Diverse Cytomegalovirus UL27 Mutations Adapt to Loss of Viral UL97 Kinase Activity under Maribavir

Abstract

In vitro resistance to maribavir (MBV), a cytomegalovirus UL97 kinase inhibitor currently in clinical trials, is known to result from viral UL97 mutations that confer moderate to high-level resistance and UL27 mutations that confer low-level resistance. To add to the four reported UL27 mutations, cytomegalovirus isolates or strains were propagated under MBV. Four clinical isolates evolved UL27 mutations, which were first detected after 8 to 30 passages under drug selection. In three separate experiments, laboratory strain T2294, which contained an exonuclease mutation, developed UL27 mutations at 10 to 12 passages under MBV. Most of these isolates and strains also developed a UL97 mutation, commonly T409M, before or after the appearance of the UL27 mutation. The passage of two laboratory strains genetically defective in UL97, in the absence of MBV, likewise resulted in UL27 mutations. The nine UL27 mutations observed included multiple instances of point, stop, and frameshift mutations, which were individually transferred to a reference CMV strain and which were shown to confer two- to threefold increases in MBV inhibitory concentrations. In contrast, seven common UL27 amino acid changes found in baseline clinical isolates conferred no MBV resistance. The mutants with UL27 mutations had slightly attenuated growth. The frequent mutation of UL27 suggests that its normal expression is mildly disadvantageous to the virus in the absence of UL97 kinase activity, whether the latter results from MBV inhibition or a genetic defect. Although the function of UL27 is unknown, it does not appear to be a direct antiviral target for MBV.

Maribavir (MBV) is a benzimidazole l-riboside inhibitor of the human cytomegalovirus (CMV) UL97 kinase (1) that is undergoing clinical trials for the prevention of posttransplantation CMV disease. Initial dose-ranging trials that showed favorable safety and toxicity profiles (21), along with indications of anti-CMV activity in human immunodeficiency virus-infected subjects (15), led to a subsequent randomized trial with stem cell transplant recipients, which showed a significant reduction in CMV infection in those given the drug prophylactically (23). As a result, further clinical trials are ongoing to assess the antiviral efficacy and safety of the drug in larger randomized trials with transplant populations.

Given the limited duration of drug administration and the few cases studied to date, there are not yet any reports of MBV resistance in CMV isolates from treated subjects. Because the antiviral mechanism of MBV differs from the mechanisms of the currently licensed systemic anti-CMV drugs ganciclovir, foscarnet, and cidofovir, all of which target the viral DNA polymerase, cross-resistance between MBV and the existing drugs is not expected and none has been reported. The CMV UL97 kinase is involved in the antiviral actions of both MBV and ganciclovir, since this kinase is responsible for the initial phosphorylation of ganciclovir that is essential for that drug to have antiviral activity (3). Well-known UL97 mutations that commonly confer ganciclovir resistance, located at codons 460, 592, 594, and 595, do not confer cross-resistance to MBV (9). The propagation of CMV isolates and strains in vitro under MBV results in UL97 mutations at or near the ATP-binding domain (codons 353, 397, 409, and 411) that confer moderate to high-level MBV resistance but not ganciclovir cross-resistance (1, 4, 7).

Separately, it has been reported that mutations in UL27, a CMV gene of unknown function, are selected after the in vitro propagation of CMV and confer a relatively lower level of MBV resistance than the known UL97 mutations. In one case, UL27 mutation L335P was identified in a viral strain resistant to other experimental drugs and further propagated under MBV (13). Three additional instances of UL27 mutations (R233S, A406V/C415stop, and W362R) were also found after the propagation of laboratory CMV strain AD169 under a carbocyclic analog of MBV, and these mutations were confirmed to confer low-grade MBV resistance by transfer to a reference CMV strain (5).

To increase the understanding of the relative frequency and diversity of UL27 mutations in CMV strains exposed to MBV in vitro and to assess the relative levels of drug resistance conferred, additional experiments were performed to propagate CMV cultures under MBV and characterize the resulting UL27 mutations. To distinguish whether UL27 mutations are a specific result of MBV exposure or whether it is an adaptation to loss of the UL97 kinase, two CMV strains that were genetically defective in UL97 were serially propagated without any added drug and were monitored for UL27 mutations.

MATERIALS AND METHODS

Viruses and cell cultures.

