Diagnostic efficacy of real-time PCR for ocular cytomegalovirus infections

Ocular infections caused by cytomegalovirus (CMV) are well-known to lead to CMV retinitis in immune-compromised patients. In immune-competent patients, CMV corneal endotheliitis and CMV uveitis have been diagnosed mainly in elderly individuals. The clinical signs of CMV can also be observed in non-HIV patients with retinitis [1].

CMV infections of the anterior segment are relatively rare and often misdiagnosed. CMV corneal endotheliitis can present as bullous keratopathy or keratitis after years of recurrent episodes. CMV uveitis can present as Fuchs heterochromic iridocyclitis or Posner-Schlossman syndrome with recurrences of elevations of the intraocular pressures (IOPs) [2]. For CMV corneal endotheliitis, the hallmark signs of the disease are endothelial cell loss, coin-shaped lesions, IOP elevations, and owl’s eye appearance of the lesions. An earlier study showed that these signs had a highly positive predictive value of 90.9% for the diagnosis of CMV [3]. For CMV uveitis, a relapsing inflammation is a well-known and characteristic sign, and it may be accompanied by endothelial cell loss. However, the diagnosis of CMV infection is difficult when made by only the clinical signs and symptoms.

PCR has been established as the standard diagnostic method for diagnosing systemic CMV infections achieving a sensitivity of 80.1% and specificity of 93% for blood samples [4]. For the diagnosis of ocular CMV infection, the amount of aqueous sample is limited, and the efficacy of quantitative real-time PCR (qPCR) has not been adequately validated. In addition, a CMV infection of the eye in immune-competent patients is relatively rare and is mainly observed in the Asian population.

Thus, the purpose of this study was to determine the efficacy of qPCR and clinical characteristics that will allow clinicians to detect and diagnose ocular CMV infections in individuals living in Japan [2] [5]. To accomplish this, we first evaluated the efficacy of CMV qPCR at five major ophthalmological institutions throughout Japan. We determined institutional or clinical factors which had affected the diagnostic efficacy of ocular CMV infection using a CMV standard and aqueous humor samples from consecutive cases of suspected ocular CMV infections. The results showed that qPCR of CMV is a very good test with excellent efficacy for diagnosing ocular CMV infections. However, knowledge of the clinical characteristics and unclassified CMV detection improves the efficacy of the diagnosis.

The results showed that the AUC for the diagnostic efficacy of CMV qPCR for aqueous samples was 0.98 for ocular CMV infections, and no other single clinical sign matched this high AUC. This indicated that qPCR for CMV is the most efficient single diagnostic method to detect ocular CMV infections, and its use allows an accurate diagnosis at the first visit when any type of ocular CMV infection is suspected.

Clinicians generally rely on the clinical signs for their diagnosis. We showed that the factors which significantly affected the diagnosis of ocular CMV infection was dependent on the disease type. The most useful clinical signs for the diagnosis were the frequency of recurrences of the corneal and anterior uveitis types of CMV infections. CMV corneal endotheliitis can be diagnosed very effectively with the clinical characteristics of the frequency of recurrences.

Suspected anterior segment viral infections, including the corneal disease types and anterior uveitis types, were associated with CMV, HSV, and VZV. We showed that CMV was the most frequent cause of these diseases. The results indicated that the exclusions of HSV and VZV were very important for the diagnosis of CMV infections. For the exclusion of HSV infections, a history of dendritic lesions or recurring stromal keratitis should be considered. A history of herpes zoster ophthalmicus highly suggests the presence of ocular VZV infection [16] [17]. Thus, the exclusion of HSV and VZV may also be achieved by the history, and DNA detection is not required. This information will benefit clinicians in the differential diagnosis of ocular CMV infections effectively. In addition, all of these Herpes virus family members are well known to shed spontaneously from healthy tissues and can be detected in diseased eye. However, we did not observe co-detection of CMV and HSV or VZV in the aqueous samples.

We showed that CMV qPCR has high specificity for diagnosing ocular CMV infections. However, we observed cases with unclassified CMV detection. In elderly subjects, more than one-half were seropositive for CMV, and the frequency of shedding of CMV in body fluids including urine and serum was approximately 7% [15]. There remain clear whether unclassified CMV detection was caused by shedding. In our ocular inflammatory disease cases, the corneal disease type was most significantly associated with the unclassified CMV detection, and its main characteristic was a history of corneal transplantation. Under these conditions, we need to be aware that presence of CMV DNA may not require treatment.

The differences in the disease type may reflect where the CMV is latently infected before causing an active infection. Thus, the identification of the tissue of CMV detection without requirement of treatment is most likely the latently infected tissue. We found a significant association of unclassified CMV detection and the corneal disease type and previous history of corneal transplantation. This suggests that the probable tissue where CMV resides latently is the cornea, and we have shown that CMV can effectively replicate in the corneal endothelial cells [8].

