cohort; P = .336, n = 10–16/group) or in the ACC (figure 3E, NBTR cohort; P = 0.316, n = 4–6/group).

Gene signatures generated from knockdown and overexpression of ADK in the HEPG2 cell line were accessed and downloaded from iLINCS and interrogated in Enrichr using Reactome 2016 to identify the biological pathways associated with ADK perturbation. The ADK KD signature was associated primarily with metabolism-related changes including energy metabolism (figure 4C). The ADK OE signature was primarily associated with immune-related pathways (figure 4C). A heat map, generated in Kaleidoscope to visualize gene expression changes of the Top 50 genes in postmortem schizophrenia datasets, is shown in figure 4D. A subset of the Top 50 genes (19/50) that compose the ADK KD and ADK OE signatures were altered more than 1.15 fold in at least 2 postmortem schizophrenia datasets in the “lookup” study analysis, demonstrating that, while ADK expression is not significantly altered in postmortem schizophrenia, a subset of the genes associated with perturbation of ADK expression and the adenosine system are altered in disease.

Discussion

Adenosine plays a number of roles in key cellular processes. Adenosine modulates glutamate and dopamine, neurotransmitter systems that are strongly implicated in schizophrenia.⁵ Adenosine also acts as an energy modulator, as energy consumption and adenosine formation are directly linked,³⁵ and bioenergetic deficits are commonly reported in schizophrenia.¹⁸ Thus, perturbation of the adenosine system, as described by both preclinical and clinical studies, may contribute to the symptoms of schizophrenia.^8,9 The adenosine hypofunction hypothesis of schizophrenia proposes that overexpression of ADK reduces extracellular adenosine levels, leading to dysregulation of adenosine neuromodulatory targets and contributing to the pathophysiology of disease.⁸ In support of this hypothesis, a mouse model of global brain ADK overexpression results in reduced extracellular adenosine and produces schizophrenia-relevant behavioral phenotypes, including cognitive deficits that are improved by ADK inhibition.^20,21 We tested this hypothesis by assaying ADK expression in the DLPFC and ACC from subjects with schizophrenia. However, our data suggests that there is no significant difference in ADK gene or protein expression in these brain regions in this disorder. Dysregulation of extracellular adenosine generating pathways, previously reported in postmortem tissue in schizophrenia,³⁶ offer an alternative mechanism for abnormalities of the adenosinergic system reported in this disorder.

ADK is expressed as a long (ADK-L) isoform with primarily nuclear expression or a short (ADK-S) variant which is expressed cytoplasmically.³⁷ ADK-S has a truncated N-terminus with the first 21 amino acids of the canonical ADK-L isoform replaced with an alternate 4 amino acid sequence.³⁸ Exon 1A of ADK-S is located in the intron between exon 1 and 2 of the canonical ADK-L sequence.³⁹ In schizophrenia, ADK-L and ADK-S splice variant expression was not significantly altered in the DLPFC. There was also no significant difference in the ratio of ADK-S to ADK-L gene expression. We have previously measured pan-ADK gene expression at the region level and in enriched populations of astrocytes and pyramidal neurons in the DLPFC and found no significant difference in expression in schizophrenia,³⁶ suggesting that measuring expression of this gene at the region level in tissue composed of a heterogeneous mix of cell types does not hamper detection of differences in ADK gene expression. In silico analysis of postmortem frontal cortex microarray and RNAseq datasets, described in figure 4B, identified no significant changes in ADK gene expression in schizophrenia nor was ADK mRNA expression altered in peripheral blood in schizophrenia subjects compared to controls.⁴⁰

ADK protein expression was not significantly different in the DLPFC or ACC in schizophrenia compared to controls. ADK-S is the primary isoform expressed in the brain³⁹ and is the main isoform measured in this study. An important consideration in postmortem studies is the potential impact of medication on the expression of ADK. In silico analysis of publically available postmortem schizophrenia microarray data suggested that antipsychotic administration may increase ADK gene expression.^30,31 As only a single subject was “off” medication in the cohort of subjects we measured ADK gene expression in (MBC cohort), it was not possible to examine the effects of antipsychotics on ADK gene expression in this group. Thus, we examined ADK gene expression in rats treated for 9 months with haloperidol-decanoate, a typical antipsychotic. There was no significant difference in expression of either ADK variant or the ratio of ADK-S/ADK-L in these animals, suggesting that antipsychotic medication does not significantly alter ADK gene expression.

