Graphical abstract

   graphic file with name fx1.jpg
   [47]Open in a new tab

Highlights

     * •
       Possible pathogenic rare missense mutations in TENM4 gene
       contribute to SCZ
     * •
       Aberrant expression of Ten-m leads to behavioral disturbances
       related to SCZ symptoms
     * •
       Ten-m affects stimulation, metabolic process, neurogenesis, and
       ATPase activity
     __________________________________________________________________

   Genetics; Human genetics; Human Genetics; Molecular genetics;
   Neurogenetics; Psychology

Introduction

   Schizophrenia (SCZ) is a severe psychiatric disorder with high
   heritability, genetic heterogeneity, and polygenic effect ([48]Kavanagh
   et al., 2015). It has been reported that SCZ shares the same risk genes
   and pathways as common mental diseases, including bipolar disorder,
   autism, depression, Alzheimer disease, mental retardation, and epilepsy
   ([49]Sullivan et al., 2012). Previous studies have confirmed that TENM4
   is associated with bipolar disorder ([50]Psychiatric GWAS Consortium
   Bipolar Disorder Working Group, 2011) ([51]Ikeda et al., 2018)
   ([52]Ivorra et al., 2014) ([53]Heinrich et al., 2013), autism
   ([54]Iossifov et al., 2014), and essential tremor ([55]Hor et al.,
   2015) ([56]Yan et al., 2020) ([57]Chao et al., 2016). In most
   genome-wide association studies (GWAS) of SCZ, the common
   single-nucleotide polymorphisms (SNPs) in TENM4 did not reach
   genome-wide significance. However, a case-control replication study
   showed that an SNP (rs12290811) located in TENM4 was associated with
   SCZ ([58]Ivorra et al., 2014). A whole-exome sequencing study in an SCZ
   family also identified one missense mutation co-segregated with the SCZ
   in TENM4 ([59]Xue et al., 2018). A genome-wide meta-analysis with
   individuals of East Asian or European ancestry identified 176 loci
   associated with SCZ, including TENM4 ([60]Lam et al., 2019). These
   studies partly provided some genetic evidence for TENM4 as an important
   candidate gene for SCZ.

   TENM4, also known as ODZ4, is a member of the teneurin protein family,
   which comprises large transmembrane proteins ([61]Baumgartner et al.,
   1994). The TENM4 gene is highly conserved among species ([62]Tucker and
   Chiquet-Ehrismann, 2006). The temporal and spatial expression of
   teneurin family proteins in Drosophila ([63]Baumgartner et al., 1994),
   zebrafish ([64]Feng et al., 2002), birds ([65]Tucker et al., 2000), and
   mice ([66]Zhou et al., 2003) showed that Tenm4 protein was highly
   expressed in the nervous system, particularly in the brain. In
   Drosophila, a study showed that Ten-m acts as a synaptic cooperative
   matching molecule between the axons of olfactory receptor neurons
   (ORNs) and the dendrites of projecting neurons (PNs) in the olfactory
   system, and Ten-m could also guide synapses to achieve specific matches
   by homophilic interactions to ensure proper matching between neurons,
   indicating Ten-m plays an important role in neuron orientation
   ([67]Hong et al., 2012). In addition, Ten-m was required for the
   organization and target selection of synapses in neurons ([68]Mosca
   et al., 2012). In mice, a mutation in Tenm4 suppressed oligodendrocyte
   differentiation and myelination of small-diameter axons ([69]Suzuki
   et al., 2012). Knockdown expression of Tenm4 in Neuro-2a cell lines
   decreased the length of individual neurites, while overexpression of
   Tenm4 promoted cellular protrusion formation and neurite outgrowth
   ([70]Suzuki et al., 2014). These studies indicated that aberrant
   expression of Ten-m might lead to mis-matched neural circuits.

   Based on these previously reported findings, we hypothesized that TENM4
   plays an important role in SCZ development and contributes to
   SCZ-related clinical symptoms, such as learning disability, sleep
   disorders, and violent behavior. Patients with SCZ often show violence,
   social withdrawal, learning and memory declines, and sleep disorders
   ([71]Dowd and Barch, 2012; [72]Cohrs, 2008; [73]Patel et al., 2014). To
   verify this hypothesis, we performed targeted sequencing of TENM4 in 68
   SCZ families to investigate the genetic contribution of TENM4 on SCZ.
   Then, to identify the role of TENM4 in SCZ, we examined the phenotypic
   characteristics by knockdown and overexpression of Ten-m in Drosophila
   and evaluated the effects of Ten-m on learning ability, sleep behavior,
   and aggressive behaviors. We also used mRNA sequencing to reveal the
   potential affected signaling pathways and molecular mechanisms of
   TENM4.

Results

Rare missense mutations in TENM4 are co-segregated with disease status in SCZ
families

   To determine whether rare mutations in TENM4 are the candidate causal
   variants of SCZ, we initially performed whole TENM4 exome sequencing in
   351 subjects from 68 SCZ families. A total of 36 mutations in TENM4
   were identified using a step-by-step depth, quality filtering method.
   Then, 17 noncoding and synonymous variants were filtered to remove, and
   19 nonsynonymous variants were kept. The 19 variants were annotated by
   ANNOVAR software (2019Oct24) with dbnsfp35c, gnomad211_exome, avsnp150,
   and intervar_20,180,118 databases to evaluate the function and allele
   frequency in the population. Given the pattern of dominant inheritance
   in the families, only rare heterozygous variants (allele frequency<1%)
   shared by the patients but not found in the normal controls in the same
   family were considered to be candidate mutations. At last, we found 5
   variants were co-segregated with SCZ in their families, which were
   p.S1745P in Family_1, p.Q2735E and p.N965S in Family_2, p.V396I in
   Family_3, and p.T1849A in Family_4 ([74]Table 1 and [75]Figure 1A). In
   addition, all the five missense mutations were predicted to be
   deleterious to the TENM4 protein function by no less than three
   software.

