Molecular Psychiatry (2009) 14, 705–718 & 2009 Nature Publishing Group All rights reserved 1359-4184/09 $32.00 www.nature.com/mp ORIGINAL ARTICLE Involvement of the PRKCB1 in autistic disorder: significant genetic association and reduced neocortical C Lintas1,2,14, R Sacco1,2,14, K Garbett3, K Mirnics3,4, R Militerni5, C Bravaccio6, P Curatolo7, B Manzi7, C Schneider8, R Melmed9, M Elia10, T Pascucci11,12, S Puglisi-Allegra11,12, K-L Reichelt13 and AM Persico1,2 1Laboratory of Molecular Psychiatry and Neurogenetics, University ‘Campus Bio-Medico’, Rome, Italy; 2Laboratory of Molecular Psychiatry and Psychiatric Genetics, Department of Experimental Neurosciences, I.R.C.C.S. ‘Fondazione Santa Lucia’, Rome, Italy; 3Department of Psychiatry, Vanderbilt University, Nashville, TN, USA; 4Kennedy Center for Research on Human Development, Vanderbilt University, Nashville, TN, USA; 5Department of Child Neuropsychiatry, II University of Naples, Naples, Italy; 6Department of Child Neuropsychiatry, University ‘Federico II’, Naples, Italy; 7Department of Child Neuropsychiatry, University ‘Tor Vergata’, Rome, Italy; 8Center for Autism Research and Education, Phoenix, AZ, USA; 9Southwest Autism Research and Resource Center, Phoenix, AZ, USA; 10Unit of Neurology and Clinical Neurophysiopathology, I.R.C.C.S. ‘Oasi Maria S.S.’, Troina (EN), Italy; 11Department of Psychology, University ‘La Sapienza’, Rome, Italy; 12Laboratory of Behavioral Neurobiology, Department of Experimental Neurosciences, I.R.C.C.S. ‘Fondazione Santa Lucia’, Rome, Italy and 13Department of Pediatric Research, Rikshospitalet, University of Oslo, Oslo, Norway

Protein C play an important role in signal transduction, regulation of gene expression and control of division and differentiation. The fsI and bII isoenzymes result from the of the PKCb gene (PRKCB1), previously found to be associated with autism. We performed a family-based association study in 229 simplex and 5 multiplex families, and a postmortem study of PRKCB1 gene expression in temporocortical gray matter (BA41/42) of 11 autistic patients and controls. PRKCB1 gene haplotypes are significantly associated with autism (P < 0.05) and have the autistic endophenotype of enhanced oligopeptiduria (P < 0.05). Temporocortical PRKCB1 gene expression was reduced on average by 35 and 31% for the PRKCB1-1 and PRKCB1-2 isoforms (P < 0.01 and < 0.05, respectively) according to qPCR. amounts measured for the PKCbII isoform were similarly decreased by 35% (P = 0.05). Decreased gene expression characterized patients carrying the ‘normal’ PRKCB1 alleles, whereas patients homozygous for the autism-associated alleles displayed mRNA levels comparable to those of controls. Whole genome expression analysis unveiled a partial disruption in the coordinated expression of PKCb-driven , including several cytokines. These results confirm the association between autism and PRKCB1 gene variants, point toward PKCb roles in altered epithelial permeability, demonstrate a significant downregulation of brain PRKCB1 gene expression in autism and suggest that it could represent a compensatory adjustment aimed at limiting an ongoing dysreactive immune process. Altogether, these data underscore potential PKCb roles in autism pathogenesis and spur interest in the identification and functional characterization of PRKCB1 gene variants conferring autism vulnerability. Molecular Psychiatry (2009) 14, 705–718; doi:10.1038/mp.2008.21; published online 4 March 2008 Keywords: autism; pervasive developmental disorders; PRKCB1; -b; temporal cortex; TGF-b

Introduction social skills, and by repetitive and stereotypic behaviors.1 Family and twin studies support strong Autism is a severe neuropsychiatric disorder char- genetic contributions to this disease.2 However, the acterized by impaired language, communication and heterogeneity of clinical symptoms and the complex- ity of underlying pathogenetic processes have until Correspondence: Dr AM Persico, Laboratory of Molecular now undermined efforts to achieve reproducible Psychiatry and Neurogenetics, University ‘Campus Bio-Medico’, genotype–phenotype correlations. On one hand, the Via Alvaro del Portillo 21, I-00128 Rome, Italy. phenotypic expression of autism-predisposing genes E-mail: [email protected] 14These authors contributed equally to this study. spans from minimal autistic traits to full-blown Received 4 August 2007; revised 23 January 2008; accepted 23 autism, identifying a broad clinical entity referred to January 2008; published online 4 March 2008 as ‘autism-spectrum disorder’ (ASD).2,3 On the other PRKCB1 gene and autistic disorder C Lintas et al 706 hand, the genetic underpinnings of autism encompass system, PKCb isoenzymes play a critical role in B-cell significant interindividual heterogeneity, numerous receptor-mediated responses, T-cell migration and contributing loci, epistasis and likely gene–environ- cytokine secretion, dendritic cell differentiation and ment interactions.2 Incomplete penetrance (that is, monocyte and macrophage functioning.20–25 In the individuals carrying autism genes, but not fulfilling gut, PKCbI and bII are expressed by epithelial cells, diagnostic criteria for the ‘affected’ status) and where they regulate intestinal permeability and phenocopies (individuals carrying no genetic predis- cell proliferation, respectively.26–28 In the kidney, position and fulfilling diagnostic criteria for ‘autism’ both PKCb isoforms are expressed by mesangial cells solely due to environmental factors) further decrease and play a major role in diabetic nephropathy and the statistical power of genetic analyses by introdu- albuminuria, possibly through oxidative stress.29–31 cing false-negative and false-positive ‘affection status’ Interestingly, many of the non-neural signs and definitions. In this complex scenario, the probability symptoms that often accompany autism strikingly of success can be maximized by employing in parallel overlap with the pathophysiological roles played by both genetic analyses and postmortem assessments of PKCb isoforms in districts outside of the CNS. brain gene expression patterns. This strategy can The chromosomal region encompassing the PRKCB1 indeed promote a better understanding of single-gene locus was initially identified using a direct identity- contributions to complex disorders, as recently by-descent mapping method as one of seven regions demonstrated for the MET gene in autism.4,5 linked to autism, each spanning only between 0.75 and Several lines of evidence suggest that autism 4.0 Mb.14 The same study described an association should be viewed as a multiorgan systemic disorder between PRKCB1 gene variants and autism both with a prenatal onset. On one hand, autism does not in Caucasian-Americans and in Mexican-Americans, solely affect the central nervous system (CNS), albeit with the two ethnic groups carrying vulnerability despite encompassing obvious neurodevelopmental alleles marked by different haplotypes.14 This result components: systemic signs and symptoms include was later not replicated in an Irish sample (see macrosomy,6 excessive intestinal permeability and Discussion).32 Still, the presence of a significant genetic nonspecific enterocolitis,7–9 immune dysreactivity9 association in two different ethnic groups, coupled and renal oligopeptiduria.10 On the other hand, the with the functional involvement of PRKCB1 in many neurodevelopmental mechanisms underlying the pathophysiological processes potentially underlying CNS abnormalities found in postmortem studies, disease-related endophenotypes, spur interest into which include reduced programmed cell death and/ additional studies of PRKCB1 in autism. In accordance or increased cell proliferation, and altered neuronal with the strategy outlined above and previously migration, differentiation and synaptogenesis, with employed successfully with the MET gene,4,5 this study the exception of the latter, are all active prenatally, was designed to replicate the initial genetic findings especially during the first trimester of pregnancy.2,11,12 in an independent sample and to extend these studies Indeed, many children later diagnosed with ASD by (a) assessing PRKCB1 gene expression in postmortem display motor abnormalities13 and/or excessive body temporocortical gray matter (BA41/42) from 11 pairs of growth6 already on the day of birth or in early ASD patients and sex-, age-, and postmortem interval- neonatal life. Therefore, genes characterized by an matched controls; (b) correlating PRKCB1 gene expres- early onset of expression and encoding sion with PRKCB1 genotypes in these postmortem brain involved in the control of cell division, differentiation samples and (c) assessing the functional consequences and migration can represent attractive candidates for of dysregulated PRKCB1 gene expression using genome- autistic disorder even if their tissue distribution wide expression array technology. Our results confirm patterns are not necessarily restricted to the CNS. the existence of a significant association between The PRKCB1 gene, located in human ch. 16p11.2, PRKCB1 gene variants and autism, describe for the first represents an interesting locus displaying expression time an association with the biochemical endopheno- patterns not restricted to the CNS and previously type defined by enhanced urinary peptide excretion found associated with autism.14 In general, protein rates, detect a significant reduction of PRKCB1 mRNA kinase C enzymes play an important role in signal and protein levels in postmortem autistic brains, reveal transduction, regulation of gene expression and a dysregulation of gene expression patterns delineating control of cell division and differentiation. The alter- a possible compensatory adjustment aimed at limiting native splicing of PRKCB1 generates two mRNA an ongoing immune dysreactive process and identify a isoforms named PRKCB1-1 and PRKCB1-2, yielding lack of downregulation in gene expression as the most the two PKCb isoenzymes bI and bII. Interestingly, likely functional correlate of PRKCB1 alleles conferring either or both isoforms are expressed in the districts autism vulnerability. most affected in autistic disorder, namely the CNS, immune system, digestive tract and kidney. In parti- Subjects and methods cular, brain PRKCB1 is expressed in hippocampus, striatum, suprachiasmatic nucleus and cerebellar Subjects recruited for the family-based association granule cells, where PKCbI influences circadian study rhythms, learning and memory, whereas PKCbII is A total of 229 simplex and 5 multiplex families with involved in fear conditioning.15–19 In the immune a non-syndromic autistic proband were recruited for

