Hepatocyte-specific glucose-6-phosphatase deficiency disturbs platelet aggregation and decreases blood monocytes upon fasting-induced hypoglycemia

Objective Glycogen storage disease type 1a (GSD Ia) is a rare inherited metabolic disorder caused by mutations in the glucose-6-phosphatase (G6PC1) gene. When untreated, GSD Ia leads to severe fasting-induced hypoglycemia. Although current intensive dietary management aims to prevent hypoglycemia, patients still experience hypoglycemic events. Poor glycemic control in GSD Ia is associated with hypertriglyceridemia, hepatocellular adenoma and carcinoma, and also with an increased bleeding tendency of unknown origin. Methods To evaluate the effect of glycemic control on leukocyte levels and coagulation in GSD Ia, we employed hepatocyte-specific G6pc1 deficient (L-G6pc−/−) mice under fed or fasted conditions, to match good or poor glycemic control in GSD Ia, respectively. Results We found that fasting-induced hypoglycemia in L-G6pc−/− mice decreased blood leukocytes, specifically proinflammatory Ly6Chi monocytes, compared to controls. Refeeding reversed this decrease. The decrease in Ly6Chi monocytes was accompanied by an increase in plasma corticosterone levels and was prevented by the glucocorticoid receptor antagonist mifepristone. Further, fasting-induced hypoglycemia in L-G6pc−/− mice prolonged bleeding time in the tail vein bleeding assay, with reversal by refeeding. This could not be explained by changes in coagulation factors V, VII, or VIII, or von Willebrand factor. While the prothrombin and activated partial thromboplastin time as well as total platelet counts were not affected by fasting-induced hypoglycemia in L-G6pc−/− mice, ADP-induced platelet aggregation was disturbed. Conclusions These studies reveal a relationship between fasting-induced hypoglycemia, decreased blood monocytes, and disturbed platelet aggregation in L-G6pc−/− mice. While disturbed platelet aggregation likely accounts for the bleeding phenotype in GSD Ia, elevated plasma corticosterone decreases the levels of proinflammatory monocytes. These studies highlight the necessity of maintaining good glycemic control in GSD Ia.

hepatomegaly and nonalcoholic fatty liver disease (NAFLD) as a consequence of excessive glycogen and lipid accumulation [1,11,12], and hepatocellular adenoma, which may progress to carcinoma [6,7,13,14]. Despite prominent hyperlipidemia, GSD Ia patients exhibit a 10% reduction in carotid intima-media thickness [15], reflecting decreased atherosclerotic lesions in the carotid arteries. Under conditions of poor glycemic control, GSD Ia patients show increased bleeding tendency, which complicates surgical procedures [16e18]. This can be prevented by continuous gastric drip feeding to maintain euglycemia for 24 h before surgery [16e18]. Several pathways have been proposed to decrease atherosclerosis and enhance bleeding in GSD Ia [16,18e20]. The exact mechanisms remain elusive, likely due to the relatively small number of GSD Ia patients included in these studies and the high level of clinical heterogeneity [21]. Hypoglycemia caused by poor glycemic control could contribute to the low incidence of atherosclerosis in GSD Ia. Hypoglycemia may decrease blood monocytes, as suggested by others [23]. Low monocyte numbers may decrease atherosclerosis and monocyte-platelet aggregates, which induce coagulation [24,25]. Hypoglycemia may also affect platelet aggregation directly [18,26]. We investigated the effects of glycemic control in GSD Ia on leukocyte levels along with coagulation and the mechanisms involved. We used a mouse model of GSD Ia with hepatocyte-specific G6pc1 deficiency (L-G6pc À/À mice) that were studied under fed or fasted conditions, to match good or poor glycemic control in GSD Ia, respectively [10].

Animals
Female B6.G6pc1 lox/lox and B6.G6pc1 lox/lox .SA CreERT2 mice were housed in a light (lights on at 7:00 AM. i.e. Zeitgeber time (ZT) 0, lights off at 7:00 PM, i.e. ZT12) and temperature (21 C) -controlled facility. Mice had free access to water and a standard chow diet (67% carbohydrates, consisting of the low glycemic index carbohydrate sources such as wheat (glycemic index 54), barley (glycemic index 28), and corn gluten, along with 23% protein and 10% fat (RMH-B, AB diets, Woerden, The Netherlands)). At 8e12 weeks of age, mice received intraperitoneal tamoxifen injections (T5648; SigmaeAldrich, St. Louis, MO, USA) (1 mg/day in 95% sunflower oil/5% ethanol) for five consecutive days to generate hepatocyte-specific G6pc1 deficient mice (L-G6pc À/À ) and littermate floxed controls (control), as described previously [27]. Female littermates were randomly assigned to experimental groups and the number of mice used for each experiment is indicated in the figure legends. No inclusion or exclusion criteria were used. Experiments were performed for 6e16 weeks after tamoxifen injections at ZT1 (fed), ZT7 (6 h fast during the inactive period or nonfasted), and in case of fasting, the next morning at ZT1 (refeeding during the inactive and active period). All animal studies were approved by the Institutional Animal Care and Use Committee from the University of Groningen under permit number AVD105002015244 and adhered to guidelines set out in the 2010/63/EU directive.

