(Val)ganciclovir is used to treat cytomegalovirus (CMV) infection following solid organ (SOT) or hematopoietic stem cell (HSCT) transplantation. Treatment failures occur, but the contribution from 39 known ganciclovir-related mutations (GRMs) in the CMV-UL97 gene remains controversial. We propose a categorization of these GRMs potentially useful when interpreting sequence analyses in clinical settings. The UL97 gene was sequenced from first/recurrent CMV infections among consecutive SOT or HSCT recipients during 2004–2009. GRMs were categorized as: Signature GRM (sGRM) if in vitro ganciclovir IC50 ratio for mutated versus wild-type virus >2 (n = 24); polymorphic GRM (pGRM) if ratio <2 (n = 15). (Val)ganciclovir treatment failure was defined as persistent viremia for 30 days or switch to foscarnet within this period. Of 99 (49 HSCT and 50 SOT) recipients with one CMV infection episode, 15 (13 HSCT and 2 SOT) experienced a total of 19 recurrent infection episodes. The prevalence of sGRM was 0% at start of first episode, whereas at start of recurrent episodes, prevalence was 37%. Only one sGRM was present at a time in individual patients. Patients with CMV containing an sGRM (vs. wild type)—but not with a pGRM—were at excess risk of treatment failure (odds ratio = 70.6 [95% CI:8.2–609.6]; p < 0.001). sGRMs emerged only following longer termed use of antiherpetic drugs and usually in recurrent CMV infection episodes. Risk of ganciclovir treatment failure was raised if an sGRM was detected.
Cytomegalovirus (CMV) infection and sometimes disease frequently complicate the course after solid organ and hematopoietic stem cell transplantations (SOT and HSCT, respectively) [1, 2]. Ganciclovir is the drug of choice to prevent and treat CMV infection with foscarnet and cidofovir as alternatives. Being a herpes virus, other antivirals like (val)acyclovir exert an antiviral selection pressure also against CMV; although the doses of (val)aciclovir necessary to prevent Herpes simplex virus and Varicella-zoster virus infection is lower than the doses required to prevent CMV infection . Although ganciclovir is able to treat most cases of CMV infection successfully, treatment failures occur either because of treatment-limiting side effects or the selection of ganciclovir-related resistant mutations (GRMs) in the viral gene targets—UL97 and UL54.
GRMs in the CMV UL97 are the best described mechanism for development of treatment failure due to resistance, and 39 GRMs in this gene have been identified and their influence on ganciclovir sensibility in vitro is well described [4-7]. However, their effect on the clinical response to ganciclovir treatment in vivo remains to be clarified . Here, we explore the emergence of various types of GRMs and how their presence affects the ability to contain CMV infection by use of ganciclovir in a large and consecutive series of transplant recipients with one or more episodes of infection. A priori, we grouped the 39 known GRMs in UL97 based on reported ratio of inhibitory concentration of ganciclovir required to reduce replication by 50% (i.e. the IC50) in viral isolates with each of the GRM versus wild type in vitro; arbitrarily we decided to group the GRM dichotomously depending on whether the ratio was above or below 2.
Materials and Methods
Consecutive patients who developed CMV infections following SOT and/or HSCT from 2004 to 2009 were considered for inclusion. The immunosuppressive regimens used reflected per protocol guidelines of each of the departments responsible for the posttransplant care. Following HSCT, cyclosporine adjusted to through blood level between 200 and 400 ng/mL is standard of care. Prednisone is added in case of graft-versus-host disease (GVHD); starting at 2 mg/kg for acute GVHD, and 1 mg/kg for chronic GVHD. After clinical response, prednisone is tapered.
Following heart transplantation, thymoglobulin 1.5 mg/kg for 3 days with 1 g methylprednisolone for 2 days followed by 125 mg for an additional 3 days is used. Prednisone 0.2 mg/kg, tapered to 0.1 mg/kg after 3 months and to zero after 1 year and Mycophenolat mofetil 1–1.5 g × 2 daily continuously. Cyclosporine is used adjusted to through blood levels between 200 and 350 ng/mL for the first 6 weeks then reduced to 150–250 ng/mL and again to 100 ng/mL.
