Aberrant chromatin landscape following loss of the H3.3 chaperone Daxx in haematopoietic precursors leads to Pu.1-mediated neutrophilia and inflammation

Defective silencing of retrotransposable elements has been linked to inflammageing, cancer and autoimmune diseases. However, the underlying mechanisms are only partially understood. Here we implicate the histone H3.3 chaperone Daxx, a retrotransposable element repressor inactivated in myeloid leukaemia and other neoplasms, in protection from inflammatory disease. Loss of Daxx alters the chromatin landscape, H3.3 distribution and histone marks of haematopoietic progenitors, leading to engagement of a Pu.1-dependent transcriptional programme for myelopoiesis at the expense of B-cell differentiation. This causes neutrophilia and inflammation, predisposing mice to develop an autoinflammatory skin disease. While these molecular and phenotypic perturbations are in part reverted in animals lacking both Pu.1 and Daxx, haematopoietic progenitors in these mice show unique chromatin and transcriptome alterations, suggesting an interaction between these two pathways. Overall, our findings implicate retrotransposable element silencing in haematopoiesis and suggest a cross-talk between the H3.3 loading machinery and the pioneer transcription factor Pu.1.

While the number of distal regions opening or closing in CMPs or GMPs compared with LT-HSCs were similar in the WT mice, substantially fewer distal regions were open in the Daxx-KO animals (Fig. 1b); most remained closed, suggesting Daxx-KO CMPs and GMPs became more restricted in their gene expression.
By overlapping accessible regions with known haematopoietic enhancers, we found more enhancers opening in KO LT-HSCs (2,509) compared with CMPs and GMPs (1,040 and 716, respectively; Fig. 1c). Based on unsupervised clustering (Fig. 1d), one set of enhancers showed increased read counts in KO LT-HSCs compared with WT LT-HSCs but reduced counts in CMPs and GMPs. Another set displayed similar counts in WT CMPs and GMPs as well as KO LT-HSCs, suggesting that chromatin features of Daxx-KO LT-HSCs may resemble myeloid-committed progenitors. The KO LT-HSCs displayed increased general opening of enhancers, including those overlapping with ERVs, in accordance with a Daxx-repressive role (Fig. 1e). ERV-overlapping enhancers were enriched in transcription factor motifs for the master regulators of haematopoiesis Pu.1 and Ets1 (Fig. 1f). Neighbouring genes displayed enrichment in immune-cell function and pathways associated with myeloid-committed progenitors (Extended Data Fig. 1d,e).
More acute effects of Daxx loss were assessed by collecting mice at 3 d post induction (d.p.i.). Like at 3 w.p.i., most changes occurred at distal regions. However, more sites displayed decreased (8,421) than increased accessibility (3,247). Despite a similar number of known haematopoietic enhancers being opened or closed (1,020 versus 1,022), more enhancers overlapping with ERVs opened (818 versus 617), with more pronounced differences at opened ERV-overlapping enhancers (Fig. 1g). RNA sequencing (RNA-seq; Extended Data Fig. 1f and Supplementary Table 1) showed increased expression of genes close to opened enhancers and ERV-overlapping enhancers (Fig. 1h,i). However, genes proximal to enhancers with reduced accessibility also showed increased expression, an effect potentially due to enhancer occupancy by transcription factors (Fig. 1i).
Ingenuity pathway analysis (IPA) and Gene Ontology suggested engagement of the double-stranded RNA (dsRNA)-recognition machinery and activation of interferon (IFN)-stimulated genes (ISGs; Fig. 1j-m and Supplementary Table 2). Key components of these pathways were indeed upregulated (for example, Irf3, Irf7, Mavs and Ddx58; Fig. 1l,m) along with selected ERV subtypes, including the Daxx targets ERV1 and IAPs, but not non-ERV RTEs ( Fig. 1n and Extended Data Fig. 1g). Enriched cell cycle-related pathways, along with upregulation of the proliferation marker Mki67 and downregulation of the quiescence gene Egr1, pointed at increased cell-cycle entry (Extended Data Fig. 1h and Supplementary Tables  1,2). Notably, upregulated genes such as Mx1, the Ifit cluster, Irf7 and Mki67 showed increased accessibility at ERV-overlapping enhancers and at Pu.1-binding sites (based on 19 ; Fig. 1o,p and Extended Data Fig. 1h), suggesting a role for Pu.1 in their regulation. Spi1 (coding for Pu.1) regulatory elements showed increased opening but its expression was unaltered (Extended Data Fig. 1h).
In agreement with the reported role for Atrx in the regulation of the telomeric repeat-containing RNA (TERRA) 10,20-23 , TERRA-binding sites (TERRA-BS) 24 were more open across all chromosomes in Daxx-KO LT-HSCs, including those at Mid1 and Erdr1, a potential HSC regulator and repressor of inflammatory skin disease [25][26][27] (Extended Data Fig. 1i,j). Chromatin opening correlated with their downregulation, as was also observed at the autosomal Wls locus (Extended Data Fig. 1j). TERRA knockdown also reduced expression of these loci, whereas Atrx loss increased their expression 23 , suggesting that Daxx and Atrx may antagonize each other for TERRA regulation.
Overall, our data suggest that Daxx loss in LT-HSCs alters the chromatin landscape at ERV-overlapping enhancers and TERRA-BS, with potential implications for induction of an IFN type I response and cell-cycle entry.

