Secreted Wnt Antagonists During Eradication of Cytomegalovirus Infection in Solid Organ Transplant Recipients



We evaluated secreted wingless (Wnt) modulators during cytomegalovirus (CMV) infection in solid organ transplant recipients (SOTr). The major findings were: (i) Plasma levels of Dickkopf-1 (DKK-1) were significantly lower in patients with CMV DNAemia above lower level of quantification at baseline. (ii) Receiver operating characteristic analysis indicated that low DKK-1 and increased secreted frizzled related protein-3 levels were predictors of poor virological outcomes during follow-up. Our findings demonstrate an imbalanced pattern of circulating secreted Wnt modulators in SOTr with poor virological outcomes following treatment for CMV disease, and may suggest a role for dysregulated Wnt signaling on viral pathogenesis during CMV infection.


area under the curve




C-reactive protein




Epstein–Barr virus


human herpes virus


human immunodeficiency virus


lower level of quantification


negative predictive value


peripheral blood mononuclear cells


positive predictive value


receiver operating characteristic


secreted frizzled related protein-3


solid organ transplant recipients




Cytomegalovirus (CMV) infection is associated with substantial morbidity in solid organ transplant recipients (SOTr), despite effective prophylaxis [1]. Antiviral treatment usually suppresses viral replication and cures CMV disease, but systemic viral replication with CMV-related disease may persist in some patients [2].

To establish infection with progressive viral replication, CMV infection activates regulatory pathways which control key cellular and immunological events [3]. Improved understanding of these may help to identify new therapeutic targets and new biomarkers of CMV-related disease. The wingless (Wnt) signaling pathway modulates several processes involved in immune responses following viral and bacterial infections [4]. This pathway is regulated by secreted antagonists or modulators including secreted frizzled related protein (sFRP) and dickkopfs (DKK) that regulate the Wnt signal [5]. Recently, in vitro studies have demonstrated that this pathway may influence CMV replication by enhanced degradation of β-catenin and decreased canonical Wnt signaling [6].

To elucidate the interaction between the Wnt signaling pathway and CMV infection, we evaluated the association between secreted Wnt antagonists (i.e. sFRP-3 and DKK-1), analyzed in stored samples from the VICTOR trial [7], and virological outcomes following treatment for CMV disease in SOTr. The study was focusing on (i) Wnt antagonists levels at baseline in relation to the presence of CMV, and (ii) changes of Wnt antagonists during treatment in relation to viral eradication (early or late).

Materials and Methods

Clinical trial

The present post hoc analysis was performed on available plasma samples obtained at baseline (n = 292) and during follow-up (days 7, 14, 21 and 49 after initiation of antiviral therapy) in the VICTOR trial ( NCT00431353) [7], comparing oral valganciclovir and intravenous ganciclovir for the treatment of CMV disease. Of these patients, 236 were verified to be CMV-DNAemia positive at baseline and 56 were shown to have CMV-DNAemia (Amplicor CMV Monitor® Test; Roche Diagnostics, Indianapolis, IN) below the lower level of quantification (LLoQ; <600 CMV-DNA copies/mL). Enrollment details are detailed elsewhere [7]. CMV disease was defined as the presence of CMV DNAemia in plasma in addition to clinical signs of disease as outlined in detail elsewhere [7]. For viral syndrome at least one of the following additional features was required: body temperature ≥38°C, new or increased significant malaise, leucopenia (white blood cell count <3500/µL), atypical lymphocytosis of ≥5% or thrombocytopenia (platelet count <100 000/µL). Tissue invasive disease was defined based on evidence of localized CMV infection, including biopsy data.

The primary outcome was eradication of CMV plasma DNAemia at day 21 while viral eradication at day 49 was a secondary outcome [7]. The clinical trial was conducted in accordance with the Declaration of Helsinki, good clinical practice guidelines and applicable local regulatory requirements. Informed written consent, covering also poststudy analyses of biological material, was obtained from all patients.

