2.1 Study design

All children who received an isolated liver transplant at <17 years of age and who were treated for biopsy-proven late ACR between January 2014 and January 2019 at Cincinnati Children's Hospital Medical Center, Cincinnati, Ohio, USA were included in the study. Late ACR was defined as occurring minimum of 6 months after transplantation. The patients were identified from our center's transplant registry utilizing ICD-10 codes for ACR and billing codes associated with the liver biopsy procedure. Patients who received a second liver transplant or a dual organ liver transplant (such as liver-kidney) were excluded from the study. Data were retrieved from the hospital-based digital patient medical records. In total, 60 biopsies from 54 patients were included in the analysis, with 1–3 episodes of late ACR occurring per patient during the study period. Patients were treated per center protocol with methylprednisolone 5 mg/kg I.V for 3 days, followed by a daily taper of 1 mg/kg each day along with increased tacrolimus trough level goals to target a range of 10–15 ng/ml. If liver tests improved sufficiently, the patient was discharged on oral prednisone at a dose of 1 mg/kg, to be tapered by 25% per week until discontinuation at week 4 post-ACR diagnosis. If liver tests did not improve, repeat methylprednisolone burst or ATG was administered per clinical judgment. If liver tests normalized, tacrolimus trough levels returned to post-transplant baseline trough goals of 3–8 ng/ml by 1 month post-ACR diagnosis.

2.2 Clinical and laboratory recorded variables

For each study subject, the following transplant-related variables were obtained via review of the electronic health record: recipient and donor age/race/ethnicity/sex, indication for transplant, recipient race/ethnicity, allograft type (whole, left lateral segment, right lobe, living donor [LD]), serostatus of Epstein-Barr virus (EBV) and cytomegalovirus (CMV) infections at the time of transplant, and history of biliary or vascular complications.

For each study subject, the following rejection-related variables were obtained via review of the electronic health record: age of recipient at rejection, time between transplant and rejection, ALT, total bilirubin, and GGT at the time of rejection diagnosis, time to ALT normalization, use of repeat methylprednisolone pulse or rabbit anti-thymocyte globulin (rATG), Banff score at the time of rejection, DSA, presence of autoantibodies, immunoglobulin G (IgG) level, history of prior or subsequent PTLD, and concurrent infection at the time of rejection. MLVI, which has been associated with risk of late ACR, was calculated by the standard deviation of tacrolimus trough measurements using at least 3 values over a 12-month time period.¹⁵ Subjects were followed for a minimum of 2 years, with a median follow-up duration of 4.5 years.

2.3 Independent pathologist review

All biopsies which had tissue available (n = 57 of 60) underwent independent review by a pathologist (D.L.) who was blinded to the initial pathology report and treatment response group. The Banff score was independently calculated for each biopsy, as were each of its components (portal inflammation: 0–3; ductulitis: 0–3; venulitis: 0–3). The degree of fibrosis was assessed by trichrome stain. Ductopenia was assessed by CK7 immunostain. C4d staining was performed in only six biopsies and was negative in all. Biopsies were uniformly assessed in a blinded fashion by D.L. for interface hepatitis (as a feature of dnAIH) and this was not observed. The degree of PC or eosinophil infiltrates in late ACR liver biopsy tissue was assessed by the following scale: 0–1 = none/rare, 2 = occasional, 3 = prominent.

2.4 Antibodies and reagents

Ten times Tris-buffered saline, Tween-20, and ProLong Gold antifade mountant with DAPI were from ThermoFisher Scientific. The Manual Opal 7-color IHC Kit was obtained from Akoya Biosciences. The following primary antibodies were used for immunofluorescence: CD4 (clone EP204, CellMarque, 1:1250 dilution, antigen retrieval in AR9 buffer), CD8 (clone SP57, Ventana, undiluted, antigen retrieval in AR9 buffer), CD20 (clone L26, Ventana, 1:2 dilution, antigen retrieval in AR6 buffer), CD68 (clone PG-M1, Abcam, 1:250 dilution, antigen retrieval in AR9 buffer), CD138 (clone B-A38, Ventana, 1:2 dilution, antigen retrieval in AR6 buffer), pan-CK (PA1-27114, Invitrogen, antigen retrieval in AR9 buffer, 1:500 dilution), AF750 goat anti-rabbit Ig (clone Ab175733, Abcam, 1:200 dilution).

