**High Plasma Branched-Chain Amino Acids Are Associated with Higher Risk of Post-Transplant Diabetes Mellitus in Renal Transplant Recipients**

**Maryse C. J. Osté 1,\*, Jose L. Flores-Guerrero 1, Eke G. Gruppen 2, Lyanne M. Kieneker 1, Margery A. Connelly 3, James D. Otvos 3, Robin P. F. Dullaart <sup>2</sup> and Stephan J. L. Bakker <sup>1</sup>**


Received: 18 November 2019; Accepted: 11 February 2020; Published: 13 February 2020

**Abstract:** Post-transplant diabetes mellitus (PTDM) is a serious complication in renal transplant recipients. Branched-chain amino acids (BCAAs) are involved in the pathogenesis of insulin resistance. We determined the association of plasma BCAAs with PTDM and included adult renal transplant recipients (≥18 y) with a functioning graft for ≥1 year in this cross-sectional cohort study with prospective follow-up. Plasma BCAAs were measured in 518 subjects using nuclear magnetic resonance spectroscopy. We excluded subjects with a history of diabetes, leaving 368 non-diabetic renal transplant recipients eligible for analyses. Cox proportional hazards analyses were used to assess the association of BCAAs with the development of PTDM. Mean age was 51.1 ± 13.6 y (53.6% men) and plasma BCAA was 377.6 ± 82.5 μM. During median follow-up of 5.3 (IQR, 4.2–6.0) y, 38 (9.8%) patients developed PTDM. BCAAs were associated with a higher risk of developing PTDM (HR: 1.43, 95% CI 1.08–1.89) per SD change (*p* = 0.01), independent of age and sex. Adjustment for other potential confounders did not significantly change this association, although adjustment for HbA1c eliminated it. The association was mediated to a considerable extent (53%) by HbA1c. The association was also modified by HbA1c; BCAAs were only associated with renal transplant recipients without prediabetes (HbA1c < 5.7%). In conclusion, high concentrations of plasma BCAAs are associated with developing PTDM in renal transplant recipients. Alterations in BCAAs may represent an early predictive biomarker for PTDM.

**Keywords:** branched chain amino acids; post-transplant diabetes mellitus; biomarker; renal transplant recipients

### **1. Introduction**

Post-transplant diabetes after transplantation (PTDM), often a result of insulin resistance and deficient insulin production [1], is a serious complication in renal transplant recipients [2]. PTDM develops in 10–20% of renal transplant recipients during the first year post-transplantation [3], although some studies reported incidences of up to 50% [4]. PTDM is an important risk factor for cardiovascular disease (CVD) and infections, contributing to impaired graft and patient survival [5–7]. Previous studies have shown that PTDM is an important risk factor for premature mortality in renal transplant recipients [8–10]. Since the main cause of death in these patients is cardiovascular-related [11–13]

and most of these patients die with a properly functioning graft [14], it is clinically relevant to identify patients that are at high risk of developing PTDM.

Non-modifiable risk factors for the development of PTDM are patient's age, race, genetic background, and family history of diabetes. On the other hand, modifiable risk factors include overweight and obesity, but also immunosuppressive medication, such as steroids and calcineurin inhibitors [5]. During the era of cyclosporine-based regimens, the largest number of incident cases of PTDM occurred beyond the first year after transplantation [2] and with current tacrolimus-based regimens, the number of incident cases of PTDM beyond the first year after transplantation is even higher [15]. Furthermore, it should be noted that if a renal transplant recipient is diagnosed with diabetes, it is considered PTDM, irrespective whether this occurs one-year post-transplantation or 10 years later [16].

Current research is focused on a better understanding of risk factors responsible for the development of PTDM. Branched chain amino acids (BCAAs), a group of three essential amino acids (i.e., valine, leucine, and isoleucine), can be obtained from diet and comprise about 15–25% of total protein intake [17]. BCAAs not only play an important role in protein metabolism, but also have metabolic functions [18]. They may stimulate protein synthesis and influence glucose homeostasis [19–21]. It is known that circulating concentrations of BCAAs are elevated in subjects with prediabetes, type 2 diabetes, metabolic syndrome, and obesity [20,22,23]. Previous studies have assessed the association of BCAAs with insulin resistance and development of type 2 diabetes [24,25]. Recently, we reported that high concentrations of BCAAs are associated with an increased risk of developing type 2 diabetes in a large prospective cohort study in the general population [26]. High concentrations of BCAAs might be the consequence of excess dietary consumption, dysbiosis of the gut microbiota, and reduced breakdown of BCAAs in skeletal muscle and adipose tissue [20,23,27,28]. It has been found that kidney transplant recipients suffer from dysbiosis of gut microbiota [29,30]. Chronic use of immunosuppressive medication, including glucocorticoids and calcineurin inhibitors, by kidney transplant recipients is moreover likely to influence skeletal muscle metabolism [31,32]. These factors could alter BCAAs in kidney transplant recipients compared to the general population and thereby contribute to development of diabetes.

Previous population-based studies have shown inverse associations between BCAAs and all-cause mortality [33–35]. Furthermore, it is known that disturbances in amino acid metabolism, particularly involving BCAAs, occur in patients with end-stage renal disease [36]. A previous study showed that levels of valine and leucine, but not isoleucine, were significantly lower in patients with stage I and II chronic kidney when compared to controls [37].

Whether BCAA plasma concentrations are associated with development of PTDM in renal transplant recipients has not yet been established. Therefore, we hypothesized that higher plasma BCAA concentrations are associated with a higher risk of developing PTDM in renal transplant recipients, as a primary endpoint. Furthermore, secondary endpoints of this study were all-cause mortality and death-censored graft failure, because these endpoints could potentially compete with development of PTDM as an endpoint.

#### **2. Materials and Methods**

#### *2.1. Study Design and Population*

In this large cross-sectional study with prospective follow-up, we included stable adult renal transplant recipients (≥18 y) with a functioning graft for at least one year after transplantation (i.e., on maintenance immunosuppression and with a stable renal function), therefore excluding patients with transient hyperglycemia post-transplantation. Between November 2008 and May 2011, patients who visited the outpatient clinic of the University Medical Center Groningen were invited to participate. Both subjects with known or apparent systemic diseases (i.e., malignancies, opportunistic infections) and subjects with a history of alcohol and/or drug addiction were excluded

from participation. Informed consent was given by 707 (86.5%) of 817 initially invited patients. We excluded patients with missing data on BCAA, resulting in 518 renal transplant recipients eligible for analyses. As previously described, we recorded age, sex, body composition and eGFR of the renal transplant recipients who did not consent [38]. Compared with participating renal transplant recipients, those who did not consent were slightly older (mean age ± SD, 58 ± 13 years versus 53 ± 13 years) and had lower eGFR (47 <sup>±</sup> 19 ml/min per 1.73 m2 versus 51 <sup>±</sup> 20 ml/min per 1.73 m2) [38]. For the analyses with PTDM, we also excluded patients with diabetes or a history of diabetes at baseline (*n* = 132). Of these 132 renal transplant recipients, 34 were diagnosed with diabetes before renal transplantation and 98 developed PTDM between time of transplantation and baseline, leaving 386 renal transplant recipients eligible for analyses (Supplementary Figure S1). The study protocol was approved by the institutional research board (METc 2008/186), which adheres to the Principles of the Declaration of Helsinki.

#### *2.2. Data Collection*

During a morning visit to the outpatient clinic, all baseline data were collected as described previously [39]. Body weight and height were measured. Body Mass Index (BMI) was calculated as weight in kilograms divided by height in meters squared (kg/m2). Systolic and diastolic blood pressure and heart rate were measured every minute for 15 minutes in a half-sitting position using a semi-automatic device (Dinamap®1846; Critikon, Tampa, FL, USA) to prevent white coat effects [40]. The average of the last three measurements was taken as blood pressure value. Information on medication was derived from patient records, whereas information on smoking behavior was obtained by questionnaire. Information on physical activity was obtained using the reliable and valid Short Questionnaire to Assess Health enhancing physical activity (SQUASH) score in time multiplied by intensity [41]. Alcohol consumption and total energy intake were measured using a reproducible, validated food frequency questionnaire (FFQ) [42], which consisted of 177 items and was updated several times. Blood samples were taken after an 8–12 h overnight fasting period in the morning after completion of 24 h urine collection. Renal transplant recipients were instructed to assure adequate urine collection. They were instructed to discard their first morning urine specimen and then collect their urine for the next 24 h, including the next morning's first specimen the day of their visit. Protein intake was measured using 24 h urinary urea excretion. Estimated Glomerular Filtration Rate (eGFR) was calculated using the serum creatinine-based Chronic Kidney Disease Epidemiology Collaboration equation [43].

#### *2.3. Quantification of BCAAs*

Plasma BCAA concentrations were measured at baseline using a Vantera Clinical Analyzer (LabCorp, Morrisville, NC, USA), a fully automated, high-throughput, 400 MHz proton (1H) nuclear magnetic resonance spectroscopy (NMR) platform. Quantification of BCAAs by NMR was validated and had been previously described in detail elsewhere [25,44]. The within-laboratory (inter-assay) and within-run (intra-assay) imprecision for NMR-measured BCAAs are described in detail previously [44]. For total BCAAs, the coefficients of variation for inter-assay and intra-assay were 1.8–6.0% and 2.1–4.4%, respectively.

#### *2.4. Clinical Endpoints*

The primary outcome of this study was PTDM, which was defined as at least one of the following criteria: symptoms of diabetes (e.g., polyuria, polydipsia, unexplained weight loss) plus a non-fasting plasma glucose concentration of ≥200mg/dL (11.1 mmol/L); fasting plasma glucose concentration (FPG) ≥126 mg/dL (7.0 mmol/L); start of antidiabetic medication; or HbA1c ≥6.5% (48 mmol/L). This definition was according to the American Diabetes Association criteria for diabetes [45], including HbA1c levels as proposed by the International Expert Panel of the international consensus meeting on PTDM [46]. The secondary outcomes of this study were all-cause mortality and death-censored graft failure. Death-censored graft failure was defined as return to hemodialysis treatment or retransplantation. All subjects received medical care at the University Medical Center Groningen alone or medical care shared with a secondary referral hospital. In accordance with the KDIGO guideline for renal transplant recipients, follow-up visits after the first year post transplantation were performed every 3 months [47]. Data on PTDM, all-cause mortality and death-censored graft failure were retrieved from patient files and verified with the corresponding nephrologist or the Municipal Personal Records Database in case of death. Endpoints were recorded until the end of September 2015. Since the outpatient program uses continuous surveillance systems, it guarantees correct and up-to-date information on patient status. No participants were lost to follow-up.

#### *2.5. Statisical Analyses*

Normal distributed data were presented as mean and standard deviation, whereas skewed distributed data were expressed as median and interquartile range. Categorical data were presented as number and percentage. Differences between diabetic and non-diabetic renal transplant recipients were compared using unpaired t-tests for normally distributed variables, Mann–Whitney U tests for skewed distributed variables and Chi-square tests for categorical variables. Differences between tertiles of total BCAA were compared using one-way ANOVA tests for normally distributed variables, Kruskal–Wallis tests for skewed distributed variables and Chi-square tests for categorical variables. Skewed distributed data were log-transformed when appropriate. Correlations between BCAAs and total energy intake, protein intake, physical activity, and HbA1c in non-diabetic renal transplant recipients were assessed using Pearson correlation coefficients.

Kaplan–Meier curves were plotted for the development of PTDM according to the highest tertile versus the two lowest tertiles of total BCAA. We performed crude and multivariable Cox proportional hazards regression analyses to assess the association of total BCAA with the development of PTDM. First, we performed crude analyses and analyses adjusted for age and sex (model 1). We further cumulatively adjusted for renal function parameters (eGFR, proteinuria, and time between transplantation and baseline) in model 2. To prevent overfitting by including too many covariates in relation to number of events [48], we adjusted for other potential confounders in additional models based on model 2. We additionally adjusted for total cholesterol and triglycerides in model 3; total energy intake, physical activity, and BMI in model 4; smoking status and alcohol consumption in model 5; prednisolone dose and trough levels of tacrolimus and cyclosporine in model 6. Total BCAA per 1 standard deviation (SD) was used as continuous variable, but also as categorical variable (highest tertile versus two lowest tertiles). Patients were censored at date of last follow-up or death. Hazards ratios and 95% CIs were given for the Cox proportional hazards analyses. Schoenfeld residuals of the BCAAs were checked and tested in STATA using the proportional hazard test by Grambsch and Therneau [49]. Furthermore, penalized splines analyses performed in R were used to visualize the association of total BCAA with the development of PTDM, adjusted for age and sex. Additionally, we evaluated potential effect modification by age, gender, BMI, eGFR and HbA1c by entering both main effects and the cross-product term in the crude model. When effect modification was observed, we proceeded with stratified analyses, with a HbA1c of 5.7–6.4% considered as prediabetic state [50,51].

In further analyses, we investigated whether plasma glucose and HbA1c could serve as mediator in the association of BCAAs and risk of PTDM. To investigate potential mediation, we performed mediation analyses using the mediation package of R [52], by which we tested significance and magnitude of mediation (see the Supplementary Materials for a detailed description). Competing risks occur when patients can experience or develop one or more events which compete with the outcome of interest [53]. To rule out competing risk of all-cause mortality with the development of PTDM, we performed competing risk analyses according to Fine and Gray [54]. We performed crude and multivariable Cox proportional hazards regression analyses to assess the association of total BCAA with the secondary outcomes all-cause mortality and death-censored graft failure. Total BCAA per 1 standard deviation (SD) was used as continuous variable, but also as categorical variable (lowest

tertile versus two highest tertiles). For the association of BCAAs with both all-cause mortality and death-censored graft failure, we evaluated the potential effect modification by diabetes.

A two-sided *p*-value of <0.05 was considered statistically significant. The main statistical analyses for the manuscript were performed using IBM Statistics SPSS version 23.0 (IBM Inc, Chicago, IL, USA). We used STATA version 11.0 (StataCorp LP, College Station, TX, USA) to check and test the Schoenfeld residuals by performing the proportional hazard test according to Grambsch and Therneau. We used R version 3.2.3 (R Foundation for Statistical Computing, Vienna, Austria) to perform penalized splines analyses and to perform mediation analyses. We used GraphPad Prism 5 (GraphPad Software Inc., La Jolla, CA, USA) to visualize the Kaplan–Meier curves for the development of PTDM.

#### **3. Results**

#### *3.1. Patient Characteristics in Whole Cohort (n* = *518)*

Mean age of overall renal transplant recipients was 52.7 ± 13.0 y and 53.7% of the participants were men. Median time between baseline measurements and transplantation was 5.0 (IQR, 1.7–11.9) years. Diabetic renal transplant recipients had significantly higher plasma concentrations of total BCAA (424.6 ± 97.9 μM) when compared with non-diabetic renal transplant recipients (377.6 ± 82.5 μM). Baseline characteristics of the overall (*n* = 518), diabetic (*n* = 132) and non-diabetic (*n* = 386) population are shown in Table 1. Non-diabetic subjects were younger, had a lower weight and BMI, had a higher physical activity score and lower heart rate when compared with diabetic RTR. Furthermore, non-diabetic subjects had lower plasma glucose, HbA1c, and triglycerides, and higher HDL cholesterol concentrations. No differences were seen in medication, except for use of statins, which was more common in the diabetic renal transplant recipients than in the non-diabetic renal transplant recipients.




**Table 1.** *Cont*.

Data are represented as mean ± SD, median (interquartile range) or *n* (%). Differences were tested by unpaired T-test or Mann–Whitney U test for continuous variables and with χ 2-test for categorical variables. RTR, renal transplant recipients; SBP, systolic blood pressure; DBP, diastolic blood pressure; eGFR, estimated glomerular filtration rate; HbA1c, hemoglobin A1c; LP-IR, lipoprotein insulin resistance index; HDL, high-density lipoprotein; LDL, low-density lipoprotein; BCAA, branched-chain amino acids.

#### *3.2. Patient Characteristics in Subgroup of Non-Diabetic Renal Transplant Recipients (n* = *386)*

For the non-diabetic renal transplant recipients the baseline characteristics according to tertiles of total BCAAs are presented in Table 2. Subjects in the highest tertile of total BCAA were more often male, consumed more alcohol, had a lower heart rate, and a lower HDL cholesterol when compared with subjects in the lowest tertile. There were no differences in transplant characteristics, renal allograft function, and glucose homeostasis. Furthermore, we found that total BCAAs were positively correlated with protein intake (*r* = 0.25, *p* = <0.001) and HbA1c (*r* = 0.12, *p* = 0.02), but not with total energy intake (*r* = –0.01, *p* = 0.82) and physical activity (*r* = 0.10, *p* = 0.06). When we divided the non-diabetic renal transplant recipients in patients with prediabetes (HbA1c ≥ 5.7%) and without prediabetes (HbA1c < 5.7%), we found a positive correlation in the prediabetic renal transplant recipients (*r* = 0.23, *p* = 0.002), but not in renal transplant recipients without prediabetes (*r* = 0.003, *p* = 0.96).





**Table 2.** *Cont*.

#### *3.3. BCAAs and Risk of Developing PTDM*

In the subgroup of non-diabetic renal transplant recipients at baseline (*n* = 386) during a median follow-up of 5.3 (IQR, 4.2–6.0) y, 38 (9.8%) subjects developed PTDM. Of the renal transplant recipients in the highest tertile of total BCAA 17.3% developed PTDM versus 8.0% in the lowest two tertiles (*p* = 0.02). The Kaplan–Meier curves for the development of PTDM according to the highest tertile versus the two lowest tertiles of total BCAA is shown in Figure 1.

**Figure 1.** Kaplan–Meier curves for the development of post-transplant diabetes mellitus (PTDM) according to the highest tertile versus the two lowest tertile of total branched chain amino acids (BCAA) in renal transplant recipients.

Cox regression analyses with total BCAA per standard deviation (SD) as continuous variable showed that higher total BCAA was associated with a higher risk of developing PTDM (HR: 1.43, 95% CI 1.08–1.89, *p* = 0.01), independent of age and sex (Table 3, model 1). After adjustment for other potential confounders, including renal function parameters, lipids, dietary and lifestyle factors, and use of medication the association did not materially change (Table 3, model 2–6). In additional Cox regression analyses with total BCAA divided in the highest tertile versus the two lower tertiles, total BCAA was again significantly associated with development of PTDM, independent of age and sex (HR: 2.07; 95% CI 1.07–3.99, *p* = 0.03). Further adjustment for potential confounders did not change the association (Table 3, model 2–6). To illustrate the association of total BCAA with development of PTDM, an age and sex adjusted penalized spline is shown in Figure 2. We found no significant effect modification by age (*p*interaction = 0.75), gender (*p*interaction = 0.17), BMI (*p*interaction = 0.31), and eGFR (*p*interaction = 0.50) in the association of total BCAA per SD with PTDM, but we did for HbA1c (*p*interaction = 0.02). We continued with stratified analyses (Supplementary Figure S2). BCAAs were associated with PTDM in renal transplant recipients without prediabetes (HbA1c < 5.7%), but not in renal transplant recipients with prediabetes (HbA1c ≥ 5.7%), independent of age, sex, eGFR, proteinuria, and time since transplantation.


**Table 3.** Association of plasma BCAAs with post-transplant diabetes mellitus in renal transplant recipients (*n* = 386).

Cox proportional hazards regression analyses were performed to assess the association of BCAAs with PTDM. Model 1: adjustment for age and sex; model 2: model 1 + adjustment for eGFR, proteinuria, and time since transplantation; model 3: model 2 + adjustment for total cholesterol and triglycerides; model 4: model 2 + adjustment for total energy intake, physical activity, and BMI; model 5: model 2 + adjustment for smoking status and alcohol intake; model 6: model 2 + adjustment for prednisolone dose and trough levels of tacrolimus and cyclosporine;. BCAA, branched chain amino acids; PTDM, post-transplant diabetes mellitus; eGFR, estimated glomerular filtration rate.

**Figure 2.** Association between plasma branched chain amino acids (BCAA) and post-transplant diabetes mellitus (PTDM) in 386 renal transplant recipients. Data were fit by a Cox regression model based on penalized splines and adjusted for age and sex. The gray area represents the 95% confidence interval.

#### *3.4. Secondary Analyses*

In mediation analyses, we found that HbA1c mediated 53% of the association between BCAAs and PTDM in renal transplant recipients, whereas plasma glucose was not a significant mediator in this association (Supplementary Table S1), after adjustment for age and sex. The results of competing risk analyses did not materially differ from those with Cox regression for the association of total BCAA per SD as continuous variable and development of PTDM (HR: 1.44, 95% CI 1.08–1.92, *p* = 0.01), adjusted for age and sex (Table 3, model 2 for comparison). Also, the analysis with total BCAA divided

in the highest tertile versus the two lower tertiles did not differ in the competing risk analysis (HR: 2.09, 95% CI 1.10–3.96, *p* = 0.02), adjusted for age and sex (Table 3, model 2 for comparison).

#### *3.5. BCAAs and Risk of All-Cause Mortality and Death-Censored Graft Failure*

In the total population of both diabetic and non-diabeticrenal transplant recipients at baseline (*n* = 518), 114 (22.0%) subjects died during a median follow-up of 5.4 (IQR, 4.7–6.2) y, whereas 65 (12.5%) subjects developed graft failure during a median follow-up of 5.3 (IQR, 4.5–6.0) y. There was no significant association between total BCAA and the individual BCAAs with all-cause mortality and death-censored graft failure (Supplementary Table S2). We found no effect modification by diabetes for the association of total BCAA with all-cause mortality (*p*interaction = 0.22) and death-censored graft failure (*p*interaction = 0.41).

#### **4. Discussion**

In this large cross-sectional study with prospective follow-up, higher concentrations of total BCAAs are associated with a higher risk of developing PTDM in renal transplant recipients. This association did not change after adjustment for relevant confounders, including age, sex, renal function parameters, lipids, dietary and lifestyle factors, and use of immunosuppressive medication. Subsequently, this association was modified by HbA1c; total BCAAs were significantly associated with PTDM in renal transplant recipients without prediabetes (HbA1c < 5.7%), but not in renal transplant recipients with prediabetes (HbA1c ≥ 5.7%). Furthermore, we show that the association between total BCAA and PTDM was mediated to a considerable extent (53%) by HbA1c. In addition, no association of total BCAAs with all-cause mortality and death-censored graft failure in renal transplant recipients was found.

It is known that BCAAs are elevated in subjects with prediabetes, type 2 diabetes, and obesity [22]. In this cohort, BCAA concentrations were elevated in diabetic renal transplant recipients when compared with non-diabetic renal transplant recipients (424.6 ± 97.9 μM vs. 377.6 ± 82.5 μM, respectively), as observed in previous studies in the general population [23,55,56]. The BCAA concentrations of the diabetic renal transplant recipients can be compared to the BCAA concentrations of 439 ± 95 μM in patients with type 2 diabetes mellitus in the general population [25]. The BCAA concentrations of the non-diabetic renal transplant recipients are comparable to the mean plasma levels of 370.3 ± 88.6 μM in a large prospective population-based cohort study [26].

