Comparison of Cytomegalovirus Terminase Gene Mutations Selected after Exposure to Three Distinct Inhibitor Compounds

ABSTRACT

Letermovir, GW275175X (a benzimidazole), and tomeglovir (Bay38-4766) are chemically unrelated human cytomegalovirus (CMV) terminase complex inhibitors that have been tested in human subjects. UL56 gene mutations are the dominant pathway of letermovir resistance, while UL89 and UL56 mutations are known to confer benzimidazole resistance. This study compares the mutations elicited by the three inhibitors during in vitro CMV propagation. GW275175X consistently selected for UL89 D344E and sometimes selected for UL89 C347S, UL89 R351H, or UL56 Q204R. Tomeglovir consistently selected for UL89 V362M and sometimes selected for UL89 N329S, T350M, H389N, or N405D or UL56 L208M, E407D, H637Q, or V639M. Adding to known and novel UL56 mutations, letermovir occasionally selected for UL89 N320H, D344E, or M359I. Recombinant phenotyping confirmed that UL89 D344E conferred 9-fold resistance (an increased 50% effective concentration [EC₅₀]) for GW275175X and increased the letermovir and tomeglovir EC₅₀s by 1.7- to 2.1-fold for the baseline virus and the UL56 Q204R, E237D, F261L, and M329T mutants. UL89 N320H and M359I conferred <2-fold letermovir resistance but 7-fold tomeglovir resistance; the N320H mutant was also 4-fold resistant to GW275175X. UL89 N329S conferred tomeglovir and letermovir cross-resistance. UL89 T350M conferred resistance to all three inhibitors. UL89 C347S conferred 27-fold GW275175X resistance. UL89 V362M and H389N conferred 98-fold and 29-fold tomeglovir resistance, respectively, without conferring cross-resistance. Thus, characteristic UL89 mutations confer substantial resistance to GW275175X and tomeglovir and are an uncommon accessory pathway of letermovir resistance. Instances of moderate cross-resistance and the proximity of the selected UL89 and UL56 mutations suggest targeting of a similar terminase functional locus involving UL56 and UL89 interaction.

INTRODUCTION

Human cytomegalovirus (CMV) is an important viral pathogen in the setting of T-cell dysfunction, like that which occurs after solid organ and hematopoietic stem cell transplantation (1). Antiviral strategies for prevention and prompt treatment of end-organ disease have effectively reduced CMV-related morbidity and mortality. All currently licensed anti-CMV drugs (ganciclovir, valganciclovir, foscarnet, and cidofovir) ultimately target the viral UL54 DNA polymerase and have well-known limitations of toxicity, drug resistance, and cross-resistance. Alternative CMV drug targets have been explored for many years. The viral terminase complex, encoded by the UL56, UL89, and UL51 genes, acting in concert with the portal protein (encoded by the UL104 gene), among others, performs the essential function of cleaving and packaging unit-length viral genomes into the viral capsid after DNA replication on a rolling-circle template (2). This terminase drug target has featured in several drug discovery programs, with three chemically unrelated inhibitors having been tested in human subjects: GW275175X (3), tomeglovir (BAY 38-4766) (4), and letermovir (5). While the first two were not developed beyond phase I (3, 6), letermovir was recently successful in a phase III trial of its use for CMV prophylaxis in stem cell transplant recipients (ClinicalTrials.gov registration no. NCT02137772), making it more relevant to gain a better understanding of the genetic mechanisms of resistance to drugs of this class and the growth fitness of resistant mutants, which influences the genetic barrier to drug resistance.

