Diabetes-Associated Myelopoiesis Drives Stem Cell Mobilopathy Through an OSM-P66shc Signaling Pathway

Page 1 of 40 Diabetes

Diabetes-associated myelopoiesis drives stem cell mobilopathy through an OSM-p66Shc signaling pathway

Mattia Albiero1,2*, Stefano Ciciliot1*, Serena Tedesco1,2, Lisa Menegazzo1, Marianna D’Anna1,2, Valentina Scattolini1,2, Roberta Cappellari1,2, Gaia Zuccolotto3,4, Antonio Rosato3,4, Andrea Cignarella2, Marco Giorgio5,6, Angelo Avogaro2, and Gian Paolo Fadini1,2

* The first two authors contributed equally

1 Veneto Institute of Molecular Medicine (VIMM), 35129 Padova, Italy 2 Department of Medicine – DIMED, University of Padova, 35128 Padova, Italy 3 Department of Surgery, Oncology and Gastroenterology, University of Padova, 35129 Padova, Italy 4 Istituto Oncologico Veneto IOV-IRCCS, 35128 Padova, Italy 5 European Institute of Oncology (IEO), 20139 Milan, Italy 6 Department of Biomedical Sciences, 35131 Padova, Italy

Corresponding author Gian Paolo Fadini Associate Professor of Endocrinology Department of Medicine, University of Padova Via Giustiniani 2, 35128 Padova, Italy Phone +39 049 8214318 Fax: +39 049 8212184 Email: [email protected] [email protected]

Running title: OSM-p66Shc regulates mobilopathy/myelopoiesis

Diabetes Publish Ahead of Print, published online April 1, 2019 Diabetes Page 2 of 40

Abstract

Diabetes impairs the mobilization of hematopoietic stem/progenitor cells (HSPCs) from the bone marrow (BM), which can worsen the outcomes of HSPC transplantation and of diabetic complications. In this study, we examined the oncostatin M (OSM) - p66Shc pathway as a mechanistic link between HSPC mobilopathy and excessive myelopoiesis. We found that streptozotocin (STZ)-induced diabetes in mice skewed hematopoiesis towards the myeloid lineage, via hematopoietic-intrinsic p66Shc. The overexpression of Osm resulting from myelopoiesis prevented HSPC mobilization after G-CSF. The intimate link between myelopoiesis and impaired HSPC mobilization after G-CSF was confirmed in human diabetes. Using cross-transplantation experiments, we found that deletion of p66Shc in the hematopoietic or non-hematopoietic system partially rescued defective HSPC mobilization in diabetes. Additionally, p66Shc mediated the diabetes-induced BM microvasculature remodeling. Ubiquitous or hematopoietic restricted Osm deletion phenocopied p66Shc deletion in preventing diabetes-associated myelopoiesis and mobilopathy. Mechanistically, we discovered that OSM couples myelopoiesis to mobilopathy by inducing Cxcl12 in BM stromal cells via non-mitochondrial p66Shc. Altogether, these data indicate that cell-autonomous activation of the OSM-p66Shc pathway leads to diabetes-associated myelopoiesis, whereas its transcellular hemato-stromal activation links myelopoiesis to mobilopathy. Targeting the OSM-p66Shc pathway is a novel strategy to disconnect mobilopathy from myelopoiesis and restore normal HSPC mobilization. Page 3 of 40 Diabetes

Diabetes is associated with low-grade inflammation, which contributes to chronic complications (1; 2). A skewed differentiation of common myeloid progenitors (CMP) translates hyperglycemia into production of pro-inflammatory cells (3). Such enhanced myelopoiesis propagates inflammation from the bone marrow (BM) to the adipose and the vasculature, leading to insulin resistance and atherosclerosis (3; 4). In parallel, mobilization of hematopoietic stem/progenitor cells (HSPCs) from the BM to peripheral blood (PB) after stimulation with granulocyte colony stimulation factor (G-CSF) is impaired in murine (5; 6) and human diabetes (7; 8), a condition termed mobilopathy (9). We herein hypothesize that myelopoesis and mobilopathy, described as two distinct pathological features of the diabetic BM, are instead mechanistically linked. Disentangling the processes linking myelopoiesis to mobilopathy has relevant clinical implications. First, pharmacologic mobilization of HSPCs is the gold standard for HSPC transplantation (10) and failure to collect robust numbers of HSPCs can delay engraftment, thereby worsening the outcome of diabetic patients undergoing transplantation (5). Second, reduction of circulating HSPCs in diabetic patients predicts the future development of micro- and macrovascular complications (11; 12). Glucose control effectively prevents myelopoiesis and partially rescues HSPC mobilization (3; 13), but many patients fail to achieve necessary glucose targets. Therefore, disconnecting mobilopathy from myelopoiesis can provide a direct therapeutic strategy to restore normal HSPC mobilization. Recent studies highlight that murine and human diabetes cause BM microvascular remodeling (14) and autonomic neuropathy (6; 15), both of which can affect HSPC traffic (16; 17). We previously found that BM denervation in diabetic mice accounts for impaired response to G-CSF and is mediated by p66Shc (6). Unlike p46 and p52, p66Shc functions both as an adaptor protein for membrane receptors and a redox enzyme. Upon phosphorylation at Ser36, p66Shc translocates to the mitochondrial intermembrane space where it catalyzes the production of hydrogen peroxide (18), contributing to processes linked to oxidative stress, including diabetic complications (19; 20). Besides sympathetic nervous system (SNS) activation, depletion of BM macrophages is a key event in the mobilization cascade induced by G-CSF, because macrophage paracrine activity sustains CXCL12 production (21). We have identified oncostatin M (OSM) as the macrophage-derived soluble factor that induces Cxcl12 expression in stromal cells, thereby antagonizing HSPC mobilization (22). OSM is a cytokine of the IL-6 family, which signals via MAP kinase and the JAK/STAT pathways, leading to pleiotropic functions including modulation of inflammation and bone formation (23; 24). In murine diabetes, excess BM macrophages result in persistent OSM signaling, inability to switch off CXCL12 levels after G-CSF, and impaired HSPC mobilization (22). Thus, OSM represents a candidate link between myelopoiesis and mobilopathy. In view of the similar benefits of p66Shc deletion and OSM inhibition on the diabetic stem cell mobilopathy (6; 22), we have hypothesized that the two pathways are mechanistically connected. In the current study, we therefore examined the interplay between OSM and p66Shc in determining the link between myelopoiesis and mobilopathy observed in experimental and human diabetes.

RESEARCH DESIGN AND METHODS Diabetes Page 4 of 40

Mice C57BL/6J wild type mice were purchased from Jackson Laboratories and established as a colony since 2001. p66Shc-/- mice were originally obtained from Pelicci’s laboratory (European Institute of Oncology, Milan, Italy), a colony was established at our facility in 2010, and mice have been backcrossed on the C57BL/6J background for >10 generations. Osm-/- mice on the C57BL/6J background were obtained from Glaxo Smith Kline (GSK, Stevenage, UK) and a colony was established since 2015. For all the experiments, we used sex- and age-matched animals. Assignment of mice to treatments or experimental groups was based on a computer generated random sequence. All animal studies were approved by the Animal Care and Use Committee of Venetian Institute of Molecular Medicine and by the Italian Health Ministry.

Humans Diabetic and non-diabetic individuals were recruited at the Division of Metabolic Diseases of the University Hospital of Padova. The protocols were approved by the local ethical committee and conducted in accordance with the Declaration of Helsinki as revised in 2000. Cross-sectional data on the association between myeloid bias and circulating HSPCs were derived from two previous studies that had been approved by the local ethics committee (6; 25). Total and differential white blood cell (WBC) count were determined in the same laboratory for both studies and CD34+ HSPC levels were quantified by flow cytometry relative to WBC. Details are given the in previous publications (6; 25). The study for G-CSF-induced mobilization was approved by the local ethics committee and is registered in ClinicalTrials.gov (NCT01102699). This was a prospective, parallel group study of direct BM stimulation with G-CSF in subjects with and without diabetes. Specific methods for quantifying blood cells and HSPCs were given in the previous publication (7). Informed consent was obtained from all participants.

Animal models Diabetes was induced in 2 months-old mice by a single intraperitoneal injection of 175 mg/kg streptozotocin (STZ). Blood glucose was measured using a FreeStyle glucometer (Abbott, IL, USA). HSPC mobilization was induced by s.c. injection of 200 g/kg/die G-CSF daily for 4 days. 3 months-old mice where treated with vehicle or carrier free recombinant mouse Oncostatin-M (495-MO/CF, R&D Systems, MN, USA) at 0.5 ug per injection every 6 hours for 48 hours before performing analysis. Total WBC count was performed using the CELL-DYN Emerald hematology analyzer (Abbott, IL, USA) on fresh EDTA-treated mouse blood.

MEF transduction MEFs were isolated from E13.5 p66Shc-/- mice after digestion with trypsin (Corning) and cultured with DMEM, 10% FBS. PINCO retroviral particles were produced from the amphotropic packaging cell line Phoenix. Cells were infected either with an empty vector, a vector encoding mouse full-length p66Shc, a vector encoding the mutants p66ShcS36A (S→A substitution at position 36) and p66ShcQQ (EE→QQ substitutions Page 5 of 40 Diabetes

at positions 132-133). P3 MEFs were infected with 3 rounds of infection with Polybrene Infection / Transfection Reagent (Sigma-Aldrich) followed by 96 hours of selection with 2 mg/ml of puromycin. Experiments were performed with p4 or p5 cells.