Table 1.

   Genetic analysis of TENM4 in SCZ
   FAM_ID Position Allele SNP AA_Change AF No. Interpro_domain
   Family_1 78,399,126 A > G Ns p.S1745P ns 7 YD repeats
   Family_2 78,369,210 G > C rs527857070 p.Q2735E 1.605 × 10^−5 6 Tox-GHH
   domain
   78,443,605 T > C rs141471364 p.N965S 0.0053 3 Carboxypeptidase-like,
   regulatory domain
   Family_3 78,574,076 C > T rs3812723 p.V396I 0.0003 3 .
   Family_4 78,383,326 T > C rs772977333 p.T1849A 0.0002 9 YD repeats
   [76]Open in a new tab

   SNP, single-nucleotide polymorphism; AA, amino acid; AF, allele
   frequency; No. the number of software that supported the damaging
   effect of the missense mutation on TENM4 protein function.

Figure 1.

   [77]Figure 1
   [78]Open in a new tab

   Genetic analysis of TENM4 in SCZ

   (A) TENM4 rare mutations were identified in four SCZ families, which
   were p.S1745P in Family_1, p.Q2735E and p.N965S in Family_2, p.V396I in
   Family_3, and p.T1849A in Family_4.

   (B) TENM4 showed nominally significant association with SCZ according
   to the previous GWAS.

   Meanwhile, to confirm the relationship between TENM4 and SCZ, we
   re-analyzed the data from the genome-wide association analysis in a
   total of 50,874 SCZ cases and 83,493 controls from the Chinese and PGC2
   genome-wide meta-analysis ([79]Li et al., 2017). We analyzed 1196 SNPs
   in the region (chr11: 78,358,876-79,156,992) and found that 39 SNPs
   showed nominally significant association with SCZ (p < 0.01,
   [80]Figure 1B).

Ten-m affects the learning abilities in Drosophila

   To investigate the potential contribution of TENM4 to SCZ, we examined
   the function of Ten-m in the Drosophila nervous system. We used the
   courtship conditioning assay, a behavioral assay that tests the
   courtship short-term memory (STM) to indicate learning ability
   ([81]Siegel and Hall, 1979). After selection at the eclosion stage,
   male flies that had never mated were solely paired with mated females
   whose cuticles had already obtained cis-vaccenyl acetate (cVA), a
   male-specific and volatile pheromone, from males on mating
   ([82]Everaerts et al., 2010). The males would be rejected by the mated
   females; then, the males formed STM, resulting in the suppression of
   courtship behavior when placed with other mated female flies over a
   period of time ([83]Siwicki et al., 2005). The reduction of suppression
   indicates a lack of STM and the damage to learning ability. We used
   courtship indices (CIs), calculated as the percentage of courtship
   duration within 10 min, and learning indices (LIs), generated by the
   relative difference between the mean CIs of naive and trained male
   flies, to quantify the STM. Our data revealed that males of control
   genotypes (elav-GAL4, UAS-Ten-m, and UAS-Ten-m-RNAi) exhibited
   courtship suppression after training compared to naive flies.
   Otherwise, there was no significant difference in CIs between naive and
   trained experimental flies (elav > UAS-Ten-m and elav > UAS-Ten-m-RNAi)
   with Ten-m aberrant expression in the nervous system, indicating that
   pan-neuronal overexpression or downexpression of Ten-m lead to lack of
   the learning ability to suppress courtship behavior ([84]Figure 2A,
   naive vs trained; elav-GAL4, p < 0.0001; UAS-Ten-m, p < 0.0001;
   UAS-Ten-m-RNAi, p < 0.0001; elav > Ten-m, p = 0.820; elav > Ten-m-RNAi,
   p = 1.000). In addition, compared with the control groups, flies with
   aberrantly expressed Ten-m pan-neuronally revealed a significant
   reduction in LIs, which highlighted the courtship STM deficits
   ([85]Figure 2B, elav-GAL4 vs elav > Ten-m, p < 0.0001, UAS-Ten-m vs
   elav > Ten-m, p < 0.0001; elav-GAL4 vs elav > Ten-m-RNAi, p < 0.0001;
   UAS-Ten-m-RNAi vs elav > Ten-m-RNAi, p < 0.0001). These results
   suggested that aberrant expression of Ten-m in the nervous system
   impairs STM.

Figure 2.

   [86]Figure 2
   [87]Open in a new tab

   Aberrant expression of Ten-m impairs courtship-conditioning learning

   (A) CIs of control males (elav-GAL4, UAS-Ten-m, UAS-Ten-m-RNAi, n = 10)
   and experimental males (elav > Ten-m and elav > Ten-m-RNAi, n = 10).
   Box-and-whisker plots of CIs show 10^th, 25^th, 50^th, 75^th, and 90^th
   centiles and means (short dashed line). N, naive; T, trained.
   Experimental males showed no difference in CIs between naive and
   trained groups, while the CIs of control males in the trained group
   were significantly lower than in the naive groups.

   (B) LIs of control males (elav-GAL4, UAS-Ten-m, UAS-Ten-m-RNAi, n = 10)
   and experimental males (elav > Ten-m and elav > Ten-m-RNAi, n = 10).
   LIs were compared using the permutation test with 100,000 random
   permutations (one-tail, H1: LI [control] > LI [experimental males]).
   LIs of experimental males were significantly higher than those in
   control males.

