Sage Journals: Discover world-class research

Abstract

Several factors have been implicated in schizophrenia (SZ), including human herpes viruses (HHV) and the adaptive immunity Human Leukocyte Antigen (HLA) genes. Here we investigated these issues in 2 complementary ways. In one analysis, we evaluated SZ-HLA and HHV-HLA associations at the level of a single allele by computing (a) a SZ-HLA protection/susceptibility (P/S) score based on the covariance between SZ and 127 HLA allele prevalences in 14 European countries, (b) estimating in silico HHV-HLA best binding affinities for the 9 HHV strains, and (c) evaluating the dependence of P/S score on HHV-HLA binding affinities. These analyses yielded (a) a set of 127 SZ-HLA P/S scores, varying by >200× (maximum/minimum), which could not be accounted for by chance, (b) a set of 127 alleles × 9 HHV best-estimated affinities, varying by >600×, and (c) a set of correlations between SZ-HLA P/S scores and HHV-HLA binding which indicated a prominent role of HHV1. In a subsequent analysis, we extended these findings to the individual person by taking into account the fact that every individual carries 12 HLA alleles and computed (a) the average SZ-HLA P/S scores of 12 randomly chosen alleles (2 per gene), an indicator of HLA-based SZ P/S for an individual, and (b) the average of the corresponding HHV estimated affinities for those alleles, an indicator of overall effectiveness of HHV-HLA binding. We found (a) that HLA protection for SZ was significantly more prominent than susceptibility, and (b) that protective SZ-HLA scores were associated with higher HHV-HLA binding affinities, indicating that HLA binding and subsequent elimination of several HHV strains may confer protection against schizophrenia.

Keywords

Human leukocyte antigen HHV1 schizophrenia immunogenetics epidemiology

Introduction

Schizophrenia is a complex and highly disabling mental disorder characterized by heterogeneous symptoms including hallucinations, delusions, disorganized speech, blunted affect, avolition, and anhedonia with a lifetime prevalence of ~1%.¹ Although the prevalence of schizophrenia is relatively low, its early onset and low remission together with high disability and psychosocial impairment result in a substantial global burden.² Accordingly, decades of research have been aimed at identifying risk and vulnerability factors to reduce the burden associated with this debilitating condition. Those efforts have documented a large genetic component to schizophrenia, with heritability estimated at around 80%.³ In addition, other factors have also been implicated in schizophrenia, including various infections as well as other environmental factors (eg, obstetric complications, birth season, cannabis use, and adverse childhood experiences).^4
-7 Indeed, considerable evidence supports a 2-hit theory that suggests that schizophrenia likely results from genetic vulnerability coupled with environmental hits during key neurodevelopmental periods.^6,8 In light of extensive research implicating exposure to infections and genetic liability in schizophrenia risk, here we focus on the influence of genes involved in the human immune response to infections, human leukocyte antigen (HLA) genes, and human herpes viruses on schizophrenia risk.

HLA and schizophrenia

HLA genes, located on chromosome 6, code for cell-surface proteins that are instrumental in immune system surveillance and elimination of foreign (eg, viral, bacterial) antigens via 2 primary pathways that work in concert to provide host protection. Class I HLA molecules (HLA-A, B, C) are expressed on nucleated cells and bind small peptides from proteolytically degraded cytosolic viruses and bacteria; those bound peptides are exported to the cell surface for presentation to CD8+ cytotoxic T cells, signaling cell destruction. Class II HLA molecules (HLA-DPB1, DQB1, DRB1) are expressed on lymphocytes and professional antigen-presenting cells (eg, macrophages, dendritic cells, and monocytes). Class II HLA presents larger peptides derived from endocytosed exogenous antigens to CD4+ T cells to facilitate B cell-mediated antibody production and adaptive immunity. More than 30 000 Class I and Class IIHLA alleles have been identified⁹; that variation maximizes population protection against infection by increasing the scope of antigens that can be bound and presented to the cell surface for destruction. Variation in HLA, particularly within the binding groove, has also been linked to variability in susceptibility to numerous human diseases.¹⁰

HLA is the most highly polymorphic region of the human genome,¹¹ differing not only across but also within populations.^12,13 The large number of polymorphisms coupled with the relatively low prevalence of schizophrenia and methodological variability across studies has resulted in inconsistent and modest associations in HLA candidate gene studies of schizophrenia¹⁴; however, early genome-wide association studies (GWAS) of schizophrenia have also documented associations with the HLA region,^15
-17 and subsequent evidence has continued to support that association.^18
-21 Furthermore, common HLA variants have been shown to influence brain abnormalities associated with schizophrenia including cerebral ventricle size,²² thalamic volume and asymmetry,²³ and hippocampal volume,²⁴ and variations in the HLA binding groove have been shown to influence treatment response with antipsychotic medications in patients with schizophrenia.²⁵ Indeed, HLA-schizophrenia associations have been one of the most commonly reported genetic associations in the literature,²⁶ though effect sizes are relatively small and findings have varied across studies due to variation between samples and geographical variation in HLA frequency.²⁷

An epidemiological approach to HLA-schizophrenia associations

HLA confers protection via pathogen elimination; consequently, HLA has evolved in concert with pathogen evolution^28,29 and has been shown to vary by population,^30,31 presumably reflecting differences in pathogen exposure. HLA-disease associations have been typically investigated using case-control studies.³² Although this design is solid, the overall results have been rather limited for at least 2 reasons. First, such studies are frequently underpowered due to difficulties in obtaining large numbers of participants, with a notable exception of the study of HLA and malaria in West Africa.³³ And second, the drawback of a small sample size is compounded by the insistence on finding HLA alleles with a very low P-value. Given that diseases are common outcomes of various factors, including pathogen insults, host responses, molecular mimicry, autoimmunity, etc., in various combinations, the objective should rather be to identify and determine the degree of involvement of different alleles in a given disease rather than the quest to find “the one” allele with a very low P-value in disease association. Such an allele, when found, would be ranked #1 and would be most probably due to a specific aspect of the disease (eg, autoimmunity) but other alleles could play smaller but still important roles (protective or susceptibility) in disease manifestation, thus effectively modulating the effect of the most prominent allele. Hence, a method that combines signed, weighted contributions of individual alleles for an overall protection/susceptibility assessment would seem more appropriate. Finally, GWAS has been used in recent studies to investigate HLA-disease associations. A limitation of this approach, as noted elsewhere,³⁴ is that HLA alleles are been detected through imputation rather than direct HLA sequencing. While more time- and cost-effective than direct sequencing, the accuracy of HLA imputation for certain alleles (eg, DRB1) has been reported to be as low as 30%.³⁵ Conversely, due to the highly polymorphic nature of HLA, direct sequencing of large enough samples to identify HLA-disease associations is often prohibitive.

Here we take an alternative immunogenetic epidemiology approach that takes advantage of the population heterogeneity of HLA and permits the identification of a high-resolution population-level HLA profile with regard to disease prevalence. Using this approach, we^36
-45 and others^30,46 have identified HLA alleles that are negatively associated with disease prevalence (ie, protective alleles) as well as HLA alleles that are positively associated with disease prevalence (ie, susceptibility alleles) in various neurological diseases,^36
-41 type 1 diabetes,⁴² and 30 types of cancer.^43
-45 In light of HLA’s involvement in pathogen elimination and autoimmunity, we presume that protective alleles facilitate pathogen elimination and that susceptibility alleles promote autoimmunity. Here we used this immunogenetic epidemiological approach above to evaluate associations between HLA frequency and schizophrenia prevalence and to derive a weighted overall measure along a protection-susceptibility continuum. Our working hypothesis is based on the following 3 assumptions: (1) a number of HLA alleles are involved in SZ, (2) the contribution of an allele to SZ can be negative (protective) or positive (susceptibility), and (3) the ultimate contribution of the HLA to SZ in a particular individual lies in the combined contributions of 12 HLA alleles carried by the individual, namely 2 of each 6 HLA classical genes (genes A, B, C of Class I and DPB1, DQB1, DRB1 of Class II); other nonclassical HLA genes (eg, DRB3, DRB4, etc.), not carried by all individuals, may contribute as well. The covariance of SZ prevalence and HLA allele frequency is a suitable quantitative estimate of the protective/susceptibility (P/S) contribution of an allele. Covariance possesses the following useful properties: (a) it can be easily calculated from the epidemiological/immunogenetic data (SZ prevalence and allele frequency), (b) it takes negative and positive values, corresponding to a protective or susceptibility contribution, (c) covariance estimates from different alleles can be added to derive and overall SZ P/S contribution of an ensemble of alleles as in the case of adding contributions of 12 alleles carried by an individual, (d) it can be normalized after division by the product of the data analyzed to yield a correlation coefficient, and, finally, (e) can be calculated from ranked data to provide a unit-free estimate of SZ P/S score. Here we use both the raw covariance and its normalized version (ie, correlation coefficient) to investigate (a) the SZ P/S contributions of various alleles and their 12-allele sets (the number of alleles carried by all individuals), and (b) the possible dependence of SZ prevalence on in silico estimated binding affinities of HLA alleles to 9 strains of human herpes virus.

Human herpes viruses (HHV) and schizophrenia

Infections, especially involving prenatal or childhood exposure, have been linked to adult schizophrenia risk purportedly via direct effects of pathogens and/or the effects of an inflammatory response to pathogens or autoimmunity on the developing brain.^47
-49 A number of pathogens have been implicated in schizophrenia including several human herpes viruses (HHV).⁵⁰ HHV are neurotropic viruses that are nearly ubiquitous and can cause lifelong infection characterized by alternating periods of latency and reactivation.^51
-53 They include herpes simplex virus 1 (HSV1/HHV1) and HHV2 (HHV2), varicella zoster virus (HHV3), Epstein-Barr virus (EBV/HHV4), cytomegalovirus (CMV/HHV5), HHV6a, HHV6b, HHV7, and HHV8. Though several studies have implicated HHV in schizophrenia, findings across studies have been highly inconsistent.^50,54 Nonetheless, intriguing links between HHV and schizophrenia exist. For instance, increased HHV antibodies have been documented in patients experiencing an initial episode of schizophrenia,^55,56 linked to higher suicide risk in patients with schizophrenia,⁵⁷ and have been associated with structural brain changes⁵⁶ and cognitive impairment^58,59 among patients with schizophrenia. In addition, studies have documented that anti-herpes treatment and anti-inflammatory medications improve schizophrenia symptoms.^55,60
-62 Furthermore, studies suggest that interactions between single nucleotide polymorphisms in the HLA region and exposure to HHV influence schizophrenia risk.^19,63,64

Consistent with studies documenting an interaction between the HLA region and HHV on schizophrenia,^19,63,64 we have hypothesized that exposure to foreign antigens, such as HHV, in the absence of HLA alleles that are capable of binding with those antigens can lead to antigen persistence, inflammation and/or autoimmunity, and downstream deleterious effects on the brain.^65
-68 Here we extend that line of research to schizophrenia, evaluating the relation between the frequency of HLA alleles and the prevalence of schizophrenia in 14 Continental Western European countries and evaluating the correlation between the HLA-schizophrenia profile and HLA-HHV binding affinity.

Methods

Prevalence of schizophrenia

Table 1 shows the population prevalence of schizophrenia in 2016 for each of the following 14 countries in Continental Western Europe (CWE): Austria, Belgium, Denmark, Finland, France, Germany, Greece, Italy, Netherlands, Portugal, Norway, Spain, Sweden, and Switzerland. Specifically, the total number of people with schizophrenia in each of the 14 CWE countries was identified from the Global Health Data Exchange,⁶⁹ a publicly available catalog of data from the Global Burden of Disease study, the most comprehensive worldwide epidemiological study of more than 350 diseases. The number of people with schizophrenia in each country was divided by the total population of each country in 2016⁷⁰ and expressed as a percentage. We have previously shown that life expectancy for these countries is virtually identical³⁶; therefore, life expectancy was not included in the current analyses.

Table 1.

Schizophrenia prevalence in the 14 CWE countries.

