Circadian rhythms of macrophages are altered by the acidic tumor microenvironment

Abstract

Tumor-associated macrophages (TAMs) are prime therapeutic targets due to their pro-tumorigenic functions, but varying efficacy of macrophage-targeting therapies highlights our incomplete understanding of how macrophages are regulated within the tumor microenvironment (TME). The circadian clock is a key regulator of macrophage function, but how circadian rhythms of macrophages are influenced by the TME remains unknown. Here, we show that conditions associated with the TME such as polarizing stimuli, acidic pH, and lactate can alter circadian rhythms in macrophages. While cyclic AMP (cAMP) has been reported to play a role in macrophage response to acidic pH, our results indicate pH-driven changes in circadian rhythms are not mediated solely by cAMP signaling. Remarkably, circadian disorder of TAMs was revealed by clock correlation distance analysis. Our data suggest that heterogeneity in circadian rhythms within the TAM population level may underlie this circadian disorder. Finally, we report that circadian regulation of macrophages suppresses tumor growth in a murine model of pancreatic cancer. Our work demonstrates a novel mechanism by which the TME influences macrophage biology through modulation of circadian rhythms.

Keywords: Circadian Rhythms; Immuno-oncology; Immunology; Macrophage; Tumor Microenvironment.

Figure 1. Macrophages of different phenotypes have distinct circadian rhythms.

Bone marrow-derived macrophages (BMDMs) were obtained from C57BL/6 mice expressing PER2-Luc. The circadian clocks of BMDMs were synchronized by a 24-h period of serum starvation in media with 0% serum, followed by a 2-h period of serum shock in media with 50% serum. BMDMs were then stimulated with either 10 ng/mL IL-4 and 10 ng/mL IL-13 (pro-resolution), or 50 ng/mL IFNγ and 100 ng/mL LPS (pro-inflammatory); or left unstimulated. (A) RNA was collected at 6 h post-synchronization, and qt-PCR was performed to assess expression of phenotype-associated genes. Shown are mean and standard error of the mean (SEM), n = 3 biological replicates. (B) Luciferase activity of BMDMs was monitored in real time by LumiCycle. Data was baseline-subtracted using the running average. Oscillation parameters of BMDMs were measured by LumiCycle Analysis. Shown are mean and standard error of the mean (SEM), n = 5 biological replicates. Data information: Statistical significance was determined by unpaired two-tailed t-test with Welch’s correction of unstimulated vs pro-inflammatory or unstimulated vs pro-resolution; *p < 0.05; **p < 0.005; ***p < 0.0005; ****p < 0.0001; ns: not significant. Experiment was replicated twice. Exact p values: (A) Arg1 0.0002 (unstimulated vs pro-resolution), 0.0005 (unstimulated vs pro-inflammatory); Retnla 0.0061 (unstimulated vs pro-resolution), 0.0387 (unstimulated vs pro-inflammatory); Chil3 0.0003 (unstimulated vs pro-resolution), 0.0278 (unstimulated vs pro-inflammatory); Il6 0.0016 (unstimulated vs pro-inflammatory); Clec10a 0.0062 (unstimulated vs pro-resolution); Mrc1 0.0026 (unstimulated vs pro-resolution); Nos2 0.0462 (unstimulated vs pro-inflammatory); Tnfa 0.0006 (unstimulated vs pro-resolution), 0.0034 (unstimulated vs pro-inflammatory); Il1b 0.0029 (unstimulated vs pro-resolution), 0.0016 (unstimulated vs pro-inflammatory). (B) Amplitude <0.0001 (unstimulated vs pro-resolution), <0.0001 (unstimulated vs pro-inflammatory); Period 0.0018 (unstimulated vs pro-resolution), 0.0299 (unstimulated vs pro-inflammatory); Damping 0.0059 (unstimulated vs pro-resolution), <0.0001 (unstimulated vs pro-inflammatory).

Figure 2. Acidic pH alters circadian rhythms of bone marrow-derived macrophages in vitro.

