The β₂-adrenergic receptor (Adrb2) entrains circadian gene oscillation and CD8⁺ T cell differentiation in response to virus infection

Abstract

Adaptive immune cells are regulated by circadian rhythms both under steady state conditions and during responses to infection. Cytolytic CD8⁺ T cells display variable responses to infection depending upon the time of day of exposure. However, the neuronal signals that entrain these cyclic behaviors remain unknown. Immune cells express various neurotransmitter receptors, and we demonstrate that selective deletion of the β₂-adrenergic receptor (Adrb2) gene perturbs the normal diurnal oscillation of clock gene expression in CD8⁺ T cells, such as Per2 and Bmal1. Consequently, their time-of-day–dependent response to vesicular stomatitis virus was dysregulated, and the diurnal development of CD8⁺ T cells into variegated populations of memory/effectors was altered in the absence of ADRB2 signaling. The diurnal fluctuations in T cell phenotypes were a distinct developmental process, independent of migration kinetics within the spleen. Thus, Adrb2 directly entrains core clock gene oscillation and regulates T cell developmental responses to virus infection as a function of time of day of pathogen exposure.

The β₂-adrenergic receptor regulates circadian gene oscillation and CD8⁺ T cell development to virus infection.

INTRODUCTION

The immune system is tightly regulated by a variety of signals that control responses to foreign pathogens while suppressing autoreactivity. In addition to the well-characterized immune regulators such as antigen (Ag) receptors and cytokines, circadian rhythms (CRs) also play a role in modulating immune responses as a function of light/dark cycles (1–4). However, the molecular pathways that control time-of-day–dependent immune responses are not well understood.

In higher vertebrates, light sensing is a powerful entrainment signal for the central clock within the brain (5). However, time regulation of organ systems outside the brain requires signals delivered to them from the brain’s central clock, the suprachiasmatic nucleus. These signals come in the form of soluble neurotransmitters that are released in a diurnal fashion to regulate peripheral clocks (6). While the oscillatory pattern of core clock gene expression is directly regulated by light inputs within the suprachiasmatic nucleus, peripheral tissue clock oscillation relies upon local and systemic neurotransmitters to provide the circadian timing of expression. There is some evidence to suggest that immune cell functions are regulated by systemic neurotransmitters. For example, lymphocyte recirculation follows a 24-hour periodic ebb and flow through peripheral lymph nodes, and this trafficking was controlled by both norepinephrine (NE) and glucocorticoids (7, 8). A variety of neurotransmitters regulates various aspects of immune cell function, including glucocorticoids, NE, and acetylcholine, all of which are generally immunosuppressive in nature (9–15). These neurotransmitters are released in a diurnal fashion to entrain the periodic oscillation of a myriad of global biological functions including immune cell regulation (8, 16, 17). Furthermore, the core clock genes, such as Bmal1 and Per/Cry, oscillate in virtually all immune cells that have been examined (18–21). However, the signaling pathways that entrain core clock gene oscillation and downstream functionality in immune cells are unknown.

Our recent study measuring transcriptional responses of T cells to viral infection uncovered a central role of the β₂-adrenergic receptor (Adrb2) in regulating multiple cytokine- and T cell effector–pathway genes (22). In addition, we observed that some unexpected pathways were differentially regulated in Adrb2 knockout (KO) CD8 T cells such as serotonin responsiveness, prion diseases, and CRs, which underscore the many aspects of neural regulation of immune function. In this study, we focused on the role of Adrb2 in regulating downstream CRs, and we uncover an important role for Adrb2 in directly entraining circadian gene oscillation as a function of light/dark cycles. The oscillatory nature of CR gene expression correlated with time-of-day developmental variegation of responding T cells as they differentiated in response to virus infection, which was markedly disrupted in the absence of Adrb2. This study demonstrates a direct role for Adrb2 in temporally regulating CR gene oscillation and functional responses in cytolytic CD8⁺ T cells.

RESULTS

Deletion of Adrb2 alters CR gene expression in CD8⁺ T cells responding to virus infection in vivo

CD8⁺ T cells express ADRB2, and we and others found that signaling through this receptor by NE or other β-agonists suppresses cytokine secretion and lytic activity in vitro and in vivo (13, 23, 24). However, the role of adrenergic signaling in regulating peripheral T cell development during infections has not been examined until recently. To this end, our recent study demonstrated a key role for Adrb2 in regulating downstream transcriptional pathways engaged in CD8⁺ T cells as they divide and respond to endogenous virus infection (22). This study revealed unique pathways regulated by adrenergic signaling in T cells, and in the current study, we focused our analysis on the CR pathway. In this experiment, published by Estrada et al. (22), purified CD8⁺ Clone4 [hemagglutinin Ag (HA)–specific TCR-Tg T cells] wild-type (WT) and Clone4 × Adrb2^−/− T cells were cotransferred to WT Balb/c recipients. The next day, animals were either injected with phosphate-buffered saline or infected with a sublethal dose of vesicular stomatitis virus expressing influenza HA (VSV-HA). Cells were purified from each of three recipient animals at each time point, and RNA sequencing analysis was performed on highly purified Clone4 CD8⁺ T cells [primary data available in GEO (Gene Expression Omnibus) at accession number GSE102478]. A focused analysis of CR pathway genes from this dataset was performed and is presented in Fig. 1.

Fig. 1. Differential core clock gene expression in Adrb2^−/− CD8⁺ T cells in response to virus infection.

Data are derived from Estrada et al. (22) comparing gene expression values [FPKM (fragments per kilobase of transcript per million mapped reads)] from adoptively transferred WT and Adrb2^−/− Clone4 CD8⁺ T cells purified from VSV-HA–infected hosts at incremental days postinfection (DPI). (A) Heatmap analysis of KEGG-annotated CR pathway genes expressed in CD8⁺ T cells at the indicated days postinfection. (B to D) Expression of Clock, Per2, and Cry2 in WT versus Adrb2^−/− CD8⁺ T cells at the indicated days postinfection. *P < 0.05 by EdgeR analysis.

