**Sensorimotor Research Utilising Immersive Virtual Reality: A Pilot Study with Children and Adults with Autism Spectrum Disorders**

### **Irene Valori 1, Rena Bayramova 2,**†**, Phoebe E. McKenna-Plumley 3,**† **and Teresa Farroni 1,\***


Received: 14 April 2020; Accepted: 26 April 2020; Published: 29 April 2020

**Abstract:** When learning and interacting with the world, people with Autism Spectrum Disorders (ASD) show compromised use of vision and enhanced reliance on body-based information. As this atypical profile is associated with motor and social difficulties, interventions could aim to reduce the potentially isolating reliance on the body and foster the use of visual information. To this end, head-mounted displays (HMDs) have unique features that enable the design of Immersive Virtual Realities (IVR) for manipulating and training sensorimotor processing. The present study assesses feasibility and offers some early insights from a new paradigm for exploring how children and adults with ASD interact with Reality and IVR when vision and proprioception are manipulated. Seven participants (five adults, two children) performed a self-turn task in two environments (Reality and IVR) for each of three sensory conditions (Only Proprioception, Only Vision, Vision + Proprioception) in a purpose-designed testing room and an HMD-simulated environment. The pilot indicates good feasibility of the paradigm. Preliminary data visualisation suggests the importance of considering inter-individual variability. The participants in this study who performed worse with Only Vision and better with Only Proprioception seemed to benefit from the use of IVR. Those who performed better with Only Vision and worse with Only Proprioception seemed to benefit from Reality. Therefore, we invite researchers and clinicians to consider that IVR may facilitate or impair individuals depending on their profiles.

**Keywords:** autism spectrum disorder; ASD; vision; proprioception; self-motion; immersive virtual reality; IVR; HMD; technology

### **1. Introduction**

Children with Autism Spectrum Disorders (ASD) can present various types of sensory atypicalities including hypersensitivity, hyposensitivity, and unique patterns of response to sensory stimuli [1], higher reliance on unimodal processing [2], and an extended (hence less precise and specialised) multisensory temporal binding window [3]. These are early symptoms that can be associated with a broad range of cascading delays and impairments [4]. Early motor development might also be affected, as it has been hypothesised that the acquisition of body knowledge develops based on our sensitivity to sensorimotor contingencies (action–consequences correspondence) and multisensory contingencies (correspondence between events in different sensory modalities) [5]. When learning a new movement, there is evidence that children with ASD are less influenced by visual feedback [6] and

that they perform better than neurotypical children when the motor learning is driven by proprioceptive input [7]. For instance, the authors asked typically developing children and children with ASD to reach a target by holding a robotic arm. In some random trials, the robotic arm was perturbed and unexpectedly influenced the children's reaching movement. In the following trial, a learning-from-error effect would lead to an altered movement, which was planned to compensate for the perturbation. The perturbation could be presented to children either through visual feedback (displacement of the cursor representing the robotic arm on the screen) or proprioceptive feedback (a force imposed on the robotic arm). Compared to typically developing children, children with ASD showed a higher sensitivity to when learning from proprioceptive feedback and a lower one when learning from visual feedback [7]. Indeed, motor learning occurs thanks to internal models of action: the association between self-generated motor commands (efferent systems) and sensory feedback from the body and the external world (afferent systems), so that it is possible to predict what would happen as the consequence of an action [6]. Information from muscle, joint, and skin receptors constitute our *proprioception*, the awareness of the position and movement of our body in space which is crucial to the production of coordinated movements [8]. Children with ASD show "an abnormal bias towards reliance on proprioceptive feedback from their own bodies, as opposed to visual feedback from the external world", which might predict impairments in motor control, social skills, and imitation ability [9] (p. 10). In learning motor sequences, adults with ASD also show deficits in the use of vision, which is the sense that neurotypical adults rely on, but preserved proprioception-driven learning [10]. Neurotypical adults have been found to experience a postural illusion (which manifests as a forward lean) when exposed to an intermittent vibratory stimulation of the posterior side of the neck, as long as vision was occluded. On the other hand, those with ASD experienced the illusion even when vision was available, demonstrating limited contribution of vision in modulating proprioception [11]. While the majority of research supports this over-reliance on proprioception, some research has contrastingly related motor impairments in ASD to an over-reliance on vision and proprioceptive deficits [12,13]. However, these studies utilised small sample sizes and limited data analyses. Meanwhile, neuroimaging research has shown associations between ASD severity and asynchronous functional connectivity between visual and motor networks in children at rest [14], reduced functional connectivity between visual areas and somatosensory motor networks, and increased connectivity between the cerebellum and sensorimotor areas in both children and adults at rest [15]. The remaining question is whether there is a general trend of over-reliance on proprioceptive over visual cues at the root of sensorimotor atypicalities in ASD. If that were the case, early interventions could potentially be aimed at increasing the reliance on vision in children with ASD, moving them away from this proprioceptively dominant processing. Such training should improve their sensorimotor functioning, potentially leading to benefits for cognitive, social, and communicative skills.

Immersive Virtual Reality (IVR) is particularly appropriate to this end as it allows for controllable input stimuli and the tracking and monitoring of individuals' actions in a safe learning situation where an individualisation of assessment and training is possible [16]. Moreover, this technology makes it possible to manipulate individual sources of sensory information (e.g., visual, vestibular, or proprioceptive) that are physiologically bound together and induce a mismatch between them to study the role of each sensory modality with respect to accuracy in different tasks [17]. For instance, we can disentangle the contribution of visual and proprioceptive inputs to body perception and movement. In this respect, the most promising IVR tools are head-mounted displays (HMDs), which block out the external world, fully immerse the user in the virtual stimulation, and foster a subjective sense of presence in the virtual world [18]. The result is physiological, emotional, and behavioural responses that are consistent with the physical existence of the virtual world [18]. Despite the broad research and intervention potential offered by HMDs, they have unique features that lead to sensorimotor interactions that do not constitute an exact corollary for real-world experience. Valori and colleagues [19] found that self-motion performance worsened in IVR conditions with vision available relative to the same conditions in

reality and indeed, the way that HMDs deliver visual information has essentially unknown effects on movement and its perception [20].

