How external signals influence the body’s internal clock protection against flu infection

We have previously shown that mice infected at dawn had 3-fold higher survival compared with mice infected at dusk (). We hypothesized that abrogation of cyclical photic cues caused by constant light exposure would abolish the time-of-day-specific protection from influenza A virus (IAV) driven by the endogenous circadian clock. By circadian terminology, ZT0 refers to the time when the lights turn on, and ZT12 is when the lights turn off; therefore, ZT23 refers to the time just before the lights go on or just before the onset of the rest phase, and ZT11 refers to the time just before the lights go off or before the onset of the active phase. As in our prior studies, C57BL/6J adult mice were infected at either ZT23 or ZT11 with IAV (H1N1; PR8- 30 PFU) and maintained in 12-hour light-dark (LD) cycling conditions. A subset of mice infected at ZT23 were moved to constant light conditions on day 4 postinfection (p.i.), remaining there throughout the study [referred to as ZT23(LL) mice] (). Light intensity for the LD and LL groups was between 270 and 300 lux. Animals were weighed and monitored daily for 14 days. Our choice of constant light exposure to start on day 4 was driven by our previous work, where we noted the weight trajectory of the ZT11 group started diverging from that of the ZT23 group by day 4, with other signs of inflammation being apparent by day 2 (). Henceforth, we referred to ZT23(LD) and ZT11(LD) as ZT23 and ZT11, respectively. Additional qualifiers will be used to define how the groups differed from the 12-hour LD cycling. ⁵ Figure 1A ⁵

Consistent with findings from our previous work, mice kept in 12-hour LD cycling and infected at ZT23 had significantly better survival (88.88% in ZT23 vs. 44.82% in ZT11;< 0.01 by Mantel-Cox log-rank test) than the mice infected at ZT11 (). However, the ZT23(LL) group had significantly lower survival [58.33% in ZT23(LL) vs. 88.88% in ZT23;< 0.05 by Mantel-Cox log-rank test] than the ZT23 group (). This was comparable to the mice infected at ZT11 [44.82% survival in ZT11 and 58.33% in ZT23(LL)]. This effect was also reflected in the weight loss trajectory, where the ZT23-infected groups are identical until day 4 p.i. (). However, thereafter, the weight loss was more pronounced in the ZT23(LL) than the ZT23 group, eventually becoming indistinguishable from that of the ZT11 group (). We also monitored these mice for sickness behavior using a previously validated scoring system (; supplemental material available online with this article;) based on activity level, behavior, and respiratory distress (,) — the higher the score, the sicker the animal. Like the ZT11 mice, the ZT23(LL) group had consistently higher scores than the ZT23 group, suggesting higher morbidity with constant light exposure (). Upon stratification by sex, we saw similar patterns in both males and females, which is consistent with our previous work (,) (). Together, these data suggest that disruption of light-dark cycling, even after the initial few days of influenza infection, abrogates the clock-driven time-of-day-specific protection from influenza-induced mortality and morbidity. P P Figure 1B Figure 1B Figure 1C Figure 1C Supplemental Table 1 ^{5 6} Figure 1D ^{5 7} Supplemental Figure 1 https://doi.org/10.1172/JCI193133DS1↗

We quantified viral titers in lung tissue to determine whether environmental light disruption affects IAV clearance. Since the disruption of light cycling was introduced on day 4 p.i., viral titers were measured on days 6 and 8 p.i. to allow the host to experience the effects of this perturbation. For this dose of the IAV, we expected days 6 and 8 p.i. to represent time points that follow peak viral titers and likely represent viral clearance. No significant difference in viral burden was observed between the ZT23 and ZT23(LL) groups on either day (). However, the ZT23 group exhibited significantly lower viral titers than the ZT11 group on day 6, consistent with our previous work, adjusted for the initial dose of the virus at infection (). The ZT23(LL) group had viral titers that fell between the ZT23 and ZT11 groups. However, by day 8, all groups had cleared the virus. These findings suggest that light-dark cycle disruption does not significantly affect IAV clearance. Figure 2A ⁵

