Joint Effects of Heat Stress and PM2.5 Exposure on Glucose Metabolism and Hepatic Insulin Signaling

Background: Heatwave events are occurring more frequently, accompanied by a signicant increase in the ambient concentration of ne particulate matter (PM 2.5 ). Epidemiological studies have suggested that heat stress or PM 2.5 exposure would impair glucose homeostasis and insulin sensitivity, but the combined effect and the exact mechanisms are not well understood. Methods: C57BL/6 mice were randomly divided into ltered air (FA), ne particulate matter (PM) group, ltered air combined with heat stress (FH) group, and ne particulate matter combined with heat stress (PH) group for a 4-week PM 2.5 exposure, followed by a 2-week heat stress exposure, via a whole-body exposure system. Systemic glucose homeostasis, insulin sensitivity, and circulating inammatory cytokines were examined. HSP72 expression and insulin signaling in the liver were measured. Results: Glucose tolerance and insulin sensitivity were impaired in response to heat stress, accompanied by lessened hepatic GLUT2 expression and inhibited insulin signaling pathway. No synergistic effects of heat stress and PM 2.5 exposure on glucose homeostasis were observed, while heat-upregulated HSP72 expression was attenuated with accumulated TNF-α induced by further PM 2.5 exposure. Conclusions: Heat stress combined with PM 2.5 exposure induced TNF-α, which could inhibit heat-elevated hepatic HSP72 expression. Elevated circulating TNF-α impaired hepatic insulin signaling and GLUT2 expression. Then, glucose homeostasis was perturbed, and insulin action was impaired.