Most notably, the extant literature seems to neglect a developmental point of view, which is only recently being addressed [21]. It seems that technology-driven peculiarities of IVR and HMDs may induce different sensorimotor effects depending on the user's developmental stage, as has been found in research with neurotypical children and adults. Indeed, when neurotypical people have to learn a walking path while wearing an HMD, adults seem not to benefit from multisensory (visual + self-motion) versus unimodal information, while children of 10−11 years old could benefit from the multisensory learning condition [22]. Therefore, we should investigate the interaction between developmental trajectories of users and the peculiarities of technologies. This would make it possible to understand the unique potentialities and limitations that IVR might have for specific populations with typical or atypical development. At the very beginning of the investigation of the potentialities and limitations related to the use of virtual reality tools for individuals with atypical developmental trajectories and sensory, motor, and cognitive atypicalities, 2D non-immersive systems were preferred due to the technological limits of IVR (graphic quality, limited field of view, temporal lag, size and weight, movement restriction, aftereffects of motion sickness, costs, and accessibility) [23]. Although almost two decades have passed, IVR has greatly improved, and HMDs are sometimes used in research and practice with neurodevelopmental disorders; to our knowledge, only one study has investigated the specific aspects of the interaction between atypical development and the atypicality of interacting with virtual environments. Simões et al. suggest that individuals with ASD may show similar social behaviours (i.e., interpersonal distance) in virtual and real environments, even though neurotypical controls differently interact with a real versus virtual person [24]. We hypothesise that HMDs have unique features that are relevant for people with ASD. This technology seems to intrinsically generate a conflict between vision and proprioception and disrupt the reliability of proprioception [19], potentially reducing its hyper-reliance in ASD. Furthermore, HMDs provide visual information that does not perfectly resemble that of the real world, and they might foster the use of the ventral visual pathway (for object qualities) rather than the dorsal pathway (for movement and spatial aspects of stimuli) [25]. This could suit the visual atypicalities of ASD, which are suggested to present impairments in the dorsal pathway [26], allowing individuals with ASD to interact with the world through the visual mechanisms that are most effective for them. However, several issues should be considered when designing virtual environments for specific purposes in sensorimotor research and interventions for individuals with ASD. Firstly, given that there are usually no binocular cues in IVR, action and perception of depth and motion will be achieved through the ventral stream, which will require much heavier input from the ventral stream than in our daily life [25]. Secondly, more research is needed regarding the role of the dorsal stream in the specific sensorimotor deficits in ASD that would be targeted by an IVR paradigm in order to provide the best possible support for the improvement of sensorimotor skills. Indeed, one of the main goals in the field of IVR technologies is to achieve near-real-life binocular motion and depth perception [27,28].

Although IVR applications for people with ASD are growing for educational, entertainment, and treatment purposes, there is a lack of knowledge about how ASD sensorimotor atypicalities and individual variability might lead to different interactive processes and outcomes. Therefore, the present study presents a method that aims at shedding initial light on the differences between moving and perceiving in reality versus IVR for children and adults with ASD. The knowledge gained through this research will be fundamentally important in informing researchers and clinicians who are using this technology with this specific population.

ASD presents a challenge for any individual involved in understanding, assessing, investigating, and treating those with the disorder. The wide variability of patient profiles requires us as researchers to struggle with methodology, embrace the uncertainty of complex phenomena, and be open, thoughtful, and modest in our research practice [29]. Given the contradictory evidence in the extant literature and the innovative aim of the present research, we adopted an exploratory, descriptive approach. As some

statisticians have recently pointed out, "rather than focusing our study reports on uncertain conclusions, we should thus focus on describing accurately how the study was conducted, what problems occurred, what data were obtained" [30] (p. 262). Therefore, the aim of this pilot is to test the feasibility of the experimental procedure with children and adults with high- and low-functioning ASD, as well as to describe data characteristics. We will highlight the importance of exploring inter- and intra-individual differences, which contain meaningful information for assessment and intervention purposes.

In sum, the aim of the present study is to investigate the extent to which the reliability of visual and proprioceptive information aids the self-motion accuracy of children and adults with ASD. To this end, we utilised a self-turn task and manipulated the way visuo-proprioceptive information was provided among unimodal and multimodal conditions. We also aim to explore whether HMD-delivered IVR, compared to equivalent real environments, affects self-motion accuracy, and to find whether the paradigm is feasible for use with this population.

### **2. Materials and Methods**

*Participants.* For this pilot study, we recruited 4 male children (8−13 years old; M = 8.7; SD = 1.2) and 5 male adults (23−39 years old; M = 28.8; SD = 8.3) with a diagnosis of ASD confirmed by their clinicians (see Table 1 for demographic information). The experiment was explained to all parties and informed consent was obtained from parents and professionals responsible for each participant. The study was conducted in accordance with the Declaration of Helsinki, and the protocol was approved by the Ethics Committee of psychology research, University of Padova (Identification code 5A539475A80B5D451B7BC863210C8A61).


**Table 1.** Participants' demographic information.

<sup>1</sup> ADHD (Attention Deficit Hyperactivity Disorder); <sup>2</sup> ODD (Oppositional Defiant Disorder); <sup>3</sup> ID (Intellectual Disability).

*Setup.* Materials and methods have been described in detail in our previous study with neurotypical children and adults [19]. The employed materials included a soundproof, 2x3 metre testing room with black interior walls where small white clouds were randomly fixed (see Figure 1), illumination, audio communication, and videotaping systems, and the HMD Oculus Gear VR 2016 (101◦ FOV, 345 g weight, 60 Hz refresh rate) interfaced with a Samsung Galaxy S7 (152 g weight) providing IVR simulations (360◦ pictures) of the testing room.

*Procedure.* Participants were asked to sit on a swivel chair fixed in the centre of the testing room. For each trial, the experimenter manually rotated the chair a certain degree (passive rotation) from a *start position* to an *end position*. After each passive rotation, participants had to rotate back to the start position (active rotation). Participants' stop position was recorded as the *return position*. The self-turn error was calculated in terms of degrees of absolute difference between the *start position* and the *return position*. Therefore, lower levels of error indicate higher accuracy.

Start, end, and return position data were manually coded by two independent raters of the video recordings. Inter-rater reliability was assessed via intra-class correlation (ICC). The intra-class correlation index (ICC) estimates an ICC = 1, with a 95% confidence interval being 1 < ICC < 1. This nearly perfect inter-coder agreement derives from the small mean difference between the two coders' values within the huge range of possible values (0–360). The mean difference between coder A and coder B is minimal (MA-B = 0.5).

**Figure 1.** The testing room.

*Experimental design and conditions.* In a within-subjects multifactorial (2 × 3) design, all participants were randomly exposed to two trials for each of six conditions (a small number of trials was used to keep the experiment as short as possible for participant comfort). The self-turn task was performed in two Environment conditions (Reality and IVR) for each of three Perception conditions (Only Proprioception, Only Vision, Vision + Proprioception). The IVR conditions involved wearing an HMD that showed 360◦ pictures of perceptually equivalent versions of the reality (R) conditions. The Only-Proprioception (P) condition removed all visual information (with a darkened room or HMD providing no input). The Only-Vision (V) condition limited the access to proprioceptively informative visual landmarks (hiding the participants' body and the room corners) in order to disrupt proprioception, while providing a proprioceptively uninformative visual texture (a pattern of small bright clouds on the walls). The intention was to disrupt proprioception via an alteration of the visual information available without making changes to the proprioceptive information arising from participants' bodies during the passive and active movements. Indeed, previous research has suggested that after being disorientated by a passive rotation in a real environment, people can still detect the position of global landmarks (the room's corners), although they were found to make huge errors in locating surrounding objects [31]. The Vision + Proprioception (VP) condition allowed the participant to access reliable visual and proprioceptive information.

In order to diversify the passive rotations, they were executed both in clockwise and counterclockwise directions, with different amplitudes. Listed below are detailed descriptions of the six experimental conditions.


All the analyses and graphical visualisations were conducted using the software R (version 3.6.1). The data were described through descriptive statistics and graphical representations, and results were interpreted from an exploratory perspective.

### **3. Results**

The first aim of this pilot is to evaluate the feasibility of the experimental procedure with children and adults, even where severe conditions are present. One of the children ("C3", 10 years old) enjoyed the swivel chair and played with it, rotating himself without complying with any verbal instruction provided. Another child ("C4", 13 years old) disliked the testing room and refused to enter it to become familiar with the environment. Data from those participants could not be collected, and the descriptive analyses therefore include seven participants.

