1.1.1 Development of pain circuits during gestation

Human brain development starts during the first weeks of gestation with a massive proliferation of neural cells peaking around the 20th week of gestation. For the nociceptive system, the late stage of gestation is considered as a critical period of development (Figure 1). Cutaneous innervation by sensory neurons starts at the 2nd trimester of gestation, when some non-nociceptive medullary reflexes are already established, but the connections between sensory neurons and the spinal cord (SC) continue to mature during the whole gestation (Konstantinidou et al., 1995). The small unmyelinated C fibers are the last sensory fibers to establish functional connections with medullary neurons. Concerning higher brain centers, the sensory and motor cortices can be distinguished at around week 27–28 (Afif et al., 2015) and the six-layered structure of the cortex can be observed as soon as 32 weeks (Kostovic & Judas, 2010). The functional connections with higher brain centers are created during the last trimester, with thalamocortical neurons starting to create synapses with cortical neurons of layer IV at around 32 weeks (Kostovic & Judas, 2010). Myelination starts during gestation, but the process continues after birth and during infancy, with a peak during the first year of post-natal life. Yet, the myelination of ascending fibers is complete at least until the thalamus at the 30th week of gestation. The capacity of the cortex to integrate sensory informations, including nociceptive ones, has been confirmed with the use of functional near-infrared spectroscopy (fNIRS) imaging in very premature infants as soon as 25 weeks of gestational age and at later age (Bartocci et al., 2006; Slater et al., 2006). Evoked cortical potentials can also be recorded in response to a stimulation of peripheral receptive field as soon as the 30th week of gestation (Slater et al., 2010). At 31 weeks, pain-triggering trigeminal odors induced the activation of the olfactory, frontal, and somatosensory cortices assessed using fNIRS, and induced pain behavior (Frie et al., 2018). Finally, the use of EEG recordings in infants showed that before 35–37 weeks, touch and noxious lance of the heel evoked non-specific neuronal activity, which transitioned to modality-specific potentials after 35–37 weeks (Fabrizi et al., 2011). Similarly, discriminative facial expression in response to non-noxious and noxious stimuli coincides with the maturation of specific brain responses and emerges between 33 and 34 weeks (Green et al., 2019).

Overall, imaging studies clearly reveal that the term newborn infant brain is able to process sensory and pain information a few days after birth. Brain activation in infant increases with the intensity of the nociceptive stimulus (Williams et al., 2015) and the regions involved in pain processing are similar to the ones activated during a pain experience in adults (Goksan et al., 2015).

In the rat, the gestational period lasts around 3 weeks, and it is considered that the developmental stage of the rat at birth is similar to the one in human at the beginning of the 3rd trimester of gestation. As in human, most pain circuits are established before birth, although some are not fully functional. To summarize (for more details, see Fitzgerald, 2005), sensory neurons in the dorsal root ganglions (DRG) are born at E12 and start innervating their peripheral and central targets, which are reached at E15 and E19 for lumbar SC A-type and C-type sensory neurons respectively. C-type neurons also seem to create functional synapses later than A-type neurons, possibly due to the maintenance of silent C-type glutamatergic sensori-spinal synapses (Baba et al., 2000). Projections of SC neurons to higher brain centers are also functional at birth and are able to modulate nociceptive responses in the newborn (Man et al., 2012).

1.1.2 Post-natal reorganization/refinement of pain circuits

If nociceptive circuits are established and functional, for most of them, as soon as the 3rd trimester of gestation in the human fetus and at birth in the rat, they are still immature and undergo strong plastic changes after birth. This is especially true at the SC level. In the newborn rat, A fibers input in the dorsal horn are predominant and more diffused compared to the adult SC. This might explain the hypersensitivity of newborn animals (i.e. they are activated at low stimulation threshold and C fibers are not functional) and could account for the extended receptor field of dorsal horn neurons. This nociceptive hypersensitivity has also been described in human newborn infants, and is even more intense in preterm infants (Cornelissen et al., 2013). During the first three post-natal weeks in rodents, A-fibers innervation tends to re-organize and withdraw from superficial dorsal horn layers I and II to localize only in deeper layers, following an activity-dependent process (Beggs et al., 2002). Chloride homeostasis is also subjected to critical developmental changes, controlling the switch from depolarizing to hyperpolarizing GABAergic currents (Ben-Ari et al., 2012; Cordero-Erausquin et al., 2005). In the SC, GABA action matures during the first three postnatal weeks (Cordero-Erausquin et al., 2005). Inhibitory processing in the SC also involves glycinergic function, which is absent in the newborn, but appears in the second week of life (Baccei & Fitzgerald, 2004). The properties of inhibitory interneurons also mature after birth, as it has been recently demonstrated with dynorphin interneurons, whose inhibitory input is weaker during the neonatal period (Brewer et al., 2020).

