1. Human Biology and Medicine
  2. Neuroscience
Download icon

Localization of spontaneous bursting neuronal activity in the preterm human brain with simultaneous EEG-fMRI

  1. Tomoki Arichi
  2. Kimberley Whitehead
  3. Giovanni Barone
  4. Ronit Pressler
  5. Francesco Padormo
  6. A David Edwards Is a corresponding author
  7. Lorenzo Fabrizi Is a corresponding author
  1. King’s College London, United Kingdom
  2. Imperial College London, United Kingdom
  3. University College London, United Kingdom
  4. Catholic University of Sacred Heart, Italy
  5. UCL-Institute of Child Health, United Kingdom
  6. Icahn School of Medicine at Mount Sinai, United States
Short Report
Cited
0
Views
469
Comments
0
Cite as: eLife 2017;6:e27814 doi: 10.7554/eLife.27814

Abstract

Electroencephalographic recordings from the developing human brain are characterized by spontaneous neuronal bursts, the most common of which is the delta brush. Although similar events in animal models are known to occur in areas of immature cortex and drive their development, their origin in humans has not yet been identified. Here, we use simultaneous EEG-fMRI to localise the source of delta brush events in 10 preterm infants aged 32–36 postmenstrual weeks. The most frequent patterns were left and right posterior-temporal delta brushes which were associated in the left hemisphere with ipsilateral BOLD activation in the insula only; and in the right hemisphere in both the insular and temporal cortices. This direct measure of neural and hemodynamic activity shows that the insula, one of the most densely connected hubs in the developing cortex, is a major source of the transient bursting events that are critical for brain maturation.

https://doi.org/10.7554/eLife.27814.001

Introduction

In animal models, spontaneous bursts of synchronized neuronal activity (known as spindle bursts) play an instructive role in key developmental processes that set early cortical circuits, including neuronal differentiation and synaptogenesis (Hanganu-Opatz, 2010; Khazipov and Luhmann, 2006). Experimental disruption of the normal occurrence and propagation of this early spontaneous activity leads to permanent loss of healthy cortical organization, such as segregation into ocular dominance columns (Xu et al., 2011) and whisker barrels (Tolner et al., 2012) in the primary visual and somatosensory cortices respectively.

Neural activity recorded in human infants during the preterm period with electroencephalography (EEG) is also characterized by intermittent high amplitude bursts known as Spontaneous Activity Transients (SATs) (Khazipov and Luhmann, 2006; André et al., 2010; Tolonen et al., 2007). SATs appear to have a crucial role in early human brain development, as their occurrence is positively correlated to brain growth during the preterm period (Benders et al., 2015). The most common of these events is the delta brush, a transient pattern characterised by a slow delta wave (0.3–1.5 Hz) with superimposed fast frequency alpha-beta spindles (8–25 Hz) (André et al., 2010; Whitehead et al., 2017). Delta brushes appear from 28 to 30 weeks PMA (Boylan et al., 2008; Lamblin et al., 1999; Niedermeyer, 2005; Vecchierini et al., 2007), have a peak incidence at 32–35 weeks PMA (André et al., 2010; Boylan et al., 2008; Lamblin et al., 1999; D'Allest and Andre, 2002; Hahn and Tharp, 2005) and disappear between 38–42 weeks PMA (Boylan et al., 2008; Hahn and Tharp, 2005). They initially have a diffuse or predominantly peri-central distribution in infants <32 weeks PMA (Lamblin et al., 1999; Boylan, 2007; Volpe, 1995), progressing to have a more temporal and occipital (but rarely frontal) topography in late preterm infants (Tolonen et al., 2007; D'Allest and Andre, 2002; Hahn and Tharp, 2005; Volpe, 1995; Watanabe et al., 1999). As with spindle bursts in animal models, delta brushes can also be elicited by external stimuli (Chipaux et al., 2013; Colonnese et al., 2010; Fabrizi et al., 2011; Milh et al., 2007) with their topographies coarsely overlying the primary sensory cortices of the relevant stimulus modality, suggesting that the activation of specific cortical regions appears on the scalp surface as different delta brush distributions.

As delta brushes are the hallmark of the preterm EEG, reviewing their incidence and morphology is an important part of the clinical neurophysiological assessment of hospitalised infants (Whitehead et al., 2017). Preterm infants with a greater incidence of delta brushes are more likely to develop normally (Biagioni et al., 1994), while diminished occurrence or atypical morphology is seen in infants with major brain lesions such as periventricular leukomalacia who later develop cerebral palsy (André et al., 2010; Watanabe et al., 1999; Conde et al., 2005; Kidokoro et al., 2006; Okumura et al., 1999; Okumura et al., 2002; Tich et al., 2007). As delta brushes should disappear at term equivalent age, the number of events can also be used to determine the severity of EEG dysmaturity, which is defined by the presence of patterns that are at least 2 weeks immature relative to an infant’s PMA ([André et al., 2010; Hahn and Tharp, 2005; Holmes and Lombroso, 1993; American Clinical Neurophysiology Society Critical Care Monitoring Committee et al., 2013) and which is associated with adverse cognitive outcome if persistent over serial recordings (Okumura et al., 2002; Holmes and Lombroso, 1993; Hayakawa et al., 1997; Lombroso, 1985).

Despite their common occurrence, developmental importance and clinical significance, existing animal and human studies are insufficient to build a model of the role of these electrophysiological events in humans, in particular because of the lack of information about their neuro-anatomical source. Whilst delta brushes can be readily identified with EEG, the localization of their source within the brain cannot be easily inferred just from the electrical potentials recorded at the scalp surface (Darvas et al., 2004). To overcome this intrinsic limitation of EEG recording, we used simultaneous EEG-fMRI to combine the temporal sensitivity of EEG with the whole brain spatial specificity of functional Magnetic Resonance Imaging (fMRI). Here, we provide the first evidence that spontaneous patterns of delta brush activity in the period preceding normal birth are associated with significant hemodynamic activity clearly localized to distinct regions within the developing cortex. We show that the most common event in the late preterm period (posterior-temporal delta brushes) are reflective of activity in the insular cortices and temporal pole. These findings provide the first evidence of a direct link between spontaneous neural and hemodynamic activity in early human life and provide a new understanding of how they relate to regional cortical function during this critical period.

Results and discussion

Simultaneous EEG-fMRI data were successfully acquired in a group of 10 infants in their late preterm period (median PMA at data collection 35 + 1 weeks, range 32 + 2 to 36 + 2 weeks; five female) during natural sleep over a median of 7.5 min (range: 3.5–10.5 min). All the infants in the study sample were clinically well at the time of study and were reported as having normal brain appearances on their structural MR images. An optimized pre-processing and analysis pipeline which incorporated an age-specific hemodynamic response function (HRF) and template brain was used for the fMRI data (detailed in the supplementary methods) (Allievi et al., 2016; Arichi et al., 2012; Arichi et al., 2010; Serag et al., 2012). Due to the confounding effects of head motion on both EEG and fMRI data, several additional steps were also taken to specifically address this issue in both the initial pre-processing and analysis phase (see supplementary methods).