(A–F) Bone marrow-derived macrophages (BMDMs) were obtained from C57BL/6 mice expressing PER2-Luc. The circadian clocks of BMDMs were synchronized by a 24-h period of serum starvation in media with 0% serum, followed by a 2-h period of serum shock in media with 50% serum. BMDMs were then cultured in media with neutral pH 7.4 or acidic media with pH 6.8 or 6.5, and stimulated with either (B, E) 10 ng/mL IL-4 and 10 ng/mL IL-13 (pro-resolution), or (C, F) 50 ng/mL IFNγ and 100 ng/mL LPS (pro-inflammatory); or (A, D) left unstimulated. (A–C). Luciferase activity was monitored in real time by LumiCycle, n = 2 biological replicates. Data from both experiments was baseline-subtracted using the running average, and oscillation parameters were measured by LumiCycle Analysis, n = 5 biological replicates. (D–F). In parallel, RNA was collected at 12, 16, 20, and 24 h post-synchronization, and qt-PCR was performed to assess oscillation of transcripts encoding core clock proteins in macrophages under acidic conditions. n = 3 biological replicates. Data information: For (A–C), top panel (LumiCycle traces), shown are mean and SEM, experiment was replicated twice. For bottom panel (LumiCycle parameter analysis), shown are mean and SEM. For (D–F), shown are mean and SEM. Experiment was replicated twice. For all panels, statistical significance was determined by unpaired two-tailed t-test with Welch’s correction of pH 7.4 vs pH 6.8 or pH 6.5 (pH 7.4 vs 6.8 was not tested in D–F). The Holm-Šídák correction for multiple t-tests was applied; *p < 0.05; **p < 0.005; ***p < 0.0005; ****p < 0.0001; ns: not significant. Exact p values: (A). Amplitude <0.0001 (pH 7.4 vs 6.8), 0.0025 (pH 7.4 vs 6.5); Period 0.0243 (pH 7.4 vs 6.8), 0.0184 (pH 7.4 vs 6.5); Damping 0.0098 (pH 7.4 vs 6.8), 0.0035 (pH 7.4 vs 6.5). (B) Amplitude <0.0001 (pH 7.4 vs 6.8), 0.0014 (pH 7.4 vs 6.5); Period 0.0039 (pH 7.4 vs 6.8), 0.002 (pH 7.4 vs 6.5); Damping 0.0022 (pH 7.4 vs 6.5). (C) Amplitude <0.0001 (pH 7.4 vs 6.8), <0.0001 (pH 7.4 vs 6.5); Period 0.0073 (pH 7.4 vs 6.8), 0.0517 (pH 7.4 vs 6.5); Damping 0.046 (pH 7.4 vs 6.5). (D) Cry1 0.005321 (pH 7.4 vs pH 6.5 CT 16), 0.03417 (pH 7.4 vs pH 6.5 CT 24). (E) Cry1 0035141 (pH 7.4 vs pH 6.5 CT 16), 0.002073 (pH 7.4 vs pH 6.5 CT 24); Nr1d1 0.025736 (pH 7.4 vs pH 6.5 CT 12), 0.036436 (pH 7.4 vs pH 6.5 CT 20). (F) Cry1 0.014339 (pH 7.4 vs pH 6.5 CT 12), 0.005812 (pH 7.4 vs pH 6.5 CT 16), 0.000525 (pH 7.4 vs pH 6.5 CT 24); Nr1d1 0.062514 (pH 7.4 vs pH 6.5 CT 12), 0.062514 (pH 7.4 vs pH 6.5 CT 16), 0.010905 (pH 7.4 vs pH 6.5 CT 24). Some ‘ns’ have been omitted from (D–F) for visual clarity.

Figure 3. Acidic pH alters circadian rhythms of peritoneal macrophages ex vivo at temporally distinct times of day.

