Complement Factor B Is a Determinant of Both Metabolic and Cardiovascular Features of Metabolic Syndrome

Supplemental Digital Content is available in the text.

The rat Cfb gene resides within the major histocompatibility region on chromosome 20p12. 18 In SHR, this region has been demonstrated to be important in blood pressure regulation, 19 serum cholesterol, adiposity, and glucose tolerance. 20,21 In this study, we knocked out Cfb in SHR to test the hypothesis that Cfb is necessary for the full expression of cardiometabolic pathophysiological traits in this model of MetS.

Methods
Detailed methods are available in the online-only Data Supplement.

Rats
Cfb −/− rats were generated using SHR/NCrl rats (Charles River, Margate, United Kingdom), by microinjecting Zinc-finger nuclease (ZFN) mRNA (Sigma), targeted to exon 6 of Cfb (target sequence: CCCCT CGGGCTCCATGaatatcTACATGGTGCTGGATG), into 1-cell stage SHR/NCrl embryos that were implanted into pseudopregnant rats. Heterozygous progeny, from a founder harboring a 19-base pair deletion in Cfb, were intercrossed to homozygosity. A search for off-target events, conducted by whole genome sequencing confirmed the 19-base pair deletion. Six additional putative mutations, analyzed by Sanger Sequencing, were determined to be false positives (Table S1). Rats were housed with free access to food and water. All procedures were performed in accordance with UK Home Office regulations.

Statistics
Unpaired t test or 2-way ANOVA (Minitab Express) were used to assess differences between genotype and treatment. All results are mean±SEM. P<0.05 was considered significant.

Generation of a Cfb Knockout Rat
Using data from a quantitative trait transcript analysis of recombinant inbred strains derived from a SHR×Brown Norway (BN-Lx/Cub) cross, 22 we identified Cfb transcript levels as uniquely and strongly correlated significantly across the recombinant inbred strains for metabolically relevant traits (glucose uptake in isolated adipocytes, r 2 =−0.65, P (adj) =0.0003; basal lipogenesis in epididymal fat, r 2 =−0.64, P (adj) =0.0002; serum high-density lipoprotein cholesterol, r 2 =−0.64, P (adj) =0.0005) and significantly differentially expressed in adipose tissue between parental strains (SHR versus Brown Norway, 1.47-fold P (adj) <0.05). Overexpression in SHR adipose tissue was confirmed by quantitative polymerase chain reaction by comparing a further insulin sensitive/normotensive Wistar Kyoto strain (WKY/NCrl; Figure  S1A). Cfb was also overexpressed in SHR left ventricle (LV), but not liver, compared with WKY ( Figure S1A). Cfb overexpression in SHR was associated with increased AP activity compared with WKY ( Figure S1B). Analysis of the Cfb gene and its adjacent region revealed 14 variants unique to SHR, not present in Brown Norway or WKY; 2 variants reside upstream of the transcription start site ( Figure S1C). To investigate the potential causative role of Cfb in the cardiometabolic traits of SHR, a 19-base pair deletion in exon 6 of the Cfb gene in the SHR germline was made using ZFNs ( Figure S1D). Abolition of Cfb expression was confirmed by quantitative polymerase chain reaction and immunoblot ( Figure S1E), and loss of Cfb function was confirmed by ablation of serum AP activity ( Figure S1F).

Glucose Homeostasis
To test whether Cfb ablation affected glucose homeostasis in SHR, oral glucose tolerance and insulin sensitivity (IVITT [intravenous insulin tolerance test]) were assessed. Fasting plasma glucose concentration in Cfb −/− was significantly lower than SHR ( Figure 1A; SHR, 4.62±0.10 versus Cfb −/− , 4.25±0.09; P=0.013). Throughout the oral glucose tolerance, blood glucose remained lower, and area under the glucose curve was significantly reduced in Cfb −/− compared with SHR; insulin concentrations were similar in both groups ( Figure 1A and 1B). Together with the G:I ratio (ratio of area under the curve of plasma glucose concentration to area under the curve of plasma insulin concentration; Figure 1C), this indicated an improvement in insulin sensitivity, further demonstrated in IVITTs by a significant 48% increase in insulin-stimulated glucose disposal (K ITT ) in Cfb −/− compared with SHR ( Figure 1D).
To determine whether transcript changes were associated with altered adipose tissue respiration, we analyzed epididymal adipocyte metabolic rate. Maximal and basal respiratory rates were significantly greater in Cfb −/− than in SHR, +Δ1.64, and +Δ1.96-fold, respectively ( Figure 2; Figure S2A). Further, reserve capacity and leak respiration were both significantly increased ( Figure S2B and S2C). However, ATP-linked respiration and ATP-generation efficiency were similar ( Figure S2D through S2E). CoxIV protein abundance-a mitochondrial marker-was similar in both Cfb −/− and SHR ( Figure S2F).
There were no differences in body temperature or activity associated with Cfb deletion (Figure S3A and S3B). September 2017