Following lung transplantation, thymoglobulin 1.5 mg/kg for 3 days with 500 mg methylprednisolone for 1 day followed by 125 mg methylprednisolone each day for 3 days is used. Prednisone 15 mg daily for 2 weeks, tapered to 5 mg daily over 4 weeks. Azathioprine 1.5 mg/kg from postoperative day 1. Ciclosporine from postoperative day 1 with target range 200–240 ng/mL for first 3 months, 150–200 for months 4–12.
Following liver transplantation, a single dose of methylprednisolone 1000 mg is given intraoperatively. In the days following the transplantation, prednisolone is tapered gradually from 200 mg on day 1 to 40 mg on day 5. For the remaining first month, 20 mg is given daily tapered to 15, 10 and 7.5 mg daily until month 6. For the next 6 months, 5 mg is given daily after which the drug is discontinued. Tacrolimus is given twice daily, aiming at through levels of 10–12 ng/mL in the first month, 8–10 ng/mL in month 2, 7–9 ng/mL in months 3–6, 6–8 ng/mL in months 7–12 and 4–6 ng/mL after 1 year. Cellcept is given twice daily at a dosage of 1000 mg continuously.
Following kidney transplantation, induction was given with either five doses of daclizumab or two doses of basiliximab and methylprednisolone 500 mg immediately before transplantation. Maintenance immunosuppression started at day 1 as triple therapy with cyclosporine, mofetil mycophenolate and steroids. Cyclosporine was dosed twice daily aiming at through levels of 150–200 ng/mL for 2 months, 100–150 ng/mL for 10 months and after 1 year for 50–100 ng/mL. Dosage of mofetil mychophenolate was 1 g twice daily. Prednisolone dosage was 40 mg daily tapered to 15 mg on day 22 and to 5 mg at month 7.
Anti-CMV chemoprophylaxis with valganciclovir was used following SOT for the first 3 months and thereafter emerging infection was treated preemptively also using valganciclovir. Following HSCT, valaciclovir (0.5 g bid) was given to prevent reactivation of Herpes simplex virus and Varicella-zoster virus, while solely a preemptive strategy toward emerging CMV infection using valganciclovir was employed. Routine surveillance for CMV infection using CMV polymerase chain reaction (PCR) as well as performed diagnostics in case of suspected infection was standard of care at all departments. Following SOT, CMV PCRs were performed weekly and sometimes longer intervals in the first 3 months following discontinuation of prophylaxis. In the following months, tests were performed in case of suspected infection. Data used for identification and characterization of patients were extracted retrospectively from electronic health records of transplant recipients. Residual samples from CMV DNA determinations stored in a central biobank were identified and used for sequencing of CMV UL97. Study approval was obtained from the Regional ethics committee.
CMV infection episodes
The start and end of the first posttransplant CMV infection episode was defined as: the time of the first positive quantifiable CMV PCR after time of the transplantation, and the time of the first of two consecutive negative CMV PCRs thereafter. Subsequent episodes of infection were numbered continuously and each had to be separated from the prior episode by at least two negative CMV PCRs.
The CMV PCR defining the start of an infection episode was selected for sequence analysis. When infection episodes were documented by two or more positive CMV PCRs, follow-up samples during treatment and the last quantifiable sample prior to the samples defining the end of the episode were also selected for sequencing.
Based on a literature search, all mutations listed in Appendix (Supporting Information) were considered to be GRM (n = 39). According to in vitro testing of ganciclovir IC50, GRMs were further categorized as polymorphic GRM (pGRM) (n = 15) (if the mutant/wild-type Ganciclovir IC50 ratio was below 2) or signature GRM (sGRM) (n = 24) (if the mutant/wild-type Ganciclovir IC50 ratio was above 2). This classification was done prior to commencing the study (see Appendix).
