Carbon dioxide regulates cholesterol levels through SREBP2

In mammals, O2 and CO2 levels are tightly regulated and are altered under various pathological conditions. While the molecular mechanisms that participate in O2 sensing are well characterized, little is known regarding the signaling pathways that participate in CO2 signaling and adaptation. Here, we show that CO2 levels control a distinct cellular transcriptional response that differs from mere pH changes. Unexpectedly, we discovered that CO2 regulates the expression of cholesterogenic genes in a SREBP2-dependent manner and modulates cellular cholesterol accumulation. Molecular dissection of the underlying mechanism suggests that CO2 triggers SREBP2 activation through changes in endoplasmic reticulum (ER) membrane cholesterol levels. Collectively, we propose that SREBP2 participates in CO2 signaling and that cellular cholesterol levels can be modulated by CO2 through SREBP2.


Introduction
A fundamental process in mammalian physiology is oxygen (O 2 ) uptake from the environment into cells in exchange of carbon dioxide (CO 2 ), a byproduct of energy generation upon aerobic respiration.Oxygen is an essential substrate for cellular metabolism and bioenergetics and is indispensable for normal physiology and survival.Consequently, mammals have developed mechanisms to sense O 2 levels and regulate O 2 consumption in order to cope with conditions of insufficient O 2 supply.A principal regulator in the response to low oxygen levels is the hypoxia-inducible factor (HIF), which participates in sensing of low oxygen levels and subsequently activates a transcriptional program that facilitates cellular adaptation to changes in oxygen levels [1][2][3].While the cellular response to oxygen levels is well characterized, relatively little is known regarding the mechanisms that participate in response to changes in CO 2 levels.It is noteworthy that CO 2 plays various critical roles in mammalian physiology including regulation of blood pH, respiratory drive, and O 2 affinity for hemoglobin [4].Under physiological conditions, arterial blood CO 2 levels are tightly maintained approximately 35 to 45 mm Hg (approximately 5%).Altered CO 2 levels are associated with the pathophysiology of various diseases such as chronic obstructive pulmonary disease (COPD) and obstructive sleep apnea (OSA) as well as impaired wound healing and fibrosis [5][6][7].
Carbon dioxide molecules are transported in the blood from body tissues to the lungs by one of 3 methods: dissolution directly into the blood, binding to hemoglobin, or carried as a bicarbonate ion.About 10% of CO 2 is dissolved in the plasma, a small fraction is bound to hemoglobin, while the majority (about 85%) is carried as a part of the bicarbonate buffer system [4].In aqueous solution, CO 2 reacts with the water to form carbonic acid (H 2 CO 3 ), which is readily buffered by the bicarbonate buffer system to maintain the pH levels within the physiological range [8].
To identify signaling pathways that regulate gene expression in response to changes in CO 2 levels, and hence participate in CO 2 sensing, we employed a cell culture setup alongside highthroughput transcriptomic and biochemical analyses.We found that CO 2 activates a distinct transcriptional response that is dependent on SREBP2, a key regulator of cholesterol biosynthesis, to regulate the expression of cholesterogenic genes and cholesterol accumulation.SREBP2 regulation by CO 2 is likely mediated by changes in endoplasmic reticulum (ER) membrane cholesterol levels.We, thus, propose that SREBP2 plays a role in cellular CO 2 signaling and that SREBP2 regulation of cholesterol levels can be modulated by changes in CO 2 levels.

The transcriptional response to low CO 2 differs from pH
To identify signaling pathways that participate specifically in CO 2 sensing and not changes in pH, we examined the global transcriptional response of cultured cells to reduction in CO 2 levels from 5% to 1%.We used special chambers with CO 2 , O 2 and temperature controls [9].Temperature, O 2 and CO 2 , levels were continuously monitored throughout the experiment with constant temperature of 37˚C and 20% O 2 .While CO 2 levels were modulated by replacing them with the inert nitrogen gas and were kept either at 5% or 1% (Fig 1A).
As aforementioned, once CO 2 reacts with aqueous solution it forms carbonic acid and acidifies it.Since the reduction in CO 2 levels from 5% to 1% resulted in alkaline condition, we also used 18 mM NaOH to alkalize the media as a control for changes that are purely pH-dependent.Importantly, under both conditions, namely 1% CO 2 or 18 mM NaOH, the media pH at 2 and 4 h post exposure was similar (approximately 7.7) and differed from that of control cells (5% CO 2 ), which maintained pH across the normal physiological range (approximately 7.3) (Fig 1B).NIH3T3 cells (a fibroblast cell line that was isolated from mouse NIH/Swiss embryos) were harvested 2 and 4 h post exposure, RNA was extracted and analyzed by RNA-sequencing.The transcriptomic analysis revealed that the transcriptional response differed between the low CO 2 exposure and the alkaline conditions, even though the pH was similar (Figs 1C and 1D and S1).Notably, principal component analysis (PCA) and unsupervised clustering analyses (Fig 1C and 1D) clearly discriminated between exposure to low CO 2 versus NaOH treatment.NaOH treatment induced a prominent effect on gene expression with 2,697 genes showing differential response (P adj.< = 0.05, |log2FC| > = 1, baseMean > = 5), with 1,320 up-and 1,377 down-regulated.While exposure to a low CO 2 level led to a milder effect on gene expression (1,328 genes with 685 up-and 643 down-regulated) (S1A and S1B Fig) .Although, both the up-and down-regulated genes overlapped between the CO 2 and NaOH groups, we found in line with the PCA and cluster analyses that a significant number of genes are uniquely altered in response to CO 2 (S1B Fig) .Overall, our analyses show that under similar alkaline pH, the transcriptional response differs between low CO 2 and NaOH treatments.Thus, supporting a distinct mechanism that is activated in response to changes in CO 2 levels to regulate gene expression.

