Transplant Accommodation in Highly Sensitized Patients: A Potential Role for Bcl-xL and Alloantibody

Authors


Abstract

Transplantation of renal allografts into recipients with circulating anti-HLA antibodies results in hyperacute rejection. In some cases, however, antibodies return without causing harm; this phenomenon has been termed ‘accommodation’. We have investigated this process in human allotransplantation.

We removed anti-HLA antibodies by immunoadsorption in seven highly sensitized dialysis patients who subsequently underwent renal transplantation. Immunohistochemistry of renal biopsies for IgG and anti-apoptotic proteins was performed. We also developed a model of ‘accommodation’ using anti-HLA antibodies eluted from sensitized patients and incubated with human umbilical vein endothelial cells (HUVECs) at different concentrations. Their effect on HUVEC phenotype was then analysed.

Anti-donor antibody returned in 4/7 patients, without evidence of hyperacute rejection. Three out of four of these ‘accommodated’ grafts showed specific endothelial up-regulation of Bcl-xL and 2/2 tested positive for endothelial IgG deposition. HUVECs incubated with subsaturating concentrations of anti-HLA antibody showed increased expression of Bcl-xL, were rendered refractory to endothelial cell activation and became resistant to complement-mediated lysis. In contrast, HUVECs incubated with saturating concentrations underwent activation and expressed low levels of Bcl-xL.

In conclusion, endothelial Bcl-xL expression defines the accommodation process in human allografts and this phenotype may be initiated by exposure of endothelium to low concentrations of anti-donor HLA antibodies.

Introduction

Allografts transplanted into ABO-incompatible or HLA-sensitized individuals usually suffer an immediate and aggressive rejection response known as hyperacute rejection (HAR). The same process affects experimental xenografts transplanted between discordant species, and has been the biggest single hurdle to clinical xenotransplantation. In clinical practice, great efforts are taken to avoid HAR by ensuring that allografts and recipients are matched for blood groups, and by obtaining a pretransplant lymphocytotoxic cross-match to identify sensitized patients with antibodies specific for donor HLA antigens. As a result, HAR is now rarely found. However, circulating antibodies against donor HLA or other non-MHC endothelial antigens may also be responsible for a delayed form of acute humoral rejection (AHR). This manifests itself with more severe episodes of rejection, is characterized by deposition of the C4d complement component and is associated with an increased incidence of graft loss (1)

HAR and AHR can be prevented by depleting antigraft antibodies from the recipient by plasmapheresis or immunoadsorption prior to transplantation (2–7). In experimental models, rejection is usual upon the return of antigraft antibody (8,9) although some allogeneic and xenogeneic organs have been observed to continue functioning despite this, a phenomenon termed graft ‘accommodation’ (2,3,6,10,11).

Sensitized patients, with high levels of circulating anti-HLA antibodies, represent a significant proportion of potential transplant recipients (12) and as a group they wait longer than average for an allograft. Some highly sensitized patients can benefit from a programme of immunoadsorption and immunosuppression in order to eliminate temporarily the anti-HLA antibodies (3,6,13). However, therapy is expensive, time-consuming, and not without risk, in particular by exposing the patients to prolonged immunosuppression. It is consequently only offered to small numbers of younger patients. Safe and effective strategies to prevent HAR and AHR or promote graft accommodation would be of enormous benefit to highly sensitized patients, and the same techniques are likely to be useful in clinical xenotransplantation.

Recently, a reliable small animal model of accommodation has been described by different groups (11,14). The phenotype of these ‘accommodated’ grafts is very distinctive; there is no evidence of rejection and the endothelial cells lining the vessels of the grafts express the products of ‘survival genes’, such as Bcl-2, Bcl-xL, and other proteins such as haemoxygenase 1 (HO-1) (15). These changes have been postulated to have important functional consequences, as expression of these proteins in endothelial cells (ECs) has an anti-inflammatory effect, rendering them resistant to apoptosis and further activation (15,16). Although there are clinical reports of accommodation in transplanted allografts (6,17), there is as yet no way to identify those grafts that have undergone this process.

