Impaired Leptin Gene Expression and Release in Cultured Preadipocytes Isolated From Individuals Born With Low Birth Weight

Increasing evidence indicates that people exposed to an adverse intrauterine environment, such as being born with a low birth weight (LBW), are at increased risk of the development of insulin resistance and type 2 diabetes (T2D) later in life (1–4). However, the molecular mechanisms underlying developmental programming of T2D are still poorly defined.

Adipose tissue plays a central role in the regulation of metabolic homeostasis through the synthesis and storage of lipids and the secretion of adipokines that can modulate metabolism in other tissues (5). Programmed changes in adipocyte differentiation potential may result in an altered endocrine function and a decreased ability to store lipids, which subsequently may be deposited in other tissues such as liver and skeletal muscle. This may lead to insulin resistance and metabolic disease (6). The process of adipocyte differentiation begins in utero and continues into adult life. Terminal adipocyte differentiation involves the coordinated expression of a subset of cell type-specific genes that gives rise to a new morphological and functional phenotype. This process is regulated by a cascade of transcriptional activators, including the CCAAT/enhancer binding proteins (C/EBPs) (7) and peroxisome proliferator–activated receptor γ (PPARγ) (8). PPARγ is considered to be the master regulator of adipogenesis and is expressed as two major isoforms, γ1 and γ2 (9). PPARγ1 is found ubiquitously in the body, whereas PPARγ2 is largely found in the adipose tissue and the large intestine (10). Maintenance of the terminally differentiated state is ensured by sustained expression of C/EBPα and the activation of a number of adipocyte-specific genes (11), one of these genes being leptin (12). Leptin is a hormone mainly produced and secreted into the circulation by adipose tissue, and it plays a key role in the chronic control of appetite, energy balance, and insulin sensitivity. Leptin is predominantly expressed in mature adipocytes (13,14), and plasma leptin levels are in general closely associated with the total fat mass in humans (15). Besides being an important adipokine, leptin can be used as a marker of terminal adipocyte differentiation (16). It has been suggested that altered programming of the mechanisms controlling leptin gene expression and secretion from adipocytes may link early nutrition and growth with altered body composition, insulin resistance, and T2D (17).

Events in early life may have lifelong consequences, and epigenetic modifications are likely to play an important role in this process. The leptin gene is located on chromosome 7 at position 31.3, with transcriptional start site at base pair (bp) 127,881,337 (12). The leptin gene promoter contains a CpG island, and during adipocyte differentiation a demethylation of distinct CpG sites in the leptin gene promoter is crucial for leptin expression (14). In this study, we hypothesized that individuals born with LBW exhibit fundamental changes in preadipocyte differentiation potential, including altered leptin expression and leptin promoter methylation.

Individuals born with LBW (birth weight below 10th percentile, n = 14) and normal birth weight (NBW) (birth weight in the 50th–90th percentile range, n = 13) were recruited through The Danish National Birth Register. All individuals were selected using highly standardized criteria in accordance with previous studies from our group (18–21). All individuals were healthy age-matched male singletons, born at term, with no family history of diabetes, without a BMI >30 kg/m², and without self-reported high physical activity level (>10 h/week). All participants received oral and written information about the experimental procedures. Anthropometric measurements were obtained, body composition was determined by a dual-energy X-ray absorptiometry scan, and fasting venous blood samples were collected. To determine in vivo insulin action, the study participants underwent a 3-h euglycemic-hyperinsulinemic clamp (80 mU ⋅ m⁻² ⋅ min⁻¹) (22). Abdominal subcutaneous adipose biopsy specimens were obtained using a Bergström biopsy needle with suction (23) after an overnight fast. In brief, the abdominal subcutaneous adipose tissue depot was localized, and the needle was inserted just beneath the skin without penetrating the muscle fascia, ensuring collection of fat tissue from the subcutaneous depot. A subgroup of the study participants (LBW = 10 participants, NBW = 9 participants) were instructed to record their food intake over 3 consecutive days prior to the collection of the fat biopsy specimens. Vitamins (C, D, B6, and E), calcium, iron, selenium, cholesterol, fibers, total fat, protein, and carbohydrates were evaluated. The study was approved by the regional ethics committee (H-A-2009–040 and H-D-2008–127), and all procedures were performed according to the Declaration of Helsinki.

