Observations on the ex situ perfusion of livers for transplantation

1Alanine transaminase

2Aspartate transaminase

3Donation after brain death

4Donation after circulatory death

5Donor risk index

6Hematoxylin and Eosin

7Hepatic artery

8Ischemic cholangiopathy

9Ischemic type biliary lesion

10United Kingdom Donor Liver Index

11Magnetic resonance cholangiopancreatography

12Normothermic ex situ liver perfusion

13Portal Vein

14United Kingdom

15Reliable metrics to determine whether a liver is viable and safe to transplant are currently not available. Normothermic perfusion of the liver affords an opportunity for viability assessment. Although there have been case reports and small series of liver transplants following ex situ normothermic perfusion,1, 2, 3, 4, 5 there are as yet no validated criteria predicting liver viability.

16This paper describes our observations on the biochemistry and perfusion characteristics of 47 human livers that were normothermically perfused, of which 22 were transplanted. We used readily available biochemical and physical measurements that could be analyzed during perfusion. From these studies, we have demonstrated criteria that may be helpful in determining viability and identified others that are not as discriminatory as has been suggested.5, 6

17Livers that had been considered unsuitable for transplantation by all UK liver centers, and high‐risk livers which we elected to assess by perfusion prior to implantation, were included. All were preserved initially with University of Wisconsin static cold storage solution.

18Approvals from the local research ethics committee and the National Health Service Blood and Transplant (NHSBT) research governance were obtained for the ex situ perfusion of human livers. Transplantation of perfused livers was also approved by our institution's new interventional procedures committee. Informed consent was given by all recipients in respect of both the liver they received and the use of normothermic ex situ perfusion.

19Livers were perfused using the Liver Assist (Organ Assist, Groningen, Netherlands).7 The perfusate comprised leucocyte‐depleted red cells (approximately 1 litre) and either 1 litre of succinylated gelatin (Gelofusine, BBraun Medical, UK) (n = 41) or Steen solution (Xvivo Perfusion, Göteborg, Sweden) (n = 6). This was supplemented with heparin, magnesium sulphate, calcium chloride, amino acids, with sodium bicarbonate added to achieve a pH > 7.2 before NESLiP began; a 2 μg/h epoprostenol infusion ran throughout the perfusion. Bile salts were not administered. Perfusion was commenced at 20°C for all livers; the perfusate was warmed to 35‐37°C over the first 10‐30 minutes. Mean hepatic artery (HA) pressure was increased from 30 mmHg at the start of perfusion to 60 mmHg at ≥35°C; portal vein (PV) pressure was increased from 4‐6 mmHg to 8‐10 mmHg over the same period. Perfusate samples were taken at 10 and 30 minutes, and every 30 minutes thereafter. Bile was sampled when it appeared in adequate volumes for assay. Bile and perfusate gases were analyzed immediately using a Cobas b221 benchtop analyzer (Roche); other biochemistry samples were frozen and stored at −70°C until assayed. Nontransplanted livers were biopsied at hourly intervals. Vascular resistance was calculated as the perfusion pressure (mmHg) divided by the flow (L/min). Table 1 and Supplementary Figure S1 detail the perfusion protocol evolution throughout the study.

20The glycogen content of frozen right lobe liver biopsies was estimated using Zhang's adaptation of an acid hydrolysis method.8, 9 Briefly, liver biopsies were homogenized with either sodium hydroxide or hydrochloric acid. The supernatant was then mixed with Glucose Assay Reagent (Sigma Aldrich) and the glucose content measured.

21Liver biopsies were placed in formalin and stained with hematoxylin and Eosin (H&E). Following perfusion nontransplanted livers were sliced transversely to include the major duct branches and formalin fixed. Stromal necrosis of septal bile ducts was scored on a scale of 1 to 4 after Hansen et al10 where Grade 1 and 2 corresponded to <25% and 25‐50% circumferential necrosis respectively and Grade 3/4 to >50% circumferential necrosis.

22A total of 47 livers, 12 from brain dead (DBD) and 35 from circulatory death (DCD) donors, were subject to NESLiP (Table 2) between January 2014 and October 2017. Of the 28 livers accepted for possible transplantation, 20 were transplanted while 2 of the 19 livers accepted with no intent to transplant were also transplanted. Median posttransplant follow‐up is 601 days (IQR 257‐752 days).

