Twenty‐four–hour normothermic perfusion of discarded human kidneys with urine recirculation

Transportable normothermic kidney perfusion for 24 hours or longer could enable viability assessment of marginal grafts, increased organ use, and improved transplant logistics. Eleven clinically declined kidneys were perfused normothermically, with 6 being from donors after brain death (median cold ischemia time 33 ± 36.9 hours) and 5 being from donors after circulatory death (36.2 ± 38.3 hours). Three kidneys were perfused using Ringer’s lactate to replace excreted urine volume, and 8 kidneys were perfused using urine recirculation to maintain perfusate volume without fluid replenishment. In all cases, normothermic perfusion either maintained or slightly improved the histopathologically assessed tubular condition, and there was effective urine production in kidneys from both donors after brain death and donors after circulatory death (2367 ± 1798 mL vs 744.4 ± 198.4 mL, respectively; P = .44). Biomarkers, neutrophil gelatinase–associated lipocalin, and kidney injury molecule‐1 were successfully detected and quantified in the perfusate. All kidneys with urine recirculation were readily perfused for 24 hours (n = 8) and exhibited physiological perfusate sodium levels (140.7 ± 1.2 mmol/L), while kidneys without urine recirculation (n = 3) achieved a reduced normothermic perfusion time of 7.7 ± 1.5 hours and significantly higher perfusate sodium levels (159.6 ± 4.63 mmol/:, P < .01). Normothermic machine perfusion of human kidneys for 24 hours appears to be feasible, and urine recirculation was found to facilitate the maintenance of perfusate volume and homeostasis.

donors, those associated with the best posttransplant outcomes, has declined. The transplantation community has therefore increasingly turned to the use of older and higher-risk donor organs, those that would not previously have been considered acceptable for transplant. Unfortunately, many such organs are not transplanted, sometimes because of clear evidence of nonviability but more often due to uncertainty as to the quality of the organ. In the past years, there has been a resurgence of interest in hypothermic machine perfusion (HMP). There is evidence that this form of preservation is superior to static cold storage (SCS), especially in the context of higher-risk and extended criteria donor organs. 1,2 Recently, the addition of oxygen delivered in some experimental perfusion systems has been shown to further improve outcomes in HMP. 3,4 To date, however, the "gold standard" in solid organ preservation in many centers remains static cold storage. Although HMP studies have shown a decrease in delayed graft function and better survival, 1,2 improved use of marginal kidneys, including from donors after cardiac death (DCDs), has remained elusive.
Normothermic machine perfusion (NMP) provides several potential advantages over both SCS and HMP by enabling normal cellular metabolism with recovery of cellular energetic status, allowing repair of reversible injury, and facilitating functional testing of the organ before transplant through the measurement of multiple perfusion and biochemical parameters during preservation.
The direct evidence that kidney preservation by NMP is superior to SCS is based on an increasing number of clinical and experimental studies from Hosgood/Nicholson and Selzner. [5][6][7][8] The preclinical studies by Hosgood/Nicholson et al demonstrated the potential for delivering therapy to the organ during normothermic perfusion, 9,10 and the authors were first to demonstrate that a short, 2-or 1hour period of normothermic perfusion before transplant after SCS yielded significant improvements in metabolic function and reduced tubular injury compared with SCS alone. 11,12 These findings were rapidly translated into human data in a series of seminal studies that showed that a 60-minute period of normothermic perfusion after SCS made it possible to transplant marginal kidneys more reliably and with greatly improved immediate graft function compared with SCS alone. [13][14][15][16] All clinical studies to date have targeted short-term durations of perfusion (1-2 hours), with the intention of recovering cellular energetics before reperfusion in the recipient. 5,13, 16 Brasile et al described successful perfusion of isolated canine and human kidneys ex vivo hypothermically at 32ºC for 48 hours. 17 However, recent work by Selzner et al in a porcine model of kidney transplant demonstrated that the use of longer, 16-hour periods of NMP after SCS is superior to both SCS alone and short-duration NMP, in terms of both tubular injury and organ function posttransplant. 18,19 Recent phase 1 and phase 3 studies in human liver transplant have further evidenced the superiority of NMP over SCS for both conventional and extended criteria donor livers, 20,21 but to date there are no published studies of prolonged NMP perfusion of human kidneys.
The aim of the present study is to evaluate for the first time the feasibility of longer-term transportable NMP in human kidneys for periods of up to 24 hours. Urine recirculation has been used to facilitate maintenance of a constant perfusate volume without additional fluid replenishment during that time period. The purpose of this study was to establish not whether urine recirculation is superior but whether it is feasible.

