Effector T_H17 cells give rise to long-lived T_RM cells that are essential for an immediate response against bacterial infection

Summary

Adaptive immunity provides life-long protection by generating central and effector memory T cells and the most recently described tissue resident memory (T_RM) cells. However, the cellular origin of CD4 T_RM cells, and their contribution to host defense remains elusive. Using IL-17A tracking-fate-mouse models we found that a significant fraction of lung CD4 T_RM cells derive from IL-17A producing effector (T_H17) cells following immunization with heat-killed Klebsiella pneumonia (Kp). These exT_H17 T_RM cells are maintained in the lung by IL-7, produced by lymphatic endothelial cells.

During a memory response, neither antibodies, γδ T cells, nor circulatory T cells are sufficient for the rapid host defense required to eliminate Kp. Conversely, using parabiosis and depletion studies, we demonstrated that exT_H17 T_RM cells play an important role in bacterial clearance. Thus, we delineate the origin and function of airway CD4 T_RM cells during bacterial infection, offering novel strategies for targeted vaccine design.

Introduction

Adaptive immunity provides life-long protection from infection by establishing an enduring reservoir of memory cells. Upon infection CD4 and CD8 T cells undergo clonal expansion followed by a subsequent contraction, and ultimately memory formation (Kaech, Wherry et al. 2002, Mueller, Gebhardt et al. 2013). While central memory T (T_CM) cells have long been known to mainly reside and circulate through secondary lymphoid tissues and effector memory (T_EM) cells to circulate through non lymphoid tissues, the identification of T_RM cells within peripheral organs has opened new concepts about tissue-specific protective immunity (Schenkel and Masopust 2014).

CD4 T_RM cells serve at the frontline at different barrier sites such as vagina (Iijima and Iwasaki 2014, Stary, Olive et al. 2015) lung (Teijaro, Turner et al. 2011, Turner, Bickham et al. 2014, Hondowicz, An et al. 2016) and skin (Glennie, Volk et al. 2017) and their main host-protective mechanism is mediated by the production of effector cytokines such as IFN-γ, IL-4.

While most of the studies have focused on the biology of virus-specific CD8 T_RM cells, basic questions regarding CD4 T_RM still remain unanswered. For example, where do these cells come from - what is the cellular origin of CD4 T_RM cells?; which environmental signals maintain them into the tissue?; and finally, are CD4 T_RM cells able to rapidly respond to a second challenge?

To address these questions, we devised an experimental plan combining the use of two complementary fate reporter mouse models, next generation T cell receptor (TCR) and single-cell sequencing approaches.

We consider that understanding CD4 T_RM cell development and function is imperative, particularly as they may represent the key target cells to design specific and efficacious vaccination against the growing number of antibiotic resistant bacterial strains emerging in the clinic. This might also help to reduce antibiotic use, avoiding the evolution of antibiotic resistant-bacteria For instance, there is currently no targeted treatment strategy for carbapenem resistant Enterobacteriaceae like Klebsiella pneumoniae (Kp). This is a leading cause of nosocomial and community-acquired gram-negative bacterial pneumonia, which results in a severe pyrogenic infection with high mortality rates (Falagas, Tansarli et al. 2014). Despite the fact that IL-17A cytokine is critical to deal with Kp infection (Moore, Moore et al. 2000, Happel, Dubin et al. 2005, Chen, McAleer et al. 2011), the function of T_H17 memory cells have been underestimated because of their short-term survival (Pepper, Linehan et al. 2010).

Considering all this, we hypothesized that CD4 T_RM cells derive from the first wave of effector cells generated during the first encounter with a pathogen. Furthermore, since CD4 T_RM cells localize at the site of immunization, we also hypothesized that some of them acquire a poised while others a more plastic status (Lee, Turner et al. 2009, Wei, Wei et al. 2009, Harrison, Linehan et al. 2019), which allow them to mount a fast and essential immune response against bacterial infection.

Here, by using an immunization-infection model with different serotypes of Kp, we show that a significant fraction of the lung long-lived CD4 T_RM cells (exT_H17cells) derive from specific-effector T_H17 cells. We observed that exT_H17 T_RM cells are maintained by IL-7 which is mainly produced by lymphatic endothelial cells in the lung. Finally we reveal that exT_H17 T_RM cells are important players that can resist an antibiotic resistant strain of Kp.

Results

T_RM cells can derive from effector cells and protect against carbapenem resistant Klebsiella pneumoniae infection

We started by characterizing the kinetics of the development of lung- T_RM cells. To this end, wild type (wt) mice were immunized twice with heat killed serotype 2 Klebsiella pneumoniae (Figure S1A) and the presence of CD4 T_RM cells was evaluated at multiple time points. An in vivo antibody (Ab) labeling technique was used to differentiate between circulatory and lung infiltrating CD4 T cells (Anderson, Mayer-Barber et al. 2014). Lung infiltrating CD4 T cells began accumulating as early as day 5 post-immunization and persisted through day 110 (Figure S1B). CD69 and CD103 have been used as markers for T_RM cells. We found that CD4 T_RM cells were CD103⁻ but characterized by high levels of CD69 expression compared with circulatory CD4 T cells (Figure S1C and D). Similar to classical CD8 and CD4 memory formation, lung infiltrating CD4 cells underwent robust expansion upon immunization, followed by an acute contraction phase. Then a stable lung T_RM CD4 population was observed during the memory phase (Figure S1E).

We next aimed to further characterize the origin of T_RM CD4 cells, a point that has still remained elusive. T_H17 cells have previously been shown to provide protection against Kp infection (Ye, Rodriguez et al. 2001, Chen, McAleer et al. 2011), however it is unclear whether these cells are short-lived and whether they contribute to the memory pool (Pepper, Linehan et al. 2010, Chen, McAleer et al. 2011, Muranski, Borman et al. 2011). Considering the plasticity of T_H17 cells and instability of IL-17A production, their main marker, one possible explanation for the unknown-origin of CD4 T_RM cells could be that T_H17 cells lose the expression of their signature cytokine, becoming “exT_H17” cells, when they mature into long-lived memory cells. We therefore used the Il17a Fate⁺ mapping mouse model (Gagliani, Amezcua Vesely et al. 2015), which allowed us to track the fate of T_H17 cells independently from the actual expression of IL-17A and from the need of in vitro re-stimulation. To generate these mice, Il17a^CRE, crossed with a R26STOP^flox/flox YFP (R26^YFP) (Hirota, Duarte et al. 2011) were further crossed with a triple-reporter mouse strain, consisting of Il17a^Katushka Il10^eGFP Foxp3^RFP alleles (Gagliani, Amezcua Vesely et al. 2015). In addition, the Fate⁺ mice allowed us to exclude Foxp3 and IL-10 producing cells from our analysis. By using these mice, we observed that Kp immunization resulted in the generation of T_H17 and exT_H17 cell populations, with the latter persisting in the lung as a population of T_RM cells for at least 110 days (Figures 1A and B).

Figure. 1: Origin and function of T_RM exT_H17 cells.

A) Dot plots show percentage of T_H17, exT_H17 and YFP^neg IL17^neg cells and B) absolute number of T_RM exT_H17 cells in lungs at different time points after immunization in Fate⁺ mice (gated on CD4 in-vitro⁺ CD4 in-vivo ^Neg TCR-β⁺ Foxp3^Neg) (n= 4 per time point). One out of 2 independent experiments is shown C) Naïve CD45.1⁺ and day 35 immunized Fate⁺ CD45.2⁺ mice were surgically joined for 3 weeks. Before sacrifice anti-CD4 antibody was injected to distinguish T_RM and circulating cells. The graph shows the absolute number of lung T_RM CD45.2⁺ and circulatory exT_H17 cells in naïve CD45.1 mice (blue) and in Fate⁺ immunize mice (red). D) Representative immunofluorescent sections of lungs from naïve (Day 0) and immunized mice (Day 14 and Day 30) showing the presence of CD4 T cells (red) and exTh17/Th17 cells (yellow) surrounding the airways only in the lungs of immunized mice. AW= airway and BV= blood vessel; blue: DAPI, grey: CD31 (n= 3 per time point) E) Parabiotic pairs were infected with live carbapenem resistant Kp after 3 weeks of surgery. The graph shows the CFU count in total lung 24 h after infection of naïve-infected (blue) or immunized-infected (red). Each dot represents one mouse. Data is cumulative of 2 experiments.

Data are represented as mean ± SEM. Mann-Whitney U test, ns (non significant); *p= 0.0286; **p = 0.0022. Data are represented as mean ± SEM.

To specifically track T_H17 cells generated in response to Kp immunization, rather than all cells with a history of IL-17A expression (Ivanov, Atarashi et al. 2009), we took advantage of an inducible Il17a fate model (Gagliani, Amezcua Vesely et al. 2015). R26^YFP mice were crossed with Il17a ^{eGFP-Cre-ERT2} mice, such that cells expressing Il17a are labeled GFP⁺YFP⁺ upon tamoxifen administration and persist as a GFP^negYFP⁺ population after the expression of Il17a is terminated. As expected, Kp immunization resulted in the induction of a newly generated T_H17 cell population (Figure S1F). Notably, 60 days after immunization most of these cells lost Il17a^eGFP expression and persisted as YFP⁺ exT_H17 cells in the lung of immunized mice (Figure S1G).

To test whether the accumulation of T_RM was associated with long-term protection, the immunized mice were infected with a second serotype of live K. pneumoniae, characterized as being resistant to carbapenems (Figure S2A). Immunized mice showed protection against carbapenem resistant Kp compared with non-immunized mice (Figure S2B).

CD4 T cells, antibodies, and γδ T cells are all important components of the immune response against Kp. However, whether memory T cells alone are sufficient to orchestrate a proper immune response, especially at the early phase of memory activation, remains unknown. Here we found that despite the absence of B cell and γδ-T cells – using genetically deficient mice - immunized animals were still more efficiently protected from acute live carbapenem resistant Kp infection than non-immunized controls (Figures S2C and D).

