Rapid In Vitro Evolution of Human Cytomegalovirus UL56 Mutations That Confer Letermovir Resistance

Abstract

Letermovir (LMV) is an experimental cytomegalovirus terminase inhibitor undergoing phase 3 clinical trials. Viral mutations have been described at UL56 codons 231 to 369 that confer widely variable levels of LMV resistance. In this study, 15 independent experiments propagating an exonuclease mutant viral strain in escalating LMV concentrations replicated 6 of the 7 published UL56 mutations and commonly elicited additional resistance-conferring mutations at UL56 codons 231, 236, 237, 244, 257, 261, 325, and 329. Mutations were first detected earlier in LMV (median, 3 passages) than in 8 parallel experiments with foscarnet (median, 15 passages). As LMV concentrations increased, the typical initial UL56 change F261L, which confers low-grade resistance, combined or was replaced with mutations conferring higher-grade resistance, eventually enabling normal viral growth in 30 μM LMV (>5,000-fold the 50% effective concentration [EC₅₀] for the wild type). At high LMV concentrations, the UL56 changes C325F/R were commonly detected, as well as a combination of changes at codons 236, 257, 329, and/or 369. Recombinant viruses containing individual UL56 mutations and combinations were constructed to confirm their resistance phenotypes and normal growth in cell culture. Several double and triple mutants showed much higher LMV resistance than the respective single mutants, particularly those including changes at both codons 236 and 257. The multiplicity of pathways to high-grade LMV resistance with minimal viral growth impact suggests a low viral genetic barrier and the need for close monitoring during treatment of active infection.

INTRODUCTION

The prevention and treatment of human cytomegalovirus (CMV) infection and disease are an important aspect of the medical care of immunosuppressed individuals. The viral DNA polymerase inhibitors ganciclovir, its oral prodrug valganciclovir, foscarnet, and cidofovir have long been used for this purpose, with generally satisfactory outcomes but also well-known limitations of toxicity, intravenous treatment complexity, and risk of drug resistance after prolonged therapy (1). Cross-resistance among current drugs may develop because they have the same DNA polymerase target (2). Therefore, priority has been given to the development of alternative CMV drug targets. The viral terminase complex, including components encoded by the CMV genes UL56, UL89, and UL51, acting in concert with UL104 and others, is responsible for the cleavage of concatemeric DNA formed during viral replication into unit-length genomes and packaging them into preformed viral capsids (3). This essential process is an attractive target for specific viral inhibition. Drug discovery programs have identified diverse chemical structures that turned out to be CMV terminase inhibitors. Earlier candidates, a benzimidazole pyranoside GW275175X (4) and a chemically unrelated compound, Bay38-4766 (tomeglovir) (5), were tested in phase I trials but were not advanced to later-stage clinical development. Resistance mutations were mapped to the gene UL89 or UL56, consistent with the viral target. More recently, letermovir (LMV; formerly AIC246) advanced to phase III clinical trials for prevention of CMV infection in stem cell transplant recipients after a successful dose-ranging phase II study (6). Published in vitro studies of LMV resistance after exposure to drug concentrations 10-fold higher than the baseline 50% effective concentration (EC₅₀) identified mutations at UL56 codons 231, 236, 241, 325, and 369, distinct from those reported for the older terminase inhibitors (7). These UL56 mutations were found to confer diverse levels of LMV resistance, ranging from a 5-fold increase in EC₅₀ for V231L to >5,000-fold for C325Y.

The objective of the present study was to conduct a series of in vitro selection experiments with increasing LMV concentrations, starting near the baseline EC₅₀, to monitor the evolution and phenotypes of resistance mutations selected under escalating drug concentrations. As with previous work on the experimental CMV antivirals maribavir (8) and cyclopropavir (9), an error-prone UL54 exonuclease domain II (codon 413) mutant was used as the baseline strain to accelerate the evolution of drug resistance mutations by 5 to 10 passages compared with wild-type virus. When this approach was first used for maribavir, the specific mutations that evolved accurately predicted those subsequently observed in treated subjects (10, 11). As an additional comparator for the current experiments, parallel selection experiments were conducted using the standard DNA polymerase inhibitor foscarnet.

