Effects of vitamin A deficiency in the postnatal mouse heart: role of hepatic retinoid stores

We demonstrate heart-wide expression of retinoic acid receptor-γ, a retinoic acid-activated nuclear receptor, and report that ventricles from adult mice with depleted hepatic vitamin A stores have altered gene expression, a significant increase in proliferating progenitor cells, and improved response to acute myocardial injury.

Abstract

To determine whether hepatic depletion of vitamin A (VA) stores has an effect on the postnatal heart, studies were carried out with mice lacking liver retinyl ester stores fed either a VA-sufficient (LRVAS) or VA-deficient (LRVAD) diet (to deplete circulating retinol and extrahepatic stores of retinyl esters). There were no observable differences in the weights or gross morphology of hearts from LRVAS or LRVAD mice relative to sex-matched, age-matched, and genetically matched wild-type (WT) controls fed the VAS diet (WTVAS), but changes in the transcription of functionally relevant genes were consistent with a state of VAD in LRVAS and LRVAD ventricles. In silico analysis revealed that 58/67 differentially expressed transcripts identified in a microarray screen are products of genes that have DNA retinoic acid response elements. Flow cytometric analysis revealed a significant and cell-specific increase in the number of proliferating Sca-1 cardiac progenitor cells in LRVAS animals relative to WTVAS controls. Before myocardial infarction, LRVAS and WTVAS mice had similar cardiac systolic function and structure, as measured by echocardiography, but, unexpectedly, repeat echocardiography demonstrated that LRVAS mice had less adverse remodeling by 1 wk after myocardial infarction. Overall, the results demonstrate that the adult heart is responsive to retinoids, and, most notably, reducing hepatic VA stores (while maintaining circulating levels of VA) impacts ventricular gene expression profiles, progenitor cell numbers, and response to injury.

NEW & NOTEWORTHY

We demonstrate heart-wide expression of retinoic acid receptor-γ, a retinoic acid-activated nuclear receptor, and report that ventricles from adult mice with depleted hepatic vitamin A stores have altered gene expression, a significant increase in proliferating progenitor cells, and improved response to acute myocardial injury.

vitamin a (VA), retinol, is an essential nutrient throughout life. It has two important functions. It is the precursor for 11-cis retinaldehyde, the chromophore that is required for vision, and for retinoic acid (RA), which acts through signaling pathways to influence gene expression (Fig. 1). A large literature has led to the well-supported view that RA plays a role in morphometric movement and tissue patterning events in the developing vertebrate embryo (34, 55). An equally large literature illustrates the role of RA in guiding the differentiation of precursor cells into different cell lineages depending on RA concentration and time of availability and RA signaling cross talk with other growth factors and cell signaling processes (2, 11, 20, 45, 50). Although it is clear that vertebrate life after birth would not be possible without dietary intake of a source of VA, the role of and mechanisms for VA involvement in maintaining postnatal cell and tissue processes are not yet understood.

Fig. 1.

Schematic showing major routes of use and storage of dietary vitamin A (VA). VA, retinol, obtained from the diet can be 1) converted to retinoic acid (RA) via a 2-step biosynthetic process catalyzed by retinol dehydrogenases (RDH) and retinaldehyde dehydrogenases (RALDH); 2) converted to retinyl esters that are stored in tissues in a reaction catalyzed by esterifying enzymes, such as lecithin retinyl acyl transferase (LRAT); or 3) converted to the retinaldehyde moiety that is the chromophore that enables vision. RAR, RA receptor.

There is an absolute requirement for VA in heart development during embryogenesis, with nutritional or biochemical deprivation resulting in abnormal heart development and death in utero (35, 55). In this context, there exists a high-fidelity system that regulates the availability and activity of RA in space and time, with functional gradients orchestrating the formation of a functional cardiovascular system (14, 34). Available evidence showing that carotenoids, RA, and nutritional VA deficiency (VAD) affect the recovery of the postnatal heart from injury supports the idea that retinoids continue to play a role in the adult cardiovascular system (10, 16, 19, 25, 31, 33, 38, 40) although potential pathways and mechanisms for effects of pro-VA and VA derivatives on postnatal heart health are far from being understood.

VA homeostasis is maintained by an elaborate multitissue system of molecular checks and balances that maintains stable serum retinoid levels and provides tissue and cell-specific access to adequate levels and species of retinoids to promote retinoid-dependent processes, including vision, cell growth and differentiation, and tissue maintenance and repair. In normal mice, dietary pro-VA (obtained from plant carotenoids) and preformed VA (obtained from animal sources) are converted to retinyl esters in the intestine and packaged into chylomicrons at an experimentally calculated ratio of ∼90% retinyl esters and 10% free retinol. Under conditions of vitamin A sufficiency (VAS), most of the retinol and retinyl esters leaving the intestine are taken up by the liver. Hepatocyte retinyl ester hydrolases hydrolyze incoming retinyl esters, and the free retinol is then reesterified by lecithin retinyl acyl transferase (LRAT) and stored in hepatic stellate cells. It is estimated that 66–75% of dietary retinoid is taken up and stored as retinyl esters in the mammalian liver (21, 26, 36, 43). By yet unknown mechanisms, these hepatic retinyl ester stores can be called upon to release retinol into the circulation bound to retinol-binding protein (RBP), presumably for delivery to and use by extrahepatic tissues, particularly during times of dietary or disease-induced VAD.

LRAT is expressed in many tissues and is the primary enzyme involved in retinol esterification in the liver. LRAT gene knockout (KO) mice have been generated by two different laboratories using two different genetic approaches. These mice develop normally and are, on a gross level, phenotypically indistinguishable from wild-type (WT) mice. Postnatal LRAT^−/− mice develop defective vision because of compromised rod and cone function brought about by a lack of adequate levels of retinoids to support regeneration of 11-cis-retinaldehyde (5, 28). Additionally, LRAT^−/− mice are more susceptible to the development of VAD than are WT mice when they are fed a diet lacking a source of VA (28, 36).

Both transgenic LRAT^−/− strains present with a similar retinoid status. The reported circulating levels of retinol in LRAT^−/− mice fed a VAS diet (LRVAS) mice are unchanged relative to WT controls (0.5–1.7 μM), but serum levels fall to negligible levels in LRAT^−/− mice fed a VAD diet (LRVAD) for 4–6 wk (≤0.13 μM). Both liver retinol and retinyl ester levels drop to negligible or nondetectable levels in LRAT^−/− animals regardless of whether the mutant mice are fed a VAS or a VAD diet. Lung retinol and retinyl ester levels also show this same precipitous decline in LRVAS and LRVAD animals (28, 36). These results confirm many previous observations showing that serum retinol levels are under tight homeostatic control and that retinol storage in the liver, lung, and eye is facilitated by LRAT.

Presently, the mechanisms by which the adult heart regulates retinol and RA under baseline conditions or in response to injury are incompletely understood. We sought to examine the effects of VA status on the postnatal heart. The LRAT^−/− mouse provides a physiological model to investigate this without induction of hepatic injury. Although gross cardiac histology appears normal in the young LRAT^−/− mouse, changes in the gene expression profiles of both right and left ventricles (RV and LV) and cell-type specific increases in the number of Sca-1 cardiac progenitors in the hearts from LRVAS mice suggest that reduced VA stores and nutritional VAD have important impacts on cardiac homeostasis, and echocardiographic data indicate that depletion of hepatic retinoid stores improves the outcome from acute myocardial injury.

METHODS

Animals

LRAT gene KO mice (LRAT^−/−) bred into a C57BL/6J background were a gift of Krzysztof Palczewski, Case Western University. Genetically matched WT control mice were obtained from the Jackson Laboratory (Bar Harbor, ME). Genotypes were confirmed by analyzing tail genomic DNA with primers in a polymerase chain reaction as described (5). Animals were housed in an animal care facility at Tennessee State University (TSU) in accordance with a protocol approved by the TSU Animal Care and Use Committee and guidelines set forward in the National Institutes of Health (NIH) Guide for the Care and Use of Laboratory Animals. The TSU colony was fed a maintenance diet (Lab Diet 5001; 15 IU VA acetate/g diet). Breeders were fed a breeder diet (Formulab diet 5008; 15 IU VA acetate/g diet). All surgical procedures for myocardial infarction (MI) and echocardiography were performed in the Cardiovascular Pathophysiology and Complications Core in the Vanderbilt University Mouse Metabolic Phenotyping Center in accordance with a protocol approved by the Vanderbilt Institutional Animal Care and Use Committee.

LRAT^−/− mice and WT mice used for flow cytometric analysis of Ki-67 in cardiac cell suspensions were fed a standard lab chow diet (Teklad Global 18% protein rodent diet supplemented with 15 IU VA acetate/g diet) at Maine Medical Center Research Institute (MMCRI) in accordance with a protocol approved by the MMCRI IACUC.

