Blood–brain barrier and intestinal epithelial barrier alterations in autism spectrum disorders

Background Autism spectrum disorders (ASD) are complex conditions whose pathogenesis may be attributed to gene–environment interactions. There are no definitive mechanisms explaining how environmental triggers can lead to ASD although the involvement of inflammation and immunity has been suggested. Inappropriate antigen trafficking through an impaired intestinal barrier, followed by passage of these antigens or immune-activated complexes through a permissive blood–brain barrier (BBB), can be part of the chain of events leading to these disorders. Our goal was to investigate whether an altered BBB and gut permeability is part of the pathophysiology of ASD. Methods Postmortem cerebral cortex and cerebellum tissues from ASD, schizophrenia (SCZ), and healthy subjects (HC) and duodenal biopsies from ASD and HC were analyzed for gene and protein expression profiles. Tight junctions and other key molecules associated with the neurovascular unit integrity and function and neuroinflammation were investigated. Results Claudin (CLDN)-5 and -12 were increased in the ASD cortex and cerebellum. CLDN-3, tricellulin, and MMP-9 were higher in the ASD cortex. IL-8, tPA, and IBA-1 were downregulated in SCZ cortex; IL-1b was increased in the SCZ cerebellum. Differences between SCZ and ASD were observed for most of the genes analyzed in both brain areas. CLDN-5 protein was increased in ASD cortex and cerebellum, while CLDN-12 appeared reduced in both ASD and SCZ cortexes. In the intestine, 75% of the ASD samples analyzed had reduced expression of barrier-forming TJ components (CLDN-1, OCLN, TRIC), whereas 66% had increased pore-forming CLDNs (CLDN-2, -10, -15) compared to controls. Conclusions In the ASD brain, there is an altered expression of genes associated with BBB integrity coupled with increased neuroinflammation and possibly impaired gut barrier integrity. While these findings seem to be specific for ASD, the possibility of more distinct SCZ subgroups should be explored with additional studies.

Despite research efforts, there are no defined explanations of how environmental triggers can lead to these neurobehavioral conditions. One hypothesis, based on the interconnectivity of the gut-brain axis, suggests that inappropriate antigen trafficking through an impaired intestinal barrier, followed by passage of antigens or activated immune complexes through a permissive bloodbrain barrier (BBB), can be part of the chain of events leading to neuroinflammation and thereby subsequent disease.  The BBB plays a critical role in the central nervous system (CNS) defense through limiting the access of circulating solutes, macromolecules, and cells that could negatively impact neuronal activity. Dysfunctions of the BBB have been associated with numerous neurological disorders, such as stroke, epilepsy, multiple sclerosis, and Parkinson's and Alzheimer's disease [47][48][49][50][51][52][53][54][55][56].
The overall goal of this study was to assess whether a dysfunctional BBB or gut barrier could contribute pathophysiologically to ASD. To address this, we conducted an in-depth molecular analysis of the components associated with the BBB and gut barrier integrity in postmortem brain tissue and small intestinal biopsies obtained from ASD subjects. In addition, we assessed changes in the BBB integrity of patients with SCZ as SCZ is a psychiatric illness in which the involvement of inflammation, immunity and altered BBB integrity have been postulated [22,23,57,58].

Postmortem tissues
Human frozen postmortem brain tissue blocks (1 to 2 cm 3 ) from the frontal cortex (Brodmann's area 45) and cerebellum (CBL) of 15 HC and eight ASD subjects were obtained from the NICHD Brain and Tissue Bank for Developmental Disorders at the University of Maryland, Baltimore, MD, USA, which is a Brain and Tissue Repository of the NIH NeuroBioBank. Additionally, the Maryland Brain Collection of the Maryland Psychiatric Research Center at the University of Maryland School of Medicine, Baltimore, USA, provided ten SCZ samples from each of the two aforementioned brain areas of interest.

