Autism Spectrum Disorder (ASD) is a developmental disorder that is characterized by three behavioral symptoms: repetitive/restrictive behaviors, communication deficits, and social deficits. The prevalence of ASD in school children from kindergarten to grade 11 has been reported to be 0.79%, with classic autism accounting for 0.25% (Lazoff et al., 2010). There has also been a significant linear increase in the prevalence of autism over the time period examined in that study (Lazoff et al., 2010), although the cause of that increase is still unclear. ASD is a genetic disorder, with a 90% concordance rate with identical twins, and a 15–20% risk of autism in siblings (Steffenburg et al., 1989; Bailey et al., 1995; Lamb, 2011). Currently, the Simon’s foundation autism gene database lists 250+ genes that have been associated with autism (Basu et al., 2009). Thus, both the genetics and behavioral aspects of the disorder are quite heterogeneous, and no single gene accounts for more than 2% of ASD cases (Abrahams and Geschwind, 2010). Autism is a highly heterogeneous disorder in humans. Autistic symptoms range from mild to severe in the three diagnostic criteria, and two autistic children might display vastly different phenotypes (Munson et al., 2008; Richler et al., 2010). There is also a lack of consistency in the human neuroanatomical findings. Looking at a specific gene, which has been shown to be associated with autism, allows one to investigate specific changes for a certain subset of autism.

The integrinβ3 (ITGβ3) gene is a subunit of the platelet- and megakaryocyte-specific heterodimeric fibrogen receptor and the widely expressed heterodynamic vitrogen receptor (Weiss et al., 2006a) located on chromosome 17. The role of the ITGβ3 gene in the brain is to control platelet function, cell-adhesion, and cell signaling (Weiss et al., 2006a). The ITGβ3 gene has been shown to be associated with both autism susceptibility (Weiss et al., 2006a; Napolioni et al., 2011) and whole brain serotonin levels (Weiss et al., 2004, 2006a). Further, ITGβ3 is also linked with an additional autism susceptibility gene, namely SLC6A4, the serotonin transporter (SERT) gene (Weiss et al., 2006b; Coutinho et al., 2007; Mei et al., 2007; Ma et al., 2010). Serotonin is a monoamine neurotransmitter that is involved in behaviors such as mood, aggression, sleep, pain sensitivity, cognition, learning, and memory (Sodhi and Sanders-Bush, 2004). Hyperserontonia is one of the oldest and most well replicated findings in autism patients, originally found in ∼30% of autism patients (Schain and Freedman, 1961). Additionally, whole blood serotonin has been identified as a marker in a number of other brain disorders such as bipolar disorder (Hughes et al., 1996), attention deficit hyperactivity disorder (ADHD; Hughes et al., 1996), mental retardation (Partington et al., 1973), and obsessive compulsive disorder (OCD; Hanna et al., 1991). While there is certainly evidence supporting ITGβ3 involvement in other disorders, particularly due to its effect on the serotonin system, it clearly has a strong relationship to autism (Weiss et al., 2006a; Coutinho et al., 2007; Ma et al., 2010; Napolioni et al., 2011).

Serotonin has also been shown to play a role in brain development prior to the time it assumes a role as a neurotransmitter in the mature brain (Lauder, 1990; Chubakov et al., 1993; Whitaker-Azmitia et al., 1996). Serotonin neurons in the brain are one of the most widely distributed neuronal systems (Sodhi and Sanders-Bush, 2004); serotonin neurons stem from four nuclei within the brain stem, the main two being the dorsal and median raphe nuclei. The dorsal raphe nuclei project axons widely to most areas of the cortex, while the medial raphe nuclei preferentially project to specific areas, such as the dentate gyrus, posterior cingulate, and entorhinal areas, as well as the parietal cortex (Kosofsky and Molliver, 1987). A number of the genes that have been associated with ASD have been linked to the serotonin system (Chugani, 2011).

