A preliminary investigation on the relationship between gut microbiota and gene expressions in peripheral mononuclear cells of infants with autism spectrum disorders

Abstract

Fecal and blood samples of infants with autism spectrum disorders (ASD) and healthy infants were analyzed to investigate the association of altered gut microbiota and ASD development. 16S rRNA gene-based sequencing found that, unlike those of healthy infants, feces of ASD infants had significantly higher and lower abundance of genera Faecalibacterium and Blautia, respectively. Moreover, DNA microarray analysis of peripheral blood mononuclear cells (PBMC) detected more highly than low expressed genes in ASD infants than in healthy infants. Gene Ontology analysis revealed that differentially expressed genes between ASD and healthy infants were involved in interferon (IFN)-γ and type-I IFN signaling pathways. Finally, strong positive correlations between expression of IFN signaling-associated genes in PBMC and fecal abundance of Faecalibacterium were found. Our results strongly suggested that altered gut microbiota in infants resulted from ASD development and was associated with systemic immunity dysregulation, especially chronic inflammation.

Graphical abstract

Altered gut microbiota in autistic infants play a role in chronic inflammation induced by interferon signaling that may cause further inflammation in central nervous system.

Key words:

• autism spectrum disorders
• Faecalibacterium
• gut microbiota
• peripheral blood mononuclear cell
• interferon signaling

Autism spectrum disorders (ASD) are complex neurophysiological disorders characterized by reduced qualitative social and communication skills but augmented ritualistic and stereotyped behavior.¹⁾ Extensive immune and neuronal dysregulation has been suggested to be involved in the development of ASD but the underlying etiology is not yet well understood.¹⁾ Recently, a greater number of ASD cases have been reported, but this increase cannot be explained only by greater awareness and improved diagnosis methods. Thus, it is likely that environmental factors²⁾ as well as genetics³⁾ play a role in the development of ASD.

Gut microbiota, consisting of more than 500 microbial species,⁴⁾ has lately attracted attention concerning the onset of ASD. Indeed, several studies reported that disruption of the gut microbiota was more frequent in individuals with ASD than it was in healthy individuals.^5–12) Emerging evidence shows that the gut microbiota affects not only nutritional and immunological functions¹³⁾ but also neural signaling¹⁴⁾ and behavioral traits.¹⁵⁾ Hence, it is very likely that the gut microbiota plays an important role in the development of ASD.

It remains unclear, however, whether altered gut microbiota is indeed a causative or deteriorating factor in ASD, or only a consequence of the disease, because no consistent trend of abnormality in the gut microbiota of individuals with ASD has been established yet. In an elegant summary of work on the association of the gut microbiota with individuals with ASD, Louis¹⁶⁾ highlighted several inconsistencies among major studies. For example, a wide variation in bacterial diversity in general, and Bacteriodetes proportions in particular, was reported across studies.¹⁶⁾ More recently, several other studies have reported altered gut microbiota in individuals with ASD,^17,18) but their data still lacked consistency with previous studies.

Due to the conflicting data currently available, there is clearly a need for further investigation of the relationship between the gut microbiota and ASD. To that end, in the present work we compared the gut microbiota composition in infants with ASD with that in healthy infants because no such study has been conducted in Asia. Furthermore, we analyzed the transcriptome of peripheral blood mononuclear cells (PBMC) from the same infants as an innovative component. In addition, we evaluated whether genes expressed at different degrees in PBMC of infants with ASD correlated with the abundance of a particular gut bacterial genus in fecal microbiota. Our results strongly suggested that altered gut microbiota in infants with ASD was not merely a consequence of the disease, but was also associated with dysregulation of systemic immunity, especially chronic inflammation in ASD.

Materials and methods

Ethics statements

The present work conformed to the code of ethics stated in the Declaration of Helsinki. The research protocol was approved by the Ethical Committee of Shiga University of Medical Science (Shiga, Japan; Approval number: 24-161). Written informed consent was obtained from the parents of the infant donors to use the fecal and blood samples for research purposes. The privacy rights of the infant donors were observed at all times.