Primary adipose precursor cells were isolated from each subcutaneous biopsy sample as previously described (24), with some modifications described below. In brief, the adipose biopsy tissue was minced into small pieces with sterile scissors, and the pieces were transferred to a 5-mL sterile-filtered digestion solution (HAM-DMEM/F12 medium containing 1 mg/mL collagenase II, and 15 mg/mL fatty acid-free BSA; Sigma) and shaken for 20 min. The cell suspension containing the stromal-vascular cell fraction was filtered through a cell strainer, and the infranatant layer was transferred to a culture flask for cell propagation.

The isolated cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM)/F12 media supplemented with 10% FBS, 1% penicillin streptomycin (PS), and 1 nmol/L fibroblast growth factor in a 5% CO₂, 37°C environment until they reached 100% confluence. Two days postconfluence was designated as day 0. To induce differentiation of the preadipocytes, cells were first cultured for 48 h with DMEM differentiation medium containing 1% PS, 3-isobutyl-1-methylxanthine (540 μmol/L), rosiglitazone (200 nmol/L), dexamethasone (0.1 μmol/L), insulin (100 nmol/L), triiodothyronine (2 nmol/L), and transferrin (10 μg/mL). After 48 h, adipocytes were cultured in DMEM differentiation medium without 3-isobutyl-1-methylxanthine. This medium was changed every 72 h until day 12, at which time the cells displayed fully mature adipocyte morphology, as evaluated by massive lipid accumulation in all cells (Oil Red O staining; Supplementary Fig. 1C). All cell cultures were checked visually during differentiation, and only cultures displaying adipocyte morphology were included in this study. Cells were harvested at 80% confluence, on days 6 and 12 of the differentiation protocol (25). After initiation of the adipogenic program, cells are termed preadipocytes (day 6) or adipocytes (day 12).

Specific cell surface markers were used to identify cells with a high adipogenic potential in the cell cultures established from the stromal-vascular cell fraction (26). CD45-negative, CD31-negative, CD90-positive, and CD166-positive cells were used to determine the number of cells with a high adipogenic potential (26–30). We cultured passage one isolated cells in DMEM/F12 medium supplemented with 10% FBS, 1% PS, and 1 nmol/L fibroblast growth factor in 5% CO₂, 37°C environment until they reached 90% confluence. Cells were harvested using TrypLE (Gibco; Life Technologies), and thereafter were washed once in wash buffer (PBS containing 2% FBS and 0.01% NaN₃) and were resuspended in staining buffer (PBS containing 2% FBS, 1% Human Serum [catalog #1001291552; Sigma] and 0.01% NaN₃). Relevant antibodies, anti-human CD45-APC, CD31-FITC, CD90-PerCP.Cy5.5, and CD166-PE (BD Pharmingen), were added to the cells and were quantified by flow cytometry using a FACSFortessa (BD Bioscience). Data analysis was performed using Kaluza software, version 1.2 (Beckman Coulter). For isotype controls, murine IgG1k-PE/APC/FITC/PerCP.Cy5.5 (R&D Systems) was applied. For compensation, single stains were made using negative control beads and positive anti-mouse IgG beads (catalog #552843; BD Bioscience).

To study adipocyte differentiation potential, mRNA levels of adipogenic transcription factors and adipogenic markers were measured. Total RNA was extracted from the adipocytes using TRIzol reagent (Invitrogen). The RNA concentration was measured spectrophotometrically using Nano Drop 1000. RNA (0.5 μg) was used for synthesis of cDNA using the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems). All gene-specific primers, except for β-actin, were designed using human specific databases (Ensembl Genome Browser) and Universal Probe Library (Roche Applied Science) (Table 1). Primers were synthesized by DNA technology, and optimization was performed before use to determine primer working concentrations. β-actin mRNA levels were evaluated by a predesigned TaqMan primer and probe (Applied Biosystems). All samples were analyzed by quantitative real-time PCR on an ABI PRISM 7900 sequence detector (Applied Biosystems). Fold changes in mRNA expression levels were calculated after normalization to β-actin. β-actin did not differ over time or between the two birth weight groups (data not shown).

To evaluate leptin and adiponectin secretion, media were collected 24 h after the cells reached 80% confluence and 24 h prior to days 6 and 12, respectively. Media and serum were analyzed using the Meso Scale Discovery platform. Leptin and adiponectin recovery was assessed by adding standard leptin and adiponectin. Meso Scale Discovery plates were analyzed on a Sector Imager 2400. The assay was performed according to the manufacturer’s protocol (without diluting samples when cell-culturing media were analyzed). All standards and samples were measured in duplicate.

