Low circulating regulatory T-cell levels after acute rejection in liver transplantation

Authors


Abstract

Immune regulatory CD4+CD25+ T cells play a crucial role in inducing and maintaining allograft tolerance in experimental models of transplantation (Tx). In humans, the effect of Tx and immunosuppression on the function and homeostasis of CD4+CD25+ regulatory T cells (Tregs) is not well characterized. In this study, the frequency of Tregs in liver transplant recipients was determined based on flow cytometric analysis of CD4, CD25, CD45RO, and cytotoxic T lymphocyte antigen (CTLA)-4 markers, and the suppressor activity of Tregs was assessed in a mixed-leukocyte reaction. A link between Tregs, acute rejection, and immune-suppressive treatment was investigated. Liver transplant recipients had significantly higher Treg levels in peripheral blood pre-Tx than healthy controls. After Tx, a significant drop in the Treg fraction was observed. This reduction of circulating Tregs was transient and was associated with immunosuppression. In recipients who did not develop rejection, a relative recovery of Treg levels was seen within the first year after Tx. Recipients who experienced an episode of steroid-treated acute rejection, however, had sustained low Treg levels. The suppressive activities of CD4+CD25+ Tregs from rejectors, nonrejectors, and healthy controls on proliferation and interferon (IFN)-γ production were indistinguishable. In conclusion, the percentage of CD4+CD25+CD45RO+CTLA-4+ quadruple-positive Tregs in peripheral blood decreases significantly after liver Tx. Treatment with methylprednisolone during Tx and for acute rejection is associated with low circulating Tregs. Despite these quantitative differences between rejectors and nonrejectors, the suppressive quality of CD4+CD25+ Tregs is identical in both groups. Liver Transpl 12:277–284, 2006. © 2006 AASLD.

Transplant tolerance, the ultimate goal in solid organ transplantation (Tx), occurs more often after Tx of the liver compared to other organs. Cessation of immune-suppressive therapy without allograft rejection has been reported to be successful in a considerable proportion of liver transplant recipients.1, 2 The exact mechanisms involved in achieving transplant tolerance remain unknown, although animal models suggest a possible role for regulatory T cells (Tregs).3, 4 In vitro studies of a distinct subset of Tregs expressing CD4, the α-chain of the IL-2 receptor (CD25), and the transcription factor Foxp3 showed that these cells do not proliferate upon stimulation, but instead suppress activation of effector T cells in a cell-contact-dependent manner.5–7 Transfer of CD4+CD25+ Tregs from animals with long-term surviving allografts to naive recipients prevents the development of allograft rejection.8 The suppressive capacity of CD4+CD25+ Tregs is not only restricted to foreign antigen-driven T-cell responses, but also entails autoreactive T-cell responses, thereby preventing the development of autoimmune diseases and maintaining peripheral tolerance to self-antigens.9, 10

In vivo studies on the mechanism of Treg-mediated suppression have shown a functional role for cytotoxic T lymphocyte antigen 4 (CTLA-4, CD152), which is constitutively expressed by CD4+CD25+ Tregs.11, 12 Interestingly, the majority of suppressive CD4+CD25+ T cells are found in the memory T-cell population, characterized by the expression of CD45RO.13 Jonuleit et al.14 have demonstrated that within the CD4+CD25+ fraction the CD45RO-positive cells, and not the CD45RO-negative cells, are anergic and have suppressive activity. These observations suggest that these cells have already encountered antigens and have acquired the phenotype of highly differentiated CD4+ T cells, distinguishing them from recently activated CD4+ T cells, which do not express CD45RO.

In the context of human liver Tx, the function and dynamics of Tregs have not been extensively studied. The aim of this study was to reveal the effect of liver Tx and immune suppression on Tregs in peripheral blood of liver transplant recipients. A relationship between the frequency of Tregs and the development of acute rejection was investigated.

Abbreviations

Tx, transplantation; Treg, regulatory T cell; IFN, interferon; CTLA, cytotoxic T lymphocyte antigen.

