CMV Infection in the Donor and Increased Kidney Graft Loss: Impact of Full HLA-I Mismatch and Posttransplantation CD8+ Cell Reduction



Despite a large body of literature, the impact of chronic cytomegalovirus (CMV) infection in donor on long-term graft survival remains unclear, and factors modulating the effect of CMV infection on graft survival are presently unknown. In this retrospective study of 1279 kidney transplant patients, we analyzed long-term graft survival and evolution of CD8+ cell population in donors and recipients by CMV serology and antigenemia status. A positive CMV serology in the donor was an independent risk factor for graft loss, especially among CMV-positive recipients (R+). Antigenemia was not a risk factor for graft loss and kidneys from CMV-positive donors remained associated with poor graft survival among antigenemia-free recipients. Detrimental impact of donor's CMV seropositivity on graft survival was restricted to patients with full HLA-I mismatch, suggesting a role of CD8+ cells. In R+ patients with positive CMV antigenemia during the first year, CD8+ cell count did not increase at 2 years posttransplantation, in contrast to R recipients. In addition, marked CD8+-cell decrease was a risk factor of graft failure in these patients. This study identifies HLA-I full mismatch and a decrease of CD8+ cell count at 2 years as important determinants of CMV-associated graft loss.


anti-lymphocyte globulins


CMV antigenemia




CMV seropositive donor


CMV seronegative donor


CMV seropositive recipient


CMV seronegative recipient


Cytomegalovirus (CMV) is the leading cause of viral infection after solid organ transplantation. This double-stranded DNA virus establishes a lifelong infection characterized by latency alternating with periodic reactivations for spreading infectious progeny in 60–100% of the adult population worldwide [1, 2]. Many studies have shown that local inflammation can induce the virus to exit of latency from dendritics cell and macrophages [3, 4], and in turn CMV reactivation promotes local inflammation [5]. The quality of the immune response to CMV has a crucial role in controlling this cycle [6, 7]. CMV infection stimulates a tremendous number of CMV-specific T cells [8]. Over time, chronic persistence of CMV causes the accumulation of an increasing number of dysfunctional CMV-specific T cells that progressively lose several of their effectors functions [9-16].

In transplant patients, drug-induced chronic immunosuppression may create a microenvironment favorable to a new equilibrium between local periodic subclinical reactivations and the host's immune response, which can be deleterious to the transplanted organ [17-19]. CMV has been implicated in chronic cardiac allograft vasculopathy [20] and bronchiolitis obliterans in lung transplant patients [21, 22]. Despite a large body of literature on CMV, the impact of transplanting kidneys from CMV-positive donor on long-term graft survival remains highly controversial [23-26]. Several studies reported that CMV was a risk factor for interstitial fibrosis/tubular atrophy [27, 28] and chronic allograft dysfunction [29, 30], while others studies did not find such effects [31, 32].

The impact of CMV on long-term kidney function may be obscured by many factors. First, because of the increased incidence of latent CMV infection with age, differentiating age- from CMV-related events is difficult, especially in evaluating kidney graft survival. Second, many studies in transplant patients relied on blood samples to survey CMV reactivation although it is now recognized that interactions between host tissues and chronic viruses are organ-specific. This is well illustrated during infection by murine γ-herpes virus 68, which is removed from the arteries by the immune system whereas other tissues are not efficiently cleared [33]. Third, reactivation of CMV infection is known to be transient and patchy, while obtaining sequential, repeated, large-sized biopsies of kidneys is not possible for ethical reasons. Finally, host factors that may modulate the long-term effects of CMV infection in transplant patients have never been documented. Therefore, the impact of transplanting kidneys from CMV-infected donors (i.e. CMV-seropositive donors) on long-term kidney graft survival may have been misjudged and needs further analysis. The aim of the present study was to determine the impact of CMV infection in the donor on graft survival and to identify host factors, which modulate this risk.

Patients and Methods


In all, 1281 consecutive renal transplantations were performed between October 1985 and December 2008 in our center (end of follow-up, December 31, 2010). Donor CMV serology was unknown for two patients, so we investigated 1279 renal transplantations. The median follow-up was 6.6 years (range 0 days to 24.6 years).