Daxx loss skews haematopoiesis towards myeloid differentiation
We next investigated the effect of Daxx loss on haematopoiesis. The overall cell numbers in the KO bone marrow (BM) were significantly reduced at 3 d.p.i. (Fig. 2a), whereas the frequencies and numbers of LT-HSCs, c-Kit + Lineage (Lin) -Sca-1 + (KLS) cells and myeloid-restricted multipotent progenitor (MPP) 3 cells (Fig. 2b) were increased. Both GMPs and lymphoid-biased MPP4 cells showed increased frequencies. The neutrophil frequency in the BM was augmented, whereas the frequency and number of B220 + cells were reduced (Fig. 2c). The B-cell defect encompassed pro-B, pre-B, and immature and mature B cells (Fig. 2d). The neutrophil frequency and number in the spleen were increased, whereas the monocyte number was reduced. B cells were unaffected and remained viable (Fig. 2e,f). At 2 w.p.i., the frequency and number of neutrophils were higher, whereas those of B cells were lower (Fig. 2g). Thus, Daxx loss causes early expansion of stem and progenitor cells, followed by increased BM and peripheral accumulation of neutrophils at the expense of B cells. To study the long-term effects of Daxx loss on haematopoiesis, we employed haematopoiesis-specific Cre lines. Daxx deletion during prenatal haematopoiesis using the Csf1rCre line failed to produce viable Daxx-KO pups (not shown), similar to a germline Daxx-KO line 28 . Haematopoiesis-specific and polyinosinic:polycytidylic acid (pI:pC)-inducible Daxx deletion in adult mice (using the Mx1Cre   Ifih1   Irf3  Ifnar2  Ifnz  Nfkb1  Mavs  Ikbke  Il1b  Ifnar1  Trim25  Fadd  Tank  Irf7  Ddx58  Traf3  Ifih1  Nfkb2  Tradd  Tnf  Ifngr2  Ifnlr1  Tbk1  Traf6  Cxcl2  Ifngr1  Il6  Traf2  Tmem173  Rnf135 Tomm70a  Fig. 2a-d) resulted in an increase in mature and immature neutrophils in the BM (Fig. 2h,i). Daxx-deficient mice displayed a trend towards lower BM cell counts but the number of white blood cells was higher in males after red-blood-cell (RBC) lysis (Extended Data Fig. 2e,f). These changes correlated with a higher frequency of Ki67 + cells in the BM as well as in HSPCs (Extended Data Fig. 2g). Similar to the RosaCre ERT2 model, the GMP frequency and number were increased (Fig. 2j,k). Although the frequency of common lymphoid progenitors (CLPs) was augmented, their total number was reduced (Fig. 2k,l); similar observations were made for LT-HSCs, and KLS, MPP3 and MPP4 cells (Extended Data Fig. 2h,i). Cytokines promoting myeloid differentiation, such as granulocyte-macrophage colony-stimulating factor (GM-CSF), interleukin (IL)-6, thrombopoietin and IL-34, were increased in plasma (Fig. 2m), correlating with higher frequencies of CD11b + , Gr-1 hi and Ly6G hi myeloid cells (Extended Data Fig. 3a). While Ly6C hi inflammatory monocytes were nearly absent (Extended Data Fig. 3a), neutrophils displayed increased number and frequency as well as diminished expression of the maturation marker Cxcr2 (Fig. 2n-p). The B-lymphocyte frequencies were reduced, with a marked drop from the pre pro-B cell stage onward (Extended Data Overall, Daxx loss markedly caused perturbations of haematopoietic differentiation compatible with a myeloid bias towards neutrophil production.

Daxx-KO mice develop neutrophilia and inflammation
Myeloid markers were increased in the peripheral blood (PB) of Daxx-KO mice, whereas the B-cell frequency decreased (Extended Data Fig. 4a). The levels of Cxcr2 were reduced in PB neutrophils, indicating a more immature status also outside of the BM (Extended Data Fig. 4b). A trend towards a reduction in the white-blood-cell count was observed (Extended Data Fig. 4c). Compared with controls, Daxx loss (Extended Data Fig. 4d-f) led to marked reduction in white pulp (Fig. 3a). The total number of spleen cells was unchanged, despite an increased frequency of myeloid cells (Fig. 3b,c). The CD41 + megakaryocyte frequency was higher (Extended Data Fig. 4a), although the number of platelets in the PB was only marginally increased (Extended Data Fig. 4c). Correlating with the reduction in white pulp, B cells were significantly reduced ( Fig. 3c,g); the decrease affected follicular but not marginal B cells and to a minor degree plasma cells (Fig. 3c,h). Spleen macrophages (Ly6C lo/− Gr-1 lo/− ) were reduced, whereas monocytes were shifted towards a Ly6C intm Gr-1 hi population ( Fig. 3d and Extended Data Fig. 4g). We observed an increase in the neutrophil number and frequency, a higher eosinophil frequency, decreased Cxcr2 levels (Fig. 3e,f) and a trend towards an increased Ki67 + cell frequency (Extended Data Fig. 4h). Expansion of CD11b + cells was observed but the F4/80 + monocytes/macrophages were unaltered ( Fig. 3i-k). In contrast to BM, erythroblast subpopulations were increased (Extended Data Fig. 4i), suggesting induction of extramedullary erythropoiesis, in accordance with normal RBC counts in the PB.
In agreement with reports linking neutrophilia and inflammation 30-33 , we observed elevated levels of several key cytokines involved in myelopoiesis and inflammation (Fig. 4a), such as GM-CSF 34 and IL-17A, the latter of which is linked to skin diseases 35,36 . Interestingly, approximately 15% of the Daxx F/F ;Mx1Cre +/− mice developed skin lesions along with splenomegaly ( Fig. 4b-e). These lesions resembled human pyoderma gangrenosum, a cutaneous neutrophilic autoinflammatory disease of unknown aetiology (Fig. 4e) 37 partly associated with myeloproliferative diseases. Similar to human pyoderma gangrenosum, the mouse lesions were characterized by thicker skin and sterile inflammation (Fig. 4e), with infiltration of neutrophils and neutrophil extracellular traps (Fig. 4f,g). The rather long latency of skin lesions correlated with a progressive increase in neutrophilia over time (Fig. 4h).
These findings suggest that progressive neutrophilia leads to systemic inflammation in Daxx-deficient mice.

Limited role of Hira in adult haematopoiesis
We next assessed the contribution of Histone cell-cycle regulator (Hira), which deposits H3.3 at euchromatin and facultative heterochromatin 38 . In pI:pC-treated Hira F/F ;Mx1Cre +/− mice (Extended Data Fig. 5a-c), no changes in erythroid marker frequencies or bone appearance were observed (Extended Data Fig. 5d,e). The frequencies of myeloid and lymphoid markers in the BM were significantly affected but markedly less than that observed in the Daxx-KO mice (Extended Data Fig. 5f,g). Although the frequency of neutrophils was significantly increased in the BM of the Hira-KO mice, it was unaffected in the spleen (Extended Data Fig. 5h,i). Unlike Daxx-deficient mice, the spleens of the Hira-KO mice were not enlarged (Extended Data Fig. 5j) and showed no or only minor changes in the myeloid, erythroid and lymphoid marker frequencies as well as organ structure (Extended Data Fig. 5k-m). No skin lesions were observed after extended monitoring. The marginal effects of Hira loss on haematopoiesis may be due to its lesser role and/or compensatory effects by Daxx given that the Daxx levels were higher in Hira-deficient cells (Extended Data Fig. 5c). The stronger phenotype described in a previous report based on Hira prenatal deletion 39 could be caused by a more prominent Hira role during embryonic haematopoiesis.