Duodenal biopsies
Biopsy specimens were taken from the second portion of the duodenum from ASD (n = 12) and control (n = 9) patients with GI symptoms undergoing esophagogastroduodenoscopy (EGD) for clinically indicated reasons. All study subjects provided written informed consent for the collection of additional biopsies, for the purpose of this study, during the procedure. This study was approved by the Partners Human Research Committee (PHRC) at Massachusetts General Hospital and performed in accordance with the ethical standards and regulatory requirements set forth by the Declaration of Helsinki. W white, AA African-American, HL Hispanic-Latino, M male, F female, HC healthy control, ASD autism spectrum disorder, GBA gut-brain axis study; GERD gastroesophageal reflux disease, GFCF gluten-free casein-free, PMI postmortem interval, ADD/ADHD attention-deficit/hyperactivity disorder, ASCVD atherosclerotic cardiovascular disease, HASCVD hypertensive arteriosclerotic cardiovascular disease, PDD/PDD-NOS pervasive developmental disorder/not otherwise specified, OCD obsessive-compulsive disorder

Gene expression profile analysis
Total RNA was isolated from samples (~100 mg) using Trizol (Life Technologies, Carlsbad, CA) and/or Directzol RNA miniprep spin columns (Zymo Research, Irvine, CA) following the manufacturers' instructions. RNA concentrations and A260/A280 and A260/A230 ratios were measured with the NanoDrop spectrophotometer (Thermo Scientific) before 2 μg total RNA were reverse transcribed using random hexamer primers and Maxima universal first strand cDNA synthesis kit #1661 (Thermo Fisher Scientific, Waltham, MA). Real-time quantitative PCR (qPCR) was used to measure gene expression levels and was performed in an iCycler96X detection system (Bio-Rad, Hercules, CA) according to the manufacturer's protocol. In brain tissues, we analyzed the expression level of genes associated with the formation, integrity, and function of the BBB and neuroinflammation, as follows: Claudin ( Table 1. The sequence specificity of synthesized primers (IDT) were determined by the online Blast program from the National Center for Biotechnology Information website (http://www.ncbi.nlm.nih.gov/blast/Blast.cgi) and compared with the GenBank database. Differences in gene expression profiles were evaluated applying the 2 (−ddCT) method [59].

Protein detection
Western blot analysis was performed as routinely described. Brain tissues (~100 mg) were homogenized on ice in lysis buffer [RIPA Buffer (Sigma, S. Louis, MO)] containing protease inhibitor cocktail (Complete Mini, Fig. 1 Altered gene expression level of TJ components in the cortex of ASD subjects. Each dot represents data from a single subject. Gene expression level is reported as 2 −ddCT with normalization of mRNA expression to the endogenous control 18S. Mean ± SEM are reported for each group. One-way ANOVA test has been used to evaluate statistical significance. *p < 0.05; **p < 0.01; ***p < 0.001 Roche). Homogenates were centrifuged at 14 K rpm for 30 min at 4°C in a benchtop centrifuge. The supernatant, representing total protein lysate, was collected, and protein concentration was determined by Lowry protein assay (DC, BioRad, Hercules, CA) using bovine serum albumin (BSA) as a protein standard. Protein lysates were mixed with loading buffer and reducing agent (both from Life Technologies, Carlsbad, CA) and heated at 99°C for 5 min. The lysates were then loaded on 4-20% Tris-Glycine gels (Life Technologies, Carlsbad, CA) and after gel electrophoresis transferred onto PVDF membranes. The membranes were blocked with 5% BSA in Tris buffered saline with 0.1% Tween 20 (TBST), and incubated with primary antibody overnight at 4°C. Actin was used as loading control while ACTA2, also known as alpha-smooth muscle actin (in this paper reported as SMA), was used for quantitative analysis normalization for tight junction (TJ) proteins. The membranes were washed three times with TBST and incubated with secondary antibody (diluted 1:5000 in 5% BSA-TBST) for 1 h at room temperature. Western blot signal was visualized using a LI-COR Odyssey infrared scan (LI-COR Biosciences, Lincoln, NE). Results were quantified with the ImageJ 1.47 software (NIH). The protein levels were