MATERIALS AND METHODS

Antiviral compounds.

Letermovir was obtained from MedChemexpress (Princeton, NJ) (catalog no. HY-15233; molecular weight [MW], 572.55) as a chiral product of >99% purity, as documented by nuclear magnetic resonance (NMR), mass spectrometry, and high-pressure liquid chromatography. A 100 mM stock solution was made in dimethyl sulfoxide (DMSO) for further serial dilution into culture media. Foscarnet was obtained as the pharmaceutical 125 mM aqueous solution (Foscavir; Astra).

Viral clones, strains, and cells.

CMV laboratory strain AD169 was modified by the insertion at US3 of a secreted alkaline phosphatase (SEAP) reporter gene driven by the CMV immediate early promoter (12). The resulting strain, T2211, was subsequently cloned as a bacterial artificial chromosome (BAC) BA1 (13). By using established recombineering technology as described previously (14, 15), the conserved UL54 exonuclease domain II residue D413 was deleted in frame, yielding BAC clone BA308, which was transfected into human foreskin fibroblasts (HFF) to generate live recombinant virus (strain T4138) that was slightly attenuated in growth and had a foscarnet-susceptible, ganciclovir-cidofovir-resistant phenotype, as previously noted for the error-prone UL54 D413A mutant (8). The D413 deletion mutant was chosen in the current study to prevent spontaneous reversion of residue 413 to the wild type after prolonged passage. All CMV cultures and antiviral assays were performed in HFF culture monolayers maintained in minimal essential medium (Earle's salts; Gibco), supplemented with 1× Glutamax-I (Gibco) and fetal bovine serum (8% when cultures were subconfluent and 3% when they were confluent).

Viral propagation in letermovir.

The exonuclease mutant T4138 was inoculated into 25-cm² HFF monolayers initially at a multiplicity of infection (MOI) of ∼0.1 and incubated for a week at a time in antiviral drugs at concentrations starting near the EC₅₀ (5 nM for LMV and 50 μM for foscarnet). At the end of the week, if viral growth was effectively inhibited (single enlarged cells, no spreading cytopathic effect [CPE]), the drug concentration was maintained; otherwise, the drug concentration was increased (1.5- to 4-fold depending on the extent of spreading CPE), and the culture was propagated by trypsinization and passage of 30% of the cell suspension to a new subconfluent HFF monolayer. This was continued for up to 30 passages and/or a maximum LMV concentration of 30 μM (threshold for cytotoxicity). Every few passages, as prompted by the emergence of uninhibited CPE in the drug, DNA extracts of infected cell suspension were prepared, amplified by PCR, and sequenced using conventional dideoxy sequencing (BigDye v3.1; Applied Biosystems) to screen for mutations at UL56 codons 1 to 500. At the final passage, DNA extracts were sequenced in the entire CMV coding sequences of UL51 (158 codons), UL56 (851 codons), and UL89 (675 codons), which represent the three well-characterized components of the CMV terminase complex (16, 17). The control foscarnet selection experiments were monitored for mutation in the UL54 sequence (codons 94 to 1000 for intermediate passages and the full 1,243 codons at the final passage). For all mutations detected at any stage, additional stored infected-cell samples from intermediate passages were extracted and sequenced to determine the timing of earliest emergence of the respective mutations, tracing back to the time point where the mutation was not detected.

Recombinant-virus construction.