White blood cell and platelet counts
Blood samples were collected by tail bleeding into EDTA-coated tubes. Total white blood cell (WBC) and platelet counts were measured using the Medonic CD620 hematology analyzer (Boule Medical, Spanga, Sweden).

Flow cytometry
Blood samples were collected by tail bleeding into EDTA-coated tubes and kept on ice. For analysis of blood leukocyte subsets, samples were kept at 4 C during the whole procedure, unless stated otherwise. Red blood cells (RBCs) were lysed for 5 min (BD Pharm Lyse; BD Bioscience, Franklin Lakes, NJ, USA) and WBCs were centrifuged, washed, and resuspended in HBSS (0.1% BSA and 0.5 mM EDTA). To assess monocytes, monocyte subsets, neutrophil, B-cell and T-cell populations, cells were stained with a cocktail of antibodies: CD45-APC-Cy7 (557659; BD Biosciences, Franklin Lakes, NJ, USA), CD115-APC (17-1152-82;  2.9. Plasma corticosterone, epinephrine, and norepinephrine Blood was collected by retro-orbital bleeding into EDTA-coated tubes (450532; Greiner Bio-One, Kremsmünster, Austria) and kept on ice. Plasma was separated by centrifugation and 5 mg reduced glutathione (G4251; SigmaeAldrich, St. Louis, MO, USA) was added per mL of plasma after centrifugation to prevent degradation of epinephrine and norepinephrine. Plasma was snap-frozen and stored at À80 C until further use. Corticosterone, epinephrine, and norepinephrine were measured by online solid phase extraction (SPE) in combination with isotope dilution liquid chromatography tandem mass spectrometry (LC-MS/MS). Corticosterone measurement was performed essentially as described by Hawley et al. [28], using corticosterone-D4 as a stable isotope-labeled internal standard on an online SPE manager in combination with a XEVO TQ mass spectrometer (Waters, Milford, MA, USA). Epinephrine and norepinephrine were analyzed as described by Van Faassen et al. [29], using epinephrine-D3 and norepinephrine-D6 as stable isotope-labeled internal standard on a Symbiosis Pharma online SPE system (Spark Holland, Emmen, The Netherlands) in combination with a Xevo TQ mass spectrometer (Waters, Milford, MA, USA).

Adrenal gland mass
Twelve weeks after tamoxifen injections, mice were sacrificed by cardiac puncture in the fed or fasted condition. Adrenal glands were collected, cleaned of surrounding fat, and weighed.

Tail vein bleeding assay
The tail vein bleeding assay was performed as described previously [30]. Briefly, mice were anesthetized with isoflurane. A 10 mm segment of the distal tip of the tail was removed using a scalpel. The tail was immediately immersed in a 50 ml tube containing 50 ml phosphate-buffered saline (PBS) prewarmed to 37 C. Time until the bleeding ceased was recorded with a maximum bleeding time of 300 s. Volume of blood loss during the bleeding assay was quantified by measuring absorbance at 550 nm using a spectrophotometer (Genesys 10S UV-Vis; Thermofisher Scientific, Waltham, MA, USA).
2.12. Prothrombin time, activated partial prothrombin time, and plasma coagulation factors Blood was collected by cardiac puncture into tubes containing 3.2% sodium citrate (9:1 v/v) (41.1506.002; Sarstedt, Nümbrecht, Germany). Samples were immediately centrifuged at RT to collect plasma. Plasma was stored at À80 C until further use. Prothrombin time (PT) and activated partial thrombin time (aPTT) were assessed on an automated coagulation analyzer (CS-2100i) with reagents (PT-Innovin for PT and ActinFS for aPTT) and protocols from the manufacturer (Siemens, Marburg, Germany). Levels of factor V (FV), VII (FVII), and FVIII (FVIII) were determined on an automated coagulation analyzer (CS2100i) based on the one-stage clotting assay with factor-deficient plasma (Siemens, Marburg, Germany). Testing was performed in accordance with the protocols from the manufacturer (Siemens). Levels are denoted as percentages of standard human plasma (Siemens, Marburg, Germany).

