The transition from adolescence to adulthood is a period when ongoing brain development coincides with a substantially increased risk of psychiatric disorders. The developmental brain changes accounting for this emergent psychiatric symptomatology remain obscure. Capitalizing on a unique longitudinal dataset that includes in vivo myelin-sensitive magnetization transfer (MT) MRI scans, we show that this developmental period is characterized by brain-wide growth in MT trajectories within both gray matter and adjacent juxtacortical white matter. In this healthy population, the expression of common developmental traits, namely compulsivity and impulsivity, is tied to a reduced growth of these MT trajectories in frontostriatal regions. This reduction is most marked in dorsomedial and dorsolateral prefrontal regions for compulsivity and in lateral and medial prefrontal regions for impulsivity. These findings highlight that psychiatric traits of compulsivity and impulsivity are linked to regionally specific reductions in myelin-related growth in late adolescent brain development.

Access optionsAccess options

Rent or Buy article

Get time limited or full article access on ReadCube.


All prices are NET prices.

Data availability

Whole-brain results are available for inspection online on Neurovault (https://neurovault.org/collections/YAHZLJRW/). Data for this specific paper have been uploaded to the Cambridge Data Repository (https://doi.org/10.17863/CAM.12959) and password protected. Our participants did not give informed consent for their measures to be made publicly available, and it is possible that they could be identified from this dataset. Access to the data supporting the analyses presented in this paper will be made available to researchers with a reasonable request to openNSPN@medschl.cam.ac.uk or the corresponding authors (G.Z. and T.U.H.).

Code availability

The custom-made SPM pipeline code for longitudinal VBM and VBQ processing is provided along with the manuscript at https://github.com/gabrielziegler/gz/tree/master/nspn_mpm_prepro_code_and_example. The code aims to be transparent regarding its applied procedures, but is not intended for clinical use. It is free, but is copyrighted software distributed under the terms of the GNU General Public Licence as published by the Free Software Foundation (either version 2, or at your option, any later version). For any questions and requests please contact G.Z.

Additional information

Journal peer review information: Nature Neuroscience thanks Lars Tjelta Westlye and other anonymous reviewer(s) for their contribution to the peer review of this work.

Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.


  1. 1.

    Gogtay, N. et al. Dynamic mapping of human cortical development during childhood through early adulthood. Proc. Natl Acad. Sci. USA 101, 8174–8179 (2004).

  2. 2.

    Sowell, E. R., Thompson, P. M., Holmes, C. J., Jernigan, T. L. & Toga, A. W. In vivo evidence for post-adolescent brain maturation in frontal and striatal regions. Nat. Neurosci. 2, 859–861 (1999).

  3. 3.

    Paus, T. Growth of white matter in the adolescent brain: myelin or axon? Brain Cogn. 72, 26–35 (2010).

  4. 4.

    Miller, D. J. et al. Prolonged myelination in human neocortical evolution. Proc. Natl Acad. Sci. USA 109, 16480–16485 (2012).

  5. 5.

    Perrin, J. S. et al. Growth of white matter in the adolescent brain: role of testosterone and androgen receptor. J. Neurosci. 28, 9519–9524 (2008).

  6. 6.

    Petanjek, Z. et al. Extraordinary neoteny of synaptic spines in the human prefrontal cortex. Proc. Natl Acad. Sci. USA 108, 13281–13286 (2011).

  7. 7.

    Whitaker, K. J. et al. Adolescence is associated with genomically patterned consolidation of the hubs of the human brain connectome. Proc. Natl Acad. Sci. USA 113, 9105–9110 (2016).

  8. 8.

    Foulkes, L. & Blakemore, S.-J. Studying individual differences in human adolescent brain development. Nat. Neurosci. 21, 315–323 (2018).

  9. 9.

    Kessler, R. C. et al. Lifetime prevalence and age-of-onset distributions of mental disorders in the World Health Organization’s World Mental Health Survey Initiative. World Psychiatry 6, 168–176 (2007).

  10. 10.

    Paus, T., Keshavan, M. & Giedd, J. N. Why do many psychiatric disorders emerge during adolescence? Nat. Rev. Neurosci. 9, 947–957 (2008).

  11. 11.

    McCarthy, H. et al. Attention network hypoconnectivity with default and affective network hyperconnectivity in adults diagnosed with attention-deficit/hyperactivity disorder in childhood. JAMA Psychiatry 70, 1329–1337 (2013).

  12. 12.

    Douaud, G. et al. A common brain network links development, aging, and vulnerability to disease. Proc. Natl Acad. Sci. USA 111, 17648–17653 (2014).

  13. 13.

    Robbins, T. W., Gillan, C. M., Smith, D. G., de Wit, S. & Ersche, K. D. Neurocognitive endophenotypes of impulsivity and compulsivity: towards dimensional psychiatry. Trends Cogn. Sci. 16, 81–91 (2012).

  14. 14.

    de Wit, S. J. et al. Multicenter voxel-based morphometry mega-analysis of structural brain scans in obsessive–compulsive disorder. Am. J. Psychiatry 171, 340–349 (2014).

  15. 15.

    Norman, L. J. et al. Structural and functional brain abnormalities in attention-deficit/hyperactivity disorder and obsessive–compulsive disorder: a comparative meta-analysis. JAMA Psychiatry 73, 815–825 (2016).

  16. 16.

    Carlisi, C. O. et al. Comparative multimodal meta-analysis of structural and functional brain abnormalities in autism spectrum disorder and obsessive–compulsive disorder. Biol. Psychiatry 82, 83–102 (2016).

  17. 17.

    Weiskopf, N., Mohammadi, S., Lutti, A. & Callaghan, M. F. Advances in MRI-based computational neuroanatomy: from morphometry to in-vivo histology. Curr. Opin. Neurol. 28, 313–322 (2015).

  18. 18.