Novel mutations and selected combinations were phenotyped by transferring them to a baseline BAC-cloned CMV strain and testing the derived recombinant live virus strain for LMV susceptibility using a standardized SEAP reporter yield reduction assay (12, 13). Similar to other BAC clones of CMV strain AD169 (18, 19), clone BA1 was discovered to have a compensatory deletion of US7 to US16 dating back to strain T2211 (12) prior to cloning. Therefore, a new baseline BAC clone, BD1, was derived from BA1, by markerless “en passant” (20) replacement of the missing US7 to US16 coding sequence that had been PCR amplified from strain AD169 (ATCC VR-538) and deletion of the redundant IRL1 to IRL13 sequences to enable the viral genome to accommodate the BAC vector and SEAP expression cassette. To reduce the possibility of residual wild-type sequences in recombinant virus progeny, a derivative of BD1 with the entire UL56 sequence deleted and replaced with an ampicillin-selectable marker (bLac) was constructed as clone BD2. This clone was used as the base for introduction of UL56 mutations by means of a transfer vector containing a Frt-Kan selectable marker, using the strategy used previously to construct large series of UL97 and UL54 mutants (13, 14). In this case, the Frt-Kan marker was introduced at the natural Bsu36I restriction site upstream of the UL56 coding sequence and incorporated into a transfer vector consisting of a plasmid clone of the EcoRI fragment of the AD169 genome that includes UL56. The residual 34-bp Frt sequence did not impair viral growth, and the baseline virus incorporating this motif was used as the reference wild-type strain for phenotyping. The desired UL56 mutations were first introduced into the transfer vector between the BamHI and BsrGI unique sites, using a suitable PCR product. Sequence-verified transfer vectors were then digested with EcoRI, and the mutant UL56 sequence was recombined into the BAC clone BD2 in the heat-induced Escherichia coli host SW105 as described previously (13, 14). The resulting kanamycin-resistant, ampicillin-sensitive colonies were checked for the expected BAC EcoRI digest pattern. The Kan selection marker was removed from a qualifying clone by arabinose-induced Flp recombinase activity in the SW105 host. The final recombinant BAC was sequenced throughout UL56 to confirm the intended mutation(s) and transfected into HFF monolayers using the transfection reagent X-tremeGENE-HP (Roche) according to the manufacturer's instructions. The resulting live virus was propagated at a low multiplicity of infection to make cell-free virus stock for phenotypic assays.

Drug susceptibility phenotyping was performed by a SEAP reporter-based yield reduction assay as used for previous studies of various CMV antiviral compounds (8, 9, 12–14). A SEAP-calibrated low-multiplicity viral inoculum (12) was inoculated into 6 wells (one row) of 24-well cluster plates containing fully confluent HFF monolayers and then incubated for 1 week with no drug and a series of 5 drug dilutions of 2-fold-increasing concentrations covering the anticipated EC₅₀ endpoint. At 1 week, samples of supernatant medium were collected for SEAP activity assays, measured as relative light units (RLU) using a chemiluminescent substrate, to determine the drug concentration associated with a reduction of SEAP activity to 50% of its baseline value in the absence of drug (EC₅₀). At least 7 assays were performed for each mutant over at least 4 separate setup dates (to allow for variation in cell culture conditions), with simultaneous controls consisting of baseline and known UL56 mutant resistant strains. Results were reported as the mean and standard deviation EC₅₀s along with the number of replicates.

Growth curves of baseline and mutant viral strains were compared as previously described (14, 21). Viral inocula with an MOI of ∼0.01 were calibrated based on comparable culture supernatant SEAP activity at 24 h. Culture media were then sampled daily at days 4 through 8, frozen, and subsequently simultaneously assayed for SEAP activity. Growth curves were plotted as mean and standard deviation RLU values based on 4 replicates.

RESULTS

Overview of mutations selected under letermovir.

Table 1 lists the observed UL56 amino acid substitutions (all resulting from single nucleotide mutations), their relative frequencies, and the timing of detection. At baseline, there were no detectable UL56 sequence differences from the reference strain AD169. Each of the 15 in vitro selection experiments evolved 2 to 8 (median, 4) UL56 mutations during serial passage as the LMV concentration was escalated from the wild-type EC₅₀ of 5 nM toward the 30 μM range. The detected mutations included 6 of the 7 previously published ones (substitutions V231L, V236M, L241P, and R369G/M/S) as well as 14 others, including the 3 most commonly detected ones (F261L, C325F, and L257I). All of the newly recognized mutations are in the UL56 codon range 231 to 329, except for a single instance of L51M, which evolved in the same time frame and subpopulation fraction as L257I and was suspected to be coincidental and not necessarily resistance related. Sequencing of the complete coding regions for the UL51 and UL89 components of the terminase complex revealed no mutations relative to strain AD169 at the end of each of the 15 selection experiments. Eight parallel selection experiments performed using foscarnet for 25 to 30 passages eventually resulted in the selection of 1 to 3 (median, 1.5) UL54 pol mutations per experiment.

TABLE 1.

