Social cognition in individuals born preterm

Faces hold a substantial value for effective social interactions and sharing. Covering faces with masks, due to COVID-19 regulations, may lead to difficulties in using social signals, in particular, in individuals with neurodevelopmental conditions. Daily-life social participation of individuals who were born preterm is of immense importance for their quality of life. Here we examined face tuning in individuals (aged 12.79 ± 1.89 years) who were born preterm and exhibited signs of periventricular leukomalacia (PVL), a dominant form of brain injury in preterm birth survivors. For assessing the face sensitivity in this population, we implemented a recently developed experimental tool, a set of Face-n-Food images bordering on the style of Giuseppe Arcimboldo. The key benefit of these images is that single components do not trigger face processing. Although a coarse face schema is thought to be hardwired in the brain, former preterms exhibit substantial shortages in the face tuning not only compared with typically developing controls but also with individuals with autistic spectrum disorders. The lack of correlations between the face sensitivity and other cognitive abilities indicates that these deficits are domain-specific. This underscores impact of preterm birth sequelae for social functioning at large. Comparison of the findings with data in individuals with other neurodevelopmental and neuropsychiatric conditions provides novel insights into the origins of deficient face processing.

. Periventricular leukomalacia (PVL) as depicted by MRI at a later stage after completed myelination in a 6-year-old male born at 32 weeks of gestation. Axial T2-weighted (T2w), FLAIR and T1-weighted (T1w) images at the level of the centrum semiovale cutting the upper part of the ventricles (top row) and at the level of the lateral ventricles (bottom row) show the characteristic gliosis as hyperintense high signal on the T2-weighted image and hypointense low signal on the T1-weighted image. The gliosis is best seen on FLAIR images, where the periventricular hyperintensities contrast with the hypointense CSF signal. The top row illustrates gliosis, and the bottom row additional tissue loss (mild ventricular dilatation with irregularly extended borders). From Ref. 5  www.nature.com/scientificreports/ understanding of others (that is not based on motion characteristics), whereas their PB peers with normal MRI scans perform much like term-born children. This deficit is related to the PVL extent in the right temporal brain area known to be heavily engaged in social cognition 50 . Along with body language, faces hold an exceptional value for a wide range of social functioning 16,21 . Only a very few indications of deficits in face processing associated with prematurity were available till last decade 51,52 . Deficient face recognition (as assessed by the Facial Memory subtest from the Test of Memory and Learning, TOMAL 53 ) is reported in 18% of 22 premature children aged 6-15 years with signs of (mostly) moderate parieto-occipital PVL identifiable on MRI scans 54 . Recent work reveals alterations in face processing as sequelae of preterm birth. PB infants examined during the first 48 h of life do not express visual preference for faces over scrambled faces as their term-born peers do 55 . Functional near-infrared spectroscopy (fNIRS) indicates that PB infants aged 6-10 months show smaller hemodynamic response in the right frontotemporal areas while watching their own mother's faces as compared to unknown faces, whereas in term-born infants, mothers' faces elicit greater hemodynamic responses than unknown faces 56 . Former preterms aged 6-8 months demonstrate atypical eye-gaze behaviour in regard to faces and social content at large: they look shorter at eyes and a face presented among non-face distractors than their term-born peers 57 . If small for gestational age, former preterms 5-15 years old perform worse on the Facial Memory subtest from the TOMAL 53 showing poorer face recognition skills 58 . However, former preterms aged 19-20 years display intact skills in terms of accuracy (albeit slower responses) for inferring emotions from dynamic faces, with some difficulties in recognizing anger at low intensity 59 . Adults (26-36 years old) who were born preterm with extremely low birthweight (< 1000 g) possess lower ability for discrimination between individual faces, whether human or monkey 60 , indicating that deficits in face processing are long-lasting and persist into adulthood.