Molecular Psychiatry PRKCB1 gene and autistic disorder C Lintas et al 707 Table 1 Genetic sample composition

Site Number of individuals Number of families Number of with autism complete trios Simplex Multiplex

II University of Naples (Naples, Italy) 51 51 — 51 I.R.C.C.S. ‘Oasi Maria S.S.’ (Troina, Italy) 37 35 1 36 I.R.C.C.S. ‘Ospedale Bambino-Gesu` ’ (Rome, Italy) 33 33 — 33 University ‘Federico II’ (Naples, Italy) 28 28 — 28 Southwest Autism Research Center (Phoenix, AZ) 27 21 3 24 II University of Rome ‘Tor Vergata’ (Rome, Italy) 22 20 1 21 U.C.B.M. (Rome, Italy) 21 21 — 21 University of Milan (Milan, Italy) 15 15 — 15 University of Turin (Turin, Italy) 4 4 — 4 A.S.L. of Rimini (Rimini, Italy) 1 1 — 1 Total sample 239 229 5 234

Table 2 Single nucleotide polymorphisms assayed in this study: (A) position on 16 relative to Build 36.2, allelic frequencies, and size of the sample genotyped; (B) linkage disequilibrium, expressed as values of D0 (range: À1to þ 1)/r2 (range: 0–1, in italics) rs number hCV ID Position on ch Allele 1 Allele 2 Allele 1 Allele 2 Number of families 16 frequency frequency genotyped (total N = 234)

(A) rs3785392 hCV11192725 23 851 984 G A 0.464992 0.535008 226 rs3785387 hCV1936137 23 869 794 A G 0.438466 0.561534 222 rs196002 hCV946275 23 870 738 A G 0.462079 0.537921 223 rs1873423 hCV11895960 23 878 178 C T 0.158919 0.841081 228

(B) rs number rs3785392 rs3785387 rs196002 rs1873423 rs3785392 — 0.905/0.736 0.712/0.495 0.349/0.027 rs3785387 0.905/0.736 — 0.654/0.389 0.367/0.033 rs196002 0.712/0.495 0.654/0.389 — 0.285/0.018 rs1873423 0.349/0.027 0.367/0.033 0.285/0.018 — this study. The composition of our clinical sample, Mental Developmental Scales, the Coloured Raven encompassing 239 autistic patients and 90 unaffected Matrices, the Bayley Developmental Scales or the siblings, is summarized in Table 1. Demographic and Leiter International Performance Scale.33 All parents clinical characteristics as well as diagnostic screening gave written informed consent for themselves and for procedures used to exclude syndromic autism have their children, using the consent form approved by been previously described.33 Briefly, patients fulfill- the I.R.B. of University Campus Bio-Medico (Rome, ing DSM-IV diagnostic criteria for Autistic Disorder1 Italy). were screened for non-syndromic autism using magnetic resonance imaging, electroencephalogram, Genotyping audiometry, urinary aminoacid, and organic acid Four single nucleotide polymorphisms (SNPs) loca- measurements, cytogenetic and fragile-X testing. ted in intron 2 of the PRKCB1 gene were genotyped, Patients with dysmorphic features were excluded including the three SNPs previously found associated even in the absence of detectable cytogenetic altera- with autism in Caucasians,14 and one additional SNP tions. Patients with sporadic seizures (that is, < 1 located approximately 18 kb upstream (Table 2A). every 6 months) were included; patients with fre- SNPs were genotyped using the TaqMan method quent seizures or focal neurological deficits were (Applied Biosystems, Foster City, CA, USA), with excluded. Autistic behaviors were assessed using the probes purchased from the manufacturer and used official Italian version of the Autism Diagnostic according to the manufacturer’s guidelines. DNA was Observation Schedule34 and the Autism Diagnostic PCR-amplified with denaturation at 95 1C for 10 min, Interview-Revised (ADI-R);35 adaptive functioning 40 cycles at 92 1C for 15 s, 60 1C for 1 min and 72 1C was assessed using the Vineland Adaptive Behavior for 45 s, followed by elongation at 72 1C for 5 min. Scales; I.Q. was determined using either the Griffith TaqMan assays were then read on a 7900HT Fast

Molecular Psychiatry PRKCB1 gene and autistic disorder C Lintas et al 708 Real-Time PCR System (Applied Biosystems), and drug therapy, puberal status and differential growth alleles were called using the SDS software (Applied rates, quantitative analyses were restricted to patients Biosystems). aged < 11 years and not taking selective 5-HT reup- take inhibitors for 5-HT blood levels, and to patients Biochemical and morphological endophenotypes aged < 16 years for head size, as in our previous Blood samples for 5-HT levels were obtained from studies.6,33,36 Statistical analyses were performed all family members and centrifuged within 20 min merging together our 210 Italian and 24 Caucasian- of venipuncture at 140 g for 25 min at 4 1C; 1 ml American families (Table 1) after population structure of supernatant (that is, platelet-rich plasma) was analyses provided no evidence of major genetic stored at À80 1C and assessed by high-performance dyshomogeneity in a subset of 155 Italian and 24 liquid chromatography, as described.36 Urinary pep- Caucasian-American patients randomly chosen one tide excretion analysis was performed by high-perfor- per family, genotyped at 90 unlinked SNPs distribu- mance liquid chromatography on the first morning ted genome-wide and analyzed using the Structure from urine samples of all family members, diluted program, as described (see Supplementary Methods).41 to 250 nm creatinine, as described.10 The total area Data are expressed as mean±s.e.m., except for head of peaks under the 215 nm absorption curve in the circumference and urinary peptide excretion rates, peptide region following the hippuric acid peak was expressed as median percentile±interquartilic range. calculated and expressed in mm2. Excess peptiduria Two-tail P-values are reported, with significance level was defined on the basis of previously published set at P < 0.05. No correction for repeated measures normality ranges determined in population con- was implemented in our family-based association trols.10 Head circumference was measured in autistic study, because (a) it has been undertaken to replicate patients and unaffected siblings by trained physicians previously published positive findings;14 (b) we docu- using a non-stretchable plastic measuring tape ment the relative genetic homogeneity of our Italian graded in millimeters, placed over the maximum and Caucasian-American patients; (c) the four SNPs fronto-occipital head perimeter; head circumference used to define a single haplotype, previously found measures were transformed into percentiles using associated with autism,14 are characterized by short sex- and age-specific standard tables, as described.6 intermarker physical distances, significant LD and non-independent marker segregation (see Results Statistical analyses section); (d) many endophenotypic measures assessed Hardy–Weinberg equilibrium was tested using the w2 in the same autistic patients are non-independent. statistic, as implemented by the HAPLOVIEW soft- Instead, statistical significance was set at P < 0.01 for ware (available at http://www.broad.mit.edu/mpg/ tests of Hardy–Weinberg equilibrium (that is, < 0.05/5 haploview/index.php)37 for the total sample and comparisons), whereas nominal P-values are reported by HWE (available at http://linkage.rockefeller.edu) to for exploratory genotype–phenotype correlations analyze separately autistic patients, mothers, fathers, regarding clinical measures, due to the large number and unaffected siblings. Family-based single-marker of clinical variables tested. and haplotype association tests were performed using the FBAT statistic (S = STX, where T is the phenotypic trait and X the marker value), as implemented by Patient brain tissue information the FBAT software package (available at http://www. Postmortem studies were performed using frozen biostat.harvard.edu/~fbat/fbat.htm), under an addi- brain tissue samples dissected from the superior tive model (option Àe).38 The FBAT statistic stems temporal gyrus (BA 41/42) of 11 patient–control pairs, from the transmission/disequilibrium test (TDT), obtained through the Autism Tissue Program from where preferential allelic transmission from hetero- the Maryland NICHD Brain Tissue Center and the zygous parents to affected offspring is tested by apply- Harvard Brain Tissue Resource Center. This neo- ing the (bÀc)2/(b þ c) statistics and the w2 (‘McNemar cortical region was chosen because it hosts well- test’).39 Quantitative traits were analyzed by quanti- documented structural and functional abnormalities tative TDT, as implemented by the FBAT software,38 in autism.42 These tissue samples largely overlap with T as the quantitative trait of interest (instead of with those employed in our recent study of the MET a dichotomic affected/unaffected status as in the pathway.5 Clinical and demographic information, TDT), and by parametric or nonparametric ANOVA, family history and autopsy reports were obtained based on genotype distributions. The HBAT command from the Autism Tissue Program web site (www. in FBAT was also employed to estimate haplotype atpportal.org) and are summarized in Table 3. The frequencies and linkage disequilibrium (LD) from presence of mental retardation (MR) was defined on pedigree data.38 TDT analyses controlling for quanti- the basis of a full-scale IQ < 70. ASD cases fulfilled tative covariates were performed using the multi- DSM-IV diagnostic criteria,1 confirmed using the nomial logistic regression model underpinning the Autism Diagnostic Interview-Revised (ADI-R).35 retrospective likelihood method implemented by the Controls were selected to match patients on sex, Unphased software package, applying the ‘confoun- age (±2 years, with the exception of ±4 years in pair der’ and ‘uncertain haplotypes’ options.40 To mini- nine) and postmortem interval, as much as possible mize the impact of confounding variables, such as (Table 3).