Resistance mutations are expected to involve the UL56 and/or the UL89 gene. These genes encode principal components of the terminase complex. Limited data exist for the older compounds, GW275175X and tomeglovir. For example, amino acid substitutions in UL56 (Q204R) and UL89 (D344E) were selected in vitro and shown to confer resistance to the benzimidazole d-ribosides 2,5,6-trichloro-1-beta-d-ribofuranosyl benzimidazole (TCRB) and its 2-bromo analog (BDCRB) (7). The level of resistance of the UL89 D344E mutant was augmented by the additional substitution UL89 A355T (8), and the double mutant showed high-level cross-resistance to the related benzimidazole pyranoside GW275175X (3). Viral culture propagation under tomeglovir selection elicited UL56 and UL89 mutations (4), but the observed amino acid substitutions, UL56 A662V and UL89 N353S and Y386H, were not individually phenotyped to show the level of drug resistance conferred. More recently, numerous UL56 amino acid changes in the codon 231 to 369 range emerged in vitro under letermovir selection to confer widely differing levels of resistance ranging from low-grade to absolute resistance, while nearly normal growth fitness was retained (9, 10). Examples include UL56 V236M, which confers ∼45-fold resistance and which was subsequently observed in a letermovir clinical trial (11); C325F/R/Y, which confer absolute resistance; and R369M, which confers moderate letermovir resistance with low-grade tomeglovir cross-resistance (9, 10).

In order to compare the mutations elicited by each terminase inhibitor and the degree of associated cross-resistance, a CMV exonuclease mutant strain was serially propagated under drug selection and emergent mutations were individually characterized by recombinant phenotyping. As in prior work (9), the error-prone exonuclease mutant was used to accelerate the evolution of resistance mutations and their discovery.

(This study was presented in part at the ASM Microbe meeting, Boston MA, June 2016.)

RESULTS

Mutations selected under letermovir exposure.

In 2 selection experiments unrelated to those previously published (9), novel amino acid substitutions UL56 N232Y and K258E were selected once each (Table 1). Previously characterized UL56 substitutions V236M, E237D, L241P, F261L, and M329T (9, 10) were also encountered. For the first time in my experience, UL89 mutations were observed under letermovir selection, with the substitutions N320H, D344E, and M359I each appearing once (Table 2). The evolution of mutations in the individual experiments is shown in Fig. 1. In experiment M141, UL89 substitution M359I added to preexisting UL56 substitution L241P by passage 13, followed by the appearance of an additional UL89 substitution, N320H, a few passages later. Sequencing of 20 PCR product clones from viral DNA at passage 26 showed 11 instances of N320H, 7 instances of M359I, and 2 clones containing neither, indicating that the two UL89 mutations occurred on separate genomes. A novel UL56 substitution (K258E) also emerged in late passages. In experiment M165, the UL89 substitution D344E (7) appeared at a late passage to add to several preexisting UL56 mutations. In both experiments M141 and M165, mutated virus was eventually able to grow well in the presence of 30 μM letermovir, or about 8,000 times the 50% effective concentration (EC₅₀) for the wild type, interestingly, without evolving a mutation at codon C325, which is the usual genetic pathway to absolute letermovir resistance (9).

TABLE 1.

Summary of UL56 mutations selected in vitro

Amino acid substitutiona	GW275175X (nb = 6)			Tomeglovir (n = 8)			Letermovir (n = 2)
	No. of exptsc	Passage no.d	Lowest concn (μM)e	No. of expts	Passage no.	Lowest concn (μM)	No. of expts	Passage no.	Lowest concn (μM)
Q204R	1	11	32
L208M				3	13	8
N232Y							1	15	0.2
K258E							1	19	8
E407D				5	4	4
H637Q				1	5	6
V639M				2	11	6
L764M				1	17	60
E769D	1	10	16

^aExcludes previously published V236M, E237D, L241P, F261L, M329T, and R369M mutations (9, 10).

^bn, total number of experiments performed.

^cNumber of experiments in which the mutation was detected.

^dMedian passage number at which the mutation was first detected.

^eLowest drug concentration at which the mutation was first detected.

TABLE 2.

Summary of UL89 mutations selected in vitro

Amino acid substitutiona	GW275175X (nb = 6)			Tomeglovir (n = 8)			Letermovir (n = 2)
	No. of exptsc	Passage no.d	Lowest concn (μM)e	No. of expts	Passage no.	Lowest concn (μM)	No. of expts	Passage no.	Lowest concn (μM)
N320H							1	19	8
N329S				1	10	4
I334V				1	5	6
D344E	6	7	4				1	24	30
C347S	4	15	30
T350M				1	17	30
R351H	1	18	30
M359I							1	13	0.8
V362M				8	12	8
H389N				1	8	8
N405D				3	5	4

^aExcludes Q11H, which was coselected with V362M in experiment M200 (Fig. 1).