Given that the viral titers were comparable in the ZT23(LL) and ZT23 groups despite significantly higher mortality in the former, we next investigated if exposure to constant light worsened lung inflammation. Consistent with our previous work (), the total bronchoalveolar lavage (BAL) cell count was higher in the mice infected at ZT11 than in those infected at ZT23 (). Interestingly, the ZT23(LL) group mice had higher total BAL cell counts than the ZT23 group, comparable to the ZT11 group (). The higher BAL count in the ZT23(LL) and ZT11 mice was accompanied by a significantly lower percentage of macrophages compared with the ZT23(LD) group (). Next, to test if a short period of LD cycling-driven circadian disruption can result in a state of heightened inflammation in the lungs, we examined BAL from naive mice exposed to constant light for 4 days. We found no differences in the total BAL cell counts between mice maintained in 12-hour LD cycling and those exposed to constant light (, A and B). Considered together, this supports the idea that the effect of short periods of light disruption is exacerbated in an infected host. The ZT23(LL) and ZT11 groups had significantly more polymorphonuclear cells in the BAL than the ZT23 group (). Furthermore, compared with the ZT23 group, on day 6 p.i., the levels of chemokines, MCP-1 and CXCL10, were significantly higher in the BAL of the ZT23(LL) and ZT11 groups (). The levels of cytokines IL-1β and IL-15 were also significantly higher in both the ZT23(LL) and ZT11 groups than in the ZT23 group, whereas the level of IL-2 was only significantly higher in the ZT23(LL) group (). Interestingly, there was no significant difference in IFN-γ levels across the 3 groups on day 6 p.i. (). Overall, the chemokine/cytokine profile of the ZT23(LL) group was aligned with that of the ZT11 group. Based on these analyses, we hypothesized that the ZT23(LL) group would have more proinflammatory immune populations than the ZT23 group. Therefore, we harvested lungs on day 8 p.i. and compared the immune populations in the 3 groups. Consistent with results from our previous work and our BAL analyses, we found that the ZT23(LL) mice had more leukocytes in the lungs than the ZT23 mice (). Because the ZT23(LL) group had higher levels of MCP-1 and CXCL10 — chemokines that promote monocyte and lymphocyte trafficking, respectively, into inflamed lungs — we enumerated these populations in the lungs (). Ly6Cinflammatory monocytes (CD45Ly6GCD11bLy6C) and T cells (CD45CD4and CD45CD8) were higher in the ZT23(LL) group than in the ZT23 group (). We posit that the intraparenchymal location and the migration patterns of T cells following influenza A infection explain why the percentage of lymphocytes in the BAL on after infection day 6 was not different for the ZT23 group versus the ZT11 or ZT23(LL) groups inB. On the other hand, flow cytometry of lung cells on day 8 revealed an elevation of T cells (). The absolute number of neutrophils (CD45Ly6GCD11b) was also higher in the ZT23(LL) group than in the ZT23 group; however, the number of B cells was comparable across the 3 groups (). Thus, the disruption of environmental lighting leads to an exaggerated inflammatory milieu, marked by high levels of chemokines and accumulation of inflammatory cell populations. ⁵ Figure 2B Figure 2B Figure 2C Supplemental Figure 2 Figure 2C Figure 2D Figure 2E Figure 2E Figure 2F Figure 2, G–K Figure 2, G–I Figure 2 Figure 2, G and H Figure 2K hi + – + hi + + + + + + +

To validate the above results, we undertook a histological assessment of lungs harvested from the 3 groups. Consistent with our previous results, using a validated scoring system (,), the ZT23 mice sustained significantly less injury than mice in the ZT11 group (, and). Lung pathology was markedly worse for the ZT23(LL) than the ZT23 group, with higher peribronchial infiltrates, perivascular infiltrates, inflammatory alveolar exudates, and epithelial necrosis (). Furthermore, we validated the results of our immunophenotyping on histology from day 8 p.i. by staining for T lymphocytes (CD3) and macrophages and monocytes (F4/80). The lungs harvested from the ZT23(LL) and ZT11 groups had a significantly higher proportion of CD3and F4/80cells per high-power field (HPF) than those of the ZT23 group (, respectively). Next, we investigated if constant light exposure would affect lung regeneration, even beyond the acute immunopathology described above. After severe lung injury, alveolar type 2 (AT2) cells promote the regeneration of the airspace or alveoli (–). We found that there were significantly fewer AT2 cells, denoted by SFTPCcells/HPF, in the lungs of the ZT23(LL) and ZT11 groups than in the ZT23 group (). This could have affected how the lungs regenerate after the initial inflammatory response. Interestingly, we previously showed that the circadian clock, especially the AT2 clock, affects mortality (), lung regeneration, and repair following IAV infection (). Although the primary driver for the poorer outcomes in the ZT23(LL) group, like those for the ZT11 group historically, is immune-mediated pathology, a loss of regenerative capacity would likely worsen recovery from IAV, leading to long-term morbidity. ^{5 6} Figure 3, A and B Supplemental Table 2 Figure 3, A and B Figure 3, C–F ^{8 11} Figure 3, G and H ^{6 7} + + + + +