Four baseline clinical CMV isolates (isolates C005, C076, C327, and C336) were from transplant recipients who had not received prior antiviral therapy and were used at three to nine passages from primary isolation. Several laboratory CMV strains modified from strain AD169, all of which contained a secreted alkaline phosphatase (SEAP) reporter gene cassette for rapid viral quantitation, have been described previously, including baseline strain T2211 (6); strain T2294, which contains UL54 pol mutation D413A that increases the DNA replication error rate (4); and strain T2266, which contains a UL97 truncation after codon 536 that results in a UL97 defective phenotype (7). Strain T2890 was derived from strain T2266 by restoring a complete UL97 sequence containing a mutation, V353A, similar to that in published strain T2807, but without other amino acid changes in UL97 that resulted from interstrain variation (7). Strain T2394 was derived from strain T2266 by restoring the full wild-type UL97 sequence and replacing codons 430 to 497 of the UL27 gene with a CMV major immediate early promoter-driven humanized Renilla green fluorescent protein expression cassette and an adjacent SwaI restriction site, analogous to the derivation of strain T2092 reported previously (5), resulting in a virus with a fluorescent cytopathic effect. Strain T2819 was independently derived by mutagenesis of bacterial artificial chromosome (BAC) clone pHB5 of CMV strain AD169 (2) to introduce the K355M mutation into the UL97 gene, thereby inactivating the kinase by disruption of its critical lysine residue (10). Mutagenesis was accomplished by using the galK selection and counterselection system with conditionally recombinogenic Escherichia coli strain SW102, according to a previously published protocol (22). The resulting K355M BAC was transfected into fibroblasts to reconstitute live mutant virus. Human foreskin fibroblasts were used for the routine propagation of virus stocks and transfection cultures; but human embryonic lung fibroblasts were used for serial propagation under MBV, serial propagation of UL97-defective strains, and yield reduction assays (8).

Propagation under MBV.

MBV (1263W94) was provided by GlaxoSmithKline and was diluted into cell culture medium from a 100 mM stock solution in dimethyl sulfoxide. Viral cultures were serially propagated in fibroblast cultures with initial MBV concentrations of 0.3 μM for clinical isolates and 0.1 μM for strain T2294; these concentrations match the baseline MBV 50% effective concentrations (EC₅₀s) published previously (4, 7). Thereafter, the MBV concentrations were increased up to 10 μM when the viral growth appeared to improve under drug and the abnormal UL97-defective cytopathic effect lessened, as described previously (4, 7). Cultures were passed weekly to fresh fibroblast monolayers as ∼20% of the infected cells from the prior passage. Strains T2266 and T2819, which were defective in UL97, were serially propagated without the use of any MBV.

Genotypic analysis of viral strains.

Aliquots of infected cells were frozen at each viral passage. After every five passages, infected cell DNA extracts were amplified by PCR and checked for their UL97 and UL27 sequences. If changes from the baseline sequences were observed, extracts of infected cells from intermediate passages were analyzed to determine the earliest passage that showed a sequence change. Sequencing was performed by PCR amplification and automated dideoxy sequencing with a BigDye Terminator fluorescent dye kit (Applied Biosystems) and an automated sequencer instrument. This technology is expected to detect the presence of a mutated sequence in the population when it is present in ∼20% of the total population (20). Estimates of the percentage of viruses with a mutated sequence in a population were made by comparison of the peak heights in the chromatogram files generated by the sequencing system.

Transfer of UL27 mutations to strain T2394.

The UL27 mutations observed were transferred to strain T2394 by using a contransfection method similar to that used to generate other strains with UL27 and UL97 mutations (4, 5, 7). Briefly, T2394 genomic viral DNA was extracted from the supernatant of infected cell cultures, digested with SwaI at a unique site within UL27 adjacent to the fluorescent reporter cassette, and cotransfected with a cloned DNA segment (AD169 nucleotides 31236 to 36674; GenBank accession number X17403) that contained the UL27 mutation(s) to be studied. Recombinant viruses were recognized by the absence of a fluorescent cytopathic effect. A small fraction of fluorescent infected cells (parental strain T2394) was sometimes seen after cotransfection, and plaque purification was performed at least one step beyond the stage at which fluorescence was no longer observed, followed by sequencing of the entire UL27 sequence to confirm the presence of the desired mutation and the absence of unintended ones.

Phenotypic assay of MBV sensitivity.