^bn, total number of experiments performed.

^cNumber of experiments in which the mutation was detected.

^dMedian passage number at which the mutation was first detected.

^eLowest drug concentration at which the mutation was first detected.

FIG 1.

Timeline of evolution of mutations during serial passage in selected experiments. The evolving amino acid substitutions observed during serial culture passage under increasing drug concentrations are listed for each experiment. Novel substitutions are indicated in bold; others have been published elsewhere (7, 9, 10). The numerical suffix after the decimal point indicates the estimated mutant subpopulation (in tenths). No suffix indicates that the full population had the mutation. Drug concentrations are listed on the top line of each experiment. Abbreviations: LMV, letermovir; 175X, GW275175X; TGV, tomeglovir.

Mutations selected under GW275175X exposure.

All 6 selection experiments consistently showed the UL89 substitution D344E at passages 5 to 10 as the first recognized change. This is the characteristic benzimidazole d-riboside resistance mutation, as previously identified (7). Additional changes observed at higher GW275175X concentrations included UL89 C347S in 4 cases and UL89 R351H in 1 case (both of these are novel substitutions) (Table 2), UL56 Q204R in 1 case, and UL56 E769D in another case (Table 1). Figure 1 shows the evolution of mutations in three GW275175X selection experiments, including multiple mutations that enabled viral growth in the presence of higher drug concentrations, such as UL89 D344E with either C347S or R351H. A previously described combination of UL89 D344E and UL56 Q204R (7) was also observed.

Mutations selected under tomeglovir exposure.

Eight experiments all ultimately selected for UL89 V362M as the signature mutation for tomeglovir resistance. Various other UL56 and UL89 substitutions were also observed beginning at passages 4 to 8, including UL56 L208M, E407D, H637Q, V639M, and L764M (Table 1) and UL89 N329S, I334V, T350M, H389N, and N405D (Table 2). The evolution of mutations under tomeglovir selection is illustrated in Fig. 1. The UL56 substitution E407D was a typical early change at passage 4 or 5 (in 4 of 8 experiments); E407D and other emergent UL89 mutations were eventually displaced by V362M, suggesting that V362M confers a higher level of tomeglovir resistance than those displaced. In experiment M200 (Fig. 1), UL89 Q11H coevolved with V362M in the same proportion and timing; because of the timing and atypical locus of this mutation, it was felt more likely to be a random variant coselected in an error-prone exonuclease mutant strain or possibly a growth-adaptive mutation. The UL56 mutations that evolved under tomeglovir selection did not overlap those known to confer letermovir resistance.

Recombinant phenotyping.

An extensive set of UL56 and UL89 variants was phenotyped after construction of recombinant bacterial artificial chromosome (BAC) clones containing single and double mutations. After reconstitution of live virus, secreted alkaline phosphatase (SEAP) yield reduction susceptibility assays were performed, and the results are shown in Table 3. All observed fold changes in EC₅₀s of 1.5 or greater for individual mutant strains were statistically significant, as reflected by P values of <0.0025 in Table 3 and the nonoverlap of the 95% confidence intervals of the EC₅₀ values. The UL89 substitution D344E was confirmed to confer 9.4-fold resistance to GW275175X, a level similar to that reported for other benzimidazole derivatives (7, 8). D344E is newly shown to decrease susceptibility to tomeglovir and letermovir, consistent with its emergence in one letermovir selection experiment (Fig. 1, experiment M165). Although the 1.7- and 1.8-fold increases in the tomeglovir and letermovir EC₅₀s, respectively, are minor, the addition of D344E to the UL56 mutations, such as the E237D, F261L, and M329T mutations, selected in the same experiment increased the EC₅₀ values of the single UL56 mutants by close to 2-fold (Table 3). When D344E was combined with UL56 Q204R, the EC₅₀s of all 3 inhibitors were also increased (Table 3), as was reported historically for other benzimidazoles (7). Another UL89 substitution, C347S, conferred a higher 27-fold increased EC₅₀ for GW275175X, while R351H conferred a level of resistance similar to that conferred by D344E.