Enzyme immunoassays

Analysis of the predictive value of DKK-1 and sFRP-3 (enzyme immunoassays, intra- and interassay coefficients of variation <10%; R&D Systems, Stillwater, MN) was performed on the pooled data from both treatment groups. No effect of three freeze and thaw cycles was observed on plasma levels of sFRP-3 or DKK-1 (n = 2, thawed three times, sFRP-3 mean CV 9.4%; DKK-1 mean CV 7.7%). Similarly, storing of the samples at bench for 4 h at room temperature had no effects of plasma levels of these antagonists (n = 2, sFRP-3 CV 6.5%; DKK-1 CV 8.9%).


Data were analyzed without replacement of missing data. Demographics and differences in sFRP-3 and DKK-1 levels were compared with the t-test, chi-square and Mann–Whitney procedures depending on distribution. Adjusted p-values were obtained using univariate general linear model. Correlations were tested with the Spearman's nonparametric rho. The area-under-the-concentration-time-curve from D0 to D21, designated area under the curve (AUC; 0–21) throughout the paper, was calculated using the trapezoidal rule using available data. Associations between sFRP-3 and DKK-1 levels at different time points (i.e. baseline, days 7, 14, 21 and AUC 0–21) with outcome (i.e. viral eradication at 21 and 49 days) was assessed by receiver operating characteristic (ROC) analysis, tested for the difference from randomness (AUC = 0.500) under the nonparametric assumption using the DeLong method [8]. The statistical measures: sensitivity, specificity, q-point (most efficient cut-off value, estimated as the average of two consecutively ordered observed test values, that maximizes the sum, specificity + sensitivity), and negative and positive predictive value (NPV and PPV) are included for all ROC calculations. Statistical analyses were performed with SPSS Statistics 17.0 for Windows (IBM Corp., Somers, NY). Parts of the ROC analyses were performed with the module pROC [9] and the module epiR [10] in R [11]. Throughout, we report two-tailed p-values and values <0.05 were considered significant.


Patient characteristics are given in Table 1.

Table 1. Demographic data of patients suspected to have CMV disease—confirmed or not confirmed by a central analysis of CMV DNAemia—and available minimum DKK-1 and sFRP-3 samples
 DNAemia positive (n = 236)DNAemia negative (n = 56)p-Value
  • CMV, cytomegalovirus; DKK-1, Dickkopf-1; EBV, Epstein–Barr virus; HHV, human herpes virus; sFRP-3, secreted frizzled related protein-3; TI, tissue invasion; Tx, transplantation.
  • aChi-square.
  • bt-Test.
  • cMann–Whitney test.
Sex (male), n (%)147 (62.3)35 (62.5)0.977a
Age (years), mean ± SD45.1 ± 13.742.5 ± 11.80.143b
Weight (kg), mean ± SD67.6 ± 15.767.6 ± 13.70.995b
Caucasian; n (%)191 (80.9)35 (62.5)0.001a
Black, n (%)5 (2.1)4 (7.1)
Oriental, n (%)25 (10.6)5 (8.9)
Hispanic, n (%)8 (3.4)9 (16.1)
Other, n (%)7 (3.0)3 (5.4)
Baseline viral load (copies/mL)
Median [25,75 perc.]18 175 [3 560–76 000]N/A
<10 000, n (%)99 (41.9)
10 000–50 000, n (%)63 (26.7)
>50 000, n (%)74 (31.4)
HLA AB-DR mismatch, n (%)152 (94.4)26. (96.3)0.686a
Incomplete information7529
Previous antiviral treatment
Any previous, n (%)112 (47.5)25 (44.6)0.704a
Prophylactic, n (%)92 (39.0)13 (23.2)0.027a
CMV D/R serotypes at time of Tx
D+/R−, n (%)45 (27.3)3 (7.7)0.066a
D+/R+, n (%)91 (55.2)29 (74.4)
D−/R+, n (%)15 (9.1)4 (10.3)
D−/R−, n (%)14 (8.5)3 (7.7)
Incomplete information7117
Time after Tx (days)
Median [25,75 perc.]70 [48–133]164 [93–1186]<0.001c
Transplanted organ
Kidney, n (%)181 (76.7)50 (89.3)0.194a
Liver, n (%)22 (9.3)3 (5.4)
Heart, n (%)18 (7.6)1 (1.8)
Lung, n (%)15 (6.4)2 (3.6)
Clinical diagnosis of CMV disease
Syndrome, n (%)136 (57.6)28 (50.0)0.301a
TI disease, n (%)100 (42.4)28 (50.0) 
HHV, n (%)43 (18.6)N/A
HHV6, n (%)22 (9.5)
HHV7, n (%)22 (9.5)
EBV, n (%)62 (27.1)
Missing information7

