Gene Expression Profiling on Global cDNA Arrays Gives Hints Concerning Potential Signal Transduction Pathways Involved in Cardiac Fibrosis of Renal Failure

Cardiac remodelling with interstitial fibrosis in renal failure, which so far is only poorly understood on the molecular level, was investigated in the rat model by a global gene expression profiling analysis. Sprague–Dawley rats were subjected to subtotal nephrectomy (SNX) or sham operation (sham) and followed for 2 and 12 weeks, respectively. Heart-specific gene expression profiling, with RZPD Rat Unigene-1 cDNA arrays containing about 27 000 gene and EST sequences revealed substantial changes in gene expression in SNX compared to sham animals. Motor protein genes, growth and differentiation markers, and extracellular matrix genes were upregulated in SNX rats. Obviously, not only genes involved in cardiomyocyte hypertrophy, but also genes involved in the expansion of non-vascular interstitial tissue are activated very early in animals with renal failure. Together with earlier findings in the SNX model, the present data suggest the hypothesis that the local renin–angiotensin system (RAS) may be activated by at least two pathways: (a) via second messengers and Gproteins (short-term signalling); and (b) via motor proteins, actins and integrins (longterm signalling). The study documents that complex hybridization analysis yields reproducible and promising results of patterns of gene activation pointing to signalling pathways involved in cardiac remodelling in renal failure. The complete array data are available via http://www.rzpd.de/cgi-bin/services/exp/viewExpressionData.pl.cgi


Introduction
Death from cardiac causes is up to 20 times more frequent in patients with renal failure than in the background population (Raine et al., 1992;US Renal Data System, 1995). Patients with renal failure are the group with the highest known cardiovascular risk. There is increasing evidence that such a risk is not unique to endstage renal failure, but may occur very early on in the course of renal disease (Sharma et al., 1996). The cardiovascular lesions include complex cardiac remodelling affecting not only contractile parenchyma with development of left ventricular hypertrophy (LVH), but also vascular and interstitial tissue Ritz, 1997, 2001).
Cardiac remodelling in renal failure is of interest not only because of its clinical relevance, but 572 K. Amann et al. also because it may provide experimental advantages, since the onset of the abnormality is clearly defined in time and primary manipulations of the heart are not required. Thus, renal failure may be a useful model for the more general problem of identifying mechanisms underlying cardiac remodelling. The exact nature of these cardiac-specific structural alterations have been characterized in detail in patients with renal failure  and particularly in the animal model of the subtotally nephrectomized rat (SNX), which is a well-accepted model of moderate renal failure (Amann and Ritz, 1997;Amann et al., , 2000Nabokov et al., 1999). Using immunohistochemistry, in situ hybridization and specific interventions, evidence for a pathophysiological role of the renin-angiotensin (RAS) and the endothelin (ET-1) system has been found (Amann et al., 2000;Nabokov et al., 1999;Wessels et al, 1999). In addition, increased expression of growth factors and alteration in the extracellular matrix has been documented .
It was the aim of the present study to: (a) investigate changes in cardiac gene expression with particular focus on fibrosis in the SNX model at an early (2 weeks) and a late time point (12 weeks), using high-throughput techniques; and (b) to confirm earlier data on activation of the abovementioned pathophysiological pathways.

Animals
Male 200 g Sprague-Dawley rats were housed in single cages at constant room temperature (20 • C) and humidity (75%) under a controlled light-dark cycle. The rats were fed a diet containing 40 g protein and 0.6 g NaCl per 100 g (Altromin Co., Lage/Lippe, Germany). After a 3 day adaptation period, the animals were randomly allotted to subtotal nephrectomy (SNX, n = 20) or sham operation (sham, n = 20), which were performed as follows. In a first operation the right kidney was removed under general anaesthesia (ketamin, xylazin) and weighed. The sham operation consisted of decapsulation of the kidney, taking special care not to damage the adrenals. After 1 week, the SNX was completed by resecting the lower and upper poles of the meanwhile hypertrophied left kidney. In order to standardize the procedure, a definite amount of cortex corresponding to 75% of the weight of the resected right kidney was removed (Amann and Ritz, 1997;Amann et al., , 2000Nabokov et al., 1999). The sham operation was performed as described above. This two-step surgical resection of the renal cortex leads to a moderate and very stable course of renal failure with no or only a very mild increase in systolic blood pressure and associated structural cardiovascular alterations, which have been extensively characterized in previous studies (Amann and Ritz, 1997;Amann et al., , 2000Nabokov et al., 1999).