TABLE 3.

Genotypes and susceptibility phenotypes of recombinant CMV strains

Strain	Genotype	GW275175X				Letermovir				Tomeglovir
		EC50a	SDb	No. of replicatesc	Fold changed	EC50	SD	No. of replicates	Fold change	EC50	SD	No. of replicates	Fold change
4190	UL56 wild type	1.1	0.28	107		3.8	0.86	100		0.44	0.07	123
4272	UL89 wild type	1.1	0.28	112		3.8	0.88	90		0.47	0.07	142
Selected under GW275175X exposure
4274	UL89 D344E	10	2.4	113	9.4	7.0	1.1	20	1.8	0.81	0.16	47	1.7
4283	UL89 C347S	30	8.1	12	27	2.4	0.43	10	0.6	0.16	0.05	9	0.3
4339	UL89 R351H	9.3	1.2	10	8.5	2.6	0.33	11	0.7	0.17	0.03	10	0.4
4372	UL56 E769D	0.72	0.17	10	0.7	3.9	1.2	14	1.0	0.33	0.03	8	0.8
4349	UL56 Q204R	4.6	0.69	8	4.2	5.8	1.3	10	1.5	1.2	0.16	13	2.7
4357	UL56 Q204R UL89 D344E	22	2.8	8	21	9.4	2.3	9	2.5	2.5	0.22	10	5.8
Selected under letermovir exposure
4263	UL56 N232Y	0.82	0.22	12	0.8	64	18	12	17	1.2	0.06	8	2.7
4249	UL56 K258E	0.72	0.15	8	0.7	52	16	12	14	0.14	0.01	8	0.3
4218	UL56 F261L	0.74	0.25	15	0.7	9.6	1.0	9	2.5	0.44	0.07	13	1.0
4313	UL56 F261L UL89 D344E	10	1.8	7	10	18	3.3	11	4.8	0.91	0.25	13	2.1
4271	UL56 M329T	0.90	0.28	12	0.8	17	5.7	12	4.5	0.46	0.11	10	1.1
4336	UL56 M329T UL89 D344E	11	2.7	8	10	30	6.1	10	8.0	0.98	0.12	10	2.2
4222	UL56 E237D	0.57	0.12	10	0.5	61	6.0	8	16	0.14	0.02	10	0.3
4314	UL56 E237D UL89 D344E	8.0	1.8	9	7.4	129	27	8	34	0.27	0.06	9	0.6
4284	UL89 N320H	4.6	0.97	10	4.2	7.0	0.44	7	1.8	3.0	0.24	11	6.5
4341	UL89 M359I	0.60	0.11	8	0.5	5.6	1.6	16	1.5	3.4	0.92	14	7.4
Selected under tomeglovir exposure
4368	UL89 I334V	1.0	0.26	9	0.9	4.0	0.82	10	1.1	0.41	0.10	10	0.9
4367	UL89 N329S	0.75	0.20	8	0.7	7.7	2.2	13	2.0	6.9	0.63	11	15
4334	UL89 T350M	5.6	1.2	11	5.1	11	2.5	10	2.8	4.1	0.66	10	8.7
4323	UL89 V362M	0.69	0.12	7	0.6	4.7	1.1	8	1.2	46	3.7	11	98
4326	UL89 H389N	0.98	0.32	16	0.9	4.1	0.75	10	1.1	14	2.4	132	29
4369	UL89 N405D	0.72	0.21	17	0.7	3.9	0.61	12	1.0	6.9	0.54	12	15
4327	UL89 I334V N405D	0.71	0.18	10	0.6	4.5	0.82	12	1.2	5.6	0.59	9	12
4337	UL56 L208M	0.71	0.13	11	0.7	3.1	0.29	8	0.8	1.5	0.36	13	3.4
4330	UL56 E407D	0.99	0.14	8	0.9	5.1	0.63	7	1.3	2.7	0.16	9	6.0
4358	UL56 H637Q	1.0	0.17	10	0.9	3.4	0.48	7	0.9	0.89	0.09	10	2.0
4359	UL56 V639M	1.0	0.36	18	0.9	4.0	0.48	8	1.0	4.6	0.42	8	10
4356	UL56 H637Q V639M	0.91	0.16	8	0.8	4.6	1.0	13	1.2	7.5	1.0	10	17
4366	UL56 L764M	1.1	0.30	9	1.0	4.3	1.2	12	1.1	0.19	0.02	8	0.4

