RNA shielding of P65 is required to potentiate oncogenic inflammation in TET2 mutated clonal hematopoiesis

Abstract

TET2 mutations (mTET2) are common genetic events in myeloid malignancies and clonal hematopoiesis (CH). These mutations arise in the founding clone and are implicated in many clinical sequelae associated with oncogenic feedforward inflammatory circuits. However, the direct downstream effector of mTET2 responsible for the potentiation of this inflammatory circuit is unknown. To address this, we performed scRNA and scATAC-seq in COVID-19 patients with and without TET2-mutated CH reasoning that the inflammation from COVID-19 may highlight critical downstream transcriptional targets of mTET2. Using this approach, we identified MALAT1, a therapeutically tractable lncRNA, as a central downstream effector of mTET2 that is both necessary and sufficient to induce the oncogenic pro-inflammatory features of mTET2 in vivo. We also elucidate the mechanism by which mTET2 upregulate MALAT1 and describe an interaction between MALAT1 and P65 which leads to RNA “shielding” from PP2A dephosphorylation thus preventing resolution of inflammatory signaling.

Introduction:

Ten Eleven Translocation 2 (TET2) is a dioxygenase that regulates normal hematopoiesis by oxidizing 5-methylcytosine residues on DNA to 5-hydroxymethylcytosine, eliciting a cascade of enzymatic reactions that results in cytosine demethylation. TET2 loss of function (LOF) somatic mutations have been detected in all myeloid neoplasms. They are the most frequent genomic event in chronic myelomonocytic leukemia (CMML) comprising 50% - 60% of all mutations (1,2). They are also seen in about 20% of AML, 20% - 35% of myelodysplastic syndromes (MDS) (3) and are the second most common genetic event in clonal hematopoiesis (CH) where they have been linked to many of the non-hematological clinical sequelae observed in patients (4–6). TET2 LOF also occurs via its enzymatic inhibition by 2 hydroxyglutarate (2HG), an oncometabolite produced by cells with mutations in the isocitrate dehydrogenase (IDH) genes (7). When present in myeloid malignancies, TET2 LOF mutations universally occur in the founding clone and result in genome-wide hypermethylation (2). No direct therapies for TET2 LOF mutations have been identified, in part, because of the challenges associated with targeting LOF mutations and because the major downstream effectors responsible for its oncogenic phenotypes have not been identified. Indeed, the broad epigenomic and transcriptomic alterations observed as a consequence of TET2 LOF mutations suggest that there may be many cooperating downstream effectors making therapeutic strategies against these similarly challenging.

Murine models of Tet2 deletion recapitulate several features of myeloid malignancies but most closely resemble CMML. These features include increased self-renewal capacity of hematopoietic stem and progenitor cells (HSPCs), splenomegaly, and peripheral monocytosis (1). The constellation of splenomegaly, myelomonocytic (CD11b+Gr-1+) expansion, and extramedullary hematopoiesis (signified by increased HSPCs in the spleen) seen in Tet2^−/− mice also defines the preleukemic myeloproliferation (PMP) phenotype which is a prerequisite for myeloid transformation in murine models (8).

Several elegant studies have demonstrated that Tet2^−/− mice and humans with TET2 LOF mutations establish an oncogenic feedforward inflammatory circuit required to develop PMP, CH, and overt myeloid malignancy (8–13). This circuit can be initiated by varied exogenous inflammatory signals such as the intestinal microbiota and atherosclerotic disease (8,14). These exogenous inflammatory signals then result in myeloid skewing of TET2-deficient HSPCs (i.e., PMP) and the aberrantly sustained secretion of proinflammatory cytokines resulting in a highly inflammatory microenvironment. TET2 LOF HSPCs then have a competitive advantage in this inflammatory context leading to their expansion and oncogenic transformation (12).

Despite increased recognition of the collaboration between inflammatory circuits and TET2 LOF myeloid transformation, the proximal downstream effectors of TET2 LOF that lead to the potentiation of this oncogenic inflammatory circuit remain largely unknown. To address this knowledge gap, we leveraged single cell (sc) RNA-seq and scATAQ-seq, and identified Metastasis Associated Lung Adenoma Transcript 1 (MALAT1), a druggable long non-coding RNA found in the nuclear speckles (15), as a critical downstream effector of TET2 LOF mediated oncogenic inflammation. We show that TET2 LOF increases MALAT1 expression via a disruption of transcriptional repression, and that Malat1 overexpression is sufficient and necessary for the development of oncogenic inflammatory features of TET2 deficiency. We further show that MALAT1 establishes the feed forward inflammatory circuits required to promote myeloid transformation by physically interacting with and attenuating PP2A-dependent P65 serine dephosphorylation. Together, these findings provide previously unknown mechanistic insights into TET2-LOF-driven leukemogenesis, unmask new mutation-specific therapeutic targets, and describe an RNA shielding effect that sustains inflammatory circuits critical to disease progression by inhibiting dephosphorylation of P65 at Ser536.

Results

TET2 Loss of Function mutations increase MALAT1 expression.

CH arises from the clonal expansion of HSPCs generally harboring a single premalignant driver mutation (6). Given the co-operation between TET2 LOF and inflammation, we reasoned that mapping the epigenetic and transcriptional landscape of CH patients in the context of SARS-CoV-2, a highly inflammatory disease process (16), would provide a uniquely informative human dataset to ascertain downstream effectors of TET2 LOF without the confounding effects of other somatic driver mutations.

We assessed peripheral blood mononuclear cells (PBMCs) from 15 patients with SARS-CoV-2 and no CH (COVID-19+/CH-) and 6 patients with SARS-CoV-2 and TET2 mutated CH (COVID-19+/TET2^MT) using Multiome (scRNA-seq + scATAC-seq) and scRNA-seq (Fig S1A). These patients were matched for their SARS-CoV-2 severity with the NIH stratification criteria (17). By pooling 73,133 cells from scRNA-seq and 20,908 from Multiome analysis and subjecting them to dimensionality reduction using UMAP, followed by cell type identification using SingleR, we identified the expected peripheral blood constituents such as B and T lymphocytes, monocytes, dendritic cells, basophils, neutrophils and HSPCs (Fig S1B).

When comparing cell type abundance, we observed a significant increase in pro-inflammatory classical CD14⁺ monocytes in COVID-19+/TET2^MT patients relative to COVID-19+/CH- patients (Fig S1C – S1E), motivating us to focus our analysis of differential gene expression on this cellular subset. As expected, a significant global transcriptional downregulation was observed. However, outside of mitochondrial RNA we surprisingly observed only two significant transcripts, MALAT1 and NEAT1, among the top 15 upregulated DEGs (Fig 1A). This observation remained even when lower mitochondrial contamination thresholds were employed and after removing low viability samples which generated adequate cDNA libraries (Fig S1F, S1G).

Figure 1: TET2 Loss of function (LOF) increases MALAT1 expression.

(A) Volcano plot showing significant differentially expressed genes (adjusted p < 0.05, Wilcoxon rank sum test) comparing CD14 monocytes from the COVID-19⁺/TET2^MT cohort to those from COVID-19⁺/ CH⁻ cohort.

Violin plots showing expression of MALAT1 (B) and NEAT1 (C) in monocyte subsets and dendritic cells. Black dot shows mean expression. ****: p <= 0.0001 (Wilcoxon rank sum test).

MALAT1 (left) and NEAT1 (right) expression from TCGA (D) and BEAT (E) AML cohorts, n= 27 TET2 LOF v 119 WT in TCGA P= 0.003 for MALAT1 and p=0.03 for NEAT1: n= 147 TET2 LOF v 439 WT in BEAT p= 0.0003 for MALAT1 and p= 0.04 for NEAT1. P value was calculated separately for MALAT1 and NEAT1 using Mann Whitney test.

(F) MALAT1 and NEAT1 expression in HEL parental and HEL cells with TET2 knocked down using T7 CRISPR RNP. Data shown are mean ± SEM of five independent experiments each ran with technical triplicates. P value was calculated separately for MALAT1 (0.005) and NEAT1 (0.08) with unpaired t test.

(G) MALAT1 and NEAT1 expression in isogenic TF-1 IDH2^R140Q mutant cell line relative to WT cells. Data shown are mean ± SEM of five experiments for MALAT1 and six for NEAT1 each ran with technical triplicates. P value was calculated separately for MALAT1 (0.003) and NEAT1 (0.3) with unpaired t test.

(H) MALAT1 and NEAT1 expression in U937 (i), THP-1 (ii), OCI-AML 3 (iii), SKM1 (iv) and HEL (v) cell lines treated with 100uM 2HG for seven days. Data shown are mean ± SEM of two separate experiments each ran with technical triplicates. P value was calculated separately for MALAT1 (U937=0.02, THP-1=0.01, HEL=0.006, OCI-AML3=0.01 and SKM-1= 0.0007) and NEAT1 (U937=0.01, THP-1=0.3, HEL=0.3, OCI-AML3=0.2 and SKM-1= 0.005) in each cell line with unpaired t test.

(I) Malat1 and Neat1 expression in immortalized Tet2^−/− murine BMNCs relative to immortalized C57BL6 BMNCs. Data shown are mean ± SEM of five separate experiments each ran with technical triplicates. P value was calculated separately for Malat1 (0.03) and Neat1 (0.03) with unpaired t test.

(J) Malat1 and Neat1 expression in BMNCs from Tet2^f/f mice compared to age and sex-matched Mx1-Cre controls. n= 7 per group. Data shown are mean ± SEM. P value was calculated separately for Malat1(0.005) and Neat1 (0.15) with unpaired t test.

(K) Malat1 and Neat1 expression in BMNCs from transgenic Tet2^−/− mice and wild type B6 mice. Data shown are mean ± SEM for n= 12 WT B6, 9 Tet2^−/− and 9 Tet2^+/−. P value was calculated separately for Malat1 (p<0.0001 WT v Tet2^−/− and p= 0.0008 WT v Tet2^+/−) and Neat1 (p= 0.02 WT v Tet2^−/− and p= 0.97 WT v Tet2^+/− using one way ANOVA with Tukey’s correction for multiple comparisons.

Both MALAT1 and NEAT1 are non-coding RNA transcripts, and both were significantly over-expressed in COVID-19+/TET2^MT when compared with COVID-19+/CH−, and when compared to previously published healthy control datasets (18). MALAT1 and NEAT1 overexpression was also observed in other cell types (Fig 1B, 1C) but not in peripherally circulating HSPCs (Fig S1H). To validate an association between TET2 LOF and MALAT1 and NEAT1 overexpression in myeloid malignancies, we measured their expression in the TCGA (Fig 1D) and BEAT (Fig 1E) AML datasets (19,20). These analyses confirmed that TET2 LOF is associated with a significant increase in MALAT1 and, to a lesser extent, NEAT1 expression compared to other genetic subtypes in human bone marrow mononuclear cells (BMNCs).

We next explored whether TET2 LOF upregulates MALAT1 and NEAT1 expression in human and murine model systems. We first observed that MALAT1, but not NEAT1, expression was increased in HEL cells with TET2 knocked down (Fig 1F, S1I), and CRISPR base-edited isogenic IDH2^R140Q mutant TF-1 human cells (Fig 1G) relative to their respective controls. We validated this finding by culturing five human leukemia cell lines (U-937, THP-1, HEL, OCI-AML3 and SKM-1) with R-2HG or PBS (untreated) for 7 days and demonstrated an increase in MALAT1 expression in all R-2HG treated cells, but variable results with NEAT1 (Fig 1Hi–v). Further, Immortalized Tet2^−/− murine BMNCs (Fig 1I); BMNCs from Tet2 floxed mice (Fig 1J); and BMNCs from transgenic Tet2^−/− mice (Fig 1K) all showed significant increases in Malat1 but variable effects on Neat1 relative to their respective controls. Collectively, these data suggest that TET2 LOF consistently induces MALAT1 but not NEAT1 overexpression in both human and murine hematopoietic cells.

TET2 LOF mutations increase MALAT1 expression via loss of EGR1 repression.