Furthermore, the physiological stimuli that initiate accommodation in vivo are not clear. There is evidence that circulating antigraft antibody is required for it to occur (10). However, in other studies, it appears that antigraft antibody depletion is necessary to allow accommodation. The hypothesis we have formulated to explain this apparent contradiction is that accommodation might only arise if endothelium is initially exposed to a low concentration of antigraft antibody. Evidence to support this hypothesis is accumulating from a xenogeneic model (18,19).

We have previously reported five highly sensitized patients who received kidney allografts without HAR after undergoing regular immunoadsorption to remove anti-HLA antibodies. Two patients were noted to have positive cross-matches in the early post-transplant period without adverse effects, indicating that their grafts had ‘accommodated’ (6). In this study we have investigated the possible mechanisms of renal transplant accommodation in an expanded cohort of such patients. We have also developed an in vitro model of accommodation using anti-HLA IgG and human umbilical vein endothelial cells (HUVECs) to explore whether low concentrations of allospecific antibodies can mediate endothelial cell phenotype changes consistent with accommodation.

Materials and Methods

Patients and transplantation

Seven highly sensitized patients (defined as those with panel reactive antibodies > 80%) were immunoadsorbed using a staphylococcal protein-A column, Citem 10 system (Excorim), with heparin and citrate anticoagulation, to achieve negative panel reactivity. Patients were immunosuppressed with prednisolone (starting dose 20 mg) and cyclophosphamide (2 mg/kg/d), for a maximum of 2 months, and then cyclophosphamide was converted to azathioprine 2 mg/kg/d if they were still being immunoadsorbed. Most patients underwent further immunoadsorption or plasmapheresis prior to transplantation. The lymphocytotoxic cross-match using donor spleen or lymph node cells was negative immediately prior to transplantation in all cases. All of the patients were transplanted with ABO-compatible, HLA-mismatched cadaveric kidneys (median HLA–antigen mismatch of 5, range 5–6).

Induction immunosuppression consisted of antilymphocyte antibody preparations (ATG or ALG) as well as intravenous azathioprine 2 mg/kg, cyclosporin A (1 mg/kg 4-hourly for the first 24–48 h), and methylprednisolone 1 g. Maintenance daily immunosuppression was initially with oral prednisolone 20 mg, azathioprine 1 mg/kg and cyclosporin A (Sandimmune 10 mg/kg in five patients, Neoral 8 mg/kg in two patients). Subsequent dosing was adjusted based on whole-blood cyclosporin levels. Delayed graft function was treated at the discretion of the attending physician, with omission or reduction of cyclosporin A, and/or the introduction of a short course of oral cyclophosphamide.

Immunohistochemistry

Renal biopsies were performed, if clinically indicated, for delayed graft function or deteriorating renal function. No protocol biopsies were performed in these patients. Samples were either snap frozen in isopentane, or fixed in buffered formalin for 24 h and then processed for histological examination. Cryostat sections were stained for immunoglobulins (IgG, IgA, IgM) and complement component C3 (Dako, UK) in two patients in whom anti-HLA antibody returned. Four-micrometre sections of paraffin-embedded tissues were cut and attached to glass slides coated with Poly L lysine (Histolab). The paraffin sections were then de-waxed, hydrated and washed in phosphate-buffered saline (PBS), pH 7.6. Paraffin-embedded sections were stained by the avidin-biotin-complex (ABC) immunoperoxidase technique, with the following primary antibodies: anti-Bcl-xL (Zymed, USA), anti-Bcl2 (Dako, UK) and anti-HO-1 (a gift from Affiniti Laboratories, UK). Second layer antibodies were swine antirabbit immunoglobulin (Dako) and rabbit antimouse immunoglobulin (Dako), developed with ABC (Dako). Antigen retrieval with microwaving in citrate buffer was employed. Briefly the sections were incubated with primary antibodies at 4°C overnight, followed by incubation with secondary layer antibodies, for 1 h, at room temperature and finally developed with ABC, counterstained with haematoxylin.

