Association of Interleukin 2 (IL-2), Interleukin 6 (IL-6), and TNF-alpha (TNFα) Gene Polymorphisms With Paranoid Schizophrenia in a Polish Population

Schizophrenia is a severe mental illness, with heritability estimated at up to 80%. Several hypotheses (genetics, neurodevelopmental, neurotransmitter, neuroimmunological) have been presented to explain the etiopathogenesis of schizophrenia.¹

Multiple linkage and association studies proved that many genes participate in the development of schizophrenia. Metaanalysis conducted by Allen et al.² confirmed the association of 24 single-nucleotide polymorphisms (SNPs) in 16 genes (COMT, DRD1, IL1B) with schizophrenia. Genes encoding cytokines seem to be good candidate genes for schizophrenia. Cytokines have important functions in the central nervous system (CNS). They function as essential mediators of cross-talk between the brain and immune system and play a key role in neuroinflammatory processes. Many of the cytokines are normally produced in the healthy brain, where they play critical roles in such aspects of neurodevelopment as neurogenesis, migration, differentiation, and synapse-formation.^3,4 They could also interact with neurons, influencing neurotransmission. For example, by binding to specific receptors on the neuron’s surface, cytokines may modulate the secretory activity of these cells in relation to catecholamines or neuropeptides.⁵ In previously published reports, concentration of various cytokines were found to be increased or decreased in patients with schizophrenia.⁶ What is more, concentration of cytokines in blood serum of schizophrenic patients may vary depending on whether the patient is in active or resting phase of the disease.⁷ Some authors suggest that schizophrenia may be associated with alterations in the Th1 (Il-2, IFN-γ)/Th2 (IL-6, IL-10) cytokine ratios (possibly induced by a viral infection), with a shift toward the Th2 system.^1,8,9

IL-2, IL-6, and TNFα mediate immune and inflammatory responses and are activating cytokines that seem to play a key role in the CNS. They are actively transported into the CNS, but are also released from activated glia cells.⁸

IL-2 is a growth factor for T-cells, NK cells, and B cells.¹⁰ In patients with schizophrenia, decreased IL-2 production and increased level of the soluble IL-2 receptor (sIL-2R) have been observed.¹¹ However, Cazzullo et al.¹² have found significantly higher production of IL-2 and INF-γ in schizophrenic patients than in controls.

IL-6 has a close functional relationship with IL-2 and TNF-α.¹³ This cytokine is produced by a variety of cell types (e.g., macrophages, monocytes, fibroblasts) and has a pleiotropic activity in various tissues. IL-6 can exhibit both pro- and anti-inflammatory properties, as demonstrated in IL-6 gene knock-out mice.¹⁴ Several studies have found the connection between increased serum concentration of IL-6 and immune abnormalities in schizophrenia.¹⁵ IL-6 is widely distributed in the brain, especially in the hippocampus and hypothalamus, and it is strongly connected with the production of neurotransmitters. IL-6 can stimulate neurons in-vitro to secrete dopamine and probably also other catecholamines.¹⁶ The dopamine hypothesis of altered dopaminergic neurotransmission is one of the most popular biochemical explanation for schizophrenia development. It is postulated that an imbalance in the redox-state of the brain may be part of the underlying pathophysiology of schizophrenia by alterations in fast-spiking interneurons. Inflammatory mediators, such as IL-6, can tip the redox balance into a pro-oxidant state.¹⁷ Moreover, high concentration of IL-6 is postulated to be associated with the duration of the disease and resistance to treatment.¹⁸

TNFα is a proinflammatory cytokine produced mainly by immune cells (i.e., macrophages and active T and B lymphocytes), but also, in the CNS, by neurons, astroglia, and microglia. This cytokine plays a central role in the immune system and, in the CNS, regulates growth and differentiation of nerve cells, modulates serotonergic neurotransmission, and regulates synaptic scaling and apoptosis.¹⁹ Gene-encoding TNFα is located within the HLA region on chromosome 6p21.3, which has been reported to be associated with susceptibility to schizophrenia.²⁰ It was shown that concentration of TNFα in whole blood from schizophrenic patients is significantly higher than in the blood of healthy individuals.²¹ However, Baker et al.²² did not reveal any differences in TNFα, or in IL-1β, IL-6, and sIL-2R serum concentration between schizophrenic patients and controls.