| MATERIALANDME THODS
Thirteen human kidney grafts were included in this study. All   A diagram and the perfusion device are shown in Figure 1A,B. The system was designed to support kidneys ex vivo for a prolonged preservation period by using perfusion with an oxygenated suspension of

| Managementoffluidvolumelossdueto urineproduction
The rate of urine production was continuously measured in all kidneys by using a liquid flow sensor (LD20 Liquid Flow Sensor; Sensirion AG).
In kidneys without urine recirculation (n = 3), perfusate volume was maintained by immediately replacing every 20-30 mL of excreted urine with an equivalent volume of Ringer's lactate, as described by previous investigators. 18,19 In the remaining kidneys (n = 10), urine was recirculated into the circuit to maintain a constant circulating volume and to avoid the electrolyte imbalance that would result from replacement with crystalloid fluids. The effect of urine recirculation in a normothermic kidney perfusion circuit has not been reported previously.

| Perfusionsolution
During preparation of the kidney for cannulation, the perfusion circuit was assembled and primed with 1 unit of packed RBCs of the same blood group as the kidney, resuspended in 250 mL of 5% human albumin solution, giving a total perfusate volume of 500 mL. During priming of the circuit, before connection of the kidney, the perfusate was supplemented with bolus doses of cefuroxime (750 mg) and calcium gluconate 10% (10 mL to counteract the calcium binding of citrate). The pH was adjusted through titration with sodium bicarbonate 8.4%

| Biochemistry
Blood gas analysis was performed by using an in-line blood gas analyzer (CDI 500, Terumo Medical Coroporation, Somerset, NJ). Glucose measurements were performed off-line using a hand-held blood gas analyzer (iSTAT, Abbot Point of Care Inc., Princeton, NJ), which was also used to further confirm Po2 and Pco2, and pH arterial and venous measurements where needed. Biomarker levels in perfusate samples were extrapolated from the standard curves.

| Histology
Core needle biopsy specimens were fixed in Millonig solution and pro-

| Statisticalanalysis
For statistical evaluation of the data (excluding ELISA), unpaired t tests (parametric) and Mann-Whitney tests (nonparametric) were performed using GraphPad Prism 7. A P-value of <.05 was considered significant. Table 1 summarizes organ retrieval characteristics and the reasons the 11 included kidneys had been declined for clinical use.

| Donorcharacteristics
For logistical reasons, discarded human kidneys offered for research typically have a longer CIT than do most of the transplanted kidneys. Figure  Information. An additional 3 kidneys (2 DCD and 1 DBD) were perfused without urine recirculation, which led to a significant shorter time on the perfusion device compared with the 10 kidneys with urine recirculation (P < .0001). All demographic parameters were similar in the 2 groups. Volume loss due to urine production was replenished 1:1 with Ringer's lactate solution.
Reasons for termination of the kidney perfusions were (i) arterial flow ≤50 mL/min, (ii) pH <7 or >7.7 measured at a Pco 2 level of 5, and/or (iii) sudden cessation of urine production.