Microcytotoxicity assay

The cross-matches were performed by a standard assay using patient's serum and donor spleen or lymph node cells. Briefly, the serum was incubated with donor cells for 30 min and then whole rabbit complement (Saxon Europe, Ltd) was added for a further hour. Donor cell death was confirmed with propidium iodide and acridine orange (Sigma, UK), and percentage of cell death assessed using an inverted fluorescence microscope. The cross-matches were also performed in the presence of dithiothreitol (DTT) to exclude IgM antibodies. Antibody specificity was tested pre- and post-transplantation in one patient by QUICKID (Quest Biomedical, UK).

Endothelial cell culture

Primary cultures and immortalized HUVECs were kindly provided by Dr G. Cockerill and Dr J. Mason, Department of Cardiovascular Medicine, and Dr F. Marreli-Berg, Department of Immunology, ICSM, Hammersmith Hospital. These were grown on 2% bovine gelatin-coated (Sigma, UK) tissue culture flasks (Nunc, Denmark) in M199 (Life Technologies Ltd, Paisley, Scotland) supplemented with 20% FCS (Globepharm, UK), 2 mm l-glutamine, 50 µg/mL penicillin and streptomycin (Life Technologies, UK), 10 µg/mL endothelial cell growth supplement (Sigma, UK), and 12 U/mL heparin (CP Pharmaceuticals Ltd, Wrexham, UK). Cell lines were HLA-typed using conventional DNA techniques. Following incubation with alloantibody, the expression of ICAM-1 was assessed, by staining with anti-ICAM-1 monoclonal antibody (ATCC, hybridoma 6.5B5), and isotype control antibody. Flow cytometry was performed on an EPICS XL flow cytometer (Coulter, UK).

Human alloantibody

Eluates from the immunoadsorption columns were collected from patient 7, and neutralized with Tris buffer to a pH of 7.3. Eluates were then dialyzed against PBS and stored at 4°C. Alloantibody was also obtained from other highly sensitized dialysis patients and control nonsensitized dialysis patients lacking any anti-HLA antibodies. These were purified on a protein A column (Pharmacia, Uppsala, Sweden) and treated as above. The alloantibody preparations were endotoxin-free and possessed a broad spectrum of anti-HLA antibodies, reacting with > 80% of cells in a panel of 40 donors representing different HLA antigens.

Western blotting

Cells were trypsinized and counted. Equal cell numbers were lysed and prepared in a standard way. Equal quantities of lysate in running buffer were loaded onto 12.5% SDS electrophoretic gels, run at 100 V for 2 h and transferred to nitrocellulose membranes (Sartorius, Germany) overnight. Membranes were blocked with 5% Marvel, in PBS with 0.01% Tween, for 1 h. Primary and secondary antibodies were made up in 5% Marvel and 0.01% Tween and exposed to the membrane for 1 h at room temperature. Washes were performed with PBS/0.01% Tween (3 × 10 min). The primary antibody used was anti-Bcl-xL (Zymed), and secondary antibody was HRP-conjugated rabbit antimouse IgG (Amersham, UK). Development was with ECL-plus (Amersham) and exposed on Kodak photographic film (supplied by Sigma).

Complement-mediated lysis

Endothelial cells were treated with alloantibody or control human immunoglobulin (BPL, UK) at equivalent low and high concentrations for variable periods of time. At the end of the assay, the cells were washed, trypsinized and plated out in a 96-well plate in triplicate. Cytotoxicity was assessed by LDH release using the CytoTox Assay (Promega) modified according to the protocol of Sepp et al. (20) with rabbit complement (Saxon Europe, Ltd).

Results

Immunoadsorption and transplantation

Seven highly sensitized patients, with panel reactive antibodies of > 80%, were immunoadsorbed. Five of these have been discussed in detail in a previous report, in which the practicality of such an approach for dealing with highly sensitized patients was described (6).

All patients had undergone previous renal transplants and received blood transfusions, and three had been pregnant. All seven patients underwent cadaveric renal transplantation. The details of the recipients and the donor organs are shown in Table 1. All the patients had positive cross-matches on historical sera, taken prior to the transplant. However, cross-matches at the time of transplantation were negative.

