Widespread nociceptive maps in the human neonatal somatosensory cortex

  1. Laura Jones  Is a corresponding author
  2. Madeleine Verriotis
  3. Robert J Cooper
  4. Maria Pureza Laudiano-Dray
  5. Mohammed Rupawala
  6. Judith Meek
  7. Lorenzo Fabrizi
  8. Maria Fitzgerald  Is a corresponding author
  1. Department of Neuroscience, Physiology & Pharmacology, University College London, United Kingdom
  2. Department of Developmental Neuroscience, University College London, United Kingdom
  3. Elizabeth Garrett Anderson Obstetric Wing, University College London Hospitals, United Kingdom

Abstract

Topographic cortical maps are essential for spatial localisation of sensory stimulation and generation of appropriate task-related motor responses. Somatosensation and nociception are finely mapped and aligned in the adult somatosensory (S1) cortex, but in infancy, when pain behaviour is disorganised and poorly directed, nociceptive maps may be less refined. We compared the topographic pattern of S1 activation following noxious (clinically required heel lance) and innocuous (touch) mechanical stimulation of the same skin region in newborn infants (n = 32) using multioptode functional near-infrared spectroscopy (fNIRS). Within S1 cortex, touch and lance of the heel elicit localised, partially overlapping increases in oxygenated haemoglobin concentration (Δ[HbO]), but while touch activation was restricted to the heel area, lance activation extended into cortical hand regions. The data reveals a widespread cortical nociceptive map in infant S1, consistent with their poorly directed pain behaviour.

Editor's evaluation

This paper is of interest to developmental neuroscientists who study the early stages of cortical maturation and specialization, particularly in the context of somatosensory and pain system development. The authors suggest that, relative to the infant touch somatotopic map, the infant nociceptive map is more widespread and poorly localised, consistent with infants' poorly directed pain behaviour.

https://doi.org/10.7554/eLife.71655.sa0

Introduction

Somatotopically organised cortical maps of activity evoked by innocuous or noxious mechanical stimulation allow us to localise our sense of touch or pain (Penfield and Boldrey, 1937; Harding-Forrester and Feldman, 2018), and may also convey computational advantages in the relay of afferent information to higher brain areas (Thivierge and Marcus, 2007). In adults, overlapping regions are involved in the cortical processing of noxious and innocuous mechanical stimulation (Kenshalo et al., 2000; Lui et al., 2008) and detailed fMRI analysis reveals a fine-grained somatotopy for nociceptive inputs in primary somatosensory (S1) cortex that are aligned with activation maps following innocuous tactile stimuli, suggesting comparable cortical representations for mechanoreceptive and nociceptive signals (Mancini et al., 2012).

A whole-body topographical map of innocuous mechanical stimulation develops in the sensorimotor cortices over the early postnatal period in rats, which represent the human final gestational trimester (Seelke et al., 2012). Distinct representations of the hands and feet can be observed from 31 weeks using fMRI (Dall’Orso et al., 2018), becoming increasingly localised by term age (Allievi et al., 2016). While haemodynamic responses to a clinically required heel lance have been recorded from 28 weeks using functional near-infrared spectroscopy (fNIRS) (Slater et al., 2006) and can be distinguished from innocuous tactile evoked brain activity in EEG recordings from 34 to 35 weeks (Fabrizi et al., 2011), the source of this activity and topographic representation of these two modalities have not been mapped, or their alignment established, in the infant cortex.

Infant pain behaviour is exaggerated and disorganised in newborn rodents and human infants (Fitzgerald, 2005; Fitzgerald, 2015; Cornelissen et al., 2013). Poor spatial tuning of nociceptive reflexes and receptive fields is a feature of the developing somatosensory system, followed by the emergence of adult organisation through activity-dependent refinement of synaptic connections (Beggs et al., 2002; Schouenborg, 2008; Koch and Fitzgerald, 2013). We hypothesised that this developmental process is reflected in ascending nociceptive signals to S1, leading to widespread S1 activation and poor spatial localisation of noxious events in early life.

To test this hypothesis, we used multioptode fNIRS to map nociceptive and innocuous mechanoreceptive activity across the infant sensorimotor cortex. fNIRS is a non-invasive measure of cerebral haemodynamic changes which can be performed at the bedside, while in skin-to-skin holding and in a naturalistic hospital setting during clinically required procedures. Using the temporal and spatial profiles of haemodynamic responses to a noxious heel lance and an innocuous touch of the hand and the heel, we show that haemodynamic activity elicited by noxious and innocuous mechanical stimuli have partially overlapping topographies in the human infant S1 cortex but that the two maps are not aligned. Noxious stimulation of the heel in the newborn evokes more widespread S1 activation than innocuous stimulation, that extends into inferior regions, normally associated with representation of the hand.

Results

Innocuous hand and heel touch evoked activity is somatotopically organised in the newborn infant S1 cortex

We first established the cortical topography of touch activation in newborn infants by mapping the extent of activation in the contralateral somatosensory (S1) cortex following innocuous mechanical stimulation (touch) of the hand and heel. Figure 1a and b shows a significant and localised increase in average concentration of oxygenated haemoglobin (Δ[HbO]) in contralateral optode channels following touch of each body area (n = 11, hand touch; n = 16 heel touch). Touch stimulation of the hand elicited significant increases in seven channels, with a maximum change (0.33 μM at 9.2 s post-stimulus) at the channel corresponding to the FCC5 position of the 10:5 placement system (Oostenveld and Praamstra, 2001), while touch of the heel elicited significant increases in seven channels, with a maximum change (0.35 μM at 15.8 s) at the channel corresponding to the CPP1 10:5 position.

Figure 1 with 2 supplements see all
Significant channel-wise haemodynamic response following innocuous (touch) and noxious (lance) mechanical stimulation of hand and heel.

Average significant change in concentration of oxygenated (Δ[HbO]) (red) and deoxygenated haemoglobin (Δ[Hb]) (blue) during (a) hand touch (n = 11), (b) heel touch (n = 16), and (c) heel lance (n = 11). Channels with significant increases in Δ[HbO] and decreases in Δ[Hb] (i.e. canonical response) during the activation period are shown with solid dark lines, channels with an inverse response only (decrease in Δ[HbO] and increase in Δ[Hb]) are shown with pale solid lines. Note the difference in the scale bar between touch and lance. The equivalent plots for non-significant changes are shown in Figure 1—figure supplement 1 and details of individual channel responses are in Figure 1—source data 1.

Figure 1—source data 1

Significant concentration changes at each channel following innocuous mechanical stimulation (touch) of the heel and the hand and following heel lance.

https://cdn.elifesciences.org/articles/71655/elife-71655-fig1-data1-v2.xlsx

The somatotopically localised increases in Δ[HbO] were accompanied by a widespread decrease in Δ[Hb] over the whole peri-rolandic area (hand: significant decreases in eight channels [peak change: –0.18 μM at 17.6 s]; heel: significant decreases in 14 channels [peak change: –0.21 μM at 11.1 s]). An inverse response (significant decrease in Δ[HbO], significant increase in Δ[Hb], or both) was mostly restricted to channels surrounding the hand and foot areas of the S1, respectively. Individual channel data is shown in Figure 1—source data 1.

Image reconstruction of the channel data (Figure 2a and b) shows that the topography of touch peak activation in the newborn infant S1 is consistent with the known adult S1 topography: the area representing the foot lies in the superomedial postcentral gyrus, while the area for the hand is more lateral and inferior (Penfield and Boldrey, 1937; Harding-Forrester and Feldman, 2018; Willoughby et al., 2020).

Image reconstruction at peak latency of the Δ[HbO] and Δ[Hb] response to an innocuous (touch) and noxious (lance) mechanical stimulation of hand and heel.

Significant changes (compared to baseline) in concentration of oxygenated (Δ[HbO]) (top row) and deoxygenated haemoglobin (Δ[Hb]) (bottom row) following (a) hand touch (n = 11), (b) heel touch (n = 16), and (c) heel lance (n = 11).

Noxious lance of the heel elicits widespread activation extending into inferior S1

We next mapped activation in the contralateral S1 following a noxious, clinically required, lance stimulus to the heel in newborn infants. The average channel response (Figure 1c) and the image reconstruction (Figure 2c) show the significant and widespread increase in Δ[HbO] following the heel lance, which extends beyond the somatotopic area for heel touch to encompass inferior areas of S1, which were associated with touch of the hand.

