Altered glial marker expression in autistic post-mortem prefrontal cortex and cerebellum

Autism spectrum disorders (ASDs) are neurodevelopmental syndromes defined by impairments in language, verbal and non-verbal communication, and restrictive/repetitive patterns of behavior [1]. ASD symptoms manifest within the first two years of life, with their severity and presentation varying considerably between individuals, thus yielding the 'spectrum’ classification. ASDs are estimated to affect 1 in 88 children in the United States, and the prevalence of ASDs is at least four times more common in males than females [2].

The etiology of autism is complex and the neurobiological mechanism(s) that result in the clinical phenotype remain to be fully understood. However, there is strong evidence that the autism phenotype ultimately results from aberrant synaptic wiring in the developing brain [3]. In particular, studies have shown that long-distance communication between disparate neocortical areas may be disrupted in ASDs, causing delays in information processing within the brain that manifest as the communication, language, and social development problems seen in children with autism [4]. Additionally, parallel research has shown that neuronal micro-circuitry within brain areas may also be disrupted in ASDs, and that this may result in local processing deficits within brain regions related to higher functioning, such as the prefrontal cortex (PFC) [5]. Underlying these circuit disruptions is a large body of evidence that has demonstrated decreased numbers of neurons (and their various subtypes) throughout the autistic brain by early childhood in post-mortem studies [6].

In addition to the body of evidence implicating aberrant local and long-distance synaptic dysfunction in ASDs, many studies have demonstrated microglial and astrocyte dysfunction in ASD brains. For instance, post-mortem pathological studies of autistic brain using immunocytochemistry (IHC) and/or stereology have identified microglial activation patterns [7‐9], and have demonstrated increased microglial cell density in multiple brain regions [8, 10]. Glial activation refers to the well characterized cascade of events that occurs upon reaction of glia to stimuli, resulting in their mobilization, release of cytokines, and ability to phagocytize, among other processes [11]; glial activation is associated with specific gene expression patterns that are distinct from 'resting’ glial gene expression [12]. Furthermore, positron emission tomography (PET) using a microglial-specific radiotracer also demonstrated microglial activation in multiple brain regions of autistic cases [13]. Additionally, studies in a Rett syndrome mouse model, a single-gene deletion disorder with autism as a component, have also demonstrated cellular microglial abnormalities [14], and a remarkable study demonstrated that autistic-like phenotypes can be partially reversed by replacing mutant Mecp2 (-/-) microglia with their respective wild-type cells [15].

Increased numbers of astrocytes, with altered cell size and branching patterns, have also been demonstrated in post-mortem autistic brains [16]. Additionally, astrocyte-specific cell marker proteins are increased in multiple autistic brain regions [17, 18]. Similar to microglial studies in ASD mouse models, astrocytes have been shown to be abnormal in a number of single-gene ASD models, including Rett [19, 20], Fragile X [21], and Tuberous Sclerosis [22]. In parallel to the aforementioned microglial study, it was also shown that replacing mutant astrocytes in Mecp2 (-/-) mice can correct some aspects of the phenotype [23].

However, it is not clear how these separate lines of evidence—one demonstrating immune/glial dysfunction in ASDs and the other implicating synaptic abnormalities—may converge into a common mechanism in the autistic brain that ultimately results in the shared clinical phenotype. Because separate studies have shown that microglia and astrocytes play critical roles in sculpting developing synapses during normal neurodevelopment [24, 25], it is reasonable to hypothesize that inherent defects or aberrant numbers of microglia and astrocytes in the developing autistic brain may be causative of the synaptic abnormalities by affecting the proper wiring of developing neuronal connections. However, because appropriately-preserved post-mortem autistic brain tissue is lacking [26], cellular-level studies assessing glial numbers and activation in human autistic brains have been limited. Moreover, quantification of cell numbers in postmortem tissue by stereology is technically challenging, further limiting the ability of researchers to assess the few appropriately-preserved tissue samples that are available. Finally, no studies have concurrently specifically assessed for microglia, astrocytes, and neurons in the same set of autistic brain samples. As a consequence, a comprehensive understanding of the relationship between glial and neuron cells in autistic brains is needed.