Association studies investigating promoter polymorphisms of IL-2, IL-6, and TNFα genes in paranoid schizophrenia are strongly limited. Taking into account the potential role of those cytokines in the ethiopathogenesis of paranoid schizophrenia and the results of the linkage studies pointing to the chromosomal regions containing the IL-2, IL-6, and TNFα genes, the present study focuses on examining the association between the IL-6 (−174G/C; rs1800795), IL-2 (−330G/T; rs2069762), and TNFα (−308G/A; rs1800629) gene promoter polymorphisms and paranoid schizophrenia in a Polish population. We hypothesized an association of the IL-2 −330 T-allele, IL-6 −174 C allele, and TNFα −308 A-allele with development of paranoid schizophrenia.

Methods

Patients and Controls

In all, 115 unrelated patients with paranoid schizophrenia (48 women [42%]; mean age 43.3 [SD: 12.6] years) were enrolled into the study. The patients were recruited from inpatients being treated at the Department and Clinic of Psychiatry, Medical University of Silesia in Katowice and the Neuropsychiatric Hospital in Lubliniec. All patients fulfilled the DSM-IV-TR (Diagnostic and Statistical Manual of Mental Disorders, 4th Edition, Text Revision) criteria for paranoid-type schizophrenia. The diagnoses were assigned based on a Structured Clinical Interview for DSM-IV Axis I Disorders–Clinical Version²³ by two independent experienced psychiatrists. Other available resources, such as clinical course, medical records, and family history information were also used. Exclusion criteria for patients were other Axis I and Axis II diagnoses, neurological illness, endocrine disorders, and autoimmune diseases. The age at onset of schizophrenia was defined as the age at first appearance of psychotic symptoms. Psychopathology at the time of admittance was monitored by the structured interview applying Positive and Negative Syndrome Scale (PANSS).²⁴ Patients were assessed to be capable of understanding the study before written consent was obtained.

The control group comprised 135 healthy unrelated individuals (43 women [32%]; mean age 41.3 [SD: 9.0] years) recruited from among volunteer blood donors of the Regional Centre of Blood Donation and Treatment in Katowice. Exclusion criteria for controls were current psychiatric problems; any other neurological disorders; family history of schizophrenia (verified by direct interview); and chronic and acute physical illness, such as infection, autoimmune, or allergic diseases.

The study population was homogeneous and composed of the Polish Caucasian residents of the Silesia region.

The study was approved by the Bioethics Committee of Medical University of Silesia.

IL-2, IL-6, and TNFα Genotyping

DNA was isolated from whole peripheral blood by the phenol-chloroform method. The quality of DNA extracts was checked by spectrophotometric measurements (Gene Quant II by LKB Pharmacia Biotech) and by electrophoresis in 1.5% agarose gel.

Polymorphism of the IL-6 gene (at position −174 G/C) was determined by the PCR-RFLP method, as described previously.²⁵

Polymorphism of the TNFα gene (−308G/A) was determined by the PCR-RFLP method. PCR reaction was carried out on a DNA template with a pair of specific primers (5′ AGGCAATAGGTTTTGAGGGCCAT 3′− forward, 5′ TCCTCCCTGCTCCGATTCCG 3′− reverse; amplicon length 107 bp), reaction mix (provided by the manufacturer), and Taq polymerase (Epicentre Biotechnologies, Madison, WI, USA) in a total volume of 25 μl. PCR cycling conditions were: 95°C for 3 minutes, followed by 35 cycles of 94°C for 30 sec, 55°C for 45 sec, and 72°C for 45 sec, and the final elongation at 72°C for 5 minutes. PCR products were digested with NcoI restriction endonuclease (MBI Fermentas). Restriction fragments were 87 and 20 bp (allele G) and uncut 107 bp fragment (allele A). For both IL-6 and TNFα genotyping, PCR products were analyzed by electrophoresis in 2% agarose gel stained with ethidium bromide. After digestion, PCR products were separated and analyzed in 3% agarose gel stained with ethidium bromide.

