C57BL/6J (B6) mice were originally purchased from the Jackson Laboratory; P14 CD45.1 transgenic mouse was a gift from Dr. Lilin Ye (Army Medical University, Chongqin); we obtained the IFN-γ receptor 1 (Ifngr1) knockout mice, which were generated by CRISPR/Cas9-mediated genome engineering, from Cyagen Biosciences (https://www.cyagen.com/cn/zh-cn/sperm-bank-live/15979). The mice were propagated and maintained under specific pathogen-free (SPF) conditions in the animal facility at Shanghai Public Health Clinical Center (SPHCC). Six-week-old female mice were used throughout the study with approval from the Institutional Animal Care and Use Committee (IACUC) of SHPCC.

The A/Shanghai/4664T/2013 (H7N9) virus was from Bio-Safety level 3 laboratory at SHPCC; the A/Chicken/Shanghai/F/98 (H9N2) virus was kindly provided by Dr. Zejun Li (Shanghai Academy of Agricultural Sciences). Recombinant Puerto Rico/8/34 (PR8, H1N1) virus expressing gp33 and LCMV Armstrong strain were kindly provided by Dr. Lilin Ye (Army Medical University, Chongqing). All the used influenza viruses were propagated in 9-day-old embryonated eggs and the 50% tissue culture infectious dose (TCID50) titers of virus stocks were determined in MDCK cells in accordance with the Reed–Muench method; the LCMV Armstrong was grown in BHK21 cells and the titers of virus stock were determined in VERO cells utilizing immune focus assay with anti-LCMV antibody (Clone: M104, Abcam). For infection, virus was adjusted with phosphate-buffered saline (PBS) to a final volume of 50 μl on ice. Two sequential models were employed in the study: in one model, naïve C57BL/6J mice were first intranasally (i.n.) infected with 150 TCID50 of H9N2 and, after an interval as indicated, i.n. exposed to 1×10⁴ TCID50 of H7N9 virus; in the other model, 5×10⁴ P14 CD8+ T cells were isolated from P14 CD45.1 transgenic mouse and seeded into naïve C57BL/6J mice, which were i.n. primed with 50 TCID50 of PR8/PR8-gp33 1 day later and after waiting for 30 days, subjected to intranasal challenge with 2×10⁵ PFU of LCMV Armstrong. The procedures related to H9N2 and H7N9 viruses were conducted in Bio-Safety Level 3 Laboratory while the handlings of other viruses were performed in Bio-Safety Level 2 Laboratory.

For preparation of RNA samples, infected lung tissues were pulverized with a mortar and pestle under liquid nitrogen and total RNA were then purified by Direct-zol RNA MiniPrep Kit (Zymo Research, Irvine, CA, USA) following extraction with RNAzol (Molecular Research Center Inc., USA). The quality of RNA was ensured by measurement on an Agilent Bioanalyzer 2100 using RNA NanoChip. RNA preparations from three lungs isolated at the same sample time point were pooled together for microarray analysis. For transcriptomic analysis of lung TRM cells derived from adoptively transferred P14 cells, lung bronchoalveolar lavage (BAL) of three mice was harvested at each assay time point, from which CD45.1+CD69+CD11a- T_RM cells were sorted through flow cytometry before pooling together (a total of 7,000–20,000 cells) for RNA extraction. Microarray analyses were performed with Affymetrix Clariom S Mouse Gene Expression Chip. Data were log2-transformed and normalized to median. GO and pathway enrichment analyses were carried out using DAVID Bioinformatics Resources. The gene expressions of IFNs and ISGs were further analyzed using Multiple Experiment Viewer (MeV).

