Assessment of established techniques to determine developmental and malignant potential of human pluripotent stem cells PDF

ARTICLE DOI: 10.1038/s41467-018-04011-3 OPEN Assessment of established techniques to determine developmental and malignant potential of human pluripotent stem cells The International Stem Cell Initiative# 1234567890():,; The International Stem Cell Initiative compared several commonly used approaches to assess human pluripotent stem cells (PSC). PluriTest predicts pluripotency through bioinformatic analysis of the transcriptomes of undifferentiated cells, whereas, embryoid body (EB) for- mation in vitro and teratoma formation in vivo provide direct tests of differentiation. Here we report that EB assays, analyzed after differentiation under neutral conditions and under conditions promoting differentiation to ectoderm, mesoderm, or endoderm lineages, are sufﬁcient to assess the differentiation potential of PSCs. However, teratoma analysis by histologic examination and by TeratoScore, which estimates differential gene expression in each tumor, not only measures differentiation but also allows insight into a PSC’s malignant potential. Each of the assays can be used to predict pluripotent differentiation potential but, at this stage of assay development, only the teratoma assay provides an assessment of plur- ipotency and malignant potential, which are both relevant to the pre-clinical safety assess- ment of PSCs. Correspondence and requests for materials should be addressed to P.W.A. (email: p.w.andrews@shefﬁeld.ac.uk) #A full list of consortium members appears at the end of the paper. NATURE COMMUNICATIONS | (2018)9:1925 | DOI: 10.1038/s41467-018-04011-3 | www.nature.com/naturecommunications 1 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-04011-3 T he capacity to differentiate into derivatives of all three PSC pluripotency. Not only do truly pluripotent cells generate a embryonic germ layers are the central deﬁning feature of very wide array of derivatives in these tumors, but they are also all pluripotent stem cells (PSC), but assessing this property often organized into organoid structures reminiscent of those that remains a challenge for human cell lines. PSC were ﬁrst recog- appear during embryonic development24. However, both the nized as embryonal carcinoma (EC) cells in teratocarcinomas, production of teratomas as xenografts, and their detailed analysis, germ cell tumors that also contain a wide array of somatic tis- which requires appropriately trained specialists, is costly and time sues1–4. In a classic experiment, using a teratocarcinoma of the consuming, and may be limited by concerns over animal welfare. laboratory mouse characterized by Stevens5 Kleinsmith and Moreover, the teratoma assay, as routinely performed, does not Pierce6 provided the ﬁrst functional demonstration of plur- yield quantitative information on lineage differentiation poten- ipotency by showing that single cells from ascites-grown tial25, although gene expression analysis of the teratomas them- embryoid bodies (EBs) could generate tumors containing EC selves can supply more deﬁnitive analysis. cells together with somatic tissues. The connection between ter- In the current International Stem Cell Initiative (ISCI) study, atocarcinoma and normal embryos was subsequently established following discussion at an ISCI workshop attended by about 100 by experiments showing that embryos transplanted to extra- members of the human PSC research community, we carried out uterine sites inevitably develop into teratomas or retransplantable a comparison of these approaches for assessing pluripotency by teratocarcinomas7,8. The discovery that murine EC cells can conducting a series of assays with human PSC lines, both ES and participate in embryonic development when transferred to early iPS cells. PluriTest was used to assess the transcriptome of the mouse embryos to give rise to chimeric mice9 led to the reali- undifferentiated cell lines. For the EB assay, we chose one widely zation that EC cells have the developmental capacity of cells of used approach, the ‘Spin EB’ system21 and used an adapted the inner cell mass. This laid the groundwork for the derivation of lineage scorecard methodology22 to assess the results. The Spin embryonic stem (ES) cells from mouse embryos10,11 and later EB method provides for control of input cell number and good from human embryos12 and of induced PSC (iPSC) from dif- cell survival, and allows for differentiation under neutral condi- ferentiated human cells13,14. tions and under well-deﬁned conditions expected to promote In assessing mouse ES or iPS cell lines, pluripotency is func- differentiation towards