Imaging evidence for disturbances in multiple learning and memory systems in persons with autism spectrum disorders

Authors


  • This article is commented on by Santosh on pages 206207 of this issue.

Dr Suzanne Goh at Columbia University Medical Center, 1051 Riverside Drive, Unit 74, New York, NY 10032, USA. E-mail: gohs@nyspi.columbia.edu

Abstract

Aim  The aim of this article is to review neuroimaging studies of autism spectrum disorders (ASD) that examine declarative, socio-emotional, and procedural learning and memory systems.

Method  We conducted a search of PubMed from 1996 to 2010 using the terms ‘autism,’‘learning,’‘memory,’ and ‘neuroimaging.’ We limited our review to studies correlating learning and memory function with neuroimaging features of the brain.

Results  The early literature supports the following preliminary hypotheses: (1) abnormalities of hippocampal subregions may contribute to autistic deficits in episodic and relational memory; (2) disturbances to an amygdala-based network (which may include the fusiform gyrus, superior temporal cortex, and mirror neuron system) may contribute to autistic deficits in socio-emotional learning and memory; and (3) abnormalities of the striatum may contribute to developmental dyspraxia in individuals with ASD.

Interpretation  Characterizing the disturbances to learning and memory systems in ASD can inform our understanding of the neural bases of autistic behaviors and the phenotypic heterogeneity of ASD.

Abbreviations
ASD

Autism spectrum disorders

MNS

Mirror neuron system

STS

Superior temporal sulcus

What this paper adds

  •  We review neuroimaging studies of ASD examining declarative, procedural, and socio-emotional learning and memory systems.
  •  A neural systems approach to understanding the phenotypic heterogeneity of ASD is adopted.
  •  We discuss how disturbances in learning and memory may relate to the core features of ASD.

In 1943, Leo Kanner1 published his seminal manuscript describing a group of children with a ‘powerful desire for aloneness and sameness.’ The key themes identified by Kanner have since been refined into a set of three nosological criteria that serve as the criterion standard for diagnosing autism spectrum disorders (ASD): disturbances in (1) social interaction, (2) communication, and (3) behavioral flexibility. How these different phenomenological criteria ‘map’ onto the brain has been a topic of intense investigation in recent years, with over 300 neuroimaging studies of ASD published in the past three decades.

Although the neural bases of the individual features of the autistic triad have yet to be clearly defined, neuroimaging studies have provided key insights into the brain disturbances present in individuals with ASD. Perhaps the most important discovery has been a trajectory of early brain overgrowth within the first years of life followed by premature growth arrest.2 This aberrant growth occurs at a critical time in human brain development, during which synaptogenesis, neuronal growth and differentiation, and myelination establish networks of local and long-range connections that are essential for normal cognitive and behavioral development. Neuroimaging studies have also identified several neural systems of interest that may be associated with certain autistic behaviors. Deficits in certain social behaviors, for example, have been associated in some studies with abnormalities of the amygdala, mirror neuron system (MNS), and superior temporal gyrus, whereas stereotyped repetitive behaviors have been associated with abnormalities of the striatum.3,4

Neuroimaging studies of ASD, however, face a number of limitations. Abnormalities of brain function are known to be present in early infancy in ASD, long before most imaging studies are performed; therefore, the findings will reflect not only congenital abnormalities but also their effect on subsequent brain development and the effects of altered interaction with the environment. The interpretation of neuroimaging studies of ASD is further constrained by a host of other factors, including substantial differences among studies in diagnostic and exclusionary criteria, age composition, the profile of comorbidities and medication use among individuals with ASD, control groups (normal vs developmentally delayed), measures of symptom severity, and imaging methodology, including different neuroanatomical definitions of brain structures. In addition, the cross-sectional design of most studies means that any conclusions about developmental trajectories will be speculative.

Genetic and molecular approaches to the study of ASD have generated important findings that may help guide the interpretation and design of neuroimaging studies. Most autism-susceptibility genes thus far identified participate either in neural migration or in synapse formation and function (i.e. NLGN3/4, NRXN1, CNTNAP2, SHANK3).5 Altered neural migration and synaptogenesis in the first years of life could establish aberrant neuronal architecture in widespread neural networks. Moreover, disturbances to molecular pathways involved in synapse formation and function could lead to ongoing problems with synapse maturation, connectivity, and stabilization.

