Genome-Wide Transcription Profile of Endothelial Cells After Cardiac Transplantation in the Rat

Authors


Corresponding author: Guttorm Haraldsen, guttorm.haraldsen@rr-research.no

Abstract

Transcriptome analyses of organ transplants have until now usually focused on whole tissue samples containing activation profiles from different cell populations. Here, we enriched endothelial cells from rat cardiac allografts and isografts, establishing their activation profile at baseline and on days 2, 3 and 4 after transplantation. Modulated transcripts were assigned to three categories based on their regulation profile in allografts and isografts. Categories A and B contained the majority of transcripts and showed similar regulation in both graft types, appearing to represent responses to surgical trauma. By contrast, category C contained transcripts that were partly allograft-specific and to a large extent associated with interferon-γ-responsiveness. Several transcripts were verified by immunohistochemical analysis of graft lesions, among them the matricellular protein periostin, which was one of the most highly upregulated transcripts but has not been associated with transplantation previously. In conclusion, the majority of the differentially expressed genes in graft endothelial cells are affected by the transplantation procedure whereas relatively few are associated with allograft rejection.

Introduction

Despite improvements in immunosuppression, allograft loss due to rejection remains an important challenge. When acute rejection occurs in the first year posttransplantation under modern protocols of immunosuppression, it usually involves alloimmune-mediated injury of graft blood vessels (1). Thereafter, solid-organ grafts fail at a rate of 3–5% per year, most of them showing evidence of transplant vasculopathy (2). In other words, acute and chronic vascular injuries have emerged as principal causes of graft loss in modern graft management (3). Although the presence of alloantigen provides the primary stimulus for allograft rejection, perioperative vascular injury appears to affect the development of acute vascular rejection episodes (4), which in turn is a risk factor for the development of transplant vasculopathy. This is perhaps not surprising since endothelial cells of the graft are exposed to a variety of insults such as the hemodynamic instability of brain dead donors, organ preservation, ischemia-reperfusion (I/R) injury and the toxicity of immunosuppressive medication (5,6). Nevertheless, the molecular mechanisms that link early graft vessel injury to subsequent rejection mediated by the host's adaptive immune system remain uncertain.

The standard technique to evaluate rejection in heart transplantation is the retrieval of endomyocardial biopsies taken at regular intervals after transplantation. Such biopsies generally show high sensitivity and specificity but have shortcomings such as reduced sensitivity associated with samples of mild rejection (7,8), as well as risks associated with the invasive procedure. Thus, there is great interest in developing noninvasive diagnostic techniques, and the proximity of graft endothelial cells to the circulation makes them an attractive putative source of soluble products amenable to analysis of blood samples.

During the last decade, investigators have partly dissected the complexity of allograft rejection by means of microarray-based genome-wide expression analyses. It has become clear, however, that restricting such analyses to whole tissue samples is a major limitation to understanding the interplay between different cell types during rejection. Moreover, with a few exceptions (9,10), most of these studies have analyzed the transcription profiles of allografts and focused less on changes associated with the surgical insult and I/R injury (11–14).

Based on these considerations, we established the endothelial cell transcriptome during transplantation and allograft rejection. Cluster analyses of the transcription profiles revealed three categories of endothelial responsiveness: categories A and B contained profiles with similar responses in isografts and allografts and most likely reflect changes associated with the surgical procedure, whereas category C contained profiles associated with allograft rejection. Interestingly, we observed that most of the modulated transcripts were not associated with the allograft response but appeared to be associated with I/R injury and tissue remodeling.

Materials and Methods

Heterotopic heart transplantation

Hearts from male Dark-Agouti-RT1a (DA) or male Lewis-RT11 rats (Harlan CPB, The Netherlands) were grafted to the neck of male Lewis rats using a modification of the cuff technique described by Heron and Olausson et al. (15,16) as detailed in the online Supporting Information.

Endothelial cell and leukocyte enrichment

Endothelial cells and leukocytes were enriched from minced and collagenase-dispersed heart tissue by two rounds of magnetic bead separation. First, we depleted the cell suspension of CD45-positive leukocytes since these cells are also CD31-positive. Second, we enriched the CD45-negative population for CD31-positive endothelial cells. A detailed protocol is provided in the online Supporting Information.