^aThe mean concentration required for 50% SEAP yield reduction (data are in nanomolar for letermovir and micromolar for the other compounds).

^bStandard deviation of EC₅₀ values.

^cNumber of replicates of testing (over at least 4 setup dates).

^dFold change in EC₅₀ compared with that for the wild type (fold changes of >1.5-fold are shown in bold, and underlined data denote cross-resistance). Fold changes of 1.5 are significant at a P value of <0.0025; all fold changes of >1.6 are significant at a P value of <1 × 10⁻⁴.

The UL89 substitution V362M conferred high-level tomeglovir resistance (a 98-fold increased EC₅₀) without conferring cross-resistance to GW275175X or letermovir. Among the variety of other UL89 substitutions also observed under tomeglovir selection, N329S, T350M, H389N, and N405D conferred 9-fold to 29-fold increased tomeglovir EC₅₀s, with the N329S mutant showing low-grade letermovir cross-resistance and the T350M mutant showing triple drug resistance (Table 3). The UL89 substitution I334V (seen with N405D in experiment M205 [Fig. 1]) phenotyped fully susceptible; I334V did not increase the tomeglovir resistance phenotype when it was combined with N405D, but it affected growth fitness (see below). Letermovir-selected UL89 substitutions N320H and M359I conferred little letermovir resistance but conferred significant cross-resistance to tomeglovir and also to GW275175X in the case of N320H (Table 3).

The UL56 Q204R substitution by itself conferred low-level resistance to GW275175X and tomeglovir and a borderline decreased susceptibility to letermovir. Other UL56 substitutions, L208M, E407D, H637Q, and V639M, including the H637Q and V639M double substitution, conferred tomeglovir resistance only, and the H637Q and V639M double mutant had greater tomeglovir resistance than either single mutant. The UL56 substitution N232Y, which was selected under letermovir exposure conferred low-grade tomeglovir cross-resistance, similar to that reported for R369M (10). UL56 substitutions L764M and E769D conferred no resistance by themselves. They appeared to be incomplete-sequence subpopulations in later passages of tomeglovir and GW275175X selection experiments, respectively, adding to the existing UL89 resistance mutations (Fig. 1). It is possible that these UL56 changes improve the growth fitness of the UL89 mutants, or they may be variants coincidentally selected with resistance mutations present in the same genome.

Comparative growth fitness.

Multicycle SEAP growth curve assays without added drug showed three categories of growth fitness, as judged by the final SEAP activity achieved in the supernatant at 8 days: wild-type levels (SEAP activities, 8.5 × 10⁵ to 10 × 10⁵ relative light units [RLU]) for the wild type and the UL56 V231L and UL89 D344E, C347S, and H389N single mutants, levels moderately reduced by ∼50% (SEAP activities, 3.5 × 10⁵ to 5.5 × 10⁵ RLU) for the UL56 K258E, UL89 N320H, T350M, and V362M single mutants, and levels markedly reduced by >75% (SEAP activities, 1.1 × 10⁵ to 2.3 × 10⁵ RLU) for the UL89 N329S, R351H, M359I, and N405D single mutants. Selected full growth curves from this data set are shown in Fig. 2. Because N405D was selected 3 times under tomeglovir exposure despite its poor growth fitness, growth curves were simultaneously determined for the N405D mutant without drug and in the presence of 2 μM tomeglovir and for the UL89 I334V N405D double mutant, as seen in experiment M205 (Fig. 1). These curves (Fig. 2) show 80% improved growth of the I334V N405D mutant at 8 days and 2.5-fold improved growth of the N405D mutant under 2 μM tomeglovir exposure compared with that of the mutant with the N405D mutation alone. The UL89 I334V change thus partially compensates for the impaired fitness of N405D, as does the presence of tomeglovir at 30% of its EC₅₀. Lesser growth enhancement of 40% to 80% at day 8 was also observed with the UL89 N320H, N329S, and M359I mutants under 1 to 2 μM tomeglovir exposure. Letermovir-resistant UL56 mutants are generally reported to have nearly wild-type growth fitness similar to that of the V231L mutant used here as a control (9, 10), but the growth of the K258E mutant was perceptibly retarded (Fig. 2).