Subjects

Six infants with ASD (experiment group) and six healthy infants (control group) in the range of 3–5 years of age and living in similar geographic regions in Japan were enrolled in this study. The infants in the ASD group were pre-categorized by the Diagnostic and Statistical Manual of Mental Disorders, 5th Edition (DSM-5) criteria and diagnosed according to the Pervasive Developmental Disorders Autism Society Japan Rating Scale (PARS)¹⁹⁾ and Modified Checklist for Autism in Toddlers (M-CHAT).²⁰⁾ No infant had a considerable gut disorder or took antibiotics or drugs to treat ASD for at least one month before the sampling period.

Collection of feces and extraction of bacterial genomic DNA

Freshly evacuated feces were aseptically collected, placed into sterilized containers (Stool Carry, Atect, Osaka, Japan) and kept at −20 °C. Containers containing the feces were transported to a laboratory within 24 h post-evacuation and stored at −80 °C until further use. Bacterial genomic DNA was extracted from the fecal samples as previously described.²¹⁾

Library preparation and DNA sequencing

Preparation of library for DNA sequencing by a Miseq desktop sequencer (Illumina, San Diego, CA, USA) was carried out as per a protocol by Illumina (https://support.illumina.com/content/dam/illumina-support/documents/documentation/chemistry_documentation/16s/16s-metagenomic-library-prep-guide-15044223-b.pdf) with minor modifications. Briefly, the V3-4 region of 16S rRNA genes in each sample was amplified by primers 341F and 805R containing 5′ overhang adapter sequence for PCR using a KAPA HiFi HotStart Ready Mix (Kapa Biosciences, Wilmington, MA, USA). The amplicon was purified by NucleoFast 96 PCR plates (Takara Bio Inc., Shiga, Japan). A second PCR was carried out using the KAPA HiFi HotStart Ready Mix (Kapa Biosciences) to attach a unique combination of dual indices (I5 and I7 index) and Illumina sequencing adapters to each sample. The amplicon of the second PCR was purified and the concentration was normalized with a SequalPrep Normalization Plate Kit (Life Technologies, Tokyo, Japan). Each of the normalized amplicons was then evenly pooled and concentrated using AMPure XP beads (Beckman Coulter, Tokyo, Japan). The size and quantity of the library were assessed with a Bioanalyzer 2100 (Agilent Technologies Japan, Tokyo, Japan) and a Library Quantification Kit for Illumina (Kapa Biosciences), respectively. The library was denatured with 0.2 N NaOH (Sigma-Aldrich, Tokyo, Japan) and combined with phiX Control (v3, Illumina; expected 20%). Eleven pM of the library combined with phiX Control was heat-denatured at 96 °C for 2 min and sequenced using a 300-bp paired-end strategy on the Miseq (Illumina) as per the manufacturer’s instructions.

Sequence data analysis

Sequences were first processed according to Fadrosh et al.²²⁾ using a modified version of the original script kindly donated by Dr B. Ma and Prof J. Ravel of the University of Maryland School of Medicine. Briefly, 8-bp index sequences at both ends extracted by Miseq Reporter software (v2.4.60) were concatenated to form a 16-bp dual-index barcode specific for each paired-end read. Next, each dual-index barcode was paired with the corresponding sample sequence that was assembled with Fast Length Adjustment of Short reads (FLASH) v1.2.11.²³⁾ De-multiplexing of the sequences according to the dual-index was carried out by Quantitative Insights Into Microbial Ecology (QIIME) v1.9.0²⁴⁾, allowing a Phred score higher than Q21, a maximum of 3 errors in barcodes and no Ns. Afterwards, sequence analysis tool USEARCH v8.0²⁵⁾ was used for further filtering (maximum expected error = 0.5), dereplication, discarding clusters less than 5 reads, and clustering OTU. Chimera was validated and removed with UCHIME version 4.2.40²⁶⁾ in reference mode using the Broad Institute’s 16S rRNA gene Gold reference database. Taxonomy assignment of resulting OTU was carried out using RDP classifier v2.10.2²⁷⁾ with the Greengenes database (published May, 2013). Default settings were used for all software unless otherwise stated.