Initial immunosuppression involved methylprednisolone, polyclonal anti-lymphocyte globulins (Thymoglobuline® or lymphoglobuline®, Imtix-Sangsat, Lyon, France) or basiliximab (Simulect®, Novartis, Rueil-Malmaison, France). The maintenance immunosuppressive regimen was azathioprine or mycophenolate mofetil, cyclosporine, tacrolimus or sirolimus and prednisone as described [34].

Renal biopsy was performed when serum creatinine raised by 20% or more over the baseline, without any other obvious cause. Date of biopsy-proven acute rejection (BPAR) was defined by the date of the biopsy.

The patients with the highest risk of CMV disease (CMV-positive donors into CMV-negative recipients [D+/R] at the time of the transplantation) received prophylaxis with a weekly infusion of immunoglobulin therapy for one month (from 1985 to 1994), aciclovir or valaciclovir (1995–2003) or valganciclovir (2004–2008) for 3 months in absence of contraindication and severe adverse events. Patients with a lower risk of CMV disease (pretransplantation CMV-positive recipients [R+]) received preemptive therapy. Since 1995, antiviral drugs were used for asymptomatic patients with CMV-Ag above 50 stained neutrophils per milliliter and those with CMV syndrome or disease, whatever the CMV-Ag values.

CMV infection diagnosis

From 1984 to 1994 diagnosis of systemic CMV infection relied on optimized CMV culture on MRC5 human cells with indirect immunofluorescence plaque staining using E13 mouse monoclonal antibody directed to CMV protein IE1 72. Stained nuclei were observed after 24 and 48 h incubation. CMV antigenemia (CMV-Ag) was added in 1995 as a more rapid and direct evaluation of CMV viremia. Briefly, after lysis of red cells, leucocytes of the patient were collected, standardized to 106 cells/mL and cytocentrifuged on a spot slide. Then, indirect immunofluorescence assay was performed with mouse monoclonal antibodies C10 and C11 directed to the early CMV protein pp65. The number of stained-polynuclear cells per 2 × 105 leucocytes was recorded. Weekly CMV-antigenemia monitoring was performed during 4 and 6 months following transplantation in R+ and D+R, respectively.

Variables studied

At the time of transplantation, the following variables were recorded: type of donor (living or deceased), CMV serology, age and gender of the recipient and donor, diabetes and weight of recipient, number of class I (-A, -B) and class II (-DR) HLA mismatches, immunosuppressive treatments, cold ischemia time and delayed graft function. The CMV serology was usually not performed after transplantation.

At the 3-month visit, the following variables were measured: systolic and diastolic arterial pressure, serum creatinine level, and proteinuria, estimated glomerular filtration rate (eGFR as measured by the Modification of Diet in Renal Disease formula in adults [35] and the Schwartz formula in children [36]). The occurrence and date of BPAR episodes were recorded.

Phenotypic analyses of lymphocytes were performed according to standard whole blood procedure using a FACStar plus (Becton Dickinson) or an EPICS-XL-MCL (Beckman Coulter) flow cytometer. CD4+, CD8+ and natural-killer (NK) cell counts were recorded at day 0 (D0) and year 2. CMV-Ag values were available from 1995 in 898 patients.

Statistical analyses

Results are expressed as percentage or mean ± SD for normally distributed variables and median and range for non-normally distributed variables. Qualitative data were compared using chi-square. Quantitative data were compared using Mann–Whitney when two groups were compared, and using ANOVA (followed by pair-wise comparisons using PSLD Fisher test) when more than two groups were compared.

Kaplan–Meier curves were used to estimate the predictive value of donor CMV serostatus (and other relevant parameters) on uncensored graft survival (death or end-stage renal disease [ESRD]), patient survival (ESRD was censored) and death-censored graft survival (equivalent to graft survival). Log-rank test was used to compare survival curves. End of follow-up was defined as death (with a functioning graft) or graft loss (i.e. dialysis or re-transplantation). Of note, only 6/1279 (0.5%) patients were lost during follow-up; in these six patients, end of follow-up was defined as the date of last visit. Parameters used as covariates were those associated (p value < 0.10) with graft loss in univariate analyses using Cox regression: we first used usual clinical and biochemical covariates (Model 1), and then added immunosuppressive medications (Model 2); we finally added donor age (Model 3) because donor age was associated with graft loss after adjustment on graft year. Results are expressed with hazard ratios (HR), 95% confidence intervals (95% CI) and p values. Statistical analyses were performed using SAS v9.1 (SAS Inst., Cary, NC). A p < 0.05 was considered statistically significant.


