Brain, Behavior, and Immunity xxx (2016) xxx–xxx
Contents lists available at ScienceDirect
Brain, Behavior, and Immunity journal homepage: www.elsevier.com/locate/ybrbi
Full-length Article
Reduced number of peripheral natural killer cells in schizophrenia but not in bipolar disorder _ Misiak a,b,⇑ Paweł Karpin´ski a, Dorota Frydecka b, Maria M. Sa˛siadek a, Błazej a b
Department of Genetics, Wroclaw Medical University, 1 Marcinkowski Street, 50-368 Wroclaw, Poland Department of Psychiatry, Wroclaw Medical University, 10 Pasteur Street, 50-367 Wroclaw, Poland
a r t i c l e
i n f o
Article history: Received 23 November 2015 Received in revised form 26 January 2016 Accepted 8 February 2016 Available online xxxx Keywords: Lymphocytes NK cells Computational deconvolution Inflammation Immunity Schizophrenia Bipolar disorder
a b s t r a c t Overwhelming evidence indicates that subthreshold inflammatory state might be implicated in the pathophysiology of schizophrenia (SCZ) and bipolar disorder (BPD). It has been reported that both groups of patients might be characterized by abnormal lymphocyte counts. However, little is known about alterations in lymphocyte proportions that may differentiate SCZ and BPD patients. Therefore, in this study we investigated blood cell proportions quantified by means of microarray expression deconvolution using publicly available data from SCZ and BPD patients. We found significantly lower counts of natural killer (NK) cells in drug-naïve and medicated SCZ patients compared to healthy controls across all datasets. In one dataset from SCZ patients, there were no significant differences in the number of NK cells between acutely relapsed and remitted SCZ patients. No significant difference in the number of NK cells between BPD patients and healthy controls was observed in all datasets. Our results indicate that SCZ patients, but not BPD patients, might be characterized by reduced counts of NK cells. Future studies looking at lymphocyte counts in SCZ should combine the analysis of data obtained using computational deconvolution and flow cytometry techniques. Ó 2016 Elsevier Inc. All rights reserved.
1. Introduction Although the immune-inflammatory concept of schizophrenia (SCZ) was proposed more than a century ago, it still serves as a rapidly evolving field of research (Khandaker et al., 2015). Epidemiological studies have revealed that prevalence of several autoimmune disorders, with exception of rheumatoid arthritis, is significantly higher in SCZ patients compared to the general population (Severance et al., 2014). Furthermore, it has been reported that prenatal infections with influenza virus, herpes simplex virus type 2, cytomegalovirus and Toxoplasma gondii increase the risk of SCZ in the offspring (Fineberg and Ellman, 2013). These findings support an epidemiological trend that SCZ patients tend to be born in winter/spring months (Davies et al., 2003). Similarly, bipolar disorder (BPD) is increasingly being recognized as a multisystem immune-inflammatory disease (Leboyer et al., 2012). To date, alterations in a number of peripheral immuneinflammatory markers have been reported in BPD and SCZ including elevated levels of pro-inflammatory cytokines (Miller et al., 2011a; Munkholm et al., 2013), overproduction of acute phase pro⇑ Corresponding author at: Department of Genetics, Wroclaw Medical University, 1 Marcinkowski Street, 50-368 Wroclaw, Poland. E-mail address:
[email protected] (B. Misiak).
teins (Maes et al., 1997; Yang et al., 2006) and increased levels of various autoantibodies (Margari et al., 2015; Pollak et al., 2015). It has long been argued that SCZ patients are characterized by a relative predominance of Th1 lymphocytes over Th2 lymphocytes (Schwarz et al., 2001). A recent meta-analysis has confirmed these findings (Guo et al., 2015). In turn, the meta-analysis performed by Miller et al. (2013) provided a broader insight into abnormal counts of blood lymphocytes observed in peripheral blood of SCZ patients. The authors revealed a significant increase in the percentage of Th lymphocytes (CD4+) and natural killer (NK) cells (CD56+) in acutely relapsed inpatients. Absolute levels of total lymphocytes, T lymphocytes (CD3+), Th lymphocytes (CD4+) and CD4+/CD8+ lymphocyte ratio were also significantly increased, and the percentage of CD3 lymphocytes was significantly decreased in drugnaïve FEP patients. The CD4/CD8 ratio appeared to be a state marker as it decreased following antipsychotic treatment in acutely relapsed patients. On the contrary, absolute levels of NK cells were found to increase during antipsychotic treatment. More recent studies have also investigated total counts of lymphocytes and monocytes in schizophrenia patients. In one study, lower total counts of peripheral blood lymphocytes have been reported (Semiz et al., 2014), while another study revealed no significant differences in the total count of lymphocytes and monocytes between patients with non-affective psychosis and healthy
http://dx.doi.org/10.1016/j.bbi.2016.02.005 0889-1591/Ó 2016 Elsevier Inc. All rights reserved.
Please cite this article in press as: Karpin´ski, P., et al. Reduced number of peripheral natural killer cells in schizophrenia but not in bipolar disorder. Brain Behav. Immun. (2016), http://dx.doi.org/10.1016/j.bbi.2016.02.005
2
´ ski et al. / Brain, Behavior, and Immunity xxx (2016) xxx–xxx P. Karpin
controls (Miller et al., 2015). Various alterations in proportions of blood leukocytes have been also reported in BPD, including increased proportions of monocytes, reduced proportions of T cells (CD3+) and cytotoxic T cells (CD3+CD8+), higher percentage of activated T cells (CD4+CD25+), the Th1/Th2 shift as well as lower percentage of Treg cells (Barbosa et al., 2015; Brambilla et al., 2014; do Prado et al., 2013; Drexhage et al., 2011). The majority of previous studies related to immune system alterations in SCZ have focused on single or few blood cell subtypes that likely limits a detection of meaningful changes occurring simultaneously in multiple cell subsets. Recent computational deconvolution approaches such as linear least-squares regression, digital sorting algorithm or linear support vector regression have provided the ability to estimate individual signal components from their mixtures (Newman et al., 2015; Shen-Orr and Gaujoux, 2013). Consequently, deconvolution methods enable to retrieve a wide range of information specific for distinct cell subsets (such as relative proportions and/or cell-specific expression profiles) directly from heterogeneous samples such as whole blood gene expression profile (Abbas et al., 2009; Newman et al., 2015; Zhong et al., 2013). Importantly, deconvolution algorithms allow to differentiate and count multiple blood cell subsets concurrently in one sample. Given that immune-inflammatory alterations may be potentially connected to shifts in blood cell subset composition, we aimed to investigate blood cell proportions quantified by means of microarray expression deconvolution using publicly available datasets from SCZ and BPD patients.