Collection of whole blood and isolation of PBMC

Peripheral blood was drawn from four of the six infants with ASD (A2-A5) and four of the six healthy infants (C2, C4-C6), and collected in Venoject II vacuum blood collection tubes (heparinized blood collection tubes) (Terumo Medical Corp., Tokyo, Japan). PBMC were isolated as previously described by Nishibayashi et al.²⁸⁾

Microarray analysis of PBMC

Total RNA was extracted and cDNA synthesized from 35 ng of total RNA as described by Ogawa et al.²⁹⁾ Complementary DNA was labeled with Cy3-dUTP (GE-Healthcare, Tokyo, Japan) with a random primer (Life Technologies) using 50 units of Klenow Polymerase (New England Biolabs, Tokyo, Japan) in the presence of 0.5 mM dATP, dGTP and dCTP, and 0.05 mM dTTP. After 3-h incubation at 37 °C, Cy3-labeled cDNA was purified with a PureLink Quick PCR Purification kit (Life Technologies). The quantity and quality of Cy3-labeled cDNA were verified using a NanoDrop Spectrophotometer (NanoDrop Technologies, Wilmington, DE, USA). Six hundred ng of each Cy3-labeled cDNA sample was fragmented by incubating with 0.55 Kunitz unit of DNase I (RNase-Free DNase Set; Qiagen) for 8 min at 37 °C. This incubation time was determined in advance so that the size of cDNA became approximately 50–150 bases. DNase I was inactivated for 2 min at 98 °C. Samples were then hybridized to the array slide [SurePrint G3 Human GE microarray, 8 × 60 K, V2; Agilent Technologies Japan] for 16 h at 65 °C. The array was washed as per the manufacturer’s instructions and scanned with an Agilent G2565CA microarray scanner. Intensity data were extracted using Feature Extraction software (Agilent).

Real-time RT-PCR

The expression of IFN-related genes (IRF7, IRF9, CXCL10 and CXCL11) in PBMC was analyzed by real-time RT-PCR. Complementary DNA was synthesized from each total RNA sample (20 ng) using a RevaTra Ace qPCR RT kit (TOYOBO, Osaka, Japan), as per the manufacturer’s instructions. Real-time PCR was conducted using a LightCycler 480 instrument (Roche Applied Science, Tokyo, Japan). Specific primers and probe sets for each gene were designed with online software available on the Universal ProbeLibrary Assay Design Center (Roche Applied Science). Sequences of the primers and probes used in this study are shown in Table S1. GAPDH and ACTB were used as the reference genes, and PCR conditions and analyses conducted for all genes were as described by Yoshikawa et al.³⁰⁾ The analysis was separately conducted twice.

Data analysis

The abundance (percentage) of each bacterial genus between groups was compared by the Student’s t-test. The Student t-test was performed by Excel^® (Microsoft Corp., Redmond, WA, USA).

Intensity data from the microarray analysis were analyzed by CLC MainWork bench v6.7 (CLC bio Japan, Tokyo, Japan). Signal intensity values were normalized by the quartile normalization method and log₂ transformed. The Student’s t-test was used to evaluate differences in the transformed values of each microarray probe between groups. The p value and the fold difference of each gene were imported into web-based pathway analysis software (iPathway), and the Gene Ontology (GO) analysis was conducted on the differentially expressed genes (−1.5 ≤ fold change ≥ 1.5, p < 0.05) to detect the processes in which they were involved. Afterwards, the correlation between the expression level (transformed value) of the differentially expressed genes involved in the GO biological processes and the abundance of bacterial genera in fecal microbiota that was significantly different between groups were assessed with the Pearson’s product-moment coefficient using freely available R software v3.1.2.

For the real-time RT-PCR analysis, the gene expression level relative to the expression level of the C2 subject was calculated by the Delta Ct method and correlation between this expression value and the abundance of Faecalibacterium in fecal microbiota was assessed as described above. The Delta Ct value was also used for the Student’s t-test analysis of the comparison between groups.

Results

Fecal microbiota in ASD infants and healthy infants

A total of 307,022 sequence reads were obtained after data processing (average 25,585.2 reads; ranged 10,344–37,647). The abundance of genera Faecalibacterium and Blautia was significantly different between infants with ASD and healthy infants (Table 1). The abundance of genus Faecalibacterium was significantly higher in the fecal microbiota of ASD infants (9.78 ± 6.75%) than it was in that of healthy infants (3.25 ± 1.61%). In contrast, the abundance of genus Blautia was significantly lower in the fecal microbiota of ASD infants (3.34 ± 4.50%) than it was in that of healthy infants (14.21 ± 10.02%).

