ARTICLE

https://doi.org/10.1038/s41467-020-17630-6 OPEN Histone deacetylase 3 controls lung alveolar macrophage development and homeostasis

Yi Yao1,2,10, Queping Liu1,2,9,10, Indra Adrianto 1,2,3,4, Xiaojun Wu1,2, James Glassbrook1,2,5, Namir Khalasawi1,2, ✉ Congcong Yin 1,2, Qijun Yi1,2, Zheng Dong 6, Frederic Geissmann 7, Li Zhou 1,2,5,8 & ✉ Qing-Sheng Mi 1,2,5,8

Alveolar macrophages (AMs) derived from embryonic precursors seed the lung before birth

1234567890():,; and self-maintain locally throughout adulthood, but are regenerated by bone marrow (BM) under stress conditions. However, the regulation of AM development and maintenance remains poorly understood. Here, we show that histone deacetylase 3 (HDAC3) is a key epigenetic factor required for AM embryonic development, postnatal homeostasis, matura- tion, and regeneration from BM. Loss of HDAC3 in early embryonic development affects AM development starting at E14.5, while loss of HDAC3 after birth affects AM homeostasis and maturation. Single-cell RNA sequencing analyses reveal four distinct AM sub-clusters and a dysregulated cluster-specific pathway in the HDAC3-deficient AMs. Moreover, HDAC3- deficient AMs exhibit severe mitochondrial oxidative dysfunction and deteriorative cell death. Mechanistically, HDAC3 directly binds to Pparg enhancers, and HDAC3 deficiency impairs Pparg expression and its signaling pathway. Our findings identify HDAC3 as a key epigenetic regulator of lung AM development and homeostasis.

1 Center for Cutaneous Biology and Immunology Research, Department of Dermatology, Henry Ford Health System, Detroit, MI 48202, USA. 2 Immunology Research Program, Henry Ford Cancer Institute, Henry Ford Health System, Detroit, MI 48202, USA. 3 Department of Public Health Sciences, Henry Ford Health System, Detroit, MI 48202, USA. 4 Center for Bioinformatics, Henry Ford Health System, Detroit, MI 48202, USA. 5 Department of Biochemistry, Microbiology, and Immunology, School of Medicine, Wayne State University, Detroit, MI 48202, USA. 6 Department of Cellular Biology and Anatomy, Augusta University, Augusta, GA 30912, USA. 7 Immunology Program, Sloan Kettering Institute, Memorial Sloan Kettering Cancer Center, New York, NY 10065, USA. 8 Department of Internal Medicine, Henry Ford Health System, Detroit, MI 48202, USA. 9Present address: Department of Pathology, Xiangya ✉ Hospital of Central South University, Changsha, Hunan 410008, China. 10These authors contributed equally: Yi Yao, Queping Liu. email: [email protected]; [email protected]

NATURE COMMUNICATIONS | (2020) 11:3822 | https://doi.org/10.1038/s41467-020-17630-6 | www.nature.com/naturecommunications 1 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-17630-6

lveolar macrophages (AMs), the resident macrophages in four unique sub-populations and that a lack of HDAC3 gives rise Alung alveoli, are important for the maintenance of to AM subset-specific pathway dysregulation. Moreover, loss of homeostasis in the airways and are involved in the HDAC3 results in dysregulated mitochondrial oxidative phos- development of a variety of pulmonary diseases1. Recent lineage- phorylation and cell survival. Mechanistically, HDAC3 directly tracing studies have demonstrated that tissue-resident macro- binds to Pparg enhancers and regulates PPAR-γ signaling phages (TRMs) originate from embryonic yolk sac (YS) erythro- during AM development. Our findings uncover HDAC3 as a key myeloid progenitors (EMPs) and can self-renew in most adult epigenetic factor in the regulation of lung AM embryonic tissues at steady state without contribution from bone marrow development and maintenance after birth. (BM) hematopoietic stem cells (HSCs)2–4. YS EMP cells differ- entiate into premacrophages (pMacs) that colonize all embryonic organs as TRMs starting at embryonic day 9.5 (E9.5)5. The YS Results EMPs also enter the fetal liver (FL) starting from E12.5, where HDAC3 is required for the embryonic development of AMs. they expand and differentiate into pMacs and/or monocytes that Before birth, F4/80hiCD11bint pMac-derived fetal macrophages5 travel to local tissues through the blood and eventually become and F4/80intCD11bhi FL monocytes sequentially colonize the tissue-specific TRMs4. Fetal monocytes begin to accumulate in developing lung around E12.5 and E14.5, respectively8. Fetal the developing lung at E14.5 and then differentiate into immature monocytes further differentiate into F4/80intCD11bint preAMs, AMs (preAMs) expressing F4/80intCD11bint, which postnatally which become CD11chiSiglec-FhiCD11blo AMs during the first mature into AMs expressing CD11chiSiglec-FhiCD11blo,4,6,7. Like week of life. To investigate the role of HDAC3 in AMs, we first the majority of other TRMs, AMs self-maintain locally at steady examined the expression pattern of HDAC3 in the lung fetal state throughout adult life4,8. Under stressed conditions, AMs are macrophages and preAMs during embryonic development, as either self-maintained by local proliferation or completely well as AMs at a young age and adulthood, using qRT-PCR. As replaced by bone marrow-derived cells in a challenge-dependent shown in Fig. 1a, HDAC3 was expressed in lung fetal macro- manner9,10. phages at E14.5 and in preAMs at E16.5 and E18.5, but its A few have been identified as requirements for AM expression was dramatically increased in AMs from the young development and homeostasis. Granulocyte macrophage colony- (P14, ~20-fold) and adult (~40-fold) mice. These results suggest stimulating factor (GM-CSF) was shown to be essential for the that HDAC3 is potentially involved in AM embryonic develop- perinatal differentiation of fetal monocytes into preAMs and for ment and maintenance after birth. the full maturation of AMs postnatally7. Transforming growth To determine whether HDAC3 regulates embryonic develop- factor-β (TGF-β) is also crucial for the embryonic differentiation ment of AMs, we generated Csf1riCreHdac3fl/fl conditional of fetal monocytes into preAMs, their postnatal maturation, as knockout (cKO) mice, in which HDAC3 is deficient in the well as the homeostasis of adult AMs11. Both GM-CSF and TGF- Csf1r-expressing myeloid cells, including TRMs and β signaling pathways induce the expression of PPAR-γ, a sig- monocytes24,25. Both HDAC3 mRNA and were efficiently nature transcription factor essential for the development of depleted (>80%) in preAMs from the HDAC3cKO newborns as AMs7,11. Macrophage identity and homeostasis require the pre- determined by qRT-PCR and flow cytometry, respectively (Fig. 1b, cise regulation of gene expression that is governed by different c). We next examined the fetal macrophages, preAMs, and fetal epigenetic mechanisms, such as DNA methylation, histone monocytes in the lung from Csf1riCreHdac3fl/fl cKO mouse modification, and chromatin structure12. However, very little is embryos and wild type (WT) littermates between E12.5 and right known about the epigenetic factors that regulate AM develop- after birth (P0). There was no significant alteration in the ment from EMPs and fetal monocytes, as well as its postnatal frequencies of F4/80hiCD11blo fetal macrophages and F4/ homeostasis. 80loCD11bhi monocytes in the lung between cKO and WT Histone deacetylases (HDACs) are enzymes that regulate gene embryos at E12.5 (Fig. 1d, e, gates shown in Supplementary expression by modifying chromatin structure through the Fig. 1a). However, a modest but significant reduction in the removal of acetyl groups from target histones or through direct frequency of fetal macrophages was observed in HDAC3cKO deacetylation of non-histone . HDACs exhibit limited embryos at E14.5 and E16.5, followed by a gradual diminish substrate specificity and rely on transcription factors with specific afterwards. On the other hand, the frequency of F4/80intCD11bint DNA-binding and/or chromatin-binding activities in order to preAM cells that emerged at E16.5 in the fetal lung was target specific genes13. Class I HDACs (HDAC1, 2, 3, 8) can dramatically decreased in HDAC3cKO compared to WT assemble into multi-component co-repressor13 or co-activator14 littermates starting at E16.5 until birth. In contrast, fetal complexes that regulate the transcription of a broad array of monocytes showed compensatorily increased distribution in the genes which fundamentally impact cellular physiology, organism absence of HDAC3 starting at E14.5. Immunofluorescence development, and disease pathogenesis15–18. Several pan HDAC staining of lung tissue sections at E18.5 further confirmed that inhibitors (HDACi) have been approved by the FDA for anti- the number of lung preAMs was significantly reduced in the tumor therapy19. More recently, HDACi was suggested for use in HDAC3cKO mice (Fig. 1f). Collectively, these results suggest that the treatment of asthma and other inflammatory lung HDAC3 is indispensable for the development of YS pMac-derived diseases20,21. HDAC3, belonging to the class I HDAC family, is lung macrophages and FL monocyte-derived preAMs during required for the macrophage inflammatory gene expression embryogenesis. program22 and serves as an epigenetic brake in macrophage Next, we investigated whether HDAC3 deficiency blocks EMP alternative activation23. However, it remains unknown whether development and differentiation into pMacs and FL monocytes in HDAC3 is involved in the ontogeny and homeostasis of AMs. the YS and FL, respectively. Interestingly, within the CD45+Lin– In this study, we aim to investigate the role of HDAC3 in the (CD45-expressing; lineage negative) population, the frequencies embryonic development and postnatal maintenance of AMs. of CD117hiF4/80– EMPs, CD117–F4/80– pMacs, and CD117–F4/ Using mice with TRM or AM lineage-specific or time-specific 80+ macrophages in the YS remained unaltered in HDAC3cKO HDAC3 deletion and the bone marrow chimeric mouse model, embryos compared to the WT at E9.5, E10.5, and E12.5 (Fig. 1g, we demonstrate that HDAC3 deficiency leads to severe impair- h, gates shown in Supplementary Fig. 1b), suggesting that ment of AM ontogeny, maintenance, maturation, and regenera- HDAC3 is dispensable for EMPs and their further differentiation tion. Single-cell RNA sequencing analyses reveal that AMs have into pMacs and YS macrophages. Furthermore, we observed

2 NATURE COMMUNICATIONS | (2020) 11:3822 | https://doi.org/10.1038/s41467-020-17630-6 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-17630-6 ARTICLE

a 60 b 1.5 P = 0.0067 d e fl/fl iCre fl/fl Hdac3 Csf1r Hdac3 Hdac3fl/fl 40 3.25 6.40 Csf1r iCreHdac3fl/fl 1.0 90

20 E12.5 MF Int *** 3 0.5 60 74.8 73.4 2 CD11b mRNA expression Relative expression 1 hi 30 0 0.0 ** fl/fl iCre F4/80 fl/fl E14.5 P14 35.3 47.7 0 E14.5 E16.5 E18.5 Adult Hdac3 Csf1r 12 14 16 18 Hdac3 58.3 39.7 c P = 9.7e–7 80 Rabbit IgG 100 100 16.4 52.9 WT 83.6% preAM 60 *** 80 KO 10.7% Int *** 75 E16.5 40

60 preAM + 50 51.0 13.5 *** CD11b

40 18.8 15.0 Int 20 25

HDAC3 19.2 31.4 Events (% of max) 20 0 F4/80 0 16 18 NB 0 E18.5 3 4 fl/fl iCre 90 01010 r fl/fl

3.06 MO HDAC3 41.8 Hdac3 Csf1 3.50 1.22 hi Hdac3 60 *** f fl/fl iCre fl/fl *** *** Hdac3 Csf1r Hdac3 26.0 46.2 *** 300 CD11b 2

P = 3.8e–4 Int 30 P0 200 F4/80 0

CD11b 29.9 9.78 12 14 16 18 NB 100 F4/80 Days of gestation Cell number/mm F4/80 0 DAPI fl/fl iCre fl/fl Hdac3 Csf1r Hdac3

g Yolk Sac Liver Hdac3fl/fl Csf1r iCreHdac3fl/fl Hdac3fl/fl Csf1r iCreHdac3fl/fl Yolk Sac

76.8 79.5 EMP h Hdac3fl/fl 4.46 5.70 N/A iCre fl/fl E9.5 Csf1r Hdac3 pMac MF Yolk Sac 14.5 12.4 Liver 100 30 100 100 28.4 27.3 53.5 59.5 MF EMP EMP pMac + 63.9 62.3 – 75 75 – 75 E10.5 – 20 50 F4/80 50 50 F4/80 F4/80 – 6.70 F4/80

8.47 hi hi – 10 25 25 12.0 10.2 73.8 75.7 25 CD117 CD117 CD117 81.9 82.8 0 CD117 0 0 0 E12.5 10 12 10 12 10 12 10 12 CD117 3.01 3.92 Days of gestation F4/80

i fl/fl iCre fl/fl j Hdac3 Csf1r Hdac3 Hdac3fl/fl Csf1r iCreHdac3fl/fl 20 E14.5 P = 0.48 P = 0.69