To evaluate end-stage differentiation, intracellular lipid droplets were stained with Oil Red O and quantified. On day 12, cells were washed with PBS and fixed with 10% Glyo Fixx for 1 h. The cells were thereafter washed twice with deionized water and once with 60% isopropanol. After washing, cells were stained with filtered Oil Red O working solution (2.1 g/L 60% isopropanol) for 1 h, then excess stain was removed by rinsing with deionized water and the cells were dried. Images of lipid droplets were visualized by light microscopy and photographed. In order to quantify lipid accumulation, Oil Red O was eluted with 100% isopropanol and the absorbance of the extracted dye was measured at 492 nm.

Genomic DNA was extracted using the DNeasy kit (Qiagen). A total of 300 ng DNA was bisulfate-converted using the EpiTect 96 Bisulfite Kit (Qiagen). PCR of the bisulfate-converted DNA was performed with the PyroMark PCR kit (Qiagen) with primers designed using the PyroMark Assay Design 2.0 software (Qiagen). Two primer assays were designed to cover the CpG sites of interest in the LEP promoter. Assay1 (forward primer: 5′-AGTTATTTTTAAATTTTTGGGAGGTATT-3′; reverse primer: 5′-ACTACTAACCCTAAACCCCCAATATAC-3′; sequencing primer: 5′- ATTTTTGGGAGGTATTTAAG-3′) covered CpG sites located at −204, −202, −200, −188, −186, and −183 bp upstream from the LEP transcription start site, and assay2 (forward primer: 5′-ATTGAGGGTTTAGGGTTAGTAGT-3′; reverse primer: 5′-ATTCCTACCAAACTCCATACCT-3′; sequencing primer: 5′-GGTAAGTAGTTATTTTGAGGG-3′) covered CpG sites −100, −95, −85, −74, −71, −62, and −51 bp upstream from the LEP transcription start site. Pyrosequencing of the PCR products was performed with the PyroMark Q96 ID instrument (Qiagen). Data were subjected to quality control and methylation quantification using the Pyrogram software, version 2.5.7.

All analysis was performed using SAS software, version 9.1.3 (SAS Institute). Parameters with data that lacked a normal distribution were logarithmically transformed prior to analysis. To evaluate adipocyte markers over time, during preadipocyte differentiation, and between the two birth weight groups, a two-way ANOVA for repeated measures was performed. The residuals obtained from the ANOVA models were evaluated, and the model was accepted only if the residuals were normally distributed. For comparisons between two groups, Student t tests were used. The Bonferroni correction was used to adjust for multiple comparisons. A nonparametric Spearman test was performed to analyze the correlation between leptin promoter DNA methylation and leptin mRNA expression. To ensure sufficient statistical power, we decided a priori to include a higher number of individuals in each group than in previous studies of adipose tissue biopsy samples from LBW and NBW individuals (31,32). In a post hoc power calculation, we had a 90% chance of detecting a 50% difference in gene expression between NBW (n = 13) and LBW (n = 14) subjects with a P value <0.05. Data in tables are presented as the mean ± SD, and data in figures are presented as the mean ± SEM. P < 0.05 was accepted as statistically significant.

The clinical characteristics of the two birth weight groups are presented in Table 2. At birth, the LBW individuals were 1.0 kg lighter than the NBW control subjects (P < 0.0001). Current weight (P < 0.01) and lean body mass (P < 0.01) were significantly lower in the LBW subjects, and they had a significantly higher waist-to-hip ratio (P < 0.05), reflecting higher abdominal adiposity. Similar differences have been observed in previous studies of LBW and NBW individuals (18–20,33–35). There were no differences in age, height, BMI, fat mass, or clamp (M value) between the two study groups. Furthermore, no differences in food intake or diet composition, with respect to energy intake, macronutrients, and micronutrients, were observed. All study participants, except one NBW control subject, were nonsmokers. Results remained the same whether including or excluding this single smoking person in the analyses.

