Chronic renal allograft damage (CAD) is manifested by a smoldering inflammatory process that leads to transplant glomerulopathy, diffuse interstitial fibrosis and tubular atrophy with loss of tubular structures. Using a Fischer 344 (RT1lvl) to Lewis (RT1l) rat renal allograft model, transcriptomic profiling and pathway mapping, we have previously shown that dynamic dysregulation of the Wnt signaling pathways may underlie progressive CAD. Retinoic acid, an important regulator of differentiation during vertebrate embryogenesis, can moderate the damage observed in this experimental model of CAD. We show here that subsets of the Hedgehog (Hh) and canonical Wnt signaling pathways are linked to the pathophysiology of progressive fibrosis, loss of cilia in epithelia and chronic dysfunction. Oral treatment with 13cis retinoic acid (13cRA) was found to selectively ameliorate the dysregulation of the Hh and canonical Wnt pathways associated with CAD, and lead to a general preservation of cilial structures. Interplay between these pathways helps explain the therapeutic effects of retinoic acid treatment in CAD, and suggests future targets for moderating chronic fibrosing organ damage.
Chronic allograft damage (CAD) is the leading cause of graft loss following kidney transplantation (1). CAD is characterized by progressive fibrosis with a variable degree of mononuclear cell infiltrate, and tubular atrophy leading to deterioration and eventual loss of renal function (2,3). This slowly developing disease results from a complex interplay between immune and nonimmune processes. Therapeutic options for CAD are limited in large part due to our restricted understanding of its pathophysiology.
The Wnt and Hedgehog (Hh) signaling pathways regulate cell growth and differentiation in various biologic settings and play key roles in tissue homeostasis and repair (4–8). We recently showed recapitulation of select aspects of Wnt signaling pathways during the progressive damage associated with CAD in a Fischer 344 to Lewis rat renal allograft model (9).
Treatment with 13cis retinoic acid (13cRA) can reduce the development of CAD in this model (10). The retinoid (RA) pathways interface with the Wnt and Hh pathways during development (11,12). While anti-inflammatory and antiproliferative actions of retinoid derivatives have been well demonstrated (10,13–15), the therapeutic effects of RA may also have an influence on the Wnt and Hh signaling pathways. To evaluate this potential mode of action, transcriptomic profiling and pathway modeling was performed on explanted rat renal allografts either undergoing chronic damage or with 13cRA therapy. Experimentally observed changes in the Wnt and Hh pathway interactions were conceptualized as subnetworks that were evaluated over time, with individual genes and their interconnections representing an experimental model of pathway interactions. The results show a dynamic interaction between RA, Wnt and Hh signaling pathways in the context of CAD, and demonstrate that RA treatment influences these pathways which parallel the reduced fibrosis seen.
Fischer 344 rats (F344 RT1lv1) served as donors and Lewis (LEW RT1l) rats as recipients for kidney transplants. Inbred male rats were purchased from Charles River Laboratories GmbH (Sulzfeld, Germany) and were used in all experiments. Animal experiments were performed according to German law on animal protection. Kidney transplantations were performed under general ether anesthesia by a modification of the approach described by Lee et al. (16) and von Toerne et al. (9). This model recapitulates relevant aspects of CAD seen in man (10,17–19). Control kidneys and renal allografts were explanted at 7, 14 and 56 days after transplantation, with and without treatment with 13cis retinoic acid provided in feed at 20 mg/kg body weight/day. Animals without treatment are defined as “placebo” controls.
Immunohistochemistry and immunofluorescence
Immunohistochemistry was performed using specific antibody reagents: mouse anti-β-catinin (610154 BD Biosciences), rabbit anti-Lef-1 (C12A5 Cell Signaling), mouse antirat Cd44H (CD44H BD Biosciences), rabbit anti-human fibronectin (A0245 DAKO), megalin/Lrp2 (gp33o clone CD7D5 was a gift from D. Kerijaschki, Vienna), rabbit anti-Indian Hedgehog (ab52919 Abcam) and mouse antiacetylated tubulin (T7451 Sigma). Immunohistochemistry was performed on paraffin embedded formaldehyde-fixed tissue sections after antigen retrieval using the Avidin Biotin Complex (ABC) method. Immunofluorescence was performed as described (20) using antibodies detailed above or additional antibodies for Indian Hedgehog (Santa Cruz 1196 with blocking peptide 1196P) and Gli2 (Acris APO926).
