The mRNA m6A reader YTHDF2 suppresses proinflammatory pathways and sustains hematopoietic stem cell function

The m6A reader YTHDF2 is a novel regulator of inflammatory pathways in HSCs. YTHDF2 loss upregulates m6A-modified proinflammatory transcripts, resulting in chronic HSC inflammatory activation, myeloid bias, and functional exhaustion.


Introduction
Emerging evidence indicates the importance of the mRNA N 6methyladenosine (m 6 A) modification in hematopoietic stem cell (HSC) biology and leukemic transformation (Vu et al., 2019). m 6 A is the most abundant internal mRNA modification, which is cotranscriptionally installed by the "m 6 A writer" complex, consisting of the METTL3/METTL14 enzymatic core and their regulator WT-associated protein (Frye et al., 2018). The modification can be reversed by m 6 A demethylases (FTO and AlkBH5), collectively called "m 6 A erasers." m 6 A-modified transcripts are recognized by "m 6 A readers," including nuclear YTHDC1 (YTH domain-containing 1) and cytoplasmic YTH domain-containing family member 1 (YTHDF1), YTHDF2, YTHDF3, and YTHDC2, which execute the outcome of m 6 A modification by promoting splicing, nuclear export, translation, or degradation. Recent studies demonstrated that loss of METTL3, METTL14, or FTO weakens acute myeloid leukemia (AML) propagation, whereas inactivation of METTL3 and METTL14 has deleterious consequences for normal HSC maintenance (Barbieri et al., 2017;Cheng et al., 2019;Li et al., 2017;Vu et al., 2017;Weng et al., 2018). Furthermore, inactivation of YTHDF2, which recognizes m 6 A-modified mRNA to mediate m 6 A-mRNA degradation Wang et al., 2015), compromises initiation and propagation of AML and results in HSC expansion (Li et al., 2018;Paris et al., 2019;Wang et al., 2018). However, the involvement of m 6 A or its readers in long-term HSC maintenance upon aging remains unknown. Here, we reveal that YTHDF2 functions to down-regulate m 6 A modified transcripts involved in the inflammatory response and maintains HSC integrity upon aging.

