Monitoring of alphatorquevirus DNA levels for the prediction of immunosuppression-related complications after kidney transplantation

2.1 Study population and setting

The present prospective, observational, cohort study was performed between November 2014 and December 2016 at the Hospital Universitario “12 de Octubre.” All adult patients with ESRD undergoing KT during this period and providing informed consent were deemed eligible for inclusion. Double organ recipients were excluded. This study was performed in accordance with the ethical standards laid down in the Declarations of Helsinki and Istanbul. The local Clinical Research Ethics Committee approved the study protocol.

2.2 Study design

Participants were enrolled at the time of KT and followed up for at least 12 months, unless graft loss (retransplantation or permanent return to dialysis) or death occurred earlier. Scheduled follow-up visits were carried out at baseline, every 2 weeks during the first 3 months, and monthly thereafter, as well as whenever clinically indicated. A number of pretransplant, perioperative, and posttransplant variables were recorded by using a standardized case report form. Plasma alphatorquevirus DNA load was quantified at baseline (ie, within 6 hours prior to the transplant procedure), day 7, and months 1, 3, 6, and 12 by a polymerase chain reaction (PCR)-based quantitative nucleic acid amplification test. Descriptions of immunosuppression and prophylaxis regimens are detailed in Supplementary Methods.

The study outcome was the occurrence of overall posttransplant serious infection (defined by the requirement of hospitalization and intravenous antimicrobial therapy) and iRAE, with the latter encompassing the occurrence of opportunistic infection (as defined below) and/or posttransplant de novo malignancy.26

2.3 Plasma alphatorquevirus DNA load quantification

Blood samples were immediately centrifuged and plasma specimens were preserved at −80°C, with aliquots thawed for the first time for the present analyses. DNA was extracted from 200 μL of plasma with the NucliSENS^® easyMAG^® automated system (bioMérieux, Marcy-l’Étoile, France), following the manufacturer's instructions. DNA loads were quantified by means of the TTV R-gene^® kit (ARGENE^® range, bioMérieux), a real-time PCR assay targeting a highly conserved segment of the 5′ untranslated region of the viral genome with >90% identity across isolates.6 Analytical performances of this assay in terms of precision (reproducibility and repeatability), correlation with the in-house PCR assay developed by Maggi et al,27 inclusivity and exclusivity, and linearity have been recently reported 28 and are detailed in Supplementary Methods. PCR amplification and amplicon detection was performed on an ABI Prism 7500 system (PE Biosystems, Foster City, CA). The viral load (in copy numbers per mL) was determined using a standard curve with known copy numbers and log₁₀-transformed for statistical analyses. The lower limit of detection (LLoD) was 167 copies/mL (95% confidence interval [CI]: 92-581) or 2.2 log₁₀ copies/mL (95% CI: 2.0-2.8), with DNA quantitation in the linear range from 2.1 × 10² to 2.1 × 10⁷ copies/mL. Specimens with undetectable DNA loads were assigned a value of 0.01 (−2.0 log₁₀) copies/mL for analysis purposes. All samples from each patient were simultaneously assayed in singlets. Although the taxonomic classification and nomenclature of the Anelloviridae family are still discussed, there is general consensus to recommend the term “alphatorquevirus” instead of “TTV” in order to encompass the different genospecies within this genus. Therefore, results of the PCR assay were referred to as plasma alphatorquevirus DNA load.

2.4 Study definitions

We defined opportunistic infection as that due to intracellular bacteria (mycobacteria, Listeria monocytogenes, and Nocardia spp.), herpesviruses (cytomegalovirus [CMV], herpes simplex virus, and varicella-zoster virus), polyomaviruses (biopsy-proven BK polyomavirus-associated nephropathy [BKVAN]), yeasts (Candida spp. and Cryptococcus spp.), molds (invasive aspergillosis and mucormycosis), and parasites (Toxoplasma gondii, Pneumocystis jirovecii, and Leishmania spp.).29 We also included the diagnosis of presumptive BKVAN (high-level polyomavirus BK replication [plasma viral load >4 log₁₀ copies/mL] at 2 time points 3 or more weeks apart) even in the absence of biopsy-proven nephropathy.30 Surgical site and urinary tract infections due to Candida spp. were excluded since such complications are usually related to previous invasive procedures, mucosal breakdown, and indwelling urinary catheters rather than to the patient's immune status. The diagnosis of CMV disease required the demonstration of CMV replication and the presence of attributable symptoms, and was categorized as viral syndrome or end-organ disease according to consensus definitions.31 Proven or probable invasive fungal infection (IFI) was defined by the European Organization on Research and Treatment in Cancer and the Mycoses Study Group criteria.32 Further study definitions are detailed in Supplementary Methods.

