PD-1 Expression during Acute Infection Is Repressed through an LSD1–Blimp-1 Axis Free

This article is featured in In This Issue, p.241

Programmed cell death–1 (PD-1) is an immunoinhibitory protein that is expressed on lymphocytes following the engagement of their Ag-specific receptor (1). During the course of an acute infection, PD-1 expression is transient on the surface of CD8 T cells, peaking during the height of an infection and returning to near baseline levels in the resulting memory T cell pool (1–5). During chronic exposure to Ag, such as that from a chronic viral infection or in certain cancers, PD-1 expression is sustained at high levels, and a state of T cell exhaustion is induced, which is characterized by severe curtailment in effector functions, including the ability to proliferate, produce cytokines, and carry out cytotoxic responses (6). T cell exhaustion is in part mediated by signaling through PD-1’s intracellular tyrosine immunoinhibitory domains following PD-1’s surface engagement with its ligands PD-L1 or PD-L2 on target cells (1, 7–9). Although Ab blockade of PD-1/PD-L1 interactions can temporarily reinvigorate immune function from the exhausted state (1), PD-1 remains stably expressed on T cells during a chronic infection (10) and is expressed even upon removal of the cell from a chronic stimulatory environment (11). The stability of PD-1 expression and inhibition of effector functions across generations of cell division suggest that an epigenetic program stably regulates the transcriptional state of the Pdcd1 locus.

PD-1 is encoded by the Pdcd1 gene. In CD8 T cells, Pdcd1 is regulated by the direct actions of transcription factors and epigenetic mechanisms (2, 10, 12). Upon TCR engagement, Pdcd1 is directly activated by a combination of transcription factors (NFATc1 and AP-1) that bind to a series of promoter-proximal elements termed conserved regions B (CR-B) and C (CR-C) (12, 13), respectively. NFAT also binds to sites at −3.7 and +17.1 kb, with respect to the transcription start site (14). Additional transcription factors appear to sustain Pdcd1 during chronic infection and include the binding of NUR77 and FOXN1 to sites located at −23 kb and CR-C, respectively (15–18). Cytokine stimulation that results in STAT3 and STAT4 activation can further induce or sustain Pdcd1 expression in mouse CD8 T cells by binding to the distal −3.7 and +17.1 sites (14). Following the cessation of the TCR signaling (e.g., via Ag/viral clearance), Pdcd1 is silenced through the binding of B lymphocyte–induced maturation protein–1 (Blimp-1) to a region between CR-C and CR-B (19). Blimp-1 binding results in the eviction of NFATc1 from CR-C (19). In addition to these mechanisms, epigenetic regulation through DNA methylation occurs across CR-C and near CR-B in CD8 T cells in both mice and humans (2, 10). In naive CD8 T cells, CpGs in the above DNA regions are consistently methylated. Upon CD8 T cell activation, the methylation is lost in a time course that parallels Pdcd1 expression. In an acute infection setting, the locus is remethylated as the infection is cleared, and PD-1 levels return to the baseline as mentioned above. By contrast, during chronic infection, DNA methylation is permanently lost and is not regained (2, 10).

In a similar manner, the accumulation and removal of activating and repressing histone modifications correlate completely with Pdcd1 expression in mouse CD8 T cells. For example, both H3K27^ac and H3K9^ac activation modifications at CR-B and CR-C correlate with Pdcd1 expression when driven by ex vivo TCR stimulation (19, 20), and H3K4^me1 is enriched when the above stimulation is coupled with IL-6 and IL-12 (STAT3/STAT4) treatment (14). However, as expression wanes during ex vivo stimulation, the repressive modifications H3K9^me3, H3K27^me3, and H4K20^me3 appear at CR-B and CR-C (19). Using the EL4 T cell line, exogenous expression of the transcriptional repressor Blimp-1 induced the appearance of all three of these repressive modifications at the Pdcd1 locus and, subsequently, silenced PD-1 expression (19). Surprisingly, Blimp-1 is also expressed during a chronic infection in exhausted CD8 T cells in which PD-1 levels are at their highest yet fails to repress PD-1 (21). The molecular mechanism for how Blimp-1 could function to repress Pdcd1 exclusively following acute inflammation is not fully clear.