BCAAs are a group of essential amino acids, comprising valine, leucine, and isoleucine, and can only be obtained from diet. They comprise about 15–25% of total protein intake [17]. Previous studies have shown that plasma BCAA levels are modifiable by a higher or lower consumption of protein. Prior work showed that higher consumption of BCAAs is significantly associated with higher plasma levels of BCAAs [57]. The correlation was moderate, but comparable to other diet-plasma biomarker correlations. It has been shown that dietary protein reduction lowers serum levels of BCAAs [58]. Recently, a randomized controlled crossover trial even showed that short term dietary reduction of BCAAs decreases postprandial insulin secretion [59]. It is known, that around 80% of dietary BCAAs reach the blood circulation [60], but circulating plasma levels of BCAAs can also be affected by their catabolism [61]. The initial site of the BCAA metabolism is skeletal muscle, because of the high branched-chain-amino-acid aminotransferase (BCAT) activity in the muscle [62]. This metabolism is sensitive to changes in the amount and composition of food. A high protein diet leads to higher concentrations of BCAAs, whereas a low protein diet lowers the plasma BCAA concentrations [62,63]. Indeed, in our study, subjects in the highest tertile of total BCAA had a higher 24 h urinary urea excretion, which is an objective measurement for total protein intake, when compared to subjects in the lowest tertile of total BCAA.

The results of the prospective analysis with PTDM are consistent with previous studies that reported the association of BCAAs with type 2 diabetes in the general population [21,22,64]. Recently, we showed in a prospective cohort study that high concentrations of BCAAs are associated with

increased risk of developing type 2 diabetes [26]. The fact that total BCAAs were significantly associated with PTDM in subjects without prediabetes, but not in prediabetic subjects suggest that alterations in total BCAAs might be an early signal of deterioration of glycemic control. A previous study has shown that elevated BCAAs levels may appear long before other markers of insulin resistance become abnormal [21]. Elevations in circulating BCAAs can occur before any alterations in insulin action are detectable. Moreover, it has been reported that plasma BCAAs might serve as a better indicator of impaired insulin resistance when compared to plasma glucose levels [65], since in patients without prediabetes the metabolic status is not deteriorated enough to alter plasma glucose levels.

The secondary outcomes, all-cause mortality and death-censored graft failure, were not associated with total BCAAs. This is in contrast to a previous study that showed an inverse association of total BCAAs and death in patients at risk for coronary artery disease [33], supporting the underlying malnutrition-inflammation syndrome hypothesis. Furthermore, the large Estonian biobank study also observed inverse associations between BCAAs and all-cause mortality [34]. Moreover, in the ADVANCE study including individuals with type 2 diabetes, low levels of leucine and valine were associated with increased all-cause mortality [35]. In our study 114/518 (22.0%) renal transplant recipients died during a median follow-up of 5.4 years, resulting in a death rate of 4.1% per year, which is slightly higher when compared to the FAVORIT trial, a large multi-center double-blind randomized controlled trial in 4110 stable renal transplant recipients (age 52 ± 9.4 years, 62.8% male at 5 years after transplantation) in which 493/4110 (12.0%) subjects died within 4.0 years of follow-up, resulting in a death rate of 3.0% per year [66]. Prior work showed that levels of valine and leucine, but not isoleucine, were significantly lower in patients with stage I and II chronic kidney disease, when compared with controls [37], suggesting potential use as a biomarker for renal dysfunction.

Currently, there are several potential mechanisms that could explain the contribution of BCAAs to the development of insulin resistance, type 2 diabetes, and PTDM, although these mechanisms are not completely understood. One mechanism proposes that BCAAs interfere with insulin signaling through activation of the mammalian target of rapamycin complex 1 (mTORC1) in skeletal muscle and serine phosphorylation of insulin receptor substrate 1 and 2, which promotes insulin resistance and can lead to the development of type 2 diabetes [20]. However, conflicting results regarding the role of BCAAs to elicit insulin resistance have been reported [67] and do question whether mTORC1 activation is sufficient or necessary in the development of insulin resistance. Others assume that BCAA dysmetabolism, especially in obesity, contributes to a rise in BCAAs, which results in accumulation of potential toxic BCAA metabolites, which could induce cellular damage [20]. These BCAA metabolites might lead to mitochondrial dysfunction and β-cell apoptosis, which is common in insulin resistance and type 2 diabetes [20]. Nevertheless, the association of total BCAAs and PTDM in our study was independent of BMI. Moreover, a previous study in the general population showed that the association of BCAAs with insulin resistance was independent of leptin and adiponectin, both valid biomarkers of adipose tissue dysfunction, when taking BMI into account [25], suggesting the association is presumably mainly driven by another mechanism.

To the best of our knowledge, this is the first study that studied the association of BCAAs with the development of PTDM in renal transplant recipients. Strengths of this study include the complete follow-up and use of clinical endpoints (PTDM, all-cause mortality, and death-censored graft failure), which are relevant in daily clinical practice. Furthermore, this study had a considerable follow-up period. Another strength is the use of stable patients who had a functioning graft for at least 1 year, which resulted in exclusion of patients with transient hyperglycemia post-transplantation, which occurs frequently and is evident in about 90% of renal transplant recipients in the first few weeks post-transplantation [68,69]. Hyperglycemia can also occur as a result of rejection therapy, infections, and other critical conditions [46]. Therefore, it is important to diagnose PTDM in stable patients (i.e., on maintenance immunosuppression, stable renal function and in absence of acute infections) [46]. This study also has several limitations. First, it is a single-center study, with a study population mainly consisting of Caucasians. As ethnicity is an independent risk factor for developing PTDM [5],

it is important to repeat this study in more diverse populations. Renal transplant recipients in the current study were included at a median of 5.0 years after transplantation. Therefore, extrapolating our results to patients in early stages after renal transplantation should be done with caution. The clinical significance or impact of the diagnosis of PTDM early or late after transplantation has yet to be determined [16]. In addition, age at time of transplantation in our cohort is lower when compared to other European cohorts [70]. Moreover, the prevalence of living donor grafts is higher in the Netherlands [71], which might also contribute to a lower age in our cohort, since younger subjects have a broader social network and therefore likely a higher chance of finding a compatible living donor at younger age. Furthermore, it is known that oral glucose tolerance tests (OGTTs) are the gold standard diagnostic tool to diagnose PTDM. Unfortunately, OGTTs were not performed, but recently it has been shown that the combined use of fasting plasma glucose and HbA1c criteria appears to be a diagnostic strategy for PTDM in stable renal transplant recipients [72]. Another limitation of this study is that only 38 subjects developed PTDM during follow-up, which led to a lack of power. Unfortunately, a comorbidity index was not available in our cohort and we do not have data on weight gain post-transplantation. It cannot be excluded that change of weight could serve as a source of bias and could spuriously strengthen or weaken the association of BCAAs with development of PTDM. Finally, longer-term intervention studies are required to determine whether BCAAs are causally related to the development of diabetes mellitus or merely act as markers of underlying pathophysiology.

#### **5. Conclusions**

In conclusion, this single-center cross-sectional assessment of BCAA in stable renal transplant recipients showed that high plasma concentrations of total BCAA are associated with a higher risk of developing PTDM during prospective follow-up. Alterations in BCAA levels might be an early signal of deterioration of glycemic control in renal transplant recipients. Further research is needed to investigate the possible mechanism/role of BCAAs in the development of post-transplant diabetes.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2077-0383/9/2/511/s1, Figure S1: Flowchart of the study, Figure S2: Stratified analyses of the association of branched chain amino acids (BCAA) and post-transplant diabetes mellitus (PTDM) in both patients with normal glucose tolerance (HbA1c < 5.7%) and prediabetes (HbA1c ≥ 5.7%) adjusted for age, sex, eGFR, proteinuria, and time since transplantation, Figure S3: Mediation analysis on the association of branched chain amino acids (BCAA) on post-transplant diabetes mellitus (PTDM), Table S1: Mediating effect of HbA1c on the association of BCAA with PTDM, Table S2: Association of BCAAs with all-cause mortality and death-censored graft failure in renal transplant recipients (*n* = 518).

**Author Contributions:** Methodology, M.C.J.O., J.L.F.-G., E.G.G., L.M.K., M.A.C., J.D.O., R.P.F.D., and S.J.L.B.; Formal analysis, M.C.J.O., J.L.F.-G., S.J.L.B., and R.P.F.D.; Resources, M.A.C. and J.D.O.; Data curation, M.C.J.O., J.L.F.-G., R.P.F.D., and S.J.L.B.; Writing—original draft preparation, M.C.J.O., J.L.F.-G., R.P.F.D. and S.J.L.B.; Writing—review and editing, M.C.J.O., J.L.F.-G., E.G.G., L.M.K., M.A.C., J.D.O., R.P.F.D. and S.J.L.B.; Supervision, S.J.LB. and R.P.F.D., Funding acquisition, S.J.L.B. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by a grant from the Dutch Top Institute Food and Nutrition (A-1003).

**Acknowledgments:** We kindly thank Else van den Berg for her contribution to patient inclusion. The cohort on which the study was based is registered at clinicaltrials.gov as "TransplantLines Food and Nutrition Biobank and Cohort Study (TxL-FN)" with number NCT02811835.

**Conflicts of Interest:** M.A.C. and J.D.O. are employees of LabCorp. All other authors of this manuscript declare no conflicts of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **The E**ff**ect of Proton Pump Inhibitor Use on Renal Function in Kidney Transplanted Patients**

### **Dominik J. G. Flothow, Barbara Suwelack, Hermann Pavenstädt, Katharina Schütte-Nütgen** † **and Stefan Reuter \*,**†

Department of Medicine D, University Hospital of Muenster, 48149 Münster, Germany;

d\_flot01@uni-muenster.de (D.J.G.F.); barbara.suwelack@ukmuenster.de (B.S.);

hermann.pavenstaedt@ukmuenster.de (H.P.); Katharina.schuette-nuetgen@ukmuenster.de (K.S.-N.)

**\*** Correspondence: sreuter@uni-muenster.de; Tel.: +49-251-83-47540

† These authors contributed to this work equally.

Received: 31 December 2019; Accepted: 15 January 2020; Published: 18 January 2020

**Abstract:** Recently, proton pump inhibitor (PPI) intake has been linked to acute kidney injury and chronic kidney disease. The objective of this study was to assess the effect of PPIs on renal function and rejection rate in kidney transplant patients. We performed a single center, retrospective analysis of 455 patients who received a kidney transplant between May 2010 and July 2015. Median follow-up time was 3.3 years. PPI prescription was assessed in half-year intervals. Primary outcome parameters were the estimated glomerular filtration rate (eGFR), change in the eGFR, and >30% and >50% eGFR decline for different time periods (up to four years post-transplantation). Our secondary outcome parameter was occurrence of biopsy proven acute rejection (BPAR) in the first two years after transplantation. Except for >30% eGFR decline from half a year to two years post-transplantation (*p* = 0.044) and change in the eGFR, >30% and >50% eGFR decline showed no association with PPI intake in our patient cohort (*p* > 0.05). Similarly, by analyzing 158 rejection episodes, BPAR showed no correspondence with mean daily PPI intake. We conclude that prolonged PPI intake has no relevant adverse effect on kidney transplant function or rejection rates. Polypharmacy, however, remains a problem in renal transplant recipients and it is thus advisable to question the necessity of PPI prescriptions when clear indications are missing.

**Keywords:** proton pump inhibitor; kidney transplantation; transplant rejection; GFR

#### **1. Introduction**

With only a single transplanted kidney and oftentimes reduced renal filtration rates, kidney transplant (KTx) recipients are particularly vulnerable to the nephrotoxic adverse effects of drugs. Care is taken to avoid such drugs that could further impair kidney function. For this reason, recent epidemiological studies that have observed a relationship between acute kidney injury (AKI), chronic kidney disease (CKD) and proton pump inhibitor (PPI) intake have been of special interest for practitioners involved in the care of KTx patients [1–5].

Furthermore, in the setting of KTx, two medication interactions of possible relevance are the interaction between PPIs and mycophenolate mofetil (MMF) and between PPIs and tacrolimus. The first may lead to decreased blood levels of the active metabolites of MMF [6–12], which may result in increased rejection rates [13–16]. Tacrolimus is known to be nephrotoxic and it is thought that interactions with PPIs may change its uptake and/or metabolism [17–19], potentially increasing tacrolimus blood concentration. This could be detrimental to kidney transplant function.

In the face of surgical stress and long-term polypharmacy, upper gastrointestinal symptoms are frequent in KTx patients [20]. Even in those patients receiving acid suppressive therapy, the risk of ulcer disease is still elevated [20]. Gastrointestinal complications have also been associated with decreased graft survival [21]. As PPIs are a very effective form of acid suppression, they are frequently given as prophylaxes among transplant recipients [20,22,23]. In our center, they are the standard of care for KTx patients.

In light of the mentioned studies and the possibility of adverse drug interactions between PPIs and mycophenolate mofetil or tacrolimus, we carried out this study. Previous PPI studies in KTx patients have focused on the possible interaction of MMF with PPIs [22,24,25] or the occurrence of other PPI intake related adverse events [23]. We retrospectively evaluated if a relationship between PPI intake and renal function could be found. We analyzed follow-up data of up to four years after transplantation. To our knowledge, no study exists to date that has specifically analyzed the changes in renal function after KTx with regard to PPI intake over a comparable time frame. Additionally, we compared rejection rates, as these may be of relevance regarding MMF and PPI interactions.

#### **2. Patients and Methods**

#### *2.1. Patients*

Prior to analysis, the data of all patients was anonymized. The local ethics committee (Ethik Kommission der Ärztekammer Westfalen-Lippe und der Medizinischen Fakultät der Westfälischen Wilhelms-Universität, No. 2014-381-f-N) approved the study. Methods in this study were carried out in accordance with the current transplantation guidelines and the Declarations of Istanbul and Helsinki. Written informed consent was given by all participants at the time of transplantation for recording their clinical data.

We herein performed an explorative, retrospective, single-center cohort study. We enrolled all patients receiving a KTx between May 2010 and July 2015 at the University Hospital Muenster. Inclusion criteria were patient age ≥18 years and PPI therapy at primary hospital discharge post-transplantation (pTx). Patients receiving multiple organ transplants remained included. The recipient and donor data was collected from the patients' electronic files. The following data was collected and used: recipient and donor age and sex; recipient body mass index (BMI); prior renal transplants; transplant under European Senior Program; donor type (living or deceased); delayed graft function (DGF; dialysis within the first week pTx); cold ischemia time; pre-transplant time dialyzed; pre-transplant arterial hypertension; pre-transplant diabetes; presence of peripheral arterial occlusive disease; cerebral arterial occlusive disease or stroke; coronary heart disease or myocardial infarction; anticoagulant prescription; antiplatelet drug prescription (including acetylsalicylic acid (ASA)); statin prescription; MMF dosage at primary discharge and two years pTx; cortisone intake at primary discharge and one year pTx; tacrolimus dose and blood level three months pTx; prior smoking history; continuation of smoking after KTx; number of human leukocyte antigen (HLA) mismatches; ABO incompatibility of transplant; induction therapy; pre-transplant donor specific antibody (DSA) occurrence; panel reactive antibodies; transplant rejection occurrence and type according to Banff criteria; eGFR from half a year to four years pTx; and data on PPI prescription. At primary discharge, none of the patients had any non-steroidal anti-inflammatory drug prescriptions, except for ASA. To improve adjustment for confounding through comorbidities, additional data was collected to calculate the Charlson comorbidity index [26] at transplantation [27].

The induction therapy was chosen according to the immunological risks of the patients. One gram of mycophenolate mofetil was given twice a day; the dosage was reduced in case of adverse events. Prednisolone was started at 500 mg intravenously (i.v.) before KTx, followed by 100 mg for three days; then reduced by 20 mg/day. A dosage of 20 mg/day was maintained until day 30 and then slowly reduced to 5 mg/day. Immunosuppressive maintenance therapy usually consisted of a calcineurin inhibitor (tacrolimus or cyclosporine A), mycophenolate sodium or mycophenolate mofetil and prednisolone.

#### *2.2. Proton Pump Inhibitors—Data Collection*

Data on the prescription of PPIs (agent and dose prescribed) were collected in half year intervals for all patients starting at primary discharge. PPI intake was assumed according to this prescription until the next interval. If data was not present for a certain interval, the prescription from the preceding interval was assumed.

Due to the favorable drug interaction profile of pantoprazole in comparison to other PPIs [28], pantoprazole is used as the PPI of choice at our center. Therefore, it has also been used as the standard PPI in this study. In order to also use data from different PPIs, equivalent doses were calculated for the two other agents (omeprazole and esomeprazole) used. We used the ratio: 40 mg pantoprazole =ˆ 20 mg omeprazole ˆ= 20 mg esomeprazole [29,30].

At our center, patients are instructed to ingest tacrolimus and MMF on an empty stomach one hour before intake of PPIs. Both tacrolimus and MMF are usually taken twice daily, while PPIs are mostly taken once per day.

#### *2.3. Group Formation*

Based on PPI intake (*n* = 363) and non-intake (*n* = 82) at half a year pTx, two patient groups were formed. These were used for a direct comparison of GFR and change thereof.

For the outcome measures >30% and >50% eGFR decline and the number of rejections, the groups 0 mg, 1–20 mg, 21–40 mg and >40 mg mean daily PPI intake were compared. The standard dose at our center is 40 mg pantoprazole, 20 mg is the common reduced dose, and above >40 mg (often 80 mg) is an elevated dose (rationale for the group formation).

#### *2.4. Outcome Measures*

Primary outcome measures were: the eGFR (at six months, one year, two years, three years and four years), change in the eGFR (from six months pTx to one year, two years, three years and four years), eGFR decline >30% and eGFR decline >50% (from six months to two years and two years to four years). All eGFR-values were calculated using the CKD-EPI (Chronic Kidney Disease Epidemiology Collaboration) equation [31].

Our secondary outcome was biopsy proven acute rejection (BPAR) in months one to six, seven to twelve and in the second year pTx. For each time period, every patient with a rejection was counted (not only a patient's first rejection). The usual indication for biopsy in our center is a rise in creatinine with no apparent cause.

#### *2.5. Statistical Analysis*

Statistical analysis was performed using IBM SPSS® Statistics 24 for Windows (IBM Corporation, Somers, NY, USA). Microsoft Excel was used for data collection, simple calculations, and graphing.

This is an explorative study and no adjustment was made for multiple testing. *p*-values ≤ 0.05 were regarded as statistically noticeable. Normally distributed continuous variables are displayed as mean ± standard deviation (SD), non-normally distributed as median and interquartile range (IQR), and categorial variables as frequencies and percentage of total. Pairwise comparisons of independent samples were performed using student's *t*-test or Welch's unequal variance *t*-test for normally distributed data and the Mann–Whitney U test was used for non-normally distributed data. For categorial variables, groups were compared using Fisher's exact test.

For the group comparison of the eGFR and change in the eGFR, the Mann–Whitney U test and multivariable linear regression were performed. Further information on model building and the included variables is found in Supplementary Materials 2A.

A possible relationship between >30% and >50% eGFR decline endpoints and mean daily PPI intake was investigated using Fischer's exact test and multivariable logistic regression analysis. Mean daily PPI intake was calculated by averaging the prescribed PPI dose at half year intervals up to half a year before the final relevant eGFR value. Information on model building and the herein included variables can be found in Supplementary Materials 2B.

Correlation of BPAR in months one to six, seven to twelve and in the second year pTx with mean daily PPI intake, respectively, was investigated univariably using Fischer's exact test and multivariably using logistic regression analysis with forward selection of confounders (same procedure as for >30% and >50% eGFR decline endpoints). A list of the variables included and further information on the testing can be found in Supplementary Materials 2C.

#### **3. Results**

#### *3.1. Patients*

A total of 511 patients were transplanted in the study period. The following exclusions were made: 37 patients as they were <18 years of age, 13 as they did not receive PPIs at primary discharge, five due to non-onset of graft function after transplantation and one due to death before primary discharge. A final number of 455 patients were included; the median follow-up time was 3.3 years (IQR, 2.2–4.9). In our cohort, 12 patients died and 10 experienced graft loss within the first year; seven died and eight graft losses occurred within the second year pTx. Most patients (96%) initially received pantoprazole as their PPI agent. Patient and donor characteristics are displayed in Table 1 (and Supplementary Materials 1).

**Table 1.** Patient characteristics at primary hospital discharge post-transplantation (additional information included in Supplementary Materials 1).


The two compared groups were formed based on PPI (proton pump inhibitor) intake (PPI Group) or non-intake (No PPI Group) at half a year post-transplantation. Results are presented as mean ± standard deviation (SD), median and interquartile range (IQR) or as absolute and relative frequencies. Abbreviations: BMI, body mass index; Tx, transplantation; MPS, enteric-coated mycophenolate sodium; MMF, mycophenolate mofetil; CCI, Charlson comorbidity index; HLA, human leukocyte antigen; ATG, Antithymocyte globulin; PRA, panel reactive antibodies. Along with HLA mismatch (*n* = 3) and Basiliximab induction (*n* = 7), four other variables have one patient with missing values.

Lost to follow-up rates in the PPI group and no PPI group were 16/363 (4.4%) and 2/82 (2.4%) at one year, 64/363 (17.6%) and 6/82 (7.3%) at two years, 152/363 (41.9%) and 28/82 (34.1%) at three years, 229/363 (63.1%) and 54/82 (65.9%) at four years, respectively.

#### *3.2. Group Comparison*

The two groups differed noticeably regarding recipient age, donor type (living or deceased), time dialyzed before transplantation, active smoking, smoking history, cerebral artery occlusive disease or stroke, coronary artery disease or myocardial infarction and Charlson index at the time of transplantation. These characteristics were favorable for the no PPI group. A noticeable difference was also found in the number of fast tacrolimus metabolizers three months pTx. (Table 1 and Supplementary Materials 1).

The PPI group showed significantly lower eGFR compared to the no PPI group at half a year, one year and two years pTx (*p* < 0.05) (multivariable). For the third and fourth year, the difference was not statistically significant (Supplementary Materials 3). The trend in the mean eGFR can be seen in Figure 1.

**Figure 1.** Trend in mean estimated glomerular filtration rate (eGFR). Mean eGFR is plotted against time in the two patient groups. Grouping is according to PPI intake or non-intake at half a year post-transplantation.

Regarding the change in the eGFR, both groups were similar over all time periods (Table 2).


**Table 2.** Results of the PPI group comparison of change in the eGFR.

Results of the comparison of the change in the eGFR, values between groups. The eGFR value measured at half a year is always used as the reference value and was subtracted from that of the later date. Groups were formed based on PPI intake or non-intake at half a year pTx (post-transplantation). For the linear regression models, the patient number is slightly reduced (<3 patients difference per test) due to missing covariables in a few patients. Abbreviation: CI, confidence interval.

#### *3.3.* >*30% and* >*50% eGFR Decline*

Multivariable logistic regression was only carried out for the outcome >30% eGFR decline from half a year to two years pTx. In all others, only univariable analysis was performed (due to the low number of events). eGFR decline >30% from half a year to two years showed statistical correlation with higher PPI doses in multivariable logistic regression (*p* = 0.044). All other eGFR decline endpoints showed no relation to mean daily PPI intake. eGFR decline >50% from two to four years showed some tendency in the same direction (*p* = 0.056). However, here, only three events occurred. Tables with the results can be viewed in Supplementary Materials 4A–D.

#### *3.4. Secondary Outcomes*

BPAR occurred in 96 patients in months one to six, in 32 patients in months seven to twelve and in 36 patients in the second year pTx. For the rejection analysis, patients who did not complete a follow up of at least 5/6 of the analyzed time points were excluded from analysis. Those with death or transplant loss with prior rejection, however, were included. Twelve patients were excluded from multivariable analyses due to missing data. The results of the analysis are presented in Figure 2A–C. Regarding rejection types, mean daily PPI intake only showed a correlation with antibody mediated rejections (AMR) in the second year pTx in univariable analysis (*p* = 0.027); multivariable (logistic regression) analysis was not feasible in this case due to zero events in one group. All other tests did not show any association.