	Country	N Schizophrenia	N total population	Schizophrenia prevalence (%)
1	Austria	30 182	8 800 000	0.3430
2	Belgium	37 116	11 300 000	0.3285
3	Denmark	16 343	5 700 000	0.2867
4	Finland	17 425	5 500 000	0.3168
5	France	205 206	64 600 000	0.3177
6	Germany	271 954	82 600 000	0.3292
7	Greece	35 645	10 800 000	0.3300
8	Italy	224 454	60 600 000	0.3704
9	Netherlands	76 112	17 000 000	0.4477
10	Norway	17 555	5 200 000	0.3376
11	Portugal	36 207	10 300 000	0.3515
12	Spain	155 797	43 300 000	0.3598
13	Sweden	33 169	9 900 000	0.3350
14	Switzerland	29 960	8 400 000	0.3567

HLA alleles

We obtained the population frequency of 127 HLA alleles from 14 Continental Western European Countries (Austria, Belgium, Denmark, Finland, France, Germany, Greece, Italy, Netherlands, Portugal, Norway, Spain, Sweden, and Switzerland), as described elsewhere.³⁹ Those 127 alleles comprised 69 Class I and 58 Class II alleles, all of which were used to analyze the schizophrenia-HLA immunogenetic scores (see below). The alleles and their mean frequencies of the individual alleles are given in Tables 2 and 3 for HLA Classes I and II, respectively.

Table 2.

The 69 HLA Class I alleles used and their mean frequencies.

Class I
Gene A			Gene B			Gene C
	Allele	Frequency		Allele	Frequency		Allele	Frequency
1	A*01:01	0.1170	1	B*07:02	0.1009	1	C*01:02	0.0370
2	A*02:01	0.2715	2	B*08:01	0.0791	2	C*03:03	0.0506
3	A*02:05	0.0122	3	B*13:02	0.0238	3	C*04:01	0.1349
4	A*03:01	0.1501	4	B*14:01	0.0091	4	C*05:01	0.0716
5	A*11:01	0.0527	5	B*14:02	0.0275	5	C*06:02	0.0829
6	A*23:01	0.0237	6	B*15:01	0.0544	6	C*07:01	0.1472
7	A*24:02	0.1051	7	B*15:17	0.0104	7	C*07:02	0.1020
8	A*25:01	0.0139	8	B*15:18	0.0043	8	C*07:04	0.0146
9	A*26:01	0.0356	9	B*18:01	0.0609	9	C*12:02	0.0160
10	A*29:01	0.0058	10	B*27:02	0.0070	10	C*12:03	0.0678
11	A*29:02	0.0315	11	B*27:05	0.0435	11	C*14:02	0.0231
12	A*30:01	0.0165	12	B*35:01	0.0690	12	C*15:02	0.0370
13	A*30:02	0.0132	13	B*35:02	0.0187	13	C*16:01	0.0303
14	A*31:01	0.0295	14	B*35:03	0.0261
15	A*32:01	0.0368	15	B*35:08	0.0111
16	A*33:01	0.0116	16	B*37:01	0.0136
17	A*33:03	0.0066	17	B*38:01	0.0276
18	A*36:01	0.0063	18	B*39:01	0.0146
19	A*68:01	0.0353	19	B*39:06	0.0069
20	A*68:02	0.0220	20	B*40:01	0.0474
			21	B*40:02	0.0212
			22	B*41:01	0.0087
			23	B*41:02	0.0056
			24	B*44:02	0.0623
			25	B*44:03	0.0431
			26	B*44:05	0.0054
			27	B*45:01	0.0090
			28	B*47:01	0.0043
			29	B*49:01	0.0220
			30	B*50:01	0.0164
			31	B*51:01	0.0640
			32	B*52:01	0.0158
			33	B*55:01	0.0129
			34	B*56:01	0.0075
			35	B*57:01	0.0278
			36	B*58:01	0.0141

Table 3.

The 58 HLA Class II alleles used and their mean frequencies.

Class II
Gene DPB1			Gene DQB1			Gene DRB1
	Allele	Frequency		Allele	Frequency		Allele	Frequency
1	DPB1 * 01:01	0.0462	1	DQB1 * 02:01	0.1514	1	DRB1 * 01:01	0.0862
2	DPB1 * 02:01	0.1486	2	DQB1 * 02:02	0.0664	2	DRB1 * 01:02	0.0202
3	DPB1 * 02:02	0.0127	3	DQB1 * 03:01	0.1925	3	DRB1 * 01:03	0.0067
4	DPB1 * 03:01	0.1273	4	DQB1 * 03:02	0.1030	4	DRB1 * 03:01	0.1233
5	DPB1 * 04:01	0.3905	5	DQB1 * 03:03	0.0444	5	DRB1 * 04:01	0.0688
6	DPB1 * 04:02	0.1206	6	DQB1 * 04:02	0.0370	6	DRB1 * 04:02	0.0118
7	DPB1 * 05:01	0.0232	7	DQB1 * 05:01	0.1181	7	DRB1 * 04:03	0.0124
8	DPB1 * 06:01	0.0162	8	DQB1 * 05:02	0.0432	8	DRB1 * 04:04	0.0248
9	DPB1 * 09:01	0.0117	9	DQB1 * 05:03	0.0314	9	DRB1 * 04:05	0.0217
10	DPB1 * 10:01	0.0176	10	DQB1 * 06:01	0.0155	10	DRB1 * 04:07	0.0099
11	DPB1 * 11:01	0.0178	11	DQB1 * 06:02	0.1138	11	DRB1 * 04:08	0.0074
12	DPB1 * 13:01	0.0174	12	DQB1 * 06:03	0.0679	12	DRB1 * 07:01	0.1090
13	DPB1 * 14:01	0.0164	13	DQB1 * 06:04	0.0339	13	DRB1 * 08:01	0.0384
14	DPB1 * 17:01	0.0212	14	DQB1 * 06:09	0.0078	14	DRB1 * 08:03	0.0048
15	DPB1 * 19:01	0.0099				15	DRB1 * 09:01	0.0186
						16	DRB1 * 10:01	0.0124
						17	DRB1 * 11:01	0.0709
						18	DRB1 * 11:02	0.0076
						19	DRB1 * 11:03	0.0096
						20	DRB1 * 11:04	0.0498
						21	DRB1 * 12:01	0.0166
						22	DRB1 * 13:01	0.0713
						23	DRB1 * 13:02	0.0433
						24	DRB1 * 13:03	0.0130
						25	DRB1 * 13:05	0.0034
						26	DRB1 * 14:01	0.0226
						27	DRB1 * 15:01	0.1104
						28	DRB1 * 15:02	0.0115
						29	DRB1 * 16:01	0.0342

SZ-HLA protection/susceptibility scores

The PSCorr scores are the Fisher z-transformed correlations between HLA schizophrenia prevalence and HLA allele frequency (both log-transformed).³⁶

PSCorr = r^{'} = atanh (r)

(1)

The PSCov scores are simply the covariance between the log-transformed SZ prevalence and HLA allele frequency. The sign of both scores indicates protection (negative) or susceptibility (positive) to schizophrenia (P/S scores), whereas their absolute value indicates the strength of their P/S effect.

HHV proteins

The amino acid sequences of surface glycoproteins from HHV 1 to 8 (1, 2, 3, 4, 5, 6A, 6B, 7, 8) were retrieved from the UniprotKB database⁷¹ and are given in the Appendix. Table 4 gives the details of these proteins and associated information regarding the number of n-mers used in the analyses of Class I and Class II alleles.

Table 4.

Attributes of the 9 HHV viruses used.

Virus	Protein	UNIPROTKB ID	Protein length (AA)
HHV-1	Envelope glycoprotein D	Q69091	394
HHV-2	Envelope glycoprotein D	P03172	393
HHV-3	Envelope glycoprotein E	Q9J3M8	623
iHHV-4	Envelope glycoprotein B	P03188	857
HHV-5	Envelope glycoprotein B	P06473	906
HHV-6A	Glycoprotein Q2	P0DOE0	214
HHV-6B	Glycoprotein Q1	Q9QJ11	516
HHV-7	Envelope glycoprotein H	P52353	690
HHV-8	Envelope glycoprotein H	F5HAK9	730

A sliding window approach^67,68 (Figure 1) was used to partition the whole sequence of each glycoprotein into n-mers for HHV-HLA analyses. Since the exact n-mer length for binding to HLA Class I and Class II molecules is not known (only its range), we used 4 n-mer lengths (2 per HLA Class) to cover their approximate optimal range, namely 9- and 10-mer for Class I and 15-and 22-mer for Class II.⁷² For each n-mer, a set of subsequences was generated (number of subsequences = length of glycoprotein—n). For all n-mers (n = 9 and 10 for Class I, and n = 15 and 22 for Class II), subsequences were collected and queried in the IEDB database (www.iedb.org) in order to identify potential epitope peptides that are recognized by and bind to HLA Class I and II surface receptor proteins. IEDB queries were performed for each of the sliding-window aggregated sequences against HLA alleles available (see below). Binding affinity predictions were obtained using the NetMHCIIpan method.⁷³ For each n-mer, a binding affinity score was predicted and reported as a percentile rank by comparing the peptide’s score against the scores of 5 million random n-mers selected from the SwissProt database ⁷¹; smaller percentile ranks indicate higher binding affinity (“good binders”). For each HHV strain and HLA allele, the lowest percentile rank (LPR) was used as the highest estimated binding affinity measure for quantitative analyses. The distribution of LPR was skewed to the left (Figure 2A) in an exponential fashion and was log-transformed (equation (2)) to normalize it (Figure 2B), as illustrated in Figures 2 and 3, respectively.

\ln (LPR) = \log_{e} (LPR)

(2)

Figure 1.

Schematic diagram to illustrate the sliding window approach for estimating exhaustively in silico the binding affinity of all consecutive 9- and 15-mer peptides. The figure refers to HHV1 for illustration but the glycoproteins of all HHV strains (Table 4) were tested, as described in the text.

Figure 2.

Frequency distributions of $LPR$ [A] and $\ln (LPR)$ [B]. N = 113 alleles × 9 HHV strains = 1017 for each distribution.

Figure 3.

Probability-probability plots of $LPR$ [A] and $\ln (LPR)$ [B] to illustrate the substantial departure of the $LPR$ distribution from normal (diagonal) and conversion to close to normal when log-transformed.

Of a total of 127 alleles available (Tables 2 and 3), 113 were used for the analyses involving HHVs because 14 HLA Class I alleles could not be modeled by the NetMHCIIpan method (A*29:01, A*33:03, A*36:01, B*14:01, B*15:17, B*15:18, B*35:02, B*35:08, B*39:06, B*41:01, B*41:02, B*44:05, B*45:01, B*47:01).

Data analysis

Schizophrenia-HLA PSCorr scores

Standard parametric (mean, standard deviation, etc.) and nonparametric statistics (median, interquartile range, etc.) were used to evaluate the distribution of scores. We used Tukey’s⁷⁴ fences to identify outlier and extreme values of the PSCorr distribution, as follows.

Negative (protective) Inner Fence = Q 1 - 1.5 IQR

(3)

Negative (protective) Outer Fence = Q 1 - 3.0 IQR

(4)

Positive (susceptibility) Inner Fence = Q 3 + 1.5 IQR

(5)

Positive (susceptibility) Outer Fence = Q 3 + 3.0 IQR

(6)

Interquartile range : IQR = Q 3 - Q 1

(7)

where Q1, Q3 are the 25th and 75th percentiles, respectively. These fences demarcated outlier values (between inner and outer fences) and extreme values (outside the outer fences). A major objective of this study was to determine whether there was a statistically significant preponderance of protective or susceptibility schizophrenia PSCorr scores overall and within (and among) HLA classes and genes. For that purpose, we used a one-sample t-test with test value = 0 and estimated the mean preponderance, its statistical significance, and the effect size.^75,76

Random permutations test for assessing the statistical significance of the schizophrenia-HLA profile

The SZ-HLA profile is a vector of length N = 127 containing the estimated PSCorr scores. We tested the null hypothesis that the assignment of PSCorr to specific alleles may be due to chance by performing the permutation test below. Let $H$ be the original (observed) SZ-HLA profile and $H^{'}$ be the profile obtained by random permutation, as follows. (1) For each allele, its observed frequencies were paired to randomly permuted prevalences of the 14 CWE countries. (2) The Pearson correlation between allele frequency and corresponding (to the permuted country) SZ prevalence was computed. (3) These correlations were Fisher z-transformed to yield a “permuted” SZ-HLA profile, $H^{'}$ . (4) Next, we computed the sum of the absolute paired differences between $H$ and $H^{'}$ : a sum of zero would indicate that the 2 profiles are the same. (5) Finally, we carried out this procedure 1 000 000 times and counted the number of times M that that sum was equal to zero, indicating that the randomly obtained profile would be the same as the observed one. Then, the ratio $w = \frac{M}{1, 000, 000}$ is the probability that the observed profile $H$ could be due to chance. In a variant of this test, we determined in 1000 runs how frequently the rank of the PSCorr scores could occur randomly, thus relaxing the requirement of an exact match.