Von Willebrand ristocetin cofactor
Blood was collected by cardiac puncture into tubes containing 3.2% sodium citrate (9:1 v/v) (41.1506.002; Sarstedt, Nümbrecht, Germany). Samples were immediately centrifuged at RT to collect plasma. Plasma was stored at À80 C until use. Von Willebrand ristocetin cofactor (vWF RCF) levels were determined on a CS2100i coagulation analyzer with the Innovance vWF activity kit (Siemens, Marburg, Germany) in accordance with the protocol from the manufacturer with all equipment and reagents from Siemens (Marburg, Germany).
2.14. Platelet aggregation To assess platelet aggregation, blood was drawn by cardiac puncture and collected into tubes containing 3.2% sodium citrate (9:1 v/v) (41.1506.002; Sarstedt, Nümbrecht, Germany). Samples were kept at RT during the whole procedure. Samples were centrifuged at 110Âg and platelet-rich plasma (PRP) was collected. PRP was diluted to a final concentration of 200*10 6 platelets per mL using platelet-poor plasma (PPP) from the same mouse. Platelet aggregation in response to 10 mM ADP was measured using light transmission aggregometry (Chrono-log Model 700; Chrono-log Corporation, PA, USA) at 37 C, stirring at 1200 rpm. Light transmission was recorded for 10 min using Aggro-Link 8 software.

Statistical analysis
All data are presented as mean AE SEM. The unpaired t-test was used to compare two datasets. To compare !3 groups, a one-way ANOVA with Bonferroni post hoc test was performed. Group size and statistical tests are reported in the figure legends. The criterion for significance was set at P < 0.05. Statistical analysis was performed using GraphPad Prism 5 (GraphPad Software; San Diego, CA, USA).

RESULTS
3.1. Fasting decreases blood leukocytes in L-G6pc À/À mice We first confirmed that similar to previous studies [27,31], hepatocytespecific G6pc1 deficiency induces hypoglycemia after a fasting period of 6 h. For this purpose, we measured blood glucose levels at Zeitgeber time (ZT) 1 (fed), ZT7 (6 h fast during the inactive period or nonfasted), and the next morning at ZT1 (refeeding during the inactive and active period). In accordance with previous studies [27], L-G6pc À/ À mice exhibited hypoglycemia (blood glucose 4.0 mM) after a 6 h fast, with average blood glucose levels of~3.6 mM ( Figure 1A).
Refeeding restored blood glucose levels to~7.9 mM (refeeding; ZT1) ( Figure 1A). L-G6pc À/À mice maintained euglycemia when they did not fast ( Figure 1B) probably because of the chow diet given to GSD Ia patients that resembles the dietary therapy. GSD Ia patients are on a dietary therapy of uncooked cornstarch that has a glycemic index of 48 [32], and as a consequence induces the slow release of glucose, preventing hypoglycemia in these patients [7]. Chow diet contains 67% carbohydrates with a glycemic index of 54 and 28 (specified in the methods section), which explains why L-G6pc À/À mice maintain euglycemia when not fasted. We observed the decrease in blood glucose levels in L-G6pc À/À mice upon a 6 h fast at all time points measured throughout the study ( Figure S1A and B).
Previous studies have revealed that prolonged fasting (20 h) leads to hypoglycemia, which decreases blood monocyte levels [23]. We thus assessed whether hepatocyte-specific G6pc1 deficiency affected blood leukocytes in the fasted and nonfasted condition. Control mice exhibited an increase in total CD45 þ blood leukocyte levels after fasting, because of circadian rhythmicity [33], whereas L-G6pc À/À mice did not ( Figure 1C). Refeeding restored blood leukocyte levels in L-G6pc À/À mice (ZT1; refeeding) ( Figure 1C). We observed no differences in blood leukocytes when L-G6pc À/À mice were not fasted and they remained euglycemic ( Figure 1D). Further analysis of leukocyte subsets in fasted animals denoted that hepatocyte-specific G6pc1 deficiency decreased total monocytes by~50%, mainly reflected by decreases in proinflammatory Ly6C hi and to a lesser extent Ly6C lo monocytes (Figure 1Ee I, K, M). Refeeding restored monocyte levels ( Figure 1I,K, M). Leukocyte levels were not different between the genotypes in the nonfasted condition ( Figure 1J, L, N). We then inquired which fasting-induced changes could account for the decrease in blood Ly6C hi monocytes in L-G6pc À/À mice during fasting. Blood Ly6C hi monocyte levels represented a strong correlation with blood glucose levels in L-G6pc À/À mice, but not in control mice upon a 6 h fast ( Fig. S2; Pearson correlation coefficient r ¼ 0.65; p ¼ 0.009), suggesting a major contribution of hypoglycemia to the decreased blood monocytes upon fasting.
