Preformed class II donor-specific antibodies are associated with an increased risk of early rejection after liver transplantation

Authors


Address reprint requests to Jacqueline G. O'Leary, M.D., M.P.H., Annette C. and Harold C. Simmons Transplant Institute, Baylor University Medical Center, 3410 Worth Street, Suite 860, Dallas, TX 75246. Telephone: 214-820-8500; FAX: 214-820-0993; E-mail: jacquelo@baylorhealth.edu

Abstract

Preformed donor-specific human leukocyte antigen antibodies (DSAs) are considered a contraindication to the transplantation of most solid organs other than the liver. Conflicting data currently exist on the importance of preformed DSAs in rejection and patient survival after liver transplantation (LT). To evaluate preformed DSAs in LT, we retrospectively analyzed prospectively collected samples from all adult recipients of primary LT without another organ from January 1, 2000 to May 31, 2009 with a pre-LT sample available (95.8% of the patients). Fourteen percent of the patients had preformed class I and/or II DSAs with a mean fluorescence intensity (MFI) ≥ 5000. Preformed class I DSAs with an MFI ≥ 5000 remained persistent in only 5% of patients and were not associated with rejection. Preformed class II DSAs with an MFI of 5000 to 10,000 remained persistent in 23% of patients, and this rate increased to 33% for patients whose MFI was ≥10,000 (P < 0.001). Preformed class II DSAs in multivariable Cox proportional hazards modeling were associated with an increased risk of early rejection [hazard ratio (HR) = 1.58; p = 0.004]. In addition, multivariate modeling showed that in comparison with no DSAs (MFI < 1000), preformed class I and/or II DSAs with an MFI ≥ 5000 were independently correlated with the risk of death (HR = 1.51; p = 0.02). Liver Transpl 19:973–980, 2013. © 2013 AASLD.

Abbreviations
AIH

autoimmune hepatitis

CNI

calcineurin inhibitor

DSA

donor-specific human leukocyte antigen antibody

HLA

human leukocyte antigen

HR

hazard ratio

IgG

immunoglobulin G

LT

liver transplantation

MELD

Model for End-Stage Liver Disease

MFI

mean fluorescence intensity

NASH

nonalcoholic steatohepatitis

NS

not significant

PBC

primary biliary cirrhosis

PSC

primary sclerosing cholangitis

The role of donor-specific human leukocyte antigen antibodies (DSAs) in liver transplantation (LT) has long been debated.[1-5] Early analyses revealed the liver's ability to absorb preformed DSAs; some were absorbed without consequence, and others resulted in preservation injury and/or vanishing bile duct syndrome.[4-8] Early clinical data based on cytotoxic crossmatching were united in leading to the conclusion of an absence of hyperacute rejection (with rare exception), but they were divided with respect to conclusions about short-term and longer term outcomes.[1, 9, 10] However, in the 1980s, graft failure rates were high and limited the analysis of long-term outcomes. Stratification by crossmatch strength in the early 1990s first elucidated increased 1-month graft failure rates in crossmatch-positive patients (29%) versus low-level crossmatch-positive and crossmatch-negative patients (16%).[2, 3, 9] In fact, “rejection was the most common cause of primary liver allograft failure” in patients with a positive crossmatch.[2]

Recent technological improvements have resulted in mounting evidence that antibody-mediated rejection occurs in LT patients; single-antigen bead analyses have found correlations between preformed DSAs and poor outcomes.[11-19] However, DSAs have not been evaluated as part of multivariable analyses, and an association of DSAs with impaired graft survival has not been consistently found in all studies.[20, 21] As a result, the precise association between DSAs and liver allograft rejection and loss is not fully understood.

Because of these disparate findings, we hypothesized that not all DSAs cause clinically meaningful pathology in liver allografts. Therefore, we evaluated class I and II DSAs separately and set a higher mean fluorescence intensity (MFI) cutoff to define positive DSAs. Although the persistence of DSAs after LT is likely important, by that time, the matching of donors to recipients has already occurred.[5, 6, 16, 22] Instead, we focused on preformed DSAs in an effort to further understand potential consequences of LT in the presence of preformed DSAs.