**Figure 2.** *Cont*.

As the possible drug interaction of MMF with PPIs was of special interest, all rejection analyses were repeated, analyzing the data of patients who had MMF at primary discharge only. None of these tests showed any significant association.

#### **4. Discussion**

Our results demonstrate that prolonged PPI intake after KTx does not lead to any meaningful decline in kidney function within the first four years after transplantation. Additionally, our analyses of rejection rates are in line with recent studies showing no relevant association between rejection rates and PPI-intake after KTx [22,24,25].

Several large epidemiological studies observed a relationship between AKI, CKD, and PPI intake [1–5]. From case studies, an association between PPI intake and interstitial nephritis was previously assumed [32]. Estimates of the impact of this finding have not been around as long [33–35]. However, the relevance and the stake of interstitial nephritis for AKI in PPI observational studies still remains unclear [35,36]. As AKI can lead to CKD, it was not surprising that an association between CKD and PPI intake was recently proposed [1]. Nevertheless, Xie and colleagues provided evidence that PPI-associated CKD even occurred in the absence of AKI [3]. The pathomechanism, however, remains unknown. Proposed mechanisms include elevation of plasma asymmetric dimethylarginine levels [37], microinflammation due to gut microbe dysbiosis [38], endothelial senescence [39] and PPI-induced hypomagnesemia [40]. In addition to hypomagnesemia, PPI intake was recently associated to be dose-dependently linked to iron deficiency and hypomagnesemia in a kidney transplant cohort from the Netherlands [41,42]. The authors speculated that the effects were associated with reduced intestinal absorption of both elements under PPI therapy. However, our center's policy is to monitor the iron status and to replace magnesium after KTx, because iron-deficiency is common and calcineurin inhibitors frequently lead to magnesium loss. Considering these findings, our results may relieve unwarranted fear when prescribing PPIs in KTx patients.

Following the Kidney Disease Improving Global Outcomes (KDIGO) guideline, a form of mycophenolate acid (MPA) together with the calcineurin inhibitor tacrolimus and low dose corticosteroid therapy is the preferred maintenance therapy after KTx at our center (Table 1 and Supplementary Materials 1) [43]. The tacrolimus target level in our center was 6–10 ng/mL from months one to three and 4–8 ng/mL for the following time.

Tacrolimus and PPI potentially interfere e.g., at the cytochrome P450 system (CYP3A) [17,18]. Another mechanism proposed is that PPI can increase the uptake of tacrolimus in the small intestine [19,44]. Both mechanisms could increase blood tacrolimus levels. Usually, these interactions are not clinically noticeable as several factors have a more profound effect on tacrolimus metabolism and exposure [45]. However, as tacrolimus can be nephrotoxic, slight increases in exposure may be relevant in the course of time [18,46]. Although analysis of tacrolimus blood levels was not the goal of this study, it is interesting to note that tacrolimus blood levels at three months pTx showed a tendency (*p* = 0.07) to be higher in the PPI group (Supplementary Materials 1). Furthermore, significantly more fast tacrolimus metabolizers were found in the no PPI group (*p* = 0.035). However, the previously mentioned Dutch KTx magnesium study found no relevant interaction between PPI and tacrolimus in their observational study [42]. Thus, these interesting findings may warrant further investigation.

It may also be worth mentioning that we did not investigate a possible effect of PPIs on the intrapatient variability of tacrolimus. It has been shown that the intrapatient variability of tacrolimus correlates with poor long-term outcomes in kidney transplant recipients [47]. It may be assumed that any drug with the possibility of interfering with tacrolimus pharmacokinetics may potentially change the intrapatient variability of tacrolimus [48]. To our knowledge, no study exists which has directly investigated a potential relationship between PPI intake and increased intrapatient variability in kidney transplant recipients. However, we suspect this effect to be minimal when adherence to our center's instruction of ingesting tacrolimus on an empty stomach is followed. As we did not see any differences regarding the eGFR changes between the groups, a clinically relevant effect of PPIs on transplant function does not seem to exist.

Mycophenolate mofetil is a prodrug that is hydrolyzed to the active metabolite MPA. It acts as a selective uncompetitive inhibitor of inosine monophosphate dehydrogenase (IMPDH) and thus inhibits de novo guanosine synthesis [6]. Pharmacokinetic studies have shown that PPIs may reduce MPA exposure in patients receiving concomitant MMF and PPI treatment, whereas the alternative drug enteric-coated mycophenolate sodium is not affected [6–12]. Decreased MPA exposure can increase rejection rates [13–16]. However, in a recent pharmacokinetic blinded cross-over study [49], this mentioned interaction was not found. In line with this and with previously published observational studies, rejection rates in our cohort were comparable between groups [22,24,25].

Previous PPI studies in KTx patients relevant to our investigation have focused on the possible interaction of MMF with PPIs. Van Boeckel and colleagues compared 125 patients taking pantoprazole with 77 patients using ranitidine [22]. The primary outcome was BPAR and secondary outcomes were creatinine and the eGFR at three months pTx. No significant differences were found in any of the outcomes. Knorr and colleagues [24] came to similar results in their comparative study of 213 patients receiving PPIs, and 390 with ranitidine. The primary outcome was BPAR in the first year pTx. Both groups had comparable rejection rates and eGFRs. Notably, in the subgroup of African American recipients (predominantly fast tacrolimus metabolizers), PPI intake and rejection rates correlated. The recent study by Patel and colleagues [25] compared rejection rates in 183 patients taking PPIs and 339 using histamine-2 receptor antagonists. The primary outcome was the incidence of acute rejection within one year pTx, but eGFR values at one month and one year were also compared between the two groups. None of these parameters showed a significant difference.

Our study adds to the existing literature as we extended the analyses of the eGFR and changes thereof to a longer time period (half a year to four years) than previous studies (analyzing three to twelve months).

A simple comparison of the eGFR reflected differences in patient characteristics (Table 1 and Supplementary Materials 1). These differences were not a result of PPI therapy as is shown by comparable changes in the eGFR in both groups during follow-up (Table 2).

Moreover, similar rejection rates in the groups taking different doses of PPI, namely 0 mg, 1–20 mg, 21–40 mg and >40 mg pantoprazole equivalent, led us to conclude that the relevance of PPI intake for MMF efficacy is at best minimal. This is in line with the previously mentioned observational studies that used different methodologies. Correction for various potentially relevant confounders did not change the results with regards to kidney function or rejection rates. In line with our observations is a recently published meta-analysis that evaluated the data of 6786 KTx patients. The authors found that PPI use was linked to hypomagnesemia, but not associated with acute rejection, graft loss, or one-year mortality [40].

One limitation of our study is the retrospective study design analyzing a limited number of patients from one center. For part of the rejection analysis and >30% and >50% eGFR decline, the inclusion of covariables in the multivariable logistic regression analyses was limited due to the low number of events. Confounding by indication proved a serious difficulty in this study. Patients with longer, higher PPI intake showed a tendency to higher comorbidity and risk factors (Table 1 and Supplementary Material 1). To cope with this, we included information on the Charlson comorbidity index. PPI therapy post-transplant is the standard of care at our center. Discontinuance of PPI medication was assumedly due to clinical evaluation or patient choice, not by standard procedure.

For the statistically observed relation of second year AMRs and >30% eGFR decline (from half a year to two years) with PPI average intake, we presume residual confounding to be the reason because in a post-hoc comparison of the groups used for the >30% eGFR decline analysis, we found significant differences for the following patient characteristics: recipient age and BMI, pre-transplant dialysis time, prior renal transplant, donor type (living or deceased), pre-transplant diabetes, ABO transplant incompatibility, cerebral arterial occlusive disease or stroke, coronary heart disease or myocardial infarction, statins, prior smoking history, continuation of smoking after KTx, and Charlson comorbidity index. All of these, apart from prior renal transplant and ABO incompatibility, showed an unfavorable tendency with higher PPI intake.

#### **5. Conclusions**

We conclude from our data that prolonged PPI therapy is safe in regard to KTx function. However, further studies into a possible interaction between PPIs and tacrolimus may be of interest. In addition, our findings highlight the importance of examining changes in the eGFR rather than single eGFR measurements in similar studies. Polypharmacy is a relevant problem in the transplant population [50] and it is always advisable to question unnecessary medication [51]. This may include PPI therapy.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2077-0383/9/1/258/s1. Supplementary Materials 1: Additional patient information; Supplementary Materials 2: Statistical analysis; Supplementary Materials 3: Results of the group comparison of the absolute eGFR; Supplementary Materials 4: Results of the >30% and >50% eGFR decline with mean daily PP-dose tests.

**Author Contributions:** D.J.G.F., K.S.-N., S.R. and conceived and designed the study, D.J.G.F. and K.S.-N. collected the data. D.J.G.F. and K.S.-N. analyzed the data. H.P. and B.S. supervised the project and discussed the data, D.J.G.F., K.S.-N. and S.R. wrote the main manuscript text. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Acknowledgments:** We are grateful to René Schmidt for his statistical advice during the course of the project. We also acknowledge support by the Open Access Publication Fund of the University of Münster.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*

### **Circulating Arsenic is Associated with Long-Term Risk of Graft Failure in Kidney Transplant Recipients: A Prospective Cohort Study**

**Camilo G. Sotomayor 1,\*,**†**, Dion Groothof 1,**†**, Joppe J. Vodegel 1, Tomás A. Gacitúa 1, António W. Gomes-Neto 1, Maryse C. J. Osté 1, Robert A. Pol 2, Catterina Ferreccio 3, Stefan P. Berger 1, Guillermo Chong 4, Riemer H. J. A. Slart 5, Ramón Rodrigo 6, Gerjan J. Navis 1, Daan J. Touw <sup>7</sup> and Stephan J. L. Bakker <sup>1</sup>**


Received: 31 December 2019; Accepted: 31 January 2020; Published: 3 February 2020

**Abstract:** Arsenic is toxic to many organ systems, the kidney being the most sensitive target organ. We aimed to investigate whether, in kidney transplant recipients (KTRs), the nephrotoxic exposure to arsenic could represent an overlooked hazard for graft survival. We performed a prospective cohort study of 665 KTRs with a functional graft ≥1 year, recruited in a university setting (2008-2011), in The Netherlands. Plasma arsenic was measured by ICP-MS, and dietary intake was comprehensively assessed using a validated 177-item food-frequency questionnaire. The endpoint graft failure was defined as restart of dialysis or re-transplantation. Median arsenic concentration was 1.26 (IQR, 1.04-2.04) μg/L. In backwards linear regression analyses we found that fish consumption (std β = 0.26; *p* < 0.001) was the major independent determinant of plasma arsenic. During 5 years of follow-up, 72 KTRs developed graft failure. In Cox proportional-hazards regression analyses, we found that arsenic was associated with increased risk of graft failure (HR 1.80; 95% CI 1.28–2.53; *p* = 0.001). This association remained materially unaltered after adjustment for donor and recipient characteristics, immunosuppressive therapy, eGFR, primary renal disease, and proteinuria. In conclusion, in KTRs, plasma arsenic is independently associated with increased risk of late graft failure.

**Keywords:** arsenic; diet; fish consumption; oxidative stress; kidney transplantation; graft failure

#### **1. Introduction**

Arsenic is toxic to many organ systems, the kidney being the most sensitive target organ [1,2]. Free radical mediated-oxidative damage is the cornerstone of arsenic-induced pathology [3]. Arsenic induces morphological alterations of mitochondria that lead to uncontrolled formation of free radicals [4], whilst it inhibits the production of glutathione that protects cells from oxidative damage, ultimately yielding irreversible cell damage [5,6]. The kidney being a major player in removal of arsenic from the system, it is also very much exposed to arsenic and therefore susceptible to arsenic-induced toxicity [7–10].

A large variety of arsenic compounds are known, divided into the elemental metal, inorganic, and organic compounds with a large variety of toxicity [1,2,11,12]. While an extraordinary cause for arsenic intake has been described as hydroarsenicism—contamination of drinking water with arsenic in the US, Chile, and Taiwan—arsenic in food is an increasingly recognized pathway of environmental exposure. Thus, upon background regional differences, arsenic exposure substantially derives from rice consumption, as well as vegetables, fruits, and herbal tea [13–19]. Of note, however, seafood is thought to be a major route for arsenic intake, followed by alcohol consumption, with the latter mainly due to contaminated wine, therewith representing an evident public health threat [20,21].

Basic and clinical evidence has linked arsenic exposure to nephrotoxicity, tubular necrosis, diffuse interstitial fibrosis, decline of kidney function, incident chronic kidney disease, and progress of native chronic kidney disease, among several other conditions such as hypercalciuria, albuminuria, and nephrocalcinosis [22–29]. Kidney transplant recipients (KTRs) are particularly vulnerability to the harmful effects of nephrotoxic agents. However, no study has been devoted to evaluating whether arsenic may be an otherwise overlooked modifiable risk factor in the post-kidney transplantation setting. The current study, therefore, aimed to identify independent environmental and system determinants of plasma arsenic levels and to evaluate the potential association of plasma arsenic levels with long-term risk of graft failure in a large cohort of well-characterized KTRs.

#### **2. Methods**

#### *2.1. Design and Study Population*

In this prospective cohort study, outpatient adult KTRs with a functioning graft ≥1 year, no alcohol or drug addiction, and without known systemic illnesses (i.e., malignancies, opportunistic infections) were invited to participate. The recruitment of patients took place at the University Medical Center Groningen between November 2008 and March 2011. In total, 817 KTRs were invited for the study, of whom 707 (87%) provided written informed consent to participate. All patients with missing plasma arsenic levels were excluded, resulting in 665 KTRs eligible for statistical analyses. Multiple imputations (*n* = 5) were used to account for missingness of data among variables other than data on circulating arsenic. The present study was approved by the Institutional Review Board (METc 2008/186) and was conducted in accordance with the Declaration of Helsinki.

The primary outcome of this study was death-censored graft failure, defined as end-stage kidney disease requiring dialysis or re-transplantation. The continuous surveillance system of the outpatient clinic of our university hospital, in which patients visit the outpatient clinic with declining frequency in accordance with the American Transplantation Society Guidelines, ensured updated information on patient status [30]. General practitioners or referring nephrologists were contacted in case the status of a patient was unknown. Endpoints were recorded until September 2015. No patients were lost to follow-up.

All KTRs were transplanted at the University Medical Center Groningen following the establishment of standard antihypertensive and immunosuppressive therapies. Relevant characteristics including recipient age, gender, cardiovascular history, and transplant-related information were extracted from patient records. Dietary intake, clinical parameters, and laboratory measurements were extensively assessed at baseline.

#### *2.2. Assessment of Dietary Intake*

Dietary intake was assessed using a validated semi-quantitative food frequency questionnaire (FFQ) developed and updated at Wageningen University [31]. The questionnaire consisted of 177 food items to record intake during the last month, taking seasonal variations into account. For each item, the frequency was expressed in times per day, week, or month. The number of servings was recorded in natural units (e.g., slice of bread or apple) or household measures (e.g., cup or spoon). The FFQ was self-administered and then checked by a trained researcher on the day of visit to the outpatient clinic. Inconsistent answers were verified with the patients. The results of the FFQ were converted into total energy and nutrient intake per day by using the Dutch Food Composition Table of 2006 [32].

#### *2.3. Clinical Parameters and Definitions*

All measurements were performed during a morning visit to the outpatient clinic. Blood pressure was determined with a semi-automatic device (Dinamap 1846, Critikon, Tampa, FL, USA), measuring every minute for 15 min. The last three measurements were averaged, following a strict protocol as described previously [33]. Body mass index (BMI) was calculated as weight in kilograms divided by height in meters squared (kg/m2), and body surface area (BSA) was estimated in meters squared (m2) by using the universally adopted formula of DuBois and DuBois [34]. Diabetes was defined as use of antidiabetic medication, fasting plasma glucose ≥ 7.0 mmol/L, and/or HbA1C higher than 6.5% [35]. Kidney function was assessed by means of estimated glomerular filtration rate (eGFR) according to the Chronic Kidney Disease Epidemiology Collaboration equation [36].

#### *2.4. Laboratory Methods and Arsenic Measurement*

Blood was drawn after a fasting period of 8–12 h, which included no medication intake. Serum high-sensitivity C-reactive protein (hs-CRP), HbA1C, triglycerides, low-density lipoprotein (LDL) cholesterol, high-density lipoprotein (HDL) cholesterol, and total cholesterol were measured using routine laboratory methods. Serum creatinine was determined using a modified version of the Jaffé method (MEGA AU 510, Merck Diagnostica, Darmstadt, Germany). Serum cystatin C was determined using Gentian Cystatin C Immunoassay (Gentian AS, Moss, Norway) on a modular analyzer (Roche Diagnostics, Mannheim, Germany). Class I and class II human leukocyte antigens (HLA) antibodies were assessed by ELISA (LATM20×5, One Lambda, Canoga Park, CA, USA) as described elsewhere [37]. According to a strict protocol, all participants were instructed to collect a 24 h urine sample the day before to their visit to the outpatient clinic. Total urinary protein concentration was determined using the Biuret reaction (MEGA AU 150, Merck Diagnostica, Darmstadt, Germany).

Arsenic plasma concentrations were assessed from EDTA plasma samples that were stored frozen at −80 ◦C. Arsenic plasma concentrations were determined using inductively coupled plasma mass spectrometry (ICP-MS, Varian 820-MS; Varian, Palo Alto, CA, USA) with a modified method for the measurement of low concentrations of heavy metals in plasma using a standard addition method. Standards were made by addition to blanc plasma known amounts of arsenic to obtain added concentrations of 0.500, 1.00, 2.00, 3.00, 4.00, and 5.00 μg/L. Control samples were made by spiking blanc plasma with known amounts of arsenic to obtain added concentrations of, respectively, 0.75 (low), 2.5 (medium), and 4.5 μg/L (high). Sample preparation consisted of diluting 100 μL sample with 1.0 mL dilution reagent. The dilution reagent contained 0.005% Triton X100, 0.005% EDTA, and 0.1 mg/L Yttrium as internal standard. Characteristics of this method are summarized in Table 1.


**Table 1.** Bias and precision of arsenic measurements.

*n*, number of control samples; SD, standard differentiation; CV, coefficient of variation.

#### *2.5. Follow-Up of Plasma Arsenic Levels in a Sample Population of the TransplantLines Cohort and Biobank Study*

Additionally, to investigate plasma arsenic levels over time, we requested follow-up plasma samples (3 months, 6 months, 1 year, and 2 years post-kidney transplantation) from 46 consecutive KTRs enrolled between February 2016 and May 2017 in the ongoing TransplantLines Prospective Cohort and Biobank Study [38]. Arsenic plasma concentrations were determined using inductively coupled plasma mass spectrometry (ICP-MS, Varian 820-MS; Varian, Palo Alto, CA, USA) with a modified method for the measurement of low concentrations of heavy metals in plasma using a standard addition method, as described hereby in the preceding section.

#### *2.6. Statistical Analyses*

Data analyses were performed using SPSS version 23.0 software (SPSS, Inc., Chicago, IL, USA) and R version 3.2.3 (R Foundation for Statistical Computing, Vienna, Austria). Continuous variables were summarized using mean (SD) for normally distributed data, whereas skewed distributed variables are given as median (IQR). Categorical variables were summarized as numbers (percentage). In all analyses, a two-sided *p* < 0.05 was considered significant. Linear regression analyses were performed to evaluate the association of baseline characteristics with arsenic concentrations, adjusted for (i) age and sex, and additionally (ii) eGFR. The assumption of homoscedasticity and normality of residual variance were verified, and a natural log-transformation was applied when appropriate. Std. β coefficients represent the difference (in SD) in arsenic per 1-SD increment in continuous characteristics or for categorical characteristics the difference (in SD) in arsenic compared to the implied reference group. In order to study, in an integrated manner, which baseline characteristics were independently associated with and were determinants of plasma arsenic, we performed forward selection of baseline characteristics according to preceding multivariable linear regression analyses (*p* for inclusion < 0.2), followed by stepwise backwards multivariable linear regression analyses (*p* for exclusion 0.05). Finally, we also performed a stepwise backwards multivariable linear regression with exclusion of eGFR in the initial model in order to isolate environmental determinants of plasma arsenic levels.

The prospective association of plasma arsenic with risk of graft failure during follow-up was examined incorporating time to event and accounting for death-censoring, by means of univariable and multivariable Cox proportional-hazards regression analyses with time-dependent covariates to calculate hazard ratios (HR) and 95% confidence intervals (CI). Schoenfeld residuals were calculated to assess whether proportionality assumptions were satisfied. Associations are shown with plasma arsenic as a continuous variable and according to tertiles of the plasma arsenic distribution. Following univariable analyses (model 1), we first performed multivariable adjustment for the most important environmental determinants of arsenic levels according to the results of our backwards linear regression analyses (model 2). To avoid overfitting, further models were performed with additive adjustments to model 2, defined as the primary multivariable model [39]. Thus, we performed additional adjustments for intake of fruits, vegetables, potato, rice, bread, and total energy intake (model 3); transplant characteristics (donor and recipient age, donor type, HLA mismatches, circulating anti-HLA class I antibodies, circulating anti-HLA class II antibodies, transplant vintage, and immunosuppressive therapy; model 4); risk factors of graft failure (eGFR, hs-CRP, systolic blood pressure, total cholesterol, and triglycerides concentration; model 5); and primary renal disease and proteinuria in model 6.

The intra-individual coefficient of variation (CV) for plasma arsenic levels in KTRs of the TransplantLines Cohort and Biobank Study was calculated using the formula CV = (SD/mean) × 100, in which SD is the standard deviation and mean is the mean value for plasma arsenic concentrations as measured in follow-up samples taken at 3 months, 6 months, 1 year, and 2 years post transplantation. Next, box plots were used to illustrate medians (interquartile range) of plasma arsenic levels during follow-up visits. Finally, significance of potential change during follow-up visits was tested using the Kruskal Wallis test.

#### **3. Results**

#### *3.1. Baseline Characteristics and Cross-Sectional Analyses*

Mean (SD) age of the 665 KTRs was 53 (13) years, of whom 383 (58%) were male. Median (IQR) plasma arsenic concentration was 1.26 (1.04–−2.04) μg/L. The baseline characteristics of the study participants along with the results of age- and sex- as well as eGFR-adjusted linear regression analyses are shown in Table 2. In stepwise backward multivariable linear regression analysis, fish consumption (β = 0.26; *p* < 0.001), eGFR (β = −0.11; *p* = 0.02), and proteinuria (std β = 0.18; *p* < 0.001) were identified as independent determinants of plasma arsenic concentrations (Table 2). If analyses were performed with eGFR excluded from the initial model, fish consumption (β = 0.27; *p* < 0.001) was identified as the only independent determinant of arsenic (Table 2).

#### *3.2. Prospective Analyses*

During a follow-up of 5 years, 72 (11%) patients developed graft failure. Chronic allograft dysfunction was the major cause of graft failure accountable for 50 (69%) of all graft failures. Other causes for graft failure included return of primary kidney disease (11%), infection (4%), acute rejection (4%), BK nephropathy (4%), vascular complications (3%), and others (4%). From low to high tertiles of the plasma arsenic distribution, 18, 25, and 29 patients developed graft failure, respectively. Prospective analyses of the association of plasma arsenic with death-censored graft failure are shown in Table 3. Multivariable-adjusted Cox proportional hazards models showed that plasma arsenic was directly associated with graft failure (HR 1.80; 95% CI 1.28–2.53, *p* = 0.001), independent of major environmental determinants of arsenic concentration, i.e., alcohol and fish consumption. In analyses with further adjustment for potential confounders, the association remained materially unchanged (Table 3). We did not find signs of a non-linear association between plasma arsenic levels and risk of death-censored graft failure (Supplementary Materials Table S1). Figure 1 illustrates the association between plasma arsenic concentration and risk of death-censored graft failure using Cox regression analyses with mean concentration of plasma arsenic as reference, adjusted for age, sex, fish intake and alcohol consumption, and in relation to the histogram of plasma arsenic distribution.