Heel lance elicited significant increases in Δ[HbO] in eight channels, with a maximum increase (0.96 μM at 14.5 s) at the channel corresponding to the CP3h 10:5 position. Four of the channels with a significant increase in Δ[HbO] following the lance also had a significant increase in Δ[HbO] following touch of the heel. Notably two channels also displayed a significant increase following touch of the hand, demonstrating a topographic overlap with both the hand and heel touch responses. The accompanying decrease in concentration of deoxygenated haemoglobin (Δ[Hb]) was widespread (significant decreases in 11 channels; peak change 1.03 μM at 9.1 s), and an inverse response was found in all channels surrounding those with a canonical response (Figures 1c and 2c).

Newborn infant nociceptive maps are more widespread and not somatotopically aligned with innocuous mechanoreceptive maps

We then compared the responses to heel touch and lance in terms of amplitude, latency and position of peak change, and extent and overlap of the overall areas of activation at peak latency. Heel touch and lance elicited a peak increase in Δ[HbO] at the same latency and location (distance between peaks = 1.65 mm, p = 0.156; difference in peak latency = 2 s [14 vs. 16 s], p = 0.358; Figure 3a–b), however this was significantly larger (maximum Δ[HbO]: 16.06 vs. 5.43 μM, p = 0.001) and wider (FWHM area: 51.41 vs. 43.79 mm2, p = 0.019) following lance stimulation. The overall area of activation was also significantly more widespread following heel lance (Δ[HbO] overall area: 455.27 vs. 259.51 mm², p = 0.009) and only partially overlapped to that following touch (Figure 3a). The increase in Δ[HbO] was significantly larger following lance compared to touch across 50% of the lance-only activation area (Figure 3a – red area, Figure 3c) and across 26% of the overlap area (Figure 3a – orange area, Figure 3c). However, there was no significant difference in Δ[HbO] changes between lance and touch across the touch-only activation area (Figure 3a – pink area). This shows that the significantly larger and wider peak response and more widespread overall area of activation following lance is not due to a greater systemic haemodynamic change following this stimulus.

Figure 3 with 1 supplement see all
Comparison of the peak and area of activation of the Δ[HbO] response to an innocuous (touch) and noxious (lance) mechanical stimulation of the heel.

(a) Overall area of significant changes in concentration of oxygenated haemoglobin (Δ[HbO]) following heel lance (red), heel touch (pink), and both (orange). Black flags demark the location of peak changes and green areas the extent of their full-width half-maximum (FWHM). (b) Statistical position of real differences between heel touch and lance in peak amplitude, FWHM, latency and location, and overall area of activation in respect to non-parametric null distributions obtained with bootstrapping and phase scrambling. (c) Comparison of response magnitude at each node. Results show the t-statistic at significantly different nodes within the areas of activation shown in (a). The equivalent plots forΔ[Hb] are shown in Figure 3—figure supplement 1 and Figure 3—source data 1.

Figure 3—source data 1

Comparison of Δ[HbO] and Δ[Hb] responses to an innocuous (touch) and noxious (lance) mechanical stimulation of the heel.

Parameter estimates for the change in oxygenated haemoglobin (Δ[HbO]) and deoxygenated haemoglobin (Δ[Hb]) following a heel lance and heel touch, and p value from statistical comparison. A two-tailed test was used (α = 0.025) for all parameters except peak location (α = 0.05). The Euclidean distance between the heel touch and heel lance peak locations has been provided rather than x, y, z coordinates of each peak. Linked to Figure 3 and Figure 3—figure supplement 1 .

https://cdn.elifesciences.org/articles/71655/elife-71655-fig3-data1-v2.xlsx

The location, but not latency, of the peak change in Δ[Hb] was significantly different following heel touch and lance (distance between peaks = 51.88 mm, p = 0.041; difference in peak latency = –19.3 s [9.5 vs. 28 s], p = 0.028), which was also significantly larger (maximum Δ[Hb]: –22.54 to –6.62 μM, p < 0.001) and wider (Δ [Hb] FWHM area: 15.31 vs. 4.63 mm2, p < 0.001) following heel lance (Figure 3—figure supplement 1). The overall area of activation was also significantly more widespread following heel lance (Δ[Hb] overall area: 475.24 vs. 95.53 mm², p < 0.001). The key difference between the two patterns of activation is that the heel touch Δ[HbO] response is limited to the areas of the S1/M1 associated with the foot, whereas the lance Δ[HbO] response extends towards other more ventral regions of S1.

Discussion

A widespread nociceptive topographic map in infant S1 that overlaps but is not aligned to the innocuous mechanoreceptive map

Somatosensory maps of cortical activity evoked by a cutaneous tactile or noxious stimulus provide a framework for localising the sense of touch or pain (Treede et al., 1999; Thivierge and Marcus, 2007). The adult primate S1 has a defined somatotopic organisation of tactile and nociceptive cortical receptive fields (Andersson et al., 1997; Kenshalo et al., 2000) including spatially precise cortical maps of Aδ and Aβ afferent fibre input (Chen et al., 2011). Human fMRI studies show that adult somatotopic maps of noxious and non-noxious mechanical stimulation substantially overlap (Lui et al., 2008) and detailed analysis reveals a fine-grained somatotopy for nociceptive inputs in S1 cortex that are highly aligned with maps of innocuous tactile stimuli, suggesting comparable cortical representations for mechanoreceptive and nociceptive signals (Mancini et al., 2012). Here, we have shown that this comparable representation is not present in the newborn infant S1 cortex.

Noxious mechanical stimulation evokes a larger peak increase in concentration of oxygenated haemoglobin (Δ[HbO]) and decrease in concentration of deoxygenated haemoglobin (Δ[Hb]) compared to innocuous stimulation of the same body area at comparable location and latency as reported elsewhere (Bartocci et al., 2006; Slater et al., 2006; Verriotis et al., 2016b). The fact that the noxious activation is greater than the touch evoked activity is presumably due to hyperaemia associated with greater depolarisation and spike activity within the activated areas. However, it does not explain the differing topography of the overall area of activation reported here. Although a lance, being a more intense stimulus, could be thought to cause more widespread cortical activation throughout the cortex, here the response to the lance is only significantly larger than touch in 50% of lance-only area of activation, 26% of the lance and touch overlap area, and nowhere in the touch-only activation area. Our results instead suggest that the response to a noxious and innocuous mechanical stimulus is not spatially aligned. Within S1 itself, the infant has a distinct somatotopic map for touch, similar to that described in adults, with the area representing the foot lying in the superomedial postcentral gyrus, and the area for the hand located more inferiorly (Penfield and Boldrey, 1937; Blake et al., 2002; Akselrod et al., 2017), consistent with previous reports in newborn infants (Dall’Orso et al., 2018). Noxious heel lance, on the other hand, evokes widespread activity within S1, peaking in a similar area of the superomedial postcentral gyrus as touch activity, but extending to the hand representation area. All stimuli elicited a significant response in the control channel which may suggest that the responses extended more posteriorly to S1, nevertheless, our data shows that the S1 somatotopic nociceptive map is not as precise as the touch map in the newborn.

It should be noted that fNIRS is an indirect measure of neuronal activation and may be confounded by systemic vascular changes such as increases in blood pressure, which may be more likely following a noxious vs. innocuous stimulus. However, we removed global changes common across channels using principal component analysis (PCA), and our results point to misaligned noxious and innocuous activity, with widespread noxious activity across the S1 specifically, but not the whole cortex as one would expect following systemic changes. Several channels/nodes did not show a significant response following the heel lance compared to baseline, and the heel lance did not have a greater response compared to heel touch at every node. Together, these findings suggest that the greater magnitude of the heel lance response is not due to systemic changes elicited by the noxious stimulus. However, as fNIRS (as with fMRI) is a measure of the vascular response following neuronal activity, the precise degree of localisation of either neuronal response is limited.