(A) Peritoneal macrophages were obtained at ZT0 or ZT12 from C57BL/6 mice expressing Per2-Luc and cultured in media with neutral pH 7.4 or acidic pH 6.5. (A) Luciferase activity was monitored in real time by LumiCycle. Data was baseline-subtracted using the running average; n = 2 biological replicates, each replicate and connecting line shown. Oscillation parameters were measured by LumiCycle Analysis; shown is the mean and SEM, n = 4 biological replicates, data pooled from 2 independent experiments. (B) Clock gene expression, in transcripts per million (TPM), of peritoneal macrophages cultured in media at pH 7.4 or pH 6.8 for 24 h, sourced from publicly available data (Data ref: GSE164697), n = 3 biological replicates [30]. (C) The magnitude of change in circadian oscillation parameters from (A) between macrophages at pH 7.4 and pH 6.5 was compared between peritoneal macrophages taken at ZT0 or ZT12. n = 4 biological replicates; data pooled from 2 independent experiments. Data information: For (A), shown are individual points and mean, experiment was replicated twice. For (B), shown is mean and SEM, For (C), shown is the mean and SEM experiment replicated twice. For all panels, statistical significance was determined by paired two-tailed t-test with Welch’s correction; *p < 0.05; **p < 0.005; ***p < 0.0005; ns: not significant. Exact p values: (A) Amplitude 0.0231 (ZT0 pH 7.4 vs pH 6.5), 0.0204 (ZT12 pH 7.4 vs pH 6.5); Period 0.0003 (ZT0 pH 7.4 vs pH 6.5), 0.0043 (ZT12 pH 7.4 vs pH 6.5); Damping 0.0104 (ZT0 pH 7.4 vs pH 6.5), 0.0317 (ZT12 pH 7.4 vs pH 6.5). (B) Arntl 0.0103; Cry1 < 0.0001; Cry2 0.0004; Tef 0.0093; Per1 0.0009; Per2 0.0001; Per3 0.0017; Nr1d1 < 0.0001; Nr1d2 0.0012; Dbp < 0.0001.

Figure 4. Lactate alters circadian rhythms in macrophages, both alone and in conjunction with acidic pH.

(A–C) Bone marrow-derived macrophages (BMDMs) were obtained from C57BL/6 mice expressing Per2-Luc. The circadian clocks of BMDMs were synchronized by a 24-h period of serum starvation in media with 0% serum, followed by a 2-h period of serum shock in media with 50% serum. BMDMs were then cultured in media with neutral pH 7.4 or acidic pH 6.5, supplemented with 0 mM or 25 mM sodium-L-lactate. (A) RNA was collected at 6 h post-treatment, and expression of pro-resolution phenotype markers Arg1 or Vegf was quantified by qtPCR. n = 6 biological replicates. (B) Luciferase activity was monitored in real time by LumiCycle. Data was baseline-subtracted using the running average, n = 4 biological replicates. Oscillation parameters were measured by LumiCycle Analysis. Data was pooled from 2 replicated experiments (pH 7.4, 0 mM sodium-L-lactate: n = 10. pH 7.4, 25 mM sodium-L-lactate, n = 10. pH 6.5, 0 mM sodium-L-lactate: n = 7. pH 6.5, 25 mM sodium-L-lactate, n = 7, all replicates were biological replicates). (C) RNA was collected at 12, 16, 20, and 24 h post-synchronization, and qt-PCR was performed to assess oscillation of transcripts encoding core clock proteins in macrophages exposed to lactate. Data was pooled from 2 independent replicated experiments. (pH 7.4, 0 mM sodium-L-lactate: n = 6. pH 7.4, 25 mM sodium-L-lactate, n = 6. pH 6.5, 0 mM sodium-L-lactate: n = 5. pH 6.5, 25 mM sodium-L-lactate, n = 6, all replicates were biological replicates). Data information: For (A), shown is the mean and SEM, experiment replicated twice. For (B), top panel (LumiCycle traces), shown is the mean and SEM, experiment was replicated twice. For bottom panel (LumiCycle parameter analysis), shown is the mean and SEM. For (C), shown is the mean and SEM. For all panels, statistical significance was determined by unpaired two-tailed t-test with Welch’s correction for pH 7.4 0 mM sodium-L-lactate vs 25 mM mM sodium-L-lactate and pH 6.5 0 mM sodium-L-lactate vs 25 mM sodium-L-lactate (other comparisons were not performed on these data). The Holm-Šídák correction for multiple t-tests was applied; *p < 0.05; **p < 0.005; ***p < 0.0005; ****p < 0.0001; ns: not significant. Exact p values: (A) Vegf 0.0045. (B) Amplitude 0.0002 (pH 7.4 0 mM sodium-L-lactate vs pH 6.5 0 mM sodium-L-lactate), 0.0015 (pH 7.4 25 mM sodium-L-lactate vs pH 6.5 25 mM sodium-L-lactate); Period <0.0001 (pH 7.4 0 mM sodium-L-lactate vs pH 6.5 0 mM sodium-L-lactate), <0.0001 (pH 7.4 0 mM sodium-L-lactate vs pH 7.4 25 mM sodium-L-lactate), 0.0209 (pH 7.4 25 mM sodium-L-lactate vs pH 6.5 25 mM sodium-L-lactate), 0.0008 (pH 6.5 0 mM sodium-L-lactate vs pH 6.5 25 mM sodium-L-lactate); Damping 0.0003 (pH 7.4 0 mM sodium-L-lactate vs pH 6.5 0 mM sodium-L-lactate), <0.0001 (pH 7.4 0 mM sodium-L-lactate vs pH 7.4 25 mM sodium-L-lactate), 0.0036 (pH 6.5 0 mM sodium-L-lactate vs pH 6.5 25 mM sodium-L-lactate). (C) Cry1 0.04153 (pH 7.4 0 mM sodium-L-lactate vs pH 7.4 25 mM sodium-L-lactate CT 20), 0.00594 (pH 7.4 0 mM sodium-L-lactate vs pH 7.4 25 mM sodium-L-lactate CT 24), 0.028433 (pH 6.5 0 mM sodium-L-lactate vs pH 6.5 25 mM sodium-L-lactate CT 24); Nr1d1 0.015989 (pH 7.4 0 mM sodium-L-lactate vs pH 7.4 25 mM sodium-L-lactate CT 20), 0.003984 (pH 6.5 0 mM sodium-L-lactate vs pH 6.5 25 mM sodium-L-lactate CT 20). Some ‘ns’ have been omitted from (C) for visual clarity.