The PCR products from the selected samples were prepared by use of two primer sets covering codons 435–645 in UL97 using the Roche (Mannheim, Germany) High-Fidelity PCR Master . The program for cycling was as follows: denaturation at 95°C for 2 min, followed by 45 cycles of 95°C for 20 s, 60°C for 30 s and 72°C for 30 s. Final elongation was done at 72°C for 4 min. Amplicons were purified (QIAquick 96 PCR Purification Kit, QIAGEN, Hilden, Germany) and then sequenced using the same primers as PCR. Sequences were analyzed with Vector NTI Advance 11 (Invitrogen, Carlsbad, CA, USA) and compared with that of the laboratory Towne strain AD169., GenBank accession numbers: FJ616285  For episodes of treatment failure, the UL54 gene was also sequenced and analyzed for the presence of GRMs at initiation of the episode.
Treatment and treatment failure
In the analyses, two different definitions of treatment failure were explored: (1) Where a treatment episode was defined as a period of (val)ganciclovir treatment initiated after or at the time of the infection start till the treatment was stopped or 30 days later whichever came first. Failure was defined as the presence of two or more positive and no negative CMV PCRs between treatment start and day 30 thereafter or a switch to foscarnet (irrespective of reason) prior to day 30. If patients were lost to follow-up prior to day 30, the survival status of patients was examined: if the patients had died within 90 days of the infection start failure was considered, otherwise success was considered (i.e. clearance of the infection was presumed). (2) Where a treatment episode was defined as a period of (val)ganciclovir treatment of 14 days and failure was considered if CMV PCR at day 14 was at a value of ≥1000 copies/mL, as defined and used in previous publications by van Der Beek et al. [8, 10].
As a well-described side effect to the use of ganciclovir is cytopenia, a switch from this drug to foscovir—defined as treatment failure in definition (1)—could be attributed to of bone marrow suppression . To address this issue, the patient file for all patients who were switched to foscovir was reviewed. If indicated in the file that the switch was made due to marrow suppression or suspicion of toxicity the patient was censored or excluded and sensitivity analyses were performed. Likewise, sensitivity analyses were performed after censoring or exclusion of patients who died prior to the end of the treatment episode.
The odds ratio of treatment failure was modeled as a repeated measure logistic regression using generalized estimation equation with robust standard errors adjusted for nonindependency of episodes.The following variables were considered for inclusion in the models: sequence analysis result sGRM/cGRM/wild-type virus, type of transplantation, gender, age, time from transplantation, time to treatment start after diagnosis of infection, viral load at treatment start, use of methylprednisolone, rituximab and/or thymoglobulin 30 days prior to diagnosis of infection until 7 days after diagnosis. Analyses were performed using SAS 9.2 (SAS Institute, Cary, NC, USA) statistical software. All reported p-values are two-sided using a level of significance of 0.05.
Recently, we documented the selection of a GRM that clinically hampered the ability to control the CMV infection in an SOT recipient resulting in a life-threatening situation due to the inability of ganciclovir to control the CMV infection because of the selection of an sGRM (A594V).
A 69-year-old male with a history of kidney transplantation was diagnosed with CMV infection (CMV PCR 500 copies/mL). At the time of diagnosis, the patient was still following a 3-month posttransplant routine course of anti-CMV chemoprophylaxis with valganciclovir. The CMV PCR was repeated with 2-week intervals and no adjustment in antiviral treatment was made. One month later, the CMV PCR value was 3400 copies/mL and valganciclovir was increased to treatment dosages. However, the infection persisted and 1 month later the patient was switched to i.v. formulation of ganciclovir. CMV DNA copy numbers kept increasing and the patients developed severe systemic malaise with central nervous system (CNS) involvement and fever. Ganciclovir-resistant CMV was suspected and the patient was switched to foscarnet and CMV PCR became undetectable 19 days after the treatment switch. Post hoc sequencing of the UL97 gene found the emergence of a mutation at codon 594 (A594V) that is known to increase the IC50 ratio (GRM vs. wild type) in vitro to 9.8  6 weeks after the onset of CMV replication that persisted thereafter.