RNA isolation and estimation of leukocyte and myocyte contamination

RNA isolation and quantitative RT-PCR are described in the online Supporting Information.

The level of leukocyte contamination in the endothelial cell fraction was estimated by comparing the CT of CD45 in the leukocyte and endothelial cell samples. Considering that the CD45 content in the leukocyte fraction is 100% and the amplification efficiency 2-fold, we estimated that a reduction of, for example, 5 cycles from the leukocyte fraction to the endothelial cell fraction corresponds to a 33-fold reduction (25) or 3% leukocyte contamination in the endothelial fraction. By this estimate, endothelial cell samples used for microarray analysis had leukocyte contamination levels in the range of 0.3–3.0% (median = 1.4%). An outlier of 6.8% in one of the Lewis control samples was considered to be caused by suboptimal amplification efficiency. Similarly, the level of myocyte contamination was estimated by comparing the CT of desmin in endothelial cell samples and whole tissue samples, indicating that the myocyte contamination in the tested samples were in the range of 0.4–2.7% (median = 0.9%).

Gene expression analysis

RNA (1.5–3.5 μg) was processed and hybridized to the GeneChip Rat Genome 230 2.0 Array (∼31,000 probes) as described in the manufacturer's ‘one-cycle’ protocol (Affymetrix, Santa Clara, CA). Endothelial cells isolated from control hearts and day 2, 3 and 4 after transplantation (post-tx) allografts and isografts were analyzed (n = 3, 24 arrays in total). In addition, one leukocyte sample was analyzed for each of days 2 and 4 post-tx, and two samples for day 3. Data were collected using the GeneChip Scanner 3000 (Affymetrix), and all data sets are available in the Gene Expression Omnibus (GSE16695).

The endothelial cell sample CEL files were preprocessed under the R Bioconductor package including normalization by robust multichip averaging (17). In addition, we removed probes with uncertain detection (i.e. >85% absent calls) or little variation (i.e. less than 2-fold difference between maximum and minimum intensity) across all experiments, obtaining 9491 probes for further analyses. To avoid identifying transcripts due to leukocyte contamination as differentially expressed, we compared all upregulated transcripts in the endothelial sample to the corresponding transcripts in the leukocyte sample and removed 1467 probes that had a higher expression value in the latter. We assumed that downregulated genes were not affected by contamination. After these preprocessing steps, we obtained 8024 probes whose intensity values were transformed to Z-scores and assigned probe annotation information from the Affymetrix NetAffx data base (as of May 2008) before using MATLAB to perform the time-series analyses.

The following analyses were designed to identify two principal populations of modulated transcripts: first, a population that is significantly affected by transplantation but not graft incompatibility (allografts vs. controls); and second, a population that is significantly affected by both transplantation and graft incompatibility (allografts vs. isografts). To identify the first population, we applied a modified version of the pairwise Fisher's linear discriminant method (18) to the 12 allograft arrays and selected the top 10% most differentially expressed probes for further studies. To identify the second population, we randomly shuffled the order of the replicate arrays within the same post-tx day, computed the relative ratios between allografts and isografts (i.e. three ratios from each day) and then applied the same Fisher's method as above. The random shuffling was repeated 100 times and the top 10% most differentially expressed probes were selected based on the average rank in these 100 tests. To obtain a list of probes only identified by the first analysis (allografts vs. controls), we subtracted the probes in the second population from the probes in the first population. Finally, to study the selected probes from both populations in more detail, Fuzzy K-nearest neighbor and stress function was used to classify probes into gene clusters based on their expression profiles (18) as described in the online Supporting Information. Clusters from the second population were divided into two categories by using Pearson's correlation to compare differential regulation, with cutoff at r = 0.3. The degree of expression coherence within clusters (noise) was analyzed by comparing the profile of each probe to the cluster center using Pearson's correlation, with cutoff set to rc= 0.6.

Patient data

Six renal allograft biopsies taken on clinical indication (elevated creatinine levels) were included, showing either no signs of rejection (n = 3) or Banff grade IB rejection (n = 3). In addition, three control biopsies were analyzed (two baseline biopsies and one taken to exclude disease). Additional information can be found in Supporting Table S1.