Figure 5. pH-induced changes in circadian rhythms are not driven by cAMP signaling alone.

(A, B) Bone marrow-derived macrophages (BMDMs) were obtained from C57BL/6 mice expressing Per2-Luc. BMDMs were cultured in media with neutral pH 7.4 or acidic pH 6.5, and (A) treated with vehicle or 20, 40, or 80 μM Forsokolin, or (B) treated with vehicle or 20, 40, 80 μM IBMX. Luciferase activity was monitored in real time by LumiCycle. Data was baseline-subtracted using the running average, n = 2 biological replicates. (C) The circadian clocks of BMDMs were synchronized by a 24-h period of serum starvation in media with 0% serum, followed by a 2-h period of serum shock in media with 50% serum. BMDMs were then cultured in media with neutral pH 7.4 or acidic pH 6.5, and treated with vehicle or 5, 10, or 15 μM MDL-12. Luciferase activity was monitored in real time by LumiCycle. Data was baseline-subtracted using the running average, and oscillation parameters were measured by LumiCycle Analysis, n = 4 biological replicates. (D) Expression of the acid sensing gene (Icer) was measured at 2 h post treatment (CT2), and gene associated with a pro-resolution phenotype (Arg1) was measured at 6 h post-treatment (CT6), n = 3 biological replicates. (E, F). (E) Phosphorylation of CREB was assessed by immunoblot at the indicated timepoints after synchronization, MDL-12 treatment, and exposure to low pH. (F) Relative pCREB levels from (E) were quantified; band intensity was normalized to total protein, n = 3 biological replicates. Data information: For (A, B), shown are individual points and mean. For (C, D, F), mean and SEM are shown. Each experiment was replicated at least two times. For (C, right panels), statistical significance was determined with two-way ANOVA to compare all pH 7.4 vs pH 6.5 data points, and for each individual MDL-12 comparison, unpaired two-tailed t-tests with Welch’s correction and the Holm-Šídák correction for multiple t-tests were applied. For (D, F), statistical significance was determined by unpaired two-tailed t-test with Welch’s correction (pH 7.4 vehicle vs pH 6.5 15 µM MDL-12 was not tested); *p < 0.05; **p < 0.005; ***p < 0.0005; ****p < 0.0001; ns: not significant. Exact P values: Amplitude 0.002346 (pH 7.4 vs pH 6.5 vehicle), 0.002346 (pH 7.4 vs pH 6.5 5 µM MDL-12), 0.01276 (pH 7.4 vs pH 6.5 10 µM MDL-12); Period <0.0001 (pH 7.4 vs pH 6.5 by two-way ANOVA), 0.005197 (pH 7.4 vs pH 6.5 vehicle), 0.003662 (pH 7.4 vs pH 6.5 5 µM MDL-12), 0.005197 (pH 7.4 vs pH 6.5 10 µM MDL-12), 0.005197 (pH 7.4 vs pH 6.5 15 µM MDL-12); Damping <0.0001 (pH 7.4 vs pH 6.5 by two-way ANOVA), 0.000307 (pH 7.4 vs pH 6.5 vehicle), 0.00119 (pH 7.4 vs pH 6.5 5 µM MDL-12), 0.000105 (pH 7.4 vs pH 6.5 10 µM MDL-12), 0.000307 (pH 7.4 vs pH 6.5 15 µM MDL-12). (D) Icer 0.0053 (pH 7.4 vehicle vs pH 6.5 vehicle); Arg1 0.0148 (pH 7.4 vehicle vs pH 7.4 15 µM MDL-12), 0.0117 (pH 7.4 vehicle vs pH 6.5 vehicle), 0.0202 (pH 6.5 vehicle vs pH 6.5 15 µM MDL-12). (F) 10 min 0.01823 (pH 7.4 vs pH 6.5 0 µM MDL-12); 2 hr 0.03913 (pH 7.4 0 µM MDL-12 vs pH 7.4 15 µM MDL-12); 6 hr 0.036106 (pH 7.4 0 µM MDL-12 vs pH 7.4 15 µM MDL-12).