7.43 7.28 15 MO (%) hi 10 CD11b

lo 5 E18.5 CD11b CD64 0 8.91 8.55 E14.5 E18.5 CD64 comparable frequencies of EMPs in the FL between the spleen, kidney, and pancreas (Supplementary Fig. 3a, b, gates HDAC3cKO and WT embryos at E10.5 and E12.5, suggesting shown in Supplementary Fig. 1a). Overall, HDAC3 deficiency that the loss of HDAC3 also does not affect EMP seeding in the does not affect EMPs, their ability to differentiate into pMacs in FL (Fig. 1g, h). In addition, the frequency of CD64loCD11bhi- the YS, or FL monocyte development, but it may play key roles in Ly6chi FL monocytes remained unaltered during the embryonic pMac-derived macrophages and FL monocyte-derived preAMs in stage (E14.5-E18.5) in the HDAC3cKO mice (Fig. 1i, j, gates the lung. shown in Supplementary Fig. 1c). HDAC3 deletion in FL fi fl monocytes, which was con rmed by qRT-PCR and ow HDAC3 is essential for the maintenance of AMs after birth. cytometry (Supplementary Fig. 2a, b), does not disrupt their We next examined the AMs from the Csf1riCreHdac3fl/fl cKO adult differentiation into TRMs in other organs, such as the brain, mice. As expected, the frequency of lung AMs (CD11chiSiglec-Fhi)

NATURE COMMUNICATIONS | (2020) 11:3822 | https://doi.org/10.1038/s41467-020-17630-6 | www.nature.com/naturecommunications 3 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-17630-6

Fig. 1 HDAC3 is required for embryonic development of AMs. a qRT-PCR analysis of HDAC3 mRNA expression in lung MFs, preAMs, or AMs from C57BL/6 mice (n = 4 for postnatal day 14 (P14), n = 3 for each other time point). The expression of mRNA (b) and protein (c) of HDAC3 in preAMs determined by qRT-PCR and flow cytometry, respectively, from n = 3 Hdac3fl/fl;Csf1riCreHdac3fl/fl newborns (P0-P1). Frequencies of HDAC3-expressing (HDAC3+) AMs are shown in c. d Representative flow cytometry plots: gated from CD45+ live cells of fetal lung. e Frequencies of lung MFs, preAMs, and MOs as in d. E12.5: n = 14 Hdac3fl/fl, n = 5 Csf1riCreHdac3fl/fl; E14.5: n = 5 Hdac3fl/fl, n = 11 Csf1riCreHdac3fl/fl, ***P = 6.3e−7 (MF), 9.0e−6 (MO); E16.5: n = 12 Hdac3fl/fl, n = 9 Csf1riCreHdac3fl/fl,**P = 7.1e−3 (MF), ***P = 2.1e−13 (preAM), 5.3e−16 (MO); E18.5: n = 14 Hdac3fl/fl, n = 19 Csf1riCreHdac3fl/fl, ***P = 5.7e–14 (preAM), 6.5e−7 (MO); newborns (NB): n = 12 Hdac3fl/fl, n = 6 Csf1riCreHdac3fl/fl, ***P = 2.0e−4 (preAM), 2.9e−5 (MO). f Lung tissue sections from E18.5 embryos as in d. Scale bars, 100 μm. Cell numbers (n = 5 Hdac3fl/fl;Csf1riCreHdac3fl/fl) are shown per tissue area (mm2). g Representative flow cytometry plots: gated from CD45+Lin‒ (lineage negative) cells in YS and CD45+ live cells in fetal liver. h Frequencies of EMPs, pMacs, and MFs as in g. E9.5: n = 13 Hdac3fl/fl, n = 11 Csf1riCreHdac3fl/fl; E10.5: n = 6 Hdac3fl/fl, n = 10 Csf1riCreHdac3fl/fl; E12.5: n = 8 Hdac3fl/fl, n = 5 Csf1riCreHdac3fl/fl. i Representative flow cytometry plots: gated from CD45+ live cells in fetal liver. j Frequencies of MO as in i. E14.5: n = 8 Hdac3fl/fl, n = 15 Csf1riCreHdac3fl/fl; E18.5: n = 6 Hdac3fl/fl, n = 8 Csf1riCreHdac3fl/fl. Each dot represents one embryo or newborn and bars represent mean ± SD of biologically-independent samples in each panel. All P values obtained by Student’s two-tailed unpaired t test. MF fetal macrophages, MO monocytes. Source data are provided as a Source Data file.

a fl/fl iCre fl/fl b fl/fl iCre fl/fl Hdac3 Csf1r Hdac3 Hdac3 Csf1r Hdac3 P = 3.1e–7 P = 1.6e–5 40 200 33.9 7.49 2 30 150

20 100

AM (%) 10 50 Cell number/mm

CD11c Siglec F 0 DAPI 0 fl/fl fl/fl iCre iCre fl/fl Siglec F fl/fl Csf1r Hdac3 Csf1r Hdac3 Hdac3 Hdac3

Cre fl/fl c d Hdac3 fl/fl CD11c Hdac3 30 60 0.6 P = 9.7e–4 P = 1.0e–8 P = 3.5e–10 ) ) 3 20 40 6 0.4

26.2 12.4 10 AM (%) 20 0.2 AM (×10 MFI (×10 Events (% of max) 0 CD11c 0 0

fl/fl Cre fl/fl Cre fl/fl iCre fl/fl Siglec F fl/fl fl/fl CD11b 3 3 3 Hdac3 Hdac3 Hdac3 Csf1r CD11c CD11c Hdac Hdac Hdac

e P = 9.5e–8 f Hdac3 fl/fl UbcCreERHdac3 fl/fl 30 40 P = 0.004 1.2 P = 0.03 )

) 30 6 3 20 0.8 20 23.4 9.48 10 AM (%) 0.4 10 AM (×10 MFI (×10 Events (% of max) 0 CD11c 0 0

fl/fl fl/fl fl/fl Cre CreER fl/fl 3 CreER fl/fl fl/fl Siglec F CD11b 3 Hdac3 Ubc Hdac Ubc Hdac3 CD11c Hdac3 Hdac3 Hdac

Fig. 2 HDAC3 is essential for the maintenance of mature AMs in the adult mice. a Flow cytometry for CD11c and Siglec-F expression of lung AMs within CD45+ live cells from n = 7 Hdac3fl/fl; Csf1riCreHdac3fl/fl adult mice. Frequencies of AMs are shown on the right. b Lung tissue sections from adult mice as in a. Scale bars, 100 μm. Cell numbers (n = 5 Hdac3fl/fl;Csf1riCreHdac3fl/fl) are shown per tissue area (mm2). c Histogram plot for CD11b expression of lung AMs from n = 7 Hdac3fl/fl;Csf1riCreHdac3fl/fl adult mice. MFI, mean fluorescence intensity. d Flow cytometry analysis of AMs from n = 10 Hdac3fl/fl, n = 17 Cd11cCreHdac3fl/fl adult mice. Frequencies and absolute numbers of AMs are shown on the right. e Histogram plot for CD11b expression of lung AMs from n = 11 Hdac3fl/fl, n = 15 Cd11ccreHdac3fl/fl adult mice. MFI of CD11b is shown on the right. f Flow cytometry analysis of AMs from n = 8 Hdac3fl/fl; UbcCreERHdac3fl/fl adult mice. Frequencies and absolute numbers of AMs are shown on the right. Each dot represents one mouse and bars represent mean ± SD of biologically-independent samples. All P values were obtained using the Student’s two-tailed unpaired t test. Source data are provided as a Source Data file. was markedly decreased in the HDAC3cKO adult mice (Fig. 2a, HDAC3. However, given that Csf1r.Cre-induced gene deletion gates shown in Supplementary Fig. 1d). This reduction was fur- occurs in the embryonic stage, the AM defects observed in the ther confirmed by immunofluorescence staining of frozen tissue Csf1riCreHdac3fl/fl cKO adult mice may be due to embryonic sections (Fig. 2b). After birth, preAMs quickly downregulate their developmental dysregulation rather than a maintenance CD11b expression and become mature AMs7. As shown in abnormality after birth. Fig. 2c, the remaining AMs in the HDAC3-deficient mice were To address this question, we next generated mice with a found to highly express CD11b. This suggests that maintenance constitutive HDAC3 deficiency in AMs postnatally by crossing of mature AMs in the adult mice is very likely dependent on Hdac3fl/fl mice with CD11cCre transgenic mice, given that CD11c

4 NATURE COMMUNICATIONS | (2020) 11:3822 | https://doi.org/10.1038/s41467-020-17630-6 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-17630-6 ARTICLE

a Donor Competitor Hdac3 fl/fl or Recipient Csf1r iCreHdac3 fl/fl B6.SJLhetero Recipient CD45.2+ 10 weeks CD45.1 Donor Competitor CD45.1+CD45.2+ B6.SJL WT CD45.2 CD45.1+

b Lung P = 1.6e–8 P = 3.5e–8 WT : Hdac3fl/fl 5 100 10 105 iCre fl/fl 4 9.51 WT : Csf1r Hdac3 10 104 90 fl/fl 3 12.1 Hdac3 10 103 0.39 0 0 90.0 80 10

5 AM (%) 0 103 104 10 0 103 104

5 Csf1r iCre 99.6 Hdac3fl/fl 7.76 0.24 0 + – + CD45.1 CD11c 0.09

Siglec F CD45.2 – CD45.2 + CD45.2 + CD45.2

Donor CD45.1 CD45.1Competitor CD45.1 Recipient

Fig. 3 Regeneration of AMs from bone marrow depends on HDAC3. The bone marrow (BM) cells from Hdac3fl/fl and Csf1ricreHdac3fl/fl mice (CD45.2+) were co-transferred with competitor BM cells from B6.SJL mice (CD45.1+) into lethally-irradiated B6.SJLhetero mice (CD45.1+CD45.2+). The lungs were harvested from the recipient mice 10 wks after reconstitution. a Schematic representation of the BM chimeric mouse working models. b Flow cytometry of CD45.1 and CD45.2 expression on CD11chiSiglec-Fhi lung AMs from n = 3 WT: Hdac3fl/fl, n = 4 WT: Csf1ricreHdac3fl/fl mice. Bars represent mean ± SD of biologically-independent samples. All P values were obtained using the Student’s two-tailed unpaired t test. Source data are provided as a Source Data file. is highly expressed in AMs after birth7. We found that the failed to reconstitute lung AMs compared to the BM from WT frequency and number of AMs were dramatically reduced in mice. These findings suggest that HDAC3 is not only important CD11cCreHdac3fl/flcKO mice compared to their WT counterparts for embryonically-derived AMs, but also required for the regen- (Fig. 2d), and that CD11b was highly expressed on AMs from eration of AMs from adult BM under stress conditions. CD11cCreHdac3fl/flcKO mice (Fig. 2e). We further generated inducible HDAC3-deficient mice by crossing Hdac3fl/fl mice with ubiquitin C promoter-driven, tamoxifen-inducible Cre (UbcCreER) HDAC3 regulates AM subset-specific gene expression profile. transgenic mice, in which tamoxifen (TAM) timely induces Recent single-cell studies have shown that heart and liver TRMs HDAC3 deletion. UbcCreERHdac3fl/fl mice (6–8 weeks old) were are heterogeneous populations at steady state27,28. To characterize treated with TAM for 5 consecutive days, then lung AMs were whether lung AMs are also heterogeneous, we sorted CD11chi- analyzed 1 week and 3–5 weeks after TAM treatment. Successful Siglec-Fhi AMs from Hdac3fl/fl WT adult mice and performed HDAC3 deletion in the targeted cells was confirmed by qRT-PCR single-cell RNA-sequencing (scRNA-seq) using the 10× Geno- (Supplementary Fig. 4a). Although normal distribution of AMs mics platform. More than 2000 cells were analyzed and T- was observed in UbcCreERHdac3fl/fl mice at 1 week post-TAM distributed stochastic neighbor embedding (t-SNE) dimension- treatment (Supplementary Fig. 4b), lack of HDAC3 led to a ality reduction analysis identified 4 major AM clusters (Fig. 4a). significant reduction in the frequency and number of lung AMs at Each AM sub-cluster has a unique gene expression pattern 3–5 weeks post-treatment (Fig. 2f). Taken together, these findings (Fig. 4b) with their signature genes (Fig. 4c). Of the AM clusters, suggest that HDAC3 also plays a key role in the maintenance of cluster 1 was the most abundant and expressed macrophage- AM homeostasis, as well as their maturation after birth at associated genes, such as Abcg1, Mrc1, and Mpeg1. Cluster 2 steady state. expressed genes such as Birc5, Top2a, Rrm2, and Stmn1, which are involved in the cell cycle and mitosis. Cluster 3 had relatively higher expression of antigen-presentation genes, such as H2-Eb1, Regeneration of AMs from the BM depends on HDAC3. AMs H2-Ab1, H2-Aa, and Cd74. Cluster 4 was enriched in many genes self-maintain at steady state independently of circulating pre- associated with immune response, such as Tnf, Il1a, Cxcl1, Cxcl2, cursors9. However, under certain inflammatory conditions or BM and Cxcl3. The cluster-specific gene expression pattern suggests transplantation following lethal whole-body irradiation, BM- that each AM cell subset may have a distinct role in AM main- derived monocytes can repopulate the AM niche9,26. We next tenance and function. sought to determine whether BM-derived AMs are also depen- In order to determine whether HDAC3 is required for the dent on HDAC3. To do this, we established a BM chimeric mouse fi fl fl maintenance of a speci c lung AM cell subset, we next performed model by co-transferring BM cells from Hdac3 / or Csf1ri- scRNA-seq on the sorted AM cells from Cd11cCreHdac3fl/flcKO fl fl + CreHdac3 / cKO donors (CD45.2 ) with competitor BM cells mice and compared the AM clusters between HDAC3cKO and from B6.SJL WT mice (CD45.1+) into the irradiated B6.SJLhetero WT mice. Although the composition of clusters 1 and 2 remained (CD45.1+CD45.2+) recipient mice (Fig. 3a). As expected, in the almost unaltered in the KO, the frequency of cluster 3 was chimeric mice, lung AMs were radiosensitive and largely replaced increased from 3.3% to 4.9%, while the frequency of cluster 4 was by BM-derived cells from both WT donor and competitor mice reduced from 4.0% to 2.4% in the absence of HDAC3 (Fig. 4d, e). (Fig. 3b). However, the BM from HDAC3cKO mice completely This suggests that the loss of HDAC3 may affect the maintenance