Plasma levels of DKK-1 and sFRP-3 in relation to the presence of CMV DNAemia at baseline

Patients with confirmed CMV DNAemia at baseline were characterized by significantly lower plasma DKK-1 compared with patients with CMV DNAemia below LLoQ (Fig. 1A). No difference between these two groups was observed for sFRP-3 at baseline (Fig. 1B). No associations were found with regard to type of organ transplanted, acute rejection (≤2 vs. >2 episodes), tissue invasive disease versus CMV syndrome, magnitude of CMV viremia (<10 000, 10 000–50 000, >50 000), CMV-IgG serological status, HLA, or −AB/DR. Adjusting for sex, age, body weight, ethnicity, type of organ transplanted, time from last transplant, tissue invasive disease versus CMV syndrome and baseline viral load still yielded a significant association between low DKK-1 and the presence of CMV DNAemia (p = 0.034), and no difference for sFRP-3 (p = 0.596).

Figure 1.

Plasma levels of (A) DKK-1 and (B) sFRP-3 in patients with detectable CMV DNAemia (n = 236) compared to patients with CMV DNAemia below LLoQ (n = 56). Data are given as median and 25th and 75th percentile. The p-value derives from the Mann–Whitney U-test. Adjusted p-values are given in the text. CMV, cytomegalovirus; DKK-1, Dickkopf-1; LLoQ, lower level of quantification; sFRP-3, secreted frizzled related protein-3.

To evaluate if DKK-1 and sFRP-3 levels were influenced by other viral infections, we examined the influence of Epstein–Barr virus (EBV), human herpes virus 6 (HHV6) and HHV7 co-infection on plasma levels of these Wnt antagonists in CMV D−/R− transplants (n = 14). DKK-1 and sFRP-3 levels at baseline did not differ to a significant extent between those with (n = 6) and those without (n = 8) concurrent infection by EBV (n = 4), HHV6 (n = 2) or HHV7 (n = 1; p = 0.345 for DKK-1 and p = 0.282 for sFRP-3), although these analyses were hampered by few observations.

While DKK-1, but not sFRP-3, was related to the presence of CMV DNAemia in SOTr, CMV serostatus (CMV seropositive, n = 73 vs. seronegative, n = 22) in healthy controls did not influence plasma levels of DKK-1 (p = 0.98) or sFRP-3 (p = 0.68).

Plasma levels of DKK-1 and sFRP-3 in relation to treatment modalities and inflammatory state

In patients with confirmed CMV DNAemia, there was no difference in sFRP-3 (p = 0.890) or DKK-1 (p = 0.270) levels at baseline (t-test), or during the initial 21 days (respectively, p = 0.734 and 0.303; multivariate p for the time–treatment interaction from repeated-measures general linear model) during the initial 21 days, between patients treated with oral valganciclovir versus intravenous ganciclovir. The description of immunosuppressive treatments and effects on outcome has been reported previously [12].There was no evidence of influence of any individual immunosuppressants (calcineurin inhibitors, mTOR inhibitors, antiproliferatives and anti-CD25 antibodies), or combination of these, on the expression of plasma DKK-1 or sFRP-3 at baseline or during follow-up (AUC 0–21), except use of calcineurin inhibitors was associated with lower levels of DKK-1 at baseline (p = 0.037). However, the association between low levels of DKK-1 at baseline and CMV DNAemia was still significant when adjusted for use of calcineurin inhibitors (p = 0.023; analysis of variance).

