Polygenic contribution to the relationship of loneliness and social isolation with schizophrenia

Andreu-Bernabeu, Álvaro; Díaz-Caneja, Covadonga M.; Costas, Javier; De Hoyos, Lucía; Stella, Carol; Gurriarán, Xaquín; Alloza, Clara; Fañanás, Lourdes; Bobes, Julio; González-Pinto, Ana; Crespo-Facorro, Benedicto; Martorell, Lourdes; Vilella, Elisabet; Muntané, Gerard; Nacher, Juan; Molto, María Dolores; Aguilar, Eduardo Jesús; Parellada, Mara; Arango, Celso; González-Peñas, Javier

doi:10.1038/s41467-021-27598-6

Download PDF

Article
Open access
Published: 10 January 2022

Polygenic contribution to the relationship of loneliness and social isolation with schizophrenia

Nature Communications volume 13, Article number: 51 (2022) Cite this article

10k Accesses
10 Citations
68 Altmetric
Metrics details

Subjects

Abstract

Previous research suggests an association of loneliness and social isolation (LNL-ISO) with schizophrenia. Here, we demonstrate a LNL-ISO polygenic score contribution to schizophrenia risk in an independent case-control sample (N = 3,488). We then subset schizophrenia predisposing variation based on its effect on LNL-ISO. We find that genetic variation with concordant effects in both phenotypes shows significant SNP-based heritability enrichment, higher polygenic contribution in females, and positive covariance with mental disorders such as depression, anxiety, attention-deficit hyperactivity disorder, alcohol dependence, and autism. Conversely, genetic variation with discordant effects only contributes to schizophrenia risk in males and is negatively correlated with those disorders. Mendelian randomization analyses demonstrate a plausible bi-directional causal relationship between LNL-ISO and schizophrenia, with a greater effect of LNL-ISO liability on schizophrenia than vice versa. These results illustrate the genetic footprint of LNL-ISO on schizophrenia.

Polygenic overlap and shared genetic loci between loneliness, severe mental disorders, and cardiovascular disease risk factors suggest shared molecular mechanisms

Article Open access 05 January 2021

Loneliness and depression: bidirectional mendelian randomization analyses using data from three large genome-wide association studies

Article 21 September 2023

Polygenic risk scores for schizophrenia and major depression are associated with socio-economic indicators of adversity in two British community samples

Article Open access 14 November 2022

Introduction

Social relationships are critical for emotional and cognitive development in social species^1,2. In fact, the scientific consensus is that the need to belong to a social group is a fundamental behaviour in humans³. Researchers have characterized both objective and perceived (i.e., loneliness) social isolation^4,5. While the former is an objective lack of social connections (interactions, contacts or relationships), the latter refers to the subjective feeling of distress associated with a lack of meaningful relationships, regardless of the amount of social contact⁶. Although isolated people often feel lonely, isolation is not always correlated with feelings of loneliness^4,5,6. However, regardless of type, both objective social isolation and loneliness are major risk factors for morbidity and mortality^6,7,8, as well as for the onset of mental disorders^{9,10,11,12,13,14}.

Most psychiatric research on loneliness and objective social isolation has associated them with depressive symptoms and major depression^14,15,16, but recently researchers have shown renewed interest in their association with psychosis^17,18,19,20. Social withdrawal and isolation are described in the early stages of schizophrenia^17,21,22, recalling the classical descriptions of pre-schizophrenia related traits by Kraepelin, Bleuler, and Conrad^23,24,25. Indeed, recent meta-analyses indicate that loneliness plays an important role in the onset and maintenance of psychotic symptoms^17,22,26. Another meta-analysis also showed a consistent association of loneliness with both positive and negative psychotic-like experiences²⁷. Moreover, there are studies suggesting that loneliness may increase subclinical paranoia in non-clinical populations²⁸. However, the causal relationships between social isolation and schizophrenia are still unclear^17,29.

Inherited biological factors could explain, at least partially, the relationship between social isolation and schizophrenia. Available evidence supports the genetic basis of loneliness and objective social isolation^30,31,32,33. A recent study used multi-trait GWAS (MTAG)³⁴, a software developed to jointly analyse different summary statistics from related traits, to assess the genetic architecture of loneliness and objective social isolation (LNL-ISO)³². The researchers combined three UK Biobank GWAS datasets of (i) perceived loneliness, (ii) a proxy of social support (combined frequency of family/friends visits and living alone), and (iii) ability to confide in someone close³². Up to 15 genome-wide significant loci and SNP-based heritability estimates (h²_SNP = 4.2%) support the contribution of common genetic variation to this social construct. This study also found a significant genetic correlation of the combined phenotype (LNL-ISO) with schizophrenia (rg = 0.17, p = 3.47 × 10⁻¹²), consistent with a previous study reporting a significant association of perceived loneliness with schizophrenia, but not with bipolar disorder³³. Schizophrenia polygenic scores also significantly predicted loneliness in an independent population sample in another study³⁵, lending further support to a shared genetic aetiology between both phenotypes.

Previous studies exploring the genetic relationship between perceived and objective social isolation and schizophrenia leave several questions unanswered, including the direction of the association, the specific biological effects of shared and non-shared predisposing variants, and the effect of additional factors on this relationship, including sex. The epidemiological and clinical presentation of psychotic disorders differs between sexes^36,37,38 and sex also seems to affect the perception of loneliness and the psychological impact of isolation, although results have been contradictory so far^39,40,41.

In this work, we aim to test the hypothesis that there is a bidirectional genetic relationship between perceived and objective social isolation and schizophrenia within a systematic and comprehensive framework (see the workflow in Fig. 1). First, we analyse loneliness and social isolation (LNL-ISO) polygenic score contribution to schizophrenia risk in an independent Spanish case-control sample (CIBERSAM case-control sample). Second, we dissect the predisposing variation to schizophrenia according to its role in LNL-ISO and analyse the polygenic risk scores, biological profiles (using brain specific functional annotations), and sex effects across each genomic partition using an SNP subsetting approach. Third, to evaluate the role of LNL-ISO in the genetic overlap between psychiatric disorders and other related traits, we study the partial correlations between schizophrenia and related phenotypes across the LNL-ISO partitions. Finally, we perform a causality analysis between LNL-ISO and schizophrenia using a two-sample Mendelian randomization approach.

**Fig. 1: Workflow of the analytic pipeline.**

Results

LNL-ISO polygenic score contribution to schizophrenia risk

We calculated polygenic scores for loneliness and isolation (PGS_LNL-ISO) using the summary statistics from the combined MTAG in the UK Biobank (UKBB) study³² based on three different traits: (i) perceived loneliness, (ii) a proxy of social support (combined frequency of family/friends visits and living alone), and (iii) ability to confide in someone close. Figure 2A shows the percentage of variance in schizophrenia risk explained by LNL-ISO (PGS_LNL-ISO) in the independent CIBERSAM case-control sample (N_SCZ = 1927; N_HC = 1561). We found that common genetic variation predisposing to LNL-ISO significantly contributed to schizophrenia risk (R² (95% CI) = 0.56% (−0.01, 1.13) at P_threshold = 0.05, p = 1.2 × 10⁻⁴). One standard deviation (s.d.) increase in PGS_LNL-ISO was associated with a 15% increase in the likelihood of belonging to the schizophrenia group (OR (95%CI) = 1.15 (1.07–1.24)). In the same target sample, LNL-ISO explained more variance in schizophrenia risk than loneliness (R² (95% CI) = 0.41% (−0.08, 0.89) at P_threshold = 0.05, p = 1.42 × 10⁻³; Fig. 2A). The contribution of PGS_LNL-ISO to schizophrenia risk was also higher than that of ability to confide (R² (95% CI) = 0.28% (−0.11, 0.67) at P_threshold = 0.05, p = 7.4 × 10⁻³) and the two measures of social support included in LNL-ISO: number of people living in household (R² (95% CI) = 0.54% (−0.02, 1.11) at P_threshold = 0.01, p = 3.14 × 10⁻⁴) and frequency of family/friends visits (R² (95% CI) = 0.42% (−0.08, 0.91) at P_threshold = 1, p = 1.2 × 10⁻³; see Supplementary Data 1B).

