Non-suicidal self-injury (NSSI), defined as the intentional and direct damage of body tissue without suicidal intent, is described as a prominent and maybe the most noticeable symptom of borderline personality disorder (BPD) However, lifetime prevalence rates of BPD range around 2.7% in the general population (1), while around 18% of adolescents report to have engaged in NSSI at least once, and repetitive NSSI is seen in around 4 and 7% of adolescents (2, 3). From an epidemiological perspective, these numbers suggest that NSSI also occurs independently of BPD in a large number of adolescents, which is why NSSI has been suggested as a distinct psychiatric disorder in the most current version of the Diagnostic and Statistical Manual [DSM-5 (4)] as condition for further study.

Still, there are commonalities between both entities. According to the DSM-5, BPD is characterized by impairments of interpersonal functioning, which can be associated to “interpersonal hypersensitivity (i.e., prone to feel slighted or insulted)” and “perceptions of others selectively biased toward negative attributes or vulnerabilities” (4). In laboratory studies, adults with BPD have repeatedly shown an increased sensitivity for social exclusion in comparison to healthy controls (5, 6). Similarly, NSSI has also been linked with impaired social interactions. For example, bullying has been shown to be a risk factor in longitudinal studies (7, 8) and patients with NSSI show elevated feelings of loneliness, even in comparison to clinical controls (9). Consequently, while BPD and NSSI do not match well in prevalence rates, sensitivity to social exclusion may play a role in both syndromes.

Investigations of the neural correlates of social exclusion [evoked by the paradigm “Cyberball” (10)] have shown increased activation in various brain regions, including the anterior cingulate cortex (ACC), anterior insula/ventrolateral prefrontal cortex (vlPFC), medial prefrontal cortex (mPFC), dorsolateral prefrontal cortex (dlPFC), and ventral striatum (11–14). While dorsal ACC, ventral striatum and anterior insula/vlPFC have been conceptualized as a network processing salience, dorsolateral and dorsomedial prefrontal regions, together with lateral parietal areas are conceptualized as a central executive network (15). The salience system has been suggested as the cerebral network where sensory data with behavioral relevance are integrated with visceral, autonomic, and hedonic signals while the central executive system might operate on the identified salience. Such operations, for example, comprise directing attention to pertinent stimuli when behavioral choices are weighed against shifting conditions, background homeostatic demands, and context (16). In adult patients with BPD, altered activation of the dorsal ACC and the dorsal mPFC was shown to relate to social exclusion, but interestingly also to social inclusion as compared to healthy controls (17). The authors suggested that this might point toward a generalized “hypermentalizing” or alarm reaction in patients with BPD in both negative and positive social situations. In a study investigating depressed adolescents with and without repetitive NSSI, we found that activation in the mPFC and vlPFC was not only different from healthy controls but also modulated by the presence or absence of NSSI (18). However, no differences during social inclusion were found across those three groups. We took these results as hint for a potentially distinct neurobiology of NSSI compared to BPD.

Given this discrepancy between patient populations despite the overlap in the clinical presentation of BPD and NSSI, we aimed at a better understanding of commonalities and differences on a neurobiological level. We further investigated this issue by recruiting patients with ongoing NSSI as a symptom with and without a diagnosis of BPD. As NSSI without BPD is most frequent during adolescence, while diagnoses of BPD are often not made before early adulthood, we recruited subjects from post puberty to early adulthood. To control for effects of age, we investigated matched healthy controls. In accord with previous finding using the same task, we hypothesized that both patient groups would show increased activation in relevant brain areas during social exclusion as compared to healthy controls. We also conjectured that only those NSSI participants with an additional BPD diagnosis would also show altered activation during social inclusion.

This study is the first to compare adolescents with ongoing NSSI without BPD diagnosis against young adults with ongoing NSSI and BPD diagnosis regarding their neural processing of social situations experimentally operationalized by the Cyberball paradigm during fMRI. While both patient groups showed significantly higher ratings on the HFS relative to their age-matched healthy control groups as an index of greater general sensitivity for social exclusion, task-specific ratings of distress experienced from the social exclusion condition of the Cyberball paradigm were significantly higher only in BPD patients relative to controls, but not in NSSI patients. Furthermore, in both rating scales, BPD patients scored higher than patients with NSSI without BPD diagnosis [see also Ref. (5, 6)].

