Runx3 Programs CD8+ T Cell Residency in Non-Lymphoid Tissues and Tumours J

Letter doi:10.1038/nature24993

Runx3 programs CD8+ T cell residency in non-lymphoid tissues and tumours J. Justin Milner1, Clara Toma1*, Bingfei Yu1*, Kai Zhang2, Kyla Omilusik1, Anthony T. Phan1, Dapeng Wang3, Adam J. Getzler3, Toan Nguyen1, Shane Crotty4,5, Wei Wang2,6,7, Matthew E. Pipkin3 & Ananda W. Goldrath1

+ Tissue-resident memory CD8 T (TRM) cells are found at common genes differentially expressed between splenic and non-lymphoid sites of pathogen exposure, where they elicit rapid and robust populations on day 7 of infection revealed two distinct gene expres- protective immune responses1,2. However, the molecular signals sion programs that segregated circulating (peripheral blood lympho- that control TRM cell differentiation and homeostasis are not fully cytes, spleen, TCM and TEM) from non-lymphoid (kidney and IELs) understood. Here we show that mouse TRM precursor cells represent P14 cells, independent of the infection time point (Fig. 1b). Lymph a unique CD8+ T cell subset that is distinct from the precursors of node or splenic KLRG1loCD127hi memory precursor cells preferentially circulating memory cell populations at the levels of gene expression give rise to circulating memory populations, whereas shorter-lived and chromatin accessibility. Using computational and pooled KLRG1hiCD127lo terminal effector cells exhibit less memory potential 3. in vivo RNA interference screens, we identify the transcription Day 7 IEL P14 cells comprising the precursors of TRM cells were tran- factor Runx3 as a key regulator of TRM cell differentiation and scriptionally distinct from splenic memory precursor cells (Fig. 1c). lo homeostasis. Runx3 was required to establish TRM cell populations This is notable, as IEL TRM cells are predominantly KLRG1 (ref. 9) and in diverse tissue environments, and supported the expression of preferentially differentiate from lymphoid-derived KLRG1lo precursors crucial tissue-residency genes while suppressing genes associated seeding non-lymphoid tissues on days 4.5–7 of infection10 (Extended 5 with tissue egress and recirculation. Furthermore, we show that Data Fig. 1a–c), consistent with studies of skin TRM cells . Thus, the TRM human and mouse tumour-infiltrating lymphocytes share a core precursor cell populations in non-lymphoid tissues are transcription- tissue-residency gene-expression signature with TRM cells that is ally distinct from circulating effector cells as well as memory precursor associated with Runx3 activity. In a mouse model of adoptive T cell cells on day 7 of infection, and most of the TRM cell transcriptional therapy for melanoma, Runx3-deficient CD8+ tumour-infiltrating program is already established at this time point, before contraction of lymphocytes failed to accumulate in tumours, resulting in greater the CD8+ T cell population. rates of tumour growth and mortality. Conversely, overexpression of As chromatin accessibility is a key determinant of cell identity and Runx3 enhanced tumour-specific CD8+ T cell abundance, delayed fate, we profiled non-lymphoid and splenic effector populations using tumour growth, and prolonged survival. In addition to establishing an assay for transposase-accessible chromatin with high-throughput Runx3 as a central regulator of TRM cell differentiation, these results sequencing (ATAC–seq) on day 7 of infection. Uniquely accessible provide insight into the signals that promote T cell residency in non- chromatin regions were identified in IEL P14 cells near genes character- lymphoid sites, which could be used to enhance vaccine efficacy or istically expressed in mature TRM cells (for example, Cd69 and Nr4a1), adoptive cell therapy treatments that target cancer. whereas genes that promote T cell re-circulation (for example, Klf2 Long-lived memory T cells provide protection from reinfection and and S1pr1) exhibited loss of accessible regions (Extended Data Fig. 2a). can serve as endogenous defenders against tumour growth3. Memory Principal component analysis (PCA) highlighted that, despite day 7 CD8+ T cell populations can be broadly segregated into circulating cen- being an ‘effector’ time point, the global chromatin landscape markedly + tral memory (TCM) and effector memory (TEM) T cells as well as tissue- differs between effector CD8 T cells located in the spleen, including resident memory (TRM) T cells that primarily reside in non-lymphoid memory precursor cells, and those located in non-lymphoid tissues 4 + tissues without egress . Circulating memory CD8 T cells and TRM cells (Fig. 1d). The unique chromatin configuration of differentiating TRM exhibit distinct gene-expression profiles5–7; however, the early transcrip- cells is consistent with the notable transcriptional differences observed tional identity of differentiating TRM cells and the signals controlling (Fig. 1a–c) and foreshadows the distinct fates of antigen-specific cells their fate are not yet fully appreciated. Here, we used an established in the spleen relative to non-lymphoid tissues. Thus, precursors of TRM infection model with P14 T cell receptor transgenic CD8+ T cells cells in non-lymphoid sites are a unique and distinct CD8+ T cell subset responsive to the lymphocytic choriomeningitis virus (LCMV) relative to effector cells in the lymphoid compartment, including the glycoprotein 33–41 peptide (GP33–41) presented by major histocompati memory precursor cell population. bility complex (MHC) class I H-2Db. In this acute infection model, Specification of CD8+ T cell fate during infection is dependent on adoptively transferred P14 cells located in non-lymphoid tissues the integrated activity of multiple transcription factors3; regulators of 6 6 11 12 on day 7 of infection began to upregulate molecules characteristic TRM cell formation include Hobit , Blimp1 , Nr4a1 , Eomes and 8 12,13 of TRM cells , including key tissue-retention molecules CD103 and T-bet . To facilitate a broader understanding of the transcriptional CD69 (Extended Data Fig. 1a). Gene expression analysis revealed that network driving TRM cell differentiation, we used a combined screening 90–96% of the genes upregulated in mature P14 TRM cells in the kidney approach, consisting of a computational strategy integrating ATAC– parenchyma or intraepithelial lymphocyte (IEL) compartment of the seq data, transcriptional profiling and personalized PageRank analysis small intestine were increased in TRM precursor cells relative to splenic to predict regulatory transcription factors, and a pooled in vivo RNA effector cells on day 7 of infection (Fig. 1a). Furthermore, analysis of interference (RNAi) loss-of-function screen targeting putative TRM cell

1Division of Biological Sciences, University of California, San Diego, La Jolla, California, USA. 2Bioinformatics and Systems Biology Graduate Program, University of California, San Diego, La Jolla, California, USA. 3Department of Immunology and Microbiology, The Scripps Research Institute, Jupiter, Florida, USA. 4Division of Vaccine Discovery, La Jolla Institute for Allergy and Immunology, La Jolla, California, USA. 5Division of Infectious Diseases, Department of Medicine, University of California, San Diego, La Jolla, California, USA. 6Department of Chemistry and Biochemistry, University of California, San Diego, La Jolla, California, USA. 7Department of Cellular and Molecular Medicine, University of California, San Diego, La Jolla, California, USA. *These authors contributed equally to this work.

Figure 1 | Computational and loss-of-function RNAi screens d, PCA of differential global chromatin accessibility of subsets on identify transcriptional regulators of TRM cell differentiation. a, Top, day 7 of infection identified by ATAC–seq analysis. e, Combinatorial comparison of gene expression of IELs (left) and kidney TRM cells (right) screening approach. TF, transcription factors. f, Transcription factors relative to TCM and TEM subsets on day 35 of LCMV infection. Red denotes with a PageRank score of ≥1.5-fold change for day 7 non-lymphoid cells genes increased in TRM relative to TCM and TEM cells; blue denotes genes compared to day 7 splenic cells are included in the heat map; proteins increased in TCM and TEM relative to TRM cells. Bottom, comparison of known to regulate TRM cell formation are in bold font. g, Enrichment of differentially expressed genes in mature TRM cells (from top panel) in cells shRNAmir constructs in IEL TRM cells relative to splenic TCM cells from from the spleen, IELs or kidney on day 7 of infection. b, Differentially the RNAi screen, reported as the average Z-score from 3 independent expressed genes between splenic, IEL and kidney populations on day 7 screens, in which each independent screen was performed by pooling of infection were compared among effector and memory CD8+ T cell DNA from sorted P14 cells from n = 15, n = 18 and n = 18 mice. Each time subsets. Populations are ordered by hierarchical clustering with Pearson point represents an individual experiment consisting of 2–3 biological correlation. c, PCA of differentially expressed genes among day 7 subsets replicates, in which cells from 2–10 mice were pooled for each and naive P14 cells. MP, memory precursors; TE, terminal effectors. replicate (a–d).

regulators identified through the computational approach (Fig. 1e). previously described role in TRM cells (Fig. 1f, Supplementary Table 1). We recently demonstrated that analysis of accessible transcription We evaluated both barrier (IEL) and non-barrier (kidney) TRM cells factor-binding motifs and target gene expression yielded insight into to reveal transcription factors important to TRM cell differentiation factors with regulatory functions in the differentiation of circulating independent of the tissue. In addition, a key strength of this com- memory CD8+ T cells14. Using this approach and the personalized putational screen is that influential roles of differentially expressed PageRank analysis15, we predicted several transcription factors with transcription factors as well as those with homogenous expression established regulatory roles in controlling TRM cell differentiation can be anticipated (Extended Data Fig. 2b). To establish functional 6 11 12 12,13 (such as Blimp1 , Nr4a1 , Eomes and T-bet ) and many with no relevance for predicted regulators of TRM cell formation identified

LM–GP33–41 abCD45.1 P14 LCMV Spleen CD45.1 P14 o.g. Figure 2 | Runx3 is essential for the Runx3 shRNA mLN Runx3 shRNA differentiation and long-term maintenance IEL >D20 IEL TRM >D20 IEL TTM + CD45.1.2 P14 CD69+ CD45.2 P14 of CD8 TRM cells. a, Congenically distinct Ctrl shRNA CD103+ IEL Ctrl shRNA CD69+ CD69+ P14 cells were transduced with retroviruses Spleen mLN IEL CD103+ IEL Spleen mLN IEL CD103+ IEL >Day 20 2 >Day 20 0 encoding Runx3 shRNAmir or a control (ctrl) –2 0 shRNAmir, mixed at a 1:1 ratio, and transferred shRNA/ 44 56 26 6

29 42 4 2 shRNA/ 71 58 96 98 56 44 73 94 CD45.1 CD45.2 –4 * –2 * to recipient mice subsequently infected with CD45.1 CD45.2 Runx3 ( ctrl shRNA) –6 LCMV, and the ratio of transduced cells was Runx3

2 –4 ( *** ctrl shRNA) * 2

log –8 evaluated on day 23 or 26 in the indicated c *** d log –6 Thy1.1 P14 Thy1.1 P14 *** tissues. mLN, mesenteric lymph nodes. Runx3fl/fl Ert2-Cre+ YFP LCMV Tamoxifen Runx3fl/fl Ert2-Cre+ YFP LCMV Tamoxifen b, Ratio of transduced transferred cells in D15 IEL T IEL TRM RM Thy1.1.2 P14 Thy1.2 D2–D5 Thy1.1.2 P14 Thy1.2 D6–D8 indicated tissues on day 32 of enteric LM– Runx3+/+ Ert2-Cre+ YFP Runx3+/+ Ert2-Cre+ YFP CD69+ CD69+ GP33–41 infection. o.g., oral gavage. c, Ratio of + + ) fl/fl + fl/fl Spleen mLN IEL CD103 IEL Spleen mLN IEL CD103 IEL Day 15

+/+ 2 transferred Runx3 Ert2-Cre YFP (Runx3 ) ) Day 7 Day 16 +/+ + +/+ +/+ 0 0 to Runx3 Ert2-Cre YFP (Runx3 ) P14 D7

41 51 20 5 /Runx3 38 41 24 10 –2 59 49 80 95 –2 cells in indicated tissues on days 6/7 and

62 59 76 90 fl/fl Thy1.1 /Runx3 –4 P = 0.1 –4 P = 0.07 fl/fl *** Thy1.2 15/16 of LCMV infection (top) or enteric –6

Runx3 –6 D16 ( *** 2 LM–GP33–41 infection (bottom). d, Ratio of –8 *** *** 18 36 4 0.6 Runx3 ( –8 fl/fl +/+

99.4 log 82 64 96 2 *** Thy1.1 –10 Runx3 to Runx3 P14 cells on day 15 Thy1.2 log *** *** of LCMV infection. e, Ratio of Runx3fl/fl to e + + LM–GP33–41 Thy1.1 P14 Tamoxifen / Thy1.1 P14 o.g. Runx3 P14 cells on day 35 of infection. Data Runx3fl/fl Ert2-Cre+ YFP Tamoxifen Runx3fl/fl Ert2-Cre+ YFP LCMV or vehicle are mean ± s.e.m of n = 5 (a), n = 3 (b), n = 3 IEL TRM D35 IEL T Thy1.1.2 P14 Thy1.2 D2–D5 Thy1.2 D16–D20 RM (day 7) and n = 6 (day 15/16) (c), n = 5 (d), and Runx3+/+ Ert2-Cre+ YFP Thy1.1.2 P14 Runx3+/+ Ert2-Cre+ YFP n = 5 (vehicle) and n = 3 (tamoxifen) (e). All CD69+ CD69+ Spleen mLN IEL CD103+ IEL Spleen IEL CD103+ IEL ) ) Day 7 Day 15 +/+ Vehicle Tamoxifen data are from one representative experiment of 1 2 +/+ 0 2 independent experiments, except in a, which D7 Veh. 0 49 51 69 31 27 73 8 92 51 48 48 /Runx3 is pooled from 2 independent experiments.