CO 2 alters the expression of genes that participate in cholesterol biosynthesis
To identify potential transcription factors that participate in gene expression regulation in response to CO 2 or NaOH, we took an advantage of our time course analysis and performed unbiased cluster analysis (Fig 2A).We identified 3 major clusters; Cluster 1: Transcripts that were monotonically down-regulated (CO 2 or NaOH; 511 and 1,102, respectively); Cluster 2:   ).(B) Pathway enrichment analysis was performed using the IPA tool for the genes included in each of the clusters for 1% CO 2 or 18 mM NaOH treatments.Presented are the top 3 enriched pathways in each cluster based on P value (for full list of pathways, see S3 Table ).(C) Schematic illustration of the cholesterol biosynthesis pathway alongside genes that were significantly affected by the treatments.Color indicates on the condition in which the genes are affected.(D) Heatmap representation of cholesterogenic genes that were significantly affected by any of the conditions presented.Data are presented as row z-score of the average expression per condition (n = 4 biological replicates).(E) Quantitative PCR analysis of cholesterogenic gene expression levels from NIH3T3 cells treated with 1% CO 2 or 18 mM NaOH (mean ± SE, n = 5 biological replicates per time point per condition, *P < 0.05, **P < 0.01, Transcripts that were up-regulated exclusively after 2 h (CO 2 or NaOH; 206 and 463, respectively); and Cluster 3: Transcripts that were monotonically up-regulated (CO 2 or NaOH; 315, and 606, respectively).Next, to uncover related biological processes affected by each treatment, we performed pathway enrichment analysis on each cluster.Remarkably, we found that cholesterol biosynthesis and its related processes are highly enriched in response to CO 2 but not to NaOH, specifically in cluster 3 which includes the monotonically up-regulated transcripts (Fig 2B).These findings indicated that low CO 2 induces the expression of genes implicated in cholesterol metabolism and that this effect is not a mere response to alkaline conditions, as it was not apparent upon NaOH treatment.This prompted us to specifically examine expression pattern of enzymes involved in de novo cholesterol biosynthesis based on our RNA-sequencing data.The vast majority of enzymes involved in different stages of cholesterol biosynthesis were up-regulated in cells exposed to low CO 2 .Notably, the induction of these transcripts was mostly absent in NaOH-treated cells (Fig 2C and 2D).Furthermore, analysis of cholesterogenic gene expression by qPCR showed that in most cases their transcript levels are specifically induced by low CO 2 levels but not upon NaOH treatment (Fig 2E).These results were in line with the above detailed RNA-sequencing analysis.A similar trend was observed in hepatocyte murine cell line (Hepa1c1) (S2A Fig) .In addition, these effects were recapitulated in primary tail fibroblasts and primary muscles, but not in primary white or brown adipocytes (S2A Fig) .Next, we examined the effect of hypercapnia, namely elevated CO 2 level, on cholesterogenic gene expression.Cells were exposed to increased CO 2 level (i.e., 10%) for 2 and 4 h and the transcript levels of cholesterogenic genes were analyzed by qPCR.Here again, O 2 level was maintained constant at 20% using our CO 2 , O 2 and temperature-controlled chambers.High CO 2 levels elicited the opposite effect to lower CO 2 levels and the expression levels of cholesterogenic genes were suppressed (S2B Fig) .Comparison of gene expression data of THP-1 monocytes exposed to 10% CO 2 [10] with our NIH3T3 cells data (1% CO 2 exposure) showed a small overlap in the responsive genes (S3A Fig) .Yet, this small group included cholesterogenic genes (e.g., Ldlr, Idi1, Insig1, Hmgcs1, Dhcr7) and their response was in line with our findings, namely 10% CO 2 repressed of their expression (e.g., Insig1, Hmgcs1) (S3B Fig).
Taken together, our analyses reveal that alteration of CO 2 levels from the physiological range modulate the expression of genes involved in cholesterol homeostasis.Reduced and elevated CO 2 levels activate and repress their expression, respectively.