The present study was directed at investigation of face tuning in PB adolescents who suffer PVL. For investigation of face tuning, we applied a recently developed tool, a set of images comprising food ingredients such as fruits and vegetables [61][62][63][64][65][66][67][68] . The Face-n-Food images to some extent border on the style of Giuseppe Arcimboldo, a genius Italian painter known for his imaginative portraits composed of fruits, vegetables, and even roasted meat (Fig. 2). The primary advantage of these images is that single components do not trigger face processing. In other words, on the Face-n-Food task, face tuning occurs spontaneously without being explicitly cued by face elements such as eyes or mouth. For seeing a face in these images, one has to perceive an image as a Gestalt (a German word that defines a configuration of elements unified into a whole in such a way that its overall properties cannot be identified from a simple sum of its parts). The other advantage of the task is the usage of unfamiliar images that is of value for research in clinical populations 69 . Here we intended to clarify (i) whether former preterms exhibiting PVL express aberrant face tuning; and (ii) whether possible shortfalls in the face sensitivity are related to other visuo-perceptual abilities or rather represent a specific deficit. For assessing visuospatial and cognitive abilities, additional tasks were administered to PB individuals. All of them represent well-established tools for neuropsychological assessment 54,70 (for description, see "Neuropsychological assessment" section in "Methods" section).

Results
The Face-n-Food images were presented to all participants one by one in a predetermined order from the least to most recognizable as a face 61 . Similar to previous work in TD individuals and in a range of neurodevelopmental and psychiatric conditions 61-68 , PB individuals described a food-plate image either in terms of food (non-face response, 0) or as a face (face response, 1). Two out of 14 PB patients completely failed on the Face-n-Food task: they did not have spontaneous face impressions even from the most face-resembling images. PB individuals   Figure 3 represents the percentage of face responses for each Face-n-Food image separately for former preterms and TD controls. Inspection of this Figure shows that former PB adolescents not only much later had a face impression, but gave overall much fewer face responses. As indicated by multiple stepwise nominal logistic regression analysis, the effect of group (TD vs. PB) is highly significant (χ 2 (1) = 96.01, p < 0.0001; effect size, r = 1.789). Remarkably, for the first five images that are less face resembling, all PB individuals provided no face responses at all. Starting from the image 5, TD participants reached very fast the ceiling level of performance. By contrast, even on the images strongly resembling a face (7 through 9), PB individuals attained about 70% of face responses, and only the most recognizable image 10 elicited 85.71% face responses. Figure 3 enables comparison of the data in PB individuals not only with TD controls, but also with earlier data in individuals with autistic spectrum disorders, ASD 64 . Multiple stepwise nominal logistic regression analysis indicates that all groups of participants significantly differ from one another (χ 2 (2) = 78.37, p < 0.0001). Most eye-catching finding is that the performance level of PB patients on the Face-n-Food task is substantially lower not only as compared with TD controls, but also with ASD individuals (χ 2 (1) = 18.38, p < 0.0001; effect size, r = 0.783; with no difference in the general intelligence quotient, GIQ scores between PB and ASD persons: t(28) = 1.124, n.s.).
No correlation was found between performance of PB individuals on the Face-n-Food task (face response rate) and the GIQ (Spearman's rho, ρ = 0.057, p = 0.85, two-tailed, n.s.). There was also a lack of correlation between face response rate and performance on the visual-perceptual tests: the Block Construction  . Percentage of face responses for each Face-n-Food image in TD controls (green), former preterms with PVL (blue), and patients with autistic spectrum disorders (ASD, violet). The image number reflects its face resemblance (1, the least recognizable as a face, through 10, the most resembling a face images from the Face-n-Food task; see Ref. 61 ; the Creative Commons Attribution [CC BY] license). Vertical bars represent 95% confidence interval, CI. The data for individuals with ASD had been reported earlier (see Ref. 64  www.nature.com/scientificreports/ no correlation was found between the face tuning and tests assessing semantic categorization: Matching by Class (ρ = 0.077, p = 0.795, n.s) and Matching by Function (ρ = − 0.077, p = 0.794, n.s.). This indicates that impaired performance on the Face-n-Food task in PB individuals originates from specific deficits in the face sensitivity rather than from other visual-perceptual or cognitive disabilities.

Discussion
By applying a novel tool for the face sensitivity examination, a recently developed Face-n-Food task 61-68 , we assessed the face tuning in individuals who were born prematurely and suffer PVL, the dominant form of brain injury in prematurity 5,54 . PVL represents a bilateral pattern of white matter lesions equally affecting both hemispheres that leaves the brain without sufficient potential for functional reorganization. The outcome indicates that in former survivors of preterm birth with PVL, the face sensitivity is substantially lower than in term-born TD peers. In PB individuals, the face sensitivity in face-like non-face images is not related to visual-perceptual organization and other cognitive abilities assessed in the course of neuropsychological examination and, therefore, the lower performance on the Face-n-Food task appears to stem from deficits in the face tuning per se, or strictly speaking, in the sensitivity to a coarse face schema.
