Atrial myxomas arise from multipotent cardiac stem cells

Abstract Aims Cardiac myxomas usually develop in the atria and consist of an acid-mucopolysaccharide-rich myxoid matrix with polygonal stromal cells scattered throughout. These human benign tumours are a valuable research model because of the rarity of cardiac tumours, their clinical presentation and uncertain origin. Here, we assessed whether multipotent cardiac stem/progenitor cells (CSCs) give rise to atrial myxoma tissue. Methods and results Twenty-three myxomas were collected and analysed for the presence of multipotent CSCs. We detected myxoma cells positive for c-kit (c-kitpos) but very rare Isl-1 positive cells. Most of the c-kitpos cells were blood lineage-committed CD45pos/CD31pos cells. However, c-kitpos/CD45neg/CD31neg cardiac myxoma cells expressed stemness and cardiac progenitor cell transcription factors. Approximately ≤10% of the c-kitpos/CD45neg/CD31neg myxoma cells also expressed calretinin, a characteristic of myxoma stromal cells. In vitro, the c-kitpos/CD45neg/CD31neg myxoma cells secrete chondroitin-6-sulfate and hyaluronic acid, which are the main components of gelatinous myxoma matrix in vivo. In vitro, c-kitpos/CD45neg/CD31neg myxoma cells have stem cell properties being clonogenic, self-renewing, and sphere forming while exhibiting an abortive cardiac differentiation potential. Myxoma-derived CSCs possess a mRNA and microRNA transcriptome overall similar to normal myocardium-derived c-kitpos/CD45neg/CD31negCSCs , yet showing a relatively small and relevant fraction of dysregulated mRNA/miRNAs (miR-126-3p and miR-335-5p, in particular). Importantly, myxoma-derived CSCs but not normal myocardium-derived CSCs, seed human myxoma tumours in xenograft’s in immunodeficient NOD/SCID mice. Conclusion Myxoma-derived c-kitpos/CD45neg/CD31neg CSCs fulfill the criteria expected of atrial myxoma-initiating stem cells. The transcriptome of these cells indicates that they belong to or are derived from the same lineage as the atrial multipotent c-kitpos/CD45neg/CD31neg CSCs. Taken together the data presented here suggest that human myxomas could be the first-described CSC-related human heart disease.


Introduction
Primary tumours of the heart are rare, with an incidence between 0.0017% and 0.19% in unselected patients at autopsy. [1][2][3] Three quarters of the tumours are benign and nearly half the benign heart tumours are myxomas. Myxoma has an annual incidence of 0.5 per million people and most commonly presents in 30-to 50-year-old adults 4 with a slight preponderance in women (65%).Cardiac myxomas are typically sporadic and isolated, while only in 5-10% are familial 5 and usually develop in the atria (75% from the left atrium and 18% from the right atrium) 6 and they are regarded as benign neoplasms in the histological conventional sense. However, the oncologic designation of benignity understates the potentially devastating effect this tumour may have on the patient. By virtue of their location, cardiac myxomas can produce myriad clinical manifestations, sometimes with fatal consequences, such as heart failure and distant embolisms of the CNS or other organs depending on their location. Therefore, these benign tumours, despite their rarity, continue to generate interest because of their clinical presentation and uncertain histogenesis.
A myxoma consists of a myxoid stroma rich in acidmucopolysaccharide with polygonal cells scattered throughout. Differentiation of myxoma cells towards neuronal, smooth muscle, endothelial, fibroblasts, and myocytes along with typical stromal cells has been described. [7][8][9] The pathogenesis of cardiac myxoma is poorly understood with early debates about whether cardiac myxomas were, indeed, neoplastic entities or organized thrombus. Recent gene expression and immunohistochemical studies have postulated that cardiac myxomas are neoplasms with the tumour cells arising most likely from multipotent mesenchymal cells yet to be identified. [10][11][12] Many studies have shown that the prototypical stem-cell properties (self-renewal, clonogenicity, and multipotent differentiation ability) are relevant to some forms of cancer. 13,14 Biologically distinct and relatively rare populations of 'tumour-initiating' cells have been identified in different form of cancers which have properties closely parallel to the three features that define normal stem cells. Malignant cells with these functional properties have been termed 'cancer stem cells' (CaSCs).