PATIENTS AND METHODS

Since 1985, sera and cells have been collected from all LT recipients at the Annette C. and Harold C. Simmons Transplant Institute before transplantation and at years 1, 2, 5, 10, 15, 20, and so forth after transplantation. At the same time, periodic protocol liver biopsy has been performed. In combination with this, clinical, laboratory, and pathology information is prospectively collected and locked into the Liver Transplant Research Database System. All patient deaths are reviewed and adjudicated by a single reviewer before they are entered into the Liver Transplant Research Database System. In addition, all patients with suspected rejection undergo liver biopsy (without serum sample acquisition) before the institution of therapy, and follow-up biopsy is always performed after treatment to verify resolution or the need for further therapy. For the purposes of this study, biopsy data were not re-reviewed.

All adult patients (age > 18 years) who underwent LT from January 1, 2000 to May 31, 2009 and had a pretransplant sample available for analysis were included in this study; after the study protocol was approved by the institutional review board retransplant and multiorgan transplant recipients were excluded. Thirteen hundred twenty-six adult patients underwent primary LT without another organ during this time period. Pretransplant samples from 1270 patients (95.8%) were anonymized and analyzed for DSAs by the Terasaki Foundation Laboratory. Additionally, 1109 patients (87.3%) had a single post-LT sample analyzed: 161 (12.7%) had no post-LT sample, 91 (7.2%) were collected 0 to 10 months after LT, 910 (71.7%) were collected >10 to 16 months after LT, and 108 (8.5%) were collected >16 months after LT.

Induction therapy, mostly with daclizumab, was given per protocol: patients with preoperative West Haven grade III or IV encephalopathy, pretransplant or posttransplant renal failure, or postoperative central nervous system toxicity from calcineurin inhibitors (CNIs) were given induction.

Human Leukocyte Antigen (HLA) Tissue Typing

All patients and donors were typed for HLA-A, HLA-B, HLA-DRB1, HLA-DRB345, and HLA-DQ with commercially available serological typing trays or molecular methods (Terasaki HLA tissue typing trays and Micro SSP or LabType SSO, respectively; One Lambda, Inc., Canoga Park, CA). Before 1998, serology was used exclusively. Since 1998, all donor class I and II HLA typing and patient class II typing have been performed with molecular methods, whereas patient class I typing is performed with serology.

HLA Immunoglobulin G (IgG) Antibody Determination

The detection of anti-HLA IgG antibodies was performed with LABScreen single-antigen class I (lot 7) and II (lot 9) beads (One Lambda). The assay was performed according to the manufacturer's protocol. A normalized MFI value ≥ 5000 was considered positive.

Statistical Analysis

Patient characteristics for the entire cohort are reported with median values and interquartile ranges (when appropriate) in Table 1. A Kaplan-Meier analysis was performed for the time to first biopsy-proven rejection and patient survival. The highest MFI of a single DSA per class was analyzed.

Table 1. Characteristics of 1270 Consecutive Primary LT Patients With Pretransplant Samples Available for Analysis Divided Into Groups by the Presence of Different Types of Preformed DSAs
 No DSA (n = 1086)Class I (n = 84)Class II (n = 50)Classes I and II (n = 50)P Value
  1. NOTE: Bolded values are significant.

  2. a

    The data are presented as medians and interquartile ranges.

  3. b

    Immunosuppression used 3 months after LT.

  4. c

    Combination induction therapy was given to 33 patients with either thymoglobulin and daclizumab (n = 28), OKT3 and daclizumab (n = 4), or OKT3 and thymoglobulin (n = 1).

Age (years)a52 (47-57)49 (44-55)53 (45-60)51 (48-58)0.06
MELD scorea16 (12-21)18 (14-23)16 (11-21)19 (15-22)0.02
Cold ischemia time (hours)a7.8 (6.1-10)8.1 (6.4-9.9)6.6 (5.4-10)7.6 (6.5-10)0.30
Warm ischemia time (minutes)a50 (43-60)49 (42-55)50 (40-56)49 (44-62)0.23
Male sex (%)71423410<0.001
Race (%)     
African American8102140.13
Hispanic15251816 
Other3146 
White74647664 
Liver disease (%)     
Hepatitis C465256400.02
PSC/PBC/AIH13192232
Alcohol15768
NASH/cryptogenic cirrhosis12121010
Metabolic/congenital1100
Secondary biliary1300
Other126610
Hepatocellular carcinoma (%)251916160.12
Donor age (years)a42 (25-56)49 (23-54)33 (20-48)40 (28-57)0.10
Donor race (%)     
White636658640.89
African American15151810 
Other22192426 
Cytomegalovirus (%)221218280.08
Rejection ever (%)454354620.06
Steroid-resistant rejection ever (%)7101280.58
CNI (%)b     
Tacrolimus655770750.32
Cyclosporine31352221
None4884 
Steroids (%)b535457750.04
Sirolimus (%)b151915150.88
Mycophenolate (%)b506257540.21
Induction (%)     
None565666780.06
OKT34542
Thymoglobulin2002
Daclizumab35343018
Combinationc3500