SREBP2 is activated in response to low CO 2 to induce the expression of cholesterogenic genes
SREBP2 is a key transcriptional regulator of genes involved in cholesterol biosynthesis [11,12].In response to changes in cholesterol levels, SREBP2 translocates from the ER to the Golgi, where subsequent cleavage occurs and the N-terminal form of SREBP2 shuttles to the nucleus and activates the expression of transcripts involved in cholesterol biosynthesis [13].Our transcription factor analysis predicted SREBP2 among the top potential transcriptional regulators for the expression of genes that are up-regulated (clusters 2 and 3) upon exposure to low CO 2 but not in response to NaOH treatment (Fig 3A and 3B).
We, therefore, hypothesized that SREBP2 is activated in response to low CO 2 to induce the expression of enzymes involved in cholesterol biosynthesis.To test this, cultured cells were exposed to low CO 2 and SREBP2 was analyzed by SDS-PAGE and immunoblot analysis.We found that the cleaved form of SREBP2 (approximately 68 kD) accumulates 2 and 4 h following exposure to low CO 2 levels (Fig 3C).This effect was specific to low CO 2 and not to alkalic pH as it was not observed in NaOH-treated cells (S4A Fig) .Biochemical nuclear-cytoplasmic fractionation further showed that the cleaved form of SREBP2 accumulates in the nucleus upon exposure to low CO 2 levels (Fig 3D).Together, our findings indicate that the SREBP2 signaling pathway is activated upon exposure to low CO 2 levels.To examine the functional consequence of SREBP2 and its downstream gene activation, we performed a time course analysis (0, 3, 6, 12, and 24 h) and measured cholesterol levels in cells cultured either at 5% or 1% CO 2 .Upon 24 h exposure to low CO 2 levels, cells accumulated cholesterol, in line with SREBP2 activation and elevated the expression of cholesterogenic genes (Fig 3E).Next, we asked whether the induction of cholesterogenic genes under low CO 2 is SREBP2-dependent.To this end, cells were transfected with either control siRNA (siNTC-Non Template Control) or siRNA against mouse SREBP2 (siSREBP2) and were exposed either to 1% CO 2 or 5% CO 2 for 4 h.As expected, SREBP2 was undetectable in siSREBP2-silenced cells under both 5% and 1% CO 2 and the basal expression levels of SREBP2 target genes was lower (Fig 3F and 3G).Control cells showed accumulation of the cleaved form of SREBP2 upon 1% CO 2 as well as induction of its cholesterogenic target genes (Fig 3F and 3G).Importantly, the induction of cholesterogenic genes was completely abolished in SREBP2 silenced cells under low CO 2 levels, indicating that the effect is SREBP2-dependent (Fig 3F).We also identified several transcripts that are induced upon low CO 2 levels in our gene expression analysis yet their induction was SREBP2-independent (S4B Fig) .It is conceivable that the response to low CO 2 levels is coordinated through the concerted action of several transcription regulators and is not exclusively SREBP2-dependent.Overall, our results suggest that low CO 2 levels elicit SREBP2 cleavage and nuclear accumulation to induce the expression of its target genes, primarily cholesterogenic genes and consequently cholesterol accumulation.

Low CO 2 activates gene expression through a sterol regulatory element
SREBP2 activates the transcription of its downstream targets by binding to a specific region on the promoter sequence known as sterol regulatory element (SRE) [14].To examine whether low CO 2 levels can activate gene expression through an SRE, we employed an SRE reporter assay.This reporter is based on the HMG-CoA synthase promoter sequence harboring SRE that drive the expression of a firefly luciferase [15].Cells were transfected with the SRE reporter and bioluminescence was continuously monitored.Consistent with the activation of SRE by SREBP2, treatment with simvastatin, which inhibits de novo cholesterol biosynthesis [16] and activates SREBP2, resulted in increased bioluminescence.This effect was suppressed upon co-administration of fatostatin (Fig 3H ), which inhibits SREBP2 ER-to-Golgi translocation [17].
Then, we tested the effect of low CO 2 on the reporter activity.In line with above-described findings, a decrease in CO 2 levels from 5% to 1% induced an increase in bioluminescence of cells expressing the wild-type reporter (pSynSRE-T-Luc) (Fig 3I ).A decrease in CO 2 levels had no effect on the bioluminescence of cells expressing either a mutant reporter (pSynSRE-Mut-T-Luc) [18] or a control luciferase reporter (pcDNA3-Luc) (Fig 3I).Consistently, an increase in CO 2 levels from 5% to 10% markedly suppressed the bioluminescence from cells expressing a wild type but not a control luciferase reporter (S5A Fig).
In our bioluminescence reporter assays, we observed an initial minor response that was not SRE-specific and was evident in the control reporters as well.This unspecific response likely stems from the effect of pH changes on bioluminescence in general [19].
Next, we employed our reporter assay to examine whether SRE activation by low CO 2 is reversible.To this end, cells expressing wild-type SRE reporter were exposed to either constant 5% as a control or interchanging 5% to 1% CO 2 levels and bioluminescence was continuously recorded.A shift in CO 2 levels from 5% to 1% increased the bioluminescence levels.This increase was reduced back to basal levels once CO 2 levels were shifted to 5% (S5B Fig) .This result indicates that CO 2 reversibly modulate SRE activation and likely SREBP2 activation.
Taken together, our results suggest that an intact SRE is sufficient for the transcriptional response to changes in CO 2 levels and the effects of CO 2 levels on it are reversible.