FIG 2.

Comparative growth curves of mutant recombinant viruses. Cell-free virus stocks were inoculated on ARPEp cells at an equal low multiplicity of infection (0.02), and culture supernatants were sampled on days 1 and 4 through 8 for SEAP activity, assayed as the number of relative light units (RLU) emitted from a chemiluminescent substrate. The growth curves shown are those for the wild-type control (WT) and recombinants with the indicated UL89 amino acid substitutions, except for one strain with UL56 substitution K258E. The growth curves for the UL89 N405D mutant were determined with and without the addition of 2 μM tomeglovir (TGV2). The error bars represent the standard deviations from 4 replicates.

DISCUSSION

In this in vitro selection study, the older terminase inhibitors GW275175X and tomeglovir selected for characteristic UL89 gene mutations, notably, D344E, C347S, and V362M, which conferred moderate to high-level drug resistance. Some UL56 mutations also evolved that conferred lower-grade resistance to these compounds but not to letermovir, except that a borderline letermovir susceptibility phenotype was observed for the Q204R mutant. In contrast, mutations conferring letermovir resistance, including all mutations known to confer high-grade resistance, map predominantly to the UL56 gene (codon 229 to 369 range), but this study revealed the occasional emergence of UL89 mutations under letermovir exposure, e.g., the same D344E mutation seen under GW275175X exposure, which had the effect of slightly decreasing letermovir susceptibility and augmenting the resistance conferred by UL56 mutation. The cross-resistance of some UL89 mutants to two or all three of the chemically unrelated inhibitors suggests that the compounds target a similar function of the terminase complex that includes participation of both UL56 and UL89 residues.

Experimental data on the structure and functional domains of the CMV terminase complex are quite limited. The mapping of drug resistance mutations may offer clues as to the sites of molecular interactions, but the wide range of residues involved in both pUL56 and pUL89 makes it unlikely that they all converge on drug-binding sites. More likely, some mutations affect drug susceptibility through allosteric, remotely transmitted conformational changes, and it is also possible that there are multiple binding sites for individual compounds. However, for each compound there is a distinct locus that confers high-grade resistance when it is mutated (UL56 C325 for letermovir, UL89 C347 for GW275175X, and UL89 V362 for tomeglovir), suggesting that these residues are the most relevant to the specific drug interaction.

The CMV UL89 terminase component includes the C-terminal nuclease domain that was expressed in vitro to obtain an experimental structure (12). Efforts to crystallize the complete UL89 protein were unsuccessful (12) and even if successful would still raise issues of interpretation because the terminase complex potentially includes dimers of UL56 and UL89 interacting among themselves and with unknown copy numbers of other proteins, such as UL51 and UL104 (2). Nevertheless, a protein structure prediction server (I-TASSER) (13) returned the pUL89 structure model shown in Fig. 3, based on the published structure of the nuclease domain and partial sequence similarities to the bacteriophage T4 gp17 terminase motor protein (a moderate confidence C score of −0.84). In this model, most of the UL89 mutations selected after exposure to the three terminase inhibitors were in relative proximity in regions putatively involved in ATPase-driven motor function or nucleic acid binding/recognition (14). More distant mutations may affect drug susceptibility by allosteric means rather than by direct displacement of inhibitor binding to a biological target, considering that low concentrations of tomeglovir markedly improved the growth of the resistant UL89 N405D mutant (and some others, but to a lesser extent) over the level observed with no drug present. This helps to explain the relatively frequent selection of UL89 N405D in the presence of low concentrations of tomeglovir (Table 1), despite the low growth fitness of this mutant (Fig. 2).