NATURE COMMUNICATIONS | (2020) 11:3822 | https://doi.org/10.1038/s41467-020-17630-6 | www.nature.com/naturecommunications 5 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-17630-6

a b Cluster 1234 40 Cluster 1 Cluster 2 Cluster 3 Cluster 4

20

0 tSNE2

–20

–40 –20 020 Expression –2 –1 012 tSNE1

c Abcg1 Birc5 H2-Eb1 Tnf d e 5 5 5 5 4 4 4 4 40 Hdac3 l/fl 3 3 3 3 Cre l/fl 2 2 2 2 Cd11c Hdac3 1 1 1 1 0 0 0 0 Hdac3fl/fl Mrc1 Top2a H2-Ab1 Egr1 5 5 5 5 20 4 4 4 4 3 3 3 3 2 2 2 2 1 1 1 1 Cd11c iCre 0 0 0 0 Hdac3fl/fl Mpeg1 Rrm2 H2-Aa Il1a 0

5 5 5 5 tSNE2 4 4 4 4 0 255075100 3 3 3 3 2 2 2 2 1 1 1 1 Log (normalized UMI counts) 0 0 0 0 –20 Cluster 1 Cluster 2 Ltc4s Stmn1 Cd74 Cxcl2 5 5 5 5 Cluster 3 4 4 4 4 3 3 3 3 Cluster 4 2 2 2 2 1 1 1 1 –40 0 0 0 0 1234 1234 1234 1 234 –20 020 Cluster tSNE1

f Cluster 2 Hdac3 l/fl Cluster 1 Cd11cCreHdac3 l/fl Phagosome Cell cycle control of maturation chromosomal 4 replication Antiproliferative role Unfolded protein of TOB in T cell 3.5 30 response Mismatch repair in Role of BRCA1 in DNA signaling 3 eukaryotes 25 damage response 2.5 20 2 15 1.5 Mitochondrial IL-8 signaling Estrogen-mediated S- Mitotic roles of polo- dysfunction 10 1 phase entry like kinase 0.5 5

0 0

LPS/IL-1 mediated Role of CHK proteins inhibition of RXR BAG2 signaling NER pathway in cell cycle function pathway checkpoint control

Cell Cycle: G2/M DNA damage checkpoint PPARα/RXRα ATM signaling Neuroinflammation regulation Activation signaling pathway Hereditary breast Protein kinase A cancer signaling signaling

Cluster 3 Cluster 4

EIF2 signaling IL-10 signaling 10 25 Th1 and Th2 activation 9 Acute phase B cell development IL-6 signaling 20 pathway 8 response signaling 7 15 6 5 Systemic lupus 10 4 Granulocyte adhesion erythematosus In B Th2 pathway TREM1 signaling 3 and diapedesis cell signaling pathway 5 2 1 0 0

Altered T cell and B Role of Dendritic cell cell signaling in Hypercytokinemia/hyp Agranulocyte adhesion maturation rheumatoid arthritis erchemokinemia in the and diapedesis pathogenesis of…

T helper cell Communication Th1 pathway Neuroinflammation differentiation between innate and signaling pathway adaptive immune cells Antigen presentation pathway Toll-like receptor signaling

Fig. 4 HDAC3 regulates AM subset-specific gene expression profiles. CD11chiSiglec-Fhi lung AMs were sorted from n = 1 Hdac3fl/fl;Cd11cCreHdac3fl/fl adult mouse for scRNA-seq analysis. a t-SNE map of individual cells clustered from 2055 Hdac3fl/fl AM cells. b Heat map representing signature genes which were most highly expressed (top 10) in each AM cluster. c Representative violin plots showing signature genes expressed by individual AM subsets. d t- SNE maps of individual cells from the Hdac3fl/fl and Cd11cCreHdac3fl/fl (3067) AM cells. e The percentile of each AM subset within total AM population between two groups. f Radar maps depicting dysregulated pathways between two groups in each AM subset. Source data are provided as a Source Data file.

6 NATURE COMMUNICATIONS | (2020) 11:3822 | https://doi.org/10.1038/s41467-020-17630-6 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-17630-6 ARTICLE of AM subsets in a cluster-specific manner. In addition, we also Fig. 7a, b) or enriched in HDAC3cKO phenotype (NES > 0, observed certain cluster-specific transcriptomic changes between FDR < 0.05) (Supplementary Fig. 8a, b). Notably, the genes WT and cKO mice (Supplementary Fig. 5). Ingenuity pathway involved in oxidative phosphorylation were significantly down- analysis (IPA) revealed that each AM cluster in the cKO mice regulated in HDAC3-deficient preAMs (Fig. 5d). Many subunits exhibited distinct dysregulated pathways compared to that in of protein complexes I-V of the , the key their WT counterpart (Fig. 4f). In cluster 1, the key disordered component of the oxidative phosphorylation process, were pathways included phagosome maturation, unfolded protein diminished in the absence of HDAC3 (Fig. 5e, f). In addition, response, mitochondria dysfunction, and BAG2 signaling path- we observed a significant decrease in Mitotracker Orange ways, which are associated with macrophage phagocytosis, fluorescence-labeled AMs from cKO, which reflects the reduced polarization, survival, and stress responses. In cluster 2, the mitochondria membrane potential and indicates impaired pathways related to cell cycle control were less enriched in the mitochondria function in HDAC3-deficient lung preAMs cKO cells. In cluster 3, several immune-responsive pathways were (Fig. 5g). Moreover, GSEA demonstrated that the absence of upregulated in the cKO cells, including E1F2 signaling, Th1/Th2 HDAC3 led to a strong enrichment of genes involved in apoptosis pathways, and dendritic cell maturation. In cluster 4, the signaling (Fig. 5h–j). Enhanced apoptosis of preAMs at E18.5 was pathways involved in cytokine signaling (i.e., IL-6 and IL-10 further confirmed by flow cytometry (Fig. 5k). These results signaling) and acute phase response signaling were less enriched suggest that HDAC3 regulates preAM mitochondrial oxidative in the cKO cells, whereas Toll-like receptor signaling, neuroin- function and cell death, which may be related to defective flammation signaling pathway, and communication between embryonic AM development in HDAC3 cKO mice. innate and adaptive immune cells, were strongly enriched in fi the cKO cells. These data reveal that HDAC3 de ciency results in HDAC3 deficiency causes impaired PPAR-γ signaling in pre- fi cluster-speci c pathway dysregulation in AMs at steady state, AMs. We next sought to identify the direct downstream targets of which may lead to distinct function disorders in each AM cluster. HDAC3 for AM development. To achieve this, we performed Interestingly, when we analyzed transcriptomes of AM fi HDAC3 ChIP-Seq analysis of lung AMs from the WT mice, populations without clustering, we identi ed a total of 186 which was further integrated with our bulk RNA-seq data. We differentially expressed genes, including 59 upregulated genes and fi found 2502 genes shared in both datasets that may be primary 127 downregulated genes in the HDAC3-de cient AMs com- targets of HDAC3 transcriptional regulation (Fig. 6a). IPA ana- pared to their WT counterparts (Supplementary Fig. 6a). IPA of lysis of these differentially expressed genes revealed that PPAR these differentially expressed genes showed that several canonical signaling was one of the canonical pathways potentially targeted pathways related to inflammation and immune responses were fi by HDAC3 (Fig. 6b). Indeed, the repression of Pparg mRNA signi cantly inhibited in the HDAC3cKO AMs (Supplementary expression was correlated with the loss of HDAC3 in HDAC3- Fig. 6b). Moreover, the signaling pathways related to macrophage deficient preAMs (Fig. 6c). ChIP-seq analysis also showed that survival and apoptosis were also remarkably impaired, including 29,30 31,32 HDAC3 can directly bind to Pparg enhancers that are located in TREM1 signaling , NF-kB pathway , and STAT3 the upstream (within −30 kilobases (kb)) or intronic (+49 kb) 33,34 fi pathway . qRT-PCR analysis further con rmed the down- regions from the transcription start sites in lung AMs from the regulation of the genes associated with these three pathways, such WT mice, but not in the HDAC3cKO mice (Fig. 6d, Supple- as Flt1, Mapk3, Map3k1, Tlr2, Fcgr2b, Tgfb1, and Tgfbr1 mentary Data 3). The binding of HDAC3 to the Pparg enhancers (Supplementary Fig. 6c). These data suggest that the inhibition was further confirmed by ChIP-qPCR using MH-S cells, a mouse of TREM1, NF-kB, and STAT3 signaling pathways might reduce AM cell line (Fig. 6e). Moreover, the GSEA of bulk RNA-seq cell survival of HDAC3cKO AMs, leading to the decreased total data revealed that the absence of HDAC3 led to a significantly number of AM cells in the absence of HDAC3 as we observed at weaker enrichment of genes involved in PPAR-γ signaling com- steady-state maintenance (Fig. 2d). On the other hand, scRNA- pared to WT (Fig. 6f, g). Lack of HDAC3 resulted in the seq also revealed that PPAR-γ was remarkably reduced in γ fi downregulation of PPAR- target genes, including Ppargc1b, HDAC3-de cient AMs (Supplementary Fig. 6a), suggesting that Acsl1, Cidec, Fabp4, Mgst3, and Nr1h3, which was further con- HDAC3 might control AM homeostasis through regulation of fi γ rmed by qRT-PCR (Fig. 6h). Collectively, these results indicate PPAR- . that HDAC3 directly targets PPAR-γ in AMs, and that loss of HDAC3 fundamentally abrogates the PPAR-γ signaling pathway during AM development. HDAC3 controls mitochondria function and survival of pre- AMs. To further obtain mechanistic insights into the embryonic γ deficiency of AMs in the absence of HDAC3, we first performed PPAR- regulates mitochondria function and survival of pre- AMs. A prior study has shown impaired development of fetal bulk RNA-seq analysis of lung preAMs from cKO and WT at γ fi E18.5. Based on the principal component analysis (PCA) of their lung preAMs in Vav1-Cre-induced, PPAR- -de cient mice at E18.57, indicating that prenatal PPAR-γ is required for AM transcriptomes, the molecular signatures showed a clear differ- iCre fl/fl ence in preAMs between cKO and WT (Fig. 5a). A total of 9074 ontogeny. Using Csf1r Pparg cKO mice, we further found a fi modest reduction of F4/80hiCD11bint YS pMac-derived lung fetal differentially expressed genes (P < 0.05) were identi ed in the int int preAMs between HDAC3cKO and WT, including 4612 down- macrophages at E14.5 and a drastic loss of F4/80 CD11b fetal liver monocyte-derived preAMs at E17.5 from Csf1riCrePpargfl/ regulated and 4462 upregulated genes (Fig. 5b, Supplementary fl Data 1). (GO) analysis further identified the cKO mice (Fig. 7a), which were consistent with the phenotype altered biological pathways in the HDAC3cKO preAMs, includ- observed in HDAC3cKO mice (Fig. 1d, e). As expected, ing multiple metabolic pathways, the electron transport chain, cell CD11chiSiglec-Fhi AMs were essentially eliminated in the Csf1ri- fl fl proliferation, and cell death (Fig. 5c, Supplementary Data 2). CrePparg / cKO adult mice (Fig. 7a), which was consistent with Most of the identified pathways are consistent with previous the previous results using Vav1Cre- and Cd11cCrePpargfl/flcKO studies in other cell types with HDAC3 deletion14,35,36. mice7. Moreover, PPAR-γ gene expression was detected in the Furthermore, gene set enrichment analysis (GSEA) identified lung fetal macrophages at E14.5, which was gradually upregulated some gene sets that were either depleted (normalized enrichment in preAMs at E16.5 and E18.5, and maintained at high levels in score, NES < 0, false discovery rate, FDR < 0.05) (Supplementary the AMs from young and adult mice (Fig. 7b). Interestingly,

NATURE COMMUNICATIONS | (2020) 11:3822 | https://doi.org/10.1038/s41467-020-17630-6 | www.nature.com/naturecommunications 7 ebaeptnil tiigo rAsdtrie by determined preAMs of staining potential) membrane ( ifrnilyepesdgns E,Nraie nihetScore. Enrichment Normalized NES, genes. expressed differentially orcinfrmlil etn rmec ls r shown. are class each from testing multiple for correction identi ilgclpoesswt h mletadjusted smallest the with processes biological red). hg olow). to (high h Hdac3 .3( 0.03 oreData Source ilgclyidpnetsmlsi ahpnl All panel. each in samples biologically-independent neigVsann fpreAMs; of staining V annexing 8 development. embryonic AM during death cell and metabolism controls HDAC3 5 Fig. ARTICLE Hdac3 Cox6c SApo fapoptosis. of plot GSEA d g a b fi fl fl Cox7b ,7.3e ), db GSEA. by ed / / etmpsoigdfeeta eeepeso ( expression gene differential showing map Heat Ranked list metric (Signal2Noise) Enrichment score (ES) –0.5 –0.4 –0.3 –0.2 –0.1 – – fl fl 0.0 5.0 2.5 0. 2. 5. 0 5 0 ; ; 0020 0040 006000 5000 4000 3000 2000 1000 0 Csf1r Csf1r MFI (×103) 'HDAC3KO' (positivelycorrelated) 0 1 2 3 4

Hdac3 Mito tracker orange ;** ); Csf1r Hits Enrichment profile − fi iCre iCre P j l/fl le.

