LRFN5 locus structure is in uenced by the individual’s sex and associated with autism

1 Background: LRFN5 is a brain-specific gene needed for synaptic development and 2 plasticity. It is the only gene in a large 5.4 Mb topologically associating domain 3 (TAD) on chromosome 14, which we term the LRFN5 locus. This locus is highly 4 conserved, but has extensive copy number variation. 5 Methods: Locus structure was studied by chromatin immunoprecipitation (chIP-on6 chip) in fibroblasts from individuals with autism and controls, supplemented with a 7 capture-HiC determination of TAD structures in a family trio. LRFN5 expression was 8 studied in foetal brain cell cultures. In addition, locus interaction was studied in four 9 large and independent cohorts by measuring deviations from Hardy-Weinberg 10 equilibrium of a common deletion polymorphism. 11 Results: We found that locus structural changes are associated with developmental 12 delay (DD) and autism spectrum disorders (ASD). In a large family, ASD in males 13 segregated with a chromosome 14 haplotype carrying a 172 kb deletion upstream of 14 LRFN5. In a fibroblast capture-HiC study on an ASD-patient-parent trio, the ASD15 susceptible haplotype (in the mother and her autistic son) had a TAD pattern different 16 from both the father and a female control. When the trimethylated histone-3-lysine-9 17 chromatin (H3K9me3) profiles in fibroblasts from control males (n=6) and females 18 (n=7) were compared, a male-female difference was observed around the LRFN5 gene 19 itself (p<0.01). Intriguingly, in three cohorts of individuals with DD (n=8757), the 20 number of heterozygotes of a common deletion polymorphism upstream of LRFN5 21 was 20-26% lower than expected from Hardy-Weinberg equilibrium. This indicates 22 early allelic interaction, and the genomic conversions from heterozygosity to wild23 type or deletion homozygosity were of equal magnitudes. In a control group of 24 medical students (n=1416), such conversions were three times more common than in 25 the DD-patient cohorts (p=0.00001). Hypothetically, such allelic interaction is needed 1 to establish monoallelic expression, which we found in foetal brain cell cultures. 2 Limitations: The male-female difference in H3K9me3 profiles was based on 3 fibroblast data from a small number of individuals, and the monoallelic expression 4 data on a single experiment. 5 Conclusions: Taken together, allelic interaction, monoallelic expression and sex6 dependent differences make the LRFN5 locus attractive for exploring the genetic basis 7 of synaptic memory and high-functioning male autism. 8

schizophrenia to an overlapping set of low-risk variants. The LRFN5 locus has only 1 been identified in a family-based GWAS study from 2009, 9 a finding not replicated in 2 more recent population-based or meta-GWAS analyses. 10,11 There are, however, other 3 results that link the LRFN5 locus to autism. A copy number variation (CNV) study of 4 regions of chromosome 2 (168,500-178,500 Mb), chromosome 13 (94,000-113,000 1 Mb), chromosome 14 (16,475-70,975 Mb), and chromosome 17 (41,195 Mb) 2 uniformly, with a median probe spacing of ~150 bp. DNA labelling, array 3 hybridization, post-hybridization washes and scanning were performed according to 4 the manufacturer's protocol: "NimbleGen Arrays User´s Guide, ChIP-chip Array", 5 v6.2 (Roche NimbleGen). In short, the chIP and input DNA (DNA from non-6 precipitated chromatin) samples were labelled with Cy5-and Cy3-conjugated random 7 nonamers, respectively. The labelled samples were purified, combined, denatured and 8 hybridized to the array for 16 hours at 42°C. After stringent washing, the array was 9 scanned using an Axon 4200AL Scanner (Molecular Devices, CA, USA) at 5-μm 10 resolution. The acquired images were analysed by DEVA v1.2 software (Roche 11 NimbleGen) creating pair reports, including raw intensities for each probe and per 12 image. From these data, ratio files were generated. For data visualization, the average 13 ratios of two replicate experiments were binned per kb, each adjusted for the number 14 of probes per bin. These data were transferred to in Excel spreadsheets for further 15 calculations and generation of plain text files in .bedgraph format, and then the data 16 was plotted against chromosomal position using the UCSC browser's custom track 17 option. 