Table 1. Relative abundance (%) of bacterial family/genera in the fecal microbiota of ASD infants and healthy infants.

Order	Family	Genus	Control (%)	ASD (%)
Bifidobacteriales	Bifidobacteriaceae	Bifidobacterium	13.74 ± 8.89	9.38 ± 5.38
Coriobacteriales	Coriobacteriaceae	Collinsella	<0.01	0.71 ± 1.58
Coriobacteriales	Coriobacteriaceae	Eggerthella	0.37 ± 0.17	0.14 ± 0.17
Bacteroidales	Bacteroidaceae	Bacteroides	18.64 ± 18.19	30.32 ± 8.21
Bacteroidales	Porphyromonadaceae	Parabacteroides	1.17 ± 1.47	1.46 ± 1.92
Bacteroidales	Rikenellaceae	Other	0.20 ± 0.50	1.71 ± 1.91
Bacteroidales	[Barnesiellaceae]	Other	<0.01	0.24 ± 0.39
Lactobacillales	Streptococcaceae	Lactococcus	0.08 ± 0.19	<0.01
Lactobacillales	Streptococcaceae	Streptococcus	6.36 ± 7.94	0.69 ± 0.55
Turicibacterales	Turicibacteraceae	Turicibacter	0.01 ± 0.02	0.09 ± 0.12
Clostridiales	Other	Other	<0.01	0.10 ± 0.17
Clostridiales	Christensenellaceae	Other	<0.01	0.03 ± 0.09
Clostridiales	Clostridiaceae	Other	0.32 ± 0.56	0.04±0.08
Clostridiales	Clostridiaceae	Clostridium	1.83 ± 1.96	1.44 ± 1.64
Clostridiales	Clostridiaceae	SMB53	0.17 ± 0.31	<0.01
Clostridiales	Lachnospiraceae	Other	6.28 ± 3.38	3.05 ± 1.93
Clostridiales	Lachnospiraceae	Anaerostipes	0.03 ± 0.05	0.40 ± 0.62
Clostridiales	Lachnospiraceae	Blautia	14.21 ± 10.02	3.34 ± 4.50
Clostridiales	Lachnospiraceae	Coprococcus	3.08 ± 2.41	1.17 ± 0.92
Clostridiales	Lachnospiraceae	Dorea	1.07 ± 0.90	1.26 ± 1.72
Clostridiales	Lachnospiraceae	Lachnospira	0.13 ± 0.23	0.26 ± 0.41
Clostridiales	Lachnospiraceae	Roseburia	2.01 ± 3.84	4.68 ± 4.92
Clostridiales	Lachnospiraceae	[Ruminococcus]	5.54 ± 4.41	2.59 ± 2.78
Clostridiales	Peptostreptococcaceae	Other	0.59 ± 0.90	0.36 ± 0.10
Clostridiales	Ruminococcaceae	Other	7.69 ± 6.85	4.84 ± 5.01
Clostridiales	Ruminococcaceae	Anaerotruncus	0.06 ± 0.07	0.07 ± 0.11
Clostridiales	Ruminococcaceae	Faecalibacterium	3.25 ± 1.61	9.78 ± 6.75
Clostridiales	Ruminococcaceae	Oscillospira	0.41 ± 0.36	1.36 ± 1.64
Clostridiales	Ruminococcaceae	Ruminococcus	2.29 ± 2.43	2.94 ± 1.79
Clostridiales	Veillonellaceae	Acidaminococcus	<0.01	0.12 ± 0.30
Clostridiales	Veillonellaceae	Dialister	0.81 ± 1.22	0.49 ± 0.49
Clostridiales	Veillonellaceae	Megamonas	<0.01	0.98 ± 2.41
Clostridiales	Veillonellaceae	Megasphaera	0.03 ± 0.07	0.36 ± 0.55
Clostridiales	Veillonellaceae	Phascolarctobacterium	<0.01	0.15 ± 0.37
Clostridiales	Veillonellaceae	Veillonella	3.58 ± 2.75	4.28 ± 5.50
Clostridiales	[Mogibacteriaceae]	Other	0.17 ± 0.27	0.03 ± 0.04
Erysipelotrichales	Erysipelotrichaceae	Other	1.92 ± 1.68	0.72 ± 0.90
Erysipelotrichales	Erysipelotrichaceae	Coprobacillus	0.02 ± 0.06	<0.01
Erysipelotrichales	Erysipelotrichaceae	[Eubacterium]	0.47 ± 0.26	0.19 ± 0.23
Burkholderiales	Alcaligenaceae	Sutterella	0.01 ± 0.01	0.98 ± 2.03
Desulfovibrionales	Desulfovibrionaceae	Bilophila	<0.01	0.05 ± 0.09
Desulfovibrionales	Desulfovibrionaceae	Desulfovibrio	0.02 ± 0.04	0.05 ± 0.13
Enterobacteriales	Enterobacteriaceae	Other	0.22 ± 0.20	1.14 ± 1.52
Enterobacteriales	Enterobacteriaceae	Klebsiella	0.67 ± 1.64	0.01 ± 0.01
Verrucomicrobiales	Verrucomicrobiaceae	Akkermansia	2.57 ± 6.22	7.96 ± 5.53