Given the role of TET2 as an epigenetic regulator (21) and the challenges associated with evaluating methylation at single-cell resolution, we leveraged our multi-omic single-cell dataset to determine the mechanism by which MALAT1 expression is increased in TET2 LOF. Because chromatin accessibility is reflective of the active enhancer/repressor and promoter structures and is strongly associated with methylation status, we compared differentially accessible peaks between COVID-19+/CH− and COVID-19+/TET2^MT demonstrating that there was decreased accessibility in cells of myeloid lineage in the COVID-19+/TET2^MT cohort as expected (Fig. S2A).

Similarly, the number of open chromatin regions (cut sites per cell) were reduced in monocytes (Fig. 2A), a strong indicator that TET2 LOF could be decreasing chromatin accessibility in these cells secondary to enhancer and promoter methylation.

Figure 2. EGR1 acts a repressor of MALAT1 expression.

(A) Violin plots showing the total number of cut sites (sum of TF-IDF normalized cut site counts) in each cell type comparing COVID-19⁺/CH⁻ and COVID-19⁺/TET2^MT cohorts. Black dots show the mean value. ns: p > 0.05, *: p <= 0.05, ****: p <= 0.0001 (Wilcoxon rank sum test).

(B) Violin plots showing the number of co-accessibility connections per peak predicted using Cicero (co-access score > 0) comparing CD14 monocytes from COVID-19⁺/CH⁻ to COVID-19⁺/TET2^MT cohorts, showing all peaks (left), peaks with CpG site (middle), and peaks without CpG site (right). Peaks with CpG sites were identified by overlapping peaks with CpG sites (EPIC array) using bedtools intersect. Black dots show the mean value. ****: p <= 0.0001 (Wilcoxon rank sum test).

(C) Volcano plot showing binding sites enrichment scores (ChromVAR) of DNA binding proteins significantly different between COVID-19⁺/CH⁻ and COVID-19⁺/TET2^MT cohorts (adjusted p value < 0.05, Wilcoxon rank sum test). Cells from each cell type were tested independently. Curated ChIP-seq peaks of DNA binding proteins assayed in myeloid cells, obtained from ReMap 2022, were used as the DNA binding sites to calculate enrichment.

(D) Violin plots showing expression of EGR1 in monocyte subsets and hematopoietic stem and progenitor cell (HSPC) populations, in the scRNA-seq data. The number of cells in each group is shown at the bottom of each violin. Black dot shows mean expression. ****: p <= 0.0001; ns: p > 0.05 (Wilcoxon rank sum test).

(E) Coverage plot showing the chromatin accessibility, gene expression (violin plots), candidate cis-regulatory elements (CCRE) annotated by the ENCODE consortium, EGR1 binding sites (from ReMap2022 - Biotype: Monocyte and Macrophage), open-chromatin peaks defined by scATAC-seq, TET2 ChIP-seq signal, CpG sites (Infinium Methylation EPIC array) and the Cicero predicted coaccessible peaks around the MALAT1 gene locus. Coaccessible peaks links that show interactions between EGR1 binding sites and MALAT1 transcription start site (TSS) are highlighted using thicker lines. reg. regulatory; Enh., Enhancer; Prox. Proximal; Dist. Distal; K4m3, H3H4me3.

Western blots showing EGR-1 knockdown and qRT-PCR showing MALAT1 and NEAT1 expression in TF-1 (F) and HEL (G) cells infected with EGR-1 shRNA and non-target shRNA (NT) control. Western blots are representative of three independent experiments and qRT-PCR data shown are mean ± SEM of four independent experiments each ran with technical triplicates.

In fact, co-accessibility analysis of cut sites within each cell type identified potential chromatin interactions, quantified as connections per peak, demonstrating a significant loss of co-accessibility in the COVID-19+/TET2^MT cohort. This loss was more prominent when analyzing connections that have CpGs associated with them, another strong indicator of the specific effect of TET2 LOF associated with methylation patterns (Fig 2B).

To understand which transcription factors were most affected by TET2 LOF-driven disruption, we performed differential enrichment analysis from published ChIP-seq datasets (22) and transcription factor motif enrichment in differentially accessible peaks from our single cell dataset. Once again, the major differences between the COVID-19+/CH− cohort and the COVID-19+/TET2^MT were found in classical monocytes and predominantly affected transcription factors highly associated with myeloid differentiation, such as EGR1 and CEBPD (Fig. 2C, S2B, S2C). Next, we compared the expression of each transcription factor observed to have differential chromatin accessibility in COVID-19+/CH− and COVID-19/TET2^MT cohorts. Among these transcription factors, EGR1 showed the greatest difference in expression in CD14 monocytes (Fig S2D, 2D) and was of particular interest because it has been demonstrated to mediate some of the downstream effects of TET2 mutations in CMML monocytes via changes in chromatin accessibility (23). Given this data and the known role of EGR1 as a repressor in monocytes and macrophages (24), we hypothesized that TET2 LOF-dependent loss of EGR1 repression at regulatory regions of MALAT1 was responsible for its increased expression. However, considering there were subsets of monocytes that showed an increase in MALAT1 expression but no change in EGR1 expression (Fig 1B, 2D), we hypothesized that EGR1’s effect was principally due to changes in chromatin accessibility rather than decreased expression.

To explore the impact of TET2 LOF on EGR1 chromatin occupancy and thus activity, we overlapped the chromatin accessibility signal of the most represented cell types (CD14 monocytes, intermediate monocytes, NK cells and CD4 T lymphocytes) in our dataset, co-accessibility connections predicted in each cohort using scATAC-seq, EGR1 binding sites from publicly available data (24,25), and the candidate cis-regulatory element (CCRE) signatures annotated by the ENCODE consortium. We also overlapped TET2 binding sites from ChIP-seq data and annotated predicted CpG islands. Using this approach, we identified EGR1 binding sites that overlap with TET2 peaks, predicted CpG islands, and cis-regulatory regions which also share co-accessibility with the MALAT1 transcription start site (TSS) (Fig. 2E) suggesting possible enhancer-promoter interactions by which EGR1 regulates MALAT1 expression. Moreover, these interactions were only observed in the COVID-19+/CH− cohort and not in the COVID-19+/TET2^MT cohort implying that the loss of EGR1 repression results in MALAT1 over expression.

EGR1 chromatin immunoprecipitation at the MALAT1 binding site in TF-1 IDH2 mutant cells and isogenic wild type controls demonstrated loss of EGR1 occupancy at MALAT1 regulatory regions in IDH2 mutant cells suggesting that these aberrantly methylated regions may directly disrupt EGR1 occupancy in TET2 LOF models (Fig S2E).

Lastly, we functionally validated EGR1 repression of MALAT1 through shRNA knockdown in two human leukemia cell lines demonstrating that downregulation of EGR1 was associated with upregulation of MALAT1 expression, but not NEAT1 (Fig. 2F, 2G). Collectively, these data suggest that EGR1 is a repressor of MALAT1 transcription and that TET2 LOF disrupts EGR1 activity leading to upregulation of MALAT1 expression.

Malat1 is necessary for TET2 LOF phenotypes.

Having established that TET2 LOF directly increases MALAT1 expression, we next annotated MALAT1 expression across normal human and murine hematopoiesis using pseudotime analysis of publicly available scRNA-seq datasets (26,27). This analysis demonstrated that MALAT1 is highest in hematopoietic stem cells with high differentiation potential, and monocytes consistent with the cellular phenotypes of TET2 LOF models (Fig S3A – S3C). Given this, we hypothesized that Malat1 may be a critical downstream effector of TET2 LOF phenotypes. We tested this hypothesis by crossing previously reported Malat1^−/− (28) and Tet2^−/− (29) mice to obtain Tet2^−/−Malat1^−/− (TKM) and Tet2^+/−Malat1^−/− (TKM+) mouse models (Fig S3D). This Malat1^−/− model was chosen because it has been demonstrated to have minimal effects on neighboring enhancers and genes including Neat1 (30).

Peripheral blood, spleen and bone marrow cells were then compared in Tet2^−/−, Tet2^+/−, TKM and TKM+ GEM models. As expected, Tet2^−/− and to a lesser extent Tet2^+/− displayed PMP defined by splenomegaly, myelomonocytic (CD11b+Gr-1+) expansion, and expansion of spleen Lin⁻Sca1⁺ckit⁺ (LSK) HSPCs at 8 weeks (1,8). However, Malat1 deletion attenuated the expansion of the myelomonocytic, as well as mature CD11b+ populations in the spleens of the TKM and TKM+ mice to levels comparable to that of wild type B6 (WT) mice. Additionally, splenomegaly and expansion of LSK cells in the spleen were also attenuated to WT levels upon deletion of Malat1 in TKM and TKM+ mice at 8 weeks (Fig 3A – 3D).

Figure 3. Malat1 is necessary for TET2-deficient phenotypes.

Quantification of CD11b+Gr-1+ myelomonocytic cells (A) and CD11b+ mature myeloid cells (B) in the spleens of 8-week-old WT, Tet2^−/−, Tet2^−/−;Malat1^−/− (TKM), Tet2^+/−, and Tet2^+/−;Malat1^−/− (TKM+) mice. Data shown are mean ± SEM. Homozygous n= 12 WT, 10 Tet2^−/−, 7 TKM. Heterozygous n= 12 WT, 9 Tet2^+/− 9 TKM+.

(C) Spleen weights of 8-week-old WT, Tet2^−/−, TKM, Tet2^+/−, and TKM+ mice. Data shown are mean ± SEM. Homozygous n= 12 WT and 7 Tet2^−/− and TKM. Heterozygous n= 12 WT and 9 Tet2^+/− and TKM+

(D) Quantification of lineage negative Sca1+ ckit+ LSK cells in the spleens of 8-week-old WT, Tet2^−/−, TKM, Tet2^+/−, TKM+ mice. Data shown are mean ± SEM. Homozygous n= 12 WT, 5 Tet2^−/− and TKM Heterozygous n= 12WT, 8 Tet2^+/− and TKM+

(E) Quantification of Ly6-C^high classical monocytes in the spleens of 8-week-old mice. Data shown are mean ± SEM Homozygous n= 12 WT, 10 Tet2^−/−, 7 TKM. Heterozygous n= 12 WT, 9 Tet2^+/− 9 TKM+

(F) Proportions of monocyte subsets in peripheral blood of 8-week-old WT Tet2^−/− and TKM mice. Each subset is expressed as a percentage of Total CD11b+CD115+ monocytes. Data shown are mean ± SEM. N= 8 WT, 6 Tet2^−/−, 6 TKM

(G) Number of colony forming units (CFU) in serial replating assay using FACS sorted bone marrow LSK cells from 8-week-old WT, Tet2^+/− and TKM+ mice. Data shown are mean ± SEM. p values shown for all genotypes are relative to Tet2^+/−

(H) Schematic for competitive bone marrow transplant experiment

(I) CD45.2 (donor) chimerism in primary (left) and secondary (right) transplant experiments using BMNCs from WT Tet2^−/− and TKM mice. Data shown are mean ± SEM for each time point.

(J) Percent Ly6-C^high classical monocytes, CD11b+Gr-1+ myelomonocytic cells, spleen weights, and spleen LSK cells in primary competitive bone marrow transplant recipients of Tet2^−/−, TKM, and WT BMNCs. Figure shows phenotypes expressed as fold change relative to WT. Data shown are mean ± SEM. n= 11 WT and TKM, 7 Tet2^−/−

(A)-(J) * p<0.05, ** p<0.01, *** p<0.001 **** p<0.0001 calculated using one-way ANOVA with Tukey’s correction for multiple comparisons for (A) –(G). Statistical significance in (I) was calculated using a two-way ANOVA.

Given the expansion of classical (MO1) monocytes in our single cell human dataset, we explored whether this was an underrecognized phenotype of TET2 mutation in humans. We assessed the frequency of classical monocytes in the peripheral blood of 114 MDS patients grouped according to TET2 mutational status (93 WT TET2 vs 21 mutated TET2). Profiling these patients by FACS revealed an expansion of classical monocytes in patients with TET2 mutations relative to patients with wild type TET2 (Fig S3E). Further, in our GEM models, Ly6-C^high monocytes constituted a significantly larger proportion of peripheral blood and spleen monocytes in Tet2^−/− but not TKM mice relative to WT mice (Fig 3E, 3F).

