T cells in non-lymphoid tissues give rise to lymph node resident memory T cells

CD69 expression is insufficient to infer residence

CD69⁺ CD8⁺ T cells are abundant in LNs from human cadavers, but scarce in SPF mice. A small frequency of virus-specific memory CD8⁺ T cells express CD69 in mouse LNs after clearance of lymphocytic choriomeningitis virus (LCMV) Armstrong infection, suggesting that SLO Trm cells represent a rare population in mice (Schenkel et al., 2014). The paucity of CD69⁺ CD8⁺ T cells in murine LNs could be due to the fact that mouse LNs were freshly isolated (e.g. human data are an artifact of a cadaverous source), genetic differences between mouse and humans, or the fact that SPF laboratory mice have considerably less microbial experience than humans. To attempt to discriminate between these possibilities, we compared CD69 expression by LN CD8⁺ T cells among SPF inbred C57Bl/6 mice, pet store mice with diverse genetics and microbial experience, and inbred C57Bl/6 mice that acquire microbial experience via co-housing with pet store mice for >2 months, as recently described (Beura et al., 2016).

Figure 1A demonstrates that few CD8⁺ T cells expressed CD69 in SPF C57Bl/6 mouse freshly isolated LNs. In contrast, in pet store mice, CD69 was expressed by about 50% of the CD44^hi (antigen-experienced) population and 30% of total LN CD8⁺ T cells, including those that expressed the LN entry receptor CD62L (Fig. 1A-B). Co-housing resulted in a similar accumulation in C57Bl/6 mice suggesting that the difference in CD69 expression among SPF mouse and human LN CD8⁺ T cells was largely driven by microbial experience rather than genetics or a post mortem artifact.

We next asked the degree to which CD69 expression was unequivocally correlated with residence. Parabiotic surgery involves the conjoining of vasculature between two inbred mice. It is a test of residence, as recirculating cells that migrate via blood will equilibrate between each parabiont, whereas Trm cells will not. 15-23 days after parabiotic surgery of CD45.1⁺ and CD45.2⁺ co-housed C57Bl/6 mice, we observed substantial mixing between LN CD8⁺ T cells (Fig. 1C-D), including among CD69⁺ cells (Fig. 1E). Further characterization revealed that moderate disequilibrium was maintained among CD62L^lo CD69⁺ CD8⁺ T cells, whereas CD62L^hi CD69⁺ CD8⁺ T cells of both CD45.1⁺ or CD45.2⁺ origin was present in similar ratios in each parabiont (Fig. 1E-F). These data suggest that some CD8 T cells may have transiently upregulated CD69 upon migration to LN in co-housed mice. Regardless, these data invite caution in interpreting residence in non-SPF organisms solely based on CD69 expression.

LN Trm cells are broadly distributed after LCMV infection

To assess LN Trm cells that do not recirculate, we turned our attention to an antigen specific model in which SLO Trm cells had previously been defined. To this end, we transferred naïve LCMV-specific CD45.1⁺ P14 TCR transgenic CD8⁺ T cells into naïve SPF C57Bl/6J mice, which were then infected with LCMV (Armstrong strain) i.p. These mice will be referred to as LCMV immune chimeras. Intraperitoneal LCMV infection has broad tropism, but is cleared within approximately one week (Beura et al., 2015). Forty five days later, lymphocytes were isolated from the indicated LNs and LCMV-specific P14 memory CD8⁺ T cells were assessed for expression of CD69 and CD62L (Fig. 2A). P14 memory CD8⁺ T cells isolated from the small intestine epithelium, spleen, and blood are shown for comparison. These data demonstrate that 1) a portion of LCMV-specific memory CD8⁺ T cells within all tested SLOs expressed CD69 long after detectable viral clearance (this was also true among endogenous LCMV-specific CD8⁺ T cells, see Fig. S1), 2) the fraction of CD69⁺ cells varied considerably among SLOs, and 3) CD69⁺ cells typically did not express CD62L in SPF mice (Fig. 2A-B). To interrogate the extent to which SLO P14 cells were stably resident, we performed parabiosis between LCMV immune chimeras containing CD90.1⁺ memory P14 and naïve mice that lacked P14 cells. Lymphocytes were harvested from LNs of both mice 30 days later, and the phenotype of memory CD90.1⁺ P14 cells in both host and naïve partner SLOs was evaluated. While CD69^- cells equilibrated between hosts, CD69⁺CD62L^lo P14 cells were absent from partner mice, indicating that this phenotype defined SLO Trm cells in this model of acute viral infection in SPF mice (Fig. 2C-D). In conclusion, after clearance of LCMV Armstrong infection, CD69⁺ CD8⁺ T cells remained in all SLOs examined, the proportions of CD69⁺ and CD69^- LCMV-specific CD8⁺ T cells varied among different SLOs, and CD69 expression correlated with residence.