Haematopoietic defects of Daxx deficiency are cell-intrinsic
To determine the nature of the phenotypic changes described earlier, we performed colony-forming unit (CFU) assays using either Lin − progenitors (HPCs) or LT-HSCs from untreated Daxx F/F ;Mx1Cre +/− and Daxx +/+ ;Mx1Cre +/− mice, which underwent spontaneous recombination in vitro, probably due to activation of IFN signalling (not shown). An increase in colonies was observed following Daxx loss in HPCs (Extended Data Fig. 6a) at passage 0 (P0). Conversely, at P1 Daxx-deficient cells gave rise to fewer colonies, suggesting that their colony-forming potential was reduced. The CFU assays with LT-HSCs showed a similar phenotype following Daxx loss (Extended Data Fig. 6b,c), with an increased number of KLS cells at P0 and reduction at P1 (Extended Data Fig. 6d). Similar changes were observed in the number of CD11b + cells and neutrophils (Extended Data Fig. 6e). Finally, transplantation of BM cells isolated from pI:pC-treated mice into congenic recipients (Extended Data Fig. 6f) suggested a reduced reconstitution capacity of Daxx-KO stem/progenitor cells, resulting in a low frequency of CD45.2 + cells in the periphery (Extended Data Fig. 6g). Despite similar white-blood-cell counts (Extended Data Fig. 6h), KO transplants displayed altered frequencies of myeloid and B cells in the BM, PB and spleen, and of KLS cells and GMPs in the BM, as observed in the donor mice (Extended Data Fig. 6i-l).
Overall, these data suggest that the aberrant blood-cell composition observed following Daxx loss is due to a cell-intrinsic defect.

Pu.1-driven myeloid signature following Daxx loss
We performed transcriptome analyses of KLS and GMP subpopulations isolated from BM chimaeras (see the 'Haematopoietic defects of Daxx deficiency are cell-intrinsic' section; stress haematopoiesis) as well as from pI:pC-treated mice collected at 3 and 24 w.p.i. Unsupervised clustering of selected master regulators and haematopoiesis markers revealed clustering according to genotype and condition as well as increased expression with time of myeloid-associated genes such as Mpo, Elane, Spi1 and Cebpa. This was accompanied by a reduction in the expression levels of lymphoid genes (Flt3, Ebf1, Irf4, Pax5 and Gfi1b) in the KO KLS cells ( Fig. 5c) but not in GMPs (Extended Data Fig. 7d). Accordingly, by comparing cells at 3 and 24 w.p.i. we found that Daxx-KO KLS cells switch from pathways compatible with increased lymphopoiesis to increased myelopoiesis and reduced B-cell generation (Fig. 5d,e and Extended Data Fig. 7e,f). This effect was even more pronounced in transplantation settings ( Fig. 5f and Extended Data Fig. 7g). Furthermore, an increasing number of Pu.1-target genes driving myeloid fate were upregulated over time and following stress, accompanied by reduced expression of lymphoid Pu.1-target genes ( Fig. 5g-i). These data are in accordance with neutrophilia becoming more pronounced with time ( Fig. 4h), which was also confirmed by pathway analysis for Daxx-KO GMP cells (Extended Data Fig. 7h).
Transcriptomics of lymphoid-and myeloid-primed multipotent progenitors (MPP4 and MPP3, respectively) combined with unsupervised clustering of myeloid marker gene and transcription factor expression showed that Daxx-KO MPP4 cells cluster together with WT MPP3 (Fig. 5j, top). A similar analysis based on lymphoid genes showed Daxx-KO MPP4 clustering with MPP3 cells, suggesting that Daxx-KO lymphoid-biased MPP4 have shifted towards a more myeloid-biased gene expression (Fig. 5j, bottom).
Deregulation of ERVs/RTEs in Daxx-KO KLS cells and GMPs became more pronounced with time and following stress, especially that of ERV1 and IAPs ( Finally, ERV-overlapping enhancers in proximity of both up-and downregulated genes were more accessible in Daxx-KO KLS cells at 3 w.p.i. (Fig. 5m,n). For instance, Spi1 and Gfi1 (which together drive neutrophilic differentiation 40 ) displayed increased enhancer accessibility at 3 w.p.i. but were upregulated only in transplantation settings. In contrast, Irf4 expression (which cooperates with Pu.1 for B-cell differentiation 41 ) was higher at 3 w.p.i. but lower at 24 w.p.i. and in transplantation settings.
Together, these findings suggest that Daxx loss alters chromatin features and Pu.1 distribution at enhancers, including those overlapping with ERVs.

Phenotypic rescue and unique changes after Daxx and Pu.1 DKO
We next investigated whether the phenotypic alterations caused by Daxx loss were dependent on Pu.1 by using a Spi1 F/F line 42 . The BM cellularity was not significantly different in Daxx-KO mice compared with Daxx and Pu.1 double-KO (DKO) mice, although the latter displayed a tendency towards higher cellularity (Fig. 7a). At 8 w.p.i., the BM of the DKO mice displayed significant reduction in the GMP and neutrophil frequencies and numbers compared with the Daxx-KO (Fig. 7b,c), whereas the B220 + cells in the BM were still compromised (Fig. 7c). Conversely, in DKO spleens, the B-cell and neutrophil frequencies and numbers were normalized ( Fig. 7d-g) along with spleen architecture and selected cytokines (at 3 and 8 w.p.i.; Fig. 7h-j). The B-cell frequency was also partially rescued in the DKO PB (Fig. 7k,l). Neutrophil and B-cell numbers and frequencies in the BM, spleen and PB were similar to WT following the loss of Pu.1 alone (Extended Data Fig. 8d,e).