    Schmierer, K., Scaravilli, F., Altmann, D. R., Barker, G. J. & Miller, D. H. Magnetization transfer ratio and myelin in postmortem multiple sclerosis brain. Ann. Neurol. 56, 407–415 (2004).

  19. 19.

    Turati, L. et al. In vivo quantitative magnetization transfer imaging correlates with histology during de- and remyelination in cuprizone-treated mice. NMR Biomed. 28, 327–337 (2015).

  20. 20.

    Callaghan, M. F., Helms, G., Lutti, A., Mohammadi, S. & Weiskopf, N. A general linear relaxometry model of R1 using imaging data. Magn. Reson. Med. 73, 1309–1314 (2015).

  21. 21.

    Campbell, J. S. W. et al. Promise and pitfalls of g-ratio estimation with MRI. NeuroImage 182, 80–96 (2018).

  22. 22.

    Raz, N. & Lindenberger, U. Only time will tell: cross-sectional studies offer no solution to the age–brain–cognition triangle: comment on Salthouse (2011). Psychol. Bull. 137, 790–795 (2011).

  23. 23.

    Paus, T. Mapping brain maturation and cognitive development during adolescence. Trends Cogn. Sci. 9, 60–68 (2005).

  24. 24.

    Natu, V. S. et al. Apparent thinning of visual cortex during childhood is associated with myelination, not pruning. Preprint at bioRxiv https://doi.org/10.1101/368274 (2018).

  25. 25.

    Burns, G. L., Keortge, S. G., Formea, G. M. & Sternberger, L. G. Revision of the padua inventory of obsessive compulsive disorder symptoms: distinctions between worry, obsessions, and compulsions. Behav. Res. Ther. 34, 163–173 (1996).

  26. 26.

    Foa, E. B. et al. The obsessive–compulsive inventory: development and validation of a short version. Psychol. Assess. 14, 485–496 (2002).

  27. 27.

    Haber, S. N. Corticostriatal circuitry. Dialogues Clin. Neurosci. 18, 7–21 (2016).

  28. 28.

    Palminteri, S. & Chevallier, C. Can we infer inter-individual differences in risk-taking from behavioral tasks? Front. Psychol. 9, 2307 (2018).

  29. 29.

    Pedroni, A. et al. The risk elicitation puzzle. Nat. Hum. Behav. 1, 803 (2017).

  30. 30.

    Frey, R., Pedroni, A., Mata, R., Rieskamp, J. & Hertwig, R. Risk preference shares the psychometric structure of major psychological traits. Sci. Adv. 3, e1701381 (2017).

  31. 31.

    Moutoussis, M. et al. Change, stability, and instability in the Pavlovian guidance of behaviour from adolescence to young adulthood. PLoS Comput. Biol. 14, e1006679 (2018).

  32. 32.

    Shahar, N. et al. Improving the reliability of model-based decision-making estimates in the two-stage decision task with reaction-times and drift-diffusion modeling. PLoS Comput. Biol. 15, e1006803 (2019).

  33. 33.

    Kiddle, B. et al. Cohort profile: The NSPN 2400 Cohort: a developmental sample supporting the Wellcome Trust NeuroScience in Psychiatry Network. Int. J. Epidemiol. 47, 18–19g (2018).

  34. 34.

    Moutoussis, M., Bentall, R. P., El-Deredy, W. & Dayan, P. Bayesian modelling of Jumping-to-Conclusions bias in delusional patients. Cognit. Neuropsychiatry 16, 422–447 (2011).

  35. 35.

    Virchow, R. Ueber das ausgebreitete Vorkommen einer dem Nervenmark analogen Substanz in den thierischen Geweben. Arch. Für Pathol. Anat. Physiol. Für Klin. Med. 6, 562–572 (1854).

  36. 36.

    Bunge, R. P. Glial cells and the central myelin sheath. Physiol. Rev. 48, 197–251 (1968).

  37. 37.

    Holmes, A. J. & Patrick, L. M. The myth of optimality in clinical neuroscience. Trends Cogn. Sci. 22, 241–257 (2018).

  38. 38.

    Rubia, K. ‘Cool’ inferior frontostriatal dysfunction in attention-deficit/hyperactivity disorder versus ‘hot’ ventromedial orbitofrontal-limbic dysfunction in conduct disorder: a review. Biol. Psychiatry 69, e69–e87 (2011).

  39. 39.

    Hauser, T. U. et al. Role of the medial prefrontal cortex in impaired decision making in juvenile attention-deficit/hyperactivity disorder. JAMA Psychiatry 71, 1165–1173 (2014).

  40. 40.

    Hauser, T. U. et al. Increased fronto-striatal reward prediction errors moderate decision making in obsessive–compulsive disorder. Psychol. Med. 47, 1246–1258 (2017).

  41. 41.

    Gillan, C. M. et al. Functional neuroimaging of avoidance habits in obsessive–compulsive disorder. Am. J. Psychiatry 172, 284–293 (2015).

  42. 42.

    Dougherty, D. D. et al. Prospective long-term follow-up of 44 patients who received cingulotomy for treatment-refractory obsessive–compulsive disorder. Am. J. Psychiatry 159, 269–275 (2002).

  43. 43.

    Figee, M. et al. Deep brain stimulation restores frontostriatal network activity in obsessive–compulsive disorder. Nat. Neurosci. 16, 386–387 (2013).

  44. 44.

    Boedhoe, P. S. W. et al. Distinct subcortical volume alterations in pediatric and adult OCD: a worldwide meta- and mega-analysis. Am. J. Psychiatry 174, 60–69 (2016).

  45. 45.

    Whelan, R. et al. Adolescent impulsivity phenotypes characterized by distinct brain networks. Nat. Neurosci. 15, 920–925 (2012).

  46. 46.

    Holmes, A. J., Hollinshead, M. O., Roffman, J. L., Smoller, J. W. & Buckner, R. L. Individual differences in cognitive control circuit anatomy link sensation seeking, impulsivity, and substance use. J. Neurosci. 36, 4038–4049 (2016).