Transcriptional profiling was performed using Affymetrix RG-U34A gene chips. For placebo controls, RNA was isolated from three nontransplanted Fischer rat control kidneys (time point zero), four transplanted kidneys for time point 7 days, five transplanted kidneys for time point 14 days and five transplanted kidneys for time point 56 days after transplantation. For RA treatment studies, RNA was isolated from four transplanted kidneys for time point 7 days, five transplanted kidneys for time point 14 days and four transplanted kidneys for time point 56 days after transplantation. RNA isolation was performed as described in Ref. 10. Microarray analysis was performed as detailed in Supporting Information and Ref. 9.
Robust multichip average (RMA) analysis
For the robust multichip average (RMA) analysis of CEL files, the program RMAExpress Version 1.0 beta 10 was used. The analysis was computed automatically using default settings. A numeric output file (.txt format) contained expression values on a normal or a logarithmic scale, after the normalization. RMA analysis was used for the Hierarchical cluster and Principle component analysis.
Hierarchical cluster (HCL) analysis
Cluster analysis was performed using the program Genesis Version 1.7.5. RMA normalized data were first log2 transformed and then normalized over genes. Hierarchical, average linkage clustering was performed over genes and arrays. The graphical output was a matrix displaying clusters of coregulated genes and arrays.
Principle component analysis (PCA)
PCA is defined mathematically as an orthogonal linear transformation that transforms the data to a new coordinate system such that the greatest variance by any projection of the data comes to lie on the first coordinate (called the first principal component), the second greatest variance on the second coordinate, and so on. PCA is theoretically the optimum transform for given data in least square terms. PCA was used for a dimensionality reduction in the data set by retaining those characteristics of the data set that contribute most to its variance, by keeping lower order principal components and ignoring higher order ones. Prior to PCA analysis RMA normalized data were further calculated against average mean, using the Genesis program, to be comparable against and equivalent to HCL data. PCA was performed using the XLStat Excel software plugin Version X.
Pathways were identified using a combination of the DAVID Bioinformatics database from the NIAID, NIH (Version 2007) (http://david.abcc.ncifcrf.gov/) (21,22) and the KEGG database from the University of Kyoto (http://www.genome.jp/kegg/kegg2.html) (23–25). To analyze alterations in pathways, a version of the individual regulatory pathways for Wnt and Hh was first defined using the KEGG database. The initial KEGG-defined pathways were then refined to include additional target genes based on current literature related to the experimental system.
Verification of mRNA expression by qPCR
One microgram of total RNA was used for cDNA synthesis by Superscript I/II reverse transcriptase (Invitrogen, Karlsruhe, Germany) with hexanucleotides as primers (Roche, Mannheim, Germany). RT-PCR products from five to eight animals per group were obtained. qPCR was performed by an ABIPrism7000 Sequence detection system (Applied Biosystems, Darmstadt, Germany) as described in Ref. 9.
Statistical analyses were performed using Graph Pad Prism 4.03. For qPCR data, a nonparametric ANOVA analysis was performed comparing the animal groups with the Kruskal–Wallis test, followed by Dunn's post-test, correcting for multiple testing. Effects of the treatment were followed by a Mann–Whitney U test. Results comparing the different groups at specific time points by the Mann–Whitney U test are displayed. p-Values less than 0.05 were considered to indicate statistically significant differences.
Transcriptomic profiling of the Fischer334 to Lewis rat transplant model of chronic renal allograft damage
In a rat model where Fischer 344 rats (F344 RT1lvl) served as donors for kidney transplants and Lewis (LEW RT1l) rats served as recipients (9,10,17) transcriptomic profiling was used to characterize changes in the Wnt, Hh and RA pathways during the dynamic progression of experimental CAD.
The model displays a well-defined acute phase of rejection presenting with endothelialitis (vascular rejection), transplant glomerulitis and interstitial rejection with tubulitis. This acute phase precedes the development of a chronic smoldering inflammation marked by progressive fibrosis and dedifferentiation (9,10,17,19,26).
