Epithelial to Mesenchymal Transition During Late Deterioration of Human Kidney Transplants: The Role of Tubular Cells in Fibrogenesis


*Corresponding author: Philip F. Halloran, phil.halloran@ualberta.ca


The hallmark of failing renal transplants is tubular atrophy and interstitial fibrosis (TA/IF). Injury to tubular epithelial cells (TEC) could contribute to fibrogenesis via epithelial–mesenchymal transition (EMT). We examined the features of EMT in renal transplants that developed TA/IF. Biopsies from 10 allograft kidneys with impaired function and TA/IF and 10 biopsies from transplants with stable function were compared to their implantation biopsies. Relative to implantation biopsies, TEC in TA/IF kidneys showed loss of epithelial markers (E-cadherin, cytokeratin) with altered distribution. Some TEC also showed new cytoplasmic expression of mesenchymal markers vimentin, S100A4, and alpha smooth muscle actin (α-SMA) and collagen synthesis marker heat shock protein (HSP-47), both in deteriorating and atrophic tubules. Double immunostaining showed coexpression of cytokeratin and vimentin, S100A4 and HSP-47, indicating intermediate stages of EMT in TA/IF. These changes were absent or much less in transplants with stable function. EMT features in the TA/IF group correlated with serum creatinine (vimentin, S100A4, HSP-47), history of T-cell-mediated rejection (cytokeratin, S100A4) and proteinuria (cytokeratin). These findings support a model in which the TEC damage induces loss of epithelial features and expression of fibroblast features, as a common pathway of deterioration by either immunologic or nonimmunologic processes.


Despite recent improvements in graft survival, the late deterioration of renal allografts remains a significant problem (1). Thus although kidney transplants are now more stable (2), about 5000 kidneys fail each year in the United States with the recipient returning to dialysis (3), making renal transplant failure a major cause of end stage renal disease. The disease processes contributing to these failures are multiple, but the main phenotype of failing renal transplants is tubular atrophy (TA) and interstitial fibrosis (IF). TA/IF developing after transplantation is evidence of nephron loss due to the burden of post-transplant immunologic and nonimmunologic injury (4). TA/IF with slow deterioration of graft function has been termed as ‘chronic allograft nephropathy’ in the Banff system of classification and graded by the chronic change scores CT CI CV (5) or simply as ‘allograft nephropathy’ (AN) (6). However, because all attempts to classify kidney transplant failure are dominated by TA and IF and carry the possibility of ambiguity, we may simply refer to the lesions as TA/IF. Many cases also demonstrate fibrous intimal thickening (FIT) of arteries (7), which can be of donor origin and does not necessarily imply previous arteritis or direct arterial injury. Indeed, FIT is common in all kidneys that lose mass and may reflect disuse due to loss of the capillary bed (8). FIT and TA/IF lesions correlate with donor age, cold ischemic time, brain death, drug toxicity as well as rejection (6). The extent of TA/IF is a prognostic marker for renal allograft failure (9), probably because it indicates lower nephron number and reduced reserve.

While the pathogenesis of fibrosis in failing kidneys is complex, it is increasingly likely that tubular epithelial cells (TECs) contribute to renal fibrogenesis (10). The TECs function as a source of fibrogenic growth factors and chemokines in the initiation of fibrogenesis, contribute to tubular atrophy by undergoing apoptosis and potentially contribute through epithelial-mesenchymal transition (EMT) (10). EMT is defined as the acquisition by epithelial cells of the phenotypic and functional properties of mesenchymal fibroblasts (11). This phenomenon occurs in organogenesis (11), malignant transformation and tumor progression (12) and as a mechanism of fibrogenesis in chronic kidney diseases (13). A role for EMT in renal fibrogenesis is suggested by studies on cultured cells (14), animal models of kidney diseases (10,15) and human nephropathies (13). Human kidney transplants with dysfunction show S100A4, an EMT maker, in association with CD8 lymphocytes (16).

To evaluate the role of TECs in fibrogenesis via EMT in human renal allografts developing TA/IF, we performed an immunohistochemistry study utilizing epithelial, mesenchymal and collagen synthesis markers to examine the relationship between TA/IF and the features associated with EMT. We studied the renal tubules for epithelial markers E-cadherin and cytokeratin staining, mesenchymal markers (vimentin, S100A4, and α-SMA) and collagen synthesis marker HSP-47. We took advantage of the implantation biopsies in our center at the time of transplantation as controls, to determine whether such features preexist in donor biopsies and used kidney biopsies from transplants with good preservation of function and morphology to see if these changes could occur in kidneys lacking TA/IF.

