ZBTB12 DNA methylation is associated with coagulation- and inflammation-related blood cell parameters: findings from the Moli-family cohort

Background Zinc finger and BTB domain-containing protein 12 (ZBTB12) is a predicted transcription factor with potential role in hematopoietic development. Recent evidence linked low methylation level of ZBTB12 exon1 to myocardial infarction (MI) risk. However, the role of ZBTB12 in the pathogenesis of MI and cardiovascular disease in general is not yet clarified. We investigated the relation between ZBTB12 methylation and several blood parameters related to cardio-cerebrovascular risk in an Italian family-based cohort. Results ZBTB12 methylation was analyzed on white blood cells from the Moli-family cohort using the Sequenom EpiTYPER MassARRAY (Agena). A total of 13 CpG Sequenom units were analyzed in the small CpG island located in the only translated ZBTB12 exon. Principal component analysis (PCA) was performed to identify groups of CpG units with similar methylation estimates. Linear mixed effect regressions showed a positive association between methylation of ZBTB12 Factor 2 (including CpG units 8, 9–10, 16, 21) and TNF-ɑ stimulated procoagulant activity, a measure of procoagulant and inflammatory potential of blood cells. In addition, we also found a negative association between methylation of ZBTB12 Factor 1 (mainly characterized by CpG units 1, 3–4, 5, 11, and 26) and white blood cell and granulocyte counts. An in silico prediction analysis identified granulopoiesis- and hematopoiesis-specific transcription factors to potentially bind DNA sequences encompassing CpG1, CpG3–4, and CpG11. Conclusions ZBTB12 hypomethylation is linked to shorter TNF-ɑ stimulated whole blood coagulation time and increased WBC and granulocyte counts, further elucidating the possible link between ZBTB12 methylation and cardiovascular disease risk. Electronic supplementary material The online version of this article (10.1186/s13148-019-0665-6) contains supplementary material, which is available to authorized users.


Background
The zinc finger and BTB domain-containing protein 12 (ZBTB12) is a predicted transcription factor belonging to the big family of methyl-CpG-binding proteins (MBPs) [1]. ZBTB12 consists of four C-terminal C2H2/ Krüppel-type zinc finger domains predicted to bind to DNA, and an N-terminal BTB (broad-complex, tram-track, and bric-a-brac) domain for protein-protein interactions [2]. ZBTB proteins are described to play a role in hematopoietic development, differentiation and lineage fate determination [3], and malignant transformation [4]. Guarrera and colleagues [5] performed a genome-wide DNA methylation analysis in white blood cells (WBC) from two European cohorts and identified a region in ZBTB12 as the top differentially methylated genomic region in patients with myocardial infarction (MI) [5]. ZBTB12 hypomethylation was associated with MI risk, and the association was more pronounced in cases with shorter time to disease [5]. Despite ZBTB12 is expressed in most human tissues (Human Protein Atlas available from www.proteinatlas.org), its function and possible role in MI pathogenesis are still unknown.
In light of ZBTB12's potential role in hematopoiesis and MI risk, we investigated the association between ZBTB12 methylation patterns in the Moli-family cohort [6] and different blood cell parameters related to coagulation, inflammation, and cardiovascular disease (CVD) risk including whole blood clotting time, plateletleukocyte mixed aggregates, and blood cell counts, previously suggested as CVD risk factors [6][7][8][9][10].

