Chronic pain is one of the conditions with the highest negative impact on quality of life throughout the world. WHO data show that globally, lower back pain and migraines constitute the two leading causes of years lived with disability (GBD 2016 Disease and Injury Incidence and Prevalence Collaborators, 2017). Unfortunately, effective treatments for chronic pain are still lacking, due in large part to the incomplete understanding of the underlying pathogenic mechanisms. A fundamental difference between pain and other percepts is that there is no pain cortex; on the contrary, pain appears to be the perceptual creation originating from the complex information flow between multiple brain areas (Singh et al., 2020). This is not surprising, given that normal pain processing functions as a signal of danger. Healthy fear learning in response to a pain-inducing stimulus requires integration of sensory information (Where am I hurt?), information about the context (Where is the pain-inducing stimulus located?), and emotional salience (How much damage can this cause?). Therefore, in pain, and in chronic pain in particular, the sensory perception is inextricably mixed with cognitive symptoms and with a powerful emotional component (Berryman et al., 2013; Porreca and Navratilova, 2017; Bell et al., 2018). Consequently, the focus of pain researchers has widened to include brain areas beyond the ones strictly involved in somato-sensory perception. Among these new areas of interest, the ventral areas of the mPFC, the prelimbic (PL) and infralimbic (IL) sub-regions, are particularly intriguing. Due to the diverse functions of the major inputs to mPFC, these regions are ideally situated to process the multi-faceted information relevant to pain perception. Furthermore, the mPFC has been shown to mediate multiple components of chronic pain including sensory (Cordeiro Matos et al., 2015; Lee et al., 2015), cognitive (such as memory and attention problems, Moriarty et al., 2011; Baker et al., 2016) and emotional (such as catastrophizing, Galambos et al., 2019) components.

Current models suggest that functional deactivation of at least part of the ventral mPFC is a major pathogenic mechanism in different chronic pain conditions. A first hint of the major involvement of the PFC in chronic pain was provided about 15 years ago, when Apkarian et al. (2004) showed that the PFC of back pain patients shows a gray matter loss that is proportional to the duration of the pain. Subsequent animal studies confirmed the mPFC involvement and provided evidence suggesting a functional deactivation of the ventral areas of the mPFC early in the pain chronification process, although some differences exist between layer 2/3 and layer 5 neurons. In layer 2/3 pyramidal cells, a decreased α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) and N-methyl-D-aspartate (NMDA) ratio, a measure of glutamatergic synaptic strength (Bredt and Nicoll, 2003; Kauer and Malenka, 2007), and decreased intrinsic excitability (Metz et al., 2009; Wang et al., 2015), together with an overall decrease in glutamate concentration in the PL (Kelly et al., 2016), suggest that both synaptic and intrinsic excitability are reduced. In layer 5 neurons, the decreased synaptic excitability (Kelly et al., 2016) and impaired excitatory cholinergic modulation (Radzicki et al., 2017) are only in part countered by a slight increase in intrinsic excitability (Cordeiro Matos et al., 2015; Wu et al., 2016). Mechanistic studies of chronic pain suggest that the ventral mPFC output exerts a critical modulatory role on pain perception, mostly through activation of descending pathways (Cheriyan and Sheets, 2018; Huang et al., 2019). This conclusion is supported by the fact that, in rodent pain models, optogenetic activation of the ventral mPFC has analgesic effect (Lee et al., 2015; Zhang et al., 2015). The cellular mechanisms of the region’s deactivation in chronic pain remain incompletely understood, but, as noted, abundant evidence suggests that synaptic mechanisms provide a major contribution to this functional state (Ji et al., 2010; Kelly and Martina, 2018). In this context, alteration of glutamatergic inputs to the mPFC appears to have a central role in the development of the chronic pain phenotype.