Nonlymphoid tissue restimulation amplifies regionalized LN Trm cells

Figure 2 demonstrated that the proportion of Trm cells varied considerably between different SLOs, but gave no clues to their ontogeny or origin. As residence provides a mechanism for compartmentalizing immunity, we asked if local restimulation could result in selective amplification of Trm cells in regional LNs. LCMV immune chimeras were generated, which contain P14 Trm cells within the female reproductive tract (FRT) (Steinert et al., 2015). P14 were restimulated by depositing gp33 peptide into the cervical lumen as described (Schenkel et al., 2013). Ovalbumin-derived SIINFEKL peptide, which does not restimulate P14, served as a negative control. When mice were examined 30-45 days after gp33 restimulation, CD69⁺ Trm cells were significantly increased within the FRT draining (iliac), but not the non-draining (cervical) LNs (Fig. 3A-C). Parabiosis experiments confirmed that these cells were indeed resident (Fig. 3D-F). Importantly, intravenous transfer of recirculating memory CD8⁺ T cells into naïve mice, followed by t.c. antigen challenge, resulted in very few CD69⁺ memory cells within draining LN (see Fig. S2).

We next assessed whether restimulation at an alternative site would also lead to accumulation of LN Trm specifically within regional LNs. We generated CD45.1⁺ OT-I (CD8⁺ TCR transgenic mice specific for ovalbumin-derived SIINFEKL peptide) immune chimeras through immunization with recombinant ovalbumin-expressing vesicular stomatitis virus (VSVova) infection i.v., which results in OT-I Trm cells within skin (Fig. 3G and S3). At least 30 days later, OT-I-activating SIINFEKL peptide was tattooed onto the left flank. Ipsilateral (left, draining) and contralateral (right, nondraining) inguinal LNs were evaluated for accumulation of CD69⁺ LN Trm cells 30-45 days after skin restimulation. Accumulation of CD69⁺CD62L^lo P14 CD8⁺ T cells was highly selective for the left flank draining LN (Fig. 3H-I and data not shown). These data reveal that local restimulation at nonlymphoid sites that contain Trm cells results in a pronounced amplification of Trm cells that patrol the draining LN, but not distant LNs.

Local restimulation increases memory CD8⁺ T cells within all regions of the draining LN

We next tested whether local restimulation impacted the total number of antigen-specific memory CD8⁺ T cells specifically within draining LNs. We first generated CD45.1⁺ OT-I immune chimeras through i.v. immunization with VSVova, and rechallenged the left flank with SIINFEKL (as in Fig. 3G). Thirty days later, we enumerated OT-I cells within contralateral (right) and ipsilateral (left) LNs by quantitative immunofluorescence microscopy (QIM) (Steinert et al., 2015) (Fig. 4A). We found that the relative abundance of OT-I CD8⁺ T cells was significantly increased within the draining LN; the population more than doubled (Fig. 4B). Closer examination revealed that each region of the LN, the follicles, paracortex, sinuses, and interfollicular area, all exhibited a similar increase in the density of memory OT-I CD8⁺ T cells specifically within the draining LN (Fig. 4B-C). These data indicate that the augmentation of SLO Trm cells, via administration of recall antigen to local nonlymphoid tissue, can substantively increase total immunosurveillance within all compartments of regional LNs.