MATERIALS AND METHODS

Patients and Healthy Controls

After receiving an informed consent, heparinized peripheral blood was obtained from 40 liver transplant recipients before and multiple time points after Tx. Ten patients developed acute rejection within 3 months of Tx and were treated with intravenous high doses of methylprednisolone (Solumedrol, 3 × 1,000 mg). Acute rejection was confirmed by histological examination of liver biopsies using the Banff classification (1997). A rejection activity index of 6 or more defined rejection, as assessed by an experienced pathologist. Steroids (Prednisone) were given to all patients, starting with 100 mg/day, and were weaned during the first 6 months after Tx. Some patients remained on maintenance steroids (n = 14). Blood samples obtained from 16 healthy volunteers (6 males, 10 females) were used as a control. The mean age of the controls was 31 yr, not significantly different from that of the patient group. General characteristics and immune-suppressive treatment of patients are summarized in Table 1.

Table 1. Patient Characteristics
PatientGenderAge (yr) at TxDiseaseInduction therapyMaintenance therapy
IL-2R blockerSteroids (Prednisone)I.V. MP (500 mg)CNISteroids (Prednisone)Acute rejection
  1. Abbreviations: AC, alcoholic cirrhosis; PBC, primary biliary cirrhosis; PSC, primary sclerosing cholangitis; HBV, chronic hepatitis B virus infection; HCV, chronic hepatitis C virus infection; Wilson, Wilson's disease; AIH, autoimmune hepatitis; CC, cryptogenic cirrhosis; Byler, Byler's disease; PLD, polycystic liver disease; I.V. MP, intravenous methylprednisolon; Tac, tacrolimus; CsA, cyclosporin A.

1M49AC++Tac++
2M40AC++Tac
3M46AC++CsA+
4M45AC++CsA+
5M58AC+++Tac+
6F56AC+++Tac
7M52AC+++Tac
8F54AC++Tac++
9M47AC+++Tac+
10M61AC+++Tac
11F49PBC++CsA++
12M37PSC+++Tac++
13F44PSC++Tac++
14F44PBC++CsA
15M34PSC++Tac
16F46PBC+++Tac+
17M38PSC+++Tac
18M53PBC+++CsA+
19F50PSC+++CsA+
20F41PBC+++Tac
21M43HBV++CsA
22M34HBV++Tac
23M45HBV+++CsA
24M38HBV+++CsA
25M59HBV+++CsA
26M54HBV+++Tac
27M59HCV+++Tac
28M61HCV++Tac+
29M55HBV++Tac+
30F52HBV++Tac
31F40APUDoma++Tac+
32M43Wilson+++CsA
33M59AIH++CsA+
34F40CC+++Tac+
35M37Byler+++Tac+
36M58CC+++Tac
37M35PLD+++Tac
38M49CC+++Tac
39M47AIH+++Tac+
40F43PLD+++Tac

Flow Cytometric Analysis

Mononuclear cells were obtained from heparinized blood by density gradient centrifugation over Ficoll-Paque plus (Amersham Biosciences, Buckinghamshire, United Kingdom). After isolation, cells were stored in 10% dimethylsulfoxide-containing medium at −180°C. For flow cytometric analysis of Tregs, the following fluorescent monoclonal antibodies were used: anti-CD4 (SK3)-PerCp-Cy5.5, -CD25 (2A3)-FITC, and -CD45RO (UCHL-1)-APC purchased from BD Pharmingen (San Diego, CA). PE-conjugated anti-CTLA-4 (BN13) was purchased from Immunotech (Marseille, France). Fluorescent mouse immunoglobulin G-1 and -2A (BD Pharmingen) antibodies were used as isotype controls.

After thawing, cells were washed twice with phosphate-buffered saline containing 0.3% bovine serum albumin and incubated for 30 minutes with CD4, CD25, and CD45RO mAbs in phosphate-buffered saline/0.3% bovine serum albumin at 4°C. Following primary incubation, cells were washed, and for staining of intracellular CTLA-4, the cells were fixed and permeabilized using the IntraPrep Reagents (Immunotech). Subsequently, the cells were washed and analyzed by flow cytometry using FACS Calibur and CELLQuest Pro software (Becton Dickinson, San Jose, CA). The percentage of Tregs was calculated as quadruple-positive cells (CD4+, CD25+, CD45RO+, and CTLA-4+) and expressed as a percentage of CD4-positive cells.

