1

Musashi-1 promotes stress-induced tumor progression through 1 recruitment of AGO2 2

Hsiao-Yun Chen1,2†, Mong-Lien Wang1,3,4,†, Benoit Laurent5, Chih-Hung Hsu6,7*, Ming-3 Teh Chen1,2,8, Liang-Ting Lin1,3, Jia Shen9, Wei-Chao Chang10, Jennifer Hsu9,10, Mien-4 Chie Hung9,10, Yi-Wei Chen2,8,11, Pin-I Huang2,11, Yi-Ping Yang1,2,4, Chung-Pin Li2,11, 5 Hsin-I Ma12, Chung-Hsuan Chen13, Wen-Chang Lin14, Shih-Hwa Chiou1,2,3,13* 6 7 1Division of Basic Research, Department of Medical Research, Taipei Veterans General 8 Hospital, Taipei, Taiwan. 9 2Institute of Clinical Medicine, School of Medicine, National Yang-Ming University, Taipei, 10 Taiwan. 11 3Institute of Pharmacology, School of Medicine, National Yang-Ming University, Taipei, 12 Taiwan. 13 4Institute of Food Safety and Health Risk Assessment, School of Pharmaceutical 14 Sciences, National Yang-Ming University, Taipei, Taiwan. 15 5Boston Children Hospital and Harvard Medical School, Boston MA, USA. 16 6Department of Public Health, and Women’s Hospital, Zhejiang University School of 17 Medicine, Hangzhou, Zhejiang, China. 18 7Division of Newborn Medicine and Epigenetics Program, Department of Medicine, 19 Boston Children’s Hospital, Boston, and Department of Cell Biology, Harvard Medical 20 School, Boston, Massachusetts, USA 21 8Department of Neurosurgery, Taipei Veterans General Hospital, Taipei, Taiwan. 22 9Department of Molecular and Cellular Oncology, University of Texas MD Anderson 23 Cancer Center, Houston, Texas, USA. 24 10Graduate Institute of Biomedical Sciences and Center for Molecular Medicine, China 25 Medical University, Taichung, Taiwan 26 11Cancer Center, Taipei Veterans General Hospital, Taipei, Taiwan. 27 12Department of Neurological Surgery, Tri-Service General Hospital and National Defense 28 Medical Center, Taipei, Taiwan. 29 13Genomic Research Center, Academia Sinica, Taipei, Taiwan. 30 14Institute of Biomedical Sciences, Academia Sinica, Taipei, Taiwan. 31 †Equal contributions (co-first) 32 33

2

* To whom correspondence should be addressed: Dr Shih-Hwa Chiou (email: 34 shchiou@vghtpe.gov.tw) and Dr. Chih-Hung Hsu (email: ch_hsu@zju.edu.cn); Shih-35 Hwa Chiou, MD, PhD, Department of Medical Research, Taipei Veterans General 36 Hospital; Institute of Pharmacology, School of Medicine, National Yang-Ming University; 37 No.201, Sec., Shih-Pai Road, Taipei, Taiwan, ROC. 38 39 Running title: Musashi-1 directs AGO2 to regulate mRNA stability in tumors. 40 41

3

Abstract 42 Carcinomatous progression and recurrence are the main therapeutic challenges 43 frequently faced by patients with refractory tumors. However, the underlined molecular 44 mechanism remains obscure. 45 Methods: We found Musashi-1 (MSI1) transported into cytosol under stress condition by 46 confocal microscopy and cell fractionation. Argonaute 2 (AGO2) was then identified as a 47 cytosolic binding partner of MSI1 by Mass Spectrametry, immunoprecipitation, and 48 recombinant protein pull-down assay. We used RNA-IP to determine the MSI1/AGO2 49 associated regions on downstream target mRNAs. Finally, we overexpressed C-terminus 50 of MSI1 to disrupt endogenous MSI1/AGO2 interaction and confirm it effects on tmor 51 progression. 52 Results: Malignant tumors exhibit elevated level of cytosolic Musashi-1 (MSI1), which 53 translocates into cytosol in response to stress and promote tumor progression. Cytosolic 54 MSI1 forms a complex with AGO2 and stabilize or destabilize its target mRNAs by 55 respectively binding to their 3´ untranslated region or coding domain sequence. Both 56 MSI1 translocation and MSI1/AGO2 binding are essential for promoting tumor 57 progression. Blocking MSI1 shuttling by either chemical inhibition or point mutation 58 attenuates the growth of GBM-xenografts in mice. Importantly, overexpression of the C-59 terminus of MSI1 disrupts endogenous MSI1/AGO2 interaction and effectively reduces 60 stress-induced tumor progression. 61 Conclusion: Our findings highlight novel molecular functions of MSI1 during stress-62 induced carcinomatous recurrence, and suggest a new therapeutic strategy for refractory 63

4

malignancies by targeting MSI1 translocation and its interaction with AGOs. 64 Key words: Musashi-1, Argonaute 2, RNA regulation, subcellular translocation, cancer, 65 tumor recurrence 66 67

5

Graphical Abstract 68 69

70 Under stress conditions, MSI1 translocates to cytosol and binds to AGO2 to promote 71 tumorigenesis. Blocking the MSI1/AGO2 interaction with MSI1 C-terminus decoy 72 suppress MSI1-dependent drug resistance and tumor progression. 73 74 75

6

Introduction 76 RNA binding proteins (RBPs) play crucial roles in various cellular processes by regulating 77 the post-transcriptional control of their mRNA targets, such as microRNA biogenesis, RNA 78 localization, translation and stability [1-6]. The RBP family of Musashi proteins, composed 79 of Musashi-1 (MSI1) and Musashi-2, exerts an essential control over multiple cellular 80 functions [7], such as the maintenance of self-renewal and pluripotency state in stem cells 81 [8]. Dysfunctions in the expression or activity of this family have been shown to lead to 82 tumorigenesis of glioblastoma (GBM) or pancreatic ductal adenocarcinoma (PDAC) [9, 83 10]. MSI1 was recently reported to directly target the 3’ untranslated region (3’ UTR) of its 84 target mRNAs to suppress their translation [11]. MSI1 also cooperates with LIN28 RBP to 85 inhibit the post-transcriptional biogenesis of miRNAs in embryonic stem cells [12]. 86 Increasing evidence points to the role of MSI1 in tumorigenesis and cancer proliferation 87 [13]. High level of MSI1 expression has been observed in several tumor tissues [9, 10, 88 14-17], and is associated with poor survival of grade III/IV gliomas patients [18]. Although 89 these studies suggest the involvement of MSI1 in malignancy, its functional roles and 90 molecular mechanisms underlying carcinomatous recurrence remain largely unknown. 91 The Argonaute (AGO) proteins, also part of the RBP family, play a central role in RNA 92 silencing processes by mediating the decay and translational inhibition of their targets 93 [19-21]. In many carcinomas, AGO2 is found to be ectopically overexpressed[19], and 94 several studies indicated that AGO2 could directly be involved in cancers progression by 95 interacting with oncogenic factors like EGFR[22]. AGO2 also responds to stress 96 stimulation by remodeling its interactions with target mRNAs and by modulating their post-97

7

transcriptional control [23]. By remodeling its occupancy on the 3´ UTR and coding 98 sequence (CDS) region of target mRNAs, AGO2 adjusts the translation rate of specific 99 group of genes [23]. However, the mechanisms by which AGO2 coordinates the 100 translation rate of specific targets in response to stresses in malignant progression are 101 still unclear. 102 In this study, we report that, in response to stress, MSI1 translocates into the cytosol 103 where it recruits AGO2 and post-transcriptionally regulates the expression of specific 104 target mRNAs. The binding of MSI1/AGO2 to the 3´ UTR of target mRNAs enhances their 105 degradation whereas binding to CDS prevents their rapid degradation. By coordinating 106 the two mechanisms, MSI1/AGO2 complex enhances tumor proliferation and ensures 107 cancer cell survival under hypoxia or chemodrug treatment. We also show that disrupting 108 MSI1/AGO2 interaction by overexpressing C-terminal region of MSI1 decreases stress-109 induced tumorigenicity. Notably, overexpression of MSI1 C-terminus increased sensitivity 110 to chemotherapeutic drugs, thus hindering the malignant progression of GBM, which 111 opens the possibility to treat tumor recurrence via tackling the MSI1/AGO2 complex 112 formation. 113 114

115

8

Results 116 Cytosolic translocation of MSI1 is essential for its pro-oncogenic effects under 117 stress 118 Overexpression of MSI1 has been reported in several tumor tissues [9, 10, 13-17]. Our 119 previous studies reported the correlation between MSI1 expression and GBM migration, 120 drug resistance, tumor progression; but little is known about the underlying mechanism 121 [24-26]. We first examined the correlation between MS1I and tumor progression by 122 immunohistochemical (IHC) staining on a small cohort of glioma patient samples. We 123 found that high levels of MSI1 expression positively correlated with grade of primary CNS 124 malignancy (lower in 12 meningioma and 14 low grade glioma but highly expressed in 33 125 grade 4 GBM; Figure S1A-B). Further, we overexpressed or depleted MSI1 in 05MG cell 126 lines to determine the proliferation, anti-apoptosis and tumor progression ability. Here, 127 overexpression of MSI1 increased the colony number, anti-apoptosis percentage and 128 tumor volume then Flag-control (Figure S1C-F), whereas depletion of MSI1 showed 129 opposite results (Figure S1G-J). Intriguingly, we also observed a significant proportion of 130 MSI1 proteins in the cytosol in recurrent glioma samples compared with the non-recurrent 131 samples (Figure 1A). We asked whether MSI1 could be dynamically regulated in 132 response to hypoxic or chemotherapeutic agents. To address this possibility, we first 133 exposed 05MG cells, human glial cells derived from a patient with glioblastoma (GBM), 134 to hypoxic treatment. Cell exposure to hypoxia did not affect the total level of MSI1, but 135 increased MSI1 levels in the cytosolic compartment (Figure 1B-C). Similar results were 136 obtained with primary GBM cells (Figure S2A-B) and pancreatic ductal adenocarcinoma 137