**Fig. 2: Polygenic score contribution of LNL-ISO (PGS_LNL-ISO) and schizophrenia (PGS_SCZ) to schizophrenia risk and heritability estimates.**

Polygenic dissection of schizophrenia by its relationship with LNL-ISO

Since PGS_LNL-ISO encompassing variants with P_LNL-ISO > 0.05 did not contribute to schizophrenia risk (R² (95% CI) = 0.052% (−0.09, 0.10) at P_threshold > 0.05, p = 0.57; Supplementary Data 1), schizophrenia summary statistics were subsetted according to their role in LNL-ISO GWAS (Supplementary Methods 4). Firstly, those variants not associated with LNL-ISO (SCZ[noLNL]; P_LNL-ISO > 0.05) were extracted. Second, variants associated with LNL-ISO (SCZ[LNL]; P_LNL-ISO < 0.05) were divided into those with a concordant sign of the allele effect in both schizophrenia and LNL-ISO (SCZ[CONC]; Beta_SCZ > 0 & Beta_LNL-ISO > 0 / Beta_SCZ < 0 & Beta_LNL-ISO < 0) and those with a discordant direction of the effect relative to schizophrenia (SCZ[DISC]; Beta_SCZ > 0 & Beta_LNL-ISO < 0/Beta_SCZ < 0 & Beta_LNL-ISO > 0; Fig. 1).

We performed PGS_scz predictions on the same schizophrenia case-control sample for the three subsets of SNPs based on the dissection of SCZ summary data according to the role in LNL-ISO: PGS_scz predictions from variants only contributing to SCZ (PGS_SCZ[noLNL]) and those contributing to both phenotypes with concordant (PGS_SCZ[CONC]) and discordant (PGS_SCZ[DISC]) effects (see Methods). Figure 2B shows the percentage of variance in schizophrenia risk explained by PGS_scz within each subset of SNPs. PGS_SCZ[CONC] explained almost four times more variance (R² = 3.94% at P_threshold = 0.5, p = 8.36 × 10⁻²⁵) than PGS_SCZ[DISC] (R² = 1.02% at P_threshold = 0.01, p = 8.43 × 10⁻⁸). PGS_SCZ comparisons across ranked deciles were also performed (Fig. 2C, Supplementary Data 2). Higher PGS_SCZ was found to be associated with SCZ risk across all described partitions (Fig. 2C)

Heritability estimates by LD-score regression (LDSR) found that variation within SCZ[CONC] showed a significant SNP-based heritability (h²_SNP) enrichment, with 3.8% of the SNPs explaining an estimated 13.1% of the h²_SNP (Enrichment(CI_95%) = 3.43 (2.86–4.01); p = 1.83 × 10⁻¹⁵; Fig. 2D; Supplementary Data 3). We found no significant heritability enrichment for SCZ[DISC] (Enrichment(CI_95%) = 1.08 (0.58–1.58); p = 0.748). Enrichment comparison of the same number of variants from SCZ[CONC] and SCZ[DISC] reflected a clear superior enrichment of concordant variants (Supplementary Fig. 2). By contrast, variants within SCZ[noLNL] harboured 65% of the SNPs and accounted for around 53.9% of the heritability, with a relative h²_SNP decrease for this annotation (Enrichment(CI_95%) = 0.81 (0.72–0.90); p = 8.12 × 10⁻⁵; Fig. 2D).

We applied partitioned heritability and LD-score regression analyses of specifically expressed genes (LDSC-SEG) within the described annotations. We observed comparable heritability enrichment profiles for SCZ[noLNL] and SCZ[CONC] across the central nervous system (CNS) and the neuronal cell type (Supplementary Data 3). The 13 brain tissues analysed displayed distinct enrichment patterns. Schizophrenia predisposing variation within SCZ[noLNL] was specifically enriched in GTEx brain cortex (p = 8.5 × 10⁻⁴) and anterior cingulate cortex (p = 5.16 × 10⁻³; Supplementary Fig. 3), while predisposing variation within SCZ[CONC] was enriched in GTEx hippocampus (p = 0.041), although the latter was not significant after FDR correction.

We assessed PGS_SCZ contribution to schizophrenia risk stratified by sex in the CIBERSAM case-control sample based on variation within SCZ[noLNL], SCZ[LNL], SCZ[CONC], and SCZ[DISC] (Supplementary Data 4). PGS_SCZ[CONC] explained significantly more variance in schizophrenia risk in females (R² (95% CI) = 2.24% (1.09, 3.38) at P_threshold = 0.1, p = 1.88 × 10⁻¹³) than in males (R² (95% CI) = 1.41% (0.60, 2.22) at P_threshold = 0.5, p = 2 × 10⁻¹³), while the opposite pattern was observed for the rest of the partitions (Supplementary Data 4). We statistically confirmed these sex-based differences using a bootstrap resampling approach comparing prediction in males and females for each genomic partition (Fig. 3, Supplementary Fig. 4).

**Fig. 3: Density plot for sex comparison of PGS_SCZ contributions to schizophrenia risk.**

Annotation-stratified genetic covariance between SCZ and related phenotypes based on LNL-ISO

We assessed covariance between predisposing genetic variation to schizophrenia and a series of neuropsychiatric disorders and related phenotypes across SCZ[noLNL], SCZ[CONC], and SCZ[DISC] using GNOVA⁴².

The majority of the disorders (MDD, ANX, ADHD, ASD, CROSS-DIS, ALC-DEP) and personality traits (NEUR, SWB, DS, PSY_EXP) tested here showed positive genetic correlation within SCZ[CONC] and negative genetic correlation within SCZ[DISC] (Fig. 4). However, BIP and OCD showed positive covariances within both genomic annotations. Therefore, alleles that increase the risk for SCZ but decrease the risk for LNL-ISO (SCZ[DISC]) are positively correlated with BIP or OCD, but negatively correlated with MDD, ASD, ADHD or ANX, providing one distinction between these two groups in their relationship with LNL-ISO. As expected, estimated correlations within SCZ[noLNL] were similar to those previously described for schizophrenia across the whole genome⁴³ (Fig. 4, Supplementary Data 5).

**Fig. 4: Annotation-stratified genetic covariance between schizophrenia and related traits.**

Mendelian randomization

We finally assessed the direction of causation between social isolation (LNL-ISO) and schizophrenia using a range of bidirectional Mendelian randomization methods (Inverse Variance Weighted (IVW)⁴⁴, Weighted Median (WM)⁴⁵, MR-Egger⁴⁶, Simple Mode (SM)⁴⁷, Weighted Mode (WM)⁴⁷, MR-PRESSO⁴⁸ and CAUSE⁴⁹). We used multiple tests to rule out horizontal pleiotropy (Table 1 and Supplementary Data 7).

Table 1 Bidirectional causal inference analyses between loneliness and isolation phenotypes and schizophrenia.

Full size table

We found evidence for a strong bidirectional causal effect of LNL-ISO on schizophrenia (IVW−β (standard error (SE)) = 1.11(0.48), p = 0.021), (WM−β (standard error (SE)) = 1.37 (0.40), p = 6.14 × 10⁻⁴) with a consistent direction of the effects across methods except in the case of MR-Egger. Although we did not detect horizontal pleiotropy with the MR-Egger intercept analysis (p = 0.36), there was evidence of heterogeneity (IVW Q-p-value = 2.94 × 10⁻⁶) (Table 1, Supplementary Fig. 5). In this scenario, the WM method, which is more robust in the presence of outliers, was preferred over the IVW method^45,50. The presence of heterogeneity also provided the most suitable explanation for the difference in the direction of Egger´s effect due to the sensitivity of this method to the presence of outliers and heterogeneity, which lead to poor causal estimates in such situations⁴⁶ (see Table 1 and Supplementary Methods 7).

Additional robust methods that eliminate outliers that may be influencing the outcome due to pleiotropy (MR-PRESSO⁴⁸) or account for both correlated and uncorrelated pleiotropy (CAUSE⁴⁹) showed comparable results to those using the WM method, with even larger effect sizes using MR-PRESSO (MR-PRESSO outlier-correction β (Sd) = 1.45(0.30), p = 0.001; CAUSE−γ (CI95%) = 0.61 (0.34, 0.89), p = 0.003) (Table 1).

We also found a causal effect of schizophrenia liability on LNL-ISO (WM−β (SE) = 0.015(0.005), p = 0.008; CAUSE−γ (CI95%) = 0.01 (0.01, 0.01), p = 0.003), with evidence of heterogeneity (IVW Q-p-value: 2.21 × 10⁻¹¹) but no indication of horizontal pleiotropy based on the MR-Egger intercept analysis (p = 0.48).