In this paper we explore the hypothesis that such abnormalities of synapse formation and function would be likely to affect multiple neural systems involved in learning and memory. Which systems are affected, in what way, and to what degree may explain in part the phenotypic heterogeneity of ASD. The aim of this paper is to review neuroimaging studies of long-term learning and memory systems in individuals with ASD. We first provide a classification of long-term learning and memory systems, and a summary of current knowledge regarding their neurobiological bases. We then address each system in turn to review studies of ASD that have correlated long-term memory functions with neuroimaging measures of the brain.

Multiple Long-Term Learning and Memory Systems

When used in everyday language, the term ‘memory’ usually refers to declarative memory, or the capacity for conscious recollection of facts and events. A vast body of experimental research in animals and humans, however, demonstrates that learning and memory comprise a wide range of relatively independent functions based within different neural systems in the brain (Fig. 1).6 Neuroimaging studies of long-term learning and memory systems in ASD have examined the declarative, procedural, and socio-emotional learning and memory systems.

Figure 1.

 A taxonomy of long-term memory systems and their neurobiological substrates.6

Declarative learning and memory

The components of declarative memory (semantic and episodic) differ in that semantic memory is concerned with factual knowledge, whereas episodic memory is concerned with recollection of previous experiences of events. The hippocampus and adjacent regions of the medial temporal lobe function in declarative learning and memory. These regions enable the binding together of multiple sensory inputs to permit representations of the relationships between constituent elements of scenes or events. The specific ability to bring together multiple sensory inputs to form complex representations has been termed ‘relational or associative memory’. The hippocampus has robust projections to prefrontal cortices, and this network may enable the use of organizational strategies in the service of declarative memory.

Procedural learning and memory

Procedural memory is a form of implicit memory that facilitates the acquisition of motor skills and habits. The earliest experimental evidence for procedural memory came through studies of the amnesic participant HM, who had undergone bilateral surgical resection of both medial temporal lobes but retained the ability to learn a hand–eye coordination skill (mirror drawing).7 A neural system involving the basal ganglia, cerebellum, and parietal and premotor cortices has been shown in animal and human studies to enable the acquisition and retention of skilled motor behaviors.8 Parietal regions (i.e. left supramarginal and angular gyri) participate in the storage of spatial representations of complex motor gestures, whereas premotor regions (i.e. supplementary motor areas) engage in the translation of these representations into motor programs. Corticostriatal circuits are important for the learning of motor sequences, and corticocerebellar circuits participate in the adaptation of movement to environmental changes.

Socio-emotional learning and memory

A wealth of human and animal research strongly implicates the amygdala in the evaluation of and response to emotionally salient stimuli. The amygdala mediates preferential attention to emotionally meaningful stimuli and enhances memory for emotionally valenced information. Although studies of primates with early bilateral amygdalectomy demonstrate a heightened fear response but otherwise typical social behaviors, humans with amygdala injury, particularly in early childhood, exhibit impaired recognition of social emotions and theory of mind.9–11 An amygdala-based learning and memory system includes connections, direct and indirect, to primary sensory and association cortices, brainstem and hypothalamus, ventral striatum, hippocampus, and orbitofrontal cortex.12

Several specific cortical regions are also believed to play a role in socio-emotional learning and memory: the MNS, the fusiform face area, and the superior temporal sulcus (STS) and surrounding cortices.

Mirror neurons are a class of neurons located within the premotor and parietal cortices that are active when an individual performs a goal-directed action or observes another individual perform a similar action. Functional imaging and neurophysiological studies in humans provide evidence for a mirror neuron network involving the STS, the pars opercularis of the inferior frontal gyrus, the inferior frontal cortex, and the inferior parietal lobule.13 Because the MNS is similarly active when an individual executes or simply observes an action, it is believed to enable not only imitation, but also our ability to link the physical and mental experiences of the self and others. In this way, imitation may be the necessary precursor for a ‘theory of mind’– the ability to recognize that others have thoughts, feelings, and intentions that are different from one’s own.