Having established the need for synchronizing the influence of entraining photic cues in maintaining the clock-driven protection from IAV in the early phase of influenza, we asked if there exists a window of vulnerability to light disruption following influenza infection. At the range of the doses used for the infection, we expected the peak viral titers to be around day 4 and clearance coinciding with day 7–8. Therefore, we infected mice at either ZT23 or ZT11 and maintained them in 12-hour LD cycling conditions until day 7 p.i. At this time, a subset of the ZT23 groups was moved to constant light, referred to as the ZT23(LL-DAY7) group, instead of day 4, as in earlier experiments (). Figure 4A

Mice from the ZT23 group had significantly lower mortality than mice infected at ZT11. However, unlike the results from constant light exposure on day 4 p.i., the time-of-day difference in outcomes was not lost when animals were moved to constant light on day 7 p.i. [Figure 4B; 90% survival in ZT23 versus 86.66% survival in ZT23(LL-DAY7) and 42.8% in ZT11]. Furthermore, the weight loss trajectory and morbidity of the ZT23(LL-DAY7) group were comparable to that of the ZT23 group (), suggesting that there was a window of vulnerability to LD disruption following influenza infection. Figure 4, C and D

To investigate if the central clock mediates the loss of time-of-day-specific protection in influenza infection in mice exposed to constant light, we analyzed the locomotor activity as a read out for the former. Since singly housing mice (necessary for locomotor activity monitoring) after IAV infection could worsen outcomes, we used naive mice in this experiment. Mice were exposed to 4 weeks of LD cycling and then moved to constant light conditions for 3 weeks. We found that the mice in LD cycling showed clear demarcation between active and rest phases, coinciding with the dark and light phases of the 24-hour cycle (). The mice exposed to constant light displayed erratic onsets of activity, eventually becoming completely arrhythmic after being exposed to constant light for 3 weeks (). Mice kept in continuous light, although arrhythmic, had comparable activity levels, supporting the idea that constant light led to a loss of rhythmicity but did not impair overall activity level (, B and C). Prolonged constant light exposure in an uninfected host is also associated with increased stress response (,). This response may include neurohumoral alteration (–), immune perturbations (–), metabolic dysregulation (,), and adverse behavioral changes (,). Many of these effects are often attributed to circadian disruption. To test whether the increased mortality observed under constant light is due to circadian-independent stress mechanisms, we repeated this experiment in clock mutant mice. Given our context-specific and short duration of LL exposure, we hypothesized that the effect of LL in our influenza model mediates its effect mainly through circadian disruption, rather than additional stress independent of circadian disruption. Supplemental Figure 3 Supplemental Figure 3A Supplemental Figure 3 ^{12 13 14 16 17 19 20 21 15 22}