In the MR analyses including the constituent phenotypes of LNL-ISO, we found comparable evidence for bidirectional causality between perceived loneliness and schizophrenia to that found for LNL-ISO (Table 1, Supplementary Figs. 5 and 6). We also found a unidirectional negative causal effect of ability to confide on schizophrenia (WM−β (SE) = −0.6 (0.19), p = 0.002), and a unidirectional negative causal effect of schizophrenia on number of people in household (WM−β (SE) = −0.011 (0.003), p = 5.86 × 10⁻³). We found no evidence of causality between the number of family/friends visits and schizophrenia.

Discussion

This work suggests the presence of genetic overlap between social isolation, measured using LNL-ISO, and schizophrenia, with a bidirectional causal relationship. We found that overlapping predisposing genetic variation with concordant effects in both phenotypes shows significant SNP-based heritability enrichment, supporting the relatively enhanced contribution of this set of variants to schizophrenia liability. We found the concordant variation to contribute more to schizophrenia risk in females and to be positively correlated with other neuropsychiatric traits. Conversely, discordant variation contributed to schizophrenia risk only in males and was negatively correlated with most neuropsychiatric traits. These results reveal the likely genomic footprint of social isolation on the heritability of schizophrenia and provide new insights about their relationship^32,35. They also support the role of LNL-ISO as a critical social trait for understanding the heterogeneity of pleiotropic genetic effects between schizophrenia and other psychiatric disorders and behavioural traits. In fact, each of the individual traits included in the composite LNL-ISO phenotype had a significant polygenic score contribution to schizophrenia risk. These results agree with separate findings of a clinical overlap between schizophrenia and both perceived loneliness and objective social disconnection and support the idea that social isolation may play a significant role in the aetiology of psychotic disorders^17,19,20,27.

Researchers have described polygenic score predictions and LD-score-based partition heritability estimates as powerful methods for evaluating the effects of genetic predisposing variation within specific subsets of variants^51,52,53. With 3.8% of SNPs explaining an estimated 13.1% of SNP-based heritability, concordant overlapping variation between both phenotypes exhibits more than a three-fold increase in heritability enrichment compared to variants not predisposing to LNL-ISO, and a much higher enrichment than most of the genome-wide annotations previously evaluated in schizophrenia⁵⁴. LDSC-SEG analyses pointed to a significant enrichment at the uncorrected level of concordant overlapping variation in the hippocampus, a brain region involved in social behaviour^55,56 and cognitive flexibility⁵⁷, which may be especially sensitive to brain inflammation caused by loneliness and isolation^10,56. In this respect, recent work has described loneliness affecting the white matter integrity of the hippocampus⁵⁸.

Despite reported sex differences in the epidemiology and clinical manifestations of psychotic disorders^36,38,59, previous studies had not found an effect of sex on genetic associations⁶⁰. By analysing the genomic overlap between schizophrenia and LNL-ISO, we did observe a differential effect of sex on polygenic contributions to schizophrenia risk. Concordant overlapping variants in SCZ and LNL-ISO accounted for a significantly greater amount of variance in schizophrenia risk in females than in males, while the opposite pattern was observed in the rest of LNL-ISO based annotations. These results are in line with recent studies suggesting a potentially higher impact of loneliness and objective social isolation on psychiatric outcomes in females than in males^26,41. This may be due to a more negative perception of social deprivation in females related to their role in modern society⁶¹ and a greater protective effect of an enriched social network in males⁶². Moreover, among patients with schizophrenia, loneliness has been described to be more prevalent in females than males⁶³. Our results suggest the existence of a social-related environment differentially affecting males and females that could be, at least in part, responsible for the different sex-stratified PGS contributions. Further studies should evaluate the impact of sex and gender differences in subjective social perception in epidemiological models.

Genetic correlations have been shown to be a very useful method for understanding shared genetic architecture and the interrelationship between disorders and related traits, despite some limitations^43,64,65,66. By evaluating annotation-stratified correlations, previous studies have described subtle structures in shared genetics between complex traits^42,67,68. In our study, we have described the impact of the genetic liability to LNL-ISO in the relationship between schizophrenia and most of the tested neuropsychiatric disorders (ASD, MDD, ANX, ADHD, ALC-DEP) and other related behavioural traits (SWB, NEUR, PSY-EXP, EA). In the majority of the disorders, schizophrenia is positively correlated within concordant overlapping variation and negatively correlated within discordant overlapping variation with LNL-ISO, thus pointing to a shared genetic impact of social isolation on comorbidity with these disorders. However, OCD and BIP have positively correlated with schizophrenia regardless of LNL-ISO based annotations, thus suggesting that the association of these disorders with schizophrenia is independent of the genetic predisposition to LNL-ISO. These results are in line with recent findings suggesting that schizophrenia, BIP, and OCD could belong to the same psychopathology factor at the genomic level⁶⁹.

The genetic relationship between schizophrenia with EA and other cognitive-related measures such as intelligence test performance has been widely studied^68,70,71. Assessing annotation-stratified genetic covariance between EA and schizophrenia, we described a negative covariance within concordant overlapping variation, while EA showed a positive correlation with schizophrenia across discordant overlapping variation and with variants only associated with schizophrenia. Our findings suggest that poor educational attainment often found in young patients with schizophrenia^72,73 could be mediated by social isolation.

Mendelian randomization analyses provided evidence of the bidirectional nature of the causal relationship between loneliness and isolation and schizophrenia liability, with greater size of the effect of LNL-ISO on schizophrenia risk than in the opposite direction. This finding of bidirectional causality between social isolation and schizophrenia was confirmed with the recently developed method CAUSE, which provides better control for correlated and uncorrelated horizontal pleiotropy⁴⁹. Our results are consistent with previous evidence suggesting that loneliness and objective social isolation could trigger both positive and negative psychotic symptoms in clinical and non-clinical populations^17,27. It could also explain the high levels of loneliness and isolation before the onset of psychosis in individuals at clinical high risk for psychosis⁷⁴. On the other hand, the described effect of schizophrenia liability on social isolation could also give an explanation to the high prevalence of loneliness in the chronic stages of psychotic illnesses^17,20,26.

Causal inferences assessing the relationships between LNL-ISO constituents and schizophrenia also found a unidirectional negative causal effect of “ability to confide” on schizophrenia, in line with recent studies describing the association of lack of confidence and loneliness with psychosis, which may be mediated by negative schemata of others^29,75. Moreover, a unidirectional negative causal effect of schizophrenia liability on the “number of people living in your household” was found, thus suggesting a possible indirect causal effect of schizophrenia genetic liability on subsequent social disconnection in participants diagnosed with schizophrenia¹⁸. This relationship is also reinforced with the significant polygenic contribution of both phenotypes to schizophrenia risk (Fig. 2A).

Previous studies assessing social determinants of poor mental health have evaluated the association of social disadvantage and their genetic determinants with the risk of psychosis^76,77. Our study adds to this previous evidence by incorporating a subjective perception to social dysfunction in psychosis from a genetic perspective. Further studies should explore the effect of subjective perception of loneliness and its association with the social defeat hypothesis with the risk of psychosis⁷⁶.

Our study was subject to several limitations. First, we used measures of loneliness and objective social isolation from the UKBB, which are based on single-question questionnaires and not on validated scales such as UCLA loneliness⁷⁸. Nevertheless, multiple research studies have previously validated binary self-reported loneliness questionnaires and found strong convergent validity with UCLA loneliness scale^1,58,78. Second, since we used discovery samples for polygenic score analysis from the UKBB population, socio-economic biases could have affected our genetic predictions to some extent^79,80. Third, partitioning the genome in order to estimate heritability enrichment in a reduced subset of SNP may have underpowered some of our analyses. Larger sample sizes in future studies could address this limitation. Fourth, we found a great degree of heterogeneity in the MR analyses. However, we implemented several complementary methods to support the robustness of our findings and report only on results that held up across all methods. Other methods for genomic dissection such as Genomic SEM⁸¹ could be used in future studies to strengthen the results presented here. Finally, the small effect sizes suggest that even if genetic variation may partially underpin the link between schizophrenia and LNL-ISO phenotypes, environmental variables are likely to play a substantial role in this association and should be explored in future epidemiological studies.

In summary, our results shed additional light on the relationship between social isolation and schizophrenia from a genetic perspective, and lend further support for the potential role of LNL-ISO in the onset and maintenance of schizophrenia and other psychotic disorders⁸². We also provide new insights into the influence of social isolation on comorbidity with other mental disorders and its interplay with behavioural traits. Given that social isolation and perceived loneliness may be modifiable, they could be targets for effective preventive interventions with a potentially substantial impact on mental health.