The fusiform face area, located on the ventral aspect of the temporal lobe, has been shown in functional magnetic resonance imaging (MRI) studies to be active in processing the structural, static properties important for facial identification. Notably, individuals with fusiform face area lesions have prosopagnosia (an impaired ability to recognize others by the face but a preserved recognition of facial expression), which differs from the pattern of deficit in individuals with ASD. The role of the fusiform face area also extends beyond the processing of faces to include, more generally, visual stimuli with which the viewer has expertise.

The STS and surrounding cortical regions play a role in processing information about facial movements and configurations, such as facial expressions, gaze shifts, mouth movements, and other body motions. This region appears to play a role in interpreting the social content of human movement.14 The STS also functions in the auditory processing of voices, with central and anterior regions showing greater specificity for processing human linguistic vocalizations.15

Long-Term Learning and Memory Systems in ASD

Declarative learning and memory

Studies of declarative memory in individuals with ASD provide evidence for selective impairment of episodic and relational memory, with relative preservation of semantic memory.16 Thus, individuals with ASD sometimes exhibit a pattern of superior performance in tasks requiring rote memory but show impaired ability to utilize context and meaning in the service of memory. This impairment in relational memory, termed ‘weak central coherence,’ is postulated to be a fundamental deficit in ASD and may be secondary to disturbances in hippocampal–prefrontal circuits.

Three neuroimaging studies have examined correlations of declarative memory function with hippocampal structure in individuals with ASD (Table I). The largest of these studies examined 45 children with ASD and a comparison group of 13 sex-matched individuals. Participants underwent structural MRI and a test of medial temporal lobe function (delayed non-matching to sample task).17 The group with ASD showed an alteration of hippocampal shape consistent with inward deformation of the subiculum, and the magnitude of deformation correlated with deficits on the delayed non-matching to sample task and with autism symptom severity (as measured by the Autism Diagnostic Interview – Revised and Autism Diagnostic Observation Schedule – Generic). A study of 14 adolescents with ASD and 18 comparison individuals, matched by age and Verbal IQ, showed significant impairment in episodic memory in the group with ASD as well as significantly greater gray matter density in the hippocampi and perihippocampal cortices; however, performance on tests of episodic memory did not correlate with the degree of abnormality of hippocampal structure.18 A study of 10 adult males with ASD and 10 comparison adult males, matched by age and Performance IQ, showed no memory impairments in the group with ASD and no correlations of hippocampal volume with measures of declarative memory or measures of autism symptom severity.19

Table I.   Neuroimaging studies of autism spectrum disorders (ASD) correlating memory function with imaging features of the brain (1996–2010)
System(s) studiedStudy (y)Diagnostic instrumentASDComparison group(s)Memory measure(s)
n (M:F)Mean age (SD)Mean IQn (M:F)Mean age (SD)Mean IQ
  1. ADI-R, Autism Diagnostic Interview-Revised; ADOS, Autism Diagnostic Observation Scale; AD, autistic disorder VIQ, Verbal IQ; NVIQ, Non-verbal IQ; PDD-NOS, pervasive developmental disorder – not otherwise specified; ASAS, Australian Scale for Asperger Syndrome; WADIC, Wing’s Autistic Disorder Interview Checklist; PIQ, Performance IQ; FSIQ, Full-scale IQ.