This observation leads to the design of the study providing result described below.
Patients and CMV infection episodes
In total, 99 (49 HSCT and 50 SOT; heart [n = 2], liver [n = 12], lung [n = 23] and kidney [n = 13]) patients consecutively diagnosed with their first episode of posttransplant CMV infection were included in the analysis. Fifteen (13 HSCT and two SOT) patients had one or more (total of n = 19) recurrent episodes of CMV infection. In total, 118 episodes of infection were evaluated and sequence analyses were performed. Patient characteristics are shown in Table 1.
Table 1. Patient characteristics
Number of included patient according to type of transplantation. CMV IgG status of donor (D) and recipient (R) at the time of the transplantation. Number of first and recurrent posttransplant CMV infection episodes.
(n = 49)
(n = 50)
(n = 99)
No. (%) male
Age (median, IQR)
42 (23; 51)
46 (33; 56)
44 (27; 55)
Donor (D)/Recipient (R) CMV serostatus, No. (%)
CMV infection episodes
Total CMV infection episodes
Results of sequence analyses
In 21/99 (21%) first CMV infection episodes and in 6/19 (31%) recurrent infection episodes, a total of four pGRMs (Q449K, H469Y, N510S, D605E) was detected in the dominant viral population (Figure 1) present when the infection was diagnosed. In individual infection episodes, only one of these four pGRMs were detected at a time, and if detected initially in the episode it was also detected in the last positive CMV PCR sample from that episode. As for the sGRMs, four other types were detected (Figure 2) in 7/19 (37%) recurrent episodes, but not in either of the 99 initial episodes (Figure 1); of note, one of the four sGRMs was similar to the one detected in the case study (A594V). Only one sGRM was detected at a time in individual infection episodes, and if detected initially in the episode it was also detected in the end of the episode. In 11/19 (58%) recurrent infection episodes, either a pGRM and/or an sGRM was detected. In only one of the seven infection episodes where an sGRM was detected, a pGRM was detected in a previous episode (and also in the episode where the sGRM was found), and a prior detected pGRM did not predict the subsequent emergence of sGRMs. In one infection episode preceding an infection episode where an sGRM emerged, this mutation was not observed.
Treatment outcome and risk factors
(Val)ganciclovir treatment failure occurred in 27/118 (23%) and 64/118 (55%) episodes according to definitions (1) and (2), respectively (see Methods section). Of the 27 treatment failures using definition (1), 18 (18%) occurred during the first episodes of infection and nine (47%) during the recurrent episodes. The components that contributed to the 27 treatment failures were in: (i) Ten patients lack of resolution of the viremia by day 30; (ii) Eleven patients switch to foscarnet and (iii) Four death while the patient continued to have viremia. The reason for the switches to forcarnet includes potential treatment-limiting toxicity to ganciclovir in six of the 11 patients. CMV contributed to all four deaths among patients who died in conjunction with the CMV infection episode; other factors that also contributed were graft versus host reaction, pulmonary Aspergillus fumigatus, bacterial sepsis and a stroke complicating an admission for bacterial and CMV infection.
After adjustment for possible confounders, presence of an sGRM in an episode of infection was associated with (val)ganciclovir treatment failure (odds ratio 70.6 [95% CI: 8.2–609.6] [p < 0.001]) when compared to episodes of infection with CMV without GRMs (called wild-type virus) (Figure 3). Conversely, the presence of a pGRM without an sGRM failed to significantly affect the risk of experiencing treatment failure. Other factors that contributed to the risk of treatment failure included a higher CMV viral load when the antiviral therapy was initiated. Of note, neither type of transplantation nor other potential confounders were significantly associated with treatment failure. These results were consistent irrespective of whether treatment failure definition 1 or 2 (see Methods) was applied. In addition, exclusion or censoring of the six patients where treatment limiting∖ toxicity to ganciclovir may have contributed to the switch to forcarnet and/or the four patients who died failed to affect the findings depicted in Figure 3.