Subjects
A total of 33 brain sections from the three subject groups were studied. No drug abuse or gunshot wounds were reported for ASD subjects. Two of the ASD individuals were on therapy with one antipsychotic and/or antidepressant. Those ASD cases whose cause of death was drowning were not considered for the brain study to eliminate the damaging effects on brain and vasculature caused by hypoxia/anoxia [60,61]. Of the ten SCZ individuals, no medical records were available. The only existing information was collected at autopsy: the postmortem toxicological analysis showed that three had used antipsychotics and/or antidepressants, one had used cocaine and one had ingested alcohol prior to death. Similar to ASD patients, no gunshot wounds at the head region were reported for SCZ subjects. The control cases had no known neurological disorder or known neuropathology. Three HC out of 15 had a history of drugs and/or alcohol abuse. Biopsies from the second portion of the duodenum were collected from the ASD (n = 12) and control (n = 8) patients with gastrointestinal symptoms. The available clinical characteristics of the analyzed subjects, together with demographic data are summarized in Table 2. Genes associated with BBB integrity and function are differentially regulated in the brain of ASD subjects To assess the integrity of the BBB, we measured the gene expression levels of the major components of the brain microvascular endothelial tight junctions (TJs) by qRT-PCR, in the cortex and CBL of HC (n = 15), ASD (n = 8), and SCZ (n = 10) subjects. In the cortex, CLDN-3, -5, and -12 and TRIC were upregulated in ASD subjects compared to controls (Fig. 1). Neither CLDN-1, OCLN, nor VE-Cad were differentially expressed in the cortex in any of the groups analyzed (not shown). Gene expression profiles of tPA, IBA-1, TSPO, MFSD2A, proinflammatory cytokines, and members of the matrix metalloprotease (MMP) family, described to be linked to BBB function disruption and/or neuroinflammation, were also analyzed. Results showed that in the cortex of ASD subjects, MMP-9 and TSPO were significantly upregulated compared to those of the HC, whereas tPA and IBA-1 appeared downregulated in SCZ compared in HC (Fig. 2). None of the pro-inflammatory cytokines analyzed were differentially regulated in ASD compared in HC. While IL-8 appeared significantly reduced in SCZ vs. HC, all cytokines evaluated were expressed at a lower level in SCZ compared to ASD (Fig. 3). IL-10 and tTG6 expressions were below detection levels. In the CBL, similar to our findings in the cortex, CLDN-5, −12 and TSPO were found to be elevated in ASD subjects when compared to HC (Fig. 4). Among the other genes analyzed, IL-1b appeared significantly higher in SCZ vs. HC (Fig. 4). As a general observation, unlike ASD, SCZ subjects clustered well together for most of the genes analyzed in both brain areas. The gene expression fold changes relative to HC (=1) and the corresponding p values are summarized in Table 3.

BBB tight junction proteins are altered in ASD and SCZ subjects
To understand whether our gene expression results correlated with levels of protein produced, we performed western blotting analysis of tissue lysates using commercially available antibodies against some TJ components in both the cortex and CBL of HC, ASD, and SCZ subjects. Consistent with the gene expression profile, we observed a significant higher level of CLDN-5 in the cortex of ASD subjects, compared in HC (Fig. 5). Interestingly and in this case in opposition to gene expression results, CLDN-12 was reduced in the cortex of ASD and also showed a reduced trend in SCZ subjects (Fig. 5).
In the CBL (Fig. 6), the production of CLDN-5 in ASD appeared slightly higher than in HC (p = 0.1), whereas its level in SCZ was reduced compared to Fig. 4 Increased claudins and inflammatory markers gene expression levels in the cerebellum of HC, ASD, and SCZ subjects. Each dot represents data from a single subject. Gene expression level is reported as 2 −ddCT with normalization of mRNA expression to the endogenous control 18S. Mean ± SEM are reported for each group. One-way ANOVA test has been used to evaluate statistical significance. *p < 0.05; **p < 0.01 that in HC (p = 0.07), although data did not reach statistical significance. No differences were observed for CLDN-12 among the groups.