In our previous report (22), EdgeR analysis (25) identified more than 320 genes that were differentially expressed between WT and Adrb2^−/− cells at any of the time points, including day 0 (d0). A comprehensive integrative pathway analysis of these constituent genes identified a variety of known effector- and cytokine-regulated pathways differentially regulated in Adrb2^−/− cells. One pathway that stood out was the Kyoto Encyclopedia of Genes and Genome (KEGG) CR pathway, which was significantly regulated at multiple time points of infection. From that dataset, an expression heatmap of known core CR genes is shown in Fig. 1A. We observed significant differential expression of the key circadian genes Per2 and Cry2 (Fig. 1, C and D), along with other core CR genes Cry1, Arntl2, Per3, and Fbxl3 (fig. S1) at various time points of infection (statistical analysis presented in fig. S2). In contrast, other CR genes such as Clock (Fig. 1B), Per1, and Dbp (fig. S1) were not differentially expressed at any time point. This experiment was a cotransfer protocol, where both WT and Adrb2^−/− responded to the virus infection simultaneously within the same WT hosts. Although the cells were harvested at different days postinfection, collection and purification of the cells were performed at ~2 hours into the light cycle [Zeitgeber 2 (ZT2), by convention] on each day. Thus, these data suggest that Adrb2 regulates the expression of some core CR genes in virus-specific CD8⁺ T cells both at the baseline and at incremental times after infection.

In addition to a marked dysregulation of a subset of CR genes in Adrb2^−/− cells, we also observed that many of the core CR genes, such as Arntl, Rora, Nfil3, Nr1d1, Nr1d2, Bhlhe40, and Dbp (fig. S1), were differentially expressed in CD8⁺ T cells on incremental days following infection. For example, Bmal1 (Arntl) was significantly suppressed at d4 postinfection and markedly increased 100-fold by d7 (fig. S2C). In contrast, Rora steadily increased through d12 compared to baseline levels before infection. These data indicate an unexpected regulation of core CR genes in CD8⁺ T cells as they divide and differentiate in response to virus infection. While this experiment was not designed to specifically measure circadian gene expression or oscillation, the dynamic dysregulation of many of the core CR genes suggests a pivotal regulation of this pathway by Adrb2. On the basis of these observations, we wished to determine whether Adrb2 directly regulated CR gene oscillation through a 24-hour 12:12 light/dark cycle.

Adrb2 controls diurnal oscillation of core CR genes and other immune-related genes in CD8⁺ T cells

To create a more physiological model for studying time-of-day–dependent T cell responses, we crossed the Adrb2-fx mouse (on C57Bl/6 background) (26) with CD8α-Cre (27), which expresses Cre late in CD8⁺ T cell development in the thymus (27). This model selectively deletes floxed targets (such as Adrb2-fx) in peripheral CD8⁺ T cells, preserving ADRB2 signaling in other cell types and tissues. Thus, the upstream oscillation of NE and the downstream response to NE remain intact in all cells except peripheral CD8⁺ T cells. To determine whether Adrb2 directly regulated core clock gene expression in CD8⁺ T cells, we bred Adrb2-fx × CD8α-Cre to the Per2::Luc mouse, which reports on PER2 protein within cells and tissues (28), facilitating the measurement of PER2 oscillation.

To assess the relative temporal expression of PER2 protein, cohorts of Per2::Luc × Adrb2^fx/fx and Adrb2^fx/fx × CD8a-Cre⁺ mice were acclimated for 3 to 5 weeks to 12:12 light/dark cycles in circadian cabinets that were staged 6 hours apart. By convention, ZT0 initiates the beginning of the light cycle. After acclimation, animals were euthanized at ZT4, ZT10, ZT16, and ZT20 corresponding to their individual cabinet time zone. Single-cell suspensions were prepared from the spleen, and CD8⁺ T cells were isolated by bead-based purification. Luciferase activity was measured from purified CD8⁺ T cells. In all measurements of oscillation, we used the CircaCompare algorithm (29) to assess whether a temporal data series was rhythmic through a 24-hour period and whether there were significant statistical differences in rhythmicity between WT and Adrb2^fx/fx × CD8a-Cre⁺ genotypes. We observed significant PER2 protein oscillation in WT CD8⁺ T cells that was phase shifted in Adrb2^fx/fx × CD8a-Cre⁺ cells (Fig. 2A). In addition to measuring PER2 protein, total RNA was isolated from purified CD8⁺ T cells, and core CR and other select transcription factor gene expression was quantified by quantitative polymerase chain reaction (qPCR). As expected, Clock was uniformly expressed at all time points and was not affected by the deletion of Adrb2 (Fig. 2E). However, in contrast to the effects on PER2 protein oscillation, we found that ablation of Adrb2 completely disrupted significant oscillation of Per2 and Per3 mRNAs with no impact on Per1 (Fig. 2, B to D). In contrast to the Per genes, Cry1 and Cry2 gene oscillation amplitude was low in WT cells, with Cry2 rhythmicity disrupted in the absence of Adrb2 (fig. S3, A and B).

Fig. 2. Adrb2 regulates periodic core clock gene oscillation in CD8⁺ T cells.

Eight- to 10-week-old WT Adrb2^fx/fx × Per2::Luc (blue circles) and Adrb2^fx/fx × CD8a-Cre⁺ × Per2::Luc (red squares) mice were acclimated for 3 weeks in light chambers to a standard 12:12 light/dark cycle staged 6 hours apart. CD8⁺ T cells were isolated from spleens at ZT4, ZT10, ZT16, and ZT22. (A) Per2 protein luciferase activity was measured by a standard luminescence assay. (B to J) Total RNA was extracted from CD8⁺ T cells, and specific gene expression for (B) Per1 (C), Per2, (D) Per3, (E) Clock, (F) Arntl (Bmal1), (G) Arntl2 (Bmal2), (H) Fbxl3, (I) Rora, and (J) Nfil3 genes were quantified by qPCR. Each symbol represents a separate animal (n = 4 to 6), and the statistical significance of rhythmicity was determined by CircaCompare. Oscillations were considered rhythmic at P < 0.05, and evaluations of rhythmicity are indicated above each graph. If oscillations in both WT and KO cells were rhythmic, then differences in mesor, phase, and amplitude were calculated and displayed in brackets above the plots. Comparisons between WT and Adrb2^fx/fx × CD8a-Cre⁺ were considered significantly different at P < 0.05. n.s., not significantly rhythmic.