Mixed-Leukocyte Reaction and IFN-γ Production

RPMI 1640 medium with L-glutamine (Bio Whittaker, Verviers, Belgium) supplemented with 10% pooled human serum (Department of Immunohematology and Bloodbank, Leiden University Medical Center, Leiden, The Netherlands), 100 μg/mL penicillin, and 100 μg/mL streptomycin (Gibco, Paisley, UK) was used for T-cell culture. After isolation of PBMCs from fresh heparinized blood obtained 1 yr after Tx, CD4+ cells were purified using the untouched CD4+ T-cell isolation kit (Miltenyi Biotec, Bergisch Gladbach, Germany). After washing with phosphate-buffered saline/0.3% bovine serum albumin, CD4+ T cells were incubated with anti-CD25 microbeads (Miltenyi Biotec) followed by a positive selection of CD4+CD25+ T cells according to the manufacturer's instructions. The CD4+CD25 fraction was used as responder cells. The purified Treg fraction contained more than 90% pure CD4+CD25+ T cells.

Donor or third-party allogeneic spleen cells were irradiated (5000 rads) and mixed (1:1 ratio) with CD4+CD25 responder T cells (5×104) in 96-well round-bottom plates. Suppression of proliferation was determined by adding increasing numbers (5 × 103, 1 × 104, or 1.5 × 104 cells) of CD4+CD25+ T cells to the mixed-leukocyte reaction. At day 4 of culture, supernatants were collected and the concentration of IFN-γ was measured by enzyme-linked immunosorbent assay (U-CyTech, Utrecht, The Netherlands). At day 5, cultures were pulsed with 1 μCi per well of [3H] thymidine (Amersham, Little Chalfont, UK) for 16 hours. Cells were harvested and proliferation was assessed by measuring radioactivity with a liquid scintillation counter. Cultures were performed in triplicate and the mean counts per minute were calculated.

Statistical Analysis

Statistical analysis of the flow cytometry data was performed using software package SPSS version 11.5 (SPSS, Chicago, IL). Significance was tested using the Mann-Whitney U test, the Wilcoxon paired test, and the Spearman's 2-tailed correlation test. P values less than 0.05 were considered significant. For the mixed-leukocyte reaction and IFN-γ production, statistical analysis was performed by analysis of the logarithmic transformation of the dependent variable with random intercept and random slope using PROC Mixed in SAS version 8.2. (SAS Institute, Cary, NC).

RESULTS

Increased Tregs in Patients with End-Stage Liver Disease

Tregs were identified in peripheral blood by flow cytometry based on the expression of CD4, CD25, CD45RO, and intracellular CTLA-4. Hereafter, quadruple-positive cells (expressing CD4, CD25, CD45RO, and CTLA-4) are referred to as Tregs. Representative dot plots of the flow cytometric analysis from a liver transplant recipient are shown in Figure 1A. Approximately 20% of the CD4+CD25+ cells express both CD45RO and CTLA-4, whereas 2% of the CD4+CD25 fraction was positive for these markers (not shown). The majority of CTLA-4-expressing CD4+CD25+ cells had the memory phenotype (CD45RO+).

Figure 1.

(A) Assessment of Tregs in peripheral blood of liver transplant patients and healthy controls on basis of CD4, CD25, CD45RO, and CTLA-4 (CD152) coexpression. Representative dot plots are shown from a liver transplant patient. The right panels were gated on the CD4+CD25+ cells in the upper left plot. The majority of CD4+CD25+ cells have a memory phenotype (CD45RO+) and approximately 20% of these cells coexpress CTLA-4. (B) Elevated levels of Tregs in patients with end-stage liver disease. The percentage of Tregs is calculated as quadruple-positive cells (CD4+, CD25+, CD45RO+, and CTLA-4+) and expressed as the percentage of CD4-positive cells. Patients with chronic liver disease (CLD) had increased Tregs in peripheral blood compared to healthy controls (HC, **, P < 0.001). Treg levels in the 4 main types of CLD were all significantly different from healthy controls, with highest levels of Tregs in bile duct diseases and viral hepatitis (*,P < 0.05). Shown is the mean ± standard error of the mean.

Patients with end-stage liver disease had significantly higher Treg percentages than healthy controls (Fig. 1B). Highest levels were seen in patients with bile duct diseases and viral hepatitis, whereas modest elevations were found in alcoholic cirrhosis and for other liver diseases.