The following fluorescence-activated cell sorting (FACS) antibodies were purchased from BD Biosciences, BioLegend, or eBioscience: PB/AF700/Percp-cy5.5 anti-CD3 (145-2C11), PB/FITC/AF488/Percp-cy5.5/APC anti-CD8a (RPA-T8), FITC/PE/APC/Percp-cy5.5 anti-CD45.1 (A20), PE anti-CD69 (H1.2F3), AmCyan/BV510/Percp-cy5.5 anti-CD103 (2E7), BV785 anti-PD1 (29F.1A12), PE/PE-cy7 anti-IFN-γ (XMG1.2), AF488 anti-CD11a (M17/4), APC/PE-cy7 anti-CD62L (MEL-14), FITC anti-CD44 (IM7), BV421/BV510 anti-CXCR3 (CXCR3-173), BV605 anti-KLRG1 (2F1), and BV421 anti-CD127 (A7R34). Tetramer APC NP366–374/H-2D(b) and Live/dead-AmCyan/BV510 were purchased respectively from MEDICAL & BIOLOGICAL LABORATORIES CO., LTD. and Invitrogen. Cell staining was performed essentially as described previously (28) and samples were acquired on the BD LSRFortessa or sorted by BD FACSaria. Data were then analyzed by FlowJo V10.4.2 (TreeStar) or BD FACSDiva™.

Infected lung was lavaged with 1 ml of PBS to collect BAL fluid. The concentrations of cytokines in BAL fluid were determined by bead-based multiplex LEGENDplex™ analysis (LEGENDplex™ Mouse Anti-Virus Response Panel, 13-plex; BioLegend) according to the manufacturer’s protocol. Cytokines measured included IFNα, IFNβ, IFNγ, TNFα, IL-1β, IL-6, IL-10, IL-12p70, GM-CSF, CCL2, CCL5, CXCL1, and CXCL10. The samples were acquired using a BD LSRFortessa™ flow cytometer and data analyses were performed with Legendplex V8.0 software (BioLegend); cytokine concentrations were expressed in pg/ml.

Anti-CD8, anti-IFN-γ, and isotype-matched control antibodies were purchased from Bio X Cell. For CD8+ T-cell depletion, 200 μg of antibody in 100 µl of PBS was delivered into each animal intranasally or intraperitoneally at 1 day before infection and repeated at day 1 and day 3 post infection. IFN-γ neutralization was performed by intranasal inoculation of 200 μg of antibody in 50 μl of PBS at indicated time points after infection.

Mice were deeply anesthetized with single intraperitoneal injection of 1% phenobarbital sodium in PBS. Lungs were rapidly removed, snap frozen in the Optimal Cutting Temperature (OCT) embedding media with liquid nitrogen, and then stored at −80°C until sectioned. Tissues blocks were cut by a CM1950 cryostat (Leica, Wetzlar, Germany) into 14-μm-thick sections, which were subsequently mounted on positively charged slides. The slides were dried for 10 min at −20°C before being stored at −80°C until further processing. RNAscope analyses were performed by using RNAscope^® Multiplex Fluorescent Kit (Advanced Cell Diagnostics, Inc, Newark, CA), following the manufacturer’s instructions. In brief, the sections were baked at 60°C for 30 min, fixed with 10% NBF (neutral buffered formalin solution) at 4°C for 1 h, and then dehydrated with increasing concentrations of ethanol (50%, 70%, and 100% ethanol), followed by protease treatment before incubation with an RNAscope probe specific to mouse IFN-γ mRNA. The sections were imaged on TissueFAXS Confocal 200 (Tissue Gnostics) and analyzed by Strata Quest 6.0X software.

First-strand complementary DNAs (cDNAs) were synthesized from 1 μg of each RNA sample by Reverse Transcription System and quantified by real-time PCR with SYBR Green PCR Mix (Promega) on an Applied Biosystems 7300 RT-PCR cycler. The relative gene expressions were calculated by the delta-delta Ct method with β-Actin used as the reference gene. Absolute quantification of viral load was obtained by extrapolating the Ct value against a copy number standard curve. The primers are shown in the supplementary table .

All statistical analyses were performed using GraphPad Prism 8. Comparison of two groups was performed using Mann–Whitney Unpaired t-test. One-way or two-way analysis of variance (ANOVA) was used to determine the difference between more than two groups, and survival was analyzed by the Kaplan–Meier method with the log-rank test. A p-value <0.05 was considered statistically significant.

