The Impact of Transplant Recipient and Donor Organ Type 2 Diabetes Polygenic Risk Scores on the Development of Early Post-Transplant Diabetes

Post-transplant diabetes mellitus (PTDM) is a complication which reduces allograft and recipient life-span. Polygenic risk scores (PRS) robustly show association with greater type 2 diabetes (T2D) development risk. We examined T2D-PRS in transplant recipients and donors using genome-wide genotyping in 1581 liver recipients, and 1555 donors and 2062 kidney recipients and 533 donors from four centers. Liver and kidney recipient T2D-PRS was associated with pre-transplant T2D and PTDM development. Liver donor, but not kidney donor, T2D-PRS was an independent risk factor for PTDM development. Inclusion of a combined liver recipient and donor T2D-PRS significantly improved PTDM prediction vs clinical characteristics-only models: AUC (95%CI): 67.6% (64.1% - 71.1%) vs. 62.3% (58.8% - 65.8%), p=0.0001. Liver recipients in the highest quintile of recipient-donor combined T2D-PRS had the greatest PTDM risk: OR (95%CI) = 3.22 (2.07 - 5.00), p=1.92E-07, compared to the lowest quintile. T2D-PRS allows identification of transplant candidates with high PTDM risk, for whom early preemptive diabetes management is warranted. Pre-transplant knowledge of donor T2D-PRS in the setting of living liver donation should optimized selection of donors to reduce PTDM. potential for practical application of the T2D-PRS based predictive model in the transplant setting, we tested whether the recipient genomics contribute to PTDM development regardless of the specific organ transplanted. Our results demonstrated that irrespective of the type of the solid-organ transplanted, recipient T2D-PRS is a predictor of PTDM. Multivariate analyses confirmed that recipient T2D-PRS, with adjustment of known prognostic clinical variables, was an independent contributor to PTDM risk in both liver and kidney transplant settings. The findings were independently confirmed in the different liver and kidney centers cohorts, regardless of the different . Details of genotyping, genotyping data quality


MAIN
Post-transplantation diabetes mellitus (PTDM), previously referred to as New Onset of Diabetes after transplant (NODAT), is a common complication following solid organ transplantation, occurring in approximately 16-44% of kidney and liver transplant recipients 1,2 . The diagnosis is established by either hemoglobin A1C (A1c) and/or fasting glucose levels, starting 45 days post-transplantation, and is observed within the first 6 months after transplantation 3,4 . PTDM shares many metabolic syndrome characteristics with T2D, including insulin resistance and decompensated insulin release, hypertriglyceridemia, obesity, hypertension, and low-grade inflammation 5,6 . The development of PTDM, and the related metabolic syndrome phenotype(s), are associated with an increase in cardiovascular disease complications, which is now one of the leading causes of recipient death after transplantation 7,8 . The modifiers causing rapid onset of PTDM include the acute exposure to diabetogenic immunosuppression therapy (IST), specifically, the exposure to high doses of steroids, maintenance calcineurin (CNI) and/or mTOR inhibitors, leading to either early transient nonsymptomatic or the persistent manifestation of PTDM [9][10][11][12] . Considering the well-established role of individuals' genetic background in the development of T2D in general populations 13,14 , we questioned whether the stress introduced by the acute exposure to diabetogenic IST post-transplant may lead to the early induction and persistent of PTDM in subsets of transplant recipients with high T2D genetic risk exposure. Another fundamental question is whether the risk of PTDM is impacted by the transfer of the organ donor T2D genetic risk, and whether this is specific to the type of solid organ transplant.
In the setting of liver transplantation this hypothesis is very relevant, as this organ plays a central role in metabolic homeostasis and is a major site for synthesis, metabolism, storage and redistribution of carbohydrate and lipids 15,16 .
Genome-wide association studies (GWAS) have uncovered contributions of inherited variants across a spectrum of common complex disease and phenotypes [17][18][19] . Polygenic risk scores (PRS) aggregate and weight the effects of many genetic variants across the human genome into a single score and have been shown to have predictive value for multiple common diseases 20 . Recent aggregated genotyping data from T2D GWAS including 74,124 T2D cases, and 824,006 controls in Europeandescent cohorts revealed a combined set of 403 common and rare independent T2D-risk genetic signals, with the derived T2D polygenic risk scores (T2D-PRS) having robust association with the development of T2D in the general population 21 . The generation of T2D-PRS that capture aspects of the etiological and clinical heterogeneity that contributes to variable clinical outcomes, provides a potential mechanism for identifying transplant candidates who are at risk of developing PTDM.
We applied this T2D-PRS in the setting of liver and kidney transplantation, with the aim of determining the association of recipient and/or donor genetics, both independently and in concert, with the development of PTDM. QuantifyingT2D-PRS in transplant candidates may be used to personalize immunosuppression, such as steroid avoidance protocols, and to risk-stratify those who may benefit from aggressive treatment such as dieting, glucose monitoring, and insulin therapy 22,23 . In the liver transplant setting it may also be possible to reduce the risk of transmission of T2D-PRS with the donated organ via better donor-recipient matching, such as when organs are offered from multiple living donors.

