Substantial Increases Occur in Serum Activins and Follistatin during Lung Transplantation

Background Lung transplantation exposes the donated lung to a period of anoxia. Re-establishing the circulation after ischemia stimulates inflammation causing organ damage. Since our published data established that activin A is a key pro-inflammatory cytokine, we assessed the roles of activin A and B, and their binding protein, follistatin, in patients undergoing lung transplantation. Methods Sera from 46 patients participating in a published study of remote ischemia conditioning in lung transplantation were used. Serum activin A and B, follistatin and 11 other cytokines were measured in samples taken immediately after anaesthesia induction, after remote ischemia conditioning or sham treatment undertaken just prior to allograft reperfusion and during the subsequent 24 hours. Results Substantial increases in serum activin A, B and follistatin occurred after the baseline sample, taken before anaesthesia induction and peaked immediately after the remote ischemia conditioning/sham treatment. The levels remained elevated 15 minutes after lung transplantation declining thereafter reaching baseline 2 hours post-transplant. Activin B and follistatin concentrations were lower in patients receiving remote ischemia conditioning compared to sham treated patients but the magnitude of the decrease did not correlate with early transplant outcomes. Conclusions We propose that the increases in the serum activin A, B and follistatin result from a combination of factors; the acute phase response, the reperfusion response and the use of heparin-based anti-coagulants.

Introduction ACTRN12609000763246). Consent was provided in writing. None of the transplant donors were from a vulnerable population and all donors or next of kin provided informed consent that was freely given.

Study Design
Serum samples from 46 of 60 patients participating in the prospective, parallel arm randomized controlled trial of RIC in patients, 18 years or older, undergoing bilateral sequential lung transplantation were used. They were stratified by diagnosis (restrictive disease, cystic fibrosis (CF) or non-CF) [24] The RIC involved 3 cycles of lower limb ischemic conditioning performed just prior to the reperfusion of the first lung transplant. Patients receiving extra-corporeal membranous oxygen and/or cardio-pulmonary bypass after surgery were excluded from the study, as there were too few to reliably estimate the effects of extra-corporeal membranous oxygen and/ or cardio-pulmonary bypass on activin A, activin B, follistatin or the other cytokines we examined.

Transplantation
Lungs retrieved from donations both after brain or cardiac death, were preserved by a single ante-grade flush using Perfadex (Vitrolife, Goteborg, Sweden). All patients received immunosuppression with tacrolimus and azathioprine pre-operatively and methyl prednisolone (1 gram) intra-operatively (500 mg at both anaesthetic induction, and first donor lung reperfusion). Patients also received 2500-5000 IU boluses of heparin to maintain activated clotting times at twice baseline with an average total dose 5000 IU of heparin per patient.

Sample collection and immunoassays
Blood samples were obtained just after anaesthesia induction, after RIC or sham treatment, and 15 minutes, 2 hours, 8 hours and 24 hours after lung reperfusion. After clotting, serum was separated and stored frozen until used. Total serum activin A, B and FS were measured by ELISA and by specific radioimmunoassay, respectively, with normal ranges based on samples from 138 normal males and females [25,26] providing a reference maximum (RM). The other cytokines were measured using the Human Th1/Th2 CBA Kit II and the Human Chemokine CBA Kit (BD Biosciences, Mountain View, California, USA), with data acquired on an LSRII cytometer and analysed using FCAP Array Software (BD Biosciences).

Statistics
Statistical analysis was performed in SPSS Statistics 22 (IBM, Chicago, USA) and graphed in GraphPad Prism 6.01 (GraphPad, California, USA). Analyses conducted were linear regressions, mixed model linear repeated measures (RIC or sham treatment, diagnosis) with an unstructured repeated covariance design, and one-way ANOVAs (diagnosis). Bonferroni post-hoc tests were used as appropriate. To correct for multiple comparisons, Bonferonni's correction was applied. Proportions in Table 1 (patient characteristics) were compared using a Chi square test, with exact statistics calculated. Other comparisons in Table 1 used a Student's ttest. Bootstrapping with 10,000 samples was used to generate 95% confidence intervals. Data presented as mean ± SEM. P<0.05 is regarded as statistically significant.