Materials and Methods

Patients and kidney tissue

Ten renal transplant recipients biopsied for renal dysfunction, whose biopsies showed TA/IF (chronic allograft nephropathy by Banff) and who had implantation biopsies available, and 10 protocol biopsies from transplants with stable graft function (used as controls) were included in this study. As standard practice, all kidneys transplanted in our center are biopsied at the time of implantation creating a total of 40 biopsies for analysis. All cases were transplanted between 1988 and 1998 at the University of Alberta Hospital. Clinical data was reviewed from the medical records and renal transplant clinical files. The study was approved by the local institutional review board of the Capital Health and the University of Alberta. Standard immunosuppression during this period was cyclosporine, azathioprine (AZA) and prednisone until 1996 when mycophenolate mofetil (MMF) replaced AZA. From 1997 to 1999, MMF was changed to AZA at 1 year in stable grafts. The patients were selected from the clinical transplant data base by the following criteria: biopsy proven diagnosis of chronic allograft nephropathy defined by Banff CT CI CV scores (5); no evidence of alloantibody-mediated rejection or endothelialitis and only minor degrees of tubulitis; and availability of paraffin embedded specimens obtained at the time of transplantation and biopsy. We excluded recurrent diseases and other specific diagnoses such as polyoma virus nephropathy. The biopsies from patients with stable function were selected on similar criteria. All patients were currently on long-term calcineurin inhibitor therapy.

Immunoperoxidase staining

Antibodies against cytokeratin (mouse monoclonal antibody, clone MNF116, Dako, Mississauga, Ontario) and E-cadherin (mouse monoclonal antibody, clone HECD-1, Zymed, San Francisco, CA) were used to detect epithelial markers. Mesenchymal markers were stained with antibodies against vimentin (mouse monoclonal antibody, clone V9, Dako, Mississauga, Ontario), alpha-smooth muscle actin (α-SMA) (mouse monoclonal antibody, clone 1A4, Dako, Mississauga, Ontario) and S100A4 (rabbit polyclonal antibody, Dako, Mississauga, Ontario). In addition, anti-heat shock protein-47 (HSP-47) mouse monoclonal antibody (clone M16.10A1, Stressgen, Victoria, British Columbia) was applied to detect collagen synthesis marker.

Immunoperoxidase staining was performed using 2 μm sections of paraffin embedded tissue. Briefly, sections were deparaffined and hydrated, then immersed in 3% H2O2 in methanol to inactivate endogenous peroxidases. Slides were pre-incubated with 20% normal goat serum to block nonspecific binding. Tissue sections were then incubated with the primary antibody and with the Envision monoclonal system (Dako, Mississauga, Ontario) as the secondary antibody. Visualization was performed using the 3, 5-diaminobenzidine (DAB) substrate kit (Dako, Mississauga, Ontario). Slides were counterstained with hematoxylin, dehydrated and mounted with Permount (Fisher Scientific, Fair Lawn, NJ). The primary antibody was substituted with isotype-matched antibody as a negative control. E-cadherin staining was done using frozen sections.

Double staining

A double staining method was performed to demonstrate the phenotypic changes of epithelial cells into mesenchymal cells, using the antibody against cytokeratins coupled with antibodies against mesenchymal markers or HSP-47. A microwave-based two color-staining technique was applied (17). Sections were treated as for the single stain from the beginning of the procedure to the development with DAB to produce a brown color, except for the use of peroxidase-conjugated goat anti-mouse IgG and mouse peroxidase anti-peroxidase complexes (PAP) (Dako, Mississauga, Ontario) instead of Envision monoclonal system. To block antibody cross reactivity and facilitate antigen retrieval, sections were treated with two rounds of microwave heating, each lasting 5 min, in 10 mM sodium citrate pH 6.0 at 2450 MHz and 800 W power. They were sequentially cooled, pre-incubated as described above and incubated with the second primary antibody followed by alkaline phosphatase-conjugated goat anti-mouse IgG and mouse alkaline phosphatase anti-alkaline phosphatase complexes (APAAP) (Dako, Mississauga, Ontario). Finally, sections were developed with Fast Red (Dako, Mississauga, Ontario) to produce a red color, counterstained with hematoxylin and mounted with Permount.