Molecular Psychiatry Table 3 Brain tissue information for patients and controls

Pair Case no.a Diagnosis Age Sex PMI (h) c Cause of death Mental Epilepsy Other features Drug therapies at time of no. (years)b retardation death

1 UMB4721 Autism 8 Male 16 Drowning Unknown No — None 2 UMB1174 Autism 7 Female 14 Sudden death, Yes Yes Coloboma iris, cortical Lamictal, Valiumd seizure heterotopias, mesial temporal sclerosis, lymphoadenopathy, recurrent infections 3 B5342 Autism 11 Female 13 Drowning Yes Yes Recurrent otitis Topamax, Lamictal, Adderall (seizure?) 4 B5569 PDD-NOS 5 Male 25.5 Drowning No No Recurrent otitis, angioedema, Prozac, melatonind food allergies 5 B5173 Autism 30 Male 20 Gastrointestinal Yes Yes Large ear lobes Dilantin, Depakote, Tranxene hemorrage bid, Cisapride, Clorazepate, folic acid, Oxcarbazepinad 6 B6337 Autism 22 Male 25 Seizure Yes Yes Intestinal lymphoadenopathy, Lamectil, Zonegran, hypertrophic spleen, recurrent Neurontin, Abilify, flax seed otitis oil, omega-3, multivitamin 7 B5321 Autism 19 Female 12.5 Traffic accident No No No No information sib 8 B5144 Autism 20 Male 23.7 Trauma Yes No — None 9 UMB4671 Autism 4 Female 13 Trauma Yes No — None 10 B6294 Autism 16 Male Unknown Seizure Unknown Yes — Topamax, Depakote, Allegra, Claritin, NuThera Multivitamin Lintas disorder C autistic and gene PRKCB1 11 B5000 Autism 27 Male 8.3 Drowning Yes No Septo-optic dysplasia, agenesis Synthroid, mannitol of the septum pellucidum,

hypothalamic and pituitary al et dysfunction 1 UMB1860 Control 8 Male 5 Cardiac No No — None arrhythmia 2 UMB1377 Control 6 Female 20 Drowning No No — None 3 UMB1407 Control 9 Female 20 Asthma No No — None 4 UMB1185 Control 4 Male 17 Drowning No No — None 5 B4211 Control 30 Male 23 Cardiac No No — None arrhythmia 6 B6221 Control 22 Male 24 Unknown No No — None 7 UMB1541 Control 20 Female 19 Head injuries No No — None 8 B3829 Control 22 Male 12 Central hepatic No No — None laceration 9 UMB1706 Control 8 Female 20 Rejection of heart No No — None transplant 10 B6207 Control 16 Male 26 Ischemic heart No No — None attack 11 B5873 Control 28 Male 23 Unknown No No — No information

a oeua Psychiatry Molecular Autism Tissue Program identifier. bMean age (±s.d.) for the autism group = 15.4±9.0, for controls = 15.7±9.2; paired-t = À0.740, 10 d.f., P = 0.476. cPMI = postmortem interval; mean PMI (±s.d.) for the autism group = 17.1±6.0, for controls = 19.0±6.0, paired-t = À0.425, 9 d.f., P = 0.681. dPharmacological therapy from the last available report, dating back to less than a year before death. 709 PRKCB1 gene and autistic disorder C Lintas et al 710 Expression analysis by real-time PCR RNA samples from six case–control pairs (no. 2, 4, 5, 7, Total RNA was extracted from all 11 case–control 8 and 11 in Table 3) provided RIN values > 6.9 and were pairs (Table 3) using the TRIzol reagent (Invitrogen, thus hybridized to HG_U133plus2 human Affymetrix Carlsbad, CA, USA) according to standard methods, GeneChips, with segmentation analysis performed and RNA quality was checked using a Bioanalyzer using Microarray Analysis Suite 5.0 (MAS5). Normal- (Agilent, Santa Clara, CA, USA). Reverse transcrip- ization of data was performed using GC-RMA.45 Data tion was performed using the QuantiTect Reverse were converted to a linear scale by log2 transformation, kit (QIAGEN, Hilden, Germany), with gene clustering performed using Genes@Work employing random hexamers as primers. The amount software.46 Gene set analysis was performed using Gene of each specific cDNA was measured using an iQ5 Set Enrichment Analysis (GSEA) software (version 2)47 Multicolor Real-Time PCR Apparatus according to a with array probe sets collapsed to gene symbol and gene standard DDCt SYBR Green protocol (BioRad, Hercules, sets generated from Biocarta gene set list (converted CA, USA).43 CyclophilinA cDNA was measured in from probe set list with manual curation), using ASD parallel with both PRKCB1 isoforms and used as a versus control phenotypes and 1000 permutations/ standard normalizer. The following isoform-specific analysis. Statistical significance was calculated using primers were designed: PRKCB1-1F AGACACCTCCA apairedStudent’st-test. False discovery rate threshold ACTTCGACAA; PRKCB1-1R CAACGATGGAGTTTG in GSEA was set at q < 0.05.48 Correlation analyses were CATTC; PRKCB1-2F AAGCTCAACGGCTATTGTGG; performed by Pearson r. PRKCB1-2R GCCATCTGCATAATCCCATC; cyclophi- linA-F GCAGACAAGGTCCCAAAG; cyclophilinA-R Results GAAGTCACCACCCTGACAC. Statistical significance was calculated using the Wilcoxon test. PRKCB1 gene variants are associated with autism The four SNPs, located in intron 2 of the PRKCB1 Western blotting locus, are in Hardy–Weinberg equilibrium both in the Western blotting was performed on 8 of the 11 case– entire data set of 795 individuals (Supplementary control pairs (no. 1–8 in Table 3) because of limited Table S1), and also when analyzing separately 239 tissue availability. Sixty micrograms of protein extract autistic patients, 234 mothers, 232 fathers, and 90 were diluted in 2 Â SDS/loading sample buffer, unaffected siblings. The same four SNPs are also in boiled for 5 min, loaded onto a 5% stacking/7.5% LD, as summarized in Table 2B; their LD pattern in separating polyacrylamide gel, run at 100 V for 2 h, our sample is superimposable to the LD pattern found and transferred onto a nitrocellulose membrane at by the International HapMap Project in their Cauca- 350 mA for 90 min. Membranes were blocked with 5% sian CEPH population (that is, Utah residents with milk in 0.1% Tris-buffered saline Tween-20 (TBS-T) ancestry from northern and western Europe; see for 1 h, washed 3 Â 10 min in 0.01% TBS-T, incubated www.hapmap.org). overnight at 4 1C with primary antibody against PRKCB1 haplotypes display a significant associa- PKCbII (sc-13149, Santa Cruz Biotechnology, Santa tion with autism, as shown in Table 4A (whole marker Cruz, CA, USA) diluted 1:600 in 0.05% TBS-T with permutation test, P < 0.05 after 32271 iterations). 5% milk. Following 3 Â 10 min washes in 0.01% TBS- Haplotype 2-2-2-1 is transmitted from heterozygous T, membranes were incubated for 1 h at room parents to their autistic offspring significantly more temperature with horseradish peroxidase-conjugated often than expected by chance (P = 0.007), at the goat antimouse IgG (sc-2005, Santa Cruz Biotechno- expense of the complementary haplotype 1-1-1-2 logy) diluted 1:5000 in 0.1% TBS-T and 3% milk. (Table 4A). Unaffected siblings display the opposite Following 3 Â 10 min washes in 0.01% TBS-T, pro- trend, with an undertransmission of haplotype 2-2-2- teins were visualized using the enhanced chemilu- 1 reaching a borderline P = 0.053 (Supplementary minescent (ECL) detection method and Hyperfilm Table S2) and yielding a highly significant difference (Amersham Biosciences, Piscataway, NJ, USA). Den- in transmission rates between autistic and non- sitometry was performed using the VersaDoc model autistic offspring (Z = 16.23, P<0.00001). Despite 4000 imaging system (BioRad). their lower statistical power, single-marker FBAT To control for protein loading, membranes were analyses also support the existence of LD between washed 3 Â 10 min in 0.01% TBS-T, and incubated the most 50 SNP, rs3785392 and the PRKCB1 with mouse AC15 anti-b-actin monoclonal antibody gene variant contributing to autism vulnerability (A5441, Sigma-Aldrich, St Louis, MO, USA) diluted (Table 4B). Superimposable trends in allelic transmis- 1:5000, and with goat antimouse IgG (AP124P, sion rates are found in Italian and Caucasian- Chemicon International, Temecula, CA, USA) diluted American families analyzed separately. No parent- 1:5000. Data are presented as PKCbII normalized to of-origin effect is consistently present (data not b-actin (mean±s.d.). Statistical significance was shown). Preliminary analyses performed on clinical determined using the Wilcoxon test. signs and symptoms suggest a possible influence of PRKCB1 alleles on the ‘presence/absence of verbal Microarray analysis or motor stereotypies at intake,’ with the ‘risk’ allele Total RNA isolation, reverse transcription and in vitro significantly blunting stereotypic behaviors. In fact, transcription were performed, as previously described.44 stereotypies were noticed at intake in 17/26 (65.4%)