When investigating associations with the inflammatory state, as reflected by plasma C-reactive protein (CRP), DKK-1 was negatively correlated with CRP as baseline value (r = −0.15, p = 0.012) and as AUC 0–21 (r = −0.18, p = 0.009). Conversely, sFRP-3 was positively correlated with CRP as baseline value (r = 0.17, p = 0.004) and as AUC 0–21 (r = 0.19, p = 0.004).

Plasma levels of DKK-1 and sFRP-3 in relation to viral outcome at day 21

The association between DKK-1 and sFRP-3 and viral outcome was assessed by ROC analysis at the individual days or as the integrated AUC 0–21. Baseline sFRP-3 was associated with failure to eradicate CMV replication at day 21 (98 [42%] of 236 patients with confirmed baseline CMV DNAemia) (AUC [95%CI] 0.61 [0.53–0.69], p = 0.004) as was AUC 0–21 (AUC [95%CI] 0.59 [0.51–0.67], p = 0.031), but not at other time points (data not shown). No association was observed between eradication at day 21 and DKK-1 at any individual time point or for the time course from days 0 to 14 (data not shown).

Plasma levels of DKK-1 and sFRP-3 in relation to viral outcome at day 49

Figure 2 shows the ROC curve for the AUC for sFRP-3 and DKK-1 from days 0 to 21. For patients who failed to eradicate the virus at day 49 (35 patients, 15%), ROC analysis identified decreased and increased AUC 0–21, for sFRP-3 (Fig. 2A) and DKK-1 (Fig. 2B), respectively, as a predictor of treatment failure at day 49. The most efficient cut-off point is indicated on the ROC curves with resulting sensitivity, specificity, PPV and NPV. The same statistics are given for the individual time points in Figure 2C, demonstrating that failure to suppress CMV replication at day 49 was associated with an increased AUC for sFRP-3 at baseline, days 7, 14, 21 and a decreased AUC for DKK-1 at day 7. Based on these cut-off points, a high sensitivity but low specificity was observed for sFRP-3, and conversely, a high specificity and low sensitivity were observed for DKK-1. This gives a high NPV, but low PPV for both Wnt modulators. When comparing the most representative ROC curves (AUC from days 0 to 21) of DKK-1 and sFRP-3 by treatment arm, the ROC AUCs were not significantly different (data not shown).

Figure 2.

ROC analysis of the Wnt antagonists as predictors of viral eradication at day 49 in the 236 patients with confirmed baseline CMV DNAemia. ROC curve of the area-under-the-concentration-time-curve from baseline to 21 days (AUC 0–21) for (A) sFRP-3 and (B) DKK-1. (C) Forest plot showing the AUC and 95%CI for plasma sFRP-3 and DKK-1 at different time points. All panels give the statistical measures AUC, sensitivity (sens), specificity (spec), positive and negative predictive value (PPV and NPV) with 95% confidence interval. The q-point is indicated with an arrow. 1, under the nonparametric assumption; 2, vs. the null hypothesis: true area = 0.5; 3, most efficient cut-off threshold (q-point). CMV, cytomegalovirus; DKK-1, Dickkopf-1; sFRP-3, secreted frizzled related protein-3; ROC, receiver operating characteristic; Wnt, wingless.


In this study, SOTr patients with positive CMV DNAemia displayed decreased circulating levels of DKK-1. Furthermore, during antiviral therapy, low DKK-1 levels and high sFRP-3 levels were predictors of treatment failure at 49 days. However, further analysis of the diagnostic performance indicated that these Wnt modulators were not viable as biomarkers for clinical use. Nonetheless, our data should encourage further investigation into modulators of Wnt signaling as potential biomarkers in this group of patients as well as further studies on the potential pathogenic implication of Wnt signaling in CMV infection.