18 Capture HiC-based LRFN5 locus TAD-structure determination 19 20 LRFN5 locus selection: The capture Hi-C (CHiC) SureSelect library was designed 21 over the genomic interval (chr14:539,000,000-47,000,000, GRCh37) using the 22 SureDesign tool from Agilent (Agilent Technologies, Santa Clara, CA). The coverage 23 was 70,5% by 159698 probes of total size 4,668 Mb and 5x tiling density. 24 Fixation of fibroblast nuclei: Capture HiC experiments were performed on dermal 25 fibroblasts from a family trio (parents and child with ASD) and a control female of the same age as the mother. Trypsinised fibroblasts were washed in PBS and then 1 transferred to a 50-ml Falcon tube and complemented with 10% FCS/PBS. 37% 2 formaldehyde was added to a final concentration of 2% and cells were fixed for 3 10 min at room temperature. Crosslinking was quenched by adding glycine (final 4 concentration; 125 mM). Fixed cells were washed twice with cold PBS and lysed 5 using fresh lysis buffer (10 mM Tris, pH 7.5, 10 mM NaCl, 5 mM MgCl2, 0.1 mM 6 EGTA with protease inhibitor) to isolate nuclei. Cell lysis was assessed 7 microscopically after 10-min incubation in ice. Nuclei were centrifuged for 5 min at 8 480g, washed once with PBS and snap frozen in liquid N2. 9 Chromosome conformation capture library preparation and sequencing: 10 3C libraries were prepared from fixed nuclei as described previously. 22 Briefly, lysis 11 buffer was removed by centrifugation at 400 g for 5 min at 4 °C, followed by 12 supernatant aspiration, snap-freezing, and pellet storage at -80 °C. Later, nuclei pellets 13 were thawed on ice, resuspended in 520 μl 1× DpnII buffer, and then incubated with 14 7.4 μl 20% SDS shaking at 900 rpm. at 37 °C for 1 hour. Next, 75 μl 20% Triton X-15 100 was added and the pellet was left shaking at 900 rpm. at 37°C for 1 hour. A 15-μl 16 aliquot was taken as a control for undigested chromatin (stored at -20°C). The 17 chromatin was digested using 40 μl 10 U/μl DpnII buffer shaking at 900 rpm. at 37°C 18 for 6 h; 40 μl of DpnII was added and samples were incubated overnight, shaking at 19 900 rpm. at 37°C. On day three, 20 μl DpnII buffer was added to the samples 20 followed by shaking for 5 more hours at 900 rpm. at 37 °C. DpnII subsequently was 21 inactivated at 65°C for 25 min and a 50-μl aliquot was taken to test digestion 22 efficiency (stored at -20°C). Next, digested chromatin was diluted in 5.1 ml H2O, 700 23 μl 10×ligation buffer (Thermo Fisher Scientific), 5 μl 30 U/μl T4 DNA ligase and 24 incubated at 16°C for 4 h while rotating. Ligated samples were incubated for a further 25 30 min at room temperature. Chimeric chromatin products and test aliquots were de-1 crosslinked overnight by adding 30 μl and 5 μl proteinase K, respectively, and 2 incubated at 65°C overnight. On the fourth day, 30 μl or 5 μl of 10 mg/ml RNase was 3 added to the samples and aliquots, respectively, and incubated for 45 min at 37°C. 4 Next, chromatin was precipitated by adding 1 volume phenol-chloroform to the 5 samples and aliquots, vigorously shaking them, followed by centrifugation at 4,000 6 rpm. at room temperature for 15 min. To precipitate aliquoted chromatin, 1 volume 7 100% ethanol and 0.1 volume 3M NaAc, pH 5.6 was added and the aliquots placed at 8 -80°C for 30 min. DNA was then precipitated by centrifugation at 5,000 rpm. for 45 9 min at 4°C followed by washing with 70% ethanol, and resuspension in 20 μl with 10 10 mM Tris-HCl, pH 7.5. To precipitate samples, extracted sample