In measuring other core CR genes, we unexpectedly found that some genes adopted alternative oscillation patterns in the absence of Adrb2. For example, Bmal1 was significantly rhythmic in both WT and KO cells but differed significantly in mesor (Fig. 2F). Likewise, Fbxl3 adopted a completely alternate phase (Fig. 2H), while RORα differed significantly in both phase and mesor (Fig. 2I) in Adrb2^fx/fx × CD8a-Cre compared to Adrb2^fx/fx control CD8⁺ T cells. In contrast, other core CR genes, such as Nfil3 (Fig. 2J), Nr1d1, and Nr1d2 (fig. S3, C and D), were arrhythmic in WT cells but adopted a unique rhythmicity in Adrb2^fx/fx × CD8a-Cre cells. In addition to core CR genes, we measured the expression of Nr4a1 (Nur77) and NFkB p105, which are not considered components of the core CR gene network but are downstream of ADRB2 signaling (30–33). Similar to Nfil3, both Nr4a1 and NFkB p105 were arrhythmic in WT cells but adopted a unique and reciprocal rhythmicity in Adrb2^fx/fx × CD8a-Cre⁺ CD8⁺ T cells (fig. S3, E and F). Collectively, we conclude from these data that the intrinsic expression of Adrb2 is required for the rhythmic oscillation of key CR genes and some downstream T cell effector genes. In the absence of Adrb2, these rhythmic patterns were either disrupted, altered, or abolished together depending upon the specific gene. Thus, Adrb2 is a critical intrinsic regulator of CR gene oscillation in CD8⁺ T cells.

Adrb2 regulates CD8⁺ T cell subset development as a function of time of day of virus infection

Having established that Adrb2 regulated the intrinsic clock gene oscillation of CD8⁺ T cells, we determined whether this regulation correlated with their functional responses to viral infection. In response to Ag stimulation, T cells rapidly divide and differentiate into variegated subpopulations that have unique and specified functions to limit pathogen spread and provide long-lived immunity. This differentiation process is guided by many factors such as costimulation, transcription, the cytokine milieu, epigenetic factors, and the strength of T cell receptor signaling (34–37). However, it is unknown whether the time of day of pathogen exposure has any impact on T cell differentiation. Numerous studies have demonstrated the circadian regulation of other CD8⁺ T cell functions such as recirculation through lymph nodes and responses to vaccination and infection in a time-of-day–dependent manner (19, 20, 38, 39). Furthermore, both glucocorticoid signaling and adrenergic signaling regulate diurnal trafficking into lymph nodes (7, 8, 40), although a CD8⁺ T cell intrinsic role for these receptors has not been demonstrated. We wished to distinguish the potential effects of ADRB2-mediated T cell differentiation from its known role in regulating trafficking by analyzing T cell responses in the spleen. We measured the proportions of most major cell types in the spleen, ranging from T and B cells to macrophages and neutrophils at incremental times of day by high-dimensional mass cytometry. For this analysis, animals were euthanized at different times of day, and spleen cells were stained with a panel of metal-conjugated antibodies followed by data collection on a Helio mass cytometer. We found that in contrast to lymph nodes, the spleen does not display such diurnal fluctuations in immune cell subsets and, while highly trafficked, remains static in the proportion of various cells throughout the light/dark cycle (Fig. 3). Thus, the homeostatic nature of splenic immune cells provides a tractable background to determine whether T cell differentiation is affected by the time of day of pathogen exposure independent of the rhythmic accumulation of various cell types throughout the light/dark cycle.

Fig. 3. Splenic immune cells do not display time-of-day–dependent oscillation frequency.

WT C57Bl/6 mice were acclimated for 3 weeks in light chambers to a standard 12:12 light/dark cycle staged 6 hours apart. Spleens were harvested at ZT0, ZT6, ZT12, and ZT18, and single-cell suspensions were stained with a panel of metal-conjugated antibodies (table S2) to profile immune cell subsets. Data were collected on a Helios mass cytometer. Data were first gated on CD45⁺ cells [common leukocyte Ag (CLA)], and secondary gating was performed with the directionality of embedded gating depicted by arrows: (A) total B220⁺/CD11b⁻ B cells, (C) CD3⁺/CD11b⁻ and CD3⁻/CD11b⁺, (D) CD3⁺/NK1.1⁺ NK T cells (natural killer T cells), (F) CD3⁺/CD8⁺ and CD3⁺/CD4⁺ T cells, (I) CD3⁺/CD4⁺/TCRβ⁺/CD25⁺ T_reg cells (regulatory T cells), (K) CD11b⁺/CD3⁻/NK1.1⁺ NK cells and CD11b⁺/CD3⁻/NK1.1⁻ cells, (M) CD11b⁺/CD3⁻/Ly6G⁻ monocytes and CD11b⁺/CD3⁻/Ly6G⁺ neutrophils, and (O) CD11b⁺/CD3⁻/Ly6G⁻/Ly6C^lo monocytes and CD11b⁺/CD3⁻/Ly6G⁻/Ly6C^hi monocytes. The percentage of each population with respect to the parent gate is graphed in (B), (E), (G), (H), (J), (L), (N), (P), and (Q). None of the populations measured were found to be significantly rhythmic in their abundance as assessed by CircaCompare.