Table 1. : Summary of patients
PatientAge (years)SexRenal diseasePrevious graftsPregnanciesYears sensitized (PRA > 80%)
133FReflux nephropathy205
243FGoodpasture's disease113
338FUnknown124
435MChronic glomerulonephritis1n/a5
539MDiabetes1n/a1.75
633FUnknown126
738FHypertension1011

In four of the patients, the cross-match became positive following transplantation, as assessed by the microcytotoxicity assay in three and flow cytometry in one. The time course showing the return of anti-donor antibodies is shown in Figure 1. In one patient, specificities of the anti-HLA antibodies were tested pre- and post-transplantation and found to be identical. One graft never functioned and was removed after 2 months, the nephrectomy specimen showing no signs of hyperacute rejection, only ischaemia and cortical necrosis. Of the other six, two grafts functioned well, but were lost to chronic rejection at 44 and 72 months; one of these patients had antibody return. The other four grafts are still functioning, with follow-up of between 4 months and 7 years, and three of these patients had antibody return. The lowest serum creatinine achieved by each of the patients is shown along with their outcome in Table 2.

Figure 1.

Return of anti-donor antibodies in patients 2, 4, 6 and 7. Direct cross-matching using microcytotoxicity (patients 2, 4 and 6) or flow cytometry (patient 7) techniques was carried out using historical and current sera. The percentages of donor cell killing or donor cell binding are plotted against the time sera was taken, in relation to the transplant.

Table 2. : Donor details/transplant outcomes
PatientHLA
mismatch
RejectionLowest serum
creatinine (μmol/L)
Graft survival
(months)
Cross-match
(days to positivity)
  • ACR, acute cellular rejection; HD, haemodialysis.

  • a

    Still functioning, latest follow-up.

15/6ACR13572
25/6ACR160442
35/6NoHD2
46/6ACR10090*10
56/6No16088*
66/6ACR13033*1
75/6Glomerulopathy2197a6

Immunohistochemistry of allograft biopsies

All seven patients underwent renal biopsies, although at different time-points. Biopsies were performed every 7 d while there was delayed graft function, or in response to an unexplained increase in serum creatinine. Control renal biopsies were from a trauma victim's nephrectomy specimen (normal kidney), pretransplant biopsy, and from patients suffering from acute cellular rejection, acute tubular necrosis, WHO class IV lupus nephritis, IgA nephropathy, thrombotic microangiopathy and postinfectious glomerulonephritis. Immunofluorescence was performed, where possible, on frozen sections. IgG and C3 were positive in two renal biopsies from two patients in whom antibody returned; positive staining coincided with the return of anti-donor antibodies and confirmed that deposition of antigraft antibodies occurred on the donor glomerular endothelium (Table 3). No evidence of endothelial IgG or C3 was found in any other biopsy specimen (data not shown).

Table 3. : Bcl-xL staining on renal biopsy specimens, from patients and disease controls
 Anti-donor antibody
(days)
Glomerular
capillaries
Peritubular
capillaries
TubulesIgG
deposition
  1. Pre-Tx, on-table pretransplantation biopsy; ATN, acute tubular necrosis; ACR, acute cellular rejection; SLE, systemic lupus erythematosus; IgA, IgA nephropathy; TMA, thrombotic microangiopathy; APIN, acute postinfectious glomerulonephritis.

Patients
1+++–-
2+(2)+++++++ND
3NDNDNDNDND
4+(10)++ND
5++ND
6+(1)+++++
7+(6)++++++
Controls
Pre-Tx+ 
ATN++ 
ACR++ 
SLE+ 
IgA++ 
TMA++ 
APIGN++ 
Normal++ 

Bcl-xL was weakly expressed on the distal tubular epithelial cells and collecting ducts in normal kidney (Figure 2A). In the biopsies of nephritic kidneys and those from transplanted kidneys showing acute tubular necrosis (ATN) or cellular rejection, this expression was increased and extended to the proximal tubules. The unique feature that was noted in three out of four of the biopsies from the patients in whom anti-donor antibody had returned was endothelial staining for Bcl-xL in glomerular capillary loops and peritubular capillaries, in addition to the marked up-regulation of tubular staining (Figure 2B–D).

Figure 2.