Perfusion fixation, RNA isolation, immunohistochemistry and in situ hybridization
After 2 and 12 weeks, respectively, the experiment was terminated by retrograde perfusion fixation via the abdominal aorta in 10 SNX and 10 sham animals, as described in detail elsewhere (Amann and Ritz, 1997;Amann et al., , 2000Nabokov et al., 1999). Five SNX and five sham animals were perfused using ice-cold NaCl for molecular biology purposes; for the remaining animals 3% glutaraldehyde was used as fixative.
Small pieces of the hearts of five animals per group and time point were homogenized, using a dismembrator (Braun Co, Melsungen, Germany). Total RNA was isolated in each case, using the single-step RNA isolation method (Chomczynski and Sacchi, 1987) with Trizol (Invitrogen Co., Germany). Poly(A) + RNA was isolated from total RNA, using Dynabeads (Dynal Co., Germany) according to the manufacturer's manual. The remaining heart was cut into transverse sections and fixed with formaldehyde or isopentane for morphological, immunohistological and molecular biological investigations. Immunohistochemical investigations were performed using the avidin-biotin method, as described in detail by . Non-radioacitve in situ hybridization for angiotensinogen, ET-1, renin and TGF-β was performed as described in detail elsewhere (Amann et al., , 2000Nabokov et al., 1999).

Rat unigene-1 cDNA array
The Rat unigene-1 array contains 27 350 cDNA clones of the Bento Soares clone collection Gene expression in uraemic cardiomyopathy 573 (University of Iowa). The cDNA products were PCR-amplified, using M13 forward and reverse standard primers, and spotted onto a 22 × 22 cm nylon membrane in a 5 × 5 pattern (RZPD Berlin). For details of the methods, see Boer et al. (2001). Each 5 × 5 field contained 12 genes spotted in duplicate and one PCR fragment representing the Escherichia coli kanamycin resistance gene. This spot was used as the 'empty' spot for background subtraction during data analysis (see below). For quality control, M13 forward and reverse primers were end-labelled with 33 P γ -ATP and hybridized to each membrane to verify that all filters from the same robot run were spotted evenly and completely. After quality control, the membranes were stripped and used for complex hybridizations after about 6 weeks. Only filters from the same robot run containing comparable concentrations of PCR products representing single genes or ESTs were used for hybridization with the two different samples (sham, SNX; see below). The individual experiments -2w (2 weeks after operation, see below) and 12w (12 weeks after operation, see below) as well as repetitions -were performed with different filter batches.

Hybridization of global rat arrays and image analysis
The whole hybridization procedure was done according to Boer et al. (2001). 500-1000 ng poly(A) + RNA was reverse transcribed using (dT)18 primer and 33P α-dCTP without amplification and purified. The labelled cDNA was hybridized to the rat unigene-1 array. The hybridization solution contained 6× SSC/5 x Denhardt's, with Cot-1 DNA (Invitrogen, Germany) and (dA)40 oligonucleotide for blocking. After exposition of the hybridized membranes, the PhosphorImager screens were scanned (Fuji FLA-3000, 100 µm resolution; Fuji BAS-reader software). The primary image analysis (estimation of nVol grey level values for each individual spot) was done using the ArrayVision software package (Interfocus), which had been adjusted to the 5 × 5 array before. The background was corrected locally in each 5 × 5 field by subtracting the empty spot signal. Normalization was done via the average signal intensity (without empty spots) on the whole membrane. The ratio of the spot-to-spot comparison was taken for further analysis; those values close to zero (smaller than 0.001) were eliminated. Since each PCR fragment was spotted twice on each membrane, and each hybridization experiment was performed twice, four ratios were entered into our database tool (Access database program, Microsoft) to make an analysis in a non-statistical manner: three out of four ratios (one outlier was accepted) had to show the same tendency to be included in the final list. For each experiment, two independent hybridizations with five animals per time point were performed.