Additional sequence analyses of the UL54 gene
Of the 27 infection episodes where treatment failure was observed, we were able to sequence the UL54 gene from 22 episodes. In one of these episodes (a recurrent episode) an sGRM was found (T503I); in this episode no GRMs in UL97 was also detected.
In the present material, four ganciclovir-related CMV UL97 mutations (so-called GRMs) were found in 21% of the first posttransplant CMV infection episodes analyzed, but all GRMs belonged in the category of polymorphic without major impact on ganciclovir susceptibility in vitro (pGRMs; i.e. a IC50 ratio [GRM vs. wild type] below 2 from published literature). When recurrence of the infection occurred GRMs were found in 58% of the episodes and both pGRMs and sGRMs (i.e. a IC50 ratio [GRM vs. wild type] above 2 from published literature) were found. The odds ratio of (val)ganciclovir treatment failure was significantly higher in episodes of infection where an sGRM was found versus episodes of infection with only wild-type virus. Conversely, the sole presence of a pGRM failed to significantly affect the risk of treatment failure.
The main findings with clinical implications in this study are (1) the association between sGRM and markedly elevated odds of treatment failure from ganciclovir, and (2) the predominant appearance of these sGRMs in recurrent episodes of infection. The association representing our first main finding was reproducible regardless which definition of failure to ganciclovir treatment was used, and collectively the results suggest that prolonged courses of therapy or switch of drug to foscovir is needed to suppress a infection with CMV harboring one of these mutations. In the present material, four types of sGRMs were found and it seems plausible that a similar outcome could be expected for other sGRMs isolated reported in the literature. Of note, we only observed a total of eight GRMs in our extensively sequenced clinical cohort, and it is unclear whether all of the 39 reported GRMs in vitro will emerge in humans, as 11 of these have only been documented in laboratory strains experimentally. Some GRMs may impact on the virus’ fitness to sustain an infection . In previous studies, the four sGRMs we observed and two additional sGRMs (H520Q and C592G) accounted for more than 80% of the ganciclovir-resistant clinical isolates which confirms the validity of our observations [14, 15].
As expected with endpoints defined by a time limit, increased viral load at the time treatment was initiated was also associated with treatment failure. The results confirm that a more established infection requires longer time to treat, why diagnosing and treating emerging infection early, when the viral load is still low, should reduce the needed treatment time in these patients .
The validity of our second main observation is supported by results from previous studies indicating an association between GRMs and recurrent episodes of CMV infection in SOT, and by the isolated finding of sGRMs in recurrent episodes of infection after HSCT [17-19]. Conversely, a recent study of treatment failure in HSCT recipients found an increased risk in first episode of infection versus recurrence, but other factors than selection of GRMs seemed to be underlying these observations .
As described in our case story, sGRMs can, if circumstances allow it, develop during the course of any episode of infection including the initial episode if viral replication is not blocked initially and completely. In our case, the infection emerged during a period of insufficiently dosed valganciclovir prophylaxis which was maintained for an additional month, thus allowing the virus to continue replicating under an antiviral selection pressure. The late emergence of the virus harboring the sGRM (it was only detected 6 weeks into the course of treatment of the infection) in our case suggests that only extensive antiviral selection pressure results in the emergence of these types of mutations. This is a surprising finding as the genetic barrier to select for either of the four sGRMs was only one as depicted in Figure 2. That is, only one mutation is required to insert the amino acid in the codon needed to make the virus resistant to ganciclovir. More detailed studies are required to better understand the relationship between the kinetics of CMV replication in various compartments of the body, drug penetration into each of these compartments and the rate of spontaneous mutation rate of the CMV polymerase (i.e. the driver of the selection of mutated virus).