Magnetic resonance imaging (MRI) in human subjects and animal models has been used quite extensively for examining volume changes in the brain. In human autism research a number of meta-analyses on human brain images have highlighted several areas of interest (Stanfield et al., 2008; Frazier and Hardan, 2009; Radua et al., 2011; Via et al., 2011). Stanfield et al. reported that the total brain, cerebral hemispheres, cerebellum, and caudate nucleus were increased in volume, whereas the corpus callosum was reduced in volume (Stanfield et al., 2008). Other more recent meta-analyses have highlighted regional differences in areas that are important for social cognition: namely the hippocampus, amygdala, and corresponding white matter tracts involved in language (Radua et al., 2011; Via et al., 2011). There are many inconsistencies in the results from multiple areas in the brain, however, and the meta-analyses describe many confounding factors, such as the genetic variability in the sample, environmental factors, IQ, and the use of high- vs. low-functioning autistic subjects. The finding of decreased volume or thinning in the corpus callosum is the most consistent finding in human autism (Stanfield et al., 2008; Frazier and Hardan, 2009). Using an animal model, such as the mouse, eliminates some of these confounding factors, as the background genetics and living environment can be controlled more easily. The link to autism in the mouse, however, is often based on the genetics and not the behavior as it is defined in the human population. Autistic behaviors in the mouse are hard to define and are as heterogeneous as they are in humans (Moy et al., 2006, 2007). Using MRI one can detect specific and highly reproducible measurements of relatively subtle volumetric differences, which are shown to be consistent with stereological findings (Lerch et al., 2008a; Spring et al., 2010). These studies have provided maps of the left/right asymmetries in the brain (Spring et al., 2010), sexual dimorphisms (Spring et al., 2007), and learning and memory changes (Lerch et al., 2011b). This same technique has also been applied to human disease models in the mouse, such as Huntington’s disease (Lerch et al., 2008a) and Alzheimer’s disease (Lerch et al., 2005, 2008b). Recently multiple studies have looked at autism related mouse models in order to probe some of the volumetric changes found in human autism in specific genetic backgrounds (Ellegood et al., 2010, 2011; Horev et al., 2011).

An ITGβ3 knockout (KO) mouse was created originally to assess the human bleeding disorder Glanzmann thrombasthenia (Hodivala-Dilke et al., 1999). Due to ITGβ3 being related to platelet function, the KO mice have deficits in platelet aggregation and clot retraction, prolonged bleeding times, and cutaneous and gastrointestinal bleeding. There is also an increase in fetal mortality for the homozygotic mice (Hodivala-Dilke et al., 1999). Recently, the ITGβ3 KO mouse model has been used to assess the behavioral deficits of the model (Carter et al., 2011). The ITGβ3 mouse showed no deficits in activity level during the open field test, or anxiety behavior during the elevated plus maze. However, they did show increased grooming behavior (a repetitive behavior) and in the three chamber social apparatus (Nadler et al., 2004; Yang et al., 2011) the ITGβ3 mice did not show a preference for social novelty (indicative of a social deficit). These behaviors define two of the three core deficits in autism. Note, however, that there is considerable overlap between the repetitive behaviors seen in OCD and autism in humans, and therefore, this increased grooming may be attributable to both an OCD and autistic phenotype.

The purpose of this study was to examine the volumetric differences in the brain of an ITGβ3 homozygous KO mouse model when compared to its corresponding wild-type (WT) mouse using high resolution MRI and detailed statistical analyses to determine the effect of ITGB3 on brain morphometry.

An 11% reduction in total brain volume was found, reflected in 53 of 62 brain regions being significantly smaller (Figure 1). The volume comparisons reported here, therefore, were measured as a relative volume (i.e., a percentage of total brain volume). Thirty-two regions were significantly different at an FDR of <0.05 when the relative volumes were compared. Selected regions are found in Table 1.

A series of coronal images highlighting the relative voxel-wise differences between the brains of the ITGβ3 KO mouse and the corresponding WT can be seen in Figure 2. These maps were created by thresholding the effect size measurement by the TFCE corrected p-value of 0.05; thus only effect sizes of significance are displayed. Numerous changes can be seen throughout the brains where the relative volume is larger (red) or smaller (blue). In the hippocampus, for example, there are drastic bilateral changes around the midline (Figures 2B–D). Further, a bilateral decrease can be seen in the dorsal raphe nuclei (Figure 2F). It should be noted that the dorsal raphe nuclei are included within the periaqueductal gray matter region in our atlas (Table 1).