Results
ZBTB12 is located on chromosome 6 and contains two CpG islands, both covering the only translated exon of the gene (EXON1, Fig. 1). Mean and standard deviation (SD) of methylation levels at the 13 ZBTB12 units studied are shown in Table 1. To identify possible connections among the ZBTB12 methylation units studied in the Moli-family cohort, we run a correlation analysis among all CpG units included in the study (Fig. 2). Since we found significant CpG unit inter-correlations, we conducted a principal component analysis (PCA) aiming at identifying common underlying components that could explain the largest part of methylation variability shared across units. Two main methylation factors emerged with PCA (Table 1), explaining a large part of gene methylation variance (86.1%). Factor 1 was characterized by high positive loadings of CpGs 3-4, 26, 1, 11, 5, 27, 18-20, and 6, and Factor 2 showed high loadings of CpGs 9-10, 21, 16, and 8 (Table 1). We first studied the association of ZBTB12 methylation factors with a number of classical CVD risk factors, including physical activity, smoke, hypertension, dyslipidemia, obesity, diabetes, and alcohol consumption. General characteristics and CVD risk factor distribution in the analyzed cohort are reported in Table 2. By studying the association between ZBTB12 methylation and these environmental factors, we observed that alcohol intake greater than 15 g/day (β = − 0.415, p = 0.0024, pFDR significant) was associated with Factor 2, while obesity (β = 0.40, p = 0.0053) and leisure-time physical activity (− 0.155, p = 0.0050) were associated with Factor 1 only with nominal significance (Table 2).
Then, we used linear mixed effect regression models to evaluate associations between ZBTB12 methylation Fig. 1 ZBTB12 structure (chr6:31899617-31901992, GRCh38/hg38 Assembly). Exon1 is indicated by a full blue box ("EXON1"). Two CpG islands are located in the gene ("CGI1" and "CGI2," depicted as light and dark green boxes, respectively). CpG islands are defined based on the formula described by Gardiner  and different blood parameters related to coagulation, inflammation, and CVD risk, namely unstimulated and TNFɑ-stimulated coagulation time (along with the resulting unstimulated-stimulated delta difference), plateletmonocyte and platelet-PMN aggregates, and blood cell counts (see Table 3). We did this through a double approach, by investigating association with methylation factors and with single CpG units. Because among the environmental variables associated with ZBTB12 methylation, only alcohol and obesity were associated with blood cell counts at p < 0.1 (data not shown), these variables were additionally included in the model as covariates to study the association between ZBTB12 methylation and blood cell counts (Tables 3 and 5). We found a significant association between Factor 2 and TNF-α-stimulated whole blood clotting time, with 16.0% increase in SD of clotting time for an increase of 1 SD in Factor 2 (β = 0.160, p = 0.0047; Table 3). The linear association between TNF-α-stimulated whole blood clotting time and Factor 2 was evident below the median (− 0.15) of adherence to the factor (Fig. 3a). Subjects with low methylation levels at factor units showed a coagulation time reduced by about half a minute (Δ-time in Fig. 3a) compared to those with higher methylation levels. The results did not significantly change when blood cell counts were added to the models as covariates.
Among blood cell count association results, WBC were associated with Factor 1, an association mainly driven by granulocytes (Table 3). In a similar fashion as for Factor 2, this association was evident at adherence to Factor 1 below the median (− 0.05), after which a plateau was reached (Fig. 3b).
DNA methylation changes at even only one CpG site can affect transcription factor (TF) binding to the DNA, influencing gene expression. Therefore, we searched for TF putative binding sites encompassing the WBC significantly associated ZBTB12 CpG sites. We found several TFs      (Fig. 4).