Face tuning in prematurity and autism. Evaluation of the outcome in PB individuals within the framework of earlier data obtained in a sample of individuals with ASD 64 of similar age, cultural background (known to affect the face sensitivity 66 ), gender (both samples are prevailed by males) and general IQ, reveals that PB individuals with PVL express a lower sensitivity to a coarse face schema in the Face-n-Food images as compared not only with TD term-born controls, but also with their peers with ASD. Although in both clinical samples, a number of patients who completely failed on the task is comparable (i.e., two out of 16 in ASD and two out of 14 in PB patients did not spontaneously recognize even the most face resembling image 10), the face recognition dynamics is different in these groups. Face recognition of the first five-six images less resembling a face is remarkably poor in PB with PVL ( Fig. 3). Only starting from image 7, PB individuals reliably recognize images as a face, reaching 85.7% recognition for the most face resembling image 10. By contrast, face recognition of ASD individuals steadily improves resulting in 87.5% face responses for each of the most recognizable images 8-10.
As ASD individuals are well-known to experience difficulties in several aspects of social cognition 17,18,35,36,71 , this discrepancy in the face sensitivity between ASD and PB patients (with substantially lower face tuning of former preterms) appears arresting. Survivors of preterm birth are reported to express autistic traits as well as difficulties in social functioning such as social withdrawal, incompetence, and communication problems 72 , or they can be also positively screened for ASD [73][74][75] . However, this cannot provide a satisfactory explanation for such a poor outcome in PB individuals. As establishing binding between even a couple of elements resembling eyes and mouth, i.e., building a coarse face scheme, is already sufficient for perception of non-face images as a face 76 , it appears stunning that PB patients are so deeply impaired on the task. Furthermore, a kind of predisposition for a coarse face schema (such as two eyes above a mouth) is believed to occur already very early in lifespan 77 : foetuses in the third trimester of pregnancy 78 , human infants [79][80][81][82] , and children aged 5-6 years 83 are reported to demonstrate a visual preference for face-like stimuli over similar images. Furthermore, even newly hatched domestic chicks (Gallus gallus) exhibit preference to a coarse face schema [84][85][86] , while non-human primates such as the rhesus monkey are likely to share face-detection machinery with humans [87][88][89] . Yet it is reported that only 7-and 8-month-old human infants (but not infants aged 5-6 months) exhibit visual preference to Archimboldo portraits over the same images presented upside-down 90 that implies a period of development of the sensitivity to faces in such images.
Impact of PVL on the face tuning. PVL lesions specifically affecting neural networks underwriting face processing might be primarily responsible for the poor outcome on the Face-n-Food task in former preterms. The periventricular area of the brain contains many interconnecting fibres, damage to which may lead to brain disintegration eliciting a number of deficits in visual perception and social cognition. Converging evidence from neuroimaging, neuropsychological and electrophysiological studies indicate that the subcortical face route, such as colliculo-pulvino-amygdalar pathway 91 , provides a neural substrate for the cortical social brain network 87,89,92,93 . Alterations of this pathway along with subcortical-cortical and cortico-cortical connectivity caused by PVL might be decisive for alterations in face processing and social cognition in PB individuals. Comparison of face tuning in PB individuals with and without PVL lesions identifiable on a MRI scan may deliver pivotal arguments for the sources of deficient face tuning. Yet one has to keep in mind that the altered brain connectivity in atypical perinatal conditions, such as Intrauterine Growth Restriction (IUGR) and prematurity itself (in former preterms without brain injury identifiable by brain imaging), have been shown to impact the brain development, affecting network connectivity in the short-and long-term and resulting in the impairment in socio-cognitive potential in childhood and adolescence [94][95][96][97][98][99] . In addition, fNIRS points to a negative correlation in PB individuals between oxyhemoglobin HbO2 measures during presentations of faces and grey matter volumes in the face-specific brain regions forming the social brain such as the fusiform gyrus and amygdala 56,100 . Earlier data indicates that PB individuals with PVL exhibit compromised ability for visual processing of body motion as compared not only with term-born peers, but also with former preterms without structural brain abnormalities detectable on MRI scans 43,101 . The same holds true for the visual social cognition (namely, event arrangement) task requiring perception and understanding of others: only former preterms suffering PVL are impaired on this task, and this deficit is related to the lesion extent in the right temporal region 50 . By affecting structural and functional brain connectivity, PVL may lead to disintegration of neural communication inside and outside of the networks supporting social cognition. In line with this assumption, in adolescents with PVL, bursts in the spectral amplitude of the magnetoencephalographic (MEG) cortical response underlying www.nature.com/scientificreports/ large-scaled neural communication occur later than in TD adolescents or are even completely absent 101 (see also 102,103 for evidence on alterations in the topography of the evoked MEG response to body motion with and without a camouflage).

