Mammalian atrial and ventricular cardiocytes display striking differences in phenotype. The bulk of atrial cardiocytes have, in addition to features commonly associated with striated cardiac muscle cell cytology, organelles that are normally found in cells engaged in the synthesis, vectorial transport and secretion of polypeptide hormones. These include an abundance of rough endoplasmic reticulum, a highly developed Golgi complex and storage granules containing hormonal products. These cytological features led to a series of past investigations that culminated in the discovery of the endocrine function of the heart [4
], which is largely mediated by the polypeptide hormones ANF and BNP.
From the data presented here it is evident that the phenotypic differences between the atria and the ventricles results from underlying differences in degree or presence/absence of several specific genes. Despite this complex background of differential expression, the microarray data reproducibility observed in different biological replicates was excellent.
The proportion of probes that were called present, as well as the number of differentially expressed genes was higher in the atria than in the ventricles. This observation could be related to a higher functional complexity of the atrial tissue, which include contraction, excitation, conduction, secretion and the presence of a profuse innervation [5
]. The detection of numerous known chamber specific transcripts [6
] supports the robustness of the GeneChip data and demonstrates that the differences in gene expression profiles observed between the atrial and ventricular tissues are not spurious in nature. A comparative analysis of the differentially expressed genes identified in other studies that are in common with our data is presented in the Additional file 5
. Several genes identified in these studies by others are in agreement with our data, even though there are experimental differences (such as species, gender and tissue selection) as well as data analysis between studies. However, none of these previous studies focused specifically on gene expression profiles relating to the endocrine function of the heart.
Based on atrial and ventricular phenotypical differences, the strategy used in the data mining process was to identify genes that had Gene Ontology annotations relating to hormone secretion, hormone activity, vesicle/granule packaging and transport. Screening for Gene Ontology terms classified as Golgi apparatus and ER cellular components was also undertaken in an effort to identify genes involved in the regulation of NP packaging and secretion. Differentially expressed transcription factors as well as transcripts coding for proteins involved in signal transduction such as ligands and receptors were also of great interest. Finally, hormone secretion can be modulated/regulated by ion channels and mechanosensors, hence genes coding for transport proteins were also of interest. Differentially expressed genes coding for proteins involved in metabolic processes as well as contractile proteins were not specifically looked at since these differences are expected to be found between the atrial and ventricular tissues but are not thought to be involved in the endocrine function of the heart. In general, functional classification of the differentially expressed genes according to their cellular components, identified in our study, parallels the results obtained by Barth et al.,[5
], which include ventricular abundance of mitochondrial genes as well as the identification of numerous atrial abundant ER and Golgi apparatus transcripts, such as Sec22 and Rab34. The larger number of probes hybridizing to cellular components involved in secretory function agrees with the phenotypic and functional differences between atrial and ventricular muscle. For example, atrial cardiocytes were more abundant in the transcripts coding for ANF and BNP than the ventricles. Correspondingly, ANF transcript abundance, which constitutes one to three percent of total atrial mRNA [10
] was found to provide for the top one percent of the GeneChip®
signal intensities. Saturation of the fluorescent signal is thought to have occurred for ANF probe sets, which explains the plateau level of 12 fold difference between atrial vs. ventricular tissues as compared to 163 fold with RT-PCR (Figure ). Oligonucleotide probe saturation can occur when the labeled cRNA is at high enough levels to result in an almost 100% probe/cRNA hybridization complex. Any other increase in expression would not be measurable due to the olignucleotide hybridization saturation. It is difficult to estimate the amount of cRNA needed to saturate the oligonulceotide probes since the specific number of oligonucleotide probes tiled onto the 112
micron surface area of the GeneChip®
is proprietary information.
Storage granule-associated proteins identified in this study include Ica1 [11
] and SGNE1. The evolutionary conserved gene Ica1, was shown to be abundant in human pancreas and heart, and moderately expressed in brain [12
]. Variable expression of 5'-untranslated region exons from the Ica1 gene was shown to be tissue specific, where cardiac tissue distinctively expressed the exon B2 transcript variant [13
]. The original observation that Ica1 expression levels were high in the heart but null in skeletal muscle lead Pietropaolo et al., [12
] to postulate that Ica1 was present in selective cells. We demonstrate here that Ica1 is expressed specifically in the atria given that all four ventricular replicates exhibited an absent detection call for this transcript (an absent call indicates that the microarray expression level for a particular transcript was below the threshold of detection).
Several studies have been able to co-localize SGNE1, a member of the secretogranin family, with neuropeptides and hormones in various tissues that are primarily composed of neuronal or endocrine cells, as well as tissues known to contain a sub-population of neuroendocrine cells [14
]. RT-PCR analysis further confirmed atrial abundance of SGNE1 (Figure ). SGNE1 has not been previously associated with the atria or the heart. Our own immunofluorescence preliminary data suggest that SGNE1 protein is localized in structures compatible with adrenergic varicosities (Figure ).
Differentially expressed transcription factors/DNA binding proteins that were more abundant in the atria than in the ventricles included numerous orphan nuclear receptors. Two out of the three Affymetrix probe sets assaying for nuclear receptor subfamily 2, group F, member 2 (Nr2f2) transcripts were identified in the atrial abundant gene list. Nr2f2 expression pattern is in agreement with previously published data (Additional file 5
) and real-time PCR analysis further validated its atrial abundance (Figure ). Other atrial abundant transcription factors include paired-like homeodomain transcription factor 2 (Pitx2), which is involved in atrial septation; T-box protein 5 (Tbx5), a known activator of the ANF promoter; and early growth response 1 (Egr1), which is up-regulated during endothelin, angiotensin II, ischemia and stretch treatments. Hairy/enhancer-of-split related with YRPW motif 1 (Hey1), as well as its activator Notch1, were more abundant in the ventricles than in the atria. Hey1 is thought to repress ANF expression in the ventricles [15
]. The known tissue specificity of several of these transcription factors further illustrates the soundness of the dataset and the identification of novel atrial abundant DNA binding proteins warrants further attention.