The early post-natal period is also critical for the maturation of endogenous modulatory controls of pain. This has been widely studied in rodents for the opioidergic system, subjected to huge post-natal maturation at both the supraspinal and spinal levels. The expression of mu opioid receptors (MORs), delta opioid receptors (DORs), and kappa opioid receptors (KORs) changes after birth, with a downregulation of MORs and DORS after birth in non-nociceptive sensory DRG neurons (Beland & Fitzgerald, 2001). In the SC, DORs are not detected until P7, while MORs and KORs are expressed at P0 in the rat, but mature after birth to reach adult levels (Rahman et al., 1998). The recruitment and modulation of rostroventral medulla or periaqueductal gray (PAG) descending controls by nociceptive input is ineffective in the rat before P8 (Schwaller et al., 2016) and surprisingly lead to an increased sensibility of the pups to mechanical stimulation at P21(Hathway et al., 2012). In the PAG, expression of the pro-opio-melanocortin gene coding for β-endorphins (and many other peptides after enzymatic cleavage) peaks at P21 (Kwok et al., 2014). Diffuse noxious inhibitory controls (DNIC), referred to as conditioned pain modulation in human (Le Bars et al., 1979), are also acquired postnatally. No DNIC can indeed be detected before P21 in the rat (Boucher et al., 1998). In human, the degree of functional connectivity between regions implicated in the descending pain modulatory system, especially between the anterior cingulate cortex and PAG, has been associated with the intensity of noxious-evoked BOLD activity in infants at term (Goksan et al., 2018).

The oxytocinergic system is known to induce significant analgesia in adult, and is of great interest in this context of early life, since it relies on the peptide oxytocin (OT), strongly involved in the neonatal period, facilitating parturition, lactation, and maternal behavior (Poisbeau et al., 2018). In rodents, OT was not detected before P1 in the pituitary and P4 in the hypothalamus (Choy & Watkins, 1979). After birth, OT neurons undergo developmental changes in their electrophysiological properties, especially during the 2nd postnatal week (Widmer et al., 1997). It is considered that the system is almost fully mature at P20 (Choy & Watkins, 1979). In human, at 15 weeks of gestation, OT can be detected in the fetal brain (Schubert et al., 1981) and adult OT cell numbers are detected in the hypothalamus at 26 weeks, even if their morphological features are still immature (Rinne et al., 1962). The evaluation of OT content in umbilical venous samples during labor suggests that the fetus is able to release OT at birth (Chard et al., 1971).

In conclusion, the nociceptive system is established and fully functional at the end of the second trimester of gestation in human. The human fetus can therefore feel pain at this stage as attested by the cortical integration of nociceptive stimulations. However, the fetus and newborn nociceptive system is still undergoing maturation and will remain extremely sensitive and plastic to environmental stressors. If exposed to an excessive number of stressful stimulations, including pain experiences, the nociceptive system will be sensitized in the long-term. This will give rise to abnormal nociceptive sensitivity and altered pain responses, in addition to other symptoms (i.e. emotions, social behaviors). This increasing body of evidence is the reason why it is critical for a better evaluation and treatment of pain at the NICU. Moreover, the comparison of rodent versus human development of the nociceptive system indicates that rodents could be a good model to study the developmental and long-term impact of early pain and stress occurring after birth.

1.2.1 The case of early life pain and its deleterious long-term consequences

According to the first major clinical studies on the topic led by Simons and colleagues in 2003 and Carbajal and colleagues in 2008, infants admitted in NICU after birth were submitted to up to 16 painful procedures per day (Carbajal et al., 2008; Simons et al., 2003). On an average stay at NICU of 14 days, it was estimated that preterm infants were submitted to 50–200 painful procedures. Among them, 2/3 were performed without appropriate pre-emptive analgesia. In the past few years, the reduction in the number of pain procedures and the use of pharmacological and non-pharmacological strategies to reduce stress and pain has become a priority in many NICUs. In 2014, it was considered that the number of painful procedures was reduced to an average of about 11 per neonate and per day (Roofthooft et al., 2014). The EPIPPAIN 2 study concluded that 76% of venipuncture, the most frequent pain procedure in the premature infant, was performed with pre-procedural analgesia (Courtois et al., 2016). Despite this better awareness, wide variations in pain evaluation, sedation, and analgesia are still observed among and between European countries (Carbajal et al., 2015).

