2224 Diabetes Volume 65, August 2016

Zhiyong Zhang,1 Louis F. Amorosa,2 Susette M. Coyle,1 Marie A. Macor,1 Morris J. Birnbaum,3 Leonard Y. Lee,1 and Beatrice Haimovich1

Insulin-Dependent Regulation of mTORC2-Akt-FoxO Suppresses TLR4 Signaling in Human Leukocytes: Relevance to

Diabetes 2016;65:2224–2234 | DOI: 10.2337/db16-0027

Leukocyte signaling in patients with systemic insulin re- (1). Modest increases in inflammatory indicators detected in sistance is largely unexplored. We recently discovered the patients have suggested that type 2 diabetes is associated presence of multiple Toll-like receptor 4 (TLR4) signaling with chronic low-grade inflammation (2,3). Human subjects intermediates in leukocytes from patients with type 2 challenged with lipopolysaccharide (LPS), a ligand of Toll-like diabetes or acute associated with cardio- receptor 4 (TLR4) (4), produce acute inflammatory re- pulmonary bypass surgery. We extend this work to show sponses that include insulin resistance (5,6). Mice adminis- that in addition to matrix metalloproteinase 9, hypoxia- tered a low dose of LPS for 1 month develop insulin inducible factor 1a, and cleaved AMPKa, patient leukocytes fi 312 resistance (7). On the other hand, TLR4-de cient mice also express IRS-1 phosphorylated on Ser , Akt phosphor- andmicetransplantedwithbonemarrowcellsfromTLR4- ylated on Thr308, and elevated TLR4 expression. Similar sig- deficient mice were more resistant to high-fat diet–induced naling intermediates were detected in leukocytes and insulin resistance than wild-type mice or mice transplanted neutrophils treated with lipopolysaccharide (LPS), a ligand withbonemarrowcellsfromwild-typemice(8).Theseand of TLR4, in vitro. In contrast, insulin, but not LPS, induced mammalian target of rapamycin complex 2 (mTORC2)– other studies have indicated that immune cells and TLR4

SIGNAL TRANSDUCTION – dependent phosphorylation of Akt on Ser473 and FoxO1/O3a signaling contribute to systemic insulin resistance (8 10). on Thr24/32 in leukocytes and neutrophils. Insulin suppressed We reported that LPS triggers rapid changes in cellu- LPS-induced responses in a dose- and time-dependent lar metabolism in human and mouse leukocytes (11,12) – manner. AS1842856, a FoxO1 inhibitor, also suppressed through a phosphoinositide 3- (PI3K) dependent TLR4 signaling. We propose that insulin is a homeostatic process that involves two metabolic regulators: AMPK regulator of leukocyte responses to LPS/TLR4 and that the and hypoxia-inducible factor 1 (HIF-1) (13). We showed signaling intermediates expressed in leukocytes of patients that LPS increases the expression of matrix metallopro- with type 2 diabetes indicate TLR4 signaling dominance teinase 9 (MMP9) in leukocytes and neutrophils and that and deficient insulin signaling. The data suggest that in- intracellular MMP9 cleaves the catalytic subunit of AMPK sulin suppresses LPS/TLR4 signals in leukocytes through (AMPKa) (13). AMPK inactivation enables dephosphory- the mTORC2-Akt-FoxO signaling axis. Better understand- lation of the regulatory-associated protein of mammalian ing of leukocyte signaling in patients with type 2 diabetes target of rapamycin (mTOR) (Raptor) on Ser792, resulting may shed new light on disease causation and progression. in mTOR complex 1 (mTORC1) activation (14,15). mTORC1 phosphorylates ribosomal S6 kinase (S6K1) on Robust inflammatory responses associated with injury or Thr389 (16,17). In addition, mTORC1 activates the tran- sepsis trigger the onset of insulin resistance in humans scription factor HIF-1, which promotes glycolysis and