–2 49 52 52 ﬂ/ﬂ

/Runx3 –2

ﬂ/ﬂ < < < –4 ** *** –4 *P 0.05, ** P 0.01, ** * P 0.005 (Student’s

Runx3 * D15 Tam. ( t-test). Symbols represent an individual mouse 35 65 58 42 23 77 2 98 2 Runx3 –6 *** *** 34 10 6 –6 * (

2 66 90 94 log Thy1.1 Thy1.1 (a–e). Thy1.2 log –8 *** *** Thy1.2

16 b + through PageRank analysis, we used an RNAi screening strategy for optimal TRM cell differentiation of H-2D GP33–41 tetramer cells to test hundreds of individual microRNA-based short hairpin RNA (Extended Data Fig. 4a–d). Taken together, these data demonstrate that (shRNAmir) constructs in parallel for the ability to promote or repress Runx3 is crucial for TRM cell differentiation and maintenance. TRM cell differentiation in vivo (Fig. 1g, Supplementary Table 2). Several transcription factors with established roles in regulating TRM 13 6 17 12,13 cells were identified (such as Nr4a1 , Blimp1 , Klf2 and T-bet ), as CD69+ + a Spleen mLN IEL CD103+ IEL well as factors with previously unknown functions in controlling CD8 2.7× 3.8× Spleen mLN 6 * ** TRM formation such as Nr4a3 and Runx3 (Fig. 1g). D8 IEL + 54 46 30 70 38 62 58 42 0.9× 1.8× + Runx3 CD69 Runx3 is a well-established regulator of CD8 T cell thymo- 4 NS * CD103+ IEL + 18 V/GFP-RV 0.4× cyte development , supports cytotoxic activity of mature CD8 *** 19,20 + D13 2 T cells , and controls CD4 T cell localization within the intestinal 56 44 38 62 71 29 77 23 CD45.1 21 Runx3-R 0 epithelium . Although little is known regarding a role for Runx3 in CD45.2 Day 8 Day 13 + bc CD8 TRM cells, both computational and functional screens identified GFP-RV Runx3-RV V fl/fl +/+ 23 19 8 4455 Runx3 as a putative regulator of TRM cell fate specification (Fig. 1f, g) D8 Core circulating Runx3Runx3Runx3-R signature e Circulating signature despite relatively uniform Runx3 expression in circulating and resident 5 15 + IEL Lung 0 NES = –2.1 0 NES = –1.91 CD8 T cell subsets (Extended Data Figs 2b, 3a). We validated a role 22 6655 1 91 –0.2 FDR q = 0 –0.2 FDR q = 0 Kid Skin –0.4 –0.4 for Runx3 through a 1:1 mixed transfer of P14 cells transduced with D13 93 Brain 1 4 e control (Cd19 shRNAmir) or Runx3 shRNAmir-encoding retroviruses CD69 CD103 IEL culating signatur WT Runx3fl/fl Runx3-RV WT into mice that were subsequently infected with LCMV (Fig. 2a). Runx3 GFP-RV Cir Runx3-RV 100 *** e shRNAmir suppressed Runx3 expression (Extended Data Fig. 3b) Residency signature 80 *** Lung 4 NES = 1.91 NES = 2.31 and impaired the formation of IEL TRM cells relative to circulating cells * Kid Enrichment scor FDR q = 0 4 60 Skin 2 FDR q = 0 121 2 40 0 (Fig. 2a and Extended Data Fig. 3c, d), consistent with the RNAi screen. equency Brain 0 Fr 20 Furthermore, Runx3 RNAi also impaired TRM cell differentiation in IEL fl/fl

0 Residency signatur WT Runx3 Runx3-RV WT D8 D13 D8 D13 Core residency the context of a localized enteric infection with Listeria monocytogenes signature CD103+ CD69+ expressing GP33–41 (LM–GP33–41) (Fig. 2b). –1 0 1 fl/fl Next, using a tamoxifen-inducible deletion approach, Runx3 Ert2- Figure 3 | Runx3 programs CD8+ T cell tissue-residency. a, Congenically + + + + + + Cre P14 (Runx3fl/fl) or Runx3 / Ert2-Cre P14 (Runx3 / ) cells were distinct P14 cells were transduced with retroviruses encoding Runx3- mixed 1:1 and transferred into host mice followed by LCMV or enteric cDNA (Runx3-RV; CD45.1+ cells) or control GFP (GFP-RV; CD45.1.2+ LM–GP33–41 infection (Fig. 2c). Runx3-deficiency resulted in a 2–6- cells), mixed at a 1:1 ratio, and transferred to recipient mice subsequently infected with LCMV. Ratios of transduced cells were evaluated on days 8 fold loss of splenocytes and minimal loss of mesenteric lymph node + + cells by day 15/16 of infection. However, Runx3-deficiency resulted in and 12/13 of infection. b, Frequency of CD69 and CD103 cells from a. a 50–150-fold loss of CD69+CD103+ T cells in both infection set- c, Left, relative expression of the core ‘circulating’ and ‘residency’ genes RM between Runx3-RV, Runx3fl/fl and Runx3+/+ CD8+ T cells. Right, gene set tings (Fig. 2c and Extended Data Fig. 3e). Moreover, delaying tamoxi enrichment analysis (GSEA). FDR, false discovery rate; NES, normalized fen treatment to days 6–8 or 16–20 of infection further emphasized a enrichment score; WT, wild type. Data are mean ± s.e.m of n = 5 mice distinct dependence of TRM cell differentiation on Runx3 (Fig. 2d) as (a, b) from one representative experiment of 2 independent experiments. well as a crucial role for Runx3 in maintaining TRM cell homeostasis *P < 0.05, ** P < 0.01, ** * P < 0.005 (Student’s t-test). NS, not significant. (Fig. 2e, Extended Data Fig. 3f). Furthermore, Runx3 was necessary Symbols represent an individual mouse (a, b).

a Circulating Residency b TIL PyMT Spleen PyMT c signature signature TIL B16 Spleen B16 Spleen TIL Kidney TRM TCM B16 melanoma TIL PyMT mammary TIL IEL T T Ctrl RM EM shRNA Runx3 49 51 15 85 shRNA ) ) P P

( ( Spleen

10 10 (tumour)

log log TIL GFP-RV Runx3-RV

PC2 (16.56%) Spleen 77% 23% 81% 19% T RM (LCMV) 34 66 71 29 27% 73% 34% 66% CD45.1 Expression change Expression change PC1 (49.89%) CD45.2 log2(TIL/spleen) log2(TIL/spleen) RV RV RV RV

0.15× 8.2× d e SpleenTIL: GFP-TIL: Runx3-SpleenTIL: TIL:GFP- Runx3-f )

1.5 3 No P14 transfer 6 e

*** e 100 *** Ctrl shRNA P = 0.0006 V)

mm 16 2 Runx3 shRNA 80 1.0 4 12 60 V/GFP-R 8 ** 40

0.5 2 culating signatur Fold change Survival (%) Fold change shRNA/ctrl shRNA) esidency signatur 4 20 e cir e r (Runx3-R mour volume (×10 Cor Cor Runx3 Tu ( 0 0 0 0 n en TIL TIL 0 6 12 18 0510 15 20 25 ple –1 0 1 –1 0 1 g S Splee Time (day) Time (day)

) No P14 transfer 3 100 P = 0.0004 h GFP-RV Mouse melanoma TIL Human melanoma TIL 16 mm Runx3-RV

2 80 0.4 NES = 1.29 0.4 NES = 1.40 12 FDR q = 0.022 FDR q = 0.013 60 0.2 0.2

8 * 40 0 0 Survival (%) 4 20

mour volume (×10 0 0 Tu 0 6 12 18 0510 15 20 25 Runx3hi Runx3lo Runx3hi Runx3lo Time (day) Time (day) + Figure 4 | CD8 TILs share transcriptional signatures with TRM cells the same approach as in c. f, g, Tumour growth and survival after adoptive and require Runx3 for tumour residency. a, Comparison of the core transfer of the indicated cell population. A log-rank (Mantel–Cox) test tissue-residency signature and core circulating signature (from was used to compare survival rates. h, GSEA of the core tissue-residency Fig. 3c) in B16 melanoma CD8+ TILs27 or PyMT mammary tumour CD8+ signature in Runx3hi versus Runx3lo TIL from single-cell RNA-seq analyses TILs27 relative to corresponding splenic cells. b, PCA of gene expression of mouse29 and human30 melanoma TILs. Data are mean ± s.e.m of of the core tissue-residency and circulating gene sets for TILs, TRM cells n = 5 (Runx3 shRNAmir) or n = 7 mice per group (Runx3-RV) from one or splenic subsets. c, d, Congenically distinct P14 cells were transduced representative experiment of 3 independent experiments (c, d) or data with retroviruses encoding Runx3 shRNAmir or Runx3-RV (CD45.1+ pooled from 3 independent experiments consisting of n = 21 (no P14), cells) and control shRNAmir or GFP-RV (CD45.1.2+ cells), mixed at a 1:1 n = 20 (control shRNAmir), n = 22 (Runx3 shRNAmir) (f), n = 10 ratio and transferred into mice with established B16-GP33–41 melanoma (no P14), n = 22 (GFP-RV), n = 21 (Runx3-RV) mice per group (g). tumours. Flow plots and graphs indicate ratio of transduced cells. *P < 0.05, ** P < 0.01, ** * P < 0.005 (Student’s t-test). Symbols represent e, Relative expression of the core tissue-residency and core circulating gene an individual mouse (d). sets in GFP-RV splenocytes, GFP-RV TILs, and Runx3-RV TILs following

Runx3 deletion also resulted in a loss of TRM cells in non-barrier tissues regulator of TRM cell specification (Fig. 3c, Supplementary Table 3). (salivary gland and kidney; Extended Data Fig. 5a, b), and optimal TRM Notably, we found that most of the core tissue-residency signature genes cell differentiation in the skin and lung parenchyma required Runx3 were upregulated in Runx3-overexpressing cells and downregulated (Extended Data Fig. 5c–h). Thus, the loss of TRM cells in a range of in Runx3-deficient cells. Conversely, the core signature of circulating non-lymphoid tissues indicated that Runx3 drives the formation of memory cells was enriched in Runx3-deficient cells and depleted from TRM cells independently of the tissue site. Furthermore, Runx3 was Runx3-overexpressing cells (Fig. 3c). Therefore, Runx3 promoted the required for maximal expression of granzyme B in TRM cells, although expression of tissue-residency signature genes and repressed genes char- cytokine production was unchanged (Extended Data Fig. 6a, b). acteristic of circulating cells. This conclusion was further corroborated Runx3-deficiency resulted in a greater frequency of annexin V+ cells by chromatin-immunoprecipitation followed by deep sequencing (Extended Data Fig. 6c, d), most prominently in the CD69+CD103+ (ChIP–seq) analysis23, indicating that Runx3 binding was enriched in TRM population; thus, the marked loss of TRM cells was at least in part both core tissue-residency and circulating genes relative to background due to a greater rate of apoptosis, as proliferation and trafficking were sites (Extended Data Fig. 8a). not affected (Extended Data Fig. 6e, f). Through evaluation of accessible Runx3-binding motifs from We next assessed whether enhanced expression of Runx3 could ATAC–seq analysis, we generated a regulatory Runx3-binding network augment TRM cell differentiation. Overexpression of Runx3 accelerated (Extended Data Fig. 8b) and found that Runx3 putatively regulates a + + IEL P14 CD69 CD103 TRM cell differentiation on day 8 of infection, distinct network of genes in differentiating IEL-TRM precursor cells but did not affect migration to the small intestine (Fig. 3a). Evidence of relative to splenic effector cells, including selective enrichment of genes enhanced TRM cell differentiation was further confirmed by the greater linked to cell adhesion and regulation of transcription factor activity. abundance of IEL TRM cells on day 12/13 of infection and enhanced In connection, Runx3 has been shown to cooperate with the transcrip- CD103 expression, consistent with a reported role for Runx3 in regu tion factor T-bet in many contexts19,24, yet T-bet is a potent suppressor 21,22 12,13 23 lating CD103 expression (Fig. 3b). Furthermore, overexpression of early TRM cell differentiation . ChIP–seq data indicated that of Runx3 also boosted TRM cell differentiation in the lung parenchyma Runx3 directly binds to multiple sites of the Tbx21 locus (encoding (Extended Data Fig. 7a–d). T-bet; Extended Data Fig. 8c), and Runx3-deficient CD8+ T cells Given that manipulation of Runx3 affected TRM cell formation in exhibited increased T-bet levels (Extended Data Fig. 8d). Tbx21 RNAi diverse tissue microenvironments, we constructed a core TRM cell in Runx3-deficient cells enhanced TRM cell numbers in the IEL com- + transcriptional signature by computational integration of CD8 TRM partment and restored CD103 and CD69 expression (Extended Data 5 gene-expression datasets from small intestine IELs, kidney, lung , Fig. 8e, f), but did not fully rescue TRM cell differentiation. These skin5 and brain7, to evaluate the hypothesis that Runx3 is a universal findings are consistent with Runx3 regulating multiple targets that

10. Masopust, D. et al. Dynamic T cell migration program provides resident influence TRM cell formation (Fig. 3c), including suppression of cano memory within intestinal epithelium. J. Exp. Med. 207, 553–564 (2010). nical genes associated with tissue egress (Extended Data Fig. 8g, h). 11. Boddupalli, C. S. et al. ABC transporters and NR4A1 identify a quiescent subset + It has been noted that CD8 tumour-infiltrating lymphocytes (TILs) of tissue-resident memory T cells. J. Clin. Invest. 126, 3905–3916 (2016). 12. Mackay, L. K. et al. T-box transcription factors combine with the cytokines can exhibit characteristics of TRM cells, and a positive prognosis has 25,26 TGF-β and IL-15 to control tissue-resident memory T cell fate. Immunity 43, been correlated with TILs that present qualities of TRM cells . As 1101–1111 (2015). Runx3 regulates core features of tissue residency (Fig. 3c), we assessed 13. Laidlaw, B. J. et al. CD4+ T cell help guides formation of CD103+ lung-resident + the transcriptional similarities of TILs and TRM cells and evaluated a memory CD8 T cells during influenza viral infection.Immunity 41, 633–645 role for Runx3 in controlling TIL accumulation. TILs isolated from (2014). 27 27 14. Yu, B. et al. Epigenetic landscapes reveal transcription factors that regulate mouse melanoma or mammary tumours shared approximately 70% CD8+ T cell differentiation.Nat. Immunol. 18, 573–582 (2017). of the core tissue-residency gene-expression program relative to splenic 15. Page, L., Brin, S., Motwani, R. & Winograd, T. The PageRank citation ranking: CD8+ T cells (Fig. 4a), and this relationship was further highlighted Bringing order to the web. Technical Report, Stanford InfoLab (1998). 16. Chen, R. et al. In vivo RNA interference screens identify regulators of antiviral through PCA (Fig. 4b). Utilizing an adoptive therapy model, Runx3- CD4+ and CD8+ T cell differentiation.Immunity 41, 325–338 (2014). RNAi or Runx3-overexpressing P14 cells were mixed with control P14 17. Skon, C. N. et al. Transcriptional downregulation of S1pr1 is required for the cells at a 1:1 ratio and transferred into mice with established mela- establishment of resident memory CD8+ T cells. Nat. Immunol. 14, 1285–1293 (2013). noma tumours expressing GP33–41 (Extended Data Fig. 9a). Runx3- 18. Taniuchi, I. et al. Differential requirements for Runx proteins in CD4 repression deficiency impaired TIL accumulation (Fig. 4c, d) without affecting and epigenetic silencing during T lymphocyte development. Cell 111, migration to the tumour (Extended Data Fig. 9b). Alternatively, Runx3- 621–633 (2002). 19. Cruz-Guilloty, F. et al. Runx3 and T-box proteins cooperate to establish the overexpression enhanced TIL abundance (Fig. 4c, d), expression of transcriptional program of effector CTLs.J. Exp. Med. 206, 51–59 (2009). granzyme B (Extended Data Fig. 9c) and certain core tissue-residency 20. Shan, Q. et al. The transcription factor Runx3 guards cytotoxic CD8+ effector genes while further suppressing core circulating genes (Fig. 4e). In clin- T cells against deviation towards follicular helper T cell lineage. Nat. Immunol. 28 18, 931–939 (2017). ical settings, TIL density strongly correlates with positive outcomes , 21. Reis, B. S., Rogoz, A., Costa-Pinto, F. A., Taniuchi, I. & Mucida, D. Mutual and we observed Runx3-deficient P14 cells were impaired in their expression of the transcription factors Runx3 and ThPOK regulates intestinal ability to control tumour growth, resulting in greater mortality (Fig. 4f). CD4+ T cell immunity. Nat. Immunol. 14, 271–280 (2013). 22. Grueter, B. et al. Runx3 regulates integrin α E/CD103 and CD4 expression Conversely, Runx3-overexpressing cells delayed tumour growth and − + + during development of CD4 /CD8 T cells. J. Immunol. 175, 1694–1705 (2005). prolonged survival (Fig. 4g). Notably, human CD8 TILs also exhibited 23. Lotem, J. et al. Runx3-mediated transcriptional program in cytotoxic enrichment of the core tissue-residency signature relative to circulating lymphocytes. PLoS One 8, e80467 (2013). CD8+ T cells25 (Extended Data Fig. 9d), and analysis of single-cell 24. Reis, B. S., Hoytema van Konijnenburg, D. P., Grivennikov, S. I. & Mucida, D. 29 Transcription factor T-bet regulates intraepithelial lymphocyte functional RNA sequencing (RNA-seq) data from mouse and human mela- maturation. Immunity 41, 244–256 (2014). noma TILs30 indicated that activated CD44+CD8+ T cells expressing 25. Djenidi, F. et al. CD8+CD103+ tumor-infiltrating lymphocytes are tumor-specific Runx3 exhibited enrichment of the tissue-residency gene-expression tissue-resident memory T cells and a prognostic factor for survival in lung + + cancer patients. J. Immunol. 194, 3475–3486 (2015). signature relative to CD44 CD8 TILs with low Runx3 expression 26. Ganesan, A.-P. et al. Tissue-resident memory features are linked to the levels (Fig. 4h). These data indicate that in both human and mouse magnitude of cytotoxic T cell responses in human lung cancer. Nat. Immunol. TILs, tissue-residency features are driven by Runx3. In connection, 18, 940–950 (2017). 27. Doedens, A. L. et al. Molecular programming of tumor-infiltrating CD8+ T cells it was recently demonstrated that human lung cancer TILs enriched and IL15 resistance. Cancer Immunol. Res. 4, 799–811 (2016). with certain qualities of TRM cells also correlated with TIL abundance 28. Gooden, M. J., de Bock, G. H., Leffers, N., Daemen, T. & Nijman, H. W. The and a positive prognosis26. Thus, the manipulation of transcription prognostic influence of tumour-infiltrating lymphocytes in cancer: a systematic factors that promote tissue residency may yield more effective TILs review with meta-analysis. Br. J. Cancer 105, 93–103 (2011). + 29. Singer, M. et al. A distinct gene module for dysfunction uncoupled from and anti-viral memory T cells by supplementing CD8 T cells with activation in tumor-infiltrating T cells.Cell 166, 1500–1511.e9 (2016). a gene-expression program that better supports features important 30. Tirosh, I. et al. Dissecting the multicellular ecosystem of metastatic melanoma by single-cell RNA-seq. Science 352, 189–196 (2016). to both TRM cells and TILs such as in situ survival, tissue retention, and repression of egress, ultimately fostering accumulation of protective T cells in tissues. Supplementary Information is available in the online version of the paper. Acknowledgements We thank all the members of the Goldrath and Pipkin Online Content Methods, along with any additional Extended Data display items and laboratories for their contributions. We also thank the Flow Cytometry Core Source Data, are available in the online version of the paper; references unique to at the La Jolla Institute for Allergy and Immunology. This study was funded in these sections appear only in the online paper. part by the UCSD Molecular Biology Cancer Fellowship (J.J.M.), the US National Institutes of Health U19AI109976 (S.C., M.E.P., A.W.G) and R01 AI095634 received 26 May; accepted 31 October 2017. (M.E.P.), California Institute for Regenerative Medicine RB5-07012 (W.W.), the Kimmelman Family Foundation and the San Diego Center for Precision Published online 6 December 2017. Immunotherapy (A.W.G.).