Given that increased self-renewal capacity of bone marrow LSKs is an established feature of TET2 LOF (1), we sought to determine the impact of Malat1 deletion on the stem cell compartment. Immunophenotypic annotation of HSPCs by FACS revealed a numerical but statistically insignificant reduction in LSK and MPP3 (myeloid progenitor) populations (Fig S3F, S3G). We then assessed whether Malat1 deletion could attenuate the increase in self renewal capacity seen with TET2 LOF by performing serial replating colony formation assays using Tet2^+/− and TKM+ LSKs. Consistent with the attenuation of PMP and classical monocytosis in TKM and TKM+ mice, Tet2^+/− LSK cells retained self-renewal capacity after five replatings while TKM+ LSK cells failed to form colonies after the 3^rd plating similar to WT LSK cells (Fig 3G).

To validate this finding in vivo, we performed competitive bone marrow transplants by admixing Tet2^−/−, TKM, or WT CD45.2 BMNCs with WT CD45.1 competitor cells in lethally irradiated F1 hybrid recipients as previously described (1) (Fig 3H). While no changes in sssself-renewal were detected in primary transplants, a partial rescue of enhanced self-renewal was observed when comparing Tet2^−/− versus TKM donor cells upon secondary transplantation with fresh CD45.1 competitor cells (Fig 3I). We also observed that recipient mice from primary transplants injected with the Tet2^−/−/WT mixture developed PMP and classical monocyte expansion while there was an absence of these phenotypes in transplant recipients injected with TKM/WT and WT/WT mixtures (Fig 3J) suggesting that Malat1 deletion in BMNCs alone is sufficient to fully attenuate oncogenic inflammatory features of TET2 LOF.

Collectively, these data suggest that Malat1 is necessary for the development of phenotypes dependent on oncogenic inflammatory circuits but not self-renewal in Tet2 LOF mutations.

Malat1 overexpression is sufficient for TET2 LOF phenotypes.

Having confirmed that Malat1 is necessary for these phenotypes in the context of TET2 LOF, we interrogated whether its overexpression alone is sufficient to recapitulate inflammatory phenotypes associated with TET2 LOF. We confirmed Malat1 overexpression (MOE) in a published transgenic mouse model (30) in BMNCs (3.3-fold increase) and splenocytes (2.7-fold increase) (Fig S4A, S4B). Upon immunophenotyping, MOE mice displayed myelomonocytic expansion in both the spleen and peripheral blood relative to control mice at 8 and 20 weeks (Fig 4A – 4E, S4C). They also developed splenomegaly relative to WT mice at both ages (Fig 4F). Finally, similar to Tet2^−/− mice, MOE mice displayed a repartitioning of peripheral blood monocytes and increased numbers of Ly6-C^high monocytes in the spleen in both age groups (Fig 4G, 4H, S4D). Interestingly, although MOE mice had increased LSK numbers in the spleen and bone marrow (Fig 4I) as well as increased MPP3 myeloid progenitors (Fig 4J), no increase in self-renewal capacity was observed in primary or secondary transplants with fresh competitor CD45.1 BMNCs (Fig 4K). However, similar to our Tet2^−/− competitive transplants, we did observe TET2 LOF phenotypes in MOE transplant recipients (Fig S4E) suggesting that exclusive overexpression of Malat1 in BMNCs reproduces oncogenic inflammatory features of TET2 LOF. Profiling of Malat1^−/− mice showed no differences in myelopoiesis or self-renewal capacity (Fig S5A – S5E). These data support that Malat1 overexpression is sufficient to recapitulate most phenotypes associated with oncogenic inflammation in TET2 LOF including the expansion of LSK cells but does not increase self-renewal capacity.

Figure 4. Malat1 overexpression is sufficient for TET2-deficient phenotypes.

Representative flow plot (A) and quantification of CD11b+Gr-1+ myelomonocytic cells in the spleen [(B), n= 11 per group for both age groups) and peripheral blood [(C), n= 8 per group at 8 weeks and 4 per group at 20 weeks] in WT and Malat1 overexpressing (MOE) mice.

(D) and (E) Quantification of CD11b+ mature myeloid cells in the spleen [(D) n= 7 per group at 8 weeks, 6 per group at 20 weeks] and peripheral blood [(E) n= 6 WT and 7 MOE for both age groups]. Data shown are mean ± SEM.

(F) Spleen weights of WT and MOE mice at 8- and 20-weeks n= 11 per group for both age groups. Data shown are mean ± SEM.

(G) Quantification of Ly6-C^high classical monocytes in the spleens of WT and MOE mice n= 7 per group for both age groups. Data shown are mean ± SEM.

(H) Representative flow cytometry plots (top) and quantification (bottom) of monocyte subpopulations in the peripheral blood of WT and MOE mice. n= 8 per group for 8 weeks and 7 per group for 20 weeks. Data shown are mean ± SEM.

(I) LSK count in Spleen (n= 8 per group) and bone marrow (n= 6 per group) of WT and MOE mice at 8 weeks Data shown are mean ± SEM.

(J) Multipotent progenitor (MPP3) counts in the bone marrow of WT and MOE mice (n= 6 per group at 8 weeks and 7 per group at 20 weeks). Data shown are mean ± SEM.

(K) CD45.2 donor chimerism in primary and secondary transplants using bone marrow mononuclear cells from 8-week-old WT and MOE mice. Data shown are mean ± SEM for each time point.

(A)-(K) * p<0.05, ** p<0.01, *** p<0.001 **** p<0.0001 calculated using unpaired t tests.

Malat1 overexpression leads to myeloid skewing in a cell extrinsic manner.

Initial mechanistic insight into how Malat1 overexpression potentiates TET2 LOF inflammatory circuits was obtained when comparing the myeloid composition of competitor and residual host cells in our competitive bone marrow transplant experiments. Surprisingly, we observed that both WT CD45.1 competitor cells and residual WT CD45.1/CD45.2 host cells admixed with MOE or Tet2^−/− CD45.2 cells had a significant expansion of Ly6-C^high monocytes and spleen LSK cells compared to competitor and residual host cells admixed with WT or TKM CD45.2 cells (Fig 5A – 5D). We reasoned that secreted factors elaborated from MOE and Tet2^−/− cells may be responsible for the development of these phenotypes in the WT cells. Given that TET2 LOF produces a pro-inflammatory milieu, we profiled 40 inflammatory cytokines in the plasma of MOE and WT mice (3 mice per group). We observed 7 significantly upregulated pro-inflammatory cytokines [KC (p=0.03), CXCL13 (p=0.02), ICAM1 (p=0.002), CXCL10 (p=0.004), complement 5a (C5a, p=0.04), TNF alpha (p= 0.007) and TREM1 (p=0.005)], and a trend towards upregulation in two other cytokines: IL-16 (p=0.05), and MCSF (p=0.07) when comparing MOE plasma to WT plasma.

Fig 5. Malat1 overexpression leads to myeloid skewing in a cell extrinsic manner.

Analysis of CD45.2 donor, CD45.1 competitor and CD45.1/CD45.2 host compartments in WT/Tet2^−/−/TKM competitive bone marrow transplant recipients. Ly6-C^high classical monocytes in peripheral blood (A) and LSK cells in spleen (B) were assessed by FACS and normalized to total live cells in each compartment. Data shown are mean ± SEM. n= 11 WT and TKM, and 7 Tet2^−/− recipients.

Analysis of CD45.2 donor, CD45.1 competitor and CD45.1/CD45.2 host compartments in WT/MOE competitive bone marrow transplant recipients. Ly6-C^high classical monocytes in peripheral blood (C) and LSK cells in spleen (D) were assessed by FACS and normalized to total live cells in each compartment. Data shown are mean ± SEM. n= 9 WT, 7 MOE. Data shown are mean ± SEM.

(E) Cytokines with significantly different levels in the plasma of 8-week-old WT and MOE mice from 40-plex cytokine array. n= 3 per group. Plasma from each mouse was analyzed independently and all results averaged and expressed as fold change relative to the WT average. Data shown are mean ± SEM.

(F) Cytokines with significantly different levels in the plasma of 8-week-old Tet2^−/− and TKM mice from 40-plex cytokine array. n= 3 per group. Plasma from each mouse was analyzed independently and all results averaged and expressed as fold change relative to TKM. Data shown are mean ± SEM.

(G) Plasma levels of IL-6 basally and upon stimulation with 100ng of LPS for 6 hours in WT and MOE mice. n= 7 per group for basal levels and n= 4 per group for stimulated mice. Data shown are mean ± SEM.

(H-K) GSEA analysis showing positive enrichment of various inflammatory pathway gene signatures.

A-G * p<0.05, ** p<0.01, *** p<0.001 **** p<0.0001.

(A)-(D) p values were calculated using two-way ANOVA. (C) – (F) p value was calculated using unpaired t tests. (G) Each cytokine was assessed independently using a one-way ANOVA.

Comparing plasma cytokine levels between Tet2^−/− and TKM mice using the same cytokine array yielded similar trends [KC (p= 0.001), CXCL13 (p= 0.02), ICAM1 (p=0.01), C5a (p=0.001), IL-16 (p= 0.005), TNF alpha (p= 0.0004) and MCSF (p=0.06)]. There were also significant increases in levels of IL-1a (p= 0.001), IL-13 (p= 0.003), CCL2 (p< 0.000001), MIG (p= 0.000009) and TIMP-1 (p= 0.02). We then compared results from the two groups (MOE vs WT, and Tet2^−/− vs TKM) to identify commonly dysregulated cytokines. Interestingly, we observed seven cytokines which were upregulated in both groups (Fig 5E, 5F) all of which are known downstream targets of NF-Kappa B (NF-κB) (31–36).

We additionally measured IL-6 with a high sensitivity ELISA basally and 6 hours after intraperitoneal injections of 100ng LPS, because baseline levels of IL-6 were expected to be below the limit of detection of our broader 40-plex assay. IL-6, a canonical downstream effector of NF-κB, was also increased in MOE plasma compared to WT plasma at baseline and after LPS stimulation (Fig 5G) analogous to published work on Tet2^−/− mice (8,12).

Ingenuity pathway analysis of differentially expressed genes from our human scRNA-seq data showed an enrichment of inflammatory signatures in COVID-19+/TET2^MT CD14 monocytes relative to COVID-19+/CH− (Fig S6A) consistent with literature on TET2 LOF. Moreover, cytokines increased in the plasma of both MOE and Tet2^−/−mice were increased at the transcript level in Tet2^+/− BMNCs relative to TKM+ (Fig S6B). Consistent with this, bulk RNA-seq data from Tet2^−/− and TKM LSK cells showed an upregulation of gene signatures associated with interferon alpha and gamma responses, LPS stimulation and TNF alpha signaling via NF-κB in Tet2^−/− mice relative to TKM mice (Fig 5H – 5K, S6C), suggesting a previously unrecognized interaction between Malat1 and NF-κB driven oncogenic inflammatory signaling.

MALAT1 expression enhances NF-Kappa B activity and sustains P65 phosphorylation.

To explore the consequence of MALAT1 expression on NF-κB activity we generated CRISPR-deleted MALAT1 knockout (MKO) and Non-target (NT) controls using the DECKO plasmid system (see methods) in parental THP-1 cells and commercially available THP-1 Dual cells expressing an NF-κB activity-linked secreted alkaline phosphatase (SEAP) reporter construct (Fig S7A, S7B). THP-1 was chosen because it is a human monocytic leukemia cell line with the highest levels of MALAT1 observed across 8 cell lines profiled (Fig S7C). We stimulated reporter cells with increasing doses of LPS and TNF alpha and identified a significant decrease in NF-κB activity in MKO cells compared to NT controls (Fig 6A, 6B).

Fig 6: MALAT1 expression enhances NF-Kappa B activity and sustains P65 phosphorylation.

NF-κB activity measured by secreted alkaline phosphatase from THP-1 dual reporter cells stimulated with LPS (A) and TNF alpha (B). Data shown are mean ± SEM of three independent experiments for LPS and two independent experiments for TNF alpha each run with four technical replicates. The area under the dose response curve was calculated for each cell line in each experiment and values compared by one way ANOVA with Tukey’s correction for multiple comparisons.

(C) Time course showing levels of phosphorylated P65 at Ser536 by western blot in THP-1 NT and MALAT1 knockout (MKO) cells stimulated with 20ng/mL of TNF alpha. Data shown are representative of three independent experiments. Vinculin was used as a loading control.

