The borderline personality disorder (BPD) is a complex and serious mental disorder characterized by a pervasive pattern of instability in affective regulation, interpersonal relationships, self image and behavioral control (Siever et al., 2002; Skodol et al., 2002a,b). BPD is characterized by progressive functional impairment, substantial treatment utilization and a mortality rate by suicide of almost 10–50 times higher than the rate in the general population (American Psychiatric Association, 2001).

The median population prevalence is estimated to be 1.6% but may be as high as 5.9%. The prevalence of BPD is about 6% in primary care settings, about 10% among individuals seen in outpatient mental health clinics, and about 20% among psychiatric inpatients (DSM-5). The prevalence of BPD may decrease in older age groups (DSM-5). Impulsivity and aggressiveness are heritable traits that may contribute to the psychopathology of BPD (Lieb et al., 2004; Siever and Weinstein, 2009).

Impulsive behavior has been conceptualized as an imbalance between the “top-down” control or “brakes” provided by the frontal cortices and excessive “bottom-up drives” triggered or signaled by limbic regions (Siever, 2008). Regarding prefrontal cortices, the dorsolateral prefrontal cortex (DLPFC) represents a key structure, along with the orbitofrontal cortex, in impulsivity control (Soloff et al., 2003; Chanen et al., 2008; Matsuo et al., 2009). Reduction of DLPFC gray matter volumes has recently been shown in BPD patients compared to healthy controls (HC; Tomoda et al., 2009; Brunner et al., 2010). Moreover, a recent PET study on BPD patients, showed that DLPFC failed to activate during the top-down cognitive control of aggression (New et al., 2007, 2009). Recent meta-analysis (Ruocco et al., 2013; Schulze et al., 2016) evidenced bilaterally reduced activation within the DLPFC in BPD patients compared to healthy subjects associated with negative emotion processing, thus confirming the PFC’s role in emotion regulation through altered connections with the insula. Finally, the interplay between DLPFC and hippocampus appears to be deeply involved memory, emotional and behavioral control in BPD (Schmahl et al., 2003, 2004; Anderson et al., 2004; New et al., 2007; Sala et al., 2009; Morandotti et al., 2013).

Literature about prefrontal functions in BPD has been controversial for many years and the cognitive profile of BPD remains to be fully clarified. Clinical theoreticians and researchers have proposed that the psychopathology of BPD is associated with disruptions in basic neurocognitive processes. BPD patients perform worse than HC in multiple neurocognitive domains (Bazanis et al., 2002; Dinn et al., 2004; Monarch et al., 2004). Other empirical evidences are inconsistent with these results and revealed a lack of specificity and stability of these reported neuropsychological deficits (Fertuck et al., 2005; LeGris and van Reekum, 2006; Silbersweig et al., 2007). A recent meta-analysis (Unoka and Richman, 2016) however highlighted impairments in neuropsychological functioning in BPD, especially in domains of decision making, memory, executive functioning, processing speed, verbal intelligence and visuospatial abilities. Moreover BPD seem to perform differently according to specific inhibitory tasks. Behavioral inhibition focuses on holding or suppression of an already selected or initiated response and thus, late control processes (Stahl et al., 2014). Stop-signal and go/no-go tasks represent the most important behavioral inhibition tasks (Aron, 2011; Swick et al., 2011). “Experimental paradigms assessing emotionally neutral impulse control in BPD have revealed inconsistent results” (Sebastian et al., 2013). Two fMRI studies focusing on behavioral inhibition in emotional context have revealed prefrontal dysfunction especially if modulated by negative emotions (Silbersweig et al., 2007; Jacob et al., 2013). For these reasons we set up a specific test that assesses information processing biases for positive and negative stimuli (see “Materials and Methods” Section).