The seven participants included here demonstrated that they understood the instructions and task after a short training period. All participants readily wore the HMD. Among them, the two children required several breaks and verbal praise for remaining focused on the task. One of them ("C1") was initially scared by the closing of the room door and by conditions performed in darkness, although he did decide to continue with the experiment. The other ("C2") found the task boring and needed to be continuously motivated. One adult ("A4") performed only the R\_P condition and then exited the room, stopping the experiment. Due to technical issues, another adult ("A1") performed the R\_VP condition twice and did not perform the IVR\_VP condition. The final dataset consisted of 24 observations from children and 50 observations from adults.

The mean self-turn error in the children's sample was 28.4 degrees (SD = 32.3), while in the adults' sample, it was 34.3 degrees (SD = 35.6). The distributions of the observed values have positive skewness, as visualised in Figure 2a,b.

(**a**)

**Figure 2.** *Cont*.

**Figure 2.** (**a**) Distributions of the observed self-turn error. Children (*nparticipants* = 2; *nobservations* = 24). (**b**) Distributions of the observed self-turn error. Adults (*nparticipants* = 5; *nobservations* = 50).

Exploring the main effect of experimental conditions, it is informative to look at individual observations, where we can appreciate that there is heterogeneity of performance (Figure 3).

**Figure 3.** Self-turn error of single observations collected by each participant among conditions (*nparticipants* = 7; *nobservations* = 74).

Means and standard deviations of self-turn error according to age group and the experimental condition are reported in Table 2.

**Table 2.** Means and standard deviations of self-turn error according to age group and the experimental condition.


Note: Standard deviations are reported in brackets. (*nparticipants* = 7; *nobservations* = 74).

(**b**)

Looking at the marginal role of perception and environment factors, we notice that those participants who perform worse in Only-Vision conditions and better in Only-Proprioception conditions seem to benefit from IVR ("A3"; "C1"; "C2"). Those who perform better with Only-Vision and worse with Only-Proprioception seem to be facilitated in Reality ("A1"; "A2"; "A5") (Figure 4a,b).

**Figure 4.** (**a**) Mean error made by each participant according to perception (marginalised over the other variables). (**b**) Mean error made by each participant according to environment (marginalised over the other variables).

Trials were equally distributed among the two possible directions (N = 37 trials in clockwise and counterclockwise directions), which do not appear to affect the self-turn error (Mclockwise = 32.5; SDclockwise = 34.3; Mcounterclockwise 32.3; SDcounterclockwise = 35.1). The amplitude of passive rotations ranges from 67.5 to 205 degrees (M = 137.2; SD = 38.5). Although the effects of amplitude are not of main interest for this study, consistently with our previous findings [19], this variable is positively correlated with self-turn error. This association seems to be qualitatively different among conditions and age groups (Figure 5a,b). Increasing amplitude appeared to reduce children's accuracy to the greatest extent in Only-Vision conditions performed in both Reality and IVR, while it reduced adults' accuracy to the greatest extent in the Vision + Proprioception condition performed in IVR. Further investigation could specifically address this topic.

**Figure 5.** (**a**) Regression lines of self-turn error according to rotation amplitude in each condition. Children (*nparticipants* = 2; *nobservations* = 24). (**b**) Regression lines of self-turn error according to rotation amplitude in each condition. Adults (*nparticipants* = 5; *nobservations* = 50).

### **4. Discussion**

This pilot study offers important initial insights regarding IVR research into the use of vision and proprioception in adults and children with ASD. The first finding with respect to feasibility is that all participants, including lower-functioning ones, readily accepted the use of HMD. Therefore, this appears to be a promising tool for research and treatment purposes in the field of severe ASD conditions, which are commonly understudied [32,33]. However, our experimental procedure requires participants to face some obstacles even when they understand the task and perform at a high level of accuracy. In this pilot study, we found that performance tended to fluctuate between within-condition trials and as such, averaging scores would make it difficult to detect an individual's best performance due to interfering factors such as emotional state, motivation, skills of behavioural management, and fluctuations in attention. Future research could adapt the experiment to build a more engaging, game-like activity and include frequent rewards for participation to create a more attractive testing environment for participants. Moreover, a detailed evaluation of within-participant

outlying performances could be run to detect the best performance the individual can show, rather than an average, which obscures these nuances.

As we only present preliminary data from a small sample, we make no inferential claims here. However, we do find this data informative for modest and cautious considerations. First of all, this methodology could show individual differences in the sensory conditions that facilitate self-motion. Moreover, we could distinguish between the individuals that may benefit more or be more impaired by using HMDs. Within the present sample, those who were facilitated by moving when proprioception was available and no vision was present also benefited from IVR. We cannot generalise this result to the whole population of individuals with ASD, but we strongly suggest that researchers and clinicians keep in mind that this technology can either facilitate or impair individuals depending on their profiles. For instance, an IVR training could be particularly effective for individuals who have reduced reliance on vision in reality. We can speculate that the limited use of external stimuli to calibrate internal body-based information might lead to early motor impairments and therefore stereotypy, which refers to restricted repetitive behaviours and interests which reduce the individuals' learning opportunities and interfere with development [34]. Therefore, future research on the potential of IVR training could select people with reduced use of vision for paradigms aimed at learning within IVR and assess outcomes such as improvements in sensorimotor functions, reduction of stereotypies, and cascading benefits on higher-order cognitive and socio-communicative abilities.

Finally, the present pilot study has some limitations, which call for future research using this promising paradigm. The first limitation is that the experimenter manually rotated the participant, and as such, although experimenters were trained to keep a similar speed and method of rotating, the rotation velocity was not perfectly consistent across trials and participants, which could potentially have influenced participants' performance. The second main limitation was the small sample size, which we plan to enlarge in future studies. This would allow us to explore the effect of other relevant factors such as age, comorbidities, and level of general functioning on individual variability. To this end, we aim to extend our measurements and assess other symptoms that could be associated with visuo-proprioceptive atypicalities, such as sensory profile, fine and gross motor abilities, severity of stereotypies and repetitive behaviours, and communicative and social skills.

The method presented here has been previously investigated with neurotypical children and adults [19]. Bayesian model comparison analyses suggested that the sensory information available and the type of environment might result in a perception x environment interaction effect. Therefore, the role of visuo-proprioceptive information might be different in the two environments. Future studies with individuals with ASD could investigate this interaction effect to explore whether different sensory strategies facilitate self-motion in either reality or IVR. Moreover, in a paper in preparation [35], we have further investigated the memory effect of the rotation amplitude (namely, the amount of information to be encoded and reproduced) of our self-turn paradigm, with findings suggesting that the encoding of own body location is facilitated when vision and proprioception are optimally integrated. Consistent with those findings, the present pilot indicates that rotation amplitude might differently affect accuracy across conditions. Our future research with people with ASD could expand on which experimental conditions are most disrupted by memory load.

There is a long way to go, and the present study is just a first indication. As of March 2020, when searching for "Vision" AND "Proprioception" AND "Autism", Scopus provides only 25 documents. Following the first experimental study published in 1983 [36], there was a gap until 2005 for the next theoretical one [37]. Further experimental research is needed to shed light on this early domain-general sensorimotor mechanism that potentially has huge implications for development.

