Proteostasis is essential for cellular health, with its disruption contributing to aging, neurodegeneration, and metabolic disorders. While the upstream proteotoxic stress-sensing and protein-folding mechanisms in the ER and cytosol are well studied, the transcriptional regulation of proteostasis remains incompletely understood, particularly concerning the temporal epigenome dynamics, chromatin landscapes, and co-regulatory networks underlying dynamic proteostasis control. Traditionally, proteotoxic stress responses were viewed as acute reactions to noxious stimuli, but recent evidence shows that many proteostasis genes exhibit ~12-hour ultradian rhythms under physiological conditions, driven by an XBP1s-dependent oscillator independent of the canonical circadian clock and cell cycle. By mapping the chromatin landscape of the murine 12-hour oscillator, we identified RBBP5—an essential subunit of the COMPASS complex responsible for H3K4 trimethylation—as a pivotal epigenetic regulator of proteostasis dynamics. In contrast, histone acetyltransferases and H3K9 acetylation were dispensable for dynamic proteostasis gene expression. RBBP5 is not only indispensable for the 12-hour oscillator but also essential for the transcriptional regulation of diverse proteotoxic stresses response, by coactivating XBP1s and promoting H3K4me3 deposition at the promoters of proteostasis genes. As a result, loss of RBBP5 sensitizes cells to proteotoxic stress in part due to impaired autophagy. Proximity labeling of H3K4me3 further uncovered a dynamic chromatin-associated proteomic architecture, including components of COMPASS, the Integrator complex and SWI/SNF remodelers, that constitutes the transcriptional response to proteotoxic stress. Together, these findings establish RBBP5 as a central regulator of proteostasis dynamics, essential for maintaining cellular resilience.

12 h RBBP5 chromatin binding is concomitant with the promoter-proximal ~12 h H3K4me3 oscillation in mouse liver

The 12 h oscillator is transcriptionally regulated by the UPR TF XBP1s in mouse liver [14,17,30], and likely under control by additional bZIP TFs, such as ATF4 and ATF6 [23]. However, decades of research on eukaryotic gene regulation demonstrated that TFs themselves are not sufficient to drive transcription initiation and elongation, and additional so-called “co-regulators” are required for full gene activation by shaping the epigenetic landscape [39–41]. Currently, the epigenetic landscape of both the 12 h oscillator and proteotoxic stress responses is poorly characterized. To identify epigenetic regulators of global proteostasis control, we first examined a published temporal hepatic epigenome dataset for histone modifications associated with the active promoters of ~12 h genes [42]. We found ~12 h rhythms of histone H3 trimethylated at lysine 4 (H3K4me3) (P = 0.024 by the cosinor2 method [43]), but not two histone acetylation markers, histone H3 acetylated at lysine 9 (H3K9Ac) and lysine 27 (H3K27Ac), at the promoters of ~12 h genes (Fig 1A and 1B). This is particularly prominent for those ~12 h genes with rhythmic expressions peaking at ZT0 and ZT12 that are enriched for proteostasis pathways [17], with H3K4me3 levels also peaking at these two time points (Fig 1A and 1B).

In eukaryotes, H3K4 is tri-methylated by the COMPASS complex, which includes regulatory subunits RBBP5, ASH2L, WDR5, DPY30, and catalytic subunits SETD1A/B/MLL1-4 (S1A Fig) [44–46]. Via writing H3K4me3, COMPASS is essential for the eviction of paused RNA polymerase II (RNAPII) and promotes transcriptional elongation [47,48]. Landscape In Silico deletion Analysis (LISA) [49] further inferred COMPASS subunits WDR5, ASH2L, DYP30, and RBBP5 as putative epigenetic regulators of hepatic ~12 h, but not circadian genes (S1B Fig). Being the commonly shared regulatory subunits for all MLL family H3K4 methyl transferase complex (including SETD1A/B and MLL1-4), RBBP5/ASH2L forms the minimal heterodimer sufficient to activate all MLL histone methyl transferase [50,51] and is essential for the assembly of the whole COMPASS complex [46,52]. Between the two essential subunits, we initially focused on RBBP5. Eigenvalue/pencil method [53] revealed that the levels of both hepatic Rbbp5 mRNA and nuclear RBBP5 protein exhibit ~12 h oscillations, with the nuclear protein level peaking at ~CT0 and ~CT12 (S1C and S1D Fig). These phases closely align with the acrophases of H3K4me3 and rhythmic gene expression (Fig 1A and 1B), supporting coordinated temporal regulation. The ~12 h oscillation of nuclear RBBP5 was further validated by four independent statistical approaches (S1D Fig), underscoring the robustness of this rhythmic pattern. Beyond the liver, pronounced ~12 h rhythms in total RBBP5 protein were also observed across multiple mice tissues—including heart, inguinal white adipose tissue (iWAT), kidney, and several brain regions (cortex, hypothalamus, and cerebellum)—with peak expression consistently occurring at CT10–14 and CT22–26 in all tissues except iWAT (S1E and S1F Fig) [54]. Together, these findings suggest that RBBP5 may be a conserved component of the mammalian ~12 h oscillator.

To further evaluate the potential role of RBBP5 in establishing the epigenetic landscape of hepatic ~12 h rhythms, we performed temporal RBBP5 ChIP-seq at a 4-h interval for a total of 48 h in the liver of male C57BL/6 mice housed under constant darkness condition fed ad libitum. Importantly, two fully independent ChIP-seq experiments were conducted using separate biological samples, with chromatin immunoprecipitation, library preparation, and sequencing carried out independently for each run. Chromatin occupancies of RBBP5 align well with nucleosome positioning downstream of the transcription start sites (TSS) and the H3K4me3 epigenome, with ‘ridges’ corresponding to 130 bp, 310 bp, 490 bp, 670 bp downstream of TSS at all time points, thus demonstrating the high quality of our ChIP-seq dataset (Fig 1C–1F). In total, we identified 7,963 hepatic RBBP5 binding sites that are within 2kb (up and downstream) of TSS of 6,451 genes (Fig 1C and S1 Table). The average binding intensities of RBBP5 displayed a robust 12 h rhythm (P = 0.0016 with RAIN [55]), with peaks occurring at CT0, CT12, CT24, and CT36 (Figs 1C–1E and S2A). This 12 h rhythm aligns with the 12 h oscillation of promoter-proximal H3K4me3 (Fig 1C–1F), but not with those of H3K9Ac or H3K27Ac (S2B Fig). This pattern is exemplified by RBBP5 promoter binding to Xbp1, one of the major UPR TFs and the transcriptional regulator of the 12 h oscillator, and to Hsph1, a canonical 12 h oscillator output gene encoding a heat shock protein (S2C Fig). Using the RAIN [55] algorithm, we identified a total of 4,667 (adjP < 0.05) RBBP5 binding sites that cycle with a ~12 h period (S2D–S2F Fig and S1 Table). Collectively, these results establish the epigenetic landscape of hepatic 12 h oscillator is marked by RBBP5-H3K4me3.

Hepatic RBBP5 cistrome coincides with that of XBP1s and hepatic ~12 h transcriptome

To investigate if RBBP5 regulates hepatic ~12 h rhythms of gene expression, we first examined the correlation between RBBP5 proximal promoter binding status and ~12 h rhythms of gene expression. As illustrated in Fig 2A, genes with proximal promoter RBBP5 binding are enriched for ~12 h rhythms of gene expression (P = 3.46e−66 by Chi-squared test). In addition, RBBP5 chromatin binding exhibits more robust ~12 h rhythmicity at sites associated with 12 h cycling genes compared to sites linked to genes lacking ~12 h rhythmicity (Fig 2B–2D). For those 2,527 ~12 h genes with RBBP5 binding, the phases of RBBP5 chromatin binding precede the phases of gene expression (P < 10e−15 by paired t test) (Fig 2B). This temporal relationship suggests that RBBP5 likely acts as a driver, rather than a consequence of ~12 h rhythms of gene expression.

We next performed Gene Ontology (GO) analysis on those hepatic genes with proximal promoter RBBP5 binding and revealed strong enrichment of proteostasis pathways including protein processing in the ER and ubiquitin-mediated proteolysis (Figs 2E and S3). Additionally, pathways involved in mRNA processing, such as mRNA splicing and mRNA surveillance, were also enriched (Figs 2E and S3). These findings align with previous observations that both mRNA and protein metabolism pathways are enriched in genes exhibiting ~12 h rhythms in both mice and humans [18,25,56]. The coupling of mRNA and protein metabolism is orchestrated by a SON-XBP1s axis and involves the liquid-liquid phase separation (LLPS) dynamics of nuclear speckles [25]. Similar pathways are also enriched in genes exhibiting both ~12 h rhythms of RBBP5 promoter binding and gene expression (Figs 2E and S3).

XBP1s is a major transcriptional factor regulating the hepatic 12 h oscillator and also exhibits a global 12 h rhythm of chromatin binding to the promoter regions peaking at CT0, 12, 24, and 36 [14,17]. Comparing the cistromes of RBBP5 and XBP1s revealed a significant overlap between the two, with 920 genes sharing RBBP5 and XBP1s proximal promoter binding (Fig 2F). Co-immunoprecipitation in the hepatic nuclear extracts at CT0 confirmed the physical interaction between XBP1s and subunits of SET1/COMPASS, including RBBP5, ASH2L, and WDR5 (Fig 2G). Motif analysis of the DNA sequences around the 6,149 RBBP5 binding sites that do not overlap with XBP1s cistrome identified binding motifs for GABPA, KLF, ATF6, NFYA, and ATF4 (Fig 2H). These findings suggest that, in addition to acting as a co-activator for XBP1s, RBBP5 may also work with other transcription factors to shape the hepatic ~12 h epigenome.

RBBP5 is an epigenetic regulator of the hepatic ~12 h oscillator, but not the BMAL1-dependent canonical core circadian clock

To establish the causality between RBBP5 and hepatic 12 h rhythms of gene expression, we generated RBBP5 liver hepatocyte-specific knockout (RBBP5 ^LKO) mice using the CRE-loxP system. Exon 10 of mouse Rbbp5, which encodes the SET/ASH2L binding domain, was flanked by loxP sites (Fig 3A). CRE-mediated deletion of Exon 10 is expected to lead to a frameshift and nonsense-mediated RNA decay of truncated Rbbp5 transcript (S4A Fig). Crossing homozygous floxed Rbbp5 mice with Albumin-CRE mice resulted in liver hepatocyte-specific deletion of RBBP5, which was confirmed by DNA genotyping, immunofluorescence (IF) against RBBP5 and western blot analysis (Figs 3B and S4B–S4D). The resulting RBBP5 ^LKO mice weighed slightly less than wild-type counterparts (S5A Fig) but maintained normal rhythmic locomotor activity and fasting-feeding cycles under both 12 h:12 h light/dark and constant darkness conditions (S5B–S5I Fig).