Baseline characteristics

Donor and recipient characteristics at the time of transplantation and at the 3-month visit are shown in Table 1 (n = 1279). In the study period, 1219 adults and 62 recipients under 18 years were transplanted. There was 1118 (86%) first transplantation. Sixty-two percent of patients were male. Recipient and donor age were respectively 45.6 ± 15.3 and 43.2 ± 15.3 years. At 3 months, mean eGFR was 54.3 ± 19 ml/min/1.73 m2. BPAR occurred in 28.3% during the first year. Patients were then analyzed separately according to donor and recipient CMV serology (Table 1). DR patients were younger than D+R+ and DR+ patients and received a kidney from younger donor than all other groups. Of note, cyclosporine was more often used in DR patients that in other patients.

Table 1. Baseline characteristics of patients with transplantation between 1985 and 2008
 All (n = 1279)D+ R (n = 277)D+ R+ (n = 284)D R+ (n = 360)D R (n = 358)
  • D+, CMV-positive donor; D, CMV-negative donor; CMV, cytomegalovirus; HLA, human leukocyte antigen; DGF, Delayed graft dysfunction; BPAR, Biopsy-proven acute rejection; eGFR, estimated creatinine clearance.
  • *p < 0.05 for comparison with D R+.
  • p < 0.05 for comparison with D+ R+.
  • §p < 0.05 for comparison with all other groups.
  • p < 0.05 for comparison with D+ R+ and D R+.
Donor characteristics
Age (years)43.2 ± 15.343.6 ± 17.646.4 ± 16.843.9 ± 15.539.7 ± 17.5§
Male sex (%)63.561.461.761.967.3
Living donor (%)
Recipient characteristics
Age (years)45.6 ± 15.344.5 ± 15.9*48.7 ± 13.747.6 ± 13.842.1 ± 16.4
Male sex (%)61.564.355.357.568.4
Diabetes mellitus (%)
Body mass index (kg/m2)24.1 ± 4.323.8 ± 4.123.6 ± 4.224. ± 4.424.2 ± 4.4
Graft characteristics
First graft (%)86.188.485.684.484.9
HLA class I AB mismatch2.7 ± 1.02.6 ± 1.12.8 ± 1.02.8 ± 1.02.7 ± 0.9
HLA class II DR mismatch1.3 ± 0.71.3 ± 0.61.3 ± 0.71.2 ± 0.71.2 ± 0.7
DGF (%)20.719.521.021.020.0
BPAR during the first year (%)28.328.225.429.230.2
Cold ischemia (hours)20.8 ± 8.321.5 ± 9.221.0 ± 8.220.6 ± 7.720.4 ± 8.1
Initial immunosuppressive regimen
Antilymphocyte globulins (%)
Interleukin 2 receptor antagonists (%)
Cyclosporine (%)§
Tacrolimus (%)24.225.627.126.418.4§
Mycophenolate mofetyl (%)65.965.069.769.260.3
Azathioprine (%)33.433.229.230.839.4
Sirolimus (%)
Clinical and biochemical variables at 3 months
Systolic arterial pressure (mmHg)138 ± 17137 ± 17139 ± 17139 ± 17137 ± 17
Diastolic arterial pressure (mmHg)80 ± 1180 ± 1180 ± 1081 ± 1080 ± 11
Serum creatinine level (µmol/L)132 ± 48134 ± 51137 ± 55130 ± 40129 ± 47
eGFR (ml/min/1.73 m2)54.3 ± 19.054.7 ± 20.951.0 ± 17.653.0 ± 15.957.7 ± 20.6
Proteinuria (g/day)0.32 ± 0.830.33 ± 0.900.33 ± 0.920.31 ± 0.560.33 ± 0.92

CMV Infection in the donor is an independent risk factor of graft loss, especially in R+ patients

Among 1279 patients, death, ESRD occurred in 150 (11.7%), 284 (22.2%), respectively, during follow-up. On univariate analysis, CMV-positive serology in the donor was associated with increased risk of uncensored graft loss (p = 0.053), with an estimated half-life of 14.5 and 17.0 years in the D+ and D groups, respectively (Figure 1A). Patient survival did not differ between the two groups (p = 0.558). However, death-censored graft survival was lower with kidneys from CMV-positive than -negative donors (Figure 1B, p = 0.005). This result remained when only CMV-exposed were analyzed (i.e. DR were excluded), as shown in Figure 1C (p = 0.023). Interestingly, by year 1 and 2, D+ patients had a lower GFR as compared to D recipients (57.7 ± 19.6 vs. 62.5 ± 19.2 ml/mn and 55.8 ± 19.9 vs. 60.8 ± 19.1 ml/mn, p < 0.0001, respectively).