Molecular Psychiatry PRKCB1 gene and autistic disorder C Lintas et al 711 Table 4 The PRKCB1 gene variant marked by haplotype 2-2-2-1 is associated with autism: (A) haplotype and (B) single-marker family-based association tests performed using FBAT, under an additive model (Àe)38

4-Marker Haplotypesa Estimated frequency N. of families S E(S) Var(S) Z P-value

(A) 2-2-2-2 0.388 128.0 164.218 156.899 50.805 1.027 0.304486 1-1-1-2 0.258 125.3 91.847 105.174 48.876 À1.906 0.056616 1-1-1-1 0.092 58.5 44.434 39.090 14.703 1.394 0.163402 2-2-1-2 0.070 45.2 25.699 27.726 13.405 À0.554 0.579796 1-1-2-2 0.062 43.2 21.570 25.451 11.209 À1.159 0.246324 2-2-2-1 0.052 37.4 32.083 23.042 11.549 2.660 0.007803 1-2-1-2 0.040 32.0 15.666 17.000 7.778 À0.478 0.632465 2-1-2-2 0.017 13.0 8.980 7.490 4.740 0.684 0.493720

(B) Marker/alleles Autistic patients

N. of families S E(S) Var(S) Z P-value

rs3785392 1 146 124.000 139.500 58.250 À2.031 0.042573 2 146 172.000 156.500 58.250 2.031 rs3785387 1 141 126.000 135.500 60.000 À1.162 0.245278 2 141 160.000 151.000 60.000 1.162 rs196002 1 135 111.000 124.000 62.250 À1.711 0.0870071 2 135 163.000 149.500 62.250 1.711 rs1873423 1 90 67.000 58.500 29.250 1.572 0.116032 2 90 119.000 127.000 29.250 À1.572

FBAT output variables: S, test statistic (that is, genotypic distribution in the offspring conditioned on affection status and parental genotypes); E(S), expected value for S; VAR(S), variance of S under the null hypothesis of no linkage and no association; N, number of informative nuclear families P-values < 0.05 are highlighted in bold. aWhole marker permutation test, P < 0.05 after 32271 iterations. patients with the 1/1 genotype at rs3785392, in 33/67 evidence of PRKCB1 gene variants affecting peptiduria (49.3%) 1/2 individuals and in 14/47 (29.8%) 2/2 or any other quantitative endophenotype could be individuals (w2 = 9.196, 2 d.f., nominal P = 0.01); found in the unaffected siblings of autistic patients for rs3785387, stereotypies were recorded at intake (Supplementary Table S4). in 17/25 (68.0%) 1/1 patients, 30/64 (46.9%) 1/2 Incorporating peptiduria as a covariate into haplo- individuals and 18/51 (35.3%) 2/2 individuals typic and single-marker analyses for affection status (w2 = 7.224, 2 d.f., nominal P = 0.03). essentially did not change their outcome: the whole marker permutation test for haplotypic analysis yielded a P = 0.054, while single-marker analyses for PRKCB1 gene variants are associated with enhanced rs3785392 and rs3785387 yielded P = 0.02 and 0.11, oligopeptiduria in autism respectively (compare with Tables 4A and 4B). These Macrocephaly/macrosomy, hyperserotoninemia and results indicate that the overtransmission of PRKCB1 enhanced peptiduria are among the best-characterized gene variants to autistic offspring is largely, though morphological and biochemical endophenotypes not entirely, independent of peptiduria. in autism.6,10,36 Performing a quantitative TDT, as implemented by the FBAT program,38 we found a significant association between parent-to-autistic PRKCB1 gene expression is significantly decreased offspring transmission rates of PRKCB1 gene variants in postmortem brains of autistic patients marked by the two most 50 SNPs, rs3785392 PRKCB1 gene expression is significantly decreased and rs3785387, and urinary peptide excretion rates in the temporal neocortex of ASD patients compared (Table 5and Supplementary Table S3). ANOVAs to their matched controls. PRKCB1 mRNA levels are based on genotype distributions provide further decreased by 26% according to oligonucleotide DNA evidence of semidominant effects for ‘risk’ allele 2 at microarray analysis (pairwise t- and rank-tests SNP rs3785387, associated with enhanced peptiduria P < 0.05). Using isoform-specific primers, quantitative (genotype 1/1, n = 20, 218.5±251 mm2; 1/2, n = 59, PCR unveils 35 and 31% mean reductions for the 277.0±220 mm2; 2/2, n = 43, 353.0±208 mm2; Kruskal– PRKCB1-1 and PRKCB1-2 isoforms, yielding P < 0.01 Wallis w2 = 6.083, 2 d.f., P<0.05). In contrast, no and < 0.05, respectively (Figure 1). In particular,

Molecular Psychiatry PRKCB1 gene and autistic disorder C Lintas et al 712 Table 5 Quantitative transmission/disequilibrium test for single nucleotide polymorphisms rs3785392 and rs3785387 on head circumference (HC), serotonin blood levels (5-HT) and urinary peptide excretion rates (PEPT) using FBAT, under an additive model (option Àe).37

Quantitative traits Marker/alleles Autistic patients

N. of families S E(S) Var(S) Z P-value

HC rs3785392 1 51 4333.000 4943.500 153049.250 À1.561 0.118636 2 51 6039.000 5428.500 153049.250 1.561 rs3785387 1 49 4494.000 4832.500 139799.750 À0.905 0.365292 2 49 5600.000 5261.500 139799.750 0.905 5-HT rs3785392 1 37 9458.000 10800.500 865583.750 À1.443 0.149027 2 37 10614.000 9271.000 865583.750 1.443 rs3785387 1 38 9811.000 10782.500 906645.750 À1.020 0.307591 2 38 11843.000 10871.000 906645.750 1.020 PEPT rs3785392 1 57 19704.000 25455.000 6414657.000 À2.271 0.0231661 2 57 36372.000 30621.000 6414657.000 2.271 rs3785387 1 58 21112.000 26607.500 7414007.750 À2.018 0.043562 2 58 36846.000 31350.500 7414007.750 2.018

E(S), expected value for S; N, number of informative nuclear families; S, FBAT test statistic; (i.e., genotypic distribution in the offspring conditioned on affection status and parental genotypes); VAR(S), variance of S under the null hypothesis of no linkage and no association. P-values < 0.05 are highlighted in bold. See Table S3 for SNPs rs196002 and rs1873423.

tive PCR. A consistent decrease in PKCbII levels was found in seven out of eight ASD individuals, compared to their matched controls (Figure 2).