A univariate analysis of predictors of rejection and impaired patient survival was undertaken. Factors with P < 0.20 were used in a stepwise multivariable model analysis to develop the final models; however, each class of DSA being evaluated was always forced into the model to determine its role in the outcome. Where indicated, some factors with known associations with the outcome were also forced into the final model. Patients with a diagnosis of hepatitis C virus were only at risk for death from hepatitis C virus while they were viremic (this risk factor was censored at the sustained virological response to therapy). Statistical significance was defined as P < 0.05. SAS 9.1 was used for all statistical analyses.

RESULTS

The characteristics of all 1270 patients with a pre-LT sample available for analysis (95.8% of all patients) are shown in Table 1, and the patients are divided by the presence or absence of preformed DSAs. Patients with preformed DSAs were more likely to be female with an autoimmune liver disease and had higher pretransplant Model for End-Stage Liver Disease (MELD) scores. Their immunosuppression was more likely to include steroids and less likely to include induction, and they were more prone to rejection.

Isolated preformed class I DSAs were found in 84 patients (6.6%), isolated preformed class II DSAs were found in 50 patients (3.9%), preformed class I and II DSAs were found in 50 patients (3.9%), and no DSAs were found in 1086 patients (85.6%). The correlation between DSAs detected by single-antigen beads and cytotoxic crossmatching is shown in Supporting Table 1. The median MFI for preformed class I DSAs was 15,793 (interquartile range = 8384-20,266), and the median MFI for preformed class II DSAs was 12,880 (interquartile range = 7663-20,129). Among the patients with preformed class I or II DSAs, 49% had more than 1 distinct class I DSA, and 50% had more than 1 distinct class II DSA (Supporting Fig. 1).

Figure 1.

Outcomes of (A) preformed class I DSAs and (B) preformed class II DSAs with a MFI ≥ 5000 shown graphically for patients with post–orthotopic LT samples available for determining persistence. (C) Number of preformed DSAs by the MFI and their risk of persistence. Persistence was defined as a MFI after orthotopic LT ≥ 5000. De novo DSAs are not shown.

For patients undergoing transplantation with preformed DSAs, Fig. 1 shows the probability of persistent DSAs after LT (on a single follow-up test, with most tests occurring at 1 year). Class I DSAs (with or without class II) with an MFI ≥ 5000 persisted 5% of the time; class II DSAs (with or without class I) with an MFI ≥ 10,000 had an increased rate of persistence (33%) in comparison with class II DSAs with an MFI of 1000 to 4999 (1.5%) or class II DSAs with an MFI between 5000 and 10,000 (23%; P < 0.001). A separate analysis (data not shown) documented that patients with preformed class I and II DSAs had the same rate of persistence on follow-up testing as those with an isolated class.

Figure 2 demonstrates that preformed class II DSAs (but not preformed class I DSAs) were associated with an increased risk of early rejection (<6 months after LT) in a Kaplan-Meier analysis (P = 0.046). The percentage of patients with rejection and the median times to rejection and steroid-resistant rejection are presented in Supporting Table 2 for patients on the basis of their pretransplant DSA status. This risk of rejection was similarly shown in Cox proportional hazards multivariable modeling: class II (but not class I) was associated with an increased risk of early rejection after we controlled for low CNI levels (tacrolimus level < 3 ng/mL or cyclosporine level < 75 ng/mL before the diagnosis of rejection), autoimmune liver disease etiology [autoimmune hepatitis (AIH), primary biliary cirrhosis (PBC), or primary sclerosing cholangitis (PSC)], recipient age and race, and steroid and mycophenolate usage at week 4 after transplantation. Although stepwise multivariable modeling was used for most factors (and eliminated the MELD score at transplant and CNI use at week 4 from the final model), low CNI levels and autoimmune liver disease were forced into these models because of their known association with rejection.

Figure 2.