23Of the 22 transplanted livers, the first 12 have been reported previously.7 One recipient died following primary nonfunction (T6), one (T12) developed early allograft dysfunction by Olthoff's criteria (day 7 INR1.6),11 and 4 recipients have developed ischemic‐type biliary lesions (ITBL) at 33 to 89 days posttransplant of whom 3 required retransplantation. None of the last 16 cases suffered reperfusion syndrome or vasoplegia, something that affected cases T2 to T6.7

24HA resistance was almost 25‐fold higher than PV resistance, with resistance being highest at the onset of perfusion and falling to reach steady state by 60 minutes; there was no difference between transplanted and nontransplanted livers (Figure 1). There was no correlation between the 60‐minute HA or PV resistance and transaminase release, a marker of hepatocellular damage.

25Alanine and aspartate transaminase (ALT and AST, respectively) concentrations in the perfusate plateaued by 2 hours in most livers; the 2‐hour ALT was therefore taken to be representative of the degree of hepatocellular damage (Figure 2). ALT and AST were closely correlated (r =.94, P <.001).

26ALT concentrations were measured in the effluent of 28 livers after flushing with compound sodium lactate (Hartmann's) solution prior to NESLiP. The effluent ALT was correlated with the ALT of the perfusate 2 hours after starting NESLiP (Figure 3A, r =.78, P <.0001; AST r =.649, P =.031). Similar data were obtained for AST (not shown). There was a correlation between the two‐hour perfusate ALT and the peak ALT in the first seven days posttransplant for the 22 transplanted livers (Figure 3B, r =.73, P =.0001). In the 12 transplanted livers where effluent ALT had been measured there was no correlation with the peak ALT posttransplant (r =.18, P =.58).

27The baseline lactate concentration varied and was higher in livers flushed with Hartmann's solution (Figure 4). Fifteen livers (32%) exhibited an initial rise in lactate within the first 30 minutes of NESLiP before falling. The concentration fell to below 2.5 mmol/L by 180 minutes in all but one (2%) liver. The peak rate of fall of lactate varied between livers from 3.5 to 21.5 mmol/L/h (2.7 to 16.3 mmol/L/kg/h, Supplementary Figure S2); there was no correlation between this and ALT or AST (data not shown).

28In spite of the absence of exogenous glucose in the perfusate, the perfusate glucose concentration rose above 10 mmol/L in all but five livers, and exceeded 40 mmol/L in one case (Figure 5). Where the glucose subsequently fell, the peak was reached by 2 hours. The rate of fall varied between 0.67 and 3.8 mmol/L/kg/h, appeared to follow zero order kinetics, and was independent of insulin, whether given as continuous infusion or as boluses of up 150units (data not shown). Insulin was therefore omitted from the last 35 NESLiPs.

29To investigate whether the glucose rise could be attributed to glycogenolysis, glycogen estimations were made on wedge biopsies of four livers (Supplementary Figure S3). Glycogen content ranged from 3.8 to 47.1 mg/g liver at the start of NESLiP, falling initially during perfusion before increasing as the perfusate glucose concentration fell.

30Twenty‐seven livers that had been cold stored in UW solution for a median 435 mins (range 222 to 1148) were flushed with a liter of Hartmann's solution prior to NESLiP. The glucose concentration of the resulting effluent ranged from 2.7 to 22.1 mmol/L (1.2 to 12.6 mmol/L/kg), and was proportional to the peak glucose during NESLiP (Supplementary Figure 4, r =.72, P <.001), suggesting that the rate of glycogenolysis was also proportional. Neither effluent glucose nor peak glucose were strongly related to cold ischemic time prior to NESLiP (r = −.11, P =.57 for effluent; r =.38, P =.04 for peak glucose). Effluent glucose showed a trend to being higher in DCD livers than DBD livers (DBD median 9.7, IQR 5.5‐11.2; median DCD 12.2, IQR 6.8‐22.1 mmol/L).

31Two of the four livers which did not generate a perfusate glucose above 10 mmol/L were interrogated with additional glucose (Figure 6). Liver T10 had a 7.5 g infusion over 90 minutes, following which the glucose was seen to fall, and liver R23 was given a 2.5 g glucose bolus, again with a fall.

32The perfusion protocol required bicarbonate administration to maintain a pH ≥ 7.2 in the HA perfusate, hence bicarbonate replacement is a useful surrogate of the liver's ability to regulate pH (Supplementary Figure 7). The median bicarbonate requirement was 10 mmol (range 0 to 60, IQR 0 to 20), with most supplementation occurring during the first two hours. Three livers required more than 30 mmol of bicarbonate, one of which (T6) was transplanted and suffered primary nonfunction.