(D) Western blot showing levels of phosphorylated P65 at Ser536 in cytoplasmic and nuclear cellular compartments basally and after stimulation with TNF alpha (15 and 60 minutes after 20ng/mL TNF alpha). Lamin A/C and HSP90 were used as nuclear and cytoplasmic loading controls respectively. Data shown are representative of three independent experiments. Densitometry was calculated for all three experiments and average values relative to unstimulated cells in each subcellular compartment indicated on the western blot.

(E) Time course showing levels of phosphorylated P65 by western blot in pooled WT, Tet2^+/−, and TKM+ BMNCS stimulated with 20ng/mL of TNF alpha. Mice used in experiment were age and sex matched. Vinculin was used as a loading control.

(F) Top left: Heat map displaying percentage of THP1 cells (NT and MALAT1 KO) with nuclear NF-κB P65 localization upon TNF alpha stimulation (20ng/ml for 30 minutes). Top right: Heat map showing similarity morphology median between nuclear image and NF-κB P65 subunit in THP1 cells (NT and MALAT1KO) after TNF alpha treatment. Positive value in THP1 WT cells after TNF alpha treatment corresponds to a high degree of correlation, indicating that NF-κB P65 subunit is largely expressed in the nucleus. Bottom: Representative images displaying nuclear and cytosolic expression of NF-κB P65 in WT and MALAT1KO THP1 cells after TNF alpha stimulation respectively.

(G) Phosphatase (PP2A) activity in THP-1 NT and MKO cells basally and upon stimulation with 20ng/mL TNF alpha for 60 minutes. Data shown are mean ± SEM of two independent experiments.

(H) western blot showing protein levels of the various PP2A subunits involved in NF-κB regulation. Vinculin was used as a loading control.

(I) Western blot showing levels of phosphorylated P65 15 and 60 minutes after stimulation with 20ng/mL of TNF alpha with and without okadaic acid, a PP2A inhibitor. Data shown are representative of two independent experiments.

(J) Densitometry differences between THP-1 NT and MKO with and without okadaic acid at 60 minutes from all experiments in (I). Data shown are mean ± SEM.

(K) RNA immunoprecipitation showing binding between P65 and MALAT1 basally and 30 minutes after stimulation with 20ng/mL of TNF alpha.

(L) Co-Immunoprecipitation experiment showing pulldown of P65 and subsequent immunoblot for PP2A B56 gamma subunit.

(M) Schematic of proposed working model

A-K * p<0.05, ** p<0.01, *** p<0.001 **** p<0.0001 calculated using unpaired t test.

P65 (Rel A) is a key transcription factor in the NF-κB pathway which is activated after phosphorylation by IKK kinase leading to the formation of heterodimers with other transcription factors and translocation to the nucleus where it activates gene expression (37). Given the central role of P65 in NF-κB signaling and the exclusive nuclear location of MALAT1, we measured P65 phosphorylation at activation-associated site Ser536 by immunoblot analysis in our in vitro models. Time course experiments demonstrated that MKO cells had a more rapid decrease in phosphorylated P65 with preserved total P65 protein levels relative to their NT controls after stimulation with both TNF alpha and LPS (Fig 6C, S7D, S7E). Consistent with the nuclear location of MALAT1, the observations made in whole cell lysates were seen in nuclear but not cytoplasmic fractions of these cells (Fig 6D). To validate the association between Malat1 and NF-κB in the context of TET2 LOF, we measured protein levels of phosphorylated P65 in BMNCs from our Tet2^+/− and TKM+ mouse models and observed a decrease in phosphorylated signal with Malat1 depletion similar to that seen in our human in vitro models (Fig 6E).

To further assess activation of P65 in our in vitro models, we measured nuclear translocation of P65 30 minutes after stimulation with TNF-alpha using a multispectral, imaging flow cytometer that enables differentiation of nuclear and cytoplasmic protein expression in thousands of cells (38). Nuclear translocation is a complex mechanism influenced by the phosphorylation state of P65 (39). Using this approach, we observed a decrease in nuclear translocation of P65 in MKO cells relative to controls as would be predicted by our observation that MALAT1 deletion decreases P65 activation-associated phosphorylation (Fig 6F, S7F).

GSEA from our murine LSK RNA-seq experiment also revealed positive enrichment of genes upregulated by P65 activity, and transcriptional targets of NF-κB in Tet2^−/− relative to TKM even in HSPCs (Fig S7G).

While IKK kinase is primarily responsible for the initial phosphorylation of P65, protein phosphatase 2A (PP2A), a ubiquitous cellular phosphatase is primarily responsible for P65 dephosphorylation (40). PP2A is a complex heterotrimeric enzyme made up of three proteins: the scaffolding subunit A, the regulatory subunit B which determines substrate specificity, and the catalytic subunit C (40).

Given that we observed no difference in initial phosphorylation in THP-1 MKO cells versus control, but a rapid loss of phosphorylated signal after activation, we reasoned that MALAT1 may prevent dephosphorylation of P65 by PP2A leading to its sustained phosphorylation and activation. To test this, we measured PP2A activity in THP-1 NT and MKO cells basally and upon stimulation with TNF alpha. We demonstrated that MKO cells had increased phosphatase activity and a higher vmax when stimulated with TNF alpha (Fig 6G, S7H) despite having comparable protein levels of PP2A subunits (Fig 6H). Further, treating THP-1 MKO cells with LB100 and Okadaic acid, two PP2A inhibitors, restored the kinetics of phosphorylated P65 in MKO cells to that of its NT control (Fig 6I, 6J, S7I). To functionally validate the role of PP2A in MALAT1-dependent NF-κB activity, we profiled gene expression of 5 cytokines downstream of NF-κB with and without a PP2A inhibitor and observed an initial decrease in gene expression in MKO cells that was rescued by inhibition of PP2A (Fig S7J, S7K).

We next hypothesized that the disruption of PP2A activity on P65 may be due to physical interference by MALAT1. RNA immunoprecipitation of total P65 basally and 30 minutes after stimulation with TNF alpha enriched MALAT1 in stimulated THP-1 NT cells but not in unstimulated cells suggesting that a physical interaction occurs at a time when P65 is undergoing nuclear translocation (41) (Fig 6K). We also generated a doxycycline-inducible system in which a biotin-like RNA aptamer was attached to the MALAT1 sequence and overexpressed in THP-1 cells (see methods). Streptavidin mediated pulldown of this MALAT1 construct showed an enrichment of total P65 in induced cells relative to uninduced with TNF alpha stimulation (Fig S7L, S7M).

Finally, co-immunoprecipitation experiments revealed greater interaction between P65 and the B56 regulatory subunit of PP2A in the absence of MALAT1 compared to controls (Fig 6L). Time course measurements revealed no change in MALAT1 expression up to 24 hours after TNF alpha stimulation (Fig S7N) suggesting that these functional experiments were not impacted by changes in MALAT1 levels. Collectively, these data suggest that MALAT1 depletion enhances PP2A activity on P65 and is likely responsible for the phosphorylation differences observed in MKO versus NT cells (Fig 6M).

Interleukin 6 (IL-6) is necessary for the effects of Malat1 overexpression.

Exogenous inflammatory stimuli leading to the elaboration of IL-6 has been demonstrated as a critical initiating factor of the oncogenic inflammatory circuit in Tet2 LOF mutations (8,12). Having established that MALAT1 expression is associated with increased NF-κB activity due to disruption of PP2A activity, in addition to the upregulation of IL-6 in our in vivo and in vitro models, and the known role of IL-6 in initiating TET2 LOF oncogenic inflammation, we sought to determine if inhibition or deletion of IL-6 signaling was sufficient to attenuate the TET2 LOF inflammatory phenotypes seen in MOE mice. To test the role of IL-6 in MOE mice pharmacologically, we profiled MOE and WT mice to confirm expansion of CD11b+Gr-1+, CD11b+ cells and Ly6-C^high monocytes in the peripheral blood at 10 weeks. After confirmation, mice from both groups were treated with weekly injections of either a neutralizing IL-6 antibody or control IgG for three weeks as previously described (8) (Fig 7A). After the injections, mice in both groups were sacrificed and CD11b+, CD11b+Gr-1+ and Ly6-C^high cells were profiled in the blood and spleen. Our assessment demonstrated an attenuation of all phenotypes in the peripheral blood of MOE mice treated with IL-6 neutralizing antibody while an expansion of these phenotypes was observed in the mice treated with control IgG (Fig 7B – 7D).

Fig 7: Interleukin 6 is necessary for the effects of Malat1 overexpression.

(A) Schematic for in vivo il-6 neutralization experiment

Pre (day 0) and post (day 28) IL-6 neutralizing antibody/IgG treatment assessment of peripheral blood CD11b+Gr-1+ cells (B), mature CD11b+ cells (C) and Ly6-C^high classical monocytosis (D) in WT and MOE mice. Data shown are mean ± SEM.

CD11b+Gr-1+ myelomonocytic cells (E), CD11b+ myeloid cells (F) and Ly6-C^high classical monocytes (G) in spleens of all mice used in IL-6 neutralizing experiment on day 28. Data shown are mean ± SEM.

(H) Spleen weights of WT, MOE, and MI-6 mice at 8 weeks. Data shown are mean ± SEM.

(I) CD11b+ cells in spleens of WT, MOE, and MI-6 mice at 8 weeks. Data shown are mean ± SEM.

(J) CD11b+Gr-1+ cells in the spleens of WT, MOE, and MI6 mice at 8 weeks. Data shown are mean ± SEM.

(K) Ly6-C^high classical monocytes in spleens of WT, MOE, and MI6 mice at 8 weeks. Data shown are mean ± SEM.

(L) Monocyte subpopulations in peripheral blood of WT, MOE, and MI6 mice at 8 weeks. Data shown are mean ± SEM.

(M) Lineage-Sca+ckit+ LSK cells in the spleens of WT, MOE and MI6 mice at 8 weeks Data shown are mean ± SEM

(A)-(K) * p<0.05, ** p<0.01, *** p<0.001 **** p<0.0001 calculated using one-way ANOVA with Tukey’s correction for multiple comparisons. n= 3 per group for IgG treatment and 3 WT, 4 MOE for IL-6 neutralizing antibody treatment in (A)-(F). n= 11 WT, 9 MOE and 7 MI-6 for all experiments in (G)-(K).

Endpoint assessment of the spleens showed MOE mice treated with neutralizing antibody had similar numbers of CD11b+Gr-1+, CD11b+ myeloid cells and Ly6-C^high monocytes as WT mice while MOE mice treated with control IgG had expansion of these populations (Fig 7E – 7G). Neutralizing antibody treatment appeared to have no significant effect on WT mice. We validated this observation genetically by crossing our MOE mouse to the previously published Il-6^−/− mouse (42) generating a Malat1 overexpressing/Il-6 knockout mouse model (MI-6), and measured myeloid, myelomonocytic and monocytic compartments at 8 weeks, a time point at which PMP and classical monocytosis was evident in MOE mice. Profiling of MI-6 mice showed they had smaller spleens (Fig 7H) and no expansion of CD11b+ (Fig 7I), CD11b+Gr-1+ cells (Fig 7J) or Ly6-C^high monocytes (Fig 7K) in the spleen. They also had comparable proportions of monocyte subsets in the peripheral blood (Fig 7L), and similar numbers of LSK cells in the spleen as WT mice (Fig 7M). This suggests IL-6 is necessary for the development of oncogenic inflammatory signals and phenotypes in MOE mice.

Discussion

Recent studies have demonstrated that aberrant potentiation of inflammation by clonally expanded monocytes and macrophages contributes to the myeloid transformation of TET2 LOF mutations as well as the establishment of premalignant conditions such as CH (8–10,12). Moreover, TET2 LOF-dependent myeloid skewing and aberrantly sustained inflammation promote the development of non-hematologic clinical sequelae such as cardiovascular disease, chronic obstructive pulmonary disease, diabetes mellitus, chronic liver disease and severe microbial infections in CH (4–6,8,14). Therefore, understanding the molecular mechanisms responsible for this myeloid skewing and enhanced inflammation, as well as developing strategies to disrupt this process are critically important to mitigate TET2 LOF-driven human disease. The present study nominates Malat1 as a central downstream effector for myeloid skewing, splenomegaly, expansion of HSPCs, and enhanced NF-κB-driven inflammation in mature Tet2 mutated hematopoietic cells. NEAT1 enrichment was also identified in our human scRNA-seq data but deprioritized because TET2 LOF was not sufficient to increase NEAT1 levels in our isogenic models. Mitochondrial gene upregulation was also observed in our samples likely because these cells were under significant cellular stress as they were harvested during severe COVID-19 infection.

