Blood RNA sequencing confirms upregulated BATF2 and FCGR1A expression in children with autism spectrum disorder


 Mutations in over 100 genes are implicated in autism spectrum disorder (ASD). DNA mutations and epigenomic modifications also contribute to ASD. Transcriptomics analysis of blood samples may offer clues for pathways dysregulated in ASD. To expand and validate published findings of RNA-sequencing (RNA-seq) studies, we performed RNA-seq of whole blood samples from a discovery cohort of eight children with ASD compared with nine age- and sex-matched neurotypical children. This revealed 10 genes with differential expression. Using real-time PCR, we compared whole blood samples from 35 children with ASD and 21 matched neurotypical children for the 10 dysregulated genes detected by RNA-seq. This revealed higher expression levels of the proinflammatory transcripts BATF2 and FCGR1A, and lower expression levels of the anti-inflammatory transcripts ISG15 and MT2A in the ASD compared to the control group. BATF2 and FCGR1A were recently reported as upregulated in blood samples of Japanese adults with ASD. Coupled with that publication, our findings support involvement of these genes in ASD phenotypes, independent of age and ethnicity. Upregulation of BATF2 and FCGR1A and downregulation of ISG15 and MT2A were reported to reduce cancer risk. Implications of the dysregulated genes for pro-inflammatory phenotypes, immunity, and cancer risk in ASD are discussed.


Introduction
Autism spectrum disorder (ASD) is the most complex and heterogonous human neurological disorder.
Inherited or de novo mutations in over 100 known genes are already implicated in ASD, yet, most incidences remain unexplained. ASD may result from detrimental epigenetic modi cations during early embryonic development 1,2 . Further contributors to the highly diverse ASD phenotypes include unique combinations of common variants in many genes (a high polygenic risk score); prenatal environmental in uences 3 ; mitochondrial de ciencies 4,5 ; a chronic pro-in ammatory state 6,7 ; psychiatric and neurologic comorbidities 8 ; and aberrant gut microbiome composition 9 . Together, the diverse phenotypes contribute to di culties in early diagnosis of autism among children, which is crucial for early treatment and parental guidance 10,11 .
The involvement of epigenetics in ASD phenotypes hinders efforts to establish improved diagnostic tools for early identi cation of ASD and its sub-phenotypes, using genomic, biochemical, and metabolomics analyses 12 . Brain imaging diagnostic tools such as fMRI have been also suggested; however, such tools still have low strati cation value and low reproducibility, and thus require further studies before their incorporation in routine pediatric neurology practice 5,13,14 . Genome-wide transcriptomic studies examine changes in gene expression that also re ect epigenetic DNA modi cations, thereby circumventing complex and costly technologies such as DNA methylome pro ling 15 . Indeed, since the completion of the human genome project, numerous studies have applied genome-wide transcriptomics, initially using microarrays and more recently using RNA-sequencing (RNA-seq) technologies, for identifying altered gene expression patterns in blood samples from individuals with ASD compared to neurotypical controls.
Venous blood samples represent an accessible, affordable, and readily available biological resource for establishing differential diagnosis, prognosis, and subtyping of complex disorders such as ASD. Indeed, the majority of transcriptomic studies in humans have utilized venous blood samples. Hence, we aimed to perform comparative RNA-seq in an independent cohort of whole blood samples from children with ASD and a neurotypical control group, to identify the top dysregulated gene transcripts, and to assess (by literature survey) if some of the detected dysregulated genes were already reported as dysregulated in ASD. Here we report that our comparative RNA-seq identi ed two upregulated mRNA transcripts, BATF2 and FCGR1A, in whole blood samples of Israeli children with ASD compared to a control group of neurotypical children. These genes were recently reported as upregulated in whole blood from Japanese adults with ASD.