Application to individuals

Since every individual carries 12 classical HLA alleles (2 of each 3 HLA Class I genes and 2 of each 3 Class II genes), an average PSCorr score can be calculated:

\begin{array}{l} ξ = \bar{PSCorr} = [r^{'} (A_{1}) + r^{'} (A_{2}) + r^{'} (B_{1}) + r^{'} (B_{2}) \\ + r^{'} (C_{1}) + r^{'} (C_{2}) + + r^{'} (D P B 1_{1}) + r^{'} (D P B 1_{2}) \\ + r^{'} (D Q B 1_{1}) + r^{'} (D Q B 1_{2}) + r^{'} (D R B 1_{1}) \\ + r^{'} (D R B 1_{2})] / 12 \end{array}

(8)

Finally, we transformed $\bar{PSCorr}$ (an average of weighted $r^{'}$ ) to the original Pearson correlation scale, so that its range would be [−1, +1]:

ξ = \tanh \bar{(PSCorr})

(9)

We obtained expected estimates of ξ using a bootstrap procedure,⁷⁷as follows. For each gene, 2 $r^{'}$ scores were drawn randomly (with replacement) from the pool of available alleles of that gene and averaged to yield a bootstrap ξ*, for a simulated “individual.” The procedure was repeated 1 million times for a total of 1000 000 ξ*. In the resulting distribution, we identified negative (protective) and positive (susceptibility) outlier and extreme values of ξ*, as described above (equations (3)–(7)). The percentage of positive (susceptibility) outliers was compared to the percentages of schizophrenia prevalence above (Table 1) using an independent samples t-test.

Application to populations

In this analysis, we extended computations above to populations by using the average allele frequencies (Table 2) as weighting factors in equation (8):

\begin{array}{l} Ξ = \bar{PSCorr}' = [r^{'} (A_{1}) f (A_{1}) + r^{'} (A_{2}) f (A_{1}) \\ + r^{'} (B_{1}) f (B_{1}) + r^{'} (B_{2}) f (B_{2}) + r^{'} (C_{1}) f (C_{1}) + \\ r^{'} (C_{2}) f (C_{2}) + + r^{'} (D P B 1_{1}) f (D P B 1_{1}) \\ + r^{'} (D P B 1_{2}) f (D P B 1_{2}) + r^{'} (D Q B 1_{1}) f (D Q B 1_{1}) + \\ r^{'} (D Q B 1_{2}) f (D Q B 1_{2}) + r^{'} (D R B 1_{1}) f (D R B 1_{1}) \\ + r^{'} (D R B 1_{2}) f (D R B 1_{2})] / 12 \end{array}

(10)

We also carried out a bootstrap analysis of Ξ (N = 1 000 000) to obtain 1 million bootstrap Ξ* values for evaluating the representation in the population of allele frequency-weighted combinations of randomly selected 12 SZ PSCorr scores.

Association of schizophrenia PSCorr scores with HHV affinities

The second objective of this study was to assess the relation between schizophrenia PSCorr scores and the in silico estimated binding affinity of the corresponding alleles with the 9 HHVs (Table 4). For this purpose, we performed 2 analyses. At the single allele level, we carried out a stepwise linear regression, where the schizophrenia PSCorr score was the dependent variable and the HLA-HHV binding affinities, ln(LPR) (equation (2)), of the 9 HHV strains were the independent variables. At the level of the individual, we took into account the fact that an individual carries 12 HLA alleles, each of which has an estimated binding affinity to the 9 HHV strains. We first computed the average of those affinities for each allele and then computed the grand average binding affinity for the whole set of the randomly chosen alleles in the bootstrap procedure above. This provided a set of 1 million pairs consisting of (a) the average PSCorr (ξ*) of the 12 randomly chosen alleles, and (b) the grand average (ζ*) of the estimated HHV affinities for that set of alleles. The correlation between ξ* and ζ* is an estimate of the aggregate association of SZ PSCorr and HHV at the level of the individual.

Schizophrenia-HLA PSCov scores

The same analyses as above were performed on PSCov scores.

Implementation of analysis procedures

The IBM-SPSS statistical package (version 27) was used for implementing standard statistical analyses, including nonparametric exploratory data analysis,⁷⁴ regression analysis, and testing of proportions. All P values reported are 2-sided. The permutation test and bootstrap procedure were implemented using FORTRAN (Geany, version 1.38, built on or after 2021-10-09) and 64-bit Mersenne Twister random number generator with a large random double-precision odd seed.

Results

The SZ-HLA immunogenetic scores PSCorr and PSCov are given in Tables 5 and 6. The 2 scores were highly significantly and positively correlated (Figure 4; r = .937, P = 3.94 × 10⁻⁵⁹, N = 127).

Table 5.

Schizophrenia Immunogenetic scores for HLA Class I alleles.

Class I
Gene A			Gene B			Gene C
Allele	PSCorr	PScov	Allele	PSCorr	PSCov	Allele	PSCorr	PSCov
A*01:01	−.248	−.002528	B*07:02	.135	.006762	C*01:02	−.702	−.019900
A*02:01	−.183	−.002039	B*08:01	.078	.002951	C*03:03	−.119	−.006297
A*02:05	−.089	−.005197	B*13:02	.015	.000647	C*04:01	−.317	−.005400
A*03:01	−.200	−.007437	B*14:01	.078	.004697	C*05:01	.131	.005611
A*11:01	.081	.002187	B*14:02	−.011	−.000632	C*06:02	.467	.013424
A*23:01	.030	.001572	B*15:01	.168	.011421	C*07:01	−.142	−.001682
A*24:02	−.018	−.000413	B*15:17	−.304	−.017833	C*07:02	.309	.013948
A*25:01	.006	.000252	B*15:18	−.757	−.044797	C*07:04	−.066	−.003011
A*26:01	.110	.004525	B*18:01	−.195	−.011494	C*12:02	−.254	−.018436
A*29:01	−.490	−.028436	B*27:02	−.956	−.037332	C*12:03	−.107	−.005263
A*29:02	.158	.015396	B*27:05	−.301	−.018373	C*14:02	−.012	−.000765
A*30:01	.208	.012826	B*35:01	−.084	−.002405	C*15:02	.084	.003699
A*30:02	.136	.010422	B*35:02	−.450	−.026727	C*16:01	.399	.044524
A*31:01	.077	.003210	B*35:03	−.205	−.011519
A*32:01	.017	.000581	B*35:08	−.872	−.051535
A*33:01	−.154	−.010631	B*37:01	.344	.014528
A*33:03	−.374	−.021057	B*38:01	−.323	−.009779
A*36:01	−.058	−.003207	B*39:01	.074	.002330
A*68:01	.148	.003772	B*39:06	−.102	−.004251
A*68:02	−.094	−.008357	B*40:01	.004	.000311
			B*40:02	−.437	−.017620
			B*41:01	−.147	−.007591
			B*41:02	−.503	−.025780
			B*44:02	.075	.001952
			B*44:03	.184	.011248
			B*44:05	−.163	−.008413
			B*45:01	−.026	−.001748
			B*47:01	−.290	−.014689
			B*49:01	−.059	−.004296
			B*50:01	−.289	−.013632
			B*51:01	.128	.019624
			B*52:01	−.510	−.023853
			B*55:01	.425	.014714
			B*56:01	−.575	−.021447
			B*57:01	.291	.010743
			B*58:01	−.421	−.018861

Table 6.

Schizophrenia immunogenetic scores for HLA Class II alleles.

Class II
Gene DPB1			Gene DQB1			Gene DRB1
Allele	PSCorr	PSCov	Allele	PSCorr	PSCov	Allele	PSCorr	PSCov
DPB1 * 01:01	−.006	−.000296	DQB1 * 02:01	−.322	−.009983	DRB1 * 01:01	−.246	−.006472
DPB1 * 02:01	.146	.002918	DQB1 * 02:02	.368	.033053	DRB1 * 01:02	−.062	−.003773
DPB1 * 02:02	−.270	−.024760	DQB1 * 03:01	.249	.012999	DRB1 * 01:03	−.103	−.005160
DPB1 * 03:01	.194	.008266	DQB1 * 03:02	−.161	−.004341	DRB1 * 03:01	.403	.013356
DPB1 * 04:01	−.209	−.003374	DQB1 * 03:03	−.160	−.008421	DRB1 * 04:01	−0243	−.020587
DPB1 * 04:02	−.378	−.008933	DQB1 * 04:02	−.190	−.011956	DRB1 * 04:02	−.020	−.001756
DPB1 * 05:01	−.152	−.006469	DQB1 * 05:01	−.094	−.001957	DRB1 * 04:03	.028	.001474
DPB1 * 06:01	−.115	−.006798	DQB1 * 05:02	−.195	−.015455	DRB1 * 04:04	.252	.015393
DPB1 * 09:01	.305	.010258	DQB1 * 05:03	.468	.024950	DRB1 * 04:05	−.055	−.005655
DPB1 * 10:01	−.484	−.025886	DQB1 * 06:01	−.313	−.020175	DRB1 * 04:07	.200	0.016376
DPB1 * 11:01	.217	.018771	DQB1 * 06:02	−.217	−.011312	DRB1 * 04:08	−.365	−.030033
DPB1 * 13:01	−.336	−.015974	DQB1 * 06:03	−.112	−.003141	DRB1 * 07:01	.167	.008761
DPB1 * 14:01	−.153	−.006537	DQB1 * 06:04	.188	.007809	DRB1 * 08:01	−.172	−.012190
DPB1 * 17:01	.326	.018901	DQB1 * 06:09	.170	.004685	DRB1 * 08:03	−.373	−.016610
DPB1 * 19:01	.058	.004375				DRB1 * 09:01	−.012	−.001582
						DRB1 * 10:01	.146	.008958
						DRB1 * 11:01	.546	.034239
						DRB1 * 11:02	−.253	−.016144
						DRB1 * 11:03	−.009	−.000537
						DRB1 * 11:04	.369	.036578
						DRB1 * 12:01	−.138	−.005480
						DRB1 * 13:01	−.145	−.003224
						DRB1 * 13:02	.143	.003480
						DRB1 * 13:03	.106	.003680
						DRB1 * 13:05	−.820	−.039492
						DRB1 * 14:01	.008	.000502
						DRB1 * 15:01	−.192	−.008125
						DRB1 * 15:02	−.330	−.015932
						DRB1 * 16:01	−.093	−.007554

Figure 4.

SZ-HLA PSCorr scores are plotted against SZ-PSCov scores. See text for details.

Permutation tests

In the permutation test, no cases were found where a random SZ prevalence – HLA allele frequency pairing (out of 1000 000 runs) matched the observed SZ-HLA PSCorr or PSCov profiles (Tables 5 and 6), thus rejecting the null hypothesis that the observed profile could be accounted for by chance (P < 1 × 10⁻⁶) for both PSCorr and PSCov scores. In addition, in the ranks version of the random permutation test, which relaxed the requirement of an exact PSCorr match, we found that in only 3/1000 runs the ranks of the observed PSCorr scores matched the ranked observed SZ-HLA PSCorr profile, thus rejecting again the null hypothesis that the observed ranks of the PSCorr scores were due to chance (P = .003); no case matching the PSCov scores profile was found, the rejecting the null hypothesis that the observed ranks of the PSCov score were due to chance (P < .001).

Frequency distributions

The ranked SZ-HLA PSCorr scores are plotted in Figure 5, and their frequency distribution and box plot are shown in Figure 6A and B, respectively; detailed statistics are given in Table 7. It can be seen that there were 3 outliers (in red), with scores lower that the lower inner fence (equation (3), −0.081). These were HLA alleles B*27:02, B*35:08, and DRB1 * 13:05, illustrated in Figure 7, with SZ-HLA PSCorr scores of −0.956, −0.870, and −0.820, respectively. The correlation coefficients were all >.5 (r = .743 for B*27:02, 0.703 for B*35:08, and 0.675 for DRB1:13:05), indicating large effect sizes.⁷⁶

Figure 5.