Results and discussion
YTHDF2-deficient HSCs activate proinflammatory pathways and lose their reconstitution capacity upon serial transplantation We have recently demonstrated that hematopoiesis-specific Vav-iCre-mediated deletion of Ythdf2 results in HSC expansion and compromises AML initiation and propagation . We used the conditional and reporter Ythdf2 fl mouse allele in which exon 2 of Ythdf2 was flanked by loxP sites and GFP was introduced after the start codon of Ythdf2, generating a fully functional GFP-YTHDF2 fusion protein (Ivanova et al., 2017). Our published data indicated that 8-12-wk-old Ythdf2 fl/fl ;Vav-iCre (conditional knockout, Ythdf2 CKO ) mice displayed largely unaffected hematopoiesis compared with control Ythdf2 fl/fl (control, Ythdf2 CTL ) mice, with modest decreases in numbers of white and red blood cells, B cells, and CD8 + T cells . Furthermore, while HSCs from 8-12-wk-old Ythdf2 CKO mice displayed myeloid bias upon primary transplantation and robustly contribute to the bone marrow (BM) HSC and progenitor cell compartments , the role of YTDHF2 in long-term HSC maintenance remains unknown. To address this, we performed secondary transplantation assays by sorting CD45.2 + Lin − c-Kit + Sca-1 + (LSK) cells from primary recipients of Ythdf2 CKO and Ythdf2 CTL HSCs  and retransplanting them (together with supporting BM cells) into lethally irradiated secondary recipient mice ( Fig. 1 A). Strikingly, Ythdf2 CKO HSCs failed to efficiently repopulate longterm multilineage hematopoiesis and stem and progenitor cell compartments in secondary recipients (Fig. 1, B-D) and displayed myeloid bias (Fig. S1 A). Thus, HSCs critically require YTHDF2 to sustain their function upon serial transplantation.
To understand the cause of the failure of Ythdf2-deficient HSCs upon serial transplantation, we next set out to determine the molecular signature of Ythdf2-deficient HSCs. Ythdf2 deficiency resulted in deregulated gene expression ( Fig. 1 E), with 81 up-regulated transcripts and 33 down-regulated transcripts (false discovery rate [FDR] < 0.05) in Ythdf2 CKO compared with Ythdf2 CTL HSCs. Strikingly, gene set enrichment analysis (GSEA) on Ythdf2 CKO versus Ythdf2 CTL HSCs revealed significant up-regulation of numerous inflammation-related processes, such as IFN-α response, IFN-γ response, inflammatory response, TNF-α signaling, and IL6/JAK/STAT3 signaling (Fig. 1 F and Fig. S1 B).
Given that proinflammatory signals are known to drive overproliferation of HSCs, ultimately resulting in loss of HSC function (Essers et al., 2009;Pietras, 2017), we aimed to confirm the proinflammatory signature in Ythdf2-deficient HSCs using additional approaches. We found that HSCs from Ythdf2 CKO mice displayed increased expression of cell surface Sca-1 protein ( Fig. 1 G), which is used as a proxy of IFN signaling (Essers et al., 2009). Furthermore, Ythdf2-deficient HSCs showed elevated levels of phosphorylated forms of STAT1 and STAT3 proteins, key mediators of IFN signaling (Fig. 1 H).
Given that Ythdf2-deficient HSCs showed up-regulation of proinflammatory pathways and displayed a myeloid bias (a hallmark of inflammation), we aimed to determine whether the primitive progenitor compartment also reflected innate immune activation in Ythdf2 CKO mice. During steady-state normal hematopoiesis, HSCs generate functionally distinct lineage-biased multipotent progenitors (MPPs), with MPP2/3 and MPP4 populations predominantly sustaining myeloid and lymphoid lineages, respectively (Pietras et al., 2015). Stressful conditions such as infection, stress, or injury induce proinflammatory stimuli, which not only activate HSCs but also cause expansion of MPPs, in particular myeloid-biased MPP2/3 populations, thereby driving myeloid differentiation at the expense of HSC integrity (Essers et al., 2009;Pietras et al., 2016;Yamashita and Passegue, 2019;Zhang et al., 2017;Zhang et al., 2016). Consistent with the activation of proinflammatory pathways, we found that 8-12-wkold Ythdf2 CKO mice displayed expansion of HSCs and a concurrent increase in all four main subpopulations, namely MPP1-4 cells ( Fig. 1 I). Thus, hematopoietic-specific deletion of Ythdf2 results in a constitutive up-regulation of the proinflammatory signature, expansion of HSCs and MPPs, and a failure of HSC function upon serial transplantation.
YTHDF2 functions to suppress m 6 A-modified proinflammatory transcripts YTHDF2 regulates the transcriptome by binding m 6 A modified transcripts to promote their decay through deadenylation Wang et al., 2014). To identify YTHDF2 target transcripts, we aimed to determine mRNAs that were methylated in normal conditions and up-regulated in Ythdf2 CKO HSCs. m 6 A sequencing (m 6 A-seq) in c-Kit + hematopoietic stem and progenitor cells was used to determine transcriptome-wide m 6 A distribution. This identified the expected distribution of m 6 A within the transcriptome (i.e., enrichment around the stop codons; Fig. 2, A and B) as well as the m 6 A consensus motif (DRACH; D = A, G, or U; R = purine; and H = A, C, or U; Fig. 2 C). Consistent with the role of YTHDF2 in promoting degradation of m 6 A-modified transcripts, we found that transcripts that contain m 6 A modification (P = 2.6 × 10 −40 , 8,253 genes) show increased expression in Ythdf2 CKO HSCs (Fig. 2 D). Ingenuity pathway analysis (IPA) indicated that up-regulated genes are enriched for transcripts related to inflammatory responses (including IFN-γ, TNF-α, IFN-α, IFN regulatory factor 7, STAT1, and TLR4; Fig. 2 Fig. S2 A). Importantly, m 6 A-seq revealed that a large proportion of these key up-regulated transcripts in HSCs are m 6 A modified, including key genes involved in inflammation such as Stat1, Il6ra, and Gadd45g (Fig. 2, E and F).

