Forty-six family trios (including six multiplex families for which both affected siblings were examined) and 14 proband-parent dyads were studied. Twenty-nine parents whose child with ASD was unable to complete testing also were examined. Thus, a total of 66 probands and 135 parents were compared with separate groups of healthy controls matched on age, gender, and non-verbal IQ to probands (n = 29) and matched on gender and non-verbal IQ to parents (n = 47; Table 1). No parent controls were related to any of the proband controls. While the parent group was not age-matched to their control counterparts, age was not related to task performance in either the adult group (r’s < .19). Participant groups were matched on nonverbal rather than verbal IQ based on data indicating that individuals with ASD show less disorder-related weaknesses in nonverbal abilities . Henceforth, controls matched to children and controls matched to parents will collectively be referred to as controls unless specified otherwise.
Testing was conducted at the University of Illinois at Chicago (n = 39) and the University of Texas Southwestern (n = 238). Individuals with ASD and their parents were recruited through community advertisements and local outpatient clinics. ASD diagnoses were confirmed using the Autism Diagnostic Observation Schedule, 2nd Edition (ADOS-2 ;), the Autism Diagnostic Interview-Revised (ADI-R ;), and expert clinical opinion based on DSM-5 criteria . Individuals with ASD were excluded if they had a known genetic disorder associated with ASD (e.g., Fragile X syndrome) or history of non-febrile seizures. All control participants were recruited through community advertisements and had a score of < 8 on the Social Communication Questionnaire (SCQ ;). Controls were excluded if they had current or past psychiatric or neurological disorders, a family history of ASD in first- or second-degree relatives, or a history of developmental disorders or severe mental illness (e.g., schizophrenia) in first-degree relatives. Two parents with elevated SCQ scores completed the ADOS, but neither met the criteria for ASD. No participant had a history of head injury resulting in loss of consciousness. To ensure that the participants could understand all the task demands, only those with a NVIQ > 60 on the Wechsler Abbreviated Scale of Intelligence (WASI ;) were included. Thirty-four participants (16 probands, 18 parents) were receiving psychotropic medication within 48 h of testing (Additional file 1). No controls were receiving any psychotropic medication within 4 weeks of participating in the study. When comparing probands/parents on-medication to those off-medication, groups did not differ on primary dependent variables (p’s > .16). Thus, all participants were included in our final analyses.
Ethics approval and consent to participate
All participants ≥ 18 years of age provided written consent, and minors provided assent and written consent was obtained from their legal guardians. Study procedures were approved by the University of Illinois at Chicago and the University of Texas Southwestern Institutional Review Boards.
Probabilistic reversal learning task
As described previously , during the probabilistic reversal learning task (PRL) task, participants were instructed to choose one of two identical stimuli (i.e., pictures of animals) positioned in different locations on the screen. Participants were reinforced (i.e., a coin appeared on the screen and placed into a money bag that kept track of total coins) on 80% of correct responses and on 20% of incorrect responses. During the acquisition phase, participants chose one of two stimulus locations until they identified the correct location on 8 of 10 consecutive trials. Then, they proceeded to the reversal phase in which the correct location was switched without warning, and participants had to identify the new correct location on 8 of 10 consecutive trials. Testing was discontinued if they did not reach the criterion within 50 trials during either phase. All participants completed two practice tests prior to PRL administration to establish test comprehension. Ten participants (7 probands, 1 parent, 2 controls) failed the acquisition phase, 14 participants (5 probands, 5 parents, 4 controls) failed the reversal phase, and 6 participants (4 probands, 4 parents) were not administered this test due to time constraints. Fifty probands, 125 parents, and 70 controls were included in the final analyses. We examined the number of errors (i.e., selecting the incorrect location) separately for acquisition and reversal phases.
Stop signal task
To examine behavioral inhibition, participants completed a stop-signal task (SST) consisting of interleaved GO and STOP trials as described previously [15, 28]. During GO trials, a target appeared to the left or right of the center, and participants responded as quickly as possible by pressing the button in the corresponding location. During STOP trials, a central STOP cue appeared at varying stop-signal delays (50–283 ms) after the GO cue, and participants were instructed to withhold their response. To ensure that participants did not delay their responses indefinitely, they received a prompt indicating “FASTER” and an “X” if they did not respond within 650 ms. The task consisted of 4 blocks of 63 trials (60% GO and 40% STOP trials) with similar ratio of GO to STOP trials in each of the 4 blocks. In order to ensure that each individual understood the instructions, participants completed a practice task consisting of 52 interleaved GO and STOP trials prior to the SST in which they had to demonstrate successful performance on 50% of STOP trials as done previously [15, 28].
Based upon our prior findings that probands show a reduced ability to proactively delay the onset of their responses and that increased slowing is associated with increased stopping success rate , baseline reaction times (RT) were measured during a task consisting of 60 GO trials administered prior to the SST. Three probands exceeded 650 ms average RT on > 50% of baseline trials and were not administered the SST. Additionally, five probands and one control failed to meet the practice criterion, and 19 individuals (8 probands, 11 controls) had scheduling issues that prevented completion of the SST. Fifty probands, 135 parents, and 64 controls were included in the final analyses. The proband participants who did not complete the SST were significantly younger than probands who completed the SST (t = 6.13, p < .001; mean age (SD) of non-completers 6.50 (2.2); mean age (SD) of completers 11.45 (3.9)). We computed the percentage of STOP trials in which participants inhibited their response and the difference in baseline GO and SST GO RTs. The order of tasks (PRL, SST) was randomly assigned to each participant.