Polymorphism of the IL-2 gene (−330T/G) was determined by allele specific PCR (AS-PCR) method. For each allele, PCR reaction was carried out on a DNA template with a pair of specific primers (forward: 5′ CTGACATGTAAGAAGCAATCTAT 3′ and reverse: 5′ CTCAGAAAATTTTCTTTGTCC 3′ for allele G, amplimer length: 215 bp; forward: 5′ TTCACATGTTCAGTGTAGTTTTAT 3′ and reverse 5′ TGTTACATTAGCCCACACTTA 3′ for allele T, amplimer length: 152 bp), reaction mix (provided by the manufacturer), and Taq polymerase (Epicentre Biotechnologies, Madison, WI, USA) in a total volume of 25 μl. PCR cycling conditions were the following: 95°C for 5 minutes, followed by 30 cycles of 95°C for 60 sec, 59°C for 60 sec and 72°C for 60 sec, and final elongation at 72°C for 5 minutes. The genotypes were determined using 2% agarose gel stained with ethidium bromide.

Statistical Analysis

Descriptive variables are presented as mean (standard deviation [SD]). The significance of the differences in allele, genotype, and haplotype frequencies, and different models of inheritance between the control and patient groups, were compared either by the χ² test or with the maximum-likelihood χ² test. The odds ratio (OR) was used as a measure of the strength of association between allele, genotype, and haplotype frequencies and schizophrenia. The associations between the distribution of genotypes for IL-6, IL-2 and TNFα gene polymorphisms and control/patient groups were assessed by log-linear analysis. The non-random associations between individual gene polymorphisms were also checked with the linkage disequilibrium analysis. The influence of different genotypes on the time of the first episode of schizophrenia was shown with the hazard curves, which have been compared by the Tarone-Ware test. To reveal associations between PANSS subscales and age at onset of schizophrenia, a Grade Correspondence Analysis was used.²⁶ The two-way table with patients as rows, and PANNS subscales and time of the first episode occurrence as columns have been transformed into non-negative functions and presented as a graph of calculated grade density. This graph is presented as a map called the “overrepresentation map” of the Spearman correlation factor ρ*, which depicts the strength of positive dependence between patients and variables. After the Grade Correspondence Analysis procedure, clustering analysis of patients and columns was done (according to best ρ*). One-way, multivariate-analysis ANOVA was then used for comparisons of PANSS subscales and time of the first episode occurrence between clusters. The variables distribution was evaluated by the Shapiro-Wilk test. Homogeneity of variances was assessed by the Levene test.

Statistical calculations were performed with Statistica software (StatSoft, Inc., 2008), Version 8.0 (www.statsoft.com), GradeStat, and SNPStats (bioinfo.inconcologia.net). All p values were two-tailed, and p <0.05 was established as statistically significant.

Results

Comparison of Genotype and Allele Distributions Between Control and Patient Groups

There were statistically significant differences in the frequency of genotypes and alleles for IL-2 and TNFα polymorphisms, but not for IL-6, between schizophrenic and control subjects. In patients, genotype TT and allele T (IL-2) and genotype AA and allele A (TNFα) were overrepresented, as compared with the control group (Table 1).

Log-linear analysis shows the significant association between genotype distribution, patients/control groups, and genes (G²=77.82, p <0.0001), as well as between genotypes and patient/control group (G²=20.26, p <0.0001), and genotypes and genes (G²=42.86, p <0.0001).

Co-dominant, dominant, recessive, and over-dominant inheritance models were also tested, and the results are presented in Table 2. All those models (except over-dominant) are statistically significant for IL-2, IL-6, and TNFα gene polymorphisms. In the co-dominant model, patients with genotypes GT-TT (IL-2) and GA-AA (TNFα) have a higher risk of developing paranoid schizophrenia, as compared with GG genotype carriers. In the dominant model, patients with genotypes GT-TT (IL-2), GC-CC (IL-6), and GA-AA (TNFα) also are at increased risk for paranoid schizophrenia as compared with GG genotypes carriers. Finally, in the recessive model, patients carrying the GC genotype (IL-6) have a higher risk of schizophrenia development than those with GG and CC genotypes (Table 2).