In line with the literature (André et al., 2010; Whitehead et al., 2017), delta brushes occurred frequently (median: 4.4/minute; range: 1.9–6.7) and with varying scalp distributions (23 in total; Figure 1 and Supplementary file 2). Nevertheless, right and left posterior-temporal delta brushes were consistently present in 10/10 and in 9/10 subjects respectively and could be associated with significant clusters of positive BOLD activity (p<0.05 with family wise error correction) in the ipsilateral insular cortex (Figure 2 and Figure 2—figure supplement 1). This result provides the first evidence that the insulae represent major locations of occurrence for these developmentally required neuronal events during our specific study window in the late preterm period. Although there are rapid changes occurring across the whole brain in human preterm development, the timing and trajectory of maturation has been shown to differ between regions (Makropoulos et al., 2016). In agreement with the idea that bursting neural activity is directly linked to brain maturation, the insular cortices in humans enter a crucial phase of their development during our study period (32–36 weeks PMA): (i) their volumetric growth trajectories (and those of the adjacent temporal lobes) accelerate (Makropoulos et al., 2016) and (ii) they establish an early pattern of dense functional and structural connectivity, which allows them to assume a prominent role as cortical hubs during infancy (Ball et al., 2014; Gao et al., 2011). As a result, the mature insulae have connections to almost all other regions of the brain, enabling them to play a versatile role in a wide range of functions including sensory and pain perception, multi-sensory integration, emotion, and cognition (Nieuwenhuys, 2012). Similarly, in primates and rodents, the insulae have a dense network of connections and play a key integrative function in sensory and behavioural processes (Butti and Hof, 2010; Mars et al., 2013; Miranda-Dominguez et al., 2014; Zingg et al., 2014). Their anatomical maturity is also more advanced in comparison to the surrounding cortex in early life (Huang et al., 2008; Kroenke et al., 2007). However, there are currently no animal developmental studies that directly address the relationship between bursting activity and maturation of this particular brain region.

Delta brushes occur with distinct topographies.

Segments of EEG recordings showing individual examples of delta brushes with the most common topographical distributions (occurred at least three times in a given subject). These rarely involved frontal and midline electrodes which are therefore omitted for illustration clarity. Right and left posterior-temporal delta brushes occurred in 10/10 and 9/10 subjects respectively, while other delta brushes were recorded in no more than two subjects. EEG traces and recording electrodes where delta brush activity was maximal are marked in red. Shaded areas represent the time of occurrence of each event.

https://doi.org/10.7554/eLife.27814.002
Figure 2 with 2 supplements see all
Localization of posterior-temporal delta brushes.

In a group of 10 preterm infants (35 + 0 weeks PMA, range 32 + 2 to 36 + 2 weeks), right posterior-temporal delta brush activity (blue) was significantly associated with BOLD clusters in the right temporal pole (z50, z55), right superior temporal lobe (x28), and the right insular cortex (z60, z62, z64, z66). Left posterior-temporal delta brush activity (red-yellow) was significantly associated with BOLD clusters in the left posterior insula (z64, z66) and left parietal operculum (z66, z68). Images show the results of a one-sample t-test (p<0.05) performed using permutation testing and corrected for family-wise error overlaid on an age-specific T2-weighted brain atlas (Serag et al., 2012).

https://doi.org/10.7554/eLife.27814.003

The importance of the preterm period for insular development in humans is further emphasized by studies showing that the degree of prematurity at birth, recreational drug use in pregnancy and late onset intra-uterine growth restriction adversely affects both insular volume and thalamo-insular connectivity at term equivalent age (Ball et al., 2012; Batalle et al., 2016; Egaña-Ugrinovic et al., 2014; Grewen et al., 2015; Salzwedel et al., 2015), with the latter being significantly correlated with cognitive outcome at 2 years of age (Ball et al., 2015). Furthermore, insular dysfunction and poor growth have been implicated in a range of psychiatric conditions, including neurodevelopmental difficulties such as autism spectrum and attention deficit hyperactivity disorders which have greater prevalence in preterm born children (Hatton et al., 2012; Johnson and Marlow, 2011).

In addition to the insulae, right-sided posterior-temporal delta brushes were associated with significant clusters of hemodynamic activity in the right superior temporal lobe and pole (Figure 2). This finding is of particular significance as these are regions where the subplate, a transient structure which is thought to play a fundamental role in the generation of spindle burst activity in animals (Tolner et al., 2012), can be qualitatively appreciated on high resolution MR images and histology (Figure 2—figure supplement 2) (Kostović et al., 2014). In human development, the subplate follows a similar trajectory to delta brush activity, reaching maximal thickness in the middle of the third trimester before disappearing in most of the brain by term equivalent age (Kostović et al., 2014). The present results therefore support a link between these functional and structural developmental features in humans.

Source localizing spontaneous delta brushes to the insulae and temporal pole does not necessary imply that this activity starts here. Spindle bursts in animals are recorded from the cortical plate (Yang et al., 2013), but are thought to be driven by spontaneous activity from the periphery (whisker pad [Yang et al., 2009], spinal cord [Inácio et al., 2016], retina [Hanganu et al., 2006] and cochlea [Johnson et al., 2011]); as well as from central pattern generators (CPGs) such as the primary motor cortex, brainstem and thalamus within the somatomotor system (for review see [Luhmann et al., 2016]). It is therefore possible that other neuronal events in these centers may also precede the occurrence of delta brushes in humans, but cannot be detected with EEG and fMRI. In this instance, activity would then be amplified by the subplate resulting in measurable electrical-hemodynamic events in the cortex. Nevertheless, it is unlikely that the insular activation we observed here resulted directly from activity in the sensory periphery as the insulae are not involved in the primary processing of visual (Lee et al., 2012), auditory (Baldoli et al., 2015) or somatosensory stimuli (Allievi et al., 2016; Arichi et al., 2010).

Bilateral and unilateral parietal, occipital and mid-temporal, but rarely frontal, delta brushes were also sporadically recorded in individual subjects, suggesting the presence of other less active sources of spontaneous activity at this developmental stage (Figure 3 and Supplementary file 2). In addition to the ipsilateral primary clusters, other areas of BOLD activity were also frequently seen in the anatomical homologue in the opposite hemisphere and occasionally in association areas of the cortex (supplementary motor area (SMA), anterior cingulate, precuneus) or deeper structures of the brain (thalamus and basal ganglia) (Supplementary file 3). Delta brushes with topographies other than posterior-temporal are more frequent earlier on in development compared to the age window studied here (Lamblin et al., 1999; Boylan, 2007; Volpe, 1995) and may represent activity from other developing brain regions which follow a different maturational trajectory. However precise localization of the sources of these events and others that were not recorded in our study group will require further work in a larger study population which spans other age ranges when these regions may be more active. Such longitudinal work may also allow an exploration of whether key features of spindle burst activity in animals are also present in humans during maturation, such as increasing propagation of neuronal activity in local and neighbouring networks (Yang et al., 2013) and regional differences in oscillatory patterns (Yang et al., 2009), as well as association of these events to the presence of the transient subplate layer.

Localization of delta brush events in a single preterm infant.