9

(PDAC) cell lines (Figure S2C-D). Addition of Leptomycin B (LMB), an inhibitor of the 138 nuclear export receptor CRM1, strongly reduced MSI1 translocation into the cytosol upon 139 hypoxic and cisplatin treatment (Figure 1B-C and Figure S2E), suggesting an active and 140 CRM1-dependent MSI1 translocation in response to environmental stress. Subcellular 141 localization is generally relied on a nuclear localization signal (NLS) and a nuclear export 142 signal (NES). Two NLS sites have been reported in the N-terminal domain of MSI1[27] 143 (Figure 1D). We identified a potential NES motif (263-LTAIPL-268) within the C-terminus 144 of MSI1 and confirmed the functionality of this motif (Figure 1D and Figure S2F). 145 Mutations in the NLS and NES motifs of MSI1 were generated; Flag-control, Flag-tagged 146 wild-type MSI1 (MSI1-wt), NES-mutant MSI1 (MSI1-NES-mut) and NLS-mutant MSI1 147 (MSI1-NLS-mut) were stably expressed in GBM cells (Figure 1E). First we confirmed no 148 differences on the protein and RNA expression levels between these stable clones 149 (Figure S3 A-B). Further, under hypoxic conditions, MSI1-NES-mut and MSI1-NLS-mut 150 remained respectively located in the nucleus and cytosol while MSI1-wt translocated into 151 the cytosol (Figure 1E and Figure S3C), suggesting that the NES motif plays an active 152 role in the subcellular translocation of MSI1 upon stress treatment. 153 We next investigated the biological consequences of MSI1 translocation. In vitro 154 functional assays showed that cells overexpressing MSI1-wt exhibited decreased 155 apoptosis, and increased proliferation and viability under hypoxia compared with Flag-156 control, MSI1-NES-mut,MSI1-NLS-mut overexpressing cells and MSI1-depleted cells. 157 (Figure S 3D-F). Consistently, in vivo studies revealed that xenografts of GBM cells 158 overexpressing MSI1-wt grew significantly bigger tumors than that of Flag-control or 159 MSI1-mutant GBM cells (Figure 1F-G). Subcellular localization of MSI1 and its mutants 160

10

in xenografts (Figure 1G-I) was consistent with that observed in Figure 1e. We further 161 explored the consequences of MSI1 shuttling in GBM under oxidative stress by treating 162 subcutaneous xenografts with cisplatin (Figure 1J). Compared with our previous 163 experiments (Figure 1F-G), cisplatin treatment enhanced tumor growth of xenografts 164 overexpressing MSI1-wt (774.365 mm3 vs 477.437 mm3 tumor volume at day 22) (Figure 165 1K-I). Mice intracranially implanted with xenografts overexpressing MSI1-wt and 166 sequentially treated with cisplatin showed an outrageous tumor invasion compared with 167 other groups (Figure 1K-I), suggesting that the nuclear-cytoplasmic shuttling of MSI1 168 strictly governs its biological function in tumorigenicity. Together, our findings showed that 169 stress-induced translocation of MSI1 is required for its pro-oncogenic functions. 170 Cytosolic MSI1 directly binds AGO2 and its target mRNAs under stress condition. 171 To address the underlying molecular mechanisms by which MSI1 shuttling promotes 172 stress-induced tumor progression, we characterized MSI1 interacting proteins by mass 173 spectrometry analysis. The Flag-tagged MSI1 protein complex in the cytosolic fraction of 174 05MG cells under normoxia or hypoxia was purified and characterized (Figure 2A and 175 Figure S4A). We found 142 proteins potentially associate with MSI1 in the cytosol. We 176 were particularly interested in those proteins that are related to stress response, such as 177 eIF3A, PABP, PKR, GCN2, and AGO2 (Figure S4B). In our immunoprecipitation data, we 178 found that the interactions of MSI1 and AGO2 sustained RNase A treatment, suggesting 179 an RNA independent manner of this RNA binding proteins interaction (Figure S4C). 180 Moreover, we found that hypoxic stress significantly enhanced the recruitment of AGO2 181 to cytosolic MSI1 in GBM and PDAC cancer cells (Figure 2B and Figure S4D-E). In vitro 182

11

binding assay confirmed the direct interaction between recombinant MSI1 and AGO2 183 (Figure 2C). Fluorescence Resonance Energy Transfer (FRET) microscopy (Figure 2D 184 and Figure S4F), and confocal microscopy in cisplatin-treated (Figure S4G) and hypoxia-185 treated (Figure S4H) cells confirmed the stress-induced interaction between MSI1 and 186 AGO2. 187 We next investigated whether AGO2 is essential for the oncogenic functions of MSI1. 188 First we showed overexpression of Flag-control, MSI1-wt, MSI1-NES-mut, or MSI1-NLS-189 mut did not change the protein and RNA levels of AGO1 and AGO2 (Figure S5A-C). 190 Knockdown and overexpression of MSI1 did not affect the protein and RNA levels of 191 AGO2, neither did knockdown of AGO2 affect the expression of MSI1 (Figure 2 E-G). 192 However, knockdown of MSI1 or AGO2 suppressed cell viability and enhanced apoptosis 193 (Figure 2H-I). We also showed that AGO2 knockdown (Figure S5D) in MSI1-194 overexpressed cells suppressed the viability (Figure 2J) and proliferation (Figure 2K) 195 through enhanced apoptosis (Figure S5E). Concomitantly, in vivo studies showed that 196 AGO2 knockdown abolished the MSI1-enhanced tumor growth (Figure S5C). These data 197 showed that AGO2 is an important downstream effector of MSI1 involved in cancer 198 development. 199 To characterize the functional roles of stress-induced AGO2-MSI1 interaction, we 200 performed RNA-binding protein immunoprecipitation sequencing (RIP-Seq) by pull-down 201 MSI1 and AGO2 respectively, and identified 336 common mRNA targets bound by 202 MSI1/AGO2 complex (Figure 2L-M). Gene Ontology (GO) analysis of the mRNAs 203 associated with both MSI1 and AGO2 showed enrichment in cell cycle progression and 204

12

apoptosis pathways (Figure 2N), consistent with the cellular phenotype we observed. We 205 further confirmed the binding of MSI1/AGO2 complex to some of the mRNA targets upon 206 hypoxia (Figure 2O and Figure S6A). Conclusively, under stress condition like hypoxia, 207 MSI1 forms a complex with AGO2 and binds to its target mRNA. 208 The MSI1/AGO2 complex regulates the stability of its target mRNA through CDS or 209 3’UTR binding. 210 We then investigated the impact of MSI1/AGO2 binding on the expression level of the 211 336 common target mRNAs. Steady mRNA levels were profiled by microarrays in control, 212 MSI1- and AGO2-knockdown cells cultured under normoxia or hypoxia. Intriguingly, we 213 observed two distinct groups of mRNA targets, with the first group (group 1) enriched in 214 apoptotic genes exhibiting increased mRNA level after knockdown of MSI1 and AGO2 215 under hypoxia, and the second group (group 2) of genes mainly involved in cell cycle 216 regulation showing opposite regulatory trend (Figure 3A). We selected three mRNA 217 targets from each group - NF2, TP53, and p21 from group 1, and CCND1, CDK4, HELLS 218 from group 2 – and evaluated their degradation rate under normoxia and hypoxia by 219 treating cells with actinomycin D (Figure 3B). In control cells, hypoxic stress decreased 220 the half-life of group1 mRNAs while increased the stability of group2 mRNAs. In contrast, 221 we observed the opposite effect in MSI1- and AGO2-knockdown cells (Figure 3B), 222 suggesting that MSI1/AGO2 complex could stabilize in response to stress a subset of 223 mRNA targets related to cell cycle (group 1) to subsequently promote tumor progression. 224 Along with this idea, MSI1/AGO2 binding could also negatively regulate the stability of 225 another subset of mRNA targets related to apoptosis (group 2) to ensure cancer cell 226

13

survival. We verified our hypothesis by qPCR and confirmed the existence of two distinct 227 types of regulation: 1) the stability of mRNA targets from group 1 decreased in response 228 to hypoxia and 2) the mRNA targets from group 2 remained expressed at similar levels 229 after hypoxia (Figure 3C). Of note, AGO2 is essential for the stress-induced and MSI1-230 mediated regulation of downstream mRNAs as AGO2-knockdown in MSI1-231 overexpressed cells abrogated the regulation of group1 and group 2 mRNA stability 232 (Figure 3D). Consistent with cytosolic MSI1-AGO2 interaction, we found that cells 233 overexpressing MSI1 mutants, with detoured subcellular localization, recaptured the 234 functional consequence on mRNA stability in a similar manor to the one caused by 235 knockdown of MSI1 in response to hypoxia (Figure 3E). Finally, we also determined the 236 protein expression level of group 1 and group 2. Intriguingly, we noticed that the group1 237 targets of NF2, TP53, and p21 were increased while the group 2 targets of CCND1, CDK4 238 and HELLS were decreased in MSI1-depleted or AGO2-depleted cells under hypoxia and 239 recovery condition (Figure 3F). The same results were also observed in the MSI1-NES-240 mut and MSI1-NLS-mut transfected cells (Figure 3G). 241 We further affirmed the phenomenon of MSI1-WT, MSI1-NLS-mut, and MSI1-NES-mut 242 overexpression under cisplatin treatment in an in vivo orthotropic xenograft mouse model 243 (Figure 4A). Orthotropic injection of cells overexpressing MSI1-wt combined with cisplatin 244 treatment led to increased size tumor, suggesting that these cells were highly resistant to 245 the cisplatin treatment (Figure 4B). In addition, growth of MSI1-NES-mut and MSI1-NLS-246 mut tumors was reduced compared to that of the parental cells, suggesting cisplatin 247 sensitivity (Figure 4B). Consistent with our previous results, the protein expression of 248 group 1 genes (NF2, p21, TP53) was decreased and the expression of group 2 genes 249

14

(CCND1 and CDK4) was upregulated in MSI1-WT tumors after cisplatin treatment (Figure 250 4C). Moreover, group 1 mRNA targets were downregulated whereas group 2 mRNA 251 targets were upregulated in MSI1-wt tumors. No changes in the expression of mRNA 252 targets were observed in the MSI1 mutant tumors (Figure 4D). Immunoprecipitation of 253 MSI1 and AGO2 from the orthotropic xenograft tumor tissue confirmed the interaction 254 between MSI1 and AOG2 in the MSI1-wt but not MSI1-NES-mut and MSI1-NLS-mut 255 tumors (Figure 4E). Together, these results indicated that the MSI1-AGO2 interaction and 256 MSI1 translocation capability governs the fate of their downstream targets as well as 257 chemodrug-resistant tumor progression of GBM. 258 MSI1/AGO2 binding to a specific location on its targets mediates distinct mRNA 259 fates 260 To decipher the molecular mechanisms by which MSI1/AGO2 complex regulates mRNA 261 stability, we carried out RIP experiments in control, MSI1- and AGO2-knockdown cells 262 under normoxia and hypoxia. Interestingly, MSI1 bound to its mRNA targets under normal 263 and hypoxic conditions while AGO2 bound to its targets only under hypoxia (Figure 5A). 264 Knockdown of AGO2 did not affect MSI1 binding to its mRNA targets whereas knockdown 265 of MSI1 hampered AGO2 recruitment (Figure 5A), suggesting a MSI1-dependent 266 recruitment of AGO2 to mRNA targets under hypoxia. We next investigated the impact of 267 MSI1 shuttling on MSI1-mRNA complex formation. To do so, we performed RIP 268 experiments with the cytosolic and nuclear fractions of cells overexpressing MSI1-wt 269 cultured under normoxia and hypoxia conditions. We showed that under normoxia, MSI1-270 mRNA complexes were in the nucleus and that upon hypoxia, they were enriched in the 271

