Enhanced Donor Antigen Presentation by B Cells Predicts Acute Cellular Rejection and Late Outcomes After Transplantation

Abstract

Background.

Enhanced B-cell presentation of donor alloantigen relative to presentation of HLA-mismatched reference alloantigen is associated with acute cellular rejection (ACR), when expressed as a ratio called the antigen presenting index (API) in an exploratory cohort of liver and intestine transplant (LT and IT) recipients.

Methods.

To test clinical performance, we measured the API using the previously described 6-h assay in 84 LT and 54 IT recipients with median age 3.3 y (0.05–23.96). Recipients experiencing ACR within 60 d after testing were termed rejectors.

Results.

We first confirmed that B-cell uptake and presentation of alloantigen induced and thus reflected the alloresponse of T-helper cells, which were incubated without and with cytochalasin and primaquine to inhibit antigen uptake and presentation, respectively. Transplant recipients included 76 males and 62 females. Rejectors were tested at median 3.6 d before diagnosis. The API was higher among rejectors compared with nonrejectors (2.2 ± 0.2 versus 0.6 ± 0.04, P value = 1.7E-09). In logistic regression and receiver-operating-characteristic analysis, API ≥1.1 achieved sensitivity, specificity, and positive and negative predictive values for predicting ACR in 99 training set samples. Corresponding metrics ranged from 80% to 88% in 32 independent posttransplant samples, and 73% to 100% in 20 independent pretransplant samples. In time-to-event analysis, API ≥1.1 predicted higher incidence of late donor-specific anti-HLA antibodies after API measurements in LT recipients (P = 0.011) and graft loss in IT recipients (P = 0.008), compared with recipients with API <1.1, respectively.

Conclusions.

Enhanced donor antigen presentation by circulating B cells predicts rejection after liver or intestine transplantation as well as higher incidence of DSA and graft loss late after transplantation

The presentation of donor antigen by B cells presents novel opportunities to further characterize, predict, and treat cellular and humoral rejection in transplant recipients. B cells present antigen to T-helper cells (Th), resulting in Th-mediated alloresponses. Th-mediated help is essential for B-cell activation and maturation to antibody-secreting plasma cells and for activation and memory formation in T-cytotoxic cells (Tc).^1-3 Activated B cells can also recruit other B cells to produce antibodies.^4,5 The CD40–CD154 interaction facilitates this communication within and between cell compartments and depends in part on the induction of CD40 on Tc and Th.^6,7 Reflecting the clinical import of these interactions, donor antigen–specific B cells, which express CD154 in lymphocyte co-culture, are enhanced during acute cellular rejection (ACR) after intestine transplantation and are strongly correlated with allospecific CD154⁺ Tc-memory cells (TcM), and with donor-specific anti-HLA antibodies (DSAs).⁸ B-cell presentation of alloantigen may be central to these observations. In mice with major histocompatibility complex (MHC) class II-deficient or HLA-DM-deficient B cells, cardiac allograft survival was enhanced and was accompanied by decreased T-cell activation and alloantibody production.⁹ HLA-DM facilitates the exchange of peptides loaded on MHC II during antigen presentation by B cells.¹⁰ Interestingly, peptides that complex more stably with MHC II are more likely to elicit Th activation, a pre-requisite for subsequent cellular and humoral alloresponses.¹⁰

We have previously described a novel assay that measures donor antigen uptake by B cells and identifies rejection-prone children with liver transplantation within a few hours.¹¹ This test system measures uptake of donor antigenic lysate in lieu of antigen presentation and is based on previous studies by others which show a direct correlation between uptake of nonallogeneic antigens and T-cell activation, the downstream effect of antigen presentation.^12-15 We have used this assay to evaluate the role of the HLA-DOA gene, which inhibits HLA-DM-mediated peptide exchange on MHC II molecules during antigen presentation by B cells.¹⁶ In genetic association studies, single nucleotide variants in the regulatory regions of HLA-DOA were associated with liver transplant (LT) rejection in children. The HLA-DOA gene inhibits antigen presentation in B cells.¹⁷