What are the major glutamatergic inputs to the ventral mPFC? The rodent PFC was originally defined as the cortical area receiving inputs from the medio-dorsal thalamus (Leonard, 1969; Krettek and Price, 1977). However, the thalamic inputs are not the only glutamatergic afferents to the ventral mPFC, and important projections originate in several other brain areas, including the contralateral mPFC, the amygdala, and the ventral hippocampus. Other inputs include the ipsilateral agranular insular cortex, which provides glutamatergic inputs to the PL (Hoover and Vertes, 2007), and the claustrum, which provides inputs to both the IL and PL. All four major inputs provide monosynaptic contacts onto dendritic spines on layer 2/3 pyramidal neurons (Little and Carter, 2012), as well as targeting neurons in deeper layers (Bacon et al., 1996; Parent et al., 2010; Collins et al., 2018; Kelly and Martina, 2018). Contacts with neurons whose cell bodies are located in deeper layers are the result of both direct innervation of these layers and of terminations onto the long apical dendrites of pyramidal cells that reach all the way to layer 1. For example, the widely branched apical tufts of layer 5 pyramidal neurons receive thalamic inputs up to layer 1 (Parent et al., 2010; Cruikshank et al., 2012).

Here we briefly summarize how the three main extra-cortical glutamatergic synaptic inputs (thalamic, hippocampal, and amygdalar) to the ventral mPFC are affected in rodent models of chronic pain and discuss how these changes may contribute to the chronic pain phenotype.

Numerous findings support the idea that the ventral mPFC is deactivated in chronic pain conditions (Ji et al., 2010; Wang et al., 2015; Kelly et al., 2016; Radzicki et al., 2017). It is reasonable to hypothesize that the decreased function of the mPFC network is mediated by the compounded action of three distinct mechanisms. The first two mechanisms that contribute to the mPFC deactivation in the chronic pain condition are synaptic: a combination of input-specific depression of the excitatory glutamatergic inputs, and potentiation of GABAergic inhibition. The reduced glutamatergic drive might explain the shortened dendrites of layer 5 mPFC neurons (Kelly et al., 2016), which, possibly in combination with the reduced length of the axon initial segments (Shiers et al., 2018), may provide a cellular basis for the grey matter reduction associated with chronic pain in patients and animal models (Apkarian et al., 2004; Seminowicz et al., 2009).

The altered synaptic inputs may also cause a more general effect, which is a widespread alteration of the excitation/inhibition (E/I) balance. E/I imbalance is regarded as a key pathogenic mechanism in numerous neurodevelopmental disorders (Deidda et al., 2015; Kim et al., 2021). Thus, it is possible that through recurrent inhibitory and excitatory circuitries the mPFC network acts as a magnifier of the effects of even minor alterations in specific inputs, so that relatively minor synaptic changes end up producing large behavioral effects.

Clearly, these synaptic effects do not take place in a vacuum but interact with the neuronal intrinsic excitability. For example, many of the thalamic input to the ACC and the PL are provided by the same thalamic neurons, which form terminals in these two different cortical areas (Kuramoto et al., 2017); yet, in chronic pain conditions the ACC is hyperexcited (Zhao et al., 2006; Xu et al., 2008; Blom et al., 2014; Singh et al., 2020) and the PL is inhibited. This finding suggests that a third mechanism, which is represented by alterations in intrinsic excitability such as increased input resistance in the ACC (Santello and Nevian, 2015) or increased action potential threshold in the PL (Wang et al., 2015), contributes to the final electrophysiological outcome of altered synaptic inputs.

As to the functional consequences of the reduced mPFC output, it is likely that the most relevant effects on the sensory components of pain perception are mediated by the descending projections to the periaqueductal gray (PAG) and to the nucleus accumbens (NAc). This idea is supported by the finding that the inhibition of the mPFC output appears particularly strong in neurons projecting to the PAG (Cheriyan and Sheets, 2018). However, activation of the mPFC projection to the nucleus accumbens has analgesic effect by itself (Lee et al., 2015), as does direct optogenetic activation of the nucleus accumbens core (Kc et al., 2020); additionally, inhibition of the PFC input to the nucleus accumbens amplifies both sensory and affective symptoms of acute pain in naïve animals (Zhou et al., 2018). Thus, the exact identity of the brain networks that relay the mPFC deactivation-mediated modulation of sensory pain remains to be established.