SLO Trm cells share some phenotypic signatures with nonlymphoid Trm cells

The phenotype of Trm cells have been defined in various nonlymphoid organs, but remain much less characterized within SLOs. To address this gap, we compared the phenotype of P14 Trm cells occupying mesenteric LN and the small intestine epithelium (SI-IEL) by flow cytometry in LCMV immune chimeras generated in SPF mice. Tcm cells were also analyzed as a basis of comparison. We found that SLO Trm cells were distinct from SI Trm cells in the expression of CD103, CXCR3, and CD44 (Fig. 5A-B). However, SLO Trm cells shared signatures in common with SI Trm cells (that also distinguished them from Tcm cell), including reduced expression levels of CD122, CD27, and Ly6C. It should be noted that expression levels of Ly6C varies on Trm cells depending on location, and that this marker does not translate to humans or many mouse strains (data not shown).

T-Distributed Stochastic Neighbor Embedding (tSNE) is an algorithm for unsupervised nonlinear dimensionality reduction that compares similarities of data points in high dimensional space, and plots them in two-dimensional space. Cells that are similar cluster closer together. We applied tSNE to our poly-phenotypic flow cytometric analysis of Trm cells from three distinct nonlymphoid organs (SI, FRT, and skin), and the Tcm cells, Tem cells, and SLO Trm cells isolated from each respective NLT draining LN (either mesenteric, iliac, or inguinal LN) on a panel including 11 phenotypic markers: CD103, CD122, Ly6C, CD27, CD62L, CXCR3, CCR9, KLRG1, CX3CR1, CD69 and CD44. For each NLT and associated draining LN, SLO Trm cells clustered close to NLT Trm cells (Fig. 5C). However, they comprised a distinct population. These data further argue against overgeneralizing perceived unifying Trm signatures (Steinert et al., 2015) by demonstrating that Trm phenotype can be quite heterogeneous. For instance, just as CD69 may not predict stable residence (Fig. 1), the absence of CD103 does not predict recirculation (Fig.5A). However, it may be possible to define a truly generalizable signature that reliably predicts stable residence. To contribute to that goal, as well as providing more in depth molecular characterization of SLO Trm cells, we performed transcriptional profiling.

SLO Trm cells expressed some gene sets in common with nonlymphoid Trm cells and others that were shared with Tcm cells

Seventy days after LCMV infection, we sorted CD69^-CD62L^hi P14 Tcm cells and CD69⁺CD62L^lo SLO P14 Trm cells from the spleens (chosen based on technical considerations because spleen contained more total P14 than individual LNs) and CD69⁺C62L^loCD103^lo NLT P14 Trm cells from the FRT of LCMV immune chimeras. Gene expression patterns were analyzed between these three distinct (Tcm, SLO Trm, FRT Trm) memory T cell subsets. We compared our dataset to a previously described transcriptional signature of residence shared by CD8⁺ Trm cells isolated from lung, skin, and the small intestine (Mackay et al., 2013). Due to cross-platform discrepancies between probe sets, we could not validate all 37 reported Trm genes. Nevertheless, and as expected, interrogated gene expression from our FRT Trm analysis was largely in accordance with this previously reported signature (Fig. 6A). SLO Trm cells shared a restricted set of NLT Trm genes, notably those involved in trafficking and differentiation (e.g., S1pr1, S1pr5, Eomes, Itga1, and Xcl1) (Fig. 6A). For some genes, SLO Trm cells more closely aligned with Tcm cells, including low expression of immune checkpoint genes Icos and Ctla4. Independent of the previously reported NLT Trm gene signature, we identified 335 genes which were commonly regulated in SLO and FRT Trm cells (Fig. 6B and Table S1). Enrichment analysis of these shared genes upregulated in both SLO and NLT Trm revealed an activated phenotype, consistent with these cells' putative role as early responders in an anamnestic response (data not shown). Relative expression of selected common Trm genes with known relevance for T cell differentiation, function, and migration are depicted in Fig. 6C.

