Differential endothelial cell gene expression by African Americans versusCaucasian Americans: a possible contribution to health disparity in vascular disease and cancer

Worldwide, coronary and cerebrovascular disease account for approximately 20% of deaths, an estimated 7.2 and 5.7 million people annually, and they are the two most common causes of death in high- and middle-income countries [1]. This proportion rises to approximately 30% if all cardiovascular disease types (for example, hypertension) are included. Even in low-income countries, cardiovascular disease is exceeded as the cause of death only by infectious diseases (in particular malaria, diarrheal diseases, tuberculosis and HIV) [1].

Within the United States there are significant health disparities between African Americans (AA) and Caucasian Americans (CA). Notably, AA have a 2.4-fold higher incidence of stroke [2] and an approximately 50% increase in prevalence of hypertension, the latter affecting approximately 31% of AA [3, 4]. This same disparity exists in the United Kingdom [5], and a local study verifies that it occurs in our own region from which the present study subjects were drawn [6]. In addition, AA display an increased prevalence of cardiovascular co-morbidities that contribute to pathogenesis in the general population [7]. For example, obesity has a 50% higher prevalence and affects approximately 45% of AA [8]. Correspondingly, AA have a two- to three-fold higher prevalence of type-2 diabetes so that it affects approximately 12% of AA [9], and they have an increased incidence of smoking, physical inactivity, and peripheral artery disease [10]. In addition to bearing the burden of a higher prevalence of cardiovascular and cerebrovascular disease, AA tend to develop such clinical diseases at a younger age than do CA (see Discussion). Disparities in cancer are addressed in the Discussion.

The debate as to what relative degree environment versus genetics causes these disparities is ongoing and vigorous.

The many factors interposed between racial identification (see Discussion) and any health issue provide a perplexing spectrum of possible non-genetic routes to disease disparity [11]. Examples include nutrition and exposures, access to health care and disparities in its delivery, social conditions and lifestyle choices, and so on. For cardiovascular health, it is telling that urbanization and its accompanying adoption of Western lifestyles are implicated in the accelerating development of hypertension and its comorbidities within Africa [12]. Likewise, the African diaspora established parallel gradients of hypertension and comorbid risk factors from West Africa to the Caribbean to North America [13, 14]. Even within the United States itself there are notable regional differences in stroke mortality rate among AA [15]. Thus, the tremendous influence of environmental factors is indisputable.

The present study, however, was conducted considering the possibility that the exaggerated burden among AA of cardiovascular disease, and perhaps even that of cancer, could stem in part from genetic determinants. Indeed, historical genetic studies indicated that ancestral continent-of-origin can be genetically identified [16]. Application of modern methods (for example, identification of single nucleotide polymorphisms [SNPs], haplotypes, copy-number variable loci, genetic diversity, frequency of non-beneficial SNPs) confirm an East African origin for modern humans with the spread of the human genome to the rest of the world [17‐19]. In turn, modern populations can carry genes that confer altered risk (higher or lower) for disease burden [20]. Indeed, newer methods such as discovery of disease association through admixture mapping have implicated specific loci influencing, for example, hypertension [21, 22], blood lipid levels [23], obesity and type-2 diabetes [24]. Such studies of AA versus CA have been extremely helpful in identifying the relationship between such loci and continent-of-origin.

Environmental and genetic influences do not occur independently; rather, gene-environment interactions contribute to disease variation [25]. Dramatic examples of this in Africa include the sickle mutation, the Duffy mutation, and APOL1 variants conferring protection against P. falciparum, and P. vivax, and Trypanosoma disease, respectively. On the other hand, gene variants that are hypothesized to have been of benefit in ancestral Africa can exert a harmful effect in modern societies. For example, a CYP3A5 allele which may have been advantageous in hot ancestral regions is now associated with salt-sensitivity hypertension in modern AA [26]. Another example is, in principle, found in the "thrifty gene" theory which posits that maximized calorie storage was advantageous ancestrally, but under modern conditions of more abundant food availability, obesity and type-2 diabetes (insulin resistance) are epidemic problems [9].