Animal Protocols for the Microarray Analysis

Age-matched, male mice were group housed in plastic cages with Alpha-Dri bedding and were maintained in their own air system in a temperature-controlled room with a 12-h:12-h light/dark cycle. Acidified water was supplied in water bottles. Custom VAD (no. CSMD130C) and VAS (no. 960220) color-coded, pelleted diets were prepared by MP BioMedical. The VAS diet was the same as the VAD diet but supplemented with 250,000 U/kg retinol palmitate (4 IU retinol equivalents/g diet) as a source of VA. WT mice were fed the VAS diet (WTVAS), and LRAT^−/− mice were fed either the VAS diet (LRVAS) or the VAD diet (LRVAD). The mice were given acidified water (pH 2.8–3.2) in water bottles and fed ad libitum. After 50 days on the diet, all animals were killed over three days.

Tissue Collection

All steps were carried out under dim yellow light. One mouse in each of the three groups was anesthetized with a ketamine/xylazine cocktail; all remaining mice were anesthetized with isofluorane. When the mouse no longer responded to a pinch test, an incision was made just below the rib cage to expose the liver. A midline cut was made to the top of the thorax. The sternum was lifted and pinned back, and the diaphragm was cut to partially expose the heart.

LV and RV.

The heart was dissected from the chest cavity and immersed in RNAlater to prevent RNA degradation. The RV were separated from the LV by cutting along the septum. The RV and LV were cut into small pieces and placed in separate tubes with fresh RNAlater. The tissues were stored in RNAlater for 1–3 days at room temperature (as recommended by the vendor), then centrifuged to remove excess fluid, and stored at −80°C until preparation of total RNA.

Liver.

One lobe of the liver was removed, rinsed in ice-cold PBS, transferred to an amber microfuge tube, snap frozen in liquid nitrogen, and stored at −80°C for high-performance liquid chromatography (HPLC) analysis.

Immunohistochemistry

Mice hearts were fixed in 4% paraformaldehyde for 4 h, submerged in 30% sucrose overnight, and cryoprotected with optimal cutting temperature compound. Cryosections (5 μM each) were permeabilized in 0.4% Triton X-100, washed in Tris-buffered saline containing 0.05% Tween 20 (TBST) and then TBS, blocked for 1 h in 0.01 M Tris·HCl, pH 7.4, 2% BSA, 2% normal goat serum, followed by overnight incubation at 4°C in 1/125 purified rabbit polyclonal antibodies raised against RA receptor-γ (RAR-γ) (7) diluted in blocking buffer, 1/100 Ki-67 antibody (cat. no. 15580; Abcam), 1/20 CD31 (cat. no. 553370; BD Pharmingen), or 1/100 α-actinin antibody (cat. no. A7811; Sigma). Detection was performed with goat anti-rabbit Alexa 488 secondary antibody (1/500, cat. no. 811034; Life Technologies) or goat anti-mouse Alexa 647 (1/500, cat. no. A-21235; Life Technologies). F-actin was stained with phalloidin coupled to Alexa 568 (1/200, cat. no. A22287; Life Technologies). Nuclei were stained with either DAPI or TO-PRO-3 (1 μM in TBS). Coverslips were mounted using Molecular Probes Prolong Gold antifade reagent (Invitrogen). Images were acquired using a Zeiss LSM 510 Meta confocal microscope and a Leica Aperio ScanScope FL digital slide viewing system with ImageScope v11 software. Images were edited, and image levels were adjusted in Adobe Photoshop CS6.

Flow Cytometry

Murine hearts were perfused with PBS/heparin to remove blood, dissected to isolate ventricular tissue, minced, and incubated with collagenase II/dispase II/DNase I/CaCl₂ for 20 min at 37°C. Cells were treated with TruStain fcX (BioLegend) to block nonspecific binding. The cells were then incubated with relevant antibodies for 20 min at 4°C, washed once with 10× volume of cold PBS/BSA/EDTA.

For intranuclear staining to detect Ki-67, cells were fixed and permeabilized using nuclear factor fixation permeabilization buffer and stained with phenylephrine (PE)-conjugated antibody against Ki-67 (16A8) or rat IgG2a-PE isotype-matched control (RTK2758) (BioLegend). Data acquisition was performed using MacsQuant Analyzer 10 (Miltenyi Biotec), and the data were analyzed with WinList 5.0 software. Nonviable cells and cell debris were excluded by using LIVE/DEAD Fixable Violet staining kit (Life Technologies). Antigen negativity was defined as having the same fluorescent intensity as the control isotype.

Extraction and Analysis of Liver Retinol and Retinyl Esters

All analytical procedures were carried out under yellow light. Approximately 0.2 g liver was homogenized in 10 volumes of ice-cold PBS with a Telmar Ultra Turax T25 homogenizer for 15 s × 2 on setting 10. The homogenate (400 μl) was mixed with an equal volume ice-cold 100% ethanol, and retinyl acetate, dissolved in ethanol, was added as an internal control for extraction efficiency. The mixture was vortexed, and three volumes hexane were added, followed by vortexing for 2 min to extract the retinoids. Following centrifugation at 1,000 g × 4°C × 5 min, the hexane layer was transferred to 0.5 ml PBS, triturated, and recentrifuged. The hexane layer was dried down under a steady stream of argon and resuspended in 0.1 ml methylene chloride. The resuspended extract was separated according to a procedure adapted from O'Byrne et al. (36). Ten-microliter extract was separated isocratically with 70% acetonitrile/15% methanol/15% methylene chloride at a flow rate of 1.5 ml/min using an Agilent 1100 HPLC system, fitted with an automatic injection sampler, guard column, Agilent Zorbax C18, 4.6 mm × 12.5 cm, wide bore column, and diode array detector set at 325 nm. Sample peaks were identified by comparing their retention times and spectra with those of authentic standards. Tissue retinoid concentrations were calculated using integrated areas under sample peaks and the best fit least squares regression curves for appropriate standards (retinol, y = 19.2x − 44.7, r² = 0.998; retinyl acetate, y = 9.5x + 23.9, r² = 0.998; retinyl palmitate, y = 12.6x + 37.1, r² = 0.996).

RNA Isolation

Total RNA was isolated from the RV and LV of the same mouse (3 mice per control and experimental group) using a modified guanidine isothiocyanate-cesium chloride centrifugation method (RV) or TRIzol RNA Isolation Reagents (Thermo Fisher Scientific, LV). Samples were treated with DNase and stored at −80°C.

Affymetrix Microarrays

Total LV and RV RNA was submitted to the GSR Microarray Core at Vanderbilt University. Quality of RNA was assessed, and only samples achieving an integrity rating of 6.8 or higher were used for transcriptome analysis (Table 1).

Table 1.

Characteristics of mice and ventricle tissue RNA used for transcriptome analyses

				Quality of Total RNA (Integrity)
Animal ID	Age at Sacrifice, months	Body Weight at Sacrifice, g	Heart/Body Weight Ratio at Sacrifice, g/g	RV	LV
WTVAS 315M	7	30	0.0033	7.4	7.6
WTVAS 233M	8	33	0.0030	7.2	7.1
WTVAS 250M	7	31	0.0032	8.0	7.7
LRVAS 293M	5	26	0.0038	7.7	7.3
LRVAS 224M	8	38	0.0035	7.2	8.0
LRVAS 247M	7	37	0.0031	7.9	7.9
LRVAD 302M	5	39	0.0032	7.7	6.8
LRVAD 230M	8	32	0.0033	7.4	7.6
LRVAD 246M	7	39	0.0040	7.8	7.3

WT, wild-type; VAS, vitamin A sufficient; LR, mice lacking liver retinyl ester stores; VAD, vitamin A deficient; RV, right ventricle; LV, left ventricle.

Sample RNA was normalized to 43.33 ng/μl and amplified with an AMBION RNA amplification kit (cat. no. 4411974) designed to generate amplified sense-strand cDNA ready for fragmentation and labeling with an Affymetrix GeneChip WT Terminal Labeling Kit (Affymetrix) using a Beckman Biomek FXP Robot. A sample (130 ng) was added to a reaction plate along with poly-A RNA spike-in controls to measure efficiency of target amplification. The reaction plate was loaded onto a robotics deck and automated first-strand cDNA synthesis, second-strand cDNA synthesis, and in vitro transcription cRNA synthesis were performed. cDNA was purified with an Ambion-WT bead cleanup kit, and eluted samples were quantified. A sample (5.5 μg) of purified cDNA products was used in fragmentation and labeling reactions. A sample (10.5 μg) of cRNA was used in reactions subjected to random primer heat denaturation, followed by second-cycle cDNA synthesis and cRNA hydrolysis with RNase H. TdT-labeled cRNA was hybridized to mouse gene 1.0 ST arrays (26,166 RefSeq transcripts covered). Signals were detected by scanning with an Affymetrix Gene-Chip Scanner using AGCC v. 1.1.