Data transformation and missing values analysis
Some results >6 SD from the mean (6 data points: 1 IL6, 2 CCL2, and 3 CXCL8) were set to missing. Data was otherwise not transformed or modified prior to, during or after analysis. Missing data was not replaced and was randomly distributed across time; however, certain assays had high rates of missing values due to the multiplex assay kit (TNFα, IFNλ, IL2 and IL4). Missing partial pressure of arterial oxygen/fraction of inspired oxygen ratios (PF) ratios result from patients being extubated after which arterial catherization was no longer available. N values should be noted when interpreting results. Missing data analysis demonstrated that multiple imputation, linear interpolation and mean of nearby points would not reliably estimate missing data, given the relatively few patients in this study. Therefore, we used a mixed model linear repeated measures design with an unstructured repeated covariance design, as noted above, enabling use of all available data in the model.
The data set needed for replication have been uploaded to Figshare and are available through the following link: http://figshare.com/article/Substantial increases occur in serum Activins and Follistatin during lung transplantation /1547913

Patient and Donor Characteristics
There were no statistically significant differences between groups who received RIC or the sham treatment (Table 1). RIC did not change the proportion of patients alive at the end of follow-up, the length of time patients spent in intensive care, or total hospital stay length. Details of the diagnostic groups comprising the patient cohort are provided in Table 1.

Serum Activin A concentrations compared to reference values
Baseline serum activin A concentrations were above the reference maximum (RM) (>0.283ng/ mL) in 17.4% of patients ( Fig 1A).

Serum activin B concentrations compared to reference values
Baseline serum activin B was above the RM concentration (>0.156ng/mL) in 21.7% of patients ( Fig 1B) The levels increased at the RIC/sham treatment and 15 minutes post-transplant timepoints; only one patient remaining below the RM. The serum activin B levels remained elevated above the RM in 68.2% of patients at 2 hours, in 67.4% at 8 hours and in 37.2% of patients at 24 hours.

Serum follistatin concentrations compared to reference values
Baseline serum FS levels were above the RM (>25.42ng/mL) in 8.7% of patients ( Fig 1C). The FS levels increased in the sample taken after the RIC/sham treatment, with the levels in 68.9% of patients being above the RM. At 15 minutes post-transplant reperfusion, 43.5% of patients were above the RM and decreased further with only 3 patients having levels above the RM at 2 hours post-transplant. FS levels were within the normal range in all patients subsequently.

Ratio of activin A to follistatin concentrations, a marker of activin A bioactivity, compared to reference values
The ratio of activin A to FS, an established surrogate measure of activin bioactivity, was greater than the RM ratio (>0.0283) in 17.4% of patients at baseline (Fig 1D). After the RIC/sham treatment, the levels in 71.1% of patients were above normal and remained elevated at 15 minutes post-transplant in 63% of patients. The levels decreased subsequently and at 2 hours posttransplant only 13.6% of patients had elevated levels, decreasing to 11.6% at 8 hours and to 7.1% at 24 hours post-transplant. No patient remained above normal for the whole study.

Ratio of activin B to follistatin concentrations, a marker of activin B bioactivity, compared to reference values
Activin B to FS ratios were greater than normal in 17.4% of patients at baseline ( Fig 1E) and increased after RIC/sham treatment, with 77% of patients above normal. This proportion decreased to 58.7% above normal at 15 minutes post-transplant, to 34.1% at 2 hours post-transplant, increasing at 8 hours post-transplant to 66.7% but by 24 hours post-transplant had decreased to 26.2% of patients above RM. Three patients remained above normal throughout the study period.

Correlations between peak values
The peak activin A level and peak activin B level from each patient were not correlated (p = 0.810). Peak activin A and peak FS levels from each patient were correlated (r 2 = 0.228, p = 0.0005) as were peak activin B and peak FS levels (p<0.0001). There were no significant correlations between the lowest PiO2/FiO2 ratio from each patient (6, 12 and 24 hour time-points) and any of the following measurements, peak serum activin A concentration (p = 0.749), peak serum activin B concentration (p = 0.104), peak serum FS concentration (p = 0.208), peak activin A/FS ratio (p = 0.315), and peak activin B/FS ratio (p = 0.594).