To identify the RBBP5-dependent oscillating hepatic transcriptome, we performed bulk RNA sequencing (RNA-Seq) analysis in the liver of male Rbbp5 ^Flox and Rbbp5 ^LKO mice at 2-h intervals for a total of 48 h under constant darkness with biological duplicates (S2 Table). To identify genes cycling with a period ranging from 6 to 36 h, we initially applied the RAIN algorithm [55] to each genotype’s temporal transcriptome [55] (S3 Table). Compared to alternative methods like JTK_CYCLE, RAIN detects rhythms with arbitrary waveforms and therefore more robustly uncovers ultradian rhythms [17]. Consistent with past studies [16,17,30,33,35], two populations of hepatic transcripts cycling with periods of ~12 h and ~24 h were identified in Rbbp5 ^Flox mice liver (Fig 3C–3E). By contrast, in Rbbp5 ^LKO mice, the ~12 h rhythms are all but abolished and the circadian oscillations remain largely intact, which include all core circadian clock genes (such as Bmal1, Per1/2/3, and Nr1d1/2) displaying robust circadian rhythms with the same phases and amplitudes as those observed in Rbbp5 ^Flox mice (Fig 3C–3F). Specifically, using RAIN with an FDR cutoff of 0.01 and 0.05, we identified a total of 3,549 and 6,025 hepatic transcripts cycling at periods between 10 and 12 h in Rbbp5 ^Flox mice, respectively. By contrast, only 61 and 813 10–12 h genes were observed in Rbbp5 ^LKO mice with the same FDR thresholds, indicating over 93% ~12 h hepatic transcriptome was abolished with hepatocyte RBBP5 ablation (S6A and S6B Fig and S3 Table). GO analysis confirmed that these RBBP5-depenent ~12 h hepatic transcriptomes are enriched in mRNA processing and proteostasis pathways, distinct from those of RBBP5-independent circadian genes enriched in lipid metabolism and phosphorylation (S6C Fig).

To assess the robustness of our findings across analytical approaches, we applied multiple orthogonal methods, including MetaCycle [57], cosinor2 [43], and the eigenvalue/pencil method [53]. While MetaCycle and cosinor2 are less sensitive than RAIN for detecting ultradian rhythms with nonsinusoidal or asymmetric waveforms, both methods nonetheless consistently revealed a marked depletion of ~12 h gene expression rhythms in Rbbp5 ^LKO mice (S7A–S7H Fig). In contrast to these approaches, the eigenvalue/pencil method does not require pre-specification of period ranges, allowing unbiased decomposition of complex temporal signals into multiple superimposed oscillatory components [53]. Revealing up to three oscillations from each gene, eigenvalue/pencil analyses also revealed prevalent ~24 h and ~12 h oscillations in Rbbp5 ^Flox mice, along with a third population cycling at a period around 8 h (S4 Table) (S8A and S8B Fig). Hepatic ~8 h oscillations were also previously reported, but their regulation and function remained elusive [28,33]. Rbbp5 ^LKO mice, on the other hand, exhibited a drastically different spectrum, characterized by the near complete loss of the ~12 h rhythms, de novo gains of many faster ultradian rhythms with periods between 6 and 7 h, and the maintenance of most circadian rhythms (S8A–S8C Fig). Similar findings were also observed for dominant oscillations (which is the one with the largest amplitude among the three superimposed oscillations for each gene) for each gene (Fig 3G and 3H). For the ~12 h genes originally identified in Rbbp5 ^Flox mice but lost in Rbbp5 ^LKO mice, the period spectrum in the Rbbp5 ^LKO mice displayed a diverse distribution. This ranged from a complete lack of rhythmicity (~39%) to shorter ultradian periods of 6–7 h, and even to longer periods exceeding 24 h (S8D Fig). RBBP5-dependent ~12 h transcriptome uncovered by the eigenvalue/pencil showed substantial overlap with the results obtained via the RAIN method. This convergence was evident both in the specific genes identified (S8E Fig) and in the enriched biological pathways (Figs 3I and S6C), thereby confirming the robustness of our findings.

Lastly, since XBP1s is a major transcriptional regulator of the hepatic 12 h oscillator [14,17], we examined how RBBP5 regulates its temporal expression. In the liver of Rbbp5 ^LKO mice, the 12 h oscillations of Xbp1s mRNA and protein were impaired, accompanied by a reduction in mean daily XBP1s expression (Fig 3J and 3K). We next performed XBP1s ChIP-qPCR at CT12, when hepatic XBP1s protein levels were comparable between WT and Rbbp5 ^LKO mice. Hepatic XBP1s occupancy at the promoters of canonical 12 h UPR genes was comparable between the two genotypes (Fig 3L). These results suggest that RBBP5 regulates the temporal expression of XBP1s but is likely not required for XBP1s chromatin recruitment, consistent with its role as a transcriptional co-regulator. Together, these results thus establish RBBP5 as a central epigenetic regulator of the hepatic 12 h oscillator, with more than 90% ~12 h hepatic transcriptome abolished in its absence.

In addition to the liver, cell-autonomous ~12 h rhythms of proteostasis gene expression can also be observed in mouse embryonic fibroblasts (MEFs) synchronized by low concentration of ER stress inducer tunicamycin (Tu) [30]. We observed a ~12 h rhythm of Rbbp5 expression in Tu-synchronized MEFs, which is independent of BMAL1 but require XBP1s (S9A Fig), indicating RBBP5 itself is also under 12 h oscillator control. To test the functional role of COMPASS in regulating the cell-autonomous ~12 h rhythms, we knocked down either Rbbp5 or Ash2l via siRNA in a previously described 12 h oscillator reporter MEFs that stably express Manf (Manf is an XBP1s-dependent 12 h oscillator output gene) promoter-driven destabilized luciferase (dluc) [17], and found that both siRNAs significantly dampen the ~12 h oscillation of luciferase activity as well as shorten the period (S9B Fig). By contrast, neither RBBP5 nor ASH2L knocking down affects the circadian oscillation of Bmal1 promoter activity in human U2OS cells in a previous genome-wide siRNA screen study [58] (S9C Fig). Collectively, our results demonstrate that RBBP5 is an integral component of the 12 h oscillator, but dispensable for the canonical circadian clock both in vivo and in vitro.

RBBP5 modulates the hepatic transcriptional response to acute proteotoxic stress in vivo

In contrast to the self-sustained oscillatory dynamics of the UPR, its regulation has been more extensively characterized as a transient response to proteotoxic stress, most notably during ER stress [12]. Upon accumulation of misfolded proteins in the ER, the UPR is activated through signaling pathways that relay information from the ER to the nucleus, culminating in the activation of three principal transcription factors: XBP1s, ATF4, and ATF6. These factors coordinate an adaptive transcriptional program that upregulates proteostasis-related genes to restore ER homeostasis (Fig 4A) [59–61]. However, unsolved chronic ER stress can also trigger maladaptive and deleterious responses, such as the activation of pro-inflammatory and immune gene expression through the TRAF2-mediated activation of AP1 (comprised of c-Fos and c-Jun), NF-κB, or STAT3 TFs (Fig 4A) [62]. Our findings that RBBP5 can co-activate XBP1s to regulate the 12 h oscillator suggest that RBBP5 is essential for the adaptive UPR TFs-mediated activation of proteostasis genes but is not required for the maladaptive activation of immune genes by immune-regulatory TFs, whose binding motifs are absent from RBBP5 genome binding sites (Fig 2F).

To test this hypothesis, we intraperitoneally injected male Rbbp5 ^Flox and Rbbp5 ^LKO mice with vehicle control or 0.05 mg/kg Tu to induce acute ER stress and performed RNA-seq on hepatic transcriptome isolated 8 h post-injection (S5 Table). This lower dose of Tu was selected for its ability to induce the UPR, but it was significantly less toxic than the more common experimental dose of 1 mg/kg—which is known to elicit hepatocellular death and severe toxicity [63]. Western blot analysis confirmed the depletion of RBBP5 and a reduction of H3K4me3 level in the liver of Rbbp5 ^LKO mice (Fig 4B). Principle component analysis (PCA) demonstrated the separation of the four groups on the transcriptome space (Fig 4C). Using an adjusted p value (FDR) of 0.10 as cut-off, we identified 139 and 45 genes significantly induced by Tu in the liver of Rbbp5 ^Flox and Rbbp5 ^LKO mice, respectively, with 21 commonly shared between the two (Figs 4D and S10A–S10G and S6 Table). In Rbbp5 ^Flox mice, these 139 genes are enriched in pathways of proteostasis and immune functions as expected, with pathways of ‘response to ER stress’ and ‘ERAD’ among the top two GO terms (Fig 4D and 4E). By contrast, in Rbbp5 ^LKO mice, these 45 genes are only enriched in immune functions, with proteostasis genes not being significantly induced by Tu (Fig 4D and 4E). In general, proteostasis genes were induced by Tu to a much greater fold in Rbbp5 ^Flox mice (such as Creld2, Xbp1, Hsp90b1), while the opposite was found for blood coagulation and immune genes in Rbbp5 ^LKO mice (such as S100a9 and Fgb) (Figs 4D, 4E, S10E, S10F and S10H). Further integrating RBBP5 ChIP-seq results, we found that, on average, the ablation of RBBP5 markedly diminished the fold induction by Tu for genes with RBBP5 promoter binding, while it had no effects for those without RBBP5 chromatin recruitment (Fig 4D and 4F). To rule out the possibility that RBBP5 regulates UPR via influencing the upstream ER stressing in the ER, we performed western blot analysis of p-IRE1α and p-PERK in the liver of Rbbp5 ^Flox and Rbbp5 ^LKO mice and did not observe any notable difference in the increase in their phosphorylated level in response to Tu between the two genotypes (Fig 4G).