UL56 mutations detected in vitro in letermovir

Amino acid substitutiona	No. of exptsb	Passagec	[LMV]d
L51M	1	5	8
V231A	1	17	240
V231L	5	14	320
V236L	4	8	160
V236M	5	13	200
E237D	1	13	80
L241P	1	10	50
T244K	2	8	34
T244R	1	15	40
L257I	7	8	20
F261C	2	26	30 μM
F261L	15	3	8
F261S	1	20	2 μM
Y321C	1	13	200
C325F	9	15	700
C325R	4	20	20 μM
M329T	2	21	20 μM
R369G	3	22	30 μM
R369M	2	3	10
R369S	4	20	5.5 μM

^aUnderlined amino acid changes are newly recognized.

^bNumber of independent selection experiments where the change was detected.

^cMedian first passage of detection of mutation.

^dMedian LMV concentration at first detection of mutation (in nanomolar units unless otherwise specified).

Timing and order of appearance of UL56 mutations.

Despite the insensitivity of standard dideoxy sequencing in detecting mutant subpopulations of <20% (22), mutation was detected in all 15 experiments within 5 passages in LMV (range, 2 to 5; median, 3), and at a median drug concentration of 8 nM (1.4-fold EC₅₀). In the 8 selection experiments with foscarnet, UL54 pol mutations were first detected much later, at passages 11 to 20 (median, 15) and at a median drug concentration of 650 μM (13-fold EC₅₀). There were differences in the relative timing of specific mutations in LMV (Table 1). For example, F261L was detected early in every experiment, while V236L and L257I were common at intermediate passages. Substitutions such as C325F, C325R, M329T, and R369G/S were usually observed at higher LMV concentrations, although C325F appeared as early as passage 3 and by passage 10 in 4 of 15 experiments, at LMV concentrations of <1 μM.

Sequential changes in viral genotype under escalating LMV concentrations for the individual experiments are detailed in Table 2, categorized by the range of drug concentrations at which they were detected. Despite the evolutionary complexity, some common patterns can be discerned. At low passages and LMV concentrations, F261L was dominant. At intermediate passages, this mutant was often overgrown by others, notably strains with the mutations V236L/M and/or L257I. Changes at both codons 236 and 257 occurred often (6 experiments), and this genotype was sustainable through a range of moderately high LMV concentrations. Finally, as LMV concentration increased into the highest (>5 μM) range, the most common pattern was for C325F/R to emerge (12 experiments) and displace other pre-existing mutants. The previously reported C325Y mutant (7) was not detected. In the minority of experiments where C325F/R did not evolve, combinations of multiple mutations such as V236L/M, L257I, M329T, and R369S also appeared to enable viral growth under high LMV concentrations. The passage number where normal-appearing viral growth occurred at LMV concentrations of >100-fold EC₅₀ ranged from 3 to 23 (median, 13).

TABLE 2.

Evolution of UL56 mutations in vitro in letermovir

Expt	No.a	UL56 mutations detected in letermovir concnb
		5–50 nM	50 nM–1 μM	1 μM–5 μM	>5 μM
M137	8	L51M (5), T244R (15), L257I (5), F261L (5)	L51M, V231A (17), T244R, L257I	L51M, V236L (20), L257I	L51M, V231A, V236L, L257I, F261C (29), R369S (25)
M138	6	T244K (5), F261L (5)	E237D (13), T244K, F261L	E237D, T244K, F261L	V231L (28), E237D, T244K, F261L, C325R (25), R369S (25)
M142	5	V231L (5), L257I (5), F261L (3)	V231L, V236L (8), L257I, F261L		V231L, V236L, L257I, C325R (13)
M143	3	F261L (2)	V231L (8), C325F (8)	C325F
M144	2	F261L (2), C325F (3)	C325F
M145	7	F261L (3)	V236L (8), V236M (8)	V236L, L257I (14), F261L, R369S (14)	V236L, F261L, C325F (15), C325R (17), R369S
M146	4	F261L (3), V236L (8)	V236L, F261L	V236L, L257I (13)	V236L, L257I, C325F (19)
M151	5	L257I (8), F261L (5)	V236M (13), L257I, F261L, Y321C (13)	V236M, L257I	V236M, L257I, F261L, M329T (19)
M152	6	F261L (3)	V231L (14), V236M (15), F261L	V231L, V236M, F261L, F261S (20)	V231L, V236M, F261L, F261S, C325F (23), R369G (25)
M162	7	V236M (3), F261L (3), R369M (3)	V236M	V236M, L257I (15), F261L	V236M, L257I, F261C, C325R (22), M329T (22), R369G (22)
M163	4	F261L (3), R369M (3)	L241P (10), R369M	V231L (16), L241P	V231L, L241P, C325F (20)
M164	3	F261L (3), L257I (5)	R369S (10)	L257I, R369S	V236M (18), R369S
M166	4	F261L (3)	T244K (10), F261L, C325F (15), R369G (10)		T244K, F261L, C325F
M169	2	F261L (3)	C325F (10)		C325F
M170	2	F261L (3)	C325F (10)		C325F