4( iCre =0.0051 PRaayi fmN xrsinlvl faotssascae ee.BMP2: genes. apoptosis-associated of levels expression mRNA of analysis qPCR P oxidative phosphorylation

Hdac3 Zero crossat3055 Rank inordereddataset Hdac3 Hdac3 Enrichment plot: = Ndufa4

l/fl AUECOMMUNICATIONS NATURE PC2 f 4.2e FDR =0 NES =–3.31 'WT' (negativelycorrelated) R-C nlsso RAepeso eeso xdtv hshrlto-soitdgenes. phosphorylation-associated oxidative of levels expression mRNA of analysis qRT-PCR Ranking metricscores fl fl PC3 / / − fl fl ,1.6e ), .* 3( mro tE18.5. at embryos i h P etmpsoigurgltdgnsrltdt ppoi identi apoptosis to related genes upregulated showing map Heat Atp5l =

− Enrichment score (ES) Ranked list metric (Signal2Noise) –0.10 –0.05 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 –5.0 –2.5 .1( 0.01 0.0 2.5 5.0 4( n 0020 0040 006000 5000 4000 3000 2000 1000 0 'HDAC3KO' (positivelycorrelated) ,1.2e ), PC1 = Ndufa8 nihetpoieHt Rankingmetricscores Hits Enrichment profile 8 b e Bmp2 Hdac3 Enrichment plot:Apoptosis – Rank inordereddataset Zero crossat3055 Ndufb5 Ndufb4 Ndufb3 Ndufb2 Ndufb1 Ndufa10 Ndufa9 Ndufa8 Ndufa7 Ndufa6 Ndufa5 Ndufa4 Ndufa3 Ndufa2 Ndufa1 3( ope ComplexII Complex I ,8.0e ), ,001( 0.011 ), Uqcrfs1 fl a 'WT' (negativelycorrelated) FDR =0.0035 NES =1.79 22)1:82 tp:/o.r/013/447001606|www.nature.com/naturecommunications | https://doi.org/10.1038/s41467-020-17630-6 | 11:3822 (2020) | / fl rnil opnn nlss(C)promdo rncitm ( transcriptome on performed (PCA) analysis component Principle P , Csf1r – ausb AESoe(n-aldFisher (one-tailed Score EASE by values n Ndufv3 Ndufv2 Ndufv1 Ndufs8 Ndufs7 Ndufs6 Ndufs4 Ndufs3 Ndufs2 Ndufs1 Ndufc2 Ndufc1 Ndufb8 Ndufb7 Ndufb6 4( ;*** ); = P iCre Il6 7 Ndufb2 auswr banduigteStudent the using obtained were values Relative expression(Log2) Hdac3 P ,006( 0.026 ), Complex III Csf1r P 080.8 –0.8 0.05). < = fl d wctmty F median MFI cytometry. ow 1.0e iCre Uqcrq Uqcrh Uqcrfs1 Uqcrc1 Uqcrb Uqcr11 Uqcr10 Sdhd Sdhc Sdhb Sdha Uqcrc2 ,1.5e ), fl/fl nihetpo foiaiepopoyainfo eeStErcmn nlss(SA of (GSEA) Analysis Enrichment Set Gene from phosphorylation oxidative of plot Enrichment i 0 Hdac3 – 4( c Tnf Csf1r – ope VComplexV Complex IV –4 –2024 eeOtlg G)ercmn ftedwrgltdadurgltdgnsin genes upregulated and downregulated the of enrichment (GO) Ontology Gene

5( Hdac3 Atp5g3 e Fold Change

;** ); iCre fl / etmpdpcigdwrgltdgnsrltdt xdtv phosphorylation oxidative to related genes downregulated depicting map Heat Sdhb fl/fl Cox8a Cox7c Cox7b Cox7a2l Cox7a2 Cox7a1 Cox6c Cox6b2 Cox6b1 Cox6a1 Cox5b Cox5a Cox4i1 Cox10 Hdac3 fl 1.6

ahdtrpeet n mroa 1. n asrpeetma Dof SD ± mean represent bars and E18.5 at embryo one represents dot Each . Hdac3 P AUECMUIAIN https://doi.org/10.1038/s41467-020-17630-6 | COMMUNICATIONS NATURE = ,2.7e ), ,2.2e ), fl/fl Casp Bg Smad Irf1 Ier Sqstm1 Pla Ifnb1 Fez1 Atf Timp Krt1 Lu Il1 Ju Isg2 Hg Ct Txni Egr Tn Tgfbr Il Il1 Pmaip1 Dnaja1 Ccnd1 Pdgfr Timp Hmox1 Satb1 F2 Cd6 Ptk Rho Ifitm Dc Cd3 Bmp Mmp 6 fl/fl h b a n m 1.1e 3 f r n n 3 f t 2 3 9 b 8 8 p 3 0 2 3 2 2 3 3 b 7 – − − 3( Atp6v1f Atp5j2 Atp5j Atp5h Atp5g3 Atp5g2 Atp5g1 Atp5f1 Atp5e Atp5d Atp5c1 Atp5b Atp5a1 Atp6v1e1 Atp6v1c1 Atp5o Atp5l 5( 4( Cd38 Cox5a Uqcrq fl oecneintensity; uorescence ;*** ); ukRAsqaayi fsre rAsfo ea ugof lung fetal from preAMs sorted of analysis RNA-seq Bulk ’ exact s c f mRNA expression j 10 ). ,4.7e ), 0 2 4 6 8

g M Biological_process gene ontology term (GO) Enrichment score Gene count Adj. Adj. Genecount Enrichmentscore Biological_process geneontologyterm(GO) Adj. Genecount Enrichmentscore Biological_process geneontologyterm(GO) at cdoiain5.66 11.50 7.87 7.43 9.93 Fatty acidoxidation Nucleoside triphosphatebiosyntheticprocess Nucleoside biosyntheticprocess Nucleoside phosphatemetabolicprocess Electron transportchain Positive regulationofbiosyntheticprocess Divalent inorganiccationhomeostasis Ion homeostasis Cell death Regulation ofmononuclearcellproliferation 11.76 Positive regulationofphosphorusmetabolicprocess Regulation ofphosphatemetabolicprocess Leukocyte cell-celladhesion System development Cell migration ATP metabolicprocess GO enrichment(upregulatedgenes) GO enrichment(downregulatedgenes) fi ’ *** P mp mRNA expression w-aldunpaired two-tailed s db SA oos(e obu)so xrsinvalues expression show blue) to (red Colors GSEA. by ed 0.0 0.5 1.0 1.5 2.0 itga ltfrMtTakrOag y (mitochondria dye Orange MitoTracker for plot Histogram

n 2 = P = B H − 4.7e

vlei AI ytm ihteBenjamini-Hochberg the with system) DAVID in -value m da

p * c 4( 5 2 3 A N.D. t

Hdac3 Cc p5 ** ** − Cox5b *** g nd 3 1 6( A t Csf1r Hdac3

p ** C 5 d38 l ** Mmp2

fl C

,5.7e ), ox Csf1r Hdac3 / ** ** 5 iCre fl

a fl/fl n ,

C Hdac3 n

= I o l6 Cre * x ** ** fl/fl 5 =

Hdac3 b ,1.3e ), C 10 ox Hdac3 − fl/fl

n T 4 6a n * ** * 4( t f 1 = Hdac3

fl/fl C Csf1r o et oredt r rvdda a as provided are data Source test. x6b ** ** 3 Cox6a1 − 1 fl

Hdac3 C 4( / ox6 iCre fl ** ** fl

turquoise; , c k / C Ccnd1 Hdac3 100K 150K fl FSC-A o 50K x 6.38 7.28 10.80 12.50 15.12 17.30 17.71 20.31 43.64 46.09 , ,1.2e ), 7 0 b * fl

n N Annexin V

/ duf = fl

+ Hdac3 Annexin V preAM (% a * **

). ) ; 4 Csf1r 20 40 60 N fl 5 0 duf / − Hdac3 k ** **

fl a

Csf1r Csf1

fl/fl 8 6( te genes: other ;

lwctmtyof cytometry Flow N

fl/fl P du iCre

Csf1r r iCre =0.0071 f 301 103 163 360 75 240 331 144 878 336 b * ** 35 33 40 145 38 72 Cox6b1 2 Hdac3 Csf1r iCre Hdac3 S 010 dhb fl/fl ** ** iCre

iCre U Hdac3 q cr Hdac3

3 fs1 Hdac3 5.28E–08 8.66E–08 1.81E–2 5.02E–12 1.27E–15 1.61E–19 9.57E–19 8.72E–20 7.62E–47 5.22E–47 1.25E–05 3.43E–07 9.93E–07 6.21E–10 6.63E–11 7.82E–12 ** ,2.4e ), fl 10 U P P / value value 4 qc fl r * ** fl .* b fl/fl q / n n The . fl fl P = = . − / = fl 4 , 3 3 NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-17630-6 ARTICLE

a b c Positive z-score z-score = 0 Negative z-score No activity pattern available Ration

Threshold –log(p-value) 0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00 2.25 Induced 10 Unchanged PPAR signaling ChIP-Seq Repressed (8276) LPS/IL-1 mediated

Inhibition of RXR function (RPKM) 5 2 2502 PXR/RXR signaling Pparg LXR/RXR signaling 0 RNA-Seq Aryl hydrocarbon (9074) Receptor signaling –5

PPARα/RXRα signaling HDAC3 KO, log

0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 –10 Ratio –10 –5 0 5 10

WT, log2(RPKM)

d Chr6: 115,300 kb – 115,550 kb 100 kb mm10 e Anti-RNA Poly II 1.2 Anti-HDAC3 Rabbit IgG WT 0.8 HDAC3

(% of input) 0.4 KO ChIP enrichment

N.D. 0.0 Pparg P parg

f g h iCre fl/fl fl/fl Enrichment plot:Targets of PPARG 3 Csf1r 2.0 0.0 fl/fl Hdac Hdac3 Hdac3 –0.1 NES = –2.30 iCre fl/fl –0.2 FDR = 0.00039 Csf1r Hdac3 Nr1h3 –0.3 1.5 –0.4 Acsl1

Enrichment score (ES) –0.5 Pex11a * –0.6 Fabp5 ** *** *** * 1.0 Mgst1

5.0 Ppargc1b 'HDAC3KO' (positively correlated) 2.5 S100a1 mRNA expression 0.5 0.0 Zero cross at 3055 Hsd11b1 –2.5 'WT' (negatively correlated) Mgst3 –5.0 N.D. N.D.