Bcl-xL staining. A. Photomicrograph of a section taken from a normal kidney stained for Bcl-xL. There is some staining of the distal tubular cells but no glomerular or endothelial staining is seen (×400). B. Photomicrograph of a section from the renal transplant biopsy taken from patient 2, stained for Bcl-xL. There is marked staining in both proximal and distal tubular cells. There is also prominent glomerular staining in the capillary loops (×400). C. Photomicrograph of the same section as that in B at higher power. This clearly shows the glomerular endothelial cell staining of Bcl-xL (arrow) (×1000). D. Photomicrograph of the same section as that in B at higher power showing the peritubular capillary endothelial cell staining of Bcl-xL (arrow) (×1000).

No glomerular or peritubular capillary endothelial staining was seen in any other control sections, nor in biopsy specimens available from the patients in whom antibody did not return (Table 3).

Neither Bcl2 nor HO-1 expression was specifically up-regulated in the accommodated grafts (data not shown), in contrast to the descriptions of accommodation in xenogeneic models (11).

In vitro model of accommodation

Alloantibody was obtained from patient 7, and from three other highly sensitized dialysis patients. Control IgG was obtained from two nonsensitized patients lacking anti-HLA antibodies. These were purified as previously described on a protein A column. A further control IgG preparation, human normal globulin (HNG), was obtained from Bio Products Laboratory (Herts, UK). Presence or absence of the antibodies for HLA antigens was confirmed by staining with HLA-transfected mouse fibroblasts and performing lymphocyte microcytotoxicity assays (data not shown). Alloantibody binding to both immortalized and primary cultures of HUVECs was assessed by flow cytometry, to establish concentrations that gave saturating (1/5–1/7 dilution) or subsaturating (1/100–1/210 dilution) binding. These two concentrations were used for all further experiments. One immortalized HUVEC line (C11STH) and two primary culture HUVECs were used in the experiments (F4 and JM1). All different HUVEC lines tested bound the anti-HLA alloantibody preparations from highly sensitized patients to some degree, presumably due to extensive cross-reactivity of the antibodies. To confirm that the EC phenotype changes were mediated by anti-HLA antibodies, we used IgG preparations from dialysis patients who did not have any anti-HLA activity, based on microcytotoxicity panel testing, as well as HNG, which was also confirmed to contain no significant titre of anti-HLA antibodies.

Effect of alloantibody on Bcl-xL expression

Following incubation with either saturating or subsaturating alloantibody for up to 5 d, EC lysates were prepared and analysed by Western blotting using an antibody specific for human Bcl-xL. IgG preparations from four different sensitized patients and controls were screened in this way. Experiments were carried out at least twice, on both immortalized and primary culture HUVECs. Alloantibody preparations from sensitized patients were associated with specific changes in Bcl-xL expression. Two different patterns of Bcl-xL expression were found in these experiments, as shown in Figure 3. When lysates from control HUVECs were found to have little Bcl-xL, expression was clearly up-regulated in those cells incubated with low concentrations of alloantibody (Figure 3A). No expression was induced in cells treated with saturating concentrations of antibody. When control HUVECs were found to have high levels of Bcl-xL expression (Figure 3B), incubation with high concentrations of alloantibody resulted in loss of Bcl-xL expression. In contrast, incubation with low concentrations of alloantibody was associated with maintenance of Bcl-xL expression. Incubation with HNG or control IgG from nonsensitized patients resulted in no specific changes in Bcl-xL expression at any time-points examined (Figure 3C). The results from all these experiments are summarized in Table 4.

Figure 3.

Western blots for Bcl-xL on human umbilical vein endothelial cells (HUVECs) treated with alloantibody from highly sensitized (A, B) and nonsensitized (C ) dialysis patients. HUVECs incubated with saturating or subsaturating alloantibody were lysed and probed for Bcl-xL. In highly sensitized patients (A, B), saturating concentrations of alloantibody down-regulated the expression of Bcl-xL. Subsaturating levels maintained or up-regulated Bcl-xL levels compared with untreated cells. In nonsensitized patients (C) no changes in Bcl-xL expression were found. A similar pattern to that in C was found in cells treated with human normal globulin (HNG) (not shown). c, control, untreated cells; SAT, saturating antibody treated cells; SS, subsaturating antibody-treated cells.

