RESEARCH ARTICLE

REGENERATIVE MEDICINE

Pharmacological targeting of kinases MST1 and MST2

augments tissue repair and regeneration
Fuqin Fan,1,2* Zhixiang He,1,2* Lu-Lu Kong,3* Qinghua Chen,1,2* Quan Yuan,4* Shihao Zhang,1,2
Jinjin Ye,1,2 Hao Liu,1,2 Xiufeng Sun,1,2 Jing Geng,1,2 Lunzhi Yuan,4 Lixin Hong,1,2 Chen Xiao,1,2
Weiji Zhang,1,2 Xihuan Sun,1,2 Yunzhan Li,1,2 Ping Wang,1,2 Lihong Huang,1,2 Xinrui Wu,1,2
Zhiliang Ji,1,2 Qiao Wu,1,2 Ning-Shao Xia,4 Nathanael S. Gray,5 Lanfen Chen,1,2 Cai-Hong Yun,3†
Xianming Deng,1,2† Dawang Zhou1,2†

Tissue repair and regenerative medicine address the important medical needs to replace damaged tissue with func-
tional tissue. Most regenerative medicine strategies have focused on delivering biomaterials and cells, yet there is
the untapped potential for drug-induced regeneration with good specificity and safety profiles. The Hippo pathway
is a key regulator of organ size and regeneration by inhibiting cell proliferation and promoting apoptosis. Kinases
MST1 and MST2 (MST1/2), the mammalian Hippo orthologs, are central components of this pathway and are, there-

Downloaded from http://stm.sciencemag.org/ by guest on October 11, 2018

fore, strong target candidates for pharmacologically induced tissue regeneration. We report the discovery of a re-
versible and selective MST1/2 inhibitor, 4-((5,10-dimethyl-6-oxo-6,10-dihydro-5H-pyrimido[5,4-b]thieno[3,2-e][1,4]
diazepin-2-yl)amino)benzenesulfonamide (XMU-MP-1), using an enzyme-linked immunosorbent assay–based
high-throughput biochemical assay. The cocrystal structure and the structure-activity relationship confirmed that
XMU-MP-1 is on-target to MST1/2. XMU-MP-1 blocked MST1/2 kinase activities, thereby activating the
downstream effector Yes-associated protein and promoting cell growth. XMU-MP-1 displayed excellent in vivo
pharmacokinetics and was able to augment mouse intestinal repair, as well as liver repair and regeneration, in
both acute and chronic liver injury mouse models at a dose of 1 to 3 mg/kg via intraperitoneal injection. XMU-
MP-1 treatment exhibited substantially greater repopulation rate of human hepatocytes in the Fah-deficient
mouse model than in the vehicle-treated control, indicating that XMU-MP-1 treatment might facilitate human
liver regeneration. Thus, the pharmacological modulation of MST1/2 kinase activities provides a novel approach
to potentiate tissue repair and regeneration, with XMU-MP-1 as the first lead for the development of targeted
regenerative therapeutics.

INTRODUCTION PGDH (SW033291) promoted tissue regeneration in mouse models

The mechanisms controlling tissue repair and regeneration have long of colon and liver injury (4). Subcutaneous injection of the drug
been of enormous interest and fascination to biologists (1, 2). Stem 1,4-dihydrophenonthrolin-4-one-3-carboxylic acid, which blocks the
cells enable the human body to develop, grow, and repair throughout degradation of hypoxia-inducible factor–1a into Swiss Webster mice
life. Although much progress has been made in exploring the thera- that do not show a regenerative phenotype, led to regenerative wound
peutic potential of induced pluripotent stem cells, embryonic stem healing after earhole punch injury (5). Administration of the recom-
cells, and mesenchymal stem cells, we are still unable to harness their binant growth factor neuregulin-1 to postnatal mice stimulated regen-
potential to repair or regenerate damaged human organs (3). Thera- eration of heart muscle cells, further improved myocardial function,
peutically enhancing tissue regeneration has proven to be challenging. and reduced the prevalence of transmural scars when mice were
However, recent studies demonstrated that several compounds capa- subjected to cryoinjury (6).
ble of promoting tissue regeneration in mice might have clinical ap- Some new targets for regenerative medicine have been identified,
plications. For example, inhibition of 15-hydroxyprostaglandin including miR302-367 and the RNA-binding protein LIN28A. Post-
dehydrogenase (15-PGDH) by a small-molecule inhibitor of 15- natal reexpression of miR302-367 reactivated the cell cycle in cardio-
myocytes by repressing Hippo signaling, resulting in reduced scar
1
State Key Laboratory of Cellular Stress Biology, Innovation Center for Cell Signaling formation after experimental myocardial infarction (7). Reexpression
Network, School of Life Sciences, Xiamen University, Xiamen, Fujian 361102, China.
2
of LIN28A, which is highly expressed during embryogenesis and in
State-Province Joint Engineering Laboratory of Targeted Drugs from Natural Products, embryonic stem cells, enhanced tissue repair in mice by increasing
Xiamen University, Xiamen, Fujian 361102, China. 3Institute of Systems Biomedicine, De-
partment of Biophysics, and Beijing Key Laboratory of Tumor Systems Biology, School of oxidative metabolism (8). Some of the above techniques need to be
Basic Medical Sciences, Peking University Health Science Center, Beijing 100191, China. used right after birth to achieve a therapeutic effect, where a critical
4
State Key Laboratory of Molecular Vaccinology and Molecular Diagnostics, National In- time point is a prerequisite. It is unknown whether the others will in-
stitute of Diagnostics and Vaccine Development in Infectious Diseases, School of Life
Sciences, and School of Public Health, Xiamen University, Xiamen, Fujian 361102, China.
duce a regenerative healing response to wounds in multiple tissue in-
5
Department of Biological Chemistry and Molecular Pharmacology, and Department of jury sites. Therefore, novel therapeutic strategies and tissue
Cancer Biology, Dana-Farber Cancer Institute, Harvard Medical School, Longwood Center engineering techniques with broader scope are needed to coax organ
2207, 360 Longwood Avenue, Boston, MA 02215, USA. recovery after injury.
*These authors contributed equally to this work.
†Corresponding author. Email: [email protected] (D.Z.); [email protected] (X.D.); The Sterile 20–like kinases MST1 and MST2 (MST1/2), the mam-
[email protected] (C.-H.Y.) malian Hippo orthologs, are key components of the Hippo signaling

www.ScienceTranslationalMedicine.org 17 August 2016 Vol 8 Issue 352 352ra108 1

RESEARCH ARTICLE

cascade that play a pivotal role in stem cell self-renewal, tissue regenera- the melting temperature (Fig. 1F). Furthermore, with increasing
tion, and organ size control (9–17). Central to this pathway is a kinase ATP concentration, XMU-MP-1 exhibited a proportional increase
cascade formed by kinases MST1/2; a scaffolding protein, Salvador/ in IC50 against MST1/2 (Fig. 1G), as well as an attenuated inhibition
WW45 (Sav); nuclear Dbf2–related family kinases LATS1 and of the MST2-mediated phosphorylation of MOB1 (fig. S3E). Together
LATS2 (LATS1/2); and an adaptor protein, MOB1. MST1/2 phos- with substrate kinetic analyses (fig. S3F), these data suggest that XMU-
phorylates and activates LATS1/2–MOB1, which then phosphorylates MP-1 is an ATP-competitive inhibitor for MST1/2.
Yes-associated protein/transcriptional coactivator with PDZ-binding Kinase selectivity of XMU-MP-1 was determined using KINO-
motif (YAP/TAZ). Phospho-YAP/TAZ is either degraded or sequest- MEscan, which profiled the inhibitor at a concentration of 1 mM
ered in the cytoplasm by the 14-3-3 protein. When the Hippo path- against a panel of 468 diverse kinases using an in vitro ATP-site com-
way is off, YAP/TAZ translocates to the nucleus and forms a functional petition binding assay (20). The KINOMEscan results are reported as
hybrid transcriptional factor with TEA domain transcription factor the percentage of the DMSO negative control signal (ctrl%) set at
(TEAD) to turn on pro-proliferative and prosurvival genes, thereby 100%, and a lower number represents higher-affinity binding. There
enabling cell proliferation. Genetic defects in this pathway in mice lead were 23 kinases with ctrl% ≤10% in total containing MST1-4, indicat-
to sustained tissue growth and can lead to cancer (12, 13, 16). Al- ing very strong inhibition. The dissociation constants (Kd) of XMU-
though Hippo kinase MST1/2 structures have been determined, MP-1 against these kinases were determined (Fig. 1H and tables S1
small-molecule MST1/2 inhibitors have yet been disclosed (18, 19). and S2). A selectivity score [S(10)] of 0.05 was determined for XMU-
Therefore, small molecules targeting Hippo signaling in a reversible MP-1, calculated by dividing the number of kinases with ctrl% ≤10%