Fasting also increases plasma VLDL-TG and VLDL-cholesterol in L-G6pc À/À mice [10]. However, elevated VLDL-cholesterol increases Ly6C hi monocytes in mice [34]. We found that fasting did not affect plasma FFAs in L-G6pc À/À mice, although plasma FFAs were elevated in the fed and fasted state, compared to controls (Fig. S3). Hence it is unlikely that lipids have contributed to the decreased blood Ly6C hi monocytes in fasted L-G6pc À/À mice. Though our data suggest that hypoglycemia is a key factor in decreasing blood monocytes, we cannot Original Article exclude that hormonal changes in fasted L-G6pc À/À mice [35] may have contributed to the decrease in monocytes. Administration of glucose by gavage during the fasting period, as carried out in wild-type mice in experiments by Jordan et al. [23], led to an even further aggravated hypoglycemia in L-G6pc À/À mice (results not shown), similar to rapid changes in blood glucose levels and rebound hypoglycemia that have been observed upon glucose administration in GSD Ia patients [7]. Therefore, we could not assess the effect of glucose on leukocyte changes in L-G6pc À/À mice directly. Neutrophils exhibited a similar response to monocytes in fasted and refed L-G6pc À/À mice, although fasting decreased neutrophils by only~26% ( Figure 1G,H,O,P). T-and B-cells were similarly affected ( Figure S4Ae F), with effects on B-cells being most pronounced. Collectively, hepatocyte-specific G6pc1 deficiency decreased circulating leukocytes upon a 6-h fast, which was restored by refeeding. All leukocyte populations were affected, and we observed major decreases in monocytes, particularly the Ly6C hi population, along with B-cells.

3.2.
Fasting increases plasma corticosterone levels in L-G6pc À/À mice We then sought to elucidate the mechanism causing the decrease in Ly6C hi monocytes in L-G6pc À/À mice upon fasting. MCP-1 enhances the recruitment of Ly6C hi monocytes from bone marrow by interacting with CCR2 on these cells [36e38]. A previous study has revealed that hypoglycemia decreases blood monocytes as a consequence of a reduction in plasma MCP-1 levels [23]. MCP-1 levels in the plasma of L-G6pc À/À mice were not affected compared to controls after a 6 h fast ( Figure S5A). M-CSF was also not affected ( Figure S5B). We observed a small increase in plasma G-CSF levels ( Figure S5C), which cannot explain the fasting-induced reduction in blood neutrophils in L-G6pc À/À mice. We then studied surface markers on monocytes and neutrophils. CCR2, CD11b, and VLA4 expression on Ly6C hi monocytes were not affected in L-G6pc À/À mice compared to controls after a 6 h fast ( Figure S5D), as were CXCR4, CD62L, CXCR2, and CD11b on neutrophils ( Figure S5E). This means hepatocyte-specific G6pc1 deficiency does not affect surface markers on monocytes and neutrophils upon fasting, suggesting no difference in the functionality of these cells. We then assessed where myeloid cells would accumulate during fasting-induced hypoglycemia in hepatocyte-specific G6pc1 deficiency. We observed an increase in monocytes in bone marrow in L-G6pc À/À mice upon fasting (Figure 2A). Although most bone marrow monocytes are Ly6C hi , the Ly6C lo population was increased, perhaps because monocytes lose their Ly6C expression upon retention in the bone marrow. Stem and progenitor cells including LSK cells, CMPs, and GMPs that give rise to myeloid cells were not affected in the bone marrow of fasted L-G6pc À/À mice ( Figure 2B), or in the spleen or liver (results not given). Livers from L-G6pc À/À mice exhibited increased neutrophils ( Figure 2C), in accordance with a previous study where this was attributed to increased hepatic expression of the neutrophil chemoattractant keratinocyte-derived chemokine (KC) that stimulates neutrophil infiltration [39,40]. The decrease in blood monocytes upon fasting in L-G6pc À/À mice may be caused by their accumulation in the bone marrow and the decrease in neutrophils because of their accumulation in the liver. We further investigated the mechanism underlying the decrease in blood Ly6C hi monocytes and their accumulation in bone marrow upon fasting-induced hypoglycemia in hepatocyte-specific G6pc1 deficiency. Hypoglycemia evokes a stress response that leads to increased plasma corticosterone levels in mice [41]. Corticosterone impairs the recruitment of monocytes from the bone marrow [42,43]. Bone marrow monocytes are mostly from the Ly6C hi subset [38]. We thus hypothesized that the decrease in Ly6C hi monocytes was caused by increased corticosterone levels as a result of hypoglycemia in fasted L-G6pc À/À mice. We measured levels of corticosterone and other stress hormones. While fasting increased plasma corticosterone levels in both genotypes, L-G6pc À/À mice presented a 2-fold higher plasma corticosterone level than controls after fasting ( Figure 3A). Adrenal gland mass was not affected after a 6 h fast, likely because the increase in plasma corticosterone was an acute effect ( Figure S6). Corticosterone increases the activity of the enzyme phenylethanolamine N-methyltransferase (PNMT), which mediates the conversion of norepinephrine into epinephrine [44]. In line with an increase in corticosterone levels, epinephrine levels were also increased by 2-fold upon fasting in L-G6pc À/À mice compared to controls ( Figure 3B), whereas norepinephrine levels were unaffected ( Figure 3C). Corticosterone and epinephrine have counter-regulatory effects on blood monocytes [42,43,45]. We observed a decrease in blood monocytes upon fasting and therefore hypothesized that the effect of corticosterone on blood monocytes is predominant, and investigated this further.