Cardiovascular Analyses
Cfb deletion reduced relative LV mass and cardiomyocyte diameter by 10% compared with SHR; however, relative heart weight was similar between genotypes ( Figure 3A and 3B; Figure S4A and S4B). Telemetrically measured systolic and diastolic blood pressures were significantly lower (−Δ7 mm Hg) in Cfb −/− than in SHR, and although heart rate was similar, rate pressure product was significantly reduced ( Figure 3C and 3D; Figure S4C through S4F). Serum aldosterone and transcripts for renal renin and hepatic angiotensinogen were all significantly reduced in Cfb −/− rats (Table S4, Figure S5A and S5B).
Early structural and functional changes in the heart were investigated using echocardiography. We confirmed that relative LV mass was significantly reduced in Cfb −/− compared with SHR; however, at this stage, LV wall thickness was not significantly different (Table S5). Functionally fractional shortening and ejection fraction were significantly increased in Cfb −/− LV compared with SHR (Table S5). Given the similar heart rate and stroke volume, cardiac output was not significantly different (Table S5).

Serum Markers of Inflammation
Given the function of Cfb in inflammatory responses, we determined the effect of Cfb −/− on Th-1 mediated inflammation by quantifying serum concentrations of cytokines (Il-2, Il-6, Il-10, granulocyte macrophage colony stimulating factor, Ifn-γ, and Tnfα). We found significant decreases in serum concentrations of Il-10 and Ifn-γ in Cfb −/− rats compared with SHR. In addition, whereas Il-6 and Tnfα were detected in SHR, the cytokines were undetectable in sera from Cfb −/− rats. Granulocyte macrophage colony stimulating factor was similar in both groups, and in neither group was Il-2 detected (Table S4).

Analysis of GWAS Hits and cis-Expression QTLs at the Human CFB Locus
To determine whether genetic variants near CFB are associated with metabolic and cardiovascular disorders relevant to MetS (Table S3), we mined the NHGRI GWAS catalog (National Human Genome Research Institute) and located 18 single-nucleotide polymorphisms (SNPs) associated with cardiometabolic traits ≤1 Mb from CFB ( Figure 5; Table S6). Six SNPs were found to be associated with type 2 diabetes mellitus, MetS, or visceral fat. Six further SNPs were related to circulating lipids. The remaining SNPs were associated with coronary heart disease and hypertension (Table S6).
We also investigated whether variants at the CFB locus are associated with CFB expression by mining GTEx datasets (the Genotype-Tissue Expression project) for CFB cis-expression quantitative trait loci (QTLs). Fifty-three SNPs were associated with CFB expression in 4 tissues ( Figure 5; Table  S7). One SNP, rs76846904, close to the HLA-DRB5 gene, is highly correlated with CFB gene expression in subcutaneous adipose tissue (effect size, 0.78; P=0.000015) and within 100 kb of GWAS hits for visceral adiposity, serum cholesterol, and coronary heart disease.
The influence of the 18 GWAS SNPs, or any of their proxies (a total of 280 SNPs), on gene expression across 9 tissues was investigated using the GTEx Portal. Four SNPs were significantly associated (false discovery rate<0.05) with CFB expression in tissues of interest ( Figure 5; Tables S6 and S7). Two SNPs, correlating with CFB expression in "adipose subcutaneous" and "artery aorta", respectively, are proxies for rs13196329 and rs2247056, which are associated with visceral fat and triglycerides in the GWAS catalog (Table, Figure 5). Two further SNPs were significantly associated with increased CFB expression in "heart LV" and correspond to the same SNP (rs805303) that is associated with increased systolic and diastolic blood pressure and hypertension in the GWAS catalog (Table; Figure 5).