To test this, we used() mice, in which the clock has been genetically disrupted globally. These mice are arrhythmic in constant darkness; henceforth, they are abbreviated as DD (). Previously, we have shown that thesemice do not have clock-driven time-of-day-specific protection against IAV (). Initially, we tested the effect of LD cycling on WT andmice to maintain consistency with our experimental design, as outlined in Figure 1. However, we found thatmice under LD conditions showed a reversal of the time of day-specific protection compared with WT mice, withmice infected at ZT11 having better outcomes thanmice infected at ZT23 (). This would suggest that, in the absence of the molecular clock, LD cycling primes processes allowing a reversal but not an abrogation of the time-of-day protection. Thus, based on our previously published work, we tested themice in DD conditions (). In circadian terminology, CT time refers to the same times as ZT but under DD conditions. Therefore, CT23 represents subjective dawn. We infectedmice and their crelittermates maintained in DD conditions. On day 4 p.i., a subset ofmice were moved from constant darkness to constant light conditions (). We found that, consistent with our previous results (),mice infected at CT23 experienced significantly higher mortality than wild-type littermates (< 0.001, Mantel-Cox log-rank test;). However, whilemice exposed to constant light still had higher mortality than the WT littermates, there was no difference in the outcomes ofmice in the 2 lighting conditions [survival of 31% inCT23 and 50% inCT23(LL-D4)], suggesting that constant light did not worsen the outcomes over that attributed to loss of the clock. iBmal1 –/– Bmal1ERT2Cre fl/fl + iBmal1 –/– iBmal1 –/– Bmal1 –/– iBmal1 –/– iBmal1 –/– Bmal1 –/– Bmal1 –/– Bmal1 –/– Bmal1 –/– P Bmal1 –/– Bmal1 –/– iBmal1 –/– iBmal1 –/– ^{23 5} Supplemental Figure 4 ⁵ Figure 4E ⁵ Figure 4F neg

To determine what pathways resulted in worse outcomes in the ZT23(LL) group, we performed transcriptional profiling on lungs harvested on day 8 p.i from both ZT23 and ZT23(LL) (= 5/ group; all females) (). Environmental light disruption following influenza infection resulted in disparate transcriptional phenotypes (). 1,125 genes were differentially expressed between the 2 groups, with 801 genes upregulated and 324 genes downregulated in the ZT23(LL) group relative to the ZT23 group. Several genes associated with the migration of leukocytes (,,) into the lung and exaggerated immune activation (,,), were upregulated in the ZT23(LL) group compared with the ZT23 group (). Further pathway analysis revealed that while those involved in innate activation (granulocyte diapedesis, IL-15 signaling, cytokine storm) were upregulated in the ZT23(LL) group relative to the ZT23 group, the most highly significant pathways involved innate to adaptive communication (communication between innate and adaptive cells) and activation of the adaptive immune system (B cell activation as in systemic lupus, altered T and B cells signaling, NFAT signaling) (). Overall, these pathways supported our findings of excessive immune cells, especially lymphocytes and monocytes, in the lungs of the ZT23(LL) group compared with the ZT23 group. Thus, even a short duration of light-based circadian disruption caused profound alteration of the immune response. n Ccr5 Ccl4 Cxcl10 Il18rap Gzma Ly6c2 Figure 5A Figure 5B Figure 5C Figure 5D