33Recordable bile production began between 30 and 150 minutes after the start of perfusion, with the delay being in part attributable to dead space in the biliary catheter (Figure 7). Bile production varied from 0 to 60 mL after four hours. There was no relation between the rate of bile production and posttransplant function, incidence of cholangiopathy in transplanted livers or degree of stromal necrosis in nontransplanted livers.

34Contrary to previous reports,12 common hepatic duct histology was not informative with loss of superficial epithelium being common, presumably due to handling and cannulation artefacts. Therefore, the technique of preparing research livers postperfusion was changed to produce transverse sections through the hilum which included septal ducts; 15 research livers were treated in this way (Figure 8).

35The peri‐biliary vascular plexus was rarely seen on H&E, and neither was hemorrhage; the perfusate used leucocyte‐depleted erythrocytes so ischemia‐related inflammation could not be graded by leucocyte infiltration, and when present leucocytes suggested infiltration in the donor prior to retrieval. Vascular endothelium was identified using a CD31 specific stain, and loss of capillary CD31‐staining corresponded to the areas of stromal necrosis. Eight livers showed grade 3 or 4 stromal necrosis, two grade 2 (one of which produced no bile), and five grade 1 (minimal) stromal necrosis; Figure 8 illustrates normal intrahepatic ducts and those with severe ischemic damage.

36Bile was analyzed at intervals during perfusion, the frequency of measurement depending on the amount of bile produced. Figure 9 shows the bile pH during perfusion. Of the 16 transplanted livers for which biliary pH was measured, the 3 that were unable to produce bile with a pH > 7.4 (livers T7, T11, and T16) developed cholangiopathy. The other transplanted livers all achieved a biliary pH > 7.5, often following an initial lower pH at the start of perfusion. Of the 11 research livers for which biliary pH was measured, four achieved a pH > 7.5 and these had minimal stromal necrosis of the major intrahepatic ducts (grade 1) on histological examination. Seven of the remaining eight livers had a bile pH < 7.5, most <7.4 and histology revealed moderate or severe stromal necrosis (grade 3 or 4) in addition to epithelial loss and loss of peribiliary glands.

37Bicarbonate, which is actively secreted by cholangiocytes,13 and glucose, which is actively resorbed by functioning cholangiocytes,14 were also measured in bile and related to the perfusate concentrations. Ducts which went on to develop cholangiopathy, and those with grade 3/4 stromal necrosis were associated with lower bile bicarbonate and a bile glucose content similar to perfusate, while ducts that had good histological appearance or did not develop cholangiopathy had a lower glucose (Figures 10 and 11).

38Validated guidelines to assess liver function and viability during normothermic ex situ liver perfusion do not exist. The available literature hints at HA and PV flow, lactate clearance,5 and bile production,6, 15 with the only clinically based evidence being derived from six human liver perfusions.5 Our observations, made on 47 livers declined for transplantation or which we considered to need pre‐implant assessment, document the behavior of less‐than‐ideal livers subject to NESLiP and identifies parameters that might be valuable in assessing livers at the margins of acceptability.

39HA and PV flow are dependent on perfusion pressure and resistance within the vascular bed, the latter being the property intrinsic to the liver. We show that resistance in HA and PV falls in the first 60 minutes of perfusion before a steady state is reached. The resistances did not appear to correlate with outcome or biochemical markers of hepatocellular injury. Resistance is commonly used as a discriminator in hypothermic kidney perfusion, although evidence for its predictive utility is poor.16, 17 Similarly there is evidence in hypothermic liver perfusion that resistance does not reflect warm ischemic injury.18

40Lactate clearance is used clinically as a marker of immediate function post–liver engraftment, and it is understandable that it has been attractive in the assessment of livers undergoing NESLiP. However, the massive capacity of a liver coupled with the small volume of the perfusion circuit (2 litres in our case) means that lactate clearance can occur with little viable hepatocellular function. Indeed, in our entire cohort of 47 high‐risk livers, only one failed to clear lactate below the 2.5 mmol/L 3‐hour threshold suggested by Mergental et al5 while the liver that suffered primary nonfunction reached this threshold within 90 minutes. The rate of lactate fall per unit liver weight may be a more appropriate measure of hepatocellular function. In addition, lactate concentration may be deceptively high in the presence of trauma or lobar ischemia,7 where lactate may be being produced by the ischemic liver segments. Hence, while a fall in lactate does not predict a viable liver, livers that were not able to clear lactate rapidly (T6, R10, Figure 4) were associated with parenchymal injury. Pre‐loading the circuit with lactate either by using a lactate containing fluid to washout the liver or using unwashed red cells, improves the utility of this test allowing rate of fall to be better assessed.