  47. 47.

    Pingault, J.-B. et al. Using genetic data to strengthen causal inference in observational research. Nat. Rev. Genet. 19, 566–580 (2018).

  48. 48.

    Lerch, J. P. et al. Studying neuroanatomy using MRI. Nat. Neurosci. 20, 314–326 (2017).

  49. 49.

    Franklin, R. J. M. & Ffrench-Constant, C. Remyelination in the CNS: from biology to therapy. Nat. Rev. Neurosci. 9, 839–855 (2008).

  50. 50.

    Sampaio-Baptista, C. et al. Motor skill learning induces changes in white matter microstructure and myelination. J. Neurosci. 33, 19499–19503 (2013).

  51. 51.

    Patton, J. H., Stanford, M. S. & Barratt, E. S. Factor structure of the Barratt impulsiveness scale. J. Clin. Psychol. 51, 768–774 (1995).

  52. 52.

    Weiskopf, N. et al. Quantitative multi-parameter mapping of R1, PD*, MT, and R2* at 3T: a multi-center validation. Front. Neurosci. 7, 95 (2013).

  53. 53.

    Callaghan, M. F. et al. Widespread age-related differences in the human brain microstructure revealed by quantitative magnetic resonance imaging. Neurobiol. Aging 35, 1862–1872 (2014).

  54. 54.

    Draganski, B. et al. Regional specificity of MRI contrast parameter changes in normal ageing revealed by voxel-based quantification (VBQ). NeuroImage 55, 1423–1434 (2011).

  55. 55.

    Jones, D. K., Knösche, T. R. & Turner, R. White matter integrity, fiber count, and other fallacies: the do’s and don’ts of diffusion MRI. NeuroImage 73, 239–254 (2013).

  56. 56.

    Donahue, C. J., Glasser, M. F., Preuss, T. M., Rilling, J. K. & Van Essen, D. C. Quantitative assessment of prefrontal cortex in humans relative to nonhuman primates. Proc. Natl Acad. Sci. USA 115, E5183–E5192 (2018).

  57. 57.

    Glasser, M. F. et al. A multi-modal parcellation of human cerebral cortex. Nature 536, 171–178 (2016).

  58. 58.

    Ashburner, J. & Ridgway, G. R. Symmetric diffeomorphic modeling of longitudinal structural MRI. Front. Neurosci. 6, 197 (2012).

  59. 59.

    Ashburner, J. A fast diffeomorphic image registration algorithm. NeuroImage 38, 95–113 (2007).

  60. 60.

    Ashburner, J. & Friston, K. J. Diffeomorphic registration using geodesic shooting and Gauss–Newton optimisation. NeuroImage 55, 954–967 (2011).

  61. 61.

    Neuhaus, J. M. & Kalbfleisch, J. D. Between- and within-cluster covariate effects in the analysis of clustered data. Biometrics 54, 638–645 (1998).

  62. 62.

    Hoffman, L., Hofer, S. M. & Sliwinski, M. J. On the confounds among retest gains and age-cohort differences in the estimation of within-person change in longitudinal studies: a simulation study. Psychol. Aging 26, 778–791 (2011).

  63. 63.

    Sliwinski, M., Hoffman, L. & Hofer, S. M. Evaluating convergence of within-person change and between-person age differences in age-heterogeneous longitudinal studies. Res. Hum. Dev. 7, 45–60 (2010).

  64. 64.

    Lash, T. L., Fox, M. P. & Fink, A. K. Applying Quantitative Bias Analysis to Epidemiologic Data (Springer, 2009).

  65. 65.

    Guillaume, B. et al. Fast and accurate modelling of longitudinal and repeated measures neuroimaging data. NeuroImage 94, 287–302 (2014).

  66. 66.

    Personal and Household Finances https://www.ons.gov.uk/peoplepopulationandcommunity/personalandhouseholdfinances/ (Office for National Statistics, accessed 17 October 2018).

  67. 67.

    Ashburner, J. & Friston, K. J. Voxel-based morphometry—the methods. NeuroImage 11, 805–821 (2000).

  68. 68.

    Gelman, A. et al. Bayesian Data Analysis 3rd edn (Chapman and Hall/CRC, 2013).

Download references


A Wellcome Trust Cambridge–UCL Mental Health and Neurosciences Network grant (095844/Z/11/Z) supported this work. R.J.D. holds a Wellcome Trust Investigator Award (098362/Z/12/Z). The UCL–Max Planck Centre is a joint initiative supported by UCL and the Max Planck Society. T.U.H. is supported by a Wellcome Sir Henry Dale Fellowship (211155/Z/18/Z), a grant from the Jacobs Foundation, the Medical Research Foundation, and a 2018 NARSAD Young Investigator grant (27023) from the Brain & Behavior Research Foundation. M.M. receives support from the UCLH NIHR BRC. P.F. is in receipt of a National Institute for Health Research (NIHR) Senior Investigator Award (NF-SI-0514-10157), and was in part supported by the NIHR Collaboration for Leadership in Applied Health Research and Care North Thames at Barts Health NHS Trust. E.T.B. is in receipt of a NIHR Senior Investigator Award, and was in part supported by the NIHR Cambridge Biomedical Research Centre. The Wellcome Centre for Human Neuroimaging is supported by core funding from the Wellcome Trust (203147/Z/16/Z). The authors thank R. Davis and FIL IT support for making large sample analysis feasible and more efficient. Thanks are also given to G. Prabhu, and to specific experts for input in relation to applied and technical methods, particularly R. Dahnke, W. Penny, G. Ridgway, M. Callaghan, N. Weiskopf, B. Draganski, J. Ashburner, C. Gaser, G. Flandin, T. Nichols, B. Guillaume, J. Bernal-Rusiel, M. Völkle, C. Driver, A. Brandmeier, F. Dick, M. Betts, G. J. Will and R. Kievit. Finally, G.Z. thanks E. Düzel for support at the DZNE. The views expressed are those of the authors and not necessarily those of the NHS, the NIHR or the Department of Health and Social Care.

