TET2 lesions enhance the aggressiveness of CEBPA-mutant AML by rebalancing GATA2 expression

The myeloid transcription factor CEBPA is recurrently biallelically mutated (i.e., double mutated; CEBPADM) in acute myeloid leukemia (AML) with a combination of hypermorphic N-terminal mutations (CEBPANT), promoting expression of the leukemia-associated p30 isoform, and amorphic C-terminal mutations. The most frequently co-mutated genes in CEBPADM AML are GATA2 and TET2, however the molecular mechanisms underlying this co-mutational spectrum are incomplete. By combining transcriptomic and epigenomic analyses of CEBPA-TET2 co-mutated patients with models thereof, we identify GATA2 as a conserved target of the CEBPA-TET2 mutational axis, providing a rationale for the mutational spectra in CEBPADM AML. Elevated CEBPA levels, driven by CEBPANT, mediate recruitment of TET2 to the Gata2 distal hematopoietic enhancer thereby increasing Gata2 expression. Concurrent loss of TET2 in CEBPADM AML induces a competitive advantage by increasing Gata2 promoter methylation, thereby rebalancing GATA2 levels. Of clinical relevance, demethylating treatment of Cebpa-Tet2 co-mutated AML restores Gata2 levels and prolongs disease latency.


INTRODUCTION 1
Most patients with CEBPA DM AML also feature additional mutations in GATA2, TET2, WT1, NRAS, FLT3, or CSF3R 9 . 23 Several of these mutations are found together with CEBPA DM more frequently than expected by the individual frequency 24 of each mutation, while other combinations are statistically underrepresented. Recent studies have shed light on the 25 molecular mechanisms underlying mutational cooperativity for some of the co-mutated genes, i.e. GATA2 14 and CSF3R 15 , 26 while mechanistic insight is still lacking for other subgroups of CEBPA DM AML. Of particular importance are mutations 27 in the gene encoding the methylcytosine dioxygenase TET2 which, by converting 5-methylcytosine to 5-28 hydroxymethylcytosine, promotes DNA demethylation. TET2 mutations (TET2 MUT ) are frequent in CEBPA DM AML cases 29 and are associated with inferior prognosis 16,17 . Moreover, loss of Tet2 has been implicated in accelerating and/or 30 aggravating hematological malignancies in combination with several other recurrent gain-of-function and loss-of-function 31 mutations 18-20 , reflecting the importance of appropriately regulated DNA demethylation in normal hematopoiesis. 32 Importantly, while Tet loss alone only mildly affects hematopoiesis with myeloid skewing and increased competitiveness 33 of HSCs 18 , as well as increased propensity of leukemic blasts to switch to a more stem-like phenotype 21 , it does not induce 34 overt leukemia per se [22][23][24] . Despite being extensively studied, mechanistic insights of how TET2 loss-of-function 35 cooperates with other aberrations has been hampered by the fact that malignant cells have been compared to their normal, 36 wild-type counterparts in many studies. 37 In the present work, we sought to overcome this limitation by comparing CEBPA-mutant AML in the presence and 1 absence of additional mutations in TET2. By combining transcriptomic and epigenomic analyses of relevant in vitro and 2 in vivo models as well as data from AML patients, we identified an intricate mechanism where TET2 loss-of-function 3 rebalances Gata2 expression levels in Cebpa DM AML, and hence drives an aggressive disease. 4

TET2 mutations impair outcome for patients with CEBPA-mutant AML 2
To validate previous reports on the spectrum of co-occurring mutations in CEBPA DM AML patients, we compiled data 3 from 557 CEBPA DM cases and evaluated the co-occurrence of other known leukemia driver mutations 3-7,17 . TET2 was the 4 second most frequently co-mutated gene, with 1 in 5 CEBPA DM cases harboring TET2 mutations ( Figure 1A; 5 Supplemental table 1). Importantly, the survival of TET2-mutant (TET2 MUT ) CEBPA DM patients was significantly lower 6 than TET2 wild-type (TET2 WT ) CEBPA DM patients (Figure 1B), consistent with previous reports 16 , while the presence of 7 TET2 mutations did not cause a higher overall number of mutations in CEBPA DM patients (Supplemental figure 1A). 8 To investigate the functional consequences of TET2 and CEBPA co-mutations, we analyzed RNA sequencing (RNA-seq) 9 data from the Beat AML dataset 1 . We identified 1546 up-and 1201 downregulated genes in patients harboring a 10 combination of CEBPA and TET2 mutations when compared to CEBPA-mutant patients with wild-type TET2 (Figure  11 1C). In line with the lower overall survival of TET2 MUT CEBPA DM patients, pathways related to inflammation, hypoxia, 12 and aggressive cancer were upregulated in CEBPA-TET2 co-mutated patients (Supplemental figure 1B). 13 These findings indicate that mutations in TET2 enhance the aggressiveness of CEBPA-mutant AML by deregulation of 14 critical cellular pathways. 15