Functionally, the contrast of social exclusion versus inclusion was significant only in the putamen of the NSSI group relative to age-matched controls and BPD patients. At the predefined level of significance, this contrast did not reveal any significant differences in the height of neural activations between BPD patients and healthy controls. Only when the level of significance was lowered for exploratory reasons, both patient groups showed a similarity in comparison to their age-matched healthy controls with relatively increased activation in the ventral ACC for the contrast of social exclusion minus inclusion. This area has previously been reported in two other studies on social exclusion (14, 29) and has functionally been related to the effect of negative social feedback (30). When contrasting social inclusion against passive viewing, a rather consistent pattern of higher activation in the BPD group compared to NSSI patients emerging in the dorsomedial, and dlPFC, and the left anterior insula was found.

Higher activation of the putamen during the exclusion relative to the inclusion and the passive watching condition was observed in the NSSI group only. This was true despite the adolescent patients’ ratings of their general sensitivity to social rejection being significantly lower than those of the BPD group. As the putamen is part of the salience-network (16), this may point toward increased neural reactivity to behaviorally relevant stimuli in adolescents with NSSI. Hyperactivation of nodes of the salience-network may be related to the perception of inconsistencies or norm violation. This may be interpreted in a way that the younger patient group may have processed the negative social feedback more intensely than healthy controls and social exclusion may have been more salient than inclusion. These findings are in line with studies reporting bullying, as severe form of social exclusion, to be a risk factor for NSSI in adolescence (7, 8). However, in the BPD group, putamen activation was indifferent across all three Cyberball conditions. Therefore, this finding should be interpreted with caution and future studies are needed for validation.

Similar to a study comparing patients with BPD to healthy controls (17), we found enhanced activation in the dmPFC in BPD patients during both the inclusion and exclusion conditions. This finding might add to the notion of generalized “hypermentalization” of BPD patients during social exclusion, as proposed by Domsalla and colleagues.

Also in line with Domsalla et al. (17), we found particularly strong neural activation to social inclusion in dorsolateral and dorsomedial prefrontal brain regions in BPD, which was not seen in the NSSI group (Figure 2). To explain this result and the additional finding of self-reports of increased feelings of being excluded and higher trait-sensitivity for exclusion in BPD patients, Domsalla et al. (17) proposed Festinger’s theory of cognitive dissonance (31). For BPD patients, social exclusion could be the situation they anticipate as a result of regularly biased processing in real life. The experimental condition of social inclusion could, therefore, be more unexpected for BPD patients who may then change their subjective perception to match expectations and to stabilize negative beliefs. Increased prefrontal activation as found by Domsalla et al. (17) and in our present study may represent a correlate of the effort to resolve this cognitive dissonance. As suggested by Etkin et al. (32) increased dorsolateral and dorsomedial prefrontal activation could be interpreted as a correlate of enhanced conflict evaluation, while increased salience-network activation as observed in adolescents with NSSI could represent enhanced conflict perception.

