Effect of chronic uremia on the transcriptional profile of the calcified aorta analyzed by RNA sequencing

Abstract

The development of vascular calcification (VC) in chronic uremia (CU) is a tightly regulated process controlled by factors promoting and inhibiting mineralization. Next-generation high-throughput RNA sequencing (RNA-seq) is a powerful and sensitive tool for quantitative gene expression profiling and the detection of differentially expressed genes. In the present study, we, for the first time, used RNA-seq to examine rat aorta transcriptomes from CU rats compared with control rats. Severe VC was induced in CU rats, which lead to extensive changes in the transcriptional profile. Among the 10,153 genes with an expression level of >1 reads/kilobase transcript/million mapped reads, 2,663 genes were differentially expressed with 47% upregulated genes and 53% downregulated genes in uremic rats. Significantly deregulated genes were enriched for ontologies related to the extracellular matrix, response to wounding, organic substance, and ossification. The individually affected genes were of relevance to osteogenic transformation, tissue calcification, and Wnt modulation. Downregulation of the Klotho gene in uremia is believed to be involved in the development of VC, but it is debated whether the effect is caused by circulating Klotho only or if Klotho is produced locally in the vasculature. We found that Klotho was neither expressed in the normal aorta nor calcified aorta by RNA-seq. In conclusion, we demonstrated extensive changes in the transcriptional profile of the uremic calcified aorta, which were consistent with a shift in phenotype from vascular tissue toward an osteochondrocytic transcriptome profile. Moreover, neither the normal vasculature nor calcified vasculature in CU expresses Klotho.

cardiovascular disease is the main cause of morbidity and mortality in patients with chronic kidney disease (CKD). Vascular calcification (VC), which was previously considered to be a passive and degenerative process, is now recognized as a cell-mediated regulated process (42, 51). VC is histologically classified into tunica intima calcification, related to atherosclerosis, and into tunica media calcification. Media calcification is predominantly observed in systemic metabolic disorders such as CKD and diabetes mellitus and as part of the aging process (7).

Cells with multilineage potential in the arterial wall (pericytes, smooth muscle cells, and adventitial myofibroblasts) may all contribute to the development of vascular calcifying diseases (7, 53, 56). Furthermore, a role of stem/progenitor cells, which are either resident in the vessel wall or circulating cells derived from the bone marrow, has been proposed (7, 23, 59).

The uremic environment causes extreme calcification stress. Calcification of the lamina media is predominant in CKD. In response to injury, contractile vascular smooth muscle cells (VSMCs) present in the lamina media become proliferative and synthesize factors generating vessel repair (41). In the context of chronic uremia, VSMCs undergo maladaptive osteochondrocytic differentiation (51). The key factors driving this phenotypic transition in CKD include elevated extracellular phosphate, which is taken up by VSMCs via sodium-dependent phosphate transporters (Pit-1 and Pit-2), uremic toxins, oxidative stress, and aldosterone (5, 7, 51).

Uremia is in addition to the predominant media calcification also characterized by atherosclerosis and intima calcification. Inflammation and atypical cell differentiation are hallmarks of the atherosclerotic lesion. Bone morphogenetic protein (BMP)-2 and BMP-4 have been proposed as mediators of endothelial inflammation (7, 13). Endothelial cells can acquire mesenchymal myofibroblast-like characteristics through an endothelial-to-mesenchymal transition (38).

Uremic serum can induce the osteogenic phenotype in mesenchymal stem cells (23).

Recently, based on the observation that the injured kidney produces circulating signals that directly affect the skeleton and vasculature, it has been proposed that uremia induces major regulatory disturbances in these organs (14). These signals may derive from recapitulation of the developmental program of nephrogenesis, reactivation of the Wnt pathway, and production of Wnt inhibitors (10, 14). Theoretically, the uremic calcified vasculature may release factors that change the physiological impact of these pathways in extravascular tissues, including the skeleton and kidneys (49).

Evidence is emerging that CKD is a state of premature aging. For instance, prelamin A accumulates in calcified arteries of children on dialysis, supporting the notion that DNA damage may be a key driver of VC in CKD (51) as well as in aging.

Additional support for the concept of uremia being a condition of premature aging came from analysis of the phenotype of Klotho knockout mice and from the observation that uremia is a state of Klotho deficiency (15). Klotho-deficient mice develop accelerated aging features and a severe vascular phenotype of calcification, arteriosclerosis, impaired endothelial function, and impaired angiogenesis. The Klotho gene is mainly expressed in the kidney, and soluble Klotho is found in serum and urine (16, 30). Several pleiotropic functions have been attributed to Klotho, including protection of the vasculature through inhibition of the high phosphate-induced VSMC calcification by inhibition of Pit-1 and Pit-2 or Wnt antagonism (17). Whether vascular cells express endogenous Klotho (29, 30, 48) is currently a matter of debate (27). Thus, it has been shown that Klotho affects VC; what is not clear is whether this effect is due to vascular membrane Klotho or due to circulating Klotho. Therefore, it is essential, with the best possible method, to examine if Klotho is expressed in the vasculature and, if so, whether the expression profile is changed in the uremic state.

In the present study, we used high-throughput RNA sequencing (RNA-seq) for a comprehensive identification of transcriptomes from aortas of normal rats and calcified aortas of uremic rats. RNA-seq technology has emerged as a powerful tool, which has improved the efficiency of gene expression profiling (36, 45, 60). RNA-seq is based on massive parallel sequencing and mapping of millions of DNA fragments derived from expressed mRNAs and provides information on the expression levels of any gene expressed in the cells or tissue samples examined. This method is sensitive in detecting rare transcripts and does not depend on the use of a specific primer or antibody (27).

Our hypothesis was that application of RNA-seq to RNA isolated from the calcified vessels in uremia could provide a more precise identification of the genes that were significantly altered in the uremic calcified aorta compared with previous approaches. Special focus was on genes related to the extracellular matrix (ECM), ossification, and Klotho. We analyzed gene expression levels in the uremic calcified vasculature and aortas from normal rats and found a large set of genes that were differentially expressed between animals with and without VC. The major shift in expression profile indicated that a dramatic change of phenotype occurred during uremic VC development. The results of this study might provide new insights into the underlying pathogenic mechanisms of uremic VC and may further guide future studies on the possible treatment or reversibility of uremic VC.

METHODS

Animals.

Inbred adult male dark agouti (DA) rats weighing 250 g (Taconic, Ejby, Denmark) were used and housed in a temperature-controlled environment with a 12:12-h light-dark cycle with food and water ad libitum.

Induction of uremia and VC.

Uremia was induced by one-step 5/6 nephrectomy (n = 11), as previously described by our laboratory (26). Rats were anesthetized with hypnorm-midazolam (Panum Institute, Copenhagen, Denmark) and given carprofen (Rimadyl, Pfizer, Copenhagen, Denmark) subcutaneously as pain relief for the following 3 days. After 1 wk of postoperative recovery, uremic rats were given a high-phosphate diet (0.9% calcium, 1.2% phosphate, and 600 IU vitamin D per kilogram of diet, Altromin Spezialfutter) to induce severe uremic CKD-mineral and bone disorders (26, 28). After 8 wk of uremia, rats were treated with vitamin D (alfacalcidol, Leo Pharmaceutical, Copenhagen, Denmark) at 30 ng intraperitoneally three times weekly. Uremic rats were then euthanized after 6 wk of vitamin D treatment and a total of 14 wk of uremia.

The well-described experimental remnant kidney model of chronic uremia with VC (39) requires both reduced kidney function and also a high-phosphate diet in addition to vitamin D. Therefore, an extra control group of normal rats given vitamin D was added (control+D) to examine the specific effect of vitamin D in our model.