Emerging avian-origin H7N9 virus is considered as a severe threat to human health due to its high mortality and pandemic-causing potential. During our course of exploration of vaccination strategy against H7N9 infection, we found in a mouse model that prior infection with low pathogenic, non-disease-causing H9N2 strain could confer a cross-protection against heterologous H7N9 challenge. This finding was made in a study where naïve C57BL/6 mice were first intranasally (i.n.) inoculated with PBS or H9N2 virus and 4 weeks later challenged with a lethal H7N9 infection ( Figure 1A ). In contrast to the PBS-primed control group, which suffered a continuous weight loss, the H9N2-primed mice exhibited only moderate weight loss at the first day post infection (dpi), thereafter stabilizing and progressing toward a full recovery that was attained at 8 dpi ( Figure 1B ). Consequently, all animals in the PBS-primed group had died by 6 dpi, whereas the H9N2-primed group showed 100% survival at the end of the 12-day observation period ( Figure 1C ). Given that the used H9N2 and H7N9 viruses share six internal genes while differing in genes encoding hemagglutinin (HA) and neuraminidase (NA) surface proteins, we speculated that the observed cross-subtype protection is most likely mediated by memory T-cell response targeting conserved internal epitopes.

To facilitate dissection of mechanism mediating the cross-protection of lung CD8+ T_RM cells, we developed a second experimental model combining adoptive cell transfer and sequential infections of two viruses that are of different types while sharing an immunodominant CD8+ T-cell epitope. In this model, as schematically illustrated in Figure 1D , CD45.1+ CD8+ T cells were isolated from P14 mice, which carries an engineered T-cell receptor specifically recognizing the CD8(H-2Db) gp33 epitope of glycoprotein of lymphocytic choriomeningitis virus (LCMV) (amino acids 33–41; KAVYNFATC), and adoptively transferred into naïve C57BL/6 mice; the recipient mice were then primed with unlethal dose (50 TCID50) of A/PR/8 H1N1 (PR8) virus or a recombinant PR8 virus with LCMV gp33 epitope inserted into the neuraminidase glycoprotein (PR8-gp33) and awaited for 30 days for memory induction before challenging with the second infection, 2 × 10⁵ plaque-forming units (PFU) of LCMV Armstrong. The lung viral load at 5 dpi after the LCMV Armstrong challenge was approximately 25-fold lower in PR8-gp33-primed mice than that in PR8-primed mice ( Figure 1E ), in line with improved lung histopathology during the first 48 h after infection, as assessed by H & E staining ( Figure 1F ). Thus, as underscored by the second model, memory CD8 + T cells deposited by primary influenza infection could act alone to protect against subsequent heterologous virus infection through engagement of a single shared epitope.

Next, we investigated the immune mechanism responsible for the H9N2 priming-induced cross-protection against H7N9. To this end, we first assessed the contribution of circulating CD8+ T cells by antibody-mediated depletion ( Figure S1A ). Depletion of circulating CD8+ T cells by intraperitoneal administration of an anti-CD8 antibody compromised but not abolished (still resulting in 100% survival) the cross-protection ( Figures 2A, B ). We also observed that transferring of H9N2-primed serum to naïve mice did not result in significant enhancement in protection against subsequent H7N9 challenge as seen with H7N9 primed serum, which is supposed to enrich for anti-H7 neutralizing antibodies ( Figure 2C ). These results together substantiated the notion that cellular response involving both systematic and resident CD8+ T cells, rather than antibody response, mainly accounts for the H9N2-mediated protection against subsequent H7N9 challenge. Following this, we analyzed the phenotype of influenza-specific memory CD8+ T cells in the bronchoalveolar lavage (BAL) at 28 dpi after H9N2 priming by flow cytometry. The surface marker profiling of NP tetramer-positive BAL CD8+ T cells was characterized by high expression of PD-1, CXCR3, CD44, CD69, and low expression of CCR7, CD25, KLRG-1, and CD127, with only a small portion of cells expressing CD103 ( Figures S2A, B ). This profiling was consistent with lung T_RM phenotype previously reported (19, 29).