Thus, obesity is substantially influenced by genetics and is now understood to be a polygenic, complex disease with very significant gene-environment interaction [27]. Likewise, the relevance of gene-environment interactions has been emphasized for hypertension [28], stroke [29], cardiovascular disease [30], as well as metabolic syndrome, diabetes and atherosclerosis [31]. Furthermore, sequence-independent examples of this have been proposed whereby environmental influences exerted during development, for example, by poor prenatal nutrition, could establish durable, even trans-generational, effects on disease risk among AA through epigenetic mechanisms [32]. At the least, it seems likely that such effects exerted in childhood can cause a child to embark on the road to eventual cardiovascular disease [28].

The vascular endothelial cell is a universal participant in vascular diseases, and for that matter in many "non-vascular" diseases, for example, immune and infectious inflammatory diseases. Therefore, the endothelial cell could be an executor in fulfilment of any potential influence of genomic effects relevant to health discrepancies. Indeed, the endothelial cell is unique in being a critical participant in, and regulator of, multiple vascular functions, as well as comprising the major biological linkage between them. Examples include inflammation biology, governance of the pro- versus anti-coagulant balance, and regulation of vascular tone, among others. These processes and their proximate regulatory mechanisms interact in complex ways, so the functional/physiologic impact of even a precisely known allelic association is not necessarily accurately predictable. Hence, verification of genomic implications on cell biology is of vital importance for understanding pathophysiology, and such information can reveal therapeutic options and even inform pharmaceutical development.

Regarding vascular disease, AA exhibit a variety of findings consistent with endothelial dysfunction, in particular abnormal nitric oxide(NO)-dependent vasorelaxation [33, 34]. Consistent with this, ethnicity affects prevalence of clinically relevant variants of endothelial nitric oxide synthase (eNOS) [35]. Also, cultured umbilical cord endothelial cells from AA babies revealed hints of eNOS malfunction (uncoupling) [36].

The present study reflects our overarching interest in the concept that genetically-influenced, inter-individual differences in endothelial cell biology contribute to the heterogeneity of clinical phenotype evident in vascular diseases. As a general concept relating genomics and health, this is an accepted and compelling model for complex diseases. Regarding endothelial biology specifically, we previously employed the present approach to reveal gene expression differences coincident with a corresponding exaggerated endothelial cell response to inflammation signaling among the subgroup of children with sickle cell anemia who develop arterial occlusive disease in the Circle of Willis at the base of the brain [37]. Similarly, the present study was enabled by the technology we previously devised [37] that allows production of robust cultures of reporter endothelial cells (BOEC, blood outgrowth endothelial cells) from peripheral blood obtained from specific, phenotypically-defined individuals, in the present case, those self-identified by race as being AA or CA. We chose this device for subject group assignment because it was the method used for seminal epidemiologic studies of stroke and hypertension prevalence among AA [2, 3], and we wished the present results to be directly relevant to such studies. Caveats regarding this approach for subject group assignment, as well as the very concept of race, are presented in the Discussion.

Eligible volunteer subjects were between 20 and 29 years of age, inclusive, because we wanted our study to focus on a young, healthy population. All subjects claimed to be healthy, to not have any known cardiovascular disease, and to not be taking heart or lipid-lowering or blood pressure medications. We chose subjects who self-identified as being CA or AA, and they were included only if they stated that both parents were also CA or AA, respectively. In Minnesota, CA are overwhelmingly of European origin, largely from central and northern Europe. We excluded potential subjects who were from coastal East Africa for two reasons. First, the significant influx of East Africans to the upper Midwest has been a recent phenomenon, largely occurring after the seminal studies of AA cardiovascular health disparity, including the one done locally. Second, the gene pool in East Africa differs somewhat from that of Central/West, sub-Saharan Africa (the ancestral origin of most African Americans), as evidenced by the former's low frequency of Duffy negative status and virtual absence of the sickle mutation [17‐19].