| Hemodynamicandmetabolic functionparameters
Perfusion parameters for the overall cohort, with or without urine recirculation, and the mean ± SD values for DBD and DCD kidneys are shown in Table 3. Physiological mean arterial pressures and flows could be achieved in both donor categories in the urine recirculation group and in 2 of the kidneys without urine recirculation. The hemodynamic parameters are shown in Figure 3A,B (arterial flow) and 3C,D (intrarenal resistance).

| Arterialpressure
The arterial pressure was regulated by changing the pump speed with the aim of maintaining a physiological mean arterial pressure between 70 and 100 mm Hg throughout the perfusion. Detailed pressure values per kidney are shown in Table 4.

| Arterialflow
Arterial flow for each single kidney is shown in Figure 3A,B and Table 4. Overall, the best arterial flow was achieved in kidney 5, a DCD kidney with a CIT of >24 hours and with a mean arterial flow of 509.6 ± 25.2 mL/min. This kidney had a significantly better flow rate during the entire perfusion of 24 hours than all other kidneys (P < .0001). In kidneys 4 and 6, kidneys with an extraordinary long CIT of >100 hours (respectively discarded because of poor perfusion and histological signs of cortical necrosis, and because of patchy perfusion 8 ), excellent arterial flows were achieved, significantly better than the flow rates in kidney 2 (P = .001; both DBD kidneys from TA B L E 1 Organ procurement parameters and reasons for discard of individual kidneys

| Intrarenalresistance
Intrarenal resistance (IRR) was calculated by dividing pressure at a specific time point by the flow at the same time point. The changes of IRR over time are shown in Figure 3C,D and Table 4. As with arterial flow, IRR was lowest in kidney 5 compared with the other perfused kidneys; mean IRR was 0.15 ± 0.01 mm Hg/mL per minute (P < .0001). The highest IRR levels could be observed in kidney 10: 1.2 ± 1.02 mm Hg/mL per minute.

| Po 2 andPco 2 levels
Mean arterial Po 2 and Pco 2 levels for each kidney are displayed in
The course of pH for all perfused kidneys is shown in Figure 4A,B and displayed in

| Glucosemeasurements
The Evaluation of glucose consumption in the case of an isolated kidney is not readily achievable, because there is not only glucose consumption but also gluconeogenesis, glycolysis, glucose filtration, and glucose reabsorption taking place at the same time.

| Lactatelevelsduringperfusion
The lactate levels for each single kidney are shown in Table 4. The
Median urine production of kidneys without urine recirculation Urine sodium levels of all except 2 kidneys, with and without urine recirculation, were low with sodium <100 mmol/L. In kidneys 4 (histologically proven cortical necrosis) and 7 (polyuric during perfusion, histologically proven acute-on-chronic hypertensive nephropathy), the urine sodium levels were the same as the perfusate values, 142 ± 3.9 mmol/L, potentially indicating tubular injury with lack of sodium reabsorption capacity. Hourly amounts of urine are displayed in Figure 5A,B for each kidney.

| Biomarkermeasurements:NGALandKIM-1
Perfusate  (Table 5) and reflects a constant release of biomarkers into the perfusate during the process of normothermic preservation in a fully closed circuit with urine recirculation.
To show differences between the kidneys, the delta concentrations of all biomarkers were calculated (value of last measurement minus value of first measurement) and the different profiles are shown in Figure S5A,B. The perfusate levels of both NGAL and KIM-1 decreased over time in the group without urine recirculation, kidneys 9-11. The exact values are shown in Table 5, and the different biomarker profiles are displayed in the Figure S5C,D.
KIM-1 could not be detected at all in kidney 9. It is worth noting that the absolute biomarker levels across the 2 groups without and with urine recirculation should not be directly compared, due to the replenishment of excreted urine with Ringer's lactate solution in 1 group.