As a repressor, Blimp-1 is known to recruit additional transcriptional repressors that result in silencing the local chromatin environment (22–24). Along with histone deacetylases HDAC1 and HDAC2 and the histone methyltransferase G9a, Blimp-1 can recruit the lysine-specific demethylase 1 (LSD1) encoded by Kdm1a (23–25). LSD1 catalyzes the removal of mono- and dimethylation modifications of H3K4 that are associated with transcriptional activation (26–28), thereby facilitating an epigenetic state of gene silencing. Given the ability of Blimp-1 to repress genes through recruitment of histone modifiers, such as LSD1, we set out to test the hypothesis that LSD1 contributes to the regulation of Pdcd1 in a Blimp-1–dependent manner. Indeed, we found that Blimp-1 was necessary to recruit LSD1 to the Pdcd1 locus and that when bound, LSD1 actively downregulates PD-1 transcription and expression. Furthermore, we found that Blimp-1 was bound to the Pdcd1 locus in both acute and chronic settings; however, LSD1 was only recruited following an acute stimulation, correlating with removal of proximal H3K4^me1/me2 modifications and appearance of a repressive epigenetic profile concurrent with Pdcd1 silencing and DNA remethylation of the locus following acute viral infection. We also found a greater frequency of PD-1–expressing tumor-infiltrating CD8 T cells (TILs) in LSD1-deficient mice compared with LSD1-sufficient mice in a melanoma model. Thus, LSD1 and Blimp1 together are responsible for resetting the epigenetic programming of the Pdcd1 locus to a resting state.

C57BL/6J mice were obtained from The Jackson Laboratory. Kdm1a^fl/fl mice (29) (provided by D. Katz at Emory University) were crossed to mice containing a Granzyme B promoter–driven Cre recombinase (B6; FVB-Tg [GMB-cre] 1Jcb) (provided by J. Jacob at Emory University) (30) and subsequently backcrossed to the C57BL/6 mouse line for four to five generations. Prdm1^fl/fl mice were provided by K. Calame (Columbia University) and similarly crossed to Granzyme B–Cre mice (19). Equal numbers of male and female mice were used in all experiments. The numbers of animals used in each experiment are provided in the figure legends. All experiments were performed in accordance with approved protocols by the Emory University Institutional Animal Care and Use Committee.

Viral stocks of lymphocytic choriomeningitis virus (LCMV) strains Armstrong and Clone-13 were generated, as previously described (31), and kindly provided by Dr. R. Ahmed (Emory University). Mice were infected with 2 × 10⁵ PFU of LCMV Armstrong i.p. or with 2 × 10⁶ PFU of LCMV Clone-13 i.v., as described (32). Viral titers were determined by plaque assay using Vero cells (CCL-81; American Type Culture Collection), as previously described (32, 33).

For analysis or ex vivo cell culture, CD8 T cells were isolated from single-cell splenocyte suspensions using the Miltenyi CD8a⁺ T Cell Isolation Kit (catolog no. 130-104-075; Miltenyi Biotec), according to the manufacture’s protocol. Where indicated, LCMV tetramer-specific cells (see below) were further purified by FACS on a BD FACSAria II (Emory School of Medicine Flow Cytometry Core) following infection, as indicated, and biochemically/molecularly analyzed immediately. For some experiments, isolated cells were cultured ex vivo in RPMI 1640 supplemented with 5% FBS, 5% bovine calf serum, 4.5 g/l glucose, 1.0 mM sodium pyruvate, 10 mM HEPES, and 100 U/ml penicillin/streptomycin. For ex vivo activation, isolated CD8 T cells were stimulated using Dynabeads Mouse T-Activator CD3/CD28 Kit (catolog no. 11453D; Thermo Fisher Scientific), according to the manufacture’s protocol, at a ratio of 2:1 bead/cell for the indicated time (24–96 h).