2. Material and methods 2.1. Human expression data and preprocessing All expression datasets were downloaded from the Gene Expression Omnibus (GEO) (Barrett et al., 2013). Detailed information on dataset accession number, platform, disease, mRNA source and application is provided in Supplementary Table 1. In general, our datasets included 199 SCZ patients, 49 BPD patients and 218 healthy controls. In addition, we used other 73 samples to validate the preprocessing and deconvolution steps. All Affymetrix arrays were obtained as CEL files and assessed for RNA quality using the ‘‘AffyRNADegradation” package (Fasold and Binder, 2013). All samples with dk < 0.45 were removed from further analysis (see (Fasold and Binder, 2012) for the details). Remaining samples were normalized with MAS5 using the ‘‘affy” package, mapped to the NCBI Entrez gene identifiers using a custom chip definition file (Brainarray, Version 19) and converted to HUGO gene symbols as suggested previously (Dai et al., 2005; Gautier et al., 2004; Newman et al., 2015). We used the ComBat algorithm implemented in the ‘‘swamp” package to correct the data for a batch effect (scan date) (Johnson et al., 2007; Lauss et al., 2013). The ComBat algorithm could be applied to GSE20300, GSE27383 and GSE46449 datasets, in which samples representing single batches were removed. For remaining Affymetrix datasets batch correction could not be used because study groups were not evenly distributed across batches or most of the batches included single samples. Due to a lack of publicly available raw data, Illumina datasets (GSE23848, GSE48072, GSE38481 and GSE38484) were downloaded as normalized matrices from GEO. Subsequently, the ‘‘WGCNA” package was used to collapse probes into a single gene measurement by selecting probes with highest mean expression across all samples (Miller et al., 2011b). Since GSE38481 and GSE38484 datasets included a unique subgroup of patients without exposition to antipsychotic treatment (drug-naïve patients), we combined antipsychotic-naive subjects and controls from both datasets using the ComBat algorithm to remove batch effects
(control samples from GSE38484 were matched for age, gender and size with controls from GSE38481). Medicated schizophrenia patients exclusively included in GSE38484 were treated as a separate dataset. Due to large technical differences, each dataset has been preprocessed and normalized separately. 2.2. Deconvolution of gene expression profiles After data normalization, we utilized the CIBERSORT algorithm (1000 iterations) and the LM22 gene signature to predict relative proportions of 21 human hematopoietic cell phenotypes (mast cells were excluded from signature) (Newman et al., 2015). Subsequently, we aggregated 21 cell subsets into 10 major cell types (B cells, plasma B cells, T CD4+, T CD8+, NK cells, gamma delta T cells, monocytes together with macrophages, dendritic cells, eosinophils and neutrophils) (Newman et al., 2015). Cell proportions were predicted in each dataset separately. 2.3. Statistics Estimated cell proportions were assessed for distribution (normal, non-normal) using the Anderson–Darling test. Homogeneity of variance was tested using the Levene’s test. Subsequently, cell proportions were compared between groups (affected individuals versus healthy controls) using the t-test (in case of normal distribution) or the Kruskal–Wallis test (in case of non-normal distribution). Bonferroni correction, taking into account the number of tested datasets, was applied to the level of significance in order to control for multiple testing. Analysis of covariance (ANCOVA) was used to test the effects of group (affected individuals vs. healthy controls) on estimated numbers of NK cells (dependent variable) after co-varying for age and sex Analysis of variance (ANOVA) was performed for one dataset (GSE48072) since all available data were categorical variables (group and sex). In case of three datasets (GSE38481, GSE38484 and GSE46449), estimated proportions of NK cells were transformed using arcsine transformation to obtain normal distribution and homogeneity of variance. All testes were two-tailed with 0.05 level of significance. STATISTICA 10 software was used to perform ANCOVA, while the rest of statistical analyses were performed using the ‘‘nortest” and ‘‘compareGroups” packages in R statistical software (Subirana et al., 2014). 3. Results 3.1. Deconvolution evaluation Although CIBERSORT has been extensively evaluated and outperformed other deconvolution methods as reported by Newman et al. (2015), we decided to conduct additional analyses to assess its performance in our study-specific setting (see Supplementary data for details). In brief, we applied CIBERSORT to GSE20300 dataset which includes 19 whole blood cell samples from a cohort of kidney transplant patients for which proportions of blood cells were provided (Shen-Orr et al., 2010). The scatter plot (Supplementary Fig. 1) shows that estimated proportions of lymphocytes (R2 = 0.77) and neutrophils (R2 = 0.74) were relatively wellcorrelated with corresponding raw data. These results were within the range of correlations reported by other authors using other deconvolution algorithms (Abbas et al., 2009; Gaujoux and Seoighe, 2013). Given that preprocessing steps may affect the relative cell proportion estimations, we assessed the reproducibility of estimated proportions in the dataset that includes 35 biological replicates (blood samples were split into two portions prior the RNA isolation step, GSE46449) (Clelland et al., 2013). After the
Please cite this article in press as: Karpin´ski, P., et al. Reduced number of peripheral natural killer cells in schizophrenia but not in bipolar disorder. Brain Behav. Immun. (2016), http://dx.doi.org/10.1016/j.bbi.2016.02.005
´ ski et al. / Brain, Behavior, and Immunity xxx (2016) xxx–xxx P. Karpin
3
Fig. 1. Comparison of proportions of lymphocyte types and neutrophils enumerated by the CIBERSORT algorithm in septic patients versus controls (GSE66099). There were significant differences in the number of all cell types between patients with sepsis and control subjects after Bonferroni correction (p 6 0.008).
