Single-cell perforin and granzyme expression reveals the anatomical localization of effector CD8⁺ T cells in influenza virus-infected mice

Virus Infection.

Specific pathogen-free female BALB/c or C57BL/6 mice (6–8 weeks old) were obtained from the Animal Resources Centre, Perth, Western Australia. Mice were anesthetized with methoxyflurane (Penthrane, Abbott), and BALB/c mice were infected intranasally with 8 × 10⁵ plaque-forming units/50 μl PBS of the reassortant virus Mem71, bearing the hemagglutinin of A/Memphis/1/71 (H3) and the neuraminidase of A/Bellamy/42 (N1) (22). C57BL/6 mice were infected similarly with 10^4.5 plaque-forming units of the H3N2 virus HKx31 (23).

Cell Preparation.

Lung parenchymal cells were obtained by mincing, collagenase/DNase I digestion, and centrifugation over a 20–55% discontinuous Percoll gradient as described (22). Cell suspensions from LN, spleen, and lung were stained for 20 min on ice with rat mAb to CD8α (53-6.72; Becton Dickinson), CD44 (IM7; PharMingen), CD11a (I21/7.7; hybridoma supernatant), and CD62L (Mel-14; hybridoma supernatant), followed by FITC-conjugated rabbit anti-rat Ig (Vector Laboratories) and Tricolor-conjugated streptavidin (Caltag, South San Francisco, CA). Alternatively, cells from C57BL/6 mice were stained with anti-CD8α mAb and a tetrameric complex of H-2D^b molecules with the HKx31 influenza virus nucleoprotein epitope ASNENMETM (NP_366–374) labeled with phycoerythrin (23). Propidium iodide was added at 1 μg/ml immediately before cell sorting. Cell suspensions were separated by using a fluorescence-activated cell sorter (FACSVantage; Becton Dickinson) with lysis ii or cellquest (Becton Dickinson) software after gating on viable lymphocytes based on forward and side scatter and exclusion of propidium iodide-stained cells. Subsets of CD44^high CD62L^low and CD44^low CD62L^high cells were collected while gating on live CD8⁺ cells. CD44^high CD62L^low cells typically constituted 8%, 7%, and 34%, and CD44^low CD62L^high cells were 16%, 7%, and 17% of live MLN, spleen, and lung CD8⁺ cells, respectively (means of three experiments). Cell purities were routinely >95% on reanalysis immediately after sorting.

⁵¹Cr-Release Cytotoxicity Assay.

Target cells of the H-2^d mastocytoma P815 were incubated at 37°C in 5% CO₂ for 1 h in DMEM with 2% FCS and 180 hemagglutinating units of Mem71 virus or with 10% FCS with or without 10 μg of the synthetic peptide TYQRTRALV, representing a major CTL epitope of the Mem71 nucleoprotein NP_147–155 (provided by David Jackson, University of Melbourne), then with 100 μCi of Na⁵¹CrO₄ (Amersham Pharmacia) for 90 min. Labeled cells were cultured at 5 × 10³ per well with serially diluted effector cells in 200-μl volumes in round-bottomed, 96-well plates. In some experiments, effector cells were incubated for 2 h at 37°C with 10 or 100 ng/ml concanamycin A (Sigma) or similarly diluted diluent (DMSO) before addition of target cells. The assay was incubated for 4 h at 37°C, after which supernates were collected for measurement of γ-emission (24). Percent specific lysis was calculated as: 100 × [(mean sample release − mean spontaneous release)/(mean maximal release B mean spontaneous release)].

Quantitative Competitive RT-PCR (QC-PCR).

RNA was isolated from fluorescence-activated cell sorter-purified samples of 10⁴ cells by using RNAzol B total RNA isolating reagent (Tel-Test, Friendswood, TX) according to the manufacturer's instructions, then reverse-transcribed by using AMV reverse transcriptase (Promega) (25). Mixtures of a fixed amount of test cDNA (1% of the sample, from the equivalent of 100 cells) were coamplified with 5-fold serial dilutions of competitor cDNA containing small deletions (25) by using primers and probes described elsewhere (16, 25, 26). After agarose electrophoresis to verify product size, amplified DNA was quantified by ELISA, using hybridization to internal FITC-labeled probes specific for either the exogenous competitor products or the endogenous DNA products. The number of mRNA molecules in the sample was determined from the equivalence point. Sensitivity of detection of endogenous and competitor cDNA was similar. QC-PCR detected 4–20 input competitor molecules by agarose gel electrophoresis and ≈4 molecules by ELISA for the products measured here. Both intra- and interassay variation was low in replicate assays of a given RNA sample, with SD generally <15% of the means.