E and
The up-regulation of m 6 A-modified transcripts involved in inflammation upon loss of YTHDF2 may be a consequence of an increase in their half-life. Given that the measurement of transcript half-life in HSCs is technically not possible, to address this, we analyzed our SLAM-seq dataset from immortalized c-Kit + cells derived from Ythdf2 CTL and Ythdf2 CKO mice , which allowed us to measure the transcriptomewide mRNA half-life (Herzog et al., 2017). We found that multiple m 6 A-modified transcripts that are up-regulated upon Ythdf2 deletion (Fig. 2 E) also have their half-lives increased in the absence of Ythdf2 (Fig. S2 B).
Aged HSCs lacking Ythdf2 undergo expansion and sustain steady-state hematopoiesis but fail to reconstitute multilineage hematopoiesis upon transplantation Proinflammatory signals are known to drive overproliferation of HSCs, promote myeloid bias, and ultimately lead to the loss of HSC activity (Essers et al., 2009;Pietras, 2017). Furthermore, proinflammatory pathways are up-regulated in aged HSCs, indicating their roles in stem cell aging (Yamashita and Passegue, 2019). We found that both young and aged HSCs expressed comparable levels of the YTHDF2 protein ( Fig. S3 A). To investigate the functional significance of YTHDF2 in unperturbed long-term multilineage hematopoiesis and maintenance of the HSC pool over a long period of time, we aged Ythdf2 CKO and Ythdf2 CTL mice for 1 yr (Fig. 3 A). Analyses of 60-65-wk-old Ythdf2 CKO mice revealed increased numbers of peripheral blood (PB) granulocytes and monocytes (i.e., myeloid bias); however, hematopoiesis was otherwise largely unaffected (Fig. 3, B and C;and Fig. S3,B and C). Furthermore, 60-65-wk-old Ythdf2 CKO mice displayed expansion of BM HSCs and primitive hematopoietic progenitor cell populations (Fig. 3 D). HSCs from these mice showed increased expression of Sca-1 ( Fig. 3 E), indicating constitutive activation of proinflammatory signals. To test the ability of the expanded aged HSCs to reconstitute multilineage hematopoiesis, we competitively transplanted HSCs from 60-65-wk-old Ythdf2 CKO and Ythdf2 CTL mice into lethally irradiated recipients (Fig. 3 A). While HSCs from aged Ythdf2 CTL mice robustly repopulated both myeloid and lymphoid lineages, HSCs from aged Ythdf2 CKO mice failed to efficiently reconstitute short-and long-term multilineage hematopoiesis in the recipients and displayed a myeloid bias ( Fig. 3 F and Fig. S1 C). The contribution of Ythdf2-deficient HSCs to the HSC compartment of the recipients was also decreased compared with that of control HSCs (Fig. 3 G), albeit this was not as striking as their inability to reconstitute multilineage hematopoiesis ( Fig. 1 F). Thus, while long-term Ythdf2 deficiency in the hematopoietic system does not derail steady-state multilineage hematopoiesis, it devastates the multilineage reconstitution capacity of HSCs. Given that the inflammatory pathways affect multiple HSC fates (Essers et al., 2009;Pietras, 2017;Pietras et al., 2014;Pietras et al., 2016), future investigations will be required to determine the exact impact of Ythdf2 deletion on HSC survival, self-renewal, differentiation capacity, and homing.
Inducible Ythdf2 deletion results in a progressive loss of lymphoid potential, with concurrent myeloid bias and subsequent HSC failure To delete Ythdf2 in adult mice in an inducible manner, we generated Ythdf2 iCKO mice in which Ythdf2 can be ablated acutely in hematopoietic cells upon administration of polyinosinicpolycytidylic acid (pIpC). Ythdf2 iCKO and Ythdf2 CTL mice were treated with pIpC and analyzed 26 and 52 wk later (Fig. 3 H). Inducible Ythdf2 deletion resulted in increased myeloid output ( Thus, inducible Ythdf2 deletion in adult mice causes a progressive loss of lymphoid output, myeloid bias upon aging, and an eventual failure of aged HSCs to reconstitute long-term multilineage hematopoiesis.

YTHDF2 protein expression in HSCs is induced by acute and chronic inflammation
Given that YTHDF2 expression is induced by stress in other contexts (Yu et al., 2019;Zhou et al., 2015) and is known to be up-regulated in immune cells during the IFN response (Rubio et al., 2018;Winkler et al., 2019), we hypothesized that YTHDF2 may itself be regulated by inflammatory stimuli in HSCs. To address this, we used Ythdf2 CTL mice harboring the GFP-YTHDF2 performed. (E) Transcript expression scatterplot from Ythdf2 CKO and Ythdf2 CTL HSCs from 8-12-wk-old mice (n = 4). Significantly dysregulated transcripts are highlighted and counted (FDR < 0.05; log 2 fold change < 1.2). (F) GSEA showing up-regulated pathways in HSCs from 8-12-wk-old Ythdf2 CKO mice compared with HSCs from Ythdf2 CTL mice. (G) Histogram and median Sca-1 fluorescence intensity in LSK CD48 − CD150 + HSCs from Ythdf2 CKO and Ythdf2 CTL mice determined by flow cytometry. Data represent mean ± SEM; **, P < 0.01 (Mann-Whitney U test); at least two independent experiments were performed.