Figure 1.

Kidneys from CMV-infected donors show lower survival, especially in R+ recipients. Estimated uncensored graft survival by donor CMV serostatus in 1279 consecutive patients with transplantation between 1985 and 2008 are represented (A) (p = 0.053). Death-censored graft survival in the same 1279 patients and in 921 CMV-exposed recipients (i.e. DR excluded) are presented in B (p = 0.005) and (C) (p = 0.023), respectively. Lower graphs represent estimated death-censored graph survival in 644 CMV-seropositive recipients (D, p = 0.008) and 635 CMV-seronegative recipients (E, p = 0.210).

We identified variables associated with risk of graft loss on univariate analysis: use of antilymphocyte globulins (1.536 [1.129–2.092], p = 0.006), mycophenolate mofetil (0.737 [0.566–0.957], p = 0.022) and cyclosporine (0.530 [0.404–0.694], p < 0.001), serum creatinine level (HR 1.010 [95% CI 1.008–1.012], p < 0.0001), proteinuria (HR per g/day 1.267 [1.184–1.357], p < 0.0001) and systolic arterial pressure (HR per 10 mmHg 1.011 [1.004–1.019], p = 0.004) at 3 months, BPAR (HR 2.660 [2.062–3.437], p < 0.0001) and date of transplantation (HR per year 0.964 [0.944–0.984], p = 0.0005). The increased risk of graft loss for the D+ group remained significant after adjustment for donor age and these variables (Table 2). In addition, acute rejection (HR 2.216 [1.670–2.941], p < 0.0001), year of transplantation (HR per year 0.934 [0.890–0.980], p = 0.0006), proteinuria (HR per g/day 1.262 [1.169–1.363], p < 0.0001) and serum creatinine (HR per µmol/L 1.007 [1.005–1.010], p < 0.0001) remained independently and significantly associated with graft survival (Supplementary Table S1). In 921 patients exposed to CMV (i.e. after exclusion of the DR transplant group), receiving a kidney from a CMV-seropositive donor remained an independent risk of graft loss (Table 2).

Table 2. CMV-positive donor is an independent risk factor of death-censored graft loss
 Univariate analysis Multivariate analysis
 HR95% CIp HR95% CIp
  1. HR, hazard ratio; 95% CI, 95% confidence interval; D+, CMV-positive donor, R+, CMV-positive recipient.
  2. Model 1: adjustment on creatinine, proteinuria, systolic blood pressure at 3 months, biopsy-proven acute rejection occurrence during the first year and date of transplantation.
  3. Model 2: model 1 + adjustment on mycophenolate mofetil, cyclosporine, anti-lymphocyte globulins use.
  4. Model 3: model 2 + adjustment on donor age.
All patients (n = 1279)1.3931.104–1.7590.005Model 11.3661.039–1.7960.026
    Model 21.3261.007–1.7440.044
    Model 31.3301.010–1.7510.042
D R excluded (n = 921)1.3911.044–1.8550.024Model 11.4421.025–2.0290.036
    Model 21.4231.011–2.0040.043
    Model 31.4171.007–1.9950.046
R+ patients (n = 644)1.5501.121–2.1440.008Model 11.5121.021–2.2390.039
    Model 21.4710.993–2.1790.054
    Model 31.4580.984–2.1610.060

Since the capacity to control CMV replication may differ according to CMV serostatus of the recipient, we estimated the risk of graft loss associated with donor CMV in relation to recipient CMV serostatus. The increased risk of death-censored graft loss observed in patients who received a graft from a CMV-positive donor was highly significant in R+ recipients (Figure 1D, p = 0.008) but not in R recipients (Figure 1E, p = 0.21). Moreover in R+ recipients, the risk was not really modified after adjustment for variables (Table 2). Therefore, CMV infection in the donor was an independent risk factor of graft loss, especially in R+ recipients before engraftment (R+).