All data were normalized using a robust multiarray average procedure. A one-way ANOVA was used to model variation in probe expression as a function of the three different treatment groups. To adjust for multiple testing, a false discovery rate (FDR) was calculated for the P values of the overall ANOVA F statistic using the p.adjust function in R (version 3.1.2) using method = “BH” (7). We chose an FDR threshold of 50% to achieve a good balance between false positives and false negatives, as this is an exploratory study. Significant pairwise comparisons between treatment groups were then determined using a P value threshold of 0.05 and an absolute fold change threshold of 1.5. Probes that were not annotated with a gene symbol were not included in heat maps. If multiple probes were annotated with the same gene symbol, one was randomly selected to include in heat maps. Heat maps were generated using the heatmap.2 method in R for all significant probes (as defined previously) for a given contrast with expression standardized with probe. Principle component analysis was performed using the prcomp method in R using the set of probes meeting the FDR < 0.5 threshold. Plots of each sample for the first three principal component values are provided for LV and RV data sets with the explained variance for each of the first three principle components included. Original and analyzed datasets are available at NCBI Gene Expression Omnibus (GEO), accession number GSE74906.

Bioinformatics Analysis

A mouse genome-wide search for RA response element (RARE) consensus motifs (44) composed of direct repeats (DR) of the hexanucleotide sequence [5′-(A/G)G(G/T)TCA-3′] separated by zero (DR0), one (DR1), two (DR2), five (DR5), or eight (DR8) nucleotides or an inverted repeat (IR0) (Table 2) was conducted using GenomePatternScan (GPS), a tool developed in the Congdon laboratory. The tool was also used to assess the presence and promiscuity of each variant in the 67 genes identified in the microarray analysis as showing significant changes in transcription in LRVAS and LRVAD heart tissue samples. Genomic data and annotations were collected from primary assembly of the Mus musculus genome, version GRCm38, Ensembl release 72 (13). GPS output provides links to the UCSC Genome Browser (http://genome.ucsc.edu/) (24) to explore the genetic context of each hit.

Table 2.

Number and percentage of 6 RARE motif variants in whole mouse genome vs. percentage of hits found in genes responsive to VA status in this study

RARE Motif	Consensus Sequence	Number of Hits Genome Wide	% of Hits Genome Wide	% of Hits Identified in 67 Genes	% of Hits Identified in 6 Selected VA-Responsive Genes*
DR0	rgktcargktca	28256	26.76%	27.88%	28.82%
DR1	rgktcanrgktca	16678	25.35%	20.05%	17.01%
DR2	rgktcannrgktca	27577	21.13%	29.72%	28.12%
DR5	rgktcannnnnrgktca	9306	9.86%	10.14%	9.49%
DR8	rgktcannnnnnnnrgktca	8383	12.68%	8.29%	8.55%
IR0	rgktcatgamcy	7854	4.23%	3.92%	8.01%
Total		98054	100.01%	100.00%	100.00%

Shown are the DNA sequences of the retinoic acid response element (RARE) queries and the number and percentage of hits returned; r = A/G; k = G/T; y = C/T; m = A/C; n = A/G/C/T.

^*See Table 10.

MI

WTVAS (n = 15, 7 male and 8 female) and LRVAS (n = 9, 5 male and 4 female) mice, 3–5 mo of age, ranging from 17–29 g, were selected for the study. Staff in the Cardiovascular Pathophysiology and Complications Core (Vanderbilt University) who carried out the investigation were blinded to the genotype and hypothesis in this experiment. Mice were anaesthetized with an intraperitoneal injection of 30 mg/kg of pentobarbital. Absence of a response to pinching the toe was used as an indicator of the appropriate level of anesthesia. Anesthetized mice were rapidly intubated and mechanically ventilated (tidal volume, 1 ml/100 g body wt; ventilation rate, 65 strokes/min) using a constant volume small animal ventilator (model 683; Harvard Apparatus). A left thoracotomy was performed at the fourth intercostal space, and the left coronary artery was ligated with a 6-0 suture loop. MI was confirmed by direct visualization of regional cyanosis of the myocardial surface distal to the suture, accompanied by S-T segment elevation on the electrocardiogram. Following successful induction of MI, the chest cavity was compressed to evacuate any air before being closed. The mice were given 0.05 mg/kg buprenorphine by subcutaneous injection immediately after surgery and then every 8–12 h for 2 days.

Echocardiography

Serial echocardiograms were performed on MI-induced WTVAS and LRVAS mice within 3 days before surgery (baseline) and 7 days post surgery. Transthoracic echocardiographic images of hearts were obtained using a Vevo 2100 Imaging System (VisualSonics) in mice immobilized under anesthesia (1.5–2.0% isoflurane). For M-mode recordings, the parasternal short-axis view was used to image the heart in two dimensions at the level of the papillary muscles. End-diastolic and end-systolic LV cavity dimensions were measured using software resident on the ultrasonograph. LV fractional shortening (FS) was calculated from M-mode-derived LV inner diameters in diastole (LVIDd) and systole (LVIDs) using the formula (LVIDd − LVIDs)/LVIDd × 100%. Imaging and interpretation of results were done by technicians who were blinded to the treatment group. Statistical analysis of the echocardiography data was performed on GraphPad using two-way ANOVA and the Bonferroni posttest for individual statistically significant differences between data sets.

RESULTS

RAR-γ Is Expressed in Postnatal Mouse Heart

WT mouse heart cryosections were reacted with an antibody directed against RAR-γ that has been extensively characterized (8). Immunoreactivity was detected in cells throughout the RV and LV of mouse heart sections (Fig. 2, A–D). The results indicate that RAR-γ is present in the nucleus of cardiomyocytes (Fig. 2, E–G, arrowhead), unidentified cells (Fig. 2, E–G, arrow), and endothelial cells (Fig. 2, H–J, arrow). RAR-γ staining was also evident in the cytosol (Fig. 2, F and I), a finding that is consistent with other reports that RARs are involved in processes outside of the nucleus (1).

Fig. 2.

RAR-γ expression in heart. Mouse heart sections were prepared as described in materials and methods. A–D: sections were stained with phalloidin (red), DAPI, or TO-PRO-3 (blue), and with an antibody directed against RAR-γ (green), and images were acquired with a Scanscope fluorescent digital slide viewing system. The images in A, B, and C are high-magnification views of the areas labeled right ventricle (RV), left ventricle 1 (LV1), and LV2, respectively, in D. E–J: images of a heart section acquired with a confocal microscope to detect nuclei (405-nm laser, E and H) and RAR-γ (488-nm laser, F and I). Merged images (405-, 488-, and 543-nm lasers) show localization of RAR-γ in the nuclei of phalloidin-stained regions (arrowhead) and unidentified cells (arrows, G, I, and J). Note, some RAR-γ staining is present in the cytosol (F and I). K: heart section reacted with antibodies directed against CD31 (green) and RAR-γ (red) and stained with DAPI (blue). Costaining of CD31-positive cells with RAR-γ antibody is indicated (arrows). L: confocal image of heart section reacted with anti-α-actinin (gray) and anti-RAR-γ (red). M: same section as shown in L, but showing nuclear localization (blue) of RAR-γ (red). Note colocalization of RAR-γ in the nuclei of α-actinin-positive cells (arrows). Scale bars represent 50 μm (C), 1 mm (D), 10 μm (G), 10 μm (K), and 25 μm (M). Images in E–J are at the scale shown in G, and images in L and M are at the scale shown in M.

Characteristics of LRVAS and LRVAD Mice

As has been reported by others (5) the LRAT^−/− mice were generally heavier and showed greater relative weight gains than age-matched and genetically matched WTVAS controls (Fig. 3). Genetic KO of the lrat gene was sufficient to deplete retinol and retinyl ester stores from the liver of postnatal LRVAS mice (Table 3). Removing all sources of VA from the diet of LRVAD mice for 50 days did not produce any of the hallmark symptoms of frank VAD, such as decreased growth rate (Fig. 3), alopecia, or ataxia. With continued feeding of a VAD diet for 6 mo, LRVAD mice exhibited a reduction in weight gain relative to LRVAS mice, which continued to show increased weight gains (unpublished laboratory observation). On the basis of this parameter, the 3-mo-old LRVAD animals used in this study represent a depleted state of VAD but not a completely deficient one.

Fig. 3.

Growth curves. Shown are means ± SD weights vs. days on the diet for wild-type mice on VA-sufficient diet (WT-VAS) (n = 7, ●), mice lacking liver retinyl ester stores fed a VAS diet (LR-VAS) (n = 6, ○), and LR mice fed a VA-deficient diet (LR-VAD) (n = 7, ▲) mice. For ease of viewing, SD bars are only shown in the positive direction. Total RNA was prepared from LV and RV tissue dissected from the same 3 mice in each group for the transcriptome analyses reported here. Note that y-axis begins at 20 g.

Table 3.

Liver retinol and retinyl ester levels

Animal Group ID	Diet	Days on Diet	n	Sex	Liver Retinol, nmol/g	Liver Retinyl Palmitate, nmol/g
WT	VAS	50	6	M	29.5 (23.1)	1371 (877)
LRAT−/−	VAS	50	5	M	nd	nd
LRAT−/−	VAD	50	6	M	nd	nd

Extracts were prepared and analyzed by high-performance liquid chromatography, as described in materials and methods. Shown are the number of animals in each sample group (n) and the mean (SD) for each measurement.