To examine time-of-day–dependent T cell differentiation, we compared the effector T cell response to VSV infection in the spleen between WT and Adrb2^fx/fx × CD8a-Cre⁺ mice. Here, cohorts of WT and Adrb2^fx/fx × CD8a-Cre⁺ mice were infected with VSV expressing chicken ovalbumin (VSV-OVA) at ZT0, ZT6, ZT12, and ZT18 within a 12:12 light/dark cycle. The VSV-OVA recombinant virus expresses chicken ovalbumin used as a traceable T cell Ag, and we infected them with a sublethal dose [1 × 10⁶ plaque-forming units (pfu)] to assess the spectrum of T cell phenotypes without the confounding effects of morbidity. All animals were euthanized on d7 at the same time of day, and splenocytes were stained with a panel of 18 fluorochrome-conjugated antibodies that included a labeled major histocompatibility complex (MHC) class I MHC-I H2-K^b tetramer loaded with the OVA 257–264 peptide to identify Ag-specific CD8⁺ T cells.

First, in assessing bulk T cells, we found no significant differences in the percentages of CD4⁺ or CD8⁺ cells as a function of time of day of virus infection (Fig. 4, A to C). The expansion of Ag-specific cells in the infected cohort was robust (Fig. 4, D and E), yet we also found no significant differences in either the percentages or total numbers of H-2K^b-OVA⁺ cells between WT and Adrb2^fx/fx × CD8a-Cre⁺ mice when infected at different times of day (Fig. 4F). Thus, the expansion of total Ag-specific cells was not regulated by the time of day at which the animals were exposed to the virus. We then tested whether the time of day of infection influenced the development of variegated T cell phenotypes that arose in response to virus infection. We first focused on CD44 and CD62L, which are common markers that distinguish bulk populations of naïve/central memory (CD44⁻/CD62L⁺), effector memory (CD44⁺/CD62L⁺), and effector cells (CD44⁺/CD62L⁻). By d7 postinfection, VSV-OVA infection increased the proportion of cells expressing CD44 within the total CD8⁺ population (Fig. 5, A and B), and we observed significant oscillatory behavior in both the CD44⁺/CD62L⁺ and CD44⁺/CD62L⁻ cells (Fig. 5, C and D). Oscillation of CD44⁺/CD62L⁺ cells was significantly disrupted in Adrb2^fx/fx × CD8a-Cre⁺ cells (Fig. 5C), while CD44⁺/CD62L⁻ cells retained their oscillatory behavior compared to WT cells (Fig. 5D). Within the Ag-specific fraction, very few H-2K^b-OVA⁺ cells were detected in uninfected spleens (Fig. 5E), and the expansion of tetramer-positive cells in response to infection drove their development largely toward CD44⁺/CD62L⁺ and CD44⁺/CD62L⁻ phenotypes (Fig. 5F). We observed significant reciprocal oscillatory behavior in the development of these populations, with CD44⁺/CD62L⁺ cells peaking at ZT6 hours postinfection and CD44⁺/CD62L⁻ cells peaking at ZT0 of infection (Fig. 5, G and H, left panels). Although these cells displayed oscillatory behavior, deletion of Adrb2 resulted in a marked phase shift in both populations (Fig. 5, G and H, right panels). Thus, Adrb2 orchestrates the development of CD8⁺ T cells in a time of day of infection-dependent manner, and these effects are independent of trafficking of total Ag-specific cells within the spleen.

Fig. 4. Expansion of Ag-specific CD8⁺ T cells as a function of time of day of infections with VSV.

WT Adrb2^fx/fx and Adrb2^fx/fx × CD8a-Cre⁺ cohorts (n = 4 to 8 per group) were acclimated for 3 weeks in cabinets in a 12:12-hour light/dark cycle staged 6 hours apart. Mice were infected with 1 × 10⁶ pfu VSV-OVA at ZT0, ZT6, ZT12, and ZT18 on d0 and allowed to recover to d7. (A) Spleen cells were stained for CD4 and CD8 and analyzed by fluorescence-activated cell sorting. (B and C) The percentages of CD4⁺ (B) and CD8⁺ (C) splenic T cells were quantified from each cohort and displayed as a function of time of day of infection. Representative dot plots of splenocytes stained for CD8 and the H-2K^b-OVA tetramer are shown for animals injected with saline (D) or infected with VSV-OVA (E). (F) The absolute numbers of splenic H-2K^b-OVA Tet⁺ cells were quantified and displayed as a function of time of day of infection. CircaCompare analysis found no statistical rhythmicity in any of the populations described above (P > 0.05).

Fig. 5. Adrb2 regulates time-of-day–dependent T cell subset development to virus infection.

WT Adrb2^fx/fx and Adrb2^fx/fx × CD8a-Cre⁺ cohorts (n = 8 per group, four females and four males) were acclimated for 3 weeks in cabinets in a 12:12-hour light/dark cycle staged 6 hours apart. Mice were injected with either saline (A and E) or infected with 1 × 10⁶ pfu VSV-OVA (B and F) at ZT0, ZT6, ZT12, and ZT18 on d0 and allowed to recover to d7. Splenocytes were stained with the H-2K^b-OVA tetramer and antibodies to CD3, CD8, CD44, and CD62L. Cells were gated on live CD3⁺/CD8⁺ cells (A and B) or gated on live CD3⁺/CD8⁺/H-2Kb-OVA Tet⁺ cells (E and F), and the proportion of CD44- and CD62L-expressing cells is shown within the quadrant gates. The percentages of CD44⁺/CD62L⁺ (C and G) and CD44⁺/CD62L⁺ (D and H) subsets were quantified and displayed as a function of ZT of infection. The statistical significance of rhythmicity was determined by CircaCompare. Oscillations were considered rhythmic at P < 0.05, and evaluations of rhythmicity are indicated above each graph. If oscillations in both WT and KO cells were rhythmic, then differences in mesor, phase, and amplitude were calculated and displayed in brackets above the plots. Comparisons between WT and Adrb2^fx/fx × CD8a-Cre⁺ were considered significantly different at P < 0.05.