Measuring the cortical haemodynamic response to innocuous and noxious mechanical stimulation in neonates fNIRS is ideally suited to a study of this kind as recording and sensory stimulation, including clinically required heel lance, can be performed at the infant cotside (Bartocci et al., 2006; Slater et al., 2010; Kashou et al., 2017; Verriotis et al., 2016b). Other methods of measuring this either do not provide sufficient spatial information and source localisation, such as EEG recording of nociceptive-related ERPs (Fabrizi et al., 2011; Jones et al., 2018) or are limited by the use of experimental ‘pinprick’ stimulators, that for ethical reasons are not actually painful, such as in fMRI studies (Goksan et al., 2015).

The change in the Δ[HbO] and Δ[Hb] following sensory stimulation is an indirect measure of neural activity: simultaneous vertex EEG and fNIRS recordings over S1 show that haemodynamic and neural responses are related in magnitude (Verriotis et al., 2016b). Following all stimuli, and consistent with the mature canonical response, channels showing a significant increase in Δ[HbO] also had a smaller decrease in Δ[Hb]. Regional overperfusion following neuronal activation, beyond that required by metabolic demands, means that less Hb is removed from the region compared to the oversupply of HbO. However, the decrease in Δ[Hb] was more widespread (but smaller in magnitude) compared to the localised increase in Δ[HbO]. This type of response, not previously reported in infants (de Roever et al., 2018), suggests that more blood is leaving the region (removing Hb) compared to the incoming supply (no significant change in Δ[HbO] in peripheral channels) due to immature regulation of cerebral blood flow (CBF). There are multiple mechanisms by which blood vessels dilate and CBF increases following neural activation, including arterial CO2 and O2 concentrations, which relax/contract the smooth muscle cells of cerebral arteries and arterioles (Kety and Schmidt, 1948), and astrocyte and pericyte activity which controls vessel diameter and the propagation of vasodilation along the vascular tree (Takano et al., 2006; Cai et al., 2018), many of which are still developing in the newborn (Pryds and Greisen, 1989; Binmöller and Müller, 1992; Fujimoto, 1995) leading to rapid changes in CBF over the first postnatal days as cerebral circulation adapts (Meek et al., 1998).

The infants in this study were held skin to skin, swaddled in their mother’s arms in a naturalistic setting, which is a major advantage of fNIRS recording over fMRI for human developmental studies of brain function. Video recording and investigator scoring confirmed that while some infant movement and maternal touching did take place and that some babies did move following the lance, these movements spanned different body parts and latencies such that any associated cortical response would be cancelled during the averaging process. Indeed, comparison of the time-series following heel touch with and without removing trials with known movement showed no significant difference (Figure 1—figure supplement 2). Furthermore, 33–50% of babies did grimace for up to 7 s following the lance, but if these facial movements mediated the response following the lance, this would have prolonged the peak or duration of the change in [HbO], while in fact the latency and time course of the response to both stimuli was the same (see Figure 1).

Differential development of somatosensory and nociceptive topographic maps

A whole-body topographical map of innocuous mechanical stimulation develops in the sensorimotor cortices over the early postnatal period in rats (Seelke et al., 2012), which represents the human final gestational trimester. In humans, distinct representations of the hands and feet can be observed from 31 weeks’ gestation, using fMRI (Dall’Orso et al., 2018), and from 28 weeks using neural activity recorded from the scalp (Donadio et al., 2018; Whitehead et al., 2018; Whitehead et al., 2019). The response to innocuous mechanical stimulation was more localised in S1 than the wider and less refined topographical map of noxious mechanical stimulation, suggesting a slower maturation of the S1 circuitry involved in nociceptive processing compared to touch processing in the infant brain.

In rodents, at every level of the developing somatosensory central nervous system, tactile processing matures before nociceptive processing (Fitzgerald, 2005; Koch and Fitzgerald, 2013; Chang et al., 2016; Chang et al., 2020; Verriotis et al., 2016a) consistent with a delayed refinement of a cortical nociceptive map. Widespread nociceptive cortical maps are consistent with infant pain behaviour, characterised by exaggerated and disorganised nociceptive reflexes in both rodent pups and human neonates (Fitzgerald, 2005; Fitzgerald, 2015), and which can fail to remove a body part from the source of pain (Waldenström et al., 2003). Nociceptive reflexes following noxious heel lance are larger in magnitude and significantly more prolonged in human infants compared to adults (Cornelissen et al., 2013) and have widespread cutaneous receptive fields that encompass the whole lower limb (Andrews and Fitzgerald, 1994). This lack of organisation could be reflected in the ascending spinothalamic and thalamocortical projections, delaying the maturation of S1 cortical nociceptive maps in the newborn. Topographic maps are established and aligned via multiple mechanisms, including molecular cues, spontaneous or sensory-dependent remodelling, and refinement. Initially, somatosensory maps are diffuse and overlapping, but in the rodent somatosensory cortex, excitatory thalamocortical afferents undergo activity-dependent refinement to sharpen these maps (Iwasato et al., 1997). Equally important is the maturation of inhibitory interneuron sensory maps which, in contrast, expand over development in an experience-dependent manner (Quast et al., 2017). Slow developmental broadening of an inhibitory nociceptive network may explain the widespread nociceptive map in S1 and also the greater amplitude of EEG noxious responses in infants compared to adults (Fabrizi et al., 2016).

Pain and the developing S1 cortex

This study highlights the importance of understanding the development of touch and pain processing in the human infant brain. The widespread S1 nociceptive topography discovered here implies that the infant S1 cortex would be unable to accurately localise noxious events and may lack the computational ability to reliably send noxious information to higher brain centres (Thivierge and Marcus, 2007; Harding-Forrester and Feldman, 2018). Heel lance is one of many skin-breaking procedures commonly performed in neonatal hospital care (Laudiano-Dray et al., 2020) and this study reveals the extent of cortical activation that follows just one such noxious procedure in the newborn. This contrasts with innocuous mechanical stimulation, such as touch, which activates a spatially restricted and somatotopically defined cortical area. Increasing evidence that repeated noxious experiences have adverse effects upon the developing brain (Ranger and Grunau, 2014; Duerden et al., 2018) underlines the importance of these results and the need for a better understanding of the mechanisms underlying the maturation of cortical nociceptive topographic maps.

Materials and methods

Participants

Thirty-two infants (35–42 gestational weeks at birth, 0–7 days of age, 12 female; Table 1) were recruited from the postnatal, special care, and high dependency wards within the neonatal unit at University College London Hospital. Infants received either (1) innocuous mechanical stimulation (touch) of the heel, (2) innocuous mechanical stimulation of the hand, or (3) a noxious mechanical stimulation (clinically required lance) of the heel. Six infants received touch stimulation of both the heel and hand. This sample size is sufficient as similar high impact works using single trial noxious stimulation or multiple mechanical stimulations have yielded significant results with group sample sizes of 5–15 (Bartocci et al., 2006) and 10–15 (Arichi et al., 2012), respectively. Ethical approval for this study was given by the NHS Health Research Authority (London – Surrey Borders) and the study conformed to the standards set by the Declaration of Helsinki. Informed written parental consent was obtained before each study.

Table 1
Infant demographics.

Demographic information about the subjects that received tactile and noxious stimuli of heel and hand.

Heel lanceHeel touchHand touchp
N111611
GA (weeks+days)39+2 (35+2–41+5)39+4 (35–42+3)39+2 (37+5–41+3)0.287
PNA (days)4 (0–7)3 (0–6)3 (0–4)0.115
Females4 (36%)6 (38%)5 (45%)0.889
Birth weight (g)3134 (2220–4072)3250 (2360–4080)3300 (2450–3754)0.774
Caesarean deliveries2 (18%)8 (50%)3 (27%)0.196
Head circumference (cm)34 (32–35.5)34.25 (31–37)34 (32.5–36)0.900
  1. Values represent median and range or proportion. GA = gestational age (weeks from the first day of the mother’s last menstrual cycle to birth); PNA = postnatal age (days since birth). No significant difference was found in any demographic parameter across the three groups (one-way ANOVA results in the last column).

