Forkhead box protein O1 (FoxO1) regulates hepatic serine protease inhibitor B1 (serpinB1) expression in a non-cell-autonomous fashion

FoxO proteins are major targets of insulin action, and FoxO1 mediates the effects of insulin on hepatic glucose metabolism. We reported previously that serpinB1 is a liver-secreted factor (hepatokine) that promotes adaptive β-cell proliferation in response to insulin resistance in the liver-specific insulin receptor knockout (LIRKO) mouse. Here we report that FoxO1 plays a critical role in promoting serpinB1 expression in hepatic insulin resistance in a non-cell-autonomous manner. Mice lacking both the insulin receptor and FoxO1 (LIRFKO) exhibit reduced β-cell mass compared with LIRKO mice because of attenuation of β-cell proliferation. Although hepatic expression of serpinB1 mRNA and protein levels was increased in LIRKO mice, both the mRNA and protein levels returned to control levels in LIRFKO mice. Furthermore, liver-specific expression of constitutively active FoxO1 in transgenic mice induced an increase in hepatic serpinB1 mRNA and protein levels in refed mice. Conversely, serpinB1 mRNA and protein levels were reduced in mice lacking FoxO proteins in the liver. ChIP studies demonstrated that FoxO1 binds to three distinct sites located ∼9 kb upstream of the serpinb1 gene in primary mouse hepatocytes and that this binding is enhanced in hepatocytes from LIRKO mice. However, adenoviral expression of WT or constitutively active FoxO1 and insulin treatment are sufficient to regulate other FoxO1 target genes (IGFBP-1 and PEPCK) but not serpinB1 expression in mouse primary hepatocytes. These results indicate that liver FoxO1 promotes serpinB1 expression in hepatic insulin resistance and that non-cell-autonomous factors contribute to FoxO1-dependent effects on serpinB1 expression in the liver.