To link the phenotypic rescue to normalization of molecular perturbations, we studied changes in transcription, chromatin landscape and histone marks in single-KO and DKO mice (Supplementary Tables 9 and 10). The PCA analysis of RNA-seq data showed that each genotype clustered separately (Fig. 8a,b). Although DEGs from different comparisons (Daxx-KO versus WT, DKO versus WT and Pu.1-KO versus WT) partially overlapped, many genotype-specific changes also occurred, particularly in KLS cells (Extended Data Fig. 9a,b). Pathway analysis based on DKO-specific and Pu.1-KO-specific DEGs in KLS cells and GMPs identified distinct biological functions and diseases, including some associated with lymphoid neoplasms (Fig. 8c,d and Extended Data Fig. 9c,d). Clustering of expression for haematopoietic transcription factors and markers, Pu.1-target genes, ISGs and the dsRNA-sensing machinery revealed both common and genotype-specific changes between Pu.1-KO and DKO cells (Extended Data Fig. 9e-g). Altered expression of selected ERV/RTE subtypes in Daxx-KO KLS cells and GMPs were in part reverted in the DKO cells (Fig. 8e,f). Finally, Pu.1-KO KLS cells and GMPs showed upregulation of some ERV/RTE subtypes and satellite repeats, suggesting a role for Pu.1 in their regulation.
On investigation of the chromatin changes, we found that 90.8% of distal regions that opened following Daxx loss became closed in DKO cells (Fig. 8g). Conversely, the vast majority of distal regions found closed in the Daxx-KO KLS cells opened in DKO KLS cells. In addition, we found DKO-specific chromatin changes: 15,587 distal regions that were open in WT and Daxx-KO KLS cells became closed in DKO cells, while 3,922 distal sites that opened in DKO cells were closed in the other genotypes. When we compared open regions in WT, Daxx-KO and DKO KLS cells, we found that DKO cells generally displayed the lowest accessibility (Fig. 8h). Many of the genes that were significantly upregulated in the Daxx-KO KLS cells at 3 w.p.i. and resided near enhancers with increased accessibility were downregulated in the DKO cells (Fig. 8i). Finally, opening of TERRA-BS at sex chromosomes in Daxx-KO KLS cells was reverted in DKO cells (Extended Data Fig. 10a). Closing of chromatin at the TERRA-BS in the vicinity of the Mid1 and Erdr1 genes in the DKO cells correlated with rescued gene expression compared with the Daxx-KO cells (Extended Data Fig. 10b).
H3K9me3 alterations found in Daxx-KO HSPCs were reverted to some extent in DKO cells (Fig. 8j). In addition, we found that there were more distal regions in DKO HSPCs that gained H3K9me3      (Fig. 8k).
Finally, we zoomed in on the locus encoding Fbp1 and Fbp2 (Fig. 8l), two critical regulators of gluconeogenesis that affect HSPC repopulation capacity via inhibition of glycolysis under the control of Setdb1 (ref. 43    cells, whereas Fbp2 remained silenced apart from transplantation settings. Fbp1 silencing was restored in DKO cells, suggesting that Pu.1 is required for its induction. This is in accordance with three main regions bound by Pu.1 at this locus (as reported in 19 ), two of which coincide with haematopoietic enhancers 18 . H3K27ac enrichment at these regions in Daxx-KO cells was reverted in DKO cells. Although H3.3 was reduced at the proximal enhancer in Daxx-KO cells, its levels were restored in DKO cells, suggesting involvement of another chaperone, such as Hira. A similar trend applied to H3K9me3 at this enhancer. The large block of H3K9me3 over the Fbp2 gene and its regulatory regions, which was previously linked to Fbp2 upregulation in HSPCs 43 following Setdb1 loss, was further enriched in Daxx-KO cells but not DKO cells. Finally, H3K27me3 was substantially reduced across the entire locus in the Daxx-KO and DKO cells, especially at the most proximal enhancer of Fbp1, suggesting that this mark may cooperate with Pu.1 loss for Fbp1 silencing.

Discussion
Together, our findings implicate the H3.3 chaperone and ERV/RTE repressor Daxx in the regulation of haematopoiesis and protection from inflammation. We propose that Daxx acts as an epigenetic barrier controlling cell plasticity in HSPCs in part via RTE control, in turn ensuring balanced cell differentiation (Extended Data Fig. 10c). Recent work in zebrafish implicates RTE-driven engagement of RNA-sensing in HSPC emergence in haematopoiesis 44 . In this respect, we showed that ERV/RTE induction correlates with the IFN type I-like response in Daxx-deficient LT-HSCs. IFN type I signalling is normally repressed in HSCs, as it may promote cell death 45 , but when activated in quiescent HSCs it causes cell-cycle entry 46 . Following acute Daxx loss, the number of LT-HSCs and other progenitor types did indeed increase. As it is unlikely that stem/progenitor cells are able to produce IFNs, ISG induction could be IFN-independent as reported in viral-infection models [47][48][49][50] . This could explain the moderate ISG induction in Daxx-KO LT-HSCs.
Our study points to a link between Daxx and the pioneer transcription factor Pu.1. Pu.1 engagement does not seem to be an immediate effect of Daxx loss in HSCs. We instead observed progressive activation of a Pu.1 myeloid programme in haematopoietic progenitors with time and following stress, along with aggravation of neutrophilia. This mirrors the response to viral infections, where HSC activation is followed by a return to quiescence and an accumulation of myeloid-biased HSCs 56 , a phenomenon that has also been observed in ageing 57 . Thus, Pu.1 may contribute to the myeloid bias within a subset of Daxx-deficient HSCs that have returned to quiescence.
Concomitant Daxx and Pu.1 loss reverts some phenotypic perturbations found in Daxx-KO mice, including a partial recovery of B-cell differentiation and spleen architecture. This would be in agreement with the reported role of Pu.1 in inhibition of terminal B-cell and plasma-cell differentiation 58 . At the molecular level, the increase in enhancer accessibility that typifies Daxx-KO KLS cells is reverted by over 90% in DKO cells. Closing of chromatin in DKO cells at TERRA-BS correlated with rescued gene expression of neighbouring genes, suggesting that perturbations at TERRA-BS and its target genes, such as Erdr1, may contribute to the phenotypes caused by Daxx loss 26 . However, DKO progenitors also show unique chromatin and transcriptome alterations that are not found in single Pu.1-or Daxx-KO mice, suggesting a genetic interaction between the two pathways. In support of this hypothesis, Daxx loss causes reduction in the levels of both Pu.1 and H3.3 at enhancers, correlating with their increased enrichment at neighbouring genes. These findings suggest that Pu.1 and H3.3 may influence their reciprocal genome distribution.