To characterize changes in the Wnt and Hh pathways in the context of RA therapy, control kidneys and renal allografts were explanted at 7, 14 and 56 days after transplantation, with and without 13cis retinoic acid (13cRA) treatment (20 mg/kg body weight/day). The transcriptome present in each kidney and time point was analyzed using Affymetrix RG-U34A gene chips (27). Explanted tissue was characterized in parallel for histological changes.
Unsupervised hierarchical clustering with Euclidean distance analysis was applied for the initial characterization of the resultant microarray data. Branching in this analysis demonstrates the general level of differential expression between the transcriptomes representing each of the data points. The progressive distance seen in branching increased over the 7-, 14- and 56-day placebo controls. The results showed a trend toward grouping of the day 56 13cRA treatment group with the normal kidneys together (Figure 1A). The two exemplary clusters shown in Figure 1A demonstrate that the control group and the 56-day treatment group are distinguishable from the other experimental groups in their expression pattern. PCA was used to further analyze the individual transcriptomes. Analysis demonstrated a separation of the experimental groups over time and with treatment. PCA demonstrated a progressive difference in overall gene expression with time within the placebo groups. Groups representing days 7 and 14 after transplantation showed the greatest difference in PC1, while PC2 separated the placebo from the treatment groups which was most pronounced on day 56. In contrast, the 56-day treatment group approximated the expression levels seen in the normal, nontransplanted control kidneys (Figure 1B). This observation at the transcriptomic level was also reflected by histology observed after treatment with 13cRA (10).
13cRA treatment moderated expression of exemplary marker genes associated with fibrosis
The expression of a select group of genes associated with the development of tissue fibrosis was verified by quantitative PCR (qPCR). This group also included relevant genes not present on the microarray. The reduced level of fibrosis in the 13cRA-treated animals (10) was reflected by a lower level of mRNA expression, relative to placebo controls, for markers of fibroblast activation including Vim, Actg2, as well as for Fsp1/S100a4 which is linked to monocyte/macrophage activation.
Tgfb1 helps drive progressive renal fibrosis (9). 13cRA significantly reduced the elevated Tgfb1 levels seen in placebo controls by day 56. The Tgfb1 target gene Serpine1/Pai was increased with CAD, but was reduced following 13cRA treatment. Expression of the Tgfb1 antagonist Bmp7-–which inversely correlates with fibrosis—showed less pronounced suppression with 13cRA treatment by day 14, and was significantly increased by day 56 compared to the placebo-treated animals. Col1a1 and Actg2 (smooth muscle actin) mRNA expression was significantly reduced by day 56, while Col3a1 expression was additionally reduced at the earliest time points in the context of treatment. Vimentin mRNA expression was also significantly reduced by day 14 treatment (Figure 2).
13cRA treatment targets the dysregulation of the Wnt and Hh pathways in CAD
The Wnt and Hh signaling pathways help regulate normal renal tubular function (28–30). Components of these pathways have been linked to a diverse set of renal disorders (9,31).
A network modeling the interaction between the Hh and Wnt pathways based on the KEGG database and BiblioSphere pathway tool was then developed. The model included the genes within the pathways, target genes and potential sites of interaction with the RA signaling pathway (Figure 3). Dynamic changes in the pathways occurring during the initiation and development of CAD and in the context of 13cRA treatment were then evaluated.
Effects of 13cRA treatment on activation of the Wnt canonical pathway
The canonical Wnt pathway is altered during progressive CAD (9). The effect of 13cRA treatment on genes linked to canonical Wnt signaling was evaluated (Figure 4). Expression of the individual Wnt ligand genes was assessed by qPCR (as only Wnt5a was present on the Affymetrix array). qPCR showed modulation of expression of seven of the 19 Wnt genes. Changes in mRNA expression of Wnt2b, Wnt3, Wnt6, Wnt7a, Wnt8b and Wnt10a mRNA (Wnt2b, Wnt8b and Wnt10a were decreased, while Wnt3, Wnt6 and Wnt7a were increased) were associated with the development of CAD. These changes were largely attenuated by 13cRA treatment in all of the Wnt ligands expressed by the tissue with effects clearly seen at the earliest time point measured (Figure 4). Wnt8b expression was significantly reduced, and Wnt6 significantly increased by day 7, while Wnt3 was significantly reduced by day 56.
Changes in central components of the canonical Wnt pathway were verified using qPCR (Figure 5A). Figure 5B illustrates the transcriptomic changes seen over time in the genes that make up the canonical Wnt signaling pathway, with and without 13cRA treatment.