Figure 1 [A] presents Kaplan-Meier estimates of age at onset of schizophrenia. Based on Kaplan-Meier analysis, one can expect that 25% of individuals will have the index episode of schizophrenia before age 26, and half of them before age 50 years. Hazard functions (HF) for schizophrenia development risk according to particular genotypes for each polymorphism are presented in Figure 1 [B–D]. For the IL-6 polymorphism, only a tendency to statistical significance between HF was observed (p=0.0533); patients with the GG genotype have lower risk of schizophrenia than those with other genotypes (lower quartiles: Q_1GG=31, Q_1GC=24, and Q_1CC=25, respectively). In case of IL-2 and TNFα polymorphisms, there were statistically significant differences in risk (related to time) of schizophrenia development according to different genotypes (p <0.0001 and p <0.05, respectively). Patients with the TT genotype for the IL-2 polymorphism have essentially higher risk of developing schizophrenia in almost all age ranges (lower quartiles, respectively: Q_1GG=36, Q_1GT=26, Q_1TT=22). Similarly, patients with the AA genotype for the TNFα polymorphism have an increased risk of schizophrenia, as compared with those with other genotypes (lower quartiles, respectively Q_1GG=26, Q_1GA=26, Q_1AA=24).

Linkage Disequilibrium Analysis and Haplotype Analysis

The linkage disequilibrium analysis has shown that there is a linkage disequilibrium between the IL-2 and IL-6 polymorphisms (D′=0.38; p <0.0001), and IL-2 and TNFα polymorphisms (D′=0.25; p <0.0001).

The results of the three-marker haplotype association tests are summarized in Table 3. Five haplotypes with frequency higher than 9% were recorded, and two of them display clinical importance. Haplotypes CTA and GTA (over-represented among the patient group) were found to increase the risk for schizophrenia significantly (OR: 4.39 and 5.9, respectively; Table 3).

Comparison of PANSS Items

All patients were assessed on the PANSS scale. To reveal associations between PANSS subscales and age at onset of schizophrenia, a novel method, called Grade Correspondence Analysis, was performed. As far as we know, this analysis is being used for the first time in genetic data analysis. The results of Grade Correspondence Analysis are presented as over-representation on the map of factor ρ and the derivative cluster analysis (Figure 2 [A]).

There are three distinct clusters indicating the associations between PANSS items and time of the first episode appearance. The first one (upper) involved patients who had the latest onset and low PANSS Negative (N) and General (G) scores. On the contrary, patients from the third cluster (lower) had the youngest age at onset and the highest mean scores for Negative and General symptoms. Patients from the second cluster have middle values of those parameters (Figure 2 [A]). Patients in the first cluster mainly carry the GGG haplotype (34.2%), whereas those in the next two clusters mainly carry the CTA haplotype (23.3% and 26.6%, respectively).

The results of ANOVA analysis are shown in Figure 2 [B]. There were statistically significant differences between clusters in Negative (p<0.0001) and General (p <0.0001) subscales, and age at onset (p <0.0001), but not on the positive (P) subscale (NS). The earlier the patients’ age at onset, the higher the mean scores for Negative and General PANSS subscales observed.

Moreover, in all clusters, statistically significant correlations were found between time of the initial episode appearance and PANNS items, as well as between PANSS General and PANSS Positive and Negative scales (Table 4).

Discussion

Schizophrenia is a multifactorial disease, with contributions from multiple susceptibility genes, epigenetic, and environmental factors. Although the exact cause of schizophrenia remains unknown, the possible role of the immune response system in the pathogenesis of schizophrenia has been indicated.¹

Currently, we attempted to establish an association between the polymorphisms in the promoter regions of IL-2 (−330G/T), IL-6 (−174G/C), and TNFα (−308G/A) genes and paranoid schizophrenia in a Polish population. To our knowledge, this is the first study investigating the combined impact of IL-2, IL-6, and TNFα gene polymorphisms on susceptibility to paranoid schizophrenia. Because of the fact that particular schizophrenia subtypes are characterized by different clinical pictures, it is reasonable to perform genetic association studies on homogenous groups of patients, especially where the impact of polymorphisms on psychopathology of schizophrenia is assessed. Genes and their polymorphisms were selected for the study based on an extensive literature review. As described in the introduction, many authors have emphasized the potential role of IL-2, IL-6, and TNFα in the development of schizophrenia.