FIG 3.

UL89 protein structure model. A ribbon representation of the top-ranking structure model created by the I-TASSER server (13). The C-terminal nuclease domain is based on the experimentally determined structure, while the remainder relies on limited peptide homology with proteins, such as the bacteriophage terminase gp17. Residues affecting drug susceptibility are labeled but were not used in model construction. Colored residues 463, 534, and 651 in the nuclease domain are involved in metal ion coordination (12).

The numerous UL89 mutations encountered in these selection experiments notably do not map to the UL89 nuclease domain (Fig. 3). Expression of this domain gave a product with nuclease activity but without site specificity (12). The structural resemblance to the HIV integrase suggests that certain metal-chelating integrase inhibitor molecules may effectively target the UL89 nuclease domain independently of the terminase inhibitors studied here (15, 16).

Mapping of UL56 resistance mutations to the corresponding structure and function is highly speculative because of the lack of homologous structures for modeling. One low-confidence model returned by I-TASSER did show the open toroid structure postulated for this terminase component (2). The range of mapped UL56 residues conferring drug resistance, including residues 204 to 208, 229 to 244, 257 to 261, 325, 369, 407, and 637 to 639, hints at possible loci of functional interactions with pUL89, given the involvement of mutations in both genes for resistance to each compound studied. Self-interacting residues are also possible, given the presumed dimeric structure of pUL56. The loci of UL56 mutation after letermovir exposure are close to some peptide motifs possibly involved in DNA binding and speculatively involved in recognition of the pac site for cleavage of unit-length genomes (17).

Validated genotype-phenotype correlations are the basis for the genotypic diagnosis of letermovir resistance at it comes into clinical use and provide guidance for the discovery of alternative terminase inhibitors. A concern about terminase inhibitors as a drug class is their genetic barrier to resistance, particularly when they are used alone for treatment, since a missed dose may result in the rapid resumption of infectious virus production, like that which occurs in vitro when terminase inhibition is withdrawn (18). When the same baseline error-prone exonuclease mutant was used, a resistance mutation developed after 5 culture passages in all 17 experiments with letermovir (15 published [9] and 2 current), 2 of 6 with GW275175X, and 5 of 8 with tomeglovir, whereas they developed in 0 of 8 experiments with foscarnet (9). Resistance can develop through a number of genetic pathways at a relatively low fitness cost, although the number of UL89 mutants whose growth was perceptibly impaired was greater than the number previously reported for the UL56 letermovir-resistant mutants (9).

Emerging resistance is likely to be an issue needing close monitoring with the clinical use of terminase inhibitors for the treatment of active CMV infection (as opposed to prophylaxis). More attention may be needed than that required with the currently licensed polymerase inhibitors. However, the development of alternative CMV antiviral targets remains a priority, in order to allow a combination treatment strategy that has been successful for HIV and hepatitis C virus. The relationships among the mutations selected under treatment with the different chemical classes of terminase inhibitors should facilitate the evaluation of derivative compounds of each class.

MATERIALS AND METHODS

Antiviral compounds.

Letermovir was commercially sourced (9). Tomeglovir was supplied as an initial sample from Michael McVoy (Virginia Commonwealth University) and additionally by eNovation Chemicals (catalog no. D388253). GW275175X was supplied by Karen Biron (GlaxoSmithKline). Compounds were prepared as 50 to 100 mM stock solutions in dimethyl sulfoxide and diluted into cell culture media as needed.

Viral strains.

A strain AD169-derived cloned recombinant virus, T4138, containing an in-frame deletion of UL54 exonuclease codon 413 (9) was used for all in vitro drug selection experiments. Recombinant phenotyping was performed in derivatives of AD169-based bacterial artificial chromosome (BAC) clone BD1, which contains a secreted alkaline phosphatase (SEAP) reporter gene for viral growth quantitation (9).

Cell cultures.