ectoderm, mesoderm or endoderm. Dif- tionally deﬁned from the PSC. However, for human PSC, be they ferentiation in teratomas was appraised by both histological ES or induced pluripotent stem cells (iPSC) cells13,14, this fun- examination and by “TeratoScore”, a computational quantitation damental assay is by the cell line’s ability, when transferred to a of gene expression data derived from teratoma tissue26. preimplantation embryo, to form to a chimeric animal in which These blinded analyses, conducted by independent experts on all of the somatic tissues and the germ line include participating PSC-derived samples in four highly experienced laboratories, cells not available. Moreover, a variety of well characterized PSC, shows that each of these methods can be used to indicate plur- from both mice and primates have only a limited ability to par- ipotency and that each is able to detect some variation in ticipate in chimera formation, even though they can differentiate developmental potential among the cell lines. The choice of which into tissues of all three germ layers in teratoma and in vitro method(s) should be used must be dictated by the biological assays15. With the advent of technologies for producing large question posed and the future use of the PSCs in question. We numbers of human PSC16,17, some destined for clinical applica- propose a schema outlining the choice of methodology for par- tions, the need for rapid and convenient assays of a speciﬁc PSC’s ticular applications. pluripotency and differentiation competence has become paramount. Results The purpose of this study was to provide an authoritative Experimental design. To compare PluriTest, EB differentiation assessment of several established alternative techniques for and teratoma, assays under conditions that would reﬂect varia- determining the developmental potential of human PSC lines. bility between laboratories and cell lines, four separate, expert The PluriTest® assay18 (www.pluritest.org), is a bioinformatics laboratories in four countries carried out these studies on each of assay in which the transcriptome of a test cell line is compared to three different, independent PSC lines and a fourth cell line, H9 the transcriptome of a large number of cell lines known to be (WA09)12, which was common to all (Supplementary Table 1). pluripotent. This test can be carried out rapidly with small All the experimental material was processed centrally, with high- numbers of cells, an important consideration in the early stages of throughput RNA sequencing (RNA-seq), quantitative real-time establishing new PSC lines. PluriTest is able to exclude cells that PCR and histology, as well as bioinformatics analyses carried out differ substantially from undifferentiated stem cells, but does not by single-specialized laboratories. In total, we compared results directly assess differentiation capacity. Complementing Plur- from 13 PSC lines (seven ESC and six iPSC lines). iTest’s focus on the undifferentiated state, various methods have been developed to monitor differentiation of the PSCs themselves in vitro, including protocols that induce spontaneous differ- Genetic integrity. It has been suggested that karyotypically var- entiation of cells in either monolayer or suspension culture, or iant PSC might be associated with persistence of undifferentiated directed differentiation under the inﬂuence of speciﬁc growth cells in xenograft tumors27,28. As an important adjunct to the factors and culture conditions that promote the emergence of differentiation studies we took several approaches to assess the particular lineages19,20. One of the most common approaches has genetic integrity of the cell lines. Prior to initiating the experi- been the use of differentiation in suspension culture, when clus- ments, the four test laboratories conﬁrmed that the cell lines they ters of cells undergo differentiation to form embryoid bodies planned to use had normal diploid karyotypes, excepting (EB), often with some internal structure apparent21. EB differ- NIBSC5, which carried a gain of the chromosome 20q amplicon entiation has also been combined with gene expression proﬁling that has been previously described29. Gene expression data also and bioinformatic quantiﬁcation of gene signatures, giving rise to permitted evaluation of the genetic integrity of the cell lines at the the pluripotency scorecard assay22. Further development of this time they were used in the experiments. Over- or under- scorecard deﬁned a panel of 96 genes that identiﬁed the differ- representation of speciﬁc regions of the genome in the undiffer- entiation capacity of a given cell line more quantitatively than the entiated PSC lines