Here, we evaluate whether B-cell uptake of alloantigen represents presentation of alloantigen and the downstream T-cell response. Furthermore, we establish the performance of this test system as a predictor of rejection and late DSA in a larger cohort of liver and nonliver allograft recipients. The appearance of DSA late after transplantation is associated with graft loss.^18,19

MATERIALS AND METHODS

Human Subjects

Ficoll-purified peripheral blood lymphocytes (PBLs) were extracted from blood samples from 3 healthy adult human subjects and a total of 138 LT or intestine transplant (IT) recipients of whom 50 rejectors and 88 nonrejectors. All subjects were sampled after institutional review board-approved informed consent, and purified PBL archived in liquid nitrogen. Archived samples for this study were selected based on availability, and proximity to a defined biopsy-proven event or a clinical visit for surveillance within a 60-d period after sampling.

Outcome Groups

The 138 total subjects included 50 who experienced rejection (R), and 88 that did not (NR). Of 50 rejection episodes in 50 subjects, 49 were biopsy-proven, and included 43 diagnosed with for cause biopsies, and 6 diagnosed with surveillance biopsies. The 50th subject was diagnosed as having clinical rejection. Rejection was diagnosed using Banff criteria.²⁰ Among 88 nonrejectors, biopsies were performed for graft dysfunction in 41 subjects, 27 for cause and 14 for surveillance. In the remaining 47 nonrejectors, graft function was stable at the time of sample collection.

Sample Cohorts for Performance Testing

The 138 subjects yielded 151 samples which were analyzed in 3 groups posttransplant training set of 99 samples from 99 subjects, 32 posttransplant validation set samples from 32 subjects, and 20 pretransplant validation set samples, of which 13 were obtained from subjects who also contributed posttransplant samples, and 7 were obtained from 7 unique subjects (Table S1, SDC, http://links.lww.com/TP/D5). From these 138 total subjects, 99 posttransplant subjects were in the training set which included 30 who experienced R and 69 who did not (NR). These subjects were used to develop a threshold predictive of rejection. The performance of this threshold was validated in (1) the remaining 32 posttransplant validation set subjects, consisting of 14 R and 18 NR and (2) pretransplant samples from 20 subjects. Among them, the 7 subjects who only provided pretransplant samples consisted of 6 R and 1 NR. The remaining 13 subjects who provided pretransplant samples had also provided posttransplant samples. These 13 subjects included 5 R and 8 NR. In all, the 20 samples assessed in the pretransplant validation cohort were obtained from 11 R and 9 NR.

B-cell Antigen Uptake, Presentation, and the T-cell Alloresponse

To understand whether an antigen uptake assay was reflective of antigen presentation, we first modeled indirect antigen presentation by incubating antigenic lysate with an HLA-nonidentical responder PBL mixture consisting of purified B cells and Th. For this experiment, (a) fluorochrome-labeled antigenic lysate labeled with carboxyfluoresciensuccinimidylester was prepared from PBLs, (b) responder B cells and Th were purified individually using Miltenyi magnetic beads, and reconstituted to make up responder PBL. Three types of responder B cells were used in each of 3 replicates of this experiment: untreated B cells, B cells pretreated with cytochalasin D, an inhibitor of antigen uptake, and B cells pretreated with primaquine, an inhibitor of antigen presentation. The frequency of CD19⁺ B cells, which express fluorochrome-labeled lysate, and CD4⁺ Th, which express CD154, were measured by flow cytometry.

Predicting Rejection in Transplant Recipients

PBLs from children with liver or intestine transplantation were tested with our previously described antigen presenting assay in which the uptake of donor and HLA-nonidentical third-party (reference) antigen was measured in parallel reaction conditions, as described.¹¹ The results were expressed as a ratio called the antigen presenting index (API).