Example of the significant hemodynamic activity correlated to less frequent delta brushes in a single preterm subject at 35 + 6 weeks PMA. (a) The occurrence of bilateral posterior-temporal delta brushes was significantly associated with well localized clusters of BOLD activity (red-yellow) in the bilateral superior temporal lobe and insulae (z55); while (b) bilateral occipital delta brushes were associated with a cluster in the medial occipital region (z33). Images show the thresholded z-statistical map with a corrected cluster significance of p<0.05 overlaid on the subject’s T2-weighted image.

https://doi.org/10.7554/eLife.27814.006

Despite the apparent absence of a tight neurovascular coupling in perinatal rodent models (Kozberg et al., 2016; Zehendner et al., 2013), we demonstrated for the first time in human infants, a clear association between a direct measure of neural (EEG) and positive functional hemodynamic activity (fMRI). Whilst in rodents, neurovascular coupling matures postnatally together with the development of long-range connectivity patterns (Kozberg et al., 2016), in humans this connectivity can be readily identified by the late preterm period, thus suggesting that neuronal and hemodynamic activity are already closely linked by this time (Allievi et al., 2016; Doria et al., 2010; White et al., 2012). This relationship is developmentally regulated across the neonatal period resulting in changing hemodynamic responses (Arichi et al., 2012) and is validated by the presence of localised positive BOLD activation in the primary auditory and somatosensory cortices following sound and passive motor stimulation respectively (Allievi et al., 2016; Arichi et al., 2010; Baldoli et al., 2015; Erberich et al., 2003).

Spontaneous activity is a fundamental feature of developing neural circuits well before the establishment of cortical layers (Luhmann et al., 2016) and refinement through experience dependent mechanisms (Khazipov and Luhmann, 2006). Our findings provide the first evidence that the most common of these neuronal events in the late preterm period are seen in the posterior temporal regions and are largely generated by the insulae and subplate. As these events are known to have an instructive function in cortical maturation in rodents (Hanganu-Opatz, 2010; Khazipov and Luhmann, 2006; Rakic and Komuro, 1995), our results suggest that these structures may play a key developmental role as a major location of these bursting events in early human life.

Materials and methods

Participants

Thirteen preterm infants (five females; studied between 32–36 weeks post-menstrual age, PMA) 5–55 days old (23 ± 17, mean ±SD) were recruited for this study from the Neonatal Unit at St Thomas’ Hospital, London (patient demographic information is detailed in Supplementary file 1). Informed written parental consent was obtained prior to each study. The research methods conformed to the standards set by the Declaration of Helsinki and were approved by the National Research Ethics Committee.

Medical case notes were reviewed and infants were assessed as clinically stable by an experienced pediatrician at the time of study. Infants were excluded if they required any respiratory support during scanning or if they were known to have a history of severe brain pathology such as extensive intraventricular hemorrhage (grade 3 with ventricular dilatation; grade 4 with parenchymal extension), birth asphyxia, focal intracerebral lesions affecting the parenchyma or white matter (such as infarction, overt hemorrhage, or multiple punctate white matter lesions), severe hydrocephalus, or congenital brain malformations.

EEG-fMRI acquisition

MR images were collected following feeding and during natural sleep on a 3-Tesla Philips Achieva scanner (Best, Netherlands) located on the Neonatal Unit. Each infant was fitted with ear protection (moulded dental putty and adhesive earmuffs (Minimuffs, Natus Medical Inc, San Carlos CA, USA)) and immobilized using a vacuum cushion (Med-Vac, CFI Medical Solutions, Fenton, MI, USA). An appropriately sized, custom-made cap containing 26–32 scalp electrodes (EasyCAP GmbH, DE) was fitted on the head of each infant prior to scanning and connected to an MR-compatible EEG system (Brain Products GmbH, DE, RRID:SCR_009443). Blood Oxygen Level Dependent (BOLD) functional MRI data (299–499 volumes) were collected using a T2*-weighted single-shot gradient echo echo-planar imaging (GRE-EPI) sequence (resolution: 2.5*2.5*3.25 mm; 21 slices; TE: 30-45msec; TR: 1500msec, FA: 60–90 degrees; SENSE factor 2). Exact synchronization between the two recording modalities was achieved by marking each MR volume acquisition on the EEG using a TTL trigger generated by the MR scanner. High resolution MPRAGE (Magnetization-prepared Rapid Gradient Echo) T1- and TSE (Turbo Spin Echo) T2-weighted MRI scans were also acquired in the same study session for registration purposes and to allow more precise anatomical localization of the identified BOLD signal changes (Merchant et al., 2009). All high resolution structural images were formally reported by a Neonatal Neuroradiologist as showing normal appearances for age. As reported in the literature, the subplate layer could be qualitatively appreciated in all of our study subjects as an area of high signal on T2 images lying just below the cortex, which was most prominently seen in the temporal poles bilaterally (Kostovic and Rakic, 1990) (Figure 2—figure supplement 2).

EEG pre-processing and analysis

Gradient artefacts caused by the MR image acquisition were filtered from the EEG data using a commercially available EEG processing software package (Analyzer 2; Brain Products, DE). EEG cardioballistic artefacts, which are typically observed in adults (Allen et al., 1998), were not present in our neonatal recordings. Three independent trained observers (KW, GB, LF) reviewed the EEG recordings and marked the occurrence of delta brush events. Delta brushes were defined as bursts of fast frequency ripples of 8–25 Hz superimposed on a slow wave of 0.3–1.5 Hz (Khazipov and Luhmann, 2006; André et al., 2010; Milh et al., 2007). Inter-rater reliability was assessed using Fleiss’ Kappa analysis (Fleiss, 1971) and resulted in a substantial agreement (median Fleiss’ Kappa 0.65 (range 0.25–0.76)). Consensus on delta brush occurrence was then reached amongst the three reviewers for each event and confirmed by a Consultant Pediatric Clinical Neurophysiologist (RP). Events were then labelled based on their field distribution as having unilateral (right – R or left – L), midline (M) or bilateral (B) frontal (F), central (C), temporal (T), parietal (Pa), posterior-temporal (PT), occipital (O), posterior-temporal occipital (PTO) or posterior quadrant (PQ) topography (Supplementary file 2). Different topographical distributions were then used as separate Explanatory Variables (EVs) in the general linear model (GLM) of the fMRI analysis (see below). Only EVs containing at least 3 events were used for analysis. Two data sets were discarded because of insufficient EEG quality (due to bridging electrodes or unsuccessful artifact removal) which made them unsuitable for reliable delta brush detection.

fMRI data pre-processing and subject level analysis

fMRI data pre-processing and analysis were performed with an optimized pipeline for studying data acquired from neonatal subjects using tools implemented in FSL (FMRIB’s software library, www.fmrib.ox.ac.uk/fsl, RRID:SCR_002823) (Allievi et al., 2016; Arichi et al., 2012; Arichi et al., 2010; Arichi et al., 2014; Arichi et al., 2013; Smith et al., 2004). Each dataset was visually reviewed to check for data quality and for overt motion artifact. BOLD contrast time-series were then truncated to exclude excessive motion signal artifact (based on the root mean square intensity difference to the center reference volume) at the beginning or at the end of the recordings. One data set was discarded as the remaining data segment did not contain more than 3 delta brushes with the same topography (i.e. belonging to the same EV).