Quantification of immunohistochemistry

Analysis was performed by counting 10 high power fields (HPFs, ×200), focusing on TECs. Expression of cytoplasmic staining for mesenchymal and collagen markers was evaluated by counting the percentage of positively stained cells per HPFs. Loss of cytoplasmic staining of epithelial markers was assessed by counting the percentage of stain-negative cells per HPFs.

Statistical analysis

The statistics were analyzed using SPSS 11.0 for Windows. Unpaired t-test was used to compare individual expression of markers across groups. Results are expressed as mean ± SD. Correlation analysis comparing clinical parameters to marker expression used Pearson's correlation coefficient for parametric data and Spearman's correlation coefficient for nonparametric data. The Mann-Whitney test was used to compare immunohistochemistry results in relation to rejection history in the TA/IF biopsies. A p-value <0.05 was considered statistically significant.


Demographic data

The clinical data from 20 renal transplant recipients are shown in Table 1, representing 10 patients in which biopsies for impaired function revealed TA/IF post-transplant and 10 control patients with stable renal function in which biopsies were obtained as part of specific protocols. All cases were on calcineurin inhibitor therapy. We excluded cases in which the major diagnosis was a specific disease such as recurrent glomerulonephritis. The serum creatinine in the TA/IF group was higher than in the protocol biopsy group, by definition. Six out of 10 patients in TA/IF group had significant proteinuria compared to none in the protocol biopsy group. Hypertension was similar in both groups. Morphologic evaluation showed arteriolar hyalinosis in 7 out of the 10 cases in the TA/IF group but only one out of the 10 cases in the protocol biopsy group.

Table 1.  Patient demographics
with TA/IF
  1. *Defined as proteinuria > 0.5 g/24 h.

  2. **Defined as treatment with antihypertensive drugs.

  3. ***Defined as new or increased hyalinosis compared with implantation biopsies.

Cadaveric/living donor7/310/0
Recipient gender (M/F)7/39/1
Donor gender (M/F)8/23/7
Recipient age (years)41.8 ± 12.744.1 ± 7.7
Donor age (years)40.1 ± 12.735.4 ± 12.8
Cold ischemic time (hours)11.5 ± 8.317.6 ± 4.4
Timing of biopsy post KT (years)4.7 ± 3.12.2 ± 1.4
Creatinine at biopsy (μmol/L)257.2 ± 82.0127.7 ± 8.5
Proteinuria (g/24 h)0.7 ± 0.60. 05 ± 0.07
Significant proteinuria (yes/no)*6/40/10
Hypertension (yes/no)**9/19/1
History of rejection (yes/no)4/61/9
Arteriolar hyalinosis (yes/no)***7/31/9


All biopsies were regraded according to the Banff 1997 classification (5) (Table 2). By definition, the implantation biopsies showed minimal pathological changes. The protocol biopsies from transplants with stable function showed minor degree of tubulitis, infiltrate, atrophy and fibrosis in some cases. Biopsies for dysfunction with TA and IF showed some infiltration and tubulitis and more arterial FIT. Quantitative evaluation confirmed the increased arteriolar hyaline thickening in the TA/IF group compared to the protocol biopsy group.

Table 2.  Banff scores in implantation biopsies, biopsies for dysfunction in transplants with TA/IF group and protocol biopsies with stable function in control transplants group
Group Banff scoreTransplants with TA/IFControl transplants
Implantation biopsy
(N = 10)
Biopsy for dysfunction
(N = 10)
Implantation biopsy
(N = 10)
Protocol biopsy
(N = 10)
  1. *Statistically significant difference (p < 0.01) comparing biopsies with dysfunction in the transplants with TA/IF group to the protocol biopsies in the control transplants group.