BM transplantation Three-months old recipient mice were treated with a myeloablative dose of total body irradiation of 10 Gy, split in two doses of 5 Gy 3 hours apart and followed by an intravenous injection of BM cells from donor mice (4107/each) isolated by flushing femurs and tibias with sterile ice-cold PBS.

Colony-forming unit assay 3104 BM cells were plated in 35 mm Petri dish containing 1 mL of methylcellulose-based medium MethoCult™ supplemented with 1% penicillin/streptomycin. After RBC lysis, 25 µL/well of PB were plated in 24-well plates containing 0.5 mL of MethoCult™ supplemented with 1% penicillin/streptomycin. Colony formation was scored after 10 days of culture. When required, murine recombinant S100A8/9 heterodimer (Biolegend) was mixed with the MethoCult™ medium.

Flow cytometry Flow cytometry was performed on BM cells or EDTA-treated peripheral blood. BM cells were isolated by flushing femurs, tibias with ice-cold MACS Separation Buffer (Miltenyi Biotec GmbH Gladbach, Germany) through a 40 M cell strainer. 100 L of PB or BM cells were incubated with antibodies for 15 minutes in the dark at room temperature. After RBC lysis, samples were resuspended in 200 L of PBS and data were acquired with a FACSCANTO (BD Biosciences) cytometer followed by analysis using FlowJo software (Tree Star Inc.).

BM derived mesenchymal stem cells (BM-MSC) Murine BM-MSC were isolated by flushing the BM of 3 months old mice and cultured in MEM-α containing 10% FBS, glutamine 2mM and penicillin-streptomycin. Passage 3 to 6 was used in all experiments. For gene expression analysis cells were treated with murine recombinant Oncostatin-M (R&D Systems) for 48 hours in serum free media.

Tissue processing Femur bones were fixed in 4% para-formaldehyde and decalcified. Bones were then washed with PBS, embedded in Killik cryostat medium (Bio-Optica, Milan, Italy) and frozen liquid nitrogen-cooled 2- methylbutane (Sigma-Aldrich). 10µm thick longitudinal femur sections were obtained with a Leica CM 1950 cryostat (Leica Biosystems S.r.l., Milan, Italy), placed on Superfrost Plus slides (J1800AMNZ, Gerhard Menzel GmbH, Braunschweig, Germany), and stored at -80°C. Diabetes Page 6 of 40

Protein phosphorylation by flow cytometry Confluent BM-MSC were treated with SCH772984 (4nM, Cayman Chemical) overnight or with Stattic (2.5 uM, Selleckchem) for 1 hour before adding recombinant OSM (R&D Systems) for 30 minutes in serum free media. Cells were detached by scraping and incubated with PE Mouse Anti-Stat3 (pY705) or PE Mouse Anti- ERK1/2 (pT202/pY204) (both from BD) in Perm Buffer III (BD Biosciences) according to the manufacture’s instruction.

Immunohistochemistry Femur sections were air-dried for 20 minutes, then incubated with blocking solution. Sections were incubated with primary antibody: anti laminin (1:50 for 4 days), anti tyrosine-hydroxylase (1:200 for 4 days), anti CD150 (1:50 for 3 days). Sections were then washed with PBS and incubated with secondary antibodies. Slides were mounted with an anti-fade aqueous mounting medium. Images were taken with a Leica DM5000B microscope, equipped with a DFC300 FX CCD camera, or with Cytell (GE Healthcare, Milan, Italy). Images were then processed with Fiji/ImageJ software (1.50, NIH, USA) or with Adobe Photoshop CS2 (9.0.2, Adobe Systems Incorporated, USA).

Morphometric measurements Vessel size and shape were measured using Fiji/ImageJ (1.50, NIH, USA). Briefly, random 500 µm2 fields from the epiphyseal and diaphyseal region (3 each, at least) of the samples were analyzed. Vessel structure was visualized by laminin staining, ROIs were manually outlined in Fiji/ImageJ. Area and shape parameters, such as circularity, were recorded. The bivariate distribution of area and circularity was visualized using the Bivariate Kernel Density Estimation (v1.0.9) in R (ver 3.1). BM innervation was determined by tyrosine- hydroxylase staining. Arterioles were identified in the whole femur section and arterioles diameter was measured with Fiji/ImageJ.

Molecular Biology RNA was isolated from flushed BM or cells by using QIAzol or with Total RNA Purification Micro Kit (Norgen Biotek) and quantified with a NanoDrop™ 2000 Spectrophotometer (Thermo Fisher Scientific, MA, USA). cDNA was synthesized using SensiFAST™ cDNA Synthesis Kit (Bioline, London, UK). qPCR was performed using SensiFAST™ SYBR® Lo-ROX Kit (Bioline, London, UK) via a QuantStudio 5 Real-Time PCR System (Thermo Fisher Scientific, MA, USA). A list of primers can be found in Table S1.

Statistical analysis Continuous data are expressed as mean±SEM, whereas categorical data are presented as percentage. Normality was checked using the Kolmogorov-Smirnov test, and non-normal data were log-transformed before analysis. Comparison between two or more groups was performed using the Student t test and ANOVA for normal variables or the Mann-Whitney U test and Kruskal-Wallis test for non-normal variables that could not be log- Page 7 of 40 Diabetes

transformed. Bonferroni adjustment was used to account for multiple testing. Linear correlations were checked using the Pearson’s r coefficient. Statistical analysis was accepted at p<0.05. Statistical analysis was performed using GraphPad Prism 6, Matlab and SPSS ver. 21.

RESULTS

Mobilopathy associates with myelopoiesis in experimental diabetes We first evaluated whether myelopoiesis and mobilopathy coexist in murine diabetes. We found that streptozotocin (STZ) induced diabetic mice had a ~2-fold expansion of PB granulocytes compared with non- diabetic mice (p<0.001; Figures 1A,B and S1), resulting in a strikingly 6-times higher granulocyte to lymphocyte (G/L) ratio (p<0.001; Figure 1C). The BM of diabetic mice contained higher numbers of granulocyte-monocyte progenitors (GMP) at the expense of CMP (Figure 1D). As a consequence, the clonogenic assay of BM cells showed an increased output of macrophage and granulocyte colonies from diabetic versus non-diabetic mice (2.2x and 2.8x, respectively; Figure 1E). The diabetic BM showed excess macrophages both in basal unstimulated conditions and after G-CSF stimulation (Figure 1F). These cells are known to produce OSM (22) and Osm gene expression in the BM of diabetic mice was indeed upregulated 7.7-times compared with that in non-diabetic mice (p=0.006; Figure 1G). Altogether, these data are consistent with exaggerated myelopoiesis and myeloid bias in diabetic mice. As the resulting overproduction of OSM can hamper mobilization (22), we evaluated if mobilopathy occurred in the same mice. Preliminary to this, we verified that the baseline PB level of HSPCs (Lin-c-Kit+Sca-1+; LKS cells) was non-significantly different in diabetic versus non-diabetic mice (Figure S2). After administering G- CSF for 4 days, HSPC level increased by 4.38-times in non-diabetic, but not in diabetic mice (Figure 1H). Such difference in the fold-change of PB-LKS cell level between diabetic and non-diabetic mice was highly significant (Figure 1I): 80% of non-diabetic mice versus 10% of diabetic mice achieved a mobilization response of at least 1.5-fold (Figure 1J). The colony-forming unit (CFU) assay from PB cells confirmed the absence of functional HSPC mobilization in diabetes (Figure S3A). The profound degree of mobilization impairment allowed us to use the G-CSF mobilization assay as a robust readout for mobilopathy in subsequent mouse experiments.

Mobilopathy associates with myelopoiesis in human diabetes We then checked whether myelopoiesis and mobilopathy occurred simultaneously in human diabetes. We first analyzed cross-sectional data of two studies wherein circulating WBC types and levels of CD34+ HSPCs were determined in the same sample (6; 25) (Table S2). In a pooled cohort of 344 subjects, diabetic patients (n=108; 74% type 2) displayed 25% lower levels of HSPCs and a 24% higher neutrophils to lymphocyte (N/L) ratio than non-diabetic individuals (Figure 2A). Higher N/L ratio and lower CD34+ HSPCs remained significantly associated with diabetes after adjusting for age, sex, BMI, hypertension, lipids, coronary artery disease, and Diabetes Page 8 of 40

8 retinopathy (Table S3). We also found a significant inverse correlation between the N/L ratio and the steady- state level of PB HSPCs (r =-0.28; Figure 2B). Considering glucose control as a continuous variable in the entire cohort, we found a significant inverse correlation between HbA1c and HSPC levels (r =-0.23; p<0.001; Figure 2C) and a direct correlation between HbA1c and the N/L ratio (r =0.21; p<0.001; Figure 2D), which persisted (both with p<0.01) after adjusting for the above mentioned confounders. These data suggest that myeloid bias is linked to a reduction in HSPC levels in human diabetes, possibly driven by hyperglycemia. Second, we evaluated whether myeloid bias was associated with mobilopathy by analyzing data from a previous prospective study wherein diabetic and non-diabetic patients (n=43) received low-dose G-CSF to test HSPC mobilization (7). The fold-change in G-CSF-induced HSPC levels versus baseline was significantly lower and the pre-G-CSF N/L ratio tended to be higher (p=0.06) in diabetic versus non-diabetic patients (Figure 2E). Remarkably, there was a significant inverse correlation between the N/L ratio and HSPC mobilization (r =-0.32; p=0.03; Figure 2F). These results cannot prove causality, because secondary analyses of previously collected cohort data can be subjected to bias and prone to false-positive signals; Nonetheless, we confirm that myelopoiesis and mobilopathy are associated in human as they are in murine diabetes.