Of special importance to the understanding of the regulation of hormone secretion by the atria are genes that encode for molecular species involved in mechanical aspect of atrial activity given that a major stimulus for ANF and BNP secretion is atrial muscle stretch. This phenomenon is referred to as "stretch-secretion coupling" [16
]. Questions still remain regarding the exact stimulus perceived by the atrial cardiocytes as well as the precise signal transduction cascade that leads to an increase in ANF and BNP secretion following stretch. Therefore, the finding that mechanoreceptors (mechanically gated ion channels) as well as mechanosensors (PLA2) are most abundant in the atria is central to the generation and the testing of hypothesis. Because some prostaglandins are known to be potent stimulators of ANF synthesis and secretion [17
], it is of interest that the mechanosensor PLA2 is abundantly expressed in the atria. Activation of PLA2 following plasma membrane stretch could cause the release of arachidonic acid, a substrate for prostaglandin synthase.
The rapid secretory response following stretch of atrial cardiocytes as compared to the response to other ANF secretion agonists such as endothelin-1 (ET-1) [18
] has long suggested that the phenomenon of stretch-secretion coupling is related to mechanosensitive ion channels. It has been suggested [19
] that KATP
channels are involved in stretch-secretion coupling. These channels are stretch-sensitive [23
], are known to couple to G proteins, including Gαo
], and have been associated with other secretory processes [26
]. The results and conclusions reached from investigations on the involvement of KATP
channels on ANF secretion have often been contradictory. An inspection of the literature suggests that a likely reason for these discrepancies may lie on differences in experimental systems (in vivo or in vitro, perfused or non perfused tissues, etc.) [19
]. Most importantly, pharmacological studies have been made with little reference to the several candidate ion channels that may be present in the atria. The present investigation shows that Kir6.1 is significantly expressed in the ventricles. The ATP binding cassette SUR2, whose expression level (1.79 fold) didn't meet the fold cut-off criteria was more abundant in the ventricles, and, while not significant, Kir6.2 and SUR1 had marginally higher expression in atria.
Three other channels had significantly higher levels of expression in the atria over the ventricles. These included the inward rectifiers Kcnj3, Kcnj5 and the background channel TREK-1. Kcnj3,5 channels, which are activated by G protein coupled receptors (GPCRs) coupled to pertussis toxin (PTX) sensitive Gi/o
proteins, exhibit mechanosensitive properties and might participate in a mechanoelectrical feedback pathway regulating ANF and BNP secretion [29
]. This view is in line with our recent finding that PTX can abolish stretch secretion coupling [31
]. The background channel TREK-1 produces an outwardly rectifying current and is a member of the 2 pore domain potassium (K2P
) channel family. TREK-1 has been characterized as a stretch-activated membrane channel. It is primarily expressed in the central nervous system but has been found in cardiac tissues [32
] as well as in specific endocrine cells, such as adrenocortical cells [33
]. Atrial abundance of TREK-1 was further validated by RT-PCR (Figure ). Immunohistochemical analysis localized TREK-1 protein in isolated adult rat atrial cardiocytes and it is suggested that TREK-1 may act as a regulator of ANF secretion [34
The complex signal transduction pathways resulting from the activation of GPCRs is in part orchestrated by an assorted population of G proteins, assembled from a various combinations of 20 Gα, 5 Gβ and 12 Gγ isoforms [2
]. Pathway analysis of the microarray data revealed differential expression of genes involved in G protein signaling pathways between atrial and ventricular tissues. The G protein Gαo
isoform 1 was found more abundant in the atria by both microarray and RT-PCR analyses and was demonstrated by immunohistochemistry to co-localize with ANF in atrial granules and in the sarcolemma [31
]. Differential expression of the Gβ and Gγ subunit isoforms was also observed between the atria and ventricles but the functional significance of the differential G protein subunit diversity is not well known.
The identification of accessory proteins involved in the modulation of G protein signal transduction has helped to gain insight into the mechanisms involved in the specificity of a cellular response from an assortment of external stimuli. RGS proteins are capable of inhibiting signaling by acting as a GTPase Accelerating/Activating Protein (GAP), which accelerate the intrinsic rate of GTP hydrolysis of the Gαi/o
]. The preponderance of genes encoding for RGS proteins appears to be tissue specific. RGS2 and RGS6 were abundantly expressed in the atria while RGS5 was found at higher levels in the ventricles. These findings are confirmed by other studies in which RGS mRNA and protein levels were measured within atrial and ventricular cardiocytes [35
In addition to the previously mentioned accessory proteins, in this work we identified the atrial specific transcript Rasd1, also known as AGS1. AGS1 was found more abundant in the atria by both microarray and RT-PCR analyses. In the absence of receptor stimulation, AGS1, a member of the Ras subgroup of small G proteins, specifically activates Gi/o
heterotrimeric signaling pathways by promoting nucleotide exchange on Gαi/o
proteins and enhancing Gβγ uncoupling from Gαi/o
]. Since stretch secretion coupling of ANF and BNP is mediated by PTX-sensitive Gαi/o
, the differential expression of AGS1, as well as the other accessory proteins for G proteins identified in this investigation, is of particular interest. Due to AGS1 opposing effects on Gi/o
mediated signaling following agonist dependent or independent GPCR activation [39
], AGS1 may help orchestrate the complex diversity and specificity of G protein signaling under different conditions.