   The Mann-Whitney U test was used for CIs. ∗∗∗p < 0.001; n.s., no
   significance.

Ten-m affects sleep behavior in Drosophila

   To determine whether Ten-m regulates sleep in Drosophila, we increased
   or decreased Ten-m expression levels using transgenes UAS-Ten-m and
   UAS-Ten-m-RNAi, respectively, driven by pan-neuronal elav-GAL4. Total
   and nighttime sleep amounts were decreased both in elav > Ten-m
   ([88]Figures 3A and 3B nighttime: elav-GAL4 vs elav > Ten-m, p <
   0.0001, UAS-Ten-m vs elav > Ten-m, p < 0.0001; total: elav-GAL4 vs elav
   > Ten-m, p < 0.0001; UAS-Ten-m vs elav > Ten-m, p = 0.002) and elav >
   Ten-m-RNAi files ([89]Figures 3G and 3H, nighttime: elav-GAL4 vs elav >
   Ten-m-RNAi, p = 0.012; UAS-Ten-m-RNAi vs elav > Ten-m-RNAi, p < 0.0001;
   total: elav-GAL4 vs elav > Ten-m-RNAi, p = 0.019; UAS-Ten-m-RNAi vs
   elav > Ten-m-RNAi, p = 0.012) compared to controls. In contrast, there
   was no change in mutants in the daytime sleep of the mutants
   ([90]Figures 3A and 3B, daytime: elav-GAL4 vs elav > Ten-m, p < 0.0001;
   UAS-Ten-m vs elav > Ten-m, p = 1.000; [91]Figures 3G and 3H, elav-GAL4
   vs elav > Ten-m-RNAi, p = 0.853; UAS-Ten-m-RNAi vs elav > Ten-m-RNAi, p
   = 0.986) which might be due to a floor effect, that is, daytime sleep
   of the controls was minimal. Further analysis revealed that the
   decreased night and overall sleep in elav > Ten-m flies was due to the
   reduced sleep bout duration but not sleep bout number ([92]Figure 3C,
   elav-GAL4 vs elav > Ten-m, p = 0.905, UAS-Ten-m vs elav > Ten-m, p =
   0.154; [93]Figure 3D, elav-GAL4 vs elav > Ten-m, p = 0.905, UAS-Ten-m
   vs elav > Ten-m, p = 0.121; [94]Figure 3E, p < 0.0001; [95]Figure 3F,
   elav-GAL4 vs elav > Ten-m, p = 0.005; UAS-Ten-m vs elav > Ten-m, p <
   0.0001). These results indicated that over-expressing Ten-m in all
   neurons would impair sleep behavior in Drosophila by reducing night
   sleep bout duration. The night and total sleep bout numbers were
   significantly increased in elav > Ten-m-RNAi flies compared to the
   control groups ([96]Figures 3I and 3J, p < 0.0001). Meanwhile, the
   night and total sleep bout durations of elav > Ten-m-RNAi groups were
   shorter than those of the control groups ([97]Figure 3K, elav-GAL4 vs
   elav > Ten-m-RNAi, p < 0.0001; UAS-Ten-m-RNAi vs elav > Ten-m-RNAi, p =
   0.001; [98]Figure 3L, p < 0.0001), suggesting the lack of Ten-m
   impaired sleep maintenance and caused sleep fragmentation in
   Drosophila. Overall, both Ten-m overexpression and downexpression
   impaired sleep maintenance and caused fragmented sleep.

Figure 3.

   [99]Figure 3
   [100]Open in a new tab

   Sleep phenotypes in Ten-m overexpressed and downexpressed mutants

   (A) Conventional sleep plots of controls (elav-GAL4, n = 13, UAS-Ten-m,
   n = 16) and experimental flies (elav > Ten-m, n = 9) in 12-hr LD
   conditions. White and black bars indicate 12-hr LD periods,
   respectively.

   (B) Total sleep amount (in 24 hr) and sleep during daytime and
   nighttime in control and experimental flies. Data are represented as
   mean ± SEM.

   (C and D) Histograms of the number of nighttime sleep bouts during
   nighttime (C) and histograms of the number of total sleep bouts (D) for
   controls and experimental flies. Data are represented as mean ± SEM.

   (E and F) Boxplots of the nighttime sleep bout duration (E) and
   boxplots of the total sleep bout duration (F) for controls and
   experimental flies. The line inside the box indicates the median; the
   upper and lower box limits the 75% and 25% quantiles, respectively;
   vertical lines above and below the box represent the 90% and 10%
   quantiles, respectively; points show the 95% and 5% outliers.

   (G–L) Conventional sleep plots of controls (elav-GAL4, n = 13, and
   UAS-Ten-m-RNAi, n = 14) and experimental flies (elav > Ten-m-RNAi, n =
   13) in 12-hr LD (G). Histograms of total sleep amount and sleep during
   daytime and nighttime (H), nighttime sleep bout number (I), nighttime
   sleep bout duration (K), total sleep bout number (J), and total sleep
   bout duration (L) for controls and experimental flies. Data are
   represented as mean ± SEM.

   One-way analysis of variance (ANOVA) followed by post hoc Tukey were
   used, except for the comparison of nonparametrically distributed data
   where the Kruskal-Wallis test was used. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p <
   0.001; n.s., no significance.