Correlations between cytokines
It is not possible to report all cytokine correlations, so we present correlations that were consistent across multiple time-points and therefore are likely to be biologically relevant in Tables 2  and 3. Activin A and B levels were mostly correlated across time, apart from after the RIC/ sham treatment and 15 minutes post-transplant. Activin A, activin B and FS were consistently correlated across all time-points, except for activin B and FS at 8 hours post-transplant. Details concerning all the correlations between activins A, activin B and follistatin and all the other cytokines measured are given in S1 Table. Effect of diagnostic grouping on serum activin A and B, follistatin and other cytokines Serum activin A concentrations were significantly affected by diagnostic grouping and by diagnostic group across time (Fig 2A). Post-hoc analysis revealed that serum activin A levels at both 2 hours and 8 hours post-transplant were significantly greater in patients with obstructive CF than in obstructive non-CF patients compared to obstructive-non-CF (p = 0.017).
Serum activin B concentrations differed across time with diagnostic group as a factor ( Fig  2B). At 2 hours post-transplant, patients with obstructive CF had significantly higher concentrations than those with obstructive non-CF (p = 0.017) vs obstructive CF) and patients with restrictive disease (p = 0.049).
Follistatin concentrations were affected by diagnostic group over time (Fig 2C). At 2 hours post-transplant FS levels were elevated in patients with obstructive CF compared to those with obstructive non-CF (p = 0.05). At 8 hours post-transplant, FS was still elevated in the obstructive CF group, compared with both obstructive non-CF (p<0.0001) and restrictive pulmonary disease (p = 0.001).
Across the 24 hrs, activin A to FS ratios were significantly different between diagnostic groups, patients with restrictive pulmonary disease had significantly higher ratios compared with obstructive non-CF (p = 0.021) (Fig 2D). Post-hoc analysis demonstrated that basal ratios were elevated significantly (p = 0.04) in patients with restrictive disease compared to those with obstructive non-CF. These remained significantly elevated at 15 minutes post-transplant (p = 0.036) and at 2 hours post-transplant (p = 0.042).
The activin B to FS ratio was not quite significantly (p = 0.053) different between diagnostic groups (Fig 2E). At the conditioning/sham time-point, ratios in patients with obstructive non-CF were elevated compared with obstructive CF (p = 0.015). This continued at 15 minutes posttransplant (p = 0.002).
The diagnostic grouping of patients identified significant correlations with some of the other cytokines (S1 Fig). There was no effect of diagnostic group on serum concentrations of  The serum levels of CXCL8-IL8 (A), CCL5 RANTES (B), CXCL9 MIG (C), CCL2 MCP1 (D), and TNF (E), measured over the 24 hour time course of this study, showed a significantly decreased response to the transplantation surgery in the CXCL 9 and CXCL10 response in patients undergoing RIC but no other significant changes (S1 and S2 Figs)

Effect of RIC on serum concentrations of activins, follistatin and other cytokines
RIC did not change serum activin A across time or at any individual time-point (Fig 3A) but reduced the average serum activin B levels over 24 hours (RIC: 0.33 ± 0.03, n = 24, Sham: 0.42 ± 0.030 n = 22p = 0.029; Fig 3B) and post-hoc analysis showed this was primarily due to decreased serum levels in the RIC exposed patients at the condition/sham time-point (p = 0.033) and the 15 minute time-point (p = 0.003) (Fig 3B). There was also a significant decrease in the response of CXCL8 and CXCL 10 in the RIC treated patients (S4 Fig). The 24 hour average serum FS level was also decreased by 15% in RIC treated compared with sham treated patients (p = 0.047) (Fig 3C). Comparing RIC treated with sham patients,  The ratio of activin A to FS and activin B to FS were mostly unaffected by RIC treatment or sham treatment (Fig 3D and 3E). RIC significantly lowered the activin B to FS ratio by 19% compared with sham treated patients only at the 15-minute post-transplant time point (p = 0.037) (Fig 3E). Predictors of PiO2/FiO2 ratios measured at 6, 12 and 24 hours In this analysis data included for each time point only includes data collected prior to that time point. Factors correlated with PiO2/FiO2 at 6 hours were ischemic time for the 1 st lung Given the low number of cases and sparse direct associations across time, further modelling was not attempted as results would almost certainly not be generalizable to other patient populations.