Baseline characteristics of 665 kidney transplant recipients (KTRs) and their association with plasma

arsenic.

**Table**

**2.**


**Table 2.** *Cont.*

aspartate low-density lipoprotein.

aminotransferase;

 eGFR, estimated glomerular filtration rate; HDL, high-density lipoprotein; HLA, human leukocyte antigens; hs-CRP,

high-sensitivity

 C-reactive protein; LDL,



Cox proportional-hazards regression analyses were performed to assess the association of plasma arsenic with risk of death-censored graft failure (number of events = 72). Associations are shown with plasma arsenic concentration as a continuous variable and according to tertiles of the plasma arsenic distribution (tertile 1: ≤1.1 μg/L; tertile 2: 1.1-1.67 μg/L; tertile 3: ≥1.67 μg/L). Model 1 is univariable. Multivariable model 2 was adjusted for fish intake and alcohol consumption. Subsequently, additive adjustment was performed for intake of fruits, vegetables, potato, rice, bread, and total energy intake (model 3); donor and recipient age, donor type, human leukocyte antigen mismatches (HLA), circulating anti-HLA class I antibodies, circulating anti-HLA class II antibodies, transplant vintage, and immunosuppressive therapy (model 4); eGFR, high-sensitivity C-reactive protein, systolic blood pressure, total cholesterol, and triglyceride concentration (model 5); primary kidney disease and proteinuria (model 6).

**Figure 1.** Association between plasma arsenic concentration and risk of death-censored graft failure using Cox regression analyses with mean concentration of plasma arsenic as reference, adjusted for age, sex, fish intake, and alcohol consumption, and in relation to the histogram of plasma arsenic distribution.

#### *3.3. Follow-up of Plasma Arsenic Levels in a Sample Population of the TransplantLines Cohort and Biobank Study*

In Supplementary Materials Figure S1 we show box plots with medians (IQR) of plasma arsenic concentration of 46 KTRs (mean age 52 <sup>±</sup> 14 years-old, eGFR 43 <sup>±</sup> 28 mL/min/1.72 m2) from the TransplantLines Prospective Cohort and Biobank Study, at different follow-up visits post-kidney transplantation. Median (interquartile range) plasma arsenic concentrations were 1.61 (1.51–1.99), 1.64 (1.52–2.05), 1.64 (1.43–1.94), and 1.59 (1.46–2.26) μg/L at 3 months, 6 months, 1 year, and 2 years post-kidney transplantation, respectively. Median (interquartile range) intra-individual coefficient of variation was 12.2% (6.7–28.7%), and we did not find signs of a significant change in plasma arsenic levels over time (*p* = 0.64).

#### **4. Discussion**

In these analyses of 665 well-characterized individuals from a Dutch cohort of KTRs, we identified fish consumption as the major environmental determinant of plasma arsenic levels. Prospective analyses showed that higher plasma arsenic levels are associated with increased long-term risk of graft failure, independent of donor and recipient characteristics, immunosuppressive therapy, eGFR, and proteinuria. These data pose arsenic as a potentially modifiable risk factor for late graft failure in KTRs, emphasizing the need for specific recommendations regarding arsenic exposure, as well as patient monitoring and management of arsenic-induced kidney injury, particularly in populations highly susceptible to nephrotoxic agents such as KTRs.

Being the major organ involved in arsenic clearance, the kidney is highly susceptible and the most sensitive target organ to arsenic exposure [1,2,9,10]. Arsenic-induced oxidative stress has been suggested to be the cornerstone of pathological mechanisms leading to kidney injury and development of chronic kidney disease [3,40]. On the one hand, decreased antioxidant capacity has been shown in individuals exposed to arsenic [41], wherein depletion of glutathione has been consistently described [5,42,43]. Of note, by protecting cells from oxidative damage, inhibition of glutathione production and subsequent glutathione depletion ultimately reverberates into increased vulnerability of cells to arsenic damage. On the other hand, it has been shown that arsenic induces morphological alterations of mitochondrial integrity that lead to uncontrolled free radical formation [4], which further feeds the circle of oxidative challenge and tissue injury. Indeed, basic and clinical evidence has linked arsenic exposure to nephrotoxicity, tubular necrosis, diffuse interstitial fibrosis, decline of kidney function, incident chronic kidney disease, and progress of native chronic kidney disease, amongst other conditions such as hypercalciuria, albuminuria, and nephrocalcinosis [22–29]. Subsequently, diminished kidney clearance of arsenic and enhanced production of reactive oxygen species longitudinally contribute to perpetuate tissue insult and progression of chronic kidney disease [22,23]. Previous studies have also shown an association between arsenic and hypertension and type 2 diabetes mellitus, both suggesting additional mechanisms for secondary kidney damage [44,45]. Ecological studies from the United States, Chile, and Taiwan have shown that arsenic exposure is associated with increased mortality from kidney disease [13–15,22,26,28,46–49]. KTRs are particularly vulnerable to harmful effects of nephrotoxic agents. End-stage kidney disease and maintenance immunosuppressive therapy are constant sources of oxidative challenge for the graft tissue, which shortens the capacity of oxidative stress defenses against additional environmental hazards. To our knowledge, the current study is the first to provide evidence of an independent prospective association between circulating arsenic levels and risk of late kidney graft failure.

Further supportive evidence for the key role of oxidative stress in arsenic-induced pathogenic mechanisms—and suggestive of potential management alternatives—was provided by the observation that co-administration of ascorbic acid and α-tocopherol to arsenic-exposed rats led to a reduction in the levels of lipid peroxidation, protein carbonyls, and hydrogen peroxide along with increased levels of reduced glutathione, ascorbic acid, and α-tocopherol. Investigation aimed to evaluating whether ascorbic acid and α-tocopherol supplementation may improve arsenic-induced altered microsomal functions in the kidney is warranted [50].

An increasing body of evidence supports that the kidney is a primary site of arsenic uptake and accumulation. Recently, X-ray fluorescence spectrometry allowed detection of arsenic accumulation, specifically at level of the kidney cortex [51]. X-ray fluorescence spectrometry may provide

comprehensive information of bioaccumulation for biomedical and toxicological research by allowing direct measurement of the distribution of arsenic at tissue, cellular, and subcellular level. Next, X-ray absorption spectroscopy has been shown to allow in vivo assessment of whole-body distribution, which is key information for the development of chelation therapies [52]. Future studies using these analytical methods may provide essential research data to understand the sequence of specific mechanisms of nephrotoxicity and deepen the understanding of the association between long-term arsenic exposure and kidney damage [51].

The current study is etiological in nature, which needs to be separated from prediction research [53]. Whereas the latter is a distinct field of epidemiologic research aimed at predicting the risk of an outcome according to a model of statistically significant predictors, which not necessarily represents causal associations, etiological studies aim to understand a certain pathway of a disease in an attempt to prevent its onset or progression [53]. Taken together, our findings and the aforementioned studies may support an etiological role of arsenic in pathways of disease that contribute to increased risk of death-censored graft failure.

Data on the average diet-derived arsenic exposure in The Netherlands are scarce. One study reported an estimated median (range) exposure of 37.8 (20.6–70.1) μg/day [54]. This was corroborated by a more recent study of Hoogenboom et al. stating that the average diet-derived arsenic exposure is <50 μg/day. In agreement with our findings, higher intake of arsenic most frequently originates from higher fish consumption [55]. A monitoring program from the Dutch Agriculture Advisory Committee (LAC), conducted in the 1980s, demonstrated that levels of arsenic in fish landed in The Netherlands varied between 0.8 and 6.8 mg/kg wet weight, showing a slight decreasing trend over time. Likewise, the arsenic levels in shrimps decreased from 4.3 to 1.3 mg/kg wet weight during that period (LAC program, 1991, in reference [41]). However, more recent data regarding arsenic-contaminated fish landed in The Netherlands are lacking and needed to evaluate strategies aiming to reduce the dietary consumption of arsenic by the population. Next, although in The Netherlands, naturally occurring arsenic concentrations in drinking water are usually below the concentrations required by the European drinking water standard (<10 μg/L in all countries, except Denmark, where it is <5 μg/L), health risks cannot be excluded at this level, and it has been recommended to optimize water supply to arsenic levels <1 μg/L [56,57].

The current study was performed in a large cohort of extensively phenotyped KTRs, allowing us to control our main findings for several potential confounders, including donor and recipient characteristics, immunosuppressive therapy, proteinuria, and eGFR. Moreover, patients were monitored for an extensive period and patient status was updated without losses to follow-up, allowing the study of the long-term association of arsenic with graft failure. Despite considerable improvement of short-term graft survival during last decades, improvement of long-term outcomes continues to lag behind, emphasizing that future advances in the field of kidney transplantation are expected from the amelioration of long-term graft attrition [58]. Systematic description of modifiable risk factors is key to promote preventive strategies particularly addressed for this population of solid organ patients.

Our study derived from a single university center from the northern part of The Netherlands, which calls for prudence to extrapolate our results to different populations regarding potential environmental arsenic contamination and exposure. Additionally, the observational design of the current study does not allow hard conclusions on causality, nor could the potentiality of reversed causation or residual confounding be eliminated, despite the substantial number of potential confounders for which we adjusted. Furthermore, the technique used in the current study does not allow different species of arsenic to be distinguished, while arsenic species have major varieties in toxicity [1–4,11,12]. Elemental arsenic is nontoxic as the metal is insoluble in bodily fluids, and inorganic species of arsenic, e.g., arsenite and arsenate, are especially toxic to humans. Organic species vary in toxicity; the most common species, monomethylarsonic acid and dimethylarsinic acid, are less toxic compared to inorganic species, and arsenobetaine and arsenosugars have a very low toxicity [1,5,9,11,59–61]. Further studies utilizing techniques with the ability to distinguish between the different species of arsenic,

e.g., high-performance liquid chromatography–inductively coupled plasma-mass spectrometry, could provide more information on the impact of the different species on graft failure in KTRs. A further limitation is that adjustment for immunological factors as potential confounders of the association was limited to adjustment for HLA matching, circulating anti-HLA class I antibodies, and circulating anti-HLA class II antibodies, since we had no data on donor-specific anti-HLA antibodies and biopsy findings. Finally, it should be acknowledged that graft failure can be the consequence of multiple, heterogenous causes. Unfortunately, in our study the numbers of cause-specific cases of death-censored graft failure was too small to allow for meaningful separate analyses [62]. Larger studies are warranted to comprehensively evaluate the association of plasma arsenic with different causes of death-censored graft failure. It should be noticed, however, that this study is the first to indicate a prospective association of arsenic with the hard endpoint graft failure, thus holding a plea for future studies which to only investigate arsenic plasma concentrations, but also take into account concentrations of arsenic in drinking water, and not only in KTRs to investigate associations with death-censored graft failure, but also in other populations, such as patients with diabetes and the general population.

#### **5. Conclusions**

In conclusion, the current study shows for the first time that circulating arsenic levels are independently associated with higher risk of late kidney graft failure, emphasizing the need for specific recommendations regarding arsenic exposure, as well as patient monitoring and management of chronic arsenic-induced kidney damage. Our findings point towards arsenic as an otherwise overlooked modifiable risk factor for adverse long-term kidney outcomes, especially in populations of vulnerability to oxidative stress challenge, *e.g.*, KTRs. Further studies are warranted to confirm our results and investigate the longitudinal association between arsenic exposure and graft failure in KTRs from populations with different dietary and environmental exposure.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2077-0383/9/2/417/s1, Figure S1: Plasma arsenic concentration of 46 kidney transplant recipients from the TransplantLines Prospective Cohort and Biobank Study [38], at different follow-up visits after transplantation, Table S1: Verification of linearity of the association between plasma arsenic and risk of death-censored graft failure.

**Author Contributions:** Conceptualization, C.G.S. and S.J.L.B.; Data curation, D.J.T.; A.W.G.-N.; Formal analysis, C.G.S., D.G., J.J.V., T.A.G., and M.C.J.O.; Funding acquisition, C.G.S. and S.J.L.B.; Investigation, C.G.S. and S.J.L.B.; Methodology, C.G.S., D.G., J.J.V., R.A.P., D.J.T., and S.J.L.B.; Project administration, S.J.L.B., A.W.G.-N.; Software, S.J.L.B.; Supervision, R.A.P., C.F., S.P.B., G.C., R.H.J.A.S., R.R., G.J.N., and D.J.T.; Writing—original draft, C.G.S., D.G., and T.A.G.; Writing—review and editing, C.G.S., D.G., J.J.V., R.A.P., C.F., S.P.B., G.C., R.H.J.A.S., R.R., G.J.N., D.J.T., and S.J.L.B. All authors have read and agreed to the published version of the manuscript.

**Funding:** This study was based on the TransplantLines Food and Nutrition Biobank and Cohort Study (TxL-FN, ClinicalTrials.gov Identifier: NCT02811835), which was funded by the Top Institute Food and Nutrition of the Netherlands (grant A-1003). Camilo G. Sotomayor is supported by a doctorate studies grant from CONICYT (F 72190118).

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Physical Activity and the Development of Post-Transplant Diabetes Mellitus, and Cardiovascular- and All-Cause Mortality in Renal Transplant Recipients**

**Oyuntugs Byambasukh 1,2,\*, Maryse C. J. Osté 3, António W. Gomes-Neto 3, Else van den Berg 3, Gerjan Navis 3, Stephan J. L. Bakker <sup>3</sup> and Eva Corpeleijn <sup>1</sup>**


Received: 30 December 2019; Accepted: 1 February 2020; Published: 3 February 2020

**Abstract:** (1) Background: Little is currently known about the health impacts of daily-life moderate-to-vigorous physical activity (MVPA) in relation to the development of post-transplant diabetes mellitus (PTDM) and the long-term survival of renal transplant recipients (RTRs). (2) Methods: We analyzed self-reported data on MVPA within non-occupational and occupational domains, estimated with the SQUASH questionnaire, from a prospective cohort study of RTRs (*n* = 650) with a functioning graft exceeding 1 year. PTDM diagnoses were based on plasma glucose levels (≥126 mg/dL), HbA1c (≥6.5%), and the use of antidiabetic medication. Mortality data were retrieved from patient files up to the end of September 2015. (3) Results: During a median follow-up period of 5.3 years, 50 patients (10%) developed PTDM and 129 (19.8%) died. Of these deaths, 53 (8.9%) were caused by cardiovascular disease. Cox regression analyses showed that higher MVPA levels among patients were associated with a lower risk of PTDM (hazard ratio (HR); 95% confidence interval (95%CI) = 0.49; 0.25–0.96, *p* = 0.04), cardiovascular- (0.34; 0.15–0.77, *p* = 0.01), and all-cause mortality (0.37; 0.24–0.58, *p* < 0.001) compared with No-MVPA patients, independently of age, sex, and kidney function parameters. Associations of MVPA with cardiovascular and all-cause mortality remained significant and materially unchanged following further adjustments made for transplant characteristics, lifestyle factors, metabolic parameters, medication use, and creatinine excretion (muscle mass). However, the association between MVPA and PTDM was no longer significant after we adjusted for metabolic confounders and glucose levels. (4) Conclusion: Higher MVPA levels are associated with long-term health outcomes in RTRs.

**Keywords:** physical activity; renal transplant recipients; transplantation; post-transplant diabetes mellitus; cardiovascular mortality; mortality

#### **1. Introduction**

Renal transplantation is a more effective treatment strategy than chronic dialysis in patients with end-stage renal disease [1]. However, post-transplant patients are at an increased risk of developing cardiometabolic diseases that lead to high morbidity and mortality among renal transplant recipients (RTRs) [2,3]. The cardiovascular mortality rate is estimated to be 10 times higher for RTRs compared with the general population [2]. Moreover, approximately 20% of RTRs develop post-transplant diabetes mellitus (PTDM) [3]. This situation necessitates an investigation aimed at developing strategies for improving the management of long-term health outcomes in RTRs.

Studies have consistently found that physical activity (PA) is a modifiable factor that contributes to reducing the risk of cardiometabolic diseases and premature mortality within the general population [4,5]. However, there is limited data on the impacts of PA on RTRs [6–10]. Studies conducted on the benefits of PA have mostly focused on the intermediate outcomes of clinical trials entailing exercise training programmes [6–8]. A few studies found that a low PA is significantly associated with substantial weight gain and with risks of cardiovascular and all-cause mortality in RTRs [11–14]. Moreover, it remains unclear whether the benefits of increased PA in relation to patients' long-term outcomes are independent of their health and transplant characteristics (i.e., kidney function and duration of pre-transplant dialysis), lifestyle factors, and use of medication [11–13].

Clinical guidelines for the general population recommend the performance of at least 150 min of moderate-to-vigorous physical activity (MVPA) per week [15]. There are no specific clinical guidelines or recommendations for RTRs. A recent position statement on exercise for solid organ transplant recipients released in 2019 recommending that it is a key step toward raising awareness of the importance of exercise training in the patients among transplant professionals [16]. The available data show that the level of daily-life PA is lower for RTRs compared with individuals within the general population [17–19]. It is unclear whether individuals within the general population as well as RTRs can attain the recommended MVPA level by engaging in different domains of daily-life activities, such as non-occupational and occupational activities. Results showed that occupational MVPA should not be included within assessments of healthy daily-life PA and should not be deemed a substitute for leisure time MVPA. Specifically, it is not known whether both non-occupational and occupational PA can contribute to the improved health of RTRs [20–22].

Therefore, we aimed to investigate the association between daily-life MVPA and the risk of developing long-term health outcomes, such as PTDM as well as cardiovascular and all-cause mortality in RTRs sampled from a large prospective study. We also examined whether these associations were independent of several variables, including age, sex, kidney function, transplant characteristics, lifestyle factors, medication use, metabolic parameters, and anthropometric measures. Moreover, we assessed the benefits of non-occupational MVPA as well as total daily-life MVPA, including occupational PA.

#### **2. Methods**

#### *2.1. Study Population*

This study was conducted in a large single-center prospective cohort of stable outpatient RTR [23,24]. A total of 817 adult RTRs who met the study's eligibility criteria, namely having a functioning graft for at least 1 year and no history of alcohol and/or drug addiction, were invited to participate in the study. Further exclusions were apparent systemic diseases, such as malignancies or active infections. In total, 707 (86.5%) RTRs signed written informed consent. Baseline data were collected between November 2008 and May 2011. We excluded 57 RTRs, whose PA questionnaires were incomplete, from the analysis, leaving a total of 650 RTRs. Subsequently, 148 and 61 RTRs, respectively, with a history of diabetes and cardiovascular diseases (CVDs) prior to undergoing transplants, were excluded from the analyses of PA and the development of PTDM and cardiovascular mortality. The study was conducted according to the Helsinki Declaration and was approved by the UMCG's review board (METc 2008/186).

#### *2.2. Measurements at Baseline*

All baseline measurements have been previously described in greater detail elsewhere [25]. Body mass index (BMI) was calculated as weight (kg) divided by height squared (m2). A semiautomatic device (Dinamap®1846; Critikon, Tampa, FL, USA) was used to measure blood pressure in a half-sitting

position and the average of the final three readings of blood pressure was used. Information on medication was derived from patient records. Daily caloric intake and alcohol consumption were calculated from a validated Food Frequency questionnaire. Information on smoking was obtained by a questionnaire. The serum creatinine-based Chronic Kidney Disease Epidemiology Collaboration equation was used to calculate the estimated glomerular filtration rate (eGFR) [26]. Creatinine excretion—a marker of muscle mass—was calculated from the 24-h urine collection as described earlier [27].

#### *2.3. Assessment of Physical Activity*

The SQUASH is a validated questionnaire used to estimate habitual physical activities performed during a normal week [28]. The SQUASH is pre-structured into four domains: commuting, leisure time and sports, household, and occupational activities. Questions consisted of three main queries: days per week, average time per day, and intensity. In this study, we used activities at the moderate (4.0–6.5 MET) to vigorous (≥6.5 MET) level. Metabolic equivalent (MET) values were assigned to activities according to Ainsworth's Compendium of Physical Activities [29]. We used the combination of leisure-time and commuting (non-occupational) moderate-to-vigorous physical activity (MVPA) minutes per week (min/week) as a measure of PA in this study, since active commuting of high intensity and longer duration is often replacing sports activities, like cycling. We did not include occupational MVPA in the main analysis because of its health benefit is not clear in the general population [20–22,30]. In an additional analysis, we investigated the association between total MVPA, including occupational MVPA, with clinical endpoints. Participants were subdivided into three categories based on their levels of non-occupational MVPA. RTRs who did not engage in PA at a moderate-to-vigorous level were deemed 'inactive' (no-MVPA), and the remaining participants (MVPA > 0) were divided into two groups based on median values of non-occupational MVPA (less active, MVPA-1 and active, MVPA-2). The MVPA min/week (median, interquartile range (IQR)) was used to define the MVPA groups (MVPA-1 and MVPA-2): 5-197 (120, 60–150) and 200–1680 (360, 260–540).

#### *2.4. Endpoints of the Study*

Endpoints of interest in this study were post-transplant diabetes mellitus (PTDM), cardiovascular and all-cause mortality. PTDM was defined according to the presence of at least one of the following criteria: diabetes symptoms (e.g., polyuria, polydipsia, or unexplained weight loss) along with a non-fasting plasma glucose concentration of ≥200 mg/dL (11.1 mmol/L); fasting plasma glucose concentration (FPG) ≥ 126 mg/dL (7.0 mmol/L); start of antidiabetic medication; or HbA1c ≥ 6.5% (48 mmol/L). This definition matched the diagnostic criteria for diabetes applied by the American Diabetes Association, including HbA1c levels, as proposed by the expert panel constituted at the international consensus meeting on PTDM [31,32]. Cardiovascular mortality and all-cause mortality were monitored through continuous surveillance conducted within the outpatient program and retrieved from patients' files up to the end of September 2015. No participants were lost to follow-up.

#### *2.5. Statistical Analysis*

The patient characteristics are expressed as means with a standard deviation for normally distributed variables or as medians with interquartile range (25th to 75th percentile) for non-normally distributed variables and numbers with percentages in case of categorical data. The differences between groups were tested using 1-way analysis of variance tests or Kruskal–Wallis tests for normally and non-normally distributed continuous variables, respectively. The frequency distributions of categorical variables were analyzed using the Pearson Chi-Square test.

