Letermovir Resistance Analysis in a Clinical Trial of Cytomegalovirus Prophylaxis for Hematopoietic Stem Cell Transplant Recipients

Abstract

Background

Letermovir (LET), a cytomegalovirus (CMV) deoxyribonucleic acid (DNA) terminase inhibitor, was recently approved for prophylaxis of CMV infection in adult CMV-seropositive recipients of allogeneic hematopoietic stem cell transplantation. Cytomegalovirus genotyping was performed to identify LET-resistance-associated variants (RAVs) among subjects in a Phase 3 trial.

Methods

The CMV UL56 and UL89 genes, encoding subunits of CMV DNA terminase, were sequenced from plasma collected from subjects with clinically significant CMV infection (CS-CMVi). Novel variants were evaluated by recombinant phenotyping to assess their potential to confer resistance to LET.

Results

Genotyping was successful for 50 of 79 LET subjects with CS-CMVi. Resistance-associated variants (encoding pUL56 V236M and C325W) were detected independently in subjects 1 and 3 who experienced CS-CMVi while receiving LET prophylaxis, and 2 other variants (encoding pUL56 E237G and R369T) were detected >3 weeks after subjects 2 and 3, respectively, had discontinued LET prophylaxis and received preemptive therapy with ganciclovir.

Conclusions

The detected incidence of CMV resistance among subjects who received LET as prophylaxis in this Phase 3 trial was low. The LET RAVs that were detected mapped to the CMV UL56 gene at positions associated with reduced susceptibility to LET based on resistance selections in cell culture.

In a Phase 3 trial of letermovir for cytomegalovirus (CMV) prophylaxis in hematopoietic stem cell transplant recipients, UL56 gene mutations conferring letermovir resistance were identified in 3 subjects among 50 who developed clinically significant CMV infection and had genotyping results.

(See the Editorial Commentary by James, on pages 1036–8.)

Among hematopoietic stem cell transplant (HSCT) recipients who are cytomegalovirus (CMV) seropositive (R⁺), without preemptive therapy (PET) 80% of recipients experience CMV infection, and 20% to 35% of them develop CMV disease [1]. In a Phase 3 randomized, double-blind, placebo-controlled trial for prevention of CMV infection in adult CMV-seropositive recipients of allogeneic HSCT, letermovir (LET) reduced clinically significant CMV infection (CS-CMVi) and was well tolerated [2].

Letermovir is a potent inhibitor of the viral enzyme CMV deoxyribonucleic acid (DNA) terminase, which cleaves newly synthesized CMV DNA into individual viral genomes and guides them into empty viral capsids [3, 4]. Inhibition of CMV terminase leads to a block of viral spread. The CMV DNA terminase complex contains the pUL56, pUL89, and pUL51 proteins, and mutations in the genes encoding these proteins have been shown to confer reduced susceptibility to LET in cell culture. Almost all LET resistance-associated variants (RAVs) map to the UL56 gene in the region encoding amino acids (AA) 229–369 [5]. The pUL56 subunit of the terminase complex is believed to play a role in DNA packaging through sequence-specific binding of DNA packaging motifs in CMV genome concatemers [6]. Inhibition of terminase is a novel mechanism for an anti-CMV drug that avoids cross-resistance to CMV DNA polymerase inhibitors [7–9]. Of 98 subjects receiving LET in a prior Phase 2b prophylaxis trial [10], only 1, who received a suboptimal 60 mg daily dose, had a mutation (encoding the pUL56 substitution V236M) that confers reduced susceptibility to LET [11]. To identify relevant CMV variants in the Phase 3 trial of LET prophylaxis in HSCT recipients, next-generation sequencing (NGS) [12, 13] was performed using plasma collected from subjects who experienced CS-CMVi through week 24 posttransplant. Replicate testing was used to exclude genotyping artifacts, and confirmed genotypic variants (GVs) were evaluated by recombinant phenotyping to assess their impact on LET susceptibility.

MATERIALS AND METHODS

Study Specimens

Details of this Phase 3 randomized, placebo-controlled clinical trial of LET (MK-8228) for the prevention of CS-CMVi in adult, CMV-seropositive allogeneic HSCT recipients have been published [2]. Subjects who developed CS-CMVi during the prophylaxis period (up to week 14 posttransplant) or follow-up period (weeks 14 to 24 posttransplant) were scheduled for blood collection at a CMVi visit immediately before the initiation of CMV PET. A sample designated for CMV DNA sequence analysis was available for approximately two thirds of subjects with CS-CMVi; for most of the remaining subjects, another sample collected close to this timepoint (±7 days) was available for genotypic analysis.

Deoxyribonucleic Acid Isolation, Polymerase Chain Reaction Amplification of UL56/UL89, and Next-Generation Sequencing

Total DNA was extracted from plasma isolated from each blood sample, and nested polymerase chain reaction (PCR) was used to amplify DNA from 3 CMV gene targets (UL56, UL89A [Exon 2] and UL89B [Exon 1]). Products were analyzed using the Illumina Miseq system and the “Athena” pipeline (designed for the analysis of amplicon-based deep sequencing data). The deduced UL56 and UL89 AA sequences were aligned with a CMV reference strain (GenBank accession number CMV_Merlin_NC006273). Differences detected at a frequency of ≥5% of the sequence reads that resulted in an AA deletion or substitution in pUL56 or pUL89 were identified as CMV GVs. Additional details are in Supplementary Methods.