Of the 45 subjects providing blood samples for this study, we subsequently excluded three AA and four CA because their BOEC cultures failed quality control testing (see below). The remaining 38 subjects (12 AA males, 9 AA females, 8 CA males, and 9 CA females) were grouped to create two subject groups, AA (n = 21) and CA (n = 17). The AA group included two individuals with sickle trait (tested for on each subject), the expected prevalence, but none with sickle cell anemia.

Subjects donated 50 to 100 ml of citrated peripheral blood that we immediately used to obtain peripheral blood mononuclear cells that were then placed in a special culture system (on collagen I; with multiple endothelial growth factors) [38]. The outgrowth BOEC from this long-term culture are mature, fully differentiated endothelial cells having unambiguous identity (by phenotypic, functional, gene expression, and electron microscopic criteria); and their population uniformity (100% BOEC) has been established [38‐41]. Although BOEC cultures ultimately reach approximately 10¹⁸-fold expansion [38], for the present study cells were expanded only to approximately 3 × 10⁷ cells (representing approximately 10⁶-fold overall expansion). BOEC were harvested four hours after the last change of culture medium and when they were 85 to 90% confluent. Each culture was meticulously performed by the same, highly-trained individual, ensuring that each step was always executed in the same manner. Culture medium components were obtained prior to project onset so that the same lot of reagents could be used for all cultures in this study. Aliquots of BOEC were taken for experimental use, quality control tests, and for cryopreservation.

Our quality control protocol is to verify that BOEC cultures are morphologically endothelial (cobblestone), that the cells are positive for three endothelial markers (CD31, VE-cadherin, P1H12), and that the cells are negative for monocyte/myeloid markers (CD14, CD45) as well as a probable marker of endothelial progenitor cells (CD133). We obtain cytogenetics analysis on each culture to ensure the BOEC are not characterized by culture-acquired clonal chromosomal abnormalities. Failure to meet the latter criterion is the reason seven subjects were excluded from data analysis.

Importantly, BOEC are not the so-called "EPC" that have been repeatedly (and largely erroneously) described in the literature. Rather, they are the progeny of a circulating, marrow-derived, transplantable, true endothelial progenitor cell [38]. Thus, BOEC themselves have never been influenced by in vivo modifiers such as blood milieu or tissue-specific signaling. Unlike other endothelial types, BOEC retain their phenotype at high degrees of expansion. Since they cryopreserve very well, aliquots from these cultures were saved at time of harvest for potential, future follow-up studies. Our previous validation studies of the present methods demonstrated that harvesting of BOEC at the degree of expansion used here takes place within a broad and safe window in which genetic instability has not ensued, but acquired endothelial phenotypes would have disappeared [37].

Sample preparation utilized BOEC trizol lysates and was always performed by the same individual and exactly as previously described [37]. For this study we used the Affymetrix U133A chip (Santa Clara, CA, USA). because our extensive prior validation experiments were performed using it. Biotin-labeled cRNA fragment samples were turned over immediately to our Microarray Core Facility which then returned raw data to us for analysis. To eliminate a source of variability, we purchased one single lot of chips in sufficient quantity for all samples from this project, so there would be no effects due to different lots of chips. We used the robust multiarray average (RMA) method to summarize expression values [42]. Data were background-adjusted, quantile-normalized (global scaling), and expression measures were summarized based on log-transformed perfect match values using median Polish algorithm. Then, we used locally weighted scatter-plot smoothing (LOWESS) to perform within-array normalization [43]. RMA and LOWESS smoothing were implemented in the software Genedata Expressionist Pro3.1PP (Basal, Switzerland).

Microarray data from this study have been deposited at the Gene Expression Omnibus, National Center for Biotechnology Information: GEO accession GSE22688, http://​www.​ncbi.​nlm.​nih.​gov/​geo/​query/​acc.​cgi?​acc=​GSE22688).