Author information

Author notes

  1. These authors contributed equally: Gabriel Ziegler, Tobias U. Hauser.

  2. A list of members appears in the Supplementary Note.


  1. Max Planck University College London Centre for Computational Psychiatry and Ageing Research, London, UK

    • Gabriel Ziegler
    • , Tobias U. Hauser
    • , Michael Moutoussis
    • , Ulman Lindenberger
    •  & Raymond J. Dolan
  2. Wellcome Centre for Human Neuroimaging, University College London, London, UK

    • Gabriel Ziegler
    • , Tobias U. Hauser
    • , Michael Moutoussis
    •  & Raymond J. Dolan
  3. Institute of Cognitive Neurology and Dementia Research, Otto-von-Guericke-University Magdeburg, Magdeburg, Germany

    • Gabriel Ziegler
  4. German Center for Neurodegenerative Diseases (DZNE), Magdeburg, Germany

    • Gabriel Ziegler
  5. Department of Psychiatry, University of Cambridge, Cambridge, UK

    • Edward T. Bullmore
    • , Ian M. Goodyer
    •  & Peter B. Jones
  6. Cambridgeshire and Peterborough National Health Service Foundation Trust, Cambridge, UK

    • Edward T. Bullmore
    • , Ian M. Goodyer
    •  & Peter B. Jones
  7. Medical Research Council/Wellcome Trust Behavioural and Clinical Neuroscience Institute, University of Cambridge, Cambridge, UK

    • Edward T. Bullmore
  8. ImmunoPsychiatry, GlaxoSmithKline Research and Development, Stevenage, UK

    • Edward T. Bullmore
  9. Research Department of Clinical, Educational and Health Psychology, University College London, London, UK

    • Peter Fonagy
  10. Center for Lifespan Psychology, Max Planck Institute for Human Development, Berlin, Germany

    • Ulman Lindenberger


  1. Search for Gabriel Ziegler in:

  2. Search for Tobias U. Hauser in:

  3. Search for Michael Moutoussis in:

  4. Search for Edward T. Bullmore in:

  5. Search for Ian M. Goodyer in:

  6. Search for Peter Fonagy in:

  7. Search for Peter B. Jones in:

  8. Search for Ulman Lindenberger in:

  9. Search for Raymond J. Dolan in:


  1. NSPN Consortium


    E.T.B., I.M.G., P.F., P.B.J., NSPN Consortium members, M.M., T.U.H. and R.J.D. designed the experiments. G.Z., T.U.H. and NSPN Consortium members performed the experiments and analyzed the data. G.Z., T.U.H., U.L. and R.J.D. wrote the paper.

    Competing interests

    E.T.B. is employed half of the time by the University of Cambridge and half of the time by GlaxoSmithKline and holds stock in GlaxoSmithKline. All other authors declare no competing interests.

    Corresponding authors

    Correspondence to Gabriel Ziegler or Tobias U. Hauser.

    Integrated supplementary information

    1. Supplementary Fig. 1 Compulsivity and impulsivity dimensions.

      (a) In our sample, we had two questionnaires available that measured obsessive-compulsive symptoms: revised Obsessive-Compulsive Inventory (OCI-R, mean ± std: 7.33 ± 7.76, range: 0–50; Foa et al., 2002, Psychol Assess) and Padua Inventory Washington State University Revision (PI-WSUR, 17.09 ± 17.00, range: 0 - 106.5; Burns et al., 1996, Behav Res Ther). To build a general compulsivity composite score, we conducted an item-level principal component analysis (n = 1543 independent subjects in all panels) and used the first principal component thereof (PC1), in a similar approach as previous studies (Gillan et al., 2016, eLife; Rouault et al., 2018, Biol Psychiatry). Scree plot of PCA presented in (b) and the highest loading items of PC1 are shown in (c). This novel compulsivity component correlated highly with the total scores of both OCI-R (d) and PI-WSUR (e). Impulsivity differences and changes were assessed using BIS (Barratt Impulsiveness Scale total score, 64.19 ± 6.81, range: 39.7 - 114.5). We found that compulsivity and impulsivity were largely independent, sharing 1.4% of common variance (f). Pearson’s correlation coefficients were provided in panels a, d-f. In addition, both dimensions were only weakly associated with other psychiatric dimensions, such as anxiety (RCMAS total; Kiddle et al., 2017, Int J Epidemiol; impulsivity r = 10, compulsivity r = 35) or depression (MFQ total; Kiddle et al., 2017, Int J Epidemiol; impulsivity r = 06, compulsivity r = .33) thus showing relatively weak associations compared to other psychiatric dimensions (for example, correlation depression and anxiety r = .88) suggesting that our results are more specific to compulsivity and impulsivity than to a general psychopathology (Alnaes et al., 2018, JAMA Psychiatry). Notably, after correction for MFQ differences across individuals, effects presented in main results were still found (data not shown). Linear-mixed effects (LME) modelling (not shown, cf. methods and supplementary information) was applied to total scores of PI-WSUR and BIS, for which longitudinal follow-ups were available in the larger behavioural NSPN sample (for PI-WSUR (BIS): 533 (805) observations from 291 (305) subjects).