Sample Testing Algorithm

Because the volume of plasma used for CMV genotyping was often limited, a testing algorithm based on previously reported patterns of LET resistance was implemented (Supplementary Figure 1). UL56 PCR was attempted first; if the amplification was successful, UL89 PCR was attempted.

Cytomegalovirus UL56 and UL89 Genotypic Variants

Observed UL56 and UL89 GVs were considered characterized if they were phenotyped as either resistant or susceptible to LET (Supplementary Tables 1 and 2). Uncharacterized GVs include the following: (1) previously reported AA substitutions with unknown impact on LET susceptibility and (2) substitutions that have not been reported previously.

Replicate Testing

To distinguish true CMV GVs from artifacts (especially in samples with low viral copy number), genotyping was repeated on novel uncharacterized variants that were observed in ≥5% but <99% of the NGS reads. Replicate testing was performed by repeating the DNA extraction from the original plasma sample, amplifying the coding sequences of UL56 and/or UL89, and sequencing each PCR product.

Recombinant Phenotyping

Recombinant phenotyping includes introduction of specific mutations into a wild-type CMV bacterial artificial chromosome clone with a secreted alkaline phosphatase (SEAP) reporter gene and testing the resulting recombinant virus for drug susceptibility by a standardized reporter-based yield reduction assay [14–16]. Infrequent substitutions at conserved AAs of CMV pUL56 or pUL89 that were identified in LET subjects were candidates for phenotypic analysis.

Human Subjects

The trial was conducted in accordance with the principles of the Declaration of Helsinki and Good Clinical Practice guidelines. The institutional review board at each center approved the trial. All of the patients provided written informed consent (Clinicaltrials.gov. NCT02137772).

RESULTS

Genotyping Analysis Population

In this Phase 3 clinical trial of LET prophylaxis, a total of 565 subjects underwent randomization; 373 received at least 1 dose of LET and 192 received at least 1 dose of placebo (PBO) (the All Randomized and Treated [ARaT] population) (Figures 1 and [2]). The genotyping analysis population (GAP) consisted of the 167 subjects in the ARaT population (79 LET and 88 PBO) who had CS-CMVi through week 24 posttransplant. A subgroup (the full analysis set [FAS] GAP) included those subjects in the GAP without detectable CMV DNA on day 1. Subjects in the GAP can be further divided into those who had CS-CMVi while receiving study medication (ie, through week 14 posttransplant) and those who had CS-CMVi during the follow-up period (weeks 14–24 posttransplant).

Figure 1.

The genotyping analysis population. Values in the boxes represent the number of subjects who met the criteria. ARaT, All Randomized and Treated; CS-CMVi, clinically significant cytomegalovirus infection; Day 1, first day the subject received study medication; LET, letermovir; PBO, placebo; W14, 14 weeks posttransplant, when subjects received the last dose of LET or PBO; W14-W24, 14–24 weeks posttransplant, the 10-week period between the last day the subject received LET or PBO and the 24-week posttransplant primary endpoint.

Cytomegalovirus UL56/UL89 Sequencing Success for Genotyping Analysis Population Subjects

For the CMV genotypic analysis, DNA was isolated from plasma and PCR amplification of the protein coding regions of the CMV UL56 and UL89 genes was attempted, followed by NGS of successful amplicons. Because volume was limiting for many plasma samples, UL56 sequencing was prioritized over UL89 sequencing. Overall, UL56 genotyping was successful for 50 of 79 (63%) LET subjects and 53 of 65 (81%) PBO subjects tested (Table 1). The difference in genotyping success for LET versus PBO subjects reflects the lower median CMV viral load of LET subjects at the CMVi visit (Supplementary Table 3). A similar association was seen when comparing viral loads of LET subjects in the ARaT GAP who failed during the 10-week post-LET follow-up (39 subjects, 82% genotyping success; median 471 CMV DNA copies/mL) versus those who failed while receiving LET (40 subjects, 45% genotyping success; median 299 CMV DNA copies/mL) (Supplementary Table 4). The rate of successful UL89 genotyping (UL89A + UL89B) among samples with NGS data available was ≥70% for LET and PBO subjects from both the FAS GAP and from those subjects who had detectable CMV DNAemia on day 1 (Table 1).

Table 1.

