Nonnucleoside Pyrrolopyrimidines with a Unique Mechanism of Action against Human Cytomegalovirus

Abstract

Based upon a prior study which evaluated a series of nonnucleoside pyrrolo[2,3-d]pyrimidines as inhibitors of human cytomegalovirus (HCMV), we have selected three active analogs for detailed study. In an HCMV plaque-reduction assay, compounds 828, 951, and 1028 had 50% inhibitory concentrations (IC₅₀s) of 0.4 to 1.0 μM. Similar results were obtained when 828 and 951 were examined by HCMV enzyme-linked immunosorbent assay (IC₅₀s = 1.9 and 0.4 μM, respectively) and when 828 was tested in a viral DNA-DNA hybridization assay (IC₅₀ = 1.3 μM). In yield-reduction assays with a low multiplicity of infection (MOI), all three compounds caused multiple log₁₀ reductions in virus titer, and the activities of these compounds were comparable to the activity of ganciclovir (GCV; IC₉₀ = 0.2 μM). In contrast to the reduction of viral titers by GCV, the reduction of viral titers by 828, 951, and 1028 decreased with increasing MOI. Cytotoxicity in human foreskin fibroblasts and KB cells ranged from 32 to >100 μM. In addition, 828 (the only compound tested) was less toxic against human bone marrow progenitor cells than GCV. Time-of-addition and time-of-removal studies established that the three pyrrolopyrimidines inhibited HCMV replication before GCV had an effect on viral DNA synthesis but after viral adsorption. Compound 828 was equally effective against GCV-sensitive and GCV-resistant HCMV clinical isolates. Combination studies with 828 and GCV showed that the effects of the two compounds on HCMV were additive but not synergistic. Taken together, the data indicate that these pyrrolopyrimidines target a viral protein that is required in an MOI-dependent manner and that is expressed early in the HCMV replication cycle.

Human cytomegalovirus (HCMV) is relatively benign in healthy individuals but can be debilitating or fatal to immunocompromised individuals such as organ transplant recipients (39, 43), neonates (7), and people infected with human immunodeficiency virus (HIV) (24, 30). Furthermore, HCMV stimulates HIV gene expression, thereby implicating HCMV as a cofactor in the progression of HIV infection (11, 48). Hence, the ability to control replication of HCMV may be important in suppressing the proliferation of HIV in afflicted individuals. In addition, intrauterine HCMV infection is the leading infectious cause of central nervous system maldevelopment in children (7).

Currently, three nucleoside analogs are available in the United States as drugs for the treatment of HCMV infections: ganciclovir (GCV) (1, 9, 41), foscarnet (PFA) (1, 8, 29), and cidofovir (CDV) (12, 18). In addition, the antisense oligonucleotide fomivirsen has been approved for direct treatment of HCMV retinitis (23). Clinically, GCV can cause suppression of bone marrow proliferation (9, 41), and the use of PFA and CDV is similarly dose limiting because of drug-induced nephrotoxicity (13, 18, 19). Furthermore, all three nucleoside analogs have poor oral bioavailabilities, and strains of HCMV resistant to the drugs are emerging (5, 15, 20, 44). Hence, the problems associated with the use of GCV, PFA, and CDV for the treatment of HCMV infections make the development of more potent and less toxic drugs for the treatment of HCMV infections a high priority.

As part of our ongoing research involving pyrrolo[2,3-d]pyrimidines as potential antiviral agents (3, 17, 34, 40, 45–47), we have described the synthesis and activities against HCMV of a number of nonnucleoside derivatives related to toyocamycin, sangivamycin, and thiosangivamycin (35). The studies in this area have been expanded, and a description of the structure-activity relationships of this class of compounds has been reported (37, 38). From this extensive study we have selected for further investigation one aliphatic analog from this series (4-amino - 7 - [(2 - methoxyethoxy)methyl]pyrrolo[2,3 - d]pyrimidine - 5-thiocarboxamide; compound 828) and two aromatic analogs, 4 - amino - 7 - (4 - methoxybenzyl)pyrrolo[2,3 - d]pyrimidine - 5 - thiocarboxamide (compound 951) and 5-cyano-4,6-diamino-7-(p-methylbenzyl)-pyrrolo[2,3-d]pyrimidine (compound 1028) (Fig. 1). This report describes the activities of these compounds against HCMV as well as their cytotoxicities for uninfected cells and examines their modes of action.