Although previously considered to be a terminally differentiated organ, the adult mammalian heart has been shown to have the ability to develop new cardiomyocytes and microvasculature throughout life. 15 Although the extent and biological significance of this cellrenewal capacity remains controversial, [16][17][18] it has been demonstrated that the adult mammalian heart harbours a population of bona fide resident endogenous cardiac stem/progenitor cells (CSCs). Such cells have been identified by us and other groups in a range of different mammalian species, including human. The CSCs were first characterized by expression of c-kit (also known as CD117), the kinase receptor for stem cell factor. 19,20 However, the adult heart contains a heterogeneous c-kit pos cell population most of which display blood and endothelial lineage-commitment. [21][22][23] The c-kit pos / CD45 pos population shows mast cells identity, while c-kit pos / CD31 pos are endothelial (precursor) cells. 22,24,25 Recently, it was demonstrated that only a small fraction of c-kit pos cardiac cells, negative for CD45 and CD31, are enriched for multipotent CSCs. 21 These cells are distributed throughout the myocardium with the highest density in the atria and apex. 19, 26 Based on the above, the objectives of this study were to assess whether c-kit pos /CD45 neg /CD31 neg cardiac cells contribute to cardiac myxoma histogenesis and whether myxoma-derived c-kit pos / CD45 neg /CD31 neg cells exhibit, in vitro and in vivo, the properties of myxoma stem cells.

Translational perspective
Primary tumours of the heart are rare, and among them, the vast majority is represented by cardiac myxomas. The typical cardiac myxoma is regarded as a benign neoplasm in a conventional sense. However, the oncologic designation of benignity understates the potentially clinical devastating effect this cardiac tumour may impose on the patient. Therefore, these benign tumours, despite their rarity, continue to generate interest because of their clinical presentation and uncertain histogenesis. The pathogenesis of cardiac myxoma is indeed poorly understood. Although there was debate in the past about whether cardiac myxoma was a neoplastic entity or an organized thrombus, recent gene expression and immunohistochemical studies established that cardiac myxoma is a neoplasm postulating that tumour cells arise most likely from multipotent mesenchymal cells, yet to be identified. Here, assessing 23 human cardiac myxomas, we detected within myxoma tissue c-kit pos / CD45 neg /CD31 neg cardiac cells expressing stemness and cardiac progenitor cell transcription factors. Some (<10%) of these c-kit pos /CD45 neg / CD31 neg cardiac myxoma cells expressed also calretinin, likely representing myxoma stromal precursor cells. More importantly, c-kit pos / CD45 neg /CD31 neg cardiac myxoma cells secrete the typical gelatinous matrix of cardiac myxoma. These c-kit pos /CD45 neg /CD31 neg cardiac myxoma cells show stem cell properties being clonogenic, self-renewing, and sphere forming. The myxoma-derived multipotent cells exhibit a transcriptome and miRNA network which is similar to that of the cardiac stem/progenitor cells (CSCs) from normal myocardium but with a relatively small fraction of dysregulated mRNAs and miRNAs (and miR-126-3p and miR-335-5p, in particular) involved in cell growth, differentiation and transformation pathways. Importantly, myxoma-derived multipotent CSCs seed human myxoma tumour in xenograft's experiments in immunodeficient mice. Therefore, for the first time, we provide the evidence of the existence of a tissue-specific myxoma-initiating stem cell in atrial myxoma. Accordingly, cardiac myxomas appear to be the first CSC-related human cardiac disease.

Methods
Detailed description of the methods is in the Supplementary material online, Methods. The authors declare that all data supporting the findings of this study are available either within the article or from the corresponding author on reasonable request.

Results
c-kit pos /CD45 neg /CD31 neg cells are present within atrial myxomas and express stemness and cardiac transcription factors Twenty-three cardiac imaging-identified myxoma tumours (Supplementary material online, Table S1) were surgically excised and processed for histology and immunohistochemistry analysis. All the tumours exhibited the typical histological cellular and extracellular matrix features of cardiac atrial myxoma (Supplementary material online, Figure S1). Dispersed within the myxoma matrix, we detected cells expressing known embryonic and adult cardiac stem/progenitor markers such as c-kit, the receptor of the stem cell factor ( Figure 1A). Total c-kit pos cardiac cells number was higher in the myxoma tissue than in the normal atrial tissue obtained from control patients (n = 10) ( Figure 1B). The significance, if any, of this difference in c-kit pos cells density is doubtful considering the differences in tissue composition.