(A) Isolated preformed class I DSAs with a MFI ≥ 5000 did not increase the probability of early rejection, whereas (B) preformed class II DSAs with a MFI ≥ 5000 did. (C,D) Multivariable Cox proportional hazards modeling of risk factors for rejection within the first 6 months after LT. *A low CNI level was considered a tacrolimus level < 3 ng/mL or a cyclosporine level < 75 ng/mL before biopsy-proven rejection.

Figure 3 demonstrates a Kaplan-Meier association between preformed class I and II DSAs and impaired patient survival. Patients with preformed DSAs were compared to those without preformed DSAs or de novo DSAs. However, these represent unadjusted analyses. A univariate analysis of the predictors of patient survival was then undertaken (Supporting Table 3). Variables with a P value < 0.20 were entered into stepwise multivariable modeling, but they were kept in the final model only if their adjusted P value was <0.05. Multivariable modeling confirmed the independent association between preformed class I and/or II DSAs and patient death [hazard ratio (HR) = 1.51] after we controlled for African American race, hepatitis C virus viremia, donor and recipient age, cytomegalovirus infection, and etiology of liver disease and censored patients at the time of de novo DSA formation (P = 0.02; Fig. 3C). For this analysis, patients with preformed DSAs were compared to patients without DSA (MFI < 1000).

Figure 3.

Survival of patients with (A) isolated preformed class I DSAs or (B) isolated preformed class II DSAs with an MFI ≥ 5000. These patients were compared to patients without preformed or de novo DSAs with a MFI ≥ 5000 (no antibodies). The Kaplan-Meier analysis was unadjusted. (C) Multivariate modeling similarly showed decreased survival for patients with preformed DSAs. Notably, the final model was based on a smaller cohort that excluded patients with DSA levels between 1000 and 4999 and censored patients at the time of de novo DSA development. *Patients who had a sustained virological response to therapy were censored for this variable at the time of the sustained virological response.

A separate Cox proportional hazards model elucidated that patients with a high (>24) MELD score before transplantation may be at increased risk for death in the presence of pretransplant class II DSAs (HR = 1.4, P = 0.06), but the MELD score did not alter the risk of rejection or steroid-resistant rejection in those with or without class I or II DSAs (Supporting Table 4A). In addition, fewer patients with preformed DSAs (65%) were alive in comparison with those without DSAs (77%; Supporting Table 4B).

Although the analysis was underpowered, the impact of induction therapy on the risk for rejection according to the presence of each class of DSAs was retrospectively evaluated (Fig. 4). Induction therapy was associated with a reduced rate of rejection in patients regardless of their preformed DSA status, but it may have abated the risk of rejection in patients with preformed class II DSAs.

Figure 4.

(A) Induction therapy decreased the risk of rejection regardless of the presence of class I DSAs with a MFI ≥ 5000. (B) Induction abated the risk of early rejection in patients with class II DSAs with a MFI ≥ 5000. However, these were uncontrolled, retrospective analyses, and they must be interpreted with caution.

DISCUSSION

A multivariable analysis of primary LT recipients has shown associations between preformed class II DSAs and early rejection and between preformed class I and/or II DSAs and inferior patient survival. This study does not eliminate a possible correlation between DSAs with a MFI of 1000 to 4999 and rejection and/or death. The MFI cutoff of 5000 was used because in early reports, a negative (<10% killing) or weakly positive (10%-30% killing) crossmatch did not affect LT outcomes, unlike a highly positive crossmatch (>30% killing).[2] Similar results were observed with flow cytometric crossmatching[23]; patients with high levels of antibodies to T cells (patient serum/normal serum ratio > 30) showed higher rates of rejection in the first month after transplantation than patients with low antibody levels (patient serum/normal serum ratio = 2-30) and patients without antibodies. Therefore, we felt it prudent to use a higher MFI for an analysis of negative outcomes.[24, 25]

Fortunately, we found that preformed class I DSAs with a MFI ≥ 5000 almost always (95%) resolved on follow-up testing after transplantation, which in our study most often occurred at 1 year. However, because we evaluated this only once after transplantation, it is not possible to know how long the DSAs were present or whether they returned at a later time after sample acquisition. However, frequent sample acquisition by Taner et al.[20] did not elucidate rebound DSAs in patients after resolution during the first year after transplantation.