To determine if RBBP5 overexpression can also amplify UPR, we tail-vein injected AAV8-TBG-GFP and AAV8-TBG-RBBP5 viruses into wild-type male mice followed by intraperitoneal injection with vehicle control or 0.025 mg/kg Tu 30 days later. Thyroxine binding globulin (TBG) promoter is liver-specific and directs RBBP5 transgene expression in hepatocytes [64]. qPCR and western blot analysis confirmed an increase of total hepatic expression of RBBP5 of ~1.3 fold in AAV8-TBG-RBBP5 group under vehicle control condition (Fig 4H and 4I). Consequently, RBBP5 hepatic overexpression significantly amplified the UPR of proteostasis genes including Manf, Hspa5 (BiP), Pdia4, Hyou1, and Syvn1 without affecting the upstream ER stress sensing (Fig 4H and 4I). In addition, we found that RBBP5 overexpression attenuated the expression of several immune and blood coagulation genes such as C3, Tifa and Fgl1 under ER stress conditions (Fig 4I).

Finally, leveraging the human gene-derived correlations across tissues (GD-CAT) dataset [65], we identified that in human liver, RBBP5 expression is also positively correlated with those involved in proteostasis (such as ribosome biogenesis, translation regulation) and mRNA metabolism (mRNA splicing and processing) and negatively associated with those implicated in immune response (acute phase response, complement activity, and immunoglobin activity) (S11 Fig). Taken together, these results demonstrate that RBBP5 modulates the transcriptional regulation of hepatic response to acute proteotoxic stress in vivo. RBBP5 activates the expression of proteostasis genes while repressing those involved in immune functions in both mice and humans.

Loss of RBBP5 compromises the cell-autonomous transcriptional response to proteotoxic stress and nutrient-deprivation, resulting in impaired autophagy and reduced cell survival

To determine whether RBBP5 regulates transcriptional responses to proteotoxic stress in a cell-autonomous manner beyond the liver, we examined MEFs, which exhibit both a robust cell-autonomous XBP1s-dependent 12 h oscillator and strong transcriptional responses to proteotoxic stress [17,30]. Rbbp5 mRNA expression was induced by Tu and transient knockdown of RBBP5 using siRNA significantly dampened the proteostasis gene expressions in response to Tu, mirroring the effects observed with XBP1s knockdown (S12A and S12B Fig). Similarly, knocking down ASH2L also markedly dampened the UPR, whereas knocking down SETD1A, one of the H3K4me3 methyltransferases in the SET1/COMPASS complex, had only modest effects, which were likely due to redundancy among the many H3K4me3 methyltransferases that exist (S12C Fig). As a negative control, we further knocked down both GCN5 and PCAF, proteins that function to write H3K9Ac as subunits in two distinct histone acetyltransferase (HAT) complexes, SAGA (Spt-Ada-Gcn5 acetyltransferase) and ATAC (Ada2a-containing) [66] and found it had no effects on the induction of Xbp1 and Manf expression in response to ER stress (S12D Fig). The lack of effects on UPR for HAT subunits is consistent with the absence of ~12 h rhythms of both H3K9Ac and H3K27Ac levels around the promoter of ~12 h proteostasis genes in mouse liver (Fig 1A and 1B). ChIP-qPCR analysis confirmed a gradual, temporal increase in the recruitment of RBBP5 to the proximal promoters of select proteostasis genes in response to ER stress, concomitant with an increase in promoter-proximal H3K4me3 levels (Fig 5A). Since XBP1s is the major TF driving UPR, we further performed IF on XBP1s and RBBP5 in MEFs and found a strong recruitment of RBBP5 to XBP1s in response to ER stress (Fig 5B and 5C).

To broadly assess the function of RBBP5 in the transcription regulation of proteotoxic stress response, we utilized orthogonal approaches to both deplete RBBP5 expression and trigger cellular stress. We took advantage of our homozygous Rbbp5 ^Flox mice by crossing them with homozygous ROSA26 Cre-ERT2 mice, then isolated and immortalized primary MEFs from homozygous F2 embryos (S12E Fig). Treating these Rbbp5 (flox/flox); ROSA26 Cre-ERT2 (+/+) MEFs with tamoxifen for 7 days resulted in the complete deletion of both floxed Rbbp5 alleles (S12F Fig) and a subsequent ~90% reduction in RBBP5 protein and a reduction of H3K4me3 levels (S12G Fig). The remaining RBBP5 protein is likely extremely long-lived and remains stable even after the deletion of underlying coding genes. To trigger ER stress, we used a different stimulus: Dithiothreitol (DTT), which disrupts protein disulfide formation in the ER [67,68]. Similar to Tu, RBBP5 depletion significantly dampened the induction of proteostasis gene expression in response to DTT (S12H Fig).

Next, we tested two additional proteotoxic stressors that are more physiologically relevant: leucine deprivation to simulate amino acid restriction, and a more severe nutrient deprivation stress by culturing cells in Earle's Balanced Salt Solution (EBSS) for 16 h. Both treatments not only trigger UPR but also activate the broader integrated stress response (ISR), resulting in the induction of autophagy [69]. RBBP5 depletion impaired the transcriptional response to both leucine deprivation and EBSS, leading to a reduced induction of autophagy-promoting genes (S13A Fig). Transcriptome profiling by RNA-seq confirmed the globally defective adaptive stress response to nutrient deprivation (EBSS) in the absence of RBBP5, where only 327 out of 2,351 genes can still be induced in RBBP5 KO MEFs (Fig 5D, 5E and S13B). GO analysis revealed that RBBP5-dependent nutrient deprivation-induced genes are strongly enriched in translation, autophagy, protein ubiquitination, and mRNA processing pathways (Fig 5F).

To investigate the physiological consequence of RBBP5 deletion, we quantified the autophagic flux by utilizing a tandem LC3 reporter mCherry-GFP-LC3 where an increase in the number of red-fluorescent cytosolic puncta indicates increased autolysosome formation and autophagic flux (S13C Fig) [69]. Consistent with the qPCR results, RBBP5 depleted MEFs failed to induce a robust general autophagy in response to leucine deprivation and further exhibited lower basal level of autophagic flux (S13D and S13E Fig). Western blot analysis of p62/SQSTM1, a marker of autophagic flux due to its degradation during autophagy [70], showed a significant reduction in wild-type, but not RBBP5 depleted cells following leucine deprivation (S13F Fig), thus confirming the defective autophagy phenotype associated with RBBP5 depletion. Since prolonged nutrient deprivation, such as 16 h of EBSS treatment, can robustly induce both general autophagy and autophagy of the ER (ER-phagy) [71], we further measured ER-phagy by utilizing a previously published ER-phagy reporter system where a stably expressed ER-targeting mCherry-RAMP4 fusion is cleaved to free mCherry during starvation-induced ER-phagy [72] (Fig 5G). As shown in Fig 5H and 5I, RBBP5 depletion prevented the cleavage of mCherry in response to EBSS, indicating an impaired ER-phagy. Western blot analysis of p62/SQSTM1 further showed undetectable levels in wild-type cells following EBSS treatment, whereas a significant level remained in RBBP5 depleted cells under the same conditions, thus confirming a reduction of the general autophagy in RBBP5 depleted cells in response to EBSS (Fig 5H). Consistent with weakened stress response and impaired autophagy, RBBP5 depleted MEFs displayed reduced cell survival under nutritional stress (S13G Fig). Taken together, our results demonstrate that RBBP5 is indispensable for regulating the cell-autonomous transcriptional responses to various proteotoxic stresses as a core subunit of the H3K4me3-writing COMPASS complex.

Proximity labeling reveals H3K4me-associated proteomic architecture of the proteotoxic stress response

H3K4me3 is widely recognized as an active promoter mark, but its precise role in transcriptional regulation remains a subject of debate. Initially, it was believed that H3K4me3 facilitated transcription initiation by recruiting the RNAPII preinitiation complex [66], However, recent evidence suggests that H3K4me3 primarily regulates RNAPII pause-release and contributes to transcriptional elongation, likely through the recruitment of Integrator Complex subunit 11 (INTS11) [47,48]. This discrepancy may stem from the context-specific functions of H3K4me3 in recruiting different transcriptional regulators across different biological processes [73]. To shed light on the potential mechanisms by which H3K4me3 modulates proteostasis stress response, we performed quantitative TurboID-based proximity labeling in MEFs (in biological triplicates per condition), utilizing anti-H3K4me3 antibody (with anti-IgG as negative control) followed by recombinant protein A-TurboID labeling in isolated nuclei of MEFs before (T = 0 h) and 2.5 h and 8 h after ER stress induction by Tu (Figs 6A and S14A) [74]. While the total level of H3K4me3 remains largely unchanged during ER stress (Fig 6B), we identified a total of 239 proteins (p < 0.1) whose recruitment to H3K4me3 peaked at 8 h post Tu treatment (S9 Table). Pathway analysis of these 239 proteins revealed expected enrichment of COMPASS subunits (Figs 6C, S14B and S14C), RNAPII subunits (S14D Fig), as well as previously identified Integrator Complex subunits (such as INTS11) (S14E Fig). The previously identified H3K4me3 demethylase KDM5A [48] revealed a much more constant recruitment to H3K4me3 during ER stress, with only a slight increase at 8 h after ER stress induction (S14C Fig).

Besides previously identified H3K4me3 writer/eraser/readers, we also revealed strong enrichment of SWI/SNF chromatin remodelers on H3K4me3 peaking at 8 h after ER stress induction (Fig 6C and 6D). These include BAF-specific (ARID1A/B), PBAF-specific (ARID2, PBRM1 and BRD7) and shared subunits (SMARCD1/2/3 and SMARCC2) (Fig 6C and 6D) [75]. The timing of SWI/SNF recruitment to H3K4me3 during ER stress closely parallels that of COMPASS and Integrator subunits but lags behind RNAPII subunits (Fig 6E). This temporal pattern suggests that SWI/SNF might potentially play a role in facilitating transcriptional elongation during stress-induced gene activation. Together, these findings demonstrate that H3K4me3 functions as a dynamic signaling hub that coordinates multiple chromatin regulators to ensure effective transcriptional adaptation to proteotoxic stress.

Ethics statement

The animal studies were carried out in accordance with the National Institutes of Health guidelines and were granted formal approval by the University of Pittsburgh’s Institutional Animal Care and Use Committee (approval number IS00013119 and IS00013119).

The generation of Rbbp5 ^(flox/flox) mice

The Rbbp5-flox allele was generated by the Mouse Embryo Services Core (University of Pittsburgh, Department of Immunology). The targeting strategy is based on the Easi-CRISPR method [99]. In brief, fertilized embryos from C57BL/6J background, produced by natural mating, were microinjected in the pronuclei with a mixture of 0.67 µM EnGen Cas9 protein (New England Biolabs, Cat.No. M0646T), two Cas9 guides RNA: Rbbp5-695rev and Rbbp5-951forw (21.23 ng/µl each) and a long single stranded oligonucleotides Rbbp5-flox-ssODN (5 ng/µl). The long single stranded oligonucleotide encoding the donor sequence was synthesized by Integrated DNA Technologies (IDT).

