Resident memory T cell development is associated with AP-1 transcription factor upregulation across anatomical niches

Tissue-resident memory T (TRM) cells play a central role in immune responses to pathogens across all barrier tissues after infection. However, the underlying mechanisms that drive TRM differentiation and priming for their recall effector function remains unclear. In this study, we leveraged both newly generated and publicly available single-cell RNA-sequencing (scRNAseq) data generated across 10 developmental time points to define features of CD8 TRM across both skin and small-intestine intraepithelial lymphocytes (siIEL). We employed linear modeling to capture temporally-associated gene programs that increase their expression levels in T cell subsets transitioning from an effector to a memory T cell state. In addition to capturing tissue-specific gene programs, we defined a consensus TRM signature of 60 genes across skin and siIEL that can effectively distinguish TRM from circulating T cell populations, providing a more specific TRM signature than what was previously generated by comparing bulk TRM to naïve or non-tissue resident memory populations. This updated TRM signature included the AP-1 transcription factor family members Fos, Fosb and Fosl2. Moreover, ATACseq analysis detected an enrichment of AP-1-specific motifs at open chromatin sites in mature TRM. CyCIF tissue imaging detected nuclear co-localization of AP-1 members Fosb and Junb in resting CD8 TRM >100 days post-infection. Taken together, these results reveal a critical role of AP-1 transcription factor members in TRM biology and suggests a novel mechanism for rapid reactivation of resting TRM in tissue upon antigen encounter.


50
Adaptive immune memory mediated by T cells is central to host defense, and our 51 appreciation of its complexity has evolved considerably over the last half century. 52 Unanticipated heterogeneity in cytokine production by memory T cells was introduced in 53 1986, and migratory heterogeneity in circulating memory T cells was introduced in 1999 54 with the description of human central and effector memory T cells (TCM, TEM) 1,2 . Over the 55 past 15 years, an additional population of resident memory T cells (TRM) has been the 56 focus of many studies. TRM cells have been shown to reside long-term in peripheral 57 tissues, rather than circulate through blood or secondary lymphoid organs, and play a 58 critical role in antigen-specific recall immune responses 3 . They also appear to play a 59 role in antagonizing tumor growth 4,5 . Studies in mice have suggested that TRM 60 populations arise from early effector T cells, which after entry and conditioning in 61 peripheral tissues alter their phenotype to maintain residency and acquire long-lasting 62 memory 6 . Recently this view has been challenged by the idea that some effector T cells 63 are already committed to a TRM lineage before they enter tissue 7 . Our ability to 64 discriminate between these possibilities and our understanding of the gene program 65 changes that lead to these unique functions remains incomplete. 66 To date, regulation of a series of key transcription factors has been associated with TRM 67 development and establishment. It has been well described that Hobit or its human 68 analog Blimp1 are key regulators of TRM maintenance that function via repression of 69 genes associated with tissue egress 8 . Runx3 has been associated with TRM 70 establishment specifically in CD8 TRM, and is responsible for a cell's responsiveness to 71 the TGFβ signals associated with TRM differentiation 9,10 . Conversely, TRM precursors  Figure 1B). 131 Surprisingly, cells from the skin across all timepoints were divided into two main 132 populations, the major one of which that was defined canonical CD8 T cell markers 133 (termed "skin1"; Cd3d, Cd8A, Trbc2) and a minor one that had additional expression of where T-cell markers were expressed on cells distinct from those expressing Mhc-II, 142 Tyrobp and CD74 -a population of single OT-I cells that expressed skin2 markers was 143 also clearly observed. (Extended data Figure 1G). Given this observation, we 144 hypothesized that our skin2 population represented a mixture of T cell/MNP doublets as 145 well as a distinct population of T cells derived from antigen-activated naïve T cells. 146 However, given our inability to separate out a pure population of skin2 T cells from the   (Figures 1E-F, Extended data figure 2F). Over 75% of cells from C8-11 184 populations were from day 10 or earlier ( Figure 1D). A small population of dLN cells 185 defined by high expression of MHC-II machinery (C12) was also present in the data.