Discussion
Our study shows that the ZBTB12 methylation profile is associated with whole blood coagulation time after TNF-ɑ stimulation and with WBC and granulocyte counts. ZBTB12 is a highly conserved gene among species, but still poorly investigated. Recently, its hypomethylation has been associated with MI risk, in two European cohorts by Guarrera and colleagues [5]. In our study, we identified common linking patterns of the 13 ZBTB12 CpG units investigated (Factor 1 and Factor 2 in Table 1) that independently affect different CVD-related blood cell characteristics.
On the one hand, ZBTB12 Factor 2 was significantly associated with both the TNF-ɑ-stimulated procoagulant activity and the time difference between unstimulated and TNF-ɑ-stimulated procoagulant activity, independently on blood cell counts. Reflecting blood procoagulant activity potential, the coagulation time is calculated as the time taken for recalcified blood to clot and is considered to be a sensitive marker of the potential clot formation and CVD risk [7]. This is because thrombus formation depends upon the procoagulant and inflammatory potential of blood cells, including monocytes, granulocytes, platelets, endothelial cells, and plasma vesicles [11]. TNF-ɑ is specifically implicated in inflammation-related thrombosis by promoting extrinsic coagulation activation. This is achieved by inducing tissue factor expression on the leukocyte surface, downregulating natural anticoagulants (protein C and heparin-antithrombin pathways) as well as thrombomodulin and the endothelial protein C receptor, while increasing platelet production, thereby enhancing thrombin formation [12].
On the other hand, ZBTB12 Factor 1 hypomethylation is associated with higher total WBC and granulocyte counts already having been associated with higher CVD risk and mortality [8,9]. Neutrophils, the largest part of granulocyte population, are also involved in the formation of neutrophil extracellular traps (NETs), known to play a role in thrombus formation [13]. Both white and red blood cells contribute to the activation of coagulation and to thrombin formation also through the action of their extracellular vesicles (EVs), additional mediators of inflammation [14]. These results suggest a second potential role of ZBTB12 in affecting myelopoiesis.
ZBTB12 expression could be regulated through the binding of myelopoiesis and hematopoiesis-specific TFs, also influenced by DNA methylation [15]. In line with this hypothesis, our data on ZBTB12 TF binding site prediction showed that the ZBTB12 CpG units 1, 3-4, and 11 are predicted to be bound by PAX-5 and p53 (units 1 and 3-4), known to be involved in hematopoiesis and B cell differentiation [16] and cell cycle arrest required for terminal myelopoiesis [17,18], and by E2F-1 (units 3-4 and 11), with a pro-apoptotic role in hematopoiesis [19] (Fig. 4). Supporting this hypothesis, ZBTB12 expression in blood cells is indeed variable across cell types and differentiation stages (data from the BLUEPRINT Consortium [20], https://blueprint.haem.cam.ac.uk/mRNA). ZBTB12, as all ZBTB proteins, could also in turn bind myelopoiesis-related genes, acting as a TF, thanks to its predicted ability of binding methylcytosine (5mC) and/or oxidized methylcytosine (oxi-mCs)-rich DNA sequences, target sequences for Zn fingers [3]. DNA methylation is a known predictor of cell specification throughout the human hematopoietic lineage [21], and other ZBTB proteins are already described to be specifically involved in granulopoiesis [22] and myeloid development in general [23].

Conclusion
Our data indicate that ZBTB12 hypomethylation (of both Factor 1 and Factor 2) that was previously associated with MI risk [5] is linked to shorter TNF-ɑ-stimulated whole blood coagulation time and increased WBC and granulocyte counts. This hitherto undescribed association with blood parameters, known to be implicated in CVD [7][8][9], further support the hypothesis of a link between ZBTB12 methylation and CVD risk. Future experimental studies should focus on the specific molecular mechanism(s) of this zinc finger protein in blood cell proliferation, maturation, and activity and its possible role in human cardiovascular disorders.

Study population
Moli-family is a family-based study which aimed to investigate the role of inflammation-mediated activation of hemostasis in CVD risk [6]. A total of 754 subjects (≥ 15 years old) were recruited from 54 extended pedigrees (23 families with and 31 control families without personal or familial history of early-onset MI). All participants were relatives of index subjects enrolled in the Moli-sani cohort study [24], which recruited 24,325 subjects randomly selected from civil registries of the Molise Region, Southern Italy, between 2005 and 2010.
In all subjects, a complete medical history and information about smoking and alcohol-drinking habits were obtained via a structured questionnaire. Height, body weight, and blood pressure were measured as described in [6,25,26].

Blood sample collection and blood functional tests
Biochemical analyses were performed in the centralized Moli-sani laboratory. Blood samples were obtained between 07:00 and 09:00 from participants who had fasted overnight and had refrained from smoking for at least 6 h. Hematological cytometric analyses were performed by the same cell counter (Coulter HMX, Beckman Coulter, IL Milan, Italy), within 1 h from venipuncture. Plateletleukocyte conjugates, platelet P-selectin, leukocyte CD11b, and L-selectin expression were measured in whole blood for the Moli-family participants, as described [27].
Whole blood procoagulant activity was measured by the coagulation time. Whole blood was incubated for 2 h at 37°C with or without tumor necrosis factor (TNF)-α (100 ng/ml). The optimal agonist concentration was previously selected on the basis of dose-response curves (not shown). At the end of incubation, whole blood coagulation time (i.e., the time taken for recalcified blood to clot) was assessed by a one-stage clotting time. Briefly, 200-μL whole blood were mixed with 100 μL 25 mM CaCl 2 , and the time to clot formation was recorded (seconds) [28].

