Article reSeArcH
high expression of MkP-related genes^26 (HSClowMkPhigh) (Fig. 2e,
Supplementary Table 2). Even within the MkP cluster, HSClowMkPhigh
cells showed increased cell-cycle gene expression and a higher fre-
quency of mutant cells, compared to HSChighMkPlow cells (P = 0.01,
Fisher’s exact test). Similarly, mutant cells within the platelet-primed
HSPCs (MkPhighTGFβlow) had higher cell-cycle gene expression than
wild-type MkPhighTGFβlow HSPCs, whereas mutant and wild-type
HSPCs in a more-quiescent state (MkPlowTGFβhigh) showed no dif-
ference in cell-cycle gene expression (Fig. 2f). These findings further
emphasize that the effect of CALR mutations is dependent on cell state,
and imparts a greater proliferative advantage in more-differentiated
cells. These data also reveal that CALR mutations skew differentiation
towards myeloid progenitors—including megakaryocytic priming—
early during haematopoiesis (Extended Data Fig. 6b).Cell identity and the effects of CALR mutations
GoT data further offer an opportunity for de novo discovery of differ-
entially expressed genes, by examining wild-type versus mutant cells
within the same subset of progenitors. Crucially, the wild-type cells
serve as an ideal comparison set, because they share all potential envi-
ronmental and patient-specific variables with the mutated cells. We
identified 198 genes that were differentially expressed between mutant
and wild-type MkPs (false-discovery-rate adjusted P < 0.1) (Fig. 3a,
Supplementary Table 3). Mutated MkPs upregulated HSPA5, whichencodes BiP, a key player in protein quality control that modulates
the activities of the three transmembrane transducers of the unfolded
protein response (UPR): PERK, IRE1 and ATF6^27. Consistently,
CALR-mutant MkPs showed upregulation of UPR genes (adjusted
P = 1.7 × 10 −^8 ) and ATF6-mediated activation of chaperone genes
(adjusted P = 0.03) (Fig. 3a, Supplementary Table 4), which provides
direct in vivo validation of previous in vitro studies that have shown an
increased UPR in CALR-mutated cells^28 ,^29. The UPR in this context
may signal endoplasmic-reticulum stress in response to misfolded
proteins, as the chaperone activity of CALR may be compromised by
the mutation^30 ,^31. Notably, among 20 differentially expressed genes
in mutant HSPCs, XBP1—an important regulator of the UPR—was
upregulated in mutant cells, which suggests that UPR activation byc20406080(^0204060)
MkP module
HSC module
log 2 (cell-cycle module)
4 567
Genotype
WT MUT
Genotype
WTMUT
79%
P = 0.01 95%
e
d
f
TGFβ module
3
6
9
02040
MkP module
Genotype
WT MUT
246
log 2 (cell-
cycle module)
2
3
4
5
6
WT MUT
2
3
4
5
WT MUT
log 2 (cell-cycle module)
P = 0.024 P = 0.13
log
(S phase module) 2
0
4
8
0
HSPC WT
log 2 (G2/M phase module)
048
HSPC MUT
0
4
8
048
48
MkP MUT
048
MkP WT
3
5
7
log
(cell-cycle module) 2
P = 4.4 × 10–4
ƔƔƔƔƔƔ
ƔƔƔƔ
ƔƔ
ƔƔ
ƔƔ
ƔƔƔƔƔƔƔƔƔƔƔƔƔƔƔƔƔƔƔƔƔƔƔƔ
2
4
6
P = 0.015
0
500
1,000
1,500
02550
Difference of mean of cell-cycle module
Platelet count
(thousands per
μl) P = 0.048, R^2 = 0.77
WT MUT
WT MUT
a
ET01
ET02
ET03
ET04
ET05
HSPCMEP
EPMkP
2.5 7.5
WT –log^10 (P value)
MUT
P = 1.3 × 10–6
0
1
2
3
4
log
(IFITM3 2
)
0
1
2
3
4
log
(IFITM2 2
)
P = 3.5 × 10–6
b
WTMUT
Fig. 2 | CALR mutations result in a higher proliferative effect on MkPs
compared to HSPCs. a, Expression of representative genes upregulated
in JAK2-mutated essential thrombocythaemia cultured cells^22 in CALR
wild-type (n = 157) versus mutant (n = 85) MkPs from a representative
sample ET01. b, Heat map of −log 10 (P value) from comparisons (between
mutant and wild-type cells) of genes expressed in JAK2-mutated essential
thrombocythaemia cells (Supplementary Table 6). c, Cell-cycle module
expression in HSPCs (n = 108 wild type versus n = 240 mutant) and
MkPs (n = 25 wild type versus n = 276 mutant) from ET03 (Extended
Data Fig. 6a). d, Platelet counts versus difference of mean cell-cycle score
(± s.e.m.) between wild-type and mutant MkPs (n = 5 samples; F-test).
e, Expression of MkP and HSC modules in MkPs from sample ET03.