TET2 deficiency accelerates Cebpa-mutant AML 16
To study the effect of TET2 mutations in CEBPA DM AML in pathophysiologically relevant in vitro and in vivo models, 17 we utilized cell and murine models in which expression of the p30 isoform is retained (Cebpa p30/p30 or Cebpa −/p30 ), while 18 the normal p42 isoform of CEBPA is completely lost 13 . Since TET2 is predominantly inactivated by loss-of-function 19 mutations 25 , we modeled TET2 mutations either by introduction of mutations with the CRISPR-Cas9 technology or by 20 conditional knockout of the Tet2 alleles. 21 First, we introduced Tet2 mutations into a murine myeloid progenitor cell model (Cebpa p30/p30 ) (Figure 2A). Tet2-targeted 22 cells displayed a selective advantage, as they outcompeted Cebpa p30/p30 cells ( Figure 2B). Detailed analysis of the Tet2 23 mutation that was associated with the proliferative advantage showed that the Tet2 locus had acquired a +1 insertion in 24 exon 3, which resulted in a downstream premature termination codon (Supplemental figure 2A-B). In line with this, 25 clones isolated from the targeted cell pool exhibited strongly reduced TET2 protein expression (Supplemental figure 26 2C). Gene expression analysis revealed that Tet2 loss in Cebpa p30/p30 cells caused downregulated expression of 916 genes, 27 while only 474 genes were upregulated ( Figure 2C). Gene set enrichment analysis (GSEA) showed higher expression of 28 MYC and E2F targets in Cebpa p30/p30 Tet2-mutated cells, consistent with their proliferative advantage (Supplemental 29 figure 2D). 30 In summary, these data show that CRISPR/Cas9-induced TET2 loss provides a competitive advantage to myeloid 31 progenitors expressing the oncogenic CEBPA variant p30. 32 Next, we wanted to assess the impact of hematopoietic expression of CEBPA p30 (Cebpa −/p30 ) with TET2-deficiency 33 (Tet2 −/− ) on AML initiation in vivo. To do so, we transplanted lethally irradiated recipient mice with BM cells derived 34 from mice with relevant allele combinations and, following hematopoietic reconstitution, induced hematopoietic-specific 35 knockout of the Cebpa WT allele and/or the Tet2 alleles ( Figure 2D). The combination of CEBPA p30 expression with 36 Tet2 loss led to an early expansion of myeloid (Mac1 + ) cells in the BM and blood compared to mice with hematopoietic 37 cells featuring either alteration on its own ( Figure 2E). Conforming to patient data and data obtained from Cebpa p30/30 1 cells, Cebpa −/p30 Tet2 −/− hematopoietic cells gave rise to AML with shorter latency than Cebpa −/p30 Tet2 +/+ cells, with a 2 median survival of 23 and 43 weeks, respectively ( Figure 2F). TET2 deficiency alone (Cebpa +/− Tet2 −/− ) did not give rise 3 to AML and cells which retained expression of the p42 isoform from one allele (Cebpa +/p30 ) only sporadically underwent 4 leukemic transformation (Figure 2F; Supplemental figure 2E). The transformed blasts expressed myeloid (Mac1 + ) and 5 granulocytic (Gr1 + ) markers, confirming myeloid origin of the leukemia (Supplemental figure 2F). The leukemias were 6 transplantable into secondary recipients, and the shorter latency of the TET2-deficient Cebpa DM AML was preserved in 7 this setting (Supplemental figure 2G-H), indicating that TET2 not only has important tumor suppressive functions 8 during malignant transformation but also during progression of AML. 9 We performed RNA-seq on Cebpa −/p30 (Tet2 WT and knockout) AML blasts to assess changes in gene expression upon 10 TET2 deficiency. Again, we found that the majority of differentially expressed genes was decreased in TET2-deficient 11 AML blasts, with 176 down-vs. 58 upregulated genes ( Figure 2G). GSEA highlighted upregulation of genes involved 12 in IL-6-JAK-STAT-signaling and hypoxia, in line with RNA-seq data from human TET2 MUT CEBPA MUT cases 13 (Supplemental figure 1B; Supplemental figure 2I). Furthermore, pathways related to cell cycle progression (G2M 14 checkpoint and E2F targets) were enriched in TET2-deficient AML, indicating increased growth upon loss of TET2, 15 consistent with the effects observed in the cell model (Supplemental figure 2D; Supplemental figure 2I). In line with 16 this, we found that a higher frequency of Cebpa −/p30 Tet2 −/− blasts expressed the proliferation marker Ki67 (Figure 2H). 