While we have shown that the ZT11 and ZT23(LL) groups both show excessive immune infiltration and subsequent tissue destruction (and), we wanted to test if the mechanism(s) underlying the poorer outcomes in both these groups were indeed similar. For this, we expanded our transcriptomic study to include 3 groups — ZT23, ZT11, and ZT23(LL) (). There were 1,304 differentially expressed (DE) genes when comparing the ZT23 group with the ZT11 group. Of these, 856 genes were upregulated and 448 genes were downregulated in the ZT11 group relative to the ZT23 group. Of the 1,304 DE genes, 739 genes were common to the comparison between ZT23 vs. ZT23(LL) (). Of these 739 shared DE genes, 587 genes were upregulated, while 151 genes were downregulated in both ZT23(LL) and ZT11 groups relative to the ZT23 group. Only one low-expression gene () was downregulated in the ZT23(LL) group and upregulated in the ZT11 group. Thus, not only was there significant overlap in the DE genes (), but their directionality was also conserved across the 2 comparisons — that of ZT23(LL) versus ZT23 and ZT11 versus ZT23, respectively. To understand how the cellular composition of the lungs from each of these 3 groups affected our transcriptomic profile, we used another dataset from day 8 p.i., where single-nuclear RNA-Seq (sn-RNA-Seq) analyses had been performed. This allowed us to use deconvolution algorithms to infer the cellular composition (). As expected, the ZT23 groups had a markedly lower proportion of immune cells than the ZT23(LL) or ZT11 groups (, A and B). Next, we compared the ZT23(LL) with the ZT11 group to further confirm the mechanistic overlap. We found only 13 DE genes with a 2-fold change in expression between these 2 groups; of these, many were circadian genes like(),, and, whose expression oscillates across the day and is expected to be disrupted by constant light (). In further analyses, as expected, various innate immune pathways, namely, "cell adhesion and diapedesis" and "the role of hypercytokinemia/hypercytokinemia in the pathogenesis of influenza," were enriched and conserved in both comparisons (). Pathways involved in adaptive immune activation were also enriched in the ZT23(LL) and ZT11 groups relative to the ZT23 group (). Genes involved in leukocyte migration (,) and activation of the immune system (,,,) were all higher in the ZT23(LL) and ZT11 groups compared with the ZT23 group. We confirmed changes in expression in some of the key genes by qPCR (). Together, these results suggested that LD-cycling perturbation following influenza impairs lymphocyte and monocyte diapedesis and results in the clock-regulated less favorable time of day. Next, we blocked the migration of leukocytes into the lungs in the ZT23(LL) groups by administering anti-leukocyte function–associated antigen (LFA) antibody on day 4 p.i., a time coinciding with their transition from LD cycling to constant light (). Interestingly, we observed that blocking leukocyte migration did not improve the outcomes in the ZT23(LL) group [; 93.3% survival in ZT23 vs. 54.5% in ZT11, vs. 65% in ZT23(LL) vs. 53.8% in ZT23(LL) with anti-LFA]. This raises 2 possibilities that are not mutually exclusive. First, some leukocyte migration may be beneficial to the host, and inhibiting all leukocyte trafficking while protecting from the effects of light disruption might make the host more vulnerable owing to the absence of the protective cell populations, at least when inhibited early on day 4 p.i. The second and the more speculative possibility is that the immune cells are merely responding to the signals emanating from the lung parenchyma in response to the light disruption; therefore, blocking leukocyte migration itself has minimal independent effect on mortality. Regardless, from these experiments, we concluded that the processes affected by light disruption are usually regulated by the clock to confer time-of-day-specific protection. Figures 2 3 Figure 6A Figure 6A Figure 6B ²⁴ Supplemental Figure 5 Figure 6C Figure 6C Figure 6C Figure 6, D–H Supplemental Figure 6A Supplemental Figure 6B Oxtr Bmal1 Arntl Dbp Npas2 Ccl4 Ccr5 Ifng Klrk1 Gzma Il18rap

While light is the primary external cue that entrains and synchronizes the central and peripheral clocks, other external cues like exercise and meal timing may significantly affect peripheral clocks (). To investigate whether the poor outcomes caused by environmental light perturbation could be rescued by using a different entrainment cue, we considered a role for food cycling (FC) in the presence of environmental light perturbation. We followed the experimental design in, where mice were infected at either ZT23 or ZT11 and a subset of mice infected at ZT23 were moved to constant light on day 4 after infection. Within the ZT23(LL) group, a subset of mice was subjected to food cycling, wherein food was provided from ZT12 to ZT0 each day for the ZT23(LL-food cycling) group, while the other had access to food ad libitum [ZT23(LL)]; all groups had unrestricted access to water (). Consistent with our previous results, the ZT23(LL) with food ad libitum group had higher mortality than the ZT23 group. As in previous experiments, the ZT23 group had significantly better survival than the ZT11 group (, 87% survival in ZT23 versus 37.5% survival in ZT11;< 0.001 by Mantel-Cox log-rank test). Likewise, the exposure to constant light significantly reduced the survival in the ZT23(LL) group [Figure 7B, 51% survival in ZT23(LL) vs. 86% survival in the ZT23 group,< 0.05 by Mantel-Cox log-rank test]. Interestingly, the mice from the ZT23(LL-food cycling) group showed rescue of the survival advantage associated with infection at ZT23 that had been lost in the ZT23(LL) group [Figure 7B, survival of 81% in ZT23(LL-food cycling) vs. 51% in ZT23(LL);< 0.05, Mantel-Cox Log-rank test]. The mice in the food-cycling group also lost less weight than those in the ZT23(LL) group (). The ZT23(LL-food cycling) groups also exhibited lower clinical scores, suggesting less morbidity than that in the ZT23(LL) group (). These findings led to the conclusion that, while LD cycling and meal timing are important zeitgebers affecting the host's circadian rhythms, the latter is hierarchically a more potent entraining factor for optimal host response to influenza infection. ²⁵ Figure 1A Figure 7A Figure 7B Figure 7C Figure 7D P P P