41We have confirmed previous observations that the glycogen content of the liver falls during the initial perfusion period, and is replenished in the later stages of NESLiP,19, 20 suggesting that most of the glucose rise seen during perfusion originates from glycogenolysis. This process probably also explains the high glucose concentrations seen in cold storage fluid (Supplementary Figure S4), necessary for anerobic ATP generation. Glycogenolysis is known to be stimulated by glucagon, epinephrine and other stress hormones,21 so may start at the time of treatment withdrawal in DCD donors, and coning in DBD donors, times characterized by high levels of such hormones; it is noteworthy that glucose levels were higher in DCD donor livers. Lactate is metabolized into glucose in an oxygen‐dependent process and this will contribute in part to the raised perfusate glucose observed at the start of NESLiP.22, 23

42The subsequent spontaneous fall in glucose during NESLiP probably in large part represented glucose entering the liver via the insulin‐independent GLUT2 bidirectional transporter, although some may be entering circulating erythrocytes via the GLUT1 transporter.24 As the glucose enters the liver it is probably incorporated into glycogen, initially in zone 3.20 The role of insulin in NESLiP is uncertain, since it is not necessary for glucose entry into the liver but may be important for other metabolic processes although its absence does not appear to have been deleterious in our perfusions. Insulin is known to inhibit glycogenolysis and stimulate glycogenesis, but high glucose concentrations may be a stronger stimulant at the levels seen during most perfusions, and we have demonstrated glycogen synthesis in the absence of exogenous insulin (Supplementary Figure S3). In our initial experiments we included an insulin infusion, and later a larger insulin bolus, but neither had an effect on glucose metabolism so insulin was thereafter omitted from the protocol. Where glucose levels did not rise during perfusion there are two possibilities, one being that there has been severe hepatocellular damage and the other that there is minimal ongoing glycogenolysis, possibly due to glycogen exhaustion. A low perfusate glucose was seen in the liver that developed primary nonfunction. We subsequently demonstrated that interrogation with a glucose challenge can demonstrate metabolic activity, and commend this as a simple test.

43In this paper, the perfusate transaminase concentrations relate to the 2‐liter circuit, and most liver perfusate concentrations had reached a steady state by 2 hours, implying no ongoing damage. Although we do not have sufficient data to support a transaminase threshold, we have successfully transplanted a liver with a perfusate ALT of 5576iU/L (AST 4963) at 2 hours, while one with an ALT of 9490iU/L (AST 13349) suffered primary nonfunction. All but 6 of the 47 livers undergoing normothermic perfusion had an ALT <6000iu/L at 2 hours. We preferred to use ALT as a marker of hepatocellular damage rather than AST, since AST may also arise from hemolysis on the circuit.

44We have shown that there is considerable transaminase release prior to NESLiP, with effluent levels correlating closely with the levels observed during perfusion. Measurement of effluent transaminase concentrations may be a useful test to perform before transplantation whenever concern exists about a liver, and may be useful to indicate those livers where further assessment and manipulation during normothermic perfusion maybe helpful. High concentrations of transaminase washing out of the liver before NESLiP suggest that significant hepatocellular damage occurred as a consequence of events during retrieval and cold storage, rather than due to reperfusion per se.25 We also suggest that future trials of NESLiP record effluent transaminase as an aid to interpreting the relative contribution of NESLiP to posttransplant transaminase levels, a common endpoint in such trials.4

45The liver regulates acid–base balance by the differential metabolism of glutamine along the hepatic lobule, and requires integrity of both zone 1 hepatocytes and the peri‐central glutamine synthetase containing hepatocytes.26, 27, 28 Three livers undergoing NESLiP were unable to regulate pH without significant or continued requirement for bicarbonate supplementation, one of which was transplanted and suffered primary nonfunction. We chose pH ≥ 7.2 as the target for our perfusions. The liver receives blood from HA, typical pH 7.4, but also from PV, typical pH 7.2 (personal observation). HA oxygenation is also greater than PV clinically, as it is in our perfusions. This is not the case in other commercial machines, so the extent to which our observations may apply on other equipment, with different perfusion conditions, remains to be seen.

