Transitional B cell cytokines risk stratify early borderline rejection after renal transplantation

Abstract

Borderline rejection (BL) in renal transplantation is associated with decreased allograft survival, yet many patients with BL maintain stable graft function. Identifying patients with early BL at risk for shortened allograft survival would allow for timely targeted therapeutic intervention aimed at improving outcomes. 851/1187 patients transplanted between 2013-18 underwent early biopsy (0-4 mos). 217/851 (25%) had BL and were compared to 387/851 without significant inflammation (NI). Serial surveillance and for-cause biopsies and seven-year follow-up were used to evaluate histological and clinical progression. To identify high-risk patients, we examined clinical/histological parameters using regression and non-linear dimensionality reduction (tSNE) and a biomarker based on peripheral blood transitional-1 B cell (T1B) IL-10/TNFα ratio. Compared to NI, early BL was associated with increased progression to late acute rejection (AR; 5-12 mos), premature interstitial fibrosis and tubular atrophy (IFTA) and decreased seven-year graft survival. However, decreased graft survival was limited to BL patients who progressed to late AR or IFTA, and was not influenced by treatment. Although tSNE clustered patients into groups based on clinical factors, the ability of these factors to risk stratify BL patients was modest. In contrast, a low T1B IL-10/TNFα ratio at 3 months identified BL patients at high risk for progression to AR (ROC AUC 0.87) and poor 7-yr graft survival (52% vs. 92%, p=0.003), while BL patients with a high ratio had similar graft survival to patients with NI (91%, p=NS). Thus, progressive early allograft inflammation manifested as BL that progresses to late AR in the first post-transplant year represents a high-risk clinical state for poor allograft outcomes. Such high-risk status can be predicted by the T1B IL-10/TNFα ratio before irreversible scarring sets in, thus allowing timely risk stratification.

Keywords: B cells; acute rejection; cytokines; kidney biopsy; transplantation.

Figure 1.. Study flow diagram.

Figure 2.. Acute and chronic Banff histological scores in patients with no significant inflammation and those with early sub-clinical (scBL) or clinical (cBL) borderline rejection.

(A). Proportion of patients with clinical vs. sub-clinical BL. (B-H). Comparison tubulitis (‘t’ score), interstitial inflammation (‘i’ score), glomerulitis (‘g’ score), peritubular capillaritis (‘ptc’ score), tubular atrophy (‘ct’ score), interstitial fibrosis (‘ci’ score), and arteriosclerosis (‘cv’ score) on index biopsy in patients with no significant inflammation (NI) vs. scBL vs. cBL. Bars represent mean ± SEM. Groups were compared by Kruskal Wallis test with Dunnett’s test for multiple group comparisons. *p<0.0001, **p=0.001, #p=0.02, ##p=0.04.

Figure 3.. Comparison of broad histological categories diagnosed on late biopsies (5–12 months) in patients with no significant inflammation (NI) and those with early borderline rejection (BL).

(A). Proportion of patients with either acute rejection (AR), or borderline rejection or no rejection on late biopsies (5–12 months) in patients with NI vs. BL. The inset compares the severity of AR defined either by Banff rejection grade ( 1A vs. 1B vs. ≥ 2/ABMR) or presence or absence of micro vascular inflammation (MVI). (B). Proportion of patients with either acute rejection (AR), or borderline rejection or no rejection on late biopsies (5–12 months) in patients with subclinical BL (scBL) vs. clinical BL (cBL). The inset compares the severity of AR defined either by Banff rejection grade ( 1A vs. 1B vs. ≥ 2/ABMR) or presence or absence of micro vascular inflammation (MVI). Rejection categories between patient groups were compared by Chi Square test.

Figure 4.. Analysis of interstitial fibrosis and tubular atrophy (IFTA) in late (5–12 months) biopsies.

