Next steps for gene-identification in primary hypertension genomics

Blood pressure (BP) genomics informs on the root origins of primary hypertension, which are still unclear. Since 2008 the field of BP genetics has changed with evidence accumulating from genome-wide association studies (GWAS). In this issue of Hypertension, Zeller et al.¹ describe a different approach to BP gene discovery by using RNA expression profiles instead of DNA variants.

Using DNA variants, twenty-four large GWA studies have been published to date, and the number of new loci is steadily increasing with a large contribution from the latest studies with up to 300,000 individuals in the discovery phase (Table 1 and Figure 1). In total ∼300 variants have now been replicated to be associated with systolic- (SBP) and diastolic blood pressure (DBP) and their phenotypic derivatives (full list and references to individual studies at www.bloodpressuregenetics.org). Some consider the glass half-full, others half-empty regarding the new knowledge gained by BP GWAS: It is clear that novel findings have been added, but at the same time much of the heritability is not yet captured by the variants identified. To date only ∼3-4% of phenotypic variance is explained by a subset of variants identified² translating to ∼6-8% of the heritability. For the hypertension clinician new results on possible causal tissues of primary hypertension and notably on the relationship of hypertension with outcomes such as renal function are of particular interest^2-4.

Table 1.

Large BP genome-wide (gw), gene-centric (gc), CardioMetabo-Chip wide (cm), and exome-wide (ex) association analyses. AD = African diaspora; n= no replication available. See www.bloodpressuregenetics.org for full references and a list of loci.

Phenotype(s)	Consortium	Ancestry	Discovery sample size	Num. new loci	Total num. of loci	Pub. year
SBP, DBP, HTN (gw)	CHARGE-BP	European	29,136	9	9	2009
SBP, DBP (gw)	Global BP-Gen	European	34,433	9	9	2009
SBP, DBP (gw)	Amish	European	7,125	1	1	2009
HTN (gw)	Global BP-Gen	European	3,320	1	1	2010
SBP, DBP, HTN (gw)	Amagasaki	Asian	1,526	0	7	2010
SBP, DBP (gw)	AGEN-BP	Asian	19,608	4	11	2011
SBP, DBP (gw)	ICBP	European	69,395	14	28	2011
MAP, PP (gw)	ICBP	European	74,064	3	12	2011
SBP, DBP (gw)	CARe	African & AD	8,591	1	1	2011
SBP, DBP, PP, MAP (gw)	Global BP-Gen	European	25,118	4	8	2011
SBP, DBP (gw)	COGENT	African & AD	29,378	3	5	2013
SBP, DBP, PP, MAP (gc)	CARe	European	61,619	2	13	2013
SBP, DBP, PP, MAP LTA (gw)	ICBP	European	46,629	3	19	2014
SBP, DBP, PP, MAP age-effect (gw)	ICBP	European	55,796	2	20	2014
SBP, DBP, PP, MAP (gc)	IBC-BP	European	87,736	11	38	2014
SBP, DBP, PP, MAP (gw)	AGEN-BP	Asian & European	99,994	12	35	2015
SBP, DBP, HTN pleiotropy (gw)	COGENT	African & AD	29,378	1	4	2015
SBP, DBP kids and adolesc. (gw)	EAGLE	European	23,689	2	2	2016
SBP, DBP (cm)	ICBP	European	201,000	17	66	2016
SBP, DBP, PP, HTN (ex)	CHARGE-BP	Trans-ethnic	146,562	31	70	2016
SBP, DBP, PP, MAP, HTN (ex)	Global BP-Gen	Trans-ethnic	192,763	30	51	2016
SBP, DBP, PP, HTN (gw)	KAISER	Trans-ethnic	321,262	44	241n	2017
SBP, DBP, PP, HTN (gw)	UKB/ICBP	European	152,259	32	102	2017
SBP, DBP, MAP, PP LTA (gw)	AGEN-BP	Asian	18,422	1	4	2017

Figure 1.

Number of BP loci discovered in the large GWAS studies (Table 1) as a function of sample size. The study name is indicated in the plot for the largest studies. The line represents the result of a linear regression of Y on X.

How can progress be made in addition to ever larger sample sizes, soon to reach close to one million participants? Extrapolating from Figure 1, an empirical number of ∼600 BP loci is expected to be apparent with one million samples, somewhat more than the expected number based on theoretical considerations⁵. Other groups concentrate on filtering the variants based on position or function which are promising avenues given the limited statistical power due to multiple-testing inflation, even using very large sample sizes. Additional valuable efforts concentrate on phenotypes with increased phenotypic precision, but SBP and DBP are hard to beat in their simplicity to ascertain, with consequently large numbers of study-participants available, their evidence from clinical trials, and their predictive power for cardiovascular events.

Zeller et al. use RNA expression levels, without considering of DNA variants, to learn more about BP, an approach that others had described before⁶: RNA expression levels themselves are utilized to find signatures associated with BP elevation, as opposed to experiments that use gene expression linked to genotype data (eSNPs) to identify likely links between variants and genes. Zeller et al. make use of 4 studies with a total sample size of 4,539 individuals for which whole blood RNA was assessed on microarrays. They identify 8 transcripts that were independently replicated to be associated with BP and that explain, collectively, 13% of the BP variance. Additionally, Zeller et al. show that 7 of these transcripts are associated with BP changes over time and one transcript (CRIP1) is associated with cardiac hypertrophy; protein levels of CRIP1 are associated with stroke; however, for all these latter findings no replication effort was undertaken.

While being much more dynamic and related to function, the use of RNA has several distinct disadvantages over DNA: First, there is little possibility of distinguishing a consequence from a cause of hypertension using only these methods. The 13% variance explained by Zeller et al. are likely to be consequence of BP elevation rather than a cause for blood pressure elevation. It is therefore of course not possible to directly compare the variance explained by DNA variants with variation explained by transcript levels. But all genes associated with BP elevation are candidates for genes driving BP elevation and these require confirmation by other methods. Second, the expression patterns of RNA transcripts vary greatly between individuals, per cell type, and also between cells of the same tissue due to cell-cell heterogeneity such as transcriptional bursting⁷. Although there are few genes present exclusively in a particular tissue or with more variation between cell-types than individuals⁸, results from one cell type can only inform partly on another cell type. The causal tissue for primary hypertension is unknown, although there are some signals for cellular components of the large vessels^2-4. As vascular cell-types are only available in low numbers, it is unclear if similar expression signatures could be identified here. Expression profiles in peripheral leucocytes, as used in the experiments of Zeller et al., might be of relevance to primary hypertension as T-cells may be involved in hypertension pathogenesis⁹. Third, DNA genotyping is remarkably accurate using modern methods. For RNA expression measures there is much greater variability and there are substantial differences between methods with an apparent advantage when RNA sequencing is used, but at significantly higher pricing¹⁰.

In summary, the article of Zeller et al. outlines one possible step towards a different approach to unravel more of blood pressure genetics. There are limitations in the method, but the approach is interesting and future experiments, run in larger sample sizes and different tissues, would likely yield additional information. RNA expression profiles in many tissues at large numbers can be used for various additional kinds of experiments and would be very valuable in the endeavor to explain primary hypertension.