Downloaded from http://stm.sciencemag.org/ by guest on October 11, 2018

and dose-dependent manner have drawn significant attention as nov- (n = 23) by the total number of kinases tested (n = 468). Cumula-
el, molecularly targeted therapeutics for regenerative medicine. tively, these studies demonstrate that XMU-MP-1 is a potent and
Here, we report the discovery of a potent and selective inhibitor of highly selective inhibitor of MST1/2 kinases.
MST1/2, 4-((5,10-dimethyl-6-oxo-6,10-dihydro-5H-pyrimido[5,4-b]
thieno[3,2-e][1,4]diazepin-2-yl)amino)benzenesulfonamide (XMU- Compound XMU-MP-1 abrogates the MST1/2-mediated
MP-1), using an enzyme-linked immunosorbent assay (ELISA)–based signaling cascade
high-throughput screen of a kinase-directed compound library. A 2.47 Å We investigated whether XMU-MP-1 was capable of inhibiting
cocrystal structure of the MST2 with XMU-MP-1 reveals the mo- MST1/2 kinase activities within cells. At concentrations ranging from
lecular basis for selectivity. The development of inhibitor-resistant al- 0.1 to 10 mM, XMU-MP-1 reduced the phosphorylation of endoge-
leles of MST1/2 in conjunction with structure-activity relationship nous MOB1, LATS1/2, and YAP in human liver carcinoma (HepG2)
(SAR) studies confirmed that the cellular effects of XMU-MP-1 are cells in a dose-dependent manner (fig. S4A). Similarly, XMU-MP-1
“on-target” to MST1/2. The pharmacokinetic properties of XMU- treatment inhibited hydrogen peroxide (H2O2)–stimulated MOB1
MP-1 were sufficient to enable inhibition of MST1/2 activities in mice, phosphorylation and MST1/2 autophosphorylation in a variety of cell
and the compound exhibited in vivo efficacy in multiple models of lines, including mouse macrophage-like cells (RAW264.7), human os-
liver and intestinal repair and regeneration. Thus, the pharmacological teosarcoma (U2OS), human colorectal adenocarcinoma (SW480), im-
manipulation of the Hippo signaling pathway using XMU-MP-1 might mortalized human retinal pigment epithelial cells (RPE1), human
provide new avenues for research in regenerative medicine and novel pleomorphic hepatocellular carcinoma (SNU-423), and HepG2, as
therapeutic options for tissue repair. well as primary mouse hepatocytes, without affecting the phosphoryl-
ation of JNK (c-Jun N-terminal kinase), which is a positive control for
H2O2 stimulation (Fig. 2A and fig. S4B).
RESULTS MST1/2 proteins are proapoptotic kinases. XMU-MP-1 treatment
prevented cell death induced by adenovirus-mediated overexpression
High-throughput screening identifies small molecules that of MST2 (fig. S4C). Another downstream effect of MST1/2 activation
inhibit kinases MST1 and MST2 is to induce phosphorylation and nuclear exit of YAP. XMU-MP-1
To identify the pharmacological modulators of MST1/2 kinase activ- treatment increased YAP nuclear translocation (Fig. 2B), leading to
ities, we designed and optimized an ELISA-based high-throughput a significant up-regulation of the YAP target genes CTGF and
screening assay using their physiological substrate MOB1 protein as CYR61 in HepG2 cells (Fig. 2C). Furthermore, in HepG2 cells over-
the reaction substrate (fig. S1, A to D). After screening an in-house expressing MST2, the enhanced phosphorylation of LATS1/2, MOB1,
compound library (~3000 compounds) designed to target the adeno- and YAP and the increased YAP cytoplasmic localization were pre-
sine triphosphate (ATP)–binding site of the kinases MST1/2, we ob- vented by XMU-MP-1 (Fig. 2D and fig. S4D). Previous studies have
tained preliminary hits that inhibited MST2 kinase activity by more shown that cell-cell contact and high cell density activate Hippo
than 50% at 10 mM. The selected hits (with MOB1 phosphorylation signaling to inhibit YAP activity by promoting cytoplasmic retention
<5%) were derived from four chemical scaffolds shown in Fig. 1A and (21). We found that YAP was exclusively located in the cytosol when
fig. S1E. HepG2 cells were grown at high density and XMU-MP-1 treatment
Through iterative rounds of structure-activity optimization, XMU- induced predominant nuclear localization of YAP, regardless of the
MP-1 (Fig. 1B and fig. S2) was identified as having the best inhibitory cell density (Fig. 2E).
activity, with IC50 values of 71.1 ± 12.9 nM and 38.1 ± 6.9 nM against To investigate whether the inhibition of kinases MST1/2 via XMU-
MST1 and MST2, respectively (Fig. 1C). XMU-MP-1 inhibited phos- MP-1 is reversible, we performed cellular washaway experiments in
phorylation of MOB1 in a dose-dependent manner (Fig. 1, D and E, HepG2 cells. One hour after washing, the cells displayed a greatly re-
and fig. S3, A to D). XMU-MP-1 also demonstrated a high-affinity duced amount of YAP in the nucleus, and the YAP subcellular
interaction with the MST2 protein, as shown by a shift of 8.5°C in distribution was restored to control levels 2 hours after washing

www.ScienceTranslationalMedicine.org 17 August 2016 Vol 8 Issue 352 352ra108 2

RESEARCH ARTICLE

Downloaded from http://stm.sciencemag.org/ by guest on October 11, 2018

Fig. 1. High-throughput screening identifies XMU-MP-1 as a potent, tathione S-transferase. (F) The thermal denaturation curve shift of MST2
ATP-competitive inhibitor of kinases MST1/2. (A) Overview of the primary (10 mM) via XMU-MP-1 (100 mM). Tm, melting transition temperature; RFU,
screening data for an in-house library. Each spot represents a compound at relative fluorescence units. (G) IC50 values of XMU-MP-1 against MST1 or
10 mM. Data are relative to the dimethyl sulfoxide (DMSO)–treated con- MST2 at different ATP concentrations. Data are means ± SD relative to
trol. (B) The chemical structure of XMU-MP-1. Synthetic steps are shown DMSO controls (n = 3). (H) The KINOMEscan result of XMU-MP-1 against
in fig. S2. (C) The inhibitory activities of XMU-MP-1 for MST1 and MST2, a panel of 468 kinases. S(10) and Kd’s of XMU-MP-1 against kinases with
measured as percentage of MOB1a phosphorylation. Data are means ± ctrl% ≤10% containing MST1-4 highlighted as blue (see tables S1 and S2
SD relative to DMSO-treated controls (n = 3). IC50, median inhibitory con- for the full list and see Materials and Methods for the detailed kinase
centration. (D and E) Immunoblot analysis of phosphorylated (p-) MOB1 in group names) were shown. Data are from one experiment representative
a cell-free reaction system at different doses of XMU-MP-1 for 15 min (D) or of three (C to H) or two (A) independent experiments. Full blots are shown
1 mM XMU-MP-1 for different incubation times (E). His, histidine; GST, glu- in fig. S11.

(Fig. 2F). Consistently, Western blotting analysis indicated that the washing (Fig. 2, G and H, and fig. S4, E and F). These data establish
XMU-MP-1 treatment greatly reduced the phosphorylation levels of that XMU-MP-1 can potently and reversibly suppress the activities of
MOB1 and YAP, which were restored to normal levels 2 hours after kinases MST1/2 and enhance their downstream YAP activation in cells.

www.ScienceTranslationalMedicine.org 17 August 2016 Vol 8 Issue 352 352ra108 3

RESEARCH ARTICLE

Fig. 2. Inhibition of
MST1/2 signaling by
XMU-MP-1 in vitro in var-
ious cell types. (A) Relative
levels of p-MOB1 and p-
MST in a variety of cell lines
or primary mouse hepato-
cytes stimulated with or
without H2O2 after treat-
ment with various con-
centrations of XMU-MP-1.
(B) Immunoblot analysis
of YAP, poly(ADP-ribose)
polymerase (PARP), and
a-tubulin, and quantifica-
tion of relative levels of
YAP in cytoplasmic (c) and
nuclear (n) fractions of hu-

Downloaded from http://stm.sciencemag.org/ by guest on October 11, 2018

man HepG2 cells treated
with various concentrations
of XMU-MP-1. (C) Real-time
quantitative polymerase
chain reaction (RT-qPCR)
analysis of the expression
levels of CTGF and CYR61
in HepG2 cells after XMU-
MP-1 treatment for 6 hours.
(D) Confocal microscopy of
YAP distribution in HepG2
cells overexpressing MST2
(or not) at about 30% con-
fluence after DMSO or
XMU-MP-1 treatment for
4 hours.Nucleiwere counter-
stained with 4′,6-diamidino-
2-phenylindole (DAPI). Cells
transfected with or without
MST2 are indicated with a
star or an arrow, respec-
tively. Scale bars, 10 mm.
(E) Confocal microscopy
of YAP distribution in
HepG2 cells cultured at
high density (100% con-
fluent) after DMSO or
XMU-MP-1 treatment for
4 hours. Nuclei were coun-
terstained with DAPI. Scale
bars, 10 mm. (F) Confocal
microscopy of YAP dis-
tribution in HepG2 cells
treated with 3 mM XMU-
MP-1 for 0 to 8 hours (top
two panels) or washed with
fresh medium after 4 hours
of treatment with XMU-
MP-1 (bottom). Cells coun-
terstained with DAPI are
inset. Scale bars, 10 mm.
(G and H) Quantification
of relative levels of p-MOB1 and p-YAP in HepG2 cells treated with various concentrations of XMU-MP-1 for 0 to 8 hours (G) or for 4 hours followed by
depleting XMU-MP-1 (5 to 8 hours) (H). Data are means ± SD relative to DMSO controls (n = 3). In (A) to (C), *P < 0.05, **P < 0.01, ***P < 0.001, Student’s
t test. All data are from one experiment, representing three independent experiments. Full blots are shown in fig. S11.