Glomerulitis (G)0.0 ± 0.00.0 ± 0.00.0 ± 0.00.0 ± 0.0
Interstitial infiltrate (I)0.1 ± 0.30.9 ± 0.30.0 ± 0.00.6 ± 0.5
Tubulitis (T)0.0 ± 0.00.7 ± 0.50.0 ± 0.00.3 ± 0.5
Vasculitis (V)0.0 ± 0.00.0 ± 0.00.0 ± 0.00.0 ± 0.0
Allograft glomerulopathy (CG)0.0 ± 0.00.1 ± 0.30.0 ± 0.00.0 ± 0.0
Interstitial fibrosis (CI)0.0 ± 0.02.2 ± 0.4*0.0 ± 0.00.7 ± 0.5
Tubular atrophy (CT)0.0 ± 0.02.2 ± 0.4*0.0 ± 0.00.7 ± 0.5
Arterial fibrous intimal thickening (CV)0.0 ± 0.01.5 ± 1.0*0.0 ± 0.00.2 ± 0.4
Arteriolar hyaline thickening (AH)0.3 ± 0.51.1 ± 1.1*0.0 ± 0.00.2 ± 0.6


By immunostaining, virtually all tubules in implantation biopsies expressed epithelial marker E-cadherin on the basolateral membrane (Figure 1A), whereas cytokeratin expression was present in the cytoplasm of TECs (Figure 1C). In contrast, in allograft TA/IF biopsies, E-cadherin expression was reduced and redistributed from the basolateral membrane to the cytoplasm (arrow) and apical membrane (arrow head) (Figure 1B) and the expression of cytokeratin was decreased (Figure 1D). The mesenchymal markers [vimentin (Figure 1E), S100A4 (Figure 1G) and α-SMA (Figure 1I)] and collagen synthesis marker [HSP-47 (Figure 1K)] showed no staining in the TECs in the implantation biopsies. New expression of vimentin (Figure 1F), S100A4 (Figure 1H) and HSP-47 (Figure 1L) was demonstrated in the cytoplasm of TECs in deteriorating and atrophic tubules of TA/IF biopsies, and in the interstitial fibroblasts. α-SMA expression was occasionally demonstrated in the TECs (star) and peritubular myofibroblasts in fibrotic areas (Figure 1J). Double immunostaining showed coexpression of cytokeratin and vimentin, S100A4 and HSP-47 in some TECs (Figure 1M–O), suggesting an intermediate stage of EMT. Some TECs, confirmed by cytokeratin expression were demonstrated in the interstitium adjacent to deteriorating or atrophic tubules, and occasionally such interstitial TECs expressed both cytokeratin and mesenchymal markers (Figure 1P).

Figure 1.

Immunohistochemistry staining shows tubular phenotypic changes in TA/IF. (A) E-cadherin expression in the basolateral membrane of tubular epithelial cells (TECs) in the implantation biopsies. (B) Reduced expression of E-cadherin and its redistribution from the basolateral membrane to the cytoplasm (arrow) and apical membrane (arrow head) in TA/IF. (C) Cytokeratin expression in the cytoplasm of TECs in the implantation biopsies. (D) Decreased expression of cytokeratin in TA/IF. Lack of staining in TECs for mesenchymal markers: vimentin (E), S100A4 (G) and α-smooth muscle actin, α-SMA (I), and collagen synthesis marker heat shock protein-47, HSP-47 (K) in the implantation biopsies. De novo expression of vimentin (F), S100A4 (H) and HSP-47 (L) in the cytoplasm of TECs in deteriorating and atrophic tubules, and in the interstitial fibroblasts. (J) Occasional expression of α-SMA in TECs (star) and peritubular myofibroblasts in fibrotic areas. Double immunostaining of cytokeratin (red) and vimentin (M), S100A4 (N), and HSP-47 (O) (brown), suggesting the intermediate stage of EMT.(P) Some TECs, confirmed by cytokeratin expression (red), are present (arrow) in the interstitium adjacent to deteriorating or atrophic tubules and express both cytokeratin and α-SMA (arrow head).

The quantitative analysis of these changes is summarized in Table 3. In the implantation biopsies from both the TA/IF and protocol transplants, virtually all TECs stained for the epithelial markers E-cadherin and cytokeratin and none expressed mesenchymal markers (vimentin and S100A4) or the collagen synthesis marker (HSP-47). TEC in renal biopsies from transplant patients with TA/IF showed significant loss of epithelial markers E-cadherin and cytokeratin compared to their implantation biopsies (p < 0.001).

Table 3.  Immunohistochemistry results in renal transplants with TA/IF group and control transplants group
Group MarkerPercent positive TECs in transplants with TA/IFPercent positive TECs in control transplants
Implantation biopsy
(N = 10)
Biopsy for dysfunction
(N = 10)

Implantation biopsy
(N = 10)
Protocol biopsy
(N = 10)

  1. *Statistically significant difference (p < 0.01) comparing implantation biopsies to corresponding groups of TA/IF or protocol biopsies.