Background
According to the European Copernicus Climate Change Service, the average temperature in Europe in 2019 was about 1.2°C higher than the average temperature in the past 40 years, which was the hottest year on record [1] . China's climate warming in the past 60 years has been particularly pronounced, with an average increase of about 0.23°C every 10 years, almost twice of the global rate [2]. There is increasing evidence that both mortality and morbidity increased as temperature rose, especially in elderly people with pre-existing conditions of cardiovascular, cerebrovascular, and respiratory diseases [3,4]. Based on an estimate of 831 million urban residents in China, every 0.5°C increase in surface temperature from the current level will result in more than 27,900 heat-related deaths each year [5]. These extreme weather episodes not only impact human health, but also in uence a variety of other global changes (e.g., population growth, urbanization, land use changes, and natural resource depletion) and potentially augment their impacts [6].
Although some critical issues of climate change remain unresolved, it is generally believed that this is mainly due to the increase in greenhouse gas emissions [7]. There is an inherent link between climate change and air pollution because greenhouse gases and air pollutants come from similar or even the same sources, such as fossil fuel combustion, which is the main source of both greenhouse gases (CH 4 , CO 2 ) and criteria air pollutants (particulate matter [PM], CO, SO X , NO X ) [8]. During a heatwave, elevated temperature is characteristically accompanied by corresponding increases in the concentrations of air pollutants, such as PM and O 3 [9]. To date, no animal study has examined the simultaneous effects of temperature and PM exposure. Although another air pollutant, O 3 , has also been shown to interact with heat stress, most of the increased risk is on the pulmonary system [10]. Therefore, increases in mortality or morbidity may feasibly be attributed to an interaction between heat and air pollutant factors. Recent studies have shown that both PM pollution and temperature change are associated with adverse health outcomes [11,12], and PM leads to the largest effect on mortality on hot days [13]. In China, the risk of total mortality, cardiovascular mortality, and respiratory mortality per 10 µg/m 3 increment in PM 10 increased more than 2-fold on high temperature days [14]. Importantly, one study suggested that the threshold of different pollutants, above which these effects become measurable, appears to be lowered during extremely hot days [15]. Therefore, it is likely that the true magnitude of the interactions between high temperature and PM pollution is underestimated. We hypothesized that heat stress combined with PM 2.5 exposure would impair glucose homeostasis and insulin sensitivity. The combined effects of PM and high temperature need to be considered and investigated to establish adequate public health policies and actions to face anticipated future changes in air quality and climate. It is widely accepted that respiratory system response during PM exposure involves in ammatory pathways [16]. Severe heat stress also leads to features of proin ammatory cytokines (e.g., IL-1β, tumor necrosis factor (TNF)-α) and hepatic oxidative stress (e.g., lipid peroxidation, glutathione oxidation) [17]. During exercises in a hot and humid environment, tissue damage was found particularly in liver that was related to in ammation, oxidative stress, and apoptosis [18]. Liver injury is a common complication of heat stress, which has been proved to be mediated by IL-1β [19]. In turn, excessive in ammatory cytokines could inhibit the production of heat shock proteins (HSP)72 [20]. In addition to cytokines, up-regulation of HSP is one of the primary cellular defense mechanisms against heat stress [21]. Studies showed that HSP72 and HSP90 were signi cantly higher in heat intolerant mice following heat exposure compared to heat tolerant mice [22]. Interestingly, these ndings suggest a substantial overlap in the mechanisms of metabolism and in ammation, which underlie progressive insulin resistance and peripheral in ammation [23]. Our previous studies revealed that PM 2.5 exposure signi cantly reduced insulin sensitivity in the liver [24,25] and disrupted the rhythms of key enzymes involved in hepatic lipid metabolism [26]. These evidences suggest that liver is an important target organ for PM 2.5 exposure.
Considering that heat stress and PM 2.5 exposure may work on the same/close targets, we therefore were particularly interested in the interactive effects of heat stress and "real world" ambient PM 2.5 exposure on liver metabolism. In this study, we systematically investigated this issue in C57BL/6 mice subjected to heat stress combined with PM 2.5 exposure.
Heat stress combined with ambient whole-body PM 2.5 inhalational exposure protocol All mice were randomly divided into 4 groups ( ltered air [FA] group, ne particulate matter [PM] group, ltered air combined with heat stress [FH] group, and ne particulate matter combined with heat stress [PH] group). Two weeks before ambient whole-body PM 2.5 inhalational protocol, 5 mice in each group were randomly selected, and telemetry sensors (DSI, ETA-F20) were implanted into abdominal cavity by surgery under anesthesia [27]. One week after the surgery, all mice were subjected to either ltered air (FA) or ambient PM 2.5 for 6 h/d, 5 d/week from April 23, 2019 to June 4, 2019, for a total duration of 6 weeks, in a set of exposure system of "Zhejiang Whole-body Exposure System 1" located on the campus of Zhejiang Chinese Medical University in Hangzhou [25]. The system consists of four temperaturecontrolled chambers. The ambient air is connected to PM chambers, in which the particles larger than 2.5 μm are removed by cyclone separator and evenly distributed in the PM chambers. The FA chambers are equipped with HEPA lter in the inlet valve to remove all the PM 2.5 in the air stream. All mice were housed in separated cages with 3-4 mice/cage in each chamber. The temperature of the heat stress chambers was raised to 37 ± 0.2°C gradually within an hour period [28,29]. A core temperature of 37°C is below the temperature (42.4°C) that has been considered as a mild heat stress [28,29] and has been shown to produce minimal pathological changes in the lungs and brain [30]. Temperature and humidity in the inhalation chambers were monitored continuously (Sup. 1 A and B). Mice are nocturnal and will be exposed to PM 2.5 and heat stress during their active period. Because air pollution levels are generally lower at night as compared to the daytime period, we need to reverse the animals' circadian cycle. Right after the 4 th week of the ambient whole-body PM 2.5 inhalation, temperature in the chamber of FH and PH was set to 37°C from 10:00 to 13:00 (3 hours) every other day (for a total of 4 days). During the 6 th week of the ambient whole-body PM 2.5 inhalation, temperature in the chamber of groups FH and PH was set to 37°C from 22:00 to 1:00 (3 hours) every other day (for a total of 4 days) (Fig. 1A). Monitoring of the FA and PM exposure environments within the two chambers was continuously recorded with an Aerosol Monitoring Meter (pDR-1500, Thermal Scienti c, China), which was zero-adjusted before use. The data of environmental ambient aerosol concentration were collected from an adjacent local EPA monitoring station, which was located at the intersection of Jiangnan Ave. and Jiangling Rd., 10 km away from the study site. The temperatures of experimental mice and 4 chambers were acquired by DSI telemetry system, which were collected and recorded every 5 minutes throughout the experiment.
Sampling and analysis of PM 2.5 in the exposure chambers To calculate PM 2.5 mass concentrations in the exposure chambers, PM 2.5 samples in the exposure chambers were collected on Te on lter membranes (Te on, 37 mm, 2 μm pore; GE healthcare, Amersham Place, UK). Te on lter membranes from PM chamber were collected daily, while the lter membranes from FA chamber were collected weekly. A microbalance (Excellence Plus XP, Mettler Toledo, Switzerland) was used to weigh the membrane before and after sampling in a climate-controlled weighing chamber. Weight gains were used to calculate the exposed concentrations in the chambers.