Control groups.

DA rats were kept in parallel with uremic rats and fed a standard diet (0.9% calcium, 0.7% phosphate, and 600 IU vitamin D per kilogram of diet, Altromin Spezialfutter). The control+D group (n = 4) was treated with alfacalcidol (30 ng ip) three times weekly in parallel with uremic rats, whereas another control group (control; n = 8) was left untreated until euthanization after 14 wk. At death, rats were anesthetized with pentobarbital (50 μg/kg ip, Nycomed-DAK, Copenhagen, Denmark), and eye blood was drawn. The aorta was dissected free from the level of the renal arteries and up to the heart, and connective tissue and blood were removed by gentle manipulation and a rinse with sterile saline. The aorta was instantly snap frozen in liquid nitrogen to minimize RNA degradation.

Biochemical measurements.

Uremia and the associated mineral metabolism disturbances were evaluated by measurements of plasma creatinine, urea, phosphate, total calcium, ionized calcium, parathyroid hormone (PTH), and fibroblast growth factor (FGF)23. Plasma phosphate, urea, creatinine, and total calcium were analyzed by Vitros 250 (Ortho-Clinical Diagnostics, Raritan, NJ), and ionized calcium was analyzed by ABL505 (Radiometer, Copenhagen, Denmark). Plasma intact FGF23 was measured by an intact FGF23 ELISA (Kainos Laboratories, Tokyo, Japan), which measured only full-length FGF23, with an intra-assay coefficient of variation of 2.5% and an interassay coefficient of variation of 5% in our laboratory (33). Plasma PTH was measured by a rat bioactive intact PTH ELISA (Immunotopics, San Clemente, CA), with an intra-assay variation of 4% and an interassay variation of 9% (18).

Histological evaluation.

Tissue sections from the abdominal aorta were obtained just above the renal arteries and from the aortic root and were examined by hematoxylin and eosin (H&E) and von Kossa staining. Calcifications examined by von Kossa were quantified on a scale from 1 to 6 by a simple scoring system. The circumferential involvement was scored from 1 to 3 (where 1 = 0–33%, 2 = 34–66%, and 3 = 67–100%), and the extent of calcification compared with the thickness of the vessel wall was scored as 1 or 2 (where 1 = <50% and 2 = >50%). The total score was calculated as the product of the two scores.

RNA isolation and real-time RT-PCR.

Quantitative RT-PCR was performed on a total of 22 target genes and 3 housekeeping genes (58) as previously described (12). The 22 genes were selected among those expected to be involved the osteochondrocytic transition. The genes covered a range from strong upregulated to strong downregulated genes and included genes with no change. Furthermore, genes of relevance for the Klotho/FGF23/FGF receptor 1 pathway were added. Aortas from six uremic rats and three control rats were used. Total RNA from the aorta was extracted with TRIzol (Sigma-Aldrich, St. Louis, MO). First-strand cDNA was synthesized from 0.5 μg RNA with the Superscript III cDNA kit (Invitrogen, Carlsbad, CA). Light cycler 480 II (Roche, Basel, Switzerland) and JumpStart (Sigma-Aldrich) were used for quantitative real-time PCR. The primers used are shown in Table 1. All samples were run in duplicate. The PCR product was confirmed on 2% agarose gels. Gene expression was normalized to the mean of three different housekeeping genes (attachment region-binding protein, β-actin, and ribosomal protein L13A).

Table 1.

PCR primers

Gene Symbol	Name	Sequence
Actb	β-Actin	Forward: 5′-GAACCCTAAGGCCAACCGTGAA-3′
		Reverse: 5′-GCGCGTAACCCTCATAGATG-3′
Alpl	Alkaline phosphatase	Forward: 5′-ATGTGGACTACCTATTGGGTCTCT-3′
		Reverse: 5′-CGTGGTCAATTCTGCCTCCTTCCA-3′
Bglap	Bone Gla protein (osteocalcin)	Forward: 5′-CCGTTTAGGGCATGTGTTGC-3′
		Reverse: 5′-CCGTCCATACTTTCGAGGCA-3′
Bmp2	Bone morphogenetic protein 2	Forward: 5′-CACAGGGACACACCAACCAT-3′
		Reverse: 5′-GCCACGATCCAGTCATTCCA-3′
Bmp7	Bone morphogenetic protein 7	Forward: 5′-CGGGAGCGGTTTGACAACGAGA-3′
		Reverse: 5′-CTCAGAAGCCCAGATGGTACGG-3′
Dmp1	Dentin matrix protein 1	Forward: 5′-GCTGTCCTGTGCTCTCCCTGTCGC-3′
		Reverse: 5′-TGACTGAGCCAAATTGCCCGTCCT-3′
Eln	Elastin	Forward: 5′-GTGGAGTTGGCCCTGGTGGTGTTA-3′
		Reverse: 5′-GCAGCCGCCTTAGCAGCAGATTT-3′
Fgf23	Fibroblast growth factor 23	Forward: 5′-GACGGAACACCCCATCAGACTATC-3′
		Reverse: 5′-CGGGCTGAAGTGATACGAT-3′
Fgfr1	Fibroblast growth factor receptor 1	Forward: 5′-CCAGATCCTGAAGACTGCTGGAGT-3′
		Reverse: 5′-CTCTTCCAGGGCTTCCAGAACGGT-3′
Fn1	Fibronectin 1	Forward: 5′-GGCCGGGGCAGATGGAAATGTGA-3′
		Reverse: 5′-GGCTGGCTGGGGGTCTCCGTGATA-3′
Ibsp	Integrin-binding sialoprotein	Forward: 5′-TACCGGCCACGCTACTTTCT-3′
		Reverse: 5′-GCGCAGCTAACTCCAACTTTC
Jun	Jun protooncogene	Forward: 5′-GGGGGAGCACTGCGGTCTGGAG-3′
		Reverse: 5′-GAGGCGTTGAGGGCATCGTCGTAG-3′
Kl	Klotho	Forward: 5′-CGTGAATGAGGCTCTGAAAGC-3′
		Reverse: 5′-GAGCGGTCACTAAGCGAATACG-3′
Mgp	Matrix Gla protein	Forward: 5′-TGGCAGCCCTGTGCTATGAATCTC-3′
		Reverse: 5′-TCCGGTTGGTGAAGGGACTGACTT-3′
Nfatc1	Nuclear factor of activated T cells, cytoplasmic, calcineurin-dependent 1	Forward: 5′-GACCGAGACCTGTGCAAGCCA-3′
		Reverse: 5′-CTGTCTTTATAATTGGAACATTG-3′
Postn	Periostin, osteoblast-specific factor	Forward: 5′-CTGCCCCGGCTATATGAGAATGGA-3′
		Reverse: 5′-CGGCGCGAAGTATGTGTAGGAC-3′
Rpl13a	Ribosomal protein L13A	Forward: 5′-CCCTCCACCCTATGACAAGA-3′
		Reverse: 5′-CCTTTTCCTTCCGTTTCTCC-3′
Rplp0	Ribosomal protein, large, P0	Forward: 5′-AAAGGGTCCTGGCTTTGTCT-3′
		Reverse: 5′-GCAAATGCAGATGGATCG-3′
Runx2	Runt-related transcription factor 2	Forward: 5′-TCCATTCCACCACGCCGCTGTCTT-3′
		Reverse: 5′-GCCTGGGAACTGCCTGGGGTCTGA-3′
Sfrp4	Secreted frizzled-related protein 4	Forward: 5′-GGTCCTTTGATGCTGACTGTAAAC-3′
		Reverse: 5′-TGGCATGAATAACATAGCTGTAG-3′
Sost	Sclerostin	Forward: 5′-GCCTCCTCAGGAACTAGAGAAC-3′
		Reverse: 5′-TACTCGGACACGTCTTTGGTG-3′
Sparc	Secreted protein, acidic, cysteine-rich (osteonectin)	Forward: 5′-TTCTTTGCGACCAAGTGCACC-3′
		Reverse: 5′-ACGTTTTTGAGCCAGTCACGC-3′
Spp1	Secreted phosphoprotein 1 (osteopontin)	Forward: 5′-CCGAGGTGATAGCTTGGCTT-3′
		Reverse: 5′-TCGGACTCCTGGCTCTTCAT-3′
Tnfsf11	TNF (ligand) superfamily, member 11 (Rankl)	Forward: 5′-CGAGCGCAGATGGATCCTAACAGA-3′
		Reverse: 5′-TCCCTTTGCACGGCCCCTTGAA-3′
Tnfrsf11b	TNF receptor superfamily, member 11b (osteoprotegerin)	Forward: 5′-GCGTGTACTGCAGCCCCGTGTG-3′
		Reverse: 5′-GAATTAGCAGGAGGCCAAGTGAGC-3′