RNA sequencing and real-time PCR validation
Whole blood samples were collected from 36 children with ASD and 21 NT children at Shaare Zedek Medical Center (Jerusalem, Israel). PBMCs were collected from 40 children with ASD and 26 NT children at the ACH (Little Rock, AR, USA.), as described in the Methods section. The demographics of the Israeli and U.S. cohorts are presented in Table 1. RNA was extracted from these blood samples. RNA from eight males with ASD (mean age 13.5±2.6 y) and nine NT males (controls; mean age 15.4±1.9 y) were applied for RNA-seq as described in the Methods. Bioinformatics analysis of the RNA-seq reads, followed by p value adjustment for genome-wide transcriptomics, identi ed 10 dysregulated genes with differential expression in whole blood in the ASD compared to the control group, p adj <0.05 (Table 2). Next, we performed real-time PCR validation for these dysregulated genes in our entire cohort of whole blood samples (36 ASD and 21 matched NT Israeli children; Table 1). Our ndings showed that four of the genes detected as dysregulated by RNA-seq were con rmed as dysregulated in the entire Israeli cohort of whole blood RNA samples (Fig. 1). We observed upregulated expression of BATF2 and FCGR1A (FD=2.03, p=0.004; FD=1.5, p=0.0013) and downregulated expression of ISG15 and MT2A (FD=0.64, p=0.0074; FD=0.74, p=0.013) in the ASD compared to the control group. Other genes found as dysregulated in RNAseq from whole blood of the two groups (Table 2) were not validated in the entire Israeli cohort; their realtime PCR ndings are shown in Supplementary Fig. S1.
Next, we assessed RNA levels of the 10 dysregulated genes detected by our RNA-seq of whole blood ( Table 2) in RNA extracted from PBMCs from a second cohort of children with ASD and NT controls (U.S. cohort; Table 1). None of these genes showed differential expression in PBMC samples from the ASD versus the control group. The NT controls of the U.S. cohort included both NT siblings of the ASD group and unrelated NT controls (Table 1). We therefore compared the RNA expression levels in PBMCs from children with ASD, separately to those of their NT siblings and to those of unrelated NT controls, for the same 10 genes. The comparison to NT siblings indicated upregulated SERPING1 (FD=1.90; p=0.023) in the children with ASD. The real-time PCR ndings from the ASD and control PBMC samples are shown in Supplementary Fig. S2 and Supplementary Table S2.

Correlations of whole blood gene expression levels with behavioral scores
We looked for possible correlations of the whole blood RNA levels of the genes detected as dysregulated in our RNA-seq, with behavioral phenotypes of the same children with ASD. We observed negative correlations between whole blood RNA levels of BATF2 or SERPING1 and the scores of the individual teacher SRS (tSRS). We observed positive correlations of RNA levels of LY6E with VABS socialization domain scores and VABS Composite scores, and of RNA levels of ISG15 with cbcl scores (Fig. 2). We also observed positive correlations of PBMC RNA levels of BATF2, MT2A, LY6E, and ISG15 with Aberrant Behavior Checklist (ABC) scores (Fig. 3).

Correlations of whole blood gene expression levels with serum endocannabinoids
Reduced levels of several endocannabinoids were reported in serum samples of children with ASD compared with controls 16 . These serum samples were from the same Israeli children from whom we collected whole blood samples and analyzed the RNA samples used in the current study. Therefore, in each of the study participants (both the ASD and control groups), we looked for correlations between blood RNA expression levels of the top dysregulated genes ( Table 2)  Further correlations for whole blood mRNA expression and serum endocannabinoid levels are listed in Supplementary Table S3. The strongest correlation observed was a negative correlation between whole blood LY6E mRNA expression levels and serum N-palmitoylethanolamine (PEA) in the NT children (PEA; r=0.7298, p=0.0004), while no such correlation was observed for the ASD group ( Supplementary Fig. S3). Further correlations between blood mRNA expression levels and serum endocannabinoid levels were detected by combining the ASD and control groups for each correlation plot ( Supplementary Fig. S4).