Experimental design

Request a detailed protocol

Brain activity (fNIRS) was recorded following a noxious (clinically required heel lance) or innocuous (touch) mechanical stimulation of the limbs at the bedside in the neonatal unit.

fNIRS recording

Request a detailed protocol

Infants wore a 21-channel array consisting of eight sources and eight detectors with inter-optode distances of 2.5–4 cm. The array was secured over the pericentral area of the scalp on the side contralateral to the stimulation with a custom-designed textile cap (EasyCap, Figure 4a). The infants’ head circumference, ear to ear lateral semi-circumference, and nasion to inion distance were measured and the cap was placed on the head by aligning specific 10/5 locations (Cz, T7). This optode arrangement provided sensitivity coverage for the somatomotor cortex contralateral to the stimulation site and of the medial part on the ipsilateral side (Figure 4d). One source-detector pair was placed at a more ventral posterior location of the scalp (P7 of the international 10/5 positioning system) (Figure 4b). This channel was sensitive to the posterior temporal lobe and worked as a control channel (Figure 4b–d). A continuous wave recording system was used with two wavelengths of source light at 780 and 850 nm and a sampling rate of 10 Hz to measure changes in oxy- and deoxy-haemoglobin concentration (Gowerlabs NTS fNIRS system).

Optode locations and sensitivity map.

(a) Typical functional near-infrared spectroscopy (fNIRS) setup on a neonate of 35+2 weeks’ gestational age (GA), 7 days’ postnatal age (PNA). (b) Channel locations according to the international 10–5 placement system (Oostenveld and Praamstra, 2001). Names of four diagonal channels (channels overlaying Cz and CPz) are not included as these were omitted from the analysis due to low signal-to-noise ratio. (c) Locations of the fNIRS sources, detectors and resulting measurement channels registered to a 39-week anatomical atlas. The central sulcus has been highlighted with a black line and the location of S1 is the gyrus posterior to this sulcus. (d) Normalised fNIRS sensitivity illustrating the spatial coverage provided by the channel arrangement in panel (c). This sensitivity map was calculated using the photon measurement density functions derived from the TOAST++ light transport modelling package.

Noxious mechanical stimulation

Request a detailed protocol

The noxious stimulus was a clinically required heel lance for blood sampling. Blade release was time-locked to the NIRS recording using an accelerometer attached to the lancet (Worley et al., 2012). The lancet was placed against the heel for at least 30 s prior to the release of the blade. This was to obtain a baseline period free from other stimulation. The heel was then squeezed 30 s after the release of the blade, again to ensure a post-stimulus period free from other stimuli. All lances were performed by the same trained nurse (MPL-D) using a disposable lancet, and standard hospital practice was followed at all times.

All infants were prone against their mother’s chest. The mother, who was inclined on a chair or bed, was instructed to avoid moving or stimulating the infant during the 1 min before and after the release of the lance.

Innocuous mechanical stimulation

Request a detailed protocol

Innocuous mechanical stimulation was delivered by light touch on the lateral edge of the infants’ palms and/or heels using a hand-held tendon hammer (ADInstruments). A piezo-electric sensor mounted on the hammer head provided a synchronising signal to the NIRS recording. A train of up to 15 touches (average = 11.1) was delivered to each limb with a variable inter-stimulus interval of 35–60 s. This resulted in an average of 11.5 heel touches (range = 7–15) and 10.6 hand touches (range = 6–13) per infant. Stimulus repetition did not cause habituation of the response to touch (no significant difference in the average time-series between the first and last 50% of the trials).

Infant movement

Request a detailed protocol

To limit cortical activation not related to the investigated sensory stimuli, body movements were minimised during the lance procedure, as infants were swaddled (wrapped securely in clothes/blankets) against the mother’s chest and the research nurse was holding the exposed foot throughout the period before and after the stimulus.

Infant movements or maternal handling were recorded on video, which was synchronised with the NIRS recording using an LED light within the frame that was activated by the release of the lance (Worley et al., 2012) or annotated at the time of study using a stopwatch. For lance trials, movements or maternal handling were separated into body parts and coded per second as either 0 (not present) or 1 (present) for the 30 s post-stimulus (Table 2). Following the lance, two babies did not move, five babies made small movements (including small or brief grimace, head nod, twitch, small hand movement), four babies made larger movements (including arms, large or prolonged grimace, nod of head), and two babies received tactile stimulation from the mother (including positioning the head, stroking the face). Following touch, body movements were present in an average of 2.9 trials in 8/16 infants (heel) and 1.4 trials in 11/11 infants (hand). Although every effort was made to minimise trials affected by body movements, these were sometimes unavoidable following a single lance trial and therefore were included for both lance and touch analysis. However, this is unlikely to have affected our results because body movements were not systematically associated to the stimulus of interest as they spanned different body parts and latencies such that any associated cortical response would be cancelled during the averaging process. Indeed, comparison of the time-series following heel and hand touch with and without removing trials with known movement showed no significant difference (Figure 1—figure supplement 2).

Table 2
Infant movements.

The number of infants who displayed movements or received tactile stimulation from their mother each second in the 30 s following lance.

Number of infant and maternal movements
Post-lance (s)HandHeadFaceFootArmMother touching faceMother touching head
11140211
21161211
31061211
41060211
51160211
61160210
70140110
80130110
90120110
100010110
110010110
120010110
130010110
140010110
150010010
160010010
170000010
180000010
190000010
200000010
210000010
220000010
230000010
240000010
250000010
260000010
270000010
280000010
290000010
300000010
  1. Each movement or stimulation was scored as present (1) or not present (0) per infant, and the value in each cell represents the total number of infants (out of 11) for whom each movement or stimulation was observed.

Data pre-processing

Request a detailed protocol

All data were pre-processed in Homer2 (Huppert and Boas, 2005). The four channels crossing over the midline had consistent poor signal quality (light intensity <0.01, SNR <2) and were not considered further (Figure 4b and c). Recordings from 1 to 8 (average: 0.74) channels in 11 (29%) trials were also removed for lance and touch stimuli. Data were then converted into optical density, motion artefacts were detected (change in amplitude >0.7 and/or change in standard deviation >15 over a 1 s time period) and then corrected using wavelet filtering (Molavi and Dumont, 2012). Instrumental drift and cardiac artefact were removed with a 0.01–0.5 Hz bandpass filter. Optical density changes recorded from all channels (likely related to stimulus-dependent systemic physiological changes) were removed using PCA [Kozberg and Hillman, 2016; Tachtsidis and Scholkmann, 2016]; one component removed for every subject. Finally, data were converted into changes in oxy- and deoxy-haemoglobin concentration (Δ[HbO] and Δ[Hb]) using the modified Beer–Lambert law (Delpy et al., 1988) with a differential path length factor of 4.39 (Wyatt et al., 1990). The continuous signal was then epoched from –5 to 30 s around the noxious and somatosensory stimuli. Somatosensory stimuli were averaged for each subject.

Signal to noise

Request a detailed protocol

The signal-to-noise ratio (SNR) for lance and for touch was calculated. Despite the peak signal for lance being larger, the estimated SNR of the peak was lower, because more touch trials were averaged in this study: PeakLance*NLancePeakTouch*NTouch= ratio of lance to touch SNR; 0.96*110.35*184= 0.67.

Channel-wise data analysis

Request a detailed protocol

Pre-processed data were then averaged across trials for each subject (except lance which only had one trial) and analysed using custom MATLAB scripts (Mathworks; version 16b). For each channel, significant changes in Δ[HbO] and Δ[Hb] were identified with a two-tailed t-test (α = 0.01) comparing each time point post-stimulus against the baseline. This baseline distribution was calculated as the mean of the individual baselines (−5 to 0 s before stimulus) according to:

1Si=1Sxi N1Si=1Sμi,1S2i=1Sσi2.

where S is the number of subjects and xi Nμi,σi2 is the baseline for subject i.

Bonferroni correction was used for multiple comparisons (17 channels × 300 samples = 5100 comparisons). Only changes (increases or decreases) which were continuously significant for at least 1 s (10% of the length of the post-lance period) were retained (Guthrie and Buchwald, 1991).

Data analysis in image space

Image reconstruction

Request a detailed protocol

The channel-wise data was used to create functional images using a cortically constrained linear reconstruction approach at each time point to obtain image time-series for each subject. The fNIRS array was registered to a 39-week gestational age anatomical mesh model with 784,391 nodes (Brigadoi et al., 2014) using tools from the AtlasViewer package (Aasted et al., 2015). Images were reconstructed using the DOT-HUB toolbox (Cooper et al., 2021) and the TOAST++ light transport modelling package (Schweiger and Arridge, 2014a, Schweiger and Arridge, 2014b), with zeroth-order Tikhonov regularisation with a regularisation hyperparameter of 0.1.