Stability of the mature cleaved form of SREBP2 is not affected by low CO 2 levels
SREBP2 translocates from the ER-to-Golgi and subsequently reaches the nucleus to induce gene expression.The exit of SREBP2 from the ER is regulated by sterol levels via SREBP cleavage-activating protein (SCAP) and insulin-induced gene (INSIG).Low ER cholesterol levels destabilize INSIG-SCAP interaction and successively enable the SREBP2-SCAP complex to translocate from the ER to Golgi where SREBP2 is cleaved [20].The mature N-terminal cleaved form of SREBP2 then shuttles to the nucleus [13] to activate gene expression as aforementioned through SRE sites on target genes [21].
Hitherto, we showed that upon low CO 2 levels SREBP2 is cleaved, the N-terminal cleaved form accumulates in the nucleus and can activate gene expression though an intact SRE site (Fig 3).To identify the signaling node though which SREBP2 is activated in response to low CO 2 levels, we systematically examined the different steps in the SREBP2 signaling pathway (S6A Fig) comparing sterol depletion with exposure to low CO 2 levels.
In the nucleus, the levels of mature cleaved form of SREBP2 are regulated by its proteasomal degradation [22] as stabilization of the nuclear form by proteasome inhibition or defective polyubiquitination actively induce its target genes [23,24].We hypothesized that low CO 2 levels might alter nuclear SREBP2 turnover and thereby induce its nuclear accumulation and target gene expression.To test this, we exogenously expressed in cultured NIH3T3 cells a FLAGtagged truncated mature SREBP2 fragment (FLAG N-SREBP2) [25], which was shown to localize in the nucleus [26].Cells were exposed either to sterol depletion upon methyl-β-cyclodextrin (MBCD) treatment or to 1% CO 2 levels.Total protein extracts were prepared and analyzed by immunoblot with either anti-SREBP2 or anti-FLAG antibody to detect the endogenous or the exogenously expressed truncated forms, respectively.Both MBCD treatment and exposure to low CO 2 induced the accumulation of the endogenous cleaved form of SREBP2 (S6B Fig) .However, neither treatment affected the levels of the exogenously expressed cleaved form (i.e., FLAG N-SREBP2) (S6C Fig) , suggesting that low CO 2 levels, similar to sterol depletion by MBCD do not affect the nuclear stability of the cleaved mature form of SREBP2.

Low CO 2 levels induce the ER-to-Golgi translocation of SREBP2
SCAP-SREBP2 ER-to-Golgi translocation is a critical step in SREBP2 activation and subsequent induction of its target genes.To examine whether the activation of SREBP2 upon low CO 2 is dependent on its ER-to-Golgi trafficking, we employed fatostatin, which pharmacologically blocks the ER-to-Golgi transport of SCAP-SREBP2 [17].Cells were exposed to either fatostatin or DMSO as control under 5% or 1% CO 2 .Low CO 2 levels induced the accumulation of the mature cleaved form of SREBP2.Importantly, this effect was blocked in the presence of fatostatin (S6D Fig) .Consistently, the induction of SREBP2 target genes in response to low CO 2 levels was eliminated in the presence of fatostatin (S6E Fig) .This result indicated that ER-to-Golgi trafficking is necessary for activation of SREBP2 by low CO 2 levels.
As aforementioned, SCAP regulates SREBP2 transport in a sterol-dependent fashion as it retains the SCAP-INSIG-SREBP2 complex in the ER membrane and inhibits the subsequent processing of SREBP2, namely, its cleavage and ER-Golgi translocation [27].We, therefore, examined whether activation of SREBP2 by low CO 2 levels is also SCAP-sensitive.We employed siRNA to knockdown SCAP and exposed control (siNTC) or SCAP knockdown (siSCAP) cells to low CO 2 levels (i.e., 1%).Low SCAP levels in cultured cells were shown to suppress SREBP2 proteolysis and expression of SREBP2 downstream target genes [28,29].The induction of SREBP2 target genes in response to low CO 2 levels was as well suppressed in SCAP-deficient cells likely due to inhibition of SREBP2-SCAP ER-to-Golgi translocation (S6F Fig) .Together, our analyses suggest that activation and induction of SREBP2 target genes upon low CO 2 levels is dependent on ER-to-Golgi trafficking and regulated by SCAP.Hence, it seems to follow the canonical pathway of SREBP2 activation as in response to low sterol levels.