Ranked list metric (Signal2Noise) 0 1000 2000 3000 4000 5000 6000 Rank in ordered dataset Fabp4 0.0 Enrichment profile Hits Ranking metric scores Cidec Acsl1 Cidec Pparg Fabp4 Mgst3 Nr1h3 Ppargc1b

Fig. 6 HDAC3 deficiency leads to impaired PPAR-γ signaling in AM development. a Venn diagram of genes identified by HDAC3 ChIP-seq in the AMs from Hdac3fl/fl adult mice overlapped with differentially expressed genes from bulk RNA-seq. b Ingenuity Pathway Analysis (IPA) showing canonical nuclear receptor pathways of the overlapped genes as in a. Threshold indicates the minimum significance level by the right-tailed Fisher’s exact test. Ratio refers to the number of molecules from the dataset that map to the pathway listed divided by the total number of molecules that map to the canonical pathway from within the IPA knowledgebase. c Scatter plot of bulk RNA-seq data showing the overlapped genes as in a from n = 3 Hdac3fl/fl WT; Csf1riCreHdac3fl/fl KO littermates. Fold change >1.5 up (red) or down (blue) and false discovery rate (FDR) < 0.05. RPKM, reads per kilobase per million. d Genome browser tracks of the Pparg highlighting ChIP-seq data at the enhancers (boxed). Hdac3fl/fl, WT; Csf1riCreHdac3fl/fl, KO. e ChIP-qPCR data representing HDAC3 binding at the Pparg enhancers in MH-S cells. Enrichment value is normalized to input measurements. Representative data from two independent experiments. f GSEA plot of targets of PPAR-γ from bulk RNA-seq. g Heat map representing downregulated genes related to PPAR-γ signaling identified by GSEA as in f. Colors (red to blue) show expression values (high to low). h qPCR analysis of mRNA expression levels of the genes associated with PPAR-γ signaling. n = 3 Hdac3fl/fl;Csf1riCreHdac3fl/fl.*P = 0.043 (Nr1h3); **P = 5.5e−3(Pparg), 3.4e−3(Ppargc1b); ***P = 4.9e−5(Acsl1), 6.4e–6 (Fabp4) by the Student’s two-tailed unpaired t test. Bars represent mean ± SD of biologically-independent samples in each panel. Source data are provided as a Source Data file.

PPAR-γ mRNA was not expressed in the TRMs of the kidney, in PPAR-γ-deficient mice (Fig. 7d–f). ChIP-qPCR analysis using liver, brain, and heart (Supplementary Fig. 9), suggesting that MH-S cells further showed an enrichment of PPAR-γ binding HDAC3-PPAR-γ axis is specifically required for the development sites in the promoter regions (within 3 kb of the transcription of lung AMs, but not for TRMs in other tissues. start sites) of oxidative phosphorylation-related genes, including We next revisited a recently-published microarray dataset7 and Atp5l, Cox5a, Cox5b, Cox6a1, Ndufa4, Sdhb, Uqcrfs1, and Uqcrq, analyzed gene expression profiles of lung preAMs from Ppargfl/fl all of which were downregulated in the absence of PPAR-γ WT and Cd11cCrePpargfl/flcKO mice at day 2 after birth. As (Fig. 7e, f), suggesting a possible direct impact of PPAR-γ on expected, the PPAR-γ signaling pathway in preAMs was severely mitochondria function-related genes in AMs (Fig. 7g). Moreover, impaired in the absence of PPAR-γ, which was consistent with the apoptosis-associated genes were highly enriched in PPAR-γ- our observations in Csf1riCreHdac3fl/flcKO mice (Fig. 7c). Inter- deficient mice compared to their WT counterparts, which was estingly, oxidative phosphorylation-related genes, which were also consistent with our observations in HDAC3cKO mice severely diminished in HDAC3cKO mice, were robustly reduced (Fig. 7h–j). These data suggest that the PPAR-γ signaling pathway

NATURE COMMUNICATIONS | (2020) 11:3822 | https://doi.org/10.1038/s41467-020-17630-6 | www.nature.com/naturecommunications 9 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-17630-6

fl/fl iCre fl/fl a b ) Pparg Csf1r Pparg 2 15 100 P = 0.04 (%)

Int 75 10 50 E14.5 43.0 30.7 CD11b

hi 5 25

F4/80 0 I 0 Relative expression (Log 80 P = 4.5e–8 (%)

lnt 60 P14 Adult E17.5 E14.5 E16.5 E18.5 Cre 40 iCre 45.5 5.48 c fl/fl CD11b fl/fl CD11b fl/fl fl/fl lnt 20 Csf1r Cd11c Hdac3 Pparg Pparg 0 Hdac3 F4/80 F4/80 1.0 P = 4.5e–4 Pex11a 30 Nr1h3 0.5 Ppargc1b 0 20 Fabp4 13.2 0.14 Adult Acsl1 –0.5 10

AM (%) Mgst1 CD11c –1.0 Cidec 0 S100a1 Siglec F fl/fl fl/fl Mgst3 Fabp5

Pparg iCre Pparg Hsd11b1

Csf1r

fl/fl Enrichment plot: Pparg oxidative_phosphorylation d f 1.5 iCre fl/fl 0.05 Csf1r Pparg NES = –1.49 0.00 –0.05 FDR = 0.053 –0.10 ** * ** * ** ** * ** * ** * * ** * * ** * ** * ** –0.15 –0.20 1.0

Enrichment score (ES) –0.25 –0.30

0.5 1.5 'KO' (positively correlated)

1.0 mRNA expression 0.5 0.0 Zero cross at 1773 –0.5 –1.0 0.0 'WT' (negatively correlated) –1.5 l 0 500 1000 1500 2000 2500 3000 3500 4000 g 3 5 a b 1 8 2 q Ranked list metric (Signal2Noise) r g 5 5 6c a b s1 r Rank in ordered dataset a tp x 6a 6b1 x dhb rf p p5 A o ox x o S c qc Enrichment profile Hits Ranking metric scores P t C C o ox C duf duf q U A C C N N U

g 8 Anti-RNA Poly II Anti-PPAR-γ 6 e Rabbit IgG Cre iCre fl/fl r fl/fl fl/fl fl/fl 4 Csf1 Cd11c

Hdac3 Pparg Pparg (% of input) Hdac3 2 Ndufs7 1.0 ChIP enrichment Atp5d N.D. N.D. N.D. N.D. N.D. 0 Cox6a1 0.5 Uqcrc1 ox5a dhb Atp5l C Cox5b ox6a1 dufa4 S qcrfs1 Uqcrq Cox10 0 C N U Ndufa3 Ndufs3 –0.5 Uqcrq h Enrichment plot: Apoptosis j 0.45 Ppargfl/fl Cox4i1 0.40 NES = 1.53 0.35 iCre fl/fl –1.0 0.30 FDR = 0.044 Csf1r Pparg Ndufv1 0.25 6.0 0.20 *** Atp5a1 0.15 0.10 Sdhd 0.05 Enrichment score (ES) 0.00 *** Ndufb7 –0.05 4.0 Sdhb *** *** Cox6b1 * Atp5g2 1.5 'KO' (positively correlated 1.0 ** Ndufs2 0.5 2.0 ** 0.0 Zero cross at 1773 –0.5 Ndufab1 mRNA expression –1.0 Ndufb5 –1.5 'WT' (negatively correlated) 0 500 1000 1500 2000 2500 3000 3500 4000 Ranked list metric (Signal2Noise) Ndufb2 Rank in ordered dataset 0.0 Ndufa8 Enrichment profile Hits Ranking metric scores f3 2 1 8 Il6 nf t 3 T Il1b Sdha A mp cnd B C Cd Uqcrfs1 Cre iCre Atp5b fl/fl fl/fl fl/fl fl/fl Atp5o i Csf1r Cd11c k Ndufc2 Hdac3 Hdac3 Pparg Pparg Sdhc 1.0 Plat HDAC3 Atp6v1c1 Bgn Ndufa5 Tgfbr3 0.5 Ndufb3 Pdgfrb Atp5j Pmaip1 0 PPAR-γ Ndufa4 Ccnd1 Ndufs1 Cd38 –0.5 Atp5c1 Egr3 Ndufb6 Atf3 –1.0 PPAR-γ signaling Cox5a Il1b Cox7b Mmp2 Atp5j2 Ifitm3 Atp5g3 Timp2 Hgf Mitochondrial Atp5l oxidative function Atp5f1 Il6 Il1a cell survial Cox6c Cd69 Atp5h Ler3 Cox7a2 Rhob Satb1 F2r Smad7 AM development serves as a key pathway that contributes to the dysregulation of Discussion mitochondrial oxidative function and cell survival in the The concept of TRMs has been refreshed in the past few years. HDAC3-deficient AMs. Taken together, our study demonstrates Recent lineage-tracing studies have revealed that adult mouse that the HDAC3-PPAR-γ axis is indispensable for normal TRMs are derived from embryonic YS EMP-derived macrophages mitochondrial oxidative function and cell survival during AM and FL monocytes rather than from bone marrow-derived pre- development (Fig. 7k). cursors, and they are self-maintained at steady state after

10 NATURE COMMUNICATIONS | (2020) 11:3822 | https://doi.org/10.1038/s41467-020-17630-6 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-17630-6 ARTICLE

Fig. 7 PPAR-γ governs mitochondria function and cell survival in AM development. a Flow cytometry analysis of lung fetal macrophages (F4/ 80hiCD11bint), preAMs (F4/80intCD11bint), and AMs (CD11chiSiglec-Fhi) from lungs of Ppargfl/fl and Csf1riCrePpargfl/fl mice at indicated ages. E14.5: n = 8 Ppargfl/fl, n = 19 Csf1riCrePpargfl/fl; E17.5: n = 10 Ppargfl/fl, n = 8 Csf1riCrePpargfl/fl; adult: n = 7 Ppargfl/fl, n = 8 Csf1riCrePpargfl/fl. b qRT-PCR analysis of PPAR-γ mRNA expression in the lung fetal macrophages (E14.5), preAMs (E16.5, E18.5), or AMs (P14, adult) isolated from C57BL/6 mice (n = 4 for P14, n = 3 for each other time point). Bulk RNA-seq data of preAMs from n = 3 Hdac3fl/fl;Csf1riCreHdac3fl/fl embryos at E18.5 were generated as in Fig. 5. Microarray data for preAMs from n = 2 Ppargfl/fl;Cd11cCrePpargfl/fl mice at day 2 after birth was obtained from GEO database (GSE60249). Heat map showing expression of genes associated with PPAR-γ signaling (c), oxidative phosphorylation (e), and apoptosis (i) from two datasets. GSEA plots presenting expression of genes associated with oxidative phosphorylation (d) and apoptosis (h). qRT-PCR analysis of mRNA expression levels of genes associated with oxidative phosphorylation (f) and apoptosis (j) in the preAMs isolated from n = 3 Ppargfl/fl;Csf1riCrePpargfl/fl embryos at E17.5. *P = 0.012 (Cox5b), 0.019 (Cox6c), 0.035 (Ndufb2), 0.02 (Tnf); **P = 7.5e−3(Atp5l), 4.4e−3(Cox5a), 3.4e−3(Uqcrq), 3.3e−3(Atf3), 3.7e−3(Il1b); ***P = 9.5e−6(Pparg), 3.0e−5(Atp5g3), 4.4e−4(Cox6a1), 3.1e–4(Cox6b1), 2.4e−4(Ndufa8), 7.0e–5(Sdhb), 6.5e−4(Uqcrfs1), 5.2e−4(Bmp2), 1.8e−4(Ccnd1), 9.1e−4(Cd38), 3.2e−4(Il6). g ChIP-qPCR data representing PPAR-γ binding at the promoter regions of the indicated oxidative phosphorylation genes in MH-S cells. Enrichment value is normalized to input measurements. Representative data from two independent experiments. k Schema of HDAC3-PPAR-γ axis in regulation of AM development. Bars represent mean ± SD of biologically-independent samples in each panel. All P values were obtained using the Student’s two-tailed unpaired t test. Source data are provided as a Source Data file.