was evaluated using expression karyotyping (e- typical histology-based teratoma assay23. The teratoma assay has Karyotyping)30. Of the 13 cell lines, only one, the ES cell line long been regarded as the ‘gold standard’ for assessing human MEL1 INSGFP/w, showed an aberrant e-karyotype containing 2 NATURE COMMUNICATIONS | (2018)9:1925 | DOI: 10.1038/s41467-018-04011-3 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-04011-3 ARTICLE a Total Median MEL1 INS GFP/w 0.75 Gene expression ratio (log2) 0.50 0.25 0.00 –0.25 –0.50 –0.75 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 2122 b 2.0 H9 Allelic ratio All Labs 1.5 1.0 HES3 MIXL1GFP/w 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 2.0 Allelic ratio 2.0 KhES-1 Allelic ratio 1.5 1.5 1.0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 1.0 16 17 18 19 20 21 22 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 2.0 GFP/w MEL1 INS Allelic ratio 2.0 Lab II 201B7 Allelic ratio 1.5 Lab I 1.5 1.0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 1.0 16 17 18 19 20 21 22 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 2.0 Allelic ratio RM3.5 2.0 TIG108-4F3 Allelic ratio 1.5 1.5 1.0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 1.0 16 17 18 19 20 21 22 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 2.0 Oxford-2 Allelic ratio 2.0 iPS(IMR90)-4 Allelic ratio 1.5 1.5 1.0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 1.0 16 17 18 19 20 21 22 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 2.0 NIBSC5 Allelic ratio 2.0 H14 Allelic ratio Lab IV Lab III 1.5 1.5 1.0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Lab I 1.0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 16 17 18 19 20 21 22 2.0 Shef3 Allelic ratio 2.0 DF19-9-11.T.H Allelic ratio 1.5 1.5 1.0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 1.0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 Chromosome Chromosome Fig. 1 Detection of chromosomal aberrations in PSC and tumors using e-Karyotyping and eSNP-karyotyping. a e-Karyotyping: each line depicts the moving average plots of global gene expression in 13 different cell lines over 300-gene bins. The gene expression of 12 cell lines (black lines) was close to the total mean, suggesting a normal karyotype. In contrast, all replicates of the MEL1 INSGFP/w (cyan) cell line showed considerable upregulation of genes from both chromosomes 12 and chromosome 17, suggesting that it harbors an additional copy of these chromosomes. b eSNP-karyotyping: detection of chromosomal aberrations in tumors using eSNP-karyotyping. Each line depicts the moving average (over 151 SNPs) of gene expression generated from RNA-seq data of tumor derived from 13 different cell lines (one plot per source cell line). Colors represent tumor replicates. Only tumors derived from MEL1 INSGFP/w and NIBSC5 show an altered allele ratio in both replicates, suggesting an aberrant karyotype with additional copies of chromosomes 17 and 12, respectively NATURE COMMUNICATIONS | (2018)9:1925 | DOI: 10.1038/s41467-018-04011-3 | www.nature.com/naturecommunications 3 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-04011-3 a b c H9(WA09) samples - all labs Lab 1 201B7 Pluripotency & Novelty Lab 1 KhES-1 Lab 2 Empirical threshold: 50 model fit Lab 3 Lab 4 Tig108-4F3 20 0 d Lab 2 e Lab 3 DF19-9-11T.H Pluripotency score HES3 MIXLGFP/w iPS(IMR90)-4 iPSC RM3.5 H14(WA14) MEL1 INS GFP/w –50 f Lab 4 Shef3 –100 NIBSC9 Oxford-2 Differentiation 1 2 3 4 Empirical threshold:1.67 Novelty score Fig. 2 Pluritest. a All PluriTest results from this study (red circles) are based on normalization to the H9 samples and were plotted on the background of the empirical density distribution of all pluripotent (red cloud) and differentiated samples (blue clouds) in the PluriTest training dataset18. b–f highlight the subsets of samples included in this study: All results from the same hPSC line (H9) cultured at each laboratory (b). Samples from Lab 1 (c), Lab 2 (d), Lab 3 (e), Lab 4 (f) are highlighted speciﬁcally. All cell lines are above the Pluripotency Score threshold (θP >= 20). Both replicates of two cell lines MEL1 INSGFP/w in d and DF19-9-11T.H in e score above the Novelty threshold (θN >= 1.67) and thus would be highlighted for further investigation. Three cell lines show larger differences between the novelty scores of their respective replicate samples 201B7 in c, RM3.5 C in d, and Oxford-2 in f extra copies of chromosomes 12 and 17 (Fig. 1a). These dis- that detects the presence of gene expression patterns usually not crepancies from the test laboratories reports for NIBSC5 and associated with human PSC. A pluripotent cell line is character- MEL1 INSGFP/w most likely reﬂect the