Polymerase chain reaction confirmed by immunohistochemistry: a two-pronged diagnostic approach in endophthalmitis


Jonathan C. Horton MD, PhD
Beckman Vision Center
University of California
San Francisco
California 94143–0730
Tel: + 1 415 476 7176
Fax: + 1 415 476 8309
Email: hortonj@vision.ucsf.edu

A 53-year-old man with adult-onset diabetes mellitus developed confusion, fever, pneumonia and loss of vision in his right eye. He was treated intravenously with cefotaxime, levofloxacin and fluconazole. His medical condition stabilized after 2 weeks and he was transferred to our institution. On examination, the right eye had minimal light perception with proptosis, conjunctival oedema and a hypopyon (Fig. 1A). Cultures of the conjunctiva, cornea, aqueous and vitreous resulted in no growth of organisms, presumably because the eye had been sterilized by prior treatment with antibiotics. Vancomycin and ceftazidime were injected intravitreally. After light perception was lost, the painful blind eye was enucleated.

Figure 1.

 (A) Panophthalmitis in the right eye. (B) Paraffin section showing a dense intraocular inflammatory reaction (haematoxylin and eosin stain). (C) Gram stain of aqueous sample, containing dispersed pigment granules and scattered, de-colorized organisms suggestive of diplococci (arrow). (D) Real-time polymerase chain reaction results for the lytA gene. DNA samples processed from paraffin-embedded specimen were run in triplicate, along with positive and negative controls. The orange line represents the threshold for a positive result. Similar results were obtained for the psaA gene. (E) Pathology specimen showing the hypopyon in the anterior chamber, with the cornea visible at upper left (haematoxylin and eosin stain). No bacterial or fungal elements were detected. (F) The same region as in (E), in a section processed for immunohistochemistry, showing Streptococcus pneumoniae antigen revealed by naphthol fast red reaction product, concentrated in macrophages.

On pathological examination the anterior chamber and vitreous were found to be filled with a dense, non-granulomatous inflammatory infiltrate (Fig. 1B). The retina was detached and necrotic. Gram, Ziehl–Neelsen and Gomori methenamine silver stains revealed no fungal or bacterial elements on high-power microscopic examination of the pathology specimen. However, review of the initial aqueous Gram stain showed sparsely distributed, de-colorized organisms resembling diplococci (Fig. 1C). This finding suggested an endogenous endophthalmitis caused by Streptococcus pneumoniae, presumably from a pulmonary source (Miller et al. 2004).

To identify S. pneumoniae as the infectious agent, slide-mounted 10-μm histological sections were deparaffinized, air dried and incubated overnight at 56 °C in 180 μl of QIAamp tissue lysis buffer and 20 μl of proteinase K (Roche Diagnostics Corp., Indianapolis, IN, USA). Extraction was performed with the QIAamp DNA mini kit (Qiagen, Inc., Valencia, CA, USA). Real-time polymerase chain reaction (PCR) with appropriate controls was carried out on the sample in triplicate (Carvalho et al. 2007). Both the lytA and psaA genes of S. pneumoniae were present (Fig. 1D).

Because PCR is extremely sensitive, it carries a risk of producing false positive results caused by contamination during sample collection or processing (Van Gelder 2001). There is even a remote risk that organisms present within the bloodstream during septicaemia, but not actually responsible for infection within the eye, can produce a false positive result. The method is also disadvantaged in that it provides no information about the locus of infection in a whole-eye pathology specimen unless microdissection is performed (Rao et al. 2006). For example, PCR may identify DNA from conjunctival flora, which may cause endophthalmitis to be erroneously attributed to organisms present outside the eye.

To eliminate the possibility of a false positive PCR result, immunohistochemistry was performed (Guarner et al. 2007). Additional slide-mounted sections were deparaffinized, rehydrated and processed in a DAKO autostainer (DakoCytomation, Inc., Carpinteria, CA, USA). The tissue was digested with proteinase K and incubated with a polyclonal antibody produced by immunizing rabbits with psaA-purified protein from S. pneumoniae, strain 22F. Figure 1E shows the hypopyon from the anterior chamber. A section processed for immunohistochemistry demonstrated abundant reactivity for the S. pneumoniae antigen in the anterior chamber (Fig. 1F) and throughout the globe. Control samples which omitted the primary antibody showed no staining.

Immunohistochemistry has not been combined previously with PCR in an ocular specimen to identify an infectious agent. It provides independent evidence for the presence of an organism identified by PCR and shows precisely which ocular tissues have been infected. It is a valuable adjunct to PCR when infectious agents cannot be identified positively on microscopy or by culture.

The use of immunohistochemistry in conjunction with PCR is feasible only when a tissue specimen is available for analysis. In this case, tissue was obtained through enucleation. However, immunohistochemistry and PCR could also be performed together on surgical specimens obtained from biopsy of ocular tissues. Because false positive results can occur in PCR, this dual approach should be considered when there is a need to improve diagnostic accuracy, especially when culture data are negative.


We thank Sherif R. Zaki, Melissa Whaley and Bernard Beall. This research was supported by an unrestricted grant from Research to Prevent Blindness.