Immunohistochemistry

Sections from fresh frozen or formalin-fixed tissue blocks were stained using standard protocols as described in the online Supporting information. Primary antibodies are listed in Supporting Table S2.

Results

Graft survival and histological evaluation

We used a rat strain combination that rejects allografts at day 6 post-tx, whereas isografts survive for more than 90 days. Accordingly, allografts showed massive infiltration of immune cells in the myocardium from day 3 post-tx, peaking at day 5, whereas isografts showed no infiltration of immune cells in the myocardium at any time point. However, a mixed leukocyte infiltration in the pericardium was present in all grafts, most likely due to surgical-induced damage.

Cluster analysis of modulated transcripts

To study the activation profile of vessels after solid organ transplantation, we enriched endothelial cells from normal hearts (controls) and at day 2, 3 and 4 post-tx for microarray analysis. These time points were selected because we were interested in the vascular changes preceding lymphocyte infiltration and graft damage, yet assuming, based on previous studies, that day 1 responses in allografts and isografts are very similar (19,20). On the other hand, allografts day 5 and 6 post-tx contain high amounts of leukocytes making it difficult to obtain pure enough endothelial cell fractions for analysis.

Bioinformatics analyses identified two principal populations of modulated genes: one that changes in the course of transplantation independent of graft incompatibility and another that responds to allograft rejection. We identified 442 probes belonging to the first population and clustered them into 20 clusters. One cluster contained nonmodulated transcripts and was excluded from further analyses. The remaining 19 clusters are in the following called Category A (CA) clusters (Supporting File), and consisted of 7 up- and 12 downregulated clusters (examples shown in Figure 1A). The second population contained 804 probes that were clustered into 30 clusters. Eleven were removed from further analyses due to outliers on day 4 or low expression coherence within clusters (rc < 0.6). The expression profiles of the remaining 19 clusters (12 and 7 containing up- and downregulated transcripts, respectively) were highly heterogeneous. Although transcripts were differentially regulated in allografts compared to isografts, some expression profiles were highly similar upon visual inspection, whereas others revealed a strong response in allografts and only a modest or even absent response in isografts. Thus, we divided the second population into two categories based on the correlation between allograft and isograft profiles: Category B transcripts showed similar expression profiles in both groups with correlation coefficients >0.3 (a total of 10 clusters (CB) listed in the Supporting File, examples shown in Figure 1B), whereas Category C transcripts showed different expression profiles in allografts compared to isografts with correlation coefficients <0.3 (a total of 9 clusters (CC) listed in the Supporting File, examples shown in Figure 1C).

Figure 1.

Cluster analysis of gene expression in graft endothelial cells. Selected clusters from Category A, B and C are shown. Cluster identity is shown in the upper or lower left corner and selected transcripts in each cluster are shown in the tables. Genes depicted in bold have been verified by immunostaining. Each gray line in the clusters represents one probe and the black line represents the cluster center with each replicate array indicated by an open square. Abbreviations: C, control/not transplanted; d2, day 2 post-tx; d3, day 3 post-tx; d4, day 4 post-tx.

Category A clusters—A global injury response

Category A clusters (similar expression profile in allografts and isografts) consisted of 433 probes representing 206 known genes (Figure 1A). Clusters with upregulated transcripts approached top levels at day 2 and remained high until day 4 (CA8, 13 and 16) or showed a slight dip in allografts on day 3 post-tx (CA2, 5, 6 and 15). Conversely, clusters with downregulated transcripts showed a persistent reduction from day 2 (CA3, 7, 11 and 18) or a slight rebound in allografts on day 3 post-tx (CA1, 4, 9, 12, 14, 17 and 19–20). Given the similar expression pattern of these transcripts in allografts and isografts, we presumed that they might represent responses to common insults such as I/R injury and other elements of surgical trauma. This hypothesis was supported by a screen of the relevant literature (Table 1). Moreover, a gene ontology analysis was conducted using the PANTHER classification system (http://www.pantherdb.org), reaching significance in cell communication, immunity and defense, as well as in signal transduction when using the known genes represented on the chip as reference (Supporting Table S3A). If using the more common reference to the whole rat genome, additional processes such as cell proliferation, cell motility and cell adhesion reached significance (Supporting Table S3B). However, we also observed that several transcripts known to be involved in processes relevant to endothelial cell behavior were not classified in these data bases, perhaps explaining why they failed to reach significance. We therefore supplemented these analyses with a screen of the relevant literature, observing that transcripts important to cell adhesion, extracellular matrix (ECM) modulation, angiogenesis, apoptosis and also cytokines and their receptors were modulated (Table 2).