Deletion of p66Shc prevents diabetes-associated myelopoiesis and mobilopathy Having shown that myelopoiesis and mobilopathy concur in murine and human diabetes, we explored the mechanisms driving their association. We first focused on p66Shc, which we previously showed to be responsible for diabetes-associated BM denervation and mobilopathy (6). BM p66Shc gene expression was >2-fold higher in diabetic versus control mice (p=0.003; Figure S4), consistent with prior data in mice and humans (26; 27). In the non-diabetic condition, we found no differences in WBC count and subtypes, G/L ratio, BM colonies and CMP/GMP progenitors, as well as BM macrophages between wild type (Wt) and p66Shc-/- mice (Figure 3A-G). However, in p66Shc-/- mice, diabetes did not increase PB granulocytes counts, and granulocytes were significantly lower than in Wt diabetic mice (Figures 3D-S5), as was the G/L ratio (Figure 3E). Furthermore, the diabetes-induced increase in myeloid CFUs (Figure 3F), CMP/GMP imbalance (Figure 3G), excess BM macrophages (Figure 3H) were completely prevented in p66Shc-/- mice. As a result, the surge in BM expression of Osm observed in diabetic Wt mice, which derives from BM macrophages (22), was absent in p66Shc-/- mice (Figure S6). The link between hyperglycemia and myelopoiesis has been attributed to the accumulation of advanced glycation end-products (AGEs) and the RAGE ligand S100A8/9 (3). Interestingly, RAGE signaling has been previously linked with downstream p66Shc activation (28). We found that S100A8/9 potentiated myelopoiesis in vitro by BM cells of Wt mice, evidenced by a 1.75-fold and a 2.0-fold increase in macrophage and granulocyte colonies, respectively (Figure 3I). However, such effect was completely abolished in p66Shc-/- BM cells (Figure 3J). In vivo treatment of non-diabetic Wt mice with S100A8/9 increased GMP and myeloid cell colonies, but such effect was not observed in p66Shc-/- mice (Figure 3K and S7). Together, these data indicate that p66Shc is required for the effects of hyperglycemia on myelopoiesis, possibly by preventing the activity of S100 proteins. Page 9 of 40 Diabetes

In agreement with our previous study (6), p66Shc deletion partially rescued HSPC mobilization in diabetic mice, as indicated by the 1.9x increase in LKS cell counts after G-CSF administration (Figure 3L). The CFU assay showed that mobilized HSPCs were functionally competent, as G-CSF increased PB hematopoietic colonies both in diabetic and non-diabetic p66Shc-/- mice (Figure S8). To dissect the hematopoietic-intrinsic and -extrinsic roles of p66Shc in regulating HSPC mobilization, we performed cross-transplantation experiments as illustrated in Figure 3M. We confirmed that BM transplantation did not impinge upon G-CSF responsiveness, as non-diabetic but not diabetic Wt mice transplanted with BM from Wt mice were able to mobilize functional HSPCs (Figure S3B,C). Wt mice transplanted with BM from p66Shc-/- mice (p66Shc-/- →Wt BMT) and rendered diabetic showed a partial restoration of HSPC increase after G-CSF stimulation. An almost identical, but partial, improvement was observed in diabetic p66Shc-/- mice transplanted with BM from Wt mice (Wt→p66Shc-/- BMT) and, in both cases, the fold-change in HSPC count was lower than that in ubiquitous p66Shc-/- diabetic mice (Figure 3N). In contrast, while diabetes increased granulocyte counts and the G/L ratio in Wt→p66Shc-/-, p66Shc-/-→Wt BMT mice were largely protected from elevation of the G/L ratio induced by diabetes (Figures 3O and S9). Knowing that myeloid-biased HSPCs reside in megakaryocytic (MK) niches (29), we analyzed MK density by staining BM section with anti-CD150. As previously noted by others (30), MKs were increased by 60% in diabetic compared with that in non-diabetic Wt mice (p=0.004). However, such an effect was not observed in p66Shc-/- mice (Figure S10), providing a further explanation for the protection of ubiquitous and hematopoietic-restricted p66Shc-/- mice from diabetes-induced myelopoiesis. Altogether, these data indicate that hematopoietic intrinsic and extrinsic mechanisms are responsible for the rescue of HSPC mobilization by p66Shc deletion, whereas prevention of myelopoiesis is hematopoietic cell- intrinsic.

p66Shc deletion improves diabetes-induced BM microvascular remodeling The partial restoration of G-CSF-induced HSPC mobilization in Wt→p66Shc-/- BMT diabetic mice suggested that deletion of p66Shc exerted protective effects on the BM stroma against hyperglycemic damage. We previously reported that p66Shc-/- mice were protected from BM sympathectomy induced by diabetes (6). We herein characterized microvascular BM remodeling in p66Shc-/- versus Wt diabetic and non-diabetic mice. In the peculiar BM microcirculation, the nutrient arterial system drains into sinusoids with capillary-size vessels and the irregular sinusoid lumen is occasionally compressed to capillary caliper (31). Using an unbiased auto- instructed procedure to score BM vessels (Figure 4A), we found that the total vascular density, numbers of arterioles and sinusoids were similar, but a significant 2.5-fold reduction in capillary-size structures was noted in diabetic versus non-diabetic Wt mice (p=0.006; Figure 4B). Remarkably, the density of BM capillary-size vessels was not reduced in p66Shc-/- diabetic versus non-diabetic mice and was higher in p66Shc-/- versus Wt diabetic mice (p=0.03). With a more detailed morphometric analysis of BM blood vessel distribution according to size and circularity, we found that diabetes also led to a reduction of larger irregular vessels, likely sinusoids, that was prevented by p66Shc deletion (Figure 4C). We then evaluated sympathetic innervation of the BM and Diabetes Page 10 of 40

10 found that tyrosine hydroxylase (Tyr-OH)+ sympathetic terminals were almost exclusively located close to arteriolar walls (Figure 4D). The percentage of innervated arterioles was significantly reduced by >2-fold in diabetic versus non-diabetic Wt mice, but not in p66Shc-/- mice (p<0.001; Figure 4E,F). Taken together, these findings indicate that deletion of p66Shc protects BM from microvascular remodeling, which can explain the partial rescue of HSPC mobilization in non-hematopoietic p66Shc-deleted mice.

Osm deletion phenocopies p66Shc deletion The hematopoietic cell-intrinsic mechanism whereby G-CSF exerts its mobilizing activity relies on suppression of BM macrophages (21). This pathway is independent from the stromal effect of G-CSF through nerve terminals, as mice sympathectomized by 6-OH dopamine showed a normal post-G-CSF suppression of BM macrophages despite being unable to mobilize HSPCs (Figure S11). This finding indicates that both hematopoietic and non-hematopoietic effects of G-CSF are required to yield a full HSPC mobilizing response and justifies the partial recovery of mobilization in Wt↔p66Shc-/- cross-transplanted animals. We previously demonstrated that antibody-mediated OSM neutralization allowed HSPC mobilization in diabetic mice by relieving the brake of CXCL12 produced by stromal cells (22). Consistent with the notion that OSM retains HSPCs in the BM niche, non-diabetic Osm-/- mice displayed higher HSPCs levels in unstimulated PB than Wt mice, which was not further increased by diabetes (Figure S2), HSPC mobilization in diabetic mice was partially rescued toward normal levels (2.2x) by genetic Osm deletion (p=0.01; Figure 5A). In addition, we observed a marked (~80%) reduction of BM macrophages in Osm-/- mice, both in the diabetic and non-diabetic condition, which was further suppressed by G-CSF (Figure 5B). This result suggests that OSM is not only a macrophage-derived paracrine factor, but it is also required for accumulation of BM macrophages in a paracrine-autocrine loop, as already seen by others in the heart (32). Indeed, diabetic Osm-/- mice had normal WBC counts (Figure 5C), but significantly lower levels of PB granulocytes compared with Wt diabetic mice (Figure S12) and the G/L ratio was restored towards the levels seen in non-diabetic mice (Figure 5D). To avoid the confounding factor of the absence of non-hematopoietic OSM in ubiquitous Osm-/- mice, we transplanted Wt mice with BM cells from Osm-/- mice and induced diabetes four weeks later (Figure 5E). After four weeks of diabetes, we tested HSPC mobilization after G-SCF treatment and found that hematopoietic-restricted Osm deletion rescued HPSC mobilization in diabetic mice towards normal levels (5.2x; p=0.03; Figure 5F). In addition, hematopoietic restricted knockout of Osm largely prevented the surge in granulocyte levels (Figures 5G and S12) and in G/L ratio induced by diabetes (Figure 5H). These data indicated that Osm deletion prevented hyperglycemia-induced myelopoiesis and mobilopathy in a hematopoietic cell-intrinsic manner. p66Shc is required for the stem cell retaining activity of OSM Since Osm-/- mice phenocopied p66Shc-/- mice in protecting against diabetes-induced myelopoiesis and mobilopathy, we hypothesized that OSM signaling required downstream p66Shc. OSM signals through heterodimers of OSMR and gp130, which elicit intracellular events leading to activation of the MAP kinase Page 11 of 40 Diabetes