Recent anatomical and clinical evidence has shown that the cerebellum, primarily considered a motor control structure, is also involved in higher cognitive functions (Ito, 1993, 2005, 2008; Ivry et al., 2002; Schmahmann, 2004; D’Angelo and Casali, 2012) and behavioral changes, such as impulsive behavior. Neuroanatomical research has shown that the cerebellum projects to the PFC through the ventrolateral thalamic nucleus (VL) including the mediodorsal thalamic nucleus (MD; Yamamoto et al., 1992; Middleton and Strick, 2001) and the reticular nucleus of the thalamus (RNT; Çavdar et al., 2002). High resolution tractography in humans in vivo has recently shown that about 40% of fiber tracts leaving the cerebellum through the superior cerebellar peduncle actually reach the PFC through the VL (Palesi et al., 2015). A fMRI study (Allen et al., 2005) demonstrated fluctuations in signal in the dentate nucleus correlated with fluctuations in cerebellar, thalamic, limbic, striatal, and cerebrocortical regions including parietal and frontal sites, with prominent coherence in DLPFC (areas 9 and 46 mainly). These anatomical and functional connections between the cerebellum and the PFC suggest that the cerebellum is involved in non-motor circuits. Additional evidence that the cerebellum plays a key role in higher-order cognitive functions and behavior comes from imaging studies altered cerebellar volumes in patients with neuropsychiatric diseases such as Attention-Deficit/Hyperactivity Disorder (ADHD) and schizophrenia (Roth and Saykin, 2004; Andreasen and Pierson, 2008; Thomann et al., 2009) and altered cerebellar metabolism in patients with obsessive compulsive disorder (Pujol et al., 2004; Nabeyama et al., 2008). Impulsivity is a core feature of these neuropsychiatric diseases (King et al., 2003). Although in the literature the reports describe increased impulsivity in BPD, the evidence about cerebellar involvement is limited and inconsistent.

In addition to neuroimaging, recently TMS has also been used to identify neural substrates of psychiatric disorders. Repetitive transcranial magnetic stimulation (rTMS) involves the delivery of trains of magnetic pulses to produce changes in cortical excitability that persist beyond the duration of the stimulus. The mechanisms through which these protocols alter local neural circuits are believed to involve processes similar to synaptic long-term potentiation (LTP) and long-term depression (LTD; Fitzgerald et al., 2003). One of the earliest rTMS studies (Pascual-Leone et al., 1994) demonstrated that impulse trains at high-frequency (>5 Hz) generally increased cortical excitability (as measured by the size of motor evokd potentials [MEPs]). These effects persisted for 3–4 min after the end of stimulation. On the contrary, rTMS at frequencies of 1 Hz or below generally decreases cortical excitability. Recent studies showed that rTMS caused 30% changes in cortical excitability persisting for about 30 min in the EEG (Thut and Pascual-Leone, 2010).

Theta-burst cerebellar rTMS has been successfully used to modulate motor timing (Del Olmo et al., 2007) and procedural learning (Torriero et al., 2011) with a reflexion on finger movement (Del Olmo et al., 2007), saccadic eye movements (Colnaghi et al., 2011), eye-blink classical conditioning (Monaco et al., 2014). These effects are likely to involve metaplasticty of cerebello-cortical connectivity, with relevant effects in progressive supranuclear palsy (Brusa et al., 2014) in dystonia (Koch et al., 2014), ataxia (Bonnì et al., 2014). Theta-burst cerebellar rTMS proved able to interfere with electrical excitablity of the PFC in a Go-No-Go task (Picazio et al., 2016). Conversely, cerebellar low-frequency rTMS increased facilitation in the primary motor cortex (M1; Oliveri et al., 2005). In the current work we reasoned that an excitatory impact on the PFC could have been exerted by cerebellar low-frequency rTMS and that this effect could impact on the Go-no-Go task execution in BPD patients.

The rTMS-induced changes in motor cortex excitability have been monitored in various studies in a pre-post stimulation design similar to this experimental study. Changes in cortical excitability pre vs. post-rTMS stimulation were interpreted as measures of rTMS-induced changes in synaptic plasticity (Fitzgerald et al., 2006). A recent TMS study (Cailhol et al., 2014) has investigated the effect of a high-frequency rTMS in 10 sessions delivered over the right DLPFC in 10 BPD patients. BPD in the rTMS group showed improvements in anger, affective instability and planning, thus confirming our hypothesis about prefrontal involvement in BPD patients’ dyscontrol. The cerebellar TMS was used in previous studies to test non-motor systems. Del Olmo and Minsk (Del Olmo et al., 2007; Minks et al., 2010) used rTMS (1 Hz, 600 pulses) over cerebellar hemispheres and studied its effect on the performance of a finger-tapping; Koch et al. (2007) studied the perception of time with TMS in two experiments; Desmond et al. (2005) tested whether disruption of the right superior cerebellum was able to impair verbal working memory performance. To the best of our knowledge there are no previous studies focusing on cerebellar TMS in personality disorders. Moreover, we had to evaluate the problem of functional cortical asymmetry in BPD, in order to define the target brain area of stimulation. Right cerebral involvement seems to be predominant in the neurobiology of the disorder (Irle et al., 2005; Sala et al., 2009; Morandotti et al., 2013). In a previous study (Barnow et al., 2009), a reduced cortical silent period (CSP) in the right cortex was found in BPD patients compared with HC. Irle et al. (2005) supported a reduced CSP in the right cerebral cortex, but not in the left side, showing a reduced right parietal cortex volume in patients with BPD. The authors hypothesized a neurodevelopmental deficit in the right hemisphere that may be linked to several traumatic life events of BDP patients. Additionally, authors observed an increased leftward asymmetry as a protective factor in BPD patients with post-traumatic stress disorder (PTSD) for the development of disabling psychotic syndromes (Irle et al., 2005).