CMV-positive donors are associated with high risk of CMV-antigenemia in CMV-positive recipients but the increased risk of graft lost is independent from CMV viremia

We wondered whether CMV infection in donors conferred an increased risk of viral replication in R+ recipients without any viral prophylaxis (n = 429). Receiving a graft from a CMV-positive donor significantly increased the risk of CMV-Ag in R+ recipients (D+/R+: 68.5% vs. D/R+: 49.6%, p < 0.0001). However, multivariate analysis revealed positive CMV-Ag was not associated with an increased risk of graft loss in patients CMV D+ patients (n = 395: HR 1.27 [0.92–1.74], p = 0.146) and in patients potentially exposed to CMV (i.e. all patients except D/R patients (n = 663: HR 1.03 [0.06–1.78], p = 0.92).

Then, we assessed risk of graft loss according to donor CMV serostatus in antigenemia-free recipients. Interestingly, as shown in Table 3, kidneys from CMV-positive donors were also associated with worse outcome in these patients, especially in R+ recipients. Overall these results show that the increased risk of graft loss associated with donor CMV seropositivity was independent from CMV viremia.

Table 3. CMV+ donor still an independent risk factor of graft loss in the absence of CMV antigenemia
CMV-Ag free patientsUnivariate analysis Multivariate analysis
HR95% CIp HR95% CIp
  1. HR, hazard ratio; 95% CI, 95% confidence interval; D+, CMV-positive donor, R+, CMV-positive recipient.
  2. Model 1: adjustment on creatinine, proteinuria, systolic blood pressure at 3 months, biopsy-proven acute rejection occurrence during the first year and date of transplantation.
  3. Model 2: model 1 + adjustment on mycophenolate mofetil, cyclosporine, anti-lymphocyte globulins use.
  4. Model 3: model 2 + adjustment on donor age.
All1.9241.174–3.1540.009Model 12.4521.424–4.2210.001
(n = 547)   Model 22.3371.354–4.0360.002
    Model 32.3211.344–4.0080.003
D R excluded2.2391.138–4.4680.020Model 13.9211.713–8.9780.001
(n = 321)   Model 23.7741.616–8.8170.002
    Model 33.7491.611–8.7220.002
R+2.4871.164–5.3140.019Model 13.3491.377–8.1480.008
(n = 216)   Model 23.2431.322–7.9590.010
    Model 33.3191.349–8.1640.009

Full HLA class I mismatch increases the detrimental impact of receiving a kidney from a CMV positive donor

The control of CMV replication in tissues relies on the recruitment of competent CMV-specific CD8+ T cells [6, 7], which, for most, is restricted to classical HLA class I expressed by recipient antigen-presenting cells. We hypothesized that interactions between recipient CMV-specific CD8+ T cells and CMV-infected donor cells could be impaired with lack of any HLA class I compatibility between the donor and recipient (four mismatches in HLA-A and -B antigens). The Figure 2 shows that the detrimental effect of receiving a kidney from a CMV-positive donor was indeed maximal in recipients with full HLA class I mismatch (upper left panel, p = 0.0004), and was even more pronounced in R+ patients (lower left panel, p = 0.0001). This detrimental effect remained after adjustment on variables of the previously described model 3 (All: HR 1.666 [1.359–5.234], p = 0.004; R+: HR 3.051 [1.012–7.254], p = 0.012). In contrast, D+ and D grafts had similar survival in patients with at least one HLA-I matching with their donor (Figure 2, right column).

Figure 2.

Partial human leukocyte antigen (HLA) class I matching protects against the detrimental effect of receive a kidney from CMV-infected donor. Death-censored graft survival in patients with full HLA class I AB mismatch (A and C) or partial HLA class I mismatch (B and D) by donor CMV serology. Upper and lower panels represent all patients (310 with full mismatch and 969 partially mismatch), and R+ recipients (n = 644), respectively. Graft survival is lower for D+ than D kidneys only in patients with full HLA class I mismatch (A, p = 0.0004; C, p = 0.0001).

At 2 years, the D+R+ recipients do not expand their CD8+ cell count in response to CMV as efficiently as did the D+R recipients

The aforementioned data suggest that the CD8+ cell response may play a role in the CMV-driven graft loss. We decided to further explore this immune response in this large cohort of patients. It has previously been reported that a large fraction of the CD8 repertoire is devoted to defense against CMV [37]. We hypothesized that the size of the CD8+ cell compartment would correlate with the anti-CMV response.