Frequency of UL56/UL89A/UL89B Next-Generation Sequencing Success Among Subjects With Clinically Significant CMV Infection Through Week 24 Posttransplant

Category	Letermovir Subjects	Placebo Subjects	Total Subjects
CS-CMVi Through Week 24 Posttransplant	79	88	167
Not CMV DNAemic on day 1 (FAS)	57	71	128
No NGS dataa	17	29	46
NGS data available	40	42	82
UL56 + UL89A + UL89B	33	36	69
UL56 + UL89A only	1	3	4
UL56 + UL89B only	1	1	2
UL56 only	5	1	6
UL89 onlyb	0	1	1
CMV DNAemic on day 1	22	17	39
No NGS data	12	6	18
NGS data available	10	11	21
UL56 + UL89A + UL89B	7	10	17
UL56 + UL89A only	1	1	2
UL56 only	2	0	2

Abbreviations: CMV, cytomegalovirus; CS-CMVi, clinically significant CMV infection; DNA, deoxyribonucleic acid; FAS, full analysis set; NGS, next-generation sequencing; UL89A, DNA sequence encoding pUL89 amino acids 297–674 (Exon 2): UL89B, DNA sequence encoding pUL89 amino acids 1–296 (Exon 1).

^aPolymerase chain reaction (PCR) amplification of UL56 was attempted but unsuccessful for letermovir subjects with “No NGS data”; for subjects who received placebo, 12 had no NGS data because PCR was attempted but unsuccessful, and 23 had no NGS data because samples designated for CMV DNA sequence analysis were not available, and CMV genotyping using other samples (Materials and Methods) was not attempted.

^bIn a deviation from the sample testing algorithm, UL56 PCR in the QiAxcel capillary electrophoresis analysis was incorrectly identified as “PCR positive” for this sample from a subject who received placebo. Aliquots of the remaining undiluted DNA sample were used for UL89A and UL89B PCR assays. When data reconciliation identified the original UL56 PCR test as unsuccessful, an aliquot of the remaining undiluted DNA from this sample was diluted 1:1 with water, and the UL56 PCR test was performed using the diluted DNA as template. PCR of UL56 using diluted DNA was also unsuccessful. Consequently, no UL89A/UL89B PCR using the diluted DNA was performed. The UL89A and UL89B PCR tests performed using undiluted DNA from this sample were successful, and the products were analyzed by NGS.

Replicate Testing Identifies Probable Polymerase Chain Reaction/Next-Generation Sequencing Artifacts

Cytomegalovirus UL56/UL89 GVs were defined as any nucleotide difference observed in ≥5% of the sequence reads that resulted in AA substitutions or deletions in pUL56 or pUL89 relative to the reference sequence. The concentration of CMV DNA in many samples collected during the CMVi visit was at or near the limit of quantitation (151 copies/mL plasma), which increases the chance of sampling bias and PCR-generated artifacts. To identify variants likely to be artifacts rather than true GVs, genotypic analysis was repeated as described in the Materials and Methods section for 9 UL56 variants and 13 UL89 variants. Only 3 of 9 UL56 variants and 0 of 13 UL89 variants were reproducibly detected (Supplementary Table 5).

Cytomegalovirus UL56/UL89 Genotypic Variants

Most of the substitutions encoded by CMV UL56/UL89 GVs in GAP subjects were known polymorphisms that have been reported in public databases; some were found previously in CMV strains that were tested for susceptibility to LET [8]. Among all subjects in the GAP with NGS results, the incidence of total CMV UL56 and UL89 GVs was similar in LET- and PBO-treated subjects, and none of the PBO subjects had a GV identified that is known to confer resistance to LET.

Novel uncharacterized pUL56 and pUL89 substitutions among the 50 LET subjects with NGS results are shown in Figure 2 and listed in Table 2. Each substitution was detected in a single subject (except pUL56 E485G and SNS445-447del, which were seen together in 4 LET subjects). Two known LET RAVs in pUL56 (V236M and C325W) were identified in different subjects (1 in each), and 2 novel pUL56 LET RAVs (E237G and R369T) were detected at AAs where different LET RAVs (E237D and R369G/M/S) had been selected previously in cell culture (Supplementary Table 1). Neither of the pUL89 substitutions previously reported to confer reduced susceptibility to LET in CMV-infected cells (N320H, D344E) was detected in the 42 LET subjects with UL89A NGS results (Figure 2B, Table 1, Supplementary Table 2).

Figure 2.

Novel uncharacterized substitutions in cytomegalovirus (CMV) pUL56 and pUL89 among letermovir (LET) subjects with CS-CMVi through week 24 posttransplant. Thick horizontal lines represent the CMV pUL56 (A) and pUL89 (B) amino acid sequences (first and last residues are shown). Boxes with light shading represent regions of amino acids sequence conservation among database entries of pUL56 (A; [26]) and pUL89 (B; [26]). Regions of sequence variability (black boxes) and the region encompassing previously characterized letermovir (LET) resistance-associated variants (RAVs) from in vitro resistance selections (red box; Supplementary Table 1) are shown in A. Substitutions that were (brown text) or were not (black text) analyzed by recombinant phenotyping are shown above the line; variants that conferred reduced susceptibility to LET (red text) are below the line. The pUL56 substitutions V236M [7], C325W [5], and E485G [8] and the SNS445-447 deletion [17] are shown in A and were phenotyped even though they were not novel. Underlined, italicized LET RAVs were first detected in plasma that was collected 32 days (E237G) or 25 days (R369T) after the last dose of LET was administered.

Table 2.