We have already reported that c-kit expression in the adult myocardial cells mainly identifies endothelial and mast cells. 21 Accordingly, most of the myxoma c-kit pos cells were CD45 pos or CD31 pos (85 ± 5%) ( Figure 1B-D), representing either cardiac mast cells and/or endothelial cells, respectively. Moreover, myxoma c-kitpos /CD45 pos cells were also tryptase pos , confirming their mast cell phenotype ( Figure 1C). A similar fraction of c-kit pos /CD45 pos / CD31 pos cells (83 ± 7%) within the total cardiac c-kit pos cells was detected in the normal right atrial tissue of the controls ( Figure 1B). On the other hand, we also identified c-kit pos cardiac myxoma cells that were negative for both CD45 and CD31 ( Figure 1E). In the adult mammalian heart, c-kit pos /CD45 neg /CD31 neg cells are highly enriched for CSCs. 19,21,24 Immunofluorescent staining revealed that c-kit pos / CD45 neg /CD31 neg myxoma cells co-express several stemness and cardiac progenitor cell transcription factors, such as Oct-4, Nkx2.5, and Isl-1 in different percentages ( Figure 1F-I), markers known to be also expressed in normal c-kit pos /CD45 neg /CD31 neg CSCs from healthy human atrial tissue. 23,27 Overall, these data raise the question of whether the c-kit pos / CD45 neg /CD31 neg myxoma cells are derived from the atrial c-kit pos / CD45 neg /CD31 neg stem/precursor cells and whether they are myxoma multipotent stem/progenitor cells.

Evidence of intermediate cell phenotype between undifferentiated cardiac progenitor cells and polygonal stromal myxoma cells
The cellular compartment of myxomas is composed of stromal cells which mainly have a polygonal structure. These cells, also known as leipidic cells, exhibit molecular and phenotypic traits of multiple cell types, such as neuronal, endothelial, vascular, and cardiac cells. 7,28,29 One of the salient characteristics of cancer stem cells (CaSCs) is that they give rise to the non-CaSCs-cells within a tumour which display a more differentiated phenotype and have less proliferative capacity compared to the CaSCs. Differentiated cells within the tumour mass are believed to be the product of the progressive abortive differentiation of undifferentiated CaSCs, which involves the formation of cell intermediates constituting the cellular heterogeneity of some neoplasias. 30 Thus, we searched for c-kit pos /CD45 neg /CD31 neg myxoma cells expressing myxoma cell markers.
Different markers have been used to specifically identify polygonal cells. 31 Recently, calretinin has been shown to mark the majority of myxoma cells. 32 Most of the myxoma polygonal cells in the myxomas reported here were positive for calretinin (74 ± 7%, Figure 1J). Importantly, 10 ± 3% of the c-kit pos /CD45 neg /CD31 neg myxoma cell cohort expressed calretinin (Figure1K and L). In contrast, calretinin was not expressed in c-kit pos CSCs from normal atrial tissue. Thus, the presence c-kit pos /CD45 neg /CD31 neg /calretinin pos cells within myxoma tissue appears to identify myxoma stromal precursor cells. This concordance potentially links the multipotent c-kit pos /CD45 neg / CD31 neg CSCs present in the normal adult myocardium with the ckit pos /CD45 neg /CD31 neg /calretinin pos myxoma cells which are likely the precursors of the myxoma leipidic cells.
Myxoma-derived c-kit pos /CD45 neg / CD31 neg cells are clonogenic and spherogenic with an abortive cardiac differentiation potential To assess the stem cell properties of putative multipotent myxoma stem cells, 33 we processed the six most recently excised out of the 23 myxomas and disaggregated them to single cells by enzymatic digestion. FACS analysis confirmed that c-kit pos cells were about 11 ± 3% of all nucleated cells the myxoma ( Figure 2A). Most of these c-kit pos cells co-expressed CD45 and CD31 (Figure 2A), confirming the immunohistochemistry data ( Figure 1C and D). Through MACS sorting we obtained c-kit pos /CD45 neg /CD31 neg myxoma cells and compared their cellular properties in vitro to normal human CSCs similarly isolated from three right atrial appendix tissues.
Normal human CSCs are efficiently clonogenic. At passage 8 (P8), they clone ( Figure 2B) with an efficiency of 13 ± 4% ( Figure 2C). A typical clone of normal hCSCs has been expanded more than 50 passages so far. After these passages, these cells retain telomerase activity and normal telomere length, without cellular and/or molecular evidence of senescence (Supplementary material online, Figure  S2A). Clonogenic normal hCSCs express mRNA and protein of recognized 'stemness' and cardiac progenitor genes such as C-KIT, MDR-1, OCT-4, NANOG, BMI-1, TERT, GATA-4, and NKX2.5 (Supplementary material online, Figure S2B Figure S2D). When grown in suspension, normal hCSCs are multipotent and give rise to pseudoembryoid bodies, also known as cardiospheres (Supplementary material online, Figure S3A). When grown in specific conditioned differentiation media, normal hCSCs differentiate into cardiomyocytes, vascular smooth muscle, and endothelial cells in vitro (Supplementary material online, Figure S3B and C).