1Department of Surgery, Rutgers Robert Wood Johnson Medical School, New This article contains Supplementary Data online at http://diabetes Brunswick, NJ .diabetesjournals.org/lookup/suppl/doi:10.2337/db16-0027/-/DC1. 2 Department of Medicine, Rutgers Robert Wood Johnson Medical School, New M.J.B. is currently affiliated with the Cardiovascular and Metabolic Research Unit, Brunswick, NJ Pfizer, Inc., Cambridge, MA. 3Institute of Diabetes, Obesity, and Metabolism, Perelman School of Medicine, © 2016 by the American Diabetes Association. Readers may use this article as University of Pennsylvania, Philadelphia, PA long as the work is properly cited, the use is educational and not for profit, and Corresponding author: Beatrice Haimovich, [email protected]. the work is not altered. Received 6 January 2016 and accepted 2 May 2016. diabetes.diabetesjournals.org Zhang and Associates 2225 proinflammatory cytokine production in both human and RESEARCH DESIGN AND METHODS mouse leukocytes treated with LPS (18,19). Because nei- Materials a 389 ther HIF-1 nor S6K1 phosphorylated on Thr was de- The following antibodies were used: anti- (Sigma- fi tected in MMP9-de cient mouse leukocytes treated with Aldrich); anti-HIF-1a, AMPKa, MMP9, IkB-a, total Akt, LPS, the mechanism by which TLR4 activates mTORC1 and TLR4 (Santa Cruz Biotechnology); and anti-Raptor might be unique (13). Ser792, total Raptor, S6K1 Thr389,AktThr308, Akt Ser473, A second complex involving mTOR, known as mTOR IRS-1 Ser312 (Ser307 in mice), IRS-1, and FoxO1/FoxO3a complex 2 (mTORC2), regulates the phosphorylation of Thr24/Thr32 (Cell Signaling Technology). Source and final 473 Akt on Ser (20,21). Activated Akt phosphorylates concentration of pharmacologic agents were as follows: FoxO1 and FoxO3a, which belong to the forkhead family LPS from Escherichia coli 0111:B4 (Sigma-Aldrich); insulin of transcription factors, on multiple threonine and ser- (Humalog; Eli Lilly), rapamycin (100 nmol/L), Torin ine residues (22). Phosphorylated FoxO proteins are ex- (50 nmol/L), and PP242 (100 nmol/L) (Tocris); and AS1842856 cluded from the nucleus and remain inactive (23). FoxO1 (100 nmol/L) (Millipore). regulates leukocyte survival (24–26) and TLR4 signaling (27,28). Although activated FoxO1 triggers an increase Human Subjects in TLR4 expression in LPS-treated Raw264.7 cells (28), The Rutgers Health Sciences Institutional Review Board it also contributes to LPS-dependent TLR4 signaling approved the study. Patients with type 2 diabetes and inactivation in these cells (28). The role of FoxO rela- patients without diabetes were recruited from endocri- tive to TLR4 signaling in human leukocytes is currently nology clinics at Rutgers Robert Wood Johnson Medical unknown. School during a scheduled visit. Written informed consent Human leukocytes express an insulin receptor (29). was obtained from all subjects before inclusion in the Insulin receptor activation triggers phosphorylation of study. Twelve patients with type 2 diabetes were recruited, insulin receptor substrate (IRS) proteins on tyrosine and their demographics are shown in Table 1. Patient clinical residues. This leads to PI3K and phosphoinositide- records are presented in Supplementary Table 1. Cardiac dependent kinase 1 activation, the latter of which phosphor- surgery patients were recruited from Rutgers Robert Wood ylates Akt on Thr308 (30). Insulin induces phosphorylation Johnson University Hospital. Inclusion and exclusion crite- of Akt on Ser473 through mTORC2, and Akt phosphor- ria were described previously (13) and are included in the ylates FoxO (21,31,32). FoxO1 phosphorylation was Supplementary Data. impaired in macrophages derived from insulin resis- Leukocytes Isolation tant obese mice (33). Despite this general frame- Blood drawn into heparin-containing tubes was sepa- work,theinsulinsignalingpathwayinleukocytesis rated into aliquots and treated with LPS (10 ng/mL) undetermined. or insulin (1 unit/mL), unless otherwise indicated. The We recently showed that leukocytes from a cohort of patients with type 2 diabetes exhibited distinct signaling intermediates (13). From these observations, we hypoth- esized that characterization of signaling intermediates Table 1—Patient characteristics expressed in patient leukocytes could have utility in deter- Without With type 2 diabetes diabetes mining disease etiology and progression. The data showed that leukocytes from patients with type 2 diabetes exhibit No. of patients 10 12 two patterns of signaling intermediates. The presence Age (years) 61 6 4616 2 of MMP9 and HIF-1a as well as cleaved AMPKa define Sex the first pattern (pattern 1), whereas the presence of Male 5 9 Akt phosphorylated on Ser473 defines the second (pattern 2). Female 5 3 2 IRS-1 phosphorylated on Ser312 and S6K1 