Statistical Methods

In 99 training set samples from 99 subjects with liver or intestine transplantation, exhaustive forward, backward, and stepwise logistic regression (LR) and receiver-operating-characteristic (ROC) curve analysis was used to identify a threshold API associated with rejection within the 60-d period after the assay, as well as sensitivity, specificity, and positive (PPV) and negative predictive values (NPV), as described previously.^21-23 Covariates in the model included gender: male and female, race: white and non-white, induction: no induction/induction, tacrolimus whole blood concentration (FKWBC). Age at transplantation was included as a continuous variable. The performance of this threshold API was validated in independent posttransplant (n = 32) and pretransplant samples (n = 7) as described earlier. We used SPSS version 27 (IBM Corporation, NY) to perform descriptive statistics, between-group comparisons with t-tests, and LR-ROC analyses. We used Stata v17 (StataCorp, College Station, TX) to perform survival analysis. We built a Cox proportional hazards stratified on organ transplanted to determine the association between API and DSA development.

RESULTS

Human Subjects

Subjects included 83 with LT, 1 with combined liver-kidney transplant, 25 with ITs, and 29 with combined liver-intestine-pancreas transplants. The indications for LT or IT are shown in Table S2 (SDC, http://links.lww.com/TP/D5).

General demographics are shown in Table S3 (SDC, http://links.lww.com/TP/D5). Briefly, the mean age at the time of the sampling in 131 subjects who provided posttransplant samples for model building was 9.7 ± 7.5 (range 0.1–27 y). Twenty-three of these 131 subjects (18%) were >18 y of age. Rejectors demonstrated elevated liver enzymes aspartate transaminase, alanine transaminase, and gamma-glutamyl transferase levels compared with NR at or within the 10-d period before blood sampling (Figure S1, SDC, http://links.lww.com/TP/D5). Rejectors had significantly higher FKWBC at the time of sampling when compared with NR (mean ± SD 10.43 ± 7.78 versus 6.69 ± 4.74, P = 0.002). The FKWBC was also significantly higher in the subjects who had API ≥1.1 when compared with subjects with API <1.1 (mean ± SD was 6.8 ± 4.7 versus 9.7 ± 8.6, P = 0.043).

B-cell Alloantigen Uptake Reflects B-cell Presentation of Alloantigen and Th Alloresponse

Uptake of HLA-nonidentical carboxyfluoresciensuccinimidylester-labeled antigenic lysate by B cells was inhibited by pretreatment of B cells with cytochalasin D, an inhibitor of antigen uptake but not by primaquin, an inhibitor of antigen presentation in 3 replicates of this experiment. However, downstream Th alloresponse measured with CD154⁺ Th cells was inhibited by B-cell pretreatment with either inhibitor. Bar diagrams in Figure 1A summarize results of 3 replicates of this experiment. Thus, antigen uptake measured by our assay is also representative of the final step of alloantigen presentation by B cells, and the downstream Th alloresponse.

FIGURE 1.

Alloantigen uptake, presentation, and Th alloresponse. A, Alloantigen uptake, presentation, and Th alloresponse. B cells purified from normal human PBL were treated with Cyto D or Prima and reconstituted with purified CD4 Th from the same subject into a responder cell mixture. Stimulation was performed with HLA-nonidentical CFSE-labeled antigenic lysate. Top, Histograms show that uptake of HLA-nonidentical CFSE-labeled antigenic lysate by B cells. Bottom, Histograms show the downstream T-cell alloresponse measured as frequencies of CD154⁺ CD4Th cells after each treatment. Bar diagrams on right summarize results from 3 replicates. B, B-cell alloantigen presentation in a rejector and nonrejector. Three scatterplots each from a rejector (upper 3 plots) and a nonrejector (lower 3 plots) show frequencies of responder B cells that present fluorochrome-labeled donor alloantigen (donor) and dye-labeled HLA-nonidentical (reference) alloantigen. The background indicates recipient B cells cultured without either alloantigen. CFSE, carboxyfluoresciensuccinimidylester; Cyto D, Cytochalasin D; PBL, peripheral blood lymphocyte; Prima, Primaquine; Th, T-helper cells.