  2. **Statistically significant difference (p < 0.01) comparing biopsies for dysfunction in the transplants with TA/IF group to the protocol biopsies in the control transplants group.

  3. NA = frozen sections for E-cadherin analysis were ‘not available’ for protocol biopsies.

E-cadherin99.4 ± 1.465.2 ± 5.8*−34.2 ± 4.499.2 ± 1.4NANA
Cytokeratin99.4 ± 1.359.4 ± 7.8*,**−40.0 ± 6.599.4 ± 1.190.0 ± 4.3*−9.4 ± 3.2
Vimentin 0.0 ± 0.030.6 ± 10.6**+30.6 ± 10.6 0.0 ± 0.0 6.3 ± 4.7*+6.3 ± 4.7
S100A4 0.0 ± 0.033.6 ± 9.4**+33.6 ± 9.4 0.0 ± 0.0 6.2 ± 4.9+6.2 ± 4.9
α-SMA 0.0 ± 0.0 1.5 ± 0.4** +1.5 ± 0.4 0.0 ± 0.0 0.1 ± 0.2+0.1 ± 0.2
HSP-47 0.0 ± 0.014.9 ± 4.4**+14.9 ± 4.4 0.0 ± 0.0 2.4 ± 2.2+2.4 ± 2.2

Some TECs in these biopsies also showed expression of mesenchymal markers vimentin and S100A4 and the collagen synthesis marker HSP-47, both in deteriorating and atrophic tubules. Relatively few tubules stained for α-SMA, but this was nevertheless increased in the TA/IF group.

The biopsies from control transplants with stable function showed minor loss of cytokeratin and some expression of vimentin compared to the corresponding implantation biopsy, but to a much lesser extent than the TA/IF biopsies.

Correlation of markers with clinical parameters

In the transplants with TA/IF, some correlations were apparent between the loss of epithelial marker loss (cytokeratin) and proteinuria, and between increased mesenchymal markers vimentin, S100A4, HSP-47 and elevated serum creatinine (Table 4). There was no correlation between donor age and the markers. In protocol biopsies from patients with stable renal function, no correlations were shown between markers and clinical parameters (serum creatinine, proteinuria and donor age) (Table 4B).

Table 4.  Correlation between markers and clinical parameters
A. In renal biopsies for dysfunction in the TA/IF group
Clinical parameter
R value (p-value)
R value (p-value)
R value (p-value)
R value (p-value)
R value (p-value)
R value (p-value)
  1. Abbreviations: Scr, serum creatinine at time of biopsy; 24 h Uprot, urine protein/24 h; NS, not significant.

SCr−0.619 (NS)−0.360 (NS)0.817 (< 0.01)0.640 (< 0.05)0.238 (NS)0.888 (< 0.01)
24 h Uprot−0.383 (NS)−0.863 (< 0.01)431 (NS)557 (NS)287 (NS)368 (NS)
Donor age−0.466 (NS)−0.604 (0.06)0.354 (NS)0.439 (NS)0.176 (NS)0.410 (NS)
B. In protocol biopsies with stable renal function
Clinical parameterR value (p value)R value (p value)R value (p value)R value (p value)R value (p value) 
SCr0.605 (NS)−0.202 (NS)−0.447 (NS)−0.366 (NS)−0.294 (NS) 
24h Uprot0.141 (NS)−0.023 (NS)0.138 (NS)0.406 (NS)0.221 (NS) 
Donor age−0.264 (NS)0.118 (NS)0.071 (NS)−0.017 (NS)−0.083 (NS) 

Comparison of patient rejection history with marker expression in transplants with TA/IF (Table 5) revealed that the biopsies with TA/IF with a history of rejection displayed less cytokeratin (p < 0.05) and more S100A4 (p < 0.05) and a trend toward more vimentin than those with no history of rejection.

Table 5.  Immunohistochemistry results in biopsies for dysfunction in the transplants with TA/IF group
Group MarkerImmunostaining of TECs (%) in Transplants with TA/IF
No rejection (N = 6)Rejection (N = 4)Change
  1. *Statistically significant difference in marker expression (p < 0.05) comparing biopsies with a history of no rejection to biopsies with a history of rejection.

