Phenotypes of cytomegalovirus genetic variants encountered in a letermovir clinical trial illustrate the importance of genotyping validation

Abstract

Emergence of drug resistance is rare after use of letermovir (LMV) as prophylaxis for post-transplant cytomegalovirus (CMV) infection. In a recent study involving renal transplant recipients, no known LMV resistance mutations were detected in those receiving LMV prophylaxis. However, uncharacterized viral amino acid substitutions were detected in LMV recipients by deep sequencing in viral subpopulations of 5% to 7%, at codons previously associated with drug resistance: UL56 S229Y (n=1), UL56 M329I (n=9) and UL89 D344Y (n=5). Phenotypic analysis of these mutations in a cloned laboratory CMV strain showed that S229Y conferred a 2-fold increase in LMV EC50, M329I conferred no LMV resistance, and D344Y knocked out viral viability that was restored after the nonviable clone was reverted to wild type D344. As in previous CMV antiviral trials, the detection of nonviable mutations, even in multiple study subjects, raises strong suspicion of genotyping artifacts and encourages the use of replicate testing for authentication of atypical mutation readouts. The non-viability of UL89 D344Y also confirms the biologically important locus of the D344E substitution that confers resistance to benzimidazole CMV terminase complex inhibitors, but does not feature prominently in LMV resistance.

Short communication

Letermovir (LMV) is a potent inhibitor of the human cytomegalovirus (CMV) terminase complex, which was initially approved for prophylaxis in stem cell recipients and more recently in renal transplant recipients (Merck, 2023). Despite concerns about a lower in vitro genetic barrier to the development of drug resistance (Chou, 2015), the efficacy of LMV as prophylaxis may be unaffected, as long as viral suppression is maintained and any significant breakthrough infections are identified and treated. In the phase 3 trial involving stem cell recipients (P001), LMV resistance was detected in only three patients of 40 who experienced CMV breakthrough among 373 who received LMV prophylaxis (Douglas et al., 2020). A subsequent single center study of 262 patients showed infrequent detection of CMV breakthough at >500 IU/mL, with one detected case of resistance (Perchetti et al., 2023). In a separate phase 3 trial in renal transplant recipients (P002), no cases of known LMV resistance were detected among recipients of LMV prophylaxis using the same deep sequencing technology as P001 (Limaye et al., 2023; Strizki et al., 2023). However, uncharacterized amino acid substitutions were noted in CMV terminase genes at codons previously associated with LMV resistance, namely UL56 S229Y (n=1), UL56 M329I (n=9) and UL89 D344Y (n=5) as small sequence subpopulations of 5 to 7% in multiple (n) patients (Merck, 2023). Since the clinical significance of these substitutions was unknown, a plan for phenotypic testing was stated in the product label and the results are reported here.

The phenotypes of substitutions UL56 S229Y and M329I, and UL89 D344Y were determined by constructing bacterial artificial chromosome (BAC) clones of reference laboratory CMV strain AD169 individually containing each of these genetic variants. The resulting live recombinant viruses were tested in cell culture for the LMV concentration required to reduce viral growth by 50% (EC50) as measured by a reporter gene, as previously described (Chou and Kleiboeker, 2022). The same methods have been used to characterize many other known LMV resistance mutations (Chou, 2015, 2017a, b). All EC50 values are calibrated to a baseline wild type control and known low-level LMV-resistant mutant UL56 V231L.

Construction of the UL56 S229Y and M329I mutants was straightforward and yielded normally growing live virus stocks on the first attempt. Phenotypic assays revealed EC50 values 2-fold increased over the baseline wild type control strain for S229Y and no increase over baseline for M329I (Table 1), while showing an 8-fold EC50 elevation for the control LMV-resistant V231L mutant as previously published (Chou and Kleiboeker, 2022). These data compare with the low-grade EC50 increases published for the in-vitro selected mutants UL56 S229F (1.8-fold)(Chou, 2017b) and M329T (4.5-fold)(Chou, 2017a).

Table 1.

Genotypes and phenotypes of UL56 mutant viruses

Strain1	UL56 Genotype2	Letermovir EC50, nM
		Mean	SD	N	Ratio
Control strains
4190	WT	2.6	0.31	18
4194	V231L	21	2.2	17	8.1
New recombinants
4583	S229Y	5.4	0.69	18	2.0
4581	M329I	2.0	0.24	16	0.77

^1.Serial number of recombinant CMV strain

^2.Amino acid substitution relative to strain AD169

WT = wild type, including silent Frt motif upstream of UL56 gene in all strains

SD = standard deviation; N = number of replicates

Ratio = EC50 of mutant virus/EC50 of wild type control

In contrast, multiple attempts to generate a live UL89 D344Y mutant virus were unsuccessful. Four separate BAC clones were constructed (Table 2), two of which introduced D344Y into the UL89 context of strain AD169. Since strain AD169 contains a sequence polymorphism at codon 345 seldom found in clinical isolates (Pilorge et al., 2014), and another common locus of polymorphism at codon 665, we also made clones that contained variants D344Y+S345A, and D344Y+S345A+D665E (relative to strain AD169). Sequence verification of the BAC clones included the entire UL89 exon 2 context of the transfer vector used to perform the recombinant mutagenesis (Chou, 2017a). To confirm that the nonviable BAC clones did not have latent defects outside the UL89 exon 2 region, two of the nonviable BAC clones were reverted from D344Y to the wild type D344 configuration using the wild type transfer vector and the revertant BAC DNAs yielded live CMV on the first transfection attempt, with wild-type LMV EC50 values (Table 2). Based on the above evidence, it was concluded that UL89 D344Y is a knockout mutation, in contrast to the D344E mutation that has long been known to confer resistance to benzimidazole terminase inhibitors (Krosky et al., 1998) and borderline decreased susceptibility to LMV (EC50 ratio 1.8), while maintaining normal growth fitness (Chou, 2017a).

