Differentiated Levels of Ganciclovir Resistance Conferred by Mutations at Codons 591 to 603 of the Cytomegalovirus UL97 Kinase Gene

ABSTRACT

Diagnostic mutations in the cytomegalovirus UL97 kinase gene are used to assess the level of associated ganciclovir resistance and therapeutic options. The best-known mutations at codons 460, 520, or 591 to 607 individually confer 5- to 10-fold-decreased ganciclovir susceptibility, except that a 3-fold decrease occurs in the case of the amino acid substitution C592G. Less common point and in-frame deletion mutations at codons 591 to 603 remain incompletely characterized. The ganciclovir susceptibilities of 17 mutants in this codon range were evaluated by use of the same recombinant phenotyping system and extensive assay replicates in two types of cell cultures. Amino acid substitutions K599E and T601M conferred no ganciclovir resistance, while A591V conferred 3.8-fold-decreased susceptibility. In-frame deletions of three or more codons conferred at least 8-fold-increased ganciclovir resistance, while the level of resistance conferred by one- or two-codon deletions varied from 4- to 10-fold, depending on their location. Measured levels of ganciclovir resistance were closely comparable when assays were performed in either fibroblasts or modified retinal epithelial cells. The significant revision of a few previously published resistance phenotypes and the new data strengthen the interpretation of genotypic testing for cytomegalovirus drug resistance.

INTRODUCTION

Prolonged treatment of human cytomegalovirus (CMV) infection with ganciclovir (GCV) or its oral prodrug, valganciclovir, may result in antiviral drug resistance when viral replication is incompletely suppressed. Without the routine availability of susceptibility testing on CMV culture isolates, the laboratory diagnosis of drug resistance is dependent on the direct detection of characteristic viral mutations in clinical specimens (1, 2). GCV resistance mutations usually develop first in the UL97 kinase gene involved in the initial phosphorylation of GCV that is necessary for its antiviral action. One of seven UL97 mutations (M460V/I, H520Q, C592G, A594V, L595S, and C603W) is initially detected in >80% of cases of GCV resistance in clinical practice, with most of the remainder consisting of less common amino acid substitutions or in-frame deletions at codons 590 to 607 or of mutations in the UL54 DNA polymerase gene (2). The canonical UL97 mutations confer 5- to 10-fold increases in the GCV concentration required for 50% viral growth inhibition (50% effective concentration [EC₅₀]), except that C592G confers a 3-fold increase. Given the limited alternative treatment options, mutations conferring lower-grade resistance may be amenable to GCV dose escalation, according to consensus treatment guidelines (3). Therefore, accurate calibration of the levels of resistance conferred by specific mutations is desirable for treatment planning purposes since the benefit of GCV dose escalation is expected to decrease as the level of resistance increases.

Although individually uncommon, UL97 in-frame codon deletions collectively occur often enough to raise questions about their associated levels of GCV resistance. Several deletions of three or more codons in the codon range 590 to 607 have been shown to confer >5-fold EC₅₀ increases (4–7), while other mutations reportedly confer EC₅₀ changes of <2-fold (5, 8), typically regarded as insignificant. Comparisons of EC₅₀ values across publications are complicated by differences in assay methodologies, and some mutations lack phenotypic characterization altogether. The current objective was to compare, using cloned viral strains and consistent methodology, the GCV susceptibility phenotypes of selected UL97 mutations in codons 591 to 603 that have been encountered in clinical specimens, including point mutations and in-frame deletions of one or more codons. The selected set included mutations with established phenotypes, mutations without a published resistance phenotype, and mutations with inconclusive susceptibility phenotypes. A secondary objective was to document the performance of an alternative cell culture system for reporter-based susceptibility assays.

RESULTS

The 16 new bacterial artificial chromosome (BAC)-cloned mutant viral strains were authenticated for the intended UL97 sequence in viral DNA amplified from infected cell extracts. Table 1 shows the genotypes and GCV susceptibility phenotypes of the baseline and mutant viruses tested in the two cell types (human foreskin fibroblasts [HFFs] and ARPEp), along with the range of previously reported GCV EC₅₀ ratios of the mutant to baseline mean value, where published. All mean EC₅₀ values were determined with sufficient replicates such that the relative standard error was <10%, thus making the 95% confidence interval of the EC₅₀s to be ±20% or less, under the prevailing culture conditions.

TABLE 1.