FIG. 1.

Structures of 4-amino-7-[(2-methoxyethoxy)methyl]pyrrolo[2,3-d]-pyrimidine-5-thiocarboxamide (compound 828), 4-amino-7-(4-methoxybenzyl)pyrrolo[2,3-d]pyrimidine-5-thiocarboxamide (compound 951), 5-cyano-4,6-diamino-7-(p-methylbenzyl)pyrrolo[2,3-d]pyrimidine (compound 1028), and GCV.

MATERIALS AND METHODS

Chemicals.

The synthesis of compounds 828 and 951 is described elsewhere (37), as is that of compound 1028 (38). GCV was kindly provided by Hoffmann-La Roche, Palo Alto, Calif. All compounds were solubilized in 100% dimethyl sulfoxide at a concentration of 10 mg/ml and were stored at −20°C. The compounds were added to cultures such that the resulting concentrations of dimethyl sulfoxide never exceeded 0.05%, by volume. Because we have previously demonstrated (36) that 828 is not stable in cell culture medium, dilutions of 828, 951, 1028, and GCV were never stored in cell culture medium but were made fresh for each experiment.

Cells and viruses.

Human foreskin fibroblast (HFF) primary cells and MRC-5 cells, a human embryonic lung cell line (ATCC CCL 171), were grown in minimal essential medium with Earle’s salts supplemented with 10% fetal bovine serum. KB cells, an established human cell line derived from an epidermoid oral carcinoma (ATCC CCL 17), were grown in minimal essential medium with Hanks’ salts supplemented with 10% calf serum. These cell lines were subcultured by conventional procedures (47) by using 0.05% trypsin plus 0.02% EDTA in a HEPES-buffered salt solution (42). All cell lines were screened periodically for mycoplasma contamination and were negative. A plaque-purified isolate, P_o, of the Towne strain of HCMV was obtained from M. F. Stinski, University of Iowa. HCMV clinical isolates sensitive (isolate BW 17517) and resistant (isolate BW 48041) to GCV were kindly provided by K. K. Biron, Glaxo Wellcome. All stocks of HCMV were prepared as described elsewhere (47).

Assays for antiviral activity.

The activities of compounds 828, 951, 1028, and GCV against the Towne strain of HCMV were determined in a number of different assays, including plaque reduction, yield reduction, and DNA-DNA probe hybridization assays and an enzyme-linked immunosorbent assay (ELISA). All HCMV plaque reduction assays were performed with monolayer cultures of HFF cells in 24-well cluster dishes (Costar, Cambridge, Mass.) as described previously (47), except that the virus inoculum (0.2 ml) contained approximately 100 PFU of HCMV per well and the compounds to be assayed were contained in the overlay medium. Protocols for HCMV yield-reduction experiments have been described previously (31). Briefly, monolayer cultures of HFF cells in 96-well culture dishes (Costar) were infected at a multiplicity of infection (MOI) of 0.5, 0.05, or 0.005 PFU/cell and were incubated in the presence of the test compounds for 6 to 7 days. Following one cycle of freezing at −76°C and thawing at 37°C, the resulting lysates were diluted, and the amount of infectious virus was quantified on new cultures of HFF cells (31). A DNA-DNA probe hybridization assay (Diagnostic Hybrids, Inc., Athens, Ohio), previously described by Danker et al. (10), was used to measure the effect of 828 and GCV on viral DNA synthesis. The activities of these compounds against HCMV also were assayed by an ELISA in MRC-5 cells, as we have described previously (37).

Cytotoxicity assays.

