A Molecular Classification of Human Mesenchymal Stromal Cells Florian Rohart1, Elizabeth Mason1, Nicholas Matigian1, Rowland
Total Page:16
File Type:pdf, Size:1020Kb
bioRxiv preprint doi: https://doi.org/10.1101/024414; this version posted August 11, 2015. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license. Rohart'et'al,'The'MSC'Signature' 1' 1' A Molecular Classification of Human Mesenchymal Stromal Cells 2' Florian Rohart1, Elizabeth Mason1, Nicholas Matigian1, Rowland Mosbergen1, 3' Othmar Korn1, Tyrone Chen1, Suzanne Butcher1, Jatin Patel2, Kerry Atkinson2, 4' Kiarash Khosrotehrani2,3, Nicholas M Fisk2,4, Kim-Anh Lê Cao3 and Christine A 5' Wells1,5* 6' 1 Australian Institute for Bioengineering and Nanotechnology, The University of 7' Queensland, Brisbane, QLD Australia 4072 8' 2 The University of Queensland Centre for Clinical Research, Herston, Brisbane, 9' Queensland, Australia, 4029 10' 3 The University of Queensland Diamantina Institute, Translational Research 11' Institute, Woolloongabba, Brisbane QLD Australia, 4102 12' 4 Centre for Advanced Prenatal Care, Royal Brisbane & Women’s Hospital, Herston, 13' Brisbane, Queensland, Australia, 4029 14' 5 Institute for Infection, Immunity and Inflammation, College of Medical, Veterinary & 15' Life Sciences, The University of Glasgow, Scotland, UK G12 8TA 16' *Correspondence to: Christine Wells, [email protected] 17' bioRxiv preprint doi: https://doi.org/10.1101/024414; this version posted August 11, 2015. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license. Rohart'et'al,'The'MSC'Signature' 2' 18' Abstract 19' Mesenchymal stromal cells (MSC) are widely used, isolated from a variety of tissues 20' and increasingly adopted for cell therapy, but the identity of these cells is poorly 21' defined and commonalities between MSC from different tissues sources is 22' controversial. Here we undertook a comprehensive review of all public MSC 23' expression studies to assess whether cells derived from different sources shared 24' any common molecular attributes. In doing so, we discovered an over-arching 25' transcriptional phenotype shared by a wide variety of MSC, freshly isolated or 26' cultured cells, and under a variety of growth conditions. We developed a modified 27' variable selection protocol that included cross platform normalisation, and 28' assessment of the selected gene stability and informativeness. A 16-gene signature 29' classified MSC with >97% accuracy, discriminating these from fibroblasts, other adult 30' stem/progenitor cell types and differentiated cells. The genes form part of a protein- 31' interaction network, and mutations in more than 65% of this network were associated 32' with Mendelian disorders of skeletal growth or metabolism. The signature and 33' accompanying datasets are provided as a community resource at 34' www.stemformatics.org resource, and the method is available from the CRAN 35' repository. 36' bioRxiv preprint doi: https://doi.org/10.1101/024414; this version posted August 11, 2015. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license. Rohart'et'al,'The'MSC'Signature' 3' 37' Introduction 38' Adult tissues maintain the capacity to be replenished as part of the normal processes 39' of homeostasis and repair. The adult stem cell hypothesis proposes that multipotent 40' cells resident in tissues are the source of this cellular renewal, and expand in 41' response to tissue injury. MSC were first isolated from bone marrow, where these 42' occupy an important stem cell niche required for reconstitution of bone and the 43' stromal compartments of marrow, and also play a supportive role in haematopoiesis 44' (1, 2). Subsequently adult stromal progenitors have been isolated and cultured from 45' most organs including placenta, heart, adipose tissue and kidneys although the 46' identity of these cells remains controversial (reviewed by (3, 4)). Specifically the 47' question of how similar cells isolated outside the bone marrow niche are is 48' unresolved, nor whether these could be considered bona fide MSC, or indeed 49' challengingly, whether MSC isolated from different tissues share any phenotypic or 50' molecular characteristics at all (3). In this light various cells described as MSC 51' (whether by name or attribution) have been reported as having quite different self- 52' renewal capacity, immunomodulatory properties or propensity to differentiate in vivo 53' (5). It has been variously argued that MSC isolated from most stromal tissues are 54' derived from perivascular progenitors (6), or recruited from the bone marrow to distal 55' tissue sites (7), or that resident stromal progenitors from different tissues must have 56' tissue-restricted phenotypes. 