Discussion
We tested the hypothesis that Cfb is necessary for the full expression of cardiometabolic pathophysiological traits in the SHR model of MetS. Through ZFN-mediated gene knockout, we showed that the Cfb-deficient (Cfb −/− ) SHR has improved glucose tolerance and insulin sensitivity, along with favorable adipose tissue distribution, adipose oxidative capacity, and reduced circulating lipids and proinflammatory cytokines compared with parental SHR. Further, Cfb −/− rats had reduced blood pressure that was associated with increased ejection fraction and fractional shortening and reduced LV mass. The human CFB locus-a gene-rich region within the major histocompatibility complex-contains several GWAS hits for cardiometabolic traits, including coronary heart disease, blood pressure, MetS, type 2 diabetes mellitus, serum lipids, and visceral fat. These colocalize with cis-expression QTLs associated with expression of CFB in subcutaneous adipose tissue and other tissues, indicating that variation in CFB expression may underlie, in part, the GWAS hits at this locus.
Glucose intolerance, insulin resistance, visceral adiposity, and dyslipidemia are the key metabolic features of MetS that increase the risk of type 2 diabetes mellitus. 23 In our study, Cfb −/− rats had reduced visceral but increased subcutaneous fat. To investigate potential molecular changes associated with favorably altered fat distribution and ameliorated glucose homeostasis in Cfb −/− rats, we investigated transcripts central to adipose tissue metabolism. Reduced EAT mass in Cfb −/− rats was because of reduced adipocyte number rather than altered adipocyte volume. Pgc1a, Cpt1, and Aco1 were upregulated in Cfb −/− rats, suggestive of increased adipocyte oxidative phosphorylation, which we confirmed by Seahorse analysis. Cfb −/− rats exhibited a marked increase in basal and maximal respiration and had a 2-fold increased reserve respiratory capacity. Taken together with the reduction in adipocyte number, the data suggest that the elevation of mitochondrial respiratory capacity may provide an adipose tissue-intrinsic mechanism for reduced fat accumulation in Cfb −/− EAT. In SAT, increased mass in Cfb −/− rats was associated with increased Fasn and reduced Pgc1a expression, consistent with the function of Fasn as an insulin-sensitive fatty acid synthase, the role of Pgc1a in stimulating fatty acid oxidation, and the known upregulation of FASN in human obesity and type 2 diabetes mellitus. 24 These changes seemed to override the increases in Aco1 and Ucp1 expression observed in Cfb −/− rats, which would be expected to reduce adipocyte mass through increased trichloroacetic acid cycle activity and thermogenesis. The redistribution of visceral to subcutaneous fat marked changes in gene expression, and adipose respiratory capacity are likely to be the key to improvements in whole-body glucose homeostasis and metabolic function in Cfb −/− rats. Reduced BAT mass in Cfb −/− rats was associated with increased Pgc1a and Slc2a4 and decreased Ucp1 and FASN expression. This fat reduction may be consistent with increased Pgc1a driving lipolysis although inhibiting fatty acid synthesis; however, further experiments in Cfb −/− rats will be required to understand the BAT energy-substrate balance resulting from Cfb deficiency.
To further investigate altered adipose function in the Cfb −/− rat, we quantified Lep and Adipoq transcripts in EAT, SAT, and BAT. Although adipose Lep expression was reduced, circulating leptin was comparable in Cfb −/− and SHR. Although incompletely explained here, this could be accounted for by differences in post-translational processing and release, or peripheral metabolism, of leptin. Despite increased Adipoq expression in SAT alone, circulating high molecular-weight adiponectin was reduced in Cfb −/− rats. Conversely, high molecular-weight adiponectin in humans is lower in obese, insulin-resistant compared with lean, insulin-sensitive individuals. 25 However, adiponectin deficiency in mice has been shown to have no effect on glucose homeostasis on a normal diet. 26,27 Further, infusion of adiponectin in high-fat fed SHRs only marginally reduced insulin levels without affecting energy expenditure or hypertension. 28 Taken together with the observed metabolic improvements, this suggests other mechanisms, besides adiponectin, drive insulin sensitization in the Cfb −/− rat.
