Genotypic analysis of cytomegalovirus retinitis poorly responsive to intravenous ganciclovir but responsive to the ganciclovir implant☆☆☆

Design 

Patients were identified by retrospective chart review. Stored vitreous from the eyes of these patients was subsequently studied in the laboratory, as detailed below. Vitreous samples from patients with CMV retinitis in which the viral genotype had been previously reported were excluded from the current study.3 This study was approved by the Human Research Committees at University of California at San Francisco, University of California at Irvine, and Emory University, and informed consent was obtained from patients participating in this study.

Methods 

We studied undiluted vitreous specimens from eight eyes of six patients with AIDS and CMV retinitis that responded poorly to intravenously administered ganciclovir but well to the ganciclovir implant. Each of these eyes failed to achieve inactive disease despite 4 weeks of induction-level therapy with intravenous ganciclovir (5 mg/kg every 12 hours), but became inactive following surgical placement of the ganciclovir implant. Lesions were considered to be inactive when the borders showed no opacification and no progression (defined as a new lesion or more than 750 μm of advancement of the border of an existing lesion, as judged by a serial clinical examinations).24 The retinitis in each of the eight eyes that were studied met three different sets of published criteria for “clinical resistance,”3, 24, 25 including those established by the AIDS Clinical Trials Group of the National Institute of Allergy and Infectious Disease.25 Study patients were identified by retrospective chart review of AIDS patients with CMV retinitis examined at University of California at Irvine Medical Center or at Emory University Medical Center. To avoid confounding variables associated with HAART and immune reconstitution we only reviewed charts of patients who were treated for CMV retinitis before 1995.

Vitreous samples from the eyes of these patients were obtained at the time of initial surgical implantation of the ganciclovir implant into the vitreous through the pars plana and stored at −70 C. Informed consent was obtained from all six patients for both implantation of the ganciclovir implant and molecular analysis of biopsied vitreous for CMV DNA. Vitreous from the eyes of 21 patients with AIDS and retinal infections other than CMV retinitis (for example, VZV retinitis, HSV retinitis, toxoplasmosis) served as negative controls. Before analysis all vitreous samples were incubated in sterile microcentrifuge tubes at 95 C for 10 minutes.

CMV strain AD169 was the source of wildtype UL97 and UL54 DNA. Control plasmid DNA with the CMV UL97 V460 resistance mutation was generated previously.3 Control plasmid DNA with the CMV UL97 Q520 resistance mutation was generated by polymerase chain reaction (PCR)-based site-directed mutagenesis.26 Control plasmid DNA with the CMV UL54 G987 resistance mutation was generated by cloning a 97 bp fragment of the CMV UL54 gene with the G987 resistance mutation (originally obtained from a patient with cidofovir resistant retinitis) into pBluescript II KS(+) (Stratagene, La Jolla, California, USA).