Cells were stained for flow cytometry in FACS buffer (PBS, 1% BSA, 1 mM EDTA) for 30 min and subsequently fixed using 1% paraformaldehyde for 30 min. Events were collected on a BD LSR II and analyzed using FloJo 9 Software. Abs used to stain cells included the following: CD4 PerCP-Cy5.5 (clone RM4.5), CD8 FITC (clone 53-6.7), CD44 allophycocyanin-Cy7 (clone IM7), CD62L Alexa Fluor 700 (clone MEL-14), CD69 PE-Cy7 (clone H1.2F3), CD127 BV510 (clone SB/199), and PD-1 PE (clone RMP1-30). Biotinylated H-2D^b MHC tetramers specific for LCMV peptides for gp33 var C41M (KAVYNFATM), gp276 (SGVENPGGYCL), and np396 (FQPQNGQFI) were obtained from the National Institutes of Health Tetramer Core Facility at Emory University and subsequently tetramerized to streptavidin-allophycocyanin (Prozyme), following their protocols (tetramers.yerkes.emory.edu).

RNA was isolated from at least three independent preparations of cells using the RNAeasy Kit (Qiagen), and cDNA was prepared from RNA libraries using SuperScript II Reverse Transcriptase (Life Technologies). RT-PCR was used to quantitate mRNA levels in technical duplicates, and values were normalized using 18S rRNA, as previously described (34).

Chromatin immunoprecipitation (ChIP) assays were performed, as previously described (19, 35). Briefly, purified cell populations were crosslinked for 10 min in 1% formaldehyde and subsequently lysed in cell lysis buffer (5 mM PIPES [pH 8.0], 85 mM KCl, 0.5% Nonidet P-40). Chromatin was extracted using nuclei lysis buffer (50 mM TRIS [pH 8.1], 10 mM EDTA, 1% SDS) and sonicated to an average length of 400–600 bp. Chromatin (5 μg) was used for immunoprecipitation reactions on protein A beads with 0.5 μg of polyclonal Abs for H3K4^me1 (catalog no. 07-436; Millipore Sigma), H3K4^me2 (catalog no. 07-030; Millipore Sigma), H3K4^me3 (catalog no. 07-473; Millipore Sigma), H3K27^ac (catalog no. 07-360; Millipore Sigma), IgG (catalog no. 12-370; Millipore Sigma), Blimp-1 (catalog no. 600-401-B52; Rockland Immunochemicals), and LSD1 (catalog no. SC-271720; Santa Cruz Biotechnology). Precipitates were quantitated by quantitative PCR and calculated as a percent of input (Supplemental Table I).

The DNA methylation content of the CR-B associated region was determined by clonal bisulfite sequencing, as previously described (2). Briefly, genomic DNA purified from CD8 T cells and bisulfite converted using the EpiTect Bisulfite Kit, as per the manufacturer’s instructions (Qiagen). Bisulfite-converted DNA was PCR amplified and cloned with the TOPO TA Cloning Kit (Life Technologies). Clones were isolated, the plasmid DNA regions were sequenced, and changes in CpG DNA methylation were determined. Data were aligned in silico using the R/Bioconductor Biostrings package and custom scripts, as previously described (36). A Fisher exact test was used to determine significance.

To quantify the levels of LSD1 protein in CD8 T cells from naive and LCMV-infected mice, CD8 T cells were isolated from single-cell splenocyte suspensions using the Miltenyi CD8a⁺ T cell Isolation Kit (catalog no.130-104-075; Miltenyi Biotec), according the manufacture’s protocol. After isolation, cells were washed in PBS and pelleted by centrifugation at 3000 RPM for 5 min at 4°C. Cells were then lysed in 1× volume of RIPA buffer (20% glycerol, 50 mM Tris [pH 8.0], 150 mM NaCl, 0.1% SDS, 0.5% sodium deoxycholate, 1% Nonidet P-40, and 1 mM PMSF) for 20 min at 4°C. The cells were then pelleted at 15,000 RPM for 10 min at 4°C, and the resulting supernatant representing the protein lysate was stored at −80°C.

For Western blotting, 60 μg of CD8 T cell lysate from naive, day 8 Clone-13–infected and day 8 Armstrong-infected C57BL/6 mice were resolved on a 9% SDS-PAGE gel and then transferred to a PVDF membrane. The membrane was cut in half and incubated overnight at 4°C in TBST (150 mM NaCl, 2 mM Tris [pH 7.4], 2 mM KCl, and 0.1% Tween 20) containing 5% nonfat dry milk with the primary Ab. The primary Abs used were LSD1 (B-9X) (1:75 dilution) (catalog no. sc-271720X; Santa Cruz Biotechnology) and β-actin (AC-15) (1:3000 dilution) (catalog no. sc-69879; Santa Cruz Biotechnology). Membranes were then incubated in TBST containing 5% nonfat dry milk with an HRP-conjugated sheep anti-mouse secondary Ab (1:3300 dilution) (catalog no. A6782; Sigma-Aldrich). To develop, membranes were treated with ECL (catalog no. RPN2106; GE Healthcare Life Sciences) and visualized with the Chemi Hi Resolution setting on a ChemiDoc MP Imagining System (Bio-Rad Laboratories).