‘Risk’ alleles at SNPs rs3785392 and rs3785387 are associated with an inability to downregulate PRKCB1 gene expression in postmortem brains of autistic patients Genomic DNA was extracted from the same neo- cortical specimens used to assess PRKCB1 gene expression and the same four SNPs were genotyped (Supplementary Table S5). PRKCB1-1 and PRKCB1-2 mRNA levels were then correlated with genotypes at SNPs rs3785392 and rs3785387, previously found Figure 1 Significant decrease in PRKCB1 gene expression associated with autism and/or peptiduria (see above). in the temporal neocortex of 11 autism-spectrum disorder Controls homozygous for ‘risk’ allele 2 at both SNPs (ASD) patients, relative to their matched controls. Levels of displayed small, nonsignificant increases in PRKCB1 mRNA encoding the PRKCB1-1 and PRKCB1-2 isoforms, gene expression levels for both isoforms (Figure 3). assessed by qPCR using isoform-specific primers, are Instead, autistic patients carrying the 2/2 genotype equally decreased. Bars indicate mean values; the hyphe- had mRNA levels comparable to those of controls, nated line corresponds to no case–control difference in whereas patients with the 1/1 or 1/2 genotypes under- mRNA levels. *P < 0.05 and **P < 0.01 for patients versus went a profound and significant downregulation of controls. gene expression (Mann–Whitney U tests, P < 0.05; Figure 3). These results indicate that homozygosity for ‘risk’ allele 2 at SNPs rs3785392 and rs3785387 is mRNA levels are decreased by 25% or more in seven associated with an inability to downregulate neo- ASD patients compared to controls, whereas the cortical PRKCB1 gene expression in autistic brains. remaining four patients displayed comparable levels These genotype–phenotype correlations certainly (Figure 1). Given the superimposable trends dis- did not stem from a skewed distribution of confound- played by the two mRNA isoforms, decreased expres- ing factors, such as MR or seizures (that is, history of sion patterns were confirmed at the protein level seizures, seizure-related death or use of antiepileptic assessing the PKCbII isoform only. Mean PKCbII drugs at the time of death; Table 3) because ‘risk’ amounts measured using Western blot and normal- alleles at SNPs rs3785392 and rs3785387 were evenly ized by b-actin were found decreased by 35% distributed among patients with and without MR or (P = 0.05), a result entirely superimposable to the seizures. Furthermore, risk alleles were equally decrease recorded at the mRNA level using quantita- associated with higher PRKCB1 mRNA levels in the

Molecular Psychiatry PRKCB1 gene and autistic disorder C Lintas et al 713

Figure 2 Decreased PKCbII protein levels in temporocortical samples of autism-spectrum disorder (ASD) individuals compared to matched controls. (a) Representative Western blots of the eight ASD case–control pairs for PKCbII and b-actin; (b) quantitative analyses of PKCbII protein levels normalized to b-actin, used as loading control. Bars indicate mean values, which differ with a P = 0.05.

Figure 3 Autistic patients (blue columns) homozygous for risk allele 2 at single nucleotide polymorphisms (SNPs) rs3785392 (N = 4) and rs3785387 (N = 3) show a statistically significant inability to downregulate PRKCB1-1 and PRKCB1-2 gene expression, as compared to patients carrying the 1/1 or 1/2 genotypes (N = 7 and 8, respectively). Controls (pink columns) carrying the 2/2 genotype at SNPs rs3785392 (N = 3) and rs3785387 (N = 2) display a nonsignificant trend. See Supplementary Table S5 for complete genotype list. *P < 0.05. presence of either seizures or MR (data not shown). and 4 for 2/2 or other genotype carriers, respectively). However, the downregulation of PRKCB1 gene expres- These analyses must be, however, viewed as merely sion may be especially pronounced in autistic exploratory, because no contrast here reaches statis- patients with MR and/or with seizures. For example, tical significance due to small sample sizes. among five patients with seizures, three individuals carrying the 1/1 or 1/2 genotypes at SNP rs3785392 Disrupted coordination of brain gene expression displayed PRKCB1-2 mRNA levels 61.3% lower patterns involving PRKCB1 in autism compared to two patients carrying the 2/2 genotype; We hypothesized that deficits in PKCb expression among patients without a history of seizures, this would putatively lead to a dysregulation in PKCb- same mean difference was reduced to 21.5% (n =2 dependent transcript networks and to significant

Molecular Psychiatry PRKCB1 gene and autistic disorder C Lintas et al 714 downstream alterations in neuronal function. To test expression patterns in postmortem neocortical tissue this hypothesis, we explored the temporocortical samples of autistic patients compared to age-, sex-, expression of PRKCB1-related transcript networks in and postmortem interval-matched controls and (d) autistic patients and matched control subjects using identifies a blunted downregulation of PRKCB1 gene oligonucleotide DNA microarray analysis. First, the expression in postmortem brains as the functional microarrays employed in this study encompass two correlate of PRKCB1 ‘risk’ alleles associated with independent probe sets for PRKCB1 (227817_at and autism. These functional abnormalities in PRKCB1 209685_s_at), which consistently demonstrated gene expression nicely parallel genetic evidence of superimposable expression differences between post- PRKCB1 gene contributions to autism pathogenesis mortem autistic and control brains (ALR = À0.44 and and lend further support to PKCb involvement in the À0.46, respectively). In addition, the two probe sets pathogenetic processes underlying autistic disorder. displayed high positive correlations (r = 0.98 and 0.86 Our family-based linkage/association study essen- for autistics and controls, respectively), further tially confirms the existence of a significant asso- validating our findings. ciation between autism and PRKCB1 gene variants Next, in a data-driven approach, we identified 1009 marked in Caucasians by polymorphisms located in gene transcripts that were highly coregulated with intron 2. The C-G-T haplotype found associated with PKCb in the brain of control subjects (r > 0.9). In autism in Caucasians by Philippi et al.14 is identical to addition, we identified 131 transcripts that (a) our risk haplotype for rs3785387 and rs196002,

displayed correlation differences of DrAUTÀCONX0.8 whereas it differs at SNP rs1873423. Considering the 0 0 with PRKCB1 between patients and controls (rAUTÀ progressive decrease of LD going from 5 to 3 in this rCON) and (b) reported a significant expression region (Table 2B), and the much stronger association difference between the autistics and control post- of 50 markers both with autism and with oligopepti- mortem temporal cortices (|ALR| > 0.585 and P < 0.05 duria in this study (Table 4 and Supplementary Table in pairwise t-test). Among these, 71 genes showed a S3), it seems likely that haplotypes associated with positive correlation with PRKCB1 levels in ASD cases autism in our sample and in Caucasian-Americans14 and an opposite or unchanged trend in controls, are essentially marking the same PRKCB1 gene vari- including the kainate 2 ionotropic glutamate receptor ant. Moreover, a comparison of LD patterns present (GRIK2) and the neuronal/epithelial high affinity in our sample and in the 171 Irish families assessed glutamate transporter (SLC1A1). Even more interest- by Yang et al.32 fosters a more cautious interpretation ingly, 60 mRNAs that were positively correlated of their allegedly negative association findings. with PRKCB1 levels in controls showed a loss of the SNPs rs3785387, rs196002 and rs1873423 display positive correlation with PRKCB1 levels in ASD cases. here D0 values ranging from 0.285 to 0.654 (Table 2B), Among others, these included genes encoding whereas D0 values range only from 0.045 to 0.195 in multiple neurodevelopmentally relevant molecules the Irish sample.32 These population genetic differen- (for example, FGF2 and FGF-R1) and cytokines (for ces can alone explain why polymorphisms employed example, TGFb2) (Figure 4). in this and in previous studies14,32 are apt to mark the Finally, we explored a knowledge-based correlation autism liability conferring PRKCB1 gene variant in between mRNA levels of PRKCB1 and genes the Italian and Caucasian-American populations, but whose expression has been shown to be under direct are uninformative in the Irish population. or indirect PKCb control.23,24,26,28–31 These results Our postmortem study has conclusively demons- revealed that coordinated expression patterns are trated a significant decrease in PRKCB1 gene expres- lost for several cytokines and enzymes involved in sion (Figures 1 and 2). Potential confounding factors oxidative stress, maintained to a large extent for are inevitably intermingled with autism in our post- others, and entirely preserved for extracellular matrix mortem sample, as they are at the phenotypic level. proteins (Table 6). In addition to the loss of corre- Approximately 75% of autistic patients indeed have lation with PKCb transcript levels, many of these MR and 25% a positive history of seizures.2 In our genes also displayed significant expression differ- postmortem sample, MR and seizures could modulate ences between autistic and control brains, including PRKCB1 allelic effects on gene expression, but do not TGFb2, MCP1, collagen IV and fibronectin 1 (Table 6). appear to produce a spurious association. Instead, this significant downregulation in gene expression is interestingly associated with ‘normal’ PRKCB1 Discussion alleles, not with alleles conferring autism liability. This study (a) replicates in an independent sample A thorough mutational search for non-conservative, the positive association between PRKCB1 gene coding mutations and for splice junction variants variants and autism previously reported by Philippi previously provided negative results.14 Recently, a et al.;14 (b) extends family-based genetic findings 593 kb region located in human ch. 16p11.2 has been by demonstrating a significant association also with shown to undergo a recurrent microdeletion or a a biochemical endophenotype characterized by en- reciprocal microduplication in approximately 1% of hanced oligopeptiduria; (c) provides novel evidence autistic patients.49,50 These genomic rearrangements, showing significantly reduced PRKCB1 mRNA and typically de novo but in some cases also inherited, are protein levels, as well as disrupted PKCb-related gene unlikely to explain our expression and association