online,
c-kit pos /CD45 neg /CD31 neg cardiac myxoma-derived cells (hereafter referred to also as myxoma-derived 'putative' CSCs) at P8 show a clonogenic capacity similar to normal hCSCs, 11 ± 2% ( Figure 2C). However, at P12 the myxoma cells start to show a significant prolongation of their doubling time ( Figure 2D) with reaches practical growth arrest at P16. This cell growth deficit was associated with a significant increase in spontaneous apoptotic cell death as revealed by TdT assay [0.1 ± 0.1% TdT pos cells in normal hCSCs compared to 5 ± 3% in ckit pos myxoma cells at P16 (P < 0.01)]. Moreover, at P12 c-kit pos / CD45 neg /CD31 neg myxoma-derived CSCs become larger with a diameter of 15 ± 2 lm when compared with 10 ± 2 lm of normal human eCSCs (P < 0.01). Interestingly, c-kit pos /CD45 neg /CD31 neg myxomaderived CSCs at P12 can still be cloned with an efficiency of 8 ± 2%.
Cloned c-kit pos /CD45 neg /CD31 neg myxoma-derived CSCs express a membrane multimarker profile ( Figure 2E) very similar to normal human c-kit pos /CD45 neg /CD31 neg CSCs (Supplementary material online, Figure S2C). Also, these clonogenic c-kit pos /CD45 neg /CD31 neg myxoma-derived CSCs express OCT-4, NANOG, BMI-1, NKX2.5, and ISL-1 ( Figure 3A) and form cardiospheres at a rate similar to normal human CSCs even though serial spherogenesis was reduced ( Figure 3B and Supplementary material online, Figure S3A). When grown in specified endothelial differentiation media, clonogenic c-kitpos /CD45 neg /CD31 neg myxoma-derived CSCs up-regulated endothelial lineage-restricted genes similarly to hCSCs ( Figure 3C and Supplementary material online, Figure S3B). However, in smooth muscle cell (SMC) differentiation media, the expression of SMC lineage markers was lower in differentiated c-kit pos /CD45 neg /CD31 neg myxoma-derived CSCs compared to normal human c-kit pos / CD45 neg /CD31 neg CSCs ( Figure 3D and Supplementary material online, Figure S3B). More importantly, when clonogenic c-kit pos / CD45 neg /CD31 neg myxoma-derived CSCs are primed for cardiomyogenic differentiation, they up-regulated cardiac myocyte transcription factors but in vitro they fail to express significant levels of the contractile cardiac muscle genes , in contrast to normal hCSCs ( Figure 3E and Supplementary material online, Figure S3C).
Overall, these data show that c-kit pos /CD45 neg /CD31 neg myxomaderived CSCs have many similarities with the normal myocardiumresident c-kit pos /CD45 neg /CD31 neg CSCs and fulfill the in vitro criteria for myxoma tumour stem cells: they are clonogenic, self-renewing, and multipotent, yet they accumulate a growth deficit over passages, which is consistent with the typical low proliferation index of cardiac myxomas. 10,34 In addition, they show a biased differentiation potential towards endothelial cells while they exhibit an abortive myogenic differentiation potential, which is also typical of myxoma tissue. 14 c-kit pos /CD45 neg /CD31 neg myxomaderived CSCs produce the typical myxoid matrix The abundant gelatinous matrix is the most impressive aspect of the morphologic appearance of cardiac myxomas. [35][36][37] This gelatinous matrix consist of glycosaminoglycans related to chondroitin sulfates and hyaluronic acid which are produced by the myxoma cells. [35][36][37] Thus, we analysed and measured glycosaminoglycan content and production of myxoma-derived c-kit pos /CD45 neg /CD31 neg CSCs at P8-12 by HPLC. Chondroitin-6-sulfate and hyaluronic acid, the main disaccharide units of glycosaminoglycans composing the gelatinous matrix, 37 were particularly abundant in the glycosaminoglycan produced and secreted by myxoma-derived c-kit pos /CD45 neg /CD31 neg CSCs ( Figure 3F) and were the most prevalent in the culture medium analysed ( Figure 3G).
These data further show that myxoma-derived c-kit pos /CD45 neg / CD31 neg CSCs behave and show the characteristics expected from myxoma tumour stem cells.