FoxO proteins are major targets of insulin action, and FoxO1 mediates the effects of insulin on hepatic glucose metabolism.
We reported previously that serpinB1 is a liver-secreted factor (hepatokine) that promotes adaptive ␤-cell proliferation in response to insulin resistance in the liver-specific insulin receptor knockout (LIRKO) mouse. Here we report that FoxO1 plays a critical role in promoting serpinB1 expression in hepatic insulin resistance in a non-cell-autonomous manner. Mice lacking both the insulin receptor and FoxO1 (LIRFKO) exhibit reduced ␤-cell mass compared with LIRKO mice because of attenuation of ␤-cell proliferation. Although hepatic expression of serpinB1 mRNA and protein levels was increased in LIRKO mice, both the mRNA and protein levels returned to control levels in LIRFKO mice. Furthermore, liver-specific expression of constitutively active FoxO1 in transgenic mice induced an increase in hepatic serpinB1 mRNA and protein levels in refed mice. Conversely, serpinB1 mRNA and protein levels were reduced in mice lacking FoxO proteins in the liver. ChIP studies demonstrated that FoxO1 binds to three distinct sites located ϳ9 kb upstream of the serpinb1 gene in primary mouse hepatocytes and that this binding is enhanced in hepatocytes from LIRKO mice. However, adenoviral expression of WT or constitutively active FoxO1 and insulin treatment are sufficient to regulate other FoxO1 target genes (IGFBP-1 and PEPCK) but not serpinB1 expression in mouse primary hepatocytes. These results indicate that liver FoxO1 promotes serpinB1 expression in hepatic insulin resistance and that non-cell-autonomous factors contribute to FoxO1-dependent effects on serpinB1 expression in the liver.
Diabetes mellitus develops when the number or function of pancreatic ␤-cells is insufficient to maintain normoglycemia. Typically, the mass of insulin-producing ␤-cells is reduced in diabetes, including the most commonly occurring types in the general population: type 1 and type 2 diabetes (1)(2)(3). Efforts to identify the molecular and cellular mechanisms that enable regeneration of pancreatic ␤-cells continue to be a major focus of diabetes research (4 -14). Competing strategies include the generation of insulin-secreting cells from pluripotent stem cells (9,10,15) or other sources (4,8,11,13,16) and the proliferation of pre-existing islet ␤-cells (5-7, 12, 17). The evolving stem cell-based approaches coupled with their oncogenic potential after transplantation suggest that this is not yet ready for therapeutic consideration (18). Current strategies have focused on harnessing the replicative activity (19,20) of ␤-cells, which is known to increase significantly in response to challenges such as insulin resistance (17,(21)(22)(23). Thus the identification of endogenous molecules that selectively enhance human ␤-cell numbers have the potential to yield safe anti-diabetic medications (24).
We have recently reported that serpinB1, a liver-derived molecule, can function as a novel ␤-cell growth factor (6). Although serpinB1 is expressed at very low levels in the liver under normal conditions, hepatic serpinB1 expression and production are considerably enhanced in models of insulin resistance, including liver-specific insulin receptor knockout (LIRKO) 7 mice, a unique model of islet ␤-cell hyperplasia (17,22). SerpinB1 also stimulates islet ␤-cell proliferation in multiple species, including zebrafish, mice, and humans, concomitant with activation of proteins in the growth factor signaling pathway (6).