The remaining 10 subject datasets were then processed using an optimized pre-processing pipeline which was implemented in FEAT (fMRI Expert Analysis Tool, v5.98), including rigid-body head motion correction (using MCFLIRT), slice-timing correction, non-brain tissue removal (using BET), spatial smoothing (Gaussian filter of full-width half-maximum [FWHM] 5 mm), global intensity normalization, and high-pass temporal filtering (cut-off 50 s) (Smith et al., 2004; Woolrich et al., 2001). As motion artifact is known to represent a key source of bias in fMRI data, residual motion and physiological noise (such as those associated with vascular or respiratory effects) were removed by performing data-denoising with MELODIC (Model-free fMRI analysis using Probabilistic Independent Component Analysis [PICA, v3.0]) (Beckmann and Smith, 2004).

Statistical analysis in FEAT was done with FMRIB’s improved linear model (FILM) with local autocorrelation correction (Woolrich et al., 2001). A general linear model (GLM) was used to perform a univariate (voxel-wise) fitting of the observed data to a linear combination of our explanatory variables (EVs). These included: (i) one EV for each delta brush topography (e.g. one EV for right posterior-temporal delta brushes, another one for left posterior-temporal delta brushes, etc.) representing the occurrence of each event convolved with a set of basis functions optimised for the preterm hemodynamic response (Arichi et al., 2012) and (ii) to further ensure that motion artifact did not affect our results, binary confound regressors to exclude each volume affected by motion (and identified as a signal outlier in the timeseries based on the root mean square intensity difference to the reference center volume) despite MELODIC denoising. The calculated t-statistical image was then converted to a z-statistical image and a threshold of 2.3 with a corrected cluster significance level of p<0.05 was then used to generate spatial maps of activated voxels on an individual subject level (Figure 2—figure supplement 1). Activation maps were then registered to the individual subject's high-resolution structural T2-weighted image using a 6 DOF rigid-body registration (FLIRT v5.5) (Jenkinson and Smith, 2001). The spatial distribution of significant clusters of BOLD activity for each subject are summarised in Supplementary file 3.

fMRI group level analysis

Individual subject activation maps corresponding to homologous delta brush topographical distribution were co-aligned to an age-specific spatio-temporal neonatal atlas using FSL's nonlinear image registration tool (FNIRT v2.0) (Serag et al., 2012). Group average functional clusters at a significance of p<0.05 were then identified using permutation testing as implemented in FSL Randomise (v2.1) (Nichols and Holmes, 2002). A non-parametric single-group t-test with threshold-free cluster enhancement (TFCE) with family-wise error correction (FWE) to correct for multiple comparisons was then used to identify study population clusters associated with left and right posterior temporal delta brush activity (Smith and Nichols, 2009).

References

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
  11. 11
  12. 12
  13. 13
  14. 14
  15. 15
  16. 16
  17. 17
    Neonatal and Paediatric Clinical Neurophysiology
    1. GB Boylan
    (2007)
    Neurophysiology of the neonatal period, Neonatal and Paediatric Clinical Neurophysiology, Churchill Livingstone Elsevier.
  18. 18
    Neonatal Cerebral Investigation
    1. GB Boylan
    2. D Murray
    3. J Rennie
    (2008)
    The normal EEG and aEEG, Neonatal Cerebral Investigation, 2nd Edition, Cambridge University Press.
  19. 19
  20. 20
  21. 21
  22. 22
  23. 23
    The Newborn Brain: Neuroscience and Clinical Applications
    1. AM D'Allest
    2. M Andre
    (2002)
    Electroencephalography, The Newborn Brain: Neuroscience and Clinical Applications, 1st Edition, Cambridge University Press.
  24. 24
  25. 25
  26. 26
  27. 27
  28. 28
  29. 29
  30. 30
  31. 31
  32. 32
    Electrodiagnosis in Clinical Neurology
    1. JS Hahn
    2. BR Tharp
    (2005)
    Neonatal and paediatric encephalography, Electrodiagnosis in Clinical Neurology, 5th Edition, Elsevier.
  33. 33
  34. 34
  35. 35
  36. 36
  37. 37
  38. 38
  39. 39
  40. 40
  41. 41
  42. 42
  43. 43
  44. 44
  45. 45
  46. 46
  47. 47
  48. 48
  49. 49
    [Electroencephalography of the premature and term newborn. Maturational aspects and glossary]
    1. MD Lamblin
    2. M André
    3. MJ Challamel
    4. L Curzi-Dascalova
    5. AM d'Allest
    6. E De Giovanni
    7. F Moussalli-Salefranque
    8. Y Navelet
    9. P Plouin
    10. MF Radvanyi-Bouvet
    11. D Samson-Dollfus
    12. MF Vecchierini-Blineau
    (1999)
    Neurophysiologie Clinique = Clinical Neurophysiology 29:123–219.
  50. 50
  51. 51
    Neonatal polygraphy in full-term and premature infants: a review of normal and abnormal findings
    1. CT Lombroso
    (1985)
    Journal of Clinical Neurophysiology : Official Publication of the American Electroencephalographic Society 2:105–155.
  52. 52
  53. 53
  54. 54
  55. 55
  56. 56
  57. 57
  58. 58
  59. 59
    Electroencephalography: Basic Principles, Clinical Applications and Related Fields
    1. E Niedermeyer
    (2005)
    Maturation of the EEG: development of waking and sleep patterns, Electroencephalography: Basic Principles, Clinical Applications and Related Fields, Philadelphia, Lippincott Williams & Wilkins.
  60. 60
  61. 61
  62. 62
  63. 63
  64. 64
  65. 65
  66. 66
  67. 67
  68. 68
  69. 69
  70. 70
  71. 71
  72. 72
    Specialised studies in the neurological evaluation (5th Edition)
    1. JJ Volpe
    (1995)
    Saunders Elsevier.
  73. 73
  74. 74
  75. 75
  76. 76
  77. 77
  78. 78
  79. 79
  80. 80
  81. 81

Decision letter

  1. Sabine Kastner
    Reviewing Editor; Princeton University, United States

In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.

Thank you for submitting your article "Localization of spontaneous bursting neuronal activity in the preterm human brain with simultaneous EEG-fMRI" for consideration by eLife. Your article has been reviewed by two peer reviewers, and the evaluation has been overseen Sabine Kastner as the Senior Editor and Reviewing Editor. The following individual involved in review of your submission has agreed to reveal his identity: Heiko J Luhmann (Reviewer #2).

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

Summary:

The topic of the paper is important and clinically relevant: Understanding the mechanisms underlying neocortical activity in the preterm human brain. Over the last years a number of insightful reports on this topic have been published (e.g. Vanhatalo lab) and we learned a lot from animal studies. However, it is still unclear where and how defined early cortical rhythms are generated. This paper is focused on the so-called δ brush (spindle burst) activity recorded by combination of EEG and fMRI in 10 preterms aged 32-36 postmenstrual weeks. The authors identified the insula and the subplate as a major source for this activity.