Gla, γ-carboxyglutamate.

RNA-seq libraries.

RNA was extracted from rat aortas followed by isolation of polyadenylated RNA to enrich for mRNA and removal of rRNA. During the subsequent cDNA synthesis and PCR amplification, amplicons were fitted with adapter sequences, allowing for sequencing on an Illumina HiSeq platform. Reads were mapped to the rat genome, and gene expression levels were calculated as reads per kilobase of exon per million mapped reads (RPKM). The data provided a detailed profile of gene expression in the different samples and identified genes differentially expressed between uremic and control rats. RNA-seq was performed on RNA extracted from the aortas of five control, four control+D, and five uremic rats (Fig. 1). Purified total RNA was treated with the Qiagen RNase-free DNase set followed by purification on miRNeasy columns (Qiagen, Germantown, MD). Two rounds of enrichment of polyadenylated RNA were then performed using the Poly(A)Purist MAG kit (Ambion, Fisher Scientific, MA) followed by RNA purification on RNAClean XP beads (Agencourt, Beckman Coulter). The RNA sample was then fragmented on ice by adding fragmentation buffer to a final concentration of 50 mM Tris·HCl (pH 8) and 5 mM MgCl₂ followed by incubation at 95°C for 3 min and 20 s. The sample was then immediately placed back on ice, and EDTA was added at 10 nM to stabilize RNA. Fragmented RNA was purified on RNAClean XP beads (Agencourt) and eluted in 13 μl diethylpyrocarbonate-treated H₂O. cDNA synthesis and ligation of an adapter to the 3′-end of the cDNA was performed (21). The first-strand synthesis and adapter ligation steps result in a cDNA pool containing adapters both at the 5′- and 3′-ends, which can be used for subsequent PCR and Illumina sequencing. Three microliters of PCR_forward (10 μM) primer (5′-AATGATACGGCGACCACCGAGATCTACACTCTTTCCCTACACGACGCT-3′), 10 μl Phusion 5 × HF buffer, 1 μl of 10 mM dNTPs, 26.5 μl H₂O, 1 μl Phusion DNA polymerase, 2.5 μl indexing primer (10 μM, 5′-CAAGCAGAAGACGGCATACGAGATNNNNNNGTGACTGGAGTTCAGACGTGTGCTCTTCCGATCT-3′; with NNNNNN indicating the index sequence), and 6 μl purified linker ligated cDNA were mixed and run with the following PCR program: 98°C for 3 min, (98°C for 80 s, 64°C for 15 s, and 72°C for 30 s) × 4, (98°C for 80 s and 72°C for 45 s) × 10, 72°C for 5 min, and hold at 4°C. The resulting sequencing libraries were quantified on a Bioanalyzer 2100 (Agilent Technologies) using a DNA1000 chip (Agilent) for smear analysis and molarity calculation. Samples were mixed in equimolar amounts, and PCR amplicons in the 200- to 600-bp size range were extracted from a 2% E-Gel SizeSelect Agarose Gel in an E-Gel Electrophoresis System (Life Technologies, Paisley, UK). The eluate was purified using Agencourt Ampure XP beads (Beckman Coulter). The final sample was again run on a DNA1000 chip (Agilent) in a Bioanalyzer 2100 for smear analysis. Finally, single-read sequencing (with a read length of 50 nt) was performed on an Illumina HiSeq2000 sequencing system (Illumina, San Diego, CA). All sequencing was performed at the Danish National DNA Sequencing Centre, University of Copenhagen.

Fig. 1.

Brief overview of the RNA sequencing (RNA-seq) methodology. Total RNA is fragmented in a MgCl₂ buffer at 95°C (fragmentation points shown by red triangles). RT-PCR is performed with a random sequence primer carrying an overhang (dark red bars) of a predefined sequence. An adapter (light green bars) is ligated onto the 3′-end of the resulting cDNA. The ligation product thus contains a known sequence at both the 5′- and 3′-ends and can be used as a template in PCR. The PCR primers also contain overhangs of a predefined sequence, and the resulting PCR product therefore contains a sequence that allows binding to Illumina sequencing flow cell adapters and amplification with standard Illumina sequencing primers. The sequencing library is then sequenced on an Illumina sequencing platform as a single read. Reads are processed and mapped to the rat genome (assembly rn5) using the spliced read aligner Tophat (22), and differentially expressed genes are identified with the Cuffdiff function included in the Cufflinks transcript assembler (57).

RNA-seq mapping and differential expression analysis.

The adapter sequence was removed from sequencing reads using cutadapt (version 1.3) (34), which required a read length of at least 25 nt after removal. Two tools from the FASTX Toolkit (44) were used. First, three positions were trimmed off the 5′-end of all reads to remove possible untemplated products of the first-strand reaction using the FASTQ Trimmer tool. Reads with >10% of positions having a quality score of <25 were then discarded using the FASTQ Quality Filter. The remaining reads were mapped to the rat genome (assembly rn5) using Tophat (version 2.0.10) (22) with library type set to fr-secondstrand, segment length set to 20 nt, and allowed segment mismatches set to 1. Ref-Seq gene annotations (46) were used to guide mapping (-G parameter). Reads mapping to rRNA, tRNA, and chrM were discarded from the analysis. Analysis of differential gene expression was performed using Cuffdiff as implemented in Cufflinks (version 2.1.1) (57) and Ref-Seq gene annotations. No effective length correction was performed, whereas a multiread correction was applied. The false discovery rate was set to 0.05, and P values of ≤0.05 were considered significant in downstream analysis.

Gene Ontology and KEGG pathway analysis.

Gene Ontology (2) cluster analysis was performed using DAVID Bioinformatics Resources (version 6.7) (19). Representative ontologies from each cluster are shown. Pathways using annotations present in the KEGG database (20) and enrichment analysis was done with DAVID Bioinformatics Resources (version 6.7). Input in both analyses were genes that changed significantly between control and uremic states with an effect size of log₂ ratio > 1 or log₂ ratio < −1, and the default DAVID rat gene set was used as the background. Thus, the same genes were used in the two analyses.

Ethics.

The experiments were performed in accordance with the Danish National Institute for Health's Guidelines for the Care and Use of Laboratory Animals, and the study was approved by the Animal Experiments Inspectorate in Denmark (Reference no. 2012-DY-2934-00023).