15

cytosol (Figure 5B, top right panel), suggesting an active translocation of MSI1-mRNA 272 complexes into the cytosol in response to hypoxia. When MSI1 failed to shuttle into the 273 cytosol (MSI1-NES-mut), the MSI1-mRNA complex remained in the nuclear compartment 274 as expected (Figure 5B, middle right panel). Surprisingly, the cytosolic mutant of MSI1 275 (MSI1-NLS-mut) was unable to bind RNA (Figure 5B, bottom panel). Our data suggested 276 that MSI1 first needs to bind its target mRNAs in the nucleus before carrying them into 277 the cytosol in response to hypoxia. Consistently, the recruitment of AGO2 to the MSI1-278 mRNA complexes occurred in the cytosol (Figure 5B, top left panel). However, MSI1-NLS-279 mut did not interact with AGO2 (Figure 5B, bottom left panel), suggesting their interaction 280 to be RNA-dependent in the cytosol. We next further characterized binding regions of 281 MSI1/AGO2 complex on its mRNA targets by performing a modified-RIP assay which 282 started with RIP to precipitated MSI1 and AGO2 along with their bound mRNAs, followed 283 by RNA fragmentation and qPCR with region-specific primers (Figure 5C). We showed 284 that, in response to hypoxia, MSI1/AGO2 complex bound the three prime untranslated 285 (3'-UTR) region of target mRNAs from group 1 while it bound the coding sequence (CDS) 286 region of those from group 2 (Figure 5C). Collectively, our data showed that upon hypoxia, 287 MSI1 together with its bound mRNA targets translocate into the cytosol where it 288 subsequently recruits AGO2 to mediate two distinct types of posttranscriptional regulation: 289 degradation of mRNA targets via binding their 3’-UTR (group 1) and stabilization of mRNA 290 targets through binding their CDS (group 2) (Figure 5D). 291 Disrupting MSI1/AGO2 interaction restrains tumor growth and alters mRNA 292 regulation 293

16

As MSI1 engages AGO2 to promote tumor progression through mRNA regulation, we 294 asked whether the disruption of MSI1/AGO2 interaction could affect the tumor growth 295 driven by cytosolic MSI1. To do so, we first mapped MSI1/AGO2 interaction using deletion 296 mutants of MSI1 (Figure 6A). We performed an in vitro binding assay by incubating the 297 purified full-length, the N-terminal or C-terminal domain of MSI1 with purified AGO2 298 protein, and found that AGO2 preferentially interacted with the C-terminal domain of MSI1 299 (Figure 6B). We next investigated whether the C-terminal domain of MSI1 (Figure 6C) 300 could act as a decoy to withdraw MSI1/AGO2 protein-protein interaction. Flag control 301 (Flag) or Flag-tagged MSI1 C-terminus (Flag-C-term) were transiently expressed in GBM 302 cells which were then subjected to immunoprecipitation against endogenous MSI1 in 303 normal and hypoxic conditions. We found that overexpression of Flag-C-term disrupted 304 the interaction between endogenous MSI1 and AGO2 (Figure 6D), confirming the 305 importance of the MSI1 C-terminal domain in this protein-protein interaction. Confocal 306 microscopy further confirmed uncoupled co-localization of endogenous MSI1 and AGO2 307 in cells transfected with Flag-C-term (Figure 6E). To precise the molecular mechanisms 308 underlying the biological effects of Flag-C-term, we performed modified-RIP assay and 309 demonstrated that Flag-C-term interfered with the recruitment of AGO2 to its mRNA 310 targets, at both 3’UTR and CDS regions (Figure 6F). We next analyzed the expression of 311 MSI1/AGO2 mRNA targets and showed that, in cells transfected with Flag-C-term, the 312 expression of mRNA targets from group 1 increased while that of targets from group 2 313 decreased after hypoxia compared to the control cells (Figure 6G). The protein 314 expression level of MSI1/AGO2 targets was increased in group 1 but decreased in groups 315 2 (Figure S7A). We further evaluated the downstream effects of MSI1/AGO2 complex 316

17

disruption by analyzing cell viability in cells transfected with Flag or Flag-C-term. Our 317 results showed a decreased percentage of viable cells after Flag-C-term expression 318 (Figure 6H) which is consistent with the decreased number of soft agar colonies (Figure 319 6I), and the increased percentage of apoptotic cells (Figure 6J). Taken together, these 320 results indicated that, by disrupting MSI1/AGO2 interaction, Flag-C-term suppressed 321 clonogenic growth and promoted apoptosis. To test the therapeutic potent of MSI1 C-322 terminus, we launched an in vivo study in which cancer cells were implanted on each 323 flank of the mice subsequently subjected to an intratumoral transfection of Flag control or 324 Flag-C-term (Figure 7A). We observed that the growth of tumors derived either from 325 MSI1- overexpressing GBM cells or MIA-PaCa2 pancreatic cancer cells was strongly 326 delayed by the administration of Flag-C-term (Figure 7B-C). Overexpressed C-terminus 327 in tumor cells disrupts endogenous MSI1/AGO2 interaction and reduces tumor volume 328 (Figure 7D-E). Collectively, our findings demonstrated that disrupting MSI1/AGO2 329 interaction with MSI1-C-term decoy suppressed tumor growth by blocking the recruitment 330 of AGO2 to its target mRNAs and subsequently altering their stability. 331 332 The MSI1/AGO2 pathway is enhanced in patients with tumor relapse 333 Cytosolic MSI1 engages AGO2 to promote stress-induced tumor growth through RNA 334 regulation. We showed that a significant proportion of MSI1 was cytosolic in samples from 335 patients with high grade glioma (Figure 1A) but it remains unclear whether MSI1/AGO2 336 pathway actively participates to tumorigenicity in patients and could also have an impact 337 on tumor recurrence. To address this point, we collected eighteen pairs of primary and 338

18

recurrent GBM samples from patients who received concurrent chemotherapeutics after 339 primary surgery (Tables S2) and analyzed the subcellular localization of MSI1 by IHC 340 staining. We showed that a significant proportion of MSI1 proteins were cytosolic in the 341 GBM recurrent samples whereas MSI1 was barely detectable in the cytosol in the primary 342 GBM samples (Figure 8A). We then collected by laser capture microdissection (LCM) 343 tissue samples from the tumor (T) and non-tumor stroma (S)[28] and analyzed the 344 expression level of MSI1 target mRNAs by qPCR. We observed that, for each pair of 345 samples, the expression of mRNA targets related to apoptosis (group 1) was decreased 346 in recurrent GBM compared to the primary GBM while the expression of the mRNA targets 347 related to cell cycle (group 2) was increased (Figure 8B). The same results were also 348 observed in a follow-up of a cohort of primary (n = 67) and recurrent (n = 32) GBM patients 349 (Figure 8C and Table S3). These results indicated that, in patients with recurrent GBM, 350 the MSI1/AGO2 pathway was enhanced to promote tumor growth and ensure cancer cell 351 survival. 352 To address whether the correlation between MSI1/AGO2 pathway and tumor recurrence 353 could be generalized to other cancer types, we collected samples from patients with 354 pancreatic ductal adenocarcinoma (PDAC) and performed IHC staining of MSI1 on non-355 recurrent (n = 18) and recurrent (n = 61) PDAC samples (Table S4). We observed that 356 around 5% of non-recurrent pancreatic samples exhibited MSI1 in the cytosol (1/18 cases; 357 data not shown) while 60% of recurrent PDAC samples (37/61 cases) displayed cytosolic 358 MSI1 (Figure 8D), suggesting that cytosolic MSI1 was associated with tumor recurrence. 359 By analyzing clinical data of the recurrent PDAC samples, we observed that patients with 360 recurrent PDAC positive for cytosolic MSI1 presented overall a lower survival than those 361

19

negative for cytosolic MSI1 (Figure 8E). We further analyzed the 37 recurrent PDAC 362 cases positive for cytosolic MSI1 and classified them based on IHC staining score 363 (IHC<0.5 or IHC>0.5). We showed that the group of patients with high score (IHC > 0.5) 364 exhibited a lower survival outcome than that with low score (IHC < 0.5) (Figure 8F-G), 365 suggesting that the level of cytosolic MSI1 could predict patient survival. Collectively, our 366 results indicated that the cytosolic MSI1/AGO2 complex interaction and downstream 367 pathway was significantly enhanced in patients with tumor relapse and could affect patient 368 survival. 369

20

Discussion 370 GBM and PDAC are highly malignant tumors with dismal prognosis and a short survival 371 time even after a completed treatment [28, 29]. Environmental stresses play a pivotal role 372 in both stem cell maintenance and tumorigenesis. Here we reported that MSI1 shuttles 373 into the cytosol under stress to form a complex with AGO2 that can stabilize or destabilize 374 its target mRNAs. Sutherland and colleagues reported in 2017 that MSI2 translationally 375 repressed Piwil1, another member of the Argonaute protein family, during male gamete 376 development [30], directly linking musashi proteins with the Argonaute family. In our study, 377 we identify the MSI1/AGO2 interaction by proteomic analysis from immunoprecipitation 378 in cancer cells (Figure S4B-C). It seems to us that the two proteins did not regulate the 379 protein expression levels of each other as knockdown of MSI1 did not change the protein 380 expression of AGO2, neither did vice versa (Figure 2E and 2G). In the case of MSI1 and 381 AGO2, the two proteins work together to determine the fate of their downstream RNA 382 targets. 383 Cytosolic MSI1 enhances tumor proliferation and cancer cell survival through this AGO2-384 dependent mRNA regulation. AGO2 is essential for MSI1 pro-tumor progression as its 385 knockdown inhibited the MSI1-enhanced tumor growth (Figure S4F). Intact functions of 386 MSI1 shuttling are essential to drive the tumorigenicity (Figure 1D-L). The sequestration 387 of MSI1 in the nucleus by disrupting the NES resulted in lack of its mRNA regulatory 388 functions while surprisingly, constitutively cytosolic MSI1 (NLS mutant) also lost its ability 389 to regulate mRNA targets. Indeed, MSI1 NLS-mutant does not bind RNA in vivo (Figure 390 5B), suggesting a model in which MSI1 must first bind its mRNA targets in the nucleus to 391

21

then relocate them into the cytoplasm. The RNA binding and sub-cellular translocation of 392 RNA-binding proteins may be regulated by post-translational modifications (PTMs)[31, 393 32]. We demonstrated that MSI1 acts as a guide to recruit AGO2 and subsequently 394 regulate the stability of their target mRNAs. The MSI1/AGO2 complex binds mRNAs 395 including genes involved in apoptosis and cell cycle progression, but exhibits two distinct 396 types of regulation. MSI1/AGO2 complex degrades mRNA targets by binding to their 397 3’UTR while it stabilizes mRNA targets by binding to their CDS (Figure 5). How can 398 MSI1/AGO2 control these two distinct modes of mRNA regulation? Target mRNAs that 399 are not engaged in translation aggregate into cytosolic organelles such as stress granules 400 (SG; for RNA stabilization) and Processing-bodies (P-bodies; for RNA degradation) [33-401 37]. AGO2 has been reported to remodel its interactions with target mRNAs and regulate 402 their post-transcriptional control under stress responses [23, 38]. In response to stress, 403 AGO2 could remodel its occupancy on the 3’UTR and CDS of target mRNAs and then 404 adjust the translation rate of specific group of genes. Although how AGO2 coordinates its 405 functions in response to stresses is still unclear, our findings highlighted a sophisticated 406 mechanism by which MSI1/AGO2 complex promote tumor progression. In response to 407 stress, the complex stabilizes a subset of mRNAs related to cell cycle to promote tumor 408 progression but also regulate the stability of another subset of mRNA targets related to 409 apoptosis to ensure cancer cell survival. Interestingly, MSI1 and AGO2 could also have 410 independent mRNA regulatory functions. We identified other subsets of genes that are 411 only regulated by either MSI1 or AGO2 (Figure 3A). These distinct regulations could rely 412 on different protein-protein interactions. For example, we showed that MSI1 is able to 413 specifically interact with other proteins in response to hypoxia (PABPC1, PKR, GCN2, 414