^aNumber of distinct mutations detected over the duration of the experiment.

^bNumbers in parentheses indicate first passage of detection of mutation (usually as a subpopulation). Underlined mutations were detected as completely mutant populations within the indicated drug concentration range. Boldface mutations (codon 325) or combinations (V236M/L + L257I) are associated with growth under drug concentrations of >1 μM.

Recombinant phenotyping of single and multiple UL56 mutations.

Of the 7 previously reported UL56 mutations that confer LMV resistance, 3 were chosen as controls for the construction of BAC-cloned mutant CMV strains to compare the measured levels of LMV resistance to published data. The substitutions V231L, L241P, and C325Y represent those reported as conferring low (∼5-fold), moderately high (∼200-fold), and very high (>5,000-fold) levels of LMV resistance, respectively. LMV EC₅₀s are shown in Table 3 and were determined using a standardized reporter-based yield reduction assay. Published data derived using a different phenotyping assay reported a comparable baseline LMV EC₅₀ of 3 to 5 nM and EC₅₀ ratios of 5, 160 to 218, and >8,000 for V231L, L241P, and C325Y, respectively (7, 23, 24). EC₅₀ ratios of >3,000 can be interpreted as absolute LMV resistance, because viral yield reduction occurs at visibly cytotoxic LMV concentrations.

TABLE 3.

Recombinant phenotyping data for newly characterized UL56 mutations

Category and amino acid substitution(s)	Recombinant virus strain	Letermovir EC50a	SD	nb	Fold changec
Controls
None (wt)	T4190	5.7 nM	1.8	43
V231L	T4194	29 nM	6.9	30	5.1
L241P	T4192	0.55 μM	0.20	15	96
C325Y	T4206	20 μM	2.0	8	>3,000
Conferring <4-fold change in LMV EC50
L51M	T4195	4.3 nM	1.2	7	0.8
V231A	T4248	12 nM	2.9	9	2.1
T244K	T4213	19 nM	4.5	21	3.3
F261L	T4218	16 nM	4.3	15	2.8
Conferring 4- to 15-fold change in LMV EC50
V236L	T4185	80 nM	12	13	14
E237D	T4222	58 nM	13	8	10
T244K F261L	T4219	47 nM	16	8	8.2
L257I	T4214	28 nM	11	22	4.9
F261C	T4245	25 nM	5.4	9	4.4
Y321C	T4251	26 nM	4.1	8	4.6
M329T	T4271	25 nM	6.4	14	4.4
Conferring 100- to 1,000-fold change in LMV EC50
E237D T244K F261L	T4217	0.59 μM	0.07	8	104
V236L L257I	T4237	1.5 μM	0.14	9	260
Conferring >1,000-fold change in LMV EC50
V236M L257I M329T	T4270	18 μM	2.3	13	>3000
C325F	T4189	21 μM	3.5	9	>3000
C325R	T4257	20 μM	1.9	11	>3000

^aMean LMV concentration reducing viral SEAP reporter growth by 50%.

^bNumber of replicates (done on at least 4 separate dates).

^cCompared with the wild-type control value.