Single-Cell RT-PCR.

Single cells were placed in 96-well plates by using the automated cell deposition unit attached to a FACS Vantage, lysed with Nonidet P-40, reverse-transcribed, and amplified by two rounds of 40-cycle nested PCR as described (16, 26). Products were agarose-electrophoresed and visualized by ethidium bromide DNA staining. Each PCR run included a set of 10-fold serially diluted cloned cDNAs for each product and a minimum of six negative control samples. The sensitivity of each reaction was typically 2.5 molecules of DNA. Data are presented only for cells that yielded a CD3ɛ PCR product in experiments in which no contamination was detected in any negative control sample.

Induction of Cytolytic Activity in the Respiratory Tract by Influenza Virus Infection.

We previously detected virus-specific cytolytic activity in freshly isolated leukocytes from lung parenchyma and bronchoalveolar lavage, but not MLN, of mice at the peak of the primary response to infection with influenza virus (5, 20). Fig. 1A shows that the cytolytic activity of lung parenchymal cells is mediated by CD8⁺ cells. Freshly isolated CD8⁺ and CD8⁻ T cells from MLN, spleen, and lung parenchyma of 7-day Mem71 influenza virus-infected BALB/c mice were incubated with virus-infected or NP_147–155 peptide-coated syngeneic target cells. Significant specific lysis was observed only in CD8⁺ cells from lung parenchyma. Lysis was mediated mainly by the Ca²⁺-dependent perforin pathway because it was markedly reduced when 2.5–10 mM EGTA was included in the ⁵¹Cr-release assay (data not shown) and when effector cells were preincubated with concanamycin A, a vacuolar-type H⁺-ATPase inhibitor that inhibits perforin-mediated cytotoxicity (27, 28) (Fig. 1B). Further separation of lung CD8⁺ cells showed that lytic activity was enriched in the CD44^high CD62L^low (activated/memory) cell fraction, which could account for all of the activity detected in the unfractionated lung population (Fig. 1C).

Expression of Perforin, Granzyme, and IFN-γ mRNAs by CD8⁺ T Cells from Infected Mice.

QC-PCR was used to measure levels of perforin, granzymes A, B, and C, and IFN-γ mRNA in freshly isolated CD8⁺ cells from MLN, spleen, and lung parenchyma. Kinetic analyses showed that perforin, granzymes A and B, and IFN-γ were expressed in all tissues, peaking at days 5–10 after influenza infection (data not shown). Granzyme C mRNA was not detected at any time in CD8⁺ cells from any of the tissues, although control granzyme C cDNA was detected with similar sensitivity to the other products measured here. Maximum levels of perforin and IFN-γ expression were comparable in all tissues whereas expression of granzymes A and B reached 5- to 60-fold-higher levels in lung than in MLN or spleen.

The relationship between cytolytic mediator expression and activation marker phenotype is shown in Fig. 2 Left. CD8⁺ cells of naive (CD44^low CD62L^high) and activated/memory (CD44^high CD62L^low, termed “activated” for brevity) phenotype were purified from MLN, spleen, and lung of 7-day virus-infected mice or from peripheral LN, spleen, and lung of uninfected control mice and analyzed by QC-PCR. Peripheral LN from control mice were used because MLN were difficult to detect in the absence of infection. Cells of naive phenotype from control mice expressed low or undetectable levels of each mRNA species except CD3ɛ, consistent with evidence that naive CD8⁺ T cells do not express these genes (16, 29, 30). Again, granzyme C was not detected in any population despite detection of control cDNA in the same assay. Expression of each of the other products was generally higher in activated LN and spleen CD8⁺ cells from infected mice than from control mice, but was comparable in activated lung parenchymal cells. Their expression was similar in the activated fraction of all three tissues from infected mice.