To test for differential expression of single genes, we utilized three approaches. (a) The Welch t-test does not require that variances of expression levels between two groups are equal, but it does assume that they are normally distributed [44]. The resulting P-value needs to be corrected for multiple comparisons, but a full Bonferroni correction is believed to be far too stringent for application to microarray data. (b) Significance Analysis of Microarray (SAM) is a nonparametric statistical procedure that employs False Discovery Rate (FDR), and it estimates FDR as an expression of false positive rate [45]. FDR is advantageous because it already adjusts for multiple hypotheses testing [46]. In this study, transcripts with FDR <.05 were considered to be significantly differentially expressed. However, to filter for a smaller field of transcripts, the results we focus on here used the much more highly stringent requirement that FDR = 0. (c) In addition, we simply examined the fold change of single gene expression for AA versus CA subject groups. All analyses of single gene expression were carried out in the statistical software R [47]. Specifically, we used the R function t-test for the Welch t-test and the R package samr for the SAM and fold change analyses.

As a positive internal control, we tested the 20 male and 18 female subjects for expression of selected genes. For the male group, 11/11 Y-linked genes showed significantly higher expression in the BOEC from male donors (FDR = 0). For the female group, 17/17 X-linked genes showed significantly higher expression in the BOEC from female donors (FDR = 0). Thus, SAM analysis robustly provided the correct answer.

To estimate the power of this study to detect differential expression of single genes, we performed a power calculation, assuming the requirement for a significance was P = 10^-5, equivalent to P = .05 corrected for 5,000 multiple comparisons. The software Java Applets for Power and Sample Size [48] was employed for power analysis for two-sample t-test. Microarray data from 27 BOEC samples from a previous study [37] were used to obtain the expected levels of expression and variance for this estimate. This estimated the power for detecting 1.5-fold differences to be 100% for transcripts at all quartile levels of expression. Power for detecting a 1.25-fold difference was (from highest to lowest expression levels): 100%, 90.5%, 64.1%, 32.9% and 67.8%.

Examination for differences between study groups in terms of biological systems pattern employed predetermined gene sets. Their construction took place prior to onset of this study, as described elsewhere in greater detail [37, 43], and was based on a variety of databases, as well as our review of the relevant literatures. These gene sets are not mutually exclusive, as we wanted them to broadly survey each given biological system, and these systems overlap biologically. For example, the shear stress responsive set includes members of multiple other biologies but which have in common shear responsiveness (see Discussion). The biological systems (number of members in a gene set) were: adhesion (146), angiogenesis (131), apoptosis (79), coagulation (152), hypoxia response (109), inflammation (117), redox signaling (83), shear stress response (156), and vasoregulation (106). The precise composition of these system-oriented gene sets is available elsewhere [49].

To determine whether pre-defined gene sets identified significant biological system differences between AA and CA, we utilized Gene Set Enrichment Analysis (GSEA). This test ranks individual genes in the set on the basis of t-statistic, examines for enrichment (in direction of either higher or lower expression) concordant with subject group assignment, and conducts many permutations to estimate the empirical nominal P-value and the FDR q-value, an expression of false positivity likelihood [50]. A gene set with FDR <.25 was considered significant, as suggested by the method's originators, and we tested our dataset in GSEA Java desktop software V2.0 which was downloaded from the authors' website [51].

Since there is no (known) way to perform a power calculation for GSEA analysis, we conducted multiple simulations in the statistical software R [47], based on the effect size and variance estimates from 27 BOEC samples from a previous study [37]. Simulation-based power estimation [52] is a generic approach and suitable for non-standard statistical techniques, such as GSEA. These allowed us to estimate that power of identifying significant changes for our pre-determined biological system gene sets (that is, P < .05 and FDR <.25) in this study approaches 100%.

This was used to verify gene expression by BOEC for authentic PSPH and the PSP-homolog (PSPHL), SOS1, AMFR, HSGP25L2G, Cxorf12, and EIF4B. These were chosen for qRT-PCR analysis because of their rank in the fold-change assessment for single gene differences. Primers and probes are displayed in Additional File 1. Methods were standard and were performed by the same individual, as previously described [37].