DeclarativeDager (2007)17ADI-R ADOS45 (38:7)47.9mo (4.2mo)AD:VIQ 46.6,
NVIQ 59, PDD-NOS:VIQ 64.8
NVIQ 74.7
13 (10:3)44.4mo (5.9mo)n/aDelayed non-matching to sample task
DeclarativeSalmond (2005)18ASAS14 (13:1)12.9y (0.7y)VIQ 10218 (6:12)12.6y (0.7y)VIQ 104Tests of episodic and semantic memory (i.e. Rivermead Behavioral Memory Test)
Declarative and socio-emotionalBoucher (2005)19WADIC10 (10:0)23y 9mo (7y 9mo)VIQ 105.5, PIQ 90.310 (10:0)24y 2mo (8y 1mo)VIQ 104.4, PIQ 97.5Tests of declarative memory and socio-emotional perception
ProceduralQiu (2010)22ADI-R ADOS32 (32:0)10.2y (1.7y)FSIQ 101.945 (45:0)10.4y (1.2y)FSIQ 116.2Florida Apraxia Screening Test
Socio-emotionalHoward (2000)23WADIC10 (10:0)Not reported (range:15.8–40.3y)n/a10 (10:0)Not reportedn/aBenton test of facial recognition; eye-gaze perception; facial emotion recognition; Warrington recognition memory task
Socio-emotionalDziobek (2010)25ADI-R27 (20:7)42y (11.3y)FSIQ 11129 (22:7)44.9y (14.6y)FSIQ 113Facial emotion recognition
Socio-emotionalNacewicz (2006)24ADI-R12 (12:0)16.8y (4.5y)n/a12 (12:0)17.0y (2.9y)n/aFacial emotion recognition
16 (16:0)14.3y (4.7y)FSIQ 9714 (14:0)13.7y (3.9y)FSIQ 122Distinguishing familiar and unfamiliar faces

The broader neuroimaging literature examining hippo-campal volume in ASD has not consistently demonstrated differences in overall hippocampal volume between ASD and comparison groups.20 Although the findings from the three studies cited above are preliminary, they suggest that more detailed anatomical and functional measures of hippocampal subregions and parahippocampal regions may show between-group differences not reflected in gross volumetric measures. These more refined measures may help to clarify the neurobiological basis of the episodic and relational memory impairment in ASD.

Procedural learning and memory

Few studies have examined procedural learning and memory in individuals with ASD; however, some evidence exists for a developmental dyspraxia in ASD – an impaired ability to learn, store, and recall spatio-temporal representations of complex movements. This feature appears to be disproportionate to the degree of impairment in basic motor control or imitation and, in one study, was shown to correlate with the severity of impairment in social interaction.21

No imaging studies of ASD have attempted to correlate performance on tasks of procedural learning and memory with structural or functional features of cortico-striatal circuits. One study, however, examined basal ganglia shape abnormalities and associations with motor praxis in 32 male children with ASD and 45 comparison children, matched by age and sex.22 In the ASD group, surface inward deformation of the bilateral anterior and posterior putamen predicted poorer praxis (as measured by a version of the Florida Apraxia Screening Test adapted for children; Table I).

The broader neuroimaging literature examining basal ganglia structure in ASD has consistently reported enlargement of the caudate nucleus, covarying for total brain volume.20 Although these studies have not examined praxis or procedural learning and memory, they have identified an association between enlargement of the striatum and the severity of repetitive behaviors.4 The relationship between repetitive behaviors and procedural learning deficits, however, remains unclear.

Socio-emotional learning and memory

Disturbances of social cognition – encompassing social perception, understanding, and behavior – are by definition universally present in ASD. Many neuroimaging studies of ASD have examined correlations of brain structure or function with ratings of social and communicative impairment, which are generally acquired by a caregiver report or by a clinician’s interpretation of observed behaviors during an office-based assessment. Only four studies have correlated anatomical features of the brain with direct measures of socio-emotional learning and memory (Table I).