Statistics.

One-way ANOVA followed by Tukey's multiple-comparison tests was used to compare means between the three groups. P values of <0.05 were considered significant. Fisher's exact test was used to test for overrepresentation of genes that shifted significantly in expression between several groups.

RESULTS

Induction of VC.

Uremic rats had significantly elevated plasma levels of creatinine and urea, as shown in Table 2. Uremic rats and alfacalcidol-treated normal rats (control+D) had significantly increased and similar levels of plasma ionized calcium, total calcium, and phosphate and suppressed PTH compared with normal control rats. Plasma intact FGF23 levels were severely enhanced in both uremic and control+D groups (Table 2). Aortic calcifications were consistently induced in all uremic rats in both the abdominal aorta and aortic root (Fig. 2). Uremic rats had a high calcification score of 4.8 ± 0.5 in the abdominal aorta and 5.0 ± 0.5 in the aortic root, whereas control+D rats only had mild signs of calcification with a patchy distribution and a score of 0.5 and 1 in only one rat. No control rats had any signs of aortic calcifications.

Table 2.

Plasma biochemistry

				P Values
	Control Group	Uremic Group	Control+D Group	Uremic vs. control group	Uremic vs. control+D group	Control vs. control+D group
Weight, g	250 ± 6	178 ± 15	211 ± 18	<0.01	NS	NS
Urea, mmol/l	6 ± 0.3	14 ± 3.1	5 ± 0.2	<0.05	<0.05	NS
Creatine, μmol/l	26 ± 1	53 ± 3	28 ± 1	<0.001	<0.001	NS
Phosphate, mmol/l	1.50 ± 0.11	2.38 ± 0.18	1.93 ± 0.13	<0.01	NS	NS
Ionized calcium, mmol/l	1.30 ± 0.01	1.59 ± 0.08	1.63 ± 0.04	<0.01	NS	<0.01
Total calcium, mmol/l	2.41 ± 0.03	2.97 ± 0.05	2.91 ± 0.07	<0.001	NS	<0.001
Parathyroid hormone, pg/ml	168 ± 40	<3	<3
Intact FGF23, pg/ml	146 ± 17	4273 ± 65	3526 ± 63	<0.001	<0.001	<0.001

Values are means ± SE.

Contol+D group, control + vitamin D-treated group; NS, not significant.

Fig. 2.

Representative histological examination by hematoxylin and eosin (HE) and von Kossa staining of aortas from control and uremic rats. No calcification was seen in the aorta from control rats. Von Kossa staining showed severe continuous medial calcification in the aorta from the uremic rat. The higher magnitude HE staining picture shows signs of osteochondrogenic transformation in the calcified area of the vessel wall from a uremic rat, with local proliferation of vascular smooth muscle cells (VSMCs) and the presence of chondrocyte-like cells (arrows). Scale bars = 500 μm and 100 μm for higher-magnitude HE staining pictures.

RNA-seq on the calcified aorta from uremic rats.

An average of 10.1 million reads was obtained per sample, 85% of which mapped to the rat genome. In total, expression levels of 17,019 genes were analyzed. Cufflinks analysis identified a total of 13,269 genes with a RPKM value of >0 and 10,153 genes with an expression level of of >1 RPKM (Fig. 3). The reads were distributed as follows: 0.5% of the genes had a RPKM value of >1,000, 8.4% of the genes had a RPKM value of >100, and 56.7% of the genes had a RPKM value of >10.

Fig. 3.

Diagram of RNA-seq results. The RNA-seq distribution of results is shown. A total of 10,153 genes with an expression level of >1 read per kilobase transcript per million mapped reads (RPKM) in at least one of the three groups were identified; 2,663 genes were differentially expressed with 47% upregulated genes and 53% downregulated genes in uremic rats relative to control + vitamin D-treated (control+D) and control rats.

Two independent sequencing libraries were built from the same uremic aorta data and analyzed as a technical control of the reproducibility of the detected expression levels. Identical expression levels were found [Pearson correlation coefficient (R) = 0.99; Fig. 4A], indicating a very low level of technical variation introduced in the experimental pipeline.

Fig. 4.

Plots of expression levels of genes expressed in aortas from uremic rats, normal control rats, and control+D rats. Plots of expression levels for genes in the different groups as obtained by Cuffdiff analysis of the Tophat mapped RNA-seq reads are shown. A: technical replicate based on expression levels generated from two independent sequencing libraries from the same uremic aorta demonstrating the low level of technical variation introduced in the experimental pipeline (Pearson's R = 0.99). B: plot of RPKM values in control and uremic rats. Red circles represent genes that significantly changed in expression level between the groups, showing that uremia had the greatest impact on the number of differentially expressed genes. Markers of osteogenesis [fibronectin (Fn1), secreted phosphoprotein 1 (Spp1), periostin (Postn), and matrix Gla protein (Mgp)] are shown among the top upregulated genes, and the aorta-related genes elastin (Eln) and tropomyosin 1a (Tpm1) are among the top downregulated genes. Mgp was highly expressed in the normal aorta and significantly upregulated in uremia, becoming the most expressed gene in the uremic rat aorta (Pearson's R = 0.59). C: plot of RPKM values in the uremic and control+D groups. Again, Mgp, Spp1, and Fn1 were highly upregulated in uremia. The correlation between the samples was higher than between control and uremia with Pearson's R = 0.68. D: plot of RPKM values in control and control+D samples. Mgp was not upregulated, whereas Spp1 was upregulated. A high level of similarity between these two conditions was found (Pearson's R = 0.97).

To compare the findings in the uremic aorta with those of the control aortas and of the aortas obtained from control rats given vitamin D, the calculated expression levels for each group were plotted against one another and correlation coefficients between groups were determined (Fig. 4, B–D). The highest correlation was found between control and control+D samples (R = 0.97). The greatest change in expression levels was between the uremic and control groups (R = 0.59), whereas a slightly higher correlation was found between the uremic and control+D groups (R = 0.68).

Of the 10,153 genes with an expression level of >1 RPKM, 26% varied significantly between uremic aortas and the two control groups; 1,905 genes were differentially expressed between uremic and control groups (false discovery rate: 0.05), 1,631 genes were differentially expressed between the uremic and control+D groups, and 885 genes were differentially expressed between the control and control+D groups.

Based on the number of differentially expressed genes, uremia had the greatest impact on gene expressions, as also indicated in the expression level plots (Fig. 4 and Tables 3–5). The higher levels of similarity between control and control+D rats compared with uremic rats was further supported by the clustering of the 14 analyzed samples in a heat map of gene expression levels (Fig. 5A). Based on both the number of genes that changed between conditions and the size of effect of these changes (Fig. 5B), uremic rats differed significantly more (P < 0.01) than control+D rats from control rats despite similar levels of plasma ionized calcium and phosphate (Table 2).

Table 3.