sensitivities of different ized as passing the PluriTest if it simultaneously exhibits a high assays for detecting low level genetic mosaicism31 and the pro- Pluripotency and a low-novelty score. If the scores of a test cell pensity of variants to overgrow the culture rapidly once they line deviate from the empirically determined Pluripotency and appear32. Consistent with this interpretation, the MEL1 INSGFP/w Novelty thresholds, the sample is ﬂagged for further investigation. is known to exhibit karyotypic instability in culture (RM, EGS, As the original PluriTest algorithm was developed for an older and AGE, unpublished results). Because of the heterogeneous cell Illumina BeadChip platform, it was adapted to a new platform composition of teratomas a different methodology is required to using the H9 samples from all four laboratories as a control for evaluate the chromosomal integrity of the cells comprising them. technical variation (Supplementary Fig. 1). Analyzing samples eSNP-karyotyping enables a direct analysis of chromosomal with the updated PluriTest script, showed that at least one aberrations by calculating the expression ratio of SNPs, making it replicate of most lines assayed passed both PluriTest criteria less sensitive to global gene expression changes between different (Fig. 2; Supplementary Fig. 1). samples33. eSNP-Karyotyping of the teratomas indicated that In the case of cell lines RM3.5 and Oxford-2, while we observed most remained karyotypically diploid, but also revealed that high-Pluripotency Scores in both replicates (Fig. 2), there was a teratomas derived from NIBSC5 had additional copies of chro- large difference in the Novelty Scores between the two replicates, mosomes 12 (and perhaps 20), and that teratomas derived from placing one replicate above the empirical threshold for the MEL1 INSGFP/w carried an additional copy of chromosome 17, Novelty Score (1.67). A similar result was obtained for one of the but not chromosome 12 (Fig. 1b). Extra copies of human chro- two replicates from the 201B7 cell line. The differences in Novelty mosomes 12, 17, and 20 are recurrent changes in cultured PSCs, score observed between replicates could be due to technical and have also been reported in human germ cell tumors29. These failures of the array hybridization, or it could reﬂect differing changes likely reﬂect a selective advantage conferred by extra extents of spontaneous differentiation in the cell line samples copies of genes on these chromosomes to cells grown either analyzed. Nevertheless, we concluded that all cell lines with one in vitro or in vivo34,35. Taken together our results suggest that replicate below the empirical Novelty Score threshold passed cultures of NIBSC5 and MEL1 INSGFP/w, but of none of the PluriTest and are predicted to have pluripotent differentiation other 11 lines, were initially mosaic containing low levels of potential in vitro and in vivo. However, the PSC lines DF19-9- variant cells. 11T.H and MEL1 INSGFP/w did not pass the empirically determined Novelty Score threshold of 1.67, thus ﬂagging them PluriTest analysis. PluriTest was used to assess the molecular for further investigation. Interestingly, the MEL1 INSGFP/w PSC similarity of the different undifferentiated cell lines to that of line did have an abnormal e-Karyotype (Fig. 1a, b), providing a other known PSC lines. RNA samples were analyzed using the possible explanation for its borderline results in PluriTest. Illumina Human HT-12 v4 Expression BeadChip and subjected to the PluriTest algorithm18. PluriTest generates two summary Scorecard analysis of embryoid body differentiation in vitro. scores from global gene expression proﬁles: a pluripotency score The participating laboratories also subjected their cell lines to a that predicts whether a cell sample is pluripotent based on the standardized embryoid body (EB)-differentiation protocol under similarity of its gene expression signature to gene expression four different conditions: neutral, without the addition of exo- proﬁles of a large collection of human PSC; and a novelty score genous growth factors that favored any particular lineage, and 4 NATURE COMMUNICATIONS | (2018)9:1925 | DOI: 10.1038/s41467-018-04011-3 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-04011-3 ARTICLE directed conditions designed to promote initial differentiation necessarily support the generation of more mature cell types. into ecto-, meso-, or endoderm lineages, respectively21. It was Lysates from the resulting EBs were examined by qRT-PCR at 0, anticipated that these protocols would be sufﬁcient to direct 4, 10, and 16 days of differentiation for expression of 190 genes differentiation toward