Unlike class I DSAs, preformed class II DSAs with higher MFIs more commonly persisted after transplantation.[26] Our study was not designed to elucidate the mechanism for clearance of DSAs versus persistence after transplantation; however, one may speculate on the basis of these and other data that the ubiquitous expression of class I on hepatocytes allows for rapid clearance in most cases.[16, 20, 27, 28] Class II expression in the liver is very limited in the absence of inflammation, so more frequent persistence may result.[29] However, changes in production, related or unrelated to immunosuppression medications, and removal via other mechanisms such as soluble class I secretion and Kupffer cell phagocytosis may also play a role.

Unlike class I DSAs, preformed class II DSAs were associated with an increased risk of early rejection. Although preformed DSAs are known to cause rejection in other solid organs, this evaluation cannot establish a causal relationship between DSAs and rejection in liver allografts.[6, 9, 16, 18, 19, 22, 29-33] Notably, DSAs may recruit cells into allografts that cause or result in rejection,[34, 35] or DSAs may simply be a marker of cellular memory that results in rejection. Because we did not re-evaluate the pathology of this cohort, we cannot exclude the possibility that patients experienced antibody-mediated rejection early after transplantation or went on to develop other types of antibody-mediated injury. Similarly, we do not have access to the precise timing of and reasons for patient sensitization, and this has resulted in an inability to correlate these factors with outcomes.

When the entire cohort was analyzed in multivariable modeling and patients with class I and II DSAs with a MFI ≥ 5000 were compared to all other patients (including those with DSAs with a MFI of 1000-4999), neither class I nor II DSAs were found to be independent predictors of patient survival. The association was strengthened but still remained insignificant when patients were censored upon the development of de novo DSAs.[36] However, only after the elimination of patients with DSAs with a MFI of 1000 to 4999 was a statistically significant association found. This likely means that some but not all DSAs with a MFI of 1000 to 4999 are clinically meaningful; however, those with an MFI ≥ 5000 are more likely to be associated with impaired survival.

In 1992, the potentially adverse impact of a cytotoxic crossmatch was known: “In such patients for whom crossmatch-negative donors may never be found … special therapeutic strategies perioperatively must be evolved.”[9] Toward this goal, OKT3 was used under the assumption that T cell help was a required component of injury in patients with DSAs.[9] Because recent Scientific Registry of Transplant Recipients data showed a survival advantage for patients who were induced,[37] we hypothesized that this resulted from its positive effect on patients with preformed DSAs. Therefore, we retrospectively analyzed the small group of patients who received induction therapy (mostly with daclizumab). This approach decreased the risk of early rejection in patients with preformed class II DSAs but did not affect overall survival in either univariate or multivariable modeling.[26] This analysis was underpowered and, therefore, should be interpreted with caution. However, randomized controlled trials are warranted to determine whether induction, preferably with cytotoxic agents, will prevent rejection in patients with preformed DSAs.

Single-antigen bead technology has markedly improved the sensitivity and specificity of DSA testing in comparison with cytotoxic crossmatching. However, not all patients with preformed DSAs experience rejection and/or die after LT. Although MFI is useful, this alone is insufficient to make these important distinctions. When HLA proteins are coupled to beads, some denaturing occurs, and epitopes that are not normally seen by the immune system in vivo are exposed. This can lead to antibody binding in vitro that is not clinically significant in vivo. In addition, not all preformed DSAs correlate with poor outcomes after LT[20, 21]; there may be subgroups of patients with worse outcomes or more antigenic epitopes that lead to an increased risk of pathological outcomes. In addition to single-antigen bead testing, other improved diagnostic tools are needed to definitively delineate DSAs with the highest association with poor outcomes. Some investigations that may be of interest include the utilization of IgG subclass testing, C1q testing, and the analysis of non-HLA donor-specific antibodies.[13, 38]

This preliminary appraisal was designed to detect associations between DSAs and adverse outcomes. In fact, preformed class II DSAs with a MFI ≥ 5000 are associated with an increased risk of early liver allograft rejection. In addition, preformed DSAs are correlated with an increased risk of patient death. At this time, heightened awareness for rejection is needed for patients with preformed class II DSAs. However, considerable work is needed to ascribe a mechanism of injury to these pathological processes and to more accurately identify patients at the highest risk for adverse outcomes. Randomized controlled trials of altered immunosuppression medications are essential, but ultimately we will need to differentiate donor-recipient pairs with preformed DSAs who are amenable to standard immunosuppression versus altered immunosuppression from those who need altered organ allocation.

Ancillary