Top 25 genes expressed in the uremic aorta

Gene Symbol	Name	Control Group, RPKM	Uremic Group, RPKM	Control+D Group, RPKM	Log2 Ratio (Uremic Group/Control Group)	P Value (Control Group Vs. Uremic Group)
Mgp	Matix Gla protein	11,135	78,680	12,731	2.82	<0.002
Bgn	Biglycan	9,950	9,630	9,296	−0.05	0.965
Spp1	Secreted phosphoprotein 1	443	6,552	1,365	3.87	<0.001
Sparc	Secreted protein, acidic, cysteine-rich (osteonectin)	8,770	6,484	5,368	−0.44	0.642
Tagln	Transgelin	11,180	6,369	9,739	−0.81	0.156
Tpm2	Tropomyosin 2, β	5,342	4,536	4,194	−0.24	0.721
Acta2	Actin, α2, smooth muscle, aorta	7,118	4,257	6,797	−0.74	0.174
Ptms	Parathymosin	6,637	3,961	5,281	−0.74	0.101
Fn1	Fibronectin 1	192	3,778	500	4.30	<0.001
Igfbp7	IGF-binding protein 7	2,837	3,128	2,657	0.14	0.816
Tmsb4x	Thymosin, β4, X chromosome	2,102	3,051	2,298	0.54	0.092
Hba1	Hemoglobin, α1	1,205	2,786	508	1.21	0.024
Vim	Vimentin	1,786	2,711	1,985	0.60	0.158
Eln	Elastin	7,389	2,704	4,693	−1.45	0.028
Cst3	Cystatin C	1,175	2,681	1,473	1.19	<0.001
Ctgf	Connective tissue growth factor	1,562	2,599	1,622	0.73	0.098
Csrp1	Cysteine- and glycine-rich protein 1	3,343	2,598	2,975	−0.36	0.508
Sod3	SOD3, extracellular	2,875	2,429	2,580	−0.24	0.685
Hba2	Hemoglobin, α2	737	2,162	351	1.55	<0.001
Myl9	Myosin, light chain 9, regulatory	2,954	2,124	2,477	−0.48	0.218
Fabp4	Fatty acid-binding protein 4, adipocyte	2,298	2,075	4,448	−0.15	0.791
Gpx3	Glutathione peroxidase 3	482	2,021	418	2.07	<0.001
Serpine	Serpin peptidase inhibitor, clade E (nexin, plasminogen activator inhibitor type 1), member 1	2,014	1,976	1,625	−0.03	0.978
Myh11	Myosin, heavy chain 11, smooth muscle	3,545	1,724	2,643	−1.04	0.087
S100a6	S100 calcium-binding protein A6	1,215	1,693	1,290	0.48	0.075

RPKM, reads per kilobase of exon per million mapped reads.

Table 5.

Top 25 downregulated genes in the uremic aorta versus control aorta (RPKM >100)

Gene Symbol	Name	Control Group, RPKM	Uremic Group, RPKM	Control+D Group, RPKM	Log2 Ratio (Uremic Group/Control Group)	P Value (Control Group Vs. Uremic Group)
Atf3	Activating transcription factor 3	257	22	153	−3.53	< 0.001
Sh3bgr	SH3 domain-binding glutamic acid-rich protein	354	37	218	−3.27	< 0.001
Hspa1b	Heat shock protein 1B	126	13	136	−3.24	< 0.001
Dbp	D site of albumin promoter (albumin D box)-binding protein	248	28	51	−3.12	< 0.001
Gstm7	Glutathione-S-transferase, μ7	154	33	79	−2.21	< 0.001
Gstm2	Glutathione-S-transferase, μ2	488	118	248	−2.05	< 0.001
Atp1b2	ATPase, Na+/K+ transporting, β2 polypeptide	140	34	88	−2.04	< 0.001
Nr1d1	Nuclear receptor subfamily 1, group D, member 1	168	42	52	−2.00	< 0.017
Dnajb1	DnaJ (heat shock protein 40) homolog, subfamily B, member 1	129	32	138	−2.00	< 0.001
Osr1	Odd-skipped related transciptionfactor 1	275	74	143	−1.89	< 0.001
Jund	Jun D protooncogene	874	236	440	−1.89	< 0.001
Pi16	Peptidase inhibitor 16	146	41	87	−1.81	< 0.001
Jun	Jun protooncogene	216	66	156	−1.71	< 0.001
Hsph1	Heat shock 105/110 protein 1	208	67	266	−1.62	< 0.001
Bag3	Bcl 2-associated athanogene 3	426	139	410	−1.62	< 0.001
Ltbp4	Latent transforming growth factor-β-binding protein 4	179	58	119	−1.61	< 0.001
Hspb1	Heat shock protein 1	1,828	605	1,763	−1.60	< 0.001
Twsg1	Twisted gastrulation BMP signaling modulator 1	225	75	169	−1.59	< 0.001
Dnajb4	DnaJ (heat shock protein 40) homolog, subfamily B, member 4	228	77	250	−1.56	< 0.001
Scd1	Stearoyl-CoA desaturase 1	122	42	273	−1.54	< 0.001
Ak3	Adenylate kinase 3	146	52	75	−1.50	< 0.001
Ddah2	Dimethylarginine dimethylaminohydrolase 2	192	70	127	−1.46	< 0.001
Eln	Elastin	7,389	2,704	4,693	−1.45	< 0.028
Ppp1r14a	Protein phosphatase 1, regulatory (inhibitor) subunit 14A	1,038	385	686	1.43	< 0.001
Rasl12	RAS-like, family 12	141	53	89	−1.42	< 0.001

Fig. 5.

Analysis of expression levels of genes in aortas from uremic rats, normal control rats, and control+D rats. A: heat map of RPKM values of genes that shifted significantly (P < 0.05) between at least two conditions. Except for control+D sample 4, which was embedded among the control samples, all samples clustered together in the three groups. Control and control+D aortas showed higher levels of similarity, whereas uremic aortas differed from the two other groups. B: cumulative frequency plot comparing the effect of vitamin D treatment in normal and uremic rats. Genes with RPKM > 1 were grouped depending on whether they changed significantly (sign) or did not change significantly (non-sign) between either control and control+D or control and uremic conditions. For each gene, log₂ ratios of RPKM values in the two states can be used as a measure of the expressional change between the two states. The absolute values of the log₂ ratios then indicate the effect size, not distinguishing between upregulation and downregulation. For all comparisons, the absolute values of the log₂ ratios were calculated and their cumulative frequencies were plotted. For both significant and nonsignificant changes, the effect on gene expression was, as a whole, more pronounced in uremic rats than in control+D rats. C: Venn diagram of significantly changed genes. For genes that changed between at least two groups, overlaps in which genes varied between the groups were identified. Most genes that shifted significantly in expression tended to do so between more than two groups (1,538 of 2,663 genes). The greatest number of significant changes was seen between the control and uremic groups (1,905 genes). The majority of genes with expressional changes between control and uremia also did change between the control+D and uremic groups. Only 176 genes changed between control and control+D groups but not compared with the uremic group. D: control+D expression levels relative to control and uremic expression levels. For all significantly changed genes, control RPKM values were transformed to 0, uremic RPKM values were transformed to 1, and the expression values of control+D genes relative to the transformed control and uremic values were calculated. The plot, for instance, gave an indication of whether genes in the control+D group changed in the same direction (values > 0) or opposite direction (values < 0) than in uremia. Similarly, values of >1 indicate that the difference between the control+D and control group is larger than the difference between the uremic and control groups. Each line in the plot represents one percentile, corresponding to 27 genes.

In general, deregulated genes were differentially expressed between more than two of the experimental groups (Fig. 5C). Thus, 1,049 of 1,905 (55%) of the genes that varied between uremic and control groups also varied between uremic and control+D groups (Fig. 5), and, conversely, 1,049 of 1,631 (64%) of the genes that varied between the control+D and control groups also varied between the uremic and control groups. This constitutes a very strong overrepresentation of genes that varied in the uremic group versus the two control groups compared with random (P < 2.2e⁻¹⁶ by Fisher;s exact test). Two hundred twenty genes (1%) differed significantly between all three groups. Again, this is more than expected by random (P < 2.2e⁻¹⁶). Half (1,320 of 2,663 genes) of all significantly changing genes had intermediate values in the control+D group relative to the control and uremic groups (Fig. 5D), significantly more than what could be expected by random. Three hundred thirty-six genes (13%) changed more in the control+D group than in the uremic group, but, in many cases, the shift in fact did go in the opposite direction when uremia was induced (values of <0 in Fig. 5D). Thus, a group of genes was significantly affected by vitamin D, but the effect was altered by uremia.