Cells positive for the surface markers CD90 and CD166, and negative for CD45 and CD31 are considered a population of cells with a high adipogenic potential (26–30). No difference in the percentage of cells with high adipogenic potential (in cell cultures established from the stromal-vascular fraction) was observed between the two groups (NBW 95.4 ± 2.2% vs. LBW 98.0 ± 0.4%; P = 0.3) (Fig. 1A–D). In addition, we found no difference in any of the early adipogenic markers in the cell cultures before differentiation between the two birth weight groups (Supplementary Fig. 1D).

To evaluate the endocrine function of cultured adipocytes, leptin mRNA expression was measured by quantitative PCR in proliferating preadipocytes, in differentiating adipocytes (day 6), and in mature adipocytes (day 12). During proliferation, before the preadipocytes enter the adipogenic program, leptin mRNA was undetectable and therefore was not included in the analyses and figures. Both leptin gene expression and protein abundance in the cell media increased significantly from day 6 to day 12 of differentiation in the two groups, reflecting adipocyte maturation (effect of time, P < 0.0001) (Fig. 2A and B). Leptin mRNA expression was significantly lower in adipocytes from LBW individuals compared with NBW individuals during differentiation into mature adipocytes (P < 0.005) (Fig. 2A). In accordance, leptin levels in the cell culture media from LBW adipocytes were markedly decreased, reflecting reduced cellular leptin release (P < 0.01) (Fig. 2B). The same tendency was observed for adiponectin, showing a nonsignificant reduction in adiponectin secretion at day 12 in the LBW group (P = 0.09) (Supplementary Fig. 1A). In vivo measurements of overnight fasting serum leptin levels in the study participants were similar between the two birth weight groups (NBW 3,405 ± 773 pg/mL vs. LBW 2,418 ± 429 pg/mL; P = 0.27) (Fig. 2C).

DNA methylation was investigated at 13 CpG sites in the promoter region of LEP situated 204–51 bp upstream of the transcription start site. The region and CpG sites were specifically chosen based on results from luciferase studies in differentiating adipocytes (14) (Fig. 3A). We found no difference in methylation degree between the two birth weight groups in proliferating preadipocytes (Fig. 3B and C). However, we found markedly increased mean DNA methylation of all 13 CpG sites in adipocytes from LBW individuals compared with NBW individuals at day 6 of differentiation (P < 0.001) as well as borderline significantly higher DNA methylation in LBW subjects compared with NBW subjects in mature adipocytes (day 12) (P = 0.06) (Fig. 3B). Interestingly, at day 6 of differentiation there was a significant downregulation (P < 0.001) of the mean DNA methylation in the NBW adipocytes compared with the proliferative stage (Fig. 3B). When specific CpG sites were investigated individually (Fig. 3C–E), LBW subjects showed significantly increased DNA methylation compared with NBW controls at two CpG sites (−74 and −62 bp) (P < 0.001) at day 6 of differentiation (Fig. 3D), after Bonferroni correction. We found a significant inverse correlation between the mean DNA methylation degree and leptin gene expression at both time points (day 6: R = −0.57, P = 0.002; day 12: R = −0.526, P = 0.01) when the two groups were evaluated together (Fig. 3F and G). When performing individual birth weight group correlations, we found a negative correlation between the mean leptin DNA methylation level and leptin mRNA expression level for both groups at both time points (R value range −0.11 to −0.70; P value range 0.01–0.7).

To evaluate adipocyte differentiation potential and cellular function, mRNA of specific adipocyte transcription factors and adipocyte markers was analyzed during differentiating. Transcription factors mRNA from the C/EBP family were detectable during differentiation but did not differ between the two groups (Fig. 4A–C). However, fatty acid binding protein 4 (FABP4) mRNA, which is a marker of adipocyte maturation, was significantly reduced in LBW adipocytes during differentiation (P < 0.05) (Fig. 4E). A similar tendency was observed for PPARγ2 mRNA expression (P = 0.05) (Fig. 4D) and for GLUT4 (P = 0.05) (Fig. 4F). There was no difference in the cytoplasmic lipid accumulation between the groups as measured by Oil Red O staining at day 12 (Supplementary Fig. 1C).

In this study, we found that preadipocytes isolated from LBW individuals showed a significantly reduced leptin gene expression and a corresponding reduced leptin release into the media when differentiated in vitro. The level of DNA methylation of the LEP promoter was significantly increased in LBW compared with control individuals during preadipocyte differentiation. The notion of a dysregulated and immature adipocyte function in LBW subjects was further supported by our finding of reduced FABP4, PPARγ2, and GLUT4 gene expression in this group. These data suggest that an adverse intrauterine environment may influence key metabolic adipocyte functions, potentially linking LBW with an increased risk of the development of T2D.