The differences between adolescents with NSSI and young adults with BPD (+NSSI) are interesting, since NSSI is one of the core symptoms of BPD. As found in previous fMRI studies (33, 34), pain (and especially tissue damage) seems to have a soothing effect in BPD patients after stress induction on a psychophysiological level and regarding neuronal processing. It was shown that the experience of painful stimuli led to increased connectivity between prefrontal and limbic regions, thus resulting in increased inhibition of limbic areas, and especially the amygdala (33, 34). The Biosocial Development Model of BPD (35) describes a vulnerability early in life, which is initially expressed by elevated impulsivity and later by increased emotional sensitivity. In this model, NSSI as a symptom could be an early indicator for the development of BPD in combination with increased sensitivity to social rejection as a second factor. NSSI and BPD might, therefore, represent a developmental continuum, starting with an increased sensitivity to social exclusion in adolescence, which generalizes to neutral or even positive social situations, thereby increasing symptom load of BPD and severity of this disorder. Correlations of behavioral results with fMRI data in this study can be seen to support this theory. The observation that those NSSI patients, who came closer to the BPD group behaviorally, did so also in terms of neurobiological effects, may support the idea that NSSI in adolescents and BPD in young adults can be seen as two expressions of a continuum. Certainly, longitudinal studies will be necessary to verify this developmental perspective. Longitudinal studies could also substantiate the knowledge whether adolescents with NSSI, who were particularly sensitive to social exclusion in this study, would be at a marked greater risk to develop BPD later on in life and would also develop a still more generalized sensitivity to social situations. From a clinical point of view, understanding why some adolescents with NSSI go on to develop the full symptomatology of BPD, while others do not, is key in tailoring interventions to patients’ needs.

When interpreting the results of this study, the following limitations have to be taken into consideration: the sample sizes of the individual groups included in this study were rather small. Therefore, significant differences may have remained undetected and results cannot be generalized easily. Replication of these findings in larger samples is, therefore, crucial. Ideally, patients with and without a diagnosis of BPD would have been of the same age. However, NSSI without concomitant BPD is very rare in young adults, while a safe diagnosis of developing BPD is comparatively infrequent in adolescents. We tried to overcome this issue by age-matched control groups. Still, a direct comparison between the two groups would have been more ideal but was not possible due to the age difference which is almost impossible to avoid. Furthermore, participants of the BPD group showed significantly higher scores of depression and were more often diagnosed with PTSD than adolescents of the NSSI group. While it is true that all of the participants in the NSSI and the BPD groups had clinically relevant levels of depression, we were able to show in a previous study of matched controls with depression but without NSSI that the effects of social exclusion were specific to NSSI, not to depression or psychiatric impairment in general (18). However, to increase validity of our findings, future replications with studies including clinical controls are necessary. Patients with BPD in this study received a larger proportion of antidepressant medication. Antidepressant medication could, therefore, have been the reason why, for example, the hyperactivation in the putamen was only seen in the NSSI group, but not in the BPD group. Therefore, this result has to be interpreted with caution. Moreover, future studies should evaluate feelings of being excluded separately for the exclusion, inclusion and control conditions as described in Ref. (6, 17). Other factors that may have influenced our results and could be interesting for future research are dissociative states during social exclusion, and differences in psychophysiological reactions.

Strengths of this study were gender- and age-matched healthy controls, by which possible age effects could be controlled for, careful psychological evaluation and the use of a robust paradigm to measure effects of social inclusion and exclusion.

In summary, the present study points toward a distinct neurobiological signature of NSSI in adolescence as compared to BPD in young adulthood. Developmental processes might be involved in a generalized sensitivity to social situations, starting with enhanced sensitivity to social exclusion in adolescence, which is correlated with NSSI and may then expand to neutral and even positive social situations with increasing BPD symptomatology. Future longitudinal studies would be necessary to elucidate this possible developmental pathway.

The study was approved by the Institutional Review Board of Ulm University, Ulm, Germany. Written informed consent was obtained from legal guardians (where applicable) and all participants. All procedures were performed according to the Declaration of Helsinki.

RB was involved in designing the study, recruited participants, conducted fMRI scans and psychological interviews, analyzed data, calculated results, and was closely involved in drafting and revising the manuscript. PP and GG gave statistical, methodological, and content advise and revised the manuscript. DN and MB recruited participants, conducted fMRI scans, analyzed data, and revised the manuscript. BA was involved in deigning the study, analyzed data, calculated results, and was closely involved in drafting and revising the manuscript.

All authors declare no conflict of interest with the potential to bias the work. PP has received research funding from the German federal ministry of research and education, the German federal agency for drugs and medical products, the German research foundation, the Baden-Wuerttemberg Foundation, Lundbeck pharmaceuticals, and the Volkswagen Foundation.