the germ layer of interest but would not (Supplementary data 2, 3) modiﬁed from the set used by a marker gene expression Ectoderm Mesoderm Endoderm Undifferentiated Mean change of 5.0 2.5 0.0 –2.5 –5.0 0 4 10 16 0 4 10 16 0 4 10 16 0 4 10 16 Time in differentiation condition (days) b Scorecard Propensity Potential (spontaneous) (directed) Lab 1 H9 (Lab 1) + + + ++ ++ ++ KhES-1 + +/– +/– ++ ++ + 207B1 + +/– +/– ++ + ++ Tig108 4f3 + + + + ++ + Lab 2 H9 (Lab 2) + +/– + ++ +++ ++ HES3/MIXL1GFP/w + + + ++ + + MEL1/INSGFP/w ++ + + ++ ++ + RM3.5C + +/– +/– ++ ++ ++ Lab 3 H9 (Lab 3) + + + + ++ + H14 + ++ + ++ + + DF19.9-11T.4 nd nd nd nd nd + iPS(IMR90)-4 ++ + ++ +++ ++ + Lab 4 H9 (Lab 4) +++ +/– + + ++ + Shef3 + + ++ ++ ++ + Oxford-2 + + + + + +/– NIBSC 5 + + ++ ++ ++ + Ecto Meso Endo Ecto Meso Endo c Ectoderm Mesoderm Endoderm 3 Lineage score: teratoma 2 1 0 0 1 2 3 0 1 2 3 0 1 2 3 Lineage score: spontaneous EB differentiation (16 days) Fig. 3 Differentiation potential and propensity in EBs. a The line plots show the mean log2 expression change (relative to day 0) of marker genes (Supplementary table 3) as a function of time and averaged over all cell lines. The expression change is shown under ectoderm conditions for ectoderm markers, mesoderm conditions for mesoderm markers, endoderm conditions for endoderm markers, and across all conditions for markers of undifferentiated cells. Shaded contours indicate the minimum/maximum observed value. b A summary table of the lineage scorecard evaluation of the “propensity” (spontaneous differentiation, left) and “potential” (directed differentiation, right) for each cell line (rows) to differentiate into the respective lineage (columns). Colors and symbols indicate increased (blue) and limited (grading of lighter blues) preference for expression of lineage- speciﬁc marker genes. +++: score >3; ++: score 2–3; +: score 1–2; +/−: score 75%) b For Teratoscore, the percentage ofectoderm, mesoderm, endoderm, and extraembryonic speciﬁc-gene expression is summarized in comparison to the mean percentage of 4 pilot, karyotypically normal teratomas: ‘+’ (the pilot expression mean) ‘++’ (similar to the pilot expression mean) c The presence of undifferentiated cells (ECL) and/or yolk sac elements (YS), assessed by both histology and by RNA-seq analysis is indicated by ‘+’, in cells that are highlighted in yellow Bock et al.22, to include genes characteristically expressed in programs, though some (KhES-1, 201B7, RM3.5C, and H9 from undifferentiated PSC, extraembryonic endoderm, trophectoderm, Labs 2 and 4) had reduced propensities to form one or both of early deﬁnitive ectoderm, mesoderm, and endoderm. For each these latter lineages, an apparent bias not recapitulated in the lineage and for undifferentiated cells, we picked an equal number teratoma assay (Table 1 below). Second, ectoderm-inducing and (n = 15) of marker genes for further analysis (Supplementary mesoderm-inducing conditions elicited strong, homogeneous Table 2), by focusing on those genes with the strongest lineage- expression signatures consistent with the expected directed speciﬁc upregulation of genes in our dataset (Methods section). lineage, while endoderm-inducing conditions elicited more These marker genes were generally more highly expressed in EBs variable responses, depending on both the cell line and on the cultured under the corresponding differentiation conditions, laboratory, a result most marked in the Oxford-2 line. Third, the while expression of markers of undifferentiated cells gradually data suggest that, overall, all cell lines were capable of dropped (Fig. 3a, Supplementary Fig. 2a). Gene expression was differentiating into representatives of all three lineages, although least variable 4 days after induction of differentiation compared there were differences in how well and how consistently the PSC to other time points (Supplementary Fig. 2b, c). lines responded to these speciﬁc differentiation cues. The lineage scorecard analysis was carried out as described previously22 but with the reﬁned gene set (Supplementary Differentiation in xenograft teratomas in vivo. Each laboratory Table 3) and with one conceptual extension: the “potential” of produced between one and three xenograft tumors from each cell cells to undergo differentiation into the three primary lineages line, by subcutaneous injection into immunodeﬁcient mice, as under directed differentiation conditions was distinguished from described in Methods section (Supplementary Table 1). Although their “propensity” to differentiate under neutral conditions. The a common protocol was suggested for tumor production, local “potential” of a cell to differentiate into a certain