RNASeq analysis reveals significant similarities and differences in the transcriptome of c-kit pos /CD45 neg / CD31 neg myxoma-derived CSCs and normal c-kit pos /CD45 neg /CD31 neg CSCs To assess the gene expression similarities and differences between the myxoma-derived 'putative' CSCs and normal human CSCs, we analysed the RNA expression profile of these two-cell populations (n = 6 clones from 2 myxoma patients and n = 9 clones from 3 control human patients) by RNASeq analysis. To increase the significance of this mRNA profile analysis, only mRNAs that had a read counts > _100 and a [fold change] > _1.5 were considered. A total of 15 321 mRNAs were analysed in this comparison and 1243 were differentially expressed in the two cell types (see Supplementary material online, Table S2). That the myxomaderived CSCs share 90% of gene expression with human CSCs highly point to a common origin for these cells. On the other hand, of the differentially regulated genes, 407 mRNA were significantly up-regulated (> _1.5 fold difference) and 836 were downregulated (-1.5 fold difference) in myxoma-derived c-kit pos / CD45 neg /CD31 neg CSCs vs. normal c-kit pos /CD45 neg /CD31 neg CSCs ( Figure 4A and Supplementary material online, Table S2).
Subsequently, we assessed the dysregulated mRNAs by bioinformatic Ingenuity pathway's analysis (IPA) to identify the specific molecular processes affected by these mRNAs. The dysregulated mRNAs were specifically involved with apoptosis signalling, cell cycle checkpoints, factors promoting cardiogenesis, stem cell pluripotency, inhibition of matrix metalloproteases, molecular mechanism of cancer, tissue factor in cancer, and telomerase signalling ( Figure 4B and Supplementary material online, Table S2). Furthermore, IPA further revealed that the dysregulated genes in myxoma-derived CSCs are components of several canonical pathways ( Figure 4C) such as HGF and IGF-1 signalling, telomerase signalling, stem cell pluripotency, and stem cell signalling among others ( Figure 4C). Also, these dysregulated mRNAs are involved with several specific biological functions ( Figure 4C), such as organ development, cell signalling, and posttranslational modification, cardiovascular disease, DNA replication, recombination and repair, cell cycle, and cancer among several others ( Figure 4C).
In summary, the RNASeq analysis clearly shows that the myxoma-derived 'putative' CSCs and the normal human CSCs are closely related and, most likely, the former are derived from the latter. In addition, the RNASeq data uncovered specific gene expression changes in myxoma-derived CSCs which are consistent with the cellular phenotype of myxoma tumours, with their abortive neuronal/cardiac differentiation and the hyperproduction of proteoglycans forming the prototypical myxoma gelatinous matrix. Taken together the phenotype and gene expression profile of the 'myxoma-derived CSCs', compared to the CSCs from normal myocardium, strongly points to them as the stem cells of cardiac myxoma.
microRNA profile of myxoma-derived c-kit pos /CD45 neg /CD31 neg CSCs MicroRNAs (miRNA), small non-coding RNAs that play critical roles in normal stem cell functions in development, adult homeostasis and disease, have emerged as important regulators of tumours and cancer stem cells. 38 Very little is yet known about microRNA regulation in the biology of adult human CSCs either in vitro and in vivo. Therefore, we assessed the microRNA expression profile in myxoma-derived c-kit pos /CD45 neg /CD31 neg CSCs to identify miRNA/mRNA networks potentially involved in their above-described gene expression dysregulation when compared to normal human CSCs.
MicroRNA profile was obtained by RNASeq from the same samples used for the mRNA transcriptome analysis above. To increase the significance of the whole miRNA profile analysis, only miRNAs that show a read count > _90 and a [fold-change] > _1.5 were considered. On this basis, we detected 66 microRNA differently regulated in the comparison. Specifically, 38 miRNA up-regulated and 28 downregulated in myxoma-derived CSCs vs. normal human CSCs ( Figure 5A and Supplementary material online, Table S3). miR-138-5p is the most down-regulated in myxoma-derived CSCs ( Figure 5B and Supplementary material online, Table S3). Down-regulation of miR-138 is observed in a variety of cancer types, which suggests that it may be involved in their pathogenesis. 39,40 miR-335 was also significantly down-regulated in myxoma c-kit pos cells ( Figure 5B and Supplementary material online, Table S3). Interestingly, miR-335 is encoded in the mesoderm-specific transcript (Mest) which is significantly modulated during the differentiation of embryonic stem cells to mesodermal cardiomyocyte precursors 41 and it is considered a potential tumour suppressor in many cancer types. 38 Furthermore, miR-355 regulates the expression of extracellular matrix components. 41,42 On the other hand, miR-126 is the most up-regulated in the myxoma-derived CSCs ( Figure 5B and Supplementary material online, Table S3). This miR mainly contributes to normal vascular development during embryogenesis and to the regenerative function of haematopoietic stem cells. [43][44][45] Recently, it has been reported that higher expression of miR-126 accelerates cancer progression by activation of mitogen-activated protein kinase and Akt, while it also suppresses tumour proliferation and tissue invasion by decreasing Crk expression. These contradictory actions suggest that miR-126 has several alternative functions specific to the type of malignancy. 46 We specifically scanned for the known and potential mRNA targets of each down-and up-regulated miRNAs through IPA 'microRNA Target Filter' using 'TargetScan Human' and 'TarBase'. The up-regulated mRNA putative targets of the down-regulated microRNAs in myxoma vs. normal CSCs and vice versa to generate the miR/mRNA networks of the two cell types ( Figure 5C). The bioinformatic analysis revealed that the miRNA/mRNA network deregulated in myxoma-derived CSCs fall into six main categories: proteoglycans in cancer, cGMP-PKG signalling pathway, Ras signalling pathway, biosynthesis of amino-acids, apoptosis of tumour cells, and cell cycle progression ( Figure 5D and E).