The cytotoxicities of compounds 828, 951, 1028, and GCV were evaluated in HFF, KB, and MRC-5 cells. The cytotoxicity produced in stationary HFF cells was estimated by visual scoring in the plaque-reduction assays of cells that were not affected by virus replication. Cytopathology was estimated at a ×35 magnification and was scored on a scale of from 0 to 4 (0 = 100% viability, 4 = 0% viability) on the day of staining for plaque counting. Cytotoxicity in logarithmically growing KB cells was determined as described previously (33). In more detailed studies, the inhibitory effect of 828 was evaluated, and population doubling times (PDTs) were determined in a KB cell growth assay (25). PDTs were calculated by means of a least-squares program by fitting the exponential portion of a growth curve. For studies of growth in KB cells, growth rates were calculated by enumeration of cells with a Coulter Counter (Coulter Electronics, Hialeah, Fla.) at 0, 24, 48, 72, and 96 h in the presence of selected concentrations of the test compound.

The toxicities of 828 and 951 were also examined in logarithmically growing MRC-5 cells. The studies were performed as described for the HCMV ELISA (37), except that the cultures were not infected with HCMV. Cell growth was determined by staining the cell sheet with crystal violet, eluting with 1% (vol/vol) HCl in ethanol, and reading the absorbance at 570 nm in a microplate reader.

To determine the potential for bone marrow toxicity, the effects of 828 and GCV on hematopoietic progenitor cell survival in vitro were evaluated. Nonadherent low-density bone marrow cells were isolated from healthy adult volunteers and were placed into methylcellulose-containing cytokines (granulocyte-macrophage colony-stimulating factor [GM-CSF], interleukin 3, and erythropoietin) with selected concentrations of test agents as described by us (26). Following a 2-week incubation, colony formation was scored as either CFU in granulocyte-macrophage cells or burst-forming units in erythroid cells. Inhibition caused by the test agents was compared to that in the no-drug controls.

Data analysis.

Dose-response relationships were used to quantify drug effects by linearly regressing the percent inhibition of parameters derived in the preceding assays (except for yield experiments) against the logarithm of the drug concentrations. For yield experiments, the logarithm of the viral titer was plotted against the logarithm of the drug concentration. Fifty percent inhibitory concentrations (IC₅₀s) and IC₉₀s (yield experiments) were calculated from the linear portions of the regression lines.

Viral adsorption studies.

Inhibition of viral adsorption was studied as follows. Five 25-cm² flasks (Costar) were seeded with HFF cells at a final concentration of 1.2 × 10⁶ cells per flask. Twenty-four hours later, the flasks were infected with HCMV at an MOI of 0.005 PFU/cell. For two flasks, 32 μM 828 was added when the cells were seeded. Drug was removed when the flasks were infected, but one of these flasks was also infected in the presence of 32 μM 828. The other was infected with drug-free medium. In addition, a third flask, which was seeded with cells in the absence of drug, was infected in the presence of 32 μM 828. For all three flasks, the inoculum was removed 1 h postinfection and the cells were rinsed by the addition of fresh medium without drug. The flasks were incubated for 7 days, an aliquot of the supernatant was removed and diluted, and the amount of infectious virus was quantified on new cultures of HFF cells (31). The virus titers derived from the experimental flasks were compared with the titers derived from two control flasks, one flask to which drug was never added and one flask to which drug was added 1 h postinfection and never removed.

Time-of-addition studies.

To investigate the effect of adding 828 and GCV at numerous times postinfection, monolayer cultures of HFF cells were seeded at a final concentration of 10⁴ cells per well in 96-well microtiter plates (Costar). The cultures were incubated overnight and were then infected with HCMV at an MOI of 0.005 PFU/cell in a total volume of 200 μl/well. At 1, 6, 12, 24, 36, 48, and 72 h postinfection, drug was added to quadruplicate wells to achieve a GCV or 828 concentration of 100, 33, 11, 3.7, 1.2, 0.41, 0.14, 0.05, 0.015, 0.005, or 0 μM. The plates were incubated for a total of 7 days after infection. Following one cycle of freezing at −76°C and thawing at 37°C, the resulting lysates were diluted and the quantity of infectious virus was determined on new cultures of HFF cells as described previously (31). Additional time-of-addition studies with 828, 951, 1028, and GCV at a single concentration of 10 μM each also were performed.

Time-of-removal studies.