17 In addition, we also observed increased proliferative capacity of Cebpa −/p30 Tet2 −/− blasts compared to Cebpa −/p30 Tet2 +/+ 18 blasts ex vivo. This difference was dependent on Tet2 status, as the TET2 co-factor Vitamin C was able to mitigate 19 proliferation of Cebpa −/p30 Tet2 +/+ but not of Cebpa −/p30 Tet2 −/− cells (Supplemental figure 2J). 20 Collectively, these data show that TET2 deficiency accelerates the establishment and progression of CEBPA p30-driven 21 AML in vivo. 22

Loss of TET2 leads to reduced Gata2 levels in Cebpa-mutant AML 23
To find conserved gene targets of the CEBPA-TET2 axis, we integrated the transcriptomic data from our in vitro and in 24 vivo models with gene expression analyses from AML patients harboring CEBPA and TET2 mutations. Three target genes 25 exhibited downregulated expression in all three data sets; FUT8, GATA2, and SIRT5 ( Figure 3A; Supplemental figure 26

3A-C). 27
Although the deregulation of these three genes was observed across species and differential experimental setups, we next 28 aimed to investigate if their decreased gene expression was a direct result of TET2 deficiency. We therefore assessed 29 chromatin accessibility and DNA methylation as a proxy for TET2 binding and activity 26 . Through assay for transposase-30 accessible chromatin sequencing (ATAC-seq), we identified 2552 differentially accessible regions in Tet2 MUT Cebpa p30/p30 31 vs. Tet2 WT Cebpa p30/p30 cells, and consistent with an activating effect of TET2, the majority of differential regions were 32 less accessible in TET2-deficient cells ( Figure 3B). Half of the ATAC-seq peaks downregulated upon Tet2 mutation 33 were located in promoters, and these regions were enriched for GATA and NFAT motifs ( Figure 3C; Supplemental 34 figure 3D). Using whole genome bisulfite sequencing (WGBS), we observed a global increase in DNA methylation in 35 Cebpa −/p30 Tet2 −/− vs. Cebpa −/p30 Tet2 +/+ AML blasts, consistent with a loss of demethylase activity in Tet2 knockout blasts 36 ( Figure 3D). Increased DNA methylation was observed in promoter regions of genes whose expression were 37 downregulated upon TET2 loss, while upregulated genes did not show any changes (+54%; Figure 3E). Strikingly, this 38 pattern was not apparent when DNA methylation was evaluated across gene bodies (Supplemental figure 3E). Non-1 regulated, neutral genes exhibited equal increase in DNA methylation across promoter and gene body ( Figure 3E; 2 Supplemental figure 3E). Thus, loss of TET2 in Cebpa DM cells caused decreased chromatin accessibility and increased 3 methylation of DNA in promoters of TET2-responsive genes, consistent with previous reports showing that TET2 binding 4 is enriched in promoters of TET2-regulated genes 27 . 5 To identify direct CEBPA-TET2 gene target(s), we evaluated the previously identified conserved candidates based on 6 changes in DNA methylation of their promoters. Out of the three target genes, only the gene encoding the transcription 7 factor GATA-binding factor 2 (GATA2) showed a gain of DNA methylation in the promoter of the gene variant 2 (Gata2 8 V2) upon TET2 deficiency (+46%; Figure 3F). In line with this, specifically the Gata2 V2 mRNA isoform was 9 downregulated in TET2-deficient Cebpa DM AML blasts (−86%; Figure 3G), while changes in mRNA expression and 10 promoter methylation of Gata2 V1 did not reach statistical significance (Figure 3F-G). 11 In summary, these analyses identify Gata2 (locus overview in Figure 3H) as a conserved target of the CEBPA-TET2 12 axis across several settings. TET2 deficiency causes increased DNA methylation of the Gata2 promoter, resulting in 13 reduced mRNA expression. 14

Moderate Gata2 reduction increases competitiveness of Cebpa-mutant AML 15