SLO Trm cells can arise from nonlymphoid T cells

We demonstrated that the local administration of recall antigen resulted in an increase in Trm cells specifically within regional LNs. We thus exploited local NLT Trm cell reactivation as an experimental approach to investigate the ontogeny of SLO Trm. Immunohistochemical analysis revealed that P14 CD8⁺ T cells appeared in the draining afferent lymphatic vessels shortly after local peptide reactivation (Fig. 7A). To test whether cells trafficked from NLT to the draining LN after local restimulation, we crossed OT-I mice to a Kaede transgenic background (Fig. 7B). The Kaede transgene allows transient positional labeling of cells because exposure to 405nm violet light induces irreversible photoconversion of the Kaede from green to red (Tomura et al., 2008). We first generated OT-I-Kaede immune chimeras via VSVova infection, then rechallenged mice in a shaved region of skin with SIINFEKL peptide delivered by tattoo gun (similar to Fig. 3G, see methods). 72 hours later, we exposed a 2mm diameter region of skin on the left flank to 405nm violet light for 5min. 12h later, we assessed skin, ipsilateral (draining) LN, and contralateral (non-draining LN) for the presence of photoconverted OT-I (Kaede-red) cells. Per expectations, a substantive fraction (∼50-75%) of OT-IxKaede memory CD8⁺ T cells isolated from flank skin had photoconverted (Fig. 7B-D). We also observed photoconverted OT-I-Kaede cells in the draining LN, but only after administration of reactivating peptide (Fig. 7B-D). This result indicates that local reactivation in skin induces migration of antigen-specific CD8⁺ T cells from NLT to the draining LN. We obtained similar results after photoconversion and antigen re-challenge in the FRT (see Fig. S4).

We showed previously that i.v. injection of titrated anti-CD90.1 antibody depletes CD90.1⁺ P14 CD8⁺ T cells from blood, spleen, and LNs, while preserving most P14 Trm cells present within the FRT (Schenkel et al., 2013). We exploited this strategy to test whether NLT Trm cells could give rise to SLO Trm cells. First, we generated CD90.1⁺ P14 LCMV immune chimeras. 2 months later, we depleted P14 cells from blood and LNs via injection of 0.75 to 1.5 mg of anti-CD90.1 antibody (Fig. 7E). Mice were then rechallenged transcervically with gp33 peptide, and rested for at least 30 days (46 days shown in Fig. 7F). We found that local peptide challenge induced the accumulation of CD62L^lo CD69⁺ P14 cells within the FRT-draining (iliac) LN. This is consistent with additional parabiosis studies indicating that nonlymphoid memory CD8⁺ T cells, very likely including NLT Trm cells (>99% of P14 CD8⁺ T cells within the FRT are Trm (Steinert et al., 2015)), give rise to a substantive fraction of SLO Trm cells (Fig. S5). For the latter, we surgically joined CD45.1 and CD90.1 P14 immune chimeric mice, confirmed equilibration in blood, then challenged both parabionts with gp33 peptide t.c. (Fig. S5). 30 days later we enumerated the CD69⁺ P14 cells of both donor and host origin in the draining node of each parabiont, which revealed a significant bias for SLO Trm cells to be of host origin, indicating derivation preferentially from pre-existing Trm cells. However, there was a minor contribution by donor cells, perhaps indicating that cells recruited transiently to NLT after local antigen challenge also contribute to the SLO Trm pool. So, these data certainly do not exclude the possibility that recirculating memory CD8⁺ T cells, possibly recruited transiently to NLT after local antigen challenge, could also contribute to the SLO Trm pool.

To confirm that NLT could be the origin of SLO Trm cells, we returned to the OT-I VSVova immune chimera model that exploited skin SIINFEKL tattooing. In this case, we grafted 1cm² of flank skin from a CD90.1⁺ OT-I immune chimera onto the upper left flank of a CD45.1⁺ OT-I immune chimera (Fig. 7G). Thus, the only source of CD90.1⁺ OT-I memory CD8⁺ T cells in recipient host mice is transplanted skin. 30 days after grafting, the grafted region was tattooed with SIINFEKL peptide. 15-20 days later, the draining (left) and contralateral (right) axillary LNs were examined for the presence of donor CD90.1⁺ OT-I cells (Fig. 7H). Skin-derived OT-I cells were observed almost exclusively in draining LN, where many expressed CD69. These data demonstrate that nonlymphoid tissues provide a source for SLO Trm cells.