Finally, dysfunction of intrinsic immunity mechanisms devoted to ERV/RTE silencing may predispose to inflammation and neoplastic       65 . It is therefore possible that DKO mice may become leukaemic at old age. Overall, this work implicates mechanisms governing accessibility to repeat elements in homeostasis of the haematopoietic system via cross-talk with pioneer transcription factors such as Pu.1. Perturbations of these mechanisms may contribute to inflammageing and predispose to leukaemogenesis.

Online content
Any methods, additional references, Nature Research reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/ s41556-021-00774-y.

Patient sample information.
We included H&E staining of skin from patients with pyoderma gangrenosum in our study, which were performed by J.W. at the University of Bonn. The histological picture of the pyoderma gangrenosum lesions was taken from a skin sample of a patient with pyoderma gangrenosum during a normal diagnostic procedure. This routine procedure includes the preparation of H&E samples. Based on German law, it is possible to take pictures from these skin samples (the patient identity is unknown and no additional investigations were performed with these samples). As a result, a 'patient consent' is not needed. However, the patient did give informed consent to perform the skin biopsy during the diagnostic procedure. This is also in accordance with the Helsinki Ethical guidelines. J.W. has ethical approval from the University of Bonn, which in principle allows the use of skin material taken for diagnostic proposes in later research (BN090/04). DAXX_recombined (5′-GCTCACGCCTTTAGTCCGAA-3′). C57BL/6N Hira mice were genotyped using the primers 2292_27 (5′-AATGGTGCTTG CTTTTGTGG-3′) and 2292_28 (5′-CCTGCTACCTTATTCTCCAGTCC -3′), and the primer 2293_30 (5′-GCATTACTTAATCCCCAGATGC-3′) was added for the analysis of recombination efficiency. Genotyping of the C57BL/6 Pu.1 mice was performed with the primers PU1_Flox_FW (5′-CTTCACTGCCCATTCATTGGCTCATCA-3′) and PU1_Flox_Rev (5′-GCTGGGGACAAGGTTTGATAAGGGAA-3′), and the primer PU1_MUT (5′-CAACCGGATCTAGACTCGAGGA-3′) was added to determine the recombination efficiency. Cre mice were genotyped according to protocols released by The Jackson Laboratory. The mice were housed under specific pathogen-free conditions at the central animal facility of the University College London and German Center for Neurodegenerative Diseases. The mice were bred and subjected to listed procedures according to protocols approved by the English (Home Office) and German authorities (Landesamt für Natur, Umwelt und Verbraucherschutz Nordrhein-Westfalen). Activation of Cre in Daxx;RosaCre ERT2 mice was induced by administration of 80 mg kg −1 tamoxifen in corn oil (Sigma-Aldrich) via oral gavage on five consecutive days (ATAC-seq was performed at 3 w.p.i.) or by administration of 100 mg kg −1 tamoxifen in corn oil via intraperitoneal injection (five consecutive days, break of two days, two consecutive days; genomic and phenotypic analyses were performed at 3 d.p.i. and 2 w.p.i.). For the induction of Cre activation in Daxx-KO, Hira-KO and Daxx and Pu.1 DKO, Mx1Cre mice received three intraperitoneal injections of 300 µg pI:pC (Sigma-Aldrich) administered every other day. The mice were usually treated at 5-11 weeks of age. Mice of both sexes were used in this study. The n value in the figure legends reflects the number of mice analysed in each experiment. Method details. Haematopoietic-cell and organ preparation. Peripheral blood samples were collected from the tail vein or by heart puncture. For the isolation of haematopoietic organs, the mice were euthanized by CO 2 exposure. Unless otherwise indicated, the mice were killed two months post treatment. Bone marrow was isolated from the femur and tibia of the hind legs. All isolated organs were directly processed for analysis, frozen for cryopreservation or fixed in neutral buffered formalin (Sigma-Aldrich) for subsequent paraffin embedding. HPCs were isolated from BM using an EasySep mouse hematopoietic progenitor cell isolation kit (STEMCELL Technologies) according to the manufacturer's instructions. If required, RBCs were lysed for 10 min in ammonium chloride (STEMCELL Technologies) and washed twice before downstream analysis.
Analysis of BM cellularity. Mice were killed two months after induction, except when indicated otherwise. BM cells were isolated from all four hind-leg bones and collected in equal volumes of PBS. The number of cells was determined using a TC20 automated cell counter (Bio-Rad). The RBCs were subsequently lysed for 10 min in ammonium chloride (STEMCELL Technologies), washed twice and collected in equal volumes of PBS. The number of cells in the samples was calculated again following after RBC lysis. If indicated, analyses were performed separately for male and female mice to control for differences in body size and weight.
Flow cytometry and cell sorting. For flow cytometry analysis, cells were isolated from the specified organs. Dead cells were stained by propidium iodide (Sigma-Aldrich) or Aqua Zombie fixable viability dye (BioLegend) according to the manufacturer's protocol. Fluorochrome-labelled antibodies, as indicated in Supplementary Table 6, were added to the cell suspension for 30 min on ice. The CD16 and CD32 receptors were blocked with CD16 and CD32 antibodies unless the cells were specifically stained for CD16/32. Intracellular staining was performed according to the manual of the Intracellular fixation and permeabilization buffer set (eBioscience). Lin − antibody panels contained antibodies to B220, Gr-1, CD11b, TER-119 and CD3e. For ATAC-seq, the Lin − antibody panel included antibodies to B220, CD19, Gr-1, CD11b, CD11c, NK1.1, TER-119 and CD3e. For RNA-seq, the Lin − antibody panel included antibodies to B220, Gr-1, CD11b, TER-119, CD8, CD4 and CD3e. For staining with multiple brilliant violet dyes, the BD Horizon brilliant stain buffer was used according to the manufacturer's instructions. For the assessment of apoptotic B cells, cells isolated from the spleen were first stained with antibodies to CD45, CD11b and B220, followed by staining with an CellEvent caspase-3/7 green flow cytometry assay kit (Invitrogen) according to the manufacturer's instructions. The cells were washed and filtered through a 70 µm filter before flow cytometric analysis using a BD LSRFortessa, BD FACSymphony, BD FACSCelesta, BD FACSAriaIII or Beckman Coulter Gallios machine. Cell sorting was performed on a BD FACSAriaIII or BD FACSAria Fusion machine using a 70 µm or 100 µm nozzle. Data were analysed using the FlowJo software. Neutrophils were gated as CD11b + Ly6G + Ly6C intm Gr-1 hi and eosinophils as CD11b + Ly6C intm SSC hi Ly6G -Gr-1 + .