All of the individual amino acid substitutions listed as newly recognized in Table 1, except T244R and F261S (single instances considered similar to T244K and F261L/C), were transferred into recombinant BAC clones, and the LMV susceptibility phenotype was determined for the derived live CMV strains, as detailed in Table 3. The 12 strains containing single UL56 point mutations were noted to have widely divergent LMV susceptibilities. As surmised, L51M was found to confer no LMV resistance, and its appearance was the presumed result of a random mutation that was coselected with L257I as the latter emerged to confer a resistant phenotype. The low-level LMV resistance conferred by F261L is consistent with the timing of its appearance in the selection experiments. A variety of single UL56 substitutions conferred a low to medium level of resistance (4- to 14-fold increased EC₅₀), while C325F and C325R conferred the same absolute LMV resistance as C325Y. A set of double and triple mutants, selected from those observed in the individual selection experiments, was constructed and tested for LMV resistance. The results (Table 3) show a strong multiplier effect of double and triple mutations. For example, the combination of F261L and T244K conferred at least twice the LMV resistance (8.2-fold) of either change alone, and adding a third change (E237D) further increased the resistance to >100-fold, thus explaining the accumulation of mutations as LMV concentrations were escalated. The frequent emergence of the V236L-L257I combination at higher LMV concentrations is also well explained by the strongly augmented level of resistance (>250-fold EC₅₀ increase) conferred by the double mutant, whereas the single mutants confer only a 5- to 14-fold-increased EC₅₀. The triple combination of mutations at codons 236, 257, and 329 as observed in experiments M151 and M162 conferred absolute LMV resistance comparable to that of the C325 mutants.

Normal growth in cell culture of UL56 mutants conferring LMV resistance.

Fig. 1 shows the comparative growth curves of representative UL56 mutants exhibiting low, medium, or high levels of resistance. Any difference in fitness of the mutant viruses compared to baseline strains is not discernible among the growth curves. This is consistent with a previous report (7).

FIG 1.

SEAP growth curves of recombinant CMV strains. Cell-free stocks of wild type (wt) and UL56 mutant virus stocks were inoculated onto human foreskin fibroblast monolayers at an MOI of ∼0.01. Culture media were assayed daily at 4 to 8 days postinoculation for SEAP activity using a chemiluminescent substrate. Each point on the growth curves is the mean and standard deviation from 4 replicates. SEAP, secreted alkaline phosphatase; RLU, relative light units; wt, control wild type with an Frt motif upstream of UL56; K355del, in-frame deletion of catalytic lysine residue resulting in loss of UL97 kinase activity and severe growth impairment, for comparison with previously published growth curves (14).

DISCUSSION

The present study recapitulated almost all of the published LMV resistance mutations in the CMV UL56 component of the terminase complex and identified many others for the first time in the codon range 231 to 369 which confer various levels of LMV resistance when present singly or in combination, as confirmed by recombinant phenotyping. LMV resistance mutations were selected rapidly at low drug concentrations relative to baseline EC₅₀. Multiple mutations evolved to sustain normal viral growth in escalating drug concentrations, eventually conferring absolute resistance to LMV from single mutations C325F/R, or less commonly from combinations of mutations.

The rate of evolution of mutations in LMV can be compared with that in other CMV antiviral compounds tested using the same in vitro selection strategy (8). The first appearance of the UL56 mutation F261L at a median of 3 passages is somewhat earlier than reported for the first mutations emerging under the UL97 inhibitor maribavir (median, 5 passages) (8) and notably earlier than with the nucleoside analog polymerase inhibitor cyclopropavir (median, 7 passages) (9) and in this study with the polymerase inhibitor foscarnet (median, 15 passages). These differences in timing likely result from the relative growth fitness of the elicited drug-resistant mutants, as foscarnet-resistant mutants have been consistently growth attenuated (14, 21, 25). Together with UL54 pol mutations reported in clinical specimens that have a nonviable phenotype (26), this suggests a higher genetic barrier to the development of foscarnet resistance. In contrast, critical UL56 residues involved in LMV binding to the terminase complex do not appear at all important for biological activity, even though LMV is remarkably potent at disrupting wild-type terminase function (24). Although published data (7) and data obtained here (Fig. 1) show no growth impairment of highly LMV-resistant mutants, a pattern of initial selection of low-resistance mutants, such as the F261L mutant, and progressive replacement by higher-resistance UL56 mutants suggests that there may be subtle differences in growth fitness that are not detectable in growth curve assays.