Molecular Psychiatry PRKCB1 gene and autistic disorder C Lintas et al 715

Figure 4 Correlation of PRKCB1 mRNA expression with levels of mRNAs encoding the glutamate receptor GRIK2, the neuronal glutamate transporter SLC1A1, FGF2 and TGFb2, in six autistic patients (red) and matched controls (green). Pearson r coefficients and ALR values are reported. findings: they are 5 Mb centromeric to the PRKCB1 regulation of gene expression seemingly fits with locus, their low frequency could hardly skew trans- several converging lines of evidence. First, PRKCB1 mission rates to the extent encountered in our asso- gene expression was strongly decreased in several ciation study and, most importantly, we found no patients, and particularly in those carrying one or two evidence of genomic rearrangements involving the ‘normal’ PRKCB1 alleles (Figures 1–3). Interestingly, PRKCB1 gene itself in five of the same brain speci- PKCb downregulation produced by either homolo- mens employed in this study and displaying gous recombination in PRKCB1 gene knockout mice here variable decreases in PRKCB1 gene expression or by selective pharmacological blockade is asso- (R Sacco, S Levy, P Levitt, AM Persico et al., article in ciated with an immunosuppression, encompassing preparation). Therefore, this PRKCB1 ‘risk’ allele most decreased monocyte recruitment and migration, redu- likely consists in an intronic variant that inactivates a ced oxidative stress and blunted gene expression for negative-feedback loop on gene expression. The exact chemokines such as ICAM-1, MCP-1 and TGFb.29–31,53 mechanism underlying this allelic effect is currently Conversely, several in vitro models indicate that the being investigated. In general, it is not unusual for selective activation of both PKCb isoforms is critical large introns, such as intron 2 of the PRKCB1 gene, to sustain oxidative stress.54,55 These proinflammatory which spans approximately 150 kb, to produce totally effects of PKCb sharply contrast with the decrease in intronic noncoding RNAs, especially frequent among PRKCB1 gene expression reported here, because the members of the ‘Regulation of trans- very same neocortical brain samples employed in this cription’ gene category.51 Alternatively, single base study display enhanced oxidative stress (L Palmieri, pair changes can indeed affect gene expression by AM Persico et al., submitted) and increased immune creating or deleting an intronic response element, gene expression.56 The latter results are in agreement enhancer or repressor,52 as recently demonstrated for with previous studies demonstrating enhanced oxida- a single change causing the loss of an Sp1 tive stress in autism57 and an ongoing immune response element in the MET gene promoter.4 Regard- reaction in a set of postmortem brain tissue samples less of the underlying mechanism, the association of of autistic patients largely non-overlapping with ours, PRKCB1 ‘risk’ alleles with a blunted adaptive down- and in the CSF of autistic children collected in vivo.58

Molecular Psychiatry PRKCB1 gene and autistic disorder C Lintas et al 716 Table 6 Pearson correlation coefficients for PRKCB1 and excretion rates increases confidence in the reliability genes whose expression is either directly or indirectly of our results and spurs further interest in the under- modulated by PKCb,23,24,26,28–31 assessed in autistic patients lying pathophysiology. Functional PRKCB1 promoter (rAUT) and in matched controls (rCON); differences in mRNA variants have been shown to be associated with levels between cases and controls, expressed as difference in diabetic nephropathy and albuminuria,60,61 mediated average log2 ratios [ALR (aut–con)], and relative paired and rank P-values; coordinated expression patterns are (A) lost for by increased renal oxidative stress, production of several cytokines and enzymes involved in oxidative stress, fibrotic cytokines such as TGFb, and consequently 29–31 (B) maintained to some extent for others and (C) largely increased glomerular permeability. Also of inter- preserved for extracellular matrix molecules est is the potential gene–gene interaction with APOE e2 alleles, which contribute to the development rAUT rCON ALR Prpval RANKpval and progression of diabetic nephropathy62,63 and are (aut-con) preferentially transmitted to autistic children in our sample.64 A potential parallel between ‘leaky kidney’ A and ‘leaky gut’ syndromes is suggested by an in vitro CTGF 0.51 À0.10 0.006 0.982 0.948 study demonstrating that PKCb and are both NADPH 0.81 À0.39 À0.005 0.979 0.879 necessary for oxidative stress to yield abnormal oxidase permeability in monolayers of human intestinal TGFb2 À0.75 0.25 0.908 0.049 0.045 26 65 VEGF À0.19 0.63 0.768 0.119 0.114 Caco-2 cells. This effect, linking calcium, oxida- tive stress,57 and PKCb to the stability of epithelial B cytoskeleton in the gut, could conceivably explain the ICAM1 0.26 0.50 0.040 0.608 0.732 abnormally elevated intestinal permeability present MCP1 À0.29 À0.42 3.213 0.047 0.026 in many autistic patients and the subsequent activa- TGFb1 À0.37 À0.52 0.658 0.193 0.247 tion of the immune system due to the absorption Endothelin-1 À0.29 À0.22 0.818 0.052 0.067 of the first foreign dietary peptides recognized by the Cox2 À0.01 À0.40 1.161 0.193 0.124 gut immune system as ‘non-self’, namely casein and gluten.7 Our current phenotypic data set does C not allow us to directly test this hypothesis. None- Collagen IV À0.58 À0.51 1.426 0.040 0.036 Collagen VI À0.87 À0.89 0.650 0.071 0.097 theless, our sample encompasses 106 families with Fibronectin 1 À0.26 0.14 0.468 0.039 0.055 mixed genotypes at SNP rs3785392, 7 families whose members all carry the ‘normal’ 1/1 genotype, and 10 families entirely composed of 2/2 ‘risk genotype’ Many genes expressing chemokines are either directly carriers. A positive history of ‘food allergies’ in or indirectly upregulated by PKCb-induced phosphor- unaffected family members of the autistic proband ylation:29–31,53 the decrease in PRKCB1 gene expres- was self-reported in 11/106 (11.6%) ‘mixed genotype’ sion documented in our study, and the increased families, 1/7 (14.3%) ‘normal genotype’ families and expression of immune genes like TGFb2 (Figure 4) 4/10 (40.0%) ‘risk genotype’ families, respectively clearly argue against PRKCB1 driving this ongoing (w2 = 7.097, 2 d.f., P < 0.05). This analysis is merely immune activation in patient samples. On the con- suggestive, but spurs interest into specific hypothesis- trary, a compensatory decrease in PRKCB1 gene driven studies aimed at determining whether there is expression occurring in ASD patients to buffer an a pathophysiological overlap between the PKCb- inappropriate activation of the immune system see- dependent processes responsible for a ‘leaky kidney’ mingly represents the most plausible scenario. In and the mechanisms contributing to the ‘leaky gut’ the latter case, any genetic variant reducing the syndrome in autism. sensitivity of PRKCB1 gene expression to this nega- In summary, this study confirms the existence of tive-feedback loop would result in an even more a genetic association between PRKCB1 gene variants persistent immune activation, sustaining oxidative and autism, demonstrates a significant association stress, and fostering developmental damage during between PRKCB1 gene variants and the quantitative early pregnancy. It will thus be extremely important endophenotype of oligopeptiduria, shows a promi- to identify the exact sequence variant conferring nent reduction in PRKCB1 gene expression in the autism vulnerability and to determine its functional temporal neocortex of autistic patients compared to correlates in vitro. Only then it will be possible to tie matched controls, demonstrates that alleles confer- the current results into the broader picture of autism ring autism vulnerability are associated with a lack pathogenesis, also because, despite its intended of downregulation in PRKCB1 gene expression, adaptive value, the decrease in PRKCB1 gene expres- and provides evidence of a partially uncoordinated sion recorded in most autistic brains could still play a expression of PRKCB1 and several PKCb-dependent negative role, for example by blocking the physio- genes involved in neurodevelopment and immune logical relocation of both FMRP and Fmr1 mRNA functions. These expression changes can currently be in dendrites following activation of metabotropic best interpreted as a compensatory adjustment aimed glutamate receptors.59 at hampering a dysregulated immune response affect- The significant association recorded here between ing the CNS of autistic patients. The negative-feed- PRKCB1 gene variants and enhanced urinary peptide back loop mediating this compensatory adjustment is