For these reasons we first focused on the circuit involving the left cerebellar cortex and the right PFC. The study should be extended in both cerebellar hemispheres in order to clarify the differences in cerebello-thalamo-cortical connections bilaterally. Cortical inhibition deficits have been demonstrated in several disorders with deficits in impulsive control (e.g., ADHD, tic disorder, Tourette syndrome) by using TMS protocol. Recent findings support an association between BPD and cortical inhibition deficits as evidenced through TMS (Barnow et al., 2009).

In this TMS study we aim at investigating whether cerebellum and PFC functions may be involved in impulsive reactions in BPD through a pre/post stimulation design.

The present study has been accepted by the local Ethical Commitee (11.09.2014) and received formal authorization from the Neurological Institute, Casimiro Mondino Foundation (17.10.2014).

The central result of this article is that cerebellar rTMS could modulate the response of BPD patients in an AGN task. Given that AGN is normally controlled by the PFC, this result also suggests the presence of prefrontal control deficits in BPD that may be linked to dyscontrolled impulsivity and sustained by anatomical and functional impairment of the DLPFC.

The comparison between accuracy and RTs in HC and BPD subjects suggests an interaction between cognitive demand and inhibitory control in BPD. In particular, the data indicate that the neuropsychological performance at AGN in BPD is especially impaired when cognitive demands are high and require complex associative capacities, as in the final double category blocks of the test. In fact the task in the last two blocks requires a higher working memory load, since the subject must keep in mind both categories to give the CR. Deficits in neuropsychological performance in BPD when working memory demands are high has been demonstrated elsewhere in our previous work (Sala et al., 2009; Lazzaretti et al., 2012). This finding supports the relationship between working memory load and impulsivity, that has been already demonstrated in literature (Hinson et al., 2003).

At baseline, patients performed particularly well in the NEGATIVE single category block (93% of CRs, see Table 2). An increased attentional process in the domain of selective attention when negative stimuli are involved and thus a deficit of inhibition of irrelevant information of aversive nature has been demonstrated in various studies (Domes et al., 2006; Soloff et al., 2015). The inability of borderline patients to disengage attention from negative stimuli may explain the good performance in the block where lack of inhibition of negative stimuli is advantageous (i.e., the target category NEGATIVE) and a worse performance where inhibition of negative stimuli is required (i.e., the block with target category POSITIVE).

The most attracting aspect of these results is that the BPD improvement after cerebellar rTMS emerged when the AGN task became more complex. Before rTMS, BDP patients performed worse than HC in the POS-INS block and in the NEG-FRU block, while after rTMS, the two groups became comparable (see Table 3). Thus cerebellar rTMS demonstrates a relevant effect in the deficitary but not in the preserved cognitive domains. The cerebellum is deeply involved in integration and coordination of different specific stimuli (attention, sensory-motor control, error detection and prediction) and this could be the reason why rTMS effect was more evident in complex associative tasks. We hypothesize that BPD patients have an altered cerebello-thalamo-cortical functional connections resulting in emotional dysregulation and disturbed impulse control. rTMS over the left cerebellum seems to interfere with existing functional connections with a facilitating effect on prefrontal inhibitory control.

The interpretation of this result in terms of neuromodulation is quite complex since there are no preliminary data about the neurophysiological effect of rTMS at 1 Hz on the cerebellum in BPD. It is known that rTMS can modulate cortical excitability in a frequency dependent manner. High frequency rTMS (≥5 Hz) was shown to induce LTP-like effects, whereas low frequency rTMS (≤1 Hz) leads to LTD-like effects (Fitzgerald et al., 2006; Thut and Pascual-Leone, 2010). Empirical data (Oliveri et al., 2005) suggest that low-frequency rTMS trains can produce plastic changes in the cerebellar cortex, similar to those reported for many cerebral areas, including motor, prefrontal and premotor cortex. Indeed, 1 Hz rTMS over cerebellar cortex produced different effects on Intracortical Facilitation (ICF) in three different studies, which may be related to different inter-stimulus intervals chosen in those studies (Fierro et al., 2007; Oliveri et al., 2005).