Nested sets of oligonucleotide primers were designed to PCR-amplify regions of the UL97 and UL54 genes with mutations known to confer resistance to ganciclovir. For amplification of portions of the UL97 gene all three nested primer sets use the same external primer pair, WQ1 and WQ2.3 Three different pairs of internal primers—(1) 5′-TCATCACGACCAGTGGAAGC-3′ and 3′-AGTCCTCTGCCCATGCCGCGC-5′, (2) 5′-GTGCGATT-ACAGCCTCAGC-3′ and 3′-AGACGCTGGGCGTG-CGCGCAAA-5′, and (3) 5′-GGGTAACGTGCTGGGC-TTTTG-3′ and 3′-CGAAGATGGTGCTTACGAGC-5′—were used to generate DNA fragments covering codons 420 to 505, 480 to 539, and 557 to 657, respectively. For amplification of portions of the UL54 gene with established ganciclovir resistance mutations we used four different sets of nested primers. These four sets of nested primers cover (1) codons 379 to 421 (Ext: 5′-CGATCGGCACCTGCGGGCA-3′, 3′-TGAGCGTCGCGAAG-ACGTTC-5′ Int: 5′-CAGGTGGGCCCAGACGT-GGA-3′, 3′-TCATGGACATATTCCACCTGAG-5′), (2) codons 492 to 539 (Ext: 5′-GCGGGTTCGGTGGTTATCGACAT-3′ and 3′- ACCACGCGCTAGACAAGTTGTGG-5ı́’; Int: 5′-CGACATGTACCCTGT-ATGCA-3′ and 3′-ACGTCCTGCGGCATAACCAC-5′), (3) codons 529 to 612 (Ext: 5′-GATGACCTG-TCTTACAAGGA-3′ and 3′-ATAGTAGAGATGGC-GGCGAC-5′ Int: 5′-CCCGCGTTGTTTCGTGGCTA-3′ and 3′-ACGACACAGTGGATTGCGAC-5′) and (4) codons 765 to 839 (Ext: 5′-CGCTAGAGAACGGCGTGACCCA-3′ and 3′-TACGATCTCGCGTGCCGC-GCCAAG-5′ Int: 5′-GGCGTGACCCACCGCT-TTGT-3′ and 3′-AGCCACCGCTGTACGATCTC-5′). For the initial round of PCR, 5 μl of vitreous was mixed with formamide (10% final concentration), dNTPs (50 μm each; Pharmacia Biotechnology, Piscataway, New Jersey, USA), 200 pM primers (WQ1 and WQ2), and 0.5 U of Vent (exo) polymerase (New England Biolabs, Beverly, Massachusetts, USA) in a total volume of 50 μl. Thermocycling parameters were 94 C for 60 to 90 seconds, 50 C for 60 to 90 seconds, and 72 C for 60 to 90 seconds for a total of 40 to 45 cycles. Five microliters of reaction product was then used for the second round of PCR using 200 pM of the appropriate internal primer pairs, dNTPs (50 μmol/l each) and 0.5U of Vent (exo) polymerase in a total volume of 50 μl. Thermocycling parameters for the second round of PCR were 94 C for 60 seconds, 50 C for 60 seconds, and 72 C for 60 seconds.

Screening of amplified CMV UL97 DNA for resistance mutations at codons 460 and 520 accomplished by restriction digest analysis of amplified CMV UL97 or UL54 DNA following a final short round of PCR amplification using intentionally mismatched (underlined bases) oligonucleotide primers. These primer sets consisted of 5′-CGACAGCTACCGACGTGCCTTTTGCAGTT-3′ and 3′-AGCTGCACTTGGGCGTGTTG-5′ to amplify the region around codon 460 and 5′-CTATCCGGATTACAACGAGCTCTG-3′ and 3′-GACGAAAGGCTGGGTACGGCGACGTC-5′ to amplify the region around codon 520. Adapted from a protocol described by Chou and associates,8 the use of these mismatched primers allowed us to introduce specific restriction digest sites into amplified UL97 DNA for the purposes of (1) differentiating wildtype from mutant UL97 DNA and (2) testing the activity of the restriction enzyme used in the assays. Specifically, 3 μl of UL97 reaction product (from the nested PCR amplification described above) was added to PCR buffer containing dNTPs (50 μmol/l each), formamide (10% final concentration), appropriate primers (200 pM each) and 0.5 U of Vent (exo) in a 50-μl volume. Thermocycling parameters for the first five cycles consisted of denaturation at 94 C for 45 seconds, annealing at 40 C for 45 seconds, and extension at 72 C for 45 seconds. This was followed by 10 cycles of denaturation at 94 C for 45 seconds, annealing at 50 C for 45 seconds, and extension at 72 C for 45 seconds. The final amplified products were subjected to overnight restriction enzyme digestion with either Hla III (for codon 460) or Alu I (for codon 520), electrophoresed through polyacrylamide and stained with ethidium bromide. Mutations that were identified by restriction digestion were confirmed by repeat amplification and double-stranded DNA sequencing. Screening of amplified CMV UL54 DNA for a G987 resistance mutation was accomplished in a similar fashion using 5′-CAAGAAACGTTACATCGGCA-3′ and 3′-CCGGAGTGCGCACTGCAGGAG-5′ as external primers and 5′-CGGCAAAGTGGAGGGCGCCT-3′, 3′-TCCCGA-GTGCGCACTGCAGG-5′ as internal primers and Hae III as the restriction enzyme.