To produce the sgRNA, a double strand linear DNA template is created by annealing of the following target specific oligonucleotides with a common primer (gRNA-Scaffold-R: 5′-AAAAAAGCACCGACTCGGTGCCACTTTTTCAAGTTGATAACGGACTAGCCTTATTTTAACTTGCTATTTCTAGCTCTAAAAC-3′) containing the full tracrRNA sequence and then PCR-amplified as previously described [100]. The sgRNA was synthesized using HiScribe T7 Quick High Yield RNA Synthesis Kit (New England Biolabs, Cat. No. E2050S).

T7-Rbbp5-695rev sequence:

TAATACGACTCACTATAGGAAGCATTCTAAGATTAAACCGTTTTAGAGCTAGAAATAGCA

T7-Rbbp5-951forw sequence:

TAATACGACTCACTATAGGAGTAACAGATAGTATTCCGAGTTTTAGAGCTAGAAATAGCA

Rbbp5-flox-ssODN sequence:

gcagttaataaacttgaagtgtttatatagaaaatacatcaggatggcaaccaaaccctttagatacctgagattttattctcttatcccaggtATAACTTCGTATAGCATACATTATACGAAGTTATgaattcttaatcttagaatgcttagtctgtttcatgctgaagataagttgtcagaactcacattgttactgatttttcgttcagtgactcgtctgttttgtcataggaaaattggagtgcatttgcaccagacttcaaagagttggatgaaaatgtagaatatgaggaaagagaatcagaatttgatattgaggatgaagataagagtgagcctgagcaaacaggtgatgcctctcagtaacagatagtattgaattcATAACTTCGTATAGCATACATTATACGAAGTTATccgaagggaaggactgtcttactcaatgctctttgaattagtgaatgcatttatttactttgtgttgtaactgtgctcaaagcttcatctgccacgtttccctcacttgtacatagttttcacagtgcttgaagagtcttactg

The injected zygotes were cultured overnight, the next day the embryos that developed to the 2-cell stage were transferred to the oviducts of pseudo pregnant CD1 female surrogates. Following optimization of the cycling conditions, the potential founder mice were genotype with the following primers pair, using Q5 High-Fidelity DNA Polymerase (NEW ENGLAND BIOLABS, Cat. no. M0491L). The body weight of male adult mice was measured weekly.

Rbbp5 genotyping forward primer: AGTTCAGAGCTGGTTTCCAAC

Rbbp5 genotyping reverse primer: AGGTAATGCCCATGTGAGCC

Biological rhythm study in mice

All mice used for biological rhythm study are in C57BL/6J background, male and between 3 and 4 months of age. Wild-type C57BL/6J mice (n = 12), Rbbp5 (flox/flox); Alb-CRE (−/−) (n = 48) and Rbbp5 (flox/flox); Alb-CRE (+/−) mice (n = 48) [Rbbp5 (flox/flox); Alb-CRE (−/−) and Rbbp5 (flox/flox); Alb-CRE (+/−) mice were littermates] were first entrained under LD12:12 conditions for 2 weeks before transferred to constant darkness for 24 h. Mice were then sacrificed via cervical dislocation at a 2 h interval [for Rbbp5 (flox/flox); Alb-CRE (−/−) and Rbbp5 (flox/flox); Alb-CRE (+/−) mice] or 4 h interval (for wild-type C57BL/6J mice) for a total of 48 h under constant darkness. Mice were fed ad libitum during the entire experiment.

Food intake and locomotor activity monitoring

Promethion Multi-plexed Metabolic Cage System was used for real-time measuring of food intake and locomotor activity. Male Rbbp5 ^Flox and Rbbp5 ^LKO mice between 3 and 4 months of age were acclimated to the chambers and ad libitum food intake was monitored for 48 h under LD12:12 followed by 48 h of constant darkness. n = 6 for Rbbp5 ^Flox and n = 8 for Rbbp5 ^LKO mice for LD12:12 study. n = 5 for Rbbp5 ^Flox and n = 3 for Rbbp5 ^LKO mice for DD study.

ER stress induction study in mice

For acute ER stress induction experiment, male littermates Rbbp5 (flox/flox); Alb-CRE (−/−) and Rbbp5 (flox/flox); Alb-CRE (+/−) mice between 5 and 7 months of age were randomly divided into two groups, and intraperitoneally injected with 0.05 mg/kg body weight of tunicamycin dissolved in 500 µl 3% DMSO in PBS or vehicle control (3% DMSO in PBS), respectively. Mice were injected with Tu between 9 and ~10 AM in the morning and 8 h later dissected for transcriptomic and western blot analysis for different tissues. The sample size is n = 3 for Rbbp5 ^Flox DMSO, n = 3 for Rbbp5 ^Flox Tu, n = 4 for Rbbp5 ^LKO DMSO, n = 4 for Rbbp5 ^LKO Tu.

For experiments involving RBBP5 liver-overexpressing mice, AAV8-TBG-EGFP and AAV8-TGB-mRBBP5 viruses were customarily designed and produced from Vector Biolab (Malvern, PA). Male wildtype C57BL/6J mice between 2 and 3 months of age were tail-vein injected with 0.5 x 10^11 genome copies of AAV8-TBG-EGFP or AAV8-TGB-mRBBP5 virus diluted in 100 µl of PBS. Thirty days later, mice injected with either viruses were randomly divided into two cohorts and intraperitoneally injected with 0.05 mg/kg body weight of tunicamycin dissolved in 500 µl 3% DMSO in PBS or vehicle control (3% DMSO in PBS). Mice were injected with Tu between 9 and ~10 AM in the morning and 8 h later dissected for transcriptomic and western blot analysis for different tissues. The sample size is n = 3 for AAV8-TBG-EGFP DMSO, n = 4 for AAV8-TBG-EGFP Tu, n = 4 for AAV8-TGB-mRBBP5 DMSO, and n = 4 for AAV8-TGB-mRBBP5 Tu.

Generation of stable autophagy reporter and ER-phagy reporter cell line

Primary MEFs were isolated from Rbbp5 (flox/flox); Rosa26 CRE (+/+) mice and immortalized with SV40 lentivirus as previously described [101]. For mCherry-EGFP-LC3B-expressing autophagy reporter MEFs, lentiviruses packaged in HEK293T cells with co-transfection of pCDH-EF1a-mCherry-EGFP-LC3B, pMD2.G and psPAX2 plasmids were used to infect Rbbp5 (flox/flox); Rosa26 CRE (+/+) MEFs with a multiplicity of infection (MOI) of 3 for three times to achieve near complete infection. For mCherry-RAMP4-expressing ER-phagy reporter MEFs, lentiviruses packaged in HEK293T cells with co-transfection of pLenti-X1-hygro-mCherry-RAMP4, pMD2.G and psPAX2 plasmids were used to infect Rbbp5 (flox/flox); Rosa26 CRE (+/+) MEFs with a multiplicity of infection (MOI) of 3. Stable MEFs were selected in the presence of 200 μg/ml hygromycin.

siRNA Transient Transfections

Immortalized MEFs isolated from wild-type C57BL/6J mice were transfected with 10µM of different siRNAs for 24~48 h with Lipofectamine RNAiMAX reagents (Life technologies) per the manufacturer’s instructions. Source of siRNA are as follows: siGENOME nontargeting siRNA pool (Dharmacon, D-001206-1305), siGENOME SMARTpool mouse Rbbp5 siRNA (Dharmacon, L-054560-01-0005), siGENOME SMARTpool mouse Ash2l siRNA (Dharmacon, L-048754-01-0005), siGENOME SMARTpool mouse Setd1a siRNA (Dharmacon, L-051358-01-0005), siGENOME SMARTpool mouse Kat2a/Gcn5 siRNA (Dharmacon, L-040665-01-0005), siGENOME SMARTpool mouse Kat2b/Pcaf siRNA (Dharmacon, L-049885-01-0005) and siGENOME SMARTpool mouse Xbp1 siRNA (Dharmacon, L-040825-00-0005). MEFs were transfected with 10 µM of different siRNAs for 48 h and treated with DMSO or 100 ng/ml of tunicamycin for 8 h.

Real-time luminescence assay

Stable Manf-dluc MEFs were transfected with nontargeting, Rbbp5, or Ash2l siRNA for 48 h in DMEM (4.5g/L glucose) supplemented with 10% FBS and treated with 25 ng/ml tunicamycin in DMEM for 2 h before subjected to real-time luminescence assay using a Lumicycle (Actimetrics). Briefly, after tunicamycin treatment, MEFs were washed with 1x PBS and cultured with DMEM (4.5 g/L glucose) supplemented with 0.1 mM luciferin and 10 mM HEPES buffer in 35 mm tissue culture dishes in the absence of serum and transferred immediately to Lumicycle for real-time luminescence analysis. Periods of oscillation were identified by embedded Periodogram function.

Immunofluorescence (IF)

Immunofluorescence was performed as previously described [25]. Briefly, liver OCT sections were fixed in cold acetone for 10 mins at −20 ºC. The sections were then air dried, rehydrated with PBS and permeabilized with PBS+ 0.1% Triton X-100. The sections were then blocked with 10% goat serum at room temperature for 1 h. Primary antibody against RBBP5 (Bethyl, A300-109A) was added to the OCT section at 1:1000 dilution overnight at 4 ºC. Next day, sections were washed three times with PBS and stained with Alexa-555 anti-rabbit secondary antibody at room temperature for 2 h. After that, the sections were washed with PBS three times and counterstained with DAPI before mounting (with ProlongGold Glass) and imaging using Leica SP8 lightening confocal microscope (Leica Microsystems).

Time-lapse microscopy for autophagy reporter MEFs

Time-lapse imaging was performed using SP8 lightening confocal microscope (Leica Microsystems) with Okolab stage top incubator to maintain constant CO₂ (5%), temperature (37 ºC) and humidity (90%). mCherry-EGFP-LC3B-expressing Rbbp5 (flox/flox); Rosa26 CRE (+/+) MEFs were plated into 8-well chamber slides in full DMEM media and treated with (to deplete RBBP5) or without tamoxifen (3 µM/ml) for seven days before replacing with full DMEM or leucine-free media counterstained with DAPI and images were taken every 2 h for a total of 16 h. The number of cytosolic autophagosomes (GFP and mCherry double positive foci) and autolysosomes (GFP negative and mCherry positive foci) per cell were quantified manually.