birth2,24,37. Although a few genes (e.g., Csf2, Pparg, and Tgfb) inflammatory cytokines and chemokines during inflammatory have been identified as requirements for AM development, little conditions. Together, our data suggest that individual AM sub- information is known about the epigenetic regulators that instruct populations might be specialized for unique functions to regulate the development and maintenance of AMs. HDAC3, belonging to lung homeostasis and antipathogen immunity. the class I HDAC family, plays a critical role in many biological Interestingly, HDAC3 deficiency resulted in unproportionable processes, including circadian rhythm17, intestinal homeostasis38, changes in the number of AM sub-cluster cells, indicating a and tissue thermogenesis14. Here, we demonstrate that HDAC3 is cluster-dependent effect of HDAC3 on AM homeostasis. More- required for embryonic development and postnatal maintenance over, the transcriptomic changes and their associated pathway of lung AMs at steady state. In addition, HDAC3 controls AM dysregulation induced by HDAC3 deletion in AMs also showed a regeneration from BM progenitors after lethal whole-body irra- cluster-specific pattern, i.e., mitochondria dysfunction in cluster diation. Our data further indicate that HDAC3 is dispensable for 1, cell cycle control disorders in cluster 2, antigen-presenting development of TRMs in the brain, spleen, kidney, and pancreas, dysfunction in cluster 3, and immune response dysregulation in suggesting that HDAC3 regulates TRM ontogeny in an organ- cluster 4. Notably, the mitochondria dysfunction found in the specific manner. abundant AM cluster (cluster 1) is consistent with our observa- As the resident macrophages in lung alveoli, AMs are long tions from bulk-RNA seq data showing impaired mitochondrial lived with proliferative self-renewal capability, and control the oxidative function and deteriorative cell death in the absence of maintenance of airway homeostasis through their capacity of HDAC3 during AM development. Given the importance of clearing surfactant and cell debris at steady state1,39. During mitochondria function in immune metabolism and cell survival respiratory infection or inflammation, AMs also function as of macrophages48, we reason that the loss of cluster 1 caused by antigen-presenting cells by transporting exogenous antigens to impaired mitochondria function largely contributes to the lung-draining lymph nodes and releasing a wide array of cyto- reduced number of total AM cells in the HDAC3cKO mice. kines and chemokines to reprogram both innate and adaptive Moreover, the disorders of cell cycle control and DNA damage immune responses40,41. However, it remains unclear whether response in the proliferating cluster (cluster 2) likely lead to less multifunctional roles of AMs are derived from AM heterogeneity proliferative capacity of this cluster, which may also partially at the individual cell level. Recent advances in single-cell tech- contribute to the reduced number of HDAC3-deficient AM cells. nology have provided a powerful tool to study tissue and cell On the other hand, loss of HDAC3 significantly affected the heterogeneity28,42–44. To address if the AMs are heterogenous at antigen-presenting function of cluster 3 and the immune- steady state and if HDAC3 regulates AM heterogeneity and responsive function of cluster 4, suggesting that HDAC3 might subset function, we performed scRNA-seq of mature AMs in the crucially control the function of both AM clusters during immune adult mice from WT and CD11cCre.HDAC3 KO mice. We found responses. Further studies are needed to assess the role of that the murine lung tissues contain four transcriptionally- HDAC3 in the functionality of individual AM subpopulations in distinct AM cell subsets. The most abundant cluster (cluster 1) the pathogenesis of lung diseases, such as asthma and cancer. accounted for over 85% of the total lung AMs and expressed high Furthermore, it remains to be determined whether fetal immature levels of genes related to lipid metabolism, including Abcg145 and preAMs have subpopulations and how HDAC3 regulates preAM Ltc4s46. Thus, this abundant AM cluster may be specialized for subsets, if any, during embryonic development. However, given pulmonary surfactant lipid clearance. It has been shown that the severe developmental defects on preAMs in the absence of adult AM network is maintained throughout life by a slow local HDAC3, we currently could not obtain enough cells from the KO proliferation of lung AMs rather than from BM precursors8. littermates for scRNA-seq analysis. Advanced technology, e.g., Interestingly, one of the minor AM clusters (cluster 2, 6%) was mass cytometry, may serve as a promising and powerful tool to characterized by higher expression of cell cycle-related genes, and address these interesting questions in the future49. may represent proliferating AM cells that slowly renew the AM Inhibition of HDACs is emerging as a promising approach to network, which has yet to be further tested. Another minor AM treat various types of malignant diseases, neurodegenerative dis- cluster (cluster 3, 3%) was enriched in many antigen- eases, inflammatory disorders, and cardiovascular diseases50. presentation-related genes and may represent the AM sub- Four pan HDACi have been approved by the FDA as anticancer population that possesses antigen-presenting function47. Fur- agents for treatment of lymphoma and multiple myeloma, while thermore, the most minor AM cluster (cluster 4, 4%) highly many other pan HDACi are currently in clinical trials for treat- expressed a variety of genes involved in immune responses and ment of lymphoma and solid tumors51. HDACi have also been may serve as the major source of AMs for secreting pro- recently suggested in the treatment of asthma and other

NATURE COMMUNICATIONS | (2020) 11:3822 | https://doi.org/10.1038/s41467-020-17630-6 | www.nature.com/naturecommunications 11 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-17630-6 inflammatory lung diseases20,21. However, due to common side GCTTCTGCTATGTCAATG-3′. The floxed allele of Hdac3fl/fl mice produced a effects of these pan HDAC inhibitors, developing drugs with high 504-bp PCR product, whereas the wild type allele resulted in a 464-bp PCR product. selectivity for individual HDAC has drawn great attention. On the other hand, different subtypes of HDACs appear to play Cell culture. MH-S cells (ATCC, CRL-2019) were cultured with RPMI 1640 with 52 4.5 g/L D-glucose (Gibco/Thermo Fisher Scientific), 10% FBS, 0.05 mM 2- disparate roles in immune defense and disease progression . fi Understanding the underlying mechanism by which individual Mercaptoethanol (Gibco/Thermo Fisher Scienti c), and 1% penicillin- streptomycin-amphotericin B at 37 °C in 5% CO2. The cell line was passaged at HDACs regulate immune cell development, maintenance, and approximately 80–90% confluence. function will help us to identify potential therapeutic treatments. In the present study, deletion of HDAC3 leads to repressed fi Isolation of embryonic and adult cells. Pregnant females were sacri ced by CO2 PPAR-γ expression in fetal immature AMs, as well as adult exposure. Embryos ranging from embryonic days E10.5-E18.5 were removed from mature AMs, as confirmed by bulk RNA-seq and scRNA-seq the uterus and washed in 4 °C phosphate-buffered saline ([PBS], Invitrogen, analyses, respectively, suggesting the essential role of the HDAC3- Carlsbad, CA). Embryos were exsanguinated through decapitation. Lungs, liver, γ brain, kidneys, spleen, and pancreas were collected, minced into tiny pieces, and PPAR- axis in the development and maintenance of AMs. In incubated for 30 min in PBS containing 1 mg/ml collagenase D (Roche, Basel, addition, our data imply that the HDAC3-PPAR-γ axis controls Switzerland), 0.01% DNase I (Worthington Biochemical Corp., Lakewood, NJ), and fetal immature AM development, at least partially through reg- 3% FBS (Hyclone, San Angelo, TX) at 37 °C. Erythrocytes from all the samples were lysed for 3 min with 0.83% NH4Cl buffer. The cells from the adult tissues were ulation of mitochondria function and cell survival. Interestingly, fi PPAR-γ binding sites are enriched near the genes associated with isolated by the same method with some modi cations. The adult lungs were γ harvested, minced, and incubated in PBS containing 1 mg/ml collagenase D oxidative phosphorylation, suggesting that PPAR- might directly (Roche), 0.01% DNase I (Worthington), and 3% FBS (Hyclone) at 37 °C for 45 min. regulate mitochondrial oxidative function in AMs. On the other The erythrocytes were lysed as described above. All the cell suspensions were hand, given lack of PPAR-γ expression in TRMs from other passed through a 70 μm cell strainer (BD Biosciences, San Jose, CA). organs, the HDAC3-PPAR-γ axis seems to be only required for TRM development in a lung-specific manner. It should be noted Flow cytometry and cell sorting. Single-cell suspensions were centrifuged at 450 × g for 7 min, resuspended in ice-cold staining buffer (1× PBS with 2% FBS), that a very recent single-cell study revealed an enrichment of fi γhi and placed in 96-well round-bottom plates. After incubation with puri ed anti- PPAR- macrophages in human lung adenocarcinoma lesions FcγRII/III antibody (clone 2.4G2) at 4 °C for 15 min, cells were stained with a 53 that likely promotes an immunosuppressive microenvironment . mixture of fluorescent surface antibodies at 4 °C for 30 min. The stained samples Thus, HDAC3i could be promising immunotherapeutic drugs to were either directly analyzed by flow cytometry or fixed with 2% formalin at 4 °C treat lung cancer patients by targeting the HDAC3-PPAR-γ axis for 20 min prior to storage at 4 °C overnight. The full list of antibodies used can be found in Supplementary Table 1. Flow cytometry was performed with FACSAria™ in lung macrophage populations, which needs to be further II or FACSCelesta™ flow cytometer (BD Biosciences). Data were acquired using BD investigated. FACSDiva software version 8.0.2 (BD Biosciences) and analyzed using FlowJo In conclusion, our data have provided evidence that 10.5.3 (BD Biosciences). + fl fl fl fl HDAC3 serves as a key epigenetic regulator to control AM The CD45 CD11chiSiglec-Fhi lung AMs of adult Hdac3 / and UbcCreERHdac3 / mice were sorted by FACS Aria II flow cytometer. The CD45+F4/80hiCD11bint fetal development at the embryonic stage, as well as in AM main- + lung macrophages at E14.5, CD45 F4/80intCD11bintpreAMs at E16.5 and E18.5, tenance and regeneration after birth. Accumulated research has CD45+CD11chiSiglec-Fhi adult lung AMs, as well as CD45+F4/80intCD11bhi FL indicated that AMs are involved in the development of cancer, monocytes at E16.5 from Hdac3fl/fl and Csf1riCreHdac3fl/fl mice were also sorted by autoimmune, and inflammatory diseases in the airway. Thus, our FACSAria™ II flow cytometer. The purity of the isolated populations was >95%. findings may shed light on HDAC3 as a potential therapeutic target for AM-based intervention in cancer and autoimmune Mitochondria function and cell apoptosis assays. The MitoTracker Orange diseases. (Invitrogen/Thermo Fisher Scientific) stock solution was prepared according to the manufacturer’s instructions. Briefly, 50 μg of lyophilized MitoTracker Orange dye was dissolved in DMSO to a final concentration of 1 mM and stored at −20 °C in Methods the dark. Cells were incubated with 50 nM MitoTracker Orange in the prewarmed fl fl Animals. Hdac3 / mice were provided by Scott W. Hiebert54. C57BL/6 (Strain RPMI 1640 medium at 37 °C with 5% CO2 for 15 min, followed by two washes with #000664), B6.SJL (Strain #002014), Cd11cCre (Strain #008068), Csf1riCre (Strain PBS. The stained cells were analyzed by flow cytometry (Ex 554 nm, Em 576 nm). #021024), UbcCreER (Strain #007001), and Ppargfl/fl (Strain #004584) mice were Cell apoptosis was measured using the Annexin V Apoptosis Detection Kit purchased from the Jackson Laboratory (Bar Harbour, ME). To generate myeloid (eFuorTM450 or APC, eBioscience/Thermo Fisher Scientific). In brief, freshly- lineage-specific HDAC3 mutant mice, we crossed Hdac3fl/fl and Csf1riCre mice isolated cells were washed once with 1× binding buffer and resuspended in 100 μL (back to a B6/C57 mouse genetic background for 6 generations). To generate of 1× Binding Buffer at 5 × 106/ml. After adding 3 μLoffluorochrome-conjugated CD11c-expressing, cell-specific HDAC3 mutant mice, we crossed Hdac3fl/fl and Annexin V, the cells were incubated at RT for 13 min. The stained cells were Cd11cCre mice. To generate inducible HDAC3 deletion mice, we crossed Hdac3fl/fl washed once with 1× binding buffer and analyzed by flow cytometry. and UbcCreER mice. To generate myeloid lineage-specific PPAR-γ-deficient mice, we crossed Ppargfl/fl and Csf1riCre mice. To generate heterozygous mice expressing Tamoxifen treatment. Tamoxifen (Sigma Aldrich, St. Louis, MO) was dissolved in both CD45.1 and CD45.2, named B6.SJLhetero, we crossed B6.SJL (CD45.1) with corn oil (Sigma Aldrich) containing 10% (vol/vol) ethanol (Thermo Fisher Sci- C57BL/6 (CD45.2) mice. entific) and was intraperitoneally administered at 50 mg/kg of mouse weight for Male and female mice from 6–18 weeks of age were used. All experiments five consecutive days. included age-matched and sex-matched littermate controls. Embryonic development was estimated considering the day of vaginal plug formation as embryonic age of 0.5 days. Bone marrow chimeras. B6.SJLHetero (CD45.1+CD45.2+) mice were lethally All mice were housed under specific pathogen-free conditions at temperatures irradiated once with 950 rads. Donor BM cells were harvested from age-matched of 20–26 °C with 30–70% humidity and a 12-h light/12-h dark cycle at Henry Ford and sex-matched CD45.2+ Hdac3fl/fl and Csf1riCreHdac3fl/fl mice, as well as B6.SJL Health System. Experimental animal protocols were performed in accordance with (CD45.1+) mice. The BM cells from Hdac3fl/fl or Csf1riCreHdac3fl/fl mice were the guidelines of the Institutional Animal Care and Use Committee. mixed with the cells from B6.SJL mice (at a 4:1 ratio) and then co-transferred to the irradiated B6.SJLHetero recipient mice (10 × 106 mixed cells/mouse). The recipient mice were sacrificed 10 wks after reconstitution. Genotyping. Genotyping of Csf1riCre55, Cd11cCre56, UbcCreER57, and Ppargfl/fl58 mice was performed in adherence to the standard PCR-based procedure. The fol- lowing primers were used for each mouse strain: Csf1riCre: forward 5′-CTAGGC RNA extraction and quantitative real-time PCR. Total RNA was extracted from CACAGAATTGAAAGATCT-3′ and reverse 5′-ATCAGCCACACCAGACACA TRMs with miRCURY RNA Isolation Kit—Cell and Plant (Exiqon). The RNA was GAGATC-3′; Cd11cCre: forward 5′-ACTTGGCAGCTGTCTCCAAG-3′ and reverse reverse-transcribed to cDNA with High Capacity cDNA Reverse Transcription Kits 5′-GCGAACATCTTCAGGTTCTG-3′; UbcCreER: forward 5′-GACGTCACCCGT (Applied Biosystems, Foster City, CA). Quantitative real-time PCR (qRT-PCR) TCTGTTG-3′ and reverse 5′-AGGCAAATTTTGGTGTACGG-3′; Ppargfl/fl:5′-CTA reactions were prepared using FastStart Universal SYBR Green Master (ROX, GTGAAGTATACTATACTCTGTGCAGCC-3′ and 5′-GTGTCATAATAAACAT Roche) and carried out using QuantStudio 7 Flex Real-Time PCR System (Applied GGGAGCATAGAAGC-3′. Hdac3fl/fl mice were genotyped using the following PCR Biosystems). Data were collected using QuantStudio 7 Flex Real-Time PCR System primer pair: 1597B, 5′-GGACACAGTCATGACCCGGTC-3′; 1133T, 5′-CTCTG software version 1.2 (Applied Biosystems) and analyzed using Microsoft Excel 2016