Methods

Samples and GWAS summary data

We used a case-control sample including 1927 schizophrenia cases (65% males) and 1,561 healthy controls (HC) (55% males) from CIBERSAM (Centro de Investigación Biomédica en Red en Salud Mental, Spain) as an independent target sample for polygenic score predictions (SCZ_CIBERSAM). All participants were genotyped as part of the Psychiatric Genomics Consortium (PGC), and passed quality control (QC filters) per PGC-SZ2 criteria⁸³. See Supplementary Methods for a detailed description. Informed consent signed by each participating subject or legal guardian and approval from the corresponding Research Ethics Committee were obtained before starting the study.

We used the following genetic summary statistics from the previous GWAS: (i) schizophrenia GWAS from the Psychiatric Genetic Consortium (PGC-SCZ2) comprising 35,476 cases and 46,839 controls⁸³, (ii)LNL-ISO combined phenotype (LNL-ISO)³² GWAS based on the combined multi-trait GWAS (MTAG) in the UKBB study, yielding an effective maximum sample size of 487,647 individuals, and (iii) the latest UKBB GWAS results for the independent loneliness and isolation traits that were included in the original LNL-ISO MTAG: (a) loneliness UKBB, (b) a proxy of social support, as measured by the frequency of family and friend visits and the number of people living in household, and (c) ability to confide in someone close to you. There is no overlap between PGC-SCZ2 and SCZ_CIBERSAM samples. Another recent schizophrenia GWAS⁸⁴, including approximately 5000 new cases and 18,000 controls to PGC2, was also used to rule out changes in risk predictions or heritability estimates compared to PGC-SCZ2 GWAS (Supplementary Data 2).

Dissection of schizophrenia summary genetic data based on LNL-ISO

First, we selected variants that were included in both schizophrenia and LNL-ISO summary data. Second, we divided schizophrenia summary statistics from the set of overlapping variants (N_SNPs = 5,658,282) into two different subsets of variants, according to their effects in the LNL-ISO (Fig. 1): those variants not associated with LNL-ISO (SCZ[noLNL]; P_LNL-ISO > 0.05; N_SNPs = 5,172,017) and those variants associated with LNL-ISO (SCZ[LNL]; P_LNL-ISO < 0.05; N_SNPs = 486,265). We selected this cutoff based on the fact that LNL-ISO-based PGS (PGS_LNL-ISO) prediction on schizophrenia in the case-control target sample from CIBERSAM at P_LNL-ISO > 0.05 was not significant (R² (CI95%) = 0.052% (−0,09,0,19) at P_threshold > 0.05, P = 0.569; Supplementary Data 1C). Third, based on the concordance or discordance of the effects of the same effect allele, we again divided SCZ[LNL] into those variants with concordant (SCZ[CONC]; N_SNPs = 269,361) or discordant (SCZ[DISC]; N_SNPs = 216,904) effects in schizophrenia and LNL-ISO (SCZ[CONC]: Beta_SCZ > 0 & Beta_LNL-ISO > 0 OR Beta_SCZ < 0 & Beta_LNL-ISO < 0; SCZ[DISC]: Beta_SCZ > 0 & Beta_LNL-ISO < 0 OR Beta_SCZ < 0 & Beta_LNL-ISO > 0). In each of the final datasets, we removed correlated SNPs due to linkage disequilibrium (LD) using PLINK 1.9 clumping algorithm (r² > 0.1; window size = 500 kb). See Supplementary Methods 4 for further details.

Polygenic score (PGS) models

We performed polygenic models based on PGC-SZ2⁸³ (PGS_SCZ) and LNL-ISO³² (PGS_LNL-ISO) GWAS summary statistics as the discovery samples, and SCZ_CIBERSAM case-control sample as the target sample (N_SCZ = 1927; N_HC = 1561). Several P thresholds were used (P < 5 × 10⁻⁸, 5 × 10⁻⁵, 1 × 10⁻³, 0.01, 0.05, 0.1, 0.2, 0.5 and 1). Genetic variants within the Major Histocompatibility Complex (MHC) were removed. We calculated standardized PGS and evaluated significance by logistic regression, using case-control status as dependent variable and sex, age, and ten first multidimensional scaling (MDS) ancestry components as covariates. Explained variance attributable to PGS was calculated as the increase in Nagelkerke’s pseudo-R² between a model with and without the PGS variable. In PGS predictions with PGC-SZ2, Nagelkerke’s pseudo-R² were converted to liability scale following the procedure proposed by Lee et al.⁸⁵ assuming a prevalence of schizophrenia in the general population ~1%⁸⁶. We applied a correction for multiple testing to all p-values. CI for the increase in R² was estimated through bootstrap resampling (N = 5000 permutations), applying the Normal Interval method, after checking the normality of the bootstrap distribution. In order to compare PGS_LNL-ISO predictions with LNL-ISO’s constituent phenotypes (loneliness UKBB, frequency of family and friend visits, number of people living in household, and ability to confide), PGS contributions of these phenotypes to schizophrenia risk were also evaluated in the same target sample.

Using the described separated subsets of variants based on their effect in LNL-ISO, we also calculated PGS_SCZ. LD-independent variants within SCZ[noLNL] (N_clumped SNPs = 169,574), SCZ[LNL] (N_clumped SNPs = 11,804), SCZ[CONC] (N_clumped SNPs = 6468) and SCZ[DISC] (N_clumped SNPs = 5,336) were used to calculate PGS on the SCZ_CIBERSAM case-control sample. We calculated standardized PGS and evaluated significance with logistic regression models as described above.

In order to assess the effect of sex on these models, we compared the explained variance in the case-control status for predisposing variation to schizophrenia within SCZ[noLNL], SCZ[LNL], SCZ[CONC] and SCZ[DISC] in females and males. Then, after bootstrap resampling (5000 permutations) of 500 schizophrenia and 500 HC subjects in each sex separately (see Supplementary Methods 3), we statistically compared the differences between the distribution of liability R² in males and females across each genomic partition with two-sided t-tests. Since no sex differences have been reported in schizophrenia overall prevalence^87,88 we considered a prevalence estimate of 1% for both sexes. We also conducted a sensitivity analysis using recent prevalence estimates in the Spanish population⁸⁹ (prevalence in males = 0.0079 and females = 0.0045), with comparable findings (Supplementary Fig. 4).

In order to understand the direction of the effect of the PGS across the different partitions (higher or lower values in SCZ patients compared to healthy controls), PGS_SCZ comparisons across ranked deciles were also performed. The target sample was first separated into ten deciles of increasing PGS. The P-threshold with the lowest p-value was selected for each partition. The phenotype values of each decile were compared to those of the reference decile (the median (5th decile) was used as a reference) one by one, with decile status as a predictor of target phenotype (5th decile was coded 0 and tested decile 1) in a logistic regression model. OR values for each comparison were estimated from regression coefficients of these decile-status predictors. Sex, age, and ten first MDS ancestry components were used as covariates.

The term “prediction” is used in relation to polygenic score models to conform to standard terminology in the field. However, these models are not used with a predictive purpose.

LD-score regression (LDSR) and heritability estimates

We calculated SNP-based heritability (h²_SNP) estimates for resulting genome partitions from dissections of schizophrenia summary genetic data based on LNL-ISO as described before: (i) SCZ[noLNL], SCZ[CONC], and SCZ[DISC] annotations; and (ii) sub-annotations from the intersection between those annotations (SCZ[noLNL] and SCZ[CONC]) and gene expression data from ten whole tissues⁶⁴, 13 brain-related tissues (Brain GTEx⁵¹), and 3 brain cell-type annotation files (neurons, astrocytes, and oligodendrocytes^51,90). The intersection with SCZ[DISC] was not included due to the low h²_SNP for this annotation. ldsc v1.0.1⁵¹, a command-line tool for estimating heritability, was used. We performed both heritability enrichment analyses across the described annotations (-h2) and one-sided t-tests to evaluate whether the cell-type enrichment in schizophrenia within a particular LNL-ISO annotation was higher than the same cell-type enrichment in schizophrenia outside the LNL-ISO annotation (-h2-cts) (see Supplementary Methods 5). Additional information on the whole procedure is described in Supplementary Methods 5.