Next, to further test whether exT_H17 cells were residing in the lung as bona fide T_RM cells, we performed parabiosis experiments with naïve wild type CD45.1⁺ and immunized Fate⁺ CD45.2⁺ mice in combination with an in vivo Ab labeling technique. Naïve CD45.1 and immunized CD45.2 mice were surgically paired and 3 to 4 weeks later we observed shared specific Kp antibodies and CD4 T cells in the circulation (Figures S2E, F and G). We observed that, although the two mice shared memory exT_H17 T cells in circulation, the large majority of exT_H17 cells were present in the lungs of the immunized Fate⁺ mice and not in the lungs of the naïve wild type mice (Figure 1C). In addition, immunofluorescent imaging of lungs from immunized mice showed that exT_H17 cells (depicted by yellow) resided in close proximity to the airways and blood vessels at 14 and 30–35 days after immunization forming a potential network of tissue resident cells (Figure 1D).

Finally, to fully explore the hypothesis that these observed resident cells act as the crucial mediators of tissue specific adaptive immunity, we conducted parabiosis studies infecting both naïve and immunized mice with antibiotic resistant Kp. Despite displaying Kp specific antibodies and circulatory memory cells, the respective naïve mice of the joined pairs failed to efficiently clear live carbapenem resistant Kp infection. In contrast, we found a very low bacterial burden in the immunized partners (Figure 1E). Moreover, we observed that when the recruitment of memory cells from lymphoid organs is impaired, using FTY720, the immunized mice still appeared to be more protected compare to naïve mice (Figure S2H).

We extended our findings to different pathogens, and found long-lived exT_H17 cells after Candida albicans and BCG infections (Figure S3), supporting the hypothesis that the longevity and presence of T_H17 effector derived T_RM cells does not only occur in the Kp setting, but they are also generated in other infections where IL17A plays an important role (Khader, Bell et al. 2007, Igyarto, Haley et al. 2011).

Taken together, these data show that effector T_H17 can give rise to T_RM cells and T_RM cells play a key protective role in the early phase of Klebsiella pneumoniae infection.

TCR specificity and transcriptome analysis of lung T_RM cells

The expansion of few clones in response to a bacterial infection is a prototypical feature of a memory response. To test whether the development of these T_RM cells follows this classical path, we sequenced the CDR3 region of the TCRβ genes of T_H17 and exT_H17 cells present in the lung 35 days after the immunization. Thus, this CDR3 sequencing also allowed us to elucidate the clonal origin of exT_H17 T_RM cells.

We found that the small population of T_H17 cells present at day 35 post-immunization was represented by highly restricted oligo-clonal expanded T cell receptor clones and these clones were shared with exT_H17 cells (Figure 2A,F), further indicating a common origin between these two populations. Then, we selected the most expanded clones among lung T_H17 cells and tracked them in all the other populations of T_RM cells in the lung and also in naive and memory cells in the spleen (Figure 2 B,C,E,G,H,J). When we evaluated the clonality of those lung T_RM cells that never expressed IL-17A (YFP^neg ), we found that they share some degree of clonality with both T_H17 and exT_H17 T_RM cells but to a lower degree compared to what T_H17 and exT_H17 T_RM cells share between one another (Figure 2 B,D,G,I). This suggests that both type of cells were generated during the effector immune response but unknown differences led them to become YFP^pos or YFP^neg cells. Finally, some of the expanded clones represented in the T_H17 cells, were also present, albeit at a very low frequency in splenic CD4 memory cells but they were not detectable in naïve CD4 T cells, denoting the generation of a specific and systemic immune response after immunization (Figure 2C,H).

Figure. 2: TCR β sequencing of Lung T_RM cells.

A, F) TCR β sequencing of lung T_RM T_H17 vs lung T_RM exT_H17 cells, B, G) lung T_RM T_H17 vs lung T_RM YFP^neg C, H) lung T_RM T_H17 vs splenic memory CD4 cells and vs splenic naive CD4 cells at day 35 after immunization in 2 different mice. D, I) TCRβ sequencing of lung exT_H17 vs lung YFP^neg at day 35 after immunization in 2 different mice. The dots plots graphs show the common shared CDR3 amino acid sequences of TCRβ chains between all the populations. The most expanded clones among T_H17 cells were colored and followed across all the comparisons between different populations. E, J) The bar graphs on the left show percentage of the most expanded clones in each population. The bar graphs on the right show the cumulative productive frequency of the most expanded clones in lung T_RM cells. The colors of the clones followed the same colors used for the dot plots graphs.

Tissue and cell-type specific expression of genes, including T_RM “core” genes, can be commonly suggestive of corresponding specific functions and cell origin. To evaluate the presence of a tissue-related specific gene profile in memory cells residing at the lung or lymph nodes we performed mRNA-sequencing. We hypothesized that based on their location, different transcriptomes will be expressed by these memory CD4 T cell populations. We profiled the transcriptome of draining lymph node (LN) naïve CD4 T cells, memory CD4 T cells and the 3 populations of T_RM cells present in the lung 35 days after immunization (T_H17, exT_H17 and YFP^neg).

We used either an unsupervised list of genes that were significantly differentially expressed (> 4 fold) (Figure 3A) or a core of T_RM signature genes (Figure 3B) (Wakim, Woodward-Davis et al. 2012, Mackay, Rahimpour et al. 2013, Pan, Tian et al. 2017) to test similarities and differences among the transcriptomes of these cell populations. We found that lung T_RM cells (T_H17 cells, ex-T_H17 and YFP^neg cells) share a similar pattern of gene expression, but different from the expression pattern of the peripheral memory LN cells, independently of the approach used to analyze them (Figures 3A,B).

Figure 3: RNA sequencing of Lung T_RM and lymph node memory and naïve CD4 populations illustrates key features of Lung T_RM populations.

A) Heatmap of genes significantly differentially expressed (>4 fold cutoff) between the lymph node and lung populations. B) Heatmap of a selected subset of genes known to be involved in tissue resident memory cell biology. C) Bar graph of gene ontology assignments of genes overexpressed 4 fold in lung T_RM cells relative to lymph node memory cells. Dashed line represent p = 0.05, Fishers exact test. The intensity of the color bar on the right indicates the amount of genes in each pathway. D) Heatmap of a selected subset of genes known to be involved in T helper subset and memory biology.

To search for different functional features among all lung T_RM cells and peripheral LN memory cells, we performed Gene Ontology (GO) analysis of genes upregulated in Lung T_RM cells using DAVID (Huang da, Sherman et al. 2009). Interestingly, GO terms showing statistically significant enrichment were mainly related with immune process such us defense response, cell to cell adhesion, cell activation, regulation of cytokine production and response to bacteria(Figure 3C). All this suggests that lung T_RM cells present a profile more focused towards a rapid immune response compared with peripheral memory cells. We then focused on the different populations of lung T_RM cells and evaluated the expression of a selected list of genes involved in specific T-helper cell subset and memory T cell biology. Interestingly, based on the expression of the genes from this list, we found that T_H17 and exT_H17 cells cluster more closely together than to the YFP^neg cells (Figure 3D). In particular, the expression patterns of Ilr1, rora, irf4 and rbpj suggested that some of the exT_H17 T_RM cells have acquired a prime signature-state that could allow them to rapidly respond producing IL-17.

Interestingly we also found that ndfip1, which was recently shown to be essential to balance the potential tissue related pathogenicity of T_H17 cells (Layman, Sprout et al. 2017), was highly expressed in exT_H17 cells. We also observed that the expression of Il7r was higher in exT_H17 cells than in YFP^neg cells. Finally, regarding the population that never activates the production of Il17a (YFP^neg cells), the expression of a T_H1/2-memory signature transcriptome (e.g. eomes, Ifng, ccr7, sell, Il4, Gata3, Ccr3, Ccr5), suggests that these cells might be a heterogeneous population composed of T_H1 and T_H2 effector derived memory cells (Figure 3D) (Chang, Palanivel et al. 2007, Bannard, Kraman et al. 2009).

Effector derived T_RM cells depend on IL-7 which is produced by lung lymphatic endothelial cells

We next interrogated the possible mechanisms that might be involved in the maintenance of exT_H17 T_RM cells in the lung. The residency and establishment of T_RM cells at peripheral tissues correlates with down-regulation of KLF2 and S1P1R, and the concomitant inability to follow a gradient of S1P and leave the tissue. (Carlson, Endrizzi et al. 2006) (Skon, Lee et al. 2013). Consistently, lower levels of KLF2 and S1PR1 expression were found in lung T_RM exT_H17 cells and in the small population of T_H17 cells compared to CD4 splenic memory or naïve CD4 T cells from Fate⁺ immunized mice (Figure 4A).

Figure 4: Lymphatic endothelial cells maintain effector derived lung T_RM cells.

A) Relative RNA expression of KLF2 and S1PR1 on day 30 in immunized mice is shown. Splenic naïve CD4 cells (black bars), splenic memory CD4 cells (red bars), lung exT_H17 cells (light blue bars) and lung T_H17 cells (dark blue bars). Technical replicates of one representative experiment out of two are shown and data is represented as mean ± SD. B) The graphs show the absolute number of total T_RM CD4 (left) and T_RM exT_H17 cells (right) in lungs (gated on CD4 in-vitro⁺ TCR β ⁺ Foxp3⁻ CD4 in-vivo^Neg) after FTY720 treatment (red circles). Non Tx mice= normal drinking water (black circles). Each dot represents one mouse. One out of 2 independent experiment is shown. C, D) Histograms show the expression of CD127 and CD122 in lung T_RM exT_H17 cells (red line) 60 days after immunization, and in total spleen CD4 cells (blue line). CD122 expression in total CD8 spleen cells is shown as a positive control (grey line) (n= 4 mice). One out of 2 independent experiments is shown E) Fate⁺ immunized mice were intra-nasally treated with anti-IL-7 Ab, or with an isotope control on day 35, 37 and 39 after immunization and the mice were sacrificed on day 40. The graph shows the absolute number of lung T_RM exT_H17 cells in the isotype ctrl treated group (black circles) and in the anti-IL7 treated group (red circles). Each dot represents one mouse. Three combined experiments are shown. F) Single Cell RNA sequencing of total lung tissue cells from immunized mice with total CD4⁺ cells and circulatory CD45⁺ removed by FACS sorting. Lung cells were profiled by droplet-based scRNA-seq (Drop-Seq). t-SNE plots show 1500 cells (dots) in a nonlinear representation of the top 50 PCs. Cells are colored by cluster. The clusters were identified by specific genes (Supplementary Figure.7). G) Representation of differentially expressed Il7 gene by cluster (y-axis) in immunized and naive conditions (x-axis). The dot size represents the fraction of cells in the cluster that express Il7; the dot color represents the intensity of Il7 expression. H) Representative immunofluorescent sections of lungs from B6-non immunized (left), IL7^GFP/+ non-immunized (middle) or day 35-immunized mice (right). Colocalization of CD4 T cells (red) and IL-7⁺lymphatic endothelial cells (Grey+ green) in the lungs of immunized mice. Blue: DAPI, grey: Lyve-1, green: IL7, red: CD4. I)The histogram shows the percentage of lung T_RM exT_H17 BrdU⁺ cells. Each dot represents one mouse. One out of 2 independent experiments is shown. Data are represented as mean ± SEM. Mann-Whitney U test ** p = 0.0082.