For the purpose of the present study, we analyzed the circulating CD8+ cell count at D0 and year-2 after transplantation by donor and recipient CMV serology (i.e. DR, D+R, DR+, D+R+) and CMV-Ag. The CD8+ cell counts, recorded for 663 at D0, was greater in R+ patients than in R patients (520 ± 270 vs. 407 ± 206/mm3, respectively; p < 0.0001). Two-year CD8+ cell counts were recorded for 164 DR (3 Ag+), 128 D+R (51 Ag+), 169 DR+ (69 Ag+) and 130 D+R+ (71 Ag+). At year 2, ANOVA analysis showed that CD8+ cell counts were different between groups presented in each panel of Figure 3. Paired-wise analyses indicated that CD8+ cell count was lowest for DR patients (vs. all other groups at year 2, p < 0.0001; Figure 3, upper panel). These data indicate that the size of the CD8+ cell compartment is driven by CMV. Interestingly, the 2-year CD8+ cell count was much lower in D+R+ than in D+R patients (599 ± 410 vs. 811 ± 647/mm3, p < 0.0001), but not different of that in DR+ (490 ± 397/mm3, p = 0.745). CMV-Ag was associated with increased number of CD8+-cells in D+R patients (1112 ± 678/mm3 vs. 530 ± 432/mm3, p < 0.0001), but not in D+R+ (605 ± 296/mm3 vs. 510 ± 224/mm3, p = 0.221) and DR+ (567 ± 318/mm3 vs. 558 ± 350/mm3, p = 0.902). In addition, the difference between D+R+Ag+ and D+RAg+ was dramatic (p < 0.0001), while no difference was observed between D+R+Ag+ and DR+Ag+ patients (p = 0.574). In addition, we found that NK and CD4+ cells that have been implicated in CMV immune response did not increase in patients exposed to CMV (Supplementary Figure S1). These data together suggested that long after the primary phase of CMV replication, D+R recipients kept a higher number of CD8+ cells as compared to D+R+ recipients, for whom renal allograft outcome was poorer.

Figure 3.

At 2 years, CMV drives a much weaker expansion in D+R+ patients as compared to D+R. Each graph represents number of CD8+ cells/mm3 at 2 year after transplantation by donor (D+/−) and recipient (R+/−) CMV serostatus. In upper panel, patients are presented by CMV serology group. In lower panel, patients are presented according to CMV-antigenemia occurrence during the first year. Mean is represented by a black bar, with the first and third quartiles. Upper and lower whiskers represent standard deviation. NS: no significant, *p < 0.0001; pair-wise PSLD Fisher test.

We next analyzed individual variations in CD8+ cell counts between D0 and year-2 posttransplantation. We had 414 patients for whom T cell subpopulation counts were recorded at both D0 and year-2. Since the size of the CD8+ cell compartment was similar in D+R+ and DR+ patients, we grouped them together for analysis of the individual evolution of the CD8+ cell subpopulation. We focused on patients who unambiguously reactivated CMV (i.e. 106 R+Ag+ and 43 D+RAg+). Spearman correlation tests showed that pretransplant CD8+ cell number was weakly correlated with the 2-year posttransplant count in R+Ag+ patients (R = 0.465, p = 0.0001) and D+RAg+ (R = 0.367, p = 0.017). As shown in Figure 4, CMV reactivation was associated with an individual increase of the CD8+ cell count in virtually all D+RAg+ patients (95%). In contrast, 36% of R+Ag+ patients showed a reduction of the CD8+ cell count after CMV reactivation. In addition, mean value of the individual variation of the CD8+ cell count was higher in D+RAg+ in comparison to R+Ag+ patients (197 ± 190 vs. 29 ± 79%, p < 0.0001). Together these data indicate that in R+ recipients the CD8+ response triggered by previous CMV-Ag was highly heterogeneous compare to D+RAg+.

Figure 4.