Two specific structures are highlighted in Figure 3: the corpus callosum (Figure 3A) and the periaqueductal gray matter (Figure 3B). The corpus callosum in the ITGβ3 KO mouse had an absolute volume difference of ∼3 mm³ or −17%, and a relative volume decrease of ∼7%. Volume changes of the corpus callosum along the midline seem to be localized in the posterior or splenium (Figure 3A). The periaqueductal gray matter had a −10% absolute volume difference which was consistent with the overall change in total brain volume, and thus showed no significant relative volume differences. However in two regions within the periaqueductal gray matter there are competing differences (Figure 3B). In the central section of this region there is a relative volume decrease of a size comparable to the decrease located in the dorsal raphe nuclei, which was also seen in Figure 2F. These competing volume differences account for the lack of relative volume difference and highlight the importance of the additional voxel-wise measurements.

A large bilateral decrease in the lateral wings of the dorsal raphe nuclei are visualized in Figure 4, indicating the connection between the ITGβ3 gene and the development of the serotonin system. The bar graph in Figure 4 displays the difference within the indicated voxel.

The hippocampal changes seem to be localized in specific regions (Figure 5), for example there is a large bilateral decrease in the CA1 region of the hippocampus. The differences also seem to follow the dentate gyrus and stratum granulosum more than the hippocampus, which is also evident in the relative volume differences for the three structures (Table 1).

The strongest differences in the cerebellum can be seen in Figure 6, which was thresholded to show only the voxels with the largest differences. Interestingly, differences are localized to two of the deep cerebellar nuclei, namely the fastigial nuclei (red arrows) and the nucleus interpositus (yellow arrow). Also there was a large decrease in the anterior cerebellar vermis (green arrow).

The most striking difference between the ITGβ3 mouse and its corresponding WT was the large difference in total brain volume (Figure 1). The 11% decrease in brain volume was larger than previous studies in mouse models related to autism. In the Neuroligin3 R451C knockin (NL3 KI) an 8% decrease in total brain volume was found (Ellegood et al., 2011). In mouse models of fragile X syndrome (FXS) and 16p11.2 deletion/duplication syndrome, both of which are relevant to autism, no significant total brain volume differences were found (Ellegood et al., 2010; Horev et al., 2011). Furthermore, neuroanatomy differences in the ITGβ3 were more widespread. Relative volume measurements in the ITGβ3 mouse found 30 of the 62 regions to be significantly different at and FDR of <5%. In the other autism studies, significant relative volume measurements at an FDR of <5% were found in 14 regions in the NL3 KI mouse, 11 in the 16p11.2 deletion, 0 in the 16p11.2 duplication, and 1 in the FXS model (Ellegood et al., 2010, 2011; Horev et al., 2011). Contrary to what is found in the ITGβ3, a well replicated finding in human autism is an increase in total brain volume. Although, the ITGβ3 mice used in this study were adult animals, and the human brain overgrowth is reported in children, which has been hypothesized to have an age-dependent effect (Amaral et al., 2008; Stanfield et al., 2008; Anagnostou and Taylor, 2011). In spite of this smaller overall brain volume, relative volume increases in the cortical gray matter were found, specifically in the frontal (+5.5%, FDR < 0.01) and parieto-temporal lobe (+5.4%, FDR < 0.001), as well as in the entorhinal cortex (+7.8%, FDR < 0.001). No significant volume change were found in the occipital lobe (−0.1%, FDR = 0.98). Interestingly, the frontal and parieto-temporal lobes are reported to be increased in the autistic children compared to controls. A similar increase was not found in the occipital lobe (Courchesne et al., 2011).