Ranked SZ-HLA PSCorr are plotted against their ranks. Notice that protective scores (blue) are more numerous and stronger than susceptibility scores (red). See text for details.

Figure 6.

(A) frequency distribution of the 127 SZ-HLA PSCorr scores. (B) boxplot of the data in Figure 5. Q1, 25th percentile; Q3, 75th percentile; IQR, interquartile range.

Table 7.

Descriptive statistics of the distribution of SZ-HLA PSCorr scores. N = 127 HLA alleles.

Mean	−0.0763
SEM	0.0249
Median	−0.0838
SD	0.2808
Minimum	−0.9559
Maximum	0.5463
Range	1.5023
IQR	0.3772
25th percentile (Q1)	−0.2463
75th percentile (Q3)	0.1308

Figure 7.

Prevalence of schizophrenia is plotted against frequency of B*27:02 (r = .743, PSCorr = −.956), B*35:01 (r = −.703, PSCorr = −.872), and DRB1 * 13:05 (r = −.675, PSCorr = −.820), as indicated. The curves are a power fits. The fits are linear between log-log transformed data, from which the correlations above are derived.

Very similar results were obtained for PSCov, with the following 2 differences. (a) There was an additional protective outlier allele (B*15:18), and (b) there were 4 outlier susceptibility alleles (C*16:01, DRB1 * 11:04, DRB1 * 11:01, DQB1 * 02:02—from highest to lower score). These results indicate a higher sensitivity of the PSCov measure.

Tests of score proportions

Since the sign of the PSCorr score is determined by the sign of PSCov score, the following results apply to both scores. (a) There was a statistically significant higher proportion of protective scores (77/127 = 60.6%; P = .017, Wilson test of single proportion). (b) The proportions of protective scores in either Class I or Class II, and in any of the 6 genes did not differ from those of the susceptibility scores.

Tests of score magnitude

With respect to PSCorr scores, a statistically significant higher magnitude of protective (vs susceptibility) scores was observed (a) overall, (b) for HLA Class I, and (c) for HLA B gene (Figure 8); no significant effects were found for Class II or any other gene. Detailed statistics for the significant effects are given in Tables 8 and 9. Similarly, the magnitude of PSCov scores was significantly higher overall, HLA Class I, and HLA B gene.

Figure 8.

Mean (±95% CI) protective preponderance of SZ-HLA scores (negative/protective – positive/susceptibility SZ-HLA score) for the groups indicated. See text for details.

Table 8.

Statistics for statistically significant preponderance of protective (negative) Schizophrenia-HLA PSCorr scores (one sample t-test, test value = 0).

	N	Mean difference	95% Confidence Interval		\|t\|	Two-tailedP value
	N	Mean difference	Lower	Upper	\|t\|	Two-tailedP value
All	127	−0.0763	−0.1256	−0.0269	3.060	.003
Class I	69	−0.1050	−0.1758	−0.0342	2.958	.004
Gene B	36	−0.1661	−0.2773	−0.0549	3.034	.005

Table 9.

Effect sizes for the results in Table 7.

	Cohen’s d	Effect size	95% Confidence Interval
	Cohen’s d	Effect size	Lower	Upper
Overall	−0.272	Small-to-Medium	−0.448	−0.094
Class I	−0.356	Small-to-Medium	−0.598	−0.112
Gene B	−0.506	Medium-to-Large	−0.850	−0.155

Application to individuals

The following results were obtained from the bootstrap analysis of the PSCorr scores (descriptive statistics in Table 10); very similar results were obtained for the PSCov scores (not illustrated). Since the SZ-HLA PSCorr scores are population-level correlations between schizophrenia prevalence and HLA allele frequencies, it is reasonable to interpret them as reflecting the degree of protection/susceptibility conferred in general on an individual of the population by their specific HLA genetic makeup. For a particular individual, the average of the PSCorr scores of the 12 HLA alleles that the individual carries (equation (8)) can serve as an overall estimate, ξ, of SZ protection/susceptibility along a continuum ranging from −1 to +1 (equation (9)). Analysis of the distribution of the bootstrapped 1 million ξ* values revealed the following. (a) The distribution was unimodal, approximating a normal distribution but skewed to the left, that is, toward negative (protection) values (Figure 9). (b) Indeed, negative (protection) values (71.2%) outnumbered significantly positive (susceptibility) values (28.8%) (P < .001, Wilson score test for a single proportion). In addition, the mean of the distribution was negative, indicating a bias toward SZ protection (P < .001, one-sample t-test against the null hypothesis that the mean = 0). (c) There were 4404 (0.4%) negative (protective) ξ* outliers and 3 extreme ones (>Tukey’s outer fence, equation (4); Figures 10 and 11A). (d) There were 3043 (0.304%) positive (susceptibility outliers (Figures 10 and 11B). These proportions differed highly significantly (Z = 15.8, P < .00001). (e) Remarkably, the percent (0.3%) of positive (susceptibility) values of ξ* above was very close to, and statistically not significantly different from, the prevalences of schizophrenia used in this study (Table 1; P = .318, independent samples t-test).

Table 10.

Descriptive statistics of the distribution of ξ*. N = 1 000 000.

Mean	−0.00197
SEM	0.000005
Median	−0.00204
SD	0.0046
Minimum	−0.0246
Maximum	0.0208
Range	0.0455
IQR	0.006169
25th percentile (Q1)	−0.005092
75th percentile (Q3)	0.001077

Figure 9.

A, frequency distribution of 1 million bootstrap ξ* values.

Figure 10.

Boxplot of the data in Figure 9. Conventions as in Figure 6B.

Figure 11.

Frequency distribution of protective (A) and susceptibility (B) ξ* values.

Application to populations

In this analysis, we computed the average Ξ of SZ-HLA PSCorr scores weighted by the corresponding allele frequencies (equation 10) to obtain the population outcome of a particular combination of HLA alleles. The results obtained are shown in Table 11; very similar results were obtained for the SZ PSCov score (data not illustrated). More specifically: (a) The distribution of Ξ*was unimodal, approximating a normal distribution, with a skew to the left (ie, toward protection values). (b) Overall, negative (protection) values (69.3%) outnumbered significantly positive (susceptibility) values (30.7%) (P < .001, Wilson score test for a single proportion). In addition, the mean of the distribution was negative, indicating a bias toward SZ protection (P < .001, one-sample t-test against the null hypothesis that the mean = 0). (c) There were 5562 (0.556%) negative (protective) Ξ* outliers and 11 extreme ones (0.0011%). (d) There were 3134 (0.313%) positive (susceptibility outliers. These proportions differed highly significantly (Z = 26.09, P < .00001). (e) Similarly to the results of ξ* above, the percent of susceptibility outliers was very close to, and statistically not significantly different from, the prevalences of schizophrenia used in this study (Table 1; P = .434, independent samples t-test).

Table 11.

Descriptive statistics of the distribution of Ξ*. N = 1 000 000.

Mean	−0.03608
SEM	0.000071
Median	−0.03510
SD	0.0709
Minimum	−0.454
Maximum	0.295
Range	0.749
IQR	0.0946
25th percentile (Q1)	−0.08281
75th percentile (Q3)	0.01176

Schizophrenia PSCorr scores and HLA-HHV binding

In this analysis, we first estimated the highest binding affinity of each HLA allele to the 9 HHV strains (Table 4). We found that HLA alleles of Class I had significantly better binding affinities than those of Class II (Figure 12, P < .001, F-test in ANOVA). The average binding affinities did not differ significantly among genes of the same HLA Class (Figure 13; P = .884 for Class I genes, and P = .492 for Class II genes, ANOVA). Next, we evaluated the possible dependence of SZ-HLA P/S scores (PSCorr and PSCov) on HHV-specific HLA-binding affinities by performing a stepwise multiple linear regression, where the SZ-HLA immunogenetic score of individual alleles was the dependent variable and the 9 corresponding HHV-specific $\ln (LPR)$ values (one per allele and HHV strain) were the independent variables. Preliminary analyses revealed that there were significant differences between the 2 HLA classes, so we performed the regression analysis above separately for HLA Class I (N = 55) and Class II (N = 58) alleles. Moreover, for each HLA Class, we performed the analysis twice, namely 1 per n-mer (9- and 10-mer for Class I, and 15- and 22-nmer for Class II). The results were clear-cut, as follows. (a) No statistically significant associations were found in any Class II analyses (15- or 22-mer) between SZ PSCorr scores and HHV ln(LPR) for any HHV strain. In contrast, (b) a statistically significant association was found in both Class I analyses (9- and 10-mer) only between PSCorr scores and HHV1 ln(LRP) for both 9-mer (Figure 14A: r = .333, P = .013, N = 55) and 10-mer AA sequences (Figure 14B: r = .339, P = .011, N = 55); the effect sizes of these correlations were medium-to-large.^75,76 Practically the same associations were found between PSCov and HHV1 ln(LPR): (a) No statistically significant associations were found in any Class II analyses (15- or 22-mer) between SZ PSCov scores and HHV ln(LPR) for any HHV strain. In contrast, (b) a statistically significant association was found in both Class I analyses (9- and 10-mer) only between PSCov scores and HHV1 ln(LRP) for both 9-mer (r = .330, P = .014, N = 55) and 10-mer AA sequences (r = .305, P = .023, N = 55).

Figure 12.

Average (±95% CI) HHV1-HLA affinities [ln(LPR)] for HLA Class I and II.

Figure 13.

Average (±95% CI) HHV1-HLA affinities [ln(LPR)] for genes of HLA Class I and II.

Figure 14.

SZ-HLA PSCorr scores are plotted against $\ln (LPR)$ for HHV1 for 9-mer (A; left panel) and 10-mer (B; right panel) analyses. See text for details.

Combined SZ-HLA PSCorr scores versus combined HLA-HHV affinities

In the previous section, we analyzed the affinities of single HLA alleles to the various HHV strains. Since an individual carries 12 HLA alleles, it is reasonable to suppose that the ultimate influence of HHV on SZ susceptibility of an individual would be reflected in the correlation between the average PSCorr ξ (equation (8)) and the average estimated 9 HHV binding affinities ζ (equation (11)). For that purpose, we ran a bootstrap procedure (see Methods) and obtained 1 000 000 bootstrap values of ξ* and ζ*. We found that these measures were positively and highly significantly correlated (r = .644, Z = 764.98, P < .001). The data from a portion of that set (N = 1000) are plotted in Figure 15. Very similar results were obtained for PSCov (data not illustrated). These results indicate a strong dependence of overall individual SZ risk on combined HHV affinities, such that a HLA makeup with overall higher binding HHV affinity would confer protection from (ie, lower the risk of) schizophrenia.

Figure 15.

One-thousand ξ* values are plotted against the corresponding averages of 9 HHV ln(LPR) values. r = .638, P < .001. See text for details.

Discussion

Methodological considerations

The SZ-HLA PSCorr and PSCov scores were used here as continuously-varying measures of schizophrenia-HLA allele covariation at the population level and not as genetic causative factors testing a specific null hypothesis; hence no statistical tests of significance or P-values were computed for individual PSCorr scores since there was no specific null hypothesis to be tested. Instead, we tested the null hypothesis that the set (vector) of the 127 SZ-HLA PSCorr scores was due to chance by performing an extensive random permutation test. The results of this test rejected the null hypothesis above (P < 1 × 10⁻⁶) as well as its more liberal version that the ranked scores were due to chance (P < .003). In accordance with these results, the set of the observed SZ-HLA PSCorr scores was employed to evaluate its possible associations with the in silico estimated binding affinities to HHV proteins, which were used in the same way, namely as continuously-varying quantitative assessments HLA-HHV associations.