Decrease of Tregs after Liver Tx

Recently, we have shown that CD4+ and CD4+CD25+ T lymphocytes within the CD3+ population were significantly reduced in peripheral blood 1 yr after liver Tx.15 Assessment of Tregs, expressing CD4, CD25, CD45RO, and CTLA-4, showed a similar reduction (Fig. 2A). Lowest levels were observed at 3 months after Tx, followed by a relative increase at 12 months and at later time points. A significant correlation was found between Treg levels within the CD4+ population before and 1 yr after Tx (Fig. 2B, P < 0.001).

Figure 2.

(A) Decrease of Tregs in peripheral blood after liver Tx. Shown are the mean percentages of CD4+CD25+CD45RO+CTLA-4+ cells within the CD4+ population of liver transplant recipients (n = 40). At all post-Tx time points Tregs were significantly decreased (*, pre vs. post, P < 0.001). Tregs decreased dramatically in the early months after liver Tx, but showed a relative increase at 1 yr. (B) Pre-Tx Treg levels correlate significantly with levels 1 yr after Tx.

Acute Rejection Is Associated with Reduced Treg Levels

A link between the level of Tregs and the occurrence of a rejection episode in the first 3 months after liver Tx was investigated. Figure 3A shows Treg levels for rejectors and nonrejectors. One year after Tx, a significantly lower percentage of Tregs within the CD4+ T-cell population was observed in patients who experienced an episode of acute rejection than in patients who did not develop allograft rejection (P = 0.005). In Figure 3B, representative dot plots of the flow cytometric analysis for a rejector and a nonrejector are shown. Within the CD4+CD25+ fraction, there was an increased proportion of CD45RO+CTLA-4+ cells in nonrejectors compared to rejectors treated with intravenous methylprednisolone. A link between Tregs levels and acute rejection was confirmed in a univariate analysis (analysis of variance, P = 0.003). After the first year, Treg levels showed a relative increase in the rejection group (P < 0.05), while remaining stable in nonrejectors. These results suggest a transient reduction of Tregs with different kinetics between rejectors and nonrejectors.

Figure 3.

(A) Acute rejection after Tx was associated with reduced Treg fractions. The mean percentages of CD4+CD25+CD45RO+CTLA-4+ cells (± standard error of the mean) are shown. Patients who experienced an episode of (methylprednisolone-treated) acute rejection (n = 10) within 3 months after Tx had significantly lower Tregs 1 yr after Tx compared to nonrejectors (n = 30). No significant differences were seen at other time points. T = months after Tx. (B) Representative dot plots are shown of CD45RO and CTLA-4 expression within the gated population of CD4+CD25+ cells for a rejector and a nonrejector at 1 yr post-Tx. Rejectors had significantly lower CD45RO+CTLA-4+ double-positive CD4+CD25+ cells than nonrejectors. All gates were based on isotype-matched control stainings.

Differential Effects of Immune-Suppressive Regimens on Tregs

To determine whether there was a correlation between immune-suppressive regimens and Tregs in peripheral blood, the relative reduction at 1 yr post-Tx in relation to individual immune suppressants was assessed (Fig. 4). All regimens included calcineurin inhibition, either cyclosporine A or FK506 (tacrolimus). Equal Treg reduction was observed in the cyclosporine or the tacrolimus groups at all time points after Tx. IL-2 receptor blockade, Basiliximab, given as induction immune-suppressive therapy, also had no impact on Treg fractions. Patients receiving steroids (Prednisone) as maintenance therapy showed a trend toward a stronger decrease at 1 yr. Both rejectors and nonrejectors received methylprednisolone (1 × 500 mg) during Tx procedure, which may explain the reduction of Tregs seen in all patients at 3 months (Fig. 3A). Additional treatment with methylprednisolone (3 × 1000 mg) for acute rejection resulted in a significantly higher decrease compared to nonrejectors (Fig. 4).

Figure 4.

Comparison of different immune-suppressive regimens with the percentage decrease of Tregs 1 yr after Tx. All patients were on treatment with a calcineurin inhibitor, either cyclosporin A (CsA, n = 12) or tacrolimus (Tac, n = 28). Steroids (Prednisone), as additional maintenance therapy, were applied to 14 of 40 patients. An IL-2 receptor blocker, Basiliximab, was given to 24 patients. Ten patients with acute rejection were treated with high doses of intravenous methylprednisolone (Solumedrol, 3 × 1,000 mg). Patients treated for acute rejection with methylprednisolone showed a more profound decrease of Tregs (*P = 0.005). Maintenance steroid treatment also showed a stronger decrease, but did not reach statistical significance. No difference was seen between CsA or Tac use, or between patients receiving an IL-2 receptor blocker and patients who did not.