22

eIF3A, eIF2C2) (Figure S4B-C) that could play a role in a different type of regulation. 415 Moreover, we showed that AGO2 specifically associates with the C-terminal region of 416 MSI1. No crystal structure of this region exists and our in silico simulation indicated that 417 the C-terminus of MSI1 is an intrinsically disordered region that has no specific structure 418 and is very flexible. We demonstrated in vitro and in vivo that C-terminal terminus of MSI1 419 could work as a decoy to disrupt the endogenous MSI1/AGO2 interaction, and to 420 efficiently block the oncogenic functions of the endogenous MSI1/AGO2 complex as well 421 as inhibit xenograft tumor growth (Figure 7). 422 MSI1 is not the only RNA-binding protein that is predominantly nuclear and can shuttle to 423 the cytoplasm upon environmental stress stimulation. The RBM45 (RNA-Binding Motif 45) 424 protein translocates from nuclear to accumulate into the cytosol in Amyotrophic Lateral 425 Sclerosis (ALS) patients [39]. This shuttling of RBM45 is used as a pathological marker 426 for the disease [39]. Cytosolic MSI1 enhances tumor proliferation and cancer cell survival, 427 and highly correlates relapse occurrence in patients with GBM and PDAC (Figure 9). We 428 could use MSI1 subcellular localization as a biomarker to further improve the diagnosis 429 and prognostic of GBM and PDAC. More importantly, significant efforts should be put in 430 the future on understanding the mechanisms of MSI1 shuttling (hypothesis of specific 431 PTMs or protein-protein interaction for example). The development of drugs that can 432 prevent MSI1 translocation from the nucleus into the cytosol represent a promising 433 therapeutic approach to decrease GBM and PDAC relapse incidence after chemotherapy. 434 This study offers an alternative possibility to decrease GBM and PDAC tumor progression 435 by using a peptide-based approach. The disruption of MSI1/AGO2 interaction with decoy 436 peptides efficiently inhibits tumor growth, suggesting that peptide-based therapy could be 437

23

of importance to decrease recurrent GBM and PDAC. Similar therapeutic strategies are 438 already developed by other groups. Chang et al. used a stapled α-helical peptide to 439 efficiently abolish p53/MDM2 interaction and induce the death of cancer cells [40]. Notably, 440 overexpression of MSI1 C-terminus increased sensitivity to chemotherapeutic drugs, thus 441 hindering the malignant progression of GBM, which opens the possibility to treat tumor 442 recurrence via tackling the MSI1/AGO2 complex formation. 443 444 445 446

24

Acknowledgment 447 The authors thanks Miss Man-Sheung Chan, Miss Fu-Ting Tsai, Miss Yi-Ching Tsai, Miss 448 Yu-Ling Ko, and Hsiao-Yun Tai (all from the Department of Medical Research, TVGH) for 449 assisting this study. This study was funded by Ministry of Science and Technology (MOST) 450 (106-2319-B-001-003, 106-2119-M-010-001, 106-3114-B-010-002, and 107-2633-B-009-451 003), Taipei Veterans General Hospital (V104E14-001/V105C-077/V106E-004/ V106C-452 001/V107C-139/V107E-002-2), Veterans General Hospitals and University System of 453 Taiwan Joint Research Program (VGHUST107-G1-6-1、VGHUST108-G1-5-3), VGH, 454 TSGH, NDMC, AS Joint Research Program (VTA105-V1-5-1, and VTA107-V1-5-1), VGH, 455 NTUH Joint Research Program (VN106-02/VN107-16), the Department of Health Cancer 456 Center Research of Excellence (MOHW105-TDUB-211-134003/105-TDU-B-211-133017, 457 Excellent Clinical Trail Center/MOHW106-TDU-B-211-113001, MOHW107-TDU-B-211-458 123001, and MOHW108-TDU-B-211-133001), NRPB Human iPSC Alliance-Core Service 459 (MOST 105-108), Yen Ching Ling medical research grant (CI-108-11), and National 460 Health Research Institutes (NHRIEX106-10621BI, and NHRI-EX107-10621BI), Taiwan. 461 This work was also financially supported by the "Center for Intelligent Drug Systems and 462 Smart Bio-devices (IDS2B) " from The Featured Areas Research Center Program within 463 the framework of the Higher Education Sprout Project by the Ministry of Education (MOE), 464 and the“Cancer Progression Research Center, MOE, National Yang-Ming University”in 465 Taiwan, 466 467

25

Reference 468 1. Gerstberger S, Hafner M, Tuschl T. A census of human RNA-binding proteins. Nat 469 Rev Genet. 2014; 15: 829-45. 470 2. Muller-McNicoll M, Neugebauer KM. How cells get the message: dynamic 471 assembly and function of mRNA-protein complexes. Nat Rev Genet. 2013; 14: 275-87. 472 3. Ascano M, Gerstberger S, Tuschl T. Multi-disciplinary methods to define RNA-473 protein interactions and regulatory networks. Curr Opin Genet Dev. 2013; 23: 20-8. 474 4. Henras AK, Soudet J, Gerus M, Lebaron S, Caizergues-Ferrer M, Mougin A, et al. 475 The post-transcriptional steps of eukaryotic ribosome biogenesis. Cell Mol Life Sci. 476 2008; 65: 2334-59. 477 5. Musunuru K. Cell-specific RNA-binding proteins in human disease. Trends 478 Cardiovasc Med. 2003; 13: 188-95. 479 6. Dreyfuss G, Kim VN, Kataoka N. Messenger-RNA-binding proteins and the 480 messages they carry. Nat Rev Mol Cell Biol. 2002; 3: 195-205. 481 7. Hadziselimovic N, Vukojevic V, Peter F, Milnik A, Fastenrath M, Fenyves BG, et al. 482 Forgetting is regulated via Musashi-mediated translational control of the Arp2/3 483 complex. Cell. 2014; 156: 1153-66. 484 8. Rentas S, Holzapfel NT, Belew MS, Pratt GA, Voisin V, Wilhelm BT, et al. Musashi-485 2 attenuates AHR signalling to expand human haematopoietic stem cells. Nature. 2016; 486 532: 508-11. 487 9. Fox RG, Park FD, Koechlein CS, Kritzik M, Reya T. Musashi signaling in stem cells 488 and cancer. Annu Rev Cell Dev Biol. 2015; 31: 249-67. 489 10. Fox RG, Lytle NK, Jaquish DV, Park FD, Ito T, Bajaj J, et al. Image-based detection 490 and targeting of therapy resistance in pancreatic adenocarcinoma. Nature. 2016; 534: 491 407-11. 492

26

11. Liu X, Yang WT, Zheng PS. Msi1 promotes tumor growth and cell proliferation by 493 targeting cell cycle checkpoint proteins p21, p27 and p53 in cervical carcinomas. 494 Oncotarget. 2014; 5: 10870-85. 495 12. Kawahara H, Okada Y, Imai T, Iwanami A, Mischel PS, Okano H. Musashi1 496 cooperates in abnormal cell lineage protein 28 (Lin28)-mediated let-7 family microRNA 497 biogenesis in early neural differentiation. J Biol Chem. 2011; 286: 16121-30. 498 13. Sureban SM, May R, George RJ, Dieckgraefe BK, McLeod HL, Ramalingam S, et 499 al. Knockdown of RNA binding protein musashi-1 leads to tumor regression in vivo. 500 Gastroenterology. 2008; 134: 1448-58. 501 14. Moreira AL, Gonen M, Rekhtman N, Downey RJ. Progenitor stem cell marker 502 expression by pulmonary carcinomas. Mod Pathol. 2010; 23: 889-95. 503 15. Kanemura Y, Mori K, Sakakibara S, Fujikawa H, Hayashi H, Nakano A, et al. 504 Musashi1, an evolutionarily conserved neural RNA-binding protein, is a versatile marker 505 of human glioma cells in determining their cellular origin, malignancy, and proliferative 506 activity. Differentiation. 2001; 68: 141-52. 507 16. Wang T, Ong CW, Shi J, Srivastava S, Yan B, Cheng CL, et al. Sequential 508 expression of putative stem cell markers in gastric carcinogenesis. Br J Cancer. 2011; 509 105: 658-65. 510 17. Shu HJ, Saito T, Watanabe H, Ito JI, Takeda H, Okano H, et al. Expression of the 511 Musashi1 gene encoding the RNA-binding protein in human hepatoma cell lines. 512 Biochem Biophys Res Commun. 2002; 293: 150-4. 513 18. Dahlrot RH, Hansen S, Herrstedt J, Schroder HD, Hjelmborg J, Kristensen BW. 514 Prognostic value of Musashi-1 in gliomas. J Neurooncol. 2013; 115: 453-61. 515 19. Meister G. Argonaute proteins: functional insights and emerging roles. Nat Rev 516 Genet. 2013; 14: 447-59. 517 20. Wee LM, Flores-Jasso CF, Salomon WE, Zamore PD. Argonaute divides its RNA 518 guide into domains with distinct functions and RNA-binding properties. Cell. 2012; 151: 519

27

1055-67. 520 21. Golden RJ, Chen B, Li T, Braun J, Manjunath H, Chen X, et al. An Argonaute 521 phosphorylation cycle promotes microRNA-mediated silencing. Nature. 2017; 542: 197-522 202. 523 22. Shen J, Xia W, Khotskaya YB, Huo L, Nakanishi K, Lim SO, et al. EGFR modulates 524 microRNA maturation in response to hypoxia through phosphorylation of AGO2. Nature. 525 2013; 497: 383-7. 526 23. Karginov FV, Hannon GJ. Remodeling of Ago2-mRNA interactions upon cellular 527 stress reflects miRNA complementarity and correlates with altered translation rates. 528 Genes Dev. 2013; 27: 1624-32. 529 24. Chen HY, Lin LT, Wang ML, Laurent B, Hsu CH, Pan CM, et al. Musashi-1 530 Enhances Glioblastoma Cell Migration and Cytoskeletal Dynamics through Translational 531 Inhibition of Tensin3. Sci Rep. 2017; 7: 8710. 532 25. Chen HY, Lin LT, Wang ML, Lee SH, Tsai ML, Tsai CC, et al. Musashi-1 regulates 533 AKT-derived IL-6 autocrinal/paracrinal malignancy and chemoresistance in 534 glioblastoma. Oncotarget. 2016; 7: 42485-501. 535 26. Chen HY, Lin LT, Wang ML, Tsai KL, Huang PI, Yang YP, et al. Musashi-1 promotes 536 chemoresistant granule formation by PKR/eIF2alpha signalling cascade in refractory 537 glioblastoma. Biochim Biophys Acta Mol Basis Dis. 2018; 1864: 1850-61. 538 27. MacNicol AM, Wilczynska A, MacNicol MC. Function and regulation of the 539 mammalian Musashi mRNA translational regulator. Biochem Soc Trans. 2008; 36: 528-540 30. 541 28. Chiou GY, Chien CS, Wang ML, Chen MT, Yang YP, Yu YL, et al. Epigenetic 542 regulation of the miR142-3p/interleukin-6 circuit in glioblastoma. Mol Cell. 2013; 52: 543 693-706. 544 29. Wang-Gillam A, Li CP, Bodoky G, Dean A, Shan YS, Jameson G, et al. 545 Nanoliposomal irinotecan with fluorouracil and folinic acid in metastatic pancreatic 546