FoxO Forkhead proteins are major targets of insulin action (25)(26)(27)(28). Insulin activates Akt, which induces phosphorylation at three highly conserved sites corresponding to Thr-24, Ser-256, and Ser-319 in human FoxO1. Akt-dependent phosphorylation leads to the translocation of FoxO proteins from the nucleus and sequestration in the cytoplasmic compartment in association with 14-3-3 proteins (29,30). Recent studies indicate that FoxO proteins promote hepatic glucose production by multiple mechanisms. In addition to enhancing the transcription of key enzymes encoded by phosphoenolpyruvate carboxykinase (pepck) and glucose 6-phosphatase (g6pase) and suppressing glycolytic and lipogenic gene expression and metabolism by direct mechanisms (30,31), studies of LIRKO and IR/FoxO1 double knockout (LIRFKO) mice indicate that liver FoxO1 also impacts the ability of insulin to regulate hepatic glucose production through extrahepatic mechanisms, including effects on the ability of insulin to suppress lipolysis in white adipose tissue (32), presumably because of production of FoxO1-dependent hepatokines.
Previously, we reported that, in addition to restoring the ability of insulin to maintain glucose homeostasis and suppress hepatic glucose production, disrupting FoxO1 expression in the liver also reverses ␤-cell hyperplasia in LIRKO mice (25). Because serpinB1 plays an important role in mediating the effects of hepatic insulin resistance on ␤-cell proliferation (6), we asked whether FoxO1 contributes to the expression of ser-pinB1 in the liver of LIRKO mice and whether these effects are mediated through direct, intrahepatic mechanisms and/or whether non-cell-autonomous factors are involved. Here we report that hepatic FoxO1, a major target of insulin action in the liver (25), plays an important role in regulating the expression of serpinB1 in the setting of hepatic insulin resistance and that this effect may involve both direct and indirect mechanisms, similar to FoxO1 regulation of hepatic glucose production.

FOXO1 regulates serpinb1 in hepatocytes
to 4-fold in LIRKO compared with IR fl/fl littermate controls, consistent with previous studies (6,17,22). In contrast, islet mass and ␤-cell proliferation were not increased in LIRFKO mice compared with floxed controls (Fig. 1, A-D), indicating that FoxO1 plays an important role in promoting ␤-cell proliferation in the setting of hepatic insulin resistance. In agreement with changes in islet ␤-cell proliferation and area, LIRKO mice exhibited 11-fold higher insulin levels compared with control mice, and this effect was also largely reversed in LIRFKO mice (Fig. 1E).
Next, we assessed the abundance of serpinB1 mRNA and protein levels in the liver of LIRKO, LIRFKO, and IR fl/fl mice. For comparison, we also assessed the expression of insulin-like growth factor-binding protein 1 (IGFBP-1), a canonical FoxO1 target gene that is directly regulated by FoxO proteins through well-characterized cis-acting FoxO-binding sites located within the proximal IGFBP-1 promoter (33). As expected, IGFBP-1 mRNA levels were increased ϳ50-fold in LIRKO livers compared with IR fl/fl or IR/FoxO1-floxed (IR fl/fl F fl/fl ) mice, whereas IGFBP-1 mRNA levels were not increased in LIRFKO-derived livers relative to IR fl/fl and IR fl/fl F fl/fl mice (Fig. 1F). Gene expression studies showed that liver serpinB1 mRNA levels also were increased ϳ70-fold in LIRKO compared with IR fl/fl mice (Fig.  1G), consistent with previous studies (6). In contrast, LIRFKO mice displayed no increase in serpinB1 gene expression compared with floxed controls, indicating that FoxO1 plays a critical role in promoting the expression of serpinB1 in the liver when hepatic insulin signaling is disrupted, suggesting effects on gene transcription, although effects on pre-mRNA or mRNA stability cannot be ruled out. Similarly, Western blot-ting showed that the level of serpinB1 protein was markedly increased in the livers of LIRKO but not LIRFKO mice compared with floxed controls (Fig. 1, H and I).