Malat1 depletion resulted in a partial rescue of enhanced self-renewal in competitive whole bone marrow transplants with Tet2^−/− BMNCs and a full rescue of enhanced self-renewal in serial replating assays with sorted Tet2^+/− LSKs. However, we observed a competitive disadvantage in similar transplants with MOE BMNCs without TET2 LOF. This suggests that while Malat1 is both necessary and sufficient for the oncogenic inflammatory features of TET2 LOF seen in progenitors and terminally differentiated cells, the increased self-renewal capacity of TET2 deficient HSPCs is regulated by a complex mechanism that may include Malat1 overexpression but also may involve other downstream effectors as seen in previous studies (12). While it is known that increased inflammation is detrimental to self-renewal capacity of stem cells (43), TET2 deficient HSPCs upregulate compensatory mechanisms such as the evasion of apoptosis that enable them to have a competitive advantage in inflammatory states (12). Collectively, this suggests that Malat1 overexpression in terminally differentiated myeloid cells results in a favorable inflammatory microenvironment that can be harnessed by TET2 mutant HSPCs, but that Malat1 is not solely responsible for a cell intrinsic fitness or self-renewal increase in these cells.

Importantly, our data suggest the exciting possibility that MALAT1 may directly “shield” P65 from PP2A and prevent its dephosphorylation. This is supported by our biochemical experiments that detected a physical interaction between MALAT1 and P65 as well as an increase in the P65 and PP2A interaction in the absence of MALAT1 after TNF alpha stimulation. This mechanism also suggests that P65 can function as an RNA binding protein, a phenomenon which has not been previously described. RNA “shielding” of P65 from dephosphorylation by MALAT1 is also consistent with the existing literature demonstrating that the potentiation of oncogenic inflammation associated with TET2 LOF results from a failure of resolution rather than an increase in its initiation (4,44,45).

Our study may also provide mechanistic insights into the numerous observations linking MALAT1 expression to inflammatory conditions such as COVID-19 infection, autoimmune diseases, solid tumor malignancies, and inflammaging (46,47). While not explored in these studies, the direct MALAT1-dependent shielding of P65 in inflammatory diseases may be a central mechanism governing how inflammation impacts hematopoiesis even in the absence of TET2 LOF. This is consistent with our observation that Malat1 overexpression alone phenocopies inflammatory hematopoiesis in vivo, a phenomenon characterized by myeloid skewing and HSPC expansion with reduced self-renewal capacity (48,49) and that severe COVID-19 infection was associated with increased MALAT1 in wild type human peripheral blood cells (Fig 1B).

Due to increasing recognition of its relevance in multiple disease processes, efforts in developing targeted therapy against MALAT1 have increased in the last decade. Early anti-MALAT1 strategies included Anti-sense oligonucleotides (ASOs), siRNAs and antisense gapmers (50). Gapmers are RNA-DNA-RNA hybrids which bind to complimentary RNA sequences and promote their degradation by recruiting RNaseH (51). Recent efforts have also focused on the development of small molecules targeting MALAT1 given the inherent advantage for clinical use of small molecules compared to ASOs. These small molecules include synthetic compounds such as compound 5, niclosamide and Tyrphostin 9, as well as naturally occurring flavonoids such as resveratrol, quercetin, and curcumin (50). While most of these small molecules target the 3’ UTR triple helix to reduce MALAT1 levels through increased degradation, compounds such as pyridostatin (PDS) aim to mitigate the effects of MALAT1 by disrupting potentially critical interactions with other molecules via its conserved RNA G-quadruplexes which are necessary for some mRNA functions (50).

In conclusion, we identify Malat1 overexpression as a critical transcriptional event in TET2 LOF which leads to increased NF-κB activity and elaboration of pro-inflammatory cytokines via disruption of PP2A activity (Fig 6M), resulting in HSPC expansion, myeloid-biased hematopoiesis, PMP, and the establishment of an oncogenic inflammatory circuit. This observation highlights that while genome wide epigenetic changes occur in these cells, discrete targetable effectors can be identified. MALAT1 can be targeted with anti-sense oligonucleotides and small molecules and therefore our observation additionally provides a viable clinically relevant strategy to attenuate myeloid skewing and enhanced inflammation across TET2 LOF-driven disease. Moreover, the fact that genetic deletion of Malat1 is tolerated in murine models (52) and no impact on normal myelopoiesis or self-renewal capacity of HSPCs was noted in our investigations (Fig S5A – S5E) suggests that a suitable therapeutic window may be present in humans. Future studies will aim to refine existing small molecule tool compounds so that these, as well as clinical grade ASOs, could be tested in human clinical trials.

Methods

Patients and sample collection

Studies involving COVID-19 and CHIP patients were carried out at the Mayo Clinic in Rochester, Minnesota according to the principles of the Belmont Report, with approval from the Mayo Clinic Institutional Review Board (IRB- 15–003786). In all cases, diagnosis was according to the 2016 iteration of the WHO classification of myeloid malignancies (53). Peripheral blood and bone marrow samples from patients were collected in EDTA tubes after written informed consent.

Studies on MDS patients were carried out at the Moffitt Cancer Center in Tampa, Florida according to the 2018 US Common Rule (45 CFR 46), with approval from the University of South Florida Institutional Review Board (IRB 19508). Peripheral blood and bone marrow were collected in heparin tubes after written informed consent. In all cases, diagnosis was according to the 2016 iteration of the WHO classification of myeloid malignancies (53).

Mouse models

All mice were bred and maintained under specified pathogen-free (SPF) conditions in the animal facility at Moffitt Cancer Center, University of South Florida with 12-hr light/dark cycle and were provided food and water ad libitum.C57BL/6J (Strain# 000664; RRID:IMSR_JAX:000664), Tet2^+/− (B6(Cg)-^Tet2tm1.2Rao/J Strain #:023359; RRID:IMSR_JAX:023359), Il6^−/− (B6.129S2-Il6t^m1Kopf/J stock Strain #:002650; RRID:IMSR_JAX:002650) and B6.SJL-Ptprc^a Pepc^b/BoyJ (Boy J CD45.1, Strain #:002014; RRID:IMSR_JAX:002014) were purchased from The Jackson Laboratory. Malat1^Tg (MOE) mice were a generous gift from Li Ma from the MD Anderson Cancer Center and Malat1^−/− (B6.Cg-Malat1<tm1.1Shna>/ShnaRbrc, BRC No. RBRC06303, Acc. No. CDB0837K, RRID:IMSR_RBRC06303) mice were obtained from the RIKEN Bioresource Center through the National Bio-Resource Project of the MEXT, Japan. All mice purchased from the Jackson laboratory were allowed to acclimate for at least two weeks before being used in experiments or for breeding. Male mice 8 to 20 weeks were used in all experiments. Mouse experiments in this study were performed in accordance with protocols approved by the Institutional Animal Care and Use Committee (IACUC) of the University of South Florida.

Established cell lines

THP-1 (TIB 202, male RRID:CVCL_0006), TF-1 parental (CRL-2003, male RRID:CVCL_0559) TF-1 IDH2^R140Q isogenic (CRL-2003IG, male RRID:CVCL_UE10), HEL (TIB-180, male RRID:CVCL_0001), U-937(CRL-1593.2, male RRID:CVCL_0007), HEK 293T (CRL-3216, female RRID:CVCL_0063) HL-60 (CCL 240, female RRID:CVCL_0002) cell lines were purchased from ATCC. SKM-1 (DSMZ, ACC547, male RRID:CVCL_0098), SET2 (DSMZ, ACC608, female RRID:CVCL_2187), and OCI-AML 3 (DSMZ, ACC582, male RRID: CVCL_1844) cell lines were generous gifts from Seongseok Yun. THP1-Dual reporter cell lines were purchased from Invivogen (Cat# thpd-nfis, male; RRID:CVCL_X599) and cultured in RPMI 1640, 2 mM L-glutamine, 25 mM HEPES, 10% fetal bovine serum, 100 μg/ml Normocin, Pen-Strep (100 U/ml-100 μg/ml) according to manufacturer’s instructions. Positive antibiotic selection was done every other passage with 10ug/ml Blasticidin and 100ug/ml Zeocin according to manufacturer’s specifications. All cell lines from ATCC were cultured at 37°C with 5% CO₂ on the basis of ATCC guidelines: (RPMI 1640 media supplemented with 10% fetal bovine serum (FBS) (Sigma-Aldrich Cat# F0926). Growth medium for TF-1 was also supplemented with 2 ng/mL of GM-CSF (R&D Systems Cat# 7954-GM-010). OCI-AML3 cell line was cultured in alpha-MEM with 10% FBS. All cell lines were regularly tested for mycoplasma (every 4 weeks), passaged when 70 – 80% confluent and split 1:10. Cells were grown for no longer than twenty passages.

Generation of modified THP-1 cell lines.

pL-CRISPR.SFFV.tRFP (Addgene plasmid 57826, RRID:Addgene_57826) and pL-CRISPR.SFFV.GFP (Addgene plasmid 57827, RRID:Addgene_57827) were donations from Benjamin Ebert (purchased from Addgene). Primers for gRNA sequences flanking the major TSS of MALAT1 (54), as well as non-specific control gRNAs, were purchased from Integrated DNA Technologies. For both MALAT1 and non-target, gRNA 1 was cloned into pL-CRISPR.SFFV.tRFP, and gRNA 2 was cloned into pL-CRISPR.SFFV.GFP as previously described (55). Briefly, each plasmid vector was digested with BsmBI restriction enzyme (New England Bio Cat# R0580) and then gel purified. The forward and reverse primers were mixed in a 1:1 ratio and phosphorylated with T4 PNK (New England Bio Cat# M0201). Primers were then ligated to the purified vector and the ligation product was used to transform Stbl3 competent cells (Invitrogen Cat# C737303). The resulting plasmids were sequenced to confirm the presence of the insert. Lentivirus for transduction of constructs was produced by mixing 5.5 μg of gRNA plasmid, 150 μl Lipofectamine 2000 (Invitrogen Cat# L3000150), 130 μl Mission Lentiviral Packaging Mix (Millipore Sigma Cat# SHP001), and 2500 μl Opti-MEM I medium (Gibco Cat# 31985070). After 15 minutes, the mixture was added to 70% confluent HEK-293T in a 100 mm dish which already contained 4 mL Opti-MEM I. After 6 hours, 6 mL of DMEM supplemented with 10% FBS was added to the dish. Medium was changed the next day. Virus was collected at 48 hours and 72 hours and filtered with a 0.45 μm syringe filter. Supernatants from 48 and 72 hours were combined, and the virus was concentrated by ultracentrifugation at 24000 rpm for 2 hours at 4°C and resuspension in 1/100 the original volume. For both MALAT1 and non-target, THP-1 cells were transduced with a 1:1 mixture of gRNA1 (tagRFP) and gRNA2 (GFP) constructs. 72 hours after transduction, cells positive for both GFP and tagRFP were sorted by FACS. Double-positive cells were then plated into 96-well plates, one cell per well. Wells were screened for TSS deletion or inversion by PCR as previously described (54). gRNAs for knocking out MALAT1 are listed in Table S1. MALAT1 knockdown was confirmed in positive clones by qRT-PCR for MALAT1 across the entire full-length transcript using SYBR primers (Integrated DNA Technologies) or Taqman assays listed in table S2.