MAS5 normalization alone or the MAS5 normalization followed by the ComBat algorithm, the estimated proportions of 21 cell subsets provided for 3 subjects (2 biological replicates each) were strongly correlated (adjusted R2 > 0.99) (Supplementary Fig. 2). Estimated proportions for other 32 biological replicates were also strongly correlated (adjusted R2 P 0.98; data not shown). Both preprocessing strategies did not alter correlations of cell proportions estimates between biological replicates. To provide a formal justification of the validity of selected deconvolution algorithm, we applied the CIBERSORT algorithm to whole blood expression data of sepsis patients, in which shifts in blood cell proportions are clinically well-defined. Recent studies have demonstrated that septic patients display an extensive depletion of peripheral gamma delta T, T CD4+, T CD8+, NK, and B lymphocytes, whereas numbers of circulating neutrophils are highly increased (Darcy et al., 2014; de Pablo et al., 2014). Therefore, we analyzed the dataset that includes 18 sepsis patients and 36 healthy controls (GSE66099). As demonstrated in Fig. 1, lymphocyte subtypes and neutrophils enumerated by CIBERSORT accurately reflected the main changes in circulating blood cells described in patients with sepsis, when compared to healthy controls. 3.2. Estimation of relative blood cell proportions in SCZ and BPD patients The CIBERSORT algorithm was applied to patients with SCZ, BPD and healthy controls to compute 10 blood cell types proportion estimates (separately for each dataset). Subsequently, the comparison of cell proportions from affected patients and proportion estimates computed in healthy controls was performed. There were no significant differences between affected and healthy individuals with respect to the majority of cell subtypes (data not shown). However, significantly lower levels of NK cells were observed in
medicated SCZ patients and, to a lower extent, in drug-naive SCZ patients in comparison with healthy controls (Table 1 and Fig. 2). Differences in the number of NK cells remained significant after Bonferroni correction (p 6 0.008) in datasets from medicated SCZ patients (GSE38484, GSE27383 and GSE48072). There were significantly more males in one dataset from medicated SCZ patients (GSE38484). In both datasets from BPD, healthy controls were significantly younger than patients. Table 2 shows results of the analysis adjusted for age and sex. There were significant and independent effects of group (affected individuals vs. healthy controls) and sex on estimates of NK cells in medicated SCZ patients (GSE38484 and GSE48072). In the dataset from drug-naïve SCZ patients (GSE38481/GSE38484), only the effect of sex was significant. Indeed, females had lower estimates of NK cells in all datasets from SCZ patients (GSE38481/GSE38484: 9.3 ± 2.1 vs. 11.9 ± 4.6, GSE38484: 7.4 ± 4.1 vs. 8.3 ± 4.3 and GSE48072: 3.6 ± 0.4 vs. 3.8 ± 0.4). Notably, there were no significant differences in the levels of NK cells between BPD patients and healthy controls across all datasets. Although significant changes in proportions of B cells, CD8, monocytes and macrophages were detected in one dataset that included medicated SCZ subjects (GSE38484), this observation was not confirmed in other datasets (GSE27383, GSE48072) from medicated SCZ patients (data not shown). Similarly, lower proportion of plasma B cells was observed only in one dataset out of two that included BPD patients (data not shown). There was no significant difference in estimates of NK cells between acutely relapsed SCZ patients and remitted individuals (data available from GSE27383, Supplementary Fig. 3). 4. Discussion To our knowledge, it is the first large analysis of multiple blood cell subsets in patients with SCZ and BPD conducted using
Please cite this article in press as: Karpin´ski, P., et al. Reduced number of peripheral natural killer cells in schizophrenia but not in bipolar disorder. Brain Behav. Immun. (2016), http://dx.doi.org/10.1016/j.bbi.2016.02.005
´ ski et al. / Brain, Behavior, and Immunity xxx (2016) xxx–xxx P. Karpin
4 Table 1 General characteristics and proportions of NK cells. Dataset
Variable
Patients
Healthy controls
p
GSE38481/GSE38484 (drug-naive SCZ patients)
Age, years Sex, M/F NK cells estimates
27.0 [24.0; 34.0] 21/8 8.8 [8.4; 11.6]
25.0 [23.0; 33.0] 33/12 11.5 [9.1; 15.2]
0.495 0.931 0.049
GSE38484 (medicated SCZ patients)
Age, years Sex, M/F NK cells estimates
40.9 ± 10.3 66/26 6.6 [3.8; 8.8]
39.3 ± 14.2 42/54 8.6 [6.1; 11.4]
0.395 <0.001* <0.001*
GSE27383 (medicated SCZ patients)
Age, years Sex, M/F NK cells estimates
N/P 43/0 12.5 ± 3.7
N/P 29/0 16.8 ± 6.2
N/A N/A 0.002*
GSE48072 (medicated SCZ patients)
Age, years Sex, M/F NK cells estimates
N/P 14/21 11.2 ± 3.5
N/P 17/14 14.2 ± 3.8
N/A 0.338 0.001*
GSE46449 (medicated BPD patients)
Age, years Sex, M/F NK cells estimates
38.0 [28.0; 49.0] 29/0 5.2 [4.0; 9.3]
29.0 [26.5; 36.0] 25/0 5.2 [3.4; 8.4]
0.006* N/A 0.548
GSE23848 (medicated BPD patients)
Age, years Sex, M/F NK cells estimates
41.5 [33.2; 45.5] 6/14 6.6 ± 2.4
26.0 [22.0; 31.5] 5/10 5.7 ± 4.1
0.014 0.833 0.497
Data expressed as mean ± SD for normal distribution, median [1st; 3rd quartile] for non-normal distribution or a number of cases for sex. N/A – not applicable. N/P – information not provided by authors. Significant differences (p 6 0.05) were marked in bold characters. * Significant differences after Bonferroni correction (p 6 0.008).