Early CMV viremia drives a long lasting expansion of the CD8+ cell compartment in all R patients, in contrast to R+ recipients in which the CD8+ response is highly heterogeneous. Distribution of patients by CD8+ cell count at day 0 (D0; x-axis) and year 2 (y-axis) for patients who had an episode of viremia during the first year posttransplantation (Ag+). R+ patients (n = 106) and D+R combinations (n = 43) are represented on the left and right panel, respectively. Black circle represents patients who received anti-lymphocyte globulins, and cross, patients who never received treatment. Black line is the bisector ([(Year 2 − D0)/D0] × 100 = 1).

Posttransplantation contraction of the CD8+ cell compartment increases the risk of graft loss in patients exposed to CMV

First, we evaluated the individual evolution of the CD8+ cell count ([(Year 2 − D0)/D0] × 100) in CMV-exposed patients. Then, we compared graft survival in patients with severe reduction in CD8+ cell count (≤10° percentile or reduction ≥40%) to others. R+ patients with such reduction showed significantly reduced graft survival (Figure 5, left panel, p = 0.003); importantly, the two groups had similar age (48.1 ± 13.2 vs. 49.7 ± 13.1 years, respectively, p = 0.58). Interestingly, severe CD8+ cell count reduction remained associate with increased risk of graft loss in 113 R+ patients who had received ALG (p = 0.014). In contrast, >40% reduction in CD8+ cell count had no impact on graft survival among DR patients (p = 0.991), which indicates that the reduction in the compartment is not by itself associated with poor graft survival. Therefore, evolution in the number of the CD8+ cell in kidney transplant patients might be a marker of the anti-CMV immune response and its reduction negatively affects graft survival.

Figure 5.

Severe reduction of CD8+ cell count after transplantation increases the risk of graft loss in R+ recipients. Graphs represent risk of death-censored graft survival by reduction in CD8+ cell count at year 2 in R+ recipients (left graph, n = 204) or DR patients (right graph, n = 121). Bold black lines indicate recipients with >40% reduction in CD8+ cell count (R+, n = 26; DR, n = 36); dashed lines, recipients without >40% reduction (R+, n = 178; DR, n = 85). CD8+ cell count reduction was associated with graft loss in R+ recipients (p = 0.003) but not DR patients (p = 0.991).


We report that latent CMV infection in the donor was an independent risk factor of kidney graft loss, especially for recipients with latent infection before engraftment. The impact of chronic CMV infection on long-term kidney graft function remains highly controversial. An epidemiologic study by Schnitzler et al. [24] reported that CMV latent infection in the donor was a risk factor of graft loss at year-2 in CMV-negative but not CMV-positive recipients. In contrast another study found that D+R+ rather than D+R transplants showed the lowest allograft survival at 3 years but this effect disappeared after adjustment for donor age [23]. Therefore, our work contrasts with these studies. However, it must be stressed that in our cohort, similarly to the aforementioned studies, the year-2 graft survival was not affected by latent CMV infection in the donor for R+ recipients (p = 0.2, data not shown), which suggests that the effects of CMV are subtle and cumulative and need a long time to become evident. Of note, our results are in line with other recent epidemiologic studies finding that CMV latency, especially in the donor, is an independent risk factor of late graft failure [25, 26], as well as other studies finding CMV infection associated with chronic allograft nephropathy [38]. Interestingly, our results suggest that R+ rather than R recipients are more susceptible to the adverse consequences of transplanting kidneys from CMV-infected donors. Therefore, we focused on this group of R+ patients.

Kidneys from CMV-infected donor carry a high-risk of viremia, as previously reported [32, 39]. Nonetheless, an episode of CMV-Ag was not a significant risk factor of graft loss, which agrees with Sagedal et al. [40]. Furthermore, the outcome was also worse for kidneys from CMV-positive donors in antigenemia-free R+ recipients. Therefore, CMV-Ag may not be a good surrogate marker of pathogenic effect of CMV over the long-term. Anyway, this high risk of viremia in R+ recipients transplanted with kidneys from CMV-infected donors suggests that the supplemental viral burden brought by the graft was not efficiently controlled by memory T cells present in CMV-seropositive patients.