+ 1. Jiang, X. et al. Skin infection generates non-migratory memory CD8 TRM cells Author Contributions J.J.M. designed and performed experiments, analysed the providing global skin immunity. Nature 483, 227–231 (2012). data and wrote the manuscript; C.T. assisted with the RNAi screen, transfections, 2. Schenkel, J. M. et al. T cell memory. Resident memory CD8 T cells trigger transductions, tissue processing, and tumour models; B.Y. performed protective innate and adaptive immune responses. Science 346, 98–101 the computational analyses and ATAC–seq experiment; K.Z. assisted with (2014). computational analyses; K.O. and T.N. assisted with tissue processing, analysis, 3. Chang, J. T., Wherry, E. J. & Goldrath, A. W. Molecular regulation of effector and and qPCR; A.T.P assisted with tissue processing, inducible deletion experiments, memory T cell differentiation.Nat. Immunol. 15, 1104–1115 (2014). and analysis; D.W. and A.J.G. helped with the inducible deletion experiments, 4. Mueller, S. N., Gebhardt, T., Carbone, F. R. & Heath, W. R. Memory T cell RNA-seq analysis, and tumour models; S.C. provided reagents and advice; subsets, migration patterns, and tissue residence. Annu. Rev. Immunol. 31, W.W. supervised the computational analysis and contributed advice; M.E.P. and 137–161 (2013). A.W.G. supervised the project, designed experiments, and wrote the manuscript. 5. Mackay, L. K. et al. The developmental pathway for CD103+CD8+ tissue- resident memory T cells of skin. Nat. Immunol. 14, 1294–1301 (2013). Author Information Reprints and permissions information is available at 6. Mackay, L. K. et al. Hobit and Blimp1 instruct a universal transcriptional www.nature.com/reprints. The authors declare no competing financial program of tissue residency in lymphocytes. Science 352, 459–463 (2016). interests. Readers are welcome to comment on the online version of the paper. 7. Wakim, L. M. et al. The molecular signature of tissue resident memory CD8 T Publisher’s note: Springer Nature remains neutral with regard to jurisdictional cells isolated from the brain. J. Immunol. 189, 3462–3471 (2012). claims in published maps and institutional affiliations. Correspondence and 8. Masopust, D., Vezys, V., Wherry, E. J., Barber, D. L. & Ahmed, R. Cutting edge: requests for materials should be addressed to M.E.P. ([email protected]) or gut microenvironment promotes differentiation of a unique memory CD8 A.W.G. ([email protected]). T cell population. J. Immunol. 176, 2079–2083 (2006). 9. Sheridan, B. S. et al. Oral infection drives a distinct population of intestinal Reviewer Information Nature thanks F. Carbone, D. Mucida, A. Schietinger resident memory CD8+ T cells with enhanced protective function. Immunity and the other anonymous reviewer(s) for their contribution to the peer review 40, 747–757 (2014). of this work.

Methods staining was subsequently performed using the Permeabilization Buffer of the Mice. Mice were maintained in specific-pathogen-free conditions in accordance Foxp3-Transcription Factor Staining Buffer Set (eBioscience). To assess cytokine + with the Institutional Animal Care and Use Committees (IACUC) of the University production, CD8 T cells were re-stimulated with the GP33–41 peptide in the of California, San Diego (UCSD) and The Scripps Research Institute, Jupiter, presence of Protein Transport Inhibitor Cocktail (eBioscience). For flow cytometry Florida (TSRI-FL). All mice were of a C57BL6/J background and bred at UCSD analysis, all events were acquired on a BD LSRFortessa X-20 or a BD LSRFortessa. and TSRI-FL or purchased from the Jackson Laboratory, including: wild-type or Cell sorting was performed on BD FACSAria or BD FACSAria Fusion instruments. P14 mice with distinct expression of the congenic molecules CD45.1, CD45.2, RNAi screening approach. We have described this screening approach in detail Thy1.1 and Thy1.2 as well as control Thy1.1+Thy1.2+ Runx3+/+ Ert2-Cre+ YFP previously16. The targeted shRNAmir library was generated on the basis of key P14 mice and Runx3 inducible deletion Thy1.1+ Runx3fl/fl Ert2-Cre+ YFP P14 genes identified from the computational screening approach as well as genes with +/+ + fl/fl + mice. Runx3 dLck-Cre YFP and Runx3 dLck-Cre YFP mice were used known roles in regulating TRM cells from literature. The library was produced by for studying polyclonal CD8+ T cell responses. Male and female mice were used cloning shERWOOD-designed shRNAmir sequences33, after PCR of synthetic for experiments, and were age and sex matched, between 1.5 and 4 months old, 97-mer oligonucleotides, into our pLMPd-Amt vector16. Purified DNA from and randomly assigned to experimental groups. The Rosa26 stop-flox enhanced sequence-verified clones was used to package retroviral particles in PLAT-E cells. yellow fluorescent protein (eYFP) reporter mice were used for all Runx3-deletion The PLAT-E cell line was obtained from Cell Biolabs and were not authenticated or experiments. Cre-mediated deletion disrupts the Runx3 DNA-binding domain in tested for mycoplasma contamination before use. For transfections, PLAT-E cells exon 4, which exists in transcripts originating from both the distal and proximal were seeded in the middle 60 wells of a 96-well flat-bottom plate at a density of promoter. Thus, both long and short Runx3 forms are inactivated in these alleles. 4 × 104–6 × 104 cells per well one day before transfection. Next, each well was indi- Naive T cell transfers, infection and treatments. Naive P14 CD8+ T cells were vidually transfected with 0.2 μg of DNA from each pLMPd-Amt clone and 0.2 μg transferred intravenously into congenically distinct sex-matched recipient mice, of pCL-Eco using TransIT-LT1 (Mirus). Retroviral supernatant was collected 36, 48 or female P14 cells were transferred into male mice. For all microarray, RNA- and 60 h after transfection, and retroviral supernatant from each well was used to seq or ATAC–seq experiments, a total of 1 × 105 P14 cells were transferred. For individually transduce in vitro activated P14 cells in 96-well round-bottom plates. co-transfer experiments, naive Thy1.1+Thy1.2+ Runx3+/+ Ert2-Cre YFP+ P14 For CD8+ T cell activation in vitro, naive CD8+ T cells from spleen and lymph cells and naive Thy1.1+ Runx3fl/fl Ert2-Cre YFP+ P14 cells were mixed 1:1 and a nodes were negatively enriched and 2 × 105 P14 cells were plated in the middle 60 total of 3 × 104 P14 cells were transferred into Thy1.2+ recipient mice. Recipient wells of 96-well round-bottom plates pre-coated with 100 μg ml −1 goat anti-hamster mice were subsequently infected intraperitoneally with 2 × 105 plaque-forming IgG (H+L, Thermo Fisher Scientific) and 1 μg ml−1 anti-CD3 (145-2C11) and units (PFU) of the Armstrong strain of LCMV or 1010 colony-forming units (CFU) of 1 μg ml−1 anti-CD28 (37.51) (both from eBioscience). Culture medium was 9 L. monocytogenes expressing GP33–41 via oral gavage one day after cell transfer. removed 18 h after activation, and replaced with retroviral supernatant supple- For induced Runx3 deletion, recipient mice were intraperitoneally treated with mented with 50 μM BME and 8 μg ml−1 polybrene (Millipore) followed by spin- 1 mg of tamoxifen diluted in sunflower oil on days 0–4, 2–5 or 6–8 of infection. fection (60 min centrifugation at 805g, 37 °C). Two hours after the spinfection, For late deletion of Runx3 (days 16–20), recipient mice were treated with 2 mg of the P14 cells were washed three times with cold PBS and 90% of each well of cells tamoxifen via oral gavage. (individually transduced with distinct retroviral constructs) was collected, pooled 5 5 For TRM precursor experiments, 1 × 10 P14 cells were transferred, recipient and 5 × 10 pooled P14 cells were transferred into recipient mice, which were then mice were infected with LCMV the next day, and KLRG1lo or KLRG1hi P14 cells infected 1 h later with 1.5 × 105 PFU of LCMV clone 13 intraperitoneally, resulting from spleens and lymph nodes were sorted on day 5 of infection. Sorted cells in an acute infection16. The remaining cells in vitro were cultured for an additional (1 × 105) were transferred into recipient mice infected 4 days previously with 24 h and either pooled for ‘input’ sequencing (6 × 105 P14 cells) or were used to + − LCMV. The number of CD62L TCM, CD62L TEM, or IEL TRM cells were evalu- test transduction efficiency of each construct using flow cytometry to detect the ated on days 20–25 of infection using flow cytometry. percentage of ametrine+ cells in each well. To distinguish vascular-associated CD8+ T cells in non-lymphoid tissues, 3 μg Twelve days after infection, spleens and small intestines were collected from of CD8α (53-6.7) conjugated to allophycocyanin (APC) eFlour780 was injected 15–18 mice and splenocytes and IEL P14 cells were processed as described above. + intravenously into mice four minutes before euthanization and organ excision. Before sorting, all IEL or splenic samples were pooled. CD62L P14 cells (TCM) CD8αneg cells were considered to be localized within non-lymphoid tissues. from the spleen as well as P14 cells from the IEL were sorted (2 × 105–6 × 105 cells Preparation of cell suspensions. Isolation of CD8+ T cells was performed total). Genomic DNA was then collected from sorted cells using the FlexiGene similarly as described31. For isolation of CD8+ T cells from the small intestine IEL kit (Qiagen). The integrated proviral passenger strand shRNAmir sequences in compartment, Peyer’s patches were removed and the intestine was cut longitudi- each cell subset were amplified from 20–100 ng total genomic DNA per reaction, nally and subsequently cut laterally into 0.5–1 cm2 pieces that were then incubated with 23–28 cycles of PCR using Ion Proton-compatible barcoded primers that with 0.154 mg ml−1 dithioerythritol (DTE) in 10% HBSS/HEPES bicarbonate for anneal to the common 5′ mir30 and shRNAmir loop sequences. Between two 30 min at 37 °C while stirring. Kidneys, salivary glands, and lungs were cut into and three replicate reactions were performed for each genomic DNA sample and pieces and digested for 30 min with 100 U ml−1 type I collagenase (Worthington) the replicates were pooled after amplification. The pooled reactions were purified in RPMI 1640, 5% FBS, 2 mM MgCl2, 2 mM CaCl2 at 37 °C while shaking. Skin using AMPure XP beads, the amplicons in each sample were quantified using a was processed similarly as previously described32, in which a 2 cm2 area of the Bioanalyzer, and then pooled in a 1:1 molar ratio for sequencing. In each replicate right flank was excised, pre-digested for 30 min at 37 °C and then enzymatically of the screen, a minimum of 2.5 million reads per sample were generated and digested with 0.7 mg ml−1 collagenase D. After enzymatic incubations (skin, lungs, retained, after filtering low-quality reads. Reads assigned to each barcode were kidneys and salivary glands), tissues were further dissociated over a 70-μm nylon aligned to a reference database of all shRNAmirs in the library using BLAST and a cell strainer (Falcon). For isolation of lymphocytes, single-cell suspensions were custom script to count the top alignment of each read and summarize the number then separated using a 44/67% Percoll density gradient. Spleens and lymph nodes of reads aligned to each shRNAmir. were processed with the frosted ends of microscope slides. Red blood cells were For analysis of shRNAmir representation in TCM cells relative to IEL TRM, the lysed with ACK buffer (140 mM NH4Cl and 17 mM Tris-base, pH 7.4). total number of reads in each of the samples was normalized, and the number of Antibodies, intracellular staining, flow cytometry and cell sorting. The reads for each shRNAmir was scaled proportionally. Subsequently, the normalized following antibodies were obtained from eBioscience: CD8α (53-6.7), CD8β number of reads in the IEL TRM cells for a given shRNAmir was divided by the (eBio H35-17.2), CD62L (MEL-14), CD127 (A7R34), KLRG1 (2F1), CD103 normalized number of reads for the same shRNAmir in the TCM sample and then (2E7), CD69 (H1.2F3), CD45.1 (A20-1.7), CD45.2 (104), Thy1.1 (OX-7, HIS51), log2 transformed. The mean and s.d. of the ratios of each of the 25 negative-control Thy1.2 (53-2.1), CCR9 (Ebio CW-1.2), CXCR3 (CXCR3-173), CD49d (R1-2), TNF shRNAmir constructs (targeting Cd19, Cd4, Cd14, Ms4a1, Cd22, Hes1, Klf12, (MP6-XT22), GzB (GB11), PD-1 (J43), Tim3 (RMT3-23), Lag3 (eBioC9B7N), Mafb, Plagl1, Pou2af1 and Smarca1) were used to calculate the Z-score for each KI-67 (SolA15), and IFNγ (XMG1.2) or from BioLegend: CD62L (MEL-14), shRNAmir construct. The screen was repeated three times and the Z-score of CD103 (2E7), CD69 (H1.2F3), CD45.1 (A20-1.7), Thy1.1 (OX-7), Thy1.2 (30-H12), each construct from each individual screen was averaged and plotted (Fig. 1g and T-bet (4B10). For analysis of apoptosis, the Annexin V Apoptosis Detection Supplementary Table 2). Certain constructs were added after the first screen or Kit was used per manufacturer instructions (eBioscience); propidium-iodide- were not detectable in one of the screens, but all constructs were successfully b negative cells were analysed for annexin V staining. The H-2D GP33–41 tetramer screened 2–3 times except for 13 constructs, which are marked by an asterisk in was obtained from the NIH Tetramer Core. For intracellular staining of cytokines Supplementary Table 2. Eighty-four per cent (21 out of 25) of all negative control or transcription factors while preserving ametrine or YFP reporter expression in shRNAmir constructs had an average Z-score between −0.9 and 0.9. transduced or Cre-YFP+ populations, cells were fixed and permeabilized through CD8+ T cell transduction, cell transfer and infection for individual analysis of a 10 min incubation with BD cytofix/cytoperm (BD Biosciences). Intracellular retroviral constructs. Activation, transfections and transductions were carried out

© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved. Letter RESEARCH as described for the RNAi screening approach except in some experiments 2 × 106 as well as naive P14 cells; spleens or IEL samples from 2–3 mice were pooled P14 cells were activated per well in 6-well plates. Congenically distinct P14 cells and 5 × 103 cells were sorted. For RNA-seq analysis of TIL, congenically distinct transduced with the Runx3.2 shRNAmir or Cd19.1 shRNAmir (control) retrovi- P14 cells were transduced with Runx3-RV or GFP-RV, mixed 1:1 and 1 × 106 cells 5 5 ruses were mixed 1:1 within 24 h of transduction and a total of 1 × 10 –5 × 10 were transferred to mice with day 7 established melanoma B16-GP33 tumours. P14 cells were transferred intravenously into recipient mice. One hour after adop- Eight days later, 1 × 103 transduced TILs or splenocytes were sorted from four tive transfer, recipient mice were infected intraperitoneally or intratracheally with mice for each replicate. For library preparation, isolation of polyA+ RNA was per- 2 × 105 PFU LCMV Armstrong or intradermally with 2 × 104 PFU clone 13. In formed as detailed online (http://www.immgen.com/Protocols/11cells.pdf). For similar experiments, P14 cells were transduced with MigR1-based retroviruses34 RNA-seq analyses of Runx3-manipulated cells, CD8+ T cells from naive Runx3+/+ that were empty (GFP-RV) or that contained Runx3 cDNA (Runx3-RV), mixed YFP+ (wild type) and Runx3fl/fl YFP+ (Runx3fl/fl) mice were enriched by negative 1:1 and transferred to recipient mice for subsequent infections. For T-bet rescue isolation and transduced (as detailed above) with a Cre cDNA-expressing retro- experiments, Thy1.2+ Runx3+/+ Ert2-Cre YFP+ P14 cells were transduced with virus (Cre-RV). Runx3-overexpressing cells were generated similarly by trans- Cd19.1 shRNAmir and Thy1.1+ Runx3fl/fl Ert2-Cre YFP+ P14 cells were transduced ducing Runx3+/+ YFP+ CD8 T cells with a Runx3 cDNA-expressing retrovirus with Tbx21.3 shRNAmir-encoding retroviruses, mixed 1:1 and transferred into (Runx3-RV). Forty-eight hours after T cell receptor activation, the CD8+ T cells recipient mice, which were infected 1 h later with LCMV Armstrong intraperito- were resuspended and re-cultured in fresh medium supplemented with 100 U ml−1 neally and treated with 1 mg tamoxifen intraperitoneally for five consecutive days rhIL-2; 24 h later, YFP+ (wild type or Runx3fl/fl) or GFP+ (Runx3-RV) were FACS- starting with the day of infection. purified and then recultured in 100 U ml−1 IL-2. The cells were expanded until Adoptive therapy tumour model. For adoptive therapy experiments, 5 × 105 day 6 by reculturing at 5 × 105 cells per millilitre every 24 h in fresh 100 U ml−1 B16-GP33 cells, which were treated for mycoplasma and authenticated in in vitro IL-2 medium. On day 6 after activation, cells were collected and total RNA was killing assays, were transplanted subcutaneously into the right flank of wild-type extracted in TRIzol. Purified RNA was depleted of ribosomal RNA and strand- mice. After tumours became palpable, 7–8 days after transplantation, in vitro specific paired-end libraries were prepared and sequenced using an Illumina expanded P14 cells were transferred intravenously. For comparison of TIL accu- Nextseq 500. Samples were generated from two biological replicates, and approxi mulation in a mixed transfer setting, naive P14 cells were activated, transduced mately 20 million paired-reads were generated per sample. Reads were mapped and expanded with 100 U ml−1 of IL-2 for 2–3 days; cells transduced with control using Tophat36 and aligned reads in transcripts were counted with HTseq37. GSEA constructs (Cd19.1 shRNAmir or GFP-RV) or experimental constructs (Runx3.2 was performed by using the GSEA module in GenePattern, and the normalized shRNAmir or Runx3-RV) were mixed 1:1 and 0.5 × 106–1 × 106 P14 cells were enrichment scores and false-discovery rate q values were determined by using the transferred intravenously. For efficacy studies, transduced cells were expanded for permutation test. 5–6 days; transduced cells were then sorted (or not sorted with a Runx3-RV and ATAC–seq was performed as described in detail previously24. Sorted cells GFP-RV transduction efficiency >83%), and 1 × 106–2.5 × 106 cells were trans- (2.5 × 104) were resuspended in 25 μl of lysis buffer and spun down 600g for ferred intravenously into mice with established B16-GP33 tumours. Tumours were 30 min at 4 °C. The nuclear pellet was resuspended in 25 μl of Tn5 transposase monitored daily and mice with ulcerated tumours or tumours exceeding 1,500 reaction mixture (Nextera DNA Sample Prep Kit, Illumina) and incubated mm3 were euthanized, in accordance with UCSD IACUC. for 30 min at 37 °C. Transposase-associated DNA was subsequently purified Quantitative PCR, microarray, RNA-seq and ATAC–seq analysis. For validation (Zymo DNA clean-up kit). For library amplification, DNA was amplified using of the Runx3-RV overexpression construct and Runx3.2 shRNAmir construct, indexing primer from Nextera kit and NEBNext High-Fidelity 2× PCR master enriched CD8+ T cells were activated, transduced, and expanded for 4–6 days in mix. Then, the amplified DNA was size-selected to fragments less than 800 bp 100 U ml−1 IL-2. Cells were sorted on ametrine (Runx3 shRNAmir or control shR- using SPRI beads. The library was sequenced using Hiseq 2500 for single- NAmir) or GFP (Runx3-RV or GFP-RV) directly into TRIzol (Life Technologies), end 50-bp sequencing to yield at least 10 million reads. We used bowtie to map and RNA was extracted per manufacturer’s specifications. Next, cDNA was syn- raw reads to the Mus musculus genome (mm10) with following parameters: thesized using Superscript II (Life Technologies) and quantitative PCR (qPCR) ‘–best -m 1’. We called peaks for each individual replicate as well as the was performed using the Stratagene Brilliant II Syber Green master mix (Agilent pooled data from the two replicates using MACS2 with a relaxed threshold Technologies). Runx3 expression levels were normalized to the housekeeping gene (P = 0.01). Hprt. We have previously validated the Tbx21.3 shRNAmir16. The following primers For the single-cell RNA-seq analysis of human30 and mouse melanoma TILs29, were used for qPCR: Runx3 forward, 5′-CAGGTTCAACGACCTTCGATT-3′, the pre-processed single-cell TIL gene expression data were downloaded from GEO Runx3 reverse, 5′-GTGGTAGGTAGCCACTTGGG-3′; Hprt forward, database with accessions GSE72056 or GSE86042, respectively. Activated CD8+ 5′-GGCCAGACTTTGTTGGATTT-3′, Hprt reverse, 5′-CAACTTGCGCT TILs (CD8α expression > 5 and CD44 expression >2) were used and classified CATCTTAGG-3′. into Runx3hi TILs, which express relatively high levels of Runx3 (Runx3 expres- On day 7 of infection, tissues from 2–3 mice were pooled and 2 × 104–3 × 104 sion > 3) and Runx3lo TILs with no Runx3 expression (Runx3 expression = 0). P14 cells from the IELs, kidney, spleen or blood were sorted into TRIzol. On day For the human TILs, melanoma #75 was used. GSEA was performed to evaluate 35 of infection, tissues from 5–10 mice were pooled and 1 × 104–2 × 104 CD62L+ enrichment of the core tissue-residency gene expression signature in Runx3hi TILs − lo TCM, CD62L TEM, kidney TRM, and IEL TRM P14 cells were sorted into TRIzol. As relative to Runx3 TILs. described previously, RNA was amplified and labelled with biotin and hybridized Computational screen: transcription factor regulatory networks and per- to Affymetrix Mouse Gene ST 1.0 micrroarrays35. Analyses were performed using sonalized PageRank analysis. Transcription factor regulatory networks and 24 GenePattern Multiplot Studio. Differentially expressed genes in IEL TRM compared PageRank analysis were performed as described previously except that gene to TCM and TEM cells as well as kidney TRM compared to TCM and TEM cells were expression and ATAC–seq data from day 7 IEL, kidney and spleen samples were identified with a fold change >1.5 and an expression value >120 (Fig. 1a). Genes used. To construct the transcription factor regulatory network, transcription with >1.5 fold change and >120 expression value between day 7 spleen, day 7 factor-binding motifs were first scanned on ATAC–seq peaks using an algorithm IELs, and day 7 kidney samples were identified (1,838 probes) and evaluated in described previously 14 and a P-value cut-off of 1 × 10−5. Then, we connected day 7 and day 35 subsets, which were ordered with Pearson correlation using the a transcription factor to a gene if the factor had any predicted binding motif HierarchicalClustering module of GenePattern (Fig. 1b); data were row centred, in the ATAC–seq peak of the nearest gene. We assembled all the interactions row normalized, and visualized with the HierarchicalClusteringViewer module between transcription factors and genes into a regulatory network. To iden- in GenePattern. tify important transcription factor regulators for TRM cell differentiation, we The core tissue-residency cells and circulating signatures were generated by performed personalized PageRank analysis in the transcription factor regula- 14 integrating differential expression (>1.5 fold change) data comparing TRM cells tory network constructed above using the pipeline described previously . The from the following tissues to circulating splenic memory cells (or splenic TCM if importance of a transcription factor is based on the quantity and quality of its both TCM and TEM datasets were available): day 35 IELs (LCMV), day 35 kidney regulated gene targets. A factor would receive a higher PageRank score if it regu parenchyma (LCMV), day 30 skin CD103+CD8+ (herpes simplex virus)5, day 30 lates more important genes, where the importance is evaluated by differential lung CD103+CD8+ (influenza virus)5, and day 20 CD103+ brain (vesicular sto- expression from microarray or RNA-seq analyses. Extended Data Fig. 2b and 7 matitis virus) ; overlapping genes upregulated in all TRM cell populations com- Supplementary Table 1 indicate the PageRank score and expression value of all prised the core tissue-residency signature (121 genes) and genes downregulated transcription factors expressed (>120 expression value) in the spleen, kidney or in all populations comprised the circulating signature (93 genes). The mouse TIL IEL cells. microarray datasets were generated previously27. Statistical analysis. Student’s t-test (two-tailed) was used for comparisons between For RNA-seq analysis of day 7 IEL, day 7 memory precursors, and day 7 terminal two groups. A log-rank (Mantel–Cox) test was used to compare survival curves. All effectors, the populations were sorted on day 7 of LCMV Armstrong infection microarray, RNA-seq, and ATAC–seq samples were performed independently in

2–3 replicates. All statistical tests were performed with GraphPad Prism software, 32. Benck, C. J., Martinov, T., Fife, B. T. & Chatterjea, D. Isolation of infiltrating and P < 0.05 was considered statistically significant. No statistical methods were leukocytes from mouse skin using enzymatic digest and gradient separation. J. Vis. Exp. 107, e53638 (2016). used to predetermine sample size. Investigators were not blinded to allocation 33. Knott, S. R. V. et al. A computational algorithm to predict shRNA potency. during experiments and outcome assessment. Mol. Cell 56, 796–807 (2014). Data availability. RNA-seq, microarray, and ATAC–seq data are available in 34. Pear, W. S. et al. Efficient and rapid induction of a chronic myelogenous the Gene Expression Omnibus (GEO) database under the SuperSeries reference leukemia-like myeloproliferative disease in mice receiving P210 bcr/ code GSE107395. Source Data are provided in the online version of the manuscript. abl-transduced bone marrow. Blood 92, 3780–3792 (1998). 35. Best, J. A. Transcriptional insights into the CD8+ T cell response to All other data are available from the corresponding author(s) upon reasonable et al. infection and memory T cell formation. Nat. Immunol. 14, 404–412 (2013). request. 36. Trapnell, C., Pachter, L. & Salzberg, S. L. TopHat: discovering splice junctions with RNA-Seq. Bioinformatics 25, 1105–1111 (2009). 31. Steinert, E. M. et al. Quantifying memory CD8 T cells reveals regionalization of 37. Anders, S., Pyl, P. T. & Huber, W. HTSeq—a Python framework to work with immunosurveillance. Cell 161, 737–749 (2015). high-throughput sequencing data. Bioinformatics 31, 166–169 (2015).