Genotypes and GCV susceptibility phenotypes of recombinant CMV strains in two cell types

Strain group and serial no.	UL97 genotypea	H587Yb	GCV assay in HFFs			GCV assay in ARPEp cells			Published ratio(s)	Reference(s)
			Mean EC50 (SD [μM])	No. of replicates	EC50 ratio (n-fold)c	Mean EC50 (SD [μM])	No. of replicates	EC50 ratio (n-fold)c
Control strains
3261	WT		1.1 (0.29)	43		1.2 (0.14)	18
3338	WT	X	1.1 (0.31)	49		1.2 (0.27)	34
3259	C592G		3.3 (0.86)	60	3.0	3.4 (0.62)	34	2.9	2.6–3.6	5, 11, 24
3252	A594V		7.0 (1.2)	33	6.4	7.7 (1.6)	38	6.6	6.4–10.1	7, 11, 39, 40
New recombinant strains
4156	A591V		4.2 (0.99)	14	3.8	4.5 (0.9)	11	3.8	1.3–1.8	5, 13
4211	A591V (BD1)		4.0 (1.1)	12	3.7	4.4 (1.2)	13	3.7
4127	591del4		10.6 (2.0)	20	9.7	12.4 (1.7)	14	10.6	6.7	7
4130	595del	X	9.7 (2.8)	22	8.5	9.5 (1.4)	10	8.0	6.7–13	7, 8
4145	595del9		10.5 (2.5)	24	9.6	10.5 (1.6)	10	9.0	8.4	4
3450	596del	X	4.6 (1.1)	18	4.0	5.6 (0.34)	10	4.7		15
4120	597del2	X	10.1 (2.1)	23	8.9	11.0 (1.3)	12	9.3
4124	597del3	X	9.9 (2.6)	24	8.7	10.9 (1.7)	11	9.1	3.8	13
3444	K599E	X	1.6 (0.39)	10	1.4	1.2 (0.20)	10	1.0		15
4146	599del	X	7.1 (2.0)	18	6.2	6.4 (1.7)	11	5.4
4126	600del		6.7 (1.8)	24	6.1	6.9 (2.0)	9	5.9	1.9	5
4122	600del2		5.4 (1.4)	21	4.9	4.6 (0.46)	12	4.0
3568	T601M	X	1.1 (0.35)	16	0.9	1.2 (0.23)	12	1.0		16
4150	601del	X	6.4 (1.6)	15	5.6	7.2 (1.7)	10	6.1		35
4131	601del2	X	6.1 (1.5)	23	5.3	6.1 (0.58)	12	5.1		36
4227	601del3		10.6 (1.2)	10	9.6	11.1 (2.3)	14	9.5	17	6

^aCodon deletion or amino acid substitution. The baseline BAC clone is BA1, except for strain 4211 (BD1; see text). WT, wild type (none).

^bThe presence of the H587Y amino acid substitution (with no impact on GCV susceptibility) is indicated (X).

^cRatio of mean EC₅₀ to that of the baseline strain. Values more than double those of previously published data are in boldface. All ratios of >2 are statistically significant (P < 1 × 10⁻⁷, Student's t test, for the mutant versus baseline mean EC₅₀ value).

The clonal ARPEp cell line overexpressing platelet-derived growth factor receptor alpha (PDGFRα) displayed growth characteristics similar to those of unmodified ARPE-19 cells, and no morphological differences were observed. Permissiveness of the cells was tested by infection with a green fluorescent protein (GFP)-expressing strain AD169 (9) virus stock, and almost all cells showed fluorescence thereafter. Several passages in the absence of puromycin did not alter these properties. Secreted alkaline phosphatase (SEAP) reporter strains used for drug susceptibility phenotyping also generated comparable supernatant SEAP activity levels in both HFF and ARPEp cultures, as observed previously in ARPE-19α cells (10). For the wild-type (strains 3261 and 3338) assay replicates listed in Table 1, mean SEAP activities with no added drug were 396 relative light units (RLU) at 24 h and 3.9 × 10⁵ RLU at 6 days (1,000-fold increase) in ARPEp cells and 724 RLU at 24 h and 4.0 × 10⁵ RLU at 6 days (600-fold increase) in HFFs.