Remarkably, but unsurprisingly, practically identical results were obtained in all analyses when SZ PSCov scores were used, instead of PSCorr scores, since these measures were strongly and highly significantly correlated (Figure 4). In a way, the PSCov score would be somewhat more appropriate because it measures pure covariation, without standardization, which, although useful for statistical significant testing, it alters the covariance estimate itself. It should be mentioned that covariance has been usefully^78,79 (and profitably⁸⁰) employed in various fields, including evolution and natural selection^78,79 and finance.⁸⁰

General

A wealth of research has documented genetic and environmental influences on schizophrenia, including exposure to infections. Here we employed an epidemiological-immunogenetic approach^30,36
-46 to explore the relations between schizophrenia and HLA at the population level by computing the covariance (and its standardized version of correlation) between schizophrenia prevalence and HLA allele frequencies in 14 CWE countries. The choice of these countries was based mainly on the availability of HLA allele frequencies and on their practically identical life expectancies. The computed SZ-HLA PSCorr (and PSCov) scores were used to (a) estimate the degree of protection (negative score) or susceptibility (positive score) conferred by specific HLA alleles on the individuals of the population, (b) evaluate in silico possible associations between those scores and the binding affinities of the same HLA alleles to proteins of 9 HHV strains, and (c) investigate the possible dependence of the SZ P/S scores on HHV-HLA binding affinities. Our findings documented a preponderance of protective effects of HLA on schizophrenia prevalence and indicated that the SZ-HLA profile was highly associated with the estimated HLA binding affinity for HHV1, such that higher HLA-HHV1 binding affinity was associated with protective SZ-HLA scores. We then extended those analyses to aggregates of SZ-HLA scores and corresponding HHV-HLA binding affinities, since every individual carries 12 HLA alleles. Indeed, we found that at this aggregate level, SZ protection conferred by HLA was highly associated with better (more effective) binding affinities to HHV. Taken together, these findings suggest that HLA binding and subsequent elimination of HHV1 confer protection against schizophrenia in individuals and populations.

HLA and schizophrenia prevalence

With regard to the association of HLA and schizophrenia, we found there were more negative associations between HLA allele frequency and schizophrenia prevalence than there were positive associations, reflecting an overall protective effect. The relative protection conferred by HLA may partially account for the low prevalence of schizophrenia. Significant protective effects were documented here for Class I HLA, and the B genes, in particular. Notably, the HLA-B gene has been previously linked to the intergenerational transmission of schizophrenia risk.⁸¹ Here, 63.9% of the HLA-B genes were protective against schizophrenia and, of the 127 alleles investigated, the most highly significant protective effect was found for HLA-B*27:02 (r′ = −0.956), an allele that has been strongly linked to risk for ankylosing spondylitis in one of the strongest HLA-disease associations known.⁸² Interestingly, this negative SZ-HLA association corresponds to the similarly negative associations between ankylosing spondylitis and schizophrenia documented in epidemiological studies.⁸³ Although there was a preponderance of protective (ie, negative) HLA-SZ associations, 39% of HLA-SZ associations were positive, suggesting susceptibility toward SZ. HLA is most commonly associated with immune-mediated/autoimmune conditions.³³ In light of substantial evidence indicating robust associations between the risk of schizophrenia and several autoimmune disorders^84
-89the present findings suggest HLA may mediate that association. Furthermore, it has been suggested that some schizophrenia endophenotypes are driven by an infectious/inflammatory and/or autoimmune component ^27,62,90
-93; the present findings suggest that HLA may influence immune-mediated schizophrenia endophenotypes.

Implications for individuals

A major outcome of this study was the derivation of a unique, continuously-varying measure, ξ, assessing the protection/susceptibility of an individual to schizophrenia, based on their specific genetic HLA makeup. The allele-based SZ-HLA PSCorr scores provided the building blocks of ξ, which is a standardized score of the average 12 PSCorr scores of the HLA alleles carried by an individual (equations (8) and (9)). A reasonable interpretation of ξ would be as a measure of schizotypy,^94
-97 such that high values of ξ would correspond with increased liability for schizophrenia, as supported by the finding that the percent of very high (outlier) positive (susceptibility) values of ξ* in our simulation was 0.3%, a percent very close to, and statistically not significantly different from, the prevalences of schizophrenia used in this study (Table 1). The same considerations apply to the negative (protective) ξ values which could be interpreted as schizophrenia-resilience⁹⁶ scores. Since ξ is a normalized score with a range [−1, +1], it can be used as a threshold for detecting individuals resilient or susceptible to schizophrenia. Specifically, a value of ξ < − 0.25 (the value of negative inner fence, equation (3)) would indicate schizophrenia resilience, whereas a value of ξ > 0.16 (the value of positive inner fence, equation (5)) would indicate schizophrenia susceptibility, with continuous gradation in both domains.

SZ-HLA PSCorr scores and in silico HHV binding affinity

Several prior studies have documented the influence of HHV1 on schizophrenia. For instance, HHV1 antibodies are elevated in patients with schizophrenia compared to healthy controls^56,98 (cf, ref⁹⁹), and HHV1 exposure is associated with brain morphological anomalies^56,61,100 and cognitive impairments^{58,61,64,100
-102} in patients with schizophrenia. Here, we did not evaluate the association between HHV and schizophrenia but rather the influence of HLA on both by examining the dependence of SZ-HLA immunogenetic scores on HHV-specific HLA-binding affinities. Those analyses revealed a significant association between the HLA-schizophrenia profile and binding affinity only for HHV1, bolstering evidence regarding the influence of HHV1 on schizophrenia. As reviewed elsewhere,⁵⁰ associations between other HHV including CMV, HHV2, and EBV with schizophrenia have been mixed. Here, at the allele level, SZ P/S scores were found to be associated with HHV1-HLA affinity. However, at the allele aggregate level, where groups of 12 alleles were considered (as carried by every individual), the association between SZ-HLA P/S scores and HHV-HLA binding affinities was much stronger, indicating the involvement of several HHV strains acting in concert. In fact, this finding is in accord with the central hypothesis of this study, namely that SZ risk/protection is the outcome of a combined effect of groups of HLA alleles and not the exclusive effect of a specific allele.

Considering the above findings as a whole, the present study suggests that the association between HHV and schizophrenia may be partially attributable to HLA composition. The formation of an antigen-HLA molecule complex is a vital step in antigen elimination and host protection against foreign antigens. Given the extremely polymorphic nature of HLA, there is large variability in the binding affinity to a given antigen across HLA alleles. In the case of poor binding affinity between certain HLA molecules and antigen epitopes, the elimination of those antigens is blocked resulting in antigen persistence which may result in inflammation, autoimmunity, and disease.⁶⁵ Indeed, we have previously discussed antigen persistence due to HLA incongruence in relation to several conditions.^{65
-68,103
-106} We documented here that the HLA-schizophrenia profile that is primarily characterized by protective effects is significantly associated with HHV binding, suggesting that HLA binding and elimination of HHV protect against schizophrenia. The converse implication, which follows from the perspective of the persistent antigen theory, is that in the absence of HLA alleles capable of binding HHV antigens with sufficient affinity and immunogenicity, such antigens may persist,¹⁰⁷ ultimately contributing to damaging effects including those associated with schizophrenia. That is, the effect of HHV1 exposure may be partially mitigated by HLA with high binding affinity. Furthermore, the combined influence of HLA composition and HHV1 exposure may partly account for the low prevalence of schizophrenia despite high HHV1 seroprevalence in Europe.¹⁰⁸

The current findings and approach may also have implications that extend beyond schizophrenia given the high degree of genetic overlap that exists between risk for schizophrenia and other psychiatric disorders.^109
-111 For instance, previous studies have documented substantial genetic overlap between schizophrenia and bipolar disorder.^109
-111 Furthermore, like schizophrenia, both infections (eg, HHV¹¹²) and HLA¹¹³ have been implicated in bipolar disorder risk, and, like schizophrenia, anti-inflammatory agents have shown promising treatment effects for bipolar disorder.¹¹⁴ The extent to which the immunogenetic profile for bipolar disorder and other psychiatric conditions overlaps with that of schizophrenia remains to be investigated and is an interesting avenue for future studies.

The “crowdfunding” nature of HLA contribution to schizophrenia (and other diseases)

In evaluating HLA contributions to schizophrenia, it is important to consider the following: (1) the HLA region is the most polymorphic of the entire human genome, (2) every individual possesses 12 HLA alleles, and (3) each HLA allele may differ in terms of the direction of association with schizophrenia (and other diseases) and the magnitude of those associations. Consequently, rather than focusing on the individual “unique” contribution of specific alleles on schizophrenia susceptibility and protection, the primary emphasis here is on relatively small contributions of individual alleles that when combined as a set contribute to overall protection/susceptibility of schizophrenia in a population and for individuals. This “crowdfunding” nature of HLA contributions to schizophrenia moves away from single gene-disease associations toward considering polygenetic contributions to disease.

Conclusions

Overall the findings of the present study imply that schizophrenia prevalence is partially attributable to the combined effects of HLA composition and HHV1 exposure in Continental Western Europe. A significant strength of this study relative to GWAS studies that rely on imputation is that the approach used here permits evaluation of the frequency of high−resolution Class I and Class II HLA alleles with schizophrenia prevalence and estimated HHV binding affinity in silico. High-resolution HLA genotyping allows for the evaluation of HLA-disease associations at the protein-level where small variations may result in discordant disease associations such as that seen here for HLA-C*07:01 which was negatively correlated with schizophrenia prevalence and C*07:02 which was positively correlated with schizophrenia prevalence. In addition, the immunogenetic epidemiological approach used here takes advantage of HLA heterogeneity to evaluate the association of a large number of HLA alleles with schizophrenia at the population level, providing a broad picture of the influence of HLA on the population prevalence of schizophrenia. Given that schizophrenia is polygenic, our approach of combining signed, weighted contributions of HLA alleles to derive a unique, continuously-varying, normalized measure could easily be extended to include contributions of others genes for a more comprehensive estimate. At the minimum, determination of the complete, high-resolution HLA makeup of individuals in a population would be needed, in addition to population frequencies of other schizophrenia-related genes, and also mutations identified in GWAS studies.

Limitations of the study

Despite these strengths, it is also prudent to consider the present findings within the context of the study’s limitations. First of all, these findings hold for the European countries studied, and their extension to other countries remains to be determined, given the geographic heterogeneity of schizophrenia prevalence,² HLA allele frequency,^30,31 and pathogen exposure.¹¹⁵ Also, our study was limited to 127 HLA alleles, the only alleles that occurred in 9 or more of the countries investigated at the time the data was obtained; still, other alleles that were not included here may be relevant to schizophrenia and HHV, especially in countries of other geographical locations. An additional consideration is that this study was limited to HHV although several bacterial and other viral infectious agents have been implicated in schizophrenia.^4
-7 Future studies evaluating the HLA-schizophrenia profile in relation to the binding affinity of HLA to other infectious agents may reveal additional HLA-mediated infectious contributors to schizophrenia.