Virus-driven T cell expansion is accompanied by rapid development into multiple phenotypes related to their functional capacities, and these unique subsets can be distinguished with select cell surface markers. We wished to determine whether the development of these CD8⁺ T cell subsets was differentially skewed on the basis of the time of day of virus infection. To measure this, we performed high-dimensional analysis of the CD44⁺/CD62L⁺ and CD44⁺/CD62L⁻ populations shown in Fig. 5 that were costained for the expression of CCR7, KLRG1, CD27, CD95, PD1, CXCR3, CD127, CD107a, CD69, and Ki67. These markers are used to identify numerous subsets of CD8⁺ T cells that fall into the terminal effectors, effector memory, and central memory along with cells having recently undergone activation, proliferation, or degranulation (41, 42). For each subset, dimensionality reduction was performed by uniform manifold approximation and projection (UMAP), and resulting populations were clustered by Phenograph (43). The analyses of total CD8⁺ T cells are shown in UMAP plots in Fig. 6 [A (CD44⁺/CD62L⁺) and F (CD44⁺/CD62L⁻)]. These clusters represent unique cellular phenotypes that are highly variegated on the basis of their differential marker expression. This is expected on the basis of a previous study characterizing the profound temporal changes that occur in CD8⁺ T cells as they develop in response to virus infection (44). The relative expression of various markers on select clusters is depicted as dot plots overlaid on the total population (Fig. 6, B to E and G to J). We quantified the relative abundance of each cluster within the total respective parent population and assessed the rhythmicity of their abundance by CircaCompare as a function of time of day of infection with VSV-OVA. Select clusters are shown for both the CD44⁺/CD62L⁺ (Fig. 6, B to E) and CD44⁺/CD62L⁻ (Fig. 6, G to J) populations to represent the diversity in phenotypes and rhythmicities observed in various populations. Although the CD44⁺/CD62L⁺ and CD44⁺/CD62L⁻ populations were clustered independently, we found that many clusters shared analogous marker expression, such as cluster 1 (CL1, orange; Fig. 6, B and G), suggesting common functionalities.

Fig. 6. Adrb2 regulates time-of-day–dependent T cell subset development to virus infection.

WT Adrb2^fx/fx and Adrb2^fx/fx × CD8a-Cre⁺ cohorts (n = 3 to 5 per group) were acclimated for 3 weeks in cabinets in a 12:12-hour light/dark cycle staged 6 hours apart. Mice were infected with 1 × 10⁶ pfu VSV-OVA at ZT0, ZT6, ZT12, and ZT18 on d0 and allowed to recover to d7. Splenocytes were stained with the H-2K^b-OVA tetramer, and a panel of 20 antibodies to cell surface markers is listed in table S3. Data were collected on a multispectral flow cytometer, and postacquisition unmixing was performed to apply compensation. Data were first Boolean gated on live CD3⁺/CD8⁺ events and then gated on CD44^+/CD62L⁺ (A) and CD44⁺/CD62L⁻ (F) populations as described in Fig. 5. The remaining markers were used to perform dimensionality reduction by UMAP followed by clustering with Phenograph. Individual clusters are color coded, and select clusters used for analysis are indicated by number in the UMAP plots in (A) and (F) for the CD44^+/CD62L⁺ and CD44⁺/CD62L⁻ populations, respectively. Select numbered clusters are displayed in (B) to (E) (CD44^+/CD62L⁺) and (G) to (J) (CD44⁺/CD62L⁻), and each colored cluster is overlaid on the total population in black. The cell surface marker expression of each cluster is represented in the overlay plots, and the percentage of each cluster is represented in the graphs on the right side as a function of time of day of infection. The statistical significance of rhythmicity was determined for each cluster by CircaCompare. Oscillations were considered rhythmic at P < 0.05, and evaluations of rhythmicity are indicated above each graph (*, significantly rhythmic).

In assessing the oscillation of each cluster, we find that some clusters (e.g., CL3, red; Fig. 6) were oscillatory in their developmental response to the virus, and their rhythmicity was similar between WT and Adrb2^fx/fx × CD8a-Cre⁺ cells (Fig. 6, C and H). In contrast, other clusters were differentially rhythmic between WT and KO cells. Within the CD44⁺/CD62L⁺ population, CL1, CL5, and CL11 displayed the time of day of infection oscillation in WT cells, and this rhythmicity was lost in the Adrb2^fx/fx × CD8a-Cre⁺ cells (Fig. 6, B, D, and E). Likewise, analogous CL1 and CL12 within the CD44⁺/CD62L⁻ population were rhythmic in WT cells but arrhythmic in the Adrb2^fx/fx × CD8a-Cre⁺ cells (Fig. 6, G and J). Other clusters such as CL1 and CL3 in the CD44⁺/CD62L⁺ population also share a common pattern of rhythmicity with the analogous phenotypes in CL1 and CL3 in the CD44⁺/CD62L⁻ population (Fig. 6, B, C, G, and H), suggesting that their development follows a common temporal mode of regulation. CD44⁺/CD62L⁺ CL11 and the analogous CD44⁺/CD62L⁻ CL12 populations displayed very similar patterns of rhythmicity in WT cells, which was lost in KO mice. These cells were unique in their high expression of the co-receptor CD27, consistent with a memory precursor phenotype (Fig. 6, E and J).

We performed a parallel analysis of CD44⁺/CD62L⁺ and CD44⁺/CD62L⁻ populations within the Ag-specific H-2K^b-OVA⁺ cells shown in Fig. 5F. We again observed significant variegated phenotypes within both CD44⁺/CD62L⁺ and CD44⁺/CD62L⁻ populations (Fig. 7, A and E), which displayed unique oscillation patterns within these clusters that were selectively regulated by Adrb2. As with total CD8⁺ cells, we find that some analogous clusters were not oscillatory in nature (CL5, yellow; Fig. 7, C and G), while others developed in a rhythmic fashion in an Adrb2-dependent manner (CL10, pink; Fig. 7D; CL7, pink; Fig. 7H). Like the CD27^hi subset identified in total CD8⁺ cells (Fig. 6), CL10 (Fig. 7D) and CL7 (Fig. 7H) within the Ag-specific subset displayed markers consistent with memory precursor cells, which persist following the resolution of infection. Not only were these clusters heterogeneous in their phenotype, their development was differentially regulated by Adrb2 and CRs. These data demonstrate that the differentiation of unique phenotypes of cells responding to the virus at specific time points is regulated by Adrb2 expressed intrinsically within CD8⁺ T cells. Thus, Adrb2 regulates the time-of-day–dependent balance of T cell phenotypes that emerge in response to virus infection, and this mode of regulation is a rhythmic mechanism to drive T cell developmental variegation.