Candidates for Phenotypic Analysis: Genotypic Variants in the CMV UL56 and UL89 Genes^a

Amino Acid Substitution Encoded by GV	Subject Number	Days LET Exposure/ Days Tx to GV	GV Detected Before (LET) or After (F/U) W14 Posttransplant	%NGS Reads With GV	Plasma CMV DNA (Copies/mL) on Sampling Date	Phenotyping Performed? (Y/N)	Comments
Substitutions in pUL56
M3Vb	15	76/125	F/U	3	5815	Y	Below validated NGS threshold
N17Dc	4	91/130	F/U	13	<151	N	Low% GV; low CMV DNA
A164V	18	85/139	F/U	99	1067	N	Present in PBO subject
V236M	1	62/84	LET	99	1076	Y	Previously characterized; LET RAV
S255L	14	96/159	F/U	100	188	Y
E237Gb	2	10/46	LET	4	<151	Y	Below validated NGS threshold; LET RAV
C325Wd	3	46/73	LET	99	28 096	Y	Previously characterized; LET RAV
R369Td,e	3	46/73	LET	e	e	Y	LET RAV
S445delc,f	7	83/152	F/U	91	<151	Y
SNS445-447delb,g	1, 8, 10h, 19h	h	h	g	g	Y	Common polymorphism
D480Ec	5	7/27	LET	25	166	N	Low %GV; low CMV DNA
E485Gb,i	1, 8, 10h, 19h	h	h	i	i	Y	Previously characterized
E485G +SNS445-447delb,g,i	1, 8, 10h, 19h	h	h	g,i	g,i	Y	Previously characterized + common polymorphism
Y575C	6	92/125	F/U	100	3706	Y
F626Sc	7	83/152	F/U	28	<151	Y
L726V	11	90/127	F/U	100	541	Y
W740Rc	4	91/130	F/U	21	<151	N	Low %GV; low CMV DNA
T775I	8	32/71	LET	100	293	Y
AVS789-791delc,j	13	85/127	F/U	88	572	Y
V814Ac	12	77/145	F/U	7	2067	Y
R816W	17	94/145	F/U	99	21 190	Y
R826Lc,k	16	90/133	F/U	5	<151	N	Low %GV; low CMV DNA
Substitutions in pUL89
E77Dl	9	97/151	F/U	99	10 921	N
I226Tc	4	91/130	F/U	6	<151	N	Low %GV; low CMV DNA
G235Sc	4	91/130	F/U	51	<151	N	Low %GV; low CMV DNA
I531Tb	14	96/159	F/U	4	207	Y	Below validated NGS threshold
L578Pc	5	7/27	LET	5	166	N	Low %GV; low CMV DNA
R629Cl	11	90/127	F/U	100	541	N

Abbreviations: CMV, a cytomegalovirus; DNA, deoxyribonucleic acid; F/U, follow-up (W14–24 posttransplant, when no study medication was administered); GV, genotypic variant; LET, letermovir; NGS, next-generation sequencing; N, no; PBO, placebo; RAV, resistance-associated variant; Tx, allogeneic stem cell transplant; Y, yes.

^aNineteen subjects (1–19) had 1 or more novel uncharacterized genotypic variant(s) in the CMV UL56 and/or UL89 gene. GVs that were not candidates for phenotypic analysis included common polymorphisms, substitutions that were previously characterized for susceptibility to LET (Supplementary Tables 1 and 2), and GVs that were also observed in 1 or more subjects who received placebo.

^bPhenotyping was performed even though the substitution did not meet the criteria for phenotyping.

^cAlthough the GV met the criteria for replicate testing, the analysis was either unsuccessful (polymerase chain reaction failed) or not performed (insufficient plasma).

^dSubject 3 had detectable CMV DNA (<151 copies/mL) on Day 1; subjects 1, 2, and 4 through 19 had undetectable CMV DNA on Day 1.

^eThe GV encoding the pUL56 R369T substitution was detected in 3 plasma samples collected from subject 3 (values are days after the subject received the last dose of LET, %NGS reads of GV, and CMV DNA copies/mL): Day 25, 12%, 7003 copies/mL; Day 82, 46%, 3878 copies/mL; Day 97, 12%, 11 178 copies/mL.

^fThe pUL56 S445del substitution was included when the pUL56 E485 + SNS445-447del substitution was phenotyped.

^gThe GV encoding the pUL56 SNS445-447del substitution was detected in 4 LET subjects (values are subject number, %NGS reads of GV, and CMV DNA copies/mL): 1, 70%, 1076 copies/mL; 8, 70%, 293 copies/mL; 10, 72%, <151 copies/mL; and 19, 75%, 304 copies/mL. The pUL56 SNS445-447del substitution was included when the pUL56 E485G + SNS445-447del substitution was phenotyped.

^hSubject 10: Days LET exposure/ Days Tx to GV = 97 of 138; GV detected after (F/U) W14 posttransplant. Subject 19: Days LET exposure/ Days Tx to GV = 18 of 46; GV detected before (LET) W14 posttransplant.