Table 1.  Response to I/R injury and interferon-γ
FunctionRegulationCategory ACategory BCategory C
  1. 1Some transcripts are represented by several probes and are therefore found in more than one category.

HypoxiaUpAceActa2Agtr1a
Oxidative Adam9Angpt2Apelin
Stress Adamts8Angptl4Bnip3
I/R injury Bmp4C1sCx3cl1/fractalkine
  BokCeruloplasminDdr2
  Ceruloplasmin1Col1a1Igfbp3
  Col1a1Cxcl10/IP-10Indo/IDO
  Col18a1E-selectinMtf1
  Daf1Fgf23P4ha1
  Edn1Fkbp5Pfkfb3
  Entpd1/CD93LbpPsmb8
  Fosl2LoxStat1
  Igfbp2Mt1aStat3
  Il6Pawr/Par-4 
  Mmp14P-selectin 
  PlvapSerpine1/PAI-1 
  PeriostinTnc 
  Psen2  
  Ptn  
  Thy1/CD90  
  Tnfrsf1a  
  Trpv1  
  Vcam1  
 DownCcnd1 calpastatin
  Cdh2 Unc5b
  Jup  
  Thbd  
  Unc5b  
IFN-γUpAceAngpt2Apelin
response Adamts7C1sC1r
  Adamts8CeruloplasminCasp4
  C4aCxcl9/MigCx3cl1/fractalkine
  CeruloplasminCxcl10/IP-10Cyp27b1
  Daf1Cxcl11/I-TACIgfbp3
  Edn1Gbp2Igtp
  Tap2Isg12Il18bp
  Upp1TncIndo/IDO
  Vcam1UbdIrgm/Lrg47/Ifi1
   WarsNub1
    Psmb8
    Psmb9
    Psmb10
    Sectm1b
    Serping1/C1INH
    Sp140
    Stat1
    Stat3
    Tap1
    Ube2l6/Ubch8
    Usp18
    Wars
 DownCcnd1  
  Cxcl12/SDF-1  
  Thbd  
Table 2.  Functional classification
FunctionRegulationCategory ACategory BCategory C
  1. 1Some transcripts are represented by several probes and are therefore found in more than one category.

AngiogenesisUpAdamts8Angpt2Apelin
  Bmp4Angptl4Igfbp3
  Col18a1Serpine1/PAI-1Il15
  Mmp14/MT1-MMP  
  Ptn  
  Tnfrsf12a/tweakr  
  Tnfrsf1a/Tnfr1  
 DownDll4Klf5Unc5b
  Fgfr1Wasf2/Wave-2Wt1
  Wt11  
ApoptosisUpBokAngpt2Apelin
  CeruloplasminAngptl4Bnip3
  Col18a1C7///Tubb2Casp4
  Il6CeruloplasminFhl2
  MycMt1aIgfbp3
  Nme2Pawr/Par-4Igtp
  Pmp22Serpine1/PAI-1Il15
  Psen2/PS-2UbdIndo/IDO
  Ril/Pdlim4 Mx1
  Thy1/CD90 Nol3
  Timp1 Ripk2
  Tnfrsf1a/Tnfr1 Serping1/C1INH
    Stat1
    Stat3
    Thy1/CD90
    Timm8a
    Usp18
 DownAldh1a1Klf5Hspb8
  Amigo2Mc4rUnc5b
  Casp2  
  Fgfr1  
  Map3k1  
  Mxi1  
  Phf17  
  Rarb  
  Tlr4  
  Tnf  
CellUpAdam9E-selectinCdh11
adhesion Fath/Fat1P-selectinCx3cl1/fractalkine
  Gja4/Cx37 Sectm1b
  Pmp22 Thy1/CD90
  Pvrl2/Nectin-2  
  Thy1/CD90  
  Vcam1  
 DownAmigo2 C1qr1/CD93
  Cdh2 Pcdh17
  Cdh23 Podxl
  Cldn11  
  Pvrl3/Nectin-3  
CytokinesUpBmp4Cxcl9/MigCcl19/ELC
and their Csf2/GM-CSFCxcl10/IP-10Cx3cl1/fractalkine
receptors Igfbp2Cxcl11/I-TACIgfbp3
  Il6Fgf23Il15
  PtnInhbb/activin BIl18bp
  Tnfrsf12a/tweakr  
  Tnfrsf1a/Tnfr1  
 DownCxcl12/SDF-1Kitl/SCFFgfr3
  Fgfr1Tgfbr1Ifnar1
  Ghr Ltbp1
  Ltbp4 Ntf3
  Tnf  
ECMUpAceAngpt2Agtr1a/AT1
modulation Adam9Angptl4Ddr2
  Adamts7Col1a1Fhl2
  Adamts8Col12a1Itih3
  Bmp4Dmbt1Nov/Ccn3
  Col1a1Fgl2P4ha1
  Col1a2Lox 
  Col3a1Pcolce 
  Col6a1Serpine1/PAI-1 
  Col6a2Tnc 
  Col18a1  
  Cthrc1  
  Edn1  
  Efemp2  
  Ehd4  
  Fgl2  
  Mmp14/MT1-MMP  
  Periostin  
  Ptn  
  Timp1  
  Tnfaip6/TSG-6  
  Vwa1  
 DownB3gnt7Lamc2Calpastatin
  Dpp4 Ltbp1
  Etv5  
  Ltbp4  
  Mmp15/MT2-MMP  
  Tnn/Tn-w  