Specificity of face tuning in prematurity.
Previous work revealed substantial (though rather specific for every single neurodevelopmental and neuropsychiatric condition) deficits in the face tuning in different patient populations (for comparative analysis, see 67 ). For example, in schizophrenia, the face tuning is substantially impaired, and this deficit is related to deficits in both social cognition and visual perceptual organization 67 , whereas individuals with major depressive disorder do not express any shortfalls in face tuning 68 . Comparison of the present outcome with earlier data in patients with Down syndrome, DS 65 , and Williams syndrome, WS 62 , sheds light on the possible origins of the deficient face tuning in former preterms. By contrast with ASD patients, both WS and DS individuals are well-known for their appetitive drive for social interaction and face fascination. Yet all three samples of patients exhibit poor tuning to coarse face information in face-like non-face images: all of them did not recognize the first five images (less resembling a face) as a face. However, individuals with DS remain on rather low face recognition rate (about 0.3) even for the most recognizable image 10 65 (Fig. 3) that can originate from some difficulties in abstract/symbolic reasoning, whereas PB individuals as well as those with WS steadily improve in their face recognition reaching 85.7% and 85%, respectively, for the most resembling a face image 10. This analysis suggests that face tuning deficits may be of similar origin in PB and WS individuals. One possible explanation is that difficulties in feature integration and Gestalt perception (perceiving images as a whole) known in both patient groups 5,62,104-107 , underlie deficient face tuning. On the same wavelength, original Arcimboldo hidden-face portraits are judged as being more ambiguous by adult TD individuals with local perceptual style 108,109 . The next step in clarification of the nature of aberrant face tuning in prematurity would be functional brain imaging during perception of face-like non-face images. Such examination, in particular, by means of MEG uncovering even fine-graded alterations in the time course and dynamics of bran activity, helps to identify potential deficiencies in neural communication of the underpinning neural networks. Another valuable approach would be using of multimodal integrative strategies that combine several sources of information (such as structural and functional brain connectivity) in relation to behavioural measures of performance 22,24 . Face tuning in prematurity and functional brain imaging. Intact communication inside and outside of the brain networks underlying face tuning is largely unidentified and the findings available are rather controversial 67 . Taking together, the findings demonstrate that (a) topography and time course of the neural circuits underpinning processing of real faces and face-like images are similar; key activation includes the occipital cortices, fusiform face area (FFA), and inferior temporal brain areas [110][111][112] ; and (b) corresponding brain activation is primarily right-hemispheric 113,114 . The right hemispheric dominance is also reported in processing of Arcimboldo-like images yielding greater functional magnetic resonance imaging (fMRI) activation, compared to Renaissance portraits and faces, in the occipito-temporal face-specific network (covering the FFA), bilateral superior and inferior parietal cortices, and the inferior frontal gyrus 109 . The right superior temporal sulcus (STS), a hub of the social brain, distinguishes real faces from face-like images 115 . Electroencephalography, EEG, suggests that already 1-to 4-day-old newborns exhibit activation in the right-lateralized network engaging lateral occipito-temporal and medial parietal areas overlapping with the face-processing circuits in adults 82 . However, fNIRS conducted in 7-8-month-old infants indicates that in response to upright Arcimboldo portraits compared with images of single vegetables used as a baseline, the concentration of oxy-Hb increases in the left (but not right) temporal area, whereas such effect is absent in response to inverted Arcimboldo images in the temporal areas of both hemispheres 90 . The findings in patients with lesions suggest that lesions to the right ocipitotemporal brain areas leave perception of Arcimboldo portraits as faces intact 116,117 , whereas left-hemispheric lesions severely affect face tuning 118 . Individuals with pre-manifest Huntington's disease (characterized by aberrant social cognition 119 ) show a right hemispheric decrease in the N170 component of event related potential, ERP, elicited by the face-like non-face images 120 .