Molecular Psychiatry PRKCB1 gene and autistic disorder C Lintas et al 717 seemingly inactivated in PRKCB1 alleles conferring 12 Miller MT, Stromland K, Ventura L, Johansson M, Bandim JM, autism vulnerability. The identification of this Gillberg C. Autism associated with conditions characterized by PRKCB1 DNA sequence variant is being actively developmental errors in early embryogenesis: a mini review. Int J Dev Neurosci 2005; 23: 201–219. pursued, to achieve a thorough understanding of the 13 Teitelbaum O, Benton T, Shah PK, Prince A, Kelly JL, Teitelbaum functional implications of PKCb in autism. P. Eshkol-Wachman movement notation in diagnosis: the early detection of Asperger’s syndrome. Proc Natl Acad Sci USA 2004; 101: 11909–11914. Acknowledgments 14 Philippi A, Roschmann E, Tores F, Lindenbaum P, Benajou A, Germain-Leclerc L et al. Haplotypes in the gene encoding protein We gratefully acknowledge Laura Gaita and Rosanna kinase c-beta (PRKCB1) on are associated with D’Oronzio for technical assistance, Francis Rousseau autism. Mol Psychiatry 2005; 10: 950–960. and Anne Philippi (IntegraGen SA, Evry, France) for 15 Bult A, Kobylk ME, Van der Zee EA. Differential expression of protein kinase C betaI (PKCbetaI) but not PKCalpha and PKCbetaII in the genotyping and the population stratification the suprachiasmatic nucleus of selected house mouse lines, and the analysis, Carlo Lenti and Roberto Rigardetto for relationship to arginine-vasopressin. Brain Res 2001; 914: 123–133. contributing to patient recruitment, all the patients 16 Ashique AM, Kharazia V, Yaka R, Phamluong K, Peterson AS, and families who generously contributed to these Ron D. Localization of the scaffolding protein RACK1 in the developing and adult mouse brain. Brain Res 2006; 1069: 31–38. studies, and the Autism Tissue Program, Harvard 17 Wu J, Song TB, Li YJ, He KS, Ge L, Wang LR. Prenatal restraint Brain Tissue Resource Center, and Maryland NICHD stress impairs learning and memory and hippocampal PKCbeta1 Brain Tissue Center for providing the brain tissue expression and translocation in offspring rats. Brain Res 2007; samples. This work was supported by the Italian 1141: 205–213. Ministry for University, Scientific Research 18 Colombo PJ, Wetsel WC, Gallagher M. Spatial memory is related to hippocampal subcellular concentrations of calcium-dependent and Technology (Programmi di Ricerca di Interesse protein kinase C isoforms in young and aged rats. Proc Natl Acad Nazionale, prot. no.2006058195), the National Sci USA 1997; 94: 14195–14199. Alliance for Autism Research (Princeton, NJ), the 19 Birikh KR, Sklan EH, Shoham S, Soreq H. Interaction of ‘read- Cure Autism Now Foundation (Los Angeles, CA) and through’ acetylcholinesterase with RACK1 and PKCbII correlates with intensified fear-induced conflict behavior. Proc Natl Acad Sci the Fondation Jerome Lejeune (Paris, France) to AMP, USA 2003; 100: 283–288. and by VUKC Startup Fund, R01 MH079299, and K02 20 Kang SW, Wahl MI, Chu J, Kitaura J, Kawakami Y, Kato RM et al. MH070786 to KM. PKCb modulates antigen receptor signaling via regulation of Btk membrane localization. EMBO J 2001; 20: 5692–5702. 21 Long A, Kelleher D, Lynch S, Volkov Y. Cutting edge: protein References kinase Cb expression is critical for export of Il-2 from T cells. J Immunol 2001; 167: 636–640. 1 American Psychiatric Association. Diagnostic and Statistical 22 Volkov Y, Long A, McGrath S, Ni Eidhin D, Kelleher D. Crucial Manual of Mental Disorders, 4th edn. American Psychiatric importance of PKCbI in LFA-1-mediated locomotion of activated Association: Washington, DC, 1994. T cells. Nat Immunol 2001; 2: 508–514. 2 Persico AM, Bourgeron T. Searching for ways out of the autism 23 Cejas PJ, Carlson LM, Zhang J, Padmanabhan S, Kolonias D, maze: genetic, epigenetic and environmental clues. Trends Lindner I et al. Protein kinase C bII plays an essential role in Neurosci 2006; 29: 349–358. dendritic cell differentiation and autoregulates its own expression. 3 Piven J, Palmer P, Jacobi D, Childress D, Arndt S. Broader autism J Biol Chem 2005; 280: 28412–28423. phenotype: evidence from a family history study of multiple- 24 Siow YL, Au-Yeung KK, Woo CW, Karmin O. Homocysteine incidence autism families. Am J Psychiatry 1997; 154: 185–190. stimulates of NADPH oxidase p47phox and 4 Campbell DB, Sutcliffe JS, Ebert PJ, Militerni R, Bravaccio C, p67phox subunits in monocytes via protein kinase Cb activation. Trillo S et al. A genetic variant that disrupts MET transcription Biochem J 2006; 398: 73–82. is associated with autism. Proc Natl Acad Sci USA 2006; 103: 25 Ma HT, Lin WW, Zhao B, Wu WT, Huang W, Li Y et al. Protein 16834–16839. kinase C b and d isoenzymes mediate cholesterol accumulation 5 Campbell DB, D’Oronzio R, Garbett K, Ebert PJ, Mirnics K, Levitt P in PMA-activated macrophages. Biochem Biophys Res Commun et al. Disruption of cerebral cortex METsignaling in autism 2006; 349: 214–220. spectrum disorder. Ann Neurol 2007; 62: 243–250. 26 Banan A, Fields JZ, Zhang Y, Keshavarzian A. Key role of PKC and 2 þ 6 Sacco R, Militerni R, Frolli A, Bravaccio C, Gritti A, Elia M et al. Ca in EGF protection of microtubules and intestinal barrier Clinical, morphological, and biochemical correlates of head against oxidants. Am J Physiol Gastrointest Liver Physiol 2001; circumference in autism. Biol Psychiatry 2007; 62: 1038–1047. 280: G828–G843. 7 D’Eufemia P, Celli M, Finocchiaro R, Pacifico L, Viozzi L, 27 Liu Y, Su W, Thompson EA, Leitges M, Murray NR, Fields AP. Zaccagnini M et al. Abnormal intestinal permeability in children Protein kinase CbII regulates its own expression in rat intestinal with autism. Acta Paediatr 1996; 85: 1076–1079. epithelial cells and the colonic epithelium in vivo. J Biol Chem 8 Wakefield AJ, Ashwood P, Limb K, Anthony A. The significance of 2004; 279: 45556–45563. ileo-colonic lymphoid nodular hyperplasia in children with 28 Yu W, Murray NR, Weems C, Chen L, Guo H, Ethridge R et al. autistic spectrum disorder. Eur J Gastroenterol Hepatol 2005; 17: Role of cyclooxygenase 2 in protein kinase C bII-mediated colon 827–836. carcinogenesis. J Biol Chem 2003; 278: 11167–11174. 9 Jyonouchi H, Geng L, Ruby A, Zimmerman-Bier B. Dysregulated 29 Koya D, Jirousek MR, Lin YW, Ishii H, Kuboki K, King GL. innate immune responses in young children with autism spectrum Characterization of protein kinase C beta isoform activation on the disorders: their relationship to gastrointestinal symptoms and gene expression of transforming growth factor-beta, extracellular dietary intervention. Neuropsychobiology 2005; 51: 77–85. matrix components, and prostanoids in the glomeruli of diabetic 10 Reichelt WH, Knivsberg AM, Nodland M, Stensrud M, Reichelt KL. rats. J Clin Invest 1997; 100: 115–126. Urinary peptide levels and patterns in autistic children from seven 30 Wu Y, Wu G, Qi X, Lin H, Qian H, Shen J et al. Protein kinase C b countries, and the effect of dietary intervention after 4 years. Dev inhibitor LY333531 attenuates intercellular adhesion molecule-1 Brain Dysfunct 1997; 10: 44–55. and monocyte chemotactic protein-1 expression in the kidney in 11 Bauman ML, Kemper TL. Neuroanatomic observations of the brain diabetic rats. J Pharmacol Sci 2006; 101: 335–343. in autism: a review and future directions. Int J Dev Neurosci 2005; 31 Ohshiro Y, Ma RC, Yasuda Y, Hiraoka-Yamamoto J, Clermont AC, 23: 183–187. Isshiki K et al. Reduction of diabetes-induced oxidative stress,