2.5 Multiparameter immunofluorescence

Five-micrometer tissue sections of formalin-fixed paraffin embedded (FFPE) liver biopsy tissue were incubated overnight at 55°C to melt the paraffin. Slides were deparaffinized with sequential incubations in Histoclear (National Diagnostics, Fisher Scientific) for 10 min at room temperature (RT) × 3, then rehydrated in ethanol, and washed in ultrapure water for 5 min at RT with gentle rotation. Antigen retrieval was performed by microwaving slides at 20% power for 15 min in the appropriate antigen retrieval buffer. Following antigen retrieval, the slides were passively cooled to RT and subsequently rinsed with ultrapure water for 5 min. Slides were incubated with a blocking reagent to inhibit nonspecific antibody binding according to the manufacturer's protocol for 10 min at RT with gentle rotation. The primary antibody was diluted in blocking reagent and incubated at 4°C overnight (in the case of anti-CD20, -CD138, -CD4, -CD68, and -CK antibodies) or for 1 h at RT (anti-CD8) in a humidified slide chamber in the dark with gentle rotation. Slides were washed twice with TBST for 5 min at RT with gentle rotation. Excess liquid was removed from each slide. Tissue sections were incubated with either anti-mouse/rabbit HRP conjugated secondary antibody for 10 min at RT, followed by two washes with TBST × 5 min. Tissues were then incubated with Opal detection fluorophores according to the manufacturer's instructions: Opal 520 (panCK), Opal 570 (CD4), Opal 650 (CD138), or Opal 690 (CD20, CD68). In the case of CD8 staining, slides were incubated for 1 h at RT with AF750-conjugated goat anti-rabbit antibody diluted 1:200 in antibody diluent. Slides were initially stained for either CD20, CD68, or CD138, followed by serial staining for CK, CD4, and CD8. Slides were coverslipped with ProLong Gold DAPI, dried, and stored at 4°C for up to 48 h until imaging.

Images were captured using a Nikon Eclipse TiE widefield microscope affixed with an Andor Xyla 4.2 megapixel 12-bit sCMOS monochromatic camera. Automated tile scanning of liver sections was done using a fully encoded motorized stage. High-speed multi-channel fluorescence imaging was acquired using a Lumencor SpectraX LED light engine. Image analysis was performed using NIS Elements Advanced Research software (Nikon).

2.6 Statistical analysis

For CD4, CD8, CD20, CD68, and CD138 cell counts, a minimum of seven portal tracts per biopsy slide were imaged. Cells were enumerated using the automated NIS Elements Advanced Research software. Mean cell counts/high-powered field (hpf) were generated for each cell type per biopsy. A total of 21 RR and 24 DR biopsies were included in the final quantification, with the remainder (6 RR and 9 DR) excluded because they did not contain a sufficient number of portal tracts for analysis. Two-tailed Student's t-test was used to determine the statistical significance between rejection response groups and immune cell counts with a p-value < .05.

Repeated measure analysis in a generalized linear mixed effect logistic regression model was conducted to determine the relationship between rejection response type and demographic factors. For stepwise variable selection, the predictors of CD4 count, CD8 count, CD20 count, CD138 count, and peak ALT were log-transformed and analyzed via a univariate logistic regression model along with the predictors of age at transplant, sex at birth, indication for transplant, history of biliary or vascular complications, recipient race, graft type, MLVI, Banff score, DSA/autoantibody presence, and time between transplant and rejection. Concurrent infection with EBV or CMV at the time of rejection was not included in the variable selection due to >50% missing data. Odds ratios, confidence intervals, and p-values were reported. A two-sided p-value < .05 was used to determine the significance of variables in all analyses. Data were analyzed using SAS.