phosphorylated BMI (kg/m )346 3296 1 on Thr389 were seen in leukocytes expressing either pat- Fasting plasma glucose (mg/dL) 101 6 5 147 6 14* tern 1 or pattern 2. The signaling intermediates asso- HbA1c ciated with pattern 1 were reproduced in leukocytes % 5.9 6 0.1 7.7 6 0.3** treated with LPS in vitro. Insulin but not LPS induced mmol/mol 41 60.7 phosphorylation of Akt on Ser473 and FoxO1/FoxO3a Cholesterol (mg/dL) 157 6 14 171 6 16 on Thr24/Thr32 in leukocytes and neutrophils and HDL cholesterol (mg/dL) 49 6 4506 2 suppressed LPS-induced TLR4 signaling in a dose- and LDL cholesterol (mg/dL) 89 6 10 96 6 14 time-dependent manner. The data suggested that TLR4 Triglycerides (mg/dL) 94 6 20 127 6 12 signaling is linked to pattern 1. We propose that leuko- Alanine aminotransferase (IU/L) 28 6 9306 6 cytes with pattern 2 are partially responsive to insulin, Aspartate aminotransferase (IU/L) 18 6 2276 3 which is no longer able to fully suppress the inflam- 6 P matory signals present in these patients’ leukocytes/ Data are mean SEM unless otherwise indicated. * = 0.024. **P = 0.002. neutrophils. 2226 Insulin and TLR4 Signaling in Human Leukocytes Diabetes Volume 65, August 2016 leukocytes were isolated as previously described (12). Ly- exhibited pattern 2. These leukocytes did not express sates containing equal protein amounts were analyzed by MMP9, HIF-1a,orcleavedAMPKa (Fig. 1A,lane3)(13). immunoblotting. The signaling intermediates associated with pattern 1 Neutrophil and Mononucleated Cell Isolation were reproduced in leukocytes and neutrophils treated with LPS in vitro to activate TLR4 (13). In the current Neutrophils and mononucleated cells (which include study, we sought to identify additional leukocyte sig- monocytes and lymphocytes) were first separated by naling components that may relate to patient insulin using Ficoll-Hypaque (Sigma-Aldrich). The mononuclear resistance. To this end, we examined whether the expression cell fraction was washed twice with PBS and then suspended in DMEM. The neutrophil fraction was separated from the and/or phosphorylation of three signaling components (Akt, IRS-1, and TLR4) previously linked to insulin resistance (34– red blood cells by using dextran (molecular weight 500,000) 36) are modified in leukocytes from the cohort of patients sedimentation. After lysis of residual red blood cells and described in Table 1 and our previous work (13). two washes with PBS, the neutrophils were suspended in Optimal Akt activation is suggested to require phos- DMEM. Isolated neutrophil and mononucleated cell suspen- phorylation on both Thr308 and Ser473 (37). Leukocytes sions were treated with LPS and/or insulin and then lysed. from 10 of the 12 patients with type 2 diabetes showed Lysates containing equal protein amounts were analyzed by Akt phosphorylated on Thr308 (Fig. 1B), as did leukocytes immunoblotting. from one patient with (subject 16: HbA1c RESULTS 6.0% [46.4 mmol/mol], fasting glucose 117 mg/dL). How- 473 Akt Phosphorylation on Thr308 and/or Ser473, IRS-1 ever, Akt phosphorylated on Ser was only detected in Phosphorylation on Ser312, and Increased TLR4 leukocytes showing pattern 2 (Fig. 1B, lanes 4, 7, 17, and Expression Are Detected in Leukocytes From Patients 21) (13), suggesting differential regulation of Akt phos- With Type 2 Diabetes phorylation on residues Thr308 and Ser473 in leukocytes of We previously characterized signaling intermediates found patients with type 2 diabetes. in peripheral blood leukocytes from 12 patients with type 2 Insulin resistance is frequently associated with either diabetes as defined by hemoglobin A1c (HbA1c) $6.4% a decline in IRS protein expression or increased IRS phos- ($46.4 mmol/mol) and/or concurrent treatment for diabe- phorylation on serine residues (38,39). Leukocytes from tes (13). The demographics of patients with and without patients with and without diabetes expressed similar IRS-1 diabetes are reported in Table 1. All 12 samples from pa- protein levels (Fig. 1B). However, phosphorylation of IRS-1 tients with type 2 diabetes exhibited S6K1 phosphorylation on Ser312 (equivalent to murine IRS-1 Ser307) was detected on Thr389 and dephosphorylated Raptor on Ser792 (12) in all 12 patients with diabetes (Fig. 1B)aswellasinthe (Fig. 1). In addition to these markers, patient leukocytes one patient with prediabetes (subject 16). expressed two predominant patterns of signaling inter- Skeletal muscle biopsy specimens and monocytes from mediates. Pattern 1, seen in leukocytes from 8 of 12 patients with type 2 diabetes express higher TLR4 protein patients, included MMP9, HIF-1a,andcleavedAMPKa levels than control subjects (34,35). These findings moti- (Fig. 1A, lane 1) (13). Leukocytes from 4 of 12 patients vated us to examine the expression of TLR4 in patient