Low CO 2 levels reduces ER cholesterol levels
The main driver of SREBP2 signaling pathway is reduction in ER cholesterol levels.Hitherto, activation of SREBP2 by low CO 2 levels followed similar steps in the canonical SREBP2 pathway as upon sterol depletion.These findings raised the following questions: (i) Does SREBP2 activation by low CO 2 levels depend on cellular cholesterol levels; and (ii) do CO 2 levels affect cellular cholesterol levels?
First, we examined if SREBP2 activation by low CO 2 levels is affected by cellular cholesterol levels.To this end, we exposed cells expressing the SRE reporter to increasing concentrations of MBCD to deplete cholesterol while shifting CO 2 levels from 5% to 1% (Fig 4A).In line with the above, both sterol depletion under 5% CO 2 as well as exposure to 1% CO 2 levels, increased the bioluminescence of the SRE reporter (Fig 4B).Up to 5 mM MBCD, we observed an additive effect in response to 1% CO 2 .Whereas in the presence of higher levels of MBCD, namely, 7 mM, low CO 2 levels elicited a very minor effect on the activation of the SRE reporter (Fig 4B).The diminished effect of low CO 2 upon elevated levels of MBCD and likely highly depleted cholesterol levels, raised the possibility that low CO 2 levels activate SREBP2 in a cholesterol-dependent manner.
To directly examine whether low CO 2 affected cellular cholesterol levels, we first quantified total cholesterol levels in cells exposed to 5% or 1% CO 2 for 4 h or upon MBCD treatment.As expected in MBCD treated cells, we observed a marked reduction in total cholesterol levels.By contrast, low or high CO 2 levels did not show any significant effect on total cellular cholesterol (Figs 4C and S7A, respectively).
SREBP2 is specifically activated in response to changes in ER sterol levels [30].Furthermore, changes in ER cholesterol levels are sufficient to activate SREBP2 even if total cholesterol levels are unaltered [31].This prompted us to examine whether CO 2 alterations specifically affect ER cholesterol levels.Hence, we repeated the above-described experiment, but this time ER membranes were isolated through differential centrifugation, with subsequent sucrose gradient and OptiPrep separation, as previously described [30].ER membrane free cholesterol and cholesterol ester content were quantified (Fig 4D and 4E).MBCD treatment significantly reduced ER cholesterol (Fig 4D ), consistent with previous reports [30,31].Remarkably, although low CO 2 levels did not affect total cellular cholesterol, we observed a substantial decrease in ER free cholesterol levels (Figs 4D and S7D).No significant effect on cholesterol esters content was detected (Figs 4E and S7D).
Cholesterol or 25-hydroxylcholesterol supplementation elevates the ER cholesterol pool that acts through SCAP-Insig binding to anchor SREBP2 in the ER and inhibit its activation (G) A schematic model; in cells under normal physiological CO 2 levels (5%) ER cholesterol levels are unaffected and SREBP2 is retained in the ER membrane.However, under low CO 2 levels (1%), ER cholesterol levels are decreased, inducing SREBP2 activation and subsequent activation of cholesterol biosynthesis related genes [32].To examine the effect of ER cholesterol pools, cells were exposed to low CO 2 in presence of cholesterol or 25-hydroxycholesterol for 4 h and the expression of cholesterogenic genes was analyzed.The transcriptional response of cholesterogenic genes to low CO 2 was abolished in presence of cholesterol or hydroxycholesterol, which further supports involvement of cholesterol levels and most likely ER cholesterol on SREBP2 activation under low CO 2 (Fig 4F).
In summary, our analyses suggest that low CO 2 specifically alters ER cholesterol, and this effect likely triggers the subsequent processing and activation of SREBP2.