Annotation-stratified genetic covariance

To examine the influence of LNL-ISO based annotations (SCZ[noLNL], SCZ[CONC], and SCZ[DISC]) on the correlation between schizophrenia and other related disorders or traits, we calculated partial correlations using GNOVA⁴² (https://github.com/xtonyjiang/GNOVA). First, we selected neuropsychiatric and related phenotypes reportedly showing significant correlations with schizophrenia and/or loneliness/social isolation phenotypes in previous studies (i.e., major depression (MDD), attention and deficit hyperactivity disorder (ADHD), autism spectrum disorder (ASD), anxiety disorder (ANX), bipolar disorder (BIP), obsessive-compulsive disorder (OCD), alcohol dependence disorder (ALC-DEP), cross-disorder phenotype (CROSS-DIS)—based on a meta-analysis across eight mental disorders with a total sample of 232,964 cases and 494,162 controls— neuroticism (NEUR), depressive symptoms (DS), subjective well-being (SWB), psychotic experiences in the general population (PSY-EXP), and educational attainment (EA)). We also included body-mass index (BMI) summary data since some researchers report that this phenotype may be influenced by LNL-ISO³² (see Supplementary Methods 6 for references). We used covariance estimates based on partial correlations restricted to SNP subsets within each annotation conducted with GNOVA⁴². Derived p-values were statistically corrected using a Benjamini–Hochberg False Discovery Rate (FDR) procedure (p_FDR < 0.05). See Supplementary Methods 6 for further details.

Two-sample Mendelian randomization

We used Mendelian Randomization (MR) to investigate the direction of the causal relationships between social isolation (LNL-ISO) and its constituents (i.e., loneliness UKBB, frequency of family visits, number of people in household, and ability to confide) with schizophrenia liability using the latest GWAS data available in MRC-IEU API resource^91,92 (https://gwas.mrcieu.ac.uk; https://mrcieu.github.io/ieugwasr/).

We selected genome-wide significant SNPs at p < 5 × 10⁻⁸ except in the case of a number of people in household due to an insufficient number of instrumental variables (IV) at this threshold. We used a p < 5 × 10⁻⁶ instead. We applied a default LD-Clumping r² threshold of 0.001 and a window of 10,000 kb. Five MR methods (i) Inverse variance-weighted method (IVW)⁴⁴, (ii) WM⁴⁵, (iii) MR-Egger⁴⁶, (iv) SM, and (v) WMo⁴⁷ were performed using the R package TwoSampleMR v.0.5.3⁹² (https://github.com/mrcieu/TwoSampleMR). We conducted sensitivity tests including heterogeneity tests (IVW and Egger Cochran′s Q statistic test)⁴⁶, leave-one-out, and pleiotropy tests using functions of the same R package. Additionally, we conducted Mendelian Randomization Pleiotropy RESidual Sum and Outlier (MR-PRESSO)⁴⁸ analyses and a novel MR latent-model method (CAUSE)⁴⁹ to further account for pleiotropy (https://jean997.github.io/cause/).

We applied a correction for multiple testing using Benjamini–Hochberg FDR (p_FDR < 0.05). See Supplementary Methods 7 for further details.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability

GWAS summary datasets used in this study have been downloaded from the following repositories: UK BIOBANK. https://www.nealelab.is/uk-biobank. MRBASE—IEU GWAS PROJECT. https://gwas.mrcieu.ac.uk/. PGC. https://www.med.unc.edu/pgc/download-results/. SSGAC. https://thessgac.com/. Summary statistics for the LNL-ISO composite (LNL-ISO) were downloaded from the following repository: (https://doi.org/10.17863/CAM.23511). Individual genotype data for the CIBERSAM consortium samples (Spain, https://www.cibersam.es/en) used here and analytic code is available from the corresponding author upon reasonable request, since the deposit of this data to a public repository is not allowed due to ethical and legal requirements at the participating centres. Functional and cell-type annotation files (Neuron, astrocytes and oligodendrocytes) for heritability analyses were downloaded from the public available LDSC repository (http://data.broadinstitute.org/alkesgroup/LDSCORE/).