Recent imaging studies suggest that an excess of neonatal painful procedures is associated with poorer neurodevelopment and can lead to a reduced head circumference. This fits well with reports showing that preterm infants display a lower white matter volume in the encephalon with abnormal functional connectivity in temporal lobes and thinner frontal/parietal cortical and subcortical grey matter structures (Brummelte et al., 2012; Ranger et al., 2013; Smith et al., 2011). Magnetic resonance imaging studies showed that early pain in very preterm infants lead to impaired thalamic maturation at 32 and 40 weeks postmenstrual age (Duerden et al., 2018), which was also confirmed at 18-month corrected age and extended to the basal ganglia and total brain volume (Schneider et al., 2018). Concerning the long-term effects, an alteration in cortical oscillatory activity has been observed in very preterm infants at school age using the ratio of gamma to alpha power, measured with magnetoencephalogragy (Doesburg et al., 2013).

One of the first publication investigating the effects of early painful insult on later pain sensitivity was the early work by Anna Taddio, who showed that neonatal circumcision significantly increased pain scores of the same children in response to vaccination at 4 or 5 months (Taddio et al., 1997). Consistent with this observation, painful procedures in early life have been associated with increased pain intensity ratings in response to venipuncture at 7.5 years (Valeri et al., 2016), with changes in thermal sensitivity at school age, (Hermann et al., 2006), and with increased pain catastrophizing (Hohmeister et al., 2009). In the UK EPICure cohort, children aged 11 years born very preterm displayed a generalized decreased sensitivity to thermal stimulation but no difference in mechanical sensitivity in the thenar eminence could be seen (Walker et al., 2009). Later, the cohort was followed-up at 18–20 years and these young adults still displayed a generalized hyposensitivity. However, in response to a prolonged noxious cold stimulus, hypersensitivity was observed in females born extremely preterm, but not in males (Walker et al., 2018).

A generalized thermal hypoalgesia was also observed in children who were submitted to neonatal surgery, associated with a decrease in mechanical and thermal sensibility next to the surgical scars (Schmelzle-Lubiecki et al., 2007; Walker et al., 2018). However, a mechanical perceptual sensitization, observed by a positive wind-up ratio, was common in the area adjacent to the scar as well as mechanical allodynia to brush (Walker et al., 2018). When subjected to a later surgery on the same area, children with major surgery during their first 3 months displayed an increased need of anesthetics and analgesics and higher pain scores in COMFORT and VAS measures (Peters et al., 2005). Moreover, a few studies described a higher susceptibility to chronic pain in ancient preterm, including migraine episodes requiring stronger pharmacological treatment (Grunau et al., 2009; Maneyapanda & Venkatasubramanian, 2005; Walker et al., 2018).

In contrast, other studies showed no differences in nociceptive threshold between children born full-term and preterm infants (Goffaux et al., 2008; Vederhus et al., 2012), but still described an alteration of DNIC in preterm infants (Goffaux et al., 2008), which was also observed in the EPICure cohort when former preterm infants were tested at 19–20 years (Walker et al., 2009).

These differences were also observed at the level of cerebral activation, measured by fMRI. In this study, tonic heat stimulation in former preterm infants at the time of adolescence induced activation of the thalamus, anterior cingulate cortex, cerebellum, basal ganglia, and PAG, which was not observed in infants born at term (Hohmeister et al., 2010).

These observations led to an increased awareness of the possible negative effects of neonatal pain and illustrated the need to reduce the number of procedure and use appropriate pain-reducing strategies. However, it is also important to note that neonatal pharmacological pain treatments may also have long-term effects on brain and nociceptive system development (McPherson & Grunau, 2014). In particular, opiates and benzodiazepines are commonly used in preterm infants during the neonatal period. Midazolam exposure during the neonatal period has been shown to impact hippocampal growth, leading to a decrease in its volume and poor cognitive outcome (Duerden et al., 2016). Morphine exposure is also common, and has been associated with decreased cerebellar volume and poor motor and cognitive outcome at an older age (Ferguson et al., 2012; Grunau et al., 2009; Zwicker et al., 2016).