Overall, the un-biased RNASeq and bioinformatics analysis of the differentially regulated miRNA networks revealed that myxomaderived CSCs (myxoma putative stem cells) have a gene expression and regulatory microRNAs profile that is consistent and concordant with the cellular features exhibited by these cells in vitro and in vivo. Therefore, these mRNA/miRNA networks have the potential to become an entry point to unravel the main molecular mechanisms underlying the myxomatous transformation and to further ascertain the causative role of CSCs in its genesis.

miR-126-3p and miR-335-5p regulate the biology of myxoma-derived CSCs in vitro
On the basis of the RNASeq-based miR-nome data above, we have then evaluated the effects of the five dysregulated miRs ( Figure 5B) on growth, clonogenic and myogenic potential of myxoma-derived CSCs (n=3) through selective gain and loss of function experiments based on the infection of lentiviral particles to transduce green fluorescent protein (GFP) together with a specific miRNA mimic or inhibitor (and a puromycin resistance cassette). At 50 MOI, we infected a large majority of myxoma-derived CSCs and with a single dose of puromycin we reached over 90% of GFP expressing cells from each specific miRmimic and -inhibitor (Supplementary material online, Figure S4A). Forty-eight hours after lentiviral infection, we accordingly obtained significant overexpression of the relative miR-mimic (Supplementary material online, Figure S4B). More importantly, cell growth analysis of myxoma-derived CSCs shows that out of the five different miR stable overexpression or suppression, miR-126-3p inhibition was the only treatment able to rescue myxoma-derived CSCs cell growth retardation at P12 ( Figure 6A). Indeed, miR-126-3p-suppressed myxomaderived CSCs had a significant higher growth kinetics when compared with control Lenti-GFP infected and transduced cells ( Figure 6B). Accordingly, clonal efficiency analysis by single-cell deposition show that miR-126-3p-inhibited myxoma-derived CSCs has higher clonogenic capacity when compared with control Lenti-GFP infected and transduced cells ( Figure 6C). Of interest, mRNA Seq analysis of myxoma-derived cells show that miR-126-3p up-regulation in these cells when compared with normal hCSCs is associated with specular down-regulation of several genes involved with cell cycle/growth and in particular of plexin-B2 (PLXNB2) (Supplementary material online, Figure S4C), which has been recently shown to play a key role in growth, survival and regenerative capabilities of normal haematopoietic and leukaemic stem and progenitor cells. 47 On the other hand, miR-335-5p inhibition was the only miRmodulating lentiviral infection that fostered robust myxoma-derived CSC myogenic commitment ( Figure 6D). Indeed, while miR-126-3p inhibition and miR-138 up-regulation resulted in the up-regulation of single myogenic genes ( Figure 6D and Supplementary material online, Figure S4D), miR-335-inhibitor treated myxoma-derived CSCs show a coherent and significant up-regulation of the main cardiac transcription factors (GATA-4, MEF2C and NKX2.5,) ensuing to an upregulation of the main contractile genes (MYH6, MYH7, TNNT2, and ACTC1) in vitro upon 14 days in myogenic differentiation media ( Figure  6D, E ). miR-335-5p was found down-regulated in unprimed cultured myxoma-derived CSCs when compared with human CSCs (see Figure 5B). Thus, miR-335-5p modulation could explain at least in part the abortive myogenic differentiation of myxoma cells. On the other hand, completely blocking miR-335-5p by a specific inhibitor could be a potential cardio-regenerative compound to enhance new myocyte formation from stem and progenitor cells and it remains mandatory to evaluate in future works the actual molecular target(s) that miR-335-p suppresses to regulate cardiomyogenic specification.
c-kit pos /CD45 neg /CD31 neg myxomaderived CSCs seed myxoma tumours in vivo Cancer-initiating (tumourigenic) cells typically form tumours after transplantation into NOD/SCID immunodeficient mice. 48 To assess the potential of myxoma-derived c-kit pos /CD45 neg /CD31 neg CSCs to seed cardiac myxomas in vivo, we transplanted 10 4 to 10 6 myxomaderived CSCs (engineered to express green fluorescent protein by lentiviral transduction), isolated from three different patients, through intramuscular injection into the quadriceps muscle of 27 NOD-SCID immunodeficient mice ( Figure 7A). Normal hCSCs (also GFP labelled) isolated from three different atrial specimens were used as controls and similarly transplanted ( Figure 7A).