Single-cell splenocyte suspensions were prepared from the spleens of day 8 LCMV Armstrong–infected LcKO and wild-type (WT) control mice. A total of 2 × 10⁶ splenocytes were incubated for 5 h at 37°C in the presence of PMA, ionomycin, and brefeldin A (5 μg/ml). After stimulation, cells were stained for surface markers for flow cytometry. The cells were then fixed and permeabilized using the Cytofix/Cytoperm Kit (catalog no. 555028; BD Biosciences), according the manufacturer’s protocol. The Abs used for intracellular cytokine staining were IL-2 PE (clone JES6-5H4), TNF-α FITC (clone MP6-XT22), and IFN-γ allophycocyanin (clone XMG1.2).

As previously shown (1, 2), infection with LCMV Armstrong induces an acute infection that results in CD8 T cells expressing high surface levels of PD-1 on day 5 postinfection, but this ultimately decreases to near-naive T cell levels 8 d following infection when virus is no longer detectable (Fig. 1A). Conversely, infection with the chronic LCMV strain Clone-13 results in PD-1 surface (Fig. 1A) and mRNA levels (2) that remain elevated during the course of the infection. To gain additional insight into epigenetic mechanisms that may be driving the above changes and differences in PD-1 expression between acute and chronic infections, we examined activating and repressive histone modifications at key regulatory elements (37) across the Pdcd1 locus (Fig. 1B). At day 8, splenic CD8 T cells were isolated and analyzed by ChIP. We observed that the −3.7 and 17.1 regions were enriched for the active H3K4 monomethylated histone mark (H3K4^me1) in Ag-specific CD8 T cells of chronically infected mice compared with both acutely infected and naive mice (Fig. 1C). Another active histone mark, H3K4^me2, was also enriched in chronically infected mice at both the CR-B and CR-C regulatory regions of the Pdcd1 locus. H3K27^ac enrichment, a mark associated with active promoters/enhancer regions (38), was higher at all sites in CD8 T cells from Clone-13–infected mice compared with those from Armstrong-infected or naive mice. ChIP for the repressive histone marks H3K9^me2 and H3K27^me3 were enriched at both CR-B and CR-C but only in Armstrong-infected mice (Fig. 1C). Taken together, these data demonstrate that at day 8 postinfection, CD8 T cells from chronically infected mice have active histone marks in the Pdcd1 locus, whereas the acutely infected mice have an enrichment for repressive marks. Moreover, these histone modifications correspond with PD-1 expression on CD8 T cells from chronically and acutely infected mice (Fig. 1A, 1C).

The above data raise the question as to whether active histone marks examined appear at an earlier timepoint during an in vivo response to an Armstrong infection that would correlate with the increased surface expression. To test this, cells were isolated at day 5 following LCMV Armstrong infection, a time point in which PD-1 expression was high (Fig. 1A) and enough activated CD8 T cells (CD44^hi CD62L^low) could be isolated for ChIP. In these experiments, H3K4^me2 and H3K27^ac were enriched over control naive CD8 T cells at all four key regulatory elements, whereas H3K4^me1 was only significantly enriched at −3.7 and 17.1 (Fig. 1D). Thus, active histone modifications at key regulatory elements are associated with PD-1 expression, and during the course of an acute infection, these marks are replaced by repressive histone modifications as expression is silenced. This raises the question of how the dynamics of histone modifications at the Pdcd1 locus are regulated.