Then, to evaluate whether a continuous supply of circulating lymphocytes is required for the maintenance of the pool of CD4 T_RM, we treated day 35 immunized mice with FTY720, an S1PR1 antagonist that blocks lymphocyte egress from lymph nodes. Decreased numbers of circulating total white blood cells were found after 2 weeks of FTY720 treatment, while, by contrast, the total number of CD4 T_RM and exT_H17 T_RM cells in the lung remained unchanged (Figure 4B). All these data pointed towards in situ mechanisms that might sustain the survival of CD4 T_RM cells. Therefore, we investigated the possible role of the cytokine milieu. Both IL-7 and IL-15 signaling are understood to be key mediators of CD4 T cell homeostatic proliferation, a mechanism which could explain the long-term and circulatory independent maintenance of T_RM (Mackay, Rahimpour et al. 2013, Iijima and Iwasaki 2014, Adachi, Kobayashi et al. 2015, Hondowicz, An et al. 2016). In line with the data obtained from the RNA sequencing experiment, we found that a fraction of T_RM exT_H17 cells expressed IL-7Rα, but they lacked IL-15Rβ expression (Figure 4C, D). To test the role of IL-7 in the maintenance of T_RM exT_H17 cells, we intranasally administered anti-IL-7 or isotype control antibodies into immunized mice. The local treatment with anti-IL-7 Abs did not diminish CD4 T cell numbers systemically (Figure S4A). However, the number of lung T_RM exT_H17 cells was reduced after IL-7 blockade (Figure 4E).

Then, considering the key role of IL-7, we also investigated the cellular source of IL-7 in the lungs of immunized mice. By performing single cell RNA-sequencing of total lung cells and unsupervised clustering (Figure S4B), we observed that first Il7 was preferentially expressed by lymphatic endothelial cells (LEC, cluster 10) and second the percentage of Il7⁺ LEC increased in immunized mice (Figure 4F,G). By using IL7^GFP/+ reporter mice, we found that lung CD4 T_RM cells are located in close proximity to IL7⁺ lymphatic endothelial cells in the lung of Klebsiella immunized mice (Figure 4H). Finally, as IL-7 is known to promote the homeostatic proliferation of T cells, Fate⁺ mice were immunized and 60 days later given bromo-2-deoxyuridine (BrdU) in the drinking water for 2 weeks. Approximately 20% of T_RM exT_H17 cells were labeled with BrdU, suggesting that homeostatic proliferation plays a role in maintaining CD4 T_RM cells (Figure 4I).

Function of effector derived T_RM cells

T_H17 cells are a distinct subset of effector CD4 T cells with a degree of plasticity (O’Shea and Paul 2010, Schlapbach, Gehad et al. 2014, DuPage and Bluestone 2016). However, whether exT_H17 T_RM cells can maintain their ability to reproduce their original cytokine profile when re-challenged with antigen remained to be addressed. If so, this could be interpreted as an evolutionary advantage to be able to maintain a population of memory cells at the previously infected barrier to provide a quick response to eventual subsequent infections. We therefore determined the expression of IL-17A and observed that only previously immunized mice show a rapid (i.e. 2 and 24 h) increase in IL-17A^Kata expression following infection with carbapenem resistant Kp (Figure 5A). One feature of T_H17 cells is their ability to recruit neutrophils (Fossiez, Djossou et al. 1996, Schwarzenberger, La Russa et al. 1998). In line with this, immunized mice showed a significantly increased number of neutrophils surrounding the bronchial lung barrier, within 2 hours after infection (Figures 5B, C).

Figure. 5: Function of T_RM exTh17 cells.

A) Dot plots show the frequencies of T_RM T_H17 cells (CD4 in-vitro⁺, CD4 in-vivo^neg, TCRβ⁺ Foxp3^neg, IL17^kata+, YFP⁺) in naïve+ infection mouse, immunized mouse (day 35), immunized mouse + infection (2 hours) and immunized mouse + infection (24 hours) (n=3 mice per group). The right graph shows the MFI of IL17^kata in T_RM T_H17 cells. B) Peroxidase immune staining of mouse Ly6B in naïve, naïve+ infected, immunized (day 35) and immunized (day 35) + infection lungs, 2 hours after Kp infection. Representative slides are shown for each group. C) Quantification of neutrophils surrounding bronchioles from B. D) Cake graph shows the percentage of INF-γ⁺ (dark grey) among the population of YFP⁺ cells 24 h after infection with live Kp (after PMA-ionomycin stimulation). Double positive cells (IL17A⁺, INF-γ⁺ middle grey) and IL17A⁺ cells (light grey) are shown separately. The graph on the right shows the percentage of INF-γ⁺ single positive cells in immunized + infection (yellow circles) and in immunized mice not infected (day 35 after immunization). E) Constructs of Fate-DTR mice. F) Scheme depicting the time plan of immunization, diphtheria toxin treatment (DT Tx) and infection of Fate-DTR mice (Top). The graph shows the CFU counts in the indicated groups. For the first 4 groups (from left to right) Fate-DTR mice were used, while for the last group Fate⁺ immunized mice were used. Each symbol represents one mouse.

Data are represented as mean ± SEM.

Kruskal-Wallis (p < 0.0001) and post-hoc Mann-Whitney U test with Bonferroni correction *p = 0.0238 **p = 0.0043 (A); one-way Anova, and post-hoc Dunnett * p< 0.05; **p < 0.005 (C); and Kruskal-Wallis (p = 0.0009) and post-hoc Mann-Whitney U test with Bonferroni correction, ns (non significant), *** p < 0.0001 (F).

Although IL-17A activity is sufficient to control spread of the bacteria, IFN-γ seems to be necessary to fully promote efficient control of Kp (Happel, Dubin et al. 2005). In parallel, it has also been shown that T_H17 cells have the ability to acquire a T_H1 like phenotype (Wei, Wei et al. 2009, Hirota, Duarte et al. 2011, Zielinski, Mele et al. 2012). We therefore wondered whether some exT_H17 cells were capable of expressing IFN-γ rather than IL-17A upon reencountering bacterial derived antigens. Our data show that a fraction of exT_H17 cells do not re-express the original cytokine IL-17A but instead IFN-γ within 24 hours after infection (Figure 5D). These data extend to T_RM CD4 T cells, the ability of some CD4 T cells to maintain a degree of flexibility.

Finally, we wanted to test whether T_RM exT_H17 cells and their several functions, are necessary to orchestrate the tissue resident memory immune response. We therefore developed a mouse model to delete cells with a history of IL17A production such as exT_H17 and their inherent ability to rapidly produce IL-17A at a desired time point. We crossed Il17a^CRE R26^YFP Foxp3^RFP with conditional diphtheria toxin receptor (DTR) mice (R26STOP^flox/flox DTR; referred to here as Fate-DTR) (Figure 5E). In this mouse model only the YFP⁺ cells express DTR and can therefore be selectively deleted at the time of diphtheria toxin (DT) treatment. We first tested the efficacy of the DT mediated depletion and observed that exT_H17 cells were depleted in the lung after three injections with the toxin (Figure S5A). As expected, γδ-exIL17 T cells were also deleted. However they did not appear to play an obvious role during memory response against Kp (Figure S2D).

We then took advantage of this model to test whether other tissue resident cells in the lung (i.e. CD4⁺ YFP^neg) are sufficient to provide a reservoir for the generation and in turn the maintenance of exT_H17 T_RM cells. Therefore, day 35 immunized mice were treated or not treated with DT and rested after 21 days (Figure S5B). Then the absolute number of exT_H17 cells and total T_RM CD4 cells were evaluated. No new generation/differentiation of exT_H17 cells was observed 21 days after DT treatment suggesting that a new wave of effector CD4 T cells is required to re-establish long-lived exT_H17 T_RM cells following depletion (Figure S5C, D and E).

Finally, immunized Fate-DTR mice were infected with live carbapenem resistant Kp and the bacteria burden was tested 24 hours later. Mice depleted of exT_H17 cells before infection were impaired in their ability to control bacterial load. As expected, the non-depleted mice, whose lungs still contained T_RM exT_H17 cells, were able to effectively control the bacterial infection (Figure 5F). These data suggest that effector derived T_H17 cells are indeed required to rapidly control this bacterial infection. Naïve depleted and naïve non-depleted mice were used as controls and both groups behaved as the immunized exT_H17 cell depleted group. Moreover, immunized Fate⁺ (non DTR) mice were used as additional controls for non-specific DT effects, showing protection after infection.

Discussion

Several studies have focused on the molecules and or transcription factors needed for the arrival and or retention of T_RM cells (Mackay, Rahimpour et al. 2013, Skon, Lee et al. 2013, Mackay, Minnich et al. 2016). However, whether T_RM cells derive from effector T cells remained unclear. Here we showed that lung infiltrating CD4 T cells became T_RM cells at later time points after Klebsiella pneumoniae immunization and protected immunized mice from carbapenem resistant Klebsiella pneumoniae infection. By using two state-of-the-art IL-17A fate mouse models, we demonstrated that effector T_H17 cells were capable of persisting as a long-lived exT_H17 population, creating a resident cellular network in the lung airways. We observed a similar phenomenon using others T_H17-induced pathogens such us Candida albicans and BCG. Using parabiosis experiments, we showed that T_H17 effector-derived T_RM cells persisted in the lung of immunized mice, despite reaching an equilibrium in the circulation of both paired parabionts. Infecting both parabionts or blocking the recruitment of circulatory cells from the periphery, CD4 T_RM cells proved the fundamental advantage to an organism of having a local memory immune response over a systemic memory response.