We adopted MVPA as a continuous and categorical variable in a Cox regression analysis. First, we tested associations of non-occupational MVPA, considered as a continuous variable, on clinical endpoints. In this analysis, MVPA (measured in min/week) was log-transformed to obtain a normal distribution. Thereafter, multivariate Cox regression analyses were performed to examine whether

higher non-occupational MVPA is associated with lower risks of PTDM, cardiovascular mortality, and all-cause mortality independently of potential confounders which are clinically known confounders in the relating associations of physical activity with long-term outcomes [6,7,12–14,33]. In these analyses, we first adjusted for age and sex (model 1) as well as kidney function parameters, including eGFR, proteinuria, the time lapse between transplantation and the baseline measures, and primary renal disease (model 2). We further adjusted model 1 for transplant characteristics (acute rejection, preemptive transplantation, and living donor status) in model 3. Similarly, we adjusted model 1 for lifestyle factors, such as smoking, alcohol consumption, and daily caloric intake (model 4); calcineurin inhibitors and prednisolone used as immunosuppressive medication (model 5); systolic blood pressure, use of antihypertensive drugs, high-density lipoprotein cholesterol (HDL), and triglycerides (model 6), BMI and waist circumference (model 7), and 24-h creatinine excretion (model 8). With regard to potential collinearity (model 7), we tested the correlation between BMI and waist circumference (*r* = 0.84, *p* < 0.001 for men and *r* = 0.81, *p* < 0.001 for women). Then we performed Cox-regression analyses using separate models adjusted for BMI and waist circumference separately (Supplementary materials, Table S1. We found no differences relating to the confounding effects of total fat and fat distribution. Finally, we included those variables in the same model. In addition, we adjusted model 1 for baseline hemoglobin A1C and fasting plasma glucose (model 9) relating to the association between MVPA and PTDM. We also adjusted model 1 relating to the association between MVPA and PTDM for diet quality (Model 10). We furthermore investigated whether diet quality might modify the association of MVPA with development of PTDM by additional inclusion of a product-term of the continuous variables of diet quality and MVPA in the concerned model, to assess potential interaction between the two. Mediterranean diet score was used as diet quality and assessed with a 177-item validated food frequency questionnaire which is described in greater detail elsewhere [33]. All models (1-10) include up to 6 variables to fulfil the rule of thumb which allows 1 variable per 7–10 events. This is now fulfilled for all analyses [34,35]. Hazard ratios were reported with 95% confidence intervals. Proportional hazard assumptions were tested using the Schoenfield residuals method developed by Grambsch and Therneau [36]. Penalized splines were constructed to visualize the association of non-occupational MVPA with PTDM as well as cardiovascular and all-cause mortality independently of age and sex.

We performed additional analyses to explore the role of work within this population by investigating the associations between total MVPA, including occupational MVPA, and non-occupational MVPA with clinical endpoints for the RTRs who worked (*n* = 322, 49.5%). Occupational status was defined using the answers for the questions related to occupational PA. If responders answered as not applicable, we considered them as unemployed. Another subgroup analysis was performed to address changes in these associations across age categories. The population was categorized as being over or under 55 years of age, based on the WHO guideline on the prevention of CVD [37]. Finally, to rule out competing mortality risks associated with the occurrence of PTDM, we conducted competing risk analyses following the procedures outlined by Fine and Gray [38].

A two-sided statistical significance was set at *p* < 0.05 for all tests. All statistical analyses were performed using SPSS software V.22 (IBM Inc., Chicago, IL, USA,) R software V.3.2.2 (R Foundation for Statistical Computing, Vienna, Austria), STATA version 13.0 (StataCorp LP, College Station, TX, USA) and Graph Pad Prism 7 (Graph Pad Software Inc., La Jolla, CA, USA).

#### **3. Results**

#### *3.1. Baseline Characteristics*

A total of 650 RTRs (men: 56.3%, mean age: 51.8 ± 13.2 years old) were examined in this study. Baseline measurements were taken 5.7 years (median value; interquartile range (IQR): 1.9–12.1 years) post-transplantation. Of the total sample of RTRs, 37.8% (*n* = 246) did not perform daily MPVA at all within any domain. The other RTRs spent a median of 200 min (IQR = 120–360 min per week) engaged in non-occupational MVPA. Table 1 shows the baseline characteristics of RTRs according to their

non-occupational MVPA levels. RTRs in the active groups (MVPA > 0) had lower values for BMI, waist circumference, and systolic blood pressure and higher creatinine excretion values compared with the values of the inactive group (no-MVPA). Moreover, higher alcohol consumption, lower concentrations of triglycerides and HDL-C, haemoglobin A1C, and less proteinuria and diabetes at the baseline level along with more 'living donors' were observed for the 'active' groups compared with the 'inactive' group. Table S2 presents the baseline characteristics of the RTRs according to the presence of clinical endpoints. Figure S1 further shows levels of daily MVPA according to the participants' ages and work status. As expected, total MVPA values, including those for occupational MVPA, were significantly higher in RTRs who were working (*n* = 322). However, when working status or age was considered, the levels of non-occupational MVPA did not differ significantly.


**Table 1.** Characteristics of the study population, according to MVPA level.


**Table 1.** *Cont.*

Data are presented as mean ± SD or median (interquartile range) and percentage (%, number). MVPA = moderate-to-vigorous physical activity, HDL-C = high-density lipoprotein cholesterol, eGFR = estimated glomerular filtration rate, A2 = angiotensin 2, ACE = angiotensin-converting-enzyme, RAAS = renin–angiotensin–aldosterone system, mTOR = mammalian target of rapamycin. \* the number of patients that do have employment.

#### *3.2. Post-Transplant Diabetes Mellitus*

A total of 50 RTRs (10%) had developed PTDM after a median follow-up period of 5.3 years (4.1–6.0 years). The multivariable Cox proportional hazard models showed that the group with the highest level of non-occupational MVPA was associated with a lower risk of PTDM (hazard ratio (HR); 95% CI = 0.49; 0.25–0.96, *p* = 0.04) compared with the no-MVPA group, independently of age, sex, and kidney function parameters (model 1, Table 2). This association remained significant after we made further adjustments for kidney function parameters, transplant characteristics, lifestyle factors, Mediterranean diet score and 24-h creatinine excretion quantities (considered as a marker of muscle mass) (models 2–4, 8, and 10). Following adjustments made for immunosuppressive medication (model 5), metabolic parameters (model 6), anthropometric measures (model 7), and baseline glucose levels (model 9), the highest level of MVPA was no longer associated with PTDM. However, when MVPA was applied as a continuous variable in the Cox regression analysis, as opposed to using groups of MVPA levels, a higher non-occupational MVPA was associated with a lower risk of PTDM independent of all of the above-mentioned confounders apart from the adjustment of metabolic parameters and glucose level.



DM = Diabetes mellitus, MVPA = moderate-to-vigorous physical activity. Model 1: adjusted for age and sex. Model 2: model 1 + adjustment for kidney function (eGFR, urinary protein excretion, time between transplantation and baseline, and primary renal disease). Model 3: model 1 + adjustment for transplant characteristics (acute rejection, pre-emptive transplantation, donor type). Model 4: model 1 + adjustment for lifestyle factors (smoking, alcohol consumption, daily caloric intake). Model 5: model 1 + adjustment for immunosuppressive medication (calcineurin inhibitors, prednisolonee). Model 6: model 1 + adjustment for lipids and blood pressure (systolic blood pressure, use of antihypertensive drugs, triglycerides, HDL-C). Model 7: model 1 + adjustment for BMI and waist circumference. Model 8: model 1 + adjustment for 24-h creatinine excretion. Model 9: model 1 + adjustment for fasting plasma glucose and HbA1c. Model 10: model 1 + adjustment for Mediterranean diet score.

#### *3.3. Cardiovascular and All-Cause Mortality*

During the follow-up period, 129 (19.8%) patients died. Of these deaths, 53 (8.9%) were caused by cardiovascular disease (CVD). In the multivariable Cox proportional hazard models, the highest level of non-occupational MVPA was associated with a lower risk of cardiovascular mortality (HR; 95% CI = 0.34; 0.15–0.77, *p* = 0.01) compared with the no-MVPA group, independently of age, sex, and kidney function parameters (model 1–2, Table 2). This association remained significant after further adjustments were made for transplant characteristics, immunosuppressive medication, metabolic parameters, and anthropometric measures (models 3, 5–7). However, the association was no longer

significant after adjusting for lifestyle factors (model 4) and creatinine excretion (model 8). Moreover, the association of MVPA with cardiovascular mortality was sustained independently of all of the potential confounders when non-occupational MVPA was applied as a continuous variable in the Cox regression (models 1–9, Table 2).

With regard to all-cause mortality, the group with the highest level of non-occupational MVPA was associated with a lower risk of all-cause mortality (HR; 95% CI = 0.37; 0.24–0.58, *p* < 0.001) compared with the no-MVPA group (model 1, Table 2). This association remained significant after we adjusted for potential confounders (models 2–8). However, the association weakened after we adjusted for transplant characteristics (model 3), metabolic parameters (model 6), and creatinine excretion (model 8). When log-transformed non-occupational MVPA was applied as a continuous variable in the Cox regression analysis, the association was independent of all of the above-mentioned confounders, apart from creatinine excretion (model 8), remaining materially unchanged.

To illustrate these associations further, age-and sex-adjusted penalized splines and the Kaplan-Meier survival curves are shown in Figures 1 and 2.

**Figure 1.** Association between non-occupational MVPA and post-transplant diabetes mellitus (PTDM), Cardiovascular (CV)mortality, and all-cause mortality in renal transplant recipients (RTRs).

**Figure 2.** Probability of survival for PTDM (**a**), cardiovascular mortality (**b**), and all-cause mortality (**c**) according to non-occupational MVPA level.

#### *3.4. Additional Analyses*

Additional analyses of the subgroup of working RTRs revealed that the inclusion of occupational PA in the estimate of MVPA resulted in the attenuation of the HRs of all of the significant and non-significant associations (Table 3). The association of non-occupational MVPA with cardiovascular and all-cause mortality was stronger compared with that of total MVPA, which includes occupational MVPA.


**Table 3.** Additional analysis on the associations of MVPA with long-term health outcomes in RTRs who are working (*n* = 322).

Total PA was the sum of non-occupational and occupational MVPA in min/week. DM = diabetes mellitus, MVPA = moderate-to-vigorous physical activity, HR=hazard ratio, CVD = cardiovascular disease, N \* = number of events. ˆ Analyses were adjusted for age, gender, and kidney function parameters. ˆˆ Analyses were adjusted for age and gender (kidney function parameters excluded in this analysis due to fulfill the rule of thumb).

Age-stratified analyses revealed that associations of MVPA with long-term health outcomes were stronger in older adults (Figure 3).


**Figure 3.** Subgroup analysis for the associations of MVPA with long-term health outcomes over age categories.

The results of the competing risk analyses showed that there was no strong influence of a competing risk of all-cause mortality on the association of MVPA with PTDM. For instance, the competing HR was 0.51 (0.30–0.92, *p* = 0.04) for the highest MVPA with PTDM after adjusting for age and sex. By comparison, the HR was 0.49 (0.25–0.96, *p* = 0.04, model 1, Table 2) when competing risks were discounted.

#### **4. Discussion**

We found that increased daily-life MVPA is associated with a reduced risk of PTDM, cardiovascular mortality, and all-cause mortality in RTRs independently of age, sex, baseline kidney function parameters, transplant characteristics, and other lifestyle habits. The association of MVPA with PTDM was affected by the adjustments we made for baseline glucose levels and metabolic parameters, but it did not seem to be affected by other potential confounders, notably anthropometric and immunosuppressive medication. The associations of MVPA with cardiovascular and all-cause mortality were not substantially affected by adjustments made for the above-mentioned confounders. These results confirm the importance of PA in the long-term healthcare management of RTRs.

Previous studies have found that PTDM is highly prevalent in RTRs [3,39]. However, data on lifestyle interventions for improving glucose tolerance or observational data on the association of increased PA with incidences of PTDM are lacking [13,40]. An intervention study showed that lifestyle modifications, including the incorporation of exercise training, improved 2-h postprandial glucose levels in RTRs who were glucose intolerant [40]. One observational study found that higher levels of PA are associated with a lower risk of glucose intolerance in RTRs [13]. However, this study entailed a cross-sectional design and did not test whether this association of PA is independent of other potential confounders. In our longitudinal study, the association of MVPA with PTDM was found to be independent of age, sex, baseline kidney function parameters, transplant characteristics, and other lifestyle factors, such as smoking, alcohol use, and diet (daily caloric intake and Mediterranean diet score). However, the association was affected by adjustments made for immunosuppressive medication, anthropometric measures (BMI and waist circumference), baseline glucose levels, and metabolic parameters. It is widely accepted that obesity is associated with the development of diabetes within the general population [41]. The use of immunosuppressive medications play a role in the development of PTDM through a pathway of stimulation of gluconeogenesis affecting increased blood glucose which can leads to insulin resistance in combination with other mechanisms [39]. However, when we applied log-transformed continuous MVPA, significant associations were observed after we adjusted for immunosuppressive medication and anthropometric measures, indicating that statistical power issues may also play a role. Thus, further large-scale studies of a longer duration should be conducted to explore whether or not MVPA is associated with PTDM independently of immunosuppressive medication and obesity. Furthermore, diet is an important factor in the development of diabetes. A previous analysis by Osté et al. for our study population showed that Mediterranean style diet predicts the development of PTDM [33]. We found that the association between MVPA and PTDM became slightly stronger when adjusted for Mediterranean diet score indicating the importance of diet, but there was no effect modification by diet quality (P-interaction = 0.147).

A previous study, investigating another sample of RTRs, found that a lower PA is strongly associated with an increased risk of cardiovascular and all-cause mortality [14]. In their Cox regression analyses, these authors found that the association was independent of potential confounders, including the history of CVD, muscle mass, and Framingham CVD risk score factors. However, they did not adjust for some clinical variables, such as kidney function and transplant characteristics (e.g., transplant vintage and donor type). Our study supports an independent association of PA with the risk of cardiovascular and all-cause mortality. Many studies have pointed to the benefit of PA within the general population in preventing premature mortality [4,5]. One of the mechanisms proposed to explain the effects of increased PA entails the improvement of all organ systems, especially the cardiovascular system. Specifically in RTRs, improved cardiovascular function is associated with improvements in kidney function. Increased physical activity can support perfusion and oxygen delivery in the kidneys. Studies have shown that higher levels of daily-life PA are associated with a lower risk of renal function decline within the general population and in patients with chronic kidney disease [42–44]. Consequently, increased PA, by improving kidney function, may be of benefit for long-term graft survival. This effect may also be due to improvements in metabolic dysfunctions, such as insulin resistance, impaired glucose tolerance, dyslipidemia, and hypertension, all of which are related to (central) adiposity [45–47]. Furthermore, a number of studies on diet analysed in-depth the effect of dietary factors on the same outcomes such as PTDM, renal function decline and mortality [23–25,33,48]. They suggest that lifestyle is very important for RTR, however, it should be noted that MVPA in daily life has not gotten that much attention. Taken together, these findings suggest that the improvement of daily-life MVPA needs to be evaluated as a therapy for improving patients' long-term survival.

Within this RTR population, MVPA levels were lower than those within the general population. In our study, 38% of RTRs were inactive (no-MVPA), whereas in the Lifelines cohort, a population-based study for which the same questionnaire (SQUASH) and comparable data processing methods were used, the prevalence of inactivity (no-MVPA) was 10% (*n* = 125,402, 40.5% males, median age of 45) [30,49]. Even in different age groups and gender, it was lower, ranging between 7.5% (*n* = 42,661, 40% of males, median age of 40) and 12.5% (*n* = 34,506, 45.6% of males, median age of 56) in the Lifelines. Lower PA levels among RTRs may be attributed to lower muscle mass (a structural abnormality) and muscle weakness (a functional abnormality) [6]. Our descriptive analysis indicated that inactive RTRs had a lower 24-h creatinine excretion value (a marker of muscle mass) compared with that of active RTRs. We also found that the duration of pre-transplantation dialysis was longer in inactive RTRs, although not significantly so. Studies concluded that low muscle mass can be caused by low PA levels [6,7,9]. This conclusion is in line with our findings, indicating that the association was slightly attenuated after we adjusted for renal factors and muscle mass but that the effect of PA remained evident. Actually a shorter time on dialysis would thus also help post-transplant health because studies showed that the level of PA declines in patients with end-stage kidney diseases and it increases after transplantation [7]. Finally, both recovery of activity after transplantation, as well as prevention of inactivity and loss of muscle mass in people with longstanding kidney disease is important for long-term health after transplantation.

A growing body of evidence is showing that occupational MVPA may have no clear benefit on health in the general population [20–22,30,49]. This was tested in our study including a specific patient population, the RTR. Even in the case of RTR, where being at work may be indicative of relatively good health, individuals who were much more active in terms of their occupational MVPA may not obtain any additional benefits for health. A clear mechanism that prevents occupational PA from generating health benefits is missing. There is always the possibility of residual confounding by factors such as sex, socioeconomic status, work-related stress, and body weight in the association between occupational PA and health outcomes [20,22,30,49]. Studies attempted to explore the possibility of residual confounding, but also found no clear association of occupational MVPA and health outcomes. Thus, we suggest that it is important to be aware that occupational MVPA should not be considered as a substitute for leisure time MVPA in RTR.

The potential benefit of PA seems to be more pronounced in older adults, a phenomenon that was described before in the general population [50]. In the general population, studies concluded that the benefit of PA can be gained more easily when there is more room for improvement, like as in older people. However, it might also be that its effects will be potentially outweighed by other, more important clinical factors (e.g., comorbidities and medication use). Therefore, we attempted to test the effect of physical activity in specific groups such as in RTR in two age groups. We found that a higher MVPA is strongly associated with the development of long-term outcomes such as PTDM and cardiovascular mortality in younger and older adults, but is especially stronger in older adults. Thus, older RTR who are able to remain active despite their longstanding condition are likely to remain relatively healthy.

The strengths of this study include its prospective design, long duration, and complete follow-up. Another strength is we included stable RTRs after transplantation and studied relevant clinical outcomes. Nevertheless, there are some limitations to our study. The observational nature of the study precludes us from drawing conclusions regarding causality. A limitation of this study was its use of self-reporting, which is subject to recall bias, for the PA assessment. However, the SQUASH questionnaire has been validated within general as well as specific populations, such as patients who have undergone total hip arthroplasty and those with ankylosing spondylitis [28,51,52]. Furthermore, PA was assessed at a single point in time. However, in RTR, after 1-year of transplantation, PA is increased by 30% and remained materially unchanged the next 5-years [18]. In this study, we included RTR > 1 year graft functioning with a median of 5.7 years post-transplantation. Another limitation is that we could not fully control for the history of all cardiometabolic diseases in the association of MVPA with all-cause

mortality. Patients with a history of diabetes or CVD before the transplantation were excluded from the analyses on the association between MVPA and PTDM or CV mortality. However, cardiometabolic diseases might be more prevalent in 'No-MVPA' group after transplantation as well. A limitation is that we could not have data on functional evaluations, like e.g., a 6 min walking test, which could have provided important information on cardiovascular efficiency. Finally, single-center nature of study, which mainly consisted of white people is unclear whether our findings can be extrapolated to other populations. It would be relevant to repeat our study in other patient populations.

#### **5. Conclusions**

Higher daily-life MVPA is associated with a reduced risk of PTDM as well as cardiovascular and all-cause mortality in RTRs, suggesting that PA has a positive influence on the long-term health management of RTRs. The associations of MVPA with cardiovascular and all-cause mortality were not substantially affected by adjustments made for potential confounders, such as age, sex, baseline kidney function parameters, transplant characteristics, lifestyle habits, metabolic parameters, anthropometric measures, and immunosuppressive medication. The association of MVPA with PTDM was affected by adjustments of metabolic parameters and glucose levels. The potentially beneficial effects of daily-life PA apply to non-occupational activities at the moderate-to-vigorous level (e.g., commuting, leisure activities, or sport). By contrast, a higher level of occupational MVPA is not directly associated with the development of long-term outcomes. The associations of non-occupational MVPA and the risk of PTDM and cardiovascular mortality were also stronger in older adults. Finally, we suggest that because of the long-term importance of PA, it should be embedded in the healthcare management of RTRs. Furthermore, large scale interventional studies are needed to test the ab initio effect of physical activity after transplantation on the development of post-transplant diabetes mellitus.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2077-0383/9/2/415/s1, Figure S1: Level of daily-life PA according to occupational status (A) and age (B), Table S1: Model 7 and separate adjustments for BMI and waist circumference in the associations of MVPA of with long-term health outcomes in RTRs. Table S2: Baseline characteristics of RTRs according to the presence of long-term health outcomes.

**Author Contributions:** Conceptualization, O.B., S.J.L.B., E.C., G.N., and E.v.d.B.; methodology, O.B., M.C.J.O., A.W.G.-N., E.v.d.B., G.N., S.J.L.B., and E.C.; validation, O.B. and E.C.; formal analysis, O.B. and M.C.J.O.; writing—original draft preparation, O.B. and E.C.; writing—review and editing, M.C.J.O., A.W.G.-N., E.v.d.B., G.N., S.J.L.B., and E.C.; supervision, S.J.L.B. and E.C.; All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by a grant from the Dutch Top Institute Food and Nutrition (A-1003).

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **The Preliminary Results of Bortezomib Used as A Primary Treatment for An Early Acute Antibody-Mediated Rejection after Kidney Transplantation—A Single-Center Case Series**

### **Aureliusz Kolonko \*, Natalia Słabiak-Bła ˙z, Henryk Karkoszka, Andrzej Wi ˛ecek and Grzegorz Piecha**

Department of Nephrology, Transplantation and Internal Medicine, Medical University of Silesia, 40-027 Katowice, Poland; nataliablaz@gazeta.pl (N.S.-B.); hkarkoszka@poczta.fm (H.K.); awiecek@sum.edu.pl (A.W.); g.piecha@outlook.com (G.P.) **\*** Correspondence: uryniusz@wp.pl; Tel.: +48-322-591-429

Received: 30 December 2019; Accepted: 12 February 2020; Published: 15 February 2020

**Abstract:** Proteasome inhibitor bortezomib has been used in the treatment of refractory cases of acute and chronic antibody-mediated rejection (AMR) in kidney transplant recipients. However, its efficacy and safety as a primary treatment for early AMR has been scarcely investigated. We herein present our preliminary experience with bortezomib- and plasmapheresis-based primary treatment for early AMR. Thirteen patients transplanted between October 2015 and September 2019 were treated (starting at median 19th post-transplant day) with bortezomib/plasmapheresis protocol for early biopsy-proven AMR. Twelve out of thirteen patients received 4 doses and one patient recieved 3 doses of bortezomib (1.3 mg/m<sup>2</sup> per dose). In 11/13 patients, 4–7 concomitant plasmapheresis sessions were performed, with or without intravenous immunoglobulin (IVIG). Of note, rituximab was not used in all study patients. The kidney graft and patient survival were 100%. The mean 3-month estimated glomerular filtration rate (eGFR) was 55.3 (95%CI: 44.9–65.8) mL/min/1.73m2, 8/13 patients completed 12-month follow-up with mean eGFR 60.4 (45.4–75.4) mL/min/1.73m2, and 6/13 patients completed a 24-month follow-up period with mean eGFR 73.9 (56.7–91.1) mL/min/1.73m2. Neutropenia < 1 G/L was observed in one patient, third or fourth grade thrombocytopenia in two patients, and eleven patients needed a blood transfusion (median: 2 units/patient). The mid-term results of a primary bortezomib-based treatment for kidney AMR showed its non-inferiority as compared to preceding regimens and acceptable safety. However, our data should be validated in a multicenter randomized trial.

**Keywords:** acute humoral rejection; first-line therapy; outcomes; proteasome inhibitor

#### **1. Introduction**

Acute antibody-mediated rejection (AMR) accounts for 20–30% of all acute rejection episodes after kidney transplantation and is often associated with poor allograft survival [1]. It rarely occurs in unsensitized patients, but may occur in up to 50% of highly sensitized recipients [2,3]. Diagnosis of AMR is based on the histopathologic features of the graft biopsy (glomerulitis, arterial-transmural lesions, thrombotic microangiopathy, etc.) and the presence of donor-specific antibodies (DSAs), with or without positive C4d staining [4]. Nowadays, most frequently used treatment modalities for AMR include plasmaphereses (PF) or immunoadsorption (IA), intravenous immunoglobulin (IVIG), anti-T-cell therapy (antithymocyte globulin, ATG), and anti-B-cell therapy (rituximab). Unfortunately, long-term results of such therapy remain suboptimal [5,6]. The rationale for PF or IA is the removal of readily available antibodies and, therefore, limiting the acute tissue damage. Currently, PF is the standard-of-care for the treatment of AMR despite the evidence uncertainty [6].