We further employed the second mouse model to dissect the contribution of systematic and lung resident memory CD8+ T cells to the cross-protective immunity conferred by prior influenza infection. We first interrogated memory P14 cells by flow cytometry at 30 days after the PR8 or PR8-gp33 priming (just before the LCMV Armstrong challenge). The PR8-gp33-primed mice showed a remarkable enrichment of P14 memory cells in the BAL, accounting for 50% of the total CD8+ population with a total number of approximately 4,000. There was also detectable accumulation of P14 cells in the spleen, albeit lower in frequency and number as compared to those in BAL. In contrast, P14 cells were barely detectable in either compartment in PR8-primed mice ( Figures 2D, E ). The BAL P14 cells were characterized by high expression of CD69, PD-1 and CXCR3 and low expression of CD103 ( Figures S2C, D ), again conforming to the known airway T_RM phenotype.

We then used two approaches to appraise the antiviral activity of PR8-gp33-induced lung memory P14 cells. In the first approach, we transferred BAL from PR8-gp33- or PR8-primed lung to naïve mice, followed by infection with LCMV Armstrong virus and determination of virus load at 5 dpi. The mice receiving PR8-gp33-primed BAL showed significantly better virus control than those receiving PR8-primed BAL, consistent with the idea that PR8-gp33 but not PR8 priming can induce the formation of lung memory P14 cell population, which is recalled by subsequent infection of LCMV Armstrong virus through recognition of the gp33 epitope and consequently afford protection ( Figure 2F ). In the second approach, we depleted circulating CD8+ T cells and lung CD8+ T_RM cells respectively by i.p. and i.n. injection of anti-CD8 antibody ( Figure S1B ), and compared the effects on the control of subsequent LCMV Armstrong challenge. An isotype-matched IgG mAb was used as the control ( Figure 2G ). Depletion of lung CD8+ T_RM cells led to approximately 30-fold increase in viral load, in sharp contrast to no observable effect shown by blockade of circulating CD8+ memory T cells ( Figure 2H ). These results, together with the aforementioned data obtained with the H9N2-primed protection against H7N9, confirmed and extended earlier findings that virus-specific lung T_RM cells, particularly those residing in airway, can effectively provide cross-protection in an antigen-dependent manner.

Previous studies on FRT-resident and skin-resident T_RM cells suggested that a major function of activated T_RM cells is triggering a state of pathogen alert in its residence by secreting IFN-γ (26, 27). A similar mechanism has been proposed for lung T_RM cells with supporting evidence mainly from previous work by McMaster et al., showing that isolated airway T_RM cells were capable of rapidly producing IFN-γ while exhibiting very limited cytolytic activity when stimulated with cognate antigen (20). However, there is lack of in vivo characterization of lung T_RM activity at the early stage of infection.

To explore the mechanistic aspects of the protective action of lung CD8+ T_RM cells, we utilized the second infection model throughout the rest of this study. We first sought to gain a better understanding of the in vivo response of lung CD8+ T_RM cells to antigen re-exposure. We thus isolated lung P14 T_RM cells from BAL at various time points within 0–48 h post LCMV Armstrong infection; RNA samples were prepared and subjected to microarray profiling ( Figure 3A ). Among the detected IFNs, IFN-γ emerges as the only IFN displaying increased mRNA levels during the entire assay period with clear upregulation being observed as early as 3 hpi, which continued to rise until peaking at 24 hpi and then decreased moderately at 48 hpi. The majority, if not all, of type I IFN genes also saw an increase in mRNA level, but it was significantly delayed as compared to IFN-γ, mostly evident at 24 hpi ( Figure 3B ). We further analyzed the mRNA dynamics of ISGs downstream of IFN and observed a substantial divergence in expression patterns. One group of ISGs, as represented by several well-known anti-influenza ISGs (Gbp3, Ifit3, Isg15, and Mx1/2), showed a kinetic pattern of mRNA expression largely resembling that of IFN-γ mRNA, suggesting that they were likely to be directly regulated by IFN-γ in an autocrine manner. These ISGs were also the most highly induced. On the opposite side, there were also ISGs showing only little to moderate induction throughout the assessment period, implicating that they are not downstream targets of IFN-γ-signaling or their activation by IFN-γ in lung CD8+ T_RM cells is suppressed by other signaling(s). Between the two groups is a group characterized by delayed kinetics of induction; we conceived that those ISGs could be downstream of IFN-γ (albeit requiring a stronger IFN-γ stimulation) or of the later produced IFN-α/β. Thus, lung T_RM cells are transcriptionally poised to upregulate IFN-γ rapidly and continually upon sensing the viral antigen, and they might be also committed to induce expression of type I IFNs at specific later time point(s) as the infection continues. On the other hand, their exhibition of strong induction of antiviral ISGs with a temporal pattern paralleling that of IFN-γ suggested that the activity of these cells might be also regulated by self-produced IFN-γ.