Measurements of blood glucose homeostasis and insulin sensitivity
Glucose tolerance test (GTT) was conducted right after the heat stress combined with PM 2.5 exposure.
Mice were fasted overnight (16 hours) and intraperitoneally injected with dextrose (2 mg/g body weight). Blood samples were collected from vena caudalis and blood glucose was measured by a FreeStyle blood glucose meter (Abbott Diabetes Care Inc., Alameda, CA, USA) at baseline, and 30, 60, 90, and 120 min after the dextrose injection. Three days after GTT, insulin sensitivity was measured by insulin tolerance test (ITT) via human regular insulin (0.5 U/kg body weight) injected intraperitoneally after 4.5 h fasting. At baseline and 30, 60, 90 and 120 minutes after insulin injection, blood glucose was measured by the same type of blood glucose meter in the same way as GTT.

Measurement of blood in ammatory biomarkers
Circulating cytokine levels were measured by CBA kit. Plasma was incubated with beads speci c for TNFα and IL-6 according to manufacturer's instructions. The total amount of cytokines was then determined using a BD Accuri C6 Plus instrument and analyzed by BD FCAP Array Software v3.0 (BD Biosciences, San Jose, CA, USA).

Immuno uorescence analysis
After the exposures, liver segments were preserved via intracardiac perfusion with 30-40 ml of normal saline containing 0.02% heparin, followed by 30-40 ml of 4% paraformaldehyde in 0.1 M potassium phosphate buffer (pH = 7.4) when mouse was under anesthesia. Fixed liver tissue was embedded in para n and sectioned into 5-μm-thick. After depara nization in xylene for 10 min followed by 100% ethanol, slices were washed three times in phosphate-buffered saline (PBS). Nonspeci c binding was blocked with 10% normal goat serum at room temperature for 30 min. An overnight incubation with rabbit anti-GLUT2 antibody (1:200) at 4°C was followed by 2 h of incubation with Alexa Fluor 488-conjugated goat anti-rabbit-IgG secondary antibody (1:200) at room temperature. The slices were washed and incubated with antifade mounting medium with DAPI as a nuclear stain. Finally, the slides were observed under a Leica uorescence microscope DM4 B (Leica Microsystems, Wetzlar, Germany).

Total protein extraction and Western blotting analysis
Liver tissue was homogenized in 250 μL of cold RIPA lysis buffer with protease inhibitors and phosphatase inhibitors to obtain total protein for 30 min and then centrifuged at 12,000 ×g for 10 min at 4 °C. Supernatants assayed for protein using a Pierce BCA Protein Assay Kit and quanti ed to same concentration of protein (3 μg/μL). Equal amounts of proteins were loaded onto 8%-12% SDSpolyacrylamide gel, which was transferred to polyvinylidene uoride membranes after electrophoresis. After blocking the nonspeci c sites, the membranes were incubated at 4°C overnight. The following primary antibodies were used in this experiment: anti-HSP72, anti-IRS-1, anti-p-IRS-1 (Ser307), anti-Akt, anti-p-Akt (Ser473), anti-GLUT2 and anti-GAPDH. The membranes were probed with a goat anti-rabbit HRP-conjugated secondary antibody at room temperature for an hour. ChemiDoc Touch ImagingSystem (Bio-Rad Laboratories, Inc. Hercules, CA, USA) was used to detect protein bands. Band densities were normalized to GAPDH in each sample.