Table 2.

Genotypes and phenotypes of UL89 BAC clones

BAC clone1	UL89 Genotype2	Transfections (N)	Live virus3	Letermovir EC50, nM
		Attempts/Live virus		Mean	SD	N	Ratio
WT control
BD54	WT	1/1	4272	2.5	0.32	6
D344Y recombinants
BD291	D344Y	2/0
BD294	D344Y	3/0
BD300	D344Y S345A	2/0
BD301	D344Y S345A D665E	1/0
WT revertants
BD296	WT from BD294	1/1	4588	2.7	0.25	4 1.1
BD302	WT from BD300	1/1	4597	3.1	0.26	1.2

^1.Bacterial artificial chromosome clone name

^2.Amino acid substitution(s), relative to strain AD169. Parent D344Y clone listed for revertants

^3.Serial number of CMV strain derived from BAC clone

WT = wild type (strain AD169), including silent Frt motif upstream of UL89 gene exon 2 in all constructs

SD = standard deviation; N = number of replicates

Ratio = EC50 of mutant virus/EC50 of wild type control

The phenotype data reported here indicate no clinical significance for these CMV variants from study P002 (Merck, 2023), despite their location at codon loci of known LMV resistance mutations. Among the 15 detections of these variants, only one (UL56 S229Y) involved any perceptible elevation of EC50, and then only a 2-fold change from baseline that is far from the 50- to >8000-fold elevation conferred by the great majority of clinically observed LMV resistance mutations (Chou and Kleiboeker, 2022). The EC50s of low-grade LMV-resistant mutants selected in vitro at nanomolar drug concentrations are well below the expected trough therapeutic concentrations (Douglas et al., 2020).

The detection of a knockout mutation UL89 D344Y recalls the many instances of sequencing artifacts encountered in CMV antiviral drug trials, whether by traditional Sanger sequencing (Chou et al., 2014; Chou et al., 2022) or by newer deep sequencing technology (Douglas et al., 2020). For example, eight UL54 DNA polymerase knockout mutations, including one detected in three patients, were not confirmed upon re-sequencing of their original clinical specimens (Chou et al., 2014). There is increased risk of sequencing artifacts in situations of low-level viral breakthrough as illustrated in trial P001, where novel mutant viral sequence subpopulations of up to 33% detected by the same deep sequencing technology as used for trial P002 were irreproducible (Douglas et al., 2020). Similar information on associated viral loads and replicate testing has not been presented for specimens involved in the current study, while the finding of LMV resistance mutations among those without known exposure to the drug raises concerns about reproducibility of other detected variants (Strizki et al., 2023). Although there is a theoretical possibility that both a knockout mutation UL89 D344Y and compensatory mutation(s) restoring viral viability exist in 5 separate patients, attention to rigor and reproducibility of genotyping readouts is recommended in view of the past experience cited. In trial P001, the deep sequencing readout pipeline was independently validated (Komatsu et al., 2019), and sufficient replicates of testing per sample (often 5 to 10) were included to reject the likelihood of quasispecies sampling error causing the fluctuating appearance of sequence subpopulations (Douglas et al., 2020). This leaves PCR amplification errors from clinical samples as the most likely source of sequencing artifacts.

Regardless of the authenticity of the UL89 D344Y knockout mutant in LMV-treated patients, this finding is interesting in confirming the essential biological function of this residue. Although its involvement in LMV antiviral activity appears to be minor, the UL89 D344E mutation that confers resistance to benzimidazole inhibitors (BDCRB and GW275175X) is often selected in vitro under those compounds (Chou, 2017a; Krosky et al., 1998). Hypothetically, residues in this vicinity of pUL89 are involved in the substrate specificity of its nuclease (Chou, 2017a).

In conclusion, the phenotypes of sequence variants UL56 S229Y and M329I, and UL89 D344Y detected as minor subpopulations in clinical trial P002 do not suggest a role in clinical LMV resistance but reinforce prior recommendations for confirmatory replicate testing of unusual mutations detected in clinical trials, especially as mixed sequence populations in low copy number specimens

• Cytomegalovirus terminase gene mutations detected as small subpopulations in a letermovir clinical trial were phenotyped

• UL56 substitutions S229Y and M329I conferred slight (2-fold) or no decreased susceptibility to letermovir

• UL89 substitution D344Y had a nonviable phenotype

• The phenotypes of these mutations do not suggest any clinical significance for outcomes of letermovir prophylaxis

• Atypical mutation readouts should be confirmed by replicate testing before proceeding with phenotypic analyses