DNA extraction and methylation analysis
Buffy coats of peripheral blood cells were isolated from whole blood samples collected in sodium citrate EDTA tubes and centrifuged at 3000 rpm for 20 min at RT. DNA extraction was done using a silica matrix-based method as described [29]. Of the 754 Moli-family participants, 623 had good quality DNA samples to perform the methylation analysis. We measured ZBTB12 methylation using the Sequenom EpiTYPER MassARRAY (Agena) platform [15]. Details of the ZBTB12 region studied (chr6: 31899847-31900326, GRCh38/hg38 Assembly) are reported by Guarrera and colleagues [5]. Bisulfite treatment was conducted on 1 μg of genomic DNA using the MethylDetector kit (Active Motif ) as described [15]. All PCR amplifications were performed in duplicate. For the CpG-specific analysis, data were discarded when the duplicate measurements had a standard deviation (SD) ≥ 5% [15,30,31]. Sequenom peaks with reference intensity > 2 and overlapping units were excluded from the analysis [15,30,31]. To exclude possible intra-plate differences, a sample of K562 DNA was carried on in each plate as an internal control.
Of the 20 CpG units included in the ZBTB12 region studied [5] (CGI1 in Fig. 1), we excluded the ones having more than 15% of missing values in the Moli-family cohort, leading to a total of 13 CpG (Table 1).

Statistical analysis
Statistical analyses were performed using SAS/STAT software (Version 9.4 for Windows©2009. SAS Institute Inc. and SAS are registered trademarks of SAS Institute Inc., Cary, NC, USA). Mean and SD were computed for continuous variables and frequencies for categorical variables. All continuous variables, including methylation data, were also transformed to z-scores (mean = 0; SD = 1).
Correlation analysis among ZBTB12 CpG units was initially conducted to discover the architecture of relationships among the methylation units studied. Then, a PCA was conducted with the aim of identifying common underlying patterns that could explain the largest part of common variance in methylation across units. PCA was conducted including the 342 individuals having all the 13 ZBTB12 CpG units successfully measured. Criteria for factor selection were eigenvalue > 1.0 as revealed by the scree test, and the interpretability of the final solution. This resulted in the identification of two main factors (Table 1), which were transformed by the orthogonal varimax rotation to keep independent latent variables for subsequent analysis [32], and then standardized. We characterized the factors using the ZBTB12 methylation sites with an absolute factor loading > 0.40. Each subject received a factor score, calculated by summing the observed methylation site values, each weighted by factor loadings.
We first studied the association between ZBTB12 methylation factors and CVD risk factors (unstimulated and TNFɑ-stimulated coagulation time along with the resulting unstimulated-stimulated delta difference, platelet-monocyte and platelet-PMN aggregates, and blood cell counts) ( Table 2), in linear mixed effect regression models adjusted for age, sex (fixed effects), and family stratification (random effect) to account for the family structure of the Moli-family cohort.
Similarly, linear mixed effect regression models were used to assess the relation of blood parameters related to CVD risk, with ZBTB12 methylation patterns (Factor 1 and Factor 2) and single CpG units. Age, gender, smoking (never-, ex-, and current smokers), and variables significantly associated with both methylation factors and specific phenotypes at p < 0.1 were treated as fixed effects, while family stratification was treated as a random effect. A false discovery rate (FDR) method (Benjamini-Hochberg) was used to adjust p values for multiple testing. A p value (pFDR) < 0.05 was considered as statistically significant. DNA methylation is cell specific and might be different among the leukocyte sub-populations, leading to false positive findings when an appropriate correction for cell count is not performed [33]. ZBTB12 was not identified as a locus with leukocyte-specific DNA methylation levels [33]. Therefore, in our analysis, we did not correct for WBC counts.

Prediction of ZBTB12 DNA binding factors
To detect potential regulatory effects of methylation at the CpG sites investigated, we searched for ZBTB12 putative binding sites, by using the PROMO software [34]. More specifically, we included in our query the regions surrounding each of the CpG sites which were found as associated with blood cell parameters in previous analyses. This allows to construct weight matrices from known binding sites extracted from the TRANSFAC DNA binding site library (version 8.3), which contains the largest available collection of DNA binding sites in eukaryotes [35,36]. The full ZBTB12 region analyzed in the methylation study (chr6: 31899847-31900326, GRCh38/ hg38 Assembly) was used as DNA sequence bait in the search. The prediction was made by focusing only on the human species and transcription factors, setting the minimum sequence similarity threshold for TF binding detection to 85%.