LSD1 encoded by Kdm1a is responsible for catalyzing the removal of both H3K4^me2 and H3K4^me1 marks to an unmethylated state, a process that has been dubbed “decommissioning enhancer,” as it can lead to gene silencing (26, 27). To test the hypothesis that LSD1 is actively recruited and responsible for downregulation of Pdcd1 gene expression following an acute infection, we crossed a Kdm1a^fl/fl mouse (29) with the Cre^GranzB mouse (30). The Kdm1a^fl/fl allele allows for Cre-mediated deletion of LSD1’s amine oxidase catalytic domain (29). The Cre^GranzB allele is induced in CD8 T cells upon T cell activation (19, 30), leading to a conditional deletion of LSD1 in activated CD8 T cells. Using the Kdm1a^fl/flCre^GranzB mouse (designated LcKO henceforth), we first determined whether an LSD1 deficiency could alter the kinetics of Pdcd1 expression in an ex vivo activation system. In this study, splenic CD8 T cells from littermate Kdm1a^fl/flCre^– controls (termed WT) and LcKO mice were purified and stimulated in culture using αCD3/CD28 beads over a 4-d period, and expression of Pdcd1 mRNA was measured by quantitative real-time PCR (qRT-PCR) at daily intervals. CD8 T cells from LcKO animals displayed significantly higher levels of Pdcd1 mRNA compared with WT littermate controls at 24 h, the peak of Pdcd1 expression by this method of stimulation (Fig. 2A). Moreover, whereas WT Pdcd1 mRNA levels decayed over the 96-h time course, Pdcd1 mRNA levels remained significantly higher in the LcKO CD8 T cells at all time points after initial stimulation. Additionally, qRT-PCR for Kdm1a mRNA levels showed that the LcKO animals had a significant reduction in LSD1 mRNA compared with WT controls, confirming efficient deletion of the floxed Kdm1a alleles following stimulation (Fig. 2B). The lag phase prior to complete deletion of Kdm1a may account for the slight (although significantly less than in WT) decrease in Pdcd1 expression after peak. Although PD-1 levels were higher at all time points in CD8 T cells from LcKO mice, the rate at which PD-1 decreased was similar between LcKO at WT cells. This could be due to additional mechanisms that have been shown to downregulate PD-1 expression, such as the loss of the PD-1 activator NFATc1 binding to its cognate site in the Pdcd1 locus or the binding of the PD-1 repressor T-bet (18, 19). However, despite the similar rate of PD-1 downregulation, these results demonstrate that LSD1 acts as a negative regulator of PD-1 expression in CD8 T cells following ex vivo stimulation.

To test if PD-1 expression is modulated by LSD1 during an acute viral infection, WT and LcKO mice were infected with LCMV Armstrong, and PD-1 expression on LCMV Ag–specific CD8 T cells was monitored by flow cytometry. At day 8 postinfection, when viral loads have been cleared and WT mice have downregulated surface PD-1 expression on LCMV Ag–specific CD8 T cells (Fig. 3A, blue), LcKO CD8 T cells retained high levels of PD-1 expression (Fig. 3A, red). Previously (19), we found that conditional deletion of Blimp-1, a transcriptional repressor that has been suggested to interact with LSD1 (23), produced a similar phenotype to LcKO mice, which is reconfirmed in this study (Fig. 3A) using Cre^GranzBPrdm1^fl/fl (BcKO) mice. The similar phenotype suggests that these enzymes may operate along the same pathway, in line with the hypothesis that Blimp-1 recruits LSD1 to mediate Pdcd1 repression. By 28 d postinfection with Armstrong, LSD1 knockout Ag–specific CD8 T cells no longer showed an increase in PD-1 expression, unlike cells from Blimp-1 knockout mice that still showed some level of PD-1 expression over their littermate (Cre^–) controls (Fig. 3A). This suggests that although LSD1 is an important determinant in the pathway responsible for decreasing PD-1 expression, it does not account for the entire level of PD-1 repression mediated by Blimp-1.

To determine if LSD1 affected PD-1 expression during a chronic LCMV infection, Ag-specific cells were isolated from WT and LcKO mice following LCMV Clone-13 infection and analyzed by flow cytometry. At day 8, WT and LcKO CD8 T cells expressed moderate levels of PD-1 (Fig. 3B), whereas at day 28 postinfection, no difference in mean fluorescence intensities was observed between WT and LcKO (Fig. 3, bottom). During chronic infection, BcKO CD8 T cells at day 8 had an overall mean fluorescent intensity that was higher, but the histogram patterns overlap with WT cells to a significant degree. Thus, at this early stage, Blimp-1 may not be contributing to Pdcd1 regulation. At the chronic infection time point (day 28), BcKO CD8 T cells expressed slightly lower levels of PD-1, which was in agreement with previous reports (21). The differential roles performed by Blimp-1 and LSD1 at day 28 of a chronic infection may indicate that their activities may be dissociated under these conditions and at these late time points.