3.3. Inhibition of corticosterone signaling prevents the fastinginduced decrease in blood Ly6C hi monocytes in L-G6pc À/À mice We then assessed whether the reduction in Ly6C hi monocytes in fasted L-G6pc À/À mice was dependent on the increase in plasma corticosterone levels. We injected mice with the glucocorticoid receptor antagonist mifepristone or vehicle at the start of the fasting period. Strikingly, mifepristone completely prevented the fasting-induced decrease in total CD45 þ leukocytes in L-G6pc À/À mice ( Figure 4A). Mifepristone also prevented the fasting-induced decrease in Ly6C hi monocytes in L-G6pc À/À mice, whereas effects on other leukocyte populations did not reach statistical significance ( Figures. 4B and S7). These data confirm our hypothesis that the decrease in Ly6C hi monocytes in fasted L-G6pc À/À mice is due to an exaggerated stress response caused by elevated plasma corticosterone.
A recent study has revealed that hypoglycemia in hepatocyte-specific G6pc1 deficient mice impairs very-low-density lipoprotein (VLDL) catabolism, which was suggested to be caused by decreased lipoprotein lipase (LPL) activity [10]. As a result, plasma triglyceride levels were elevated [10]. Corticosterone may inhibit LPL activity [46,47]. We thus assessed whether the increase in plasma VLDL-triglycerides (VLDL-TG) in fasted L-G6pc À/À mice was dependent on plasma corticosterone levels. Mifepristone prevented the increase in plasma VLDL-TG upon fasting in L-G6pc À/À mice ( Figure S8A and B), suggesting that elevated corticosterone levels contribute to fastinginduced hypertriglyceridemia in GSD Ia.

Fasting increases bleeding time in L-G6pc
À/À mice The increased bleeding tendency in GSD Ia has been attributed to the decrease in von Willebrand factor (vWF) [19], disturbed platelet aggregation [16,18], and increased blood pressure [48]. Myeloid cells such as monocytes and neutrophils form leukocyte-platelet aggregates that contribute to coagulation [24]. Hence, a fasting-induced decrease in myeloid cells may reduce the formation of leukocyte-platelet aggregates, impair blood coagulation, and increase bleeding time. To test this hypothesis, we first assessed bleeding time in L-G6pc À/À mice in the fed, fasted, nonfasted, and refed condition using the tail vein bleeding assay. Similar to observations in GSD Ia patients with poor glycemic control, fasting increased bleeding time in L-G6pc À/À mice ( Figure 5A). Further confirming these data, L-G6pc À/À mice exhibited an increase in blood loss per minute upon fasting compared to controls ( Figure 5B). The increased bleeding time and blood volume loss per minute were restored upon refeeding. Bleeding time and blood volume Original Article loss per minute were not affected in the fed and nonfasted states. Collectively, these data show that fasting prolongs bleeding time in L-G6pc À/À mice. We then investigated whether prolonged bleeding time upon fasting was caused by a decrease in monocyte-aggregates and neutrophilaggregates because of a fasting-induced reduction in myeloid cells in L-G6pc À/À mice. The percentage of monocytes and neutrophils that formed aggregates with platelets (CD41 þ ) was not affected by hepatocyte-specific G6pc1 deficiency ( Figure S9A and B). This indicates that the capacity of monocytes and neutrophils to form aggregates with platelets did not change under fasting conditions. Nonetheless, the absolute decrease in these cell populations could contribute to less myeloid cell-platelet aggregates and increased bleeding during fasting in hepatocyte-specific G6pc1 deficiency.