Discussion
Upregulated expression of BATF2 and FCGR1A in Israeli children with ASD con rms ndings in Japanese adults with ASD The upregulated expression levels of both BATF2 and FCGR1A in our Israeli cohort of whole blood samples from children with ASD ( Fig. 1) corroborate ndings from a recent Japanese RNA-seq study in whole blood from adults with ASD 17 . To our knowledge our study provides the rst validation of a genome-wide RNA-seq study with ASD blood samples. Our literature search of RNA-seq studies (Table 3; see Methods) revealed large variation in ndings from earlier transcriptomic studies (using either RNA-seq or RNA microarray technologies) in whole blood samples of individuals with ASD compared to matched NT controls. Notably, each of the dysregulated genes listed in our literature survey was mentioned in only a single study (or a single meta-analysis in regard to the meta-analysis by Lee et al., 2019 18 ). Therefore, our current study appears to be the rst validation of any dysregulated gene in blood samples from individuals diagnosed with ASD. Moreover, our validation was done on children with ASD (Table 1), in contrast to adults with ASD in the mentioned Japanese study 17 .
Our RNA-seq detected SERPING1 as the top upregulated gene in whole blood from children with ASD (FD=3.4962; P adj =0.0072; Table 2). SERPING1 was among the upregulated genes in whole blood from Japanese adults with ASD 17 . Yet, our real-time PCR experiments could not validate this nding in the entire whole blood samples ( Supplementary Fig. S1). Likewise, SERPING1 expression was similar in PBMC-derived RNA samples from children with ASD and NT children (Supplementary Table S2). Nonetheless, our real-time PCR experiments in PBMC-derived RNA samples (our U.S. cohort) indicated upregulated expression of SERPING1 in children with ASD compared with their NT siblings (Supplementary Table S2). These ndings exemplify an advantage of including NT siblings of children with ASD in autism research studies.
Our ndings thus suggest a central role in ASD for the upregulated genes identi ed, independent of age and ethnicity. Further studies, including transcriptomic studies with brain tissues from ASD animal models, are needed to elucidate the relevance of the dysregulated genes for ASD behavioral scores ( Fig. 2  & Fig. 3) and their implications for ASD phenotypes.

Dysregulated ASD genes and cancer
All four genes that were found in our study to present dysregulated mRNA transcript levels in blood from children with ASD have been investigated mostly in the context of cancer. Both BATF2 and FCGR1A, detected in our study as upregulated in blood samples of children with ASD, code for cancer protective proteins. BATF2 was shown to have an antitumor effect in a mouse model through upregulation of IL-12 p40 in tumor-associated macrophages, leading to CD8+ T-cell activation and tumor accumulation 19 .
Among other cancers, BATF2 was demonstrated as a tumor suppressor of gastric cancer 20 , glioblastoma 21 , and esophageal squamous cell carcinoma 22 . Higher tumor FCGR1A expression correlated with improved prognosis in laryngeal cancer 23 , as well as in cervical cancer and melanoma, in which it was associated with increased tumor in ltration of CD4+ and CD8+ T cells and dendritic cells 24 .
Additionally, higher expression levels of both ISG15 and MT2A, found in this study as downregulated in blood samples from children with ASD, were reported to be associated with worse cancer prognosis. The protein coded by ISG15 (interferon-stimulated gene 15 ubiquitin like modi er) was implicated in autophagy, exosome secretion, DNA repair, and immune modulation pathways; it is also a known tumor promoter by suppressing immune cell tumor in ltration 25 . ISG15 was shown to drive tumorigenesis and metabolic plasticity of pancreatic cancer, suggesting that its inhibition may be a treatment option for pancreatic cancer 26 . Higher ISG15 expression was also associated with poor prognosis in breast cancer 27 . The protein coded by MT2A, metallothionein 2A, is the major metallothionein in humans, and serves as a chelator of intracellular zinc ions and protects cells against free radicals. MT2A is upregulated in most cancers, and contributes to their chemotherapy resistance by chelation of zinc and platinum-containing drugs and by its action on p53 zinc-dependent activity. MT2A upregulation results in p53 misfolding secondary to zinc chelation, while low cellular MT2A levels allow proper p53 function as a genome stability guardian 28, 29 . Lastly, downregulated C1 Inhibitor (encoded by SERPING1) was shown to increase cancer risk 30,31 . Hence, the upregulated SERPING1 observed in sub-cohorts of this study may also contribute to reduced cancer risk in children with ASD.
Taken together, the upregulation of both BATF2 and FCGR1A, and the downregulation of both ISG15 and MT2A, as we detected in blood samples from children with ASD, all suggest a reduced risk of cancer. Indeed, a huge reduction in cancer risk (OR=0.06; 95% CI: 0.02, 0.19; p<0.0001) was reported among children with ASD aged 0 to 14 years compared with matched controls 32 . These authors compared cancer rates in 1,837 individuals with ASD and in 9,336 controls in the registry of the University of Iowa Hospitals and Clinics. They observed that the large gap in cancer rates between individuals with ASD and controls was lower at older ages, being only 2-fold less among individuals with ASD aged above 55 years compared with controls.
Our ndings on upregulated BATF2 and FCGR1A, and downregulated ISG15 and MT2A (or possibly some of these genes) in children with ASD thus seem to agree with the ndings of the Darbro et al.2016 32 epidemiologic survey. Albeit, we did not identify similarly large studies on cancer risk among children with ASD. The only other epidemiologic study reporting reduced cancer risk among individuals with ASD was smaller (91 individuals with ASD and 6,186 sex-and birth-year controls), and was based on death records, thus on older individuals. For all ages combined, it reported a 4.3-fold reduced risk of death from metastatic cancer compared with controls 33 . However, an earlier study reported 1.95-fold higher cancer incidence among males with ASD based on a Taiwanese cancer registry; the elevated cancer risk was particularly high (3.58-fold) for individuals with ASD aged 15-19 years 34 . Additionally, higher cancer mortality (OR=1.80) among individuals with ASD was reported in a study on premature mortality 35 . Yet, the latter study did not include breakdown of death by age. Hence the cancer risk among individuals with ASD remains controversial. Epidemiologic studies with larger cohorts are required to assess the cancer risk among children with ASD compared with NT children.