Figure 6. Clock correlation distance (CCD) and weighted gene co-expression network analysis (WGCNA) provide evidence of circadian disorder in murine tumor-associated macrophages.

(A) RNAseq datasets (see Methods) of WT peritoneal macrophages (pMac), BMAL1 KO peritoneal macrophages, and tumor-associated macrophages (TAMs) were analyzed for expression of Arg1 and Crem (WT pMac: n = 5. BMAL1KO pMac: n = 4. TAMs: n = 10, all replicates are biological replicates). (B) A pan-tissue murine reference control was used for clock correlation distance (CCD). (C, D). (C) Clock correlation distance (CCD) analysis was performed and (D) statistical analysis to compare CCD scores was performed by calculating delta CCD. For BMDMs, n = 72 biological replicates, for all other conditions, see (A). (E) Weighted gene co-expression network analysis (WGCNA) was performed on the indicated circadian clock genes using data from (A–D). Data information: For (A), mean and SEM are shown. Statistical significance was determined with Ordinary one-way ANOVA with Tukey’s multiple comparisons test, ****p < 0.0001 for each indicated comparison. For (D), *p < 0.05 by delta CCD analysis compared to control group only (pMacs WT). For (E), *p < 0.05 by WGCNA for significant covariance by Pearson correlation, and blank squares are not significant. Exact p values: (D) 0.00802 (pMacs WT vs pMacs BMAL1KO), 0.007826 (pMacs WT vs TAMs).

Figure 7. Clock correlation distance (CCD) and weighted gene co-expression network analysis (WGCNA) provide evidence of circadian disorder in human tumor-associated macrophages.

(A) Clock correlation distance (CCD) analysis was performed using RNAseq datasets of macrophages from tumor (TAMs, n = 40 patients), tumor-adjacent tissue from NSCLC patients (n = 34 patients), and alveolar macrophages from healthy donors (n = 24 donors) (see Methods for information on these samples). TAM samples were subset by median Crem expression into Crem high TAM samples (TAMs Crem high, n = 20 patients) and Crem low TAM samples (TAMs Crem low, n = 20 patients). (B) Statistical analysis to compare CCD scores was performed by calculating delta CCD, with p < 0.05 being deemed significantly different from the control group. (C) Weighted gene co-expression network analysis (WGCNA) was performed on the indicated circadian clock genes using data from (A, B). Data information: For (B), p value shown is by delta CCD analysis compared to control group. For (C), *p < 0.01 by WGCNA for significant covariance by Pearson correlation, and blank squares are not significant.

Figure 8. Heterogeneity in circadian rhythms of cells within a population can lead to circadian disorder observed by CCD.

(A–C) Increasingly desynchronized populations were modeled using an RNAseq data set of WT peritoneal macrophages (n = 12 biological replicates) taken at 4-h intervals across two days (see Methods). (A) A schematic of the populations used in experimental design. (B, C) (B) Clock correlation distance (CCD) analysis was performed and (C) statistical analysis to compare CCD scores was performed by calculating delta CCD. Data information: *p < 0.05 by delta CCD analysis compared to control group. Exact p values: (C) 0.016442 (Synchronized vs 12 h Desynchronized).

Figure 9. There is heterogeneity in expression of circadian clock genes within the tumor-associated macrophage population.