28

cancer after previous gemcitabine-based therapy (NAPOLI-1): a global, randomised, 547 open-label, phase 3 trial. Lancet. 2016; 387: 545-57. 548 30. Sutherland JM, Sobinoff AP, Fraser BA, Redgrove KA, Siddall NA, Koopman P, et 549 al. RNA binding protein Musashi-2 regulates PIWIL1 and TBX1 in mouse 550 spermatogenesis. J Cell Physiol. 2018; 233: 3262-73. 551 31. Lovci MT, Bengtson MH, Massirer KB. Post-Translational Modifications and RNA-552 Binding Proteins. Adv Exp Med Biol. 2016; 907: 297-317. 553 32. Lee EK. Post-translational modifications of RNA-binding proteins and their roles in 554 RNA granules. Curr Protein Pept Sci. 2012; 13: 331-6. 555 33. Decker CJ, Parker R. P-bodies and stress granules: possible roles in the control of 556 translation and mRNA degradation. Cold Spring Harb Perspect Biol. 2012; 4: a012286. 557 34. Buchan JR, Parker R. Eukaryotic stress granules: the ins and outs of translation. 558 Mol Cell. 2009; 36: 932-41. 559 35. Anderson P, Kedersha N. RNA granules: post-transcriptional and epigenetic 560 modulators of gene expression. Nat Rev Mol Cell Biol. 2009; 10: 430-6. 561 36. Parker R, Sheth U. P bodies and the control of mRNA translation and degradation. 562 Mol Cell. 2007; 25: 635-46. 563 37. Lopez-Orozco J, Pare JM, Holme AL, Chaulk SG, Fahlman RP, Hobman TC. 564 Functional analyses of phosphorylation events in human Argonaute 2. Rna. 2015; 21: 565 2030-8. 566 38. Detzer A, Engel C, Wunsche W, Sczakiel G. Cell stress is related to re-localization 567 of Argonaute 2 and to decreased RNA interference in human cells. Nucleic Acids Res. 568 2011; 39: 2727-41. 569 39. Collins M, Riascos D, Kovalik T, An J, Krupa K, Krupa K, et al. The RNA-binding 570 motif 45 (RBM45) protein accumulates in inclusion bodies in amyotrophic lateral 571 sclerosis (ALS) and frontotemporal lobar degeneration with TDP-43 inclusions (FTLD-572

29

TDP) patients. Acta Neuropathol. 2012; 124: 717-32. 573 40. Chang YS, Graves B, Guerlavais V, Tovar C, Packman K, To KH, et al. Stapled 574 alpha-helical peptide drug development: a potent dual inhibitor of MDM2 and MDMX for 575 p53-dependent cancer therapy. Proc Natl Acad Sci U S A. 2013; 110: E3445-54. 576 577 578 579 580

30

Figure legend 581 Figure 1. Translocation of MSI1 into the cytosol correlates with tumor progression 582 and cell proliferation under stress conditions. 583 (A) IHC staining for MSI1 in primary (n = 67) and recurrent (n = 32) GBM. Magnifying 584 power: 200× (top) and 600× (bottom). Quantitation of cells expressing cytoplasmic or 585 nuclear MSI1 is shown in the bar graph on the right. (B) 05MG cells pre-treated with or 586 without nuclear export inhibitor leptomycinB (LMB) (10 ng/mL, 2 hr) under normoxia or 587 hypoxia conditions for 24 hr were subjected to anti-MSI1 (green) immnunostaining and 588 DAPI (blue) nuclear counterstaining. Images were acquired from Carl Zeiss confocal 589 microscope system. The intensity of green fluorescence in nuclear and cytosol was 590 quantified and shown as relative percentage in the graph at the right. (C) Total protein (T), 591 nuclear (N), and cytoplasmic (C) fractionations of 05MG cells under normoxia or hypoxia 592 (24 hr) in the presence or absence of LMB (10 ng/mL) were subjected to immunoblotting 593 with MSI1, Lamin A/C (nuclear internal control) and GAPDH (cytosolic control) antibodies. 594 (D) A schematic presentation showing the mutation sites in the NLS (orange) and NES 595 (red) motifs of human MSI1. All constructs were sub-cloned into p-3xFlag-Myc-CMV 596 expression vector. (E) 05MG cells stably transfected with the Flag-control, Flag-tagged 597 MSI1-wt, MSI1-NES-mut, or MSI1-NLS-mut were subjected to normoxia (N) or hypoxia 598 (H) treatment for 24 hr and then immunostained with anti-Flag antibody (green). Images 599 were acquired from Carl Zeiss confocal microscope system, and the quantification of 600 fluorescent intensity in the nuclear and cytosolic compartments was shown as relative 601 percentage in the graph at the left. (F) Null mice were subcutaneously transplanted with 602

31

05MG/Flag-control, 05MG/MSI1-wt, 05MG/MSI1-NES-mut or 05MG/MSI1-NLS-mut cells. 603 Tumor size was measure with a caliper at the indicated time points. The 05MG/MSI1-604 NES-mut and 05MG/MSI1-NLS-mut tumors showed similar growth curves as the 605 05MG/Flag-control cells, while the 05MG/MSI1-wt tumor grew much more rapidly. N = 6. 606 **P < 0.05 vs. 05MG/Flag-controlcells. (G) Xenograft tumors were excised (top), and 607 tumor tissues were subjected to immunostaining to evaluate the expression and 608 distribution of Flag-control, Flag-tagged MSI1-wt, MSI1-NES-mut, and MSI1-NLS-mut 609 proteins (bottom). Images were acquired from Carl Zeiss confocal microscope system. 610 (H) The intensity of green fluorescence in nuclear and cytosol was quantified and shown 611 as relative percentage in the graph at the right. (I) Tumor tissue were harvested and 612 homogenized. Whole-tumor lysates were subjected to Western blot analysis. Data 613 represent the mean ± S.D. of three independent experiments performed in triplicate. (J) 614 A schematic depicting the experimental design for subcutaneously transplanted. (K-L) 615 Null mice were subcutaneously transplanted with 05MG/Flag-control, 05MG/MSI1-wt, 616 05MG/MSI1-NES-mut, or 05MG/MSI1-NLS-mut cells. Two days after the tumor size 617 reached 50 mm3, mice started cisplatin (20 mg/kg) or PBS administered via tail-vein 618 injection for total 3 times with 2-day interval. The tumor size was measured with a caliper 619 at the indicated time points. Xenograft tumors were excised 40 days after DDP treatment. 620 N = 6, **P < 0.05. 621 Figure 2. MSI1 interacts with AGO2 and binds to their common downstream mRNA 622 targets under hypoxia. 623 (A) A schematic illustrating the procedure for identifying hypoxia-induced binding partners 624

32

of MSI1. (B) Co-immunoprecipitation of endogenous AGO2 with MSI1 antibody in the 625 cytosol or nuclear fraction of 05MG cells under hypoxia for indicated period of time. (C) 626 In vitro binding assay of purified baculovirus-expressed His-tagged AGO2 and Flag-627 tagged MSI1 proteins. (D) 05MG cells-expressing FRET pairs of MSI1-orange and AGO2-628 GFP were bleached at the region of interest (ROI) indicated by yellow boxes. Unbleached 629 controls (pre-bleach) were also shown in parallel. Left, representative images of MSI1 630 (orange) and AGO2 (green) before and after photobleaching experiments. Right, 631 quantification of FRET photobleaching experiments was performed by calculating FRET 632 efficiencies for the FRET pairs MSI1 (orange)-AGO2 (green). (E-G) Western blotting 633 confirmed endogenous expression levels of MSI1, AGO2, HIF-1α and actin in MSI1-634 knockdown cells (E), AGO2-knockdown cells (F), and MSI1-overexpressed cells (G). (H) 635 05MG with AGO2-knockdown were subjected to an MTT viability assay. The relative fold 636 change of the number of viable cells in each day was presented in the graph. (I) The 637 percentage of apoptotic cells of control, MSI1-knockdown and AGO2-knockdown cells 638 were determined by external Annexin-V under normoxiac and hypoxic conditions.(J) 639 05MG/Flag-control, 05MG/Flat-MSI1-wt, and 05MG/MSI1-wt with AGO2-knockdown 640 (Flag-MSI1/shAGO2) were subjected to an MTT viability assay. The relative fold change 641 of the numbers of viable cells in each day was presented in the graph. (K) 05MG/Flag-642 control, 05MG/MSI1-wt and 05MG/MSI1-wt/shAGO2 cells were subjected to colony 643 formation assay for 10 days, and the numbers of colony were quantitated by ImageJ 644 software. (L) Flow-chart of preparing RNA-binding protein immunoprecipitation (RIP) 645 samples for NGS analysis (RIP-seq). (M) Intersection of mRNA targeted by MSI1 and 646 AGO2. (N) Gene ontology (GO) enrichment analysis was conducted by DAVID software 647

33

according to the category of biological processes. Benjamini ≤ 0.05 is selected as 648 interesting GO. The GO accession, name, and the corresponding p-value were shown in 649 the graph. (O) 05MG cells pre-treated with or without LBM (10 ng/mL, 2 hours) and 650 cultured in hypoxia condition for 24 hours were stained for TP53 and CCND1 mRNAs 651 (TAS-cy5, cherry-red), MSI1 (green), and AGO2 (red). Merged images of co-localization 652 of MSI1/AGO2/mRNA (white) by confocal microscopy are shown. 653 Figure 3. MSI1/AGO2 mediate the stability of downstream mRNA targets. 654 (A) Parental and MSI1- or AGO2-knockdown cells under normoxia and hypoxia 655 conditions were subjected to a gene expression microarray. Bioinformatics analysis of the 656 microarray data with focus on the 336 common targets of MSI1 and AGO2 identified by 657 RIP-Seq showed the hierarchical clustering these common targets in the heat map. The 658 red and green colors respectively indicate the differentially up or downregulated genes. 659 Each group were done in three distinct biological replicates and the means signals were 660 transformed to the log2 scale. (B) Actinomycin D (Act. D, 5 μg/ml) was added to parental 661 and MSI1- or AGO2-knockdown cells for the indicated times. We compared the half-life 662 distribution of TP53, NF2, CDKN1A, CCND1, CDK4 and HELLS mRNA levels between 663 parental and MSI1- or AGO2-knockdown cells. The RNA expression levels are shown 664 below each the respective box-plots. (C, E) Parential and MSI1- or AGO2-knockdown 665 05MG cells, as well as MSI1-wt, MSI1-NES-mut, and MSI1-NLS-mut transfected 05MG 666 cells were treated with normoxia or hypoxia conditions for 24 hours. Purified RNA was 667 subjected to RT-PCR with primers specific to TP53, NF2, CDKN1A, CCND1, CDK4 and 668 HELLS. The mRNA levels under hypoxia were normalized by that under normoxia and 669