In CMV retinitis, the amount of CMV DNA was significantly higher than in CMV corneal endotheliitis or anterior uveitis (Fig. 1). Generally, CMV retinitis is associated with systemic CMV infections, and the amount of CMV DNA in the blood is known to be a significant sign for poor prognosis and increased mortality [18] [19]. Transmission to the retina is considered to occur by transmission from the blood through the retinal vascular endothelium [20]. CMV retinitis is generally observed in immune compromised patients, and the reduced anti-viral responses may permit unrestricted viral proliferation. This may explain the higher amount of CMV DNA in CMV retinitis patients, and why presumable latency or unclassified detection was not observed for the retinitis type of CMV disease.

In the anterior uveitis type of CMV disease, the AUC for the number of recurrences and IOP elevations was lower than that for the corneal disease type. Such differences may be caused by strain differences of the CMV [21]. Currently, it remains unclear whether the endotheliitis type and the anterior uveitis type of CMV are caused by different CMV strains. However, Oka et al. reported no significant difference in the amount of viral load or viral protein profiles for these two types of diseases [22].

The quantitative aspects of CMV qPCR have not been well appreciated for the aqueous humor. Some researchers have concluded that a positive or negative detection of the DNA of CMV was sufficient information. However, we conclude that the viral load of CMV is also very important information for clinicians to determine the treatment protocol as is well known for systemic CMV infections [19]. For example, the viral load of CMV was significantly associated with the endothelial cell loss and glaucoma medications [2] [7]. Thus, a higher viral load will predict refractoriness or advancement of the stage of CMV infection. When treating anterior segment inflammations due to CMV, anti-viral drugs are used. After anti-CMV viral drugs are applied, the amount of CMV decreases, and the level generally becomes undetectable [23]. However, the clinical signs of endotheliitis or uveitis, and the IOP elevation can promptly resolve before the aqueous CMV amount becomes completely negative. A decrease in the viral load in the presence of low levels in the aqueous humor suggests a need of prolonged use of anti-viral drugs. However, the decision on whether to continue the use of antivirals is very difficult to make without tracking the viral load. In addition, the resistance to ganciclovir occurs in up to 10% of the cases when anti-viral treatment duration becomes longer in systemic infections [19]. Tracking of the amount of viral DNA not responding to the antiviral treatment will become an important sign to consider drug-resistant infections.

To detect the presence or absence of CMV, the qPCR facilities had equally efficient diagnostic efficacy. This was somewhat unexpected because the five facilities used different primers, reagents, and equipment. Importantly, the copy numbers by each institutional assay were different by up to fourfold for the same sample. This can be due to different target and amplification protocols, or DNA standards of copy numbers. In our hands, the amplification efficacy of different qPCR methods for CMV did not appear to cause significant changes of the copy numbers. Indeed, incorrect amplification is easily noticed by technician and by trouble shooting. However, problematic quality of DNA standard for copy number calculations is very difficult to be recognized. Although this can be tested by limiting dilution of standard to one copy of DNA, maintaining this high sensitivity is difficult for routine laboratory work. During trouble shooting processes, we noticed a decay or improper preparation of the DNA standards that can lead to overestimations of the template DNA concentration. This can cause significantly increased calculated copy numbers. This further supports the importance of validated DNA standards.

Thus, we propose the use of standardized international unit, which theoretically corresponds to one copy of genome. Generally, anterior segment CMV infections are rare, and the pooling of the data and comparisons of case series was required to accurately determine the etiology of the disease and the therapeutic strategy. Thus, reporting the viral loads in IUs can facilitate more accurate assessments of such trials, and efficient integration of data by meta-analysis will help determine the efficacy.

There are several limitations in our study. The efficacy of qPCR for clinical samples and incidence of ocular CMV infection were assessed using case series which were mainly referred to us for diagnosis and may not represent the exact incidence. In addition, the results of a non-selected population might differ. After this assessment of performance of qPCR, the protocol of PCR in each facility may have been updated to further improve efficacy of PCR for quality control purpose.

For the current study, the sensitivity of CMV qPCR was 98 IU as the limit of detection which is higher than the theoretical limit of detection of PCR (three copies) [6]. However, when 98 IU was used as the cut off value, no difference in the efficacy of qPCR was observed in the different institutions (Table 2). In addition, patients with less than cut off were not required for anti-CMV treatment.

Sensitivity of qPCR for ocular CMV infections was very high. However, CMV DNA was not always detected at the first visit, and some cases required repeated examinations of CMV DNA for the final diagnosis. Clinicians need to be aware that qPCR becomes negative when inflammations are not intense.

In conclusion, qPCR is a very effective diagnostic test; however, the CMV DNA amount depends on the clinical characteristics. Thus, determining the clinical characteristics will facilitate a more accurate diagnosis than simply relying on costly CMV qPCR testing.