Treg-Mediated Suppressive Activity Is Not Different between Rejectors and Nonrejectors

The suppressive activity of Tregs 1 yr after Tx was determined from rejectors and nonrejectors in a mixed-leukocyte reaction using donor and third-party allogeneic spleen cells. Recipient T cells showed a hyporesponsiveness against donor spleen cells compared to third party, even after depletion of CD4+CD25+ T cells (data not shown). Because of this low antidonor response, further suppression of this response by regulatory CD4+CD25+ T cells could not be determined. Figure 5 shows suppression of the third-party immune response. Suppression of proliferation (Fig. 5A) and cytokine production (Fig. 5B) of effector CD4+CD25 T cells from rejectors, nonrejectors, and healthy controls was comparable on a cell-for-cell basis, indicating that the suppressive activity of Tregs is not affected by continuous immune suppression.

Figure 5.

Comparable suppressor activity by Tregs from rejectors, nonrejectors, and healthy controls. One year after Tx, CD4+CD25+ T–cell-mediated suppression was determined in a mixed-leukocyte reaction for patients who experienced acute rejection (n = 3), patients who did not reject (n = 3), and healthy controls (n = 3). Irradiated third-party spleen cells (5 × 104) were used for stimulation and mixed with responder CD4+CD25 T cells (5 × 104). Subsequently, purified CD4+CD25+ T cells were added in increasing numbers: 5 × 103 (0.1), 1 × 104 (0.2), and 1.5 × 104 (0.3). (A) CD4+CD25+ T cells inhibit the allogeneic T-cell response mediated by CD4+CD25 T cells (P < 0.001). After 5 days of culture, T-cell proliferation was determined. Relative proliferation was calculated with the means of triplicate determinations. In rejectors, nonrejectors, and healthy controls, inhibition of proliferation was similar (P = not significant). (B) Tregs inhibit IFN-γ production by effector CD4+CD25 T cells. Shown are the relative mean IFN-γ concentrations for 3 individuals in each group, as determined by enzyme-linked immunosorbent assay.

DISCUSSION

The role of Tregs in experimental models of transplant tolerance is well established. In the current study we found that after human liver Tx the percentage of Tregs expressing CD4, CD25, CD45RO, and CTLA-4 changes dramatically. Levels of these Tregs in peripheral blood significantly decreased after Tx. In a previous study we found that the total CD4+ fraction within the T-cell population was also reduced in the first year after Tx.15 Taken this finding into consideration, the absolute reduction of Tregs was even more profound than shown in Figure 2A. Significant differences in Treg levels were found between patients who experienced acute rejection versus nonrejectors (Fig. 3A). Accordingly, acute rejection was associated with significantly lower Foxp3 mRNA levels 1 yr after Tx (data not shown). Despite the quantitative changes, the suppressive activity of isolated CD4+CD25+ T cells from liver transplant recipients, both rejectors and nonrejectors, was comparable to that of healthy controls (Fig. 5). On a cell-for-cell basis, the inhibition of T-cell proliferation and IFN-γ production upon allogeneic stimulation was similar for rejectors and nonrejectors; however, the levels of Tregs in blood were 3 times lower in rejectors (Fig. 3A), indicating a higher overall suppression in nonrejectors.

The direct T-cell response to donor antigens after Tx, as determined by a mixed-leukocyte reaction, was low or undetectable in most recipients. There are a number of nonmutually exclusive mechanisms that can be responsible for this state of donor-specific hyporesponsiveness. These mechanisms include T-cell anergy, mixed chimerism, deletion of reactive T cells, or active immune regulation by donor-specific Tregs. The hyporesponsiveness is more frequent in liver Tx compared to heart Tx.16 In our experience, depletion of CD4+CD25+ cells did not overcome the hyporesponsiveness in the mixed-leukocyte reaction (data not shown), suggesting that other CD25 Treg populations17, 18 or other mechanisms than immune regulation might be in play. To be able to determine the suppressive capacity of Tregs after Tx, we used third-party cells as stimulus in the mixed-leukocyte reaction. It is well established that the response to unrelated, third-party allo-antigens is maintained after Tx. This way we were able to demonstrate that the suppressive activity of CD4+CD25+ Tregs was maintained after Tx and comparable between rejectors and nonrejectors (Fig. 5). This, however, does not exclude a possible role for other significant regulatory populations, like CD8+CD28 suppressor T cells, in clinical transplant tolerance.19–21