Therefore, the purpose of this study was two-fold. First, we sought to determine if microglia, astrocyte, and neuron-specific markers were altered in post-mortem autistic brain tissue, in order to further investigate the role of glia in ASDs. Then, we determined if glial and neuronal cell-type specific marker expression patterns are consistent with known cellular-level findings, because gene expression studies of post-mortem human brain are often easier to perform than cell-level studies, and therefore this approach may serve as a valuable 'screening’ assay to infer relative cell proportions.

In this study, we compared internally-normalized mRNA expression levels of microglial, astrocyte, and neuronal cell-type specific marker genes in post-mortem brain tissue from patients with autism and healthy controls. Our results provide further evidence for a role of glia in autism pathology, and suggest that assessment of glial cell-type specific markers may serve as a proxy for relative cellular numbers or activation patterns.

The average post-mortem interval (PMI) was not significantly different between autistic and control tissue samples (Table 1; ASD = 17.9 h, ctrl = 13.2 h, P = 0.16). This remained true after sub-stratifying by brain region (ASD PFC = 19.4 h, ctrl PFC = 13.2 h, P = 0.28; and ASD cerebellum = 16.0 h, ctrl cerebellum = 13.25 h, P = 0.48). RNA isolated from post-mortem brain tissue was generally of high quality, and the RNA Integrity Number (RIN) was not significantly different between autism and controls (ASD = 5.84, ctrl = 6.18, P = 0.67). The RIN was also not significantly different after sub-stratifying by brain region (ASD PFC = 4.92, ctrl PFC = 5.80, P = 0.13; and ASD cerebellum = 7.00, Ctrl cerebellum = 6.65, P = 0.85).

In the PFC, quantification of microglial markers demonstrated significantly increased expression in autistic samples of TREM2, DAP12, and CX3CR1, but not AIF1 (Figure 1). The expression of TREM2 was highest of all microglial markers, approximately 1.75-fold higher in autism brain tissue than controls (P = 0.0016). The levels of CX3CR1 and DAP12 were 1.50-fold (P = 0.0092) and 1.34-fold (P = 0.0086) higher in autistic samples relative to controls, respectively. Similarly, the expression of astrocyte marker GFAP was significantly higher in autistic brains (1.70-fold, P = 0.0049). Conversely, however, both the pan-neuronal marker NEFL, and the GABAergic interneuron-specific marker PVA, were significantly lower in autistic samples compared to controls (0.68-fold, P = 0.0034; and 0.52-fold, P = 0.0020, respectively).

In post-mortem cerebellum, the expression of astrocyte marker GFAP was also significantly higher in autism samples than in healthy controls (2.63-fold, P = 0.0022; Figure 2). In contrast, the expression of microglial markers TREM2, DAP12, CX3CR1, and AIF1 were lower in autism tissue than in control tissue, with fold changes of 0.780 (P = 0.0056), 0.797 (P = 0.0083), 0.659 (P = 0.0029), and 0.808 (P = 0.0052), respectively. Expression of neuronal markers PVA and NEFL were also lower in autism samples than in control samples (0.862-fold, P = 0.033; and 0.798-fold, P = 0.013, respectively), as was found in the PFC.

While there have been multiple studies assessing RNA expression levels in autistic tissue, here we report on the expression of microglial, astrocyte, and neuron-specific cell markers concurrently in two regions of autistic brains. Our results show that glial-specific markers demonstrate altered expression in autistic brains. Moreover, the expression pattern of these cell-type specific markers parallels previous findings of glial cell number and activation patterns in autistic brain assessed via cell-level techniques (discussed below). Therefore, this approach may be a useful alternative for assessing activation and/or cell numbers in post-mortem brain tissue studies of ASD patients.

Microglia cell-marker research is still a relatively new area, and thus the markers used to quantify microglial cell number and activation are still debated. To address this issue, we used four different markers that are putatively microglial-specific. Our results demonstrate that in the PFC, there is increased expression of all microglial markers assessed, although AIF1 did not reach statistical significance. However, previous reports have shown that AIF1 expression in the brain is low [32], potentially contributing to this result. The finding of increased microglial cell markers in autistic PFC is in agreement with a number of studies that have found increased numbers and activation of microglia in autistic brains. For instance, Morgan et al. showed increased microglial density in dorsolateral PFC grey matter of ASD brains via IHC and stereology [8], and they also demonstrated that microglia are more closely associated with neurons in autistic dorsolateral PFC than in controls [9]. Similarly, Tetreault et al. also demonstrated increased microglial density in the frontoinsular and visual cortex of autistic brains as compared to controls [10]. Additionally, a number of studies have specifically identified microglial activation in autistic frontal cortex, through both PET radiotracer imaging [13] and IHC/cytokine profiling approaches [7]. Our results provide further support to the growing body of evidence demonstrating increased microglial numbers and activation in autistic PFC, and our cell marker gene expression results are largely in concordance with these cell level studies.