Ten-m affects the aggressive behavior in Drosophila

   We examined the aggressive behavior of Ten-m and Ten-m-RNAi flies using
   different GAL4 drivers. The latency to initiate fighting was
   significantly reduced ([101]Figure 4A p < 0.0001; [102]Figure 4B,
   elav-GAL4 vs elav > Ten-m-RNAi, p = 0.009; UAS-Ten-m-RNAi vs elav >
   Ten-m-RNAi, p < 0.0001), and the frequency was increased
   ([103]Figure 4A, p < 0.0001; [104]Figure 4B, elav-GAL4 vs elav >
   Ten-m-RNAi, p = 0.832; UAS-Ten-m-RNAi vs elav > Ten-m-RNAi, p = 0.001)
   in the flies with aberrantly expressed Ten-m pan-neuronally compared to
   the control groups. Male-male aggression is promoted by cVA in
   Drosophila via Or67d and Or65a receptors expressed by ORNs ([105]Wang
   and Anderson, 2010). ORNs form synapses with PNs and local interneurons
   (LNs) in the antennal lobe ([106]Das et al., 2008). Mz699-GAL4 is
   expressed in 90% of all ventral PNs, and 80% of these Mz699-PNs have
   been identified as GABAergic neurons, which are related to aggression
   ([107]Lai et al., 2008) ([108]Alekseyenko et al., 2019), and Mz19 PNs
   are inherently connected to Or67d ORNs ([109]Hong et al., 2012). We
   examined the aggressive behavior of UAS-Ten-m and UAS-Ten-m-RNAi
   transgenic flies using Mz699-GAL4, Mz19-GAL4, and LN1-GAL4, and latency
   and frequency changed sharply between the experimental and the control
   groups ([110]Figures 4A and 4B, p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001;
   n.s., no significance.). These results reveal that aberrant expression
   of Ten-m in Drosophila would lead flies to become more aggressive.

Figure 4.

   [111]Figure 4
   [112]Open in a new tab

   Ten-m modulates aggression

   (A) The latency and frequency of male flies overexpressing Ten-m in all
   neurons (elav > Ten-m, n = 14) or in specific neurons such as Mz699,
   Mz19, and LN1 (Mz699>Ten-m, n = 14, Mz19>Ten-m, n = 15, LN1>Ten-m, n =
   9) compared to the controls (elav-GAL4, n = 15, UAS-Ten-m, n = 11,
   Mz699-GAL4, n = 12, Mz19-GAL4, n = 10, LN1-GAL4, n = 10). Fighting
   latencies of experimental flies were significantly longer than those in
   controls, and fighting frequencies of experimental flies were
   significantly higher than those in controls. Data are represented as
   mean ± SEM.

   (B) The latency and frequency of male flies downexpressing Ten-m in all
   neurons (elav > Ten-m-RNAi, n = 12) or in specific neurons such as
   Mz699, Mz19, and LN1 (Mz699>Ten-m-RNAi, n = 12, Mz19>Ten-m-RNAi, n =
   10, LN1>Ten-m-RNAi, n = 13) compared to the controls (elav-GAL4, n =
   15, UAS-Ten-m-RNAi, n = 11, Mz699-GAL4, n = 12, Mz19-GAL4, n = 10,
   LN1-GAL4, n = 10). Fighting latencies of experimental flies were
   significantly longer than those of controls, and fighting frequencies
   of experimental flies were significantly higher than those of controls.
   Data are represented as mean ± SEM.

   One-way ANOVA followed by post hoc Tukey were used, except for the
   comparison of nonparametrically distributed data where the
   Kruskal-Wallis test was used. ∗∗p < 0.01, ∗∗∗p < 0.001; n.s., no
   significance.

RNA sequencing reveals the potential molecular mechanism

   Through RNA sequencing of brain tissues in Drosophila, differentially
   expressed genes (DEGs) under the conditions of Ten-m overexpression or
   downexpression were compared ([113]Figure 5A), and the functional
   classification of these DEGs was carried out. A total of 164
   significant DEGs were identified between elav > Ten-m-RNAi and
   elav-GAL4 groups, including 83 upregulated genes and 81 downregulated
   genes ([114]Figure 5B, [115]Table S3). Among DEGs in the elav >
   Ten-m-RNAi vs elav-GAL4, Gene Ontology (GO) enrichment analysis was
   related to stimulation, such as biological process (BP) terms linked to
   response to light stimulus and detection of external stimulus
   ([116]Figure 5C, [117]Table S5). We also revealed that 130 genes were
   differentially expressed in elav > Ten-m vs elav-GAL4 groups, and 85
   genes were upregulated, and 45 genes were downregulated
   ([118]Figure 5D, [119]Table S2). In the elav > Ten-m vs elav-GAL4, the
   significantly enriched GO terms were those involved in metabolic
   processes such as the nucleoside monophosphate metabolic process
   ([120]Figure 5E, [121]Table S4). As shown in the Venn diagram
   ([122]Figure 5F), we found there were 55 overlapped DEGs with both
   overexpressed and decreased expression of Ten-m. The clustering heatmap
   analysis of 55 overlapped significant DEGs indicated a contrasting gene
   expression profile between experimental groups. elav > Ten-m-RNAi and
   elav > Ten-m had a very similar profile of gene expression, revealing
   the same high and low expression trends while showing a different
   profile against elav-GAL4 ([123]Figure 5G). GO enrichment analysis of
   overlapped significant DEGs showed that the only BP term was
   neurogenesis, and the only molecular function (MF) term was ATPase
   activity ([124]Figure 5H, [125]Table S6).

Figure 5.