lineage was circumstances mandated some modiﬁcations to this protocol in deﬁned as the lineage score at 16 days of directed differentiation each case, particularly with respect to the particular strains of culture conditions. That is, ectoderm induction was used for mice used as hosts. After cutting each tumor into several pieces, ectoderm marker proﬁling, mesoderm induction for mesoderm approximately half of them were randomly selected for histology, markers, and endoderm induction for endoderm markers. The while the other half was processed to provide RNA for RNA-seq “propensity” (or inherent bias) of a cell line to undergo and TeratoScore analysis. differentiation was calculated from the lineage scores (Methods All PSC-derived tumors were classiﬁed as teratomas, since each section) of all marker sets after 16 days in neutral differentiation contained tissues identiﬁed as derivatives of the three germ layers conditions. (Fig. 4a, b). Overall, a median of 10% (range, 5–30%) of the Scorecard analysis resulted in three key observations (Fig. 3b, differentiated tissues observed were of endodermal derivation, Supplementary Fig. 2a, b). First, in neutral culture conditions all 40% (range, 10–60%) represented tissues of mesodermal origin cell lines had the propensity to upregulate ectoderm markers, but and 45% (range, 10–80%) represented tissues of ectodermal all cell lines also initiated mesoderm and endoderm expression origin (Table 1 and Fig. 4c). Cells from all three embryonic germ 6 NATURE COMMUNICATIONS | (2018)9:1925 | DOI: 10.1038/s41467-018-04011-3 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-04011-3 ARTICLE P End Sq End Mes End Mes End Ect a b c H9-ESC ESC iPSC Tissue Lab1 Lab2 Lab3 Lab4 Lab1 Lab2 Lab3 Lab4 Lab1 Lab2 Lab3 Lab4 Ectoderm Neural 4/4 3/3 1/3 2/3 3/3 5/6 5/5 5/5 6/6 3/3 3/3 3/3 pig. epith. 4/4 3/3 1/3 2/3 4/6 3/5 1/5 4/6 1/3 1/3 1/3 sq. epith. 2/3 2/3 4/6 2/3 choroid pl. 4/4 3/3 Mesoderm Cartilage 3/3 1/3 1/3 1/3 4/6 3/5 1/5 4/6 1/3 2/3 3/3 Bone 1/3 1/3 3/6 2/3 Stroma 2/3 Fat 2/3 Mesenchy. 4/4 2/3 1/3 3/6 2/5 3/6 1/3 Muscle 2/3 Endoderm Glands 4/4 1/3 2/3 3/3 2/6 2/5 3/5 2/6 1/3 1/3 1/3 Ducts 3/3 2/3 1/3 3/6 3/5 3/6 Intestine 1/3 2/5 2/5 EB N N YS N ECL d ECL e f ECL YS ECL YS g h i Fig. 4 Histological evaluation of three embryonic germ layers and undifferentiated EC-Like and yolk sac elements in xenograft tumors. a Mucus secreting intestinal-like epithelium (End-endoderm), neural tube rosettes (Ect-ectoderm), and intervening stroma (Mes-mesoderm) (×240). b Intestinal-like epithelium (End-endoderm), surrounded by connective tissue, smooth muscle and fat cells (Mes-mesoderm). The left outer rim of mesoderm is lined by intestinal-like epithelium (End-endoderm). To the left there is pigmented epithelium (P), corresponding to retina (Ect-ectoderm), and a nest of glycogen rich squamous epidermal cells (Sq) (×120). c A summary of tissue types recorded per individual tumor piece surveyed from each laboratory; at least two pieces of each tumor were examined. d Lower magniﬁcation view of a teratoma containing undifferentiated stem cells (EC-Like, ECL), identiﬁed as embryonal carcinoma-like (ECL) cells, neural tube-like rosettes (N) and non-descript stromal cells (×120). e Higher magniﬁcation of the same xenograft. Undifferentiated ECL cells (ECL) are arranged into anastomosing cords. Dark dot-like cells are undergoing apoptosis. Compare the loosely structured chromatin of the ECL cells with the dark nuclei containing condensed chromatin in the neural rosettes (N) (×240). f Two embryoid bodies (EB) forming tubes lined by ECL cells, separated by a space from the surrounding yolk sac epithelium (YS). Both embryoid bodies contain prominent apoptotic bodies. Note the loosely textured yolk sac (YS) corresponding to the connective tissue that runs between the yolk sac and the blastocyst (magma reticulare) of early human embryos (×120). g Antibody to OCT3/4 staining ECL cell nuclei. h Antibody to the zinc-ﬁnger protein ZBTB16 reacts with the nuclei of yolk sac cells around three cylinders of ECL cells. i Antibody to SALL4 staining ECL cell nuclei and also the yolk sac (YS) cells in their vicinity layers were found in the teratomas, derived from both ES mesoderm, and glandular, ductal and intestine tissue for most and iPS cell lines produced by each of the laboratories. Although of the endoderm (Fig. 4c). all teratomas contained derivatives of the three embryonic Some teratomas also contained areas of undifferentiated cells, germ layers, in fact only a fairly narrow range of tissues was which we designated as embryonal carcinoma-like (ECL) cells, routinely identiﬁed. Neural tube-like structures, pigmented some exhibited areas of yolk