The functional significance of selected IL-2, IL-6, and TNFα gene polymorphisms has been previously determined. An IL-2 genetic T→G polymorphism is located 330 bp upstream to the transcription start-site (within the region containing the NFAT transcription factor binding site).²⁷ Hoffmann et al.²⁸ reported that the GG genotype is associated with a high level of IL-2, whereas TT and GT genotypes are linked to reduction in the production of this cytokine. Four polymorphisms in the promoter region of the IL-6 gene have been described: at positions −597 G/A, −572 G/C, −373 A/G, and −174 G/C. The promoter region from −180 to −123 is crucial for transcriptional induction of IL-6 gene in response to several factors such as viruses or cytokines such as IL-1, TNFα.²⁹ Polymorphism of −174 G/C has an influence on the IL-6 expression level, which is increased by the presence of allele G, as compared with the allele C. However, some authors reported no linkage between the concentration of IL-6 and the polymorphism at position −174.³⁰ A TNFα promoter polymorphism at position -308 (−308G/A) influenced gene expression, and allele A is, according to some authors, associated with a higher level of TNFα production.³¹ However, other published reports indicate that allele A reduces TNFα expression or there is no difference in the TNFα mRNA levels between −308A and −308G alleles.^32,33

Epidemiological studies indicate that maternal infections (influenza, rubella, toxoplasmosis) during the first and the early second trimesters of pregnancy are linked to greater risk for schizophrenia development.^34–36 The precise neurobiological mechanism explaining those increased risks in relation to infections is not currently known. The role for cytokines and an impaired immune response to these infections during the critical period of brain development is commonly considered.³⁷ Results from studies conducted among pregnant women and in animal models indicate that the concentrations of cytokines such IL-1β, IL-6, and TNFα, which regulate normal brain development, are increased after infection.³⁸ Moreover, dysregulation of maternal cytokines may be responsible for behavioral deficits and/or cognitive functioning in offspring.³ Taking into account that mRNAs expression for cytokines in the CNS is developmentally regulated, altered level of cytokines, including IL-6 and TNFα after maternal infections may lead to abnormal cortical development, and, ultimately, schizophrenia.³⁸ We hypothesized that IL-6 and TNFα gene polymorphisms that contribute to changes in cytokine levels may impair the immune response to infections and also affect the development of the normal brain.

In this study, we have found statistically significant differences in the frequency distribution of genotypes and alleles for the IL-2 polymorphism between paranoid schizophrenia patients and controls. The presence of genotype TT and allele T correlates with an increasing risk of paranoid schizophrenia. Our results are in line with the results of Schwartz et al.,³⁹ who found a significant association of the IL-2 −330 TT genotype with schizophrenia in a German population.

It is known that IL-1, IL-6, and TNFα protect neurons against the toxic effects of such factors as β-amyloid peptide or N-methyl-d-aspartic acid (NMDA).⁴⁰ On the other hand, studies conducted in animal models have shown that chronic stimulation of TNFα exert a negative influence on the neurons’ viability.⁴¹ McGuire et al.⁴² have found that TNFα is toxic to embryonic mesencephalic dopaminergic neurons. Currently, we have found statistically significant differences in genotype and allele layout for the TNFα polymorphism between patient and control groups. Genotype AA and allele A (potentially connected with higher production of this cytokine) were over-represented among schizophrenic patients when compared with controls. Our results correspond with those observed by other authors, with an increased level of TNFα in serum and also an increased TNFα mRNA level in the prefrontal cortex of patients with schizophrenia.^43,44 Polymorphism −308G/A in the TNFα gene has also been examined in other populations. An association between this polymorphism and schizophrenia was recorded, for example, in the Italian population.⁴⁵ However, most studies reported no association between TNFα −308G/A polymorphism and schizophrenia.^46–48

Similar to Liu et al.,⁶ we did not find any significant differences in the occurrence of IL-6 −174 genotypes between patient and control groups. In our previous study, we found frequent occurrence of the C allele in both schizophrenic men and women, as compared with healthy individuals, but these differences were not statistically significant.²⁵

In the next stage, we conducted haplotype analysis to determine whether combinations of specific alleles is associated with greater risk of developing schizophrenia. It was found that haplotypes CTA and GTA correlate with increasing risk (4.4 times and 5.9 times, respectively) of paranoid schizophrenia development in this Polish population. Kampman et al.⁴⁷ investigated the interaction between the polymorphisms in TNFα (−308G/A) and EGF (61A/G) genes. They found that schizophrenic patients with a combination of EGF A/A and TNF-α G/A or A/A genotypes are characterized by earlier disease appearance than patients not having such configuration. Our results showed that in the group of patients with the earliest age at onset of schizophrenia, haplotype CTA was the most frequent.