The changes in transcript levels were verified by immunohistochemical analyses of protein expression for a few selected genes in each category (Figure 2). The presence of a pericardial infiltrate is a well-known response in heterotopic grafts (2) and a potential source of error in whole tissue analysis, and was therefore avoided when evaluating the selected transcripts. Vascular cell adhesion molecule 1 (VCAM1), a well-established marker of endothelial cell activation, showed similar induction levels in allografts and isografts. Immunostaining for VCAM1 was negative in control hearts, weakly positive in a few scattered vessels in day 1 grafts and further increased on day 2 post-tx. Staining intensities were further enhanced in days 3 and 4 allografts (Figure 2) and the strongly positive vessels were often associated with areas of infiltrating cells. In isografts, the strongly positive vessels were more evenly distributed and staining intensities were similar on day 2, 3 and 4 post-tx.

Figure 2.

Regulation at the protein level in rat hearts. Immunostaining for selected gene products and phosphorylated STAT1 (last row) in control tissue (DA, left column), day 3 allografts (center columns) and day 3 isografts (right column). Protein of interest is shown in red or gray scale, vessel marker in green (CD31 or VE-cadherin) and nuclear marker (Hoechst) in blue. Scale bar = 50 μm.

Many of the differentially regulated transcripts encoded structural components of the ECM, modulators of ECM–cell interactions and vascular remodeling/angiogenesis (Table 2). Given that several transcripts involved in TGF-β (transforming growth factor beta) signaling were modulated, we assessed the relevance of these observations by immunostaining for the downstream product periostin, one of the most strongly upregulated transcripts in our analysis and a secreted matricellular protein that is associated with tissue regeneration after injury (21). Periostin was strongly induced in the interstitial space on day 3 and 4 post-tx and more prominent in allografts than in isografts (Figure 2). However, there were large differences in staining intensity between replicates as also seen in the microarray data (cluster CA15 in the Supporting File). Human kidney grafts with and without acute cellular rejection also showed a strong increase of periostin interstitial deposition compared to control kidneys (data not shown), pointing to a possible relevance to clinical transplantation. Kidney biopsies were preferred to cardiac biopsies because they contain vessels of several categories as opposed to heart biopsies that mostly contain capillaries.

Category B clusters—Trauma response enhanced by alloresponses

Clusters in Category B had profiles similar to those in Category A (compare Figure 1A and B) and consisted of 156 probes representing 83 known genes. These transcripts were also assigned to I/R injury (Table 1) and other responses seen in Category A (Table 2), as well as to interferon- and macrophage-mediated immunity (Supporting Table S3A), indicating that the stronger transcriptional increase seen in Category B allografts compared to isografts might be due to injury responses further increased by alloresponses (22).