Footnotes

Appendix

Amino acid sequences of the 9 HHV strains (Table 4). Strain labels are from Uniprot (https://www.uniprot.org/uniprotkb/)

MGGAAARLGAVILFVVIVGLHGVRSKYALVDASLKMADPNRFRGKDLPVLDQLTDPPGVRRVYHIQAGLPDPFQPPSLPITVYYAVLERACRSVLLNAPSEAPQIVRGASEDVRKQPYNLTIAWFRMGGNCAIPITVMEYTECSYNKSLGACPIRTQPRWNYYDSFSAVSEDNLGFLMHAPAFETAGTYLRLVKINDWTEITQFILEHRAKGSCKYALPLRIPPSACLSPQAYQQGVTVDSIGMLPRFIPENQRTVAVYSLKIAGWHGPKAPYTSTLLPPELSETPNATQPELAPEDPEDSALLEDPVGTVAPQIPPNWHIPSIQDAATPYHPPATPNNMGLIAGAVGGSLLAALVICGIVYWMRRHTQKAPKRIRLPHIREDDQPSSHQPLFY

MGRLTSGVGTAALLVVAVGLRVVCAKYALADPSLKMADPNRFRGKNLPVLDRLTDPPGVKRVYHIQPSLEDPFQPPSIPITVYYAVLERACRSVLLHAPSEAPQIVRGASDEARKHTYNLTIAWYRMGDNCAIPITVMEYTECPYNKSLGVCPIRTQPRWSYYDSFSAVSEDNLGFLMHAPAFETAGTYLRLVKINDWTEITQFILEHRARASCKYALPLRIPPAACLTSKAYQQGVTVDSIGMLPRFIPENQRTVALYSLKIAGWHGPKPPYTSTLLPPELSDTTNATQPELVPEDPEDSALLEDPAGTVSSQIPPNWHIPSIQDVAPHHAPAAPSNPGLIIGALAGSTLAVLVIGGIAFWVRRRAQMAPKRLRLPHIRDDDAPPSHQPLFY

MGTVNKPVVGVLMGFGIITGTLRITNPVRASVLRYDDFHIDEDKLDTNSVYEPYYHSDHAESSWVNRGESSRKAYDHNSPYIWPRNDYDGFLENAHEHHGVYNQGRGIDSGERLMQPTQMSAQEDLGDDTGIHVIPTLNGDDRHKIVNVDQRQYGDVFKGDLNPKPQGQRLIEVSVEENHPFTLRAPIQRIYGVRYTETWSFLPSLTCTGDAAPAIQHICLKHTTCFQDVVVDVDCAENTKEDQLAEISYRFQGKKEADQPWIVVNTSTLFDELELDPPEIEPGVLKVLRTEKQYLGVYIWNMRGSDGTSTYATFLVTWKGDEKTRNPTPAVTPQPRGAEFHMWNYHSHVFSVGDTFSLAMHLQYKIHEAPFDLLLEWLYVPIDPTCQPMRLYSTCLYHPNAPQCLSHMNSGCTFTSPHLAQRVASTVYQNCEHADNYTAYCLGISHMEPSFGLILHDGGTTLKFVDTPESLSGLYVFVVYFNGHVEAVAYTVVSTVDHFVNAIEERGFPPTAGQPPATTKPKEITPVNPGTSPLLRYAAWTGGLAAVVLLCLVIFLICTAKRMRVKAYRVDKSPYNQSMYYAGLPVDDFEDSESTDTEEEFGNAIGGSHGGSSYTVYIDKTR

MTRRRVLSVVVLLAALACRLGAQTPEQPAPPATTVQPTATRQQTSFPFRVCELSSHGDLFRFSSDIQCPSFGTRENHTEGLLMVFKDNIIPYSFKVRSYTKIVTNILIYNGWYADSVTNRHEEKFSVDSYETDQMDTIYQCYNAVKMTKDGLTRVYVDRDGVNITVNLKPTGGLANGVRRYASQTELYDAPGWLIWTYRTRTTVNCLITDMMAKSNSPFDFFVTTTGQTVEMSPFYDGKNKETFHERADSFHVRTNYKIVDYDNRGTNPQGERRAFLDKGTYTLSWKLENRTAYCPLQHWQTFDSTIATETGKSIHFVTDEGTSSFVTNTTVGIELPDAFKCIEEQVNKTMHEKYEAVQDRYTKGQEAITYFITSGGLLLAWLPLTPRSLATVKNLTELTTPTSSPPSSPSPPAPSAARGSTPAAVLRRRRRDAGNATTPVPPTAPGKSLGTLNNPATVQIQFAYDSLRRQINRMLGDLARAWCLEQKRQNMVLRELTKINPTTVMSSIYGKAVAAKRLGDVISVSQCVPVNQATVTLRKSMRVPGSETMCYSRPLVSFSFINDTKTYEGQLGTDNEIFLTKKMTEVCQATSQYYFQSGNEIHVYNDYHHFKTIELDGIATLQTFISLNTSLIENIDFASLELYSRDEQRASNVFDLEGIFREYNFQAQNIAGLRKDLDNAVSNGRNQFVDGLGELMDSLGSVGQSITNLVSTVGGLFSSLVSGFISFFKNPFGGMLILVLVAGVVILVISLTRRTRQMSQQPVQMLYPGIDELAQQHASGEGPGINPISKTELQAIMLALHEQNQEQKRAAQRAAGPSVASRALQAARDRFPGLRRRRYHDPETAAALLGEAETEF

MESRIWCLVVCVNLCIVCLGAAVSSSSTSHATSSTHNGSHTSRTTSAQTRSVYSQHVTSSEAVSHRANETIYNTTLKYGDVVGVNTTKYPYRVCSMAQGTDLIRFERNIICTSMKPINEDLDEGIMVVYKRNIVAHTFKVRVYQKVLTFRRSYAYIYTTYLLGSNTEYVAPPMWEIHHINKFAQCYSSYSRVIGGTVFVAYHRDSYENKTMQLIPDDYSNTHSTRYVTVKDQWHSRGSTWLYRETCNLNCMLTITTARSKYPYHFFATSTGDVVYISPFYNGTNRNASYFGENADKFFIFPNYTIVSDFGRPNAAPETHRLVAFLERADSVISWDIQDEKNVTCQLTFWEASERTIRSEAEDSYHFSSAKMTATFLSKKQEVNMSDSALDCVRDEAINKLQQIFNTSYNQTYEKYGNVSVFETSGGLVVFWQGIKQKSLVELERLANRSSLNITHRTRRSTSDNNTTHLSSMESVHNLVYAQLQFTYDTLRGYINRALAQIAEAWCVDQRRTLEVFKELSKINPSAILSAIYNKPIAARFMGDVLGLASCVTINQTSVKVLRDMNVKESPGRCYSRPVVIFNFANSSYVQYGQLGEDNEILLGNHRTEECQLPSLKIFIAGNSAYEYVDYLFKRMIDLSSISTVDSMIALDIDPLENTDFRVLELYSQKELRSSNVFDLEEIMREFNSYKQRVKYVEDKVVDPLPPYLKGLDDLMSGLGAAGKAVGVAIGAVGGAVASVVEGVATFLKNPFGAFTIILVAIAVVIITYLIYTRQRRLCTQPLQNLFPYLVSADGTTVTSGSTKDTSLQAPPSYEESVYNSGRKGPGPPSSDASTAAPPYTNEQAYQMLLALARLDAEQRAQQNGTDSLDGQTGTQDKGQKPNLLDRLRHRKNGYRHLKDSDEEENV

MHFLVVYILIHFHAYRGMAALPLFSTLPKITSCCDSYVVINSSTSVSSLISTCLDGEILFQNEGQKFCRPLTDNRTIVYTMQDQVQKPLSVTWMDFNLVISDYGRDVINNLTKSAMLARKNGPRYLQMENGPRYLQMETRISDLFRHECYQDNYYVLDKKLQMFYPTTHSNELLFYPSEATLPSPWQEPPFSSPWPEPTFPSRWYWLLLNYTNY

MRPPRRSAPILVCAISMATALSNATVHRDAGTVESTPPPDDEDNYTAKYYDDSIYFNIYDGTNPTPRRRTLPEIISKFSTSEMSRLGGLKAFVPVDYTPTTTLEDIEDLLNYAICDDNSCGCLIETEARXMFGDIIICVPLSAESRGVRNLKSRIMPMGLSQILSSGLGLHFSLLYGAFGSNYNSLAYMERLKPLTAMTAIAFCPMTSKLELRQNYRLEKARXNLIVNIELLKIQNHGGQTIKTLTSFAIVRKDSDGQDWETCTRFASVSIEDILRSKPAANGTCCPPRDVHHDRPTLQSSNSWTRTEYFEPWQDVVDAYVPINDNHCPNDSYVVFQTLQGHEWCSRLNKNDTKNYLSSVLAFKNALYETEELMETIGMRLASQILSLVGQRGTSIRNIDPAIVSALWHSLPEKLTTTNIKYDIASPTHMSPALXTIFIQTGTSKQRFRNAGLLMVNNIFTVQARYSKQNMFEKKIYGYEHLGQALCEGGHVFYNPRDVYFQNIKMAATEPTVVRT

MYFYINSLLLIVSINGWKHWNILNSSICVNEKTNQTIIQPGLITFNFHDYNETRVYQIPKCLFGYTFVSNLFDSVNFDESFDQYKHRITRFFNPSTEKAVKIYAQKFQTNIKPVSHTKTITVSFLPLFYEKDVYFANVSEIRKLYYNQYICTLSNGLTDYLFPITERCVMRHYNYLNTVFMLALTPSFFIISVETGMDDVVFIFGNVSRIFFKAPFRKSSFIYRQTVSDDLLLITKKTTIERFYPFLKIDFLDDIWKQNYDISFLIAKFNKLATVYIMEGFCGKPVNKDTFHLMFLFGLTHFLYSTRGDGLLPLLEILNTHQSIITMGRFLEKCFKMTKSHLLYPEMEKLQNFQLVDYSYITSDLTIPISAKLAFLSLADGRIVTVPQNKWKEIENNIETLYEKHKLFTNLTQPERANLFLLSEIGNSLVFQEKIKRKIHVLLASLCNPLEMYFWTHMLDNVMDIETMFSPCATATRKDLTQRVVNNILSYKNLDAYTNKVMNTLSVYRKKRLDMFKSISCVSNEQAAFLTLPNITYTISSKYILAGTSFSVTSTVISTTIIITVVPLNSTCTPTNYKYSVKNIKPIYNISSHDCVFCESLVVEYDDIDGIIQFVYIMDDKQLLKLIDPDTNFIDVNPRTHYLLFLRNGSVFEITALDLKSSQVSIMLVLLYLIIIIIVLFGIYHVFRLF

MQGLAFLAALACWRCISLTCGATGALPTTATTITRSATQLINGRTNLSIELEFNGTSFFLNWQNLLNVITEPALTELWTSAEVAEDLRVTLKKRQSLFFPNKTVVISGDGHRYTCEVPTSSQTYNITKGFNYSALPGHLGGFGINARLVLGDIFASKWSLFARDTPEYRVFYPMNVMAVKFSISIGNNESGVALYGVVSEDFVVVTLHNRSKEANETASHLLFGLPDSLPSLKGHATYDELTFARNAKYALVAILPKDSYQTLLTENYTRIFLNMTESTPLEFTRTIQTRIVSIEARRACAAQEAAPDIFLVLFQMLVAHFLVARGIAEHRFVEVDCVCRQYAELYFLRRISRLCMPTFTTVGYNHTTLGAVAATQIARVSATKLASLPRSSQETVLAMVQLGARDGAVPSSILEGIAMVVEHMYTAYTYVYTLGDTERKLMLDIHTVLTDSCPPKDSGVSEKLLRTYLMFTSMCTNIELGEMIARFSKPDSLNIYRAFSPCFLGLRYDLHPAKLRAEAPQSSALTRTAVARGTSGFAELLHALHLDSLNLIPAINCSKITADKIIATVPLPHVTYIISSEALSNAVVYEVSEIFLKSAMFISAIKPDCSGFNFSQIDRHIPIVYNISTPRRGCPLCDSVIMSYDESDGLQSLMYVTNERVQTNLFLDKSPFFDNNNLHIHYLWLRDNGTVVEIRGMYRRRAASALFLILSFIGFSGVIYFLYRLFSILY

Author Contributions

LMJ and SAC contributed to data retrieval. APG contributed to data analysis. LMJ and APG wrote the paper. LMJ, SAC, and APG edited and approved the paper.

Funding:

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Partial funding for this study was provided by the University of Minnesota (the Kunin Chair in Women’s Healthy Brain Aging, the Brain and Genomics Fund, the McKnight Presidential Chair of Cognitive Neuroscience, and the American Legion Brain Sciences Chair) and the U.S. Department of Veterans Affairs. The sponsors had no role in the current study design, analysis or interpretation, or in the writing of this paper. The contents do not represent the views of the U.S. Department of Veterans Affairs or the United States Government.

Declaration of conflicting interests:

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data Availability

The datasets generated and analyzed in the current study are publicly available.

Ethical Approval

This article does not contain any studies with human participants performed by any of the authors.

ORCID iDs

Lisa M James

Apostolos P Georgopoulos

Significance Statement

Schizophrenia prevalence covaries with the population frequencies of human leukocyte antigen (HLA) alleles, which, in turn, bind with various affinities to human herpes viruses. HLA alleles protective against schizophrenia had higher binding affinities to HHV, suggesting that HLA binding and subsequent elimination of HHV may confer protection against schizophrenia.

References

Saha

Chant

Welham

McGrath

A systematic review of the prevalence of schizophrenia. PLoS Med. 2005;2:e141.

Charlson

Ferrari

Santomauro

, et al. Global epidemiology and burden of schizophrenia: findings from the Global Burden of Disease Study 2016. Schizophr Bull. 2018;44:1195-1203.