As the plasma cells are the main source of antibody production, the proteasome inhibitor, bortezomib (Velcade®, Milennium Pharmaceuticals, Cambridge, MA, USA), had been introduced into the AMR treatment, mostly as a rescue therapy in refractory cases [7–10]. In general, kidney graft outcomes were significantly better in acute than chronic AMR [8,10–12]. In some reports, such an adjuvant therapy was effective in decreasing DSAs and stabilizing kidney graft function in mid-term observation [7,9]. Nonetheless, the rate of graft loss during longer follow-up periods was high and the excretory function of still-functioning transplanted organs was markedly decreased [7,9,13,14]. As early as 2010, Walsh et al. reported the first use of bortezomib along with a single dose of rituximab in two patients as a primary therapy in early post-transplant AMR, with a rapid DSAs elimination and excellent renal function at five and six months of observation [15]. Later on, Waiser et al. compared the effect of bortezomib- versus rituximab-based AMR therapies, but both study subgroups contained acute and chronic AMR cases with a substantial age difference between groups [16]. Since then, only a few case reports have been published [17,18]. The existing patophysiologic evidence of the potential mechanisms leading to the therapeutic effect of bortezomib indicates its ability for causing apoptosis of antibody-producing plasma cells, blocking the secretion of class IgG antibodies against human leukocyte antigens (anti-HLA) [19], and decreasing the number of plasma cells within the graft [20]. In order to maximize the bortezomib efficacy in AMR treatment, in all previously reported protocols, each dose was administered after plasmapheresis session, which aimed to decrease the amount of circulating antibodies. Therefore, we herein present the largest-to-date cohort of kidney transplant recipients with early AMR diagnosed in graft biopsy and primarily treated with bortezomib-based therapy without concomitant rituximab administration.

#### **2. Methods**

#### *2.1. Study Group*

We present a retrospective observational study, including all consecutive kidney graft recipients (KTRs) from our center, transplanted between October 2015 and September 2019 and diagnosed with early biopsy-proven AMR. All patients received their organs from deceased donors and all had negative complement-dependent cytotoxicity (CDC) crossmatch performed immediately prior to transplantation. Flow cytometric crossmatches were not performed. In our country, only patients with negative CDC crossmatch are listed for the transplant centre in order to choose the recipients of organs procured from the donor. Based on the routine pre-transplant screening and the last Luminex results, the presence of DSA with mean fluorescence intensity (MFI) ≥ 5000 eliminates the potential kidney transplant candidate from the ongoing procedure. A routine kidney graft early protocol biopsy was introduced at our center in the second half of 2015. Since than, all patients who were diagnosed with AMR were then assigned to primary therapy, which included bortezomib. The study was conducted in accordance with the Declaration of Helsinki. As this drug is not registered for the AMR therapy, the Bioethic Committee of the Medical University of Silesia was consulted and all patients gave their informed consent for the off-label use of bortezomib in their therapy. Notably, the guidelines for immunosuppressive therapy after kidney transplantation, issued by the Polish Transplant Society, have allowed the use of bortezomib in KTRs with AMR since 2012 [21].

#### *2.2. Immunosuppressive Protocol*

The standard immunosuppressive protocol included tacrolimus 2 × 0.1 mg/kg twice daily (with target through level 7–12 ng/mL) and mycophenolate mofetil 750 mg twice daily, both started immediately prior to operating procedure, and steroids, starting with the dose of 500 mg of methylprednisolone intravenously (i.v.) during the operation. Induction therapy was based on the rabbit antithymocyte globulin (rATG) (Thymoglobuline®, Genzyme Europe B.V., Amsterdam, Holland) in immunologically high-risk recipients (maximum panel reactive antibodies (PRA) titer > 25% and/or the presence of pre-transplant DSAs) or the anti-interleukin 2 receptor blocker basiliximab (Simulect®, Novartis Europharm Europe, Dublin, Ireland). Pre-transplant anti-HLA antibodies were evaluated by solid-phase assays using bead arrays and a Luminex platform. Pre- and post-transplant DSAs were determined using a single-antigen bead assay and results were expressed as mean fluorescence intensity (MFI). Among 4 patients who did not receive rATG induction, in one of those patients, the pre-transplant DSA were undetermined and IL-2RB was given, one patient did not received rATG or IL-2RB induction due to the lack of information concerning the presence of DSA at the time of transplantation, and 2 others had their DSA only in HLA class I and in a relatively low titers (i.e., 1992 and 541), so IL-2RB was used. Each administration of rATG was preceded by metyloprednisolone 125 mg i.v., paracetamol 1.0 g i.v., and an antihistaminic drug, and the first dose was started preoperatively and then continued with intermittent dosing based on lymphocyte count. Additionally, routine fluconazole (100 mg), valgancyclovir (labeled dose adjusted to the kidney graft function), and sulfamethoxazole-trimethoprim (2 × 480 mg) prophylaxis was given in rATG-treated patients.

#### *2.3. Primary Treatment Protocol of AMR*

After diagnosis, bortezomib (four doses, each 1.3 mg/m2) was administered subcutaneously the day after the PF session. Concomitantly, 4–8 PF sessions were performed every second/third day, with a plasma exchange rate of 2.0–2.5× patient's plasma volume. In patients who did not receive rATG induction, the routine valgancyclovir and sulfamethoxazole-trimethoprim prophylaxis was started prior to bortezomib therapy. Additionally, antibiotic prophylaxis with piperacillin/tazobactam was started and continued during the bortezomib/plasmapheresis therapy. Immediately before every subcutaneous administration of bortezomib, above i.v. metyloprednisolone, paracetamol, and an antihistaminic drug were given as a premedication.

#### *2.4. Kidney Graft Function and Protocol Biopsies*

Kidney graft estimated glomerular filtration rate (eGFR) was calculated based on MDRD (Modification of Diet in Renal Disease) formulation at the 3rd, 12th, and 24th post-transplant month.

Kidney graft protocol biopsies were performed usually at the 8th–11th post-transplant day. All biopsies were evaluated by one experienced pathologist according to the revised Banff classification [4,22–24]. Each kidney biopsy specimen was routinely stained for hematoxyllin and eosin, PAS, Masson trichrome, and silver methenamine. Additionally, SV40 antigen staining specific for polyoma BK virus infection and von Kossa staining for the presence of calcium-phosphate deposits within the tubular lumen or interstitium were performed. We also analyzed histologic signs of potential calcineurin inhibitor nephrotoxicity. Immunohistochemistry was routinely performed (CD4, CD8, CD20, CD68, and C4d) and described semi-quantitatively based on the grade of infiltration (as scattered cells, foci, clusters, groups, or diffused infiltration). AMR was diagnosed based on the following criteria: (1) the presence of histologic signs of microvascular injury (glomerulitis, peritubular capillaritis (PTC-itis), arteriitis, acute tubular injury/necrosis, (2) positive C4d staining, and (3) presence of DSAs. During the follow-up period, control protocol biopsies were performed and analyzed in 9 patients.

#### *2.5. Statistics*

Statistical analysis was performed using the Statistica software (StatSoft Polska, Cracow, Poland). Values were presented as means with 95% confidence interval (CI) or medians with Q25–Q75 quartile values. The comparison of kidney graft function before and after bortezomib treatment was performed using the Student's *t*-test. *p*-values below 0.05 were considered as statistically significant.

#### **3. Results**

#### *3.1. Baseline Characteristics*

Thirteen KTRs (7 males and 6 females) with early acute AMR treated with bortezomib-based primary therapy were analyzed. Their demographic and clinical characteristics are presented in Table 1. The mean recipient age was 53 years (minimum 30, maximum 68), mean body mass index (BMI) was 26.4 (95%CI: 24.0–28.7), and the median dialysis vintage before transplantation was 43 (IQR: 27–64) months. The history of previous blood transfusions was positive in nine, negative in one, and unknown in two patients. Out of six females, four reported past pregnancies. Only three patients had historical and two patients had the last pre-transplant panel-reactive antibodies (PRA) ≥ 25%, whereas nine patients presented positive results of virtual PRA, calculated based on the Eurotransplant Reference Laboratory HLA database version 2.0. Pre-transplant DSAs were present in twelve patients, with median MFI 10,706 (IQR: 2741–11,415) (Table 2). Induction therapy was used in twelve patients, including rATG in nine and basiliximab in three patients, respectively.

#### *3.2. AMR Diagnosis and Treatment*

AMR was diagnosed based on the first protocol biopsy, performed at a median 10 (IQR: 9–10) post-transplant day (Table 2). Due to technical constraints, the histopatologic biopsy results were available after 2–5 working days. In the majority of patients, the CD4, CD8, and CD68 infiltration was predominantly seen, whereas CD20 staining revealed only single cells or scattered foci, except in three patients, in whom, CD20 clusters and/or groups were described (Figure 1). Hence, a primary AMR treatment started at median 19th post-transplant day. In eight patients, all three Banff criteria of AMR were fulfilled. In the next four patients, the suspected diagnosis was C4d-negative AMR. In one patient, results of pre-transplant single-antigen bead assay were not available, but the screening test for class II anti-HLA antibodies was positive (Table 2). Immediately after AMR diagnosis, the primary treatment was started as described above. One patient received only 3 out of 4 planned doses of bortezomib due to the observed gastrointestinal side effects. In the first three patients, PF treatment with fresh frozen plasma (FFP) was completed. In the next eight patients, FFP together with 5% human albumin was used, in the 1:1 volume proportion. The last two patients were treated with a modified protocol (bortezomib, ATG with total dose 5 mg/kg and IVIG single dose 1g/kg) as they initially received basiliximab induction, whereas we were not able to plan and perform PF sessions at that time. Besides, dialysis therapy was required in five patients before and partially also during the AMR treatment.

**Figure 1.** Microphotograph presenting typical histologic findings in kidney transplant recipients with early antibody-mediated rejection. (**A**) Hematoxylin and eosin, magnification 100×. Interstitial edema (tubules are not back to back) with inflammatory infiltrates—asterisks, acute tubular injury (ATI) with the flattening of epithelium cells with the absence of brush border, acute tubular necrosis (ATN) with the tubular basement membranes denuded of epithelial cells, peritubular capillaritis (PTC-itis)—arrowheads, and hypoperfused glomeruli with microthrombi—arrow. (**B**) A diffused C4d staining pattern around peritubular capillaries. Magnification 200×. (**C**) Immunostaining demonstrating CD68-positive macrophages within glomeruli (a), interstitial space (b), and peritubular capillaries (c). Magnification 100×.


**Table1.**Baselinedemographicandclinicalcharacteristicsofkidneytransplantrecipientswithbortezomib-treatedantibody-mediatedrejection donor-specific antibodies; MFI, mean fluorescence intensity; GNC, glomerulonephritis; PNC, pyelonephritis; ADPKD, autosomal dominant polycystic kidney disease; DM, diabetesmellitus; rATG, rabbit antithymocyte globulin; IL-2RB, interleukin-2 receptor blocker; SGF, slow graft function; IGF, immediate graft function; DGF, delayed graft function;non available.

 n/a,



AMR, acute antibody-mediated rejection; AMR susp, the suspected AMR; Bx, protocol biopsy; POD, post-operative day; DSA, donor-specific antibodies; TCMR, T-cell mediated rejection;CNI,calcineurininhibitortoxicity.

#### *3.3. Kidney Graft Function and Survival*

After bortezomib-based primary AMR therapy, kidney graft function improved in all patients (serum creatinine concentration decreased from mean 4.6 (2.6–6.6) mg/dL before to 1.6 (1.3–1.9) mg/dL after treatment; *p* < 0.001). Mean 3-month serum creatinine concentration (SCr) was 1.35 ± 0.3 mg/dL (eGFR 55.3, 95%CI: 44.9–65.8 mL/min/1.73m2), 8/13 patients completed 12-month follow-up with mean SCr 1.2±0.3 mg/dL (eGFR 60.4, 95%CI: 45.4–75.4 mL/min/1.73m2), and 6/13 patients completed a 24-month follow-up period with mean SCr 1.0 <sup>±</sup> 0.2 mg/dL (eGFR 73.9, 95%CI: 56.7–91.1 mL/min/1.73m2) (Table 2). In post-treatment control protocol biopsies (Figure 2), four patients presented normal histology, one patient showed the partial resolution of microvascular inflammation, one patient presented mild signs of acute tubular necrosis, and one specimen was inadequate for histopatologic diagnosis. The signs of acute humoral rejection (C4d-) were still present in one patient. One patient was transferred to other transplant center and the results of her control biopsy are unknown.

**Figure 2.** Microphotograph presenting histological findings after bortezomib-based therapy of acute antibody-mediated rejection. (**A**) Hematoxylin and eosin, magnification 200×. Patent capillary lamina without signs of microangiopathy or glomerulitis (a). Prolapse of capillary tuft into the lumen of proximal tubule (b). Small foci of interstitial inflammatory infiltrates (c). (**B**) Negative (a) and nonspecific (b) peritubular capillary C4d staining. Magnification 200×. (**C**) Small interstitial foci of CD68+ cells (a). Single CD68+ cells within glomeruli (b). Magnification 100×.

In the follow-up period of median 21 (IQR: 6–30) months, both patient and kidney graft survival was 100%. Post treatment DSAs were determined in nine patients and were absent in three of them. Median MFI was significantly lower (1373 (IQR: 0–3046)) than prior to treatment.

#### *3.4. Treatment Safety*

Despite the preceded induction therapy with rATG in nine patients treated with bortezomib/plasmapheresis AMR protocol approximately 2–3 weeks later, the treatment was generally well tolerated. In one patient, the last dose of bortezomib was cancelled due to gastrointestinal toxicity. Additionally, in one patient, due to the serious hemorrhage after each of the first two PF sessions and the need of reoperation due to large hematoma, the next six PF had to be performed using citrate to avoid heparin administration. In one patient, we observed ascites of unkonwn reason, which resolved thereafter, and in another, the nasal ulceration was noticed. One patient developed a urinary tract infection during AMR treatment, and another patient presented a reccurrent upper respiratory tract infection within a few post-transplant months, which finally resolved thereafter. Based on the laboratory parameters, neutropenia < 1000 cells/μL was observed in one patient, third or fourth grade thrombocytopenia (<50,000 cells/μL) was observed in two patients, and ten patients needed a blood transfusion (median: 2 units/patient). Finally, in seven patients, the transient mild elevation of liver function tests was noted. No other serious adverse events, including neurotoxicity, opportunistic infections, or malignancy, were observed during the follow-up period.

#### **4. Discussion**

In this study, we presented our results of the first-line treatment of early post-kidney transplantation AMR based on the administration of bortezomib accompanied by plasma exchange and/or rATG and IVIG. Those early and mid-term outcomes, quantified by kidney graft function, seem to be adequate, being significantly better than previously reported outcomes in KTRs with the diagnosis of early AMR, in whom the first-line treatment was based on ATG, plasmaphereses, IVIG, and/or rituximab, and then the rescue treatment with bortezomib was applied. Additionally, the overall safety profile during and after the AMR treatment was acceptable.

The use of proteasome inhibitor in the treatment of early or late AMR was postulated because it targets antibody-producing plasma cells [25]. Notably, the vast majority of AMR episodes are diagnosed during the later post-transplant period, as pre-transplant DSA titers are increasing or de novo DSAs are produced as a consequence of substantial HLA mismatch, immunosuppressive regimen minimization, non-adherence which is increasing over time, and other relevant factors [26,27]. Hence, until now, the main evidence regarding the effectiveness of multimodal AMR treatment in KTRs is based on late AMR episodes, where its success rate is only moderate. Moreover, bortezomib was usually used as a second-line therapy, after the failure of the initial treatment. Regarding its potential utility as a first-line medication, the literature evidence is scarce. To date, after the first report of two cases [15], the same group published the more comprehensive study, including ten patients who received bortezomib-based primary AMR treatment [11]. However, in all these patients, this primary treatment consisted of bortezomib, plasmapheresis, and a single dose of rituximab. The only previous evidence of the sole effect of bortezomib/plasmapheresis as a primary AMR treatment is the comparison of ten patients treated with bortezomib-based regimen with the historic control group of nine patients who received the rituximab-based AMR regimen [16]. The 18-month graft survival was 6/10 in the bortezomib group, much worse than in our present report. However, they diagnosed AMR episodes based on the indication, not protocol biopsies, which may suggest later diagnosis and treatment as compared to our study. Secondly, there were acute and chronic AMR cases mixed together in both groups and the exact post-transplant timing of AMR therapy diagnosis and treatment initiation was not given. Additionally, all PF sessions were performed using only 4% albumin, whereas in our cohort, a fresh frozen plasma constituted approximately 50% of total PF exchange volume. These particular differences, especially those involving the timing of bortezomib therapy initiation, may partly explain the considerable difference in outcomes of kidney grafts after an AMR episode.

We decided to introduce the protocol based on the four labeled doses of bortezomib associated with the concomitant plasmapheresis sessions. In two patients without plasmaphereses, rATG and IVIG were administered instead. Of note, rixutimab was not used in study patients. At the time, as well as the negative literature concerning the efficacy of rituximab in AMR treatment, we also kept in mind the relatively high cost of our previously used rituximab protocol (approximately \$1352 USD/dose in patients with 1.8 m<sup>2</sup> of body surface). Instead, the cost of bortezomib is negligible (~\$44 USD/dose in a given patient). Our present case series results suggest an adequate effectiveness of such protocol. In fact, the mean serum creatinine concentration after 12 and 24 months was just about optimal. Of note, the previously reported kidney graft outcomes after the refractory AMR treatment were noticeably worse [7,9,13,14], with several graft losses and suboptimal kidney graft excretory function. We can only hypothesize that some specific properties of bortezomib may condition its effect in the early AMR. As was shown by Perry et al. [19], bortezomib, but not rATG or rituximab, completely abolished anti-HLA antibody production against all HLA specifities. Besides, it also induced the significant increase of apoptotic plasma cells' percentage in vitro [19]. It is possible that the removal of circulating antibodies by PF results in a rebound of their production, thereby enhancing the sensitivity to proteasome [28]. It could partly explain the efficacy of our primary approach to early post-transplant AMR, with bortezomib, but not rituximab, given immediately as a first-line agent accompanied by PF. As the 12-month serum creatinine concentration was shown to be a good predictor of the long-term

kidney graft survival [29], we may expect that the routine primary AMR therapy including bortezomib could also optimize the long-term results of kidney transplantations in this group of patients.

When analyzing the safety features, it is worth to notice that the spectrum of potentially bortezomib-related hematologic and gastrointestinal disturbances is generally similar to those observed after rATG induction, which was received by the majority of patients presented in this study. Thus, we may assume that the use of bortezomib did not increase the risk of adverse events in subjects who received both medications. Only one patient with leukopenia after initial rATG induction later presented aggravated leukopenia and neutropenia during the bortezomib/plasmapheresis treatment. Besides, in four patients who were not treated with rATG, but received bortezomib, we observed only mild leukopenia in two, thrombocytopenia in two, and slight elevation of liver enzymes activities in seven. Overall, the observed adverse events spectrum in our cohort was similar to the previously reported abnormalities obtained during the treatment of refractory AMR [7,9]. In patients with multiple myeloma, neurotoxicity often limits the use of bortezomib [30]. In this case series, however, we did not observe neuropathy associated with bortezomib use.

We are aware about the study limitations, namely its retrospective character and the low number of patients. However, we present the largest case series of KTRs with early diagnosed AMR, primarily treated with bortezomib. Heterogeneity of AMR treatment regimen is the another limitation, with 2/13 patients treated without plasmaphereses (but with rATG and IVIG instead), and with plasmaphereses performed in the first three patients only with sole albumin supplementation. Nevertheless, rituximab was not used in these patients, whereas the labeled dose of bortezomib was the core of the primary AMR therapy. Also, the last few patients had a shorter follow-up period and a lack of post-treatment biopsies and assessment of DSAs. Nevertheless, 6/13 patients completed the 24-month and 8/13 completed the 12-month observation period. Finally, the authors considered the lack of control group as a serious limitation.

In conclusion, the mid-term observation of primary bortezomib-based treatment of early post-kidney transplant AMR showed its non-inferiority as compared to previously proposed regimens and acceptable safety profile. It could encourage clinicians to perform early protocol biopsies, especially in patients with high immunological risk, and initiate such a therapeutic protocol as a first-line treatment. However, our data should be confirmed in a larger, multicenter randomized trial.

**Author Contributions:** Conceptualization, A.K. and G.P.; Methodology, G.P.; Formal analysis, A.K.; Investigation, A.K., H.K., and G.P.; Data curation, A.K. and N.S.-B.; Writing—original draft preparation, A.K. and H.K.; Writing—review and editing, G.P. and A.W. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Altered Gut Microbial Fermentation and Colonization with** *Methanobrevibacter smithii* **in Renal Transplant Recipients**

**Tim J. Knobbe 1,\*, Rianne M. Douwes 1, Daan Kremer 1, J. Casper Swarte 1,2, Michele F. Eisenga 1, António W. Gomes-Neto 1, Marco van Londen 1, Frans T. M. Peters 2, Hans Blokzijl 2, Ilja M. Nolte 3, Wouter H. Hendriks 4, Hermie J. M. Harmsen <sup>5</sup> and Stephan J. L. Bakker <sup>1</sup>**


Received: 23 December 2019; Accepted: 10 February 2020; Published: 14 February 2020

**Abstract:** Renal transplant recipients (RTRs) often suffer from posttransplant diarrhea. The observed dysbiosis in RTR may influence the fermentation processes in the gut. In this study, we aimed to investigate whether fermentation differs between RTRs and healthy controls (HCs), by measuring breath H2 and CH4 concentrations. Additionally, we determined the fecal presence of the methanogen *Methanobrevibacter smithii* (*M. smithii*), which plays a main role in the process of methanogenesis. Data from the TransplantLines Biobank and Cohort Study (NCT03272841) was used. A total of 142 RTRs and 77 HCs were included. Breath H2 concentrations in RTRs were not significantly different from HCs. Breath CH4 concentrations in RTRs were significantly lower compared with HCs (median [interquartile range (IQR)] 7.5 [3.9–10.6] ppm vs. 16.0 [8.0–45.5] ppm, *p* < 0.001). *M. smithii* was less frequently present in the feces of RTRs compared to HCs (28.6% vs. 86.4% resp., *p* < 0.001). Our findings regarding the altered methanogenesis in the gut of RTRs show similarities with previous results in inflammatory bowel disease patients. These findings provide novel insight into the alterations of fermentation after renal transplantation, which may contribute to understanding the occurrence of posttransplant diarrhea.

**Keywords:** posttransplant diarrhea; methanogenesis; *Methanosphaera stadtmanae*; mucins; sulfatereducing bacteria

#### **1. Introduction**

Renal transplantation is the preferred treatment for patients with end-stage renal disease [1–3]. Part of its success has been made possible by improved therapeutic options, such as ameliorations in surgical techniques and perioperative care [4]. Despite the success of transplantation, the burden of morbidity in renal transplant recipients (RTRs) remains high [5].