2.5 Statistical analysis

Quantitative data were shown as the mean ± standard deviation (SD) or the median with interquartile ranges (IQR). Qualitative variables were expressed as absolute and relative frequencies. Categorical variables were compared using the χ² test. Student t test or Mann-Whitney U test were applied for continuous variables. Repeated measures were compared with the Student t test for paired samples or the Wilcoxon test. Pearson's correlation coefficient or Spearman's rank correlation coefficient were used to investigate the correlation between continuous variables.

The optimal cutoff values (ie, that with the highest value for the combined sensitivity and specificity) of plasma alphatorquevirus DNA loads to predict study outcomes were identified by the Youden's index 33 in the area under receiver operating characteristics (auROC) curve. In the absence of an external validation cohort, the selected cutoff values were evaluated by bootstrap simulation, which estimated how good the predictive performance of the test (ie, having a plasma DNA load below or above the threshold) would be on a hypothetical set of new patients. To this aim, 1000 bootstrap samples of equal size were generated from the study population by sampling with replacement. Time-to-first-event curves were plotted by the Kaplan-Meier method. Intergroup differences in cumulative incidence curves were compared with the log-rank test, and multivariate Cox regression models were used to evaluate the association between plasma DNA loads and study outcomes. Multicollinearity was assessed using variance inflation factors. Associations were given as hazard ratios (HRs) and 95% CIs.

Plasma alphatorquevirus DNA load area under curve (AUC) was calculated by means of the trapezoid rule.34 As previously described,15, 35 we estimated the DNA load doubling time according to the formula: dt = (t₂-t₁) × [Ln₂/Ln(q₂/q₁)], where q₁ and q₂ represent the plasma loads (in copies/mL) at the time of the first and second positive PCR result, respectively, and (t₂-t₁) the interval (in days) between both dates. Only episodes in which there was an increase between the first and second alphatorquevirus DNA load values of >3-fold were considered for analysis.

All the significance tests were 2-tailed. The threshold for significance was set at a P  < .05. Statistical analysis was performed using SPSS version 20.0 (IBM Corp., Armonk, NY) and Prism version 6.0 (GraphPad Software Inc., La Jolla, CA).

3.1 Characteristics of the study population

We included 221 KT recipients, whose clinical characteristics are shown in Table 1. The median follow-up was 494 days (IQR: 434-542). Five patients (2.3%) experienced graft loss after a median posttransplant interval of 41 days (IQR: 18-260.5), whereas 2 patients (0.9%) died (1-year survival rate: 98.0%). These patients were censored for follow-up at these points.

Regarding the occurrence of study outcomes, 128 patients (57.9%) developed a total of 287 episodes of posttransplant infection (incidence rate: 2.67 episodes per 1000 transplant-days). The clinical and microbiological characteristics are detailed in Supplementary Results (Table S1). The median interval from transplantation to the first episode was 37.5 days (IQR: 14-99.3). Fifty-one patients (23.1%) had 65 episodes of iRAE (incidence rate: 0.61 episodes per 1000 transplant-days) (Table 2). The median interval to the first episode was 78 days (IQR: 39-235).

3.2 Kinetics of plasma alphatorquevirus DNA load

The total number of monitoring points for plasma alphatorquevirus DNA was 997 (median of 5 points per patient [IQR: 4-5]). Only 2.4% (24/997) samples were below the LLoD (ie, undetectable load), with rates ranging from 3.7% (7/187) at baseline to 0.6% (1/177) at month 6. There was a progressive increase in viral load from baseline (2.9 ± 1.6 log₁₀ copies/mL), to peak at month 6 (5.7 ± 1.9 log₁₀ copies/mL) and slightly decreased at month 12 (5.0 ± 2.1 log₁₀ copies/mL) (P  < .05 for all paired comparisons) (Figure 1). Accordingly, viral loads peaked in most patients at month 6 (51.6% [114/221]).