References

Hughes, M. E., Waite, L. J., Hawkley, L. C. & Cacioppo, J. T. A short scale for measuring loneliness in large surveys: results from two population-based studies. Res. Aging 26, 655–672 (2004).
Article PubMed PubMed Central Google Scholar
Shankar, A., Hamer, M., McMunn, A. & Steptoe, A. Social isolation and loneliness: relationships with cognitive function during 4 years of follow-up in the English longitudinal study of ageing. Psychosom. Med. 75, 161–170 (2013).
Article PubMed Google Scholar
Baumeister, R. F. & Leary, M. R. The need to belong: desire for interpersonal attachments as a fundamental human motivation. Psychol. Bull. 117, 497–529 (1995).
Article CAS PubMed Google Scholar
Perissinotto, C. M. & Covinsky, K. E. Living alone, socially isolated or lonely—what are we measuring? J. Gen. Intern. Med 29, 1429–1431 (2014).
Article PubMed PubMed Central Google Scholar
Cacioppo, J. T., Hawkley, L. C., Norman, G. J. & Berntson, G. G. Social isolation. Ann. N. Y. Acad. Sci. 1231, 17–22 (2011).
Article ADS PubMed PubMed Central Google Scholar
Holt-Lunstad, J., Smith, T. B., Baker, M., Harris, T. & Stephenson, D. Loneliness and social isolation as risk factors for mortality: a meta-analytic review. Perspect. Psychol. Sci. 10, 227–237 (2015).
Article PubMed Google Scholar
Cacioppo, J. T., Cacioppo, S., Capitanio, J. P. & Cole, S. W. The neuroendocrinology of social isolation. Annu. Rev. Psychol. 66, 733–767 (2015).
Article PubMed Google Scholar
Steptoe, A., Shankar, A., Demakakos, P. & Wardle, J. Social isolation, loneliness, and all-cause mortality in older men and women. PNAS 110, 5797–5801 (2013).
Article ADS CAS PubMed PubMed Central Google Scholar
Beutel, M. E. et al. Loneliness in the general population: prevalence, determinants and relations to mental health. BMC Psychiatry 17, 97 (2017).
Article PubMed PubMed Central Google Scholar
Hawkley, L. C. & Cacioppo, J. T. Loneliness matters: a theoretical and empirical review of consequences and mechanisms. Ann. Behav. Med. 40, 218–227 (2010).
Article PubMed Google Scholar
Heinrich, L. M. & Gullone, E. The clinical significance of loneliness: a literature review. Clin. Psychol. Rev. 26, 695–718 (2006).
Article PubMed Google Scholar
Matthews, T. et al. Lonely young adults in modern Britain: findings from an epidemiological cohort study. Psychol. Med 49, 268–277 (2019).
Article PubMed Google Scholar
Meltzer, H. et al. Feelings of loneliness among adults with mental disorder. Soc. Psychiatry Psychiatr. Epidemiol. 48, 5–13 (2013).
Article PubMed Google Scholar
Peerenboom, L., Collard, R. M., Naarding, P. & Comijs, H. C. The association between depression and emotional and social loneliness in older persons and the influence of social support, cognitive functioning and personality: a cross-sectional study. J. Affect. Disord. 182, 26–31 (2015).
Article CAS PubMed Google Scholar
Adams, K. B., Sanders, S. & Auth, E. A. Loneliness and depression in independent living retirement communities: risk and resilience factors. Aging Ment. Health 8, 475–485 (2004).
Article CAS PubMed Google Scholar
Cacioppo, J. T., Hughes, M. E., Waite, L. J., Hawkley, L. C. & Thisted, R. A. Loneliness as a specific risk factor for depressive symptoms: cross-sectional and longitudinal analyses. Psychol. Aging 21, 140–151 (2006).
Article PubMed Google Scholar
Michalska da Rocha, B., Rhodes, S., Vasilopoulou, E. & Hutton, P. Loneliness in psychosis: a meta-analytical review. Schizophrenia Bull. 44, 114–125 (2018).
Article Google Scholar
Green, M. F. et al. Social disconnection in schizophrenia and the general community. Schizophrenia Bull. 44, 242–249 (2018).
Article Google Scholar
Lim, M. H., Gleeson, J. F. M., Alvarez-Jimenez, M. & Penn, D. L. Loneliness in psychosis: a systematic review. Soc. Psychiatry Psychiatr. Epidemiol. 53, 221–238 (2018).
Article PubMed Google Scholar
Eglit, G. M. L., Palmer, B. W., Martin, A. S., Tu, X. & Jeste, D. V. Loneliness in schizophrenia: construct clarification, measurement, and clinical relevance. PLoS One 13, e0194021 (2018).
Article PubMed PubMed Central Google Scholar
van der Werf, M., van Winkel, R., van Boxtel, M. & van Os, J. Evidence that the impact of hearing impairment on psychosis risk is moderated by the level of complexity of the social environment. Schizophr. Res. 122, 193–198 (2010).
Article PubMed Google Scholar
Bucci, P. et al. Premorbid academic and social functioning in patients with schizophrenia and its associations with negative symptoms and cognition. Acta Psychiatr. Scand. 138, 253–266 (2018).
Article CAS PubMed Google Scholar
Kraepelin, E., Robertson, G. M. & Barclay, R. M. Dementia Praecox and Paraphrenia (Chicago Medical Book Co., 1919).
Mishlove, M. & Chapman, L. J. Social anhedonia in the prediction of psychosis proneness. J. Abnorm. Psychol. 94, 384–396 (1985).
Article CAS PubMed Google Scholar
Conrad, K. Die Beginnende Schizophrenie; Versuch Einer Gestaltanalyse Des Wahns (Thieme, 1958).
Solmi, M. et al. Factors associated with loneliness: an umbrella review of observational studies. J. Affect. Disord. 271, 131–138 (2020).
Article PubMed Google Scholar
Chau, A. K. C., Zhu, C. & So, S. H.-W. Loneliness and the psychosis continuum: a meta-analysis on positive psychotic experiences and a meta-analysis on negative psychotic experiences. Int. Rev. Psychiatry 31, 471–490 (2019).
Article PubMed Google Scholar
Gollwitzer, A., Wilczynska, M. & Jaya, E. S. Targeting the link between loneliness and paranoia via an interventionist-causal model framework. Psychiatry Res. 263, 101–107 (2018).
Article PubMed Google Scholar
Lamster, F., Lincoln, T. M., Nittel, C. M., Rief, W. & Mehl, S. The lonely road to paranoia. a path-analytic investigation of loneliness and paranoia. Compr. Psychiatry 74, 35–43 (2017).
Article PubMed Google Scholar
Gao, J. et al. Genome-wide association study of loneliness demonstrates a role for common variation. Neuropsychopharmacol 42, 811–821 (2017).
Article CAS Google Scholar
Matthews, T. et al. Social isolation, loneliness and depression in young adulthood: a behavioural genetic analysis. Soc. Psychiatry Psychiatr. Epidemiol. 51, 339–348 (2016).
Article PubMed PubMed Central Google Scholar
Day, F. R., Ong, K. K. & Perry, J. R. B. Elucidating the genetic basis of social interaction and isolation. Nat. Commun. 9, 2457 (2018).
Article ADS PubMed PubMed Central Google Scholar
Abdellaoui, A. et al. Phenome-wide investigation of health outcomes associated with genetic predisposition to loneliness. Hum. Mol. Genet. https://doi.org/10.1101/468835 (2018).
Turley, P. et al. Multi-trait analysis of genome-wide association summary statistics using MTAG. Nat. Genet. 50, 229–237 (2018).
Article CAS PubMed PubMed Central Google Scholar
Abdellaoui, A. et al. Predicting loneliness with polygenic scores of social, psychological and psychiatric traits. Genes, Brain Behav. 17, e12472 (2018).
Article CAS Google Scholar
Riecher-Rössler, A., Butler, S. & Kulkarni, J. Sex and gender differences in schizophrenic psychoses—a critical review. Arch. Women’s Ment. Health https://doi.org/10.1007/s00737-018-0847-9 (2018).
Szymanski, S. et al. Gender differences in onset of illness, treatment response, course, and biologic indexes in first-episode schizophrenic patients. Am. J. Psychiatry 152, 698–703 (1995).
Article CAS PubMed Google Scholar
Ochoa, S., Usall, J., Cobo, J., Labad, X. & Kulkarni, J. Gender differences in schizophrenia and first-episode psychosis: a comprehensive literature review. Schizophr. Res. Treatment. https://doi.org/10.1155/2012/916198 (2012).
Barreto, M. et al. Personality and Individual Differences Vol. 169 (Elsevier, 2020).
Galderisi, S., Bucci, P., Üçok, A. & Peuskens, J. No gender differences in social outcome in patients suffering from schizophrenia. Eur. Psychiatry 27, 406–408 (2012).
Article CAS PubMed Google Scholar
Liu, H., Zhang, M., Yang, Q. & Yu, B. Gender differences in the influence of social isolation and loneliness on depressive symptoms in college students: a longitudinal study. Soc. Psychiatry Psychiatr. Epidemiol. 55, 251–257 (2020).
Article PubMed Google Scholar
Lu, Q. et al. A powerful approach to estimating annotation-stratified genetic covariance via GWAS summary statistics. Am. J. Hum. Genet 101, 939–964 (2017).
Article CAS PubMed PubMed Central Google Scholar
Bulik-Sullivan, B. et al. An atlas of genetic correlations across human diseases and traits. Nat. Genet. 47, 1236–1241 (2015).
Article CAS PubMed PubMed Central Google Scholar
Davey Smith, G. & Hemani, G. Mendelian randomization: genetic anchors for causal inference in epidemiological studies. Hum. Mol. Genet. 23, R89–R98 (2014).
Article CAS PubMed PubMed Central Google Scholar
Bowden, J., Davey Smith, G., Haycock, P. C. & Burgess, S. Consistent estimation in Mendelian randomization with some invalid instruments using a weighted median estimator. Genet. Epidemiol. 40, 304–314 (2016).
Article PubMed PubMed Central Google Scholar
Bowden, J., Davey Smith, G. & Burgess, S. Mendelian randomization with invalid instruments: effect estimation and bias detection through Egger regression. Int J. Epidemiol. 44, 512–525 (2015).
Article PubMed PubMed Central Google Scholar
Hartwig, F. P., Davey Smith, G. & Bowden, J. Robust inference in summary data Mendelian randomization via the zero modal pleiotropy assumption. Int J. Epidemiol. 46, 1985–1998 (2017).
Article PubMed PubMed Central Google Scholar
Verbanck, M., Chen, C.-Y., Neale, B. & Do, R. Detection of widespread horizontal pleiotropy in causal relationships inferred from Mendelian randomization between complex traits and diseases. Nat. Genet. 50, 693–698 (2018).
Article CAS PubMed PubMed Central Google Scholar
Morrison, J., Knoblauch, N., Marcus, J. H., Stephens, M. & He, X. Mendelian randomization accounting for correlated and uncorrelated pleiotropic effects using genome-wide summary statistics. Nat. Genet. 52, 740–747 (2020).
Article CAS PubMed PubMed Central Google Scholar
Burgess, S. et al. Guidelines for performing Mendelian randomization investigations. Wellcome Open Res. 4, 186 (2019).
Article PubMed Google Scholar
Finucane, H. K. et al. Heritability enrichment of specifically expressed genes identifies disease-relevant tissues and cell types. Nat. Genet 50, 621–629 (2018).
Article CAS PubMed PubMed Central Google Scholar
Wray, N. R. et al. Research review: polygenic methods and their application to psychiatric traits. J. Child Psychol. Psychiatry 55, 1068–1087 (2014).
Article PubMed Google Scholar
Rammos, A., Gonzalez, L. A. N., Weinberger, D. R., Mitchell, K. J. & Nicodemus, K. K. The role of polygenic risk score gene-set analysis in the context of the omnigenic model of schizophrenia. Neuropsychopharmacology 44, 1562–1569 (2019).
Article CAS PubMed PubMed Central Google Scholar
Finucane, H. K. et al. Partitioning heritability by functional annotation using genome-wide association summary statistics. Nat. Genet. 47, 1228–1235 (2015).
Article CAS PubMed PubMed Central Google Scholar
Montagrin, A., Saiote, C. & Schiller, D. The social hippocampus. Hippocampus 28, 672–679 (2018).
Article PubMed Google Scholar
Biggio, F. et al. Social enrichment reverses the isolation-induced deficits of neuronal plasticity in the hippocampus of male rats. Neuropharmacology 151, 45–54 (2019).
Article CAS PubMed Google Scholar
Anacker, C. & Hen, R. Adult hippocampal neurogenesis and cognitive flexibility—linking memory and mood. Nat. Rev. Neurosci. 18, 335–346 (2017).
Article CAS PubMed PubMed Central Google Scholar
Spreng, R. N. et al. The default network of the human brain is associated with perceived social isolation. Nat. Commun. 11, 6393 (2020).
Article ADS CAS PubMed PubMed Central Google Scholar
Häfner, H. et al. The epidemiology of early schizophrenia: influence of age and gender on onset and early course. Br. J. Psychiatry 164, 29–38 (1994).
Article Google Scholar
Consortium, S. W. G. of the P. G., Ripke, S., Walters, J. T. & O’Donovan, M. C. Mapping genomic loci prioritises genes and implicates synaptic biology in schizophrenia. Preprint at medRxiv https://doi.org/10.1101/2020.09.12.20192922 (2020).
Benenson, J. F. et al. Social exclusion: more important to human females than males. PLoS One 8, e55851 (2013).
Article ADS CAS PubMed PubMed Central Google Scholar
Stokes, J. & Levin, I. Gender differences in predicting loneliness from social network characteristics. J. Pers. Soc. Psychol. 51, 1069–1074 (1986).
Article CAS PubMed Google Scholar
Badcock, J. C. et al. Loneliness in psychotic disorders and its association with cognitive function and symptom profile. Schizophrenia Res. 169, 268–273 (2015).
Article Google Scholar
Bulik-Sullivan, B. K. et al. LD Score regression distinguishes confounding from polygenicity in genome-wide association studies. Nat. Genet. 47, 291–295 (2015).
Article CAS PubMed PubMed Central Google Scholar
Lee, J. J., McGue, M., Iacono, W. G. & Chow, C. C. The accuracy of LD Score regression as an estimator of confounding and genetic correlations in genome-wide association studies. Genet Epidemiol. 42, 783–795 (2018).
Article PubMed PubMed Central Google Scholar
Ni, G., Moser, G., Wray, N. R. & Lee, S. H. Estimation of genetic correlation via linkage disequilibrium score regression and genomic restricted maximum likelihood. Am. J. Hum. Genet 102, 1185–1194 (2018).
Article CAS PubMed PubMed Central Google Scholar
Zhang, Y. et al. Local genetic correlation analysis reveals heterogeneous etiologic sharing of complex traits. Genome Biol. https://doi.org/10.1101/2020.05.08.084475 (2020).
Lam, M. et al. Pleiotropic meta-analysis of cognition, education, and schizophrenia differentiates roles of early neurodevelopmental and adult synaptic pathways. Am. J. Hum. Genet 105, 334–350 (2019).
Article CAS PubMed PubMed Central Google Scholar
Waldman, I. D., Poore, H. E., Luningham, J. M. & Yang, J. Testing structural models of psychopathology at the genomic level. World Psychiatry 19, 350–359 (2020).
Article PubMed PubMed Central Google Scholar
Escott-Price, V. et al. Genetic liability to schizophrenia is negatively associated with educational attainment in UK Biobank. Mol. Psychiatry 25, 703–705 (2020).
Article PubMed Google Scholar
Le Hellard, S. et al. Identification of gene loci that overlap between schizophrenia and educational attainment. Schizophr. Bull. 43, 654–664 (2017).
PubMed Google Scholar
Hakulinen, C. et al. The association between early-onset schizophrenia with employment, income, education, and cohabitation status: nationwide study with 35 years of follow-up. Soc. Psychiatry Psychiatr. Epidemiol. 54, 1343–1351 (2019).
Article PubMed Google Scholar
Dickson, H. et al. Academic achievement and schizophrenia: a systematic meta-analysis. Psychological Med. 50, 1949–1965 (2020).
Article Google Scholar
Robustelli, B. L., Newberry, R. E., Whisman, M. A. & Mittal, V. A. Social relationships in young adults at ultra high risk for psychosis. Psychiatry Res 247, 345–351 (2017).
Article PubMed Google Scholar
Müller, H. et al. Negative schemata about the self and others and paranoid ideation in at-risk states and those with persisting positive symptoms. Early Inter. Psychiatry 12, 1157–1165 (2018).
Article Google Scholar
Selten, J.-P., van der Ven, E., Rutten, B. P. F. & Cantor-Graae, E. The social defeat hypothesis of schizophrenia: an update. Schizophr. Bull. 39, 1180–1186 (2013).
Article PubMed PubMed Central Google Scholar
Allardyce, J. & Boydell, J. Review: the wider social environment and schizophrenia. Schizophr. Bull. 32, 592–598 (2006).
Article PubMed PubMed Central Google Scholar
Russell, D. W. UCLA loneliness scale (Version 3): reliability, validity, and factor structure. J. Personal. Assess. 66, 20–40 (1996).
Article CAS Google Scholar
Fry, A. et al. Comparison of sociodemographic and health-related characteristics of UK Biobank participants with those of the general population. Am. J. Epidemiol. 186, 1026–1034 (2017).
Article PubMed PubMed Central Google Scholar
Haworth, S. et al. Apparent latent structure within the UK Biobank sample has implications for epidemiological analysis. Nat. Commun. 10, 333 (2019).
Article ADS PubMed PubMed Central Google Scholar
Grotzinger, A. D. et al. Genomic structural equation modelling provides insights into the multivariate genetic architecture of complex traits. Nat. Hum. Behav. 3, 513–525 (2019).
Article PubMed PubMed Central Google Scholar
Bennett, J. C., Surkan, P. J., Moulton, L. H., Fombonne, E. & Melchior, M. Childhood social isolation and psychotic experiences in young adulthood: a community based study. Eur. Child Adolesc. Psychiatry 29, 1003–1010 (2020).
Article PubMed Google Scholar
Schizophrenia Working Group of the Psychiatric Genomics Consortium. Biological insights from 108 schizophrenia-associated genetic loci. Nature 511, 421–427 (2014).
Article ADS PubMed Central Google Scholar
Pardiñas, A. F. et al. Common schizophrenia alleles are enriched in mutation-intolerant genes and in regions under strong background selection. Nat. Genet 50, 381–389 (2018).
Article PubMed PubMed Central Google Scholar
Lee, S. H., Goddard, M. E., Wray, N. R. & Visscher, P. M. A better coefficient of determination for genetic profile analysis. Genet Epidemiol. 36, 214–224 (2012).
Article PubMed Google Scholar
McGrath, J., Saha, S., Chant, D. & Welham, J. Schizophrenia: a concise overview of incidence, prevalence, and mortality. Epidemiol. Rev. 30, 67–76 (2008).
Article PubMed Google Scholar
Charlson, F. J. et al. Global epidemiology and burden of schizophrenia: findings from the global burden of disease study 2016. Schizophr. Bull. 44, 1195–1203 (2018).
Article PubMed PubMed Central Google Scholar
Perälä, J. et al. Lifetime prevalence of psychotic and bipolar I disorders in a general population. Arch. Gen. Psychiatry 64, 19–28 (2007).
Article PubMed Google Scholar
Orrico-Sánchez, A., López-Lacort, M., Muñoz-Quiles, C., Sanfélix-Gimeno, G. & Díez-Domingo, J. Epidemiology of schizophrenia and its management over 8-years period using real-world data in Spain. BMC Psychiatry 20, 149 (2020).
Article PubMed PubMed Central Google Scholar
Cahoy, J. D. et al. A transcriptome database for astrocytes, neurons, and oligodendrocytes: a new resource for understanding brain development and function. J. Neurosci. 28, 264–278 (2008).
Article CAS PubMed PubMed Central Google Scholar
Ruth Mitchell, E. MRC IEU UK Biobank GWAS pipeline version 2. data.bris https://doi.org/10.5523/bris.pnoat8cxo0u52p6ynfaekeigi (2019).
Hemani, G. et al. The MR-Base platform supports systematic causal inference across the human phenome. eLife 7, e34408 (2018).
Article PubMed PubMed Central Google Scholar