The 16 newly constructed recombinant strains listed in Table 1 all grew normally, consistent with published information that the entire range of codons 591 to 607 can be deleted without measurable impact on viral growth (5). For these mutants, the mean SEAP values for each strain (in ARPEp cells) without added drug averaged 380 RLU at 24 h (range, 216 to 536 RLU) and 4.0 × 10⁵ RLU at 6 days (range, 2.7 × 10⁵ to 5.9 × 10⁵ RLU), an 860- to 1,500-fold increase per strain, similar to the corresponding wild-type values listed above.

Control baseline and mutant strains containing amino acid substitutions C592G and A594V had GCV EC₅₀s that calibrated well to previously published ranges, irrespective of the cell type used for assay. The baseline GCV EC₅₀ of 1.1 μM in HFFs matches previous data obtained from the same reporter-based phenotyping system since its inception (10–12). Unlike nonclonal ARPE-19α cells, where the baseline GCV EC₅₀ was 2.4 μM (10), the new ARPEp cells gave a GCV EC₅₀ of 1.2 μM, almost identical to that in standard HFF cultures. The routinely used resistant control strain containing C592G gave the expected 3-fold increase in GCV EC₅₀, while another resistant control strain, A594V, gave an ∼6.5-fold increase in GCV EC₅₀, which is within the published range.

Single codon deletions tested included 595, 596, 599, 600, and 601. Among these, the most resistant was a deletion of codon 595 (595del), with a GCV EC₅₀ ratio of 8 to 8.5, whereas deletion of the adjacent codon 596 conferred a lesser EC₅₀ ratio of 4 to 4.7, and single codon deletions at codons 599 to 601 gave EC₅₀ ratios of about 6 (Table 1). In the two cases where phenotypes have been published previously, the 595del data are consistent (ratios of 6.7 to 13 [7, 8]), but the EC₅₀ ratio for 600del is more than twice that previously reported (5).

Multiple codon deletions tested included those starting at codon 591, 595, 597, 600, or 601. All deletions of three or more codons (starting at 591, 595, 597, and 601) conferred greater than 8-fold increases in GCV EC₅₀ ratios, comparable with available prior data except for that for deletion of three codons starting at 597 (597del3), which was published as 3.8-fold (13). Deletions of two codons starting at 597, 600, or 601 gave an EC₅₀ ratio of ∼9-fold for 597del2 and of 4- to 5.3-fold for the others. No previous data exist for these mutants.

Newly phenotyped amino acid substitutions K599E and T601M were found to confer no GCV resistance as single mutants. Two independent A591V recombinants were constructed, one derived from baseline clone BA1, as for all other strains in Table 1, and a second one based on clone BD1, which differs from BA1 in the restoration of missing US7 to US16 sequences and compensatory deletion of the extra copy of genes RL1 through RL13 (14). The BD1-derived baseline strain T4198 has the same GCV EC₅₀ as the BA1-derived strain T3261 (Table 1), whether assayed in HFFs (1.1 μM) (10) or in ARPEp cells (1.2 μM ± 0.31; 11 replicates). Both BA1- and BD1-derived A591V mutant strains showed GCV EC₅₀ ratios of 3.7 to 3.8 in both cell types in over 40 total replicates of testing, making this mutant equivalent to the well-known adjacent C592G as a marker for low-grade GCV resistance, which represents a significant difference from previously reported EC₅₀ ratios of 1.3 or 1.8 for A591V (5, 13).

DISCUSSION

Extensive replicates of testing in a well-documented phenotyping system were used to calibrate the levels of GCV resistance conferred by a range of less common UL97 mutations. In codons 591 to 603, several in-frame deletions of three or more codons conferred >8-fold-increased GCV EC₅₀s, while deletions of one or two codons had a more variable effect, ranging from 4- to 9-fold increases. Three mutations showed a significantly more GCV-resistant phenotype than previously published, with implications for clinical interpretation and treatment algorithms.

Findings with K599E and T601M confirm that not all amino acid substitutions in codon 590 to 607 range confer GCV resistance (15, 16). These findings are the same as those previously observed for substitutions E596D, N597D, K599R, L600I, and D605E (11, 12, 17, 18) and unlike those for E596G/Y and K599T, which were reported to confer GCV resistance (5, 17, 19). Except for the common UL97 sequence polymorphism D605E, all of these substitutions are rarely encountered (2).