and JAK-STAT3/5 pathways (33). Shc proteins cooperate with other adaptor proteins to transduce membrane receptor signals to MAPK. We previously found that both STAT3 and MAPK are required for Cxcl12 induction by OSM in BM stromal cells (22). Here, we hypothesized that the adaptor function of p66Shc is required for OSMR signal transduction to MAPK to induce Cxcl12 (model shown in Figure 6A). We found that the ability of OSM (30 ng/ml; ~1 µM) to stimulate Cxcl12 gene expression in BM-derived stromal cells (3.6x) was completely abolished in the absence of p66Shc (Figure 6B). The concentration was of OSM was chosen based on a previous dose-effect curve (22). In contrast to p46 and p52, p66Shc acts as both an adaptor protein for signaling cascades and a mitochondrial redox protein (18). To dissect whether mitochondrial function of p66Shc was required for OSM signaling and Cxcl12 induction, we transfected p66Shc-/- mouse embryonic fibroblasts (MEFs) with an empty vector or vectors encoding Wt p66, 36SerAla mutated p66 (which cannot translocate to mitochondria), or a catalytically inactive p66 (p66qq), and then treated MEFs with OSM: Cxcl12 induction by OSM in p66Shc-/- MEFs was rescued by expression of Wt, Ser36 mutated, or catalytically inactive p66Shc but not empty vector (Figure 6C), suggesting that the adaptor function, and not the mitochondrial function, of p66Shc was required for OSM signaling. In addition, we found that activation of ERK by OSM was abolished in BM-derived stromal cells from p66Shc-/- mice, while activation of STAT3 was unaffected (Figure 6D). This set of experiments is in line with the model depicted in Figure 6E, where p66Shc is recruited to OSMR and cooperates to activate the MAPK pathway, which, along with STAT3 activation via gp130, is needed to induce Cxcl12. Finally, to gather in vivo evidence that OSM signaling requires p66Shc, we treated mice with systemic OSM injections. Gene expression of Cxcl12 in the BM was significantly induced in Wt (4.5x; p=0.01) and in Osm-/- mice (7.9x; p=0.01), but not in p66Shc-/- mice (0.8x; p=0.77; Figure 7F). In parallel, the >2-fold higher levels of HSPCs observed in steady state basal condition in Osm-/- and in p66Shc-/- mice could be significantly suppressed by OSM injection in Osm-/- mice (p=0.04) but not in p66Shc-/- mice (Figure 6G). These data support the concept that regulation of HSPC trafficking by OSM via Cxcl12 requires p66Shc. Furthermore, in both Wt and Osm-/- mice, injection of OSM increased circulating granulocytes and reduced lymphocytes, thereby increasing 2-fold the G/L ratio, but this effect was absent in p66Shc-/- mice (Figure 6H), demonstrating that the effect of OSM on myelopoiesis is also dependent on p66Shc.

DISCUSSION

Defective HSPC mobilization in response to G-CSF is a consistent finding in experimental and human diabetes (8), but the underlying causes are incompletely understood. Our new data indicate that mobilopathy is intimately linked with myelopoiesis, an underlying driver of diabetes-associated inflammation. We herein Diabetes Page 12 of 40

12 show a novel OSM-p66Shc signaling pathway that is overactive in diabetes against HSPC mobilization via hematopoietic cell-intrinsic and extrinsic mechanisms (Figure 7). OSM is produced by myeloid inflammatory cells that are exceedingly present in the diabetic BM as part of the enhanced myelopoiesis induced by hyperglycemia (22). In turn, OSM signal transduction is activated in BM stromal cells via non-mitochondrial p66Shc to induce CXCL12 production, thereby retaining HSPCs in the BM niche (Figure 6). Notably, p66Shc also mediates microvascular remodeling of the diabetic BM that can jeopardize HSPC traffic (Figure 4). G- CSF exerts its mobilizing function by acting on hematopoietic cells and on the BM stroma (21). Remarkably, both hematopoietic and non-hematopoietic p66Shc deletion was needed to restore HSPC mobilization response to G-CSF in diabetes (Figure 3). Hematopoietic-restricted p66Shc deletion partially rescued mobilization in diabetic mice, along with inhibition of diabetes-induced MK expansion and myelopoiesis. Hyperglycemia- driven myelopoiesis arises from the skewed hematopoiesis stimulated by RAGE ligands (3), an effect that we found requires p66Shc. At the same time, hematopoietic Osm deletion prevented diabetes-induced myelopoiesis and the ability of OSM to stimulate myelopoiesis also required p66Shc. The striking similarities between Osm-/- and p66Shc-/- mice is indeed consistent with the notion that OSM couples myelopoiesis with mobilopathy via p66Shc. These data together indicate that activation of the OSM-p66Shc pathway drives diabetes-associated myelopoiesis in a cell-autonomous way, whereas its transcellular hemato-stromal activation links myelopoiesis to mobilopathy. Understanding HSPC mobilization unresponsiveness to G-CSF has clinical implications for patients undergoing HSPC collection for transplantation purposes (8). Thus, interrupting the OSM-p66Shc pathway provides a therapeutic strategy in conditions of poor HSPC mobilization, like diabetes. Diabetic stem cell mobilopathy precedes reduction of steady-state PB HSPCs in human diabetes (7), which in turn has been linked with worsening of diabetic complications (11; 12). Mobilopathy preceded reduction of PB HSPCs also in diabetic mice (Figures 1 and S2). However, since our new data reveal a causal link between myelopoiesis and mobilopathy, future studies should better clarify if diabetes outcomes are more related to alterations in blood inflammatory cells, circulating stem cells, or stem cell mobilization. In summary, we provide evidence that an overactive OSM-p66Shc pathway couples diabetes-associated myelopoiesis with HSPC mobilopathy. In addition to rescuing HSPC mobilization, tackling this pathway in the BM could provide a new avenue for the improvements of the diabetes-related inflammation and complication risk.

ACKNOWLEDGEMENTS

Author contribution Page 13 of 40 Diabetes

M.A., S.C., A.A. and G.P.F. designed the research. M.A, S.C., S.T., L.M., V.S., R.C., A.R., G.Z.; A.C. and M.G. performed the research. M.A., S.C.; S.T., V.S., and G.P.F. analyzed the data. M.A., S.C., A.A. and G.P.F. wrote the manuscript. All authors reviewed and edited the manuscript.

Guarantor statement G.P.F. is the guarantor of the work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Conflict of interest disclosure M.A., S.C. and G.P.F. are the inventors of a patent, hold by the University of Padova, on the use of pharmacologic oncostatin-M inhibition to allow stem cell mobilization.

Funding The study was supported by the following grants: European Foundation for the Study of Diabetes (EFSD)/Novartis 2013 grant to GPF; EFSD/Lilly 2016 grant to GPF; Ministry of University and Education PRIN grant 2015 to GPF; Italian Diabetes Society/Lilly grant 2017 to GPF; Fondazione Cariplo 2016-0922 to GPF.

References 1. Donath MY, Shoelson SE: Type 2 diabetes as an inflammatory disease. Nat Rev Immunol 2011;11:98-107 2. Devaraj S, Glaser N, Griffen S, Wang-Polagruto J, Miguelino E, Jialal I: Increased monocytic activity and biomarkers of inflammation in patients with type 1 diabetes. Diabetes 2006;55:774-779 3. Nagareddy PR, Murphy AJ, Stirzaker RA, Hu Y, Yu S, Miller RG, Ramkhelawon B, Distel E, Westerterp M, Huang LS, Schmidt AM, Orchard TJ, Fisher EA, Tall AR, Goldberg IJ: Hyperglycemia promotes myelopoiesis and impairs the resolution of atherosclerosis. Cell Metab 2013;17:695-708 4. Nagareddy PR, Kraakman M, Masters SL, Stirzaker RA, Gorman DJ, Grant RW, Dragoljevic D, Hong ES, Abdel-Latif A, Smyth SS, Choi SH, Korner J, Bornfeldt KE, Fisher EA, Dixit VD, Tall AR, Goldberg IJ, Murphy AJ: Adipose tissue macrophages promote myelopoiesis and monocytosis in obesity. Cell Metab 2014;19:821-835 5. Ferraro F, Lymperi S, Mendez-Ferrer S, Saez B, Spencer JA, Yeap BY, Masselli E, Graiani G, Prezioso L, Rizzini EL, Mangoni M, Rizzoli V, Sykes SM, Lin CP, Frenette PS, Quaini F, Scadden DT: Diabetes impairs hematopoietic stem cell mobilization by altering niche function. Sci Transl Med 2011;3:104ra101 6. Albiero M, Poncina N, Tjwa M, Ciciliot S, Menegazzo L, Ceolotto G, Vigili de Kreutzenberg S, Moura R, Giorgio M, Pelicci P, Avogaro A, Fadini GP: Diabetes causes bone marrow autonomic neuropathy and impairs stem cell mobilization via dysregulated p66Shc and Sirt1. Diabetes 2014;63:1353-1365 7. Fadini GP, Albiero M, Vigili de Kreutzenberg S, Boscaro E, Cappellari R, Marescotti M, Poncina N, Agostini C, Avogaro A: Diabetes impairs stem cell and proangiogenic cell mobilization in humans. Diabetes Care 2013;36:943-949 8. Fadini GP, DiPersio JF: Diabetes mellitus as a poor mobilizer condition. Blood Rev 2018;32:184-191 9. DiPersio JF: Diabetic stem-cell "mobilopathy". N Engl J Med 2011;365:2536-2538 Diabetes Page 14 of 40