Figure 1—Signaling intermediates expressed in leukocytes from patients with type 2 diabetes. Blood samples were obtained from patients without diabetes (N) (n = 10) and patients with type 2 diabetes (2) (n = 12) seen at Rutgers Robert Wood Johnson Medical School clinics. The researchers were blinded to the clinical characteristics of the study participants (Table 1) while the samples were being analyzed. *, a healthy control sample used multiple times; ▼, a patient without diabetes with an HbA1c of 6% (46.4 mmol/mol) and fasting glucose of 117 mg/dL. A: Signaling intermediates detected in leukocytes from a patient without diabetes (lane 2) and representative patients with type 2 diabetes showing pattern 1 (lane 1) or pattern 2 (lane 3). B: Samples presented in the same order as those in Zhang et al. (13) to enable direct comparison. Leukocytes were isolated, lysed, and analyzed by immunoblotting to determine the expression levels and phosphor- ylation state of the specified proteins. Immunoblotting for actin was used to confirm equal protein loading. p-S, phosphorylation on serine; p-T, phosphorylation on threonine. diabetes.diabetesjournals.org Zhang and Associates 2227 leukocytes. Increased TLR4 expression was detected in 10 detected within 10 min of LPS stimulation, and the phos- of the 12 leukocyte samples of patients with type 2 di- phorylation signal remained strong for at least 4 h (Fig. abetes (Fig. 1B). 2B). Though initially weak, phosphorylation of IRS-1 on Ser312 was seen by 30 min (Fig. 2B). LPS also triggered a Phosphorylation of Akt on Thr308 but Not Ser473, significant increase in TLR4 expression. These data suggest Phosphorylation of IRS-1 on Ser312, and Increased TLR4 Expression Are Detected in Leukocytes Treated that the signaling intermediates associated with pattern 1 With LPS In Vitro are governed by a common mechanism linked to TLR4. We have previously shown that signaling intermediates Insulin Suppresses LPS-Induced TLR4 Signaling in expressed in leukocytes from patients with type 2 di- Leukocytes abetes were reproduced in leukocytes treated with LPS in Having observed Akt phosphorylation on Thr308 but not vitro (13) (Fig. 2A), implicating TLR4 in the cascade of Ser473 in our LPS model, we sought to identify clinical aberrant signaling events associated with clinical insulin factors determining the presence of Ser473 phosphoryla- resistance. We therefore sought to determine whether tion in leukocytes. Examination of patient medical records LPS/TLR4 signaling could also account for the phosphor- revealed that the four patients with type 2 diabetes ylation of Akt on Thr308 and Ser473, the phosphorylation expressing Akt phosphorylated on Ser473 were treated of IRS-1 on Ser312, and/or the increase in TLR4 expres- with insulin (Supplementary Table 1). Because insulin sion seen in leukocytes from patients with type 2 dia- triggers phosphorylation of Akt on Ser473 and Thr308 betes. Akt phosphorylated on Thr308 but not Ser473 was and S6K1 on Thr389 in various model cell systems (36),