Notes: Data are presented as mean ± SD. Bold text indicates that taxa are significantly different between ASD infants and healthy infants (p < 0.05).

Differentially expressed genes in PBMC between infants with ASD and healthy infants

There were 1,056 more highly expressed genes in ASD infants than in healthy infants. In contrast, expression of 517 more genes was significantly low in ASD infants than they were in healthy infants.

With regard to the GO analysis of the differentially expressed genes, a considerable number of GO biological processes associated with response to viruses (six of 20) were enriched. The top 20 GO biological processes significantly enriched are listed in Table 2. Processes apart from the top 20 were further analyzed to detect processes associated with interferon, as some studies reported an increase in interferon (IFN)-α or IFN-γ in infants with ASD.^31–33) Three and five processes associated with IFN-γ and type I IFN, respectively, were found to be significantly enriched (Table 3). Furthermore, interferon regulatory factor (IRF)7 and IRF9 genes were involved in seven of the eight processes associated with interferon.

Table 2. GO pathways enriched for the differentially expressed genes between ASD infants and healthy infants.

ID	Pathway name	significance*	Total genes**	p value
GO:0009615	Response to virus	27	296	9.E-06
GO:0051607	Defense response to virus	21	199	1.E-05
GO:0006952	Defense response	75	1412	3.E-04
GO:0050688	Regulation of defense response to virus	10	77	4.E-04
GO:0046456	Icosanoid biosynthetic process	8	51	4.E-04
GO:1901570	Fatty acid derivative biosynthetic process	8	51	4.E-04
GO:0045071	Negative regulation of viral genome replication	7	40	5.E-04
GO:0009954	Proximal/distal pattern formation	6	30	6.E-04
GO:0006826	Iron ion transport	8	54	6.E-04
GO:0098542	Defense response to other organism	26	365	7.E-04
GO:0006636	Unsaturated fatty acid biosynthetic process	8	59	1.E-03
GO:0019050	Suppression by virus of host apoptotic process	2	2	1.E-03
GO:0033668	Negative regulation by symbiont of host apoptotic process	2	2	1.E-03
GO:0039526	Modulation by virus of host apoptotic process	2	2	1.E-03
GO:0052041	Negative regulation by symbiont of host programmed cell death	2	2	1.E-03
GO:0052490	Negative regulation by organism of programmed cell death in other
	Organism involved in symbiotic interaction	2	2	1.E-03
GO:0061324	Canonical Wnt signalling pathway involved in positive regulation of
	Cardiac outflow tract cell proliferation	2	2	1.E-03
GO:1901963	Regulation of cell proliferation involved in outflow tract morphogenesis	2	2	1.E-03
GO:0060548	Negative regulation of cell death	43	721	1.E-03
GO:0048525	Negative regulation of viral process	9	75	1.E-03

Note: Only the top 20 pathways are shown.