A591V has not been documented as a polymorphism in baseline UL97 sequences from untreated subjects but has occasionally been detected since the early use of GCV (5, 20, 21), including the emergence of A591V as a new mutation after GCV therapy (22). The first published phenotype was from plaque reduction assays showing a GCV EC₅₀ of 7.9 μM, which was 1.3-fold increased over control wild-type EC₅₀s (5) but above the cutoff of 6 μM originally proposed for classifying GCV resistance (23). Based on the EC₅₀ ratio, A591V has been omitted from lists of GCV resistance mutations (3, 13). More recently, it was reported that A591V alone conferred a borderline GCV EC₅₀ ratio of 1.8, but it markedly increased the GCV EC₅₀ when combined with UL54 DNA polymerase mutations (13). Many replicates of testing of two independent A591V mutants simultaneously with the well-known UL97 C592G now indicate that the two substitutions should be regarded as conferring similar low-grade GCV resistance levels, and the combined effect of A591V and UL54 mutations (13) should be seen as analogous to the effect documented with combined C592G and UL54 mutations (12, 24).

The single- and multiple-codon in-frame deletion mutations phenotyped in this study do not include all the various deletions that have been encountered but offer guidance for interpretation of those yet to be phenotyped, according to the location and number of deleted codons. The two cases (600del and 597del3) where the current phenotypes are more GCV resistant (Table 1) than previously reported by plaque reduction assays (5, 13) reflect the same technical considerations as those for A591V. There are well-documented difficulties in the standardization of plaque assays across studies because of limited replicates of testing (often no more than three), compounded by variable operator expertise and culture conditions leading to fluctuating baseline and mutant EC₅₀ values that greatly affect calculated ratios (25).

A secondary objective of this study was to compare the assay performance of a cloned retinal epithelial cell line overexpressing platelet-derived growth factor receptor alpha (PDGFRα). Fortuitously, these ARPEp cells gave absolute and relative EC₅₀ values equivalent to those obtained in HFFs even though GCV EC₅₀ values are expected to vary according to cell culture conditions affecting GCV activity, such as its final phosphorylation by cellular kinases and competition with intracellular dGTP concentrations. Because of robust and sustained proliferation less affected by senescence, full permissiveness for AD169-derived strains, and equivalency of EC₅₀ determinations, the ARPEp cells are a technically advantageous replacement for HFFs for reporter-based antiviral susceptibility phenotyping.

The clinical purpose of having accurate GCV resistance phenotypes is to assess the associated degree of impairment of antiviral efficacy. UL97 mutations alone, even those that knock out kinase activity entirely (26), do not confer more than moderate ∼15-fold elevations in the GCV EC₅₀, probably because a minimal level of GCV phosphorylation is accomplished by cellular enzymes alone. Some UL97 mutations confer low-level (<5-fold) increased GCV EC₅₀s, most notably C592G and now also A591V, presumably reflecting the extent to which they decrease GCV phosphorylation (27). UL54 polymerase mutations provide additional biochemical pathway(s) of GCV resistance and when combined with UL97 mutations lead to high-level (>15-fold, often >20-fold) GCV resistance (12, 13, 24, 28). GCV dose escalation has been proposed as an interim therapeutic adjustment in cases of lower-grade (29) and possibly moderate GCV resistance (30). There is no proven connection between GCV EC₅₀ values obtained in the present assay system and the therapeutic response to the usual peak and trough plasma GCV levels of ∼25 μM and ∼6 μM with standard dosing (31, 32). However, these plasma levels are roughly at the boundaries of the EC₅₀s categorized herein as high, moderate, and low levels. GCV dose escalation may be appropriate where high-level GCV resistance (combined UL97 and UL54 mutation) is not present and is more likely to succeed where the level of GCV resistance is low. Because of the major influence of host factors on outcomes, this treatment approach should be carefully monitored for toxicity, viral loads, disease progression, and the emergence of additional resistance mutations.

MATERIALS AND METHODS

Mutations selected for phenotyping.

All mutations selected for phenotyping have been detected in one or more clinical specimens. UL97 mutations selected as controls were those with amino acid substitutions C592G and A594V (11). Previously phenotyped mutations included A591V (5, 13), 591del4 (in-frame deletion of four consecutive codons beginning with 591 [7, 33]), 595del (8), 595del9 (4), 597del3 (13), 600del (5), and 601del3 (6). Newly phenotyped mutations included 596del (15), 597del2, K599E (15), 599del, 600del2 (34), T601M (16), 601del (35), and 601del2 (36). The deletion mutations were selected to represent single- and multiple-codon deletions starting from different parts of the codon range 591 to 603.