Increased pore-forming CLDNs and decreased barrierforming TJ components expression in the intestine of ASD patients
To corroborate our hypothesis of a possible involvement of a dysfunctional intestinal barrier in the pathogenesis of ASD, we analyzed small intestinal biopsies of ASD and HC subjects for TJ defects (Fig. 7). Our results showed that 75% (9 out of 12) of the samples analyzed had a decreased expression of at least one barrier-forming TJ component (CLDN-1, OCLN, TRIC), whereas about 66% (eight out of 12) showed increased levels of pore-forming CLDNs, described to increase intestinal permeability (CLDN-2, CLDN-10, CLDN-15) ( [62][63][64][65]).

Discussion
To the best of our knowledge, this is the first study addressing the molecular signature of BBB dysfunctions in ASD and SCZ in human samples. In spite of the obvious small sample number, due to the nature of the samples and the consequent procurement difficulty, we observed statistically significant differences among the study groups. However, it is also worthwhile to acknowledge that the small sample number may have affected some otherwise possibly significant results, which could have emerged if a larger number of tissues had been available.
Our molecular analysis of the BBB integrity and function shows an altered BBB in the ASD subjects evaluated. Indeed, the increased gene expression of MMP-9 we detected in ASD subjects supports our hypothesis of an impaired BBB, most likely associated with neuroinflammation. Several pieces of evidence indicate that MMP-9 secretion induces BBB disruption and this is an important step in the development of inflammatory diseases of the nervous system [66][67][68][69][70][71][72][73][74]. Although we did not observe elevated gene expression levels of proinflammatory cytokines in the ASD subjects analyzed in this study, we found a significantly high expression of TSPO in the ASD brain. TSPO increased expression level in activated microglia and reactive astrocytes [75][76][77][78][79][80][81] is   associated with brain injury and neuroinflammation, making TSPO a reliable marker of brain inflammation. Of the four claudins (i.e., CLDN-1, -3, -5 and -12) that to date are thought to be incorporated in the BBB [82][83][84][85], we found that two were significantly more expressed in the ASD brain as compared in HC. CLDN-5, the major and more abundant cell adhesion molecule of TJs in brain endothelial cells, is expressed 5.1 and 3.6 times higher in the cortex and CBL of ASD subjects, respectively, than in those of HC (Table 3). Similarly, CLDN-12 is expressed about 7.7 and 3.6 times more in the ASD cortex and CBL, respectively, than in the control group. CLDN-3, on the other hand, appears elevated by about 2.7 and 3.0 times in the cortex of both SCZ and ASD subjects, respectively, compared in that of HC. TRIC, another component of the TJs associated with increased epithelial tightness and decreased permeability to macromolecules [86], appears significantly elevated in the ASD cortex, suggesting either a more competent BBB or a compensatory mechanism for disrupted endothelial barrier integrity.
At the protein level, CLDN-5 is elevated in the ASD cortex, consistent with its mRNA expression, whereas, on the other hand, CLDN-12 is significantly reduced. An increased gene expression of sealing TJ components (CLDN-5, -12, and -3 and TRIC) should imply a tighter BBB and a better control in the diffusion of macromolecules. However, our data on MMP9 and TSPO expression compatible with ongoing neuroinflammation suggest a compensatory mechanism to repair a compromised BBB. Another plausible explanation for the CLDN-5 data is that, once produced, the protein is not capable of being incorporated into the TJ complex because of impaired trafficking to the endothelial cell-cell interface or because the protein has some mutations that prevent it from establishing homophilic and/or heterophilic protein-protein interactions, leading to a sustained compensatory gene expression and protein accumulation. On the other hand, the discrepancy between gene upregulation and decreased protein expression observed for CLDN-12 might suggest the existence of a compensatory mechanism for post-translational destruction/degradation of the protein itself. While the sealing function of CLDN-5 in BBB TJs is well established [83,87], that of CLDN-3 and -12 are less understood. CLDN-1 and -3 sequences are highly homologous to CLDN-5, suggesting similar properties. Indeed, several studies strongly suggest that CLDN-3 contributes to BBB tightness [85,[88][89][90]. Conversely, CLDN-12 sequence is quite dissimilar from those of CLDN-1 and -3, and this might indicate a different function. Piontek et al. [91] showed that CLDN-12 cannot form homophilic trans-interactions or homopolymeric strands in vitro, suggesting that "heterophilic cis-and/or trans-interactions with another TJ-protein are necessary for the incorporation of CLDN-12 in strands." It may be hypothesized that the increased CLDN-5 we observed in ASD subjects might be compensatory to a decreased CLDN-12 in a heterophilic cis-and/or trans-interaction between the two claudins.