Mapperley et al.
Journal of Experimental Medicine 7 of 11 YTHDF2 suppresses inflammation to sustain HSCs https://doi.org/10.1084/jem.20200829 display up-regulated proinflammatory pathways) undergo expansion under steady-state conditions, additional inflammatory stimuli abrogate this phenotype. Finally, to investigate whether Ythdf2 deficiency in HSCs had any impact on their short-term reconstitution potential upon pIpC administration, we competitively transplanted equal numbers of HSCs from pIpC-treated Ythdf2 CKO and Ythdf2 CTL mice into lethally irradiated recipients (Fig. 4 F). HSCs lacking Ythdf2 displayed decreased short-term myeloid and B-lymphoid reconstitution capacity, indicating that inflammation temporarily weakened Ythdf2-deficient HSC activity (Fig. 4 H). Thus, our data indicate that YTHDF2 is induced by inflammation in HSCs and suggest that YTHDF2 may function to protect HSCs from excessive proinflammatory signals. The mechanisms through which YTHDF2 is induced by inflammation, and its precise involvement in protecting HSCs in this context, remain important open questions meriting further investigations.
Collectively, YTHDF2 expression is induced in HSCs by inflammatory stimuli and its deletion results in constitutive activation of proinflammatory pathways, consistent with myeloid bias. While YTHDF2 is not essential for long-term steady-state unperturbed hematopoiesis upon aging, Ythdf2 deletion causes progressive HSC expansion and failure of HSCs to reconstitute multilineage hematopoiesis. Because the activation of inflammatory pathways in HSCs by infections or injury causes their proliferation and myeloid cell production, but in the longer term compromises HSC activity (Esplin et al., 2011;Essers et al., 2009;Pietras, 2017;Takizawa et al., 2017), HSCs must use mechanisms to counteract these signals and protect their long-term integrity.
While several studies indicated that proinflammatory signals are suppressed in HSCs at different levels of gene expression, including cell signaling, gene transcription, and RNA editing (Hartner et al., 2009;King et al., 2011;Sato et al., 2009), the functional significance of dynamic RNA modifications in this important aspect of HSC biology remains poorly understood. We suggest that the m 6 A-dependent mRNA decay pathway is a novel mechanism through which HSCs suppress inflammation and sustain HSC integrity. Interrogation of up-regulated proinflammatory genes upon Ythdf2 deficiency in genetic rescue approaches is warranted to further support this notion. Given our findings that YTHDF2 itself is induced by inflammation, we suggest that YTHDF2 may function in a negative feedback loop to prevent an excessive inflammatory response detrimental to HSC function. Thus, our work sets the stage for further investigations focusing on how YTHDF2 regulates HSC fates during inflammation upon acute and chronic infection. In closing, our data position YTHDF2 as an essential repressor of proinflammatory pathways in HSCs and highlight the key significance of m 6 A mRNA modification in long-term HSC maintenance upon aging.
All experiments involving mice were performed under University of Edinburgh Veterinary oversight with UK Home Office authorization. All control mice were littermates, and treatment groups were randomized among littermates.

Syngeneic transplantation assays
Transplant recipient CD45.1 + /CD45.2 + mice were lethally irradiated with two doses of 5.5 Gy administered ≥4 h apart at an pIpC administration For gene deletion via activation of the Mx1-Cre promotor, mice were injected i.p. every other day with 300 µg pIpC (GE Healthcare) for a total of six doses over 11 d, as previously described (Guitart et al., 2017;Guitart et al., 2013;Kranc et al., 2009;Paris et al., 2019), and gene deletion was assayed by YTHDF2-GFP fusion protein fluorescence. To test the effect of innate immune activation, mice were injected i.p. with 300 µg pIpC (GE Healthcare) as described per experiment.

CD117 (c-Kit) enrichment
Enrichment for cells expressing CD117 (c-Kit) was performed using CD117 MicroBeads and LS columns from Miltenyi Biotec, according to the manufacturer's instructions.