deconvolution approach. The method, we used in this study (CIBERSORT), is based on machine learning approach that allows to provide estimates of various cell subsets on the basis of gene signatures (reference expression profiles of marker genes for analyzed cell subsets) and the tissue expression profiles (Newman et al., 2015). In the present study, we found significantly lower number of NK cells in medicated SCZ patients using cell proportions computed from genome-wide expression data. These findings were also observed, to a lesser extent, in drug-naive SCZ patients and
were not significant after controlling for age and sex. Females had significantly lower numbers of NK cells in all datasets from SCZ patients. In contrast, depletion of NK cells was not detected in datasets from BPD patients. Numerous studies on lymphocyte distribution in SCZ have provided inconsistent results (Steiner et al., 2010). Some authors have reported increased numbers of NK cells in SCZ patients during acute relapse in comparison with healthy controls (Sasaki et al., 1994; Schattner et al., 1996), while other authors have observed decreased counts of NK cells
Fig. 2. Strip charts and box plots illustrating NK cells proportions predicted for affected individuals (SCZ and BPD patients) and healthy controls. WB – whole blood; PMBCs – peripheral blood mononuclear cells.
Please cite this article in press as: Karpin´ski, P., et al. Reduced number of peripheral natural killer cells in schizophrenia but not in bipolar disorder. Brain Behav. Immun. (2016), http://dx.doi.org/10.1016/j.bbi.2016.02.005
´ ski et al. / Brain, Behavior, and Immunity xxx (2016) xxx–xxx P. Karpin
5
Table 2 Effects of diagnostic group on NK cells estimates after co-varying for age and sex.
GSE38481 GSE38484 GSE48072 GSE46449 GSE23848
and GSE38484 (drug-naïve SCZ patients) (medicated SCZ patients) (medicated SCZ patients) (medicated BPD patients) (medicated BPD patients)
Group sex
Group (patients/controls)
Age
F
p
F
p
F
p
F
p
2.681 13.812 9.771 0.533 0.628
0.106 <0.001 0.003 0.467 0.434
0.631 0.020 – 0.016 0.269
0.429 0.886 – 0.899 0.607
4.521 6.041 4.370 – 0.264
0.037 0.015 0.041 – 0.611
0.147 0.284 2.750 – 0.229
0.702 0.595 0.102 – 0.635
Sex
Missing statistics and a lack of results from one dataset (GSE27383) are due to a lack of information on age and/or sex. Significant effects (p < 0.05) were marked in bold characters.
(Sperner-Unterweger et al., 1999) or no significant difference between patients and healthy controls (Steiner et al., 2010; Yovel et al., 2000). However, a recent meta-analysis of studies looking at lymphocytes counts in SCZ showed a significant increase in the percentage of NK cells in acutely relapsed inpatients (Miller et al., 2013). In addition, no significant differences in NK cells count between stable SCZ patients and healthy controls have been reported (Mazzarello et al., 2004; Rudolf et al., 2004; SpernerUnterweger et al., 1999). Similarly, studies on NK cells activity (NKA) in SCZ show little agreement in their results. While few studies have reported lower or higher NKA, the majority of studies have found no significant difference in NKA between SCZ patients and healthy controls (for review see (Yovel et al., 2000)). It has been shown that NKA is higher among SCZ patients in comparison with other subgroups of patients (BPD patients and individuals with personality disorders) (Yovel et al., 2000), while there was no difference in NK cells count between SCZ patients, BPD patients and healthy controls (Torres et al., 2009). Decrease in NK cells was observed in patients with mania (Abeer et al., 2006), while in stable bipolar outpatients there was no difference in the number of NK cells in comparison with healthy controls (Breunis et al., 2003). No association was observed between NKA and current mood state among euthymic, manic and depressed BPD patients (Breunis et al., 2003). In our study, the analysis of one dataset from SCZ patients revealed that lower numbers of NK cells appeared to be a trait marker as these alterations were present in acutely relapsed patients and remitted individuals. In the majority of longitudinal studies, no significant treatment effects on the number of NK cells have been reported (Bilici et al., 2003; Sasaki et al., 1994; Steiner et al., 2010). However, in one study, initially low number of NK cells was normalized in the course of treatment (SpernerUnterweger et al., 1999). Similarly, in one study, NKA on admission was significantly lower than after treatment (Sasaki et al., 1994). We believe that discrepancies in NK counts observed in previous studies may result from technical artefacts introduced during material processing for flow cytometry (FCM) or fluorescenceactivated cell sorting (FACS). First, both methods require a long time interval between blood draw and measurement, which may induce secondary alterations in gene expression. Moreover, there are numerous reports demonstrating that preparation of cells for FACS may have negative effects on RNA integrity and/or cells viability (Nilsson et al., 2014; Russell et al., 2013; Werthén et al., 2001). It has been widely reported that FCM measurements might be confounded by various antigens and compounds present in plasma (Bukowska-Straková et al., 2006; Liwski et al., 2014). Contrary to FCM and FACS, deconvolution measurements are based on whole blood samples that have been directly drawn to RNA stabilization solution, which ensures minimal gene expression changes related to sample processing (Duale et al., 2014). It should be noted; however, that precision and sensitivity of deconvolution approaches in enumerating various cell populations strongly