3.1 Patient demographics and cohort characteristics

Patients who had received an isolated liver transplant at <17 years of age and treated for biopsy-proven late ACR between January 2014 and 2019 were stratified into rapid responders (RR) or delayed responders (DR) based on ALT normalization within 30 days of diagnosis (RR) or longer than 30 days (DR). A total of 60 rejection episodes in 54 patients were included in the analysis with 1–3 episodes of rejection per patient. Thirty-three rejection episodes were DR (55%) and the remainder were RR. There were no statistically significant differences in baseline immunosuppression usage between the DR and RR groups, with the majority of patients on tacrolimus monotherapy and only eight patients on sirolimus monotherapy at the time of ACR (not shown). All patients were followed for a minimum of 2 years. There was one graft loss in the RR group for biliary stricture resulting in re-transplantation, and one patient in the DR group died due to tumor recurrence. The DR group reached the threshold value of ALT <50 on an average of 102 days after treatment, while the RR group normalized ALT by an average of 19 days post-diagnosis (Table 1). There were no statistically significant differences between DR and RR rejection episodes in age of donor or recipient or time between transplant and ACR, although there was a trend toward patients with DR being older at the time of rejection with a longer time between transplant and rejection (Table 1). Similarly, there were no differences between DR and RR patients in the categories of recipient sex, donor/recipient race, or transplant indication with the majority of subjects transplanted for biliary atresia (BA) in both groups (Table 1), and only one patient was transplanted for autoimmune hepatitis. Although both vascular and biliary complications have been associated with increased risk for rejection, these surgical complications occurred at similar rates in both DR and RR rejection episodes. As expected, DR rejection episodes were more likely to be treated with prolonged courses of IV steroids (8 DR vs. 3 RR), and all of the steroid-resistant rejection episodes requiring treatment with rATG (n = 3) occurred in the DR group (Table 1). No association was identified between rejection treatment response and CMV or EBV seroconversions, which were previously reported to confer risk for rejection (Table 1).³⁴ Although ACR episodes predispose to future rejection episodes, there were no statistically significant differences in prior rejection episodes or rejection episodes observed during the follow-up period in DR versus RR subjects (Table 1).

3.2 Serum GGT level, but not ALT, bilirubin, Banff score, or MLVI, is associated with delayed rejection treatment response

Rejection Activity Index (RAI) or Banff score was developed in adult liver transplant recipients undergoing early rejection, with three factors graded on a scale of 0–3: portal inflammation, bile ductulitis, and venulitis.^{27, 35} Similar to previous work,²⁷ the Banff score on the original histopathology report achieved similar severity scores in both DR and RR groups (OR 0.953, CI 0.657–1.381, p = .795, Table 2). There were no statistically significant differences found in Banff total, Banff sub-score, or fibrosis score between DR and RR subjects (Table 3). Although centrilobular infiltrates have been associated with late ACR,¹⁰ there were no statistically significant differences in either the presence or extent of centrilobular infiltrates in DR versus RR subjects (Table 3). Surprisingly and in contrast to prior studies,³⁵ the log-transformed peak ALT value (a biomarker of rejection severity) was also not significantly associated with treatment response (Table 2, OR 1.130). Because cholestatic hepatitis is associated with steroid-refractory rejection,²⁴ we analyzed total bilirubin and GGT levels at diagnosis in DR versus RR subjects. There was a trend in increased bilirubin in DR subjects compared to RR, although this was not statistically significant (p = .06; Figure 1). However, serum GGT mean levels (Figure 1; p = .05 by Student's t-test) and log-transformed GGT values (Table 2) were increased in DR compared to RR subjects, with each 10-fold increase in GGT level resulting in OR 5.242 of being DR (p = .03).