   [126]Figure 5
   [127]Open in a new tab

   RNA sequencing of brain tissues with aberrant expression of Ten-m in
   Drosophila

   (A) FPKM of controls (elav-GAL4) and experimental flies (elav > Ten-m
   and elav > Ten-m-RNAi). Data are represented as mean ± SD. The
   Kruskal-Wallis test was used for comparison. Compared to the control
   groups, the experimental groups had increased or decreased the
   expression of Ten-m.

   (B) Volcano plots of DEGs from elav > Ten-m-RNAi vs elav-GAL4 groups.
   Red denotes high expression while blue denotes low expression. In
   total, 164 significant DEGs were identified, including 83 upregulated
   genes and 81 downregulated genes.

   (C) GO enrichment analysis of DEGs in the elav > Ten-m-RNAi group. The
   DEGs were enriched in stimulation-associated pathways, including light
   stimulus and detection of external stimulus.

   (D) Volcano plots of DEGs from elav > Ten-m vs elav-GAL4 groups. Red
   denotes high expression while blue denotes low expression. In total,
   130 significant DEGs were identified, including 85 upregulated genes
   and 45 downregulated genes.

   (E) GO enrichment analysis of DEGs in the elav > Ten-m group. The DEGs
   were enriched in metabolism-associated pathways, such as the nucleoside
   monophosphate metabolic process.

   (F) Venn diagram showing the intersection of significant DEGs observed
   in elav > Ten-m and elav > Ten-m-RNAi groups. There were 55 overlapped
   DEGs for both decreased and overexpressed Ten-m.

   (G) Heatmap of overlapped significant DEGs. Hierarchical clustering was
   performed to highlight the DEGs, indicating a contrasting gene
   expression profile between experimental groups. Red denotes high
   expression while blue denotes low expression.

   (H) GO enrichment analysis of overlapped significant DEGs in elav >
   Ten-m and elav > Ten-m-RNAi groups. The only BP term was neurogenesis,
   and the only MF term was ATPase activity.

Discussion

   To date, hundreds of genes related to SCZ genes have been identified
   from GWAS, exome sequencing studies, transcriptomic studies, and
   epigenetic studies according to the SCZ database (SZDB,
   [128]http://www.szdb.org/SZDB/index.html) ([129]Sanders and Gill, 2007;
   [130]Khavari and Cairns, 2020; [131]Xu et al., 2011; [132]Chen et al.,
   2015; [133]Farrell et al., 2015). Previous studies have also provided
   amount of evidence for the polygenic architecture of SCZ ([134]Smeland
   et al., 2020). Many common genetic variants and rare functional
   mutations make up the polygenic architecture of SCZ ([135]Henriksen
   et al., 2017). In SCZ, common and rare genetic variants are associated
   with disorders of biological processes related to neurodevelopment,
   neuroexcitability, synaptic function, and the immune system
   ([136]Ukkola-Vuoti et al., 2019; [137]Boyajyan et al., 2015; [138]Jiang
   et al., 2020). In this study, we focused on rare functional mutations
   with minor allele frequency (MAF) < 0.01 in the candidate gene, TENM4.
   We identified five rare functional mutations of TENM4 gene in SCZ
   families, suggesting TENM4 contributes to SCZ. Interestingly, common
   SNPs in TENM4 only presented a nominally significant association with
   SCZ and did not reach the level of genome-wide significance, which
   might be because most mutations identified in TENM4 were missense
   mutations. Unlike the loss of function mutations, most missense
   mutations have a reduced effect on gene function, and some mutations
   might even not affect gene function. It is not easily to distinguish
   these mutations from the true functional missense mutations, but the
   GWAS results revealed that there might still be a genetic influence of
   TENM4 on SCZ. We have not performed the sequencing in patients with
   sporadic SCZ because the limited sample size would affect the results a
   lot, and the very big sample size may find the association between rare
   functional mutations.

   TENM4 is known to be involved in mental illness and cognition, but
   there has been no behavioral analysis of TENM4 in animal models. Thus,
   we used the Drosophila model to demonstrate the potential mechanism of
   SCZ caused by TENM4 mutations. We observed reduced STM with abnormal
   neuronal expression of Ten-m. Intellectual disability and learning
   disabilities are prevalent in patients with SCZ. SCZ is also associated
   with a substantial cognitive impairment that affects working memory,
   processing speed, attention, and verbal memory ([139]Dickinson et al.,
   2007). It is also difficult for patients with SCZ to learn from
   positive feedback ([140]Dowd and Barch, 2012). Patients with SCZ often
   reveal social withdrawal, and the possibility of violence is increased
   ([141]Lamberti, 2007). Sleep and circadian rhythm disorders occur in
   many psychiatric patients, including SCZ ([142]Wulff et al., 2010).
   Furthermore, sleep disturbances occur frequently in patients with SCZ
   ([143]Cohrs, 2008), including shorter duration, sleep fragmentation,
   insomnia, or other sleeping disorders ([144]Monti and Monti, 2004). We
   found that the aberrant expression of Ten-m could cause sleep reduction
   and increased aggressiveness, which exhibit an SCZ-like phenotype.
   Thus, we inferred that TENM4 is a promising candidate gene, which might
   be related to some SCZ symptoms in SCZ, such as high aggression levels,
   sleep disorders, and learning ability related to cognitive deficits.

   Overexpression and downexpression of Ten-m both caused night-sleep
   reduction in Drosophila, but the effect was not the same.
   Overexpression of Ten-m caused reduced sleep duration, and both
   overexpression and downexpression of Ten-m caused sleep fragmentation.
   We speculated that these effects were caused by different mechanisms,
   which need to be studied. The GWAS of actigraphic sleep phenotypes
   indicated that wake after sleep onset is associated with TENM4
   ([145]Spada et al., 2016).