    2. Supplementary Fig. 2 Image processing and analysis workflow.

      (a) Overview of applied longitudinal qMRI preprocessing workflow using SPM (and CAT toolbox, r1318) to normalize in standardized space (within-subject: Ashburner & Ridgway, 2012, Front Neurosci; between-subject: geodesic shooting, Ashburner & Friston, 2009, NeuroImage; 2011, NeuroImage) and smoothing (via tissue-weighted smoothing: Draganski et al., 2011, NeuroImage) of myelin-sensitive magnetization transfer (MT) saturation maps. First, we performed symmetric diffeomorphic registration for longitudinal MRI (Ashburner & Ridgway, 2012, Front Neurosci), combining nonlinear diffeomorphic and rigid-body registration and a correction for intensity inhomogeneity artefacts. The registration model creates a midpoint image for each subject and the corresponding deformation fields for every individual scan. Second, we applied CAT segmentation to each subject’s midpoint image, which assumes every voxel to be drawn from an unknown mixture of gray matter (GM), white matter (WM), and cerebrospinal fluid (CSF) tissue classes. The applied segmentation algorithm performs a partial volume estimation (PVE) to account for mixed voxels with two tissue types. Third, nonlinear template generation and image registration to MNI space was performed using the individual midpoint GM and WM tissue maps and diffeomorphic registration using geodesic shooting (Ashburner & Friston, 2011, NeuroImage). Quantitative MT maps from all time-points were normalized to the MNI space using (within- and between-subjects) transformations obtained in previous steps. Fourth, since our study aimed at local statistical analysis of voxel-based trajectories of quantitative MT parameters within GM and WM tissue classes, the normalized MT maps were smoothed using previously established (Draganski et al., 2011, NeuroImage) tissue-weighted-smoothing with a Gaussian kernel of 6 mm full width at half maximum (FWHM). Fifth, in order to avoid biasing results by image artefacts (for example due to movement, segmentation or normalization errors), the MPM maps were carefully checked manually before and after longitudinal registration by an expert. Additionally, the data was quality checked using a during scan motion proxy (Castella et al., 2018, Magn Reson Med; cf. Supplementary Fig. 8) and statistical covariance-based sample inhomogeneity measures (as implemented in the CAT toolbox) to exclude subjects with extremal overall deviation of quantitative values due to acquisition or processing artefacts. All processed quantitative MT and morphometric maps for subsequent longitudinal modelling steps were obtained using the above steps. (b) Schematic illustration of longitudinal voxel-based qMRI analysis in relation to developmental effects and compulsivity/impulsivity dimensions using SPM Sandwich Estimator toolbox r1.2.10 (SwE, Guillaume et al., 2014, NeuroImage). For more compact visualization of results, projections of template-space statistical (non-) parametric z-maps were performed on surface reconstruction of the same voxel-based template (gray and white matter tissue classes) used for normalization. (c) Comparing SwE toolbox output for parametric and non-parametric results for effect of developmental growth of MT within cortical regions (main results Fig. 1). Top row shows parametric inference correcting for multiple comparisons using voxelwise/peak-level FDR (p < .05, statistical Z-maps from one-sided Wald-test for MT increase with time/visits or mean age of subject). Middle row illustrates same FDR corrected results obtained using SwE toolbox under wild bootstrapping with 999 straps. Bottom row applies clusterwise FWE correction extent threshold k = 1147 (p = .05) with initial cluster forming threshold .001. The differences of FWE cluster and FDR voxelwise correction are expected, generally results were found to be similar in our ~500 scan sample suggesting validity of parametric inference scheme. SwE TFCE-based inference was found to be more sensitive than FWE (not shown).

    3. Supplementary Fig. 3 Longitudinal and cross-sectional increase, sexual dimorphism and nonlinearities of the myelin marker during transitioning into adulthood.

      (a) We show longitudinal increase of myelin-sensitive MT over study visits/time, from baseline to follow-up (mean (std) 1.3 (0.32) years). We observed more emphasized change of MT especially in cingulate cortex (max z-value voxel [z = 6.03, p = .004, voxelwise FDR, p < .001, clusterwise FWE] in posterior cingulate [MNI: 13 −48 30], n = 497/288 scans/subjects apply for panels a-d, 51.7% female) and lateral fronto-parieto-temporal gray matter regions. All statistical maps in panels a-d were obtained from one-sided Wald-tests using longitudinal sandwich estimator for Voxel-Based-Quantification analysis (VBQ). Here, surface projections for Z-maps (p < .05, voxelwise FDR) are shown, testing for positive main effect of study visit/time. (b) We show MT increases with higher age_mean of participants (14–25 years) in widespread cortical areas (max z-value [z = 7.33, p = .005, voxelwise FDR, p = .001 clusterwise FWE] in posterior superior temporal gyrus [63 −36 11]). Here age_mean refers to mean of each individual’s age over all visits of the longitudinal study. The effect of age_mean is independent from the effect of changes over time/visits and estimated separately, while the former is purely cross-sectional and the latter is purely longitudinal. We find that cross-sectional effects of age_mean are linked to local MT differences in fronto-parieto-temporal cortical gray matter. Notably, the observed differences of longitudinal and cross-sectional effects have been previously discussed in context of Simpson’s paradox (Kievit et al., 2013, Front Psychol). In addition to statistical processes (design, cohort differences, selective attrition, sampling bias, ground truth between/within-subject variability ratio) any image processing pipelines might affect between- and within-subjects differentially and result in regional sensitivity differences (Bernal-Rusiel et al., 2013, NeuroImage). Here, surface projections for z-maps (p < .05, FDR) are shown, one-sided Wald-test for positive effects of age_mean on MT. (c) Females have a significantly higher myelin-sensitive MT in bilateral insula (max z-value [z = 6.32, p = .012 voxelwise FDR, p = .001 clusterwise FWE] in left posterior insula [−43 −10 8]), amygdala, superior frontal gyrus, anterior and midcingulate cortex, middle frontal gyrus and lateral temporal cortex. We test for females over males in the same longitudinal model as shown in Figs. 1, 2. In contrast to cortical regions only very minor significant sex-related differences observed in cortical and striatal adjacent white matter regions (not shown). All presented z-maps are peak/voxel level corrected, p < .05 (FDR). Right panel shows a plot of myelin-sensitive MT in females/males with adjusted data (gray/black) and age-related models (red/orange, effect of interest: intercept, time/visit, age_mean, sex, sex by age_mean, that is assuming major trend is linear) in insular peak (for fixed other covariates, interactions etc.). X-axis: age refers to the chronological age of each individual at each visit (cf. supplemental experimental procedures). (d) Testing nonlinearities in terms of a deceleration of MT changes (that is an age-dependent reduction of individual slopes) during the developmental transition into adulthood. In particular, we observe minor tendencies to deceleration of within-subject MT growth with higher ages, in posterior cingulate (max z-value [z = 3.45, p = .0003 unc.] in right PCC [13 −48 30]) and orbitofrontal superficial white matter (z-value [z = 3.62, p = .0002 unc.] in right OFC [33 30 −13]) testing for a negative time/visit by age_mean interaction (for fixed covariates, time/visits, age_mean, sex etc.). Shown z-maps are peak/voxel level uncorrected, p < .005. Right panel shows a plot of peak cluster in orbitofrontal white matter MT with females/males data (gray/black) and full model predictions (red/orange, all predictors). X-axis: age refers to the chronological age of each individual at each visit. Left panel shows zoom on estimated longitudinal model predictions (red/orange for females/males) suggesting the age-related linear decreases of rate of within-subject change.