3.5. Fasting does not decrease coagulation factors in L-G6pc À/À mice We next investigated other mechanisms accounting for the fastinginduced increase in the bleeding time of L-G6pc À/À mice. Since coagulation factors are mainly produced by the liver [49], where G6pc1 is highly expressed [50], we first assessed whether coagulation factors secreted by the liver and downstream pathways were affected by hepatocyte-specific G6pc1 deficiency. Prothrombin (PT) time, reflecting the intrinsic coagulation pathway that is dependent on factor VII and the common pathway consisting of factors X, II (thrombin), and I (fibrinogen), was not affected ( Figure 6A). Activated partial thromboplastin time (aPTT), reflecting the extrinsic coagulation pathway that depends on factors XI, XII, VIII, and IX, was also not affected ( Figure 6B). Plasma levels of coagulation factors V (FV) and  VIII (FVIII) were increased in L-G6pc À/À mice in the fed and fasted condition, while factor VII (FVII) was not affected (Figure 6CeE). The increases in FV and FVIII would increase coagulation, and hence, these increases are unlikely to explain the bleeding phenotype of L-G6pc À/À mice. It has been described that GSD Ia patients exhibited decreased vWF levels [19]. On the contrary, we found that L-G6pc À/À mice exhibited increased vWF levels upon fasting ( Figure 6F). The increase in plasma vWF and FVIII could be the consequence of the fasting-induced increase in plasma corticosterone [51e53]. Similar to FV and FVIII, vWF would increase coagulation. Therefore, these increases cannot explain the prolonged bleeding phenotype in fasted L-G6pc À/À mice. Hence, the bleeding phenotype in L-G6pc À/À mice upon fasting is not due to decreases in vWF, FVIII, or coagulation factors produced by the liver.
3.6. Fasting disturbs platelet aggregation in L-G6pc À/À mice Even though the number of patients was limited (between 2 and 7 per study), it has been reported that GSD Ia patients exhibit disturbed platelet aggregation [16e18]. This could account for the prolonged bleeding time in GSD Ia. We thus assessed whether fasting affected platelet aggregation in L-G6pc À/À mice. Blood platelet counts were not different between genotypes in both the fed and fasted condition ( Figure 7A), but studies on platelet function did reveal distinct differences. Initially, ADP-induced stimulation of platelet-rich plasma from both controls and L-G6pc À/À mice led to aggregation in both the fed and fasted condition ( Figure 7B). Strikingly, after maximum aggregation had been reached, platelets isolated from fasted mice, but not from non-fasted L-G6pc À/À mice, showed~50% platelet disaggregation ( Figure 7B,C). Hence, hepatocyte-specific G6pc1 deficiency disturbs platelet aggregation after fasting. Among all mechanisms that we have explored, it is most likely that disturbed platelet aggregation accounts for the prolonged bleeding time in L-G6pc À/À mice. We anticipate that this is also the main mechanism accounting for the bleeding tendency observed in hypoglycemic GSD Ia patients.

DISCUSSION
Our findings reveal that hepatocyte-specific G6pc1 deficiency in mice increases plasma corticosterone levels, decreases blood leukocytes, and disturbs platelet aggregation upon fasting. These results are most likely the consequence of fasting-induced hypoglycemia. The fastinginduced decrease in leukocytes in hepatocyte-specific G6pc1 deficiency was mainly reflected by a decrease in Ly6C hi monocytes and was prevented by the glucocorticoid receptor antagonist mifepristone.  Furthermore, we found that hepatocyte-specific G6pc1 deficiency prolonged bleeding time upon fasting due to disturbed platelet aggregation. Previous studies have recognized a relationship between hyperglycemia and monocytosis [54], and our data also indicate a similar relationship under conditions of fasting-induced hypoglycemia in hepatocyte-specific G6pc1 deficiency, reflected by decrease in blood monocyte levels and disturbed platelet aggregation. A relationship between glucose levels and blood monocytes in the setting of transient intermittent hyperglycemia independent of diabetes, and prolonged fasting has been shown previously [54]. Four injections of glucose over a period of 6 h increase blood monocyte levels, particularly the Ly6C hi population in wild-type mice [54], and a fasting period of 20 h decreases blood monocytes in wild-type mice, with reversal by glucose administration [23]. Similarly, a 19 h fasting period in humans decreases plasma glucose levels and blood monocytes [23]. Several studies thus indicate a positive relationship between blood monocytes and plasma glucose levels [23,54], sharing similarities with our observations in a mouse model of hepatic GSD Ia. While transient intermittent hyperglycemia increases blood monocytes because of elevated S100A8/A9 secretion by neutrophils as a consequence of glucose transporter (GLUT)-1 mediated glucose uptake [54], the decrease in blood monocytes upon hypoglycemia during Original Article 10 prolonged fasting is considered to be secondary to a decrease in MCP-1 [23]. MCP-1 interacts with its receptor CCR2 on bone marrow monocytes to promote their egress from the bone marrow into blood [36e38]. We did not detect differences in plasma MCP-1 upon fastinginduced hypoglycemia in hepatocyte-specific G6pc1 deficiency, but we did observe an increase in plasma corticosterone and epinephrine. Corticosterone decreases blood monocytes, but epinephrine has the opposite effect [45]. In diabetes, injections of insulin increase epinephrine levels [55] and blood monocytes [45], and presumably as a consequence thereof, patients with type 2 diabetes exhibit high levels of blood monocytes [22], although hyperglycemia also contributes to this effect [22]. Nonetheless, the question remains as to why monocytes were not increased as a consequence of the increased epinephrine in mice with hepatocyte-specific G6pc1 deficiency, similar to effects of insulin injections in patients with diabetes. We assume that this relates to diabetic patients having impaired adrenocorticotrophic hormone responses to hypoglycemia, leading to the suppression of cortisol production by the adrenal gland [56]. As such, cortisol levels are not upregulated during hypoglycemia as a consequence of insulin injections [56], whereas epinephrine levels do increase, leading to elevated blood monocyte levels. Our studies showed that fastinginduced hypoglycemia in hepatocyte-specific G6pc1 deficiency increases plasma levels of both corticosterone and epinephrine, and the decrease in blood monocytes suggests that the effect of corticosterone is predominant. Indeed, we found that the corticosterone receptor antagonist mifepristone reversed the fasting-induced decrease in Ly6C hi monocytes in L-G6pc À/À mice. We mainly observed differences in the Ly6C hi monocyte subset, presumably because corticosterone impairs the egress of monocytes from the bone marrow [42,43,57], and most bone marrow monocytes are Ly6C hi [38]. This is consistent with previous studies in rats demonstrating that acute stress or administration of corticosterone decreases blood monocytes within 2 h [42,58]. B-cells were also reduced by fasting in L-G6pc À/À mice. Previous studies have shown that blood B-cell levels are controlled by corticosterone, at least when corticosterone levels are elevated for a long period (72 h) [59]. Mifepristone tended to reverse the decrease in B-cells in L-G6pc À/À mice, although this did not reach statistical significance. Previous studies have also revealed that myocardial infarction (MI) leads to activation of the sympathetic nervous system, increasing norepinephrine production which acts on b 3 -adrenoceptors in bone marrow and promotes stem cell mobilization within 48 h of MI [60]. We did not observe any differences in norepinephrine levels or on hematopoietic stem and progenitor cells in the bone marrow, indicating that this mechanism did not play a role in L-G6pc À/À mice. GSD Ia patients display elevated levels of cortisol, the human homolog of corticosterone [61]. This was observed in the fed state, under conditions of good glycemic control. This effect was attributed to the production of NADPH in hepatocytes resulting from G6P accumulation. NADPH is a co-factor required for 11b-HSD1 reductase activity [62], resulting in the conversion of inactive glucocorticoids (cortisone in humans and 11-dehydrocortisone in mice) into active glucocorticoids (cortisol and corticosterone in humans and mice, respectively) [63]. GSD Ia patients exhibit increased conversion of administered cortisone into cortisol after overnight dexamethasone suppression [62], consistent with increased 11b-HSD1 reductase activity and elevated plasma cortisol levels [61]. Although L-G6pc À/À mice do accumulate G6P in the liver, 11b-HSD1 reductase activity was not affected [62]. This is likely the consequence of 11b-HSD1 reductase activity being 20 times higher in mice than in humans [62,64], and may explain why plasma corticosterone is not elevated in L-G6pc À/À mice in the fed state.