To investigate when the pyrrolopyrimidines and GCV begin to have their antiviral effects, time-of-removal studies were performed. Subconfluent monolayers of HFFs were seeded at 10⁴ cells per well in 96-well microtiter plates and the plates were incubated overnight. The cells then were infected with HCMV at an MOI of 0.005 PFU/cell. Following a 1-h adsorption, 10 μM 828, 951, 1028, or GCV was added to separate cultures. At 3, 6, 9, 12, 24, 36, 48, and 60 h following the addition of drug, the medium was aspirated, washed three times with either Hanks’ balanced salt solution or minimal essential medium (E), and then replaced with drug-free medium. The flasks were incubated for a total of 7 days after infection, at which point the supernatant was diluted and the titer of infectious virus was determined with new cultures of HFF cells.

Analysis of drug interactions.

A drug combination assay was performed by the HCMV ELISA procedure described by us (37) and by a three-dimensional method (MacSynergy II) to analyze drug-drug interactions developed by Prichard and Shipman (32). Briefly, data derived from quintuplicate plates were used to construct dose-response surfaces. Theoretical additive interactions were calculated from the dose-response curves for each drug used individually. This calculated surface was subtracted from the experimentally determined dose-response surface to reveal regions of nonadditive activity. Interpretation of the data is as follows: If the resulting plane appeared as a horizontal plane at 0% inhibition, the interactions between the two drugs are additive. Depressions in the plane indicate antagonism, whereas peaks above the plane indicate synergistic interactions between the two drugs. Confidence intervals (95%) around each of the points that defined the dose-response surface were calculated from the quintuplicate data to provide limits for the experimental dose-response surface. If the upper confidence limits of the experimental data were less than the calculated additive surface, antagonism would be considered significant at that confidence level. Conversely, if the lower confidence limits of the experimental data were greater than the calculated additive surface, the synergy would be considered significant. Finally, if the calculated additive surface were contained within the confidence limits, the interaction would be considered additive.

RESULTS

Activity against HCMV.

As we reported previously, the pyrrolo[2,3-d]pyrimidine nucleoside analogs 828, 951, and 1028 (Fig. 1) are active against HCMV (37, 38). Compounds 828, 951, and 1028 exhibited potent activities in plaque-reduction assays against HCMV (Table 1). Similar results were obtained for 828 and 951 in an HCMV ELISA. GCV, in contrast, was less active in ELISAs than in plaque-reduction assays. The activities of 828 and GCV were also examined in DNA-DNA hybridization probe assays. Compound 828 was slightly more active than GCV in two separate experiments (Table 1).

TABLE 1.

Activities against HCMV and cytotoxicities of compounds 828, 951, and 1028 and GCV

Compound	IC50 (μM)a
	HCMV activity			Cytotoxicity in the following cells:
	Plaque	ELISA	DNA hybridization	HFF	MRC-5	KB
828b	1.0 ± 0.6	1.9 ± 0.6	1.3 ± 0.7	≥100	>100	>100
951c	0.4 ± 0.2	0.4 ± 0.2	NDd	49	80	36
1028e	0.8 ± 1.0	ND	ND	32	ND	90
GCVf	8.7 ± 6.9	20 ± 9.0	3.7 ± 0.5	>100	>100	≥200

^aAntiviral and cytoxicity assays are described in the text. 

^bHCMV activity results are the averages ± standard deviations of five, four, and two experiments, respectively; cytotoxicity assay results are averages of four, two, and nine experiments, respectively. 

^cHCMV activity results are the averages ± standard deviations of four and two experiments, respectively; cytotoxicity assay results are the averages of three, two, and five experiments, respectively. 

^dND, not done. 

^ePlaque-reduction assay results are the averages ± standard deviations of 28 experiments; cytotoxicity assay results are the averages of two and three experiments, respectively. 

^fHCMV activity results are the averages ± standard deviations of 54, 3, and 2 experiments, respectively; cytotoxicity assay results are the averages of 102, 2, and 2 experiments, respectively. 

Effects on uninfected cells.

The effects of 828, 951, and 1028 on the growth of HFF cells, MRC-5 cells, and KB cells were examined to measure drug cytotoxicity (Table 1). All three pyrrolopyrimidines showed a reasonable separation between antiviral activity and cytotoxicity. In HFF cells, 828 and GCV were the least toxic, while 951 and 1028 had measurable cytotoxicities. In MRC-5 cells, 951 again showed greater cytotoxicity than 828 and GCV. In KB cells, 951 was the most cytotoxic, followed by 1028, 828, and GCV.