10. Schmitz N, Linch DC, Dreger P, Goldstone AH, Boogaerts MA, Ferrant A, Demuynck HM, Link H, Zander A, Barge A: Randomised trial of filgrastim-mobilised peripheral blood progenitor cell transplantation versus autologous bone-marrow transplantation in lymphoma patients. Lancet 1996;347:353-357 11. Rigato M, Bittante C, Albiero M, Avogaro A, Fadini GP: Circulating Progenitor Cell Count Predicts Microvascular Outcomes in Type 2 Diabetic Patients. J Clin Endocrinol Metab 2015;100:2666-2672 12. Fadini GP, Rigato M, Cappellari R, Bonora BM, Avogaro A: Long-term Prediction of Cardiovascular Outcomes by Circulating CD34+ and CD34+CD133+ Stem Cells in Patients With Type 2 Diabetes. Diabetes Care 2017;40:125-131 13. Fadini GP, Sartore S, Schiavon M, Albiero M, Baesso I, Cabrelle A, Agostini C, Avogaro A: Diabetes impairs progenitor cell mobilisation after hindlimb ischaemia-reperfusion injury in rats. Diabetologia 2006;49:3075-3084 14. Spinetti G, Cordella D, Fortunato O, Sangalli E, Losa S, Gotti A, Carnelli F, Rosa F, Riboldi S, Sessa F, Avolio E, Beltrami AP, Emanueli C, Madeddu P: Global remodeling of the vascular stem cell niche in bone marrow of diabetic patients: implication of the microRNA-155/FOXO3a signaling pathway. Circ Res 2013;112:510-522 15. Dang Z, Maselli D, Spinetti G, Sangalli E, Carnelli F, Rosa F, Seganfreddo E, Canal F, Furlan A, Paccagnella A, Paiola E, Lorusso B, Specchia C, Albiero M, Cappellari R, Avogaro A, Falco A, Quaini F, Ou K, Rodriguez-Arabaolaza I, Emanueli C, Sambataro M, Fadini GP, Madeddu P: Sensory neuropathy hampers nociception-mediated bone marrow stem cell release in mice and patients with diabetes. Diabetologia 2015;58:2653-2662 16. Itkin T, Gur-Cohen S, Spencer JA, Schajnovitz A, Ramasamy SK, Kusumbe AP, Ledergor G, Jung Y, Milo I, Poulos MG, Kalinkovich A, Ludin A, Kollet O, Shakhar G, Butler JM, Rafii S, Adams RH, Scadden DT, Lin CP, Lapidot T: Distinct bone marrow blood vessels differentially regulate haematopoiesis. Nature 2016;532:323-328 17. Katayama Y, Battista M, Kao WM, Hidalgo A, Peired AJ, Thomas SA, Frenette PS: Signals from the sympathetic nervous system regulate hematopoietic stem cell egress from bone marrow. Cell 2006;124:407-421 18. Giorgio M, Migliaccio E, Orsini F, Paolucci D, Moroni M, Contursi C, Pelliccia G, Luzi L, Minucci S, Marcaccio M, Pinton P, Rizzuto R, Bernardi P, Paolucci F, Pelicci PG: Electron transfer between cytochrome c and p66Shc generates reactive oxygen species that trigger mitochondrial apoptosis. Cell 2005;122:221-233 19. Cosentino F, Francia P, Camici GG, Pelicci PG, Luscher TF, Volpe M: Final common molecular pathways of aging and cardiovascular disease: role of the p66Shc protein. Arterioscler Thromb Vasc Biol 2008;28:622-628 20. Vono R, Fuoco C, Testa S, Pirro S, Maselli D, Ferland McCollough D, Sangalli E, Pintus G, Giordo R, Finzi G, Sessa F, Cardani R, Gotti A, Losa S, Cesareni G, Rizzi R, Bearzi C, Cannata S, Spinetti G, Gargioli C, Madeddu P: Activation of the Pro-Oxidant PKCbetaII-p66Shc Signaling Pathway Contributes to Pericyte Dysfunction in Skeletal Muscles of Patients With Diabetes With Critical Limb Ischemia. Diabetes 2016;65:3691-3704 21. Chow A, Lucas D, Hidalgo A, Mendez-Ferrer S, Hashimoto D, Scheiermann C, Battista M, Leboeuf M, Prophete C, van Rooijen N, Tanaka M, Merad M, Frenette PS: Bone marrow CD169+ macrophages promote the retention of hematopoietic stem and progenitor cells in the mesenchymal stem cell niche. J Exp Med 2011;208:261-271 22. Albiero M, Poncina N, Ciciliot S, Cappellari R, Menegazzo L, Ferraro F, Bolego C, Cignarella A, Avogaro A, Fadini GP: Bone Marrow Macrophages Contribute to Diabetic Stem Cell Mobilopathy by Producing Oncostatin M. Diabetes 2015;64:2957-2968 23. West NR, Hegazy AN, Owens BMJ, Bullers SJ, Linggi B, Buonocore S, Coccia M, Gortz D, This S, Stockenhuber K, Pott J, Friedrich M, Ryzhakov G, Baribaud F, Brodmerkel C, Cieluch C, Rahman N, Muller-Newen G, Owens RJ, Kuhl AA, Page 15 of 40 Diabetes

Maloy KJ, Plevy SE, Keshav S, Travis SPL, Powrie F: Oncostatin M drives intestinal inflammation and predicts response to tumor necrosis factor-neutralizing therapy in patients with inflammatory bowel disease. Nat Med 2017;23:579-589 24. Guihard P, Danger Y, Brounais B, David E, Brion R, Delecrin J, Richards CD, Chevalier S, Redini F, Heymann D, Gascan H, Blanchard F: Induction of osteogenesis in mesenchymal stem cells by activated monocytes/macrophages depends on oncostatin M signaling. Stem Cells 2012;30:762-772 25. Fadini GP, Bonora BM, Marcuzzo G, Marescotti MC, Cappellari R, Pantano G, Sanzari MC, Duran X, Vendrell J, Plebani M, Avogaro A: Circulating Stem Cells Associate With Adiposity and Future Metabolic Deterioration in Healthy Subjects. J Clin Endocrinol Metab 2015;100:4570-4578 26. Costantino S, Paneni F, Mitchell K, Mohammed SA, Hussain S, Gkolfos C, Berrino L, Volpe M, Schwarzwald C, Luscher TF, Cosentino F: Hyperglycaemia-induced epigenetic changes drive persistent cardiac dysfunction via the adaptor p66(Shc). Int J Cardiol 2018;268:179-186 27. Pagnin E, Fadini G, de Toni R, Tiengo A, Calo L, Avogaro A: Diabetes induces p66shc gene expression in human peripheral blood mononuclear cells: relationship to oxidative stress. J Clin Endocrinol Metab 2005;90:1130-1136 28. Cai W, He JC, Zhu L, Chen X, Striker GE, Vlassara H: AGE-receptor-1 counteracts cellular oxidant stress induced by AGEs via negative regulation of p66shc-dependent FKHRL1 phosphorylation. Am J Physiol Cell Physiol 2008;294:C145- 152 29. Pinho S, Marchand T, Yang E, Wei Q, Nerlov C, Frenette PS: Lineage-Biased Hematopoietic Stem Cells Are Regulated by Distinct Niches. Dev Cell 2018;44:634-641 e634 30. Kraakman MJ, Lee MK, Al-Sharea A, Dragoljevic D, Barrett TJ, Montenont E, Basu D, Heywood S, Kammoun HL, Flynn M, Whillas A, Hanssen NM, Febbraio MA, Westein E, Fisher EA, Chin-Dusting J, Cooper ME, Berger JS, Goldberg IJ, Nagareddy PR, Murphy AJ: Neutrophil-derived S100 calcium-binding proteins A8/A9 promote reticulated thrombocytosis and atherogenesis in diabetes. J Clin Invest 2017;127:2133-2147 31. De Bruyn PP, Breen PC, Thomas TB: The microcirculation of the bone marrow. Anat Rec 1970;168:55-68 32. Lorchner H, Poling J, Gajawada P, Hou Y, Polyakova V, Kostin S, Adrian-Segarra JM, Boettger T, Wietelmann A, Warnecke H, Richter M, Kubin T, Braun T: Myocardial healing requires Reg3beta-dependent accumulation of macrophages in the ischemic heart. Nat Med 2015;21:353-362 33. Tanaka M, Miyajima A: Oncostatin M, a multifunctional cytokine. Rev Physiol Biochem Pharmacol 2003;149:39-52 Diabetes Page 16 of 40

FIGURE LEGENDS

Figure 1 – Mobilopathy and myelopoiesis in experimental diabetes. Panels (A) through (G) report the comparison between diabetic (n=16) and non-diabetic control (n=12) mice in total white blood cells (WBC, A), absolute counts of lymphocyte (Lympho), monocytes (Mono), and granulocytes (Granulo) (B), as well as the granulocyte-to-lymphocyte (G/L) ratio (C). D) Comparison of FACS-defined common myeloid progenitors (CMP) and granulocyte-macrophage progenitors (GMP). E) Results of the colony-forming unit (CFU) assay from bone marrow (BM) cells (GEMM, granulocyte-erythroid-macrophage-Megakaryocyte colonies; GM, granulocyte-macrophage colonies; M, macrophage colonies; G, granulocyte colonies). F) Percentages of BM macrophages over total BM cells in diabetic versus control mice in the unstimulated and G-CSF stimulated conditions. G) Gene expression of oncostatin M (Osm) in the BM of diabetic versus control mice. *p<0.05 for the comparison between diabetic and control mice. Panels (H) through (J) illustrate HSPC mobilization in diabetic versus non-diabetic mice. HSPCs, defined as Lin-c-kit+Sca-1+ (LKS), were quantified before and after G-CSF administration and reported as fold-change from baseline in non-diabetic control mice (n=20) and in diabetic mice (n=20). Individual lines, indicating single mice, are shown along with the average fold-change (95% C.I.) for each time-point and the respective p-values (H). I) Comparison of the fold-change in LKS cell levels after G-CSF between control and diabetic mice. J) Comparison of the percentage of mice achieving a mobilization response of at least 1.5-fold in the diabetic versus non-diabetic control condition. Histograms indicate mean ± SEM. Box plots indicate median with interquartile range, and whiskers indicate range.