Histology
Tibias and femurs were dissected and cleaned before fixation in 4% paraformaldehyde for 24 h. Fixed bones were decalcified, embedded in paraffin, and sectioned. Sections were stained with H&E and mounted. Whole sections were imaged using a Vectra Polaris slide scanner, and representative images were taken at 10× and 20× magnification.
RNA-seq 100 HSCs per sample were sorted into 0.4% Triton X-100, RNase inhibitor, 10 mM dNTPs, and 10 µM Oligo dT before RNA-seq by SMART-Seq2 in the Single Cell Genomics Facility at the Medical Research Council Weatherall Institute of Molecular Medicine (University of Oxford). Alignments to the GRCm38 mouse genome were performed using HISAT2 (Kim et al., 2015), and further analysis was performed in R. Counts were assigned to genes using the Rsubread package, and quality control analysis of the samples was performed with the SingleCellExperiment and scater packages. Differentially expressed genes were identified using likelihood ratio tests computed by the edgeR package. The RNA-seq dataset was deposited in GEO (accession no. GSE142019).
m 6 A meRIP-seq m 6 A meRIP-seq library preparation was performed as previously described (Lin et al., 2016) from Ythdf2 CTL c-Kit-enriched cells. Reads from two biological replicates were aligned to the mouse GRCm38 genome using HISAT2 (Kim et al., 2015), and peaks were called using MACS2, version 2.1.1 with the nomodel flag (Zhang et al., 2008). To analyze the distribution of m 6 A-associated reads along the transcripts, treat_pileup bedgraph file outputs from MACS2 were converted to bigWig format and used as input for the computeMatrix function of the deepTools package, version 2.5.5 (Ramírez et al., 2014). Motif enrichment was done using Homer, version 4.10, selecting a motif length of six nucleotides. Background regions were generated by shuffling peaks along the transcriptome using the shuffleBed tool from the bedtools suite, version 2.28 (Quinlan and Hall, 2010). The GVIZ bioconductor package was used for peak visualization (Hahne and Ivanek, 2016). m 6 A meRIP-seq dataset was deposited in GEO (accession no. GSE142020).
GSEA and IPA GSEA was performed using GSEA software version 3.0 with 1,000 permutations and default parameters. Gene differential expression, computed by the edgeR package in R, was ranked by moderated t statistics, which takes into account variability between genes in the ranking. Ranked genes were compared with gene lists in the Hallmark subset of the MSigDB database, version 7.0. IPA was performed using the Core Analysis Function offered by Qiagen's Ingenuity Pathway Analysis software (Krämer et al., 2014). The interrogated RNA-seq and mass spectrometry datasets were filtered for adjusted P values of differential expression (FDR < 0.05), and the threshold for significant activation or inhibition was defined by an absolute Z-score value >2.

SLAM-seq analysis
The SLAM-seq dataset  was analyzed using the SlamDunk pipeline (Herzog et al., 2017).

Mass spectroscopy analysis
Immortalized Ythdf2 CKO and Ythdf2 CTL c-Kit-enriched cells  were pelleted, lysed, and processed according to (Wiśniewski and Rakus, 2014). Tryptic peptides were analyzed on a QExactive+ mass spectrometer connected to an Ultimate Ultra3000 chromatography system packed with 1.8 µm uChrom and separated by an acetonitrile gradient. Mass spectra were analyzed using MaxQuant against the Uniprot mouse database proteome, ID UP000000589.

Quantification and statistical analysis
Statistical analyses were performed using Prism 6 software (GraphPad Software). P values were calculated using a twotailed Mann-Whitney U test unless stated otherwise.
Online supplemental material Fig. S1 shows the frequency of B cells, T cells, granulocytes, and monocytes within the CD45.2 + donor-derived population of PB, BM, and spleen from transplant recipients shown in Figs. 1 B; and 3 F, I, and K. Also shown are GSEA plots from RNA-seq of Ythdf2 CKO and Ythdf2 CTL HSCs from Fig. 1 F. Fig. S2 shows IPA of RNA-seq and mass spectrometry from Ythdf2 CKO and Ythdf2 CTL cells and mRNA decay curves for transcripts of interest within Ythdf2 CKO and Ythdf2 CTL c-Kit + cells. Fig. S3 shows data to support our conclusions in aged mice (YTHDF2-GFP mean fluorescence is not significantly altered with age; aged mice have normal BM cellularity and histology); percentage of monocytes, granulocytes, CD4 + T cells, CD8 + T cells, B cells, and GFP + cells in PB from mice 1 yr after Ythdf2 deletion; and total blood counts, numbers of differentiated cells, and numbers of HSCs for mice 1 yr after Ythdf2 deletion.