ⁱThe GV encoding the pUL56 E485G substitution was detected in 4 LET subjects (values are subject number, %NGS reads of GV, and CMV DNA copies/mL): 1, 99%, 1076 copies/mL; 8, 99%, 293 copies/mL; 10, 100%, <151 copies/mL; and 19, 100%, 304 copies/mL. The GV encoding the pUL56 E485G substitution was previously identified as a LET-susceptible GV [8].

^jGenotyping of the CMV UL56 gene from this plasma sample from subject 13 identified deletions of the codons for amino acids A786-A787-A788, and GVs encoding V790A and S791A pUL56 substitutions. Phenotypic analysis of the pUL56 AVS789-791del substitution was used to characterize the impact of these variants on susceptibility to LET.

^kThe GV encoding the pUL56 R826L substitution was seen in 5% of the NGS reads in the plasma sample from subject 16 but also in <5% of the NGS reads in many LET and PBO subjects. This GV was likely an NGS artifact.

^lBecause the pUL89 E77D and R629C substitutions are distal from known LET-resistant variants in pUL89 (N320H and D344E [14]), they were not analyzed by recombinant phenotyping.

Drug Susceptibility of Recombinant Cytomegalovirus Strains With pUL56 and pUL89 Substitutions

Recombinant phenotyping was performed to characterize the impact of selected pUL56 and pUL89 substitutions or deletions on susceptibility to anti-CMV drugs and viral growth. Twenty-seven variants (21 in pUL56, 6 in pUL89) were candidates for phenotyping (Figure 2, Table 2), and 18 were evaluated. Among these were 3 substitutions encoded by GVs that were below the validated NGS threshold of 5% (pUL56 M3V, 3%; pUL56 E237G, 4%; and pUL89 I531T, 4%) and 4 pUL56 GVs (V236M, C325W, SNS445-447del, and E485G) that were not novel: V236M and C325W are previously characterized LET RAVs [5, 7], SNS445-447del is a common polymorphism [17], and E485G was previously identified in a CMV isolate with confirmed susceptibility to LET [8].

The mean LET half-maximal effective concentration (EC₅₀) value and the fold-shift in mean LET EC₅₀ relative to a UL56/UL89 wild-type CMV reference strain were determined for 16 recombinant viruses (Table 3). The LET EC₅₀ values for the reference strains in these experiments were comparable to previously reported values [7, 15, 18]. Eleven recombinant CMV strains bearing novel uncharacterized substitutions and/or deletions had no significant shift in susceptibility to LET (EC₅₀ ratios ranged from 0.9 to 1.1); another strain (with a F626S substitution in pUL56) was more sensitive to LET, with an EC₅₀ that was one half the wild-type value. Cytomegalovirus recombinant strains with the pUL56 V236M, E237G, C325W, or R369T substitutions were less susceptible to LET, with EC₅₀ shifts of 50-, 13-, 8262-, and 52-fold, respectively. Susceptibility of 3 strains to the CMV DNA polymerase inhibitors ganciclovir (GCV), foscarnet (FOS), and cidofovir (CDV) was also measured, and there was a small but statistically significant shift (2.1-fold) in the GCV EC₅₀ for the mutant CMV strain with the pUL56 E237G substitution (Table 3). This strain was similar to wild-type in susceptibility to FOS (1.3-fold) and CDV (1.4-fold). There were no significant changes in susceptibility to GCV/FOS/CDV for mutant CMV strains with either the pUL56 V236M [7], C325W, or R369T substitution.

Table 3.

Drug Susceptibility of Recombinant CMV Strains With pUL56 and pUL89 Substitutions

	Letermovira		CMV DNA Polymerase Inhibitorsb
Substitution	Mean EC50 ± SD (nM)	Fold EC50 Shift Relative to Wild-Type (Letermovir)	Mean EC50 ± SD (Ganciclovir; µM)	Fold EC50 Shift Relative to Wild-Type (Ganciclovir)	Mean EC50 ± SD (Foscarnet; µM)	Fold EC50 Shift Relative to Wild-Type (Foscarnet)	Mean EC50 ± SD (Cidofovir; µM)	Fold EC50 Shift Relative to Wild-Type (Cidofovir)
pUL56 M3V	1.95 ± 0.2	1.0	N.D.	N.D.	N.D.	N.D.	N.D.	N.D.
pUL56 V236M	102 ± 12.7	50.2	N.D.	N.D.	N.D.	N.D.	N.D.	N.D.
pUL56 E237G	34.5 ± 3.89	13.0	2.55 ± 0.45	2.1	46 ± 11	1.3	0.35 ± 0.03	1.4
pUL56 S255L	2.35 ± 0.36	1.1	N.D.	N.D.	N.D.	N.D.	N.D.	N.D.
pUL56 C325W	18 420 ± 920	8262.0	1.64 ± 0.19	1.2	36.71 ± 4.19	1.1	0.33 ± 0.07	1
pUL56 R369T	107 ± 14.1	52.0	1.85 ± 0.42	1.5	40 ± 9.7	1.1	0.3 ± 0.04	1.2
pUL56 SNS445-447 deletion pluspUL56 E485Gc	2.02 ± 0.26	1.0	N.D.	N.D.	N.D.	N.D.	N.D.	N.D.
pUL56 E485G	1.88 ± 0.13	0.9	N.D.	N.D.	N.D.	N.D.	N.D.	N.D.
pUL56 Y575C	2.09 ± 0.49	0.9	N.D.	N.D.	N.D.	N.D.	N.D.	N.D.
pUL56 F626S	1.30 ± 0.33	0.5	N.D.	N.D.	N.D.	N.D.	N.D.	N.D.
pUL56 L726V	2.59 ± 0.37	1.0	N.D.	N.D.	N.D.	N.D.	N.D.	N.D.
pUL56 T775I	2.22 ± 0.39	1.0	N.D.	N.D.	N.D.	N.D.	N.D.	N.D.
pUL56 AVS789-791 deletion	2.58 ± 0.46	1.0	N.D.	N.D.	N.D.	N.D.	N.D.	N.D.
pUL56 V814A	2.65 ± 0.51	1.0	N.D.	N.D.	N.D.	N.D.	N.D.	N.D.
pUL56 R816W	2.00 ± 0.27	0.9	N.D.	N.D.	N.D.	N.D.	N.D.	N.D.
pUL89 I531T	2.02 ± 0.28	1	N.D.	N.D.	N.D.	N.D.	N.D.	N.D.