Because LcKO cells have increased levels of PD-1 at day 8 after LCMV Armstrong infection, we sought to determine whether LSD1-deficient CD8 T cells displayed markers of an exhausted phenotype. To test this, cytokine expression for these cells was compared with WT. CD8 T cells from LcKO mice produced increased levels of TNF-α and IL-2 but had decreased levels of IFN-γ (Fig. 3C). This demonstrates that, despite the increased levels of PD-1 expressed by LcKO cells, these cells do not appear to be exhausted. To rule out the possibility that the observed increase in PD-1 levels was not due to persistent Ag, we measured the viral titers from LcKO and WT mice at day 8 following Armstrong infection and found that both LSD1-deficient and -sufficient mice cleared the infection by this time point (data not shown). In line with this, there was no difference in the frequency of Ag-specific cells between the LcKO and WT mice (Fig. 3D). Taken together, these results demonstrate that, despite increased levels of PD-1, LSD1-deficient CD8 T cells are functional and capable of clearing virus effectively and that elevated PD-1 levels are not a result of persistent Ag.

To determine if LSD1 interacts directly with the Pdcd1 locus concurrent with PD-1 downregulation, and whether it is recruited there by Blimp-1, ChIP was performed on ex vivo αCD3/CD28 stimulated splenic CD8 T cells isolated from LcKO and BcKO animals as well as the corresponding Cre^– controls (WT). At day 4, when PD-1 expression has nearly returned to baseline levels (Fig. 2A), LSD1 was bound to site 2 [Blimp-1 binding site (39)] and CR-B region in WT mice (Fig. 4A). This indicates a direct interaction with the locus and suggests that LSD1 itself directly represses Pdcd1 transcription. However, LSD-1 binding was not observed in BcKO mice concurrent with a loss of Blimp-1 at the Pdcd1 locus (Fig. 4A, 4B), indicating that Blimp-1 is necessary to recruit LSD1 to the region. Importantly, LSD1 binding was minimally observed at the locus in LcKO cells, indicating that the conditional deletion was efficient (Fig. 4A). However, Blimp-1 was still bound in LcKO mice (Fig. 4B), indicating that Blimp-1 is capable of independently interacting with Pdcd1. These data demonstrate that LSD1 interacts directly with the Pdcd1 locus, corresponding with suppression of Pdcd1 expression ex vivo, and requires Blimp-1.

Blimp-1 binds directly to Pdcd1 following an acute infection and acts as a transcriptional repressor (19). However, during a chronic infection, Blimp-1 is expressed at even higher levels and was correlated with maximal PD-1 expression (21). To determine if Blimp-1 interacted directly with Pdcd1 during chronic inflammation in vivo, ChIP was performed on virus-specific CD8 T cells from naive, day 8 acutely infected (Armstrong), or day 8 chronically infected (Clone-13) mice. As we previously reported, Blimp-1 was bound to Pdcd1 following acute infection (19). LSD1 was also bound at both site-2 and CR-B, correlating with its repressive role (Fig. 5A). Additionally, Blimp-1 was bound at day 8 during a chronic infection (Fig. 5A). Importantly, in this setting, LSD1 was not recruited to the region, despite the fact that Blimp-1 was bound. To rule out the possibility of differential expression levels of LSD-1 between viral infections, we analyzed previously published (40) RNA microarray data and observed no differences (Fig. 5B). In agreement with the RNA data, there was also no observed difference in LSD1 protein level between acute and chronic viral infections at day 8 (Fig. 5C). This suggests that although Blimp-1 interacts with the Pdcd1 locus in CD8 T cells during a chronic infection setting, it fails to downregulate PD-1 levels because LSD1 is not recruited.