Next, we examined the effects of lung CD8+ T_RM cell activation on the surrounding tissues by comparing mRNA expression profiles of PR8-gp33-primed vs. PR8-primed lungs collected at the same time points as described above for lung CD8+ T_RM cell isolation. The gene expression pattern of IFN-γ in PR8-gp33-primed lungs was highly similar to that observed with lung CD8+ T_RM cells, featuring a continuous rise between 3 and 24 hpi and a sign of decline at 48 hpi ( Figure 3C ). Such similarity, together with the fact that PR8-primed lungs showed little induction of IFN-γ mRNA at 12 hpi and 24 hpi relative to 0 hpi ( Figure 3D ), was consistent with the notion that lung CD8+ T_RM cells are the major source of IFN-γ in the lung at the early phase of infection. The analysis of ISGs mRNA expression provided further evidence: for PR8-gp33-primed lungs, a wide array of ISGs were strongly induced by 24 hpi and remained at similar levels at 48 hpi, with the induction of a large portion of them being already discernable at 12 hpi ( Figure 3C ); for PR8-primed lungs, robust induction of a broad spectrum of ISGs was also observed, but it was delayed until 48 hpi ( Figure 3D ). The concomitant earlier induction of IFN-γ and ISGs in PR8-gp33-primed lungs relative to PR8-primed lungs was better visualized in the heatmap constructed using the gene expression ratio between the two lung types ( Figure 3E ). Two differences in the expression profiling of ISGs should be noted between PR8-gp33-primed lungs and lung CD8+ T_RM cells: (1) For the group of ISGs showing high induction in both contexts, lung T_RM cells displayed faster kinetics, which could be explained by the fact that IFN-γ secreted by lung T_RM cells can be immediately captured by these cells while requiring time to reach lung cells at a distance; (2) the spectrum of highly induced ISGs was broader in PR8-gp33-primed lungs than lung CD8+ T_RM cells, indicating that activated lung CD8+ T_RM cells elicit ISG expression program in the lung mainly through acting on lung cells other than on themselves.

We subsequently verified the above microarray data by using quantitative RT-PCR to measure mRNA levels of IFN-γ and four representative antiviral ISGs including Gbp2, Mx1, Oas1a, and Isg15. The measurements of PR8-gp33-primed lung samples showed a course of induction shared by the five mRNAs that was overall consistent with that revealed by microarray: the mRNA induction was readily detectable at 12 hpi, peaking at 24 hpi before declining to some extent at 48 hpi (except for Mx1, whose induction peaked at 48 hpi). By comparison, with PR8 priming, IFN-γ mRNA saw no substantial elevation throughout the 48-h observation period while the degree of the induction of individual ISG mRNA was significantly lower at 12–24 hpi but became comparable or even higher at 48 hpi ( Figures 4A, B ). We further analyzed the expression of IFN-γ mRNA in situ using RNAscope hybridization assay. The enumeration of fluorescent (IFN-γ mRNA positive) dots indicated 24 hpi as the time point when the presence of lung CD8+ T_RM cells resulted in maximal induction of IFN-γ at the transcriptional level, corroborating the microarray and RT-PCR analyses ( Figures 4C, D ). Collectively, these data demonstrated rapid induction of IFN-γ as a main in vivo characteristic of activated lung T_RM cells, which might underlie the ability of these cells to promote an antiviral ISG program in the lung that occurs earlier than during primary response.