Stabilization of beta-catenin is an integral component of the activated canonical Wnt signaling pathway. While beta-catenin was not regulated at the mRNA level, nuclear accumulation of beta-catenin protein in progressive CAD was verified by immunohistochemistry and immunoflorescence in endothelia of peritubular capillaries, but not in tubular epithelia probably due to its low abundance in nuclei (Figure 6A and Ref. 9). This accumulation was reduced with 13cRA treatment (Figure 6A). 13cRA treatment also modulated expression of the Wnt associated transcription factors Lef1 and Tcf1. Bmp4, cMyc, Fn1, Cd44, Mmp7 and Nos2 are downstream target genes of the canonical Wnt pathway. While Bmp4 was reduced, the other target genes increased with CAD (9) (Figure 5B). The elevated levels of Cd44 and Fn1 in CAD returned to control kidney levels by day 56 with 13cRA treatment, as did Myc mRNA levels. For Cd44 and Tn1 the decrease in expression during 13cRA administration is shown in Supporting Figure 1.
13cRA treatment influences the Hedgehog pathway
The actions of retinoic acid signaling during vertebrate development include effects on the Hh pathway (32). Due to its crucial role in the maintenance of normal tissue homeostasis including the cilia structures found on tubular epithelial cells, a disruption in Hh signaling is likely to contribute to the renal tubular injury seen during CAD.
Expression of the Hh ligand Desert Hedgehog (Dhh) was increased, while Indian Hedgehog (Ihh) and Sonic Hedgehog (Shh) were decreased with progressive CAD, although expression of Ihh protein appeared to increase in interstitial mononuclear cells during CAD (data not shown). The receptor Ptch1, a reliable indicator for Hh pathway induction (33), was increased with CAD. The Shh adaptor Lrp2/megalin (and Wnt pathway gene) was downregulated with progressive tissue damage. Three members of the Gli family of transcription factors are associated with the Hh pathway (Gli1, Gli2 and Gli3) (34). Gli1 and Gli2 are described as activators of Hh, while Gli3 is thought to act as a repressor (35). A moderate increase in Gli1 mRNA (qPCR) was seen with CAD, while Gli2 and Gli3 mRNA levels were reduced over time (Figure 7A).
The Bmps belong to a superfamily of growth factors that connect the Hh pathway and TGFB signaling with fibrotic processes (36,37). The Hh target genes Bmp4 and Bmp7 were downregulated during CAD (Figure 7A), while secreted frizzled-related protein 1 (Sfrp1), an inhibitor that regulates both Hh and canonical Wnt signaling (38), was increased by day 14 (Figure 7A). With 13cRA treatment, the dysregulation of Ihh, Shh and Lrp2 observed during progressive CAD was prevented, and the CAD-associated reduction in Bmp7 mRNA was reversed by day 56. No 13cRA effects were observed on Gli1 expression; however, Sfrp1 mRNA was significantly reduced during all time points in the treatment groups. The Gli2 downregulation associated with CAD was also moderated by treatment (Figure 7A). A schematic model was used to summarize the transcriptomic changes that occur at 7, 14 and 56 days in the Hh pathway during the initiation and progression of CAD, and in the course of 13cRA treatment (Figure 7B).
Alterations in the expression of megalin/Lrp2, Ihh and Gli2 protein were further verified by immunohistochemistry. The protein expression seen mirrored the mRNA data (Figure 8A). Megalin was localized largely to the tubular epithelial/brush boarder region in normal kidneys and was lost with progressive damage. Megalin expression was maintained with treatment. The expression of Ihh protein localized mainly to tubular epithelia in normal kidney. Protein expression decreased with increasing tubular damage, but this effect was attenuated by 13cRA treatment (data not shown).
Protein expression of Gli2 was also verified by immunofluorescence. The glomerulus in normal kidney was negative for protein expression, while day 56 allografts showed a focal Gli2 protein expression in tubules with segment differentiation. Treated kidneys showed a distribution that paralleled that seen in the control kidneys (Figure 8B (a–c)).
RA treatment prevents the loss of primary cilia structures during CAD
Cilia are microtubule-based organelles that act as signaling centers and mechanosensors of fluid flow through the lumen of renal tubules (28,39). The Hh and Wnt signaling pathways are associated with proper function of cilia (28). Thus, primary cilia may represent an important downstream target of CAD (29,30).