We analyzed a single-cell RNAseq dataset of tumor-associated macrophages (see Methods for information on these samples). (A) Unbiased clustering was performed to identify TAM subpopulations. (B) Differential gene expression analysis was performed on these TAM clusters, and expression of significantly different circadian genes was plotted along with Crem. (C) Crem expression of macrophages in TAM clusters was measured. (D) TAMs were subset by Crem expression into Crem high TAMs and Crem low TAMs, and these groups were overlaid on the UMAP plot shown in (A). (E) Differential gene expression analysis was performed on Crem high vs Crem low, and expression of significantly different circadian genes was plotted. Data information: All differentially expressed genes or groups of clusters in (B), (C), and (E) are adj.p < 0.005 by limma implementation of the Wilcoxon Rank-Sum test. Exact P value: (C) 2.71 E-27 (Crem-low clusters vs Crem-high clusters). Exact P values for (B, E) are available in Appendix Fig. S4.

Figure 10. A functional circadian clock in macrophages can influence tumor growth in a murine model of PDAC.

(A) Levels of Bmal1 in bone marrow-derived macrophages (BMDMs) from WT or BMAL1 KO mice were assessed by immunoblot. (B) To confirm functional disruption of the circadian clock, peritoneal macrophages or BMDMs were obtained from WT or BMAL1 KO mice expressing PER2-Luc and cultured in vitro with D-luciferin. Luciferase activity was monitored in real time by LumiCycle, n = 2 biological replicates. (C, D) (C) Illustration and (D) results of in vivo tumor growth experiment. Bone marrow-derived macrophages (BMDMs) obtained from WT or BMAL1 KO mice were subcutaneously co-injected with KCKO cells into the flank of WT mice. Tumor growth was measured by caliper, n = 20 individual mice, 10 male and 10 female. Data information: For (A, B), experiment was replicated twice, and for (B), shown are individual points and mean. For (D), shown is the mean and SEM. Statistical significance determined at each time point by multiple Mann–Whitney tests with two-stage step-up (Benjamini, Krieger, and Yekutieli) correction for multiple testing; *p < 0.05; **p < 0.005; ***p < 0.0005 with q < 0.05 for all tests; ns: not significant. Experiment was replicated twice. Exact p values: (D) 0.00321 (16 days), 0.000696 (24 days), 0.016898 (26 days).

Figure EV1. The PER2-Luciferase reporter system enables real-time monitoring of circadian rhythms of macrophages.

(A) A schematic of the Per2-Luciferase (Per2-Luc) luciferase reporter system. (B) A schematic of the synchronization protocol in which the circadian clocks of bone marrow-derived macrophages (BMDMs) derived from C57BL/6 mice expressing Per2-Luc were synchronized by a 24-h period of serum starvation in media with 0% serum, followed by a 2-h period of serum shock in media with 50% serum. (C) BMDMs were then cultured in RPMI/10% FBS supplemented with D-luciferin at circadian time (CT) 0. Luciferase activity of BMDMs was monitored in real time by LumiCycle, n = 2 biological replicates. (D, E) Protein and RNA were collected at the indicated times post-serum shock to assess (D) cAMP signaling by p-CREB levels and (E) expression of Icer, n = 3 biological replicates. Data information: For (C–E), experiments were replicated twice. For (C), shown are individual points and mean. For (E), mean and SEM are shown, and statistical significance determined by unpaired two-tailed t-test with Welch’s correction; *p < 0.05; **p < 0.005; ***p < 0.0005; ****p < 0.0001; ns: not significant. Exact p values: (E) 0.0039 (0 h vs 1 h), 0.0006 (0 h vs 2 h), <0.0001 (0 h vs 4 h).

Figure EV2. Macrophages sense and respond to an acidic extracellular environment when cultured in vitro in media with acidic pH.

(A, B) Bone marrow-derived macrophages (BMDMs) were obtained from C57BL/6 mice expressing PER2-Luc. BMDMs were cultured in media with pH 7.4 or acidic media with pH 6.5, and stimulated with either 10 ng/mL IL-4 and 10 ng/mL IL-13 (pro-resolution), or 50 ng/mL IFNγ and 100 ng/mL LPS (pro-inflammatory); or left unstimulated. RNA was collected at 2 h post-treatment, and qt-PCR was performed to assess expression of genes associated with (A) phenotype or (B) acid sensing in macrophages. For both panels, n = 3 biological replicates. Data information: Shown are mean and SEM. Statistical significance determined by two-tailed t-test with Welch’s correction. The Holm-Šídák correction for multiple t-tests were applied; *p < 0.05; **p < 0.005; ***p < 0.0005; ****p < 0.0001; ns: not significant. Experiments were replicated twice. Exact p values: (A) Arg1 0.024372 (pH 7.4 vs pH 6.5 unstimulated); Vegf 0.000403 (pH 7.4 vs pH 6.5 unstimulated), 0.011077 (pH 7.4 vs pH 6.5 pro-resolution), 0.048413 (pH 7.4 vs pH 6.5 pro-Inflammatory); Nos2 0.008509 (pH 7.4 vs pH 6.5 pro-inflammatory). (B) Icer < 0.0001 (pH 7.4 vs pH 6.5 unstimulated), 0.01749 (pH 7.4 vs pH 6.5 pro-resolution).