We expanded our study to determine whether antipsychotic administration effects ADK protein expression. ADK expression was not significantly different in the DLPFC or ACC in schizophrenia subjects who were “on” antipsychotic medication at the time of death compared to those who were “off” medication. Overall, ADK gene or protein expression is not significantly altered in the DLPFC or ACC in schizophrenia nor does antipsychotic administration appear to effect ADK expression.

Interestingly, while ADK overexpression is purported to result in schizophrenia-relevant symptoms, recent gene studies suggest an ADK loss of function mutation is associated with schizophrenia. A single schizophrenia patient with a deletion in the ADK region was identified in a copy number variant (CNV) study.⁴¹ In a follow-up case report, the loss of function variant of ADK resulted in low ADK mRNA and suggests that the CNV in the ADK region results in susceptibility to schizophrenia and other ADK-deficiency-related phenotypes in this patient.⁴⁰ Gene signatures generated following knockdown or overexpression of ADK in vitro are associated with primarily metabolic and immune-related pathways, respectively. Schizophrenia is associated with deficits in energy metabolism¹⁹ and abnormalities in neuroimmune pathways, with reports of elevated inflammatory markers in a subset of schizophrenia subjects.⁴² A subset of the Top 50 genes that compose the ADK KD and ADK OE signatures were also altered in postmortem schizophrenia “lookup” analysis. The effects of dysregulation of the adenosine system on energy metabolism and neuroimmune pathways in schizophrenia has yet to be elucidated.

Overall, our data suggest that the reduced extracellular adenosine levels that are posited to contribute to dysregulation of glutamate and dopamine transmission systems and the symptoms of schizophrenia,^7,8 are not due to overexpression of ADK. Other adenosine metabolism pathways are altered in schizophrenia and may lead to reduced availability of extracellular adenosine.³⁶ Ectonucleoside triphosphate diphosphohydrolase (ENTPD 1) and ENTPD 2, a primarily glial enzyme, convert ATP to ADP and AMP and have decreased gene expression in enriched populations of astrocytes in the DLPFC in schizophrenia. Altered ENTPD expression may result in reduced generation of AMP, the substrate for adenosine. Others have also reported reduced ENTPD enzyme activity in the striatum in schizophrenia.⁴³ Additionally, we found increased gene expression levels of adenosine deaminase, an enzyme that catabolises adenosine to inosine, and reduced levels of the adenosine transporter, equilibrative nucleoside transporter 1 (ENT1), in enriched populations of pyramidal neurons in schizophrenia. Taken together, these data suggest that generation of extracellular adenosine may be reduced in schizophrenia, in line with the adenosine hypofunction hypothesis of schizophrenia, but driven by cell-specific, non-ADK dependent mechanisms.

Limitations

It should be noted that ADK localization shifts dramatically during development from neuronal expression of the ADK-L isoform in early development to astrocytic expression of the ADK-S isoform during postnatal development.^24,39,44 Gestational knockout of ADK in mice results in significant social and contextual learning impairments that are not seen when astrocytic ADK is knocked out in early adulthood.⁴⁵ This is in line with human data where patients with mutations that result in ADK deficiency have severe developmental delays and neurological impairment.^46,47 As schizophrenia can be considered a neurodevelopmental disorder,⁴⁸ aberrant ADK activity during development may contribute to the development of this disorder but may not be detected in postmortem studies of adult tissue.

ADK may be rapidly degraded in postmortem tissue leading to overall depression of ADK levels. However, we found no significant association between PMI and ADK levels nor was the protein expression of an additional astrocyte marker associated with ADK, GFAP,⁴⁹ significantly altered in control and schizophrenia subjects in this study.

In summary, we examined ADK expression in 2 different brain regions, in tissue obtained from 3 brain bank collections, and found no significant change in gene or protein expression of ADK in schizophrenia. Thus, hypofunction of adenosine in schizophrenia does not appear to be driven by overexpression of ADK.

Funding

Lindsey Brinkmeyer Schizophrenia Research Fund (National Institute of Mental Health R01 MH094445).