186
Unlike the cells with a similar expression profile in skin that were filtered out, these 187 represented a small fraction of the total data (n = 449 cells, 0.71%) (Extended data 188 figure 2E). Given their small representation and our inability to distinguish doublets, this 189 population was not explored further.  T cell subsets with skin-homing features defined in lymph-node. 204 We sought to assess in which tissue compartment (dLN versus skin) and at which time 205 point does T cell subset differentiation start emerging through the course of T cell 206 development. Our first two timepoints revealed little heterogeneity, with 96% of Naïve 207 day 0 cells being found in C4 and 98% of day 2 cells belonging to C5 (Extended data 208 Figure 2E). Interestingly cell subset diversification was first observed at day 5 across 209 three distinct clusters (C8, C9 and C10) representing 85% of day 5 cells from the dLN 210 (Extended data Figure 2E, 3A). C8 was enriched in interferon-response genes (Ifit1, 211 Isg20) as well as Btg1, a gene associated with T cell quiescence 34 (Figure 2A, 212 Extended data Figure 3B) Figure 2E). To understand the 219 relationship between these dLN populations (C8, C9 and C10) and the defined skin cell  predicted rate of cellular proliferation should be considered when modeling cellular 233 trajectories, we calculated a growth rate based on expression of genes associated with 234 the cell cycle and apoptosis ( Figure 2C). The predicted growth rates across the 13 235 clusters and 10 timepoints were fairly uniform with the exception of C5, which was 236 largely made up of day 2 cells and most associated a higher growth rate (mean log 237 growth rate of C5: 0.64, mean log growth rate of all other clusters: -0.13). However, 238 given that this cluster represented only small a fraction of our dataset (7,130 cells, 239 11.3%), we opted for modeling the data with a uniform growth rate.

264
To better understand genes that are associated with TRM differentiation, skin T cells 265 were subclustered independently for further downstream analysis. Additionally, we 266 included in our analysis a previously published dataset that used similar time course 267 kinetics to analyze the differentiation of siIEL TRM cells using a LCMV infection model 12 . 268 We leveraged these two temporal datasets to better define the distinct and shared   and siIEL TRM cells, the majority of dLN TCIRC genes from the VACV model were shared 307 with spleen TCIRC genes from the LCMV model (n = 208 genes, 77%) (Extended data 308 Figure 5B). Notably, the majority the dLN-specific genes in the spleen time-course were 309 most highly expressed at the latest timepoints (Extended data figure 5C). This would 310 suggest that, while strict statistical criteria did not allow them to be considered TCIRC 311 genes in the spleen, many may still be involved in TCIRC differentiation in dLN and 312 spleen. In contrast, several TCIRC-specific genes found in the spleen of LCMV-infected 313 mice were not up-regulated in the dLN TCIRC of VACV-infected mice (n = 565 genes) 314 (Extended data Figure 5B). This could plausibly be attributed to distinct cellular  Skin and siIEL T cells from VACV and LCMV models share core transcriptional 319 programs essential to resident memory T cell development. 320 One of our goals in modeling TRM development across multiple anatomical 321 compartments was to curate a consensus TRM gene signature. While many of the 322 defined TRM genes were distinct in our two anatomical niches, we found 71 genes were   Cebpb) that could successfully distinguish resident from circulating T cell subsets, 510 further validating our analytical strategy. Future work will be needed to characterize the 511 role of these newly defined targets in TRM development.

512
As noted above, this consensus TRM signature included genes associated with 513 responses to hypoxia as well as those classified as immediate early response genes.

514
The immediate early genes were hallmarked by AP-1 transcription factor members, 515 whose importance in TRM differentiation across tissues was also highlighted in our                  Source code for data analysis will be available on GitHub upon publication of this study.