Dysregulated ASD genes and immunity
Among the common phenotypic features observed in ASD is innate immune system dysregulation, leading to a chronic pro-in ammatory state 6,36 . The innate immune pathways affected in ASD include signaling mediated via cytokines, hepatocyte growth factor receptor, microglia, and the complement system. These suggest a role for aberrant immune function in the broad ASD phenotypes 37 . A recent RNA-seq study of whole blood from adults with ASD found dysregulated transcription of genes implicated in innate and adaptive immunity. These included upregulated expression of BATF2 and FCGR1A 17 , as con rmed in our current study of whole blood from children with ASD. The consequences to the immune system, of dysregulation in children with ASD of the four genes observed in our study (Fig. 1), is discussed in the above section. Notably, BATF2 was shown to promote in ammation in response to lipopolysaccharides or infection 38 , while ISG15 is known to promote anti-in ammatory pathways 39,40 . Thus, the upregulation of BATF2 mRNA, as well as the downregulation of ISG15 mRNA observed in our study supports the involvement of the pro-in ammatory phenotypes that have often been observed in ASD 6,7,36 .
Lack of validation of the genes that were dysregulated in whole blood RNA-seq, in PBMC samples Our real-time PCR experiments did not validate any of the 10 dysregulated genes detected by our RNA-seq of whole blood (Israeli cohort; Table 1) in RNA extracted from PBMCs of a second cohort of children with ASD and NT controls (U.S. cohort; Table 1). Nonetheless, comparing children with ASD to their NT siblings indicated upregulated SERPING1 in the PBMCs of those with ASD (Supplementary Table S2). The real-time PCR ndings from the PBMC samples, as presented in Supplementary Fig. S2 and compared with Fig. 1, suggest that the source of the other dysregulated transcripts detected in our RNA-seq of whole blood RNA (Table 2) mostly represent neutrophil RNA. This is because these cells (which represent the major source of blood-derived RNA) are depleted during isolation of PBMCs from whole blood (neutrophils have higher density than PBMCs, and are removed during PBMC separation, together with erythrocytes). This conclusion is rational considering that neutrophils are the key players in in ammation 41,42 , and that individuals with ASD often display pro-in ammatory phenotypes 6,7,36 . Indeed, in ammatory signaling and reactive oxygen species mediators were shown to be upregulated in neutrophils of children with ASD 43 .

Correlations of whole blood gene expression levels with serum endocannabinoids
Recent years have seen growing interest in studying the use of cannabinoid drugs for treating behavioral and social de cits of individuals with ASD 44 . The endocannabinoid system was reported to be dysregulated in various animal models of ASD 45 . Children with ASD were reported to have lower serum endocannabinoids 16 , and rare mutations in endocannabinoid pathway genes were implicated in some persons with ASD 46 . Circulating endocannabinoids are derived from multiple tissues 47 . However, plasma endocannabinoid levels were demonstrated to re ect brain concentrations 48 . Hence, the correlations reported here for blood expression of some of the dysregulated genes with serum endocannabinoids (Fig. 4) can shed light on the pathophysiology of ASD. However, studies in animal models of ASD, which allow measurements of endocannabinoid levels and transcriptomics in brain tissues during different stages of pre-and postnatal development 49,50 , are required for exploring these correlations. The endocannabinoid PEA was reported to display anti-in ammatory 13,51 , antiepileptic 52 , and antineuropathic 53 properties. The LY6E cell surface protein was shown to be upregulated by several in ammatory cytokines, including interferons, TNF-alpha, and IL-1 alpha 54 . As expected, we found a strong negative correlation between the anti-in ammatory endocannabinoid PEA and the proin ammatory transcript LY6E in NT children (Supplementary Table S3, r=-0.73, p=0.0004). This expected negative correlation was not observed in children with ASD, thus demonstrating another example of dysregulated immunity in ASD. Further studies are needed to clarify the relation of the dysregulated blood transcriptomics to the reduced serum endocannabinoids observed in children with ASD 16 .

