Targeting Integrated Stress Response by ISRIB combined with imatinib attenuates STAT5 signaling and eradicates therapy-resistant Chronic Myeloid Leukemia cells

Integrated Stress Response (ISR) facilitates cellular adaptation to variable environmental conditions by reprogramming cellular response. Activation of ISR was reported in neurological disorders and solid tumours, but its function in hematological malignancies remains largely unknown. Previously we showed that ISR is activated in chronic myeloid leukemia (CML) CD34+ cells, and its activity correlates with disease progression and imatinib resistance. Here we demonstrate that inhibition of ISR by small molecule ISRIB, but not by PERK inhibitor GSK2656157, restores sensitivity to imatinib and eliminates CM Blast Crisis (BC) D34+ resistant cells. We found that in Patient Derived Xenograft (PDX) mouse model bearing CD34+ imatinib/dasatinib-resistant CML blasts with PTPN11 gain-of-function mutation, combination of imatinib and ISRIB decreases leukemia engraftment. Furthermore, genes related to SGK3, RAS/RAF/MAPK, JAK2 and IFNγ pathways were downregulated upon combined treatment. Remarkably, we confirmed that ISRIB and imatinib combination decreases STAT5 phosphorylation and inhibits expression of STAT5-target genes responsible for proliferation, viability and stress response. Thus, our data point to a substantial effect of imatinib and ISRIB combination, that results in transcriptomic deregulation and eradication of imatinib-resistant cells. Our findings suggest such drug combination might improve therapeutic outcome of TKI-resistant leukemia patients exhibiting constitutive STAT5 activation.


Introduction 68
Chronic myeloid leukemia (CML) which is driven by oncogenic BCR-ABL1 tyrosine kinase, is an example of a 69 disease that is successfully treated with molecular targeted therapy. Introduction of imatinib significantly improved 70 CML treatment, patients' life expectancy and overall survival 1,2 . However, although imatinib shows remarkable 71 clinical efficacy in the chronic phase, the effects in advanced phases are short-lived, complete remissions are 72 rare, and relapse occurs often [3][4][5] . Many patients show primary or secondary resistance to imatinib or second 73 generation tyrosine kinase inhibitors (TKIs), such as dasatinib, nilotinib or bosutinib. The resistance originates in 74 majority from cellular intrinsic mechanisms. Apart from BCR-ABL1 point mutations (e.g. T315I) which affect drug 75 binding affinity 6,7 , the BCR-ABL1 gene amplification or clonal evolution may lead to relapse driven by both BCR-76 ABL1-dependent and -independent mechanisms. 77 The most recognized pathways responsible for resistance are mediated by activation of JAK2/STAT5, 78 RAS/RAF/MAPK or PI3K/Akt/mTOR 3,8,9 . They activate proliferation, anti-apoptotic response and survival, 79 cytokine and growth factors signaling, altogether strongly promoting resistance to treatment and disease relapse. 80 Therefore, targeting these pathways is one of the current strategies for eradication of resistant cells [10][11][12] . 81 Previously, we identified that the PERK-eIF2 pathway related to Integrated Stress Response (ISR) is activated in 82 CD34+ CML-BP cells 13 . ISR is a highly conserved signaling responsible for cell adaptation and survival upon 83 stress conditions [14][15][16][17] . This is achieved by phosphorylation of the eukaryotic translation initiation factor eIF2 84 remodelling of translation 18 and transcription of stress response effector genes, including CHOP and GADD34, 85 which are ISR markers. 86 Under physiological conditions, the ISR is one of the mechanisms sustaining homeostatic balance in a healthy 87 cell. Cancer cells can utilize ISR to survive and develop drug resistance. Previous reports demonstrated that ISR 88 is active in solid tumors in which it correlates with hypoxia and metastasis 19 . However, ISR has not been deeply 89 studied in leukemia. Since recognized, ISR is proposed as a therapeutic target in cancer [20][21][22] . Nevertheless, no 90 efficient and specific strategy has been proposed still, especially for hematological malignancies. 91 We report here that inhibition of ISR signaling by small molecule ISRIB combined with imatinib has potential to 92 eradicate imatinib-resistant CML-BP cells. We show that such treatment specifically changes gene expression 93 profile and inhibits oncogenic STAT5 signaling. Therefore the combination of ISRIB and imatinib was identified as 94 SQuIRE pipeline for alignment of reads, transcript assembly and quantification, and differential gene expression 165 analysis, respectively. Differentially expressed genes with false discovery rate (FDR) < 0.05 were reported here. 166 Principal component analysis of all samples (11 replicates in total from 4 conditions) based on gene expression 167 6 data (transcripts per kilobase million or TPM) was performed with 29 . The Clust tool 30 was used for co-expressed 168 gene clusters identification across all samples. The default normalization procedure of Clust for RNA-seq TPM 169 data (quantile normalization followed by log2-transformation and Z-score normalization, code "101 3 4") was 170 applied. gProfiler 31 was utilized for the simultaneous functional enrichment analysis of the genes from all clusters 171 in multi-query mode. The RNA-Seq data from this publication have been deposited to the NCBI GEO repository 172 (https://www.ncbi.nlm.nih.gov/geo) and can be accessed with the dataset identifier GSE171853. 173 174