a D7 IEL c KLRG1hi b KLRG1lo 77 0 75 0 KLRG1hi D5 LCMV Spleen LCMV 99 30 Tem 0 0 P14 cells D4 LCMV CD49d CXCR3 CCR9 Recipient Tcm 51 2 49 0 Trm KLRG1 1 mLN CD44 100 lo *** *** * KLRG1 3 0 KLRG1hi 80 KLRG1lo y 39 6 41 0 60 lo Kidney KLRG1 ) 15x hi 40 10 KLRG1 6 4 14 0 Frequenc 29x ** 20 12x 6x 5 ** 30 2 27 0 27x * ** 0 3 SG CD49dhi CXCR3hi CCR9hi 4 * 13 3 3 2 17 4 16 0 2 ansferrred Cells (Log β IEL 1 Tr 1 Tcm TemTcm Tem Trm 33 1199 # CD8 CD8α i.v. KLRG1 CD69 CD103 Spleen mLN IEL Extended Data Figure 1 | KLRG1lo cells preferentially give rise to infection. c, Top, schematic of experimental design. KLRG1lo and KLRG1hi TRM cells. a, Left, representative flow cytometric gating strategy for P14 cells were sorted from spleens and lymph nodes on day 5 of LCMV distinguishing P14 cells located in non-lymphoid tissues after intravenous infection and transferred into recipient mice infected 4 days previously administration of CD8α in LCMV-infected mice. Right, in vitro activated with LCMV. Bottom, TCM, TEM and TRM P14 cells were enumerated on P14 cells were transferred to recipient mice and infected with LCMV; the days 20 or 25 of infection using flow cytometry. Data are mean ± s.e.m frequency of CD69+ and CD103+ P14 cells among KLRG1hi and KLRG1lo of n = 5 mice (a, b) or n = 3–4 mice (c) from one representative of 2 on day 7 of infection is indicated. b, Frequency of CCR9, CXCR3 and independent experiments. *P < 0.05, ** P < 0.01, ** * P < 0.005 (Student’s CD49d in KLRG1lo and KLRG1hi cells in the IEL compartment on day 7 of t-test). Symbols represent an individual mouse (c).

a b 3 Nfatc1

) 5 D7 Spleen 3 2 Stat1 Nfkb2 Runx3 4 (x10 1 Nr4a2 D7 IEL Stat5b 0 S1pr1 Prdm1 3 Nr4a1 ) 2 Stat4

) Nfatc2 Nr4a3 2 4 2 Jun D7 Spleen Irf4 ) 2 (x10 1 D7 IEL Fosb 0 Klf2 -2 -1 1 234 56 Gata3 -1 ) D7 Spleen 3 4 Smad3 D7 IEL/D7Spleen

Expression (AU 2 Nr3c1 -2 PageRank score (Log D7 IEL (x10 Stat6 Bach2 Eomes 0 -3 Cd69 Tbx21 Stat3 -4 )

D7 Spleen 3 6 Arid3b 3 Expression (Log )

D7 IEL (x10 2 D7 IEL/D7 Spleen 0 Nr4a1 L Putative regulator, differentially expressed Tcm D7 IE Putative regulator, independent of expression Spleen D35 IEL D35 D7

Extended Data Figure 2 | Representative ATAC–seq peaks and putative TRM cell regulators identified through PageRank analysis. a, ATAC–seq analysis of the indicated loci on day 7 of infection (left) and corresponding gene expression (right). b, Personalized PageRank score and gene expression of transcription factors, with selected factors highlighted.

a bcCon shRNA CD45.1 P14 Input LCMV 5000 1.00 Runx3 shRNA Runx3 shRNA 4000 48 D12 IEL Trm 0.75 CD45.2 P14 52 3000 Con shRNA CD45.1 CD45.2 Spleen 0.50 Day 12 mLN

+ ) 2 2000 CD69 2 IEL expression (AU) + g 0.25 Spleen mLN IEL CD103 IEL 1 CD69+ +

1000 (Lo RNA/ CD103 IEL unx3 h 0 R

0 Relative to control shRNA 0

45 69 33 18 x3 -1

Fold Change Runx3 Expression shRNA *

n Tcm Tem

Naive CD45.1 55 31 67 82

D7 IEL Ru SpleenD7 PBL D35 IEL -2 *** D7 Kidney CD45.2 Con D35 Kidney -3 ***

d e LCMV LM-GP +/+ Day 12 Con shRNA 33-41 Runx3 Runx3 shRNA Runx3fl/fl Spleen IEL * *** Spleen IEL Spleen IEL 100 *** *** *** 0011 80 5088811 403 92 100 80 +/+ Con y Runx3 80 shRNA y 0 0 60 0 2 1 0

uenc 60 0041 40 40 11 0 54 1111 3067 0 40 Runx3 Freq fl/fl Runx3 Frequenc shRNA 20 20 0 0 1 0 CD69 0 0 CD69 0 0 CD103 CD69+ CD103+ CD103 LCMV LM-GP33 LCMV LM-GP33 IEL CD69+ CD103+

f Vehicle Tamoxifen Runx3+/+ Runx3fl/fl Spleen IEL Spleen IEL 9 0 3 92 11 0 7 86 100 * ** Runx3+/+ 80 0 3 0 6 60 8 0 3 9900 4 0 32 46 40 fl/fl Runx3 Frequency 20

CD69 0 6 0 4 0 CD103 VehTam VehTam CD69+ CD103+ + + Extended Data Figure 3 | Runx3-deficiency impairs IEL TRM cell quantification of the frequency of CD69 and CD103 cells of control formation. a, Runx3 mRNA levels from indicated cells determined by shRNAmir or Runx3 shRNAmir cells (right) from experimental schematic microarray analyses. b, Relative Runx3 mRNA expression of in vitro in c. e, Representative flow cytometry plots and quantification of the cultured cells transduced with retroviruses encoding control shRNAmir or frequency of CD69+ and CD103+ cells from Fig. 2c. f, Representative Runx3 shRNAmir, measured by qPCR. c, Congenically distinct P14 cells flow cytometry plots and quantification of the frequency of CD69+ and were transduced with retroviruses encoding Runx3 shRNAmir or control CD103+ cells from Fig. 2e. Data are mean ± s.e.m and representative of shRNAmir, mixed at a 1:1 ratio, and transferred to recipient mice that were two independent experiments (b) with n = 5 (c, d), n = 5 (LM–GP33–41) or subsequently infected with LCMV. Representative flow cytometry plots n = 6 (LCMV) (e), and n = 5 (vehicle) or n = 3 (tamoxifen) (f). *P < 0.05, (bottom, left) and quantification of the ratio of P14 cells transduced with ** P < 0.01 ** * P < 0.005 (Student’s t-test). Symbols represent an individual Runx3 shRNAmir or control shRNAmir in indicated tissues on day 12 of mouse (c–f). infection (bottom, right). d, Representative flow cytometry plots (left) and

abCD69+ cdRunx3+/+ + fl/fl Spleen IEL CD103 IEL ) Runx3

1.9x 4.6x 8.5x 10 2.4x 29x 158x Spleen IEL

3.4 3.0 1.9 + ** *** 5 *** ** *** 6 +/+ 1 0 52 40 80 Runx3 100 + +/+ (Log fl/fl +/+ Runx3 + 5 Runx3 Runx3

4 + +

tramer 75 60 4 0 0 Te

tramer 3

tramer 3 50 40 Te 2.0 0.9 0.1

3 0 65 5 33-41

2 Te 2 fl/fl % CD69 fl/fl % CD103

33-41 Runx3 1 Runx3 25 20 33-41 1 % GP GP

CD69 0 0 0 0 0 0 + + CD103 Cre-YFP Spleen IEL CD69 # GP Spleen IEL CD69 IEL IEL CD103+ IEL CD103+ IEL + Extended Data Figure 4 | Runx3-deficiency impairs IEL TRM cell flow cytometry plots and quantification of the frequency of CD69 and formation in a polyclonal setting. a, Representative flow cytometry plot CD103+ cells. Data are mean ± s.e.m with n = 4 (Runx3+/+) or n = 5 b fl/fl fl/fl of H-2D GP33–41 tetramer staining of lymphocytes from Runx3 (Runx3 ) mice pooled from two independent experiments. ** P < 0.01, + + + + dLck-Cre YFP and Runx3 / dLck-Cre YFP mice on day 12 of LCMV ** * P < 0.005 (Student’s t-test). Symbols represent an individual infection (gated on total lymphocytes). b, Quantification of the proportion mouse (b, d). (left) and absolute number (right) of tetramer+ cells. c, d, Representative

abCD69+ Day 16 Spleen ) Thy1.1 P14 Spleen mLN Kidney SG IEL CD103+ IEL 2 0 mLN fl/fl + LCMV Tamoxifen Kidney

Runx3 Ert2-Cre YFP (log -2 SG +/+ Trm IEL 51 18 36 9 5 40.6 -4 *** CD69+ Thy1.1.2 P14 49 Thy1.2 D2 - D5 82 64 91 95 96 99.4 Thy1.1 +/+ + Thy1.1 -6 *** CD103+ IEL

Runx3 Ert2-Cre YFP /Runx3 Thy1.2 Thy1.2 fl/fl -8 *** *** -10 Runx3 ***

cdCon shRNA e Con shRNA Con Runx3 Runx3 shRNA Runx3 shRNA CD45.1.2 P14 shRNA shRNA Con shRNA + i.d. LCMV 1.2 * 10 * 30 *

D12 Skin Trm + 8 Spleen 0.9 + 81 55 20 CD45.1 P14 CD45.2 6 Runx3 shRNA ansduced Cells 0.6 Tr 4 %CD69

%CD103 10 β Skin 0.3 2 77 27

CD8 0 0 0 Ametrine Spleen Skin Skin Skin Frequency

f g Con Runx3 h shRNA shRNA CD45.1.2 P14 + i.t. LCMV Con shRNA Con shRNA Spleen Con shRNA Runx3 shRNA Runx3 shRNA D12-13 Lung Trm 63 37 CD45.1 P14 CD45.2 2.5 * ** 50 Runx3 shRNA * 80

2.0 + 40 Lung + 60 57 30 1.5 30 ansduced Cells 40 Tr 1.0

+ 20 %CD69

CD69 %CD103

β CD103+ 0.5 10 20 49 19 Lung CD8 0 0 0 Ametrine Spleen Lung CD69+ Lung

Frequency Lung CD103+ Lung + + Extended Data Figure 5 | Runx3 is required for TRM cell formation in e, Frequency of CD69 and CD103 cells. f, Schematic for experimental diverse non-lymphoid tissues. a, Schematic of experimental design. design. g, Representative flow cytometry plots (left) and quantification b, Representative flow cytometry plots (left) and quantification (right) (right) of the ratio of transduced cells in the lung parenchyma relative to of the ratio of Runx3fl/fl to Runx3+/+ P14 cells (gated on YFP-Cre+ cells) the spleen for control shRNAmir or Runx3 shRNAmir P14 cells on in lymphoid and non-lymphoid compartments on days 15/16 of LCMV day 12 of an intratracheal LCMV infection. h, Frequency of CD69+ infection (as in Fig. 2d but including salivary gland (SG) and kidney and CD103+ cells. Data are mean ± s.e.m and representative of two populations). c, Schematic for experimental design. d, Representative flow independent experiments with n = 6 (b), or data pooled from two cytometry plots (left) and quantification (right) of the ratio of transduced individual experiments with n = 6 per group (c–h). *P < 0.05, ** P < 0.01, cells in the skin relative to the spleen for control shRNAmir or Runx3 ** * P < 0.005 (Student’s t-test). Symbols represent an individual mouse shRNAmir P14 cells on day 12 of an intradermal LCMV infection. (b, d, e, g, h).

a b Spleen IEL Day 12 Con shRNA Con shRNA Con shRNA Runx3 shRNA Runx3 shRNA 123 333 Runx3 shRNA 80 * Con 100 shRNA * 60 + 0.4% + 75 Spleen B 47 40 0.1% 40 0 11 0 28 50

% Gz Runx3 ** ** 52% 20 α IEL shRNA % Cytokine 25 26% 55 47 0 TNF 0 GzB Spleen IEL IFNγ α γ α α γ α γ γ IFN TNF IFN TNF TNF IFN TNF IFN Spleen IEL cd *** +/+ 50 Con shRNA 80 *** Runx3 12% + Runx3fl/fl V Runx3 shRNA 11% *** +

Spleen 40 Spleen V 14% 11% 60 *** 30 ** 22% 17% n.s. IEL IEL 40 35% Annexin 20 29% Annexin 24% % 20 10 20% % CD69+CD103+ IEL CD69+CD103+ IEL 44% 36% 0 0 Annexin V Spleen IEL CD69+ Annexin V Spleen IEL CD69+ CD103+ IEL CD103+ IEL

e f Con shRNA Input Day 6 Day 14 CD45.1 P14 Runx3 shRNA Runx3 shRNA LCMV D5 Host 100 48 D6 Spleen 18h + 75 CD45.2 P14 52 CD45.1 Con shRNA CD45.2 Spleen 50 n.s. % KI-67 ) 2 mLN 10

/ Spleen mLN IEL 1 IEL 5

(Log 0 0 Spleen IEL Spleen IEL

NA 18h shR

R 52 45 50 3 -1 48 55 50 CD45.1

Runx -2 CD45.2

Con sh -3

Extended Data Figure 6 | Runx3-deficiency enhances TRM cell On day 6 of infection, splenocytes were collected and retransferred to apoptosis but does affect trafficking or proliferation. a, Representative day 5 infected host mice and 18 h later spleen, mesenteric lymph nodes flow cytometry histogram of granzyme B (GzB) staining (left) and and small intestine were obtained to assess trafficking. Representative quantification of frequency of GzB+ cells on day 12 or 14 of infection. flow cytometry plots (bottom, left) and quantification of the ratio b, Representative flow cytometry plots (left) and quantification (right) of of P14 cells transduced with control shRNAmir to Runx3 shRNAmir the frequency of IFNγ- and TNF-producing control shRNAmir or Runx3 (bottom, right) in indicated tissues 18 h after transfer. f, Frequency of shRNAmir P14 cells on day 6 of LCMV infection, restimulated with Ki-67+ control shRNAmir or Runx3 shRNAmir transduced P14 cells in GP33–41 peptide. c, d, Representative histograms and quantification of a mixed transfer setting on days 6 and 12 or 14 of LCMV infection. Data annexin V+ cells from shRNAmir mixed transfers on day 14 of LCMV are mean ± s.e.m and representative of two independent experiments with infection (c) or from day 8 Runx3fl/fl and Runx3+/+ mixed P14 transfers in n = 5 (a), n = 3 (b), n = 5 (c), n = 6 (d), n = 4 (e), and n = 3 on day 6 or which tamoxifen was administered on days 2–5 of LCMV infection (d). n = 4 on day 14 (f) except d is pooled from two independent experiments. e, Congenically distinct P14 cells were transduced with control shRNAmir *P < 0.05, ** P < 0.01, ** * P < 0.005 (Student’s t-test). n.s., not significant. or Runx3 shRNAmir encoding retroviruses, mixed at a 1:1 ratio, and Symbols represent an individual mouse (a–f). transferred to recipient mice that were subsequently infected with LCMV.