Measurement of gene expression by real-time RT-PCR
RT-PCR was performed using RNA extracted from mouse liver tissue. Total RNA was isolated with RNAiso Plus according to the manufacturer's protocol. It was then converted to cDNA using High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Carlsbad, California, USA) and analyzed by RT-PCR on QuantStudio 6 Flex system using SYBR Green (ThermoFisher, Waltham, MA, USA). The expression level for each gene was calculated using the ΔCt method relative to β-actin. The sequences of all primers used are listed in Table 1.

Statistical nalysis
All data were analyzed with Prism v.8.3.0 (GraphPad Inc., USA). Results are expressed as mean ± standard error of mean (SEM) unless noti ed otherwise. P values of less than 0.05 were considered statistically signi cant. Statistically signi cant values were compared using a two-way ANOVA followed by bonferroni's multiple comparisons test.  Table 2.

Results
Metabolic characteristics of heat stress combined with PM 2.5 exposure in mice Six weeks of PM 2.5 or heat stress exposure resulted in no differences in the body weights or average food intake among the four groups (Fig. 2, A and B). Right after the heat stress combined with PM 2.5 exposure, GTT was performed. Three days after GTT, ITT was measured. Compared with the FA group, heat stress signi cantly elevated blood glucose levels in FH mice at 30-and 60-min during GTT (Fig. 2C). Compared with the PM group, heat stress combined with PM 2.5 exposure demonstrated signi cant elevation of blood glucose levels in PH mice at 30-min during GTT (Fig. 2C). Compared with the FA group, heat stress combined with PM 2.5 exposure signi cantly elevated blood glucose levels in FH mice at 30-and 60-min during GTT (Fig. 2C). Compared with the FA group, heat stress impaired insulin sensitivity in FH mice at 90-min during ITT (Fig. 2E). Compared with the PM group, heat stress combined with PM 2.5 exposure also signi cantly impaired insulin sensitivity in PH mice at 60-, 90-, and 120-min during ITT (Fig. 2E). Compared with the FA group, heat stress combined with PM 2.5 impaired insulin sensitivity in FH mice at 90-and 120-min during ITT (Fig. 2E). However, compared with FH group, heat stress combined with PM 2.5 exposure induced no further impairment in GTT or ITT (Fig. 2, C and E). Area under the curves (AUC) analysis of GTT and ITT con rmed these ndings (Fig. 2, D and F). These results indicated that heat stress or heat stress combined with PM 2.5 exposure, but not PM 2.5 exposure alone, impaired glucose homeostasis and insulin sensitivity.
Heat stress combined with PM 2.5 exposure elevated circulating TNF-α and decreased expression of HSP72 in the liver To comprehensively understand the effects of heat stress and PM 2.5 on organ in ammation, hepatic proin ammatory cytokine production was examined with RT-PCR and Cytometric Bead Array kit. After 6 weeks of exposure, although there was no signi cant difference in the expression pattern of IL-1β, IL-6, or TNF-α among the four groups, the PM and FH mice exhibited a light increase in mRNA and circulating levels of IL-6 and TNF-α (Fig. 3A). Interestingly, compared with PM mice or FA mice, the circulating level of TNF-α was signi cantly higher in PH mice. Although no signi cant increase was shown, the circulating levels of TNF-α was about 63.7% higher in PH mice than that in FH mice (Fig. 3B). As shown in Fig 3C, heat stress alone induced signi cant increase in HSF72 in the liver, while signi cant downregulation in HSP72 was observed between FH and PH mice. Furthermore, heat stress combined with PM 2.5 exposure resulted in no difference in the amount of HSP72 protein compared with PM 2.5 exposure (Fig. 3C). We further investigated the HSF1/HSPA1A axis (critical for the transcription of HSP72) in the liver by RT-PCR.
The levels of HSF1 mRNA and Hspa1a mRNA showed no differences among groups (Fig. 3D). Marked histologic structural damages were observed in FH and PH mice, while normal light microscope appearance was better preserved in FA and PM mice (Sup. 2).
Heat stress combined with PM 2.5 exposure attenuated hepatic GLUT2 protein expression and insulin signaling.
As shown in Fig. 4A, hepatic GLUT2 expression was examined with immuno uorescence assay.
Quantitative analysis of MFI (mean uorescence intensity) indicated that the expression of hepatic GLUT2 was attenuated by PM 2.5 , heat stress, or heat stress combined with PM 2.5 exposure (Fig. 4B).
Similar inhibition on GLUT2 expression was observed at protein level (Fig. 4C). To further con rm the effects of heat stress or PM 2.5 exposure, Slc2a2 (transcription factor for GLUT2) was examined by RT-PCR. Although no signi cant alteration in response to PM 2.5 exposure, PM mice showed a trend toward downregulation of Slc2a2 in the liver compared to FA. Furthermore, both FH and PH mice exhibited signi cant decrease in its mRNA expression (Fig. 4D). To further con rm the inhibitory effect of heat stress and/or PM 2.5 exposure on hepatic insulin signaling, phosphorylated levels of IRS-1 at Ser307 were further examined. Compared to FA group, heat stress exposure signi cantly elevated the level of p-IRS-1 Ser307 . It was the same case when heat stress combined with PM 2.5 exposure compared to the PM group (P = 0.05) (Fig. 4E). Akt is the activated downstream of the IRS family, which is necessary for GLUT2 function [31,32]. To further con rm the effects of heat stress and/or PM 2.5 exposure, phosphorylated levels of Akt at Ser473 were examined. Compared to FA group, p-Akt Ser473 was signi cantly inhibited in response to heat stress or heat stress combined with PM 2.5 exposure. However, there was no statistical difference at the phosphorylated levels of Akt between FA and PM mice, FH and PH mice (Fig. 4F).