sac elements, and some contained epithelium and squamous epithelium accounted for most cells in some areas organized into EB like structures (Fig. 4d–f). ectoderm, cartilage, connective tissue, and bone for most The histological identiﬁcation of the ECL was conﬁrmed by NATURE COMMUNICATIONS | (2018)9:1925 | DOI: 10.1038/s41467-018-04011-3 | www.nature.com/naturecommunications 7 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-04011-3 immunostaining for expression of OCT3/4 (POU5F1) (Fig. 4g) layers, and into extraembryonic elements, such as yolk sac12, or and the yolk sac cells by immunostaining for ZBTB16 (Fig. 4h)36. the PSCs may proliferate in which case they may be noted as ECL As expected, SALL4 expression was found in both yolk sac and cells, suggesting a potential malignant phenotype. In the clinical ECL cells37,38 (Fig. 4i). The initiating PSCs in teratomas may pathology of germ cell tumors (GCT), embryonal carcinoma and differentiate into derivatives of mature elements of all three germ yolk sac elements are frequently found in malignant a Lab 1 ESC Lab 2 iPSC 0.20 Lab 3 Lab 4 0.15 0.10 0.05 0.00 KhESC-1.T1 KhESC-1.T3 KhESC-1.T2 347 HES3/MIXL1GFP/w.T2 Shef3.T1b Shef3.T2 H9.T1 346 HES3/MIXL1GFP/w.T1 348 HES3/MIXL1GFP/w.T3 H9.T3 iPS(IMR90)-4.T2 iPS(IMR90)-4.T1 DF19-9-11.T.H.T1 TIG108-4F3.T1 TIG108-4F3.T2 H14.T1 H14.T2 H9.T1 RM3.5L.T1 RM3.5L.T2 iPS(IMR90)-4.T3 DF19-9-11.T.H.T3 MEL1 INSGFP/w.T3 MEL1 INSGFP/w.T2 MEL1 INSGFP/w.T1 RM3.5L.T3 207B1.T1 Shef3.T1a H9.T3 Oxford-2.T1 Oxford-2.T2 NIBSC5.T2b NIBSC5.T1 NIBSC5.T2a H9.T1 RNA-seq 340 342 341 362 363 376 356 370 369 374 344 345 372 373 355 353 352 371 375 351 350 349 354 343 361 377 359 360 368 364 366 357 identifier 12 10,000 30 30 TSG: 7.1 Tig108 4f3 TIG108-4f3 b Shef3 c d DF-19-9-11T.4 DF-19-9-11T.H KhES-1 20 20 IMR90-4 iPS(IMR90)-4 HES3/MIXL1GFP/w 10 10 Shef3 Shef3 Shef3 30 30 10 H14 TSG: 6.5 KhES-1 20 20 iPSC RM3.5 1000 Average % expression of original tissue / lineage 10 10 % Expression undifferentiated markers 30 30 TSG: 4.0 8 20 20 10 10 TeratoScore grade 100 30 30 TSG: 3.8 KhES-1 20 20 6 10 10 30 30 TSG: 6.5 20 20 10 4 10 10 30 30 TSG: 0.7 HES3/MIXL1GFP/w 20 20 10 10 2 1 30 30 TSG: 135 All teratomas 20 20 10 10 0 CPNS Skin Adip Heart Kdny Mscl Bld Gut Liver Lung Panc Plac Undf Ect Mes End ExEm Tissues Teratomas Tissues Teratomas Ect Mes End 8 NATURE COMMUNICATIONS | (2018)9:1925 | DOI: 10.1038/s41467-018-04011-3 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-04011-3 ARTICLE teratocarcinomas, (reviewed in ref. 4), while yolk sac and expression of these marker genes also clustered in a principal immature neural elements are commonly associated with component analysis (Supplementary Fig. 3). Of these lines, malignant transformation in teratomas of childhood39–42. It has TIG108 4f3 had been previously classiﬁed as ‘differentiation been proposed that the experimental teratomas produced by both defective’, and KhES-1 as ‘intermediate defective’, in an assay that mouse and human ES and iPS cells are more akin to GCT of the assessed the persistence of undifferentiated, OCT3/4+ (POU5F1) newborn (type 1 GCT), than to those of the adult (type 2 GCT)43. cells after a deﬁned period of speciﬁc neural induction in vitro48. This distinction correlates with the diploid or near diploid In concert with that report we noted that TIG108 4f3 tumors karyotypes of most ES and iPS cells, in contrast to the grossly showed higher levels of the stem cell markers than KhES-1 aneuploid karyotypes of human EC cells from adult germ cell tumors (Fig. 5b). Teratomas derived from ﬁve of these seven cell tumors. In the experimental teratomas at hand we ﬁnd it lines were found by histological analysis to contain ECL cells noteworthy that even when these potentially malignant elements (KhES-1 and TIG108 4f3) and/or yolk sac cells (KhES-1, TIG108 were observed, robust differentiation into tissues derivatives of all 4f3, H14, DF19-9-11T.H, Shef3) (Table 1). Overall, these results three germ layers was also seen within the same tumor. Histologic suggest that many of the teratomas contained differentiated evidence alone does not permit a deﬁnitive conclusion as to derivatives of extra-embryonic membranes and potentially ECL whether the ﬁnding of ECL cells intermingled with yolk sac cells. elements is indicative of the malignant potential of a subset of the The TeratoScore algorithm enables the use of teratoma gene PSCs tested, but it is certainly a cause for concern. Although, expression to provide a quantitative analysis of the ability of a most teratoma histological sections include differentiated struc- PSC line to differentiate26. This analysis quantiﬁes tissue-speciﬁc- tures from the three embryonic