Because CMV induces a massive accumulation of CMV-specific CD8+ T cells [13, 37], which have a crucial role in control of CMV replication [6], we analyzed the circulating CD8+ cell subpopulation by donor and recipient CMV serostatus and the occurrence of CMV-Ag. We found that CMV had a long-lasting and profound impact on recipients' circulating pool of CD8+ but not CD4+ or NK cells in D+R patients. Despite the large body of literature dealing with anti-CMV T cell response in transplantation, none has compared the long-term evolution of the CD8+ T cell count by CMV serostatus and in correlation with graft survival in such large number of patients (n = 414). Interestingly, among patients in whom CMV-Ag developed in the posttransplant period, only CMV-seronegative recipients showed marked expansion of the CD8+ cell count until at least 2 years as compared with CMV-positive recipients. This might cause poor ability to control chronic CMV replication and may account for the serious effects of CMV chronic infection in these recipients. Interestingly, in accordance with this interpretation we found that posttransplantation reduction of the CD8+ cell count increased the risk of graft loss in patients exposed to CMV. We might envision that the poor ability to increase the CD8 count in response to CMV reactivation might indicate a process of clonal exhaustion. The impact of clonal exhaustion on the control of chronic viral replication is well known [41]. Although exhaustion of CMV-specific T cells is debated, many studies have reported impairment of CMV-specific T cells over the long term, in older-aged infected individuals or in transplant patients [10-16]. Furthermore, a recent study reported that programmed cell death-1 (PD-1) signature was higher in CMV-specific T cells than in naïve or influenza-specific T cells [9]. The role of CMV-associated T cell impairment in chronic organ-specific dysfunction has not yet been investigated. Most of the studies of anti-CMV immune response in transplantation involved small series of patients, which did not allow for reliably assessing long-term graft survival. Interestingly, we have just completed a large clinical study (n = 1318 patients) in two independent kidney and lung transplant cohorts aimed at analyzing the association between graft survival and polymorphism of the PDCD-1 gene, a crucial gene in viral-induced T cell clonal exhaustion [42]. Results indicate that a polymorphism in this gene was a risk factor of CMV-associated graft loss in both cohorts of patients. Interestingly, this polymorphism also correlated with posttransplantation CD8+ cell counts only in patients exposed to CMV. We should acknowledge that the retrospective design and the lack of analysis of the specific anti-CMV response constitute important limitations of our work. Nonetheless, we feel that these results along with those of the genetic study should encourage further exploration of the role of CMV-specific CD8+ cell exhaustion in long term graft survival. The fact that we did not assess intragraft CMV replication by sensitive techniques in biopsies might appear as a limitation. However in our opinion, this strategy has important limitations for many reasons, intrinsic to the biology of the virus. CMV reactivations within a tissue are known to be transient and patchy and may leave behind durable and nonspecific histological scars, making it difficult to establish a causal relationship by histopathology or even PCR tests. In addition, reactivation occurring after the first year, when biopsies are seldom performed, will remain largely unrecognized. Hence, strategies based on analyzing intragraft viral replication in kidneys besides being costly and invasive are likely to lead to misclassification bias and failure to evidence the role of CMV in long-term graft survival. If indeed risk of severe acute infection (i.e. efficiently prevented by drug prophylaxis) is greater for naïve kidney recipients (R) than R+ recipients, our results suggest that in contrast, these naive recipients might have better control over chronic infections. This finding highlights that acute and chronic infections are distinct pathogenic processes with different rules for both the pathogen and the host, as was shown in other experimental models [41].

Finally, we identified full HLA-I mismatch as a major risk factor in the detrimental effects associated with CMV seropositivity of the donor. This result might be explained by the fact that memory CD8+ T cells from R+ recipients are restricted to their HLA class I and are therefore poorly able to clear viral replication in full HLA-I mismatched target cells of donors, especially endothelial cells. Indeed, those recipients may have some difficulties to generate new CMV-specific CD8+ T cells able to recognize CMV peptides presented by fully allogeneic class I. So, poor ability of T cells to control intragraft CMV replication could favor graft injuries especially the development of vasculopathy. Therefore, our data suggest that full HLA-I mismatch should be avoided in R+ patients receiving a kidney from a CMV-infected donor.

In summary, we report that positive CMV serology in the donor was an independent risk factor for graft loss, especially in R+ recipients with full HLA class I mismatch, even in antigenemia-free patients. Moreover, R+ recipients with marked posttransplantation reduction of the CD8+ cell count had the lowest graft survival. Theses results may lead to improve organ allocation strategies in order to prevent CMV-associated chronic allograft dysfunction.


We thank Laura Smales for editing the manuscript.

Funding source: None.


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