   SCZ is related to severe abnormal mutations in genes essential for
   synaptic plasticity ([146]Gulsuner et al., 2020). Neural circuits of
   the human brain are composed of approximately 100 million neurons
   intercommunicated by synapses which are eventually responsible for
   behavior and brain function ([147]Li et al., 2016). Ten-m affects the
   synapses morphological structure in Drosophila, suggesting that the
   aberrant expression of TENM4 may cause changes in synapse morphological
   structure which could lead to SCZ.

   RNA sequencing showed that the effect of Ten-m on stimulation and the
   metabolic process might cause behavioral abnormalities. The sensitivity
   to cVA in males plays an important role in learning, memory, and
   aggressive behavior, while the sensory perception of light stimulation
   affects sleep in Drosophila. In addition, the imbalance of the
   nucleoside pathway and metabolism significantly affects the
   neuromodulatory function significantly ([148]Ipata et al., 2011). The
   GO enrichment analysis of overlapped significant DEGs revealed
   neurogenesis and ATPase activity. Moreover, hippocampal neurogenesis
   and synaptic functions are crucial for learning and memory ([149]Nabavi
   et al., 2014). Aberrant subcortical neurogenesis often occurs in SCZ
   ([150]Hoffmann et al., 2019). The abnormal regulation of energy
   metabolism leads to changes in the synthesis and release of ATP.
   Studies have shown that ATP is the main stimulant of microglial and can
   be used as a signal to mediate the growth of microglial and the
   communication of nerve cells, which would affect the function of
   neurons and synapses ([151]Davalos et al., 2005). Thus, the molecular
   mechanism of TENM4 on SCZ might be mainly through affecting
   neurogenesis and regulating ATPase activity.

Limitations of the study

   There are still some limitations in our study. For example, we need a
   big sample size of sporadic SCZ patients to perform the target
   sequencing to find the association between rare functional mutations.
   More functional experiments should be set to find more direct
   pathological evidence in SCZ. The molecular mechanism of TENM4 on SCZ
   remains unclear.

STAR★Methods

Key resources table

   REAGENT or RESOURCE SOURCE IDENTIFIER
   Biological samples
     __________________________________________________________________

   Human blood SJTU BIO-X psychiatric sample bank
   [152]http://www.bio-x.cn/
     __________________________________________________________________

   Deposited data
     __________________________________________________________________

   Raw data of Target sequencing This paper Mendeley Data, V2,
   [153]https://doi.org/10.17632/6w5zmn6j8h.2.
   mRNA sequencing analyzed data This paper Mendeley Data, V1,
   [154]https://doi.org/10.17632/hy3bh2z8x8.1
     __________________________________________________________________

   Experimental models: Organisms/strains
     __________________________________________________________________

   D. melanogaster: elav-GAL4 Bloomington Drosophila Stock Center
   flybase_[155]FBst0000458
   D. melanogaster: Mz19-GAL4 Bloomington Drosophila Stock Center
   flybase_FBal0155865
   D. melanogaster: Mz699-GAL4 Bloomington Drosophila Stock Center
   flybase_FBal0066073
   D. melanogaster: UAS-Ten-m Bloomington Drosophila Stock Center
   flybase_[156]FBst0041569
   D. melanogaster: UAS-Ten-m-RNAi Bloomington Drosophila Stock Center
   flybase_FBtp0052660
   D. melanogaster: w;Cyo/sco;MKRS/Tb Kyoto Stock Center
   flybase_[157]FBst0305835
   D. melanogaster: LN1-GAL4 Kyoto Stock Center flybase_[158]FBst0302813
     __________________________________________________________________

   Oligonucleotides
     __________________________________________________________________

   Primers for TENM4, see [159]supplemental information
   Such as:
   GACCCTGTTTGT GGATGTGGA This paper N/A
     __________________________________________________________________

   Software and algorithms
     __________________________________________________________________

   Burrows-Wheeler Aligner (BWA) [160]Li and Durbin, 2009
   [161]http://bio-bwa.sourceforge.net/
   Samtools [162]Li et al. (2009) [163]http://samtools.sourceforge.net/
   Genome Analysis Toolkit (GATK) [164]McKenna et al. (2010)
   [165]http://www.broadinstitute.org/gsa/wiki/index.php/The_Genome_Analys
   is_Toolkit
   ANNOVAR [166]Wang et al. (2010)
   [167]https://annovar.openbioinformatics.org/en/latest/
   IGV [168]Robinson et al. (2011) [169]http://www.igv.org/
   Trimmomatic [170]Bolger et al. (2014)
   [171]http://www.usadellab.org/cms/index.php?page=trimmomatic
   TopHat2 [172]Kim et al. (2013)
   [173]http://ccb.jhu.edu/software/tophat/manual.shtml
   Cufflinks [174]Trapnell et al. (2012)
   [175]http://cole-trapnell-lab.github.io/cufflinks/
   Bowtie2 [176]Langmead and Salzberg (2012)
   [177]http://bowtie-bio.sourceforge.net/bowtie2/index.shtml
   HTSeq [178]Anders et al. (2015)
   [179]https://htseq.readthedocs.io/en/master/
   DESeq2 [180]Love et al. (2014)
   [181]http://bioconductor.org/packages/release/bioc/html/DESeq2.html
   [182]Open in a new tab

Resource availability

Lead contact

   Further information and requests for resources and reagents should be
   directed to and will be fulfilled by the Lead Contact, Daizhan Zhou
   ([183]zhoudaizhan@sjtu.edu.cn).