The quantitative changes seen in Tregs after Tx might be a result of immune-suppressive therapy. A recent study showed reduced levels of CD4+CD25high cells in immune-suppressed liver Tx recipients compared to recipients who were free of immune suppression.22 All patients in our study were treated with calcineurin inhibitors (cyclosporin A or tacrolimus). Calcineurin inhibitors are potent inhibitors of T-cell activation by blocking IL-2 production. As IL-2 is essential for Treg function and survival,23 blocking IL-2 production by calcineurin inhibitors may thereby negatively affect the Treg homeostasis, and therefore hamper tolerance induction.24–29 In addition, a recent study by Baan et al.30 showed reduced induction of Foxp3 in the presence of calcineurin inhibitors (both cyclosporin A and tacrolimus) in the mixed-leukocyte reaction. On the other hand, in our study group treatment with anti-IL-2 receptor antibody (Basiliximab) had no clear effect on Treg levels at 3 months or 1 yr after Tx (Fig. 4).

In the first 3 months after Tx the intensive immune-suppressive treatment directly following Tx may have caused the strong drop in Treg percentages. All patients received methylprednisolone (Solumedrol) and high doses of steroids (Prednisone). It is well known that steroids like methylprednisolone induce apoptosis of lymphocytes. Apoptosis of Tregs could explain the significant reduction of circulating Tregs in the first months after Tx and the sustained reduction in methylprednisolone-treated rejectors. At later time points after Tx (1-4 yr), the percentage of Tregs showed a relative increase in patients who were free of acute rejection, suggesting that the reduction of Tregs is transient. In patients who experienced an episode of rejection, however, Treg levels remained low throughout the first year and only showed a relative increase at 2 yr or later. Additionally, patients receiving steroid maintenance therapy also showed reduced Treg levels after Tx (Fig. 4), though this did not reach statistical significance due to small sample size. Taken together, these findings further support the negative and transient effect of steroid treatment on circulating Tregs. This is consistent with an earlier report that methylprednisolone inhibits spontaneous acceptance of the liver in a rat allo-Tx model, which might be linked to Tregs.31 Moreover, in 2 other models of allograft tolerance methylprednisolone also inhibited acceptance of the transplanted organ.27, 32

Interestingly, patients with end-stage liver disease have an increased fraction of circulating Tregs compared to healthy controls. Recent studies have shown that hepatitis B and C virus infections are associated with an increase of Tregs in peripheral blood.33, 34 Also in our study group, patients with viral hepatitis had 3 times higher Treg levels than controls. Furthermore, we found that high pre-Tx Tregs levels correlated with relatively high Treg levels post-Tx (Fig. 2B). This correlation suggests that high levels of Tregs, linked to the underlying disease, determine the level of Tregs at 1 yr post-Tx. Consistently, patients with viral hepatitis did not develop acute rejection (Table 1), a result that may involve the increased Treg levels seen in these patients. It is well documented that patients with hepatitis B infection have a lower incidence of acute rejection.35 On the contrary, patients with primary biliary cirrhosis or primary sclerosing cholangitis also showed increased Tregs, yet rejection was seen in 4 of 10 patients. This could be a consequence of an impaired Treg function, as seen in some autoimmune diseases.36

Understanding the role of Tregs in the immune response after allo-Tx may contribute to the identification of patients who acquired operational tolerance, and therefore would not need immune-suppressive treatment. Indeed, there is some evidence that pediatric patients who acquired operational tolerance after liver Tx have increased Treg levels compared to patients on immunosuppression.22 Our study demonstrates an association between the homeostasis of Tregs after Tx and acute rejection, illustrating the current dilemma of immune-suppressive treatment—on the one hand preventing rejection, but on the other hand inhibiting Tregs and thereby possibly interfering with the development of transplant tolerance.

Acknowledgements

The authors thank Patrick Boor, Katja Segeren, and Atilla Zahiri for technical support. We thank Betina Hansen for performing statistical analysis and Scot Henry for critically reading the manuscript.

Ancillary