Immunoblot

Nuclear extracts were made from liver according to a previously published protocol [102]. Whole cell lysates were lysed in RIPA buffer as previously described [25]. Protein concentrations were determined by Bradford assays (Bio-Rad), and aliquots were snap-frozen in liquid nitrogen and stored at −80 °C until usage. Immunoblot analyses were performed as described previously [80]. Briefly, 25 ~50 µg proteins separated by 4~20% gradient SDS-PAGE gels (Biorad) were transferred to nitrocellulose membranes, blocked in TBST buffer supplemented with 5% bovine serum albumin (BSA) or 5% fat-free milk and incubated overnight with primary antibody anti-PERK (Cell signaling, #3192), anti-phospho-PERK (Thr908) (Thermo Fisher, MA5–15033), anti-ATF4 (Cell signaling, #11815), anti-IRE1α (Cell signaling, #3294), anti-phospho-IRE1α (Ser724) (ABclonal, AP0878), anti-XBP1s (Biolegend, 658802), anti-RBBP5 (Bethyl, A300-109A), anti-RBBP5 (D3I6P) (Cell signaling, #13171), anti-WDR5 (Cell signaling, #13,105), anti-ASH2L (Bethyl, A300-489A), anti-ASH2L (Bethyl, A300-112A), anti-SETD1A (Diagenode, CS-117-100), anti-SETD1B (Diagenode, CS-118-100), anti-MLL1-c (D6G8N) (Cell signaling, #14197), anti-mCherry (E5D8F) (Cell signaling, #43590), anti-p62/SQSTM1 (Santa Cruz, sc-28359) and anti-β-ACTIN (Cell signaling, #12620) at 4 °C. Blots were incubated with an appropriate secondary antibody coupled to horseradish peroxidase at room temperature for 1 h, and reacted with ECL reagents per the manufacturer’s (Thermo) suggestion and detected by Biorad ChemiDoc MP Imaging System.

Co-immunoprecipitation (Co-IP)

Nuclear extracts (NE) were made from liver according to our published protocol [102]. Protein concentrations were determined by Bradford assays (Bio-Rad), and aliquots were snap-frozen in liquid nitrogen and stored at −80 °C until usage. 200 μg of NE was used for per IP. 5 μg of different antibodies or control IgG were coupled to 1.5 mg of magnetic Dynabeads (Life Technologies) for every IP using Dynabeads Antibody Coupling Kit (Life Technologies) per manufactures’ protocol. Co-IP was essentially carried out as previously described [80], except that coupled antibody Dynabeads were added to the NEs for incubation. The antibodies used for Co-IP are as follows: anti-XBP1s (Biolegend, 658802).

qRT-PCR

Total mRNA was isolated from murine embryonic fibroblasts (MEFs) or mouse liver with PureLink RNA mini kit (Life Technologies) with additional on-column DNase digestion step to remove genomic DNA per the manufacturer’s instructions. Reverse transcription was carried out using 5 µg of RNA using Superscript III (Life Technologies) per the manufacturer’s instructions. For gene expression analyses, cDNA samples were diluted 1/30-fold (for all other genes except for 18sRNA) and 1/900-fold (for 18sRNA). qPCR was performed using the SYBR green system with sequence-specific primers. All data were analyzed with 18S or β-actin as the endogenous control. qPCR primer sequences are as follows and all primers span introns, except for primers for quantifying pre-mRNAs:

Mouse Rbbp5 forward primer: CAGAGCCTATGCAGAAGCTGCA

Mouse Rbbp5 reverse primer: CTTCACCAGGTTGCCAATGCTC

Mouse Ddit3 forward primer: GGAGGTCCTGTCCTCAGATGAA

Mouse Ddit3 reverse primer: GCTCCTCTGTCAGCCAAGCTAG

Mouse Creld2 forward primer: CTACACCAAGGAGAGTGGACAG

Mouse Creld2 reverse primer: TCTGTCTCCTCAAAGCCTTCCG

Mouse Hspa5 forward primer: TGTCTTCTCAGCATCAAGCAAGG

Mouse Hspa5 reverse primer: CCAACACTTCCTGGACAGGCTT

Mouse Dnajb11 forward primer: TGTGACCGTCTCACTGGTTGAG

Mouse Dnajb11 reverse primer: CCCTTTCTTCCACAGCTTGGCT

Mouse Pdia4 forward primer: TGGGCTCTTTCAGGGAGATGGT

Mouse Pdia4 reverse primer: GGGAGACTTTCAGGAACTTGGC

Mouse Hsp90b1 forward primer: GTTTCCCGTGAGACTCTTCAGC

Mouse Hsp90b1 reverse primer: ATTCGTGCCGAACTCCTTCCAG

Mouse Syvn1 forward primer: CCAACATCTCCTGGCTCTTCCA

Mouse Syvn1 reverse primer: CAGGATGCTGTGATAAGCGTGG

Mouse Gdf15 forward primer: ACTCAGTCCAGAGGTGAGATTG

Mouse Gdf15 reverse primer: GGGGCCTAGTGATGTCCCAG

Mouse Il1a forward primer: ACGGCTGAGTTTCAGTGAGACC

Mouse Il1a reverse primer: CACTCTGGTAGGTGTAAGGTGC

Mouse Il1b forward primer: TGGACCTTCCAGGATGAGGACA

Mouse Il1b reverse primer: GTTCATCTCGGAGCCTGTAGTG

Mouse Fga forward primer: GGATTCTAACTCACTGACCAGGA

Mouse Fga reverse primer: CCTCAGGATCTCAATTCTGCGC

Mouse C3 forward primer: CGCAACGAACAGGTGGAGATCA

Mouse C3 reverse primer: CTGGAAGTAGCGATTCTTGGCG

Mouse Tifa forward primer: AGCAAGAAAACCAGTTTGATGGTAG

Mouse Tifa reverse primer: GCAACAGGAACTGATACTCTCCG

Mouse Fgl1 forward primer: GGAAACTGTGCTGAGGAAGAGC

Mouse Fgl1 reverse primer: TCCGTTTCTGCCCTGTAGGAAC

Mouse Fgb forward primer: TGACACCTCCATCAAGCCGTAC

Mouse Fgb reverse primer: GGTCCCATTTCCTGCCAAAGTC

Mouse Il6 forward primer: CTTCCATCCAGTTGCCTTCT

Mouse Il6 reverse primer: CTCCGACTTGTGAAGTGGTATAG

Mouse Herpud1 forward primer: acctgagccgagtctaccc

Mouse Herpud1 reverse primer: aacagcagcttcccagaataaa

Mouse Nbr1 forward primer: GGATTTAAAGCACCTCCTGATTCC

Mouse Nbr1 reverse primer: ATTGGGTCCCACTCAGGTCT

Mouse Sesn2 forward primer: CTGGCCGAACTCATCCAAG

Mouse Sesn2 reverse primer: CTGCCTCATGCGTTCCATC

Mouse Sqstm1 forward primer: AACAGATGGAGTCGGGAAAC

Mouse Sqstm1 reverse primer: AGACTGGAGTTCACCTGTAGA

Mouse Stbd1 forward primer: CTGGGAAGTAGATGGGAAAGTG

Mouse Stbd1 reverse primer: TCGGTGTTTGTGGTGTAGTG

Mouse Sirt1 forward primer: TGACAGAACGTCACACGCCA

Mouse Sirt1 reverse primer: ATTGTTCGAGGATCGGTGCCA

Mouse Gcn5 forward primer: TTGATTGAGCGCAAACAGGC

Mouse Gcn5 reverse primer: CAGCCTGTCTCTCGAATGCC

Mouse Pcaf forward primer: CGTGAAGAAGGCGCAGTTG

Mouse Pcaf reverse primer: CAGGACTCCTCTGCCTTGC

Mouse Ash2l forward primer: gctgtgtctgctagtgggaac

Mouse Ash2l reverse primer: catcttgctgcttacgcttg

Mouse Setd1a forward primer: gggtagcaccccctactctc

Mouse Setd1a reverse primer: gggtttgaaggaggttgaagt

Mouse Eif2ak3 forward primer: ccttggtttcatctagcctca

Mouse Eif3ak3 reverse primer: atccagggaggggatgat

Mouse total Xbp1 forward primer: gggtctgctgagtcc

Mouse total Xbp1 reverse primer: cagactcagaatctgaagagg

Mouse Sec23b forward primer: tgaccaaactggacttctgga

Mouse Sec23b reverse primer: aaagaatctcccatcaccatgt

Mouse Manf forward primer: gacagccagatctgtgaactaaaa

Mouse Manf reverse primer: tttcacccggagcttcttc

Mouse Hyou1 forward primer: GAGGCGAAACCCATTTTAGA

Mouse Hyou1 reverse primer: GCTCTTCCTGTTCAGGTCCA

Mouse Atf4 forward primer: CCACTCCAGAGCATTCCTTTAG

Mouse Atf4 reverse primer: CTCCTTTACACATGGAGGGATTAG

Mouse Atf6 forward primer: CATGAAGTGGAAAGGACCAAATC

Mouse Atf6 reverse primer: CAGCCATCAGCTGAGAATTAGA

Mouse 18s RNA forward primer: ctcaacacgggaaacctcac

Mouse 18s RNA reverse primer: cgctccaccaactaagaacg

Mouse β-actin forward primer: aaggccaaccgtgaaaagat

Mouse β-actin reverse primer: gtggtacgaccagaggcatac

RNA-Seq to identify RBBP5-dependent 12 h hepatic transcriptome

Mouse liver tissues were collected from Rbbp5 ^Flox (n = 2) and Rbbp5 ^LKO (n = 2) mice at 2 h intervals for a total of 48 h under constant darkness conditions. Total RNA was isolated from mouse liver with miRNeasy Mini Kit (Qiagen) per the manufacturer’s instructions. RNA was assessed for quality using an Agilent TapeStation 4150/Fragment Analyzer 5300 and RNA concentration was quantified on a Qubit FLEX fluorometer. Libraries were generated with the Illumina Stranded mRNA Library Prep kit (Illumina: 20040534) according to the manufacturer’s instructions. Briefly, 200 ng of input RNA was used for each sample. Following adapter ligation, 12 cycles of indexing PCR were completed, using IDT for Illumina RNA UD Indexes (Illumina: 20040553-6). Library quantification and assessment were done using a Qubit FLEX fluorometer and an Agilent TapeStation 4150/Fragment Analyzer 5300. The prepared libraries were pooled and sequenced using NoveSeq 6000 (Illumina), generating an average of 40 million 2 × 100 bp paired end reads per sample. RNA-Seq library preparation and sequencing were performed at UPMC Genome Center. Raw RNA-seq FASTQ files were analyzed by FastQC for quality control. Adaptors and low-quality reads were filtered by Trimmomatic [103]. Then the processed reads were aligned by HISAT2 [104] against mouse reference mm10. Gene and isoform FKPM values were calculated by Cufflinks. Since the background gene expression level is FPKM = 0.1 in mouse liver [17], only those genes with average gene expression across 48 hours in Rbbp5 ^Flox mice larger than 0.1 were included for rhythm-identification analysis. Averaged FPKM values at each time were used for cycling transcripts identification by the eigenvalue/pencil method. Raw temporal data were subject to polynomial detrend (n = 2) first, and superimposed oscillations were identified using the previously described eigenvalue/pencil method on the detrended dataset [17]. Specifically, three oscillations were identified from each gene. Criterion for circadian gene is period between 21 h and 27 h, decay rate between 0.9 and 1.1; for ~12 h genes: period between 10 h and 13 h, decay rate between 0.9 and 1.1. The relative amplitude is calculated via dividing the amplitude by the mean expression value of each gene. All analysis were performed in MatlabR2017A. Heat maps were generated by Gene Cluster 3.0 and TreeView 3.0 alpha 3.0 using log2 mean-normalized values. RAIN analysis was performed as previously described in Bioconductor (3.4) (http://www.bioconductor.org/packages/release/bioc/html/rain.html) [55] using the full detrended dataset with duplicated at each time point.