LRAT, lecithin retinyl acyl transferase; nd, not detectable.

VAD Does Not Affect the Size or Gross Morphology of the Postnatal Heart

There were no statistically significant differences in heart weight or heart/body weight ratios between WTVAS (n = 6), LRVAS (n = 7), or LRVAD (n = 7) animals, as indicated by a Kruskal-Wallis one-way ANOVA on ranks. Histochemical analysis of LRVAS and LRVAD hearts did not reveal gross alterations in heart morphology or tissue integrity (data not shown).

Affymetrix Microarray Analyses

Differential gene expression was assessed by an Affymetrix microarray analysis of RNA extracted from the RV and LV of three age-matched WTVAS, three LRVAS, and three LRVAD mice, for a total of 18 microarrays. With the use of the selection criteria described in materials and methods, the number of probes that were downregulated or upregulated in all three experimental groups (LRVAS, LRVAD, and WTVAS control tissues) in LV and RV are shown in Fig. 4A. On the basis of the current gene annotations, excluding duplicates, these probe sets map to 81 distinct genes. Principal component analysis (PCA) of probes found to be significant at FDR < 0.5 was used to visualize sample separation attributable to variability in the differentially expressed genes in the LV (Fig. 4B) and RV (Fig. 4C). The PCA indicated that outcomes segregated into discrete groupings corresponding to sample treatments.

Fig. 4.

Overview of gene expression analysis. A: bar graph showing number of genes found to be significantly different in cardiac ventricular tissues from WTVAS, LRVAS, and LRVAD mice. Shades represent the section of the heart in which significant differences in gene expression were found (RV only, black; LV only, medium gray; both RV and LV, light gray). B: principle component (PC) plot of the first 3 PCs accounting for 99.8% of the variance for all 114 significant probes meeting the false discovery rate (FDR) < 0.5 threshold in the overall ANOVA in the LV. C: PC plot of the first 3 PCs accounting for 99.7% of the variance for all 82 significant probes meeting the FDR < 0.5 threshold in the overall ANOVA in the RV.

LV and RV Have Distinct Gene Expression Profiles

Hierarchical clustering of probes detected as significantly different indicated clear differences in the expression of transcripts in the LV and RV of WTVAS, LRVAS, and LRVAD animals (Fig. 5). Relative to WT controls, a total of 53 genes showed either increased or decreased expression in RV and LV from LRVAS animals, but only 12 of these genes were affected in both ventricles. Similarly, only 13 of 68 total genes were altered in both the RV and LV of LRVAD animals (Fig. 4A).

Fig. 5.

RV vs. LV heat map. Hierarchical clustering of all probes detected as significantly different in either RV or LV of WT mice using the following criteria: FDR < 0.5 for overall ANOVA P value; P < 0.05 and absolute fold change > 1.5 for at least 1 of the 3 contrasts; the probe is annotated with a gene symbol; and the probe represents a unique gene symbol (or if multiple probes were found to be significant for a given gene symbol, 1 was randomly selected).

Genetic and Nutritional Changes in VA Status Alter Gene Expression Profiles in Mouse Heart

Heat maps illustrating relative expression levels of probes found to be significantly different in ventricles of experimental animals relative to WTVAS controls are shown in Figs. 6 and 7. Twenty-eight probes were significantly altered in LRVAS RV (Fig. 6A) and 31 in LRVAD RV (Fig. 7A) animals relative to WTVAS RV controls. Thirty-seven probes were significantly altered in LRVAS LV (Fig. 6B) and 50 in LRVAD LV (Fig. 7B) animals relative to WTVAS LV controls.

Fig. 6.

Heat map depicting specific effect of LRAT knockout on gene expression. Hierarchical clustering of significant probes for LRVAS vs. WTVAS contrast (FDR < 0.5 for overall ANOVA; absolute fold change > 1.5 and P < 0.05 for contrast) for RV (A) and LV (B).

Fig. 7.

Heat map depicting specific effect of combined LRAT knockout with nutritional deficiency in VA. Hierarchical clustering of significant probes for LRVAD vs. WTVAS contrast (FDR < 0.5 for overall ANOVA; absolute fold change > 1.5 and P < 0.05 for contrast) for RV (A) and LV (B).

Genes Associated with Retinoid Metabolism

Three genes associated with retinoid metabolism were observed to show differences in expression in the experimental animals in the microarray screen (Table 4). There was a 2–3-fold increase in the expression of the truncated lrat transcript, which retains expression in the LRAT KO, in the RV and LV of LRVAS and LRVAD mice. The probe for retinaldehyde reductase (dhrs3) was significantly decreased in LRVAD RV, but not LV (Table 4), whereas the probe for cellular RBP III (rbp7) exhibited a graded increase in LRVAS (up 1.7-fold) and LRVAD (up 2.3-fold) RV (Table 4).

Table 4.

Genes involved in retinoid metabolism identified in microarray screen

	Right Ventricle		Left Ventricle
Gene	LRVAS vs. WTVAS	LRVAD vs. WTVAS	LRVAS vs. WTVAS	LRVAD vs. WTVAS
Dhrs3, a retinaldehyde reductase	−1.2 (0.02)	−1.5 (0.0004)	ns	ns
	FDR = 0.37	FDR = 0.37
Lrat, transfers ester to retinol	3.0 (7.9 e−7)	2.6 (1.6 e−6)	2.7 (5.3 e−7)	2.0 (3.9 e−6)
	FDR = 0.008	FDR = 0.008	FDR = 0.007	FDR = 0.007
Rbp7, cellular retinol binding protein III	1.7 (0.009)	2.3 (0.001)	ns	ns
	FDR = 0.56	FDR = 0.56

Values in unbolded font failed 1 of 3 parameters for significance but are included to show trends. Values represent significant fold changes in transcript expression relative to WTVAS control tissues.

ns, not significantly different than WTVAS; FDR, false discovery rate.

In Silico Analysis of Significant Genes Identified in the Microarray for Putative Retinoid Response Elements

Canonical RA signaling is facilitated by RA activation of RAR-RXR heterodimers that bind to DNA RARE, which then either upregulate or downregulate gene transcription. A search of the mouse genome was carried out with DR0, DR1, DR2, DR5, DR8, and IR0 RARE consensus motifs. A total of 98,054 sites were detected in the regions between the beginning and end of the gene and nearest upstream and downstream neighboring genes and, in some cases, introns. RARE variants representing all six motifs were detected, with the number of DR0, DR1, and DR2 sites accounting for 73% of the hits (Table 2).

Fourteen probes identified in the microarray screen that are the products of uncharacterized genes were removed from further analysis. The remaining 67, which encode known proteins (Tables 4–8), were analyzed for the presence of putative RARE. Fifty-seven of these genes were found to have RARE (Table 9). From 1 to 39 RARE variants were found either upstream or downstream from a gene or in an intron. No patterns were observed in the number or type of RARE relative to gene function, VA responsiveness, or direction of change in expression in the microarray. All but two genes (ctse and lect1) had at least one DR1 or one DR2 RARE, and in most cases (32/57) both motifs were present. All three genes involved in retinoid metabolism (Table 4) had both DR1 and DR2 motifs (Table 9).

Table 8.

Genes not known to be associated with cardiovascular function whose transcripts were significantly affected by VA status in LV