Materials availability

   Materials used or generated in this study will be available upon
   reasonable request, and a material transfer agreement may be required.

Experimental model and subject details

Patients and samples

   In this study, all individuals with SCZ were interviewed by at least
   two independent psychiatrists. Diagnoses were made strictly according
   to the Diagnostic and Statistical Manual of Mental Disorders (DSM)-IV
   criteria, and patients with other neurological, mental, or psychiatric
   disorders, or with history of drug use were excluded. All participants
   provided written informed consent. Approval for our study was received
   from the Ethics Committee of Human Genetic Resources of Shanghai Jiao
   Tong University and local hospitals.

   For the family-based sequencing study, 68 SCZ families with no less
   than 3 patients each were selected, and 351 samples were collected for
   targeted sequencing. There are 183 males and 168 females, and the
   average age was 39. All blood samples were collected during 1998-2005
   in Shanghai, Shandong, Hebei, Anhui, Gansu, Heilongjiang and Guangxi
   provinces in China.

Fly lines

   The fly lines used can be found in [184]Table S1. Other genotypes were
   obtained by fly crossing. All procedures were approved by the Ethics
   Committee of Animal Research of Shanghai Jiao Tong University.

   The larvae and adults were kept on a standard sugar-yeast-cornmeal
   medium under a 12-hour light/dark (LD) cycle at 25°C with 60% humidity.
   A list of fly lines used can be found in [185]Table S1. Two genotypes
   of fly stocks were obtained by crossing other genotypes, elav>Ten-m and
   elav>Ten-m-RNAi. We used balanced lethal genotype flies:
   w;Cyo/sco;MKRS/Tb. The genotype includes four types of lethal alleles,
   which cause the death of the organism that carries them. To obtain
   elav>Ten-m flies, firstly, UAS-Ten-m flies were crossed with
   w;Cyo/sco;MKRS/Tb flies. We selected presenting Cyo and MKRS phenotypes
   of male flies from the offspring to cross with female elav;+;+ flies,
   then selected elav;Cyo/+;UAS-Ten-m/+ males. Simultaneously, female
   elav;+;+ flies were crossed with w;Cyo/sco;MKRS/Tb flies, and we
   selected presenting Cyo and MKRS phenotypes of male flies from the
   offspring to cross with female elav;+;+ flies, then selected
   elav;Cyo/+;MKRS/+ females to cross with the elav;Cyo/+;UAS-Ten-m/+
   males. Finally, we obtained elav>Ten-m flies. The same methods were
   used to obtain elav>Ten-m-RNAi flies. Other fly genotypes were obtained
   by crossing parental flies. More than five generations with the w^1118
   strain outcrossing were performed to standardize the background.

   In crossing experiment, we used 20 mature females and 10 mature males.
   In courtship conditioning assay, we used 10 males and 10 females which
   were selected at the eclosion stage. In aggression assay, we used only
   males which were kept separately for 7 days after eclosion stage. In
   sleep and activity assay, we used only virgin females which were
   collected immediately after eclosion stage.

Method details

Targeted sequencing of TENM4

   Primer3 was used to design 70 target-specific primer pairs for all
   exons and exon-intron boundaries of TENM4 ([186]Untergasser et al.,
   2012). The primers were synthesized with common adapter sequences at
   their 5’ ends as previously described ([187]Forshew et al., 2012). All
   70 primer pairs were divided into 6 multiplex primer pools. The library
   construction was performed essentially as previously described,
   including pre-amplification of the tagged gene amplicons, generation of
   a barcoded DNA library, quantification and clean-up ([188]Forshew
   et al., 2012). The final DNA libraries were determined by Qubit 2100
   (Thermo Scientific, USA) and Bioanalyzer (Illumina, USA) and then
   sequenced on the Illumina HiSeq X-Ten System (Illumina, USA) ([189]Wu
   et al., 2017).

   The targeted next-generation sequencing data were filtered by quality
   analysis and PCR primers and splices, and the paired 150 bp-long reads
   were mapped to hg19 human reference genome using the Burrows-Wheeler
   Aligner (BWA) software ([190]Li and Durbin, 2009). Samtools was used
   for sequence sorting, gene location labeling, and format transformation
   ([191]Li et al., 2009). Genome Analysis Toolkit (GATK) was used to
   identify mutation, filter and select comment to generate a VCF file
   ([192]McKenna et al., 2010) . All variants were annotated by ANNOVAR
   software (2019Oct24) ([193]Wang et al., 2010), and the potential
   effects on the protein function were predicted using 9 software (SIFT,
   PolyPhen2, LRT, MutationTaster, MutationAssessor, FATHMM, PROVEAN,
   MetaSVM, and MetaLR). All functional mutations were visually confirmed
   one-by-one using IGV software (v2.8.6) ([194]Robinson et al., 2011).

Courtship conditioning assay

   A courtship conditioning assay was used to test courtship behavior and
   learning memory in Drosophila. Males and females were selected at the
   eclosion stage and raised in Drosophila standard culture medium. Male
   flies were kept separately and females were kept in groups of 10-15
   flies. Before the experiments, selected females were pre-mated by
   housing with 15-20 w^1118 males per group for 18 hours. Males were
   divided into naive group and trained group. Male flies in trained
   groups aged up to 4-7 days after eclosion were solely paired with two
   mated females 1h for training, then the mated females were removed for
   half an hour before the test. Individual males from the control or
   trained groups were placed into the test chamber, 15 mm in diameter by
   7 mm deep, and paired with two mated females. The period of experiments
   lasted for 10 minutes and were recorded by a video camera. The
   courtship index (CI) was calculated as the percentage of male courtship
   duration within 10 minutes. The learning index (LI) was calculated
   using the mean CIs, LI = [CI [naive] – CI [trained]]/CI [naive]
   ([195]Keleman et al., 2012). As the distributions of LIs are different
   between groups, LIs were used for the permutation test with 100,000
   random permutations to compare between groups, and the data were
   processed with R software. One-tail testing against the null hypothesis
   was used in the comparison. The Mann–Whitney U test was used for CIs.