SEAP yield reduction assays were performed as described previously (6, 8). Cell-free virus stock was inoculated onto 6-day-old human embryonic lung fibroblast monolayers in 24-well culture plates at a multiplicity of infection (MOI) of 0.01 to 0.02 and cultured for 6 days under a range of MBV concentrations in six wells per assay. The drug concentration required to reduce the supernatant SEAP activity by 50% (EC₅₀), assayed by using a chemiluminescent substrate, was determined by curve fitting and was used to compare the sensitivities of strains to the sensitivity of baseline strain T2211. To allow for variations in culture conditions and to provide statistical significance for relatively small differences in drug sensitivity, at least 12 total assay replicates performed on at least four separate assay dates were used to calculate a mean EC₅₀ and the standard error. Simultaneous assays were performed with control strain T2211 on all setup dates to obtain appropriate EC₅₀s for the combinations of cell cultures and drug concentrations used. Strain T2890 (UL97 mutation V353A) was used as a resistant control strain that was expected to confer an ∼15-fold increased MBV EC₅₀ over that for strain T2211 (7).

Comparative growth rates of mutants with UL27 mutations.

Multicycle viral growth curves were assessed by serial measurement of the SEAP activity in the supernatants of 24-well cultures of human embryonic lung fibroblasts at days 1, 4, 5, 6, 7, and 8 after the inoculation of a specific strain at an MOI of 0.01 to 0.02 (6, 7). Since the effective initial inoculum has an important influence on the subsequent growth curve and the SEAP activity at 24 h postinoculation is a measure of the MOI (6), care was taken to compare the growth curves of virus stocks that were inoculated into wells of the same batch of cultured fibroblasts and that showed similar SEAP activities at 24 h. Strain T2266 was used as a severely growth-attenuated UL97-deficient control strain and was inoculated at a slightly higher MOI. The mean SEAP activity and standard deviation in four replicate wells were used to construct the growth curves.

RESULTS

UL27 and UL97 mutations selected after serial passage.

Table 1 summarizes the temporal evolution of the UL27 and UL97 mutations in the strains and isolates serially passaged under MBV, including the passage number and the MBV concentration at which each mutation was first detected. For the four clinical isolates, a mutation was observed in UL27 before one was observed in UL97 in three instances (experiments M12, M17, and M20). These occurred at passages 8, 22, and 30, respectively, under relatively low concentrations of MBV (1 μM or less). The appearance of an initial UL27 mutation did not resolve the abnormal UL97-defective cytopathic effect depicted previously (4), even with several further passages under MBV, unless an additional UL97 mutation emerged. In experiments M12 and M17, mutants with the UL27 mutation became the only detectable sequence population eight passages after its first appearance; but in experiment M20, mutants with the UL27 mutation were still only ∼50% of the sequence population eight passages later, by which time a UL97 mutation had evolved. For the fourth clinical isolate (experiment M18), UL97 mutation V353A appeared first and was among the mutations described in a previous report (7). This was followed three passages later by the appearance of a UL27 mutation when the drug concentration was increased from 4 μM to 10 μM. Mutants with this UL27 mutation remained ∼60% of the sequence population after six additional passages.

TABLE 1.

UL27 and UL97 mutations selected under MBV

Expt no.	CMV source	Name	UL97			UL27
			Mutation	Passage no.	MBV concn (μM)	Mutation	Passage no.	MBV concn (μM)
M12	Clinical isolate	C327	T409M	18	0.3	A269T	8	0.3
M17	Clinical isolate	C336				W362stop	22	0.3
M18	Clinical isolate	C076	V353A	9	4	del301-311a	12	10
M20	Clinical isolate	C005	T409M	37	4	218-delCb	30	1
M27	Lab strain	T2294	T409M	7	3	W153R	10	10
M28	Lab strain	T2294	T409M	4	0.5	E22stop	12	10
M30	Lab strain	T2294	T409M	5	1	V353E	11	10
M19	Lab strain	T2266	Trunc536c		0	L193F	21	0
M24	Lab strain	T2819	K355M		0	L426F	30	0

^aDeletion of codons 301 to 310 and the first two bases of codon 311, resulting in a frameshift mutation.

^bDeletion of a single C base at codon 218, resulting in a frameshift mutation.

^cDeletion of codons 537 through the end of UL97, resulting in a premature stop codon.

Experiments M27, M28, and M30 were part of a series of six experiments (experiments M25 to M30) already reported in connection with UL97 mutations that evolved rapidly under MBV in error-prone strain T2294, which contained an exonuclease mutation (4). No UL27 mutations were observed through passage 20 in the other three experiments (experiments M25, M26, and M29); the mutants in each of these experiments had evolved multiple UL97 mutations that conferred high-grade MBV resistance before then (4). In each of experiments M27, M28, and M30, the UL27 mutations evolved three to six passages after the appearance of UL97 mutation T409M. In experiment M27, mutants with the UL27 mutation were the only detectable sequence population after three further passages, whereas for experiments M28 and M30, mutants with the UL27 mutation remained an incomplete population after eight or more further passages.