aqueous phases were 11 mixed with 7 ml H2O, 1 ml 3M NaAc, pH 5.6, and 35 ml 100% ethanol. Following 12 incubation at -20°C for at least 3 h, precipitated chromatin was isolated by 13 centrifugation at 5,000 rpm. for 45 min at 4 °C. The chromatin pellet was washed 14 with 70% ethanol and further centrifuged at 5,000 rpm. for 15 min at 4°C. Finally, 3C 15 library chromatin pellets were dried at room temperature and resuspended in 10 mM 16 Tris-HCl, pH 7.5. To check the 3C library, 600 ng were loaded on a 1% gel together 17 with the undigested and digested aliquots. The 3C library was then sheared using a 18 corresponding to the peak of the locus-TAD and the LRFN5 promoter) were analysed 23 by both DNA sequencing after bisulphite treatment, and custom designed MLPA tests 24 (using the P-300 kit from MRC-Holland, and following the manufacturer's protocol).
Still, no sex differences were found. All these methylation studies were done on 1 leukocyte DNA from peripheral blood samples. 2 LRFN5 expression in foetal brain cell cultures 3 Human neural stem cell lines were established from 12-week foetal brain cell cultures 4 using a conditionally immortalization methodology, as described. 23 STROC05, 5 STROC08 and STROC11 are striatal cell lines that are identical and from the same 6 dissection, just different isolated clones. An Illumina SNP array was used to identify 7 heterozygous SNPs in LRFN5, and the LRFN5 cDNA sequence of polyclonal brain 8 cells was then compared to the cDNA sequence of these cell lines. 9 SNP array analysis of a deletion polymorphism upstream of LRFN5

10
The allele frequencies of a common 60 kb deletion, chr14(GRCh37):g.41609383-11 41669664, detected by at least 12 oligonucleotides on the Affymetrix 6.0 SNP array 12 and by at least 18 oligonucleotides on the Affymetrix CytoScan SNP array 13 (Affymetrix, ThermoFischer Scientific, USA), were determined in four cohorts of 14 individuals: three ascertained due to developmental disorders, and one group of 15 mostly medical students (Table 2). All cohorts were anonymized. Three of the cohorts 16 (patient cohort I and III and the student cohort) had their genomic copy number status 17 determined by the Affymetrix 6.0 array, while patient cohort II was tested by the 18 Affymetrix CytoScan array. 19

Statistical analysis 20
Basic and on-line statistical tools were used to investigate the statistical significance 21 of the following findings: the segregation of the A-haplotype in the autism family: 22 Fisher Exact Test (p = 0.0476); the female-male comparison in the table of Figure 3: 23 equilibrium and observed allelic distribution in the patient cohorts in Table 2: Chi-1 square test (p < 0.00001); and the difference between the allelic distribution in the 2 patient cohorts and the student cohort in Table 2: Chi-square test (p < 0.00001). 3

RESULTS 4
LRFN5 locus structural changes can be associated with developmental 5 delay and autism 6 Our attention was drawn to the LRFN5 locus in 2007 when a girl with intellectual 7 disability (ID) that gave no social contact was found to have two de novo 8 chromosome changes: a small 2q31.1 deletion and a balanced 14;21-translocation. 24 9 One translocation breakpoint was in the LRFN5 locus at 14q21.1, and the other on the 10 acrocentric 21p arm. To explore the consequences of these chromosome aberrations 11 for chromatin structure, we performed ChIP-on-chip experiments on patient skin 12 fibroblasts. The results were compared to a non-autistic boy with another LRFN5 13 locus translocation: a de novo t(6;14)(q26;q21.1) causing Coffin-Sirin syndrome 14 because ARID1B was disrupted on chromosome 6 (Supplementary Figure 1). He 15 speaks well and has good social function. A clear difference in the LRFN5 locus 16 chromatin profiles was only seen in the autistic girl. Possibly this is because her 17 translocation fused the LRFN5 locus to an acrocentric p-arm, and a trimethylated 18 