Discussion
The adverse health effects of heat stress combined with PM 2.5 exposure, which is demonstrated with epidemiological studies, has not been investigated in animal studies. In the present study, we used C57BL/6 mice to investigate the impact of heat stress combined with PM 2.5 exposure on insulin signaling in the liver. Our major ndings include: a) Impaired glucose tolerance and insulin sensitivity in response to heat stress; b) Inhibited hepatic GLUT2 expression mediated by insulin signaling due to heat stress; c) No synergistic effects of heat stress and PM 2.5 exposure on glucose homeostasis.
HSP72 is a major stress-inducible chaperone protein, which folds newly translated and misfolded proteins correctly and stabilizes or degrades mutant proteins [33]. Overexpression of HSP72 protects cells from stresses, including heat stress and oxidative stress [34]. Recent clinical and experimental evidences suggest a complex interaction between heat stress and systemic in ammatory response syndrome, of which cytokines have been implicated as adverse mediators [35]. Consistent with it, the protective expression of HSP72 in the liver and the circulating level of TNF-α were upregulated by heat stress exposure in the present study. However, despite of signi cantly increased circulating TNF-α, hepatic expression of HSP72 decreased in presence of heat stress combined with PM 2.5 exposure. Studies have shown that excessive expression of in ammatory factor TNF-α could inhibit the HSF1/HSPA1A axis, which is critical for the transcription of HSP72 [36,37]. Thus, protective upregulation of HSP72 induced by heat could be offset by PM 2.5 exposure plus heat stress-enhanced TNF-α secretion. However, the exact mechanism and the interaction between TNF-α and HSP72 under heatwaves, merits further investigation.
Previous studies have clearly shown that insulin sensitivity could be improved by HSP72 induction via heat treatment (42°C for 30 min), pharmacologic intervention, and transgenic overexpression [23,38,39].
On the other hand, mice with HSP72 deletion were prone to be obese and glucose intolerant [40], further supporting the role of HSP72 in metabolic regulation. However, something contrary between HSP72 and glucose metabolic disorder was observed. In addition to systemic insulin sensitivity, the molecules in hepatic insulin signaling pathway were also explored. GLUT2 expression is critical for the control of glucose-sensitive genes, and inactivation of GLUT2 in the liver impaired glucose-stimulated insulin secretion [41]. GLUT2 expression was signi cantly inhibited by heat stress or heat stress combined with PM 2.5 exposure, which was regulated by its transcription factor Slc2a2. The alteration of GLUT2 levels were mostly in line with the changes of p-IRS-1 Ser307 and p-Akt Ser473 , downstream of classic insulin signaling pathway [42]. Thus, heat stress itself may impair hepatic insulin signaling and GLUT2 expression, which interfered glucose transport to hepatocytes and resulted in the elevation of blood glucose in GTT and ITT. We noticed that the changes of hepatic GLUT2 were inconsistent with the results of GTT and ITT measurements, particularly for PM mice. The ITT and GTT are mainly affected by three key insulin target tissues: skeletal muscle, liver, and white adipose tissue [43]. Our previous study found that exposure to PM 2.5 signi cantly reduced the insulin sensitivity of the three key insulin target tissues, but the sensitivity of each tissue was quite different [44]. Excessive in ammatory cytokines have been shown to impair insulin sensitivity [45,46] and inhibit the production of HSP72 [20] as well. The failure of HSP72 to protect insulin sensitivity may be due to the offset of insulin sensitivity by heat-or heat plus PM 2.5 exposure-induced TNF-α secretion and direct inhibition of in ammation on HSP72 expression itself.
Different from our previous studies, we found no impairment in glucose homeostasis and insulin sensitivity in PM 2.5 exposed mice [24,26]. We suppose that it may be due to insu cient PM 2.5 exposure duration (6-week duration) [44,47]. Even though, inhibited GLUT2 expression and a trend toward decrease in p-Akt Ser473 was shown, indicating the presence of PM 2.5 exposure induced adverse effects. Our component analysis showed that metals and inorganic ions were the most abundant components in PM 2.5 samples. Researchers also pointed out that PM 2.5 -bound metal could affect not only the respiratory system but also metabolism (hepatic urea synthesis and fat deposits) [48,49]. It makes us to speculate that the toxic effect of PM 2.5 under current situation (concentration and components) may be not so strong as that in previous studies. Although epidemiological survey demonstrated that PM concentration increased simultaneously during a heatwave [9], no further toxic effects of PM 2.5 exposure upon heat stimulation were observed. The trend toward increase in circulation TNF-α and attenuation of hepatic HSP72 expression indicated the presence of, albeit little, interactive action in response to heat stress combined with PM 2.5 exposure. Thus, the exact adverse effects of multiple environmental factor exposure await further investigations with more sophisticated designs.