GFP-RV medLN a b c Lung d Runx3-RV CD45.1 P14 GFP-RV medLN CD69+CD103+ Lung 8 Runx3-RV + i.t. LCMV 40 60 Runx3-RV Runx3 Lung 4 6.8x D12 Lung Trm V *** CD45.1.2 P14 17 5 Expression V Control 6 CD45.2 * 30 25 Lung 3 GFP-RV GFP-RV + 20 52 48 + 4 2.6x 3 20 V/GFP-R Runx3 2 15 + * 11 12 CD69 10 + % CD69 2 CD103 1 Runx3-RV % CD103 10 78 22 5 Lung Runx3-R 16 CD45.1 0 0 CD69 Relative to GFP-R 0 0 CD45.2 CD103 Lung Lung Fold Change + + Extended Data Figure 7 | Runx3 overexpression enhances lung TRM CD69 CD103 lung parenchyma population on day 12 or 13. differentiation. a, Runx3 mRNA expression of in vitro cultured cells d, Representative flow cytometry plots (left) and quantification (right) of transduced with GFP-RV or Runx3-RV. b, Schematic for experimental the frequency of CD69+ and CD103+ P14 cells in the lung parenchyma. design of intratracheal LCMV infection. c, Representative flow cytometry Data are mean ± s.e.m and representative of one of two independent plots (left) and quantification (right) of the ratio of GFP-RV or Runx3-RV experiments (a) and n = 4 per group (c,d). *P < 0.05, ** * P < 0.005 cells in the mediastinal lymph nodes (medLN), lung parenchyma, or (Student’s t-test). Symbols represent an individual mouse (c, d).

b a IEL Spleen Spleen Runx3 IEL Trm Runx3 Binding Network Cell-cell adhesion Binding Network Atp10a Kcnj16 Mrpl15 Zfp69 Gm1647 Fam168b Tr ib1 Rpf2 Ptpn12 Fbxl3 Pdpk1 % Runx3 Targets Srgap2 St8sia1 F730043M19Rik Gm43223 Dtnb Lfng Gm4890 Gm16754 Suox Gm42545 Gm10287 Te sc Unk Mpp5 Cass4 F2r Clasp1 LbrErlin2 Gm42912 Rpl29 Gm10805 Rwdd1 Nr3c1 Wdr37 RP23-479K9.6 Srd5a3 Slc39a14 Gm12915 Tbl1xr1 Gm24169 Fignl1 Gm2822 Kctd3 Gna11 Ip6k2 Zcchc18 Cdc14a Pigk Tmem154 Tnfsf8 Cdkal1 RP24-266E2.1 Card6 Arhgap15 Eif2ak3 Ggta1 2310009B15Rik A930006K02Rik Gpr137c Haus2 Sel1l Ssb Blm Etohd2 Transcription factor activity Cmpk2 Ifng BC003331 Ctnnbip1 Atl2 Armc10Uqcrfs1Bsg Tbc1d4 Edrf1 Raf1 Ts pan13 Ywhag 2500002B13Rik Nd fip2 9530068E07Rik Gm25735Gm23101 Cd2 Rsbn1l Vps54 Tr bv17 Gas2l3 Ppcdc Birc5 Vgll4 Sgk1 0610040F04Rik Ppp1r16b Rtkn2 Zfp292 Gm44428 Gm44065 Smim20 Cnrip1 Gm10640 Gm44663 Abhd1 Cenpn Dnmbp Mir1901 Depdc1b Abhd15 A430035B10Rik Txlnb Bmp2k Nrg4 Cptp Atp10d Tmem71 RP23-413D14.7 Atp6v0b Arl15 Pmepa1 Zfp507 Gm20545 Pdlim5 Sox5 Sipa1l1 Pik3r1 Gm10392 Hcn1Gm44663 Gm32381 Rap1b Rb1 Pgam5 1700066M21RikDuox2 RP23-348E18.5 Hif1an Ksr1 Rasa4 Cenpn Rnf19a Tr am1 Phf14 Gm9889 Ear10 Vps37b Znhit6 Gm4208 Parvg Nfkb1 Ksr1 015 30 45 60 75 Gm24523 Rrad Cers4 Ston2 Irf8 1810019N24Rik Tnfrsf1b Tbc1d23 Herpud2 Ldlrap1 Gm15897 Cyhr1 Dzip3 Inhba2900060B14RikFos Plek 4930465M20RiTprak 1 Fndc3a Gm15856 Homer1 Gm42759 Gm24754 Lrrc27 Mmadhc Rgs1 Gm12890 Galc Wwox Gtf2b Hsd17b11Cox5a CenpWdr7f 5 Tr p53inp1 VimKlhdc4 Csgalnact2 Vim RP23-403G9.6 Gltp Mir6980Map3k8Eif2akTjp13 Fcgr3 Usp32 Tr bv17 Gm22710Tr ip13 Cass4 Samd13RuPde7andc3b Cyld Ta nc2 RP24-532A16.2 Carnmt1 Klhl2 Ccnd3 Rapgef6 Pcm1 Senp8 Phf20 Ercc4 A330023F24RikMob3b Kdm6a Upf3b Prkar1a Rasa2 Depdc1b Bloc1s2 RP24-83E15.3 Smad7 Daglb Atf6b Gm13006 Tw f2 Atg7 Chaf1aRgcc Sh3gl1 Mir652 Tr im2 Park7 Coro7 Tbc1d23 Nfatc2Gm22574 S100a3 Rnf169 1700021A07RikNdufa12 Styk1 Egr3 mt-Nd6 Inpp4b Gm33882 Atp8a2 Ta rs2 Tmem123Ikbip Hook1 Ralgps2 Cd44 Rabgap1 Gm9484 Gm37696 Lrrc8c Smpdl3a Cntrob Gzmb Mx1 Abcc4 Gm26656 Tm9sf3 Cops8 Tgfbrap1 HdgfrpCbl3 Gm585Gm370045 Mef2a Gm12364 Bcas2 Gm29395 AtGm25846p2b4 RP23-451G9.2Gm15232 Stxbp3 Med13 Eif3h 3300005D01Rik Kif2a Gm37760Setbp1Gm28706Ptger4P2rx3Iqgap2 Phgdh Tmem268 A730015C16Rik Lrmp Fam187b Pbx4 Mtus1 Dph2 Stx7Slc12a7 Khnyn Gm1647 Ado Tr im17 Fosl2 Gas2l3 1700022E09Rik Il6st Mctp2 Stt3b 4930402H24Rik Rap1aAnkmy2 Asap1 Mir6359Mex3c Cgref1 Il2ra Coro7 Atp5o Tr bv3 Nfkbiz Fam192a Immp2l Te cpr1 Athl1 Gm12967 Cpsf3l Gm24041 Mrvi1 P2ry12 Atp2b1 5430427O19Rik Panx1 Mrvi1 Ikbkb Gtf2a2CersTmem594 Iqsec1 Khnyn Tm4sf5 Nd fip1 Gm3500Tr ac 0 Odf2l Tr im33 Cbx2 Gm43051 Gm25534 Sertad2 Tmem261 Kbtbd11 Chaf1b Tbc1d30 1700082M22Rik Skil Uap1 Gm10617 Zfp637 Edem3 Pml Gm139775033430I15RikSmarca5 Pcca Camta1 Dyrk2 Ddx43 Gm22768 Rbpj Pxylp1Gm13481 Cmas Ly6c2 Gm29509 Ankrd28 Fam173b Idh1 Gm37930 Sipa1l1Gm44310 Oxr1 Pkp4 Gm26756 5031414D18Rik Arsk Siae Brd9 Atp2b1 Pus7 Mir92-1Cast Gm44592RP24-130A20.Nup107 1 Gm45041 Spata24 Elovl5 Ccr9 Gm44698 Klhdc4 Gm10638 Lcp1 4930539J05RikTiam1 Nrbp1 Serpinb1b D16Ertd472e Tmf1 Batf Hs6st1 Gltscr2 Nek8 2700060E02Rik Phip 4930539J05Rik Fignl1 Dph2 Etohd2 Gm25014 Thap6Nek7 Ino80dosGm43379 Glrx3BC100451 Fgl2 Pts Cbx7 SetmarSgms1 Mtx3Isca1 Rgs1Calm2 Calcoco1 Pacs2 Tmbim4 Ms4a4c Retsat Tr bv2 Runx2os3 Tbcel Gm24445 Tigar Ube2q2Ccdc90b Lin54 Lpxn Trib1 Strn4 Klf13 Zcchc18 Fdx1 Slmo2 Hspa4Tr af1 Tnfrsf9 Fndc3a Ccdc148 Ifng Lbh Ts pan6 Edaradd Sox5os3 Lrrc8c Gm44083 Gm7645 Ta cc1 Protein phosphorylation Trm Core Gls Smim3 Tbkbp1 Sep15 Dcaf12 Gpatch2 Lgals8 Abcc12 Rpp40 Gm22798 Il2rb Creb1 Nqo1 Mbp AbhdPdlim4 5 Gm22929 Tmem71 Unc13a Sypl Gm24041 Clasp2BfspGm43372 9 Dot1l Nr4a3 Mir212 Psmg1 A 1 Sp4 Styk1 Naaa Fam26f Spata13 Tmem229b Purg Atg10 Zfp36l2 Ntpcr Acvr1 Pcnp Tmbim4 Myc NdLcp2fip2 Mier1 Dopey1Pcx GgpsZkscan11 Dclk1 Akap11 Tpm3 Abtb2 Lrrc49 Gm28017Lurap1 Itga1 Kifc2 3830406C13Rik Siae Sh3gl1 Ubxn6 Iqgap2 Tram1 Colgalt1 Gm25016Prmt9 Lpin2 Ccr2 Cmpk2 Gm20705 Il21r Ap1m1 Car5b Rab8b 2700060E02Rik Rsad2 TrGm1398dc 2 Mir1955 Gm1359Phlpp1Pdlim4 1 Olfr524 Rhoa Gns Fam3c Atp5k Gm2065Sytl2 8 Anxa1 Rbl2 Gpatch2 Gcc2 Lpcat1 Gm26656 Ston2 Ubn2 Gm6990 Gm24433RP23-397L11.6 Gm42545 Lss Efr3Car5ab Gm16014 CytipDcaf17 Fez2 E130304I02Rik Ikzf5 Gzmk Platr22 Akap13 Tgfbr2 Gramd3 Gm25669 Prkcz Vrk1 Yy1 Prpf4b Pls1 Fam96b Ang Ccdc80 Zeb2 Gm13616Arid5b RP24-230C18.3Rasa4 Irak2 Tr ib2 Mpc1 Ly75 Gm30382 Tmem229b Tmem123 0610030E20Rik Gm13919 Gtf3c1 Mtmr1 Gm15567Minpp1 Pik3ap1 RP23-397L11.6 Gbe1 Bcl11b Ccnj Mrps27 Hcn1 Gm43646 Ccdc191 Zfp217 4632427E13Rik Rnpep Gm22024 Inpp4b Lcp1 Appl2 Nans Vgll4 Mei4 Cyp26b1 March3MeMir132x3c Ifngr1 Cxadr Gm8705 Arhgef12 Lhpp Gm37033 Gtf2a1 Rpl15-ps5 Tc tex1d1 Reck 4933423P22RikWdr4Tw8 f2 Fos Olfml3Bre Spi1 Wipi1 Isca1 Stk4 Tbcel Klf5 Setd7 S100a3 Kbtbd1Mtss11 Calm2 Gm16225 mt-Nd6 Gm26720 Gm22768 Batf RuNupl1nx2os3 Abhd11os Gm24433 Gm10489 Rufy1 Gm10044 1810019N24Rik Sik3 Ddx21 Gm13293 Gm38137 Fez2 Auh Ctdp1 Plekha6 Gm12709 Runx3 Wdr45b Sgk1 Gm33882Cd8b1 Mbtps2mt-Tt Dbr1 Wdr45b Psph Zbtb21 Cobll1 Gm11683 Tjp2 Bard1 Acot11 Gm25846 Rchy1 Cttn Lrrc8d Capg Tnfaip3 Ube2cbp Gm19280 Gm23802 Glipr1 Tpt1-ps6 Vsir4930558J18Rik 5830416I19Rik 5830418P13RikGm6990Csgalnact2 Foxc2 Gm28100 Uap1 Gm15856 Mtmr7 Cblb Il6st Laptm5 Inhba Nrip1 Ralgps1 Gm30238 Sec11cGm29336 Gm17509 Ets1 Vps39Olfr524 Gm10617 Gpr65 Kctd5 Jakmip1 Grb7 Dip2b Arhgap15Cd27 Gm24298 Ubn2 mt-Tp Ddx21 Fyn Clic5 Umps Neurl3 XdhZfp36l2 Wdr37 Chd9 Cpne3 Ttyh2 Oxct1 Idh1 Rfesd Golga7Snx12 Lamtor5 Snora15 Zkscan1 Fundc1 Te x9 Spin1 Necab3 Camk2d Lclat1Ints10 Lurap1Panx1 Ascc3Antxr2 Pmepa1Inpp5d Reps1 Ccdc148 Dennd1a Mycbp2 Oaz2-ps Prkcz Chaf1aFam105a Ret Nacc1 1810043G02Rik Klrd1 Rap1b RP24-286J21.7 Gm5049 Gmeb2 Tbcd Klf6 Brd7 Kctd1 Dennd4a Runx3 Gm37587TrRuibnx2os22Acadl Gpr160 Fundc1 St8sia1 4930542C12Rik Psip1 Gm45054 Gm28556 Rplp1 Vwa8 Ywhag Mir31 Vipr2 Smg6 mt-Te Cct6a Eya2 Slmo2 Oplah Gm16565 Gm11527 Il2ra Zfp507RP23-394G23.2Cep41 Arid1b Circ. Core Egr3 Nfia Tmem2 Tc eanc2 Bcorl1 Mettl22 Tbl1xr1 Pde4dCalcoco1 Dyrk2 Bms1 Tigit Slc44a5 Mirlet7i Gm29509 PtchGpc5Rrp11 2 Mettl20Xpr1 Serpinb9 Dusp10 Gm9888 Spata17 Phf14 Lmna Kctd3 Prr5l Brd4 Anp32a Rabgap1l Gm4429Mir8092 Dnajc117 Tbc1d30 Gm11928Tnfsf14 Gm19585 Lgals1 Phlpp1 Gm16225 Ctsc Jazf1 Gm4340Gm174550 Cmya5 Dazl 1810030O07RikVmp1 1700082M22RiSlc14a1k Ptcd2 4930542C12Rik Kti12 Gm43372 Ya e1d1 Tnfrsf9 RP23-394G23.2 Casp8 Gna13 Gm18604 Gm2452AL845457.3 1 Usp10 Mctp2 Gm15861Ica1 Sgms1 Dclk1 Arhgap26 Kazald1 Gm37084 Papln 1700022E09Rik Clasp1 GzmbDnmbDpp3pDopey1 Klhdc1 Gm15794 Atcayos Pcdhga8 Myo19 Plscr4 Il18r1 Gm29441 Athl1 Nrp1 Arhgap11a Zfp91 Colgalt1 Tr ak1 Nek7 TrPtpraf3c Nfatc2 Tr af4 Tnfaip3Rab13 Mpc1 mt-Te Smpdl3aMarcH3f3h3 b Klra9 Cenpw Cenpw Ta gap Atp6v0a1 Rwdd3 SntbTnfrsf1b1 Ggta1 2810474O19Rik Stk25 Entpd5 Ddx43Rbl1 Gm23633 Nsmaf Gpbp1mt-Tp Gm11529AL603834.Gm20501 3 Jmjd1cRP23-403G9.6 Srp68 Amigo2C2cd5 Ctss Acy1 Snx14 Atp2b4 Gtf2a1 Gm15999 Akap13 Fto Runx3 Ar Herpud2 Gpaa1Snx33 Gm11934 Cast Rab10os Cep63 Clcf1Gm13429 1700084E18Rik Pygl Rapgef1 Cell cycle Samhd1 Hectd2 Vcp AcadmAnapc1 Ltbp1 Gm9484Sdhb Zcrb1 Tigit Qrsl1 5330438D12Rik Atrnl1 Ikbkb Dok2 Ts pan2 ChacGm438172 Cacnb4 Kti1Gns2 Bak1 Cdadc1 Sdhb Odf2l Nr4a3 Ppp2r5a Arid5a Dhrs7b Accs Ankmy2 Fam57a MsraRbms1 Sesn3 Tmem30a Gm22024Cnrip1 Tnip3 Tc ea2 Fosl2 Srrm2 Lrrc23 Ccm2 Suclg2 Gm23011Abtb2 4931406P16Rik Filip1 Zfp1 Brwd1 Edrf1Ta cc1 Ssrp1 Ints2 Mir6538 Iqsec1 Pign Cited2 Pde7a Fam3c Mbd2 Rftn1 Ust Tbl1x Unk Gm11795 E230016M11Rik Zfp644 B4galt7 CblbCcdc22 Fam105a Ercc4 Arv1 Hspa4 Arhgap26 Gm29395 Tc ta RP23-382B3.Stam2 1 Eif3j1 Wipf1 Zfp821 Klra7 Hacl1 Gm43331 Eif3h Sec11c Tatdn1 Nob1 Yae1d1 Dcaf12RP23-413D14.6 Ankrd44 Gm16133 Map4k4 Fbxl13 Pls1 Chrm4 Gm11529 Gm28536 Ctss Spata7 LzH3f3btfl1 Sptbn1 Plac8 Creld1 Lhpp Asap1 Tr im17 Psmg1 Gm16188 Meis3 Cpsf3Frmd4alF2rl2Gm24754Gm37027Heg1 Zg16 Frmd4b Rtkn2 Trmt1 Arid4bApbb1ip Igf2r Katna1 Fbxw7 Foxc2 Usp48 Lmf1 Unc13a Spata17 Ifitm10 Cdc14a Acap1 Zfp54 Cy fip2 Rps23 Gm26720 Zfp91 RP23-451G9.2 S100a2 Ndufa12Mx1 Tmem261 Poglut1 Gm5735 Gm20379 Tigar Cln5 Svil Gm15999 Cd160 Oxsm Kat6b Brix1 Swap70 Apobec1 Ndst2 Acox1 Btg3 Myo6 Fyco1 Gmip Senp7 Dock10Hic1 4930465M20Rik Gm1361Nat96 Gm23764 Gm15549Gm16619 CregRpap2 3 Gm4364Rassf36 Gm43223 Pdcd1 Tr ps1 Tnrc6a Pold2 Elf4 Gm24574 Gm29977Tc tex1d1 Gzmk Gm382001700064M15Rik Gm9970 Te x9 4933406C10Rik Pdcd1 Myef24833407H14Rik Tc eanc2 Dnah17 Gm7726 Tmem42 Vasp Acsl5 FlotIkzf15 Tle4 Rybp Cptp Derl3 4930554C24RikSugt1 Gna12 Rnls Mcl1 Nrp1 Hectd2 Unc5b MgllMtmr7 Lmf1 Pttg1 Setbp1 Aim1 Epsti1Gm24447Rpl21 Atp5oCd44 Plek Mgat5 Abcc4Sirt1 Arv1 Aicda Adipor2 StatPlatr24 2 RP23-413D14.5 Gm29540 Rnpep Rabgap1 SpGm43246red2 Gm8113Anp32bCcdc6 Ahnak Background 2900076G11RiApoh k Asxl1 Cpne3 Stxbp3 Rps3a3To x Xpr1 Ppp2r5cD8Ertd738e Gtf2a2 LpxnPim1 Lrpprc Mir7007 Gm45029 Gm19585 Zak Ifrd1 4933425M03Rik Gm10655 Birc5 Klra7 4930554C24Rik Gm20705 Zfp637 Usp48 Zfp800 GmnnCxcr6 CandHn1 rnpa3 Cux1 Acot11 Dip2a Rasl11b Gm43039 Gm44321 Ust Pde11a Myl4 Gm13994 Id3 Golm1 1110037F02RikAhi1 Micu1 Gm2822 Brip1os Gm15232 1810011H11RikFbxw7 Ttk Scd2 Cln5 Tbcd Spag1 E230015B07RikPabpc1 Klra1 Gm2528Gm167312 Gm6223 Mtg2 BC051019 Dhrs7bCoa5 Bach1 Pigw Mitf Gm5533 Gm43177Ripk4 Gm44019 Serp2 Ts c22d3 Dazl Bcl10 Oaz2-ps Blk Ptpn12 Gm26141 Tr at1 Lonp2 Mocs1 Wdr1Gm370399 Gm7882 Txlnb Rplp1 Stx11 Src Tc p1 Itga1 Klre1 Gm5432 Atp6v0bTr p53i13Mir7688 Fhl2 Hid1 Serpinb9 Pim2 Gm13025 Vrk1 Elovl5 Gnptab Rchy1 Zfp395 Sp4 Zfp397 Tpst2 Nck2 Oxct1 Cdk5rap2 Lamtor5 Tr im25 C030005K06Rik Slc25a5 Gng2 Padi3 Slc14a1 Tmem30a Gm1745Ropn1l0 Map3k2 Cd200r3 Gm12260Pigk Gm11714 S100a2 Ptch1 Ak2 Gm20091 2010204K13Rik Stat3 Ccdc153Chn2 Gtf2irAtp1a1d2 Slc35f2 Ppp2r5a Lrrc8d Fhad1 Fzd9 Fhit Mir5135 Mir599 Gm37520 1600020E01Rik Gsg2 Mina mt-Tt Spice1 Mrps27 Hpcal1 Bzw1 Gm255Phf209 Tmsb10 Gm10584 Cmas Fbxo22 Strn4 Pacs2 Gm43672Gm27671Gm44019 Gcsam Sox17 Nf2 Gm12227 Vcp Gm6223 Fam96b Zswim6 Mrpl17 Hn1l Arl5a To x LrrkTmem38b1 RP24-122D14.3Mocs1 Gm42684 Tmtc4Tsc22d3 Bbof1 Klra5 Gm43631 Ifitm10 Itga4 Diaph3 1700025M24Rik Tmem2 Ar RasaBa1 rd1 Ube2cbp Zfp318 Pxylp1Fam160a2 Hdgfrp2 Adgrv1 Dap Tle4 Cd226Hnrnph1 Ttk Klf13 Dcaf17 Cebpb 2310043M15Rik Smurf1 Ntan1 Numbl Tr im25Afap1l2 Bcas2 Zfp644 Kcnj16 Arhgap25 E230016M11Rik Rab30 Protein dephosphorylation Pank1 Gm9888 RP24-464B16.1 Gata3 Mina Abca2 Gm11795 Etv6 Rrad Gm25794 Lrp8os3 Mms22l Efcab9Zfp326 Mdfic Spice1 Zfp292 Gm14009 Ta b2Gm11655 HdacSusd64 Ppp1r16b 9230114J08Rik Dleu2 Rasa3 Exosc44930563I02Rik Snd1 Gm26839 Wasf2 Mapre2 Gm26549Brwd1 Ak5 Set Cd7 Gm20091 AI467606 Rere B230208H11Rik Cnr2 Gm38223 Ts pyl3 Gdpd5 Sema4b Siah2 Tr pm6 Fam168b Gm15498 Nat9 C330018D20RiGlipr1k Arhgap11a Gm43380 Hsbp1 Gm26517 Cxcr4 Nsmaf Xrcc6bp1Tmem59 Dock5 Gm11683Cipc Hspa4l 5031434O11RikKlf2 Gm25459 Wwox BlmhCradd Mycbp2 Nck1 Klhl8 Nfia RP24-286J21.7 Dapk2 Ttc28 Tr ps1 Lrguk Gm19024 Madd Stk24 Bnip2 Srgap2 Brip1os Sel1l Eri2 Sema4d Stum Eya2Sppl2a Ccr4 Mir1931 Themis Gm44152 Zfp950E130311K13RiHnrnpfk Zfp316 Ttc13 Runx3 Gm13039 Ndufv2Gnptab Gm12742 Card6 Glt28d2 Duoxa2 Foxp1 Cep85l Bmp2k Snx18 Akap11 Ica1 Npm3 H2afy RpapCmklr12 Ccdc64 Camk2g Cdc42 Celf2 Gm1171Asxl3 1 Gm37240 1700017B05Rik Rfx3 Gm45153 Smpd5 Arl5b Gm20458 Fam207a D16Ertd472e Gm29540 Med30 Gm22574 Foxd2os Gm11975 A730015C16Rik4930594M17Rik Nfatc1 Gm13618 Gm12709 Golim4 Ctsc Itpr3 Abhd17b Chl1 Prex1 Gm37419 Usp39 Extl3 Mir146 Gm12708 Gm10190 Btg2 Gm43701 Tmem165 Map3k1 RP23-403I3.7 Ado Slc47a1 Fam160a2 Sash3 Mbd1 Rnf123 Gm13412 Haao Derl3 Cbfa2t3 Usp44 Sra1 Stt3b Gm29402 Gm9889 Itga4 Gm20670 0 1 23 GO Fold Enrichment c d Runx3 ChIPseq * +/+ 15 *** Runx3+/+ 10 Runx3 Con shRNA