www.ScienceTranslationalMedicine.org 17 August 2016 Vol 8 Issue 352 352ra108 4

RESEARCH ARTICLE

Cocrystal structure and SAR reveal that XMU-MP-1 is We next explored whether the pharmacological inhibition of MST1/2
on-target to MST2 would similarly potentiate liver repair and regeneration in vivo. The
To study how XMU-MP-1 recognizes the MST1/2 kinases, we pharmacokinetic properties of XMU-MP-1 were first evaluated in
determined the protein/inhibitor complex structure via cocrystallization. Sprague-Dawley rats administered a single intravenous or oral dose.
MST2 in this complex adopted the typical active conformation closely XMU-MP-1 exhibited favorable pharmacokinetics with a half-life of
resembling the wild-type MST1 crystal structure [Protein Data Bank 1.2 hours, an area under the curve of 1035 (h·ng)/ml, and a bio-
(PDB) 3COM; fig. S5A] as evidenced by the extended activation loop availability of 39.5% (table S3). In pharmacodynamic experiments,
(A-loop) and the phosphorylated Thr174 and Thr180 residues (residues the maximal phosphorylation inhibition of MOB1 and YAP was
177 and 183 in MST1) (Fig. 3A). In contrast, the previously reported achieved between 1.5 and 6 hours after intraperitoneal dosing with
apo-bound and AMP-PNP–bound kinase–dead mutant MST2 D146N XMU-MP-1 (1 mg/kg) (Fig. 4A). A dose escalation study of XMU-
structures (PDB 4LG4 and 4LGD) were in the inactive conformation MP-1 revealed that the phosphorylation of MOB1 in liver tissue
(fig. S5B) (18, 19). XMU-MP-1 bound deeply into the ATP-binding was blocked at a minimal dose (1 mg/kg, intraperitoneally) (Fig. 4B).
pocket, sandwiched between the N- and C-lobes, and closely associated Liver-specific deletion of MST1/2 using Alb-Cre (Mst1fl/flMst2fl/fl–
with the hinge of the kinase (Fig. 3B and fig. S5C). The pyrimido[4,5-b] Alb-Cre) resulted in a two- to threefold increase in liver/body weight
thieno[3,2-e][1,4]diazepine scaffold adopted a bent conformation that ratio and overexpansion of bile duct–like liver progenitor cells (oval
optimally fit the hydrophobic Val41 and Leu153 side chains; the phenyl cells) compared to control wild-type mice (Mst1fl/flMst2fl/fl) within
ring of the compound contacted the side chain of Leu33 through a hy- 2 months (Fig. 4C). Wild-type mice treated with XMU-MP-1 daily for

Downloaded from http://stm.sciencemag.org/ by guest on October 11, 2018

drophobic interaction. The compound was well accommodated in the 2 months had moderately increased ratios (around 20 to 30%) of liver/
ATP-binding pocket and interacted with the bulky Met gatekeeper body weight and very low levels of oval cell expansion, as indicated by
(residue 99) through its methyl moiety (Fig. 3B and fig. S5C). staining for the marker cytokeratin 19 (CK19) (Fig. 4, D and E). These
The sulfonamide moiety of XMU-MP-1 interacted with the kinase treated mice did not exhibit any cancer growth and had no detectable
through at least two interactions, including the direct polar interaction levels of a-fetoprotein, a marker for hepatocellular carcinoma. More-
with the Asp109 side chain and the indirect interaction with a Tyr101 over, the liver/body weight ratio and the percentage of CK19-positive
side-chain hydroxyl through a water bridge. Because of these interactions, and bromodeoxyuridine (BrdU)–positive cells were restored to
the conformation of the sulfonamide group was fixed, as shown by the normal levels after 2 weeks of XMU-MP-1 withdrawal (Fig. 4, D
well-defined electron density (fig. S5D). Notably, unlike other kinases, the and E). Furthermore, mice treated with XMU-MP-1 at a threefold
MST1/2 kinases have a C-terminal helix (aJ) located near the sulfonamide higher dose (3 mg/kg) daily for 7 days showed equivalent daily
group, which might provide extra determinants for further optimization weights and physical activity compared with those treated with a ve-
of the compound (for instance, to increase the binding potency or selec- hicle control, showing no significant changes in cell proliferation
tivity through careful tailoring of the tail group of the compound). among various tissues (Fig. 4F) and serum chemistry (table S4). Thus,
To corroborate the functional relevance of these interactions and to XMU-MP-1 was well tolerated, and the mice appeared healthy with
develop inhibitor-resistant alleles of the kinase, we performed muta- no signs of distress in the tolerability experiments.
genesis and biochemical studies. Introduction of single mutations at To investigate the effect of XMU-MP-1 on liver regeneration, we
Y101F or D109A did not compromise kinase activity; however, the treated wild-type mice with XMU-MP-1 once a day before a two-
inhibition potency (IC50) of XMU-MP-1 was remarkably reduced thirds partial hepatectomy followed by daily treatment for 7 days.
by ~10-fold to 361.1 nM for MST2 (Y101F) and by ~27-fold to Similarly, XMU-MP-1 treatment significantly decreased the phos-
1040 nM for MST2 (D109A) relative to wild-type kinase (Fig. 3C). phorylation levels of MOB1 and YAP (fig. S9A). After partial hep-
For the double mutation Y101F/D109A, an even higher IC50 of atectomy, the mice treated with XMU-MP-1 exhibited substantially
1678 nM was observed (Fig. 3C). Consistent with these biochemical enhanced liver regeneration. Between postoperative days 1 and 7,
data, the similar inhibitory effects of XMU-MP-1 in HepG2 cells the number of Ki67-positive cells increased rapidly in the liver and
transfected with various MST1/2 mutations were observed (Fig. 3, was significantly higher in the XMU-MP-1–treated mice than in
D and E, and figs. S6 and S7). These findings confirm the importance controls (Fig. 5A). Moreover, the number of nuclear YAP-positive
of interactions identified by the structural studies and further demon- cells and the expression levels of YAP target gene Ctgf were signifi-
strate that XMU-MP-1 is on-target to MST1/2 kinase. cantly higher in the XMU-MP-1–treated mice than in controls (fig.
Additional SAR studies of XMU-MP-1 (1a) revealed the key S9, B and C). The XMU-MP-1–treated mice also increased their liv-
chemical features for efficacy (fig. S8, A and B). The importance of er/body weight ratio faster than controls, although the two groups
the free hydrogens of -NH2 in sulfonamide (1a) was demonstrated were similar by day 6 (Fig. 5B). XMU-MP-1, therefore, seems to have
by the gradual loss of efficacy when replacing the amide with cyclo- a strong effect in vivo when hepatocytes are actively proliferating.
propanylamine (1b) and 4-methylpiperazine (1c). Without the oxygen Previous studies showed that YAP is involved in proliferation and
substitution of sulfur, the biochemical activity was also greatly reduced differentiation during postnatal liver development (24, 25). To assess
(1e versus 1d). The cellular efficacies of these analogs to inhibit MOB1 the synergistic effect of XMU-MP-1 with proliferative signals in post-
phosphorylation under H2O2 stimulation (fig. S8C) and YAP cyto- natal mice, 10-day-old mice were treated daily with XMU-MP-1 for
plasmic retention (fig. S8D) in HepG2 cells were compromised, 3 weeks. XMU-MP-1 treatment significantly increased the number
consistent with in vitro biochemical activities (Fig. 2, A and B). of CK19- and BrdU-positive cells, as well as the liver/body weight
ratio, by 15 to 20% compared to that of vehicle control (Fig. 5, C and
XMU-MP-1 promotes liver repair and regeneration in mice D). Moreover, the XMU-MP-1–treated MST1/2 double-knockout
Liver-specific deletion of MST1/2 using a Cre-expressing adenovirus (Mst1fl/flMst2fl/fl–Alb-Cre) mice did not increase the number of
leads to immediate and pronounced liver overgrowth in mice (22, 23). Ki67-positive cells 36 hours after partial hepatectomy compared to

www.ScienceTranslationalMedicine.org 17 August 2016 Vol 8 Issue 352 352ra108 5

RESEARCH ARTICLE

Next, we sought to determine the effects of

XMU-MP-1 on facilitating the repopulation of
human hepatocytes using Fah−/−/Rag2−/−/
Il2rg−/− (FRG) mice, which are excellent re-
cipients of human hepatocyte xenografts
(26, 27). These mice, when removed from
NTBC [(2-(2-nitro-4-trifluoromethylbenzoyl)-
1,3-cyclohexanedione)], a drug that inhibits
the formation of hepatotoxic levels of fuma-
ryl acetoacetate, undergo liver failure unless
they receive intrasplenic injection of hepato-
cytes. The FRG mice received human hepa-
tocytes, followed by either vehicle control or
XMU-MP-1, and engraftment was evalu-
ated by histology, immunohistochemistry,
and serum human albumin level (Fig. 5, F
to H). Compared with the vehicle-treated