Construction of mutant recombinant CMV strains.

Baseline CMV strain AD169 was cloned as bacterial artificial chromosome (BAC) BA1 containing a secreted alkaline phosphatase (SEAP) reporter gene for viral quantitation (11). Control BAC clones and derived strains representing wild-type baseline virus and UL97 mutants C592G and A594V were the same as reported previously (11). For the remaining 15 mutations, including those previously phenotyped in this laboratory using nonclonal strains, new mutant BAC clones and derived strains were constructed as previously described for the control strains (11). A second A591V mutant was derived using the same methods from a separate BD1 clone of strain AD169 (10, 14). Some baseline and mutant strains also contain the UL97 amino acid variation H587Y that has been shown not to affect the GCV susceptibility phenotype (12).

Cell cultures.

Standard cell cultures for viral propagation, cell-free stock preparation, and yield reduction assays consisted of human foreskin fibroblasts (HFFs), as used extensively for reporter-based drug resistance phenotyping in this laboratory (10, 12, 37). Recent experience with ARPE-19α cells, which are retinal epithelial cells transduced to overexpress the platelet-derived growth factor receptor alpha (PDGFRα), suggested that this cell type offers full permissiveness for CMV laboratory strain AD169 (10, 38). Because the ARPE-19α cells previously used (10, 38) were not clonal, a new cell line, ARPEp, was created by transducing normal ARPE-19 cells with a vector overexpressing PDGFRα. The vector was constructed by inserting the PDGFRα coding sequence from cDNA clone pSPORT6 (clone 5205969; GE Dharmacon) into the retrovirus plasmid pCMMP-MCS-IRES-Puro (plasmid 36592; Addgene) using enzymes AgeI and NotI. To generate pseudotyped retrovirus particles, the vector was cotransfected into 293T cells with a plasmid expressing retroviral Gag/Pol/Rev and a plasmid expressing vesicular stomatitis virus G protein (VSV-G), using Lipofectamine 2000 reagent (Invitrogen). Cell culture supernatants containing retrovirus particles were collected at 48 h posttransfection and used to transduce ARPE-19 cells (ATCC CRL-2302). After transduction, the cells were serially diluted in 96-well culture plates and cultured in the presence of puromycin, and a single-cell clone was isolated by limiting dilution, expanded, and stored for future use. These ARPEp cells were grown in Dulbecco's minimal essential medium with 4.5 g/liter glucose and 8% to 10% fetal bovine serum, the latter reduced to 3% after inoculation with infectious CMV.

Phenotypic assays.

Assays for GCV EC₅₀s were performed by reporter-based yield reduction assays as described previously (11, 12), using culture supernatant SEAP activity as a measure of viral growth. Twenty-four-well plates seeded with HFFs or ARPEp cells were used 2 to 4 days after cells reached confluence. Rows of six wells were inoculated with 0.3 ml of the cell-free viral stock to be tested at a dilution needed to achieve a multiplicity of infection (MOI) of 0.02, as calibrated by titration of plaque counts against culture supernatant SEAP activity at 24 h postinoculation (12). After 90 min, inocula were removed from each row and replaced with maintenance medium containing no drug (control well) or GCV at 5 concentrations increasing 2-fold to about 4 times the expected EC₅₀ value. Culture supernatant SEAP activity in the no-drug well was assayed at 24 h and 6 to 7 days postinoculation using a chemiluminescent substrate, as described previously (12). The SEAP activity from the no-drug well at the later time point (6 to 7 days) was used to quantify the viral growth from which a 50% reduction (EC₅₀) was determined by fitting the readings from the various drug concentrations to an exponential curve. Assays were set up with wild-type and well-known GCV-resistant control strains (C592G and/or A594V) in the same batches of cell cultures. Quality control criteria for valid assays (11, 12) included the following: 24-h SEAP activity compatible with an MOI of 0.02 ± 0.01, final supernatant SEAP activity without drug of 2 × 10⁵ relative light units (RLU), curve fit of the yield reduction with a correlation coefficient (R²) of >0.89, and EC₅₀ value within the range of the first four drug concentrations used. Additionally, EC₅₀ values from simultaneously tested control strains had to be in a range of 0.6 to 1.6 times their known values. At least nine replicates of testing over four separate setup dates were performed for each strain and cell type for improved statistical significance.