Fig. 7. Adrb2 regulates time-of-day–dependent Ag-specific T cell subset development to virus infection.

WT Adrb2^fx/fx and Adrb2^fx/fx × CD8a-Cre⁺ cohorts (n = 3 to 5 per group) were acclimated for 3 weeks in cabinets in a 12:12-hour light/dark cycle staged 6 hours apart. Mice were infected with 1 × 10⁶ pfu VSV-OVA at ZT0, ZT6, ZT12, and ZT18 on d0 and allowed to recover to d7. Splenocytes were stained with the H-2K^b-OVA tetramer, and a panel of 20 antibodies to cell surface markers is listed in table S3. Data were collected on a multispectral flow cytometer, and postacquisition unmixing was performed to apply compensation. Data were first Boolean gated on live CD3⁺/CD8⁺/H-2K^b-OVA Tet⁺ events and then gated on CD44^+/CD62L⁺ (A) and CD44⁺/CD62L⁻ (E) populations as described in Fig. 5. The remaining markers were used to perform dimensionality reduction by UMAP followed by clustering with Phenograph. Individual clusters are color coded, and select clusters used for analysis are indicated by number in the UMAP plots in (A) and (E) for the CD44^+/CD62L⁺ and CD44⁺/CD62L⁻ populations, respectively. Select numbered clusters are displayed in (B) to (D) (CD44^+/CD62L⁺) and (F) to (H) (CD44⁺/CD62L⁻), and each colored cluster is overlaid on the total population in black. The cell surface marker expression of each cluster is represented in the overlay plots, and the percentage of each cluster is represented in the graphs on the right side as a function of time of day of infection. The statistical significance of rhythmicity was determined for each cluster by CircaCompare. Oscillations were considered rhythmic at P < 0.05, and evaluations of rhythmicity are indicated above each graph (*, significantly rhythmic).

DISCUSSION

In this study, we report two major findings that link adrenergic signaling to the entrainment of oscillations in CD8⁺ T cell gene expression and responses to viral infection. We found that the intrinsic expression of Adrb2 is required for normal timely CR-dependent gene expression, and this oscillation mirrors the robust development of unique T cell phenotypes as a function of time of day of infection with VSV. Both innate and adaptive immune cells display rhythmic behavior such as homeostatic recirculation through lymph nodes and time-of-day–dependent responses to pathogens (4, 38, 45–48). Most studies of CR regulation of immune function in mice have focused on innate immune cell responses to endotoxin (45, 49–54). However, recent studies have focused on CR regulation of adaptive T and B cells (19, 20, 55) and have even linked Adrb2 to this regulation (7, 40, 56). In mice, trafficking of B and T cells into lymph nodes follows a circadian oscillation, with both B and T cells optimally immigrating into lymph nodes toward the end of the light cycle (38). Immigration to LNs is regulated by diurnal oscillations in glucocorticoids and NE, and deletion of Adrb2 partially blocks the oscillatory nature of recirculation (7, 40). In some circumstances, oscillatory behavior in one cell type is directed by reciprocal oscillations in others. For example, circadian oscillation within dendritic cells regulates rhythmic responses to vaccine Ag in T cells by modulating Ag processing (57). A similar indirect mechanism of CR-dependent antitumor response by CD8⁺ T cells has also been observed (58). T cells also display direct CR-dependent behavior (21), which has been revealed by disrupting the core clock by genetic ablation of Bmal1 (19, 20). Although various neurotransmitters such as glucocorticoids and NE have been shown to regulate these CR-dependent immune processes, their mechanism of control remained elusive.

Sympathetic nerves secrete NE in response to pathogenic organisms [reviewed in (59)]. Neurons themselves express various Toll-like receptors, enabling them to respond directly to certain pathogen-associated molecular patterns (60). Both viral and bacterial infections elicit bursts of NE secretion from sympathetic neurons, and pathogen-associated molecular patterns such as lipopolysaccharide drive NE release within seconds upon exposure (61, 62). Soluble neurotransmitters play a key role in regulating the light/dark entrainment of CR-dependent biological functions. In studying the role of epinephrine and NE in immune cell regulation, we found a remarkable immunosuppressive function for NE through ADRB2 signaling that potently inhibited the inflammatory processes of CD8⁺ T cells (13) as well as innate macrophages and dendritic cells (14). In the course of these studies, we found an unexpected role for Adrb2 in directly regulating the CR core gene transcriptional pathway (22), and a detailed analysis of this finding is presented in the current study. Seven of the core CR genes belonging to the KEGG CR pathway were found to be significantly differentially expressed in Adrb2-deficient CD8⁺ T cells on one or more days following VSV infection. This experiment also revealed a dysregulation of some genes before infection that remained differentially expressed at different days after infection, such as Per2. This may suggest that both intrinsic Adrb2 signaling and extrinsic factors from viral infection affect CR gene expression (63). Regardless, we found that the loss of Adrb2 intrinsically within CD8⁺ T cells significantly altered the expression of several key CR genes as they divided in response to infection.