Sullivan

Kendler

Neale

MC.

Schizophrenia as a complex trait: evidence from a meta-analysis of twin studies. Arch Gen Psychiatry. 2003;60:1187-1192.

Brown

AS.

The environment and susceptibility to schizophrenia. Prog Neurobiol. 2011;93:23-58.

Rapoport

Giedd

Gogtay

Neurodevelopmental model of schizophrenia: update 2012. Mol Psychiatry. 2012;17:1228-1238.

Davis

Eyre

Jacka

, et al. A review of vulnerability and risks for schizophrenia: beyond the two hit hypothesis. Neurosci Biobehav Rev. 2016;65:185-194.

Robinson

Bergen

SE.

Environmental risk factors for schizophrenia and bipolar disorder and their relationship to genetic risk: current knowledge and future directions. Front Genet. 2021;12:686666.

Bayer

Falkai

Maier

Genetic and non-genetic vulnerability factors in schizophrenia: the basis of the “two hit hypothesis”. J Psychiatr Res. 1999;33:543-548.

Robinson

Barker

Georgiou

Cooper

Flicek

Marsh

SGE

. IPD-IMGT/HLA Database. Nucleic Acids Res. 2020;48:D948-D955.

10.

Dendrou

Petersen

Rossjohn

Fugger

HLA variation and disease. Nat Rev Immunol. 2018;18:325-339.

11.

Trowsdale

Knight

JC.

Major histocompatibility complex genomics and human disease. Annu Rev Genomics Hum Genet. 2013;14:301-323.

12.

Heath

Gut

Brennan

, et al. Investigation of the fine structure of European populations with applications to disease association studies. Eur J Hum Genet. 2008;16:1413-1429.

13.

Buhler

Nunes

Nicoloso

Tiercy

Sanchez-Mazas

The heterogeneous HLA genetic makeup of the Swiss population. PLoS One. 2012;7:e41400.

14.

Wright

Nimgaonkar

Donaldson

Murray

RM.

Schizophrenia and HLA: a review. Schizophr Res. 2001;47:1-12.

15.

Purcell

Wray

Stone

JL.

Common polygenic variation contributes to risk of schizophrenia and bipolar disorder. Nature. 2009;460:748-752. doi:10.1038/nature0818

16.

Shi

Levinson

Duan

, et al. Common variants on chromosome 6p22. 1 are associated with schizophrenia. Nature. 2009;460:753-757. doi:10.1038/nature08192

17.

Stefansson

Ophoff

Steinberg

, et al. Common variants conferring risk of schizophrenia. Nature. 2009;460:744-747. doi:10.1038/nature08186

18.

The Schizophrenia Psychiatric Genome-Wide Association Study (GWAS) Consortium. Genome-wide association study identifies five new schizophrenia loci. Nat Genet. 2011;43:969-976. doi:10.1038/ng.940

19.

Bamne

Wood

Chowdari

, et al. Evaluation of HLA polymorphisms in relation to schizophrenia risk and infectious exposure. Schizophr Bull. 2012;38:1149-1154.

20.

Irish Schizophrenia Genomics Consortium and the Wellcome Trust Case Control Consortium 2. Genome-wide association study implicates HLA-C*01:02 as a risk factor at the major histocompatibility complex locus in schizophrenia. Biol Psychiatry. 2012;72:620-628.

21.

Sekar

Bialas

de Rivera

, et al. Schizophrenia risk from complex variation of complement component 4. Nature. 2016;520:177-183. doi:10.1038/nature16549

22.

Agartz

Brown

Rimol

, et al. Common sequence variants in the major histocompatibility complex region associate with cerebral ventricular size in schizophrenia. Biol Psychiatry. 2011;70:696-698.

23.

Brucato

Guadalupe

Franke

Fisher

Francks

A schizophrenia-associated HLA locus affects thalamus volume and asymmetry. Brain Behav Immun. 2015;46:311-318. doi:10.1016/j.bbi.2015.02.021

24.

Walters

Rujescu

Franke

, et al. The role of the major histocompatibility complex region in cognition and brain structure: a schizophrenia GWAS follow-up. Am J Psychiatr. 2013;170:877-885.

25.

Le Clerc

Taing

Fond

, et al. A double amino-acid change in the HLA-A peptide-binding groove is associated with response to psychotropic treatment in patients with schizophrenia. Transl Psychiatry. 2015;5:e608-e608.

26.

Legge

Santoro

Periyasamy

Okewole

Arsalan

Kowalec

Genetic architecture of schizophrenia: a review of major advancements. Psychol Med. 2021;51:2168-2177.

27.

Kodavali

Watson

Prasad

, et al. HLA associations in schizophrenia: Are we re-discovering the wheel? Am J Med Genet A. 2014;165:19-27. doi:10.1002/ajmg.b.32195

28.

Sanchez-Mazas

Lemaître

Currat

Distinct evolutionary strategies of human leucocyte antigen loci in pathogen-rich environments. Philos Trans R Soc Lond B Biol Sci. 2012;367:830-839.

29.

Salamon

Klitz

Easteal

, et al. Evolution of HLA class II molecules: Allelic and amino acid site variability across populations. Genetics. 1999;152:393-400.

30.

Garamszegi

LZ.

Global distribution of malaria-resistant MHC-HLA alleles: the number and frequencies of alleles and malaria risk. Malar J. 2014;13:349.

31.

Singh

Kaul

Khan

A comparative review of HLA associations with hepatitis B and C viral infections across global populations. World J Gastroenterol. 2007;13:1770-1787. doi:10.3748/wjg.v13.i12.1770

32.

Blackwell

Jamieson

Burgner

HLA and infectious diseases. Clin Microbiol Rev. 2009;22:370-385.

33.

Hill

Yates

Allsopp

, et al. Human leukocyte antigens and natural selection by malaria. Philos Trans R Soc Lond B Biol Sci. 1994;346:379-385. doi:10.1098/rstb.1994.0155.

34.

Hollenbach

Oksenberg

JR.

The immunogenetics of multiple sclerosis: A comprehensive review. J Autoimmun. 2015;64:13-25. doi:https://doi.org/10.1016/j.jaut.2015.06.010

35.

Vlachopoulou

Lahtela

Wennerström

, et al. Evaluation of HLA-DRB1 imputation using a Finnish dataset. Tissue Antigens. 2014;83:350-355. doi:10.1111/tan.12343

36.

James

Georgopoulos

The human leukocyte antigen (HLA) DRB1*13:02 allele protects against dementia in continental Western Europe. J Neurol Neuromed. 2019;4:1-6. doi:10.29245/2572.942X/2019/5.1253

37.

James

Georgopoulos

Tri-allelic Human Leukocyte Antigen (HLA) protection against dementia. J Neurol Neuromed. 2019;5:12-17. doi:10.29245/2572.942X/2019/1.1261

38.

James

Georgopoulos

Shared human leukocyte antigen (HLA) coverage in dementia and Parkinson’s disease. J Neurol Neuromed. 2020;5:45-54. doi:10.29245/2572.942X/2020/3.1275

39.

James

Georgopoulos

AP.

Immunogenetic epidemiology of dementia and Parkinson’s disease in 14 continental European countries: shared human leukocyte antigen (HLA) profiles. Journal Immunol Sci. 2021;5:16-26. doi:10.29245/2578-3009/2021/2.1209

40.

James

Georgopoulos

AP.

Immunogenetic epidemiology of multiple sclerosis in 14 continental Western European countries. J Immunol Sci. 2021;5:40-46. doi:10.29245/2578.3009/2021/2.1216

41.

James

Georgopoulos

AP.

Immunogenetic epidemiology of motor neuron diseases in 14 continental Western European countries. J Immunol Sci. 2021;5:22-28. doi:http://dx.doi.org/10.29245/2578-3009/2021/3.1221

42.

James

Georgopoulos

AP.

Immunogenetic epidemiology of Type 1 diabetes in 14 Continental Western European countries. J Immunol Sci. 2021;5:29-35. doi:10.29245/2578-3009/2021/3.1219

43.

Georgopoulos

James

LM.

Immunogenetic association of 127 Human Leukocyte Antigen (HLA) alleles with 30 cancers in continental Western European countries. Journal Immunol Sci. 2022;6:6-17. doi:10.29245/2578-3009/2022/2.1227

44.

James

Georgopoulos

AP.

Protective and susceptibility wffects of Human Leukocyte Antigen on melanoma prevalence and their implications for predicting checkpoint blockade immunotherapy. J Immunol Sci. 2022;6:25-35.

45.

James

Georgopoulos

AP.

Immunogenetic profiles and associations of breast, cervical, ovarian, and uterine cancers. Cancer Inform. 2023;22:2211769351221148588. doi:10.1177/11769351221148588

46.

Fainboim

Cañero Velasco

Marcos

, et al. Protracted, but not acute, hepatitis A virus infection is strongly associated with HLA-DRB*1301, a marker for pediatric autoimmune hepatitis. Hepatology. 2001;33:1512-1517. doi:10.1053/jhep.2001.24562

47.

Brown

Derkits

EJ.

Prenatal infection and schizophrenia: a review of epidemiologic and translational studies. Am J Psychiatr. 2010;167:261-280.

48.

Khandaker

Zimbron

Dalman

Lewis

Jones

PB.

Childhood infection and adult schizophrenia: a meta-analysis of population-based studies. Schizophr Res. 2012;139:161-168.

49.

Khandaker

Zimbron

Lewis

Jones

PB.

Prenatal maternal infection, neurodevelopment and adult schizophrenia: a systematic review of population-based studies. Psychol Med. 2013;43:239-257. doi:10.1017/S0033291712000736

50.

Arias

Sorlozano

Villegas

, et al. Infectious agents associated with schizophrenia: a meta-analysis. Schizophr Res. 2012;136:128-136.

51.

Staras

Dollard

Radford

Flanders

Pass

Cannon

MJ.

Seroprevalence of cytomegalovirus infection in the United States, 1988-1994. Clin Infect Dis. 2006;43:1143-1151. doi:10.1086/508173.51

52.

Wald

Corey

. Persistence in the population: epidemiology, transmission. In: Arvin

Campadelli-Fiume

Mocarski

, et al., eds. Human Herpesviruses: Biology, Therapy, and Immunoprophylaxis. Cambridge University Press; 2007. https://www.ncbi.nlm.nih.gov/books/NBK47447

53.

McQuillan

Kruszon-Moran

Flagg

Paulose-Ram

Prevalence of herpes simplex virus type 1 and type 2 in persons aged 14–49: United States, 2015–2016. NCHS Data Brief, no 304. Hyattsville, MD: National Center for Health Statistics; 2018.

54.

Yolken

Torrey

EF.

Viruses, schizophrenia, and bipolar disorder. Clin Microbiol Rev. 1995;8:131-145.

55.

Torrey

Leweke

Schwarz

, et al. Cytomegalovirus and schizophrenia. CNS Drugs. 2006;20:879-885.

56.

Prasad

Shirts

Yolken

Keshavan

Nimgaonkar

VL.

Brain morphological changes associated with exposure to HSV1 in first-episode schizophrenia. Mol Psychiatry. 2007;12:105-11313, 1. doi:10.1038/sj.mp.4001915

57.

Dickerson

Origoni

Schweinfurth

LAB

, et al. Clinical and serological predictors of suicide in schizophrenia and major mood disorders. J Nerv Ment Dis. 2018;206:173-178.

58.

Dickerson

Boronow

Stallings

Origoni

Ruslanova

Yolken

RH.

Association of serum antibodies to herpes simplex virus 1 with cognitive deficits in individuals with schizophrenia. Arch Gen Psychiatry. 2003;60:466-472.

59.

Shirts

Prasad

Pogue-Geile

Dickerson

Yolken

Nimgaonkar

Antibodies to cytomegalovirus and herpes simplex virus 1 associated with cognitive function in schizophrenia. Schizophr Res. 2008;106:268-274. doi:10.1016/j.schres.2008.07.011

60.

Dickerson

Boronow

Stallings

Origoni

Yolken

RH.

Reduction of symptoms by valacyclovir in cytomegalovirus-seropositive individuals with schizophrenia. Am J Psychiatr. 2003;160:2234-2236.

61.

Prasad

Watson

Dickerson

Yolken

Nimgaonkar

VL.