We have chosen to focus primarily on leptin because it exerts a pivotal regulatory role in the control of energy homeostasis, and because leptin is a marker of a mature and functional adipocyte. Decreased leptin production in the cultured preadipocytes could link the physiological functions of the subcutaneous adipose tissue in LBW individuals with the development of insulin resistance and T2D. One of the main functions of leptin is to regulate appetite via specific leptin receptors in the hypothalamus, and a decreased leptin production and/or action is a main mechanism to increase hunger and, subsequently, food intake (15). In a newborn baby with LBW, increased food intake and weight gain may indeed be considered a beneficial survival mechanism. Nonetheless, rapid catch-up growth as a result of reduced leptin production/action may increase the risk of adiposity and T2D in humans with LBW in the long term (36).

Interestingly, we did not find any downregulation of leptin in plasma from the adult LBW individuals. However, the in vivo serum leptin levels do not only reflect the intrinsic and preprogrammed capability of the individual adipocyte to produce and release leptin, but are also controlled by a number of external regulatory mechanisms (37,38). Moreover, plasma leptin levels correlate with the amount of body fat, adipocyte cell size, and the distribution of fat into the subcutaneous or visceral depots (39). In a recent study, we found a diminished upregulation of leptin in response to short-term high-fat overfeeding in LBW subjects (35). It raised the possibility of a preprogrammed fundamental defect in preadipocytes, leading to impaired leptin secretion, which may play a role in the development of obesity in LBW individuals irrespective of the fasting plasma level.

The leptin promoter region contains a CpG island that is subjected to dynamic methylation during adipocyte maturation (14). This methylation process could be affected by the fetal environment in which the adipocyte was formed, and thereby could influence the adipocyte leptin gene expression and its secretion (40). The role of DNA methylation in transcriptional regulation of tissue-specific genes appears to be the primary silencing mechanism for genes with a CpG-rich promoter (41,42). In the proliferative state, there was no difference in leptin promoter methylation between the groups. However, during differentiation (day 6) the NBW group had a significantly lower degree of DNA methylation and an overall increased leptin mRNA expression as well as leptin secretion compared with the LBW group. A demethylation in the leptin promoter region of preadipocytes has been suggested to be required for leptin gene expression (14), which was also observed in the NBW group. The absence of any downregulation in degree of methylation during early differentiation may explain or contribute to the lack of increase of leptin gene expression during differentiation in LBW subjects, thereby linking the fetal environment to an adverse adipocyte endocrine function. Further studies are needed to understand the downstream events responsible for altered leptin gene promoter methylation in preadipocytes from LBW subjects, including the primary events occurring during fetal life.

The relationship between small changes in DNA methylation and their influence on gene expression has still not been fully elucidated. In this study, the change in methylation degree of the LEP promoter region ranged from 0 to 28% at the specific CpG sites studied (data not shown), and with a mean DNA methylation change ranging from 1.3 to 6.3%. Thus, the DNA methylation changes found in the LEP promoter may seem relatively modest compared with the observed changes in leptin gene and protein expression. However, the decreased leptin gene expression and secretion may represent a cumulative effect of changes in methylation on several CpG sites over time. Melzner et al. (14) showed that demethylation and thereby full activity of this region of the LEP promoter was critical for leptin gene expression, and that methylation of four of our investigated CpG sites (sites at −202, −188, −186, and −51 bp) was particularly important for LEP promoter activity during adipocyte differentiation. Importantly, we documented a statistically significant inverse correlation between LEP promoter methylation and leptin gene expression in the combined cohort of LBW and NBW subjects. Accordingly, there is evidence to suggest that the leptin gene methylation and gene expression findings are associated, and that the findings are biologically relevant in the context of developmental programming of T2D in LBW subjects. Besides DNA methylation, epigenetic regulation also includes histone modifications, which potentially could be an additional regulatory factor of leptin gene expression (43,44).

To evaluate key adipocyte functions in LBW versus NBW subjects, we measured a range of selected markers known to be essential for adipocyte differentiation and metabolism. We found no difference between the groups in any of the transcription factors belonging to the C/EBP family, representing some of the earliest differentiation markers. However, we found borderline significant reductions in the gene expression levels of PPARγ2 and GLUT4, and a significant reduction in FABP4, which is a marker of late differentiation, in the LBW subjects. These reductions in gene expression observed during late differentiation may be a result of subtle transcriptional defects occurring at earlier time points during preadipocyte differentiation in LBW subjects. These findings support the theory of a more immature adipocyte cell type in the LBW subjects.