Sex impact on the face tuning in prematurity. As predominantly male PB patients had been enrolled in the present work, one of the study's limitations is that it left possible sex differences beyond attention. It is known that males are at a 14-20% higher risk of premature birth 121-123 and its complications, being more vulnerable to white matter injury 5,124,125 . In the face sensitivity, female superiority has been observed by administering the Face-n-Food task to a homogeneous group of university students 61 . Yet in adult individuals with major depression, no gender differences occur in the face tuning 68 . The female brain is reported to be more responsive to face-like non-face images such as clocks or backpacks eliciting a face impression, with a greater activation in the right STS and Brodmann area 22, though sex differences are absent at earlier stages of face processing 112 . Therefore, the rough face schema appears to be sex-independently hardwired in the typically developing brain. Future work is required to elucidate possible sex differences in the face sensitivity of former preterms both at behavioural and brain levels. However, most likely, sex differences will be camouflaged by a more influential impact of PVL, and, therefore, clarification of this issue appears more plausible in former preterms without signs of severe brain abnormalities.

Résumé
Although a course face schema is thought to be hardwired in the brain, individuals who were born preterm with signs of brain injury (PVL) exhibit substantial deficits in the face tuning not only compared with typically developing controls but also with individuals with ASD. The lack of associations between the face sensitivity and Face-n-Food task. The Face-n-Food task was administered to participants. The task is described in detail elsewhere 61 . In brief, for this task, all participants were presented with a set of images, one by one, in the predetermined order from the least to most resembling a face (images 1 to 10). This order was determined in the previous study with TD volunteers 61 . This fixed order had been used, because once seen as a face, Face-n-Food images are often processed with a strong face-dominating bias. On each trial, participants had to perform a spontaneous recognition task: they were asked to briefly describe what they saw. Their reports were recorded, and then analysed by independent experts. For further data processing, the responses were coded as either non-face (0) or face (1) report. No immediate feedback was provided. To avoid time pressure that can potentially cause stress and negative emotional and physiological reactions blocking cognitive processes, there was no time limit on the task. With each participant, the testing procedure lasted no longer than 10-20 min.
Neuropsychological assessment. Several additional tasks directed at assessment of visuospatial and cognitive abilities were routinely administered to PB individuals. All of them represent well-established tools for neuropsychological assessment 54,70 : (1)  images of objects presented under unusual lighting conditions) and photographs of the same objects presented in a conventional manner. Participants are asked to recognize each object. (5) The Imagery Figure (IF) test contains 20 out of the 128 items of the Birmingham Object Recognition Battery 131 , defined as easy in the original battery, and seemed appealing to children. Ten of them represent depictions of objects and the rest ten are imaginary fanciful images combining features of two objects or animals (e.g., a camel with a goose's head). Participants have to indicate whether each item represents an animal or object that really exists. (6) In addition, two tests (described below) are aimed at assessment of semantic categorization 129,130 . Both the Matching by Class and the Matching by Function tests consist of 20 out of the 64 items of the original test (the Birmingham Object Recognition Battery 131 ). In ten out of 20 selected items, participants are asked to match objects belonging to the same class, and in the rest to match objects that are functionally related. Each item consists of the target image and two images, one that matches the target and a distractor. For each participant on each test, z-scores were computed by using the following formula: 'raw score' minus 'average score' (for participants of this age) divided by SD (for participants of this age).

Data analysis.
At outset, all data sets were routinely analysed for normality of distribution by using Shapiro-Wilk tests with subsequent usage of either parametric (for normally distributed data) or, otherwise, nonparametric statistics. For not normally distributed data sets, additionally to means and SDs, Mdns and 95% CIs are reported throughout the text.