histone-3-lysine-9 (H3K9me3) spreading effect from FBXO33 and centromeric could 19 be seen on derivative chromosome 14, while more trimethylated histon-3-lysin-4 20 (H3K4me3) and less trimethylated histon-3-lysin-27 (H3K27me3) was seen in the 21 LRFN5 locus on derivative chromosome 21 (Supplementary Figure 1). In addition, an 22 apparent "compensatory" H3K4me3-effect corresponding to the 2.6 Mb 2q31.1 23 deletion could also be seen. However, this unexpected and unprecedented effect could be 2q31.1-locus specific as it was not found in an individual with a 2.2 Mb 1 chromosome 17q23.1 deletion (Supplementary Figure 2). These two unique 14;21-2 translocation patients initiated our LRFN5 research, and we noted that autistic 3 behaviour was only seen in the individual with chromatin changes to this locus. 4 Of more general relevance for LRFN5 function is the number of individuals with 5 DD/ASD and copy number variants (CNVs) in the LRFN5 locus in the DECIPHER 6 database (n=30) and our own records (n=3) ( Supplementary Figures 3 and 4, pluss  7 reference 25 ). Only one small deletion containing LRFN5 itself was registered as de 8 novo. Because many CNVs are inherited from a seemingly unaffected parent, most 9 have been regarded as non-pathogenic. Nevertheless, if this truly is an ASD 10 susceptibility locus, even reduced penetrance should not have prevented that from 11 being discovered. However, if sex-of-origin also matters for autism susceptibility, an 12 ASD link could easily be missed. 13 Autism in males segregated with a specific LRFN5 locus haplotype 14 inherited from their mothers 15 High-functioning ASD was diagnosed in two pairs of brothers from two families from 16 the same geographical region (Table 1; fam #1and fam #2). These four males have 17 been seen by the senior author both as children and adults, and their autism phenotype 18 is strikingly similar. They all had a 172 kb deletion just upstream of the LRFN5 19 promoter and the same locus haplotype (called the A-haplotype in Table 1), inherited 20 from their normal mothers. The families were too distant to know about any 21 relatedness. The A-haplotype was also found in a normal maternal uncle, inherited 22 from his mother. Of the nine individuals sharing the A-haplotype in these families, 23 4/5 males had ASD, and 4/4 SRY-negatives (three females and one XX-male) were normal. This may suggest that the A-haplotype increased ASD susceptibility in males 1 in this family (Fisher Exact Test p=0.0476). We also found overlapping deletions on 2 different haplotypes in four other individuals or families, ascertained by copy-number 3 high-resolution SNP-array testing because of developmental delay, ID, ASD or 4 schizophrenia (Table 1). This could indicate that the A-haplotype is linked to male 5 ASD-susceptibility, not the 172 kb deletion per se. An effect of the deletion is not 6 excluded, but then it must be haplotype dependent. The 172 kb deletion does not 7 contain enhancer-like chromatin profiles or lncRNAs. 8 The structure of the LRFN5 locus is influenced by the sex of the 9 individual 10 To explore if the A-haplotype influenced TAD-structure of the LRFN5 locus, we 11 performed a capture HiC-experiment on skin fibroblasts from a family trio (boy with 12 ASD and his unaffected parents), with a normal unrelated female as control ( Figure  13 1). Three patterns emerged: In the autistic boy and his mother, both sharing the A-14 haplotype, the whole-locus 5.4 Mb mega-TAD had three sub-TADs with boundaries 15 at the LRFN5 promoter (arrow A in Figure 1) and the middle of the LRFN5-16 downstream gene desert (arrow B in Figure 1). The small red diamond just 17 centromeric to the LRFN5 promoter marks the 172 kb deletion. In the father, only two 18 TADs could be discerned inside the mega-TAD. No distinct "B-junction" could be 19 seen. In the control female, only the mega-TAD was distinct; an "A-junction" was 20 diffuse if at all present (Figure 1). 