Neuroanatomical observations in autism has shown reduced neuronal cell size and increased cellular packing density in the hippocampus, subiculum, entorhinal cortex, amygdala, mammilary body, anterior cingulated gyrus, and septum (Bauman and Kemper, 2005). Of these seven regions with notable cellular differences in human autism five had significant relative volume differences in the ITGβ3 mouse, hippocampus (−4%), subiculum (−5%), entorhinal cortex (+8%), amygdala (+7%), and mammilary bodies (+10%); one had no significant difference, the septum; and one was not examined as the cingulate gyrus is included within our corpus callosum region. The reduced neuronal cell size and increased pack density may only account for those regions that were smaller in the ITGβ3 mouse, however. The reported neuroanatomical assessment of these structures could certainly be the cause of these volume difference seen in the ITGβ3 brains. However, it would require an extensive histological examination to determine what the specific driving force is behind these changes, which was beyond the scope of this study. Similarly, within the hippocampus the CA1 regions were noted after golgi staining to have a decrease complexity and the extent of dendritic arbors (Bauman and Kemper, 2005), this could be the reason for the bilateral decrease found in the CA1, shown in Figure 5. The entorhinal cortex has also been linked to schizophrenia and found to be reduced in volume (Falkai et al., 2000). Schizophrenia and autism have a number of overlapping genetic findings (neurexin1 for example; Voineskos et al., 2012), so there might be a similar mechanism causing the difference, although an increase in the size of the entorhinal cortex was found in the ITGβ3 mouse.

Many white matter structures were smaller in the ITGβ3 mouse. Sixteen of the 23 white matter structures in the atlas had significantly smaller relative volumes compared to the WT at FDR < 10%. White matter differences have become common findings in human autism, with the theory that children with autism undergo a period of abnormal white matter development (Ecker et al., 2012). White matter deficits in autism have often been thought of as atypical or under-connectivity (Geschwind and Levitt, 2007), and therefore the large decreases found in the white matter in this model would add weight to that finding. The corpus callosum, the largest most well formed white matter structure in the mouse, had an absolute volume decrease of 17% compared to the WT, which corresponded to a relative volume decrease of 7.1%. This is consistent with a number of volumetric findings in human autism (Cody et al., 2002; Stanfield et al., 2008; Verhoeven et al., 2010). This difference has also been seen in the Neuroligin3 R451C knockin (NL3 KI) mouse model, also associated with autism (Ellegood et al., 2011). In the NL3 KI mouse, the changes in the corpus callosum along the midline were localized in the posterior region, which is consistent with the ITGβ3 mouse seen in Figure 3A. These differences in the posterior part of the corpus callosum have also been reported in human autism populations (Egaas et al., 1995; Piven et al., 1997).

A number of regions affected in the ITGβ3 mouse are closely related to the serotonin system and human autism. Perhaps the most interesting finding was the bilateral decrease in the lateral wings of the dorsal raphe nuclei (Figure 4). The initial formation of the serotonergic neurons is in the brain stem, specifically in the raphe nuclei (Takahashi et al., 1986). From these nuclei the serotonin neurons project to a number of different areas in the brain (Sodhi and Sanders-Bush, 2004). Changes in these nuclei, therefore, could have long lasting effects on many other regions of the brain. Specifically, a number of serotonergic projections from the raphe nuclei travel to the cortex, which could be the cause of the relative volume increases in three of the four cortical structures (Table 1). Also, one of the serotonergic projections to the hippocampus originates from the lateral wings of the dorsal raphe nuclei (Jacobs and Azmitia, 1992), which is the exact area outlined in Figure 4. The smaller dorsal raphe nuclei, therefore, may be causing a corresponding decrease in size in the hippocampus. The hippocampus has also been long reported as a region of interest in autism (Aylward et al., 1999). In this current study, the hippocampal region as a whole is divided into three regions, namely the dentate gyrus, the stratum granulosum, and the residual hippocampus. The residual hippocampus in the ITGβ3 mouse has a relative volume decrease of −4.3%, the dentate gyrus decreased by −7.8%, and the stratum granulosum decreased by −11.8%. The dentate gyrus and stratum granulosum are more affected than the residual hippocampus, and this is also seen in Figure 5. Yan et al. (1997) determined that a reduction of serotonin in the early postnatal period could result in changes in the morphology of the dentate granule cells, particularly in the synaptic spine density. In fact dendritic spine density was reduced by 27% in all ages in that study. This change in the morphology in the dentate granule cells may account for the stronger effect in dentate gyrus than the rest of the hippocampus, but it remains to be seen whether that is the only effect causing the difference.