    4. Supplementary Fig. 4 Macrostructural volume changes, age-related differences of brain volume and nonlinearities in late adolescence and young adulthood.

      (a) Developmental volume changes and differences using longitudinal Voxel-Based Morphometry (VBM) analysis and sandwich estimator modelling. Shown are projections of statistical Z-maps (from one-sided Wald-test) testing for volume shrinkage (negative effect of study time/visits or age of subject) in cortical (top-row) and subcortical (2nd row) gray matter regions. Widespread cortical shrinkage was observed in lateral temporal (max z-value voxel [z = 9.02, p = .002 voxelwise FDR, p=.001 clusterwise FWE] in middle temporal gyrus [MNI: 65 −42 4], n = 494/285 scans/subjects apply for panels a-c, 51.9% female), inferior parietal and lateral frontal as well as cingulate cortex. Subcortical gray matter shrinkage was especially found in cerebellum (not shown) (max z-value voxel [z = 6.00, p = .003 voxelwise FDR, p = .001 clusterwise FWE] in cereb. exterior [20 −50 −13]), ventral striatum (regional peak voxel [z-value = 5.64, p = .001, voxelwise FDR, p = .001 clusterwise FWE] in [−7 13 −10]), posterior striatum, pallidum and thalamus. All shown Z-maps are peak/voxel level corrected, p < .05 (FDR). SPM results can be found in Supplementary Table 4. (b) Local volume trajectories show widespread nonlinearities during transitioning into adulthood. In particular, we observed an age-dependent deceleration of the (within-subject) shrinkage in posterior cingulate cortex (max z-value voxel [z = 5.18, p = .005 voxelwise FDR, p = .001 clusterwise FWE] in [4 −56 16]), precuneus, anterior prefrontal cortex, middle cingulate cortex, insula, inferior parietal cortex, lateral temporal regions, frontal pole, superior frontal sulcus, inferior frontal gyrus, and orbital gyrus. Statistical maps were obtained using a one-sided Wald-test for a positive time/visit by age_mean (mean age across visits) interaction (for fixed covariates, time/visits, age_mean, sex etc.). Left data plot shows gray matter volume in females/males with data (gray/black) and full model predictions (red/orange, all predictors) in posterior cingulate cortex. X-axis: age refers to the chronological age of each individual at each visit). Right data plot shows same plot with adjusted data (removing effects of covariates of no interest) and model predictions (red/orange, effect of interest: intercept, time/visit, age_mean, sex, age_mean by sex, age_mean squared, age_mean squared by sex, age_mean by time, age_mean by time by sex). (c) Developmental volume expansion in cortical and subcortical white matter regions. Shown are projections of statistical Z-maps testing for volume expansion (one-sided Wald-test for positive effects of study time/visits or age of subject) in cortical (left) and core (right) white matter areas. Widespread volume expansion was found in cortex adjacent white matter as well as in core white matter with strongest effects in bilateral pyramidal motor tracts (max z-value voxel [z-value = 4.95, p = .001 voxelwise FDR, p = .001 clusterwise FWE] in left upper pyramidal tracts [−19 −19 46]) and prefrontal white matter (max regional z-value voxel [z = 4.65, p = .001 voxelwise FDR, p = .006 clusterwise FWE] in right anterior medial prefrontal cortex [13 42 −5]). SPM results are shown in Supplementary Table 4. Left data plot shows adjusted data (removing effects of covariate of no interest) and model predictions (red/orange, effect of interest: intercept, time/visits, sex, sex by time) in pyramidal tract peak. Right panel shows corresponding plot in prefrontal white matter (red/orange, effects of interest: intercept, age_mean, sex, age_mean by sex, age_mean squared, age_mean squared by sex), a region showing strongest cross-sectional age-related growth (regional peak, z-value = 4.5, p = .017 voxelwise FDR). Y-axis: modulated local tissue volume (VBM); x-axis: time of scan in years relative to each subject’s mean age over all visits. age_mean refers to each individual’s mean age over all visits. (cf. methods and supplementary notes on modelling).

    5. Supplementary Fig. 5 Additional impulsivity white matter effects, analysis of specific compulsivity sub-scores PI-WSUR and OCI-R.