Cardiac structural alterations in the rat model of subtotal nephrectomy
We found evidence for activation of fibrotic and hypertrophic pathways at different time points after the induction of renal failure in the rat model of subtotal nephrectomy.
It is well known from several previous studies using exactly the same animal model that after 2 weeks of renal failure no increase in blood pressure and only a moderate increase in left ventricular weight are observed, whereas activation of interstitial cells is already present (Amann et al., 1994). Afterwards, left ventricular weight increases progressively, leading to marked LVH after 8 and 12 weeks (Amann et al., , 2000Nabokov et al., 1999). The animals do not, however, develop marked hypertension or anaemia, and plasma renin activation is low, which is in contrast to other models of renal failure (Amann et al., 1994(Amann et al., , 2000. Treatment of subtotally nephrectomized rats using sympatholytic agents provided some indirect evidence of a potential activation of the sympathetic nervous system; there is no definite data, however . At 8 weeks, LVH of renal failure is accompanied by a significant increase in myocyte diameter, activation and expansion of the non-vascular interstitial tissue, a decrease in capillary density and an increase in wall thickness of small intramyocardial arterioles ( Figure 1A, B; Ritz, 1997, 2001;Amann et al., , 2000Nabokov et al., 1999). Due to this time dependency of cardiac lesions, we decided to investigate an early (2 weeks) and a late (12 weeks) time point.

Analysis of differentially regulated genes
We were especially interested in extracellular matrix (ECM) genes and in molecules possibly involved in promoting ECM formation. Therefore, for data mining from gene expression profiling experiments, the genes were grouped as follows: 1. Members of the renin-angiotensin system (RAS) as the potentially most important participating hormone system. 2. Members of the group of ECM genes. 3. Genes involved in the regulation of cell junctions (adhesion molecules or structural proteins). 4. Genes involved in cell signalling (G-proteins, MAP/ERK cascade, second messenger). 5. Members of the cytoskeleton (structural proteins, motor proteins).
Genes and ESTs belonging to these groups were classified according to the GeneCards database (Weizmann Institute; http://bioinformatics.weizmann.ac.il/cards/) and the literature (Figure 2). Altogether, about 400 regulated genes potentially involved in the above-described cardiac lesions could be classified into the selected gene families.
The renin-angiotensin system (RAS) and downstream pathways are activated in uraemic cardiac hypertrophy We found an early upregulation of the endothelin B (ET B ) receptor in the hypertrophic heart after SNX ( Figure 2); this finding was confirmed by in situ hybridization, documenting increased Gene expression in uraemic cardiomyopathy 575 ET-1 mRNA in the heart of SNX after 12 w ( Figure 3H), as well as an earlier study using PCR (Amann et al., 2000). In addition, using in situ hybridization we found increased expression of renin  and angiotensinogen RNA in the heart of SNX ( Figure 3D) compared to controls ( Figure 3C), indicating activation of the local RAS. Moreover, expression of TGF-β mRNA was markedly higher in the heart of SNX rats ( Figure 3F) compared to controls ( Figure 3E).