Regulation of hepatic serpinB1 in FoxO transgenic and FoxO knockout mice
To better characterize the effects of FoxO1 on serpinB1 expression, we also examined hepatic expression of serpinB1 (and IGFBP-1) in WT mice and FoxO1 transgenic mice (Tgn), which express a constitutively active form of FoxO1 (TSS/A-FoxO1) in the liver (34). WT and Tgn mice were fasted for 18 h overnight and sacrificed 6 h after refeeding, when the function of endogenous FoxO proteins is suppressed but TSS/A-FoxO1 remains active in the liver of Tgn mice (34). As shown in Fig. 2A, liver IGFBP-1 expression was dramatically enhanced in refed Tgn mice compared with the WT, as reported previously (34). SerpinB1 mRNA levels were also markedly increased (ϳ20fold) in the liver of Tgn versus WT mice, and Western blotting confirmed that serpinB1 protein levels were substantially increased in livers from refed Tgn versus WT mice (Fig. 2, C and  D). These results show that FoxO1 is sufficient to promote increased serpinB1 expression in the liver in vivo, similar to the effect of FoxO1 on IGFBP-1 expression.
In a related experiment, we also examined liver-specific Fox-OKO mice in which the expression of FoxO1, FoxO3, and FoxO4 was selectively disrupted in the liver to determine whether endogenous FoxO proteins contribute to the expression of serpinB1 in the liver under conditions where the insulin receptor has not been deleted. Nine-to ten-week-old male FoxOKO and floxed littermate controls (FoxO fl/fl ) were fasted

FoxO1 binds physically to sites upstream of serpinb1
Next, we sought to investigate whether FoxO1 binds directly to specific binding sites in the 5Ј upstream region of serpinb1 (NC_000079.6). Using the publicly available browser (https:// ecrbrowser.dcode.org (43)), 8 we observed that serpinb1 is located on mouse chromosome 13 between serpinb1c and wrnip1 (GRCm38.p4 chr13: 32,842,092-32,851,185, complement). Alignment of genomic sequences across mouse, human, and other species revealed that the 10-kb region upstream of serpinb1 contains conserved intergenic binding sites for several transcription factors, including three highly conserved upstream regions (which we named R1, R2, and R3) located ϳ9 kb upstream from the serpinB1 transcription initiation site. These sites share 75% sequence homology between mouse and human and contain putative FoxO-binding sites (Fig. 3A, arrows indicate localization of putative FoxO-binding sites). In particular, we identified two sites (R1 and R2) in upstream regions located at chr13: 32,859,879 -32,860,181 in addition to a third site (R3) located at the chr13: 32,861,943-32,862,290 genomic region located ϳ9 kb upstream of the serpinB1 transcription initiation site. In contrast, putative FoxO-binding sites were not identified in the proximal serpinb1 promoter region.
Based on these data, we performed ChIP assays to test whether FoxO1 binds to these predicted sites in hepatocytes isolated from 4-month-old IR fl/fl and LIRKO mice. Following cross-linking, chromatin isolation and sonication, and precipitation with a specific antibody against FoxO1 or the IgG control, the enrichment of DNA fragments containing FoxObinding sites was quantified by real-time PCR using two