Statistical analysis 175
Data were analysed using GraphPad Prism (GraphPad Software, La Jolla, CA, USA) Single comparisons were 176 tested using unpaired Student's t-tests for normal distributed samples or Mann-Whitney-U tests when normal 177 distribution was not given. One-way or two-way ANOVA was applied for multiple comparison analysis, with 178 Bonferroni's multiple comparison post-test. For RT-qPCR unpaired Student's t-test with Welch's correction was 179 applied. P values < 0.05 were estimated as significant (*p<0.05; **p <0.005; ***p<0.0005). Data are presented as 180 mean ± SD. 181 182

GENETIC ISR INHIBITION SENSITIZES CML CELLS TO IMATINIB IN VITRO 184
To study the impact of Integrated Stress Response globally, ISR was inhibited by targeting the main regulatory 185 hub -eIF2. This was achieved by expression of non-phosphorylable (S51A) eIF2 form (visible as additional 186 band on western blot), followed by overexpression of shRNA against eIF23'UTR (S51A shUTR) to inhibit 187 expression of endogenous wt eIF2, leading altogether to complete lack of eIF2 phosphorylation (Fig. 1A, 188 detailed procedure of generation of genetically modified cells is provided in Supplementary Information). Both 189 generated cell lines had unaffected levels of PERK, an UPR kinase acting upstream of eIF2and expressed 190 GFP necessary for FACS sorting (Fig. 1A). The functional influence on ISR confirmed by detection of mRNAs 191 encoding ISR markers CHOP and GADD34 showed that inhibition of the eIF2 phosphorylation attenuates 192 dynamics of the ISR activation (Fig. S1A). In addition, inhibition of the eIF2 phosphorylation itself decreased cell 193 viability (Fig. 1B), and associated with increased basal GADD34 and CHOP mRNA levels indicating stress-194 induced cell death (Fig. S1B). This indicated that the lack of ISR pathway itself is cytotoxic for CML cells. 195 Furthermore, imatinib-induced apoptosis was higher in S51A and further increased in S51A shUTR cells, 196 compared to wt (Fig. 1C). This implies that indeed K562 cells utilize the eIF2 phosphorylation-dependent 197 mechanism and that ISR inhibition sensitizes CML cells to imatinib.

ISRIB, BUT NOT GSK157, SENSITIZES CML CELLS TO IMATINIB IN VIVO 199
Even if the genetic approaches are useful, the pharmacological inhibition still gives the highest possibility for the 200 clinical applications targeting the signaling pathways. Thus, we tested two ISR inhibitors: GSK2656157 (GSK157) 201 and ISRIB ( Fig. 2A). GSK157 is an ATP-competitive inhibitor of PERK kinase, which stops the PERK-dependent 202 ISR activation. Small molecule ISRIB blocks the eIF2-P-dependent downstream signaling and inhibits the 203 executive part of ISR, without the cytotoxic effects 32-35 . Both drugs have not been tested in leukemia, including 204 CML. Pre-conditioning of K562 CML cells with either GSK157 or ISRIB, followed by ISR induction by thapsigargin, 205 revealed that both ISR inhibitors significantly reduced expression of CHOP and GADD34 mRNAs in leukemia 206 cells (Fig. 2B, C). 207 The results obtained in vitro ( A short and aggressive 7-day regimen was applied to test the beneficial effects, followed by treatment with 231 imatinib or ISRIB alone or with drug combination (experimental scheme - Fig. 4A). All variants showed noticeable 232 but not significant decrease in the spleen weight (Fig. 4B). To estimate the short-term engraftment, the 233 percentage of human CD45+ (hCD45+) was detected within the bone marrow or spleen populations (Fig. 4C, 4D, 234 4E; Fig. S2). Combination of imatinib and ISRIB significantly decreased percentage of hCD45+ cells in the bone 235 marrow, showing a 2-to 3-fold lower level compared to the treatment with imatinib or ISRIB alone. In addition, the 236 combined treatment treatment decreased the engraftment into the spleen (which is considered as a secondary 237 niche), compared to imatinib alone (Fig. 4E). These results showed that the combined treatment eradicates 238 resistant blasts and decreases leukemia engraftment, therefore confirming the synergistic effect of imatinib and 239 ISRIB. 240