We sought to compare and contrast the clonality and the whole transcriptome of lung T_RM cells vs periphery memory and naïve cells. Here, we found that lung T_RM cells comprise 3 distinct populations: a) exT_H17 cells, derived from effector T_H17 cells generated during effector immune response b) a low percentage of T_H17 cells and c) YFP^neg cells, or cells that despite their never having been IL-17A producers, also became T_RM cells. Moreover, the TCR-seq analysis revealed a common clonal origin of T_H17 and exT_H17 T_RM cells with highly frequent clones that first expanded, then contracted and finally, generated a bona fide T_H17 T_RM response. The frequency of YFP^neg T_RM clones is in line with some previous studies where a T_H1 population was also generated during the first immune response against Klebsiella pneumoniae (Happel, Dubin et al. 2005). Additionally, our whole-transcriptomic analysis revealed similarities among all lung T_RM cells, distinguishing them from peripheral memory cells. Nevertheless, some effector derived exT_H17 cells still maintained a “T_H17 prone -memory status” (Muranski, Borman et al. 2011, Youngblood, Hale et al. 2017) which may explain their fast response. A more detailed analysis of YFP^neg T_RM cells unveiled possible distinct origins for this population. Based on their T_H1/ T_H2 and T_CM related transcriptome and in line with previous human studies (Sathaliyawala, Kubota et al. 2013, Becattini, Latorre et al. 2015), we speculate that some YFP^neg T_RM cells may have derived from effector/effector memory T_H1 and T_H2 cells, and some from central memory cells. However, more studies using polyclonal and combined cytokine-fate-mapping systems are needed to clarify the origin of this heterogeneous population of cells. Previous studies, using TCR-sequencing or skin injections of TCR transgenic T cells have suggested that central memory CD8 T cells are the the origin of skin CD8 T_RM cells (Gaide, Emerson et al. 2015, Kirsch, Watanabe et al. 2015). Our results show that CD4 T_RM cells were also able to derive from the first wave of effector cells generated at the beginning of the immune response.

The mechanisms elucidated for CD8 T_RM cell maintenance indicate that cytokines, transcription factors and even some free-fatty acid binding proteins could be involved in these processes (Adachi, Kobayashi et al. 2015, Mackay, Wynne-Jones et al. 2015, Schenkel, Fraser et al. 2016, Pan, Tian et al. 2017). Nevertheless, little is known about CD4 T_RM cell maintenance. We hypothesized that based on the tissue and the environment where T_RM cells are located, different mechanisms will operate to maintain their survival. In our models we observed that there was no contribution from peripheral memory cells when it come to maintenance of exT_H17 cells in the lung. By antibody-blocking and single-cell-RNA-seq experiments, we demonstrated that IL-7 derived from lymphatic endothelial cells was needed to maintained the survival and location of lung exT_H17 T_RM cells. Our results highlight the importance of stromal cells in the maintenance of T_RM exT_H17 cells, resembling similar mechanism by which IL7 support other cells in the bone marrow (Cordeiro Gomes, Hara et al. 2016).

Another aspect that we experimentally addressed is the mechanisms whereby T_RM cells induce protection during infection. Our in-vivo results are in line with the “primed status” of T_RM cells, usually view as a frontline branch of the adaptive immune system, that “seeds” mucosal tissues. The kinetic experiments suggested that some of the lung T_RM exT_H17 cells transcribe high levels of Il17a as rapidly as 2 hours after challenge with carbapenem resistant Klebsiella pneumoniae infection. Our data are in line with previous studies where CD4 T_RM cells express IFN-γ or IL-4 in their membrane, or quickly release these cytokines after viral, parasite, or house dust mite re-challenge to better control infections or contribute to the pathology associated with asthma (Teijaro, Turner et al. 2011, Iijima and Iwasaki 2014, Hondowicz, An et al. 2016). In addition some of the lung T_RM exT_H17 cells also maintain the flexibility to produce only IFN-γ, the other key cytokine for an efficient response to Klebsiella pneumoniae. The plasticity of T_H17 cells is well characterized (Wei, Wei et al. 2009, Hirota, Duarte et al. 2011, Zielinski, Mele et al. 2012) but for the first time our data show that the capacity to adapt to the environment is maintained also in T_RM exT_H17 cells.

We also hypothesized that the tissue-environment created by the reactivation of T_RM exT_H17 cells commits neutrophils to extravasate and rapidly clear the bacteria. One potential caveat of the functional experiments in which we used diphtheria toxin to delete both T_RM exT_H17 and T_H17 cells is that also IL17A-producing γδ T cells were deleted. However, using knock out mice for γδ T cells we showed that γδ T cells do not appear to play a key role during the memory response, therefore supporting a major role for T_RM exT_H17 and T_H17 cells during this memory phase. In contrast, our data suggest that γδ T cells play a key role during the first response to the infection in a naïve mouse (Figure S2D). Our data therefore support the hypothesis that T_RM T_H17 cells and IL-17A γδ T respond to this pathogen at different times of exposure; however further studies are required to fully dissect the “division of labor” between these two type of IL-17A producer cells during Klebsiella infection. All this is also the case for circulatory Il-17A producer memory cells.

Taken together, our results indicate for the first time that a long-lived exT_H17 T_RM population originates from polarized T_H17 effector cells. This population is maintained by non-stromal lymphatic endothelial cells that produce IL-7. Furthermore, we underline the importance of having an effector derived T_RM cell population as frontline cells that together with other T_RM cells mediate the rapid clearance of threatening strains of antibiotic resistant bacterial infections. We suggest that this cell population plays a central role in the cellular basis of immunological memory.

Finally, our study may hopefully pave the way for future studies in human tissues to design better strategies aimed to deliver more effective vaccines. Hospitalized patients infected with carbapenem-resistant Klebsiella pneumoniae, do not respond to carbapenem antibiotic treatment and thus represent a serious clinical problem. Our data provide basic knowledge that may provide a possible solution to such a problem: vaccination for prophylaxis of carbapenem-resistant Kp especially in patients at risk.

STAR METHODS

CONTACT FOR REAGENT AND RESOURCE SHARING

Further information and requests for resources and reagents should be directed to Maria Carolina Amezcua Vesely (caroamezcuavesely@gmail.com), Nicola Gagliani (n.gagliani@uke.edu) and Richard Flavell (richard.flavell@yale.edu).

Mice

C57BL/6 mice 8–12 weeks of age were obtained from Charles River (USA).

As previously described (1) the Fate⁺ model was obtained by crossing the original Foxp3^RFP IL-10^eGFP IL-17A^Kata mice with IL-17a^CRE+/+ R26YFP^+/+ mice (kindly provided by Prof B. Stockinger). The iFate mice were generated in our Lab (Gagliani, Amezcua Vesely et al. 2015). The iFate mice were used in heterozygosis for the IL-17a^{IRES-eGFP-CRE-ERT2} allele (Gagliani, Amezcua Vesely et al. 2015).

ROSA26iDTR were acquired from the Jackson laboratories (007900) and crossed with IL17^{CRE +/+} FoxP3 RFP^+/+ R26STOP^flox/flox YFP^+/+ to obtain the Fate-DTR mice.

Heterozygous IL7GFP/+ mice (Miller C. N. et al, Int Immunol 2013) were provided Joao Pereira. Animal procedures were approved by the Institutional Animal Care and Use Committee of Yale University

Heat killed Klebsiella pneumoniae

K. pneumoniae (Kp) ATCC strain 43816, serotype 2 was expanded over night at 37 C in nutrient broth. 200 ul of the bacterial suspension were added to 50 ml of nutrient broth and grown for 3.5/4 hr at 37 C, allowing the culture to reach log phase. The concentration of Kp was determined by measuring the absorbance at 600 nm. Bacteria were pelleted by centrifugation at 4,500 rpm for 8 min, washed twice in sterile PBS, and re-suspended in 5 ml of sterile PBS. To heat killed the bacteria we incubated the suspension for 2 hr at 70 C. Then we calculate the amount of protein by Biorad kit (DC protein assay) and keep our stock of 1mg/ml of heat killed Kp at −20 C. Cultures in solid nutrient broth confirm that the bacteria were death.

Immunization and infection with Klebsiella pneumoniae

For immunizations, 40 ug of heat killed Kp (43816) were intranasally administered into anesthetized mice (Ketamine/Xylazine (100 mg/kg and 10 mg/kg respectively)) on day 0 and 7 (Supplementary figure 1A).

For infections Klebsiella pneumoniae BAA-2146 (NDM1+) (ATCC) was expanded over night at 37 C in nutrient broth. 200 ul of the bacterial suspension were added to 50 ml of nutrient broth and grown for 3.5/4 hr at 37 C, allowing the culture to reach log phase. The concentration of Kp was determined by measuring the absorbance at 600 nm (OD₆₀₀= 2 resemble 2×10⁹ bacteria/ml). Bacteria were pelleted by centrifugation at 4,500 rpm for 8 min, washed twice in sterile PBS, and re-suspended at the desired concentration with sterile PBS. 5×10⁵ bacteria (40 ul) were intranasally administered on anesthetized mice. 24 hours after infection mice were sacrificed. The lungs were re-suspended in 500 ul of sterile PBS and homogenized using gentleMACS dissociator. Different dilutions of lungs were plated in agar nutrient broth. CFU numbers were counted 24 hours later (Supplementary figure 3A).

In vivo labeling of circulating CD4 T cells

To distinguish between circulating CD4⁺ T cells and CD4⁺ T cells infiltrating into the lung parenchyma, we injected fluorochrome labeled antibody specific for CD4 (CD4-A700) into the bloodstream of mice. Briefly, naive or immunized mice were injected intravenously with 5 μg Alexa 700 conjugated anti-CD4 Ab (clone RM4–5) in PBS, and after 7 min, mice were euthanized. For in vitro CD4 staining, Pacific Blue conjugated anti-CD4 Ab (clone RM4–4) was used. In all the experiments blood staining controls were used and around 96–99% of CD4⁺ T cells were labeled.