Molecular Psychiatry PRKCB1 gene and autistic disorder C Lintas et al 718 fibrotic cytokine expression, and renal dysfunction in protein 49 Kumar RA, Karamohamed S, Sudi J, Conrad DF, Brune C, Badner JA kinase Cbeta-null mice. Diabetes 2006; 55: 3112–3120. et al. Recurrent 16p11.2 microdeletions in autism. Hum Mol Genet 32 Yang MS, Cochrane L, Conroy J, Hawi Z, Fitzgerald M, Gallagher L 2007; 17: 628–638. et al. Protein kinase C-b1 gene variants are not associated with 50 Weiss LA, Shen Y, Korn JM, Arking DE, Miller DT, Fossdal R autism in the Irish population. Psychiatr Genet 2007; 17: 39–41. et al. Association between microdeletion and microduplication at 33 Conciatori M, Stodgell CJ, Hyman SL, O’Bara M, Militerni R, 16p11.2 and autism. N Engl J Med 2008; 358: 667–675. Bravaccio C et al. Morphogenetic effect of the HOXA1 A218G 51 Nakaya HI, Amaral PP, Louro R, Lopes A, Fachel AA, Moreira YB polymorphism on head circumference in patients with autism. et al. Genome mapping and expression analyses of human intronic Biol Psychiatry 2004; 55: 413–419. noncoding RNAs reveal tissue-specific patterns and enrichment 34 Lord C, Rutter M, DiLavore PC, Risi S. ADOS, Autism Diagnostic in genes related to regulation of transcription. Genome Biol 2007; Observation Schedule. Western Psychological Services: Los Angeles, 8: R43. 2002 (Italian version ed. by Tancredi R, Saccani M, Persico AM, 52 Lin CY, Vega VB, Thomsen JS, Zhang T, Kong SL, Xie M et al. Parrini B, Igliozzi R and Faggioli R. Organizzazioni Speciali: Whole-genome cartography of estrogen receptor alpha binding Florence, 2005). sites. PLoS Genet 2007; 3: e87. 35 Rutter M, Le Couter A, Lord C. ADI-R, Autism Diagnostic 53 Leitges M, Schmedt C, Guinamard R, Davoust J, Schaal S, Stabel S Interview—Revised. Western Psychological Services: Los Angeles, et al. Immunodeficiency in protein kinase cb-deficient mice. 2003 (Italian version ed. by Faggioli R, Saccani M, Persico AM, Science 1996; 273: 788–791. Tancredi R, Parrini B and Igliozzi R. Organizzazioni Speciali: 54 Dimitri P, Domenicotti C, Nitti M, Vitali A, Borghi R, Cottalasso D Florence, 2005). et al. Oxidative stress induces increase in intracellular amyloid 36 Persico AM, Pascucci T, Puglisi-Allegra S, Militerni R, Bravaccio beta-protein production and selective activation of bIandbII C, Schneider C et al. Serotonin transporter promoter variants do PKCs in NT2 cells. Biochem Biophys Res Commun 2000; 268: not explain the hyperserotoninemia in autistic children. Mol 642–646. Psychiatry 2002; 7: 795–800. 55 Domenicotti C, Paola D, Vitali A, Nitti M, d’Abramo C, Cottalasso 37 Barrett JC, Fry B, Maller J, Daly MJ. Haploview: analysis and D et al. Glutathione depletion induces of rat hepatocytes visualization of LD and haplotype maps. Bioinformatics 2005; 21: through activation of protein kinase C novel isoforms and 263–265. dependent increase in AP-1 nuclear binding. Free Radic Biol 38 Horvath S, Xu X, Lake SL, Silverman EK, Weiss ST, Laird NM. Med 2000; 29: 1280–1290. Family-based tests for associating haplotypes with general 56 Garbett KA, Ebert PJ, Mitchell A, Lintas C, Manzi B, Mirnics K phenotype data: application to asthma genetics. Genet Epidemiol et al. Immune transcriptome alterations in the temporal cortex of 2004; 26: 61–69. subjects with autism. Neurobiol Dis 2008 (in press). 39 Spielman RS, McGinnis RE, Ewens WJ. Transmission test for 57 Chauhan A, Chauhan V. Oxidative stress in autism. Pathophysio- linkage disequilibrium: the insulin gene region and insulin- logy 2006; 13: 171–181. dependent diabetes mellitus (IDDM). Am J Hum Genet 1993; 52: 58 Vargas DL, Nascimbene C, Krishnan C, Zimmerman AW, Pardo CA. 506–516. Neuroglial activation and neuroinflammation in the brain of 40 Dudbridge F. Pedigree disequilibrium tests for multilocus haplo- patients with autism. Ann Neurol 2005; 57: 67–81. types. Genet Epidemiol 2003; 25: 115–121. 59 Antar LN, Afroz R, Dictenberg JB, Carroll RC, Bassell GJ. 41 Pritchard JK, Stephens M, Donnelly P. Inference of population Metabotropic glutamate receptor activation regulates fragile x structure using multilocus genotype data. Genetics 2000; 155: mental retardation protein and FMR1 mRNA localization 945–959. differentially in dendrites and at synapses. J Neurosci 2004; 24: 42 Zilbovicius M, Meresse I, Chabane N, Brunelle F, Samson Y, 2648–2655. Boddaert N. Autism, the superior temporal sulcus and social 60 Araki S, Ng DP, Krolewski B, Wyrwicz L, Rogus JJ, Canani L et al. perception. Trends Neurosci 2006; 29: 359–366. Identification of a common risk haplotype for diabetic nephro- 43 Bustin SA. Absolute quantification of mRNA using real-time pathy at the protein kinase C-b1 (PRKCB1) gene locus. J Am Soc reverse transcription polymerase chain reaction assays. J Mol Nephrol 2003; 14: 2015–2024. Endocrinol 2000; 25: 169–193. 61 Araki S, Haneda M, Sugimoto T, Isono M, Isshiki K, Kashiwagi A 44 Arion D, Sabatini M, Unger T, Pastor J, Alonso-Nanclares L, et al. Polymorphisms of the protein kinase C-b gene (PRKCB1) Ballesteros-Yanez I et al. Correlation of transcriptome profile with accelerate kidney disease in type 2 diabetes without overt electrical activity in temporal lobe epilepsy. Neurobiol Dis 2006; proteinuria. Diabetes Care 2006; 29: 864–868. 22: 374–387. 62 Araki S, Koya D, Makiishi T, Sugimoto T, Isono M, Kikkawa R et al. 45 Irizarry RA, Hobbs B, Collin F, Beazer-Barclay YD, Antonellis KJ, APOE polymorphism and the progression of diabetic nephropathy Scherf U et al. Exploration, normalization, and summaries of high in Japanese subjects with type 2 diabetes: results of a prospective density oligonucleotide array probe level data. Biostatistics 2003; observational follow-up study. Diabetes Care 2003; 26: 2416–2420. 4: 249–264. 63 Araki S, Moczulski DK, Hanna L, Scott LJ, Warram JH, Krolewski 46 Lepre J, Rice JJ, Tu Y, Stolovitzky G. Genes@Work: an efficient AS. APOE polymorphisms and the development of diabetic algorithm for pattern discovery and multivariate feature selection nephropathy in type 1 diabetes: results of case-control and in gene expression data. Bioinformatics 2004; 20: 1033–1044. family-based studies. Diabetes 2000; 49: 2190–2195. 47 Subramanian A, Tamayo P, Mootha VK, Mukherjee S, Ebert BL, 64 Persico AM, D’Agruma L, Zelante L, Militerni R, Bravaccio C, Gillette MA et al. Gene set enrichment analysis: a knowledge- Schneider C et al. Enhanced APOE2 transmission rates in families based approach for interpreting genome-wide expression profiles. with autistic probands. Psychiatr Genet 2004; 14: 73–82. Proc Natl Acad Sci USA 2005; 102: 15545–15550. 65 Krey JF, Dolmetsch RE. Molecular mechanisms of autism: a 48 Storey JD, Tibshirani R. Statistical significance for genomewide possible role for Ca2 þ signaling. Curr Opin Neurobiol 2007; 17: studies. Proc Natl Acad Sci USA 2003; 100: 9440–9445. 112–119.

Supplementary Information accompanies the paper on the Molecular Psychiatry website (http://www.nature.com/mp)

Molecular Psychiatry