For the eigenvalue method, every gene consists of multiple superimposed oscillations. Therefore, we define a circadian gene as any gene that exhibits a superimposed circadian rhythm, regardless of its relative amplitude compared to other superimposed oscillations. Similar definitions apply to 12 h genes. Under this definition, a gene can be both a circadian and 12 h-cycling gene. By comparison, we define a dominant circadian gene as any gene whose superimposed circadian rhythm has the largest amplitude among all oscillations. Similar definitions also apply to 12 h genes. Under this definition, dominant circadian and dominant 12 h genes are mutually exclusive.

RNA-seq for identifying RBBP5-dependent acute transcriptional response to ER stress

Mouse liver tissues were collected from Rbbp5 ^Flox and Rbbp5 ^LKO mice intraperitoneally treated with or without Tu for 8 h. Total mRNA was isolated from mouse liver with PureLink RNA mini kit (Life Technologies) with additional on-column DNase digestion step to remove genomic DNA per the manufacturer’s instructions. Detailed procedures for transcriptome sequencing were described previously [105] and were supported by the Genomics and Systems Biology Core of the Pittsburgh Liver Research Center. Briefly, the mRNA samples per mouse were processed into short-read libraries using the Bio-Rad SEQuoia Dual Indexed Primers. Next, the circularization was performed using the Element Biosciences Adept Library Compatibility Kit v1.1 based on the manufacturer’s instructions, followed by the quantification using qPCR with the provided standard. Finally, the libraries were measured by the Element Biosciences AVITI system to sequence paired end reads with 75 bp each. The raw FASTQ files from RNA-seq were trimmed by Trimmomatic [103] to filter out low-quality reads. The surviving reads were then aligned to the mouse mm10 reference genome by STAR aligner [106]. Fragments Per Kilobase of transcript per Million mapped reads (FPKM) values per library were quantified by the tool Cufflinks [107]. Log 2 normalized fold induction and adjusted p values between Rbbp5 ^Flox DMSO and Rbbp5 ^LKO DMSO, Rbbp5 ^Flox Tu and Rbbp5 ^LKO Tu, Rbbp5 ^Flox DMSO and Rbbp5 ^Flox Tu, and Rbbp5 ^LKO DMSO and Rbbp5 ^LKO Tu were calculated by using the iDEP (integrated Differential Expression and Pathway analysis) online application [108].

Chromatin Immunoprecipitation (ChIP)-Seq and ChIP-qPCR

ChIP for RBBP5 was performed using anti-RBBP5 antibody (Bethyl, A300-109A) as previously described [17]. Briefly, mouse liver samples were submerged in PBS + 1% formaldehyde, cut into small (~1 mm³) pieces with a razor blade and incubated at room temperature for 15 min. Fixation was stopped by the addition of 0.125 M glycine (final concentration). The tissue pieces were then treated with a TissueTearer and finally spun down and washed twice in PBS. Chromatin was isolated by the addition of lysis buffer, followed by disruption with a Dounce homogenizer. The chromatin was enzymatically digested with MNase. Genomic DNA (Input) was prepared by treating aliquots of chromatin with RNase, Proteinase K and heated for reverse-crosslinking, followed by ethanol precipitation. Pellets were resuspended and the resulting DNA was quantified on a NanoDrop spectrophotometer. An aliquot of chromatin (30 μg) was precleared with protein A agarose beads (Invitrogen). Genomic DNA regions of interest were isolated using 4 μg of antibody. Complexes were washed, eluted from the beads with SDS buffer, and subjected to RNase and proteinase K treatment. Crosslinking was reversed by incubation overnight at 65 °C, and ChIP DNA was purified by phenol-chloroform extraction and ethanol precipitation. The DNA libraries were prepared at the University of Pittsburgh and sequenced at Gene by Gene, Ltd and Novagene per standard protocols. DNA libraries were prepared with Ovation Ultralow V2 DNA-Seq library preparation kit (NuGen) using 1ng input DNA. The size selection for libraries was performed using SPRIselect beads (Beckman Coulter) and purity of the libraries was analyzed using the High Sensitivity DNA chip on Bioanalyzer 2100 (Agilent). The prepared libraries pooled and sequenced using Nova-Seq 6000 (Illumina), generating ~30 million 75 bp single-end reads per sample for the first repeat and sequenced by Nova-Seq (Illumina), generated ~50 million 150 bp pair-end reads for the second repeat. ChIP-qPCR for MEFs was essentially performed the same way as previously described with anti-RBBP5 (Bethyl, A300-109A) and anti-H3K4me3 (Active Motif, 61379) antibodies, except that the MEFs were directly fixed with 1% formaldehyde before subject to nuclei isolation and chromatin immunoprecipitation. The primers used for ChIP-qPCR are as follows:

Gene desert forward primer: GCAACAACAACAGCAACAATAAC

Gene desert reverse primer: CATGGCACCTAGAGTTGGATAA

Manf promoter forward primer: CCCTTAAATGGGTCAACGTCTC

Manf promoter reverse primer: GGCGCTAAACCCAAGGAAA

Xbp1 promoter forward primer: TCCGTACGGTGGGTTAGAT

Xbp1 promoter reverse primer: ACCTTCTTCTGTGCCTGTG

Hyou1 promoter forward primer: CCGTGTGGGTACGTCCT

Hyou1 promoter reverse primer: GACGGCTGCTCCATCCT

Sesn2 promoter forward primer: ACAGGAGGCCGGGACTA

Sesn2 promoter reverse primer: CTGGGCTGAAAGGAGTGTCTAT

Negative control region 1 forward primer: TTCTGTCCCTGCAGTTCTTTC

Negative control region 1 reverse primer: GGCTGTGGTGTCTTGAGATTAG

Negative control region 2 forward primer: ACGTGATCCTAAGACCTCTAGT

Negative control region 2 reverse primer: GGGTAGCACTCTCCTTAGTTTC

ChIP-Seq analysis

The sequences identified were mapped to the mouse genome (UCSC mm10) using BOWTIE function in Galaxy Deeptool (https://usegalaxy.org/) [109]. Only the sequences uniquely mapped with no more than 2 mismatches were kept and used as valid reads. PCR duplicates were also removed. Peak calling was carried out by MACS2 (version 2.1.1.20160309) in Galaxy (options --mfold 5, 50 --p value 1e−4), on each ChIP-seq file against input. To account for the different sequencing depths between samples, the BAM files generated from MACS2 were RPKM normalized to sequencing depth using the bamCoverage function in Galaxy Deeptool and the bigwig files were generated accordingly. The relative intensity of each RBBP5 binding site is further calculated via the computeMatrix function with the RPKM normalized bigwig files and bed files from the peak calling as inputs by calculating the area under the curve. Temporal RBBP5 binding intensities from the two independent runs were processed separately, with each dataset detrended using a second-order polynomial followed by Z-score normalization. The normalized signals were then used for 12 h rhythm detection using RAIN.

Proximity labeling

Proximity labeling was performed using Protein A-TurboID as previously described [74]. Briefly, target cells were grown to 90%–100% confluency in 15 cm dishes, harvested in PBS using a cell scraper, and pelleted by centrifugation. Nuclei were isolated by resuspending the cell pellet in ice-cold HLB buffer, incubating on ice, and washing through sequential centrifugation steps. Isolated nuclei were incubated with digitonin buffer containing ~2 µg of IgG or H3K4me3 antibody (Active Motif, 61379) for 20 min at room temperature with shaking, followed by incubation with 1.4 µg of Protein A-TurboID for 30 min at 4 °C. After washing, biotinylation was induced by incubating samples in biotinylation buffer (containing ATP, MgCl₂, and 500 µM biotin) for 10 min at 37 °C. Nuclei were lysed in RIPA buffer at 4 °C overnight, sonicated, and centrifuged to collect the supernatant. Streptavidin affinity purification was performed by incubating lysates with M-270 streptavidin beads for 2 h at room temperature or overnight at 4 °C. Beads were washed extensively with RIPA buffer, and bound proteins were eluted by boiling at 100 °C for 15 min in 70 µL of 1.5X LDS buffer. Samples were clarified by centrifugation and magnetic separation.

For protein identification, half of each sample was separated ~1.5 cm on a 10% Bis-Tris Novex mini-gel (Invitrogen) using the MES buffer system. The gel was stained with Coomassie, and each lane was excised into 10 equally sized segments. Gel pieces were processed using a DigestPro robot (CEM) with the following protocol: washes with 25 mM ammonium bicarbonate and acetonitrile, reduction with 10 mM dithiothreitol at 60 °C, alkylation with 50 mM iodoacetamide at room temperature, digestion with trypsin (Promega) at 37 °C for 4 h, and quenching with formic acid. The resulting peptides were analyzed by nano LC-MS/MS using a Waters M-class HPLC system interfaced with a ThermoFisher Fusion Lumos mass spectrometer. Peptides were loaded onto a trapping column and eluted over a 75 µm analytical column at 350 nL/min with a 30-min gradient. MS and MS/MS spectra were acquired in the Orbitrap at resolutions of 60,000 and 15,000 FWHM, respectively. Data were searched using Mascot against the SwissProt Mouse database with a 10 ppm peptide mass tolerance and a 20 ppm fragment mass tolerance. Search parameters included trypsin digestion, fixed carbamidomethylation of cysteines, and variable modifications such as oxidation (M), acetylation (protein N-term), deamidation (NQ), and pyro-glu (N-term Q). Data were processed using Scaffold software with a 1% false discovery rate (FDR) at both the protein and peptide levels, requiring at least two unique peptides per protein.