As LSD1 is responsible for removing the histone marks H3K4^me1 and H3K4^me2, we set out to determine if LcKO mice possessed increased levels of these marks at the Pdcd1 locus. To test this, LcKO and Cre^– control (WT) mice were infected with LCMV Armstrong for 8 d. Ag-specific (pooled gp33, gp276, np396) CD8 T cells were isolated by FACS and subjected to ChIP for H3K4^me1 and H3K4^me2 at key regulatory regions across the Pdcd1 locus. CD8 T cells from LcKO mice showed enrichment for H3K4^me1 at the −3.7 and +17.1 regions compared with WT cells (Fig. 5D). Enrichment for H3K4^me2 was also observed in the LcKO CD8 T cells at the CR-B, CR-C, and −3.7 regulatory elements. The increased levels of these active histone marks correlated with the elevated levels of PD-1 expressed by LcKO CD8 T cells and mirrors the histone pattern observed in the PD-1–expressing cells at day 8 during chronic infection (Fig. 1C). Together, these results demonstrate that LSD1 is responsible for removing the activating histone marks H3K4^me1 and H3K4^me2 from the Pdcd1 locus and suggest that LSD1 downregulates PD-1 expression by facilitating the removal of active histone marks after being recruited to locus by Blimp-1 during an acute viral infection.

We previously showed that during the course of LCMV Armstrong infection, Ag-specific CD8 T cells initially lose CpG methylation at a region near CR-B (2). As the infection wanes and PD-1 levels return to near baseline, DNA methylation in that region reappears. Thus, the loss of CpG methylation at the Pdcd1 locus is correlated with gene expression, and reappearance of methylation is concurrent with gene silencing. Because de novo DNA methylation is dependent on an H3K4^me0 methylation state (41, 42) and LSD1 has been shown to catalyze the removal of H3K4^me1 and H3K4^me2 (Fig. 5C) as well as be important for global DNA methylation (43), we sought to determine if remethylation of the Pdcd1 was altered in LcKO mice during an acute infection. In accordance with the remethylation observed in WT animals at day 8 after LCMV Armstrong infection (2), WT control mice showed a relative abundance of CpG methylation across the CR-B region of Pdcd1 (Fig. 6). In contrast, LSD1-deficient CD8 T cells, which have elevated levels of PD-1 expression at this time (Fig. 3), showed minimal remethylation of the locus, with a majority of alleles analyzed exhibiting either no methylation or only a handful of methylated CpGs across the region (Fig. 6). This suggests that the inability to remove H3K4 methylation in LSD1-deficient CD8 T cells inhibits the acquisition of DNA methylation, which leads to a failure to fully silence PD-1, thereby providing a further mechanism for retaining prolonged gene expression.

To investigate whether LSD1 deletion had an effect on PD-1 expression and disease progression in another model, we employed the B16 melanoma model (44). LcKO and Cre^– littermate controls were injected with 1 × 10⁶ tumor cells, and mice were checked for weight and tumor size daily. Compared with each other, LcKO and WT mouse weights did not statistically differ throughout the course of the experiment; however, tumor growth appeared to be statistically lower in the mice with LSD-deficient CD8 T cells (Fig. 7A). Examination of the TILs showed that the LcKO mice had a higher frequency of PD-1^hi cells compared with WT control animals, suggesting that LSD1 deficiency plays a role in the expression of PD-1 in a nonviral inflammatory model (Fig. 7B).

Multiple recent studies have highlighted the important effects of dynamic epigenetic regulation in driving immune responses (16, 45–47). In this study, we demonstrate that LSD1 is a novel epigenetic repressor of PD-1 expression. Ex vivo–stimulated LSD1-deficient CD8 T cells displayed increased levels of PD-1 expression compared with WT cells. LCMV-specific CD8 T cells also displayed increased PD-1 expression levels at day 8 following acute infection but not chronic infection. Blimp-1 was identified as necessary to recruit LSD1 to Pdcd1-regulatory regions using LSD1 and Blimp-1 conditional knockout mice. Our results show that LSD1 is capable of binding at the CR-B region as well as the Blimp-1 binding site located 575 bp upstream. These data could suggest that LSD1 is recruited to the locus through multiple factors binding to discrete sites or that LSD1 accumulates across a locus once it is recruited. In both cases, such an event could result in the decommissioning of the active histone marks, as shown in this study for Pdcd1. Although Blimp-1 associates with Pdcd1 in both acute (PD-1 silencing) and chronic (PD-1 permissive) inflammatory environments, LSD-1 is only recruited to the locus in the acute infection, providing a mechanistic epigenetic switch for the differential expression of PD-1 in the two infectious systems. Furthermore, the association of LSD1 with Pdcd1 during resolution of an acute infection correlates with, and is necessary for, the disappearance of the active histone modifications targeted by LSD1: H3K4^me1 and H3K4^me2. Additionally, the actions of LSD1 and the removal of these histone marks correlate with the remethylation of the CR-B upstream, regulatory region, and silencing of PD-1 protein expression.