The cerebellum has also been implicated in human autism, in particular the cerebellar vermis (Courchesne et al., 1988), which has been found to be underdeveloped or decreased in size. Further, a number of changes have been found in the cellular organization of the cerebellum, with reported findings of reductions in the density of Purkinje cells (Palmen et al., 2004; Bauman and Kemper, 2005). Volumetric changes in the cerebellum have been studied quite extensively using MRI in human autism often finding increases in the overall volume of the cerebellum, which was reported to be consistent with the total brain volume change (Amaral et al., 2008). This is in contrast to what was found in the ITGβ3 mouse where the cerebellar changes are quite drastic, with relative volume decreases in the arbor vita of the cerebellum (the white matter part, which included the deep cerebellar nuclei) of −12.8% and the cerebellar cortex of −9.0%. While the cerebellum as a whole was drastically different in the ITGβ3 mouse, a very strong decrease was found in the size of the anterior cerebellar vermis (Figure 6, green arrow). Furthermore, a number a large changes were also localized to two of the deep cerebellar nuclei (Figure 6), namely the fastigial nuclei (red arrows) and the nucleus interpositus (yellow arrow). These two deep cerebellar nuclei were also implicated in FXS (Ellegood et al., 2010), but in that study the difference was more subtle.

Behaviorally, the ITGβ3 KO mouse shows a lack of preference for social novelty (Carter et al., 2011), in the study by Carter et al. (2011) they state that it would be tempting to label this deficit as a hippocampal-associated memory deficit and they indicate that previous work has suggested that social memory may be more dependent on the amygdala and the olfactory bulbs (Ferguson et al., 2001; Adolphs, 2009; Carter et al., 2011). With the differences reported here in the amygdala, hippocampus, and olfactory bulbs (+8% relative volume), it may in fact be a combination of all three regions that is causing this change in social novelty. Also reported in that study was an increase in grooming behavior in novel environments (Carter et al., 2011), and while it may be tempting to label this as an OCD phenotype, there is a considerable amount of overlap between the repetitive behaviors seen in OCD and autism in humans (Zandt et al., 2009). Clinically, the distinction is often based on the sophistication of the compulsion or obsession, which may be inaccessible in the mouse (Zandt et al., 2007). The ITGβ3 could be modeling aspects of both conditions and yield insight into the genetic basis of repetitive behaviors. The increased grooming behavior was thought to implicate a cortico-striatal circuit in mice, and has been reported previously in mouse models of OCD (Welch et al., 2007; Shmelkov et al., 2010). In the Shmelkov et al. (2010) study, the mice had abnormalities in the striatal anatomy, specifically a decrease in volume, and an over activation of the orbitofrontal cortex. Similarly, the striatum in the ITGβ3 KO mouse was also decreased by −3.7% in relative volume and may be the cause of this increased grooming seen in both models. A larger relative volume of the frontal cortex was also found in the ITGβ3 mice (+5.5%), which could be due to the over activation reported in the orbitofrontal cortex in the OCD mice.

Future histological examination of these mice is required to answer some of the open ended questions left by this study. Specifically, what is reason behind the smaller dorsal raphe nuclei? Is it a cellular loss that is causing the difference? Is there a loss in the number of projections from the nuclei? With regards to the hippocampus, are the decreases in the dentate gyrus caused by the morphological change in the synaptic spine density as was hypothesized? Or is there an unknown factor contributing to this difference? The strong changes in the deep cerebellar nuclei are intriguing, specifically as they are related to previous findings in a Fragile X mouse model. In that study it was hypothesize that neuronal loss may contribute to the changes in the nuclei (Ellegood et al., 2010), is that the same for the ITGβ3 mouse model?

In summary, the ITGβ3 mouse model has a number of interesting changes which are related to both autism and the serotonin system in general, and based on the behavioral findings in these mice may also have relevance to other brain disorders, such as OCD. The ITGβ3 KO mouse also appears to have a more severe anatomical phenotype that previously examined mouse models related to autism. The changes in the lateral wings of the dorsal raphe nuclei are particularly intriguing, since that nucleus is so integral to the development of many different brain regions. Further investigation on the cause of some of these differences is required; however, this study was an important first step in the investigation of the neuroanatomy of the ITGβ3 mouse model.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.