G-proteins, second messengers and motor proteins as potential signal mediators
Since the 7-transmembrane receptor ET B is coupled to further effector systems by nucleotide regulatory proteins (Douglas and Ohlstein, 1997), we focused on this group of proteins in our analysis and also on second messengers as further downstream molecules. We found downregulation of cdc42 after 2 w, and upregulation of two RAC clones (12 w) of RHO B (2 w) and ESTs similar to RHO C and RHO A (12 w). This makes sense, because focal adhesion assembly is mainly mediated by cdc42, RAC and members of the Ras superfamily of small GTP-binding proteins and other effectors (Cau et al., 2001;Krendel et al., 2002;Schoenwaelder and Burridge, 1999). C-JUN, another signalling molecule, was not differentially regulated in the present model. P38 mitogenactivated protein kinase (MAPK ), an important member of the cytoplasmic mitogen-activated protein kinase/extracellular-signal regulated kinases (MAP/ERK ), showed controversial results: One clone was upregulated after 2 w, which would fit to the well known ET B receptor-dependent activation (Cattaruzza et al., 2001), whereas another clone was downregulated after 2 and 12 w, respectively. A marked regulation of protein kinase C and associated molecules (20 clones), inositol kinases (eight clones), phosphatases (five clones) and phospholipase C (11 clones) points to a major role of these genes in LVH.
There is convincing evidence that motor proteins are involved in the formation of focal adhesions, integrin assembly and ECM formation . We obtained some expected results from our complex hybridizations: 30 myosin clones were differentially regulated, 10 clones (four of them encoding myosin light chains) downregulated after 2 w, 13 clones upregulated and eight clones downregulated after 12 w. Of the 12 dynein clones, three went down after 2 w and four after 12 w (three of them are ESTs), whereas five were upregulated after 12 w. With kinesin, the result was as follows: two clones (kinesin-related protein) down in 2 w, eight clones up in 12 w, two clones down in 12 w. The predominant upregulation of some of the cytoskeleton-related proteins at 12 w might be initiated by LIM proteins (Khurana et al., 2002).

The extracellular matrix (ECM) accumulates during cardiac remodelling
As a central linker protein between the cytoplasm and the extracellular matrix, we found integrin-β1 to be upregulated, especially in the 12 w sample (Figure 2). This result was confirmed by immunohistochemistry ( Figure 3A, B; . 28 clones of the 128 ECM-linked clones listed in Figure 1 code for collagen subunits or enzymes involved in collagen turnover. Most genes were upregulated after 2 w, and no further change was detected after 12 w. However, single clones were upregulated after 12 w, one being a collagenase (UMCase). This indicates that collagen turnover continues at a somewhat lower level after 2 w. Two procollagens even appear to be downregulated after 2 and 12 w, respectively, whereas another clone (procollagen C-proteinase enhancer protein, PCOLCE) was upregulated after 2 and 12 w. A number of proteoglycans, another important group involved in ECM metabolism, were also upregulated as early as 2 w after SNX. Several SPARC clones were downregulated, which is in agreement with other findings (Schoenwaelder and Burridge, 1999). Other mediators of ECM formation found to be upregulated were tissue inhibitor of metalloproteinase 3 (TIMP3) at 2 w and 12 w, matrix metalloproteinase 7 (MMP7) at 2 w, laminins (eight clones) at 12 w, plectin (three clones) at 12 w, cell adhesion kinase β at 12 w, and N-cadherins and catenins.

Discussion
Renal failure is associated with complex cardiac remodelling that occurs very early on in the course of the disease (Parfrey et al., 1996;Stefanski et al., 1996) and it is characterized by LVH with accompanying changes in the myocardial composition, i.e. interstitial fibrosis and impaired microcirculation. Activation and expansion of cardiac interstitial tissue is an important hallmark of cardiac remodelling in uraemic patients and in rats with renal failure (Amann and Ritz, 1997;Amann et al., 1994;Mall et al., 1990). Using electron microscopy, activation and enlargement of interstitial fibroblasts was seen as early as 2 weeks after SNX (Amann  al., 1994). Interestingly, this was not seen in other experimental models of cardiac hypertrophy, i.e. the spontaneously hypertensive rat (SHR) or the 1C-2K rat with renovascular hypertension (Amann et al., 1994;. The pathophysiology of these cardiac alterations is still not clear; however, earlier studies found some evidence for a pathophysiological role of the RAS and the ET system Ritz, 1997, 2001;Wessels et al., 1999).
In the present study, we investigated differential gene expression in the course of cardiac remodelling in experimental renal failure using cDNA array technology. We found an early upregulation of several groups of genes known to be involved in ECM production and remodelling, i.e. collagens, proteoglycans, laminin and connected linker molecules. Some of the results of the present study are in agreement with earlier immunohistochemical and molecular biological studies in the heart of SNX animals of renal failure Wessels et al., 1999), or were confirmed by other assays (Figure 3).