Gene	LRVAS vs. WTVAS	LRVAD vs. WTVAS	Function
Genes associated with energy balance
Atp5e, ATP synthase, H+ transporting, mitochondrial F1 complex	1.4	1.5	Encodes a subunit of mitochondrial ATP synthase
Genes associated with immunity and inflammation
Cd24a antigen	1.8*	2.5*
Cd38 antigen		−1.6†
Cd209a antigen	−5.5	−5.6	RA induces Cd209
Clec2e, C-type lectin domain family 2, member e	1.8	2.0	Lectin-like receptor expressed in myeloid cells and NK cells
Colec11, collectin subfamily member 11	−1.8	−1.7	COLEC11 genes encode soluble pattern recognition molecules in the lectin pathway of complement
Ifi44l	2.2	2.6	Interferon-induced protein 44 like
Klrb1b	1.4	1.5	Killer cell lectin-like receptor subfamily B member 1B
Lect1	−2.8	−2.9	Leukocyte cell-derived chemotaxin
Genes associated with cell proliferation
Mki67, antigen identified by monoclonal antibody Ki-67		1.4‡	Cellular marker for proliferation
Genes associated with chromosome organization
H2afj, H2A histone family, member J	1.3	1.5
Mcm6, minichromosome maintenance deficient 6	1.9	2.2
Rsl1, regulator of sex-limited protein 1	−2.1	−2.2	Zinc finger regulatory transcription repression factor; regulates sex- and tissue-specific promoter methylation and dynamic hormone-responsive chromatin configuration
Genes associated with growth, differentiation, and maintenance
Filip1l, filamin A-interacting protein 1-like	−1.1	−1.5	Regulator of the antiangiogenic activity on endothelial cells
Rerg, RAS-like, estrogen-regulated, growth inhibitor	2.2	2.1	In cell lines, overexpression leads to reduced proliferation rate, colony formation, and tumorigenic potential
Genes encoding receptors, channels, and signaling proteins
Gpr111	−1.5	−1.5	G protein-coupled receptor 111
Lgr6	2.1	2.0	Leucine-rich repeat-containing G protein-coupled receptor 6
Mrgprf	−1.5	−1.7	MAS-related GPR, member F
Tmc7	1.7	1.5	Transmembrane channel-like gene family 7
Genes encoding enzymes
Mboat2	−1.9	−1.8	Membrane bound O-acyltransferase domain containing 2
Gdpd3	−33.5	−35.9	Glycerophosphodiester phosphodiesterase domain containing 3
Glo1	−1.6	−1.6	Glyoxalase 1
Hgsnat	−1.4	−1.5	Heparan-α-glucosaminide N-acetyltransferase
Hnmt	−2.4	−2.5	Histamine N-methyltransferase
Mut	−1.6	−1.7	Methylmalonyl-coenzyme A mutase
Pm20d1	−1.6	−1.8	Peptidase M20 domain containing 1
Vma21, VMA21 vacuolar H+-ATPase homolog (S. cerevisiae)	−1.2	−1.5
Genes encoding other proteins
Ctse, cathepsin E	1.9	1.6
Fbxo44, F-box protein 44	−3.4	−3.1
Park2, Parkinson disease (autosomal recessive, juvenile) 2	−1.6	−2.1	Parkinson disease (autosomal recessive, juvenile) 2
Pkd2l2, polycystic kidney disease 2-like 2	−1.4	−1.5
Rsph3a, radial spoke 3A homolog (Chlamydomonas)	1.6	1.9
Zfp738/945, zinc finger proteins	−1.5	−1.7

Bold values are significantly affected by VA status in the left ventricle. Values in unbolded font failed 1 of 3 tests for significance.

^*Cd24a had an FDR = 0.67 and P < 0.040 for LRVAS and P < 0.007 for LRVAD.

^†Cd38 had an FDR = 0.59 and P < 0.003 for LRVAD.

^‡Mki67 had an FDR = 0.67 and P < 0.032 for LRVAD.

Table 9.

Genes and RARE motif number and type

Gene	DR0	DR1	DR2	DR5	DR8	IR0	Total
angpt2	2	1	2			1	6
aoah	4		2	2		1	9
atp5E			1				1
casq2		2	5				7
cd209a	3		2	1			6
cd24a	2	1	1				4
cd38	2	2	2		1		7
clec2e		1					1
colec11		1	2		1		4
cox18	5		5		1	1	12
ctse	1						1
dhrs3	2	1	2	1	1		7
Elovl7	2	1	2				5
fbxo44	2		1		1		4
filip1l	3	3	3	1	1		11
glo1			4				4
gpr111		2					2
hal			1	1			2
hgsnat			1		1		2
hn1l		1	1				2
hnmt	3	2	2	6		1	14
Ifi44l			2	1		1	4
Itpripl2		3	1	1			5
klrb1b	1	1	1				3
lect1	1			1	1		3
lgr6	2	1	4	1	1	1	10
limch1	2	5	3	1	1	1	13
lrat		1	1	1	3		6
mboat2	2		5		8		15
mcm6		1					1
Mett2l21d	2	2	1	1			6
mgp	2	1	1				4
mki67	11	11	8	4	3	2	39
mut	10	8	11	2	1	1	33
nppb		2	1				3
npr1	2	1	2	1	1		7
npy	5	6	2	4	1		18
park2	11	7	10	6		5	39
pdpn	6		2	1	1		10
pfkfb1		1					1
pkd2l2	3	1	2	1		1	8
pkig	2		1				3
pm20d1	2	1		1	1		5
ptgfr	2	1	6		1		10
rbp7	1	1	1		2		5
rerg	4	1	4		1	1	11
rsg1		1					1
rsl1			1				1
rsph3a	2	1					3
rsph3b	3	1	3				7
sfrp2	4	3	6	3	2		18
tcea3	3		1		1		5
tmc7		3	1				4
ubiad1	5	2	2	1			10
vma21	1	2	3	1			7
zfp87			1				1
zfp945	1		3				4
Totals	121	87	129	44	36	17	434

Number and type of RARE motifs detected in flanking intergenic sequence to the next known gene and intragenic sequence of the target gene for the 67 genes identified in the microarray analysis as having significantly different expression levels in LRVAS and LRVAD RV or LV relative to WTVAS controls are shown.

Six of the 57 genes found to contain RARE were selected for a more detailed analysis of the types and locations of RARE upstream and downstream of the gene (Table 10). The selected genes include the three retinoid metabolism genes (lrat, dhrs3, and rbp7), a gene that has been shown to be negatively regulated by RA (mgp) (12), a gene that has been shown to be positively regulated by RA (angpt2) (48), and a cell proliferation marker (mKi67). Dhrs3 is induced by RA (56), lrat is differentially regulated by RA (37, 57), and rbp7 is increased under states of VAD (9). The direction of the changes in expression levels of these transcripts in LRVAD ventricles, as determined by the microarray analysis, was consistent with a state of VAD, with dhrs3 (Table 4) and angpt2 (Table 6) mRNAs exhibiting significantly decreased expression levels in LRVAD ventricles and rbp7 (Table 4) and mgp (Table 6) showing increased expression levels in both LRVAS and LRVAD tissues relative to WTVAS controls.

Table 10.

Detailed analysis of RARE in 6 genes known to respond to VA status

Gene ID and Location	Chromosome Index	RARE Motif	Distance to Gene	Motif Direction	Region of DNA
lrat, chromosome 3 DNA anti-sense strand (−)	82949816	DR8	45844	Anti-sense	Upstream of gene
	82953173	DR8	49201	Sense	Upstream of gene
	82996879	DR8	92907	Anti-sense	Upstream of gene
	82971302	DR5	67330	Sense	Upstream of gene
	82967915	DR2	63943	Sense	Upstream of gene
	82972031	DR0	68059	Sense	Upstream of gene
dhrs3, chromosome 4 DNA sense strand (+)	144894881	DR8		Anti-sense	Intron
	144898374	DR5		Sense	Intron
	144873589	DR2	19237	Sense	Upstream of gene
	144961926	DR2	34708	Sense	Downstream of gene
	144899328	DR1		Anti-sense	Intron
	144944809	DR0	17591	Sense	Downstream of gene
	144948438	DR0	21220	Sense	Downstream of gene
Mgp, chromosome 6 DNA anti-sense strand (−)	136860020	DR2	12415	Anti-sense	Upstream of gene
	136869018	DR1	3417	Anti-sense	Upstream of gene
	136884827	DR0	9023	Anti-sense	Downstream of gene
	136900034	DR0	24230	Anti-sense	Downstream of gene
angpt2, chromosome 8 DNA anti-sense strand (−)	18687334	DR5	2928	Sense	Downstream of gene
	18723223	DR2		Anti-sense	Intron
	18751784	DR2	10223	Anti-sense	Upstream of gene
	18658582	DR1	31680	Anti-sense	Downstream of gene
	18689848	DR1	414	Anti-sense	Downstream of gene
	18774561	DR1	33000	Sense	Upstream of gene
	18816277	DR1	74716	Anti-sense	Upstream of gene
	18695286	DR0		Anti-sense	Intron
	18769633	DR0	28072	Sense	Downstream of gene
	18779102	IR0	37541	Sense/anti-sense	Downstream of gene
rbp7, chromosome 4 DNA anti-sense strand (−)	149450825	DR8		Anti-sense	Intron
	149463185	DR8	8208	Anti-sense	Upstream of gene
	149431573	DR2	18113	Anti-sense	Downstream of gene
	149464578	DR1	9601	Anti-sense	Upstream of gene
	149462743	DR0	7766	Sense	Upstream of gene
mki67, chromosome 7 DNA anti-sense strand (−)	135963069	DR8	246691	Anti-sense	Upstream of gene
	136216758	DR8	500380	Anti-sense	Upstream of gene
	136771227	DR8	1054849	Anti-sense	Upstream of gene
	136251024	DR5	534646	Anti-sense	Upstream of gene
	136461059	DR5	744681	Sense	Upstream of gene
	136560751	DR5	844373	Sense	Upstream of gene
	136637986	DR5	921608	Sense	Upstream of gene
	135802380	DR2	86002	Sense	Upstream of gene
	135978900	DR2	262522	Anti-sense	Upstream of gene
	136441382	DR2	725004	Sense	Upstream of gene
	136544626	DR2	828248	Sense	Upstream of gene
	136670230	DR2	953852	Anti-sense	Upstream of gene
	136711388	DR2	995010	Anti-sense	Upstream of gene
	136743036	DR2	1026658	Sense	Upstream of gene
	136817972	DR2	1101594	Anti-sense	Upstream of gene
	135788600	DR1	72222	Sense	Upstream of gene
	135814575	DR1	98197	Sense	Upstream of gene
	135897296	DR1	180918	Anti-sense	Upstream of gene
	135915535	DR1	199157	Anti-sense	Upstream of gene
	135956189	DR1	239811	Sense	Upstream of Gene
	136132050	DR1	415672	Anti-sense	Upstream of gene
	136137176	DR1	420798	Anti-sense	Upstream of gene
	136292546	DR1	576168	Sense	Upstream of gene
	136330137	DR1	613759	Sense	Upstream of gene
	136409785	DR1	693407	Anti-sense	Upstream of gene
	136814240	DR1	1097862	Anti-sense	Upstream of gene
	135687597	DR0	2190	Sense	Downstream of gene
	135791844	DR0	75466	Anti-sense	Upstream of gene
	135795717	DR0	79339	Sense	Upstream of gene
	135870969	DR0	154591	Anti-sense	Upstream of gene
	135886470	DR0	170092	Anti-sense	Upstream of gene
	136032369	DR0	315991	Sense	Upstream of gene
	136040808	DR0	324430	Anti-sense	Upstream of gene
	136112258	DR0	395880	Anti-sense	Upstream of gene
	136149013	DR0	432635	Anti-sense	Upstream of gene
	136445138	DR0	728760	Anti-sense	Upstream of gene
	136792870	DR0	1076492	Sense	Upstream of gene
	136506955	IR0	790577	Sense/anti-sense	Upstream of gene
	136527792	IR0	811414	Sense/anti-sense	Upstream of gene