21 These subtle capture-HiC differences point to variation in TAD structure and that the 22 autistic boy inherited his mother's structure. Hypothetically, this specific structure 23 (with a "B-junction") could be A-haplotype dependent and the one linked to autism in this family. We also noted that the father and the control female had TAD differences 1 (Figure 1). 2 To explore if this could reflect differences in their chromatin profiles, chIP-on-chip-3 generated chromatin data (H3K4me3, H3K27me3 and H3K9me3) on fibroblasts from 4 four family members, five control males and seven control females, were compared. 5 A sex difference was found for the H3K9me3 profiles, i.e. heterochromatin protein-1 6 (HP1)-associated constitutive heterochromatin, but only corresponding to the LRFN5 7 gene itself and the region downstream to the H3K4me3/H3K27me3 signals marking 8 the peak of the mega-TAD (Figure 2). In this area, males had more heterochromatin 9 than females. This difference was significant, but individual differences in the degree 10 of this effect should be noted (Figure 2). In fact, the groups are overlapping (Figure  11 3). In contrast, the sex difference was zero in a region 1 Mb upstream (Figure 3). 12 The four family members, including the mother and the SRY-negative brother of the 13 index boy that shared the A-haplotype (Table 1) To explore if we could test large numbers of individuals for sex-related or autism-21 related differences in this region, we performed a pilot MLPA-based study on 22 anonymized blood leukocyte DNA to see if these sex difference in chromatin profiles 23 was also reflected in methylation differences of a CpG located at the conserved CTCF binding site in the LRFN5 promoter (chr14:42,069,895-42,069,949, hg19). An 1 average methylation degree of 0.19 (CI 0.13-0.25) in control males (n=16) and control 2 females (n=6) were found with no sex difference, and the same average level (0.20) 3 was found in patients investigated because of non-ID associated autism (n=14). In a 4 separate experiment, using the Illumina 450K BeadChip methylation test, we found 5 an average degree of CpG methylation of 0.14 at the LRFN5 promoter (19 CpGs were 6 interrogated) and 0.52 in the LRFN5 gene itself (8 CpGs were interrogated) in 10 7 other control individuals, still without sex difference. This shows that the difference in 8 fibroblast H3K9me3 profiles at the LRFN5 gene is not reflected in differences in CpG 9 methylation of leukocyte DNA in the same region. 10

Allelic interaction must be frequent in the LRFN5 locus 11
If the LRFN5 locus structure is critical for brain function, how is it maintained when 12 copy number changes are so common? A possible answer to this question was 13 unexpectedly found upon examining the allelic distribution of a common 60 kb 14 deletion in the middle of the LRFN5-upstream gene desert (chr14:41,609,383-15 41,669,664 (hg19); Table 2 (position of deletion indicated in Figure 4). We examined 16 three large and independent cohorts of individuals (two from Norway and one from 17 the Netherlands) who had been investigated with a high-density SNP-array because of 18 a developmental disorder, usually including variable degrees of developmental delay 19 (DD). In addition, one Dutch cohort (n=1,416) of mainly medical students served as a 20 control. The deletion's minor allele frequency (MAF) was around 8.5% in the Dutch 21 population, 15% in the Norwegian population, and 14% in gnomAD 22 (gnomad.broadinstitute.org). 26 When comparing the observed allelic distribution in 23 these cohorts with the expected distribution as per Hardy-Weinberg equilibrium, there were too few heterozygotes in all cohorts. We found that the loss of heterozygote 1 wild-type (wt) / deletion (del) was of equal magnitude in both directions, i.e. to wt/wt 2 and to del/del. In the three patient cohorts, this loss was 20%, 24% and 26%, 3 respectively, but in the student cohort, it was 80% (Table 2). This surprising 4 difference was not a technical artefact. 