In expanded cytotoxicity studies, the effects of 828 on KB cell growth and of 828 and GCV on human bone marrow progenitor cell colony formation were determined. The PDT in KB cells increased from 22 h in control cultures to 30 h in cultures treated with 828 (data not shown). In human bone marrow progenitor cells, 828 was significantly less toxic than GCV. The IC₅₀s of GCV were 3.5 μM (95% confidence interval, 2.0 to 6.2 μM) for colony formation by granulocyte-macrophages and 30 μM (95% confidence interval, 14 to 64 μM) for colony formation by erythrocytes. In contrast, the IC₅₀s of 828 were 35.3 μM (95% confidence interval, 25 to 50 μM) for granulocyte-macrophages and >100 μM for erythrocytes.

MOI affects activity against HCMV.

Initial HCMV yield-reduction experiments with 828 showed that the compound had little or no effect. This was surprising, since it was quite active in plaque-reduction assays. One possible explanation was the difference in the MOIs used in these two assays. To examine the effect of MOI on 828 activity, yield-reduction assays were performed at three different MOIs. Figure 2 shows dose-response curves for GCV and 828 at MOIs of 0.5, 0.05, and 0.005 PFU/cell. The slopes of the dose-response curves for GCV were similar, indicating that the MOI did not alter the effectiveness of the drug. In contrast, the slopes of the dose-response curves for 828 were very different, with little antiviral activity observed at an MOI of 0.5 PFU/ml but with good activity present at an MOI of 0.005 (Fig. 2).

FIG. 2.

Effect of MOI on the antiviral activities of 828 and GCV. Subconfluent monolayers of HFF cells were infected at MOIs of 0.005, 0.05, and 0.5 PFU/cell and were incubated in the presence of the test compounds for 7 days. Infectious virus present at that time in culture supernatants was quantitated by plaque-reduction assay with HFF cells as described in the text. Similar results were obtained in three separate experiments.

Performance of HCMV yield-reduction assays with GCV, 828, 951, and 1028 at three different MOIs explored this MOI dependency further. Table 2 shows typical results. In three experiments, each performed in triplicate (including the experiment whose results are shown in Table 2), there was no statistically significant difference among the IC₉₀s of GCV as determined by paired t tests. This confirmed earlier observations that GCV did not act in an MOI-dependent manner. In contrast, in each of the three experiments, there was a statistically significant difference in the IC₉₀s of 828 at MOIs of 0.5 and 0.05 (P < 0.01 in all experiments). The difference between the IC₉₀s of 828 at MOIs of 0.05 and 0.005 was also statistically significant in two of three experiments (P < 0.0001, P = 0.001, and P = 0.155), respectively). For 951, the difference in IC₉₀s at MOIs of 0.5 and 0.05 was statistically significant (P < 0.0001 in two experiments performed in triplicate), while the IC₉₀s at MOIs of 0.05 and 0.005 were not significantly different in either experiment. This indicated that 951 also acted in an MOI-dependent manner, but its activity may be somewhat less MOI dependent than that of 828. Two experiments performed in triplicate with 1028 also demonstrated that this compound is MOI dependent in its activity against HCMV but may not be as MOI dependent as 828. In both experiments with 1028 there was a statistically significant difference between the IC₉₀s at MOIs of 0.5 and 0.05 (P < 0.0001 and P = 0.016, respectively) but the difference between the IC₉₀s at MOIs of 0.05 and 0.005 was significant only in one experiment (P = 0.03 and P = 0.7, respectively).

TABLE 2.

Effect of MOI on activities of 828 and GCV against HCMV

Compound	IC90 (μM) at the following MOI (PFU/cell)a:
	0.005	0.05	0.5
828b	0.4	24.1	52.7
951c	0.2	0.5	20.7
1028c	0.7	2.4	12.6
GCVd	0.7	1.4	0.4

^aThe IC₉₀s were interpolated from dose-response curves constructed with seven drug concentrations in triplicate. The IC₉₀s presented here are typical of those from a single experiment. Similar results were observed in one other experiment for 951 and 1028 and two other experiments for 828 and GCV. 