^* Significance (p < 0.05) was calculated by comparing the number of differentially expressed genes between groups.

^** Total number of genes assigned to each pathway. Differentially expressed genes were imported into iPathway software.

Table 3. Interferon-related GO pathways enriched for the differentially expressed genes between ASD infants and healthy infants.

ID	Pathway name	Significance*	Total genes**	p value
GO:0034341	Response to interferon-gamma	11	117	0.003
GO:0060333	Interferon-gamma-mediated signaling pathway	8	74	0.005
GO:0071346	Cellular response to interferon-gamma	9	97	0.008
GO:0045351	Type I interferon biosynthetic process	3	13	0.01
GO:0060337	Type I interferon signaling pathway	7	73	0.016
GO:0071357	Cellular response to type I interferon	7	73	0.016
GO:0034340	Response to type I interferon	7	74	0.017
GO:0035456	Response to interferon-beta	3	18	0.025

^* Significance (p < 0.05) was calculated by comparing the number of differentially expressed genes between groups.

^** Total number of genes assigned to each pathway. Differentially expressed genes were imported into iPathway software.

Correlation between differentially expressed genes in PBMC and the abundance of bacterial genera in fecal microbiota

Three and two of eight differentially expressed genes involved in the GO process named IFN-γ-mediated signaling pathway (GO:0060333) significantly correlated with the abundance of Faecalibacterium and Blautia, respectively (Table 4). In addition, three of eight differentially expressed genes showed a correlation that tended to be significant (p < 0.1) with either genus (Table 4). Interestingly, although six of seven genes involved in the GO process named type I interferon signaling pathway (GO:0060337) significantly correlated with the abundance of genus Faecalibacterium, only IRF7 correlated with the abundance of Blautia (Table 4).

Table 4. Correlation between the expression level of genes involved in interferon signaling pathways in PBMC and the abundance of Faecalibacterium and Blautia in fecal microbiota.

GO process		Interferon gamma mediated signalling pathway (GO:0060333)
Genes		IRF7	IRF9	SP100	HLA-DQA1	HLA-DQB1	CAMK2G	HCK	PML
Fold difference (vs. healthy infants)		3.38	1.81	1.94	−1.68	−2.17	−1.92	−2.56	2.50
Correlation (r) §	Faecalibacterium	0.76*	0.75*	0.43	−0.61	−0.63†	−0.71†	−0.67†	0.74*
	Blautia	0.84†	−0.29	−0.67†	0.81*	0.67†	0.55	0.64†	−0.42
GO process		Type I interferon signalling pathway (GO:0060337)
Genes		IRF7	IRF9	SP100	IFITM1	IFIT1	ISG20	MX1
Fold difference (vs. healthy infants)		3.38	1.81	1.94	1.99	10.23	3.68	5.42
Correlation (r) §	Faecalibacterium	0.76*	0.75*	0.43	0.80*	0.80*	0.82*	0.74*
	Blautia	−0.84†	−0.29	−0.67†	−0.58	−0.48	−0.49	−0.64†

Notes: Fold differences of the differentially expressed genes between groups involved in GO:0060333 and GO:0060337 are shown. §Pearson’s r between gene expression levels in PBMC and the fecal abundance of Faecalibacterium or Blautia was calculated by R.

^* p < 0.05.

^† p < 0.1.

Real-time PCR analysis

Normalization by housekeeping genes GAPDH and ACTB yielded very similar data. Thus, in the present work only data normalized by GAPDH are shown.

The expression of all IFN signaling-associated genes evaluated by real-time RT-PCR was higher in ASD infants than it was in healthy infants, and exactly as it was detected by the microarray analysis. IRF7 was expressed significantly higher in ASD infants than it was in healthy infants, and the expression of three other genes tended to be higher also in ASD infants (p < 0.1). Finally, the expression of all genes significantly correlated with the abundance of genus Faecalibacterium (Fig. 1).

Fig. 1. Correlation of gene expression levels in PBMC with the abundance of genus Faecalibacterium in fecal microbiota.