Immunostaining for E-selectin, also a well-established marker of endothelial cell activation, showed that it was generally absent in control tissues but strongly induced in scattered vessels post-tx (Figure 2), corresponding well with the transcriptional values.

The matricellular protein tenascin-C (TN-C, Table 2), another highly upregulated ECM transcript, affects cell adhesion and migration (23) and is induced at sites of inflammation, repair and regeneration (24). Immunostaining revealed that it was mostly absent in control tissues but upregulated in all transplants as early as day 1 post-tx (Figure 2). TN-C was seen in scattered endothelial cells (Figure 2, allograft detail) but the staining intensity was even stronger in the interstitial space, most likely due to its secretory nature.

The chemokine (C-X-C motif) ligand 11 (CXCL11/I-TAC) is thought to be important during allograft rejection (25,26). Immunostaining showed weak-to-absent signal in control tissue with the exception of a few larger positive vessels in isograft controls. In isografts, there was absent or weak focal expression in capillaries. By contrast, allografts showed strong staining in capillaries at day 2 to 4 post-tx (Figure 2) and in infiltrating cells (leukocytes and macrophages) at days 3 and 4 post-tx. CXCL11 was also present in smooth muscle cells in small arteries on all days in both graft types (Figure 2, control inlet).

Category C clusters—Rejection-associated responses

Category C clusters contained 234 probes representing 125 known genes that were predominantly or exclusively regulated in allografts (Figure 1C). Clusters with upregulated transcripts either reached a plateau on day 2 (CC11), or showed a steady day-by-day increase (CC1, 3, 13, 22 and 27) as opposed to the majority of clusters in Categories A and B. Clusters with downregulated transcripts showed a slight rebound on day 3 post-tx (CC20, 25 and 30). Although gene ontology analyses failed to identify significant biological processes in category C when using the known genes on the chip as reference (Supporting Table S3A), several processes reached significance when using the whole rat genome as reference. Among them were cell communication, signal transduction, immunity and defense, cell structure and motility as well as apoptosis (Supporting Table S3B).

Moreover, a large number of transcripts were associated with IFN-γ-induced responses (Table 1) (27). Accordingly, we observed a strongly increased phosphorylation of the transcription factor STAT1 (signal transducer and activator of transcription 1) in allograft but not isograft vessels (Figure 2), compatible with the assumption that only allografts are sufficiently exposed to this cytokine or other activators of STAT1. Most transcripts in category C were preferentially expressed in allografts but to some extent also induced in isografts, and we observed that the IFN-γ-inducible enzyme indoleamine-2,3-dioxygenase (IDO1/Indo) (28,29) appeared to be specifically induced in allograft vessels. However, efforts to confirm the data in our rat model were fruitless because the polyclonal antibody, in disagreement with our array data and the current literature in human and rodent heart (29–32), stained all vessels in control hearts, and several other anti-IDO1 reagents gave inconclusive results (data not shown). On the other hand, immunostaining with several antibodies in human tissue revealed that IDO1 was undetectable in vessels in the control kidney but expressed in scattered vessels in all kidney grafts (Figure 3B). Weak staining in a few vessels was also found in the baseline biopsies, perhaps due to I/R events (data not shown) (33).

Figure 3.

IDO1 expression in human kidney grafts. Immunostaining for IDO1 in normal kidney (left column) and in transplanted kidney without rejection (center) and with rejection (right column). IDO1-positive vessels (red or gray scale) are indicated by arrows. Vessel marker is shown in green (CD34) and nuclear marker (Hoechst) in blue. Scale bar = 50 μm.

Discussion

This study was designed to establish how endothelial cells in solid organ grafts respond to the transplantation procedure and the allogeneic response of the host. It revealed that transcriptional changes associated with the surgical procedure rather than the allograft incompatibility were the major feature of endothelial cell activation. Surgical trauma is partly caused by I/R injury, thought to be mediated by the stabilization of hypoxia-inducible factor 1 alpha (34) and the subsequent enhanced production of reactive oxygen species. These events have in several studies been shown to increase the expression of a large variety of mediators reflecting endothelial cell activation (35,36) and were accordingly among the transcripts found in Categories A and B (Table 1), such as several leukocyte-endothelial cell adhesion molecules, collagens and angiogenic mediators. Although previous studies addressing the cyto- and chemokine profiles of whole-graft tissues have made similar conclusions at time points earlier than 48 h (19,20,37), our observation that this is also true for an isolated cell population up to 4 days post-tx is novel (Table 2).