On the basis of these initial findings, we directly measured Adrb2-mediated CR gene oscillation within CD8⁺ T cells throughout a 24-hour light/dark cycle under steady-state conditions. We found that the intrinsic expression of Adrb2 on CD8⁺ T cells regulates the diurnal oscillation of many core CR genes and downstream CR gene targets. Adrb2-mediated CR gene regulation was selective, as not all CR genes were affected by the absence of ADRB2 signaling such as Per1. Although oscillation of the core CR genes Per2 and Per3 was disrupted, we observed several genes that became oscillatory in expression or adopted a unique phase or amplitude in the absence of Adrb2. There are several interpretations of these data. First, the curious acquisition of alternate frequencies by other genes could be explained by the continued response to other entrainment neurotransmitter receptors, such as the glucocorticoid receptor and the acetylcholine receptor, which are still intact and available for signaling in Adrb2-deficient cells. Thus, in the absence of Adrb2, continued oscillatory signaling through other entrainment receptors may artificially promote alternate expression patterns in downstream CR genes. Alternatively, there may be an intrinsic dysregulation that occurs when the normal oscillation frequency of one factor becomes disturbed in the absence of Adrb2. For example, a direct alteration in Per2/3 oscillation in the absence of Adrb2 may indirectly affect Bmal, as these two factors reciprocally regulate each other (64). Last, it is also likely that posttranslational mechanisms, including phosphorylation and ubiquitination, are also under the direct control of ADRB2 signaling, which may affect both mRNA and protein levels of select CR genes (65). In the case of Per2, we found that the PER2 protein, while phase shifted, remained oscillatory in Adrb2-deficient cells, even though oscillation of Per2 mRNA was lost. As both transcriptional regulation and posttranslational regulation of CR genes and gene products operate interdependently to create oscillation, it is possible that ADRB2 signaling regulates both mechanisms to differentially modulate CR gene expression and protein stability. In conclusion, we find that Adrb2 is critical for regulating the oscillation of CR genes in CD8⁺ T cells under homeostasis.

Rhythmic CR gene expression is known to drive hundreds of downstream gene targets in an oscillatory manner over the course of a day, and this oscillation in immune cells likely explains differences in responses to pathogens as a function of time of day of infection. In the case of influenza, mice infected at the end of the light cycle (ZT11) were more susceptible to the pathological effects of flu than those infected at the beginning of the cycle (ZT23), even though viral titers were not significantly altered (39). As the excessive release of pro-inflammatory cytokines drives pathological outcomes, cellular immunity is likely the key factor in controlling circadian-regulated responses. Thus, our second major finding demonstrates the role for Adrb2 in regulating CR-dependent diversification of T cell subsets that develop in response to virus infection. CD8⁺ T cells divide rapidly and express CD44 in response to virus infection. Approximately half of these responders down-regulate CD62L as they commit to terminal effector phenotypes. We observed a profound time-of-day–dependent shift in the proportion of cells that committed to the CD44⁺/CD62L⁺ and CD44⁺/CD62L⁻ phenotypes, which was significantly altered in the absence of Adrb2. The effect was particularly pronounced in Ag-specific cells, which displayed a marked reciprocal oscillation phase between CD44⁺/CD62L⁺ and CD44⁺/CD62L⁻ when animals were infected at the ZT6 time point, and was significantly phase shifted in Adrb2-ko cells. The time-of-day abundance of these select phenotypes is unlikely to be explained by selective ingress and egress into the spleen, as none of the total parent populations of responding cells differed in proportion either at the steady state or in response to different times of day of infection. Rather, we observed a remarkable time-of-day–dependent development of unique T cell subsets that displayed lineage marker expression of various phenotypes, which were selectively dysregulated in the absence of Adrb2. On a more detailed level, we found that conventional short-lived effector cells (KLRG1^hi/CD27^lo/PD1^hi) developed and expanded in a manner independent of the time of day of infection and displayed no significant rhythmicity, and their expansion was not affected by the loss of Adrb2. In contrast, recently activated and potentially degranulated cells within the total CD8⁺ population (CD107a^hi/Ki67^hi) displayed rhythmicity, peaking when animals were infected between ZT12 and ZT18, and their oscillation was also not regulated by Adrb2. Notably, in WT mice, a population of CD27^hi/KLRG1^int/CD95^lo cells peaked in their expansion at ZT12 to ZT18, while their rhythmic development was lost in the absence of Adrb2. These cells developed in both CD44⁺/CD62L⁺ and CD44⁺/CD62L⁻ compartments and within the total CD8⁺ and Ag-specific H-2Kb-OVA⁺ populations. These cells most closely match the memory-precursor effector phenotype with high expression of CD27 and low expression of KLRG1 (66). Thus, the rhythmic development of these varied effectors differs from each other, and their oscillation was selectively dependent on Adrb2. It is yet unclear how this selective development occurs as very little is known about how complex phenotypes arise at each cell division after activation. These populations could be driven by differential expression of Adrb2, selective cytokine expression influencing Ag-presenting cell function, or possibly time-of-day–dependent proliferation kinetics, none of which are mutually exclusive. Nonetheless, our findings demonstrate a complex and selective role for Adrb2 in orchestrating time-of-day–dependent T cell phenotype development in response to infection.

The differential rhythmic nature we see in cellular phenotypes between WT and KO cells mirrors the types of rhythmic patterns we observed in gene expression. Some genes were rhythmic in WT but not in KO and conversely for others. It is likely that the loss of Adrb2 allows other entrainment pathways to dominate and create alternative rhythms for some genes and for some developmental cell types. Recent studies have shown that influenza A infection follows a clear CR-dependent inflammatory response in the lungs, with more severe inflammation occurring at later ZTs of infection (39, 67). Because of the critical involvement of CD8⁺ T cells in general responses to viral infection, we propose that CD8⁺ T cell regulation by Adrb2 will control CR-dependent responses and inflammation. Our previous studies described a potent role for Adrb2 in acutely suppressing CD8⁺ T cell cytokine secretion and lytic activity (13). It is possible that some of the immunosuppressive activities of Adrb2 signaling are independent of its regulation of CR gene oscillations, as both systemic and local levels of NE follow a diurnal pattern of secretion by sympathetic neurons that innervate secondary lymphoid tissues. Epinephrine and NE levels rise and fall systemically in a diurnal fashion, and real-time measurements in rats show a peak of epinephrine and NE in serum at the end of the dark cycle (68). We propose that this normal oscillation of ADRB2 ligands directly entrains intrinsic CR gene oscillation in CD8⁺ T cells. Bmal1 and Per2 represent the core factors on either side of the clock pendulum, both of which are dysregulated in their periodicity in the absence of Adrb2. Because whole-body deletion of Bmal1 leads to enhanced flu-mediated lung inflammation (69), we predict that Bmal1, and perhaps Per2, may also regulate the magnitude and timing of memory T cell responses to influenza in a manner that is dependent on entrainment by ADRB2 signaling. The current study opens unique areas to explore the mechanisms underlying time-of-day–dependent responses to infection and vaccination. Future studies will identify specific CR gene–dependent pathways in CD8⁺ T cells that regulate both their early effector cell development and later memory cell diversification and responses to recall infection.

