In the present study we examined a group of latent factors identified by performing factor analyses on 15 MetS or ECHO traits. The seven identified latent factors reduced the complexity of this large number of phenotypes. "Lipids-INS" factor was mainly contributed to by HLDC, TG and INS. "BMI-INS" was another factor identified from MetS domains. From the ECHO variables two latent factors were identified, "LV wall thickness" with main contributions from LVMI, PWT, RWT and MWS, and "LV geometry" with main contributors LVMI, LVID and RWT. "BP" factor, primarily a combination of SBP and DBP, was strongly evidenced in African Americans. However, when factor analysis with no rotation was applied in the whites, BP combined with "LV wall thickness" and "LV geometry" to form two new latent variables "BP-LV geometry" a combination of SBP, DBP, LVMI, PWT and LVID and "BP-LV dimension wall thickness" a combination of SBP, DBP, LVID, RWT and PP/SV. We look at these factors as MetS, ECHO, and a combination of MetS-ECHO latent factors. The presence of BP as a connector between MetS and ECHO is consistent with the increased risk of CV disease associated with hypertension. Chinali et al [21
] found that abnormal LV geometry and function are related to the MetS, with increased BP being the MetS component most strongly associated with cardiac abnormalities [22
]. Therefore, combined MetS-ECHO domains with BP contributions in the HyperGEN whites extend previous findings.
Two QTLs, one from MetS domains, and one from ECHO were prominent. The chromosome 12 QTL for the "BMI-INS" factor was replicated across ethnic groups, although the LOD scores were larger in African Americans. Its location coincides with linkage reports at 12q24.2 for non-insulin dependent DM, rheumatoid arthritis, and multiple sclerosis [23
]. In this area the ACACB
gene located on 12q24.1 may be a candidate. This gene reported in murine studies, has been implicated in controlling mitochondrial fat oxidation and is considered a regulator of energy expenditure [25
]. Although Wilson et al [26
] reported a finding for fat mass in 12q24, their peak location is about 30–35 cM apart from our "BMI-INS" peak finding.
Of particular note are the results in the region around the marker AFM031XA5 (D16S402) at 16q24.2-q24.3 for the "LV wall thickness" in whites. This location marked by D16S402, has been found to be linked with cadherin 13 (CDH13
) gene. CDH13
, which is also called heart cadherin
, is believed to be a calcium dependent mediator of cell-cell interaction in the heart and acts as negative regulator of neural cell growth. Joshi et al [27
] claimed cadherin
may have multiple signaling functions in vascular remodeling and may protect endothelial cells from oxidative stress-induced apoptosis. The CDH13
gene is about 1.2 M bp long, has 14 exons and encodes for a protein with 713 amino acids. This gene is most highly expressed in the heart. The AFM031XA5 marker represents a sequence between 81,851,181–81,851,356 (bp), precisely in the non-coding 5th
intronic region. We hypothesize that the CDH13
polymorphism, involved in calcium mediated cell-cell adhesion, can influence calcium deposition in the heart. Such a hypothesis is supported also by the Bella et al [12
] study of valve calcification in the HyperGEN where a significant QTL (3.2 LOD score) was located on chromosome 16 relatively close to the above marker. In contrast, Mayosi et al [13
] findings on chromosome 12 and 16 on LV mass do not overlap with our findings. We have a putative QTL located at 47 cM (1.99 LOD score) on chromosome 12 for LV Geometry factor (with contributions of LVID (+), LVMI (+) and RWT (-)) in whites. Theirs, (2.19 LOD score) is located at 75 cM based on the Marshfield genetic map. Our finding on chromosome 16 for LV wall thickness in whites (with contributions of LVID (+), RWT (+), and LVMI (+)) reaches a local maximum of 2.81 LOD score at 113 cM. Theirs is located at the start of chromosome 16 (18.07 cM based on the Marshfield map), and reached a local maximum of 1.85 LOD score. Such differences may arise for several reasons; our study is ascertained for hypertension [14
], larger in sample, our factor scores are composite traits compared to theirs for LV mass, to mention a few.