Our cohort of SCZ individuals did not show major changes in the genes analyzed compared to HC, except for an increased IL-1b in the CBL and significantly reduced levels of tPA and IBA-1 in the cortex. Whereas we do not know the physiological relevance of downregulated IBA-1, our results are in agreement with data from other studies indicating activation of proinflammatory networks and reduced tPA activity in SCZ [92][93][94][95][96][97][98][99][100][101][102], this latter feature leading to a dysfunctional coagulation system and increased risk of thrombotic events in SCZ patients [101]. Other genes analyzed in SCZ appeared to follow a trend similar to ASD subjects (i.e., increased CLDN-3 and reduced CLDN-12 protein level in the cortex). Furthermore, although non-statistically significant (p = 0.07), we did observe a trend toward reduced CLDN-5 protein level in the SCZ CBL as well, suggesting a possibly leaking BBB. Previous clinical and postmortem research in SCZ has suggested that an impaired BBB could be a contributing factor to the pathogenesis of the disease in only a subgroup of SCZ patients [103][104][105]. It is important to mention that the SCZ and ASD subjects studied were not selected on the basis of specific immune biomarkers and, therefore, may represent a mixed group of patients, which thereby limits the statistical significance of some of our findings. A sample stratification approach based on specific immunological features might be a better way of analyzing pathways of interest, and this may be particularly true for SCZ, a disorder characterized by marked clinical and neurobiological heterogeneity. Alternatively, SCZ physiopathology may lie in subtle changes that chronically affect brain function.
As for the intestine, due to the high heterogeneity intrinsic of ASD, our goal was to elucidate the TJ component profile from a functional perspective. Toward this aim, we have analyzed the expression level of TJ elements dividing them into two major functional groups: pore-forming (CLDN-1, OCLN, TRIC) and (See figure on previous page.) Fig. 7 Increased pore-forming claudins and decreased barrier-forming TJ components expression in the small intestine of HC and ASD subjects. Gene expression levels of TJ components in duodenal biopsies from HC (n = 9) and ASD patients (n = 12). CLDN-2, -10 and/or -15 levels are higher in eight out of 12 ASD samples, compared in controls. CLDN-1, OCLN, and/or TRIC levels are decreased in nine out of 12 ASD samples over controls. Each graph represents single patient results. Data are expressed as fold change over the averaged controls barrier-forming components (CLDN-2, -10, -15). These results, showing increased expression levels of poreforming (66% of the ASD samples) and decreased levels of barrier-forming (75% of the ASD samples) TJ components in the duodenal samples, suggest an impaired gut barrier and serve as a proof of concept to support the hypothesis of a gut-brain axis dysfunction in a subgroup of ASD patients. According to this hypothesis, non-self antigens crossing a damaged intestinal barrier elicit a local and/or systemic inflammatory reaction that, Fig. 8 Gene expression in the cortex and cerebellum (CBL) of HC, ASD, and SCZ subjects clustered by functional category. Results are represented as scatter plots where each dot represents data obtained from one subject sample. a Cortex BBB sealing properties group includes CLDN-3, -5, and -12, TRIC, and OCLN. b CBL "BBB sealing properties" group includes CLDN-3, -5, and -12, TRIC, and OCLN. c Cortex BBB function associated markers group includes tPA, MMP2/9, and MSFD2A. d CBL BBB function associated markers group includes tPA and MMP-2/9. e Cortex Inflammation group includes IL-1b, IL-6, and IL-8; TSPO; and IBA-1. f CBL Inflammation group includes IL-1b, IL-6, and IL-8; TSPO; IBA-1. Gene expression level is reported as 2 −ddCT with normalization of mRNA expression to the endogenous control 18S. Mean ± SEM are reported for each group. One-way ANOVA test has been used to evaluate statistical significance. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001 associated with a breach of the BBB, may lead to ASD in genetically predisposed subjects.
This study addressed for the first time, at the molecular level, some relevant and still open questions regarding the gut-brain axis in neurobehavioral disorders and furthermore demonstrated that both the BBB and gut barrier may be affected in ASD. However, these findings present with a few limitations. The first lies in the postmortem nature of the samples with potential confounding factors deriving from tissue preservation and the personal history of each subject, including exposure to medications, health conditions, cause of death, and postmortem interval (PMI). Our statistical analysis of the PMI showed no differences among groups suggesting that PMI might not have affected our results. Similarly, data from medical records or postmortem toxicology analysis showed that about the same number of subjects in ASD and SCZ groups were on antipsychotics/antidepressants, so it is unlikely that the differences we observed among groups could be explained by the effects of the drugs. For clarity, a few studies addressing the impact of antipsychotics on the BBB function do suggest a disturbance effect on its sealing property [106,107], whereas other studies on the blood-cerebrospinal (BCS) fluid barrier show divergent results [103,108]. Furthermore, there are numerous studies reporting the suppressing effect of antipsychotics on the blood cytokine profile in SCZ and ASD [98,[109][110][111][112][113][114][115][116][117][118], as well as on microglia activation [119,120], that could explain our data on IBA-1 and the cytokine gene expression profiles. However, our IBA-1 results are well in agreement with the published studies on postmortem brain tissues showing no significant changes in its expression in ASD [121].
The second limitation of this study is represented by its descriptive nature that does not allow us to draw conclusions on the causality of a breach of the intestinal/BBB integrity supporting neuroinflammation and ASD and SCZ symptoms. However, our data is in line with animal studies suggesting causality between loss of barrier function and neurobehavioral changes compatible with ASD [122][123][124]. Finally, as already mentioned, a third limitation is represented by the small number of samples analyzed due to limited availability which might impact on the statistical significance of some of our results. However, given the complexity of the studies performed and the difficulty associated with the procurement of these samples, our study represents a solid foundation to justify a more robust sample collection and analysis to further investigate the gut-brain axis in ASD.