Previously cryopreserved BOEC from all of the subjects were re-established in culture. BOEC at 90% confluence were harvested for whole cell lysate preparation in phosphate-free buffer using three freeze-thaw cycles. L-phosphoserine phosphatase activities were measured by the release of inorganic phosphate (P_i) from L-phosphoserine, using an assay somewhat modified from that of Cho et al. [53]. The method is described in greater detail in Additional File 2. The phosphatase activity of each sample was expressed as μmol P_i/μg protein in total cell lysate. We utilized the same method to assess in vitro activity against substrate L-phosphoserine of authentic PSP (obtained from abcam, Cambridge, MA, USA) and PSP-homolog (synthesized in the Oligonucleotide and Protein Synthesis Facility in our Biomedical Genetics Center core facility).

Of the genes shown in Table 1, "PSPH" stands out because it exhibits the highest differential expression among AA subjects (4.08-fold increase, FDR = 0, P = 4 × 10 ^-9 ) (Figure 1A). However, only two of the Affymetrix probes for PSPH are specific for authentic PSPH, while the others additionally detect the known homolog, PSPHL. Indeed, qRT-PCR revealed that expression of authentic PSPH was not significantly different for AA versus CA, and it was very modest (Figure 1B). In contrast, expression of PSPHL was significantly elevated (P = .015) for the AA subjects (Figure 1C), and this entirely accounts for the false positive array result for PSPH. Interestingly, qRT-PCR analysis showed that PSPHL was undetectable in 12/17 CA donors but only 1/21 AA subjects.

The homolog, PSPHL, has partial sequence identity to PSPH, and the 31-residue overlap region contains some, but not all, of those involved in substrate hydrolysis [54, 55]. The PSPHL product is thus unlikely to display enzymatic activity. Consistent with published data [54], authentic PSP exhibited half-activity on L-phosphoserine at 2 μM, while the PSP-homolog had no activity. To screen for possible inhibitory activity by the homolog, we examined the L-phosphoserine phosphatase activity of AA versus CA BOEC. Despite PSPHL having higher levels in AA BOEC (Figure 1C), the L-phosphoserine phosphatase activities of BOEC from AA and CA were equivalent: 1.16 ± 0.22 and 1.21 ± 0.31 μmole P_i/μg total cell lysate, respectively. The amount of endogenous free P_i in AA BOEC (2.80 ± 0.77 μmole P_i/μg) was slightly, but not significantly, higher than in CA BOEC (2.29 ± 0.93 μmole P_i/μg).

We used qRT-PCR to examine (previously-cryopreserved) BOEC from all study subjects for expression of several genes from Table 1, and fold-changes were 1.26 for authentic-PSPH (P = NS) and 3.96 for PSPHL (P = .0154) for AA versus CA. As predicted from the small fold-changes detected by microarray, qRT-PCR was unable to confirm a significant difference for the generally similar fold-changes for SOS1, AMFR, HSGP25L2G, Cxorf12, or EIF4B.

To consider whole biological systems relevant to endothelial biology, we used GSEA to test BOEC from AA versus CA for evidence that any of nine specific systems tended to have altered expression: adhesion, apoptosis, angiogenesis, coagulation, hypoxia response, inflammatory signaling, redox response, shear stress response, and vasoregulation [37, 49]. This comparison revealed a significant whole-system differential expression in one biological system, shear stress response genes. This was evidenced not only by P = .027 and FDR = .14, but also by the fact that these indicators are, simply by inspection, unambiguously different from the values for the other eight systems (Table 2). The component composition of the shear response gene set is provided in Table 3 which also indicates which members of this set actually contributed to identification of this particular biological system as being different for AA versus CA (that is, those exhibiting "core enrichment"). The expanded names of these gene identifiers are provided in Additional File 5. All genes within the core enrichment group were found to have changed in direction of increased expression.