^bThe IC₉₀s observed for 828 at these three MOIs are statistically different from one another (P ≤ 0.05). 

^cThe IC₉₀ observed at an MOI of 0.5 is significantly different from IC₉₀s observed at MOIs of 0.05 and 0.005 (P < 0.0001 for 951; P < 0.05 for 1028). 

^dThe IC₉₀s of GCV at these three MOIs are not statistically different from one another. 

Viral adsorption studies.

Viral adsorption studies with 828 demonstrated that viral titers were dramatically reduced when the compound was added after viral adsorption but not when 828 was present before and during adsorption and then removed (Table 3). If drug was present when the cells were seeded and/or during inoculation and later washed out, little effect on HCMV titers was observed compared to the effect of the control. There did appear to be an effect when drug was present only during seeding. However, the effect was slight compared with the effect of drug added 1 h postinfection. In addition, there was no significant effect when drug was present both during seeding and during infection. This suggested that 828 does not function primarily by blocking absorption of the virus to the cell.

TABLE 3.

Effect of 828 on viral titer when 828 was present before, during, or after viral adsorption

Addition of 32 μM 828 at the following times:	Mean ± SDa HCMV titerb	% of no-drug control
Drug not present	(3.5 ± 1.2) × 105	100
When cells were seeded	(1.4 ± 0.4) × 105	40
When cells were seeded and up to 1 h postinfection	(2.8 ± 1.4) × 105	80
During infection up to 1 h postinfection	(2.4 ± 0.6) × 105	69
Beginning 1 h postinfection	(1.1 ± 0.4) × 103	0.3

^aSD, standard deviation. 

^bPFU/ml. 

Time-of-addition studies.

Time-of-addition studies were performed to examine the effect of adding GCV or 828 up to 72 h postinfection. Drug was added to cells at selected times postinfection and the HCMV titer was determined. As described above, 828 gave a dose-response curve with marked effects when it was added at zero time. In contrast, it was almost without effect when it was added at 72 h postinfection (Fig. 3). Addition at intermediate times gave results between these two extremes. GCV, on the other hand, produced multiple log₁₀ reductions in viral titers when it was added up to 3 days postinfection. In separate time-of-addition experiments, 951 and 1028 gave results similar to those for 828 (data not shown). Thus, 828, 951, and 1028 act at similar times in the viral lytic cycle, suggesting that they target a viral function that acts before replication of viral DNA.

FIG. 3.

Effect of time of addition of 828 and GCV on HCMV yield. Subconfluent monolayers of HFF cells were infected at an MOI of 0.005 and treated with test compound at 1 (○), 3 (●), 6 (□), 9 (■), 12 (▵), 24 (▴), 36 (◊), 48 (⧫), or 72 (▿) h postinfection. All plates were incubated for a total of 7 days. Infectious progeny virus was quantified by plaque-reduction assay on HFF cells as described in the text. Similar results were obtained in two separate experiments.

Time-of-removal studies.

Studies to examine the effect of removal of the pyrrolopyrimidines at various times postinfection were performed to further elucidate when the drugs act in the virus lytic cycle. As with the time-of-addition studies, time-of-removal experiments were run for 7 days to allow two rounds of viral replication, due to the MOI dependence of the pyrrolopyrimidines. All compounds were fully active when they were present for 7 days. Removal of GCV prior to viral DNA synthesis resulted in partial inhibition of HCMV replication (Fig. 4), most likely due to residual GCV-triphosphate in the cells. In contrast, the pyrrolopyrimidines were nearly fully functional if they were removed as early as 36 h postinfection, even in this multicycle growth assay. In fact, the aliphatic analog 828 appeared to act somewhat earlier than the aromatic analogs 951 and 1028. These observations are consistent with the results of the time-of-addition studies, which showed that all three of the pyrrolopyrimidines acted before GCV in the HCMV lytic cycle. The data also show that this early inhibition was not reversible, suggesting either tight-binding inhibition of a viral target or inhibition of an immediate-early or early viral protein required for subsequent replication steps.

FIG. 4.