depends on algorithm itself and quality of input data (signatures and expression profiles) (Gaujoux and Seoighe, 2013). For example, estimates of NK cells proportions revealed in our study differ between datasets (see Table 1 and Fig. 2) that likely reflects various biological materials, sample size, chemistry, scanning systems and statistics used to derive input data. However, we have demonstrated by re-analyzing sepsis dataset that differences in cell proportions enumerated by the CIBERSORT algorithm accurately reflected relevant proportion shifts (Darcy et al., 2014; de Pablo et al., 2014). Importantly, it has been demonstrated that 90% of 10 blood cell subtypes could be enumerated with a significant correlation with flow cytometry measurements (including NK cells) (Newman et al., 2015). Therefore, deconvolution approaches have likely the potential to provide broader insights than those obtained from FCM and FACS (Yadav and De, 2015). Our findings might be relevant to the pathophysiology of SCZ and should be discussed in frame of a crosstalk between nervous and immune systems (Kelley and McCusker, 2014; Ziemssen and Kern, 2007). Indeed, the central and peripheral nervous systems are able to modulate immune responses by releasing several neurotransmitters, neuropeptides, hormones, and cytokines. Several lines of evidence indicate that dopamine is one of catecholamines playing a central role in neuromodulation of the immune system. Both primary and secondary lymphoid organs contain high levels of dopaminergic nerve terminals (Mignini et al., 2003). Moreover, many circulating immune cells, including NK cells (CD3-CD56+), constitutively express dopamine receptors (McKenna et al., 2002; Zhao et al., 2013). Deregulation in NK cell numbers and NKA in SCZ patients may be related to altered central dopaminergic activity. Chronic administration of dopamine agonists has been shown to decrease in abundance of NK cells (Harms et al., 2012; Jankowski et al., 2010). Specifically, it has been observed that stimulation of D2-like receptors may attenuate NK cells activity, while activation of D1-like receptors may enhance NK cell activity (Zhao et al., 2013). Similarly, it has been suggested that inhibition of effector functions of NK cells might appear due to selective upregulation of the D5 receptor (Mikulak et al., 2014). Conversely, parkinsonian patients, who suffer from reduced central dopaminergic levels, have been reported to exhibit increased NK activity associated with disease duration (Mihara et al., 2008) and symptom severity (Niwa et al., 2012). Taking into account these lines of evidence, decreased counts of NK cells might be perceived as a surrogate measure of central dopaminergic activity. Our study has several limitations that should be addressed. First, the number of drug-naive SCZ subjects was relatively small and thus results obtained from this subgroup of patients could not be replicated in a separate dataset. Secondly, we were not able to obtain raw data for the majority of SCZ and BPD patients. Therefore, controlling for important factors such as data normalization and removal of batch effects was limited. However, the fact that a significant depletion of NK cells was consistently observed across datasets obtained by different platforms and pre-processed with
Please cite this article in press as: Karpin´ski, P., et al. Reduced number of peripheral natural killer cells in schizophrenia but not in bipolar disorder. Brain Behav. Immun. (2016), http://dx.doi.org/10.1016/j.bbi.2016.02.005
´ ski et al. / Brain, Behavior, and Immunity xxx (2016) xxx–xxx P. Karpin
6
different techniques should be considered as an added value of our study. It should be stressed that we had no access to detailed information on sociodemographic characteristics, clinical manifestation and pharmacological treatment regimens. Therefore, we were not able to comprehensively address effects of these variables on the number of NK cells in SCZ and BPD patients. Indeed, it is important to note that a difference in the number of NK cells between drugnaïve SCZ patients and healthy controls was not significant after correction for multiple testing and after controlling for age and sex. These findings would suggest that decreased levels of NK cells in SCZ appear due to medication effects. However, there were also no significant differences in the number of NK cells between acutely relapsed SCZ patients and healthy controls. In addition, overlapping treatment regimens are often used in both SCZ and BPD patients. Finally, a lack of information on the use of various medications further limits unequivocal conclusions about medication effects. It should also be noted that a number of differences in age and sex between patients and healthy controls were detected across datasets. Our adjusted analysis revealed that female sex, but not age, might be related to lower number of NK cells. Therefore, it is most likely that differences in age, observed in datasets from BPD patients, did not confound estimates of NK cells. Similarly, the difference in sex distribution towards a predominance of males in one dataset from SCZ patients most likely did not bias our results since female sex appeared to be associated with lower numbers of NK cells. Finally, it should be noted that although deconvolution methods enable to observe multiple cell subsets simultaneously, their ability to discriminate closely related cell populations is limited (Shen-Orr and Gaujoux, 2013). Thus, we were unable to precisely establish proportions of various subsets of NK cells (Poli et al., 2009). In conclusion, our preliminary results suggest that reduced number of NK cells might appear in SCZ patients but not in BPD patients. More studies are required to confirm these findings in drug-naïve SCZ patients. Future studies should also combine the analysis of expression data from PBMCs with flow cytometry measurements in order to disentangle discordant results reported in previous studies. It might be also recommended to include the measures of central dopaminergic activity in studies looking at alterations in the number of NK cells in SCZ patients. Conflict of interest None to declare. Acknowledgments _ Misiak is supThis research received no specific funding. Błazej ported by the START 2015 scholarship provided by the Foundation for Polish Science. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.bbi.2016.02.005. References Abbas, A.R., Wolslegel, K., Seshasayee, D., Modrusan, Z., Clark, H.F., 2009. Deconvolution of blood microarray data identifies cellular activation patterns in systemic lupus erythematosus. PLoS ONE 4, e6098. Abeer, El-Sayed, A., Ramy, H.A., 2006. Immunological changes in patients with mania: changes in cell mediated immunity in a sample from Egyptian patients. Egypt. J. Immunol. 13, 79–85. Barbosa, I.G., Rocha, N.P., Assis, F., Vieira, E.L., Soares, J.C., Bauer, M.E., Teixeira, A.L., 2015. Monocyte and lymphocyte activation in bipolar disorder: a new piece in the puzzle of immune dysfunction in mood disorders. Int. J. Neuropsychopharmacol., 18
Barrett, T., Wilhite, S.E., Ledoux, P., Evangelista, C., Kim, I.F., Tomashevsky, M., Marshall, K.A., Phillippy, K.H., Sherman, P.M., Holko, M., Yefanov, A., Lee, H., Zhang, N., Robertson, C.L., Serova, N., Davis, S., Soboleva, A., 2013. NCBI GEO: archive for functional genomics data sets – update. Nucleic Acids Res. 41, D991–D995. Bilici, M., Tekelioglu, Y., Efendioglu, S., Ovali, E., Ulgen, M., 2003. The influence of olanzapine on immune cells in patients with schizophrenia. Prog. Neuropsychopharmacol. Biol. Psychiatry 27, 483–485. Brambilla, P., Bellani, M., Isola, M., Bergami, A., Marinelli, V., Dusi, N., Rambaldelli, G., Tansella, M., Finardi, A.M., Martino, G., Perlini, C., Furlan, R., 2014. Increased M1/decreased M2 signature and signs of Th1/Th2 shift in chronic patients with bipolar disorder, but not in those with schizophrenia. Transl. Psychiatry 4, e406. Breunis, M.N., Kupka, R.W., Nolen, W.A., Suppes, T., Denicoff, K.D., Leverich, G.S., Post, R.M., Drexhage, H.A., 2003. High numbers of circulating activated T cells and raised levels of serum IL-2 receptor in bipolar disorder. Biol. Psychiatry 53, 157–165. Bukowska-Straková, K., Baran, J., Gawlicka, M., Kowalczyk, D., 2006. A false expression of CD8 antigens on CD4+ T cells in a routine flow cytometry analysis. Folia Histochem. Cytobiol. 44, 179–183. Clelland, C.L., Read, L.L., Panek, L.J., Nadrich, R.H., Bancroft, C., Clelland, J.D., 2013. Utilization of never-medicated bipolar disorder patients towards development and validation of a peripheral biomarker profile. PLoS ONE 8, e69082. Dai, M., Wang, P., Boyd, A.D., Kostov, G., Athey, B., Jones, E.G., Bunney, W.E., Myers, R.M., Speed, T.P., Akil, H., Watson, S.J., Meng, F., 2005. Evolving gene/transcript definitions significantly alter the interpretation of GeneChip data. Nucleic Acids Res. 33, e175. Darcy, C.J., Woodberry, T., Davis, J.S., Piera, K.A., McNeil, Y.R., Chen, Y., Yeo, T.W., Weinberg, J.B., Anstey, N.M., 2014. Increased plasma arginase activity in human sepsis: association with increased circulating neutrophils. Clin. Chem. Lab. Med. 52, 573–581. Davies, G., Welham, J., Chant, D., Torrey, E.F., McGrath, J., 2003. A systematic review and meta-analysis of Northern Hemisphere season of birth studies in schizophrenia. Schizophr. Bull. 29, 587–593. de Pablo, R., Monserrat, J., Prieto, A., Alvarez-Mon, M., 2014. Role of circulating lymphocytes in patients with sepsis. Biomed. Res. Int. 2014, 671087. do Prado, C.H., Rizzo, L.B., Wieck, A., Lopes, R.P., Teixeira, A.L., Grassi-Oliveira, R., Bauer, M.E., 2013. Reduced regulatory T cells are associated with higher levels of Th1/TH17 cytokines and activated MAPK in type 1 bipolar disorder. Psychoneuroendocrinology 38, 667–676. Drexhage, R.C., Hoogenboezem, T.H., Versnel, M.A., Berghout, A., Nolen, W.A., Drexhage, H.A., 2011. The activation of monocyte and T cell networks in patients with bipolar disorder. Brain Behav. Immun. 25, 1206–1213. Duale, N., Lipkin, W.I., Briese, T., Aarem, J., Rønningen, K.S., Aas, K.K., Magnus, P., Harbak, K., Susser, E., Brunborg, G., 2014. Long-term storage of blood RNA collected in RNA stabilizing Tempus tubes in a large biobank – evaluation of RNA quality and stability. BMC Res. Notes 7, 633. Fasold, M., Binder, H., 2012. Estimating RNA-quality using GeneChip microarrays. BMC Genomics 13, 186. Fasold, M., Binder, H., 2013. AffyRNADegradation: control and correction of RNA quality effects in GeneChip expression data. Bioinformatics 29, 129–131. Fineberg, A.M., Ellman, L.M., 2013. Inflammatory cytokines and neurological and neurocognitive alterations in the course of schizophrenia. Biol. Psychiatry 73, 951–966. Gaujoux, R., Seoighe, C., 2013. Cell Mix: a comprehensive toolbox for gene expression deconvolution. Bioinformatics 29, 2211–2212. Gautier, L., Cope, L., Bolstad, B.M., Irizarry, R.A., 2004. Affy – analysis of Affymetrix GeneChip data at the probe level. Bioinformatics 20, 307–315. Guo, J., Liu, C., Wang, Y., Feng, B., Zhang, X., 2015. Role of T helper lymphokines in the immune-inflammatory pathophysiology of schizophrenia: systematic review and meta-analysis. Nord. J. Psychiatry 69, 364–372. Harms, R., Morsey, B., Boyer, C.W., Fox, H.S., Sarvetnick, N., 2012. Methamphetamine administration targets multiple immune subsets and induces phenotypic alterations suggestive of immunosuppression. PLoS ONE 7, e49897. Jankowski, M.M., Ignatowska-Jankowska, B., Glac, W., Swiergiel, A.H., 2010. Cocaine administration increases CD4/CD8 lymphocyte ratio in peripheral blood despite lymphopenia and elevated corticosterone. Int. Immunopharmacol. 