Discussion
Alterations in CO 2 levels (hypocapnia or hypercapnia) have been increasingly linked to various pathologies [6,[33][34][35], yet the molecular mechanisms that are implicated in the response to changes in CO 2 remain elusive.
In the present study, we show that a decrease in CO 2 levels activate a distinct gene expression program that differs from mere pH changes (e.g., NaOH treatment).These findings support the presence of a specific mechanism that respond to changes in CO 2 levels.Furthermore, we show that SREBP2 participates in CO 2 signaling to regulate the expression of its target genes, primarily genes of cholesterol biosynthesis.Of note, CO 2 is in equilibrium with HCO 3 -, hence, we cannot conclude whether the observed cellular response to altered CO 2 is due to molecular CO 2 or to changes in bicarbonate levels.Dissecting the effect of CO 2 per se from associated change in bicarbonates is expected to be challenging in view of their rapid equilibrium in physiological systems.This issue can be potentially addressed by using out-of-equilibrium CO 2 /HCO 3 -solutions [36].In addition, manipulation of CO 2 levels in vivo in animal models are extremely challenging due to various homeostatic mechanisms that rapidly act to maintain the equilibrium between CO 2 , bicarbonate, and pH levels.Interestingly, in pancreatic cancer cells SREBP2 induces the expression of cholesterogenic genes in response to extracellular acidic condition [37].Consistently, our results show that alkaline conditions per se (i.e., NaOH treatment), unlike exposure to low carbon dioxide levels, elicit a minor effect on the expression of SREBP2 target genes in non-cancerous cells.Hence, it appears that SREBP2 can be activated in response to various stimuli, namely low carbon dioxide levels and acidic conditions.Since low carbon dioxide levels are associated with alkaline and not acidic conditions, it further supports our conclusion that low CO 2 levels activate SREBP2 through distinct mechanism that is not necessarily pH-related.It is noteworthy that different cell types might respond differently to pH or CO 2 levels.Previous reports showed that carbon dioxide regulates different signaling pathways such as NFκB, Wnt, and TGFβ signaling, as well as circadian rhythms in different cell types [10,[38][39][40].We also show that SREBP2 is activated in response to low CO 2 levels in fibroblast, hepatocytes, and muscles but not in adipocytes.Hence, it is conceivable that the response to CO 2 is conducted through myriad of signaling pathways, some of which are cell-type specific.Our gene expression analysis identified HIF-1α and YAP as potential candidates that participate in the response to CO 2 .Indeed, hypoxia regulates cholesterol metabolism through HIF-1α [41], yet the involvement of HIF-1α in conjunction with SREBP2 in response to CO 2 was hitherto never tested.Likewise, recent evidence points towards functional interaction between YAP and SREBP in regulation of lipid metabolism [42,43]; however, its relevance to CO 2 remains unknown.
through SRE region on their gene promoter.The data underlying the graphs shown in the figure is included in S1 Data.Graphical illustrations were generated with BioRender.com.AUC, area under curve; ER, endoplasmic reticulum; MBCD, methyl-beta-cyclodextrin; SRE, sterol regulatory element.https://doi.org/10.1371/journal.pbio.3002367.g004 Importantly, a role of CO 2 in the control of cholesterol homeostasis has not been previously reported.Interestingly, cellular cholesterol has been shown in vitro to regulate CO 2 permeability in different cell types [44,45].In these studies, each cell type exhibited different CO 2 permeability rate depending on its cholesterol content [45].This raises the intriguing possibility of a mechanism whereby changes in CO 2 levels regulate cholesterol biosynthesis through SREBP2 to regulate cell membrane cholesterol content and control CO 2 permeability in response to environmental changes.
Although, CO 2 is generated as a byproduct of cellular enzymatic reactions, CO 2 is also consumed as a carbon source in the conversion of acetyl-CoA to malonyl-CoA.Acetyl-CoA serves as a key precursor for fatty acid and cholesterol biosynthesis pathways that are major lipid building blocks for cell membranes.SREBPs control the flux of acetyl-CoA into fatty acid and mevalonate synthetic pathways [46].Reduction of extracellular CO 2 might limit the abundance of intracellular CO 2 and would shift the flux of acetyl-CoA towards cholesterol biosynthesis.This shift in substrate supply may serve to support the increased expression of cholesterogenic enzymes by SREBP2.
At the molecular level, our findings suggest that under physiological growth conditions (i.e., 5% CO 2 ), SREBP2 is retained in the ER membrane.Low CO 2 reduces ER cholesterol levels and triggers SREBP2 translocation from ER to Golgi, where SREBP2 is cleaved.The cleaved, transcriptionally active form of SREBP2 (N-SREBP2) enters the nucleus and activates transcription of cholesterol biosynthetic enzymes through SRE region on the gene promoter (Fig 4G).It remains unclear how CO 2 modulates the ER lipids composition, and future studies are expected to address the underlying molecular mechanisms.
In summary, we propose that SREBP2 participates in cellular CO 2 signaling and that SREBP2 regulation of cholesterol levels can be modulated by changes in CO 2 levels.

Ethics statement
All animal experiments and procedures were conducted in conformity with and approval of the Weizmann Institute Animal Care and Use Committee (IACUC) guidelines, working within the anti-cruelty law (experiments on animals) of 1994 as stated by the Ministry of Health of the Israeli Parliament.Experiments were done in accordance with these specific applications: 05730621-1 and 01700223-2.

Sterol depletion
Sterol depletion from the cells was performed by MBCD (Sigma) treatment.Cells were seeded in media B at density 300,000 cells/3 cm dish.On day 4, cells were washed with PBS twice to remove residual serum and cells were incubated in DMEM (Gibco, 21063-029) without serum, supplemented with either 2 mM MBCD (100 mM stock in water) or vehicle control for 2 h at 37˚C and 5% CO 2 .Next, cells were washed twice with PBS and cultured in media B without serum.