46We found no evidence to support the notion that bile production affected outcome following transplantation. One liver that suffered primary nonfunction following transplant had an intermediate bile production, and livers that developed cholangiopathy were among the highest and lowest bile producers. In contrast to quantity, the pH of bile appears to provide a marker of biliary integrity. Our experience in a limited number of perfusions suggests that livers which are able to produce bile with a pH > 7.5 in the first 4 hours of perfusion do not exhibit histological features of extensive stromal necrosis of the major septal ducts or suffer ischemic‐type biliary lesions when transplanted; the converse applies when the pH ≤ 7.4. Necrosis of the stroma of the larger septal ducts was taken as a histological marker of irreversible bile duct damage in this study, and its presence was highlighted by the absence of capillaries staining with CD31.

47Bile should normally contain little glucose, since while canalicular bile has a concentration similar to plasma, cholangiocytes reabsorb glucose from bile resulting in a low glucose,14 and this provided an additional marker of bile duct viability. A bile pH < 7.4 and glucose <10 mmol/L lower than perfusate glucose suggested significant stromal injury. Although it is possible that low perfusate pH may affect the ability to generate a pH > 7.5, this did not appear to be the case although high glucose gradients did pose more of a challenge to cholangiocyte reabsorption. Larger studies will be needed to validate the predictive role of biliary pH and glucose, and remove any bias in our observations.

48Absence of a commercial source of clinical grade bile salts meant we were not able to include them in our protocol. The use of bile salts will affect bile production and may affect its pH and the incidence of cholangiopathy, but initial reports from a multicenter trial where they were used suggest cholangiopathy also occurs even when taurocholate is present (David Nasralla, personal communication). Perfusions in excess of 8 hours in the absence of bile salts have been shown to result in biliary damage in pigs,29 hence we limited our perfusions to under 8 hours.

49The incidence of cholangiopathy in transplanted livers (4/22 = 18%) and grade 3 or 4 septal duct stromal necrosis in nontransplanted livers (8/15 = 53%) suggests that NESLiP, per se, does not prevent biliary damage. This may be because the damage relates to events prior to commencement of NESLiP. However, cholangiopathy has not been reported in DCD livers undergoing hypothermic oxygenated perfusion (HOPE).30 It is possible that this difference relates to biliary susceptibility to reperfusion injury by reactive oxygen species, and its mitigation by slow introduction of oxygen that characterises HOPE;31, 32 equally, the high resistance and consequent low initial arterial flows that characterise normothermic reperfusion may result in further warm ischemic injury to the ducts. The etiology of biliary injury during NESLiP need further evaluation, and will be aided by our identification of biochemical markers of cholangiocyte damage.

50Other biochemical observations, not highlighted, may also provide corroborative evidence of function and damage. Urea rose during perfusion signifying metabolic activity. Perfusate potassium rose initially in all perfusions, and continued to rise where the transaminase release was high but fell in the livers exhibiting less damage and better metabolic activity (data not shown).

51As this is an observational study, where the perfusion protocol evolved over time as experience accrued, there are limitations that need to be kept in mind. The significance of lower perfusate oxygenation on the incidence of vasoplegia and post reperfusion syndrome has been discussed elsewhere.7 We evaluated Steen solution as a replacement for Gelofusine in six perfusions but it was more expensive and the results were not superior. Recognition of a possible association of biliary pH with bile duct damage resulted in a number of livers being declined due to a low bile pH and high bile glucose. Table 3 shows the criteria we currently use to determine whether or not the liver is safe to transplant, but we fully acknowledge that early in our experience we turned down livers that we would now consider viable. Similarly, Table 3 is not meant to represent definitive criteria, but rather illustrates our current “comfort zone.”

52In summary, this report of the 47 normothermic perfusions we have undertaken has identified patterns of biochemical response to ischemia and reperfusion. The redundancy of metabolic activity in a liver perfused on a small circuit, together with its remarkable regenerative capacity which permits significant hepatocellular injury to be tolerated, confound attempts to define thresholds. Nevertheless, we do believe that we may have demonstrated a simple marker of biliary integrity, as well as casting doubt on the significance of bile production and thresholds of portal and hepatic artery flows as absolute markers of viability.

53The authors of this manuscript have conflicts of interest to disclose as described by the American Journal of Transplantation. Lucy Randle now works for OrganOx, who make the Metra normothermic liver perfusion device; Andrew Butler is part owner of a patent on the circuit used by the OrganOx Metra device. The other authors of this manuscript have no conflicts of interest to disclose.