Effect of time of removal of pyrrolopyrimidines or GCV on HCMV yield. Subconfluent monolayers of HFF cells were infected at an MOI of 0.005 and were treated with 10 μM GCV (○), 828 (●), 951 (□), or 1028 (■). At the indicated times, the cells were rinsed to remove drug and fresh medium was added; infected cultures without a drug were treated identically to serve as controls. Incubation with fresh medium was continued through 7 days, and HCMV titers were determined.

Activities against clinical isolates of HCMV.

The activities of 828 and GCV against a matched pair of HCMV clinical isolates sensitive (isolate BW 17517) and resistant (isolate BW 48041) to GCV were determined in two separate plaque-reduction experiments. GCV resistance in BW 48041 is due to a change from leucine to serine at amino acid 595 in UL97 (6). Both GCV and 828 were very active against BW 17517 (IC₅₀s = 4.5 ± 0.9 and 0.85 ± 0.07 μM, respectively). However, for strain BW 48041, GCV had an IC₅₀ of 44 ± 8.5 μM, whereas 828 had an IC₅₀ of 0.3 ± 0.1 μM. Thus, there was no evidence of cross-resistance between GCV and 828. This is consistent with evidence from other experiments that suggest the modes of action of the pyrrolopyrimidines are different from that of GCV.

Studies with drug combinations.

Because 828 and GCV act at different times in the viral replication cycle, the use of these two compounds in combination could potentiate the effect of each one alone. The effects of the two compounds on HCMV replication were measured by the ELISA and were analyzed by two methods. Figure 5 presents the resulting data as a family of dose-response curves for 828 at GCV concentrations of 0 to 100 μM. Figure 5 shows that the 828 dose-response curves shift to lower drug concentrations with increasing concentrations of GCV. This establishes that the two compounds interacted in an additive or synergistic manner. To differentiate between these two possibilities, the data were analyzed by using MacSynergy II, which indicated that the two compounds interact in an additive manner.

FIG. 5.

HCMV replication in the presence of combinations of 828 and GCV. Subconfluent monolayers of HFF cells were infected at an MOI of 0.005 and were treated in quintuplicate with the noted concentrations of 828 and with 0 (○), 1.2 (●), 3.7 (□), 11 (■), 33 (▵), and 100 (▴) μM GCV. After incubation for 7 days, HCMV was detected by ELISA and dose-response curves were constructed.

DISCUSSION

In the present study, the nature of the activities of nonnucleoside toyocamycin and thiosangivamycin analogs against HCMV has been examined. By all measures of HCMV activity tested, 828, 951, and 1028 were potent inhibitors of HCMV. By plaque-reduction, yield-reduction, and viral-DNA hybridization assays and by ELISA, the activities of these compounds were comparable or superior to that of GCV at low MOIs (0.005 PFU/cell). Toxicity studies in a variety of mammalian cell lines demonstrated that the antiviral activities of the compounds were well separated from toxicity for uninfected cells. Furthermore, bone marrow toxicity studies demonstrated that 828 was less toxic than GCV, and previous studies showed that up to 100 μM 828 did not inhibit incorporation of [³H]uridine and [³H]deoxythymidine in uninfected HSB-2 cells (37).

We have also demonstrated previously that 828 is converted in cell culture medium to the 5-carbonitrile analog (compound 830), with a half-life of 50 h (36). Hence, the biological data for 828 most likely are the result of the effects of a mixture of 828 and 830. Because 830 is inactive against HCMV, with no biological activity observed at concentrations of 100 μM (38), the antiviral activity of 828 is likely greater than that which we observed. Furthermore, since several 5-thioamide-substituted pyrrolo[2,3-d]pyrimidines are inactive against HCMV but still undergo conversion to the nitrile (36), the conversion of 828 to 830 is probably not essential for biological activity. By extension, we would also expect 951 to be converted to its corresponding carbonitrile, which is also inactive against HCMV (37). In contrast, 1028 should be stable since it contains a carbonitrile rather than a thioamide group.