Figure 2—LPS- and insulin-induced responses in human leukocytes. A, B, and E: Blood samples were treated with LPS for the indicated time. The sample shown in panel B, lane 8, was treated with insulin (1 unit/mL) for 60 min and served as a positive control. C, D, and F: Blood samples were treated with insulin (1 unit/mL) for the indicated time. The sample shown in panel D, lane 8, was treated with LPS (10 ng/mL) for 90 min and served as a positive control. G: Leukocytes from patients with type 2 diabetes exhibiting either pattern 1 or pattern 2 were analyzed for Akt phosphorylation on Ser473 and FoxO1/FoxO3a phosphorylation on Thr24/Thr32. The patient IDs correspond to those presented in Fig. 1B. H: Blood samples were obtained from a CPB patient on the morning of the surgery, in the recovery room, and on days 1–4 postsurgery. Similar data were obtained for two other patients. D, day; P, presurgery; p-S, phosphorylation on serine; p-T, phosphorylation on threonine; R, recovery room. 2228 Insulin and TLR4 Signaling in Human Leukocytes Diabetes Volume 65, August 2016 we examined whether insulin can induce similar responses intermediates associated with pattern 1 returned to in leukocytes. Leukocytes treated with insulin in vitro baseline, whereas Akt phosphorylated on Ser473 and exhibited a transient increase in S6K1 phosphorylation FoxO1/FoxO3a on Thr24/Thr32 appeared. Thus, the on Thr389 that peaked by 60 min (Fig. 2C). Insulin also leukocytes exhibited pattern 1 while the patient was in- triggered transient increases in Akt phosphorylation on sulin resistant. The signaling mediators associated with Thr308 and Ser473 (Fig. 2D). Akt remained phosphorylated patterns 1 and 2 detected in patients with type 2 diabetes on Ser473 for at least 120 min, whereas the phosphoryla- (Figs. 1B and 2G) and in patients after CPB surgery (Fig. tion on Thr308 decreased below the detection level after 2H) are presented in Table 2. 90 min. 473 Insulin Suppresses LPS-Induced Responses In general, Akt phosphorylated on Ser phosphory- in a Dose- and Time-Dependent Manner lates FoxO proteins, which are transcription factors. Studies have suggested that even a single meal high in fat Phosphorylated FoxO proteins are excluded from the nu- can trigger release of LPS from the gut into the circulation cleus and therefore inactive. Human neutrophils express (41). Because insulin levels increase after food intake, it is FoxO1, FoxO3a, and FoxO4 (24). The presence of Akt 473 possible that human leukocytes are exposed periodically phosphorylated on Ser in leukocytes treated with in- and transiently to LPS plus insulin. For this reason, we sulin suggested that insulin might also induce phosphor- next examined the combined effects of LPS and insulin on ylation of FoxO proteins. By using an antibody that reacts leukocyte signaling. Although LPS-induced signals were with both FoxO1 and FoxO3a when phosphorylated on A B 24 32 detected within minutes (Fig. 2 and ), combined treat- Thr and Thr , respectively, we determined that in- ment with LPS and insulin delayed the appearance of sulin induces FoxO1/FoxO3a phosphorylation in leuko- TLR4 signaling mediators until at least 90 min (Fig. 3A F cytes (Fig. 2 ). LPS failed to induce a similar response and B). Furthermore, the appearance of TLR4 signaling E (Fig. 2 ) mediators after ;90 min coincided with a decline in Akt fi This nding raised the question about whether FoxO1 phosphorylation on Ser473 and FoxO1/FoxO3a on Thr24/Thr32. and FoxO3a are phosphorylated in the leukocytes of In complementary dose-dependent studies, leukocytes patients with type 2 diabetes. To this end, we reanalyzed were treated for 1 h with decreasing insulin concentra- B 8 of the 12 samples shown in Fig. 1 . The patient ID tions ranging from 1 to 0.05 units/mL (Fig. 3C and D)in G numbers shown in Fig. 2 correspond to those in Fig. combination with LPS (10 ng/mL). Control samples were B 1 . FoxO1 and FoxO3a were not phosphorylated in leuko- treated with insulin alone for 1 or 2 h. Akt phosphoryla- cytes with pattern 1. Two of the four samples with tion on Ser473, but not on Thr308, was detected in leuko- pattern 2 showed no FoxO1/FoxO3a phosphorylation cytes treated for 2 h with insulin alone (1 unit/mL) (Fig. (subjects 7 and 4). The other two exhibited weak phos- 3D, lane 4). As the phosphorylation of Akt on Ser473 de- phorylation (subjects 17 and 21). Of note, although sub- clined in leukocytes treated with insulin at a concentra- ject samples 7 and 4 exhibited elevated TLR4 expression, tion ,0.25 units/mL, TLR4 signaling intermediates samples 17 and 21 did not. Collectively, these data in- became visible (Fig. 3C and D, lanes 8 and 9). These studies dicate that the phosphorylation of FoxO1/FoxO3a is im- paired in leukocytes from patients with type 2 diabetes. Regardless of diabetes history, the majority of patients Table 2—Summary of signaling intermediates detected in who undergo cardiopulmonary bypass (CPB) surgery expe- leukocytes from patients without and with type 2 diabetes rience an acute phase of insulin resistance immediately with pattern 1 or pattern 2 after surgery and require insulin infusion to maintain With type 2 diabetes glucose within a normal range (40). We reported that Pattern 1 Pattern 2 leukocytes from patients who underwent CPB surgery ex- Without (8 of 12 (4 of 12 press signaling intermediates postsurgery similar to those diabetes patients) patients) seen in leukocytes from patients with type 2 diabetes (13). MMP9 expressed N Y N Here, we studied temporal changes in Akt, IRS-1, and AMPKa cleaved N Y N FoxO1/FoxO3a phosphorylation in leukocytes from pa- p-Raptor Ser792 YN N tients who underwent CPB. Signaling intermediates seen a in leukocytes from one of three patients studied are pre- HIF-1 expressed N Y N 389 sented in Fig. 2H. Robust Akt phosphorylation on Thr308 p-S6K1 Thr NY Y and IRS-1 phosphorylation on Ser312 were seen between p-Akt Thr308 NYYorN days 1 and 3 postsurgery (Fig. 2H), whereas total Akt and p-Akt Ser473 NN Y IRS-1 expression remained constant throughout. S6K1 p-FoxO1/FoxO3a phosphorylation on Thr389 was detected on days 1–3post- Thr24/Thr32 NNYorN surgery, and this coincided with the presence of cleaved p-IRS-1 Ser312 NY Y a 792 AMPK , Raptor dephosphorylated on Ser ,andMMP9 Increased TLR4 N Y Y or N expression (13). Of note, as glucose homeostasis im- N, no; p, phosphorylated; Y, yes. proved by days 3 and 4, the expression of the signaling diabetes.diabetesjournals.org Zhang and Associates 2229

Figure 3—Insulin suppresses LPS-induced signaling in leukocytes in a time- and dose-dependent manner. A and B: Blood samples were treated with LPS (10 ng/mL) plus insulin (1 unit/mL) for the indicated time. The sample in panel B, lane 8, was treated with LPS alone (10 ng/mL) for 90 min and served as a positive control. C and D: Blood samples were untreated (lane 1), treated with LPS for 2 h (lane 2), or treated with insulin (1 unit/mL) for 1 h (lane 3) or 2 h (lane 4). The samples shown in lanes 5–9 were treated for 1 h with LPS (10 ng/mL) in combination with insulin at the indicated concentration. p-S, phosphorylation on serine; p-T, phosphorylation on threonine; U, unit; UN, untreated.

establish that insulin can suppress leukocyte responses to addition, neutrophils and leukocytes isolated from pa- LPS in a time- and dose-dependent manner. tients after CPB surgery displayed similar signaling inter- mediates (13). In the current study, we first sought to LPS and Insulin-Induced Responses Seen in determine whether LPS might also induce, as seen in Leukocytes Are Reproduced in Neutrophils leukocytes (Fig. 2A), Akt phosphorylation on Thr308 and Neutrophils, lymphocytes, and monocytes constitute, re- IRS-1 phosphorylation on Ser312 in isolated neutrophils spectively, ;50–65%, ;30%, and ;5% of all leukocytes (Fig. 4A). We then asked whether neutrophils respond to in human blood. We showed that the temporal changes in insulin. As shown in Fig. 4A, insulin triggered Akt phos- MMP9, HIF-1a, and AMPKa expression seen in leukocytes phorylation on Ser473 and Thr308 and S6K1 phosphoryla- treated with LPS are also seen in neutrophils (13). In tion on Thr389 in neutrophils and suppressed LPS-induced 2230 Insulin and TLR4 Signaling in Human Leukocytes Diabetes Volume 65, August 2016