Another interesting property of our data is the observation that several of the differentially regulated genes are involved in modulation of the ECM (Table 2), an observation well in line with previous analyses of murine kidney transplants (22). Among the transcripts involved in matrix modulation, we verified the induction of the matricellular proteins TN-C and periostin, and identified good correlations between the microarray data and the immunostaining. Enhanced TN-C expression has been observed in renal allografts (38), whereas periostin-induction in transplantation is a novel observation that also demonstrated the usefulness of our single-cell population approach. In combination, the joint upregulation of TN-C and periostin may facilitate an increased influx of cells eventually causing ECM modulation and fibrosis in allografts. Thus, despite their upregulation in isografts, which can be viewed as a stage of successful repair that goes back to baseline levels, it is possible that the allocation of TN-C to category B (as well as other mediators in this category) reflects changes related to the enhanced upregulation of injury and repair-associated transcripts in allografts observed by Famulski et al. (22). This is further supported by the clinical observation that early rejection episodes or a persisting low-grade inflammation may lead to fibrosis and transplant vasculopathy (39).

Although we prefer the interpretation that the similar gene expression pattern observed in allografts and isografts is at least partly driven by I/R injury, there is currently no available technique for orthotopic cardiac transplantation in rodents and lack of pressure load in heterotopically transplanted hearts remains a possible contributing factor. However, the expression of periostin and TN-C in human kidney grafts indicate that the injury-related and ECM-modulating transcripts are not related to a model artifact but rather a stereotyped response of injured/rejecting organs.

The expression profile associated with allograft responses in our study contained a large proportion of transcripts associated with the effects of IFN-γ such as IDO1. This assumption was further emphasized by the allograft-specific phosphorylation of STAT1 in vessels. However, it should be kept in mind that other cytokines also signal via STAT1 and that several targets known to be IFN-γ-responsive appeared to be either aberrantly regulated in the endothelial cells or affected by other signals (Supporting Table S4).

Relatively few of the most strongly modulated transcripts (about 30%) were associated with rejection, but this does not exclude an important role of endothelial cells in allograft responses. One may be the vascular expression of intercellular adhesion molecule 1, which has been shown to affect graft survival but was removed from our data set after leukocyte-related filtration, another the presentation of chemokines that enable firm adhesion of lymphocytes on the endothelial cell surface (40). Furthermore, recent evidence suggests that tissue damage associated with surgery leads to release of damage-associated molecules from endothelial cells, such as alarmins that subsequently affect the course of the immune response (3,41). Such posttranscriptional regulated release can not be assessed by measuring transcriptional changes.

Although the Gene Ontology analyses (Supporting Table S3) implies that some transcripts may originate from a small proportion of cardiomyocytes as evidenced by the remaining small levels of desmin transcripts in our data set, we noted that several transcripts thought to be myocyte-derived are also expressed in cultured endothelial cells (GEO profiles). This uncertainty and also our experience that increased efforts to purify endothelial cells drive them away from their native state (42–44), underscores the importance of verifying microarray data from single-cell populations with multiparameter immunofluorescence analyses of whole tissue.

In conclusion, this study demonstrates that vessels respond strongly to the surgical trauma of transplantation and that the majority of the modulated transcripts are associated with I/R responses and tissue remodeling. Nevertheless, a distinct group of transcripts was associated with the development of an allograft response, partly reflecting a response to IFN-γ.

Acknowledgments

We would like to thank Vivi Bull Stubberud at the Division of Surgery and Hogne Røed Nilsen, Aaste Aursjø, Vigdis Wendel, Linda Manley, Kathrine Hagelsteen and Linda I. Solfjell at LIIPAT, Institute of Pathology for excellent technical assistance.

This work was supported by a grant from the Norwegian Foundation for Health and Rehabilitation to BM, by a grant from the Norwegian Health Association to HB and ML, and by grants from South-Eastern Norway Regional Health Authority to MK and GH.

Disclosures

None.

Ancillary