To assess whether a genetic deficiency in UL97 expression was sufficient to elicit UL27 mutations in the absence of MBV exposure, two UL97-knockout constructs were serially propagated. Strains T2266 and T2819 both have the expected UL97-defective cytopathic appearance (4) and evolved UL27 mutations after 21 to 30 passages without added drug. Ten passages later, mutants with UL27 mutations were estimated to be 40 to 50% of the sequence population.

Genotypic analysis of UL27 and UL97 mutations.

The UL27 mutations detected in the nine experiments (Table 1) consisted of five point mutations, two stop mutations, and two frameshift mutations, with no individual mutation detected in more than one experiment; and they differed from all four previously reported UL27 mutations (5, 13). Each of the stop and frameshift mutations interrupted the translation of UL27 at or before codon 362, including one mutation (codon 22 stop) that essentially knocked out the entire UL27 gene. The diversity of UL27 mutations is in contrast to the UL97 mutations selected in vitro under MBV (1, 4, 7), which so far have been shown to involve only point mutations at codons 353, 397, 409, and 411. The two new instances of the UL97 T409M mutation reported here augment its status as the most common mutation selected under MBV under the cell culture conditions used in this study (4).

Recombinant phenotyping of UL27 mutations.

To confirm that the nine UL27 mutations observed affect sensitivity to MBV, they were transferred to strain T2394, creating the mutant recombinant viruses listed in Table 2. Mutations were transferred individually, except in the case of experiment M17, in which seven common baseline amino acid changes in UL27 that occurred in the clinical isolate were transferred, with (strain T2898) and without (strain T2899) the addition of the UL27 mutation W362stop. After plaque purification and sequence verification, many yield-reduction phenotyping assays were performed with each recombinant virus and gave mean MBV EC₅₀s for the mutants with UL27 mutations that were increased two- to threefold over those for control strains T2211 and T2899 (Table 2). This level of MBV resistance is consistent with that previously reported for mutants with UL27 mutations by use of a different yield reduction assay (5). Importantly, point mutations L193F and L426F, which resulted from the propagation of genetically UL97-defective strains without MBV, conferred similar levels of MBV resistance. The low-grade resistance conferred by UL27 mutations is considerably less than that conferred by UL97 mutations (9- to >200-fold), as exemplified by strain T2890 (4, 7). The seven UL27 amino acid changes seen in the baseline clinical isolates (5; unpublished data) and represented in strain T2899 did not confer any MBV resistance.

TABLE 2.

Genotypes and MBV sensitivities of recombinant CMV strains

Strain	Gene	Mutation	Other UL27 changesa	MBV EC50 (μM)b	Fold change in sensitivityc
T2211	UL27	None		0.095 ± 0.002
T2899	UL27	None	N289D, D298G, N300G, P307L	0.084 ± 0.002	0.9
			V310A, D351N, I367V
T2890	UL97	V353A		1.48 ± 0.06	16
T2981	UL27	W153R		0.16 ± 0.01	1.7
T2767	UL27	L193F		0.24 ± 0.02	2.6
T2867	UL27	A269T		0.19 ± 0.01	2
T3053	UL27	V353E		0.20 ± 0.01	2.1
T3037	UL27	L426F		0.21 ± 0.01	2.2
T3020	UL27	E22stop		0.19 ± 0.008	2
T2898	UL27	W362stop	N289D, D298G, N300G, P307L	0.21 ± 0.008	2.2
			V310A, D351N, I367V
T3019	UL27	218delC	Nine changes beyond codon 218d	0.23 ± 0.01	2.5
T2931	UL27	301-311del	Frame shift at codon 311e	0.29 ± 0.02	3.1

^aAmino acid changes compared with the sequence of laboratory strain AD169 (GenBank accession number X17403).

^bMean and standard error of at least 12 determinations performed on at least four separate days.

^cThe MBV EC₅₀ compared with that for baseline strain T2211. The ratio is slightly higher if the EC₅₀ is compared with that for strain T2899.

^dChanges beyond codon 218 are not relevant because of a single C base deletion/frame shift at codon 218.

^eDeletion of codons 301 to 310 and first two bases of codon 311, resulting in a frame shift.

Comparative growth rates of mutants with UL27 mutations.