10, 1229– 1234. Johnson, W.E., Li, C., Rabinovic, A., 2007. Adjusting batch effects in microarray expression data using empirical Bayes methods. Biostatistics 8, 118–127. Kelley, K.W., McCusker, R.H., 2014. Getting nervous about immunity. Semin. Immunol. 26, 389–393. Khandaker, G.M., Cousins, L., Deakin, J., Lennox, B.R., Yolken, R., Jones, P.B., 2015. Inflammation and immunity in schizophrenia: implications for pathophysiology and treatment. Lancet Psychiatry 2, 258–270. Lauss, M., Visne, I., Kriegner, A., Ringnér, M., Jönsson, G., Höglund, M., 2013. Monitoring of technical variation in quantitative high-throughput datasets. Cancer Inform. 12, 193–201. Leboyer, M., Soreca, I., Scott, J., Frye, M., Henry, C., Tamouza, R., Kupfer, D.J., 2012. Can bipolar disorder be viewed as a multi-system inflammatory disease? J. Affect. Disord. 141, 1–10. Liwski, C.J., Padley, D.J., Gustafson, M.P., Winters, J.L., Gastineau, D.A., Jacob, E.K., 2014. Discordant CD34+ cell results in peripheral blood and hematopoietic progenitor cell-apheresis product: implications for clinical decisions and impact on patient treatment. Transfusion 54, 541–544. Maes, M., Delange, J., Ranjan, R., Meltzer, H.Y., Desnyder, R., Cooremans, W., Scharpe, S., 1997. Acute phase proteins in schizophrenia, mania and major depression: modulation by psychotropic drugs. Psychiatry Res. 66, 1–11.
Please cite this article in press as: Karpin´ski, P., et al. Reduced number of peripheral natural killer cells in schizophrenia but not in bipolar disorder. Brain Behav. Immun. (2016), http://dx.doi.org/10.1016/j.bbi.2016.02.005
´ ski et al. / Brain, Behavior, and Immunity xxx (2016) xxx–xxx P. Karpin Margari, F., Petruzzelli, M.G., Mianulli, R., Campa, M.G., Pastore, A., Tampoia, M., 2015. Circulating anti-brain autoantibodies in schizophrenia and mood disorders. Psychiatry Res. 230, 704–708. Mazzarello, V., Cecchini, A., Fenu, G., Rassu, M., Dessy, L.A., Lorettu, L., Montella, A., 2004. Lymphocytes in schizophrenic patients under therapy: serological, morphological and cell subset findings. Ital. J. Anat. Embryol. 109, 177–188. McKenna, F., McLaughlin, P.J., Lewis, B.J., Sibbring, G.C., Cummerson, J.A., BowenJones, D., Moots, R.J., 2002. Dopamine receptor expression on human T- and Blymphocytes, monocytes, neutrophils, eosinophils and NK cells: a flow cytometric study. J. Neuroimmunol. 132, 34–40. Mignini, F., Streccioni, V., Amenta, F., 2003. Autonomic innervation of immune organs and neuroimmune modulation. Auton. Autacoid Pharmacol. 23, 1–25. Mihara, T., Nakashima, M., Kuroiwa, A., Akitake, Y., Ono, K., Hosokawa, M., Yamada, T., Takahashi, M., 2008. Natural killer cells of Parkinson’s disease patients are set up for activation: a possible role for innate immunity in the pathogenesis of this disease. Parkinsonism Relat. Disord. 14, 46–51. Mikulak, J., Bozzo, L., Roberto, A., Pontarini, E., Tentorio, P., Hudspeth, K., Lugli, E., Mavilio, D., 2014. Dopamine inhibits the effector functions of activated NK cells via the upregulation of the D5 receptor. J. Immunol. 193, 2792–2800. Miller, B.J., Buckley, P., Seabolt, W., Mellor, A., Kirkpatrick, B., 2011a. Meta-analysis of cytokine alterations in schizophrenia: clinical status and antipsychotic effects. Biol. Psychiatry 70, 663–671. Miller, B.J., Gassama, B., Sebastian, D., Buckley, P., Mellor, A., 2013. Meta-analysis of lymphocytes in schizophrenia: clinical status and antipsychotic effects. Biol. Psychiatry 73, 993–999. Miller, B.J., Kandhal, P., Rapaport, M.H., Mellor, A., Buckley, P., 2015. Total and differential white blood cell counts, high-sensitivity C-reactive protein, and cardiovascular risk in non-affective psychoses. Brain Behav. Immun. 45, 28–35. Miller, J.A., Cai, C., Langfelder, P., Geschwind, D.H., Kurian, S.M., Salomon, D.R., Horvath, S., 2011b. Strategies for aggregating gene expression data: the collapseRows R function. BMC Bioinformatics 12, 322. Munkholm, K., Brauner, J.V., Kessing, L.V., Vinberg, M., 2013. Cytokines in bipolar disorder vs. healthy control subjects: a systematic review and meta-analysis. J. Psychiatr. Res. 47, 1119–1133. Newman, A.M., Liu, C.L., Green, M.R., Gentles, A.J., Feng, W., Xu, Y., Hoang, C.D., Diehn, M., Alizadeh, A.A., 2015. Robust enumeration of cell subsets from tissue expression profiles. Nat. Methods 12, 453–457. Nilsson, H., Krawczyk, K.M., Johansson, M.E., 2014. High salt buffer improves integrity of RNA after fluorescence-activated cell sorting of intracellular labeled cells. J. Biotechnol. 192 (Pt A), 62–65. Niwa, F., Kuriyama, N., Nakagawa, M., Imanishi, J., 2012. Effects of peripheral lymphocyte subpopulations and the clinical correlation with Parkinson’s disease. Geriatr. Gerontol. Int. 12, 102–107. Poli, A., Michel, T., Thérésine, M., Andrès, E., Hentges, F., Zimmer, J., 2009. CD56bright natural killer (NK) cells: an important NK cell subset. Immunology 126, 458–465. Pollak, T.A., Beck, K., Irani, S.R., Howes, O.D., David, A.S., McGuire, P.K., 2015. Autoantibodies to central nervous system neuronal surface antigens: psychiatric symptoms and psychopharmacological implications. Psychopharmacology. Rudolf, S., Schlenke, P., Broocks, A., Peters, M., Rothermundt, M., Arolt, V., Kirchner, H., 2004. Search for atypical lymphocytes in schizophrenia. World J. Biol. Psychiatry 5, 33–37. Russell, J.N., Clements, J.E., Gama, L., 2013. Quantitation of gene expression in formaldehyde-fixed and fluorescence-activated sorted cells. PLoS ONE 8, e73849.