PRIMARY TKI-RESISTANT BLASTS 243
To investigate the molecular effects of the double treatment, RNA-seq was performed on FACS-sorted hCD45+ 244 CML cells isolated from the PDX bone marrow. Principal component analysis (PCA) indicated that cells treated 245 with imatinib and ISRIB are transcriptionally distinct (Fig. 5A). This was confirmed by hierarchical clustering of 246 significantly changed genes between pairs of tested conditions (treatment vs control), and supported by the 247 Pearson correlation values, which showed higher correlation between sole ISRIB and sole imatinib treatment 248 compared to control (r = 0.69), than between each of the single treatments and the combined imatinib+ISRIB 249 treatment compared to control (r = 0.32 and 0.37, respectively; Fig. 5B). The SGK3 and SNURF/SNRPN genes 250 regulating alternative RNA processing were identified as significantly downregulated upon the double treatment. 251 Upregulated genes in majority encoded proteins regulating transcription and RNA processing. 252 To identify genes responsible for the increased sensitivity, the gene expression profiles for imatinib versus 253 imatinib + ISRIB were compared. In addition to the previously described (Fig. 5B), genes encoding proteins from 254 the small GTP-binding RAS superfamily (RGPD5 and RGPD8) were significantly downregulated (Fig. 5C, for all 255 treatment combinations see Fig. S3A). 256 Since genes that are co-expressed are often co-regulated, clusters of co-expressed genes (C0-C12) across all 257 variants of treatment were identified (all genes included, regardless of their statistical significance of change in 258 expression) (Fig. 5D). Clusters with the highest number of genes represented the groups in which drug 259 combination led to either sharp decrease (C0, C1) or increase (C5, C6) of gene expression (Fig. 5D, 5E). Those 260 four clusters represented about 72% of all detected genes (Fig. 5E). This clearly indicated that the gene 261 expression pattern for ISRIB + imatinib combination is specific and different from the other treatment conditions. 262 9

SIGNALING 265
To predict the cellular mechanisms altered by combined treatment, all 13 defined gene clusters underwent the 266 functional enrichment analysis. The C0 and C1 clusters which included genes downregulated upon combined 267 treatment, were significantly enriched in terms related to RAS/RAF/BRAF/MAPK signalling (Fig. 6, marked in red,  268 and Fig S5; for all clusters see Fig. S3). Specifically, for Ras and MAPK signaling (detailed member genes in While SGK3 gene encoding serine/threonine-protein kinase SGK3 was significantly downregulated after the 277 combined treatment (Fig. 5B), expression of the SGK3 interaction partners, selected based on the interaction 278 partner datasource: BioGRID, IntAct (EMBL-EBI) and APID databases (see Supplementary Information), showed 279 rather moderate inhibition upon combined treatment (Fig. S6). Among the downregulated genes, we found 280 GSK3what may suggest its regulatory connection with SGK3 and specific downregulation upon combined 281 treatment. 282 Altogether, these results showed that genes related to oncogenic pro-leukemic signaling were downregulated 283 upon combination of imatinib and ISRIB, presumably enhancing targeting of leukemia cells by imatinib. 284