) fl/fl ) 2 fl/fl Runx3 2 Runx3 shRNA 0 Naive Runx3 8 0 12 6 9 Activated Con shRNA 4 6 Tbx21 Runx3 shRNA 2 3 Tbet MFI (x1 Tbet MFI (x1 0 0 Cells (% Max) T-bet

+ - + CD69 e Ametrine Ametrinetrine IEL CD103+ IEL ) Thy1.1 P14 2 0 Ametrine- fl/fl + LCMV +

Tamoxifen (log Ametrine Runx3 Ert2-Cre YFP IELIEL

Tbx21 shRNA 3 97 25 75 +/+ Trm -4 Thy1.2 D0 - D4 Thy1.1.2 P14 + +/+ + CCD69D69 /Runx3 Runx3 Ert2-Cre YFP + fl/fl -8 Con shRNA CD103CD103 * 0 100 17 83 IEL Thy1.1 Thy1.2 Runx3 -12 ***

f - + Ametrine Ametrine * +/+ 80 100 Runx3 16 47 13 45 ** * Runx3fl/fl Runx3+/+ *** 80 +

60 + 10 11 60 40 33 7 29 42 40 % CD69 fl/fl % CD103 9 Runx3 20 20 0 2 CD6 0 0 CD103 Ametrine- Ametrine+ Ametrine- Ametrine+

Runx3 ChIPseq ) 2 2 ) 2 Klf2

Naive (log 2

(log 1 S1pr1 1 Ccr7

Activated / WT V/ WT l

fl/f 0 0 Klf2 -1

Runx3 -1 Runx3-R -2 Expression Fold Change Expression Fold Change -3 Extended Data Figure 8 | Runx3 regulates distinct gene programs in 1:1 and transferred into recipient mice subsequently infected with LCMV. circulating cells versus tissue resident cells and operates upstream of Recipient mice were treated with tamoxifen on days 0–4 of infection. T-bet in programming IEL TRM cell differentiation. a, Percentage of Representative flow cytometry plots (middle) and quantification of the genes of the core tissue-residency signature, core circulating signature, ratio of untransduced (ametrine−) Runx3+/+ and Runx3fl/fl P14 cells or background sites that exhibit direct Runx3 binding by ChIP–seq and the ratio of transduced (ametrine+) Runx3+/+ control shRNAmir to analysis23. b, Left, predicted Runx3 binding network, generated from Runx3fl/fl Tbx21 shRNAmir (right) were evaluated on day 12 of LCMV ATAC–seq analysis, in IEL P14 cells and splenic P14 cells on day 7 of infection. f, Representative flow cytometry plots (left) and quantification infection. Red indicates genes putatively regulated by Runx3 in IEL cells; (right) of the frequency of CD69+ and CD103+ cells. g, Runx3 ChIP–seq grey indicates genes putatively regulated by Runx3 in splenic cells. Right, of the Klf2 locus in naive and activated CD8+ T cells23. h, Fold change in Gene Ontology (GO) enrichment analysis of gene sets in the predicted gene expression of Klf2, S1pr1 and Ccr7 in Runx3fl/fl and Runx3-RV cells Runx3 binding network in each tissue. c, Runx3 ChIP–seq of the Tbx21 relative to Runx3+/+ wild-type cells, from RNA-seq analysis consisting of locus in naive and activated CD8+ T cells from ref. 23. d, Representative two replicates per sample. Data are mean ± s.e.m and representative of one flow cytometry histograms (left) and mean fluorescent intensity (MFI) of two independent experiments with n = 6 (Runx3fl/fl) or n = 4 (Runx3 quantification (right) of T-bet expression in splenic P14 cells on day 8 of shRNAmir) (d) and n = 4 (e, f) per group. *P < 0.05, ** P < 0.01 infection. e, Schematic for experimental design (left) in which Runx3+/+ ** * P < 0.005 (Student’s t-test). Symbols represent an individual Ert2-Cre+ YFP were transduced with control shRNAmir and Runx3+/+ mouse (d–f). Ert2-Cre+ YFP P14 cells were transduced with Tbx21 shRNAmir, mixed

a B16-GP33 b melanoma tumor 1.5 Day 0 CD45.1 P14 Input TIL Runx3 shRNA 1.0 CD45.2 48 52 18h 53 47 CD45.1.2 P14 CD45.1.2 P14 D7 B16-GP33 CD45.1 CD45.1 0.5 Con shRNA or Con shRNA CD45.2 CD45.2 shRNA/Con shRNA GFP-RV 1:1 Day 7 0 TIL CD45.1 P14 Runx3 Runx3 shRNA or Runx3-RV Day 15-20 TIL accumulation

c d 784 1788 811 392 560 41 Con shRNA Core Residency Signature Core Circulating Signature 405 732 552 318 442 36 Runx3 shRNA 0.6 NES = 2.53 Spleen NES = -1.83 FDR q = 0 0.4 0 FDR q = 0 PD-1 Tim3 Lag3 IFNγ GzB TNFa 0.2 -0.2

685 427 464 399 1825 362 GFP-RV 0 -0.4 1059 775 663 345 7237 367 Runx3-RV Spleen

TIL PBMC TIL PBMC PD-1Tim3 Lag3 IFNγ GzB TNFa Extended Data Figure 9 | Runx3-deficiency does not impair trafficking control shRNAmir, Runx3 shRNAmir, GFP-RV, or Runx3-RV TILs to the tumour but does affect the effector phenotype of TIL. a, Schematic in mixed transfer settings. Control P14 splenocytes were included in of adoptive therapy experimental design. b, Congenically distinct P14 histograms for reference. d, Gene set enrichment analysis of the core cells were transduced with retroviruses encoding Runx3 shRNAmir or tissue-residency and core circulating gene signatures in human lung CD8+ control shRNAmir, mixed at a 1:1 ratio, and transferred into mice with TILs relative to corresponding CD8+ PBMCs25. Data are mean ± s.e.m and established B16-GP33 melanoma tumours. Eighteen hours after transfer, combined of two independent experiments with n = 5 mice per group (b) tumours were collected to assess the ratio of Runx3 shRNAmir to control or representative of two independent experiments with n = 3–6 per shRNAmir P14 cells. c, Representative flow cytometry histograms of group (b). Symbols represent an individual mouse (b).