GATA2 is an essential transcription factor for hematopoietic cells and has profound effects on HSC maintenance. Altogether, our data suggest that loss of TET2 in Cebpa DM AML causes a moderate decrease in Gata2 expression, which 3 in turn increases the competitive fitness of the leukemia. Hence, this indicates that TET2 and GATA2 mutations are 4 partially redundant in CEBPA DM AML, providing a mechanistic rationale for the mutational spectra observed in this AML 5 entity. 6 Increased CEBPA p30 binding to the Gata2 distal hematopoietic enhancer drives expression of Gata2 via TET2 7 We next asked if GATA2 expression is dependent on CEBPA mutational status. To this end, we exploited published 8 transcriptomics data from human and mouse CEBPA DM AML 11 . GATA2 expression was increased in human CEBPA DM 9 leukemic granulocyte/monocyte progenitors (GMPs) compared to GMPs from healthy donors (+77%; Supplemental 10 figure 5A). Correspondingly, Gata2 was upregulated in murine Cebpa p30/p30 leukemic GMPs as compared to normal 11 GMPs (+43%; Figure 5A). Since CEBPA is known to exert its transcription factor activity by binding to enhancers and 12 thereby promote gene expression 36 , we assessed binding of CEBPA to the crucial Gata2 distal hematopoietic enhancer 13 (G2DHE; −77 kb in mouse) that governs Gata2 expression in hematopoietic stem and progenitor cells including 14 GMPs 11,37 . Notably, we found substantially increased levels of CEBPA bound to the G2DHE in Cebpa p30/p30 leukemic 15 GMPs compared to normal counterparts (+147%; Figure 5B Figure 5E). Combined, these data suggest that CEBPA binding to the G2DHE is important for promoting 26 Gata2 expression in Cebpa DM AML. Further, the G2DHE has been shown to primarily regulate expression of the 27 hematopoietic specific Gata2 variant 2 (V2) 38,39 , conforming with our data that particularly the Gata2 V2 promoter 28 displayed an increase in DNA methylation and that the Gata2 V2 mRNA was downregulated in TET2-deficient Cebpa DM 29 AML blasts (Figure 3F-G). In light of these findings, we asked whether elevated CEBPA level, and not the CEBPA mutation(s) per se, drives the 1 selective pressure for GATA2-and/or TET2 loss in AML to achieve moderate GATA2 levels that are optimal for leukemia 2 growth. We therefore stratified AML cases in the Beat AML cohort 1 based on CEBPA expression and assessed their 3 GATA2 and TET2 mutational status. Indeed, the frequency of GATA2 and/or TET2 mutations was three-fold higher in 4 CEBPA HIGH AML compared to the CEBPA LOW samples (Figure 5J). In line with previous data showing a hypermorphic 5 effect of CEBPA DM 11 , the CEBPA HIGH group contained the majority of the CEBPA-mutant cases in the cohort (82 and 6 100% of CEBPA SM and CEBPA DM , respectively), while none of the cases in the CEBPA LOW group were CEBPA-mutated. 7 In conclusion, our data suggest that elevated CEBPA binding to the G2DHE, driven by the hypermorphic effect of 8 Cebpa NT , increases TET2-mediated demethylation of the Gata2 promoter, which leads to elevated Gata2 levels in 9 Cebpa DM AML. In this context, Cebpa DM AML cells gain a competitive advantage by loss of TET2, which in turn 10 promotes an increase in DNA methylation at the Gata2 promoter resulting in the rebalancing of Gata2 levels. Cebpa −/p30 Tet2 +/+ blasts ( Figure 6C). Importantly, a longer intermittent 5-AZA treatment prolonged the survival of mice 23 transplanted with Cebpa −/p30 Tet2 −/− blasts (median survival +22%; Figure 6D-E), while it did not affect disease latency 24 of mice transplanted with Cebpa −/p30 Tet2 +/+ blasts. 25 In summary, we show that the demethylating agent 5-AZA can restore Gata2 expression levels in TET2-deficient 26 Cebpa DM AML to that of TET2-proficient Cebpa DM AML, and concomitantly reduce leukemic burden and prolong 27 survival of mice transplanted with TET2-deficient Cebpa DM leukemic blasts. 28

DISCUSSION 1