Downloaded from http://stm.sciencemag.org/ by guest on October 11, 2018

controls, XMU-MP-1 significantly increased
serum human albumin levels within 6 weeks
(Fig. 5F). XMU-MP-1–treated livers were
more greatly repopulated than controls, as
assessed by H&E staining and the liver en-
zyme marker fumaryl acetoacetate hydrolase
(FAH) (Fig. 5G). XMU-MP-1 treatment re-
sulted in a significant increase in the ratio of
BrdU+/FAH+ cells, indicating enhanced pro-
liferation rate of human hepatocytes (Fig.
5G). Similar numbers of terminal deoxynu-
cleotidyl transferase–mediated deoxyuridine
triphosphate nick end labeling (TUNEL)–
positive cells in nonhuman (murine) hepato-
cyte areas (FAH−) suggested that XMU-MP-1
did not promote Fah-deficient hepatocyte
survival (Fig. 5G). However, there were more
nuclear YAP in Fah-deficient liver cells trea-
ted with XMU-MP-1, indicating the inhibition
of Hippo signaling (Fig. 5H). In addition to
FAH, the indicators of functional hepato-
cytes, AAT, CYP3A4, CYP2D6, and CK18,
Fig. 3. XMU-MP-1 is on-target to MST2. (A) Overall structure of the MST2/XMU-MP-1 complex (PDB were equivalent in both controls and XMU-
5DH3) (stereo view). The MST2 protein and the compound are shown as slate cartoons and sticks, re- MP-1–treated animals (Fig. 5H).
spectively. (B) XMU-MP-1 interacts with MST2 at the ATP-binding site. The MST2 kinase is shown as a slate
cartoon except for the P-loop, which is a Ca trace (thin line), to enable a better view of the compound. TheXMU-MP-1 attenuates
aC-helix, A-loop, and P-loop are red, orange, and deep blue (thin line), respectively. The side chains of the
acetaminophen-induced liver injury
residues of Met99, Tyr101, Leu33, Val41, Leu153, Asp109, Lys56, and Glu70 are shown as sticks. The compound Acetaminophen (APAP) overdose is the lead-
is shown as sticks, with C (yellow), O (red), S (orange), and N (blue) atoms. (C) IC50 values for the inhibition
ing cause of acute liver failure worldwide
of wild-type MST2 (WT) or mutant MST2 (Y101F, D109A, or both Y101F and D109A) by XMU-MP-1 for (28). Currently, N-acetylcysteine (NAC), an
15 min. Data are means ± SD relative to DMSO controls (n = 3). (D) p-MOB1 in HepG2 cells expressing
antioxidant, is the antidote for APAP toxicity.
Flag-tagged various forms of MST2 as indicated or control vectors after DMSO or XMU-MP-1 treatment
However, NAC is effective only for patients
for 3 hours. Data are means ± SD relative to DMSO controls (n = 3). ns, not significant; *P < 0.05, **P < 0.01,
Student’s t test. (E) Confocal microscopy of the distribution of YAP in HepG2 cells overexpressing Flag- who present within hours of an acute over-
tagged various forms of MST2, as indicated, after DMSO or XMU-MP-1 treatment for 4 hours. Nuclei were dose (29). In addition, prolonged (>24 hours)
counterstained with DAPI. Cells transfected with or without MST2 (WT or mutant, as indicated) are treatment with NAC can be toxic and delay
labeled with a star or an arrow, respectively. Scale bars, 10 mm. Data in (C) to (E) are from one experimentliver regeneration in patients with APAP
representative of three independent experiments. hepatotoxicity (30). NAC efficiently blocks
APAP-induced injury in mice when NAC
those of vehicle-treated MST1/2-null mice (Fig. 5E and fig. S9D), in- treatment occurs within 1.5 hours of APAP administration (31).
dicating that liver overgrowth was due to the inhibition of MST1/2 by MST1/2 proteins are proapoptotic kinases. Thus, blocking MST1/2
XMU-MP-1 in vivo. might ameliorate APAP-induced liver cell death.

www.ScienceTranslationalMedicine.org 17 August 2016 Vol 8 Issue 352 352ra108 6

RESEARCH ARTICLE

Downloaded from http://stm.sciencemag.org/ by guest on October 11, 2018

Fig. 4. Inhibition of MST1/2 signaling by XMU-MP-1 in mouse liver. sections (n = 3) (D) or liver/body weight ratio (n = 6) (E) of wild-type mice
(A and B) Immunoblots and quantification of p-MOB1 and p-YAP in lysates treated with a vehicle or XMU-MP-1 (1 mg/kg) daily for 2 months, or after
of livers from wild-type mice treated with XMU-MP-1 (1 mg/kg) over time 2 weeks of XMU-MP-1 withdrawal. ND, not determined. Data are means ±
(0 to 12 hours) (A) or different doses of XMU-MP-1 for 3 hours (B). Data are SD. P values were determined by Student’s t test. Scale bars, 100 mm.
means ± SD relative to DMSO controls (n = 3). *P < 0.05, **P < 0.01, ***P < (F) Ki67 staining in tissue sections from wild-type mice treated with a
0.001 versus control, Student’s t test. GAPDH, glyceraldehyde-3-phosphate vehicle or XMU-MP-1 (3 mg/kg) daily for 7 days. Quantification of the Ki67-
dehydrogenase. (C) Hematoxylin and eosin (H&E) or immunohistochemical positive cells in various tissues. Data are means ± SD (n = 6). P values were
(IHC) staining of the oval cell marker CK19 in liver sections of Mst1fl/flMst2fl/fl determined by Student’s t test. Scale bars, 100 mm. Data are from one ex-
and Mst1fl/flMst2fl/fl–Alb-Cre mice. Scale bars, 100 mm. (D and E) H&E and periment representative of three (A to C) or two (D to F) independent
IHC staining and quantification of CK19- and BrdU-positive cells in liver experiments. Full blots are shown in fig. S11.

www.ScienceTranslationalMedicine.org 17 August 2016 Vol 8 Issue 352 352ra108 7

RESEARCH ARTICLE

Downloaded from http://stm.sciencemag.org/ by guest on October 11, 2018

Fig. 5. XMU-MP-1 promotes human and mouse liver cell proliferation. serum human albumin (hALB) levels. Two weeks after transplantation, animals
(A and B) Ki67-positive cells in liver sections (A) or the liver/body weight ratio were treated with a vehicle or XMU-MP-1 (1 mg/kg) once per day. Data are
(B) of partially hepatectomized mice treated with a vehicle or XMU-MP-1 means ± SD (n = 9). (G) H&E staining and IHC staining of FAH, BrdU, and TUNEL
(1 mg/kg) twice per day. Data are means ± SD (n = 5). (C and D) H&E stain- in FRG livers. H, human hepatocytes. Quantifications of the ratio of BrdU+/FAH+
ing and IHC staining of CK19 and BrdU (C), as well as quantifications of cells and the number of TUNEL-positive cells in nonhuman hepatocyte region
CK19- and BrdU-positive cells in liver sections (n = 3) or the liver/body weight (FAH− cells) in liver sections of mice treated with a vehicle or XMU-MP-1
ratio of postnatal mice (n = 4) (D) treated with a vehicle or XMU-MP-1 (1 mg/kg) (1 mg/kg) once per day for 28 days. Data are means ± SD (n = 3). Scale bars,
once per day for 21 days. Data are means ± SD. Scale bars, 100 mm. (E) Ki67- 200 mm. (H) IHC staining of YAP and human hepatic functional proteins cy-
positive cells in liver sections of partially hepatectomized Mst1fl/flMst2fl/fl or tochrome P-450 (CYP) 3A4, CYP2D6, a-1 antitrypsin (AAT), and CK18 in FRG
Mst1fl/flMst2fl/fl–Alb-Cre mice treated with a vehicle or XMU-MP-1 (1 mg/kg) livers. Scale bars, 50 mm. *P < 0.05, **P < 0.01, ***P < 0.001 versus control,
twice per day for 2 days. Data are means ± SD (n = 3). (F) The repopulation of unless otherwise indicated, Student’s t test. Data are from one experiment
transplanted human hepatocytes in FRG livers was determined by measuring representative of three (A to E) or two (F to H) independent experiments.

www.ScienceTranslationalMedicine.org 17 August 2016 Vol 8 Issue 352 352ra108 8

RESEARCH ARTICLE

NAC (250 mg/kg) administered at 1.5 hours, but not at 2.5 or marily by helping to maintain colonocyte proliferation in damaged
3.5 hours after APAP administration, protected mice from liver cen- mucosa.
trilobular necrosis (Fig. 6A and fig. S10A). Extending this therapeutic
window, XMU-MP-1 administered within 2.5 hours after APAP ad- XMU-MP-1 ameliorates chronic liver injury
ministration effectively prevented the continual expansion of centri- Next, we evaluated the regenerative effect of XMU-MP-1 in chronic
lobular necrotic lesions in APAP-treated mouse livers (Fig. 6B and liver injury, which is of great clinical interest. The two most commonly
fig. S10, B to D). NAC treatment alone at 1.5 hours after APAP ad- used models of experimental chronic liver injury are iterative toxic
ministration treated about 80% of mice, whereas the combination of damage (for example, elicited by CCl4 intoxication) and bile duct
NAC and XMU-MP-1 (1 mg/kg) at 1.5 hours after APAP administra- ligation (35). Chronic liver injury was induced by the administration
tion protected 100% of mice from APAP-induced death (Fig. 6A). The of CCl4 twice per week for 4 weeks (Fig. 7A) or by bile duct ligation
combination of NAC and XMU-MP-1, or XMU-MP-1 treatment (Fig. 7B). Liver fibrosis is the common result of chronic hepatic injury.
alone, effectively extended the protective effects to 2.5 hours after Liver fibrosis and cell death were assessed by Sirius Red and TUNEL
APAP administration, as shown by the increased survival rate from staining, respectively. In the model of CCl4-induced liver chronic in-
~40% (NAC alone or vehicle control) to ~75% (Fig. 6B), as well as jury, mice treated with XMU-MP-1 (1 mg/kg) daily for 10 days
the smaller liver centrilobular necrotic lesions compared with the con- (starting at day 30 after toxic insult) had a small amount of collagen
trol groups (Fig. 6C and fig. S10D). Furthermore, compared with the deposition as well as few dead cells, whereas mice treated with vehicle
vehicle controls, all treatments given at either 1.5 or 2.5 hours after control demonstrated bridging fibrosis and significantly more TUNEL-