Abbreviations: CMV, cytomegalovirus; DNA, deoxyribonucleic acid; EC₅₀, 50% effective concentration; ND, not determined; SD, standard deviation.

^aLET EC₅₀ values for the UL56/UL89 wild-type recombinant strain in 3 experiments were 2.03 ± 0.26, 2.23 ± 0.3, and 2.57 ± 0.47 nM.

^bFor the CMV DNA polymerase inhibitors, wild-type strain EC₅₀ values in 2 experiments were 1.23 ± 0.21 and 1.34 ± 0.16 µM (ganciclovir), 35 ± 4.9 and 32.78 ± 4.97 µM (foscarnet), and 0.25 ± 0.03 and 0.33 ± 0.07 µM (cidofovir).

^cBecause GVs encoding the pUL56 SNS445-447del and E485G substitutions were detected together in 4 LET subjects (Table 2), a single CMV recombinant strain bearing both changes in UL56 was constructed and tested for susceptibility to LET. The pUL56 S445del substitution was included when the pUL56 E485G + SNS445-447del substitution was phenotyped.

Growth of Recombinant Cytomegalovirus With Novel pUL56/pUL89 Substitutions

Two of the novel uncharacterized variants (pUL56 F626S and pUL89 I531T) had a measurable growth defect. Using SEAP activity in the supernatants from >12 replicate cell cultures infected with either recombinant CMV strain as a marker for viral replication, mean relative light units were reduced by 14% and 20% (for pUL56 F626S and pUL89 I531T, respectively) on day 6 postinfection compared with wild-type reference strains (data not shown). The other recombinant CMV strains with novel uncharacterized variants in pUL56 or pUL89 grew as well as the wild-type reference strains.

Clinical Virology Profile of Study Subjects With Letermovir Resistance-Associated Variants

Details about the 3 subjects who had variants in UL56 that conferred reduced susceptibility to LET are presented in Figure 3. Additional genotyping of plasma collected before or after the CMVi visit was attempted as part of an effort to characterize the emergence or persistence of LET RAVs in these subjects.

Figure 3.

Profile of 3 subjects with letermovir (LET) resistance-associated variants (RAVs). (A) Subject 1; (B) Subject 2; (C) Subject 3. Bars along the top represent anti-cytomegalovirus (CMV) medications ( = LET, 240 mg oral; = ganciclovir, 650 mg intravenous (I.V.); = ganciclovir, 500 mg I.V.; = ganciclovir, 400 mg I.V.; = valganciclovir, 900 mg oral; = valganciclovir, 450 mg oral). The length of each bar indicates the duration of therapy; gaps reflect periods when no anti-CMV medication was administered. Horizontal lines indicate the limit of CMV deoxyribonucleic acid (DNA) detection (dashed line, 91 copies/mL) and quantitation (dotted line, 151 copies/mL). Symbols reflect the outcome from CMV UL56 genotypic analysis: genotyping was either attempted and successful (); attempted and unsuccessful (); or not attempted (). Text labels adjacent to selected symbols indicate specific pUL56 LET RAVs; the percentage of next-generation sequencing reads in the CMV DNA sequence that encodes the mutation(s) is shown.

The CMV GV encoding the pUL56 V236M substitution was found in DNA obtained from subject 1 (Figure 3A) while the subject was receiving LET prophylaxis (62 days after the first dose was administered). Plasma from subject 1 had 1076 copies of CMV DNA/mL when this GV was detected in 99% of the NGS reads. No CMV DNA was detected on study day 1; genotyping of plasma collected on other days was not successful (Figure 3A). Subject 1 did not receive LET as scheduled on study days 2, 3, 4, 8, and 9— all other doses were administered until LET was discontinued, and PET with valganciclovir (VGCV) was initiated on study day 62. The pUL56 V236M substitution was previously identified in a Ph2b study of LET [11] and in cell-culture selections for CMV mutants with reduced susceptibility to LET—in vitro, this mutation confers a 22- to 50-fold increase in the LET EC₅₀ [7, 11, 15, 19].