57' The question of ontogeny aside, there is little consensus on whether MSC from 58' differing tissue sources share common functional attributes. Most human studies 59' have been conducted on very small numbers of donors, so it is difficult to dissect 60' donor-donor heterogeneity from source heterogeneity. Donor-donor variation is a bioRxiv preprint doi: https://doi.org/10.1101/024414; this version posted August 11, 2015. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license. Rohart'et'al,'The'MSC'Signature' 4' 61' major contributor to differences in MSC growth and differentiation capacity, and 62' clonal variation is evident even when derived from the same bone marrow (8, 9). 63' Consequently there is little agreement in the literature on definitive molecular or 64' cellular phenotypes of human cultured MSC, whether from bone marrow or other 65' sources. 66' The lack of definitive markers for human MSC is a major barrier to understanding 67' genuine similarities, or resolving differences between various cell sources or 68' subsets. Modern molecular classification tools are needed for the characterisation of 69' MSC ex vivo and in vivo. Here we describe a sophisticated integrative transcriptome 70' analysis of public MSC datasets to assess how similar these cells are, and describe 71' the major classes of MSC captured in the literature to date. 72' 73' Results 74' A 16-gene MSC signature defines an overarching MSC phenotype. 75' To address whether MSC shared a common molecular phenotype that could 76' distinguish them from other stromal or progenitor cells, we systematically reviewed 77' all of the publicly available transcriptome data for presumed MSC, identifying 120 78' potential datasets that were derived from a wide variety of tissues and age groups, 79' but 35/120 datasets failed our QC criteria for data quality and were excluded from 80' the study. We assessed the accompanying phenotypic data of the remaining 85 81' datasets carefully (Supplementary Tables S1-S5) for immunophenotype and ex vivo 82' differentiation, as determined by the International Society for Cellular Therapy (10). A 83' ‘gold standard’ sample set was assembled, consisting of 125 MSC samples from 16 bioRxiv preprint doi: https://doi.org/10.1101/024414; this version posted August 11, 2015. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license. Rohart'et'al,'The'MSC'Signature' 5' 84' independently derived datasets. The ‘Gold standard’ MSC were primarily derived 85' from bone marrow, but also included a variety of adult, neonatal and fetal stromal 86' sources, and these were compared to 510 non-MSC samples from primary human 87' tissues, including cultured fibroblasts, haematopoietic cells and pluripotent stem cell 88' lines (Supplementary Table S1, S2). 89' We derived a novel cross-study framework to test whether we could find similarities 90' between the MSC in our training set despite tissue, platform or laboratory 91' differences. Our approach, described in Figure 1A, included a cross-platform 92' normalisation step (11), and a modified multivariate discriminant analysis that 93' included steps to evaluate the stability of gene selection when datasets were 94' subsampled as well as steps to evaluate the informativeness of the variables that 95' contributed to each component (Figure 1B, Supplementary Figure S1). 96' This identified 16 genes (Figure 1C) that collectively formed a ‘signature’, which 97' across 4 components grouped bone-marrow derived MSC with MSC from other 98' sources, and provided a high degree of discrimination between MSC and non-MSC 99' cell types (Figure 1D, Supplementary Figure S1). The accuracy of the signature 100' (Table 1) was 97.8%, calculated as the percentage of correctly classified samples in 101' 200 subsamplings of an internal test set. 102' 103' Figure 1. Identifying the MSC signature. 104' A) Workflow summarizing our improved implementation of the sPLS-DA and 105' derivation of a stable 16 gene MSC classifier; 106' B) Choosing the most informative minimal gene set on each component by testing 107' the benefit of including more genes. Each dot is a gene set, ordered along the x-axis bioRxiv preprint doi: https://doi.org/10.1101/024414; this version posted August 11, 2015. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license. Rohart'et'al,'The'MSC'Signature' 6' 108' by decreasing stability (frequency of selection).