Lymphocyte Preparation and Flow Cytometry

Spleen, mediastinal lymph nodes (medLNs), and bone marrow (BM) were removed from immunized (day 35) or naive euthanized mice, placed into RPMI media supplemented with 5% fetal calf serum (FCS), and passed through a cell strainer (70mm). For isolation of lung infiltrates, cells were collected after 1 hr digestion in RPMI media supplemented with 5% FCS, 1 mg/ml Collagenase D (Roche), and 10 mg/ml DNase I (Sigma) at 37 C. Homogenates were passed through a cell strainer, and ammonium-chloride-potassium lysing buffer was used. Immune cells were separated with a Percoll gradient by centrifugation at 450 g for 20 min. Cells were removed from the interface, washed and stained with the indicated FACS antibodies. Cells were acquired on an LRS II Flow Cytometer (BD Biosciences), and data were analyzed with FlowJo software (Tree Star).

For intracellular IL-17A and IFN-γ staining in Figure 4C, cells were re-stimulated for 3 hr with phorbol 12-myristate 13-acetate (20 ng/ml) and ionomycin (1 mg/ml) in the presence of monensin (1 mg/ml). Then cells were fixed with BD fixation protocol and perm for 4 min with NP-40. Cells were washed and stained with intracellular antibodies describe above.

The gating strategy shown in Supplementary Figure 1 was preceded by gating on the typical physical parameters of the lymphocyte (FSC, SSC), exclusion of the doublets and by gating on CD4 in-vitro⁺ and TCRβ⁺ cells.

The gating strategy shown in Figure 1,4,5 and Supplementary figure 2, 5 and 8 was preceded by gating on the typical physical parameters of the lymphocyte, exclusion of the doublets and by gating on CD4 in-vitro⁺, TCRβ⁺, Foxp3^{RFP neg}.

In all the experiments the gating strategy of every time point was based on negative and single positive controls.

BrdU labeling

For analysis of homeostatic proliferation, Fate⁺ mice were given BrdU (0.8 mg/ml; Sigma-Aldrich)/1% sucrose in the drinking water ad libitum. The water was carefully protected from light and change every other day. After 15 days, organs were collected, processed and cells were stained with surface Abs and incorporated BrdU was detected with a BrdU flow kit (BD Pharmingen) according to the manufacturer’s specifications.

Tamoxifen, FTY720 and Diphtheria toxin treatments

Tamoxifen (Sigma) was dissolved in corn or sesame oil (Fluka, Sigma). Naïve or immunized Inducible IL17A Fate mice were i.p. injected with 2 mg of tamoxifen on day 3, 7 and 10 after immunization.

Fate⁺ mice were immunized with heat killed Kp. 35 days later mice were given 4 μg/ml of FTY720 (Cayman Chemical Company) in the drinking water for two weeks.

For the FTY720 and infections studies, the mice were given FTY720 in the drinking water for 14 days before infection. After the infection the mice kept under FTY720 treatment for 1 more day and then sacrificed (Supp Fig 2H)

Diphtheria toxin was administered i.p. at 50 μg/kg body weight on day 35, 37 and 39 after immunization. On day 45, mice were infected i.n. with Kp BAA-2146 and sacrificed 24 hr after infection.

Antibodies and chemicals administered to the mice

All antibodies for flow cytometry analysis were from Biolegend.

Diphteria toxin, BrdU and Tamoxifen were from Sigma.

See KRT table for the details.

KEY RESOURCES TABLE.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Mouse IgG2a anti-mouse CD45.1 PECy7 clone A20	Biolegend	110730
Rat IgG2a anti- mouse CD19 APC clone 6D5	Biolegend	115512
Armenian Hamster IgG anti-mouse TCR gd APC, clone GL3	Biolegend	118116
Armenian Hamster IgG anti-CD69 PECy7 clone H1.2F3	Biolegend	104511
Armenian Hamster IgG anti-CD69 APC clone H1.2F3	Biolegend	104513
Armenian Hamster IgG anti-mouse CD69 APCCy7 clone H1.2F3	Biolegend	104525
Armenian Hamster IgG anti-mouse TCRb PECy7 clone H57-597	Biolegend	109222
Armenian Hamster IgG anti-mouse TCRb APC clone H57-597	Biolegend	109212
Rat IgG2a anti-mouse CD4 Pacific Blue clone RM4/4	Biolegend	116008
Rat IgG2a anti-mouse CD4 Alexa 700 clone RM4/5	Biolegend	100536
Rat IgG2a anti-mouse CD127PeCy7 clone A7R34	Biolegend	135014
Armenian Hamster IgG anti-mouse CD103 APC clone 2E7	Biolegend	121414
Mouse IgG2a anti-mouse CD45.2 Alexa 700 clone 104	Biolegend	109822
Rat IgG1 anti-mouse IL17a Pacific blue TC11-18H10.1	Biolegend	506925
Rat IgG1 anti mouse IFNg APC clone XMG1.2	BD PHARMIGEN	554413
Rat IgG2a anti-mouse CD122 clone 5H4	Biolegend	105905
Purified anti-mouse CD16/CD32 clone 93	eBioscience	14-0161-85
Rat IgG1 anti-mouse Lyve-1 eFlour 660 clone ALY7	Invitrogen, Thermo Fisher	50-0443-82
Rat IgG2a anti-mouse CD31 Alexa 488 clone 390	Biolegend	102413
Rat IgG2a anti-mouse CD31 Alexa 488 clone MEC13.3	Biolegend	102514
Anti GFP (polyclonal antibody)	invitrogen	A11122
Donkey anti Rabbit IgG Alexa 488	Invitrogen, Thermo Fisher	A21206
Mouse IgG2b Anti mouse/human IL7 clone M25	Bio X cell	BE0048
Mouse IgG2b isotype control, unknown specificity clone MCP-11	Bio X cell	BE0086
Rat IgG2b anti-mouse/human CD44 PeCy7 clone IM7	Biolegend	103030
Rat IgG2a anti-mouse CD62L APC clone MEL-14	Biolegend	104412
Bacterial and Virus Strains
Klebsiella pneumoniae 43816	ATCC	ATCC 43816
Klebsiella pneumoniae BAA-2146 (NDM1+)	ATCC	ATCC BAA-2146
Candida albicans SC5314 strain	Daniel Kaplan	{Igyarto, 2011 #39}
BCG	Lauren Cohn	Yale PCCSM
Chemicals, Peptides, and Recombinant Proteins
Tamoxifen	Sigma-Aldrich	Cat#T5648
FTY720	Cayman Chemichal Company	162359-56-0
Diphtheria toxin	Sigma-Aldrich	D0564
BrdU	Sigma-Aldrich	B5002
Collagenase D	Roche/ Sigma-Aldrich	11088882001
DNase I	Sigma	9003989
Percoll	GE Healthcare	17089101
UltraPure™ BSA (50 mg/mL)	ThermoFisher	AM2616
Ficol 20%	Sigma-Aldrich	F5415-50ML
Sarkosyl	Sigma-Aldrich	L7414-10ML
UltraPure™ 0.5M EDTA	Life Tech	15575-020
Trizma® hydrochloride solution, 2M Tris ph 7,5	Sigma-Aldrich	T2944-100ML
RNAse Inhibitor NxGwn	Lucigen	Lucigen
Maxima H Minus Reverse Transcriptase	ThermoFisher	EP0753
1M DTT (Dithiothreitol)	Teknova	D0060
1H,1H,2H,2H-Perfluorooctan-1-ol	Synquest	Synquest
Critical Commercial Assays
BrdU flow kit	BD Pharmingen	552598
ELISA plates for KP titers	Corning	3590
Nextera XT 96	Illumina	FC-131-1002
NextSeq® 500 High Output v2 Kit (75 cycles)	Illumina	FC 404 2005
HotStart ReadyMix, 500 × 25 μL reactions	KAPA	KM2602
RNeasy Plus Micro Kit	Qiagen	74034
Deposited Data
RNA Sequencing	This paper	GEO: GSE130446
Single cell RNA Sequencing	This Paper	GEO: GSE131450
TCR Sequencing	This Paper	https://clients.adaptivebiotech.com
Experimental Models: Organisms/Strains
Mouse: C57Bl/6	Yale University	N/A
Mouse: Fate+	Gagliani N. et al 2015	N/A
Mouse: iFate	Gagliani N. et al 2015	N/A
Mouse: ROSA26iDTR	The Jackson Laboratories	007900
Mouse: IL7GFP/+	Yale University	N/A
Oligonucleotides
Primer for quantitative RT-PCR: Hprt	Sigma--Aldrich	NM 013556
Primer for quantitative RT-PCR: Klf2) (F, cattgcaactgggaaggatg; R, aaagggtctgtgacctgtgtg)	Sigma--Aldrich
Primer for quantitative RT-PCR: S1pr1 (F, accttccgcaagaacatctc; R, ttgcagcccacatctaacag)	Sigma--Aldrich
DropSeq Barcoded Bead SeqB 5’ –Bead–Linker-TTTTTTTAAGCAGTGGTATCAACGCAGAGTACJJJJJJJJJJJJNNNNNNNNTTTTTTTTTTTTTTTTTTTTTTTTTTTTTT-3’	ChemeGenes, Macosko et al., 2015	Cat# Macosko-2011-10
Template_Switch_Oligo AAGCAGTGGTATCAACGCAGAGTGAATrGrGrG	Macosko et al., 2015	IDT
SMART PCR primer AAGCAGTGGTATCAACGCAGAGT	Macosko et al., 2015	Sigma
New-P5-SMART PCR hybrid oligo AATGATACGGCGACCACCGAGATCTACACGCCTGTCCGCGGAAGCAGTGGTATCAACGCAGAGT* A*C	Macosko et al., 2015	IDT
Custom Read 1 primer GCCTGTCCGCGGAAGCAGTGGTATCAACGCAGAGTAC	Macosko et al., 2015	Sigma
Software and Algorithms
Prism 6.0	GraphPad	https://www.graphpad.com
FlowJo	Tree Star	https://www.flowjo.com/solutions/flowjo/downloads
Leica LAS X software	to be completed this
Ask Jun or who did the analysis to complete this
FastQC	Babraham Institute	https://www.bioinformatics.babraham.ac.uk/projects/fastqc/
Tophat2	Johns Hopkins/University of Washington	Trapnell et. Al. Genome Biology, 2013
Cufflinks	Trapnell Lab	Trapnell et. al. 2012 Nature Protocols
Drop-seq tools	Macosko et al. 2015	http://mccarrolllab.com/dropseq/
R	R foundation for statistical computing	https://www.R-project.org
CummeRbund R Package	MIT/Harvard University	http://compbio.mit.edu/cummeRbund/
Pheatmap R package	Raivo Kolde (CRAN)	https://cran.r-project.org/web/packages/pheatmap/pheatmap.pdf
Rtsne R package	Jesse H. Krijthe	https://github.com/jkrijthe/Rtsne
DBSCAN R package	Michael Hahsler	https://github.com/mhahsler/dbscan
Excel	Microsoft	www.microsoft.com/Microsoft/Excel
Other
PDMS co-flow microfluidic droplet generation device	Nanoshift	Cat# Drop-Seq
100-micron cell strainer	VWR	Cat# 21008-950
40-micron cell strainer	VWR	Cat#21008-949
Fuchs-Rosenthal hemocytometer	Incyto	#DHC-F01
Droplet generation oil	Bio-Rad	Cat# 186-4006