Figure 2 – Myelopoiesis and mobilopathy in human diabetes. A) Comparison of circulating CD34+ HSPCs and the neutrophil-to-lymphocyte (N/L) ratio in a pooled cohort of non-diabetic (n=236) and diabetic (n=108) patients (*p<0.05). B) Linear correlation between HSPC levels and the N/L ratio. Regression coefficients are reported for the entire cohort (along with p-values) and for the non-diabetic and diabetic patients separately. C,D) Linear correlation between HbA1c and HSPC levels (C) or the N/L ratio (D): the regression line with its 95% C.I. are shown along with the regression coefficients and p-values. E) Comparison between diabetic and non-diabetic patients in the increase (fold-change) in HSPC levels after G-CSF (*p<0.05) and in baseline N/L ratio. F) Linear correlation between the baseline N/L ratio and the increase (fold-change) in HSPC levels after G-CSF stimulation. Regression coefficients are reported for the entire cohort (along with p-value) and for the non-diabetic and diabetic patients separately. Histograms indicate mean ± SEM.

Figure 3 – p66Shc deletion protects from diabetes-induced myelopoiesis and mobilopathy. Myelopoiesis was evaluated in non-diabetic and diabetic wild type and p66Shc-/- mice (n>10/group, unless specified), by comparing peripheral blood white blood cell (WBC) count (A), WBC types (B-D), the granulocyte-to- lymphocyte (G/L) ratio (E), the bone marrow (BM) cell colony forming unit (CFU) assay (F), FACS-defined BM common myeloid progenitors (CMP) and granulocyte-macrophage progenitors (GMP; G), and the percentage of BM macrophages (H, n=5/group). *p<0.05 diabetic versus Ctrl; †p<0.05 p66Shc-/- versus wild- Page 17 of 40 Diabetes

type. I,J) CFU assay performed using BM cells from wild type (I) or p66Shc-/- (J) mice, which were stimulated ex vivo with 2 µg/ml S100A8/9 (*p<0.05 for the comparison of with untreated controls cells). K) Percentage of FACS-defined GMP in the BM of non-diabetic wild type and p66Shc-/- mice treated with vehicle or S100A8/9 (20 µg/kg twice a day for 3 days). L) Mobilization of HSPCs was evaluated in p66Shc-/- diabetic and non-diabetic mice (n=5/group) by enumerating circulating Lin-c-kit+Sca-1+ (LKS) cells. *p<0.05 in post- G-CSF versus baseline M) Schematic representation of the generation of hematopoietic and non-hematopoietic p66Shc-/- mice. N) The fold-change with 95% C.I. of LKS cells (calculated as post-G-CSF divided by pre-G- CSF levels) in wild type diabetic mice, p66Shc-/- diabetic mice, and in diabetic mice with crossed BM transplantation (*significantly different from 1.0, denoted by the dashed line indicating no effect; n=4- 5/group). O) The change in G/L ratio induced by diabetes is plotted for wild type mice, p66Shc-/- mice, and in mice with crossed BM transplantation (*significantly different from 1.0, denoted by the dashed line). The annotation under panels N and O indicates the genotype of the host (receiver mice) or the BM (donor mice).

Figure 4 – p66Shc deletion ameliorates bone marrow microvascular remodeling in diabetes. A) Bone marrow (BM) sections were stained with Hoechst (total cellularity) and anti-laminin to evaluate the microvasculature: based on specific thresholds, vascular items were scored as arterioles, sinusoids, or small vessels of capillary caliper. B) The four aligned panels report the numbers of any vessel, arterioles, sinusoids, and capillary-size structures per field in diabetic and non-diabetic wild type and p66Shc-/- mice (*p<0.05 diabetic versus non- diabetic; † p<0.05 versus wild type). C) Kernell density plot of vascular items scored based on size (x-axis) and circularity (y-axis): the area of the plot identified by the dashed box contains large irregular items (likely sinusoids), which was reduced by diabetes in wild type, but not in p66Shc-/- mice. D) BM sections were stained with Hoechst, anti-laminin (blood vessels) and anti-tyrosine hydroxylase (Tyr-OH), a marker of sympathetic nerve fibers. A representative example from a non-diabetic Wt mouse is shown to illustrate the pattern of Tyr- OH staining. E) Number of innervated arterioles / field and the fraction of innervated arterioles (F) over the total number of arterioles (*p<0.05 versus non-diabetic control). Histograms indicate mean ± SEM, with superimposed individual data points (n=5 / group).

Figure 5 – Osm deletion protects from diabetes-induced myelopoiesis and mobilopathy. A) Mobilization of HSPCs defined as Lin-c-kit+Sca-1+ (LKS) cells, induced by G-CSF, in non-diabetic and in diabetic Osm-/- mice (*p<0.05 versus baseline, n=5). B) Percentage of BM macrophages over total BM cellularity in diabetic and non-diabetic wild type (same as Figure 1F) and Osm-/- mice in unstimulated (Unst.) and G-CSF stimulated conditions (*p<0.05 versus control; †versus unstimulated; ‡versus wild-type; n=8-10/group). C,D) Total white blood cell (WBC) count (C) and granulocyte-to-lymphocyte (G/L) ratio (D) in non-diabetic and STZ diabetic wild type and Osm-/- mice (statistics marks as in panel B, n>10/group). E) Schematic representation of the bone marrow transplantation (BMT) experiment to generate hematopoietic restricted Osm-deleted mice. F) Mobilization of HSPCs in non-diabetic and diabetic hematopoietic-restricted Osm-/- mice (*p<0.05 versus baseline; n=5). G,H) Total white blood cell count (G), and G/L ratio (H) in diabetic and non-diabetic wild type Diabetes Page 18 of 40

18 mice with Osm-/- bone marrow. Histograms indicate mean ± SEM with superimposed individual data points, where appropriate.

Figure 6 – Signaling of OSM requires p66Shc. A) Hypothetical model wherein recruitment of p66Shc to OSM receptor (OSMR), instead of migration to mitochondria, is required for OSM to regulate Cxcl12 expression. B) Gene expression of Cxcl12 in bone marrow (BM)-derived mesenchymal stem cells (MSCs) isolated from wild type (Wt) or p66Shc-/- and treated with OSM 30 ng/mL or vehicle (control; *p<0.05 versus control; n=5/condition). C) Gene expression of Cxcl12 in mouse embryonic fibroblasts (MEFs) isolated from p66Shc-/- mice and transfected with an empty vector or vector encoding for wild type p66Shc (p66WT), serine 36 mutated p66Shc (p66S36A), or catalytically inactive p66Shc (p66qq) and treated with OSM or vehicle (control); *p<0.05 versus control; n=4/condition. D) Phosphorylation of ERK1/2 on threonine 202 or tyrosine 204 and phosphorylation of STAT3 on tyrosine 705 was evaluated by flow cytometry in wild type and p66Shc-/- MSCs treated with vehicle (Ctrl), mouse OSM with and without an inhibitor of ERK (SCH772984) or STAT (Stattic), respectively (a representative experiment of 3 replicates is shown). E) Signaling model wherein p66Shc recruited to the OSMR is required for ERK activation, but not for JAK-STAT activation via gp130/OSMR, although both ERK and STAT are required for OSM to induce Cxcl12. F) Gene expression of Cxcl12 in the BM of wild type (Wt), Osm-/-, and p66Shc-/- mice treated with vehicle (control) or OSM 0.5 µg every 6 h for 48 h (p<0.05 versus control). G) Levels of HSPCs, defined as Lin-c-kit+Sca-1+ (LKS) cells, in wild type (Wt), Osm-/-, and p66Shc-/- mice treated with vehicle (control) or OSM (*p<0.05 versus control). H) Fold-change of the granulocyte-to-lymphocyte (G/L) ratio in wild type (Wt), Osm-/-, and p66Shc-/- mice treated with vehicle (control) or OSM (p<0.05 versus control). Histograms indicate mean ± SEM with superimposed individual data points for each experiment.

Figure 7. Schematic representation of the link between myelopoiesis and mobilopathy exerted by the OSM- p66Shc signaling pathway. CMP, common myeloid progenitors; PMN, polymorphonuclear cells; M, macrophages; OSM, oncostatin M; OSMR, OSM receptor; HSPCs, hematopoietic stem/progenitor cell. M, mitochondrion; N, nucleus. Red bullets marked with “+” denote stimulatory effects. Page 19 of 40 Diabetes

Figure 1

4.38 A B C 2.5 D 0.8 H 10 10 p<10-4 15 15 Control (1.71-7.26) 10 p=0.86 Diabetic * l) l 2.0 * 8 8 

 0.6 8 / / 3 10 10 3 1.5 6 6 6 0.75 * 0.4 (0.33-1.00) * 1.0 4 41.00 4 1.00

5 5 of cells % LKS cells LKS LKScells 0.2 (0.79-1.14) cells LKS (0.77-1.20) (fold change) (fold (fold change) (fold Cells x 10 x Cells 0.5 * 2 2 change) (fold 2 WBCs(x10 0 0 0.0 0.0 0 0 Granulo / lymphocyte ratio lymphocyte / Granulo 0 l CMP GMP c lo o Baseline G-CSF Baseline G-CSF no u tr e ntrol o lin SF o abeti M an C F C r Con ne S Di Lympho G Diabetic Non diabetic controlG- mice Diabeticli mice Base se a G-C Control Bp<10-5

) -5 Diabetic 10 p<10 100 150 F 2.0 G CT 4 I J >1.5-fold

E  Control - *

basal) 8 Diabetic 80 1.5-fold 1.5 * 3 100 vs 6 60 1.0 2 4 * 40 CFU, n CFU, 50 2 %of mice 0.5 1 20 * 0 * mobilization cell LKS (fold change (fold 0 (%) macrophages BM 0.0 0 0 l (2 expression Osm BM l c c M M M G ro i rol ti ol tic M G Unstimulated G-CSF et tr Tota nt be ab ia iabe GE Cont Di Co D Con D Diabetes Page 20 of 40

Figure 2

A 600 2.5 * B 1500 Non-diabetic (r = -0.22) 2.0 Diabetic (r = -0.28) All (r = -0.24; p<0.001) 400 * 1000 1.5 HSPCs HSPCs + + 1.0 200 ratio N/L 500 (/10^6 cells) (/10^6 (/10^6 cells) (/10^6 CD34 CD34 0.5

0 0.0 0 0 1 2 3 4 5 6 tic e b N/L ratio Diabetic Dia Non-diabetic Non-diabetic

C 1500 r = -0.23; p < 0.001 D 6.0 r = 0.21; p < 0.001 95% C.I. 95% C.I.