Histologically, appearance of myxoma tumour tissue identified by Alcian Blue-PAS staining with the typical myxoma matrix with scattered cells, was directly proportional to the transplanted cell number ( Figure 7B and C). However, palpable masses were evident only in two out of nine mice at 36 weeks after injection of 10 6 cells ( Figure 7D). Importantly, scattered cells within the formed myxoid matrix stained positive for GFP, establishing their derivation from the injected myxoma-derived c-kit pos /CD45 neg /CD31 neg CSCs ( Figure 7E). Most of these GFP cells (68 ± 5%) co-stained for Calretinin ( Figure 7F), showing that myxoma-derived c-kit pos / CD45 neg /CD31 neg CSCs differentiated in the typical myxoma 'leipidic' cells in vivo. In striking contrast, the c-kit pos /CD45 neg /CD31 neg hCSCs from the normal controls were unable to form any palpable mass or any histological detectable tumour in any of the injected mice, independently of the cell number administered ( Figure  7C and G).
Overall, the data presented in this study was strongly supportive of the notion that the myxoma-derived c-kit pos /CD45 neg /CD31 neg multipotent cells are closely related (derived from?) the myocardial CSCs and that they are the 'tumour-initiating cells' of the myxoma. The results of the transplantation of these cells into immunosuppressed mice further support this notion and clearly show that these myxoma-derived c-kit pos /CD45 neg /CD31 neg CSCs seed de novo myxoma tumours in vivo (Take home figure).

Discussion
Data on mammalian and especially human cardiac neoplasias is very scarce because the myocardium is a privileged tissue which for still unknown reasons very rarely develops cancer. Therefore, information about the pathogenesis of heart neoplasias is particularly valuable to further understand its resistance to neoplastic transformation and the potential origin of those rare appearances. The main findings of the present study are: (i) c-kit pos /CD45 neg /CD31 neg cells, which are the characteristic phenotype of the multipotent CSCs, are present within atrial myxomas and fulfill all the criteria expected for 'myxoma initiating/stem cells'; (ii) The myxoma-derived c-kit pos / CD45 neg /CD31 neg multipotent cell exhibit a transcriptome and miRNA network which is similar to that of the CSCs from normal myocardium but with a relatively small fraction of dysregulated mRNAs and miRNAs (miR-126-3p and miR-335, in particular) involved in cell growth, differentiation and transformation pathways; (iii) myxoma-derived initiating/stem cells in vitro and in vivo produce the typical myxoid matrix; (iv) when transplanted into immunosuppressed mice, the myxoma-derived initiating/stem cells, in contrast to the control CSCs, produce myxoma-like tumours with a polysaccharide composition, calretinin expression, and cellular distribution characteristic of the human atrial myxomas.
Cardiac myxoma is a relative rare and histologically benign gelatinous mass, which despite its benignity has significant clinical relevance due its common presentation. 3 Indeed, myxomas can create lifethreatening haemodynamic instability, due the ability of the tumour progressively or acutely to obstruct pulmonary or venous drainage, to impair flow across the atrioventricular valves causing a filling defect 6,8 and/or embolism, due to tumour fragmentation, which typically cause cerebral embolism and transient/persistent neurological dysfunction. 6 While clinical course of cardiac myxomas has been very precisely assessed, there is still considerable uncertainty as to their cell origin. The tissue cell composition of myxomas is typically heterogeneous, Take home figure The main findings of the study showing that cardiac myxomas arise from multipotent c-kit pos /CD45 neg /CD31 neg myxoma tumour initiating cells and the working hypothesis of their direct derivation from normal c-kit pos /CD45 neg /CD31 neg cardiac stem/progenitor cells transformed by specific miRNA modulation.
with cells expressing protein markers specific to various cell lineages including epithelial, endothelial, myogenic, myofibroblast, neural and neuro-endocrine, often within the same mass. 10 The presence of CSX cardiac homeobox transcripts on its own suggested an endothelial derivation, 7 while the detection of both NKX2.5 and eHAND transcription factors within leipidic cells, suggested a myogenic cell origin 34 and variable expression of other markers that suggest neural or neuro-endocrine origins. 34 Therefore, it has been suggested that most likely cardiac myxomas derive from a pluripotent/multipotent mesenchymal stem cell or an endothelial progenitor cell located around the fossa ovalis and surrounding endocardium. 10,34 Nevertheless, the precise nature of these putative stem/progenitor cell is yet unknown.