Programmed changes in preadipocyte differentiation potential and metabolic functions could potentially result in a decreased ability to store lipids, an event associated with insulin resistance and metabolic disease (6). However, we found no significant impairment of the capability to store fat in vitro as estimated by Oil Red O staining in the preadipocytes from LBW subjects at day 12. Metabolic challenges, like fat overfeeding, may be required to reveal an impaired capability of adipocytes to retain and store fat in LBW subjects. In vivo studies of LBW subjects support the idea of fundamental changes in lipid metabolism including increased fasting lipolysis (20) as well as increased nocturnal fat oxidation, potentially associated with an inability to retain fat in the subcutaneous adipose tissue depot (45). A previous study reported reduced numbers of muscle satellite stem cells in mice subjected to prenatal undernutrition (46). Indeed, in vivo fat tissue preadipocyte number could also theoretically be altered and thereby contribute to the increased risk of T2D in LBW subjects (46). Importantly, in the current study our findings in cultured preadipocytes cannot be explained by a reduced number of cells present in vivo, given that the established cultures from all individuals contained the same number of cells with a high adipogenic potential as determined by fluorescence-activated cell sorter analysis. Accordingly, the difference in leptin secretion is likely to represent an intrinsic functional impairment in the preadipocytes obtained from LBW subjects. The extent to which there may be a reduced content (density) of preadipocytes in the subcutaneous tissue in LBW subjects in vivo, in addition to functional impairment of leptin transcription, needs to be addressed in future studies.

In accordance with previous studies, LBW individuals have an altered fat distribution compared with NBW controls (33–35), which is reflected in this study as a higher waist-to-hip ratio in LBW individuals. However, no correlation between waist-to-hip ratio and leptin methylation, mRNA expression, and protein secretion was found in the two study groups. The differences in leptin readouts remained after correction for differences in waist-to-hip ratio.

In summary, we showed that leptin mRNA and secreted protein were downregulated in cultured preadipocytes derived from LBW subjects. Furthermore, cultured preadipocytes from the LBW subjects showed increased DNA methylation of several CpG sites in the proximal leptin promoter during differentiation. These findings, combined with the observation of decreased FABP4, PPARγ2, and GLUT4 gene expression in LBW preadipocytes, together support the idea of dysfunctional preadipocyte functions and a relatively more immature fat cell type in young adult LBW subjects. Further studies are needed to establish the extent to which these findings may contribute to increased risk of the development of T2D in LBW subjects.

Funding. This study was funded by the Rigshospitalet, the Danish PhD Schools of Metabolism & Endocrinology, the Danish Council for Independent Research, the Novo Nordisk Foundation, the Danish Strategic Research Council, European Foundation for the Study of Diabetes, and the Aase and Ejnar Danielsens Fond. The Centre of Inflammation and Metabolism is supported by a centre grant from the Danish National Research Foundation. B.K.P. and J.F.P.W. are partners of the UNIK Project: Food, Fitness & Pharma for Health and Disease, supported by the Danish Ministry of Science, Technology, and Innovation.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. N.S.S. edited and revised the manuscript, drafted the manuscript, prepared the figures, interpreted the results of the experiments, analyzed the data, performed the experiments, and designed the cell study. C.B. edited and revised the manuscript, drafted the manuscript, interpreted the results of the experiments, analyzed the data, performed the experiments, and designed the cell study. L.G. edited and revised the manuscript, analyzed the data, and performed the experiments. B.M. and S.W.J. edited and revised the manuscript, performed the experiments, and designed the human cohort study. H.S.S. edited and revised the manuscript, prepared the figures, analyzed the data, and performed the experiments. C.S. edited and revised the manuscript, interpreted the results of the experiments, and designed the cell study. J.F.P.W. edited and revised the manuscript and designed the human cohort study. B.K.P. edited and revised the manuscript and interpreted the results of the experiments. A.V. edited and revised the manuscript, drafted the manuscript, interpreted the results of the experiments, analyzed the data, designed the human cohort study, and designed the cell study. A.V. is the guarantor of this 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.