API Threshold of ≥1.1 Predicts ACR in Transplant Recipients

In LR analysis, the optimal model identified API as a single variable predictor of rejection in the training set. ROC curve analysis identified an API of 1.06 as having the optimal balance of sensitivity and specificity for prediction of rejection. For corroboration, API measurements from the 54 LT and 45 IT recipients in the training set were subjected to separate LR and ROC analyses, using the same covariates applied to the combined analyses. The optimal model once again identified API as the best predictor of rejection outcome at or above threshold values of 1.06 and 1.1085, respectively. We chose the threshold API of 1.1 at or above which rejection is predicted as a common threshold for all organs and applied it to 32 posttransplant and 20 pretransplant validation samples to determine test performance.

Test Performance

A higher proportion of B cells presented donor antigen relative to reference alloantigen, leading to a significantly greater API among rejectors compared with nonrejectors (2.2 ± 0.2 versus 0.6 ± 0.40, P value = 1.7E-09, Figure 1B) in the combined cohort of 151 training and validation set samples from 138 LT and IT recipients. LR and ROC analysis showed that API ≥1.1 predicted ACR in the 99 training set samples with sensitivity, specificity, PPV, and NPV and corresponding 95% confidence intervals (CIs) of 93% (76-99), 97% (89-100), 93% (76-99), and 97% (88-99), respectively. In 32 independent posttransplant blinded validation set samples, respective performance metrics for the threshold API ≥1.1 were 86% (57-98), 83% (59-96), 80% (52-96), and 88% (64-99), respectively. In 20 pretransplant samples, respective performance metrics for the threshold API ≥1.1 were 73% (39-94), 100% (55-100), 100% (52-100), and 75% (43-95), respectively (Table 1; Figure S2, SDC, http://links.lww.com/TP/D5).

TABLE 1.

Summary of test performance measured in the various cohorts based on the API threshold of ≥1.1

Group	Sensitivity		Specificity		PPV		NPV
	Estimate	95% CI	Estimate	95% CI	Estimate	95% CI	Estimate	95% CI
Training set	0.93 (28/30)	(0.76-0.99)	0.97 (67/69)	(0.89 –1.00)	0.93 (28/30)	(0.76–0.99)	0.97 (67/69)	(0.88–0.99)
Validation set	0.86 (12/14)	(0.57-0.98)	0.83 (15/18)	(0.59-0.96)	0.80 (12/15)	(0.52-0.96)	0.88 (15/17)	(0.64-0.99)
Validation set (pretransplant)	0.73 (8/11)	(0.39-0.94)	1.00 (9/9)	(0.55-1.00)	1.00 (8/8)	(0.52-1.00)	0.75 (9/12)	(0.43-0.95)

Unless otherwise stated, posttransplant samples were used in the training set and validation set.

API, antigen presenting index; CI, confidence interval; NPV, negative predictive value; PPV, positive predictive value.

Enhanced Donor Antigen Presentation by B Cells Precedes Future Development of DSA