Recently, the relevant role that cells with typical stemness properties play in the development of various cancer types has been pointed out. 49 These cells, like the stem cells of normal tissues, are characterized by their clonogenicity, multipotency, and self-renewal capacity. Malignant cells exhibiting these properties have been then defined as cancer stem cells (CaSCs) or tumour-initiating cells. Presumably, the CaSCs originate from mutations in normal stem/progenitor cells which when multiply with or without aborted differentiation induce the formation and development of some primary tumours.
Independently, various research groups have shown a regenerative process in the adult mammalian heart based, at least in part, on the existence of CSCs which can be identified and harvested from human heart tissues, as well as the different mammalian species studied. 19,21,50 Unfortunately, ongoing research on CSC biology and regenerative potential has been jeopardized by recent scandals and controversies. 51,52 Moreover, several genetic cell-fate mapping studies in mice have claimed that endogenous regenerative potential of CSCs is minimal. 16,53,54 These cell-fate mapping studies have been shown to fail to test the fate of the CSCs 17,18,55,56 and, consequently, the precise contribution of adult CSCs to endogenous cardiomyocyte replacement in homeostasis or after injury remains controversial. However, the evidence that myocardial tissue harbours cells which in vitro and in vivo are self-renewing, clonogenic, and multipotent showing CSC potential is uncontroverted.
On this basis, through cell and molecular characterizations of cardiac myxoma tissue we have shown here that these tumours harbour a cohort of c-kit pos /CD45 neg /CD31 neg cardiac cells. A cell type which in normal mammalian tissue is highly enriched for CSCs. 21 In myxomas, these cells are characterized by the expression of transcription factors and proteins typical of both cardiac progenitor as well as differentiated cardiac lineages. In addition, a small fraction of these cells within or isolated from the myxoma mass also express calretinin, a neuronal marker, which is highly expressed in polygonal myxoma cells. 34 Furthermore, at early passages, the c-kit pos /CD45 neg /CD31 neg myxoma-derived CSCs possess proliferative and clonogenic capacity similar to normal human CSCs. However, myxoma-derived CSCs, while secreting the typical mucopolysaccharides that compose the myxoma gelatinous mass in vivo, 34 progressively lose their selfrenewal capability and accumulate cell-cycle retardation and increased apoptotic rate through increasing passages in vitro. In addition, myxoma-derived CSCs exhibit an abortive differentiation potential in vitro whereby they undergo a multi-linage differentiation programme failing to specify into definitive cardiac cell lineages.
Although hampered by the difficulty and limitation of working with samples of human tissue which makes the data presented here mainly descriptive and falling short of being able to establish direct cause-effect relationship(s) (as it is also the case for most work based on primary human tissue), our findings strongly point to c-kit pos /CD45 neg / CD31 neg myxoma-derived CSCs as the progeny of the CSCs present in the normal myocardium and as the initiating/stem cell of cardiac myxomas which in addition to their gelatinous structure are characterized by a low proliferative index and the presence of a heterogeneous abortively differentiated cell phenotypes. This conclusion is further strengthened by the fact that xenotransplantation of c-kit pos / CD45 neg /CD31 neg myxoma-derived cells into immunodeficient mice seed the formation and growth of myxoma in vivo, highly suggesting that these cells are myxoma-initiating/stem cells.
Finally, from a clinical perspective, diagnosis of a cardiac myxoma is standardly obtained by advanced cardiac imaging 57, 58 . Surrogate biomarkers (like circulating microRNAs) 59 may help to optimize even at early stages the diagnosis and treatment modalities. Therefore, it remains highly tempting to speculate that miR-126 up-regulation in myxoma initiating CSCs, as detected here, could indeed represent a highly valuable serum biomarker for developing myxoma tumours. In addition, although surgical removal of the tumour mass is a curative therapy, in the cases of myxoma relapses, local administration of specific miR mimics/inhibitors (see Figure 6) to locally target c-kit pos / CD45 neg /CD31 neg myxoma tumour initiating cells could be theoretically envisioned as a specific therapy for myxoma recurrences.

Conclusions
Overall, our findings strongly support the hypothesis that cardiac myxomas are a stem cell-derived tumour arising from mesenchymal multipotent resident CSCs. Thus, it is proposed that atrial myxomas are a CSC-derived cardiac disease.

Supplementary material
Supplementary material is available at European Heart Journal online.