12-18 SNP-array oligonucleotides covered the 5 60 kb deletion, and >5 in a row is usually sufficient for correct calling. Furthermore, 6 if this was due to missed wt/del calling, the high number of del/del homozygotes 7 would still be incompatible with Hardy-Weinberg equilibrium. Also, when examining 8 SNP array results from our diagnostic routine, we found that there was an excess of 9 homozygosity regions >1 Mb in the LRFN5 locus compared to many other regions in 10 the genome. We are not aware of any other potential explanations for these 11 observations than an early mitotic allelic conversion event with a frequency of at least 12 1 in 5 individuals in the DD group and as high as 4 in 5 individuals in the student 13 group. This is far beyond expectation. Maybe this allelic interaction is needed to 14 establish monoallelic expression. In human fetal brain cell cultures from the striatum, 15 we did indeed find evidence indicating monoallelic LRFN5 expression 16 (Supplementary Figure 5). 17

DISCUSSION 18
Despite extensive research, it has remained elusive why autism, and especially the 19 higher-functioning variants, is more common in males than in females. 2 Here we 20 show that the synaptic regulation and maintenance gene LRFN5, situated in the 21 middle of a conserved gene desert capable of forming a 5.4 Mb mega-TAD, may be 22 part of the answer to this question. The three main reasons for this are the large family 23 with remotely related pairs of brothers with a very similar form of higher-functioning autism sharing the same maternally inherited LRFN5 locus haplotype (Table 1), the 1 sex-influenced differences in locus chromatin 2D structure in fibroblasts from normal 2 males and females (Figure 2 and 3), and the allelic interaction that must take place in 3 early embryonic development, as evidenced by the striking deviation from Hardy-4 Weinberg equilibrium of a 60 kb deletion polymorphism (Table 2). 5 There are eight putative SOX9 and SRY binding sites flanking the LRFN5 locus; 3+2 6 SOX9 sites and 1+2 SRY sites (Figure 2 and Supplementary Figure 6; TFBS_conc 7 track in the UCSC browser). This could have importance for generation of the 8 observed male/female chromatin difference (Figures 2 and 3). The mechanism behind 9 the early locus interaction is more difficult to explain. We are unaware of such 10 frequent allelic interaction and locus conversion in any other part of the genome, e.g. 11 the frequency of inter-chromosomal gene conversion in gene families with more than 12 two alleles has been estimated to be around 0.2%. 27 Gene conversion events are more 13 common towards the 3'UTR end of protein-coding genes, and this is believed to be 14 due to RNA transcription aiding the process. 27 Maybe LINC02315, an RNA gene 15 upstream of LRFN5 ending in the common 60 kb deletion polymorphism, has a 16 similar role (Figure 4). 17 The most interesting question is why frequent locus homozygotisation by gene 18 conversions occurs. We hypothesize that this process is needed to establish stochastic 19 monoallelic expression, advantageous for fine-tuning LRFN5 expression. Given the 20 high conversion frequencies, allelic interaction is probably the rule, and allelic 21 conversion a consequence. Exploratory foetal brain expression data suggests that 22 monoallelic expression does occur in humans (Supplementary Figure 5), but we do 23 not know if this is always the case or if it is stochastic. The next question is why fine-tuning of LRFN5 expression is so critical that the gene, on top of having a 1.9 kb 1 5'UTR encoded by exons 1 and 2, needs to be framed by 2-3 Mb of conserved, 2 presumed regulatory, gene deserts. Maybe the difference in locus conversion 3 frequencies between a student cohort (4 in 5 were converted to homozygosity) and 4 three DD-cohorts (1 in 4 were converted to homozygosity) provides a clue (Table 2). 