Download references

Acknowledgements

This work was supported by the Spanish Ministry of Science and Innovation. Instituto de Salud Carlos III (SAM16PE07CP1, PI16/02012, PI17/00997, PI19/01024, PI20/00721), co-financed by ERDF Funds from the European Commission, “A way of making Europe”, CIBERSAM. Madrid Regional Government (B2017/BMD-3740 AGES-CM-2), European Union Structural Funds. European Union Seventh Framework Program under grant agreements FP7-4-HEALTH-2009-2.2.1-2-241909 (Project EU-GEI), FP7- HEALTH-2013-2.2.1-2-603196 (Project PSYSCAN), and FP7- HEALTH-2013-2.2.1-2-602478 (Project METSY); and European Union H2020 Program under the Innovative Medicines Initiative 2 Joint Undertaking (grant agreement No 115916, Project PRISM, and grant agreement No 777394, Project AIMS-2-TRIALS), Fundación Familia Alonso, Fundación Alicia Koplowitz, and Fundación Mutua Madrileña. We are grateful to Paloma Pantoja for her assistance in the design of the figures. A.A.-B. holds a Rio Hortega Grant from Instituto de Salud Carlos III (CM20/00114). C.M.D.-C. holds a Juan Rodés Grant from Instituto de Salud Carlos III (JR19/00024).