Bioluminescence recordings
Unless indicated otherwise, for bioluminescence recordings cells were seeded in media B at density of 300,000 cells per 3-cm culture dish.Next day, cells were transiently transfected with one of the plasmids containing pSynSRE-T-Luc (750 ng), pSynSRE-Mut-T-Luc (750 ng), or pcDNA3-Luc (500 ng) as detailed above.After 48 h from transfection, the culture medium was replaced with media B supplemented with 100 nM D-Luciferin (Promega, USA) and bioluminescence was recorded continuously with LumiCycle32 recorder (Actimetrics, USA) in a 37˚C, 5% CO 2 incubator.After 24 h (once the luciferase signal reached a stationary phase), CO 2 levels were shifted to 1%, 10%, or kept at 5% CO 2 as control.Bioluminescence data were extracted using the LumiCycle Analysis software (Actimetrics, USA).The relative bioluminescence was calculated by normalizing the raw counts to the 10 h pretreatment average value.

Protein extraction, gel electrophoresis, and immunoblotting
Whole cell lysate was prepared as previously described [9].For nuclear and cytoplasmic fraction, cell pellets were resuspended in lysis buffer (HEPES 10 mM (pH 7.5), 10 mM KCl, 0.1 mM EDTA, 0.5% Noniodate 40, 1 mM DTT, PMSF 0.5 mM) supplemented with protease inhibitor cocktail (Sigma) and allowed to swell on ice for 15 to 20 min with intermittent mixing.Tubes were vortexed (10 s) to disrupt the cell membrane and then centrifuged at 12,000 g at 4˚C for 30 s.The supernatant was stored at −80˚C till further use as cytoplasmic extract.The pelleted nuclei were washed twice with 1 ml lysis buffer and was resuspended in nuclear extraction buffer (20 mM HEPES (pH 7.5), 400 mM NaCl, 1 mM EDTA, and 1 mM PMSF) with protease inhibitor cocktail and then incubated on ice for 30 min.Nuclear extracts were collected by centrifugation at 12,000 g for 15 min at 4˚C.The protein concentration of the cytoplasmic and nuclear extract was quantified using BCA protein assay kit (Thermo Scientific, USA).Finally, samples were heated at 95˚C for 5 min in Laemmli sample buffer and analyzed by SDS-PAGE and immunoblot.SREBP2 antibody that was used in our study (Anti-SREBP2, Clone 22D5, MABS1988, Lot # 3272232, Merck) recognizes the N-terminal region of murine SRE-binding protein 2. Details of the antibodies used are listed under S8 Table.

ER membrane isolation
Cells were seeded in 15-cm culture dish at density of 2 × 10 6 cells per dish.On day 4, cells were subjected to different treatments as indicated.Next, cells were washed with cold PBS, scrapped in 2 ml PBS, and collected in 15 ml tube.The suspension was centrifuged at 500 g for 10 min to obtain cell pellet, snap frozen in liquid nitrogen, and stored at −80˚C until further use.Isolation of ER membranes was performed with minor modification as previously described [30].Cell pellet were homogenized with glass dounce (15 to 25 rounds) in cold lysis buffer (50 mM Tris-HCl (pH 7.5), 150 mM NaCl, 15% sucrose) containing protease inhibitor cocktails.A small aliquot of homogenate was stored as whole cell lysate (fraction A) (S7B Fig) .The lysates were centrifuged at 3,000 g for 10 min to yield nuclear pellet and supernatant (fraction B).
Nuclear pellets were lysed with nuclear extraction buffer (20 mM HEPES (pH 7.5), 400 mM NaCl, 1 mM EDTA, and 1 mM DTT) and stored as fraction C. Further, the supernatant was diluted to 3 ml with lysis buffer and loaded on discontinuous sucrose gradient which was set in SW41 tube (Beckman) by overlaying the following sucrose solutions all in the above lysis buffer: 2 ml 45%, 4 ml 30%, 3 ml of the diluted supernatant in 15% sucrose, and 1 ml 7.5%.The tubes were ultra-centrifuged in SW41Ti rotor (Beckman) at 100,000 g for 60 min and allowed to slow down without application of a break.The 2 bands of membranes were clearly visible, upper light membrane fraction (Interphase between 15% and 7.5%) were collected and marked as fraction D and heavy membrane fraction (interphase between 45% and 30% sucrose) were collected in another tube as fraction E. The collected fractions at each stage were analyzed by immunoblot with relevant organelle markers as indicated (S7C Fig) .Further, purification of heavy membrane fraction (fraction E) was performed with OptiPrep-Density gradient medium (Sigma).Fraction E from the above sucrose gradient was loaded at the bottom of SW41 tube and subsequently, overlaid with dilutions of OptiPrep-Density gradient medium.Discontinuous OptiPrep gradient was generated by underlying in sequence form bottom to top-1 ml fraction E, 2.5 ml each of 25%, 23%, 21%, 19% OptiPrep media diluted in ice cold tris-buffer (50 mM Tris-HCl (pH 7.5), 150 mM NaCl) and equilibrate for 2 h at 4˚C.After incubation, tubes were ultra-centrifuged at a speed of 110,000 g for 120 min.After centrifugation, OptiPrep fractions were collected from top to bottom of the tube (approximately 900 μl each fraction) and fractions were run on the SDS-PAGE with marker protein for ER membrane and the fraction showing no organelle contamination (fraction no 5) was used for lipidomic analysis.