12 NATURE COMMUNICATIONS | (2020) 11:3822 | https://doi.org/10.1038/s41467-020-17630-6 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-17630-6 ARTICLE

(Microsoft, Redmond, Washington). The full list of primers used can be found in Bulk RNA-Seq and ChIP-Seq data processing.TheRNA-Seqreadswerealigned Supplementary Table 2. Samples were normalized to GAPDH expression. to the mouse GRCm38/mm10 reference genome and expression levels for each sample were quantified and normalized using Biomedical Genomics Workbench 5.0 (Qiagen). Gene ontology (GO) biological process enrichment Immunofluorescence. A piece of lung or liver freshly isolated from adult mice or analysis was performed using DAVID Bioinformatics Resources 6.860. Gene Set embryos at E18.5 was immediately frozen in OCT compound for cryostat sec- Enrichment Analysis (GSEA) was performed using GSEA software 4.0.361,62. tioning (5 μm per section). The frozen tissue sections were fixed with acetone for Further pathway analysis was performed using the Ingenuity Pathway Analysis 15 min at −20 °C. After rinsing with PBS for 5 min, samples were blocked with ([IPA], Qiagen Bioinformatics, version 42012434). Raw ChIP-seq reads were anti-FcγRII/III antibody at RT for 15 min followed by staining at 4 °C overnight quality checked using the software FastQC version v0.10.163 and processed with PE-labeled anti-Siglec-F (adult lung) or PE-labeled anti-F4/80 (fetal lung) using BBDuk (v36.19) and seqtk (v1.2-r94, https://github.com/lh3/seqtk)to antibody. The samples were then stained with 1 μg/ml DAPI at RT for 1 min and trim the adapters and low-quality bases, respectively. The trimmed reads were mounted under a coverslip with one drop of IMMU-MOUNT (Thermo Fisher then aligned to the mouse mm10 genome sequence using Bowtie2 (v2.2.5)64 and Scientific). Specimens were visualized on an Olympus FSX100 fluorescence only uniquely-matching reads were retained. Mapping results of each ChIP and microscope (Olympus Scientific Solutions Americas Corp, Waltham, MA) at ×10 the input sample were subjected to ChIP-enriched peak calling using the magnification. At least 5 independent images were collected from each specimen. MACS265 (v2.1.1.20160309). Peak annotation was conducted using ChIPseeker Images were collected using FSX-BSW (version 03.02.12) software (Olympus). The (v1.24.0)66. cell density of AMs (Siglec-F+) or preAMs/TRMs (F4/80+) was measured based on cell numbers counted in the fields using cellSens Dimensions Imaging Software scRNA-seq data processing. Paired-end read counts for all known murine genes version 1.15 (Olympus). underwent barcode and unique molecular identifier (UMI) deconvolution. They were subsequently aligned to the mm10 reference genome using the 10× Cell RangerTM [v2.1] pipeline67,68. A total of 2000–3000 cells were captured, with Bulk RNA sequencing. To achieve optimal library quality, cDNA synthesis and 53,000–73,000 mean reads per cell, and 1200–1300 median genes per cell. Datasets preamplification for total RNA of the sorted lung preAMs from WT or Csf1ri- were subsequently analyzed using R-3.3.2 package Seurat’s Canonical Correlation Cre fl/fl Hdac3 cKO embryos at E18.5 were performed using the SMART-Seq v4 Ultra Analysis (CCA) workflow to characterize the differences between pooled samples69. ’ Low Input RNA Kit (Clontech, Mountain View, CA) following the manufacturer s The following quality control metrics were employed. To identify potential fl fi instructions. Brie y, cDNA was pre-ampli ed by a thermal cycler (Applied Bio- doublets, cells expressing uncharacteristically high numbers of genes (>2500) were – – system) using 12 14 cycles and sheered into 150 400 bp fragments using a Bior- excluded (double the median number of genes). Low-quality cells were excluded uptor Pico sonicator (Diagenode, Denville, NJ). The DNA ends were repaired using based on a low number of genes detected (<300) or having high mitochondria ™ End-It DNA End-Repair Kit (Epicentre, Charlotte, SC). The end-repaired DNA genetic content (>5.0%). In addition, uninteresting sources of variation within the fragments were ligated to adapters by T4 ligase (New England Bioloabs, Ipswich, data were removed. Genes removed include sex-specific genes (Xist and all Y ′ ′ → ′ MA) after a single A was added at 3 by Klenow Fragment (3 5 Exo-) (New genes), ribosomal structural proteins as identified by gene ontology England Bioloabs). MinElute Reaction Cleanup Kit (Qiagen, Gaithersburg, MD) was term GO:0003735 and the RPG: Ribosomal Protein Gene database70, non-coding used for DNA cleanup after each step. Illumina P5 and P7 primers with indicated rRNAs, Hbb, and genes expressed in <3 cells. fi fi adapters (Supplementary Table 3) were used for nal library ampli cation. The The global-scaling normalization method, LogNormalize, was employed to – fi fi fi 150 400 bp nal product was puri ed on a 2% E-gel (Thermo Fisher Scienti c). normalize gene expression measurements of each cell by the total expression, Sequencing was performed by the DNA sequencing core facility at University of multiplying this by a factor of 10,000, followed by log-transformation. Highly Michigan using a 50 bp single end read setup on the Illumina Hiseq 4000 platform. variable genes in each data analysis were identified, and the intersecting top 500 genes in each dataset were used for clustering and downstream analyses. Datasets underwent scaling based on nUMI and percent mitochondria gene content. The Single-cell RNA sequencing. Single-cell sequencing libraries were generated by an number of CCs (1:15) used to cluster cells was determined by manual inspection of ′ established protocol using the 10× Genomics Chromium Single Cell 3 Reagent Kit scree plots. Further pathway analysis was performed using IPA. (v2 Chemistry) and Chromium Single Cell Controller59. Briefly, 5000 FACS-sorted cells were loaded into each reaction for gel bead-in-emulsion (GEM) generation and cell barcoding. Reverse transcription of the GEM (GEM-RT) was performed Statistics and reproducibility. No statistical method was utilized to pre- ’ using a Veriti™ 96-Well Fast Thermal Cycler, (Applied Biosystems) at 53 °C for 45 determinesamplesize.Students unpaired two-tailed t tests were performed min, 85 °C for 5 min, and a 4 °C hold. cDNA amplification was performed after using GraphPad Prism 8.4.3 software (GraphPad, La Jolla, CA) unless otherwise fi fi fi * GEM-RT cleanup with Dynabeads MyOne Silane (Thermo Fisher Scientific) using speci ed. Statistical signi cance is displayed as: N.S., not signi cant; P <0.05; ** *** the same thermocycler (98 °C for 3 min, 98 °C for 15 s, 67 °C for 20 s, 72 °C for 1 P < 0.01; P <0.001. min, followed by a 12 cycle repeat at 72 °C for 1 min and a 4 °C hold). Amplified Bulk RNA-seq (three biologically independent samples/group), scRNA-seq cDNA was cleaned up with SPRIselect Reagent Kit (Beckman Coulter, Brea, CA) (single sample/group), and ChIP-seq (single sample/group) were performed one fl that was followed by a library construction procedure, including fragmentation, time. Immuno uorescence, bone marrow chimera, ChIP-qPCR and qRT-PCR end repair, adapter ligation, and library amplification. An Agilent 2100 Bioanalyzer experiments were performed twice independently with similar results. All other (Agilent, Santa Clara, CA) was used for library quality control. Libraries were experiments were performed at least three times independently with similar results. sequenced on an Illumina HiSeq4000 using a paired-end flow cell: Read 1, 26 cycles; i7 index, 8 cycles; Read 2, 98 cycles. Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

ChIP sequencing. Cell pellets from freshly-isolated AMs (1.38 × 106 and 0.77 × 106 cells for WT and cKO, respectively) of adult Hdac3fl/fl WT (n = 13) or Csf1ri- Data availability CreHdac3fl/fl cKO mice (n = 26) were directly frozen at −80 °C. The samples were All ChIP-seq, bulk RNA-seq, and scRNA-seq data reported here have been deposited in then submitted to EpiGentek (Farmingdale, NY) for ChIP, library preparation, and the Gene Expression Omnibus (GEO) under accession number SuperSeries GSE122533. sequencing. Briefly, the cells were fixed with 1% formaldehyde and chromatin was The data of PPAR-γ microarray (GSE602497) had been previously disclosed in GEO by isolated using a ChromaFlashTM Chromatin Extraction Kit. Chromatin was other researches. DAVID Bioinformatics Resources 6.860 and RPG: Ribosomal Protein sheared using the Episonic2000 Sonicatin System. ChIP was performed using an Gene database70 are both publicly accessible. The data supporting this study are available antibody against HDAC3 (Abcam, #ab7030, polyclonal, 2 µg/500 µl assay buffer) in the Article, Supplementary Information, Source Data, or available from the authors on the chromatin, and input DNA (without immunoprecipitation) was used as upon reasonable requests. The reporting summary and editorial checklist for this article background. After adapter linking, the library was size selected (100–300 bp) and are available as a Supplementary file. Source data are provided with this paper. PCR amplified. Ten nanomolar of each sample library was provided for next generation sequencing on a HiSeq 4000. Received: 14 November 2019; Accepted: 3 July 2020;

ChIP-qPCR. The ChIP assay was performed on 80–90% confluent cultures using an EZ-ChIP kit (Millipore, Burlington, MA, #17-371) and 1 μg of anti-RNA Poly- merase II antibody (Millipore, #05-623B, clone CTD4H8, 1 μg/ml), 2 μg of anti- HDAC3 antibody (Abcam, #ab32369, clone Y415, 2 μg/ml), 1 μg of anti-PPAR-γ antibody (Cell Signaling, Danvers, MA, #2443 S, clone 81B8, 1 μg/ml), and 1 μgof rabbit IgG (Abcam, #ab171870, polyclonal, 1 μg/ml) per reaction. DNA-relative References fi enrichment was determined by normalization to an input genomic DNA. All ChIP 1. Hussell, T. & Bell, T. J. Alveolar macrophages: plasticity in a tissue-speci c – experiments were obtained from independent chromatin preparations, and all context. Nat. Rev. Immunol. 14,81 93 (2014). quantitative real-time PCR reactions were performed in duplicates for each sample 2. Schulz, C. et al. A lineage of myeloid cells independent of Myb and on each amplicon. Primers for the ChIP-qPCR are listed in Supplementary Table 2. hematopoietic stem cells. Science 336,86–90 (2012).

NATURE COMMUNICATIONS | (2020) 11:3822 | https://doi.org/10.1038/s41467-020-17630-6 | www.nature.com/naturecommunications 13 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-17630-6