Aggression assay

   Pairs of male flies were introduced into a fighting chamber, which was
   13 mm in diameter by 7 mm in height. We used 20 mm × 20 mm glass slides
   to cover the chamber. Fly behaviors were observed and recorded for
   15 minutes ([196]Zhou et al., 2008).The aggressive behaviors of male
   flies were evaluated in two aspects: latency and frequency. Latency is
   the duration between putting experimental flies into the chambers and
   the start of the first agonistic encounter while frequency is the
   number of aggressive behaviors of male flies, such as lunges, within
   15 minutes. ([197]Chen et al., 2002).

Sleep and activity assay

   All experimental flies were virgin females collected immediately after
   eclosion stage; 15 flies of each genotype were required for parallel
   experiments. The sleep and activity of the experimental flies were
   monitored by the Drosophila Activity Monitor (DAM) system (TriKinetics,
   USA). The selected flies were raised in standard food vials until they
   were monitored, then they were moved into 65 mm × 5 mm glass tubes
   (Wuhan Yihong, China) containing solely 5% sucrose (Sangon Biotech,
   China) and 2% agar (Sangon Biotech, China). Subsequently, glass tubes
   with experimental flies were placed in the DAM system which was set in
   12-hour LD conditions for the test. LD conditions were light-on at 9 am
   and light-off at 9 pm. Data collection started after the flies were
   acclimated to the environment for 3 to 5 days([198]Song et al., 2017).
   Data records of experimental activities were binned in one-minute
   periods in LD for three days. Flies were regarded to be sleeping when
   they were inactive for at least 5 minutes. Flies were implied to have
   died when their whole period of activity was less than 10 minutes, and
   the data were excluded. Flies’ sleep duration and parameters related to
   sleep behavior were analyzed by R software. Conventional sleep plots
   were portrayed by calculating the average amount of sleep duration
   within 30 minutes of experimental flies.

RNA extraction and sequencing

   The flies were placed in 1 mL TRIzol (Invitrogen, USA), and we used
   forceps to rapidly break down the brains of the adult flies rapidly.
   Almost 20 heads of each genotype were saved in a 1.5 mL microcentrifuge
   tube with 200 μL of TRIzol. There were 3-4 replicates for each
   genotype. RNA extraction and purification were performed using TRIzol.
   The RNA was eluted with nuclease-free water, then a NanoDrop 2000
   (Thermo Scientific, USA) was used to evaluate of RNA purity and
   concentration.

   RNA-sequencing was performed on 3-4 replicates of RNA samples from each
   of elav-GAL4, elav>Ten-m-RNAi, and elav>Ten-m groups. The sample
   libraries were prepared according to the standard process of the TruSeq
   RNA Sample Prep Kit v2 (Illumina, USA). The final libraries were
   determined by Qubit 2100 (Thermo Scientific, USA) and Bioanalyzer
   (Illumina, USA) and paired-end sequencing was performed on an Illumina
   HiSeq X-Ten System (Illumina, USA), with a read length of 2 × 150 bp.

   Adapters and low-quality reads were removed from the raw data using
   Trimmomatic (v0.36) ([199]Bolger et al., 2014). TopHat (v2.1.1) was
   used to align the RNA-seq reads to a reference genome (BDGP5.25) with
   the short-read aligner Bowtie2 (v2.3.4) ([200]Kim et al., 2013)
   ([201]Langmead and Salzberg, 2012). Cufflinks (v2.2.1) was used to
   analyze the BAM files, including assembling the transcripts from the
   aligned reads and estimating their abundances using a Fragments Per
   Kilobase of transcript per Million mapped reads (FPKM)-based algorithm
   ([202]Trapnell et al., 2012). HTSeq (v0.10) was used to generate the
   read count files for each gene ([203]Anders et al., 2015). Then, DESeq2
   (v1.36.0) was used to identify DEGs, which were selected with log2
   (fold change) >1 or log2 (fold change) <-1 and with statistical
   significance (adjusted p-value <0.05) ([204]Love et al., 2014). GO and
   KEGG pathway enrichment analysis of DEGs was performed using the
   Database for Annotation, Visualization and Integrated Discovery (DAVID)
   functional annotation tool (v6.8).

Quantification and statistical analysis

   In courtship conditioning assay, the courtship index (CI) was
   calculated as the percentage of male courtship duration within
   10 minutes. The learning index (LI) was calculated using the mean CIs,
   LI = [CI naive – CI trained]/CI naive ([205]Keleman et al., 2012). As
   the distributions of LIs are different between groups, LIs were used
   for the permutation test with 100,000 random permutations to compare
   between groups, and the data were processed with R software. One-tail
   testing against the null hypothesis was used in the comparison. The
   Mann–Whitney U test was used for CIs.

   In aggression assay and sleep assay, all behavior data analyses were
   performed using Sigma Plot (v14.0) and SPSS (v22.0). For multiple
   comparisons, one-way ANOVA followed by post hoc Tukey were used, except
   for the comparison of nonparametrically distributed data where the
   Kruskal–Wallis test was used.

Acknowledgments