34

shown as relative value in the chart. (D) 05MG/MSI1-wt and 05MG/MSI1-wt/shAGO2 670 cells were treated with normoxia or hypoxia conditions for 24 hours. Purified RNA was 671 subjected to quantitative RT-PCR with primers specific to TP53, NF2, CDKN1A, CCND1, 672 CDK4 and HELLS. The mRNA levels under hypoxia were normalized by that under 673 normoxia and shown as relative value in the chart. * P<0.05. (F) 05MG with AGO2- or 674 MSI1-knockdown, and (G) MSI1-wt, MSI1-NES-mut, or MSI1-NLS-mut overexpression 675 under normoxia or hypoxia for 24 hours, or recovery from hypoxia condition for 1 and 6 676 hours, were subjected to Western blotting with antibody of NF2, p53, p21, cyclin D1, 677 CDK4, HELLS, AGO2, MSI1, Flag and actin. 678 Figure 4. MSI1 interacts with AGO2 in the cytosol and promote tumor growth. 679 (A) Schematic illustration presenting the experimental design of xenograft orthotropic 680 tumor model. SCID mice were orthotopically transplanted with 05MG/Flag-control, 681 05MG/MSI1-wt, 05MG/MSI1-NES-mut, or 05MG/MSI1-NLS-mut GFP cells. Twenty days 682 after transplantation, mice were administered 3 times with 2-day interval of Cisplatin (20 683 mg/kg) or PBS via tail-vein injection. (B) Xenograft tumors were excised 20 days after 684 Cisplatin treatment. Representative images of GFP-positive tumors are shown. N = 6, **P 685 < 0.05. (C) Xenograft orthotropic tumor tissue were sectioned and subjected to IHC to 686 evaluate Flag-tagged tumor, NF2, p53, p21, cyclinD1, CDK4 and ki67 expression levels. 687 (D) Tumors tissues (five of each group) were harvested and homogenized. Whole-tumor 688 lysates were analyzed by qPCR. (E) Co-immunoprecipitation of endogenous AGO2 with 689 Flag-tagged MSI1 using Flag antibody in 05MG/MSI1-WT, 05MG/MSI1-NES-mut, and 690 05MG/MSI1-NLS-mut xenograft tumors tissues. 691 Figure 5. Differential regulation of mRNAs by MSI1/AGO2 complex under hypoxia. 692

35

(A) Endogenous MSI1 or AGO2 was immunoprecipitated in MSI1 or AGO2 knockdown 693 05MG cell with anti-MSI1 or anti-AGO2 antibody. Western blot of the immunoprecipitation 694 (IP) confirmed the MSI1/AGO2 interaction in hypoxia-treated parental cells but not in 695 MSI1 or AGO2 knockdown cells (top). Total RNAs isolated from IP were subjected to NF2, 696 TP53, CCND1, and HELLS mRNA quantitation by using qPCR with specific primer. 697 Quantification of mRNA expression levels experiments by normalization with IgG control. 698 Data represent the mean ± S.D. of three independent experiments performed in triplicate. 699 * P<0.05 vs IgG signal. (B) Nuclear and cytosolic fractions of 05MG/MSI1-wt, 700 05MG/MSI1-NES-mut, and 05MG/MSI1-NLS-mut cells were subjected to the 701 immunoprecipitation with Flag antibodies to pull down the complexes interacting with 702 Flag-tagged MSI1. Left, immunoprecipitates were subjected to Western blot to assess the 703 binding between AGO2 and full-length or mutated MSI1. Right, total RNAs isolated from 704 the immunoprecipitated complexes were analyzed by qRT-PCR for NF2, TP53, CCND1, 705 and HELLS mRNA levels. Fold change in mRNA levels was normalized to IgG-706 precipitated controls. Data represent the mean ± S.D. of three independent experiments 707 performed in triplicate. * P<0.05 vs IgG signal. (C) Modified-RIP analysis of the binding 708 regions of MSI1 and AOG2 on the target mRNAs. RIP were performed with anti-MSI1 or 709 anti-AGO2 followed by RNA fragmentation and qPCR of NF2, TP53, CCND1, and HELLS 710 coding sequence (CDS) and 3´ UTR. The schematic illustration showed the relative 711 locations of each pair of qPCR primers for CDS and 3’UTR regions. MSI1 or AGO2 712 palindromic-binding sequence exists within the peak. Quantification of fold changes of the 713 signals were normalized to IgG-precipitated controls. This experiments were done in three 714 distinct biological replicates. (D) A schematic illustrating the fate of mRNA determined by 715

36

the MSI1-AGO2 complex. MSI1-AGO2 regulates RNA stability of specific RNAs to sustain 716 tumor growth under stress in two ways: 1) MSI1-AGO2 facilitates tumor suppressor gene 717 mRNA decay to prevent stress-induced cell death (likely through the conventional UTR 718 binding followed by post-transcriptional repression) and 2) MSI1-AGO2 stabilizes and 719 protects mRNA of cell cycle genes to promote prompt translation upon stress removal 720 (likely through the CDS binding and subsequent aggregation in stress granules). 721 Figure 6. The C-terminal of MSI1 is critical for AGO2 binding and cell viability. 722 (A) Schematic presentation of the constructs of full-length, C-terminus, and N-terminus 723 of MSI1 as well as wild-type AGO2 for purifying recombinant proteins. (B) Pull-down 724 assay with recombinant MSI1 and AGO2 proteins showed that the C-terminus of MSI1 is 725 essential for the direct MSI1/AGO2 interaction. (C) A Schematic illustrating the Flag-726 control, full-length and C-terminal (C-term) fragment of human MSI1. The construct of 727 MSI1 C-term was sub-cloned into p-3×Flag-Myc-CMV expression vector. (D) Cells 728 transient transfected with Flag-control or Flag-tagged MSI1 C-term (Flag-C-term) were 729 subjected to co-immunoprecipitation assay for endogenous AGO2 and MSI1 protein-730 protein interaction. Transfection of the Flag-C-term blocked hypoxia-induced MSI1/AGO2 731 binding. (E) Cells transfected with Flag-control or Flag-C-term were analyzed under 732 confocal microscopy for the subcellular co-localization of MSI1 (Red) and AGO2 (Green). 733 (F) Flag-control and Flag-C-term transfected cells were subjected to an modified-RIP 734 assay using anti-MSI1 or anti-AGO2 antibodies, followed by RNA fragmentation and qRT-735 PCR analysis to determine the fold change enrichment of the coding sequence (CDS) 736 and 3´ UTR of the NF2, TP53, CCND1 and HELLS mRNAs. Quantification of the fold 737

37

changes of binding signals was performed by normalizing IP signals to IgG-precipitated 738 controls. The peaks indicated MSI1 or AGO2 palindromic-binding sequence. Flag-C-term 739 blocked the binding of AGO2 but not MSI1 to target sequence in mRNAs. (G) Flag-control 740 and Flag-C-term transfected cells were subjected to normoxia or hypoxia for 24 hr. 741 Purified total RNA was subjected to RT-PCR using primers specific for NF2, TP53, 742 CDKN1A, CCND1, CDK4, and HELLS. The mRNA levels under hypoxia were normalized 743 with that under normoxia and presented as relative fold changes in the chart. (H) 05MG 744 cells transiently transfected with Flag control or Flag-tagged MSI1 C-term were subjected 745 to an MTT viability assay. The relative fold change of the number of viable cells in each 746 day was presented in the graph. (I) Flag-control and Flag-C-term transfected cells were 747 subjected to colony formation assay for 10 days and quantitated by ImageJ software. (J) 748 The percentage of apoptotic cells of Flag-control and Flag-C-term transfected cells was 749 determined by external Annexin-V under normoxiac and hypoxic conditions. 750 Figure 7. The C-terminal of MSI1 suppress xenograft tumor growth in GBM and 751 PDAC animal models. 752 (A) A schematic presentation showing the design of animal experiment with in vivo 753 delivery of Flag-C-term (10 µg using in vivo-jetPEI in vivo nucleic acid delivery reagent). 754 Xenograft tumor size was monitored from day 2 after injection of Flag-control or Flag-C-755 term. (B) Immunocompromised mice were subcutaneously transplanted with 05MG/Flag-756 MSI1 stable cells. Two days after tumor size reached 50 mm3, mice were intratumorally 757 injected with 10 µg of Flag-control or Flag-C-term for 3 rounds with 2-day intervals. Tumor 758 size was then monitored for 22 days. The expression of MIS1-C-term in the xenograft 759

38

tumor tissue was assessed by Western blot. (C) Immunocompromised mice were 760 subcutaneously transplanted with MIA-PaCa2 cells. Two days after tumor size reached 761 50 mm3, mice were intratumorally injected with Flag-control or Flag-C-term for 3 rounds 762 with 2 days interval. Tumor size was then monitored for 22 days. The expression of MIS1-763 C-term in the xenograft tumor tissue was assessed by Western blot. (D) Tumor tissue 764 were harvested and homogenized for Western blotting. Whole-tumor lysate were 765 subjected to co-immunoprecipitation assay for endogenous AGO2 and MSI1 protein-766 protein interaction. Data represent the mean ± S.D. of three independent experiments 767 performed in triplicate. Tumor tissues of (E) 05MG/GBM cells and (F) MIA-768 PaCa2/pancreatic cells from MSI1-C-term injected xenografts were immunostained with 769 anti-Flag antibodies to observer the expression of MSI1-C-term in the tumors. 770 Figure 8. Cytosolic MSI1 expression associates with GBM relapse and PDAC 771 recurrence in patients. 772 (A) MSI1 expression was examined by IHC in 18 paired primary and recurrent GBM 773 tissues. Three representative cases (Pt 1 to 3) were presented. Boxes highlighting MSI1 774 expression pattern. (B) qPCR analysis of NF2, TP53, p21, CCND1, CDK4, and HELLS 775 mRNA expression levels in microdissected tumor (T) and stroma (S) samples from the 18 776 paired primary and recurrent GBM specimens. All mRNA expression levels in T parts 777 were first normalized by that in respective S counterparts, and then the total 36 expression 778 levels (primary and recurrent) of each mRNA were rated as percentile from 0% (green) to 779 100% (red). A heat map shows the relative mRNA expression levels between paired 780 primary and recurrent GBM tissue. (C) qPCR analysis of NF2, TP53, p21, CCND1, CDK4, 781