Adoptive transfers and infections

P14 immune chimeras were generated by transferring 5×10⁴ P14 CD8⁺ T cells into naive C57BL/6J mice followed by infection with 2×10⁵ plaque-forming units (PFU) LCMV Armstrong via intraperitoneal (i.p.) injection one day later. OT-I immune chimeras were generated by transferring 5×10⁴ naive OT-I CD8⁺ T cells into C57BL/6 mice. Mice were infected with 1×10⁶ PFU Vesicular Stomatitis Virus (VSV) expressing chicken ovalbumin (VSVova) via intravenous (i.v.) route.

Intravascular staining, lymphocyte isolation and phenotyping

An intravascular staining method was used to discriminate between cells present in the vasculature from cells in the tissue parenchyma as described previously. Animals were injected i.v. with biotin/fluorochrome-conjugated anti-CD8α through the tail vein. Three minutes' post-injection, animals were sacrificed, and tissues were harvested as described (Anderson et al., 2014). Isolation of cells from skin was done as described (Collins et al., 2016). A 1-2 cm² area of skin was harvested after shaving. The tissue was chopped into small pieces and incubated with RPMI+5% FBS containing collagenase type-III (3mg/ml) and Dnase I (5mg/ml) at 37°C with constant shaking for 1h. After t he 1h incubation, skin pieces were further dissociated using a gentlemacs dissociator (Miltenyi Biotec) and filtered twice through a 70 μm mesh before staining. Where indicated, epidermis was separated from dermis as previously described (Beura et al., 2018). Briefly, epidermal sheets were prepared from flank skin by affixing of the epidermis to slides with an optically clear double-sided adhesive (Thorlabs). Slides were incubated in 2.5mg/ml Dispase II (Roche) in PBS for 1 h at 37 °C after which dermis was physically removed. Lymphocyte isolation from other tissues and SLOs was performed as described (Thompson et al., 2016). Isolated lymphocytes were surface-stained with indicated antibodies. See Key Resources Table for the complete list of antibodies used for flow cytometry staining. The stained samples were acquired using LSRII or LSR Fortessa flow cytometers (BD) and analyzed with FlowJo software (Treestar).

Tissue freezing, immunofluorescence and microscopy

Harvested murine tissues were fixed in 2% paraformaldehyde for 2hrs before being treated with 30% sucrose overnight for cryoprotection. The sucrose treated tissue was embedded in tissue freezing medium OCT and frozen in an isopentane liquid bath. Frozen blocks were cut in a Leica cryostat to prepare 7 μm-thick sections. Slides were stained and immunofluorescence microscopy was performed using a Leica DM6000 B microscope (Schenkel et al., 2013; Steinert et al., 2015). Counterstaining with either 4′,6-Diamidino-2-Phenylindole dihydrochloride (DAPI) (Sigma Aldrich) or sytox green (Thermofisher) was done to detect nuclei. Enumeration of various cell types was performed using ImageJ64 as described before (Schenkel et al., 2013; Steinert et al., 2015) except the QIM procedures are semi-automated. Images underwent gaussian filtering, thresholding and local maxima detection to identify individual cells. Antibodies and chemical used for staining are mentioned in the Key Resources Table.