E-cadherin66.7 ± 5.563.0 ± 6.3−3.7
Cytokeratin63.8 ± 5.652.9 ± 5.9*−10.9
Vimentin26.1 ± 10.737.4 ± 6.4+11.4
S100A428.3 ± 8.241.4 ± 3.8*+13.1
α-SMA1.5 ± 0.41.6 ± 0.4+0.1
HSP-4713.3 ± 4.517.3 ± 3.5+4.0


The present studies explored the emergence of EMT makers in human kidney transplants manifesting impaired function and TA/IF and biopsies for dysfunction. Immunohistochemistry studies of biopsies taken at the time of renal allograft dysfunction showed loss of epithelial markers (E-cadherin, cytokeratin), new expression of mesenchymal markers (vimentin, S100A4, α-SMA) and a collagen synthesis marker (HSP-47) by TECs in deteriorating and atrophic tubules. Moreover, double immunostaining using antibody against cytokeratins and mesenchymal markers or HSP-47 demonstrated coexpression of both epithelial and mesenchymal markers in some TECs, suggesting ongoing EMT. Furthermore, TECs (defined by cytokeratin expression) were demonstrated in the interstitium adjacent to deteriorating or atrophic tubules, and occasionally these TECs expressed both cytokeratin and mesenchymal markers. While there are many difficulties in establishing the existence and importance of EMT in vivo, the present findings are consistent with the migration of TECs across the tubular basement membrane to the renal interstitium, and suggest that TECs contribute to fibrogenesis during late deterioration in kidney transplants, through the processes of EMT.

The loss of epithelial markers is a crucial step in EMT (18). E-cadherin is a well-characterized adhesive junction protein and plays an essential role in maintaining the structural integrity and polarity of epithelial cells (11). The reduction in E-cadherin eliminates adherens junctions formation, disturbs cell polarity and may lead to increased proliferation and transdifferentiation (18). The decrease in E-cadherin and its redistribution from the basolateral membrane to the cytoplasm and the apical membrane observed in our study suggests the loss of polarity of TECs. Perturbed epithelial adhesion is followed by the loss of cytoskeletal integrity and the acquisition of mesenchymal protein expression (19). In this present study, the disappearance of intermediate filaments of the keratin type (cytokeratin) accompanies decrease in E-cadherin, de novo expression of mesenchymal markers and is in agreement with studies in human native renal biopsies (13) and cultured mouse tubular cells (20). Changes in the expression profiles are controlled by TGF-β1, a profibrotic cytokine. It downregulates E-cadherin and induces the expression of mesenchymal markers in the in vitro model of EMT (19,21).

New expression of mesenchymal markers in deteriorating and atrophic tubules indicates that TECs can undergo phenotypic change toward fibroblast-like cells during fibrogenesis in late deteriorating renal allografts. Vimentin is an intermediate filament protein expressed only in mesenchymal cells and has been used as a marker for EMT. In the early stage of development, TECs derived from the metanephrogenic mesenchyma express vimentin but vimentin is lost when cytokeratin expression is turned on in well-developed TECs (22). Thus, TECs in the normal adult kidney do not express vimentin. Studies in human native kidney biopsies demonstrated the expression of vimentin in TECs in fibrotic areas, suggesting a role of EMT in fibrogenesis (13).

S100A4 is emerging as a marker for EMT in studies on human tissue including kidney transplants (16), similar to oncological studies of tumor progression (23). S100A4 is the human homologue of mouse Fsp-1, a mesenchymal marker, the expression of which is associated with mesenchymal cell shape, cell motility and potential metastasis of cancer cells (24). Fsp-1 expression is induced early by TGF-β and epithelial growth factor in EMT, and is responsible for the transformation of TECs into fibroblasts (25). Implantation biopsies expressed no S100A4 in TECs, whereas low levels of expression were demonstrable in protocol biopsies from transplants with good function [compatible with previous reports of low levels of fibrosis in protocol biopsies (26)]. However, its expression increases significantly in biopsies with TA/IF. We also observed increased S100A4 in kidneys with history of rejection compared to transplants with TA/IF. In the study of acute renal allograft rejection, not only was S100A4 expressed in TECs that frequently associated with infiltrating CD8+ T cells, but intratubular T cells could induce TECs to transform into proliferative fibroblasts and migrate into interstitial areas (16,27). Since S100A4 expression in TEC is positively regulated either by TGF-β alone or by its T-cell membrane-bound form (16), it is plausible that T- cell-mediated rejection directly induces TEC to express S100A4, but it also seems likely that S100A4 is a consequence of other stimuli that cause EMT.