Medication level variation index is calculated by averaging minimum of 3 tacrolimus drug levels over a minimum 12-month period to obtain the standard deviation.¹⁵ MLVI has been associated with medication non-adherence and is a risk factor for rejection.¹⁵ However, in our cohort, the log-transformed MLVI surprisingly showed a trend toward inverse relationship with rejection treatment response, with increased MLVI having OR 0.597 of being DR to treatment (p = .06, Table 2). These data suggest that MLVI may be inversely correlated with rejection treatment response.

3.3 Increased relative presence of autoantibodies and DSA correlate with rejection treatment response

Acute cellular rejection, especially when it occurs late after transplant, can be difficult to distinguish from dnAIH, an entity that is histologically defined by interface hepatitis and the presence of anti-nuclear (ANA) and anti-smooth muscle antibodies (ASMA) arising in patients without a history of autoimmune liver disease.^29-31 DnAIH has more recently been re-defined as plasma-cell-rich rejection and tends to resolve much more slowly relative to T-cell-mediated rejection.⁹ Additionally, de novo DSA are a risk factor for pediatric late liver transplant rejection.^{11, 18} However, whether DSA are associated with rejection severity or treatment response in pediatric liver transplant recipients is not known. As such, we set out to determine whether the presence of autoantibodies or de novo DSA was associated with an increased risk of DR. As shown in Table 4, 55.17% of DR rejection episodes had DSA positivity measured within 3 months of rejection diagnosis as compared to 39% of RR rejection episodes. Similarly, nine DR rejection episodes were concurrently positive for autoantibodies (34%) as compared to one RR episode which was positive for autoantibodies (5%; Table 4). As shown in Table 2, patients with DSA measured within 3 months of rejection episode had a trend toward increased DR risk (OR 1.683, CI = 0.433–6.537, p = .358). Importantly, ANA and/or ASMA autoantibodies were more prevalent in patients with DR (Table 4). Univariate analysis showed a trend toward DR with an OR 9.136 (CI = 0.876–95.228, p = .064). Blinded pathologist review of DR versus RR liver biopsies did not reveal differences in the degree of PC infiltrates (Table 3) or interface hepatitis (not shown) between DR and RR rejection episodes, suggesting that DR subjects did not meet the criteria for dnAIH.

3.4 Inflammatory milieu at rejection diagnosis is associated with rejection treatment response

We next tested the hypothesis that the inflammatory milieu at the time of diagnosis of rejection was associated with rejection treatment response. Seven biopsies (RR = 3, DR = 4) were excluded from immunofluorescence analysis due to insufficient portal tracts (<7) for analysis from available archived FFPE liver tissue. A total of 21 RR and 24 DR FFPE 5 μm liver biopsy sections were stained sequentially with anti-CD20, -CD4, -CK, and -CD8 antibodies with the manual seven-color Opal IHC kit to allow simultaneous imaging of five parameters, including the DAPI nuclear stain. Serial sections from the same biopsy were used to image anti-CD138 and anti-CD68, respectively. A representative image from one RR and one DR liver biopsy is shown in Figure 2. As shown in Figure 3, increased portal infiltration by CD20⁺ and CD8⁺ cells occurred in DR rejection (p = .04, .05, respectively). There were no differences in CD4⁺ T cells, CD138⁺ PCs, or CD68⁺ macrophages in portal infiltrates in DR compared to RR biopsies (Figure 3). None of the cell counts reached a significance level in univariate analysis (Table 5).

The following terms were included in multiple regression analysis: log-scale CD4, CD8, CD20, and CD138 cell counts, log-scale peak ALT, age at transplant, recipient sex, transplant indication, history of biliary complications, and history of vascular complications. No significant term was identified at 0.10 level in the multiple regression model.