We also tested the hypothesis that deletion of Cfb in SHR would affect the expression of SHR cardiovascular phenotypes. In this study, we showed that Cfb −/− rats had reduced systolic and diastolic blood pressure, reduced LV mass and cardiomyocyte diameter, and an abrogated isoproterenolinduced increase in rate pressure product. These alterations represent a marked amelioration in several of the key cardiovascular features of MetS manifested in SHR. Cardiometabolic genome-wide association study (GWAS) hits and cis-eQTLs (quantitative trait loci) located in the human the complement factor B (CFB) locus. Eighteen relevant cardiometabolic single-nucleotide polymorphisms (SNPs) located <1 Mb from the boundaries of the human CFB gene (upper; red). Twenty-six SNPs were retrieved from the GTEx Portal that were found to be significantly associated with CFB expression (P<0.05), blue SNPs are associated with a significant negative effect, whereas red SNPs are associated with a significant positive effect. Four SNPs (with 1 overlapping) were determined to be correlated to both CFB expression, as well as being GWAS hits for relevant cardiometabolic traits (lower; red/blue). See Table S8 for a list of genes located in the CFB locus.
The reduction in blood pressure was associated with reductions in renin-angiotensin system components, suggesting that Cfb may have a direct effect, yet unexplained, on this system, mediating blood pressure and subsequently LV mass. Although Cfb deletion leads to lower blood pressure in SHR, our experiments do not distinguish whether Cfb is responsible for increasing above or maintaining basal blood pressure. Further detailed experiments are required to distinguish these 2 possible mechanisms.
To gain further insight into the molecular changes caused by Cfb deficiency in the heart, we investigated the effect of Cfb deletion on cardiomyogenic genes (ie, Nppa, Nppb, Myh6, Myh7, Acta1, and Camk2d), which are activated in response to stress. 29 Our study showed that despite reduced LV mass, Camk2d expression was significantly increased in saline-treated Cfb −/− . CaMKII (calcium/calmodulin-dependent protein kinase type 2) is proposed to regulate inflammation (Cfb, Tnfa, and Il-6) and cardiomyogenesis in response to hypertension-related pressure overload, β-adrenergic agonists, or myocardial infarction-induced cell injury. 30 Thus, Cfb may contribute to both cardiac inflammation and hypertrophy in response to stress, possibly through regulation of cardiomyogenic gene expression. For example, we showed complete or near complete abrogation in Cfb −/− rats of the isoproterenol-stimulated increase in Acta1 and Nppb expression seen in SHR. Further, Nppa expression was increased in both salineand isoproterenol-treated Cfb −/− . Therefore, independent of blood pressure, the lack of compensatory Acta1 upregulation and the favourable Actc1:Acta1 ratio 31 indicate that the Cfb −/− LV may be partially protected from compensatory cytoskeletal changes associated with LV dysfunction. Equally, abrogation of Nppb expression in the presence of isoproterenol indicates that the Cfb −/− LV is partly protected from stress. Further, upregulation of Nppa in Cfb −/− rats may, in part, contribute to the observed reduction in cardiomyocyte diameter and LV mass. Taken together, in Cfb −/− rats, upregulation of Nppa and abrogation of Acta1 expression in the presence of isoproterenol may indicate a blood pressure-independent mechanism for preserving LV function.
In addition to glucose metabolism and hypertension, we assessed the concentration of circulating lipids and Th-1 cytokines and showed reduced cholesterol and triglycerides, as well as reduced proinflammatory cytokines in Cfb −/− rats. Some of the metabolic and immune parameters that we measured here have also been measured in a Cfb −/− mouse, although no cardiovascular measurements have been reported. Like the Cfb −/− rat, the Cfb −/− mouse lacks AP activity and has reduced Tnfα, Il-6, and Ifn-γ. 32,33 Although having some immune similarities to the Cfb −/− rat, Cfb −/− mice compared with WT mice are more glucose intolerant and have higher circulating triglycerides. 34 The differences between these 2 models could be because of several reasons, including genetic background affecting metabolism differently, the use of high-fat diet in the mouse studies to elicit a phenotype, and the presence of 2 protein-coding Cfb transcripts in the mouse, whereas rats and humans have only one. On a high-fat diet, Ldlr −/− /Cfb −/− mice showed protection against atherosclerosis, 35 which is distinct from the amelioration in metabolic and cardiovascular phenotypes that we observed here. However, the 2 studies combined strongly encourage further investigation of Cfb as a target for protection from the development of cardiovascular disease.