Potential links from the RAS to the extracellular matrix via G-proteins, second messengers and cytoskeleton
In a previous study using in situ hybridization we found increased cardiac renin, angiotensinogen and preproendothelin mRNA after induction of renal failure (Amann and Ritz, 1997;Amann et al., 2000). An important role of the RAS in cardiovascular remodelling, in particular cardiac fibrosis, has been documented by experimental and clinical studies (Brilla, 2000). In the present study, we found evidence for an early upregulation of the ET B receptor in the hearts of SNX after 2 w. The ET B receptor is predominantly located on endothelial cells and is known to be upregulated in LVH in a G protein-dependent manner (Douglas and Ohlstein, 1997;Zolk et al., 2002).
Therefore, as shown for the kidney, several specific ET responses apparently exist in the heart (Nambi et al., 2001) which represent not only a prohypertrophic stimulus for cardiomyocytes but also stimulate fibroblasts (Tsuruda et al., 2002) and vascular smooth muscle cells (VSMC; Giulumian et al., 2002), as indicated by earlier studies using ET-receptor blockers (Amann et al., 2000;Nabokov et al., 1999).
We found strong evidence for upregulation of several second messenger molecules, i.e. protein kinase C and associated molecules, the phosphatidylinositol pathway, phosphoinositide 3-kinase (PI3K) and phosphoinositide 5-kinase (PI5K), which are known to be connected with actin filament formation (Hartwig et al., 1995;Keely et al., 1997). Interestingly, an increased interstitial expression of integrin-β1 after SNX (Figure 2;  confirms its central role in ECM formation and also its mediating function by binding to actin stress fibres.
Since we do not find a remarkable upregulation of actin genes, it remains to be investigated whether actin stress fibre formation can directly be correlated to the increase of second messengers and integrins that we found. We also postulate an important role for phosphorylation and dephosphorylation of phosphatidylinositol in regulation of cardiac remodelling in LVH. This hypothesis will need to be confirmed using more specialized phosphorylation assays.

Motor proteins play a potential role in signal translation
There is strong experimental evidence that motor proteins are involved in the formation of focal adhesions, integrin assembly and ECM formation, e.g. it has been found that myotonic dystrophy kinase related cdc42-binding kinase (MRCK) phosphorylates myosin light chains (Leung et al., 1998). Myosin light chain phosphorylation promotes both myosin filament assembly and actinactivated myosin ATPase activity . These effects result in bundling of actin filaments. The resulting tension is transmitted to integrins, which cluster in a characteristic fashion (Chrzanowska-Wodnicka and . As expected, it is mainly the expression of myosin genes and of genes coding for associated proteins, as well as other motor proteins, that are regulated in our experiments. We are currently not able to address any signalling function for dynein and kinesin, but they function as transporters along microtubules and are co-localized with or linked to actin fibres via further linker molecules, respectively (Oakley and Brunette, 1995).

K. Amann et al.
Two potential pathways are leading to ECM formation in hearts of renal failure Whereas most of the collagens and proteoglycans are upregulated during the first days after operation, in the 12 w group a striking upregulation of laminins, together with integrins, is noted. In particular, integrin is thought to be a central player because of its influence on RHO family GTPases. Structural cytoskeletal proteins again are mediators of integrin clustering and microtubule polymerization is affected by RHO A (Schoenwaelder and Burridge, 1999). We conclude from our results that at least two pathways for ECM remodelling are operative. They both start with activation of the RAS. Subsequently, either G-proteins and second messengers are activated (short-term signalling) or motor proteins, actins and integrins are upregulated (long-term signalling; Figure 4).