Downloaded from http://stm.sciencemag.org/ by guest on October 11, 2018

APAP treatment significantly decreased serum alanine transaminase positive cells, suggesting that inhibition of MST1/2 by XMU-MP-1 is
(ALT) levels at 24 hours after APAP treatment (Fig. 6C). Of the treat- capable of partially rescuing fibrosis (Fig. 7C). XMU-MP-1 treatment
ments provided at 2.5 hours after APAP administration, however, only exhibited a similar effect in the model of duct ligation–induced chronic
XMU-MP-1 alone or the combination of NAC and XMU-MP-1 signif- liver injury, with MST1/2 inhibition reducing fibrosis and cell prolif-
icantly decreased the mortality rate (Fig. 6A). eration (Fig. 7D).
Genetic disruption of the components of the Hippo pathway has
been shown to result in eventual tumor formation (12, 13). Never-
theless, mice with APAP-induced injury that were treated daily with DISCUSSION
XMU-MP-1 (1 mg/kg) for 1 week did not exhibit cancerous pheno-
types for more than 10 months when examined by gross necropsy. Pharmacological manipulation of the Hippo pathway could open
They, like control mice, which were subjected to a sublethal dose doors to facile tissue regeneration with drugs, rather than complex
of APAP followed by the treatment of PBS or NAC, remained biomaterial and cell therapies. MST1/2 kinases, for instance, play ma-
healthy (n = 10 per group). In addition, wild-type mice treated dai- jor roles in organ growth control. Our present study demonstrates that
ly with XMU-MP-1 at the same dose (1 mg/kg) but for a longer XMU-MP-1 is a selective and reversible small-molecule MST1/2 in-
period of time (2 months) did not exhibit any cancer growth. These hibitor with in vivo efficacy in four different rodent models of liver
results indicate that the pharmacological blockage of MST1/2 activ- injury, attenuating Hippo signaling and augmenting tissue repair. Ad-
ities can halt the negative sequelae of APAP-induced liver injury ditionally, this new agent could be used to study the mechanisms of
and increase the animal survival rate beyond that obtained with Hippo signaling–mediated pathobiology in other experimental models
NAC alone—the current clinical standard of care. of organ disease and damage.
Genetic disruption of the components of the Hippo pathway
XMU-MP-1 protects mice from DSS-induced colitis results in YAP activation and, eventually, tumor formation; thus, de-
To determine whether the MST1/2 inhibitor XMU-MP-1 promotes veloping tissue-regenerative therapeutics by targeting Hippo signaling
regeneration in intestinal tissues where Hippo signaling plays an is a daunting challenge (12, 13). Recent studies demonstrated that the
important role (32, 33), we examined the effects of XMU-MP-1 downstream effector protein YAP of kinases MST1/2 could be a
in the colons of mice treated for 7 days with oral dextran sodium promising target for cancer prevention (36, 37). One study showed
sulfate (DSS), an agent that induces the colon tissue ulcerations similar that the enantiopure organoruthenium might inhibit MST1 kinase
to those found in humans with ulcerative colitis. Wild-type mice treated in vitro using a nonphysiological substrate of MST1/2 kinase, but
daily with XMU-MP-1 showed a marked resistance to DSS-induced the authors did not provide further evidence about whether it could
colitis compared with controls. XMU-MP-1–treated mice showed sig- target MST2 kinase or about inhibitor selectivity and in vivo efficacy
nificantly less weight loss (Fig. 6D) and suppression of colitis symp- (38). With XMU-MP-1 in hand, we demonstrated that pharmaco-
toms, with substantially less diarrhea and rectal bleeding, as assessed logical inhibition of MST1/2 can augment tissue repair and regenera-
by the disease activity index (DAI) (Fig. 6E) (34). tion after injuries such as partial hepatectomy, APAP-induced liver
To further characterize the mechanism through which MST1/2 hepatotoxicity, and chemically induced liver chronic injury and colitis.
inhibition protects the colon epithelium, we assessed the effects of Healthy or injured mice treated with XMU-MP-1 (1 mg/kg) daily for
XMU-MP-1 on the cell proliferation in the colon crypts. The 2 months or 1 week, respectively, did not exhibit cancerous growth.
numbers of nuclear YAP-positive cells and BrdU incorporation To some extent, these results differ from the MST1/2 double-knockout
or Ki67-positive proliferating cells were significantly higher in the phenotype, suggesting that remaining MST1/2 activities in wild-type
XMU-MP-1 treatment group than in controls (Fig. 6F). Mice trea- mice treated with XMU-MP-1 are sufficient to control the cell over-
ted daily with XMU-MP-1 (1 mg/kg) for 1 week in the model of proliferation that persists in intact tissues. Previous studies showed
DSS-induced injury did not exhibit cancerous phenotypes for more that Hippo signaling is suppressed to some extent during tissue re-
than 10 months. Thus, MST1/2 inhibition confers protection pri- generation (25, 39). Thus, XMU-MP-1 treatment could facilitate

www.ScienceTranslationalMedicine.org 17 August 2016 Vol 8 Issue 352 352ra108 9

RESEARCH ARTICLE

Downloaded from http://stm.sciencemag.org/ by guest on October 11, 2018

Fig. 6. XMU-MP-1 promotes liver and intestinal tissue repair and re- (D and E) Body weight (D) and DAI (E) over days 1 to 21 of mice with colitis
generation after acute injury. (A) The mortality of wild-type mice injected induced by DSS (days 1 to 7) that were also treated with a vehicle (n = 12)
intraperitoneally with vehicle (n = 11), NAC (n = 10), XMU-MP-1 (n = 11), or or XMU-MP-1 (1 mg/kg) (n = 12) daily. Data are means ± SD. Plots of daily
NAC + XMU-MP-1 (n = 11) at 1.5 or 2.5 hours after the oral administration of weights were graphed as the percentage of the day 1 value (D). Plots of daily
APAP (400 mg/kg). P values versus control were determined by Mantel-Cox DAI are shown in (E). *P < 0.05 compared between the vehicle-treated and
test. (B) TUNEL staining of liver sections at indicated post-APAP time points XMU-MP-1–treated groups at the same time point via Student’s t test. (F) IHC
from wild-type mice treated 2.5 hours after a sublethal dose of APAP (200 mg/kg) staining of YAP, BrdU, and Ki67 in the colon sections of wild-type mice after
administration. Scale bars, 200 mm. (C) Serum ALT levels at 24 hours after APAP the oral administration of DSS for 7 days after vehicle or XMU-MP-1 (1 mg/kg)
administration among wild-type mice injected intraperitoneally with vehicle each day. BrdU- and Ki67-positive cells in the colon sections were quantified
(n = 7), NAC (n = 5 at 1.5 hours; n = 4 at 2.5 hours), XMU-MP-1 (n = 7 at 1.5 hours; in tissue sections from individual animals. Data are means ± SD (n = 10).
n = 4 at 2.5 hours), or NAC + XMU-MP-1 (n = 5 at 1.5 hours; n = 4 at 2.5 hours) P values were determined by Student’s t test. Scale bars, 100 mm. All data are
after APAP administration. P values were determined by Mantel-Cox test. from one experiment representative of three independent experiments.

the damaged tissue repair and regeneration but not affect intact, Liver fibrosis occurs as an attempt to limit tissue damage in re-
healthy organs. Moreover, we demonstrated that the inhibitory ef- sponse to chronic liver injury. However, progressive fibrosis eventu-
fect of XMU-MP-1 on MST1/2 is reversible. Thus, XMU-MP-1 appears ally results in cirrhosis. Antifibrotic therapies should inhibit matrix
to be safe as a targeted therapeutic to treat tissue injury. deposition, collagen synthesis, and hepatic stellate cell activation. A

www.ScienceTranslationalMedicine.org 17 August 2016 Vol 8 Issue 352 352ra108 10

RESEARCH ARTICLE

Fig. 7. XMU-MP-1
potentiates tissue re-
pair and regeneration
after chronic liver inju-
ry. (A and B) Schematics
for CCl4-induced (A) and
bile duct ligation (BDL)–
induced (B) chronic liver
injury and fibrosis evalu-
ation in mice. (C) Sirius
Red and TUNEL staining
andcorrespondingquan-
tifications in liver sec-
tions 10 days after CCl4
administration. Injured
mice were treated with
a vehicle or XMU-MP-1
(1 mg/kg) for 10 days.