Conclusions
In conclusion, our results seem to point to a dysfunctional gut-brain axis associated with neuroinflammation in ASD. When clustered together by functional groups (barrier properties, pro-inflammatory markers, and enzymatic activity), our data support the notion that in ASD, there is a differential regulation of the pathways associated with our hypothesis of a gut-brain axis dysfunction involving the intestinal barrier, BBB integrity/ function, and neuroinflammation (Fig. 8). Certainly, more in-depth molecular studies, possibly in animal or in vitro cell culture models, are necessary to understand the specific mechanisms behind the BBB disturbance we have herein reported. These studies will allow for patient stratification and personalized interventions (precision medicine) to target specific pathways involved in the pathogenesis of ASD. the specific collection was approved by the Maryland Psychiatric Research Center (MPRC), Maryland Brain Collection Steering Commitee, and the NIH NeuroBioBank Tissue Access Commitee. The specimens were collected under specific protocols, and the family consent was collected for the tissue to be used for scientific purposes including publication. Study procedures for the collection of the duodenal biopsies were reviewed and approved by the Partners Human Research Committee (PHRC) at Massachusetts General Hospital and performed in accordance with the Declaration of Helsinki. The PHRC operates in full compliance with all applicable federal, state, and local laws and regulations to protect the rights and welfare of the human subjects. Written informed consent was obtained from all the subjects at the time of enrollment.