### **5. Conclusions**

The present pilot study offers preliminary insights into how the self-motion accuracy of children and adults with ASD is affected by individual differences in the way they rely on vision and proprioception, and in how they interact with real environments and IVR. Preliminary results suggest that inter-individual variability in sensorimotor functioning has a meaningful impact on the possibility for people with the heterogeneous conditions of ASD to be facilitated by perceiving, moving, and therefore learning in IVR. Importantly, this research also found this paradigm and the use of an HMD to be acceptable and feasible with the present sample, indicating good potential for future research utilising these methods.

**Author Contributions:** Conceptualisation, I.V., P.E.M.-P., R.B. and T.F.; methodology, I.V., P.E.M.-P., R.B. and T.F.; validation, I.V., P.E.M.-P., R.B. and T.F.; formal analysis, I.V.; investigation, I.V., P.E.M.-P., and T.F.; resources, T.F.; data curation, I.V.; writing—original draft preparation, I.V., P.E.M.-P., R.B.; writing—review and editing, I.V., P.E.M.-P., R.B. and T.F.; visualisation, I.V.; supervision, T.F.; project administration, I.V., T.F.; funding acquisition, T.F. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by Beneficentia Stiftung Foundation.

**Acknowledgments:** Many thanks to Marco Godeas and Carlo Marzaroli for their amazing technical support. Our gratitude to Associazione Pro Musica—Public music school and Comune di Ruda (Udine, Italy) for hosting our laboratory and supporting our activities in the last few years. Thanks to F.lli Budai S.r.l. for building the experimental room: a generous gift for which we are very grateful. Last but not least, our projects on ASD greatly benefited from the contributions of brilliant students: Elena Cassano, Miriam Cerasole, Gloria Chiera, Andrea Marchina; Giulia Monsignore. Gigantic thanks to the participants and their families, for trusting us.

**Conflicts of Interest:** The authors declare no conflict of interest.

### **References**


**Data Availability:** All data files are available from the OSF public repository at the following URL (https://osf.io/dyf2t/?view\_only=746a9829df784d4f9be1312f4e0aa716).

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **P-cresol Alters Brain Dopamine Metabolism and Exacerbates Autism-Like Behaviors in the BTBR Mouse**

**Tiziana Pascucci 1,2, Marco Colamartino 1,2, Elena Fiori 1,2,3, Roberto Sacco 4, Annalisa Coviello 1, Rossella Ventura 1,2, Stefano Puglisi-Allegra 5, Laura Turriziani <sup>6</sup> and Antonio M. Persico 6,\***


Received: 12 March 2020; Accepted: 9 April 2020; Published: 13 April 2020

**Abstract:** *Background:* Autism Spectrum Disorder (ASD) is a neurodevelopmental disorder characterized by deficits in social interaction/communication, stereotypic behaviors, restricted interests, and abnormal sensory-processing. Several studies have reported significantly elevated urinary and foecal levels of *p*-cresol in ASD children, an aromatic compound either of environmental origin or produced by specific gut bacterial strains. *Methods:* Since *p*-cresol is a known uremic toxin, able to negatively affect multiple brain functions, the present study was undertaken to assess the effects of a single acute injection of low- or high-dose (1 or 10 mg/kg i.v. respectively) of *p*-cresol in behavioral and neurochemical phenotypes of BTBR mice, a reliable animal model of human ASD. *Results:* P-cresol significantly increased anxiety-like behaviors and hyperactivity in the open field, in addition to producing stereotypic behaviors and loss of social preference in BTBR mice. Tissue levels of monoaminergic neurotransmitters and their metabolites unveiled significantly activated dopamine turnover in amygdala as well as in dorsal and ventral striatum after *p*-cresol administration; no effect was recorded in medial-prefrontal cortex and hippocampus. *Conclusion:* Our study supports a gene x environment interaction model, whereby *p*-cresol, acting upon a susceptible genetic background, can acutely induce autism-like behaviors and produce abnormal dopamine metabolism in the reward circuitry.

**Keywords:** autism spectrum disorder (ASD); biomarker; *p*-cresol; mouse social behavior; dopamine

### **1. Background**

Autism Spectrum Disorder (ASD) is a neuropsychiatric disorder that begins early in childhood and is characterized by deficits in social interaction and communication, repetitive behaviors, restricted interests, and abnormal sensory processing [1]. The incidence of ASD has dramatically risen during the last few decades, reaching the rate of 1 affected in 58 children [2], making autism one of the most widespread disorders in child neuropsychiatry [3,4]. Both genetic and environmental factors contribute to the pathogenesis of ASD [5,6]. A wide variety of environmental factors have been hypothesized to contribute to ASD pathogenesis, but conclusive evidence has been reached for a small minority, including prenatal infections, some medications (valproic acid, thalidomide, misoprostol, selective serotonin reuptake inhibitors), pesticides, and air pollutants, among others [7].

The complexity of ASD has spurred interest into patient subgrouping strategies, either based on endophenotyping or on biomarkers. Endophenotypes represent familial, heritable and quantitative traits associated with a complex disorder [8,9]. Biomarkers are associated with the disease without necessarily displaying heritability and familiarity; rather, they merely tag for the presence/absence of the disease due to environmental or pathophysiological links, not necessarily of a genetic nature [9]. A reliable set of autism biomarkers could foster earlier and more reliable diagnoses, predict developmental trajectories and treatment response, and identify individuals at high-risk, eventually leading to the establishment of preventive health care strategies, contributing to dissect ASD into more discrete clinical entities, and perhaps even revealing unknown causes of autism, at least in some cases [9].

In recent years, targeted and unbiased metabolomic studies have unveiled a set of potential ASD biomarkers, i.e., small urinary molecules significantly elevated in autistic children [10,11]. Among urinary solutes, *p*-cresol was found to be significantly elevated in autistic children compared to sex- and age-matched controls up until age 8, in two independent samples recruited in Italy and France [12,13]. This finding was later replicated measuring foecal *p*-cresol levels [14,15]. Using an unbiased approach, mass spectrometry-based urinary metabolomics detected *p*-cresol among the 20 solutes best able to differentiate small ASD children from matched controls [11]. Interestingly, elevated urinary *p*-cresol levels were significantly associated with chronic constipation in autistic children, pointing toward slow intestinal transit time as one the main factors allowing greater gut absorption of potentially neuroactive compounds, such as *p*-cresol [16]. The identification of *p*-cresol and of its metabolite *p*-cresylsulphate as two well-known neuroactive uremic toxins poses the question whether, aside from representing a potentially valuable biomarker, the consistent elevation of urinary *p*-cresol detected in young autistic children with chronic constipation may contribute to the clinical severity of their ASD [17]. Preliminary data point toward possible correlations between urinary *p*-cresol concentrations and ASD severity measured using the Childhood Autism Rating Scale (CARS) [12]. Multiple mechanisms could account for the negative influences of *p*-cresol on neural function, ranging from membrane depolarization and increased susceptibility to seizures [18], to decreased Na+-K<sup>+</sup> ATPase activity [19], to blunted conversion of dopamine (DA) to norepinephrine (NE) due to inhibition of dopamine-β-hydroxylase [20].