We detected 31 genes that met our most stringent criterion for differential expression significance, that FDR = 0, with 4 changed in the AA > CA direction, and 27 changed in the AA <CA direction (Table 1). All transcripts that showed a significant degree of change at the much more commonly utilized threshold of FDR <.05 are displayed in Additional File 3 (21 transcripts for 20 genes for AA >CA) and Additional File 4 (200 transcripts for 184 genes for AA <CA).

Many of the genes shown to be differentially expressed at the thresholds of FDR = 0 (Table 1) and FDR <.05 (Additional Files 3 and 4) have direct relevance to endothelial cell biology. While we do not wish to overstate their potential importance, the following discussion of some from Table 1 (presented in order of their listing therein) attempts to illustrate possible biological roles.

Our results revealed no significant difference between our study groups in NOS3 expression: AA >CA by 1.31-fold; t-test P = .071; FDR = not significant. Thus, the eNOS-related differences observed both clinically and experimentally (see Introduction) perhaps reflect acquired changes (e.g., enzyme uncoupling) rather than alterations in the endothelial expression level of the protein. As noted above, one such possibility would perhaps derive from the diminished level of DNM2 presented above.

Our second analysis examined whether any of nine biological systems relevant to endothelial cell biology would display a tendency toward altered expression for AA versus CA. Indeed, on a systems basis, AA had significantly altered expression of the gene set representing the shear response biological system, a major determinant of endothelial phenotype and vascular homeostasis (Table 2). The component members within this gene set which contributed to this result are shown in Table 3.

Laminar shear stress induces endothelial "quiescence" (anti-inflammatory, anti-thrombotic, anti-adhesive, anti-angiogenic, anti-oxidant, anti-atherosclerotic, and with optimal vasoregulatory balance), while areas of low or disturbed shear stress exhibit opposite changes [70‐73]. This actually is oversimplified, since experimental data indicate that the induced results also can differ depending upon whether the flow is pulsatile or oscillatory. In any case, shear responsiveness is mediated partly by its induced levels of transcription factor KLF2 (Kruppel-like factor-2), a "master switch" that regulates perhaps a third of the approximately 1,000 shear responsive genes. For example, the 47/156 members of the shear responsiveness gene set that accounted for the observed difference between AA and CA are depicted in Figure 2, which additionally highlights those that are directly influenced by KLF2. The genes that share the property of shear responsiveness actually are participants in multiple endothelial cell biologies as is summarized in limited fashion in Table 4. For example, some of the chemokines that promote development of atherosclerotic lesions are shear responsive [74], for example, CCL2 (monocyte chemoattractant protein 1) which is one of the genes contributing to the present implication of shear responsiveness (Table 3). Notably, KLF2 and NFκB down-regulate each other as significant regulators in orchestrating endothelial inflammatory phenotype.

The present study cannot identify whether the changes observed for AA would yield a net beneficial or net harmful effect on vessel wall biology, or whether they would even alter biological shear responsiveness. The increased expression among AA was inexplicably observed for genes that, in the normal process of establishing a quiescent endothelium, both increase in response to KLF2 (for example, NOS3, PLAT, THBD) and decrease in response to it (for example, EDN1, F3, TGFB1, JUNB, VCAM1) (Table 3, Figure 2). The explanation for this is not known at this time. Interestingly, there was no concurrent differential expression of the two major shear response "master switches," KLF2 (fold-change 1.05, P = .646, FDR = .678) or Nrf2 (1.06-fold change, P = .408, FDR = .588). However, we emphasize that the present method would be insensitive to, for example, a KLF2 SNP that alters its function but not expression.