Gene ontology related to the calcified and normal aorta.

The genes that strongly varied significantly between uremic and control aortas (at least 2-fold change in either direction) were grouped according to their gene ontologies. A Gene Ontology Cluster Analysis (Table 6) showed that significantly deregulated genes were enriched for ontologies related to the ECM, response to wounding, response to organic substance, carbohydrate binding, and ossification (Fig. 6). The same set of genes was subjected to a pathway analysis (Table 7), where the pathways that were disturbed in uremia involved ECM proteins (e.g., ECM-receptor interaction) and the Jak/STAT and MAPK signal transduction pathways.

Table 6.

Gene Ontology cluster analysis of genes deregulated in uremic rat aortas

Cluster	Annotation Type	Enrichment Score	Cluster Members	Gene Count	P Value
1	Cellular component	19.79	Extracellular region part	95	8.80e−24
			Extracellular region	134	9.10e−24
			Proteinaceous extracellular matrix	38	1.00e−11
2	Biological process	13.43	Response to wounding	57	6.60e−13
			Defense response	51	9.90e−11
			Inflammatory response	36	3.00e−10
3	Biological process	8.03	Response to organic substance	92	1.30e−13
			Response to extracellular stimulus	36	1.90e−6
			Response to vitamin	15	4.90e−3
4	Molecular function	6.77	Carbohydrate binding	36	1.50e−5
			Pattern binding	20	1.30e−5
			Glycosaminoglycan binding	17	1.90e−4
5	Biological process	6.51	Ossification	21	2.50e−6
			Skeletal system development	30	6.30e−5
			Osteoblast differentiation	10	1.80e−3

The five most enriched clusters of Gene Ontology catagories are shown.

Fig. 6.

Venn diagram and relative expression in three Gene Ontology categories. This is similar to the analysis shown in Fig. 5 but was performed specifically on genes belonging to the indicated Gene Ontology groups. Left: Venn diagram of significantly changed genes. As with the analysis of all genes (Fig. 5), most genes in the categories changed in uremic aortas with large overlaps in which genes changed in the comparisons between conditions. Right: GO categories showing control+D expression levels relative to control and uremic expression levels. For all significantly changed genes, control RPKM values were transformed to 0, uremic RPKM values were transformed to 1, and the expression values of control+D genes relative to the transformed control and uremic values were calculated. Many genes had intermediate expression levels in control+D aorta relative to control and uremic aortas.

Table 7.

KEGG pathways perturbed in uremic rat aortas

Pathway		Gene Count	P Value	Benjamini Corrected
1	Extracellular matrix-receptor interaction	15	3.7e−5	5.6 e−3
2	Complement and coagulation cascades	13	1.4 e−4	1.1 e−2
3	Cell adhesion molecules	18	9.9 e−4	4.9 e−2
4	Jak-STAT signaling pathway	17	1.4 e−3	5.0 e−2
5	Focal adhesion	21	1.5 e−3	4.5 e−2
6	MAPK signaling pathway	25	3.1 e−3	7.6 e−2
7	Hematopoietic cell lineage	11	4.9 e−3	1.0 e−1
8	Viral myocarditis	11	1.1 e−2	1.8 e−1
9	Systemic lupus erythematosus	11	1.3 e−2	2.0 e−1
10	Cardiac muscle contraction	10	1.5 e−2	2.0 e−1

The 10 pathways most enriched for deregulated genes are shown.

Individually affected genes as presented by RNA-seq.

The top 25 genes expressed in aortas from uremic rats are shown in Table 3. The gene with the highest expression level was matrix γ-carboxyglutamate protein (Mgp), which increased sevenfold compared with the control. Seven genes within the top 25 were significantly upregulated compared with both control groups. These included genes for the inhibition of mineralization, osteopontin [secreted phosphoprotein 1 (Spp1) and Mgp], fibronectin (Fn1; a regulator of osteoblastic differentiation), and the antioxidant glutathione peroxidase 3 (Gpx3).

The top 25 upregulated genes in the uremic aorta are shown in Table 4. Many of the upregulated genes are of relevance to osteogenic transformation, tissue calcification, and Wnt modulation [e.g., Fn1, Spp1, cartilage oligomeric matrix protein (Comp), periostin (Postn), Mgp, latent transforming growth factor (TGF)-β binding protein 2 (Ltbp2), Ankh, and sclerostin (Sost)]. Among other genes of relevance for tissue calcification, but not among the top 25 upregulated genes, solute carrier (Slc)20a1, which encodes Pit-1, significantly increased in uremia (RPKM: 32 to 60, P < 0.001). Similarly, and not shown in Table 4, is the expression of secreted frizzled-related protein 4 (Sfrp4), an important modulator of Wnt signaling, which increased from a RPKM value of 17 to 39 in uremia (P < 0.001). Expression of Bmp4 was significantly increased in uremia (RPKM: from 40 to 80, P < 0.001). Similarly, TGF-β receptor 1 (TGFBR1) gene expression significantly increased (RPKM: 18 to 31, P < 0.001). The smooth muscle cell markers smooth muscle aortic α-actin (Acta2), smooth muscle myosin (Myh11; Table 3), and caldesmon (Cald1; RPKM: 1191 to 1044, not significant) did not change expression levels in uremic aortas, whereas calponin 1 had decreased levels (RPKM: 1,093 to 547; P < 0.001).

Table 4.

Top 25 upregulated genes in the uremic aorta versus control aorta (RPKM >100)

Gene Symbol	Name	Control Group, RPKM	Uremic Group, RPKM	Control+D Group, RPKM	Log2 Ratio (Uremic Group/Control Group)	P Value (Control Group Vs. Uremic Group)
Cyp2e1	Cytochrome P-450, family 2, subfamily E, polypeptide 1	2	133	4	6.15	< 0.001
Fn1	Fibronectin 1	192	3,778	500	4.30	< 0.001
Spp1	Secreted phosphoprotein 1	443	6,552	1,365	3.89	< 0.001
Comp	Cartilage oligomeric matrix protein	20	299	33	3.87	< 0.001
Cd74	Cd74 molecule, major histocompatibility complex, class II invariant chain	28	358	37	3.66	< 0.001
Sost	Sclerostin	12	139	33	3.59	< 0.001
Postn	Periostin, osteoblast-specific factor	87	810	200	3.21	< 0.001
Lyz2	Lysozyme 2	34	263	28	2.97	< 0.001
Mgp	Matrix Gla protein	11,135	78,680	12,731	2.82	< 0.002
Ltbp2	Latent transforming growth factor-β-binding protein 2	167	901	268	2.43	< 0.001
Ankh	ANKH inorganic pyrophosphate transport regulator	25	125	33	2.30	< 0.001
Serping1	Serpin peptidase inhibitor, clade G (C1 inhibitor), member 1	140	651	142	2.22	< 0.001
Mfge8	Milk fat globule-EGF factor 8 protein	329	1,499	498	2.19	< 0.001
Car3	Carbonic anhydrase 3	140	629	177	2.17	< 0.001
Fxyd5	FXYD domain-containing ion transport regulator 5	95	422	165	2.15	< 0.001
Gpx3	Glutathione peroxidase 3	482	2,021	418	2.07	< 0.001
Retn	Resistin	139	536	204	1.94	< 0.001
Vcam1	Vascular cell adhesion molecule 1	80	293	184	1.88	< 0.001
Tmsb10	Thymosin, β10	29	104	38	1.87	< 0.001
Ano1	Anoctamin 1, calcium-activated chloride channel	36	129	53	1.86	< 0.001
Timp1	Tissue inhibitor of metallopeptidase 1	36	129	55	1.83	< 0.002
Lbp	Lipopolysaccharide-binding protein	68	211	50	1.63	< 0.001
Cdh13	Cadherin 13	51	153	59	1.57	< 0.001
Hba2	Hemoglobin, α2	737	2,162	351	1.55	< 0.001
Emp1	Epithelial membrane protein 1	52	151	61	1.53	< 0.001