Our study examined both male and female mice, and similar findings were reported in both sexes (). All experiments were conducted with age- (10–20 weeks) and sex-matched mice. Male and female mice were used in approximately equal proportions for individual experiments unless specified. Given that there were no differences in the overall phenotype with constant light exposure, only female mice were used for the RNA-Seq experiment to minimize any variability. Supplemental Figure 1

Specific pathogen–free 8- to 16-week-old C57BL/6J mice (both males and females) were placed in circadian cabinets on reverse light-dark cycles as described previously (). This allowed us to simultaneously infect mice or harvest tissues across all experimental groups. After 4–6 weeks of acclimatization in these cabinets, mice were anesthetized lightly with isoflurane prior to being infected intranasally with mouse-adapted influenza virus (PR8). Given the use of circadian cabinets, mice were simultaneously infected at either ZT23 or ZT11 and maintained in 12-hour light-dark cycle. The dose chosen was likely to cause >50% mortality when mice were infected at ZT11. Where relevant, on days 4 or 7 p.i., a subset of ZT23-LD mice was moved to constant light until the end of the study. Animals were weighed and scored based on a previously validated scoring criterion each day. For a subset of the ZT23 LL mice, the food was cycled to a 12-hour period, coinciding with their active and rest phases, which had them remain in 12-hour LD conditions. Food trays were manually removed for the food cycling experiments. The food trays in non-food-cycled study groups were handled but not removed to control for any stress associated with this procedure. C57BL/6J mice were either purchased from The Jackson Laboratory or bred in-house. ⁵

Inducible Bmal1mice were generated using 2-month-oldmice treated with 5 mg (in 50 μL) tamoxifen via oral gavage daily for 5 consecutive days ().littermates treated with tamoxifen were used as controls. They were exposed to constant darkness for 1–2 days before infection with IAV at CT23. Both males and females were used in equal proportions in the mentioned experiment. These mice were originally purchased from The Jackson Laboratory and then bred up for experiments. –/– Bmal1ERT2Cre fl/fl + Bmal1Cre fl/fl neg ⁵

Lungs were harvested and homogenized in 0.1% PBS-gelatin. Influenza virus was detected by using MDCK cells (gift from Scott Hensley's group at the University of Pennsylvania; originally purchased from ATCC, PTA-6500). MDCK cells were overlaid with 1:10 dilutions of the homogenized lungs and incubated at 37°C; 5% COfor 1 hour to allow the virus to infect the MDCK cells. Thereafter, 175 μL of tissue culture medium supplemented with 2 μg/mL of TPCK-treated trypsin was added, and cells were incubated for another 72 hours at 37°C; 5% CObefore viral titration. After 72 hours, hemagglutination of turkey red blood cells (RBCs) was tested by collecting 50 μL of the medium from the infected MDCK plate to detect and quantify the virus. The hemagglutination of RBCs confirmed the presence of the virus particles (). 2 2 ⁵

The trachea of mice was cannulated, and lungs were gently lavaged. The supernatant from the first pass was collected and flash-frozen for ELISA. The cells from all 4 passes were pooled and counted using a Nexcelcom cell counter. Cytospins of BAL cells were stained with Hemacolor (Sigma-Aldrich, 65044-93). The lungs were perfused with PBS, followed by inflation with 10% buffered formalin and then fixing in 10% buffered formalin for 24–48 hours. The pathology core at Children's Hospital of Philadelphia paraffin-embedded and stained the lungs for H&E stain. Using a previously validated scoring method (), scoring was performed with scorers remaining blinded to the group assignment. Briefly, 5 lung fields selected at random at ×20 magnification were scored based on a scale of 0–3 for the following: (a) peribronchial infiltrates, (b) perivascular infiltrates, (c) alveolar exudates, and (d) epithelial damage of medium-sized airways. ⁵