5 Hypothetically, if the LRFN5 locus is a genetic basis for synaptic memory, then both 6 allelic interaction and conversion could be advantageous for this function. This fits 7 well with LRFN5's role as a protein involved in synapse strength and dynamics, 15,20,28 8 and it could have relevance for the photographic memory of details that some autistic 9 individuals may have. 10 The concept of autism-related risk haplotypes of the LRFN5 locus, variable locus 11 structure influenced by the individual's sex, and early allelic homozygotisation, fits 12 well with the complex pattern of ASD inheritance. 4 In the families with the ASD-13 susceptible A-haplotype described here, all ASD males were born to carrier mothers, 14 and one non-penetrant male was also recorded (Table 1). Of note, this male was hemi-15 methylated at rs144497930, i.e. at the major TAD-peak around the CTCF binding site 16 in Figure 4 (Supplementary Table 1). Other family members and control individuals 17 were fully methylated at this CpG. Maybe the epigenetic pattern predisposing to 18 autism was not established during his early embryogenesis. In a "locus-interaction-19 determines-ASD-susceptibility" model, this makes sense and would also explain why 20 rare deletions recorded in DECIPHER may be pathogenic despite inheritance from a 21 presumed normal male or female parent (Supplementary Figure 3). Such deletions 22 could interfere with allelic interaction and epigenetic regulation. It should also be 23 noted that while many different deletions may be seen in the LRFN5 locus 24 (Supplementary Figure 6), duplications are rare, also in DECIPHER (Supplementary Figure 4). Finally, the whole LRFN5 locus or the LRFN5 gene itself can be disrupted 1 or deleted, apparently without (additional) phenotypic consequences, as indicated 2 from the DECIPHER database, a gnomAD pLI of 0.56, 26 and our own diagnostic 3 genomic copy-number records. This is as expected if stochastic monoallelic 4 expression is the rule. 5

LIMITATIONS 6
The capture-HiC and chromatin chIP-on-chip experiments were all done on human 7 skin fibroblasts, not brain cells, which is the relevant tissue for LRFN5 expression. As 8 the TAD (3D) structure and probably also the chromatin (2D) structure can be activity 9 dependent, the relevance of these finding for LRFN5 locus structure in brain cells is 10 yet unknown, but TAD structure tends to be conserved between cell types. The size of 11 the control cohort exploring sex-related differences in H3K9me3 profiles of the 12 LRFN5 locus is small (6 males +7 females), and these data need to be confirmed in an 13 independent and preferably larger cohort to be certain of a sex difference. Hopefully, 14 an easier method to investigate such changes will be developed. Finally, the 15 monoallelic expression data (Supplementary Figure 5) is from a single experiment. 16

CONCLUSIONS 17
Our work suggests that LRFN5 regulation could be Y-chromosome-dependent and 18 complex with a strikingly high frequency of early embryonic allelic interaction with 19 locus conversions. Structural variants in this locus are linked to autism, and we 20 speculate that this locus is a significant part of the genetic basis of synaptic memory.         Capture-HiC results of a family trio (index male with ASD, his mother and father) and a control female (of the same age as index' mother). Arrows indicate TAD junctions (A and B). The small diamond just centromeric to TAD junction A (corresponds to the LRFN5 promoter) marks the 172 kb familial deletion. Top: LRFN5 locus H3K9me3 chIP-on-chip pro les from index with ASD, his 46,XX17 DSD non-autistic brother, his normal mother and his normal father (top four lanes). Middle: Pro les from 5 control males.
Bottom: 1 Pro les from 7 control females. Note the pro le variability around the LRFN5 gene in contrast to the anking H3K9me3-enriched domains.