FOXO1 regulates serpinb1 in hepatocytes
independent sets of primers that were designed for each of the three upstream regions containing a predicted FoxObinding site or for the proximal serpinb1 promoter, which did not contain a predicted high-affinity FoxO-binding site, as a control.
Preliminary studies of the ChIP assay were performed using an antibody against H3K9Ac, a posttranslationally modified form of histone 3 known to interact with the promoter region of gapdh. H3K9Ac was bound to the gapdh promoter after immunoprecipitation with anti-H3K9Ac antibody, and the relative enrichment was similar when the ChIP assay was conducted on control or LIRKO hepatocytes (Fig. 3B). We also confirmed that FoxO1 binding to the proximal promoter regions of several known FoxO1 target genes, including pepck, g6pase, and igfbp-1, which contain known FoxO-binding sites, was increased in hepatocytes from LIRKO mice compared with the floxed control (Fig. 3, C-E).
In contrast, we did not detect FoxO1 binding to unrelated sequences in the proximal promoter of serpinb1, as the relative chromatin enrichment containing this region of serpinb1 promoter was similar after immunoprecipitation with anti-FoxO1 using two different primer sets (PP1 and PP2) in both control and LIRKO hepatocytes (Fig. 3, F and G). However, ChIP experiments demonstrated significant binding (2.4-to 2.9-fold) of FoxO1 to the predicted FoxO-binding sites in regions R1 (Fig. 3, H and I), R2 (Fig. 3, J and K), and R3 (Fig. 3, L and M) upstream of serpinb1, as assessed by two primer sets, and the recruitment of FoxO1 to these binding sites was increased significantly further in hepatocytes from LIRKO mice compared with the control by 2.3-to 3.9-fold. Together, these data indicate that FoxO1 binds directly to multiple conserved FoxO1-binding sites located upstream of serpinb1 in hepatocytes and that FoxO1 binding to these sites is increased when insulin receptor signaling is disrupted in the liver.

Effects of insulin and FoxO1 on serpinb1 gene expression in hepatocytes in vitro
We next asked whether insulin and/or FoxO1 regulate ser-pinb1 gene expression in WT hepatocytes in a cell-autonomous manner. Mouse primary hepatocytes isolated from 4-monthold male C57BL6/J mice were stabilized in DMEM with 5 mM glucose and 0.1% fetal bovine serum (FBS) and then treated with DMEM with 5 mM glucose with or without 100 nM insulin for 1, 2, 4, 8, 12, or 24 h, followed by RNA extraction and measurement of serpinB1 and IGFBP-1 transcripts. IGFBP-1 mRNA levels were reduced by 50% after 2-h incubation with insulin and continued to decline and remained suppressed throughout the 24-h incubation period (Fig. 4A), indicating that insulin effectively regulates the expression of IGFBP-1, a canonical FoxO1 target gene, in primary hepatocytes under these conditions. In contrast, insulin treatment had no significant effect on serpinb1 gene expression at any time point in the study (Fig. 4B), suggesting that the effects of insulin on hepatocyte serpinB1 expression may not be cell-autonomous and that other factors also are required.
To determine whether FoxO1 is sufficient to promote ser-pinb1 gene expression in isolated hepatocytes, we transduced primary hepatocytes with adenovirus-expressing GFP alone (Ad-GFP), GFP plus WT FoxO1 (Ad-FoxO1(WT)), or GFP plus a constitutively active form of FoxO1 (Ad-FoxO1(TSS/A)), and gene expression was assessed 48 h later for serpinb1 and three other genes known to be regulated by FoxO1 (igfbp-1, pepck, and g6pase). At a multiplicity of infection (m.o.i.) of either 10 ( Fig. 4, C-F) or 100 (Fig. 4, G-J) adenoviral particles per hepatocyte, transfection with Ad-FoxO1(WT) and Ad-FoxO1(TSS/A) increased the expression of IGFBP-1, PEPCK, and Glc-6-Pase compared with Ad-GFP, confirming that FoxO1 is sufficient to promote the expression of these target genes in hepatocytes in a cell-autonomous manner and that other factors required for transactivation of by FoxO1 (35) are present in these hepatocytes. In contrast, transfection with either Ad-FoxO1(WT) or Ad-FoxO1(TSS/A) failed to increase the expression of serpinB1 (Fig. 4, C-J), suggesting that non-cell-autonomous factors may also be required for the regulation of serpinB1 by FoxO proteins in the liver.
To test whether the lack of activation of serpinB1 expression was due to an absence of binding of FoxO1 to the upstream binding sites in isolated hepatocytes, we also performed ChIP assays in primary hepatocytes infected with either Ad-GFP, Ad-FoxO1(WT), or Ad-FoxO1(TSS/A) and with an adenovirus expressing a DNA binding-defective form of FoxO1, Ad-FoxO1(H215R), which contains a point mutation within helix 3 of the DNA-binding domain that disrupts sequencespecific binding to the FoxO-binding site. As before, an antibody against H3K9Ac was used for the ChIP assay (Fig. 5A), and genes whose expression is controlled by FoxO1 and that contain known FoxO-binding sites within their proximal promoters (pepck or g6pase) were included as positive controls. As shown in Fig. 5, B and C, WT and TSSS/A-FoxO1 are recruited to the PEPCK and Glc-6-Pase promoters whereas FoxO1 (H215R) is not, indicating that this binding is mediated through the FoxO1 DNA-binding domain. Similarly, adenoviral infection of either Ad-FoxO1(WT) or Ad-FoxO1(TSS/A) enhanced FoxO1-binding to the genomic DNA at distinct upstream sites of the serpinB1 gene (Fig. 5, D-F). In contrast, binding was not enhanced by the expression of DNA binding-defective (H215R) FoxO1 (Fig. 5, D-F). These results indicate that FoxO1 is recruited to these sites through interactions with the FoxO1 DNA-binding domain. Taken together with gene expression studies (Fig. 4), these results also indicate that recruitment of FoxO1 to these sites is not sufficient to promote serpinb1 gene expression in isolated hepatocytes.
Together, these data suggest the novel concept that, in contrast to igfbp-1, pepck, and g6pase, insulin treatment and FoxO1 are not sufficient to modulate serpinb1 gene expression in isolated primary hepatocytes despite evidence supporting sequence-specific recruitment of FoxO1 to its binding sites upstream of the serpinb1 gene and that non-cell-autonomous effects of FoxO1 may also contribute to the regulation of hepatic serpinb1 expression by insulin in vivo.