The studies summarized above spur interest into testing *p*-cresol for behavioral effects in animals carrying a genetic predisposition toward autism-like behaviors. Despite several difficulties in developing rodent models with autistic features [21,22], to date, environmental, genetic, and lesion murine models reproducing autism-like behaviors have been developed [22–26]. The present study aims to assess the effects of acute *p*-cresol in a well-established inbred murine model of ASD, the BTBR mouse [23,27,28]. A single low dose of *p*-cresol (1 mg/kg) significantly raises anxiety and hyperactivity, two frequent ASD comorbidities, while acute administration of a higher dose (10 mg/kg i.v.) also exacerbates core symptoms of ASD, blunting interest in a conspecific intruder and enhancing stereotypic behaviors. Brain region-specific neurochemical analyses link these behaviors to parallel, dose-dependent increases in DA turnover in the AMY, nucleus accumbens (NAc) and dorsal caudate putamen (CP).

### **2. Methods**

### *2.1. Animals*

Every precaution was taken to minimize animal suffering and the number of animals used. For this study, only BTBR T+tf/J male mice were used. Parental strains were obtained from the Jackson Laboratories (Bar Harbor, ME, USA). After weaning at postnatal day (PND) 28, animals were housed 4 per standard breeding cage with food and water ad libitum on a 12:12 h dark:light cycle (lights on

07:00 a.m.–07:00 p.m.). Only male mice were included in the study to avoid possible variability, due to hormonal fluctuations in female mice. Behavioral experiments were carried at PND 60–70 and were performed on the second part of the day (h 01:00 p.m.–06:00 p.m.). Behavioral tests were performed blind to treatment. Mice were habituated to the behavioral testing room for 1 hour before starting the experiment. Tests were conducted in a sound-attenuated room and recorded through a camera (SSCDC378P, Sony, Tokyo, Japan) connected to a computer. Video were analyzed using the EthoVision video tracking software and the Observer XT program (Noldus information technology, Wageningen, The Netherlands) for automatic and manual recording, respectively.

All groups (CNTR, PC1 and PC10) were submitted to the elevated plus maze, open field motor test, object recognition test [29], and three-chamber social interaction test [30,31], in this order. Behavioral testing was performed 15 min after receiving a *p*-cresol/saline injection. Animals were sacrificed by rapid decapitation 100 min after the injection, heads were frozen and brains were removed and prepared for biochemical assay [32,33].

All experiments of this study were approved by the ethics committee of the Italian Ministry of Health and therefore conducted under license/approval ID #: 10/2011-B, according with Italian regulations on the use of animals for research (legislation DL 116/92) and NIH guidelines on animal care.

### *2.2. P-cresol Treatment*

P-cresol was purchased from Sigma-Aldrich (St. Louis, MO, USA), dissolved in saline (0.9% NaCl) and the two different doses (1 or 10 mg/kg) were intravenously delivered by tail vein injection through a micro-cannula to reduce the stress of manipulation. Mice were randomly assigned to experimental groups: (a) naïve, (b) saline-treated controls, and (c) animals that received *p*-cresol 1 mg/kg (P-C1) or (d) *p*-cresol 10 mg/kg (P-C10). Since no difference was recorded between naïve and saline-treated animals, they were grouped together and defined as "control group" (CNTR). Behavior was tested 15 min after the injection.

### *2.3. Elevated Plus Maze*

Emotional reactivity and anxiety-like behaviors were measured using the Elevated Plus Maze, a gray plexiglass apparatus with two open arms (27 × 5 cm) and two enclosed arms (27 × 5 × 15 cm) extending from a central platform (5 × 5 cm).

Animals were individually tested for 5 min, and the total number of entries in the open and closed arms, the percentage of entries in the open arms [(open entries/open + closed entries) × 100] and percentage of time spent in the open arms [(time in open arms/time in open + closed arms) × 100] were automatically analyzed using the EthoVision software.

### *2.4. Open Field Test*

The apparatus consists in a circular open field, 60 cm in diameter and 20 cm in height. Mice were individually introduced in the empty apparatus and left free to explore the arena for 30 min. Videos from each 30-min Open Field Test session were recorded. Distance travelled (cm) and speed (cm/s) were automatically analyzed using the EthoVision software.

### *2.5. Object Recognition Test*

The apparatus is the same as for the Open Field Test (Figure 1C). Each mouse was individually submitted to three 6-minute sessions (Open Field, Pre-Test and Test sessions). At the end of each session, the animal was returned to its home cage for 3 min. All sessions were videotaped and analyzed by an experimenter trained to the Noldus Observer XT event coding software.

**Figure 1.** P-cresol enhances anxiety-like behaviors, stereotypies, locomotor parameters and hinders social preference in BTBR mice. (**A**) Total entries, % of time spent and entries in open arms in the Elevated Plus Maze. (**B**) Distance travelled and speed in the Open Field Test after acute *p*-cresol treatment. (**C**) Schematic representation of the Object Recognition Test. (**D**) Time spent grooming during the first session of the Object Recognition Test. (**E**) Time spent exploring the novel or familiar object during the test session of the Object Recognition Test. (**F**) Schematic representation of the three-chamber Social Interaction Test. (**G**) Time in object and subject zones during the Social Interaction Test session. (**H**) Time spent in contact with the object or with the social intruder during the Social Interaction Test. Results are shown as mean ± sem. \*, \*\*, \*\*\* *p* < 0.05, *p* < 0.01, *p* < 0.001 P-C1 or P-C10 vs. CNTR. ˆˆ *p* < 0.01 P-C10 vs. P-C1, ## *p* < 0.01 old vs. new, §§, §§§ *p* < 0.01, *p* < 0.001 subject vs. object.

During the Open Field session, each mouse was left free to explore the arena for 6 min and time spent grooming was measured.

During the pre-Test session, the mouse was introduced in the arena containing two identical objects (A1 and A2: two identical black plastic cylinders of 8 cm in height and 4 cm in diameter, horizontally fixed to a rectangular base), as shown in Figure 1C, and left free to explore. Total time spent exploring two identical objects (A1 and A2) was measured and analyzed.

For the Test session, both objects were substituted, one with object A3, identical to the previous objects, and the other with the new object B (a red and gray plastic spool: 8 cm in height and 5 m in diameter). Object recognition was evaluated by comparing total time spent exploring the novel (B) vs. the familiar (A3) object.

### *2.6. Three-chamber Social Interaction Test*

The apparatus was a three-chamber box made in plexiglass (Figure 1F). Two transparent partitions (23 cm in height) with removable openings divided the box into three identical rectangular chambers (60 cm × 40 cm). The two external chambers contained two perforated plexiglass cylinders, used to enclose stranger BTBR mice. The test consisted in two 10 min sessions, encompassing the Habituation session and the Sociability Test session. Immediately after the Habituation session the animal was confined to the center chamber while an unfamiliar strain-, sex-, and age-matched adult intruder (subject) or an object were placed inside the cylinders. Videos were recorded and analyzed both automatically and manually, using the EthoVision and Observer XT programs. Time spent in each chamber, time spent in contact with the two cylinders, distance travelled and speed were recorded and analyzed.