In contrast, our cerebellar results show significantly lower expression of all four microglial cell specific markers in autistic brains. While other studies have identified microglial activation in the cerebellum of autistic tissue [33], no study has attempted to specifically quantify microglial cells in the cerebellum using the markers assessed here, and therefore histopathologic studies in the cerebellum are needed to confirm these findings. One report described increased microglial cell activation in the cerebellum, assessed via HLA-DR staining in the white matter and granular cell layer of the cerebellum [7], and another showed increased microglial activation throughout the brain (although most prominently in the cerebellum) using an in vivo PET metabolic radiotracer [13]. However, HLA-DR expression in human microglial cells has been shown to be highly variable between individuals, and its expression actually decreases upon cytokine stimulation [34]. Moreover, as we previously discussed [35], the cerebellum is anatomically and physiologically unique; thus metabolic and pathological findings in the cerebellum must be interpreted with caution. Furthermore, our tissue samples from cerebellum contained all three layers of the cerebellar cortex, as opposed to the molecular layer only. Until direct histologic assessment of these microglial markers are performed in post-mortem autistic cerebellum, our results must be interpreted cautiously. However, our results do suggest that there are significant differences in microglial cell markers between ASD PFC and cerebellum, perhaps reflecting differences in microglial activation and/or cell numbers between these areas in ASD brain.

In both the PFC and the cerebellum, there was significantly increased expression of the astrocyte-specific marker GFAP in autistic brains. This trend was most prominent in the cerebellum, where GFAP expression was over two-fold higher in ASD brains than in healthy controls. Our findings parallel those of previous studies, which have shown increased expression of GFAP protein in the cerebellum and cortex of patients of autism through IHC staining, western blotting, and mRNA expression [7, 17, 36, 37]. While studies have not been done to quantify astrocyte numbers in the autistic cerebellum, our results and those of previous studies provide evidence for astroglial reaction in autism.

Interestingly, we also found significantly decreased expression of the pan-neuronal marker NEFL in both the PFC and the cerebellum of autistic brains. This result is also supported by previous studies, which have shown decreased NEFL mRNA expression in the anterior cingulate gyrus, motor cortex, and thalamus of autistic brains [38]. However, cell-level studies in autistic brain have produced conflicting results about neuron numbers. While a large body of evidence has suggested there is a loss of neurons in many areas of autistic brains, as recently reviewed in [39], other studies have shown that young autistic brains may have 70% more neurons in the PFC [40]. Importantly, though, is the age of the patient at time of death, as longitudinal studies have suggested that early brain overgrowth in ASDs quickly reverses to a phenotype of neuronal loss [41]. Consequently, the older age of patients in this study may bias our findings towards the neuronal loss spectrum of the disease.

Similarly, we found significant decreases in the GABAergic interneuron-specific marker PVA in both the PFC and cerebellum of autistic samples. Despite many studies demonstrating decreased GABAergic components across different areas of the autistic brain, as reviewed in [42], the one pathological analysis of parvalbumin-positive interneurons in ASD did not identify differences in the autistic cerebellum [43]. However, this study only assessed the molecular layer of the cerebellar cortex, whereas our tissue samples contained all three layers. Additionally, while parvalbumin interneurons have been shown to be unchanged in the autistic posterior cingulate cortex and fusiform gyrus [44], and increased in the autistic hippocampus [45], they have not been directly assessed in the autistic PFC.