The GV encoding the pUL56 E237G substitution was identified in subject 2 on study day 42. Before this GV was detected, the subject experienced CS-CMVi and discontinued LET prophylaxis on study day 10 (Figure 3B). There were <151 copies of CMV DNA/mL in the plasma sample collected on study day 42, and E237G was detected in 4% of the CMV UL56 NGS reads. Subject 2 received VGCV therapy from study days 42 to 65, and plasma CMV DNA levels were either <151 copies/mL or undetectable through week 48 posttransplant. Prior cell-culture selections for LET-resistant mutants had identified a different substitution at the same position (pUL56 E237D) that confers a 21-fold reduction in susceptibility to LET [15].

Two substitutions associated with reduced susceptibility to LET were identified in subject 3 (Figure 3C). Cytomegalovirus DNA was detected but not quantifiable (<151 copies/mL) when this subject began LET prophylaxis. The UL56 GV encoding C325W was first detected in 99% of the NGS reads in a plasma sample with 969 copies of CMV DNA/mL collected on study day 39 (during LET prophylaxis). The CMV DNAemia reached a peak of 28096 copies/mL on study day 46, at which point LET prophylaxis was discontinued and PET was initiated. The UL56 GV encoding C325W was still detected 94 days after subject 3 received the last dose of LET; however, the proportion of CMV DNA sequences with this LET RAV decreased in successive episodes of CMV DNAemia. The mutations encoding pUL56 C325W and 3 other pUL56 C325 variants (C325F/R/Y) have been described in recent publications [5, 7, 15]. The second GV in subject 3, encoding pUL56 R369T, was first detected in 12% of the NGS reads in a sample with 7003 CMV DNA copies/mL on study day 71 (25 days after discontinuation of LET prophylaxis). It was also detected on study days 128 (46% NGS reads, 3878 copies/mL) and 143 (12% NGS reads, 11 178 copies/mL), but not in samples collected on study days 39, 44, 46, 49, 58, or 65. Other substitutions at pUL56 R369 (R369G/M/S) that confer a decrease in susceptibility to LET in cell culture have been described previously [7].

DISCUSSION

In this analysis of samples from the Phase 3 study of LET prophylaxis, almost all of the novel uncharacterized GVs in the CMV UL56 and UL89 genes encoding DNA terminase subunits that were found among LET subjects were polymorphisms with no impact on susceptibility to LET. UL51 genotyping was not performed because there was no evidence when the resistance analysis plan was developed that a GV in UL51 could affect LET susceptibility. A single UL51 mutation has recently been reported in only 2 of many experiments where CMV was exposed to LET in cell culture [20].

Of 373 subjects who received LET prophylaxis in the Phase 3 trial, 79 of whom met the primary endpoint, 4 different GVs that impact susceptibility to LET were identified in 3 subjects among 50 that had genotyping data available. These 3 subjects were also among the 18 successfully genotyped out of 40 who developed CS-CMVi while on LET prophylaxis (Figure 1, Supplementary Table 4). Lack of adherence to the dosing regimen (subject 1) or detectable CMV DNA in plasma at the start of therapy (subject 3) may have contributed to breakthrough of CMV strains with the pUL56 V236M or C325W substitutions, respectively. Mutations encoding the pUL56 E237G (subject 2) and R369T (subject 3) substitutions were first detected in a minority of the CMV UL56 sequences (4% and 12%, respectively), in samples that were collected >3 weeks after LET prophylaxis was discontinued, and the subjects had received preemptive therapy with GCV/VGCV. These 4 substitutions shift in vitro susceptibility to LET to widely different degrees (V236M, 50-fold; E237G, 13-fold; C325W, 8262-fold; and R369T, 52-fold).

An important genotyping quality control initiative of this study was to show that NGS reads of minority subpopulations in low copy number specimens are a risk for false-positive detection of mutations (Supplementary Table 5). Assessment of the significance of GVs in clinical trials should include evidence of their reproducible detection, especially for minority subpopulations.

Because LET RAVs were infrequent in this trial, additional clinical experience is needed to better understand the nature and incidence of LET resistance. Clinically significant drug resistance is expected to be infrequent in a prophylaxis trial where subjects are taken off study drug soon after viral breakthrough, and the present data do not directly translate to LET treatment situations. Recent reports have identified UL56 mutations encoding C325F or C325Y in recipients of hematopoietic stem cell or solid organ transplants who received LET as salvage therapy [21–23]. For a patient who received LET prophylaxis and experienced breakthrough CMV with evidence of a UL56 mutation encoding C325F, there was no information about CMV DNAemia at the time LET prophylaxis was initiated 20 days after transplantation [24]. The pUL56 R369T substitution was not detected in subject 3 until 25 days after discontinuation of LET prophylaxis, but other pUL56 R369 substitutions (G/M/S) have been found in CMV mutants exposed to LET in cell-culture resistance selections [7], and both pUL56 R369G and R369S have been reported in clinical specimens obtained from patients who have received LET [25]. The pUL56 V236M substitution was seen in 1 subject enrolled in the Phase 2b and 1 subject in the Phase 3 trial of LET, but both subjects experienced LET-dosing interruptions during the first 2 weeks of prophylaxis ([10, 11]; Figure 3A; CMD, 2017, unpublished results). Although V236M has also been detected in recent diagnostic genotypic testing, clinical details for these patients have not been reported [25]. Finally, the clinical significance of the other LET RAV from the Phase 3 trial (an E237G substitution in pUL56) is uncertain—this was a minority variant (only 4% of the CMV UL56 sequence) detected in a sample with <151 copies of CMV DNA/mL 32 days after the last dose of LET prophylaxis had been administered. The pUL56 E237G substitution confers a 13-fold shift in the LET EC₅₀ and a 2-fold shift in the GCV EC₅₀ in cell culture. The subject with this variant responded to GCV/VGCV therapy, and no other instances of this mutation in clinical use of LET have been reported.