The results from this work also show that in a melanoma cancer model, the TILs from the LcKO mice displayed higher PD-1 levels compared with WT controls. This matches the observation seen in the acute viral infection in which activated CD8 T cells deficient in LSD1 expressed higher levels of PD-1. Despite the increased presence of PD-1^hi cells, the overall health (as measured by weight loss) of the conditional knockout animals was not significantly different from in the WT mice. However, LcKO mice had smaller tumors across the time course. This could be because of LSD1 controlling other genes that also show enhanced expression when LSD1 is deleted in this model. In line with this, deletion of LSD1 in plasmablasts resulted in the upregulation of 471 genes compared with WT cells (28). Pharmacological inhibition of epigenetic modifiers, including LSD1, is effective in treating some cancers as a method to directly inhibit genes aberrantly expressed in the cancer itself (48). Thus, although the model used in our research does not affect LSD1 within the tumor cells, the higher PD-1 on TILs and potential subsequent cellular exhaustion could complicate this treatment strategy.

Blimp-1, a protein associated with B cell maturation into Ab-secreting plasma cells, is a critical inhibitor of Pdcd1 transcription following acute T cell stimulation (19). Increases in Blimp-1 mRNA and protein following Ag clearance are associated with concurrent decreases in PD-1. Paradoxically, if Ag persists and stimulation through the TCR continues, Blimp-1 levels nonetheless increase further (21). Whereas in an acute infection, the CD8 T cells from BcKO mice showed increased PD-1, suggesting an inhibitory function, and the same deletion resulted in modestly lower levels of PD-1 in a chronic infection, suggesting that Blimp-1 acted as an activator. In this study, Blimp-1 was shown to be bound at Pdcd1 in both infection modalities, thereby precluding the possibility that the changing function of Blimp-1 was mediated indirectly through binding to other target genes. Instead, recruitment of LSD1, which was dependent on the presence of Blimp-1, was found to be unique to acute infections.

Many mechanisms could potentially explain the failure of Blimp-1 to recruit LSD1 in a chronic infection setting. Splice variants of Blimp-1 have been shown both to be involved in different timing of expression (49) and to have alternative functions (50). These include variants that alter Blimp-1’s ability to recruit additional transcription factors while preserving its ability to bind DNA (21, 51). Other biochemical mechanisms could also be involved, including posttranslational modifications of either Blimp-1 or LSD1. Furthermore, other factors binding to locus could sterically hinder or aid LSD1 recruitment. Another mechanism explaining the differential binding of LSD1 during acute and chronic infection could be expression levels of LSD1; however, analysis of published transcriptomics data sets (40) suggest that this is not the case.

For Pdcd1, LSD1 provides an important mechanistic link between the dynamics of histone code modifications, DNA methylation, and gene expression of this critical immune regulatory locus. Dynamic molecular events at the Pdcd1 locus are mirrored in chromatin accessibility patterns and DNA methylation patterns that change as CD8 T cells differentiate from naive to effector, memory, and exhausted cells at Pdcd1 and many other T cell expressed genes (2, 10, 15, 16, 46, 47). Although this study focused on the effects of epigenetics on determining the expression of a single immune-related gene, LSD1 is most likely to have additional consequences on regulating CD8 T cell differentiation and gene expression. Irrespective of LSD1’s additional roles in modulating CD8 T cell gene expression, the work presented in this study establishes a clear mechanism for LSD1 to differentially regulate the expression levels of PD-1 during acute and chronic viral infections.

We thank members of the laboratory for helpful critique and comments during the course of this work. We thank Royce Butler for expert animal husbandry. We thank Dr. Rafi Ahmed for helpful discussions regarding the work and for providing LCMV viral strains, the National Institutes of Health Tetramer Core Facility for providing LCMV-specific tetramers, and also R. Karaffa and K. Fife for cell sorting from the Emory University School of Medicine Flow Cytometry Core.

The authors have no financial conflicts of interest.