Many other genes remain to be investigated
Some upregulated ECM genes discussed here were also identified by expression profiling of human hypertrophied cardiac tissue, e.g. myosin regulatory light chains, troponin and desmin (Hwang et al., 2000); it should be mentioned, however, that many other molecules appear to be regulated and have to be investigated in further detail. Examples include the TIMP and MMP proteins, which were shown to be important regulators and mediators of renal fibrosis via induction of aldosterone and angiotensin II (ANGII) (Papakonstantinou et al., 2001). They were not addressed in the present experiment and have to be investigated in further studies.
Furthermore, members of other gene families are obviously regulated and are potentially involved in cardiac remodelling as shown by us, and others (Hwang et al., 2000(Hwang et al., , 2002. CD59 is an 18-20 kDa GPI-anchored membrane protein that functions as a key regulator of the terminal step of the complement activation cascade. It restricts binding of C9 to the C5b-8 complex, thereby preventing the formation of the membrane attack complex. The rat analogue of human CD59 (protectin) is expressed in the sarcolemmal membranes of normal cardiomyocytes as an important sarcolemmal inhibitor of the complement membrane attack complex (MAC; Vakeva et al., 1994); it is lost, or downregulated, after ischaemic or hypoxic injury to the heart (Venugopal et al., 2001). In addition, in experimental studies it was shown that infusion or overexpression of CD59 offers a significant protection against complement-mediated myocardial injuries (Chakraborti et al., 2000;Fisicaro et al., 2000).
Lipoprotein lipase (LPL) of either VSMC or macrophage origin is an endothelial-bound enzyme that is rate determining for the clearance of triacylglycerol-rich lipoproteins. It is known to accelerate atherosclerosis (Esenabhalu et al., 2002;Wilson et al., 2001), is upregulated in cardiomyocytes by insulin (Ewart et al., 1999) and its heparin-releasable LPL activity was higher in the heart of rats with moderate, but not severe, diabetes (Rodrigues et al., 1997).
Heat shock proteins (HSP s) are well known for their ability to 'protect' the structure and function of native macromolecules, particularly as they traffic across membranes. Some of them (HSP 27, HSP 72, HSP60) have been shown to be upregulated after cardiac ischaemic/hypoxic injury in order to offer cardioprotection against ischaemic injury by induction of antiapoptotic effects via interaction with bax and/or bak pathways (Kirchhoff et al., 2002;Lin et al., 2001).
Plasminogen activator inhibitor-1 (PAI-1) is a major anti-fibrinolytic glycoprotein thought to promote vascular growth and fibrosis during various pathophysiological situations (Venugopal et al., 2001). In particular, it was shown that aldosterone interacts with ANG II to increase PAI-1 expression in vitro and in vivo. PAI-1, by inhibiting the production of plasmin from plasminogen, tips the balance in favour of ECM accumulation, thereby promoting fibrosis (Brown et al., 2002).
Ubiquitin, a so-called repair-related protein, is upregulated in the heart early after injury (Ishikawa et al., 2000), particularly in post-ischaemic recovery (Sharma et al., 1996). In addition, the ubiquitin/proteasome system represents the cell's major tool for extralysosomal protein degradation, regulating a variety of different processes, such as proliferation, differentiation, cell cycling and apoptosis. Because both negative and positive regulators of proliferation and apoptosis undergo proteasomal degradation in a tightly regulated and temporally controlled fashion, the 26S proteasome can play opposite roles in the regulation of proliferation and apoptosis (Naujokat and Hoffmann, 2002).
Calcyclin, a member of the S100 protein family of Ca-binding proteins, is expressed in human epithelial cells and fibroblasts of several organs, indicating that it is related either to proliferation rate or secretion activity (Kuznicki et al., 1992). Although not much is known about the specific role of calcyclin in the heart, Ca-binding proteins play a critical role in cardiomyocyte function regulating sarcoplasmic reticulum Ca 2+ handling (McMahon et al., 2002).
We plan to investigate the signalling pathways leading to ECM formation in more detail. Currently, we are performing cDNA sub-arrays on glass slides and nylon membranes, respectively, containing clones which represent about 1000 genes potentially involved in LVH, and about 100 'housekeeping' genes as controls. With this array, which contains only resequenced clones, we are more flexible than with the global array, in that we can reach a higher sensitivity and can repeat the same experiment several times, which should enable us to detect changes even in genes with low expression.
We further intend to investigate samples obtained at other time points in the course of renal failure, to better characterize the time course. Furthermore, the similarities or differences to other models of LVH and the effect of various treatments will be studied. Proteomic assays, especially those looking for protein modifications such as phosphorylations, are under way and should allow us to obtain further insights into the molecular processes involved in cardiac remodelling in renal failure.