Immunofluorescence staining

Lungs were inflated with a perfusion buffer (1% PFA; 50% OCT; 1x PBS). Incubations at 4°C with sucrose gradients were done sequentially (10% sucrose 1 hour, 20% sucrose 1 hour, 30 % sucrose 1 hour) and then the lungs were flash frozen in Tissue Freezing Medium (Jung). Cryosections (8μm) were mounted on glass slides, fixed with 4% PFA and blocked with αCD32/CD16 (BioLegend) at a dilution of 1/200. Staining was performed for 2h at room temperature for primary antibodies and 1h at room temperature for secondary antibodies. The following antibodies were used: rabbit anti-GFP (Life technologies) was used at 1/100, followed by polyclonal goat anti-rabbit-Alexa Fluor 488 (Life Technologies) at 1/2000 for detecting YFP. Anti-CD4-Pe (BioLegend) and anti-CD31-APC (BioLegend) were used at 1/70.

The images were captured by a confocal microscope at room temperature, using 40x and 10x lenses and were processed using Volocity 6.3 software (Perkin Elmer). For IL7 detection in IL7^GFP/+ mice cryosections (40 μm) were mounted on glass slides, fixed with 4% PFA and blocked with αCD32/CD16 (BioLegend) at a dilution of 1/200. Staining was performed for 2h at room temperature for primary anti-GFP antibody and 1h at room temperature for secondary antibodies. The following antibodies have been used rabbit anti-GFP (Thermo/A11122), and donkey anti rabbit IgG-A488 (Thermo/A21206). The images were captured by a confocal microscope (Leica TCS SP5), using 40x and were processed using Leica LAS X software.

Real time PCR

RNA from cells was isolated using TRIzol LS reagent (Life Technology) according to the manual. RNA was subjected to reverse transcription with SuperScript II (Invitrogen) with oligo(dT) primer. cDNA was semi quantified using commercially available primer/probe sets from Applied Biosystems. Samples were analyzed with the change in cycle threshold method. Results were normalized to hypoxanthine phosphoribosyltransferase (Hprt).

K.pneumonia antibody titers

Flat bottom high binding 96-well plates (Costar) were coated with 10 ug/ml of heat killed Kp in PBS at 4°C overnight, and then washed with PBS-Tween 20 0.05%. Each plate was blocked with PBS plus 1 % BSA (blocking buffer) overnight at 4°C. Serum samples were diluted in blocking buffer and incubated for 2 h at room temperature. Detection of IgM or IgG was done using goat anti-mouse IgM-Biotin or anti-mouse IgG-Biotin (Southern Biotech) for 1 h at room temperature followed by incubation with streptavidin coupled to Horseradish peroxidase (Vector laboratories) for 30 minutes at room temperature. The reaction was developed with 3, 3´, 5, 5´-tetramethylbenzidine substrate-chromogen (BD Bioscience) and stopped with 1N H2SO4. Anti-Kp titers were determined by serial dilutions of plasma and the titer was defined as the highest dilution that yielded an OD at 450 nm that was 2-fold greater than the background level.

Cutaneous Candida albicans infection

C. albicans (kindly provided by J. Craft and D. Kaplan, (Yale University and University of Pittsburg respectively) was grown in YPAD medium at 30° C until the OD₆₀₀ reached 1.5–1.8. After washing with sterile PBS, C. albicans were counted using a hemocytometer.

Mice were anesthetized with ketamine and xylazine (100/10 mg/kg body weight), shaved on the back with electric clipper, then chemically depilated with Nair hair remover lotion (Church & Dwight) for 3 minutes and cleaned. The stratum corneum was removed with 20 strokes with 220 grit sandpaper (3M). 2×10⁸ C. albicans in 50 ul of sterile PBS was applied to the skin. 30 days after infection mice were sacrificed and 1 cm² of infected skin was excised, minced and then digested for 1 h at 37 °C in RPMI medium containing liberase (400 μg/ml; Roche) collagenase D (1 mg/ml; Roche) and DNAse I (10 mg/ml Sigma) with continuous stirring. Cells were separated via a Percoll gradient by centrifugation at 450 g for 20 min. Cells were then removed from the interface, washed, stained with specifics abs and analyzed by FACS.

Bacillus Calmette-Guérin (BCG) infection

Mycobacterium bovis BCG (kindly provided by Dr Lauren Cohn, Yale School of Medicine, Pulmonary, Critical Care and Sleep Medicine) was grown in Middlebrook 7H9 broth base (Difco) supplemented with 10% Bacto Middlebrook OADC enrichment (Remel) for 2 weeks at 37°C. Mice were infected intranasaly with 1×10⁶ viable bacteria in 20ul of PBS or with PBS alone as a control. The final concentration of viable bacteria was enumerated by plate counts of CFU with Middlebrook 7H10 agar (Difco) supplemented with 0.5% glycerol (Difco) and 10% Bacto Middlebrook OADC enrichment.

After 30 days of infection mice were sacrificed, the lungs were harvested and processed as described above for the experiments with K. pneumoniae.

Parabiosis

In short the mice were anesthetized with a mixture of Ketamine/Xylazine (100 mg/kg and 10 mg/kg respectively). After shaving the corresponding lateral aspects of each mouse, matching skin incisions were made from behind the ear to the hip and sutured together with nylon and collagen (5–0,ETHICON, VICRYL*) absorbable suture. Then these areas were clipped with 7-mm stainless-steel wound clips (Reflex7).

RNA sequencing

All cells used for bulk RNA-sequencing were obtained from mice immunized with heat-killed Kp (described above) 35 days post-immunization. Purified lymphocyte populations were obtained from lymph nodes as described above, lymphocytes populations from lung tissue were obtained via collagenase digestion and percoll separation as described above. Purified RNA was obtained using Qiagen Kits.

Libraries for RNA sequencing were generated using Illumina TruSeq mRNA-Seq Sample Prep Kits. Library concentration and quality was assessed using a Bioanalyzer, and sequencing performed on an Illumina HiSeq 2500 at the Yale Center for Genomic Analysis.

Analysis performed using the Tuxedo suite, with splice junction mapping performed using TopHat and transcript quantification using Cufflinks (Trapnell C. et al, Nat. Protoc 2012). Downstream analysis was performed in R using CummeRbund (http://compbio.mit.edu/cummeRbund/) analysis software, heatmaps were produced using the package (https://cran.r-project.org/web/packages/pheatmap/pheatmap.pdf) “pheatmap” in R. Gene ontology analysis was performed using DAVID (Huang da W et al, Nat. Protoc 2009).

Single cell RNA sequencing

Naive or immunized (day 35) C57BL/6 mice were injected intravenously with 2.5 ug Alexa 700 conjugated anti-CD45.2 Ab in PBS, and after 7 min, mice were euthanized. For in vitro CD45.2 staining, PECy7 conjugated anti-CD45.2 Ab was used. For isolation of total lung cells, cells were collected after 1 hr digestion in RPMI media supplemented with 5% FCS, 1 mg/ml Collagenase D (Roche), 10 mg/ml DNase I (Sigma) and 0.1 mg/ml Dispase II (Sigma) at 37 C 250 rpm. Homogenates were passed through a cell strainer, and ammonium-chloride-potassium lysing buffer was used. After in vitro staining with PECy7 anti-CD45.2 and Pacific blue anti-CD4 Abs, CD45.2 in vivo ^neg, CD4 ^neg cells were sorted into PBS plus 0.01% BSA. The cells were diluted to concentration of 100 cells/μl and 1 ml aliquots were used as input to the Drop-seq protocol. Cells were processed for Drop-seq within ∼30 min of collection.

Drop-seq was performed as described previously (Macosko E. Z. et al Cell 2015, Shekhar K. et al, Cell 2016) with minor modifications. The beads were purchased from ChemGenes Corporation, Wilmington MA (catalog number Macosko201110) and the PDMS co-flow microfluidic droplet generation device was generated by Nanoshift, Emeryville CA. The cDNA was generated using Template Switch Oligo and the populations of 5,000 beads (∼150 cells) were separately amplified for 16 cycles of PCR (conditions identical to those previously described) with SMART PCR primer and PCR products were purified by the addition of 0.6x AMPure XP beads (Beckman Coulter). For each conditions the cells from multiple mice were pooled together. Immunized group consisted of cells from 3 mice and naïve condition consisted of 5 mice. Each sample was collected by Drop-seq. For immunized conditions two 1 ml collections were performed and for naïve mouse one collection was performed. For immunized group the cDNA from an estimated 4000 cells was prepared and tagmented by Nextera XT using 1000 pg of cDNA input, and the custom primers New-P5-SMART-hybrid [3] and Nextera XT primers - N701 and N702 (Illumina). For naïve group, cDNA from an estimated 2000 cells was used as input into the Nextera XT tagmentation with New-P5-SMART-hybrid oligoand N704. The concentration and the quality of libraries was measured with Qubit HS DNA assay and Agilent TapeStation Three libraries were sequenced on the Illumina NextSeq 500 using 2.0 pM in a volume of 1.3 ml HT1, and 2 ml of 0.3 μM Custom Read 1 primer (3) for priming of read 1. Read 1 was 20 bp; read 2 (paired end) was 60 bp. Single cell RNA-seq data was processed as described previously (Macosko E. Z. et al Cell 2015) with Drop-seq tools. After the generation of the digital expression matrix, data was normalized in terms of log (CPM / 10 + 1), where CPM stands for counts per million (meaning that the sum of all gene-levels is equal 1,000,000). Subsequently, to visualize the cell subpopulations in two dimensions, PCA (principal component analysis) followed by t-SNE (t-Distributed Stochastic Neighbor Embedding), a non-linear dimension reduction method, were applied on the normalized data. DBSCAN (Density-based spatial clustering of applications with noise) was then used on the t-SNE coordinates to investigate cell subpopulations. Marker genes for each cluster of cells were identified based on binary classifiers trained with Logistic Regression with the R package “glm” (https://www.R-project.org/.). In order to analyze behaviors of specific genes in different clusters and conditions (Fig.4F), fraction of cells expressing the gene of interest and average expression values among expressing cells were calculated.