Mutational cooperativity is a fundamental driver of cancer development, progression, and aggressiveness. For CEBPA DM 2 AML, co-occurring lesions have been found in genes such as GATA2, TET2, WT1, FLT3, and CSFR3. While the 3 mechanistic basis for the cooperation between CEBPA and GATA2/CSFR3 mutations has been investigated using mouse 4 models 14,15 , we have very little insights into why other lesions, such as those in TET2, are overrepresented in CEBPA DM 5 AML. Here, we show that TET2 loss-of-function in CEBPA DM AML leads to an aggressive disease phenotype by 6 rebalancing the increased and suboptimal levels of GATA2 that are induced by hypermorphic CEBPA NT mutations driving 7 CEBPA-p30 isoform expression (see model in Figure 7A). Specifically, loss of TET2 binding to the hematopoietic 8 specific G2DHE enhancer results in increased DNA methylation in the promoter region of the hematopoietic-specific 9 Gata2 isoform (Gata2 V2). This proleukemic effect of TET2 loss can be reversed by the demethylating agent 5-10 azacytidine, suggesting that this could be a potential treatment option in CEBPA DM TET2 MUT patients. Altogether, our 11 work proposes that CEBPA-mutant AMLs acquire additional lesions in genes such as GATA2 and TET2 to reestablish 12 balanced GATA2 levels that permit leukemia development and progression. frequently found in TET2-deficient CEBPA DM compared to TET2-proficient CEBPA DM AML. Finally, we showed that 31 mutations in GATA2 and TET2 are overrepresented in AML cases with high CEBPA expression. This supports the notion 32 that unfavorable, high GATA2 levels in AML promoted by the CEBPA-TET2 axis are not limited to CEBPA DM AML, 33 but also include cases where CEBPA expression is high for other reasons. Further, this model also suggests that a major 34 proleukemic effect of TET2 deficiency is to rebalance GATA2 levels in the context of CEBPA DM AML (Illustrated in 35 Figure 7B). 36 GATA2 expression is mainly driven by the conserved G2DHE in normal myeloid progenitors and leukemic blasts by 1 promoting expression from the hematopoietic specific Gata2 V2 promoter 37,38,42,52 . Our data demonstrate that CEBPA 2 plays a key role in regulating G2DHE activity. Specifically, we show that the hypermorphic effects of CEBPA DM 11 , and 3 experimental models thereof, result in increased GATA2 expression compared to CEBPA WT , and that CEBPA deficiency 4 resulted in reduced Gata2 levels. Secondly, we observed increased CEBPA binding to the G2DHE in Cebpa DM AML 5 compared to normal progenitors and found that deletion or mutagenesis of the CEBPA-bound region of the enhancer 6 resulted in lower expression of Gata2 in Cebpa DM cells. In further support of a role of CEBPA, the G2DHE is highly leads to impaired upregulation of myeloid-specific genes upon Cebpa induction, with corresponding increased promoter 23 methylation 65 . Also, in TET2 MUT or Tet2 −/− leukemia an enrichment of CEBP-motifs at or near hypermethylated CpGs 24 was observed 26,66 . Importantly, AML with silenced CEBPA is associated with DNA hypermethylation, a feature that is 25 not present in CEBPA DM AML, which may suggest a broader function of CEBPA in recruitment of TET2 67 . 26 Given the well-known issues with current TET2 antibodies, we have been unable to ChIP TET2 in our cells, despite 27 numerous attempts 68 . Despite this caveat, we conclude that a growing body of evidence is supporting an important role 28 of CEBPA in recruitment of TET2 to DNA, promoting demethylation and transcription of target genes. 29 While our findings suggest that GATA2 MUT and TET2 MUT both converge at rebalancing the increased expression of GATA2 30 in CEBPA DM AML, patients with CEBPA DM and GATA2 MUT have a more favorable prognosis 16,30-32,47 than patients 31 harboring the CEBPA DM and TET2 MUT combination 16,17 . This suggests that while GATA2 deregulation plays an important 32 role in leukemogenesis in the CEBPA MUT context, TET2 deficiency may likely contribute to malignancy through 33 additional mechanisms that shall remain subject of another study. Of clinical interest, we find that TET2 deficiency 34 renders Cebpa DM AML sensitive to 5-AZA and that TET2-deficient cells lose their proliferative advantage over TET2-35 proficient cells following 5-AZA treatment. In agreement with TET2-dependent Gata2 expression, our and previous 36 results show that 5-AZA treatment derepresses Gata2 expression in TET2-deficient cells 42 . Intriguingly, CEBPA CT 37 mutations have recently been reported to sensitize AML to treatment with hypomethylating agents by disrupting the 1 inhibitory interaction with DNMT3A mediated by the wild-type CEBPA bZIP domain 69 . Taken together, this suggests 2 that demethylating agents could be a particularly feasible treatment option in CEBPA DM TET2 MUT patients. 