Figure 4—Insulin suppresses LPS-induced signaling in neutrophils but not mononucleated cells in a dose-dependent manner. After isolation, neutrophils (A) and mononucleated cells (B) were treated with LPS, insulin, or a combination of the two as described in Fig. 2C and D. All the samples shown had equal protein amounts. In panel B, lane 10 was loaded with lysates of leukocytes treated with LPS or insulin (for the blots probed for p-S473 Akt and p-T24/32 FoxO1/FoxO3a) and served as a positive control. p-S, phosphorylation on serine; p-T, phosphorylation on threonine; U, unit; UN, untreated.

signaling in a dose-dependent manner. Both LPS and in- AS1842856 (Fig. 5B). These data identify FoxO1 as a sulin failed to trigger similar responses in mononucleated critical upstream regulator of TLR4 signaling in human blood cells (Fig. 4B). leukocytes. The residual phosphorylation of IRS-1 on Ser312 detected in leukocytes treated with AS1842856, LPS and Insulin Signaling to mTORC1 and mTORC2 in Torin, or PP242 (Fig. 4) suggests that an additional FoxO1/ Leukocytes FoxO3a-independent regulatory mechanism contributes to We further explored the LPS and insulin signaling pathways thisphosphorylationevent.Theworkingmodelsarepre- in leukocytes. mTORC1 phosphorylates ribosomal S6K1 on sented in Fig. 5C. Thr389 (16,17). The detection of S6K1 phosphorylated on Thr389 in leukocytes treated with either LPS or insulin (Fig. Insulin Regulates Leukocyte TLR4 Expression and 2) suggested that both pathways activate mTORC1. Treat- Sensitivity to LPS ment of leukocytes with the mTORC1 inhibitor rapamycin TLR4 expression was markedly increased in the majority before treatment with LPS blocked the phosphorylation of of the type 2 diabetes samples studied and in leukocytes S6K on Thr389 and the phosphorylation of Akt on Thr308 treated with LPS in vitro (Figs. 1–3). Leukocytes from (Fig. 5A, lane 3). Torin and PP242, which inhibit both patients after CPB also exhibited a significant increase mTORC1 and mTORC2, reproduced the effect of rapamycin in TLR4 expression postsurgery (Fig. 6A). We next asked (42). The phosphorylation of IRS-1 on Ser312 was par- whether the increase in TLR4 expression in leukocytes tially reduced, but TLR4 expression was not affected by alters their sensitivity to LPS by using patient leukocytes rapamycin, Torin, or PP242. We then examined the before and after CPB surgery. Leukocyte activation trig- effect of these inhibitors on leukocyte responses to insulin. gers IkBa degradation, enabling translocation of nuclear Rapamycin, Torin, and PP242 suppressed the phosphoryla- factor-kB to the nucleus (9). We used IkBa degrada- tion of Akt on Thr308 and the phosphorylation of S6K tion as a proxy for assessing TLR4 sensitivity. Leukocytes on Thr389. Akt phosphorylation on Ser473,inducedby obtained either before or after CPB surgery showed IkBa insulin, was inhibited by both Torin and PP242, whereas degradation within 5 min of LPS treatment in vitro (Fig. rapamycin had no effect. These studies show that the 6B). However, although IkBa degradation in leukocytes increases in TLR4 expression and IRS-1 phosphoryla- from three patients required LPS at 10 ng/mL presurgery, tion on Ser312 induced by LPS are partially mTORC1/ a concentration of 2–5 ng/mL was sufficient to trigger mTORC2 independent. IkBa degradation in leukocytes from the same patients To assess the role of FoxO1 relative to leukocyte TLR4 postsurgery (Fig. 6C). These data indicate that the in- signaling, we next used a FoxO1 inhibitor, AS1842856 crease in TLR4 expression after CPB surgery correlates (43). LPS failed to induce MMP9, AMPKa cleavage, or the with a two- to fivefold increase in leukocyte sensitivity increase in TLR4 expression in leukocytes pretreated with to LPS. Consistent with the observation that FoxO1 diabetes.diabetesjournals.org Zhang and Associates 2231

Figure 5—LPS and insulin signaling pathways in leukocytes. A: Blood samples were not treated (2) or pretreated for 1 h (+) with rapamycin, Torin, or PP242. The samples were next treated for 90 min with LPS (lanes 2–5) or for 60 min with insulin (lanes 6–9). B: Blood samples were not treated (2) or treated for 1 h (+) with AS1842856 (AS), a FoxO1 inhibitor. The samples were then untreated or treated with LPS for the indicated time. Leukocytes were isolated, lysed, and analyzed by immunoblotting. C: The working models for insulin and TLR4 signaling in leukocytes. Insulin activates mTORC1, triggering phosphorylation of S6K1 on Thr389 and Akt on Thr308. Insulin also activates mTORC2. This leads to phosphorylation of Akt on Ser473 and FoxO1/FoxO3a on Thr24/Thr32. TLR4 activation triggers an increase in MMP9 expression through FoxO proteins (51). MMP9 cleaves AMPKa. AMPK inactivation enables Raptor dephosphorylation on Ser792 and mTORC1 activation. mTORC1 regulates the phosphorylation of S6K1 on Thr389 and Akt on Thr308. FoxO1/FoxO3a also upregulates the expression of TLR4. TLR4 triggers IRS-1 phosphorylation on Ser312 by a mechanism that is only partially regulated by FoxO1/FoxO3a. When insulin and LPS are combined, insulin-dependent activation of mTORC2, Akt phosphorylation on Ser473, and FoxO1/FoxO3a phosphorylation on Thr24/Thr32 suppresses TLR4 signaling in a dose- and time-dependent manner. P, phosphorylation; p-S, phosphorylation on serine; p-T, phosphorylation on threonine; Rapa, rapamycin.

regulates the expression of TRL4 in Raw264.7 cells (28), suppresses TLR4 signaling but also diminishes leukocyte the FoxO1 inhibitor AS1842856 suppressed TLR4 responsiveness to LPS. expression in leukocytes treated with LPS (Fig. 5B). Fur- thermore, insulin suppressed the expression of TLR4 in a DISCUSSION time- and dose-dependent fashion (Fig. 6D and E). To- To gain a better understanding of leukocyte signaling in gether, these data demonstrate that insulin not only the context of insulin resistance, we studied leukocytes 2232 Insulin and TLR4 Signaling in Human Leukocytes Diabetes Volume 65, August 2016