Assessment of peak changes in Δ[HbO] and Δ[Hb]

Request a detailed protocol

Image space analysis was conducted separately from channel space. We first assessed if hand and foot touch and foot lance evoked a significant response in Δ[HbO] and Δ[Hb], by comparing their peak change to baseline. To do that, we averaged the image time-series and found the latency at which the maximum change occurred anywhere across the cortex. For each node, significant changes at that latency in Δ[HbO] and Δ[Hb] were assessed with a two-tailed t-test (α = 0.01) comparing the 5 s window around the identified peak latency against the 5 s baseline. As in channel-wise analysis peak signal and baseline distributions were estimated by concatenating the individual baselines (−5 to 0 s pre-stimulus) and peak windows (−2.5 to 2.5 s around peak latency). Bonferroni correction was used for multiple comparisons (784,391 nodes = 784,391 comparisons). To display these results, we: (1) reconstructed an image using the average channel-wise data within the 5 s window around the peak latency (averaged in time) for hand touch, heel touch, and lance and (2) masked this image according to the result of the statistical test above.

Comparison of peak changes in Δ[HbO] and Δ[Hb] between heel lance and touch

Request a detailed protocol

Next, we compared the significant changes in Δ[HbO] and Δ[Hb] between the heel lance and touch conditions in two ways: (1) comparing the magnitude of the response at peak latency at every node using a Student’s t-test, (2) comparing the characteristics of the peak response (overall area of activation and latency, location, and spread of the peak change in Δ[HbO] and Δ[Hb]) using a non-parametric test.

Overall area of activation was defined by starting at the node with the peak change and continually expanding to include neighbouring nodes that (1) were connected by three face edges and (2) had a significant change from baseline. Difference in peak location was the Euclidean distance between the peaks. Spread of the peak was defined as the part of the overall area of activation around the peak where changes in Δ[HbO] (or Δ[Hb]) were at least half of the peak change (full-width half-maximum, FWHM).

The non-parametric null distribution was derived by calculating these differences between randomly selected sets of surrogate image time-series (bootstrapping on surrogate data). We here describe how we obtained surrogate image time-series and then how we conducted bootstrapping.

Each individual recording (i.e image time-series) can be considered as the linear sum of a signal of interest (i.e. the response to the stimulus) and a stationary random noise component. The assumption is that the signal is the same in each recording while the noise changes (van Drongelen, 2007). Therefore, if we were to conduct another recording on another subject, the new data would be the linear sum of the same signal that we find in the original data but different random noise. Creating surrogate data consists in generating new random noise to add to the signal estimated from our data. To do this we: (1) estimate the signal by averaging across individual recordings (i.e. subjects) in response to the same stimulus modality; (2) isolate the noise in our data by subtracting this estimate from each recording; (3) phase-randomise each noise time-series. Phase-randomisation is applied independently to each node time-series in the frequency domain. This means that the phase component of the complex-valued signal is rotated at each frequency by an independent random variable chosen from the uniformly distributed range of 0 and 2π (Theiler et al., 1992). At the end of this process, we have a new set of surrogate noise time-series.

To generate the full non-parametric null distribution against which to compare our data, we used bootstrapping. To estimate each sample of the null distribution, we calculated the differences in overall area of activation and peak amplitude, latency, position, and FWHM between two random sets of surrogate data without any systematic difference. To create the random sets, we: (1) pooled together all the newly obtained surrogate noise time-series; (2) added the grand average (across lance and touch) signal (as we do not want systematic differences between sets to estimate a null distribution); (3) randomly split (with repetition) these surrogate data into two sets. We repeated this 1000 times in order to obtain the full non-parametric null distribution (bootstrapping). An experimental difference outside the 95% confidence interval was considered significant (p < 0.05).

Data availability

All raw data files are open access and are available to download from Figshare (https://doi.org/10.6084/m9.figshare.13252388.v3).

The following data sets were generated

References

  1. Book
    1. Harding-Forrester S
    2. Feldman DE
    (2018) Somatosensory maps
    In: Harding-Forrester S, editors. Handbook of Clinical Neurology. Elsevier. pp. 73–102.
    https://doi.org/10.1016/B978-0-444-63622-5.00004-8
  2. Conference
    1. Huppert T
    2. Boas DA
    (2005)
    HomER: Hemodynamic evoked response NIRS data analysis GUI
    Martinos Cent Biomed Imaging.
  3. Book
    1. Kozberg M
    2. Hillman E
    (2016) Chapter 10 - Neurovascular coupling and energy metabolism in the developing brain. In: Progress in Brain Research
    In: Masamoto K, Hirase H, Yamada K, editors. New Horizons in Neurovascular Coupling: A Bridge Between Brain Circulation and Neural Plasticity. Elsevier. pp. 213–242.
    https://doi.org/10.1016/bs.pbr.2016.02.002
    1. Pryds O
    2. Greisen G
    (1989)
    Effect of PaCO2 and Haemoglobin Concentration on Day to Day Variation of CBF in Preterm Neonates
    Acta Paediatrica 78:33–36.
    1. van Drongelen W
    (2007)
    Signal Processing for Neuroscientists
    55–70, Signal Averaging, Signal Processing for Neuroscientists, Academic Press, 10.1016/B978-012370867-0/50004-8.

Decision letter

  1. Chris I Baker
    Senior and Reviewing Editor; National Institute of Mental Health, National Institutes of Health, United States

Our editorial process produces two outputs: i) public reviews designed to be posted alongside the preprint for the benefit of readers; ii) feedback on the manuscript for the authors, including requests for revisions, shown below. We also include an acceptance summary that explains what the editors found interesting or important about the work.

Decision letter after peer review:

Thank you for submitting your article "Widespread nociceptive maps in the human neonatal somatosensory cortex" for consideration by eLife. Your article has been reviewed by 2 peer reviewers, and the evaluation has been overseen by Chris Baker as Reviewing and Senior Editor. The reviewers have opted to remain anonymous.

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

Essential revisions:

Please fully address the concerns and suggestions of the reviewers as laid out in their reviews below. Specifically:

1) Address the statistical concerns raised by both reviewers with new analyses. In particular, this will require including direct statistical testing of the two maps (R1) as well as addressing concerns around circularity (R1) and the different numbers of trials for each condition (R2).

2) Reconsider use of the term "widespread" to describe responses to the painful stimuli (R1 and R2). Terms such as "mislocalized" or "disorganized" may be more appropriate.

3) Treat motion equivalently for the two conditions in the analyses (R2).

Reviewer #1 (Recommendations for the authors):

1. Differences in regions of activations constituting touch maps and nociceptive maps are never actually statistically tested:

See the public review for an outline of the issue. The suggested solution is to directly compare the noxious and touch evoked activations and demonstrate statistically significant differences between them.

2. Nociceptive maps are equally as widespread as touch maps:

See the public review for an outline of the issue. The distinction between "mislocalisation" versus "widespread" needs to be clarified, as the claim that nociceptive maps are widespread is ubiquitous and currently not sufficiently contextualised relative to the touch maps. Or at least the apparent tension between this core claim and the lack of difference in area of activation for the noxious and non-noxious stimulus needs explicit discussion. Of course, the "mislocalisation" claim would have to be backed up by a demonstration of statistically significant differences between the maps, which is currently not provided (see comment 1 above).

3. The analyses that produce areas of activation in source space (image space) are circular:

See the public review for an outline of the issue. The core message of this manuscript is that infants' nociceptive maps are more poorly localised than touch maps. This could be demonstrated in sensor space without any need to do source reconstruction. As outlined in comment 1 above, a direct statistical comparison of noxious evoked and touch evoked activity in sensor space would demonstrate the core message the authors wish to convey, and does not need source reconstruction meaning the circular analysis can be omitted altogether. And as mentioned above, all statistically significant results found in source space are not novel, so very little would be lost in keeping the analysis and results in sensor space and simple establishing statistically significant differences between the nociceptive and touch maps.

4. Limitations of the fNRIS technique and of the link between properties of nociceptive maps (widespread/mislocalisation) and poorly directed behaviour are not adequately discussed:

See the public review for an outline of the issue. This major limitation of the fNIRS technique requires much more serious and explicit discussion.