The reviewers and editors thought that this paper has the potential to make an insightful contribution. However, due to the lack of important methods details, the study could not be fully evaluated at this stage. In order to fully evaluate the study, it will be necessary to make the following essential revisions. Please also note that there are other important, more detailed concerns that can be extracted from the appended reviews.

Essential revisions:

1) Scholarship: Both reviewers thought that the literature regarding δ brushes is not well considered, and the broader significance of the findings needs to be clarified.

2) The development of insula needs to be discussed in relation to other brain regions, and importantly the authors need to explain how they source localized this region relative to others.

3) Both reviewers noted that numerous methods details were missing, and due to that, the work could not fully be evaluated. Please include a thorough methods description that will permit other investigators to replicate your findings.

4) Alternative explanations need consideration (and trivial explanations such as scanner noise need to be ruled out).

Reviewer #1:

This paper is looking at a difficult population to study, preterm infants, and recording simultaneous EEG and fMRI. This by itself is noteworthy and the questions that they are pursuing are also interesting in terms of understanding aspects of early brain development. Data from such studies are valuable.

My enthusiasm for this paper is somewhat tempered, however, by a few points. First the review of literature is rather light on δ brushes, as they are citing primarily their own review paper. As these are transient neurophysiological events, apparently elicited by sensory stimuli, the uninformed reader would like to know 'so what?' What is their relevance other than the general statement of their absence or their persistence being associated with injury or adverse outcome?

They discuss how the insular cortices are in a critical phase of development in this late preterm period. But is not the rest of the brain as well undergoing massive development? Is the development in the insulae distinct from other brain areas at this time? This needs to be further discussed. As the scanner is a noisy environment, could the localisation of these sources be related to auditory sensory effects?

In terms of their methods there needs to be a greater detail on the fMRI processing pipeline. For example, it is unusual to use the basic default options in FEAT and particularly with preterm brains. There needs to be more discussion and detail on the motion correction, as this is a big issue with infant scanning. Basically, a lot more detail on the methods is needed. Also, the infants scanned were selected how? Were they all in-patients (which suggest that they may not have been healthy)? Some caution needs to be taken in the interpretation as it is a small N.

The authors are to be commended on the work and the approach; some further information is needed prior to being acceptable for publication, so that the full understanding of the methods, effects and their relevance is more readily accessible

Reviewer #2:

The topic of the paper is important and clinically relevant. I have the following questions and concerns:

1) EEG and fMRI have limitations in their spatial and temporal resolution. It is not easy to identify functional connections and "hub" regions or even layers (as the subplate) with these techniques and some important structures in this context cannot be studied (spinal cord, retina, cochlea). From animal studies (rodents and monkeys) we know that spontaneous δ brush/ spindle burst activity in newborn neocortex is driven by the sensory periphery (e.g. retinal bursts). In newborn rodent somatosensory cortex spontaneous activity is triggered by motor patterns (CPGs) in motor cortex, spinal cord, brainstem (see your reference Luhmann et al., 2016 for discussion of this issue). So all these previous studies on rodents and primates do not support the conclusion of the present paper.

2) How do the authors exclude that the insula is just co-activated by the activity patterns generated in the retina, spinal cord etc.?

3)Results and Discussion, second paragraph: What is the evidence that the insula in newborn rodent cortex or preterm primate neocortex is so well connected and may fulfil the role of a hub region?

4) In newborn rodent cortex and some EEG studies on preterm human cortex, spindle bursts are initially local (columnar) events, which at later stages propagate to neighbouring regions (e.g. papers by JW Yang et al). What is the developmental pattern in the present manuscript?

5)Results and Discussion, third paragraph: The subplate certainly does not drive the spindle bursts in vivo, it acts as a relay or amplifier for the activity coming from subcortical inputs.

6) How can the authors identify the subplate? Does the spatial resolution of the methods allow this major conclusion (Results and Discussion, last paragraph)?

7) The authors only studied "spontaneous" activity? Is the insula involved in evoked δ brush activity?

8) Did the authors observe in their EEG recordings regional differences in the δ brush properties?

https://doi.org/10.7554/eLife.27814.011

Author response

Essential revisions:

1) Scholarship: Both reviewers thought that the literature regarding δ brushes is not well considered, and the broader significance of the findings needs to be clarified.

We have now extended the literature review regarding δ brushes throughout the manuscript. This has included (i) emphasizing their clinical importance in the introduction and (ii) better placing our work within the current knowledge of early spontaneous neuronal activity. Overall, we have added and discussed references to 42 extra relevant papers from animal models and humans studies. In particular see response 1 to reviewer 1 and responses 1, 3-8 to reviewer 2.

2) The development of insula needs to be discussed in relation to other brain regions, and importantly the authors need to explain how they source localized this region relative to others.

We have discussed the development of the insulae in relation to other brain regions and extended consideration of the importance of the preterm period for their development. We have also clarified that the insula is specifically the source/location of posterior-temporal δ brushes (and not of those with other topographies), as the resting hemodynamic activity of this brain region only is linked to their occurrence (as demonstrated by our analysis). We have also stressed the difference between neuronal activity source and generator: with respect to our study,the insula is the location of the recorded activity, but may not necessarily be the location from where it started (which may be in the periphery or from central patterns generators which cannot be monitored with current non-invasive recordings (see response 1 to reviewer 2)). Moreover, we would also like to emphasize that posterior-temporal δ brushes are the most common neuronal events in our study period, but this does not exclude that at other developmental stages other brain areas may be more active and sources of other type of δ brushes (e.g. pericentral and occipital). Please see the response 2 to reviewer 2 and responses 2, 3 and 7 to reviewer 2.

3) Both reviewers noted that numerous methods details were missing, and due to that, the work could not fully be evaluated. Please include a thorough methods description that will permit other investigators to replicate your findings.

We have now provided a more thorough method description. See responses 4 and 5 to reviewer 1 and response 6 to reviewer 2.

4) Alternative explanations need consideration (and trivial explanations such as scanner noise need to be ruled out).

We have now discussed alternative explanations and ruled out trivial ones. See response 3 to reviewer 1 and responses 2,5 and 7 to reviewer 2.

Reviewer #1:

This paper is looking at a difficult population to study, preterm infants, and recording simultaneous EEG and fMRI. This by itself is noteworthy and the questions that they are pursuing are also interesting in terms of understanding aspects of early brain development. Data from such studies are valuable.

My enthusiasm for this paper is somewhat tempered, however, by a few points.

First the review of literature is rather light on δ brushes, as they are citing primarily their own review paper. As these are transient neurophysiological events, apparently elicited by sensory stimuli, the uninformed reader would like to know 'so what?' What is their relevance other than the general statement of their absence or their persistence being associated with injury or adverse outcome?

We originally took advantage of our recent review to comply with the word limit for an eLife short report (1500 words). We have now expanded our literature review about δ brushes to make their relevance clearer to a wide scientific audience (Introduction, first and second paragraphs).