Corresponding author(s): Goldrath, A.G. Initial submission Revised version Final submission Life Sciences Reporting Summary Nature Research wishes to improve the reproducibility of the work that we publish. This form is intended for publication with all accepted life science papers and provides structure for consistency and transparency in reporting. Every life science submission will use this form; some list items might not apply to an individual manuscript, but all fields must be completed for clarity. For further information on the points included in this form, see Reporting Life Sciences Research. For further information on Nature Research policies, including our data availability policy, see Authors & Referees and the Editorial Policy Checklist.

` Experimental design 1. Sample size Describe how sample size was determined. Based on previous and preliminary studies within our lab, we predicted the reported sample sizes would be sufficient to ensure adequate power. For Figures 2 and 3, we expected to see a 50-75% difference in Runx3-deficient Trm compared with control Trm cells in shRNA and KO models or Runx3-RV compared to GFP-RV cells; therefore, a sample size of n=3-8 was chosen to allow determination of at least a 50% reduction in Trm (t-test, α set at 0.05). For Figure 4 d,e we also expected to see 50-75% difference in Runx3 shRNA TIL compared with control shRNA TIL or Runx3-RV compared to GFP-RV TIL; therefore, a sample size of n=3-7 was chosen to allow determination of at least a 50% reduction in TIL accumulation (t-test, α set at 0.05). For Figure 4 g,h, we expected to see a 20-50% difference in tumor size and mortality between Runx3 shRNA and control shRNA groups or Runx3-RV and GFP-RV groups; therefore, a sample size of 10-21 was chosen (t-test, α set at 0.05 or Log-rank test). Figure 1 sample sizes were chosen to achieve a sufficient cell number after sorting for subsequent processing, based on previous experiments within the lab. 2. Data exclusions Describe any data exclusions. No data were excluded from analyses except in adoptive transfer/LCMV infection experiments, recipient mice that rejected transferred P14 cells (<~5%) were excluded (Figures 2-3). 3. Replication Describe whether the experimental findings were All attempts at replication were successful. reliably reproduced. 4. Randomization Describe how samples/organisms/participants were Mice were chosen at random for each group prior to all adoptive cell transfers and allocated into experimental groups. all tumor transplants. 5. Blinding Describe whether the investigators were blinded to For all mixed transfer experiments, blinding is not relevant. For tumor growth group allocation during data collection and/or analysis. assessments, the investigator was aware of the cell type the was transferred into tumor-bearing mice. Note: all studies involving animals and/or human research participants must disclose whether blinding and randomization were used. June 2017

1 6. Statistical parameters nature research | life sciences reporting summary For all figures and tables that use statistical methods, confirm that the following items are present in relevant figure legends (or in the Methods section if additional space is needed). n/a Confirmed

The exact sample size (n) for each experimental group/condition, given as a discrete number and unit of measurement (animals, litters, cultures, etc.) A description of how samples were collected, noting whether measurements were taken from distinct samples or whether the same sample was measured repeatedly A statement indicating how many times each experiment was replicated The statistical test(s) used and whether they are one- or two-sided (note: only common tests should be described solely by name; more complex techniques should be described in the Methods section) A description of any assumptions or corrections, such as an adjustment for multiple comparisons The test results (e.g. P values) given as exact values whenever possible and with confidence intervals noted A clear description of statistics including central tendency (e.g. median, mean) and variation (e.g. standard deviation, interquartile range) Clearly defined error bars

See the web collection on statistics for biologists for further resources and guidance.

` Software Policy information about availability of computer code 7. Software Describe the software used to analyze the data in this The PageRank analysis utilized in Figure 1 was described in detail previously (Yu et study. al., Nat. Immunol, 2017).

For manuscripts utilizing custom algorithms or software that are central to the paper but not yet described in the published literature, software must be made available to editors and reviewers upon request. We strongly encourage code deposition in a community repository (e.g. GitHub). Nature Methods guidance for providing algorithms and software for publication provides further information on this topic.

` Materials and reagents Policy information about availability of materials 8. Materials availability Indicate whether there are restrictions on availability of All unique materials used are readily available from the authors. unique materials or if these materials are only available for distribution by a for-profit company. June 2017

2 9. Antibodies nature research | life sciences reporting summary Describe the antibodies used and how they were validated The antibodies are described below. All antibodies were purchased from for use in the system under study (i.e. assay and species). eBioscience unless specified. All antibodies were validated by manufacturer.

Antibody Clone Color Catalog # Lot #

CD8a 53-6.7 eFluor780 470081-82 4322567

CD8b eBio H35-17.2 BV421 (eFluor450) 480083-82 E16107-105 CD8b eBio H35-17.2 PE 12-0083-82 4287758

CD62L MEL-14 BV510 (BioLegend) 104441 B213054 CD62L MEL-14 PE-Cy7 25-0621-82 4277103 CD62L MEL-14 Fitc 11-0621-82 4278965

CD127 A7R34 PE-Cy7 25-1271-82 E07599-1635

KLRG1 2F1 PB (eFluor450) 48-5893-82 4271587 KLRG1 2F1 APC 17-5893-82 4323183

CD103 2E7 Fitc 11-1031-85 E00455-1634 CD103 2E7 PE 12-1031-83 4303133 CD103 2E7 Percp Cy5.5 (BioLegend) 121412 B212948

CD69 H1.2F3 Percp Cy5.5 45-0691-82 4313339 CD69 H1.2F3 BV711 (BioLegend) 104537 B240847

CD45.1 A20-1.7 BV785 (BioLegend) 110743 B231586 CD45.1 A20-1.7 PB (eFluor450) 480453-82 4313590

CD45.2 104 APC 17-0454-82 4290825 CD45.2 104 Percp Cy5.5 45-0454-82 4277873

Thy1.1 OX-7 PB (BioLegend) 202529 B231585 Thy1.1 OX-7 PE-Cy7 250900-82 E07586-1633

Thy1.2 53-2.1 APC 17-0902-82 E07187-1635 Thy1.2 30-H12 BV785 (BioLegend) 105331 B229101

CCR9 eBio CW-1.2 PE-Cy7 25-1991-82 4278641

CXCR3 CXCR3-173 PE 12-1831-82 4299803

CD49d R1-2 Fitc 11-0492-85 E003441630

T-bet 4B10 PE-Cy7 (BioLegend) 644823 B214293

TNFa MP6-XT22 APC 17-7321-82 E07384-1631

GzB GB11 PE MHGB04 1850394 GzB GB11 APC MHGB05 1884625

PD-1 J43 APC-Cy7 47-9985-82 4324436

Tim3 RMT3-23 PE 12-5870-82 E01844-1634

Lag3 eBio C9B7N Percp Cy5.5 46-2231-82 4295768 June 2017

KI-67 SolA15 PB (eFluor450) 48-5698-82 4297555

3 10. Eukaryotic cell lines nature research | life sciences reporting summary a. State the source of each eukaryotic cell line used. B16 mouse melanoma cells expressing B16gp33-41 were used, and were gifted by Alain Lamarre (INRS).

b. Describe the method of cell line authentication used. B16-GP33 has been authenticated in our lab, as they form melanoma tumors and using P14 T cells in killing assays to confirm GP33 expression..

c. Report whether the cell lines were tested for B16-GP33 cells were treated for mycoplasma contamination prior to use. mycoplasma contamination.

d. If any of the cell lines used are listed in the database No commonly misidentified cell lines were used. of commonly misidentified cell lines maintained by ICLAC, provide a scientific rationale for their use.

` Animals and human research participants Policy information about studies involving animals; when reporting animal research, follow the ARRIVE guidelines 11. Description of research animals Provide details on animals and/or animal-derived All mice were of a C57BL6/J background and bred at UCSD and TSRI-FL or materials used in the study. purchased from the Jackson Laboratory, including: WT or P14 mice with distinct expression of the congenic molecules CD45.1, CD45.1.2, CD45.2, Thy1.1, Thy1.1.2, and Thy1.2 as well as control Thy1.2+ Runx3+/+Ert2-Cre+YFP P14 mice and Runx3 inducible deletion Thy1.1+Runx3fl/flErt2-Cre+YFP P14 mice. Runx3+/+dLck-Cre+YFP and Runx3fl/fldLck-Cre+YFP mice were used for studying polyclonal CD8+ T cell responses. The Rosa26 stop-flox eYFP reporter mice were used for all Runx3- deletion experiments. Both male and female mice were used, and all mice were used at 6-20 wks of age.

Policy information about studies involving human research participants 12. Description of human research participants Describe the covariate-relevant population The study did not involve human research participants. characteristics of the human research participants. June 2017

4 June 2017 nature research | flow cytometry reporting summary 1

s of Final submission Revised version m nylon cell strainer (Falcon). For isolation of m nylon cell strainer (Falcon). For isolation μ Initial submission Corresponding author(s):Corresponding Goldrath, A.W. enzymatically digested with 0.7 mg/mL collagenase D. After enzymatic enzymatically digested with 0.7 mg/mL glands), tissues were further incubations (skin, lungs, kidneys, salivary dissociated over a 70 then separated using a 44/67% lymphocytes, single-cell suspensions were nodes were processed with Percoll density gradient. Spleens and lymph blood cells were lysed with ACK the frosted ends of microscope slides. Red pH 7.4). buffer (140 mM NH4Cl and 17 mM Tris-base, lymphocyte (IEL) compartment, Peyer’s patches were removed and the lymphocyte (IEL) compartment, Peyer’s cut laterally into intestine was cut longitudinally and subsequently with 0.154mg/mL 0.5-1cm2 pieces that were then incubated bicarbonate for 30min at 37°C dithioerythritol (DTE) in 10% HBSS/HEPES lungs were cut into pieces and while stirring. Kidneys, salivary glands and I collagenase (Worthington) in digested for 30min with 100 U/mL type CaCl2 at 37°C while shaking. Skin RPMI 1640, 5% FBS, 2mM MgCl2, 2mM 50) in which a 2cm2 area of the was processed similarly as described (ref 30min at 37°C and then right flank was excised, pre-digested for X-20 or a BD LSRFortessa. The software used for collecting was BDFACS Diva software and for analyzing FlowJo software was used. The purity of sorted samples were typically >98% pure. To check purity, an aliquot of sorted cells was analyzed or in cases where cells were sorted directly into Trizol, an additional aliquot of cells were sorted and purity was checked. For Figure 2 and 3, CD8a+ cells were gated on, then ametrine (shRNA lymphocytes followed by singlet discrimination gates. For Figure 1, all sorted cells from the mLN, spleen or blood were sorted based on CD8a+ and congenic markers followed by CD127, KLRG1 or CD62l as indicated. For IEL or kidney populations CD8a IV negative CD8b+ cells were gated on and congenic markers were used for sorting.

identical markers).

the flow cytometry data. populations within post-sort fractions.

Data presentation Methodological details

4. A numerical value for number of cells or percentage (with statistics) is provided. for number of cells or percentage (with 4. A numerical value 3. All plots are contour plots with outliers or pseudocolor plots. plots with outliers or pseudocolor 3. All plots are contour 2. The axis scales are clearly visible. Include numbers along axes only for bottom left plot of group (a 'group' is an analysi axes only for bottom left plot of group clearly visible. Include numbers along 2. The axis scales are 1. The axis labels state the marker and fluorochrome used (e.g. CD4-FITC). the marker and fluorochrome used 1. The axis labels state

5. Describe the sample preparation. the small intestine intraepithelial For isolation of CD8+ T cells from

6. Identify the instrument used for data collection.6. Identify the instrument used for data For flow cytometry analysis, all events were acquired on a BD LSRFortessa 7. Describe the software used to collect and analyze

8. Describe the abundance of the relevant cell 9. Describe the gating strategy used. consisted of FSCxSSC gating of For all analyses, gating schematics For all flow cytometry data, confirm that: For all flow cytometry ` Form fields will expand as needed. Please do not leave fields blank. Please do not will expand as needed. Form fields Flow Cytometry Reporting Summary Reporting Cytometry Flow `

June 2017

nature research | flow cytometry reporting summary 2

experiments) or YFP (knockout experiments) then congenic markers markers congenic then experiments) (knockout or YFP experiments) on to were gated or Thy1.2) Thy1.1.2 Thy1.1, CD45.1.2, CD45.2, (CD45.1, distinct congenically analysis of Subsequent populations. mixed distinguish iodide propidium CD69, of CD103, levels expression included populations V. and Annexin in Figure 2 except strategy was used as 3, the same gating For Figure Gating and sorting of ametrine+. cells were GFP+ instead transduced (p. 18). is discussed in Methods Fig. 3 RNAseq data strategy for was of mixed transfer populations in tumors For Figure 4, analysis P14 cells for Figure 2. Transduced and expanded performed as described or GFP reporter expression for efficacy were sorted on Ametrine experiments. Tick this box to confirm that a figure exemplifying the gating strategy is provided in the Supplementary Information. strategy is provided in the Supplementary that a figure exemplifying the gating Tick this box to confirm cOrrecTIONS & AmeNdmeNTS

ERRaTum doi:10.1038/nature25445 Erratum: Runx3 programs CD8+ T cell residency in non-lymphoid tissues and tumours J. Justin milner, clara Toma, Bingfei Yu, Kai Zhang, Kyla Omilusik, Anthony T. Phan, dapeng Wang, Adam J. Getzler, Toan Nguyen, Shane crotty, Wei Wang, matthew e. Pipkin & Ananda W. Goldrath

Nature 552, 253–257 (2017); doi:10.1038/nature24993 In this Letter, owing to errors introduced during the proofreading process, the words ‘infection with’ were missing from the sentence “Furthermore, Runx3 RNAi also impaired TRM cell differentiation in the context of a localized infection with enteric Listeria monocytogenes expressing GP33–41 (LM–GP33–41) (Fig. 2b).” In addition, in Fig. 1a, the x-axis label for the bottom right graph should have read “Expression change log2(D7 kid/D7 TCM)” rather than “Expression change log2(D35 kid/D35 TCM)”. In Fig. 1e, the arrow pointing from the spleen to TCM should have been enlarged and aligned with the arrow above, and in the heat map in Fig. 1f ‘Irf4’ should have been non-italic upright font. These errors have all been corrected in the online versions of the Letter. Supplementary Information to this Corrigendum shows the original uncorrected Fig. 1, for transparency.

Supplementary Information is available in the online version of this Corrigendum.