In addition, routing nursing procedures, which could be expected to be painless, can still induce a strong discomfort and stress in the newborn infant. Clustered care are used as a strategy to reduce stress, by grouping routine procedures (changing the diaper, measuring abdominal girth, taking the axillary temperature, cleaning the mouth with gauze and sterile water…) to allow longer periods of rest. However, some studies showed that these routine procedures can induce behavioral responses assimilated to pain. This is for example the case of weighing, which led to a mean score around 4 of 7 on the neonatal infant pain scale (Catelin et al., 2005). Clustered care also induced an increased pain response measured using the brow bulge item of the neonatal facial coding system, and increased NIDCAP stress behavioral cues, although the effect on facial responses seemed lower for clustered care than for a pain procedure (Holsti et al., 2005). When performed just before a painful procedure, clustered care (Holsti et al., 2006) or handling and immobilization (Porter et al., 1998) can also increase the pain response due to that procedure, as demonstrated with an increase in facial response during blood collection.

Altogether, these data show that early life pain is a major burden in newborn infant, in particular in very preterm infants who are more frequently exposed to invasive or skin breaking procedures due to their medical status and care requirements. Clinical observations suggest that neonatal pain strongly affects the development and function of the nociceptive system. Most of the time, a generalized hyposensitivity to normal stimulation, associated with an increased sensitivity to nociceptive stimulation, was observed while allodynia was also detected in some cases. These differences could be explained by differences in the amount of pain associated with the early life procedures, in the type of pain, by the context of exposure to pain, or by the adequacy of associated pain relieving treatment. The timing of nociceptive stimulation may also be an important factor. As animal models allow controlling for type of pain modality, number of procedure, and other environmental factors such as maternal separation, they could be really helpful to better understand the mechanisms underlying the long-term effects observed in human.

1.2.2 Non-nociceptive stress can also impact the development of the nociceptive system

Since the last few years, clinicians aimed to improve clinical care of newborn infants, to decrease pain stimulation as much as possible, and to generalize the safe use of anesthetic and analgesic even for “small” painful procedures in the newborn. However, non-painful ELS is yet another issue, especially in the case of preterm birth where newborn infants can be subjected to various sensory disturbances which could be detrimental for later brain functions: excessive visual or auditory stimulation, numerous uses of antiseptic solutions on the skin, or lack of sensory clues from the parents (as depicted in Figure 2). Despite the maturation state of their sensory system, very preterm infants are able to detect small variations in lights levels in the incubator, as small as 10–50 lux (Zores et al., 2015). Additionally, acoustic stimulation, from a minimum signal-to-noise ratio threshold of 5–10 dBA, can have short-term effects on different physiologic and behavioral responses, including a decrease in the cerebral oxygenation of very preterm infants (Kuhn et al., 2012) and frequent disruptions in their sleep (Kuhn et al., 2013). These environmental variations should be taken into consideration, as it was estimated that very preterm infants are submitted to an average of 3 increases in light level and 7 sound peaks per hour, as well as an average of olfactory and/or trigeminal stimulation of 44–60 times per day, which increased to 146–206 times a day at peak exposure (Kuhn et al., 2011, 2012; Zores et al., 2015).

Besides, newborns kept in incubators are impacted for their early interactions with the parents and this can change for example the amount of eye contact and touch (Korja et al., 2012). Some studies also showed that the duration of parental visit decreased along the days of stay in NICU, but that the holding frequencies of the newborn by the parents increased (Reynolds et al., 2013) or that mothers of preterm infants tend to report longer duration of affectionate holding (Korja et al., 2008). Parent–child interactions are critical for the good development of the child, and the absence or the alteration of parental behavior could induce strong cognitive, social, or autonomic deficits, as it will be described later in this review. However, the architectural design of the NICU with single family room and the implementation of an infant and family-centered care model of care can positively enhance parental presence and involvement (Kuhn et al., 2018; Lester et al., 2016).