Double-strand DNA sequencing was performed using a fluorescent dye terminator sequencing kit (AMpliTaqFS, Perkin-Elmer ABI) and an automated DNA sequencer (ABI-Prism Model 377; Applied Biosystems, Foster City, California, USA). Amplified PCR products were purified before sequencing using the QIAEX II kit (Qiagen, Chatsworth, California, USA).

Results 

Analysis of UL97 CMV DNA amplified from vitreous 

Targeted regions of the CMV UL97 gene encompassing codons 420 to 505, 480 to 539, and 557 to 657 were successfully amplified by PCR from the vitreous of all eight study eyes and analyzed for resistance mutations by a combination of DNA sequencing and restriction digestion. Of the eight vitreous samples tested, only one contained CMV DNA with a UL97 resistance mutation, a C to T mutation at base 1781, predicting an alanine to valine mutation at codon 594 (see Table 1). This mutation was confirmed by repeat amplification and double-stranded DNA sequencing. Using the same methods no CMV DNA was amplified from the vitreous of 21 control eyes from patients with AIDS and retinal infections other than CMV retinitis.

Table 1. Summary of Genotypic Analysis of CMV DNA Amplified From Study Eyes

	Case	Eye	UL97 Target Codons			UL54 Target Codons
			460	520	590–607	379–421	492–612	765–839	987
	1	OD	WT	WT	WT	WT	NA	WT	WT
		OS	WT	WT	WT	G406S†	WT	WT	WT
	2	OD	WT	WT	WT	WT	WT	WT	WT
		OS	WT	WT	WT	N408S†	WT	WT	WT
	3	OD	WT	WT	A594V*	WT	WT	WT	WT
	4	OS	WT	WT	WT	WT	WT	WT	WT
	5	OD	WT	WT	WT	WT	WT	WT	WT
	6	OS	WT	WT	WT	WT	WT	NA	A987G*

A = alanine; CMV = cytomegalovirus; G = glycine; N = asparagine; NA = not amplifiable; V = valine; WT = wild-type; S = serine. A594V = alanine to valine mutation at codon 594; A987G = alanine to glycine mutation at codon 987; G406S = glycine to serine mutation at codon 406; N408S = asparagine to serine mutation at codon 408.

* These mutations have been confirmed by marker transfer to cause resistance. † These mutations have not yet been confirmed by marker transfer to cause resistance.

Discussion 

In this study we performed genotypic analysis of eight vitreous samples from six patients with CMV retinitis refractory to intravenous ganciclovir but responsive to the ganciclovir implant. Three major points emerge from our analysis: (1) CMV retinitis that is poorly responsive to intravenous ganciclovir may be caused by virus with wild-type genotype as well as by virus with resistance mutations; (2) the ganciclovir implant can be effective therapy for CMV retinitis caused by virus with certain UL97 and UL54 resistance mutations, although it may not be effective for all known resistance mutations; and (3) CMV UL54 genotypes can differ between eyes in patients with bilateral retinitis.

Both wild-type CMV and CMV with resistance mutations can cause retinitis that is poorly responsive to intravenous ganciclovir. Vitreous from only two of the eight eyes with CMV retinitis that we tested in this study had established resistance mutations in either the UL97 or UL54 gene. Vitreous from an additional two eyes had UL54 mutations which, based on their location, are likely to confer ganciclovir resistance. It is possible that the virus in the remaining four eyes had ganciclovir resistance mutations that have yet to be described. However, we believe this is unlikely since Chou and associates9 have shown that mutations at just four UL97 codons (460, 520, 594, and 595) account for approximately 90% of ganciclovir-resistant CMV (IC50 seconds > 6.0 μmol/l; 1.6 μg/ml) isolated from blood and urine.9 Our assays were clearly designed to detect mutations at these four codons as well as many others. In those eyes in which we were unable to find CMV UL97 and UL54 resistance mutations it is likely that intravenous therapy with ganciclovir was unsuccessful as a consequence of poor intraocular drug delivery.