5. Linking comments 1, 2, and 4 to the initial hypothesis:

The hypothesis stated in the introduction is as follows: "We hypothesised that this developmental process is reflected in ascending nociceptive signals to SI, leading to widespread cortical activation and poor spatial localisation of noxious events in early life". The assessment of the poor localisation and widespread activation of noxious evoked activity is assessed and explicitly interpreted relative to touch evoked activity: "Noxious stimulation of the heel in the newborn evokes a more widespread cortical activation than innocuous stimulation, that extends into inferior regions of S1, normally associated with representation of the hand.".

However, using statistical tests to directly compare the noxious evoked activity to the touch evoked activity, there was neither a statistically significant difference between the noxious and touch peak localisations (i.e. peak locations do not differ) nor their extent of cortical activations (i.e. areas of activations do not differ). This subset of results seem to directly go against the initial hypothesis and interpreted conclusions.

A much clearer link between the tests run, results provided, and hypothesis put forward needs to be outlined, as it is currently very unclear why some results (e.g. activations assessed relative to baseline) should be considered supportive of the initial hypothesis but not indicative of fNIRS-related haemodynamic confounds (e.g. non-neural cardiovascular response signals such as tissue-damage evoked blood pressure changes), while other results (e.g. direct statistical comparisons between noxious and touch evoked activations) that seem to contradict the initial hypothesis are not considered as such by the authors.

Reviewer #2 (Recommendations for the authors):

In order to address the issue of repetition suppression in the touch condition, it would be of interest to see either an analysis of change in the touch responses across trials, or a comparison between touch and nociceptive trials based on only the first touch trial.

There is some small irregularity in the reporting of motion in the pain condition. The N for this condition in Table 1 is 11, but when reporting motion the authors state that 2 participants didn't move, 6 made small movements, and 4 made large movements.

Perhaps I am misreading the analysis procedure, but if the authors really are averaging responses to the touch/pain conditions and using a single, mean value for each time point in statistical tests (compared in a one-sample t-test to the baseline distribution?), then I do believe these analyses should be adjusted to retain variability across participants in response to the stimuli of interest.

I was also surprised by the information on Bonferroni corrections. The authors were really multiplying their p values by thousands (channel-wise analyses) or hundreds of thousands (image-based analyses) of comparisons and arriving at corrected p values less than p = .001, in tests with 11 infants and a single trial per infant? I would not expect this sort of effect size in infant fNIRS research. Am I misunderstanding how a Bonferroni correction would be applied here?

[Editors’ note: further revisions were suggested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "Widespread nociceptive maps in the human neonatal somatosensory cortex" for further consideration by eLife. Your revised article has been evaluated by Chris Baker (Senior Editor).

The manuscript has been improved but the reviewers raised some remaining concerns and made some good suggestions. There are some remaining issues that need to be addressed, as outlined below:

– Please show the S1 ROI as suggested by R1

– In the revision, you show that there is no discernible impact of motion on the heel touch trials in response to a concern raised by R2. However, this does not necessarily solve the issue raised. The suggestion to treat both trial types equally (as raised in the first round of review) should be followed.

– Both reviewers and myself had a hard time following exactly how the image space analyses are conducted and whether the concern about circularity has been addressed or not. It sounds as if different nodes or timepoints are being selected on the basis of the responses observed, which still has the potential to introduce a selection bias (i.e. circularity). Please clarify and expand the text in describing the analyses.

– As noted by R2, the text at the top of page 5 is confusing. More generally, the descriptions in the text tend to be quite dense and complicated. Please revise the section noted and consider simplifying the text throughout to make the descriptions of the analyses and the corresponding conclusions easier to follow.

– As suggested by R2, please consider providing the preprocessed data in addition to the raw files.

Reviewer #1 (Recommendations for the authors):

This manuscript has improved since initial submission, and the authors have appropriately addressed the points in my previous review. I have one final suggestion that I think would greatly help the reader and improve the manuscript further.

It would be very helpful to see an outline of the S1 area, as defined and used in this study. The brain surface onto which the activity has been projected during image reconstruction is very smooth (see Figures2-4); it is not easy to make out with any confidence where any of the key landmarks are e.g. central sulcus, precentral sulcus, postcentral sulcus. In all of your source space maps, only two brain orientations are used: view from above and view from behind. I would suggest adding an additional supplementary figure of these two brain orientations with the S1 area clearly delineated. Or alternatively, putting in a fourth panel into Figure 4 to delineate the area, as Figure 4 already contains quite a bit of localisation/orientation information (regarding optode setup), so seems a sensible place for defining the ROI of interest i.e. area S1.

Having this area clearly defined and detailing how it was defined (e.g. anatomically or functionally defined; manually or automatically outlined; use of an established functional or structural atlas, etc) will really help readers interpret the activity maps more accurately.

Reviewer #2 (Recommendations for the authors):

In general, the authors made clear and reasonable edits to their manuscript that addressed many of the concerns with the original paper. I do, however, think there are a few points that still need attention:

– One of my original concerns was that motion-contaminated trials were dropped in the heel touch condition but not the heel lance condition. In their revision, the authors compare heel touch data with and without motion-contaminated trials and find no significant differences, but I'm not sure this is the right way to deal with this concern; instead I still think motion should be treated the same across the two conditions. Large artifacts introduced by motion could fail to create systematic differences in heel touch data with and without motion, but still account for apparent differences between lance and touch data. If motion artifacts aren't a significant concern to include in the lance condition because they average one another out, that approach ought to be fine for the heel touch condition as well. If artifacts make the heel touch data too noisy, despite the greater number of trials, then their presence in the single-trial lance data ought to be treated as a serious source of contamination in the lance data and excluded from both.

– The authors dealt with Reviewer 1's concern about circularity in the image space analysis by using the image space peak, rather than the source space peak, to select the time period for statistical testing. If anything, I think this makes the circularity problem worse. Selecting a subset of the data based on one feature (proximity to peak concentration values), and then using those data in a subsequent statistical test to characterize the functional properties of the cortical location in question is the issue; the data used to select need to be independent of the data used to test. This seems like a tricky problem to solve here, but maybe a leave-one-subject-out procedure, in which the authors iterate through leaving out one subject when selecting the peak, extract response data from that subject, and then move onto the next iteration in order to obtain their test data would help? (Though I agree with Reviewer 1's original point that there's nothing wrong with just analyzing the data in source space, and omitting the image space analysis entirely).

– I was confused by the new text at the top of page 5, which (as far as I can tell) states that at every location where there was a significant effect of lance > baseline, there was also a significant lance > heel touch effect; and conversely that at every location where touch > baseline, there was also a touch > lance effect. Given that there was overlap in the effects of lance > baseline and touch > baseline, how could this possibly be true? Maybe this is not what the authors were trying to say, and they only meant that there are some regions in which there is a significant lance> touch effect and others in which there is a significant touch>lance effect (as shown in Figure 3c). If that's the case, I would suggest rewording this new text, as it was difficult to parse.

– I went to check the data files, and it seems that the authors only provide the raw.nirs files and demographic characteristics in the repository. Reconstructing their preprocessing stream would be extremely time consuming; I suggest they should also provide the data in the form used for statistical tests.

https://doi.org/10.7554/eLife.71655.sa1

Author response

Essential revisions:

Please fully address the concerns and suggestions of the reviewers as laid out in their reviews below. Specifically:

1) Address the statistical concerns raised by both reviewers with new analyses. In particular, this will require including direct statistical testing of the two maps (R1) as well as addressing concerns around circularity (R1) and the different numbers of trials for each condition (R2).

We have now added a direct comparison of the HbO response amplitude between lance and heel touch at each node. The results have been added to Figure 3 and are discussed in the relevant methods and Results sections.

We have removed the latency restriction from the source-space analysis (i.e. using the peak latency identified from the channel-space) and updated the results appropriately; the results have not changed as a consequence of this.

We have now included a brief description of analysis which demonstrates that there was no significant habituation following the repeated heel touch stimulation. We have provided a figure of the resulting time-series in our response to reviewers which can be added as a supplementary image if required.

2) Reconsider use of the term "widespread" to describe responses to the painful stimuli (R1 and R2). Terms such as "mislocalized" or "disorganized" may be more appropriate.