Discussion
Diabetes mellitus is characterized by an absolute or relative scarcity of functional pancreatic islet insulin-secreting ␤-cells. Therapies that could increase pancreatic ␤-cell function or mass could potentially reverse the disease and halt its devastat-

FOXO1 regulates serpinb1 in hepatocytes
ing complications (36). The last decade has witnessed a substantial effort in devising methods to generate pancreatic ␤-cells (4 -12, 15-17). Previous studies have shown that compensatory ␤-cell proliferation is a common response to insulin resistance in a variety of settings, including obesity (21). Islet mass and ␤-cell proliferation are markedly increased in LIRKO mice, and we have shown that circulating endogenous factors produced by the liver (hepatokines) contribute to this effect, including serpinB1, which we found to have pancreatic ␤-cell proliferation-promoting activity in multiple species (6,17). Although hepatic expression and circulating levels of serpinB1 are markedly elevated in LIRKO mice, little is known regarding specific mechanisms involved in regulating the expression of serpinB1 in the liver. Here we show that FoxO1 plays an important role in promoting serpinB1 expression in vivo in the setting of hepatic insulin resistance.

FOXO1 regulates serpinb1 in hepatocytes
Previously, we reported that knocking out FoxO1 in the liver is sufficient to restore the ability of insulin to maintain glucose homeostasis and prevent ␤-cell hyperplasia in LIRKO mice (25). Here we show that the effect of knocking out FoxO1 in the liver on ␤-cell mass reflects changes in ␤-cell proliferation and the expression of serpinB1, indicating that effects of hepatic insulin resistance on serpinB1 expression and ␤-cell proliferation are FoxO1-dependent. Further, studies in transgenic mice expressing a constitutively active form of FoxO1 in the liver demonstrate that FoxO1 is also sufficient to enhance hepatic serpinB1 expression in mice in which insulin signaling has not been disrupted in the liver. Further, studies in liver-specific FoxO knockout mice show that endogenous FoxO proteins (including FoxO1, FoxO3, and FoxO4) contribute to hepatic expression of serpinB1 under fasting conditions, when insulin levels and effects are suppressed. Taken together, these results demonstrate that FoxO proteins promote the expression of ser-pinB1 in the liver and support the concept that FoxO-dependent effects on the expression of serpinB1 and possibly other factors may contribute to increased ␤-cell proliferation and mass in the setting of hepatic insulin resistance.
We also found that the regulation of serpinB1 by FoxO1 and insulin differs from that of several known FoxO target genes in the liver. Previous studies have emphasized cell-autonomous effects of FoxO proteins on gene expression because of recruitment to cis-acting elements through sequence-specific interactions with FoxO-binding sites often located within the proximal promoters of target genes. In this study, computer analysis identified several FoxO1-binding sites located ϳ9 kb upstream of the serpinb1 gene, and ChIP studies confirmed that FoxO1 is recruited to these upstream sites. In contrast, no putative FoxO-binding sites were identified in the serpinb1 proximal promoter, and ChIP studies showed that FoxO1 is also not recruited to the proximal promoter region of the serpinb1 gene through other mechanisms, e.g. protein-interactions. Adenoviral expression of WT and constitutively active FoxO1 was sufficient to enhance, and insulin treatment was sufficient to suppress, the expression of IGFBP-1, PEPCK, and Glc-6-Pase, consistent with previous studies, but not serpinB1, in isolated hepatocytes. These findings suggest that other factors also are required for FoxO proteins to effectively promote the expression of serpinB1 and suggest the novel concept that FoxO1 and insulin regulate serpinB1 expression in the liver in a non-cellautonomous fashion.
It is interesting to note that these differences in the regulation of serpinB1 versus IGFBP-1, PEPCK, and Glc-6-Pase correspond to differences in the organization of FoxO-binding sites in the 5Ј upstream regions of the serpinb1, igfbp-1, pepck, and g6pase genes. Functional FoxO-binding sites are most often located in the proximal promoter region and are found within several hundred base pairs of the transcription initiation site of the igfbp-1, pepck, and g6pase genes, where FoxO1 is able to bind and function together with other and trans-acting factors to stimulate gene expression directly. In this study, ChIP assays confirmed that FoxO1 interacts with the proximal promoter region of the igfbp-1, pepck, and g6pase genes and that this interaction is increased when the insulin receptor is knocked out in hepatocytes, consistent with previous studies indicating that FoxO1 directly contributes to insulin regulation of these genes. In contrast, ChIP studies showed that FoxO1 does not interact with elements located within the proximal serpinb1 promoter but, instead, that FoxO1 is recruited to several conserved FoxO-binding sites located further upstream of the serpinb1 gene, and this binding is also enhanced when the insulin receptor is knocked out. Furthermore, the binding of FoxO1 to this upstream region increases when the con-