CDR3 TCRβ sequencing

Spleen and lung were removed from immunized mouse. Tissue resident cells from the lung were isolated by cell sorting CD4^{in-vitro+ in-vivo neg} (YFP^neg, exTh17 and Th17). Memory and naïve CD4 T cells were isolated from spleen of the same mouse by cell sorting.

DNA extraction, amplification and TCRβ-chain sequencing was conducted by Adaptive Biotechnologies based on the ImmunoSEQ platform (Carlson C. S. et al. Nat Commun 2013)

Statistical Analysis

Statistical analyses were calculated in Prism 6.0 (Graphpad Software). Accordingly to the experimental set-up and the distribution of the means (normal versus non parametric) we used ANOVA or Kruskal-Wallis followed by Dunnett and Mann - Whitney U test, respectively, for the multiple comparisons; p <0.05 was considered significant. Bonferroni correction was used to counteract the problem in case of multiple comparisons.

Supplementary Material

Supplemental Figure 1

Figure S1. Related to Figure 1: A) Scheme of immunization with heat killed Klebsiella pneumoniae (serotype K2) on day 0 and 7. At different time points C57BL/6 mice were sacrificed and the presence of CD4 T_RM was evaluated using an in-vivo antibody labeling technique. B) Zebra plots show accumulation of infiltrating/tissue resident CD4 in-vitro⁺ TCR-β⁺, CD4 in-vivo^neg, Foxp3 ^neg T cells in lung after immunization with heat killed K. pneumoniae in C57BL/6 mice at different time points. In-vivo labeling delineates infiltrating/resident (CD4 in-vivo^neg, CD4 in-vitro⁺) and circulating (CD4 in-vivo⁺, CD4 in-vitro⁺) polyclonal lung CD4 T cell subsets. C) CD69 and D) CD103 expression in infiltrating/ T_RM (blue line) and circulatory (red line) CD4⁺ TCR β⁺ T cells. E) Absolute number of lung CD4 infiltrating/resident (CD4 in-vitro⁺, CD4 in-vivo^neg, TCRb⁺) T cells at different time points after immunization (n= 3 per time point). One out of 2 independent experiments is shown. F) The dot plots show the frequency of lung exT_H17 cells in non-immunized (left) and day 14 immunized inducible IL17A Fate mice (right). Both groups of mice were treated with tamoxifen. G) The dot plots show the frequency of lung exT_H17 cells in non-immunized (left) and day 60 immunized inducible IL17A Fate mice (right). Both groups of mice were treated with tamoxifen. The bottom histogram shows the expression of CD69 in exT_H17 cells of immunized + tamoxifen mouse. The bar graphs on the right shows the frequency of exT_H17 cells in the lung of non-immunized and immunized mice (Day 60). Each dot represents one mouse. Representative dot plots are shown (n: 3 mice per group). One out of 3 experiments is shown. Data are represented as mean ± SEM. Mann-Whitney U test, *p=0.0357

Supplemental Figure 2

Figure S2. Related to Figure 1: A) Scheme of immunization with heat killed Klebsiella pneumoniae (K2 serotype) on day 0 and 7 and infection with Cre (carbapenem-resistant enterobacteriaceae) Kp (NDMI⁺) on day 35. 24 hours later the mice where sacrificed. B) The graphs show the amount of Cre Kp (NDMI⁺) CFU after 24 hs of infection of C57BL/6 WT naïve mice (blue circles) or immunized mice (red squares). C and D) The graphs show the amount of Cre Kp (NDMI⁺) CFU after 24 hs of infection. WT, uMT and TCRγδ KO mice were immunized at day 0 and 7 with heat killed Kp serotype 2 (ATCC 43816) and on day 35 infected with live 5×10⁵ CFU of Cre Kp (NDMI⁺ ATCC BAA 2146). Each dot represents one mouse. Data is cumulative of 4 experiments for C, and 2 experiments in D. E) The graphs show the titer of specific antibodies against Kp, IgM and IgG before parabiosis in naïve and day 30 immunized mice. 30 days after parabiosis the titer of specific antibodies against Kp, IgM and IgG were measure in naïve and immunized mice. F) Naïve CD45.1⁺ and day 35 immunized CD45.2⁺ mice were surgically joined for 3 weeks. Before sacrifice anti CD4 antibody was injected to distinguish tissue infiltrating/resident and circulating cells. Dot plots showing blood full chimerism 3 weeks after surgery. G) Dot plots showing the percentage of CD4⁺ CD45.2⁺ cells in the lung of naïve CD45.1 mouse (left) and in the lung of immunized CD45.2⁺ mouse (right). H) The graphs show the amount of Cre Kp (NDMI⁺) CFU after 24 hs of infection. Immunized mice were given FTY720 in the drinking water from day 35 until the day of sacrificed (14 days) (Immunized + FTY720). Naïve and Immunized mice drinking normal water were also infected and used as controls.

Data are represented as mean ± SEM and ± SD.

Mann-Whitney U test (B), Kruskal-Wallis (p = 0.0004) and post-hoc Mann-Whitney U test with Bonferroni correction (C), Kruskal-Wallis (p < 0.0001) and post-hoc Mann- Whitney U test with Bonferroni correction (D). Kruskal-Wallis (p = 0.0134) and post-hoc Mann Whitney U test with Bonferroni correction (H).

Supplemental Figure 3

Figure S3. Related to Figure 1: Fate+ mice where infected in the skin with C. albicans. A, B) Dot plots show the percentage and B) absolute numbers of Th17 and exTh17 after infection at day 7 and day 30. Non infected Fate⁺ mice where used as controls (day 0). C) Histograms show the percentage of skin exTh17 cells that express CD103 (left graph, and CD69 (right graph) at day 7 (blue histogram) and day 30 (red histogram) after infection. Grey histograms represents negative control. D) Dot plots show the frequency of T_RM exT_H17 cells in the lung of naïve (top) or BCG infected Fate⁺ mice after 30 days of infection (bottom). The histogram shows the expression of CD69 in lung exT_H17 cells (blue line) and YFP negative cells (grey line). Representative dot plots are shown (n:3 mice per group).

Supplemental Figure 4

Figure S4. Related to Figure 4: A) Fate+ immunized mice were intra-nasally treated with anti-IL-7 Ab, or with an isotope control on day 35, 37 and 39 after immunization and the mice were sacrificed on day 40. The graph shows the absolute number of spleen (left) and dLN (right) exT_H17 cells in the isotype ctrl treated group (black circles) and in the anti-IL-7 treated group (red circles). Each dot represents one mouse. Two combined experiments are shown. Data are represented as mean ± SEM. Mann-Whitney U Test ns (non significant). B) The Box plots show the specific genes for each cluster, with cluster labels on the x-axis; normalized expression values on the y-axis. The same color codes as in Figure 4F were used. Three markers are shown for each cluster.

Supplemental Figure 5

Figure S5. Related to Figure 5: Immunized Fate-DTR mice were either treated (Tx) with diphtheria toxin (DT) (Figure 5E) or with PBS and were sacrificed 7 days after the last DT treatment. A) Absolute number of Lung T_RM cells (CD4 in vitro⁺, CD4 in vivo^neg TCRβ⁺, FoxP3^neg) and γδT cells. The circles and triangles represent YFP positive (exT_H17, exγδIL17) and YFP negative cells respectively. Blue symbols and green symbols represent immunized PBS Tx or immunized DT Tx mice respectively. Each symbol represents one mouse (Immu n=4; Imm+DT n=6). Showing one out of 2 representative experiments. Data are represented as mean ± SEM. Mann-Whitney U test, ns (non-significant). B) Immunized Fate-DTR mice were either treated (Tx) with diphtheria toxin (DT) on Day 45, 47 and 49 or with PBS and were sacrificed 21 days after the last DT treatment. C) Representative dot plots showing the percentage of YFP negative and YFP positive cells (Gate CD4 in vitro⁺, CD4 in vivo^neg TCRβ⁺, FoxP3^neg) (Imm=4; Imm+DT n=5). Each dot plot represents one mouse. D) Absolute number of YFP⁺ (exT_H17) Lung T_RM cells. Data are represented as mean ± SEM. Mann- Whitney U test. E) Absolute number of YFP^neg Lung T_RM cells. Data are represented as mean ± SEM. Mann- Whitney U test, ns(non-significant).

Table 1:

This table contains the list of the T_RM core genes (Figure 3B) and of T helper core genes related genes (Figure 3D). These lists were used to generate the cluster analysis shown in Figure 3B and 3D respectively.