Other genetic linkage findings of interest included the MetS and ECHO factors, for "BP-LV wall thickness" and "BP-LV geometry" in whites (Table ). For the "BP-LV geometry" factor a QTL on chromosome 3 at marker D3S1262 (201.3 cM) showed a LOD of 2.0 in the same location where an abdominal obesity MetS QTL was reported. [28
] A QTL for "BP-LV wall thickness" factor with peak LOD score of 2.8 located on marker D3S1311 (225 cM), in proximity to DLG1
gene which is reported to have a role in the epithelial differentiation and regulation of smooth muscle [29
]. A LOD score of 2.3 was found for "LV geometry" after Varimax rotation, in a region that includes the FGF12
gene (fibroblast growth factor 12) which binds to the C terminus of the cardiac voltage-gated sodium channel Na(v)1.5 and modulates the properties of the channel [31
MetS related factors showed potentially important QTLs for African Americans. For example, "BMI-INS" factor showed highly significant results on 2p22-2p21 and on 12q24.2, as did the "BP" factor on 19q13.1 (Table ). Chen et al [32
] have previously reported strong linkage evidence in West Africans for percent body fat on chromosome 2 at 72.6 cM, relatively close to our 2p22-2p21 QTL finding. A possible candidate gene located at 2p21, close to the position of our linkage markers, is the LSL gene, which controls leptin serum levels. Candidate regions linking to our chromosome 19 BP QTL region have also been reported as for example by Bielinski et al. [33
] for pulse pressure. Also Cooper et al [34
] described an SBP QTL for Nigerians marked with D19S246 at 78.1 cM (about 7 cM distant from our finding) on the Marshfield map. Finally, in our study LOD scores increased when excluding DM subjects, which can be a reflection of the effect of genetic heterogeneity caused by diabetes.
Our new findings differ in several regards with the previous analysis of MetS in the HyperGEN study [9
]. First, the majority of the selected risk factors differ. The previous study had 11 risk variables to characterize only MetS; while the current one includes MetS variables and variables that additionally embrace cardiac geometry and function. Second, the sample sizes are not the same, and vary because of elimination due to missing values per variable and inability to include subjects from the 5th
HyperGEN center that did not participate in the ECHO study. Third, only 50–57% of the total original risk factors' variance was explained by the latent factors accepted in the model. Finally, the identified factors differ across studies in terms of the contribution of specific risk factors. Nevertheless when it comes to the "Lipids-INS" factor, (where a key contribution in African Americans was from HDLC, TG, and INS), the suggestive QTL found in the previous study on chromosome 11q24, is replicated in this study with a similar LOD score, but with a shift of the peak location (from 131.3 cM to 113.1 cM). Combined traits analysis can discover QTL locations that affect more than one trait, i.e. with pleiotropic effects. The 11q23-24 human genome location is well known for a cluster of genes (APOA1, APOC3, APOA4
, and APOA5
), which effect the lipid levels as well as is associated with DM and heart disease [35
]. In contrast, the use of two methodologies with and without Varimax rotation it might artificially increase the chances to obtain significant linkage results. Rao and Gu [36
] showed that for a set of 400 genome wide markers, if the LOD score threshold is relaxed to ~1.75, one expects a tolerance of 1 false-positive per genome. It is expected that this threshold, used by us in reporting results, may achieve a better "balance" of the types of the statistical errors. However, in this study we emphasized LOD score results at the level of above 2.5 and 3. Also the arrangements of risk factors into MetS (and ECHO) factors lessen multiple comparison issues [37
]. Another problem is the fact that we did not have a full evidence of replication about the QTLs in African-Americans and whites. Hirschhorn et al [38
] have shown through simulations that a QTL explaining 20% of variance can produce strong signal in one linkage analysis, but it can be undetectable in another one, simply for the reason of sampling variation. An additional possibility is that a common causal genetic mutation in one population might be rare in another population.
We expected not only that factor scores to provide an opportunity to detect pleiotropic gene effects. In addition, when traits closely related such as SBP and DBP form a separate factor they provide an increased power to detect putative QTLs compared with single trait analysis. For example, analyzing separately SBP and DBP produced weak LOD scores in the peak found on chromosome 19. When analyzed with the factor scores in African Americans we found a LOD score of 2.67, 2.18, 2.97, and 1.98 depending on the rotation method used and including or excluding T2D subjects. Nevertheless we do not know if the BP factor findings are true or type I error results. Consequently, further elucidation of the precise location of the causative polymorphisms of the identified linkage peaks is being processed with an abundant number of SNPs.