RNA isolation and microarray hybridization

CD8⁺ CD90.1⁺ P14 Tcm (CD62L^hi CD69^-) and SLO Trm (CD62L^lo CD69⁺) populations were sorted from spleen of LCMV immune chimeras using a BD FACSAria II. For isolation of Trm cells from Female reproductive tract (FRT), tissues were cut into small pieces in RPMI 1640 containing 5% FBS, 2 mM MgCl₂, 2 mM CaCl₂, and 0.5 mg/ml type IV collagenase (Sigma-Aldrich) and incubated for 1 h at 37°C with stirring. After enzymatic digestion, the remaining tissue pieces were mechanically disrupted using a gentleMACs dissociator (Miltenyi Biotec, San Diego, CA). The resulting single-cell suspensions was used as a source for sorting CD69⁺C62L^loCD103^lo Trm cells from FRT. Sorted cells were homogenized using QIAshredder columns (Qiagen) and RNA was extracted using an RNeasy kit (Qiagen) per the manufacturer's instructions. RNA quality was assessed using the Agilent Model 2100 Bioanalyzer. Following quality control, total RNA samples were processed using the GeneChip® 3′ IVT Pico Kit (Affymetrix). Samples were hybridized and loaded onto Affymetrix Genechip Mouse Genome 430 2.0 arrays.

Microarray analysis

All the raw data files obtained by the Affymetrix scanner passed the data quality control steps. The probe data were then preprocessed and normalized with RMA (Robust Multiarray Average) from the “affy” R/Bioconductor package (Gautier et al., 2004). Processed probe expression was then collapsed to gene expression using the Collapse DataSet function in Gene Set Enrichment Analysis (GSEA) with the following settings: collapsing mode for probe sets ≥ 1 gene = Max_probe and Chip platforms = GENE_SYMBOL.chip (Mootha et al., 2003). Prior to performing differential expression analysis, genes with average expression across all samples less than the background expression level (expression < 3.5) were filtered out. Differential gene expression analysis was performed using the LIMMA package in R (Ritchie et al., 2015), and FDR-adjusted p-values were used to select statistically significant genes (FDR cut-off = 0.05). Gene Ontology term enrichment analysis was performed for differentially expressed gene sets using the “GOseq” package in R (Young et al., 2010).

In vivo antibody treatment, local Trm reactivation, and photoconversion

Circulating Thy1.1⁺ P14 CD8⁺ T cells were depleted by injecting 0.75 to 1.5 μg of titrated Thy1.1 (HIS51, eBioscience) antibody i.p. as previously described (Schenkel et al., 2013). For local Trm reactivation experiments, 50μg of the indicated peptides was delivered t.c. as described (Schenkel et al., 2013) in a volume of 30μl by modified gel loading pipet. To reactivate OT-I CD8⁺ Trm cells positioned in skin, a 2cm² area of the flank skin was shaved and 0.5 μg of SIINFEKL peptide was applied using a tattoo gun. PBS was used in control animals. Peptide tattooing was performed on only one side of mouse (ipsilateral) and the contralateral side is treated as a control. For photoconversion of skin Kaede OT-I Trm cells, a reactivated area of skin (5 mm²) was exposed to violet light (405nm ± 5nm) using a Fiber Coupled Violet LED light source (Optogenetics-LED-Violet, Prizmatix) for 5 minutes. The tattooed and photoconversion sites were identical. The surrounding skin was covered with aluminum foil. For photoconversion of cells in the mouse uterus, the violet LED light source was modified to be connected to a 43mm long cannula (with a core diameter 500μm, NA:0.63) using zirconia mating sleeves. The cannula was gently inserted into the vagina of a sedated animal and guided through the cervical canal and stably positioned inside the uterine horn for the duration of photo exposure (5 minutes).

Parabiosis and skin transplant surgeries

Parabiosis surgery was done as described (Steinert et al., 2015). Parabiosed mice were then allowed to rest for 14–30 days before experiments. Equilibration was confirmed in the peripheral blood prior to Trm cell reactivation. Skin transplant was performed as described earlier (Silva and Sundberg, 2013). Briefly, depilated flank skin was harvested from donor CD90.1/OT-I immune chimeric mice. The recipient graft bed was prepared by excising a ∼1 cm² piece of skin from the upper left flank. The donor skin was sutured on to the graft site using silk sutures (Sofsilk, Covidien) and a band aid was used to keep the graft in place. The band aid and sutures were removed 7-10 days' post-surgery and the graft was allowed to heal for at least 30 days before peptide challenge.