Primary cilia were visualized by immunohistochemistry and immunofluorescence using an antibody directed against acetylated tubulin, which revealed the cilia axoneme (Figure 9) (40). By day 56 following transplantation, the placebo-treated kidneys showed a complete disruption of the cilia axoneme which paralleled the pronounced dysregulation of the Hh and Wnt signaling pathways seen. By contrast, the 13cRA treated animals showed good preservation of cilia structure on day 56 of treatment, displaying structures that resembled those of normal kidney (Figure 9).
Orally administered 13cRA dramatically reduces chronic renal damage in the Fischer-Lewis transplant model (10). However, this treatment reduces, but does not eliminate, the mononuclear infiltrate of the interstitium and tubules. This suggests that in addition to its anti-inflammatory effects, other RA driven actions contribute to the therapeutic effects seen (10). An influence of 13cRA treatment on the Hh and Wnt signaling pathways was established in vivo. The results suggest a dynamic regulatory network linking these three regulatory pathways in the context of chronic organ damage (Figure 10 and Table 1).
Table 1. A summary of the general expression of Hedgehog and Wnt pathway associated genes found to be differentially regulated with CAD, and the influence of 13cRa treatment on gene expression
Significant effect of treatment
Canonical Wnt pathway
Through binding to its nuclear receptors, RA signaling has complex and pleiotropic functions during vertebrate development (32). There are three RXR (alpha, beta and gamma) and three RARs receptors (alpha, beta and gamma). RAR heterodimerises with RXR. These dimers bind to response elements in the regulatory regions of target genes and moderate their transcription (32,41).
The Wnt pathways help control cell differentiation and polarity. Canonical Wnt is activated through stabilization of beta-catenin and binding to the so-called T-cell specific transcription factor (Tcf). The beta-catenin-Tcf complex then moderates Wnt target gene transcription (38). A direct link between retinoids, nuclear receptors and Wnt pathways during development has been established (11,12,42,43). Nuclear receptors including PPARγ and the RAR receptor are known to bind beta-catenin and through this regulate the activation of Wnt target genes by competing with Tcf/Lef1 for DNA binding (43,44).
13cRA treatment resulted in increased expression of the lignds Wnt2b and Wnt10a, and decreased expression of Wnt3, Wnt6, Wnt7a and Wnt8b. Modification of the Wnt pathway via RA signaling has been previously shown to help tilt cartilage matrix homeostasis toward catabolism (45). Treatment appeared to lead to a reduction in the fibrosis-promoting aspect of the Wnt canonical pathway as evidenced by the effects on Cd44, Fn1, Nos2, Mmp7 and Myc relative to untreated kidneys. Downregulation of Tcf1 observed during CAD was attenuated by 13cRA treatment, while increased Lef1 expression seen in CAD was reduced with RA. These effects of13cRA on gene expression were often seen at the earliest time point tested.
Aberrant modulation of Hh (and Wnt) signaling pathways is linked to specific human diseases (5,7,31). Hh signaling helps control tissue homeostasis in the adult (46,47). Disruption of Hh signaling can impair wound healing (48). Expression of the Hh ligands Ihh and Shh, the Hh-linked receptor Lrp2, and Hh target genes Bmp4 and Bmp7 were decreased with CAD, while the Hh ligand Dhh, the receptor Ptch1, transcription factor Gli1 and Sfrp1 were increased with CAD. Gli1-dependent upregulation of Sfrp1 provides a direct link to the Wnt pathway where Sfrp1 acts as an upstream pathway inhibitor (6,49). Expression levels of the Hh transcription factors Gli2 and Gli3—mediators of Ihh signaling—were found to be decreased with progressive damage. In the case of Shh, Dhh, Ptch, Gli1, Gli3, Sfrp1, Bmp4 and Bmp7, the first alterations in mRNA expression were identified by the 7-day time point.
Lrp2, also referred to as megalin, is an endocytotic scavenger receptor expressed on the brush border of the renal proximal tubule (50). Lrp2 is a member of a family of receptors with structural similarities to the low density lipoprotein receptor. Decreased renal Lrp2 expression has been associated with diverse renal diseases characterized by proteinuria and defective tubular reuptake of filtered proteins (50,51). Lrp2/megalin mRNA and protein levels were dramatically reduced with progression of damage. This reduction in Lrp2/Megalin accompanying CAD was prevented by 13cRA treatment. 13cRA treatment was seen to reverse, or block the development of many of the CAD associated effects on the Hh signaling pathway.