A study of 10 adult males with ASD and 10 comparison males, matched by age and Performance IQ, examined correlations of amygdala volume with measures of socio-emotional learning and memory (i.e. tasks of facial emotion recognition, eye direction detection, and face recognition).19 The ASD group performed significantly worse on these tasks and had significantly larger amygdalae than the comparison group, but performance on tests of socio-emotional learning and memory did not correlate with the degree of amygdala enlargement. Another study of 10 males with ASD and 10 sex-matched comparison children detected significantly worse performance in the group with ASD on tests of memory for faces, eye-gaze perception, and recognition of facial emotion.23 The group with ASD also exhibited a significantly increased volume of the amygdalae. Another study examined two separate groups with ASD (and a matched comparison group) to assess correlations of amygdala volume with tests of socio-emotional learning and memory.24 In the first group, 12 males with ASD and 12 age- and sex-matched comparison children underwent structural MRI and a task that required them to distinguish neutral from emotional facial expressions. No significant differences were detected in amygdala volumes or in performance on the emotion recognition task between groups; however, in the ASD group, smaller amygdalae predicted a significantly slower judgment time for emotional expressions. In the second group, 16 males with ASD and 14 age- and sex-matched comparison children underwent structural MRI and a task requiring differentiation of familiar and unfamiliar faces. Individuals in the group with ASD were less accurate in their judgment of familiar and unfamiliar faces, and amygdalae were significantly smaller in the group with ASD than in the comparison group. In the only study to examine both amygdala volumes and the cortical thickness of the fusiform gyrus, 27 adults with ASD and 29 comparison individuals, matched by age, sex, and Full-scale IQ, underwent structural MRI and a test of facial emotion recognition.25 No difference in amygdala volume was detected between the groups; however, the group with ASD showed a significantly increased cortical thickness of the left fusiform gyrus, which predicted impairments on the facial emotion recognition task.

Overall, the differences in age composition, the use of different measures of socio-emotional memory, and the variability in imaging methodology and in the profile of comparison groups make the findings from these studies difficult to reconcile. Age composition, in particular, has been shown to influence findings related to amygdala volume in ASD, with younger individuals showing enlargement and older individuals showing normal or reduced volumes.20 This pattern has been postulated to be due to overgrowth and hyperactivity of the amygdala during childhood and subsequent atrophy.24 No study has yet correlated direct measures of socio-emotional memory with imaging features of the MNS or STS; however, these regions play a role in several important aspects of human social behavior, and abnormalities of both structure and function have been detected in imaging studies of individuals with ASD.3

Conclusion

Genetic and molecular studies suggest that the pathophysiological mechanisms involved in ASD are likely to alter synapse formation and function. Such mechanisms could lead to widespread, enduring disturbances in functioning across multiple neural systems, and consequently potential impairment in the acquisition of a wide range of skills, including social, communicative, motor, and intellectual skills. Despite a rapid growth in the number of neuroimaging studies of ASD in recent years, those examining learning and memory systems in ASD are few in number. These studies have employed a variety of measures of learning and memory, and, so far, have examined correlations with structural but not functional imaging measures of the brain. The early literature supports the following preliminary hypotheses: (1) abnormalities of hippocampal subregions may contribute to autistic deficits in episodic and relational memory; (2) disturbances to an amygdala-based network (which may include the fusiform gyrus, superior temporal cortex, and MNS) could play a role in autistic deficits in socio-emotional learning and memory; and (3) abnormalities of the striatum may contribute to developmental dyspraxia in individuals with ASD.

How deficits in learning and memory relate to the core features of ASD remains unclear. Could deficits in procedural learning and memory at an early age lead to a restricted behavioral repertoire and repetitive behaviors? Would deficits in episodic and relational memory predispose to an autistic pattern of restricted interests? A better understanding of the links among brain, memory, and behavior will require studies that examine all three levels of disturbances: the neurobiological abnormalities, neuropsychological impairments, and behavioral manifestations of autism.

Neuroimaging studies of learning and memory systems in ASD will also need to take into account interactions among multiple memory systems. Evidence from animal and human studies suggests that many learning tasks activate multiple memory systems in parallel, and that these interactions may be competitive in nature.26 Thus, damage to a given memory system can produce enhanced learning in another system. This phenomenon could help to explain the prodigious ‘splinter’ abilities observed in some autistic savants. The neurobiological mechanisms by which these systems interact are not yet known, but could involve direct anatomical projections between systems or indirect neuromodulatory influences of other brain structures.

Elucidating the disturbances to multiple learning and memory systems in ASD could have several potential implications for both research and clinical practice. Identifying patterns of neural and cognitive abnormalities could help to facilitate the discovery of biologically-based ASD subtypes that would replace the behaviorally defined subtypes currently in use. A clear understanding of the learning and memory deficits could suggest therapies targeted at remediation of these specific deficits. Understanding the profile of relative strengths and weaknesses in learning and memory may ultimately aid in the development of optimal therapeutic interventions.

Ancillary