(A). Comparison of cumulative IFTA scores in patients with no significant inflammation (NI) vs. those with early borderline rejection (BL). (B). Comparison of IFTA categories with no or minimal IFTA (IFTA scores 0–0.5), mild IFTA (IFTA score 1–2) and ≥ moderate IFTA (IFTA score>2) in patients with NI vs BL. (C-D). Comparison of cumulative IFTA scores and IFTA categories in NI vs. BL patients with no IFTA on their index biopsy. (E-F). Comparison of cumulative IFTA scores and IFTA categories in BL patients with no late acute rejection (AR) vs. BL patients with late AR. (G). Comparison of proportion of NI vs. BL patients who had either late AR and/or IFTA. Error bars represent SEM. Groups were compared by Mann Whitney U test (A, C, E) or Chi squared test (B, D, F, G).

Figure 5.. Early borderline rejection (BL) and long-term outcomes.

(A). Patient survival (mortality) in patients with BL vs. those with no significant inflammation (NI). (B). Death censored graft survival in patients with BL vs. NI. (C). Death censored graft survival in patients with either BL or NI with or without histological progression to late acute rejection (AR) by one year. (D). Death censored graft survival in patients with BL categorized based on their progression to late AR with or without microvascular inflammation (MVI) compared to patients with NI. (E). Death censored graft survival in patients with either BL or NI with or without IFTA by one year. Survival analysis was performed by Kaplan Meier method and survival curves were compared by Log rank test. * p=0.001 NI vs. BL ➔ AR; ** p<0.0001 vs. NI, p=0.001 vs. BL ➔ no AR, p=0.003 vs. BL ➔ AR without MVI; Δ p=0.0007 vs. NI; p=0.06 vs. no IFTA.

Figure 6.. Clinical risk stratification of early borderline rejection (BL).

(A). Clustering of patients with BL into 5 distinct neighborhoods using T-distributed Stochastic Neighborhood Embedding (t-SNE). (B). Spread of the donor type across the five t-SNE neighborhoods. (C). Cold ischemia time and the IT and the five t-SNE neighborhoods. (D). Kidney donor profile index (KDPI) and the five t-SNE neighborhoods. (E). Grouping of Neighborhoods 1–2 and 3–5 (given baseline similarities). (F). Proportion of patients with either acute rejection (AR), or borderline rejection or no rejection on late biopsies (5–12 months) in N1–2 vs. N35 patients. (G). Comparison of cumulative IFTA scores in N1–2 vs. N3–5 patients. (H). Death censored graft survival in N1–2 vs. N3–5 patients. Rejection categories between patient groups were compared by Chi Square test. Cumulative IFTA scores were compared by Mann Whitney U test. Survival analysis was performed by Kaplan Meier method and survival curves were compared by Log rank test.

Figure 7.. T1B IL-10/TNFα ratio and Risk stratification of early borderline rejection (BL).

(A). Comparison of B cell IL-10/TNFα ratio in patients with no significant inflammation (NI) compared to those with BL that did not progress histologically to late acute rejection (AR) or who did progress to late AR. (B). Comparison of TrB IL-10/TNFα ratio in patients with no significant inflammation (NI) compared to those with BL that did not progress histologically to late acute rejection (AR) or who did progress to late AR. (C). Comparison of TrB IL-10/TNFα ratio in patients with no significant inflammation (NI) compared to those with BL that did not progress histologically to late acute rejection (AR) or who did progress to late AR. (D). ROC-AUC for the prediction of late AR in patients with BL. Sensitivity, specificity, positive predictive value and negative predictive value were calculated based on the coordinates of the ROC analysis. (E). Percentage of high- vs. low-risk BL patients (based on the T1B IL-10/TNFα ratio) with the diagnosis of AR vs. borderline rejection vs. resolution of inflammation on the late biopsies (5–12mos). (F). Death censored graft survival in NI vs. BL patients who were further categorized as high-risk (T1B cytokine ratio<1.3) or low-risk (T1B cytokine ratio ≥1.3) based on T1B IL-10/TNFα ratio. Error bars represent mean ± SEM. Patient groups were compared by Kruskal Wallis test with Dunnett’s test for multiple comparisons (A-C) or by Chi-square test (E). Survival analysis was performed by Kaplan Meier method and survival curves were compared by Log rank test. *p=0.04, ** p=0.4, ***p=0.0045, ^p=0.003, ^^p=0.9, ^^^p=0.001, #p<0.0001, ##p>0.99, ###p<0.0001, Δ p<0.0001, ΔΔ p=.003 (vs. NI-LR) and p=0.02 (vs. BL-LR).