Table 6.

Transcripts identified in microarray screen associated with cardiovascular function

Gene	LRVAS vs. WTVAS	LRVAD vs. WTVAS	Function
Angpt2	−1.4 (0.003) FDR = 0.42	−1.5 (0.0006) FDR = 0.42	Angiopoietin 2 is significantly increased in acute decompensated heart failure and is a predictor of poor outcome (41)
Casq2	−1.1 (0.10) FDR = 0.43	−1.5 (0.0007) FDR = 0.43	Calsequestrin 2 buffers sarcoplasmic reticulum Ca2+ ions and regulates ryanodine receptor (42)
Mgp	1.6 (0.002) FDR = 0.46	1.6 (0.0012) FDR = 0.46	Matrix Gla protein is a calcification inhibitor protein; may have a role in left ventricular dysfunction in patients with symptomatic aortic stenosis (52)
Npr1	−1.4 (0.003) FDR = 0.50	−1.5 (0.002) FDR = 0.50	Natiuretic peptide receptor 1
Npy	1.5 (0.006) FDR = 0.50	1.8 (0.0011) FDR = 0.50	Neuropeptide Y is expressed in heart neurons and endothelial cells; its extensive actions are mediated through NPY1 and NPY6 receptors; the Y1 receptor induces vasoconstriction and regulates gene expression in hypertrophying cardiomyocytes (18)
S100a1	−1.5 (0.0006) FDR = 0.34	−1.6 (0.0005) FDR = 0.34	S100 calcium-binding protein A1 is the most abundant member of the calcium-binding S100 protein family in myocardial tissue; serves important roles in energy balance, myofilament sliding, myofilament calcium sensibility, titin-actin interaction, apoptosis, and cardiac remodeling (17)

Shown here are results observed in only the left ventricle. Values in bold represent significant fold change in transcript expression relative to WTVAS control tissues; P values are indicated in parentheses.

ns, not significantly different than WTVAS.

Values in unbolded font failed 1 of 3 parameters for significance but are included to show trends.

Genes Associated with Cardiovascular Function/Development

Nine transcripts that encode proteins associated with cardiovascular function met all three parameters for statistical significance in the microarray screen; three additional transcripts met two of the three parameters and are reported because of their relevance to adult heart health or previous association with changes in VA status (Tables 5 and 6). The transcript showing the greatest effect was for histidine ammonia lyase (HAL), a gene involved in histidine catabolism, whose loss of function has been associated with a decreased risk of incident coronary heart disease (54). The transcript for HAL was downregulated 13-fold and 5-fold in LRAT^−/− LV and RV, respectively, with no apparent additional effect induced by absence of VA from the diet (Table 5). The angiopoietin 2 (angpt2) transcript, which encodes an antagonist of angiopoietin 1 and is also increased in the setting of cardiovascular disease, such as heart failure (41), showed progressive downregulation in LRVAS and LRVAD LV (Table 6).

Table 5.

Transcripts identified in microarray screen associated with cardiovascular function

	Right Ventricle		Left Ventricle
Gene	LRVAS vs. WTVAS	LRVAD vs. WTVAS	LRVAS vs. WTVAS	LRVAD vs. WTVAS	Function
Hal	−5.3 (0.0007)	−4.8 (0.0009)	−13.4 (3.5 e−7)	−13.6 (3.4 e−7)	Histidine ammonia lyase; loss of function is associated with increased histidine levels, which are inversely correlated with risk of chronic heart disease (54)
	FDR = 0.39	FDR = 0.39	FDR = 0.005	FDR = 0.005
Kcnj14	−1.9 (0.0003)	−1.5 (0.003)	−2.2 (0.0007)	−1.5 (0.01)	Potassium inwardly rectifying channel; regulates membrane excitability
	FDR = 0.35	FDR = 0.35	FDR = 0.46	FDR = 0.46
Nppa	ns	−2.4 (0.006)	ns	ns	Common marker of heart failure
		FDR = 0.67
Nppb	ns	−2.7 (0.03)	ns	ns	Common marker of heart failure
		FDR = 0.81
Pkig	2.5 (0.0001)	2.4 (0.0002)	2.5 (1.6 e−5)	2.4 (2.2 e−5)	Protein kinase inhibitor, γ; vascular endothelial cells express isoforms of protein kinase A inhibitor (30)
	FDR = 0.27	FDR = 0.27	FDR = 0.0.08	FDR = 0.08
Rbp7	1.7 (0.008)	2.3 (0.001)			CRBP III; VAD induces expression of this protein in heart vascular endothelial cells (9)
	FDR = 0.56	FDR = 0.56

Shown here are results observed in both RV and LV. Values in bold represent significant fold change in transcript expression relative to WTVAS control tissues; P values are indicated in parentheses. Values in unbolded font failed 1 of 3 parameters for significance but are included to show trends.

ns, not significantly different than WTVAS.

Transcripts encoding natriuretic and vasoactive peptides induced by multiple cardiac stressors (nppa and nppb) were downregulated twofold in LRVAD RV but showed no significant changes in expression in LRVAS RV (Table 5). Notably, the transcript for natriuretic peptide receptor 1 (NPR1), which modulates the activities of NPPA and NPPB, showed progressive downregulation in LRVAS and LRVAD LV (Table 6). In contrast, the transcript for neuropeptide Y (NPY), a peptide expressed in cardiac neurons and endothelial cells, was elevated in both LRVAS and LRVAD LV (Table 6).

An effect on calcium homeostasis in the heart was suggested by the relative decrease in expression of transcripts encoding calsequestrin 2 (casq2) and S100a1 in LRVAS and LRVAD LV. Mgp mRNA, which encodes a matrix Gla protein that inhibits calcification, showed increased expression in LV from LRVAS and LRVAD animals relative to controls (Table 6).

As discussed above, rbp7 encodes cellular RBP III (CRBP III). This protein is known to be induced in heart vascular endothelial cells by VAD (9), and, consistent with this finding, the rbp7 transcript exhibited a graded increase in expression in the RV of animals fed a VAS diet (LRVAS, up 1.7-fold) and those fed a VAD diet (LRVAD, up-2.3 fold). No such increases were observed in the LV (Table 5).

Kvnj14, which encodes a potassium inwardly rectifying channel, was downregulated twofold in both the RV and LV, with tissues from LRVAS animals showing a slightly greater effect than those from LRVAD mice. The transcript for a protein kinase inhibitor (pkig) that normally shows relatively high expression in the heart was elevated more than twofold in LRVAS and LRVAD RV and LV (Table 5).

Nine of the 12 genes associated with cardiovascular function have RARE motifs either in the region upstream of or downstream of the next known gene (Table 9). No RARE were found in the vicinity of the potassium channel (kcnj14), natriuretic peptide (nppa, although a RARE was found in the region of the nppb gene), or the calcium binding protein (S100a1) genes.

Other Transcripts That Were Significantly Affected in the Adult Mouse Heart by VA Status

The transcripts of an additional 59 genes not included in Tables 4–6 are listed in Tables 7 and 8. Three genes encode proteins involved in energy balance; nine genes encode proteins involved in immunity; nine genes encode receptors, channels, or signaling proteins; fifteen genes encode enzymes; five genes encode proteins associated with proliferation, growth, or differentiation. Curiously, gdpd3, the transcript showing the greatest change in expression in the microarray analysis (down 25–27-fold in LRVAS and LRVAD RV and down 34–35-fold in LRVAS and LRVAD LV), encodes glycerophosphodiesterase domain containing 3, a gene product of unknown function and no RARE signature.