Downloaded from http://stm.sciencemag.org/ by guest on October 11, 2018

Scale bars, 100 mm.
(D) H&E, Sirius Red,
and TUNEL staining
and corresponding
quantifications in liver
sections after 30 days
of BDL. Injured mice
were treated with a vehi-
cle or XMU-MP-1 (1 mg/
kg) for 10 days. L, lesions.
Scale bars, 200 mm. P
values were determined
by Student’s t test. Data
in (C)and(D)are means ±
SD (n = 3) from one ex-
periment representa-
tive of two independent
experiments.

www.ScienceTranslationalMedicine.org 17 August 2016 Vol 8 Issue 352 352ra108 11

RESEARCH ARTICLE

previous study showed that macrophage depletion in mice during in- blinded to genotype and/or treatment arm. For drug treatments, age-
jury resulted in a marked loss of stellate cells and matrix deposition and gender-matched animals were randomly assigned to treatment
(40); thus, macrophages are likely essential for regeneration. Addition- arms with about equivalent numbers in each group. Box-and-whisker
ally, reactive oxygen species could augment fibrogenesis (41). We pre- plots were identified using RStudio-defined outliers (shown as circles),
viously found that MST1/2 kinases are essential for macrophage but all data points were used in statistical analyses.
activation and that the elimination of MST1/2 from the liver resulted
in increased expression of enzymes that scavenge oxidants (25, 42). General synthesis of XMU-MP-1 and its analogs
Another study showed that the Hippo downstream effector protein MST1/2 kinase inhibitor XMU-MP-1 and its analogs were synthesized
YAP promoted stellate cell activation (43). The complex interplay be- from a starting mixture of methyl 3-aminothiophene-2-carboxylate,
tween MST1/2 kinase activity and cellular activity during regeneration, diisopropylethylamine, and 2,4-dichloro-5-nitropyrimidine in isopro-
evidenced by these contradictory findings, might explain why pharma- panol, as described in Supplementary Methods (figs. S2 and S8A). For
cological inhibition of MST1/2 results in only a partial rescue of liver biochemical assays and cellular studies, XMU-MP-1 was dissolved in
fibrosis in mice. Thus, more work is needed to determine the antifi- DMSO (stock concentration, 10 mM). For animal treatments, XMU-
brotic mechanisms of XMU-MP-1 and to further optimize the dosage MP-1 was dissolved in 0.1% citric acid aqueous solution containing
to achieve better antifibrotic efficacy. 20% Kolliphor HS 15.
Together, our results show that XMU-MP-1 is a valuable chemical
probe to study the Hippo signaling pathway, and it provides a starting In vitro and in vivo kinase inhibition assays

Downloaded from http://stm.sciencemag.org/ by guest on October 11, 2018

point for medicinal chemistry efforts aimed at developing therapeutics For the in vivo inhibition assays, human embryonic kidney (HEK)
that target the Hippo-signaling cascade. The major challenges in hu- 293T cells were transfected with 0.5 mg of empty plasmid or pCMV
man hepatocyte transplantation are the limited supply of donor plasmids expressing various forms of Flag-tagged full-length MST1
organs and the lack of effective methods to improve engraftment or MST2 kinase in 12-well plates each. Twenty-four hours after
and proliferation of donor hepatocytes after transplantation. Targeting transfection, cells were treated with the indicated doses of XMU-
Hippo signaling by XMU-MP-1 or other therapeutics (7) might MP-1 for 3 hours. Cell lysates were analyzed via immunoblotting
support hepatocyte expansion after transplant in patients. The future with the indicated antibodies. For the in vitro kinase inhibition assays,
success of such trial would lead to autologous live cell therapy. Al- recombinant GST-tagged MOB1a and various forms of recombinant
though our studies involve several mouse models, in addition to liver His-tagged full-length MST1 or MST2 kinase were expressed and pu-
and intestinal tissues, the loss of MST1/2 in other tissues, including the rified from Escherichia coli. The enzyme, ATP, and GST-MOB1 con-
pancreas, skin, and heart, also promotes tissue growth (11). It will be sumption were kept consistent with the previously optimized
of interest to determine the protective effect of XMU-MP-1 on injuries conditions. The assays were performed with the indicated doses of
to those tissues. In summary, pharmacological manipulation of Hippo XMU-MP-1 in the kinase assay buffer for 30 min at 30°C followed
signaling unlocks new opportunities for regeneration of multiple tis- by SDS–polyacrylamide gel electrophoresis and immunoblot analyses.
sues that, to date, have seen minimal success in clinical approaches
centered on cell- and biomaterial-based therapies. KINOMEscan profiling of XMU-MP-1
XMU-MP-1 was profiled against a panel of 468 kinases using KINO-
MEscan technology, an active-site–dependent competition-binding as-
MATERIALS AND METHODS say at 1 mM (DiscoverX Corp.). The KINOMEscan selectivity score is
a quantitative measure of a compound’s selectivity (20). It is calculated
Study design by dividing the number of kinases that bind to the compound by the
The purpose of this study was to explore a therapeutic regenerative total number of kinases tested. The results are reported as “control%”
medicine path that pharmacologically targets the Hippo pathway ki- (ctrl%) in which lower numbers represent higher-affinity binding;
nases MST1/2. We developed an ELISA-based screen approach to ctrl% = (test compound signal − positive control signal)/(negative control
identify potential inhibitors and designed in vitro studies in murine signal − positive control signal) × 100, where the negative control =
and human cell lines, as well as primary murine hepatocytes, and in DMSO (ctrl% = 100%) and the positive control = control compound
vivo efficacy studies in four mouse models of acute and chronic tissue (ctrl% = 0%); S(10) = (number of kinases with ctrl% ≤10%)/(number
injuries: partial hepatectomy, hepatocyte repopulation in FRG mice, of kinases tested). ctrl% <10% means very strong inhibition, and ctrl%
APAP-induced hepatotoxicity, and DSS-induced colitis (all described >70% means very weak inhibition. Kinome illustration reproduced
in Supplementary Methods). At least 3 to 4 biological replicates were courtesy of Cell Signaling Technology (www.cellsignal.com). The ki-
used for each biochemical analysis, whereas a sample size of at least 6 nase group or individual kinase names in the current study includes
to 12 biological replicates per group was used for animal testing to TK (tyrosine kinase), TKL (TK-like), STE (homologs of yeast Sterile 7,
achieve a 90% power. The detailed sample size for each animal study Sterile 11, and Sterile 20 kinases), AGC [protein kinase A, G, and C
determined the mean value and SD (s) and defined the difference in families], CAMK (calcium/calmodulin-dependent protein kinase),
means (d) of the measurement marker, using an online calculator CK1 (casein kinase 1), and CMGC [cyclin-dependent kinase (CDK),
(http://hedwig.mgh.harvard.edu/sample_size/js/js_parallel_quant. mitogen-activated protein kinase (MAPK), glycogen synthase kinase 3
html), assuming 90% power (1 − b), a 5% significance level (a), and a (GSK3), and CDC2-like kinase (CLK) families].
two-sided test.
Data collection occurred for a predetermined period of time, as Real-time quantitative polymerase chain reaction
dictated by literature- or core facility–based standard, and no exclu- One microgram of total RNA from the liver tissue or cells was reverse-
sion criteria were applied. All analyses were performed by examiners transcribed with oligo(dT) and SuperScript III Reverse Transcriptase

www.ScienceTranslationalMedicine.org 17 August 2016 Vol 8 Issue 352 352ra108 12

RESEARCH ARTICLE

(Invitrogen). RT-qPCR was performed using a Bio-Rad iQ SYBR D. Mikkola, J. Naidoo, S. Wei, H.-H. Tai, S. L. Gerson, J. M. Ready, B. Posner, J. K. V. Willson,
S. D. Markowitz, Inhibition of the prostaglandin-degrading enzyme 15-PGDH potentiates
Green Supermix Kit and the Bio-Rad iCycler iQ system (Bio-Rad).
tissue regeneration. Science 348, aaa2340 (2015).
All runs were accompanied by the internal control Gapdh gene. The 5. Y. Zhang, I. Strehin, K. Bedelbaeva, D. Gourevitch, L. Clark, J. Leferovich, P. B. Messersmith,
samples were run in triplicate and normalized to GAPDH using a DD E. Heber-Katz, Drug-induced regeneration in adult mice. Sci. Transl. Med. 7, 290ra92 (2015).
cycle threshold–based algorithm to provide arbitrary units represent- 6. B. D. Polizzotti, B. Ganapathy, S. Walsh, S. Choudhury, N. Ammanamanchi, D. G. Bennett,
ing relative expression levels. The primer sequences for specific genes C. G. dos Remedios, B. J. Haubner, J. M. Penninger, B. Kühn, Neuregulin stimulation of
cardiomyocyte regeneration in mice and human myocardium reveals a therapeutic
are shown in table S6. window. Sci. Transl. Med. 7, 281ra45 (2015).
7. Y. Tian, Y. Liu, T. Wang, N. Zhou, J. Kong, L. Chen, M. Snitow, M. Morley, D. Li, N. Petrenko,
Statistical analysis S. Zhou, M. Lu, E. Gao, W. J. Koch, K. M. Stewart, E. E. Morrisey, A microRNA-Hippo
All statistical analyses were performed using Prism 5 (GraphPad pathway that promotes cardiomyocyte proliferation and cardiac regeneration in mice.
Sci. Transl. Med. 7, 279ra38 (2015).
Software). The liver weights and liver/body weight ratios, BrdU or
8. N. Shyh-Chang, H. Zhu, T. Yvanka de Soysa, G. Shinoda, M. T. Seligson, K. M. Tsanov, L. Nguyen,
TUNEL labeling count, relative mRNA levels, and quantification of J. M. Asara, L. C. Cantley, G. Q. Daley, Lin28 enhances tissue repair by reprogramming cellular
immune blots were tabulated graphically with error bars metabolism. Cell 155, 778–792 (2013).
corresponding to means ± SD and compared using two-tailed Stu- 9. S. J. Rawat, J. Chernoff, Regulation of mammalian Ste20 (Mst) kinases. Trends Biochem. Sci.
dent’s t test. Body weight changes and DAI scores in the DSS model 40, 149–156 (2015).
10. J. Avruch, D. Zhou, J. Fitamant, N. Bardeesy, F. Mou, L. R. Barrufet, Protein kinases of the
with the different treatment arms were compared across the time Hippo pathway: Regulation and substrates. Semin. Cell Dev. Biol. 23, 770–784 (2012).
course of study using two-tailed paired-samples Student’s t test. Sur-