39

and HELLS mRNA levels in a group of primary (N = 67) and recurrent (N = 32) GBM 782 tissues (*P < 0.01). P values were estimated by a log-rank test. (D) 61 recurrent PDAC 783 patient samples were collected and stained for MSI1 by IHC. 3 representative cases 784 showed positive stain of cytosolic MSI1. (E) Survival analysis of the cytosolic MSI1-785 positive (cytosol-positive; N = 37) and cytosolic MSI1-negative (cytosol-negative; N = 24) 786 recurrent PDAC patients indicates that cytosolic MSI1-positive patients have poorer 787 survival outcome than cytosolic MSI1-negative patients. (F) In the 37 cytosol-positive 788 PDAC cases, the expression level of cytosolic MSI1 were evaluated by IHC score. In the 789 20 cases with cytosolic MSI1 IHC score < 0.5, 13 cases survived over 10 months after 790 recurrence; while in the 17 cases with cytosolic MSI1 IHC score > 0.5, only 2 cases 791 survived over 10 months after recurrence. P = 0.001; Chi-square = 10.80. (G) Post-792 recurrent survival analysis of the two groups (IHC score > 0.5 and IHC score < 0.5) of 793 cytosol-positive PDAC patients. 794 795

N0

50

100

150Cytosol fractionNuclear fraction

Perc

enta

ge o

f cel

ls

* #

H

N+LMB

H+LMB

AN N+LMB H H+LMB

MSI

1D

API

Mer

ge

B

C

D

GBM

MSI1Hypoxia(h)

GAPDHLamin

Control LMBT C N T C N

024024024 024024024

Cytosol fractionNuclear fraction

0

50

100

150

GBM

Recurr

ent-G

BM

Rec

urre

nt-G

BM

P1 P2

P3 P4

Figure 1

*

N H N H N H N H

Flag

DAP

IM

erge

Flag MSI1-wt MSI1-NES-mut MSI1-NLS-mut

MSI1-wt263 268

L T A I P L

C-term1 58 61 107 113 361

K R S R R T K K I FMSI1-NES-mut

1RRM1 RRM2 C-term

263 268 361

A T A A P AMSI1-NLS-mut

1C-term

58 61 107 113 361

A A S A A T A A I F

(3x) Flag-control

E

Flag MSI1-wt MSI1-NES-mut

MSI1-NLS-mut

Flag

DAPI

Merg

e

0 5 10 15 20 25 30 350

200

400

600

800

1000

1200

Day

FlagMSI1-wtMSI1-NES-mutMSI1-NLS-mut

F G H

Euthanasia(Day 22)

Tumor 50mm³(Day 0)

0

200

400

600

800

1000

Time(day)

2 6 8 10 12 14 16 18 20 22

Flag

MSI1-wt NES-mut NLS-mut

Flagadd Cisplatin

add Cisplatin

J K L

**

**

Flag

GAPDH

Lamin A/C

T C N

Flag MSI1-wt

NES-mut

NLS-m

ut

T C N T C N T C N

0

50

100

150Cytosol fractionNuclear fraction

Flag

MSI1-wt

NES-mut

NLS-m

ut

I

Perc

enta

ge o

f cel

ls

Aver

age

Inte

nsity

CisplatinT.V. injection

Tum

or v

olum

e (m

m³)

Tum

or v

olum

e (m

m³)

FlagFlag add CisplatinMSI1-wt add CisplatinNES-mut add CisplatinNLS-mut Cisplatin

Pre-bleach Post-bleach

MSI

1-O

rang

eAG

O2-

GFP

Pre-bleach Post-bleachN H, 24hr

Pre-bleach Post-bleachH, 16hr

* *

*

A

B

D

C

O

MSI1GAPDH

LaminA/C

Cytosol

Hypoxia(h)

Nuclear

Input

IgG IP Input

IgG IP0

AGO224 0 24 0 24 0 24 0 24 0 24

--+

++

+--

--+

++His-AGO2

Flag-MSI1His-AGO2Flag-MSI1

Input

IP:AGO2

+--

Figure 2

Flag-control

150

100

50

0Num

ber o

f col

onie

s

FlagMSI1FlagMSI1/shAGO2

* #

L

J K

MSI1

CCND1mRNA MSI1 AGO2 Merge

Norm

oxia

Merge TP53 mRNA MSI1 AGO2 Merge Merge

Hypo

xiaHy

poxia

+LM

B

Norm

oxia

Hypo

xiaHy

poxia

+LM

B

MParental cells

Normoxia Hypoxia

Anti-MSI1 or Anti-AGO2 antibodybind RNA/protein complex

Eluted complex,RNA isolated

NGS analysis

Cell cycle processAnti-apoptosis

Hexose catabolic process

Regulation of apoptosis

Generation of precusor metabolicCellular carbohydrate catabolic process

Microtubule cytosleleton organizationCell redox homostasis

Regulation of organelle organizationIntracellular transport

Glucose catabolic processCell cycle

Mitotic cell cycle

0.000 0.002 0.004 0.006 0.008 0.010(Benjamini)

362 263336

AGO2 target mRNA: 698 MSI1 target mRNA: 599

AGO2/MSI1 target mRNA: 336

N

AGO2

MSI1Actin

HIF-1α

N N HHsiC

on.

siAGO2

AGO2

FlagActin

HIF-1α

N N HHFlag Flag

MSI1

AGO2

MSI1Actin

HIF-1α

N N HHsiC

on.

siMSI1E F G

150

100

50

0Num

ber o

f col

onie

s

*

siCon.siMSi1siAGO2siMSI1/siAGO2

* *

6543210

54321

Cel

l via

bilit

y (fo

ld)

Time(day)

Flag-controlFlag-MSI1

*Flag-MSI1/shAGO2

#

H I

50

20

10

0

30

40

siCon.siMSi1siAGO2

Prec

enta

ge o

f pos

itive

ce

lls (A

nnxi

n V)

Normoxia Hypoxia

**

50

40

30

20

10

0FRET

effi

cien

cy (%

)

H, 16h

r

H, 24h

rN

B C E

0.0

0.5

1.0

1.5

Expr

essi

on le

vel o

f mR

NA

(Hyp

oxia

/Nor

mox

ia) siMSI1

NF2TP53p2

1

CCND1CDK4

HELLS

* * *

0.0

0.5

1.0

1.5

Expr

essi

on le

vel o

f mR

NA

(Hyp

oxia

/Nor

mox

ia) siAGO2

NF2TP53p2

1

CCND1CDK4

HELLS

* * *

siCon.

NF2TP53p2

1

CCND1CDK4

HELLS

0.0

0.5

1.0

1.5

Expr

essi

on le

vel o

f mR

NA

(Hyp

oxia

/Nor

mox

ia)

* **

* * *

0.0

0.5

1.0

1.5

MSI1-NES-mut

Expr

essi

on le

vel o

f m

RN

A (H

/N)

NF2TP53 p2

1

CCND1CDK4

HELLS

NF2TP53 p2

1

CCND1CDK4

HELLS

0.0

0.5

1.0

1.5

MSI1-wt

*

Expr

essi

on le

vel o

f m

RN

A (H

/N)

* *

0.0

0.5

1.0

1.5

MSI1-NLS-mutEx

pres

sion

leve

l of

mR

NA

(H/N

)

NF2TP53 p2

1

CCND1CDK4

HELLS

* * *

Time half-life, TP53 mRNA

05

10152025

Time(

hrs)

siCon. siMSI1 siAGO2

* **

N H H HN N0

5

10

15

20

Tim

e(hr

s)

Time half-life, CDK4 mRNA

siCon. siMSI1 siAGO2

** *

N H H HN N

Time half-life, p21 mRNA

05

10152025

Time(

hrs)

siCon. siMSI1 siAGO2

* **

N H H HN N

Tim

e(hr

s)

Time half-life, HELLS mRNA

0

5

10

15

20

siCon. siMSI1 siAGO2

* * *

N H H HN N

Time half-life, NF2 mRNA

05

10152025

Time(

hrs)

siCon. siMSI1 siAGO2

* **

N H H HN N

Group 1

Tim

e(hr

s)

05

10152025

Time half-life, CCND1 mRNA

siCon. siMSI1 siAGO2

**

N H H HN N

Group 2

siAG

O2

siM

SI1

log2 Fold change

0.8

0.4

0

-0.8

-0.4

siC

ON

.

A

D

FlagMSI1

NF2TP53p2

1

CCND1CDK4

HELLS

0.0

0.5

1.0

1.5

Expr

essi

on le

vel o

f mR

NA

(Hyp

oxia

/Nor

mox

ia)

* **

FlagMSI1/KD AGO2

NF2TP53p2

1

CCND1CDK4

HELLS

0.0

0.5

1.0

1.5

Expr

essi

on le

vel o

f mR

NA

(Hyp

oxia

/Nor

mox

ia)

* **

F G

log2 Fold change

siAG

O2

siM

SI1

siC

ON

.

Figure 3

NF2

p53

p21

cyclinD1

CDK4

HELLS

Flag

Actin

N H R1WT NES-mut NLS-mut

R6N H R1 R6 N H R1 R6NF2

p53

p21

cyclinD1

CDK4

HELLS

AGO2

MSI1

N H R1siCon. siMSI1 siAGO2

R6N H R1 R6 N H R1 R6

Actin

*

B

CNF2p53p21cyclinD1CDK4ki67 FlagMSI1

Flag

MSI

1-w

tN

ES-m

utN

LS-m

utFl

ag

Cis

plat

in

NF2 TP53 P21 CCND1 CDK4 HELLS

Flag

MSI1-wt add CisplatinMSI1-NES-mut add CisplatinMSI1-NLS-mut add Cispaltin

Flag add Cisplatin

Relat

ive ex

pres

sion l

evel 3.0

2.0

1.0

0.0*

* *

***

*

*

*

** *

*** *

* *** ***** ***

A

Orthotropic(Brain)

Actin

AGO2

Flag

Input IgG IP:Flag

NLS

-mut

MSI

1-w

t

NES

-mut

NLS

-mut

MSI

1-w

t

NES

-mut

NLS

-mut

MSI

1-w

t

NES

-mut

D

NLS

-mut

NES

-mut

MSI

1-w

tFl

agFl

ag

Cis

plat

in

E

Figure 4

20 daysEuthanasia

40days

Cisplatin

CisplartinT.V. injection

Orthotropic (Brain)

0 24 0 24 0 24Inp

utIgG IP

A B

*

*

***

* **

**

**

NF2, IP: MSI1

C-1 C-2 C-3 C-4 C-5 U-1 U-2 U-3 U-4 U-50

20

40

60

Fold

enr

ichm

ent

(rela

tive

to Ig

G IP

)