Patients often experience gastrointestinal complaints such as diarrhea, which is associated with premature kidney allograft failure and mortality, and which affects quality of life [5,6]. This posttransplant diarrhea is believed to be non-infectious and induced by the use of medication [5,7]. Recently, a study in RTRs showed that dysbiosis in the gut might cause or contribute to this posttransplant diarrhea [7]. Lee et al. demonstrated in this study that the gut microbiota diversity of RTRs with diarrhea was significantly lower than in RTRs without diarrhea. In addition, RTRs with diarrhea had a lower diversity of commensal bacterial taxa in the gut, creating a dysfunctional metabolic state. These commensal bacterial taxa are important for the degradation of complex molecules such as complex carbohydrates. During this degradation, among many other molecules, short-chain fatty acids are produced, which contributes to overall gut health [8,9]. It has been proposed that posttransplant diarrhea might be the consequence of a diminished ability to digest complex sugars [7]. A proportion of complex polymers such as fibers escape digestion and absorption in the small bowel. These complex polymers are then fermented to short-chain fatty acids (acetate, butyrate and propionate) and gases (hydrogen (H2) and carbon dioxide (CO2)) [10].

In order to maintain fermentation, it is essential that H2 concentration is reduced by H2-consuming microorganisms [11]. H2 can be used as an electron donor in sulfate respiration, methanogenesis or acetogenesis to produce hydrogen sulfide (H2S), methane (CH4) and acetate, respectively [12]. Production of H2S is most favorable, followed by the production of CH4 and acetate, respectively. However, for the production of H2S, the presence of sulfate is necessary [13]. The production of CH4 is performed by archaea. *Methanobrevibacter smithii* (*M. smithii*) and *Methanosphaera stadtmanae (M. stadtmanae)* are the two methanogens usually detected in the human gut. *M. smithii* is the predominant methanogen in the human colon [14]. Next to H2,, formate can be used for the methanogenesis as well [15]. The produced CH4 and the remaining H2 are excreted in breath and flatus. Therefore, both gases can be measured in exhaled breath [13]. Measuring breath CH4 concentrations is a simple way to investigate the metabolism of intestinal methanogens, since no significant catabolism elsewhere in the human body has been observed [12]. The presence of *M. smithii* can be measured in the feces, as has previously been performed in studies investigating patients suffering from inflammatory bowel disease (IBD) [12,16].

The dysbiosis in RTRs may influence the fermentation in the gut and the processes following fermentation, possibly leading to or contributing to posttransplant diarrhea. To gain more insight into pathogenesis of this diarrhea, we aimed to investigate the fermentation and methanogenesis in the gut in RTRs. Firstly, we aimed to investigate whether breath H2 and CH4 concentrations differ between RTRs and HCs. Secondly, we aimed to investigate whether the presence of *M. smithii* in feces differs between RTRs and HCs, and finally we aimed to identify the determinants of CH4 production.

#### **2. Methods**

#### *2.1. Study Population*

For this study we used data from the TransplantLines Biobank and Cohort Study (ClinicalTrials.gov identifier: NCT03272841). A detailed description of the study design, inclusion and exclusion criteria has been described previously [17]. In addition to the standard protocol, we measured breath H2 and CH4 concentrations and analyzed the presence of *M. smithii* in feces for the current study. (Potential) living organ donors were used as a healthy control group for comparison. Our inclusion period was between February and December 2017. The study protocol has been approved by the Institutional Review Board (METc 2014/077) (METc UMCG), adheres to the UMCG Biobank Regulation, and is in accordance with the WMA Declaration of Helsinki and the Declaration of Istanbul [17].

#### *2.2. Patient Comorbidities*

Diabetes mellitus was defined according to the guidelines of the American Diabetes Association [18]. The estimated glomerular filtration rate (eGFR) was calculated using the serum creatinine-based chronic

kidney disease epidemiology collaboration (CKD-EPI) formula. Data regarding the history of allograft rejection and primary renal disease before transplantation were retrieved from patients' medical files.

#### *2.3. Breath H2 and CH4 Concentration Measurement*

For H2 and CH4 measurements, breath samples were collected using a 50 cc syringe with an opening of 6 mm in diameter at approximately 40 cc with a 3-way-stopcock. Subjects were instructed to inhale normally and exhale fully in the syringe, with the plunger set at 50 cc and the 3-way stopcock open. After full expiration, the opening was immediately closed by the subject's finger, the plunger was set to 30 cc and the 3-way stopcock was closed. This resulted in breath samples that were not diluted by environmental air. Two breath samples were taken subsequently per study subject. Breath samples were analyzed within 12 h after sample collection. H2, CH4 and CO2 measurements were performed using a solid-state gas-chromatography device (Breathtracker SC, QuinTron Instrument Company, Inc., Milwaukee, WI, USA). The device separates the components by the basic principle of gas chromatography, using room air as the carrier gas, which is pumped through the system by an internal circulating pump. H2 and CH4 are separated from all other reducing gases and from each other, and are carried past a solid-state sensor [19]. The sensors are reported to be affected only by reducing gases, so it is unaffected by other gases in the sample; it can also employ a CO2 correction factor [19]. The analytical sensitivity is 1 ppm for H2 and CH4 and 0.1% for CO2. The Breathtracker has a linear analytical range of 2–150 ppm for H2, 2–75 ppm for CH4 and 1000–70,000 ppm for CO2. To ensure reliable breath measurements, study subjects were not allowed to smoke for at least one hour before the sample collection [20].

#### *2.4. M. Smithii Measurement in Feces*

Fecal samples were collected the day prior to the TransplantLines visit, using a FecesCatcher (TAG Hemi VOF, Zeijen, The Netherlands) and were immediately frozen after collection. The feces samples were transported in cold storage to the TransplantLines visit, and immediately stored at −80 °C (−112 ◦F) [17]. After thawing, DNA was extracted with the RBB and Qiagen method, as performed by Yu et al. with modifications described by de Goffau et al. [21,22]. To measure the quantity of *M. smithii*, real-time quantitative reverse transcription polymerase chain reaction (RT-PCR) (7500 real time PCR system, applied Biosystems, Thermo Fisher Scientific, Waltham, USA) was performed. Primers were taken as described by Johnston et al., and differentiation between *M. smithii* and other organisms in the sample was assessed using *nifH* genes [23]. The number of *nifH* genes are equal to the number of *M. smithii*, since only one gene of *nifH* is present in each *M. smithii* [24]. Analyses were performed using the Taqman machine and processed using SDSShell (Applied Biosystems, Thermo Fisher Scientific, Waltham, USA). The quantifiable presence of *M. smithii* was determined using a cycle threshold value. Values < 40 cycles were regarded as positive, and values ≥40 were regarded as negative. For analyses, CT-values ≥40 were regarded as negative and concentrations of *M. smithii* in these patients were regarded as 0 *M. smithii*/gram feces. A detailed method description is attached in Supplementary File 1.

#### *2.5. Statistical Analyses*

Statistical analyses were performed using the Statistical Package for the Social Sciences (SPSS) version 23.0 (IBM corp.; Armonk, NY, USA). In all analyses, *p* < 0.05 was regarded as statistically significant. Categorical variables are presented as *n* (%), normally distributed variables as mean ± standard deviation (SD) and non-normally distributed variables as median [interquartile range]. Normality was assessed using Q–Q plots. Differences between groups with normally distributed variables were assessed using independent T-tests. Non-normally distributed data were compared using the Mann–Whitney U test. Comparison of categorical variables was performed using a chi-square test for groups with *n* ≥ 5 and a Fisher's exact test for groups with *n* < 5. For all other tests and visualizations, the mean of the duplicate measurements of the breath H2 and CH4 concentration in breath was used. To correct for environmental CH4, 2 ppm was subtracted from each breath

CH4-measurement [25,26]. Possible determinants of breath CH4 were identified using univariable linear regression. All variables with a *p*-value <0.2 were included in a multivariable linear regression model run backward to identify the determinants of breath CH4 production. Because H2 is used by *M. smithii* for the conversion to CH4, an interaction term of H2 and *M. smithii* was added in the analysis. Log10 transformations were performed if necessary to reach conditions in all performed analyses.

#### **3. Results**

We included 219 study subjects, of whom 142 (64.8%) were RTRs and 77 (35.2%) were HCs. Among RTRs, 91 (64.1%) were male, and the mean age was 56.3 ± 13.7 years. Among HCs, 39 (50.6%) were male, and the mean age was 56.4 ± 10.6 years. Baseline characteristics are shown in Table 1. A Consort Flow diagram is presented to provide an overview of subgroups that were used in different analyses (Figure 1). Breath H2 concentrations of the RTRs were not significantly different compared with HCs (Table 1). The RTRs had, however, lower breath CH4 concentrations compared to the HCs (7.5 [3.9–10.6] ppm vs. 16.0 [8.0–45.5] ppm, *p* < 0.001). Data distributions of breath H2 and CH4 concentrations are shown in Supplementary File 2. Raw data are shown in the Supplementary data.


**Table 1.** Baseline characteristics.


**Table 1.** *Cont*.

Data are presented as mean ± standard deviation (SD), median with interquartile ranges (IQRs) or number with percentages (%). Abbreviations: BMI, body mass index; eGFR, estimated glomerular filtration rate; CH4, methane; H2, hydrogen; HbA1c, hemoglobin A1c; *M. smithii*, *Methanobrevibacter smithii.*

**Figure 1.** Consort flow diagram. Abbreviations: CH4, methane; H2, hydrogen; *M. smithii*, *Methanobrevibacter* smithii.

#### *3.1. M. Smithii in Feces*

The feces of 98 study subjects was not available for analysis. *M. smithii* abundance was analyzed in the feces samples of 77 RTRs and 44 HCs (i.e., 121 of 219 study subjects, see Figure 1). Among the RTRs, 22 (28.6%) had quantifiable concentrations of *M. smithii* in their stool samples. Among HCs, 38 (86.4%) had quantifiable concentrations of *M. smithii* in their feces samples. The median abundance of *M. smithii* in the feces of those study subjects was 5.9 <sup>×</sup> 10<sup>7</sup> [1.2 <sup>×</sup> 106–8.9 <sup>×</sup> 108] per gram feces. A quantifiable concentration of *M. smithii* was significantly less frequently observed in RTRs compared to HCs ((22 (28.6%) vs. 38 (86.4%) resp.; *p* < 0.001) (Table 1). In addition, the abundance of *M. smithii* was positively correlated with breath CH4 concentrations (*r* = 0.69, *p* < 0.001).

#### *3.2. Determinants of Breath CH4*

Determinants of breath CH4 were analyzed using linear regression analysis in all 219 study subjects, and these results are presented in Table 2. Breath H2 and the presence of a quantifiable abundance of *M. smithii* in feces were associated with higher breath CH4 concentrations (standardized beta (st. β) 0.57, *p* < 0.001 and st. β 0.94, *p* < 0.001 resp.). A negative interaction was found between both determinants on breath CH4 (st. β −0.51, *p* = 0.001), indicating that in the presence of *M. smithii* the magnitude of the correlation between H2 and CH4 in breath decreases from overt to virtually absent (*r* = 0.88, *p* < 0.001 vs. *r* = 0.09, *p* = 0.5 resp., Figure 2). In addition, the use of mycophenolate mofetil was associated with a lower breath CH4 concentration (st. β −0.18, *p* = 0.014). The described determinants explained 55.0% of the total variation in breath CH4 concentrations.


**Table 2.** Linear regression analysis of log10 breath CH4 concentration.

Abbreviations: BMI, body mass index; eGFR, estimated glomerular filtration rate; CH4, methane; H2, hydrogen; *M. smithii*, *Methanobrevibacter smithii*; St. β, standardized beta. *R*<sup>2</sup> = 0.550. \* Run backwards.

**Figure 2.** Scatterplot of log10 breath H2 and CH4 concentration by presence of *M. smithii* in feces. There is a difference between the relation between H2 and CH4 in subjects with and without *M. smithii*. Abbreviations: CH4, methane; H2, hydrogen; *M. smithii*, *Methanobrevibacter smithii*. *N* = 121. Pearson correlation in the absence of *M. smithii, r* = 0.88, *p* < 0.001. Pearson correlation in the presence of *M. smithii, r* = 0.09, *p* = 0.5.

#### **4. Discussion**

We have shown that although no significant difference in breath H2 concentration was found between RTRs and HCs, breath CH4 concentrations were significantly lower in the RTRs compared with the HCs. In addition, we found a significantly lower presence of *M. smithii* in the feces of RTRs compared with HCs. Breath H2 and the presence of *M. smithii* in feces were associated with higher breath CH4 concentrations. Moreover, the association between breath H2 and CH4 concentrations disappeared in presence of *M. smithii* in feces. Finally, mycophenolate mofetil was associated with a lower breath CH4 concentration.

The reduced breath CH4 concentration in RTRs compared to HCs which we observed might be explained by the reduced presence of *M. smithii* in the feces of RTRs, since *M. smithii* is the most abundant methanogen in the human gut [12]. One reason for the lower prevalence of *M. smithii* in the feces of RTRs may be the result of an increased presence or activity of sulfate-reducing bacteria (SRB). It has been described that a high concentration of either methanogens or sulfate-reducing bacteria is present in the feces of healthy individuals. These two groups of microorganisms appear to be competing for H2, with the prevailing group becoming the predominant organism [27,28]. However, since no mechanism of direct competition between SRB, methanogens and acetogens has been observed, at this point it is impossible to predict any dominance of one of these hydrogenotrophs [29]. If the gut in RTRs is more colonized with SRB, or if these SRB are more active, more hydrogen sulfide (H2S) is produced. H2S is highly toxic to the colonocytes and impairs their metabolic function, especially the butyrate oxidation [30]. Butyrate has a known anti-inflammatory effect and several other health-promoting functions [31]. The presence of butyrate in the lumen and the oxidation by colonocytes are both involved in the regulation of water and sodium absorption from the colon [30]. SRB and the consequent disturbance of butyrate oxidation is believed to play a key role in the pathogenesis of IBD [12]. If SRB are indeed more present in RTRs, this might also be an explanatory factor for the occurrence of any of the gastrointestinal complaints of RTRs [16], especially since the butyrate concentration in RTRs seems to be lower due to the reduced prevalence of bacteria taxa that produce butyrate [7].

In addition, more colonization with SRB, and consequently more H2S, may diminish the positive effects of the butyrate in the gut of RTRs. However, the presence of SRB was not measured in this study. Although our results regarding the decreased presence of *M. smithii* do support this hypothesis, future studies will have to further test this hypothesis.

Another reason for the lower prevalence of *M. smithii* in feces might be a lower presence of mucins in the gut of RTRs. We observed no significant correlation between breath H2 and CH4 concentrations in the presence of *M. smithii*. Therefore, the produced CH4 by *M. smithii* may be derived from endogenous substrates such as mucins, formate or other unknown substrates [12,32]. Importantly, for mucins it has not yet been settled whether they contribute to methane production or rather inhibit it, or under which circumstances stimulation may shift towards inhibition [32,33]. Mucins cover the epithelium and form a protective layer in the gut, thereby providing a protective layer against pathogenic organisms [34]. Deficiencies of mucin in the intestinal barrier are associated with an abnormal mucosal inflammatory response, which is present in IBD [34]. The role of mucins in the fermentation processes in RTRs has, to our best knowledge, never been investigated.

We also observed a strong correlation between breath H2 and CH4 concentrations in the absence of *M. smithii* in feces (Figure 2). Possibly, other methanogens that flourish in the absence of *M. smithii* are more dependent upon H2 concentrations. One of these methanogens might be *M. stadtmanae*, an archaeon that is the second most common archaeon in the healthy gut after *M. smithii* [35]. It is known that the CH4 production by *M. stadtmanae* is highly dependent upon the presence of H2 and methanol [36,37]. An increased prevalence of *M. stadtmanae* has been observed previously in IBD patients in a study by Lecours et al. [38]. Interestingly, it has been reported that *M. stadtmanae* can induce an inflammatory cytokine response from monocyte-derived dendritic cells, which may contribute to pathological conditions in the gut [39]. In order to gain more insight into gut health in RTRs, the prevalence of *M. stadtmanae* needs to be further investigated [12,39,40].

In addition, our study shows that patients using mycophenolate mofetil exhale lower concentrations of CH4. Previous studies have shown that mycophenolate mofetil is associated with gastrointestinal injury and diarrhea, although any underlying mechanisms are incompletely understood [6]. Lower CH4 levels are also associated with diarrhea [12]. Future studies may investigate whether methanogenesis plays a role in the association between mycophenolate mofetil and diarrhea.

Our observations are in line with previous studies in IBD patients [12,41]. This is interesting, since RTRs and IBD patients have similarities: both groups suffer from intestinal dysbiosis, often have diarrhea and often need to take immunosuppressive medication [38,42–44]. Scanlan et al. observed a significantly lower presence of methanogen-positive feces samples in patients suffering from ulcerative colitis compared to healthy controls (24% versus 48%). In addition, a lower presence of methanogens in patients suffering from Crohn's disease was observed (30% versus 48%), although this association was not statistically significant [41]. In another study by Ghavami et al., significantly higher amounts of *M. smithii* were found in the feces samples of HCs compared to IBD patients [16]. Our results suggest that the reduced colonization with *M. smithii*, and possibly the methanogenesis of IBD patients, might be comparable to RTRs.

It is known that CH4 reduces inflammation, oxidative stress and apoptosis in the human body [45]. Our findings show significantly lower breath CH4 concentrations in RTRs compared to HCs, while the protective properties of CH4 appear especially important in RTRs in the context of (prevention of) renal rejection, inflammation and high levels of oxidative stress [46]. Future studies may further investigate the associations of (breath) CH4 concentrations with patient outcomes, such as renal rejection. If the hypotheses regarding the protective properties of CH4 are confirmed, the relatively low CH4 levels in RTR may be a therapeutic target, since CH4 concentrations in the body can be increased iatrogenically by inhalation or injection [47,48].

No difference in breath H2 concentration was found in the current study. This is in line with other studies: the matter of hydrogenotrophics in the gut is highly complex, and is dependent upon many variables [49].

It is a limitation of our study that we did not measure mucin concentrations or potential colonization by SRB or *M. stadtmanae*. In addition, we did not measure H2 and CH4 concentrations in flatus, although it has been found that the concentration of both gases are higher in flatus than in breath when the concentrations are high [50]. No data regarding menopausal status was available in our study, although it is known that menopausal status does affect the gut microbiome [51]. Additionally, we did not measure breath H2 and CH4 concentrations and the abundance of *M. smithii* in the feces of RTRs before transplantation. Further limitations of our study are that it was performed in a single center, and that our RTRs were included at different time points after transplantation. Another limitation of this study is that for logistical reasons we were unable to analyze *M. smithii* in the feces of all our study subjects. In addition, although we measured the breath H2 concentrations in the morning, shortly after breakfast, it was in a non-fasting state. Finally, the current study uses cross-sectional data, and therefore no conclusions regarding causal relationships can be drawn.

#### **5. Conclusions**

To the best of our knowledge, this is the first study to investigate both breath and feces samples in RTRs. The study shows that breath CH4 concentration and the prevalence of *M. smithii* in feces are significantly lower in RTRs compared to HCs. Our findings regarding the altered methanogenesis in the gut of RTRs show significant similarities with previous results in IBD patients. We observed that in the absence of *M. smithii*, breath CH4 production is highly dependent on H2 concentration, while this is not the case in the presence of *M. smithii*. Apparently, methanogenesis differs significantly depending on presence of *M. smithii*. Finally, the use of mycophenolate mofetil was associated with methanogenesis. These findings provide novel insight into the alterations of fermentation after renal transplantation, which may contribute to the occurrence of posttransplant diarrhea. In addition, this study has raised important hypotheses regarding the potential role of SRB and *M. stadtmanae* in post-transplant diarrhea. Future studies are needed to investigate the role of SRB and *M. stadtmanae.* Additionally, future research may study whether altered methanogenesis is associated with clinical outcomes, such as posttransplant diarrhea.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2077-0383/9/2/518/s1, Supplementary File 1: extended description of the methods used to quantify the number of M. smithii in fecal samples, Supplementary File 2: data distributions of breath H2 and CH4 concentrations, Supplementary data: raw data.

**Author Contributions:** Data curation, T.J.K., R.M.D., M.F.E., A.W.G.-N., M.v.L., J.C.S. and S.J.L.B.; formal analysis T.J.K., R.M.D., D.K. and S.J.L.B.; methodology T.J.K., R.M.D., M.F.E., A.W.G.-N., M.v.L. and S.J.L.B.; writing—original draft preparation, T.J.K. and D.K.; writing—review and editing, T.J.K., R.M.D., D.K., J.C.S., M.F.E., A.W.G.-N., M.v.L., F.T.M.P., H.B., I.M.N., W.H.H., H.J.M.H. and S.J.L.B.; supervision, S.J.L.B. All authors have read and agreed to the published version of the manuscript.

**Funding:** R.M. Douwes is supported by NWO/TTW in a partnership program with DSM, Animal Nutrition and Health, The Netherlands; grant number: 14939.

**Conflicts of Interest:** The authors declare no conflict of interest. The funder had no role in the design of the study; in the collection, analyses, or interpretation of data, in the writing of the manuscript, or in the decision to publish the results.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **C3d-Positive Preformed DSAs Tend to Persist and Result in a Higher Risk of AMR after Kidney Transplants**

**Sooin Choi 1,**†**, Kyo Won Lee 2,**†**, Jae Berm Park 2, Kyunga Kim 3, Hye-Ryeon Jang 4, Wooseong Huh <sup>4</sup> and Eun Suk Kang 5,\***


Received: 23 December 2019; Accepted: 28 January 2020; Published: 30 January 2020

**Abstract:** C3d-binding assays have been introduced as methods for the prediction of the presence of complement-binding functional antibodies; however, the prognostic value of C3d-positive preformed donor-specific antibodies (pDSAs) has not been fully evaluated. In this study, we performed a retrospective investigation of the association of pDSAs and their C3d-binding capacity with one-year clinical outcomes. pDSAs were defined as donor-specific antibodies (DSAs) that were produced before kidney transplants (KTs) (pre-pDSAs) or within the first four weeks after KTs, owing to rebound immune response (post-pDSAs). Of 455 adult KT recipients, pre-pDSAs and post-pDSAs were found in 56 (12.3%) and 56 (12.3%) recipients, respectively, and C3d-positive post-pDSAs were found in 13 recipients (2.9%) in total. Approximately half of the C3d-negative pre-pDSAs (37/73, 50.7%) disappeared after transplantation; however, all C3d-positive pre-pDSAs (8/8, 100%) persisted after transplantation despite desensitization (*p* = 0.008). C3d-positive pDSAs were significantly associated with a higher incidence and risk of AMR (*p* < 0.001, OR 94.467–188.934). Identification of the C3d-binding activity of pDSAs before and early after KT is important for predicting the persistence of pDSAs and the risk of AMR induced by the presence of pDSAs.

**Keywords:** kidney transplant (KT); donor-specific antibodies (DSA); C3d-binding assay; antibodymediated rejection (AMR)

#### **1. Introduction**

Donor-specific human leukocyte antigen antibodies (DSAs) are a critical factor in kidney transplantation (KT), as antibody-mediated rejection (AMR) induced by the binding of DSAs to the allograft represents a major post-transplant complication. If DSAs are detected before transplantation, desensitization procedures, such as the administration of rituximab or intravenous immunoglobulin (IVIG), or plasmapheresis, may be performed to reduce the DSA titer and lower the risk of AMR [1].

Single antigen bead-based antibody identification (SAB) assays are generally used as the standard for DSA monitoring. However, the clinical relevance of all the detected DSAs remains unclear because the presence of DSAs does not always correlate with complement-mediated cytotoxicity crossmatching and may not induce AMR. Recently, C1q- and C3d-binding assays were introduced as methods for predicting the presence of complement-binding functional antibodies; however, the prognostic value of these tests remains controversial [2–8]. In particular, the clinical significance of preformed DSAs (pDSAs) with complement-binding activities has been not fully evaluated.