Figure EV3. Exposure to cancer cell supernatant further modulates circadian rhythms in addition to pH-driven changes.

Bone marrow-derived macrophages (BMDMs) were obtained from C57BL/6 mice expressing Per2-Luc. The circadian clocks of BMDMs were synchronized by a 24-h period of serum starvation in media with 0% serum, followed by a 2-h period of serum shock in media with 50% serum. BMDMs were then cultured in RPMI with neutral pH 7.4 or acidic pH 6.5, or in KCKO supernatant at pH 6.5 or pH-adjusted to pH 7.4. Luciferase activity was monitored in real time by LumiCycle. Data was baseline-subtracted using the running average, and oscillation parameters were measured by LumiCycle Analysis, n = 4 biological replicates. Data information: Shown is the mean and SEM. Statistical significance determined by unpaired two-tailed t-test with Welch’s correction; *p < 0.05; **p < 0.005; ***p < 0.0005; ****p < 0.0001; ns: not significant. Exact p values: Amplitude 0.0216 (pH 7.4 RPMI vs pH 7.4 KCKO sup), 0.0008 (pH 7.4 RPMI vs pH 6.5 RPMI), 0.0105 (pH 7.4 KCKO sup vs pH 6.5 KCKO sup), 0.0007 (pH 6.5 RPMI vs pH 6.5 KCKO sup); Period <0.0001 (pH 7.4 RPMI vs pH 7.4 KCKO sup), 0.0002 (pH 7.4 RPMI vs pH 6.5 RPMI), 0.0004 (pH 7.4 KCKO sup vs pH 6.5 KCKO sup), <0.0001 (pH 6.5 RPMI vs pH 6.5 KCKO sup); Damping <0.0001 (pH 7.4 RPMI vs pH 7.4 KCKO sup), 0.0002 (pH 7.4 RPMI vs pH 6.5 RPMI).

Figure EV4. Acidic pH alters circadian rhythms in macrophages in the absence of prior serum starvation followed by serum shock.

Bone marrow-derived macrophages (BMDMs) were obtained from C57BL/6 mice expressing Per2-Luc. BMDMs were cultured in media with neutral pH 7.4 or acidic pH 6.8 or 6.5. Luciferase activity was monitored in real time by LumiCycle (pH 7.4: n = 3. pH 6.8: n = 2. pH 6.5: n = 3). Data was baseline-subtracted using the running average, and oscillation parameters were measured by LumiCycle Analysis (pH 7.4: n = 5. pH 6.8: n = 4. pH 6.5: n = 5, data pooled from 2 individual experiments). Data information: For LumiCycle traces, shown are individual points and mean. For LumiCycle analysis, shown is the mean and SEM. Statistical significance determined by unpaired two-tailed t-test with Welch’s correction (pH 6.8 and 6.5 comparison not tested); **p < 0.005; ***p < 0.0005; ****p < 0.0001. Exact p values: Amplitude 0.0035 (pH 7.4 vs pH 6.8), 0.0004 (pH 7.4 vs pH 6.5); Period 0.0044 (pH 7.4 vs pH 6.8), <0.0001 (pH 7.4 vs pH 6.5).

Figure EV5. Heterogeneity in circadian rhythms of cells within a population can lead to an altered circadian clock gene network in samples.

Increasingly desynchronized populations were modeled using an RNA-seq data set of WT peritoneal macrophages taken at 4-h intervals across two days, n = 12 biological replicates (see Fig. 8 and Methods). Weighted gene co-expression network analysis (WGCNA) was performed. Data information: *p < 0.01 by WGCNA for significant covariance by Pearson correlation, and blank squares are not significant.