As far as we know, this is also the first study that examines the impact of haplotypes of IL-2, IL-6, and TNFα gene polymorphisms on psychopathological symptoms in patients with paranoid schizophrenia. Because particular schizophrenia subtypes are characterized by different clinical pictures, it is reasonable to perform genetic association studies on homogenous groups of patients, for example, only those with a paranoid schizophrenia diagnosis, as in our case. An association was observed between improvement in the PANSS Negative and General scores, but not in PANSS Positive scores, and CTA haplotype. These results suggest that the CTA haplotype is more likely to contribute to the development of Negative and General symptoms of paranoid schizophrenia. Although IL-6 (−174G/C) polymorphism does not increased susceptibility to schizophrenia, it appears to be important in the context of psychopathology, as in patients with higher intensity of Negative and General PANSS symptoms, the co-presence of allele C (not G) with alleles T and A (for IL-2 -330G/T and TNFα −308G/A polymorphisms) was observed.

Dopamine and serotonin play a major role in mediating the psychotic symptoms of schizophrenia. The dopamine hypothesis of schizophrenia postulates that positive symptoms of schizophrenia may result from excess dopaminergic neurotransmission, particularly in mesolimbic and striatal brain regions, whereas dopaminergic deficits in prefrontal brain regions are responsible for the negative symptoms.⁴⁹ According to some authors, IL-2 and IL-6 may be associated with positive and negative symptoms arising in patients with schizophrenia. IL-2 increases dopamine turnover in the prefrontal cortex, whereas IL-6 induces higher activity of serotonin and mesocortical dopamine in the hippocampus and prefrontal cortex.⁵⁰ Licinio et al.⁵¹ postulated that IL-2 may cause increase in dopamine neurotransmission in some schizophrenic patients. There is a positive correlation between IL-2 and homovanillic acid (HVA) levels and between HVA and positive symptoms in schizophrenic patients. In contrast, some authors have also suggested a significant inverse relationship between IL-2 level and the PANSS Positive subscale.⁵⁰ We have found genotype TT and allele T (connected with lower production of IL-2) with a higher frequency in schizophrenic patients, and this may partly explain decreased dopamine turnover and the severity of negative symptoms in patients with the CTA haplotype. Taking into account that IL-6 also enhance dopamine turnover in the frontal cortex, the additional presence of the allele C for IL-6 polymorphism (potentially associated with decreased IL-6 expression) may also contribute to the higher intensity of negative symptoms. However, due to the fact that IL-6 has both pro- and anti-inflammatory activities, and molecular mechanism of transcriptional regulation of IL-6 encoding gene is intricate, the results we obtained are difficult to interpret. Such interpretation is also hampered by the fact that cytokines frequently exhibit redundant and pleiotropic effects. Moreover, specific combinations of cytokines may act synergistically or antagonistically, depending on the state of the target cells, and the combination of doses and timing sequence of cytokine release.⁴

It is postulated that schizophrenia with predominance of negative symptoms is associated with a hypo-dopaminergic state in the prefrontal cortex. According to some authors, TNFα may stimulate the catecholaminergic system, but chronic TNFα release has the opposite effect.⁵² We have found genotype AA and allele A (connected with higher production of TNFα) for TNFα −308G/A polymorphism, similarly to the CTA haplotype, to have a higher frequency in schizophrenic patients, and CTA haplotype was dominant in patients with the highest Negative and General symptoms scores. Naudin et al.⁵³ previously detected higher level of TNFα in schizophrenic patients, but this did not correspond to the clinical status.

To sum up, the T allele and homogenous TT genotype of −330G/T polymorphism in the IL-2 gene and A allele and AA genotype of −308G/A polymorphism in the TNFα gene may correlate with increasing risk of paranoid schizophrenia development in Polish population, but research on a larger patient group is needed.

Haplotypes CTA and GTA, respectively, are associated with a 4.4- and 5.9-fold increased risk of paranoid schizophrenia development.

Our results support the hypothesis that variations in promoters of IL-2, IL-6 and TNFα genes may contribute to the development and clinical course of paranoid schizophrenia.

The main limitation of the current study is that the relatively small sample size limits the generalizability of our findings. Future replication studies are required to confirm our observations.