Gene ontology analysis

DAVID (Version 2021) [110] (https://david.ncifcrf.gov) was used to perform Gene Ontology analyses. Briefly, gene names were first converted to DAVID-recognizable IDs using Gene Accession Conversion Tool. The updated gene list was then subject to GO analysis using all Homo Sapiens as background and with Functional Annotation Chart function. GO_BP_DIRECT, KEGG_PATHWAY, or UP_KW_BIOLOGICAL_PROCESS were used as GO categories. Only GO terms with a p value less than 0.05 were included for further analysis. In addition, for Fig 2C, enrichment analyses were performed using the clusterProfiler R package (v4.14.0). For each of the four gene sets, gene symbols were first converted to Entrez IDs using the bitr function with the mouse annotation database org.Mm.e.g.,db. Genes that failed to map (~3%–6% per group) were excluded. Over-representation analysis (ORA) for GO Biological Process terms was performed using the enrichGO function, and KEGG pathway enrichment was performed using the enrichKEGG function with organism code “mmu”. For both analyses, the complete set of annotated mouse genes (112,288 Entrez IDs) was used as the statistical background universe. P-values were adjusted for multiple testing using the Benjamini–Hochberg (BH) method, and terms with adjusted p-value < 0.05 and q-value < 0.2 were considered significant. Results were visualized as dot plots using ggplot2, with dot size representing the gene ratio (fraction of input genes annotated to each term) and dot color representing the −log10(adjusted p-value).

Motif analysis

Motif analysis was performed with the SeqPos motif tool (version 0.590) [111] embedded in Galaxy Cistrome using all motifs within the mouse reference genome mm10 as background. LISA analysis was performed using webtool (http://lisa.cistrome.org/).

Statistical analysis

Data were analyzed and presented with GraphPad Prism software (version 10.0.0). Exact p values were calculated for all statistical analyses unless they were smaller than 1.0e−15, which is the lower limit of detection in GraphPad Prism. In such cases, p  <  1.0e−15 is reported. All measurements were taken from distinct biological samples. Plots show individual data points and bars at the mean and ± the standard error of the mean (SEM). Individual tests used were provided at the end of each figure legend. Generally, two-tailed Student’s t-tests were used to compare differences between two groups, except in Fig 4I where strong prior evidence supports a unidirectional change and a subsequent utilization of one-tailed Student t test. Two-way ANOVA was used to compare the difference between two treatments with one additional variable (such as different proteostasis genes), such as data in Fig 4D. However, for data with unequal sample size, mixed effects analyses were used in lieu of Two-way ANOVA, such as data in S3A Fig. Cosinor2 was used to assess the presence of 12-h rhythms in different epigenetic marks (Fig 1B), as it is specifically designed for the analysis of short and sparsely sampled time-series data [43].

S1 Fig. ~12 h RBBP5 expression in mouse liver.

(A) COMPASS writes H3K4me3 using S-adenosyl methionine (SAM) as the substrate. (B) LISA revealing putative TFs and co-regulators for hepatic top 500 most robust (the smallest p values by RAIN analysis) circadian (y-axis) and ~12 h (x-axis) genes. (C) Temporal murine hepatic Rbbp5 expression assayed by RNA-seq (n = 2 biologically independent samples per time point). Period is calculated by the eigenvalue/pencil method. (D) western blot and quantification (n = 2 biologically independent samples per time points) of temporal nuclear RBBP5 level in mouse liver at different CTs normalized to total Ponceau S staining intensity. Period is calculated by eigenvalue/pencil method and is further tested against four different methods for the existence of ~12 h rhythm with respective P values shown. (E) Temporal expression of total RBBP5 proteins in different mice tissues previously reported in [54]. (F) Heatmap demonstrates the P values for having 12 h rhythm of RBBP5 expression in each tissue tested by four different methods. Data: Mean ± S.E.M. All data were collected from mice housed under constant darkness conditions with different times indicated by CT. Raw html file for B is available in S2 Raw Image file. Numerical values are available in S1 Data.

https://doi.org/10.1371/journal.pbio.3004048.s001

(JPG)

S2 Fig.  ~12 h RBBP5 cistrome in mouse liver.

(A) Heatmap showing two independent repeats of temporal RBBP5 ChIP-seq with each row corresponding to the same gene region. (B) Heatmap of RBBP5 (repeat 1), H3K9Ac and H3K27Ac chromatin occupancy 1kb ± of TSS of 6,451 genes. (C) Snapshot of target genes selected for alignment of hepatic RBBP5 (two repeats), H3K4me3, H3K9Ac, and H3K27Ac at different CTs. (D) Heatmap of temporal RBBP5 binding intensity (average of the two repeats) for 7,963 binding sites, along with –log₂ transformed adjP value for having ~12 h rhythm by RAIN. Those 4,667 ~12 h RBBP5 (adjP < 0.05) sites were highlighted. (E, F) Cumulative distribution of the number of ~12 h and circadian RBBP5 binding sites ranked by p value (E) or adj P value/false discovery rate (FDR) (F). All data were collected from mice housed under constant darkness conditions with different times indicated by CT. Numerical values are available in S1 Data.

https://doi.org/10.1371/journal.pbio.3004048.s002

(JPG)

S3 Fig. Hepatic RBBP5 cistrome is associated with proteostasis and mRNA metabolic genes.

Dot plot for GO analysis demonstrating the −log₁₀ transformed adjP value and gene counts of different enriched pathways for different genes. Numerical values are available in S2 Data.

https://doi.org/10.1371/journal.pbio.3004048.s003

(JPG)

S4 Fig. The generation of Rbbp5 ^LKO mice.

(A, B) Expected (A) and actual genotyping result (B) for different mice. (C, D) western blot (C) and quantification (D) of RBBP5 in different tissues in Rbbp5 (fl/fl); Alb-CRE (+/−) and Rbbp5 (fl/fl); Alb-CRE (−/−) mice. n = 2 biologically independent samples per genotype. Data: Mean ± S.E.M. Statistical tests used: unpaired two-tailed Student t test for D. Numerical values are available in S3 Data.

https://doi.org/10.1371/journal.pbio.3004048.s004

(JPG)

S5 Fig. Liver hepatocyte-specific deletion of RBBP5 does not alter rhythmic locomotor activity nor fasting-feeding cycles in mice.

(A) Average body weight of male Rbbp5 ^Flox and Rbbp5 ^LKO mice at different ages. Mean ± 95% confidence interval. n = 3 ~ 71 mice for Rbbp5 ^Flox and n = 6 ~ 54 mice for Rbbp5 ^LKO mice for each week. (B) Real-time locomotor activity monitoring in Rbbp5 ^Flox and Rbbp5 ^LKO mice under 12 hr light/12 hr dark conditions. (C) Averaged measurements within the light and dark phase as described in B. (D) Real-time measurement of food intake in Rbbp5 ^Flox and Rbbp5 ^LKO mice under 12 hr light/12 hr dark conditions. (E) Averaged measurements within the light and dark phase as described in D. n = 6 for Rbbp5 ^Flox and n = 8 for Rbbp5 ^LKO mice. (F) Real-time locomotor activity monitoring in Rbbp5 ^Flox and Rbbp5 ^LKO mice under constant dark conditions. (G) Averaged measurements within the rest and active phase as described in F. (H) Real-time measurement of food intake in Rbbp5 ^Flox and Rbbp5 ^LKO mice under constant dark conditions. (I) Averaged measurements within the rest and active phase as described in H. n = 5 for Rbbp5 ^Flox and n = 3 for Rbbp5 ^LKO mice. Data: Mean ± S.E.M for C, G, E and I. Statistical tests used: unpaired two-tailed Student t test for C, G, E and I. Mixed effects analysis for A. Data from F, G, H and I were collected from mice housed under the constant darkness condition with different times indicated by CT. Data from B, C, D, and E were collected from mice housed under the 12 h light on/12 h light off condition, with ZT0 and ZT12 indicating light on and light off time, respectively. Numerical values are available in S3 Data.

https://doi.org/10.1371/journal.pbio.3004048.s005

(JPG)

S6 Fig. RBBP5 is an epigenetic regulator of the hepatic ~12 h oscillator, but not the canonical ~ 24 h circadian clock.

(A) A table listing the number of ~12 h and ~24 h circadian genes in Rbbp5 ^Flox and Rbbp5 ^LKO mice with different statistical thresholds. (B) Heat map of 6,025 ~12 h gene expression (FDR < 0.05) in the liver of Rbbp5 ^Flox and Rbbp5 ^LKO mice, with 5,611 of them lost in Rbbp5 ^LKO mice, along with −log₁₀ transformed adjP values for having 10−12 h rhythm by RAIN. Log₂ normalized fold change of average gene expression across 48 hours between Rbbp5 ^Flox and Rbbp5 ^LKO mice was also shown. (C) Heat map summary of GO analysis demonstrating the −log₁₀ transformed P values of different enriched pathways for different genes. All data were collected from mice housed under constant darkness conditions with different times indicated by CT. Numerical values are available in S3 Data.

https://doi.org/10.1371/journal.pbio.3004048.s006

(JPG)

S7 Fig. MetaCycle and cosinor2 methods reveal RBBP5-dependent hepatic  ~12 h transcriptome.

(A-D) MetaCycle method: Histograms showing the period distributions of all rhythmic genes uncovered from Rbbp5 ^Flox and Rbbp5 ^LKO mice with P value (A) and adjP values (B) cut off (<0.05). Cumulative distribution of 10 ~12 h genes with different FDR cut-offs (C) and number of 10-12 h and 22h-28h genes in Rbbp5 ^Flox and Rbbp5 ^LKO mice with FDR cutoff of 0.05 (D). (E-H) Cosinor2 method: Histograms showing the period distributions of all rhythmic genes uncovered from Rbbp5 ^Flox and Rbbp5 ^LKO mice with P value (E) and adjP values (F) cut off (<0.05). Cumulative distribution of 10 ~ 12 h genes with different FDR cut-offs (G) and number of 10-12 h and 22h-28h genes in Rbbp5 ^Flox and Rbbp5 ^LKO mice with FDR cutoff of 0.05 (H). Numerical values are available in S3 Data.

https://doi.org/10.1371/journal.pbio.3004048.s007

(JPG)

S8 Fig. Eigenvalue/pencil method analysis of RBBP5-dependent hepatic ~12 h transcriptome.