Notes: Gene expression levels of IRF7 (A), IRF9 (B), CXCL10 (C) and CXCL11 (D) in healthy infants (squares) and ASD infants (triangles) were assessed by real-time PCR and their correlation with the abundance of Faecalibacterium in fecal microbiota evaluated. Correlations (r) and statistical significance were determined by Pearson’s product-moment correlation. The correlation (r) and P values are shown in the figure.

Discussion

ASD are putatively caused by multiple genetic and non-genetic etiologies.¹⁾ Etiological differences between ASD infants and healthy infants have been evaluated from various aspects including genetic mutations,³⁾ brain activity,³⁴⁾ neuronal³³⁾ and immune functions,³⁵⁾ metabolism³⁶⁾ and gut microbiota composition.¹⁶⁾ In this study, we evaluated the differences between ASD infants and healthy infants by focusing on the gut microbiota and its relation with transcriptomic profiles of PBMC, namely, genes involved in interferon signaling.

The number of samples from ASD infants used in the present work may seem to be relatively small in comparison with the numbers usually reported in clinical trials. The reason for the small number of samples was due to the fact that consent from Japanese parents of autistic infants to analyze their fecal and peripheral blood samples was difficult to obtain. We attribute this parental unwillingness to participate to inherent social and cultural characteristics of Japanese society. For example, in Japan, emotional and mental problems are still stigmatized.³⁷⁾ As a result, these problems are frequently hidden from and denied to outsiders.³⁸⁾ Nonetheless, we are confident that, although preliminary, the data obtained from statistically significant correlations between the Faecalibacterium abundance and gene expression levels in infant fecal and blood samples, respectively, were sufficiently solid to support our conclusions.

The abundance of two bacterial genera in fecal microbiota was significantly different between ASD infants and healthy infants. Furthermore, compared with healthy infants, ASD infants had a significantly higher abundance of genus Faecalibacterium, but a significantly lower abundance of genus Blautia. Faecalibacterium prausnitzii is the only species known to belong to genus Faecalibacterium³⁹⁾ and thus, has been mostly regarded as commensal or even beneficial.⁴⁰⁾ Moreover, recent studies showed that this species produces microbial anti-inflammatory molecule, an anti-inflammatory protein,^41,42) and that a decrease of F. prausnitzii is associated with human inflammatory disorders.⁴⁰⁾ Thus, a high population of this genus in fecal samples of ASD infants in the present study was unexpected. Nonetheless, this genus seems to be not always beneficial, especially in infants. For example, Hansen et al.⁴³⁾ found that F. prausnitzii was more abundant in infants with Crohn’s disease than it was in control infants, and Balamurugan et al.⁴⁴⁾ reported a higher number of F. prausnitzii in obese infants than in infants with normal weight. Gut inflammation is a symptom often observed in experimental models of obesity.⁴⁵⁾ Recently, Song et al.⁴⁶⁾ also found an enrichment of F. prausnitzii and a reduced number of anti-inflammatory effect-inducing bacteria in the fecal samples of patients suffering from atopic dermatitis, a chronic skin inflammation disease. Therefore, it is likely that higher abundance of Faecalibacterium in ASD infants in this study may have resulted in gut inflammation, at least to some extent.

It is important to mention that the evidence of abnormality detected in the gut microbiota of ASD infants in this study was also inconsistent with previous human studies investigating the gut microbiota in ASD infants^{5–12,17,18)} and even in a mouse model.⁴⁷⁾ For example, no studies have suggested an increase in Faecalibacterium abundance in ASD infants. In addition, the genus Blautia consists of bacteria re-clustered from Clostriudium cluster XIVa,⁴⁸⁾ and no relation linking this bacterial group with ASD has been reported yet. Hence, to investigate whether abnormalities in the gut microbiota were plausibly involved in the etiologies of ASD or only a consequence of biased eating habits and food preferences often reported in ASD infants,⁴⁹⁾ we evaluated the correlation between the expression levels of the differentially expressed genes in PBMC and the abundance of Faecalibacterium and Blautia.