Downloaded from http://stm.sciencemag.org/ by guest on October 11, 2018

11. F.-X. Yu, B. Zhao, K.-L. Guan, Hippo pathway in organ size control, tissue homeostasis, and
vival data were analyzed by the Kaplan-Meier statistical method. P < cancer. Cell 163, 811–828 (2015).
0.05 was considered statistically significant. 12. K. F. Harvey, X. Zhang, D. M. Thomas, The Hippo pathway and human cancer. Nat. Rev. Cancer
13, 246–257 (2013).
13. R. Johnson, G. Halder, The two faces of Hippo: Targeting the Hippo pathway for regen-
erative medicine and cancer treatment. Nat. Rev. Drug Discov. 13, 63–79 (2014).
SUPPLEMENTARY MATERIALS 14. D. Pan, The Hippo signaling pathway in development and cancer. Dev. Cell 19, 491–505 (2010).
15. S. Piccolo, S. Dupont, M. Cordenonsi, The biology of YAP/TAZ: Hippo signaling and beyond.
www.sciencetranslationalmedicine.org/cgi/content/full/8/352/352ra108/DC1 Physiol. Rev. 94, 1287–1312 (2014).
Materials and Methods 16. F. Zanconato, M. Cordenonsi, S. Piccolo, YAP/TAZ at the roots of cancer. Cancer Cell 29,
Fig. S1. Identification, characterization, and optimization of a potent and selective inhibitor of MST1/2. 783–803 (2016).
Fig. S2. General synthetic scheme of XMU-MP-1 and its analogs. 17. Q. Zeng, W. Hong, The emerging role of the Hippo pathway in cell contact inhibition,
Fig. S3. Identification, characterization, and optimization of a potent and selective inhibitor of MST1/2. organ size control, and cancer development in mammals. Cancer Cell 13, 188–192 (2008).
Fig. S4. Effects of XMU-MP-1 on the MST1/2-mediated signaling cascade in HepG2 liver cells. 18. L. Ni, S. Li, J. Yu, J. Min, C. A. Brautigam, D. R. Tomchick, D. Pan, X. Luo, Structural basis for
Fig. S5. The complex structure of MST2/XMU-MP-1. autoactivation of human Mst2 kinase and its regulation by RASSF5. Structure 21, 1757–1768
Fig. S6. Effects of XMU-MP-1 on wild-type or mutated MST2-mediated signaling cascade in (2013).
human HepG2 cells. 19. L. Ni, Y. Zheng, M. Hara, D. Pan, X. Luo, Structural basis for Mob1-dependent activation of
Fig. S7. Effects of XMU-MP-1 on wild-type or mutated MST1-mediated signaling cascade in the core Mst-Lats kinase cascade in Hippo signaling. Genes Dev. 29, 1416–1431 (2015).
human HepG2 cells. 20. M. W. Karaman, S. Herrgard, D. K. Treiber, P. Gallant, C. E. Atteridge, B. T. Campbell, K. W. Chan,
Fig. S8. Inhibition of MST2 by various XMU-MP-1 analogs. P. Ciceri, M. I. Davis, P. T. Edeen, R. Faraoni, M. Floyd, J. P. Hunt, D. J. Lockhart, Z. V. Milanov,
Fig. S9. Effects of XMU-MP-1 on MST1/2 signaling-mediated liver regeneration in mice. M. J. Morrison, G. Pallares, H. K. Patel, S. Pritchard, L. M. Wodicka, P. P. Zarrinkar, A quan-
Fig. S10. XMU-MP-1 stimulates MST1/2 signaling-mediated liver tissue repair and regeneration titative analysis of kinase inhibitor selectivity. Nat. Biotechnol. 26, 127–132 (2008).
in mice. 21. B. Zhao, X. Wei, W. Li, R. S. Udan, Q. Yang, J. Kim, J. Xie, T. Ikenoue, J. Yu, L. Li, P. Zheng, K. Ye,
Fig. S11. Full scans of Western blots for Figs. 1 (D and E), 2B, and 4 (A and B), and figs. S2A and A. Chinnaiyan, G. Halder, Z.-C. Lai, K.-L. Guan, Inactivation of YAP oncoprotein by the Hippo
S3 (C and D). pathway is involved in cell contact inhibition and tissue growth control. Genes Dev. 21,
Fig. S12. Full scans of Western blots for figs. S3E and S4. 2747–2761 (2007).
Fig. S13. Full scans of Western blots for fig. S6. 22. D. Zhou, C. Conrad, F. Xia, J.-S. Park, B. Payer, Y. Yin, G. Y. Lauwers, W. Thasler, J. T. Lee,
Fig. S14. Full scans of Western blots for fig. S7. J. Avruch, N. Bardeesy, Mst1 and Mst2 maintain hepatocyte quiescence and suppress hepa-
Fig. S15. Full scans of Western blots for figs. S8C and S9A. tocellular carcinoma development through inactivation of the Yap1 oncogene. Cancer Cell 16,
Table S1. Full list of KINOMEscan profiling data of XMU-MP-1. 425–438 (2009).
Table S2. Kinase profiling data of XMU-MP-1. 23. H. Wu, L. Wei, F. Fan, S. Ji, S. Zhang, J. Geng, L. Hong, X. Fan, Q. Chen, J. Tian, M. Jiang, X. Sun,
Table S3. Pharmacokinetics of XMU-MP-1 in rats. C. Jin, Z.-Y. Yin, Q. Liu, J. Zhang, F. Qin, K.-H. Lin, J.-S. Yu, X. Deng, H.-R. Wang, B. Zhao,
Table S4: Blood chemistry in tolerability studies. R. L. Johnson, L. Chen, D. Zhou, Integration of Hippo signalling and the unfolded
Table S5. Crystallographic data collection and refinement statistics for MST2 (residues 16–313) protein response to restrain liver overgrowth and tumorigenesis. Nat. Commun. 6, 6239 (2015).
bound to XMU-MP-1. 24. S. Septer, G. Edwards, S. Gunewardena, A. Wolfe, H. Li, J. Daniel, U. Apte, Yes-associated
Table S6. The sequences of primers used in this study. protein is involved in proliferation and differentiation during postnatal liver development.
Data S1. Tabulated raw data. (Microsoft Excel format) Am. J. Physiol. Gastrointest. Liver Physiol. 302, G493–G503 (2012).
25. H. Wu, Y. Xiao, S. Zhang, S. Ji, L. Wei, F. Fan, J. Geng, J. Tian, X. Sun, F. Qin, C. Jin, J. Lin, Z.-Y. Yin,
T. Zhang, L. Luo, Y. Li, S. Song, S.-C. Lin, X. Deng, F. Camargo, J. Avruch, L. Chen, D. Zhou, The
Ets transcription factor GABP is a component of the Hippo pathway essential for growth and
REFERENCES AND NOTES antioxidant defense. Cell Rep. 3, 1663–1677 (2013).
26. H. Azuma, N. Paulk, A. Ranade, C. Dorrell, M. Al-Dhalimy, E. Ellis, S. Strom, M. A. Kay, M. Finegold,
1. G. C. Gurtner, S. Werner, Y. Barrandon, M. T. Longaker, Wound repair and regeneration. M. Grompe, Robust expansion of human hepatocytes in Fah−/−/Rag2−/−/Il2rg−/− mice. Nat.
Nature 453, 314–321 (2008). Biotechnol. 25, 903–910 (2007).
2. S. A. Eming, P. Martin, M. Tomic-Canic, Wound repair and regeneration: Mechanisms, 27. K.-D. Bissig, T. T. Le, N.-B. Woods, I. M. Verma, Repopulation of adult and neonatal mice with
signaling, and translation. Sci. Transl. Med. 6, 265sr6 (2014). human hepatocytes: A chimeric animal model. Proc. Natl. Acad. Sci. U.S.A. 104, 20507–20511 (2007).
3. A. Trounson, C. McDonald, Stem cell therapies in clinical trials: Progress and challenges. 28. M. J. Hodgman, A. R. Garrard, A review of acetaminophen poisoning. Crit. Care Clin. 28,
Cell Stem Cell 17, 11–22 (2015). 499–516 (2012).
4. Y. Zhang, A. Desai, S. Y. Yang, K. B. Bae, M. I. Antczak, S. P. Fink, S. Tiwari, J. E. Willis, 29. I. M. Whyte, B. Francis, A. H. Dawson, Safety and efficacy of intravenous N-acetylcysteine
N. S. Williams, D. M. Dawson, D. Wald, W.-D. Chen, Z. Wang, L. Kasturi, G. A. Larusch, for acetaminophen overdose: Analysis of the Hunter Area Toxicology Service (HATS)
L. He, F. Cominelli, L. Di Martino, Z. Djuric, G. L. Milne, M. Chance, J. Sanabria, C. Dealwis, database. Curr. Med. Res. Opin. 23, 2359–2368 (2007).