Table 4. : Summary of changes in human umbilical vein endothelial cell (HUVEC) Bcl-xL expression following incubation with different IgG preparations
Patient IgGBcl-xL expression compared with control
SubsaturatingSaturating
  1. +, up-regulated; –, down-regulated; 0, unchanged.

  2. (Results are representative of experiments performed twice on both primary and immortalised HUVECs). The different effect on Bcl-xL expression between subsaturating and saturating IgG preparations from sensitized patients is significant (p < 0.01) using Mann–Whitney U-test. HNG, human normal globulin.

Highly sensitized
1+0
2+0
30
4+0
Non-sensitized
100
200
HNG00

Effect of alloantibody on ICAM-1 expression

Saturating and subsaturating concentrations of alloantibody were added to HUVEC cultures and incubated for up to 5 d. Cells were then washed and stained for expression of ICAM-1. Saturating doses of alloantibody were found to up-regulate ICAM-1 expression, whereas subsaturating doses did not (Figure 4). Maximum up-regulation was seen after 24 h incubation. After 4 or 5 d, ICAM-1 expression had returned to control levels.

Figure 4.

ICAM-1 expression on human umbilical vein endothelial cells (HUVECs) following incubation with alloantibody. Graph showing the changes in ICAM-1 expression on HUVECs following treatment with saturating or subsaturating concentrations of alloantibody over 5 d. Control cells were not treated with antibody.

Re-stimulation experiments were then performed. HUVECs were incubated with saturating or subsaturating concentrations of alloantibody. After 4 d, levels of ICAM-1 expression on those cells incubated with saturating concentrations had fallen near to those on cells incubated with subsaturating concentrations (Figure 5). These cells were then re-stimulated with saturating concentrations of alloantibody for 24 h. HUVECs previously incubated with saturating concentrations of alloantibody up-regulated expression of ICAM-1 again, whereas those previously incubated with subsaturating concentrations failed to increase ICAM-1 expression after 24 h (Figure 5).

Figure 5.

Changes in ICAM-1 expression following re-stimulation with alloantibody. Expression of ICAM-1 on human umbilical vein endothelial cells (HUVECs) pretreated with either saturating or subsaturating alloantibody concentrations after 4 d and following re-stimulation with saturating alloantibody for a further 24 h. Pretreatment with saturating alloantibody results in further ICAM-1 expression following re-stimulation. However, subsaturating alloantibody pretreatment prevents ICAM-1 up-regulation following re-stimulation.

Effect of alloantibody on complement-mediated lysis

The sensitivity of ECs to complement-mediated lysis was assessed. Cells were incubated with either subsaturating or saturating concentrations of antibody, before being exposed to the anticlass I monoclonal antibody W6/32 and rabbit complement. Prior incubation with alloantibody did not inhibit the maximal binding of W6/32. Cells treated with subsaturating concentrations of alloantibody for 4 d were resistant to complement-mediated lysis, whereas those incubated with saturating concentrations were as sensitive as controls (Figure 6). Control IgGs produced no such change, suggesting that antibody binding to HLA molecules was required for this effect.

Figure 6.

A. Complement-mediated lysis of human umbilical vein endothelial cells (HUVECs) treated with different concentrations of alloantibody. HUVECs treated with subsaturating levels of alloantibody were significantly less susceptible to complement-mediated lysis (p < 0.05) than those treated with saturating concentrations, and were also less susceptible than untreated control cells. B. Complement-mediated lysis of HUVECs treated with control human normal globulin (HNG). Concentrations of human immunoglobulin comparable to the alloantibody were used but did not alter HUVEC susceptibility to complement-mediated lysis, compared with untreated control cells.

Discussion

Renal transplantation in the presence of anti-donor cytotoxic antibodies is associated with a high rate of immediate graft failure due to HAR (21,22). That antibodies are responsible for this phenomenon is clear from early experiments in animal allograft models in which transferred sensitized serum was capable of inducing accelerated graft loss (23). HAR is now rarely seen because of routine pretransplant cross-matching.