We found a significant although weak direct correlation between recipient age and plasma alphatorquevirus DNA loads at baseline (r = 0.263; P  = .0002) and day 7 (r = 0.151; P  = .029), but not thereafter. CMV-seronegative recipients had lower alphatorquevirus loads as compared to those who were seropositive (2.3 ± 1.6 vs 2.9 ± 1.6 log₁₀ copies/mL; P  = .028). Regarding posttransplant factors, induction therapy with anti-thymocyte globulin (ATG) was associated with higher plasma alphatorquevirus DNA loads through the first 6 posttransplant months, with significant differences at month 3 (5.5 ± 1.8 vs 4.8 ± 1.9 log₁₀ copies/mL; P  = .018) (Figure S1). We found no impact of the administration of anti-CMV prophylaxis or the primary immunosuppression regimen. Twenty-two patients (10.0%) were converted to an mTOR inhibitor after a median interval of 217 days (IQR: 117.5-306.8). Plasma alphatorquevirus DNA loads did not differ between this subgroup of patients and the rest of the cohort either (data not shown).

3.3 Correlation between alphatorquevirus DNA loads and immune parameters

In order to better characterize the potential of alphatorquevirus DNA kinetics as a surrogate marker for immunosuppression, we analyzed the relationship between this parameter and absolute counts of peripheral blood lymphocyte populations. At posttransplant month 1, there were significant inverse correlations between the plasma alphatorquevirus DNA load and concurrently measured CD3⁺ (r = −0.238; P  = .017), CD4⁺ (r = −0.241; P  = .015), and CD8⁺ T cell counts (r = −0.240; P  = .016). Such correlations were even more evident by month 3, particularly for CD3⁺ (r = −0.347; P  < .0001) and CD4⁺ T cell counts (r = −0.330; P  < .001) (Figure S2).

3.4 Alphatorquevirus DNA loads at single points and outcomes

First, we explored the association between plasma alphatorquevirus DNA loads at discrete time points and study outcomes. We found no significant differences in baseline loads between patients with or without posttransplant infection (2.9 ± 1.6 vs 2.8 ± 1.6 log₁₀ copies/mL; P  = .854) or iRAE (3.1 ± 1.5 vs 2.8 ± 1.6 log₁₀ copies/mL; P  = .369). Such lack of significance persisted when plasma loads were measured at day 7, either for infection (3.2 ± 1.7 vs 3.1 ± 1.7 log₁₀ copies/mL; P  = .697) or iRAE (3.5 ± 1.6 vs 3.0 ± 1.7 log₁₀ copies/mL; P  = .148).

Nevertheless, differences emerged thereafter, when most recipients have reached a stable level of immunosuppression. Plasma alphatorquevirus DNA loads at month 1 were higher among patients who subsequently developed posttransplant infection as compared to those remaining free from this complication (4.6 ± 1.3 vs 3.8 ± 1.9 log₁₀ copies/mL; P  = .023). Such a difference was also observed between patients developing or not developing iRAE (4.9 ± 1.2 vs 3.9 ± 1.8 log₁₀ copies/mL; P  = .009) (Figure 2A). Likewise, peak DNA loads during the preceding periods were significantly higher in patients who developed iRAE beyond month 3 (5.6 ± 1.3 vs 4.7 ± 1.8 log₁₀ copies/mL; P  = .007) or month 6 (6.8 ± 2.0 vs 5.8 ± 1.7 log₁₀ copies/mL; P  = .012) (Figure 2B), with a nonsignificant trend at the latter point for posttransplant infection (6.3 ± 1.9 vs 5.8 ± 1.7 log₁₀ copies/mL; P  = .097).

In view of its potential utility for guiding clinical decisions early after transplantation, we further analyzed the discriminative value of alphatorquevirus DNA loads at month 1. The auROCs for predicting infection and iRAE were 0.624 (95% CI: 0.517-0.732; P  = .029) and 0.704 (95% CI: 0.588-0.820; P  = .002), with optimal cutoff values set at 3.15 and 4.56 log₁₀ copies/mL, respectively. The presence of plasma alphatorquevirus DNA loads above these thresholds was associated with higher cumulative incidences of infection (log-rank P  = .009) and iRAE (log-rank P  = .0006) (Figure 3). The predictive performance of both cutoff values estimated through 1000 bootstrap samples is detailed in Table 3. Such associations remained significant after multivariate adjustment, both for posttransplant infection (adjusted HR: 2.88; 95% CI: 1.13-7.36; P  = .027) (Table S2) and iRAE (adjusted HR: 5.17; 95% CI: 2.01-13.33; P  = .001) (Table S3).