We have been careful throughout the manuscript to only refer to a ‘widespread’ response specifically within the S1 cortex (and equally refer to the touch response being localised within the S1) and we feel the term ‘widespread’ is more easily understood than ‘mislocalised’. Although the foot touch and lance HbO responses have the same size area of activation, the response to foot touch is localised within the S1, but extends beyond this into the motor cortex. However, the response to lance does not extend into the same area of the motor cortex, and instead extends into the entire S1. Thus, although the two maps are misaligned, the more important point here is that nociceptive maps are widespread in terms of S1 activation. We have made small clarifications throughout the manuscript to ensure there is no confusion regarding our use of the word widespread.

3) Treat motion equivalently for the two conditions in the analyses (R2).

We have performed additional analyses comparing the heel touch data with and without removing the trials with known movement. A sentence summarising this analysis and the results have been added to the methods section. No significant differences were found, suggesting that this difference between the heel touch and lance trials did not contribute significantly to the present findings. We have included a figure of the resulting channel-space time-series in our response to the reviewers and this can be added as a supplementary figure if needed.

Reviewer #1 (Recommendations for the authors):

1. Differences in regions of activations constituting touch maps and nociceptive maps are never actually statistically tested:

See the public review for an outline of the issue. The suggested solution is to directly compare the noxious and touch evoked activations and demonstrate statistically significant differences between them.

We have now included this analysis in the manuscript: ‘In regions significantly active following the lance stimulus, direct comparison at each node shows that the response following the heel lance was indeed larger than the response following the heel touch, conversely, the response following the heel lance was significantly smaller in the region only active following the heel touch’. We have also added an image of these results to figure 3.

2. Nociceptive maps are equally as widespread as touch maps:

See the public review for an outline of the issue. The distinction between "mislocalisation" versus "widespread" needs to be clarified, as the claim that nociceptive maps are widespread is ubiquitous and currently not sufficiently contextualised relative to the touch maps. Or at least the apparent tension between this core claim and the lack of difference in area of activation for the noxious and non-noxious stimulus needs explicit discussion. Of course, the "mislocalisation" claim would have to be backed up by a demonstration of statistically significant differences between the maps, which is currently not provided (see comment 1 above).

We have been careful throughout the manuscript to only refer to a ‘widespread’ or ‘localised’ response specifically within the S1 cortex. Although the heel touch and lance HbO responses have the same size area of activation, the response to heel touch is localised within the S1 (but extends beyond this into the motor cortex), and the heel lance response is widespread within the S1. The spread within the S1 is the important point here, rather than the size of the overall responses. We have also added the requested analysis to help further this point (see above) and have made small edits to clarify throughout the manuscript.

3. The analyses that produce areas of activation in source space (image space) are circular:

See the public review for an outline of the issue. The core message of this manuscript is that infants' nociceptive maps are more poorly localised than touch maps. This could be demonstrated in sensor space without any need to do source reconstruction. As outlined in comment 1 above, a direct statistical comparison of noxious evoked and touch evoked activity in sensor space would demonstrate the core message the authors wish to convey, and does not need source reconstruction meaning the circular analysis can be omitted altogether. And as mentioned above, all statistically significant results found in source space are not novel, so very little would be lost in keeping the analysis and results in sensor space and simple establishing statistically significant differences between the nociceptive and touch maps.

We have removed the latency restriction from the source-space analysis (i.e. we find the peak response in source-space rather than use the peak latency identified from the channel-space) and updated the results appropriately; the results have not changed as a consequence of this.

4. Limitations of the fNRIS technique and of the link between properties of nociceptive maps (widespread/mislocalisation) and poorly directed behaviour are not adequately discussed:

See the public review for an outline of the issue. This major limitation of the fNIRS technique requires much more serious and explicit discussion.

It is possible that systemic responses due to an increase in blood pressure may still be an issue following a lance stimulus. However, we believe that this is not a contributing factor in our results. If there was any remaining systemic response across the scalp following the heel lance, then we would expect all nodes to have a larger response following the lance compared to heel touch, or at least all channels/nodes to show a significant increase from baseline. However, despite the majority of the S1 (Figures 2c and 3a) being significantly active following the heel lance, there are still 9/17 channels that do not have a significant increase in HbO from baseline, and a region of the cortex which is more active following heel touch stimulation.

Moreover, although the control channel had a significant increase from baseline following the heel lance, this is not evidence that the heel lance elicited widespread systemic changes, as the heel touch also elicited a response in the control channel.

The overall areas of activation are the same for both stimuli, but they are not aligned. Although we use the term ‘widespread’, we are careful to only say as such within the S1, and not widespread across the whole cortex. Only a widespread response across the whole cortex could be viewed as being confounded by systemic changes. We have now added the discussion of systemic changes to the manuscript.

5. Linking comments 1, 2, and 4 to the initial hypothesis:

The hypothesis stated in the introduction is as follows: "We hypothesised that this developmental process is reflected in ascending nociceptive signals to SI, leading to widespread cortical activation and poor spatial localisation of noxious events in early life". The assessment of the poor localisation and widespread activation of noxious evoked activity is assessed and explicitly interpreted relative to touch evoked activity: "Noxious stimulation of the heel in the newborn evokes a more widespread cortical activation than innocuous stimulation, that extends into inferior regions of S1, normally associated with representation of the hand.".

However, using statistical tests to directly compare the noxious evoked activity to the touch evoked activity, there was neither a statistically significant difference between the noxious and touch peak localisations (i.e. peak locations do not differ) nor their extent of cortical activations (i.e. areas of activations do not differ). This subset of results seem to directly go against the initial hypothesis and interpreted conclusions.

A much clearer link between the tests run, results provided, and hypothesis put forward needs to be outlined, as it is currently very unclear why some results (e.g. activations assessed relative to baseline) should be considered supportive of the initial hypothesis but not indicative of fNIRS-related haemodynamic confounds (e.g. non-neural cardiovascular response signals such as tissue-damage evoked blood pressure changes), while other results (e.g. direct statistical comparisons between noxious and touch evoked activations) that seem to contradict the initial hypothesis are not considered as such by the authors.

Please see previous responses.

We have now added the discussion of systemic changes to the manuscript and altered the hypothesis to state that we expected widespread activation within the S1 specifically.

Reviewer #2 (Recommendations for the authors):

In order to address the issue of repetition suppression in the touch condition, it would be of interest to see either an analysis of change in the touch responses across trials, or a comparison between touch and nociceptive trials based on only the first touch trial.

In order to address this, we have carried out a statistical comparison of the first and last 50% of foot touch trials at each sample post-lance across all channels. Similar comparison was made between heel touch data with and without removing trials with known movement. We did not find any significant differences for either comparison, and the average time series can be seen in Author response image 1. Solid lines represent the average and dashed lines represent the standard deviation. We have now added these results to the manuscript and the image can be included as a supplementary if required.

Author response image 1

There is some small irregularity in the reporting of motion in the pain condition. The N for this condition in Table 1 is 11, but when reporting motion the authors state that 2 participants didn't move, 6 made small movements, and 4 made large movements.

Thank you for noting this error. This should be 5 made small movements.

Perhaps I am misreading the analysis procedure, but if the authors really are averaging responses to the touch/pain conditions and using a single, mean value for each time point in statistical tests (compared in a one-sample t-test to the baseline distribution?), then I do believe these analyses should be adjusted to retain variability across participants in response to the stimuli of interest.

We apologise for the confusion as this was an error in the manuscript; we averaged across trials not subjects. We have changed the relevant section in the methods.

‘Pre-processed data were then averaged across trials (except lance which only had one trial) and analysed using custom MATLAB scripts (Mathworks; version 16b).’

I was also surprised by the information on Bonferroni corrections. The authors were really multiplying their p values by thousands (channel-wise analyses) or hundreds of thousands (image-based analyses) of comparisons and arriving at corrected p values less than p = .001, in tests with 11 infants and a single trial per infant? I would not expect this sort of effect size in infant fNIRS research. Am I misunderstanding how a Bonferroni correction would be applied here?

The p-values provided in Figure 1 – source data 1 are indeed the corrected p-values, which were calculated by multiplying each p value (one per sample post-lance) by the number of comparisons (5100). Within the periods of significant change compared to baseline, we have provided the minimum and maximum p value.

[Editors’ note: further revisions were suggested prior to acceptance, as described below.]