Figure 6—An increase in leukocyte TLR4 expression enhances leukocyte responsiveness to LPS. A: Blood samples were obtained from three CPB patients presurgery, in the recovery room, and on day 1 postsurgery. B: Blood samples obtained from a CPB patient presurgery and on day 1 postsurgery were treated with LPS (10 ng/mL) for the indicated time. C: Blood samples obtained from three CPB patients presurgery and on day 1 postsurgery were treated for 5 min with LPS at the indicated concentration. D: Blood samples were treated with LPS plus insulin for the indicated time. E: Blood samples were untreated (lane 1), treated with LPS for 2 h (lane 2), or treated with insulin (1 unit/mL) for 1 h (lane 3) or 2 h (lane 4). The samples shown in lanes 5–9werefirst treated for 1 h with insulin at the indicated concentration and then for 1 h with LPS (10 ng/mL). Leukocytes were isolated, lysed, and analyzed by immunoblotting. D1, day 1; P, presurgery; R, recovery room; U, unit; UN, untreated. from two patient cohorts: patients with type 2 diabetes an increase in TLR4, and AMPKa cleavage, all of which are (chronic insulin resistance) and patients undergoing CPB induced by LPS, indicating that FoxO1 is an essential medi- surgery (acute insulin resistance). In parallel studies, we ator of leukocyte TLR4 signaling. In line with this finding, analyzed responses of leukocytes and neutrophils treated others have shown that FoxO1 bound multiple enhancer- in vitro with LPS and/or insulin, in combination with like elements in the TLR4 gene and increased the expression well-characterized pharmacologic response modulators. The of TLR4 in LPS-treated Raw264.7 cells (28). When taken as a leukocytes from the cohort of 12 patients with type 2 whole, the current data provide compelling evidence that diabetes exhibited two patterns of signaling intermedi- pattern 1 reflects TLR4 signaling. Others reported that ates. Pattern 1 included MMP9, cleaved AMPKa, HIF-1a, the concentration of TLR4 ligands is higher in the blood and enhanced TLR4 expression. Leukocytes with pattern of patients with type 2 diabetes than in patients without 1 expressed Akt phosphorylated on Thr308 but not Ser473. diabetes or healthy control subjects (34,44). Persistent Three additional signaling intermediates, Raptor dephos- presence of TLR4 ligands could sustain chronic TLR4 sig- phorylated on Ser792, S6K1 phosphorylated on Thr389, naling activation in leukocytes from patients with type 2 and IRS-1 phosphorylated on Ser312, were detected in diabetes. leukocytes with pattern 1 and pattern 2. Patient leuko- To our knowledge, the current data are the first to cytes obtained on days 1–3 after CPB surgery also exhibited demonstrate that insulin activates a canonical signaling pattern 1. We showed that plasma post-CPB surgery con- pathway that includes mTORC2, Akt phosphorylated on tained a factor that induced TLR4 activation in healthy Thr308 and Ser473, S6K1 phosphorylated on Thr389, donor leukocytes (13). Furthermore, the entire repertoire and FoxO1/FoxO3a phosphorylated on Thr24/Thr32 of signaling mediators associated with pattern 1 was in human leukocytes and neutrophils. The detection of Akt reproduced in leukocytes treated with LPS in vitro. The phosphorylated on Ser473 in leukocytes from patients with FoxO1 inhibitor AS1842856 suppressed MMP9 expression, type 2 diabetes receiving insulin suggests that leukocytes diabetes.diabetesjournals.org Zhang and Associates 2233 respond to insulin in vivo. Analyses of CPB patient leukocytes that insulin was able to suppress a subset but not the led to a similar conclusion. Acute insulin resistance develops entire panel of inflammatory signaling mediators in these in patients after CPB, requiring insulin infusion for several patients’ leukocytes. days postsurgery. While being treated with insulin on days In summary, the current data provide novel insight 1–3 post-CPB surgery, patient leukocytes exhibited the entire into chronic human leukocyte inflammatory responses repertoire of signaling mediators associated with pattern 1. and compelling evidence that TLR4 signaling is activated By days 3 and 4, as patients regained insulin sensitivity, in leukocytes from patients with type 2 diabetes. Studies pattern 1 was replaced by Akt phosphorylated on Ser473 have suggested that food intake, particularly when rich in and FoxO1/FoxO3a phosphorylated on Thr24/Thr32.Thus, fat, might trigger an increase in TLR4 ligand abundance in both the in vitro and the in vivo data suggest that insulin blood (50). If such physiologic responses exist, leukocytes triggers Akt phosphorylation on Ser473 and FoxO1/FoxO3a might be exposed frequently and simultaneously to insu- phosphorylation on Thr24/Thr32 in leukocytes and neutro- lin and TLR4 ligands. The dynamic and highly competitive phils and that in the presence of these signaling mediators, interaction between insulin and TLR4 signaling pathways TLR4 signals are suppressed. The data also suggest that in- in leukocytes could serve an important homeostatic func- sulin fails to suppress TLR4 signals in patients with insulin tion, namely, to prevent unwanted leukocyte activation. resistance. A number of studies have suggested that insulin Chronic exposure to low-dose TLR4 ligands could contrib- has anti-inflammatory effects, which is consistent with the ute to the derailment of such physiologic responses, ulti- current findings. Of note, the administration of insulin to mately leading to insulin resistance and type 2 diabetes subjects challenged with LPS suppressed inflammatory re- because leukocyte activation increases reliance on sys- sponses in vivo (45). Prior studies that used mouse embry- temic glucose availability. Under conditions of chronic onic fibroblasts and bone marrow–derived dendritic cells low-grade inflammation, leukocytes most likely contribute showed that TLR4 not only activates FoxO1 but also triggers to systemic insulin resistance to ensure the glucose avail- FoxO1 inactivation through mTORC2 and Akt phosphory- ability essential for their survival. lation on Ser473 (46).Inmarkedcontrast,TLR4signalingin leukocytes is limited to FoxO1 activation, indicating that FoxO1 inactivation in the context of TLR4 signaling requires Funding. The study was supported in part by funds from the New Jersey Health Foundation (to L.Y.L. and B.H.). input from another signaling mediator, such as insulin. Duality of Interest. No potential conflicts of interest relevant to this article The current study also established that TLR4 signaling were reported. fi 308 induces speci c phosphorylation of Akt on Thr . Al- Author Contributions. Z.Z. contributed to the study design, performance 308 though phosphorylation of Akt on residue Thr or of experiments, and preparation of the manuscript. L.F.A. and L.Y.L. contributed 473 Ser enhances the catalytic activity of Akt by $10- to the patient recruitment and data analysis and interpretation. S.M.C. and M.A.M. fold, it is generally believed that the catalytic activity of contributed to the patient recruitment and data analysis. M.J.B. contributed to the data Akt is regulated synergistically and requires both Thr308 interpretation. B.H. contributed to the study design, data analysis and interpretation, and Ser473 phosphorylation for optimal Akt activation and preparation of the manuscript. B.H. is the guarantor of this work and, as such, (30,37). Akt phosphorylation on Ser473 is assumed to rep- had full access to all the data in the study and takes responsibility for the integrity of resent Akt activation. However, in non–small cell lung the data and accuracy of the data analyses. 308 cancer cells, the phosphorylation of Akt at Thr rather References 473 than at Ser showed a better correlation with Akt acti- 1. Black PR, Brooks DC, Bessey PQ, Wolfe RR, Wilmore DW. Mechanisms of vation (47). Studies in human platelets have similarly insulin resistance following injury. Ann Surg 1982;196:420–435 473 demonstrated that the phosphorylation of Akt on Ser 2. Duncan BB, Schmidt MI, Pankow JS, et al.; Atherosclerosis Risk in Com- was dispensable for Akt phosphorylation on Thr308 and munities Study. Low-grade systemic inflammation and the development of type 2 Akt1 activity (48). Although in many instances PI3K acti- diabetes: the atherosclerosis risk in communities study. Diabetes 2003;52:1799– vates Akt, which then activates mTORC1, the current data 1805 suggest that mTORC1 regulates the phosphorylation of 3. Gregor MF, Hotamisligil GS. Inflammatory mechanisms in obesity. Annu Rev Akt on Thr308 in leukocytes treated with LPS or insulin. Immunol 2011;29:415–445 The role of Akt phosphorylated on Thr308 in leukocytes is 4. Hoshino K, Takeuchi O, Kawai T, et al. Cutting edge: Toll-like receptor 4 fi unclear. (TLR4)-de cient mice are hyporesponsive to lipopolysaccharide: evidence for – Leukocytes from 4 of 12 patients with type 2 diabetes TLR4 as the Lps gene product. J Immunol 1999;162:3749 3752 5. Agwunobi AO, Reid C, Maycock P, Little RA, Carlson GL. Insulin resistance exhibited pattern 2, which included Akt phosphorylated 473 a a and substrate utilization in human endotoxemia. J Clin Endocrinol Metab 2000; on Ser but not MMP9, HIF-1 , or cleaved AMPK . The – 24 32 85:3770 3778 phosphorylation of FoxO1/FoxO3a on Thr /Thr in 6. Lowry SF. Human endotoxemia: a model for mechanistic insight and these samples was either trace or absent. Given that FoxO1 therapeutic targeting. Shock 2005;24(Suppl. 1):94–100 phosphorylation/inactivation was impaired in liver, adi- 7. Cani PD, Amar J, Iglesias MA, et al. Metabolic endotoxemia initiates obesity pose, and muscle from insulin resistant mice (49), the and insulin resistance. Diabetes 2007;56:1761–1772 current data suggest that this might also apply to the 8. Saberi M, Woods NB, de Luca C, et al. Hematopoietic cell-specific deletion leukocytes of patients with type 2 diabetes. Because of toll-like receptor 4 ameliorates hepatic and adipose tissue insulin resistance in the four patients were treated with insulin, it is plausible high-fat-fed mice. Cell Metab 2009;10:419–429 2234 Insulin and TLR4 Signaling in Human Leukocytes Diabetes Volume 65, August 2016