COMBINATION OF ISRIB AND IMATINIB INHIBITS STAT5 SIGNALING IN CML CELLS 286
Transcriptomic data indicated that the combined treatment can downregulate oncogenic RAS/RAF/MAPK, JAK2, 287 SGK3 and IFN signalling. In addition, genes that are mediators of JAK2/STAT5 signalling were attenuated (Fig.  288   6, Fig. S4, S5). This indicated that the combined treatment might inhibit the STAT5 pathway. 289 To obtain direct evidence that treatment with imatinib + ISRIB shows synergistic effect, STAT5 phosphorylation 290 was assessed in K562 and LAMA84 cell lines, which were shown to activate the above signaling pathways 13,38 . 291 To better visualize the effects, strong ISR response in vitro was activated by thapsigargin. In both cell lines, 292 combination of ISRIB with imatinib decreased STAT5 phosphorylation detected by western blot (Fig. 7A, 7B), and 293 confirmed by phospho-flow cytometry (Fig. 7C). On the other hand, ISRIB alone did not change, whereas 294 imatinib alone only partially decreased phosphorylation of STAT5, compared to double treatment, with effectivity 295 lower in K562 cells, which were more resistant. Conversely, the significant additive effect (estimated by the 296 phosphorylation levels) of the combined therapy combined to imatinib alone was not observed for other pro-297 leukemic related regulators such as: AKT, mTOR, S6K, SGK3, GSK3 or ERK (Fig. S7). So, the genetic data 298 indicating downregulation of the SGK3 -GSK3 link were not confirmed in vitro. Interestingly, inhibition of AKT 299 and ERK phosphorylation by imatinib ( Fig. S7A and S7F, respectively), associated with decreased BCR-ABL1 300 activity (Fig. S8A), but not BCR-ABL1 protein level (Fig. S8B), indicated that either pAKT or pERK are not 301 involved in acquiring the BCR-ABL1-independent resistance. 302 The results presented in Fig. 7A-7D imply that the combined treatment attenuates the STAT5-dependent 303 signaling. To test this, the fold change analysis of the STAT5 target genes expression was performed within the 304 success of imatinib in the CML-CP treatment, the disease is still not fully curable and eradication of all leukemic 317 cells is not efficient. Imatinib intolerance or primary resistance occurs, as well as many patients develop 318 secondary resistance due to activation of signaling pathways, including JAK/STAT5, GSK3 or RAS/MEK/ERK 319 3,8,9 . Importantly, such activation might occur in a BCR-ABL1-independent manner, thus upon imatinib treatment 320 of even BCR-ABL1 non-mutated cells, those oncogenic pathways still remain active. Therefore, one of the current 321 strategies to eradicate leukemic blasts, is to target BCR-ABL1 together with oncogenic signaling pathways, to 322 resensitize cells to TKIs 5,39-41 . 323 Here we provide evidence that inhibition of Integrated Stress Response by ISRIB combined together with imatinib 324 might significantly break the resistance by targeting both, the stress response adaptative signaling as well as the 325 STAT5-dependent intrinsic signaling. This can result in effective elimination of imatinib-refractory cells in CML. 326 Unexpectedly, only ISRIB but not another ISR inhibitor -GSK157 belonging to the PERK inhibitors family, was 327 effective in vivo. This is consistent with recent studies of amyotrophic lateral sclerosis which showed similar data 328 indicating that ISRIB but not GSK157 inhibitor, was more effective and improved neuronal survival 42 . Such effect 329 can be a result of an eIF2α phosphorylation-independent effects 43 , moderate specificity of GSK157, as its affinity 330 to RIPK1 was shown to be significantly higher than to PERK kinase 44  Mechanistically, we have discovered that the combined treatment inhibits STAT5 phosphorylation and decreases 343 expression of STAT5 target genes, that regulate proliferation, apoptosis and stress response. Targeting STAT5, 344 which is an oncogenic signaling in imatinib resistant forms of CML 3,9,56 , effectively overcomes resistance and 345 eradicates leukemic cells [57][58][59] . The experimental therapy proposed by us, not only inhibits ISR but also attenuates 346 the STAT5-dependent signaling in CML. It is to note, that the overactivated STAT5 has also been detected in 347 other hematopoietic malignancies, such as non-CML chronic myeloproliferative disorders correlating with JAK2 348 V617F mutation 60 or Flt3-ITD positive AML 61 . Therefore, it is worth considering that the proposed strategy might 349 be effective also in other blood disorders. 350 In striking contrast, even if downregulation of related genes was observed in the transcriptomic analysis, the 351 mTOR, SGK3, GSK3, AKT and ERK activity was not specifically targeted by the double treatment in vitro. 352 Notably, even though the regulatory ISR-SGK3 link was shown in glioma 62 , and our transcriptomic data indicated 353 SGK3 downregulation by the combined treatment, this was not confirmed in the model studies in vitro. On the 354 other hand, pAKT and pERK, together with BCR-ABL1 activity were inhibited already by imatinib alone, and not 355 further downregulated by drug combination. Therefore, those pathways were probably not responsible for the 356 resistant phenotype. Nevertheless, in other leukemias in which the resistance developed due to AKT or ERK 357 overactivation, such effect might help to eradicate the resistant blasts. 358 Interestingly, differential expression of genes responsible for the immune modulation (visible even in the xenograft 359 model, which excludes involvement of T and B lymphocytes, but still encompasses functional myeloid cells) 360 suggests possible involvement of the immune system remodelling in the therapeutic outcome. This data support 361 the idea of targeting the innate immune system or immune checkpoints in myeloid malignancies, including CML 12 63-65 . Thus, even though experiments were performed in immunodeficient (lacking adaptive, lymphocyte-mediated 363 response) mice, signaling and functional effects related to the innate immune responses (mediated by e.g. 364 macrophages) were possibly functional leading to the observed changes. Although interesting, this has to be 365 verified in subsequent studies using the syngenic mouse model. 366 In conclusion, we discovered a novel strategy to break the resistance and eradicate imatinib-refractory CML 367 blasts, which is based on therapeutic combination of ISR inhibitor ISRIB together with imatinib. We postulate that 368 such strategy can improve therapeutic outcomes in CML patients showing TKI resistance related to overactivated 369 STAT5 and stress adaptation signaling. Possibly, a similar approach based on ISRIB combined with a typical 370 chemotherapy may also be applied to other hematological malignancies with constitutively activated STAT5 371 signaling and STAT5-dependent resistance.