Rat Cfb resides in chromosome 20p12, a region previously found to be important in the regulation of blood pressure, glucose homeostasis, and adiposity in SHR. [18][19][20][21] We propose that Cfb, at least in SHR, plays a major part in the development of key features of MetS that are linked to 20p12. However, given that the SHR.1N congenic that covers 20p12 has a reduction of 20 mm Hg, other genes in the region may also contribute. 19 The location of human CFB and the syntenic region to the rat gene is on human 6p21.33. 18 We located 18 SNPs with genome-wide significant associations to cardiometabolic traits ≤1 Mb from CFB. Several GWAS hits in the region were associated with type 2 diabetes mellitus and components of MetS. Two SNPs, rs13196329 and rs2247056, were correlated with visceral fat, triglycerides, and CFB expression. Further, 1 SNP, rs805303, was significantly positively correlated with systolic and diastolic blood pressure, and hypertension, as well as with increased CFB expression. These results suggest that CFB expression associated with these SNPs may be causally linked to accumulation of visceral fat, circulating lipids, and development of hypertension in humans.
In addition to altering complement activity, Cfb ablation reduced proinflammatory cytokines Ifn-γ, Il-6, and Tnfα whose elevated levels are associated with hypertension, obesity, and insulin resistance. 36,37 Further, chronic low-grade inflammation and overactivation of the innate immune system are now recognized causes of type 2 diabetes mellitus, 4,5 with clinical trials for therapeutic targets against inflammatory pathways for the treatment of diabetes mellitus and cardiovascular disease currently underway. 38 CFB indicates Complement factor b; FDR, false discovery rate; GWAS, genome-wide association study; QTL, quantitative trait locus; SNP, single-nucleotide polymorphisms; and TSS, transcription start site.
*P value (FDR), P value after adjustment for false discovery rate. †Slope of the correlation curve between SNP and CFB expression.
Compounds that target CFB already exist, and taken together with the findings in our study, suggest that CFB has significant potential as a novel target for treatment of metabolic disease 39,40 This is the first study to report the widespread amelioration of metabolic and cardiovascular phenotypes through deletion of an alternative complement pathway gene in a model of MetS. Cfb deletion improves glucose homeostasis, adipose distribution and function, lowers blood pressure and reduces cardiac hypertrophy, protecting against LV stress. Together with our analysis of the human CFB region for cardiometabolic traits, we conclude that CFB expression and function may directly or indirectly regulate multiple metabolic and cardiovascular processes in health and disease in the rat and in humans.

Perspectives
CFB is elevated in human cohorts with type 2 diabetes mellitus and cardiovascular disease, although a causal relationship has yet to be established. We identified alterations in Cfb expression as a possible cause of hypertension and insulin resistance in the SHR. Cfb knockout rats have improved glucose homeostasis linked to favorable alterations in adipose tissue distribution and function and reduced blood pressure and LV mass suggesting new adipose tissue-intrinsic and blood pressure-independent mechanisms for SHR insulin resistance and cardiac hypertrophy. SNPs in human CFB are associated both with hypertension and visceral adiposity and with CFB gene expression, suggesting that genetic variation in CFB may, in part, explain the genetic associations at the human CFB locus. Further studies are required to establish whether overexpression of adipose tissue Cfb alone is the prime determinant of MetS traits. Clinical trials are presently being undertaken to test the therapeutic effects of CFB inhibitors and to investigate AP components as causal factors in human diseases related to overactivity of the innate immune system. Given the findings in this study, CFB may also be a valid therapeutic target to treat or prevent progression of human MetS.