Highly sensitized patients, who have high titres of anti-HLA antibodies, are problematic to transplant due to the difficulty of locating a donor organ that does not provoke a positive cross-match. Furthermore, anti-HLA antibodies may provoke a severe form of acute humoral rejection, which is associated with poor graft outcome and characterized by C4d deposition in the peritubular capillaries (1,24). However, depletion of these antibodies prior to transplantation using plasmapheresis or immunoadsorption can create a window of time in which a transplant may be performed without subsequent HAR (2–7). Recurrence of antibody generally results in HAR (8,9) or AHR (1). There are, however, a number of examples, from both animal studies and human transplantation of antigraft antibodies returning without causing rejection (2,3,6,10,11). This phenomenon has been termed graft ‘accommodation’ (25,26). The finding was originally described in ABO-incompatible grafts (2,17) and subsequently in allo-sensitized patients' allografts (3,6). Little insight into the mechanisms underlying accommodation was gained until recently, with the resurgence of experimental xenografting, in which HAR was proving to be the largest single hurdle to overcome.

We have successfully transplanted seven highly sensitized patients, following immunoadsorption, to remove their anti-HLA antibodies. In one patient (patient 3) there was primary nonfunction, and the graft was removed after 2 months. Although there was no evidence of hyperacute rejection on the nephrectomy specimen, it is not possible to exclude that this represented an example of antibody-mediated rejection, since such cases may present with acute tubular necrosis and primary nonfunction (24). In four of the transplanted patients, there was a return of the anti-donor antibodies, demonstrated by a positive direct cross-match; two of these grafts were examined for antibody deposition and were found to have IgG as well as C3 complement component, deposited on the renal transplant endothelium. Despite this, all four grafts functioned well (and the creatinines the patients achieved are shown in Table 1).

This represents the first description of alloantibody deposition on transplant endothelium without evidence of vascular rejection. Staining of the transplant biopsies showed specific up-regulation of Bcl-xL in three of the four grafts in which there was a return of anti-donor antibodies. None of the control biopsies from normal or diseased native kidneys, or from transplanted kidneys, showed this pattern. The close relationship between Bcl-xL up-regulation and accommodation suggests that the latter is an active process, and not merely a passive effect of antibody depletion at the time of ischaemia-reperfusion injury. If recovery from the ischaemia-reperfusion in the absence of antibody were the mechanism by which accommodation occurred, we would have expected up-regulation of Bcl-xL in all cases of antibody depletion, which we did not observe. Thus Bcl-xL up-regulation in the microvascular endothelial cells appears to be a specific marker of accommodation in human renal allografts. This is in complete accordance with descriptions of an accommodated phenotype in animal models of xenotransplantation (11), in which accommodated grafts show endothelial cell up-regulation of anti-apoptotic proteins Bcl-xL, Bcl-2, A-20 and HO-1. These changes correlated with decreased EC activation in vitro. A recent report suggests that EC up-regulation of members of the Bcl family may confer a protective anti-inflammatory effect by inhibiting NF-κB (27). Thus endothelial cell activation, as measured by expression of adhesion molecules, is intimately linked to the level of expression of the Bcl family of proteins.

Previous studies in ABO-incompatible and highly sensitized patients had suggested that lowering the antigraft antibody titre prior to transplantation, followed by maintenance of low levels for some days following transplantation, would allow the grafts to survive even if antibody titres subsequently rose. In xenogeneic model systems, similar results were achieved by depleting complement or antibody (10) or by neutralizing the antigraft antibodies with soluble antigen (28). However, in one xenogeneic model, if antibody was completely absent at the time of transplantation, and suppressed following transplantation by cyclophosphamide, accommodation was never achieved; instead, return of antibody following cessation of cyclophosphamide was always accompanied by graft rejection (10). This suggests that some antigraft antibody is required at the time of transplantation for accommodation to occur and is the basis of our hypothesis that low concentrations of antigraft antibodies may mediate accommodation. We have tested this previously in a xenogeneic system (18,19). Using a model with cultured HUVECs and anti-HLA IgG antibodies isolated from highly sensitized patients, we have here presented evidence to support the hypothesis in a completely allogeneic system. HUVECs incubated with alloantibody from sensitized patients for 5 d showed a sustained or increased level of expression of Bcl-xL, became resistant to activation, as measured by up-regulation of ICAM-1, and acquired resistance to complement-mediated lysis, consistent with the notion that ‘accommodated’ endothelial cells are protected from the effects of activated complement components.

