CRISPR Gene Editing in Stem Cells

reduced overall efficiency (in contrast to low-risk MDS) (Figures

2C and S3D–S3F). On the other hand, MDS/AML-iPSCs gave

rise to CD45

HPCs with efficiencies comparable to those of

normal cells but failed to differentiate further and retained

CD34 expression until day 18 and beyond (Figures 2C and

S3D–S3F). Reciprocally, CD90 expression, normally lost by

day 10–12 of differentiation, was retained by low-risk MDS,

high-risk MDS, and MDS/AML cells in a stage-specific manner

(Figures 2D, S3G, S4A, and S4B).

We found that megakaryocyte progenitors with the CD41a

CD45 surface phenotype that give rise to CFU-Mk colonies

emerge in these cultures on or before day 8 in normal cells (Fig-

ures 2E–2G, S4B, and S4C). Strikingly, this population was

severely decreased already in preleukemic cells and effectively

abolished in MDS (Figures 2E, S4B, and S4C). While normal

iPSCs gave rise to all types of hematopoietic colonies in methyl-

cellulose cultures (Figures S5A and S5B), iPSCs from all disease

stages exhibited reduced clonogenicity, with erythroid and mul-

tilineage colonies (burst-forming unit-erythrocyte [BFU-E] and

colony-forming unit-granulocyte, erythrocyte, monocyte, mega-

karyocyte [CFU-GEMM]) primarily affected already in preleuke-

mic and, more so, in low-risk MDS cells, while high-risk MDS

generated very few or no colonies (Figures 3A and 3B). MDS/

AML cells gave rise exclusively to myeloid colonies composed

mostly of immature cells (Figures 3A and 3B). Morphologic

assessment of HPCs on day 14 of differentiation and of more

mature cells from methylcellulose cultures revealed dysplastic

changes, which were milder and restricted to the erythroid line-

age in preleukemic and low-risk MDS and more widespread and

affecting all lineages in high-risk MDS cells (Figures 3B and S5C).

Finally, we measured the growth rate and viability of HPCs

derived from the different disease stage iPSCs. Low-risk MDS

showed a mild decrease in growth rate and viability, which was

much more pronounced in high-risk MDS, whereas growth and

viability of MDS/AML cells was completely restored to normal

levels (Figures 3C, 3D, and S5D–S5F).

MDS/AML-Derived Hematopoietic Cells Give Rise to

Serially Transplantable Leukemia

To assess in vivo engraftment potential, we then transplanted

day 8–16 HPCs derived from iPSCs of the various disease stages

into

NOD/SCID/IL-2R

(NSG) mice (Figure 4A). As expected

from many previous studies, HPCs derived from normal iPSCs

showed no detectable engraftment (Vo and Daley, 2015) (Figures

4B and 4C). Similarly, HPCs from MDS iPSCs (both low-risk

and high-risk) did not exhibit engraftment potential (Figures 4B

and 4C). In contrast, MDS/AML-HPCs showed high levels (up

to 80%) of human engraftment in multiple animals (Figures 4B

and 4C). The transplantable cells showed features of myeloid

leukemia, including a predominantly myeloid immunophenotype

and infiltration of the bone marrow and spleen by immature hu-

man CD45

cells with blast-like morphology, also found in the

peripheral blood, which could be transplanted into secondary

recipients (Figures 4D–4H). The latter readily succumbed to an

AML-like disease within 3 weeks of transplantation.

In summary, our phenotypic analyses show that iPSCs derived

from distinct disease stages across the myeloid malignancy

spectrum capture hematopoietic phenotypes of graded severity

and/or stage specificity that together delineate a phenotypic

roadmap to myeloid transformation, ultimately leading to a

fulminant serially transplantable myeloid leukemia (Figure 4I;

Table S3).

Transcriptomes of Disease-Stage-Specific iPSC-

Derived HPCs Recapitulate Features of Disease

Progression

We performed RNA sequencing (RNA-seq) of sorted CD34

cells

from three normal lines (two iPSC lines and the H1 hESC line) and

two low-risk MDS, three high-risk MDS, and three MDS/AML

iPSC lines (Figures 5 and S6A). By examining the gene expres-

sion profile among the different disease stages, we found a clear

clustering of samples according to disease status by principal-

component analysis, with the first principal component sepa-

rating normal from MDS and the second separating the AML

from the MDS samples (Figure 5A). Differential expression ana-

lyses identified 472 upregulated and 329 downregulated, 868

upregulated and 284 downregulated, and 760 upregulated

and 439 downregulated genes among AML versus normal,

high-risk MDS versus normal, and low-risk MDS versus normal,

respectively (log2FC > 3 or log2FC < 3 and adjusted P value <

0.05; Figure S6B). Hierarchical clustering of all lines based on

these differentially expressed genes recapitulated progression

from normal to MDS/AML (Figure 5B). Based on analysis of

Gene Ontology (GO) categories, genes involved in positive regu-

lation of apoptosis and negative regulation of cell proliferation

became upregulated at the transition from normal to low-risk

MDS and subsequently downregulated upon transformation

to MDS/AML, in agreement with our phenotypic analyses (Fig-

ure 5C). Similarly, negative regulation of differentiation was a

category upregulated early on. By gene set enrichment analysis

(GSEA), we identified significantly enriched disease-specific and

shared functional pathways in the iPSC-derived HPCs repre-

senting the three different disease stages (Figure S6C; Table

S4). Notably, both the high-risk MDS- and MDS/AML- iPSC-

derived HPCs were significantly enriched for the high-risk MDS

deletion 7q gene set, consistent with both their respective dis-

ease state and their specific genetic makeup (Figure 5D). Addi-

tionally, the MDS/AML-iPSC-HPCs showed specific enrichment

for a gene set found in a subset of human AML patients that had a

worse clinical prognosis and contained chromosome 7 abnor-

malities (Valk et al., 2004) (Figure 5E). Overall, these data suggest

that gene expression programs found in HPCs derived from our

iPSC panel recapitulate disease progression and capture gene

expression signatures derived from primary samples from pa-

tients with myeloid malignancies.

Modeling Disease Stage Transitions

We next asked if this model and the phenotypes characterized

therein could guide modeling transitions between disease

stages, as readouts for disease progression or reversal. We first

analyzed an iPSC line derived from the high-risk MDS line MDS-

2.13 after spontaneous correction of the del(7q) (Kotini et al.,

2015) (Figure 6A). Following correction of the del(7q), this line

only harbors a

SRSF2

P95L mutation (and a

PHF6

mutation of

uncertain significance). Since the

SRSF2

P95L mutation is an

early event in MDS and alone not sufficient for the development

of MDS, the corrected line (MDS-2.A3C) would be predicted to

capture a preleukemic stage (Papaemmanuil et al., 2013).

Cell Stem Cell

, 315–328, March 2, 2017

319