Recently, Angelova et al [6] demonstrated in vitro that CMV infection can dysregulate canonical Wnt signaling in fibroblasts and trophoblasts, through effects on endogenous β-catenin. This was associated with decreased transcriptional activity in downstream genes directly regulated by β-catenin, such as DKK-1 [6]. Thus, our in vivo data, showing decreased circulating DKK-1 at baseline in the presence of CMV DNAemia, could reflect a suppressive effect of CMV on canonical Wnt signaling with reduced transcriptional activity on downstream targets such as DKK-1, although no stepwise association was observed with magnitude of viremia. Low DKK-1 was also associated with enhanced systemic inflammation as reflected by CRP both at baseline and during follow-up, and it is possible that CMV infection and enhanced inflammation could contribute to decreased circulating DKK-1 in CMV-infected SOTr patients. Alternatively, low DKK-1 could be a risk factor for the development of CMV infections. However, the descriptive nature of our paper makes it hard to conclude if DKK-1 is a marker or mediator of CMV pathophysiology.

While CMV infection was associated with decreased DKK-1 at baseline, ROC analysis identified increased sFRP-3 at baseline and during time course (i.e. AUC) as a predictor of virological treatment failure at day 21. Attenuated canonical Wnt signaling enhances human immunodeficiency virus (HIV) replication in peripheral blood mononuclear cells (PBMC) [13], and it is possible that sFRP-3 could be involved in viral pathogenesis by antagonizing Wnt ligands and thus impairing Wnt signaling. The association between increased sFRP-3 and viral outcome was stronger at day 49 with significant associations with virological treatment failure at all time points in addition to the integrated time-course data. In addition, DKK-1 at day 7, 14 DKK-1 and for the integrated time course was identified as a predictor of viral outcome. The stronger association between these Wnt modulators and 49 versus 21 day virological treatment failure could indicate that the imbalance between these antagonists is restricted to individuals who respond poorly to anti-CMV therapy, potentially identifying a subgroup of patients who may benefit from therapy targeting Wnt activity. However, based on the calculated best cut-off values (i.e. q points) the data do not support DKK-1 or sFRP-3 as reliable biomarkers since although the probability of virologic failure at day 49 in patients with DKK-1 values above the thresholds, and/or sFRP-3 values below the thresholds—that is, “negative” test results—is low, the probability that a “positive” test value predicts failure is quite low.

Although these data do not support DKK-1 or sFRP-3 as viable biomarkers suitable for clinical practice, they suggest that alterations in the balance of Wnt antagonists are associated with a higher probability of late failures and should encourage further investigation into Wnt signaling and CMV pathophysiology in SOTr. Furthermore, sFRP-3 and DKK-1 levels were unmodified by antiviral therapy and no differences in their levels were observed with regard to common immunosuppressive treatments, although the study was not designed to address these effects. Thus, Wnt signaling may represent a treatment target for individualized patient management in patients who do not respond to conventional CMV therapy. Induction of canonical Wnt signaling with lithium chloride, which could counteract the suppressive effects of sFRP-3 and increase DKK-1 transcription, led to a potent reduction in HIV replication in PBMC in vitro, and in vivo [13]. However, in relation to CMV infection, this hypothesis will have to be further examined in both in vivo and in vitro studies. Also, there are a multitude of other Wnt modulators, both in the DKK and in the sFRP family, as well as others such as sclerostin, R-spondins and Wnt inhibitory factor 1 that merit further investigation.

In conclusion, low baseline circulating DKK-1 levels were observed in a large population of SOTr with CMV infection, and decreased DKK-1 and elevated sFRP-3 levels were predictors of poor viral outcome during long-term follow-up. These data may suggest a role for dysregulated Wnt signaling in the outcomes of CMV infection. However, in vitro studies in relevant cell types as well as in vivo studies in even larger study population and animal models are needed to further clarify any potential role of DKK-1 and sFRP-3 in CMV infection.


The VICTOR study and its primary publications would not have been possible without the contribution of the late Mark Pescovitz, transplant surgeon and polymath, whose insight and expertise we sadly miss.


The authors of this manuscript have no conflicts of interest to disclose as described by the American Journal of Transplantation.