The variety of mutations detected per selection experiment in LMV as drug concentrations were escalated suggests many alternative genetic pathways for achieving a level of resistance matching any level of drug exposure. Selection experiments under other drugs as described above rarely select for more than 3 distinct mutations per experiment, but the median number of mutations in LMV was 4, and the number ranged up to 8 (Table 2). A point mutation at residue C325 appears to have the unique role of conferring absolute LMV resistance with preserved fitness, requiring no further viral genetic adaptation or diversification. More UL56 mutations evolved in the absence of C325 changes, but all remained in the codon range 231 to 369 as earlier noted (7), with some combinations of changes conferring high-level or absolute resistance. No sequence changes were noted in other terminase component genes UL51 and UL89, indicating that these are not preferential loci for LMV resistance. However, we are monitoring additional LMV selection experiments that did not evolve C325 changes to see if the diversity of selected variants occasionally includes changes in UL51, UL89, and UL104.

Recombinant phenotyping data confirmed that various levels of LMV resistance were conferred by the newly recognized UL56 mutations (Table 3), ranging from low for F261L to absolute for C325F/R, but commonly in the 4- to 15-fold range for single mutations. Transfers of combinations of 2 or 3 mutations at codons as observed in specific selection experiments showed a strong multiplier effect on the level of LMV resistance, which was notable with combinations of mutations at codons 236, 257, and 329. Table 2 suggests that although they were not specifically phenotyped, combinations of R369G/S and other mutations in the codon range 236 to 257 also have this effect. Moderate resistance mutations in the same gene combining to confer very high resistance to the same drug are not a feature of standard CMV polymerase inhibitors.

Clinical correlation is pending to assess the in vitro impression of a low genetic barrier to the development of LMV resistance. Available in vitro data predict that drug resistance may emerge during ongoing CMV replication more quickly with maribavir, and even more so with LMV, than with traditional polymerase inhibitors. Limited information is available at present from clinical trials and treatment studies. In the failed phase III maribavir prophylaxis trials, genotypic drug resistance was not detected (27), but the study drug was stopped soon after detection of breakthrough viral infection. Two case reports exist of the relatively early onset of maribavir resistance after attempted salvage therapy (10, 11), along with 5 additional recent unpublished cases among 35 treated subjects (28), all involving the same mutations at UL97 codons 409 and 411 originally observed in vitro. For LMV, one case of breakthrough infection during the phase II prophylaxis trial was recently reported to develop UL56 V236M (29) at 7 weeks, a resistance mutation which was previously detected in vitro (7) and also is well represented in Table 1.

Translation of in vitro drug resistance data into clinical practice involves factors such as interstrain sequence polymorphism, the cell culture assays for EC₅₀s, and the effective intracellular antiviral concentrations achieved after therapeutic doses. For LMV, the UL56 codons 231 to 369, which include the most common resistance mutations, are distinct from the codon ranges (e.g., 425 to 476) where sequence polymorphisms are prevalent (30), and rare polymorphisms noted at codons 242 and 327 actually confer reduced LMV EC₅₀s (31). Variation in cell culture conditions resulted in EC₅₀s for maribavir that differed 100-fold for the same viral strain (32), but this was tested for LMV without finding a significant effect of cell type or viral inoculum (24). Comparison of in vitro EC₅₀s with therapeutic LMV concentrations is limited by the paucity of published pharmacokinetic data (33). Better information is needed relevant to the higher LMV doses used in the current phase III trials to determine if the drug levels achieved in vivo are sufficient to suppress the replication of some of the resistant mutants in Table 3.

In a prophylaxis setting, it is plausible that with the nanomolar in vitro antiviral potency of LMV, adequate doses may maintain complete viral suppression (29) without a high incidence of breakthrough infection and consequent drug resistance. It is more probable that the treatment of high-grade viremia or CMV end-organ disease will result in the rapid selection of LMV resistance mutations. Genotypic resistance testing may be needed sooner than is typically recommended for the standard polymerase inhibitors (1). Use of suitable antiviral drug combinations may be warranted to decrease the likelihood of LMV resistance in therapeutic use.