      (a) Impulsivity was found to be associated with a reduced growth of MT in subcortical white matter areas, especially adjacent to striatum (internal capsule), anterior core white matter and close to pyramidal tracts. We show projections of Z-maps from one-sided Wald-test of a negative impulsivity by time/visit interaction (p < .05, voxelwise FDR, n = 497/288 scans/subjects and test applies to panels a-b, 51.7% female). (b) More extensive presentation of main results in Fig. 4c. Subjects with higher impulsivity show reduced myelin-sensitive MT especially in anterior medial and lateral (IFG, MFG), insular and parieto-temporal areas. We present voxel-based analysis Z-maps (p < .05, voxelwise FDR) projected on cortical white matter and striatum, directed testing for negative main effect of impulsivity. SPM results are shown in Supplementary Table 5. We account for effects of visit/time, age_mean, sex, socioeconomic status, interactions and confounds. (c) Post-hoc specificity analysis within regions observed in main compulsivity analysis using two separate compulsivity sub-scores PI-WSUR and OCI-R and linear mixed effects modelling of local MT (averaged within 6 mm sphere around peak effects presented in main results Fig. 3a) in left SFG (left column) and ACC (right column). We explore effects of scores and their respective time/visit interactions with otherwise age_mean, time/visit, sex, socioeconomic status, interactions, and confounds entered as fixed-effects covariates (n = 413/217 scans/subjects). Rows illustrate model’s trait by time/visit interaction and respective statistics of fixed effects coefficients (t-value, p-value, two-sided, df = 400) suggesting sub-score-dependent rates of local MT change over study visits for PI-WSUR (top row) and OCI-R (bottom row). Plots show adjusted longitudinal data and model prediction (including effects of interest intercept, time/visit, score by time interaction) with model trajectory illustrated in more red to yellow colours for low to higher values of each sub-score (z = [−2,1,0,1,2]). Statistics suggest that OCI-R score difference are predictive of local MT changes in those areas while PI-WSUR shows similar, but less strong tendencies. Y-axis: MT; x-axis: time of scan in years relative to each subject’s mean age over all visits. All plots show effects of interest and MT data adjusted for effects of no interest (covariates and confounds).

    6. Supplementary Fig. 6 Exploring specificity of peak effects with respect to each trait dimension.

      We show results of linear-mixed modelling of local MT (averaged within 6 mm sphere around peak effects presented in main results Figs. 3a, b and 4a) in left SFG, amPFC, left anterior insula, right IFG, left ACC (anterior midcingulate) and left ventral striatum depicted in columns. We use a design matrix accounting simultaneously for both compulsivity and impulsivity regressors and their respective time/visit interactions (n = 497/288 scans/subjects with both covariates available, 51.7% female). Otherwise, age_mean, visit, sex, socioeconomic status, interactions, and confounds are entered as fixed effects covariates. Rows illustrate model’s trait by time/visit interaction and respective statistics from fixed effect coefficient (t-value, p-value, two-sided, df = 479) suggesting trait-dependent rates of local MT change over study visits for compulsivity (top row) and impulsivity (bottom row). Plots show adjusted longitudinal data and model prediction (including effects of interest intercept, time/visit, trait by time interaction) with model trajectory illustrated in more red to yellow colours for low to higher scores of each trait (z = [−2,1,0,1,2]). Statistics suggest specificity of effects reported in Fig. 3a, b and Fig. 4a for each dimension in all (except one) peak effect region even when accounting for variations of both traits simultaneously. The insula shows attenuated growth as (additive linear) function of both traits when controlling for variations of the other. Y-axis: MT; x-axis: time of scan in years relative to each subject’s mean age over all visits. All plots show effects of interest (time) and data adjusted for effects of no interest (covariates and confounds).

    7. Supplementary Fig. 7 Global frontal myelination in relation to impulsivity and compulsivity. Voxel-wise analysis of correlated brain-behavioral changes over study visits.

      (a) Reduced frontal MT growth rates with higher compulsivity (C, t = −2.494, p = .013, two-sided, df = 427) and impulsivity (I, t = −3.268, p = .0012, two-sided, df = 474) scores observed (illustrated with coloured median split group trajectories in left and right panel) using linear-mixed effects modelling testing time/visit by trait interaction (accounting for covariates such as subject’s mean age, sex, interactions etc., n = 497/288 scans/subjects in a-b, t (p) values from corresponding fixed-effects coefficients). Frontal, development-independent MT (across gray and adjacent white matter) at baseline distinguishes compulsivity (left panel) and impulsivity (right panel) with a ‘hypo-myelination’ in the latter dimension (t = 2.298, p = .022, two-sided, df = 474), but no baseline-effect in the former trait (t = 1.025, p = .30, two-sided, df = 474). Voxel-based analyses reveal similar effects: while impulsivity differences were linked to quantitative MT deficits (Fig. 4c) that further widens with development, compulsivity suggests a local decrease of a subtle myelin-related head-start (Fig. 3c). X-axis: time of scan in years relative to each subject’s mean age over all visits (cf. supplemental methods). Both dimensions showed a reduction in myelin-related longitudinal growth (impulsivity by time interactions: t = −2.795, p = .005, two-sided, df = 421; compulsivity by time: t = −1.99, p = .047, two-sided, df = 421). (b) Global growth rate was found to be associated with patterns of both compulsivity and impulsivity. Linear-mixed effects modelling of global frontal MT trajectories with continuous trait scores showed additive effects of both risk scores on MT growth (no interaction, p > .31). Panel shows brain trajectories with post-hoc median splits to illustrate these associations. High-risk subjects (red) scoring high on both scores (+: above median) express a cessation of MT development, while low-risk subjects (blue, -: below median) show the most pronounce MT growth. Subjects with mixed patterns (purple: high compulsive & low impulsive, green: low compulsive & high impulsive) exhibit intermediate MT growth rates. This suggests that frontal developmental myelin-related growth patterns are significant indicator of traits in terms of expressing higher values on the considered dimensions. (c) All presented results in Fig. 3 & 4 were focused on brain correlates of individual variability of trait compulsivity and impulsivity as defined in methods section. Here we focus on brain correlates of developmental changes of impulsivity in terms of the longitudinal progression of BIS total scores over study visits/time. As suggested by Guillaume et al. (2014, NeuroImage) time-varying behavioural scores were decomposed in purely within- and between subjects components and entered as regressors in voxel-wise modelling of myelin sensitive MT (in addition to covariates time/visits, age_mean, sex and confounds, n = 376/188 scans/subjects). All statistical maps were obtained from longitudinal sandwich estimator for VBQ analysis. Here, surface projections for Z-maps (p < .01 unc., one-sided) are shown, testing for negative effect of within-subject changes of BIS. We observed focal tendencies for individual BIS growth (impulsivity) being inversely related to MT growth over visits (peak in anterior inferior frontal gyrus, z-value = 3.67, p = .00013 unc.). That suggests that the reduced growth rate is more strongly expressed in subjects who manifest an accentuated impulsivity increase over study visits, such that subjects who manifest an even more restricted growth in myelin become more impulsive.