Most importantly, all of these changes were only seen when HUVECs were incubated with subsaturating concentrations of alloantibody, whereas incubation with saturating concentrations mediated EC activation. An apparent anomaly in our results is due to the fact that, in some experiments, resting HUVECs expressed Bcl-xL, whereas in others (the majority) resting HUVECs were Bcl-xL-negative (see Figure 3). We have no explanation for this. The pattern of Bcl-xL expression after incubation with subsaturating concentrations of alloantibody was clear in those experiments where resting Bcl-xL expression was absent, but more difficult to interpret when resting Bcl-xL was present. However, it is clear from all our experiments that levels of Bcl-xL expression were greater in cells incubated with subsaturating compared with saturating alloantibody concentrations (see Table 4).

Others have achieved similar changes in ECs that render the cells resistant to subsequent injury. In many instances, treatment with toxic substances at low doses or for prolonged periods of time induces a state of resistance to subsequent higher doses of toxins. Nath et al. (29) showed that exposure of rat kidneys to high levels of haemoglobin induces injury, but low doses cause HO-1 up-regulation and confer resistance to further injurious stimulus. Similarly, we have previously shown that porcine ECs treated with low doses of human immunoglobulin, containing xenoreactive antibodies, are protected in a time-dependent fashion from complement-mediated lysis and activation, whereas high doses of antibody conferred no such protection (18). In all these cases there was a shift from an activated and pro-inflammatory phenotype to a protected one.

Our results with saturating concentrations of alloantibody were predictable from previous studies; others have shown that anti-HLA antibody can induce EC activation, as judged by ICAM-1 up-regulation (30), and that prior up-regulation of the endothelial proteins Bcl-xL and HO-1 confers protection from alloantibody-induced activation (31). The signalling pathways that mediate such events are known to involve phosphorylation of intracellular tyrosine residues (32). We have confirmed altered phosphotyrosine patterns in HUVECs incubated with saturating concentrations of alloantibody (data not shown), but found that such changes do not occur after incubation with subsaturating concentrations. This suggests that alternative signalling pathways must be involved in the up-regulation of Bcl-xL after ligation of HLA-class I by subsaturating concentrations of alloantibody. These ongoing studies will form the focus of a future report.

Based on our findings and studies in animal models (11,31) we speculate that, in vivo, the specific up-regulation of Bcl-xL in allograft microvasculature may be crucially central to the development of the accommodated phenotype, by conferring resistance to subsequent antibody-mediated injury. Protection from apoptosis and cellular activation by Bcl-xL expression may also render the ECs a less favourable surface for thrombosis, since healthy ECs present a less thrombogenic surface than apoptotic cells (33), and resting ECs bind activated platelets less avidly than activated ECs (34).

In summary, this is the first description of a specific ‘accommodated’ phenotype in transplanted human renal allografts. Our findings are consistent with the hypothesis that accommodation is an active process induced by anti-donor antibodies. Specifically, we postulate that the endothelium of renal allografts, transplanted into patients who have undergone plasma exchange or immunoadsorption immediately prior to transplantation, will be initially exposed to low titres of anti-HLA antibodies, which will in turn initiate a series of protective changes in the endothelium which manifest as accommodation. Central to this process is the specific up-regulation of Bcl-xL in the renal microvasculature. We envisage that these observations will provide the basis for further study and that understanding the molecular mechanisms that mediate accommodation may allow the development of specific therapies to promote accommodation in allografts prior to transplantation.

Acknowledgments

ADS is an MRC Training Fellow. We are grateful to the staff in the tissue-typing laboratory, Hammersmith Hospital, for carrying out the microcytotoxicity assays and the HUVEC HLA-typing.

Ancillary