To examine the mechanism of action of 828 on the replication cycle of HCMV, we performed a number of comparative studies with GCV, an inhibitor of viral DNA replication (4), as a control. The viral adsorption study established that 828 acts after viral adsorption, and the time-of-addition and time-of-removal studies demonstrated that 828 acts prior to when GCV acts. Compounds 951 and 1028 appeared to act slightly later than 828, raising the possibility that the aliphatic analog (compound 828) acts somewhat differently from the aromatic analogs (compounds 951 and 1028). Regardless, the data strongly suggest that these nonnucleoside pyrrolo[2,3-d]pyrimidine analogs all act via a similar mechanism early in the replication cycle.

All of these time-of-addition and time-of-removal experiments were performed with 0.005 PFU/cell because the pyrrolopyrimidines are most effective at a low MOI. After one round of viral replication, an input MOI of 0.005 resulted in low levels of replication even in the no-drug control. Therefore, virus from these experiments was not harvested until 7 days postinfection. This allowed two rounds of viral replication and, due to the MOI-dependent activity of the pyrrolopyrimidines, resulted in an exaggeration of the effect of the drug. For example, because 828 was essentially inactive at an MOI of 0.5 PFU/cell, if enough virus were produced during the first round of replication to cause an MOI of 0.5 or higher for the second round of replication, the virus will replicate as though no drug were present. In contrast, if addition of drug at a particular time postinfection strongly inhibited viral replication during the first lytic cycle, the MOI for the second cycle will remain in a range at which the drug is active. Therefore, the drug would continue to strongly inhibit viral replication during the second lytic cycle as well. By the same reasoning, drug effects at times of partial inhibition in the time-of-removal experiments may also have appeared to be less than they actually were.

In contrast to GCV, the activities of the pyrrolopyrimidines against HCMV decreased as a function of increasing viral load. These results demonstrate that the reduction in titer by 828 at 0.005 PFU/cell was due to an inhibitory effect on a viral process. If 828 simply produced a toxic effect on a cellular process, reductions in virus titer at higher MOIs would have been observed, especially at higher concentrations of drug.

We have not definitively established the target of action of nonnucleoside pyrrolo[2,3-d]pyrimidines. However, because these drugs are active earlier in the replication cycle than GCV, which targets viral DNA polymerase, the target of the pyrrolopyrimidines may be an immediate-early protein. Because the pyrrolopyrimidines are different in so many ways from GCV (MOI dependence, time of activity, and activity against GCV-resistant virus), it is reasonable to assume that these drugs do not inhibit the viral polymerase, which is a target for all the currently approved HCMV therapies. The pyrrolopyrimidines share several of these characteristics with other heterocyclic inhibitors of HCMV replication. Thiazolopyrimidines (21) and RPR CMV423 (2) also inhibit HCMV in an MOI-dependent manner and act early in the replication cycle, thereby raising the possibility of a common viral target. However, the viral target(s) of these compounds has also not yet been identified.

The MOI dependence of the pyrrolopyrimidines is consistent with the hypothesis that they target a viral protein that is required in an MOI-dependent manner, such as IE1 (16). Alternatively, they could target a viral protein that is transported intercellularly, as the herpes simplex virus type 1 VP22 is (14). Greaves and Mocarski (16) have demonstrated that the HCMV IE1 protein is required in an MOI-dependent manner; virus lacking IE1 requires 2 to 3 PFU/cell rather than 1 PFU/cell in order to replicate (16). IE1 is also appealing as a target for this series of compounds because it has serine kinase activity (28). Sangivamycin, a ribosyl pyrrolopyrimidine structurally similar to the compounds tested in the present study, is known to inhibit protein kinase C, a cellular serine threonine protein kinase (22, 27). It is possible that structural differences between sangivamycin and 828, 951, and 1028 lessen or remove activity against protein kinase C but result in inhibition of a virally induced protein kinase, such as IE1, instead.

In summary, we have demonstrated that 828, 951, and 1028 are potent nonnucleoside inhibitors of HCMV that act earlier in the viral replication cycle than GCV. In addition, the toxicity of 828 is comparable to that of GCV for tissue culture cells but is less toxic than GCV for bone marrow progenitor cells. Consequently, these compounds may be useful antiviral agents because of their selective inhibition of HCMV, their unique site of action, and their activities against HCMV isolates resistant to GCV.