CFU assay with re-plating. For the CFU assays, HPCs were isolated from BM as described in the 'Haematopoietic-cell and organ preparation' section. The HPCs were plated (15,000 cells per 35 mm dish) in MethoCult M3231 medium supplemented with 100 ng ml −1 SCF, 10 ng ml −1 GM-CSF, 10 ng ml −1 IL-3 and 10 ng ml −1 IL-6 (all STEMCELL Technologies) to promote myeloid-cell formation. The cells were cultivated for 1 week at 37 °C. Next, the colonies were counted under a microscope and 15,000 cells were re-plated under the same conditions. After another week of cultivation at 37 °C, the colonies were again counted under a microscope.
Western blotting. Cells were washed in PBS and lysed in RIPA buffer with cOmplete protease inhibitor cocktail (Roche). The lysed samples were sonicated for 10 s and the protein concentrations were determined using a Pierce protein assay kit (Thermo Fisher) according to the manufacturer's instructions. The proteins were separated by SDS-PAGE and blotted on a nitrocellulose membrane. Blocking was performed for 1 h in 5% non-fat dry milk (PBS-T) or 4% BSA (TBS-T) at room temperature and the membranes were incubated overnight with primary antibodies at 4 °C. The primary antibodies were detected by horseradish peroxidase-conjugated secondary antibodies and subsequently visualized using the ChemiDoc XRS+ imaging system (Bio-Rad). The antibodies used are specified in Supplementary Table 12. Normalization of protein levels was performed by densitometry (ImageLab software, Bio-Rad).
Immunofluorescence and H&E staining. For immunofluorescence imaging, cells were spun down using a Cytospin 4 (Thermo Fisher Scientific) cytocentrifuge at 600 r.p.m. for 5 min. Frozen sections were cut to 7-15 µm sections using a CryoStar NX70 (Thermo Fisher Scientific) cryostat. The cells and tissue sections were fixed in 4% paraformaldehyde for 20 min, washed three times in PBS-T (PBS with 0.1% Triton X-100) and permeabilized for 15 min using 0.3% Triton X-100. Blocking was performed for 1 h in 5% goat serum and the samples were incubated with primary antibodies overnight at 4 °C. The antibodies are listed in Supplementary  Table 12. After washing in PBS-T, the samples were incubated with secondary antibodies for 90 min at room temperature, washed 3× in PBS-T and the nuclei were stained for 10 min in 1 µg ml −1 Hoechst 33342 (Thermo Fisher Scientific), followed by additional wash steps. The cells and tissue sections were mounted using Roti-Mount Aqua (ROTH) medium. Images were obtained on a Zeiss LSM700 or LSM800 laser scanning microscope. For the H&E staining, organs were isolated from mice, fixed in 10% neutral buffered formalin (Sigma-Aldrich) and subsequently processed for paraffin embedding. Tissue sections were cut to 5-15 µm using a CUT5062 microtome (SLEE medical). The sections were washed twice for 5 min in xylene, followed by a 5 min incubation in a 90:10 xylene:ethanol mix. Subsequently, rehydration steps were performed in absolute ethanol (2×), 95% ethanol, 80% ethanol and 70% ethanol, 5 min each, followed by a brief rinse in water. The samples were incubated with haematoxilin (Life Technologies) for 2 min, followed by rinsing in water and 0.25% ammonia water. The slides were incubated for 1 min in 95% ethanol, followed by a 2 min stain in alcoholic eosin Y (Sigma-Aldrich). Dehydration was performed in 95% ethanol and absolute ethanol (3×; 1 min each), followed by incubation in xylene (3×; 1 min each). The sections were mounted with DPX (Sigma-Aldrich). Images were taken on a Zeiss Epi-Scope microscope.
Cytokine analysis. Blood samples were obtained from mice and spun at 1,000g for 10 min at 4 °C to collect the plasma. Cytokine concentrations were determined using a LEGENDplex mouse inflammation panel (BioLegend, 740446) and mouse HSC panel (BioLegend, 740677) according to the manufacturer's recommendations. Analysis was performed on a BD FACSymphony machine and using the LEGENDplex data analysis software (BioLegend).
Northern blotting. RNA was isolated from BM cells using a Direct-zol RNA miniprep plus kit (Zymo Research) according to the manufacturer's protocol. Subsequently, DNA was digested using a Turbo DNA-free kit (Thermo Fisher) following the manufacturer's protocol. The RNA was recovered and concentrated using an RNA clean and concentrator-5 kit (Zymo Research). Northern blot analysis was performed according to standard procedures. Briefly, 5 µg of total RNA was run on a 1% denaturing agarose gel. The RNA was transferred to a Hybond-N+ membrane (GE Healthcare) by capillary transfer and fixed by ultraviolet-light crosslinking. The membrane was pre-hybridized at 42 °C for 20 min and then hybridized in PerfectHyb plus hybridization buffer (Sigma) containing 1 × 10 6 cpm ml −1 of 32 P-labelled TERRA DNA probe 5′-(TAACCC)5-3′ and 0.1 mg ml −1 herring sperm DNA (Thermo Fisher). Hybridization was carried out overnight at 42 °C. The following day, the membrane was washed once in low stringency buffer (2×SSC and 0.1% SDS) at room temperature for 5 min and twice in high stringency buffer (0.5×SSC and 0.1% SDS) at 42 °C for 20 min. The membranes were exposed to muliautoradiography film at −80 °C. After exposure, the membrane was stripped in boiling stripping buffer (0.1% SDS and 5 mM EDTA) and re-hybridized using a GAPDH probe (5′-GTAGACCCACGACATACTCAGCACCGGCCTCACCCCATT-3′) as a loading control.