Paraffin-embedded lung sections were stained for macrophages and monocytes (F4/80) and lymphocyte (CD3) populations using immunohistochemistry (IHC) by the CHOP pathology core. The primary antibody used for macrophage staining was F4/80 (D2S9R) XP Rabbit mAB (Cell Signaling; 70076) and Recombinant anti-CD3 epsilon antibody (EPR20752) (abcam; ab215212) was used for staining the lymphoid population. All stained slides were digitally scanned at ×20 using an Aperio CS-O slide scanner (Leica Biosystems). The lungs were annotated and analyzed using the Nuclear v9 macros from Aperio Imagescope software for F4/80 and CD3 analysis. Quantitation was performed using the Aperio Imagescope software. Data represent percentage positive nuclei (for F4/80 and CD3) per HPF. +

For immunofluorescence staining, the paraffin-embedded lung sections were stained with primary antibody SFTPC (1:100, anti-pro surfactant protein, rabbit, AB3786 Millipore) at 4°C overnight, followed by secondary antibody donkey anti-rabbit IgG (H+L) Alexa Fluor 488 (A21206; Invitrogen Antibodies, Thermo Fisher) at 1:250 dilution. The slides were incubated with DAPI (0.2 μg/mL) (Thermo Fisher; 1738176) for 10 minutes. VectaMount AQ (Vector Laboratories Inc.; H-5501) was used to mount the slides. The histological and cytological scoring of all the samples was performed in a blinded fashion. To quantify SFTPCcells after influenza, images were captured at ×20 on a Leica DM6 fluorescent microscope, and then cells were counted from >4 random sections per sample. The Cell Counter plug-in on ImageJ (NIH) was used to count SFTPCcells. In total, >1,000 cells were counted from= 3–5 sections per group. + + n

BAL from the first pass as described above on day 6 p.i. or lung homogenate in 0.1% gelatin on day 8 p.i. was sent for ELISA analysis to the Human Immunology Core at the University of Pennsylvania for 9-plex cytokine array analysis. MCP-1, IP-10, IL-2, IL-1β, IL-15, and TNF-α were measured using a custom ELISA kit (Milliplex Mouse Cytokine/Chemokine Magnetic Bead Panel; MCYTOMAG-70K-09C; Millipore SIGMA). Chemokines and cytokines were measured in 25 μL of BAL per sample in duplicates, and the protocol for 9-plex Luminex Performance Assay was performed according to the manufacturer's protocol (Millipore SIGMA). FLEXMAP 3D (Bio-Rad) was used to acquire data. Luminex x Potent 4.2 and Bio-Plex Manager 6.1 software (Bio-Rad) were used to analyze data.

Wireless infrared motion sensors (Actimetrics) were used to measure locomotor activity of singly housed mice before IAV infection. Once the uninfected mice were acclimatized to the circadian cabinets, their activity was measured in 12-hour light-dark cycles for 4 weeks. After 4 weeks, the mice were exposed to constant light for another 3 weeks, and actogram data were collected. Parameters of the animals' rhythms, like amplitude, period, phase, average activity count, and percentage of variance, were analyzed using Clocklab software.

The bulk transcriptomic analyses were performed on female mice only, since sex-stratification confirmed that male and female mice exhibited equivalent circadian patterns in inflammatory () and repair-associated phenotypes (). Similar results were also seen in the current model (). Restricting to a single sex reduced biological variability while preserving interpretability. TRIzol (Life Technologies) was used to extract and isolate RNA from the inferior lobe of the mouse lung. The RNeasy Mini Elute Clean Up Kit (Qiagen) was used to purify the RNA further as per the manufacturer's protocol. The NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies Inc.) was used to assess the quality and quantity of the extracted RNA. Only samples with RINS >7 after QC were sent for sequencing to Azenta Life Sciences. Sequencing was done using an Illumina NovaSeq X with Poly A selection to generate 2x 150 strand-specific paired-end reads. Twenty million paired-end reads were obtained per sample. A mouse reference genome on an in-house resampling-based normalization and quantification pipeline aligned the samples. They were further compared with existing gene annotations (ENSEMBL), and loci and isoforms were identified. A false discovery rate–based control for multiple testing was used to identify differentially expressed genes. Finally, to assess fully the effects on key pathways and mediators, Ingenuity and GSEA were used. ^{5 7} Supplemental Figure 1