9. Yamamoto Y, Gaynor RB. IkappaB : key regulators of the NF-kappaB 32. Brunet A, Bonni A, Zigmond MJ, et al. Akt promotes cell survival by phos- pathway. Trends Biochem Sci 2004;29:72–79 phorylating and inhibiting a Forkhead transcription factor. Cell 1999;96:857–868 10. Arkan MC, Hevener AL, Greten FR, et al. IKK-beta links inflammation to 33. Su D, Coudriet GM, Hyun Kim D, et al. FoxO1 links insulin resistance to obesity-induced insulin resistance. Nat Med 2005;11:191–198 proinflammatory cytokine IL-1beta production in macrophages. Diabetes 2009; 11. Zhang Z, Lowry SF, Guarente L, Haimovich B. Roles of SIRT1 in the acute 58:2624–2633 and restorative phases following induction of inflammation. J Biol Chem 2010; 34. Dasu MR, Devaraj S, Park S, Jialal I. Increased toll-like receptor (TLR) 285:41391–41401 activation and TLR ligands in recently diagnosed type 2 diabetic subjects. Di- 12. Haimovich B, Zhang Z, Calvano JE, et al. Cellular metabolic regulators: novel abetes Care 2010;33:861–868 fl – indicators of low-grade in ammation in humans. Ann Surg 2014;259:999 1006 35. Reyna SM, Ghosh S, Tantiwong P, et al. Elevated toll-like receptor 4 ex- a 13. Zhang Z, Amorosa LF, Coyle SM, et al. Proteolytic cleavage of AMPK and pression and signaling in muscle from insulin-resistant subjects. Diabetes 2008; intracellular MMP9 expression are both required for TLR4-mediated mTORC1 57:2595–2602 activation and HIF-1a expression in leukocytes. J Immunol 2015;195:2452– 36. Boucher J, Kleinridders A, Kahn CR. Insulin receptor signaling in normal and 2460 insulin-resistant states. Cold Spring Harb Perspect Biol 2014;6:6 14. Towler MC, Hardie DG. AMP-activated protein kinase in metabolic control 37. Alessi DR, Andjelkovic M, Caudwell B, et al. Mechanism of activation of and insulin signaling. Circ Res 2007;100:328–341 by insulin and IGF-1. EMBO J 1996;15:6541–6551 15. Laplante M, Sabatini DM. mTOR signaling in growth control and disease. 38. Friedman JE, Ishizuka T, Liu S, et al. Reduced insulin receptor signaling in Cell 2012;149:274–293 the obese spontaneously hypertensive Koletsky rat. Am J Physiol 1997;273: 16. Pullen N, Thomas G. The modular phosphorylation and activation of p70s6k. E1014–E1023 FEBS Lett 1997;410:78–82 39. Hirosumi J, Tuncman G, Chang L, et al. A central role for JNK in obesity and 17. Pearson RB, Dennis PB, Han JW, et al. The principal target of rapamycin- – induced p70s6k inactivation is a novel phosphorylation site within a conserved insulin resistance. Nature 2002;420:333 336 hydrophobic domain. EMBO J 1995;14:5279–5287 40. Doenst T, Wijeysundera D, Karkouti K, Zechner C, Maganti M, Rao V, 18. Cramer T, Yamanishi Y, Clausen BE, et al. HIF-1alpha is essential for Borger MA: Hyperglycemia during cardiopulmonary bypass is an independent risk myeloid cell-mediated inflammation. Cell 2003;112:645–657 factor for mortality in patients undergoing cardiac surgery. J Thorac Cardiovasc 19. Rius J, Guma M, Schachtrup C, et al. NF-kappaB links innate immunity to Surg 2005;130:1144 the hypoxic response through transcriptional regulation of HIF-1alpha. Nature 41. Harte AL, Varma MC, Tripathi G, et al. High fat intake leads to acute 2008;453:807–811 postprandial exposure to circulating endotoxin in type 2 diabetic subjects. Di- 20. Jacinto E, Loewith R, Schmidt A, et al. Mammalian TOR complex 2 controls abetes Care 2012;35:375–382 the actin and is rapamycin insensitive. Nat Cell Biol 2004;6:1122– 42. Thoreen CC, Kang SA, Chang JW, et al. An ATP-competitive mammalian 1128 target of rapamycin inhibitor reveals rapamycin-resistant functions of mTORC1. 21. Sarbassov DD, Guertin DA, Ali SM, Sabatini DM. Phosphorylation and J Biol Chem 2009;284:8023–8032 regulation of Akt/PKB by the rictor-mTOR complex. Science 2005;307:1098– 43. Nagashima T, Shigematsu N, Maruki R, et al. Discovery of novel forkhead 1101 box O1 inhibitors for treating type 2 diabetes: improvement of fasting glycemia in 22. Tzivion G, Dobson M, Ramakrishnan G. FoxO transcription factors; regula- diabetic db/db mice. Mol Pharmacol 2010;78:961–970 tion by AKT and 14-3-3 proteins. Biochim Biophys Acta 2011;1813:1938–1945 44. Pussinen PJ, Havulinna AS, Lehto M, Sundvall J, Salomaa V. Endotoxemia is 23. Biggs WH 3rd, Meisenhelder J, Hunter T, Cavenee WK, Arden KC. Protein associated with an increased risk of incident diabetes. Diabetes Care 2011;34: kinase B/Akt-mediated phosphorylation promotes nuclear exclusion of the 392–397 winged helix transcription factor FKHR1. Proc Natl Acad Sci U S A 1999;96: 45. Dandona P, Ghanim H, Bandyopadhyay A, et al. Insulin suppresses endotoxin- – 7421 7426 induced oxidative, nitrosative, and inflammatory stress in humans. Diabetes Care 24. Crossley LJ. Neutrophil activation by fMLP regulates FOXO (forkhead) 2010;33:2416–2423 transcription factors by multiple pathways, one of which includes the binding of 46. Brown J, Wang H, Suttles J, Graves DT, Martin M. Mammalian target of FOXO to the survival factor Mcl-1. J Leukoc Biol 2003;74:583–592 rapamycin complex 2 (mTORC2) negatively regulates Toll-like receptor 4-mediated 25. Jonsson H, Allen P, Peng SL. Inflammatory arthritis requires Foxo3a to inflammatory response via FoxO1. J Biol Chem 2011;286:44295–44305 prevent Fas ligand-induced neutrophil apoptosis. Nat Med 2005;11:666–671 47. Vincent EE, Elder DJ, Thomas EC, et al. Akt phosphorylation on Thr308 but 26. Tothova Z, Kollipara R, Huntly BJ, et al. FoxOs are critical mediators of not on Ser473 correlates with Akt protein kinase activity in human non-small cell hematopoietic stem cell resistance to physiologic oxidative stress. Cell 2007;128: lung cancer. Br J Cancer 2011;104:1755–1761 325–339 48. Moore SF, Hunter RW, Hers I. mTORC2 -mediated Akt 27. Peng SL. Foxo in the immune system. Oncogene 2008;27:2337–2344 (protein kinase B) serine 473 phosphorylation is not required for Akt1 activity in 28. Fan W, Morinaga H, Kim JJ, et al. FoxO1 regulates Tlr4 inflammatory pathway signalling in macrophages. EMBO J 2010;29:4223–4236 human platelets [published correction appears in J Biol Chem 2011;286:31062]. – 29. Walrand S, Guillet C, Boirie Y, Vasson MP. Insulin differentially regulates J Biol Chem 2011;286:24553 24560 monocyte and polymorphonuclear neutrophil functions in healthy young and el- 49. Guo S. Insulin signaling, resistance, and the metabolic syndrome: insights derly humans. J Clin Endocrinol Metab 2006;91:2738–2748 from mouse models into disease mechanisms. J Endocrinol 2014;220:T1–T23 30. Alessi DR, James SR, Downes CP, et al. Characterization of a 50. Erridge C, Attina T, Spickett CM, Webb DJ. A high-fat meal induces low- 3-phosphoinositide-dependent protein kinase which phosphorylates and grade endotoxemia: evidence of a novel mechanism of postprandial inflam- activates protein kinase Balpha. Curr Biol 1997;7:261–269 mation. Am J Clin Nutr 2007;86:1286–1292 31. Guertin DA, Stevens DM, Thoreen CC, et al. Ablation in mice of the mTORC 51. Li H, Liang J, Castrillon DH, DePinho RA, Olson EN, Liu ZP. FoxO4 regulates components raptor, rictor, or mLST8 reveals that mTORC2 is required for sig- tumor necrosis factor alpha-directed smooth muscle by activating naling to Akt-FOXO and PKCalpha, but not S6K1. Dev Cell 2006;11:859–871 matrix metalloproteinase 9 gene transcription. Mol Cell Biol 2007;27:2676–2686