7
Sasaki, T., Nanko, S., Fukuda, R., Kawate, T., Kunugi, H., Kazamatsuri, H., 1994. Changes of immunological functions after acute exacerbation in schizophrenia. Biol. Psychiatry 35, 173–178. Schattner, A., Cori, Y., Hahn, T., Sirota, P., 1996. No evidence for autoimmunity in schizophrenia. J. Autoimmun. 9, 661–666. Schwarz, M.J., Chiang, S., Muller, N., Ackenheil, M., 2001. T-helper-1 and T-helper-2 responses in psychiatric disorders. Brain Behav. Immun. 15, 340–370. Semiz, M., Yildirim, O., Canan, F., Demir, S., Hasbek, E., Tuman, T.C., Kayka, N., Tosun, M., 2014. Elevated neutrophil/lymphocyte ratio in patients with schizophrenia. Psychiatr. Danub. 26, 220–225. Severance, E.G., Yolken, R.H., Eaton, W.W., 2014. Autoimmune diseases, gastrointestinal disorders and the microbiome in schizophrenia: more than a gut feeling. Schizophr. Res. Shen-Orr, S.S., Gaujoux, R., 2013. Computational deconvolution: extracting cell type-specific information from heterogeneous samples. Curr. Opin. Immunol. 25, 571–578. Shen-Orr, S.S., Tibshirani, R., Khatri, P., Bodian, D.L., Staedtler, F., Perry, N.M., Hastie, T., Sarwal, M.M., Davis, M.M., Butte, A.J., 2010. Cell type-specific gene expression differences in complex tissues. Nat. Methods 7, 287–289. Sperner-Unterweger, B., Whitworth, A., Kemmler, G., Hilbe, W., Thaler, J., Weiss, G., Fleischhacker, W.W., 1999. T-cell subsets in schizophrenia: a comparison between drug-naive first episode patients and chronic schizophrenic patients. Schizophr. Res. 38, 61–70. Steiner, J., Jacobs, R., Panteli, B., Brauner, M., Schiltz, K., Bahn, S., Herberth, M., Westphal, S., Gos, T., Walter, M., Bernstein, H.G., Myint, A.M., Bogerts, B., 2010. Acute schizophrenia is accompanied by reduced T cell and increased B cell immunity. Eur. Arch. Psychiatry Clin. Neurosci. 260, 509–518. Subirana, I., Sanz, H., Vila, J., 2014. Building bivariate tables: the compareGroups package for R. J. Stat. Softw. Torres, K.C., Souza, B.R., Miranda, D.M., Nicolato, R., Neves, F.S., Barros, A.G., Dutra, W.O., Gollob, K.J., Correa, H., Romano-Silva, M.A., 2009. The leukocytes expressing DARPP-32 are reduced in patients with schizophrenia and bipolar disorder. Prog. Neuropsychopharmacol. Biol. Psychiatry 33, 214–219. Werthén, M., Sellborn, A., Källtorp, M., Elwing, H., Thomsen, P., 2001. In vitro study of monocyte viability during the initial adhesion to albumin- and fibrinogencoated surfaces. Biomaterials 22, 827–832. Yadav, V.K., De, S., 2015. An assessment of computational methods for estimating purity and clonality using genomic data derived from heterogeneous tumor tissue samples. Brief Bioinform. 16, 232–241. Yang, Y., Wan, C., Li, H., Zhu, H., La, Y., Xi, Z., Chen, Y., Jiang, L., Feng, G., He, L., 2006. Altered levels of acute phase proteins in the plasma of patients with schizophrenia. Anal. Chem. 78, 3571–3576. Yovel, G., Sirota, P., Mazeh, D., Shakhar, G., Rosenne, E., Ben-Eliyahu, S., 2000. Higher natural killer cell activity in schizophrenic patients: the impact of serum factors, medication, and smoking. Brain Behav. Immun. 14, 153–169. Zhao, W., Huang, Y., Liu, Z., Cao, B.B., Peng, Y.P., Qiu, Y.H., 2013. Dopamine receptors modulate cytotoxicity of natural killer cells via cAMP-PKA-CREB signaling pathway. PLoS ONE 8, e65860. Zhong, Y., Wan, Y.W., Pang, K., Chow, L.M., Liu, Z., 2013. Digital sorting of complex tissues for cell type-specific gene expression profiles. BMC Bioinformatics 14, 89. Ziemssen, T., Kern, S., 2007. Psychoneuroimmunology-cross-talk between the immune and nervous systems. J. Neurol. 254 (Suppl. 2), II8–II11.
Please cite this article in press as: Karpin´ski, P., et al. Reduced number of peripheral natural killer cells in schizophrenia but not in bipolar disorder. Brain Behav. Immun. (2016), http://dx.doi.org/10.1016/j.bbi.2016.02.005