When uremic with control+D groups were compared, all osteogenic markers were consistently less upregulated in the control+D group, and when the control+D group was compared with the control group, only four of these genes (Spp1, Sost, Fn1, and Postn) were significantly deregulated and all markedly less upregulated than when the uremic group was compared with the control and control+D groups. Cytochrome P-450 (Cyp)2e1 was the most upregulated gene by uremia, increasing 71-fold compared with the control group and 32-fold compared with the control+D group.

The present data clearly showed that the Klotho gene is not expressed in the rat aorta. When the reads mapping to the Klotho gene were analyzed, expression of Klotho was only detected at negligible levels in the RNA-seq experiments in both normal and calcified aortas from uremic rats. Importantly, the pattern of mapping revealed that the majority of these few reads were not at all corresponding to exons. Figure 7 shows the positions of all reads from the control rat aorta mapping to the Klotho gene. Most of the small number of reads map to intronic regions. For comparison, reads mapping to the highly expressed Mgp gene are shown.

Fig. 7.

The Klotho gene is not expressed in the rat aorta. Mapping of reads to the rat (rn5) genome for Klotho and Mgp are shown. A representative example of RNA-seq reads for Klotho and Mgp from control rat aorta mapping to the rat genome are shown. The Klotho gene was not expressed. The density of reads along the genome is shown along the top row. Note the difference in scale (max density for Klotho is 2 vs. 19,625 for Mgp). Also shown are the positions of all reads mapping to the Klotho gene in the control rat aorta. Most of the small number of reads mapped to intronic regions. For comparison, reads mapping to the highly expressed Mgp gene are shown.

The top 25 downregulated genes in aortas from uremic rats are shown in Table 5. The most down-regulated gene was activating transcription factor 3 (Atf3), which decreaed >10-fold. The smooth muscle cell marker elastin (Eln) was the gene with the highest absolute decrease, and, at the same time, cathepsin S (Ctss), which mediates Eln degradation, was strongly increased (RPKM: from 19 to 66, P < 0.001) in uremia. In addition, the Wnt modulator Sfrp2 and Twisted gastrulation protein homolog 1 (Twsg1), which are both BMP signaling modulator genes, were strongly downregulated in uremic aortas (P < 0.001; data not shown), whereas extracellular antagonists to BMP, follistatin (RPKM: 4 to 12, P < 0.001) and Gremlin1 (RPKM: 1 to 5, P < 0.001), were significantly upregulated.

RNA-seq and quantitative RT-PCR.

Overall, there was a high correlation between the expression levels detected with RNA-seq and quantitative PCR (Fig. 8). Except for integrin-binding sialoprotein IBSP, all genes with a threshold cycle (C_t) of >30 had a RPKM value of <1, indicating very low expression of these genes. Of the 22 target genes examined, 6 genes had a C_t value of >30 cycles and were not detected (Klotho, Bmp7, and Fgf23) or only detected at very low levels by RNA-seq (Fig. 8). The shift in gene expression seen in RNA-seq was confirmed by RT-PCR, both regarding the direction of change (12 of 16 genes) and, to a large extent, the degree of change (Fig. 9). In three of the four inconsistent genes, the shift in gene expression did not reach significance in either of the methods. In the last gene, namely, Eln, there was great dispersion of the PCR results, which could explain the divergence.

Fig. 8.

Correlation of gene expression levels in the rat aorta by RNA-seq and RT-PCR. Overall, the two methods were in agreement. Genes with a threshold cycle (C_t) value of >30 cycles were not expressed (including Klotho) or expressed at very low levels according to RNA-seq (see text).

Fig. 9.

Gene ratios for PCR and RNA-seq. The shift in gene expression levels in uremia demonstrated by RT-PCR and the corresponding RNA-seq results is shown. Average RT-PCR values for controls were set to 1. Average expression in uremic aortas relative to control aortas is indicated by the short horizontal line. For comparison, the average expression level in uremic aortas (relative to control aortas) according to RNA-seq is indicated by the black “x.” The shift in gene expression seen in RNA-seq was confirmed by RT-PCR both regarding the direction of change and, to a large extent, the degree of change.

DISCUSSION

The present study describes, for the first time, the transcriptome-wide differential expression profiling in normal and calcified rat aorta models using RNA-seq. Uremic rats had severe aortic calcifications compared with control+D rats, which had a very low calcification score, and control rats, which had no signs of calcification.

Major changes were found in the transcriptional profiles between the three groups, with 2,665 genes showing significant differences in expression levels. Expression profiles for control and control+D groups shared a high level of similarity compared with the uremic group, underlining a dramatic impact of uremia on the calcified aorta transcriptome.

Interestingly, the most enriched Gene Ontology clusters were related to ECM remodeling and bone formation. Similarly, the pathways most perturbed by uremia, according to the number of deregulated genes, showed a strong tendency to involve ECM remodelling, e.g., ECM-receptor interactions and cell adhesion molecules. Although little is known about the function of many of the proteins that are encoded by a number of the genes involved in matrix mineralization and VC, genes like Mgp and Postn, which were among the most upregulated genes in the uremic aorta, are known for their crucial impact on bone formation and mineralization. These findings represent evidence for an implication of ECM genes, although the relationship to matrix mineralization in VC still is unknown (1, 32).

Mgp was the gene with highest expression level in the uremic group, and it increased sevenfold compared with the control. High Mgp mRNA expression and protein (Mgp) accumulation in the calcified arterial wall have been previously found (52). Mgp is a strong inhibitor of VC, with extensive arterial calcifications observed in Mgp-null mice, and therefore the massive increase in expression level is most likely a defensive mechanism (32). It has, however, also been shown that Mgp is a BMP-2 regulatory protein, and while low levels of Mgp inhibit, high levels strongly enhance the osteoinductive effect of BMP-2 (62). Regulation of BMP-4 by Mgp via a mechanism involving activin kinase receptor (ALK)-1 has similarly been proposed, and a reciprocal regulation with ALK-1 inducing Mgp promotor activity in vascular endothelial cells has also previously been demonstrated (61). Expression of Postn has not been previously shown in the calcified vasculature, but it has been found to be upregulated after balloon catheter damage to arteries (31). Strongly upregulated Ltbp2 is also an ECM protein principally expressed by all cells of mesenchymal origin. Its function is largely unknown, but it has been found to surround chondrocytes (63), and Ltbp2 might have a negative regulatory impact on the assembly of elastic fibers (54).