In this study, the finding that HSP-47 is expressed in the cytoplasm of deteriorating and atrophic tubules provides support for the contention that TECs play a role in fibrogenesis via EMT. New tubular expression of collagen-specific stress protein (HSP-47), a marker of collagen synthesis, also indicates the functional changes of TECs and a potential role of TECs in fibrogenesis. In a previous study, HSP-47 expression in interstitial cells (myofibroblasts) correlated with the degree of interstitial fibrosis in late deteriorating renal allografts (28). Collagen composition and assembly regulates epithelial-mesenchymal transition (21).

On the other hand, the expression of α-SMA (marker for myofibroblast) in TECs was uncommon in our study. In other studies, myofibroblasts were demonstrated adjacent to deteriorating and atrophic tubules (13). It is possible that the switch to vimentin and S100A4 expression occurs earlier than expression of α-SMA during EMT, and may explain the difficulty of using α-SMA as a marker to demonstrate EMT in TECs. The interstitial expression of α-SMA has been identified as a prognostic indicator for disease progression in both human and experimental glomerulonephritis (29,30), and could emerge as a useful marker in transplants. The degree to which α-SMA-positive cells in renal disorders are derived from TECs is not clear. While the present studies show only a few identifiable TECs expressing α-SMA, we cannot exclude the possibility that TEC-derived cells express α-SMA after they have lost their identifiable epithelial markers, once they have entered the interstitium.

While the population studied here is too small to define clinical correlations of EMT markers, the present observations serve as a training set indicating that useful correlations may emerge in a larger test set. Proteinuria correlated with cytokeratin loss and elevated creatinine with vimentin, S100A4 and HSP-47. Post-transplant nonimmune and immune stresses are associated with late deterioration of kidney transplants (6,31). Proteinuria can be directly tubulotoxic or stimulate the secretion of chemokines or cytokines by TECs leading to injury and later interstitial fibrosis (10). The damaged TECs may react by proliferation or hypertrophy in the early stages of injury and undergo either apoptosis or EMT in the later phases, leading to TA/IF (10,32). Long-term CNI effects are suggested by the hyalinosis of the afferent arterioles (33), although we lack direct evidence that CNI effects contribute to EMT in this study. CNIs can be directly tubulotoxic, but the predominant mechanism of renal injury is through chronic renal vasoconstriction (34). Again, damaged TECs may transdifferentiate into mesenchymal fibroblasts or undergo apoptosis (35) leading to TA/IF in renal allografts.

Because formal proof that epithelial cells can migrate and transdifferentiate into cells that contribute to interstitial fibrosis is lacking, we must regard the present findings as suggestive but not definitive. One problem is the relative paucity of transitional forms, but even then such cells (e.g. the double stained cells with cytokeratin and mesenchymal markers) could have alternative explanations. For example, cytokeratin positive interstitial cells might be derived from deteriorating tubules that have lost basement membrane, or from primitive mesenchymal cells that have acquired cytokeratin. What we need to prove is that TECs migrate across the basement membrane into the interstitium and contribute to fibrogenesis, and a method of quantifying the contribution of this phenomenon in total renal atrophy and fibrosis.

Despite these reservations, the importance of studying the EMT features associated with damaged kidney transplants is that it opens opportunities for intervention. The potential treatments to slow progression of EMT and TA/IF could include careful minimization of CNI, or reduction of proteinuria with ACE inhibitors or angiotensin II receptor antagonists. Preclinical studies of novel therapeutic strategies such as hepatocyte growth factor (36), or bone morphogenetic protein-7 (37) for blocking TGF-β/Smad signaling, the main pathway regulating EMT (38), could lead to clinical applications to slow or arrest nephron loss after renal injury. Given the large number of kidney transplants at risk, the present study provides encouragement for the exploration of these possibilities.


We thank Konrad S. Famulski, Ph.D. (University of Alberta, Canada) and Deborah E. James, Ph.D. (University of Alberta, Canada) for their critical reading of this manuscript. We are grateful for grant support from Genome Canada, the Canadian Institutes of Health Research, the Roche Organ Transplant Research Foundation, the Kidney Foundation of Canada and The Muttart Foundation.