Illuminating uveitis: metagenomic deep sequencing identifies common and rare pathogens

Abstract

Background: Ocular infections remain a major cause of blindness and morbidity worldwide. While prognosis is dependent on the timing and accuracy of diagnosis, the etiology remains elusive in ~50 % of presumed infectious uveitis cases. The objective of this study is to determine if unbiased metagenomic deep sequencing (MDS) can accurately detect pathogens in intraocular fluid samples of patients with uveitis.

Methods: This is a proof-of-concept study, in which intraocular fluid samples were obtained from five subjects with known diagnoses, and one subject with bilateral chronic uveitis without a known etiology. Samples were subjected to MDS, and results were compared with those from conventional diagnostic tests. Pathogens were identified using a rapid computational pipeline to analyze the non-host sequences obtained from MDS.

Results: Unbiased MDS of intraocular fluid produced results concordant with known diagnoses in subjects with (n = 4) and without (n = 1) uveitis. Samples positive for Cryptococcus neoformans, Toxoplasma gondii, and herpes simplex virus 1 as tested by a Clinical Laboratory Improvement Amendments-certified laboratory were correctly identified with MDS. Rubella virus was identified in one case of chronic bilateral idiopathic uveitis. The subject's strain was most closely related to a German rubella virus strain isolated in 1992, one year before he developed a fever and rash while living in Germany. The pattern and the number of viral identified mutations present in the patient's strain were consistent with long-term viral replication in the eye.

Conclusions: MDS can identify fungi, parasites, and DNA and RNA viruses in minute volumes of intraocular fluid samples. The identification of chronic intraocular rubella virus infection highlights the eye's role as a long-term pathogen reservoir, which has implications for virus eradication and emerging global epidemics.

Keywords: Metagenomic deep sequencing; Pathogen discovery; Rubella virus; Uveitis.

Fig. 1

Pathogen identification based on abundance and background subtraction. Organisms in each sample are plotted as a function of matched read pairs per million read pairs (rM) at the species level based on nucleotide (nt) alignment. For an organism to be considered a potential pathogen, it must have known pathogenic potential and have >20 rM (above dashed line). For subject 3, H. hammondi is a eukaryotic organism closely related to T. gondii. It is expected that a small fraction of sequences originating from T. gondii will align to other closely related organisms. Abbreviations: sp, species; H. hammondi, Hammondia hammondi; T. gondii, Toxoplasma gondii; HSV-1, herpes simplex virus-1; C. neoformans, Cryptococcus neoformans; P. melaninogenica, Prevotella melaninogenica; V. dahliae, Verticillium dahliae; S. erythraea, Saccharopolyspora erythraea; S. saprophyticus, Staphylococcus saprophyticus

Fig. 2

Clinical course and ocular findings of a 40-year-old man with bilateral, idiopathic chronic anterior and intermediate uveitis. a Subject 6’s clinical course spanning 22 years. b Shows different colored irises (heterochromia) between the right and left eyes (top panels) and transillumination defects that are prominent in the left eye because of iris atrophy (lower panels). c Shows diffused aggregates of inflammatory cells (keratic precipitates; red arrows) on the endothelium of the cornea. Abbreviations: HSV, herpes simplex virus; VZV, varicella zoster virus; CMV, cytomegalovirus; PCR, polymerase chain reaction; RE, right eye; LE, left eye; MMR, measles/mumps/rubella vaccine; MTX, methotrexate; Rx, treatment

Fig. 3

Identification of rubella virus (RV) by metagenomic deep sequencing (MDS). a Illustrates how the 9688 nucleotide paired-end sequence reads obtained from sequencing the RNA extracted from subject 6’s aqueous fluid aligned to the most closely matched RV genome (GenBank DQ388280.1): 99.3 % of the total RV genome is represented. Positions of synonymous (black vertical lines) and non-synonymous (red vertical lines) variants are shown. Of the 149 substitutions, 107 were synonymous and 42 were non-synonymous. Of the 42 non-synonymous mutations, 25 occurred within the coding region for the E1 and E2 glycoproteins. Per unit length, the number of non-synonymous mutations in the E1 and E2 proteins was 6.3-fold higher than in the non-structural proteins. The cyan marker above the E1 gene represents the 739-nucleotide sequence window recommended by the World Health Organization (WHO) for RV genotyping. b Phylogenetic analysis of subject 6’s RV strain obtained from MDS with 32 WHO reference strains, GUZ_GER92 (Stuttgart strain), and the RV27/3 vaccine strain, demonstrating that the subject’s RV sequence was most closely related to the genotype 1G viruses and not the vaccine strain