Lungs were harvested on day 8 p.i. Nuclear extraction and next-generation sequencing libraries were prepared using the 10x Genomics single cell 3′ v4 (polyA) reagent kit following the manufacturer's protocol at Azenta Life Sciences. Libraries were uniquely indexed, pooled, and sequenced on an Illumina sequencer. Libraries were sequenced in a paired-end, dual-index sequencing run targeting 20,000 mean reads per cell. Demultiplexing and mapping of reads to the mouse mm10 genome to generate feature barcode matrices was performed using the Cell Ranger pipeline (v9.0.0, 10x Genomics) with the "include introns" option. Feature barcode count matrices were corrected for ambient RNA contamination using SoupX (v1.6.2), and adjusted expression matrices were imported into Seurat (v5.0.3) within the R statistical computing environment (v4.4.0). Nuclei expressing less than 7,000 features, between 800 and 40,000 unique molecular indexes, and less than 10% mitochondrial content were retained. Potential doublets were detected and removed using scDblFinder (v1.18.0). The filtered dataset was log normalized, scaled while regressing out mitochondrial content, and the top 2,000 highly variable features were utilized as input for PCA. The top 20 PCs were used downstream in uniform manifold approximation and projection (UMAP) and graph-based clustering using the "FindClusters" function within Seurat and a resolution of 0.8. Subsequent clusters and cell types were annotated using a publicly available lung single-cell RNA-Seq dataset by Niethamer et al. () as a reference within the singleR package (v2.6.0) and compared with additional canonical marker genes. ⁵⁵

To determine the cell-type abundances in the bulk RNA-Seq datasets, we employed a reference sn-RNA-Seq dataset (available in the Gene Expression Omnibus under accession GSE296722) of mouse lung at day 8 after influenza infection. This reference dataset had been processed using the cellranger v9.0.0 pipeline using the mm10 genome, and cell types were annotated using a published mouse lung dataset, also following influenza infection with the same viral strain. The MuSiC method was then used to deconvolve the bulk RNA-Seq datasets by cell type (). Because the MuSiC method uses replicate reference datasets, we randomly subsampled each cell type into 2 "replicate" data types before running MuSiC. After deconvolution, cell types were grouped into endothelial, epithelial, mesenchymal, or immune cells. ²⁴

As described previously, the lungs were digested using DNAse II (Roche) and Liberase (II) at 37°C for 30 minutes after perfusing the lungs with PBS through the right ventricle. The dissociated tissue was then passed through a 70 μm cell strainer, followed by centrifugation and RBC lysis. After washing and resuspending the cells in PBS, 3 × 10cells were blocked with anti-CD16/32 antibody and first stained with Live/Dead fixability dyes for 30 minutes. Following washing, the cells were stained with desired antibodies on ice for 30 minutes. Cells were fixed in 2% paraformaldehyde. Flow cytometric data were obtained using an LSR Fortessa cytometer and analyzed using FlowJo software (Tree Star Inc.). All cells were pregated on size as singlets, live cells. All subsequent gating was done on CD45cells only. 6 +

Mice were either purchased from The Jackson Laboratory or bred in-house. The reported findings summarize the results from 3–6 independent experiments.

GraphPad (Prism V9 and V10) was used for all statistical analyses. In experiments with more than 2 groups and normally distributed data, a 1-way or 2-way ANOVA was performed. Šídák's correction was used for multiple comparisons. When data were not normally distributed, we used Mann-Whitney or Kruskal-Wallis tests. Most data are represented as mean ± SEM, except for data obtained from objective histological scoring, which was presented as median ± interquartile range.values of less than 0.05 were considered significant. P

Approval from the Children's Hospital of Philadelphia Institutional Animal Care and Use Committee was obtained for all animal studies, and the stipulations were met according to the(National Academies Press, 2011). Guide for the Care and Use of Laboratory Animals

The sn-RNA-Seq data are available at the NCBI GEO (accession GSE288858). All other data are available as source data in thefile. Supporting Data Values