Limitations
Our current study has several limitations. First, the small cohorts do not enable separate analysis for males and females. As all the participants in the RNA-seq discovery cohort and the majority of the participants in the entire Israeli cohort were male, the relevance of our ndings to female children with ASD needs to be clari ed. Second, the whole blood samples (Israeli cohort) and PBMC samples (U.S. cohort) were collected separately, prior to conducting the RNA-seq project; hence, we did not have both whole blood and freshly separated PBMCs from the same individuals. Since the RNA-seq project, its analysis, and real-time PCR validation experiments were conducted from October 2020 to September 2021, recruitment of the children was affected by the Covid-19 pandemic and related lockdowns. Additionally, the children with ASD in the U.S. cohort were on average four years younger than those of the Israeli cohort (Table 1). Hence, our ndings on lack of validation for the whole blood dysregulated genes in PBMC samples from children with ASD require con rmation with a study that compares whole blood and PBMC derived RNAs from the same participants. Lastly, the relevance of our transcriptomic ndings for early ASD diagnosis is uncertain, as only a few children in both the Israeli and the U.S.
cohorts were under age 4 years, the most crucial period for early ASD diagnosis 10,11 .
In light of the above mentioned reservations, we conclude that validation in larger cohorts, which will ideally include blood samples from younger children, both males and females, are essential for assessing the potential diagnostic and prognostic values of the dysregulated genes detected in our current study. The implications for the upregulated blood transcription of BATF2 and FCGR1A, and downregulated transcription of ISG15 and MT2A, for the distinctive immune system phenotypes in autism, as well as the controversial published ndings on lower cancer risk among children with ASD, merit further studies.

Participants
The study comprised an Israeli and a U. Whole blood (Israeli cohort) samples were collected into Tempus TM Blood RNA tubes (Applied Biosystems™ Catalog number 4342792, Thermo Fisher Scienti c, MA, USA) and tubes were frozen immediately at minus 80°C until the RNA extractions. For PBMC sample preparation (U.S. cohort), whole blood samples were collected in EDTA Vaccutainer TM tubes, and PBMCs were separated from fresh blood samples as described 55 and stored at minus 80°C until the RNA extractions.
RNA extraction RNA was extracted from whole blood samples using the Tempus™ Spin RNA Isolation Kit (Invitrogen™, Thermo Fisher Scienti c, MA, USA) by following the manufacturer's protocol. RNA quality was determined by an automated electrophoresis process using the TapeStation system (Agilent, CA, USA). Samples with high quality RNA were speci ed as those with RIN (RNA integrity number) >8. Nucleic acid quantitation was carried out by Qubit™ RNA HS Assay Kit (Invitrogen™, Thermo Fisher Scienti c, MA, USA). The samples selected for sequencing were diluted to 1 ug/ml.
RNA was extracted from PBMC samples as described previously 55 . Due to low RNA amounts, these samples were not prepared for sequencing, and were applied for assessing the expression of the top transcripts detected by our RNA-seq discovery cohort in PBMC (real-time PCR).
Library preparation, RNA sequencing, and data processing RNA samples were shipped in dry ice to Macrogen Europe BV (Amsterdam, Netherlands) for poly-A mRNA sequencing. Libraries were prepared using TrueSeq stranded total RNA LT sample prep kit (Illumina, CA, USA). RNA sequencing was performed as a paired-end read on an Illumina True-Seq platform.

Competing interests
The authors declare no competing interests or other interests that might be perceived to in uence the results and/or discussion reported in this article.

Data availability
The datasets generated during the current study are available from the corresponding authors on reasonable request.   Figure 1 Real-time qPCR validation for whole blood RNA expression levels in ASD and control children (Israeli cohort). Box plots show mean ± SEM mRNA levels for ASD vs. neurotypical control whole blood samples. Outliers were removed and analysis was done using a non-parametric Mann Whitney test. FD and p values are shown for the genes with differential expression in ASD vs. neurotypical controls.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.