1000 4.0 HSPCs +

500 ratio N/L 2.0 (/10^6 cells) (/10^6 CD34

0 0.0 0 2 4 6 8 10 12 14 0 2 4 6 8 10 12 14 HbA1c (%) HbA1c (%)

E 5.0 2.5 p=0.06 F 4.0 Non-diabetic (r = -0.45) 4.0 2.0 Diabetic (r = -0.47) 3.0 All (r = -0.32; p=0.03) 3.0 1.5 HSPCs HSPCs 2.0 + + 2.0 1.0 N/L ratio N/L * 1.0 CD34 CD34 1.0 0.5

0.0 0.0 0.0 (fold change after G-CSF) change after (fold (fold change after G-CSF) change after (fold c c c 0.0 1.0 2.0 3.0 4.0 tic ti ti ti e e e be b N/L ratio iab iab d Dia -d Dia n Non- No Page 21 of 40 Diabetes

Figure 3

5 5 L M -/- 3.5x* p66Shc 4 4 Hematopoieic basal) BMT p66Shc-/- vs 3 3 1,9x* Wt STZ Diabetic 2 2 (4 wks) CFU / mL / CFU p66Shc-/- 1 1 Control LKS cell mobilization cellLKS

(fold change (fold vehicle 0 0 BMT Baseline G-CSF Baseline G-CSF e F Wt Non-hematopoietic lin -/- Non-diabetice controlCS mice STZ diabetic mice p66Shc as G- B

N 4.0 O *** 4.0 ** 3.0 3.0

2.0 2.0 G/L ratio G/L 1.0 1.0 LKS cell mobilization cellLKS (fold change vs basal) vs (foldchange 0.0 95%C.I.) (fold-change, 0.0 Host Wt KO Wt KO Host Wt KO Wt KO BM Wt Wt KO KO BM Wt Wt KO KO Diabetes Page 22 of 40

Figure 4 Page 23 of 40 Diabetes

Figure 5

Control * vs Ctrl Wild type A 3.1x* B C -/- D 5 2.0 Diabetic † vs Unst. 15 Osm 1.5 ‡ vs Wt 4 * l) 

basal) 1.5 2,2x* / * 3 10 1.0 vs 3 1.0 †* 2 5 ratio G/L 0.5 ‡ 0.5 † 1 ‡ ‡ †‡ †‡ WBCs(x10 LKS cell mobilization cell LKS BM macrophages (%) macrophages BM (fold change (fold 0 0.0 0 0.0 Baseline G-CSF Baseline G-CSF Unst. G-CSF Unst. G-CSF Ctrl Diabetic Ctrl Diabetic Non-diabetic control mice STZ diabetic mice Wild type Osm-/- (from Fig. 1F)

F 15 7,7x* G 20 H 1.5 E l)  basal)

/ 15 OSM-/- C57 5,2x* Diabetic 10 3 1.0 STZ (n=5) vs BMT chimera 10 (4 wks) (4 wks) w/t C57 5 ratio G/L 0.5 chimera 5 (n=5x2) Control chimera WBCs(x10

vehicle (n=5) mobilization cell LKS (fold change (fold 0 0 0.0 Baseline G-CSF Baseline G-CSF ol tic ol tr etic be Non-diabetic control mice STZ diabetic mice Contr Dia Con Diab Diabetes Page 24 of 40

Figure 6 Page 25 of 40 Diabetes

Figure 7 Diabetes Page 26 of 40

Online Appendix, page 1

ONLINE APPENDIX

Figure S1 – Percentages of WBC types in diabetic versus control mice. *p<0.05 versus control. The histogram indicates mean and SEM.

100 Control Diabetic 80 60 * * 40 Percentage 20 0

ono M ranulo Lympho G Page 27 of 40 Diabetes

Online Appendix, page 2

Figure S2 – Basal levels of circulating HSPCs. HSPCs, defined as Lin-c-Kit+Sca-1+ (LKS) cells were enumerated by flow cytometry in non-diabetic control (Ctrl) and in diabetic mice either wild type (Wt), Osm- /- or p66Shc-/-. *p<0.05 for the indicated comparison between non-diabetic Osm-/- and Wt mice. The histogram indicates mean and SEM, with superimposed numbers of observations in each group.

Ctrl Diabetic

1200 * 1000

800 600

400

200 =18 =12 =10 Basal LKS cells / ml LKS/ cells Basal 0 n =19 n =22 n n n =9 n Wt Osm-/- p66Shc-/- Diabetes Page 28 of 40

Online Appendix, page 3

Figure S3 – HSPC mobilization in diabetes using CFU assay. A) Colony-forming units (CFU) from peripheral blood cells were quantified before and after G-CSF stimulation in wild type (Wt) non-diabetic control mice and in mice with streptozotocin (STZ)-induced diabetes. B) HSPC mobilization was also assessed as Lin-Sca-1+c-kit+ (LKS) cells in Wt mice transplanted with bone marrow (BM) cells from Wt mice and kept as non-diabetic controls or rendered diabetic by injection of streptozotocin (STZ). C) The change in CFU / mL of blood after G-CSF, indicating mobilization of functional HSPCs, is shown for Wt mice transplanted with BM cells from Wt mice. The change in LKS and CFU numbers is illustrated by the box and whisker plots along with individual data points. Box plots indicate median with interquartile range, while whiskers indicate range.

A 15001500p<0.001 1500 p=0.33 1500

10001000 1000 1000

500 500 CFU / mL / CFU 500 mL / CFU CFU / mL / CFU 500 CFU / mL / CFU

0 0 0 0 Baselinee F G-CSF Baseline G-CSF lin e Non-diabetic G-CScontrol mice STZne diabetic mice Bas G-CSF Baseli B 10 p<0.001 10 p=0.23 8 8

6 6

4 4 LKScells LKS cells LKS 2 2 (fold change vs basal) vs change (fold (fold change vs basal) vs change (fold 0 0 Baseline G-CSF Baseline G-CSF Non-diabetic control mice STZ diabetic mice

C 5000 p=0.003 5000 p=0.64 4000 4000

3000 3000

2000 2000 CFU / mL / CFU CFU / mL / CFU 1000 1000

0 0 Baseline G-CSF Baseline G-CSF Non-diabetic control mice STZ diabetic mice Page 29 of 40 Diabetes

Online Appendix, page 4

Figure S4 – Induction of p66Shc in the diabetic BM. Gene expression of p66Shc was determined in unfractionated BM of non-diabetic control mice and in streptozotocin (STZ) diabetic mice at four weeks of hyperglycemia (n=8/group; p=0.003). Histograms indicate mean ± SEM and individual dots indicate single animals. ) ) p=0.003 CT CT 0.03  

0.02

0.01

0.00 p66Shc expression ( expression p66Shc p66Shc expression ( expression p66Shc c tici et iabeab Z d Non diabeticSTSTZ di Diabetes Page 30 of 40

Online Appendix, page 5

Figure S5 – Percentages of WBC in wild-type and p66Shc-/- non diabetic (A) and diabetic (B) mice. *p<0.05 versus control. The histogram indicates mean and SEM.

A Non diabetic control 100 Wild-type p66Shc-/- 80 60 40 Percentage 20 0

Mono Lympho Granulo

B Diabetic 100 Wild-type p66Shc-/- 80 * 60 40 * Percentage 20 0

Mono Lympho Granulo Page 31 of 40 Diabetes

Online Appendix, page 6

Figure S6 – Osm gene expression in wild-type and p66Shc-/- non-diabetic control (Ctrl) and diabetic mice. *p<0.05 versus Ctrl; †p<0.05 versus wild-type. Histogram indicates mean and SEM.

Wild type p66Shc-/- )

CT 5  - 4 * 3

2 expression (2 expression 1

Osm † 0 BM Ctrl Diabetic Diabetes Page 32 of 40

Online Appendix, page 7

Figure S7 – Effects of S100A8/9 treatment in vivo. Non-diabetic wild type and p66Shc-/- mice were treated with vehicle or with S100A8/9 protein 20 µg/kg twice a day for 3 days. The graph on top shows the total number of colonies, which are explored below for each group by colony type. *p<0.05 for the indicated comparison; n.s., not statistically significant. Histograms represent mean and SEM with superimposed individual data points.