When the input MOI was calibrated by using the SEAP activities at 24 h postinoculation (6) to ensure the use of comparable viral inocula, the growth of mutants with UL27 mutations was slightly attenuated compared with the growth of wild-type strains T2211 and T2899; but their growth was attenuated much less so than that of UL97-deficient strain T2266, which exhibited the severe growth deficiency characteristic of UL97 genetic knockouts (18). A representative selection of growth curves is shown in Fig. 1. The growth retardation of the mutant strains with UL27 mutations was most notable at day 6, and the accumulated growth eventually approached that of the wild-type strains, unlike the situation with strain T2266. These results are similar to those previously published for mutants with UL27 mutations (5) and determined by a different growth assay. The deletion of large parts of UL27 has much less of an effect on viral growth than the truncation of UL97, and point mutations in either gene that are selected under MBV have only a slight effect on viral growth (5, 7).

FIG. 1.

SEAP growth curves for selected CMV mutant strains with UL27 mutations and controls. Calibrated viral inocula (6) were used to achieve 24-h SEAP readings of 590, 640, 773, 707, 725, and 1,340 for strains T2899, T2211, T2867, T2931, T2898, and T2266, respectively. Cultures of each strain were set up in quadruplicate, and supernatants were sampled at days 1 and 4 to 8. The mean SEAP activities and standard deviations on each day were plotted on the basis of light output from a chemiluminescent substrate (6). The rank order of the growth curves is not explainable by the relative 24-h SEAP readings that reflect the MOI of input virus. wt, wild type.

DISCUSSION

The results presented here indicate that the exposure of CMV to MBV frequently leads to diverse viral UL27 point, stop, and frameshift mutations that confer low-grade MBV resistance. Mutations that interrupt the expression of UL27 and the evolution of similar UL27 mutations in genetically UL97-deficient viruses without MBV exposure suggest that these mutations are an adaptation to the loss of UL97 kinase activity rather than to the possibility that UL27 acts as an independent antiviral target for MBV. The pattern of UL27 mutation is very different from that observed in UL97, in which a limited number of point mutations that confer high-level MBV resistance cluster around the predicted ATP-binding site of the kinase molecule, indicating a probable locus for binding of the drug (3, 4).

There is no understanding of the biological role of UL27 mutations that emerge under conditions of UL97 kinase deficiency. The normal function of the human CMV UL27 gene remains unknown, despite the finding that the corresponding M27 gene in murine CMV appears to be an interferon antagonist (24); a similar function was not observed for UL27 (16). UL27 encodes a 608-residue protein with a putative nuclear localization signal near residue 500 (5) but little recognizable homology to proteins of known function. The observed abundance of UL27 stop and frameshift mutations suggests that the normal nuclear localization and expression of this gene product are mildly detrimental to viral growth in the absence of the UL97 kinase. UL27 possibly acts as a negative kinase modulator that becomes less desirable when the UL97 kinase is absent and cellular kinases must play compensatory roles in viral growth. A clearer understanding of compensatory changes for the loss of the UL97 kinase requires an assessment of the relative importance of its various known and potential substrates, such as the viral UL44 polymerase accessory protein (14, 17), the UL83 tegument protein (12), the cellular retinoblastoma protein (11, 19), and possibly, the UL27 protein on the basis of the sequence context of some if its serine residues (5). A lack of phosphorylation of the UL27 protein in the absence of UL97 kinase activity could also be a factor favoring UL27 mutations. The normal functions of UL27 are not essential in cell culture, since UL27-defective mutants grow relatively well.

The significance of UL27 mutations in the clinical use of MBV also remains uncertain. No CMV isolates from subjects who have received prolonged MBV treatment in clinical trials are yet available for study. By themselves, UL27 mutations probably do not confer enough MBV resistance to preclude therapeutic use of the drug, although such mutations may facilitate the subsequent emergence of UL97 mutations, as in experiments M12 and M20. UL27 mutations also appear to add to existing single UL97 mutations, such as V353A or T409M, which already confer 15- to 80-fold increased MBV resistance (experiments M18, M27, M28, and M30), possibly further increasing the level of drug resistance toward the very high levels (a >150-fold increase in the EC₅₀) seen with certain UL97 mutations and combinations.

This study triples the number of UL27 mutations that have been documented to confer decreased sensitivity to MBV and which need to be included in the databases used for the genotypic analysis of resistance to this anti-CMV drug, which is in advanced-stage clinical trials. Further data collection and recombinant phenotyping are needed, especially with isolates from treated subjects, to refine the genetic basis for the detection of MBV resistance.