3B Genes:	exTh17	Th17	YFPNeg	LN_Memory	LN_Naive
Adam4	0	0	0.320789	0.575025	0
Adam8	52.8695	79.9729	27.94	0.673972	0.0200892
Aff3	0.580873	3.71542	5.53656	15.2608	12.2403
Ahr	0.511897	0.153789	0.91247	0.14691	0.109491
Bach2	7.6402	3.55421	9.10501	15.1167	15.8146
Bad	17.214	56.5555	14.1857	35.4653	13.7499
Bag3	11.7366	18.8748	11.8528	6.20421	0.251834
Bax	134.856	186.44	124.816	94.9991	105.179
Pydc4	9.30679	12.8171	31.455	97.7354	167.008
Bcl2	21.678	4.01872	23.4818	13.0457	33.4956
Bcl2l1	63.6311	43.4141	49.2159	11.7069	7.9072
Bcl2l11	22.2084	25.458	43.2721	3.31532	5.27807
Bgn	0.153109	0	0.750593	0.0358505	0.0651365
Btla	5.06223	1.99226	8.02059	30.191	17.5035
Ccl2	8.51415	8.28079	14.359	0	0
Ccl3	7.84485	42.8973	22.7538	6.72655	0
Ccl4	14.1719	52.2169	43.754	23.2497	0.482154
Ccl9	63.8326	164.825	98.0097	0.123146	0.198933
Ccr6	129.722	219.057	73.7499	48.9827	3.1776
Ccr7	87.2397	146.459	180.812	169.882	475.339
Ccr8	261.797	84.2079	257.666	17.7674	10.1766
Ccr9	0.958746	9.66609	12.415	4.53051	2.09285
Cd244	1.8863	5.25569	3.37893	0	0
Cd27	5.82642	0.380267	69.1739	138.294	195.475
Cd36	7.3075	8.63759	10.9358	1.23546	0.00996698
Cd38	10.7758	8.08544	14.6698	24.1699	0.0334365
Cd44	201.492	190.488	153.509	49.3359	4.08609
Cd55	17.2911	29.0783	17.7628	21.3558	17.7869
Cd86	25.0885	77.6358	43.0791	21.0034	3.11128
Cdh1	1.31217	4.00539	3.04901	0.427185	0
Chn2	2.67461	0.312715	5.34619	11.643	2.7623
Cmah	3.73635	1.03427	6.62836	15.0482	17.4606
Col1a2	0.0219386	0	0.0710211	0.032134	0.0069603
Ctla4	368.756	607.455	164.797	14.004	3.93766
Cx3cr1	3.78348	1.17633	10.6619	0.361805	0
Cxcl10	29.6531	61.5554	33.4849	11.5606	0
Cxcl9	0.90074	20.6316	0.36246	0.00499948	0
Cxcr5	8.7829	8.84221	18.2728	76.7941	4.2784
Dapl1	0.932281	5.33548	7.99692	173.396	633.422
Dcn	0	0	0.914805	0.0336434	0
Dock8	21.1722	14.3952	19.1589	26.1727	25.5088
Dtx1	0.695137	0.893031	1.32471	17.8169	29.5047
Egr1	14.9579	10.3611	23.5496	10.9145	16.0509
Elovl7	0.517024	0	0.281797	0.174125	0.954252
Emp1	49.2645	27.1846	26.5431	4.09328	0
Emp2	0.11062	1.05647	0.687951	0.36806	0.0398707
Eomes	0.0681423	0	1.14329	5.34343	0.92676
Fabp4	17.3874	25.1788	15.7998	1.01768	0
Fabp5	84.2297	189.829	73.4023	24.3898	9.41517
Fam65b	18.7707	14.7219	25.0517	46.3984	53.1738
Fcgr2b	20.23	79.6408	48.9514	11.4566	0.0334294
Fgf13	0.455421	0.312654	3.65175	5.58606	8.98842
Fn1	9.05567	8.40408	19.9454	0.0214201	0
Fos	235.1	198.882	230.727	3.95908	2.52353
Fosl2	100.714	81.0534	77.6971	0.352715	1.47108
Fut11	15.3131	3.50319	14.8282	23.3813	25.3917
Gadd45b	122.324	201.893	110.498	11.4633	8.4944
Jun	22.0643	17.863	12.1299	4.80137	6.43026
Gramd4	1.72103	0.67884	4.98049	4.93037	9.64409
Gzmb	18.7029	99.3538	94.3884	1.19967	0.023602
Gzmc	0.282313	3.6255	2.38057	0.748125	0.444278
Gzmf	0	0	0.0301648	0	0
Gzmm	2.87235	0.0843436	1.47468	0.344779	0.238187
Havcr2	7.53343	12.14	14.0544	0.110619	0
Hpgds	2.85422	0.0147194	2.32165	5.22678	2.13475
Hspa1a	2.64218	0.194103	5.92863	0.0135915	0.0727638
Icam2	56.9577	27.1573	76.6911	93.8391	159.506
Icos	1075.2	1560.24	671.817	116.596	27.4734
Ifitm2	350.219	648.966	495.109	55.0214	2.57103
Ifitm3	623.629	1101.36	686.01	35.3957	5.86796
Ifng	32.2278	77.2645	56.581	11.8064	0.0700584
Il21r	163.905	164.941	124.276	30.473	80.1092
Il2ra	27.8064	24.54	12.574	2.35074	0.519221
Il2rb	367.872	287.678	254.038	68.891	7.21834
Il4ra	91.4554	70.6956	70.9335	21.4862	36.4428
Il6ra	1.37339	1.79369	4.38283	8.29798	7.57756
Il7r	335.133	252.975	273.643	199.617	236.145
Inpp4b	45.4048	22.6503	47.6784	66.223	93.3096
Irf4	12.2116	12.439	7.17245	9.9391	0.622295
Itga1	2.73594	2.57965	3.75423	3.95989	3.80478
Itga4	18.595	22.8322	23.6606	24.1741	5.17317
Itga5	2.60504	1.50476	6.23736	0.113039	0.183608
Itgae	8.73839	5.66242	7.85881	2.06011	0.120663
Itgax	8.70462	7.33254	19.2738	0.530394	0.0060763
Junb	519.941	316.791	264.682	8.64708	7.50741
Klf2	56.9603	46.9767	62.6953	50.6301	97.2444
Klf3	10.0573	6.92346	17.4549	20.5219	17.6243
Klre1	0.684562	0.209167	11.1576	0.0199333	0
Klrg1	2.74481	7.64255	14.9285	2.47855	0
Lef1	21.9672	11.2608	57.3993	132.123	569.071
Rorc	19.8199	18.5966	5.64957	3.03501	0.00894821
Litaf	351.864	375.107	301.062	20.0902	14.1632
Lpl	25.9159	36.5378	66.9348	0.148803	0.0903724
Ly6c1	27.1465	39.665	23.3922	60.9024	215.388
Ly6c2	158.341	270.914	237.915	65.2445	11.929
Ly86	87.9592	207.231	119.331	133.772	4.16583
Mmp2	0.746712	0.678074	1.91878	0.00957691	0
Mtor	5.84036	2.30455	4.38287	5.31338	2.17745
Nr4a1	1.62349	0.721332	0.47182	0	0
Nr4a2	37.8142	22.288	39.5706	1.67953	0.261677
Nr4a3	46.3848	57.0615	35.3366	1.70282	2.97483
Otud5	9.44183	9.82215	8.19139	11.4316	14.9012
Pdcd1	300.375	581.258	330.001	95.7609	11.6902
Qpct	76.9008	32.715	45.8005	13.8748	9.32429
Rasgrp2	39.5044	34.499	63.085	39.1931	141.16
Rgs1	665.603	1212.67	515.116	42.5125	3.26151
Rgs2	629.887	980.827	540.658	31.0992	41.3607
Rgs5	0	0	0	0.0561055	0
Rhob	11.313	13.6287	8.61273	0.779562	0.178227
Rorc	1.68187	6.05918	1.6502	0.0995869	0
S1pr1	87.8161	11.8297	110.401	67.4614	158.542
S1pr5	0.637731	3.18697	1.06187	0	0
Samd3	0.0965664	0	1.92481	4.35896	0
Sell	23.8693	37.8471	55.3416	132.099	359.333
Sidt1	2.46918	0.248779	7.23938	12.0723	20.1752
Sik1	31.1668	26.2517	21.7094	0.660721	1.15682
Skil	98.5397	49.9617	75.9529	11.3679	22.9549
Slamf6	8.51641	5.2898	29.9464	109.845	40.485
Sox13	0.108177	0	0.102517	0.193807	0.151553
Tcf7	80.1438	9.87503	125.223	342.105	470.009
Tigit	249.141	295.606	363.296	90.7212	1.56796
Timp3	0.167794	0.0116843	0.704462	0	0
Tlr1	0.715386	2.3093	1.51154	8.08907	6.75228
Tmem123	109.709	137.621	95.2131	54.0762	45.6117
Tnf	16.8345	19.3047	13.2811	5.62487	1.46383
Tnfsf10	3.9295	1.64918	3.09357	5.26483	4.59722
Tnfsf8	3.40005	6.29636	3.5233	2.65015	0.3118
Tnfsf9	7.0645	26.5352	14.5197	9.14896	0.524815
Traf4	12.3083	2.54328	7.55939	1.90512	3.18627
Usp33	2.92806	0.138583	3.50379	7.30786	4.49258
Vps37b	401.929	327.77	339.738	12.2259	22.9063
Xcl1	5.17385	23.1569	55.1631	67.1408	1.80594
Zfp683	1.61235	0.268927	1.46049	0.227627	2.15995
3D Genes:	exTh17	Th17	YFPNeg
Ifng	32.2278	77.2645	56.581
Ccl4	14.1719	52.2169	43.754
Gzmb	18.7029	99.3538	94.3884
Ccr7	87.2397	146.459	180.812
Pglyrp1	34.8062	69.1724	92.365
Sell	23.8693	37.8471	55.3416
Gzma	25.0986	54.0095	137.584
Il4	0.186542	2.89691	8.88272
Gata3	38.1817	16.3644	56.3495
Ccl5	340.647	311.403	3383.38
Il12rb2	2.6791	0.619294	18.42
Slamf6	8.51641	5.2898	29.9464
Eomes	0.0681423	0	1.14329
Axl	9.38972	7.5961	27.1029
Cd27	5.82642	0.380267	69.1739
Il23r	14.8404	7.62962	5.81795
Il7r	335.133	252.975	273.643
Ndfip1	288.928	86.7409	142.731
Rorc	1.68187	6.05918	1.6502
Il17a	111.35	552.221	23.8903
Il17f	318.783	2087.35	73.346
Il1r1	12.759	16.0709	4.58026
Rora	193.351	221.482	106.429
Irf4	12.2116	12.439	7.17245
Rbpj	45.8433	43.616	24.5309