3.5 Areas under curve for plasma alphatorquevirus DNA and outcomes

We explored the correlation between the cumulative magnitude of alphatorquevirus DNAemia, estimated through the AUC for log₁₀ plasma DNA load, and study outcomes. The AUCs between baseline and month 1 (AUC_0-30) were significantly higher among patients with posttransplant infection (5.1 ± 1.7 vs 4.6 ± 1.7 log₁₀ copies/mL; P  = .046) or iRAE (5.4 ± 1.4 vs 4.7 ± 1.7 log₁₀ copies/mL; P  = .015) beyond that point. Likewise, the AUCs to month 6 (AUC_0-180) were also higher among patients subsequently developing posttransplant infection (8.8 ± 1.3 vs 7.9 ± 1.6 log₁₀ copies/mL; P  = .032) or iRAE (9.1 ± 1.2 vs 7.9 ± 1.5 log₁₀ copies/mL; P  = .023) (Figure 4).

3.6 Kinetics of alphatorquevirus DNA loads and outcomes

Previous studies have suggested that TTV replication kinetics mirrors more accurately the state of immunosuppression than the viral load at a given point.15, 36 Thus, we investigated whether dynamic changes in alphatorquevirus loads correlates with posttransplant outcomes by separately analyzing the trajectory (ascending or nonascending [ie, stable or decreasing] slope) and magnitude (viral doubling time) of change in plasma alphatorquevirus DNA loads between 2 consecutive monitoring points.

Patients showing an increasing slope of change in alphatorquevirus DNA loads between day 7 and month 1 were more likely to subsequently develop posttransplant infection compared to those with nonascending kinetics (57.3% [47/82] vs 18.8% [3/16]; P  = .005). A similar nonsignificant trend was also observed for iRAE (26.8% [22/82] vs 6.2% [1/16]; P  = .108). Increasing kinetics of alphatorquevirus DNA load between both points acted as an independent predictor for posttransplant infection (adjusted HR: 4.29; 95% CI: 1.32-14.04; P  = .016) (Table S4), with significant differences in terms of cumulative incidence (log-rank P  = .013) (Figure 5). No comparable associations were observed for any of the remaining time intervals, including that immediately after transplantation (ie, from baseline to day 7). This finding is concordant with the sigmoidal-shaped model proposed for TTV DNA kinetics in lung transplant (LT) recipients, in which the increase in viral load exhibits a delay of ≈15 days after the initiation of immunosuppression, followed by an almost linear increase between days 15 and 45 and a progressive stabilization thereafter.15 Figure S3 depicts illustrative examples of increasing dynamics of alphatorquevirus DNA loads and relevant posttransplant events.

The lowest doubling time for alphatorquevirus DNA load across different time intervals was observed between day 7 and month 1 (median: 4.9 days [IQR: 3.3-7.6]) (Table S5), in accordance with the aforementioned sigmoidal-shaped course. Doubling times through the first month were lower among patients who received ATG induction, either between baseline and day 7 (4.0 [IQR: 2.1-6.5] vs 7.1 [IQR: 4.3-17.1] days; P < .0001) or between day 7 and month 1 (4.0 [IQR: 2.8-6.1] vs 6.3 [IQR: 3.6-9.1] days; P  = .020) (Figure S4). In view of this significant interaction, we separately analyzed alphatorquevirus doubling times according to the type of induction therapy. There were no differences among ATG-treated patients who did or did not develop posttransplant infection or iRAE. However, doubling times between day 7 and month 1 were lower for patients who did not receive ATG and developed posttransplant infection as compared to those remaining free from this complication (5.5 [IQR: 3.5-8.4] vs 7.3 [IQR: 5.3-22.4] days; P  = .070) (Figure S5).

3.7 Alphatorquevirus DNA loads and graft rejection

Finally, we analyzed the correlation between plasma alphatorquevirus DNA loads and graft rejection. In concordance with the presumed nature of this variable as a marker of immunosuppression, baseline loads were lower (suggesting a higher level of immunocompetence) among patients who developed acute rejection during the first 90 posttransplant days (1.7 ± 2.3 vs 2.9 ± 1.6 log₁₀ copies/mL; P  = .035). In addition, the cumulative incidence of rejection was significantly higher among patients with undetectable DNA at baseline (28.6% [2/7] vs 3.3% [6/180]; P  = .030). After multivariate adjustment, higher plasma alphatorquevirus DNA loads at baseline remained as a protective factor for the development of acute graft rejection (adjusted HR [per 1-log₁₀ copies/mL increase]: 0.69; 95% CI: 0.49 - 0.97; P  = .034) (Table S6).