(A, B) Distributions of all (A) and dominant (B) oscillations uncovered by the eigenvalue/pencil method in Rbbp5 ^Flox and Rbbp5 ^LKO mice. (C) Heat map of all ~12 h gene expressions (or lack thereof) in Rbbp5 ^Flox and Rbbp5 ^LKO mice with both raw data and superimposed ~12 h rhythms shown. 4,522 ~12 h genes were abolished, 238 ~12 h genes dampened, and 392 ~12 h genes were enhanced in Rbbp5 ^LKO mice, respectively. Log₂ normalized fold change of average gene expression across 48 hours between Rbbp5 ^Flox and Rbbp5 ^LKO mice was also shown. (D) The periods distribution in Rbbp5 ^LKO mice of those ~12 h genes originally identified in Rbbp5 ^Flox mice but lost in Rbbp5 ^LKO mice. (E) Venn diagram comparing unique and common ~12 h transcriptomes uncovered by the eigenvalue/pencil and RAIN methods with two different FDR cut-offs. All data were collected from mice housed under constant darkness conditions with different times indicated by CT. Numerical values are available in S3 Data.

https://doi.org/10.1371/journal.pbio.3004048.s008

(JPG)

S9 Fig. RBBP5 is a cell-autonomous epigenetic regulator of the ~12 h oscillator, but not the canonical ~ 24h circadian clock.

(A) qPCR analysis of Rbbp5 expression in Tu (25ng/ml)-synchronized MEFs with scrambled, Bmal1 and Xbp1 siRNAs. RAIN [55] was used to test for the existence of 12 h rhythms with P values shown (n = 3 ~ 12 biologically independent samples per time point). (B) Representative raw (left) and polynomial detrended (middle) real-time luminescence of MEFs expressing Manf promoter-driven dluc [17] transfected with different siRNAs and quantified periods (n = 4 ~ 5 biologically independent samples) (right). (C) Real-time luminescence traces of Bmal1-dluc U2OS cells transfected with different siRNAs as reported in [58] (n = 13, 2 and 2 for scrambled, RBBP5 and ASH2L siRNA, respectively). Data: Mean ± S.E.M. Statistical tests used: RAIN for A, unpaired two-tailed Welch's t test for B, and unpaired two-tailed Student t test for C. Numerical values are available in S3 Data.

https://doi.org/10.1371/journal.pbio.3004048.s009

(JPG)

S10 Fig. RBBP5 regulates the hepatic transcriptional response to proteotoxic stress.

(A−D) Volcano plot illustrating the log ₂ normalized fold change versus −log2 transformed adjusted p values for different comparisons. (E) Scatter plot comparing the log ₂ transformed fold change of gene expression by Tu in Rbbp5 ^Flox (x- axis) and Rbbp5 ^LKO (y-axis) mice. (F) Scatter plot comparing the −log ₁₀ transformed adjusted p values for gene expression induced by Tu in Rbbp5 ^Flox (x- axis) and Rbbp5 ^LKO (y-axis) mice. (G) Venn diagram comparing distinct and common genes induced by Tu in Rbbp5 ^Flox and Rbbp5 ^LKO mice with FDR < 0.1. (H) Heat map summary of GO analysis demonstrating the −log₁₀ transformed P values of different enriched pathways for different groups of genes. Numerical values are available in S4 Data.

https://doi.org/10.1371/journal.pbio.3004048.s010

(JPG)

S12 Fig. RBBP5 is required for transcriptional responses to ER stresses.

(A) Western blot of RBBP5 in MEFs transfected with scrambled or Rbbp5 siRNA for 48 h. (B) qPCR of different genes in MEFs transfected with scrambled, Rbbp5 or Xbp1 siRNAs and treated with DMSO or Tu (100 ng/ml) for 8 h (n = 2~4 biologically independent samples). (C) qPCR of different genes in MEFs transfected with scrambled, Ash2l or Setd1a siRNAs and treated with DMSO or Tu (100 ng/ml) for 8 h (n = 2 ~ 4 biologically independent samples). (D) qPCR of different genes in MEFs transfected with scrambled, or the combined Gcn5 and Pcaf siRNAs and treated with DMSO or Tu (100 ng/ml) for 8 h (n = 2~6 biologically independent samples). (E, F) Isolation and immortalization of Rbbp5 (fl/fl) ROSA26-CreERT2 (+/+) MEFs (E), which were treated with vehicle (WT) or tamoxifen (Rbbp5 KO) for 5 or 7 days and genotyping was performed (F) to confirm the deletion of floxed Rbbp5 allele. (G) Rbbp5 (fl/fl) ROSA26-CreERT2 (+/+) MEFs were treated with tamoxifen for 5 or 7 days and western blot analysis was performed. (H) qPCR of different genes in Rbbp5 (fl/fl) ROSA26-CreERT2 (+/+) MEFs treated with vehicle (WT) or tamoxifen (Rbbp5 KO) followed by treatment with vehicle control or 1 mM DTT for 5 hours (n = 3 biologically independent samples). Data: Mean ± S.E.M. Statistical tests used: Two-way ANOVA for B, C and H, unpaired two-tailed Student t test for D. Numerical values are available in S4 Data.

https://doi.org/10.1371/journal.pbio.3004048.s012

(JPG)

S13 Fig. RBBP5 is required for transcriptional responses to nutrition deprivation.

(A) qPCR analysis of Rbbp5 (fl/fl) ROSA26-CreERT2 (+/+) MEFs pre-treated with (RBBP5 KO) or without (WT) tamoxifen and then treated with full DMEM media, leucine-free media or EBSS for 16 hours (n = 3 ~ 6 biologically independent samples). (B) PCA plot of transcriptome of ROSA26-CreERT2 (+/+) MEFs treated with vehicle (WT) or tamoxifen (Rbbp5 KO) followed by treatment with full DMEM media or EBSS for 16 hours. (C) Diagram of the mCherry-GFP-LC3 autophagy reporter. (D, E) Representative images (D) and quantification (E) of autophagic flux in mCherry-GFP-LC3-expressing Rbbp5 (fl/fl) ROSA26-CreERT2 (+/+) MEFs treated with vehicle (WT) or tamoxifen (Rbbp5 KO) followed by treatment with full DMEM media or leucine-free media for 16 hours (n = 18~26 cells per condition). (F) western blot of different proteins in Rbbp5 (fl/fl) ROSA26-CreERT2 (+/+) MEFs treated with vehicle (WT) or tamoxifen (Rbbp5 KO) followed by treatment with full DMEM media or leucine free media for 16 hours. (G) Percentage of cells normalized to before treatment for mCherry-GFP-LC3-expressing Rbbp5 (fl/fl) ROSA26-CreERT2 (+/+) MEFs treated with vehicle (WT) or tamoxifen (Rbbp5 KO) followed by treatment with full DMEM media, leucine-free media or EBSS for up to 16 hours (n = 3~4 biologically replicates per condition). Data: Mean ± S.E.M. Statistical tests used: Two-way ANOVA for A and G, unpaired two-tailed Student t test for E. Numerical values are available in S4 Data.

https://doi.org/10.1371/journal.pbio.3004048.s013

(JPG)

S14 Fig. Proximity labeling reveals H3K4me-associated proteomic architecture of the proteotoxic stress response.

(A) Imperial blue staining of total biotinylated proteins in MEF treated with 100ng/ml Tu for 0, 2.5, or 8 h before subject to anti-IgG and anti-H3K4me3 antibody-based proximity labeling. Unspecific bands (in a different gel) denoted by asterisks are excluded from mass spectrometry analysis by gel removal. (B) A diagram illustrating common and specific subunits of mammalian COMPASS complex. (C–E) Heatmap of fold enrichment of different subunits in COMPASS (C), RNAP II (D) or Integrator Complex (E) (over IgG control) at different times after ER stress. Numerical values are available in S4 Data.

https://doi.org/10.1371/journal.pbio.3004048.s014

(JPG)

S1 Table. FPKM quantification of temporal hepatic RBBP5 ChIP-seq in wild-type mouse liver.

Tab1: Quantification for repeat 1. Tab 2: Quantification for repeat 2. Tab 3: RAIN quantification for 12 h cycling RBBP5 peaks. Tab 4: Eigenvalue/pencil for 12 h cycling RBBP5 peaks. Tab 5: RAIN quantification for 24 h cycling RBBP5 peaks.

https://doi.org/10.1371/journal.pbio.3004048.s015

(XLSX)

S4 Table. Eigenvalue/pencil analysis of oscillating hepatic transcriptome identified in Rbbp5 ^Flox and Rbbp5 ^LKO mice.

Tab 1: All oscillations identified in Rbbp5 ^Flox mice. Tab 2: All dominant oscillations identified in Rbbp5 ^Flox mice. Tab 3: All 10~13 h oscillations identified in Rbbp5 ^Flox mice. Tab 4: All dominant 10~13 h oscillations identified in Rbbp5 ^Flox mice. Tab 5: All oscillations identified in Rbbp5 ^LKO mice. Tab 6: All dominant oscillations identified in Rbbp5 ^LKO mice. Tab 7: All 10~13 h oscillations identified in Rbbp5 ^LKO mice. Tab 8: All dominant 10~13 h oscillations identified in Rbbp5 ^LKO mice.

https://doi.org/10.1371/journal.pbio.3004048.s018

(XLSX)

S1 Raw Images. Original images of western blot gels shown in Figs 2G, 3K, 4B, 4G, 4H, 5H, 6B, S1D, S4B, S4C, S12A, S12F, S12G, S13F, and S14A.

https://doi.org/10.1371/journal.pbio.3004048.s024

(PDF)

S3 Data. Original numerical values for Figs 3C–3L, S4D, S5A–S5I, S6B, S6C, S7A, S7B, S7C, S7E, S7F, S7G, S8A–S8D, and S9A–S9C.

https://doi.org/10.1371/journal.pbio.3004048.s028

(XLSX)

S4 Data. Original numerical values for Figs 4B–4I, 5A, 5C, 5D, 5E, 5F, 5I, 6D, 6E, S10A–S10F, S10H, S12B, S12C, S12D, S12H, S13A, S13B, S13E, S13G, and S14C–S14E.

https://doi.org/10.1371/journal.pbio.3004048.s029

(XLSX)