Creation of Immortalized Murine Cells

Bone marrow cells harvested from Tet2^−/− and C57BL/6 experimental mice were immortalized for in vitro differentiation assays as previously described (56). Briefly, cells were transduced with an estrogen-regulated Hoxb8-containing retrovirus and maintained in RPMI supplemented with 10% FBS, 50 ng/mL m-SCF, and 0.5 μM beta-estradiol.

qRT-PCR

Total RNA was extracted using the Quick-RNA miniprep kit (Zymo Research Cat# R1055). 1ug of total RNA for each sample was reverse transcribed to cDNA using the High-Capacity RNA-to-cDNA^™ Kit (Thermo Fisher Scientific Cat# 4387406) and RT-PCR done with Applied Biosystems PowerUp SYBR Green Master Mix (Thermo Fisher Scientific, Cat# A25777). Each qRT-PCR experiment was carried out in at least three biological replicates each with three technical replicates per sample. Cytokine expression in WT, Tet2^+/− and TKM+ mice was assessed using Applied Biosystems TaqMan Fast Advanced Master Mix (Thermo Fisher Scientific Cat# 4444963). For cytokine gene expression in THP-1 NT and MKO cells, Cells were pretreated with DMSO or Okadaic acid (50nM) for three hours before stimulating with 100ng/mL of LPS for 15 hours. 18s was used as a reference gene for experiments comparing MALAT1 and NEAT1 expression in wild type and TET2 LOF groups. Primer sequences for various transcripts as well as cytokines are listed in table S3. Taqman assays for assessing murine cytokine gene expression are listed in Table S4. Gapdh was used as the reference gene for all qRT-PCR experiments involving murine samples.

Flow cytometry

Staining was performed in 5 mL polystyrene test tubes with 2 washes per step. Red blood cells in bone marrow, spleen and peripheral blood samples were lysed by incubating the samples for 5 minutes in Ammonium Chloride Potassium (ACK) lysis buffer. Cells were then washed and stained with LIVE DEAD violet fixable dye (Thermo Fisher, Cat# L34964) for thirty minutes after which they were fixed with 1.6% formaldehyde for ten minutes at room temperature. Cells were then washed and incubated with mouse FcR blocking reagent (Miltenyi Biotec Cat# 130-092-575, RRID:AB_2892833) for ten minutes after which they were incubated with fluorophore-conjugated antibodies for thirty minutes in BD brilliant Stain Buffer (BD Biosciences Cat# 566349, RRID:AB_2869750) and washed. For panels that included biotin labelled antibodies, cells were then stained with Brilliant Violet 711 streptavidin (BioLegend Cat# 405241) for twenty minutes and washed again before acquisition. Antibodies used are as follows: APC-Cy7 CD45.2 (BD Biosciences Cat# 560694, RRID:AB_1727492), BUV737 CD45.1(BD Biosciences Cat# 564574, RRID:AB_2738850), FITC CD45.1 (BioLegend Cat# 110706, RRID:AB_313495), Brilliant Violet 711 CD19 (BioLegend Cat# 115555, RRID:AB_2565970), BUV 395 CD11b (BD Biosciences Cat# 563553, RRID:AB_2738276), PE-Cy7 CD115 (BioLegend Cat# 135524, RRID:AB_2566460), Brilliant Violet 510 Gr-1 (BD Biosciences, Cat# 56304 RRID AB_2722496), Alexa Fluor 488 Ly6-C (BioLegend Cat# 128022, RRID:AB_10639728) BUV395 Sca1(BD Biosciences Cat# 563990, RRID:AB_2738527) CD117/ckit (BioLegend Cat# 135106, RRID:AB_1877215), PE Dazzle CD48 (BioLegend Cat# 103438, RRID:AB_2650823) Brilliant Violet 785 CD150 (BioLegend Cat# 115937, RRID:AB_2565962), APC CD135 (BD Biosciences Cat# 560718, RRID:AB_1727425) Alexa Fluor 700 CD34 (BD Biosciences Cat# 560518, RRID:AB_1727471), biotin anti-mouse lineage panel (CD3, Gr-1, CD11b, B220 and Ter119) (BioLegend Cat# 133307, RRID:AB_11124348). The following gating strategies were used: monocyte subpopulations: single cells were gated on using FSC-A, SSC-A, FSC-W and SSC-W. Live cells were gated from single cells. CD45.2 positive cells were gated on and subsequently Lineage negative cells were gated on (CD3−, Ter119− CD19−, B220−). Monocytes defined as CD11b+CD115+ were selected and then grouped into classical, non-classical and intermediate monocytes using their expression of Ly6-C (Fig S4D). Myelomonocytic cells were defined as CD11b+Gr-1+ Lineage negative cells (Fig S4C). LSK analysis: Live, single CD45.2+ cells were gated on as described prior. Lineage negative cells (CD3−, Ter119−, Gr-1-, CD11b−, B220−, CD19−) were then selected and CD117 (ckit)+ SCa1+ cells were gated on lineage negative cells (LSK). CD48, CD150, CD34 and CD135 were then used to define multipotent progenitor populations (MPP3: Lin− ckit+ Sca1+ CD48+ CD150− CD135− CD34+).

Cell Sorting for RNA sequencing

Femurs and Tibias were harvested from 8-week-old male mice and bone marrow was extracted by centrifugation. Red blood cells were lysed with ACK buffer for five minutes after which cells were washed with PBS. Lineage negative cells were then selected with LS columns (Miltenyi Biotec 130-042-401) using a direct lineage cell depletion kit (Miltenyi Biotec 130-110-470) following the manufacturer’s instructions. Zombie NIR (BioLegend Cat# 423105) was used as viability stain. Cells were then incubated with mouse FcR blocking reagent for ten minutes, stained with BUV395 Sca1 (BD Biosciences Cat# 563990, RRID:AB_2738527), BB700 CD117 (BD Biosciences Cat# 566414, RRID:AB_2744269) and sorted by FACS. Sorted LSK cells were washed with PBS and total RNA was extracted with Qiagen RNeasy micro kit (Qiagen Cat# 74004). RNA samples were processed with the Takara SMARTer Stranded Total RNA-Seq Kit v2 Pico and sequenced in one NovaSeq-6000 SP-200 sequencing run (101×101).

Colony Formation Assays

Sca1⁺ ckit⁺ LSK cells were sorted from bone marrow of C57BL6, Tet2^+/− and Tet2^+/−; Malat1^−/− (TKM+) mice as described above and seeded at a density of 2000 cells/replicate into cytokine supplemented methylcellulose medium (Methocult M3434, Stem Cell Technologies Cat# 03434) in 6 well Smart dish (Stem Cell Technologies Cat# 2737). Colonies propagated in culture were scanned and scored at day 7 using STEMvision (Stem Cell Technologies Cat# 22008) and re-plated (10,000 cells/replicate) for a total of 5 platings (7, 14, 21, 28 and 35 days).

Competitive Bone Marrow Transplants (BMT)

For primary competitive BMT, freshly dissected femurs and tibias were isolated from C57BL/6, MOE, Malat1^−/−, Tet2^−/− and TKM donor mice as well as Boy J CD45.1 mice. Bone marrow was isolated by centrifugation of cut bones and red blood cells were lysed in ammonium chloride-potassium bicarbonate (ACK) lysis buffer for 5 minutes and washed. After centrifugation, cells were re-suspended in sterile 0.9% saline, filtered through a cell strainer and counted. CD45.2 donor BM cells from donor mice were mixed with BoyJ CD45.1 competitor BM donor cells in a 1:1 ratio. Recipient animals (F1, CD45.2/CD45.1) were lethally irradiated (two doses of 500 rads, 4 hours apart) on the day of transplant. 1 million cells from the 1:1 CD45.2/CD45.1 mixture was injected intravenously (via tail vein or retro-orbital injection) into recipient mice. For secondary competitive BMT, CD45.2 donor BM cells from primary competitive BMT recipients were negatively selected using the MojoSort Mouse CD45.1 Selection Kit (BioLegend Cat# 480017) and then mixed with fresh BoyJ CD45.1 competitor BM cells in a 4:1 ratio. 5 million total cells were injected into secondary recipients (4 ×10⁶ CD45.2 and 1 ×10⁶ fresh competitor). Chimerism analysis for progressive engraftment was done on peripheral blood obtained by submandibular bleeds every 4 weeks using APC-Cy7 CD45.2 (BD Biosciences Cat# 560694, RRID:AB_1727492) and FITC CD45.1 (BioLegend Cat# 110706, RRID:AB_313495). At endpoint recipient mice were euthanized and bone marrow, spleen and peripheral blood were analyzed by FACS. CD45.1 population in endpoint studies was determined using BUV737 CD45.1 (BD Biosciences Cat# 564574, RRID:AB_2738850)

Western blotting

20ng/mL of recombinant human TNF alpha (Stem Cell Technologies Cat# 78068) and 100ng/mL Lipopolysaccharides (LPS) (Sigma Aldrich Cat# L5293–2mL) were used for stimulation in all western blot experiments. Cells were treated with 4uM of LB100 (Selleck Chemicals Cat# S7537) or 50nM of Okadaic acid (Thermo Fisher Scientific Cat#J60155.IAM) for 3 hours prior to stimulation in PP2A inhibitor rescue experiments. Cells were stimulated for the stated periods of time then spun down and washed in ice cold PBS. Cells were then lysed in RIPA buffer with Halt protease and phosphatase inhibitors (Life Technologies, Cat# 78440) on ice for 30 minutes followed by centrifugation at 14,000 rpm for 15 minutes at 4°C.

Lysed samples were normalized by protein concentration determined using the Pierce BCA assay (Fisher Scientific Cat# 23227). Normalized samples were denatured in Laemmli Sodium dodecyl sulfate loading buffer at 100°C for five minutes then loaded onto bis-tris gels (Bio-Rad Cat# 1610173). Transfer to PVDF membrane (Fisher Scientific Cat# IPVH00010) was done overnight after which membranes were blocked with 5% nonfat milk (Genesee Scientific Cat# 20–241) in TBST for one hour. Primary incubation was done overnight at 4°C with a 1:1000 dilution after which appropriate secondary antibody (Jackson ImmunoResearch Labs Cat# 115-035-003, RRID:AB_10015289 or Jackson ImmunoResearch Labs Cat# 111-035-003, RRID:AB_2313567) was incubated at 1:2000 for an hour at room temperature. Films were then developed using a film processor for radiography films. Primary antibodies used are as follows: Phospho-NF-κB p65 (Ser536) (93H1) Rabbit mAb (Cell Signaling Technology Cat# 3033, RRID:AB_331284), NF-κB p65 (D14E12) XP Rabbit mAb (Cell Signaling Technology Cat# 8242, RRID:AB_10859369), total IκBα (44D4) Rabbit mAb (Cell Signaling Technology Cat# 4812, RRID:AB_10694416), Phospho-IκBα (Ser32/36) (5A5) Mouse mAb (Cell Signaling Technology Cat# 9246, RRID:AB_2267145), PP2A C Subunit Antibody (Cell Signaling Technology Cat# 2038, RRID:AB_2169495), PP2A A Subunit (81G5) Rabbit mAb (Cell Signaling Technology Cat# 2041, RRID:AB_2168121), HSP90 (C45G5) Rabbit mAb (Cell Signaling Technology Cat# 4877, RRID:AB_2233307) Lamin A/C Lamin A/C Antibody (Cell Signaling Technology Cat# 2032, RRID:AB_2136278), and EGR1 (15F7) Rabbit mAb, (Cell Signaling Technology Cat# 4153, RRID:AB_2097038). Mouse monoclonal Anti-Vinculin antibody (Sigma-Aldrich Cat# V9131, RRID:AB_477629), Monoclonal Anti-β-Actin antibody (Sigma-Aldrich Cat# A5441, RRID:AB_476744), and PP2A-B56-γ Antibody (F-10) (Santa Cruz Biotechnology Cat# sc-374380, RRID:AB_10989746).