Growth of recombinant CMV strains bearing the UL56 mutations encoding V236M, E237G, C325W, or R369T substitutions was not greatly attenuated in cell culture [5, 7]. The impact of these mutations on fitness in vivo is not known because the 3 study subjects with these LET RAVs were given preemptive therapy with GCV and/or VGCV, and their CMVis were effectively reduced or eliminated. In subject 3, the proportion of CMV sequences encoding the pUL56 C325W RAV was lower in several samples collected weeks after LET dosing was discontinued (Figure 3C); although this change may reflect decreased fitness of the mutant virus, it could have resulted from superinfection or reactivation of a CMV strain that is wild-type at UL56 codon 325.

More than 30 LET RAVs have been identified in cell-culture selections for resistance, with mutations that map predominantly to the CMV UL56 gene [5, 14, 20]. Many of these mutations conferred only a modest (<10-fold) change in the LET EC₅₀. In contrast, the CMV mutations identified in patients who have received LET and had CMV breakthrough are associated with some of the largest LET EC₅₀ increases relative to a wild-type reference strain (Table 3, Supplementary Table 1). These differences between mutants observed in cell culture and those detected in patients are not unexpected, since the in vitro resistance selections involved serial passage of CMV stocks with incrementally increasing concentrations of LET, starting near the EC₅₀, and CMV strains in patients who receive the recommended dose of LET are exposed to initial and sustained concentrations of LET that are many multiples above the EC₅₀.

CONCLUSIONS

Overall, CMV breakthrough associated with LET resistance was infrequent in this Phase 3 trial of LET prophylaxis in HSCT recipients. The few LET RAVs that were detected mapped to the CMV UL56 gene at positions associated with significantly reduced susceptibility to LET based on resistance selections in cell-culture models of CMVi.

Supplementary Data

Supplementary materials are available at The Journal of Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes

Acknowledgments. We thank Christopher Assaid, Cyrus Badshah, Joan Butterton, Jill Cairnes, Nancy Castro, Lei Chen, Michael Citron, Amy Espeseth, Jessica Flynn, Jay Grobler, Nick Kartsonis, Laurie MacDonald, Yoshihiko Murata, Kendra Sarratt, Anita Shaw, Nicole Stauffer, Walter Straus, Bo Wei, Carolee Welebob, and Wendy Yeh from Merck & Co., Inc., (Kenilworth, NJ) for technical assistance and/or helpful comments on the manuscript; Monique R. K. S. Bansraj, Esmeralda D. C. Bosman, Ilona C. E. van Haaften, Angela S. Hoogenboom, Michiel T. Weber, and Leontine I. van der Well from DDL for technical assistance; and Ibironke Addy, Peter Lischka, Helga Rübsamen-Schaeff, and Holger Zimmerman from AiCuris for helpful discussions. Ronald J. Ercolani and L. Elizabeth Satterwhite from Portland VA Medical Center for technical assistance.

Financial support. Funding for this research was provided by Merck Sharp & Dohme Corp., a subsidiary of Merck & Co., Inc., (Kenilworth, NJ) and National Institutes of Health Grant AI116635 (to S. C.) and supported by use of Department of Veterans Affairs facilities and resources for recombinant phenotyping.

Potential conflicts of interest. C. M. D., R. B., D. H., R. L., D. L., M. M., D. N., V. T., H. W., and J. S. are employees of Merck Sharp & Dohme Corp., a subsidiary of Merck & Co., Inc., (Kenilworth, NJ) and may own stock/stock options in Merck & Co., Inc., (Kenilworth, NJ). L.-J. v. D. reports a financial relationship with DDL Diagnostic Laboratory, outside the submitted work. S. C. reports a Cooperative Research & Development Agreement from Merck & Co., Inc.; a grant from the National Institutes of Health; and a Cooperative Research and Development Agreement from Shire, a Takeda company, outside the submitted work. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

Presented in part: BMT Tandem, 2018, Salt Lake City, Utah; European Congress of Clinical Microbiology and Infectious Diseases, 2018, Madrid, Spain; International Conference on Antiviral Research, 2018, Porto, Portugal; International Conference on Antiviral Research, 2019, Baltimore, Maryland.