DSA measurements were available for 85 of 138 subjects with posttransplant API measurements. These 85 recipients included 32 with an API ≥1.1 indicative of increased presentation of donor antigen, and 53 children with API <1.1. B-cell API was measured at median 411 (5–5817) d after transplantation. In recipients with API ≥1.1, the incidence of DSA was higher at 20 of 32, 63% versus 16 of 53, 32%, P = 0.006, Fisher exact test, compared with those with API <1.1. Time to DSA in the respective groups was lower with median 2401 (392–3539) d versus 3122 (825–8282) d, in recipients with high API compared with those with API <1.1 (P < 0.01, Kaplan-Meier [K-M] test). In the subcohort of 55 LT recipients, those with API ≥1.1 had a higher incidence, 19 of 28 (68%) versus 10 of 27 (37%), and earlier detection of DSA at 2989 (1359–6991) d versus 3204 (825–8282) d in those with API <1.1 (P = 0.011, K-M test). Among IT recipients, DSAs were present in 1 of 4 (25%) recipients with API >1.1 and 6 of 26 (23%) recipients with API <1.1 (P = NS). In Kaplan–Meier analysis, greater risk of DSA development was seen in those with API >1.1 among all recipients (P < 0.01) and those receiving livers alone (P = 0.011; Figure 2; Figure S3, SDC, http://links.lww.com/TP/D5). Cox proportional hazards model stratified by graft type showed that the risk of DSA development over time is 2.29 times higher in recipients with API >1.1 compared with recipients with API <1.1 (95% CI, 1.15-4.59, P = 0.019).

FIGURE 2.

Test performance. Upper row, Dot plots show API values for rejectors and nonrejectors in (A) all posttransplant training set samples, (B) all posttransplant validation set samples, and (C) all pretransplant validation set samples. Lower row, K-M plots show time to late DSA measurements (D) 85 of 131 total study subjects for whom DSA measurements were available, and (E) 55 LT recipients. F, K-M plot shows graft loss rate in 52 IT recipients. Each plot shows 2 subgroups, recipients with API <1.1 (green) and those with API ≥ 1.1 (red). API, antigen presenting index; DSA, donor-specific anti-HLA antibody; K-M, Kaplan-Meier; LT, liver transplant.

Enhanced Donor Antigen Presentation by B Cells Is Associated With Graft Loss in IT Recipients

Of 131 LT and IT recipients who contributed training and validation sets of posttransplant samples, 20 experienced graft loss. Causes of graft loss included acute on chronic rejection in 13, death with functioning grafts in 3, acute refractory rejection and graft-versus-host disease in 2, acute rejection and infection in 1, and hepatic artery thrombosis in 1. These graft losses occurred in 9 of 45 recipients with API ≥1.1 and 10 of 86 with API <1.1, at median (range) 1626 (190–6250) d versus 2390 (228–4661) d, respectively (P = 0.235, NS, K-M test). In Kaplan–Meier analysis, the risk of graft loss was greater for IT recipients with API >1.1 (P < 0.01). Graft losses occurred earlier at median 727 (190–6250) d in 7 of 13 (54%) recipients with API ≥1.1, compared with median 2571 (228–4661) d in 9 of 39 (23%) recipients with API <1.1 (P = 0.008, K-M test; Figure 2; Figure S3, SDC, http://links.lww.com/TP/D5). In the LT subcohort, graft loss occurred in 2 of 32 (6%) recipients with API ≥1.1 and 2 of 47 (4%) recipients with API <1.1 (P = NS).

DISCUSSION

This study expands on our previous findings by showing that alloantigen uptake is representative of alloantigen presentation by B cells and the downstream T-cell response (Figure 1A). The uptake of alloantigen was associated with enhanced Th activation resulting in higher frequencies of CD154⁺ Th cells. The antigen presentation process encompasses antigen uptake and presentation.^12-15 Inhibition of either process with cytochalasin and primaquine, respectively, resulted in decreased frequencies of CD154⁺ Th cells, compared with those seen in the absence of these inhibitors.