germ layers, an elaborate analysis gene expression within heterogeneous PSC-derived teratomas, of cell type would require further experimental investigation. thus providing an estimation of teratoma tissue composition. RNA samples were extracted from 44 of the 58 tumors Since TeratoScore was originally designed for microarray analysis, prepared. Of these, 35 samples passed the quality control tests for it was adapted in this study to analyze RNA-seq data. Similar to RNA quantity, purity and integrity required for RNA-seq and the original TeratoScore calculation, a 100-gene signature was further analysis. These samples represented tumors derived from created by identifying genes expressed in teratomas and speciﬁc 12 different independent cell lines (6 ESC, 6 iPSC), as well as to tissues representing derivatives of the three embryonic germ tumors derived from H9 in three of the four laboratories. An layers and the extra-embryonic membranes (Methods section; initial unbiased hierarchical clustering of all gene expression data Supplementary Data 4). Comparing expression of these genes in a was performed (Fig. 5a). Although tumors derived from the same teratoma to their respective expression level in normal tissues cell line sometimes clustered together, sometimes they did not, provides an estimate for the existence of cells from each tissue and data from tumors of the different cell lines, be they ES- or within the tumor, as well as a lineage expression proportion. By iPSC-derived, as well as those from H9-derived tumors, even if calculating the expression values from the different lineages, the from the same laboratory, were scattered throughout the TeratoScore provides a uniﬁed grade that weighs the different dendrogram suggesting there was no obvious laboratory effect tissue-speciﬁc expression within a teratoma and provides an (Fig. 5a). estimate of the ability of a PSC line to differentiate (Methods To assess whether there were residual undifferentiated PSC section). As expected, each individual normal tissue yielded a within the teratomas, we queried the RNA-seq datasets for high-expression level of its speciﬁc cell type and lineage expression of ten undifferentiated PSC markers (Supplementary (Supplementary Fig. 4), yet a low uniﬁed TeratoScore grade Table 3). These marker genes were initially selected based on (Fig. 5c). In contrast, teratomas show a relatively high score for all results previously published by the ISCI44 and include several cell types (Supplementary Fig. 4) and lineages, and also higher markers also expressed by yolk sac endoderm cells45–47. When TeratoScore grades (Fig. 5c). A TeratoScore grade >10 was expression of these genes in the teratomas was compared to that deemed sufﬁcient to determine that a given tumor was initiated of cultured undifferentiated PSC, their mean expression was from a PSC line capable of differentiating toward derivatives of found to be 2.5% of that in the PSCs (Supplementary Table 3). three germ layers in a relatively evenly distributed fashion, since Nevertheless, eleven teratoma samples, originating from no normal tissue exceeded this threshold (Fig. 5c). However, of seven different PSC lines (KhES-1, TIG108 4f3, RM3.5, the 35 teratomas tested, six samples originating from three PSC H14, DF19-9-11T.H, IPS(IMR90)-4, Shef3), exhibited substan- lines (Shef3, KhES-1, and HES3 MIXL1GFP/w) did not reach this tially higher average expression levels of these 10 markers, threshold (Fig. 5c). A closer look at the expression patterns from suggesting the presence of undifferentiated PSCs and/or yolk these teratoma samples revealed higher expression of neuroecto- sac elements (Fig. 5b). Those teratomas showing elevated dermal markers in KhES-1 and HES3 MIXL1GFP/w compared to Fig. 5 Teratoma RNA-seq expression data analysis. a Unsupervised hierarchical clustering analysis of RNA-seq expression of teratomas from four different laboratories (calculated using complete linkage and Spearman correlation distance). Tumors from the same laboratory appear in the same color. Label numbers (T1, T2, etc.) indicate teratoma replicates. Speciﬁc RNA-seq sample identiﬁers are indicated below the sample names. b Mean relative expression of human undifferentiated PSC/yolk sac markers within teratomas and normal tissues calculated with respect to their expression in PSCs. Eleven teratomas (highlighted by colored dots) showed an expression greater than teratoma overall average (2.5%). c TeratoScore grades, calculated f

Assessment of established techniques to determine developmental and malignant potential of human pluripotent stem cells PDF

Document Details

Tags

Related

Summary

Full Transcript

Upgrade to continue