Our findings however also underline that the emergence of sGRMs during the course of a first CMV infection after transplantation is a rare event. In the seven infection episodes where an sGRM was detected in our subsequent more comprehensive analysis, the sGRM was detected only in recurrent episodes of infection. Of note, if present the sGRM could be detected in the first positive CMV PCR sample of the episode suggesting that it was selected prior to commencing treatment for the recurrent infection. The emergence of these sGRM's initial in the course of the recurrent infection hence suggest that selection may also occur in-between infection episodes—at least between infection episodes that can be detected by a positive CMV PCR from plasma. Recurrent infection episodes occur typically more frequently in patients with more extensive degrees of immunosuppression and hence more frequently in HSCT than in SOT recipients (in our material 13 of the 15 patients [87%] with one or more recurrent episodes were HSCT recipients, where 49% of the patients with a first episode received this type of transplantation) . More extensive immunosuppression renders patients at excess risk of experiencing reactivation of multiple other types of herpes viruses, and many of these patients are typically using chemoprophylaxis with (val)acyclovir which has some—although not sufficient—antiviral activity also against CMV . Hence, the combination of poorer immunosurveillance and more extensive use of antiviral chemoprophylaxis aimed at other herpes viruses as a consequence of the more advanced immunosuppression may explain our observations.
The clinical implication from the discussion above is that a CMV drug resistance test should be considered if CMV episodes do not resolve on correctly dosed (val)ganciclovir. The probability that such a test detects the presence of treatment-limiting viral drug resistance mutations is particularly high in recurrent episodes of infection or if an episode was not initially treated with ganciclovir in appropriate therapeutic doses.
Although not part of the primary focus of this paper, we did also sequence the CMV UL54 gene in the first CMV PCR positive sample from 22 of the 27 episodes where (val)ganciclovir treatment failure was observed, and found one sGRM at codon 503 (T503I). The in vitro IC50 ratio for this mutation has been reported to be 2.9 . Further studies are required to more fully understand the possible contribution from GRMs in UL54 to (val)ganciclovir treatment failures, and until then, our data do not support sequencing of this gene as part of routine care.
The limitations of this study include the retrospective design which could cause variations in monitoring frequencies and follow-up in HSCT versus SOT recipients. Plasma concentrations of ganciclovir were not measured and inadequate dosing or failure to adapt to renal function could attribute to a prolonged treatment response. We used Sanger sequencing where a mutation can only be detected if present in more than approximately 20% of the virus population. Hence, it is possible that mutations in minority virus variants at the time of sampling remained undetected, and only deep sequencing techniques can fully describe the kinetics of the emergence of sGRMs. Finally, we used two different rather diverse approaches to define treatment failure and found consistent results for our main finding. However, currently there is no consensus on how to define treatment failure in CMV infection and the definitions vary by as much as persistence of viremia from 2 to 8 weeks after initiation of treatment [8, 22, 23]. In order to optimize guidelines on how to interpret evolution of CMV viral load during course of treatment, more evidence variation in clinical and virological (i.e. resistance selection) outcomes from applying various types of definitions applied in routine practice and in controlled trials is urgently required. Collaboration among transplant units with different approaches and with collectively large number of patients will be required to resolve this.
In conclusion, irrespective of which type of transplant and/or prophylactic strategy is used, our findings suggest that when observing clinical treatment failure to (val)ganciclovir in an episode of posttransplantation CMV infection, the selection of an sGRM should be suspected and sequencing of the UL97 gene can confirm this. Switch to foscarnet should be considered if an sGRM is found. The probability that an sGRM explains failure to respond to (val)ganciclovir is higher in recurrent than in first infection episodes, and in particular if other potential contribution causes can be ruled out including inadequate dosing, malabsorption or noncompliance to valganciclovir.
The manuscript was not prepared or funded by a commercial organization. The authors of this manuscript have no conflicts of interest to disclose as described by the American Journal of Transplantation.