In our previous study, we found that an API, which measured the presentation of donor antigen relative to presentation of HLA-nonidentical third-party or reference alloantigen, distinguished rejectors from nonrejectors in a modest cohort of LT or IT recipients.¹¹ Here, we further show using LR and ROC analysis that a threshold API of ≥1.1 can identify LT or IT recipients who develop ACR within the 60-d posttest period with a sensitivity, specificity, and PPV and NPV approaching or exceeding 80% in the validation set of posttransplant samples (Figure S2, SDC, http://links.lww.com/TP/D5). For 20 pretransplant samples, the prediction or rejection within the 60-d posttransplant period approaches or exceeds 75% for each performance metric. Applied to posttransplant samples for each organ system separately, an API of ≥1.1 has PPV and NPV of 77% and 92%, respectively, for LT recipients, and 100% and 80%, respectively, for IT recipients (Figure S4 and Table S4, SDC, http://links.lww.com/TP/D5). These findings have clinical application for the rapid assessment of rejection risk, if the test system is used with clinical and other laboratory parameters, and not by itself. The test system and the reporting of the API can be completed in 4–6 h. This can allow the caregiver to develop a treatment plan, while waiting for confirmatory biopsies, or validate the need for a confirmatory biopsy.

The added clinical benefits of measuring B-cell alloantigen presentation in the clinic may include identification of recipients at risk to develop DSA late after transplantation, and those who are at risk for late graft loss (Figure 2; Figure S3, SDC, http://links.lww.com/TP/D5). In our study, the API threshold of ≥1.1 predicted a 2.2-fold greater risk of developing late DSA, ≥1 y after API measurements (Figure 2; Figure S3, SDC, http://links.lww.com/TP/D5). Further there was no significant difference between HLA match/mismatch between donor–recipient who developed DSA versus who did not develop DSA (Table S5, SDC, http://links.lww.com/TP/D5).

This threshold was also associated with a significantly higher incidence of late graft loss among IT recipients, who are susceptible to this event because of the highly immunogenic intestine allograft. LT recipients did not demonstrate this association between B cells that present donor and graft loss, likely because of the liver’s regenerative ability in the face of immunological injury. The added benefit of predicting who might develop late DSA can aid in the selection of maintenance regimens in those with an API ≥1, and who may have experienced T cell–mediated coupled with circulating DSA. Such individuals are at risk for recurrent episodes of rejection. A high API may also stratify high-risk IT recipients into those who need more or less immunosuppression. After intestine transplantation, graft survival has yet to approach those of other organs.

We acknowledge the obvious limitations of our cross-sectional study, a predominantly pediatric population, and the lack of serial measurements to describe a longitudinal relationship between changing immunosuppression levels in the course of routine care. Additional studies can address these limitations. Enhanced donor antigen presentation in pretransplant samples from rejectors mirrors similar relationships of pretransplant T-cell alloreactivity with rejection outcomes²³ but cannot be adequately explained by prior sensitization due to transfusions, pregnancies, or a previous transplant. Cross-sensitization by prior exposure to viral antigens are applicable to T cells and have not been demonstrated for B cells.^24,25 Also, in a study of renal transplant recipients, sensitized patients demonstrated enhanced numbers of memory B cells, whereas nonsensitized patients did not.²⁶ Possible explanations may lie in interactions between B cells, dendritic cells, and T cells^27,28 but will require further study. Thus, for the rare cohort of subjects in this study, pretransplant samples can serve as a reasonable cohort to test validity of the predictive threshold. In support of this assumption, we find that among 13 subjects who also contributed posttransplant samples for assessment of posttransplant performance, pretransplant performance of an API of ≥1.1 consisted of sensitivity of 80% (4 of 5) and specificity of 100% (8 of 8). In the 7 unique subjects who only provided pretransplant samples, pretransplant performance of an API of ≥1.1 consisted of sensitivity of 67% (4 of 6) and specificity of 100% (1 of 1).

In conclusion, donor antigen presentation by B cells is associated with ACR with clinically acceptable sensitivity and specificity and may have the added benefit of predicting the development of DSA and late graft loss after liver or intestine transplantation.

ACKNOWLEDGMENTS

Techniques are described in patent US10222374, Assignee University of Pittsburgh, Licensee Plexision, Inc, Pittsburgh, PA. Raizman-Hainey Endowed Fund, Hillman Foundation of Pittsburgh.