    8. Supplementary Fig. 8 Assessing the effects of during-scan motion on local MT and on effects of interest such as development, sex, compulsivity and impulsivity.

      We assessed a retrospective proxy for during-scan motion based on Multi-Parameter Mapping R2* exponential decay model residual in white matter areas (see Castella et al., 2018, Magn Reson Med; and www.hmri.info for details). The motion proxy (denoted by SDR2*) has been show to accurately reflect absolute motion measures using prospective motion correction. Beyond visual and covariance based quality checks, SDR2* was used to remove 10% of scans with highest motion artefacts before analysis. (a) In resulting clean sample, average SDR2* of each participant correlates with motion parameters during an independent resting-state fMRI scan of same individuals (r = .36, two-sided Pearson’s correlation, p = e-9 from corresponding t-distribution, n = 316 independent subjects) suggesting its validity. (b) Across all scans and individuals the motion proxy was found to be positively associated with increased focal MT in right inferior cortical areas (peak right lingual and fusiform gyrus, z-value = 5.02, p = .016, one-sided Wald-test, voxelwise FDR, n = 497/288 scans/subjects applies for panels b-c). No areas with negative effects or trends of motion were observed. (c) To avoid spurious MT effects induced by motion variability, SDR2* was included as a time-varying (scan by scan) covariate during all presented main analyses. Comparing developmental growth (left column) and sex (right column) effects on myelin-sensitive MT without (top row) and with (bottom row) motion regressor (p < .05, voxelwise FDR). Accounting (linearly) for motion did not affect observed developmental increase of MT (over age or time/visits) and even increased statistical strength of sex effects suggesting females having higher MT in insular and frontal areas (cf Supplementary Fig. 3c, one-sided Wald-test for females over males, accounting for all other covariates such as age and visit etc.) (d) Linear mixed-effects modelling of SDR2* over all scans and subjects revealed no change of motion over visits (left plot, t = 1.47, p = .14, two-sided, df = 564, n = 573/316 scans/subjects, statistics from fixed-effects coefficient) or with age of subject (right plot, t = 0.09, p = .93, df = 564). This supports that presented developmental effects on myelin-sensitive MT (Fig. 1 and Supplementary Fig. 3) are unlikely to be biased by motion. In general, we observed higher motion (t = −5, p = e-6, two-sided, df = 564) in males compared to females. Consequently, accounting for motion induced variability analyses reveals more accurate estimation of sexual dimorphism. (e) Absence of significant compulsivity by time/visit interaction (left panel, t = −0.23, p = .81, two-sided, df = 513, n = 525/273 scans/subjects) and main effects (right panel, t = 1.37, p = .17, two-sided, df = 513) on SDR2* supports independence of main findings in Fig. 3 and motion-related variability. (f) Subjects with higher impulsivity showed more positive change of SDR2* from baseline to follow up scans (t = 1.98, p = .048, two-sided, df = 561, n = 573/316 scans/subjects) but no generally higher motion (t = −0.26, p = .73, two-sided, df = 561). Given the observed positive association of SDR2* and MT above, this suggests that reduced growth of MT over visits in higher impulsive subjects might be regionally underestimated, that is an even more pronounced reduction of growth is likely in some areas. Therefore, motion proxy SDR2* was included in all main analysis and Fig. 4 is likely to represent effects independent from motion-induced artefacts. Plots b & d-f show effects of interest and motion proxy data adjusted for effects of no interest (that is covariates and confounds). Notably, a slightly larger initial sample with 573 native space scans (before spatial processing) with available motion parameters was used to conduct analyses in panels a & d–f.

    9. Supplementary Fig. 9 Longitudinal intelligence development in our study and relation to compulsivity and impulsivity dimensions. Control analysis of main finding using various additional covariates.

      (a) Linear mixed-effects modelling of unstandardized WASI intelligence matrix subscore over all scans and subjects revealed increase of abilities over visits (left plot, t = 4.8, p = e-5, df = 516, statistics from fixed-effects coefficient, two-sided, n = 522/292 observations/subjects in a and c) and with higher age of subject (right plot, t = 2.4, p = .017, df = 516). This supports that cognitive ability differences and changes might be a hidden variable for compulsivity/impulsivity related associations presented in the main findings. However, this is unlikely since our study revealed no evidence for intelligence being related to either compulsivity (b) (n = 481/252 observations/subjects, df = 472, two-sided) or impulsivity (c) (n = 522/292 observations/subjects, df = 513, two-sided), supporting that our main finding can be considered rather independent of cognitive development in late adolescents. All plots show effects of interest and data adjusted for effects of no interest (that is covariates and confounds). (d) Here we illustrate the main finding of reduced longitudinal growth of cortical MT with higher compulsivity (top row) and impulsivity (bottom row) (main results Figs. 3a and 4a) when additionally accounting for between-subject covariates and their time/visit interactions for alcohol consumption, substance consumption, and ethnicity during longitudinal Sandwich Estimator modelling. Last two columns show similar accounting for fully time-varying WASI intelligence sub-scores matrix and vocabulary as covariates. Here, surface projections for Z-maps (p < .05, voxelwise FDR, one-sided Wald-test) are shown, testing for negative trait by visit/time interaction (age_mean, sex, socioeconomic status, interactions and confounds fixed in all analyses, n = 497/288 scans/subjects).

    Supplementary information

    1. Supplementary Information

      Supplementary Figures 1–9, Supplementary Tables 1–5, Supplementary Note, and NSPN Consortium member list.

    2. Reporting Summary

    About this article

    Publication history