Table 7.

Genes not associated with cardiovascular function whose transcripts were significantly affected by VA status in RV

Gene	LRVAS vs. WTVAS	LRVAD vs. WTVAS	Function
Genes associated with cell adhesion/interaction/cytoskeletal
Limch1, LIM and calponin homology domains 1	1.4	1.6	Actin binding domain
Pdpn, podoplanin	−1.5	−1.5	Type-I integral membrane glycoprotein with diverse distribution in human tissues
Genes associated with energy balance
Cox18, cytochrome c oxidase assembly protein 18	−1.5	−1.5	Assembly of mitochondrial oxidative phosphorylation protein
Pfkfb1, 6-phosphofructo-2-kinase	1.4	2.0	Glycolysis
Genes Associated with Immune System and Function
Retnla, resistin like α	−1.9	−3.7	Immunity, inflammation
Genes associated with growth and differentiation
Lgr6, leucine-rich repeat containing G protein-coupled receptor 6	3.1	3.3	Receptor for R-spondins that potentiates the canonical Wnt signaling pathway and acts as a marker of multipotent stem cells in the epidermis
Tcea3, transcription elongation factor A (SII), 3	−1.5	−1.5	Suppression of Tcea3 shifts mouse embryonic stem cells from pluripotency into enhanced mesoderm development, and vasculogenesis factors are activated
Genes encoding receptors, channels, and signaling proteins
Itpripl2	1.2	1.5	Inositol 1,4,5-triphosphate receptor interacting protein
Olfr872	−4.6	−4.7	Olfactory receptor 872
Ptgfr	−1.6	−1.4	Prostaglandin F receptor
Rsg1	−1.5	−1.4	REM2 and RAB-like small GTPase 1
Tmc7	1.5	1.7	Transmembrane channel-like gene family 7
Genes encoding enzymes
Aoah	−1.7	−1.4	Acyloxyacyl hydrolase
Mboat2	−2.7	−2.2	Membrane bound O-acyltransferase domain containing 2
Atp10d	−1.5	−1.6	ATPase, class V, type 10D
Gdpd3	−27.3	−25.1	Glycerophosphodiester phosphodiesterase domain containing 3
Glo1	−1.6	−1.6	Glyoxalase 1
Mettl21d	1.5	1.5	Methyltransferase like 21D
Ubiad1	−1.7	−1.4	UbiA prenyltransferase domain containing 1
Genes encoding other proteins
Elovl7, ELOVL family member 7	−3.4	−3.9	Elongation of long-chain fatty acids
Hn1l	−1.6	−1.6	Hematological and neurological expressed 1-like
Park2	−1.6	−2.1	Parkinson's disease (autosomal recessive, juvenile) 2
Pkd2l2	−1.7	−1.7	Polycystic kidney disease 2-like 2
Rsph3a/b	1.8	2.1	Radial spoke 3A homolog (Chlamydomonas)
Sfrp2	−1.7	−1.4	Secreted frizzled-related protein 2
Zfp87	−1.5	−1.5	Zinc finger protein 87

Values in bold font are significantly affected by VA status in RV. Values in unbolded font failed 1 of 3 tests for significance.

Effect of LRAT Inactivation on Cardiac Cell Proliferation

VA status is known to affect the balance between the mitotic potential and differentiation status of regenerative adult tissues (3). To assess whether reduced VA stores may affect the proliferation of cardiac cells, we performed immunohistochemical staining for Ki-67 antigen, a marker for cell proliferation. Figure 8, A–C shows the specific intranuclear localization of Ki-67. The low number of Ki-67-positive cells in cardiac tissue and the variation in the level of Ki-67 protein expression make it difficult to accurately quantitate the number of cells expressing proliferation marker, using immunofluorescence staining of tissue sections. To enable rapid quantification of Ki-67-positive cells, we used a multiparametric flow cytometry analysis of cardiac cell suspensions isolated from hearts of LRVAS and WTVAS mice. Because it is known that the number of cardiac myocytes undergoing proliferation is extremely low in adult heart (49), the analysis focused on nonmyocyte cells. Figure 8D illustrates the gating strategy that was used to identify Ki-67-positive cells using an isotype-matched control antibody. We found that cell suspensions obtained from LRVAS hearts were characterized by a higher number of Ki-67-positive cells compared with WTVAS (Fig. 8, D and F). After analysis of Ki-67-positive population, we identified subpopulations of proliferating cells as CD31^posSca-1^pos endothelial cells (Fig. 8E, top, right), Sca-1^posCD31^neg cardiac progenitors (bottom, right), and Sca-1^negCD31^neg double negative cells (bottom, left) Further analysis revealed that the number of proliferating Sca-1^posCD31^neg cardiac progenitors was significantly higher in LRVAS compared with WTVAS (Fig. 8, E and G). Consequently, cell counts revealed 1.6-fold more Sca-1^posCD31^neg cells in the LRVAS cell suspensions compared with controls (74,000 ± 8,000 vs. 46,000 ± 2,000 cells/ventricle for LRVAS and WTVAS, respectively; P = 0.012, unpaired 2-tailed t-test). No difference in the numbers of proliferating Sca-1^negCD31^neg cells or endothelial cells was found between LRVAS and WTVAS control (Fig. 8, H and I). The total numbers of Sca-1^negCD31^neg stromal cells or CD31^pos endothelial cells that expressed Ki-67 were not significantly different.

Fig. 8.

Analysis of Ki-67 protein expression. A–C: representative images of mouse heart sections stained with DAPI (blue), Ki-67 (green), and phalloidin (red). The arrows indicate Ki-67-immunopositive nuclei; scale bar = 10 μm. D: representative cytofluorographic outlier contour plots of Ki-67 protein intranuclear expression in myocyte-free cell suspension obtained from hearts of LRAT^−/− (knockout, KO, top) and WT (bottom) mice; dead cells, cell debris, and CD45^pos immune cells were excluded from analysis. E: contour and dot plots demonstrating the presence of subpopulations of CD31^pos endothelial cells (upper right quadrant), Sca1^posCD31^neg cardiac progenitors (lower right quadrant), and Sca1^negCD31^neg stromal cells within Ki-67^posCD45^neg cell population from LRATKO (top) and WT (bottom) animals. F–I: numbers of total Ki-67^pos cells (F) and Ki-67^pos:Sca-1^posCD31^neg (G), Sca-1^negCD31^neg (H), and CD31^pos (I) cell subpopulation are presented as means ± SE of 4 animals in each group; P values are indicated, unpaired 2-tailed t-test. Number of cells expressing Ki-67 was calculated from total percentage of Ki-67^pos cell, percentage of cell subpopulation within Ki-67 population, and total number of cells (837 ± 17 and 844 ± 32 × 10⁵ cell/ventricles for LRAT^−/− and WT mice, respectively; P = 0.845, unpaired 2-tailed t-test).

Effect of VA Status on Response to MI

At baseline, LRVAS and WTVAS mice had similar cardiac systolic function and structure as measured by echocardiography (Fig. 9). By 1 wk after myocardial injury induced by surgical MI, LRVAS mice had less adverse myocardial remodeling, as demonstrated by increased FS% and decreased LVIDd compared with WTVAS (Fig. 9, A and B). Interestingly, LRVAS mice demonstrated a greater degree of hypertrophy in the noninfarcted myocardium (LVPWd) compared with WTVAS at 1 wk post MI (Fig. 9C).

Fig. 9.

Myocardial infarction (MI) in WTVAS and LRVAS mice. Echocardiography analysis of WTVAS (●, n = 15) and LRVAS (○, n = 9) mice at baseline and 1 wk after MI. A: there was no difference between WTVAS and LRVAS in baseline cardiac systolic function as measured by fractional shortening (FS%). MI is associated with reduced FS% at 1 wk for both genotypes (P < 0.0001) although LRVAS reduced the effect of MI on FS% (*P = 0.037 by 2-way ANOVA). B: there were no differences between WTVAS and LRVAS in LV size as measured by LV inner diameters in diastole (LVIDd) at baseline. At 1 wk after MI, LVIDd was significantly increased in WTVAS (*P < 0.001 vs. baseline), but not LRVAS (P > 0.05 vs. baseline). C: there was no difference in diastolic LV wall thickness (LVPWd) at baseline between WTVAS and LRVAS. At 1 wk after MI, the noninfarcted LVPWd was increased in LRVAS (*P < 0.05 vs. baseline), but not WTVAS (P > 0.05 vs. baseline).