3 In conclusion, our results reveal that GATA2 is a conserved target of the CEBPA-TET2 mutational axis in CEBPA DM AML 4 and we propose an intricate mechanism by which elevated CEBPA p30 levels mediate recruitment of TET2 to regulatory 5 regions of the Gata2 gene to promote its expression. We demonstrate that increased GATA2 levels are disadvantageous 6 to CEBPA DM leukemic cells and that this can be counteracted by TET2 loss thus providing an explanation for the co-7 occurrence of CEBPA and TET2 lesions in AML. Finally, increased Gata2 promoter methylation, inflicted by TET2 8 deficiency, can be restored by demethylating 5-AZA treatment, thereby providing entry points for the development of 9 rational targeted therapies in AML patients with these mutations. Competitive shRNA-knockdown: C57BL6/6J.SJL recipients (female, 10-12 weeks old) were sub-lethally irradiated 22 (500 cGy) 12-24h prior to being intravenously injected with a 1:1 mix of Cebpa p30/p30 cells 13 transduced with shRNA 23 targeting Gata2 (Supplemental table 6) or with control-shRNA as previously described 83 . The ratio of Gata2-or control-24 shRNA-GFP + to control-shRNA-YFP + cells was analyzed by flow cytometry four weeks later. 25 5-azacytidine treatment: C57BL6/6J.SJL recipients (female, 10-12 weeks old) were sub-lethally irradiated (500 cGy) 26 12-24h prior to being intravenously injected with 1×10 5 thawed live BM cells from moribund secondary recipient mice. 27 The mice were given Ciprofloxacin in the drinking water to prevent infections 3 weeks post-irradiation. The mice received 28 intraperitoneal injections with the demethylating agent 5-azacytidine (2.5 mg/kg/day in saline; #A2385 Sigma-Aldrich) 29 at days 6-10 and 20-24 post-transplantation. The time of the BM cell injection was set as time-point zero for the survival 30 study and mice were monitored and euthanized when moribund. To evaluate the effects of short-term 5-azacytidine 31 treatment, recipient mice were treated at days 13-15 and euthanized 24 hours after the last injection. BM was collected 32 for FACS, and sorted cells were frozen for subsequent analysis. After aspiration of the supernatant, Cebpa p30/p30 cells were seeded at a density of 0.5-1×10 5 cells/cm 2 . The transduction 25 was repeated the following day, and the cells were cultured for 24 h prior to FACS sorting of transduced (GFP + /YFP + ) 26 cells on a BD FACSAria TM III (BD Bioscience). The efficiency of shRNA-mediated gene expression knockdown was 27 assessed with qPCR and cells were used for transplantation and assessment of their competitiveness in vivo. 28 using Pfu Turbo Cx Hotstart DNA polymerase (#600410 Agilent) with primers targeting a part of the CpG island in the 10 Gata2 V2 promoter region (Supplemental table 5). After verification of their correct size, PCR products were cloned 11 using Zero Blunt Topo PCR Cloning kit (#450245 Invitrogen) and single colonies were picked and amplified. Plasmid 12 DNA was isolated using NucleoSpin Plasmid EasyPure (#740727.250 Macherey-Nagel), the correct insert size was 13 verified after cleavage with restriction enzyme EcoRI (#R0101 New England Biolabs) and sent for Sanger sequencing 14 using the M13 primer provided with the cloning kit. 15

Statistics 16
Data were analyzed for significance using parametric tests, with prior log-transformation if necessary to achieve normal 17 distribution. Normality was evaluated by Shapiro-Wilk test. Two-group analyses were done using unpaired two-tailed t-18 test. Multiple-group analyses were done with one-way-ANOVA followed by multiple comparisons correction using 19 Dunnett when comparing to a reference group, or two-way-ANOVA followed by multiple comparisons correction using 20 Šídák test when comparing two independent factors across four groups. Data sets that did not pass normality tests were 21 analyzed by Kruskal-Wallis test followed by multiple comparisons correction using Dunn's test. Survival curves were 22 analyzed using Mantel-Cox Log-rank test. To compare distributions Wilson/Brown binominal test was used. P-values 23 <0.05 were considered statistically significant. Data was analyzed using GraphPad Prism (v. 9). Data is shown as 24 mean±SEM unless otherwise stated. 25

Data availability 26
The data generated in this study is publicly available in Gene Expression Omnibus