MATERIALS AND METHODS

Animal care

All mice used in this study were handled according to the guidelines of the Institutional Animal Care and Use Committee at the University of Texas Southwestern Medical Center. WT C57Bl/6J and CD8α-Cre transgenic mice (27) were purchased from the Jackson Laboratory. Adrb2^fx/fx animals were provided by G. Karsenty from Columbia University (26). Adrb2^fx/fx and Cd8αCre⁺ mice were bred to obtain Adrb2^fx/fx Cd8αCre⁺ mice, and Cre⁻ littermates were used for controls. The PER2::LUC^+/+ mice (catalog number 006852) were received from J. Takahashi (28). The Adrb2^fx/fx Cd8αCre⁺ and PER2::LUC^+/+ mice were bred to obtain PER2::LUC^+/−Adrb2^fx/fx Cd8αCre⁺ mice. All mice were housed and bred in specific pathogen–free facilities under a standard 12:12-hour light/dark cycle with ad libitum feeding.

Circadian cabinets

Timed gene expression experiments and infections were performed in a Biosafety Level 2–level biohazard containment facility equipped with circadian cabinets. For both steady-state measurements and timed infections, animals were cohoused in cages placed in light/dark cabinets (Actimetrics). Each cabinet was programmed to a 12:12-hour light/dark cycle differing by 6 hours with respect to ZT0, initiating the light phase. The cabinets maintained a constant ambient temperature of 22°C with constant air recirculation.

Infections

For standard VSV infections, WT and Adrb2^fx/fx Cd8αCre⁺ mice were infected by intraperitoneal injection with 100 μl of either 10⁶ pfu VSV-OVA (70) or saline. Animals were monitored and weighed daily until the termination of the experiment.

Luciferase assay

CD8 T cells were isolated from spleens of mice using the CD8⁺ T cell purification kit according to the manufacturer’s instructions (Miltenyi Biotech). Luciferase activity was measured in purified CD8 T cell lysates with the luciferase assay system (Promega). Purified CD8 cells were lysed using 1× cell lysis buffer, and 100 μl of luciferase assay substrate was added to 20 μl of cell lysate in luminometer tubes. The tubes were vortexed briefly, and luciferase activity was measured with a TD 20/20 luminometer (Turners) with the 2-s measurement delay followed by a 10-s measurement read for luciferase activity.

Reverse transcription polymerase chain reaction analyses

Total RNA was isolated from CD8⁺ T cells with the Qiagen RNA extraction kit, and 40 to 100 ng of RNA was used to perform reverse transcription using the ABI High Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Foster City, CA). The cDNA was used as a template for the qPCR reactions using Bio-Rad’s SYBR Green supermix using the QuantStudio 7 Flex Real-Time PCR System (Applied Biosystems, Foster City, CA). The primers used to quantify mRNA are listed in table S1. Rn18s and Eif2a were used as reference genes in qPCRs, and the average changes in both references were used to calculate relative specific mRNA differences between samples. The relative gene expression differences were calculated using the 2^−ΔΔCt method (71).

Flow and mass cytometry

Spleen cells were prepared by homogenizing spleen tissue through mesh screens followed by red blood cell lysis treatment (22). Single-cell suspensions of mouse splenocytes were used for staining in all flow cytometry measurements. For immune cell profiling by mass cytometry, splenocytes were stained for cell surface proteins with metal-conjugated antibodies listed in table S2. Cells were analyzed on a Helios mass cytometer (Standard Biotools). Data were batch normalized, and postacquisition processing was performed to remove residual events (72). Downstream analysis was performed on the OMIQ data analysis pipeline (www.omiq.ai).

For standard cell surface phenotyping, splenocytes were first stained with the MHC-I tetramer (H-2K^b-OVA, NIH Tetramer Core, Atlanta, GA) in RPMI media for 30 min at room temperature. The cells were then stained with a cocktail of cell surface antibodies (BioLegend; table S3) in 0.5% bovine serum albumin/phosphate-buffered saline and analyzed on an Aurora multispectral flow cytometer (Cytek Biosciences). Data were compensated by unmixing in FlowCyte Software (Cytek Biosciences) followed by dimensionality reduction by UMAP and clustering with Phenograph (73) within the OMIQ data analysis pipeline.

Statistical analysis

Statistical tests were performed using GraphPad Prism software. One-way or two-way analysis of variance (ANOVA) was used followed by a Bonferroni post test for pairwise comparisons within the groups. A Student’s two-tailed t test was used for simple pairwise comparisons; differences with P > 0.05 were considered significant. The rhythmic pattern of oscillation and the differences in mesor, phase, and amplitude were quantified by the CircaCompare (29) statistical tool available on Github (https://github.com/RWParsons/circacompare).

For high-dimensional flow and mass cytometry data, all populations were either identified by Boolean gating (cytokine-secreting cells) or by clustering (cell surface phenotypes). For clustering, data were first subjected to dimensionality reduction by UMAP and then clustered with the Phenograph algorithm (43) at k = 50. SAM (statistical analysis of microarray) tools of OMIQ were used to perform statistical comparisons between groups and between time zones. Data from statistically significant comparisons were then analyzed for rhythmicity with CircaCompare. The rhythmicity of data was graphed in Prism software with a third-order polynomial function to represent temporal oscillation.

Supplementary Materials

The PDF file includes:

Figs. S1 to S3

Tables S1 to S3

Legend for data file S1

Other Supplementary Material for this manuscript includes the following:

Data file S1