ER-lipidomic analysis
Cholesterol and cholesterol ester were identified and quantified using multi-dimensional mass spectrometry-based shotgun lipidomic analysis [55].In brief, each 300 μl ER suspension sample was accurately transferred to a disposable glass culture test tube.A pre-mixture of internal standards (IS) was added prior to conducting lipid extraction for quantification of the targeted lipid species based on the protein content of individual ER suspension.Lipid extraction was performed using a modified Bligh and Dyer procedure [56], and each lipid extract was reconstituted in chloroform:methanol (1:1, v:v) at a volume of 100 μl/300 μl ER suspension samples.
For shotgun lipidomics, lipid extract was further diluted to a final concentration of approximately 500 fmol total lipids per μl.Mass spectrometric analysis was performed on a triple quadrupole mass spectrometer (TSQ Altis, Thermo Fisher Scientific, USA) and a Q Exactive mass spectrometer (Thermo Scientific, USA), both of which were equipped with an automated nanospray device (TriVersa NanoMate, Advion Bioscience Ltd., Ithaca, NY) as described [57].Identification and quantification of cholesterol and cholesterol ester were performed using an automated software program [58,59].Data processing (e.g., ion peak selection, baseline correction, data transfer, peak intensity comparison, and quantitation) was performed as described [59].The results were normalized to volume of ER suspension (pmol/100 μl ER suspension).

Total cholesterol quantification
Lipid extraction and cholesterol quantification were performed using a Total Cholesterol Assay Kit Fluorometrically (Cell Biolabs, STA-390) according to the manufacturer's protocol.

Statistics
All the statistical analyses were performed using Excel, Python, and GraphPad prism (Version 9.1.0.221).Specific information on sample sizes, statistical significance, and variance measures is provided in the relevant figure legends.Significantly expressed genes are defined based on the difference in expression between each of the time points (i.e., T0, T2, T4) per condition, based on the following criteria: P adj.< = 0.05, |log2FC| > = 1, baseMean > = 5.Normalized data of significant genes in each of the conditions was clustered using Python's scikit-learn KMeans function.PCA analysis was performed using R's prcomp function (scale = TRUE) of stats package.

Fig 1 .
Fig 1. AU : AbbreviationlistshavebeencompiledforthoseusedinFigs1to4:Pleaseverifythatallentriesarecorrect: The transcriptional response to low CO 2 levels compared to NaOH treatment.(A) A schematic depiction of the experimental design.NIH3T3 cells were cultured in media B (as detailed in method) at 37˚C with 5% CO 2 and 20% oxygen.On day 4, cells were either untreated or treated with 18 mM NaOH or shifted to a special incubator (Coy Labs, USA) with 1% CO 2 and 20% oxygen.Cells and media were collected 2 and 4 h post treatment.RNA was extracted and analyzed by RNA sequencing (n = 4 for each time point per condition).(B) pH measurements of the growth media (mean ± SE, n = 3 biological replicates for each time point per condition, ***P < 0.001, nonsignificant (ns), two-sided Student's t test).(C) PCA.(D) Heatmap representation of genes that were significantly altered (see Methods) between time points within conditions.Data are presented as row z-scores of the expression per condition.See also S1 Fig and S1 Table.The data underlying the graphs shown in the figure is included in S1 Data.Graphical illustrations were generated with BioRender.com.PCA, principal component analysis.https://doi.org/10.1371/journal.pbio.3002367.g001

Fig 2 .
Fig 2. Low CO 2 levels specifically induce the expression of genes related to cholesterol biosynthesis.(A) K-means unsupervised clustering of significant genes for each of the conditions.Black line represents the mean z-score (for gene lists see S2 Table).(B) Pathway enrichment analysis was performed using the IPA tool for the genes included in each of the clusters for 1% CO 2 or 18 mM NaOH treatments.Presented are the top 3 enriched pathways in each cluster based on P value (for full list of pathways, see S3 Table).(C) Schematic illustration of the cholesterol biosynthesis pathway alongside genes that were significantly affected by the treatments.Color indicates on the condition in which the genes are affected.(D) Heatmap representation of cholesterogenic genes that were significantly affected by any of the conditions presented.Data are presented as row z-score of the average expression per condition (n = 4 biological replicates).(E) Quantitative PCR analysis of cholesterogenic gene expression levels from NIH3T3 cells treated with 1% CO 2 or 18 mM NaOH (mean ± SE, n = 5 biological replicates per time point per condition, *P < 0.05, **P < 0.01,