Recent research has focused on the potential effects of rTMS and tACS on cortical oscillatory activity in a frequency specific manner (Alagapan et al., 2016; Peng and Tang, 2016). Alongside the LPT/LTD mechanisms, rTMS could modify cortical excitability and plasticity inducing modulations of ongoing oscillations, which are deeply involved in functional brain networks. In particular, rTMS in the alpha frequency band seems to vary sensory detection, perception and performance (Klimesch et al., 2003; Thut et al., 2011).

The way by which temporally patterned non invasive stimulations alter cortical oscillatory dynamics is still controversial. In particular, high-frequency rTMS (10 Hz) induces a transient synchronized activity for delta (δ) and theta (θ) rhythms while low-frequency rTMS (1–5 Hz) shows the opposite effect of de-synchronizing low-frequency brain rhythms (Thut et al., 2003; Fuggetta and Noh, 2013).

The mechanistic understanding of underlying changes in brain activity should be further analyzed, in order to create more targeted stimulation designs (Fröhlich, 2015). For example, Farzan et al. (2016) recently demonstrated that iTBS on right lateral CrusI/II and Lobules VIIA/VIIB subregions increased the complexity of brain signal across multiple time-scales in cortical areas corresponding to the network stimulated through cerebellum.

Moreover, it has been demonstrated by MRI that the activity changes induced by rTMS are not restricted to the directly stimulated area, but involve functionally connected remote areas (Paus et al., 2001; Strafella et al., 2001). Such cross-modal plasticity-like effects may be transferred via direct cortico-cortical connections, indirectly via multi-sensory association areas, or via subcortical interplay at the thalamic level. Thus, it is possible that cerebellar rTMS on the posterior lateral lobe, a “cognitive” part of the cerebellum, has an effect on the PFC through the cerebello-thalamo-cortical fiber tracts (Palesi et al., 2015). Our results confirm that low-frequency rTMS has a facilitatory effect on the PFC, as hypothesized in the “Introduction” Section.

The neurophysiological mechanisms are not easily predictable though, and interpretation of reported results is even more complex when considering indirect measures of cortical excitability such as a neurocognitive performance instead of electromyographic (EMG) activity. It is possible that low-frequency rTMS induced a transient depression of Purkinje cell excitability, thereby disinhibiting cells in the dentate nucleus, the site of origin of the output fibers directed toward the cereberal cortex. Dentate cell disihinibition could have resulted in a facilitation of the thalamus and PFC.

It is known that “these cortico-cerebellar loops are involved in the identification of errors and novelty (sensory prediction) and can trigger automatic corrections, promote learning and redirect attention” (D’Angelo and Casali, 2012). Accordingly, cerebellar alterations affecting the cerebello-cortical loops may lead not only to the well known motor abnormalities but also to behavioral, cognitive and affective alterations, usually correlated to frontal regions and limbic areas. The cerebellum is fundamental for contextualizing external and internal stimuli and coordinating their spatio-temporal evolution, generating coherent ensemble activities. Therefore, dysfunction of the cerebellar circuits and of information reentry toward the frontal and parietal cortex may contribute to preventing the formation of coherent and contextualized behaviors. Additionally, the cerebellum is critical for revealing differences between predictions elaborated by the cortex and the stimuli conveyed by the senses. Thus, dysfunction of the cortico-cerebellar circuits may affect the detection of novelty and impair attention switching (D’Angelo, 2011). These complex functions (sensory prediction and novelty detection) seem to be impaired in BPD patients and are widely described by clinicians. For this reason a study focused on cortico-cerebellar projections’ functionality could help us to clarify the neurophysiological correlates of BPD patients.

To the best of our knowledge, this could be the first study investigating if BPD patients show differences in cerebellar rTMS parameters compared with HC.

GZDV wrote the protocol, performed patients selection and diagnostic evaluation, performed TMS tests and wrote the body of the article. RM contributed to patient selection and text writing. JM performed TMS tests and data analysis. NC engineered the stimulation/recording system and the data analysis programs. DB contributed to healthy control selection. EC and FB coordinated the patients facility and diagnostic operations. EDA coordinated the work and ethical committee application, supervised the neurophysiological part and the final version of the text.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.