The increase in plasma cortisol in the fed state does not exclude that GSD Ia patients might exhibit high cortisol levels upon fasting-induced hypoglycemia, similar to mice with hepatocyte-specific G6pc1 deficiency. Hypoglycemia in GSD Ia patients correlates with hyperlipidemia, in particular, increased VLDL-TGs [7,10,65,66]. Cortisol increases VLDL-TG plasma levels [67,68]. We found that the glucocorticoid receptor antagonist mifepristone suppressed the fastinginduced increase in plasma VLDL-TG in L-G6pc À/À mice, consistent with a role for this antagonist in promoting VLDL-TG clearance, as suggested [69]. These data indicate that increased plasma cortisol caused by poor glycemic control contributes to high plasma VLDL-TG levels in GSD Ia patients [46,47]. Despite elevated VLDL-TG levels, a study in 9 GSD Ia patients has revealed that carotid intima-media thickness was not increased in GSD Ia patients compared to controls [15]. This effect can be explained as VLDL particles being too large to enter the vessel wall, similar to a setting of Apolipoprotein C3 overexpression in mice [70]. Nonetheless, our data reveal that in a setting of poor glycemic control, levels of blood proinflammatory Ly6C hi monocytes are decreased. Ly6C hi monocytes infiltrate into atherosclerotic lesions [71]. Monocytosis enhances atherosclerosis, as revealed by studies in animal models [71] and humans [72]. The outcome of our studies suggests that conditions of poor glycemic control in GSD Ia patients may reduce the levels of proinflammatory monocytes. This could contribute to the decrease in atherosclerosis in GSD Ia. We also found that fasted L-G6pc À/À mice exhibit prolonged bleeding time in the tail vein bleeding assay, replicating the enhanced bleeding tendency in GSD Ia patients with poor glycemic control [16e18]. Various mechanisms have been proposed for this observation, including decreases in vWF [19], disturbed platelet aggregation [16,18], and increased blood pressure [48]. Because GSD Ia is a rare disorder, studies typically include small numbers of patients (between 2 and 7 per study), and hence, the bleeding phenotype in GSD Ia has not been evaluated systematically. We found that some coagulation factors secreted by the liver, one of the main organs affected in GSD Ia, were increased in L-G6pc À/À mice upon fasting, as were FVIII and vWF; however, these increases would rather have led to increased coagulation and thus cannot explain the bleeding phenotype in GSD Ia. Strikingly, we observed that only in the fasted condition, L-G6pc À/À mice showed disturbed platelet aggregation. This was reflected bỹ 50% platelet disaggregation in response to ADP after maximum aggregation had been reached. This indicates that the secondary wave of platelet aggregation, which is dependent on the release of ADP and ATP stored in the dense granules of platelets [73,74], is absent in hepatocyte-specific G6pc1 deficiency. Therefore, the platelet activation process and recruitment of platelets cannot be amplified and a solid platelet plug cannot be formed [73e76]. This is commonly observed in patients with dense granule storage pool disease [77], which is characterized by an almost complete absence of dense granule ADP and ATP [78,79] and a bleeding phenotype [79]. In accordance with these observations, it has been reported that platelets of 2 GSD Ia patients contained less ADP and ATP during hypoglycemia [18]. Continuous gastric drip feeding to maintain normoglycemia corrected ADP and ATP levels along with the disturbed platelet aggregation phenotype [18]. Together with our results in mice with hepatocytespecific G6pc1 deficiency, these findings indicate that disturbed platelet aggregation in GSD Ia is the consequence of low levels of ATP and ADP in the platelet dense granules due to hypoglycemia. In addition, we cannot exclude that the decrease in myeloid cells in L-G6pc À/À mice may have led to a decrease in absolute numbers of monocyte-and neutrophil-platelet aggregates, which could contribute to increased bleeding in fasted L-G6pc À/À mice compared to controls.
Several studies have shown that glucose levels affect platelet aggregation. Platelet deficiency of GLUT 1 and GLUT 3, which mediate glucose uptake, decreases platelet activation in response to proteaseactivated receptor 4 peptide (PAR4) and prolongs bleeding time in mice [80,81]. These observations support our finding that bleeding time is increased in fasting-induced hypoglycemia in L-G6pc À/À mice. However, in contrast to findings in GSD Ia [16e18], acute hypoglycemia during an insulin stress test in humans increases platelet aggregation in response to ADP without affecting platelet ATP and ADP content [82]. The exact mechanism for this finding is unclear. Furthermore, acute insulin-induced hypoglycemia has been demonstrated to increase vWF levels in humans [83,84], likely the consequence of the hypoglycemia-induced increase in corticosterone levels [51,52] or endothelial dysfunction [84,85], and enhancing platelet aggregation. In sum, several mechanisms affecting platelet aggregation or disaggregation, including the one identified in GSD Ia, are regulated by blood glucose levels. A limitation of our study is the use of adult, hepatocyte-specific G6pc1 deficient mice. Whole body G6pc1 À/À mice require daily glucose injections to survive, they rarely live longer than 3 months, and cannot be used to investigate the effects of G6pc1 deficiency during fasting [11,86,87]. Although kidney-specific and intestinal-specific G6pc1 deficient mice do not become hypoglycemic upon fasting [88e90], we cannot exclude that in GSD Ia patients, loss of G6pc activity in kidney and intestine contributes to the bleeding phenotype under conditions of poor glycemic control.

CONCLUSION
We have shown that hepatocyte-specific G6pc1 deficiency increases plasma corticosterone levels upon fasting, which decreases Ly6C hi monocytes in the blood. Although direct effects of poor glycemic control on cortisol or leukocyte levels have not been evaluated in GSD Ia patients, this would be of interest, especially because increased plasma cortisol may contribute to elevated VLDL-TG under conditions of poor glycemic control in GSD Ia. In addition, we found that hepatocyte-specific G6pc1 deficiency disturbs platelet aggregation during fasting-induced hypoglycemia, accounting for the bleeding phenotype in GSD Ia. Together, these results highlight the clinical necessity of maintaining good glycemic control in GSD Ia.