FOXO1 regulates serpinb1 in hepatocytes
stitutively active form of FoxO1 (TSS/A) is expressed but not the DNA binding-defective FoxO1 (H215R) mutant. Taken together, these results indicate that the recruitment of FoxO1 to these upstream binding sites involves sequence-specific interaction with FoxO-binding sites but that is not sufficient to enhance the expression of serpinb1 in isolated hepatocytes and that other factors are also required for driving the transcription of the serpinb1 gene in the liver when insulin signaling is impaired.
Recent studies indicate that the extrahepatic effects of FoxO1 are also important in mediating other effects of hepatic insulin resistance, including the ability of insulin to regulate hepatic glucose production (HGP). Studies in LIRKO and LIRFKO mice indicate that insulin regulates HGP in a FoxO1dependent fashion (25,37), and yet insulin and FoxO1 have limited effects on glucose production in isolated hepatocytes (37). This apparent discrepancy has been resolved by the recognition that the ability of insulin to suppress lipolysis in white adipose tissue (WAT) and the flux of nonesterified fatty acids to the liver, where fatty acid oxidation supports and promotes gluconeogenesis, is critical for the ability of insulin to suppress HGP (38). Titchenell et al. (32) reported that the ability of insulin to suppress WAT lipolysis is impaired when insulin signaling is disrupted in the liver and that knocking out FoxO1 in the liver restores the ability of insulin to suppress WAT lipolysis and, thereby, regulate HGP. Thus, in addition to its direct, cell-autonomous effects on the expression of gluconeogenic and glycolytic genes in the liver, hepatic FoxO1 also impacts the ability of insulin to regulate HGP through indirect, non-cell-autonomous mechanisms. Based on our results, it is interesting to speculate that FoxO1-dependent mechanisms promoting the expression of liver serpinB1 may also involve non-cell-autonomous effects and communication with factors derived from other cells or tissues in the setting of hepatic insulin resistance. We observed that other genes, including the soluble leptin receptor and growth hormone receptor, are also regulated by FoxO1 in the liver in vivo but do not respond to transfection with FoxO1 expression vectors in isolated hepatocytes. 9 Together, these observations support the novel concept that FoxO1 contributes to the regulation of hepatic gene expression and metabolism through both cell-autonomous and non-autonomous mechanisms.
In summary, we found that FoxO1-dependent mechanisms promote the expression of serpinB1 when insulin signaling is disrupted in the liver, concomitant with effects on ␤-cell proliferation. FoxO proteins are required and sufficient to promote serpinB1 gene expression in vivo and interact with FoxO binding sites in multiple conserved binding sites upstream of the serpinb1 gene. However, unlike other FoxO target genes, neither insulin nor FoxO1 is sufficient to regulate the expression of serpinB1 in primary hepatocytes, suggesting that non-cell-autonomous factors are also involved in mediating the effects of insulin and FoxO1 on the expression of serpinB1 in the liver. These results support the broader concept that FoxO proteins function in the context of a complex regulatory network that includes both cell-autonomous and non-autonomous mechanisms involved in the control of metabolism. A potential role of other regulatory pathways and transcription factors contributing to mediating the expression of serpinB1 cannot be ruled out and requires further study. A better understanding of the mechanism(s) mediating the effects of insulin and FoxO proteins and potentially other transcription factors on the expression of serpinB1 may provide additional insight into mechanisms contributing to the regulation of gene expression and metabolism in hepatic insulin resistance and provide opportunities to selectively target hepatic serpinB1 expression to increase the number of functional ␤-cells in patients with diabetes.

Animals
Mouse studies were performed at the Jesse Brown Veterans Affairs Medical Center after approval by its institutional animal care committee. Male liver-specific transgenic mice expressing a modified form of human FoxO1, in which all three Akt phosphorylation sites (Thr-24, Ser-256, and Ser-319) were replaced with alanine residues (TSS/A-FoxO1), were crossed with female FVB/N WT female mice (Harlan Laboratories) to generate transgenic and WT littermate controls for studies (34). Liver-specific FoxOKO mice, in which the alleles for FoxO1, FoxO3, and FoxO4 are disrupted in the liver, were generated by crossing FoxO fl/fl mice (from Dr. Ron DePinho) with albumin-Cre (The Jackson Laboratory) mice (39). LIRKO and LIRFKO mice were generated previously by crossing albumin-Cre (The Jackson Laboratory) with IR fl/fl mice (provided by C. R. Kahn, Harvard Medical School) and FoxO1 fl/fl mice (provided by Ronald DePinho, MD Anderson Cancer Center), as described previously (25). Mice were housed on a 12-h:12-h light:dark cycle with lights off at 18:00 and provided standard chow. Plasma insulin was detected by ELISA (Crystal Chem).

Fasting/feeding studies
To evaluate the ability of FoxO1 to promote the expression of SerpinB1 in the liver, 10-week-old male FoxO1 transgenic mice and WT littermates were fasted for 18 h overnight, refed at 9 a.m., and sacrificed 6 h after refeeding, when the function of endogenous FoxO proteins is suppressed in the liver and the transgenic TSS/A-FoxO1 remains active, similar to previous studies (34,39). Conversely, to evaluate the role of endogenous FoxO proteins in promoting the expression of serpinB1 in the liver, liver-specific FoxO KO mice and floxed littermates were sacrificed at the end of an 18-h overnight fast, when insulin levels are low and endogenous FoxO proteins are active in the liver, as before (39).