Further classification of differentially expressed genes by their gene ontologies pointed toward a shift in phenotype in the calcifying cells of the uremic aorta toward an osteoblast or chondroblast-like cell type. A number of the most upregulated genes, Spp1, Comp, and Fn1, are chondroosteoblastic lineage markers, and increased expression has been previously shown in the calcified vasculature. Increased Fn1 expression has been found in the aorta in an animal model of transplant arteriosclerosis (47) and might induce VSMC calcification and osteoblastic differentiation via the ERK pathway (8). Spp1 and Comp have been demonstrated in the calcified human vasculature (3). However, surprisingly, when we examined the expression of smooth muscle cell markers in the present investigation, several genes, including Acta2, Myh11, and Cald1, did not change expression levels in uremic aortas, and only a few, such as calponin 1, had decreased levels. Previously, in an in vitro setting, Alves et al. (1) compared the expression of VSMCs and osteoblasts in site-by-site experiments. They found no overlap in the transcriptional program of calcifying VSMCs and osteoblasts and further demonstrated that VSMC calcifications did not depend on the downregulation of smooth muscle cell contractile markers. The present in vivo models further support the hypothesis that calcifying VSMCs use the same mechanisms as osteoblasts to mineralize, while still keeping their own identity, and thereby opening up for potential reversibility of VC.

One possibility is that osteogenic progenitor cells invade the calcified aorta, but this can unfortunately not be addressed in our model, due to lack of a uniform phenotyping definition (59). Some markers of circulating osteogenic cells, such as CD44 and platelet-derived growth factor receptor-α (data not shown) were, however, present with a significantly increased expression in the uremic aorta.

A number of the differentially expressed genes are not ECM or bone markers but have been linked to vasculopathy. The gene most affected by uremia was Cyp2e1. A change in the expression of Cyp2e1 has not been previously reported in uremia but has been shown to be the most upregulated gene in the aorta from diabetic rats (50). The Cyp2e1-induced decrease in 20-HETE production could contribute to decreased vascular contractility, a phenomenon that also might contribute to uremic VC. Ankh is a transmembrane protein that increases efflux of the calcification inhibitor pyrophosphate, and loss of function mutations in the Ankh gene result in ectopic calcification (11). The adipokine resistin has been proposed to play a role in endothelial dysfunction and has a proliferative effect on VSMCs through the ERK-pathway and induces expression of Vcam1 (40). Vcam1, which participates in the initiation of atherosclerosis, was strongly upregulated in uremic aortas in the present study (6).

The recent concept of disruption in the system biology in CKD by Fang et al. (10) and Hruska et al. (14) is based on observations that injured kidneys produce circulating signals that directly affect the vasculature and skeleton. Kidney injury leads to increased levels of Wnt inhibitors, dickkopf WNT signaling pathway inhibitor 1, and Sclerostin in the systemic circulation, affecting bone remodelling and vascular function. In the present study, expression of Sost was highly increased in the calcified aortas, and it might in CKD potentially contribute to increased circulating Sclerostin.

Based on the present results, our hypothesis is that the vasculature, as a result of VC in uremia, may release factors that change the physiological role of Wnt in extravascular tissues, including in the skeleton and kidneys (49). It remains to be established whether the calcifying vasculature contributes to the disruption in system biology in uremia.

A second renal repair factor, which circulates in elevated levels in CKD, is activin A. Inhibition of activin signaling blocks vascular calcification and renal fibrosis in CKD (14). In the present study, Inhba, which codes for the β A subunit (activin A is a dimer composed of two β A subunits) was significantly increased in uremic aortas (data not shown). Furthermore, expression of the TGFBR1 gene, which codes for an alternative type 1 receptor downstream the activin type 2A receptor, was increased and thus contributes to the proposed importance of activin signaling in VC. Gene expressions of the related pathway of BMPs were also deregulated in uremic rats with increased expression levels of BMP4 and of BMP's extracellular antagonists, Sost, follistatin, Gremlin, and BMP receptor 2, whereas a strong downregulation of Twsg1 was shown, which further underlines the potential importance of activin/BMP signaling in VC (9).

JunD knockout mice display a vascular phenotype resembling accelerated endothelial aging with an impairment of endothelium-dependent vasorelaxation, disruption of mitochondrial function, and increased oxidative stress (43). Expression of JunD was significantly decreased in uremic aortas, supporting the concept of uremia being a state of premature aging. JunD protooncogene expression has not been previously demonstrated in the uremic calcified vasculature, but recently an age-dependent downregulation of JunD has been shown in the endothelium (43).

Klotho is an evolutionarily highly conserved protein involved in longevity, and overexpression of Klotho extends the lifespan of mice (25). Furthermore, Klotho regulates the signaling of cytokines and growth factors of importance for vascular function, endothelial integrity, and resistance to oxidative stress, such as IGF-I, TGF-β, and nitric oxide (4, 15, 25, 55). It is important to determine if a deficiency of Klotho in patients with uremia also contributes to reduced longevity and the many complications, including VCs, increased rate of cardiovascular morbidity, and death. Klotho protein is present in membrane-bound and soluble forms, with the latter generated by shedding of the extracellular domain of membrane Klotho (24). The present results clearly established that the Klotho gene is not expressed in the rat aorta. Some studies have indicated the presence of Klotho mRNA and protein in VSMCs with either decreased or increased expression in uremia, and some have found that Klotho was induced in the adventitia of the uremic rat aorta (17, 29, 37, 48), whereas other investigations found no signs of Klotho being expressed in the aorta (35). The RT-PCR method used for the detection of Klotho mRNA might be too sensitive for the relevant demonstration of Klotho, as shown in the results of the present study, and factors related to poor specificity and sensitivity of the available Klotho antibodies for protein detection may all critically account for the discrepant findings in the actual literature, as recently discussed by our group (27). Thus, Klotho is not acting in a paracrine/autocrine mode being directly involved in the uremic vasculopathy or being indirectly involved as a coreceptor for FGF23. Increasing evidence points toward the kidneys being the only source of circulating Klotho (16, 30). The present results exclude the vasculature as an additional source and hereby support that notion. The vasoprotective effects of Klotho might therefore probably be exerted by circulating soluble kidney-derived Klotho, and the decrease in the kidney Klotho levels in uremia might, as such, potentially contribute to the aging vascular phenotype in CKD.

In conclusion, this is the first application of RNA-seq used to study the changes that occur in the transcriptional profile of the aorta when connected with VC in uremia. It has been demonstrated that a large set of genes are differentially expressed between animals with and without VC. The dramatic change in phenotype that occurs during the development of uremic VCs is connected to a major shift in expression profiles of genes related to ECM regulation, osteogenic transformation, and Wnt modulation. It is hypothesized that the vasculature, as a result of calcification in uremia, may release factors that change the physiological roles of Wnt in extravascular tissues, including the skeleton. The present results clearly demonstrate that the Klotho gene is not expressed in the rat aorta, in either in the uremic or normal aorta.

GRANTS

This work was supported by the University of Copenhagen, The Lundbeck Foundation, grant R126-2012-12320, and The Danish Kidney Foundation.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author(s).

AUTHOR CONTRIBUTIONS

Author contributions: J.L.R., E.G., E.L., and K.O. conception and design of research; J.L.R., E.G., M.L.M., J.H.-B., J.V., C.B.A., and E.L. performed experiments; J.L.R., E.G., M.L.M., J.H.-B., J.V., C.B.A., E.L., and K.O. analyzed data; J.L.R., E.G., M.L.M., J.H.-B., J.V., C.B.A., E.L., and K.O. interpreted results of experiments; J.L.R., E.G., M.L.M., J.H.-B., C.B.A., E.L., and K.O. prepared figures; J.L.R., E.G., E.L., and K.O. drafted manuscript; J.L.R., E.G., M.L.M., J.H.-B., J.V., C.B.A., E.L., and K.O. edited and revised manuscript; J.L.R., E.G., M.L.M., J.H.-B., J.V., C.B.A., E.L., and K.O. approved final version of manuscript.