RNA immunoprecipitation

THP-1 cells (RRID:CVCL_0006) were stimulated with 20ng/mL of TNF alpha for thirty minutes. Stimulated and non-stimulated cells were washed with PBS and crosslinked in 1% formaldehyde for 10 mins at room temp. Cells were then quenched with 0.125M glycine for 5 mins, pelleted and washed with PBS. Cell pellets were resuspended in 2ml PBS, 2 ml nuclear isolation buffer (1.28 M sucrose, 40 mM Tris pH 7.5, 20 mM MgCl2, 4% Triton X-100) and 6ml water. The mixture was kept on ice for 20 mins with frequent mixing after which cells were pelleted by centrifugation at 2,500 × g for 15 mins at 4°C. Cell pellets were resuspended in 1ml RIP buffer (150 mM KCl, 25 mM Tris pH 7.4, 5 mM EDTA, 0.5 mM DTT, 0.5% NP40, 100 U/ml RNAse inhibitor and Protease inhibitors)

Chromatin was sheared by sonication and mixture spun at 13,000 rpm for 15 mins at 4°C to pellet membranes and debris. 5ug of NF-κB p65 (D14E12) XP Rabbit mAb (Cell Signaling Technology Cat# 8242, RRID:AB_10859369) or Rabbit (DA1E) mAb IgG XP^® Isotype Control (Cell Signaling Technology Cat# 3900, RRID:AB_1550038) was added and samples were incubated overnight at 4°C. Protein A/G magnetic beads (Thermo Fisher Scientific Cat# 88802) were then added, and samples incubated for an hour at 4° with rotation. Bead were washed 5 times in RIP buffer then resuspended in 100ul RIP buffer and incubated at 70°C for 1 hour to reverse crosslinks. Beads were then resuspended in 100 ul Proteinase K buffer (10 mM Tris-HCl, pH 8.0, 1 mM EDTA, 0.5% SDS) and 5ul Proteinase K and incubated at 65°C for 1 hr. RNA was isolated using RNeasy micro kit (Qiagen Cat# 74004) according to manufacturer’s instructions. cDNA conversion and qRT-PCR were carried out as described above and results for basal and stimulated samples were normalized to their respective IgG samples.

Chromatin Immunoprecipitation (ChIP)

Chromatin Immunoprecipitation was done using the SimpleChIP^® Plus Sonication Chromatin IP Kit (Cell Signaling Technology Cat #56383) according to manufacturer’s instructions. Briefly, 2 ×10⁷ cells were crosslinked in 1% formaldehyde for ten minutes and then quenched with glycine for 5 minutes. Cells were then centrifuged at 4°C for 5 minutes at 1000 × g and washed twice with PBS. Cell pellets were then lysed with sonication lysis buffer and sonication nuclear lysis buffer after which chromatin was fragmented by sonication. After confirmation of sufficient fragmentation by gel electrophoresis, samples were diluted in ChIP buffer and divided into two portions. One half was incubated with 2ug of Rabbit IgG (Cell Signaling Technology Cat# 2729, RRID:AB_1031062) and the other with EGR1 antibody (Cell Signaling Technology Cat# 4154, RRID:AB_2097035) overnight at 4°C with rotation. Magnetic beads were then added and allowed to incubate for two hours with rotation. Beads were then washed three times with low salt wash buffer and once with high salt wash buffer each wash taking 5 minutes. DNA was then eluted in elution buffer at 65°C for thirty minutes with gentle vortexing after which magnetic beads were removed and crosslinks in eluted DNA were reversed by heating at 65°C for two hours with proteinase k. DNA was then purified using supplied columns and buffers according to manufacturer’s instructions and binding assessed by qRT-PCR using PowerUP SYBR green master mix. Histone H3 (Cell Signaling Technology Cat# 4620, RRID:AB_1904005) was used as a positive control in all experiments. Primer sequences for targeted regions on MALAT1 are listed in table S5.

Co-Immunoprecipitation

THP-1 NT and MKO cells were stimulated with 20ng/mL TNF alpha for thirty minutes and then washed with PBS. Unstimulated cells were incubated in normal growth medium. Cells were then lysed with Pierce IP Lysis buffer (Thermofisher scientific Cat# 87787) containing protease and phosphatase inhibitors for 15 minutes then spun at 14000rpm for 15 minutes. Protein concentrations were quantified using Pierce BCA assay and lysates precleared with Protein A/G magnetic beads. 1mg of protein for each condition was incubated with either NF-κB p65 (D14E12) XP Rabbit mAb (Cell Signaling Technology Cat# 8242, RRID:AB_10859369) or control IgG overnight after which Protein A/G magnetic beads were added for two hours. The beads were then washed with IP lysis buffer five times and bound proteins eluted. Proteins pulled down were assessed by western blotting as described prior.

MALAT1 pulldown assay

The system was generated by cloning the published optimized RNA aptamer S1 (4xS1) (57) to the full length MALAT1 transcript. Lentiviral particles were then made from the generated construct using the two-vector packaging system, and THP-1 cells infected by spinfection. Induction doses of doxycycline (Fisher Scientific, Cat# AC446060050) were optimized, and nuclear speckle expression was confirmed by microscopy in MALAT1 knockout cells. For the pulldown, 15 million THP-1 cells were induced with 1ug of doxycycline for 24 hours after which they were stimulated with TNF alpha. There was an un-induced but TNF alpha-stimulated control for all experiments. Cells were then crosslinked with 1% formaldehyde in PBS for 10 minutes, spun down and resuspended in 0.125M glycine for 5 minutes to quench the formaldehyde. The cells were washed three times in ice cold PBS, centrifuging at 350 × g after each wash. After washing, the cell pellets were resuspended in 0.5mL SA-RNP buffer ( 20mM Tris HCL [pH 7.5], 150mM NaCl, 1.5mM Mgcl₂, 2mM DTT and 2mM vanadylribonucleosid) supplemented with 100U/mL SUPERase-In RNase inhibitor and protease and phosphatase inhibitors. The mixture was then incubated at 4°C with rotation to lyse the cells. The lysates were then sonicated, resting on ice between each round of sonication, after which they were incubated with 100uL avidin agarose solution (Thermo Fisher Scientific, cat# 20219) for 10 minutes at 4°C. Membranes and debris were pelleted by centrifugation at 13,000rpm for 15 minutes at 4°C. and supernatants saved. Supernatants were pre-cleared with 50uL of yeast RNA at 4°C for 3 hours with rotation. 50uL of pre-cleared supernatant was saved as input for each condition. The remainder of the supernatants were incubated with 100uL of streptavidin-coated magnetic beads (Thermo Fisher Scientific, Cat# 65605D) overnight at 4°C with rotation. After this beads were washed 5 times with 1mL of SA-RNP buffer, resuspended in Laemmli Sodium dodecyl sulfate loading buffer and boiled for 5 minutes at 100°C. Proteins were then analyzed by western blot as described previously.

Stimulation of THP-1 Dual cells

NF-KB activity was assessed using Quanti-Blue reagent (Invivogen Cat# rep-qbs) according to manufacturer’s instructions. Briefly, THP-1 WT (RRID:CVCL_0006), NT and MKO Dual cells were plated in media free of antibiotics overnight a day before assay was carried out. On the day of the assay cells were plated in fresh media at a concentration of 500,000 cells/ml and stimulated with stated doses of LPS for 6 hours or TNF alpha for 4 hours. After stimulation 20uL of supernatant was placed in 4 wells of a 96-well dish for each condition and 180uL of reconstituted Quanti-Blue reagent was added to each well. Color development was monitored over 6 hours and absorbance was read at 630nm with a microplate reader. Blank measurements were subtracted from all readings and the average of four readings for each condition normalized to average unstimulated values.

Nuclear and cytoplasmic fractionation

Fractionation of protein lysates was done using the NE-PER Nuclear and Cytoplasmic Extraction Reagents (Thermo Fisher Scientific, Cat# 78833) according to the manufacturer’s instructions. Briefly, Cells were stimulated with 20ng/mL of TNF alpha for 15 and 60 minutes and then pelleted and washed with PBS. Cytoplasmic fractions were obtained by incubating with CER I and CER II with protease and phosphatase inhibitors according to manufacturer’s instructions and spun at maximum speed for 5 minutes. Following this nuclear fractions were obtained from samples using NER with protease and phosphatase inhibitors for forty minutes vortexing for 15 seconds at maximum speed every ten minutes, then spinning at maximum speed for five minutes. Total protein was quantified using Pierce BCA assay and equal amounts of denatured protein loaded unto 7.5% gels. Protein levels were then assessed by western blot as described above. Lamin A/C and HSP90 were used as loading controls for nuclear and cytoplasmic fractions respectively.

Nuclear Translocation in THP-1 cells

NF-κB nuclear translocation in THP1 (RRID:CVCL_0006) cells upon TNFα stimulation was determined by imaging flow cytometry analysis as described before (38). Briefly, THP1 NT or MALAT1KO cells were stimulated with TNF alpha (20ng/ml) for 30 minutes before being harvested, stained with Live/Dead Near IR fluorescent reactive dye for 10 minutes. Cells were then permeabilized in BD Cytofix/Cytoperm buffer (fixation/permeabilization kit, Cat No. 554714, BD Bioscience) for 20 minutes and incubated with NF-κB p65 (D14E12) XP Rabbit mAb (Cell Signaling Technology Cat# 8242, RRID:AB_10859369) followed by AF-647 conjugated Donkey anti-rabbit IgG (BioLegend Cat# 406414, RRID:AB_2563202) in 1X BD Perm/Wash buffer. DAPI (nuclear marker) was added immediately before cell acquisition in ImageStream MarkII imaging cytometer (Amnis, WA). Image wizard tool in the IDEAS software (Amnis, WA) was employed to determine the degree of NF-κB P65 nuclear localization based on the similarity morphology score between nuclear image using DAPI and protein of interest (NF-κB P65). Cell population with positive similarity morphology value is considered to have NF-κB P65 highly expressed in the nucleus and were gated out to determine population percentage with nuclear NF-κB.

Phosphatase assay

THP-1 NT and MKO cells were stimulated with 20ng of recombinant human TNF alpha (Stem Cell Technologies 78068) for one hour. Stimulated and unstimulated cells were centrifugated at 300 × g for 5 minutes and pellets washed with TBS (Sigma Aldrich Cat# T5030-50TAB). Cells were lysed with RIPA buffer and protein quantified using Pierce BCA assay (Fisher Scientific Cat# 23227). Phosphatase activity was measured in 2.5ug of total protein for each condition using the 17–127 | Ser/Thr Phosphatase Assay Kit 1 (K-R-pT-I-R-R), Ser/T (EMD Millipore Cat# 17–127) according to manufacturer’s instructions. Results were normalized to unstimulated cells for each condition.

LPS stimulation, Cytokine array and ELISA

Plasma was obtained from peripheral blood of mice by centrifugation at 2000 × g for 15 minutes at 4°C for all experiments. Basal levels of cytokines were measured using the Proteome Profiler Mouse Cytokine Array Kit, Panel A (R&D Systems Cat# ARY006) and IL-6 was measured using the IL-6 Mouse ELISA kit, high sensitivity (Thermo Fisher Scientific Cat# BMS603HS). For stimulated studies, age and sex-matched MOE and WT mice were injected with 100ng of LPS in 200uL of sterile 0.9% saline intraperitoneally and monitored for adverse effects such as hypothermia. After 6 hours mice were sacrificed, and plasma was obtained by centrifugation at 2,000 × g for twenty minutes at 4°C. Grossly hemolyzed samples were discarded and IL-6 levels were measured in plasma using the Mouse IL-6 Quantikine ELISA Kit (R&D systems Cat# M6000B).

In Vivo Il-6 neutralization experiment

Peripheral blood was obtained from 10-week-old C57BL/6 and MOE mice via submandibular bleeds on day 0 and relative expansion of CD11b+Gr-1+ cells and Ly6-C^high classical monocytes in MOE mice was confirmed by flow cytometry as described above. C57BL/6 and MOE mice were randomly assigned into treatment and control groups. 1g of Il-6 neutralizing antibody (Bio X Cell Cat# BE0046, RRID:AB_1107709) or IgG control (Bio X Cell, Cat# BE0088, RRID:AB_1107775) was given in 200uL of InVivoPure pH 7.0 Dilution Buffer (Bio X Cell Cat# IP0070) every 7 days, for 21 days (day 7, 14 and 21) via intraperitoneal injection as previously described (8) starting on day 7 to allow mice to recover from bleeds. Seven days after the last injection the mice were euthanized, and peripheral blood and spleen were analyzed by flow cytometry.

shRNA Knockdown of EGR1

Lentiviral particles were made from shRNA lentiviral vector targeting human EGR1 (Sigma-Aldrich MISSION shRNA Target Clone ID TRCN000027385, target sequence CATCTCTCTGAACAACGAGAA) using the two-vector packaging system. Non-target PLKO.1 shRNA viral particles were also made with the same system using the TRC non-target control lentiviral vector (Addgene Cat# 10879 RRID: Addgene_10879). HEL and TF-1 cells were infected with EGR1 shRNA and non-Target shRNA viral particles using spinfection (1000 × g for 90 minutes at 37° then rest for 60 minutes at 37°). Growth medium was changed after this, and cells allowed to grow overnight. Infected cells were selected with puromycin (2ug/mL for TF-1 and 4ug/mL for HEL) after overnight incubation. Knockdown of EGR1 in selected cells was confirmed by western blot.