They discuss how the insular cortices are in a critical phase of development in this late preterm period. But is not the rest of the brain as well undergoing massive development? Is the development in the insulae distinct from other brain areas at this time? This needs to be further discussed.

We agree with the reviewer that the entire human brain is certainly undergoing extensive developmental changes during the preterm period and we apologise if we gave a different impression. However, development within specific brain regions has been found to occur at different trajectories and our findings suggest that the distribution of δ brush activity may reflect this maturation. Δ brushes with topographies other than posterior-temporal (perhaps representing activity from other developing brain regions) occur less frequently during the studied developmental juncture and are more frequent earlier on (Volpe 1995, Lamblin, Andre et al. 1999, Boylan 2007). In keeping with our findings, insular growth is particularly rapid during our specific study age window in comparison to other areas (Makropoulos, Aljabar et al. 2016). In addition, there is converging evidence that the insulae is specifically vulnerable to external insults during our study period as both its connectivity and volume have been found to be altered following drug use in pregnancy, late onset intra-uterine growth restriction, and preterm birth (Egana-Ugrinovic, Sanz-Cortes et al. 2014, Grewen, Salzwedel et al. 2015, Salzwedel, Grewen et al. 2015, Batalle, Munoz-Moreno et al. 2016).

We have amended the manuscript to acknowledge that rapid development is occurring across the whole brain in the preterm period (Results and Discussion, second paragraph), have clarified that our results may reflect the different developmental trajectories of specific brain areas (Results and Discussion, sixth paragraph) and added the references about specific insular vulnerability (Results and Discussion, third paragraph).

As the scanner is a noisy environment, could the localisation of these sources be related to auditory sensory effects?

MRI scanners are indeed a noisy environment, although 3 levels of sound protection are provided for the infants (dental putty in the external auditory meatus, ear muffs, vacuum cushion around the head). Moreover, the factors that cause the noise (the helium pump cooling system, gradient coil vibration and other hardware aspects) are ongoing, regular and most importantly, do not covary with the events of interest. Scanner noise will therefore not contribute to activity related to specific neural events occurring at intermittent and irregular times such as those identified in the current study (Moelker and Pattynama 2003).

We can be further certain that the patterns of activity identified in our study are not related to the scanner noise, as functional responses to experimental auditory stimuli in preterm infants have been localised to the superior temporal gyrus and not the insula with a number of different imaging modalities. These include: auditory evoked δ brushes measured with EEG at the mid-temporal and not posterior temporal electrodes (Chipaux, Colonnese et al. 2013), hemodynamic responses over the temporal regions measured with Near Infrared Spectroscopy (NIRS) (Zaramella, Freato et al. 2001, Mahmoudzadeh, Dehaene-Lambertz et al. 2013) and bilateral BOLD responses in the superior temporal lobes measured with fMRI (Baldoli, Scola et al. 2015).

In terms of their methods there needs to be a greater detail on the fMRI processing pipeline. For example, it is unusual to use the basic default options in FEAT and particularly with preterm brains. There needs to be more discussion and detail on the motion correction, as this is a big issue with infant scanning. Basically, a lot more detail on the methods is needed.

The reviewer is correct in their assertion that the analysis of neonatal fMRI data needs careful adaptation and we apologize if there was insufficient detail about the processing methods in the manuscript. We are grateful for the opportunity to provide more technical details on our fMRI processing pipeline which has already been successfully applied with neonatal subjects (including preterm infants) to localize somatosensory responses (Arichi, Moraux et al. 2010, Arichi, Fagiolo et al. 2012, Allievi, Arichi et al. 2016), olfactory responses (Arichi, Gordon-Williams et al. 2013), and following brain injury (Arichi, Counsell et al. 2014).

We would like to clarify that whilst the analysis was performed using FSL’s FEAT package, the processing steps and analysis settings have all been optimized (and if necessary, customized) for the neonatal population. The first stages of the analysis pipeline do indeed involve the same pre-processing steps as those used in a standard adult analysis as these are necessary to deal with potential sources of bias which are inherent to all fMRI data regardless of the study population. These include: (i) high pass filtering of the data (to remove a slow “drift” of the signal due to a gradual change in the steady state of tissue magnetization during data acquisition), (ii) slice-timing correction (due to the fact that data is acquired in sequential slices across a given volume), (iii) non-brain tissue removal (so that only activity inside the brain is analysed), and (iv) spatial smoothing (to increase signal-to-noise-ratio). After that, specific adaptations for studying neonatal subjects are applied. These included taking into account the longer lag of the preterm hemodynamic responses through the use of a population specific hemodynamic response function (HRF) in the GLM (Arichi, Fagiolo et al. 2012), an optimized registration pipeline which used both affine and non-linear methods, an age-specific template brain (Serag, Aljabar et al. 2012), and non-parametric permutation methods for the group analysis given the clear uncertainty about the distribution of data across the wider population of preterm infants.

The reviewer is also correct that appropriately dealing with the effects of head motion is crucial in fMRI analysis and is of particular relevance when studying a neonatal population. For this reason, we used extremely strict criteria for rejecting both EEG and fMRI data which was corrupted by motion and used further correction steps during the analysis itself in addition to rigid-body re-alignment of each volume (as in a “standard” fMRI analysis). This included truncating fMRI data sets based on the calculated root mean squared intensity difference to a reference and using a binary confound regressor in the GLM for signal outliers generated by head motion (Power, Barnes et al. 2012). We also used Independent Component Analysis (ICA) to identify non-linear effects on the fMRI data generated by head motion and physiological movements (cardiovascular pulsation and respiratory movements), thus allowing the removal of specific artifacts and clear sources of bias.

We have now added more information in the main body of the manuscript to highlight the additional steps used to deal with head motion and optimized analysis (Results and Discussion, first paragraph and additions in subsection “fMRI data pre-processing and subject level analysis”).

Also, the infants scanned were selected how? Were they all in-patients (which suggest that they may not have been healthy)? Some caution needs to be taken in the interpretation as it is a small N.

All the infants included in the study group were recruited and scanned during the preterm period and were all in-patients (either on the postnatal ward or special care baby unit) when studied. However, their admission was simply to allow provision of food and warmth and the main clinical priority was to establish regular oral feeding schedule and monitor weight gain. Immediately prior to data acquisition, all the infants were assessed by an experienced Paediatrician and were adjudged to be clinically stable for scanning. None of the infants studied required respiratory support during data acquisition.

Given that the aim of the study was to spatially localize bursting neuronal events in a normally developing brain, we specifically excluded infants with identified (or potential) abnormal large-scale anatomy (localized brain injury, diagnosed congenital disorder or chromosomal syndrome), abnormal white matter microstructure (a clinical history of birth asphyxia, sepsis, severe intra-uterine growth restriction, chronic lung disease), or altered baseline brain activity (meningoencephalitis, encephalopathy, a clinical history of seizures or medication known to alter consciousness levels). In addition to the simultaneous EEG-fMRI data, we also acquired high resolution structural MRI which was fully reported by a Neonatal Neuroradiologist. All the structural images acquired from our study population were reported as normal for age.

We have now clarified in the manuscript that all the infants recruited were clinically well at the time of study and had normal appearances on their structural MR images (Results and Discussion, first paragraph and subsection “EEG-fMRI acquisition”).