| Histologyresults
Tissue samples were obtained from all kidneys after 24 hours/at the end of perfusion. Acute tubular injury, of different severity Vacuolation is a nonspecific finding that might be attributed to osmotically active perfusate compounds, such as mannitol (this might be the cause in our case), and to sustained ischemic damage.
In this particular case, KIM-1 immunoexpression was observed immunohistochemistry results for all kidneys are displayed in Table 6. Three of our consecutive normothermic kidney perfusions were with volume replenishment (using Ringer's lactate), as performed by previous investigators, 17,18 to provide a "standard" comparison to our urine recirculation group. In these kidney perfusions, it was not possible to maintain a physiological pH after 4-6 hours of perfusion.

| D ISCUSS I ON
The most significant difference was the higher sodium perfusate levels in the group without urine recirculation. Despite the small sample size, urine recirculation seems to be an optimal tool to keep a physiological perfusate composition for long-term normothermic kidney perfusion.
In Further investigations will be required to determine whether a physiological perfusate sodium level or other urine metabolites are essential for maintenance of acid-base balance.
Compared with results from normothermic liver perfusion, 25 lactate did not seem to correlate with a poor or good performance of the kidney during NMP-not in the urine recirculation group or in the group with urine replenishment. According to the literature, it is most likely that lactate levels are related with active glucose metabolism in the kidney-both glycolysis and gluconeogenesis. [26][27][28][29] The lactate metabolism of the native human kidney plays a major role and the renal cortex appears to be the most important lactateconsuming organ after the liver. 26 Bartlett et al showed years ago that the renal function affects lactate and glucose metabolism. 27 The kidney contributes to glucose homeostasis via different pathways: gluconeogenesis, glucose filtration, glucose reabsorption, and glucose consumption. 28 There seems to be a corticomedullary glucose-lactate recycling system. The medulla consumes glucose via active glycolysis and generates lactate. The cortex has the ability to take up the lactate released by the medulla and uses it for oxidation and gluconeogenesis. 27 Furthermore, Bartlett et al demonstrated in their early publication that lactate production also correlates with the urine flow rate and sodium resorption. However, they stated that lactate consumption did not correlate with renal function. 27 Although the numbers in our study are too low to demonstrate a potential correlation with urine flow rate and sodium resorption, we also could not see any correlation between perfusion parameters and histology results. short-term normothermic kidney perfusion. Higher levels of urine NGAL, but not KIM-1, were associated with a raised donor creatinine before organ retrieval. 31 Interestingly, in our study, kidney 4, which appeared to be the one with the most significant ischemic injury, lowest urine output, and cortical necrosis on biopsy, had the greatest increase in NGAL. In the Hosgood analysis, KIM-1 levels were not associated with perfusion parameters or renal function in the donor. 25 An increase of KIM-1 in a closed perfusion circuit could be of interest in future, but the numbers in the present study are insufficient to draw robust conclusions.
Alongside its potential role as an early biomarker of acute tubular injury, 32 it has been suggested that the shedding of KIM-1 in the kidney undergoing regeneration is an active process that allows dedifferentiated tubular cells to scatter on denuded patches of the basement membrane and reestablish a continuous epithelial layer. 33  TA B L E 6 Histology results and KIM-1 immunohistochemistry, tubular condition fibroelastosis, or arteriolar hyalinosis) was not associated with the development of acute tubular injury after NMP. In all except 1 kidney, the baseline tubular condition was preserved or improved. Only in kidney 2, a poorly perfused DBD kidney from an elderly donor, did the tubular condition deteriorate (vacuolation).
We have demonstrated, for the first time, that NMP of the human kidney is feasible for 24 hours and that the recycling of urine is an effective method of maintaining perfusate homeostasis and acid-base stability. The preliminary data on perfusate biomarkers are encouraging and require corroboration with posttransplant data in due course. The ability to maintain (and possibly improve) the condition of donor kidneys of marginal quality for long enough to carry out viability assessment could increase the feasibility to exploit this important source of donor organs.

ACK N OWLED G M ENTS
The authors gratefully acknowledge funding and support from the