Patient Populations:
The study characteristics of four cohorts, two liver transplant cohorts and two kidney transplant cohorts, are shown in Table 1 Relationship between T2D-PRS and PTDM in liver transplant recipients: After removing recipients with pre-transplant T2D, both recipient and donor T2D-PRS were found to be independently associated with PTDM risk in liver transplant recipients ( Table 3)

Relationship between T2D-PRS and PTDM in kidney transplant recipients:
In the kidney cohort, recipient T2D-PRS was significantly associated with PTDM development (Table 4). Recipient clinical variables including age and BMI, also had substantial influence on PTDM development. Analyses of the kidney cohort was limited to recipients of European ancestry since > 98% patients enrolled in Leuven are of European ancestry, and the results should thus be interpreted in this setting only. The limited donor data available from the DeKAF cohort indicates that kidney donor genomics, and/or clinical variables does not have impact PTDM development (Table 4). We tested two clinical scenarios, in which modification of induction and maintenance immunosuppression results in 30% or 60% reduction of PTDM in the recipients among the highest risk categories (quintile 4 and 5) based on the combined donor-recipient T2D-PRS. These interventions targeting those patients at highest genetic risk would be expected to lower the overall incidence of PTDM in the entire recipient population, from the observed 28.1%, to the expected 23.8% and 19.5% respectively (Supplementary Table 2).