### *2.7. Biochemical Assay*

Biochemical assays were performed as previously described [32,33]. Briefly, frozen brains were fixed vertically on the freezing microtome pate. Punches were obtained from 300 μm-thick brain slices (coronal sections). Stainless steel tubes of 0.8, 1.0, or 1.5 mm inside diameter were used. Coordinates were measured as follows: medial pFC, two slices from section 80 to section 130 (1.5 mm tube); NAc, three slices from section 151 to section 201 (1.0 mm tube); CP, 4 slices from section 151 to section 230 (1.5 mm tube); AMY, 5 slices from section 251 to section 350 (0.8 and 1.0 mm tube); HIP, 3 slices from section 301 to section 350 (0.8 and 1.0 mm tube; including CA1, CA2 and CA3 fields). Punches were stored in liquid nitrogen until the day of analysis. Frozen tissues were then weighed and homogenized in 0.05 M HClO4. Homogenates were centrifuged at 14,000 rpm for 20 min at 4 ◦C. Tissue levels of DA, NE, 5-HT and their metabolites were assessed using HPLC. The HPLC system consists of an Alliance (Waters) system and a coulometric detector (ESA Model 5200A Coulochem II) provided with a 5011 high sensitivity analytical cell and a 5021 conditioning cell, the potential being set at 0.450 mV and 0.100 mV, respectively. A Nova-Pack Phenyl column and a Sentry Guard Nova-Pack pre-column were purchased from Waters Assoc. Flow rate was 1 ml/min. The mobile Phase consisted of 3% methanol in 0.1 M Na-phosphate buffer pH 3.0, 0.1 mM, Na2EDTA and 0.5 mM 1-octane sulphonic acid Na salt.

### *2.8. Statistical Analysis*

Behavioral parameters recorded in the Elevated Plus Maze and Open Field Test were analyzed using one-way ANOVAs to detect group effects (three levels: CNTR, P-C1, P-C10), followed by a post-hoc Duncan's test. For the Object Recognition Test, the total time spent exploring the familiar (A3) vs. the novel (B) object during the test session were analyzed by two-way ANOVA for repeated measures ("group", three levels: CNTR, P-C1, P-C10 as between factor; "object", two levels: A3 and B as within factor). Simple effect analysis of the factor "object" was also performed within each group. Similarly, for the Social Interaction Test time spent in each chamber and time spent in contact with the two cylinders were analyzed by two-way ANOVA for repeated measures ("group" three levels: CNTR, P-C1, P-C10 as between factor; "zone", two levels: object and subject as within factor). Distance

travelled and speed by treatment group were analyzed using one-way ANOVA, followed by Duncan's post-hoc test. Data are presented as mean ± sem.

One-way ANOVAs, followed by a post-hoc Duncan's test, were used for statistical analysis of the effects of treatment (three levels: CNTR, P-C1, P-C10) for each amine and metabolite (ng/g wet weight) within each brain region.

### **3. Results**

### *3.1. P-cresol Enhances Anxiety-like Behaviors in BTBR Mice*

The Elevated Plus Maze test is based on the natural inclination of mice to avoid open, elevated and bright places, in spite of their tendency to actively explore novel environments. Results are shown in Figure 1A (CNTR, *n* = 10; P-C1, *n* = 8; P-C10, *n* = 8 mice). The percentage of time spent in the open arms by the CNTR group (17.13%) is consistent with previous studies [34]. P-cresol (1 and 10 mg/kg) profoundly decreases the percentage of time spent in the open arms (F2,23 = 10.632; *p* < 0.001), without significantly affecting the total number of entries (F2,23 = 1.187; *p* = 0.32) and the percentage of entries in the open arm (F2,23 = 1.644; *p* = 0.21). Hence, both low and high *p*-cresol doses increase anxiety-like behaviors in BTBR mice tested using the Elevated Plus Maze.

### *3.2. Locomotor Activity is Enhanced by p-cresol in the Open Field Test*

Results from the Open Field Test are displayed in Figure 1B (CNTR, *n* = 10; P-C1, *n* = 9; P-C10, *n* = 7). Both low- and high-dose *p*-cresol significantly enhanced distance travelled (F2,23 = 5.826; *p* < 0.01) and speed (F2,23 = 5.914; *p* < 0.01) compared to control mice, already yielding hyperactivity at low *p*-cresol doses.

### *3.3. P-cresol Enhances Motor Stereotypies without Modifying Object Recognition and Discrimination Behaviors*

During the first Object Recognition Test session (Figure 1C), time spent grooming was measured (CNTR, *n* = 8; P-C1, *n* = 7; P-C10, *n* = 7). Figure 1D shows that the P-C10 group spent significantly more time self-grooming compared with controls and P-C1 animals (F2,19 = 18.12; *p* < 0.001), who do not differ from each other. A partial dose-dependent shift from hyperactivity to stereotyped behaviors was thus recorded.

Time spent exploring two identical objects during the Pretest session of the Object Recognition Test did not differ between controls and treatment groups (mean ± sem: CNTR = 80.27 ± 6.59; PC-1 = 88.09 ± 6.25; PC-10 = 67.55 ± 11.92; F2,23 = 1.426 *p* = 0.264, data not shown), demonstrating unchanged interest in object exploration. Similar results were obtained during the Test session (Figure 1E), indicating that *p*-cresol does not significantly influence the ability to discriminate novel vs. familiar objects (F2,19 = 0.897; *p* = 0.424).

### *3.4. High Dose p-cresol Thwarts Preference for Social Interaction*

Behavioral results from the three-chamber Social Interaction Test are displayed in Figure 1G,H (CNTR, *n* = 6; P-C1, *n* = 7; P-C10, *n* = 7). No treatment effect was recorded on general motor activity neither during the habituation session (distance travelled: F2,16 = 3.342; *p* = 0.054; speed: F2,16 = 1.544; *p* = 0.237; time spent in each chamber: F2,16 = 0.276; *p* = 0.763), nor during the Sociability Test session (distance travelled, F2,16 = 1.504; *p* = 0.243; speed: F2,16 = 1.572; *p* = 0.229; time spent in each chamber F2,16 = 0.164; *p* = 0.85) (Figure 1G). Time spent sniffing the cylinders did not differ during habituation (F2,16 = 0.263; *p* = 0.77), whereas a significant treatment effect was recorded during the Sociability Test over time spent in contact with the cylinders containing subject vs. object (F2,16 = 6.241; *p* < 0.01). In fact, CNTR and low-dose cresol-treated animals (P-C1) maintained a significant preference for the social stimulus, while high-dose cresol-treated animals (P-C10) lost their social preference, spending

the same amount of time sniffing the two cylinders containing either the conspecific intruder or the object (Figure 1H).

### *3.5. P-cresol Enhances Dopamine Metabolism in NAc, CP and AMY*

Neurochemical data concerning brain levels of monoamines and their metabolites assessed in medial pFC, HIPP, AMY, CP and NAc are summarized in Table 1 and Figure 2 (CNTR, *n* = 9; P-C1, *n* = 6; P-C10, *n* = 6). Significant treatment effects were recorded in NAc, CP and AMY on levels of DA (NAc F3,18 = 21.358; *p* < 0.001; CP: F3,15 = 13.028; *p* < 0.001; AMY: F3,15 = 3.267; *p* < 0.05), HVA (CP: F3,15 = 8.988; *p* < 0.001; NAc: F3,18 = 6.649; *p* < 0.01), and DOPAC (NAc: F3,18 = 9.886; *p* < 0.001; CP: F3,15 = 5.851; *p* < 0.001; AMY: F3,15 = 3.482; *p* < 0.05) (Figure 2B). DA turnover was largely enhanced in NAc and CP and only by high-dose *p*-cresol (P-C10); whereas in AMY, both low- and high-dose *p*-cresol were equally effective (Figure 2B). No significant change was recorded for norepinephrine and 5-HIAA, whereas 5-HT levels were increased only in the CP following the higher dose of *p*-cresol (F2,16 = 8.927; *p* < 0.01) (Table 1). No treatment effect was detected in medial pFC and HIPP for any monoamine or metabolite level (Table 1).