Discussion
This study establishes that substantial increases in serum concentrations of activin A, B and FS occur during lung transplantation surgery. Since activin A is a key cytokine regulating proinflammatory and pro-fibrotic processes [27,28], this study raises the possibility that the substantial increases in activin A and B could compromise the function of transplanted lungs. Only further long term functional studies will answer this concern and determine if there is a need to change procedures, medications or other interventions that could decrease the magnitude of these increases in activins A and B.
The increases in basal activin A and B levels in these patients likely arise from the pathophysiological processes causing the need for lung transplantation. Both clinical and animal studies demonstrate that high activin A levels cause inflammation and the resultant fibrosis compromises lung function [27,29]. Elevated serum activin A levels occur in patients with a variety of inflammation-based lung disorders such as CF, bleomycin-induced fibrosis, allergybased inflammatory models and idiopathic pulmonary fibrosis [27][28][29][30] and are consistent with the capacity of elevated activin A levels to cause hepatic and renal fibrosis [16,31].
The rapid increase in serum activin A concentrations associated with lung transplantation partly reflects the increase triggered by the acute phase response to surgery [18,32,33]. The same mechanisms are likely to be the cause of the novel finding of elevated serum activin B levels. Given that the magnitude of the increases of activin A and B surpass those associated with the expected acute phase response, we suspected that other factors were involved.
The most likely cause of the magnitude of the rapid increase in activin A and B is the use of heparin anti-coagulation. 5,000-10,000 units of heparin are administered to the transplant recipient shortly after anaesthesia induction. The transplant donor also receives a substantial dose of heparin during transplant donation (up to 25,000IU given to brain dead donors), and additional heparin is administered to the transplanted organ (up to 25,000IU).
In sheep, we demonstrated that heparin injections led to a rapid dose-related increase in serum activin A and FS [20] due to the negatively charged heparin releasing FS and activin A bound to cell surface proteoglycans via a series of positively charged amino acids (arginine) in the FS sequence [19]. This binding site is also the mechanism by which FS-bound activin A is targeted to lysosomal degradation [15,34]. The capacity of heparin to release FS and activin A was also shown in patients undergoing cardiac procedures using heparin [21,22]. Using a sheep model of cardiopulmonary bypass, Chen et al [21] showed that increases in serum activin A and FS did not occur when non-heparin anticoagulants were used. It is possible that the magnitude of the increase in activin A is in fact greater since all patients received significant doses of methyl prednisolone as a prelude to transplantation. Previous studies have shown that glucocorticoids will decrease activin A levels [11].
We propose that heparin use and the acute phase response are the likely cause of the novel finding that serum activin B levels also increase at this time point. Further studies are necessary to determine the pathophysiological actions of the elevated activin B concentrations. While the early functional capacity of the transplanted lung did not show any correlation with the magnitude of the activin A and B released during transplant surgery, further long term studies are needed to clarify this issue.
RIC significantly decreased the 24 hour average levels of activin B and FS, and seemed to blunt the maximal rise of activin B and FS. Further, the area under the curve in Fig 3, representing the 24 hour activin A levels are lower but not significantly and when combined with the decrease in activin B post-RIC, may explain the lower levels of FS during the same period, as both activins are determinants of FS production [11,18]. However, these changes did not affect the lowest P/F ratio or the average P/F ratio across the 24 hours. These findings warrant a larger study of RIC will be required to confirm its effects on the release profile of activin A, B and FS. Apart for activins A and B, the only other cytokine to be affected by RIC was CXCL9 that shows a marked decrease in response to RIC.
We note that of all the cytokines measured, the activins A and B are consistently the two cytokines which increase very early in lung transplantation surgery, a pattern similar to the activin A response to LPS [13]. These observation support the view that activin A rapidly responds to the IRI during the initial reperfusion of the transplanted lung and continues in those patients randomized to receive RIC and is the likely initiator of the inflammatory response [13]. As discussed earlier, the magnitude of this response is augmented by the use of heparin. Further studies are required to determine if activin B also has the capability to initiate inflammatory responses [13]. Since 1μg of FS, given 1 hour prior to an LPS challenge, by binding and neutralizing the actions of activin A, markedly changed the cytokine response to LPS such as halving the release of TNFα, emphasizes the capacity of activin A to act as a master controller of the cytokine response. The rapid response of both activin A and B to IRI also supports the concept that they are the "initiators" of this response. As such, changing the type of anti-coagulant used could substantially reduce the release of the activins. Alternatively, administering FS just prior and during the reperfusion of a transplant could minimize any resultant functional impairment of the transplant.
The peak activin A levels released during lung transplantation were of greater magnitude than those associated with death in patients with septicaemia or with higher mortality in patients in intensive care units with acute respiratory failure [25,35]. It is possible that the concomitant release of FS may protect the transplant and other tissues from the damaging actions of activin A that include apoptosis of hepatocytes and B lymphocytes [36,37]. The latter may affect the morbidity associated with transplantation surgery, outcomes that could be clarified by comparing the use of heparin and non-heparin based anti-coagulants.
As heparin is administered to most donors prior to organ collection for transplantation, the use of non-heparin anticoagulants offers a simple way of resolving this issue. Alternatively, administration of FS to the donor at lung collection or adding it to preservation solutions may decrease organ damage. Further, FS use during perfusion of lungs on a rig prior to transplantation may also be worthy of assessment to preserve lung function [6].
The large increases in activin A and B during transplant surgery are clearly influenced by many factors including the cause of the patient's lung injury, the acute phase response and the dose of heparin administered during the entire transplant surgery. Further studies to assess the effects of heparin versus non-heparin anti-coagulants during lung transplantation will clarify these issues and determine if a change in anti-coagulant practice can improve patient outcomes.  Table. The correlations between the activin A, activin B and follistatin levels, their ratios and all the other cytokines measured are detailed in S1 Table. (DOCX)