The manuscript has been improved but the reviewers raised some remaining concerns and made some good suggestions. There are some remaining issues that need to be addressed, as outlined below:

– Please show the S1 ROI as suggested by R1

We have highlighted the central sulcus to the brain in Figure 4 so that it is now clear which optodes were sensitive to the postcentral gyrus and so used to delineate the somatosensory cortex. We have also modified the key in Figures 1 (channel space analysis) to include the 10-5 labels so the relationship between channel space and cortical surface is clearer.

– In the revision, you show that there is no discernible impact of motion on the heel touch trials in response to a concern raised by R2. However, this does not necessarily solve the issue raised. The suggestion to treat both trial types equally (as raised in the first round of review) should be followed.

This has now been done and the new results reported in the manuscript. Treating both trial types equally had increased the significance of the data and this is discussed in lines 220-224. The impact of motion is also added to the Methods (lines 430-439)

– Both reviewers and myself had a hard time following exactly how the image space analyses are conducted and whether the concern about circularity has been addressed or not. It sounds as if different nodes or timepoints are being selected on the basis of the responses observed, which still has the potential to introduce a selection bias (i.e. circularity). Please clarify and expand the text in describing the analyses.

The initial concern arose because we originally selected time windows to analyse in image space from results from channel space analysis. To address this concern about circularity we have separated channel and image space analysis completely. In image space we assess the significance of peak change against baseline. Importantly the peak change was searched across ANY node and latency avoiding any bias. We clarified this section of the Methods (lines 487-488 and 495-519).

– As noted by R2, the text at the top of page 5 is confusing. More generally, the descriptions in the text tend to be quite dense and complicated. Please revise the section noted and consider simplifying the text throughout to make the descriptions of the analyses and the corresponding conclusions easier to follow.

This section has been rewritten and the analysis described in simpler terms (lines 158-173)

– As suggested by R2, please consider providing the preprocessed data in addition to the raw files.

The preprocessed data will be available upon reasonable request to Dr Laura Jones, laura.a.jones@ucl.ac.uk.data

https://doi.org/10.7554/eLife.71655.sa2

Article and author information

Author details

  1. Laura Jones

    Department of Neuroscience, Physiology & Pharmacology, University College London, London, United Kingdom
    Contribution
    Data curation, Formal analysis, Investigation, Methodology, Project administration, Writing - original draft, Writing - review and editing
    For correspondence
    laura.a.jones@ucl.ac.uk
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-5755-4977
  2. Madeleine Verriotis

    Department of Neuroscience, Physiology & Pharmacology, University College London, London, United Kingdom
    Present address
    Department of Developmental Neuroscience, University College London, Great Ormond Street Institute of Child Health, London, United Kingdom
    Contribution
    Formal analysis, Investigation, Methodology
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-3019-0370
  3. Robert J Cooper

    Department of Developmental Neuroscience, University College London, London, United Kingdom
    Contribution
    Formal analysis, Methodology, Resources, Validation
    Competing interests
    No competing interests declared
  4. Maria Pureza Laudiano-Dray

    Department of Neuroscience, Physiology & Pharmacology, University College London, London, United Kingdom
    Contribution
    Investigation, Supervision
    Competing interests
    No competing interests declared
  5. Mohammed Rupawala

    Department of Neuroscience, Physiology & Pharmacology, University College London, London, United Kingdom
    Contribution
    Formal analysis
    Competing interests
    No competing interests declared
  6. Judith Meek

    Elizabeth Garrett Anderson Obstetric Wing, University College London Hospitals, London, United Kingdom
    Contribution
    Clinical supervision, Conceptualization, Formal analysis, Funding acquisition, Methodology, Supervision, Writing - review and editing
    Competing interests
    No competing interests declared
  7. Lorenzo Fabrizi

    Department of Neuroscience, Physiology & Pharmacology, University College London, London, United Kingdom
    Contribution
    Formal analysis, Funding acquisition, Investigation, Methodology, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-9582-0727
  8. Maria Fitzgerald

    Department of Neuroscience, Physiology & Pharmacology, University College London, London, United Kingdom
    Contribution
    Conceptualization, Funding acquisition, Methodology, Project administration, Supervision, Writing - original draft, Writing - review and editing
    For correspondence
    m.fitzgerald@ucl.ac.uk
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-4188-0123

Funding

Medical Research Council (MR/M006468/1)

  • Judith Meek
  • Lorenzo Fabrizi
  • Maria Fitzgerald

Medical Research Council (MR/L019248/1)

  • Lorenzo Fabrizi

Engineering and Physical Sciences Research Council (EP/N025946/1)

  • Robert J Cooper

Medical Research Council (MR/S003207/1)

  • Judith Meek
  • Lorenzo Fabrizi
  • Maria Fitzgerald

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

This work was funded by the Medical Research Council UK (MR/M006468/1, MR/L019248/1, and MR/S003207/1). RJC is funded by EPSRC Fellowship EP/N025946/1.

Ethics

Human subjects: Ethical approval for this study was given by the NHS Health Research Authority (London - Surrey Borders) and the study conformed to the standards set by the Declaration of Helsinki. Informed written parental consent was obtained before each study (REC no: 11/LO/0350; NIHR Portfolio Study ID: 12036). Separate media consent was obtained from the parent to use a photo of their child in academic publications (Figure 4a).

Senior and Reviewing Editor

  1. Chris I Baker, National Institute of Mental Health, National Institutes of Health, United States

Version history

  1. Received: June 25, 2021
  2. Preprint posted: July 30, 2021 (view preprint)
  3. Accepted: April 22, 2022
  4. Accepted Manuscript published: April 22, 2022 (version 1)
  5. Version of Record published: May 10, 2022 (version 2)

Copyright

© 2022, Jones et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 838
    Page views
  • 211
    Downloads
  • 5
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Laura Jones
  2. Madeleine Verriotis
  3. Robert J Cooper
  4. Maria Pureza Laudiano-Dray
  5. Mohammed Rupawala
  6. Judith Meek
  7. Lorenzo Fabrizi
  8. Maria Fitzgerald
(2022)
Widespread nociceptive maps in the human neonatal somatosensory cortex
eLife 11:e71655.
https://doi.org/10.7554/eLife.71655

Further reading

    1. Cell Biology
    2. Neuroscience
    Elisabeth Jongsma, Anita Goyala ... Collin Yvès Ewald
    Research Article Updated

    The amyloid beta (Aβ) plaques found in Alzheimer’s disease (AD) patients’ brains contain collagens and are embedded extracellularly. Several collagens have been proposed to influence Aβ aggregate formation, yet their role in clearance is unknown. To investigate the potential role of collagens in forming and clearance of extracellular aggregates in vivo, we created a transgenic Caenorhabditis elegans strain that expresses and secretes human Aβ1-42. This secreted Aβ forms aggregates in two distinct places within the extracellular matrix. In a screen for extracellular human Aβ aggregation regulators, we identified different collagens to ameliorate or potentiate Aβ aggregation. We show that a disintegrin and metalloprotease a disintegrin and metalloprotease 2 (ADM-2), an ortholog of ADAM9, reduces the load of extracellular Aβ aggregates. ADM-2 is required and sufficient to remove the extracellular Aβ aggregates. Thus, we provide in vivo evidence of collagens essential for aggregate formation and metalloprotease participating in extracellular Aβ aggregate removal.

    1. Computational and Systems Biology
    2. Neuroscience
    Marjorie Xie, Samuel P Muscinelli ... Ashok Litwin-Kumar
    Research Article Updated

    The cerebellar granule cell layer has inspired numerous theoretical models of neural representations that support learned behaviors, beginning with the work of Marr and Albus. In these models, granule cells form a sparse, combinatorial encoding of diverse sensorimotor inputs. Such sparse representations are optimal for learning to discriminate random stimuli. However, recent observations of dense, low-dimensional activity across granule cells have called into question the role of sparse coding in these neurons. Here, we generalize theories of cerebellar learning to determine the optimal granule cell representation for tasks beyond random stimulus discrimination, including continuous input-output transformations as required for smooth motor control. We show that for such tasks, the optimal granule cell representation is substantially denser than predicted by classical theories. Our results provide a general theory of learning in cerebellum-like systems and suggest that optimal cerebellar representations are task-dependent.