The cognitive and affective components of the chronic pain syndrome are likely mediated by different mPFC projections, and the correlation between mPFC firing and cognitive performance appears different in control and pain conditions. For example, disruption of the ventral hippocampus to PFC connectivity was reported in a model of inflammatory pain (Ma et al., 2019). Additionally, theta-rhythm connectivity between the (dorsal) hippocampus and the mPFC is reduced in neuropathic pain, and this is correlated with performance on a working memory task (Cardoso-Cruz et al., 2013). As expected, in control conditions, optogenetic inhibition of PL firing and of hippocampus-mPFC coherence disrupts working memory performance. In pain animals, however, optogenetic inhibition of PL pyramidal cells restores normal theta rhythms and coherence and improves working memory (Cardoso-Cruz et al., 2019a). As the chronic pain phenotype (both neuropathic and inflammatory) is characterized by reduced mPFC excitability and disrupted ventral hippocampus–PL connectivity, the fact that pyramidal neuron inhibition improves cognitive performance in pain animals is unexpected and suggests a context where synaptic modulation of pyramidal cell activity, possibly through E/I imbalance (discussed above) is more important than simple firing frequency. Although the detailed network effects of this optogenetic modulation on PL output remain to be clarified, it is interesting that this manipulation does not directly affect pain measures (Cardoso-Cruz et al., 2019b). Another group, however, recently showed that modulation of the input from the ventral hippocampus to the mPFC has a significant impact on pain perception, as chemogenetic activation of the ventral CA1-IL pathway alleviated spontaneous pain in a model of inflammatory pain (Ma et al., 2019). Thus, these data suggest that input-specific effects are central to the mPFC role in pain modulation. At the same time, the role of the ventral hippocampus in pain perception also remains unclear, because in a rodent model of early stage neuropathic pain neither optogenetic nor pharmacological activation of the ventral hippocampus produced analgesic effects (Wei et al., 2021). Multiple factors may explain these differences, including the specific pain model used, the pain duration, the strength, duration and pattern of the hippocampal stimulation, and the type of neurons stimulated.

Thus, as expected for a complex percept such as pain, the brain mechanisms involved are extremely complicated. An additional layer of complexity is provided by the fact that although both the PL and IL are inactivated in chronic pain, it is unclear whether their relative function level is equally affected. For example, a recent study in a neuropathic pain model found that the balance between excitatory and inhibitory effects caused by activation of the BLA input to the mPFC is decreased in the PL, but not the IL (Cheriyan and Sheets, 2020). This is not a trivial point because the PL and IL have different functions. For example, in coding aversion responses, they are believed to play opposite roles, with the PL promoting aversion learning and the IL facilitating extinction of aversive responses (Gourley and Taylor, 2016). Whether similar differences exist in the roles that the IL and PL play in the pain phenotype remains unclear. The picture, however, is even more complex, as we must also consider the temporal structure of mPFC rearrangement in the pain phenotype. Many of the changes we have discussed take place relatively early in the course of the disease (in rodents, within 1–2 weeks from the onset of the peripheral injury). This is a particularly interesting timeframe, because it likely represents the transition from acute to chronic pain. Yet, brain networks continue to change in the presence of neuropathic injuries, either continuing along their disease trajectory or due to adaptive responses (Ren et al., 2021). Finally, the pain phenotype also exhibits important sex dimorphism, including in mPFC-specific tasks (Shiers et al., 2018). Accordingly, recent data show that sex-specific differences are apparent in pain-associated effects on GABAergic modulation in the rodent Mpfc (Jones and Sheets, 2020). Whether these sex-specific differences in mPFC function are relevant for the sex dimorphism of the chronic pain phenotype will likely represent an intriguing field of research in the near future.

CK focused on researching hippocampal inputs. TJ focused on amygdalar inputs. MM focused on thalamic inputs and drafted the manuscript. All authors contributed to the final version.

CK is employed by company Aptinyx Inc. The remaining 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.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.