Given the possibility that the level of secreted IFN protein might be regulated by post-transcriptional mechanism, we assessed the protein levels of IFN-γ and two type I interferons, IFN-α and IFN-β, in the BALs using cytometric bead array (CBA) assay. The PR8-gp33-primed group started to show enhanced level of BAL IFN-γ protein at 12 hpi as compared to the PR8-primed group; a larger difference between the two groups was observed at 24 hpi and 48 hpi as the BAL IFN-γ protein levels increased progressively in the PR8-gp33-primed group while being largely unchanged in their PR8-primed counterpart ( Figure 5A ). The inconsistency between sustained accumulation of BAL IFN-γ protein and decline of lung-localized IFN-γ mRNA from 24 hpi to 48 hpi in PR8-gp33-primed animal could be reconciled by delayed IFN-γ protein synthesis and/or that IFN-γ protein is more stable than its mRNA. Unlike IFN-γ, type I interferons showed unchanged or rather decreased protein level in the BAL of PR8-gp33-primed animal compared with PR8-primed animals ( Figure 5A ). These data strengthened the notion that lung CD8+ T_RM cells rapidly upregulate IFN-γ mRNA and consequently increase the amount of secreted IFN-γ upon activation.

IFNs are known to have paradoxical functions in the control and pathogenesis of viral infection (29). Besides the well-known viral inhibition activity, they might stimulate production of pro-inflammatory cytokines and their prolonged action could impair tissue recovery after viral clearance (30). Thus, we also assessed the temporal expression of two important inflammatory cytokines, IL-6 and MCP-1, in the BAL using CBA assay. In the absence of lung CD8+ T_RM cells (PR8-primed animal), both cytokines were lowly expressed at 12 hpi and then underwent a sharp rise to reach peak level at 24 hpi, thereafter declining to low level at 48 hpi. Under the action of lung CD8+ T_RM cells (PR8-gp33-primeed animals), the high induction of both cytokines at 24 hpi was pronouncedly suppressed ( Figure 5B ). These data suggested that activated lung CD8+ T_RM cells may facilitate lung recovery after infection by acting to restrain the elevation of proinflammatory cytokines.

As all the above data lend support for the notion that activated lung CD8+ T_RM cells rapidly secrete IFN-γ to induce timely expression of antiviral ISGs in the lung for control of viral infection, we sought to conduct a direct examination. To this end, we adopted a modified version of the second infection model, wherein an IFN-γ-targeting neutralization antibody or an isotype matched control antibody was intranasally administered both on 1 day before and at 6 h after the LCMV Armstrong challenge ( Figure 5C ). The antibody-mediated blockade of IFN-γ had no detectable effect on the number and frequency of P14 cells in the BAL fluid and their surface marker expression ( Figures S3A, B ), but it nearly completely abolished the transcriptional induction of four of the seven representative ISGs examined, namely, Gbp2, Gbp3, Mx1, and Isg15 ( Figure 5D ). For the other three ISGs, including Oas1a, Mx2 and Ifit1, their expression was unaffected by IFN-γ neutralization, indicative of them being possibly induced by other IFNs ( Figure 5E ). Indeed, the BAL levels of IFN-α and IFN-β proteins, as assessed by CBA assay, showed no significant changes upon IFN-γ blockade; thus, they and/or other type I IFNs may account for the IFN-γ-independent ISG induction ( Figure 5F ). Our CBA assessments also found that the expression of both IL-6 and MCP-1, two cytokines suppressed in the presence of activated lung T_RM, was substantially elevated upon IFN-γ blockade, though the time point for most significant upregulation varied ( Figure 5G ). These results solidified the notion that IFN-γ facilitates the early action of lung T_RM in sculpting a potent tissue-wide ISGs program while keeping the inflammatory response under control to minimize collateral tissue damages. They also surprisingly revealed that part of this ISG expression program is independent of IFN-γ, suggesting the potential involvement of type I IFNs.

Finally, we examined the IFN-γ dependence of lung T_RM cells in protection against viral infection. The IFN-γ signaling was blocked in the second infection model using either IFN-γ neutralization antibody as described above, or IFN-γ receptor 1 (Ifngr1) knockout mice. In both contexts, the absence of IFN-γ signaling resulted in a significant, nearly 3-fold increase in lung viral load of LCMV Armstrong, thus validating the mediating role of IFN-γ in the antiviral activity of lung CD8+ T_RM cells ( Figures 5H, I ).