3. Perdiguero, E. G. et al. The origin of tissue-resident macrophages: when an 34. Fletcher, J. S. et al. STAT3 inhibition reduces macrophage number and tumor erythro-myeloid progenitor is an erythro-myeloid progenitor. Immunity 43, growth in neurofibroma. Oncogene 38, 2876–2884 (2019). 1023–1024 (2015). 35. Summers, A. R. et al. HDAC3 is essential for DNA replication in 4. Hoeffel, G. et al. C-Myb(+) erythro-myeloid progenitor-derived fetal hematopoietic progenitor cells. J. Clin. Invest. 123, 3112–3123 (2013). monocytes give rise to adult tissue-resident macrophages. Immunity 42, 36. Thapa, P., Romero Arocha, S., Chung, J. Y., Sant’Angelo, D. B. & Shapiro, V. 665–678 (2015). S. Histone deacetylase 3 is required for iNKT cell development. Sci. Rep. 7, 5. Mass, E. et al. Specification of tissue-resident macrophages during 5784 (2017). organogenesis. Science 353, aaf4238 (2016). 37. Ginhoux, F. et al. Fate mapping analysis reveals that adult microglia derive 6. Kopf, M., Schneider, C. & Nobs, S. P. The development and function of lung- from primitive macrophages. Science 330, 841–845 (2010). resident macrophages and dendritic cells. Nat. Immunol. 16,36–44 (2015). 38. Alenghat, T. et al. Histone deacetylase 3 coordinates commensal-bacteria- 7. Schneider, C. et al. Induction of the nuclear receptor PPAR-gamma by the dependent intestinal homeostasis. Nature 504, 153–157 (2013). cytokine GM-CSF is critical for the differentiation of fetal monocytes into 39. Remmerie, A. & Scott, C. L. Macrophages and lipid metabolism. Cell alveolar macrophages. Nat. Immunol. 15, 1026–1037 (2014). Immunol. 330,27–42 (2018). 8. Guilliams, M. et al. Alveolar macrophages develop from fetal monocytes that 40. Vecchiarelli, A. et al. Role of human alveolar macrophages as antigen- differentiate into long-lived cells in the first week of life via GM-CSF. J. Exp. presenting cells in Cryptococcus neoformans infection. Am. J. Respir. Cell Mol. Med. 210, 1977–1992 (2013). Biol. 11, 130–137 (1994). 9. Hashimoto, D. et al. Tissue-resident macrophages self-maintain locally 41. Blank, F. et al. Size-dependent uptake of particles by pulmonary antigen- throughout adult life with minimal contribution from circulating monocytes. presenting cell populations and trafficking to regional lymph nodes. Am. J. Immunity 38, 792–804 (2013). Respir. Cell Mol. Biol. 49,67–77 (2013). 10. Machiels, B. et al. A gammaherpesvirus provides protection against allergic 42. Briggs, J. A. et al. The dynamics of gene expression in vertebrate asthma by inducing the replacement of resident alveolar macrophages with embryogenesis at single-cell resolution. Science 360, aar5780 (2018). regulatory monocytes. Nat. Immunol. 18, 1310–1320 (2017). 43. Wu, X. et al. Research techniques made simple: single-cell RNA sequencing 11. Yu, X. et al. The cytokine TGF-beta promotes the development and and its applications in dermatology. J. Invest. Dermatol. 138, 1004–1009 homeostasis of alveolar macrophages. Immunity 47, 903–912 e904 (2017). (2018). 12. Amit, I., Winter, D. R. & Jung, S. The role of the local environment and 44. Loyher, P. L. et al. Macrophages of distinct origins contribute to tumor epigenetics in shaping macrophage identity and their effect on tissue development in the lung. J. Exp. Med. 215, 2536–2553 (2018). homeostasis. Nat. Immunol. 17,18–25 (2016). 45. Schmitz, G., Langmann, T. & Heimerl, S. Role of ABCG1 and other ABCG 13. Clark, M. D. et al. Structural insights into the assembly of the histone family members in lipid metabolism. J. Lipid Res. 42, 1513–1520 (2001). deacetylase-associated Sin3L/Rpd3L corepressor complex. Proc. Natl Acad. 46. Lam, B. K. & Austen, K. F. Leukotriene C4 synthase: a pivotal enzyme in Sci. USA 112, E3669–3678 (2015). cellular biosynthesis of the cysteinyl leukotrienes. Prostaglandins Other Lipid 14. Emmett, M. J. et al. Histone deacetylase 3 prepares brown adipose tissue for Mediat. 68-69, 511–520 (2002). acute thermogenic challenge. Nature 546, 544–548 (2017). 47. Weinberg, D. S. & Unanue, E. R. Antigen-presenting function of alveolar 15. Dovey, O. M., Foster, C. T. & Cowley, S. M. Histone deacetylase 1 (HDAC1), macrophages: uptake and presentation of Listeria monocytogenes. J. Immunol. but not HDAC2, controls embryonic stem cell differentiation. Proc. Natl Acad. 126, 794–799 (1981). Sci. USA 107, 8242–8247 (2010). 48. Angajala, A. et al. Diverse roles of mitochondria in immune responses: novel 16. Guan, J. S. et al. HDAC2 negatively regulates memory formation and synaptic insights into immuno-metabolism. Front. Immunol. 9, 1605 (2018). plasticity. Nature 459,55–60 (2009). 49. Yao, Y. et al. CyTOF supports efficient detection of immune cell subsets from 17. Feng, D. et al. A circadian rhythm orchestrated by histone deacetylase 3 small samples. J. Immunol. Methods 415,1–5 (2014). controls hepatic lipid metabolism. Science 331, 1315–1319 (2011). 50. Yoon, S. & Eom, G. H. HDAC and HDAC inhibitor: from cancer to 18. Moser, M. A., Hagelkruys, A. & Seiser, C. Transcription and beyond: the role cardiovascular diseases. Chonnam Med. J. 52,1–11 (2016). of mammalian class I lysine deacetylases. Chromosoma 123,67–78 (2014). 51. Suraweera, A., O’Byrne, K. J. & Richard, D. J. Combination therapy with 19. Eckschlager, T., Plch, J., Stiborova, M. & Hrabeta, J. Histone deacetylase histone deacetylase inhibitors (HDACi) for the treatment of cancer: achieving inhibitors as anticancer drugs. Int. J. Mol. Sci. 18, 1414 (2017). the full therapeutic potential of HDACi. Front. Oncol. 8, 92 (2018). 20. Royce, S. G. & Karagiannis, T. C. Histone deacetylases and their inhibitors: 52. Roche, J. & Bertrand, P. Inside HDACs with more selective HDAC inhibitors. new implications for asthma and chronic respiratory conditions. Curr. Opin. Eur. J. Med. Chem. 121, 451–483 (2016). Allergy Clin. Immunol. 14,44–48 (2014). 53. Lavin, Y. et al. Innate immune landscape in early lung adenocarcinoma by 21. van den Bosch, T., Kwiatkowski, M., Bischoff, R. & Dekker, F. J. Targeting paired single-cell analyses. Cell 169, 750–765 e717 (2017). transcription factor lysine acetylation in inflammatory airway diseases. 54. Knutson, S. K. et al. Liver-specific deletion of histone deacetylase 3 disrupts Epigenomics 9, 1013–1028 (2017). metabolic transcriptional networks. EMBO J. 27, 1017–1028 (2008). 22. Chen, X. et al. Requirement for the histone deacetylase Hdac3 for the 55. Deng, L. et al. A novel mouse model of inflammatory bowel disease links inflammatory gene expression program in macrophages. Proc. Natl Acad. Sci. mammalian target of rapamycin-dependent hyperproliferation of colonic USA 109, E2865–2874 (2012). epithelium to inflammation-associated tumorigenesis. Am. J. Pathol. 176, 23. Mullican, S. E. et al. Histone deacetylase 3 is an epigenomic brake in 952–967 (2010). macrophage alternative activation. Genes Dev. 25, 2480–2488 (2011). 56. Caton, M. L., Smith-Raska, M. R. & Reizis, B. Notch-RBP-J signaling controls 24. Gomez Perdiguero, E. et al. Tissue-resident macrophages originate from yolk- the homeostasis of CD8- dendritic cells in the spleen. J. Exp. Med. 204, sac-derived erythro-myeloid progenitors. Nature 518, 547–551 (2015). 1653–1664 (2007). 25. Yao, Y. et al. Micro RNAs are required for Langerhans cell, skin- and lung- 57. Ruzankina, Y. et al. Deletion of the developmentally essential gene ATR in resident macrophage ontogeny. J. Allergy Clin. Immunol. 142, 976–978 (2018). adult mice leads to age-related phenotypes and stem cell loss. Cell Stem Cell 1, 26. Perdiguero, E. G. & Geissmann, F. The development and maintenance of 113–126 (2007). resident macrophages. Nat. Immunol. 17,2–8 (2016). 58. He, W. et al. Adipose-specific peroxisome proliferator-activated receptor 27. Dick, S. A. et al. Self-renewing resident cardiac macrophages limit adverse gamma knockout causes insulin resistance in fat and liver but not in muscle. remodeling following myocardial infarction. Nat. Immunol. 20,29–39 (2019). Proc. Natl Acad. Sci. USA 100, 15712–15717 (2003). 28. MacParland, S. A. et al. Single cell RNA sequencing of human liver reveals 59. Zheng, G. X. et al. Massively parallel digital transcriptional profiling of single distinct intrahepatic macrophage populations. Nat. Commun. 9, 4383 (2018). cells. Nat. Commun. 8, 14049 (2017). 29. Yuan, Z. et al. Triggering receptor expressed on myeloid cells 1 (TREM-1)- 60. Huang da, W., Sherman, B. T. & Lempicki, R. A. Systematic and integrative mediated Bcl-2 induction prolongs macrophage survival. J. Biol. Chem. 289, analysis of large gene lists using DAVID bioinformatics resources. Nat. Protoc. 15118–15129 (2014). 4,44–57 (2009). 30. Yuan, Z. et al. TREM-1-accentuated lung injury via miR-155 is inhibited by 61. Subramanian, A. et al. Gene set enrichment analysis: a knowledge-based LP17 nanomedicine. Am. J. Physiol. Lung Cell Mol. Physiol. 310, L426–438 approach for interpreting genome-wide expression profiles. Proc. Natl Acad. (2016). Sci. USA 102, 15545–15550 (2005). 31. Pagliari, L. J., Perlman, H., Liu, H. & Pope, R. M. Macrophages require 62. Mootha, V. K. et al. PGC-1alpha-responsive genes involved in oxidative constitutive NF-kappaB activation to maintain A1 expression and phosphorylation are coordinately downregulated in human diabetes. Nat. mitochondrial homeostasis. Mol. Cell Biol. 20, 8855–8865 (2000). Genet. 34, 267–273 (2003). 32. Hill, A. A. et al. Activation of NF-kappaB drives the enhanced survival of 63. Kidder, B. L., Hu, G. & Zhao, K. ChIP-Seq: technical considerations for adipose tissue macrophages in an obesogenic environment. Mol. Metab. 4, obtaining high-quality data. Nat. Immunol. 12, 918–922 (2011). 665–677 (2015). 64. Langmead, B., Trapnell, C., Pop, M. & Salzberg, S. L. Ultrafast and memory- 33. Liu, H. et al. Serine phosphorylation of STAT3 is essential for Mcl-1 efficient alignment of short DNA sequences to the . Genome expression and macrophage survival. Blood 102, 344–352 (2003). Biol. 10, R25 (2009).

14 NATURE COMMUNICATIONS | (2020) 11:3822 | https://doi.org/10.1038/s41467-020-17630-6 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-17630-6 ARTICLE

65. Zhang, Y. et al. Model-based analysis of ChIP-Seq (MACS). Genome Biol. 9, Competing interests R137 (2008). The authors declare no competing interests. 66. Yu, G., Wang, L. G. & He, Q. Y. ChIPseeker: an R/Bioconductor package for ChIP peak annotation, comparison and visualization. Bioinformatics 31, Additional information 2382–2383 (2015). 67. Macosko, E. Z. et al. Highly parallel genome-wide expression profiling of Supplementary information is available for this paper at https://doi.org/10.1038/s41467- individual cells using nanoliter droplets. Cell 161, 1202–1214 (2015). 020-17630-6. 68. Magella, B. et al. Cross-platform single cell analysis of kidney development shows stromal cells express Gdnf. Dev. Biol. 434,36–47 (2018). Correspondence and requests for materials should be addressed to L.Z. or Q.-S.M. 69. Butler, A., Hoffman, P., Smibert, P., Papalexi, E. & Satija, R. Integrating single- cell transcriptomic data across different conditions, technologies, and species. Peer review information Nature Communications thanks the anonymous reviewer(s) for Nat. Biotechnol. 36, 411–420 (2018). their contribution to the peer review of this work. 70. Nakao, A., Yoshihama, M. & Kenmochi, N. RPG: the ribosomal protein gene database. Nucleic Acids Res. 32, D168–170 (2004). Reprints and permission information is available at http://www.nature.com/reprints Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Acknowledgements We thank Dr. Hongmei Peng, Dr. Yu Wang, Dr. Jun Zhang, Dr. Xuan Wang, and Dr. Yi Yang for helping prepare samples. We thank Dr. Keji Zhao from the Lung and Blood Open Access Institute, National Institutes of Health, for valuable technical support in RNA-Seq. This This article is licensed under a Creative Commons study is partially supported by National Institutes of Health grants Attribution 4.0 International License, which permits use, sharing, R01AR069681, R61AR076803, R01AI119041 (Q-S.M.), R01AR087659 (L.Z.), and the adaptation, distribution and reproduction in any medium or format, as long as you give Henry Ford Immunology Program grants (T71016, Q-S.M.; T71017, L. Z.). appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless Author contributions indicated otherwise in a credit line to the material. If material is not included in the ’ Q-S.M., Y.Y., and L.Z. designed the study; Y.Y. performed majority of the experiments article s Creative Commons license and your intended use is not permitted by statutory and drafted the manuscript; Q.L. performed most of the embryonic experiments; I.A., regulation or exceeds the permitted use, you will need to obtain permission directly from J.G., and X.W. analyzed sequencing data; X.W. prepared cDNA libraries for sequencing; the copyright holder. To view a copy of this license, visit http://creativecommons.org/ N.K. assisted in qRT-PCR and flow cytometry work; C.Y. and Q.Y. maintained mouse licenses/by/4.0/. colonies and performed genotyping; Q.-S.M., L.Z., Y.Y., and Q.L. analyzed the data; Z.D. supplied materials; F.G. contributed to the interpretation of the results; Q-S.M. and L.Z. © The Author(s) 2020 supervised the overall study and finalized the manuscript.

NATURE COMMUNICATIONS | (2020) 11:3822 | https://doi.org/10.1038/s41467-020-17630-6 | www.nature.com/naturecommunications 15