The ADI-R and ADOS were used to confirm clinical diagnoses and assess social-communication abnormalities and RRBs in probands. The ADI-R is a semi-structured caregiver interview used to characterize current and past clinical symptoms of ASD, including social abnormalities, communication impairments, and RRBs. The ADOS is a semi-structured assessment of social-communication impairments and RRBs. For each measure, higher scores represent more severe ASD symptoms.
In order to determine if PRL and SST performance covaried with subclinical ASD traits in parents of children with ASD, each parent completed the self-report version of the Broad Autism Phenotype-Questionnaire (BAP-Q ;). The BAP-Q quantifies the severity of subclinical features of ASD, including social aloofness, pragmatic communication deficits, and rigid personality traits. As recently indicated, parental BAP is a useful tool to create phenotypically distinct subgroups of families of children with ASD . Parents’ scores for each subdomain were compared against published norms . As previously done , parents who scored above the identified BAP cutoffs on any subdomain were classified as “BAP+ parents,” and those who did not exceed any subdomain cutoff were categorized as “BAP− parents” (Table 2). Relative to prior studies, our sample demonstrated similar percentages of parents exceeding cutoffs for aloof (16%), pragmatic communication (25%), and rigid personality subscales (25%) or showing at least one BAP feature (~ 30% [2, 3, 32];). Probands with at least one BAP+ parent were categorized as “probands of BAP+ parents”; all other probands were categorized as “probands of BAP− parents.” Only two probands had both parents classified as BAP+ parents. Notably, probands of BAP+ parents and probands of BAP− parents did not differ on ADOS or ADI ratings (p’s > .22; Table 2). Ten parents did not complete the BAP-Q due to time constraints. Thirty-two BAP+ parents, 83 BAP− parents, 16 probands of BAP+ parents, and 30 probands of BAP− parents completed the PRL test. Thirty-four BAP+ parents, 91 BAP− parents, 18 probands of BAP+ parents, and 32 probands of BAP− parents completed the SST.
Each dependent variable was age-adjusted to account for non-linear relations between age and task performance as in previous work . An inverse regression function was fit to data from healthy controls (combined proband control and parent control groups) from the current study to provide estimates of expected performance based on each participant’s age as in previous studies . Then, the difference between each participant’s actual performance and their age-adjusted expected value was calculated, creating a deviation score for each participant for each dependent variable (Additional file 2). Deviation scores were converted to Z-scores based on the sample mean and standard deviation of all controls, with negative Z-scores denoting worse performance than expected given the individual’s age. For example, a negative Z-score for either PRL reversal phase errors or SST errors would indicate the participant is making more errors than would be expected given their chronological age, and a negative Z-score for SST RT slowing would reflect reduced RT slowing than expected. Distributions for each of our primary cognitive control outcomes for each subgroup are shown in Additional file 4.
Separate ANOVAs were used to examine each age-adjusted dependent variable (Z-score) with group (proband vs parent vs control) as the between-subject factor. Significant effects were probed with planned pairwise comparisons using Bonferroni corrections for multiple comparisons. Due to possible impact of including six multiplex families, we removed one proband from each of these families (at random) and conducted all the analyses a second time. Results were not substantively different, so all probands were included in the final analyses. PRL reversal phase errors were not normally distributed (kurtosis: proband = 1.235; parent = 4.465; control = 6.269), so non-parametric Kruskal-Wallis H tests were conducted.
In order to determine whether parents demonstrating subclinical ASD features and their offspring demonstrated greater cognitive control issues than controls, we conducted separate ANOVAs comparing probands of BAP+ parents, probands of BAP− parents, and controls as well as BAP+ parents, BAP− parents, and controls on each dependent variable. Initial analyses of individual tasks included any participant who completed that task. Follow-up analyses including only the subset of individuals who completed all tasks were not substantively different (proband = 39, parent = 125, control = 57; Additional file 3).
To estimate the familiality of behavioral flexibility and inhibition deficits in family trios, Sequential Oligogenetic Linkage Analysis Routines (SOLAR) was used . This analysis approach provides estimates of familiality (h2) representing the proportion of variance in PRL or SST performance accounted for by family membership. Maximum likelihood estimates were used to compare a model in which performance is explained by family membership relative to a model in which family membership is not considered.
In order to examine inter-relationships between behavioral flexibility and response inhibition in probands, parents, and controls, separate non-parametric Spearman correlations were conducted for each group. For probands only, we examined the relationships between behavioral flexibility and inhibition deficits with ADI-R (ADI-R ;) and ADOS-2 (ADOS-2 ;) ratings of ASD symptoms. The revised algorithms for modules 1–3  and module 4  were used. In order to determine whether deficits are more severe for probands who have parents with subclinical traits, we conducted the same analyses separately for probands of BAP+ parents and probands of BAP− parents. To reduce type I error rates, we only considered relationships significant if |r| > .50 or p < .01.