Renal Bmp7, a member of the TGF-β superfamily, disappears early in fibrogenic renal diseases (52). A direct stimulatory effect of retinoic acid on Bmp7 expression has been previously described (53). Bmp4 and Bmp7, which were downregulated with experimental CAD, returned to normal levels with 13cRA treatment. Bmp biology also provides a link between the Hh and Wnt pathways. Mesenchymally produced Bmp can induce Ihh in gastrointestinal epithelial cells, which was found to lead to the induction of Wnt2B and Bmp4 (54). Ihh has also been described as an antagonist of Wnt signaling (55). The Sfrps act as soluble modulators of Wnt signaling. Sfrp1, a target of the Hh pathway (49), was upregulated by 13cRA treatment.
Ihh protein was verified in proximal tubules with intense staining of the brush border, distal tubules and collecting duct, which was reduced with progressive damage. In the context of continuous 13cRA treatment the Ihh staining pattern returned to that seen in control kidneys (data not shown). Gli2 protein expression could also be verified by immunofluorescence. Glomeruli in normal kidneys were negative, while Gli2 protein was found in proximal tubules. In the context of CAD, a focal expression of Gli2 was seen in tubules with segment differentiation. 13cRA treatment lead to a protein distribution that mirrored that seen in normal kidney (Figure 8B).
Primary cilium helps regulate and maintain cellular and tissue homeostasis by acting as a sensory apparatus that signals through the Hh and Wnt pathways. Cilia are found in proximal and distal tubules, the macula densa and the collecting duct. Polycystic kidney disease, a common genetic cause of chronic renal failure, is a ciliopathy that arises from abnormalities in the primary cilium leading to perturbation of the signaling pathways that regulate renal epithelial cell growth and differentiation. Here we show that disruption of cilia also underlies aspects of the pathophysiology of CAD, but that RA treatment effectively prevents disruption of cilia in the renal tubular regions. Immunohistochemical and immunofluoresence staining for acetylated tubulin identified ciliary axoneme in the proximal tubules of normal kidneys and allografts. CAD led to a general disruption of ciliary structure, while treatment preserved the cilary axoneme. These results are supported by a recent report showing that RA signaling can be beneficial in enhancing morphological and functional aspects of regenerating cilia in the sinus mucosa (56).
During embryonic development RA acts as a major integrator of the signaling pathways that control apoptosis, proliferation and differentiation (32,57,58). A series of sequential roles for Hh has been established in pancreas development where cross talk between Hh and Bmp and retinoic acid signaling pathways controls pancreatic specification and differentiation. RA was shown to directly downregulate the Hh signaling components ptc1 and smo (59).
The Wnt and Hh pathways are active in normal renal tissue where they help maintain tissue homeostasis (60,61). Wnt biology helps control cell polarity, proliferation and other processes. Reactivation and dysregulation of Wnt pathways underlies chronic fibrosis. Hh signaling is important in development and wound healing. We show here that Hh signaling is also associated with CAD.
A direct interaction between Wnt, Hh and RA signaling pathways underlie fundamental aspects of development and organogensis (11,12,57,62). Similar effects may also underlie the biology seen by RA treatment in this model of CAD. Whether these are direct or indirect effects of RA on the individual components of these pathways will require a more detailed systematic study.
During renal allograft rejection, the chronic recruitment of leucocytes such as T cells and macrophages can promote tissue damage. These cells may also help drive the disruption of Wnt/Hh signaling network by communicating the need to affect tissue repair by paracrine production of Wnt ligands. Activation of the retinoic pathway appears to override part of this program by influencing the dysregulation of the Wnt and Hh signaling pathways. The RAR/RXR pathways may thus represent important targets for therapeutic intervention in chronic tissue fibrosis.
This work was supported by the Deutsche Forschungsgemeinschaft (SFB 938 and GRK 880/3 to H.-J.G., NE 468/2-4, SFB 571 to P.J.N.).
The authors of this manuscript have no conflicts of interest to disclose as described by the American Journal of Transplantation.