RNA-seq of cells isolated from transplanted animals. CD45.2 + KLS and GMP (Lin − c-Kit + Sca-1 − CD16/32 + CD34 + ) populations were sorted from the BM of transplanted mice. Total RNA was isolated from samples containing between 4,700 and 246,000 cells using TRIzol reagent (Invitrogen) and a miRNeasy micro kit (Qiagen) according to the manufacturer's protocol. The precipitated RNA was resuspended in RNase-free water and the RNA quantity and quality (RINe) were assessed via the HS RNA analysis screen tape assay on a 4200 TapeStation system (Agilent Technologies). The total RNA was converted into double-stranded complementary DNA libraries as the template for high-throughput sequencing using the Ovation SoLo RNA-seq kit (NuGEN Technologies). Briefly, the DNA was digested and first-strand DNA was synthesized using random hexamers. After second-strand synthesis, the ends were repaired, followed by adaptor ligation and PCR pre-amplification. NuGEN proprietary AnyDeplete probes were added to selectively block fragments originating from ribosomal RNA from amplification in the subsequent PCR reaction. Size-selection and purification of cDNA fragments of approximately 200-500 bp in length was performed using AMPure XP beads (Beckman Coulter). The size distribution of the cDNA libraries was measured using the HS D1000 DNA assay on a 4200 TapeStation system (Agilent Technologies). The cDNA libraries were quantified using a KAPA library quantification kit (Kapa Biosystems). After cluster generation on a cBot (Illumina), the libraries were sequenced in a paired-end 2 × 101 bp run on a HiSeq 1500 system (Illumina) using TruSeq v3 chemistry and de-multiplexed using CASAVA version 1.8.4. The experiment was run in duplicate.

RNA-seq of cells collected at steady state. LT-HSCs (Lin
and KLS cells were sorted from BM 3 d.p.i. (RosaCreER mice: LT-HSCs), 3 w.p.i. (Mx1Cre mice: GMPs and KLS, MPP3 and MPP4 cells) or 24 w.p.i. (Mx1Cre mice: KLS cells and GMPs). Total RNA was isolated from 2,000-5,000 cells using a NucleoSpin RNA XS plus kit (Macherey-Nagel) according to the manufacturer's instructions. The RNA quantity and quality (RINe) were assessed using the HS RNA analysis screen tape assay on a 4200 TapeStation system (Agilent Technologies). Total RNA was converted into double-stranded cDNA libraries as a template for high-throughput sequencing using the SMARTer stranded total RNA-seq kit v.2-pico input mammalian kit (TaKaRa Bio) according to the manufacturer's instructions. Briefly, after first-strand cDNA synthesis with SMARTer oligonucleotides, barcoded adaptors for Illumina sequencing were added via limited-cycle PCR. The PCR products were purified and ribosomal cDNA was then depleted. The remaining cDNA fragments were further amplified by PCR (15 cycles) with primers universal to all libraries. After purifying the resulting PCR products, the size distribution of the cDNA libraries was measured using the HS 5000 DNA assay on a 4200 TapeStation system (Agilent Technologies) and the cDNA concentrations were determined using a Qubit dsDNA HS assay kit (Thermo Fisher Scientific). After cluster generation, the samples were sequenced either as single-end 1 × 100 bp or paired-end 2 × 100 bp on a NovaSeq 6000 system (Illumina). The experiments were run in duplicate or triplicate.
ATAC-seq of RosaCreER mice at 3 w.p.i. For ATAC-seq, RosaCreER Daxx mice were killed 3 weeks after tamoxifen treatment. The BM was isolated from the femur and tibia, followed by RBC lysis and sorting of CMP (Lin -CD45.2 + c-Kit + Sca-1 -CD16/32 -CD34 + ) and GMP populations following antibody staining. Enrichment of c-Kit + cells using CD117 MicroBeads (Miltenyi Biotec) was performed before sorting LT-HSCs (Lin -CD45.2 + c-Kit + Sca-1 + CD48 -CD150 + ). Between 8,000 and 50,000 flow-sorted cells were collected in flow buffer (PBS, 2% FBS and 2 mM EDTA) and immediately processed following previously published protocols 67 . Briefly, the sorted cells were spun down at 500g for 5 min at 4 °C, washed once in cold PBS and spun down in 50 μl cold lysis buffer (10 mM Tris-HCl pH 7.4, 10 mM NaCl, 3 mM MgCl 2 and 0.1% IGEPAL CA-630) at 500g for 10 min at 4 °C. Immediately thereafter the tagmentation reaction was started by adding Nextera's Tn5 transposase (TDE1) in reaction buffer. The transposition reaction mix was incubated for 30 min at 37 °C, followed by DNA purification using a MinElute PCR purification kit (Qiagen). ATAC-seq libraries were generated from tagmented DNA by PCR amplification using Nextera index primers (Illumina) according to the manufacturer's protocol. The ATAC-seq libraries were purified using a PCR purification MinElute kit (Qiagen) and quantified using KAPA library quantification kits (Kapa Biosystems) and a D1000 assay on a Tapestation 4200 (Agilent). The libraries were sequenced in a paired-end 2 × 101 bp run on a HiSeq 1500 system (Illumina) using TruSeq v3 chemistry and de-multiplexed using CASAVA version 1.8.4. In addition, ATAC-seq libraries from CMP and GMP cells were sequenced as a single-read 75 bp, rapid run on the HiSeq 1500 system using HiSeq Rapid v.2 chemistry. The resulting data were also de-multiplexed using CASAVA version 1.8.4. The experiment was run in duplicates.
Statistics and reproducibility. Statistical analyses were performed using GraphPad Prism (GraphPad Software) or the R program. Summarized views on data that underlie the statistical tests as well as the exact P values are available in the source data. The statistical details for each experiment are also provided in the figure legends. Statistical tests were run as two-sided tests, when appropriate. Most experiments were performed independently at least twice; details are provided in the figure legends. The genomic assay data (Figs. 1,5,6,8 and Extended Data Figs. 1,7-10) are based on two or three biological replicates. The experimental data in Figs. 2g,i,k,p, 3b,f, 4e,f, 7a,d,i and Extended Data Figs. 2d, 4b,c, 8b,d,e are based on biological replicates (n is given in the figure legends) that were not repeated independently on a different day.