*Brain Sci.* **2020** , *10*, 233

hippocampus,

 amygdala, caudate

**Table 1.**

Neurochemical

 analysis of monoamine

 and metabolite levels (ng/g wet weight) assessed in medial prefrontal cortex,

**Figure 2.** P-cresol enhances tissue levels of dopamine and its metabolites in the amygdala, caudate putamen and nucleus accumbens of BTBR mice. (**A**) Tissue levels of DA, DOPAC, HVA, NE, 5-HT and 5-HIAA, measured in medial pFC, NAc, CP, HIP, AMY. (**B**) Tissue levels of DA, DOPAC, HVA, measured in NAc, CP and AMY. CNTR, *n* = 9–10; P-C1, *n* = 6, P-C10 *n* = 6. Data are expressed as mean ± sem ng/g wet weight. \*, \*\*, \*\*\* *p* < 0.05, 0.01, 0.001 P-C1 or P-C10 vs. CNTR group. ##, ### *p* < 0.01, 0.001 P-C10 vs. P-C1 (treatment effect) by Duncan's post-hoc test following one-way ANOVAs. Abbreviations: AMY: Amygdala; CP: Caudate Putamen; DA: dopamine; DOPAC: 3,4-Dihydroxyphenylacetic acid; HIP: Hippocampus; HVA: Homovanillic acid; NAc: Nucleus Accumbens; pFC: preFrontal Cortex.

### **4. Discussion**

In the present study, acute *p*-cresol administration to BTBR mice, a reliable animal model of ASD [23,27,28], elicited autism-like behaviors and enhanced dopaminergic turnover both in the AMY, and in the dorsal and ventral striatum. Importantly, behavioral abnormalities elicited by *p*-cresol in BTBR mice strikingly resemble core symptoms and co-morbid disorders clinically observed in human autistic individuals. On the one hand, excessive interest in objects over social interaction and stereotypic behaviors represent two of the hallmarks of an ASD diagnosis in humans [1]. Additionally, hyperactivity and anxiety are among the most frequent co-morbidities in autistic patients, with ADHD and anxiety disorders being diagnosed in 33%–37% and in 39.6% of ASD cases, respectively [35,36]. BTBR mice are an inbred strain spontaneously displaying autism-like behaviors [23,27,28]. These behavioral abnormalities likely stem from strain-specific genetic underpinnings involving neurodevelopmental genes, like kynurenine 3-hydroxylase (*Kmo*), Disrupted in Schizophrenia (*Disc1*) and exostosin 1 (*Ext1*) [28]. The induction of hyperactivity in the Open Field Test, but not in the 3-chamber Social Interaction Test, most likely represents only an apparent contradiction, because the more interesting social interaction apparatus is able to engage motivated exploratory behaviors in mice that can "cover" the spontaneous hyperactivity visible in the Open Field Test. In addition, differences in session duration between the two tests (30 min in the Open Field Test vs. 10 minutes in the Social Interaction Test) can further influence the expression of hyperactivity in treated BTBR. Instead, a large body of literature reports a lack of sociability in BTBR using the three-chambered social approach, although data showing that BTBR control mice display significant sociability [37–40] or a non-significant preference for subject exploration are also present (see Figure 1B in ref. [40], Figure 3B in ref. [41], and Figure 3B in ref. [42]). One possible explanation for these discrepancies is that genetically-driven ASD-like behaviors in the BTBR strain may spontaneously be under threshold and may emerge to a different extent depending upon experimental manipulations, handling or treatments [37]. Furthermore, discrepancies due to different choice of intruder (conspecific vs. different strain) in the Social Interaction Test cannot be excluded (in present study we used a BTBR conspecific intruder). Baseline control behavioral parameters recorded in our BTBR mice in the Elevated Plus Maze, Object Recognition Test and Social Interaction Test are absolutely in line with previous studies from our lab [29,32,43,44] and are coherent with the overall literature [45–47], although absolute values predictably differ, likely due to differences in housing environment, animal handling, and test settings. Finally, blunted social preference in the three-chamber test could conceivably stem from enhanced anxiety rather than reflecting a real social interaction deficit. While we cannot exclude contributions by anxiety to this behavior, the emotional reaction of BTBR mice to the objects during pre-test and test sessions of the Object Recognition Test did not differ between groups, as all groups spent the same time exploring objects. Most importantly, both low- and high-dose *p*-cresol produced anxiety-like behaviors in the Elevated Plus Maze. Therefore, if anxiety played a pivotal role in reducing social preference, the lower *p*-cresol dose should have also been effective. In summary, our results collectively support a gene x environment interaction model, whereby, acting upon a susceptible genetic background, *p*-cresol triggers anxiety and hyperactivity at a low dose, while boosting core autism-like symptoms at the higher dose.

Behavioral abnormalities are paralleled by neurochemical alterations, mainly involving the dopaminergic turnover. This interpretation is in line with long-standing evidence of dopamine-β-hydroxylase inhibition by *p*-cresol [20] and with the proportionate increase in DA and its metabolites, supporting increased DA accumulation, release and catabolism (both intra- and extra-cellular). However, the measurable, albeit non-significant, increase in NE recorded in several brain regions displaying increased DA and its metabolites (Table 1) indicates that enhanced DA synthesis may also contribute to cresol-induced dopaminergic imbalance. On the one hand, levels of DA and its metabolites were dose-dependently increased in the ventral and dorsal striatum, where only the higher *p*-cresol dose was effective (Figure 2B). On the other hand, dose-independent effects were recorded in the AMY, where low- and high-dose *p*-cresol were equally effective in boosting DA turnover (Figure 2B). This regional distribution and dose-dependency fit well with the pattern

of behavioral abnormalities recorded in these same animals. Low- and high-dose *p*-cresol were equally effective in reducing time spent in the open arms at the Elevated-Plus Maze and in enhancing locomotor activity (Figure 1A,B). Instead, only high-dose *p*-cresol significantly increased stereotypic behaviors and blunted social interaction(Figure 1D,H). This trend resembles the effects of acute amphetamine in rodents, yielding hyperactivity at low doses and stereotypic behaviors (sniffing and grooming) at higher doses [48,49]. Drosophila melanogaster carrying the ASD-associated hDAT ΔN336 variant, which impairs DA uptake while sparing DA efflux, displays behavioral abnormalities that are strikingly overlapping with those recorded here following acute *p*-cresol—namely increased fear, impaired social interactions, and enhanced locomotion [50]. Modest increases in 5-HT levels parallel the much larger changes observed in levels of dopamine and its metabolites (Table 1). We cannot exclude synergistic serotoninergic contributions to cresol-induced behavioral effects, since 5-HT transporter KO mice display at least some autism-like behaviors, including social deficits and increased anxiety [51]. However, changes in brain 5-HT levels are relatively minor compared to changes in DA and never reach statistical significance, except in the striatum following high-dose *p*-cresol (Table 1). Furthermore, changes in 5-HIAA levels are even more modest, and there is only partial overlap between serotoninergic neurochemical parameters and behavioral changes. Collectively, serotoninergic contributions to cresol-induced behavioral abnormalities may seemingly play a secondary role at best. Instead, our data strongly reinforce the "dopamine hypothesis" of ASD [52], pointing toward the existence in autistic brains of two distinct dopaminergic activation thresholds: a lower threshold in the AMY to boost anxiety and hyperactivity, and a higher threshold in ventral and dorsal striatum to produce stereotypic behaviors and to divert motivational drives from interaction with conspecific animals to inanimate objects. D1 receptor activation or D2 receptor knock-out in the dorsal striatum have been shown to yield autistic-like behaviors in mice [53]. In line with this evidence, BTBR mice display blunted DRD2 signaling and responsiveness to extracellular DA in the presence of preserved DRD2 mRNA and protein levels [54]. On the other hand, comparable DRD1 expression and responsiveness to DA was recorded in BTBR and in C57Bl6 mice [54]. Altogether, much of the current literature on the motivational circuitry in ASD underscores reward-processing deficits towards social and monetary incentives [55,56]. Instead, results displayed in Figure 1H promote a more balanced view, whereby reduced DA activation by social stimuli may be seemingly paired with preserved or even enhanced DA activation by exposure to inanimate objects or by sensory self-stimulation [57–59]. Future experiments will have to extend the present findings, identifying the receptor and signaling pathways mediating the dopaminergic effects recorded in our experiments, and to explore whether the activation of DA turnover by *p*-cresol contributes to favoring LTP-based synaptic plasticity in the NAc [60], possibly fostering "addictive" attitudes towards routines, objects, or absorbing interests including internet and videogames.