Conclusions
In summary, our data showed that heat stress combined with PM 2.5 exposure induced TNF-α, which could inhibit heat-elevated hepatic HSP72 expression. Elevated circulating TNF-α impaired hepatic insulin signaling and GLUT2 expression. Then, glucose homeostasis was perturbed, and insulin action was impaired. The above results provide a new perspective for revealing the etiology of liver metabolism abnormality from the viewpoint of insulin resistance induced by heat stress combined with PM 2.5 .   Metabolic characteristics of mice in response to heat stress combined with PM2.5 exposure A-B, Weekly body weight and average food intake during the exposure period. C-D, GTT and GTT analysis with AUC. E-F ITT and ITT analysis with AUC. In C and E, * P < 0.05 when FH compared to FA group; # P < 0.05 when PH compared to PM group; § P < 0.05 when PH compared to FA group. In D and F, * P < 0.05; ** P < 0.01. n = 12-14. Circulating cytokines and hepatic expression of HSP72 in response to heat stress combined with PM2.5 exposure. A, mRNA levels of IL-1β, IL-6, and TNF-α in the liver. B, Circulating levels of IL-6 and TNF-α. C, mRNA levels of HSF1 and Hspa1a in the liver. D, Representative bands and analyzed protein levels of HSP72 in the liver at the end of the exposure. * P < 0.05; ** P < 0.01. n = 5-6. Hepatic GLUT2 expression and insulin signaling in response to heat stress combined with PM2.5 exposure. A, Immuno uorescence staining for GLUT2 (green and arrows) and DAPI (blue) in the liver after 6 weeks of exposure in FA, PM, FH, and PH groups. Images were obtained under the same conditions at a magni cation of 200× across treatment groups. Scale bar, 50 μm. B, Quantitative analysis of the MFI of GLUT2 protein by immuno uorescence assay. C, Representative bands and analyzed protein levels of