DISCUSSION

The results of this study strongly suggest that the adult mouse heart is a retinoid-responsive organ. The presence and broad expression of RAR-γ in the nucleus and cytosol of cells present in both the RV and LV indicates that the heart has the potential to respond to canonical and noncanonical modes of RA signaling. In support of this interpretation, transcripts encoded by genes with RARE that are positively regulated by RA (41, 56) were decreased in LRVAD ventricles (angpt2 and dhrs3), and transcripts encoded by genes with RARE that are negatively regulated by RA or VA status (9, 52) were significantly increased in both LRVAS and LRVAD ventricles (mgp and rbp7) relative to WTVAS controls. These responses are consistent with a prediction that both a deficiency in VA stores and/or a lack of VA in the diet will lead to RA deficiency in the heart. Thus, although neither the LRAT^−/− mice fed a VAS diet nor the LRAT^−/− mice fed a VAD diet for 50 days exhibited systems-level responses that are hallmarks of frank VAD (e.g., weight loss, alopecia, ataxia), the animals on both diets did exhibit subtle, but significant, responses at the genetic level.

The retinoid profile and composition of chylomicrons leaving the intestine of LRVAS mice are altered such that, rather than a 90:10 ratio of retinyl ester to free retinol in chylomicrons, LRVAS enterocytes produce chylomicrons with an estimated composition of 40% retinyl ester:60% free retinol (36). These results indicate that, relative to WT controls, there is more chylomicron-bound retinol available in the circulation of LRAT^−/− animals. This change in the profile of circulating retinol, together with the severe reduction in retinol stores, underscores a need for LRAT^−/− mice to compensate in some way to rid the animal of excess retinol. It is likely that these unknown compensatory processes occur at the tissue level because circulating retinol levels are unchanged in LRAT^−/− animals relative to WT controls (28, 36). Under normal conditions, the complex esterification, ester hydrolysis, and reesterification of retinol in intestine, liver, and extrahepatic tissues modulate retinol availability to prevent adverse effects of excess retinaldehyde or RA on tissue homeostasis and gene expression. Liu et al. (29) have reported that cyp26a1 levels are elevated in LRVAS liver. This finding has two relevant implications. The gene encoding cyp26a1 is positively regulated by RA, and the encoded protein clears RA from tissues by oxidizing it to more polar, nontoxic metabolites. On the basis of these RA-related gene expression and functional properties of cyp26a1, the authors have suggested that RA production may actually be upregulated in some LRVAS mouse tissues. In studies reported here, there was no elevation in cytochrome mRNA expression or other evidence for increased RA availability in the heart; in fact, the data are more in line with a state of local VAD.

The lrat construct used to produce the transgenic LRAT^−/− mice has a sequence interruption downstream from the protein start site such that a truncated mRNA transcript is still expressed in LRAT^−/− mice, but functional LRAT protein is not produced (5). The microarray result showed that expression of the truncated transcript was elevated in the RV and LV of both LRVAS and LRVAD animals relative to WTVAS controls. The lrat gene has a RARE, and it appears to have a differential response to RA that is cell specific and context dependent (37, 57). The increased expression of the lrat transcript may represent a response to local RA availability and/or a compensatory response to decreased VA stores brought about by a deficiency in LRAT (23). Further investigation is needed to assess whether the increase in truncated lrat transcript levels is physiologically significant.

Although the FDR cut-off criterion was not met for rbp7 (FDR = 0.56), the gene did show a noteworthy, graded increase in expression in LRVAS and LRVAD ventricles (Tables 4 and 5). This result is consistent with reports showing that VAD leads to increased expression of CRBP III in heart vascular endothelial cells (9). Additionally, the in silico studies reported here provide new evidence that the upstream and downstream regions of rbp7 harbor RARE (Tables 9 and 10).

The flow cytometric analysis of cardiac cell suspensions showed that depletion of hepatic VA stores is associated with increased Ki-67 protein expression in Sca-1^posCD31^neg cells and enhanced accumulation of these cells in hearts of LRVAS animals. We and others have shown previously that Sca-1^posCd31^neg cardiac cells represented a subpopulation of cardiac progenitors in adult hearts, characterized by a capability to differentiate toward endothelial cells and cardiac myocytes (6, 32, 46, 51, 53). These cells play a protective role in MI (47, 53).

Although our group studied effects of VA status on adult heart in animals with VA in the circulation, but depleted in liver and heart, our results have similarities to those of Lin et al. (27), who studied effects of RA on embryonic heart development in Raldh2^−/− mice. Lin and colleagues observed that short-term RA supplementation of the RA-deficient mutant mice was associated with increased numbers of Ki-67⁺ early cardiac progenitor cells in the ventricular compact zone accompanied by reduced cardiomyocyte expansion and expression of nppa, a specific differentiation marker (27). Interestingly, the results of our microarray analysis indicate a more than twofold decrease in the relative expression levels of nppa in the RV of LRVAD mice (Table 5). These findings support the idea that RA inhibits the expansion of progenitor cell pools while promoting cell differentiation into functional phenotypes (3). Additionally, these contrasting effects of RA are operational in tissues derived from different embryonic germ layers during both embryogenesis and postnatal life.

The increased number of Sca-1-positive cardiac progenitors in LRVAS ventricle tissue was accompanied by decreased numbers of Sca-1-negative stromal cells, which suggests that these two subpopulations may be interdependent. Sca-1 plays a crucial role in lineage commitment, and lack of Sca-1 expression is associated with decreased Wnt signaling and cardiomyogenic differentiation of cardiac progenitors (4). The role of VA in the regulation of cardiac progenitor transition into Sca-1-negative stromal cells has not been previously described. An appreciation for the physiological relevance of VA status to the proliferation of cardiac progenitors in the adult heart is novel and likely important from both a basic scientific and clinical perspective.

Our results showing a positive effect of reduced hepatic VA stores on myocardial response to injury were unexpected given previous reports in the literature showing that dietary VA depletion contributed to adverse ventricular remodeling after MI and that supplementation with RA ameliorated this effect (16). It has been noted that hepatic stellate cells release retinol stores into the circulation after injury to the heart brought about by either ischemia produced by experimental MI (39) or drug-induced cardiomyopathy (15), and RALDH2 (an enzyme responsible for RA biosynthesis) levels are elevated in the heart on days 1–7 following surgically induced MI (22), suggesting that VA is marshaled from stores and RA levels are increased in the injured heart, presumably to promote healing or repair. However, our findings support a different view. We speculate that increased availability of RA might impede cardiac progenitor cell expansion during early stages of heart injury when repopulation of injured tissue with cells or cell products that support repair and/or reduce inflammation may be important. It was shown that cardiac Sca-1^posCD31^neg progenitors produce high levels of proangiogenic cytokines and growth factors, thus improving myocardial recovery after ischemic injury (51, 53). Our finding that Sca-1^posCD31^neg cardiac progenitors show a relative increase in LRVAS ventricles supports this prediction. Huang et al. (22) observed that C/EBP proteins mediate injury-induced activation of Raldh2 in the epicardium. When these investigators inhibited C/EBP signaling, elevated RALDH2 expression was prevented, and there was reduced immune infiltration, reduced fibrosis, and preserved cardiac function after MI (22). Further mechanistic studies are needed to explain the role of VA status in acute and chronic myocardial injury and repair.

Taken together, the findings reported here provide new evidence that depletion of hepatic VA stores affects the heart at a molecular level. The presence of RAR-γ is consistent with the view that the adult heart is a retinoid-responsive organ. The in silico analysis showing RARE in some but not all genes identified in the microarray analysis raises the possibility that both direct canonical RA signaling and indirect VA or RA signaling may be involved in directing gene expression in the adult heart. Furthermore, the physiological data indicate that VA status significantly impacts the response of the heart to injury.

GRANTS

This work was supported by a grant from NIH/NIGMS/MBRS/SCORE S 06 GM 008092 to M. Asson-Batres, the American Heart Association AHAGIA and NIH U01 HL100398 to D. Sawyer, and NIH K01 HL121045 to C. Galindo. Flow cytometry experiments were carried out in the Progenitor Cell Analysis Core Facility at MMCRI, which is supported by NIH/NIGMS P20GM103465, COBRE in Stem Cell Biology and Regenerative Medicine (D. Wojchowski, P.I.). Cardiac surgery and echocardiography were conducted in the Mouse Metabolic Phenotyping Center, which is supported by NIH Grant DK59637.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

M.A.A.-B. and D.B.S. conception and design of research; M.A.A.-B., S.V.R., O.T., C.B.C., S.M., T.-L.T., and C.L.G. performed experiments; M.A.A.-B., S.V.R., O.T., C.W.D., C.B.C., C.R.L., S.M., C.R.-E., C.L.G., A.J.F.-L., and D.B.S. analyzed data; M.A.A.-B., S.V.R., O.T., C.B.C., C.R.L., S.M., C.R.-E., A.J.F.-L., and D.B.S. interpreted results of experiments; M.A.A.-B., S.V.R., O.T., C.W.D., and A.J.F.-L. prepared figures; M.A.A.-B. drafted manuscript; M.A.A.-B., S.V.R., O.T., C.W.D., C.B.C., C.R.L., S.M., C.R.-E., A.J.F.-L., and D.B.S. edited and revised manuscript; M.A.A.-B., S.V.R., O.T., C.W.D., C.B.C., C.R.L., A.J.F.-L., and D.B.S. approved final version of manuscript.