Islet studies
For studies of ␤-cell proliferation, 10-week-old freely fed male LIRKO, LIRFKO, IR fl/fl , and IR fl/fl FoxO1 fl/fl mice were sacrificed between 10 and 11 a.m. by decapitation following brief sedation with isoflurane. Following collection of cervical blood, the liver was freeze-clamped with aluminum tongs precooled with liquid N 2 for storage at Ϫ80°C, and the pancreas was fixed 9 T. Unterman and W. Zhang, unpublished observations.

FOXO1 regulates serpinb1 in hepatocytes
overnight in Z-Fix (Anatech) and embedded in paraffin prior to sectioning in the University of Illinois Histology Core as before (25). For analysis of islet mass, tissue sections (4 m thick) were probed with anti-insulin antibody (Abcam) and then scanned at ϫ200 with an Aperio Scanscope CS for quantification of total tissue and islet cross-sectional area using ImageScope software (Aperio). For analysis of ␤-cell proliferation, sections were immunostained with anti-Ki67 (BD Biosciences) and antiinsulin (Abcam) antibodies. Cell counting was performed manually in a blinded fashion by a single observer. Insulin ϩ and Ki67 ϩ ␤-cells were identified using a fluorescence microscope (Olympus). Insulin ϩ cells showing nuclear 4Ј,6diamidino-2-phenylindole staining were considered ␤-cells. Insulin ϩ cells showing nuclear colocalized staining for 4Ј,6diamidino-2-phenylindole ϩ and Ki67 ϩ were counted as proliferating ␤-cells as described previously (6,17), and at least 1200 ␤-cells per pancreas were counted. At sacrifice, mice were briefly sedated with isoflurane prior to decapitation. The liver was snap-frozen in liquid nitrogen and then stored at Ϫ80°C until analysis.

Hepatocytes and adenoviral vectors
As we described previously (17), hepatocytes were isolated from male mice by collagenase digestion via portal vein perfusion. Hepatocytes were washed twice in hepatocyte wash medium (Invitrogen) and then seeded in collagen-coated 6-well plates (BD BioCoat) at a density of 10 6 cells/well in DMEM containing 25 mM glucose and 10% FBS (v/v) for 4 -6 h before further treatment. For studies of hepatocytes from LIRKO and IR fl/fl mice, plated cells were rinsed with PBS and refed with DMEM, 25 mM glucose, and 10% FBS and harvested 24 h later for ChIP analysis. For studies of insulin effects on gene expression, plated hepatocytes from WT C57Bl/6 mice were rinsed with PBS and then refed with DMEM plus 5 mM glucose and 0.1% FBS for 16 h prior to treatment with DMEM containing 5 mM glucose with or without 100 nM insulin for 1-24 h. For studies of FoxO with adenoviral vectors, plated hepatocytes from WT C57Bl/6 were rinsed with PBS and refed with DMEM, 25 mM glucose, 10% FBS and transfected with adenoviral vectors (10 -100 infectious units/cell) for 16 h when cells were refed with DMEM, 25 mM glucose, and 10% FBS. Hepatocytes were harvested 24 h later for analysis of gene expression or ChIP studies. Adenoviral vectors, which express GFP alone or GFP plus WT FoxO1, TSS-A FoxO1, or H215R FoxO1 were described previously (34,40).

Western blotting
Tissue samples were lysed in M-PER buffer (Thermo Fisher Scientific), and total protein concentration was measured by BCA assay (Pierce). Samples were resuspended in Laemmli buffer with ␤-mercaptoethanol, boiled, and resolved by SDS-PAGE. Proteins were transferred onto nitrocellulose membranes, blocked in PBS containing 5% BSA and 0.1% Tween 20, and incubated with rabbit antiserum to human SerpinB1 (42), GAPDH (Cell Signaling Technology, 5174P), or mouse antiserum to actin (Santa Cruz Biotechnology, sc-1615) or ␤-tubulin (Santa Cruz Biotechnology, sc-5274). Secondary goat anti-rab-FOXO1 regulates serpinb1 in hepatocytes bit (Santa Cruz Biotechnology, sc-2054) or goat anti-mouse (Santa Cruz Biotechnology, sc-2055) was used thereafter. Protein bands were quantified using ImageJ software.

Statistical analysis
All data are presented as mean Ϯ S.E. Data were analyzed using unpaired, two-tailed Student's t test. p Ͻ 0.05 was considered statistically significant.