Vehicle S100A8/9 100 * n.s. 80

40 BM CFU, n CFU, BM 20

0 Wild type p66Shc-/-

Wild type mice p66Shc-/- mice

100 Vehicle 100 Vehicle * + S100A8/9 + S100A8/9 80 80 60 60

* n CFU, CFU, n CFU, 40 40 20 20 0 0 l l M M M G M M M G M G M G Tota Tota GE GE Page 33 of 40 Diabetes

Online Appendix, page 8

Figure S8 – Mobilization of functional HSPCs in p66Shc-/- mice. The colony forming unit (CFU) assay of peripheral blood cells collected before and after granulocyte colony stimulation factor (G-CSF) stimulation in non-diabetic and diabetic p66Shc-/- mice (n=3/group). Median and range, along with lines indicating single mice and fold-change versus basal are reported. *p<0.05 versus baseline.

1500 17,5x*

1000 12,8x*

500 CFU / mL / CFU

0 Baseline G-CSF Baseline G-CSF Non-diabetic control mice STZ diabetic mice Diabetes Page 34 of 40

Online Appendix, page 9

Figure S9 – White blood cell types in Wt ↔ p66Shc-/- cross-transplantation experiments. In each panel diabetic mice are compared to non-diabetic controls (*p<0.05). Panel A shows data from wild type (Wt) mice which received a bone marrow transplant (BMT) from Wt mice. Panel B shows data from p66Shc-/- mice transplanted with BM cells from Wt mice. Panel C shows data from Wt mice transplanted with BM cells from p66Shc-/- mice. Histograms indicate mean and SEM.

A) BMT Wt → Wt 15 100 Control 0.5

l) Diabetic *  80

/ * 0.4

3 10 60 0.3 40 0.2 5 * Percentage 20 0.1 WBCs(x10 0 0 Granulo / lymphocyte ratio lymphocyte / Granulo 0.0 ol tic tr e Mono ControlDiabetic Lympho Granulo Con Diab

B) BMT Wt → p66Shc-/- 15 100 Control 1.0 l)

 Diabetic

/ 80 0.8 * 3 10 60 * 0.6 * 40 5 0.4 Percentage 20 0.2 WBCs(x10 0 0

Granulo / lymphocyte ratio lymphocyte / Granulo 0.0 ol lo ol tr etic no tic b pho nu tr e Mo Con ia ym ra D L G Con Diab C) BMT p66Shc-/- → Wt

20 100 Control 1.0

l) Diabetic  80 / 15 0.8 3 60 0.6 10 40 0.4

5 Percentage 20 0.2 WBCs(x10 0

0 ratio lymphocyte / Granulo 0.0 ol no lo tr etic pho b nu ia ym Mo ra Con D L G ControlDiabetic Page 35 of 40 Diabetes

Online Appendix, page 10

Figure S10 – Megakaryopoiesis in diabetic and non-diabetic wild type and p66Shc-/- mice. Bone marrow sections were stained with Hoechst, laminin (red), and CD150 to visualize megakaryocytes (MK, green), which were quantified per microscopic field (*p<0.05 versus non diabetic control; †p<0.05 versus wild type). White arrowheads indicate MK. Histograms show mean and SEM, with superimposed individual data points.

Control Diabetic

Wild type type p66Shc-/- Wild 50 * 40 †

30 - / - 20

10 p66Shc

Megakariocytes / field / Megakariocytes 0 100 µm Ctrl Diabetic Hoechst Laminin CD150 Diabetes Page 36 of 40

Online Appendix, page 11

Figure S11 – Bone marrow macrophages and HSPC mobilization in sympathectomized mice. Mice sympathectomized by 6-OH dopamine (6-OHDA) were compared with mice treated with vehicle. A) 6-OHDA effectively reduced expression of the sympathetic fiber marker tyrosine hydroxylase (Tyr-OH), suggesting they had been sympathectomized (*p<0.05 versus control). B) The percentage of BM macrophages over total BM cellularity in sympathectomized mice was similar to that of control mice at baseline, and BM macrophages were equally suppressed by G-CSF treatment in both groups of mice. C) Mobilization of Lin-c-kit+Sca-1+ (LKS) cells after administration of G-CSF was impaired in mice sympathectomized with 6-OHDA (*p<0.05 versus baseline). Histograms indicate mean ± SEM.

Aa Bb Basal Cc Basal Post G-CSF Post G-CSF 1.5 0.6 8.0 * 6.0 1.0 0.4 4.0

0.5 0.2 LKScells (fold change) (fold (fold-change) * * * 2.0 Tyr-OH expression Tyr-OH

0.0 (%) macrophages BM 0.0 0.0 Control 6-OHDA Control 6-OHDA rol DA ont C OH 6- Page 37 of 40 Diabetes

Online Appendix, page 12

Figure S12 – White blood cell types in global (B) and hematopoietic restricted (B) Osm knockout versus wild type mice (A). In each panel, diabetic mice (red columns) are compared with non-diabetic controls (white columns). *p<0.05 versus control. BMT, bone marrow transplantation.

Control C A 15 Wild type B 15 Osm-/- 15 Osm-/-  Wt BMT Diabetic l l l    / / / 3 3 3 10 10 10 * * 5 5 5 Cells x 10 x Cells Cells x 10 x Cells Cells x 10 x Cells

0 0 0 o o o o ho h ho p on nul p ono nul M a mp Mono anulo M ym r r ym ra L G Ly G L G Diabetes Page 38 of 40

Online Appendix, page 13

Table S1 – List of primers.

Gene Forward Primer (5’->3’) Reverse Primer (5’->3’)

Osm AGCCCTATATCCGCCTCCAA GTGTGTCCTCACTGGGGAAG

Ubc GCCCAGTGTTACCACCAAGA CCCATCACACCCAAGAACA

Cxcl12 CGGGTCAATGCACACTTGTC GAGCCAACGTCAAGCATCTG

Shc1 GTCCGACTACCCTGTGTTCCTT CAGCAGGATTGGCCAGCTT (p66Shc)

Tyrosine CACCTATGCACTCACCCGAG CCAGTACACCGTGGAGAGTT hydroxilase Page 39 of 40 Diabetes

Online Appendix, page 14

Table S2 – Characteristics of the cross-section substudy. Continuous variables are presented as mean ± standard deviation, whereas categorical variables are presented as percentage. HDL, high density lipoprotein. LDL, low density lipoprotein. ACEi, angiotensin converting enzyme inhibitors. ARBs, angiotensin receptor blockers. CCB, calcium channel blockers. OHA, oral hypoglycemic agents.

Characteristic Non-diabetic Diabetic P Number 236 108 - Type 1 / type 2 diabetes - 28 / 80 - Age 45.8±7.5 57.3±11.7 <0.001 Sex male, % 34.7 64.8 <0.001 Body mass index, kg/m2 24.3±3.7 28.3±5.4 <0.001 Overweight, % 28.4 33.3 0.354 Obese, % 8.0 38.0 <0.001 Active smoking, % 21.6 14.8 0.140 Hypertension, % 25.4 75.0 <0.001 Systolic blood pressure, mm Hg 129.7±15.0 133.8±16.6 0.025 Diastolic blood pressure, mm Hg 83.3±10.2 80.7±10.2 0.026 Total cholesterol, mg/dL 207.3±38.7 187.0±36.7 <0.001 HDL cholesterol, mg/dL 61.8±16.1 51.5±17.7 <0.001 LDL cholesterol, mg/dL 127.6±35.8 108.9±29.2 <0.001 Triglycerides, mg/dL 89.4±76.3 133.2±87.4 <0.001 Diabetes duration, years - 10.9±10.3 - HbA1c, % 4.8±0.3 8.1±1.6 <0.001 (mmol/mol) (29±2) (65±13) Coronary artery disease, % 0.0 10.2 <0.001 Retinopathy, % 0.0 29.6 <0.001 Medications, % ACEi / ARBs 5.9 61.1 <0.001 CCB 2.5 13.9 <0.001 Beta-blockers 1.3 17.6 <0.001 Anti-platelet 0.8 31.5 <0.001 Statin 2.1 58.3 <0.001 Insulin 0.0 45.4 <0.001 OHA 0.0 47.2 <0.001 CD34+ HSPCs / 106 470.5±187.8 386.5±199.4 <0.001 N/L ratio 1.74±0.58 2.09±0.81 <0.001 Diabetes Page 40 of 40

Online Appendix, page 15

Table S3 – Results of the multiple regression analysis. To adjust for concomitant risk factors and conditions illustrated in Table S1, a multiple linear regression analysis was run with the co-variates illustrated below and CD34+ cell count (/106 events) or the neutrophil / lymphocyte (N/L) ratio as dependent variables. The non- standardized B coefficients with standard error (SE) are shown along with p values.

CD34+ cell count N/L ratio Variable B ± SE p B ± SE p Age, years -1.7±1.3 0.167 -0.01±0.005 0.030 Sex, male 37.7±23.3 0.107 0.04±0.06 0.658 BMI, kg/m2 7.5±2.8 0.007 0.01±0.01 0.420 Hypertension (Yes vs no) -31.6±26.3 0.230 -0.04±0.10 0.719 Total cholesterol, mg/dl 0.2±0.3 0.081 -0.001±0.003 0.258 HDL cholesterol, mg/dl -0.7±0.8 0.405 0.001±0.003 0.621 Triglycerides, mg/dl 0.1±0.2 0.671 0.00±0.001 0.784 Coronary artery disease (yes vs no) -14.1±61.1 0.817 0.04±0.22 0.856 Retinopathy (yes vs no) 72.6±42.9 0.092 -0.14±0.15 0.323 Diabetes (yes vs no) -41.1±11.9 0.001 0.55±0.12 <0.001