Author information

These authors contributed equally: Álvaro Andreu-Bernabeu and Javier González-Peñas.

Authors and Affiliations

Department of Child and Adolescent Psychiatry, Institute of Psychiatry and Mental Health, Hospital General Universitario Gregorio Marañón, Madrid, Spain
Álvaro Andreu-Bernabeu, Covadonga M. Díaz-Caneja, Lucía De Hoyos, Carol Stella, Xaquín Gurriarán, Clara Alloza, Mara Parellada, Celso Arango & Javier González-Peñas
Instituto de Investigación Sanitaria Gregorio Marañón (IiSGM), Madrid, Spain
Álvaro Andreu-Bernabeu, Covadonga M. Díaz-Caneja, Lucía De Hoyos, Carol Stella, Xaquín Gurriarán, Clara Alloza, Mara Parellada, Celso Arango & Javier González-Peñas
CIBERSAM, Centro Investigación Biomédica en Red Salud Mental, Madrid, Spain
Covadonga M. Díaz-Caneja, Lourdes Fañanás, Julio Bobes, Ana González-Pinto, Benedicto Crespo-Facorro, Lourdes Martorell, Elisabet Vilella, Gerard Muntané, Juan Nacher, María Dolores Molto, Eduardo Jesús Aguilar, Mara Parellada, Celso Arango & Javier González-Peñas
School of Medicine, Universidad Complutense, Madrid, Spain
Covadonga M. Díaz-Caneja, Mara Parellada & Celso Arango
Instituto de Investigación Sanitaria (IDIS) de Santiago de Compostela, Complexo Hospitalario Universitario de Santiago de Compostela (CHUS), Servizo Galego de Saúde (SERGAS), Santiago de Compostela, Galicia, Spain
Javier Costas & Xaquín Gurriarán
Department of Evolutionary Biology, Ecology and Environmental Sciences, Faculty of Biology, University of Barcelona, Barcelona, Spain
Lourdes Fañanás
Faculty of Medicine and Health Sciences−Psychiatry, Universidad de Oviedo, ISPA, INEUROPA, Oviedo, Spain
Julio Bobes
BIOARABA Health Research Institute, OSI Araba, University Hospital, University of the Basque Country, Vitoria, Spain
Ana González-Pinto
Hospital Universitario Virgen del Rocío, Department of Psychiatry, Universidad de Sevilla, Sevilla, Spain
Benedicto Crespo-Facorro
Hospital Universitari Institut Pere Mata, IISPV, Universitat Rovira i Virgili, Reus, Spain
Lourdes Martorell, Elisabet Vilella & Gerard Muntané
Neurobiology Unit, Department of Cell Biology, Interdisciplinary Research Structure for Biotechnology and Biomedicine (BIOTECMED), University of Valencia, Valencia, 46100, Spain
Juan Nacher
Department of Genetics, University of Valencia, Campus of Burjassot, Valencia, Spain
María Dolores Molto
Department of Medicine, Universitat de València, Valencia, Spain
María Dolores Molto & Eduardo Jesús Aguilar
Fundación Investigación Hospital Clínico de Valencia, INCLIVA, 46010, Valencia, Spain
Eduardo Jesús Aguilar

Authors

Álvaro Andreu-Bernabeu
View author publications
You can also search for this author in PubMed Google Scholar
Covadonga M. Díaz-Caneja
View author publications
You can also search for this author in PubMed Google Scholar
Javier Costas
View author publications
You can also search for this author in PubMed Google Scholar
Lucía De Hoyos
View author publications
You can also search for this author in PubMed Google Scholar
Carol Stella
View author publications
You can also search for this author in PubMed Google Scholar
Xaquín Gurriarán
View author publications
You can also search for this author in PubMed Google Scholar
Clara Alloza
View author publications
You can also search for this author in PubMed Google Scholar
Lourdes Fañanás
View author publications
You can also search for this author in PubMed Google Scholar
Julio Bobes
View author publications
You can also search for this author in PubMed Google Scholar
Ana González-Pinto
View author publications
You can also search for this author in PubMed Google Scholar
Benedicto Crespo-Facorro
View author publications
You can also search for this author in PubMed Google Scholar
Lourdes Martorell
View author publications
You can also search for this author in PubMed Google Scholar
Elisabet Vilella
View author publications
You can also search for this author in PubMed Google Scholar
Gerard Muntané
View author publications
You can also search for this author in PubMed Google Scholar
Juan Nacher
View author publications
You can also search for this author in PubMed Google Scholar
María Dolores Molto
View author publications
You can also search for this author in PubMed Google Scholar
Eduardo Jesús Aguilar
View author publications
You can also search for this author in PubMed Google Scholar
Mara Parellada
View author publications
You can also search for this author in PubMed Google Scholar
Celso Arango
View author publications
You can also search for this author in PubMed Google Scholar
Javier González-Peñas
View author publications
You can also search for this author in PubMed Google Scholar

Contributions

J.G.P., A.A.-B., J.C., and C.M.D.-C. contributed to the conception and design of the study. J.G.P. designed the analytic pipeline. J.G.P. and A.A.-B. were involved in data analysis and writing the paper. C.M.D.-C., J.C., G.M., L.d.H., B.C.-F., E.V., and C.A. edited the paper. The rest of the authors (X.G., C.A., L.F., J.B., A.G.P., L.M., J.N., M.D.M., E.J.A., and M.P.) were involved in patient recruitment and data collection. All the authors have reviewed and approved the final version of the manuscript.

Corresponding author

Correspondence to Javier González-Peñas.

Ethics declarations

Competing interests

C.A has been a consultant to or has received honoraria or grants from Acadia, Angelini, Gedeon Richter, Janssen Cilag, Lundbeck, Minerva, Otsuka, Roche, Sage, Servier, Shire, Schering Plough, Sumitomo Dainippon Pharma, Sunovion, and Takeda. B.C.-F. has received honoraria (advisory board and educational lectures) and travel expenses from Takeda, Menarini, Angelini, Teva, Otsuka, Lundbeck, and Johnson & Johnson. He has also received unrestricted research grants from Lundbeck. Xaquín Gurriarán has received a grant from Fundación Instituto Roche. C.M.D.-C. has received honoraria from AbbVie, Sanofi, and Exeltis. The rest of the authors declare no competing interests.

Peer review information

Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work. Peer reviewer reports are available.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary Information

Peer Review File

Description of Additional Supplementary Files

Supplementary Data 1

Supplementary Data 2

Supplementary Data 3

Supplementary Data 4

Supplementary Data 5

Supplementary Data 6

Supplementary Data 7

Reporting Summary

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions

About this article

Cite this article

Andreu-Bernabeu, Á., Díaz-Caneja, C.M., Costas, J. et al. Polygenic contribution to the relationship of loneliness and social isolation with schizophrenia. Nat Commun 13, 51 (2022). https://doi.org/10.1038/s41467-021-27598-6

Download citation

Received: 18 December 2020
Accepted: 26 November 2021
Published: 10 January 2022
DOI: https://doi.org/10.1038/s41467-021-27598-6

This article is cited by

Estimating the direct effects of the genetic liabilities to bipolar disorder, schizophrenia, and behavioral traits on suicide attempt using a multivariable Mendelian randomization approach
- Brenda Cabrera-Mendoza
- Necla Aydin
- Renato Polimanti
Neuropsychopharmacology (2024)
Identifying novel proteins underlying loneliness by integrating GWAS summary data with human brain proteomes
- Xiaojing Gu
- Meng Dou
- Wei Zhang
Neuropsychopharmacology (2023)
Loneliness and depression: bidirectional mendelian randomization analyses using data from three large genome-wide association studies
- David A. Sbarra
- Ferris A. Ramadan
- Yann C. Klimentidis
Molecular Psychiatry (2023)

Comments

By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.