Urinary and foecal levels of *p*-cresol have been consistently found elevated in autistic children compared to typically developing controls [11–16]. Preliminary evidence suggests that high urinary *p*-cresol may be clinically associated with greater autism severity and history of behavioral regression [12,17]. P-cresol is not produced by human cells, which lack *p*-hydroxyphenylacetate decarboxylase (pHPAD), the final enzyme of tyrosine transformation into *p*-cresol [17]. Hence, urinary *p*-cresol is either absorbed through the skin, the gut and the lungs from a variety of environmental sources (listed in Table 2 in ref. [17]), or it is produced by gut bacterial strains able to express pHPAD. The primary origin of urinary *p*-cresol elevation in autistic children remains to be determined, as does the reason for its normalization after age 8. However, its association with chronic constipation and longer intestinal transit time supports greater *p*-cresol absorption through the gut, while no association with the "leaky gut" was observed [16]. Chronic constipation thus likely represents a broad, non-specific facilitator of neurotoxic effects exerted by environmental and gut-derived compounds.

The present results raise further interest into *p*-cresol, not only as an ASD biomarker but also as a potential contributor to autism pathogenesis, by boosting DA turnover in specific brain regions of autistic individuals. P-cresol is certainly not the only neuroactive exogenous compound produced by gut bacteria and able to negatively affect behavior. Propionic acid, a short chain fatty acid produced by anaerobic gut bacteria including Clostridia and Propionibacteria, has been shown to produce a variety of behavioral, immune, mitochondrial effects in rodent models closely resembling human ASD [61]. Studies of urinary and foecal levels of propionic acid in autistic children compared to typically developing controls have yielded conflicting results [14,15]. Nonetheless, this compound could indeed play a pathoplastic role in specific patient subgroups, which need to be better defined at the clinical level. Meanwhile, additional tryptophan-derived gut bacterial compounds were found significantly elevated in the urines of autistic children, namely indolyl 3-acetic acid, indoxyl sulfate, and indolyl lactate [11]. These compounds have not yet been thoroughly assessed for possible neuroactive behavioral effects.

### **5. Limitations**

The main limitation of the present study is the lack of a reversal experiment, showing that abnormal behaviors are corrected by administering dopamine receptor antagonists. Due to practical constraints, sample sizes of BTBR mice are relatively small, but 4–5 different litters were used for behavioral experiments and behavioral data appear reasonably consistent among different litters. In fact, all significant differences between control vs. cresol-treated animal mean values displayed in Figure 1 are at least three times larger than inter-litter S.E.M.s per each sample, with the sole exception of the Social Interaction Test (object vs. subject contact time) were P-C10 and controls differ 2.47 times the interlitter S.E.M. values of controls. Repetitive behaviors/restricted interests were assessed only by measuring stereotypic motor activity in the open field test, and not by applying specific tests designed to quantify mouse behaviors corresponding more closely to this diagnostic criterion. Locomotor activity data could have provided additional information if broken down into bins of 3–5 min, allowing an assessment of how quickly the mice habituate to the open field, and the time course of *p*-cresol effects. Finally, urinary baseline levels of endogenous *p*-cresol should be measured and compared among different inbred mouse strains because, if particularly elevated in BTBR mice, they could promote their autism-like phenotypic features and contribute to the behavioral abnormalities induced by exogenous *p*-cresol administration. In addition to addressing these limitations, our follow-up study will involve in parallel both the hypersociable C57Bl/6 mice and the ASD model BTBR mice, to further test the hypothesis that the behavioral abnormalities exacerbated by acute *p*-cresol are the result of a BTBR-specific gene x environment interaction.

### **6. Conclusions**

This study demonstrates that acute *p*-cresol administration to an animal model of ASD induces behavioral abnormalities closely resembling core symptoms of ASD and comorbidities frequently observed in autistic individuals. These results underscore the importance of gene x environment interaction models, able to merge genetic predisposition and evidence-based environmental exposure to specific neurotoxic compounds into a unitary scenario. From a mechanistic standpoint, these results move the field beyond well-established paradigms in the autism literature, such as the imbalance between glutamate and GABA to explain insistence on sameness and the co-morbidity with epilepsy [62], or the role of 5-HT in reference to hyperserotonemia, disruption of circadian rhythmicity, neuroinflammation and neuronal excitability [63–65]. In a complementary view, they point toward critical dopaminergic roles in autistic symptoms as being relevant as stereotypic behaviors, hyperactivity, anxiety and motivational drive towards inanimate objects. Thirdly, urinary gut-derived neurotoxic compounds, such as *p*-cresol, could serve as useful ASD biomarkers, whose specificity now deserves to be assessed in samples of young non-autistic children affected with chronic constipation. Finally, the correction of chronic constipation and microbiota transfer therapy represent two reasonable and testable approaches, aimed at partly ameliorating autistic behaviors by reducing the absorption of neurotoxic compounds of environmental origin or derived from specific gut-bacterial strains [66]. Studies addressing the efficacy of these therapeutic approaches will largely benefit from parallel

assessments of urinary biomarkers, such as *p*-cresol and other gut-derived compounds, in order to provide mechanistic insights into their effects on the longitudinal time course of autistic symptoms.

**Availability of data and materials:** The datasets used and/or analyzed during the current study are available from the corresponding authors on reasonable request.

**Author Contributions:** Conceptualization, T.P., R.S., S.P.-A., and A.M.P.; Data curation, M.C., E.F., A.C., R.V. and L.T.; Formal analysis, E.F., A.C. and R.V.; Funding acquisition, T.P.; Investigation, M.C., A.C. and R.V.; Methodology, M.C., R.S., A.C. and R.V.; Supervision, T.P., S.P.-A. and A.M.P.; Writing—original draft, T.P., E.F. and A.M.P.; Writing—review and editing, T.P., LT. and A.M.P. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by Ministry for Education, University and Research (Grant n. RBFR10RZ0N\_002) to TP and RV and by the Italian Ministry of Health (Grants n. NET-2013-02355263 and CCR-2017-9999901) to AMP.

**Conflicts of Interest:** The authors declare that they have no competing interests.

### **Abbreviations**


### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