In vitro selection experiments were performed in the same human foreskin fibroblasts (HFF) described previously (9) and cultured in Eagle minimal essential medium with glutamine and 3% to 8% fetal bovine serum supplementation. Reporter-based SEAP yield reduction assays were performed in ARPEp cells, which are retinal epithelial cells transduced to overexpress platelet-derived growth factor receptor alpha (PDGFRα), rendering the cells fully permissive for laboratory strains, such as AD169 (19). ARPEp cells were maintained in Dulbecco minimal essential medium with 4.5 g/liter glucose, 2.5 mM glutamine, and 3% to 8% fetal bovine serum.

In vitro propagation under antiviral compound treatment.

In vitro propagation under antiviral compound exposure was performed as previously described for letermovir (9), using the exonuclease mutant strain T4138 in HFF and starting concentrations of 5 nM letermovir or 0.5 to 1 μM GW275175X or tomeglovir. These concentrations are close to their respective EC₅₀ values. Passage of ∼30% of infected cells to a fresh HFF culture was performed weekly with the escalation of the drug concentration as tolerated until active viral growth was observed under selection with 30 μM letermovir or 60 μM GW275175X or tomeglovir. At every few passages, DNA was extracted from infected cells and standard dideoxy sequencing (BigDye Terminator cycle sequencing kit; Applied Biosystems) was performed on PCR products representing the CMV UL56 gene and both exons of the CMV UL89 gene, as described previously (9). Sequencing chromatograms and data were aligned to the strain AD169 reference sequence. An estimate of the mutant population in the case of mixed peaks was made by comparing the relative peak heights of the various base calls.

Recombinant phenotyping.

Uncharacterized amino acid substitutions in the UL56 and UL89 genes were phenotyped after construction of recombinant viruses as previously described (9). To facilitate the construction and purification of various mutant clones of UL56 and UL89 exon 2, a transfer vector containing an Flp recombination target (FRT)-delimited kanamycin resistance (kan) marker was inserted upstream of each gene region and the desired UL56 or UL89 mutation was introduced into the transfer vector. This was recombined into a bacterial artificial chromosome (BAC) clone lacking the corresponding UL56- or UL89 exon 2-coding region (to prevent carryover of the wild-type sequence). Double mutant BAC clones of UL89 and UL56 were constructed by recombining a UL56 mutant transfer vector into a parental BAC clone containing UL89 D344E and a deletion of the UL56-coding sequence. After removal of the kan marker by induction of Flp recombinase, the recombinant BAC was transfected into HFF or ARPEp cells to recover live mutant virus, the UL56 or UL89 exon 2 sequences of which were checked throughout for the intended sequence. A cell-free virus stock was prepared and used to determine the drug concentrations required to reduce viral growth by 50% (EC₅₀) by assay of the secreted alkaline phosphatase (SEAP) reporter in the culture supernatant of ARPEp cells in 24-well culture plates (9, 19). Baseline wild-type strains (with the FRT motif upstream of UL56 or UL89) and resistant control strains (with UL56 V231L or UL89 D344E or H389N) were included for the calibration of each setup. For each mutant tested, at least 7 replicates of the EC₅₀ assays were performed on at least 4 separate setup dates to control for variation in cell culture conditions. The statistical significance of differences in mean EC₅₀ values was assessed by the Student t test (two-tailed, unequal standard deviations [SDs]) for mutant and wild-type control strains set up on the same dates and with the same batches of cell cultures. The 95% confidence intervals for EC₅₀ values were calculated from the mean EC₅₀, number of replicates (n), and SDs listed in Table 1 as mean ± $(t\times \text{SD})/\sqrt{n}$, where t (value available as Microsoft Excel function TINV using parameters of 0.05 and n−1) ranges from 2.45 (for n = 7) to 1.98 (for n = 100).

Comparative in vitro growth fitness.

The comparative in vitro growth fitness of different strains was assessed from the SEAP growth curves as previously published (9). After inoculation at a comparable multiplicity of infection (∼0.02), as calibrated by the SEAP activity in the culture supernatant at 24 h, further aliquots of culture supernatants were collected on each of days 4 through 8 postinoculation and assayed for SEAP activity as a reporter-based measure of viral growth.