Human leukocyte antigen (HLA) antibodies at a titer below the SAB assay cut-off level or diluted across multiple beads that share target epitopes may not be appropriately detected in SAB assays. The titer of those cryptic antibodies can be elevated by immunological memory response shortly after KT [9]. Wiebe et al. suggested that false-negative pDSAs and their increased titer after transplantation, due to memory B cell activation, may create the false impression of de novo DSA (dnDSA) early post-transplantation. They suggested that no dnDSA was detected prior to 6 months when using two strict definitions: (1) all historic and current samples were DSA-negative, with an MFI cut-off of 300 and special attention to grouped epitopes; (2) no AMR in protocol biopsies at 6 months after transplantation [10]. Therefore, we thought that DSAs that were found only before KT should not be defined as pDSAs. In this study, we considered all of the DSAs produced within one month following KT as pDSAs and then investigated the production of pDSAs in pre- and post-transplantation and evaluated their effect on the occurrence of acute rejection and clinical outcome associated with their C3d-binding activity.

#### **2. Materials and Methods**

#### *2.1. Study Population*

Of the 560 adult recipients who underwent KT between January 2013 and July 2017 at the Samsung Medical Centre, Seoul, Korea, 455 patients (279 men and 176 women) were included in this study. Multi-organ, ABO-incompatible, or combined kidney and bone marrow transplantation cases were excluded (Figure 1). Recipients who underwent desensitization owing to a high level of panel reactive antibodies (PRA) (over 50%) without DSAs were excluded. All cases were negative for complement-dependent cytotoxicity (CDC), crossmatched on the day of transplantation. DSA status was monitored pre-transplantation (within 1 month before KT) and post-transplantation (at 1 and 4 weeks after KT). The recipients were divided into four groups according to the presence of DSAs before (pre-pDSA) and after transplantation (post-pDSA): Group 1, the pDSA-negative group (recipients without pre- and post-pDSAs); Group 2, the cryptic pDSA rebound group (comprising recipients with post-pDSA only); Group 3, the pDSA reversed group (recipients with pre-pDSA only); and Group 4, the pDSA persistent group (recipients with both pre- and post-pDSAs). To determine the effect of C3d-binding capacity on clinical outcomes, the post-pDSA positive groups (Groups 2 and 4) were subdivided according to their C3d-binding capacity. A schematic of the study is shown in Figure 2. This study was approved by the Institutional Review Board of Samsung Medical Centre, Seoul, Korea (SMC-2016-07-140-003), and the requirement for the subjects' informed consent was waived.

**Figure 1.** Study population and recipient groups according to donor-specific antibody (DSA) presence and C3d-binding capacity. Pre-pDSA, DSA confirmed before transplantation; Post-pDSA, DSA confirmed within 1 month of KT; KT, kidney transplantation; ABOi, ABO-incompatible; pDSA, preformed donor-specific HLA antibody; PRA, panel-reactive antibody; CKBMT, combined kidney and bone marrow transplantation.

**Figure 2.** The study scheme showing definitions of recipient groups and a summary of clinical outcome follow up. All DSAs presented within 1 month before KT and produced within 1 month after KT were considered as preformed DSAs (pDSAs). Pre-pDSA, DSAs confirmed before transplantation; Post-pDSA, DSAs confirmed at 1 week and/or 4 weeks after KT; KT, kidney transplantation; pDSA, preformed donor-specific HLA antibody; SAB, single antigen bead-based antibody identification assay; eGFR, estimated glomerular filtration rate.

#### *2.2. Desensitization and Immunosuppression*

The desensitization protocol consisted of the administration of rituximab (Genentech Inc., San Francisco, CA, USA), treatment with intravenous immunoglobulin (IVIG) (Green Cross, Seoul, Korea), and plasmapheresis by using a COBE Spectra (Gambro BCR, Lakewood, CO, USA) before transplantation. In recipients with a pre-pDSA mean fluorescence intensity (MFI) of ≥2,500 by SAB assay, all three desensitization protocols were performed; in recipients with a low pre-pDSA MFI of

<2,500 by SAB assay, only rituximab was administered. For the induction of immunosuppression, recombinant anti-thymocyte globulin (rATG) (Genzyme, Cambridge, MA, USA) was administered on Day 0; it was administered post-transplantation on Days 1 and 2. Basiliximab (Simulect, Novartis, Basel, Switzerland) was used to induce immunosuppression in recipients without pre-pDSA. Desensitization and immunosuppression protocol details have been described elsewhere [11].

#### *2.3. Immunologic Assays*

HLA-A, -B, -C, -DRB1, and -DQB1 loci HLA typing for donors and recipients was performed by using polymerase chain reaction with sequence-specific primer (PCR-SSP) (One Lambda, Canoga Park, CA, USA) or reverse sequence-specific oligonucleotide probes (rSSOP) (Immucor, Peachtree Corners, GA, USA).

Anti-HLA antibody measurements were performed at 1 and 4 weeks after KT. Anti-HLA class I and II IgG antibodies were tested by using a Luminex bead-based detection assay. The LABScreen Mixed kit (One Lambda, Waltham, MA, USA) was used to screen for class I and II antibodies in conjunction with the HLA Fusion software v3.0 (One Lambda, Canoga Park, CA, USA). Sera that were positive in the screening test were subsequently tested for HLA antibody specificities and the presence of DSA using the LIFECODES LSA Class I and Class II SAB kit (Immucor, Stamford, CT, USA) in accordance with the manufacturer's recommendations; the results were analyzed by using Match-It software v1.2 (Immucor, Norcross, GA, USA). All sera were subjected to SAB treatment with 50 mM dithiothreitol (DTT) for 30 min. Antibody-positive results were assigned when more than two criteria were calculated from background MFI, control MFI was calculated, and the normalization factors recommended by the manufacturer were met.

The complement-binding capacity of DSAs in sera stored at −70 ◦C was measured using the LIFECODES C3d assay kit (Immucor, Stamford, CT, USA) in accordance with the manufacturer's instructions.

#### *2.4. Data Collection and Statistical Methods*

Data describing patient characteristics and their clinical outcomes were obtained from medical records. Categorical variables were summarized by number and percentage (%) and compared among groups by using Fisher's exact test or the chi-square test as appropriate. For continuous variables, the Shapiro–Wilks test was performed beforehand to examine the normality of distributions, and then summarized with mean (SD, standard deviation) or median (IQR, interquartile range) and compared among groups using one-way ANOVA or the Kruskal–Wallis test according to the normality of their distribution.

#### 2.4.1. Recipient Characteristics

Data on demography, underlying diseases, transplantation conditions, conditioning, and immunosuppression regimens were included in the analysis.

#### 2.4.2. HLA Antibody Characterization

Preformed DSAs (pDSAs) were defined as DSAs that were produced before KT (pre-pDSAs), including Groups 3 and 4, or within the first 4 weeks after KT due to rebound immune response (post-pDSAs), such as Group 2. One of our hypotheses was that the cryptic pDSA rebound group (Group 2) would be useful for the assessment of the effects of pDSAs that were undetectable before transplantation using current antibody tests, and patients were therefore not subjected to the pre-transplant desensitization, in contrast to the sensitization in the pre-pDSA positive groups (Groups 3 and 4). Receiver operating curves (ROC) were plotted to assess SAB MFI performance in an effort to predict the C3d-binding activities of HLA antibodies. Optimal cut-offs exhibiting maximal sensitivity and specificity (Youden index) were obtained for risk assessment of the presence of DSAs.

#### 2.4.3. Clinical Outcome

To assess clinical outcomes, graft rejection rate, rejection-free survival, and graft function were evaluated. Graft biopsy was performed on Day 14 and 1 year post-KT, or whenever there was clinical suspicion of acute rejection. Biopsy results up to 400 days after KT were included in the analysis because the protocol-mandated biopsy schedule was delayed for some patients due to hospital or patient circumstances. Acute cellular rejection (ACR) and AMR were diagnosed in accordance with the Banff Criteria 2013 [12]. The rates of ACR and AMR occurrence were compared among recipient groups using Fisher's exact test or the chi-square test, as appropriate. The univariable logistic regression was repeatedly used for the four-group comparison (Groups 1, 2, 3, and 4) and also for two- or foursubgroup comparisons: Group 1, Group 4 C3d (-) subgroup, and Group 4 C3d (+) subgroup. Wald's chi-square test was used for pairwise comparison with Bonferroni's correction. Rejection-free survival rates were estimated by using the Kaplan–Meier method, and the four groups were compared via the stratified log-rank test. Graft function was evaluated using the estimated glomerular filtration rate (eGFR) at 1, 3, 6, 9, and 12 months post-KT. eGFR was calculated using the modification of diet in renal disease (MDRD) study equation. Generalized estimating equation (GEE) analyses were applied to repeated measurements of eGFR. *p* < 0.05 was considered statistically significant.

#### 2.4.4. Statistical Software

Statistical analyses were computed by using SAS v9.4 (SAS Institute, Cary, NC, USA) and SPSS v22.0 (IBM, Armonk, NY, USA). Analyse-it v5.10 (Analyse-it Software, Leeds, UK) was used for graphical analyses.

#### **3. Results**

#### *3.1. Recipient Characteristics*

Fifty-six recipients (12.3%) had detectable DSAs prior to KT (re-pDSA; Figure 1). The patients were divided into four groups according to their pre- and post-KT DSA status: the pDSA negative group (Group 1; *n* = 380, 83.5%), the cryptic pDSA rebound group (Group 2; *n* = 19, 4.2%), a pDSA reversed group (Group 3; *n* = 19, 4.2%), and a pDSA persistent group (Group 4; *n* = 37, 8.1%). All recipients with pre-pDSA (Groups 3 and 4) underwent desensitization. The recipient characteristics for each group are summarized in Table 1.

In pre-transplantation, seven recipients (12.5%) had C3d-positive pDSAs; however, the number of recipients having C3d-positive pDSAs increased to 13 recipients (23.2%) within the first month after KT. Five recipients exhibited persistently C3d-positive pDSAs before and after transplantation. HLA classes of total and C3d (+) DSAs in Groups 2, 3, and 4 are summarized in Table 2. Class II DSAs were assessed with limited loci, as –DR and -DQB1. In contrast with the higher frequencies of class I HLAs in pre-transplantation, class II HLAs were more frequent in post-transplantation, not only among total DSAs but also among C3d (+) DSAs. However, there were no significant differences in the distribution of post-pDSA class and SAB MFI between Groups 2 and 4 (*p* = 1.000; *p* = 0.327 in class I and *p* = 0.882 in class II, respectively; Supplementary Table S1).


**Table 1.** Patient characteristics.

anti-thymocyte

 globulin; CsA, cyclosporine A; MMF,

Mycophenolate

 mofetil; PD, prednisolone;

 FK, FK506; PRA, panel reactive antibody.

#### *J. Clin. Med.* **2020**, *9*, 375


**Table2.**HLAclassesandcomplementbindingactivitiesofpreandpost-transplantationdonor-specificHLAantibodiesinrecipient

\* Class II NA, not applicable.

#### *3.2. HLA Antibody Characteristics, Including Complement Binding Capacities*

In total, 105 pDSAs derived from 75 recipients (Groups 2, 3, and 4) were identified; the median number per recipient was 1.4 (range: 1–6) (Supplementary Table S2). Of these pDSAs, 81 (77.1%) were persistent pre-pDSA, and 24 (22.9%) were rebound cryptic pDSAs that were newly produced within 4 weeks post-KT. C3d-binding capacities were observed in 9.9% (8/81) of pre-pDSA and 20.6% (14/68) of post-pDSA. Among the 73 C3d-negative pre-pDSAs derived from 49 recipients, 50.7% (37/73) became negative (pDSA reversed), 43.8% (32/73) persisted as C3d-negative, and 5.5% (4/73) became C3d-positive after transplantation. Pre-pDSA SAB MFIs were significantly higher in Group 4 than in Group 3 (*p* < 0.001). Importantly, all 8 C3d-positive pre-pDSAs identified in 7 recipients persisted after transplantation, although 25% (2/8) became C3d-negative, which was significant when compared with the C3d-negative pre-pDSAs (*p* = 0.008; Figure 3, red lines).

**Figure 3.** Production and MFI changes of pDSAs within 4 weeks of kidney transplantation in different patient groups. The red lines indicate cases of C3d-positive pDSAs and the grey lines highlight cases of C3d-negative pDSAs. pDSA, preformed donor-specific HLA antibody; MFI, mean fluorescence intensity.

To determine the SAB MFI cut-offs for HLA antibodies predicting C3d-binding activities, all 1515 HLA antibodies from 112 recipients were analyzed (Figure 4). The numbers of observed HLA antibodies to A, B, C, DR, DQB1, and DPB1 were 322, 526, 71, 370, 159, and 67, respectively. SAB MFIs of C3d-positive antibodies were significantly higher than those of C3d-negative antibodies in all loci. For class I antibodies, 15.1% (139/919) of the HLA antibodies were C3d-positive, and their median SAB MFI was 9429 (IQR: 5457–16,016), whereas that of the C3d-negative antibodies was 1988 (IQR: 1271–3190). The optimal cut-off value of the total class I loci for predicting C3d-binding activities was 7797, and the area under the curve (AUC) was 0.908. For class II antibodies, 42.6% (254/596) of the HLA antibodies were C3d-positive, and their median SAB MFI was 10,341 (IQR: 6693–14,207), whereas that of the C3d-negative antibodies was 1,711 (IQR: 1002–3401). The optimal cut-off value of total class II loci was 4460, and the AUC was 0.914 (Supplementary Table S3, Supplementary Figure S1).

#### *3.3. Rejection Episodes and Graft Function*

Overall, 177 recipients (38.9%) were diagnosed with rejection episodes, including ACR and AMR, with an incidence of 168 (36.9%) and 13 (2.9%), respectively; in addition, 4 (1.7%) recipients were diagnosed with both ACR and AMR (Table 3). The incidence of ACR among the four groups was not significantly different. In contrast, the incidence of AMR among the four groups was significantly different. In both Groups 2 and 4, recipients with C3d-positive post-pDSA exhibited a significantly higher incidence of AMR (2/4, 50.0% and 3/9, 33.3%, respectively) than recipients with C3d-negative post-pDSA (1/15, 6.7% and 4/28, 14.3%, respectively). The odds ratio (OR) of AMR risk was significantly increased in the Group 2 C3d (+) subgroup and both C3d (-) and (+) subgroups in Group 4 compared with Group 1 (Table 4). Although it had marginal statistical significance (adjusted *p* = 0.0876), the C3d-positive subgroup, but not the C3d-negative subgroup in Group 2, exhibited quite different OR (OR = 0.056 and OR = 0.778, respectively) compared with Group 3.

The 1-year AMR-free survival was also significantly different among all groups (Figure 5a), and it was the lowest in the C3d-positive subgroup of Group 2, the cryptic pDSA rebound cases (Figure 5b).

During the maximum 400 day follow-up period (median 17.5, IQR 12.0–330.0), graft failure was not observed in any recipients, and the eGFR differences observed among the groups were not significantly different at any time point (*p* = 0.575; Supplementary Figure S2).

**Figure 4.** Mean fluorescence intensity distribution of 1522 Class I and II HLA antibodies according to their loci and C3d-binding capacities (**a**,**b**). Box plot, 1st to 3rd quartile range and whiskers extend to the furthest observation within 1.5× interquartile range from the quartiles. SAB, single antigen bead-based antibody identification assay.

**Figure 5.** AMR-free survival according to the presence of pre-and post-pDSAs and C3d-binding capacities. The incidences of AMR were significantly different among the groups (**a**) and when considering the presence of post-pDSA C3d-binding capacities (**b**) (*p*<0.001, respectively). The statistical difference between the Kaplan–Meier survival curves was evaluated by using the log-rank test, and *p*-values of <0.05 were considered statistically significant. AMR, antibody-mediated rejection; pDSA, preformed donor-specific HLA antibody.

#### **4. Discussion**

Both preformed DSAs present before (pre-pDSAs) and early after KT (post-pDSAs) were associated with the risk of AMR when they had C3d-binding activities. In particular, C3d-positive pre-pDSAs tended to persist after transplantation, despite the pretransplant desensitization.

#### *4.1. Cryptic DSAs*

pDSAs that were produced within 1 month of transplantation were considered as cryptic pDSAs with anamnestic reactions [10]. Using this study design, we compared the clinical effect of pDSAs with or without pre-transplantation desensitization. Our results indicated that the presence of rebound pDSA was primarily associated with AMR, followed by that of persistent pDSA, reversed pDSA, and negative pDSA.

The incidence of newly produced DSAs after transplantation, which usually occurs within 1 year of KT, with a variable median time between 6 months and 4.6 years, has been reported in 13% to 30% of pDSA-negative recipients before KT [10,13,14]. Several studies have analyzed the effect of early produced DSAs, which were defined as those produced within 1 year of transplantation [15,16], but it remained unclear whether this was dnDSA or pDSA [16]. King et al. reported that DSAs that were produced sooner than 1 month after transplantation exerted more pronounced effects on recipients and allografts than the effects of those produced later [16]. Cryptic DSAs may exist below the detection limit of the current antibody tests and can be induced rapidly by anamnestic reactions after re-stimulation by the donor graft. This is in contrast with dnDSAs, which develop gradually through the primary immune reaction associated with an indirect pathway after encountering new alloantigens [9,10]. Anamnestic reactions lead to the production of C3d-positive pDSAs, and this was related to the highest incidence of AMR in this study. Therefore, further studies regarding the management of cryptic pDSAs are necessary.

#### *4.2. Persistent pDSAs*

pDSA persistence is known to be associated with AMR. Kimball et al. reported a higher AMR incidence in persistent pDSA groups than in negatively converted groups (43% and 3%, respectively) [17]. In addition, Marfo et al. demonstrated that recipients with persistent pDSAs experienced more acute and chronic rejection (*p* = 0.006) than recipients with reversed pDSA [18]. The independent risk factor associated with persistence of pDSA was pre-transplant MFI, as shown by Redondo-Pachon et al., and class II DSAs persisted more frequently [19]. Similarly, we found that persistent pDSAs possessed higher MFIs. Although the risk of AMR was not statistically different between the persistent pDSA group and reversed pDSA group, it tended to be higher in the persistent pDSA group. The C3d-positivity (33.3%, 3/9) of persistent pDSA resulted in an increased risk of AMR (OR = 9.000, *p* = 0.0781) compared with C3d-negative group (OR = 3.000, *p* = 0.3439).


\*


*\* p*-values of < 0.05 were considered statistically significant; bold letters indicate significant results; adjusted *p*-values were computed with Bonferroni's correction for multiple comparisons. pDSA, preformed donor-specific HLA antibody; CI, confidence interval.

**Table 3.**

Comparison

 of the occurrence of rejection in recipient groups.

#### *4.3. Complement-Binding Capacities and Clinical Outcome*

SAB was designed to detect all IgG antibody isotypes, irrespective of their complement-binding capacity. IgG3 was the most potent complement binder among the subclasses of IgG and significantly affected the occurrence of rejection and graft loss after transplantation via IgG3-induced C1q-binding [20]. Honger et al. reported that IgG pre-pDSA was composed of 39% IgG1 and/or IgG3, 7% IgG2 and/or IgG4, and a 54% mixture of both complement-binding and weak/non-complementbinding subclasses [21]. These findings suggested that not all SAB-positive antibodies promoted complement activation, rejection, and graft loss. Therefore, two modified SAB assays targeting different complement derivatives were recently introduced (the C1q- and C3d-binding assays). dnDSAs harboring C1q-binding capacity were known to affect clinical outcomes such as graft survival, acute AMR, and transplant glomerulopathy [8,22–27], but the role of pDSA in AMR or poor graft survival prediction is controversial [28–31]. C3d-positive dnDSA increased the risk of graft loss, AMR, proteinuria, C4d histological staining, and rapid progression to graft dysfunction [3,6,32,33]; however, C3d-positive pre-pDSA was reported not to increase the risk of graft failure significantly [34]. In the present study, we performed a C3d-binding assay, rather than a C1q-binding assay, for two reasons. First, targeting C3d is more relevant because C3d is derived after the initiation of the complement activation cascade, so it may reflect more of the functional aspects of antibodies [7]. Second, less data are available on the significance of C3d-positive pDSAs than that of C1q-positive pDSAs. We also found that C3d-positive pDSAs tended to persist after KT, and were associated with higher AMR incidence, regardless of desensitization.

The threshold of SAB MFI, which was correlated with C1q- and C3d-binding capacities, has been reported [3,7], with the purpose of estimating the risk of identified antibodies in a timely manner before carrying out the subsequent complement binding assay. The suggested cut-off MFIs for predicting C3d-binding capacities in previous reports ranged from 4225–17,057 and 8356–15,027 for HLA class I and class II antibodies, respectively [6,32,35]. Class II HLA antibodies were reported to possess higher MFIs than those of class I in regard to C3d-binding capacity. In this study, the median MFI of C3d-positive antibodies also tended to be higher in class II than those of class I, but the optimal cut-off value of class I (MFI, 7797; sensitivity, 61.9%; specificity, 97.9%) was higher than that of class II (MFI, 4460; sensitivity, 88.9%; specificity, 83.9%). Such a discrepancy may be due to different sensitivities and specificities, as determined by different studies or the limited number of HLA antibodies analyzed in our study. Kamburova et al. reported that 95% of C3d-positive antibodies exhibit SAB MFI values of 4000 or more, but only 56% of antibodies exhibiting an MFI of 4000 or more possessed C3d-binding capacity. Based on this, they suggested that the C3d-binding capacity was correlated with the SAB MFI; however, positivity cannot be completely predicted based on SAB MFI [34].

This study has a few limitations. First, the definition of pDSAs, which included cryptic pDSAs, was not verified by using donor-specific mBCs, and the possibility of rapidly produced dnDSAs cannot be excluded. Second, the number of recipients who progressed to AMR was small, although the study was able to determine statistical significance from these data. Finally, the contribution of HLA-DP DSAs could not be estimated, as donor and recipient DP typing was not performed.

#### **5. Conclusions**

The monitoring of pDSA persistence, particularly that with C3d-binding capacities that occurred despite desensitization, and of DSA production immediately following KT likely reflects that elevated cryptic pDSAs, due to the anamnestic response, are critical for the prediction of AMR. This approach would aid the initiation of timely therapeutic intervention to reduce the risk of DSAs.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2077-0383/9/2/375/s1, Table S1: Donor-specific antibody characteristics in Group 2 and 4 recipients, Table S2: MFI of 105 DSAs from 75 recipients, Table S3: Distribution of mean fluorescence intensity of 1,515 anti-HLA antibodies in a single antigen bead assay according to their C3d-binding capacity and optimal cut-offs to predict C3d-binding capacity Figure S1: Receiver operating curves analysis of the mean fluorescence intensity of single antigen bead-based antibody

*J. Clin. Med.* **2020**, *9*, 375

identification assay performance in an effort to predict complement binding capability in class I (a) and class II (b); Figure S2: eGFR differences observed among the groups were not significantly different. eGFR was calculated by the modification of diet in renal disease (MDRD) study equation.

**Author Contributions:** E.S.K. conceived the study; S.C., K.W.L., J.B.P., H.-R.J., W.H., and E.S.K. conducted the research described; C.S. and K.K. performed the statistical analysis; all authors contributed to writing the final manuscript. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Acknowledgments:** The authors wish to acknowledge the contributions of Bok Ryo Kim, R.N., and the technical support of Jun Woo Park, M.T.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