We examined the correlated genes by focusing on IFN-related signaling processes enriched by the GO analysis, and found that, unlike Blautia abundance, Faecalibacterium abundance was strongly correlated with a greater number of differentially expressed genes involved in both the interferon-γ-mediated signaling pathway and the type I interferon signaling pathway, but in particular in the latter. Although conflicting reports exist,^50,51) a higher concentration of IFN-α or IFN-γ in cerebrospinal fluid or plasma/serum and higher production of IFN-γ by blood cells have been reported in several studies.^31–33) It is also of particular interest that large doses of IFN-α given to children in cancer treatment cause similar symptoms to those of autism, namely, withdrawing, lost interest in their surroundings and non-communication.⁵²⁾

To further confirm the correlation between Faecalibacterium abundance in fecal microbiota and the genes associated with IFN signaling, the expression level of IRF7 and IRF9 was measured by real-time PCR as they would be plausible markers for activation of type I interferon signaling.^53,54) Genes CXCL10 and CXCL11 were also evaluated because they were (a) induced by interferon signaling^55–57) and, (b) expressed higher than 4-fold in ASD infants than they were in healthy infants in microarray analyses (although this expression was not statistically significant). In addition, an elevated concentration of CXCL10 or CXCL11, but especially the former, has been previously reported in serum or cerebrospinal fluid of individuals with several neuroimmunological disorders, including ASD.^33,58–60) The expression level of IRF7 and IRF9 evaluated by real-time PCR showed a strong positive correlation with the abundance of Faecalibacterium in fecal microbiota, which was consistent with our results from the microarray analysis. Moreover, the expression level of CXCL10 and CXCL11 also significantly correlated with Faecalibacterium abundance. Accordingly, the expression of all genes evaluated by real-time PCR was higher in PBMC of ASD infants than it was in those of healthy infants.

Taken all the results together, three scenarios can be suggested. First, as suggested in previous studies,⁶¹⁾ immune dysfunction may be a common feature in ASD infants and is likely associated with chronic inflammation because IFN has been suggested to be involved in chronic, not acute inflammation.^62,63) Second, the gut microbiota of ASD infants may be typically different than that of healthy infants of same age, because unlike healthy infants, ASD infants in the present study had high and low abundance of genera Faecalibacterium and Blautia, respectively. Third, the abundance of Faecalibacterium and Blautia, especially the former, may play a role in the dysfunction of systemic immunity as their abundance in fecal microbiota significantly correlated with the expression level of many genes in PBMC, including the genes associated with IFN signaling.

As suggested by other studies,⁶⁴⁾ increased gut permeability is likely involved in the mechanism by which abnormal gut microbiota affects gene expression in PBMC and consequently induces systemic immune dysfunctions, but further investigation of this mechanism was beyond the scope of this study. Nonetheless, it can be theorized that high abundance of Faecalibacterium in the gut microbiota of infants may indicate a mild to moderate gut inflammation, and this condition may be indicative of increased gut permeability as well.⁶⁵⁾ Furthermore, increased gut permeability reportedly results in higher antigenic load,⁶⁴⁾ which may have induced dysfunction of systemic immunity including an inflammatory response of PBMC.

Conclusion

In conclusion, in the present study, we demonstrated that alterations of the gut microbiota play an important role in the dysfunction of systemic immunity and thus the gut microbiota could be at least considered a deteriorating factor of ASD. Although the number of infants evaluated in this study was relatively small, all statistically significant correlations found between the abundance of Faecalibacterium and gene expression levels in PBMC strongly support our conclusions. Further study is required to reveal the exact link between an abnormal gut microbiota and ASD, and the effects caused by abnormal gut microbiota on infants with ASD.

Author contributions

R.I., Y.S., C.S., T.S., and T.T. designed research, Y.S., and C.S. collected clinical samples, R.I., and T.T. performed the experiments, R.I., and M.O. analyzed data, and R.I., and G.A.R. wrote the manuscript. All authors have read and approved the final manuscript.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Funding

This study was supported in part by The Morinaga Foundation for Health & Nutrition.

Supplemental material

The supplemental material for this paper is available at http://dx.doi.org/10.1080/09168451.2016.1222267.

Acknowledgment

The authors thank Dr T. Tsuruta, R. Nishibayashi and M. Nakatani at Kyoto Prefectural University for their technical assistance. The authors would also like to thank Drs B. Ma and M. Humphreys and Prof J. Ravel of the University of Maryland School of Medicine for their assistance in the analysis of sequencing data.