Reporting Summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
All sequencing data that support the findings of this study can be found at Gene Expression Omnibus under the accession number GSE119309. An overview of the genomics studies run as part of this study can be found in Supplementary  Table 11. Previously published sequencing data that were re-analysed here are available under the accession codes GSE60101 and GSE79180 (SRR2062971 and SRR2062968). The GTF and FASTA files used for Bioinformatics analysis (mm10, GENCODE release M14) can be downloaded from GENCODE (https:// www.gencodegenes.org/mouse/release_M14.html). All other data supporting the findings of this study are available from the corresponding author on reasonable request. Source data are provided with this paper. Fig. 2 | Daxx-deficient BM displays increased proliferation and altered stem/progenitor cell frequencies and numbers. a, Genotyping of Daxx +/+ ;Mx1Cre +/and Daxx F/F ;Mx1Cre +/mice untreated (-) or treated with pI:pC (+). Arrows indicate the wild-type, floxed and recombined alleles (n = 5 independent experiments). b, Daxx mRNA levels upon pI:pC treatment in bone marrow. Shown is relative expression of Daxx mRNA over Tbp mRNA levels (n = 3 mice per genotype). c, IF of Daxx in bone marrow, scale bar = 10 µm (n = 4 mice per genotype, two independent experiments). Inlay image zoomed on representative nuclei. d, Western blot of Daxx, alpha-tubulin and H3.3 in Lineage-negative and Lineage-positive bone marrow cells (n = 2 mice per genotype). e, Bone marrow cellularity counts before and after RBC lysis in male mice (n = 5 mice per gender, non-parametric Mann-Whitney test, boxand-whiskers plot: min to max (whiskers), 25 th and 75 th percentile and median). f, Bone marrow cellularity counts before and after RBC lysis in female mice (n = 5 mice per gender, non-parametric Mann-Whitney test, box-and-whiskers plot: min to max (whiskers), 25 th and 75 th percentile and median). g, Flow cytometry analysis of Ki-67 + cells in BM and HPCs (n = 5 mice per genotype, repeated in two independent experiments, non-parametric Mann-Whitney test). h,i, Flow cytometry analysis of HSC and MPP populations (n = 3 mice per genotype, repeated in two independent experiments, Student's t-test). BM, bone marrow; HPC, haematopoietic progenitor cells; HSC, haematopoietic stem cell; MPP, multipotent progenitors. Data in box plots are mean and min to max. ns, not significant; * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001. Exact p-values and numerical source data can be found in the accompanying source data. Unprocessed blots provided in Source data. Fig. 3 | Daxx deficiency results in myeloid/lymphoid imbalance in the bone marrow. a, Flow cytometry analysis of myeloid cell surface markers (n = 6 mice per genotype, repeated in three independent experiments). b, Flow cytometry analysis of lymphoid cell surface markers (n = 6 mice per genotype, repeated in three independent experiments). c, Frequencies of B cell progenitors (n = 6 mice per genotype, repeated in three independent experiments. d, Western blot of Daxx, H3.3 and corresponding β-actin control in isolated B220 + cells from bone marrow (n = 3 mice per genotype, repeated in two independent experiments). e, Bones isolated from two Daxx +/+ ;Mx1Cre and Daxx F/F ;Mx1Cre mice. f, Frequencies of erythroblast populations (n = 6 mice per genotype, repeated in three independent experiments. Data in box plots are mean and min to max. ns, not significant; * P < 0.05,d ** P < 0.01, *** P < 0.001, **** P < 0.0001, Wilcoxon rank test. Exact p-values and numerical source data can be found in the accompanying source data. Source image file provided in Source data. Fig. 6 | The haematopoietic phenotype of Daxx deletion is due to a cell-intrinsic defect. a, Number of colonies after first plating (P0) and first passage (P1) using haematopoietic progenitor cells (n = 5 mice per genotype, repeated in three independent experiments, box-and-whiskers plot: min to max, 25 th to 75 th percentile and median). b, Number of colonies after first plating (P0) and first passage (P1) using LT-HSCs (n = 2 mice per genotype). c, Number of cells after first plating (P0) and first passage (P1) using LT-HSCs (n = 2 mice per genotype). d, Number of KLS cells and e, number of CD11b + cells and neutrophil-like cells (Ly6C + ,Ly6G + ) after first plating (P0) of LT-HSCs and first passage (P1; n = 2 mice per genotype). f, Schematic representation of the experimental set-up to generate bone marrow chimeras. g, Frequencies of CD45.1 + and CD45.2 + cells in transplanted mice (n = 8 mice per genotype). h, WBC counts in total bone marrow (n = 8 mice per genotype, box-and-whiskers plot: min to max, 25 th to 75 th percentile and median). i, Flow cytometry analysis in peripheral blood of transplanted mice at two different time points: 1 st : 4-5 weeks post transplantation; 2 nd pooled data from two independent experiments collected at 8 or 12 weeks post transplantation (n = 8 mice per genotype). j, Frequencies of mature CD45.2 + blood cells in bone marrow of recipient mice (n = 8 mice per genotype, repeated in three independent experiments). k, Frequencies of CD45.2 + KLS and GMPs in bone marrow (n = 4 mice per genotype, repeated in two independent experiments). l, Flow cytometry analysis of CD45.2 + mature blood cells in spleen of recipient mice (n = 8 mice per genotype, repeated in three independent experiments). BM, bone marrow; PB, peripheral blood; KLS, c-Kit pos , lineage neg , Sca-1 pos ; GMP, granulocyte-monocyte progenitor; P0, first plating; P1, first passage. Data in box plots are mean and min to max. ns, not significant; * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001, Wilcoxon rank test. Exact p-values and numerical source data can be found in the accompanying source data. Fig. 7 | Gene expression changes in Daxx-deficient haematopoietic progenitors. a, PCA of the top-500 most variable genes for GMPs. b, Trajectory of gene expression of KLS collected at 3 w.p.i., 24 w.p.i. or from transplanted animals. c, Boxplot showing mean pseudotime for each sample group for GMPs (n = 2 mice per genotype). d, Heatmap showing scaled expression of selected transcription factors and regulators involved in blood cell differentiation (n = 2 mice per genotype). e, Top 5 haematologic biofunctions and diseases altered at 3 w.p.i. in KLS Daxx KO cells. f, Top 5 haematologic biofunctions and diseases altered at 24 w.p.i. in KLS Daxx KO cells. g, Top 5 haematologic biofunctions and diseases altered in KLS Daxx KO cells isolated from transplanted animals. h, Top 5 haematologic biofunctions and diseases altered in GMP Daxx KO cells collected at 3 w.p.i., 24 w.p.i. and from transplanted animals. i, Boxplots of mean normalized read counts of ERV subtypes and satellite repeats (n = 2 mice per genotype) in GMP cells. j, Heatmap showing scaled expression of interferon response genes and the dsRNA recognition machinery for GMPs. Data in box plots are min to max and mean. Numerical source data can be found in the accompanying source data. μ μ