In addition to potentially serving as markers of glial cell numbers and/or activation, many of the genes assessed have specific putative biological relevance to ASDs themselves. For instance, in brain, Dap12 (also known as TYROBP) encodes a microglial-specific transmembrane signaling polypeptide [46]. The encoded protein acts as an activating signal transduction element with known roles in brain myelination and inflammation [47]. Its receptor, TREM2, encodes a membrane protein that functions in modulating the brain’s immune response via production of constitutive cytokines, and is critical for activating microglial phagocytosis [48]. Rare mutations within these two genes have been associated with Nasu-Hakola disease [49]. Nasu-Hakola disease is characterized neurologically by new onset psychiatric and cognitive symptoms in the fourth decade of life, evolving to memory loss and cognitive decline resembling Alzheimer’s disease [50]. Interestingly, single nucleotide polymorphisms in this receptor pathway were also recently linked to a significantly increased risk of Alzheimer’s disease, and are thought to relate to the inability of microglia to properly remove neurodegenerative debris such as beta-amyloid [51]. It is intriguing to speculate that defects in this same pathway in neurodevelopmental disorders such as ASDs may also result in defects in microglial phagocytosis, but in the developmental context it is the inability to prune overabundant synapses, as opposed to neurodegenerative debris, that results in the behavioral phenotype.

Similarly, CX3CR1, also known as the fractalkine receptor, encodes a protein receptor on the surface of microglia that binds to the chemokine CX3CL1 (also called fractalkine); fractalkine functions to induce microglial migration and adhesion during phagocytosis [52]. This pathway was recently demonstrated to play a critical role in the developing brain of mice allowing for migration of microglia to their synaptic targets, where phagocytosis and synaptic refinement occur. CX3CR1 knockout mice had more synapses on cortical neurons than wild-type mice, and displayed subtle neurological deficits [25].

AIF1, also known as IBA1, encodes a protein that is consistently up-regulated in expression during microglial activation, and therefore has been used to discriminate between resting and activated microglia [53]. The AIF1 gene is located within the highly variable locus containing parts of the major histocompatibility complex, which itself has been linked to ASDs and dysfunctional microglia phagocytosis [54].

Glial fibrillary acidic protein (GFAP) is an intermediate filament protein that is expressed by astrocytes, and as discussed above, has been previously shown to be up-regulated in autistic brains as was also shown here. Interestingly, decreased expression of GFAP has been reported in other neurodevelopmental disorders, such as Schizophrenia and bipolar disorder [55]. GFAP, like all intermediated filament proteins, is important in maintaining the cellular cytoskeleton in astrocytes. The cytoskeleton plays a number of key functional roles in addition to maintaining cell shape, including inter-cellular communication, mitosis, and cellular migration during phagocytosis. Mutations in GFAP are responsible for Alexander’s disease, a rare disorder characterized by severe developmental delay, increased head size, and seizures [56].

Overall, our findings demonstrate that the autistic brain by mid-childhood has molecular changes consistent with increased astrocyte expression and decreased neuronal and interneuron expression in both the PFC and cerebellum, with PFC-specific increased microglial marker expression. Furthermore, the specific glial molecules found to be abnormal in autistic brains are intimately involved in glial mobilization and phagocytosis pathways, they have previously been shown to be critical for normal neurodevelopment, and are known to be causative of other rare neurodevelopmental phenotypes. These findings support the notion that a complex interplay between glial dysfunction and neurogenesis may underlie the clinical manifestations of ASDs.

This study has a number of limitations of note. Foremost is the relatively modest sample size. Unfortunately, post-mortem human brain research in general is hampered by the lack of accessibility to tissue samples, and pediatric samples in particular are scarce [26]. Therefore, replication with a large number of samples will be important. However, we chose qRT-PCR techniques in this pilot study because of the increased sensitivity compared to whole-genome microarray or sequencing approaches, and therefore some aspects of the small sample size limitation are addressed. Secondly, due to the inter-individual heterogeneity of the brain, and in the brain-banking methodologies used in distinguishing areas of post-mortem brain tissue, it cannot be assumed that all samples will derive from the exact same anatomic site within the PFC or cerebellum; this limitation is largely unavoidable. Lastly, the approach of using cell-type specific marker expression as a proxy for cell number and/or activation still needs verification, by assessing them concurrently with traditional histopathologic/stereology analysis. However, the concordance of our results with previously published studies, and the scarcity of appropriate ASD brain tissue and technical expertise, suggest this may be a valuable and simple alternative 'screening’ approach.

In summary, assessment of glial numbers and activation in autism post-mortem brain research is hampered by the scarcity of appropriately-preserved tissue, and the technical challenge of traditional stereotactic methods. We show that glial and neuron cell-type specific markers have mRNA expression patterns that parallel known cellular aberrations in ASDs. Our results provide further evidence that glial cells may play a role in the pathogenesis of ASDs, and suggest that assessing for glial cell-type specific marker expression may represent a viable approach to relatively quantify glial cell patterns in ASD post-mortem research.