www.ScienceTranslationalMedicine.org 17 August 2016 Vol 8 Issue 352 352ra108 13

RESEARCH ARTICLE

30. R. Yang, K. Miki, X. He, M. E. Killeen, M. P. Fink, Prolonged treatment with N-acetylcystine 42. F. Mou, M. Praskova, F. Xia, D. Van Buren, H. Hock, J. Avruch, D. Zhou, The Mst1 and Mst2
delays liver recovery from acetaminophen hepatotoxicity. Crit. Care 13, R55 (2009). kinases control activation of rho family GTPases and thymic egress of mature thymocytes.
31. Y. Xie, A. Ramachandran, D. G. Breckenridge, J. T. Liles, M. Lebofsky, A. Farhood, H. Jaeschke, J. Exp. Med. 209, 741–759 (2012).
Inhibitor of apoptosis signal-regulating kinase 1 protects against acetaminophen-induced liver 43. I. Mannaerts, S. B. Leite, S. Verhulst, S. Claerhout, N. Eysackers, L. F. R. Thoen, A. Hoorens,
injury. Toxicol. Appl. Pharmacol. 286, 1–9 (2015). H. Reynaert, G. Halder, L. A. van Grunsven, The Hippo pathway effector YAP controls
32. D. Zhou, Y. Zhang, H. Wu, E. Barry, Y. Yin, E. Lawrence, D. Dawson, J. E. Willis, S. D. Markowitz, mouse hepatic stellate cell activation. J. Hepatol. 63, 679–688 (2015).
F. D. Camargo, J. Avruch, Mst1 and Mst2 protein kinases restrain intestinal stem cell prolifer-
ation and colonic tumorigenesis by inhibition of Yes-associated protein (Yap) overabundance. Acknowledgments: We thank J. Avruch for his discussions and comments on the manuscript.
Proc. Natl. Acad. Sci. U.S.A. 108, E1312–E1320 (2011). Funding: This work was supported by the National Basic Research Program (973) of China
33. J. Cai, N. Zhang, Y. Zheng, R. F. de Wilde, A. Maitra, D. Pan, The Hippo signaling pathway [2015CB910502 (to L.C.) and 2012CB917202 (to C.-H.Y.)], the National Natural Science Foundation
restricts the oncogenic potential of an intestinal regeneration program. Genes Dev. 24, of China [31270918, U1505224, and J1310027 (to D.Z.); 81422045, U1405223, and 21272195
2383–2388 (2010). (to X.D.); 81372617, 81422018, and U1405225 (to L.C.); 31270769 (to C.-H.Y.); and 81472229 (to
34. H. S. Cooper, S. N. Murthy, R. S. Shah, D. J. Sedergran, Clinicopathologic study of dextran L.H.)], the 111 Projects (B12001 and B06016), the Fundamental Research Funds for the Central
sulfate sodium experimental murine colitis. Lab. Invest. 69, 238–249 (1993). Universities of China-Xiamen University [CXB2014004 (to Z.J.), 20720140551 (to L.C.),
35. J. P. Iredale, Models of liver fibrosis: Exploring the dynamic nature of inflammation and 20720160064 and 2013121032 (to X.D.), and 2013121034 and 20720140537 (to D.Z.)]. The
repair in a solid organ. J. Clin. Invest. 117, 539–548 (2007). funders had no role in the study design, data collection and analysis, decision to publish, or
36. S. Jiao, H. Wang, Z. Shi, A. Dong, W. Zhang, X. Song, F. He, Y. Wang, Z. Zhang, W. Wang, preparation of the manuscript. Author contributions: D.Z., X.D., C.-H.Y., and L.C. conceived
X. Wang, T. Guo, P. Li, Y. Zhao, H. Ji, L. Zhang, Z. Zhou, A peptide mimicking VGLL4 the project. D.Z., L.C., X.D., and F.F. performed data analysis/statistics. X.D., N.S.G., and Z.H.
function acts as a YAP antagonist therapy against gastric cancer. Cancer Cell 25, 166–180 conceived and performed chemical synthesis of XMU-MP-1 and its analogs and small-molecule
(2014). structure determination. L.-L.K. and C.-H.Y. designed and performed cocrystal structure study of

Downloaded from http://stm.sciencemag.org/ by guest on October 11, 2018

37. Y. Liu-Chittenden, B. Huang, J. S. Shim, Q. Chen, S.-J. Lee, R. A. Anders, J. O. Liu, D. Pan, MST2 with XMU-MP-1. D.Z. and L.C. conceived the biological study. F.F., S.Z., J.Y., C.X., Xihuan Sun,
Genetic and pharmacological disruption of the TEAD–YAP complex suppresses the oncogenic L. Huang, X.W., and Z.J. performed biochemical assay. F.F., J.G., P.W., L. Hong, and Q.W. performed
activity of YAP. Genes Dev. 26, 1300–1305 (2012). cellular experiments. Q.C., S.Z., H.L., Xiufeng Sun, W.Z., and Y.L. performed animal model study. D.Z.,
38. R. Anand, J. Maksimoska, N. Pagano, E. Y. Wong, P. A. Gimotty, S. L. Diamond, E. Meggers, Q.Y., N.-S.X., and L.Y. designed and performed the human hepatocyte in vivo study. C.-H.Y., X.D.,
R. Marmorstein, Toward the development of a potent and selective organoruthenium L.C., and D.Z. co-wrote the paper. All authors edited the manuscript. Competing interests: A
mammalian sterile 20 kinase inhibitor. J. Med. Chem. 52, 1602–1611 (2009). patent has been filed with Chinese patent application no. 201610121108.5. Data and materials
39. W. E. Naugler, B. D. Tarlow, L. M. Fedorov, M. Taylor, C. Pelz, B. Li, J. Darnell, M. Grompe, availability: The structure was deposited in the PDB with ID 5DH3.
Fibroblast growth factor signaling controls liver size in mice with humanized livers. Gastroenterology
149, 728–740.e15 (2015). Submitted 12 January 2016
40. J. S. Duffield, S. J. Forbes, C. M. Constandinou, S. Clay, M. Partolina, S. Vuthoori, S. Wu, R. Lang, Accepted 24 July 2016
J. P. Iredela, Selective depletion of macrophages reveals distinct, opposing roles during liver Published 17 August 2016
injury and repair. J. Clin. Invest. 115, 56–65 (2005). 10.1126/scitranslmed.aaf2304
41. J. Geng, X. Sun, P. Wang, S. Zhang, X. Wang, H. Wu, L. Hong, C. Xie, X. Li, H. Zhao, Q. Liu,
M. Jiang, Q. Chen, J. Zhang, Y. Li, S. Song, H.-R. Wang, R. Zhou, R. L. Johnson, K.-Y. Chien, Citation: F. Fan, Z. He, L.-L. Kong, Q. Chen, Q. Yuan, S. Zhang, J. Ye, H. Liu, X. Sun, J. Geng,
S.-C. Lin, J. Han, J. Avruch, L. Chen, D. Zhou, Kinases Mst1 and Mst2 positively regulate L. Yuan, L. Hong, C. Xiao, W. Zhang, X. Sun, Y. Li, P. Wang, L. Huang, X. Wu, Z. Ji, Q. Wu, N.-S. Xia,
phagocytic induction of reactive oxygen species and bactericidal activity. Nat. Immunol. N. S. Gray, L. Chen, C.-H. Yun, X. Deng, D. Zhou, Pharmacological targeting of kinases MST1 and
16, 1142–1152 (2015). MST2 augments tissue repair and regeneration. Sci. Transl. Med. 8, 352ra108 (2016).

www.ScienceTranslationalMedicine.org 17 August 2016 Vol 8 Issue 352 352ra108 14

Pharmacological targeting of kinases MST1 and MST2 augments tissue repair and
regeneration
Fuqin Fan, Zhixiang He, Lu-Lu Kong, Qinghua Chen, Quan Yuan, Shihao Zhang, Jinjin Ye, Hao Liu, Xiufeng Sun, Jing
Geng, Lunzhi Yuan, Lixin Hong, Chen Xiao, Weiji Zhang, Xihuan Sun, Yunzhan Li, Ping Wang, Lihong Huang, Xinrui
Wu, Zhiliang Ji, Qiao Wu, Ning-Shao Xia, Nathanael S. Gray, Lanfen Chen, Cai-Hong Yun, Xianming Deng and Dawang
Zhou

Sci Transl Med 8, 352ra108352ra108.

DOI: 10.1126/scitranslmed.aaf2304

Downloaded from http://stm.sciencemag.org/ by guest on October 11, 2018

Drug-induced regeneration
Popping a pill to repair an organ may eventually become reality. Turning away from conventional scaffolds,
materials, and cell-based regenerative medicine strategies, Fan and colleagues sought a small molecule that could
specifically target a critical signaling molecule in the Hippo pathway. Loss of kinases in this pathway, MST1/2,
increases cell proliferation during development; thus, the authors hypothesized that inhibiting their activity in
mature organs could help repair any damage. They discovered a drug, XMU-MP-1, that blocked MST1/2 activity
and found that it promoted liver repair and regeneration in four different mouse models of acute and chronic
injuries, including acetaminophen-induced injury, which is a common cause of liver failure worldwide. Such a
pharmacological strategy could make tissue regeneration easier for many, compared to complex biomaterial and
cell therapies.

ARTICLE TOOLS http://stm.sciencemag.org/content/8/352/352ra108

SUPPLEMENTARY http://stm.sciencemag.org/content/suppl/2016/08/15/8.352.352ra108.DC1
MATERIALS

RELATED http://stm.sciencemag.org/content/scitransmed/8/357/357ra122.full
CONTENT
http://science.sciencemag.org/content/sci/356/6342/1020.full
http://stke.sciencemag.org/content/sigtrans/11/547/eaaj1757.full

REFERENCES This article cites 43 articles, 12 of which you can access for free
http://stm.sciencemag.org/content/8/352/352ra108#BIBL

PERMISSIONS http://www.sciencemag.org/help/reprints-and-permissions

Use of this article is subject to the Terms of Service

Science Translational Medicine (ISSN 1946-6242) is published by the American Association for the Advancement of
Science, 1200 New York Avenue NW, Washington, DC 20005. 2017 © The Authors, some rights reserved; exclusive
licensee American Association for the Advancement of Science. No claim to original U.S. Government Works. The
title Science Translational Medicine is a registered trademark of AAAS.