Pie charts of wild-type versus mutant cell frequencies in HSClowMkPhigh
(n = 121 cells) and HSChighMkPlow (n = 28 cells) populations. Fisher’s
exact test, two-sided. f, Expression of TGFβ and MkP modules in HSPCs
from sample ET01, and cell-cycle score in HSPCs in MkPhighTGFβlow
(n = 127 wild type versus n = 41 mutant) and MkPlowTGFβhigh (n = 105
wild type versus n = 15 mutant) populations. P values for a–c, f are from a
two-sided Wilcoxon rank-sum test.
ER stress
UPR
ATF6
c d
IRE1
ER stress
Unspliced
XBP1 mRNA
Spliced
XBP1 mRNA
Spliced region
5 ′... ... 3 ′ 5 ′... ...^3 ′
GoT
Barcode
Unspliced
XBP1 cDNA
Spliced
XBP1 cDNA
5 ′
1.3 kb
3 ′ UTRe4
e
0
Density^10
EP
MkP
IMP
NP
HSPC
0
1
2
3
4
5
0
Pseudotime
UMI
MUT
WT
XBP1s
XBP1s/XBP1u
XBP1s/XBP1u
2
4
6
log
(NF- 2
κB module)
6.5
7.5
8.5
log
(anti-apoptosis 2
module)
f
Cell cycleHSC
TGFβ
HSPC1HSPC2HSPC3 Expression
High
Low
P = 5.3 × 10–^4
P = 0.001
P = 7.1 × 10–^4 P = 5.5 × 10–^6
P = 0.24
P = 2.1 × 10–^4
(n = 143) (n = 418) (n = 486)
a MkP HSPC
NFKBIA
CXCL2
CD69
LST1
SERPINB1
HPGDS
XBP1
TSC22D3
KLF2 SOD2
VIM
CALR
RHOH
ANKRD28
0
2
4
6
8
–1 01
Fold change (expressed in log 2 )
HSPD1
CALR
CCT5
HSP90B1
TUBB4B
HSP90AA1
HSPA8
HSPA5
NUDC
CHAF1A
PPP1R14ALAT
LDHA
JUN
TXNIP
SOX4
FCER1A
0
2
4
6
8
–1 0 1
Fold change (expressed in log 2 )
–log
(adjusted 10
P value)
–log
(adjusted 10
P-value)
NF-κB signalling
Adj P = 0.03
Higher in WT Higher in MUT b Higher in WT Higher in MUT
ATF6-mediated UPR
Adj P = 0.03
Unfolded protein binding
Adj P = 1.7 × 10–8
PIM2
MkP HSPC
0
5
10
15
WT MUT
P = 2.3 × 10–4
MkP
P = 0.0064
0
2
4
6
WT MUT
HSPC
IRE1–XBP1
4
5
3
4
5
P = 5.3 × 10–10
P = 1.9 × 10–6
WT MUT
WT MUT
log
(XBP1 module) 2
PERK
0
1
2
3
0
2
4
P = 0.062
P = 0.99
WT MUT
WT MUT
log
(PERK module) 2
3
4
5
3
4
WT MUT
5
P < 10–10
P = 5.9 × 10–7
WT MUT
log
(ATF6 module) 2
WT
MUT
Total XBP1
WT
MUT
Fig. 3 | Transcriptional effects of CALR mutation are dependent on
cell identity. a, b, Differentially expressed genes between mutant and
wild-type MkPs (a) and between mutant and wild-type HSPCs (b) across
ET01–ET05 samples (Supplementary Table 6). P values combined using
Fisher’s combined test with Benjamini–Hochberg adjustment. Key gene-
set enrichments are shown (hypergeometic test, Methods). c, Expression
of genes upregulated in UPR branches in MkPs (n = 442 wild type
versus n = 640 mutant) and HSPCs (n = 1,704 wild type versus n = 613
mutant) from samples ET01–ET05. P values from likelihood ratio tests
of linear mixed model, with and without mutation status (Methods). ER,
endoplasmic reticulum. d, Schematic of GoT applied to the XBP1 splice
site. e, Left, local regression of total and spliced XBP1 (XBP1s) expression
in progenitor cells from samples ET03 and ET04 (n = 1,308 wild
type versus n = 1,514 mutant; shading denotes 95% confidence interval).
Right, ratio of XBP1s to unspliced XBP1 (XBP1u) in MkPs (n = 115
wild type versus n = 248 mutant) and HSPCs (n = 489 wild type versus
n = 302 mutant). f, Expression of NF-κB pathway and anti-apoptosis genes
in HSPC1 (n = 116 wild type versus n = 27 mutant), HSPC2 (n = 365
wild type versus n = 53 mutant) and HSPC3 (n = 381 wild type versus
n = 105 mutant) from sample ET01. P values for e, f are from a two-sided
Wilcoxon rank-sum test.
18 JUlY 2019 | VOl 571 | NAtUre | 357