Reviewer #2:

The topic of the paper is important and clinically relevant. I have the following questions and concerns:

1) EEG and fMRI have limitations in their spatial and temporal resolution. It is not easy to identify functional connections and "hub" regions or even layers (as the subplate) with these techniques and some important structures in this context cannot be studied (spinal cord, retina, cochlea). From animal studies (rodents and monkeys) we know that spontaneous δ brush/ spindle burst activity in newborn neocortex is driven by the sensory periphery (e.g. retinal bursts). In newborn rodent somatosensory cortex spontaneous activity is triggered by motor patterns (CPGs) in motor cortex, spinal cord, brainstem (see your ref Luhmann et al., 2016 for discussion of this issue). So all these previous studies on rodents and primates do not support the conclusion of the present paper.

We think that this conclusion may arise from the different value that the concept of “generators” has in animal models compared to human non-invasive imaging. As you correctly point out, in the animal world, “generators” are those cells or group of cells where the activity is originating. On the other hand, in the non-invasive imaging world, “generators” is often used to refer to those centres whose activation produces the activity recorded on the scalp, but are not necessarily the point where the activity started from. To clarify, we cannot exclude that there may be neuronal activity in the thalamus, other central pattern generators or driven by the sensory periphery (even if this latter is unlikely, see reply to question 2) which cannot be picked up with EEG or fMRI, but that is then amplified by the subplate resulting in measurable electrical-hemodynamic activity in the insulae. However, with our non-invasive analysis, we can tell that the source of the posterior-temporal δ brushes, independently of what triggered it, is in the insulae, in the same way that spindle bursts have their source in, for example, the barrel cortex, even if they are driven by the thalamus or by the sensory periphery.

We have now replace the word “generator” with “source” or “location” when referring to human studies and we have added a new paragraph in the Results and Discussion to make this difference explicit (fifth paragraph).

2) How do the authors exclude that the insula is just co-activated by the activity patterns generated in the retina, spinal cord etc.?

As you correctly assert, by nature of our non-invasive methodology and the constraints of studying fragile preterm human infants, we cannot definitively rule out that the insulae were co-activated by peripherally generated activity. However, numerous imaging studies with this population have shown that different forms of sensory stimulation induce activity which can be reliably localized to their respective primary sensory cortices. These include visual responses in the occipital lobe (Lee, Donner et al. 2012), auditory responses in the temporal lobe (Mahmoudzadeh, Dehaene-Lambertz et al. 2013, Baldoli, Scola et al. 2015), and passive movement responses in the primary sensori-motor cortices (Arichi, Moraux et al. 2010, Allievi, Arichi et al. 2016). We therefore feel it is unlikely that the activation we identified could have been generated in peripheral regions (such as the retina) as across all of these studies, additional patterns of activation were never described within the insulae.

However, we agree that there remains uncertainty about the origin of the recorded events as discussed in the new Results and Discussion paragraph (fifth paragraph).

3)Results and Discussion, second paragraph: What is the evidence that the insula in newborn rodent cortex or preterm primate neocortex is so well connected and may fulfil the role of a hub region?

Thank you for raising this point, we agree that it needs clarification. The insulae have a dense network of connections and are thought to play a key integrative role in sensory and behavioural processing in primates and even rodents (where it is not operculized and therefore its gross anatomical appearance is very different) (Butti and Hof 2010, Mars, Sallet et al. 2013, Miranda-Dominguez, Mills et al. 2014, Zingg, Hintiryan et al. 2014). Although there is evidence across these species that neurogenesis and cortical maturation is more advanced in the insulae in comparison to the surrounding cortex in early life (Kroenke, Van Essen et al. 2007, Huang, Yamamoto et al. 2008), to our knowledge detailed characterization of the development of regional cortical connectivity has never been done.

Given that the aim of our study was to source localize EEG bursting events in human preterm infants, our discussion was intended only to reflect the existing literature which supports that the insulae are key hub regions specifically in human infants (Gao, Gilmore et al. 2011, Ball, Aljabar et al. 2014). As similar developmental work has not been done in animals, we agree that it would be inappropriate to interpret our results in that context.

We have now clarified in the manuscript that our Results and Discussion with respect to the insula apply only to our study population of human preterm infants and have emphasized that further work is clearly needed to understand how bursting events may relate to the development of insula connectivity in animals (second paragraph).

4) In newborn rodent cortex and some EEG studies on preterm human cortex, spindle bursts are initially local (columnar) events, which at later stages propagate to neighbouring regions (e.g. papers by JW Yang et al). What is the developmental pattern in the present manuscript?

While developmental changes in the propagation properties of spontaneous and evoked neuronal events have been well characterised in rodents, very little is known about this in the human preterm brain. Using fMRI, we have previously shown that the spatial distribution of evoked and spontaneous somatosensory hemodynamic activity in human infants undergoes similar changes to those described by your group in newborn rodents (Doria, Beckmann et al. 2010, Allievi, Arichi et al. 2016). The activity is initially localised in a small area of the primary somatosensory cortex, which at later stages propagates to other brain regions (neighbouring and not) in the same and opposite hemisphere resulting in a more adult-like activation pattern. Concurrently, the relative incidence of spontaneous bilateral neuronal bursts appears to increase over the equivalent period to the last trimester of human gestation (Lombroso 1979, Anderson, Torres et al. 1985, Koolen, Dereymaeker et al. 2016), before plateauing in the immediate perinatal period (Koolen, Dereymaeker et al. 2014). On the other hand, specific regional δ brush propagation has never been characterised, potentially because it occurs at an anatomical and functional scale that cannot be resolved with low-density EEG. Therefore, this very interesting point unfortunately cannot be addressed with the present sample population and experimental set-up as changes in spontaneous activity propagation happen over a significantly longer post-menstrual age window (over several weeks in humans) than the one studied here and at an extremely refined spatial level.

We have now added a sentence to highlight the need of longer longitudinal studies to address this point (Results and Discussion, sixth paragraph).

5)Results and Discussion, third paragraph: The subplate certainly does not drive the spindle bursts in vivo, it acts as a relay or amplifier for the activity coming from subcortical inputs.

Thank you for pointing this out. We have now clarified that even though the subplate has been shown to be necessary for the generation of the spindle bursts, there is no evidence to suggest that it drives them in the new Discussion paragraph (fifth paragraph). We have also removed this sentence: “The subplate is thought to drive spindle burst activity in animals and its ablation results in loss of bursting events and permanent disruption of cortical maps” and rephrased the preceding sentence as: “This finding is of particular significance as these are regions where the subplate, a transient structure which is thought to play a fundamental role in the generation of spindle burst activity in animals (Tolner, Sheikh et al. 2012),…”

6) How can the authors identify the subplate? Does the spatial resolution of the methods allow this major conclusion (Results and Discussion, last paragraph)?