HypoxiaNormoxia

TP53, IP: MSI1

C-1 C-2 C-3 C-4 C-5 U-1 U-2 U-3 U-40

20

40

60

U-5

CCND1, IP: MSI1

C-1 C-2 C-3 C-4 U-1 U-2 U-3 U-40

20

40

60 HELLS, IP: MSI1

C-1 C-2 C-3 C-4 C-5 U-1 U-2 U-3 U-4 U-50

20

40

60

TP53, IP: AGO2

C-1 C-2 C-3 C-4 C-5 U-1 U-2 U-3 U-40

20

40

60

U-5

CCND1, IP: AGO2

C-1 C-2 C-3 C-4 U-1 U-2 U-30

20

40

60 HELLS, IP: AGO2

C-1 C-2 C-3 C-4 C-5 U-1 U-2 U-3 U-4 U-50

20

40

60NF2, IP: AGO2

C-1 C-2 C-3 C-4 C-5 U-1 U-2 U-3 U-4 U-50

20

40

60

Fold

enr

ichm

ent

(rela

tive

to Ig

G IP

)

HypoxiaNormoxia

HypoxiaNormoxia

HypoxiaNormoxia

HypoxiaNormoxia

HypoxiaNormoxia

HypoxiaNormoxia

HypoxiaNormoxia

3’

421 1490 2697

5’

TP53 mRNA

3’5’

210 1097 4261

CCND1 mRNA

3’

443 1020 5820CDS(C) 3’UTR(U)

5’

NF2 mRNA

3’

156 540 2672

5’

HELLS mRNA

1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 51 2 3 4 1 2 3 4

CDS(C) 3’UTR(U) CDS(C) 3’UTR(U) CDS(C) 3’UTR(U)

AGO2MSI1AAAAAm7G

Cytosol

MSI1AAAAAm7G

AGO2MSI1AAAAAm7G CDS

3’UTR:degradation

CDS:Stability

MSI1AAAAAm7G

Nuclear

Cytosol

MSI1AAAAAm7G

MSI1AAAAAm7G

Nuclear

AGO2

AGO2

Group 1 Group 2

**

*

*

Hypoxia

Input

IgG IP:MSI1

IP:AGO2

siAGO2siMSI1

MSI1AGO2

Normoxia

Actin

--+

-- --+-- -- +

-- --+-- -- +

-- --+-- -- +

-- +-- --

0

5

10

15

20

25 NormoxiaHypoxia

siC siM siA siC siM siA siC siM siA siC siM siA

0

5

10

15

20

25 NormoxiaHypoxia

** **

*

*

*

***

*

*

* * * *

05

10152025 Normoxia

Hypoxia

05

10152025 Normoxia

Hypoxia

* ** *

C N C N C NNF2 TP53 CCND1

C NCDK4

C N C N C N C N

* * * *

Input

IgG IP:MSI1

IP:AGO2

--+

-- --+-- -- +

-- --+-- -- +

-- --+-- -- +

-- +-- --

C

Figure 5

NF2 TP53 CCND1 CDK4

NF2 TP53 CCND1 HELLS

0 24 0 24 0 24Inp

utIgG IP

MSI1-wt

Hypoxia(hr)AGO2

FlagMSI1GAPDH

LaminA/C

Cytosol NuclearMSI1-wt

Hypoxia(hr)AGO2

FlagMSI1GAPDH

LaminA/C

0 24 0 24 0 24Inp

utIgG IP

0 24 0 24 0 24Inp

utIgG IP

MSI1-NES-mutCytosol Nuclear

MSI1-NES-mut

Hypoxia(hr)AGO2

FlagMSI1GAPDH

LaminA/C

0 24 0 24 0 24Inp

utIgG IP

0 24 0 24 0 24Inp

utIgG IP

MSI1-NLS-mutCytosol Nuclear

Fold

enr

ichm

ent

Fold

enr

ichm

ent

05

10152025

C N C N C NC N

NormoxiaHypoxia

NF2 TP53 CCND1 CDK4

MSI1-NES-mut

Fold

enr

ichm

ent

Parental cells, IP: MSI1

Fold

enr

ichm

ent

(rela

tive

to Ig

G IP

)

Parental cells, IP: AGO2

Fold

enr

ichm

ent

(rela

tive

to Ig

G IP

)

siC siM siA siC siM siA siC siM siA siC siM siANF2 TP53 CCND1 HELLS

D

F

G

D

H

6543210

54321

Cel

l via

bilit

y (fo

ld)

Time(day)

Flag-controlFlag-C-term *

Flag-C-termFlag-control

30

20

10

0% o

f pos

itive

cel

l (A

nnex

in V

)

HypoxiaNormoxia

*

Flag-C-termFlag-control

150

100

50

0Num

ber o

f col

onie

s *

E

0.0

0.5

1.0

1.5

Expr

essi

on le

vel o

f mR

NA

(Hyp

oxia

/Nor

mox

ia)

NF2TP53 p2

1

CCND1CDK4

HELLS

Flag-C-termFlag-control

NF2TP53p2

1CCND1

CDK4

HELLS

0.0

0.5

1.0

1.5

Expr

essi

on le

vel o

f mR

NA

(Hyp

oxia

/Nor

mox

ia)

* ** * **

Figure 6A

Flag-C-term C-term361

FlagMSI1 C-term1 361

RRM1 RRM2

Flag-controlFlag-C-term

AGO2MSI1Flag

N H N H N H N H N H N H-- -- + + -- -- + + -- -- + +

----++----++----++

Input

IgG IP:MSI1

B

C

AGO2 MSI1 MergeTransfected Flag-control

AGO2 MSI1 MergeTransfected Flag-C-term

Actin

Fold

enr

ichm

ent

(rela

tive

to Ig

G IP

) Fo

ld e

nric

hmen

t (re

lativ

e to

IgG

IP)

NF2, IP: MSI1

C-1 C-2 C-3 C-4 C-5 U-1 U-2 U-3 U-4 U-50

20

40

60

Flag-C-termFlag-control

**

3’443 1020 5820CDS(C) 3’UTR(U)

5’

NF2 mRNA

1 2 3 4 5 1 2 3 4 5 3’

421 1490 2697

5’

TP53 mRNA

1 2 3 4 5 1 2 3 4 5

CDS(C) 3’UTR(U)

TP53, IP: MSI1

C-1 C-2 C-3 C-4 C-5 U-1 U-2 U-3 U-40

20

40

60

U-5

**Flag-C-termFlag-control

NF2, IP: AGO2

C-1 C-2 C-3 C-4 C-5 U-1 U-2 U-3 U-4 U-50

20

40

60

*Flag-C-termFlag-control

TP53, IP: AGO2

C-1 C-2 C-3 C-4 C-5 U-1 U-2 U-3 U-40

20

40

60

U-5

*Flag-C-termFlag-control

Group 1

3’5’

210 1097 4261

CCND1 mRNA

1 2 3 4 1 2 3 4

CDS(C) 3’UTR(U)

3’

156 540 2672

5’

HELLS mRNA

1 2 3 4 5 1 2 3 4 5

CDS(C) 3’UTR(U)

CCND1, IP: MSI1

C-1 C-2 C-3 C-4 U-1 U-2 U-3 U-40

20

40

60

**

Flag-C-termFlag-control

HELLS, IP: MSI1

C-1 C-2 C-3 C-4 C-5 U-1 U-2 U-3 U-4 U-50

20

40

60

**Flag-C-termFlag-control

CCND1, IP: AGO2

C-1 C-2 C-3 C-4 U-1 U-2 U-30

20

40

60

*

U-4

Flag-C-termFlag-control

HELLS, IP: AGO2

C-1 C-2 C-3 C-4 C-5 U-1 U-2 U-3 U-4 U-50

20

40

60

*

Flag-C-termFlag-control

Group 2

Fold

enr

ichm

ent

(rela

tive

to Ig

G IP

) Fo

ld e

nric

hmen

t (re

lativ

e to

IgG

IP)

MSI1-C-term C-term

MSI1-FL C-termRRM1 RRM2

RRM1 RRM2MSI1-N-term

SUMO

SUMO

SUMO

His

His

HisM

SI1-

C-te

rm

MSI

1-FL

MSI

1-N

-term

AGO

2

MSI

1-C

-term

MSI

1-FL

MSI

1-N

-term

Input IP:AGO2IP:AGO2IB:His-ab

AGO2

MSI1-FL

MSI1-N-term

MSI1-C-term

190130

10070

55

40

35

25

I

J

Flag-control

B

A

CC

on.

C-te

rm.

Flag DAPI Merge

GBM:05MG

Flag DAPI Merge

Pancreatic:MIA-PaCa2

Con

.C

-term

.

Time after injection (day)

Flag-ControlFlag-C-term

0

200

400

600

800

Tum

or v

olum

e(m

m³)

GBM:05MG-MSI1-OE

2 4 6 8 10 12 14 16 18 20 22

Con. C-term

**

Con.C-te

rm

Flag

Actin

Flag-ControlFlag-C-term

Tum

or v

olum

e(m

m³)

Con. C-term

**

100

200

300

400

500

0

Pancreatic:MIA-PaCa2

Con.C-te

rm

Flag

Actin

MSI1

Euthanasia(Day 22)(Day 0) Day2 Day6 Day10

Implantation Intratumor transfection

Tumor 50mm³

D

E

F

-- + -- + -- +--+ -- + -- +

-- + -- + -- +--+ -- + -- +

Input

IgG IP: MSI1

Input

IgG IP: MSI1

Flag-contorlFlag-C-term

AGO2

MSI1

Actin

Null mice No1 Null mice No2

Figure 7

Right: Flag-controlLeft: Flag-C-term

Time after injection (day)2 4 6 8 10 12 14 16 18 20 22

A B

G

Percentile0 100

Pancreatic: Cytosol positive

Patient 1 Patient 2 Patient 3

Recurrent GBM (2⁰ operation)

Patient 1 Patient 2 Patient 3

GBM (1⁰ operation)

Patient 1 Patient 2 Patient 3

Cytosol negativeCytosol positive

After recurrent survival time (months)

Perc

ent s

urvi

val (

%)

0 10 20 300

50

100

150

D E

F

Survival > 10 month

Survival < 10 month

MSI1IHC>0.5

MSI1IHC<0.5

2

15

13

7Total 1720

P value = 0.001 ,Chi-square value 10.80

Cytosolic positive

Perc

ent s

urvi

val (

%)

0

50

100

150 MSI1, IHC > 0.5MSI1, IHC < 0.5

After recurrent survival time (months)0 10 20 25155

C

*CCND1

020406080

Rel

ativ

e ex

pres

sion

(2-D

ct)

GBM rGBM

*

CDK4

020406080

Rel

ativ

e ex

pres

sion

(2-D

ct)

GBM rGBM

NF2

*

020406080

Rel

ativ

e ex

pres

sion

(2-D

ct)

GBM rGBM

*TP53

020406080

Rel

ativ

e ex

pres

sion

(2-D

ct)

GBM rGBM

*P21

020406080

Rel

ativ

e ex

pres

sion

(2-D

ct)

GBM rGBM

020406080 *

HELLS

Rel

ativ

e ex

pres

sion

(2-D

ct)

GBM rGBM

NF2

TP53

P21

CCND1

CDK4

HELLS

Patient NO.

GBM

1 2 3 4 7 85 6 91011121314 17181516 1 2 3 4 7 85 6 91011121314 17181516

Recurrent GBM

Group 1

Group 2

Figure 8