Additional differences between control and infected mice were apparent when the impact of infection on cell composition was considered (Fig. 2 Right). The absolute number and frequency of CD8⁺ cells with a naive phenotype were lower in all tissues of infected mice than control mice, whereas the number and frequency of lung cells with an activated phenotype were 4.4- and 2.7-fold higher, respectively. The majority of other CD8⁺ cells in all populations were cells with intermediate CD44 and high CD62L levels (data not shown). Activated cells were about 6-fold more frequent in the lung than in MLN or spleen of infected mice. Multiplication of average mRNA levels per cell (Fig. 2 Left) by cell numbers (Fig. 2 Right) showed that total expression was higher in the activated CD8⁺ population in lung compared with MLN of infected mice (perforin, 5.2-fold; granzyme A, 33-fold; granzyme B, 41-fold; IFN-γ, 41-fold). The data suggest that the difference in lytic activity between MLN, spleen, and lung shown in Fig. 1 can be attributed, at least in part, to the higher frequency of activated cells in lung.

Single-Cell Analysis of Perforin, Granzyme, and IFN-γ mRNA Expression.

We previously showed that two-round nested RT-PCR can detect cytokine, perforin, and granzyme transcripts at high frequency in single T cells after activation in vitro (16, 31, 32). We therefore used nested RT-PCR to determine whether activated CD8⁺ T cells from lymphoid tissues and lung differed in their patterns of expression of perforin, granzymes A–C, and IFN-γ at the single-cell level. CD44^high CD62L^low CD8⁺ cells were isolated from MLN and lung of 7-day virus-infected mice and assayed without restimulation in vitro (Fig. 3). Individual cells from both tissues expressed various combinations of mRNAs, as we have observed previously in CD8⁺ T cells activated in vitro (16). Each product was found both with and without the other product, and many cells expressed granzymes A and/or B without perforin. Frequencies of perforin, granzyme A, granzyme C, and IFN-γ were similar between the two tissues, and only granzyme B was expressed at a significantly higher frequency in lung than in MLN (2.6-fold, P < 0.009 by z test). As in Fig. 2, however, activated cells comprised a smaller fraction of the CD8⁺ population in MLN (8%) than in lung (36%). Although the PCR method used in this analysis did not quantify mRNA levels per cell, the granzyme C PCR products obtained from 6–8% of cells in Fig. 3 (and some cells in Fig. 4) were of low abundance, suggesting that mRNA levels per positive cell were substantially lower for this granzyme than for the other mediators.

The specificity of the cells analyzed in Fig. 3 was not known. We therefore used NP_366–374/H-2D^b tetramers to detect specific T cells in a similar model of infection with HKx31 influenza virus in C57BL/6 mice (23). These tetramers correspond to one of two major CD8⁺ T cell specificities defined in this response (23, 33–35). At days 7 and 8 of infection, as in the Mem71 system used above, freshly isolated lung parenchymal cells displayed strong lytic activity against syngeneic target cells infected with HKx31 virus or loaded with NP_366–374, whereas those from MLN and spleen did not (data not shown). Tetramer-binding CD8⁺ cells were purified from the MLN (0.2–1.4% of CD8⁺ cells in three experiments), spleen (0.5–2%), and lung parenchyma (3–5%) of C57BL/6 mice 8 days after infection and assayed for perforin, granzyme, and IFN-γ expression (Fig. 4). The proportion of cells positive for one or more products was higher among tetramer-binding than tetramer-nonbinding cells in MLN (84% vs. 41%), spleen (84% vs. 11%), and lung (95% vs. 74%). Frequencies of tetramer-binding cells expressing perforin, granzyme B, or IFN-γ were similar in all tissues whereas granzyme A was more frequently expressed in the lung and spleen than in MLN (1.8-fold, P < 0.009 by z test, and 1.5-fold, P < 0.05, respectively). Granzyme C was rarely expressed. By contrast, there were marked differences between the tetramer-nonbinding cells from the three tissues, with 30–50% of lung cells but very few MLN or spleen cells expressing perforin, granzyme B, or IFN-γ. These data suggest that the tetramer-nonbinding cells from MLN and spleen were mainly naive or resting cells whereas most of those from lung were activated. Tetramer-nonbinding lung cells are expected to include cells specific for other influenza virus epitopes (34) as well as bystander cells.

Collectively, these single-cell analyses showed that a high proportion of freshly isolated, activated, or epitope-specific CD8⁺ T cells from all tissues expressed perforin, granzyme A, granzyme B, and/or IFN-γ in various combinations. In addition to increased frequencies of granzyme A or B expression, the main differences noted in the lung compared with MLN were modest increases in the frequencies of positive cells (P = 0.05 for tetramer-binding cells), particularly those expressing multiple products such as the predicted cytolytic phenotype of perforin with granzymes A and/or B (Fig. 5).