The subplate can be easily appreciated visually on high resolution T2-weighted structural MR images (Vasung, Lepage et al. (2016), Widjaja, Geibprasert et al. (2010) and also see Figure 2—figure supplement 2 for examples from our study), but it is currently not possible to clearly delineate its boundaries (Kostovic and Rakic 1990, Kostovic, Jovanov-Milosevic et al. 2014). As a result, it has never been precisely mapped onto a template image such as that used for our group analysis. Our suggestion that the temporal poles represent a key area where the subplate remains prominent during our study period is therefore based on qualitative MRI appearances which have been validated by post-mortem histological human studies (Kostovic and Jovanov-Milosevic 2006).

As it was not possible to definitively localize our identified clusters of activity to the subplate onto a population template in the group analysis, our intention in this manuscript was only to highlight that the activity was seen in a region where the subplate can be qualitatively appreciated in individual subjects in this period. In future work, we will aim to study this important issue in more detail by extending our study population to include more infants at a younger PMA and incorporating methods which may allow clearer delineation of the subplate on a group level.

We have now clarified that the activity was localized in a region where the subplate is prominent (Results and Discussion, fourth paragraph and subsection “EEG-fMRI acquisition”). We have also included Figure 2—figure supplement 2 and added a sentence explaining the need for further work to specifically answer this question (Results and Discussion, sixth paragraph.

7) The authors only studied "spontaneous" activity? Is the insula involved in evoked δ brush activity?

This is currently unknown as no studies of sensory evoked activity with combined EEG and fMRI have ever been conducted in preterm neonates. However, we know from our own work and that of others that (i) somatosensory (Allievi, Arichi et al. 2016), auditory (Baldoli, Scola et al. 2015) or visual stimulation (Lee, Donner et al. 2012) does not evoke hemodynamic activity in the preterm insulae and that (ii) sensory evoked δ brushes seem more prevalent over the relevant sensory areas (Milh, Kaminska et al. 2007, Colonnese, Kaminska et al. 2010, Fabrizi, Slater et al. 2011, Chipaux, Colonnese et al. 2013). Taken together these results suggest that the insulae are not involved in sensory evoked δ brushes.

This is addressed in the new Results and Discussion paragraph (fifth paragraph).

8) Did the authors observe in their EEG recordings regional differences in the δ brush properties?

We did not visually observe any differences, however we are planning a quantitative analysis in a future EEG only experiment as there is a single report that shows regional differences in the frequency content of δ brushes in very preterm infants (up to 32 weeks PMA) (Anderson, Torres et al. 1985). For the present paper, only the time of occurrence and topography of a δ brush episode was necessary to build the explanatory variables for the fMRI analysis. For example, if a pericentral δ brush had a different frequency power spectrum than a posterior-temporal δ brush, this would not have influenced the current analysis as the two would have been already separated in two different explanatory variables because of their topography.

We added a new sentence proposing to address this point as future work (Results and Discussion, sixth paragraph).

https://doi.org/10.7554/eLife.27814.012

Article and author information

Author details

  1. Tomoki Arichi

    1. Centre for the Developing Brain, Division of Imaging Sciences and Biomedical Engineering, King’s College London, London, United Kingdom
    2. Department of Bioengineering, Imperial College London, London, United Kingdom
    Contribution
    Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Visualization, Methodology, Writing—original draft, Project administration, Writing—review and editing, Data acquisition
    Competing interests
    No competing interests declared
    ORCID icon 0000-0002-3550-1644
  2. Kimberley Whitehead

    Department of Neuroscience, Physiology and Pharmacology, University College London, London, United Kingdom
    Contribution
    Formal analysis, Writing—review and editing, Data acquisition
    Competing interests
    No competing interests declared
  3. Giovanni Barone

    1. Centre for the Developing Brain, Division of Imaging Sciences and Biomedical Engineering, King’s College London, London, United Kingdom
    2. Department of Pediatrics, Catholic University of Sacred Heart, Rome, Italy
    Contribution
    Formal analysis, Writing—review and editing, Data aquisition
    Competing interests
    No competing interests declared
  4. Ronit Pressler

    Clinical Neurosciences, UCL-Institute of Child Health, London, United Kingdom
    Contribution
    Formal analysis, Writing—review and editing
    Competing interests
    No competing interests declared
  5. Francesco Padormo

    1. Centre for the Developing Brain, Division of Imaging Sciences and Biomedical Engineering, King’s College London, London, United Kingdom
    2. Translational and Molecular Imaging Institute, Icahn School of Medicine at Mount Sinai, New York, United States
    Contribution
    Methodology, Writing—review and editing
    Competing interests
    No competing interests declared
  6. A David Edwards

    1. Centre for the Developing Brain, Division of Imaging Sciences and Biomedical Engineering, King’s College London, London, United Kingdom
    2. Department of Bioengineering, Imperial College London, London, United Kingdom
    Contribution
    Conceptualization, Resources, Funding acquisition, Writing—review and editing
    For correspondence
    ad.edwards@kcl.ac.uk
    Competing interests
    No competing interests declared
    ORCID icon 0000-0003-4801-7066
  7. Lorenzo Fabrizi

    Department of Neuroscience, Physiology and Pharmacology, University College London, London, United Kingdom
    Contribution
    Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Visualization, Methodology, Writing—original draft, Project administration, Writing—review and editing, Data acquisition
    For correspondence
    l.fabrizi@ucl.ac.uk
    Competing interests
    No competing interests declared
    ORCID icon 0000-0002-9582-0727

Funding

Medical Research Council (MR/L019248/1)

  • Kimberley Whitehead
  • Lorenzo Fabrizi

National Institute for Health Research

  • Tomoki Arichi

Academy of Medical Sciences

  • Tomoki Arichi

Medical Research Council (MR/P008712/1)

  • Tomoki Arichi

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

Acknowledgements

The authors acknowledge support from the Department of Health via the National Institute for Health Research (NIHR) comprehensive Biomedical Research Centre award to Guy’s and St Thomas’ NHS Foundation Trust in partnership with King’s College London and King’s College Hospital NHS Foundation Trust. TA was supported by an Academic Clinical Lectureship from the NIHR, a Starter Grant from the Academy of Medical Sciences (AMS) and a Medical Research Council (MRC) Clinician Scientist Fellowship (MR/P008712/1). LF and KW were supported by a MRC Career Development Award (MR/L019248/1). The authors also thank Professor Maria Fitzgerald, Professor Jo Hajnal, and Dr Robert Störmer for invaluable discussion and technical support throughout the study. We are also extremely grateful to the patients and families who participated in the study.

Ethics

Human subjects: Informed consent, and consent to publish, was obtained from the parents of all subjects enrolled in the study. National Research Ethics Committee approval was obtained from the West London REC (12/LO/1247). All of the research methods conformed to the standard set by the Declaration of Helsinki.

Reviewing Editor

  1. Sabine Kastner, Reviewing Editor, Princeton University, United States

Publication history

  1. Received: April 15, 2017
  2. Accepted: August 3, 2017
  3. Version of Record published: September 12, 2017 (version 1)

Copyright

© 2017, Arichi 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

  • 469
    Page views
  • 76
    Downloads
  • 0
    Citations

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

Comments

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)

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

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

Further reading

    1. Cell Biology
    2. Genes and Chromosomes
    Wahid A Mulla et al.
    Research Article Updated