Yet, this dilemma may derive, in part, from the fact that BOEC in this study perhaps would have had a relaxed shear response system, as opposed to a shear-modulated one. Unfortunately, existing shear responsiveness studies are not very enlightening regarding this concept because they generally have described changes induced by a shear regime rather than possible inter-individual differences in the starting phenotype. We speculate that the present data, in suggesting a tonic increase in multiple shear responsive genes for AA subjects, may be hinting at some underlying, functional variation of the apparatus that senses shear stress and induces phenotypic changes. We emphasize, however, that there is no direct evidence for this at the present time. Shear responsiveness is complex, incompletely understood, and involves multiple steps: mechanosensitivity, trans-membrane mechanotransduction, chemical mechanotransduction, downstream signalling and feedback loops, and ultimately transcription and translation [70]. We are currently utilizing the banked, cryopreserved BOEC from the AA and CA subjects in the present study to examine their gene expression shear responsiveness in a flow chamber system, and also to further delineate reason(s) for any difference.

Specific examples of dietary salt and lipid influences were noted in the Introduction. Emerging data emphasize the common occurrence of cardiovascular disease despite an absence of customary risk factors and suggest the importance of coincidental concurrence of genetic variability (genetic high risk) and environmental factors such as tobacco smoke exposure (environmental high risk) in premature and/or accelerated disease genesis [30]. An illustrative example of this in the general population, inherited variation in APOE alleles, is provided by the large role that oxidative stress probably plays in cardiovascular disease genesis. Compared to the majority of individuals having the ε3 isoform (gene frequency approximately 0.77), the fewer individuals having the ε2 isoform (gene frequency approximately 0.08) have the highest levels of blood APOE and lower cholesterol levels and protection from cardiovascular disease and stroke. Conversely, those with the ε4 isoform (gene frequency approximately 0.15) have the lowest APOE levels and higher cholesterol. However, the disease risk elevation observed for ε4 individuals is largely seen in those having environmental exposure to tobacco smoke [30]. Furthermore, the antioxidant capability of APOE shows apoE2 > apoE3 >apoE4, and the latter individuals exhibit higher levels of lipid hydroperoxides, which are harmful to endothelium [30]. In linking genetic makeup with dietary habits and environmental exposure, this provides a dramatic example of gene-environment interaction and the contextual complexity of its role as a determinant of vascular disease phenotype. As noted elsewhere: 'Genes load the gun, but the environment pulls the trigger' (attributed to Dr. Elliot Joslin) [30].

Finally, as noted in the Introduction, health disparities occur worldwide and in many forms. Undoubtedly, some reflect mostly environmental/social factors and others are more dependent upon genetic factors. However, a genetic admixture at some level is a universal characteristic of the human population, so the present study is widely relevant in the general sense. Admixture mapping studies for disease association loci are in the earlier stages of development, so at this time they are applied to more dramatic cases of genetic admixture in populations. One example, derived from an admixture a few hundred years (approximately six to seven generations) ago, is described here. Another prominent admixture occurred in western China, where the Uyghurs display an approximately 50/50 admixture of Asian and European backgrounds [75]. This example occurred more remotely, perhaps approximately 100 generations ago, and is likely explained by the fact that the Uyghurs have lived astride the historical Silk Road. Therefore, it most likely represents a version of admixture that is more historically representative of the human experience, admixture accompanying commerce or migration.

Quite separately from cardiovascular health, endothelial cells play a fundamental role in cancer biology since tumor growth is rate-dependent on angiogenesis [59]. AA are known to have cancers (for example, of prostate in men [56, 62] and breast in women [57, 76]) that exhibit greater aggressiveness, present in more advanced stages, and entail higher mortality rates, compared to their behavior in CA. Of great relevance to the theme of the present study, an examination of breast tumor microenvironment suggested that it exhibits a greater degree of angiogenesis (largely a host response to tumor presence) in AA compared to CA [57]. And in the general population, increased likelihood of invasive and metastatic breast cancer (previously shown to be influenced by the degree of tumor angiogenesis) is associated with specific vascular endothelial growth factor alleles [76], as is the aggressiveness of prostate cancer in men [62]. Therefore, the approach used in the present study may well be of use to further identify the influences of genetics and continent-of-origin on the endothelial biology of angiogenesis insofar as it contributes to cancer cell behavior.