Exposure to herpes simplex virus type 1 and cognitive impairments in individuals with schizophrenia. Schizophr Bull. 2012;38:1137-1148.

62.

Müller

Inflammation in schizophrenia: pathogenetic aspects and therapeutic considerations. Schizophr Bull. 2018;44:973-982.

63.

Kim

Shirts

Dayal

, et al. Are exposure to cytomegalovirus and genetic variation on chromosome 6p joint risk factors for schizophrenia? Ann Med. 2007;39:145-153.

64.

Shirts

Kim

Reich

, et al. Polymorphisms in MICB are associated with human herpes virus seropositivity and schizophrenia risk. Schizophr Res. 2007;94:342-353.

65.

James

Georgopoulos

Persistent antigens hypothesis: the human leukocyte antigen (HLA) connection. JNeurolNeuromed 2018;3:27-31. doi:10.29245/2572.942X/2018/6.1235

66.

James

Georgopoulos

AP.

Human leukocyte antigen as a key factor in preventing dementia and associated apolipoprotein E4 risk. Front Aging Neurosci. 2019;11:82.

67.

Charonis

James

Georgopoulos

AP.

In silico assessment of binding affinities of three dementia-protective human leukocyte antigen (HLA) alleles to nine human herpes virus antigens. Curr Res Transl Med. 2020;68:211-216. doi:10.1016/j.retram.2020.06.002

68.

James

Charonis

Georgopoulos

AP.

Association of dementia Human Leukocyte Antigen (HLA) profile with human herpes viruses 3 and 7: an

in-silico investigation. Journal Immunol Sci. 2021;5:7-14. doi:10.29245/2578-3009/2021/3.1218

69.

Global Burden of Disease Collaborative Network. Global Burden of Disease Study 2016 (GBD 2016) Results. Seattle: Institute for Health Metrics and Evaluation (IHME), 2020. Accessed July 5, 2021. http://ghdx.healthdata.org/gbd-results-tool.

70.

Population Reference Bureau. 2016 world population data sheet with a special focus on human needs and sustainable resources. Population Reference Bureau, Washington, DC, 2016. Accessed February 5, 2019. https://www.prb.org/2016−world−population−data−sheet/

71.

The Uniprot Consortium. 2019. https://www.uniprot.org/

72.

Blum

Wearsch

Cresswell

Pathways of antigen processing. Annu Rev Immunol. 2013;31:443-473.

73.

Reynisson

Alvarez

Paul

Peters

Nielsen

NetMHCpan-4.1 and NetMHCIIpan-4.0: improved predictions of MHC antigen presentation by concurrent motif deconvolution and integration of MS MHC eluted ligand data. Nucleic Acids Res. 2020;48:W449-W454. doi:10.1093/nar/gkaa379

74.

Tukey

JW.

Exploratory Data Analysis. Pearson; 1977.

75.

Cohen

Statistical Power Analysis for the Behavioral Sciences. 2nd ed. Erlbaum; 1988.

76.

Cohen

A power primer. Psychol Bull. 1992;112:155-159.

77.

Efron

Tibshirani

An Introduction to the Bootstrap. Chapman & Hall; 1993.

78.

Price

GR.

Selection and covariance. Nature. 1970;227:520-521. doi:10.1038/227520a0

79.

Frank

SA.

George Price’s contributions to evolutionary genetics. J Theor Biol. 1995;175:373-388.

80.

Markowitz

Portfolio selection. J Finance. 1952;7:77-91.

81.

Childs

Sobel

Palmer

Sinsheimer

JS.

Detection of intergenerational genetic effects with application to HLA-B matching as a risk factor for schizophrenia. Hum Hered. 2011;72:161-172.

82.

Sheehan

NJ.

The ramifications of HLA-B27. J R Soc Med. 2004;97:10-14.

83.

Sellgren

Frisell

Lichtenstein

Landen

Askling

The association between schizophrenia and rheumatoid arthritis: A nationwide population-based Swedish study on intraindividual and Familial Risks. Schiz Bull. 2014;40:1552-1559. doi:10.1093/schbul/sbu054

84.

Eaton

Byrne

Ewald

, et al. Association of schizophrenia and autoimmune diseases: linkage of Danish national registers. Am J Psychiatr. 2006;163:521-528.

85.

Benros

Mortensen

Eaton

WW.

Autoimmune diseases and infections as risk factors for schizophrenia. Ann N Y Acad Sci. 2012;1262:56-66. doi:10.1111/j.1749-6632.2012.06638.x

86.

Benros

Nielsen

Nordentoft

Eaton

Dalton

Mortensen

PB.

Autoimmune diseases and severe infections as risk factors for schizophrenia: a 30-year population-based register study. Am J Psychiatr. 2011;168:1303-1310.

87.

Benros

Pedersen

Rasmussen

Eaton

Nordentoft

Mortensen

PB.

A nationwide study on the risk of autoimmune diseases in individuals with a personal or a family history of schizophrenia and related psychosis. Am J Psychiatr. 2014;171:218-226.

88.

Chen

Chao

Chen

, et al. Prevalence of autoimmune diseases in in-patients with schizophrenia: nationwide population-based study. Br J Psychiatry. 2012;200:374-380.

89.

Ungprasert

Wijarnpreecha

Cheungpasitporn

Patients with psoriasis have a higher risk of schizophrenia: a systematic review and meta-analysis of observational studies. J Postgrad Med. 2019;65:141-145.

90.

Kirkpatrick

Miller

BJ.

Inflammation and schizophrenia. Schizophr Bull. 2013;39:1174-1179.

91.

Feigenson

Kusnecov

Silverstein

SM.

Inflammation and the two-hit hypothesis of schizophrenia. Neurosci Biobehav Rev. 2014;38:72-93.

92.

Khandaker

Cousins

Deakin

Lennox

Yolken

Jones

PB.

Inflammation and immunity in schizophrenia: implications for pathophysiology and treatment. Lancet Psychiatry. 2015;2:258-270.

93.

Al-Diwani

AAJ

Pollak

Irani

Lennox

. Psychosis: an autoimmune disease? Immunology. 2017;152:388-401. doi:10.1111/imm.12795

94.

Meehl

PE.

Schizotaxia, schizotypy, schizophrenia. Am Psychol. 1962;17:827-838.

95.

Kwapil

Barrantes-Vidal

Schizotypy: looking back and moving forward. Schizophr Bull. 2015;41 Suppl 2(Suppl 2):S366-S373. doi:10.1093/schbul/sbu186

96.

Barrantes-Vidal

Grant

Kwapil

TR.

The role of schizotypy in the study of the etiology of schizophrenia spectrum disorders. Schizophr Bull. 2015;41(Suppl 2):S408-S416. doi:10.1093/schbul/sbu191

97.

Fonseca-Pedrero

Debbané

Rodríguez-Testal

Cohen

Docherty

Ortuño-Sierra

Schizotypy: the way ahead. Psicothema. 2021;33:16-27.

98.

Mohagheghi

Eftekharian

Taheri

Alikhani

MY.

Determining the IgM and IgG antibodies titer against HSV1, HSV2 and CMV in the serum of schizophrenia patients. Hum Antibodies. 2019;26:87-93. doi:10.3233/HAB-170325

99.

Delisi

Smith

Hamovit

, et al. Herpes simplex virus, cytomegalovirus and Epstein-barr virus antibody titres in sera from schizophrenic patients. Psychol Med. 1986;16:757-763. doi:10.1017/S0033291700011764

100.

Schretlen

Vannorsdall

Winicki

, et al. Neuroanatomic and cognitive abnormalities related to herpes simplex virus type 1 in schizophrenia. Schizophr Res. 2010;118:224-231.

101.

Tucker

Bertke

AS.

Assessment of cognitive impairment in HSV-1 positive schizophrenia and bipolar patients: systematic review and meta-analysis. Schizophr Res. 2019;209:40-47. doi:10.1016/j.schres.2019.01.001

102.

Dickerson

Schroeder

Nimgaonkar

Gold

Yolken

The association between exposure to herpes simplex virus type 1 (HSV-1) and cognitive functioning in schizophrenia: A meta-analysis. Psychiatry Res. 2020;291:113157. doi:10.1016/j.psychres.2020.113157

103.

Charonis

James

Georgopoulos

In silico analysis of the binding affinities of antigenic epitopes of vaccines administered to Gulf war veterans to specific HLA Class II alleles protective for Gulf war Illness. J Neurol Neuromed. 2019;4:23-30. doi:10.29245/2572.942X/2019/5.1254

104.

Charonis

James

Georgopoulos

AP.

SARS-CoV-2 in silico binding affinity to human leukocyte antigen (HLA) Class II molecules predicts vaccine effectiveness across variants of concern (VOC). Sci Rep. 2022;12:18074-18110. doi:10.1038/s41598-022-11956-5

105.

James

Tsilibary

Charonis

Georgopoulos

Commentary: COVID-19 and the path to Immunity. Journal Immunol Sci. 2020;4:31-33. doi:10.29245/2578-3009/2020/4.1201

106.

James

Georgopoulos

AP.

At the root of 3 “long” diseases: persistent antigens inflicting chronic damage on the brain and other organs in Gulf war illness, Long-COVID-19, and chronic fatigue syndrome. Neurosci Insights. 2022;17:26331055221114817.

107.

Griffin

DE.

Why does viral RNA sometimes persist after recovery from acute infections?

PLoS Biol. 2022;20:e3001687.

108.

Yousuf

Ibrahim

Harfouche

Abu-Hijleh

Abu-Raddad

Herpes simplex virus type 1 in Europe: systematic review, meta-analyses and meta-regressions. BMJ Glob Health. 2020;5:e002388.

109.

Brainstorm Consortium. Analysis of shared heritability in common disorders of the brain. Science. 2018;360:eaap8757.

110.

Zhang

Sjölander

Ploner

Bulik

Bergen

SE.

Novel disease associations with schizophrenia genetic risk revealed in ~400,000 UK Biobank participants. Mol Psychiatry. 2022;27:1448-1454.

111.

Zheutlin

Dennis

Karlsson Linnér

, et al. Penetrance and pleiotropy of polygenic risk scores for schizophrenia in 106,160 patients across four health care systems. Am J Psychiatr. 2019;176:846-855.

112.

Oliveira

Oliveira-Maia

Tamouza

Brown

Leboyer

Infectious and immunogenetic factors in bipolar disorder. Acta Psychiatr Scand. 2017;136:409-423. doi:10.1111/acps.12791

113.

Tamouza

Oliveira

Etain

, et al. HLA genetics in bipolar disorder. Acta Psychiatr Scand. 2018;138:464-471.

114.

Rosenblat

JD.

Targeting the immune system in the treatment of bipolar disorder. Psychopharmacology. 2019;236:2909-2921.

115.

James

Harfouche

Welton

, et al. Herpes simplex virus: global infection prevalence and incidence estimates, 2016. Bull World Health Organ. 2020;98:315-329.

Schizophrenia,Human Leukocyte Antigen (HLA),and Herpes Viruses: Immunogenetic Associations at the Population Level

Abstract

Keywords

Introduction

HLA and schizophrenia

An epidemiological approach to HLA-schizophrenia associations

Human herpes viruses (HHV) and schizophrenia

Methods

Prevalence of schizophrenia

HLA alleles

SZ-HLA protection/susceptibility scores

HHV proteins

Data analysis

Schizophrenia-HLA PSCorr scores

Random permutations test for assessing the statistical significance of the schizophrenia-HLA profile

Application to individuals

Application to populations

Association of schizophrenia PSCorr scores with HHV affinities

Schizophrenia-HLA PSCov scores

Implementation of analysis procedures

Results

Permutation tests

Frequency distributions

Tests of score proportions

Tests of score magnitude

Application to individuals

Application to populations

Schizophrenia PSCorr scores and HLA-HHV binding

Combined SZ-HLA PSCorr scores versus combined HLA-HHV affinities

Discussion

Methodological considerations

General

HLA and schizophrenia prevalence

Implications for individuals

SZ-HLA PSCorr scores and in silico HHV binding affinity

The “crowdfunding” nature of HLA contribution to schizophrenia (and other diseases)

Conclusions

Limitations of the study

Footnotes

Appendix

Author Contributions

Funding:

Declaration of conflicting interests:

Data Availability

Ethical Approval

ORCID iDs

Significance Statement

References