It is highly unlikely that the four mutations that we identified in the eyes of our study patients were false positives. First, we routinely assay negative controls simultaneously with clinical specimens. Second, using the same assays we failed to detect CMV DNA in 21 vitreous samples from the eyes of AIDS patients with retinal infections other than CMV retinitis. Third, using the same assays we detected UL97 resistance mutations in less than 1% of eyes with newly diagnosed CMV retinitis.29

Our data indicate that, in addition to the management of CMV retinitis caused by wild-type virus, the ganciclovir implant may be useful in the management of CMV retinitis caused by some drug-resistant viral strains. This observation is consistent with the higher levels of ganciclovir that are achieved in the vitreous during therapy with the implant compared with that found in the course of intravenous therapy. Kuppermann and associates16 demonstrated a mean intravitreal ganciclovir concentration of 0.93 μg/ml (3.64 μmol/l)in patients with AIDS who were being actively treated with maintenance intravenous ganciclovir therapy for CMV retinitis. This concentration of ganciclovir is lower than that needed to achieve 50% plaque reduction of many wild-type isolates of CMV.30, 31, 32, 33 In contrast, the mean concentration of intravitreal ganciclovir achieved with the implant is 4.1 μg/ml (16.1 μmol/l), a concentration that exceeds the IC50 for most moderately resistant isolates of CMV.19, 20

The presence of different UL54 genotypes in the right and left eyes of the two study patients with bilateral CMV retinitis is consistent with earlier work in which we demonstrated different CMV UL97 and gB genotypes in patients with bilateral disease.3, 15 It is also consistent with studies of nonocular tissues in which patients with AIDS have been shown to be simultaneously infected with multiple CMV strains, with different strains isolated from different tissues.14 These findings highlight the problem in assuming that CMV propagated from blood or urine accurately reflects the viral genotype in the eye. They also raise the question as to whether drug-resistant mutations may be locally selected for in the unique microenvironment of the eye, as a consequence of subtherapeutic antiviral drug delivery from systemic administration.16, 18

In this study we were successful at amplifying all of the targeted regions of the CMV UL97 gene but only approximately 90% of the targeted regions of the UL54 gene from eight study eyes with CMV retinitis. The most likely explanation for our inability to amplify all targeted regions was that the patients in this study were all being aggressively treated for CMV retinitis at the time vitreous samples were obtained, thereby decreasing the ocular viral load. We have previously shown that it is more difficult to amplify CMV DNA from eyes of treated patients than from eyes with newly diagnosed retinitis.34 A second possible explanation is that the targeted regions of viral DNA in some of the vitreous samples may have had polymorphisms at one or more of the primer binding sites, leading to inefficient PCR amplification.

This study highlights a problem with the current clinical terminology used to describe CMV retinitis that fails to respond to a standard regimen of antiviral therapy. The terms “clinically resistant,” “drug resistant,” and “rapidly relapsing” are often used interchangeably and without formal definition in describing such cases. Clearly, the use of such terms in the description of the cases presented in this study could lead to inaccurate conclusions about the phenotype or genotype of the responsible virus. In this manuscript we have adopted the terminology “poorly responsive to” in an attempt to define a specific clinical response independent of viral genotype or phenotype. We realize that there are problems associated with the use of this terminology as well but hope that this opens up a constructive dialogue on this subject.

In conclusion, CMV retinitis that is poorly responsive to intravenous ganciclovir therapy but responsive to the ganciclovir implant may be caused by both wild-type virus as well as virus with UL97 or UL54 resistance mutations. The greater effectiveness of the ganciclovir implant over intravenous administration in these cases is most likely due to the higher intraocular drug levels that are achieved with the implant, levels that are higher than the IC50 for many moderately resistant CMV isolates.