DISCUSSION
PTDM is recognized as a common complication of solid organ transplantation which requires intensive patient management. The clinical risk factors are well characterized, and overlap with some of the well-known variables associated with T2D in the general population such as age, gender, and BMI 27,28 . Unique to the transplant population is the rapid development of diabetes post-transplant, which is attributed to the use of multiple diabetogenic drugs, specifically steroids and calcineurin inhibitors [9][10][11][12] . Previous work identified PTDM as independent predictor of mortality, and there is a significant association with the development of cardiovascular events 7,8 . It is likely that novel risk stratification tools, incorporating predictive clinical and genetic variables, are needed to predict the risk of PTMD development, and to assist in the design of strategies to reduce the clinical expression.
Recent advances in the identification and characterization of DNA polymorphims associated with individual clinical predisposition to T2D, provide integrated PRS that have the potential to influence clinical management. Aggregated genotyping data from European-descent GWAS of 74,124 T2D cases, and 824,000 controls, imputed using high density reference panels, revealed a combined set of 403 highly significant T2D-risk genetic variant signals 21 . Other large population studies identify similar signals, confirming the potential of T2D-PRS in predicting the development of disease 21,29 and their association with cardiovascular, renal, and neuropathy complications 21,29 . Our study applied these T2D-PRS approaches in the transplant setting, hypothesizing that the PTDM clinical phenotype is rapidly induced by predisposing recipient clinical variables and immunosuppression drugs, in genetically susceptible recipients. Consequently, identifying individuals at risk could lead to the development of preemptive therapeutic strategies, and/or donor-selection, to prevent diabetes complications.
As a proof of principle, our study tested whether the T2D-PRS can identify solid-organ transplant candidates who develop T2D prior to their transplantation. Analyses of subjects with pre-transplant T2D vs non pre-transplant T2D demonstrated strong association with T2D status, consistent with the original GWAS meta-analyses results from which the T2D-PRS was derived 21 .
To examine the potential for practical application of the T2D-PRS based predictive model in the We next determined whether the donor T2D-PRS and/or clinical phenotype should be considered as risk exposures for PTDM development. Donor T2D-PRS was found to be associated with PTDM in liver transplant, but not kidney transplant, demonstrating unique contribution of the donor liver for PTDM development. This is not surprising, considering the central role of the liver in the regulation of glucogenesis, glucose metabolism and in insulin clearance, as well as regulation of lipogenesis 15,16 .
There was no evidence that the liver or the kidney donor clinical phenotype is associated with the development of PTDM in this study. This observation leads us to conclude that unique to the liver transplant setting, the donor T2D-PRS is as an independent risk variable, and that the risk score should include combined impact of the recipient and donor genomics. The limited data from the kidney cohort suggested that the donor genomics does not impact PTDM outcomes, and that the predictive score may be determined by the recipient genomics alone.
Previous studies suggest that induction and maintenance immunosuppressive drugs, including the use of high dose steroids and CNI early after transplantation, account for 74% of the risk for PTDM 30,31 . Consequently, it is reasonable to hypothesize that these medications are most likely promoting and accelerating the manifestation of PTDM in the genetically susceptible recipients. The relatively rapid appearance of the clinical PTDM phenotype provides an opportunity to test whether T2D-PRS can be utilized in the preemptive design of intervention strategies to reduce the development of the clinical complication. The approach should include modification of immunosuppression, such as steroid-free immunosuppression induction protocols, substituting CNI with costimulatory blockade, and aggressive monitoring and early treatment of PTDM, that can significantly reduce the observed incidence in the entire transplant population [34][35][36] . Adjustment of immunosuppression must balance the risk of rejection versus the benefit of reduction in PTDM incidence [37][38][39] . Consequently, the clinical application should be first tested in selected recipients who have increased risk of PTDM development, and who are at low rejection risk.
To demonstrate the potential impact, we proposed scenarios in which only recipients with highest combined donor/recipient T2D-PRS quintiles undergo modification in post-transplant management, and calculated the outcomes if a portion of these selected recipients would respond well by avoiding the development of PTDM. Targeting high T2D-PRS upper quintile recipients, and reduction of the incidence of PTDM in this subset of recipients, will significantly reduce the incidence of PTDM in overall liver transplant populations.
Unique to the liver transplantation is the question whether living donor T2D-PRS should be tested prior to donation, and whether D-R genomic matching should be a consideration. This is far more relevant in the pediatric liver transplant setting, where polygenic risk variants transferred with the transplanted allograft, may impact the long term survival of the allograft and the child. We would advocate for non-immune related genetic matching, aiming to avoid identifiable risk exposure to the development of this complication, specifically when large pools of living liver donors are available.
There are several limitations of this study. First, the PRS used in this study derived from studying T2D in general population, not in transplant specific setting where additional exogenous factors such as stress from surgery and medication which may interact with additional genetic variants for the development of PTDM. GWAS of large transplant cohorts may reveal additional genetic variations contributing specifically to PTDM and incorporation of additional PTDM-specific genetic variants into T2D-PRS would likely enhanced its ability to predict the risk of PTDM. Secondly, we have limited power to detect the effect from kidney donors due to the low PTDM incidence rate in the DeKAF cohort. Our null observation of kidney donor genetic contribution to PTDM would require to be confirmed in other kidney cohorts with larger sample sizes, or with higher PTDM incidence.
Our study demonstrates the importance, and the potential application, of PRS in solid-organ transplantation. Using T2D-PRS as a clinical predictive assay could lead to personalized treatment strategies, aiming to reduce the occurrence of PTDM. While the study focuses on the risk exposure for the development of PTDM, the same approach could be used to explore the impact of recipient and donor genetics on the development of other significant morbid complications, that influence long term recipient survival e.g. chronic kidney disease.

METHODS
The study included two liver and two kidney transplant cohorts, transplanted at nine North American and one European centers. The first liver cohort, included recipients and their respective donors, The studies were approved by the Institutional Review Boards at each of the respective institutions.
The participants signed informed consents prior to transplantation, and at the time of organ donation.
Genotyping: Genotyping was conducted as previously described using a custom genome-wide genotyping tool, the Affymetrix Axiom Transplant Array, which was tailored with content for transplantation outcomes 41 . Quality control and assurance of these datasets was performed in accordance with the iGeneTRAiN GWAS pipelines 42 . Details of genotyping, genotyping data quality control, imputation, and the determination of racial clusters using principle components (PCs) can be found in the supplemental document.
Statistical Analysis: We used the published summary statistics of a set of 403 T2D-associated, independent risk signals from a recent T2D GWAS including 74,124 T2D cases, 824,006 controls to calculate PRS 21 . Genotyping data quality control and imputation procedure were detailed in the supplementary information. SNPs with imputation info score > 0. 8     .69E-06 *Recipients dropped out of this analysis dues to missing covariates such as missing recipient BMI (n=28 in kidney recipients), missing donor BMI (n=2 in liver recipients, n=35 in kidney recipients), missing recipient sex (n=3 in kidney recipients), missing donor sex (n=7 in kidney recipients), missing recipient age (n=3 in kidney recipients), and missing donor age (n=10 in kidney recipients).  0.65 * Only Recipients without pre-transplant DM were included in the analysis. Donor-recipient pairs dropped out of this analysis dues to missing recipient BMI (n=23), donor BMI (n=28), recipient sex (n=3), donor sex (n=6), recipient age (n=3), and donor age (n=8).