T7 CRISPR RNP knockdown of TET2

Cas 9 ribonucleoprotein (RNP) with a nuclear localizing signal was purchased from PNA Bio (Cat # CP01) and used according to the manufacturer’s instructions. Briefly, three guide RNA scaffolds targeting TET2 were incubated separately with Cas9 ribonucleoprotein to generate a pre-formed complex. The complex was then transfected into HEL cells via electroporation using the SF Cell Line 4D-Nucleofector^™ X Kit L (Lonza, Cat # V4XC-2012). After electroporation, dead cells were removed using a dead cell removal kit (Miltenyi Biotec Cat #130-090-101) and live cells allowed to expand. TET2 knockdown was confirmed by western blot and the guide RNA with the best knockdown (gRNa3) was chosen. gRNA sequences are listed in Table S6

Bulk RNA-seq Data Analysis

The quality of the raw reads was assessed using the FastQC program v0.12.1 (Andrews, Krueger et al. 2010 RRID:SCR_014583) and visualized using MultiQC v1.7(58) (RRID:SCR_014982). Low-quality bases (Phred RRID:SCR_001017, score less than 30) and adapters were removed by trimming using the Trimmomatic v0.39 (RRID:SCR_011848) in pair-end mode (ILLUMINACLIP:TruSeq3-PE-2.fa:2:40:15:8:true, HEADCROP:10, LEADING:5, TRAILING:5 SLIDINGWINDOW:4:5, MINLEN:25)(59). High-quality reads were mapped against the primary coding sequences of mouse(60). The coding sequences (gencode.vM31.transcripts.fa) were obtained from the Gencode (RRID:SCR_014966) database (https://www.gencodegenes.org/mouse/release_M21.html#) and aligned using Salmon v1.4.0(61). The count reads were analyzed using DRPPM-EASY, an Expression Analysis App hosted in R ShinY (https://github.com/shawlab-moffitt/DRPPM-EASY-ExprAnalysisShinY, RRID:SCR_001626). Differentially expressed genes (DEGs) identification was done using the DESeq2 package(62) (RRID:SCR_000154) with false a discovery rate (FDR) of P < 0.05 and log₂ fold-change of ≥1.5 or ≤ −1.5 from DRPPM-EASY application including Gene Set Enrichment Analysis (GSEA) analyses Analysis (RRID:SCR_003199).

Single-cell RNA sequencing Evolutionary Trajectory Analysis

Human (one individual’s CD34+ enriched bone marrow mononuclear cells available in the Palantir tutorial from Setty et al (26) and murine (LSK cells from three mice, after removing cell cycle effects in data and code provided from Dahlin, Hamey et al (27)) single-cell RNA sequencing (scRNA-seq) datasets of HSC populations were subjected to evolutionary trajectory analysis to understand how MALAT1 expression changes during hematopoiesis using Palantir (26), a trajectory-detection algorithm that aligns the individuals cells along pseudotime to explore changes in expression over pseudotime. For each dataset, the scRNA-seq count matrix was loaded, normalized, and log-transformed. Principal component projections and diffusion maps were generated to reduce the dimensionality of the dataset. scRNA-seq data are inherently sparse due to the experimental setup, so imputation with MAGIC (63), [Markov Affinity-based Graph Imputation of Cells algorithm (RRID:SCR_006406)], was performed to infer the missing expression levels across all the cells. The cell expressing the maximum amount of CD34 was set as the start cell and terminal cells were found using expression of other HSC markers. The Palantir algorithm was then run on the imputed expression matrix to calculate the differentiation potentials for each lineage starting with the start cell identified. Differential potentials were compared across a variety of HSC markers for the human and murine datasets to identify the trajectory of MALAT1 during normal hematopoiesis. The following canonical HSC genes were chosen for evaluation: CD34: Marker for stem and precursor cells (64), MPO: Marker for early myeloid lineage (65), GATA1: Marker for erythroid lineage (64), CSF1R: Marker for dendritic cell lineage (66), ITGA2B (CD41): Marker for megakaryocyte lineage (67), CEBPA: Marker for neutrophil lineage (68), JUN: Marker for myelomonocytic lineage (69).

Visualization of the evolutionary trajectories and gene expression was performed with t-SNE (for human dataset used per-calculated embeddings from tutorial (26) or force-directed graph embeddings (for murine analysis used pre-calculated Gephi force directed embeddings from tutorial accompany Dahlin et al (27). The specific datasets and python code (RRID:SCR_024202) (via Jupyter Notebook) to generate these trajectories is available online at: https://github.com/mcfefa/MALAT1-scRNAseq.

Single-cell data analysis

Single-cell RNA-seq: Sequenced reads from the droplet libraries were processed using 10x Genomics Cell Ranger v6.0.1(70) (Cite doi:10.1038/ncomms14049). The reads were aligned to the pre-built human reference transcriptome GRCh38 - v2020-A (July 7, 2020) provided by 10x Genomics. Read trimming, alignment, UMI counting, and cell calling were performed by Cell Ranger. Doublet prediction was done using Scrublet v0.2.1(71) with default parameters. Downstream processing was done using Seurat v4.0.4(72) (RRID:SCR_024202). Count matrices from all samples were combined and batch-corrected using Seurat v4 integration method. Count matrix from each sample was log-normalized, scaled to mean 0 and variance 1, and dimensionality reduced using PCA on the top 3000 variable genes across all samples. The reciprocal PCA and reference-based integration options were applied in the anchor finding step due to large data size. Four patient samples, one from each sex and each condition, were chosen randomly as references, and PCs 1–50 were used for the reference-based integration. The integrated data included a total of 79,422 cells from 21 patient samples. Cells with more than 50% of reads mapped to mitochondrial genes, those with less than 200 unique genes detected and those that were predicted as doublets by Scrublet were removed. After QC filtering the number of cells was reduced to 73,133. Genes that were not expressed in at least 3 cells were also removed from downstream analysis. Uniform manifold approximation and projection (UMAP) was made using the top 50 PCs obtained by running PCA on the integrated (batch-corrected) gene expression matrix. Cell type identification was done with SingleR v1.10.0(73) (RRID:SCR_023120) using immune data from celldex v1.6.0 as reference(74)

Single-cell Multiome data analysis

Sequenced reads from the gene expression (GEX) and DNA accessibility (ATAC, RRID:SCR_015980) droplet libraries of the Multiome assay were processed using 10x Genomics Cell Ranger ARC v2.0.0. The reads were aligned to the pre-built human reference genome GRCh38 - v2020-A-2.0.0 (May 3, 2021) provided by 10x Genomics. Read trimming, alignment, duplicate marking (ATAC), UMI counting (GEX), peak calling (ATAC) and joint cell calling were performed by Cell Ranger. Downstream processing was done using Seurat v4.0.4 (RRID:SCR_024202) and Signac v1.4.0(75). GEX and ATAC count matrices were integrated (batch-corrected) independently using Seurat (RRID:SCR_024202). Sample GEX count matrices were integrated following the same steps as used for the scRNA-seq data. Default options (CCA and pairwise anchor-finding) were used in the integration anchor finding step, since the data size was smaller than the scRNA-seq data. To merge the ATAC data from all samples, a common peak set was created by merging peaks from all samples using the reduce function from the R package GenomicRanges (RRID:SCR_000025). Peaks that were smaller than 20 base pairs or larger than 10000 base pairs after merging were removed. The count matrix for each sample with the new common peak set was then recalculated using Signac. The count matrices were normalized using Term Frequency - Inverse Document Frequency (TF-IDF) and dimensionality reduced using singular value decomposition (SVD) using only peaks with non-zero counts in at least 20 cells. The samples were then integrated using Seurat V4 (RRID:SCR_024202) integration with the reciprocal LSI (rLSI) method used on LSI components 2 to 50 (since the first LSI component correlates with sequencing depth) in the pairwise anchor finding step. The UMAP was calculated using the integrated LSI components 2 to 50. Seurat’s weighted nearest neighbor (WNN) algorithm was used on principal components 1 to 50 (GEX) and integrated LSI components 2 to 50 (ATAC) together to obtain a combined UMAP projection of both GEX and ATAC counterparts of the complete scMultiome dataset. Cells with more than 50% of reads mapped to mitochondrial genes, those with less than 200 unique genes detected (GEX), those with less than 200 unique peaks detected (ATAC) and those with transcription start site (TSS) enrichment score (as calculated by Signac) less than 1 were removed. Cell type identification was done using Azimuth algorithm to map the scMultiome GEX data to the scRNA-seq data since they were generated from the same cohort. The labels were then transferred from the scRNA-seq data to the single-cell Multiome data.

Differential gene expression analysis

Differential gene expression testing was done on the log-normalized counts using Seurat’s FindMarkers function with default parameters unless specified otherwise. The statistical test applied was Wilcoxon Rank Sum test with p-values adjusted using Bonferroni correction based on the total number of genes in the dataset with an adjusted p-value of 0.05. Differential gene expression testing comparing COVID-19⁺/TET2^MT versus COVID-19⁺/CH⁻ was done for each cell type independently. To make sure that the results were not driven by a single patient sample we applied a leave-one-out approach on all tests where we removed cells from one sample at a time redoing the tests and keeping only the genes that passed the adjusted p-value threshold in all tests. Pathway analysis was done using the Ingenuity Pathway Analysis, (RRID:SCR_008653) platform on differentially expressed genes.

Differential DNA accessibility and motif analysis

The differentially accessible peaks were identified by comparing the TF-IDF normalized scATAC cut-site counts of any two pair of cell populations using Seurat’s FindMarkers function. Here, the method used was the logistic regression framework along with testing for the number of fragments in peaks as a latent variable. Motif enrichments in peaks were estimated using ChromVAR (76) enrichment scores on the JASPAR2020 (77) motif matrix set. Enrichment of scATAC cut-sites in ChIP-seq peaks of DNA binding proteins (representing their binding sites) were also estimated using ChromVAR by providing ChIP-seq peaks from ReMap2022 (25) as input. Differential enrichment of binding sites was then calculated using FindMarkers function with default parameters and the leave-one-out approach. Coaccessibility scores between pairs of peaks were calculated using Cicero v1.3.5 (78). To calculate the number of coaccessible connections per peak we considered pairs of peaks with scores above 0, as predicted connections.

Materials and Correspondence

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Eric Padron (Eric.Padron@moffitt.org).

Data availability

The scRNA-seq and Multiome (GEX + ATAC) datasets generated and analyzed in this study have been deposited into the NCBI Gene Expression Omnibus (GEO, RRID:SCR_005012) data base (https://www.ncbi.nlm.nih.gov/geo/) with accession number GSE210435. Code and scripts used for analysis are made available in the GitHub repository https://github.com/LabFunEpi/TET2_analysis. Murine LSK bulk RNA-seq datasets generated in this study have been deposited into the NCBI Gene Expression Omnibus (GEO, RRID:SCR_005012) database (https://www.ncbi.nlm.nih.gov/geo/) with accession number GSE243492.

Data Analysis and Statistical Analysis

P values for MALAT1 expression in publicly available datasets was calculated using Mann Whitney test. All murine experiments were conducted with age and sex matched wild-type controls to account for experimental variabilities. P values for in vivo experiments was calculated using an unpaired students t test when comparing means of two groups, or one way ANOVA with Tukey’s correction for more than two groups. For competitive transplant experiments, means of peripheral blood chimerisms were compared with a two-way ANOVA (GraphPad Prism 10.0, RRID:SCR_002798). Error bars indicate the standard error of mean (SEM). For analysis of NF-κB activity assay in reporter (Dual) THP-1 cells, the area under the dose response curve was calculated for each cell line in each experiment. The values obtained in all experiments were compared using one-way ANOVA with Tukey’s correction.

Statement of Significance.

This work identifies MALAT1 as a requisite downstream effector of oncogenic feedforward inflammatory circuits necessary for the development of TET2-mutated CH and fulminant myeloid malignancy. We elucidate a novel mechanism by which MALAT1 “shields” P65 from dephosphorylation to potentiate this circuit and nominate MALAT1 inhibition as a future therapeutic strategy.