1.2.3 How to deal with pain and stress in the newborn? Focus on skin-to-skin contact and its long-term effects

Pain prevention, a careful assessment of pain by trained staff and appropriate pain management are key factors to prevent the long-term consequences of neonatal pain. Apart from the recommendations on the use of antalgics for painful procedures, non-pharmacological interventions are more and more encouraged and have the advantage to be devoid of adverse effects. They include non-nutritive sucking, the use of sweet solution, skin-to-skin-contact or even aquatic physical therapy (Golianu et al., 2007). Child and family-centered strategies are also developed to limit painful and stressful procedures. These include, on top of an appropriate pain management, an unlimited parental access to NICU, psychological support for the parents, postural support for the newborn, breast feeding and lactation support, sleep protection, and skin-to-skin contact (Roue et al., 2017). The later, also called kangaroo mother care (KMC), is of particular interest as it restores the physical interactions between the parents and the newborn. It helps to maintain the physiological state of the infant, by increasing body temperature, regulating respiration, and improving sleep. KMC is an effective strategy to reduce stress in the newborn, decreasing corticosterone levels and cardiac frequency, but also to activate OT release in the parents (Cong et al., 2015), decrease their stress (Cho et al., 2016) and increase the amount of maternal milk (Hurst et al., 1997). Besides these positive effects and other medical benefits (Conde-Agudelo & Diaz-Rossello, 2016), KMC has beneficial effects on neurodevelopment, as seen when used in NICU in high incomes countries (Feldman et al., 2002), or when used continuously in middle income countries as part of the KMC program (Charpak et al., 2017).

Different studies showed that KMC in the newborn is beneficial against pain, as summarized in a recent Cochrane meta-analysis (Johnston et al., 2017). KMC decreased pain behaviors and the autonomic responses to heel stick or intramuscular injections in the preterm and very preterm infant (Castral et al., 2008; Cong et al., 2009; Johnston et al., 2003, 2009; Ludington-Hoe et al., 2005), reduced brain activation measured by fNIRS during venipuncture (Olsson et al., 2016), and the magnitude of noxious-related cortical activity measured with EEG in response to heel-lance (Jones et al., 2020). It is important to note that kangaroo care can also be provided by fathers; however, Johnston and colleagues showed that maternal kangaroo care is more effective to decrease pain during heel lance procedures (Johnston et al., 2011). Maternal voice and touch were proposed as key determinant of the effect of KMC. However, recorded maternal voice was not sufficient to decrease pain behavior in the newborn infant (Johnston et al., 2007). It did not change the autonomic response to pain when associated with maternal contact but was beneficial on recovery time (Johnston et al., 2012). Mother–infant interactions also promote the release of OT in the newborn and which could at least partly explain the analgesic effects of KMC (Poisbeau et al., 2018; Vittner et al., 2018).

If it is now clear that KMC can induce an efficient analgesia in the newborn, the long-term effects of KMC are currently under investigation and suggest that it has positive effects on brain development (Charpak et al., 2017), function of motor circuits (Schneider et al., 2012), stress response and lead to better cognitive and executive functions, as well as a better organized sleep (Feldman et al., 2014). Based on all these results, KMC constitutes a promising strategy to prevent the deleterious consequences of early pain and stress on later pain responses, but this has not been investigated yet and other studies will be needed to answer this question in the next few years.

In conclusion, NICU environment is associated with numerous pain and stressful procedures but also environmental stress, which can be deleterious for newborn neurodevelopment. This early experience of neonatal pain and stress may affect the peripheral processing and central modulation of nociceptive signal, leading to altered behavioral responses to pain at later age. This awareness is now accompanied by a reduction in the number of painful procedures, a better identification of possible stressors in most NICUs and efforts to lessen preterm infant's exposure to them. In particular, efficient and careful use of analgesic or development of non-pharmacological techniques to reduce pain and stress. These techniques have proven some efficacy in the short term, but might still be protective against the deleterious long-term consequences on pain responses.

1.3.1 Consequences of maternal separation on the nociceptive system

Many studies evaluated the long-term effects of MS on pain behavior and identified altered pain responses in response to different type of stimulation. An overview of MS long-term effects on the nociceptive system is presented in Figure 3. Numerous hypotheses have been proposed to explain the development of abnormal visceral and somatic sensitivity in MS animals. The origin seems to be multiple, going from peripheral, spinal and supraspinal anatomical and functional neuronal alterations in the pups, to a change in the composition of the microbiota or in immunological agents which can interact with the nervous system.

1.3.2 Other ELS models and their consequences on the nociceptive system